From: tf2spi Date: Mon, 13 Oct 2025 19:39:31 +0000 (-0400) Subject: Finish Miss(le) Hell, the abomination X-Git-Url: http://git.misomosispi.com/?a=commitdiff_plain;h=c41b74142b078440454cc7c26fe74182bde1b9bb;p=miss-le-hell.git Finish Miss(le) Hell, the abomination --- diff --git a/.gitignore b/.gitignore index 259148f..92babd2 100644 --- a/.gitignore +++ b/.gitignore @@ -30,3 +30,14 @@ *.exe *.out *.app + +# Build and distribution directories +build/ +dist/ + +# Sysroot from your device +sysroot/ + +# You can have your personal shell scripts with '.i.sh' extensions +*.i.sh +*.i.bash diff --git a/README.md b/README.md index 41c9a1c..7f2d6d1 100644 --- a/README.md +++ b/README.md @@ -1,2 +1,75 @@ # hbatbp Hanging by a Thread: Break(ing)point + +## Building + +You can build the APK with the following commands +```sh +# This builds the APK and must be executed in current directory +# +# Needs ANDROID_HOME and JAVA_HOME variables as well as correct sdkmanager packages installed +# If you don't have them installed, the script will print out what is needed +scripts/android/build.sh + +# Cause the APK to emit a SIGSTOP on entry +# and a debugger will emit SIGCONT to continue +WAIT=1 scripts/android/build.sh + +# Build a release build +# WARNING: This takes a really, really long time because it turns on LTO and builds for all archs +scripts/android/build.sh release +``` + +This APK is named ``app-signed.apk`` and can then be installed like this +```sh +adb install build/android/app-signed.apk +``` + +Or the convenience shell scripts can be executed +``` +# Build, install, and debug in one run +# In addition, have the app wait for the debugger +WAIT=1 scripts/android/build.sh && scripts/android/install.sh && scripts/android/debug.sh + +# Convenience script with a nice logcat filter for debugging +scripts/android/log.sh +``` + +In addition, you can also build for Windows +``` +# SDL3's VC development release (SDL3-devel-3.x.x-VC.zip) must be installed at $SDL3_HOME +# If the SDL3_HOME environment variable is not defined, define it for this script +# +# clang is automatically searched for in Visual Studio 2022 Community +# but you can change scripts/windows/toolchain.sh to find clang somewhere else. +# It would be ideal to look across various VS years, distributions, and versions +# for clang, but I haven't bothered with this yet. +# +# Like the Android build script, must also be executed in current directory +scripts/windows/build.sh +``` + +It can then be run directly like this +``` +build/windows/app.exe +``` + +## Current Pain Points + +As I'm developing small test games for ``hbatbp``'s engine, +these are some general pain points that stand out in development + +* The tooling and support for common things like fonts, atlases, etc., is poor (if existant) +* Many GFX attributes and descriptions are missing or underdeveloped + - Uniforms, for example, would be nice so that we could do camera math in the vertex shader +* Most GFX objects are non-existant and textures are the only one you can have multiple of +* There's no well-developed linear algebra (``linalg``) library yet + - This would simplify sprite scaling and rotation in spritebatches + - This would simplify games programming in general +* Lack of a good handle/index allocation structure + - This is part of why it's been hard to improve core GFX and add new GFX objects + - Makes it easier to swap between AOS and SOA structures where appropriate + - Makes it easier for games to allocate massive arrays and manage them easily +* Lack of allocators + - Bump would make it easier to manage allocations and serve as a good temp allocator +* Lack of good UI functions diff --git a/assets/binary/LiberationSerif-Regular.ttf b/assets/binary/LiberationSerif-Regular.ttf new file mode 100644 index 0000000..d7622a3 Binary files /dev/null and b/assets/binary/LiberationSerif-Regular.ttf differ diff --git a/assets/binary/atlas.png b/assets/binary/atlas.png new file mode 100644 index 0000000..882207d Binary files /dev/null and b/assets/binary/atlas.png differ diff --git a/assets/binary/explosion.wav b/assets/binary/explosion.wav new file mode 100644 index 0000000..4e00306 Binary files /dev/null and b/assets/binary/explosion.wav differ diff --git a/assets/binary/missle_shot.wav b/assets/binary/missle_shot.wav new file mode 100644 index 0000000..5418f42 Binary files /dev/null and b/assets/binary/missle_shot.wav differ diff --git a/assets/text/font.frag b/assets/text/font.frag new file mode 100644 index 0000000..27201a5 --- /dev/null +++ b/assets/text/font.frag @@ -0,0 +1,9 @@ +#version 300 es +precision mediump float; +uniform sampler2D tex; +in vec2 uv; +out vec4 frag_color; +void main() { + float bright = texture(tex, uv).r; + frag_color = vec4(bright, bright, bright, min(bright * 255.0, 1.0)); +} diff --git a/assets/text/texture.frag b/assets/text/texture.frag new file mode 100644 index 0000000..b10ef19 --- /dev/null +++ b/assets/text/texture.frag @@ -0,0 +1,8 @@ +#version 300 es +precision mediump float; +uniform sampler2D tex; +in vec2 uv; +out vec4 frag_color; +void main() { + frag_color = texture(tex, uv); +} diff --git a/assets/text/texture.vert b/assets/text/texture.vert new file mode 100644 index 0000000..b1ae347 --- /dev/null +++ b/assets/text/texture.vert @@ -0,0 +1,8 @@ +#version 300 es +in vec4 position; +in vec2 uv0; +out vec2 uv; +void main() { + gl_Position = position; + uv = uv0; +} diff --git a/core/afx.c b/core/afx.c new file mode 100644 index 0000000..747bd62 --- /dev/null +++ b/core/afx.c @@ -0,0 +1,235 @@ +#include "afx.h" +#include "os.h" + +// Miniaudio, while technically counting more as platform code, +// does enough of what we need to just use it in core! +#define STB_VORBIS_HEADER_ONLY +#include "../platform/stb_vorbis.c" +#include "../platform/miniaudio.h" + +static struct allocator *afx_allocator = &os_allocator_default; + +#define VFS_TO_OS(f) ((union os_file){{(f)}}) +#define OS_TO_VFS(f) ((ma_vfs_file)(f)._i._i) +static ma_result onOpen(ma_vfs* pVFS, const char *pFilePath, ma_uint32 openMode, ma_vfs_file *p) +{ + (void)pVFS, (void)openMode; + union os_file fp = os_url_openc(pFilePath, afx_allocator); + *p = OS_TO_VFS(fp); + return *p ? MA_SUCCESS : MA_ERROR; +} +static ma_result onOpenW(ma_vfs* pVFS, const wchar_t* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + (void)pVFS, (void)pFilePath, (void)openMode, (void)pFile; + return MA_ERROR; +} +static ma_result onClose(ma_vfs* pVFS, ma_vfs_file file) +{ + (void)pVFS; + os_file_close(VFS_TO_OS(file)); + return MA_SUCCESS; +} +static ma_result onRead(ma_vfs* pVFS, ma_vfs_file file, void* pDst, size_t sizeInBytes, size_t* pBytesRead) +{ + (void)pVFS; + ssize_t sz = os_file_read(VFS_TO_OS(file), pDst, sizeInBytes); + if (sz >= 0) + *pBytesRead = sz; + else + return MA_ERROR; + return MA_SUCCESS; +} +static ma_result onWrite(ma_vfs* pVFS, ma_vfs_file file, const void* pSrc, size_t sizeInBytes, size_t* pBytesWritten) +{ + (void)pVFS; + ssize_t sz = os_file_write(VFS_TO_OS(file), pSrc, sizeInBytes); + if (sz >= 0) + *pBytesWritten = sz; + else + return MA_ERROR; + return MA_SUCCESS; +} +static ma_result onSeek(ma_vfs* pVFS, ma_vfs_file file, ma_int64 offset, ma_seek_origin origin) +{ + (void)pVFS; + int whence = -1; + switch (origin) + { + case ma_seek_origin_start: whence = OS_SEEK_SET; break; + case ma_seek_origin_current: whence = OS_SEEK_CUR; break; + case ma_seek_origin_end: whence = OS_SEEK_END; break; + } + int64_t ofs = os_file_seek(VFS_TO_OS(file), offset, whence); + return (ofs >= 0) ? MA_SUCCESS : MA_ERROR; +} +static ma_result onTell(ma_vfs* pVFS, ma_vfs_file file, ma_int64* pCursor) +{ + (void)pVFS; + int64_t ofs = os_file_tell(VFS_TO_OS(file)); + if (ofs >= 0) { + *pCursor = ofs; + return MA_SUCCESS; + } + return MA_ERROR; +} +static ma_result onInfo(ma_vfs* pVFS, ma_vfs_file file, ma_file_info* pInfo) +{ + (void)pVFS; + int64_t size = os_file_size(VFS_TO_OS(file)); + if (size < 0) + return MA_ERROR; + pInfo->sizeInBytes = size; + return MA_SUCCESS; +} +static ma_vfs_callbacks fs_cbs = { + onOpen, + onOpenW, + onClose, + onRead, + onWrite, + onSeek, + onTell, + onInfo, +}; + +// For ma_allocation_callbacks +static void *onMalloc(size_t sz, void *pUserData) +{ + return os_malloc_emulate_a(sz, (struct allocator *)pUserData); +} +static void *onRealloc(void *p, size_t sz, void *pUserData) +{ + return os_realloc_emulate_a(p, sz, (struct allocator *)pUserData); +} +static void onFree(void *p, void *pUserData) +{ + os_free_emulate_a(p, (struct allocator *)pUserData); +} + +// Okay, miniaudio stuff out of the way, initialize the audio context +struct afx_ctx { + ma_engine _e; + ma_resource_manager _m; +}; +static struct blob g_afx_ctx = {}; + +static void afx_once(void) +{ + ma_engine *e = (ma_engine *)0; + ma_resource_manager *m = (ma_resource_manager *)0; + + struct blob b = os_alloc((struct allocator *)0, sizeof(struct afx_ctx), 0); + struct afx_ctx *ctx = (struct afx_ctx *)b.memory; + if (!ctx) + goto afx_once_fail; + + ma_resource_manager_config rmconfig = ma_resource_manager_config_init(); + rmconfig.allocationCallbacks = (ma_allocation_callbacks){ + afx_allocator, + onMalloc, + onRealloc, + onFree, + }; + rmconfig.pVFS = (ma_vfs *)&fs_cbs; + if (ma_resource_manager_init(&rmconfig, &ctx->_m) != MA_SUCCESS) + goto afx_once_fail; + m = &ctx->_m; + + ma_engine_config econfig = ma_engine_config_init(); + econfig.pResourceManager = m; + if (ma_engine_init(&econfig, &ctx->_e) != MA_SUCCESS) + goto afx_once_fail; + e = &ctx->_e; + + g_afx_ctx = b; + return; + +afx_once_fail: + if (e) + ma_engine_uninit(e); + if (m) + ma_resource_manager_uninit(m); + os_free((struct allocator *)0, b); +} +static union os_once g_once = OS_ONCE_INIT; +static struct afx_ctx *afx_ctx(void) +{ + return (struct afx_ctx *)g_afx_ctx.memory; +} + +struct afx_sound_i { + ma_sound _i; + struct blob _b; +}; + +#define AFX_TO_MA(s) ((ma_sound *)&(s)._i->_i) +#define MA_TO_AFX(s) ((union afx_sound){(struct afx_sound_i *)s}) +union afx_sound afx_sound_construct(struct stringc url, int flags) +{ + union afx_sound s = {}; + struct afx_ctx *ctx = afx_ctx(); + if (!ctx) + return s; + + struct blob b = os_alloc(afx_allocator, sizeof(*s._i), 0); + s._i = (struct afx_sound_i *)b.memory; + if (!s._i) + return s; + s._i->_b = b; + + int ma_flags = 0; + if (flags & AFX_SOUND_STREAM) ma_flags |= MA_SOUND_FLAG_STREAM; + if (flags & AFX_SOUND_DECODE) ma_flags |= MA_SOUND_FLAG_DECODE; + char *nullterm = os_string_nullterm_make(url, afx_allocator); + ma_result result = ma_sound_init_from_file(&ctx->_e, nullterm, ma_flags, (ma_sound_group *)0, (ma_fence *)0, &s._i->_i); + if (nullterm && nullterm != url.data) + os_free_emulate_a(nullterm, afx_allocator); + if (result != MA_SUCCESS) { + os_free(afx_allocator, b); + s._i = (struct afx_sound_i *)0; + } + if (flags & AFX_SOUND_LOOP_NOW) afx_sound_loop(s); + if (flags & AFX_SOUND_PLAY_NOW) afx_sound_restart(s); + return s; +} + +void afx_sound_destruct(union afx_sound s) +{ + if (!s._i) return; + ma_sound_uninit(&s._i->_i); + os_free(afx_allocator, s._i->_b); +} + +void afx_sound_restart(union afx_sound s) +{ + if (!s._i) return; + ma_sound_seek_to_pcm_frame(AFX_TO_MA(s), 0); + ma_sound_start(AFX_TO_MA(s)); +} + +void afx_sound_stop(union afx_sound s) +{ + if (!s._i) return; + ma_sound_stop(AFX_TO_MA(s)); +} + +void afx_sound_loop(union afx_sound s) +{ + if (!s._i) return; + ma_sound_set_looping(AFX_TO_MA(s), 1); +} + +void afx_start(void) +{ + os_once_run(&g_once, afx_once); + struct afx_ctx *ctx = afx_ctx(); + if (!ctx) return; + ma_engine_start(&ctx->_e); +} + +void afx_stop(void) +{ + struct afx_ctx *ctx = afx_ctx(); + if (!ctx) return; + ma_engine_stop(&ctx->_e); +} diff --git a/core/core_callbacks.c b/core/core_callbacks.c new file mode 100644 index 0000000..c8e82a1 --- /dev/null +++ b/core/core_callbacks.c @@ -0,0 +1,86 @@ +#include "core_callbacks.h" +#include "game_callbacks.h" +#include "os.h" +#include "gfx.h" +#include "afx.h" +#include "input.h" + +static union gfx_ctx g_gfx_ctx; +void core_callback_entry(int argc, char **argv) { + OSLOGD("Logging entry!"); + (void)os_time(OS_TIME_TICK); // One call needed to initialize ticks + game_callback_entry(argc, argv); +} + +// If working correctly, should regenerate all resources from gfx_ctx +void core_callback_gfx_startup(union platform_window window) +{ + OSLOGD("Logging gfx startup!"); + g_gfx_ctx = gfx_startup(window, g_gfx_ctx); + + // Want to start up audio again + afx_start(); + + game_callback_gfx_startup(); +} + +// If working correctly, should mark all resources from gfx_ctx as invalid +void core_callback_gfx_shutdown(union platform_window window) +{ + (void)window; + OSLOGD("Logging gfx shutdown!"); + game_callback_gfx_shutdown(); + g_gfx_ctx = gfx_shutdown(window); + + // Want to stop audio effects + afx_stop(); +} + +void core_callback_event(union platform_event *event, union platform_window window) +{ + if (event->type == PLATFORM_EVENT_NONE) + return; + + switch (event->type) { + case PLATFORM_EVENT_FINGER_DOWN: + case PLATFORM_EVENT_FINGER_UP: + case PLATFORM_EVENT_FINGER_MOTION: + case PLATFORM_EVENT_KEY_UP: + case PLATFORM_EVENT_KEY_DOWN: + case PLATFORM_EVENT_MOUSE_BUTTON_UP: + case PLATFORM_EVENT_MOUSE_BUTTON_DOWN: + case PLATFORM_EVENT_MOUSE_MOTION: + case PLATFORM_EVENT_GAMEPAD_CONNECTED: + case PLATFORM_EVENT_GAMEPAD_DISCONNECTED: + case PLATFORM_EVENT_GAMEPAD_BUTTON_DOWN: + case PLATFORM_EVENT_GAMEPAD_BUTTON_UP: + case PLATFORM_EVENT_GAMEPAD_AXIS_MOTION: + input_inject_platform_event(event, window); + break; + default: + return; + } +} + +void core_callback_variable(void) +{ + // TODO: Setup delta time + // TODO: Keep track of time to do a fixed update +} + +void core_callback_render(union platform_window window) +{ + // Always do updates here + game_callback_variable(); + + // Only do a render stage if the window gfx is valid + // TODO: It's useful for game code to be able to render to + // textures instead of the screen for post-processing, + // saving images of the game, etc..., so make wrappers + // for render_begin, render_end, and gfx_present. + if (!platform_gfx_valid(window)) return; + platform_gfx_render_begin(window); + game_callback_render(); + platform_gfx_render_end(window); + platform_gfx_present(window); +} diff --git a/core/gfx.c b/core/gfx.c new file mode 100644 index 0000000..7226a93 --- /dev/null +++ b/core/gfx.c @@ -0,0 +1,664 @@ +#include "gfx.h" +#include "os.h" + +// Various graphics resources +struct gfx_shader_i +{ + union platform_gfx_shader platform; + struct gfx_shader_desc desc; + struct blob vertex_url; + struct blob fragment_url; +}; + +struct gfx_texture_i +{ + int width, height; + union platform_gfx_texture platform; + struct gfx_texture_desc desc; + struct blob url; +}; + +struct gfx_pipeline_i +{ + union platform_gfx_pipeline platform; + struct gfx_pipeline_desc desc; +}; + +struct gfx_sampler_i +{ + union platform_gfx_sampler platform; + struct gfx_sampler_desc desc; +}; + +struct gfx_font_i +{ + union platform_gfx_font platform; + struct gfx_font_desc desc; + struct blob url; + struct blob bmglyphs; + union gfx_texture texture; + float fontsize; +}; + +// TODO: Most memory in a gfx resource are for regeneration +// which is very rarely used, so this should be split into +// a resource and a "regen" structure for better caching +union gfx_resource_i +{ + struct gfx_texture_i texture; + struct gfx_shader_i shader; + struct gfx_pipeline_i pipeline; + struct gfx_sampler_i sampler; + struct gfx_font_i font; +}; + +struct gfx_resource_list_i +{ + union gfx_resource_i resource; + uint32_t prev; + uint32_t next; +}; + +enum gfx_resource_type +{ + GFX_RESOURCE_NONE = 0, + GFX_RESOURCE_TEXTURE = 1, + GFX_RESOURCE_SHADER = 2, + GFX_RESOURCE_PIPELINE = 3, + GFX_RESOURCE_SAMPLER = 4, + GFX_RESOURCE_FONT = 5, +}; + +// One global graphics context since we don't create multiple windows +#define GFX_RESOURCE_BITS_IDX 16 +#define GFX_RESOURCE_BITS_IDX_GET(h) ((h) & ((1 << GFX_RESOURCE_BITS_IDX) - 1)) +#define GFX_RESOURCE_BITS_GEN 8 +#define GFX_RESOURCE_BITS_GEN_GET(h) (((h) >> GFX_RESOURCE_BITS_IDX) & ((1 << GFX_RESOURCE_BITS_GEN) - 1)) +#define GFX_RESOURCE_BITS_TYP 4 +#define GFX_RESOURCE_BITS_TYP_GET(h) (((h) >> (GFX_RESOURCE_BITS_IDX + GFX_RESOURCE_BITS_GEN)) & ((1 << GFX_RESOURCE_BITS_TYP) - 1)) +struct gfx_ctx_i +{ + struct handle_table handles; + struct gfx_resource_list_i resources[1 << GFX_RESOURCE_BITS_IDX]; + union platform_window window; + uint32_t regen_list; +}; +// We have one global graphics context because we don't support multiple windows yet +// Handles declared separately so that .data sections don't blow up... +static struct handle_table g_ctx_data_handles = HANDLE_TABLE_STATIC(GFX_RESOURCE_BITS_IDX, GFX_RESOURCE_BITS_GEN); +struct gfx_ctx_i g_ctx_data = {}; +struct gfx_ctx_i *g_ctx = (struct gfx_ctx_i *)0; +static struct allocator *gfx_allocator = &os_allocator_default; + +// TODO: Handle functions extremely, wildly thread unsafe. Make it thread safe. +// NOTE: Handles don't keep track of context, so, if and when we support multiple contexts, +// then you would have to explicitly set the correct context and then use the handle +static union gfx_resource_i *gfx_resource_from_handle(uint32_t h, uint8_t type) +{ + struct gfx_ctx_i *ctx = g_ctx; + uint32_t index = 0; + if (ctx) + index = handle_table_check(&ctx->handles, h); + if (!index || GFX_RESOURCE_BITS_TYP_GET(h) != type) + return (union gfx_resource_i *)0; + return &(ctx->resources[index].resource); +} +static uint32_t gfx_handle_alloc(uint8_t type) +{ + struct gfx_ctx_i *ctx = g_ctx; + uint32_t h = 0; + if (ctx) + h = handle_table_alloc(&ctx->handles); + if (!h) + return 0; + h |= (type << (GFX_RESOURCE_BITS_IDX + GFX_RESOURCE_BITS_GEN)); + + // Initialize the resource + struct gfx_resource_list_i *node = &ctx->resources[GFX_RESOURCE_BITS_IDX_GET(h)]; + memset(node, 0, sizeof(*node)); + + // Add it to the regen list + uint32_t head_h = ctx->regen_list; + if (head_h) { + struct gfx_resource_list_i *head = &ctx->resources[GFX_RESOURCE_BITS_IDX_GET(head_h)]; + uint32_t prev_h = head->prev; + if (prev_h) { + struct gfx_resource_list_i *prev = &ctx->resources[GFX_RESOURCE_BITS_IDX_GET(prev_h)]; + prev->next = h; + } + head->prev = h; + node->prev = prev_h; + } + node->next = head_h; + ctx->regen_list = h; + return h; +} +static void gfx_handle_free(uint32_t h) +{ + struct gfx_ctx_i *ctx = g_ctx; + if (!ctx || handle_table_free(&ctx->handles, h) < 0) + return; + + // Remove it from the regen list + struct gfx_resource_list_i *node = &ctx->resources[GFX_RESOURCE_BITS_IDX_GET(h)]; + uint32_t prev_h = node->prev; + uint32_t next_h = node->next; + if (prev_h) + ctx->resources[GFX_RESOURCE_BITS_IDX_GET(prev_h)].next = next_h; + if (next_h) + ctx->resources[GFX_RESOURCE_BITS_IDX_GET(next_h)].prev = prev_h; + + // Deinitialize the resource + memset(node, 0, sizeof(*node)); +} + +static struct gfx_texture_i *gfx_texture_from_handle(uint32_t handle) +{ + return &gfx_resource_from_handle(handle, GFX_RESOURCE_TEXTURE)->texture; +} +static struct gfx_shader_i *gfx_shader_from_handle(uint32_t handle) +{ + return &gfx_resource_from_handle(handle, GFX_RESOURCE_SHADER)->shader; +} +static struct gfx_pipeline_i *gfx_pipeline_from_handle(uint32_t handle) +{ + return &gfx_resource_from_handle(handle, GFX_RESOURCE_PIPELINE)->pipeline; +} +static struct gfx_sampler_i *gfx_sampler_from_handle(uint32_t handle) +{ + return &gfx_resource_from_handle(handle, GFX_RESOURCE_SAMPLER)->sampler; +} +static struct gfx_font_i *gfx_font_from_handle(uint32_t handle) +{ + return &gfx_resource_from_handle(handle, GFX_RESOURCE_FONT)->font; +} + +// TODO: For regeneration functions, do we want them to take a gfx_ctx_i? +// We already assume that we have a global context that we can reference. +static void gfx_regen_texture(struct gfx_ctx_i *ctx, struct gfx_texture_i *texture) +{ + if (texture->platform._i) + return; + + // NOTE: Do not remove this check + // gfx_font_construct will use empty URLs for textures + // so that they regenerated in a custom manner instead of like this + struct gfx_texture_desc *desc = &texture->desc; + if (!desc->url.len) + return; + + union os_file fp = os_url_open(desc->url, gfx_allocator); + struct blob contents = os_file_read_all(fp, gfx_allocator); + struct image img = os_image_decode(contents.memory, contents.size, gfx_allocator); + texture->width = img.width; + texture->height = img.height; + + // If the image width or height is 0, there's nothing to draw + if (img.width != 0 && img.height != 0) + texture->platform = platform_gfx_texture_construct(ctx->window, &(struct platform_gfx_texture_desc){ + .pixels = img.pixels, + .width = img.width, + .height = img.height, + .bytesize = img.bytesize, + .format = img.format, + }); + else + os_loge("Image failed to read or has no pixels!"); + + os_free(gfx_allocator, img._i); + os_free(gfx_allocator, contents); + os_file_close(fp); +} +static void gfx_regen_font(struct gfx_ctx_i *ctx, struct gfx_font_i *font) +{ + // TODO: Regenerate texture and font + (void)ctx; + (void)font; + + if (font->platform._i) + return; + + struct gfx_font_desc *desc = &font->desc; + if (!desc->url.len) + return; + + union os_file fp = os_url_open(desc->url, gfx_allocator); + struct blob contents = os_file_read_all(fp, gfx_allocator); + os_file_close(fp); + if (!contents.memory) + return; + font->platform = platform_gfx_font_construct(ctx->window, &(struct platform_gfx_font_desc){ + .a = gfx_allocator, + .data_free = gfx_allocator, + .data = contents, + }); + // If the allocation succeeds, then font owns the data now + if (!font->platform._i) + os_free(gfx_allocator, contents); +} +static void gfx_regen_shader(struct gfx_ctx_i *ctx, struct gfx_shader_i *shader) +{ + if (shader->platform._i) + return; + + struct gfx_shader_desc *desc = &shader->desc; + if (!desc->vertex_url.len || !desc->fragment_url.len) + return; + + // TODO: Ensure vertex and fragment null terminated + union os_file vfp = os_url_open(desc->vertex_url, gfx_allocator); + union os_file ffp = os_url_open(desc->fragment_url, gfx_allocator); + struct blob vertex = os_file_read_all_extra(vfp, gfx_allocator, 1); + if (vertex.memory) + ((char *)vertex.memory)[vertex.size-1] = 0; + struct blob fragment = os_file_read_all_extra(ffp, gfx_allocator, 1); + if (fragment.memory) + ((char *)fragment.memory)[fragment.size-1] = 0; + shader->platform = platform_gfx_shader_construct(ctx->window, &(struct platform_gfx_shader_desc){ + .vertex = vertex.memory ? vertex.memory : "", + .fragment = fragment.memory ? fragment.memory : "", + }); + os_free(gfx_allocator, fragment); + os_free(gfx_allocator, vertex); + os_file_close(ffp); + os_file_close(vfp); +} +static void gfx_regen_pipeline(struct gfx_ctx_i *ctx, struct gfx_pipeline_i *pipeline) +{ + if (pipeline->platform._i) + return; + + struct gfx_pipeline_desc *desc = &pipeline->desc; + struct gfx_shader_i *shader = gfx_shader_from_handle(desc->shader._i); + if (!shader) + return; + + // If shader has already been regenerated, then gfx_regen_shader is a noop. + // Otherwise, it will regenerate the platform shader that we need if successful + gfx_regen_shader(ctx, shader); + if (!shader->platform._i) + return; + pipeline->platform = platform_gfx_pipeline_construct(ctx->window, &(struct platform_gfx_pipeline_desc){ + .shader = shader->platform, + }); +} +static void gfx_regen_sampler(struct gfx_ctx_i *ctx, struct gfx_sampler_i *sampler) +{ + if (sampler->platform._i) + return; + + struct gfx_sampler_desc *desc = &sampler->desc; + (void)desc; + sampler->platform = platform_gfx_sampler_construct(ctx->window, &(struct platform_gfx_sampler_desc){ + }); +} +static void gfx_regen_resources(struct gfx_ctx_i *ctx, union platform_window window) +{ + ctx->window = window; + uint32_t h = ctx->regen_list; + while (h) { + struct gfx_resource_list_i *node = &ctx->resources[GFX_RESOURCE_BITS_IDX_GET(h)]; + union gfx_resource_i *resource = &node->resource; + switch (GFX_RESOURCE_BITS_TYP_GET(h)) { + case GFX_RESOURCE_SHADER: gfx_regen_shader(ctx, &resource->shader); break; + case GFX_RESOURCE_TEXTURE: gfx_regen_texture(ctx, &resource->texture); break; + case GFX_RESOURCE_PIPELINE: gfx_regen_pipeline(ctx, &resource->pipeline); break; + case GFX_RESOURCE_SAMPLER: gfx_regen_sampler(ctx, &resource->sampler); break; + case GFX_RESOURCE_FONT: gfx_regen_font(ctx, &resource->font); break; + } + h = node->next; + } +} +union gfx_ctx gfx_startup(union platform_window window, union gfx_ctx prev) +{ + struct gfx_ctx_i *ctx = prev._i; + if (!ctx) { + g_ctx = &g_ctx_data; + prev._i = g_ctx; + ctx = prev._i; + + ctx->handles = g_ctx_data_handles; + ctx->window = window; + } else { + gfx_regen_resources(ctx, window); + } + return prev; +} +union gfx_ctx gfx_shutdown(union platform_window window) +{ + // TODO: Assert that the platform window matches one in context + (void)window; + struct gfx_ctx_i *ctx = g_ctx; + if (!ctx) + return (union gfx_ctx){}; + + uint32_t h = ctx->regen_list; + while (h) { + struct gfx_resource_list_i *node = &ctx->resources[GFX_RESOURCE_BITS_IDX_GET(h)]; + union gfx_resource_i *resource = &node->resource; + switch (GFX_RESOURCE_BITS_TYP_GET(h)) { + case GFX_RESOURCE_SHADER: + if (resource->shader.platform._i) { + platform_gfx_shader_destruct(ctx->window, resource->shader.platform); + resource->shader.platform = (union platform_gfx_shader){}; + } + break; + case GFX_RESOURCE_TEXTURE: + if (resource->texture.platform._i) { + platform_gfx_texture_destruct(ctx->window, resource->texture.platform); + resource->texture.platform = (union platform_gfx_texture){}; + } + break; + case GFX_RESOURCE_PIPELINE: + if (resource->pipeline.platform._i) { + platform_gfx_pipeline_destruct(ctx->window, resource->pipeline.platform); + resource->pipeline.platform = (union platform_gfx_pipeline){}; + } + break; + case GFX_RESOURCE_SAMPLER: + if (resource->sampler.platform._i) { + platform_gfx_sampler_destruct(ctx->window, resource->sampler.platform); + resource->sampler.platform = (union platform_gfx_sampler){}; + } + break; + case GFX_RESOURCE_FONT: + if (resource->font.platform._i) { + platform_gfx_font_destruct(ctx->window, resource->font.platform); + resource->font.platform = (union platform_gfx_font){}; + } + if (resource->font.bmglyphs.memory) { + os_free(gfx_allocator, resource->font.bmglyphs); + memset(&resource->font.bmglyphs, 0, sizeof(resource->font.bmglyphs)); + } + } + h = node->next; + } + + ctx->window = (union platform_window){}; + return (union gfx_ctx){ ctx }; +} +union gfx_shader gfx_shader_construct(const struct gfx_shader_desc *desc) +{ + struct gfx_ctx_i *ctx = g_ctx; + uint32_t h = gfx_handle_alloc(GFX_RESOURCE_SHADER); + if (!h) + return (union gfx_shader){}; + + struct gfx_shader_i *shader = gfx_shader_from_handle(h); + shader->fragment_url = os_strdup(desc->fragment_url, gfx_allocator); + shader->vertex_url = os_strdup(desc->vertex_url, gfx_allocator); + memcpy(&shader->desc, desc, sizeof(shader->desc)); + shader->desc.vertex_url = BLOB2STRC(shader->vertex_url); + shader->desc.fragment_url = BLOB2STRC(shader->fragment_url); + gfx_regen_shader(ctx, shader); + return (union gfx_shader){h}; +} +void gfx_shader_destruct(union gfx_shader s) +{ + struct gfx_ctx_i *ctx = g_ctx; + struct gfx_shader_i *shader = gfx_shader_from_handle(s._i); + if (!shader) + return; + platform_gfx_shader_destruct(ctx->window, shader->platform); + os_free(gfx_allocator, shader->vertex_url); + os_free(gfx_allocator, shader->fragment_url); + gfx_handle_free(s._i); +} + +union gfx_pipeline gfx_pipeline_construct(const struct gfx_pipeline_desc *desc) +{ + struct gfx_ctx_i *ctx = g_ctx; + uint32_t h = gfx_handle_alloc(GFX_RESOURCE_PIPELINE); + if (!h) + return (union gfx_pipeline){}; + + struct gfx_pipeline_i *pipeline = gfx_pipeline_from_handle(h); + memcpy(&pipeline->desc, desc, sizeof(pipeline->desc)); + gfx_regen_pipeline(ctx, pipeline); + return (union gfx_pipeline){h}; +} +void gfx_pipeline_destruct(const union gfx_pipeline p) +{ + struct gfx_ctx_i *ctx = g_ctx; + struct gfx_pipeline_i *pipeline = gfx_pipeline_from_handle(p._i); + if (!pipeline) + return; + if (pipeline->platform._i) + platform_gfx_pipeline_destruct(ctx->window, pipeline->platform); + gfx_handle_free(p._i); +} +void gfx_pipeline_bind(union gfx_pipeline p) +{ + struct gfx_ctx_i *ctx = g_ctx; + struct gfx_pipeline_i *pipeline = gfx_pipeline_from_handle(p._i); + if (pipeline && pipeline->platform._i) + platform_gfx_pipeline_bind(ctx->window, pipeline->platform); +} + +union gfx_texture gfx_texture_construct(const struct gfx_texture_desc *desc) +{ + struct gfx_ctx_i *ctx = g_ctx; + uint32_t h = gfx_handle_alloc(GFX_RESOURCE_TEXTURE); + if (!h) + return (union gfx_texture){}; + + struct gfx_texture_i *texture = gfx_texture_from_handle(h); + memcpy(&texture->desc, desc, sizeof(texture->desc)); + texture->url = os_strdup(desc->url, gfx_allocator); + texture->desc.url = BLOB2STRC(texture->url); + gfx_regen_texture(ctx, texture); + return (union gfx_texture){h}; +} +void gfx_texture_destruct(union gfx_texture t) +{ + struct gfx_ctx_i *ctx = g_ctx; + struct gfx_texture_i *texture = gfx_texture_from_handle(t._i); + if (!texture) + return; + if (texture->platform._i) + platform_gfx_texture_destruct(ctx->window, texture->platform); + os_free(gfx_allocator, texture->url); + gfx_handle_free(t._i); +} +int gfx_texture_width(union gfx_texture t) +{ + // Return 1 to avoid divide by 0 situations + struct gfx_texture_i *texture = gfx_texture_from_handle(t._i); + if (!texture) + return 1; + return texture->width; +} +int gfx_texture_height(union gfx_texture t) +{ + // Return 1 to avoid divide by 0 situations + struct gfx_texture_i *texture = gfx_texture_from_handle(t._i); + if (!texture) + return 1; + return texture->height; +} +struct rectanglei32 gfx_texture_rect(union gfx_texture texture) +{ + return (struct rectanglei32){0, 0, gfx_texture_width(texture), gfx_texture_height(texture)}; +} + +union gfx_sampler gfx_sampler_construct(const struct gfx_sampler_desc *desc) +{ + struct gfx_ctx_i *ctx = g_ctx; + uint32_t h = gfx_handle_alloc(GFX_RESOURCE_SAMPLER); + if (!h) + return (union gfx_sampler){}; + + struct gfx_sampler_i *sampler = gfx_sampler_from_handle(h); + memcpy(&sampler->desc, desc, sizeof(sampler->desc)); + gfx_regen_sampler(ctx, sampler); + return (union gfx_sampler){h}; +} + +void gfx_sampler_destruct(union gfx_sampler s) +{ + struct gfx_sampler_i *sampler = gfx_sampler_from_handle(s._i); + if (!sampler) + return; + struct gfx_ctx_i *ctx = g_ctx; + if (sampler->platform._i) + platform_gfx_sampler_destruct(ctx->window, sampler->platform); + gfx_handle_free(s._i); +} + +union gfx_font gfx_font_construct(const struct gfx_font_desc *desc) +{ + struct gfx_ctx_i *ctx = g_ctx; + uint32_t h = gfx_handle_alloc(GFX_RESOURCE_FONT); + if (!h) + return (union gfx_font){}; + + // Zero-ing out the texture is enough to make sure + // that it's in a good state and that it's not regenerated + // via URL so that we can regenerate it later in gfx_regen_font. + // GFX resources are always initialized to 0 when allocated. + uint32_t tex_h = gfx_handle_alloc(GFX_RESOURCE_TEXTURE); + if (!tex_h) { + gfx_handle_free(h); + return (union gfx_font){}; + } + + struct gfx_font_i *font = gfx_font_from_handle(h); + font->texture = (union gfx_texture){tex_h}; + font->fontsize = 12.0f; + font->url = os_strdup(desc->url, gfx_allocator); + memcpy(&font->desc, desc, sizeof(font->desc)); + font->desc.url = BLOB2STRC(font->url); + gfx_regen_font(ctx, font); + return (union gfx_font){h}; +} +void gfx_font_destruct(union gfx_font f) +{ + struct gfx_font_i *font = gfx_font_from_handle(f._i); + if (!font) + return; + struct gfx_ctx_i *ctx = g_ctx; + if (font->platform._i) + platform_gfx_font_destruct(ctx->window, font->platform); + os_free(gfx_allocator, font->url); + gfx_handle_free(f._i); +} +// Lazily recreate the texture so that, if user wants to specify codepoint ranges +// and font sizes beforehand, they can do so and generate the texture once +static void gfx_font_texture_lazy(struct gfx_font_i *font) +{ + struct gfx_ctx_i *ctx = g_ctx; + if (!ctx) + return; + + // If there are bitmap glyphs, then we have the texture already + if (font->bmglyphs.memory) + return; + + // TODO: We want to calculate how much memory we need dynamically from codepoint ranges + // For now, we only support ASCII characters and allocate that much memory + font->bmglyphs = os_alloc(gfx_allocator, sizeof(struct platform_gfx_font_bmglyph) * (0x7e - 0x20), 0); + if (!font->bmglyphs.memory) + return; + + struct image img = platform_gfx_font_texture(ctx->window, font->platform, font->fontsize, font->bmglyphs.memory); + if (img.width == 0 && img.height == 0) { + os_free(gfx_allocator, font->bmglyphs); + memset(&font->bmglyphs, 0, sizeof(font->bmglyphs)); + return; + } + + struct gfx_texture_i *texture = gfx_texture_from_handle(font->texture._i); + if (texture) { + texture->width = img.width; + texture->height = img.height; + texture->platform = platform_gfx_texture_construct(ctx->window, &(struct platform_gfx_texture_desc){ + .pixels = img.pixels, + .width = img.width, + .height = img.height, + .bytesize = img.bytesize, + .format = img.format, + }); + } + os_free(platform_gfx_font_allocator(ctx->window, font->platform), img._i); +} +union gfx_texture gfx_font_texture(union gfx_font f) +{ + struct gfx_font_i *font = gfx_font_from_handle(f._i); + if (!font) + return (union gfx_texture){}; + gfx_font_texture_lazy(font); + return font->texture; +} +struct gfx_font_bmglyph gfx_font_bmglyph_get(union gfx_font f, int rendered_codepoint, int kerning_codepoint) +{ + struct gfx_font_i *font = gfx_font_from_handle(f._i); + if (!font) + return (struct gfx_font_bmglyph){}; + + // TODO: Assuming ASCII again, stop that, get some help + gfx_font_texture_lazy(font); + if (!font->bmglyphs.memory || rendered_codepoint < 0x20 || rendered_codepoint > 0x7e) + return (struct gfx_font_bmglyph){}; + + (void)kerning_codepoint; + struct platform_gfx_font_bmglyph *glyph = &((struct platform_gfx_font_bmglyph *)font->bmglyphs.memory)[rendered_codepoint - 0x20]; + return (struct gfx_font_bmglyph){ + .x = glyph->x0, + .y = glyph->y0, + .width = (glyph->x1 - glyph->x0), + .height = (glyph->y1 - glyph->y0), + .xoffset = (int16_t)glyph->xoff, + .yoffset = (int16_t)-(glyph->yoff + (float)(glyph->y1 - glyph->y0)), + .xadvance = (int16_t)glyph->xadvance, + }; +} +struct gfx_font_info gfx_font_info_get(union gfx_font f) +{ + struct gfx_font_i *font = gfx_font_from_handle(f._i); + struct gfx_ctx_i *ctx = g_ctx; + if (!font || !ctx) + return (struct gfx_font_info){}; + struct platform_gfx_font_info info = platform_gfx_font_info_get(ctx->window, font->platform); + return (struct gfx_font_info){ + .fontsize = font->fontsize, + .ascent = (float)info.ascent, + .descent = (float)info.descent, + .linegap = (float)info.linegap, + }; +} +void gfx_font_size_set(union gfx_font f, float fontsize) +{ + struct gfx_font_i *font = gfx_font_from_handle(f._i); + if (!font) + return; + // If font size changes, texture is invalid. Invalidate the glyphs and regenerate the texture. + if (fabsf(fontsize - font->fontsize) > 1e-4f) { + font->fontsize = fontsize; + os_free(gfx_allocator, font->bmglyphs); + memset(&font->bmglyphs, 0, sizeof(font->bmglyphs)); + } +} + +void gfx_vertices_draw(union gfx_texture t, union gfx_sampler s, const void *vertices, size_t bytelen, size_t vertex_count) +{ + struct gfx_texture_i *texture = gfx_texture_from_handle(t._i); + struct gfx_sampler_i *sampler = gfx_sampler_from_handle(s._i); + if (!texture || !sampler) + return; + if (texture->platform._i && sampler->platform._i && vertices && bytelen) + platform_gfx_vertices_draw(g_ctx->window, texture->platform, sampler->platform, vertices, bytelen, vertex_count); +} + +// Debug/prototyping functions +union gfx_ctx _gfx_ctx_global(void) +{ + return (union gfx_ctx){ g_ctx }; +} +union platform_window _gfx_ctx_window(union gfx_ctx ctx) +{ + return ctx._i->window; +} +struct allocator *_gfx_allocator(void) +{ + return gfx_allocator; +} diff --git a/core/input.c b/core/input.c new file mode 100644 index 0000000..f849a74 --- /dev/null +++ b/core/input.c @@ -0,0 +1,131 @@ +#include "input.h" +#include "os.h" +#include "linalg.h" + +static struct input_callbacks g_input_callbacks = {}; +void input_set_callbacks(const struct input_callbacks *cbs) +{ + if (!cbs) + return; + memcpy(&g_input_callbacks, cbs, sizeof(g_input_callbacks)); +} + +static struct input_motion pixels_to_motion(struct vector2f32 pixels, struct vector2f32 delta, union platform_window window) +{ + int wwidth = platform_window_width(window); + int wheight = platform_window_height(window); + return (struct input_motion){ + // normalized + .normalized = (struct vector2f32){ + ((pixels.x / wwidth) - 0.5f) * 2.0f, + -((pixels.y / wheight) - 0.5f) * 2.0f, + }, + .normdelta = (struct vector2f32){ + ((delta.x / wwidth) - 0.5f) * 2.0f, + -((delta.y / wheight) - 0.5f) * 2.0f, + }, + // TODO: This is so horribly wrong, but we don't use this yet. + .inches = pixels, + }; +} + +// TODO: More button flags for higher-valued buttons +// TODO: Button flags associated by device rather than global +// NOTE: Don't make BUTTON_MAX > 30 since shift loop checks if g_mouse_buttons <= (1 << i) +#define BUTTON_MAX 30 +static uint32_t g_mouse_buttons = 0; +void input_inject_platform_event(const union platform_event *event, union platform_window window) +{ + (void)window; + switch (event->type) { + case PLATFORM_EVENT_MOUSE_BUTTON_UP: + case PLATFORM_EVENT_MOUSE_MOTION: + case PLATFORM_EVENT_MOUSE_BUTTON_DOWN: + case PLATFORM_EVENT_FINGER_DOWN: + case PLATFORM_EVENT_FINGER_MOTION: + case PLATFORM_EVENT_FINGER_UP: + if (g_input_callbacks.on_fmotion) { + int addition = 0; + uint32_t device = event->finger.device; + uint32_t finger = event->finger.id + 1; + float pressure = event->finger.pressure; + struct vector2f32 pixels = event->finger.pixels; + + // Calculate how many fingers were removed or added + switch (event->type) { + case PLATFORM_EVENT_MOUSE_BUTTON_DOWN: + if (!event->button.index || event->button.index > BUTTON_MAX) return; + g_mouse_buttons |= (1 << event->button.index); + finger = PSEUDOFINGER_MOUSE_BUTTON_BASE + event->button.index; + pixels = event->button.pixels; + device = event->button.which; + pressure = 1.0f; + FALLTHROUGH; + case PLATFORM_EVENT_FINGER_DOWN: + addition = 1; + break; + case PLATFORM_EVENT_MOUSE_MOTION: + finger = PSEUDOFINGER_MOUSE_MOTION; + pixels = event->motion.pixels; + device = event->motion.which; + pressure = 1.0f; + FALLTHROUGH; + case PLATFORM_EVENT_FINGER_MOTION: + addition = 0; + break; + case PLATFORM_EVENT_MOUSE_BUTTON_UP: + if (!event->button.index || event->button.index > BUTTON_MAX) return; + g_mouse_buttons &= ~(1 << event->button.index); + finger = PSEUDOFINGER_MOUSE_BUTTON_BASE + event->button.index; + pixels = event->button.pixels; + device = event->button.which; + FALLTHROUGH; + case PLATFORM_EVENT_FINGER_UP: + addition = -1; pressure = 0.0f; + break; + } + if (!finger) + return; + // TODO: Actually implement delta for fingers + struct input_motion motion = pixels_to_motion(pixels, (struct vector2f32){}, window); + g_input_callbacks.on_fmotion(device, finger, &motion, pressure, addition); + if (finger == PSEUDOFINGER_MOUSE_MOTION) + for (uint32_t i = 0; g_mouse_buttons >= (1 << i); i++) + if (g_mouse_buttons & (1 << i)) + g_input_callbacks.on_fmotion(device, PSEUDOFINGER_MOUSE_BUTTON_BASE + i, &motion, 1.0f, 0); + } + break; + case PLATFORM_EVENT_KEY_UP: + case PLATFORM_EVENT_KEY_DOWN: + if (event->key.sym == PLATFORM_KEY_NONE || event->key.repeat) + return; + if (g_input_callbacks.on_key) + g_input_callbacks.on_key(event->key.which, event->key.sym, + event->type == PLATFORM_EVENT_KEY_DOWN); + break; + case PLATFORM_EVENT_GAMEPAD_CONNECTED: + case PLATFORM_EVENT_GAMEPAD_DISCONNECTED: + if (g_input_callbacks.on_gstate) + g_input_callbacks.on_gstate(event->gdevice.which, + event->type == PLATFORM_EVENT_GAMEPAD_CONNECTED); + break; + case PLATFORM_EVENT_GAMEPAD_BUTTON_DOWN: + case PLATFORM_EVENT_GAMEPAD_BUTTON_UP: + if (event->gbutton.button == PLATFORM_GBUTTON_NONE) + return; + if (g_input_callbacks.on_gbutton) + g_input_callbacks.on_gbutton(event->gbutton.which, event->gbutton.button, + event->type == PLATFORM_EVENT_GAMEPAD_BUTTON_DOWN); + break; + case PLATFORM_EVENT_GAMEPAD_AXIS_MOTION: + if (event->gaxis.axis == PLATFORM_GAXIS_NONE) + return; + if (g_input_callbacks.on_gaxis) + g_input_callbacks.on_gaxis(event->gaxis.which, event->gaxis.axis, event->gaxis.value); + // TODO: For axises, consider having configurations of deadzones + // and provide the game functions to set these. + // TODO: Consider also making triggers buttons and simulate this + // by using deadzones to see if the button is pressed or unpressed + break; + } +} diff --git a/core/os.c b/core/os.c new file mode 100644 index 0000000..83ddc54 --- /dev/null +++ b/core/os.c @@ -0,0 +1,560 @@ +#include "platform.h" +#include "os.h" +#include + +struct os_timespec os_time(int type) +{ + struct platform_os_timespec platform_ts = {}; + struct os_timespec ts = {}; + int platform_type = type; + platform_ts.err = -1; + ts.err = -1; + if (type == OS_TIME_TICK) + platform_type = OS_TIME_MONO; + platform_ts = platform_os_time(platform_type); + if (platform_ts.err < 0) + return ts; + + // Various sanity checks on platform_ts + if (platform_ts.nsec < 0) { + platform_ts.nsec += 1000000000; + platform_ts.sec -= 1; + } else if (platform_ts.nsec >= 1000000000) { + platform_ts.nsec -= 1000000000; + platform_ts.sec += 1; + } + // Very wildly broken platform implementation + // that really needs fixing at that point + if (platform_ts.nsec < 0 || platform_ts.nsec >= 1000000000) + return ts; + + // TODO: + // monotonic time might not always be monotonic on platforms + // we should do our best here to make sure it is monotonic + // but we should try our best... later. + ts.sec = platform_ts.sec; + ts.nsec = platform_ts.nsec; + ts.err = 0; + static struct os_timespec os_tick = { + INT64_MIN, + INT32_MIN, + -1, + }; + if (type == OS_TIME_TICK) { + if (os_tick.err == -1) { + os_tick.sec = ts.sec; + os_tick.nsec = ts.nsec; + os_tick.err = 0; + } + ts.sec -= os_tick.sec; + ts.nsec -= os_tick.nsec; + if (ts.nsec < 0) { + ts.sec -= 1; + ts.nsec += 1000000000; + } + } + return ts; +} +float os_time2f(struct os_timespec ts) +{ + return (float)ts.sec + (float)(ts.nsec / 1e9f); +} +double os_time2d(struct os_timespec ts) +{ + return (double)ts.sec + (double)(ts.nsec / 1e9); +} +float os_timef(int type) +{ + return os_time2f(os_time(type)); +} +double os_timed(int type) +{ + return os_time2d(os_time(type)); +} + +size_t os_getpagesize(void) +{ + static size_t pagesize = 0; + if (pagesize == 0) { + // Get the actual platform result + size_t result = platform_os_getpagesize(); + + // Sanity check + // Must be at least 1 + if (result == 0) + result = 1; + + // Must be actually aligned + size_t aligned = result; + while (aligned & (aligned - 1)) + aligned &= (aligned - 1); + + // If it wasn't aligned before, round up + if (aligned != result) + result = (aligned << 1); + + // Now assign the result + pagesize = result; + } + return pagesize; +} + +int os_string_nullterm(struct stringc s) +{ + // If the string has 0 length, any pointer is possible + // Better to not try to play games with pointers at that point + if (s.len == 0) + return OS_STRING_NULLMAYBE; + + // First, check that there's no null terminator + for (size_t i = 0; i < s.len; i++) + if (!s.data[i]) + return OS_STRING_NULLEARLY; + size_t pagesize = os_getpagesize(); + + // If not page-aligned, then we can read one more byte + // If that one more byte is a null, then it's null termed + const char *end = s.data + s.len; + int status = OS_STRING_NULLMAYBE; + if (((uintptr_t)end & (pagesize - 1)) && !*end) + status = OS_STRING_NULLTERMED; + return status; +} + +char *os_string_nullterm_make(struct stringc s, struct allocator *a) +{ + int result = os_string_nullterm(s); + if (result == OS_STRING_NULLEARLY) + return (char *)0; + + char *nullterm = (char *)s.data; + if (result == OS_STRING_NULLMAYBE) { + nullterm = os_malloc_emulate_a(s.len + 1, a); + if (!nullterm) + return (char *)0; + + memcpy(nullterm, s.data, s.len); + nullterm[s.len] = 0; + } + return nullterm; +} + +void os_log_tag_v(int level, const char *tag, const char *fmt, va_list ap) +{ + return platform_os_log_v(level, tag, fmt, ap); +} + +void os_log_tag(int level, const char *tag, const char *fmt, ...) +{ + va_list ap; + va_start(ap, fmt); + os_log_tag_v(level, tag, fmt, ap); + va_end(ap); +} + +static const size_t os_alignment_default = 16; +struct blob os_realloc(struct allocator *a, struct blob prev, size_t size, size_t alignment) +{ + struct blob b = {}; + if (alignment == 0) + alignment = os_alignment_default; + if (alignment == 0 || (alignment & (alignment - 1))) + return b; + if (size == 0) { + b.memory = OS_ZERO_SIZE_PTR; + return b; + } + if (!a) + a = &os_allocator_default; + if (a->fns && a->fns->realloc_fn) + b = a->fns->realloc_fn(a, prev, size, alignment); + return b; +} +void os_free(struct allocator *a, struct blob mem) +{ + if (!mem.memory || mem.memory == OS_ZERO_SIZE_PTR) + return; + if (!a) + a = &os_allocator_default; + if (a->fns && a->fns->free_fn) + a->fns->free_fn(a, mem); +} +void os_free_all(struct allocator *a) +{ + if (!a) + a = &os_allocator_default; + if (a->fns && a->fns->free_all_fn) + a->fns->free_all_fn(a); +} +struct blob os_frealloc(struct allocator *a, struct blob prev, size_t size, size_t alignment) +{ + struct blob b = os_realloc(a, prev, size, alignment); + return b; +} +struct blob os_alloc(struct allocator *a, size_t size, size_t alignment) +{ + struct blob b = {}; + return os_realloc(a, b, size, alignment); +} + +struct os_malloc_chunk { + struct blob header; + char data[]; +}; +static inline struct os_malloc_chunk *os_malloc_chunk_from_p(void *p) +{ + return (struct os_malloc_chunk *)((struct blob *)p - 1); +} +struct blob os_blob_from_emulate(void *p) +{ + return os_malloc_chunk_from_p(p)->header; +} +void *os_realloc_emulate_a(void *p, size_t sz, struct allocator *a) +{ + // Check for overflows + size_t alloc_sz = sz + sizeof(struct blob); + if (alloc_sz < sz) return (void *)0; + + struct blob current = {}; + if (p) current = os_blob_from_emulate(p); + struct blob blob_new = os_realloc(a, current, alloc_sz, 0); + if (!blob_new.memory) return (void *)0; + + struct os_malloc_chunk *chunk = blob_new.memory; + chunk->header = blob_new; + return chunk->data; +} +void *os_malloc_emulate_a(size_t sz, struct allocator *a) +{ + return os_realloc_emulate_a((void *)0, sz, a); +} +void os_free_emulate_a(void *p, struct allocator *a) +{ + struct blob current = {}; + if (p) current = os_blob_from_emulate(p); + os_free(a, current); +} +void *os_realloc_emulate(void *p, size_t sz) +{ + return os_realloc_emulate_a(p, sz, (struct allocator *)0); +} +void *os_malloc_emulate(size_t sz) +{ + return os_malloc_emulate_a(sz, (struct allocator *)0); +} +void os_free_emulate(void *p) +{ + os_free_emulate_a(p, (struct allocator *)0); +} + +int os_strcmp_prefix(const char *s, const char *prefix) +{ + return strncmp(s, prefix, strlen(prefix)); +} +struct blob os_strdup(struct stringc s, struct allocator *a) +{ + struct blob b = os_alloc(a, s.len, 1); + if (b.memory) + memcpy(b.memory, s.data, s.len); + return b; +} +struct blob os_cstrdup(const char *s, struct allocator *a) +{ + return os_strdup(CSTR2STRC(s), a); +} + +void os_file_close(union os_file fp) +{ + if (!fp._i._i) return; + return platform_os_file_close(fp._i); +} +ssize_t os_file_read(union os_file fp, void *buf, size_t size) +{ + if (!fp._i._i) return -1; + return platform_os_file_read(fp._i, buf, size); +} +ssize_t os_file_write(union os_file fp, const void *buf, size_t size) +{ + if (!fp._i._i) return -1; + return platform_os_file_write(fp._i, buf, size); +} +int64_t os_file_seek(union os_file fp, int64_t delta, int from) +{ + if (!fp._i._i) return -1; + return platform_os_file_seek(fp._i, delta, from); +} +int64_t os_file_size(union os_file fp) +{ + if (!fp._i._i) return -1; + + // TODO: + // This is the dumb, stdio-like way of getting the file size. + // The correct way is to do a stat and get the file size but + // this is unimplemented right now (sadge...) + int64_t save = os_file_seek(fp, 0, OS_SEEK_CUR); + if (save < 0) + return -1; + int64_t size = os_file_seek(fp, 0, OS_SEEK_END); + os_file_seek(fp, save, OS_SEEK_SET); + if (size < 0) + return -1; + return size; +} + +union os_file os_asset_openc(int type, const char *filename) +{ + union os_file fp = {}; + if (!filename) + return fp; + // TODO: dynamically allocate path + char path[256]; + const char *prefix = (type == ASSET_BINARY) ? "binary" : "text"; + snprintf(path, sizeof(path), "%s/%s", prefix, filename); + fp._i = platform_os_asset_open(path); + return fp; +} +union os_file os_asset_open(int type, struct stringc filename, struct allocator *a) +{ + union os_file fp = {}; + char *nullterm = os_string_nullterm_make(filename, a); + if (!nullterm) + return fp; + fp = os_asset_openc(type, nullterm); + if (nullterm != filename.data) + os_free_emulate_a(nullterm, a); + return fp; +} + +struct blob os_file_read_all_extra(union os_file fp, struct allocator *a, size_t extra) +{ + struct blob b = {}; + + // TODO: Getting file size is common, so abstract this + // This also fails if file is not seekable + int64_t size = os_file_seek(fp, 0, OS_SEEK_END); + int64_t status = size; + if (size < 0) + return b; + status = os_file_seek(fp, 0, OS_SEEK_SET); + if (status < 0) + return b; + + int64_t alloc_size = size + extra; + if (size > (INT32_MAX - 1)) + return b; + b = os_alloc(a, alloc_size, 1); + if (!b.memory) + return b; + for (ssize_t readtotal = 0, readlen = 0; readtotal < size;) { + readlen = os_file_read(fp, (unsigned char *)b.memory + readtotal, size - readtotal); + if (readlen <= 0) { + os_free(a, b); + b = (struct blob){}; + break; + } + readtotal += readlen; + } + return b; +} + +// TODO: +// Paths should be cleaned and normalized before passing to the platform +// Paths should always be a POSIX path with the platform turning it into a useful OS path +union os_file os_url_open(struct stringc url, struct allocator *a) +{ + union os_file fp = {}; + if (!url.data || !url.len) + return fp; + + // This is... Ew... and has common functionality that can be refactored + struct stringc prefix = STRLIT("app://"); + const char *iter = url.data; + size_t iter_len = url.len; + if (iter_len >= prefix.len && !memcmp(iter, prefix.data, prefix.len)) { + iter += prefix.len; + iter_len -= prefix.len; + + int type = -1; + if (prefix = STRLIT("binary/"), (iter_len >= prefix.len && !memcmp(iter, prefix.data, prefix.len))) { + iter += prefix.len; + iter_len -= prefix.len; + type = ASSET_BINARY; + } else if (prefix = STRLIT("text/"), (iter_len >= prefix.len && !memcmp(iter, prefix.data, prefix.len))) { + iter += prefix.len; + iter_len -= prefix.len; + type = ASSET_TEXT; + } + if (type >= 0) + fp = os_asset_open(type, (struct stringc){iter, iter_len}, a); + else + OSLOGE("URL '"STRFMT"' represents neither binary nor text asset!", STRARGS(url)); + } + + // Nothing else implemented + return fp; +} +union os_file os_url_openc(const char *url, struct allocator *a) +{ + return os_url_open(CSTR2STRC(url), a); +} + +struct image os_image_decode(const void *mem, size_t size, struct allocator *a) +{ + return platform_os_image_decode(mem, size, a); +} + +// Default allocator +static struct blob realloc_fn(struct allocator *a, struct blob prev, size_t size, size_t align) +{ + (void)a; + + // TODO: Don't ignore alignment + (void)align; + + struct blob b = {}; + b.memory = realloc(prev.memory, size); + if (b.memory) { + b.size = size; + b.alignment = align; + b.capacity = size; + } + return b; +} + +static void free_fn(struct allocator *a, struct blob mem) +{ + (void)a; + free(mem.memory); +} + +struct allocator_functions os_allocator_default_functions = { + realloc_fn, + free_fn, + + // libc malloc APIs have no concept of freeing everything + (void (*)(struct allocator *))0, +}; +struct allocator os_allocator_default = {&os_allocator_default_functions}; + +static struct blob arena_realloc_fn(struct allocator *a, struct blob prev, size_t size, size_t align) +{ + struct blob b = {}; + if (size <= prev.capacity && ((uintptr_t)prev.memory & (align - 1)) == 0) { + b = prev; + b.size = size; + return b; + } + + // TODO: Don't assume arena memory itself is aligned + struct arena *rena = (struct arena *)a; + size_t ptr_new = rena->ptr; + if (prev.memory == &rena->b.memory[rena->ptr_last]) + ptr_new = rena->ptr_last; + ptr_new += align - 1; + if (ptr_new < (align - 1)) + return b; + ptr_new &= ~(align - 1); + if (rena->b.capacity - ptr_new < size) + return b; + rena->ptr_last = ptr_new; + b.memory = &rena->b.memory[ptr_new]; + b.size = size; + b.alignment = align; + b.capacity = size; + rena->ptr = ptr_new + size; + return b; +} +static void arena_free_fn(struct allocator *a, struct blob mem) +{ + struct arena *rena = (struct arena *)a; + if (mem.memory == &rena->b.memory[rena->ptr_last]) + rena->ptr = rena->ptr_last; + // Otherwise, allocator doesn't know how to free it +} +static void arena_free_all_fn(struct allocator *a) +{ + struct arena *rena = (struct arena *)a; + rena->ptr_last = 0; + rena->ptr = 0; +} + +struct allocator_functions os_allocator_arena_functions = { + arena_realloc_fn, + arena_free_fn, + arena_free_all_fn, +}; +struct allocator os_allocator_arena = {&os_allocator_arena_functions}; + +// TODO We don't have a default temporary allocator right now +struct allocator_functions os_allocator_default_temp_functions = {}; +struct allocator os_allocator_default_temp = {&os_allocator_default_temp_functions}; + +uint32_t handle_table_alloc(struct handle_table *t) +{ + // If there's nothing in the free list, get a new entry from arena + uint32_t handle = 0; + if (t->head == 0) { + if (t->top == 0) + t->top = 1; + if (t->top >= (1 << t->bits_idx)) + return handle; + handle = t->top++; + t->table[handle] = 0; + return handle; + } + + // Pop an entry from the head of the free list + uint32_t index_mask = (1 << t->bits_idx) - 1; + uint32_t gen_mask = ((1 << t->bits_gen) - 1) << t->bits_idx; + uint32_t index = t->head & index_mask; + handle = t->table[index]; + t->head = handle & index_mask; + handle = (handle & ~index_mask & gen_mask) | index; + return handle; +} +uint32_t handle_table_check(struct handle_table *t, uint32_t handle) +{ + // Get generation and index from handle + uint32_t index_mask = (1 << t->bits_idx) - 1; + uint32_t gen_mask = ((1 << t->bits_gen) - 1) << t->bits_idx; + uint32_t index = handle & index_mask; + uint32_t gen = handle & ~index_mask & gen_mask; + + // Get the generation of entry and check it against handle + // TODO: Should probably log an index failure + if (index >= t->top) + return 0; + uint32_t entry = t->table[index]; + uint32_t gen_check = entry & ~index_mask & gen_mask; + + // TODO: Should probably log a generation check failure + if (gen_check != gen) + return 0; + return index; +} +int handle_table_free(struct handle_table *t, uint32_t handle) +{ + uint32_t index = handle_table_check(t, handle); + if (!index) + return -2; + uint32_t index_mask = (1 << t->bits_idx) - 1; + + // Increment the generation and push it on free list + t->table[index] = ((t->table[index] & ~index_mask) | t->head) + (index_mask + 1); + t->head = index; + return 0; +} +// Convenience function for reallocing an already allocated entry +uint32_t handle_table_realloc(struct handle_table *t, uint32_t handle_prev) +{ + return (handle_prev == 0 || handle_table_free(t, handle_prev) == -1) + ? 0 : handle_table_alloc(t); +} +// Clear the entire table, but don't zero everything out +// so that later virtual pages not used yet don't get allocated +void handle_table_clear(struct handle_table *t) +{ + memset(t->table, 0, sizeof(*t->table) * t->top); + t->top = 1; + t->head = 0; +} diff --git a/game/game_callbacks.c b/game/game_callbacks.c new file mode 100644 index 0000000..a78d0ea --- /dev/null +++ b/game/game_callbacks.c @@ -0,0 +1,542 @@ +#include "game_callbacks.h" +#include "os.h" +#include "gfx.h" +#include "afx.h" +#include "input.h" +#include "sprite.h" + +// Exported for sprite.c +union gfx_sampler g_sampler; + +static union gfx_pipeline g_pipeline; +static union gfx_shader g_shader; +static union afx_sound g_missle_shot; +static union afx_sound g_explosion_sound; + +// The text that we render is always a single 8-bit channel, +// unlike our assets which are currently RGBA, so we need a +// different fragment shader and pipeline. +// We can use the same vertex shader, ofc. +static union gfx_shader g_font_shader; +static union gfx_pipeline g_font_pipeline; +static union gfx_font g_font; + +static float timeprev; +static float timedelta; +static float reset_tick = 0.0f; + +#define FLAG_LEFT 1 +#define FLAG_RIGHT 2 +#define FLAG_UP 4 +#define FLAG_DOWN 8 +#define SHIP_FIRERATE 5.0f +#define EXPLOSION_TIME 0.2f +#define PELLET_WIDTH 16.0f +#define EXPLOSION_WIDTH 128.0f +#define EXPLOSION_RADIUS (EXPLOSION_WIDTH * 0.45f) +#define MISSLE_WIDTH 32.0f +#define SHIP_WIDTH 64.0f +#define SHIP_RADIUS (SHIP_WIDTH * 0.3f) +#define RETICLE_WIDTH 64.0f +#define PELLET_SPAWN_BASE 0.5f + +// TODO: Define in a more general math file +#define EPSILON 0.001f + +#define MISSLE_RECT (struct rectanglei32){0, 0, 32, 32} +#define PELLET_RECT (struct rectanglei32){0, 32, 32, 32} +#define RETICLE_RECT (struct rectanglei32){0, 64, 32, 32} +#define EXPLOSION_RECT (struct rectanglei32){0, 96, 32, 32} +#define SHIP_RECT (struct rectanglei32){0, 128, 32, 32} +static union gfx_texture g_atlas; + +static struct { + struct vector2f32 pos; + float firetime; + uint32_t firefinger; + struct vector2f32 firepos; + uint32_t movefinger; + struct vector2f32 movedelta; + float destroyed; +} g_ship = {}; +static uint32_t g_flags; +static float g_pellet_spawn_time; +static struct vector2f32 g_jiggle; +static struct rectanglef32 g_camera; + +#define CAM_DIM_X 450.0f +#define CAM_DIM_Y 800.0f +#define JIGGLE_MAX 0.05f +#define PROJ_X_HIGH (CAM_DIM_X * (1.0f + JIGGLE_MAX)) +#define PROJ_X_LOW (-PROJ_X_HIGH) +#define PROJ_Y_HIGH (CAM_DIM_Y * (1.0f + JIGGLE_MAX)) +#define PROJ_Y_LOW (-PROJ_Y_HIGH) + +enum projectile_type { + PROJ_NONE, + PROJ_PELLET, + PROJ_MISSLE, + PROJ_EXPLOSION, +}; + +// Projectiles in our bullet hell are fire-and-forget, +// so they don't return handles. They just are there +// and will do their own thing without intervention. +// +// This is nice programmatically because then we can then move and +// compact projectiles without having to update references +// +// TODO: +// In future games, we want to update the engine so that +// we can handle returning handles better +struct projectile { + uint32_t type; + union { + struct vector2f32 pos; + struct { + struct vector2f32 pos; + float speed; + float rot; + float distance_left; + } missle; + struct { + struct vector2f32 pos; + struct vector2f32 vel; + } pellet; + struct { + struct vector2f32 pos; + float timeleft; + } explosion; + }; +}; + +struct projectiles { + int count; + size_t alloc_count; + struct blob alloc_mem; + struct projectile *data; +}; +static struct projectiles g_friendly = {}; +static struct projectiles g_enemy = {}; +static struct projectile *add_projectile(struct projectiles *projectiles) +{ + if ((size_t)projectiles->count >= projectiles->alloc_count) { + size_t alloc_count = projectiles->alloc_count << 1; + if (alloc_count < 4096) + alloc_count = 4096; + size_t projsize = sizeof(struct projectile); + struct blob mem = os_realloc((struct allocator *)0, projectiles->alloc_mem, projsize * alloc_count, 0); + if (!mem.memory) + return (struct projectile *)0; + projectiles->alloc_mem = mem; + projectiles->data = (struct projectile *)mem.memory; + projectiles->alloc_count = alloc_count; + } + int idx = projectiles->count++; + if (idx == 0) { + memset(projectiles->data, 0, sizeof(*projectiles->data)); + idx = projectiles->count++; + } + return &projectiles->data[idx]; +} +static void add_missle(struct projectiles *projectiles, struct vector2f32 pos, float speed, float rot, float distleft) +{ + struct projectile *p = add_projectile(projectiles); + if (!p) + return; + p->type = PROJ_MISSLE; + p->missle.pos = pos; + p->missle.speed = speed; + p->missle.rot = rot; + p->missle.distance_left = distleft; + afx_sound_restart(g_missle_shot); +} +static void add_explosion(struct projectiles *projectiles, struct vector2f32 pos) +{ + struct projectile *p = add_projectile(projectiles); + if (!p) + return; + p->type = PROJ_EXPLOSION; + p->explosion.pos = pos; + p->explosion.timeleft = EXPLOSION_TIME; + afx_sound_restart(g_explosion_sound); +} +static void add_pellet(struct projectiles *projectiles, struct vector2f32 pos, struct vector2f32 vel) +{ + struct projectile *p = add_projectile(projectiles); + if (!p) + return; + p->type = PROJ_PELLET; + p->pellet.pos = pos; + p->pellet.vel = vel; +} +static void remove_projectile(struct projectiles *projectiles, int proj) +{ + if (proj <= 0 || proj >= projectiles->count) + return; + projectiles->data[proj] = projectiles->data[--projectiles->count]; +} +static void update_projectiles(struct projectiles *projectiles) +{ + for (int i = 0; i < projectiles->count; i++) { + struct projectile *p = &projectiles->data[i]; + switch (p->type) { + case PROJ_MISSLE: { + float distance = p->missle.speed * timedelta; + v2f32addself(&p->pos, v2f32muls(v2f32sincos(p->missle.rot), distance)); + p->missle.distance_left -= distance; + if (p->missle.distance_left <= 0.0f) { + struct vector2f32 pos = p->pos; + remove_projectile(projectiles, i); + add_explosion(projectiles, pos); + continue; + } + } break; + case PROJ_EXPLOSION: + p->explosion.timeleft -= timedelta; + if (p->explosion.timeleft <= 0.0f) { + remove_projectile(projectiles, i); + } else { + float scale = p->explosion.timeleft / EXPLOSION_TIME; + struct vector2f32 r = v2muls(v2f32rand1signed(), scale); + v2f32addself(&g_jiggle, r); + + // TODO: Yes, I know. O(N**2) terribleness and assumes that + // explosions only come from friendlies. It works for now. + for (int i = 0; i < g_enemy.count; i++) { + struct projectile *enemy = &g_enemy.data[i]; + if (enemy->type != PROJ_PELLET) + continue; + struct vector2f32 delta = v2f32sub(enemy->pos, p->pos); + // TODO: Yes, I also know this isn't exact. I'll fix that... Maybe. + if (v2f32mag2(delta) < (EXPLOSION_RADIUS * EXPLOSION_RADIUS)) + remove_projectile(&g_enemy, i); + } + } + break; + case PROJ_PELLET: + v2f32addself(&p->pellet.pos, v2f32muls(p->pellet.vel, timedelta)); + struct vector2f32 bounds = {(CAM_DIM_X + PELLET_WIDTH) / 2.0f, (CAM_DIM_Y + PELLET_WIDTH) / 2.0f}; + if (p->pellet.pos.x < -bounds.x || p->pellet.pos.x > bounds.x + || p->pellet.pos.y < -bounds.y || p->pellet.pos.y > bounds.y) { + remove_projectile(projectiles, i); + } else if (g_ship.destroyed <= EPSILON) { + struct vector2f32 delta = v2f32sub(p->pellet.pos, g_ship.pos); + if (v2f32mag2(delta) < SHIP_RADIUS * SHIP_RADIUS) { + g_ship.destroyed = EPSILON * 2.0f; + add_explosion(&g_friendly, p->pellet.pos); + remove_projectile(projectiles, i); + // Assumes all pellets are enemies, which is true now + } + } + break; + } + } +} +static void render_projectiles(struct projectiles *projectiles) +{ + for (int i = 0; i < projectiles->count; i++) { + struct projectile *p = &projectiles->data[i]; + float rotation = 0.0f; + struct vector2f32 size = {}; + struct rectanglei32 rect = {}; + + switch (p->type) { + case PROJ_PELLET: + rect = PELLET_RECT; + size = (struct vector2f32){PELLET_WIDTH, PELLET_WIDTH}; + break; + case PROJ_MISSLE: + rect = MISSLE_RECT; + size = (struct vector2f32){MISSLE_WIDTH, MISSLE_WIDTH}; + rotation = p->missle.rot; + break; + case PROJ_EXPLOSION: + rect = EXPLOSION_RECT; + size = (struct vector2f32){EXPLOSION_WIDTH, EXPLOSION_WIDTH}; + break; + } + struct sprite s; + s = (struct sprite){ + .tex = rect, + .space = { p->pos.x, p->pos.y, size.x, size.y, }, + .rotation = rotation, + }; + sprite_draw_one(g_atlas, &s); + } +} + +static void update_ship(void) +{ + if (g_ship.destroyed > EPSILON) { + g_ship.destroyed += timedelta; + return; + } + + struct vector2f32 vel = {}; + float mag = 256.0f; + if (g_flags & FLAG_LEFT) + vel.x -= mag; + if (g_flags & FLAG_RIGHT) + vel.x += mag; + if (g_flags & FLAG_UP) + vel.y += mag; + if (g_flags & FLAG_DOWN) + vel.y -= mag; + + g_ship.pos = v2f32add(g_ship.pos, v2f32muls(vel, timedelta)); + float xbound = (CAM_DIM_X - SHIP_WIDTH) / 2; + float ybound = (CAM_DIM_Y - SHIP_WIDTH) / 2; + if (g_ship.pos.x < -xbound) + g_ship.pos.x = -xbound; + if (g_ship.pos.x > xbound) + g_ship.pos.x = xbound; + if (g_ship.pos.y < -ybound) + g_ship.pos.y = -ybound; + if (g_ship.pos.y > ybound) + g_ship.pos.y = ybound; + + if (g_ship.firetime > 0.0f) + g_ship.firetime -= timedelta; + if (g_ship.firefinger) { + if (g_ship.firetime <= 0.0f) { + struct vector2f32 delta = v2f32sub(g_ship.firepos, g_ship.pos); + float mag = v2mag(delta); + if (mag > 1e-4) { + float angle = v2atan(delta); + add_missle(&g_friendly, + g_ship.pos, + 600.0f, + angle, + mag + ); + } + g_ship.firetime = 1.0f / SHIP_FIRERATE; + } + float scale = g_ship.firetime * SHIP_FIRERATE; + struct vector2f32 r = v2muls(v2f32rand1signed(), scale); + + v2f32addself(&g_jiggle, r); + } +} +static void render_ship(void) +{ + if (g_ship.destroyed > EPSILON) + return; + + struct sprite s = { + .tex = SHIP_RECT, + .space = { g_ship.pos.x, g_ship.pos.y, SHIP_WIDTH, SHIP_WIDTH }, + }; + sprite_draw_one(g_atlas, &s); + if (g_ship.firefinger) { + s = (struct sprite) { + .tex = RETICLE_RECT, + .space = { g_ship.firepos.x, g_ship.firepos.y, RETICLE_WIDTH, RETICLE_WIDTH }, + }; + sprite_draw_one(g_atlas, &s); + } +} + +static void reset_game() +{ + g_friendly.alloc_count = 0; + g_friendly.count = 0; + g_enemy.alloc_count = 0; + g_enemy.count = 0; + memset(&g_ship, 0, sizeof(g_ship)); + g_flags = 0; + g_pellet_spawn_time = PELLET_SPAWN_BASE; + reset_tick = os_timef(OS_TIME_TICK); +} +static void on_key(uint32_t keyboard, uint32_t key, bool down) +{ + (void)keyboard; + uint32_t bit = 0; + switch (key) { + case KEY_W: + bit = FLAG_UP; + break; + case KEY_S: + bit = FLAG_DOWN; + break; + case KEY_A: + bit = FLAG_LEFT; + break; + case KEY_D: + bit = FLAG_RIGHT; + break; + case KEY_R: + reset_game(); + break; + } + if (down) + g_flags |= bit; + else + g_flags &= ~bit; +} +static void on_fmotion(uint32_t device, uint32_t finger, const struct input_motion *motion, float pressure, int addition) +{ + (void)device, (void)pressure; + struct vector2f32 pos = v2f32mul(motion->normalized, (struct vector2f32){ + CAM_DIM_X / 2.0f, CAM_DIM_Y / 2.0f + }); + if (addition == 1) { + if (g_ship.destroyed > 2.0f * EXPLOSION_TIME) { + reset_game(); + return; + } + struct vector2f32 delta = v2sub(pos, g_ship.pos); + float mag = v2mag(delta); + if (mag < (SHIP_WIDTH * 1.2f) && finger < PSEUDOFINGER_MOUSE_BUTTON_BASE && !g_ship.movefinger) { + g_ship.movefinger = finger; + g_ship.movedelta = delta; + } else { + g_ship.firefinger = finger; + } + } else if (addition == -1) { + if (finger == g_ship.firefinger) + g_ship.firefinger = 0; + if (finger == g_ship.movefinger) + g_ship.movefinger = 0; + } + if (g_ship.firefinger == finger) + g_ship.firepos = pos; + else if (g_ship.movefinger == finger) + g_ship.pos = v2f32sub(pos, g_ship.movedelta); +} + +void game_callback_entry(int argc, char **argv) { + // Args useless for now since we only handle mobile + (void)argc; + (void)argv; + + // Temporary tests + struct handle_table handles = HANDLE_TABLE_STATIC(4,1); + uint32_t h1 = handle_table_alloc(&handles); + uint32_t h2 = handle_table_alloc(&handles); + OSLOGI("Handles from handle table... %08x,%08x", h1, h2); + handle_table_free(&handles, h1); + h1 = handle_table_alloc(&handles); + h2 = handle_table_realloc(&handles, h2); + OSLOGI("Handles from handle table realloc'd... %08x,%08x", h1, h2); + h1 = handle_table_realloc(&handles, h1); + h2 = handle_table_realloc(&handles, h2); + OSLOGI("Handles from handle table realloc'd again (generation should wrap)... %08x,%08x", h1, h2); + handle_table_clear(&handles); + OSLOGI("Handle free after table clear should fail... %d, %d", + handle_table_free(&handles,h1), + handle_table_free(&handles,h2)); + h1 = handle_table_alloc(&handles); + h2 = handle_table_alloc(&handles); + OSLOGI("Handles from handle table after clear... %08x,%08x", h1, h2); + + // Set the input callbacks for the game and seed randomness + input_set_callbacks(&(struct input_callbacks){ + .on_key = on_key, + .on_fmotion = on_fmotion, + }); + srand((int32_t)os_timed(OS_TIME_WALL)); + timeprev = os_timef(OS_TIME_TICK); +} + +// If working correctly, this is called once and the user +// no longer needs to worry about recreating resources destroyed +// when the window is created or destroyed +static void gfx_once(void) +{ + OSLOGD("Logging gfx once!"); + g_shader = gfx_shader_construct(&(struct gfx_shader_desc){ + .vertex_url = STRLIT("app://text/texture.vert"), + .fragment_url = STRLIT("app://text/texture.frag"), + }); + g_pipeline = gfx_pipeline_construct(&(struct gfx_pipeline_desc){ + .shader = g_shader, + }); + g_sampler = gfx_sampler_construct(&(struct gfx_sampler_desc){ + }); + g_atlas = gfx_texture_construct(&(struct gfx_texture_desc){ + .url = STRLIT("app://binary/atlas.png"), + }); + g_font = gfx_font_construct(&(struct gfx_font_desc){ + .url = STRLIT("app://binary/LiberationSerif-Regular.ttf"), + }); + gfx_font_size_set(g_font, 24.0f); + g_font_shader = gfx_shader_construct(&(struct gfx_shader_desc){ + .vertex_url = STRLIT("app://text/texture.vert"), + .fragment_url = STRLIT("app://text/font.frag"), + }); + g_font_pipeline = gfx_pipeline_construct(&(struct gfx_pipeline_desc){ + .shader = g_font_shader, + }); + + g_missle_shot = afx_sound_construct(STRLIT("app://binary/missle_shot.wav"), AFX_SOUND_FX); + g_explosion_sound = afx_sound_construct(STRLIT("app://binary/explosion.wav"), AFX_SOUND_FX); +} + +void game_callback_gfx_startup(void) +{ + static union os_once g_gfx_once = OS_ONCE_INIT; + os_once_run(&g_gfx_once, gfx_once); +} + +void game_callback_gfx_shutdown(void) +{ +} + + +void game_callback_variable(void) +{ + g_camera = (struct rectanglef32){0.0f, 0.0f, CAM_DIM_X, CAM_DIM_Y}; + v2f32clampself(&g_jiggle, v2f32s(-1.0f), v2f32s(1.0f)); + v2f32mulselfs(&g_jiggle, CAM_DIM_X * JIGGLE_MAX); + rectf32moveself(&g_camera, g_jiggle); + g_jiggle = (struct vector2f32){}; + sprite_camera(g_camera); + + // Generally update all the entities + update_projectiles(&g_enemy); + update_projectiles(&g_friendly); + update_ship(); + + // Occasionally spawn enemy bullets + if ((g_pellet_spawn_time -= timedelta) < 0.0f) { + float dstx_sector = (rand() % 7 - 3) / 3.5f; + float srcx_sector = (rand() % 7 - 3) / 3.5f; + float scale = ((os_timef(OS_TIME_TICK) - reset_tick) / 10.0f + 1.0f); + + struct vector2f32 pos = (struct vector2f32){ + srcx_sector * (CAM_DIM_X / 2.0f), + CAM_DIM_Y / 2.0f + }; + float vely = -128.0f - scale * 256.0f * (rand() % RAND_MAX) / (float)RAND_MAX; + float dstx = dstx_sector * (CAM_DIM_X / 2.0f); + struct vector2f32 dst = {dstx, -CAM_DIM_Y / 2.0f}; + float slope_inv = (dst.x - pos.x) / -CAM_DIM_Y; + float velx = vely * slope_inv; + + pos.y = CAM_DIM_Y / 2.0f; + add_pellet(&g_enemy, pos, (struct vector2f32){velx, vely}); + g_pellet_spawn_time = PELLET_SPAWN_BASE / scale; + } + + timedelta = os_timef(OS_TIME_TICK) - timeprev; + timeprev += timedelta; + if (timedelta > 0.1f) + timedelta = 0.1f; +} + +void game_callback_render(void) +{ + // Bind the pipeline for sprites and render + gfx_pipeline_bind(g_pipeline); + render_projectiles(&g_enemy); + render_projectiles(&g_friendly); + render_ship(); + sprite_flush(); + + // Now bind the pipeline for fonts and render + gfx_pipeline_bind(g_font_pipeline); + text_draw(g_font, STRLIT("Hello World!\nHow are you?"), (struct vector2f32){64.0f,64.0f}); + sprite_flush(); +} diff --git a/game/sprite.c b/game/sprite.c new file mode 100644 index 0000000..0721b49 --- /dev/null +++ b/game/sprite.c @@ -0,0 +1,123 @@ +#include "sprite.h" +#include "os.h" + +#define MAX_BATCH_SIZE 4095 +static struct spritebatch +{ + size_t count; + union gfx_texture tex; + struct sprite_vertex batch[MAX_BATCH_SIZE]; +} g_spritebatch = {}; + +// See game_callbacks.c for this +extern union gfx_sampler g_sampler; +static struct rectanglef32 g_camera = {}; + +void sprite_flush(void) +{ + size_t vlen = g_spritebatch.count; + size_t vbytes = vlen * sizeof(*g_spritebatch.batch); + union gfx_texture tex = g_spritebatch.tex; + g_spritebatch.count = 0; + g_spritebatch.tex = (union gfx_texture){}; + if (!tex._i || !vlen) + return; + gfx_vertices_draw(tex, g_sampler, g_spritebatch.batch, vbytes, vlen); +} + +void sprite_vertices_draw(union gfx_texture tex, const struct sprite_vertex *verts, size_t bytesize) +{ + if (g_spritebatch.tex._i != tex._i) { + sprite_flush(); + g_spritebatch.tex = tex; + } + size_t vcount = bytesize / sizeof(*verts); + for (size_t i = 0; i < vcount; i++) { + if (g_spritebatch.count >= MAX_BATCH_SIZE) { + sprite_flush(); + g_spritebatch.tex = tex; + } + g_spritebatch.batch[g_spritebatch.count++] = verts[i]; + } +} + +void sprite_draw_one(union gfx_texture tex, const struct sprite *sprite) +{ + struct vector2f32 spacewh = { + sprite->space.width, + sprite->space.height + }; + struct matrix2f32 rotm = m2f32rot(sprite->rotation); + struct vector2f32 xy[4] = { + {-spacewh.x / 2, -spacewh.y / 2}, + {spacewh.x / 2, -spacewh.y / 2}, + {-spacewh.x / 2, spacewh.y / 2}, + {spacewh.x / 2, spacewh.y / 2} + }; + struct vector2f32 uv[4] = { + {(float)sprite->tex.x, (float)(sprite->tex.y + sprite->tex.height)}, + {(float)(sprite->tex.x + sprite->tex.width), (float)(sprite->tex.y + sprite->tex.height)}, + {(float)sprite->tex.x, (float)sprite->tex.y}, + {(float)(sprite->tex.x + sprite->tex.width), (float)sprite->tex.y}, + }; + struct rectanglei32 texrect = gfx_texture_rect(tex); + for (size_t i = 0; i < 4; i++) { + // Convert to UV + v2f32divself(&uv[i], (struct vector2f32)rectdims(texrect)); + + // Apply rotation + m2f32mulv2f32self(rotm, &xy[i]); + + // Now apply translations + scales from the camera + v2f32addself(&xy[i], v2f32sub(rectf32pos(sprite->space), rectf32pos(g_camera))); + v2f32divself(&xy[i], v2f32muls(rectf32dims(g_camera), 0.5f)); + } + struct sprite_vertex vertices[6] = { + {xy[0], uv[0]}, + {xy[1], uv[1]}, + {xy[2], uv[2]}, + {xy[1], uv[1]}, + {xy[2], uv[2]}, + {xy[3], uv[3]}, + }; + sprite_vertices_draw(tex, vertices, sizeof(vertices)); +} + +void sprite_draw(union gfx_texture tex, const struct sprite *s, size_t bytesize) { + size_t scount = bytesize / sizeof(*s); + for (size_t i = 0; i < scount; i++) + sprite_draw_one(tex, &s[i]); +} + +void text_draw(union gfx_font f, struct stringc text, struct vector2f32 loc) +{ + union gfx_texture t = gfx_font_texture(f); + struct gfx_font_info info = gfx_font_info_get(f); + float pxscale = info.fontsize / (info.ascent - info.descent); + float line_height = pxscale * (info.ascent - info.descent + info.linegap); + float xbase = loc.x; + for (size_t i = 0; i < text.len; i++) { + if (text.data[i] == '\n') { + loc.y -= line_height; + loc.x = xbase; + continue; + } + + struct gfx_font_bmglyph glyph = gfx_font_bmglyph_get(f, text.data[i], 0); + if (glyph.xadvance) { + struct sprite s = { + .tex = {glyph.x, glyph.y, glyph.width, glyph.height}, + .space = {loc.x + glyph.xoffset, loc.y + glyph.yoffset, glyph.width, glyph.height}, + }; + s.space.x += s.space.width / 2.0f; + s.space.y += s.space.height / 2.0f; + sprite_draw_one(t, &s); + loc.x += glyph.xadvance; + } + } +} + +void sprite_camera(struct rectanglef32 cam) +{ + memcpy(&g_camera, &cam, sizeof(g_camera)); +} diff --git a/game/sprite.h b/game/sprite.h new file mode 100644 index 0000000..6319dbb --- /dev/null +++ b/game/sprite.h @@ -0,0 +1,37 @@ +#ifndef HBAT_SPRITE_H +#define HBAT_SPRITE_H + +#include "types.h" +#include "gfx.h" + + +struct sprite_vertex { + struct vector2f32 xy; + struct vector2f32 uv; +}; + +struct sprite { + struct rectanglei32 tex; + struct rectanglef32 space; + float rotation; +}; + +// Spritebatch functions + +// Custom Vertices draw function +void sprite_vertices_draw(union gfx_texture tex, const struct sprite_vertex *verts, size_t bytesize); + +// Flush spritebatch buffer +void sprite_flush(void); + +// Sprites are always drawn in relation to a texture and never by themselves +void sprite_draw(union gfx_texture tex, const struct sprite *s, size_t bytesize); +void sprite_draw_one(union gfx_texture tex, const struct sprite *s); + +void text_draw(union gfx_font font, struct stringc text, struct vector2f32 loc); + + +// Set the 2D camera to use in the spritebatch for translation + scaling +void sprite_camera(struct rectanglef32 cam); + +#endif diff --git a/include/afx.h b/include/afx.h new file mode 100644 index 0000000..b3320db --- /dev/null +++ b/include/afx.h @@ -0,0 +1,31 @@ +#ifndef HBAT_AFX_H +#define HBAT_AFX_H + +#include "types.h" + +enum afx_sound_flag { + AFX_SOUND_STREAM = 1, + AFX_SOUND_DECODE = 2, + AFX_SOUND_PLAY_NOW = 4, + AFX_SOUND_LOOP_NOW = 8, + + // Common flags used for sound effects and background music respectively + AFX_SOUND_FX = AFX_SOUND_DECODE, + AFX_SOUND_BG = AFX_SOUND_STREAM | AFX_SOUND_PLAY_NOW | AFX_SOUND_LOOP_NOW, +}; + +struct afx_sound_i; +union afx_sound { struct afx_sound_i *_i; }; + +// TODO: add "sound groups" to afx so that sounds can be processed differently +// (i.e. Master vs. SFX vs. Background volumes) +union afx_sound afx_sound_construct(struct stringc url, int flags); +void afx_sound_destruct(union afx_sound sound); +void afx_sound_restart(union afx_sound sound); +void afx_sound_stop(union afx_sound sound); +void afx_sound_loop(union afx_sound sound); +void afx_sound_unloop(union afx_sound sound); + +void afx_start(void); +void afx_stop(void); +#endif diff --git a/include/compiler.h b/include/compiler.h new file mode 100644 index 0000000..c0f1885 --- /dev/null +++ b/include/compiler.h @@ -0,0 +1,33 @@ +#ifndef HBAT_COMPILER_H +#define HBAT_COMPILER_H + +// VA_ARGS are annoying so have the one way to rule them all +#define VAOPT(...) __VA_OPT__(, ) __VA_ARGS__ + +// C and C++ do things differently +#ifdef __cplusplus +#define LICAST(s, t) t +#define THREADLOCAL thread_local +#define ALIGNOF(x) alignof(x) +#define ALIGNAS(x) alignas(x) +#else +#define LICAST(s, t) (s t) +#define THREADLOCAL __thread +#define ALIGNOF(x) __alignof__(x) +#define ALIGNAS(x) __attribute__((aligned(x))) +#endif + +// Not required, but some people like this cast style +#define CAST(t) (t) + +// Very architecture-specific +// It's 4096 for all architectures we care about but +// only use this as a last resort when OS APIs fail +#ifndef PAGE_SIZE +#define PAGE_SIZE 4096 +#endif + +// Fallthrough macro +#define FALLTHROUGH __attribute__((fallthrough)) + +#endif diff --git a/include/core_callbacks.h b/include/core_callbacks.h new file mode 100644 index 0000000..24f53f7 --- /dev/null +++ b/include/core_callbacks.h @@ -0,0 +1,23 @@ +#ifndef HBAT_CORE_CALLBACKS_H +#define HBAT_CORE_CALLBACKS_H + +#include "platform.h" +#include "types.h" + +// Initialization function with argc, argv for desktop +void core_callback_entry(int argc, char **argv); + +// Important because, on mobile, graphics contexts get invalidated +// so the game needs to invalidate and recreate resources again +void core_callback_gfx_startup(union platform_window window); +void core_callback_gfx_shutdown(union platform_window window); + +// When a platform event comes (like input), respond using this callback +void core_callback_event(union platform_event *event, union platform_window window); + +// This will also call the game's update function +// and will only call the game's render function +// if the window has valid gfx on it +void core_callback_render(union platform_window window); + +#endif diff --git a/include/game_callbacks.h b/include/game_callbacks.h new file mode 100644 index 0000000..f935061 --- /dev/null +++ b/include/game_callbacks.h @@ -0,0 +1,15 @@ +#ifndef HBAT_GAME_CALLBACKS_H +#define HBAT_GAME_CALLBACKS_H + +#include "platform.h" +#include "types.h" + +// Corresponding functions in core_callbacks.h +void game_callback_entry(int argc, char **argv); +void game_callback_gfx_startup(void); +void game_callback_gfx_shutdown(void); +void game_callback_event(void); // TODO: Want a cross-platform event here +void game_callback_variable(void); // TODO: Fixed update callback +void game_callback_render(void); + +#endif diff --git a/include/gfx.h b/include/gfx.h new file mode 100644 index 0000000..1690e01 --- /dev/null +++ b/include/gfx.h @@ -0,0 +1,117 @@ +#ifndef HBAT_GFX_H +#define HBAT_GFX_H + +#include "platform.h" +#include "linalg.h" + +enum gfx_vertex_format { + GFX_VERTEX_NONE = 0, + GFX_VERTEX_FLOAT = 1, + + GFX_VERTEX_V1 = GFX_VERTEX_FLOAT, + GFX_VERTEX_V2, + GFX_VERTEX_V3, + GFX_VERTEX_V4, + + GFX_VERTEX_M1X1 = GFX_VERTEX_FLOAT, + GFX_VERTEX_M2X2, + GFX_VERTEX_M3X3, + GFX_VERTEX_M4X4, +}; + +enum gfx_sampler_format { + GFX_SAMPLER_NONE = 0, + GFX_SAMPLER_LINEAR = 1, +}; + +struct gfx_ctx_i; +union gfx_ctx { struct gfx_ctx_i *_i; }; + +// In order to conserve resources, descriptions refer to +// URLs rather than actual contents so that they can be +// unloaded and reloaded again on mobile without taking +// up a lot of space +struct gfx_shader_desc { + struct stringc vertex_url; + struct stringc fragment_url; +}; +struct gfx_shader_i; +union gfx_shader { uint32_t _i; }; + +struct gfx_pipeline_desc { + union gfx_shader shader; +}; +struct gfx_pipeline_i; +union gfx_pipeline { uint32_t _i; }; + +struct gfx_texture_desc { + struct stringc url; +}; +union gfx_texture { uint32_t _i; }; + +struct gfx_sampler_desc { +}; +struct gfx_sampler_i; +union gfx_sampler { uint32_t _i; }; + +// Font rendering has an unusual way of managing textures +// that cannot be easily or efficiently replicated via gfx_texture, +// so gfx_font is born to bridge that gap. +struct gfx_font_desc { + struct stringc url; +}; +// This models a similar structure to BMFont's char +struct gfx_font_bmglyph { + uint16_t x,y,width,height; + int16_t xoffset,yoffset,xadvance; +}; +struct gfx_font_info { + float fontsize; + float ascent; + float descent; + float linegap; +}; +struct gfx_font_i; +union gfx_font { uint32_t _i; }; + +union gfx_ctx gfx_startup(union platform_window window, union gfx_ctx prev); +union gfx_ctx gfx_shutdown(union platform_window window); + +union gfx_shader gfx_shader_construct(const struct gfx_shader_desc *desc); +void gfx_shader_destruct(union gfx_shader shader); + +union gfx_pipeline gfx_pipeline_construct(const struct gfx_pipeline_desc *desc); +void gfx_pipeline_destruct(union gfx_pipeline pipeline); +void gfx_pipeline_bind(union gfx_pipeline pipeline); + +union gfx_texture gfx_texture_construct(const struct gfx_texture_desc *desc); +void gfx_texture_destruct(union gfx_texture texture); +int gfx_texture_width(union gfx_texture texture); +int gfx_texture_height(union gfx_texture texture); +struct rectanglei32 gfx_texture_rect(union gfx_texture texture); + +// TODO: I'm not a fan of the fact that there are samplers in the first place. +// An API that just takes global sampler desc and sets it as opposed to +// creating a sampler to store somewhere is an improvement IMO. +// +// Something like gfx_sampler_set(const struct gfx_sampler_desc *desc); +union gfx_sampler gfx_sampler_construct(const struct gfx_sampler_desc *desc); +void gfx_sampler_destruct(union gfx_sampler sampler); + +union gfx_font gfx_font_construct(const struct gfx_font_desc *desc); +void gfx_font_destruct(union gfx_font font); +union gfx_texture gfx_font_texture(union gfx_font font); // TODO: Optionally suggest a texture size to make more efficient use of space. +struct gfx_font_bmglyph gfx_font_bmglyph_get(union gfx_font font, int rendered_codepoint, int kerning_codepoint); +// TODO: Getting multiple glyphs at once from a string is a more usual and practical case... Put that case in. +struct gfx_font_info gfx_font_info_get(union gfx_font font); +void gfx_font_size_set(union gfx_font font, float fontsize); + +void gfx_vertices_draw(union gfx_texture texture, union gfx_sampler sampler, const void *vertices, size_t bytelen, size_t vertices_count); + +// This is mostly a debug + prototyping definitions for platform gfx functions +// You shouldn't need any of these in a release build +union gfx_ctx _gfx_ctx_global(void); +union platform_window _gfx_ctx_window(union gfx_ctx ctx); +struct allocator *_gfx_allocator(void); + +#endif diff --git a/include/input.h b/include/input.h new file mode 100644 index 0000000..b58e2ce --- /dev/null +++ b/include/input.h @@ -0,0 +1,141 @@ +#ifndef HBAT_INPUT_H +#define HBAT_INPUT_H + +#include "types.h" +#include "platform.h" + +enum gaxis_type { + GAXIS_NONE = PLATFORM_GAXIS_NONE, + GAXIS_LEFTX = PLATFORM_GAXIS_LEFTX, + GAXIS_LEFTY = PLATFORM_GAXIS_LEFTY, + GAXIS_RIGHTX = PLATFORM_GAXIS_RIGHTX, + GAXIS_RIGHTY = PLATFORM_GAXIS_RIGHTY, + GAXIS_LEFT_TRIGGER = PLATFORM_GAXIS_LEFT_TRIGGER, + GAXIS_RIGHT_TRIGGER = PLATFORM_GAXIS_RIGHT_TRIGGER, +}; +enum gbutton_type { + GBUTTON_NONE = PLATFORM_GBUTTON_NONE, + GBUTTON_SOUTH = PLATFORM_GBUTTON_SOUTH, + GBUTTON_EAST = PLATFORM_GBUTTON_EAST, + GBUTTON_WEST = PLATFORM_GBUTTON_WEST, + GBUTTON_NORTH = PLATFORM_GBUTTON_NORTH, + GBUTTON_BACK = PLATFORM_GBUTTON_BACK, + GBUTTON_GUIDE = PLATFORM_GBUTTON_GUIDE, + GBUTTON_START = PLATFORM_GBUTTON_START, + GBUTTON_LEFT_STICK = PLATFORM_GBUTTON_LEFT_STICK, + GBUTTON_RIGHT_STICK = PLATFORM_GBUTTON_RIGHT_STICK, + GBUTTON_LEFT_SHOULDER = PLATFORM_GBUTTON_LEFT_SHOULDER, + GBUTTON_RIGHT_SHOULDER = PLATFORM_GBUTTON_RIGHT_SHOULDER, + GBUTTON_DPAD_UP = PLATFORM_GBUTTON_DPAD_UP, + GBUTTON_DPAD_DOWN = PLATFORM_GBUTTON_DPAD_DOWN, + GBUTTON_DPAD_LEFT = PLATFORM_GBUTTON_DPAD_LEFT, + GBUTTON_DPAD_RIGHT = PLATFORM_GBUTTON_DPAD_RIGHT, + GBUTTON_MISC1 = PLATFORM_GBUTTON_MISC1, + GBUTTON_RIGHT_PADDLE1 = PLATFORM_GBUTTON_RIGHT_PADDLE1, + GBUTTON_LEFT_PADDLE1 = PLATFORM_GBUTTON_LEFT_PADDLE1, + GBUTTON_RIGHT_PADDLE2 = PLATFORM_GBUTTON_RIGHT_PADDLE2, + GBUTTON_LEFT_PADDLE2 = PLATFORM_GBUTTON_LEFT_PADDLE2, + GBUTTON_TOUCHPAD = PLATFORM_GBUTTON_TOUCHPAD, + GBUTTON_MISC2 = PLATFORM_GBUTTON_MISC2, + GBUTTON_MISC3 = PLATFORM_GBUTTON_MISC3, + GBUTTON_MISC4 = PLATFORM_GBUTTON_MISC4, + GBUTTON_MISC5 = PLATFORM_GBUTTON_MISC5, + GBUTTON_MISC6 = PLATFORM_GBUTTON_MISC6, +}; +enum key_sym { + KEY_NONE = PLATFORM_KEY_NONE, + KEY_A = PLATFORM_KEY_A, + KEY_B = PLATFORM_KEY_B, + KEY_C = PLATFORM_KEY_C, + KEY_D = PLATFORM_KEY_D, + KEY_E = PLATFORM_KEY_E, + KEY_F = PLATFORM_KEY_F, + KEY_G = PLATFORM_KEY_G, + KEY_H = PLATFORM_KEY_H, + KEY_I = PLATFORM_KEY_I, + KEY_J = PLATFORM_KEY_J, + KEY_K = PLATFORM_KEY_K, + KEY_L = PLATFORM_KEY_L, + KEY_M = PLATFORM_KEY_M, + KEY_N = PLATFORM_KEY_N, + KEY_O = PLATFORM_KEY_O, + KEY_P = PLATFORM_KEY_P, + KEY_Q = PLATFORM_KEY_Q, + KEY_R = PLATFORM_KEY_R, + KEY_S = PLATFORM_KEY_S, + KEY_T = PLATFORM_KEY_T, + KEY_U = PLATFORM_KEY_U, + KEY_V = PLATFORM_KEY_V, + KEY_W = PLATFORM_KEY_W, + KEY_X = PLATFORM_KEY_X, + KEY_Y = PLATFORM_KEY_Y, + KEY_Z = PLATFORM_KEY_Z, +}; + +// Fingers which aren't actually fingers but are useful to model as such, +// like mouse motions and buttons. +enum pseudofinger { + // You can use this constant to indicate the lack of a finger pressed + // You must not change this value so that zero-initialized structs + // or falsy checks automatically use this value + PSEUDOFINGER_NONE = 0, + + // Mouse motion is a very special finger + // It will never emit a down or up event, but it will emit move events, so + // it's usually ignored in applications that only care about pressed fingers + PSEUDOFINGER_MOUSE_MOTION = 0x7fffff00, + + // Mouse buttons + PSEUDOFINGER_MOUSE_BUTTON_BASE = PSEUDOFINGER_MOUSE_MOTION, + PSEUDOFINGER_MOUSE_BUTTON_LEFT = PSEUDOFINGER_MOUSE_BUTTON_BASE + 1, + PSEUDOFINGER_MOUSE_BUTTON_MIDDLE = PSEUDOFINGER_MOUSE_BUTTON_BASE + 2, + PSEUDOFINGER_MOUSE_BUTTON_RIGHT = PSEUDOFINGER_MOUSE_BUTTON_BASE + 3, + PSEUDOFINGER_MOUSE_BUTTON_X1 = PSEUDOFINGER_MOUSE_BUTTON_BASE + 4, + PSEUDOFINGER_MOUSE_BUTTON_X2 = PSEUDOFINGER_MOUSE_BUTTON_BASE + 5, + // X3, X4, ..., X251 + PSEUDOFINGER_MOUSE_BUTTON_X252 = PSEUDOFINGER_MOUSE_BUTTON_BASE + 255, +}; + +struct input_motion { + // Normalized with respect to the center, but you can have values outside of [-1, 1]. + // This means that the pointer is moving outside of the current window. + struct vector2f32 normalized; + + // TODO: + // normdelta is unimplemented because I didn't need it for recent games + // but it's important for most games to have this. How you should do this + // is to not add a delta to platform events, but instead track inputs that use + // this, like fingers, and then provide the deltas that way. + // + // TODO: + // When you track inputs, whenever the window goes out of focus or in focus + // or during gfx startup or shutdown, you should clear out any input state + // and emit up events for any keys, fingers, etc., that are down. + struct vector2f32 normdelta; + + // Correlates to normalized and normdelta but describes inches from top-left instead + // Useful in calculations where pointer/mouse sensitivity matters + // + // TODO: Also implement this by getting the DPI of the screen + struct vector2f32 inches; + struct vector2f32 inchdelta; +}; + +struct input_callbacks { + void (*on_gstate)(uint32_t gamepad, bool connected); + void (*on_gbutton)(uint32_t gamepad, uint32_t button, bool down); + void (*on_gaxis)(uint32_t gamepad, uint32_t axis, float normalized); + void (*on_key)(uint32_t keyboard, uint32_t key, bool down); + + // on_fmotion is a very overloaded function + // If fingers added is +1, the finger is down, and if it is -1, it is up (if 0, no up nor down event) + // In addition, there are pseudofingers like mouse buttons and motion. + // Note that psi is for pressure, but that no device really supports that kind of sensitivity yet, + // So instead treat it as some device-independent pressure unit. + void (*on_fmotion)(uint32_t device, uint32_t finger, const struct input_motion *motion, float psi, int fingers_added); +}; +void input_set_callbacks(const struct input_callbacks *cbs); + +void input_inject_platform_event(const union platform_event *ev, union platform_window window); + +#endif diff --git a/include/linalg.h b/include/linalg.h new file mode 100644 index 0000000..ad55850 --- /dev/null +++ b/include/linalg.h @@ -0,0 +1,249 @@ +#ifndef HBAT_LINALG_H +#define HBAT_LINALG_H +#include "compiler.h" +#include "types.h" +#include +#include + +#define V2F32FMT "(%f,%f)" +#define V3F32FMT "(%f,%f,%f)" +#define V4F32FMT "(%f,%f,%f,%f)" +struct vector2f32 { + float x, y; +}; +struct vector3f32 { + float x, y, z; +}; +struct vector4f32 { + float x, y, z, w; +}; + +#define V2I32FMT "(%d,%d)" +#define V3I32FMT "(%d,%d,%d)" +#define V4I32FMT "(%d,%d,%d,%d)" +struct vector2i32 { + int32_t x, y; +}; +struct vector3i32 { + int32_t x, y, z; +}; +struct vector4i32 { + int32_t x, y, z, w; +}; + +#define V2ARGS(v2) (v2).x, (v2).y +#define V3ARGS(v3) (v3).x, (v3).y, (v3).z +#define V4ARGS(v4) (v4).x, (v4).y, (v4).z, (v4).w + +// NOTE: +// Rectangle's x,y can either represent the center or the top-left corner. +// In most cases, rectanglef32 uses center and rectanglei32 uses top-left +// because they usually correspond to world and texcoord space +// respectively where such coordinates are more useful in context. +#define RECTF32FMT "[%f,%f,%f,%f]" +struct rectanglef32 { + float x, y, width, height; +}; +#define RECTI32FMT "[%d,%d,%d,%d]" +struct rectanglei32 { + int32_t x, y, width, height; +}; +#define RECTARGS(r) (r).x, (r).y, (r).width, (r).height + +// Column-major order matricies +#define M2F32FMT "(%f,%f)(%f,%f)" +struct matrix2f32 { + struct vector2f32 x, y; +}; +struct matrix3f32 { + struct vector3f32 x, y, z; +}; +struct matrix4f32 { + struct vector4f32 x, y, z, w; +}; +#define M2F32ARGS(m) (m).x.x, (m).y.x, (m).x.y, (m).y.y + +// Vector operations, generic ones provided as macros, specialized ones as inline functions + +// Vector arithmetic +#define v2add(v1, v2) {(v1).x + (v2).x, (v1).y + (v2).y} +#define v3add(v1, v2) {(v1).x + (v2).x, (v1).y + (v2).y, (v1).z + (v2).z} +#define v4add(v1, v2) {(v1).x + (v2).x, (v1).y + (v2).y, (v1).z + (v2).z, (v1).w + (v2).w} +static inline struct vector2f32 v2f32add(struct vector2f32 v1, struct vector2f32 v2) +{ + return LICAST(struct, vector2f32)v2add(v1, v2); +} + +#define v2addself(v1, v2) (v1)->x += (v2).x; (v1)->y += (v2).y +static inline void v2f32addself(struct vector2f32 *v1, struct vector2f32 v2) +{ + v2addself(v1, v2); +} + +#define v2sub(v1, v2) {(v1).x - (v2).x, (v1).y - (v2).y} +#define v3sub(v1, v2) {(v1).x - (v2).x, (v1).y - (v2).y, (v1).z - (v2).z} +#define v4sub(v1, v2) {(v1).x - (v2).x, (v1).y - (v2).y, (v1).z - (v2).z, (v1).w - (v2).w} +static inline struct vector2f32 v2f32sub(struct vector2f32 v1, struct vector2f32 v2) +{ + return LICAST(struct, vector2f32)v2sub(v1, v2); +} + +#define v2subself(v1, v2) (v1)->x += (v2).x; (v1)->y += (v2).y +static inline void v2f32subself(struct vector2f32 *v1, struct vector2f32 v2) +{ + v2subself(v1, v2); +} + +#define v2mul(v1, v2) {(v1).x * (v2).x, (v1).y * (v2).y} +static inline struct vector2f32 v2f32mul(struct vector2f32 v1, struct vector2f32 v2) +{ + return LICAST(struct, vector2f32)v2mul(v1, v2); +} +#define v2mulself(v1, v2) (v1)->x *= (v2).x; (v1)->y *= (v2).y +static inline void v2f32mulself(struct vector2f32 *v1, struct vector2f32 v2) +{ + v2mulself(v1, v2); +} + +#define v2div(v1, v2) {(v1).x / (v1).x, (v1).y / (v2).y} +static inline struct vector2f32 v2f32div(struct vector2f32 v1, struct vector2f32 v2) +{ + return LICAST(struct, vector2f32)v2div(v1, v2); +} +#define v2divself(v1, v2) (v1)->x /= (v2).x; (v1)->y /= (v2).y +static inline void v2f32divself(struct vector2f32 *v1, struct vector2f32 v2) +{ + v2divself(v1, v2); +} + + +#define v2clamp(v, l, h) {\ + (v).x < (l).x ? (l).x : ((v).x > (h).x ? (h).x : (v).x),\ + (v).y < (l).y ? (l).y : ((v).y > (h).y ? (h).y : (v).y),\ +} +static inline struct vector2f32 v2f32clamp(struct vector2f32 v, struct vector2f32 l, struct vector2f32 h) +{ + return LICAST(struct, vector2f32)v2clamp(v, l, h); +} +#define v2clampself(v, l, h) \ + (v)->x = (v)->x < (l).x ? (l).x : ((v)->x > (h).x ? (h).x : (v)->x);\ + (v)->y = (v)->y < (l).y ? (l).y : ((v)->y > (h).y ? (h).y : (v)->y) +static inline void v2f32clampself(struct vector2f32 *v, struct vector2f32 l, struct vector2f32 h) +{ + v2clampself(v, l, h); +} + +#define v2mag2(v) ((v).x * (v).x + (v).y * (v).y) +#define v2mag(v) sqrtf(v2mag2(v)) +#define v3mag2(v) ((v).x * (v).x + (v).y * (v).y + (v).z * (v).z) +#define v3mag(v) sqrtf(v3mag2(v)) +#define v4mag2(v) ((v).x * (v).x + (v).y * (v).y + (v).z * (v).z + (v).w * (v).w) +#define v4mag(v) sqrtf(v4mag2(v)) +static inline float v2f32mag2(struct vector2f32 v) +{ + return (float)v2mag2(v); +} +static inline float v2f32mag(struct vector2f32 v) +{ + return (float)v2mag(v); +} + +#define v2atan(v) atan2f((v).y, (v).x) +static inline float v2f32atan(struct vector2f32 v) +{ + return v2atan(v); +} + +// Vector with scalar ops +#define v2s(s) {(s), (s)} +static inline struct vector2f32 v2f32s(float s) +{ + return LICAST(struct, vector2f32)v2s(s); +} + +#define v2adds(v, s) {(v).x + (s), (v).y + (s)} + +#define v2subs(v, s) {(v).x - (s), (v).y - (s)} +static inline struct vector2f32 v2f32subs(struct vector2f32 v, float s) +{ + return LICAST(struct, vector2f32)v2subs(v, s); +} + +#define v2muls(v, s) {(v).x * (s), (v).y * (s)} +static inline struct vector2f32 v2f32muls(struct vector2f32 v, float s) +{ + return LICAST(struct, vector2f32)v2muls(v, s); +} +#define v2mulselfs(v, s) (v)->x *= s; (v)->y *= s +static inline void v2f32mulselfs(struct vector2f32 *v, float s) +{ + v2mulselfs(v, s); +} + +#define v2divs(v, s) {(v).x / (s), (v).y / (s)} +static inline struct vector2f32 v2f32divs(struct vector2f32 v, float s) +{ + return LICAST(struct, vector2f32)v2divs(v, s); +} + +static inline struct vector2f32 v2f32rand1(void) +{ + return LICAST(struct, vector2f32){ + (float)rand() / (float)RAND_MAX, + (float)rand() / (float)RAND_MAX, + }; +} + +static inline struct vector2f32 v2f32rand1signed(void) +{ + return LICAST(struct, vector2f32){ + (float)(rand() - RAND_MAX / 2) / (float)(RAND_MAX / 2), + (float)(rand() - RAND_MAX / 2) / (float)(RAND_MAX / 2), + }; +} + +#define v2sincos(a) {cos(a), sin(a)} +static inline struct vector2f32 v2f32sincos(float a) +{ + return LICAST(struct, vector2f32)v2sincos(a); +} + +// Matrix operations +#define m2mulv2(m, v) {(m).x.x * (v).x + (m).y.x * (v).y, (m).x.y * (v).x + (m).y.y * (v).y} +static inline struct vector2f32 m2f32mulv2f32(struct matrix2f32 m, struct vector2f32 v) +{ + return LICAST(struct, vector2f32)m2mulv2(m, v); +} +static inline void m2f32mulv2f32self(struct matrix2f32 m, struct vector2f32 *v) +{ + *v = m2f32mulv2f32(m, *v); +} + +static inline struct matrix2f32 m2f32rot(float a) +{ + float c = cos(a), s = sin(a); + return LICAST(struct, matrix2f32){ + .x = { c, s}, + .y = {-s, c}, + }; +} + +// Rectangle operations +#define rectmoveself(r, v) (r)->x += (v).x; (r)->y += (v).y +static inline void rectf32moveself(struct rectanglef32 *r, struct vector2f32 v) +{ + rectmoveself(r, v); +} + +#define rectpos(r) {(r).x, (r).y} +#define rectdims(r) {(r).width, (r).height} +static inline struct vector2f32 rectf32pos(struct rectanglef32 r) +{ + return LICAST(struct, vector2f32)rectpos(r); +} +static inline struct vector2f32 rectf32dims(struct rectanglef32 r) +{ + return LICAST(struct, vector2f32)rectdims(r); +} + +#endif diff --git a/include/os.h b/include/os.h new file mode 100644 index 0000000..9d5d2fb --- /dev/null +++ b/include/os.h @@ -0,0 +1,201 @@ +#ifndef HBAT_OS_H +#define HBAT_OS_H + +#include +#include +#include +#include +#include +#include "compiler.h" +#include "types.h" +#include "platform.h" + +// Synchronization primitives +union os_once { union platform_os_once _i; }; +#define OS_ONCE_INIT LICAST(union, os_once){PLATFORM_OS_ONCE_INIT} +static inline void os_once_run(union os_once *o, void (*run_fn)(void)) +{ + platform_os_once_run(&o->_i, run_fn); +} + +// Nice OS goodies +struct os_timespec { + int64_t sec; + int32_t nsec; + int32_t err; +}; +enum os_time_type { + // Must correspond to PLATFORM definitions + // for easy mapping to platform functions + OS_TIME_WALL = PLATFORM_OS_TIME_WALL, + OS_TIME_MONO = PLATFORM_OS_TIME_MONO, + OS_TIME_TICK = PLATFORM_OS_TIME_TICK__UNUSED, +}; +size_t os_getpagesize(void); +struct os_timespec os_time(int type); +float os_timef(int type); +double os_timed(int type); +float os_time2f(struct os_timespec ts); +double os_time2d(struct os_timespec ts); +int64_t os_time_cmp_ns(struct os_timespec lhs, struct os_timespec rhs); +int64_t os_time_cmp(struct os_timespec lhs, struct os_timespec rhs); + +// Because memory is allocated by the OS on a page-level basis, +// so long as the pointer is not page-aligned, we can read one +// byte past the end of the buffer to see if it has a null terminator. +// +// For NULLEARLY, the string contains a null character itself +// For NULLTERMED, the string is definitely null terminated +// For NULLMAYBE, it either surely isn't or address is page aligned +enum os_string_nullterm_result +{ + OS_STRING_NULLEARLY = -1, + OS_STRING_NULLTERMED = 0, + OS_STRING_NULLMAYBE = 1, +}; +int os_string_nullterm(struct stringc s); + +// Use os_string_nulltermed above to determine if allocation is needed +// If a null pointer returned, allocation failed or NULLEARLY returned +// If s.data is returned, then it's already null terminated and no allocation was made +// If any other pointer is returned, it was allocated with os_malloc_emulate_a +char *os_string_nullterm_make(struct stringc s, struct allocator *a); + +// Assets can often be stored and read differently depending on +// whether it's binary or text. Text assets are usually compressed +// and binary assets are not compressed. +enum asset_type { + ASSET_BINARY = 0, + ASSET_TEXT = 1, +}; +enum os_file_seek { + // Must correspond to PLATFORM definitions + // for easy mapping to platform functions + OS_SEEK_SET = PLATFORM_OS_SEEK_SET, + OS_SEEK_CUR = PLATFORM_OS_SEEK_CUR, + OS_SEEK_END = PLATFORM_OS_SEEK_END, +}; +struct os_stat { + int64_t size; +}; +union os_file { union platform_os_file _i; }; +void os_file_close(union os_file fp); +ssize_t os_file_read(union os_file fp, void *buf, size_t size); +ssize_t os_file_write(union os_file fp, const void *buf, size_t size); +int64_t os_file_seek(union os_file fp, int64_t delta, int from); +int64_t os_file_size(union os_file fp); +// TODO: int os_file_stat(union os_file fp, struct os_stat stat); +static inline int64_t os_file_tell(union os_file fp) +{ + return os_file_seek(fp, 0, OS_SEEK_CUR); +} + +// File APIs that require a null-terminated C string +union os_file os_asset_openc(int type, const char *filename); + +// Generic URL open API so that resources can be read from anywhere. +// This shouldn't support network accesses, though, +// just different ways of addressing the local filesystem. +union os_file os_url_openc(const char *url, struct allocator *a); + +// File + URL APIs that take a string + length instead +// An allocation might be needed to add a null terminator +union os_file os_asset_open(int type, struct stringc filename, struct allocator *a); +union os_file os_url_open(struct stringc filename, struct allocator *a); + +// Will read all the file and can add extra space (like for a null terminator) +struct blob os_file_read_all_extra(union os_file fp, struct allocator *a, size_t extra); +static inline struct blob os_file_read_all(union os_file fp, struct allocator *a) +{ + return os_file_read_all_extra(fp, a, 0); +} + +// Image loading functions +struct image os_image_decode(const void *mem, size_t size, struct allocator *a); + +// TODO: File IO on the filesystem +// TODO: URL IO so that applications can load resources generically (see gfx.h for an example) + +enum log_verbosity { + LOG_NONE = 0, + LOG_FATAL = 1, + LOG_ERROR, + LOG_WARNING, + LOG_INFO, + LOG_VERBOSE, + LOG_DEBUG, + LOG_TRACE, +}; + +// Can define a tag on the command line for os_log* to use +// if you define it as a macro beforehand. Otherwise, it'll +// just use this default that you set here. +#ifndef OSLOG_TAG_DEFAULT +#define OSLOG_TAG_DEFAULT "oslog" +#endif + +// If you want, you can use a custom tag with os_log_tag* +// but I don't recommend this because it can make it harder +// and unintuitive for people to find and filter your logs in, +// say, syslogs or logcat. You should stick with os_log(v). +void os_log_tag(int verbosity, const char *tag, const char *fmt, ...); +void os_log_tag_v(int verbosity, const char *tag, const char *fmt, va_list ap); +#define os_log(V, FMT, ...) os_log_tag(V, OSLOG_TAG_DEFAULT, FMT VAOPT(__VA_ARGS__)) +#define os_log_v(V, FMT, AP) os_log_tagv(V, OSLOG_TAG_DEFAULT, FMT, AP) + +// A null allocator can be provided, in which case the global allocator is used +#define OS_ZERO_SIZE_PTR ((void *)(uintptr_t)-64) +struct blob os_alloc(struct allocator *a, size_t size, size_t alignment); +struct blob os_realloc(struct allocator *a, struct blob prev, size_t size, size_t alignment); +struct blob os_frealloc(struct allocator *a, struct blob prev, size_t size, size_t alignment); +void os_free(struct allocator *a, struct blob mem); +void os_free_all(struct allocator *a); + +// Most C libraries expect a malloc-like API that has a global allocator +// and returns pointers rather than blobs so emulate with a global allocator +struct blob os_blob_from_emulate(void *p); +void *os_realloc_emulate_a(void *p, size_t sz, struct allocator *a); +void *os_malloc_emulate_a(size_t sz, struct allocator *a); +void os_free_emulate_a(void *p, struct allocator *a); +void *os_realloc_emulate(void *p, size_t sz); +void *os_malloc_emulate(size_t sz); +void os_free_emulate(void *p); + +int os_strcmp_prefix(const char *s, const char *prefix); +struct blob os_cstrdup(const char *cstr, struct allocator *a); +struct blob os_strdup(struct stringc s, struct allocator *a); + +#define OSLOG_TAG(V, TAG, FMT, ...) os_log_tag(V, TAG, FMT VAOPT(__VA_ARGS__)) +#define OSLOG(V, FMT, ...) OSLOG_TAG(V, OSLOG_TAG_DEFAULT, FMT VAOPT(__VA_ARGS__)) + +#define OSLOGF(FMT, ...) OSLOG(LOG_FATAL, FMT VAOPT(__VA_ARGS__)) +#define OSLOGE(FMT, ...) OSLOG(LOG_ERROR, FMT VAOPT(__VA_ARGS__)) +#define OSLOGW(FMT, ...) OSLOG(LOG_WARNING, FMT VAOPT(__VA_ARGS__)) +#define OSLOGI(FMT, ...) OSLOG(LOG_INFO, FMT VAOPT(__VA_ARGS__)) +#define OSLOGV(FMT, ...) OSLOG(LOG_VERBOSE, FMT VAOPT(__VA_ARGS__)) +#define OSLOGD(FMT, ...) OSLOG(LOG_DEBUG, FMT VAOPT(__VA_ARGS__)) +#define OSLOGT(FMT, ...) OSLOG(LOG_TRACE, FMT VAOPT(__VA_ARGS__)) + +// You can use this style of naming as well +#define os_logf(FMT, ...) os_log(LOG_FATAL, FMT VAOPT(__VA_ARGS__)) +#define os_loge(FMT, ...) os_log(LOG_ERROR, FMT VAOPT(__VA_ARGS__)) +#define os_logw(FMT, ...) os_log(LOG_WARNING, FMT VAOPT(__VA_ARGS__)) +#define os_logi(FMT, ...) os_log(LOG_INFO, FMT VAOPT(__VA_ARGS__)) +#define os_logv(FMT, ...) os_log(LOG_VERBOSE, FMT VAOPT(__VA_ARGS__)) +#define os_logd(FMT, ...) os_log(LOG_DEBUG, FMT VAOPT(__VA_ARGS__)) +#define os_logt(FMT, ...) os_log(LOG_TRACE, FMT VAOPT(__VA_ARGS__)) + +// Standard allocator, also used for malloc emulation +extern struct allocator os_allocator_default; + +// Standard temporary allocator +extern struct allocator os_allocator_default_temp; + +// Handle table functions +uint32_t handle_table_alloc(struct handle_table *t); +uint32_t handle_table_check(struct handle_table *t, uint32_t handle); +int handle_table_free(struct handle_table *t, uint32_t handle); +uint32_t handle_table_realloc(struct handle_table *t, uint32_t handle_prev); +void handle_table_clear(struct handle_table *t); + +#endif diff --git a/include/platform.h b/include/platform.h new file mode 100644 index 0000000..6dcb75f --- /dev/null +++ b/include/platform.h @@ -0,0 +1,328 @@ +#ifndef HBAT_PLATFORM_H +#define HBAT_PLATFORM_H +#include +#include "types.h" +#include "linalg.h" + +// Synchronization functions +// +// TODO: +// We have to include system headers here because OSes +// have specific macros for primitive initialization. +// If porting to other OSes, use ifdefs to switch this. +#ifdef _WIN32 +// TODO: Do we want to put windows.h into a pch? +// it is a very, very big header file that +// takes a while to include +#define WIN32_LEAN_AND_MEAN +#include +#include +union platform_os_once { INIT_ONCE _i; }; +#define PLATFORM_OS_ONCE_INIT LICAST(union, platform_os_once){(INIT_ONCE)INIT_ONCE_STATIC_INIT} +void platform_os_once_run(union platform_os_once *once, void (*run_fn)(void)); +#else +#include +#include +union platform_os_once { pthread_once_t _i; }; +#define PLATFORM_OS_ONCE_INIT LICAST(union, platform_os_once){(pthread_once_t)PTHREAD_ONCE_INIT} +void platform_os_once_run(union platform_os_once *once, void (*run_fn)(void)); +#endif + +// Other very general OS functions +struct platform_os_timespec { + int64_t sec; + int32_t nsec; + int32_t err; +}; +enum platform_os_time_type { + PLATFORM_OS_TIME_WALL, + PLATFORM_OS_TIME_MONO, + + // Defined for the sake of mapping easily to OS_TIME_* + // and guaranteeing a unique identifier, but should + // remain unused in platform code. + PLATFORM_OS_TIME_TICK__UNUSED, +}; +size_t platform_os_getpagesize(void); +struct platform_os_timespec platform_os_time(int type); + +// Window functions +struct platform_window_i; +union platform_window { struct platform_window_i *_i; }; +int platform_window_width(union platform_window window); +int platform_window_height(union platform_window window); + +// Shamelessly ripped from SDL events so that it's easier +// to port between this event and SDL's events and so less +// information is lost in the transition +enum platform_event_type { + PLATFORM_EVENT_NONE = 0, + PLATFORM_EVENT_FINGER_DOWN, + PLATFORM_EVENT_FINGER_MOTION, + PLATFORM_EVENT_FINGER_UP, + PLATFORM_EVENT_KEY_UP, + PLATFORM_EVENT_KEY_DOWN, + PLATFORM_EVENT_MOUSE_BUTTON_DOWN, + PLATFORM_EVENT_MOUSE_BUTTON_UP, + PLATFORM_EVENT_MOUSE_MOTION, + PLATFORM_EVENT_GAMEPAD_CONNECTED, + PLATFORM_EVENT_GAMEPAD_DISCONNECTED, + PLATFORM_EVENT_GAMEPAD_AXIS_MOTION, + PLATFORM_EVENT_GAMEPAD_BUTTON_DOWN, + PLATFORM_EVENT_GAMEPAD_BUTTON_UP, +}; +enum platform_gaxis_type { + PLATFORM_GAXIS_NONE = 0, + PLATFORM_GAXIS_LEFTX, + PLATFORM_GAXIS_LEFTY, + PLATFORM_GAXIS_RIGHTX, + PLATFORM_GAXIS_RIGHTY, + PLATFORM_GAXIS_LEFT_TRIGGER, + PLATFORM_GAXIS_RIGHT_TRIGGER, +}; +enum platform_gbutton_type { + PLATFORM_GBUTTON_NONE = 0, + PLATFORM_GBUTTON_SOUTH, + PLATFORM_GBUTTON_EAST, + PLATFORM_GBUTTON_WEST, + PLATFORM_GBUTTON_NORTH, + PLATFORM_GBUTTON_BACK, + PLATFORM_GBUTTON_GUIDE, + PLATFORM_GBUTTON_START, + PLATFORM_GBUTTON_LEFT_STICK, + PLATFORM_GBUTTON_RIGHT_STICK, + PLATFORM_GBUTTON_LEFT_SHOULDER, + PLATFORM_GBUTTON_RIGHT_SHOULDER, + PLATFORM_GBUTTON_DPAD_UP, + PLATFORM_GBUTTON_DPAD_DOWN, + PLATFORM_GBUTTON_DPAD_LEFT, + PLATFORM_GBUTTON_DPAD_RIGHT, + PLATFORM_GBUTTON_MISC1, + PLATFORM_GBUTTON_RIGHT_PADDLE1, + PLATFORM_GBUTTON_LEFT_PADDLE1, + PLATFORM_GBUTTON_RIGHT_PADDLE2, + PLATFORM_GBUTTON_LEFT_PADDLE2, + PLATFORM_GBUTTON_TOUCHPAD, + PLATFORM_GBUTTON_MISC2, + PLATFORM_GBUTTON_MISC3, + PLATFORM_GBUTTON_MISC4, + PLATFORM_GBUTTON_MISC5, + PLATFORM_GBUTTON_MISC6, +}; +enum platform_key_sym { + PLATFORM_KEY_NONE = 0, + PLATFORM_KEY_A = 'a', + PLATFORM_KEY_B = 'b', + PLATFORM_KEY_C = 'c', + PLATFORM_KEY_D = 'd', + PLATFORM_KEY_E = 'e', + PLATFORM_KEY_F = 'f', + PLATFORM_KEY_G = 'g', + PLATFORM_KEY_H = 'h', + PLATFORM_KEY_I = 'i', + PLATFORM_KEY_J = 'j', + PLATFORM_KEY_K = 'k', + PLATFORM_KEY_L = 'l', + PLATFORM_KEY_M = 'm', + PLATFORM_KEY_N = 'n', + PLATFORM_KEY_O = 'o', + PLATFORM_KEY_P = 'p', + PLATFORM_KEY_Q = 'q', + PLATFORM_KEY_R = 'r', + PLATFORM_KEY_S = 's', + PLATFORM_KEY_T = 't', + PLATFORM_KEY_U = 'u', + PLATFORM_KEY_V = 'v', + PLATFORM_KEY_W = 'w', + PLATFORM_KEY_X = 'x', + PLATFORM_KEY_Y = 'y', + PLATFORM_KEY_Z = 'z', +}; +enum platform_mouse_button_code { + PLATFORM_MOUSE_BUTTON_NONE = 0, + PLATFORM_MOUSE_BUTTON_LEFT = 1, + PLATFORM_MOUSE_BUTTON_MIDDLE = 2, + PLATFORM_MOUSE_BUTTON_RIGHT = 3, +}; + +// A lot of these events model the events in SDL3 +// but are generally leaner since we rely on ourselves +// to generate things like timestamps and deltas +struct platform_event_finger { + int type; + unsigned int device; + unsigned int id; + struct vector2f32 pixels; + float pressure; +}; +struct platform_event_key { + int type; + unsigned int which; + int sym; + bool repeat; + uint8_t reserved1; + uint16_t reserved2; +}; +struct platform_event_mouse_motion { + int type; + unsigned int which; + struct vector2f32 pixels; +}; +struct platform_event_mouse_button { + int type; + unsigned int which; + struct vector2f32 pixels; + uint8_t index; + uint8_t clicks; + uint16_t reserved1; +}; +struct platform_event_gamepad_device { + int type; + unsigned int which; +}; +struct platform_event_gamepad_axis { + int type; + unsigned int which; + uint8_t axis; + uint8_t reserved1; + int16_t value; +}; +struct platform_event_gamepad_button { + int type; + unsigned int which; + uint8_t button; + uint8_t reserved1; + uint16_t reserved2; +}; +union platform_event { + int type; + struct platform_event_finger finger; + struct platform_event_key key; + struct platform_event_mouse_button button; + struct platform_event_mouse_motion motion; + struct platform_event_gamepad_device gdevice; + struct platform_event_gamepad_axis gaxis; + struct platform_event_gamepad_button gbutton; +}; + +// GFX resource types +// Sokol uses a lot of integer IDs so a lot of these unions +// will also have an integer id to more easily work with that +struct platform_gfx_pipeline_i; +union platform_gfx_pipeline { struct platform_gfx_pipeline_i *_i; uintptr_t sgid; }; +struct platform_gfx_shader_i; +union platform_gfx_shader { struct platform_gfx_shader_i *_i; uintptr_t sgid; }; +struct platform_gfx_texture_i; +union platform_gfx_texture { struct platform_gfx_texture_i *_i; uintptr_t sgid; }; +struct platform_gfx_sampler_i; +union platform_gfx_sampler { struct platform_gfx_sampler_i *_i; uintptr_t sgid; }; +struct platform_gfx_font_i; +union platform_gfx_font { struct platform_gfx_font_i *_i; }; + +// GFX resource descriptions +struct platform_gfx_shader_desc { + const char *vertex; + const char *fragment; +}; +struct platform_gfx_pipeline_desc { + union platform_gfx_shader shader; +}; +struct platform_gfx_texture_desc { + const void *pixels; + size_t width; + size_t height; + size_t bytesize; + int format; +}; +struct platform_gfx_sampler_desc { +}; + +// Fonts are special. They're not actual graphics objects +// that show up in most APIs, but small wrappers around +// the font file and the generated texture, so they need +// userspace allocators for themselves, and also the data +// so that they can own it and free it after finishing. +struct platform_gfx_font_desc { + struct allocator *a; + struct allocator *data_free; + struct blob data; +}; +struct platform_gfx_font_bmglyph { + // NOTE: This looks a lot like stbtt_packedchar. + // This is on purpose so that it's easier to use this in stbtt_Pack*. + // If you change this structure, make sure to change stb_impl.c as well + unsigned short x0,y0,x1,y1; + float xoff,yoff,xadvance; + float xoff2,yoff2; +}; +struct platform_gfx_font_info { + int ascent; + int descent; + int linegap; +}; + +// Gfx functions, all taking a window +int platform_gfx_valid(union platform_window window); +void platform_gfx_startup(union platform_window window); +void platform_gfx_shutdown(union platform_window window); +void platform_gfx_present(union platform_window window); + +void platform_gfx_render_begin(union platform_window window); +void platform_gfx_render_end(union platform_window window); + +union platform_gfx_shader platform_gfx_shader_construct(union platform_window window, const struct platform_gfx_shader_desc *desc); +union platform_gfx_shader platform_gfx_shader_construct(union platform_window window, const struct platform_gfx_shader_desc *desc); + +union platform_gfx_shader platform_gfx_shader_construct(union platform_window window, const struct platform_gfx_shader_desc *desc); +void platform_gfx_shader_destruct(union platform_window window, union platform_gfx_shader p); + +union platform_gfx_pipeline platform_gfx_pipeline_construct(union platform_window window, const struct platform_gfx_pipeline_desc *desc); +void platform_gfx_pipeline_destruct(union platform_window window, union platform_gfx_pipeline p); +void platform_gfx_pipeline_bind(union platform_window window, union platform_gfx_pipeline pipeline); + +union platform_gfx_texture platform_gfx_texture_construct(union platform_window window, const struct platform_gfx_texture_desc *desc); +void platform_gfx_texture_destruct(union platform_window window, union platform_gfx_texture texture); + +union platform_gfx_sampler platform_gfx_sampler_construct(union platform_window window, const struct platform_gfx_sampler_desc *desc); +void platform_gfx_sampler_destruct(union platform_window window, union platform_gfx_sampler sampler); + +union platform_gfx_font platform_gfx_font_construct(union platform_window window, const struct platform_gfx_font_desc *desc); +void platform_gfx_font_destruct(union platform_window window, union platform_gfx_font font); +struct image platform_gfx_font_texture(union platform_window window, union platform_gfx_font font, float fontsize, struct platform_gfx_font_bmglyph *glyphs); // TODO: should take valid codepoint ranges, assuming ASCII for now (0x20 ( ) - 0x7e (~)) +struct platform_gfx_font_info platform_gfx_font_info_get(union platform_window window, union platform_gfx_font font); +struct allocator *platform_gfx_font_allocator(union platform_window window, union platform_gfx_font font); + +void platform_gfx_vertices_draw(union platform_window window, union platform_gfx_texture texture, union platform_gfx_sampler sampler, const void *data, size_t bytelen, size_t vertices_count); + +// OS File functions +enum platform_os_file_mode { + PLATFORM_MODE_R = 1, +}; +enum platform_os_file_seek { + // Really should remain aliases for SEEK_* + // functions just in case those are mistakenly + // used instead of these constants, both in + // cross-platform and platform-dependent code + PLATFORM_OS_SEEK_SET = SEEK_SET, + PLATFORM_OS_SEEK_CUR = SEEK_CUR, + PLATFORM_OS_SEEK_END = SEEK_END, +}; +struct platform_os_file_i; +union platform_os_file { struct platform_os_file_i *_i; intptr_t fd; }; +void platform_os_file_close(union platform_os_file fp); +ssize_t platform_os_file_read(union platform_os_file fp, void *buf, size_t size); +ssize_t platform_os_file_write(union platform_os_file fp, const void *buf, size_t size); +int64_t platform_os_file_seek(union platform_os_file fp, int64_t delta, int from); + +// Various open functions we expect +// union platform_os_file platform_os_fs_open(const char *fname, int flags, int mode); +union platform_os_file platform_os_asset_open(const char *filename); +struct blob platform_os_asset_read_all(int type, const char *filename, struct allocator *a); + +// Image decoding functions +struct image platform_os_image_decode(const void *mem, size_t size, struct allocator *a); + +// Log functions +void platform_os_log_v(int level, const char *tag, const char *fmt, va_list ap); + +#endif diff --git a/include/types.h b/include/types.h new file mode 100644 index 0000000..2016e44 --- /dev/null +++ b/include/types.h @@ -0,0 +1,95 @@ +#ifndef HBAT_STRUCTS_H +#define HBAT_STRUCTS_H +#include "compiler.h" + +// Standard structures and types used across code base + +// Use all the POSIX types instead of defining own +#include +#include +#include +#include + +#ifdef _WIN32 +typedef intptr_t ssize_t; +#else +#include +#endif + +struct blob { + void *memory; + size_t size; + size_t capacity; + size_t alignment; +}; +#define BLOB_FROM_STATIC(m) LICAST(struct, blob){(m),sizeof(m),sizeof(m),ALIGNOF(m)} +struct allocator; +struct allocator_functions { + struct blob (*realloc_fn)(struct allocator *a, struct blob prev, size_t size, size_t align); + void (*free_fn)(struct allocator *a, struct blob mem); + + // Frees up all blobs allocated, which is useful for temporary allocators + void (*free_all_fn)(struct allocator *a); +}; +struct allocator { + struct allocator_functions *fns; + // Append anything else you want after that... +}; +struct arena { + struct allocator a; + struct blob b; + size_t ptr; + size_t ptr_last; +}; +extern struct allocator os_allocator_arena; +#define ARENA_FROM_BLOB(b) LICAST(struct, arena){os_allocator_arena, (b), 0, 0} +struct handle_table { + uint32_t *table; + uint32_t top; + uint32_t head; + uint16_t bits_idx; + uint16_t bits_gen; +}; +#define HANDLE_TABLE_STATIC(bits_idx,bits_gen) LICAST(struct, handle_table){(uint32_t[1<<(bits_idx)]){},1,0,(bits_idx),(bits_gen)} + +struct slice { + void *data; + size_t bytelen; +}; +struct slicec { + const void *data; + size_t bytelen; +}; +struct string { + char *data; + size_t len; +}; +struct stringc { + const char *data; + size_t len; +}; +#define SLICONST(slice) LICAST(struct, slicec){(slice).data, (slice).bytelen} +#define STRLIT(sliteral) LICAST(struct, stringc){sliteral, sizeof(sliteral) - 1} +#define BLOB2STR(b) LICAST(struct, string){(b).memory, (b).size} +#define BLOB2STRC(b) LICAST(struct, stringc){(b).memory, (b).size} +#define CSTR2STR(cstr) LICAST(struct, string){(cstr), strlen(cstr)} +#define CSTR2STRC(cstr) LICAST(struct, stringc){(cstr), strlen(cstr)} +#define STRCONST(s) LICAST(struct, stringc){s.data, s.len} +#define STRFMT "%.*s" +#define STRARGS(s) (s).len, (s).data + +enum image_format { + IMAGE_FORMAT_RGBA8, + IMAGE_FORMAT_RGBX8, + IMAGE_FORMAT_R8, +}; +struct image { + struct blob _i; + void *pixels; + size_t bytesize; + size_t width; + size_t height; + int format; +}; + +#endif diff --git a/logo/hdpi/explosion.png b/logo/hdpi/explosion.png new file mode 100644 index 0000000..0b079c7 Binary files /dev/null and b/logo/hdpi/explosion.png differ diff --git a/logo/ldpi/explosion.png b/logo/ldpi/explosion.png new file mode 100644 index 0000000..0b079c7 Binary files /dev/null and b/logo/ldpi/explosion.png differ diff --git a/logo/mdpi/explosion.png b/logo/mdpi/explosion.png new file mode 100644 index 0000000..0b079c7 Binary files /dev/null and b/logo/mdpi/explosion.png differ diff --git a/platform/README.md b/platform/README.md new file mode 100644 index 0000000..21337b6 --- /dev/null +++ b/platform/README.md @@ -0,0 +1,8 @@ +# HBATBP Android Sources + +Android platform files needed to load the main loop for our game + +## TODO + +Document some options in AndroidManifest that need to be changed +on Android SDK/NDK changes as well as app-specific changes diff --git a/platform/android.c b/platform/android.c new file mode 100644 index 0000000..521ff98 --- /dev/null +++ b/platform/android.c @@ -0,0 +1,362 @@ +#include +#include +#include +#include +#include +#include +#include + +#include "platform.h" +#include "core_callbacks.h" +#include "os.h" + +#define _LOG(prio, fmt, ...) ((void)__android_log_print((prio), (OSLOG_TAG_DEFAULT), (fmt)__VA_OPT__(, ) __VA_ARGS__)) +#define LOGE(fmt, ...) _LOG(ANDROID_LOG_ERROR, (fmt)__VA_OPT__(, ) __VA_ARGS__) +#define LOGW(fmt, ...) _LOG(ANDROID_LOG_WARN, (fmt)__VA_OPT__(, ) __VA_ARGS__) +#define LOGI(fmt, ...) _LOG(ANDROID_LOG_INFO, (fmt)__VA_OPT__(, ) __VA_ARGS__) + +void platform_os_log_v(int level, const char *tag, const char *fmt, va_list ap) +{ + int android_level = ANDROID_LOG_ERROR; + switch (level) + { + case LOG_FATAL: + case LOG_ERROR: android_level = ANDROID_LOG_ERROR; break; + case LOG_WARNING: android_level = ANDROID_LOG_WARN; break; + case LOG_INFO: android_level = ANDROID_LOG_INFO; break; + case LOG_VERBOSE: android_level = ANDROID_LOG_INFO; break; + case LOG_DEBUG: android_level = ANDROID_LOG_DEBUG; break; + case LOG_TRACE: android_level = ANDROID_LOG_VERBOSE; break; + default: return; + } + (void)__android_log_vprint(android_level, tag, fmt, ap); +} + +struct platform_window_i { + EGLDisplay display; + EGLSurface surface; + EGLContext context; + struct android_app *app; + int width, height; +}; +static struct android_app *global_app; + +// TODO: Just assets for now, we eventually want files +// AAsset fopencookie functions +ssize_t aasset_read(void *cookie, void *buf, size_t size) +{ + if (size > INT_MAX || !cookie) return -1; + return AAsset_read((void *)cookie, buf, size); +} +ssize_t aasset_write(void *cookie, const void *buf, size_t size) +{ + // Assets are totally read-only! + (void)cookie, (void)buf, (void)size; + return -1; +} +off_t aasset_seek(void *cookie, off_t off, int whence) +{ + if (!cookie) return -1; + return AAsset_seek((AAsset *)cookie, off, whence); +} +int aasset_close(void *cookie) +{ + if (!cookie) return -1; + AAsset_close((AAsset *)cookie); + return 0; +} +// TODO: Figure out why this prototype isn't showing up +// and declare this prototype correctly in the meanwhile +FILE *funopen(void *cookie, void *readfn, void *writefn, void *seekfn, void *closefn); +union platform_os_file platform_os_asset_open(const char *path) +{ + union platform_os_file fp = {}; + if (!global_app) { + LOGE("No Android App yet?"); + return fp; + } + + // Open asset + AAssetManager *manager = global_app->activity->assetManager; + AAsset *asset = AAssetManager_open(manager, path, AASSET_MODE_BUFFER); + if (!asset) { + LOGE("Asset '%s' failed to open", path); + return fp; + } + + FILE *stdiofp = funopen(asset, + aasset_read, + aasset_write, + aasset_seek, + aasset_close + ); + if (!stdiofp) { + LOGE("fopencookie of asset failed!"); + AAsset_close(asset); + } + fp._i = (struct platform_os_file_i *)stdiofp; + return fp; +} + +int platform_window_width(union platform_window window) +{ + return window._i ? window._i->width : 0; +} +int platform_window_height(union platform_window window) +{ + return window._i ? window._i->height : 0; +} +void platform_gfx_present(union platform_window window) +{ + if (platform_gfx_valid(window)) + eglSwapBuffers(window._i->display, window._i->surface); +} +int platform_gfx_valid(union platform_window window) +{ + return window._i && window._i->context != EGL_NO_CONTEXT; +} + +static void platform_window_i_clear(struct platform_window_i *data) +{ + // Reasonable defaults + memset(data, 0, sizeof(*data)); + data->context = EGL_NO_CONTEXT; + data->surface = EGL_NO_SURFACE; + data->display = EGL_NO_DISPLAY; +} + +static void platform_window_i_shutdown(struct platform_window_i *data) +{ + union platform_window w = { ._i = data }; + core_callback_gfx_shutdown(w); + platform_gfx_shutdown(w); + if (data->display != EGL_NO_DISPLAY) { + if (data->context != EGL_NO_CONTEXT) + eglDestroyContext(data->display, data->context); + if (data->surface != EGL_NO_SURFACE) + eglDestroySurface(data->display, data->surface); + eglTerminate(data->display); + } + platform_window_i_clear(data); +} + +static void platform_window_i_startup(struct platform_window_i *data, struct android_app *app) +{ + platform_window_i_clear(data); + LOGI("Attempting to initialize platform data..."); + + data->display = eglGetDisplay(EGL_DEFAULT_DISPLAY); + eglInitialize(data->display, (void *)0, (void *)0); + + const EGLint attribs_desired[] = { + EGL_SURFACE_TYPE, EGL_WINDOW_BIT, + EGL_RENDERABLE_TYPE, EGL_OPENGL_ES3_BIT, + EGL_BLUE_SIZE, 8, + EGL_GREEN_SIZE, 8, + EGL_RED_SIZE, 8, + EGL_NONE + }; + int configs_count = 0; + EGLConfig *configs = (EGLConfig *)0; + EGLConfig config_chosen = (EGLConfig)0; + if (eglChooseConfig(data->display, attribs_desired, (EGLConfig *)0, 0, &configs_count) + && (configs = (EGLConfig *)malloc(configs_count * sizeof(*configs))) + && eglChooseConfig(data->display, attribs_desired, configs, configs_count, &configs_count)) { + for (int i = 0; i < configs_count; i++) { + EGLConfig cfg = configs[i]; + EGLint r,g,b,d; + if (eglGetConfigAttrib(data->display, cfg, EGL_RED_SIZE, &r) + && eglGetConfigAttrib(data->display, cfg, EGL_RED_SIZE, &g) + && eglGetConfigAttrib(data->display, cfg, EGL_RED_SIZE, &b) + && eglGetConfigAttrib(data->display, cfg, EGL_RED_SIZE, &d) + && r == 8 && g == 8 && b == 8 && d == 0) { + config_chosen = cfg; + break; + } + } + if (!config_chosen && configs_count != 0) + config_chosen = configs[0]; + } + free(configs); + if (!config_chosen) { + LOGW("Could not choose EGL Config"); + goto init_fail; + } + + data->app = app; + data->surface = eglCreateWindowSurface(data->display, config_chosen, data->app->window, (void *)0); + EGLint context_attribs[] = { + EGL_CONTEXT_CLIENT_VERSION, 3, + EGL_NONE + }; + data->context = eglCreateContext(data->display, config_chosen, (void *)0, context_attribs); + if (!eglMakeCurrent(data->display, data->surface, data->surface, data->context)) { + LOGW("Could not make context current!"); + goto init_fail; + } + + eglQuerySurface(data->display, data->surface, EGL_WIDTH, &data->width); + eglQuerySurface(data->display, data->surface, EGL_HEIGHT, &data->height); + + platform_gfx_startup((union platform_window){data}); + core_callback_gfx_startup((union platform_window){data}); + LOGI("Platform data initialization successful!"); + return; +init_fail: + LOGW("Failed to initialize platform data!"); + platform_window_i_shutdown(data); +} + +static void onAppCmd(struct android_app *app, int32_t cmd) +{ + switch (cmd) { + case APP_CMD_INIT_WINDOW: + // A new window is created, associate a renderer with it. You may replace this with a + // "game" class if that suits your needs. Remember to change all instances of userData + // if you change the class here as a reinterpret_cast is dangerous this in the + // android_main function and the APP_CMD_TERM_WINDOW handler case. + if (app->window) { + struct platform_window_i *data = (struct platform_window_i *)malloc(sizeof(*data)); + if (data) { + memset(data, 0, sizeof(*data)); + platform_window_i_startup(data, app); + app->userData = (void *)data; + } + } + break; + case APP_CMD_TERM_WINDOW: + // The window is being destroyed. Use this to clean up your userData to avoid leaking + // resources. + // + // We have to check if userData is assigned just in case this comes in really quickly + if (app->userData) { + platform_window_i_shutdown(app->userData); + app->userData = (void *)0; + } + break; + default: + break; + } +} + +static int32_t onInputEvent(struct android_app *app, AInputEvent *e) +{ + (void)app; + union platform_event ev; + + switch (AInputEvent_getType(e)) { + case AINPUT_EVENT_TYPE_MOTION: { + size_t i = 0, len; + int32_t action_fields = AMotionEvent_getAction(e); + int32_t action = action_fields & AMOTION_EVENT_ACTION_MASK; + int32_t pointer_idx = (action_fields & AMOTION_EVENT_ACTION_POINTER_INDEX_MASK) + >> AMOTION_EVENT_ACTION_POINTER_INDEX_SHIFT; + switch (action) { + // When Android introduced multi-touch, they introduced pointer down and up actions + // instead of as down and up actions for what I assume to be compatibility reasons. + // We want to see these events so keep this. + case AMOTION_EVENT_ACTION_POINTER_DOWN: + case AMOTION_EVENT_ACTION_POINTER_UP: + i = pointer_idx; + len = i + 1; + ev.type = (action == AMOTION_EVENT_ACTION_POINTER_UP) + ? PLATFORM_EVENT_FINGER_UP + : PLATFORM_EVENT_FINGER_DOWN; + break; + // Single-touch action down and up action base case + case AMOTION_EVENT_ACTION_DOWN: + case AMOTION_EVENT_ACTION_UP: + len = AMotionEvent_getPointerCount(e); + ev.type = (action == AMOTION_EVENT_ACTION_UP) + ? PLATFORM_EVENT_FINGER_UP + : PLATFORM_EVENT_FINGER_DOWN; + break; + // Move action for motion event base case + case AMOTION_EVENT_ACTION_MOVE: + len = AMotionEvent_getPointerCount(e); + ev.type = PLATFORM_EVENT_FINGER_MOTION; + break; + default: + return 0; + } + ev.finger.device = AInputEvent_getDeviceId(e); + for (; i < len; i++) { + ev.finger.id = AMotionEvent_getPointerId(e, i); + ev.finger.pixels.x = AMotionEvent_getX(e, i); + ev.finger.pixels.y = AMotionEvent_getY(e, i); + ev.finger.pressure = AMotionEvent_getPressure(e, i); + void *w = (void *)0; + if (!app) + app = global_app; + if (app) + w = app->userData; + core_callback_event(&ev, (union platform_window){ + (struct platform_window_i *)w + }); + } + return 0; + } + default: + LOGI("I got some input event?"); + return 0; + } +} + +/*! + * This the main entry point for a native activity + */ +static void dummy_entry(void) +{ + // If instructed to wait for a debugger, use SIGSTOP + // and then the debugger will release SIGSTOP +#if WAIT_FOR_DEBUGGER == 1 + raise(SIGSTOP); +#endif + char *argv[1] = {NULL}; core_callback_entry(0, argv); +} +static union platform_os_once dummy_entry_once = PLATFORM_OS_ONCE_INIT; +void android_main(struct android_app *pApp) +{ + platform_os_once_run(&dummy_entry_once, dummy_entry); + LOGI("Hello from android_main!"); + + // Register an event handler for Android events + global_app = pApp; + pApp->onAppCmd = onAppCmd; + pApp->onInputEvent = onInputEvent; + + // This sets up a typical game/event loop. It will run until the app is destroyed. + do { + // Process all pending events before running game logic. + bool done = false; + while (!done) { + // 0 is non-blocking. + int timeout = 0; + int events; + struct android_poll_source *pSource; + int result = ALooper_pollOnce(timeout, 0, &events, (void **)&pSource); + switch (result) { + case ALOOPER_POLL_TIMEOUT: + //[[clang::fallthrough]]; + case ALOOPER_POLL_WAKE: + // No events occurred before the timeout or explicit wake. Stop checking for events. + done = true; + break; + case ALOOPER_EVENT_ERROR: + //aout << "ALooper_pollOnce returned an error" << std::endl; + break; + case ALOOPER_POLL_CALLBACK: + break; + default: + if (pSource) { + pSource->process(pApp, pSource); + } + } + } + + // Do the updates and rendering + core_callback_render((union platform_window){(struct platform_window_i *)pApp->userData}); + } while (!pApp->destroyRequested); + + global_app = (struct android_app *)0; +} diff --git a/platform/miniaudio.h b/platform/miniaudio.h new file mode 100644 index 0000000..53499f9 --- /dev/null +++ b/platform/miniaudio.h @@ -0,0 +1,93468 @@ +/* +Audio playback and capture library. Choice of public domain or MIT-0. See license statements at the end of this file. +miniaudio - v0.11.22 - 2025-02-24 + +David Reid - mackron@gmail.com + +Website: https://miniaud.io +Documentation: https://miniaud.io/docs +GitHub: https://github.com/mackron/miniaudio +*/ + +/* +1. Introduction +=============== +To use miniaudio, include "miniaudio.h": + + ```c + #include "miniaudio.h" + ``` + +The implementation is contained in "miniaudio.c". Just compile this like any other source file. You +can include miniaudio.c if you want to compile your project as a single translation unit: + + ```c + #include "miniaudio.c" + ``` + +miniaudio includes both low level and high level APIs. The low level API is good for those who want +to do all of their mixing themselves and only require a light weight interface to the underlying +audio device. The high level API is good for those who have complex mixing and effect requirements. + +In miniaudio, objects are transparent structures. Unlike many other libraries, there are no handles +to opaque objects which means you need to allocate memory for objects yourself. In the examples +presented in this documentation you will often see objects declared on the stack. You need to be +careful when translating these examples to your own code so that you don't accidentally declare +your objects on the stack and then cause them to become invalid once the function returns. In +addition, you must ensure the memory address of your objects remain the same throughout their +lifetime. You therefore cannot be making copies of your objects. + +A config/init pattern is used throughout the entire library. The idea is that you set up a config +object and pass that into the initialization routine. The advantage to this system is that the +config object can be initialized with logical defaults and new properties added to it without +breaking the API. The config object can be allocated on the stack and does not need to be +maintained after initialization of the corresponding object. + + +1.1. Low Level API +------------------ +The low level API gives you access to the raw audio data of an audio device. It supports playback, +capture, full-duplex and loopback (WASAPI only). You can enumerate over devices to determine which +physical device(s) you want to connect to. + +The low level API uses the concept of a "device" as the abstraction for physical devices. The idea +is that you choose a physical device to emit or capture audio from, and then move data to/from the +device when miniaudio tells you to. Data is delivered to and from devices asynchronously via a +callback which you specify when initializing the device. + +When initializing the device you first need to configure it. The device configuration allows you to +specify things like the format of the data delivered via the callback, the size of the internal +buffer and the ID of the device you want to emit or capture audio from. + +Once you have the device configuration set up you can initialize the device. When initializing a +device you need to allocate memory for the device object beforehand. This gives the application +complete control over how the memory is allocated. In the example below we initialize a playback +device on the stack, but you could allocate it on the heap if that suits your situation better. + + ```c + void data_callback(ma_device* pDevice, void* pOutput, const void* pInput, ma_uint32 frameCount) + { + // In playback mode copy data to pOutput. In capture mode read data from pInput. In full-duplex mode, both + // pOutput and pInput will be valid and you can move data from pInput into pOutput. Never process more than + // frameCount frames. + } + + int main() + { + ma_device_config config = ma_device_config_init(ma_device_type_playback); + config.playback.format = ma_format_f32; // Set to ma_format_unknown to use the device's native format. + config.playback.channels = 2; // Set to 0 to use the device's native channel count. + config.sampleRate = 48000; // Set to 0 to use the device's native sample rate. + config.dataCallback = data_callback; // This function will be called when miniaudio needs more data. + config.pUserData = pMyCustomData; // Can be accessed from the device object (device.pUserData). + + ma_device device; + if (ma_device_init(NULL, &config, &device) != MA_SUCCESS) { + return -1; // Failed to initialize the device. + } + + ma_device_start(&device); // The device is sleeping by default so you'll need to start it manually. + + // Do something here. Probably your program's main loop. + + ma_device_uninit(&device); + return 0; + } + ``` + +In the example above, `data_callback()` is where audio data is written and read from the device. +The idea is in playback mode you cause sound to be emitted from the speakers by writing audio data +to the output buffer (`pOutput` in the example). In capture mode you read data from the input +buffer (`pInput`) to extract sound captured by the microphone. The `frameCount` parameter tells you +how many frames can be written to the output buffer and read from the input buffer. A "frame" is +one sample for each channel. For example, in a stereo stream (2 channels), one frame is 2 +samples: one for the left, one for the right. The channel count is defined by the device config. +The size in bytes of an individual sample is defined by the sample format which is also specified +in the device config. Multi-channel audio data is always interleaved, which means the samples for +each frame are stored next to each other in memory. For example, in a stereo stream the first pair +of samples will be the left and right samples for the first frame, the second pair of samples will +be the left and right samples for the second frame, etc. + +The configuration of the device is defined by the `ma_device_config` structure. The config object +is always initialized with `ma_device_config_init()`. It's important to always initialize the +config with this function as it initializes it with logical defaults and ensures your program +doesn't break when new members are added to the `ma_device_config` structure. The example above +uses a fairly simple and standard device configuration. The call to `ma_device_config_init()` takes +a single parameter, which is whether or not the device is a playback, capture, duplex or loopback +device (loopback devices are not supported on all backends). The `config.playback.format` member +sets the sample format which can be one of the following (all formats are native-endian): + + +---------------+----------------------------------------+---------------------------+ + | Symbol | Description | Range | + +---------------+----------------------------------------+---------------------------+ + | ma_format_f32 | 32-bit floating point | [-1, 1] | + | ma_format_s16 | 16-bit signed integer | [-32768, 32767] | + | ma_format_s24 | 24-bit signed integer (tightly packed) | [-8388608, 8388607] | + | ma_format_s32 | 32-bit signed integer | [-2147483648, 2147483647] | + | ma_format_u8 | 8-bit unsigned integer | [0, 255] | + +---------------+----------------------------------------+---------------------------+ + +The `config.playback.channels` member sets the number of channels to use with the device. The +channel count cannot exceed MA_MAX_CHANNELS. The `config.sampleRate` member sets the sample rate +(which must be the same for both playback and capture in full-duplex configurations). This is +usually set to 44100 or 48000, but can be set to anything. It's recommended to keep this between +8000 and 384000, however. + +Note that leaving the format, channel count and/or sample rate at their default values will result +in the internal device's native configuration being used which is useful if you want to avoid the +overhead of miniaudio's automatic data conversion. + +In addition to the sample format, channel count and sample rate, the data callback and user data +pointer are also set via the config. The user data pointer is not passed into the callback as a +parameter, but is instead set to the `pUserData` member of `ma_device` which you can access +directly since all miniaudio structures are transparent. + +Initializing the device is done with `ma_device_init()`. This will return a result code telling you +what went wrong, if anything. On success it will return `MA_SUCCESS`. After initialization is +complete the device will be in a stopped state. To start it, use `ma_device_start()`. +Uninitializing the device will stop it, which is what the example above does, but you can also stop +the device with `ma_device_stop()`. To resume the device simply call `ma_device_start()` again. +Note that it's important to never stop or start the device from inside the callback. This will +result in a deadlock. Instead you set a variable or signal an event indicating that the device +needs to stop and handle it in a different thread. The following APIs must never be called inside +the callback: + + ```c + ma_device_init() + ma_device_init_ex() + ma_device_uninit() + ma_device_start() + ma_device_stop() + ``` + +You must never try uninitializing and reinitializing a device inside the callback. You must also +never try to stop and start it from inside the callback. There are a few other things you shouldn't +do in the callback depending on your requirements, however this isn't so much a thread-safety +thing, but rather a real-time processing thing which is beyond the scope of this introduction. + +The example above demonstrates the initialization of a playback device, but it works exactly the +same for capture. All you need to do is change the device type from `ma_device_type_playback` to +`ma_device_type_capture` when setting up the config, like so: + + ```c + ma_device_config config = ma_device_config_init(ma_device_type_capture); + config.capture.format = MY_FORMAT; + config.capture.channels = MY_CHANNEL_COUNT; + ``` + +In the data callback you just read from the input buffer (`pInput` in the example above) and leave +the output buffer alone (it will be set to NULL when the device type is set to +`ma_device_type_capture`). + +These are the available device types and how you should handle the buffers in the callback: + + +-------------------------+--------------------------------------------------------+ + | Device Type | Callback Behavior | + +-------------------------+--------------------------------------------------------+ + | ma_device_type_playback | Write to output buffer, leave input buffer untouched. | + | ma_device_type_capture | Read from input buffer, leave output buffer untouched. | + | ma_device_type_duplex | Read from input buffer, write to output buffer. | + | ma_device_type_loopback | Read from input buffer, leave output buffer untouched. | + +-------------------------+--------------------------------------------------------+ + +You will notice in the example above that the sample format and channel count is specified +separately for playback and capture. This is to support different data formats between the playback +and capture devices in a full-duplex system. An example may be that you want to capture audio data +as a monaural stream (one channel), but output sound to a stereo speaker system. Note that if you +use different formats between playback and capture in a full-duplex configuration you will need to +convert the data yourself. There are functions available to help you do this which will be +explained later. + +The example above did not specify a physical device to connect to which means it will use the +operating system's default device. If you have multiple physical devices connected and you want to +use a specific one you will need to specify the device ID in the configuration, like so: + + ```c + config.playback.pDeviceID = pMyPlaybackDeviceID; // Only if requesting a playback or duplex device. + config.capture.pDeviceID = pMyCaptureDeviceID; // Only if requesting a capture, duplex or loopback device. + ``` + +To retrieve the device ID you will need to perform device enumeration, however this requires the +use of a new concept called the "context". Conceptually speaking the context sits above the device. +There is one context to many devices. The purpose of the context is to represent the backend at a +more global level and to perform operations outside the scope of an individual device. Mainly it is +used for performing run-time linking against backend libraries, initializing backends and +enumerating devices. The example below shows how to enumerate devices. + + ```c + ma_context context; + if (ma_context_init(NULL, 0, NULL, &context) != MA_SUCCESS) { + // Error. + } + + ma_device_info* pPlaybackInfos; + ma_uint32 playbackCount; + ma_device_info* pCaptureInfos; + ma_uint32 captureCount; + if (ma_context_get_devices(&context, &pPlaybackInfos, &playbackCount, &pCaptureInfos, &captureCount) != MA_SUCCESS) { + // Error. + } + + // Loop over each device info and do something with it. Here we just print the name with their index. You may want + // to give the user the opportunity to choose which device they'd prefer. + for (ma_uint32 iDevice = 0; iDevice < playbackCount; iDevice += 1) { + printf("%d - %s\n", iDevice, pPlaybackInfos[iDevice].name); + } + + ma_device_config config = ma_device_config_init(ma_device_type_playback); + config.playback.pDeviceID = &pPlaybackInfos[chosenPlaybackDeviceIndex].id; + config.playback.format = MY_FORMAT; + config.playback.channels = MY_CHANNEL_COUNT; + config.sampleRate = MY_SAMPLE_RATE; + config.dataCallback = data_callback; + config.pUserData = pMyCustomData; + + ma_device device; + if (ma_device_init(&context, &config, &device) != MA_SUCCESS) { + // Error + } + + ... + + ma_device_uninit(&device); + ma_context_uninit(&context); + ``` + +The first thing we do in this example is initialize a `ma_context` object with `ma_context_init()`. +The first parameter is a pointer to a list of `ma_backend` values which are used to override the +default backend priorities. When this is NULL, as in this example, miniaudio's default priorities +are used. The second parameter is the number of backends listed in the array pointed to by the +first parameter. The third parameter is a pointer to a `ma_context_config` object which can be +NULL, in which case defaults are used. The context configuration is used for setting the logging +callback, custom memory allocation callbacks, user-defined data and some backend-specific +configurations. + +Once the context has been initialized you can enumerate devices. In the example above we use the +simpler `ma_context_get_devices()`, however you can also use a callback for handling devices by +using `ma_context_enumerate_devices()`. When using `ma_context_get_devices()` you provide a pointer +to a pointer that will, upon output, be set to a pointer to a buffer containing a list of +`ma_device_info` structures. You also provide a pointer to an unsigned integer that will receive +the number of items in the returned buffer. Do not free the returned buffers as their memory is +managed internally by miniaudio. + +The `ma_device_info` structure contains an `id` member which is the ID you pass to the device +config. It also contains the name of the device which is useful for presenting a list of devices +to the user via the UI. + +When creating your own context you will want to pass it to `ma_device_init()` when initializing the +device. Passing in NULL, like we do in the first example, will result in miniaudio creating the +context for you, which you don't want to do since you've already created a context. Note that +internally the context is only tracked by it's pointer which means you must not change the location +of the `ma_context` object. If this is an issue, consider using `malloc()` to allocate memory for +the context. + + +1.2. High Level API +------------------- +The high level API consists of three main parts: + + * Resource management for loading and streaming sounds. + * A node graph for advanced mixing and effect processing. + * A high level "engine" that wraps around the resource manager and node graph. + +The resource manager (`ma_resource_manager`) is used for loading sounds. It supports loading sounds +fully into memory and also streaming. It will also deal with reference counting for you which +avoids the same sound being loaded multiple times. + +The node graph is used for mixing and effect processing. The idea is that you connect a number of +nodes into the graph by connecting each node's outputs to another node's inputs. Each node can +implement its own effect. By chaining nodes together, advanced mixing and effect processing can +be achieved. + +The engine encapsulates both the resource manager and the node graph to create a simple, easy to +use high level API. The resource manager and node graph APIs are covered in more later sections of +this manual. + +The code below shows how you can initialize an engine using it's default configuration. + + ```c + ma_result result; + ma_engine engine; + + result = ma_engine_init(NULL, &engine); + if (result != MA_SUCCESS) { + return result; // Failed to initialize the engine. + } + ``` + +This creates an engine instance which will initialize a device internally which you can access with +`ma_engine_get_device()`. It will also initialize a resource manager for you which can be accessed +with `ma_engine_get_resource_manager()`. The engine itself is a node graph (`ma_node_graph`) which +means you can pass a pointer to the engine object into any of the `ma_node_graph` APIs (with a +cast). Alternatively, you can use `ma_engine_get_node_graph()` instead of a cast. + +Note that all objects in miniaudio, including the `ma_engine` object in the example above, are +transparent structures. There are no handles to opaque structures in miniaudio which means you need +to be mindful of how you declare them. In the example above we are declaring it on the stack, but +this will result in the struct being invalidated once the function encapsulating it returns. If +allocating the engine on the heap is more appropriate, you can easily do so with a standard call +to `malloc()` or whatever heap allocation routine you like: + + ```c + ma_engine* pEngine = malloc(sizeof(*pEngine)); + ``` + +The `ma_engine` API uses the same config/init pattern used all throughout miniaudio. To configure +an engine, you can fill out a `ma_engine_config` object and pass it into the first parameter of +`ma_engine_init()`: + + ```c + ma_result result; + ma_engine engine; + ma_engine_config engineConfig; + + engineConfig = ma_engine_config_init(); + engineConfig.pResourceManager = &myCustomResourceManager; // <-- Initialized as some earlier stage. + + result = ma_engine_init(&engineConfig, &engine); + if (result != MA_SUCCESS) { + return result; + } + ``` + +This creates an engine instance using a custom config. In this particular example it's showing how +you can specify a custom resource manager rather than having the engine initialize one internally. +This is particularly useful if you want to have multiple engine's share the same resource manager. + +The engine must be uninitialized with `ma_engine_uninit()` when it's no longer needed. + +By default the engine will be started, but nothing will be playing because no sounds have been +initialized. The easiest but least flexible way of playing a sound is like so: + + ```c + ma_engine_play_sound(&engine, "my_sound.wav", NULL); + ``` + +This plays what miniaudio calls an "inline" sound. It plays the sound once, and then puts the +internal sound up for recycling. The last parameter is used to specify which sound group the sound +should be associated with which will be explained later. This particular way of playing a sound is +simple, but lacks flexibility and features. A more flexible way of playing a sound is to first +initialize a sound: + + ```c + ma_result result; + ma_sound sound; + + result = ma_sound_init_from_file(&engine, "my_sound.wav", 0, NULL, NULL, &sound); + if (result != MA_SUCCESS) { + return result; + } + + ma_sound_start(&sound); + ``` + +This returns a `ma_sound` object which represents a single instance of the specified sound file. If +you want to play the same file multiple times simultaneously, you need to create one sound for each +instance. + +Sounds should be uninitialized with `ma_sound_uninit()`. + +Sounds are not started by default. Start a sound with `ma_sound_start()` and stop it with +`ma_sound_stop()`. When a sound is stopped, it is not rewound to the start. Use +`ma_sound_seek_to_pcm_frame(&sound, 0)` to seek back to the start of a sound. By default, starting +and stopping sounds happens immediately, but sometimes it might be convenient to schedule the sound +the be started and/or stopped at a specific time. This can be done with the following functions: + + ```c + ma_sound_set_start_time_in_pcm_frames() + ma_sound_set_start_time_in_milliseconds() + ma_sound_set_stop_time_in_pcm_frames() + ma_sound_set_stop_time_in_milliseconds() + ``` + +The start/stop time needs to be specified based on the absolute timer which is controlled by the +engine. The current global time in PCM frames can be retrieved with +`ma_engine_get_time_in_pcm_frames()`. The engine's global time can be changed with +`ma_engine_set_time_in_pcm_frames()` for synchronization purposes if required. Note that scheduling +a start time still requires an explicit call to `ma_sound_start()` before anything will play: + + ```c + ma_sound_set_start_time_in_pcm_frames(&sound, ma_engine_get_time_in_pcm_frames(&engine) + (ma_engine_get_sample_rate(&engine) * 2); + ma_sound_start(&sound); + ``` + +The third parameter of `ma_sound_init_from_file()` is a set of flags that control how the sound be +loaded and a few options on which features should be enabled for that sound. By default, the sound +is synchronously loaded fully into memory straight from the file system without any kind of +decoding. If you want to decode the sound before storing it in memory, you need to specify the +`MA_SOUND_FLAG_DECODE` flag. This is useful if you want to incur the cost of decoding at an earlier +stage, such as a loading stage. Without this option, decoding will happen dynamically at mixing +time which might be too expensive on the audio thread. + +If you want to load the sound asynchronously, you can specify the `MA_SOUND_FLAG_ASYNC` flag. This +will result in `ma_sound_init_from_file()` returning quickly, but the sound will not start playing +until the sound has had some audio decoded. + +The fourth parameter is a pointer to sound group. A sound group is used as a mechanism to organise +sounds into groups which have their own effect processing and volume control. An example is a game +which might have separate groups for sfx, voice and music. Each of these groups have their own +independent volume control. Use `ma_sound_group_init()` or `ma_sound_group_init_ex()` to initialize +a sound group. + +Sounds and sound groups are nodes in the engine's node graph and can be plugged into any `ma_node` +API. This makes it possible to connect sounds and sound groups to effect nodes to produce complex +effect chains. + +A sound can have its volume changed with `ma_sound_set_volume()`. If you prefer decibel volume +control you can use `ma_volume_db_to_linear()` to convert from decibel representation to linear. + +Panning and pitching is supported with `ma_sound_set_pan()` and `ma_sound_set_pitch()`. If you know +a sound will never have its pitch changed with `ma_sound_set_pitch()` or via the doppler effect, +you can specify the `MA_SOUND_FLAG_NO_PITCH` flag when initializing the sound for an optimization. + +By default, sounds and sound groups have spatialization enabled. If you don't ever want to +spatialize your sounds, initialize the sound with the `MA_SOUND_FLAG_NO_SPATIALIZATION` flag. The +spatialization model is fairly simple and is roughly on feature parity with OpenAL. HRTF and +environmental occlusion are not currently supported, but planned for the future. The supported +features include: + + * Sound and listener positioning and orientation with cones + * Attenuation models: none, inverse, linear and exponential + * Doppler effect + +Sounds can be faded in and out with `ma_sound_set_fade_in_pcm_frames()`. + +To check if a sound is currently playing, you can use `ma_sound_is_playing()`. To check if a sound +is at the end, use `ma_sound_at_end()`. Looping of a sound can be controlled with +`ma_sound_set_looping()`. Use `ma_sound_is_looping()` to check whether or not the sound is looping. + + + +2. Building +=========== +miniaudio should work cleanly out of the box without the need to download or install any +dependencies. See below for platform-specific details. + +Note that GCC and Clang require `-msse2`, `-mavx2`, etc. for SIMD optimizations. + +If you get errors about undefined references to `__sync_val_compare_and_swap_8`, `__atomic_load_8`, +etc. you need to link with `-latomic`. + + +2.1. Windows +------------ +The Windows build should compile cleanly on all popular compilers without the need to configure any +include paths nor link to any libraries. + +The UWP build may require linking to mmdevapi.lib if you get errors about an unresolved external +symbol for `ActivateAudioInterfaceAsync()`. + + +2.2. macOS and iOS +------------------ +The macOS build should compile cleanly without the need to download any dependencies nor link to +any libraries or frameworks. The iOS build needs to be compiled as Objective-C and will need to +link the relevant frameworks but should compile cleanly out of the box with Xcode. Compiling +through the command line requires linking to `-lpthread` and `-lm`. + +Due to the way miniaudio links to frameworks at runtime, your application may not pass Apple's +notarization process. To fix this there are two options. The first is to compile with +`-DMA_NO_RUNTIME_LINKING` which in turn will require linking with +`-framework CoreFoundation -framework CoreAudio -framework AudioToolbox`. If you get errors about +AudioToolbox, try with `-framework AudioUnit` instead. You may get this when using older versions +of iOS. Alternatively, if you would rather keep using runtime linking you can add the following to +your entitlements.xcent file: + + ``` + com.apple.security.cs.allow-dyld-environment-variables + + com.apple.security.cs.allow-unsigned-executable-memory + + ``` + +See this discussion for more info: https://github.com/mackron/miniaudio/issues/203. + + +2.3. Linux +---------- +The Linux build only requires linking to `-ldl`, `-lpthread` and `-lm`. You do not need any +development packages. You may need to link with `-latomic` if you're compiling for 32-bit ARM. + + +2.4. BSD +-------- +The BSD build only requires linking to `-lpthread` and `-lm`. NetBSD uses audio(4), OpenBSD uses +sndio and FreeBSD uses OSS. You may need to link with `-latomic` if you're compiling for 32-bit +ARM. + + +2.5. Android +------------ +AAudio is the highest priority backend on Android. This should work out of the box without needing +any kind of compiler configuration. Support for AAudio starts with Android 8 which means older +versions will fall back to OpenSL|ES which requires API level 16+. + +There have been reports that the OpenSL|ES backend fails to initialize on some Android based +devices due to `dlopen()` failing to open "libOpenSLES.so". If this happens on your platform +you'll need to disable run-time linking with `MA_NO_RUNTIME_LINKING` and link with -lOpenSLES. + + +2.6. Emscripten +--------------- +The Emscripten build emits Web Audio JavaScript directly and should compile cleanly out of the box. +You cannot use `-std=c*` compiler flags, nor `-ansi`. + +You can enable the use of AudioWorkets by defining `MA_ENABLE_AUDIO_WORKLETS` and then compiling +with the following options: + + -sAUDIO_WORKLET=1 -sWASM_WORKERS=1 -sASYNCIFY + +An example for compiling with AudioWorklet support might look like this: + + emcc program.c -o bin/program.html -DMA_ENABLE_AUDIO_WORKLETS -sAUDIO_WORKLET=1 -sWASM_WORKERS=1 -sASYNCIFY + +To run locally, you'll need to use emrun: + + emrun bin/program.html + + + +2.7. Build Options +------------------ +`#define` these options before including miniaudio.c, or pass them as compiler flags: + + +----------------------------------+--------------------------------------------------------------------+ + | Option | Description | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_WASAPI | Disables the WASAPI backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_DSOUND | Disables the DirectSound backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_WINMM | Disables the WinMM backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_ALSA | Disables the ALSA backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_PULSEAUDIO | Disables the PulseAudio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_JACK | Disables the JACK backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_COREAUDIO | Disables the Core Audio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_SNDIO | Disables the sndio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_AUDIO4 | Disables the audio(4) backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_OSS | Disables the OSS backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_AAUDIO | Disables the AAudio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_OPENSL | Disables the OpenSL|ES backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_WEBAUDIO | Disables the Web Audio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_CUSTOM | Disables support for custom backends. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_NULL | Disables the null backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_ONLY_SPECIFIC_BACKENDS | Disables all backends by default and requires `MA_ENABLE_*` to | + | | enable specific backends. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_WASAPI | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the WASAPI backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_DSOUND | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the DirectSound backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_WINMM | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the WinMM backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_ALSA | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the ALSA backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_PULSEAUDIO | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the PulseAudio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_JACK | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the JACK backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_COREAUDIO | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the Core Audio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_SNDIO | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the sndio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_AUDIO4 | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the audio(4) backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_OSS | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the OSS backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_AAUDIO | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the AAudio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_OPENSL | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the OpenSL|ES backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_WEBAUDIO | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the Web Audio backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_CUSTOM | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable custom backends. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ENABLE_NULL | Used in conjunction with MA_ENABLE_ONLY_SPECIFIC_BACKENDS to | + | | enable the null backend. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_DECODING | Disables decoding APIs. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_ENCODING | Disables encoding APIs. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_WAV | Disables the built-in WAV decoder and encoder. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_FLAC | Disables the built-in FLAC decoder. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_MP3 | Disables the built-in MP3 decoder. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_DEVICE_IO | Disables playback and recording. This will disable `ma_context` | + | | and `ma_device` APIs. This is useful if you only want to use | + | | miniaudio's data conversion and/or decoding APIs. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_RESOURCE_MANAGER | Disables the resource manager. When using the engine this will | + | | also disable the following functions: | + | | | + | | ``` | + | | ma_sound_init_from_file() | + | | ma_sound_init_from_file_w() | + | | ma_sound_init_copy() | + | | ma_engine_play_sound_ex() | + | | ma_engine_play_sound() | + | | ``` | + | | | + | | The only way to initialize a `ma_sound` object is to initialize it | + | | from a data source. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_NODE_GRAPH | Disables the node graph API. This will also disable the engine API | + | | because it depends on the node graph. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_ENGINE | Disables the engine API. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_THREADING | Disables the `ma_thread`, `ma_mutex`, `ma_semaphore` and | + | | `ma_event` APIs. This option is useful if you only need to use | + | | miniaudio for data conversion, decoding and/or encoding. Some | + | | families of APIs require threading which means the following | + | | options must also be set: | + | | | + | | ``` | + | | MA_NO_DEVICE_IO | + | | ``` | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_GENERATION | Disables generation APIs such a `ma_waveform` and `ma_noise`. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_SSE2 | Disables SSE2 optimizations. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_AVX2 | Disables AVX2 optimizations. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_NEON | Disables NEON optimizations. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_NO_RUNTIME_LINKING | Disables runtime linking. This is useful for passing Apple's | + | | notarization process. When enabling this, you may need to avoid | + | | using `-std=c89` or `-std=c99` on Linux builds or else you may end | + | | up with compilation errors due to conflicts with `timespec` and | + | | `timeval` data types. | + | | | + | | You may need to enable this if your target platform does not allow | + | | runtime linking via `dlopen()`. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_USE_STDINT | (Pass this in as a compiler flag. Do not `#define` this before | + | | miniaudio.c) Forces the use of stdint.h for sized types. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_DEBUG_OUTPUT | Enable `printf()` output of debug logs (`MA_LOG_LEVEL_DEBUG`). | + +----------------------------------+--------------------------------------------------------------------+ + | MA_COINIT_VALUE | Windows only. The value to pass to internal calls to | + | | `CoInitializeEx()`. Defaults to `COINIT_MULTITHREADED`. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_FORCE_UWP | Windows only. Affects only the WASAPI backend. Will force the | + | | WASAPI backend to use the UWP code path instead of the regular | + | | desktop path. This is normally auto-detected and should rarely be | + | | needed to be used explicitly, but can be useful for debugging. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ON_THREAD_ENTRY | Defines some code that will be executed as soon as an internal | + | | miniaudio-managed thread is created. This will be the first thing | + | | to be executed by the thread entry point. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_ON_THREAD_EXIT | Defines some code that will be executed from the entry point of an | + | | internal miniaudio-managed thread upon exit. This will be the last | + | | thing to be executed before the thread's entry point exits. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_THREAD_DEFAULT_STACK_SIZE | If set, specifies the default stack size used by miniaudio-managed | + | | threads. | + +----------------------------------+--------------------------------------------------------------------+ + | MA_API | Controls how public APIs should be decorated. Default is `extern`. | + +----------------------------------+--------------------------------------------------------------------+ + + +3. Definitions +============== +This section defines common terms used throughout miniaudio. Unfortunately there is often ambiguity +in the use of terms throughout the audio space, so this section is intended to clarify how miniaudio +uses each term. + +3.1. Sample +----------- +A sample is a single unit of audio data. If the sample format is f32, then one sample is one 32-bit +floating point number. + +3.2. Frame / PCM Frame +---------------------- +A frame is a group of samples equal to the number of channels. For a stereo stream a frame is 2 +samples, a mono frame is 1 sample, a 5.1 surround sound frame is 6 samples, etc. The terms "frame" +and "PCM frame" are the same thing in miniaudio. Note that this is different to a compressed frame. +If ever miniaudio needs to refer to a compressed frame, such as a FLAC frame, it will always +clarify what it's referring to with something like "FLAC frame". + +3.3. Channel +------------ +A stream of monaural audio that is emitted from an individual speaker in a speaker system, or +received from an individual microphone in a microphone system. A stereo stream has two channels (a +left channel, and a right channel), a 5.1 surround sound system has 6 channels, etc. Some audio +systems refer to a channel as a complex audio stream that's mixed with other channels to produce +the final mix - this is completely different to miniaudio's use of the term "channel" and should +not be confused. + +3.4. Sample Rate +---------------- +The sample rate in miniaudio is always expressed in Hz, such as 44100, 48000, etc. It's the number +of PCM frames that are processed per second. + +3.5. Formats +------------ +Throughout miniaudio you will see references to different sample formats: + + +---------------+----------------------------------------+---------------------------+ + | Symbol | Description | Range | + +---------------+----------------------------------------+---------------------------+ + | ma_format_f32 | 32-bit floating point | [-1, 1] | + | ma_format_s16 | 16-bit signed integer | [-32768, 32767] | + | ma_format_s24 | 24-bit signed integer (tightly packed) | [-8388608, 8388607] | + | ma_format_s32 | 32-bit signed integer | [-2147483648, 2147483647] | + | ma_format_u8 | 8-bit unsigned integer | [0, 255] | + +---------------+----------------------------------------+---------------------------+ + +All formats are native-endian. + + + +4. Data Sources +=============== +The data source abstraction in miniaudio is used for retrieving audio data from some source. A few +examples include `ma_decoder`, `ma_noise` and `ma_waveform`. You will need to be familiar with data +sources in order to make sense of some of the higher level concepts in miniaudio. + +The `ma_data_source` API is a generic interface for reading from a data source. Any object that +implements the data source interface can be plugged into any `ma_data_source` function. + +To read data from a data source: + + ```c + ma_result result; + ma_uint64 framesRead; + + result = ma_data_source_read_pcm_frames(pDataSource, pFramesOut, frameCount, &framesRead); + if (result != MA_SUCCESS) { + return result; // Failed to read data from the data source. + } + ``` + +If you don't need the number of frames that were successfully read you can pass in `NULL` to the +`pFramesRead` parameter. If this returns a value less than the number of frames requested it means +the end of the file has been reached. `MA_AT_END` will be returned only when the number of frames +read is 0. + +When calling any data source function, with the exception of `ma_data_source_init()` and +`ma_data_source_uninit()`, you can pass in any object that implements a data source. For example, +you could plug in a decoder like so: + + ```c + ma_result result; + ma_uint64 framesRead; + ma_decoder decoder; // <-- This would be initialized with `ma_decoder_init_*()`. + + result = ma_data_source_read_pcm_frames(&decoder, pFramesOut, frameCount, &framesRead); + if (result != MA_SUCCESS) { + return result; // Failed to read data from the decoder. + } + ``` + +If you want to seek forward you can pass in `NULL` to the `pFramesOut` parameter. Alternatively you +can use `ma_data_source_seek_pcm_frames()`. + +To seek to a specific PCM frame: + + ```c + result = ma_data_source_seek_to_pcm_frame(pDataSource, frameIndex); + if (result != MA_SUCCESS) { + return result; // Failed to seek to PCM frame. + } + ``` + +You can retrieve the total length of a data source in PCM frames, but note that some data sources +may not have the notion of a length, such as noise and waveforms, and others may just not have a +way of determining the length such as some decoders. To retrieve the length: + + ```c + ma_uint64 length; + + result = ma_data_source_get_length_in_pcm_frames(pDataSource, &length); + if (result != MA_SUCCESS) { + return result; // Failed to retrieve the length. + } + ``` + +Care should be taken when retrieving the length of a data source where the underlying decoder is +pulling data from a data stream with an undefined length, such as internet radio or some kind of +broadcast. If you do this, `ma_data_source_get_length_in_pcm_frames()` may never return. + +The current position of the cursor in PCM frames can also be retrieved: + + ```c + ma_uint64 cursor; + + result = ma_data_source_get_cursor_in_pcm_frames(pDataSource, &cursor); + if (result != MA_SUCCESS) { + return result; // Failed to retrieve the cursor. + } + ``` + +You will often need to know the data format that will be returned after reading. This can be +retrieved like so: + + ```c + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_channel channelMap[MA_MAX_CHANNELS]; + + result = ma_data_source_get_data_format(pDataSource, &format, &channels, &sampleRate, channelMap, MA_MAX_CHANNELS); + if (result != MA_SUCCESS) { + return result; // Failed to retrieve data format. + } + ``` + +If you do not need a specific data format property, just pass in NULL to the respective parameter. + +There may be cases where you want to implement something like a sound bank where you only want to +read data within a certain range of the underlying data. To do this you can use a range: + + ```c + result = ma_data_source_set_range_in_pcm_frames(pDataSource, rangeBegInFrames, rangeEndInFrames); + if (result != MA_SUCCESS) { + return result; // Failed to set the range. + } + ``` + +This is useful if you have a sound bank where many sounds are stored in the same file and you want +the data source to only play one of those sub-sounds. Note that once the range is set, everything +that takes a position, such as cursors and loop points, should always be relatvie to the start of +the range. When the range is set, any previously defined loop point will be reset. + +Custom loop points can also be used with data sources. By default, data sources will loop after +they reach the end of the data source, but if you need to loop at a specific location, you can do +the following: + + ```c + result = ma_data_set_loop_point_in_pcm_frames(pDataSource, loopBegInFrames, loopEndInFrames); + if (result != MA_SUCCESS) { + return result; // Failed to set the loop point. + } + ``` + +The loop point is relative to the current range. + +It's sometimes useful to chain data sources together so that a seamless transition can be achieved. +To do this, you can use chaining: + + ```c + ma_decoder decoder1; + ma_decoder decoder2; + + // ... initialize decoders with ma_decoder_init_*() ... + + result = ma_data_source_set_next(&decoder1, &decoder2); + if (result != MA_SUCCESS) { + return result; // Failed to set the next data source. + } + + result = ma_data_source_read_pcm_frames(&decoder1, pFramesOut, frameCount, pFramesRead); + if (result != MA_SUCCESS) { + return result; // Failed to read from the decoder. + } + ``` + +In the example above we're using decoders. When reading from a chain, you always want to read from +the top level data source in the chain. In the example above, `decoder1` is the top level data +source in the chain. When `decoder1` reaches the end, `decoder2` will start seamlessly without any +gaps. + +Note that when looping is enabled, only the current data source will be looped. You can loop the +entire chain by linking in a loop like so: + + ```c + ma_data_source_set_next(&decoder1, &decoder2); // decoder1 -> decoder2 + ma_data_source_set_next(&decoder2, &decoder1); // decoder2 -> decoder1 (loop back to the start). + ``` + +Note that setting up chaining is not thread safe, so care needs to be taken if you're dynamically +changing links while the audio thread is in the middle of reading. + +Do not use `ma_decoder_seek_to_pcm_frame()` as a means to reuse a data source to play multiple +instances of the same sound simultaneously. This can be extremely inefficient depending on the type +of data source and can result in glitching due to subtle changes to the state of internal filters. +Instead, initialize multiple data sources for each instance. + + +4.1. Custom Data Sources +------------------------ +You can implement a custom data source by implementing the functions in `ma_data_source_vtable`. +Your custom object must have `ma_data_source_base` as it's first member: + + ```c + struct my_data_source + { + ma_data_source_base base; + ... + }; + ``` + +In your initialization routine, you need to call `ma_data_source_init()` in order to set up the +base object (`ma_data_source_base`): + + ```c + static ma_result my_data_source_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) + { + // Read data here. Output in the same format returned by my_data_source_get_data_format(). + } + + static ma_result my_data_source_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) + { + // Seek to a specific PCM frame here. Return MA_NOT_IMPLEMENTED if seeking is not supported. + } + + static ma_result my_data_source_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) + { + // Return the format of the data here. + } + + static ma_result my_data_source_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) + { + // Retrieve the current position of the cursor here. Return MA_NOT_IMPLEMENTED and set *pCursor to 0 if there is no notion of a cursor. + } + + static ma_result my_data_source_get_length(ma_data_source* pDataSource, ma_uint64* pLength) + { + // Retrieve the length in PCM frames here. Return MA_NOT_IMPLEMENTED and set *pLength to 0 if there is no notion of a length or if the length is unknown. + } + + static ma_data_source_vtable g_my_data_source_vtable = + { + my_data_source_read, + my_data_source_seek, + my_data_source_get_data_format, + my_data_source_get_cursor, + my_data_source_get_length + }; + + ma_result my_data_source_init(my_data_source* pMyDataSource) + { + ma_result result; + ma_data_source_config baseConfig; + + baseConfig = ma_data_source_config_init(); + baseConfig.vtable = &g_my_data_source_vtable; + + result = ma_data_source_init(&baseConfig, &pMyDataSource->base); + if (result != MA_SUCCESS) { + return result; + } + + // ... do the initialization of your custom data source here ... + + return MA_SUCCESS; + } + + void my_data_source_uninit(my_data_source* pMyDataSource) + { + // ... do the uninitialization of your custom data source here ... + + // You must uninitialize the base data source. + ma_data_source_uninit(&pMyDataSource->base); + } + ``` + +Note that `ma_data_source_init()` and `ma_data_source_uninit()` are never called directly outside +of the custom data source. It's up to the custom data source itself to call these within their own +init/uninit functions. + + + +5. Engine +========= +The `ma_engine` API is a high level API for managing and mixing sounds and effect processing. The +`ma_engine` object encapsulates a resource manager and a node graph, both of which will be +explained in more detail later. + +Sounds are called `ma_sound` and are created from an engine. Sounds can be associated with a mixing +group called `ma_sound_group` which are also created from the engine. Both `ma_sound` and +`ma_sound_group` objects are nodes within the engine's node graph. + +When the engine is initialized, it will normally create a device internally. If you would rather +manage the device yourself, you can do so and just pass a pointer to it via the engine config when +you initialize the engine. You can also just use the engine without a device, which again can be +configured via the engine config. + +The most basic way to initialize the engine is with a default config, like so: + + ```c + ma_result result; + ma_engine engine; + + result = ma_engine_init(NULL, &engine); + if (result != MA_SUCCESS) { + return result; // Failed to initialize the engine. + } + ``` + +This will result in the engine initializing a playback device using the operating system's default +device. This will be sufficient for many use cases, but if you need more flexibility you'll want to +configure the engine with an engine config: + + ```c + ma_result result; + ma_engine engine; + ma_engine_config engineConfig; + + engineConfig = ma_engine_config_init(); + engineConfig.pDevice = &myDevice; + + result = ma_engine_init(&engineConfig, &engine); + if (result != MA_SUCCESS) { + return result; // Failed to initialize the engine. + } + ``` + +In the example above we're passing in a pre-initialized device. Since the caller is the one in +control of the device's data callback, it's their responsibility to manually call +`ma_engine_read_pcm_frames()` from inside their data callback: + + ```c + void playback_data_callback(ma_device* pDevice, void* pOutput, const void* pInput, ma_uint32 frameCount) + { + ma_engine_read_pcm_frames(&g_Engine, pOutput, frameCount, NULL); + } + ``` + +You can also use the engine independent of a device entirely: + + ```c + ma_result result; + ma_engine engine; + ma_engine_config engineConfig; + + engineConfig = ma_engine_config_init(); + engineConfig.noDevice = MA_TRUE; + engineConfig.channels = 2; // Must be set when not using a device. + engineConfig.sampleRate = 48000; // Must be set when not using a device. + + result = ma_engine_init(&engineConfig, &engine); + if (result != MA_SUCCESS) { + return result; // Failed to initialize the engine. + } + ``` + +Note that when you're not using a device, you must set the channel count and sample rate in the +config or else miniaudio won't know what to use (miniaudio will use the device to determine this +normally). When not using a device, you need to use `ma_engine_read_pcm_frames()` to process audio +data from the engine. This kind of setup is useful if you want to do something like offline +processing or want to use a different audio system for playback such as SDL. + +When a sound is loaded it goes through a resource manager. By default the engine will initialize a +resource manager internally, but you can also specify a pre-initialized resource manager: + + ```c + ma_result result; + ma_engine engine1; + ma_engine engine2; + ma_engine_config engineConfig; + + engineConfig = ma_engine_config_init(); + engineConfig.pResourceManager = &myResourceManager; + + ma_engine_init(&engineConfig, &engine1); + ma_engine_init(&engineConfig, &engine2); + ``` + +In this example we are initializing two engines, both of which are sharing the same resource +manager. This is especially useful for saving memory when loading the same file across multiple +engines. If you were not to use a shared resource manager, each engine instance would use their own +which would result in any sounds that are used between both engine's being loaded twice. By using +a shared resource manager, it would only be loaded once. Using multiple engine's is useful when you +need to output to multiple playback devices, such as in a local multiplayer game where each player +is using their own set of headphones. + +By default an engine will be in a started state. To make it so the engine is not automatically +started you can configure it as such: + + ```c + engineConfig.noAutoStart = MA_TRUE; + + // The engine will need to be started manually. + ma_engine_start(&engine); + + // Later on the engine can be stopped with ma_engine_stop(). + ma_engine_stop(&engine); + ``` + +The concept of starting or stopping an engine is only relevant when using the engine with a +device. Attempting to start or stop an engine that is not associated with a device will result in +`MA_INVALID_OPERATION`. + +The master volume of the engine can be controlled with `ma_engine_set_volume()` which takes a +linear scale, with 0 resulting in silence and anything above 1 resulting in amplification. If you +prefer decibel based volume control, use `ma_volume_db_to_linear()` to convert from dB to linear. + +When a sound is spatialized, it is done so relative to a listener. An engine can be configured to +have multiple listeners which can be configured via the config: + + ```c + engineConfig.listenerCount = 2; + ``` + +The maximum number of listeners is restricted to `MA_ENGINE_MAX_LISTENERS`. By default, when a +sound is spatialized, it will be done so relative to the closest listener. You can also pin a sound +to a specific listener which will be explained later. Listener's have a position, direction, cone, +and velocity (for doppler effect). A listener is referenced by an index, the meaning of which is up +to the caller (the index is 0 based and cannot go beyond the listener count, minus 1). The +position, direction and velocity are all specified in absolute terms: + + ```c + ma_engine_listener_set_position(&engine, listenerIndex, worldPosX, worldPosY, worldPosZ); + ``` + +The direction of the listener represents it's forward vector. The listener's up vector can also be +specified and defaults to +1 on the Y axis. + + ```c + ma_engine_listener_set_direction(&engine, listenerIndex, forwardX, forwardY, forwardZ); + ma_engine_listener_set_world_up(&engine, listenerIndex, 0, 1, 0); + ``` + +The engine supports directional attenuation. The listener can have a cone the controls how sound is +attenuated based on the listener's direction. When a sound is between the inner and outer cones, it +will be attenuated between 1 and the cone's outer gain: + + ```c + ma_engine_listener_set_cone(&engine, listenerIndex, innerAngleInRadians, outerAngleInRadians, outerGain); + ``` + +When a sound is inside the inner code, no directional attenuation is applied. When the sound is +outside of the outer cone, the attenuation will be set to `outerGain` in the example above. When +the sound is in between the inner and outer cones, the attenuation will be interpolated between 1 +and the outer gain. + +The engine's coordinate system follows the OpenGL coordinate system where positive X points right, +positive Y points up and negative Z points forward. + +The simplest and least flexible way to play a sound is like so: + + ```c + ma_engine_play_sound(&engine, "my_sound.wav", pGroup); + ``` + +This is a "fire and forget" style of function. The engine will manage the `ma_sound` object +internally. When the sound finishes playing, it'll be put up for recycling. For more flexibility +you'll want to initialize a sound object: + + ```c + ma_sound sound; + + result = ma_sound_init_from_file(&engine, "my_sound.wav", flags, pGroup, NULL, &sound); + if (result != MA_SUCCESS) { + return result; // Failed to load sound. + } + ``` + +Sounds need to be uninitialized with `ma_sound_uninit()`. + +The example above loads a sound from a file. If the resource manager has been disabled you will not +be able to use this function and instead you'll need to initialize a sound directly from a data +source: + + ```c + ma_sound sound; + + result = ma_sound_init_from_data_source(&engine, &dataSource, flags, pGroup, &sound); + if (result != MA_SUCCESS) { + return result; + } + ``` + +Each `ma_sound` object represents a single instance of the sound. If you want to play the same +sound multiple times at the same time, you need to initialize a separate `ma_sound` object. + +For the most flexibility when initializing sounds, use `ma_sound_init_ex()`. This uses miniaudio's +standard config/init pattern: + + ```c + ma_sound sound; + ma_sound_config soundConfig; + + soundConfig = ma_sound_config_init(); + soundConfig.pFilePath = NULL; // Set this to load from a file path. + soundConfig.pDataSource = NULL; // Set this to initialize from an existing data source. + soundConfig.pInitialAttachment = &someNodeInTheNodeGraph; + soundConfig.initialAttachmentInputBusIndex = 0; + soundConfig.channelsIn = 1; + soundConfig.channelsOut = 0; // Set to 0 to use the engine's native channel count. + + result = ma_sound_init_ex(&soundConfig, &sound); + if (result != MA_SUCCESS) { + return result; + } + ``` + +In the example above, the sound is being initialized without a file nor a data source. This is +valid, in which case the sound acts as a node in the middle of the node graph. This means you can +connect other sounds to this sound and allow it to act like a sound group. Indeed, this is exactly +what a `ma_sound_group` is. + +When loading a sound, you specify a set of flags that control how the sound is loaded and what +features are enabled for that sound. When no flags are set, the sound will be fully loaded into +memory in exactly the same format as how it's stored on the file system. The resource manager will +allocate a block of memory and then load the file directly into it. When reading audio data, it +will be decoded dynamically on the fly. In order to save processing time on the audio thread, it +might be beneficial to pre-decode the sound. You can do this with the `MA_SOUND_FLAG_DECODE` flag: + + ```c + ma_sound_init_from_file(&engine, "my_sound.wav", MA_SOUND_FLAG_DECODE, pGroup, NULL, &sound); + ``` + +By default, sounds will be loaded synchronously, meaning `ma_sound_init_*()` will not return until +the sound has been fully loaded. If this is prohibitive you can instead load sounds asynchronously +by specifying the `MA_SOUND_FLAG_ASYNC` flag: + + ```c + ma_sound_init_from_file(&engine, "my_sound.wav", MA_SOUND_FLAG_DECODE | MA_SOUND_FLAG_ASYNC, pGroup, NULL, &sound); + ``` + +This will result in `ma_sound_init_*()` returning quickly, but the sound won't yet have been fully +loaded. When you start the sound, it won't output anything until some sound is available. The sound +will start outputting audio before the sound has been fully decoded when the `MA_SOUND_FLAG_DECODE` +is specified. + +If you need to wait for an asynchronously loaded sound to be fully loaded, you can use a fence. A +fence in miniaudio is a simple synchronization mechanism which simply blocks until it's internal +counter hit's zero. You can specify a fence like so: + + ```c + ma_result result; + ma_fence fence; + ma_sound sounds[4]; + + result = ma_fence_init(&fence); + if (result != MA_SUCCESS) { + return result; + } + + // Load some sounds asynchronously. + for (int iSound = 0; iSound < 4; iSound += 1) { + ma_sound_init_from_file(&engine, mySoundFilesPaths[iSound], MA_SOUND_FLAG_DECODE | MA_SOUND_FLAG_ASYNC, pGroup, &fence, &sounds[iSound]); + } + + // ... do some other stuff here in the mean time ... + + // Wait for all sounds to finish loading. + ma_fence_wait(&fence); + ``` + +If loading the entire sound into memory is prohibitive, you can also configure the engine to stream +the audio data: + + ```c + ma_sound_init_from_file(&engine, "my_sound.wav", MA_SOUND_FLAG_STREAM, pGroup, NULL, &sound); + ``` + +When streaming sounds, 2 seconds worth of audio data is stored in memory. Although it should work +fine, it's inefficient to use streaming for short sounds. Streaming is useful for things like music +tracks in games. + +When loading a sound from a file path, the engine will reference count the file to prevent it from +being loaded if it's already in memory. When you uninitialize a sound, the reference count will be +decremented, and if it hits zero, the sound will be unloaded from memory. This reference counting +system is not used for streams. The engine will use a 64-bit hash of the file name when comparing +file paths which means there's a small chance you might encounter a name collision. If this is an +issue, you'll need to use a different name for one of the colliding file paths, or just not load +from files and instead load from a data source. + +You can use `ma_sound_init_copy()` to initialize a copy of another sound. Note, however, that this +only works for sounds that were initialized with `ma_sound_init_from_file()` and without the +`MA_SOUND_FLAG_STREAM` flag. + +When you initialize a sound, if you specify a sound group the sound will be attached to that group +automatically. If you set it to NULL, it will be automatically attached to the engine's endpoint. +If you would instead rather leave the sound unattached by default, you can specify the +`MA_SOUND_FLAG_NO_DEFAULT_ATTACHMENT` flag. This is useful if you want to set up a complex node +graph. + +Sounds are not started by default. To start a sound, use `ma_sound_start()`. Stop a sound with +`ma_sound_stop()`. + +Sounds can have their volume controlled with `ma_sound_set_volume()` in the same way as the +engine's master volume. + +Sounds support stereo panning and pitching. Set the pan with `ma_sound_set_pan()`. Setting the pan +to 0 will result in an unpanned sound. Setting it to -1 will shift everything to the left, whereas ++1 will shift it to the right. The pitch can be controlled with `ma_sound_set_pitch()`. A larger +value will result in a higher pitch. The pitch must be greater than 0. + +The engine supports 3D spatialization of sounds. By default sounds will have spatialization +enabled, but if a sound does not need to be spatialized it's best to disable it. There are two ways +to disable spatialization of a sound: + + ```c + // Disable spatialization at initialization time via a flag: + ma_sound_init_from_file(&engine, "my_sound.wav", MA_SOUND_FLAG_NO_SPATIALIZATION, NULL, NULL, &sound); + + // Dynamically disable or enable spatialization post-initialization: + ma_sound_set_spatialization_enabled(&sound, isSpatializationEnabled); + ``` + +By default sounds will be spatialized based on the closest listener. If a sound should always be +spatialized relative to a specific listener it can be pinned to one: + + ```c + ma_sound_set_pinned_listener_index(&sound, listenerIndex); + ``` + +Like listeners, sounds have a position. By default, the position of a sound is in absolute space, +but it can be changed to be relative to a listener: + + ```c + ma_sound_set_positioning(&sound, ma_positioning_relative); + ``` + +Note that relative positioning of a sound only makes sense if there is either only one listener, or +the sound is pinned to a specific listener. To set the position of a sound: + + ```c + ma_sound_set_position(&sound, posX, posY, posZ); + ``` + +The direction works the same way as a listener and represents the sound's forward direction: + + ```c + ma_sound_set_direction(&sound, forwardX, forwardY, forwardZ); + ``` + +Sound's also have a cone for controlling directional attenuation. This works exactly the same as +listeners: + + ```c + ma_sound_set_cone(&sound, innerAngleInRadians, outerAngleInRadians, outerGain); + ``` + +The velocity of a sound is used for doppler effect and can be set as such: + + ```c + ma_sound_set_velocity(&sound, velocityX, velocityY, velocityZ); + ``` + +The engine supports different attenuation models which can be configured on a per-sound basis. By +default the attenuation model is set to `ma_attenuation_model_inverse` which is the equivalent to +OpenAL's `AL_INVERSE_DISTANCE_CLAMPED`. Configure the attenuation model like so: + + ```c + ma_sound_set_attenuation_model(&sound, ma_attenuation_model_inverse); + ``` + +The supported attenuation models include the following: + + +----------------------------------+----------------------------------------------+ + | ma_attenuation_model_none | No distance attenuation. | + +----------------------------------+----------------------------------------------+ + | ma_attenuation_model_inverse | Equivalent to `AL_INVERSE_DISTANCE_CLAMPED`. | + +----------------------------------+----------------------------------------------+ + | ma_attenuation_model_linear | Linear attenuation. | + +----------------------------------+----------------------------------------------+ + | ma_attenuation_model_exponential | Exponential attenuation. | + +----------------------------------+----------------------------------------------+ + +To control how quickly a sound rolls off as it moves away from the listener, you need to configure +the rolloff: + + ```c + ma_sound_set_rolloff(&sound, rolloff); + ``` + +You can control the minimum and maximum gain to apply from spatialization: + + ```c + ma_sound_set_min_gain(&sound, minGain); + ma_sound_set_max_gain(&sound, maxGain); + ``` + +Likewise, in the calculation of attenuation, you can control the minimum and maximum distances for +the attenuation calculation. This is useful if you want to ensure sounds don't drop below a certain +volume after the listener moves further away and to have sounds play a maximum volume when the +listener is within a certain distance: + + ```c + ma_sound_set_min_distance(&sound, minDistance); + ma_sound_set_max_distance(&sound, maxDistance); + ``` + +The engine's spatialization system supports doppler effect. The doppler factor can be configure on +a per-sound basis like so: + + ```c + ma_sound_set_doppler_factor(&sound, dopplerFactor); + ``` + +You can fade sounds in and out with `ma_sound_set_fade_in_pcm_frames()` and +`ma_sound_set_fade_in_milliseconds()`. Set the volume to -1 to use the current volume as the +starting volume: + + ```c + // Fade in over 1 second. + ma_sound_set_fade_in_milliseconds(&sound, 0, 1, 1000); + + // ... sometime later ... + + // Fade out over 1 second, starting from the current volume. + ma_sound_set_fade_in_milliseconds(&sound, -1, 0, 1000); + ``` + +By default sounds will start immediately, but sometimes for timing and synchronization purposes it +can be useful to schedule a sound to start or stop: + + ```c + // Start the sound in 1 second from now. + ma_sound_set_start_time_in_pcm_frames(&sound, ma_engine_get_time_in_pcm_frames(&engine) + (ma_engine_get_sample_rate(&engine) * 1)); + + // Stop the sound in 2 seconds from now. + ma_sound_set_stop_time_in_pcm_frames(&sound, ma_engine_get_time_in_pcm_frames(&engine) + (ma_engine_get_sample_rate(&engine) * 2)); + ``` + +Note that scheduling a start time still requires an explicit call to `ma_sound_start()` before +anything will play. + +The time is specified in global time which is controlled by the engine. You can get the engine's +current time with `ma_engine_get_time_in_pcm_frames()`. The engine's global time is incremented +automatically as audio data is read, but it can be reset with `ma_engine_set_time_in_pcm_frames()` +in case it needs to be resynchronized for some reason. + +To determine whether or not a sound is currently playing, use `ma_sound_is_playing()`. This will +take the scheduled start and stop times into account. + +Whether or not a sound should loop can be controlled with `ma_sound_set_looping()`. Sounds will not +be looping by default. Use `ma_sound_is_looping()` to determine whether or not a sound is looping. + +Use `ma_sound_at_end()` to determine whether or not a sound is currently at the end. For a looping +sound this should never return true. Alternatively, you can configure a callback that will be fired +when the sound reaches the end. Note that the callback is fired from the audio thread which means +you cannot be uninitializing sound from the callback. To set the callback you can use +`ma_sound_set_end_callback()`. Alternatively, if you're using `ma_sound_init_ex()`, you can pass it +into the config like so: + + ```c + soundConfig.endCallback = my_end_callback; + soundConfig.pEndCallbackUserData = pMyEndCallbackUserData; + ``` + +The end callback is declared like so: + + ```c + void my_end_callback(void* pUserData, ma_sound* pSound) + { + ... + } + ``` + +Internally a sound wraps around a data source. Some APIs exist to control the underlying data +source, mainly for convenience: + + ```c + ma_sound_seek_to_pcm_frame(&sound, frameIndex); + ma_sound_get_data_format(&sound, &format, &channels, &sampleRate, pChannelMap, channelMapCapacity); + ma_sound_get_cursor_in_pcm_frames(&sound, &cursor); + ma_sound_get_length_in_pcm_frames(&sound, &length); + ``` + +Sound groups have the same API as sounds, only they are called `ma_sound_group`, and since they do +not have any notion of a data source, anything relating to a data source is unavailable. + +Internally, sound data is loaded via the `ma_decoder` API which means by default it only supports +file formats that have built-in support in miniaudio. You can extend this to support any kind of +file format through the use of custom decoders. To do this you'll need to use a self-managed +resource manager and configure it appropriately. See the "Resource Management" section below for +details on how to set this up. + + +6. Resource Management +====================== +Many programs will want to manage sound resources for things such as reference counting and +streaming. This is supported by miniaudio via the `ma_resource_manager` API. + +The resource manager is mainly responsible for the following: + + * Loading of sound files into memory with reference counting. + * Streaming of sound data. + +When loading a sound file, the resource manager will give you back a `ma_data_source` compatible +object called `ma_resource_manager_data_source`. This object can be passed into any +`ma_data_source` API which is how you can read and seek audio data. When loading a sound file, you +specify whether or not you want the sound to be fully loaded into memory (and optionally +pre-decoded) or streamed. When loading into memory, you can also specify whether or not you want +the data to be loaded asynchronously. + +The example below is how you can initialize a resource manager using it's default configuration: + + ```c + ma_resource_manager_config config; + ma_resource_manager resourceManager; + + config = ma_resource_manager_config_init(); + result = ma_resource_manager_init(&config, &resourceManager); + if (result != MA_SUCCESS) { + ma_device_uninit(&device); + printf("Failed to initialize the resource manager."); + return -1; + } + ``` + +You can configure the format, channels and sample rate of the decoded audio data. By default it +will use the file's native data format, but you can configure it to use a consistent format. This +is useful for offloading the cost of data conversion to load time rather than dynamically +converting at mixing time. To do this, you configure the decoded format, channels and sample rate +like the code below: + + ```c + config = ma_resource_manager_config_init(); + config.decodedFormat = device.playback.format; + config.decodedChannels = device.playback.channels; + config.decodedSampleRate = device.sampleRate; + ``` + +In the code above, the resource manager will be configured so that any decoded audio data will be +pre-converted at load time to the device's native data format. If instead you used defaults and +the data format of the file did not match the device's data format, you would need to convert the +data at mixing time which may be prohibitive in high-performance and large scale scenarios like +games. + +Internally the resource manager uses the `ma_decoder` API to load sounds. This means by default it +only supports decoders that are built into miniaudio. It's possible to support additional encoding +formats through the use of custom decoders. To do so, pass in your `ma_decoding_backend_vtable` +vtables into the resource manager config: + + ```c + ma_decoding_backend_vtable* pCustomBackendVTables[] = + { + &g_ma_decoding_backend_vtable_libvorbis, + &g_ma_decoding_backend_vtable_libopus + }; + + ... + + resourceManagerConfig.ppCustomDecodingBackendVTables = pCustomBackendVTables; + resourceManagerConfig.customDecodingBackendCount = sizeof(pCustomBackendVTables) / sizeof(pCustomBackendVTables[0]); + resourceManagerConfig.pCustomDecodingBackendUserData = NULL; + ``` + +This system can allow you to support any kind of file format. See the "Decoding" section for +details on how to implement custom decoders. The miniaudio repository includes examples for Opus +via libopus and libopusfile and Vorbis via libvorbis and libvorbisfile. + +Asynchronicity is achieved via a job system. When an operation needs to be performed, such as the +decoding of a page, a job will be posted to a queue which will then be processed by a job thread. +By default there will be only one job thread running, but this can be configured, like so: + + ```c + config = ma_resource_manager_config_init(); + config.jobThreadCount = MY_JOB_THREAD_COUNT; + ``` + +By default job threads are managed internally by the resource manager, however you can also self +manage your job threads if, for example, you want to integrate the job processing into your +existing job infrastructure, or if you simply don't like the way the resource manager does it. To +do this, just set the job thread count to 0 and process jobs manually. To process jobs, you first +need to retrieve a job using `ma_resource_manager_next_job()` and then process it using +`ma_job_process()`: + + ```c + config = ma_resource_manager_config_init(); + config.jobThreadCount = 0; // Don't manage any job threads internally. + config.flags = MA_RESOURCE_MANAGER_FLAG_NON_BLOCKING; // Optional. Makes `ma_resource_manager_next_job()` non-blocking. + + // ... Initialize your custom job threads ... + + void my_custom_job_thread(...) + { + for (;;) { + ma_job job; + ma_result result = ma_resource_manager_next_job(pMyResourceManager, &job); + if (result != MA_SUCCESS) { + if (result == MA_NO_DATA_AVAILABLE) { + // No jobs are available. Keep going. Will only get this if the resource manager was initialized + // with MA_RESOURCE_MANAGER_FLAG_NON_BLOCKING. + continue; + } else if (result == MA_CANCELLED) { + // MA_JOB_TYPE_QUIT was posted. Exit. + break; + } else { + // Some other error occurred. + break; + } + } + + ma_job_process(&job); + } + } + ``` + +In the example above, the `MA_JOB_TYPE_QUIT` event is the used as the termination +indicator, but you can use whatever you would like to terminate the thread. The call to +`ma_resource_manager_next_job()` is blocking by default, but can be configured to be non-blocking +by initializing the resource manager with the `MA_RESOURCE_MANAGER_FLAG_NON_BLOCKING` configuration +flag. Note that the `MA_JOB_TYPE_QUIT` will never be removed from the job queue. This +is to give every thread the opportunity to catch the event and terminate naturally. + +When loading a file, it's sometimes convenient to be able to customize how files are opened and +read instead of using standard `fopen()`, `fclose()`, etc. which is what miniaudio will use by +default. This can be done by setting `pVFS` member of the resource manager's config: + + ```c + // Initialize your custom VFS object. See documentation for VFS for information on how to do this. + my_custom_vfs vfs = my_custom_vfs_init(); + + config = ma_resource_manager_config_init(); + config.pVFS = &vfs; + ``` + +This is particularly useful in programs like games where you want to read straight from an archive +rather than the normal file system. If you do not specify a custom VFS, the resource manager will +use the operating system's normal file operations. + +To load a sound file and create a data source, call `ma_resource_manager_data_source_init()`. When +loading a sound you need to specify the file path and options for how the sounds should be loaded. +By default a sound will be loaded synchronously. The returned data source is owned by the caller +which means the caller is responsible for the allocation and freeing of the data source. Below is +an example for initializing a data source: + + ```c + ma_resource_manager_data_source dataSource; + ma_result result = ma_resource_manager_data_source_init(pResourceManager, pFilePath, flags, &dataSource); + if (result != MA_SUCCESS) { + // Error. + } + + // ... + + // A ma_resource_manager_data_source object is compatible with the `ma_data_source` API. To read data, just call + // the `ma_data_source_read_pcm_frames()` like you would with any normal data source. + result = ma_data_source_read_pcm_frames(&dataSource, pDecodedData, frameCount, &framesRead); + if (result != MA_SUCCESS) { + // Failed to read PCM frames. + } + + // ... + + ma_resource_manager_data_source_uninit(&dataSource); + ``` + +The `flags` parameter specifies how you want to perform loading of the sound file. It can be a +combination of the following flags: + + ``` + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_DECODE + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING + ``` + +When no flags are specified (set to 0), the sound will be fully loaded into memory, but not +decoded, meaning the raw file data will be stored in memory, and then dynamically decoded when +`ma_data_source_read_pcm_frames()` is called. To instead decode the audio data before storing it in +memory, use the `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_DECODE` flag. By default, the sound file will +be loaded synchronously, meaning `ma_resource_manager_data_source_init()` will only return after +the entire file has been loaded. This is good for simplicity, but can be prohibitively slow. You +can instead load the sound asynchronously using the `MA_RESOURCE_MANAGER_DATA_SOURCE_ASYNC` flag. +This will result in `ma_resource_manager_data_source_init()` returning quickly, but no data will be +returned by `ma_data_source_read_pcm_frames()` until some data is available. When no data is +available because the asynchronous decoding hasn't caught up, `MA_BUSY` will be returned by +`ma_data_source_read_pcm_frames()`. + +For large sounds, it's often prohibitive to store the entire file in memory. To mitigate this, you +can instead stream audio data which you can do by specifying the +`MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM` flag. When streaming, data will be decoded in 1 +second pages. When a new page needs to be decoded, a job will be posted to the job queue and then +subsequently processed in a job thread. + +The `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING` flag can be used so that the sound will loop +when it reaches the end by default. It's recommended you use this flag when you want to have a +looping streaming sound. If you try loading a very short sound as a stream, you will get a glitch. +This is because the resource manager needs to pre-fill the initial buffer at initialization time, +and if you don't specify the `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING` flag, the resource +manager will assume the sound is not looping and will stop filling the buffer when it reaches the +end, therefore resulting in a discontinuous buffer. + +For in-memory sounds, reference counting is used to ensure the data is loaded only once. This means +multiple calls to `ma_resource_manager_data_source_init()` with the same file path will result in +the file data only being loaded once. Each call to `ma_resource_manager_data_source_init()` must be +matched up with a call to `ma_resource_manager_data_source_uninit()`. Sometimes it can be useful +for a program to register self-managed raw audio data and associate it with a file path. Use the +`ma_resource_manager_register_*()` and `ma_resource_manager_unregister_*()` APIs to do this. +`ma_resource_manager_register_decoded_data()` is used to associate a pointer to raw, self-managed +decoded audio data in the specified data format with the specified name. Likewise, +`ma_resource_manager_register_encoded_data()` is used to associate a pointer to raw self-managed +encoded audio data (the raw file data) with the specified name. Note that these names need not be +actual file paths. When `ma_resource_manager_data_source_init()` is called (without the +`MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM` flag), the resource manager will look for these +explicitly registered data buffers and, if found, will use it as the backing data for the data +source. Note that the resource manager does *not* make a copy of this data so it is up to the +caller to ensure the pointer stays valid for its lifetime. Use +`ma_resource_manager_unregister_data()` to unregister the self-managed data. You can also use +`ma_resource_manager_register_file()` and `ma_resource_manager_unregister_file()` to register and +unregister a file. It does not make sense to use the `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM` +flag with a self-managed data pointer. + + +6.1. Asynchronous Loading and Synchronization +--------------------------------------------- +When loading asynchronously, it can be useful to poll whether or not loading has finished. Use +`ma_resource_manager_data_source_result()` to determine this. For in-memory sounds, this will +return `MA_SUCCESS` when the file has been *entirely* decoded. If the sound is still being decoded, +`MA_BUSY` will be returned. Otherwise, some other error code will be returned if the sound failed +to load. For streaming data sources, `MA_SUCCESS` will be returned when the first page has been +decoded and the sound is ready to be played. If the first page is still being decoded, `MA_BUSY` +will be returned. Otherwise, some other error code will be returned if the sound failed to load. + +In addition to polling, you can also use a simple synchronization object called a "fence" to wait +for asynchronously loaded sounds to finish. This is called `ma_fence`. The advantage to using a +fence is that it can be used to wait for a group of sounds to finish loading rather than waiting +for sounds on an individual basis. There are two stages to loading a sound: + + * Initialization of the internal decoder; and + * Completion of decoding of the file (the file is fully decoded) + +You can specify separate fences for each of the different stages. Waiting for the initialization +of the internal decoder is important for when you need to know the sample format, channels and +sample rate of the file. + +The example below shows how you could use a fence when loading a number of sounds: + + ```c + // This fence will be released when all sounds are finished loading entirely. + ma_fence fence; + ma_fence_init(&fence); + + // This will be passed into the initialization routine for each sound. + ma_resource_manager_pipeline_notifications notifications = ma_resource_manager_pipeline_notifications_init(); + notifications.done.pFence = &fence; + + // Now load a bunch of sounds: + for (iSound = 0; iSound < soundCount; iSound += 1) { + ma_resource_manager_data_source_init(pResourceManager, pSoundFilePaths[iSound], flags, ¬ifications, &pSoundSources[iSound]); + } + + // ... DO SOMETHING ELSE WHILE SOUNDS ARE LOADING ... + + // Wait for loading of sounds to finish. + ma_fence_wait(&fence); + ``` + +In the example above we used a fence for waiting until the entire file has been fully decoded. If +you only need to wait for the initialization of the internal decoder to complete, you can use the +`init` member of the `ma_resource_manager_pipeline_notifications` object: + + ```c + notifications.init.pFence = &fence; + ``` + +If a fence is not appropriate for your situation, you can instead use a callback that is fired on +an individual sound basis. This is done in a very similar way to fences: + + ```c + typedef struct + { + ma_async_notification_callbacks cb; + void* pMyData; + } my_notification; + + void my_notification_callback(ma_async_notification* pNotification) + { + my_notification* pMyNotification = (my_notification*)pNotification; + + // Do something in response to the sound finishing loading. + } + + ... + + my_notification myCallback; + myCallback.cb.onSignal = my_notification_callback; + myCallback.pMyData = pMyData; + + ma_resource_manager_pipeline_notifications notifications = ma_resource_manager_pipeline_notifications_init(); + notifications.done.pNotification = &myCallback; + + ma_resource_manager_data_source_init(pResourceManager, "my_sound.wav", flags, ¬ifications, &mySound); + ``` + +In the example above we just extend the `ma_async_notification_callbacks` object and pass an +instantiation into the `ma_resource_manager_pipeline_notifications` in the same way as we did with +the fence, only we set `pNotification` instead of `pFence`. You can set both of these at the same +time and they should both work as expected. If using the `pNotification` system, you need to ensure +your `ma_async_notification_callbacks` object stays valid. + + + +6.2. Resource Manager Implementation Details +-------------------------------------------- +Resources are managed in two main ways: + + * By storing the entire sound inside an in-memory buffer (referred to as a data buffer) + * By streaming audio data on the fly (referred to as a data stream) + +A resource managed data source (`ma_resource_manager_data_source`) encapsulates a data buffer or +data stream, depending on whether or not the data source was initialized with the +`MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM` flag. If so, it will make use of a +`ma_resource_manager_data_stream` object. Otherwise it will use a `ma_resource_manager_data_buffer` +object. Both of these objects are data sources which means they can be used with any +`ma_data_source_*()` API. + +Another major feature of the resource manager is the ability to asynchronously decode audio files. +This relieves the audio thread of time-consuming decoding which can negatively affect scalability +due to the audio thread needing to complete it's work extremely quickly to avoid glitching. +Asynchronous decoding is achieved through a job system. There is a central multi-producer, +multi-consumer, fixed-capacity job queue. When some asynchronous work needs to be done, a job is +posted to the queue which is then read by a job thread. The number of job threads can be +configured for improved scalability, and job threads can all run in parallel without needing to +worry about the order of execution (how this is achieved is explained below). + +When a sound is being loaded asynchronously, playback can begin before the sound has been fully +decoded. This enables the application to start playback of the sound quickly, while at the same +time allowing to resource manager to keep loading in the background. Since there may be less +threads than the number of sounds being loaded at a given time, a simple scheduling system is used +to keep decoding time balanced and fair. The resource manager solves this by splitting decoding +into chunks called pages. By default, each page is 1 second long. When a page has been decoded, a +new job will be posted to start decoding the next page. By dividing up decoding into pages, an +individual sound shouldn't ever delay every other sound from having their first page decoded. Of +course, when loading many sounds at the same time, there will always be an amount of time required +to process jobs in the queue so in heavy load situations there will still be some delay. To +determine if a data source is ready to have some frames read, use +`ma_resource_manager_data_source_get_available_frames()`. This will return the number of frames +available starting from the current position. + + +6.2.1. Job Queue +---------------- +The resource manager uses a job queue which is multi-producer, multi-consumer, and fixed-capacity. +This job queue is not currently lock-free, and instead uses a spinlock to achieve thread-safety. +Only a fixed number of jobs can be allocated and inserted into the queue which is done through a +lock-free data structure for allocating an index into a fixed sized array, with reference counting +for mitigation of the ABA problem. The reference count is 32-bit. + +For many types of jobs it's important that they execute in a specific order. In these cases, jobs +are executed serially. For the resource manager, serial execution of jobs is only required on a +per-object basis (per data buffer or per data stream). Each of these objects stores an execution +counter. When a job is posted it is associated with an execution counter. When the job is +processed, it checks if the execution counter of the job equals the execution counter of the +owning object and if so, processes the job. If the counters are not equal, the job will be posted +back onto the job queue for later processing. When the job finishes processing the execution order +of the main object is incremented. This system means the no matter how many job threads are +executing, decoding of an individual sound will always get processed serially. The advantage to +having multiple threads comes into play when loading multiple sounds at the same time. + +The resource manager's job queue is not 100% lock-free and will use a spinlock to achieve +thread-safety for a very small section of code. This is only relevant when the resource manager +uses more than one job thread. If only using a single job thread, which is the default, the +lock should never actually wait in practice. The amount of time spent locking should be quite +short, but it's something to be aware of for those who have pedantic lock-free requirements and +need to use more than one job thread. There are plans to remove this lock in a future version. + +In addition, posting a job will release a semaphore, which on Win32 is implemented with +`ReleaseSemaphore` and on POSIX platforms via a condition variable: + + ```c + pthread_mutex_lock(&pSemaphore->lock); + { + pSemaphore->value += 1; + pthread_cond_signal(&pSemaphore->cond); + } + pthread_mutex_unlock(&pSemaphore->lock); + ``` + +Again, this is relevant for those with strict lock-free requirements in the audio thread. To avoid +this, you can use non-blocking mode (via the `MA_JOB_QUEUE_FLAG_NON_BLOCKING` +flag) and implement your own job processing routine (see the "Resource Manager" section above for +details on how to do this). + + + +6.2.2. Data Buffers +------------------- +When the `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM` flag is excluded at initialization time, the +resource manager will try to load the data into an in-memory data buffer. Before doing so, however, +it will first check if the specified file is already loaded. If so, it will increment a reference +counter and just use the already loaded data. This saves both time and memory. When the data buffer +is uninitialized, the reference counter will be decremented. If the counter hits zero, the file +will be unloaded. This is a detail to keep in mind because it could result in excessive loading and +unloading of a sound. For example, the following sequence will result in a file be loaded twice, +once after the other: + + ```c + ma_resource_manager_data_source_init(pResourceManager, "my_file", ..., &myDataBuffer0); // Refcount = 1. Initial load. + ma_resource_manager_data_source_uninit(&myDataBuffer0); // Refcount = 0. Unloaded. + + ma_resource_manager_data_source_init(pResourceManager, "my_file", ..., &myDataBuffer1); // Refcount = 1. Reloaded because previous uninit() unloaded it. + ma_resource_manager_data_source_uninit(&myDataBuffer1); // Refcount = 0. Unloaded. + ``` + +A binary search tree (BST) is used for storing data buffers as it has good balance between +efficiency and simplicity. The key of the BST is a 64-bit hash of the file path that was passed +into `ma_resource_manager_data_source_init()`. The advantage of using a hash is that it saves +memory over storing the entire path, has faster comparisons, and results in a mostly balanced BST +due to the random nature of the hash. The disadvantages are that file names are case-sensitive and +there's a small chance of name collisions. If case-sensitivity is an issue, you should normalize +your file names to upper- or lower-case before initializing your data sources. If name collisions +become an issue, you'll need to change the name of one of the colliding names or just not use the +resource manager. + +When a sound file has not already been loaded and the `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC` +flag is excluded, the file will be decoded synchronously by the calling thread. There are two +options for controlling how the audio is stored in the data buffer - encoded or decoded. When the +`MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_DECODE` option is excluded, the raw file data will be stored +in memory. Otherwise the sound will be decoded before storing it in memory. Synchronous loading is +a very simple and standard process of simply adding an item to the BST, allocating a block of +memory and then decoding (if `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_DECODE` is specified). + +When the `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC` flag is specified, loading of the data buffer +is done asynchronously. In this case, a job is posted to the queue to start loading and then the +function immediately returns, setting an internal result code to `MA_BUSY`. This result code is +returned when the program calls `ma_resource_manager_data_source_result()`. When decoding has fully +completed `MA_SUCCESS` will be returned. This can be used to know if loading has fully completed. + +When loading asynchronously, a single job is posted to the queue of the type +`MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER_NODE`. This involves making a copy of the file path and +associating it with job. When the job is processed by the job thread, it will first load the file +using the VFS associated with the resource manager. When using a custom VFS, it's important that it +be completely thread-safe because it will be used from one or more job threads at the same time. +Individual files should only ever be accessed by one thread at a time, however. After opening the +file via the VFS, the job will determine whether or not the file is being decoded. If not, it +simply allocates a block of memory and loads the raw file contents into it and returns. On the +other hand, when the file is being decoded, it will first allocate a decoder on the heap and +initialize it. Then it will check if the length of the file is known. If so it will allocate a +block of memory to store the decoded output and initialize it to silence. If the size is unknown, +it will allocate room for one page. After memory has been allocated, the first page will be +decoded. If the sound is shorter than a page, the result code will be set to `MA_SUCCESS` and the +completion event will be signalled and loading is now complete. If, however, there is more to +decode, a job with the code `MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_BUFFER_NODE` is posted. This job +will decode the next page and perform the same process if it reaches the end. If there is more to +decode, the job will post another `MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_BUFFER_NODE` job which will +keep on happening until the sound has been fully decoded. For sounds of an unknown length, each +page will be linked together as a linked list. Internally this is implemented via the +`ma_paged_audio_buffer` object. + + +6.2.3. Data Streams +------------------- +Data streams only ever store two pages worth of data for each instance. They are most useful for +large sounds like music tracks in games that would consume too much memory if fully decoded in +memory. After every frame from a page has been read, a job will be posted to load the next page +which is done from the VFS. + +For data streams, the `MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC` flag will determine whether or +not initialization of the data source waits until the two pages have been decoded. When unset, +`ma_resource_manager_data_source_init()` will wait until the two pages have been loaded, otherwise +it will return immediately. + +When frames are read from a data stream using `ma_resource_manager_data_source_read_pcm_frames()`, +`MA_BUSY` will be returned if there are no frames available. If there are some frames available, +but less than the number requested, `MA_SUCCESS` will be returned, but the actual number of frames +read will be less than the number requested. Due to the asynchronous nature of data streams, +seeking is also asynchronous. If the data stream is in the middle of a seek, `MA_BUSY` will be +returned when trying to read frames. + +When `ma_resource_manager_data_source_read_pcm_frames()` results in a page getting fully consumed +a job is posted to load the next page. This will be posted from the same thread that called +`ma_resource_manager_data_source_read_pcm_frames()`. + +Data streams are uninitialized by posting a job to the queue, but the function won't return until +that job has been processed. The reason for this is that the caller owns the data stream object and +therefore miniaudio needs to ensure everything completes before handing back control to the caller. +Also, if the data stream is uninitialized while pages are in the middle of decoding, they must +complete before destroying any underlying object and the job system handles this cleanly. + +Note that when a new page needs to be loaded, a job will be posted to the resource manager's job +thread from the audio thread. You must keep in mind the details mentioned in the "Job Queue" +section above regarding locking when posting an event if you require a strictly lock-free audio +thread. + + + +7. Node Graph +============= +miniaudio's routing infrastructure follows a node graph paradigm. The idea is that you create a +node whose outputs are attached to inputs of another node, thereby creating a graph. There are +different types of nodes, with each node in the graph processing input data to produce output, +which is then fed through the chain. Each node in the graph can apply their own custom effects. At +the start of the graph will usually be one or more data source nodes which have no inputs and +instead pull their data from a data source. At the end of the graph is an endpoint which represents +the end of the chain and is where the final output is ultimately extracted from. + +Each node has a number of input buses and a number of output buses. An output bus from a node is +attached to an input bus of another. Multiple nodes can connect their output buses to another +node's input bus, in which case their outputs will be mixed before processing by the node. Below is +a diagram that illustrates a hypothetical node graph setup: + + ``` + >>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Data flows left to right >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> + + +---------------+ +-----------------+ + | Data Source 1 =----+ +----------+ +----= Low Pass Filter =----+ + +---------------+ | | =----+ +-----------------+ | +----------+ + +----= Splitter | +----= ENDPOINT | + +---------------+ | | =----+ +-----------------+ | +----------+ + | Data Source 2 =----+ +----------+ +----= Echo / Delay =----+ + +---------------+ +-----------------+ + ``` + +In the above graph, it starts with two data sources whose outputs are attached to the input of a +splitter node. It's at this point that the two data sources are mixed. After mixing, the splitter +performs it's processing routine and produces two outputs which is simply a duplication of the +input stream. One output is attached to a low pass filter, whereas the other output is attached to +a echo/delay. The outputs of the low pass filter and the echo are attached to the endpoint, and +since they're both connected to the same input bus, they'll be mixed. + +Each input bus must be configured to accept the same number of channels, but the number of channels +used by input buses can be different to the number of channels for output buses in which case +miniaudio will automatically convert the input data to the output channel count before processing. +The number of channels of an output bus of one node must match the channel count of the input bus +it's attached to. The channel counts cannot be changed after the node has been initialized. If you +attempt to attach an output bus to an input bus with a different channel count, attachment will +fail. + +To use a node graph, you first need to initialize a `ma_node_graph` object. This is essentially a +container around the entire graph. The `ma_node_graph` object is required for some thread-safety +issues which will be explained later. A `ma_node_graph` object is initialized using miniaudio's +standard config/init system: + + ```c + ma_node_graph_config nodeGraphConfig = ma_node_graph_config_init(myChannelCount); + + result = ma_node_graph_init(&nodeGraphConfig, NULL, &nodeGraph); // Second parameter is a pointer to allocation callbacks. + if (result != MA_SUCCESS) { + // Failed to initialize node graph. + } + ``` + +When you initialize the node graph, you're specifying the channel count of the endpoint. The +endpoint is a special node which has one input bus and one output bus, both of which have the +same channel count, which is specified in the config. Any nodes that connect directly to the +endpoint must be configured such that their output buses have the same channel count. When you read +audio data from the node graph, it'll have the channel count you specified in the config. To read +data from the graph: + + ```c + ma_uint32 framesRead; + result = ma_node_graph_read_pcm_frames(&nodeGraph, pFramesOut, frameCount, &framesRead); + if (result != MA_SUCCESS) { + // Failed to read data from the node graph. + } + ``` + +When you read audio data, miniaudio starts at the node graph's endpoint node which then pulls in +data from its input attachments, which in turn recursively pull in data from their inputs, and so +on. At the start of the graph there will be some kind of data source node which will have zero +inputs and will instead read directly from a data source. The base nodes don't literally need to +read from a `ma_data_source` object, but they will always have some kind of underlying object that +sources some kind of audio. The `ma_data_source_node` node can be used to read from a +`ma_data_source`. Data is always in floating-point format and in the number of channels you +specified when the graph was initialized. The sample rate is defined by the underlying data sources. +It's up to you to ensure they use a consistent and appropriate sample rate. + +The `ma_node` API is designed to allow custom nodes to be implemented with relative ease, but +miniaudio includes a few stock nodes for common functionality. This is how you would initialize a +node which reads directly from a data source (`ma_data_source_node`) which is an example of one +of the stock nodes that comes with miniaudio: + + ```c + ma_data_source_node_config config = ma_data_source_node_config_init(pMyDataSource); + + ma_data_source_node dataSourceNode; + result = ma_data_source_node_init(&nodeGraph, &config, NULL, &dataSourceNode); + if (result != MA_SUCCESS) { + // Failed to create data source node. + } + ``` + +The data source node will use the output channel count to determine the channel count of the output +bus. There will be 1 output bus and 0 input buses (data will be drawn directly from the data +source). The data source must output to floating-point (`ma_format_f32`) or else an error will be +returned from `ma_data_source_node_init()`. + +By default the node will not be attached to the graph. To do so, use `ma_node_attach_output_bus()`: + + ```c + result = ma_node_attach_output_bus(&dataSourceNode, 0, ma_node_graph_get_endpoint(&nodeGraph), 0); + if (result != MA_SUCCESS) { + // Failed to attach node. + } + ``` + +The code above connects the data source node directly to the endpoint. Since the data source node +has only a single output bus, the index will always be 0. Likewise, the endpoint only has a single +input bus which means the input bus index will also always be 0. + +To detach a specific output bus, use `ma_node_detach_output_bus()`. To detach all output buses, use +`ma_node_detach_all_output_buses()`. If you want to just move the output bus from one attachment to +another, you do not need to detach first. You can just call `ma_node_attach_output_bus()` and it'll +deal with it for you. + +Less frequently you may want to create a specialized node. This will be a node where you implement +your own processing callback to apply a custom effect of some kind. This is similar to initializing +one of the stock node types, only this time you need to specify a pointer to a vtable containing a +pointer to the processing function and the number of input and output buses. Example: + + ```c + static void my_custom_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) + { + // Do some processing of ppFramesIn (one stream of audio data per input bus) + const float* pFramesIn_0 = ppFramesIn[0]; // Input bus @ index 0. + const float* pFramesIn_1 = ppFramesIn[1]; // Input bus @ index 1. + float* pFramesOut_0 = ppFramesOut[0]; // Output bus @ index 0. + + // Do some processing. On input, `pFrameCountIn` will be the number of input frames in each + // buffer in `ppFramesIn` and `pFrameCountOut` will be the capacity of each of the buffers + // in `ppFramesOut`. On output, `pFrameCountIn` should be set to the number of input frames + // your node consumed and `pFrameCountOut` should be set the number of output frames that + // were produced. + // + // You should process as many frames as you can. If your effect consumes input frames at the + // same rate as output frames (always the case, unless you're doing resampling), you need + // only look at `ppFramesOut` and process that exact number of frames. If you're doing + // resampling, you'll need to be sure to set both `pFrameCountIn` and `pFrameCountOut` + // properly. + } + + static ma_node_vtable my_custom_node_vtable = + { + my_custom_node_process_pcm_frames, // The function that will be called to process your custom node. This is where you'd implement your effect processing. + NULL, // Optional. A callback for calculating the number of input frames that are required to process a specified number of output frames. + 2, // 2 input buses. + 1, // 1 output bus. + 0 // Default flags. + }; + + ... + + // Each bus needs to have a channel count specified. To do this you need to specify the channel + // counts in an array and then pass that into the node config. + ma_uint32 inputChannels[2]; // Equal in size to the number of input channels specified in the vtable. + ma_uint32 outputChannels[1]; // Equal in size to the number of output channels specified in the vtable. + + inputChannels[0] = channelsIn; + inputChannels[1] = channelsIn; + outputChannels[0] = channelsOut; + + ma_node_config nodeConfig = ma_node_config_init(); + nodeConfig.vtable = &my_custom_node_vtable; + nodeConfig.pInputChannels = inputChannels; + nodeConfig.pOutputChannels = outputChannels; + + ma_node_base node; + result = ma_node_init(&nodeGraph, &nodeConfig, NULL, &node); + if (result != MA_SUCCESS) { + // Failed to initialize node. + } + ``` + +When initializing a custom node, as in the code above, you'll normally just place your vtable in +static space. The number of input and output buses are specified as part of the vtable. If you need +a variable number of buses on a per-node bases, the vtable should have the relevant bus count set +to `MA_NODE_BUS_COUNT_UNKNOWN`. In this case, the bus count should be set in the node config: + + ```c + static ma_node_vtable my_custom_node_vtable = + { + my_custom_node_process_pcm_frames, // The function that will be called process your custom node. This is where you'd implement your effect processing. + NULL, // Optional. A callback for calculating the number of input frames that are required to process a specified number of output frames. + MA_NODE_BUS_COUNT_UNKNOWN, // The number of input buses is determined on a per-node basis. + 1, // 1 output bus. + 0 // Default flags. + }; + + ... + + ma_node_config nodeConfig = ma_node_config_init(); + nodeConfig.vtable = &my_custom_node_vtable; + nodeConfig.inputBusCount = myBusCount; // <-- Since the vtable specifies MA_NODE_BUS_COUNT_UNKNOWN, the input bus count should be set here. + nodeConfig.pInputChannels = inputChannels; // <-- Make sure there are nodeConfig.inputBusCount elements in this array. + nodeConfig.pOutputChannels = outputChannels; // <-- The vtable specifies 1 output bus, so there must be 1 element in this array. + ``` + +In the above example it's important to never set the `inputBusCount` and `outputBusCount` members +to anything other than their defaults if the vtable specifies an explicit count. They can only be +set if the vtable specifies MA_NODE_BUS_COUNT_UNKNOWN in the relevant bus count. + +Most often you'll want to create a structure to encapsulate your node with some extra data. You +need to make sure the `ma_node_base` object is your first member of the structure: + + ```c + typedef struct + { + ma_node_base base; // <-- Make sure this is always the first member. + float someCustomData; + } my_custom_node; + ``` + +By doing this, your object will be compatible with all `ma_node` APIs and you can attach it to the +graph just like any other node. + +In the custom processing callback (`my_custom_node_process_pcm_frames()` in the example above), the +number of channels for each bus is what was specified by the config when the node was initialized +with `ma_node_init()`. In addition, all attachments to each of the input buses will have been +pre-mixed by miniaudio. The config allows you to specify different channel counts for each +individual input and output bus. It's up to the effect to handle it appropriate, and if it can't, +return an error in it's initialization routine. + +Custom nodes can be assigned some flags to describe their behaviour. These are set via the vtable +and include the following: + + +-----------------------------------------+---------------------------------------------------+ + | Flag Name | Description | + +-----------------------------------------+---------------------------------------------------+ + | MA_NODE_FLAG_PASSTHROUGH | Useful for nodes that do not do any kind of audio | + | | processing, but are instead used for tracking | + | | time, handling events, etc. Also used by the | + | | internal endpoint node. It reads directly from | + | | the input bus to the output bus. Nodes with this | + | | flag must have exactly 1 input bus and 1 output | + | | bus, and both buses must have the same channel | + | | counts. | + +-----------------------------------------+---------------------------------------------------+ + | MA_NODE_FLAG_CONTINUOUS_PROCESSING | Causes the processing callback to be called even | + | | when no data is available to be read from input | + | | attachments. When a node has at least one input | + | | bus, but there are no inputs attached or the | + | | inputs do not deliver any data, the node's | + | | processing callback will not get fired. This flag | + | | will make it so the callback is always fired | + | | regardless of whether or not any input data is | + | | received. This is useful for effects like | + | | echos where there will be a tail of audio data | + | | that still needs to be processed even when the | + | | original data sources have reached their ends. It | + | | may also be useful for nodes that must always | + | | have their processing callback fired when there | + | | are no inputs attached. | + +-----------------------------------------+---------------------------------------------------+ + | MA_NODE_FLAG_ALLOW_NULL_INPUT | Used in conjunction with | + | | `MA_NODE_FLAG_CONTINUOUS_PROCESSING`. When this | + | | is set, the `ppFramesIn` parameter of the | + | | processing callback will be set to NULL when | + | | there are no input frames are available. When | + | | this is unset, silence will be posted to the | + | | processing callback. | + +-----------------------------------------+---------------------------------------------------+ + | MA_NODE_FLAG_DIFFERENT_PROCESSING_RATES | Used to tell miniaudio that input and output | + | | frames are processed at different rates. You | + | | should set this for any nodes that perform | + | | resampling. | + +-----------------------------------------+---------------------------------------------------+ + | MA_NODE_FLAG_SILENT_OUTPUT | Used to tell miniaudio that a node produces only | + | | silent output. This is useful for nodes where you | + | | don't want the output to contribute to the final | + | | mix. An example might be if you want split your | + | | stream and have one branch be output to a file. | + | | When using this flag, you should avoid writing to | + | | the output buffer of the node's processing | + | | callback because miniaudio will ignore it anyway. | + +-----------------------------------------+---------------------------------------------------+ + + +If you need to make a copy of an audio stream for effect processing you can use a splitter node +called `ma_splitter_node`. This takes has 1 input bus and splits the stream into 2 output buses. +You can use it like this: + + ```c + ma_splitter_node_config splitterNodeConfig = ma_splitter_node_config_init(channels); + + ma_splitter_node splitterNode; + result = ma_splitter_node_init(&nodeGraph, &splitterNodeConfig, NULL, &splitterNode); + if (result != MA_SUCCESS) { + // Failed to create node. + } + + // Attach your output buses to two different input buses (can be on two different nodes). + ma_node_attach_output_bus(&splitterNode, 0, ma_node_graph_get_endpoint(&nodeGraph), 0); // Attach directly to the endpoint. + ma_node_attach_output_bus(&splitterNode, 1, &myEffectNode, 0); // Attach to input bus 0 of some effect node. + ``` + +The volume of an output bus can be configured on a per-bus basis: + + ```c + ma_node_set_output_bus_volume(&splitterNode, 0, 0.5f); + ma_node_set_output_bus_volume(&splitterNode, 1, 0.5f); + ``` + +In the code above we're using the splitter node from before and changing the volume of each of the +copied streams. + +You can start and stop a node with the following: + + ```c + ma_node_set_state(&splitterNode, ma_node_state_started); // The default state. + ma_node_set_state(&splitterNode, ma_node_state_stopped); + ``` + +By default the node is in a started state, but since it won't be connected to anything won't +actually be invoked by the node graph until it's connected. When you stop a node, data will not be +read from any of its input connections. You can use this property to stop a group of sounds +atomically. + +You can configure the initial state of a node in it's config: + + ```c + nodeConfig.initialState = ma_node_state_stopped; + ``` + +Note that for the stock specialized nodes, all of their configs will have a `nodeConfig` member +which is the config to use with the base node. This is where the initial state can be configured +for specialized nodes: + + ```c + dataSourceNodeConfig.nodeConfig.initialState = ma_node_state_stopped; + ``` + +When using a specialized node like `ma_data_source_node` or `ma_splitter_node`, be sure to not +modify the `vtable` member of the `nodeConfig` object. + + +7.1. Timing +----------- +The node graph supports starting and stopping nodes at scheduled times. This is especially useful +for data source nodes where you want to get the node set up, but only start playback at a specific +time. There are two clocks: local and global. + +A local clock is per-node, whereas the global clock is per graph. Scheduling starts and stops can +only be done based on the global clock because the local clock will not be running while the node +is stopped. The global clocks advances whenever `ma_node_graph_read_pcm_frames()` is called. On the +other hand, the local clock only advances when the node's processing callback is fired, and is +advanced based on the output frame count. + +To retrieve the global time, use `ma_node_graph_get_time()`. The global time can be set with +`ma_node_graph_set_time()` which might be useful if you want to do seeking on a global timeline. +Getting and setting the local time is similar. Use `ma_node_get_time()` to retrieve the local time, +and `ma_node_set_time()` to set the local time. The global and local times will be advanced by the +audio thread, so care should be taken to avoid data races. Ideally you should avoid calling these +outside of the node processing callbacks which are always run on the audio thread. + +There is basic support for scheduling the starting and stopping of nodes. You can only schedule one +start and one stop at a time. This is mainly intended for putting nodes into a started or stopped +state in a frame-exact manner. Without this mechanism, starting and stopping of a node is limited +to the resolution of a call to `ma_node_graph_read_pcm_frames()` which would typically be in blocks +of several milliseconds. The following APIs can be used for scheduling node states: + + ```c + ma_node_set_state_time() + ma_node_get_state_time() + ``` + +The time is absolute and must be based on the global clock. An example is below: + + ```c + ma_node_set_state_time(&myNode, ma_node_state_started, sampleRate*1); // Delay starting to 1 second. + ma_node_set_state_time(&myNode, ma_node_state_stopped, sampleRate*5); // Delay stopping to 5 seconds. + ``` + +An example for changing the state using a relative time. + + ```c + ma_node_set_state_time(&myNode, ma_node_state_started, sampleRate*1 + ma_node_graph_get_time(&myNodeGraph)); + ma_node_set_state_time(&myNode, ma_node_state_stopped, sampleRate*5 + ma_node_graph_get_time(&myNodeGraph)); + ``` + +Note that due to the nature of multi-threading the times may not be 100% exact. If this is an +issue, consider scheduling state changes from within a processing callback. An idea might be to +have some kind of passthrough trigger node that is used specifically for tracking time and handling +events. + + + +7.2. Thread Safety and Locking +------------------------------ +When processing audio, it's ideal not to have any kind of locking in the audio thread. Since it's +expected that `ma_node_graph_read_pcm_frames()` would be run on the audio thread, it does so +without the use of any locks. This section discusses the implementation used by miniaudio and goes +over some of the compromises employed by miniaudio to achieve this goal. Note that the current +implementation may not be ideal - feedback and critiques are most welcome. + +The node graph API is not *entirely* lock-free. Only `ma_node_graph_read_pcm_frames()` is expected +to be lock-free. Attachment, detachment and uninitialization of nodes use locks to simplify the +implementation, but are crafted in a way such that such locking is not required when reading audio +data from the graph. Locking in these areas are achieved by means of spinlocks. + +The main complication with keeping `ma_node_graph_read_pcm_frames()` lock-free stems from the fact +that a node can be uninitialized, and it's memory potentially freed, while in the middle of being +processed on the audio thread. There are times when the audio thread will be referencing a node, +which means the uninitialization process of a node needs to make sure it delays returning until the +audio thread is finished so that control is not handed back to the caller thereby giving them a +chance to free the node's memory. + +When the audio thread is processing a node, it does so by reading from each of the output buses of +the node. In order for a node to process data for one of its output buses, it needs to read from +each of its input buses, and so on an so forth. It follows that once all output buses of a node +are detached, the node as a whole will be disconnected and no further processing will occur unless +it's output buses are reattached, which won't be happening when the node is being uninitialized. +By having `ma_node_detach_output_bus()` wait until the audio thread is finished with it, we can +simplify a few things, at the expense of making `ma_node_detach_output_bus()` a bit slower. By +doing this, the implementation of `ma_node_uninit()` becomes trivial - just detach all output +nodes, followed by each of the attachments to each of its input nodes, and then do any final clean +up. + +With the above design, the worst-case scenario is `ma_node_detach_output_bus()` taking as long as +it takes to process the output bus being detached. This will happen if it's called at just the +wrong moment where the audio thread has just iterated it and has just started processing. The +caller of `ma_node_detach_output_bus()` will stall until the audio thread is finished, which +includes the cost of recursively processing its inputs. This is the biggest compromise made with +the approach taken by miniaudio for its lock-free processing system. The cost of detaching nodes +earlier in the pipeline (data sources, for example) will be cheaper than the cost of detaching +higher level nodes, such as some kind of final post-processing endpoint. If you need to do mass +detachments, detach starting from the lowest level nodes and work your way towards the final +endpoint node (but don't try detaching the node graph's endpoint). If the audio thread is not +running, detachment will be fast and detachment in any order will be the same. The reason nodes +need to wait for their input attachments to complete is due to the potential for desyncs between +data sources. If the node was to terminate processing mid way through processing its inputs, +there's a chance that some of the underlying data sources will have been read, but then others not. +That will then result in a potential desynchronization when detaching and reattaching higher-level +nodes. A possible solution to this is to have an option when detaching to terminate processing +before processing all input attachments which should be fairly simple. + +Another compromise, albeit less significant, is locking when attaching and detaching nodes. This +locking is achieved by means of a spinlock in order to reduce memory overhead. A lock is present +for each input bus and output bus. When an output bus is connected to an input bus, both the output +bus and input bus is locked. This locking is specifically for attaching and detaching across +different threads and does not affect `ma_node_graph_read_pcm_frames()` in any way. The locking and +unlocking is mostly self-explanatory, but a slightly less intuitive aspect comes into it when +considering that iterating over attachments must not break as a result of attaching or detaching a +node while iteration is occurring. + +Attaching and detaching are both quite simple. When an output bus of a node is attached to an input +bus of another node, it's added to a linked list. Basically, an input bus is a linked list, where +each item in the list is and output bus. We have some intentional (and convenient) restrictions on +what can done with the linked list in order to simplify the implementation. First of all, whenever +something needs to iterate over the list, it must do so in a forward direction. Backwards iteration +is not supported. Also, items can only be added to the start of the list. + +The linked list is a doubly-linked list where each item in the list (an output bus) holds a pointer +to the next item in the list, and another to the previous item. A pointer to the previous item is +only required for fast detachment of the node - it is never used in iteration. This is an +important property because it means from the perspective of iteration, attaching and detaching of +an item can be done with a single atomic assignment. This is exploited by both the attachment and +detachment process. When attaching the node, the first thing that is done is the setting of the +local "next" and "previous" pointers of the node. After that, the item is "attached" to the list +by simply performing an atomic exchange with the head pointer. After that, the node is "attached" +to the list from the perspective of iteration. Even though the "previous" pointer of the next item +hasn't yet been set, from the perspective of iteration it's been attached because iteration will +only be happening in a forward direction which means the "previous" pointer won't actually ever get +used. The same general process applies to detachment. See `ma_node_attach_output_bus()` and +`ma_node_detach_output_bus()` for the implementation of this mechanism. + + + +8. Decoding +=========== +The `ma_decoder` API is used for reading audio files. Decoders are completely decoupled from +devices and can be used independently. Built-in support is included for the following formats: + + +---------+ + | Format | + +---------+ + | WAV | + | MP3 | + | FLAC | + +---------+ + +You can disable the built-in decoders by specifying one or more of the following options before the +miniaudio implementation: + + ```c + #define MA_NO_WAV + #define MA_NO_MP3 + #define MA_NO_FLAC + ``` + +miniaudio supports the ability to plug in custom decoders. See the section below for details on how +to use custom decoders. + +A decoder can be initialized from a file with `ma_decoder_init_file()`, a block of memory with +`ma_decoder_init_memory()`, or from data delivered via callbacks with `ma_decoder_init()`. Here is +an example for loading a decoder from a file: + + ```c + ma_decoder decoder; + ma_result result = ma_decoder_init_file("MySong.mp3", NULL, &decoder); + if (result != MA_SUCCESS) { + return false; // An error occurred. + } + + ... + + ma_decoder_uninit(&decoder); + ``` + +When initializing a decoder, you can optionally pass in a pointer to a `ma_decoder_config` object +(the `NULL` argument in the example above) which allows you to configure the output format, channel +count, sample rate and channel map: + + ```c + ma_decoder_config config = ma_decoder_config_init(ma_format_f32, 2, 48000); + ``` + +When passing in `NULL` for decoder config in `ma_decoder_init*()`, the output format will be the +same as that defined by the decoding backend. + +Data is read from the decoder as PCM frames. This will output the number of PCM frames actually +read. If this is less than the requested number of PCM frames it means you've reached the end. The +return value will be `MA_AT_END` if no samples have been read and the end has been reached. + + ```c + ma_result result = ma_decoder_read_pcm_frames(pDecoder, pFrames, framesToRead, &framesRead); + if (framesRead < framesToRead) { + // Reached the end. + } + ``` + +You can also seek to a specific frame like so: + + ```c + ma_result result = ma_decoder_seek_to_pcm_frame(pDecoder, targetFrame); + if (result != MA_SUCCESS) { + return false; // An error occurred. + } + ``` + +If you want to loop back to the start, you can simply seek back to the first PCM frame: + + ```c + ma_decoder_seek_to_pcm_frame(pDecoder, 0); + ``` + +When loading a decoder, miniaudio uses a trial and error technique to find the appropriate decoding +backend. This can be unnecessarily inefficient if the type is already known. In this case you can +use `encodingFormat` variable in the device config to specify a specific encoding format you want +to decode: + + ```c + decoderConfig.encodingFormat = ma_encoding_format_wav; + ``` + +See the `ma_encoding_format` enum for possible encoding formats. + +The `ma_decoder_init_file()` API will try using the file extension to determine which decoding +backend to prefer. + + +8.1. Custom Decoders +-------------------- +It's possible to implement a custom decoder and plug it into miniaudio. This is extremely useful +when you want to use the `ma_decoder` API, but need to support an encoding format that's not one of +the stock formats supported by miniaudio. This can be put to particularly good use when using the +`ma_engine` and/or `ma_resource_manager` APIs because they use `ma_decoder` internally. If, for +example, you wanted to support Opus, you can do so with a custom decoder (there if a reference +Opus decoder in the "extras" folder of the miniaudio repository which uses libopus + libopusfile). + +A custom decoder must implement a data source. A vtable called `ma_decoding_backend_vtable` needs +to be implemented which is then passed into the decoder config: + + ```c + ma_decoding_backend_vtable* pCustomBackendVTables[] = + { + &g_ma_decoding_backend_vtable_libvorbis, + &g_ma_decoding_backend_vtable_libopus + }; + + ... + + decoderConfig = ma_decoder_config_init_default(); + decoderConfig.pCustomBackendUserData = NULL; + decoderConfig.ppCustomBackendVTables = pCustomBackendVTables; + decoderConfig.customBackendCount = sizeof(pCustomBackendVTables) / sizeof(pCustomBackendVTables[0]); + ``` + +The `ma_decoding_backend_vtable` vtable has the following functions: + + ``` + onInit + onInitFile + onInitFileW + onInitMemory + onUninit + ``` + +There are only two functions that must be implemented - `onInit` and `onUninit`. The other +functions can be implemented for a small optimization for loading from a file path or memory. If +these are not specified, miniaudio will deal with it for you via a generic implementation. + +When you initialize a custom data source (by implementing the `onInit` function in the vtable) you +will need to output a pointer to a `ma_data_source` which implements your custom decoder. See the +section about data sources for details on how to implement this. Alternatively, see the +"custom_decoders" example in the miniaudio repository. + +The `onInit` function takes a pointer to some callbacks for the purpose of reading raw audio data +from some arbitrary source. You'll use these functions to read from the raw data and perform the +decoding. When you call them, you will pass in the `pReadSeekTellUserData` pointer to the relevant +parameter. + +The `pConfig` parameter in `onInit` can be used to configure the backend if appropriate. It's only +used as a hint and can be ignored. However, if any of the properties are relevant to your decoder, +an optimal implementation will handle the relevant properties appropriately. + +If memory allocation is required, it should be done so via the specified allocation callbacks if +possible (the `pAllocationCallbacks` parameter). + +If an error occurs when initializing the decoder, you should leave `ppBackend` unset, or set to +NULL, and make sure everything is cleaned up appropriately and an appropriate result code returned. +When multiple custom backends are specified, miniaudio will cycle through the vtables in the order +they're listed in the array that's passed into the decoder config so it's important that your +initialization routine is clean. + +When a decoder is uninitialized, the `onUninit` callback will be fired which will give you an +opportunity to clean up and internal data. + + + +9. Encoding +=========== +The `ma_encoding` API is used for writing audio files. The only supported output format is WAV. +This can be disabled by specifying the following option before the implementation of miniaudio: + + ```c + #define MA_NO_WAV + ``` + +An encoder can be initialized to write to a file with `ma_encoder_init_file()` or from data +delivered via callbacks with `ma_encoder_init()`. Below is an example for initializing an encoder +to output to a file. + + ```c + ma_encoder_config config = ma_encoder_config_init(ma_encoding_format_wav, FORMAT, CHANNELS, SAMPLE_RATE); + ma_encoder encoder; + ma_result result = ma_encoder_init_file("my_file.wav", &config, &encoder); + if (result != MA_SUCCESS) { + // Error + } + + ... + + ma_encoder_uninit(&encoder); + ``` + +When initializing an encoder you must specify a config which is initialized with +`ma_encoder_config_init()`. Here you must specify the file type, the output sample format, output +channel count and output sample rate. The following file types are supported: + + +------------------------+-------------+ + | Enum | Description | + +------------------------+-------------+ + | ma_encoding_format_wav | WAV | + +------------------------+-------------+ + +If the format, channel count or sample rate is not supported by the output file type an error will +be returned. The encoder will not perform data conversion so you will need to convert it before +outputting any audio data. To output audio data, use `ma_encoder_write_pcm_frames()`, like in the +example below: + + ```c + ma_uint64 framesWritten; + result = ma_encoder_write_pcm_frames(&encoder, pPCMFramesToWrite, framesToWrite, &framesWritten); + if (result != MA_SUCCESS) { + ... handle error ... + } + ``` + +The `framesWritten` variable will contain the number of PCM frames that were actually written. This +is optionally and you can pass in `NULL` if you need this. + +Encoders must be uninitialized with `ma_encoder_uninit()`. + + + +10. Data Conversion +=================== +A data conversion API is included with miniaudio which supports the majority of data conversion +requirements. This supports conversion between sample formats, channel counts (with channel +mapping) and sample rates. + + +10.1. Sample Format Conversion +------------------------------ +Conversion between sample formats is achieved with the `ma_pcm_*_to_*()`, `ma_pcm_convert()` and +`ma_convert_pcm_frames_format()` APIs. Use `ma_pcm_*_to_*()` to convert between two specific +formats. Use `ma_pcm_convert()` to convert based on a `ma_format` variable. Use +`ma_convert_pcm_frames_format()` to convert PCM frames where you want to specify the frame count +and channel count as a variable instead of the total sample count. + + +10.1.1. Dithering +----------------- +Dithering can be set using the ditherMode parameter. + +The different dithering modes include the following, in order of efficiency: + + +-----------+--------------------------+ + | Type | Enum Token | + +-----------+--------------------------+ + | None | ma_dither_mode_none | + | Rectangle | ma_dither_mode_rectangle | + | Triangle | ma_dither_mode_triangle | + +-----------+--------------------------+ + +Note that even if the dither mode is set to something other than `ma_dither_mode_none`, it will be +ignored for conversions where dithering is not needed. Dithering is available for the following +conversions: + + ``` + s16 -> u8 + s24 -> u8 + s32 -> u8 + f32 -> u8 + s24 -> s16 + s32 -> s16 + f32 -> s16 + ``` + +Note that it is not an error to pass something other than ma_dither_mode_none for conversions where +dither is not used. It will just be ignored. + + + +10.2. Channel Conversion +------------------------ +Channel conversion is used for channel rearrangement and conversion from one channel count to +another. The `ma_channel_converter` API is used for channel conversion. Below is an example of +initializing a simple channel converter which converts from mono to stereo. + + ```c + ma_channel_converter_config config = ma_channel_converter_config_init( + ma_format, // Sample format + 1, // Input channels + NULL, // Input channel map + 2, // Output channels + NULL, // Output channel map + ma_channel_mix_mode_default); // The mixing algorithm to use when combining channels. + + result = ma_channel_converter_init(&config, NULL, &converter); + if (result != MA_SUCCESS) { + // Error. + } + ``` + +To perform the conversion simply call `ma_channel_converter_process_pcm_frames()` like so: + + ```c + ma_result result = ma_channel_converter_process_pcm_frames(&converter, pFramesOut, pFramesIn, frameCount); + if (result != MA_SUCCESS) { + // Error. + } + ``` + +It is up to the caller to ensure the output buffer is large enough to accommodate the new PCM +frames. + +Input and output PCM frames are always interleaved. Deinterleaved layouts are not supported. + + +10.2.1. Channel Mapping +----------------------- +In addition to converting from one channel count to another, like the example above, the channel +converter can also be used to rearrange channels. When initializing the channel converter, you can +optionally pass in channel maps for both the input and output frames. If the channel counts are the +same, and each channel map contains the same channel positions with the exception that they're in +a different order, a simple shuffling of the channels will be performed. If, however, there is not +a 1:1 mapping of channel positions, or the channel counts differ, the input channels will be mixed +based on a mixing mode which is specified when initializing the `ma_channel_converter_config` +object. + +When converting from mono to multi-channel, the mono channel is simply copied to each output +channel. When going the other way around, the audio of each output channel is simply averaged and +copied to the mono channel. + +In more complicated cases blending is used. The `ma_channel_mix_mode_simple` mode will drop excess +channels and silence extra channels. For example, converting from 4 to 2 channels, the 3rd and 4th +channels will be dropped, whereas converting from 2 to 4 channels will put silence into the 3rd and +4th channels. + +The `ma_channel_mix_mode_rectangle` mode uses spacial locality based on a rectangle to compute a +simple distribution between input and output. Imagine sitting in the middle of a room, with +speakers on the walls representing channel positions. The `MA_CHANNEL_FRONT_LEFT` position can be +thought of as being in the corner of the front and left walls. + +Finally, the `ma_channel_mix_mode_custom_weights` mode can be used to use custom user-defined +weights. Custom weights can be passed in as the last parameter of +`ma_channel_converter_config_init()`. + +Predefined channel maps can be retrieved with `ma_channel_map_init_standard()`. This takes a +`ma_standard_channel_map` enum as its first parameter, which can be one of the following: + + +-----------------------------------+-----------------------------------------------------------+ + | Name | Description | + +-----------------------------------+-----------------------------------------------------------+ + | ma_standard_channel_map_default | Default channel map used by miniaudio. See below. | + | ma_standard_channel_map_microsoft | Channel map used by Microsoft's bitfield channel maps. | + | ma_standard_channel_map_alsa | Default ALSA channel map. | + | ma_standard_channel_map_rfc3551 | RFC 3551. Based on AIFF. | + | ma_standard_channel_map_flac | FLAC channel map. | + | ma_standard_channel_map_vorbis | Vorbis channel map. | + | ma_standard_channel_map_sound4 | FreeBSD's sound(4). | + | ma_standard_channel_map_sndio | sndio channel map. http://www.sndio.org/tips.html. | + | ma_standard_channel_map_webaudio | https://webaudio.github.io/web-audio-api/#ChannelOrdering | + +-----------------------------------+-----------------------------------------------------------+ + +Below are the channel maps used by default in miniaudio (`ma_standard_channel_map_default`): + + +---------------+---------------------------------+ + | Channel Count | Mapping | + +---------------+---------------------------------+ + | 1 (Mono) | 0: MA_CHANNEL_MONO | + +---------------+---------------------------------+ + | 2 (Stereo) | 0: MA_CHANNEL_FRONT_LEFT
| + | | 1: MA_CHANNEL_FRONT_RIGHT | + +---------------+---------------------------------+ + | 3 | 0: MA_CHANNEL_FRONT_LEFT
| + | | 1: MA_CHANNEL_FRONT_RIGHT
| + | | 2: MA_CHANNEL_FRONT_CENTER | + +---------------+---------------------------------+ + | 4 (Surround) | 0: MA_CHANNEL_FRONT_LEFT
| + | | 1: MA_CHANNEL_FRONT_RIGHT
| + | | 2: MA_CHANNEL_FRONT_CENTER
| + | | 3: MA_CHANNEL_BACK_CENTER | + +---------------+---------------------------------+ + | 5 | 0: MA_CHANNEL_FRONT_LEFT
| + | | 1: MA_CHANNEL_FRONT_RIGHT
| + | | 2: MA_CHANNEL_FRONT_CENTER
| + | | 3: MA_CHANNEL_BACK_LEFT
| + | | 4: MA_CHANNEL_BACK_RIGHT | + +---------------+---------------------------------+ + | 6 (5.1) | 0: MA_CHANNEL_FRONT_LEFT
| + | | 1: MA_CHANNEL_FRONT_RIGHT
| + | | 2: MA_CHANNEL_FRONT_CENTER
| + | | 3: MA_CHANNEL_LFE
| + | | 4: MA_CHANNEL_SIDE_LEFT
| + | | 5: MA_CHANNEL_SIDE_RIGHT | + +---------------+---------------------------------+ + | 7 | 0: MA_CHANNEL_FRONT_LEFT
| + | | 1: MA_CHANNEL_FRONT_RIGHT
| + | | 2: MA_CHANNEL_FRONT_CENTER
| + | | 3: MA_CHANNEL_LFE
| + | | 4: MA_CHANNEL_BACK_CENTER
| + | | 4: MA_CHANNEL_SIDE_LEFT
| + | | 5: MA_CHANNEL_SIDE_RIGHT | + +---------------+---------------------------------+ + | 8 (7.1) | 0: MA_CHANNEL_FRONT_LEFT
| + | | 1: MA_CHANNEL_FRONT_RIGHT
| + | | 2: MA_CHANNEL_FRONT_CENTER
| + | | 3: MA_CHANNEL_LFE
| + | | 4: MA_CHANNEL_BACK_LEFT
| + | | 5: MA_CHANNEL_BACK_RIGHT
| + | | 6: MA_CHANNEL_SIDE_LEFT
| + | | 7: MA_CHANNEL_SIDE_RIGHT | + +---------------+---------------------------------+ + | Other | All channels set to 0. This | + | | is equivalent to the same | + | | mapping as the device. | + +---------------+---------------------------------+ + + + +10.3. Resampling +---------------- +Resampling is achieved with the `ma_resampler` object. To create a resampler object, do something +like the following: + + ```c + ma_resampler_config config = ma_resampler_config_init( + ma_format_s16, + channels, + sampleRateIn, + sampleRateOut, + ma_resample_algorithm_linear); + + ma_resampler resampler; + ma_result result = ma_resampler_init(&config, NULL, &resampler); + if (result != MA_SUCCESS) { + // An error occurred... + } + ``` + +Do the following to uninitialize the resampler: + + ```c + ma_resampler_uninit(&resampler); + ``` + +The following example shows how data can be processed + + ```c + ma_uint64 frameCountIn = 1000; + ma_uint64 frameCountOut = 2000; + ma_result result = ma_resampler_process_pcm_frames(&resampler, pFramesIn, &frameCountIn, pFramesOut, &frameCountOut); + if (result != MA_SUCCESS) { + // An error occurred... + } + + // At this point, frameCountIn contains the number of input frames that were consumed and frameCountOut contains the + // number of output frames written. + ``` + +To initialize the resampler you first need to set up a config (`ma_resampler_config`) with +`ma_resampler_config_init()`. You need to specify the sample format you want to use, the number of +channels, the input and output sample rate, and the algorithm. + +The sample format can be either `ma_format_s16` or `ma_format_f32`. If you need a different format +you will need to perform pre- and post-conversions yourself where necessary. Note that the format +is the same for both input and output. The format cannot be changed after initialization. + +The resampler supports multiple channels and is always interleaved (both input and output). The +channel count cannot be changed after initialization. + +The sample rates can be anything other than zero, and are always specified in hertz. They should be +set to something like 44100, etc. The sample rate is the only configuration property that can be +changed after initialization. + +The miniaudio resampler has built-in support for the following algorithms: + + +-----------+------------------------------+ + | Algorithm | Enum Token | + +-----------+------------------------------+ + | Linear | ma_resample_algorithm_linear | + | Custom | ma_resample_algorithm_custom | + +-----------+------------------------------+ + +The algorithm cannot be changed after initialization. + +Processing always happens on a per PCM frame basis and always assumes interleaved input and output. +De-interleaved processing is not supported. To process frames, use +`ma_resampler_process_pcm_frames()`. On input, this function takes the number of output frames you +can fit in the output buffer and the number of input frames contained in the input buffer. On +output these variables contain the number of output frames that were written to the output buffer +and the number of input frames that were consumed in the process. You can pass in NULL for the +input buffer in which case it will be treated as an infinitely large buffer of zeros. The output +buffer can also be NULL, in which case the processing will be treated as seek. + +The sample rate can be changed dynamically on the fly. You can change this with explicit sample +rates with `ma_resampler_set_rate()` and also with a decimal ratio with +`ma_resampler_set_rate_ratio()`. The ratio is in/out. + +Sometimes it's useful to know exactly how many input frames will be required to output a specific +number of frames. You can calculate this with `ma_resampler_get_required_input_frame_count()`. +Likewise, it's sometimes useful to know exactly how many frames would be output given a certain +number of input frames. You can do this with `ma_resampler_get_expected_output_frame_count()`. + +Due to the nature of how resampling works, the resampler introduces some latency. This can be +retrieved in terms of both the input rate and the output rate with +`ma_resampler_get_input_latency()` and `ma_resampler_get_output_latency()`. + + +10.3.1. Resampling Algorithms +----------------------------- +The choice of resampling algorithm depends on your situation and requirements. + + +10.3.1.1. Linear Resampling +--------------------------- +The linear resampler is the fastest, but comes at the expense of poorer quality. There is, however, +some control over the quality of the linear resampler which may make it a suitable option depending +on your requirements. + +The linear resampler performs low-pass filtering before or after downsampling or upsampling, +depending on the sample rates you're converting between. When decreasing the sample rate, the +low-pass filter will be applied before downsampling. When increasing the rate it will be performed +after upsampling. By default a fourth order low-pass filter will be applied. This can be configured +via the `lpfOrder` configuration variable. Setting this to 0 will disable filtering. + +The low-pass filter has a cutoff frequency which defaults to half the sample rate of the lowest of +the input and output sample rates (Nyquist Frequency). + +The API for the linear resampler is the same as the main resampler API, only it's called +`ma_linear_resampler`. + + +10.3.2. Custom Resamplers +------------------------- +You can implement a custom resampler by using the `ma_resample_algorithm_custom` resampling +algorithm and setting a vtable in the resampler config: + + ```c + ma_resampler_config config = ma_resampler_config_init(..., ma_resample_algorithm_custom); + config.pBackendVTable = &g_customResamplerVTable; + ``` + +Custom resamplers are useful if the stock algorithms are not appropriate for your use case. You +need to implement the required functions in `ma_resampling_backend_vtable`. Note that not all +functions in the vtable need to be implemented, but if it's possible to implement, they should be. + +You can use the `ma_linear_resampler` object for an example on how to implement the vtable. The +`onGetHeapSize` callback is used to calculate the size of any internal heap allocation the custom +resampler will need to make given the supplied config. When you initialize the resampler via the +`onInit` callback, you'll be given a pointer to a heap allocation which is where you should store +the heap allocated data. You should not free this data in `onUninit` because miniaudio will manage +it for you. + +The `onProcess` callback is where the actual resampling takes place. On input, `pFrameCountIn` +points to a variable containing the number of frames in the `pFramesIn` buffer and +`pFrameCountOut` points to a variable containing the capacity in frames of the `pFramesOut` buffer. +On output, `pFrameCountIn` should be set to the number of input frames that were fully consumed, +whereas `pFrameCountOut` should be set to the number of frames that were written to `pFramesOut`. + +The `onSetRate` callback is optional and is used for dynamically changing the sample rate. If +dynamic rate changes are not supported, you can set this callback to NULL. + +The `onGetInputLatency` and `onGetOutputLatency` functions are used for retrieving the latency in +input and output rates respectively. These can be NULL in which case latency calculations will be +assumed to be NULL. + +The `onGetRequiredInputFrameCount` callback is used to give miniaudio a hint as to how many input +frames are required to be available to produce the given number of output frames. Likewise, the +`onGetExpectedOutputFrameCount` callback is used to determine how many output frames will be +produced given the specified number of input frames. miniaudio will use these as a hint, but they +are optional and can be set to NULL if you're unable to implement them. + + + +10.4. General Data Conversion +----------------------------- +The `ma_data_converter` API can be used to wrap sample format conversion, channel conversion and +resampling into one operation. This is what miniaudio uses internally to convert between the format +requested when the device was initialized and the format of the backend's native device. The API +for general data conversion is very similar to the resampling API. Create a `ma_data_converter` +object like this: + + ```c + ma_data_converter_config config = ma_data_converter_config_init( + inputFormat, + outputFormat, + inputChannels, + outputChannels, + inputSampleRate, + outputSampleRate + ); + + ma_data_converter converter; + ma_result result = ma_data_converter_init(&config, NULL, &converter); + if (result != MA_SUCCESS) { + // An error occurred... + } + ``` + +In the example above we use `ma_data_converter_config_init()` to initialize the config, however +there's many more properties that can be configured, such as channel maps and resampling quality. +Something like the following may be more suitable depending on your requirements: + + ```c + ma_data_converter_config config = ma_data_converter_config_init_default(); + config.formatIn = inputFormat; + config.formatOut = outputFormat; + config.channelsIn = inputChannels; + config.channelsOut = outputChannels; + config.sampleRateIn = inputSampleRate; + config.sampleRateOut = outputSampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_flac, config.channelMapIn, sizeof(config.channelMapIn)/sizeof(config.channelMapIn[0]), config.channelCountIn); + config.resampling.linear.lpfOrder = MA_MAX_FILTER_ORDER; + ``` + +Do the following to uninitialize the data converter: + + ```c + ma_data_converter_uninit(&converter, NULL); + ``` + +The following example shows how data can be processed + + ```c + ma_uint64 frameCountIn = 1000; + ma_uint64 frameCountOut = 2000; + ma_result result = ma_data_converter_process_pcm_frames(&converter, pFramesIn, &frameCountIn, pFramesOut, &frameCountOut); + if (result != MA_SUCCESS) { + // An error occurred... + } + + // At this point, frameCountIn contains the number of input frames that were consumed and frameCountOut contains the number + // of output frames written. + ``` + +The data converter supports multiple channels and is always interleaved (both input and output). +The channel count cannot be changed after initialization. + +Sample rates can be anything other than zero, and are always specified in hertz. They should be set +to something like 44100, etc. The sample rate is the only configuration property that can be +changed after initialization, but only if the `resampling.allowDynamicSampleRate` member of +`ma_data_converter_config` is set to `MA_TRUE`. To change the sample rate, use +`ma_data_converter_set_rate()` or `ma_data_converter_set_rate_ratio()`. The ratio must be in/out. +The resampling algorithm cannot be changed after initialization. + +Processing always happens on a per PCM frame basis and always assumes interleaved input and output. +De-interleaved processing is not supported. To process frames, use +`ma_data_converter_process_pcm_frames()`. On input, this function takes the number of output frames +you can fit in the output buffer and the number of input frames contained in the input buffer. On +output these variables contain the number of output frames that were written to the output buffer +and the number of input frames that were consumed in the process. You can pass in NULL for the +input buffer in which case it will be treated as an infinitely large +buffer of zeros. The output buffer can also be NULL, in which case the processing will be treated +as seek. + +Sometimes it's useful to know exactly how many input frames will be required to output a specific +number of frames. You can calculate this with `ma_data_converter_get_required_input_frame_count()`. +Likewise, it's sometimes useful to know exactly how many frames would be output given a certain +number of input frames. You can do this with `ma_data_converter_get_expected_output_frame_count()`. + +Due to the nature of how resampling works, the data converter introduces some latency if resampling +is required. This can be retrieved in terms of both the input rate and the output rate with +`ma_data_converter_get_input_latency()` and `ma_data_converter_get_output_latency()`. + + + +11. Filtering +============= + +11.1. Biquad Filtering +---------------------- +Biquad filtering is achieved with the `ma_biquad` API. Example: + + ```c + ma_biquad_config config = ma_biquad_config_init(ma_format_f32, channels, b0, b1, b2, a0, a1, a2); + ma_result result = ma_biquad_init(&config, NULL, &biquad); + if (result != MA_SUCCESS) { + // Error. + } + + ... + + ma_biquad_process_pcm_frames(&biquad, pFramesOut, pFramesIn, frameCount); + ``` + +Biquad filtering is implemented using transposed direct form 2. The numerator coefficients are b0, +b1 and b2, and the denominator coefficients are a0, a1 and a2. The a0 coefficient is required and +coefficients must not be pre-normalized. + +Supported formats are `ma_format_s16` and `ma_format_f32`. If you need to use a different format +you need to convert it yourself beforehand. When using `ma_format_s16` the biquad filter will use +fixed point arithmetic. When using `ma_format_f32`, floating point arithmetic will be used. + +Input and output frames are always interleaved. + +Filtering can be applied in-place by passing in the same pointer for both the input and output +buffers, like so: + + ```c + ma_biquad_process_pcm_frames(&biquad, pMyData, pMyData, frameCount); + ``` + +If you need to change the values of the coefficients, but maintain the values in the registers you +can do so with `ma_biquad_reinit()`. This is useful if you need to change the properties of the +filter while keeping the values of registers valid to avoid glitching. Do not use +`ma_biquad_init()` for this as it will do a full initialization which involves clearing the +registers to 0. Note that changing the format or channel count after initialization is invalid and +will result in an error. + + +11.2. Low-Pass Filtering +------------------------ +Low-pass filtering is achieved with the following APIs: + + +---------+------------------------------------------+ + | API | Description | + +---------+------------------------------------------+ + | ma_lpf1 | First order low-pass filter | + | ma_lpf2 | Second order low-pass filter | + | ma_lpf | High order low-pass filter (Butterworth) | + +---------+------------------------------------------+ + +Low-pass filter example: + + ```c + ma_lpf_config config = ma_lpf_config_init(ma_format_f32, channels, sampleRate, cutoffFrequency, order); + ma_result result = ma_lpf_init(&config, &lpf); + if (result != MA_SUCCESS) { + // Error. + } + + ... + + ma_lpf_process_pcm_frames(&lpf, pFramesOut, pFramesIn, frameCount); + ``` + +Supported formats are `ma_format_s16` and` ma_format_f32`. If you need to use a different format +you need to convert it yourself beforehand. Input and output frames are always interleaved. + +Filtering can be applied in-place by passing in the same pointer for both the input and output +buffers, like so: + + ```c + ma_lpf_process_pcm_frames(&lpf, pMyData, pMyData, frameCount); + ``` + +The maximum filter order is limited to `MA_MAX_FILTER_ORDER` which is set to 8. If you need more, +you can chain first and second order filters together. + + ```c + for (iFilter = 0; iFilter < filterCount; iFilter += 1) { + ma_lpf2_process_pcm_frames(&lpf2[iFilter], pMyData, pMyData, frameCount); + } + ``` + +If you need to change the configuration of the filter, but need to maintain the state of internal +registers you can do so with `ma_lpf_reinit()`. This may be useful if you need to change the sample +rate and/or cutoff frequency dynamically while maintaining smooth transitions. Note that changing the +format or channel count after initialization is invalid and will result in an error. + +The `ma_lpf` object supports a configurable order, but if you only need a first order filter you +may want to consider using `ma_lpf1`. Likewise, if you only need a second order filter you can use +`ma_lpf2`. The advantage of this is that they're lighter weight and a bit more efficient. + +If an even filter order is specified, a series of second order filters will be processed in a +chain. If an odd filter order is specified, a first order filter will be applied, followed by a +series of second order filters in a chain. + + +11.3. High-Pass Filtering +------------------------- +High-pass filtering is achieved with the following APIs: + + +---------+-------------------------------------------+ + | API | Description | + +---------+-------------------------------------------+ + | ma_hpf1 | First order high-pass filter | + | ma_hpf2 | Second order high-pass filter | + | ma_hpf | High order high-pass filter (Butterworth) | + +---------+-------------------------------------------+ + +High-pass filters work exactly the same as low-pass filters, only the APIs are called `ma_hpf1`, +`ma_hpf2` and `ma_hpf`. See example code for low-pass filters for example usage. + + +11.4. Band-Pass Filtering +------------------------- +Band-pass filtering is achieved with the following APIs: + + +---------+-------------------------------+ + | API | Description | + +---------+-------------------------------+ + | ma_bpf2 | Second order band-pass filter | + | ma_bpf | High order band-pass filter | + +---------+-------------------------------+ + +Band-pass filters work exactly the same as low-pass filters, only the APIs are called `ma_bpf2` and +`ma_hpf`. See example code for low-pass filters for example usage. Note that the order for +band-pass filters must be an even number which means there is no first order band-pass filter, +unlike low-pass and high-pass filters. + + +11.5. Notch Filtering +--------------------- +Notch filtering is achieved with the following APIs: + + +-----------+------------------------------------------+ + | API | Description | + +-----------+------------------------------------------+ + | ma_notch2 | Second order notching filter | + +-----------+------------------------------------------+ + + +11.6. Peaking EQ Filtering +------------------------- +Peaking filtering is achieved with the following APIs: + + +----------+------------------------------------------+ + | API | Description | + +----------+------------------------------------------+ + | ma_peak2 | Second order peaking filter | + +----------+------------------------------------------+ + + +11.7. Low Shelf Filtering +------------------------- +Low shelf filtering is achieved with the following APIs: + + +-------------+------------------------------------------+ + | API | Description | + +-------------+------------------------------------------+ + | ma_loshelf2 | Second order low shelf filter | + +-------------+------------------------------------------+ + +Where a high-pass filter is used to eliminate lower frequencies, a low shelf filter can be used to +just turn them down rather than eliminate them entirely. + + +11.8. High Shelf Filtering +-------------------------- +High shelf filtering is achieved with the following APIs: + + +-------------+------------------------------------------+ + | API | Description | + +-------------+------------------------------------------+ + | ma_hishelf2 | Second order high shelf filter | + +-------------+------------------------------------------+ + +The high shelf filter has the same API as the low shelf filter, only you would use `ma_hishelf` +instead of `ma_loshelf`. Where a low shelf filter is used to adjust the volume of low frequencies, +the high shelf filter does the same thing for high frequencies. + + + + +12. Waveform and Noise Generation +================================= + +12.1. Waveforms +--------------- +miniaudio supports generation of sine, square, triangle and sawtooth waveforms. This is achieved +with the `ma_waveform` API. Example: + + ```c + ma_waveform_config config = ma_waveform_config_init( + FORMAT, + CHANNELS, + SAMPLE_RATE, + ma_waveform_type_sine, + amplitude, + frequency); + + ma_waveform waveform; + ma_result result = ma_waveform_init(&config, &waveform); + if (result != MA_SUCCESS) { + // Error. + } + + ... + + ma_waveform_read_pcm_frames(&waveform, pOutput, frameCount); + ``` + +The amplitude, frequency, type, and sample rate can be changed dynamically with +`ma_waveform_set_amplitude()`, `ma_waveform_set_frequency()`, `ma_waveform_set_type()`, and +`ma_waveform_set_sample_rate()` respectively. + +You can invert the waveform by setting the amplitude to a negative value. You can use this to +control whether or not a sawtooth has a positive or negative ramp, for example. + +Below are the supported waveform types: + + +---------------------------+ + | Enum Name | + +---------------------------+ + | ma_waveform_type_sine | + | ma_waveform_type_square | + | ma_waveform_type_triangle | + | ma_waveform_type_sawtooth | + +---------------------------+ + + + +12.2. Noise +----------- +miniaudio supports generation of white, pink and Brownian noise via the `ma_noise` API. Example: + + ```c + ma_noise_config config = ma_noise_config_init( + FORMAT, + CHANNELS, + ma_noise_type_white, + SEED, + amplitude); + + ma_noise noise; + ma_result result = ma_noise_init(&config, &noise); + if (result != MA_SUCCESS) { + // Error. + } + + ... + + ma_noise_read_pcm_frames(&noise, pOutput, frameCount); + ``` + +The noise API uses simple LCG random number generation. It supports a custom seed which is useful +for things like automated testing requiring reproducibility. Setting the seed to zero will default +to `MA_DEFAULT_LCG_SEED`. + +The amplitude and seed can be changed dynamically with `ma_noise_set_amplitude()` and +`ma_noise_set_seed()` respectively. + +By default, the noise API will use different values for different channels. So, for example, the +left side in a stereo stream will be different to the right side. To instead have each channel use +the same random value, set the `duplicateChannels` member of the noise config to true, like so: + + ```c + config.duplicateChannels = MA_TRUE; + ``` + +Below are the supported noise types. + + +------------------------+ + | Enum Name | + +------------------------+ + | ma_noise_type_white | + | ma_noise_type_pink | + | ma_noise_type_brownian | + +------------------------+ + + + +13. Audio Buffers +================= +miniaudio supports reading from a buffer of raw audio data via the `ma_audio_buffer` API. This can +read from memory that's managed by the application, but can also handle the memory management for +you internally. Memory management is flexible and should support most use cases. + +Audio buffers are initialized using the standard configuration system used everywhere in miniaudio: + + ```c + ma_audio_buffer_config config = ma_audio_buffer_config_init( + format, + channels, + sizeInFrames, + pExistingData, + &allocationCallbacks); + + ma_audio_buffer buffer; + result = ma_audio_buffer_init(&config, &buffer); + if (result != MA_SUCCESS) { + // Error. + } + + ... + + ma_audio_buffer_uninit(&buffer); + ``` + +In the example above, the memory pointed to by `pExistingData` will *not* be copied and is how an +application can do self-managed memory allocation. If you would rather make a copy of the data, use +`ma_audio_buffer_init_copy()`. To uninitialize the buffer, use `ma_audio_buffer_uninit()`. + +Sometimes it can be convenient to allocate the memory for the `ma_audio_buffer` structure and the +raw audio data in a contiguous block of memory. That is, the raw audio data will be located +immediately after the `ma_audio_buffer` structure. To do this, use +`ma_audio_buffer_alloc_and_init()`: + + ```c + ma_audio_buffer_config config = ma_audio_buffer_config_init( + format, + channels, + sizeInFrames, + pExistingData, + &allocationCallbacks); + + ma_audio_buffer* pBuffer + result = ma_audio_buffer_alloc_and_init(&config, &pBuffer); + if (result != MA_SUCCESS) { + // Error + } + + ... + + ma_audio_buffer_uninit_and_free(&buffer); + ``` + +If you initialize the buffer with `ma_audio_buffer_alloc_and_init()` you should uninitialize it +with `ma_audio_buffer_uninit_and_free()`. In the example above, the memory pointed to by +`pExistingData` will be copied into the buffer, which is contrary to the behavior of +`ma_audio_buffer_init()`. + +An audio buffer has a playback cursor just like a decoder. As you read frames from the buffer, the +cursor moves forward. The last parameter (`loop`) can be used to determine if the buffer should +loop. The return value is the number of frames actually read. If this is less than the number of +frames requested it means the end has been reached. This should never happen if the `loop` +parameter is set to true. If you want to manually loop back to the start, you can do so with with +`ma_audio_buffer_seek_to_pcm_frame(pAudioBuffer, 0)`. Below is an example for reading data from an +audio buffer. + + ```c + ma_uint64 framesRead = ma_audio_buffer_read_pcm_frames(pAudioBuffer, pFramesOut, desiredFrameCount, isLooping); + if (framesRead < desiredFrameCount) { + // If not looping, this means the end has been reached. This should never happen in looping mode with valid input. + } + ``` + +Sometimes you may want to avoid the cost of data movement between the internal buffer and the +output buffer. Instead you can use memory mapping to retrieve a pointer to a segment of data: + + ```c + void* pMappedFrames; + ma_uint64 frameCount = frameCountToTryMapping; + ma_result result = ma_audio_buffer_map(pAudioBuffer, &pMappedFrames, &frameCount); + if (result == MA_SUCCESS) { + // Map was successful. The value in frameCount will be how many frames were _actually_ mapped, which may be + // less due to the end of the buffer being reached. + ma_copy_pcm_frames(pFramesOut, pMappedFrames, frameCount, pAudioBuffer->format, pAudioBuffer->channels); + + // You must unmap the buffer. + ma_audio_buffer_unmap(pAudioBuffer, frameCount); + } + ``` + +When you use memory mapping, the read cursor is increment by the frame count passed in to +`ma_audio_buffer_unmap()`. If you decide not to process every frame you can pass in a value smaller +than the value returned by `ma_audio_buffer_map()`. The disadvantage to using memory mapping is +that it does not handle looping for you. You can determine if the buffer is at the end for the +purpose of looping with `ma_audio_buffer_at_end()` or by inspecting the return value of +`ma_audio_buffer_unmap()` and checking if it equals `MA_AT_END`. You should not treat `MA_AT_END` +as an error when returned by `ma_audio_buffer_unmap()`. + + + +14. Ring Buffers +================ +miniaudio supports lock free (single producer, single consumer) ring buffers which are exposed via +the `ma_rb` and `ma_pcm_rb` APIs. The `ma_rb` API operates on bytes, whereas the `ma_pcm_rb` +operates on PCM frames. They are otherwise identical as `ma_pcm_rb` is just a wrapper around +`ma_rb`. + +Unlike most other APIs in miniaudio, ring buffers support both interleaved and deinterleaved +streams. The caller can also allocate their own backing memory for the ring buffer to use +internally for added flexibility. Otherwise the ring buffer will manage it's internal memory for +you. + +The examples below use the PCM frame variant of the ring buffer since that's most likely the one +you will want to use. To initialize a ring buffer, do something like the following: + + ```c + ma_pcm_rb rb; + ma_result result = ma_pcm_rb_init(FORMAT, CHANNELS, BUFFER_SIZE_IN_FRAMES, NULL, NULL, &rb); + if (result != MA_SUCCESS) { + // Error + } + ``` + +The `ma_pcm_rb_init()` function takes the sample format and channel count as parameters because +it's the PCM variant of the ring buffer API. For the regular ring buffer that operates on bytes you +would call `ma_rb_init()` which leaves these out and just takes the size of the buffer in bytes +instead of frames. The fourth parameter is an optional pre-allocated buffer and the fifth parameter +is a pointer to a `ma_allocation_callbacks` structure for custom memory allocation routines. +Passing in `NULL` for this results in `MA_MALLOC()` and `MA_FREE()` being used. + +Use `ma_pcm_rb_init_ex()` if you need a deinterleaved buffer. The data for each sub-buffer is +offset from each other based on the stride. To manage your sub-buffers you can use +`ma_pcm_rb_get_subbuffer_stride()`, `ma_pcm_rb_get_subbuffer_offset()` and +`ma_pcm_rb_get_subbuffer_ptr()`. + +Use `ma_pcm_rb_acquire_read()` and `ma_pcm_rb_acquire_write()` to retrieve a pointer to a section +of the ring buffer. You specify the number of frames you need, and on output it will set to what +was actually acquired. If the read or write pointer is positioned such that the number of frames +requested will require a loop, it will be clamped to the end of the buffer. Therefore, the number +of frames you're given may be less than the number you requested. + +After calling `ma_pcm_rb_acquire_read()` or `ma_pcm_rb_acquire_write()`, you do your work on the +buffer and then "commit" it with `ma_pcm_rb_commit_read()` or `ma_pcm_rb_commit_write()`. This is +where the read/write pointers are updated. When you commit you need to pass in the buffer that was +returned by the earlier call to `ma_pcm_rb_acquire_read()` or `ma_pcm_rb_acquire_write()` and is +only used for validation. The number of frames passed to `ma_pcm_rb_commit_read()` and +`ma_pcm_rb_commit_write()` is what's used to increment the pointers, and can be less that what was +originally requested. + +If you want to correct for drift between the write pointer and the read pointer you can use a +combination of `ma_pcm_rb_pointer_distance()`, `ma_pcm_rb_seek_read()` and +`ma_pcm_rb_seek_write()`. Note that you can only move the pointers forward, and you should only +move the read pointer forward via the consumer thread, and the write pointer forward by the +producer thread. If there is too much space between the pointers, move the read pointer forward. If +there is too little space between the pointers, move the write pointer forward. + +You can use a ring buffer at the byte level instead of the PCM frame level by using the `ma_rb` +API. This is exactly the same, only you will use the `ma_rb` functions instead of `ma_pcm_rb` and +instead of frame counts you will pass around byte counts. + +The maximum size of the buffer in bytes is `0x7FFFFFFF-(MA_SIMD_ALIGNMENT-1)` due to the most +significant bit being used to encode a loop flag and the internally managed buffers always being +aligned to `MA_SIMD_ALIGNMENT`. + +Note that the ring buffer is only thread safe when used by a single consumer thread and single +producer thread. + + + +15. Backends +============ +The following backends are supported by miniaudio. These are listed in order of default priority. +When no backend is specified when initializing a context or device, miniaudio will attempt to use +each of these backends in the order listed in the table below. + +Note that backends that are not usable by the build target will not be included in the build. For +example, ALSA, which is specific to Linux, will not be included in the Windows build. + + +-------------+-----------------------+--------------------------------------------------------+ + | Name | Enum Name | Supported Operating Systems | + +-------------+-----------------------+--------------------------------------------------------+ + | WASAPI | ma_backend_wasapi | Windows Vista+ | + | DirectSound | ma_backend_dsound | Windows XP+ | + | WinMM | ma_backend_winmm | Windows 95+ | + | Core Audio | ma_backend_coreaudio | macOS, iOS | + | sndio | ma_backend_sndio | OpenBSD | + | audio(4) | ma_backend_audio4 | NetBSD, OpenBSD | + | OSS | ma_backend_oss | FreeBSD | + | PulseAudio | ma_backend_pulseaudio | Cross Platform (disabled on Windows, BSD and Android) | + | ALSA | ma_backend_alsa | Linux | + | JACK | ma_backend_jack | Cross Platform (disabled on BSD and Android) | + | AAudio | ma_backend_aaudio | Android 8+ | + | OpenSL ES | ma_backend_opensl | Android (API level 16+) | + | Web Audio | ma_backend_webaudio | Web (via Emscripten) | + | Custom | ma_backend_custom | Cross Platform | + | Null | ma_backend_null | Cross Platform (not used on Web) | + +-------------+-----------------------+--------------------------------------------------------+ + +Some backends have some nuance details you may want to be aware of. + +15.1. WASAPI +------------ +- Low-latency shared mode will be disabled when using an application-defined sample rate which is + different to the device's native sample rate. To work around this, set `wasapi.noAutoConvertSRC` + to true in the device config. This is due to IAudioClient3_InitializeSharedAudioStream() failing + when the `AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM` flag is specified. Setting wasapi.noAutoConvertSRC + will result in miniaudio's internal resampler being used instead which will in turn enable the + use of low-latency shared mode. + +15.2. PulseAudio +---------------- +- If you experience bad glitching/noise on Arch Linux, consider this fix from the Arch wiki: + https://wiki.archlinux.org/index.php/PulseAudio/Troubleshooting#Glitches,_skips_or_crackling. + Alternatively, consider using a different backend such as ALSA. + +15.3. Android +------------- +- To capture audio on Android, remember to add the RECORD_AUDIO permission to your manifest: + `` +- With OpenSL|ES, only a single ma_context can be active at any given time. This is due to a + limitation with OpenSL|ES. +- With AAudio, only default devices are enumerated. This is due to AAudio not having an enumeration + API (devices are enumerated through Java). You can however perform your own device enumeration + through Java and then set the ID in the ma_device_id structure (ma_device_id.aaudio) and pass it + to ma_device_init(). +- The backend API will perform resampling where possible. The reason for this as opposed to using + miniaudio's built-in resampler is to take advantage of any potential device-specific + optimizations the driver may implement. + +BSD +--- +- The sndio backend is currently only enabled on OpenBSD builds. +- The audio(4) backend is supported on OpenBSD, but you may need to disable sndiod before you can + use it. + +15.4. UWP +--------- +- UWP only supports default playback and capture devices. +- UWP requires the Microphone capability to be enabled in the application's manifest (Package.appxmanifest): + + ``` + + ... + + + + + ``` + +15.5. Web Audio / Emscripten +---------------------------- +- You cannot use `-std=c*` compiler flags, nor `-ansi`. This only applies to the Emscripten build. +- The first time a context is initialized it will create a global object called "miniaudio" whose + primary purpose is to act as a factory for device objects. +- Currently the Web Audio backend uses ScriptProcessorNode's, but this may need to change later as + they've been deprecated. +- Google has implemented a policy in their browsers that prevent automatic media output without + first receiving some kind of user input. The following web page has additional details: + https://developers.google.com/web/updates/2017/09/autoplay-policy-changes. Starting the device + may fail if you try to start playback without first handling some kind of user input. + + + +16. Optimization Tips +===================== +See below for some tips on improving performance. + +16.1. Low Level API +------------------- +- In the data callback, if your data is already clipped prior to copying it into the output buffer, + set the `noClip` config option in the device config to true. This will disable miniaudio's built + in clipping function. +- By default, miniaudio will pre-silence the data callback's output buffer. If you know that you + will always write valid data to the output buffer you can disable pre-silencing by setting the + `noPreSilence` config option in the device config to true. + +16.2. High Level API +-------------------- +- If a sound does not require doppler or pitch shifting, consider disabling pitching by + initializing the sound with the `MA_SOUND_FLAG_NO_PITCH` flag. +- If a sound does not require spatialization, disable it by initializing the sound with the + `MA_SOUND_FLAG_NO_SPATIALIZATION` flag. It can be re-enabled again post-initialization with + `ma_sound_set_spatialization_enabled()`. +- If you know all of your sounds will always be the same sample rate, set the engine's sample + rate to match that of the sounds. Likewise, if you're using a self-managed resource manager, + consider setting the decoded sample rate to match your sounds. By configuring everything to + use a consistent sample rate, sample rate conversion can be avoided. + + + +17. Miscellaneous Notes +======================= +- Automatic stream routing is enabled on a per-backend basis. Support is explicitly enabled for + WASAPI and Core Audio, however other backends such as PulseAudio may naturally support it, though + not all have been tested. +- When compiling with VC6 and earlier, decoding is restricted to files less than 2GB in size. This + is due to 64-bit file APIs not being available. +*/ + +#ifndef miniaudio_h +#define miniaudio_h + +#ifdef __cplusplus +extern "C" { +#endif + +#define MA_STRINGIFY(x) #x +#define MA_XSTRINGIFY(x) MA_STRINGIFY(x) + +#define MA_VERSION_MAJOR 0 +#define MA_VERSION_MINOR 11 +#define MA_VERSION_REVISION 22 +#define MA_VERSION_STRING MA_XSTRINGIFY(MA_VERSION_MAJOR) "." MA_XSTRINGIFY(MA_VERSION_MINOR) "." MA_XSTRINGIFY(MA_VERSION_REVISION) + +#if defined(_MSC_VER) && !defined(__clang__) + #pragma warning(push) + #pragma warning(disable:4201) /* nonstandard extension used: nameless struct/union */ + #pragma warning(disable:4214) /* nonstandard extension used: bit field types other than int */ + #pragma warning(disable:4324) /* structure was padded due to alignment specifier */ +#elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wpedantic" /* For ISO C99 doesn't support unnamed structs/unions [-Wpedantic] */ + #if defined(__clang__) + #pragma GCC diagnostic ignored "-Wc11-extensions" /* anonymous unions are a C11 extension */ + #endif +#endif + + +#if defined(__LP64__) || defined(_WIN64) || (defined(__x86_64__) && !defined(__ILP32__)) || defined(_M_X64) || defined(__ia64) || defined(_M_IA64) || defined(__aarch64__) || defined(_M_ARM64) || defined(__powerpc64__) || defined(__ppc64__) + #define MA_SIZEOF_PTR 8 +#else + #define MA_SIZEOF_PTR 4 +#endif + +#include /* For size_t. */ + +/* Sized types. */ +#if defined(MA_USE_STDINT) + #include + typedef int8_t ma_int8; + typedef uint8_t ma_uint8; + typedef int16_t ma_int16; + typedef uint16_t ma_uint16; + typedef int32_t ma_int32; + typedef uint32_t ma_uint32; + typedef int64_t ma_int64; + typedef uint64_t ma_uint64; +#else + typedef signed char ma_int8; + typedef unsigned char ma_uint8; + typedef signed short ma_int16; + typedef unsigned short ma_uint16; + typedef signed int ma_int32; + typedef unsigned int ma_uint32; + #if defined(_MSC_VER) && !defined(__clang__) + typedef signed __int64 ma_int64; + typedef unsigned __int64 ma_uint64; + #else + #if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wlong-long" + #if defined(__clang__) + #pragma GCC diagnostic ignored "-Wc++11-long-long" + #endif + #endif + typedef signed long long ma_int64; + typedef unsigned long long ma_uint64; + #if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic pop + #endif + #endif +#endif /* MA_USE_STDINT */ + +#if MA_SIZEOF_PTR == 8 + typedef ma_uint64 ma_uintptr; +#else + typedef ma_uint32 ma_uintptr; +#endif + +typedef ma_uint8 ma_bool8; +typedef ma_uint32 ma_bool32; +#define MA_TRUE 1 +#define MA_FALSE 0 + +/* These float types are not used universally by miniaudio. It's to simplify some macro expansion for atomic types. */ +typedef float ma_float; +typedef double ma_double; + +typedef void* ma_handle; +typedef void* ma_ptr; + +/* +ma_proc is annoying because when compiling with GCC we get pedantic warnings about converting +between `void*` and `void (*)()`. We can't use `void (*)()` with MSVC however, because we'll get +warning C4191 about "type cast between incompatible function types". To work around this I'm going +to use a different data type depending on the compiler. +*/ +#if defined(__GNUC__) +typedef void (*ma_proc)(void); +#else +typedef void* ma_proc; +#endif + +#if defined(_MSC_VER) && !defined(_WCHAR_T_DEFINED) +typedef ma_uint16 wchar_t; +#endif + +/* Define NULL for some compilers. */ +#ifndef NULL +#define NULL 0 +#endif + +#if defined(SIZE_MAX) + #define MA_SIZE_MAX SIZE_MAX +#else + #define MA_SIZE_MAX 0xFFFFFFFF /* When SIZE_MAX is not defined by the standard library just default to the maximum 32-bit unsigned integer. */ +#endif + + +/* Platform/backend detection. */ +#if defined(_WIN32) || defined(__COSMOPOLITAN__) + #define MA_WIN32 + #if defined(MA_FORCE_UWP) || (defined(WINAPI_FAMILY) && ((defined(WINAPI_FAMILY_PC_APP) && WINAPI_FAMILY == WINAPI_FAMILY_PC_APP) || (defined(WINAPI_FAMILY_PHONE_APP) && WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP))) + #define MA_WIN32_UWP + #elif defined(WINAPI_FAMILY) && (defined(WINAPI_FAMILY_GAMES) && WINAPI_FAMILY == WINAPI_FAMILY_GAMES) + #define MA_WIN32_GDK + #else + #define MA_WIN32_DESKTOP + #endif +#endif +#if !defined(_WIN32) /* If it's not Win32, assume POSIX. */ + #define MA_POSIX + + /* + Use the MA_NO_PTHREAD_IN_HEADER option at your own risk. This is intentionally undocumented. + You can use this to avoid including pthread.h in the header section. The downside is that it + results in some fixed sized structures being declared for the various types that are used in + miniaudio. The risk here is that these types might be too small for a given platform. This + risk is yours to take and no support will be offered if you enable this option. + */ + #ifndef MA_NO_PTHREAD_IN_HEADER + #include /* Unfortunate #include, but needed for pthread_t, pthread_mutex_t and pthread_cond_t types. */ + typedef pthread_t ma_pthread_t; + typedef pthread_mutex_t ma_pthread_mutex_t; + typedef pthread_cond_t ma_pthread_cond_t; + #else + typedef ma_uintptr ma_pthread_t; + typedef union ma_pthread_mutex_t { char __data[40]; ma_uint64 __alignment; } ma_pthread_mutex_t; + typedef union ma_pthread_cond_t { char __data[48]; ma_uint64 __alignment; } ma_pthread_cond_t; + #endif + + #if defined(__unix__) + #define MA_UNIX + #endif + #if defined(__linux__) + #define MA_LINUX + #endif + #if defined(__APPLE__) + #define MA_APPLE + #endif + #if defined(__DragonFly__) || defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) + #define MA_BSD + #endif + #if defined(__ANDROID__) + #define MA_ANDROID + #endif + #if defined(__EMSCRIPTEN__) + #define MA_EMSCRIPTEN + #endif + #if defined(__ORBIS__) + #define MA_ORBIS + #endif + #if defined(__PROSPERO__) + #define MA_PROSPERO + #endif + #if defined(__NX__) + #define MA_NX + #endif + #if defined(__BEOS__) || defined(__HAIKU__) + #define MA_BEOS + #endif + #if defined(__HAIKU__) + #define MA_HAIKU + #endif +#endif + +#if defined(__has_c_attribute) + #if __has_c_attribute(fallthrough) + #define MA_FALLTHROUGH [[fallthrough]] + #endif +#endif +#if !defined(MA_FALLTHROUGH) && defined(__has_attribute) && (defined(__clang__) || defined(__GNUC__)) + #if __has_attribute(fallthrough) + #define MA_FALLTHROUGH __attribute__((fallthrough)) + #endif +#endif +#if !defined(MA_FALLTHROUGH) + #define MA_FALLTHROUGH ((void)0) +#endif + +#ifdef _MSC_VER + #define MA_INLINE __forceinline + + /* noinline was introduced in Visual Studio 2005. */ + #if _MSC_VER >= 1400 + #define MA_NO_INLINE __declspec(noinline) + #else + #define MA_NO_INLINE + #endif +#elif defined(__GNUC__) + /* + I've had a bug report where GCC is emitting warnings about functions possibly not being inlineable. This warning happens when + the __attribute__((always_inline)) attribute is defined without an "inline" statement. I think therefore there must be some + case where "__inline__" is not always defined, thus the compiler emitting these warnings. When using -std=c89 or -ansi on the + command line, we cannot use the "inline" keyword and instead need to use "__inline__". In an attempt to work around this issue + I am using "__inline__" only when we're compiling in strict ANSI mode. + */ + #if defined(__STRICT_ANSI__) + #define MA_GNUC_INLINE_HINT __inline__ + #else + #define MA_GNUC_INLINE_HINT inline + #endif + + #if (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 2)) || defined(__clang__) + #define MA_INLINE MA_GNUC_INLINE_HINT __attribute__((always_inline)) + #define MA_NO_INLINE __attribute__((noinline)) + #else + #define MA_INLINE MA_GNUC_INLINE_HINT + #define MA_NO_INLINE __attribute__((noinline)) + #endif +#elif defined(__WATCOMC__) + #define MA_INLINE __inline + #define MA_NO_INLINE +#else + #define MA_INLINE + #define MA_NO_INLINE +#endif + +/* MA_DLL is not officially supported. You're on your own if you want to use this. */ +#if defined(MA_DLL) + #if defined(_WIN32) + #define MA_DLL_IMPORT __declspec(dllimport) + #define MA_DLL_EXPORT __declspec(dllexport) + #define MA_DLL_PRIVATE static + #else + #if defined(__GNUC__) && __GNUC__ >= 4 + #define MA_DLL_IMPORT __attribute__((visibility("default"))) + #define MA_DLL_EXPORT __attribute__((visibility("default"))) + #define MA_DLL_PRIVATE __attribute__((visibility("hidden"))) + #else + #define MA_DLL_IMPORT + #define MA_DLL_EXPORT + #define MA_DLL_PRIVATE static + #endif + #endif +#endif + +#if !defined(MA_API) + #if defined(MA_DLL) + #if defined(MINIAUDIO_IMPLEMENTATION) || defined(MA_IMPLEMENTATION) + #define MA_API MA_DLL_EXPORT + #else + #define MA_API MA_DLL_IMPORT + #endif + #else + #define MA_API extern + #endif +#endif + +#if !defined(MA_STATIC) + #if defined(MA_DLL) + #define MA_PRIVATE MA_DLL_PRIVATE + #else + #define MA_PRIVATE static + #endif +#endif + + +/* SIMD alignment in bytes. Currently set to 32 bytes in preparation for future AVX optimizations. */ +#define MA_SIMD_ALIGNMENT 32 + +/* +Special wchar_t type to ensure any structures in the public sections that reference it have a +consistent size across all platforms. + +On Windows, wchar_t is 2 bytes, whereas everywhere else it's 4 bytes. Since Windows likes to use +wchar_t for its IDs, we need a special explicitly sized wchar type that is always 2 bytes on all +platforms. +*/ +#if !defined(MA_POSIX) && defined(MA_WIN32) +typedef wchar_t ma_wchar_win32; +#else +typedef ma_uint16 ma_wchar_win32; +#endif + + + +/* +Logging Levels +============== +Log levels are only used to give logging callbacks some context as to the severity of a log message +so they can do filtering. All log levels will be posted to registered logging callbacks. If you +don't want to output a certain log level you can discriminate against the log level in the callback. + +MA_LOG_LEVEL_DEBUG + Used for debugging. Useful for debug and test builds, but should be disabled in release builds. + +MA_LOG_LEVEL_INFO + Informational logging. Useful for debugging. This will never be called from within the data + callback. + +MA_LOG_LEVEL_WARNING + Warnings. You should enable this in you development builds and action them when encountered. These + logs usually indicate a potential problem or misconfiguration, but still allow you to keep + running. This will never be called from within the data callback. + +MA_LOG_LEVEL_ERROR + Error logging. This will be fired when an operation fails and is subsequently aborted. This can + be fired from within the data callback, in which case the device will be stopped. You should + always have this log level enabled. +*/ +typedef enum +{ + MA_LOG_LEVEL_DEBUG = 4, + MA_LOG_LEVEL_INFO = 3, + MA_LOG_LEVEL_WARNING = 2, + MA_LOG_LEVEL_ERROR = 1 +} ma_log_level; + +/* +Variables needing to be accessed atomically should be declared with this macro for two reasons: + + 1) It allows people who read the code to identify a variable as such; and + 2) It forces alignment on platforms where it's required or optimal. + +Note that for x86/64, alignment is not strictly necessary, but does have some performance +implications. Where supported by the compiler, alignment will be used, but otherwise if the CPU +architecture does not require it, it will simply leave it unaligned. This is the case with old +versions of Visual Studio, which I've confirmed with at least VC6. +*/ +#if !defined(_MSC_VER) && defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L) + #include + #define MA_ATOMIC(alignment, type) _Alignas(alignment) type +#else + #if defined(__GNUC__) + /* GCC-style compilers. */ + #define MA_ATOMIC(alignment, type) type __attribute__((aligned(alignment))) + #elif defined(_MSC_VER) && _MSC_VER > 1200 /* 1200 = VC6. Alignment not supported, but not necessary because x86 is the only supported target. */ + /* MSVC. */ + #define MA_ATOMIC(alignment, type) __declspec(align(alignment)) type + #else + /* Other compilers. */ + #define MA_ATOMIC(alignment, type) type + #endif +#endif + +typedef struct ma_context ma_context; +typedef struct ma_device ma_device; + +typedef ma_uint8 ma_channel; +typedef enum +{ + MA_CHANNEL_NONE = 0, + MA_CHANNEL_MONO = 1, + MA_CHANNEL_FRONT_LEFT = 2, + MA_CHANNEL_FRONT_RIGHT = 3, + MA_CHANNEL_FRONT_CENTER = 4, + MA_CHANNEL_LFE = 5, + MA_CHANNEL_BACK_LEFT = 6, + MA_CHANNEL_BACK_RIGHT = 7, + MA_CHANNEL_FRONT_LEFT_CENTER = 8, + MA_CHANNEL_FRONT_RIGHT_CENTER = 9, + MA_CHANNEL_BACK_CENTER = 10, + MA_CHANNEL_SIDE_LEFT = 11, + MA_CHANNEL_SIDE_RIGHT = 12, + MA_CHANNEL_TOP_CENTER = 13, + MA_CHANNEL_TOP_FRONT_LEFT = 14, + MA_CHANNEL_TOP_FRONT_CENTER = 15, + MA_CHANNEL_TOP_FRONT_RIGHT = 16, + MA_CHANNEL_TOP_BACK_LEFT = 17, + MA_CHANNEL_TOP_BACK_CENTER = 18, + MA_CHANNEL_TOP_BACK_RIGHT = 19, + MA_CHANNEL_AUX_0 = 20, + MA_CHANNEL_AUX_1 = 21, + MA_CHANNEL_AUX_2 = 22, + MA_CHANNEL_AUX_3 = 23, + MA_CHANNEL_AUX_4 = 24, + MA_CHANNEL_AUX_5 = 25, + MA_CHANNEL_AUX_6 = 26, + MA_CHANNEL_AUX_7 = 27, + MA_CHANNEL_AUX_8 = 28, + MA_CHANNEL_AUX_9 = 29, + MA_CHANNEL_AUX_10 = 30, + MA_CHANNEL_AUX_11 = 31, + MA_CHANNEL_AUX_12 = 32, + MA_CHANNEL_AUX_13 = 33, + MA_CHANNEL_AUX_14 = 34, + MA_CHANNEL_AUX_15 = 35, + MA_CHANNEL_AUX_16 = 36, + MA_CHANNEL_AUX_17 = 37, + MA_CHANNEL_AUX_18 = 38, + MA_CHANNEL_AUX_19 = 39, + MA_CHANNEL_AUX_20 = 40, + MA_CHANNEL_AUX_21 = 41, + MA_CHANNEL_AUX_22 = 42, + MA_CHANNEL_AUX_23 = 43, + MA_CHANNEL_AUX_24 = 44, + MA_CHANNEL_AUX_25 = 45, + MA_CHANNEL_AUX_26 = 46, + MA_CHANNEL_AUX_27 = 47, + MA_CHANNEL_AUX_28 = 48, + MA_CHANNEL_AUX_29 = 49, + MA_CHANNEL_AUX_30 = 50, + MA_CHANNEL_AUX_31 = 51, + MA_CHANNEL_LEFT = MA_CHANNEL_FRONT_LEFT, + MA_CHANNEL_RIGHT = MA_CHANNEL_FRONT_RIGHT, + MA_CHANNEL_POSITION_COUNT = (MA_CHANNEL_AUX_31 + 1) +} _ma_channel_position; /* Do not use `_ma_channel_position` directly. Use `ma_channel` instead. */ + +typedef enum +{ + MA_SUCCESS = 0, + MA_ERROR = -1, /* A generic error. */ + MA_INVALID_ARGS = -2, + MA_INVALID_OPERATION = -3, + MA_OUT_OF_MEMORY = -4, + MA_OUT_OF_RANGE = -5, + MA_ACCESS_DENIED = -6, + MA_DOES_NOT_EXIST = -7, + MA_ALREADY_EXISTS = -8, + MA_TOO_MANY_OPEN_FILES = -9, + MA_INVALID_FILE = -10, + MA_TOO_BIG = -11, + MA_PATH_TOO_LONG = -12, + MA_NAME_TOO_LONG = -13, + MA_NOT_DIRECTORY = -14, + MA_IS_DIRECTORY = -15, + MA_DIRECTORY_NOT_EMPTY = -16, + MA_AT_END = -17, + MA_NO_SPACE = -18, + MA_BUSY = -19, + MA_IO_ERROR = -20, + MA_INTERRUPT = -21, + MA_UNAVAILABLE = -22, + MA_ALREADY_IN_USE = -23, + MA_BAD_ADDRESS = -24, + MA_BAD_SEEK = -25, + MA_BAD_PIPE = -26, + MA_DEADLOCK = -27, + MA_TOO_MANY_LINKS = -28, + MA_NOT_IMPLEMENTED = -29, + MA_NO_MESSAGE = -30, + MA_BAD_MESSAGE = -31, + MA_NO_DATA_AVAILABLE = -32, + MA_INVALID_DATA = -33, + MA_TIMEOUT = -34, + MA_NO_NETWORK = -35, + MA_NOT_UNIQUE = -36, + MA_NOT_SOCKET = -37, + MA_NO_ADDRESS = -38, + MA_BAD_PROTOCOL = -39, + MA_PROTOCOL_UNAVAILABLE = -40, + MA_PROTOCOL_NOT_SUPPORTED = -41, + MA_PROTOCOL_FAMILY_NOT_SUPPORTED = -42, + MA_ADDRESS_FAMILY_NOT_SUPPORTED = -43, + MA_SOCKET_NOT_SUPPORTED = -44, + MA_CONNECTION_RESET = -45, + MA_ALREADY_CONNECTED = -46, + MA_NOT_CONNECTED = -47, + MA_CONNECTION_REFUSED = -48, + MA_NO_HOST = -49, + MA_IN_PROGRESS = -50, + MA_CANCELLED = -51, + MA_MEMORY_ALREADY_MAPPED = -52, + + /* General non-standard errors. */ + MA_CRC_MISMATCH = -100, + + /* General miniaudio-specific errors. */ + MA_FORMAT_NOT_SUPPORTED = -200, + MA_DEVICE_TYPE_NOT_SUPPORTED = -201, + MA_SHARE_MODE_NOT_SUPPORTED = -202, + MA_NO_BACKEND = -203, + MA_NO_DEVICE = -204, + MA_API_NOT_FOUND = -205, + MA_INVALID_DEVICE_CONFIG = -206, + MA_LOOP = -207, + MA_BACKEND_NOT_ENABLED = -208, + + /* State errors. */ + MA_DEVICE_NOT_INITIALIZED = -300, + MA_DEVICE_ALREADY_INITIALIZED = -301, + MA_DEVICE_NOT_STARTED = -302, + MA_DEVICE_NOT_STOPPED = -303, + + /* Operation errors. */ + MA_FAILED_TO_INIT_BACKEND = -400, + MA_FAILED_TO_OPEN_BACKEND_DEVICE = -401, + MA_FAILED_TO_START_BACKEND_DEVICE = -402, + MA_FAILED_TO_STOP_BACKEND_DEVICE = -403 +} ma_result; + + +#define MA_MIN_CHANNELS 1 +#ifndef MA_MAX_CHANNELS +#define MA_MAX_CHANNELS 254 +#endif + +#ifndef MA_MAX_FILTER_ORDER +#define MA_MAX_FILTER_ORDER 8 +#endif + +typedef enum +{ + ma_stream_format_pcm = 0 +} ma_stream_format; + +typedef enum +{ + ma_stream_layout_interleaved = 0, + ma_stream_layout_deinterleaved +} ma_stream_layout; + +typedef enum +{ + ma_dither_mode_none = 0, + ma_dither_mode_rectangle, + ma_dither_mode_triangle +} ma_dither_mode; + +typedef enum +{ + /* + I like to keep these explicitly defined because they're used as a key into a lookup table. When items are + added to this, make sure there are no gaps and that they're added to the lookup table in ma_get_bytes_per_sample(). + */ + ma_format_unknown = 0, /* Mainly used for indicating an error, but also used as the default for the output format for decoders. */ + ma_format_u8 = 1, + ma_format_s16 = 2, /* Seems to be the most widely supported format. */ + ma_format_s24 = 3, /* Tightly packed. 3 bytes per sample. */ + ma_format_s32 = 4, + ma_format_f32 = 5, + ma_format_count +} ma_format; + +typedef enum +{ + /* Standard rates need to be in priority order. */ + ma_standard_sample_rate_48000 = 48000, /* Most common */ + ma_standard_sample_rate_44100 = 44100, + + ma_standard_sample_rate_32000 = 32000, /* Lows */ + ma_standard_sample_rate_24000 = 24000, + ma_standard_sample_rate_22050 = 22050, + + ma_standard_sample_rate_88200 = 88200, /* Highs */ + ma_standard_sample_rate_96000 = 96000, + ma_standard_sample_rate_176400 = 176400, + ma_standard_sample_rate_192000 = 192000, + + ma_standard_sample_rate_16000 = 16000, /* Extreme lows */ + ma_standard_sample_rate_11025 = 11025, + ma_standard_sample_rate_8000 = 8000, + + ma_standard_sample_rate_352800 = 352800, /* Extreme highs */ + ma_standard_sample_rate_384000 = 384000, + + ma_standard_sample_rate_min = ma_standard_sample_rate_8000, + ma_standard_sample_rate_max = ma_standard_sample_rate_384000, + ma_standard_sample_rate_count = 14 /* Need to maintain the count manually. Make sure this is updated if items are added to enum. */ +} ma_standard_sample_rate; + + +typedef enum +{ + ma_channel_mix_mode_rectangular = 0, /* Simple averaging based on the plane(s) the channel is sitting on. */ + ma_channel_mix_mode_simple, /* Drop excess channels; zeroed out extra channels. */ + ma_channel_mix_mode_custom_weights, /* Use custom weights specified in ma_channel_converter_config. */ + ma_channel_mix_mode_default = ma_channel_mix_mode_rectangular +} ma_channel_mix_mode; + +typedef enum +{ + ma_standard_channel_map_microsoft, + ma_standard_channel_map_alsa, + ma_standard_channel_map_rfc3551, /* Based off AIFF. */ + ma_standard_channel_map_flac, + ma_standard_channel_map_vorbis, + ma_standard_channel_map_sound4, /* FreeBSD's sound(4). */ + ma_standard_channel_map_sndio, /* www.sndio.org/tips.html */ + ma_standard_channel_map_webaudio = ma_standard_channel_map_flac, /* https://webaudio.github.io/web-audio-api/#ChannelOrdering. Only 1, 2, 4 and 6 channels are defined, but can fill in the gaps with logical assumptions. */ + ma_standard_channel_map_default = ma_standard_channel_map_microsoft +} ma_standard_channel_map; + +typedef enum +{ + ma_performance_profile_low_latency = 0, + ma_performance_profile_conservative +} ma_performance_profile; + + +typedef struct +{ + void* pUserData; + void* (* onMalloc)(size_t sz, void* pUserData); + void* (* onRealloc)(void* p, size_t sz, void* pUserData); + void (* onFree)(void* p, void* pUserData); +} ma_allocation_callbacks; + +typedef struct +{ + ma_int32 state; +} ma_lcg; + + +/* +Atomics. + +These are typesafe structures to prevent errors as a result of forgetting to reference variables atomically. It's too +easy to introduce subtle bugs where you accidentally do a regular assignment instead of an atomic load/store, etc. By +using a struct we can enforce the use of atomics at compile time. + +These types are declared in the header section because we need to reference them in structs below, but functions for +using them are only exposed in the implementation section. I do not want these to be part of the public API. + +There's a few downsides to this system. The first is that you need to declare a new struct for each type. Below are +some macros to help with the declarations. They will be named like so: + + ma_atomic_uint32 - atomic ma_uint32 + ma_atomic_int32 - atomic ma_int32 + ma_atomic_uint64 - atomic ma_uint64 + ma_atomic_float - atomic float + ma_atomic_bool32 - atomic ma_bool32 + +The other downside is that atomic pointers are extremely messy. You need to declare a new struct for each specific +type of pointer you need to make atomic. For example, an atomic ma_node* will look like this: + + MA_ATOMIC_SAFE_TYPE_IMPL_PTR(node) + +Which will declare a type struct that's named like so: + + ma_atomic_ptr_node + +Functions to use the atomic types are declared in the implementation section. All atomic functions are prefixed with +the name of the struct. For example: + + ma_atomic_uint32_set() - Atomic store of ma_uint32 + ma_atomic_uint32_get() - Atomic load of ma_uint32 + etc. + +For pointer types it's the same, which makes them a bit messy to use due to the length of each function name, but in +return you get type safety and enforcement of atomic operations. +*/ +#define MA_ATOMIC_SAFE_TYPE_DECL(c89TypeExtension, typeSize, type) \ + typedef struct \ + { \ + MA_ATOMIC(typeSize, ma_##type) value; \ + } ma_atomic_##type; \ + +#define MA_ATOMIC_SAFE_TYPE_DECL_PTR(type) \ + typedef struct \ + { \ + MA_ATOMIC(MA_SIZEOF_PTR, ma_##type*) value; \ + } ma_atomic_ptr_##type; \ + +MA_ATOMIC_SAFE_TYPE_DECL(32, 4, uint32) +MA_ATOMIC_SAFE_TYPE_DECL(i32, 4, int32) +MA_ATOMIC_SAFE_TYPE_DECL(64, 8, uint64) +MA_ATOMIC_SAFE_TYPE_DECL(f32, 4, float) +MA_ATOMIC_SAFE_TYPE_DECL(32, 4, bool32) + + +/* Spinlocks are 32-bit for compatibility reasons. */ +typedef ma_uint32 ma_spinlock; + +#ifndef MA_NO_THREADING + /* Thread priorities should be ordered such that the default priority of the worker thread is 0. */ + typedef enum + { + ma_thread_priority_idle = -5, + ma_thread_priority_lowest = -4, + ma_thread_priority_low = -3, + ma_thread_priority_normal = -2, + ma_thread_priority_high = -1, + ma_thread_priority_highest = 0, + ma_thread_priority_realtime = 1, + ma_thread_priority_default = 0 + } ma_thread_priority; + + #if defined(MA_POSIX) + typedef ma_pthread_t ma_thread; + #elif defined(MA_WIN32) + typedef ma_handle ma_thread; + #endif + + #if defined(MA_POSIX) + typedef ma_pthread_mutex_t ma_mutex; + #elif defined(MA_WIN32) + typedef ma_handle ma_mutex; + #endif + + #if defined(MA_POSIX) + typedef struct + { + ma_uint32 value; + ma_pthread_mutex_t lock; + ma_pthread_cond_t cond; + } ma_event; + #elif defined(MA_WIN32) + typedef ma_handle ma_event; + #endif + + #if defined(MA_POSIX) + typedef struct + { + int value; + ma_pthread_mutex_t lock; + ma_pthread_cond_t cond; + } ma_semaphore; + #elif defined(MA_WIN32) + typedef ma_handle ma_semaphore; + #endif +#else + /* MA_NO_THREADING is set which means threading is disabled. Threading is required by some API families. If any of these are enabled we need to throw an error. */ + #ifndef MA_NO_DEVICE_IO + #error "MA_NO_THREADING cannot be used without MA_NO_DEVICE_IO"; + #endif +#endif /* MA_NO_THREADING */ + + +/* +Retrieves the version of miniaudio as separated integers. Each component can be NULL if it's not required. +*/ +MA_API void ma_version(ma_uint32* pMajor, ma_uint32* pMinor, ma_uint32* pRevision); + +/* +Retrieves the version of miniaudio as a string which can be useful for logging purposes. +*/ +MA_API const char* ma_version_string(void); + + +/************************************************************************************************************************************************************** + +Logging + +**************************************************************************************************************************************************************/ +#include /* For va_list. */ + +#if defined(__has_attribute) + #if __has_attribute(format) + #define MA_ATTRIBUTE_FORMAT(fmt, va) __attribute__((format(printf, fmt, va))) + #endif +#endif +#ifndef MA_ATTRIBUTE_FORMAT +#define MA_ATTRIBUTE_FORMAT(fmt, va) +#endif + +#ifndef MA_MAX_LOG_CALLBACKS +#define MA_MAX_LOG_CALLBACKS 4 +#endif + + +/* +The callback for handling log messages. + + +Parameters +---------- +pUserData (in) + The user data pointer that was passed into ma_log_register_callback(). + +logLevel (in) + The log level. This can be one of the following: + + +----------------------+ + | Log Level | + +----------------------+ + | MA_LOG_LEVEL_DEBUG | + | MA_LOG_LEVEL_INFO | + | MA_LOG_LEVEL_WARNING | + | MA_LOG_LEVEL_ERROR | + +----------------------+ + +pMessage (in) + The log message. +*/ +typedef void (* ma_log_callback_proc)(void* pUserData, ma_uint32 level, const char* pMessage); + +typedef struct +{ + ma_log_callback_proc onLog; + void* pUserData; +} ma_log_callback; + +MA_API ma_log_callback ma_log_callback_init(ma_log_callback_proc onLog, void* pUserData); + + +typedef struct +{ + ma_log_callback callbacks[MA_MAX_LOG_CALLBACKS]; + ma_uint32 callbackCount; + ma_allocation_callbacks allocationCallbacks; /* Need to store these persistently because ma_log_postv() might need to allocate a buffer on the heap. */ +#ifndef MA_NO_THREADING + ma_mutex lock; /* For thread safety just to make it easier and safer for the logging implementation. */ +#endif +} ma_log; + +MA_API ma_result ma_log_init(const ma_allocation_callbacks* pAllocationCallbacks, ma_log* pLog); +MA_API void ma_log_uninit(ma_log* pLog); +MA_API ma_result ma_log_register_callback(ma_log* pLog, ma_log_callback callback); +MA_API ma_result ma_log_unregister_callback(ma_log* pLog, ma_log_callback callback); +MA_API ma_result ma_log_post(ma_log* pLog, ma_uint32 level, const char* pMessage); +MA_API ma_result ma_log_postv(ma_log* pLog, ma_uint32 level, const char* pFormat, va_list args); +MA_API ma_result ma_log_postf(ma_log* pLog, ma_uint32 level, const char* pFormat, ...) MA_ATTRIBUTE_FORMAT(3, 4); + + +/************************************************************************************************************************************************************** + +Biquad Filtering + +**************************************************************************************************************************************************************/ +typedef union +{ + float f32; + ma_int32 s32; +} ma_biquad_coefficient; + +typedef struct +{ + ma_format format; + ma_uint32 channels; + double b0; + double b1; + double b2; + double a0; + double a1; + double a2; +} ma_biquad_config; + +MA_API ma_biquad_config ma_biquad_config_init(ma_format format, ma_uint32 channels, double b0, double b1, double b2, double a0, double a1, double a2); + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_biquad_coefficient b0; + ma_biquad_coefficient b1; + ma_biquad_coefficient b2; + ma_biquad_coefficient a1; + ma_biquad_coefficient a2; + ma_biquad_coefficient* pR1; + ma_biquad_coefficient* pR2; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_biquad; + +MA_API ma_result ma_biquad_get_heap_size(const ma_biquad_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_biquad_init_preallocated(const ma_biquad_config* pConfig, void* pHeap, ma_biquad* pBQ); +MA_API ma_result ma_biquad_init(const ma_biquad_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_biquad* pBQ); +MA_API void ma_biquad_uninit(ma_biquad* pBQ, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_biquad_reinit(const ma_biquad_config* pConfig, ma_biquad* pBQ); +MA_API ma_result ma_biquad_clear_cache(ma_biquad* pBQ); +MA_API ma_result ma_biquad_process_pcm_frames(ma_biquad* pBQ, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_biquad_get_latency(const ma_biquad* pBQ); + + +/************************************************************************************************************************************************************** + +Low-Pass Filtering + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double cutoffFrequency; + double q; +} ma_lpf1_config, ma_lpf2_config; + +MA_API ma_lpf1_config ma_lpf1_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency); +MA_API ma_lpf2_config ma_lpf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, double q); + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_biquad_coefficient a; + ma_biquad_coefficient* pR1; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_lpf1; + +MA_API ma_result ma_lpf1_get_heap_size(const ma_lpf1_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_lpf1_init_preallocated(const ma_lpf1_config* pConfig, void* pHeap, ma_lpf1* pLPF); +MA_API ma_result ma_lpf1_init(const ma_lpf1_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_lpf1* pLPF); +MA_API void ma_lpf1_uninit(ma_lpf1* pLPF, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_lpf1_reinit(const ma_lpf1_config* pConfig, ma_lpf1* pLPF); +MA_API ma_result ma_lpf1_clear_cache(ma_lpf1* pLPF); +MA_API ma_result ma_lpf1_process_pcm_frames(ma_lpf1* pLPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_lpf1_get_latency(const ma_lpf1* pLPF); + +typedef struct +{ + ma_biquad bq; /* The second order low-pass filter is implemented as a biquad filter. */ +} ma_lpf2; + +MA_API ma_result ma_lpf2_get_heap_size(const ma_lpf2_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_lpf2_init_preallocated(const ma_lpf2_config* pConfig, void* pHeap, ma_lpf2* pHPF); +MA_API ma_result ma_lpf2_init(const ma_lpf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_lpf2* pLPF); +MA_API void ma_lpf2_uninit(ma_lpf2* pLPF, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_lpf2_reinit(const ma_lpf2_config* pConfig, ma_lpf2* pLPF); +MA_API ma_result ma_lpf2_clear_cache(ma_lpf2* pLPF); +MA_API ma_result ma_lpf2_process_pcm_frames(ma_lpf2* pLPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_lpf2_get_latency(const ma_lpf2* pLPF); + + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double cutoffFrequency; + ma_uint32 order; /* If set to 0, will be treated as a passthrough (no filtering will be applied). */ +} ma_lpf_config; + +MA_API ma_lpf_config ma_lpf_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order); + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_uint32 lpf1Count; + ma_uint32 lpf2Count; + ma_lpf1* pLPF1; + ma_lpf2* pLPF2; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_lpf; + +MA_API ma_result ma_lpf_get_heap_size(const ma_lpf_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_lpf_init_preallocated(const ma_lpf_config* pConfig, void* pHeap, ma_lpf* pLPF); +MA_API ma_result ma_lpf_init(const ma_lpf_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_lpf* pLPF); +MA_API void ma_lpf_uninit(ma_lpf* pLPF, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_lpf_reinit(const ma_lpf_config* pConfig, ma_lpf* pLPF); +MA_API ma_result ma_lpf_clear_cache(ma_lpf* pLPF); +MA_API ma_result ma_lpf_process_pcm_frames(ma_lpf* pLPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_lpf_get_latency(const ma_lpf* pLPF); + + +/************************************************************************************************************************************************************** + +High-Pass Filtering + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double cutoffFrequency; + double q; +} ma_hpf1_config, ma_hpf2_config; + +MA_API ma_hpf1_config ma_hpf1_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency); +MA_API ma_hpf2_config ma_hpf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, double q); + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_biquad_coefficient a; + ma_biquad_coefficient* pR1; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_hpf1; + +MA_API ma_result ma_hpf1_get_heap_size(const ma_hpf1_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_hpf1_init_preallocated(const ma_hpf1_config* pConfig, void* pHeap, ma_hpf1* pLPF); +MA_API ma_result ma_hpf1_init(const ma_hpf1_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hpf1* pHPF); +MA_API void ma_hpf1_uninit(ma_hpf1* pHPF, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_hpf1_reinit(const ma_hpf1_config* pConfig, ma_hpf1* pHPF); +MA_API ma_result ma_hpf1_process_pcm_frames(ma_hpf1* pHPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_hpf1_get_latency(const ma_hpf1* pHPF); + +typedef struct +{ + ma_biquad bq; /* The second order high-pass filter is implemented as a biquad filter. */ +} ma_hpf2; + +MA_API ma_result ma_hpf2_get_heap_size(const ma_hpf2_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_hpf2_init_preallocated(const ma_hpf2_config* pConfig, void* pHeap, ma_hpf2* pHPF); +MA_API ma_result ma_hpf2_init(const ma_hpf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hpf2* pHPF); +MA_API void ma_hpf2_uninit(ma_hpf2* pHPF, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_hpf2_reinit(const ma_hpf2_config* pConfig, ma_hpf2* pHPF); +MA_API ma_result ma_hpf2_process_pcm_frames(ma_hpf2* pHPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_hpf2_get_latency(const ma_hpf2* pHPF); + + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double cutoffFrequency; + ma_uint32 order; /* If set to 0, will be treated as a passthrough (no filtering will be applied). */ +} ma_hpf_config; + +MA_API ma_hpf_config ma_hpf_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order); + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_uint32 hpf1Count; + ma_uint32 hpf2Count; + ma_hpf1* pHPF1; + ma_hpf2* pHPF2; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_hpf; + +MA_API ma_result ma_hpf_get_heap_size(const ma_hpf_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_hpf_init_preallocated(const ma_hpf_config* pConfig, void* pHeap, ma_hpf* pLPF); +MA_API ma_result ma_hpf_init(const ma_hpf_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hpf* pHPF); +MA_API void ma_hpf_uninit(ma_hpf* pHPF, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_hpf_reinit(const ma_hpf_config* pConfig, ma_hpf* pHPF); +MA_API ma_result ma_hpf_process_pcm_frames(ma_hpf* pHPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_hpf_get_latency(const ma_hpf* pHPF); + + +/************************************************************************************************************************************************************** + +Band-Pass Filtering + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double cutoffFrequency; + double q; +} ma_bpf2_config; + +MA_API ma_bpf2_config ma_bpf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, double q); + +typedef struct +{ + ma_biquad bq; /* The second order band-pass filter is implemented as a biquad filter. */ +} ma_bpf2; + +MA_API ma_result ma_bpf2_get_heap_size(const ma_bpf2_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_bpf2_init_preallocated(const ma_bpf2_config* pConfig, void* pHeap, ma_bpf2* pBPF); +MA_API ma_result ma_bpf2_init(const ma_bpf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_bpf2* pBPF); +MA_API void ma_bpf2_uninit(ma_bpf2* pBPF, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_bpf2_reinit(const ma_bpf2_config* pConfig, ma_bpf2* pBPF); +MA_API ma_result ma_bpf2_process_pcm_frames(ma_bpf2* pBPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_bpf2_get_latency(const ma_bpf2* pBPF); + + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double cutoffFrequency; + ma_uint32 order; /* If set to 0, will be treated as a passthrough (no filtering will be applied). */ +} ma_bpf_config; + +MA_API ma_bpf_config ma_bpf_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order); + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 bpf2Count; + ma_bpf2* pBPF2; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_bpf; + +MA_API ma_result ma_bpf_get_heap_size(const ma_bpf_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_bpf_init_preallocated(const ma_bpf_config* pConfig, void* pHeap, ma_bpf* pBPF); +MA_API ma_result ma_bpf_init(const ma_bpf_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_bpf* pBPF); +MA_API void ma_bpf_uninit(ma_bpf* pBPF, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_bpf_reinit(const ma_bpf_config* pConfig, ma_bpf* pBPF); +MA_API ma_result ma_bpf_process_pcm_frames(ma_bpf* pBPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_bpf_get_latency(const ma_bpf* pBPF); + + +/************************************************************************************************************************************************************** + +Notching Filter + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double q; + double frequency; +} ma_notch2_config, ma_notch_config; + +MA_API ma_notch2_config ma_notch2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double q, double frequency); + +typedef struct +{ + ma_biquad bq; +} ma_notch2; + +MA_API ma_result ma_notch2_get_heap_size(const ma_notch2_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_notch2_init_preallocated(const ma_notch2_config* pConfig, void* pHeap, ma_notch2* pFilter); +MA_API ma_result ma_notch2_init(const ma_notch2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_notch2* pFilter); +MA_API void ma_notch2_uninit(ma_notch2* pFilter, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_notch2_reinit(const ma_notch2_config* pConfig, ma_notch2* pFilter); +MA_API ma_result ma_notch2_process_pcm_frames(ma_notch2* pFilter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_notch2_get_latency(const ma_notch2* pFilter); + + +/************************************************************************************************************************************************************** + +Peaking EQ Filter + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double gainDB; + double q; + double frequency; +} ma_peak2_config, ma_peak_config; + +MA_API ma_peak2_config ma_peak2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double q, double frequency); + +typedef struct +{ + ma_biquad bq; +} ma_peak2; + +MA_API ma_result ma_peak2_get_heap_size(const ma_peak2_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_peak2_init_preallocated(const ma_peak2_config* pConfig, void* pHeap, ma_peak2* pFilter); +MA_API ma_result ma_peak2_init(const ma_peak2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_peak2* pFilter); +MA_API void ma_peak2_uninit(ma_peak2* pFilter, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_peak2_reinit(const ma_peak2_config* pConfig, ma_peak2* pFilter); +MA_API ma_result ma_peak2_process_pcm_frames(ma_peak2* pFilter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_peak2_get_latency(const ma_peak2* pFilter); + + +/************************************************************************************************************************************************************** + +Low Shelf Filter + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double gainDB; + double shelfSlope; + double frequency; +} ma_loshelf2_config, ma_loshelf_config; + +MA_API ma_loshelf2_config ma_loshelf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double shelfSlope, double frequency); + +typedef struct +{ + ma_biquad bq; +} ma_loshelf2; + +MA_API ma_result ma_loshelf2_get_heap_size(const ma_loshelf2_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_loshelf2_init_preallocated(const ma_loshelf2_config* pConfig, void* pHeap, ma_loshelf2* pFilter); +MA_API ma_result ma_loshelf2_init(const ma_loshelf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_loshelf2* pFilter); +MA_API void ma_loshelf2_uninit(ma_loshelf2* pFilter, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_loshelf2_reinit(const ma_loshelf2_config* pConfig, ma_loshelf2* pFilter); +MA_API ma_result ma_loshelf2_process_pcm_frames(ma_loshelf2* pFilter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_loshelf2_get_latency(const ma_loshelf2* pFilter); + + +/************************************************************************************************************************************************************** + +High Shelf Filter + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double gainDB; + double shelfSlope; + double frequency; +} ma_hishelf2_config, ma_hishelf_config; + +MA_API ma_hishelf2_config ma_hishelf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double shelfSlope, double frequency); + +typedef struct +{ + ma_biquad bq; +} ma_hishelf2; + +MA_API ma_result ma_hishelf2_get_heap_size(const ma_hishelf2_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_hishelf2_init_preallocated(const ma_hishelf2_config* pConfig, void* pHeap, ma_hishelf2* pFilter); +MA_API ma_result ma_hishelf2_init(const ma_hishelf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hishelf2* pFilter); +MA_API void ma_hishelf2_uninit(ma_hishelf2* pFilter, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_hishelf2_reinit(const ma_hishelf2_config* pConfig, ma_hishelf2* pFilter); +MA_API ma_result ma_hishelf2_process_pcm_frames(ma_hishelf2* pFilter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_uint32 ma_hishelf2_get_latency(const ma_hishelf2* pFilter); + + + +/* +Delay +*/ +typedef struct +{ + ma_uint32 channels; + ma_uint32 sampleRate; + ma_uint32 delayInFrames; + ma_bool32 delayStart; /* Set to true to delay the start of the output; false otherwise. */ + float wet; /* 0..1. Default = 1. */ + float dry; /* 0..1. Default = 1. */ + float decay; /* 0..1. Default = 0 (no feedback). Feedback decay. Use this for echo. */ +} ma_delay_config; + +MA_API ma_delay_config ma_delay_config_init(ma_uint32 channels, ma_uint32 sampleRate, ma_uint32 delayInFrames, float decay); + + +typedef struct +{ + ma_delay_config config; + ma_uint32 cursor; /* Feedback is written to this cursor. Always equal or in front of the read cursor. */ + ma_uint32 bufferSizeInFrames; + float* pBuffer; +} ma_delay; + +MA_API ma_result ma_delay_init(const ma_delay_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_delay* pDelay); +MA_API void ma_delay_uninit(ma_delay* pDelay, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_delay_process_pcm_frames(ma_delay* pDelay, void* pFramesOut, const void* pFramesIn, ma_uint32 frameCount); +MA_API void ma_delay_set_wet(ma_delay* pDelay, float value); +MA_API float ma_delay_get_wet(const ma_delay* pDelay); +MA_API void ma_delay_set_dry(ma_delay* pDelay, float value); +MA_API float ma_delay_get_dry(const ma_delay* pDelay); +MA_API void ma_delay_set_decay(ma_delay* pDelay, float value); +MA_API float ma_delay_get_decay(const ma_delay* pDelay); + + +/* Gainer for smooth volume changes. */ +typedef struct +{ + ma_uint32 channels; + ma_uint32 smoothTimeInFrames; +} ma_gainer_config; + +MA_API ma_gainer_config ma_gainer_config_init(ma_uint32 channels, ma_uint32 smoothTimeInFrames); + + +typedef struct +{ + ma_gainer_config config; + ma_uint32 t; + float masterVolume; + float* pOldGains; + float* pNewGains; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_gainer; + +MA_API ma_result ma_gainer_get_heap_size(const ma_gainer_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_gainer_init_preallocated(const ma_gainer_config* pConfig, void* pHeap, ma_gainer* pGainer); +MA_API ma_result ma_gainer_init(const ma_gainer_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_gainer* pGainer); +MA_API void ma_gainer_uninit(ma_gainer* pGainer, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_gainer_process_pcm_frames(ma_gainer* pGainer, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_result ma_gainer_set_gain(ma_gainer* pGainer, float newGain); +MA_API ma_result ma_gainer_set_gains(ma_gainer* pGainer, float* pNewGains); +MA_API ma_result ma_gainer_set_master_volume(ma_gainer* pGainer, float volume); +MA_API ma_result ma_gainer_get_master_volume(const ma_gainer* pGainer, float* pVolume); + + + +/* Stereo panner. */ +typedef enum +{ + ma_pan_mode_balance = 0, /* Does not blend one side with the other. Technically just a balance. Compatible with other popular audio engines and therefore the default. */ + ma_pan_mode_pan /* A true pan. The sound from one side will "move" to the other side and blend with it. */ +} ma_pan_mode; + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_pan_mode mode; + float pan; +} ma_panner_config; + +MA_API ma_panner_config ma_panner_config_init(ma_format format, ma_uint32 channels); + + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_pan_mode mode; + float pan; /* -1..1 where 0 is no pan, -1 is left side, +1 is right side. Defaults to 0. */ +} ma_panner; + +MA_API ma_result ma_panner_init(const ma_panner_config* pConfig, ma_panner* pPanner); +MA_API ma_result ma_panner_process_pcm_frames(ma_panner* pPanner, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API void ma_panner_set_mode(ma_panner* pPanner, ma_pan_mode mode); +MA_API ma_pan_mode ma_panner_get_mode(const ma_panner* pPanner); +MA_API void ma_panner_set_pan(ma_panner* pPanner, float pan); +MA_API float ma_panner_get_pan(const ma_panner* pPanner); + + + +/* Fader. */ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; +} ma_fader_config; + +MA_API ma_fader_config ma_fader_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate); + +typedef struct +{ + ma_fader_config config; + float volumeBeg; /* If volumeBeg and volumeEnd is equal to 1, no fading happens (ma_fader_process_pcm_frames() will run as a passthrough). */ + float volumeEnd; + ma_uint64 lengthInFrames; /* The total length of the fade. */ + ma_int64 cursorInFrames; /* The current time in frames. Incremented by ma_fader_process_pcm_frames(). Signed because it'll be offset by startOffsetInFrames in set_fade_ex(). */ +} ma_fader; + +MA_API ma_result ma_fader_init(const ma_fader_config* pConfig, ma_fader* pFader); +MA_API ma_result ma_fader_process_pcm_frames(ma_fader* pFader, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API void ma_fader_get_data_format(const ma_fader* pFader, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate); +MA_API void ma_fader_set_fade(ma_fader* pFader, float volumeBeg, float volumeEnd, ma_uint64 lengthInFrames); +MA_API void ma_fader_set_fade_ex(ma_fader* pFader, float volumeBeg, float volumeEnd, ma_uint64 lengthInFrames, ma_int64 startOffsetInFrames); +MA_API float ma_fader_get_current_volume(const ma_fader* pFader); + + + +/* Spatializer. */ +typedef struct +{ + float x; + float y; + float z; +} ma_vec3f; + +typedef struct +{ + ma_vec3f v; + ma_spinlock lock; +} ma_atomic_vec3f; + +typedef enum +{ + ma_attenuation_model_none, /* No distance attenuation and no spatialization. */ + ma_attenuation_model_inverse, /* Equivalent to OpenAL's AL_INVERSE_DISTANCE_CLAMPED. */ + ma_attenuation_model_linear, /* Linear attenuation. Equivalent to OpenAL's AL_LINEAR_DISTANCE_CLAMPED. */ + ma_attenuation_model_exponential /* Exponential attenuation. Equivalent to OpenAL's AL_EXPONENT_DISTANCE_CLAMPED. */ +} ma_attenuation_model; + +typedef enum +{ + ma_positioning_absolute, + ma_positioning_relative +} ma_positioning; + +typedef enum +{ + ma_handedness_right, + ma_handedness_left +} ma_handedness; + + +typedef struct +{ + ma_uint32 channelsOut; + ma_channel* pChannelMapOut; + ma_handedness handedness; /* Defaults to right. Forward is -1 on the Z axis. In a left handed system, forward is +1 on the Z axis. */ + float coneInnerAngleInRadians; + float coneOuterAngleInRadians; + float coneOuterGain; + float speedOfSound; + ma_vec3f worldUp; +} ma_spatializer_listener_config; + +MA_API ma_spatializer_listener_config ma_spatializer_listener_config_init(ma_uint32 channelsOut); + + +typedef struct +{ + ma_spatializer_listener_config config; + ma_atomic_vec3f position; /* The absolute position of the listener. */ + ma_atomic_vec3f direction; /* The direction the listener is facing. The world up vector is config.worldUp. */ + ma_atomic_vec3f velocity; + ma_bool32 isEnabled; + + /* Memory management. */ + ma_bool32 _ownsHeap; + void* _pHeap; +} ma_spatializer_listener; + +MA_API ma_result ma_spatializer_listener_get_heap_size(const ma_spatializer_listener_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_spatializer_listener_init_preallocated(const ma_spatializer_listener_config* pConfig, void* pHeap, ma_spatializer_listener* pListener); +MA_API ma_result ma_spatializer_listener_init(const ma_spatializer_listener_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_spatializer_listener* pListener); +MA_API void ma_spatializer_listener_uninit(ma_spatializer_listener* pListener, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_channel* ma_spatializer_listener_get_channel_map(ma_spatializer_listener* pListener); +MA_API void ma_spatializer_listener_set_cone(ma_spatializer_listener* pListener, float innerAngleInRadians, float outerAngleInRadians, float outerGain); +MA_API void ma_spatializer_listener_get_cone(const ma_spatializer_listener* pListener, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain); +MA_API void ma_spatializer_listener_set_position(ma_spatializer_listener* pListener, float x, float y, float z); +MA_API ma_vec3f ma_spatializer_listener_get_position(const ma_spatializer_listener* pListener); +MA_API void ma_spatializer_listener_set_direction(ma_spatializer_listener* pListener, float x, float y, float z); +MA_API ma_vec3f ma_spatializer_listener_get_direction(const ma_spatializer_listener* pListener); +MA_API void ma_spatializer_listener_set_velocity(ma_spatializer_listener* pListener, float x, float y, float z); +MA_API ma_vec3f ma_spatializer_listener_get_velocity(const ma_spatializer_listener* pListener); +MA_API void ma_spatializer_listener_set_speed_of_sound(ma_spatializer_listener* pListener, float speedOfSound); +MA_API float ma_spatializer_listener_get_speed_of_sound(const ma_spatializer_listener* pListener); +MA_API void ma_spatializer_listener_set_world_up(ma_spatializer_listener* pListener, float x, float y, float z); +MA_API ma_vec3f ma_spatializer_listener_get_world_up(const ma_spatializer_listener* pListener); +MA_API void ma_spatializer_listener_set_enabled(ma_spatializer_listener* pListener, ma_bool32 isEnabled); +MA_API ma_bool32 ma_spatializer_listener_is_enabled(const ma_spatializer_listener* pListener); + + +typedef struct +{ + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_channel* pChannelMapIn; + ma_attenuation_model attenuationModel; + ma_positioning positioning; + ma_handedness handedness; /* Defaults to right. Forward is -1 on the Z axis. In a left handed system, forward is +1 on the Z axis. */ + float minGain; + float maxGain; + float minDistance; + float maxDistance; + float rolloff; + float coneInnerAngleInRadians; + float coneOuterAngleInRadians; + float coneOuterGain; + float dopplerFactor; /* Set to 0 to disable doppler effect. */ + float directionalAttenuationFactor; /* Set to 0 to disable directional attenuation. */ + float minSpatializationChannelGain; /* The minimal scaling factor to apply to channel gains when accounting for the direction of the sound relative to the listener. Must be in the range of 0..1. Smaller values means more aggressive directional panning, larger values means more subtle directional panning. */ + ma_uint32 gainSmoothTimeInFrames; /* When the gain of a channel changes during spatialization, the transition will be linearly interpolated over this number of frames. */ +} ma_spatializer_config; + +MA_API ma_spatializer_config ma_spatializer_config_init(ma_uint32 channelsIn, ma_uint32 channelsOut); + + +typedef struct +{ + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_channel* pChannelMapIn; + ma_attenuation_model attenuationModel; + ma_positioning positioning; + ma_handedness handedness; /* Defaults to right. Forward is -1 on the Z axis. In a left handed system, forward is +1 on the Z axis. */ + float minGain; + float maxGain; + float minDistance; + float maxDistance; + float rolloff; + float coneInnerAngleInRadians; + float coneOuterAngleInRadians; + float coneOuterGain; + float dopplerFactor; /* Set to 0 to disable doppler effect. */ + float directionalAttenuationFactor; /* Set to 0 to disable directional attenuation. */ + ma_uint32 gainSmoothTimeInFrames; /* When the gain of a channel changes during spatialization, the transition will be linearly interpolated over this number of frames. */ + ma_atomic_vec3f position; + ma_atomic_vec3f direction; + ma_atomic_vec3f velocity; /* For doppler effect. */ + float dopplerPitch; /* Will be updated by ma_spatializer_process_pcm_frames() and can be used by higher level functions to apply a pitch shift for doppler effect. */ + float minSpatializationChannelGain; + ma_gainer gainer; /* For smooth gain transitions. */ + float* pNewChannelGainsOut; /* An offset of _pHeap. Used by ma_spatializer_process_pcm_frames() to store new channel gains. The number of elements in this array is equal to config.channelsOut. */ + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_spatializer; + +MA_API ma_result ma_spatializer_get_heap_size(const ma_spatializer_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_spatializer_init_preallocated(const ma_spatializer_config* pConfig, void* pHeap, ma_spatializer* pSpatializer); +MA_API ma_result ma_spatializer_init(const ma_spatializer_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_spatializer* pSpatializer); +MA_API void ma_spatializer_uninit(ma_spatializer* pSpatializer, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_spatializer_process_pcm_frames(ma_spatializer* pSpatializer, ma_spatializer_listener* pListener, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_result ma_spatializer_set_master_volume(ma_spatializer* pSpatializer, float volume); +MA_API ma_result ma_spatializer_get_master_volume(const ma_spatializer* pSpatializer, float* pVolume); +MA_API ma_uint32 ma_spatializer_get_input_channels(const ma_spatializer* pSpatializer); +MA_API ma_uint32 ma_spatializer_get_output_channels(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_attenuation_model(ma_spatializer* pSpatializer, ma_attenuation_model attenuationModel); +MA_API ma_attenuation_model ma_spatializer_get_attenuation_model(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_positioning(ma_spatializer* pSpatializer, ma_positioning positioning); +MA_API ma_positioning ma_spatializer_get_positioning(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_rolloff(ma_spatializer* pSpatializer, float rolloff); +MA_API float ma_spatializer_get_rolloff(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_min_gain(ma_spatializer* pSpatializer, float minGain); +MA_API float ma_spatializer_get_min_gain(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_max_gain(ma_spatializer* pSpatializer, float maxGain); +MA_API float ma_spatializer_get_max_gain(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_min_distance(ma_spatializer* pSpatializer, float minDistance); +MA_API float ma_spatializer_get_min_distance(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_max_distance(ma_spatializer* pSpatializer, float maxDistance); +MA_API float ma_spatializer_get_max_distance(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_cone(ma_spatializer* pSpatializer, float innerAngleInRadians, float outerAngleInRadians, float outerGain); +MA_API void ma_spatializer_get_cone(const ma_spatializer* pSpatializer, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain); +MA_API void ma_spatializer_set_doppler_factor(ma_spatializer* pSpatializer, float dopplerFactor); +MA_API float ma_spatializer_get_doppler_factor(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_directional_attenuation_factor(ma_spatializer* pSpatializer, float directionalAttenuationFactor); +MA_API float ma_spatializer_get_directional_attenuation_factor(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_position(ma_spatializer* pSpatializer, float x, float y, float z); +MA_API ma_vec3f ma_spatializer_get_position(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_direction(ma_spatializer* pSpatializer, float x, float y, float z); +MA_API ma_vec3f ma_spatializer_get_direction(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_set_velocity(ma_spatializer* pSpatializer, float x, float y, float z); +MA_API ma_vec3f ma_spatializer_get_velocity(const ma_spatializer* pSpatializer); +MA_API void ma_spatializer_get_relative_position_and_direction(const ma_spatializer* pSpatializer, const ma_spatializer_listener* pListener, ma_vec3f* pRelativePos, ma_vec3f* pRelativeDir); + + + +/************************************************************************************************************************************************************ +************************************************************************************************************************************************************* + +DATA CONVERSION +=============== + +This section contains the APIs for data conversion. You will find everything here for channel mapping, sample format conversion, resampling, etc. + +************************************************************************************************************************************************************* +************************************************************************************************************************************************************/ + +/************************************************************************************************************************************************************** + +Resampling + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRateIn; + ma_uint32 sampleRateOut; + ma_uint32 lpfOrder; /* The low-pass filter order. Setting this to 0 will disable low-pass filtering. */ + double lpfNyquistFactor; /* 0..1. Defaults to 1. 1 = Half the sampling frequency (Nyquist Frequency), 0.5 = Quarter the sampling frequency (half Nyquest Frequency), etc. */ +} ma_linear_resampler_config; + +MA_API ma_linear_resampler_config ma_linear_resampler_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut); + +typedef struct +{ + ma_linear_resampler_config config; + ma_uint32 inAdvanceInt; + ma_uint32 inAdvanceFrac; + ma_uint32 inTimeInt; + ma_uint32 inTimeFrac; + union + { + float* f32; + ma_int16* s16; + } x0; /* The previous input frame. */ + union + { + float* f32; + ma_int16* s16; + } x1; /* The next input frame. */ + ma_lpf lpf; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_linear_resampler; + +MA_API ma_result ma_linear_resampler_get_heap_size(const ma_linear_resampler_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_linear_resampler_init_preallocated(const ma_linear_resampler_config* pConfig, void* pHeap, ma_linear_resampler* pResampler); +MA_API ma_result ma_linear_resampler_init(const ma_linear_resampler_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_linear_resampler* pResampler); +MA_API void ma_linear_resampler_uninit(ma_linear_resampler* pResampler, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_linear_resampler_process_pcm_frames(ma_linear_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut); +MA_API ma_result ma_linear_resampler_set_rate(ma_linear_resampler* pResampler, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut); +MA_API ma_result ma_linear_resampler_set_rate_ratio(ma_linear_resampler* pResampler, float ratioInOut); +MA_API ma_uint64 ma_linear_resampler_get_input_latency(const ma_linear_resampler* pResampler); +MA_API ma_uint64 ma_linear_resampler_get_output_latency(const ma_linear_resampler* pResampler); +MA_API ma_result ma_linear_resampler_get_required_input_frame_count(const ma_linear_resampler* pResampler, ma_uint64 outputFrameCount, ma_uint64* pInputFrameCount); +MA_API ma_result ma_linear_resampler_get_expected_output_frame_count(const ma_linear_resampler* pResampler, ma_uint64 inputFrameCount, ma_uint64* pOutputFrameCount); +MA_API ma_result ma_linear_resampler_reset(ma_linear_resampler* pResampler); + + +typedef struct ma_resampler_config ma_resampler_config; + +typedef void ma_resampling_backend; +typedef struct +{ + ma_result (* onGetHeapSize )(void* pUserData, const ma_resampler_config* pConfig, size_t* pHeapSizeInBytes); + ma_result (* onInit )(void* pUserData, const ma_resampler_config* pConfig, void* pHeap, ma_resampling_backend** ppBackend); + void (* onUninit )(void* pUserData, ma_resampling_backend* pBackend, const ma_allocation_callbacks* pAllocationCallbacks); + ma_result (* onProcess )(void* pUserData, ma_resampling_backend* pBackend, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut); + ma_result (* onSetRate )(void* pUserData, ma_resampling_backend* pBackend, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut); /* Optional. Rate changes will be disabled. */ + ma_uint64 (* onGetInputLatency )(void* pUserData, const ma_resampling_backend* pBackend); /* Optional. Latency will be reported as 0. */ + ma_uint64 (* onGetOutputLatency )(void* pUserData, const ma_resampling_backend* pBackend); /* Optional. Latency will be reported as 0. */ + ma_result (* onGetRequiredInputFrameCount )(void* pUserData, const ma_resampling_backend* pBackend, ma_uint64 outputFrameCount, ma_uint64* pInputFrameCount); /* Optional. Latency mitigation will be disabled. */ + ma_result (* onGetExpectedOutputFrameCount)(void* pUserData, const ma_resampling_backend* pBackend, ma_uint64 inputFrameCount, ma_uint64* pOutputFrameCount); /* Optional. Latency mitigation will be disabled. */ + ma_result (* onReset )(void* pUserData, ma_resampling_backend* pBackend); +} ma_resampling_backend_vtable; + +typedef enum +{ + ma_resample_algorithm_linear = 0, /* Fastest, lowest quality. Optional low-pass filtering. Default. */ + ma_resample_algorithm_custom, +} ma_resample_algorithm; + +struct ma_resampler_config +{ + ma_format format; /* Must be either ma_format_f32 or ma_format_s16. */ + ma_uint32 channels; + ma_uint32 sampleRateIn; + ma_uint32 sampleRateOut; + ma_resample_algorithm algorithm; /* When set to ma_resample_algorithm_custom, pBackendVTable will be used. */ + ma_resampling_backend_vtable* pBackendVTable; + void* pBackendUserData; + struct + { + ma_uint32 lpfOrder; + } linear; +}; + +MA_API ma_resampler_config ma_resampler_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut, ma_resample_algorithm algorithm); + +typedef struct +{ + ma_resampling_backend* pBackend; + ma_resampling_backend_vtable* pBackendVTable; + void* pBackendUserData; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRateIn; + ma_uint32 sampleRateOut; + union + { + ma_linear_resampler linear; + } state; /* State for stock resamplers so we can avoid a malloc. For stock resamplers, pBackend will point here. */ + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_resampler; + +MA_API ma_result ma_resampler_get_heap_size(const ma_resampler_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_resampler_init_preallocated(const ma_resampler_config* pConfig, void* pHeap, ma_resampler* pResampler); + +/* +Initializes a new resampler object from a config. +*/ +MA_API ma_result ma_resampler_init(const ma_resampler_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_resampler* pResampler); + +/* +Uninitializes a resampler. +*/ +MA_API void ma_resampler_uninit(ma_resampler* pResampler, const ma_allocation_callbacks* pAllocationCallbacks); + +/* +Converts the given input data. + +Both the input and output frames must be in the format specified in the config when the resampler was initialized. + +On input, [pFrameCountOut] contains the number of output frames to process. On output it contains the number of output frames that +were actually processed, which may be less than the requested amount which will happen if there's not enough input data. You can use +ma_resampler_get_expected_output_frame_count() to know how many output frames will be processed for a given number of input frames. + +On input, [pFrameCountIn] contains the number of input frames contained in [pFramesIn]. On output it contains the number of whole +input frames that were actually processed. You can use ma_resampler_get_required_input_frame_count() to know how many input frames +you should provide for a given number of output frames. [pFramesIn] can be NULL, in which case zeroes will be used instead. + +If [pFramesOut] is NULL, a seek is performed. In this case, if [pFrameCountOut] is not NULL it will seek by the specified number of +output frames. Otherwise, if [pFramesCountOut] is NULL and [pFrameCountIn] is not NULL, it will seek by the specified number of input +frames. When seeking, [pFramesIn] is allowed to NULL, in which case the internal timing state will be updated, but no input will be +processed. In this case, any internal filter state will be updated as if zeroes were passed in. + +It is an error for [pFramesOut] to be non-NULL and [pFrameCountOut] to be NULL. + +It is an error for both [pFrameCountOut] and [pFrameCountIn] to be NULL. +*/ +MA_API ma_result ma_resampler_process_pcm_frames(ma_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut); + + +/* +Sets the input and output sample rate. +*/ +MA_API ma_result ma_resampler_set_rate(ma_resampler* pResampler, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut); + +/* +Sets the input and output sample rate as a ratio. + +The ration is in/out. +*/ +MA_API ma_result ma_resampler_set_rate_ratio(ma_resampler* pResampler, float ratio); + +/* +Retrieves the latency introduced by the resampler in input frames. +*/ +MA_API ma_uint64 ma_resampler_get_input_latency(const ma_resampler* pResampler); + +/* +Retrieves the latency introduced by the resampler in output frames. +*/ +MA_API ma_uint64 ma_resampler_get_output_latency(const ma_resampler* pResampler); + +/* +Calculates the number of whole input frames that would need to be read from the client in order to output the specified +number of output frames. + +The returned value does not include cached input frames. It only returns the number of extra frames that would need to be +read from the input buffer in order to output the specified number of output frames. +*/ +MA_API ma_result ma_resampler_get_required_input_frame_count(const ma_resampler* pResampler, ma_uint64 outputFrameCount, ma_uint64* pInputFrameCount); + +/* +Calculates the number of whole output frames that would be output after fully reading and consuming the specified number of +input frames. +*/ +MA_API ma_result ma_resampler_get_expected_output_frame_count(const ma_resampler* pResampler, ma_uint64 inputFrameCount, ma_uint64* pOutputFrameCount); + +/* +Resets the resampler's timer and clears its internal cache. +*/ +MA_API ma_result ma_resampler_reset(ma_resampler* pResampler); + + +/************************************************************************************************************************************************************** + +Channel Conversion + +**************************************************************************************************************************************************************/ +typedef enum +{ + ma_channel_conversion_path_unknown, + ma_channel_conversion_path_passthrough, + ma_channel_conversion_path_mono_out, /* Converting to mono. */ + ma_channel_conversion_path_mono_in, /* Converting from mono. */ + ma_channel_conversion_path_shuffle, /* Simple shuffle. Will use this when all channels are present in both input and output channel maps, but just in a different order. */ + ma_channel_conversion_path_weights /* Blended based on weights. */ +} ma_channel_conversion_path; + +typedef enum +{ + ma_mono_expansion_mode_duplicate = 0, /* The default. */ + ma_mono_expansion_mode_average, /* Average the mono channel across all channels. */ + ma_mono_expansion_mode_stereo_only, /* Duplicate to the left and right channels only and ignore the others. */ + ma_mono_expansion_mode_default = ma_mono_expansion_mode_duplicate +} ma_mono_expansion_mode; + +typedef struct +{ + ma_format format; + ma_uint32 channelsIn; + ma_uint32 channelsOut; + const ma_channel* pChannelMapIn; + const ma_channel* pChannelMapOut; + ma_channel_mix_mode mixingMode; + ma_bool32 calculateLFEFromSpatialChannels; /* When an output LFE channel is present, but no input LFE, set to true to set the output LFE to the average of all spatial channels (LR, FR, etc.). Ignored when an input LFE is present. */ + float** ppWeights; /* [in][out]. Only used when mixingMode is set to ma_channel_mix_mode_custom_weights. */ +} ma_channel_converter_config; + +MA_API ma_channel_converter_config ma_channel_converter_config_init(ma_format format, ma_uint32 channelsIn, const ma_channel* pChannelMapIn, ma_uint32 channelsOut, const ma_channel* pChannelMapOut, ma_channel_mix_mode mixingMode); + +typedef struct +{ + ma_format format; + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_channel_mix_mode mixingMode; + ma_channel_conversion_path conversionPath; + ma_channel* pChannelMapIn; + ma_channel* pChannelMapOut; + ma_uint8* pShuffleTable; /* Indexed by output channel index. */ + union + { + float** f32; + ma_int32** s16; + } weights; /* [in][out] */ + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_channel_converter; + +MA_API ma_result ma_channel_converter_get_heap_size(const ma_channel_converter_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_channel_converter_init_preallocated(const ma_channel_converter_config* pConfig, void* pHeap, ma_channel_converter* pConverter); +MA_API ma_result ma_channel_converter_init(const ma_channel_converter_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_channel_converter* pConverter); +MA_API void ma_channel_converter_uninit(ma_channel_converter* pConverter, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_channel_converter_process_pcm_frames(ma_channel_converter* pConverter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount); +MA_API ma_result ma_channel_converter_get_input_channel_map(const ma_channel_converter* pConverter, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_channel_converter_get_output_channel_map(const ma_channel_converter* pConverter, ma_channel* pChannelMap, size_t channelMapCap); + + +/************************************************************************************************************************************************************** + +Data Conversion + +**************************************************************************************************************************************************************/ +typedef struct +{ + ma_format formatIn; + ma_format formatOut; + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_uint32 sampleRateIn; + ma_uint32 sampleRateOut; + ma_channel* pChannelMapIn; + ma_channel* pChannelMapOut; + ma_dither_mode ditherMode; + ma_channel_mix_mode channelMixMode; + ma_bool32 calculateLFEFromSpatialChannels; /* When an output LFE channel is present, but no input LFE, set to true to set the output LFE to the average of all spatial channels (LR, FR, etc.). Ignored when an input LFE is present. */ + float** ppChannelWeights; /* [in][out]. Only used when mixingMode is set to ma_channel_mix_mode_custom_weights. */ + ma_bool32 allowDynamicSampleRate; + ma_resampler_config resampling; +} ma_data_converter_config; + +MA_API ma_data_converter_config ma_data_converter_config_init_default(void); +MA_API ma_data_converter_config ma_data_converter_config_init(ma_format formatIn, ma_format formatOut, ma_uint32 channelsIn, ma_uint32 channelsOut, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut); + + +typedef enum +{ + ma_data_converter_execution_path_passthrough, /* No conversion. */ + ma_data_converter_execution_path_format_only, /* Only format conversion. */ + ma_data_converter_execution_path_channels_only, /* Only channel conversion. */ + ma_data_converter_execution_path_resample_only, /* Only resampling. */ + ma_data_converter_execution_path_resample_first, /* All conversions, but resample as the first step. */ + ma_data_converter_execution_path_channels_first /* All conversions, but channels as the first step. */ +} ma_data_converter_execution_path; + +typedef struct +{ + ma_format formatIn; + ma_format formatOut; + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_uint32 sampleRateIn; + ma_uint32 sampleRateOut; + ma_dither_mode ditherMode; + ma_data_converter_execution_path executionPath; /* The execution path the data converter will follow when processing. */ + ma_channel_converter channelConverter; + ma_resampler resampler; + ma_bool8 hasPreFormatConversion; + ma_bool8 hasPostFormatConversion; + ma_bool8 hasChannelConverter; + ma_bool8 hasResampler; + ma_bool8 isPassthrough; + + /* Memory management. */ + ma_bool8 _ownsHeap; + void* _pHeap; +} ma_data_converter; + +MA_API ma_result ma_data_converter_get_heap_size(const ma_data_converter_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_data_converter_init_preallocated(const ma_data_converter_config* pConfig, void* pHeap, ma_data_converter* pConverter); +MA_API ma_result ma_data_converter_init(const ma_data_converter_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_converter* pConverter); +MA_API void ma_data_converter_uninit(ma_data_converter* pConverter, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_data_converter_process_pcm_frames(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut); +MA_API ma_result ma_data_converter_set_rate(ma_data_converter* pConverter, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut); +MA_API ma_result ma_data_converter_set_rate_ratio(ma_data_converter* pConverter, float ratioInOut); +MA_API ma_uint64 ma_data_converter_get_input_latency(const ma_data_converter* pConverter); +MA_API ma_uint64 ma_data_converter_get_output_latency(const ma_data_converter* pConverter); +MA_API ma_result ma_data_converter_get_required_input_frame_count(const ma_data_converter* pConverter, ma_uint64 outputFrameCount, ma_uint64* pInputFrameCount); +MA_API ma_result ma_data_converter_get_expected_output_frame_count(const ma_data_converter* pConverter, ma_uint64 inputFrameCount, ma_uint64* pOutputFrameCount); +MA_API ma_result ma_data_converter_get_input_channel_map(const ma_data_converter* pConverter, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_data_converter_get_output_channel_map(const ma_data_converter* pConverter, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_data_converter_reset(ma_data_converter* pConverter); + + +/************************************************************************************************************************************************************ + +Format Conversion + +************************************************************************************************************************************************************/ +MA_API void ma_pcm_u8_to_s16(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_u8_to_s24(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_u8_to_s32(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_u8_to_f32(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s16_to_u8(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s16_to_s24(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s16_to_s32(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s16_to_f32(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s24_to_u8(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s24_to_s16(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s24_to_s32(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s24_to_f32(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s32_to_u8(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s32_to_s16(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s32_to_s24(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_s32_to_f32(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_f32_to_u8(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_f32_to_s16(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_f32_to_s24(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_f32_to_s32(void* pOut, const void* pIn, ma_uint64 count, ma_dither_mode ditherMode); +MA_API void ma_pcm_convert(void* pOut, ma_format formatOut, const void* pIn, ma_format formatIn, ma_uint64 sampleCount, ma_dither_mode ditherMode); +MA_API void ma_convert_pcm_frames_format(void* pOut, ma_format formatOut, const void* pIn, ma_format formatIn, ma_uint64 frameCount, ma_uint32 channels, ma_dither_mode ditherMode); + +/* +Deinterleaves an interleaved buffer. +*/ +MA_API void ma_deinterleave_pcm_frames(ma_format format, ma_uint32 channels, ma_uint64 frameCount, const void* pInterleavedPCMFrames, void** ppDeinterleavedPCMFrames); + +/* +Interleaves a group of deinterleaved buffers. +*/ +MA_API void ma_interleave_pcm_frames(ma_format format, ma_uint32 channels, ma_uint64 frameCount, const void** ppDeinterleavedPCMFrames, void* pInterleavedPCMFrames); + + +/************************************************************************************************************************************************************ + +Channel Maps + +************************************************************************************************************************************************************/ +/* +This is used in the shuffle table to indicate that the channel index is undefined and should be ignored. +*/ +#define MA_CHANNEL_INDEX_NULL 255 + +/* +Retrieves the channel position of the specified channel in the given channel map. + +The pChannelMap parameter can be null, in which case miniaudio's default channel map will be assumed. +*/ +MA_API ma_channel ma_channel_map_get_channel(const ma_channel* pChannelMap, ma_uint32 channelCount, ma_uint32 channelIndex); + +/* +Initializes a blank channel map. + +When a blank channel map is specified anywhere it indicates that the native channel map should be used. +*/ +MA_API void ma_channel_map_init_blank(ma_channel* pChannelMap, ma_uint32 channels); + +/* +Helper for retrieving a standard channel map. + +The output channel map buffer must have a capacity of at least `channelMapCap`. +*/ +MA_API void ma_channel_map_init_standard(ma_standard_channel_map standardChannelMap, ma_channel* pChannelMap, size_t channelMapCap, ma_uint32 channels); + +/* +Copies a channel map. + +Both input and output channel map buffers must have a capacity of at least `channels`. +*/ +MA_API void ma_channel_map_copy(ma_channel* pOut, const ma_channel* pIn, ma_uint32 channels); + +/* +Copies a channel map if one is specified, otherwise copies the default channel map. + +The output buffer must have a capacity of at least `channels`. If not NULL, the input channel map must also have a capacity of at least `channels`. +*/ +MA_API void ma_channel_map_copy_or_default(ma_channel* pOut, size_t channelMapCapOut, const ma_channel* pIn, ma_uint32 channels); + + +/* +Determines whether or not a channel map is valid. + +A blank channel map is valid (all channels set to MA_CHANNEL_NONE). The way a blank channel map is handled is context specific, but +is usually treated as a passthrough. + +Invalid channel maps: + - A channel map with no channels + - A channel map with more than one channel and a mono channel + +The channel map buffer must have a capacity of at least `channels`. +*/ +MA_API ma_bool32 ma_channel_map_is_valid(const ma_channel* pChannelMap, ma_uint32 channels); + +/* +Helper for comparing two channel maps for equality. + +This assumes the channel count is the same between the two. + +Both channels map buffers must have a capacity of at least `channels`. +*/ +MA_API ma_bool32 ma_channel_map_is_equal(const ma_channel* pChannelMapA, const ma_channel* pChannelMapB, ma_uint32 channels); + +/* +Helper for determining if a channel map is blank (all channels set to MA_CHANNEL_NONE). + +The channel map buffer must have a capacity of at least `channels`. +*/ +MA_API ma_bool32 ma_channel_map_is_blank(const ma_channel* pChannelMap, ma_uint32 channels); + +/* +Helper for determining whether or not a channel is present in the given channel map. + +The channel map buffer must have a capacity of at least `channels`. +*/ +MA_API ma_bool32 ma_channel_map_contains_channel_position(ma_uint32 channels, const ma_channel* pChannelMap, ma_channel channelPosition); + +/* +Find a channel position in the given channel map. Returns MA_TRUE if the channel is found; MA_FALSE otherwise. The +index of the channel is output to `pChannelIndex`. + +The channel map buffer must have a capacity of at least `channels`. +*/ +MA_API ma_bool32 ma_channel_map_find_channel_position(ma_uint32 channels, const ma_channel* pChannelMap, ma_channel channelPosition, ma_uint32* pChannelIndex); + +/* +Generates a string representing the given channel map. + +This is for printing and debugging purposes, not serialization/deserialization. + +Returns the length of the string, not including the null terminator. +*/ +MA_API size_t ma_channel_map_to_string(const ma_channel* pChannelMap, ma_uint32 channels, char* pBufferOut, size_t bufferCap); + +/* +Retrieves a human readable version of a channel position. +*/ +MA_API const char* ma_channel_position_to_string(ma_channel channel); + + +/************************************************************************************************************************************************************ + +Conversion Helpers + +************************************************************************************************************************************************************/ + +/* +High-level helper for doing a full format conversion in one go. Returns the number of output frames. Call this with pOut set to NULL to +determine the required size of the output buffer. frameCountOut should be set to the capacity of pOut. If pOut is NULL, frameCountOut is +ignored. + +A return value of 0 indicates an error. + +This function is useful for one-off bulk conversions, but if you're streaming data you should use the ma_data_converter APIs instead. +*/ +MA_API ma_uint64 ma_convert_frames(void* pOut, ma_uint64 frameCountOut, ma_format formatOut, ma_uint32 channelsOut, ma_uint32 sampleRateOut, const void* pIn, ma_uint64 frameCountIn, ma_format formatIn, ma_uint32 channelsIn, ma_uint32 sampleRateIn); +MA_API ma_uint64 ma_convert_frames_ex(void* pOut, ma_uint64 frameCountOut, const void* pIn, ma_uint64 frameCountIn, const ma_data_converter_config* pConfig); + + +/************************************************************************************************************************************************************ + +Data Source + +************************************************************************************************************************************************************/ +typedef void ma_data_source; + +#define MA_DATA_SOURCE_SELF_MANAGED_RANGE_AND_LOOP_POINT 0x00000001 + +typedef struct +{ + ma_result (* onRead)(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); + ma_result (* onSeek)(ma_data_source* pDataSource, ma_uint64 frameIndex); + ma_result (* onGetDataFormat)(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); + ma_result (* onGetCursor)(ma_data_source* pDataSource, ma_uint64* pCursor); + ma_result (* onGetLength)(ma_data_source* pDataSource, ma_uint64* pLength); + ma_result (* onSetLooping)(ma_data_source* pDataSource, ma_bool32 isLooping); + ma_uint32 flags; +} ma_data_source_vtable; + +typedef ma_data_source* (* ma_data_source_get_next_proc)(ma_data_source* pDataSource); + +typedef struct +{ + const ma_data_source_vtable* vtable; +} ma_data_source_config; + +MA_API ma_data_source_config ma_data_source_config_init(void); + + +typedef struct +{ + const ma_data_source_vtable* vtable; + ma_uint64 rangeBegInFrames; + ma_uint64 rangeEndInFrames; /* Set to -1 for unranged (default). */ + ma_uint64 loopBegInFrames; /* Relative to rangeBegInFrames. */ + ma_uint64 loopEndInFrames; /* Relative to rangeBegInFrames. Set to -1 for the end of the range. */ + ma_data_source* pCurrent; /* When non-NULL, the data source being initialized will act as a proxy and will route all operations to pCurrent. Used in conjunction with pNext/onGetNext for seamless chaining. */ + ma_data_source* pNext; /* When set to NULL, onGetNext will be used. */ + ma_data_source_get_next_proc onGetNext; /* Will be used when pNext is NULL. If both are NULL, no next will be used. */ + MA_ATOMIC(4, ma_bool32) isLooping; +} ma_data_source_base; + +MA_API ma_result ma_data_source_init(const ma_data_source_config* pConfig, ma_data_source* pDataSource); +MA_API void ma_data_source_uninit(ma_data_source* pDataSource); +MA_API ma_result ma_data_source_read_pcm_frames(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); /* Must support pFramesOut = NULL in which case a forward seek should be performed. */ +MA_API ma_result ma_data_source_seek_pcm_frames(ma_data_source* pDataSource, ma_uint64 frameCount, ma_uint64* pFramesSeeked); /* Can only seek forward. Equivalent to ma_data_source_read_pcm_frames(pDataSource, NULL, frameCount, &framesRead); */ +MA_API ma_result ma_data_source_seek_to_pcm_frame(ma_data_source* pDataSource, ma_uint64 frameIndex); +MA_API ma_result ma_data_source_seek_seconds(ma_data_source* pDataSource, float secondCount, float* pSecondsSeeked); /* Can only seek forward. Abstraction to ma_data_source_seek_pcm_frames() */ +MA_API ma_result ma_data_source_seek_to_second(ma_data_source* pDataSource, float seekPointInSeconds); /* Abstraction to ma_data_source_seek_to_pcm_frame() */ +MA_API ma_result ma_data_source_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_data_source_get_cursor_in_pcm_frames(ma_data_source* pDataSource, ma_uint64* pCursor); +MA_API ma_result ma_data_source_get_length_in_pcm_frames(ma_data_source* pDataSource, ma_uint64* pLength); /* Returns MA_NOT_IMPLEMENTED if the length is unknown or cannot be determined. Decoders can return this. */ +MA_API ma_result ma_data_source_get_cursor_in_seconds(ma_data_source* pDataSource, float* pCursor); +MA_API ma_result ma_data_source_get_length_in_seconds(ma_data_source* pDataSource, float* pLength); +MA_API ma_result ma_data_source_set_looping(ma_data_source* pDataSource, ma_bool32 isLooping); +MA_API ma_bool32 ma_data_source_is_looping(const ma_data_source* pDataSource); +MA_API ma_result ma_data_source_set_range_in_pcm_frames(ma_data_source* pDataSource, ma_uint64 rangeBegInFrames, ma_uint64 rangeEndInFrames); +MA_API void ma_data_source_get_range_in_pcm_frames(const ma_data_source* pDataSource, ma_uint64* pRangeBegInFrames, ma_uint64* pRangeEndInFrames); +MA_API ma_result ma_data_source_set_loop_point_in_pcm_frames(ma_data_source* pDataSource, ma_uint64 loopBegInFrames, ma_uint64 loopEndInFrames); +MA_API void ma_data_source_get_loop_point_in_pcm_frames(const ma_data_source* pDataSource, ma_uint64* pLoopBegInFrames, ma_uint64* pLoopEndInFrames); +MA_API ma_result ma_data_source_set_current(ma_data_source* pDataSource, ma_data_source* pCurrentDataSource); +MA_API ma_data_source* ma_data_source_get_current(const ma_data_source* pDataSource); +MA_API ma_result ma_data_source_set_next(ma_data_source* pDataSource, ma_data_source* pNextDataSource); +MA_API ma_data_source* ma_data_source_get_next(const ma_data_source* pDataSource); +MA_API ma_result ma_data_source_set_next_callback(ma_data_source* pDataSource, ma_data_source_get_next_proc onGetNext); +MA_API ma_data_source_get_next_proc ma_data_source_get_next_callback(const ma_data_source* pDataSource); + + +typedef struct +{ + ma_data_source_base ds; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_uint64 cursor; + ma_uint64 sizeInFrames; + const void* pData; +} ma_audio_buffer_ref; + +MA_API ma_result ma_audio_buffer_ref_init(ma_format format, ma_uint32 channels, const void* pData, ma_uint64 sizeInFrames, ma_audio_buffer_ref* pAudioBufferRef); +MA_API void ma_audio_buffer_ref_uninit(ma_audio_buffer_ref* pAudioBufferRef); +MA_API ma_result ma_audio_buffer_ref_set_data(ma_audio_buffer_ref* pAudioBufferRef, const void* pData, ma_uint64 sizeInFrames); +MA_API ma_uint64 ma_audio_buffer_ref_read_pcm_frames(ma_audio_buffer_ref* pAudioBufferRef, void* pFramesOut, ma_uint64 frameCount, ma_bool32 loop); +MA_API ma_result ma_audio_buffer_ref_seek_to_pcm_frame(ma_audio_buffer_ref* pAudioBufferRef, ma_uint64 frameIndex); +MA_API ma_result ma_audio_buffer_ref_map(ma_audio_buffer_ref* pAudioBufferRef, void** ppFramesOut, ma_uint64* pFrameCount); +MA_API ma_result ma_audio_buffer_ref_unmap(ma_audio_buffer_ref* pAudioBufferRef, ma_uint64 frameCount); /* Returns MA_AT_END if the end has been reached. This should be considered successful. */ +MA_API ma_bool32 ma_audio_buffer_ref_at_end(const ma_audio_buffer_ref* pAudioBufferRef); +MA_API ma_result ma_audio_buffer_ref_get_cursor_in_pcm_frames(const ma_audio_buffer_ref* pAudioBufferRef, ma_uint64* pCursor); +MA_API ma_result ma_audio_buffer_ref_get_length_in_pcm_frames(const ma_audio_buffer_ref* pAudioBufferRef, ma_uint64* pLength); +MA_API ma_result ma_audio_buffer_ref_get_available_frames(const ma_audio_buffer_ref* pAudioBufferRef, ma_uint64* pAvailableFrames); + + + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_uint64 sizeInFrames; + const void* pData; /* If set to NULL, will allocate a block of memory for you. */ + ma_allocation_callbacks allocationCallbacks; +} ma_audio_buffer_config; + +MA_API ma_audio_buffer_config ma_audio_buffer_config_init(ma_format format, ma_uint32 channels, ma_uint64 sizeInFrames, const void* pData, const ma_allocation_callbacks* pAllocationCallbacks); + +typedef struct +{ + ma_audio_buffer_ref ref; + ma_allocation_callbacks allocationCallbacks; + ma_bool32 ownsData; /* Used to control whether or not miniaudio owns the data buffer. If set to true, pData will be freed in ma_audio_buffer_uninit(). */ + ma_uint8 _pExtraData[1]; /* For allocating a buffer with the memory located directly after the other memory of the structure. */ +} ma_audio_buffer; + +MA_API ma_result ma_audio_buffer_init(const ma_audio_buffer_config* pConfig, ma_audio_buffer* pAudioBuffer); +MA_API ma_result ma_audio_buffer_init_copy(const ma_audio_buffer_config* pConfig, ma_audio_buffer* pAudioBuffer); +MA_API ma_result ma_audio_buffer_alloc_and_init(const ma_audio_buffer_config* pConfig, ma_audio_buffer** ppAudioBuffer); /* Always copies the data. Doesn't make sense to use this otherwise. Use ma_audio_buffer_uninit_and_free() to uninit. */ +MA_API void ma_audio_buffer_uninit(ma_audio_buffer* pAudioBuffer); +MA_API void ma_audio_buffer_uninit_and_free(ma_audio_buffer* pAudioBuffer); +MA_API ma_uint64 ma_audio_buffer_read_pcm_frames(ma_audio_buffer* pAudioBuffer, void* pFramesOut, ma_uint64 frameCount, ma_bool32 loop); +MA_API ma_result ma_audio_buffer_seek_to_pcm_frame(ma_audio_buffer* pAudioBuffer, ma_uint64 frameIndex); +MA_API ma_result ma_audio_buffer_map(ma_audio_buffer* pAudioBuffer, void** ppFramesOut, ma_uint64* pFrameCount); +MA_API ma_result ma_audio_buffer_unmap(ma_audio_buffer* pAudioBuffer, ma_uint64 frameCount); /* Returns MA_AT_END if the end has been reached. This should be considered successful. */ +MA_API ma_bool32 ma_audio_buffer_at_end(const ma_audio_buffer* pAudioBuffer); +MA_API ma_result ma_audio_buffer_get_cursor_in_pcm_frames(const ma_audio_buffer* pAudioBuffer, ma_uint64* pCursor); +MA_API ma_result ma_audio_buffer_get_length_in_pcm_frames(const ma_audio_buffer* pAudioBuffer, ma_uint64* pLength); +MA_API ma_result ma_audio_buffer_get_available_frames(const ma_audio_buffer* pAudioBuffer, ma_uint64* pAvailableFrames); + + +/* +Paged Audio Buffer +================== +A paged audio buffer is made up of a linked list of pages. It's expandable, but not shrinkable. It +can be used for cases where audio data is streamed in asynchronously while allowing data to be read +at the same time. + +This is lock-free, but not 100% thread safe. You can append a page and read from the buffer across +simultaneously across different threads, however only one thread at a time can append, and only one +thread at a time can read and seek. +*/ +typedef struct ma_paged_audio_buffer_page ma_paged_audio_buffer_page; +struct ma_paged_audio_buffer_page +{ + MA_ATOMIC(MA_SIZEOF_PTR, ma_paged_audio_buffer_page*) pNext; + ma_uint64 sizeInFrames; + ma_uint8 pAudioData[1]; +}; + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_paged_audio_buffer_page head; /* Dummy head for the lock-free algorithm. Always has a size of 0. */ + MA_ATOMIC(MA_SIZEOF_PTR, ma_paged_audio_buffer_page*) pTail; /* Never null. Initially set to &head. */ +} ma_paged_audio_buffer_data; + +MA_API ma_result ma_paged_audio_buffer_data_init(ma_format format, ma_uint32 channels, ma_paged_audio_buffer_data* pData); +MA_API void ma_paged_audio_buffer_data_uninit(ma_paged_audio_buffer_data* pData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_paged_audio_buffer_page* ma_paged_audio_buffer_data_get_head(ma_paged_audio_buffer_data* pData); +MA_API ma_paged_audio_buffer_page* ma_paged_audio_buffer_data_get_tail(ma_paged_audio_buffer_data* pData); +MA_API ma_result ma_paged_audio_buffer_data_get_length_in_pcm_frames(ma_paged_audio_buffer_data* pData, ma_uint64* pLength); +MA_API ma_result ma_paged_audio_buffer_data_allocate_page(ma_paged_audio_buffer_data* pData, ma_uint64 pageSizeInFrames, const void* pInitialData, const ma_allocation_callbacks* pAllocationCallbacks, ma_paged_audio_buffer_page** ppPage); +MA_API ma_result ma_paged_audio_buffer_data_free_page(ma_paged_audio_buffer_data* pData, ma_paged_audio_buffer_page* pPage, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_paged_audio_buffer_data_append_page(ma_paged_audio_buffer_data* pData, ma_paged_audio_buffer_page* pPage); +MA_API ma_result ma_paged_audio_buffer_data_allocate_and_append_page(ma_paged_audio_buffer_data* pData, ma_uint32 pageSizeInFrames, const void* pInitialData, const ma_allocation_callbacks* pAllocationCallbacks); + + +typedef struct +{ + ma_paged_audio_buffer_data* pData; /* Must not be null. */ +} ma_paged_audio_buffer_config; + +MA_API ma_paged_audio_buffer_config ma_paged_audio_buffer_config_init(ma_paged_audio_buffer_data* pData); + + +typedef struct +{ + ma_data_source_base ds; + ma_paged_audio_buffer_data* pData; /* Audio data is read from here. Cannot be null. */ + ma_paged_audio_buffer_page* pCurrent; + ma_uint64 relativeCursor; /* Relative to the current page. */ + ma_uint64 absoluteCursor; +} ma_paged_audio_buffer; + +MA_API ma_result ma_paged_audio_buffer_init(const ma_paged_audio_buffer_config* pConfig, ma_paged_audio_buffer* pPagedAudioBuffer); +MA_API void ma_paged_audio_buffer_uninit(ma_paged_audio_buffer* pPagedAudioBuffer); +MA_API ma_result ma_paged_audio_buffer_read_pcm_frames(ma_paged_audio_buffer* pPagedAudioBuffer, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); /* Returns MA_AT_END if no more pages available. */ +MA_API ma_result ma_paged_audio_buffer_seek_to_pcm_frame(ma_paged_audio_buffer* pPagedAudioBuffer, ma_uint64 frameIndex); +MA_API ma_result ma_paged_audio_buffer_get_cursor_in_pcm_frames(ma_paged_audio_buffer* pPagedAudioBuffer, ma_uint64* pCursor); +MA_API ma_result ma_paged_audio_buffer_get_length_in_pcm_frames(ma_paged_audio_buffer* pPagedAudioBuffer, ma_uint64* pLength); + + + +/************************************************************************************************************************************************************ + +Ring Buffer + +************************************************************************************************************************************************************/ +typedef struct +{ + void* pBuffer; + ma_uint32 subbufferSizeInBytes; + ma_uint32 subbufferCount; + ma_uint32 subbufferStrideInBytes; + MA_ATOMIC(4, ma_uint32) encodedReadOffset; /* Most significant bit is the loop flag. Lower 31 bits contains the actual offset in bytes. Must be used atomically. */ + MA_ATOMIC(4, ma_uint32) encodedWriteOffset; /* Most significant bit is the loop flag. Lower 31 bits contains the actual offset in bytes. Must be used atomically. */ + ma_bool8 ownsBuffer; /* Used to know whether or not miniaudio is responsible for free()-ing the buffer. */ + ma_bool8 clearOnWriteAcquire; /* When set, clears the acquired write buffer before returning from ma_rb_acquire_write(). */ + ma_allocation_callbacks allocationCallbacks; +} ma_rb; + +MA_API ma_result ma_rb_init_ex(size_t subbufferSizeInBytes, size_t subbufferCount, size_t subbufferStrideInBytes, void* pOptionalPreallocatedBuffer, const ma_allocation_callbacks* pAllocationCallbacks, ma_rb* pRB); +MA_API ma_result ma_rb_init(size_t bufferSizeInBytes, void* pOptionalPreallocatedBuffer, const ma_allocation_callbacks* pAllocationCallbacks, ma_rb* pRB); +MA_API void ma_rb_uninit(ma_rb* pRB); +MA_API void ma_rb_reset(ma_rb* pRB); +MA_API ma_result ma_rb_acquire_read(ma_rb* pRB, size_t* pSizeInBytes, void** ppBufferOut); +MA_API ma_result ma_rb_commit_read(ma_rb* pRB, size_t sizeInBytes); +MA_API ma_result ma_rb_acquire_write(ma_rb* pRB, size_t* pSizeInBytes, void** ppBufferOut); +MA_API ma_result ma_rb_commit_write(ma_rb* pRB, size_t sizeInBytes); +MA_API ma_result ma_rb_seek_read(ma_rb* pRB, size_t offsetInBytes); +MA_API ma_result ma_rb_seek_write(ma_rb* pRB, size_t offsetInBytes); +MA_API ma_int32 ma_rb_pointer_distance(ma_rb* pRB); /* Returns the distance between the write pointer and the read pointer. Should never be negative for a correct program. Will return the number of bytes that can be read before the read pointer hits the write pointer. */ +MA_API ma_uint32 ma_rb_available_read(ma_rb* pRB); +MA_API ma_uint32 ma_rb_available_write(ma_rb* pRB); +MA_API size_t ma_rb_get_subbuffer_size(ma_rb* pRB); +MA_API size_t ma_rb_get_subbuffer_stride(ma_rb* pRB); +MA_API size_t ma_rb_get_subbuffer_offset(ma_rb* pRB, size_t subbufferIndex); +MA_API void* ma_rb_get_subbuffer_ptr(ma_rb* pRB, size_t subbufferIndex, void* pBuffer); + + +typedef struct +{ + ma_data_source_base ds; + ma_rb rb; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; /* Not required for the ring buffer itself, but useful for associating the data with some sample rate, particularly for data sources. */ +} ma_pcm_rb; + +MA_API ma_result ma_pcm_rb_init_ex(ma_format format, ma_uint32 channels, ma_uint32 subbufferSizeInFrames, ma_uint32 subbufferCount, ma_uint32 subbufferStrideInFrames, void* pOptionalPreallocatedBuffer, const ma_allocation_callbacks* pAllocationCallbacks, ma_pcm_rb* pRB); +MA_API ma_result ma_pcm_rb_init(ma_format format, ma_uint32 channels, ma_uint32 bufferSizeInFrames, void* pOptionalPreallocatedBuffer, const ma_allocation_callbacks* pAllocationCallbacks, ma_pcm_rb* pRB); +MA_API void ma_pcm_rb_uninit(ma_pcm_rb* pRB); +MA_API void ma_pcm_rb_reset(ma_pcm_rb* pRB); +MA_API ma_result ma_pcm_rb_acquire_read(ma_pcm_rb* pRB, ma_uint32* pSizeInFrames, void** ppBufferOut); +MA_API ma_result ma_pcm_rb_commit_read(ma_pcm_rb* pRB, ma_uint32 sizeInFrames); +MA_API ma_result ma_pcm_rb_acquire_write(ma_pcm_rb* pRB, ma_uint32* pSizeInFrames, void** ppBufferOut); +MA_API ma_result ma_pcm_rb_commit_write(ma_pcm_rb* pRB, ma_uint32 sizeInFrames); +MA_API ma_result ma_pcm_rb_seek_read(ma_pcm_rb* pRB, ma_uint32 offsetInFrames); +MA_API ma_result ma_pcm_rb_seek_write(ma_pcm_rb* pRB, ma_uint32 offsetInFrames); +MA_API ma_int32 ma_pcm_rb_pointer_distance(ma_pcm_rb* pRB); /* Return value is in frames. */ +MA_API ma_uint32 ma_pcm_rb_available_read(ma_pcm_rb* pRB); +MA_API ma_uint32 ma_pcm_rb_available_write(ma_pcm_rb* pRB); +MA_API ma_uint32 ma_pcm_rb_get_subbuffer_size(ma_pcm_rb* pRB); +MA_API ma_uint32 ma_pcm_rb_get_subbuffer_stride(ma_pcm_rb* pRB); +MA_API ma_uint32 ma_pcm_rb_get_subbuffer_offset(ma_pcm_rb* pRB, ma_uint32 subbufferIndex); +MA_API void* ma_pcm_rb_get_subbuffer_ptr(ma_pcm_rb* pRB, ma_uint32 subbufferIndex, void* pBuffer); +MA_API ma_format ma_pcm_rb_get_format(const ma_pcm_rb* pRB); +MA_API ma_uint32 ma_pcm_rb_get_channels(const ma_pcm_rb* pRB); +MA_API ma_uint32 ma_pcm_rb_get_sample_rate(const ma_pcm_rb* pRB); +MA_API void ma_pcm_rb_set_sample_rate(ma_pcm_rb* pRB, ma_uint32 sampleRate); + + +/* +The idea of the duplex ring buffer is to act as the intermediary buffer when running two asynchronous devices in a duplex set up. The +capture device writes to it, and then a playback device reads from it. + +At the moment this is just a simple naive implementation, but in the future I want to implement some dynamic resampling to seamlessly +handle desyncs. Note that the API is work in progress and may change at any time in any version. + +The size of the buffer is based on the capture side since that's what'll be written to the buffer. It is based on the capture period size +in frames. The internal sample rate of the capture device is also needed in order to calculate the size. +*/ +typedef struct +{ + ma_pcm_rb rb; +} ma_duplex_rb; + +MA_API ma_result ma_duplex_rb_init(ma_format captureFormat, ma_uint32 captureChannels, ma_uint32 sampleRate, ma_uint32 captureInternalSampleRate, ma_uint32 captureInternalPeriodSizeInFrames, const ma_allocation_callbacks* pAllocationCallbacks, ma_duplex_rb* pRB); +MA_API ma_result ma_duplex_rb_uninit(ma_duplex_rb* pRB); + + +/************************************************************************************************************************************************************ + +Miscellaneous Helpers + +************************************************************************************************************************************************************/ +/* +Retrieves a human readable description of the given result code. +*/ +MA_API const char* ma_result_description(ma_result result); + +/* +malloc() +*/ +MA_API void* ma_malloc(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks); + +/* +calloc() +*/ +MA_API void* ma_calloc(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks); + +/* +realloc() +*/ +MA_API void* ma_realloc(void* p, size_t sz, const ma_allocation_callbacks* pAllocationCallbacks); + +/* +free() +*/ +MA_API void ma_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks); + +/* +Performs an aligned malloc, with the assumption that the alignment is a power of 2. +*/ +MA_API void* ma_aligned_malloc(size_t sz, size_t alignment, const ma_allocation_callbacks* pAllocationCallbacks); + +/* +Free's an aligned malloc'd buffer. +*/ +MA_API void ma_aligned_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks); + +/* +Retrieves a friendly name for a format. +*/ +MA_API const char* ma_get_format_name(ma_format format); + +/* +Blends two frames in floating point format. +*/ +MA_API void ma_blend_f32(float* pOut, float* pInA, float* pInB, float factor, ma_uint32 channels); + +/* +Retrieves the size of a sample in bytes for the given format. + +This API is efficient and is implemented using a lookup table. + +Thread Safety: SAFE + This API is pure. +*/ +MA_API ma_uint32 ma_get_bytes_per_sample(ma_format format); +static MA_INLINE ma_uint32 ma_get_bytes_per_frame(ma_format format, ma_uint32 channels) { return ma_get_bytes_per_sample(format) * channels; } + +/* +Converts a log level to a string. +*/ +MA_API const char* ma_log_level_to_string(ma_uint32 logLevel); + + + + +/************************************************************************************************************************************************************ + +Synchronization + +************************************************************************************************************************************************************/ +/* +Locks a spinlock. +*/ +MA_API ma_result ma_spinlock_lock(volatile ma_spinlock* pSpinlock); + +/* +Locks a spinlock, but does not yield() when looping. +*/ +MA_API ma_result ma_spinlock_lock_noyield(volatile ma_spinlock* pSpinlock); + +/* +Unlocks a spinlock. +*/ +MA_API ma_result ma_spinlock_unlock(volatile ma_spinlock* pSpinlock); + + +#ifndef MA_NO_THREADING + +/* +Creates a mutex. + +A mutex must be created from a valid context. A mutex is initially unlocked. +*/ +MA_API ma_result ma_mutex_init(ma_mutex* pMutex); + +/* +Deletes a mutex. +*/ +MA_API void ma_mutex_uninit(ma_mutex* pMutex); + +/* +Locks a mutex with an infinite timeout. +*/ +MA_API void ma_mutex_lock(ma_mutex* pMutex); + +/* +Unlocks a mutex. +*/ +MA_API void ma_mutex_unlock(ma_mutex* pMutex); + + +/* +Initializes an auto-reset event. +*/ +MA_API ma_result ma_event_init(ma_event* pEvent); + +/* +Uninitializes an auto-reset event. +*/ +MA_API void ma_event_uninit(ma_event* pEvent); + +/* +Waits for the specified auto-reset event to become signalled. +*/ +MA_API ma_result ma_event_wait(ma_event* pEvent); + +/* +Signals the specified auto-reset event. +*/ +MA_API ma_result ma_event_signal(ma_event* pEvent); + + +MA_API ma_result ma_semaphore_init(int initialValue, ma_semaphore* pSemaphore); +MA_API void ma_semaphore_uninit(ma_semaphore* pSemaphore); +MA_API ma_result ma_semaphore_wait(ma_semaphore* pSemaphore); +MA_API ma_result ma_semaphore_release(ma_semaphore* pSemaphore); +#endif /* MA_NO_THREADING */ + + +/* +Fence +===== +This locks while the counter is larger than 0. Counter can be incremented and decremented by any +thread, but care needs to be taken when waiting. It is possible for one thread to acquire the +fence just as another thread returns from ma_fence_wait(). + +The idea behind a fence is to allow you to wait for a group of operations to complete. When an +operation starts, the counter is incremented which locks the fence. When the operation completes, +the fence will be released which decrements the counter. ma_fence_wait() will block until the +counter hits zero. + +If threading is disabled, ma_fence_wait() will spin on the counter. +*/ +typedef struct +{ +#ifndef MA_NO_THREADING + ma_event e; +#endif + ma_uint32 counter; +} ma_fence; + +MA_API ma_result ma_fence_init(ma_fence* pFence); +MA_API void ma_fence_uninit(ma_fence* pFence); +MA_API ma_result ma_fence_acquire(ma_fence* pFence); /* Increment counter. */ +MA_API ma_result ma_fence_release(ma_fence* pFence); /* Decrement counter. */ +MA_API ma_result ma_fence_wait(ma_fence* pFence); /* Wait for counter to reach 0. */ + + + +/* +Notification callback for asynchronous operations. +*/ +typedef void ma_async_notification; + +typedef struct +{ + void (* onSignal)(ma_async_notification* pNotification); +} ma_async_notification_callbacks; + +MA_API ma_result ma_async_notification_signal(ma_async_notification* pNotification); + + +/* +Simple polling notification. + +This just sets a variable when the notification has been signalled which is then polled with ma_async_notification_poll_is_signalled() +*/ +typedef struct +{ + ma_async_notification_callbacks cb; + ma_bool32 signalled; +} ma_async_notification_poll; + +MA_API ma_result ma_async_notification_poll_init(ma_async_notification_poll* pNotificationPoll); +MA_API ma_bool32 ma_async_notification_poll_is_signalled(const ma_async_notification_poll* pNotificationPoll); + + +/* +Event Notification + +This uses an ma_event. If threading is disabled (MA_NO_THREADING), initialization will fail. +*/ +typedef struct +{ + ma_async_notification_callbacks cb; +#ifndef MA_NO_THREADING + ma_event e; +#endif +} ma_async_notification_event; + +MA_API ma_result ma_async_notification_event_init(ma_async_notification_event* pNotificationEvent); +MA_API ma_result ma_async_notification_event_uninit(ma_async_notification_event* pNotificationEvent); +MA_API ma_result ma_async_notification_event_wait(ma_async_notification_event* pNotificationEvent); +MA_API ma_result ma_async_notification_event_signal(ma_async_notification_event* pNotificationEvent); + + + + +/************************************************************************************************************************************************************ + +Job Queue + +************************************************************************************************************************************************************/ + +/* +Slot Allocator +-------------- +The idea of the slot allocator is for it to be used in conjunction with a fixed sized buffer. You use the slot allocator to allocate an index that can be used +as the insertion point for an object. + +Slots are reference counted to help mitigate the ABA problem in the lock-free queue we use for tracking jobs. + +The slot index is stored in the low 32 bits. The reference counter is stored in the high 32 bits: + + +-----------------+-----------------+ + | 32 Bits | 32 Bits | + +-----------------+-----------------+ + | Reference Count | Slot Index | + +-----------------+-----------------+ +*/ +typedef struct +{ + ma_uint32 capacity; /* The number of slots to make available. */ +} ma_slot_allocator_config; + +MA_API ma_slot_allocator_config ma_slot_allocator_config_init(ma_uint32 capacity); + + +typedef struct +{ + MA_ATOMIC(4, ma_uint32) bitfield; /* Must be used atomically because the allocation and freeing routines need to make copies of this which must never be optimized away by the compiler. */ +} ma_slot_allocator_group; + +typedef struct +{ + ma_slot_allocator_group* pGroups; /* Slots are grouped in chunks of 32. */ + ma_uint32* pSlots; /* 32 bits for reference counting for ABA mitigation. */ + ma_uint32 count; /* Allocation count. */ + ma_uint32 capacity; + + /* Memory management. */ + ma_bool32 _ownsHeap; + void* _pHeap; +} ma_slot_allocator; + +MA_API ma_result ma_slot_allocator_get_heap_size(const ma_slot_allocator_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_slot_allocator_init_preallocated(const ma_slot_allocator_config* pConfig, void* pHeap, ma_slot_allocator* pAllocator); +MA_API ma_result ma_slot_allocator_init(const ma_slot_allocator_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_slot_allocator* pAllocator); +MA_API void ma_slot_allocator_uninit(ma_slot_allocator* pAllocator, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_slot_allocator_alloc(ma_slot_allocator* pAllocator, ma_uint64* pSlot); +MA_API ma_result ma_slot_allocator_free(ma_slot_allocator* pAllocator, ma_uint64 slot); + + +typedef struct ma_job ma_job; + +/* +Callback for processing a job. Each job type will have their own processing callback which will be +called by ma_job_process(). +*/ +typedef ma_result (* ma_job_proc)(ma_job* pJob); + +/* When a job type is added here an callback needs to be added go "g_jobVTable" in the implementation section. */ +typedef enum +{ + /* Miscellaneous. */ + MA_JOB_TYPE_QUIT = 0, + MA_JOB_TYPE_CUSTOM, + + /* Resource Manager. */ + MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER_NODE, + MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER_NODE, + MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_BUFFER_NODE, + MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER, + MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER, + MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_STREAM, + MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_STREAM, + MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_STREAM, + MA_JOB_TYPE_RESOURCE_MANAGER_SEEK_DATA_STREAM, + + /* Device. */ + MA_JOB_TYPE_DEVICE_AAUDIO_REROUTE, + + /* Count. Must always be last. */ + MA_JOB_TYPE_COUNT +} ma_job_type; + +struct ma_job +{ + union + { + struct + { + ma_uint16 code; /* Job type. */ + ma_uint16 slot; /* Index into a ma_slot_allocator. */ + ma_uint32 refcount; + } breakup; + ma_uint64 allocation; + } toc; /* 8 bytes. We encode the job code into the slot allocation data to save space. */ + MA_ATOMIC(8, ma_uint64) next; /* refcount + slot for the next item. Does not include the job code. */ + ma_uint32 order; /* Execution order. Used to create a data dependency and ensure a job is executed in order. Usage is contextual depending on the job type. */ + + union + { + /* Miscellaneous. */ + struct + { + ma_job_proc proc; + ma_uintptr data0; + ma_uintptr data1; + } custom; + + /* Resource Manager */ + union + { + struct + { + /*ma_resource_manager**/ void* pResourceManager; + /*ma_resource_manager_data_buffer_node**/ void* pDataBufferNode; + char* pFilePath; + wchar_t* pFilePathW; + ma_uint32 flags; /* Resource manager data source flags that were used when initializing the data buffer. */ + ma_async_notification* pInitNotification; /* Signalled when the data buffer has been initialized and the format/channels/rate can be retrieved. */ + ma_async_notification* pDoneNotification; /* Signalled when the data buffer has been fully decoded. Will be passed through to MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_BUFFER_NODE when decoding. */ + ma_fence* pInitFence; /* Released when initialization of the decoder is complete. */ + ma_fence* pDoneFence; /* Released if initialization of the decoder fails. Passed through to PAGE_DATA_BUFFER_NODE untouched if init is successful. */ + } loadDataBufferNode; + struct + { + /*ma_resource_manager**/ void* pResourceManager; + /*ma_resource_manager_data_buffer_node**/ void* pDataBufferNode; + ma_async_notification* pDoneNotification; + ma_fence* pDoneFence; + } freeDataBufferNode; + struct + { + /*ma_resource_manager**/ void* pResourceManager; + /*ma_resource_manager_data_buffer_node**/ void* pDataBufferNode; + /*ma_decoder**/ void* pDecoder; + ma_async_notification* pDoneNotification; /* Signalled when the data buffer has been fully decoded. */ + ma_fence* pDoneFence; /* Passed through from LOAD_DATA_BUFFER_NODE and released when the data buffer completes decoding or an error occurs. */ + } pageDataBufferNode; + + struct + { + /*ma_resource_manager_data_buffer**/ void* pDataBuffer; + ma_async_notification* pInitNotification; /* Signalled when the data buffer has been initialized and the format/channels/rate can be retrieved. */ + ma_async_notification* pDoneNotification; /* Signalled when the data buffer has been fully decoded. */ + ma_fence* pInitFence; /* Released when the data buffer has been initialized and the format/channels/rate can be retrieved. */ + ma_fence* pDoneFence; /* Released when the data buffer has been fully decoded. */ + ma_uint64 rangeBegInPCMFrames; + ma_uint64 rangeEndInPCMFrames; + ma_uint64 loopPointBegInPCMFrames; + ma_uint64 loopPointEndInPCMFrames; + ma_uint32 isLooping; + } loadDataBuffer; + struct + { + /*ma_resource_manager_data_buffer**/ void* pDataBuffer; + ma_async_notification* pDoneNotification; + ma_fence* pDoneFence; + } freeDataBuffer; + + struct + { + /*ma_resource_manager_data_stream**/ void* pDataStream; + char* pFilePath; /* Allocated when the job is posted, freed by the job thread after loading. */ + wchar_t* pFilePathW; /* ^ As above ^. Only used if pFilePath is NULL. */ + ma_uint64 initialSeekPoint; + ma_async_notification* pInitNotification; /* Signalled after the first two pages have been decoded and frames can be read from the stream. */ + ma_fence* pInitFence; + } loadDataStream; + struct + { + /*ma_resource_manager_data_stream**/ void* pDataStream; + ma_async_notification* pDoneNotification; + ma_fence* pDoneFence; + } freeDataStream; + struct + { + /*ma_resource_manager_data_stream**/ void* pDataStream; + ma_uint32 pageIndex; /* The index of the page to decode into. */ + } pageDataStream; + struct + { + /*ma_resource_manager_data_stream**/ void* pDataStream; + ma_uint64 frameIndex; + } seekDataStream; + } resourceManager; + + /* Device. */ + union + { + union + { + struct + { + /*ma_device**/ void* pDevice; + /*ma_device_type*/ ma_uint32 deviceType; + } reroute; + } aaudio; + } device; + } data; +}; + +MA_API ma_job ma_job_init(ma_uint16 code); +MA_API ma_result ma_job_process(ma_job* pJob); + + +/* +When set, ma_job_queue_next() will not wait and no semaphore will be signaled in +ma_job_queue_post(). ma_job_queue_next() will return MA_NO_DATA_AVAILABLE if nothing is available. + +This flag should always be used for platforms that do not support multithreading. +*/ +typedef enum +{ + MA_JOB_QUEUE_FLAG_NON_BLOCKING = 0x00000001 +} ma_job_queue_flags; + +typedef struct +{ + ma_uint32 flags; + ma_uint32 capacity; /* The maximum number of jobs that can fit in the queue at a time. */ +} ma_job_queue_config; + +MA_API ma_job_queue_config ma_job_queue_config_init(ma_uint32 flags, ma_uint32 capacity); + + +typedef struct +{ + ma_uint32 flags; /* Flags passed in at initialization time. */ + ma_uint32 capacity; /* The maximum number of jobs that can fit in the queue at a time. Set by the config. */ + MA_ATOMIC(8, ma_uint64) head; /* The first item in the list. Required for removing from the top of the list. */ + MA_ATOMIC(8, ma_uint64) tail; /* The last item in the list. Required for appending to the end of the list. */ +#ifndef MA_NO_THREADING + ma_semaphore sem; /* Only used when MA_JOB_QUEUE_FLAG_NON_BLOCKING is unset. */ +#endif + ma_slot_allocator allocator; + ma_job* pJobs; +#ifndef MA_USE_EXPERIMENTAL_LOCK_FREE_JOB_QUEUE + ma_spinlock lock; +#endif + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_job_queue; + +MA_API ma_result ma_job_queue_get_heap_size(const ma_job_queue_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_job_queue_init_preallocated(const ma_job_queue_config* pConfig, void* pHeap, ma_job_queue* pQueue); +MA_API ma_result ma_job_queue_init(const ma_job_queue_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_job_queue* pQueue); +MA_API void ma_job_queue_uninit(ma_job_queue* pQueue, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_job_queue_post(ma_job_queue* pQueue, const ma_job* pJob); +MA_API ma_result ma_job_queue_next(ma_job_queue* pQueue, ma_job* pJob); /* Returns MA_CANCELLED if the next job is a quit job. */ + + + +/************************************************************************************************************************************************************ +************************************************************************************************************************************************************* + +DEVICE I/O +========== + +This section contains the APIs for device playback and capture. Here is where you'll find ma_device_init(), etc. + +************************************************************************************************************************************************************* +************************************************************************************************************************************************************/ +#ifndef MA_NO_DEVICE_IO +/* Some backends are only supported on certain platforms. */ +#if defined(MA_WIN32) + #define MA_SUPPORT_WASAPI + + #if defined(MA_WIN32_DESKTOP) /* DirectSound and WinMM backends are only supported on desktops. */ + #define MA_SUPPORT_DSOUND + #define MA_SUPPORT_WINMM + + /* Don't enable JACK here if compiling with Cosmopolitan. It'll be enabled in the Linux section below. */ + #if !defined(__COSMOPOLITAN__) + #define MA_SUPPORT_JACK /* JACK is technically supported on Windows, but I don't know how many people use it in practice... */ + #endif + #endif +#endif +#if defined(MA_UNIX) && !defined(MA_ORBIS) && !defined(MA_PROSPERO) + #if defined(MA_LINUX) + #if !defined(MA_ANDROID) && !defined(__COSMOPOLITAN__) /* ALSA is not supported on Android. */ + #define MA_SUPPORT_ALSA + #endif + #endif + #if !defined(MA_BSD) && !defined(MA_ANDROID) && !defined(MA_EMSCRIPTEN) + #define MA_SUPPORT_PULSEAUDIO + #define MA_SUPPORT_JACK + #endif + #if defined(__OpenBSD__) /* <-- Change this to "#if defined(MA_BSD)" to enable sndio on all BSD flavors. */ + #define MA_SUPPORT_SNDIO /* sndio is only supported on OpenBSD for now. May be expanded later if there's demand. */ + #endif + #if defined(__NetBSD__) || defined(__OpenBSD__) + #define MA_SUPPORT_AUDIO4 /* Only support audio(4) on platforms with known support. */ + #endif + #if defined(__FreeBSD__) || defined(__DragonFly__) + #define MA_SUPPORT_OSS /* Only support OSS on specific platforms with known support. */ + #endif +#endif +#if defined(MA_ANDROID) + #define MA_SUPPORT_AAUDIO + #define MA_SUPPORT_OPENSL +#endif +#if defined(MA_APPLE) + #define MA_SUPPORT_COREAUDIO +#endif +#if defined(MA_EMSCRIPTEN) + #define MA_SUPPORT_WEBAUDIO +#endif + +/* All platforms should support custom backends. */ +#define MA_SUPPORT_CUSTOM + +/* Explicitly disable the Null backend for Emscripten because it uses a background thread which is not properly supported right now. */ +#if !defined(MA_EMSCRIPTEN) +#define MA_SUPPORT_NULL +#endif + + +#if defined(MA_SUPPORT_WASAPI) && !defined(MA_NO_WASAPI) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_WASAPI)) + #define MA_HAS_WASAPI +#endif +#if defined(MA_SUPPORT_DSOUND) && !defined(MA_NO_DSOUND) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_DSOUND)) + #define MA_HAS_DSOUND +#endif +#if defined(MA_SUPPORT_WINMM) && !defined(MA_NO_WINMM) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_WINMM)) + #define MA_HAS_WINMM +#endif +#if defined(MA_SUPPORT_ALSA) && !defined(MA_NO_ALSA) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_ALSA)) + #define MA_HAS_ALSA +#endif +#if defined(MA_SUPPORT_PULSEAUDIO) && !defined(MA_NO_PULSEAUDIO) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_PULSEAUDIO)) + #define MA_HAS_PULSEAUDIO +#endif +#if defined(MA_SUPPORT_JACK) && !defined(MA_NO_JACK) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_JACK)) + #define MA_HAS_JACK +#endif +#if defined(MA_SUPPORT_COREAUDIO) && !defined(MA_NO_COREAUDIO) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_COREAUDIO)) + #define MA_HAS_COREAUDIO +#endif +#if defined(MA_SUPPORT_SNDIO) && !defined(MA_NO_SNDIO) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_SNDIO)) + #define MA_HAS_SNDIO +#endif +#if defined(MA_SUPPORT_AUDIO4) && !defined(MA_NO_AUDIO4) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_AUDIO4)) + #define MA_HAS_AUDIO4 +#endif +#if defined(MA_SUPPORT_OSS) && !defined(MA_NO_OSS) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_OSS)) + #define MA_HAS_OSS +#endif +#if defined(MA_SUPPORT_AAUDIO) && !defined(MA_NO_AAUDIO) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_AAUDIO)) + #define MA_HAS_AAUDIO +#endif +#if defined(MA_SUPPORT_OPENSL) && !defined(MA_NO_OPENSL) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_OPENSL)) + #define MA_HAS_OPENSL +#endif +#if defined(MA_SUPPORT_WEBAUDIO) && !defined(MA_NO_WEBAUDIO) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_WEBAUDIO)) + #define MA_HAS_WEBAUDIO +#endif +#if defined(MA_SUPPORT_CUSTOM) && !defined(MA_NO_CUSTOM) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_CUSTOM)) + #define MA_HAS_CUSTOM +#endif +#if defined(MA_SUPPORT_NULL) && !defined(MA_NO_NULL) && (!defined(MA_ENABLE_ONLY_SPECIFIC_BACKENDS) || defined(MA_ENABLE_NULL)) + #define MA_HAS_NULL +#endif + +typedef enum +{ + ma_device_state_uninitialized = 0, + ma_device_state_stopped = 1, /* The device's default state after initialization. */ + ma_device_state_started = 2, /* The device is started and is requesting and/or delivering audio data. */ + ma_device_state_starting = 3, /* Transitioning from a stopped state to started. */ + ma_device_state_stopping = 4 /* Transitioning from a started state to stopped. */ +} ma_device_state; + +MA_ATOMIC_SAFE_TYPE_DECL(i32, 4, device_state) + + +#ifdef MA_SUPPORT_WASAPI +/* We need a IMMNotificationClient object for WASAPI. */ +typedef struct +{ + void* lpVtbl; + ma_uint32 counter; + ma_device* pDevice; +} ma_IMMNotificationClient; +#endif + +/* Backend enums must be in priority order. */ +typedef enum +{ + ma_backend_wasapi, + ma_backend_dsound, + ma_backend_winmm, + ma_backend_coreaudio, + ma_backend_sndio, + ma_backend_audio4, + ma_backend_oss, + ma_backend_pulseaudio, + ma_backend_alsa, + ma_backend_jack, + ma_backend_aaudio, + ma_backend_opensl, + ma_backend_webaudio, + ma_backend_custom, /* <-- Custom backend, with callbacks defined by the context config. */ + ma_backend_null /* <-- Must always be the last item. Lowest priority, and used as the terminator for backend enumeration. */ +} ma_backend; + +#define MA_BACKEND_COUNT (ma_backend_null+1) + + +/* +Device job thread. This is used by backends that require asynchronous processing of certain +operations. It is not used by all backends. + +The device job thread is made up of a thread and a job queue. You can post a job to the thread with +ma_device_job_thread_post(). The thread will do the processing of the job. +*/ +typedef struct +{ + ma_bool32 noThread; /* Set this to true if you want to process jobs yourself. */ + ma_uint32 jobQueueCapacity; + ma_uint32 jobQueueFlags; +} ma_device_job_thread_config; + +MA_API ma_device_job_thread_config ma_device_job_thread_config_init(void); + +typedef struct +{ + ma_thread thread; + ma_job_queue jobQueue; + ma_bool32 _hasThread; +} ma_device_job_thread; + +MA_API ma_result ma_device_job_thread_init(const ma_device_job_thread_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_device_job_thread* pJobThread); +MA_API void ma_device_job_thread_uninit(ma_device_job_thread* pJobThread, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_device_job_thread_post(ma_device_job_thread* pJobThread, const ma_job* pJob); +MA_API ma_result ma_device_job_thread_next(ma_device_job_thread* pJobThread, ma_job* pJob); + + + +/* Device notification types. */ +typedef enum +{ + ma_device_notification_type_started, + ma_device_notification_type_stopped, + ma_device_notification_type_rerouted, + ma_device_notification_type_interruption_began, + ma_device_notification_type_interruption_ended, + ma_device_notification_type_unlocked +} ma_device_notification_type; + +typedef struct +{ + ma_device* pDevice; + ma_device_notification_type type; + union + { + struct + { + int _unused; + } started; + struct + { + int _unused; + } stopped; + struct + { + int _unused; + } rerouted; + struct + { + int _unused; + } interruption; + } data; +} ma_device_notification; + +/* +The notification callback for when the application should be notified of a change to the device. + +This callback is used for notifying the application of changes such as when the device has started, +stopped, rerouted or an interruption has occurred. Note that not all backends will post all +notification types. For example, some backends will perform automatic stream routing without any +kind of notification to the host program which means miniaudio will never know about it and will +never be able to fire the rerouted notification. You should keep this in mind when designing your +program. + +The stopped notification will *not* get fired when a device is rerouted. + + +Parameters +---------- +pNotification (in) + A pointer to a structure containing information about the event. Use the `pDevice` member of + this object to retrieve the relevant device. The `type` member can be used to discriminate + against each of the notification types. + + +Remarks +------- +Do not restart or uninitialize the device from the callback. + +Not all notifications will be triggered by all backends, however the started and stopped events +should be reliable for all backends. Some backends do not have a good way to detect device +stoppages due to unplugging the device which may result in the stopped callback not getting +fired. This has been observed with at least one BSD variant. + +The rerouted notification is fired *after* the reroute has occurred. The stopped notification will +*not* get fired when a device is rerouted. The following backends are known to do automatic stream +rerouting, but do not have a way to be notified of the change: + + * DirectSound + +The interruption notifications are used on mobile platforms for detecting when audio is interrupted +due to things like an incoming phone call. Currently this is only implemented on iOS. None of the +Android backends will report this notification. +*/ +typedef void (* ma_device_notification_proc)(const ma_device_notification* pNotification); + + +/* +The callback for processing audio data from the device. + +The data callback is fired by miniaudio whenever the device needs to have more data delivered to a playback device, or when a capture device has some data +available. This is called as soon as the backend asks for more data which means it may be called with inconsistent frame counts. You cannot assume the +callback will be fired with a consistent frame count. + + +Parameters +---------- +pDevice (in) + A pointer to the relevant device. + +pOutput (out) + A pointer to the output buffer that will receive audio data that will later be played back through the speakers. This will be non-null for a playback or + full-duplex device and null for a capture and loopback device. + +pInput (in) + A pointer to the buffer containing input data from a recording device. This will be non-null for a capture, full-duplex or loopback device and null for a + playback device. + +frameCount (in) + The number of PCM frames to process. Note that this will not necessarily be equal to what you requested when you initialized the device. The + `periodSizeInFrames` and `periodSizeInMilliseconds` members of the device config are just hints, and are not necessarily exactly what you'll get. You must + not assume this will always be the same value each time the callback is fired. + + +Remarks +------- +You cannot stop and start the device from inside the callback or else you'll get a deadlock. You must also not uninitialize the device from inside the +callback. The following APIs cannot be called from inside the callback: + + ma_device_init() + ma_device_init_ex() + ma_device_uninit() + ma_device_start() + ma_device_stop() + +The proper way to stop the device is to call `ma_device_stop()` from a different thread, normally the main application thread. +*/ +typedef void (* ma_device_data_proc)(ma_device* pDevice, void* pOutput, const void* pInput, ma_uint32 frameCount); + + + + +/* +DEPRECATED. Use ma_device_notification_proc instead. + +The callback for when the device has been stopped. + +This will be called when the device is stopped explicitly with `ma_device_stop()` and also called implicitly when the device is stopped through external forces +such as being unplugged or an internal error occurring. + + +Parameters +---------- +pDevice (in) + A pointer to the device that has just stopped. + + +Remarks +------- +Do not restart or uninitialize the device from the callback. +*/ +typedef void (* ma_stop_proc)(ma_device* pDevice); /* DEPRECATED. Use ma_device_notification_proc instead. */ + +typedef enum +{ + ma_device_type_playback = 1, + ma_device_type_capture = 2, + ma_device_type_duplex = ma_device_type_playback | ma_device_type_capture, /* 3 */ + ma_device_type_loopback = 4 +} ma_device_type; + +typedef enum +{ + ma_share_mode_shared = 0, + ma_share_mode_exclusive +} ma_share_mode; + +/* iOS/tvOS/watchOS session categories. */ +typedef enum +{ + ma_ios_session_category_default = 0, /* AVAudioSessionCategoryPlayAndRecord. */ + ma_ios_session_category_none, /* Leave the session category unchanged. */ + ma_ios_session_category_ambient, /* AVAudioSessionCategoryAmbient */ + ma_ios_session_category_solo_ambient, /* AVAudioSessionCategorySoloAmbient */ + ma_ios_session_category_playback, /* AVAudioSessionCategoryPlayback */ + ma_ios_session_category_record, /* AVAudioSessionCategoryRecord */ + ma_ios_session_category_play_and_record, /* AVAudioSessionCategoryPlayAndRecord */ + ma_ios_session_category_multi_route /* AVAudioSessionCategoryMultiRoute */ +} ma_ios_session_category; + +/* iOS/tvOS/watchOS session category options */ +typedef enum +{ + ma_ios_session_category_option_mix_with_others = 0x01, /* AVAudioSessionCategoryOptionMixWithOthers */ + ma_ios_session_category_option_duck_others = 0x02, /* AVAudioSessionCategoryOptionDuckOthers */ + ma_ios_session_category_option_allow_bluetooth = 0x04, /* AVAudioSessionCategoryOptionAllowBluetooth */ + ma_ios_session_category_option_default_to_speaker = 0x08, /* AVAudioSessionCategoryOptionDefaultToSpeaker */ + ma_ios_session_category_option_interrupt_spoken_audio_and_mix_with_others = 0x11, /* AVAudioSessionCategoryOptionInterruptSpokenAudioAndMixWithOthers */ + ma_ios_session_category_option_allow_bluetooth_a2dp = 0x20, /* AVAudioSessionCategoryOptionAllowBluetoothA2DP */ + ma_ios_session_category_option_allow_air_play = 0x40, /* AVAudioSessionCategoryOptionAllowAirPlay */ +} ma_ios_session_category_option; + +/* OpenSL stream types. */ +typedef enum +{ + ma_opensl_stream_type_default = 0, /* Leaves the stream type unset. */ + ma_opensl_stream_type_voice, /* SL_ANDROID_STREAM_VOICE */ + ma_opensl_stream_type_system, /* SL_ANDROID_STREAM_SYSTEM */ + ma_opensl_stream_type_ring, /* SL_ANDROID_STREAM_RING */ + ma_opensl_stream_type_media, /* SL_ANDROID_STREAM_MEDIA */ + ma_opensl_stream_type_alarm, /* SL_ANDROID_STREAM_ALARM */ + ma_opensl_stream_type_notification /* SL_ANDROID_STREAM_NOTIFICATION */ +} ma_opensl_stream_type; + +/* OpenSL recording presets. */ +typedef enum +{ + ma_opensl_recording_preset_default = 0, /* Leaves the input preset unset. */ + ma_opensl_recording_preset_generic, /* SL_ANDROID_RECORDING_PRESET_GENERIC */ + ma_opensl_recording_preset_camcorder, /* SL_ANDROID_RECORDING_PRESET_CAMCORDER */ + ma_opensl_recording_preset_voice_recognition, /* SL_ANDROID_RECORDING_PRESET_VOICE_RECOGNITION */ + ma_opensl_recording_preset_voice_communication, /* SL_ANDROID_RECORDING_PRESET_VOICE_COMMUNICATION */ + ma_opensl_recording_preset_voice_unprocessed /* SL_ANDROID_RECORDING_PRESET_UNPROCESSED */ +} ma_opensl_recording_preset; + +/* WASAPI audio thread priority characteristics. */ +typedef enum +{ + ma_wasapi_usage_default = 0, + ma_wasapi_usage_games, + ma_wasapi_usage_pro_audio, +} ma_wasapi_usage; + +/* AAudio usage types. */ +typedef enum +{ + ma_aaudio_usage_default = 0, /* Leaves the usage type unset. */ + ma_aaudio_usage_media, /* AAUDIO_USAGE_MEDIA */ + ma_aaudio_usage_voice_communication, /* AAUDIO_USAGE_VOICE_COMMUNICATION */ + ma_aaudio_usage_voice_communication_signalling, /* AAUDIO_USAGE_VOICE_COMMUNICATION_SIGNALLING */ + ma_aaudio_usage_alarm, /* AAUDIO_USAGE_ALARM */ + ma_aaudio_usage_notification, /* AAUDIO_USAGE_NOTIFICATION */ + ma_aaudio_usage_notification_ringtone, /* AAUDIO_USAGE_NOTIFICATION_RINGTONE */ + ma_aaudio_usage_notification_event, /* AAUDIO_USAGE_NOTIFICATION_EVENT */ + ma_aaudio_usage_assistance_accessibility, /* AAUDIO_USAGE_ASSISTANCE_ACCESSIBILITY */ + ma_aaudio_usage_assistance_navigation_guidance, /* AAUDIO_USAGE_ASSISTANCE_NAVIGATION_GUIDANCE */ + ma_aaudio_usage_assistance_sonification, /* AAUDIO_USAGE_ASSISTANCE_SONIFICATION */ + ma_aaudio_usage_game, /* AAUDIO_USAGE_GAME */ + ma_aaudio_usage_assitant, /* AAUDIO_USAGE_ASSISTANT */ + ma_aaudio_usage_emergency, /* AAUDIO_SYSTEM_USAGE_EMERGENCY */ + ma_aaudio_usage_safety, /* AAUDIO_SYSTEM_USAGE_SAFETY */ + ma_aaudio_usage_vehicle_status, /* AAUDIO_SYSTEM_USAGE_VEHICLE_STATUS */ + ma_aaudio_usage_announcement /* AAUDIO_SYSTEM_USAGE_ANNOUNCEMENT */ +} ma_aaudio_usage; + +/* AAudio content types. */ +typedef enum +{ + ma_aaudio_content_type_default = 0, /* Leaves the content type unset. */ + ma_aaudio_content_type_speech, /* AAUDIO_CONTENT_TYPE_SPEECH */ + ma_aaudio_content_type_music, /* AAUDIO_CONTENT_TYPE_MUSIC */ + ma_aaudio_content_type_movie, /* AAUDIO_CONTENT_TYPE_MOVIE */ + ma_aaudio_content_type_sonification /* AAUDIO_CONTENT_TYPE_SONIFICATION */ +} ma_aaudio_content_type; + +/* AAudio input presets. */ +typedef enum +{ + ma_aaudio_input_preset_default = 0, /* Leaves the input preset unset. */ + ma_aaudio_input_preset_generic, /* AAUDIO_INPUT_PRESET_GENERIC */ + ma_aaudio_input_preset_camcorder, /* AAUDIO_INPUT_PRESET_CAMCORDER */ + ma_aaudio_input_preset_voice_recognition, /* AAUDIO_INPUT_PRESET_VOICE_RECOGNITION */ + ma_aaudio_input_preset_voice_communication, /* AAUDIO_INPUT_PRESET_VOICE_COMMUNICATION */ + ma_aaudio_input_preset_unprocessed, /* AAUDIO_INPUT_PRESET_UNPROCESSED */ + ma_aaudio_input_preset_voice_performance /* AAUDIO_INPUT_PRESET_VOICE_PERFORMANCE */ +} ma_aaudio_input_preset; + +typedef enum +{ + ma_aaudio_allow_capture_default = 0, /* Leaves the allowed capture policy unset. */ + ma_aaudio_allow_capture_by_all, /* AAUDIO_ALLOW_CAPTURE_BY_ALL */ + ma_aaudio_allow_capture_by_system, /* AAUDIO_ALLOW_CAPTURE_BY_SYSTEM */ + ma_aaudio_allow_capture_by_none /* AAUDIO_ALLOW_CAPTURE_BY_NONE */ +} ma_aaudio_allowed_capture_policy; + +typedef union +{ + ma_int64 counter; + double counterD; +} ma_timer; + +typedef union +{ + ma_wchar_win32 wasapi[64]; /* WASAPI uses a wchar_t string for identification. */ + ma_uint8 dsound[16]; /* DirectSound uses a GUID for identification. */ + /*UINT_PTR*/ ma_uint32 winmm; /* When creating a device, WinMM expects a Win32 UINT_PTR for device identification. In practice it's actually just a UINT. */ + char alsa[256]; /* ALSA uses a name string for identification. */ + char pulse[256]; /* PulseAudio uses a name string for identification. */ + int jack; /* JACK always uses default devices. */ + char coreaudio[256]; /* Core Audio uses a string for identification. */ + char sndio[256]; /* "snd/0", etc. */ + char audio4[256]; /* "/dev/audio", etc. */ + char oss[64]; /* "dev/dsp0", etc. "dev/dsp" for the default device. */ + ma_int32 aaudio; /* AAudio uses a 32-bit integer for identification. */ + ma_uint32 opensl; /* OpenSL|ES uses a 32-bit unsigned integer for identification. */ + char webaudio[32]; /* Web Audio always uses default devices for now, but if this changes it'll be a GUID. */ + union + { + int i; + char s[256]; + void* p; + } custom; /* The custom backend could be anything. Give them a few options. */ + int nullbackend; /* The null backend uses an integer for device IDs. */ +} ma_device_id; + +MA_API ma_bool32 ma_device_id_equal(const ma_device_id* pA, const ma_device_id* pB); + + +typedef struct ma_context_config ma_context_config; +typedef struct ma_device_config ma_device_config; +typedef struct ma_backend_callbacks ma_backend_callbacks; + +#define MA_DATA_FORMAT_FLAG_EXCLUSIVE_MODE (1U << 1) /* If set, this is supported in exclusive mode. Otherwise not natively supported by exclusive mode. */ + +#ifndef MA_MAX_DEVICE_NAME_LENGTH +#define MA_MAX_DEVICE_NAME_LENGTH 255 +#endif + +typedef struct +{ + /* Basic info. This is the only information guaranteed to be filled in during device enumeration. */ + ma_device_id id; + char name[MA_MAX_DEVICE_NAME_LENGTH + 1]; /* +1 for null terminator. */ + ma_bool32 isDefault; + + ma_uint32 nativeDataFormatCount; + struct + { + ma_format format; /* Sample format. If set to ma_format_unknown, all sample formats are supported. */ + ma_uint32 channels; /* If set to 0, all channels are supported. */ + ma_uint32 sampleRate; /* If set to 0, all sample rates are supported. */ + ma_uint32 flags; /* A combination of MA_DATA_FORMAT_FLAG_* flags. */ + } nativeDataFormats[/*ma_format_count * ma_standard_sample_rate_count * MA_MAX_CHANNELS*/ 64]; /* Not sure how big to make this. There can be *many* permutations for virtual devices which can support anything. */ +} ma_device_info; + +struct ma_device_config +{ + ma_device_type deviceType; + ma_uint32 sampleRate; + ma_uint32 periodSizeInFrames; + ma_uint32 periodSizeInMilliseconds; + ma_uint32 periods; + ma_performance_profile performanceProfile; + ma_bool8 noPreSilencedOutputBuffer; /* When set to true, the contents of the output buffer passed into the data callback will be left undefined rather than initialized to silence. */ + ma_bool8 noClip; /* When set to true, the contents of the output buffer passed into the data callback will not be clipped after returning. Only applies when the playback sample format is f32. */ + ma_bool8 noDisableDenormals; /* Do not disable denormals when firing the data callback. */ + ma_bool8 noFixedSizedCallback; /* Disables strict fixed-sized data callbacks. Setting this to true will result in the period size being treated only as a hint to the backend. This is an optimization for those who don't need fixed sized callbacks. */ + ma_device_data_proc dataCallback; + ma_device_notification_proc notificationCallback; + ma_stop_proc stopCallback; + void* pUserData; + ma_resampler_config resampling; + struct + { + const ma_device_id* pDeviceID; + ma_format format; + ma_uint32 channels; + ma_channel* pChannelMap; + ma_channel_mix_mode channelMixMode; + ma_bool32 calculateLFEFromSpatialChannels; /* When an output LFE channel is present, but no input LFE, set to true to set the output LFE to the average of all spatial channels (LR, FR, etc.). Ignored when an input LFE is present. */ + ma_share_mode shareMode; + } playback; + struct + { + const ma_device_id* pDeviceID; + ma_format format; + ma_uint32 channels; + ma_channel* pChannelMap; + ma_channel_mix_mode channelMixMode; + ma_bool32 calculateLFEFromSpatialChannels; /* When an output LFE channel is present, but no input LFE, set to true to set the output LFE to the average of all spatial channels (LR, FR, etc.). Ignored when an input LFE is present. */ + ma_share_mode shareMode; + } capture; + + struct + { + ma_wasapi_usage usage; /* When configured, uses Avrt APIs to set the thread characteristics. */ + ma_bool8 noAutoConvertSRC; /* When set to true, disables the use of AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM. */ + ma_bool8 noDefaultQualitySRC; /* When set to true, disables the use of AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY. */ + ma_bool8 noAutoStreamRouting; /* Disables automatic stream routing. */ + ma_bool8 noHardwareOffloading; /* Disables WASAPI's hardware offloading feature. */ + ma_uint32 loopbackProcessID; /* The process ID to include or exclude for loopback mode. Set to 0 to capture audio from all processes. Ignored when an explicit device ID is specified. */ + ma_bool8 loopbackProcessExclude; /* When set to true, excludes the process specified by loopbackProcessID. By default, the process will be included. */ + } wasapi; + struct + { + ma_bool32 noMMap; /* Disables MMap mode. */ + ma_bool32 noAutoFormat; /* Opens the ALSA device with SND_PCM_NO_AUTO_FORMAT. */ + ma_bool32 noAutoChannels; /* Opens the ALSA device with SND_PCM_NO_AUTO_CHANNELS. */ + ma_bool32 noAutoResample; /* Opens the ALSA device with SND_PCM_NO_AUTO_RESAMPLE. */ + } alsa; + struct + { + const char* pStreamNamePlayback; + const char* pStreamNameCapture; + int channelMap; + } pulse; + struct + { + ma_bool32 allowNominalSampleRateChange; /* Desktop only. When enabled, allows changing of the sample rate at the operating system level. */ + } coreaudio; + struct + { + ma_opensl_stream_type streamType; + ma_opensl_recording_preset recordingPreset; + ma_bool32 enableCompatibilityWorkarounds; + } opensl; + struct + { + ma_aaudio_usage usage; + ma_aaudio_content_type contentType; + ma_aaudio_input_preset inputPreset; + ma_aaudio_allowed_capture_policy allowedCapturePolicy; + ma_bool32 noAutoStartAfterReroute; + ma_bool32 enableCompatibilityWorkarounds; + ma_bool32 allowSetBufferCapacity; + } aaudio; +}; + + +/* +The callback for handling device enumeration. This is fired from `ma_context_enumerate_devices()`. + + +Parameters +---------- +pContext (in) + A pointer to the context performing the enumeration. + +deviceType (in) + The type of the device being enumerated. This will always be either `ma_device_type_playback` or `ma_device_type_capture`. + +pInfo (in) + A pointer to a `ma_device_info` containing the ID and name of the enumerated device. Note that this will not include detailed information about the device, + only basic information (ID and name). The reason for this is that it would otherwise require opening the backend device to probe for the information which + is too inefficient. + +pUserData (in) + The user data pointer passed into `ma_context_enumerate_devices()`. +*/ +typedef ma_bool32 (* ma_enum_devices_callback_proc)(ma_context* pContext, ma_device_type deviceType, const ma_device_info* pInfo, void* pUserData); + + +/* +Describes some basic details about a playback or capture device. +*/ +typedef struct +{ + const ma_device_id* pDeviceID; + ma_share_mode shareMode; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_channel channelMap[MA_MAX_CHANNELS]; + ma_uint32 periodSizeInFrames; + ma_uint32 periodSizeInMilliseconds; + ma_uint32 periodCount; +} ma_device_descriptor; + +/* +These are the callbacks required to be implemented for a backend. These callbacks are grouped into two parts: context and device. There is one context +to many devices. A device is created from a context. + +The general flow goes like this: + + 1) A context is created with `onContextInit()` + 1a) Available devices can be enumerated with `onContextEnumerateDevices()` if required. + 1b) Detailed information about a device can be queried with `onContextGetDeviceInfo()` if required. + 2) A device is created from the context that was created in the first step using `onDeviceInit()`, and optionally a device ID that was + selected from device enumeration via `onContextEnumerateDevices()`. + 3) A device is started or stopped with `onDeviceStart()` / `onDeviceStop()` + 4) Data is delivered to and from the device by the backend. This is always done based on the native format returned by the prior call + to `onDeviceInit()`. Conversion between the device's native format and the format requested by the application will be handled by + miniaudio internally. + +Initialization of the context is quite simple. You need to do any necessary initialization of internal objects and then output the +callbacks defined in this structure. + +Once the context has been initialized you can initialize a device. Before doing so, however, the application may want to know which +physical devices are available. This is where `onContextEnumerateDevices()` comes in. This is fairly simple. For each device, fire the +given callback with, at a minimum, the basic information filled out in `ma_device_info`. When the callback returns `MA_FALSE`, enumeration +needs to stop and the `onContextEnumerateDevices()` function returns with a success code. + +Detailed device information can be retrieved from a device ID using `onContextGetDeviceInfo()`. This takes as input the device type and ID, +and on output returns detailed information about the device in `ma_device_info`. The `onContextGetDeviceInfo()` callback must handle the +case when the device ID is NULL, in which case information about the default device needs to be retrieved. + +Once the context has been created and the device ID retrieved (if using anything other than the default device), the device can be created. +This is a little bit more complicated than initialization of the context due to its more complicated configuration. When initializing a +device, a duplex device may be requested. This means a separate data format needs to be specified for both playback and capture. On input, +the data format is set to what the application wants. On output it's set to the native format which should match as closely as possible to +the requested format. The conversion between the format requested by the application and the device's native format will be handled +internally by miniaudio. + +On input, if the sample format is set to `ma_format_unknown`, the backend is free to use whatever sample format it desires, so long as it's +supported by miniaudio. When the channel count is set to 0, the backend should use the device's native channel count. The same applies for +sample rate. For the channel map, the default should be used when `ma_channel_map_is_blank()` returns true (all channels set to +`MA_CHANNEL_NONE`). On input, the `periodSizeInFrames` or `periodSizeInMilliseconds` option should always be set. The backend should +inspect both of these variables. If `periodSizeInFrames` is set, it should take priority, otherwise it needs to be derived from the period +size in milliseconds (`periodSizeInMilliseconds`) and the sample rate, keeping in mind that the sample rate may be 0, in which case the +sample rate will need to be determined before calculating the period size in frames. On output, all members of the `ma_device_descriptor` +object should be set to a valid value, except for `periodSizeInMilliseconds` which is optional (`periodSizeInFrames` *must* be set). + +Starting and stopping of the device is done with `onDeviceStart()` and `onDeviceStop()` and should be self-explanatory. If the backend uses +asynchronous reading and writing, `onDeviceStart()` and `onDeviceStop()` should always be implemented. + +The handling of data delivery between the application and the device is the most complicated part of the process. To make this a bit +easier, some helper callbacks are available. If the backend uses a blocking read/write style of API, the `onDeviceRead()` and +`onDeviceWrite()` callbacks can optionally be implemented. These are blocking and work just like reading and writing from a file. If the +backend uses a callback for data delivery, that callback must call `ma_device_handle_backend_data_callback()` from within its callback. +This allows miniaudio to then process any necessary data conversion and then pass it to the miniaudio data callback. + +If the backend requires absolute flexibility with its data delivery, it can optionally implement the `onDeviceDataLoop()` callback +which will allow it to implement the logic that will run on the audio thread. This is much more advanced and is completely optional. + +The audio thread should run data delivery logic in a loop while `ma_device_get_state() == ma_device_state_started` and no errors have been +encountered. Do not start or stop the device here. That will be handled from outside the `onDeviceDataLoop()` callback. + +The invocation of the `onDeviceDataLoop()` callback will be handled by miniaudio. When you start the device, miniaudio will fire this +callback. When the device is stopped, the `ma_device_get_state() == ma_device_state_started` condition will fail and the loop will be terminated +which will then fall through to the part that stops the device. For an example on how to implement the `onDeviceDataLoop()` callback, +look at `ma_device_audio_thread__default_read_write()`. Implement the `onDeviceDataLoopWakeup()` callback if you need a mechanism to +wake up the audio thread. + +If the backend supports an optimized retrieval of device information from an initialized `ma_device` object, it should implement the +`onDeviceGetInfo()` callback. This is optional, in which case it will fall back to `onContextGetDeviceInfo()` which is less efficient. +*/ +struct ma_backend_callbacks +{ + ma_result (* onContextInit)(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks); + ma_result (* onContextUninit)(ma_context* pContext); + ma_result (* onContextEnumerateDevices)(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData); + ma_result (* onContextGetDeviceInfo)(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo); + ma_result (* onDeviceInit)(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture); + ma_result (* onDeviceUninit)(ma_device* pDevice); + ma_result (* onDeviceStart)(ma_device* pDevice); + ma_result (* onDeviceStop)(ma_device* pDevice); + ma_result (* onDeviceRead)(ma_device* pDevice, void* pFrames, ma_uint32 frameCount, ma_uint32* pFramesRead); + ma_result (* onDeviceWrite)(ma_device* pDevice, const void* pFrames, ma_uint32 frameCount, ma_uint32* pFramesWritten); + ma_result (* onDeviceDataLoop)(ma_device* pDevice); + ma_result (* onDeviceDataLoopWakeup)(ma_device* pDevice); + ma_result (* onDeviceGetInfo)(ma_device* pDevice, ma_device_type type, ma_device_info* pDeviceInfo); +}; + +struct ma_context_config +{ + ma_log* pLog; + ma_thread_priority threadPriority; + size_t threadStackSize; + void* pUserData; + ma_allocation_callbacks allocationCallbacks; + struct + { + ma_handle hWnd; /* HWND. Optional window handle to pass into SetCooperativeLevel(). Will default to the foreground window, and if that fails, the desktop window. */ + } dsound; + struct + { + ma_bool32 useVerboseDeviceEnumeration; + } alsa; + struct + { + const char* pApplicationName; + const char* pServerName; + ma_bool32 tryAutoSpawn; /* Enables autospawning of the PulseAudio daemon if necessary. */ + } pulse; + struct + { + ma_ios_session_category sessionCategory; + ma_uint32 sessionCategoryOptions; + ma_bool32 noAudioSessionActivate; /* iOS only. When set to true, does not perform an explicit [[AVAudioSession sharedInstace] setActive:true] on initialization. */ + ma_bool32 noAudioSessionDeactivate; /* iOS only. When set to true, does not perform an explicit [[AVAudioSession sharedInstace] setActive:false] on uninitialization. */ + } coreaudio; + struct + { + const char* pClientName; + ma_bool32 tryStartServer; + } jack; + ma_backend_callbacks custom; +}; + +/* WASAPI specific structure for some commands which must run on a common thread due to bugs in WASAPI. */ +typedef struct +{ + int code; + ma_event* pEvent; /* This will be signalled when the event is complete. */ + union + { + struct + { + int _unused; + } quit; + struct + { + ma_device_type deviceType; + void* pAudioClient; + void** ppAudioClientService; + ma_result* pResult; /* The result from creating the audio client service. */ + } createAudioClient; + struct + { + ma_device* pDevice; + ma_device_type deviceType; + } releaseAudioClient; + } data; +} ma_context_command__wasapi; + +struct ma_context +{ + ma_backend_callbacks callbacks; + ma_backend backend; /* DirectSound, ALSA, etc. */ + ma_log* pLog; + ma_log log; /* Only used if the log is owned by the context. The pLog member will be set to &log in this case. */ + ma_thread_priority threadPriority; + size_t threadStackSize; + void* pUserData; + ma_allocation_callbacks allocationCallbacks; + ma_mutex deviceEnumLock; /* Used to make ma_context_get_devices() thread safe. */ + ma_mutex deviceInfoLock; /* Used to make ma_context_get_device_info() thread safe. */ + ma_uint32 deviceInfoCapacity; /* Total capacity of pDeviceInfos. */ + ma_uint32 playbackDeviceInfoCount; + ma_uint32 captureDeviceInfoCount; + ma_device_info* pDeviceInfos; /* Playback devices first, then capture. */ + + union + { +#ifdef MA_SUPPORT_WASAPI + struct + { + ma_thread commandThread; + ma_mutex commandLock; + ma_semaphore commandSem; + ma_uint32 commandIndex; + ma_uint32 commandCount; + ma_context_command__wasapi commands[4]; + ma_handle hAvrt; + ma_proc AvSetMmThreadCharacteristicsA; + ma_proc AvRevertMmThreadcharacteristics; + ma_handle hMMDevapi; + ma_proc ActivateAudioInterfaceAsync; + } wasapi; +#endif +#ifdef MA_SUPPORT_DSOUND + struct + { + ma_handle hWnd; /* Can be null. */ + ma_handle hDSoundDLL; + ma_proc DirectSoundCreate; + ma_proc DirectSoundEnumerateA; + ma_proc DirectSoundCaptureCreate; + ma_proc DirectSoundCaptureEnumerateA; + } dsound; +#endif +#ifdef MA_SUPPORT_WINMM + struct + { + ma_handle hWinMM; + ma_proc waveOutGetNumDevs; + ma_proc waveOutGetDevCapsA; + ma_proc waveOutOpen; + ma_proc waveOutClose; + ma_proc waveOutPrepareHeader; + ma_proc waveOutUnprepareHeader; + ma_proc waveOutWrite; + ma_proc waveOutReset; + ma_proc waveInGetNumDevs; + ma_proc waveInGetDevCapsA; + ma_proc waveInOpen; + ma_proc waveInClose; + ma_proc waveInPrepareHeader; + ma_proc waveInUnprepareHeader; + ma_proc waveInAddBuffer; + ma_proc waveInStart; + ma_proc waveInReset; + } winmm; +#endif +#ifdef MA_SUPPORT_ALSA + struct + { + ma_handle asoundSO; + ma_proc snd_pcm_open; + ma_proc snd_pcm_close; + ma_proc snd_pcm_hw_params_sizeof; + ma_proc snd_pcm_hw_params_any; + ma_proc snd_pcm_hw_params_set_format; + ma_proc snd_pcm_hw_params_set_format_first; + ma_proc snd_pcm_hw_params_get_format_mask; + ma_proc snd_pcm_hw_params_set_channels; + ma_proc snd_pcm_hw_params_set_channels_near; + ma_proc snd_pcm_hw_params_set_channels_minmax; + ma_proc snd_pcm_hw_params_set_rate_resample; + ma_proc snd_pcm_hw_params_set_rate; + ma_proc snd_pcm_hw_params_set_rate_near; + ma_proc snd_pcm_hw_params_set_buffer_size_near; + ma_proc snd_pcm_hw_params_set_periods_near; + ma_proc snd_pcm_hw_params_set_access; + ma_proc snd_pcm_hw_params_get_format; + ma_proc snd_pcm_hw_params_get_channels; + ma_proc snd_pcm_hw_params_get_channels_min; + ma_proc snd_pcm_hw_params_get_channels_max; + ma_proc snd_pcm_hw_params_get_rate; + ma_proc snd_pcm_hw_params_get_rate_min; + ma_proc snd_pcm_hw_params_get_rate_max; + ma_proc snd_pcm_hw_params_get_buffer_size; + ma_proc snd_pcm_hw_params_get_periods; + ma_proc snd_pcm_hw_params_get_access; + ma_proc snd_pcm_hw_params_test_format; + ma_proc snd_pcm_hw_params_test_channels; + ma_proc snd_pcm_hw_params_test_rate; + ma_proc snd_pcm_hw_params; + ma_proc snd_pcm_sw_params_sizeof; + ma_proc snd_pcm_sw_params_current; + ma_proc snd_pcm_sw_params_get_boundary; + ma_proc snd_pcm_sw_params_set_avail_min; + ma_proc snd_pcm_sw_params_set_start_threshold; + ma_proc snd_pcm_sw_params_set_stop_threshold; + ma_proc snd_pcm_sw_params; + ma_proc snd_pcm_format_mask_sizeof; + ma_proc snd_pcm_format_mask_test; + ma_proc snd_pcm_get_chmap; + ma_proc snd_pcm_state; + ma_proc snd_pcm_prepare; + ma_proc snd_pcm_start; + ma_proc snd_pcm_drop; + ma_proc snd_pcm_drain; + ma_proc snd_pcm_reset; + ma_proc snd_device_name_hint; + ma_proc snd_device_name_get_hint; + ma_proc snd_card_get_index; + ma_proc snd_device_name_free_hint; + ma_proc snd_pcm_mmap_begin; + ma_proc snd_pcm_mmap_commit; + ma_proc snd_pcm_recover; + ma_proc snd_pcm_readi; + ma_proc snd_pcm_writei; + ma_proc snd_pcm_avail; + ma_proc snd_pcm_avail_update; + ma_proc snd_pcm_wait; + ma_proc snd_pcm_nonblock; + ma_proc snd_pcm_info; + ma_proc snd_pcm_info_sizeof; + ma_proc snd_pcm_info_get_name; + ma_proc snd_pcm_poll_descriptors; + ma_proc snd_pcm_poll_descriptors_count; + ma_proc snd_pcm_poll_descriptors_revents; + ma_proc snd_config_update_free_global; + + ma_mutex internalDeviceEnumLock; + ma_bool32 useVerboseDeviceEnumeration; + } alsa; +#endif +#ifdef MA_SUPPORT_PULSEAUDIO + struct + { + ma_handle pulseSO; + ma_proc pa_mainloop_new; + ma_proc pa_mainloop_free; + ma_proc pa_mainloop_quit; + ma_proc pa_mainloop_get_api; + ma_proc pa_mainloop_iterate; + ma_proc pa_mainloop_wakeup; + ma_proc pa_threaded_mainloop_new; + ma_proc pa_threaded_mainloop_free; + ma_proc pa_threaded_mainloop_start; + ma_proc pa_threaded_mainloop_stop; + ma_proc pa_threaded_mainloop_lock; + ma_proc pa_threaded_mainloop_unlock; + ma_proc pa_threaded_mainloop_wait; + ma_proc pa_threaded_mainloop_signal; + ma_proc pa_threaded_mainloop_accept; + ma_proc pa_threaded_mainloop_get_retval; + ma_proc pa_threaded_mainloop_get_api; + ma_proc pa_threaded_mainloop_in_thread; + ma_proc pa_threaded_mainloop_set_name; + ma_proc pa_context_new; + ma_proc pa_context_unref; + ma_proc pa_context_connect; + ma_proc pa_context_disconnect; + ma_proc pa_context_set_state_callback; + ma_proc pa_context_get_state; + ma_proc pa_context_get_sink_info_list; + ma_proc pa_context_get_source_info_list; + ma_proc pa_context_get_sink_info_by_name; + ma_proc pa_context_get_source_info_by_name; + ma_proc pa_operation_unref; + ma_proc pa_operation_get_state; + ma_proc pa_channel_map_init_extend; + ma_proc pa_channel_map_valid; + ma_proc pa_channel_map_compatible; + ma_proc pa_stream_new; + ma_proc pa_stream_unref; + ma_proc pa_stream_connect_playback; + ma_proc pa_stream_connect_record; + ma_proc pa_stream_disconnect; + ma_proc pa_stream_get_state; + ma_proc pa_stream_get_sample_spec; + ma_proc pa_stream_get_channel_map; + ma_proc pa_stream_get_buffer_attr; + ma_proc pa_stream_set_buffer_attr; + ma_proc pa_stream_get_device_name; + ma_proc pa_stream_set_write_callback; + ma_proc pa_stream_set_read_callback; + ma_proc pa_stream_set_suspended_callback; + ma_proc pa_stream_set_moved_callback; + ma_proc pa_stream_is_suspended; + ma_proc pa_stream_flush; + ma_proc pa_stream_drain; + ma_proc pa_stream_is_corked; + ma_proc pa_stream_cork; + ma_proc pa_stream_trigger; + ma_proc pa_stream_begin_write; + ma_proc pa_stream_write; + ma_proc pa_stream_peek; + ma_proc pa_stream_drop; + ma_proc pa_stream_writable_size; + ma_proc pa_stream_readable_size; + + /*pa_mainloop**/ ma_ptr pMainLoop; + /*pa_context**/ ma_ptr pPulseContext; + char* pApplicationName; /* Set when the context is initialized. Used by devices for their local pa_context objects. */ + char* pServerName; /* Set when the context is initialized. Used by devices for their local pa_context objects. */ + } pulse; +#endif +#ifdef MA_SUPPORT_JACK + struct + { + ma_handle jackSO; + ma_proc jack_client_open; + ma_proc jack_client_close; + ma_proc jack_client_name_size; + ma_proc jack_set_process_callback; + ma_proc jack_set_buffer_size_callback; + ma_proc jack_on_shutdown; + ma_proc jack_get_sample_rate; + ma_proc jack_get_buffer_size; + ma_proc jack_get_ports; + ma_proc jack_activate; + ma_proc jack_deactivate; + ma_proc jack_connect; + ma_proc jack_port_register; + ma_proc jack_port_name; + ma_proc jack_port_get_buffer; + ma_proc jack_free; + + char* pClientName; + ma_bool32 tryStartServer; + } jack; +#endif +#ifdef MA_SUPPORT_COREAUDIO + struct + { + ma_handle hCoreFoundation; + ma_proc CFStringGetCString; + ma_proc CFRelease; + + ma_handle hCoreAudio; + ma_proc AudioObjectGetPropertyData; + ma_proc AudioObjectGetPropertyDataSize; + ma_proc AudioObjectSetPropertyData; + ma_proc AudioObjectAddPropertyListener; + ma_proc AudioObjectRemovePropertyListener; + + ma_handle hAudioUnit; /* Could possibly be set to AudioToolbox on later versions of macOS. */ + ma_proc AudioComponentFindNext; + ma_proc AudioComponentInstanceDispose; + ma_proc AudioComponentInstanceNew; + ma_proc AudioOutputUnitStart; + ma_proc AudioOutputUnitStop; + ma_proc AudioUnitAddPropertyListener; + ma_proc AudioUnitGetPropertyInfo; + ma_proc AudioUnitGetProperty; + ma_proc AudioUnitSetProperty; + ma_proc AudioUnitInitialize; + ma_proc AudioUnitRender; + + /*AudioComponent*/ ma_ptr component; + ma_bool32 noAudioSessionDeactivate; /* For tracking whether or not the iOS audio session should be explicitly deactivated. Set from the config in ma_context_init__coreaudio(). */ + } coreaudio; +#endif +#ifdef MA_SUPPORT_SNDIO + struct + { + ma_handle sndioSO; + ma_proc sio_open; + ma_proc sio_close; + ma_proc sio_setpar; + ma_proc sio_getpar; + ma_proc sio_getcap; + ma_proc sio_start; + ma_proc sio_stop; + ma_proc sio_read; + ma_proc sio_write; + ma_proc sio_onmove; + ma_proc sio_nfds; + ma_proc sio_pollfd; + ma_proc sio_revents; + ma_proc sio_eof; + ma_proc sio_setvol; + ma_proc sio_onvol; + ma_proc sio_initpar; + } sndio; +#endif +#ifdef MA_SUPPORT_AUDIO4 + struct + { + int _unused; + } audio4; +#endif +#ifdef MA_SUPPORT_OSS + struct + { + int versionMajor; + int versionMinor; + } oss; +#endif +#ifdef MA_SUPPORT_AAUDIO + struct + { + ma_handle hAAudio; /* libaaudio.so */ + ma_proc AAudio_createStreamBuilder; + ma_proc AAudioStreamBuilder_delete; + ma_proc AAudioStreamBuilder_setDeviceId; + ma_proc AAudioStreamBuilder_setDirection; + ma_proc AAudioStreamBuilder_setSharingMode; + ma_proc AAudioStreamBuilder_setFormat; + ma_proc AAudioStreamBuilder_setChannelCount; + ma_proc AAudioStreamBuilder_setSampleRate; + ma_proc AAudioStreamBuilder_setBufferCapacityInFrames; + ma_proc AAudioStreamBuilder_setFramesPerDataCallback; + ma_proc AAudioStreamBuilder_setDataCallback; + ma_proc AAudioStreamBuilder_setErrorCallback; + ma_proc AAudioStreamBuilder_setPerformanceMode; + ma_proc AAudioStreamBuilder_setUsage; + ma_proc AAudioStreamBuilder_setContentType; + ma_proc AAudioStreamBuilder_setInputPreset; + ma_proc AAudioStreamBuilder_setAllowedCapturePolicy; + ma_proc AAudioStreamBuilder_openStream; + ma_proc AAudioStream_close; + ma_proc AAudioStream_getState; + ma_proc AAudioStream_waitForStateChange; + ma_proc AAudioStream_getFormat; + ma_proc AAudioStream_getChannelCount; + ma_proc AAudioStream_getSampleRate; + ma_proc AAudioStream_getBufferCapacityInFrames; + ma_proc AAudioStream_getFramesPerDataCallback; + ma_proc AAudioStream_getFramesPerBurst; + ma_proc AAudioStream_requestStart; + ma_proc AAudioStream_requestStop; + ma_device_job_thread jobThread; /* For processing operations outside of the error callback, specifically device disconnections and rerouting. */ + } aaudio; +#endif +#ifdef MA_SUPPORT_OPENSL + struct + { + ma_handle libOpenSLES; + ma_handle SL_IID_ENGINE; + ma_handle SL_IID_AUDIOIODEVICECAPABILITIES; + ma_handle SL_IID_ANDROIDSIMPLEBUFFERQUEUE; + ma_handle SL_IID_RECORD; + ma_handle SL_IID_PLAY; + ma_handle SL_IID_OUTPUTMIX; + ma_handle SL_IID_ANDROIDCONFIGURATION; + ma_proc slCreateEngine; + } opensl; +#endif +#ifdef MA_SUPPORT_WEBAUDIO + struct + { + int _unused; + } webaudio; +#endif +#ifdef MA_SUPPORT_NULL + struct + { + int _unused; + } null_backend; +#endif + }; + + union + { +#if defined(MA_WIN32) + struct + { + /*HMODULE*/ ma_handle hOle32DLL; + ma_proc CoInitialize; + ma_proc CoInitializeEx; + ma_proc CoUninitialize; + ma_proc CoCreateInstance; + ma_proc CoTaskMemFree; + ma_proc PropVariantClear; + ma_proc StringFromGUID2; + + /*HMODULE*/ ma_handle hUser32DLL; + ma_proc GetForegroundWindow; + ma_proc GetDesktopWindow; + + /*HMODULE*/ ma_handle hAdvapi32DLL; + ma_proc RegOpenKeyExA; + ma_proc RegCloseKey; + ma_proc RegQueryValueExA; + + /*HRESULT*/ long CoInitializeResult; + } win32; +#endif +#ifdef MA_POSIX + struct + { + int _unused; + } posix; +#endif + int _unused; + }; +}; + +struct ma_device +{ + ma_context* pContext; + ma_device_type type; + ma_uint32 sampleRate; + ma_atomic_device_state state; /* The state of the device is variable and can change at any time on any thread. Must be used atomically. */ + ma_device_data_proc onData; /* Set once at initialization time and should not be changed after. */ + ma_device_notification_proc onNotification; /* Set once at initialization time and should not be changed after. */ + ma_stop_proc onStop; /* DEPRECATED. Use the notification callback instead. Set once at initialization time and should not be changed after. */ + void* pUserData; /* Application defined data. */ + ma_mutex startStopLock; + ma_event wakeupEvent; + ma_event startEvent; + ma_event stopEvent; + ma_thread thread; + ma_result workResult; /* This is set by the worker thread after it's finished doing a job. */ + ma_bool8 isOwnerOfContext; /* When set to true, uninitializing the device will also uninitialize the context. Set to true when NULL is passed into ma_device_init(). */ + ma_bool8 noPreSilencedOutputBuffer; + ma_bool8 noClip; + ma_bool8 noDisableDenormals; + ma_bool8 noFixedSizedCallback; + ma_atomic_float masterVolumeFactor; /* Linear 0..1. Can be read and written simultaneously by different threads. Must be used atomically. */ + ma_duplex_rb duplexRB; /* Intermediary buffer for duplex device on asynchronous backends. */ + struct + { + ma_resample_algorithm algorithm; + ma_resampling_backend_vtable* pBackendVTable; + void* pBackendUserData; + struct + { + ma_uint32 lpfOrder; + } linear; + } resampling; + struct + { + ma_device_id* pID; /* Set to NULL if using default ID, otherwise set to the address of "id". */ + ma_device_id id; /* If using an explicit device, will be set to a copy of the ID used for initialization. Otherwise cleared to 0. */ + char name[MA_MAX_DEVICE_NAME_LENGTH + 1]; /* Maybe temporary. Likely to be replaced with a query API. */ + ma_share_mode shareMode; /* Set to whatever was passed in when the device was initialized. */ + ma_format format; + ma_uint32 channels; + ma_channel channelMap[MA_MAX_CHANNELS]; + ma_format internalFormat; + ma_uint32 internalChannels; + ma_uint32 internalSampleRate; + ma_channel internalChannelMap[MA_MAX_CHANNELS]; + ma_uint32 internalPeriodSizeInFrames; + ma_uint32 internalPeriods; + ma_channel_mix_mode channelMixMode; + ma_bool32 calculateLFEFromSpatialChannels; + ma_data_converter converter; + void* pIntermediaryBuffer; /* For implementing fixed sized buffer callbacks. Will be null if using variable sized callbacks. */ + ma_uint32 intermediaryBufferCap; + ma_uint32 intermediaryBufferLen; /* How many valid frames are sitting in the intermediary buffer. */ + void* pInputCache; /* In external format. Can be null. */ + ma_uint64 inputCacheCap; + ma_uint64 inputCacheConsumed; + ma_uint64 inputCacheRemaining; + } playback; + struct + { + ma_device_id* pID; /* Set to NULL if using default ID, otherwise set to the address of "id". */ + ma_device_id id; /* If using an explicit device, will be set to a copy of the ID used for initialization. Otherwise cleared to 0. */ + char name[MA_MAX_DEVICE_NAME_LENGTH + 1]; /* Maybe temporary. Likely to be replaced with a query API. */ + ma_share_mode shareMode; /* Set to whatever was passed in when the device was initialized. */ + ma_format format; + ma_uint32 channels; + ma_channel channelMap[MA_MAX_CHANNELS]; + ma_format internalFormat; + ma_uint32 internalChannels; + ma_uint32 internalSampleRate; + ma_channel internalChannelMap[MA_MAX_CHANNELS]; + ma_uint32 internalPeriodSizeInFrames; + ma_uint32 internalPeriods; + ma_channel_mix_mode channelMixMode; + ma_bool32 calculateLFEFromSpatialChannels; + ma_data_converter converter; + void* pIntermediaryBuffer; /* For implementing fixed sized buffer callbacks. Will be null if using variable sized callbacks. */ + ma_uint32 intermediaryBufferCap; + ma_uint32 intermediaryBufferLen; /* How many valid frames are sitting in the intermediary buffer. */ + } capture; + + union + { +#ifdef MA_SUPPORT_WASAPI + struct + { + /*IAudioClient**/ ma_ptr pAudioClientPlayback; + /*IAudioClient**/ ma_ptr pAudioClientCapture; + /*IAudioRenderClient**/ ma_ptr pRenderClient; + /*IAudioCaptureClient**/ ma_ptr pCaptureClient; + /*IMMDeviceEnumerator**/ ma_ptr pDeviceEnumerator; /* Used for IMMNotificationClient notifications. Required for detecting default device changes. */ + ma_IMMNotificationClient notificationClient; + /*HANDLE*/ ma_handle hEventPlayback; /* Auto reset. Initialized to signaled. */ + /*HANDLE*/ ma_handle hEventCapture; /* Auto reset. Initialized to unsignaled. */ + ma_uint32 actualBufferSizeInFramesPlayback; /* Value from GetBufferSize(). internalPeriodSizeInFrames is not set to the _actual_ buffer size when low-latency shared mode is being used due to the way the IAudioClient3 API works. */ + ma_uint32 actualBufferSizeInFramesCapture; + ma_uint32 originalPeriodSizeInFrames; + ma_uint32 originalPeriodSizeInMilliseconds; + ma_uint32 originalPeriods; + ma_performance_profile originalPerformanceProfile; + ma_uint32 periodSizeInFramesPlayback; + ma_uint32 periodSizeInFramesCapture; + void* pMappedBufferCapture; + ma_uint32 mappedBufferCaptureCap; + ma_uint32 mappedBufferCaptureLen; + void* pMappedBufferPlayback; + ma_uint32 mappedBufferPlaybackCap; + ma_uint32 mappedBufferPlaybackLen; + ma_atomic_bool32 isStartedCapture; /* Can be read and written simultaneously across different threads. Must be used atomically, and must be 32-bit. */ + ma_atomic_bool32 isStartedPlayback; /* Can be read and written simultaneously across different threads. Must be used atomically, and must be 32-bit. */ + ma_uint32 loopbackProcessID; + ma_bool8 loopbackProcessExclude; + ma_bool8 noAutoConvertSRC; /* When set to true, disables the use of AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM. */ + ma_bool8 noDefaultQualitySRC; /* When set to true, disables the use of AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY. */ + ma_bool8 noHardwareOffloading; + ma_bool8 allowCaptureAutoStreamRouting; + ma_bool8 allowPlaybackAutoStreamRouting; + ma_bool8 isDetachedPlayback; + ma_bool8 isDetachedCapture; + ma_wasapi_usage usage; + void* hAvrtHandle; + ma_mutex rerouteLock; + } wasapi; +#endif +#ifdef MA_SUPPORT_DSOUND + struct + { + /*LPDIRECTSOUND*/ ma_ptr pPlayback; + /*LPDIRECTSOUNDBUFFER*/ ma_ptr pPlaybackPrimaryBuffer; + /*LPDIRECTSOUNDBUFFER*/ ma_ptr pPlaybackBuffer; + /*LPDIRECTSOUNDCAPTURE*/ ma_ptr pCapture; + /*LPDIRECTSOUNDCAPTUREBUFFER*/ ma_ptr pCaptureBuffer; + } dsound; +#endif +#ifdef MA_SUPPORT_WINMM + struct + { + /*HWAVEOUT*/ ma_handle hDevicePlayback; + /*HWAVEIN*/ ma_handle hDeviceCapture; + /*HANDLE*/ ma_handle hEventPlayback; + /*HANDLE*/ ma_handle hEventCapture; + ma_uint32 fragmentSizeInFrames; + ma_uint32 iNextHeaderPlayback; /* [0,periods). Used as an index into pWAVEHDRPlayback. */ + ma_uint32 iNextHeaderCapture; /* [0,periods). Used as an index into pWAVEHDRCapture. */ + ma_uint32 headerFramesConsumedPlayback; /* The number of PCM frames consumed in the buffer in pWAVEHEADER[iNextHeader]. */ + ma_uint32 headerFramesConsumedCapture; /* ^^^ */ + /*WAVEHDR**/ ma_uint8* pWAVEHDRPlayback; /* One instantiation for each period. */ + /*WAVEHDR**/ ma_uint8* pWAVEHDRCapture; /* One instantiation for each period. */ + ma_uint8* pIntermediaryBufferPlayback; + ma_uint8* pIntermediaryBufferCapture; + ma_uint8* _pHeapData; /* Used internally and is used for the heap allocated data for the intermediary buffer and the WAVEHDR structures. */ + } winmm; +#endif +#ifdef MA_SUPPORT_ALSA + struct + { + /*snd_pcm_t**/ ma_ptr pPCMPlayback; + /*snd_pcm_t**/ ma_ptr pPCMCapture; + /*struct pollfd**/ void* pPollDescriptorsPlayback; + /*struct pollfd**/ void* pPollDescriptorsCapture; + int pollDescriptorCountPlayback; + int pollDescriptorCountCapture; + int wakeupfdPlayback; /* eventfd for waking up from poll() when the playback device is stopped. */ + int wakeupfdCapture; /* eventfd for waking up from poll() when the capture device is stopped. */ + ma_bool8 isUsingMMapPlayback; + ma_bool8 isUsingMMapCapture; + } alsa; +#endif +#ifdef MA_SUPPORT_PULSEAUDIO + struct + { + /*pa_mainloop**/ ma_ptr pMainLoop; + /*pa_context**/ ma_ptr pPulseContext; + /*pa_stream**/ ma_ptr pStreamPlayback; + /*pa_stream**/ ma_ptr pStreamCapture; + } pulse; +#endif +#ifdef MA_SUPPORT_JACK + struct + { + /*jack_client_t**/ ma_ptr pClient; + /*jack_port_t**/ ma_ptr* ppPortsPlayback; + /*jack_port_t**/ ma_ptr* ppPortsCapture; + float* pIntermediaryBufferPlayback; /* Typed as a float because JACK is always floating point. */ + float* pIntermediaryBufferCapture; + } jack; +#endif +#ifdef MA_SUPPORT_COREAUDIO + struct + { + ma_uint32 deviceObjectIDPlayback; + ma_uint32 deviceObjectIDCapture; + /*AudioUnit*/ ma_ptr audioUnitPlayback; + /*AudioUnit*/ ma_ptr audioUnitCapture; + /*AudioBufferList**/ ma_ptr pAudioBufferList; /* Only used for input devices. */ + ma_uint32 audioBufferCapInFrames; /* Only used for input devices. The capacity in frames of each buffer in pAudioBufferList. */ + ma_event stopEvent; + ma_uint32 originalPeriodSizeInFrames; + ma_uint32 originalPeriodSizeInMilliseconds; + ma_uint32 originalPeriods; + ma_performance_profile originalPerformanceProfile; + ma_bool32 isDefaultPlaybackDevice; + ma_bool32 isDefaultCaptureDevice; + ma_bool32 isSwitchingPlaybackDevice; /* <-- Set to true when the default device has changed and miniaudio is in the process of switching. */ + ma_bool32 isSwitchingCaptureDevice; /* <-- Set to true when the default device has changed and miniaudio is in the process of switching. */ + void* pNotificationHandler; /* Only used on mobile platforms. Obj-C object for handling route changes. */ + } coreaudio; +#endif +#ifdef MA_SUPPORT_SNDIO + struct + { + ma_ptr handlePlayback; + ma_ptr handleCapture; + ma_bool32 isStartedPlayback; + ma_bool32 isStartedCapture; + } sndio; +#endif +#ifdef MA_SUPPORT_AUDIO4 + struct + { + int fdPlayback; + int fdCapture; + } audio4; +#endif +#ifdef MA_SUPPORT_OSS + struct + { + int fdPlayback; + int fdCapture; + } oss; +#endif +#ifdef MA_SUPPORT_AAUDIO + struct + { + /*AAudioStream**/ ma_ptr pStreamPlayback; + /*AAudioStream**/ ma_ptr pStreamCapture; + ma_mutex rerouteLock; + ma_aaudio_usage usage; + ma_aaudio_content_type contentType; + ma_aaudio_input_preset inputPreset; + ma_aaudio_allowed_capture_policy allowedCapturePolicy; + ma_bool32 noAutoStartAfterReroute; + } aaudio; +#endif +#ifdef MA_SUPPORT_OPENSL + struct + { + /*SLObjectItf*/ ma_ptr pOutputMixObj; + /*SLOutputMixItf*/ ma_ptr pOutputMix; + /*SLObjectItf*/ ma_ptr pAudioPlayerObj; + /*SLPlayItf*/ ma_ptr pAudioPlayer; + /*SLObjectItf*/ ma_ptr pAudioRecorderObj; + /*SLRecordItf*/ ma_ptr pAudioRecorder; + /*SLAndroidSimpleBufferQueueItf*/ ma_ptr pBufferQueuePlayback; + /*SLAndroidSimpleBufferQueueItf*/ ma_ptr pBufferQueueCapture; + ma_bool32 isDrainingCapture; + ma_bool32 isDrainingPlayback; + ma_uint32 currentBufferIndexPlayback; + ma_uint32 currentBufferIndexCapture; + ma_uint8* pBufferPlayback; /* This is malloc()'d and is used for storing audio data. Typed as ma_uint8 for easy offsetting. */ + ma_uint8* pBufferCapture; + } opensl; +#endif +#ifdef MA_SUPPORT_WEBAUDIO + struct + { + /* AudioWorklets path. */ + /* EMSCRIPTEN_WEBAUDIO_T */ int audioContext; + /* EMSCRIPTEN_WEBAUDIO_T */ int audioWorklet; + float* pIntermediaryBuffer; + void* pStackBuffer; + ma_result initResult; /* Set to MA_BUSY while initialization is in progress. */ + int deviceIndex; /* We store the device in a list on the JavaScript side. This is used to map our C object to the JS object. */ + } webaudio; +#endif +#ifdef MA_SUPPORT_NULL + struct + { + ma_thread deviceThread; + ma_event operationEvent; + ma_event operationCompletionEvent; + ma_semaphore operationSemaphore; + ma_uint32 operation; + ma_result operationResult; + ma_timer timer; + double priorRunTime; + ma_uint32 currentPeriodFramesRemainingPlayback; + ma_uint32 currentPeriodFramesRemainingCapture; + ma_uint64 lastProcessedFramePlayback; + ma_uint64 lastProcessedFrameCapture; + ma_atomic_bool32 isStarted; /* Read and written by multiple threads. Must be used atomically, and must be 32-bit for compiler compatibility. */ + } null_device; +#endif + }; +}; +#if defined(_MSC_VER) && !defined(__clang__) + #pragma warning(pop) +#elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) + #pragma GCC diagnostic pop /* For ISO C99 doesn't support unnamed structs/unions [-Wpedantic] */ +#endif + +/* +Initializes a `ma_context_config` object. + + +Return Value +------------ +A `ma_context_config` initialized to defaults. + + +Remarks +------- +You must always use this to initialize the default state of the `ma_context_config` object. Not using this will result in your program breaking when miniaudio +is updated and new members are added to `ma_context_config`. It also sets logical defaults. + +You can override members of the returned object by changing it's members directly. + + +See Also +-------- +ma_context_init() +*/ +MA_API ma_context_config ma_context_config_init(void); + +/* +Initializes a context. + +The context is used for selecting and initializing an appropriate backend and to represent the backend at a more global level than that of an individual +device. There is one context to many devices, and a device is created from a context. A context is required to enumerate devices. + + +Parameters +---------- +backends (in, optional) + A list of backends to try initializing, in priority order. Can be NULL, in which case it uses default priority order. + +backendCount (in, optional) + The number of items in `backend`. Ignored if `backend` is NULL. + +pConfig (in, optional) + The context configuration. + +pContext (in) + A pointer to the context object being initialized. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Unsafe. Do not call this function across multiple threads as some backends read and write to global state. + + +Remarks +------- +When `backends` is NULL, the default priority order will be used. Below is a list of backends in priority order: + + |-------------|-----------------------|--------------------------------------------------------| + | Name | Enum Name | Supported Operating Systems | + |-------------|-----------------------|--------------------------------------------------------| + | WASAPI | ma_backend_wasapi | Windows Vista+ | + | DirectSound | ma_backend_dsound | Windows XP+ | + | WinMM | ma_backend_winmm | Windows XP+ (may work on older versions, but untested) | + | Core Audio | ma_backend_coreaudio | macOS, iOS | + | ALSA | ma_backend_alsa | Linux | + | PulseAudio | ma_backend_pulseaudio | Cross Platform (disabled on Windows, BSD and Android) | + | JACK | ma_backend_jack | Cross Platform (disabled on BSD and Android) | + | sndio | ma_backend_sndio | OpenBSD | + | audio(4) | ma_backend_audio4 | NetBSD, OpenBSD | + | OSS | ma_backend_oss | FreeBSD | + | AAudio | ma_backend_aaudio | Android 8+ | + | OpenSL|ES | ma_backend_opensl | Android (API level 16+) | + | Web Audio | ma_backend_webaudio | Web (via Emscripten) | + | Null | ma_backend_null | Cross Platform (not used on Web) | + |-------------|-----------------------|--------------------------------------------------------| + +The context can be configured via the `pConfig` argument. The config object is initialized with `ma_context_config_init()`. Individual configuration settings +can then be set directly on the structure. Below are the members of the `ma_context_config` object. + + pLog + A pointer to the `ma_log` to post log messages to. Can be NULL if the application does not + require logging. See the `ma_log` API for details on how to use the logging system. + + threadPriority + The desired priority to use for the audio thread. Allowable values include the following: + + |--------------------------------------| + | Thread Priority | + |--------------------------------------| + | ma_thread_priority_idle | + | ma_thread_priority_lowest | + | ma_thread_priority_low | + | ma_thread_priority_normal | + | ma_thread_priority_high | + | ma_thread_priority_highest (default) | + | ma_thread_priority_realtime | + | ma_thread_priority_default | + |--------------------------------------| + + threadStackSize + The desired size of the stack for the audio thread. Defaults to the operating system's default. + + pUserData + A pointer to application-defined data. This can be accessed from the context object directly such as `context.pUserData`. + + allocationCallbacks + Structure containing custom allocation callbacks. Leaving this at defaults will cause it to use MA_MALLOC, MA_REALLOC and MA_FREE. These allocation + callbacks will be used for anything tied to the context, including devices. + + alsa.useVerboseDeviceEnumeration + ALSA will typically enumerate many different devices which can be intrusive and not user-friendly. To combat this, miniaudio will enumerate only unique + card/device pairs by default. The problem with this is that you lose a bit of flexibility and control. Setting alsa.useVerboseDeviceEnumeration makes + it so the ALSA backend includes all devices. Defaults to false. + + pulse.pApplicationName + PulseAudio only. The application name to use when initializing the PulseAudio context with `pa_context_new()`. + + pulse.pServerName + PulseAudio only. The name of the server to connect to with `pa_context_connect()`. + + pulse.tryAutoSpawn + PulseAudio only. Whether or not to try automatically starting the PulseAudio daemon. Defaults to false. If you set this to true, keep in mind that + miniaudio uses a trial and error method to find the most appropriate backend, and this will result in the PulseAudio daemon starting which may be + intrusive for the end user. + + coreaudio.sessionCategory + iOS only. The session category to use for the shared AudioSession instance. Below is a list of allowable values and their Core Audio equivalents. + + |-----------------------------------------|-------------------------------------| + | miniaudio Token | Core Audio Token | + |-----------------------------------------|-------------------------------------| + | ma_ios_session_category_ambient | AVAudioSessionCategoryAmbient | + | ma_ios_session_category_solo_ambient | AVAudioSessionCategorySoloAmbient | + | ma_ios_session_category_playback | AVAudioSessionCategoryPlayback | + | ma_ios_session_category_record | AVAudioSessionCategoryRecord | + | ma_ios_session_category_play_and_record | AVAudioSessionCategoryPlayAndRecord | + | ma_ios_session_category_multi_route | AVAudioSessionCategoryMultiRoute | + | ma_ios_session_category_none | AVAudioSessionCategoryAmbient | + | ma_ios_session_category_default | AVAudioSessionCategoryAmbient | + |-----------------------------------------|-------------------------------------| + + coreaudio.sessionCategoryOptions + iOS only. Session category options to use with the shared AudioSession instance. Below is a list of allowable values and their Core Audio equivalents. + + |---------------------------------------------------------------------------|------------------------------------------------------------------| + | miniaudio Token | Core Audio Token | + |---------------------------------------------------------------------------|------------------------------------------------------------------| + | ma_ios_session_category_option_mix_with_others | AVAudioSessionCategoryOptionMixWithOthers | + | ma_ios_session_category_option_duck_others | AVAudioSessionCategoryOptionDuckOthers | + | ma_ios_session_category_option_allow_bluetooth | AVAudioSessionCategoryOptionAllowBluetooth | + | ma_ios_session_category_option_default_to_speaker | AVAudioSessionCategoryOptionDefaultToSpeaker | + | ma_ios_session_category_option_interrupt_spoken_audio_and_mix_with_others | AVAudioSessionCategoryOptionInterruptSpokenAudioAndMixWithOthers | + | ma_ios_session_category_option_allow_bluetooth_a2dp | AVAudioSessionCategoryOptionAllowBluetoothA2DP | + | ma_ios_session_category_option_allow_air_play | AVAudioSessionCategoryOptionAllowAirPlay | + |---------------------------------------------------------------------------|------------------------------------------------------------------| + + coreaudio.noAudioSessionActivate + iOS only. When set to true, does not perform an explicit [[AVAudioSession sharedInstace] setActive:true] on initialization. + + coreaudio.noAudioSessionDeactivate + iOS only. When set to true, does not perform an explicit [[AVAudioSession sharedInstace] setActive:false] on uninitialization. + + jack.pClientName + The name of the client to pass to `jack_client_open()`. + + jack.tryStartServer + Whether or not to try auto-starting the JACK server. Defaults to false. + + +It is recommended that only a single context is active at any given time because it's a bulky data structure which performs run-time linking for the +relevant backends every time it's initialized. + +The location of the context cannot change throughout it's lifetime. Consider allocating the `ma_context` object with `malloc()` if this is an issue. The +reason for this is that a pointer to the context is stored in the `ma_device` structure. + + +Example 1 - Default Initialization +---------------------------------- +The example below shows how to initialize the context using the default configuration. + +```c +ma_context context; +ma_result result = ma_context_init(NULL, 0, NULL, &context); +if (result != MA_SUCCESS) { + // Error. +} +``` + + +Example 2 - Custom Configuration +-------------------------------- +The example below shows how to initialize the context using custom backend priorities and a custom configuration. In this hypothetical example, the program +wants to prioritize ALSA over PulseAudio on Linux. They also want to avoid using the WinMM backend on Windows because it's latency is too high. They also +want an error to be returned if no valid backend is available which they achieve by excluding the Null backend. + +For the configuration, the program wants to capture any log messages so they can, for example, route it to a log file and user interface. + +```c +ma_backend backends[] = { + ma_backend_alsa, + ma_backend_pulseaudio, + ma_backend_wasapi, + ma_backend_dsound +}; + +ma_log log; +ma_log_init(&log); +ma_log_register_callback(&log, ma_log_callback_init(my_log_callbac, pMyLogUserData)); + +ma_context_config config = ma_context_config_init(); +config.pLog = &log; // Specify a custom log object in the config so any logs that are posted from ma_context_init() are captured. + +ma_context context; +ma_result result = ma_context_init(backends, sizeof(backends)/sizeof(backends[0]), &config, &context); +if (result != MA_SUCCESS) { + // Error. + if (result == MA_NO_BACKEND) { + // Couldn't find an appropriate backend. + } +} + +// You could also attach a log callback post-initialization: +ma_log_register_callback(ma_context_get_log(&context), ma_log_callback_init(my_log_callback, pMyLogUserData)); +``` + + +See Also +-------- +ma_context_config_init() +ma_context_uninit() +*/ +MA_API ma_result ma_context_init(const ma_backend backends[], ma_uint32 backendCount, const ma_context_config* pConfig, ma_context* pContext); + +/* +Uninitializes a context. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Unsafe. Do not call this function across multiple threads as some backends read and write to global state. + + +Remarks +------- +Results are undefined if you call this while any device created by this context is still active. + + +See Also +-------- +ma_context_init() +*/ +MA_API ma_result ma_context_uninit(ma_context* pContext); + +/* +Retrieves the size of the ma_context object. + +This is mainly for the purpose of bindings to know how much memory to allocate. +*/ +MA_API size_t ma_context_sizeof(void); + +/* +Retrieves a pointer to the log object associated with this context. + + +Remarks +------- +Pass the returned pointer to `ma_log_post()`, `ma_log_postv()` or `ma_log_postf()` to post a log +message. + +You can attach your own logging callback to the log with `ma_log_register_callback()` + + +Return Value +------------ +A pointer to the `ma_log` object that the context uses to post log messages. If some error occurs, +NULL will be returned. +*/ +MA_API ma_log* ma_context_get_log(ma_context* pContext); + +/* +Enumerates over every device (both playback and capture). + +This is a lower-level enumeration function to the easier to use `ma_context_get_devices()`. Use `ma_context_enumerate_devices()` if you would rather not incur +an internal heap allocation, or it simply suits your code better. + +Note that this only retrieves the ID and name/description of the device. The reason for only retrieving basic information is that it would otherwise require +opening the backend device in order to probe it for more detailed information which can be inefficient. Consider using `ma_context_get_device_info()` for this, +but don't call it from within the enumeration callback. + +Returning false from the callback will stop enumeration. Returning true will continue enumeration. + + +Parameters +---------- +pContext (in) + A pointer to the context performing the enumeration. + +callback (in) + The callback to fire for each enumerated device. + +pUserData (in) + A pointer to application-defined data passed to the callback. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Safe. This is guarded using a simple mutex lock. + + +Remarks +------- +Do _not_ assume the first enumerated device of a given type is the default device. + +Some backends and platforms may only support default playback and capture devices. + +In general, you should not do anything complicated from within the callback. In particular, do not try initializing a device from within the callback. Also, +do not try to call `ma_context_get_device_info()` from within the callback. + +Consider using `ma_context_get_devices()` for a simpler and safer API, albeit at the expense of an internal heap allocation. + + +Example 1 - Simple Enumeration +------------------------------ +ma_bool32 ma_device_enum_callback(ma_context* pContext, ma_device_type deviceType, const ma_device_info* pInfo, void* pUserData) +{ + printf("Device Name: %s\n", pInfo->name); + return MA_TRUE; +} + +ma_result result = ma_context_enumerate_devices(&context, my_device_enum_callback, pMyUserData); +if (result != MA_SUCCESS) { + // Error. +} + + +See Also +-------- +ma_context_get_devices() +*/ +MA_API ma_result ma_context_enumerate_devices(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData); + +/* +Retrieves basic information about every active playback and/or capture device. + +This function will allocate memory internally for the device lists and return a pointer to them through the `ppPlaybackDeviceInfos` and `ppCaptureDeviceInfos` +parameters. If you do not want to incur the overhead of these allocations consider using `ma_context_enumerate_devices()` which will instead use a callback. + +Note that this only retrieves the ID and name/description of the device. The reason for only retrieving basic information is that it would otherwise require +opening the backend device in order to probe it for more detailed information which can be inefficient. Consider using `ma_context_get_device_info()` for this, +but don't call it from within the enumeration callback. + + +Parameters +---------- +pContext (in) + A pointer to the context performing the enumeration. + +ppPlaybackDeviceInfos (out) + A pointer to a pointer that will receive the address of a buffer containing the list of `ma_device_info` structures for playback devices. + +pPlaybackDeviceCount (out) + A pointer to an unsigned integer that will receive the number of playback devices. + +ppCaptureDeviceInfos (out) + A pointer to a pointer that will receive the address of a buffer containing the list of `ma_device_info` structures for capture devices. + +pCaptureDeviceCount (out) + A pointer to an unsigned integer that will receive the number of capture devices. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Unsafe. Since each call to this function invalidates the pointers from the previous call, you should not be calling this simultaneously across multiple +threads. Instead, you need to make a copy of the returned data with your own higher level synchronization. + + +Remarks +------- +It is _not_ safe to assume the first device in the list is the default device. + +You can pass in NULL for the playback or capture lists in which case they'll be ignored. + +The returned pointers will become invalid upon the next call this this function, or when the context is uninitialized. Do not free the returned pointers. + + +See Also +-------- +ma_context_enumerate_devices() +*/ +MA_API ma_result ma_context_get_devices(ma_context* pContext, ma_device_info** ppPlaybackDeviceInfos, ma_uint32* pPlaybackDeviceCount, ma_device_info** ppCaptureDeviceInfos, ma_uint32* pCaptureDeviceCount); + +/* +Retrieves information about a device of the given type, with the specified ID and share mode. + + +Parameters +---------- +pContext (in) + A pointer to the context performing the query. + +deviceType (in) + The type of the device being queried. Must be either `ma_device_type_playback` or `ma_device_type_capture`. + +pDeviceID (in) + The ID of the device being queried. + +pDeviceInfo (out) + A pointer to the `ma_device_info` structure that will receive the device information. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Safe. This is guarded using a simple mutex lock. + + +Remarks +------- +Do _not_ call this from within the `ma_context_enumerate_devices()` callback. + +It's possible for a device to have different information and capabilities depending on whether or not it's opened in shared or exclusive mode. For example, in +shared mode, WASAPI always uses floating point samples for mixing, but in exclusive mode it can be anything. Therefore, this function allows you to specify +which share mode you want information for. Note that not all backends and devices support shared or exclusive mode, in which case this function will fail if +the requested share mode is unsupported. + +This leaves pDeviceInfo unmodified in the result of an error. +*/ +MA_API ma_result ma_context_get_device_info(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo); + +/* +Determines if the given context supports loopback mode. + + +Parameters +---------- +pContext (in) + A pointer to the context getting queried. + + +Return Value +------------ +MA_TRUE if the context supports loopback mode; MA_FALSE otherwise. +*/ +MA_API ma_bool32 ma_context_is_loopback_supported(ma_context* pContext); + + + +/* +Initializes a device config with default settings. + + +Parameters +---------- +deviceType (in) + The type of the device this config is being initialized for. This must set to one of the following: + + |-------------------------| + | Device Type | + |-------------------------| + | ma_device_type_playback | + | ma_device_type_capture | + | ma_device_type_duplex | + | ma_device_type_loopback | + |-------------------------| + + +Return Value +------------ +A new device config object with default settings. You will typically want to adjust the config after this function returns. See remarks. + + +Thread Safety +------------- +Safe. + + +Callback Safety +--------------- +Safe, but don't try initializing a device in a callback. + + +Remarks +------- +The returned config will be initialized to defaults. You will normally want to customize a few variables before initializing the device. See Example 1 for a +typical configuration which sets the sample format, channel count, sample rate, data callback and user data. These are usually things you will want to change +before initializing the device. + +See `ma_device_init()` for details on specific configuration options. + + +Example 1 - Simple Configuration +-------------------------------- +The example below is what a program will typically want to configure for each device at a minimum. Notice how `ma_device_config_init()` is called first, and +then the returned object is modified directly. This is important because it ensures that your program continues to work as new configuration options are added +to the `ma_device_config` structure. + +```c +ma_device_config config = ma_device_config_init(ma_device_type_playback); +config.playback.format = ma_format_f32; +config.playback.channels = 2; +config.sampleRate = 48000; +config.dataCallback = ma_data_callback; +config.pUserData = pMyUserData; +``` + + +See Also +-------- +ma_device_init() +ma_device_init_ex() +*/ +MA_API ma_device_config ma_device_config_init(ma_device_type deviceType); + + +/* +Initializes a device. + +A device represents a physical audio device. The idea is you send or receive audio data from the device to either play it back through a speaker, or capture it +from a microphone. Whether or not you should send or receive data from the device (or both) depends on the type of device you are initializing which can be +playback, capture, full-duplex or loopback. (Note that loopback mode is only supported on select backends.) Sending and receiving audio data to and from the +device is done via a callback which is fired by miniaudio at periodic time intervals. + +The frequency at which data is delivered to and from a device depends on the size of its period. The size of the period can be defined in terms of PCM frames +or milliseconds, whichever is more convenient. Generally speaking, the smaller the period, the lower the latency at the expense of higher CPU usage and +increased risk of glitching due to the more frequent and granular data deliver intervals. The size of a period will depend on your requirements, but +miniaudio's defaults should work fine for most scenarios. If you're building a game you should leave this fairly small, whereas if you're building a simple +media player you can make it larger. Note that the period size you request is actually just a hint - miniaudio will tell the backend what you want, but the +backend is ultimately responsible for what it gives you. You cannot assume you will get exactly what you ask for. + +When delivering data to and from a device you need to make sure it's in the correct format which you can set through the device configuration. You just set the +format that you want to use and miniaudio will perform all of the necessary conversion for you internally. When delivering data to and from the callback you +can assume the format is the same as what you requested when you initialized the device. See Remarks for more details on miniaudio's data conversion pipeline. + + +Parameters +---------- +pContext (in, optional) + A pointer to the context that owns the device. This can be null, in which case it creates a default context internally. + +pConfig (in) + A pointer to the device configuration. Cannot be null. See remarks for details. + +pDevice (out) + A pointer to the device object being initialized. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Unsafe. It is not safe to call this function simultaneously for different devices because some backends depend on and mutate global state. The same applies to +calling this at the same time as `ma_device_uninit()`. + + +Callback Safety +--------------- +Unsafe. It is not safe to call this inside any callback. + + +Remarks +------- +Setting `pContext` to NULL will result in miniaudio creating a default context internally and is equivalent to passing in a context initialized like so: + + ```c + ma_context_init(NULL, 0, NULL, &context); + ``` + +Do not set `pContext` to NULL if you are needing to open multiple devices. You can, however, use NULL when initializing the first device, and then use +device.pContext for the initialization of other devices. + +The device can be configured via the `pConfig` argument. The config object is initialized with `ma_device_config_init()`. Individual configuration settings can +then be set directly on the structure. Below are the members of the `ma_device_config` object. + + deviceType + Must be `ma_device_type_playback`, `ma_device_type_capture`, `ma_device_type_duplex` of `ma_device_type_loopback`. + + sampleRate + The sample rate, in hertz. The most common sample rates are 48000 and 44100. Setting this to 0 will use the device's native sample rate. + + periodSizeInFrames + The desired size of a period in PCM frames. If this is 0, `periodSizeInMilliseconds` will be used instead. If both are 0 the default buffer size will + be used depending on the selected performance profile. This value affects latency. See below for details. + + periodSizeInMilliseconds + The desired size of a period in milliseconds. If this is 0, `periodSizeInFrames` will be used instead. If both are 0 the default buffer size will be + used depending on the selected performance profile. The value affects latency. See below for details. + + periods + The number of periods making up the device's entire buffer. The total buffer size is `periodSizeInFrames` or `periodSizeInMilliseconds` multiplied by + this value. This is just a hint as backends will be the ones who ultimately decide how your periods will be configured. + + performanceProfile + A hint to miniaudio as to the performance requirements of your program. Can be either `ma_performance_profile_low_latency` (default) or + `ma_performance_profile_conservative`. This mainly affects the size of default buffers and can usually be left at its default value. + + noPreSilencedOutputBuffer + When set to true, the contents of the output buffer passed into the data callback will be left undefined. When set to false (default), the contents of + the output buffer will be cleared the zero. You can use this to avoid the overhead of zeroing out the buffer if you can guarantee that your data + callback will write to every sample in the output buffer, or if you are doing your own clearing. + + noClip + When set to true, the contents of the output buffer are left alone after returning and it will be left up to the backend itself to decide whether or + not to clip. When set to false (default), the contents of the output buffer passed into the data callback will be clipped after returning. This only + applies when the playback sample format is f32. + + noDisableDenormals + By default, miniaudio will disable denormals when the data callback is called. Setting this to true will prevent the disabling of denormals. + + noFixedSizedCallback + Allows miniaudio to fire the data callback with any frame count. When this is set to false (the default), the data callback will be fired with a + consistent frame count as specified by `periodSizeInFrames` or `periodSizeInMilliseconds`. When set to true, miniaudio will fire the callback with + whatever the backend requests, which could be anything. + + dataCallback + The callback to fire whenever data is ready to be delivered to or from the device. + + notificationCallback + The callback to fire when something has changed with the device, such as whether or not it has been started or stopped. + + pUserData + The user data pointer to use with the device. You can access this directly from the device object like `device.pUserData`. + + resampling.algorithm + The resampling algorithm to use when miniaudio needs to perform resampling between the rate specified by `sampleRate` and the device's native rate. The + default value is `ma_resample_algorithm_linear`, and the quality can be configured with `resampling.linear.lpfOrder`. + + resampling.pBackendVTable + A pointer to an optional vtable that can be used for plugging in a custom resampler. + + resampling.pBackendUserData + A pointer that will passed to callbacks in pBackendVTable. + + resampling.linear.lpfOrder + The linear resampler applies a low-pass filter as part of its processing for anti-aliasing. This setting controls the order of the filter. The higher + the value, the better the quality, in general. Setting this to 0 will disable low-pass filtering altogether. The maximum value is + `MA_MAX_FILTER_ORDER`. The default value is `min(4, MA_MAX_FILTER_ORDER)`. + + playback.pDeviceID + A pointer to a `ma_device_id` structure containing the ID of the playback device to initialize. Setting this NULL (default) will use the system's + default playback device. Retrieve the device ID from the `ma_device_info` structure, which can be retrieved using device enumeration. + + playback.format + The sample format to use for playback. When set to `ma_format_unknown` the device's native format will be used. This can be retrieved after + initialization from the device object directly with `device.playback.format`. + + playback.channels + The number of channels to use for playback. When set to 0 the device's native channel count will be used. This can be retrieved after initialization + from the device object directly with `device.playback.channels`. + + playback.pChannelMap + The channel map to use for playback. When left empty, the device's native channel map will be used. This can be retrieved after initialization from the + device object direct with `device.playback.pChannelMap`. When set, the buffer should contain `channels` items. + + playback.shareMode + The preferred share mode to use for playback. Can be either `ma_share_mode_shared` (default) or `ma_share_mode_exclusive`. Note that if you specify + exclusive mode, but it's not supported by the backend, initialization will fail. You can then fall back to shared mode if desired by changing this to + ma_share_mode_shared and reinitializing. + + capture.pDeviceID + A pointer to a `ma_device_id` structure containing the ID of the capture device to initialize. Setting this NULL (default) will use the system's + default capture device. Retrieve the device ID from the `ma_device_info` structure, which can be retrieved using device enumeration. + + capture.format + The sample format to use for capture. When set to `ma_format_unknown` the device's native format will be used. This can be retrieved after + initialization from the device object directly with `device.capture.format`. + + capture.channels + The number of channels to use for capture. When set to 0 the device's native channel count will be used. This can be retrieved after initialization + from the device object directly with `device.capture.channels`. + + capture.pChannelMap + The channel map to use for capture. When left empty, the device's native channel map will be used. This can be retrieved after initialization from the + device object direct with `device.capture.pChannelMap`. When set, the buffer should contain `channels` items. + + capture.shareMode + The preferred share mode to use for capture. Can be either `ma_share_mode_shared` (default) or `ma_share_mode_exclusive`. Note that if you specify + exclusive mode, but it's not supported by the backend, initialization will fail. You can then fall back to shared mode if desired by changing this to + ma_share_mode_shared and reinitializing. + + wasapi.noAutoConvertSRC + WASAPI only. When set to true, disables WASAPI's automatic resampling and forces the use of miniaudio's resampler. Defaults to false. + + wasapi.noDefaultQualitySRC + WASAPI only. Only used when `wasapi.noAutoConvertSRC` is set to false. When set to true, disables the use of `AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY`. + You should usually leave this set to false, which is the default. + + wasapi.noAutoStreamRouting + WASAPI only. When set to true, disables automatic stream routing on the WASAPI backend. Defaults to false. + + wasapi.noHardwareOffloading + WASAPI only. When set to true, disables the use of WASAPI's hardware offloading feature. Defaults to false. + + alsa.noMMap + ALSA only. When set to true, disables MMap mode. Defaults to false. + + alsa.noAutoFormat + ALSA only. When set to true, disables ALSA's automatic format conversion by including the SND_PCM_NO_AUTO_FORMAT flag. Defaults to false. + + alsa.noAutoChannels + ALSA only. When set to true, disables ALSA's automatic channel conversion by including the SND_PCM_NO_AUTO_CHANNELS flag. Defaults to false. + + alsa.noAutoResample + ALSA only. When set to true, disables ALSA's automatic resampling by including the SND_PCM_NO_AUTO_RESAMPLE flag. Defaults to false. + + pulse.pStreamNamePlayback + PulseAudio only. Sets the stream name for playback. + + pulse.pStreamNameCapture + PulseAudio only. Sets the stream name for capture. + + pulse.channelMap + PulseAudio only. Sets the channel map that is requested from PulseAudio. See MA_PA_CHANNEL_MAP_* constants. Defaults to MA_PA_CHANNEL_MAP_AIFF. + + coreaudio.allowNominalSampleRateChange + Core Audio only. Desktop only. When enabled, allows the sample rate of the device to be changed at the operating system level. This + is disabled by default in order to prevent intrusive changes to the user's system. This is useful if you want to use a sample rate + that is known to be natively supported by the hardware thereby avoiding the cost of resampling. When set to true, miniaudio will + find the closest match between the sample rate requested in the device config and the sample rates natively supported by the + hardware. When set to false, the sample rate currently set by the operating system will always be used. + + opensl.streamType + OpenSL only. Explicitly sets the stream type. If left unset (`ma_opensl_stream_type_default`), the + stream type will be left unset. Think of this as the type of audio you're playing. + + opensl.recordingPreset + OpenSL only. Explicitly sets the type of recording your program will be doing. When left + unset, the recording preset will be left unchanged. + + aaudio.usage + AAudio only. Explicitly sets the nature of the audio the program will be consuming. When + left unset, the usage will be left unchanged. + + aaudio.contentType + AAudio only. Sets the content type. When left unset, the content type will be left unchanged. + + aaudio.inputPreset + AAudio only. Explicitly sets the type of recording your program will be doing. When left + unset, the input preset will be left unchanged. + + aaudio.noAutoStartAfterReroute + AAudio only. Controls whether or not the device should be automatically restarted after a + stream reroute. When set to false (default) the device will be restarted automatically; + otherwise the device will be stopped. + + +Once initialized, the device's config is immutable. If you need to change the config you will need to initialize a new device. + +After initializing the device it will be in a stopped state. To start it, use `ma_device_start()`. + +If both `periodSizeInFrames` and `periodSizeInMilliseconds` are set to zero, it will default to `MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_LOW_LATENCY` or +`MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_CONSERVATIVE`, depending on whether or not `performanceProfile` is set to `ma_performance_profile_low_latency` or +`ma_performance_profile_conservative`. + +If you request exclusive mode and the backend does not support it an error will be returned. For robustness, you may want to first try initializing the device +in exclusive mode, and then fall back to shared mode if required. Alternatively you can just request shared mode (the default if you leave it unset in the +config) which is the most reliable option. Some backends do not have a practical way of choosing whether or not the device should be exclusive or not (ALSA, +for example) in which case it just acts as a hint. Unless you have special requirements you should try avoiding exclusive mode as it's intrusive to the user. +Starting with Windows 10, miniaudio will use low-latency shared mode where possible which may make exclusive mode unnecessary. + +When sending or receiving data to/from a device, miniaudio will internally perform a format conversion to convert between the format specified by the config +and the format used internally by the backend. If you pass in 0 for the sample format, channel count, sample rate _and_ channel map, data transmission will run +on an optimized pass-through fast path. You can retrieve the format, channel count and sample rate by inspecting the `playback/capture.format`, +`playback/capture.channels` and `sampleRate` members of the device object. + +When compiling for UWP you must ensure you call this function on the main UI thread because the operating system may need to present the user with a message +asking for permissions. Please refer to the official documentation for ActivateAudioInterfaceAsync() for more information. + +ALSA Specific: When initializing the default device, requesting shared mode will try using the "dmix" device for playback and the "dsnoop" device for capture. +If these fail it will try falling back to the "hw" device. + + +Example 1 - Simple Initialization +--------------------------------- +This example shows how to initialize a simple playback device using a standard configuration. If you are just needing to do simple playback from the default +playback device this is usually all you need. + +```c +ma_device_config config = ma_device_config_init(ma_device_type_playback); +config.playback.format = ma_format_f32; +config.playback.channels = 2; +config.sampleRate = 48000; +config.dataCallback = ma_data_callback; +config.pMyUserData = pMyUserData; + +ma_device device; +ma_result result = ma_device_init(NULL, &config, &device); +if (result != MA_SUCCESS) { + // Error +} +``` + + +Example 2 - Advanced Initialization +----------------------------------- +This example shows how you might do some more advanced initialization. In this hypothetical example we want to control the latency by setting the buffer size +and period count. We also want to allow the user to be able to choose which device to output from which means we need a context so we can perform device +enumeration. + +```c +ma_context context; +ma_result result = ma_context_init(NULL, 0, NULL, &context); +if (result != MA_SUCCESS) { + // Error +} + +ma_device_info* pPlaybackDeviceInfos; +ma_uint32 playbackDeviceCount; +result = ma_context_get_devices(&context, &pPlaybackDeviceInfos, &playbackDeviceCount, NULL, NULL); +if (result != MA_SUCCESS) { + // Error +} + +// ... choose a device from pPlaybackDeviceInfos ... + +ma_device_config config = ma_device_config_init(ma_device_type_playback); +config.playback.pDeviceID = pMyChosenDeviceID; // <-- Get this from the `id` member of one of the `ma_device_info` objects returned by ma_context_get_devices(). +config.playback.format = ma_format_f32; +config.playback.channels = 2; +config.sampleRate = 48000; +config.dataCallback = ma_data_callback; +config.pUserData = pMyUserData; +config.periodSizeInMilliseconds = 10; +config.periods = 3; + +ma_device device; +result = ma_device_init(&context, &config, &device); +if (result != MA_SUCCESS) { + // Error +} +``` + + +See Also +-------- +ma_device_config_init() +ma_device_uninit() +ma_device_start() +ma_context_init() +ma_context_get_devices() +ma_context_enumerate_devices() +*/ +MA_API ma_result ma_device_init(ma_context* pContext, const ma_device_config* pConfig, ma_device* pDevice); + +/* +Initializes a device without a context, with extra parameters for controlling the configuration of the internal self-managed context. + +This is the same as `ma_device_init()`, only instead of a context being passed in, the parameters from `ma_context_init()` are passed in instead. This function +allows you to configure the internally created context. + + +Parameters +---------- +backends (in, optional) + A list of backends to try initializing, in priority order. Can be NULL, in which case it uses default priority order. + +backendCount (in, optional) + The number of items in `backend`. Ignored if `backend` is NULL. + +pContextConfig (in, optional) + The context configuration. + +pConfig (in) + A pointer to the device configuration. Cannot be null. See remarks for details. + +pDevice (out) + A pointer to the device object being initialized. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Unsafe. It is not safe to call this function simultaneously for different devices because some backends depend on and mutate global state. The same applies to +calling this at the same time as `ma_device_uninit()`. + + +Callback Safety +--------------- +Unsafe. It is not safe to call this inside any callback. + + +Remarks +------- +You only need to use this function if you want to configure the context differently to its defaults. You should never use this function if you want to manage +your own context. + +See the documentation for `ma_context_init()` for information on the different context configuration options. + + +See Also +-------- +ma_device_init() +ma_device_uninit() +ma_device_config_init() +ma_context_init() +*/ +MA_API ma_result ma_device_init_ex(const ma_backend backends[], ma_uint32 backendCount, const ma_context_config* pContextConfig, const ma_device_config* pConfig, ma_device* pDevice); + +/* +Uninitializes a device. + +This will explicitly stop the device. You do not need to call `ma_device_stop()` beforehand, but it's harmless if you do. + + +Parameters +---------- +pDevice (in) + A pointer to the device to stop. + + +Return Value +------------ +Nothing + + +Thread Safety +------------- +Unsafe. As soon as this API is called the device should be considered undefined. + + +Callback Safety +--------------- +Unsafe. It is not safe to call this inside any callback. Doing this will result in a deadlock. + + +See Also +-------- +ma_device_init() +ma_device_stop() +*/ +MA_API void ma_device_uninit(ma_device* pDevice); + + +/* +Retrieves a pointer to the context that owns the given device. +*/ +MA_API ma_context* ma_device_get_context(ma_device* pDevice); + +/* +Helper function for retrieving the log object associated with the context that owns this device. +*/ +MA_API ma_log* ma_device_get_log(ma_device* pDevice); + + +/* +Retrieves information about the device. + + +Parameters +---------- +pDevice (in) + A pointer to the device whose information is being retrieved. + +type (in) + The device type. This parameter is required for duplex devices. When retrieving device + information, you are doing so for an individual playback or capture device. + +pDeviceInfo (out) + A pointer to the `ma_device_info` that will receive the device information. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Unsafe. This should be considered unsafe because it may be calling into the backend which may or +may not be safe. + + +Callback Safety +--------------- +Unsafe. You should avoid calling this in the data callback because it may call into the backend +which may or may not be safe. +*/ +MA_API ma_result ma_device_get_info(ma_device* pDevice, ma_device_type type, ma_device_info* pDeviceInfo); + + +/* +Retrieves the name of the device. + + +Parameters +---------- +pDevice (in) + A pointer to the device whose information is being retrieved. + +type (in) + The device type. This parameter is required for duplex devices. When retrieving device + information, you are doing so for an individual playback or capture device. + +pName (out) + A pointer to the buffer that will receive the name. + +nameCap (in) + The capacity of the output buffer, including space for the null terminator. + +pLengthNotIncludingNullTerminator (out, optional) + A pointer to the variable that will receive the length of the name, not including the null + terminator. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Unsafe. This should be considered unsafe because it may be calling into the backend which may or +may not be safe. + + +Callback Safety +--------------- +Unsafe. You should avoid calling this in the data callback because it may call into the backend +which may or may not be safe. + + +Remarks +------- +If the name does not fully fit into the output buffer, it'll be truncated. You can pass in NULL to +`pName` if you want to first get the length of the name for the purpose of memory allocation of the +output buffer. Allocating a buffer of size `MA_MAX_DEVICE_NAME_LENGTH + 1` should be enough for +most cases and will avoid the need for the inefficiency of calling this function twice. + +This is implemented in terms of `ma_device_get_info()`. +*/ +MA_API ma_result ma_device_get_name(ma_device* pDevice, ma_device_type type, char* pName, size_t nameCap, size_t* pLengthNotIncludingNullTerminator); + + +/* +Starts the device. For playback devices this begins playback. For capture devices it begins recording. + +Use `ma_device_stop()` to stop the device. + + +Parameters +---------- +pDevice (in) + A pointer to the device to start. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Safe. It's safe to call this from any thread with the exception of the callback thread. + + +Callback Safety +--------------- +Unsafe. It is not safe to call this inside any callback. + + +Remarks +------- +For a playback device, this will retrieve an initial chunk of audio data from the client before returning. The reason for this is to ensure there is valid +audio data in the buffer, which needs to be done before the device begins playback. + +This API waits until the backend device has been started for real by the worker thread. It also waits on a mutex for thread-safety. + +Do not call this in any callback. + + +See Also +-------- +ma_device_stop() +*/ +MA_API ma_result ma_device_start(ma_device* pDevice); + +/* +Stops the device. For playback devices this stops playback. For capture devices it stops recording. + +Use `ma_device_start()` to start the device again. + + +Parameters +---------- +pDevice (in) + A pointer to the device to stop. + + +Return Value +------------ +MA_SUCCESS if successful; any other error code otherwise. + + +Thread Safety +------------- +Safe. It's safe to call this from any thread with the exception of the callback thread. + + +Callback Safety +--------------- +Unsafe. It is not safe to call this inside any callback. Doing this will result in a deadlock. + + +Remarks +------- +This API needs to wait on the worker thread to stop the backend device properly before returning. It also waits on a mutex for thread-safety. In addition, some +backends need to wait for the device to finish playback/recording of the current fragment which can take some time (usually proportionate to the buffer size +that was specified at initialization time). + +Backends are required to either pause the stream in-place or drain the buffer if pausing is not possible. The reason for this is that stopping the device and +the resuming it with ma_device_start() (which you might do when your program loses focus) may result in a situation where those samples are never output to the +speakers or received from the microphone which can in turn result in de-syncs. + +Do not call this in any callback. + + +See Also +-------- +ma_device_start() +*/ +MA_API ma_result ma_device_stop(ma_device* pDevice); + +/* +Determines whether or not the device is started. + + +Parameters +---------- +pDevice (in) + A pointer to the device whose start state is being retrieved. + + +Return Value +------------ +True if the device is started, false otherwise. + + +Thread Safety +------------- +Safe. If another thread calls `ma_device_start()` or `ma_device_stop()` at this same time as this function is called, there's a very small chance the return +value will be out of sync. + + +Callback Safety +--------------- +Safe. This is implemented as a simple accessor. + + +See Also +-------- +ma_device_start() +ma_device_stop() +*/ +MA_API ma_bool32 ma_device_is_started(const ma_device* pDevice); + + +/* +Retrieves the state of the device. + + +Parameters +---------- +pDevice (in) + A pointer to the device whose state is being retrieved. + + +Return Value +------------ +The current state of the device. The return value will be one of the following: + + +-------------------------------+------------------------------------------------------------------------------+ + | ma_device_state_uninitialized | Will only be returned if the device is in the middle of initialization. | + +-------------------------------+------------------------------------------------------------------------------+ + | ma_device_state_stopped | The device is stopped. The initial state of the device after initialization. | + +-------------------------------+------------------------------------------------------------------------------+ + | ma_device_state_started | The device started and requesting and/or delivering audio data. | + +-------------------------------+------------------------------------------------------------------------------+ + | ma_device_state_starting | The device is in the process of starting. | + +-------------------------------+------------------------------------------------------------------------------+ + | ma_device_state_stopping | The device is in the process of stopping. | + +-------------------------------+------------------------------------------------------------------------------+ + + +Thread Safety +------------- +Safe. This is implemented as a simple accessor. Note that if the device is started or stopped at the same time as this function is called, +there's a possibility the return value could be out of sync. See remarks. + + +Callback Safety +--------------- +Safe. This is implemented as a simple accessor. + + +Remarks +------- +The general flow of a devices state goes like this: + + ``` + ma_device_init() -> ma_device_state_uninitialized -> ma_device_state_stopped + ma_device_start() -> ma_device_state_starting -> ma_device_state_started + ma_device_stop() -> ma_device_state_stopping -> ma_device_state_stopped + ``` + +When the state of the device is changed with `ma_device_start()` or `ma_device_stop()` at this same time as this function is called, the +value returned by this function could potentially be out of sync. If this is significant to your program you need to implement your own +synchronization. +*/ +MA_API ma_device_state ma_device_get_state(const ma_device* pDevice); + + +/* +Performs post backend initialization routines for setting up internal data conversion. + +This should be called whenever the backend is initialized. The only time this should be called from +outside of miniaudio is if you're implementing a custom backend, and you would only do it if you +are reinitializing the backend due to rerouting or reinitializing for some reason. + + +Parameters +---------- +pDevice [in] + A pointer to the device. + +deviceType [in] + The type of the device that was just reinitialized. + +pPlaybackDescriptor [in] + The descriptor of the playback device containing the internal data format and buffer sizes. + +pPlaybackDescriptor [in] + The descriptor of the capture device containing the internal data format and buffer sizes. + + +Return Value +------------ +MA_SUCCESS if successful; any other error otherwise. + + +Thread Safety +------------- +Unsafe. This will be reinitializing internal data converters which may be in use by another thread. + + +Callback Safety +--------------- +Unsafe. This will be reinitializing internal data converters which may be in use by the callback. + + +Remarks +------- +For a duplex device, you can call this for only one side of the system. This is why the deviceType +is specified as a parameter rather than deriving it from the device. + +You do not need to call this manually unless you are doing a custom backend, in which case you need +only do it if you're manually performing rerouting or reinitialization. +*/ +MA_API ma_result ma_device_post_init(ma_device* pDevice, ma_device_type deviceType, const ma_device_descriptor* pPlaybackDescriptor, const ma_device_descriptor* pCaptureDescriptor); + + +/* +Sets the master volume factor for the device. + +The volume factor must be between 0 (silence) and 1 (full volume). Use `ma_device_set_master_volume_db()` to use decibel notation, where 0 is full volume and +values less than 0 decreases the volume. + + +Parameters +---------- +pDevice (in) + A pointer to the device whose volume is being set. + +volume (in) + The new volume factor. Must be >= 0. + + +Return Value +------------ +MA_SUCCESS if the volume was set successfully. +MA_INVALID_ARGS if pDevice is NULL. +MA_INVALID_ARGS if volume is negative. + + +Thread Safety +------------- +Safe. This just sets a local member of the device object. + + +Callback Safety +--------------- +Safe. If you set the volume in the data callback, that data written to the output buffer will have the new volume applied. + + +Remarks +------- +This applies the volume factor across all channels. + +This does not change the operating system's volume. It only affects the volume for the given `ma_device` object's audio stream. + + +See Also +-------- +ma_device_get_master_volume() +ma_device_set_master_volume_db() +ma_device_get_master_volume_db() +*/ +MA_API ma_result ma_device_set_master_volume(ma_device* pDevice, float volume); + +/* +Retrieves the master volume factor for the device. + + +Parameters +---------- +pDevice (in) + A pointer to the device whose volume factor is being retrieved. + +pVolume (in) + A pointer to the variable that will receive the volume factor. The returned value will be in the range of [0, 1]. + + +Return Value +------------ +MA_SUCCESS if successful. +MA_INVALID_ARGS if pDevice is NULL. +MA_INVALID_ARGS if pVolume is NULL. + + +Thread Safety +------------- +Safe. This just a simple member retrieval. + + +Callback Safety +--------------- +Safe. + + +Remarks +------- +If an error occurs, `*pVolume` will be set to 0. + + +See Also +-------- +ma_device_set_master_volume() +ma_device_set_master_volume_gain_db() +ma_device_get_master_volume_gain_db() +*/ +MA_API ma_result ma_device_get_master_volume(ma_device* pDevice, float* pVolume); + +/* +Sets the master volume for the device as gain in decibels. + +A gain of 0 is full volume, whereas a gain of < 0 will decrease the volume. + + +Parameters +---------- +pDevice (in) + A pointer to the device whose gain is being set. + +gainDB (in) + The new volume as gain in decibels. Must be less than or equal to 0, where 0 is full volume and anything less than 0 decreases the volume. + + +Return Value +------------ +MA_SUCCESS if the volume was set successfully. +MA_INVALID_ARGS if pDevice is NULL. +MA_INVALID_ARGS if the gain is > 0. + + +Thread Safety +------------- +Safe. This just sets a local member of the device object. + + +Callback Safety +--------------- +Safe. If you set the volume in the data callback, that data written to the output buffer will have the new volume applied. + + +Remarks +------- +This applies the gain across all channels. + +This does not change the operating system's volume. It only affects the volume for the given `ma_device` object's audio stream. + + +See Also +-------- +ma_device_get_master_volume_gain_db() +ma_device_set_master_volume() +ma_device_get_master_volume() +*/ +MA_API ma_result ma_device_set_master_volume_db(ma_device* pDevice, float gainDB); + +/* +Retrieves the master gain in decibels. + + +Parameters +---------- +pDevice (in) + A pointer to the device whose gain is being retrieved. + +pGainDB (in) + A pointer to the variable that will receive the gain in decibels. The returned value will be <= 0. + + +Return Value +------------ +MA_SUCCESS if successful. +MA_INVALID_ARGS if pDevice is NULL. +MA_INVALID_ARGS if pGainDB is NULL. + + +Thread Safety +------------- +Safe. This just a simple member retrieval. + + +Callback Safety +--------------- +Safe. + + +Remarks +------- +If an error occurs, `*pGainDB` will be set to 0. + + +See Also +-------- +ma_device_set_master_volume_db() +ma_device_set_master_volume() +ma_device_get_master_volume() +*/ +MA_API ma_result ma_device_get_master_volume_db(ma_device* pDevice, float* pGainDB); + + +/* +Called from the data callback of asynchronous backends to allow miniaudio to process the data and fire the miniaudio data callback. + + +Parameters +---------- +pDevice (in) + A pointer to device whose processing the data callback. + +pOutput (out) + A pointer to the buffer that will receive the output PCM frame data. On a playback device this must not be NULL. On a duplex device + this can be NULL, in which case pInput must not be NULL. + +pInput (in) + A pointer to the buffer containing input PCM frame data. On a capture device this must not be NULL. On a duplex device this can be + NULL, in which case `pOutput` must not be NULL. + +frameCount (in) + The number of frames being processed. + + +Return Value +------------ +MA_SUCCESS if successful; any other result code otherwise. + + +Thread Safety +------------- +This function should only ever be called from the internal data callback of the backend. It is safe to call this simultaneously between a +playback and capture device in duplex setups. + + +Callback Safety +--------------- +Do not call this from the miniaudio data callback. It should only ever be called from the internal data callback of the backend. + + +Remarks +------- +If both `pOutput` and `pInput` are NULL, and error will be returned. In duplex scenarios, both `pOutput` and `pInput` can be non-NULL, in +which case `pInput` will be processed first, followed by `pOutput`. + +If you are implementing a custom backend, and that backend uses a callback for data delivery, you'll need to call this from inside that +callback. +*/ +MA_API ma_result ma_device_handle_backend_data_callback(ma_device* pDevice, void* pOutput, const void* pInput, ma_uint32 frameCount); + + +/* +Calculates an appropriate buffer size from a descriptor, native sample rate and performance profile. + +This function is used by backends for helping determine an appropriately sized buffer to use with +the device depending on the values of `periodSizeInFrames` and `periodSizeInMilliseconds` in the +`pDescriptor` object. Since buffer size calculations based on time depends on the sample rate, a +best guess at the device's native sample rate is also required which is where `nativeSampleRate` +comes in. In addition, the performance profile is also needed for cases where both the period size +in frames and milliseconds are both zero. + + +Parameters +---------- +pDescriptor (in) + A pointer to device descriptor whose `periodSizeInFrames` and `periodSizeInMilliseconds` members + will be used for the calculation of the buffer size. + +nativeSampleRate (in) + The device's native sample rate. This is only ever used when the `periodSizeInFrames` member of + `pDescriptor` is zero. In this case, `periodSizeInMilliseconds` will be used instead, in which + case a sample rate is required to convert to a size in frames. + +performanceProfile (in) + When both the `periodSizeInFrames` and `periodSizeInMilliseconds` members of `pDescriptor` are + zero, miniaudio will fall back to a buffer size based on the performance profile. The profile + to use for this calculation is determine by this parameter. + + +Return Value +------------ +The calculated buffer size in frames. + + +Thread Safety +------------- +This is safe so long as nothing modifies `pDescriptor` at the same time. However, this function +should only ever be called from within the backend's device initialization routine and therefore +shouldn't have any multithreading concerns. + + +Callback Safety +--------------- +This is safe to call within the data callback, but there is no reason to ever do this. + + +Remarks +------- +If `nativeSampleRate` is zero, this function will fall back to `pDescriptor->sampleRate`. If that +is also zero, `MA_DEFAULT_SAMPLE_RATE` will be used instead. +*/ +MA_API ma_uint32 ma_calculate_buffer_size_in_frames_from_descriptor(const ma_device_descriptor* pDescriptor, ma_uint32 nativeSampleRate, ma_performance_profile performanceProfile); + + + +/* +Retrieves a friendly name for a backend. +*/ +MA_API const char* ma_get_backend_name(ma_backend backend); + +/* +Retrieves the backend enum from the given name. +*/ +MA_API ma_result ma_get_backend_from_name(const char* pBackendName, ma_backend* pBackend); + +/* +Determines whether or not the given backend is available by the compilation environment. +*/ +MA_API ma_bool32 ma_is_backend_enabled(ma_backend backend); + +/* +Retrieves compile-time enabled backends. + + +Parameters +---------- +pBackends (out, optional) + A pointer to the buffer that will receive the enabled backends. Set to NULL to retrieve the backend count. Setting + the capacity of the buffer to `MA_BUFFER_COUNT` will guarantee it's large enough for all backends. + +backendCap (in) + The capacity of the `pBackends` buffer. + +pBackendCount (out) + A pointer to the variable that will receive the enabled backend count. + + +Return Value +------------ +MA_SUCCESS if successful. +MA_INVALID_ARGS if `pBackendCount` is NULL. +MA_NO_SPACE if the capacity of `pBackends` is not large enough. + +If `MA_NO_SPACE` is returned, the `pBackends` buffer will be filled with `*pBackendCount` values. + + +Thread Safety +------------- +Safe. + + +Callback Safety +--------------- +Safe. + + +Remarks +------- +If you want to retrieve the number of backends so you can determine the capacity of `pBackends` buffer, you can call +this function with `pBackends` set to NULL. + +This will also enumerate the null backend. If you don't want to include this you need to check for `ma_backend_null` +when you enumerate over the returned backends and handle it appropriately. Alternatively, you can disable it at +compile time with `MA_NO_NULL`. + +The returned backends are determined based on compile time settings, not the platform it's currently running on. For +example, PulseAudio will be returned if it was enabled at compile time, even when the user doesn't actually have +PulseAudio installed. + + +Example 1 +--------- +The example below retrieves the enabled backend count using a fixed sized buffer allocated on the stack. The buffer is +given a capacity of `MA_BACKEND_COUNT` which will guarantee it'll be large enough to store all available backends. +Since `MA_BACKEND_COUNT` is always a relatively small value, this should be suitable for most scenarios. + +``` +ma_backend enabledBackends[MA_BACKEND_COUNT]; +size_t enabledBackendCount; + +result = ma_get_enabled_backends(enabledBackends, MA_BACKEND_COUNT, &enabledBackendCount); +if (result != MA_SUCCESS) { + // Failed to retrieve enabled backends. Should never happen in this example since all inputs are valid. +} +``` + + +See Also +-------- +ma_is_backend_enabled() +*/ +MA_API ma_result ma_get_enabled_backends(ma_backend* pBackends, size_t backendCap, size_t* pBackendCount); + +/* +Determines whether or not loopback mode is support by a backend. +*/ +MA_API ma_bool32 ma_is_loopback_supported(ma_backend backend); + +#endif /* MA_NO_DEVICE_IO */ + + + +/************************************************************************************************************************************************************ + +Utilities + +************************************************************************************************************************************************************/ + +/* +Calculates a buffer size in milliseconds (rounded up) from the specified number of frames and sample rate. +*/ +MA_API ma_uint32 ma_calculate_buffer_size_in_milliseconds_from_frames(ma_uint32 bufferSizeInFrames, ma_uint32 sampleRate); + +/* +Calculates a buffer size in frames from the specified number of milliseconds and sample rate. +*/ +MA_API ma_uint32 ma_calculate_buffer_size_in_frames_from_milliseconds(ma_uint32 bufferSizeInMilliseconds, ma_uint32 sampleRate); + +/* +Copies PCM frames from one buffer to another. +*/ +MA_API void ma_copy_pcm_frames(void* dst, const void* src, ma_uint64 frameCount, ma_format format, ma_uint32 channels); + +/* +Copies silent frames into the given buffer. + +Remarks +------- +For all formats except `ma_format_u8`, the output buffer will be filled with 0. For `ma_format_u8` it will be filled with 128. The reason for this is that it +makes more sense for the purpose of mixing to initialize it to the center point. +*/ +MA_API void ma_silence_pcm_frames(void* p, ma_uint64 frameCount, ma_format format, ma_uint32 channels); + + +/* +Offsets a pointer by the specified number of PCM frames. +*/ +MA_API void* ma_offset_pcm_frames_ptr(void* p, ma_uint64 offsetInFrames, ma_format format, ma_uint32 channels); +MA_API const void* ma_offset_pcm_frames_const_ptr(const void* p, ma_uint64 offsetInFrames, ma_format format, ma_uint32 channels); +static MA_INLINE float* ma_offset_pcm_frames_ptr_f32(float* p, ma_uint64 offsetInFrames, ma_uint32 channels) { return (float*)ma_offset_pcm_frames_ptr((void*)p, offsetInFrames, ma_format_f32, channels); } +static MA_INLINE const float* ma_offset_pcm_frames_const_ptr_f32(const float* p, ma_uint64 offsetInFrames, ma_uint32 channels) { return (const float*)ma_offset_pcm_frames_const_ptr((const void*)p, offsetInFrames, ma_format_f32, channels); } + + +/* +Clips samples. +*/ +MA_API void ma_clip_samples_u8(ma_uint8* pDst, const ma_int16* pSrc, ma_uint64 count); +MA_API void ma_clip_samples_s16(ma_int16* pDst, const ma_int32* pSrc, ma_uint64 count); +MA_API void ma_clip_samples_s24(ma_uint8* pDst, const ma_int64* pSrc, ma_uint64 count); +MA_API void ma_clip_samples_s32(ma_int32* pDst, const ma_int64* pSrc, ma_uint64 count); +MA_API void ma_clip_samples_f32(float* pDst, const float* pSrc, ma_uint64 count); +MA_API void ma_clip_pcm_frames(void* pDst, const void* pSrc, ma_uint64 frameCount, ma_format format, ma_uint32 channels); + +/* +Helper for applying a volume factor to samples. + +Note that the source and destination buffers can be the same, in which case it'll perform the operation in-place. +*/ +MA_API void ma_copy_and_apply_volume_factor_u8(ma_uint8* pSamplesOut, const ma_uint8* pSamplesIn, ma_uint64 sampleCount, float factor); +MA_API void ma_copy_and_apply_volume_factor_s16(ma_int16* pSamplesOut, const ma_int16* pSamplesIn, ma_uint64 sampleCount, float factor); +MA_API void ma_copy_and_apply_volume_factor_s24(void* pSamplesOut, const void* pSamplesIn, ma_uint64 sampleCount, float factor); +MA_API void ma_copy_and_apply_volume_factor_s32(ma_int32* pSamplesOut, const ma_int32* pSamplesIn, ma_uint64 sampleCount, float factor); +MA_API void ma_copy_and_apply_volume_factor_f32(float* pSamplesOut, const float* pSamplesIn, ma_uint64 sampleCount, float factor); + +MA_API void ma_apply_volume_factor_u8(ma_uint8* pSamples, ma_uint64 sampleCount, float factor); +MA_API void ma_apply_volume_factor_s16(ma_int16* pSamples, ma_uint64 sampleCount, float factor); +MA_API void ma_apply_volume_factor_s24(void* pSamples, ma_uint64 sampleCount, float factor); +MA_API void ma_apply_volume_factor_s32(ma_int32* pSamples, ma_uint64 sampleCount, float factor); +MA_API void ma_apply_volume_factor_f32(float* pSamples, ma_uint64 sampleCount, float factor); + +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_u8(ma_uint8* pFramesOut, const ma_uint8* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_s16(ma_int16* pFramesOut, const ma_int16* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_s24(void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_s32(ma_int32* pFramesOut, const ma_int32* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_f32(float* pFramesOut, const float* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_copy_and_apply_volume_factor_pcm_frames(void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount, ma_format format, ma_uint32 channels, float factor); + +MA_API void ma_apply_volume_factor_pcm_frames_u8(ma_uint8* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_apply_volume_factor_pcm_frames_s16(ma_int16* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_apply_volume_factor_pcm_frames_s24(void* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_apply_volume_factor_pcm_frames_s32(ma_int32* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_apply_volume_factor_pcm_frames_f32(float* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor); +MA_API void ma_apply_volume_factor_pcm_frames(void* pFrames, ma_uint64 frameCount, ma_format format, ma_uint32 channels, float factor); + +MA_API void ma_copy_and_apply_volume_factor_per_channel_f32(float* pFramesOut, const float* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float* pChannelGains); + + +MA_API void ma_copy_and_apply_volume_and_clip_samples_u8(ma_uint8* pDst, const ma_int16* pSrc, ma_uint64 count, float volume); +MA_API void ma_copy_and_apply_volume_and_clip_samples_s16(ma_int16* pDst, const ma_int32* pSrc, ma_uint64 count, float volume); +MA_API void ma_copy_and_apply_volume_and_clip_samples_s24(ma_uint8* pDst, const ma_int64* pSrc, ma_uint64 count, float volume); +MA_API void ma_copy_and_apply_volume_and_clip_samples_s32(ma_int32* pDst, const ma_int64* pSrc, ma_uint64 count, float volume); +MA_API void ma_copy_and_apply_volume_and_clip_samples_f32(float* pDst, const float* pSrc, ma_uint64 count, float volume); +MA_API void ma_copy_and_apply_volume_and_clip_pcm_frames(void* pDst, const void* pSrc, ma_uint64 frameCount, ma_format format, ma_uint32 channels, float volume); + + +/* +Helper for converting a linear factor to gain in decibels. +*/ +MA_API float ma_volume_linear_to_db(float factor); + +/* +Helper for converting gain in decibels to a linear factor. +*/ +MA_API float ma_volume_db_to_linear(float gain); + + +/* +Mixes the specified number of frames in floating point format with a volume factor. + +This will run on an optimized path when the volume is equal to 1. +*/ +MA_API ma_result ma_mix_pcm_frames_f32(float* pDst, const float* pSrc, ma_uint64 frameCount, ma_uint32 channels, float volume); + + + + +/************************************************************************************************************************************************************ + +VFS +=== + +The VFS object (virtual file system) is what's used to customize file access. This is useful in cases where stdio FILE* based APIs may not be entirely +appropriate for a given situation. + +************************************************************************************************************************************************************/ +typedef void ma_vfs; +typedef ma_handle ma_vfs_file; + +typedef enum +{ + MA_OPEN_MODE_READ = 0x00000001, + MA_OPEN_MODE_WRITE = 0x00000002 +} ma_open_mode_flags; + +typedef enum +{ + ma_seek_origin_start, + ma_seek_origin_current, + ma_seek_origin_end /* Not used by decoders. */ +} ma_seek_origin; + +typedef struct +{ + ma_uint64 sizeInBytes; +} ma_file_info; + +typedef struct +{ + ma_result (* onOpen) (ma_vfs* pVFS, const char* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile); + ma_result (* onOpenW)(ma_vfs* pVFS, const wchar_t* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile); + ma_result (* onClose)(ma_vfs* pVFS, ma_vfs_file file); + ma_result (* onRead) (ma_vfs* pVFS, ma_vfs_file file, void* pDst, size_t sizeInBytes, size_t* pBytesRead); + ma_result (* onWrite)(ma_vfs* pVFS, ma_vfs_file file, const void* pSrc, size_t sizeInBytes, size_t* pBytesWritten); + ma_result (* onSeek) (ma_vfs* pVFS, ma_vfs_file file, ma_int64 offset, ma_seek_origin origin); + ma_result (* onTell) (ma_vfs* pVFS, ma_vfs_file file, ma_int64* pCursor); + ma_result (* onInfo) (ma_vfs* pVFS, ma_vfs_file file, ma_file_info* pInfo); +} ma_vfs_callbacks; + +MA_API ma_result ma_vfs_open(ma_vfs* pVFS, const char* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile); +MA_API ma_result ma_vfs_open_w(ma_vfs* pVFS, const wchar_t* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile); +MA_API ma_result ma_vfs_close(ma_vfs* pVFS, ma_vfs_file file); +MA_API ma_result ma_vfs_read(ma_vfs* pVFS, ma_vfs_file file, void* pDst, size_t sizeInBytes, size_t* pBytesRead); +MA_API ma_result ma_vfs_write(ma_vfs* pVFS, ma_vfs_file file, const void* pSrc, size_t sizeInBytes, size_t* pBytesWritten); +MA_API ma_result ma_vfs_seek(ma_vfs* pVFS, ma_vfs_file file, ma_int64 offset, ma_seek_origin origin); +MA_API ma_result ma_vfs_tell(ma_vfs* pVFS, ma_vfs_file file, ma_int64* pCursor); +MA_API ma_result ma_vfs_info(ma_vfs* pVFS, ma_vfs_file file, ma_file_info* pInfo); +MA_API ma_result ma_vfs_open_and_read_file(ma_vfs* pVFS, const char* pFilePath, void** ppData, size_t* pSize, const ma_allocation_callbacks* pAllocationCallbacks); + +typedef struct +{ + ma_vfs_callbacks cb; + ma_allocation_callbacks allocationCallbacks; /* Only used for the wchar_t version of open() on non-Windows platforms. */ +} ma_default_vfs; + +MA_API ma_result ma_default_vfs_init(ma_default_vfs* pVFS, const ma_allocation_callbacks* pAllocationCallbacks); + + + +typedef ma_result (* ma_read_proc)(void* pUserData, void* pBufferOut, size_t bytesToRead, size_t* pBytesRead); +typedef ma_result (* ma_seek_proc)(void* pUserData, ma_int64 offset, ma_seek_origin origin); +typedef ma_result (* ma_tell_proc)(void* pUserData, ma_int64* pCursor); + + + +#if !defined(MA_NO_DECODING) || !defined(MA_NO_ENCODING) +typedef enum +{ + ma_encoding_format_unknown = 0, + ma_encoding_format_wav, + ma_encoding_format_flac, + ma_encoding_format_mp3, + ma_encoding_format_vorbis +} ma_encoding_format; +#endif + +/************************************************************************************************************************************************************ + +Decoding +======== + +Decoders are independent of the main device API. Decoding APIs can be called freely inside the device's data callback, but they are not thread safe unless +you do your own synchronization. + +************************************************************************************************************************************************************/ +#ifndef MA_NO_DECODING +typedef struct ma_decoder ma_decoder; + + +typedef struct +{ + ma_format preferredFormat; + ma_uint32 seekPointCount; /* Set to > 0 to generate a seektable if the decoding backend supports it. */ +} ma_decoding_backend_config; + +MA_API ma_decoding_backend_config ma_decoding_backend_config_init(ma_format preferredFormat, ma_uint32 seekPointCount); + + +typedef struct +{ + ma_result (* onInit )(void* pUserData, ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend); + ma_result (* onInitFile )(void* pUserData, const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend); /* Optional. */ + ma_result (* onInitFileW )(void* pUserData, const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend); /* Optional. */ + ma_result (* onInitMemory)(void* pUserData, const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend); /* Optional. */ + void (* onUninit )(void* pUserData, ma_data_source* pBackend, const ma_allocation_callbacks* pAllocationCallbacks); +} ma_decoding_backend_vtable; + + +typedef ma_result (* ma_decoder_read_proc)(ma_decoder* pDecoder, void* pBufferOut, size_t bytesToRead, size_t* pBytesRead); /* Returns the number of bytes read. */ +typedef ma_result (* ma_decoder_seek_proc)(ma_decoder* pDecoder, ma_int64 byteOffset, ma_seek_origin origin); +typedef ma_result (* ma_decoder_tell_proc)(ma_decoder* pDecoder, ma_int64* pCursor); + +typedef struct +{ + ma_format format; /* Set to 0 or ma_format_unknown to use the stream's internal format. */ + ma_uint32 channels; /* Set to 0 to use the stream's internal channels. */ + ma_uint32 sampleRate; /* Set to 0 to use the stream's internal sample rate. */ + ma_channel* pChannelMap; + ma_channel_mix_mode channelMixMode; + ma_dither_mode ditherMode; + ma_resampler_config resampling; + ma_allocation_callbacks allocationCallbacks; + ma_encoding_format encodingFormat; + ma_uint32 seekPointCount; /* When set to > 0, specifies the number of seek points to use for the generation of a seek table. Not all decoding backends support this. */ + ma_decoding_backend_vtable** ppCustomBackendVTables; + ma_uint32 customBackendCount; + void* pCustomBackendUserData; +} ma_decoder_config; + +struct ma_decoder +{ + ma_data_source_base ds; + ma_data_source* pBackend; /* The decoding backend we'll be pulling data from. */ + const ma_decoding_backend_vtable* pBackendVTable; /* The vtable for the decoding backend. This needs to be stored so we can access the onUninit() callback. */ + void* pBackendUserData; + ma_decoder_read_proc onRead; + ma_decoder_seek_proc onSeek; + ma_decoder_tell_proc onTell; + void* pUserData; + ma_uint64 readPointerInPCMFrames; /* In output sample rate. Used for keeping track of how many frames are available for decoding. */ + ma_format outputFormat; + ma_uint32 outputChannels; + ma_uint32 outputSampleRate; + ma_data_converter converter; /* Data conversion is achieved by running frames through this. */ + void* pInputCache; /* In input format. Can be null if it's not needed. */ + ma_uint64 inputCacheCap; /* The capacity of the input cache. */ + ma_uint64 inputCacheConsumed; /* The number of frames that have been consumed in the cache. Used for determining the next valid frame. */ + ma_uint64 inputCacheRemaining; /* The number of valid frames remaining in the cache. */ + ma_allocation_callbacks allocationCallbacks; + union + { + struct + { + ma_vfs* pVFS; + ma_vfs_file file; + } vfs; + struct + { + const ma_uint8* pData; + size_t dataSize; + size_t currentReadPos; + } memory; /* Only used for decoders that were opened against a block of memory. */ + } data; +}; + +MA_API ma_decoder_config ma_decoder_config_init(ma_format outputFormat, ma_uint32 outputChannels, ma_uint32 outputSampleRate); +MA_API ma_decoder_config ma_decoder_config_init_default(void); + +MA_API ma_result ma_decoder_init(ma_decoder_read_proc onRead, ma_decoder_seek_proc onSeek, void* pUserData, const ma_decoder_config* pConfig, ma_decoder* pDecoder); +MA_API ma_result ma_decoder_init_memory(const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder); +MA_API ma_result ma_decoder_init_vfs(ma_vfs* pVFS, const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder); +MA_API ma_result ma_decoder_init_vfs_w(ma_vfs* pVFS, const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder); +MA_API ma_result ma_decoder_init_file(const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder); +MA_API ma_result ma_decoder_init_file_w(const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder); + +/* +Uninitializes a decoder. +*/ +MA_API ma_result ma_decoder_uninit(ma_decoder* pDecoder); + +/* +Reads PCM frames from the given decoder. + +This is not thread safe without your own synchronization. +*/ +MA_API ma_result ma_decoder_read_pcm_frames(ma_decoder* pDecoder, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); + +/* +Seeks to a PCM frame based on its absolute index. + +This is not thread safe without your own synchronization. +*/ +MA_API ma_result ma_decoder_seek_to_pcm_frame(ma_decoder* pDecoder, ma_uint64 frameIndex); + +/* +Retrieves the decoder's output data format. +*/ +MA_API ma_result ma_decoder_get_data_format(ma_decoder* pDecoder, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); + +/* +Retrieves the current position of the read cursor in PCM frames. +*/ +MA_API ma_result ma_decoder_get_cursor_in_pcm_frames(ma_decoder* pDecoder, ma_uint64* pCursor); + +/* +Retrieves the length of the decoder in PCM frames. + +Do not call this on streams of an undefined length, such as internet radio. + +If the length is unknown or an error occurs, 0 will be returned. + +This will always return 0 for Vorbis decoders. This is due to a limitation with stb_vorbis in push mode which is what miniaudio +uses internally. + +For MP3's, this will decode the entire file. Do not call this in time critical scenarios. + +This function is not thread safe without your own synchronization. +*/ +MA_API ma_result ma_decoder_get_length_in_pcm_frames(ma_decoder* pDecoder, ma_uint64* pLength); + +/* +Retrieves the number of frames that can be read before reaching the end. + +This calls `ma_decoder_get_length_in_pcm_frames()` so you need to be aware of the rules for that function, in +particular ensuring you do not call it on streams of an undefined length, such as internet radio. + +If the total length of the decoder cannot be retrieved, such as with Vorbis decoders, `MA_NOT_IMPLEMENTED` will be +returned. +*/ +MA_API ma_result ma_decoder_get_available_frames(ma_decoder* pDecoder, ma_uint64* pAvailableFrames); + +/* +Helper for opening and decoding a file into a heap allocated block of memory. Free the returned pointer with ma_free(). On input, +pConfig should be set to what you want. On output it will be set to what you got. +*/ +MA_API ma_result ma_decode_from_vfs(ma_vfs* pVFS, const char* pFilePath, ma_decoder_config* pConfig, ma_uint64* pFrameCountOut, void** ppPCMFramesOut); +MA_API ma_result ma_decode_file(const char* pFilePath, ma_decoder_config* pConfig, ma_uint64* pFrameCountOut, void** ppPCMFramesOut); +MA_API ma_result ma_decode_memory(const void* pData, size_t dataSize, ma_decoder_config* pConfig, ma_uint64* pFrameCountOut, void** ppPCMFramesOut); + +#endif /* MA_NO_DECODING */ + + +/************************************************************************************************************************************************************ + +Encoding +======== + +Encoders do not perform any format conversion for you. If your target format does not support the format, and error will be returned. + +************************************************************************************************************************************************************/ +#ifndef MA_NO_ENCODING +typedef struct ma_encoder ma_encoder; + +typedef ma_result (* ma_encoder_write_proc) (ma_encoder* pEncoder, const void* pBufferIn, size_t bytesToWrite, size_t* pBytesWritten); +typedef ma_result (* ma_encoder_seek_proc) (ma_encoder* pEncoder, ma_int64 offset, ma_seek_origin origin); +typedef ma_result (* ma_encoder_init_proc) (ma_encoder* pEncoder); +typedef void (* ma_encoder_uninit_proc) (ma_encoder* pEncoder); +typedef ma_result (* ma_encoder_write_pcm_frames_proc)(ma_encoder* pEncoder, const void* pFramesIn, ma_uint64 frameCount, ma_uint64* pFramesWritten); + +typedef struct +{ + ma_encoding_format encodingFormat; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_allocation_callbacks allocationCallbacks; +} ma_encoder_config; + +MA_API ma_encoder_config ma_encoder_config_init(ma_encoding_format encodingFormat, ma_format format, ma_uint32 channels, ma_uint32 sampleRate); + +struct ma_encoder +{ + ma_encoder_config config; + ma_encoder_write_proc onWrite; + ma_encoder_seek_proc onSeek; + ma_encoder_init_proc onInit; + ma_encoder_uninit_proc onUninit; + ma_encoder_write_pcm_frames_proc onWritePCMFrames; + void* pUserData; + void* pInternalEncoder; + union + { + struct + { + ma_vfs* pVFS; + ma_vfs_file file; + } vfs; + } data; +}; + +MA_API ma_result ma_encoder_init(ma_encoder_write_proc onWrite, ma_encoder_seek_proc onSeek, void* pUserData, const ma_encoder_config* pConfig, ma_encoder* pEncoder); +MA_API ma_result ma_encoder_init_vfs(ma_vfs* pVFS, const char* pFilePath, const ma_encoder_config* pConfig, ma_encoder* pEncoder); +MA_API ma_result ma_encoder_init_vfs_w(ma_vfs* pVFS, const wchar_t* pFilePath, const ma_encoder_config* pConfig, ma_encoder* pEncoder); +MA_API ma_result ma_encoder_init_file(const char* pFilePath, const ma_encoder_config* pConfig, ma_encoder* pEncoder); +MA_API ma_result ma_encoder_init_file_w(const wchar_t* pFilePath, const ma_encoder_config* pConfig, ma_encoder* pEncoder); +MA_API void ma_encoder_uninit(ma_encoder* pEncoder); +MA_API ma_result ma_encoder_write_pcm_frames(ma_encoder* pEncoder, const void* pFramesIn, ma_uint64 frameCount, ma_uint64* pFramesWritten); + +#endif /* MA_NO_ENCODING */ + + +/************************************************************************************************************************************************************ + +Generation + +************************************************************************************************************************************************************/ +#ifndef MA_NO_GENERATION +typedef enum +{ + ma_waveform_type_sine, + ma_waveform_type_square, + ma_waveform_type_triangle, + ma_waveform_type_sawtooth +} ma_waveform_type; + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_waveform_type type; + double amplitude; + double frequency; +} ma_waveform_config; + +MA_API ma_waveform_config ma_waveform_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, ma_waveform_type type, double amplitude, double frequency); + +typedef struct +{ + ma_data_source_base ds; + ma_waveform_config config; + double advance; + double time; +} ma_waveform; + +MA_API ma_result ma_waveform_init(const ma_waveform_config* pConfig, ma_waveform* pWaveform); +MA_API void ma_waveform_uninit(ma_waveform* pWaveform); +MA_API ma_result ma_waveform_read_pcm_frames(ma_waveform* pWaveform, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_waveform_seek_to_pcm_frame(ma_waveform* pWaveform, ma_uint64 frameIndex); +MA_API ma_result ma_waveform_set_amplitude(ma_waveform* pWaveform, double amplitude); +MA_API ma_result ma_waveform_set_frequency(ma_waveform* pWaveform, double frequency); +MA_API ma_result ma_waveform_set_type(ma_waveform* pWaveform, ma_waveform_type type); +MA_API ma_result ma_waveform_set_sample_rate(ma_waveform* pWaveform, ma_uint32 sampleRate); + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + double dutyCycle; + double amplitude; + double frequency; +} ma_pulsewave_config; + +MA_API ma_pulsewave_config ma_pulsewave_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double dutyCycle, double amplitude, double frequency); + +typedef struct +{ + ma_waveform waveform; + ma_pulsewave_config config; +} ma_pulsewave; + +MA_API ma_result ma_pulsewave_init(const ma_pulsewave_config* pConfig, ma_pulsewave* pWaveform); +MA_API void ma_pulsewave_uninit(ma_pulsewave* pWaveform); +MA_API ma_result ma_pulsewave_read_pcm_frames(ma_pulsewave* pWaveform, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_pulsewave_seek_to_pcm_frame(ma_pulsewave* pWaveform, ma_uint64 frameIndex); +MA_API ma_result ma_pulsewave_set_amplitude(ma_pulsewave* pWaveform, double amplitude); +MA_API ma_result ma_pulsewave_set_frequency(ma_pulsewave* pWaveform, double frequency); +MA_API ma_result ma_pulsewave_set_sample_rate(ma_pulsewave* pWaveform, ma_uint32 sampleRate); +MA_API ma_result ma_pulsewave_set_duty_cycle(ma_pulsewave* pWaveform, double dutyCycle); + +typedef enum +{ + ma_noise_type_white, + ma_noise_type_pink, + ma_noise_type_brownian +} ma_noise_type; + + +typedef struct +{ + ma_format format; + ma_uint32 channels; + ma_noise_type type; + ma_int32 seed; + double amplitude; + ma_bool32 duplicateChannels; +} ma_noise_config; + +MA_API ma_noise_config ma_noise_config_init(ma_format format, ma_uint32 channels, ma_noise_type type, ma_int32 seed, double amplitude); + +typedef struct +{ + ma_data_source_base ds; + ma_noise_config config; + ma_lcg lcg; + union + { + struct + { + double** bin; + double* accumulation; + ma_uint32* counter; + } pink; + struct + { + double* accumulation; + } brownian; + } state; + + /* Memory management. */ + void* _pHeap; + ma_bool32 _ownsHeap; +} ma_noise; + +MA_API ma_result ma_noise_get_heap_size(const ma_noise_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_noise_init_preallocated(const ma_noise_config* pConfig, void* pHeap, ma_noise* pNoise); +MA_API ma_result ma_noise_init(const ma_noise_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_noise* pNoise); +MA_API void ma_noise_uninit(ma_noise* pNoise, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_noise_read_pcm_frames(ma_noise* pNoise, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_noise_set_amplitude(ma_noise* pNoise, double amplitude); +MA_API ma_result ma_noise_set_seed(ma_noise* pNoise, ma_int32 seed); +MA_API ma_result ma_noise_set_type(ma_noise* pNoise, ma_noise_type type); + +#endif /* MA_NO_GENERATION */ + + + +/************************************************************************************************************************************************************ + +Resource Manager + +************************************************************************************************************************************************************/ +/* The resource manager cannot be enabled if there is no decoder. */ +#if !defined(MA_NO_RESOURCE_MANAGER) && defined(MA_NO_DECODING) +#define MA_NO_RESOURCE_MANAGER +#endif + +#ifndef MA_NO_RESOURCE_MANAGER +typedef struct ma_resource_manager ma_resource_manager; +typedef struct ma_resource_manager_data_buffer_node ma_resource_manager_data_buffer_node; +typedef struct ma_resource_manager_data_buffer ma_resource_manager_data_buffer; +typedef struct ma_resource_manager_data_stream ma_resource_manager_data_stream; +typedef struct ma_resource_manager_data_source ma_resource_manager_data_source; + +typedef enum +{ + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM = 0x00000001, /* When set, does not load the entire data source in memory. Disk I/O will happen on job threads. */ + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_DECODE = 0x00000002, /* Decode data before storing in memory. When set, decoding is done at the resource manager level rather than the mixing thread. Results in faster mixing, but higher memory usage. */ + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC = 0x00000004, /* When set, the resource manager will load the data source asynchronously. */ + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT = 0x00000008, /* When set, waits for initialization of the underlying data source before returning from ma_resource_manager_data_source_init(). */ + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_UNKNOWN_LENGTH = 0x00000010, /* Gives the resource manager a hint that the length of the data source is unknown and calling `ma_data_source_get_length_in_pcm_frames()` should be avoided. */ + MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING = 0x00000020 /* When set, configures the data source to loop by default. */ +} ma_resource_manager_data_source_flags; + + +/* +Pipeline notifications used by the resource manager. Made up of both an async notification and a fence, both of which are optional. +*/ +typedef struct +{ + ma_async_notification* pNotification; + ma_fence* pFence; +} ma_resource_manager_pipeline_stage_notification; + +typedef struct +{ + ma_resource_manager_pipeline_stage_notification init; /* Initialization of the decoder. */ + ma_resource_manager_pipeline_stage_notification done; /* Decoding fully completed. */ +} ma_resource_manager_pipeline_notifications; + +MA_API ma_resource_manager_pipeline_notifications ma_resource_manager_pipeline_notifications_init(void); + + + +/* BEGIN BACKWARDS COMPATIBILITY */ +/* TODO: Remove this block in version 0.12. */ +#if 1 +#define ma_resource_manager_job ma_job +#define ma_resource_manager_job_init ma_job_init +#define MA_JOB_TYPE_RESOURCE_MANAGER_QUEUE_FLAG_NON_BLOCKING MA_JOB_QUEUE_FLAG_NON_BLOCKING +#define ma_resource_manager_job_queue_config ma_job_queue_config +#define ma_resource_manager_job_queue_config_init ma_job_queue_config_init +#define ma_resource_manager_job_queue ma_job_queue +#define ma_resource_manager_job_queue_get_heap_size ma_job_queue_get_heap_size +#define ma_resource_manager_job_queue_init_preallocated ma_job_queue_init_preallocated +#define ma_resource_manager_job_queue_init ma_job_queue_init +#define ma_resource_manager_job_queue_uninit ma_job_queue_uninit +#define ma_resource_manager_job_queue_post ma_job_queue_post +#define ma_resource_manager_job_queue_next ma_job_queue_next +#endif +/* END BACKWARDS COMPATIBILITY */ + + + + +/* Maximum job thread count will be restricted to this, but this may be removed later and replaced with a heap allocation thereby removing any limitation. */ +#ifndef MA_RESOURCE_MANAGER_MAX_JOB_THREAD_COUNT +#define MA_RESOURCE_MANAGER_MAX_JOB_THREAD_COUNT 64 +#endif + +typedef enum +{ + /* Indicates ma_resource_manager_next_job() should not block. Only valid when the job thread count is 0. */ + MA_RESOURCE_MANAGER_FLAG_NON_BLOCKING = 0x00000001, + + /* Disables any kind of multithreading. Implicitly enables MA_RESOURCE_MANAGER_FLAG_NON_BLOCKING. */ + MA_RESOURCE_MANAGER_FLAG_NO_THREADING = 0x00000002 +} ma_resource_manager_flags; + +typedef struct +{ + const char* pFilePath; + const wchar_t* pFilePathW; + const ma_resource_manager_pipeline_notifications* pNotifications; + ma_uint64 initialSeekPointInPCMFrames; + ma_uint64 rangeBegInPCMFrames; + ma_uint64 rangeEndInPCMFrames; + ma_uint64 loopPointBegInPCMFrames; + ma_uint64 loopPointEndInPCMFrames; + ma_uint32 flags; + ma_bool32 isLooping; /* Deprecated. Use the MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING flag in `flags` instead. */ +} ma_resource_manager_data_source_config; + +MA_API ma_resource_manager_data_source_config ma_resource_manager_data_source_config_init(void); + + +typedef enum +{ + ma_resource_manager_data_supply_type_unknown = 0, /* Used for determining whether or the data supply has been initialized. */ + ma_resource_manager_data_supply_type_encoded, /* Data supply is an encoded buffer. Connector is ma_decoder. */ + ma_resource_manager_data_supply_type_decoded, /* Data supply is a decoded buffer. Connector is ma_audio_buffer. */ + ma_resource_manager_data_supply_type_decoded_paged /* Data supply is a linked list of decoded buffers. Connector is ma_paged_audio_buffer. */ +} ma_resource_manager_data_supply_type; + +typedef struct +{ + MA_ATOMIC(4, ma_resource_manager_data_supply_type) type; /* Read and written from different threads so needs to be accessed atomically. */ + union + { + struct + { + const void* pData; + size_t sizeInBytes; + } encoded; + struct + { + const void* pData; + ma_uint64 totalFrameCount; + ma_uint64 decodedFrameCount; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + } decoded; + struct + { + ma_paged_audio_buffer_data data; + ma_uint64 decodedFrameCount; + ma_uint32 sampleRate; + } decodedPaged; + } backend; +} ma_resource_manager_data_supply; + +struct ma_resource_manager_data_buffer_node +{ + ma_uint32 hashedName32; /* The hashed name. This is the key. */ + ma_uint32 refCount; + MA_ATOMIC(4, ma_result) result; /* Result from asynchronous loading. When loading set to MA_BUSY. When fully loaded set to MA_SUCCESS. When deleting set to MA_UNAVAILABLE. */ + MA_ATOMIC(4, ma_uint32) executionCounter; /* For allocating execution orders for jobs. */ + MA_ATOMIC(4, ma_uint32) executionPointer; /* For managing the order of execution for asynchronous jobs relating to this object. Incremented as jobs complete processing. */ + ma_bool32 isDataOwnedByResourceManager; /* Set to true when the underlying data buffer was allocated the resource manager. Set to false if it is owned by the application (via ma_resource_manager_register_*()). */ + ma_resource_manager_data_supply data; + ma_resource_manager_data_buffer_node* pParent; + ma_resource_manager_data_buffer_node* pChildLo; + ma_resource_manager_data_buffer_node* pChildHi; +}; + +struct ma_resource_manager_data_buffer +{ + ma_data_source_base ds; /* Base data source. A data buffer is a data source. */ + ma_resource_manager* pResourceManager; /* A pointer to the resource manager that owns this buffer. */ + ma_resource_manager_data_buffer_node* pNode; /* The data node. This is reference counted and is what supplies the data. */ + ma_uint32 flags; /* The flags that were passed used to initialize the buffer. */ + MA_ATOMIC(4, ma_uint32) executionCounter; /* For allocating execution orders for jobs. */ + MA_ATOMIC(4, ma_uint32) executionPointer; /* For managing the order of execution for asynchronous jobs relating to this object. Incremented as jobs complete processing. */ + ma_uint64 seekTargetInPCMFrames; /* Only updated by the public API. Never written nor read from the job thread. */ + ma_bool32 seekToCursorOnNextRead; /* On the next read we need to seek to the frame cursor. */ + MA_ATOMIC(4, ma_result) result; /* Keeps track of a result of decoding. Set to MA_BUSY while the buffer is still loading. Set to MA_SUCCESS when loading is finished successfully. Otherwise set to some other code. */ + MA_ATOMIC(4, ma_bool32) isLooping; /* Can be read and written by different threads at the same time. Must be used atomically. */ + ma_atomic_bool32 isConnectorInitialized; /* Used for asynchronous loading to ensure we don't try to initialize the connector multiple times while waiting for the node to fully load. */ + union + { + ma_decoder decoder; /* Supply type is ma_resource_manager_data_supply_type_encoded */ + ma_audio_buffer buffer; /* Supply type is ma_resource_manager_data_supply_type_decoded */ + ma_paged_audio_buffer pagedBuffer; /* Supply type is ma_resource_manager_data_supply_type_decoded_paged */ + } connector; /* Connects this object to the node's data supply. */ +}; + +struct ma_resource_manager_data_stream +{ + ma_data_source_base ds; /* Base data source. A data stream is a data source. */ + ma_resource_manager* pResourceManager; /* A pointer to the resource manager that owns this data stream. */ + ma_uint32 flags; /* The flags that were passed used to initialize the stream. */ + ma_decoder decoder; /* Used for filling pages with data. This is only ever accessed by the job thread. The public API should never touch this. */ + ma_bool32 isDecoderInitialized; /* Required for determining whether or not the decoder should be uninitialized in MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_STREAM. */ + ma_uint64 totalLengthInPCMFrames; /* This is calculated when first loaded by the MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_STREAM. */ + ma_uint32 relativeCursor; /* The playback cursor, relative to the current page. Only ever accessed by the public API. Never accessed by the job thread. */ + MA_ATOMIC(8, ma_uint64) absoluteCursor; /* The playback cursor, in absolute position starting from the start of the file. */ + ma_uint32 currentPageIndex; /* Toggles between 0 and 1. Index 0 is the first half of pPageData. Index 1 is the second half. Only ever accessed by the public API. Never accessed by the job thread. */ + MA_ATOMIC(4, ma_uint32) executionCounter; /* For allocating execution orders for jobs. */ + MA_ATOMIC(4, ma_uint32) executionPointer; /* For managing the order of execution for asynchronous jobs relating to this object. Incremented as jobs complete processing. */ + + /* Written by the public API, read by the job thread. */ + MA_ATOMIC(4, ma_bool32) isLooping; /* Whether or not the stream is looping. It's important to set the looping flag at the data stream level for smooth loop transitions. */ + + /* Written by the job thread, read by the public API. */ + void* pPageData; /* Buffer containing the decoded data of each page. Allocated once at initialization time. */ + MA_ATOMIC(4, ma_uint32) pageFrameCount[2]; /* The number of valid PCM frames in each page. Used to determine the last valid frame. */ + + /* Written and read by both the public API and the job thread. These must be atomic. */ + MA_ATOMIC(4, ma_result) result; /* Result from asynchronous loading. When loading set to MA_BUSY. When initialized set to MA_SUCCESS. When deleting set to MA_UNAVAILABLE. If an error occurs when loading, set to an error code. */ + MA_ATOMIC(4, ma_bool32) isDecoderAtEnd; /* Whether or not the decoder has reached the end. */ + MA_ATOMIC(4, ma_bool32) isPageValid[2]; /* Booleans to indicate whether or not a page is valid. Set to false by the public API, set to true by the job thread. Set to false as the pages are consumed, true when they are filled. */ + MA_ATOMIC(4, ma_bool32) seekCounter; /* When 0, no seeking is being performed. When > 0, a seek is being performed and reading should be delayed with MA_BUSY. */ +}; + +struct ma_resource_manager_data_source +{ + union + { + ma_resource_manager_data_buffer buffer; + ma_resource_manager_data_stream stream; + } backend; /* Must be the first item because we need the first item to be the data source callbacks for the buffer or stream. */ + + ma_uint32 flags; /* The flags that were passed in to ma_resource_manager_data_source_init(). */ + MA_ATOMIC(4, ma_uint32) executionCounter; /* For allocating execution orders for jobs. */ + MA_ATOMIC(4, ma_uint32) executionPointer; /* For managing the order of execution for asynchronous jobs relating to this object. Incremented as jobs complete processing. */ +}; + +typedef struct +{ + ma_allocation_callbacks allocationCallbacks; + ma_log* pLog; + ma_format decodedFormat; /* The decoded format to use. Set to ma_format_unknown (default) to use the file's native format. */ + ma_uint32 decodedChannels; /* The decoded channel count to use. Set to 0 (default) to use the file's native channel count. */ + ma_uint32 decodedSampleRate; /* the decoded sample rate to use. Set to 0 (default) to use the file's native sample rate. */ + ma_uint32 jobThreadCount; /* Set to 0 if you want to self-manage your job threads. Defaults to 1. */ + size_t jobThreadStackSize; + ma_uint32 jobQueueCapacity; /* The maximum number of jobs that can fit in the queue at a time. Defaults to MA_JOB_TYPE_RESOURCE_MANAGER_QUEUE_CAPACITY. Cannot be zero. */ + ma_uint32 flags; + ma_vfs* pVFS; /* Can be NULL in which case defaults will be used. */ + ma_decoding_backend_vtable** ppCustomDecodingBackendVTables; + ma_uint32 customDecodingBackendCount; + void* pCustomDecodingBackendUserData; +} ma_resource_manager_config; + +MA_API ma_resource_manager_config ma_resource_manager_config_init(void); + +struct ma_resource_manager +{ + ma_resource_manager_config config; + ma_resource_manager_data_buffer_node* pRootDataBufferNode; /* The root buffer in the binary tree. */ +#ifndef MA_NO_THREADING + ma_mutex dataBufferBSTLock; /* For synchronizing access to the data buffer binary tree. */ + ma_thread jobThreads[MA_RESOURCE_MANAGER_MAX_JOB_THREAD_COUNT]; /* The threads for executing jobs. */ +#endif + ma_job_queue jobQueue; /* Multi-consumer, multi-producer job queue for managing jobs for asynchronous decoding and streaming. */ + ma_default_vfs defaultVFS; /* Only used if a custom VFS is not specified. */ + ma_log log; /* Only used if no log was specified in the config. */ +}; + +/* Init. */ +MA_API ma_result ma_resource_manager_init(const ma_resource_manager_config* pConfig, ma_resource_manager* pResourceManager); +MA_API void ma_resource_manager_uninit(ma_resource_manager* pResourceManager); +MA_API ma_log* ma_resource_manager_get_log(ma_resource_manager* pResourceManager); + +/* Registration. */ +MA_API ma_result ma_resource_manager_register_file(ma_resource_manager* pResourceManager, const char* pFilePath, ma_uint32 flags); +MA_API ma_result ma_resource_manager_register_file_w(ma_resource_manager* pResourceManager, const wchar_t* pFilePath, ma_uint32 flags); +MA_API ma_result ma_resource_manager_register_decoded_data(ma_resource_manager* pResourceManager, const char* pName, const void* pData, ma_uint64 frameCount, ma_format format, ma_uint32 channels, ma_uint32 sampleRate); /* Does not copy. Increments the reference count if already exists and returns MA_SUCCESS. */ +MA_API ma_result ma_resource_manager_register_decoded_data_w(ma_resource_manager* pResourceManager, const wchar_t* pName, const void* pData, ma_uint64 frameCount, ma_format format, ma_uint32 channels, ma_uint32 sampleRate); +MA_API ma_result ma_resource_manager_register_encoded_data(ma_resource_manager* pResourceManager, const char* pName, const void* pData, size_t sizeInBytes); /* Does not copy. Increments the reference count if already exists and returns MA_SUCCESS. */ +MA_API ma_result ma_resource_manager_register_encoded_data_w(ma_resource_manager* pResourceManager, const wchar_t* pName, const void* pData, size_t sizeInBytes); +MA_API ma_result ma_resource_manager_unregister_file(ma_resource_manager* pResourceManager, const char* pFilePath); +MA_API ma_result ma_resource_manager_unregister_file_w(ma_resource_manager* pResourceManager, const wchar_t* pFilePath); +MA_API ma_result ma_resource_manager_unregister_data(ma_resource_manager* pResourceManager, const char* pName); +MA_API ma_result ma_resource_manager_unregister_data_w(ma_resource_manager* pResourceManager, const wchar_t* pName); + +/* Data Buffers. */ +MA_API ma_result ma_resource_manager_data_buffer_init_ex(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source_config* pConfig, ma_resource_manager_data_buffer* pDataBuffer); +MA_API ma_result ma_resource_manager_data_buffer_init(ma_resource_manager* pResourceManager, const char* pFilePath, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_buffer* pDataBuffer); +MA_API ma_result ma_resource_manager_data_buffer_init_w(ma_resource_manager* pResourceManager, const wchar_t* pFilePath, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_buffer* pDataBuffer); +MA_API ma_result ma_resource_manager_data_buffer_init_copy(ma_resource_manager* pResourceManager, const ma_resource_manager_data_buffer* pExistingDataBuffer, ma_resource_manager_data_buffer* pDataBuffer); +MA_API ma_result ma_resource_manager_data_buffer_uninit(ma_resource_manager_data_buffer* pDataBuffer); +MA_API ma_result ma_resource_manager_data_buffer_read_pcm_frames(ma_resource_manager_data_buffer* pDataBuffer, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_resource_manager_data_buffer_seek_to_pcm_frame(ma_resource_manager_data_buffer* pDataBuffer, ma_uint64 frameIndex); +MA_API ma_result ma_resource_manager_data_buffer_get_data_format(ma_resource_manager_data_buffer* pDataBuffer, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_resource_manager_data_buffer_get_cursor_in_pcm_frames(ma_resource_manager_data_buffer* pDataBuffer, ma_uint64* pCursor); +MA_API ma_result ma_resource_manager_data_buffer_get_length_in_pcm_frames(ma_resource_manager_data_buffer* pDataBuffer, ma_uint64* pLength); +MA_API ma_result ma_resource_manager_data_buffer_result(const ma_resource_manager_data_buffer* pDataBuffer); +MA_API ma_result ma_resource_manager_data_buffer_set_looping(ma_resource_manager_data_buffer* pDataBuffer, ma_bool32 isLooping); +MA_API ma_bool32 ma_resource_manager_data_buffer_is_looping(const ma_resource_manager_data_buffer* pDataBuffer); +MA_API ma_result ma_resource_manager_data_buffer_get_available_frames(ma_resource_manager_data_buffer* pDataBuffer, ma_uint64* pAvailableFrames); + +/* Data Streams. */ +MA_API ma_result ma_resource_manager_data_stream_init_ex(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source_config* pConfig, ma_resource_manager_data_stream* pDataStream); +MA_API ma_result ma_resource_manager_data_stream_init(ma_resource_manager* pResourceManager, const char* pFilePath, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_stream* pDataStream); +MA_API ma_result ma_resource_manager_data_stream_init_w(ma_resource_manager* pResourceManager, const wchar_t* pFilePath, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_stream* pDataStream); +MA_API ma_result ma_resource_manager_data_stream_uninit(ma_resource_manager_data_stream* pDataStream); +MA_API ma_result ma_resource_manager_data_stream_read_pcm_frames(ma_resource_manager_data_stream* pDataStream, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_resource_manager_data_stream_seek_to_pcm_frame(ma_resource_manager_data_stream* pDataStream, ma_uint64 frameIndex); +MA_API ma_result ma_resource_manager_data_stream_get_data_format(ma_resource_manager_data_stream* pDataStream, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_resource_manager_data_stream_get_cursor_in_pcm_frames(ma_resource_manager_data_stream* pDataStream, ma_uint64* pCursor); +MA_API ma_result ma_resource_manager_data_stream_get_length_in_pcm_frames(ma_resource_manager_data_stream* pDataStream, ma_uint64* pLength); +MA_API ma_result ma_resource_manager_data_stream_result(const ma_resource_manager_data_stream* pDataStream); +MA_API ma_result ma_resource_manager_data_stream_set_looping(ma_resource_manager_data_stream* pDataStream, ma_bool32 isLooping); +MA_API ma_bool32 ma_resource_manager_data_stream_is_looping(const ma_resource_manager_data_stream* pDataStream); +MA_API ma_result ma_resource_manager_data_stream_get_available_frames(ma_resource_manager_data_stream* pDataStream, ma_uint64* pAvailableFrames); + +/* Data Sources. */ +MA_API ma_result ma_resource_manager_data_source_init_ex(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source_config* pConfig, ma_resource_manager_data_source* pDataSource); +MA_API ma_result ma_resource_manager_data_source_init(ma_resource_manager* pResourceManager, const char* pName, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_source* pDataSource); +MA_API ma_result ma_resource_manager_data_source_init_w(ma_resource_manager* pResourceManager, const wchar_t* pName, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_source* pDataSource); +MA_API ma_result ma_resource_manager_data_source_init_copy(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source* pExistingDataSource, ma_resource_manager_data_source* pDataSource); +MA_API ma_result ma_resource_manager_data_source_uninit(ma_resource_manager_data_source* pDataSource); +MA_API ma_result ma_resource_manager_data_source_read_pcm_frames(ma_resource_manager_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_resource_manager_data_source_seek_to_pcm_frame(ma_resource_manager_data_source* pDataSource, ma_uint64 frameIndex); +MA_API ma_result ma_resource_manager_data_source_get_data_format(ma_resource_manager_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_resource_manager_data_source_get_cursor_in_pcm_frames(ma_resource_manager_data_source* pDataSource, ma_uint64* pCursor); +MA_API ma_result ma_resource_manager_data_source_get_length_in_pcm_frames(ma_resource_manager_data_source* pDataSource, ma_uint64* pLength); +MA_API ma_result ma_resource_manager_data_source_result(const ma_resource_manager_data_source* pDataSource); +MA_API ma_result ma_resource_manager_data_source_set_looping(ma_resource_manager_data_source* pDataSource, ma_bool32 isLooping); +MA_API ma_bool32 ma_resource_manager_data_source_is_looping(const ma_resource_manager_data_source* pDataSource); +MA_API ma_result ma_resource_manager_data_source_get_available_frames(ma_resource_manager_data_source* pDataSource, ma_uint64* pAvailableFrames); + +/* Job management. */ +MA_API ma_result ma_resource_manager_post_job(ma_resource_manager* pResourceManager, const ma_job* pJob); +MA_API ma_result ma_resource_manager_post_job_quit(ma_resource_manager* pResourceManager); /* Helper for posting a quit job. */ +MA_API ma_result ma_resource_manager_next_job(ma_resource_manager* pResourceManager, ma_job* pJob); +MA_API ma_result ma_resource_manager_process_job(ma_resource_manager* pResourceManager, ma_job* pJob); /* DEPRECATED. Use ma_job_process(). Will be removed in version 0.12. */ +MA_API ma_result ma_resource_manager_process_next_job(ma_resource_manager* pResourceManager); /* Returns MA_CANCELLED if a MA_JOB_TYPE_QUIT job is found. In non-blocking mode, returns MA_NO_DATA_AVAILABLE if no jobs are available. */ +#endif /* MA_NO_RESOURCE_MANAGER */ + + + +/************************************************************************************************************************************************************ + +Node Graph + +************************************************************************************************************************************************************/ +#ifndef MA_NO_NODE_GRAPH +/* Must never exceed 254. */ +#ifndef MA_MAX_NODE_BUS_COUNT +#define MA_MAX_NODE_BUS_COUNT 254 +#endif + +/* Used internally by miniaudio for memory management. Must never exceed MA_MAX_NODE_BUS_COUNT. */ +#ifndef MA_MAX_NODE_LOCAL_BUS_COUNT +#define MA_MAX_NODE_LOCAL_BUS_COUNT 2 +#endif + +/* Use this when the bus count is determined by the node instance rather than the vtable. */ +#define MA_NODE_BUS_COUNT_UNKNOWN 255 + + +/* For some internal memory management of ma_node_graph. */ +typedef struct +{ + size_t offset; + size_t sizeInBytes; + unsigned char _data[1]; +} ma_stack; + + +typedef struct ma_node_graph ma_node_graph; +typedef void ma_node; + + +/* Node flags. */ +typedef enum +{ + MA_NODE_FLAG_PASSTHROUGH = 0x00000001, + MA_NODE_FLAG_CONTINUOUS_PROCESSING = 0x00000002, + MA_NODE_FLAG_ALLOW_NULL_INPUT = 0x00000004, + MA_NODE_FLAG_DIFFERENT_PROCESSING_RATES = 0x00000008, + MA_NODE_FLAG_SILENT_OUTPUT = 0x00000010 +} ma_node_flags; + + +/* The playback state of a node. Either started or stopped. */ +typedef enum +{ + ma_node_state_started = 0, + ma_node_state_stopped = 1 +} ma_node_state; + + +typedef struct +{ + /* + Extended processing callback. This callback is used for effects that process input and output + at different rates (i.e. they perform resampling). This is similar to the simple version, only + they take two separate frame counts: one for input, and one for output. + + On input, `pFrameCountOut` is equal to the capacity of the output buffer for each bus, whereas + `pFrameCountIn` will be equal to the number of PCM frames in each of the buffers in `ppFramesIn`. + + On output, set `pFrameCountOut` to the number of PCM frames that were actually output and set + `pFrameCountIn` to the number of input frames that were consumed. + */ + void (* onProcess)(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut); + + /* + A callback for retrieving the number of input frames that are required to output the + specified number of output frames. You would only want to implement this when the node performs + resampling. This is optional, even for nodes that perform resampling, but it does offer a + small reduction in latency as it allows miniaudio to calculate the exact number of input frames + to read at a time instead of having to estimate. + */ + ma_result (* onGetRequiredInputFrameCount)(ma_node* pNode, ma_uint32 outputFrameCount, ma_uint32* pInputFrameCount); + + /* + The number of input buses. This is how many sub-buffers will be contained in the `ppFramesIn` + parameters of the callbacks above. + */ + ma_uint8 inputBusCount; + + /* + The number of output buses. This is how many sub-buffers will be contained in the `ppFramesOut` + parameters of the callbacks above. + */ + ma_uint8 outputBusCount; + + /* + Flags describing characteristics of the node. This is currently just a placeholder for some + ideas for later on. + */ + ma_uint32 flags; +} ma_node_vtable; + +typedef struct +{ + const ma_node_vtable* vtable; /* Should never be null. Initialization of the node will fail if so. */ + ma_node_state initialState; /* Defaults to ma_node_state_started. */ + ma_uint32 inputBusCount; /* Only used if the vtable specifies an input bus count of `MA_NODE_BUS_COUNT_UNKNOWN`, otherwise must be set to `MA_NODE_BUS_COUNT_UNKNOWN` (default). */ + ma_uint32 outputBusCount; /* Only used if the vtable specifies an output bus count of `MA_NODE_BUS_COUNT_UNKNOWN`, otherwise be set to `MA_NODE_BUS_COUNT_UNKNOWN` (default). */ + const ma_uint32* pInputChannels; /* The number of elements are determined by the input bus count as determined by the vtable, or `inputBusCount` if the vtable specifies `MA_NODE_BUS_COUNT_UNKNOWN`. */ + const ma_uint32* pOutputChannels; /* The number of elements are determined by the output bus count as determined by the vtable, or `outputBusCount` if the vtable specifies `MA_NODE_BUS_COUNT_UNKNOWN`. */ +} ma_node_config; + +MA_API ma_node_config ma_node_config_init(void); + + +/* +A node has multiple output buses. An output bus is attached to an input bus as an item in a linked +list. Think of the input bus as a linked list, with the output bus being an item in that list. +*/ +typedef struct ma_node_output_bus ma_node_output_bus; +struct ma_node_output_bus +{ + /* Immutable. */ + ma_node* pNode; /* The node that owns this output bus. The input node. Will be null for dummy head and tail nodes. */ + ma_uint8 outputBusIndex; /* The index of the output bus on pNode that this output bus represents. */ + ma_uint8 channels; /* The number of channels in the audio stream for this bus. */ + + /* Mutable via multiple threads. Must be used atomically. The weird ordering here is for packing reasons. */ + ma_uint8 inputNodeInputBusIndex; /* The index of the input bus on the input. Required for detaching. Will only be used within the spinlock so does not need to be atomic. */ + MA_ATOMIC(4, ma_uint32) flags; /* Some state flags for tracking the read state of the output buffer. A combination of MA_NODE_OUTPUT_BUS_FLAG_*. */ + MA_ATOMIC(4, ma_uint32) refCount; /* Reference count for some thread-safety when detaching. */ + MA_ATOMIC(4, ma_bool32) isAttached; /* This is used to prevent iteration of nodes that are in the middle of being detached. Used for thread safety. */ + MA_ATOMIC(4, ma_spinlock) lock; /* Unfortunate lock, but significantly simplifies the implementation. Required for thread-safe attaching and detaching. */ + MA_ATOMIC(4, float) volume; /* Linear. */ + MA_ATOMIC(MA_SIZEOF_PTR, ma_node_output_bus*) pNext; /* If null, it's the tail node or detached. */ + MA_ATOMIC(MA_SIZEOF_PTR, ma_node_output_bus*) pPrev; /* If null, it's the head node or detached. */ + MA_ATOMIC(MA_SIZEOF_PTR, ma_node*) pInputNode; /* The node that this output bus is attached to. Required for detaching. */ +}; + +/* +A node has multiple input buses. The output buses of a node are connecting to the input busses of +another. An input bus is essentially just a linked list of output buses. +*/ +typedef struct ma_node_input_bus ma_node_input_bus; +struct ma_node_input_bus +{ + /* Mutable via multiple threads. */ + ma_node_output_bus head; /* Dummy head node for simplifying some lock-free thread-safety stuff. */ + MA_ATOMIC(4, ma_uint32) nextCounter; /* This is used to determine whether or not the input bus is finding the next node in the list. Used for thread safety when detaching output buses. */ + MA_ATOMIC(4, ma_spinlock) lock; /* Unfortunate lock, but significantly simplifies the implementation. Required for thread-safe attaching and detaching. */ + + /* Set once at startup. */ + ma_uint8 channels; /* The number of channels in the audio stream for this bus. */ +}; + + +typedef struct ma_node_base ma_node_base; +struct ma_node_base +{ + /* These variables are set once at startup. */ + ma_node_graph* pNodeGraph; /* The graph this node belongs to. */ + const ma_node_vtable* vtable; + ma_uint32 inputBusCount; + ma_uint32 outputBusCount; + ma_node_input_bus* pInputBuses; + ma_node_output_bus* pOutputBuses; + float* pCachedData; /* Allocated on the heap. Fixed size. Needs to be stored on the heap because reading from output buses is done in separate function calls. */ + ma_uint16 cachedDataCapInFramesPerBus; /* The capacity of the input data cache in frames, per bus. */ + + /* These variables are read and written only from the audio thread. */ + ma_uint16 cachedFrameCountOut; + ma_uint16 cachedFrameCountIn; + ma_uint16 consumedFrameCountIn; + + /* These variables are read and written between different threads. */ + MA_ATOMIC(4, ma_node_state) state; /* When set to stopped, nothing will be read, regardless of the times in stateTimes. */ + MA_ATOMIC(8, ma_uint64) stateTimes[2]; /* Indexed by ma_node_state. Specifies the time based on the global clock that a node should be considered to be in the relevant state. */ + MA_ATOMIC(8, ma_uint64) localTime; /* The node's local clock. This is just a running sum of the number of output frames that have been processed. Can be modified by any thread with `ma_node_set_time()`. */ + + /* Memory management. */ + ma_node_input_bus _inputBuses[MA_MAX_NODE_LOCAL_BUS_COUNT]; + ma_node_output_bus _outputBuses[MA_MAX_NODE_LOCAL_BUS_COUNT]; + void* _pHeap; /* A heap allocation for internal use only. pInputBuses and/or pOutputBuses will point to this if the bus count exceeds MA_MAX_NODE_LOCAL_BUS_COUNT. */ + ma_bool32 _ownsHeap; /* If set to true, the node owns the heap allocation and _pHeap will be freed in ma_node_uninit(). */ +}; + +MA_API ma_result ma_node_get_heap_size(ma_node_graph* pNodeGraph, const ma_node_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_node_init_preallocated(ma_node_graph* pNodeGraph, const ma_node_config* pConfig, void* pHeap, ma_node* pNode); +MA_API ma_result ma_node_init(ma_node_graph* pNodeGraph, const ma_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_node* pNode); +MA_API void ma_node_uninit(ma_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_node_graph* ma_node_get_node_graph(const ma_node* pNode); +MA_API ma_uint32 ma_node_get_input_bus_count(const ma_node* pNode); +MA_API ma_uint32 ma_node_get_output_bus_count(const ma_node* pNode); +MA_API ma_uint32 ma_node_get_input_channels(const ma_node* pNode, ma_uint32 inputBusIndex); +MA_API ma_uint32 ma_node_get_output_channels(const ma_node* pNode, ma_uint32 outputBusIndex); +MA_API ma_result ma_node_attach_output_bus(ma_node* pNode, ma_uint32 outputBusIndex, ma_node* pOtherNode, ma_uint32 otherNodeInputBusIndex); +MA_API ma_result ma_node_detach_output_bus(ma_node* pNode, ma_uint32 outputBusIndex); +MA_API ma_result ma_node_detach_all_output_buses(ma_node* pNode); +MA_API ma_result ma_node_set_output_bus_volume(ma_node* pNode, ma_uint32 outputBusIndex, float volume); +MA_API float ma_node_get_output_bus_volume(const ma_node* pNode, ma_uint32 outputBusIndex); +MA_API ma_result ma_node_set_state(ma_node* pNode, ma_node_state state); +MA_API ma_node_state ma_node_get_state(const ma_node* pNode); +MA_API ma_result ma_node_set_state_time(ma_node* pNode, ma_node_state state, ma_uint64 globalTime); +MA_API ma_uint64 ma_node_get_state_time(const ma_node* pNode, ma_node_state state); +MA_API ma_node_state ma_node_get_state_by_time(const ma_node* pNode, ma_uint64 globalTime); +MA_API ma_node_state ma_node_get_state_by_time_range(const ma_node* pNode, ma_uint64 globalTimeBeg, ma_uint64 globalTimeEnd); +MA_API ma_uint64 ma_node_get_time(const ma_node* pNode); +MA_API ma_result ma_node_set_time(ma_node* pNode, ma_uint64 localTime); + + +typedef struct +{ + ma_uint32 channels; + ma_uint32 processingSizeInFrames; /* This is the preferred processing size for node processing callbacks unless overridden by a node itself. Can be 0 in which case it will be based on the frame count passed into ma_node_graph_read_pcm_frames(), but will not be well defined. */ + size_t preMixStackSizeInBytes; /* Defaults to 512KB per channel. Reducing this will save memory, but the depth of your node graph will be more restricted. */ +} ma_node_graph_config; + +MA_API ma_node_graph_config ma_node_graph_config_init(ma_uint32 channels); + + +struct ma_node_graph +{ + /* Immutable. */ + ma_node_base base; /* The node graph itself is a node so it can be connected as an input to different node graph. This has zero inputs and calls ma_node_graph_read_pcm_frames() to generate it's output. */ + ma_node_base endpoint; /* Special node that all nodes eventually connect to. Data is read from this node in ma_node_graph_read_pcm_frames(). */ + float* pProcessingCache; /* This will be allocated when processingSizeInFrames is non-zero. This is needed because ma_node_graph_read_pcm_frames() can be called with a variable number of frames, and we may need to do some buffering in situations where the caller requests a frame count that's not a multiple of processingSizeInFrames. */ + ma_uint32 processingCacheFramesRemaining; + ma_uint32 processingSizeInFrames; + + /* Read and written by multiple threads. */ + MA_ATOMIC(4, ma_bool32) isReading; + + /* Modified only by the audio thread. */ + ma_stack* pPreMixStack; +}; + +MA_API ma_result ma_node_graph_init(const ma_node_graph_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_node_graph* pNodeGraph); +MA_API void ma_node_graph_uninit(ma_node_graph* pNodeGraph, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_node* ma_node_graph_get_endpoint(ma_node_graph* pNodeGraph); +MA_API ma_result ma_node_graph_read_pcm_frames(ma_node_graph* pNodeGraph, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_uint32 ma_node_graph_get_channels(const ma_node_graph* pNodeGraph); +MA_API ma_uint64 ma_node_graph_get_time(const ma_node_graph* pNodeGraph); +MA_API ma_result ma_node_graph_set_time(ma_node_graph* pNodeGraph, ma_uint64 globalTime); + + + +/* Data source node. 0 input buses, 1 output bus. Used for reading from a data source. */ +typedef struct +{ + ma_node_config nodeConfig; + ma_data_source* pDataSource; +} ma_data_source_node_config; + +MA_API ma_data_source_node_config ma_data_source_node_config_init(ma_data_source* pDataSource); + + +typedef struct +{ + ma_node_base base; + ma_data_source* pDataSource; +} ma_data_source_node; + +MA_API ma_result ma_data_source_node_init(ma_node_graph* pNodeGraph, const ma_data_source_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source_node* pDataSourceNode); +MA_API void ma_data_source_node_uninit(ma_data_source_node* pDataSourceNode, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_data_source_node_set_looping(ma_data_source_node* pDataSourceNode, ma_bool32 isLooping); +MA_API ma_bool32 ma_data_source_node_is_looping(ma_data_source_node* pDataSourceNode); + + +/* Splitter Node. 1 input, many outputs. Used for splitting/copying a stream so it can be as input into two separate output nodes. */ +typedef struct +{ + ma_node_config nodeConfig; + ma_uint32 channels; + ma_uint32 outputBusCount; +} ma_splitter_node_config; + +MA_API ma_splitter_node_config ma_splitter_node_config_init(ma_uint32 channels); + + +typedef struct +{ + ma_node_base base; +} ma_splitter_node; + +MA_API ma_result ma_splitter_node_init(ma_node_graph* pNodeGraph, const ma_splitter_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_splitter_node* pSplitterNode); +MA_API void ma_splitter_node_uninit(ma_splitter_node* pSplitterNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +/* +Biquad Node +*/ +typedef struct +{ + ma_node_config nodeConfig; + ma_biquad_config biquad; +} ma_biquad_node_config; + +MA_API ma_biquad_node_config ma_biquad_node_config_init(ma_uint32 channels, float b0, float b1, float b2, float a0, float a1, float a2); + + +typedef struct +{ + ma_node_base baseNode; + ma_biquad biquad; +} ma_biquad_node; + +MA_API ma_result ma_biquad_node_init(ma_node_graph* pNodeGraph, const ma_biquad_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_biquad_node* pNode); +MA_API ma_result ma_biquad_node_reinit(const ma_biquad_config* pConfig, ma_biquad_node* pNode); +MA_API void ma_biquad_node_uninit(ma_biquad_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +/* +Low Pass Filter Node +*/ +typedef struct +{ + ma_node_config nodeConfig; + ma_lpf_config lpf; +} ma_lpf_node_config; + +MA_API ma_lpf_node_config ma_lpf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order); + + +typedef struct +{ + ma_node_base baseNode; + ma_lpf lpf; +} ma_lpf_node; + +MA_API ma_result ma_lpf_node_init(ma_node_graph* pNodeGraph, const ma_lpf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_lpf_node* pNode); +MA_API ma_result ma_lpf_node_reinit(const ma_lpf_config* pConfig, ma_lpf_node* pNode); +MA_API void ma_lpf_node_uninit(ma_lpf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +/* +High Pass Filter Node +*/ +typedef struct +{ + ma_node_config nodeConfig; + ma_hpf_config hpf; +} ma_hpf_node_config; + +MA_API ma_hpf_node_config ma_hpf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order); + + +typedef struct +{ + ma_node_base baseNode; + ma_hpf hpf; +} ma_hpf_node; + +MA_API ma_result ma_hpf_node_init(ma_node_graph* pNodeGraph, const ma_hpf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hpf_node* pNode); +MA_API ma_result ma_hpf_node_reinit(const ma_hpf_config* pConfig, ma_hpf_node* pNode); +MA_API void ma_hpf_node_uninit(ma_hpf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +/* +Band Pass Filter Node +*/ +typedef struct +{ + ma_node_config nodeConfig; + ma_bpf_config bpf; +} ma_bpf_node_config; + +MA_API ma_bpf_node_config ma_bpf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order); + + +typedef struct +{ + ma_node_base baseNode; + ma_bpf bpf; +} ma_bpf_node; + +MA_API ma_result ma_bpf_node_init(ma_node_graph* pNodeGraph, const ma_bpf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_bpf_node* pNode); +MA_API ma_result ma_bpf_node_reinit(const ma_bpf_config* pConfig, ma_bpf_node* pNode); +MA_API void ma_bpf_node_uninit(ma_bpf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +/* +Notching Filter Node +*/ +typedef struct +{ + ma_node_config nodeConfig; + ma_notch_config notch; +} ma_notch_node_config; + +MA_API ma_notch_node_config ma_notch_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double q, double frequency); + + +typedef struct +{ + ma_node_base baseNode; + ma_notch2 notch; +} ma_notch_node; + +MA_API ma_result ma_notch_node_init(ma_node_graph* pNodeGraph, const ma_notch_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_notch_node* pNode); +MA_API ma_result ma_notch_node_reinit(const ma_notch_config* pConfig, ma_notch_node* pNode); +MA_API void ma_notch_node_uninit(ma_notch_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +/* +Peaking Filter Node +*/ +typedef struct +{ + ma_node_config nodeConfig; + ma_peak_config peak; +} ma_peak_node_config; + +MA_API ma_peak_node_config ma_peak_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double q, double frequency); + + +typedef struct +{ + ma_node_base baseNode; + ma_peak2 peak; +} ma_peak_node; + +MA_API ma_result ma_peak_node_init(ma_node_graph* pNodeGraph, const ma_peak_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_peak_node* pNode); +MA_API ma_result ma_peak_node_reinit(const ma_peak_config* pConfig, ma_peak_node* pNode); +MA_API void ma_peak_node_uninit(ma_peak_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +/* +Low Shelf Filter Node +*/ +typedef struct +{ + ma_node_config nodeConfig; + ma_loshelf_config loshelf; +} ma_loshelf_node_config; + +MA_API ma_loshelf_node_config ma_loshelf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double q, double frequency); + + +typedef struct +{ + ma_node_base baseNode; + ma_loshelf2 loshelf; +} ma_loshelf_node; + +MA_API ma_result ma_loshelf_node_init(ma_node_graph* pNodeGraph, const ma_loshelf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_loshelf_node* pNode); +MA_API ma_result ma_loshelf_node_reinit(const ma_loshelf_config* pConfig, ma_loshelf_node* pNode); +MA_API void ma_loshelf_node_uninit(ma_loshelf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +/* +High Shelf Filter Node +*/ +typedef struct +{ + ma_node_config nodeConfig; + ma_hishelf_config hishelf; +} ma_hishelf_node_config; + +MA_API ma_hishelf_node_config ma_hishelf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double q, double frequency); + + +typedef struct +{ + ma_node_base baseNode; + ma_hishelf2 hishelf; +} ma_hishelf_node; + +MA_API ma_result ma_hishelf_node_init(ma_node_graph* pNodeGraph, const ma_hishelf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hishelf_node* pNode); +MA_API ma_result ma_hishelf_node_reinit(const ma_hishelf_config* pConfig, ma_hishelf_node* pNode); +MA_API void ma_hishelf_node_uninit(ma_hishelf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +typedef struct +{ + ma_node_config nodeConfig; + ma_delay_config delay; +} ma_delay_node_config; + +MA_API ma_delay_node_config ma_delay_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, ma_uint32 delayInFrames, float decay); + + +typedef struct +{ + ma_node_base baseNode; + ma_delay delay; +} ma_delay_node; + +MA_API ma_result ma_delay_node_init(ma_node_graph* pNodeGraph, const ma_delay_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_delay_node* pDelayNode); +MA_API void ma_delay_node_uninit(ma_delay_node* pDelayNode, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API void ma_delay_node_set_wet(ma_delay_node* pDelayNode, float value); +MA_API float ma_delay_node_get_wet(const ma_delay_node* pDelayNode); +MA_API void ma_delay_node_set_dry(ma_delay_node* pDelayNode, float value); +MA_API float ma_delay_node_get_dry(const ma_delay_node* pDelayNode); +MA_API void ma_delay_node_set_decay(ma_delay_node* pDelayNode, float value); +MA_API float ma_delay_node_get_decay(const ma_delay_node* pDelayNode); +#endif /* MA_NO_NODE_GRAPH */ + + +/* SECTION: miniaudio_engine.h */ +/************************************************************************************************************************************************************ + +Engine + +************************************************************************************************************************************************************/ +#if !defined(MA_NO_ENGINE) && !defined(MA_NO_NODE_GRAPH) +typedef struct ma_engine ma_engine; +typedef struct ma_sound ma_sound; + + +/* Sound flags. */ +typedef enum +{ + /* Resource manager flags. */ + MA_SOUND_FLAG_STREAM = 0x00000001, /* MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM */ + MA_SOUND_FLAG_DECODE = 0x00000002, /* MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_DECODE */ + MA_SOUND_FLAG_ASYNC = 0x00000004, /* MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC */ + MA_SOUND_FLAG_WAIT_INIT = 0x00000008, /* MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT */ + MA_SOUND_FLAG_UNKNOWN_LENGTH = 0x00000010, /* MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_UNKNOWN_LENGTH */ + MA_SOUND_FLAG_LOOPING = 0x00000020, /* MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING */ + + /* ma_sound specific flags. */ + MA_SOUND_FLAG_NO_DEFAULT_ATTACHMENT = 0x00001000, /* Do not attach to the endpoint by default. Useful for when setting up nodes in a complex graph system. */ + MA_SOUND_FLAG_NO_PITCH = 0x00002000, /* Disable pitch shifting with ma_sound_set_pitch() and ma_sound_group_set_pitch(). This is an optimization. */ + MA_SOUND_FLAG_NO_SPATIALIZATION = 0x00004000 /* Disable spatialization. */ +} ma_sound_flags; + +#ifndef MA_ENGINE_MAX_LISTENERS +#define MA_ENGINE_MAX_LISTENERS 4 +#endif + +#define MA_LISTENER_INDEX_CLOSEST ((ma_uint8)-1) + +typedef enum +{ + ma_engine_node_type_sound, + ma_engine_node_type_group +} ma_engine_node_type; + +typedef struct +{ + ma_engine* pEngine; + ma_engine_node_type type; + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_uint32 sampleRate; /* Only used when the type is set to ma_engine_node_type_sound. */ + ma_uint32 volumeSmoothTimeInPCMFrames; /* The number of frames to smooth over volume changes. Defaults to 0 in which case no smoothing is used. */ + ma_mono_expansion_mode monoExpansionMode; + ma_bool8 isPitchDisabled; /* Pitching can be explicitly disabled with MA_SOUND_FLAG_NO_PITCH to optimize processing. */ + ma_bool8 isSpatializationDisabled; /* Spatialization can be explicitly disabled with MA_SOUND_FLAG_NO_SPATIALIZATION. */ + ma_uint8 pinnedListenerIndex; /* The index of the listener this node should always use for spatialization. If set to MA_LISTENER_INDEX_CLOSEST the engine will use the closest listener. */ +} ma_engine_node_config; + +MA_API ma_engine_node_config ma_engine_node_config_init(ma_engine* pEngine, ma_engine_node_type type, ma_uint32 flags); + + +/* Base node object for both ma_sound and ma_sound_group. */ +typedef struct +{ + ma_node_base baseNode; /* Must be the first member for compatibility with the ma_node API. */ + ma_engine* pEngine; /* A pointer to the engine. Set based on the value from the config. */ + ma_uint32 sampleRate; /* The sample rate of the input data. For sounds backed by a data source, this will be the data source's sample rate. Otherwise it'll be the engine's sample rate. */ + ma_uint32 volumeSmoothTimeInPCMFrames; + ma_mono_expansion_mode monoExpansionMode; + ma_fader fader; + ma_linear_resampler resampler; /* For pitch shift. */ + ma_spatializer spatializer; + ma_panner panner; + ma_gainer volumeGainer; /* This will only be used if volumeSmoothTimeInPCMFrames is > 0. */ + ma_atomic_float volume; /* Defaults to 1. */ + MA_ATOMIC(4, float) pitch; + float oldPitch; /* For determining whether or not the resampler needs to be updated to reflect the new pitch. The resampler will be updated on the mixing thread. */ + float oldDopplerPitch; /* For determining whether or not the resampler needs to be updated to take a new doppler pitch into account. */ + MA_ATOMIC(4, ma_bool32) isPitchDisabled; /* When set to true, pitching will be disabled which will allow the resampler to be bypassed to save some computation. */ + MA_ATOMIC(4, ma_bool32) isSpatializationDisabled; /* Set to false by default. When set to false, will not have spatialisation applied. */ + MA_ATOMIC(4, ma_uint32) pinnedListenerIndex; /* The index of the listener this node should always use for spatialization. If set to MA_LISTENER_INDEX_CLOSEST the engine will use the closest listener. */ + + /* When setting a fade, it's not done immediately in ma_sound_set_fade(). It's deferred to the audio thread which means we need to store the settings here. */ + struct + { + ma_atomic_float volumeBeg; + ma_atomic_float volumeEnd; + ma_atomic_uint64 fadeLengthInFrames; /* <-- Defaults to (~(ma_uint64)0) which is used to indicate that no fade should be applied. */ + ma_atomic_uint64 absoluteGlobalTimeInFrames; /* <-- The time to start the fade. */ + } fadeSettings; + + /* Memory management. */ + ma_bool8 _ownsHeap; + void* _pHeap; +} ma_engine_node; + +MA_API ma_result ma_engine_node_get_heap_size(const ma_engine_node_config* pConfig, size_t* pHeapSizeInBytes); +MA_API ma_result ma_engine_node_init_preallocated(const ma_engine_node_config* pConfig, void* pHeap, ma_engine_node* pEngineNode); +MA_API ma_result ma_engine_node_init(const ma_engine_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_engine_node* pEngineNode); +MA_API void ma_engine_node_uninit(ma_engine_node* pEngineNode, const ma_allocation_callbacks* pAllocationCallbacks); + + +#define MA_SOUND_SOURCE_CHANNEL_COUNT 0xFFFFFFFF + +/* Callback for when a sound reaches the end. */ +typedef void (* ma_sound_end_proc)(void* pUserData, ma_sound* pSound); + +typedef struct +{ + const char* pFilePath; /* Set this to load from the resource manager. */ + const wchar_t* pFilePathW; /* Set this to load from the resource manager. */ + ma_data_source* pDataSource; /* Set this to load from an existing data source. */ + ma_node* pInitialAttachment; /* If set, the sound will be attached to an input of this node. This can be set to a ma_sound. If set to NULL, the sound will be attached directly to the endpoint unless MA_SOUND_FLAG_NO_DEFAULT_ATTACHMENT is set in `flags`. */ + ma_uint32 initialAttachmentInputBusIndex; /* The index of the input bus of pInitialAttachment to attach the sound to. */ + ma_uint32 channelsIn; /* Ignored if using a data source as input (the data source's channel count will be used always). Otherwise, setting to 0 will cause the engine's channel count to be used. */ + ma_uint32 channelsOut; /* Set this to 0 (default) to use the engine's channel count. Set to MA_SOUND_SOURCE_CHANNEL_COUNT to use the data source's channel count (only used if using a data source as input). */ + ma_mono_expansion_mode monoExpansionMode; /* Controls how the mono channel should be expanded to other channels when spatialization is disabled on a sound. */ + ma_uint32 flags; /* A combination of MA_SOUND_FLAG_* flags. */ + ma_uint32 volumeSmoothTimeInPCMFrames; /* The number of frames to smooth over volume changes. Defaults to 0 in which case no smoothing is used. */ + ma_uint64 initialSeekPointInPCMFrames; /* Initializes the sound such that it's seeked to this location by default. */ + ma_uint64 rangeBegInPCMFrames; + ma_uint64 rangeEndInPCMFrames; + ma_uint64 loopPointBegInPCMFrames; + ma_uint64 loopPointEndInPCMFrames; + ma_sound_end_proc endCallback; /* Fired when the sound reaches the end. Will be fired from the audio thread. Do not restart, uninitialize or otherwise change the state of the sound from here. Instead fire an event or set a variable to indicate to a different thread to change the start of the sound. Will not be fired in response to a scheduled stop with ma_sound_set_stop_time_*(). */ + void* pEndCallbackUserData; +#ifndef MA_NO_RESOURCE_MANAGER + ma_resource_manager_pipeline_notifications initNotifications; +#endif + ma_fence* pDoneFence; /* Deprecated. Use initNotifications instead. Released when the resource manager has finished decoding the entire sound. Not used with streams. */ + ma_bool32 isLooping; /* Deprecated. Use the MA_SOUND_FLAG_LOOPING flag in `flags` instead. */ +} ma_sound_config; + +MA_API ma_sound_config ma_sound_config_init(void); /* Deprecated. Will be removed in version 0.12. Use ma_sound_config_2() instead. */ +MA_API ma_sound_config ma_sound_config_init_2(ma_engine* pEngine); /* Will be renamed to ma_sound_config_init() in version 0.12. */ + +struct ma_sound +{ + ma_engine_node engineNode; /* Must be the first member for compatibility with the ma_node API. */ + ma_data_source* pDataSource; + MA_ATOMIC(8, ma_uint64) seekTarget; /* The PCM frame index to seek to in the mixing thread. Set to (~(ma_uint64)0) to not perform any seeking. */ + MA_ATOMIC(4, ma_bool32) atEnd; + ma_sound_end_proc endCallback; + void* pEndCallbackUserData; + ma_bool8 ownsDataSource; + + /* + We're declaring a resource manager data source object here to save us a malloc when loading a + sound via the resource manager, which I *think* will be the most common scenario. + */ +#ifndef MA_NO_RESOURCE_MANAGER + ma_resource_manager_data_source* pResourceManagerDataSource; +#endif +}; + +/* Structure specifically for sounds played with ma_engine_play_sound(). Making this a separate structure to reduce overhead. */ +typedef struct ma_sound_inlined ma_sound_inlined; +struct ma_sound_inlined +{ + ma_sound sound; + ma_sound_inlined* pNext; + ma_sound_inlined* pPrev; +}; + +/* A sound group is just a sound. */ +typedef ma_sound_config ma_sound_group_config; +typedef ma_sound ma_sound_group; + +MA_API ma_sound_group_config ma_sound_group_config_init(void); /* Deprecated. Will be removed in version 0.12. Use ma_sound_config_2() instead. */ +MA_API ma_sound_group_config ma_sound_group_config_init_2(ma_engine* pEngine); /* Will be renamed to ma_sound_config_init() in version 0.12. */ + +typedef void (* ma_engine_process_proc)(void* pUserData, float* pFramesOut, ma_uint64 frameCount); + +typedef struct +{ +#if !defined(MA_NO_RESOURCE_MANAGER) + ma_resource_manager* pResourceManager; /* Can be null in which case a resource manager will be created for you. */ +#endif +#if !defined(MA_NO_DEVICE_IO) + ma_context* pContext; + ma_device* pDevice; /* If set, the caller is responsible for calling ma_engine_data_callback() in the device's data callback. */ + ma_device_id* pPlaybackDeviceID; /* The ID of the playback device to use with the default listener. */ + ma_device_data_proc dataCallback; /* Can be null. Can be used to provide a custom device data callback. */ + ma_device_notification_proc notificationCallback; +#endif + ma_log* pLog; /* When set to NULL, will use the context's log. */ + ma_uint32 listenerCount; /* Must be between 1 and MA_ENGINE_MAX_LISTENERS. */ + ma_uint32 channels; /* The number of channels to use when mixing and spatializing. When set to 0, will use the native channel count of the device. */ + ma_uint32 sampleRate; /* The sample rate. When set to 0 will use the native channel count of the device. */ + ma_uint32 periodSizeInFrames; /* If set to something other than 0, updates will always be exactly this size. The underlying device may be a different size, but from the perspective of the mixer that won't matter.*/ + ma_uint32 periodSizeInMilliseconds; /* Used if periodSizeInFrames is unset. */ + ma_uint32 gainSmoothTimeInFrames; /* The number of frames to interpolate the gain of spatialized sounds across. If set to 0, will use gainSmoothTimeInMilliseconds. */ + ma_uint32 gainSmoothTimeInMilliseconds; /* When set to 0, gainSmoothTimeInFrames will be used. If both are set to 0, a default value will be used. */ + ma_uint32 defaultVolumeSmoothTimeInPCMFrames; /* Defaults to 0. Controls the default amount of smoothing to apply to volume changes to sounds. High values means more smoothing at the expense of high latency (will take longer to reach the new volume). */ + ma_uint32 preMixStackSizeInBytes; /* A stack is used for internal processing in the node graph. This allows you to configure the size of this stack. Smaller values will reduce the maximum depth of your node graph. You should rarely need to modify this. */ + ma_allocation_callbacks allocationCallbacks; + ma_bool32 noAutoStart; /* When set to true, requires an explicit call to ma_engine_start(). This is false by default, meaning the engine will be started automatically in ma_engine_init(). */ + ma_bool32 noDevice; /* When set to true, don't create a default device. ma_engine_read_pcm_frames() can be called manually to read data. */ + ma_mono_expansion_mode monoExpansionMode; /* Controls how the mono channel should be expanded to other channels when spatialization is disabled on a sound. */ + ma_vfs* pResourceManagerVFS; /* A pointer to a pre-allocated VFS object to use with the resource manager. This is ignored if pResourceManager is not NULL. */ + ma_engine_process_proc onProcess; /* Fired at the end of each call to ma_engine_read_pcm_frames(). For engine's that manage their own internal device (the default configuration), this will be fired from the audio thread, and you do not need to call ma_engine_read_pcm_frames() manually in order to trigger this. */ + void* pProcessUserData; /* User data that's passed into onProcess. */ +} ma_engine_config; + +MA_API ma_engine_config ma_engine_config_init(void); + + +struct ma_engine +{ + ma_node_graph nodeGraph; /* An engine is a node graph. It should be able to be plugged into any ma_node_graph API (with a cast) which means this must be the first member of this struct. */ +#if !defined(MA_NO_RESOURCE_MANAGER) + ma_resource_manager* pResourceManager; +#endif +#if !defined(MA_NO_DEVICE_IO) + ma_device* pDevice; /* Optionally set via the config, otherwise allocated by the engine in ma_engine_init(). */ +#endif + ma_log* pLog; + ma_uint32 sampleRate; + ma_uint32 listenerCount; + ma_spatializer_listener listeners[MA_ENGINE_MAX_LISTENERS]; + ma_allocation_callbacks allocationCallbacks; + ma_bool8 ownsResourceManager; + ma_bool8 ownsDevice; + ma_spinlock inlinedSoundLock; /* For synchronizing access to the inlined sound list. */ + ma_sound_inlined* pInlinedSoundHead; /* The first inlined sound. Inlined sounds are tracked in a linked list. */ + MA_ATOMIC(4, ma_uint32) inlinedSoundCount; /* The total number of allocated inlined sound objects. Used for debugging. */ + ma_uint32 gainSmoothTimeInFrames; /* The number of frames to interpolate the gain of spatialized sounds across. */ + ma_uint32 defaultVolumeSmoothTimeInPCMFrames; + ma_mono_expansion_mode monoExpansionMode; + ma_engine_process_proc onProcess; + void* pProcessUserData; +}; + +MA_API ma_result ma_engine_init(const ma_engine_config* pConfig, ma_engine* pEngine); +MA_API void ma_engine_uninit(ma_engine* pEngine); +MA_API ma_result ma_engine_read_pcm_frames(ma_engine* pEngine, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_node_graph* ma_engine_get_node_graph(ma_engine* pEngine); +#if !defined(MA_NO_RESOURCE_MANAGER) +MA_API ma_resource_manager* ma_engine_get_resource_manager(ma_engine* pEngine); +#endif +MA_API ma_device* ma_engine_get_device(ma_engine* pEngine); +MA_API ma_log* ma_engine_get_log(ma_engine* pEngine); +MA_API ma_node* ma_engine_get_endpoint(ma_engine* pEngine); +MA_API ma_uint64 ma_engine_get_time_in_pcm_frames(const ma_engine* pEngine); +MA_API ma_uint64 ma_engine_get_time_in_milliseconds(const ma_engine* pEngine); +MA_API ma_result ma_engine_set_time_in_pcm_frames(ma_engine* pEngine, ma_uint64 globalTime); +MA_API ma_result ma_engine_set_time_in_milliseconds(ma_engine* pEngine, ma_uint64 globalTime); +MA_API ma_uint64 ma_engine_get_time(const ma_engine* pEngine); /* Deprecated. Use ma_engine_get_time_in_pcm_frames(). Will be removed in version 0.12. */ +MA_API ma_result ma_engine_set_time(ma_engine* pEngine, ma_uint64 globalTime); /* Deprecated. Use ma_engine_set_time_in_pcm_frames(). Will be removed in version 0.12. */ +MA_API ma_uint32 ma_engine_get_channels(const ma_engine* pEngine); +MA_API ma_uint32 ma_engine_get_sample_rate(const ma_engine* pEngine); + +MA_API ma_result ma_engine_start(ma_engine* pEngine); +MA_API ma_result ma_engine_stop(ma_engine* pEngine); +MA_API ma_result ma_engine_set_volume(ma_engine* pEngine, float volume); +MA_API float ma_engine_get_volume(ma_engine* pEngine); +MA_API ma_result ma_engine_set_gain_db(ma_engine* pEngine, float gainDB); +MA_API float ma_engine_get_gain_db(ma_engine* pEngine); + +MA_API ma_uint32 ma_engine_get_listener_count(const ma_engine* pEngine); +MA_API ma_uint32 ma_engine_find_closest_listener(const ma_engine* pEngine, float absolutePosX, float absolutePosY, float absolutePosZ); +MA_API void ma_engine_listener_set_position(ma_engine* pEngine, ma_uint32 listenerIndex, float x, float y, float z); +MA_API ma_vec3f ma_engine_listener_get_position(const ma_engine* pEngine, ma_uint32 listenerIndex); +MA_API void ma_engine_listener_set_direction(ma_engine* pEngine, ma_uint32 listenerIndex, float x, float y, float z); +MA_API ma_vec3f ma_engine_listener_get_direction(const ma_engine* pEngine, ma_uint32 listenerIndex); +MA_API void ma_engine_listener_set_velocity(ma_engine* pEngine, ma_uint32 listenerIndex, float x, float y, float z); +MA_API ma_vec3f ma_engine_listener_get_velocity(const ma_engine* pEngine, ma_uint32 listenerIndex); +MA_API void ma_engine_listener_set_cone(ma_engine* pEngine, ma_uint32 listenerIndex, float innerAngleInRadians, float outerAngleInRadians, float outerGain); +MA_API void ma_engine_listener_get_cone(const ma_engine* pEngine, ma_uint32 listenerIndex, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain); +MA_API void ma_engine_listener_set_world_up(ma_engine* pEngine, ma_uint32 listenerIndex, float x, float y, float z); +MA_API ma_vec3f ma_engine_listener_get_world_up(const ma_engine* pEngine, ma_uint32 listenerIndex); +MA_API void ma_engine_listener_set_enabled(ma_engine* pEngine, ma_uint32 listenerIndex, ma_bool32 isEnabled); +MA_API ma_bool32 ma_engine_listener_is_enabled(const ma_engine* pEngine, ma_uint32 listenerIndex); + +#ifndef MA_NO_RESOURCE_MANAGER +MA_API ma_result ma_engine_play_sound_ex(ma_engine* pEngine, const char* pFilePath, ma_node* pNode, ma_uint32 nodeInputBusIndex); +MA_API ma_result ma_engine_play_sound(ma_engine* pEngine, const char* pFilePath, ma_sound_group* pGroup); /* Fire and forget. */ +#endif + +#ifndef MA_NO_RESOURCE_MANAGER +MA_API ma_result ma_sound_init_from_file(ma_engine* pEngine, const char* pFilePath, ma_uint32 flags, ma_sound_group* pGroup, ma_fence* pDoneFence, ma_sound* pSound); +MA_API ma_result ma_sound_init_from_file_w(ma_engine* pEngine, const wchar_t* pFilePath, ma_uint32 flags, ma_sound_group* pGroup, ma_fence* pDoneFence, ma_sound* pSound); +MA_API ma_result ma_sound_init_copy(ma_engine* pEngine, const ma_sound* pExistingSound, ma_uint32 flags, ma_sound_group* pGroup, ma_sound* pSound); +#endif +MA_API ma_result ma_sound_init_from_data_source(ma_engine* pEngine, ma_data_source* pDataSource, ma_uint32 flags, ma_sound_group* pGroup, ma_sound* pSound); +MA_API ma_result ma_sound_init_ex(ma_engine* pEngine, const ma_sound_config* pConfig, ma_sound* pSound); +MA_API void ma_sound_uninit(ma_sound* pSound); +MA_API ma_engine* ma_sound_get_engine(const ma_sound* pSound); +MA_API ma_data_source* ma_sound_get_data_source(const ma_sound* pSound); +MA_API ma_result ma_sound_start(ma_sound* pSound); +MA_API ma_result ma_sound_stop(ma_sound* pSound); +MA_API ma_result ma_sound_stop_with_fade_in_pcm_frames(ma_sound* pSound, ma_uint64 fadeLengthInFrames); /* Will overwrite any scheduled stop and fade. */ +MA_API ma_result ma_sound_stop_with_fade_in_milliseconds(ma_sound* pSound, ma_uint64 fadeLengthInFrames); /* Will overwrite any scheduled stop and fade. */ +MA_API void ma_sound_set_volume(ma_sound* pSound, float volume); +MA_API float ma_sound_get_volume(const ma_sound* pSound); +MA_API void ma_sound_set_pan(ma_sound* pSound, float pan); +MA_API float ma_sound_get_pan(const ma_sound* pSound); +MA_API void ma_sound_set_pan_mode(ma_sound* pSound, ma_pan_mode panMode); +MA_API ma_pan_mode ma_sound_get_pan_mode(const ma_sound* pSound); +MA_API void ma_sound_set_pitch(ma_sound* pSound, float pitch); +MA_API float ma_sound_get_pitch(const ma_sound* pSound); +MA_API void ma_sound_set_spatialization_enabled(ma_sound* pSound, ma_bool32 enabled); +MA_API ma_bool32 ma_sound_is_spatialization_enabled(const ma_sound* pSound); +MA_API void ma_sound_set_pinned_listener_index(ma_sound* pSound, ma_uint32 listenerIndex); +MA_API ma_uint32 ma_sound_get_pinned_listener_index(const ma_sound* pSound); +MA_API ma_uint32 ma_sound_get_listener_index(const ma_sound* pSound); +MA_API ma_vec3f ma_sound_get_direction_to_listener(const ma_sound* pSound); +MA_API void ma_sound_set_position(ma_sound* pSound, float x, float y, float z); +MA_API ma_vec3f ma_sound_get_position(const ma_sound* pSound); +MA_API void ma_sound_set_direction(ma_sound* pSound, float x, float y, float z); +MA_API ma_vec3f ma_sound_get_direction(const ma_sound* pSound); +MA_API void ma_sound_set_velocity(ma_sound* pSound, float x, float y, float z); +MA_API ma_vec3f ma_sound_get_velocity(const ma_sound* pSound); +MA_API void ma_sound_set_attenuation_model(ma_sound* pSound, ma_attenuation_model attenuationModel); +MA_API ma_attenuation_model ma_sound_get_attenuation_model(const ma_sound* pSound); +MA_API void ma_sound_set_positioning(ma_sound* pSound, ma_positioning positioning); +MA_API ma_positioning ma_sound_get_positioning(const ma_sound* pSound); +MA_API void ma_sound_set_rolloff(ma_sound* pSound, float rolloff); +MA_API float ma_sound_get_rolloff(const ma_sound* pSound); +MA_API void ma_sound_set_min_gain(ma_sound* pSound, float minGain); +MA_API float ma_sound_get_min_gain(const ma_sound* pSound); +MA_API void ma_sound_set_max_gain(ma_sound* pSound, float maxGain); +MA_API float ma_sound_get_max_gain(const ma_sound* pSound); +MA_API void ma_sound_set_min_distance(ma_sound* pSound, float minDistance); +MA_API float ma_sound_get_min_distance(const ma_sound* pSound); +MA_API void ma_sound_set_max_distance(ma_sound* pSound, float maxDistance); +MA_API float ma_sound_get_max_distance(const ma_sound* pSound); +MA_API void ma_sound_set_cone(ma_sound* pSound, float innerAngleInRadians, float outerAngleInRadians, float outerGain); +MA_API void ma_sound_get_cone(const ma_sound* pSound, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain); +MA_API void ma_sound_set_doppler_factor(ma_sound* pSound, float dopplerFactor); +MA_API float ma_sound_get_doppler_factor(const ma_sound* pSound); +MA_API void ma_sound_set_directional_attenuation_factor(ma_sound* pSound, float directionalAttenuationFactor); +MA_API float ma_sound_get_directional_attenuation_factor(const ma_sound* pSound); +MA_API void ma_sound_set_fade_in_pcm_frames(ma_sound* pSound, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInFrames); +MA_API void ma_sound_set_fade_in_milliseconds(ma_sound* pSound, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInMilliseconds); +MA_API void ma_sound_set_fade_start_in_pcm_frames(ma_sound* pSound, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInFrames, ma_uint64 absoluteGlobalTimeInFrames); +MA_API void ma_sound_set_fade_start_in_milliseconds(ma_sound* pSound, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInMilliseconds, ma_uint64 absoluteGlobalTimeInMilliseconds); +MA_API float ma_sound_get_current_fade_volume(const ma_sound* pSound); +MA_API void ma_sound_set_start_time_in_pcm_frames(ma_sound* pSound, ma_uint64 absoluteGlobalTimeInFrames); +MA_API void ma_sound_set_start_time_in_milliseconds(ma_sound* pSound, ma_uint64 absoluteGlobalTimeInMilliseconds); +MA_API void ma_sound_set_stop_time_in_pcm_frames(ma_sound* pSound, ma_uint64 absoluteGlobalTimeInFrames); +MA_API void ma_sound_set_stop_time_in_milliseconds(ma_sound* pSound, ma_uint64 absoluteGlobalTimeInMilliseconds); +MA_API void ma_sound_set_stop_time_with_fade_in_pcm_frames(ma_sound* pSound, ma_uint64 stopAbsoluteGlobalTimeInFrames, ma_uint64 fadeLengthInFrames); +MA_API void ma_sound_set_stop_time_with_fade_in_milliseconds(ma_sound* pSound, ma_uint64 stopAbsoluteGlobalTimeInMilliseconds, ma_uint64 fadeLengthInMilliseconds); +MA_API ma_bool32 ma_sound_is_playing(const ma_sound* pSound); +MA_API ma_uint64 ma_sound_get_time_in_pcm_frames(const ma_sound* pSound); +MA_API ma_uint64 ma_sound_get_time_in_milliseconds(const ma_sound* pSound); +MA_API void ma_sound_set_looping(ma_sound* pSound, ma_bool32 isLooping); +MA_API ma_bool32 ma_sound_is_looping(const ma_sound* pSound); +MA_API ma_bool32 ma_sound_at_end(const ma_sound* pSound); +MA_API ma_result ma_sound_seek_to_pcm_frame(ma_sound* pSound, ma_uint64 frameIndex); /* Just a wrapper around ma_data_source_seek_to_pcm_frame(). */ +MA_API ma_result ma_sound_seek_to_second(ma_sound* pSound, float seekPointInSeconds); /* Abstraction to ma_sound_seek_to_pcm_frame() */ +MA_API ma_result ma_sound_get_data_format(ma_sound* pSound, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_sound_get_cursor_in_pcm_frames(ma_sound* pSound, ma_uint64* pCursor); +MA_API ma_result ma_sound_get_length_in_pcm_frames(ma_sound* pSound, ma_uint64* pLength); +MA_API ma_result ma_sound_get_cursor_in_seconds(ma_sound* pSound, float* pCursor); +MA_API ma_result ma_sound_get_length_in_seconds(ma_sound* pSound, float* pLength); +MA_API ma_result ma_sound_set_end_callback(ma_sound* pSound, ma_sound_end_proc callback, void* pUserData); + +MA_API ma_result ma_sound_group_init(ma_engine* pEngine, ma_uint32 flags, ma_sound_group* pParentGroup, ma_sound_group* pGroup); +MA_API ma_result ma_sound_group_init_ex(ma_engine* pEngine, const ma_sound_group_config* pConfig, ma_sound_group* pGroup); +MA_API void ma_sound_group_uninit(ma_sound_group* pGroup); +MA_API ma_engine* ma_sound_group_get_engine(const ma_sound_group* pGroup); +MA_API ma_result ma_sound_group_start(ma_sound_group* pGroup); +MA_API ma_result ma_sound_group_stop(ma_sound_group* pGroup); +MA_API void ma_sound_group_set_volume(ma_sound_group* pGroup, float volume); +MA_API float ma_sound_group_get_volume(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_pan(ma_sound_group* pGroup, float pan); +MA_API float ma_sound_group_get_pan(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_pan_mode(ma_sound_group* pGroup, ma_pan_mode panMode); +MA_API ma_pan_mode ma_sound_group_get_pan_mode(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_pitch(ma_sound_group* pGroup, float pitch); +MA_API float ma_sound_group_get_pitch(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_spatialization_enabled(ma_sound_group* pGroup, ma_bool32 enabled); +MA_API ma_bool32 ma_sound_group_is_spatialization_enabled(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_pinned_listener_index(ma_sound_group* pGroup, ma_uint32 listenerIndex); +MA_API ma_uint32 ma_sound_group_get_pinned_listener_index(const ma_sound_group* pGroup); +MA_API ma_uint32 ma_sound_group_get_listener_index(const ma_sound_group* pGroup); +MA_API ma_vec3f ma_sound_group_get_direction_to_listener(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_position(ma_sound_group* pGroup, float x, float y, float z); +MA_API ma_vec3f ma_sound_group_get_position(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_direction(ma_sound_group* pGroup, float x, float y, float z); +MA_API ma_vec3f ma_sound_group_get_direction(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_velocity(ma_sound_group* pGroup, float x, float y, float z); +MA_API ma_vec3f ma_sound_group_get_velocity(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_attenuation_model(ma_sound_group* pGroup, ma_attenuation_model attenuationModel); +MA_API ma_attenuation_model ma_sound_group_get_attenuation_model(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_positioning(ma_sound_group* pGroup, ma_positioning positioning); +MA_API ma_positioning ma_sound_group_get_positioning(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_rolloff(ma_sound_group* pGroup, float rolloff); +MA_API float ma_sound_group_get_rolloff(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_min_gain(ma_sound_group* pGroup, float minGain); +MA_API float ma_sound_group_get_min_gain(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_max_gain(ma_sound_group* pGroup, float maxGain); +MA_API float ma_sound_group_get_max_gain(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_min_distance(ma_sound_group* pGroup, float minDistance); +MA_API float ma_sound_group_get_min_distance(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_max_distance(ma_sound_group* pGroup, float maxDistance); +MA_API float ma_sound_group_get_max_distance(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_cone(ma_sound_group* pGroup, float innerAngleInRadians, float outerAngleInRadians, float outerGain); +MA_API void ma_sound_group_get_cone(const ma_sound_group* pGroup, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain); +MA_API void ma_sound_group_set_doppler_factor(ma_sound_group* pGroup, float dopplerFactor); +MA_API float ma_sound_group_get_doppler_factor(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_directional_attenuation_factor(ma_sound_group* pGroup, float directionalAttenuationFactor); +MA_API float ma_sound_group_get_directional_attenuation_factor(const ma_sound_group* pGroup); +MA_API void ma_sound_group_set_fade_in_pcm_frames(ma_sound_group* pGroup, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInFrames); +MA_API void ma_sound_group_set_fade_in_milliseconds(ma_sound_group* pGroup, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInMilliseconds); +MA_API float ma_sound_group_get_current_fade_volume(ma_sound_group* pGroup); +MA_API void ma_sound_group_set_start_time_in_pcm_frames(ma_sound_group* pGroup, ma_uint64 absoluteGlobalTimeInFrames); +MA_API void ma_sound_group_set_start_time_in_milliseconds(ma_sound_group* pGroup, ma_uint64 absoluteGlobalTimeInMilliseconds); +MA_API void ma_sound_group_set_stop_time_in_pcm_frames(ma_sound_group* pGroup, ma_uint64 absoluteGlobalTimeInFrames); +MA_API void ma_sound_group_set_stop_time_in_milliseconds(ma_sound_group* pGroup, ma_uint64 absoluteGlobalTimeInMilliseconds); +MA_API ma_bool32 ma_sound_group_is_playing(const ma_sound_group* pGroup); +MA_API ma_uint64 ma_sound_group_get_time_in_pcm_frames(const ma_sound_group* pGroup); +#endif /* MA_NO_ENGINE */ +/* END SECTION: miniaudio_engine.h */ + +#ifdef __cplusplus +} +#endif +#endif /* miniaudio_h */ + + +/* +This is for preventing greying out of the implementation section. +*/ +#if defined(Q_CREATOR_RUN) || defined(__INTELLISENSE__) || defined(__CDT_PARSER__) +#define MINIAUDIO_IMPLEMENTATION +#endif + +/************************************************************************************************************************************************************ +************************************************************************************************************************************************************* + +IMPLEMENTATION + +************************************************************************************************************************************************************* +************************************************************************************************************************************************************/ +#if defined(MINIAUDIO_IMPLEMENTATION) || defined(MA_IMPLEMENTATION) +#ifndef miniaudio_c +#define miniaudio_c + +#include +#include /* For INT_MAX */ +#include /* sin(), etc. */ +#include /* For malloc(), free(), wcstombs(). */ +#include /* For memset() */ + +#include +#include +#if !defined(_MSC_VER) && !defined(__DMC__) + #include /* For strcasecmp(). */ + #include /* For wcslen(), wcsrtombs() */ +#endif +#ifdef _MSC_VER + #include /* For _controlfp_s constants */ +#endif + +#if defined(MA_WIN32) + #include + + /* + There's a possibility that WIN32_LEAN_AND_MEAN has been defined which will exclude some symbols + such as STGM_READ and CLSCTL_ALL. We need to check these and define them ourselves if they're + unavailable. + */ + #ifndef STGM_READ + #define STGM_READ 0x00000000L + #endif + #ifndef CLSCTX_ALL + #define CLSCTX_ALL 23 + #endif + + /* IUnknown is used by both the WASAPI and DirectSound backends. It easier to just declare our version here. */ + typedef struct ma_IUnknown ma_IUnknown; +#endif + +#if !defined(MA_WIN32) +#include +#include /* select() (used for ma_sleep()). */ +#include +#endif + +#ifdef MA_NX +#include /* For nanosleep() */ +#endif + +#include /* For fstat(), etc. */ + +#ifdef MA_EMSCRIPTEN +#include +#endif + + +/* Architecture Detection */ +#if !defined(MA_64BIT) && !defined(MA_32BIT) +#ifdef _WIN32 +#ifdef _WIN64 +#define MA_64BIT +#else +#define MA_32BIT +#endif +#endif +#endif + +#if !defined(MA_64BIT) && !defined(MA_32BIT) +#ifdef __GNUC__ +#ifdef __LP64__ +#define MA_64BIT +#else +#define MA_32BIT +#endif +#endif +#endif + +#if !defined(MA_64BIT) && !defined(MA_32BIT) +#include +#if INTPTR_MAX == INT64_MAX +#define MA_64BIT +#else +#define MA_32BIT +#endif +#endif + +#if defined(__arm__) || defined(_M_ARM) +#define MA_ARM32 +#endif +#if defined(__arm64) || defined(__arm64__) || defined(__aarch64__) || defined(_M_ARM64) +#define MA_ARM64 +#endif + +#if defined(__x86_64__) || defined(_M_X64) +#define MA_X64 +#elif defined(__i386) || defined(_M_IX86) +#define MA_X86 +#elif defined(MA_ARM32) || defined(MA_ARM64) +#define MA_ARM +#endif + +/* Intrinsics Support */ +#if (defined(MA_X64) || defined(MA_X86)) && !defined(__COSMOPOLITAN__) + #if defined(_MSC_VER) && !defined(__clang__) + /* MSVC. */ + #if _MSC_VER >= 1400 && !defined(MA_NO_SSE2) /* 2005 */ + #define MA_SUPPORT_SSE2 + #endif + /*#if _MSC_VER >= 1600 && !defined(MA_NO_AVX)*/ /* 2010 */ + /* #define MA_SUPPORT_AVX*/ + /*#endif*/ + #if _MSC_VER >= 1700 && !defined(MA_NO_AVX2) /* 2012 */ + #define MA_SUPPORT_AVX2 + #endif + #else + /* Assume GNUC-style. */ + #if defined(__SSE2__) && !defined(MA_NO_SSE2) + #define MA_SUPPORT_SSE2 + #endif + /*#if defined(__AVX__) && !defined(MA_NO_AVX)*/ + /* #define MA_SUPPORT_AVX*/ + /*#endif*/ + #if defined(__AVX2__) && !defined(MA_NO_AVX2) + #define MA_SUPPORT_AVX2 + #endif + #endif + + /* If at this point we still haven't determined compiler support for the intrinsics just fall back to __has_include. */ + #if !defined(__GNUC__) && !defined(__clang__) && defined(__has_include) + #if !defined(MA_SUPPORT_SSE2) && !defined(MA_NO_SSE2) && __has_include() + #define MA_SUPPORT_SSE2 + #endif + /*#if !defined(MA_SUPPORT_AVX) && !defined(MA_NO_AVX) && __has_include()*/ + /* #define MA_SUPPORT_AVX*/ + /*#endif*/ + #if !defined(MA_SUPPORT_AVX2) && !defined(MA_NO_AVX2) && __has_include() + #define MA_SUPPORT_AVX2 + #endif + #endif + + #if defined(MA_SUPPORT_AVX2) || defined(MA_SUPPORT_AVX) + #include + #elif defined(MA_SUPPORT_SSE2) + #include + #endif +#endif + +#if defined(MA_ARM) + #if !defined(MA_NO_NEON) && (defined(__ARM_NEON) || defined(__aarch64__) || defined(_M_ARM64)) + #define MA_SUPPORT_NEON + #include + #endif +#endif + +/* Begin globally disabled warnings. */ +#if defined(_MSC_VER) + #pragma warning(push) + #pragma warning(disable:4752) /* found Intel(R) Advanced Vector Extensions; consider using /arch:AVX */ + #pragma warning(disable:4049) /* compiler limit : terminating line number emission */ +#endif + +#if defined(MA_X64) || defined(MA_X86) + #if defined(_MSC_VER) && !defined(__clang__) + #if _MSC_VER >= 1400 + #include + static MA_INLINE void ma_cpuid(int info[4], int fid) + { + __cpuid(info, fid); + } + #else + #define MA_NO_CPUID + #endif + + #if _MSC_VER >= 1600 && (defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 160040219) + static MA_INLINE unsigned __int64 ma_xgetbv(int reg) + { + return _xgetbv(reg); + } + #else + #define MA_NO_XGETBV + #endif + #elif (defined(__GNUC__) || defined(__clang__)) && !defined(MA_ANDROID) + static MA_INLINE void ma_cpuid(int info[4], int fid) + { + /* + It looks like the -fPIC option uses the ebx register which GCC complains about. We can work around this by just using a different register, the + specific register of which I'm letting the compiler decide on. The "k" prefix is used to specify a 32-bit register. The {...} syntax is for + supporting different assembly dialects. + + What's basically happening is that we're saving and restoring the ebx register manually. + */ + #if defined(MA_X86) && defined(__PIC__) + __asm__ __volatile__ ( + "xchg{l} {%%}ebx, %k1;" + "cpuid;" + "xchg{l} {%%}ebx, %k1;" + : "=a"(info[0]), "=&r"(info[1]), "=c"(info[2]), "=d"(info[3]) : "a"(fid), "c"(0) + ); + #else + __asm__ __volatile__ ( + "cpuid" : "=a"(info[0]), "=b"(info[1]), "=c"(info[2]), "=d"(info[3]) : "a"(fid), "c"(0) + ); + #endif + } + + static MA_INLINE ma_uint64 ma_xgetbv(int reg) + { + unsigned int hi; + unsigned int lo; + + __asm__ __volatile__ ( + "xgetbv" : "=a"(lo), "=d"(hi) : "c"(reg) + ); + + return ((ma_uint64)hi << 32) | (ma_uint64)lo; + } + #else + #define MA_NO_CPUID + #define MA_NO_XGETBV + #endif +#else + #define MA_NO_CPUID + #define MA_NO_XGETBV +#endif + +static MA_INLINE ma_bool32 ma_has_sse2(void) +{ +#if defined(MA_SUPPORT_SSE2) + #if (defined(MA_X64) || defined(MA_X86)) && !defined(MA_NO_SSE2) + #if defined(MA_X64) + return MA_TRUE; /* 64-bit targets always support SSE2. */ + #elif (defined(_M_IX86_FP) && _M_IX86_FP == 2) || defined(__SSE2__) + return MA_TRUE; /* If the compiler is allowed to freely generate SSE2 code we can assume support. */ + #else + #if defined(MA_NO_CPUID) + return MA_FALSE; + #else + int info[4]; + ma_cpuid(info, 1); + return (info[3] & (1 << 26)) != 0; + #endif + #endif + #else + return MA_FALSE; /* SSE2 is only supported on x86 and x64 architectures. */ + #endif +#else + return MA_FALSE; /* No compiler support. */ +#endif +} + +#if 0 +static MA_INLINE ma_bool32 ma_has_avx() +{ +#if defined(MA_SUPPORT_AVX) + #if (defined(MA_X64) || defined(MA_X86)) && !defined(MA_NO_AVX) + #if defined(_AVX_) || defined(__AVX__) + return MA_TRUE; /* If the compiler is allowed to freely generate AVX code we can assume support. */ + #else + /* AVX requires both CPU and OS support. */ + #if defined(MA_NO_CPUID) || defined(MA_NO_XGETBV) + return MA_FALSE; + #else + int info[4]; + ma_cpuid(info, 1); + if (((info[2] & (1 << 27)) != 0) && ((info[2] & (1 << 28)) != 0)) { + ma_uint64 xrc = ma_xgetbv(0); + if ((xrc & 0x06) == 0x06) { + return MA_TRUE; + } else { + return MA_FALSE; + } + } else { + return MA_FALSE; + } + #endif + #endif + #else + return MA_FALSE; /* AVX is only supported on x86 and x64 architectures. */ + #endif +#else + return MA_FALSE; /* No compiler support. */ +#endif +} +#endif + +static MA_INLINE ma_bool32 ma_has_avx2(void) +{ +#if defined(MA_SUPPORT_AVX2) + #if (defined(MA_X64) || defined(MA_X86)) && !defined(MA_NO_AVX2) + #if defined(_AVX2_) || defined(__AVX2__) + return MA_TRUE; /* If the compiler is allowed to freely generate AVX2 code we can assume support. */ + #else + /* AVX2 requires both CPU and OS support. */ + #if defined(MA_NO_CPUID) || defined(MA_NO_XGETBV) + return MA_FALSE; + #else + int info1[4]; + int info7[4]; + ma_cpuid(info1, 1); + ma_cpuid(info7, 7); + if (((info1[2] & (1 << 27)) != 0) && ((info7[1] & (1 << 5)) != 0)) { + ma_uint64 xrc = ma_xgetbv(0); + if ((xrc & 0x06) == 0x06) { + return MA_TRUE; + } else { + return MA_FALSE; + } + } else { + return MA_FALSE; + } + #endif + #endif + #else + return MA_FALSE; /* AVX2 is only supported on x86 and x64 architectures. */ + #endif +#else + return MA_FALSE; /* No compiler support. */ +#endif +} + +static MA_INLINE ma_bool32 ma_has_neon(void) +{ +#if defined(MA_SUPPORT_NEON) + #if defined(MA_ARM) && !defined(MA_NO_NEON) + #if (defined(__ARM_NEON) || defined(__aarch64__) || defined(_M_ARM64)) + return MA_TRUE; /* If the compiler is allowed to freely generate NEON code we can assume support. */ + #else + /* TODO: Runtime check. */ + return MA_FALSE; + #endif + #else + return MA_FALSE; /* NEON is only supported on ARM architectures. */ + #endif +#else + return MA_FALSE; /* No compiler support. */ +#endif +} + +#if defined(__has_builtin) + #define MA_COMPILER_HAS_BUILTIN(x) __has_builtin(x) +#else + #define MA_COMPILER_HAS_BUILTIN(x) 0 +#endif + +#ifndef MA_ASSUME + #if MA_COMPILER_HAS_BUILTIN(__builtin_assume) + #define MA_ASSUME(x) __builtin_assume(x) + #elif MA_COMPILER_HAS_BUILTIN(__builtin_unreachable) + #define MA_ASSUME(x) do { if (!(x)) __builtin_unreachable(); } while (0) + #elif defined(_MSC_VER) + #define MA_ASSUME(x) __assume(x) + #else + #define MA_ASSUME(x) (void)(x) + #endif +#endif + +#ifndef MA_RESTRICT + #if defined(__clang__) || defined(__GNUC__) || defined(_MSC_VER) + #define MA_RESTRICT __restrict + #else + #define MA_RESTRICT + #endif +#endif + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + #define MA_HAS_BYTESWAP16_INTRINSIC + #define MA_HAS_BYTESWAP32_INTRINSIC + #define MA_HAS_BYTESWAP64_INTRINSIC +#elif defined(__clang__) + #if MA_COMPILER_HAS_BUILTIN(__builtin_bswap16) + #define MA_HAS_BYTESWAP16_INTRINSIC + #endif + #if MA_COMPILER_HAS_BUILTIN(__builtin_bswap32) + #define MA_HAS_BYTESWAP32_INTRINSIC + #endif + #if MA_COMPILER_HAS_BUILTIN(__builtin_bswap64) + #define MA_HAS_BYTESWAP64_INTRINSIC + #endif +#elif defined(__GNUC__) + #if ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 3)) + #define MA_HAS_BYTESWAP32_INTRINSIC + #define MA_HAS_BYTESWAP64_INTRINSIC + #endif + #if ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)) + #define MA_HAS_BYTESWAP16_INTRINSIC + #endif +#endif + + +static MA_INLINE ma_bool32 ma_is_little_endian(void) +{ +#if defined(MA_X86) || defined(MA_X64) + return MA_TRUE; +#else + int n = 1; + return (*(char*)&n) == 1; +#endif +} + +static MA_INLINE ma_bool32 ma_is_big_endian(void) +{ + return !ma_is_little_endian(); +} + + +static MA_INLINE ma_uint32 ma_swap_endian_uint32(ma_uint32 n) +{ +#ifdef MA_HAS_BYTESWAP32_INTRINSIC + #if defined(_MSC_VER) + return _byteswap_ulong(n); + #elif defined(__GNUC__) || defined(__clang__) + #if defined(MA_ARM) && (defined(__ARM_ARCH) && __ARM_ARCH >= 6) && !defined(MA_64BIT) /* <-- 64-bit inline assembly has not been tested, so disabling for now. */ + /* Inline assembly optimized implementation for ARM. In my testing, GCC does not generate optimized code with __builtin_bswap32(). */ + ma_uint32 r; + __asm__ __volatile__ ( + #if defined(MA_64BIT) + "rev %w[out], %w[in]" : [out]"=r"(r) : [in]"r"(n) /* <-- This is untested. If someone in the community could test this, that would be appreciated! */ + #else + "rev %[out], %[in]" : [out]"=r"(r) : [in]"r"(n) + #endif + ); + return r; + #else + return __builtin_bswap32(n); + #endif + #else + #error "This compiler does not support the byte swap intrinsic." + #endif +#else + return ((n & 0xFF000000) >> 24) | + ((n & 0x00FF0000) >> 8) | + ((n & 0x0000FF00) << 8) | + ((n & 0x000000FF) << 24); +#endif +} + + +#if !defined(MA_EMSCRIPTEN) +#ifdef MA_WIN32 +static void ma_sleep__win32(ma_uint32 milliseconds) +{ + Sleep((DWORD)milliseconds); +} +#endif +#ifdef MA_POSIX +static void ma_sleep__posix(ma_uint32 milliseconds) +{ +#ifdef MA_EMSCRIPTEN + (void)milliseconds; + MA_ASSERT(MA_FALSE); /* The Emscripten build should never sleep. */ +#else + #if (defined(_POSIX_C_SOURCE) && _POSIX_C_SOURCE >= 199309L) || defined(MA_NX) + struct timespec ts; + ts.tv_sec = milliseconds / 1000; + ts.tv_nsec = milliseconds % 1000 * 1000000; + nanosleep(&ts, NULL); + #else + struct timeval tv; + tv.tv_sec = milliseconds / 1000; + tv.tv_usec = milliseconds % 1000 * 1000; + select(0, NULL, NULL, NULL, &tv); + #endif +#endif +} +#endif + +static MA_INLINE void ma_sleep(ma_uint32 milliseconds) +{ +#ifdef MA_WIN32 + ma_sleep__win32(milliseconds); +#endif +#ifdef MA_POSIX + ma_sleep__posix(milliseconds); +#endif +} +#endif + +static MA_INLINE void ma_yield(void) +{ +#if defined(__i386) || defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64) + /* x86/x64 */ + #if (defined(_MSC_VER) || defined(__WATCOMC__) || defined(__DMC__)) && !defined(__clang__) + #if _MSC_VER >= 1400 + _mm_pause(); + #else + #if defined(__DMC__) + /* Digital Mars does not recognize the PAUSE opcode. Fall back to NOP. */ + __asm nop; + #else + __asm pause; + #endif + #endif + #else + __asm__ __volatile__ ("pause"); + #endif +#elif (defined(__arm__) && defined(__ARM_ARCH) && __ARM_ARCH >= 7) || defined(_M_ARM64) || (defined(_M_ARM) && _M_ARM >= 7) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6T2__) + /* ARM */ + #if defined(_MSC_VER) + /* Apparently there is a __yield() intrinsic that's compatible with ARM, but I cannot find documentation for it nor can I find where it's declared. */ + __yield(); + #else + __asm__ __volatile__ ("yield"); /* ARMv6K/ARMv6T2 and above. */ + #endif +#else + /* Unknown or unsupported architecture. No-op. */ +#endif +} + + +#define MA_MM_DENORMALS_ZERO_MASK 0x0040 +#define MA_MM_FLUSH_ZERO_MASK 0x8000 + +static MA_INLINE unsigned int ma_disable_denormals(void) +{ + unsigned int prevState; + + #if defined(_MSC_VER) + { + /* + Older versions of Visual Studio don't support the "safe" versions of _controlfp_s(). I don't + know which version of Visual Studio first added support for _controlfp_s(), but I do know + that VC6 lacks support. _MSC_VER = 1200 is VC6, but if you get compilation errors on older + versions of Visual Studio, let me know and I'll make the necessary adjustment. + */ + #if _MSC_VER <= 1200 + { + prevState = _statusfp(); + _controlfp(prevState | _DN_FLUSH, _MCW_DN); + } + #else + { + unsigned int unused; + _controlfp_s(&prevState, 0, 0); + _controlfp_s(&unused, prevState | _DN_FLUSH, _MCW_DN); + } + #endif + } + #elif defined(MA_X86) || defined(MA_X64) + { + #if defined(__SSE2__) && !(defined(__TINYC__) || defined(__WATCOMC__) || defined(__COSMOPOLITAN__)) /* <-- Add compilers that lack support for _mm_getcsr() and _mm_setcsr() to this list. */ + { + prevState = _mm_getcsr(); + _mm_setcsr(prevState | MA_MM_DENORMALS_ZERO_MASK | MA_MM_FLUSH_ZERO_MASK); + } + #else + { + /* x88/64, but no support for _mm_getcsr()/_mm_setcsr(). May need to fall back to inlined assembly here. */ + prevState = 0; + } + #endif + } + #else + { + /* Unknown or unsupported architecture. No-op. */ + prevState = 0; + } + #endif + + return prevState; +} + +static MA_INLINE void ma_restore_denormals(unsigned int prevState) +{ + #if defined(_MSC_VER) + { + /* Older versions of Visual Studio do not support _controlfp_s(). See ma_disable_denormals(). */ + #if _MSC_VER <= 1200 + { + _controlfp(prevState, _MCW_DN); + } + #else + { + unsigned int unused; + _controlfp_s(&unused, prevState, _MCW_DN); + } + #endif + } + #elif defined(MA_X86) || defined(MA_X64) + { + #if defined(__SSE2__) && !(defined(__TINYC__) || defined(__WATCOMC__) || defined(__COSMOPOLITAN__)) /* <-- Add compilers that lack support for _mm_getcsr() and _mm_setcsr() to this list. */ + { + _mm_setcsr(prevState); + } + #else + { + /* x88/64, but no support for _mm_getcsr()/_mm_setcsr(). May need to fall back to inlined assembly here. */ + (void)prevState; + } + #endif + } + #else + { + /* Unknown or unsupported architecture. No-op. */ + (void)prevState; + } + #endif +} + + +#ifdef MA_ANDROID +#include + +int ma_android_sdk_version() +{ + char sdkVersion[PROP_VALUE_MAX + 1] = {0, }; + if (__system_property_get("ro.build.version.sdk", sdkVersion)) { + return atoi(sdkVersion); + } + + return 0; +} +#endif + + +#ifndef MA_COINIT_VALUE +#define MA_COINIT_VALUE 0 /* 0 = COINIT_MULTITHREADED */ +#endif + + +#ifndef MA_FLT_MAX + #ifdef FLT_MAX + #define MA_FLT_MAX FLT_MAX + #else + #define MA_FLT_MAX 3.402823466e+38F + #endif +#endif + + +#ifndef MA_PI +#define MA_PI 3.14159265358979323846264f +#endif +#ifndef MA_PI_D +#define MA_PI_D 3.14159265358979323846264 +#endif +#ifndef MA_TAU +#define MA_TAU 6.28318530717958647693f +#endif +#ifndef MA_TAU_D +#define MA_TAU_D 6.28318530717958647693 +#endif + + +/* The default format when ma_format_unknown (0) is requested when initializing a device. */ +#ifndef MA_DEFAULT_FORMAT +#define MA_DEFAULT_FORMAT ma_format_f32 +#endif + +/* The default channel count to use when 0 is used when initializing a device. */ +#ifndef MA_DEFAULT_CHANNELS +#define MA_DEFAULT_CHANNELS 2 +#endif + +/* The default sample rate to use when 0 is used when initializing a device. */ +#ifndef MA_DEFAULT_SAMPLE_RATE +#define MA_DEFAULT_SAMPLE_RATE 48000 +#endif + +/* Default periods when none is specified in ma_device_init(). More periods means more work on the CPU. */ +#ifndef MA_DEFAULT_PERIODS +#define MA_DEFAULT_PERIODS 3 +#endif + +/* The default period size in milliseconds for low latency mode. */ +#ifndef MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_LOW_LATENCY +#define MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_LOW_LATENCY 10 +#endif + +/* The default buffer size in milliseconds for conservative mode. */ +#ifndef MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_CONSERVATIVE +#define MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_CONSERVATIVE 100 +#endif + +/* The default LPF filter order for linear resampling. Note that this is clamped to MA_MAX_FILTER_ORDER. */ +#ifndef MA_DEFAULT_RESAMPLER_LPF_ORDER + #if MA_MAX_FILTER_ORDER >= 4 + #define MA_DEFAULT_RESAMPLER_LPF_ORDER 4 + #else + #define MA_DEFAULT_RESAMPLER_LPF_ORDER MA_MAX_FILTER_ORDER + #endif +#endif + + +#if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wunused-variable" +#endif + +/* Standard sample rates, in order of priority. */ +static ma_uint32 g_maStandardSampleRatePriorities[] = { + (ma_uint32)ma_standard_sample_rate_48000, + (ma_uint32)ma_standard_sample_rate_44100, + + (ma_uint32)ma_standard_sample_rate_32000, + (ma_uint32)ma_standard_sample_rate_24000, + (ma_uint32)ma_standard_sample_rate_22050, + + (ma_uint32)ma_standard_sample_rate_88200, + (ma_uint32)ma_standard_sample_rate_96000, + (ma_uint32)ma_standard_sample_rate_176400, + (ma_uint32)ma_standard_sample_rate_192000, + + (ma_uint32)ma_standard_sample_rate_16000, + (ma_uint32)ma_standard_sample_rate_11025, + (ma_uint32)ma_standard_sample_rate_8000, + + (ma_uint32)ma_standard_sample_rate_352800, + (ma_uint32)ma_standard_sample_rate_384000 +}; + +static MA_INLINE ma_bool32 ma_is_standard_sample_rate(ma_uint32 sampleRate) +{ + ma_uint32 iSampleRate; + + for (iSampleRate = 0; iSampleRate < sizeof(g_maStandardSampleRatePriorities) / sizeof(g_maStandardSampleRatePriorities[0]); iSampleRate += 1) { + if (g_maStandardSampleRatePriorities[iSampleRate] == sampleRate) { + return MA_TRUE; + } + } + + /* Getting here means the sample rate is not supported. */ + return MA_FALSE; +} + + +static ma_format g_maFormatPriorities[] = { + ma_format_s16, /* Most common */ + ma_format_f32, + + /*ma_format_s24_32,*/ /* Clean alignment */ + ma_format_s32, + + ma_format_s24, /* Unclean alignment */ + + ma_format_u8 /* Low quality */ +}; +#if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic pop +#endif + + +MA_API void ma_version(ma_uint32* pMajor, ma_uint32* pMinor, ma_uint32* pRevision) +{ + if (pMajor) { + *pMajor = MA_VERSION_MAJOR; + } + + if (pMinor) { + *pMinor = MA_VERSION_MINOR; + } + + if (pRevision) { + *pRevision = MA_VERSION_REVISION; + } +} + +MA_API const char* ma_version_string(void) +{ + return MA_VERSION_STRING; +} + + +/****************************************************************************** + +Standard Library Stuff + +******************************************************************************/ +#ifndef MA_ASSERT +#define MA_ASSERT(condition) assert(condition) +#endif + +#ifndef MA_MALLOC +#define MA_MALLOC(sz) malloc((sz)) +#endif +#ifndef MA_REALLOC +#define MA_REALLOC(p, sz) realloc((p), (sz)) +#endif +#ifndef MA_FREE +#define MA_FREE(p) free((p)) +#endif + +static MA_INLINE void ma_zero_memory_default(void* p, size_t sz) +{ + if (p == NULL) { + MA_ASSERT(sz == 0); /* If this is triggered there's an error with the calling code. */ + return; + } + + if (sz > 0) { + memset(p, 0, sz); + } +} + + +#ifndef MA_ZERO_MEMORY +#define MA_ZERO_MEMORY(p, sz) ma_zero_memory_default((p), (sz)) +#endif +#ifndef MA_COPY_MEMORY +#define MA_COPY_MEMORY(dst, src, sz) memcpy((dst), (src), (sz)) +#endif +#ifndef MA_MOVE_MEMORY +#define MA_MOVE_MEMORY(dst, src, sz) memmove((dst), (src), (sz)) +#endif + +#define MA_ZERO_OBJECT(p) MA_ZERO_MEMORY((p), sizeof(*(p))) + +#define ma_countof(x) (sizeof(x) / sizeof(x[0])) +#define ma_max(x, y) (((x) > (y)) ? (x) : (y)) +#define ma_min(x, y) (((x) < (y)) ? (x) : (y)) +#define ma_abs(x) (((x) > 0) ? (x) : -(x)) +#define ma_clamp(x, lo, hi) (ma_max(lo, ma_min(x, hi))) +#define ma_offset_ptr(p, offset) (((ma_uint8*)(p)) + (offset)) +#define ma_align(x, a) (((x) + ((a)-1)) & ~((a)-1)) +#define ma_align_64(x) ma_align(x, 8) + +#define ma_buffer_frame_capacity(buffer, channels, format) (sizeof(buffer) / ma_get_bytes_per_sample(format) / (channels)) + +static MA_INLINE double ma_sind(double x) +{ + /* TODO: Implement custom sin(x). */ + return sin(x); +} + +static MA_INLINE double ma_expd(double x) +{ + /* TODO: Implement custom exp(x). */ + return exp(x); +} + +static MA_INLINE double ma_logd(double x) +{ + /* TODO: Implement custom log(x). */ + return log(x); +} + +static MA_INLINE double ma_powd(double x, double y) +{ + /* TODO: Implement custom pow(x, y). */ + return pow(x, y); +} + +static MA_INLINE double ma_sqrtd(double x) +{ + /* TODO: Implement custom sqrt(x). */ + return sqrt(x); +} + + +static MA_INLINE float ma_rsqrtf(float x) +{ + #if defined(MA_SUPPORT_SSE2) && !defined(MA_NO_SSE2) && (defined(MA_X64) || (defined(_M_IX86_FP) && _M_IX86_FP == 2) || defined(__SSE2__)) + { + /* + For SSE we can use RSQRTSS. + + This Stack Overflow post suggests that compilers don't necessarily generate optimal code + when using intrinsics: + + https://web.archive.org/web/20221211012522/https://stackoverflow.com/questions/32687079/getting-fewest-instructions-for-rsqrtss-wrapper + + I'm going to do something similar here, but a bit simpler. + */ + #if defined(__GNUC__) || defined(__clang__) + { + float result; + __asm__ __volatile__("rsqrtss %1, %0" : "=x"(result) : "x"(x)); + return result; + } + #else + { + return _mm_cvtss_f32(_mm_rsqrt_ss(_mm_set_ps1(x))); + } + #endif + } + #else + { + return 1 / (float)ma_sqrtd(x); + } + #endif +} + + +static MA_INLINE float ma_sinf(float x) +{ + return (float)ma_sind((float)x); +} + +static MA_INLINE double ma_cosd(double x) +{ + return ma_sind((MA_PI_D*0.5) - x); +} + +static MA_INLINE float ma_cosf(float x) +{ + return (float)ma_cosd((float)x); +} + +static MA_INLINE double ma_log10d(double x) +{ + return ma_logd(x) * 0.43429448190325182765; +} + +static MA_INLINE float ma_powf(float x, float y) +{ + return (float)ma_powd((double)x, (double)y); +} + +static MA_INLINE float ma_log10f(float x) +{ + return (float)ma_log10d((double)x); +} + + +static MA_INLINE double ma_degrees_to_radians(double degrees) +{ + return degrees * 0.01745329252; +} + +static MA_INLINE double ma_radians_to_degrees(double radians) +{ + return radians * 57.295779512896; +} + +static MA_INLINE float ma_degrees_to_radians_f(float degrees) +{ + return degrees * 0.01745329252f; +} + +static MA_INLINE float ma_radians_to_degrees_f(float radians) +{ + return radians * 57.295779512896f; +} + + +/* +Return Values: + 0: Success + 22: EINVAL + 34: ERANGE + +Not using symbolic constants for errors because I want to avoid #including errno.h + +These are marked as no-inline because of some bad code generation by Clang. None of these functions +are used in any performance-critical code within miniaudio. +*/ +MA_API MA_NO_INLINE int ma_strcpy_s(char* dst, size_t dstSizeInBytes, const char* src) +{ + size_t i; + + if (dst == 0) { + return 22; + } + if (dstSizeInBytes == 0) { + return 34; + } + if (src == 0) { + dst[0] = '\0'; + return 22; + } + + for (i = 0; i < dstSizeInBytes && src[i] != '\0'; ++i) { + dst[i] = src[i]; + } + + if (i < dstSizeInBytes) { + dst[i] = '\0'; + return 0; + } + + dst[0] = '\0'; + return 34; +} + +MA_API MA_NO_INLINE int ma_wcscpy_s(wchar_t* dst, size_t dstCap, const wchar_t* src) +{ + size_t i; + + if (dst == 0) { + return 22; + } + if (dstCap == 0) { + return 34; + } + if (src == 0) { + dst[0] = '\0'; + return 22; + } + + for (i = 0; i < dstCap && src[i] != '\0'; ++i) { + dst[i] = src[i]; + } + + if (i < dstCap) { + dst[i] = '\0'; + return 0; + } + + dst[0] = '\0'; + return 34; +} + + +MA_API MA_NO_INLINE int ma_strncpy_s(char* dst, size_t dstSizeInBytes, const char* src, size_t count) +{ + size_t maxcount; + size_t i; + + if (dst == 0) { + return 22; + } + if (dstSizeInBytes == 0) { + return 34; + } + if (src == 0) { + dst[0] = '\0'; + return 22; + } + + maxcount = count; + if (count == ((size_t)-1) || count >= dstSizeInBytes) { /* -1 = _TRUNCATE */ + maxcount = dstSizeInBytes - 1; + } + + for (i = 0; i < maxcount && src[i] != '\0'; ++i) { + dst[i] = src[i]; + } + + if (src[i] == '\0' || i == count || count == ((size_t)-1)) { + dst[i] = '\0'; + return 0; + } + + dst[0] = '\0'; + return 34; +} + +MA_API MA_NO_INLINE int ma_strcat_s(char* dst, size_t dstSizeInBytes, const char* src) +{ + char* dstorig; + + if (dst == 0) { + return 22; + } + if (dstSizeInBytes == 0) { + return 34; + } + if (src == 0) { + dst[0] = '\0'; + return 22; + } + + dstorig = dst; + + while (dstSizeInBytes > 0 && dst[0] != '\0') { + dst += 1; + dstSizeInBytes -= 1; + } + + if (dstSizeInBytes == 0) { + return 22; /* Unterminated. */ + } + + + while (dstSizeInBytes > 0 && src[0] != '\0') { + *dst++ = *src++; + dstSizeInBytes -= 1; + } + + if (dstSizeInBytes > 0) { + dst[0] = '\0'; + } else { + dstorig[0] = '\0'; + return 34; + } + + return 0; +} + +MA_API MA_NO_INLINE int ma_strncat_s(char* dst, size_t dstSizeInBytes, const char* src, size_t count) +{ + char* dstorig; + + if (dst == 0) { + return 22; + } + if (dstSizeInBytes == 0) { + return 34; + } + if (src == 0) { + return 22; + } + + dstorig = dst; + + while (dstSizeInBytes > 0 && dst[0] != '\0') { + dst += 1; + dstSizeInBytes -= 1; + } + + if (dstSizeInBytes == 0) { + return 22; /* Unterminated. */ + } + + + if (count == ((size_t)-1)) { /* _TRUNCATE */ + count = dstSizeInBytes - 1; + } + + while (dstSizeInBytes > 0 && src[0] != '\0' && count > 0) { + *dst++ = *src++; + dstSizeInBytes -= 1; + count -= 1; + } + + if (dstSizeInBytes > 0) { + dst[0] = '\0'; + } else { + dstorig[0] = '\0'; + return 34; + } + + return 0; +} + +MA_API MA_NO_INLINE int ma_itoa_s(int value, char* dst, size_t dstSizeInBytes, int radix) +{ + int sign; + unsigned int valueU; + char* dstEnd; + + if (dst == NULL || dstSizeInBytes == 0) { + return 22; + } + if (radix < 2 || radix > 36) { + dst[0] = '\0'; + return 22; + } + + sign = (value < 0 && radix == 10) ? -1 : 1; /* The negative sign is only used when the base is 10. */ + + if (value < 0) { + valueU = -value; + } else { + valueU = value; + } + + dstEnd = dst; + do + { + int remainder = valueU % radix; + if (remainder > 9) { + *dstEnd = (char)((remainder - 10) + 'a'); + } else { + *dstEnd = (char)(remainder + '0'); + } + + dstEnd += 1; + dstSizeInBytes -= 1; + valueU /= radix; + } while (dstSizeInBytes > 0 && valueU > 0); + + if (dstSizeInBytes == 0) { + dst[0] = '\0'; + return 22; /* Ran out of room in the output buffer. */ + } + + if (sign < 0) { + *dstEnd++ = '-'; + dstSizeInBytes -= 1; + } + + if (dstSizeInBytes == 0) { + dst[0] = '\0'; + return 22; /* Ran out of room in the output buffer. */ + } + + *dstEnd = '\0'; + + + /* At this point the string will be reversed. */ + dstEnd -= 1; + while (dst < dstEnd) { + char temp = *dst; + *dst = *dstEnd; + *dstEnd = temp; + + dst += 1; + dstEnd -= 1; + } + + return 0; +} + +MA_API MA_NO_INLINE int ma_strcmp(const char* str1, const char* str2) +{ + if (str1 == str2) return 0; + + /* These checks differ from the standard implementation. It's not important, but I prefer it just for sanity. */ + if (str1 == NULL) return -1; + if (str2 == NULL) return 1; + + for (;;) { + if (str1[0] == '\0') { + break; + } + if (str1[0] != str2[0]) { + break; + } + + str1 += 1; + str2 += 1; + } + + return ((unsigned char*)str1)[0] - ((unsigned char*)str2)[0]; +} + +MA_API MA_NO_INLINE int ma_strappend(char* dst, size_t dstSize, const char* srcA, const char* srcB) +{ + int result; + + result = ma_strncpy_s(dst, dstSize, srcA, (size_t)-1); + if (result != 0) { + return result; + } + + result = ma_strncat_s(dst, dstSize, srcB, (size_t)-1); + if (result != 0) { + return result; + } + + return result; +} + +MA_API MA_NO_INLINE char* ma_copy_string(const char* src, const ma_allocation_callbacks* pAllocationCallbacks) +{ + size_t sz; + char* dst; + + if (src == NULL) { + return NULL; + } + + sz = strlen(src)+1; + dst = (char*)ma_malloc(sz, pAllocationCallbacks); + if (dst == NULL) { + return NULL; + } + + ma_strcpy_s(dst, sz, src); + + return dst; +} + +MA_API MA_NO_INLINE wchar_t* ma_copy_string_w(const wchar_t* src, const ma_allocation_callbacks* pAllocationCallbacks) +{ + size_t sz = wcslen(src)+1; + wchar_t* dst = (wchar_t*)ma_malloc(sz * sizeof(*dst), pAllocationCallbacks); + if (dst == NULL) { + return NULL; + } + + ma_wcscpy_s(dst, sz, src); + + return dst; +} + + + +#include +static ma_result ma_result_from_errno(int e) +{ + if (e == 0) { + return MA_SUCCESS; + } +#ifdef EPERM + else if (e == EPERM) { return MA_INVALID_OPERATION; } +#endif +#ifdef ENOENT + else if (e == ENOENT) { return MA_DOES_NOT_EXIST; } +#endif +#ifdef ESRCH + else if (e == ESRCH) { return MA_DOES_NOT_EXIST; } +#endif +#ifdef EINTR + else if (e == EINTR) { return MA_INTERRUPT; } +#endif +#ifdef EIO + else if (e == EIO) { return MA_IO_ERROR; } +#endif +#ifdef ENXIO + else if (e == ENXIO) { return MA_DOES_NOT_EXIST; } +#endif +#ifdef E2BIG + else if (e == E2BIG) { return MA_INVALID_ARGS; } +#endif +#ifdef ENOEXEC + else if (e == ENOEXEC) { return MA_INVALID_FILE; } +#endif +#ifdef EBADF + else if (e == EBADF) { return MA_INVALID_FILE; } +#endif +#ifdef ECHILD + else if (e == ECHILD) { return MA_ERROR; } +#endif +#ifdef EAGAIN + else if (e == EAGAIN) { return MA_UNAVAILABLE; } +#endif +#ifdef ENOMEM + else if (e == ENOMEM) { return MA_OUT_OF_MEMORY; } +#endif +#ifdef EACCES + else if (e == EACCES) { return MA_ACCESS_DENIED; } +#endif +#ifdef EFAULT + else if (e == EFAULT) { return MA_BAD_ADDRESS; } +#endif +#ifdef ENOTBLK + else if (e == ENOTBLK) { return MA_ERROR; } +#endif +#ifdef EBUSY + else if (e == EBUSY) { return MA_BUSY; } +#endif +#ifdef EEXIST + else if (e == EEXIST) { return MA_ALREADY_EXISTS; } +#endif +#ifdef EXDEV + else if (e == EXDEV) { return MA_ERROR; } +#endif +#ifdef ENODEV + else if (e == ENODEV) { return MA_DOES_NOT_EXIST; } +#endif +#ifdef ENOTDIR + else if (e == ENOTDIR) { return MA_NOT_DIRECTORY; } +#endif +#ifdef EISDIR + else if (e == EISDIR) { return MA_IS_DIRECTORY; } +#endif +#ifdef EINVAL + else if (e == EINVAL) { return MA_INVALID_ARGS; } +#endif +#ifdef ENFILE + else if (e == ENFILE) { return MA_TOO_MANY_OPEN_FILES; } +#endif +#ifdef EMFILE + else if (e == EMFILE) { return MA_TOO_MANY_OPEN_FILES; } +#endif +#ifdef ENOTTY + else if (e == ENOTTY) { return MA_INVALID_OPERATION; } +#endif +#ifdef ETXTBSY + else if (e == ETXTBSY) { return MA_BUSY; } +#endif +#ifdef EFBIG + else if (e == EFBIG) { return MA_TOO_BIG; } +#endif +#ifdef ENOSPC + else if (e == ENOSPC) { return MA_NO_SPACE; } +#endif +#ifdef ESPIPE + else if (e == ESPIPE) { return MA_BAD_SEEK; } +#endif +#ifdef EROFS + else if (e == EROFS) { return MA_ACCESS_DENIED; } +#endif +#ifdef EMLINK + else if (e == EMLINK) { return MA_TOO_MANY_LINKS; } +#endif +#ifdef EPIPE + else if (e == EPIPE) { return MA_BAD_PIPE; } +#endif +#ifdef EDOM + else if (e == EDOM) { return MA_OUT_OF_RANGE; } +#endif +#ifdef ERANGE + else if (e == ERANGE) { return MA_OUT_OF_RANGE; } +#endif +#ifdef EDEADLK + else if (e == EDEADLK) { return MA_DEADLOCK; } +#endif +#ifdef ENAMETOOLONG + else if (e == ENAMETOOLONG) { return MA_PATH_TOO_LONG; } +#endif +#ifdef ENOLCK + else if (e == ENOLCK) { return MA_ERROR; } +#endif +#ifdef ENOSYS + else if (e == ENOSYS) { return MA_NOT_IMPLEMENTED; } +#endif +#ifdef ENOTEMPTY + else if (e == ENOTEMPTY) { return MA_DIRECTORY_NOT_EMPTY; } +#endif +#ifdef ELOOP + else if (e == ELOOP) { return MA_TOO_MANY_LINKS; } +#endif +#ifdef ENOMSG + else if (e == ENOMSG) { return MA_NO_MESSAGE; } +#endif +#ifdef EIDRM + else if (e == EIDRM) { return MA_ERROR; } +#endif +#ifdef ECHRNG + else if (e == ECHRNG) { return MA_ERROR; } +#endif +#ifdef EL2NSYNC + else if (e == EL2NSYNC) { return MA_ERROR; } +#endif +#ifdef EL3HLT + else if (e == EL3HLT) { return MA_ERROR; } +#endif +#ifdef EL3RST + else if (e == EL3RST) { return MA_ERROR; } +#endif +#ifdef ELNRNG + else if (e == ELNRNG) { return MA_OUT_OF_RANGE; } +#endif +#ifdef EUNATCH + else if (e == EUNATCH) { return MA_ERROR; } +#endif +#ifdef ENOCSI + else if (e == ENOCSI) { return MA_ERROR; } +#endif +#ifdef EL2HLT + else if (e == EL2HLT) { return MA_ERROR; } +#endif +#ifdef EBADE + else if (e == EBADE) { return MA_ERROR; } +#endif +#ifdef EBADR + else if (e == EBADR) { return MA_ERROR; } +#endif +#ifdef EXFULL + else if (e == EXFULL) { return MA_ERROR; } +#endif +#ifdef ENOANO + else if (e == ENOANO) { return MA_ERROR; } +#endif +#ifdef EBADRQC + else if (e == EBADRQC) { return MA_ERROR; } +#endif +#ifdef EBADSLT + else if (e == EBADSLT) { return MA_ERROR; } +#endif +#ifdef EBFONT + else if (e == EBFONT) { return MA_INVALID_FILE; } +#endif +#ifdef ENOSTR + else if (e == ENOSTR) { return MA_ERROR; } +#endif +#ifdef ENODATA + else if (e == ENODATA) { return MA_NO_DATA_AVAILABLE; } +#endif +#ifdef ETIME + else if (e == ETIME) { return MA_TIMEOUT; } +#endif +#ifdef ENOSR + else if (e == ENOSR) { return MA_NO_DATA_AVAILABLE; } +#endif +#ifdef ENONET + else if (e == ENONET) { return MA_NO_NETWORK; } +#endif +#ifdef ENOPKG + else if (e == ENOPKG) { return MA_ERROR; } +#endif +#ifdef EREMOTE + else if (e == EREMOTE) { return MA_ERROR; } +#endif +#ifdef ENOLINK + else if (e == ENOLINK) { return MA_ERROR; } +#endif +#ifdef EADV + else if (e == EADV) { return MA_ERROR; } +#endif +#ifdef ESRMNT + else if (e == ESRMNT) { return MA_ERROR; } +#endif +#ifdef ECOMM + else if (e == ECOMM) { return MA_ERROR; } +#endif +#ifdef EPROTO + else if (e == EPROTO) { return MA_ERROR; } +#endif +#ifdef EMULTIHOP + else if (e == EMULTIHOP) { return MA_ERROR; } +#endif +#ifdef EDOTDOT + else if (e == EDOTDOT) { return MA_ERROR; } +#endif +#ifdef EBADMSG + else if (e == EBADMSG) { return MA_BAD_MESSAGE; } +#endif +#ifdef EOVERFLOW + else if (e == EOVERFLOW) { return MA_TOO_BIG; } +#endif +#ifdef ENOTUNIQ + else if (e == ENOTUNIQ) { return MA_NOT_UNIQUE; } +#endif +#ifdef EBADFD + else if (e == EBADFD) { return MA_ERROR; } +#endif +#ifdef EREMCHG + else if (e == EREMCHG) { return MA_ERROR; } +#endif +#ifdef ELIBACC + else if (e == ELIBACC) { return MA_ACCESS_DENIED; } +#endif +#ifdef ELIBBAD + else if (e == ELIBBAD) { return MA_INVALID_FILE; } +#endif +#ifdef ELIBSCN + else if (e == ELIBSCN) { return MA_INVALID_FILE; } +#endif +#ifdef ELIBMAX + else if (e == ELIBMAX) { return MA_ERROR; } +#endif +#ifdef ELIBEXEC + else if (e == ELIBEXEC) { return MA_ERROR; } +#endif +#ifdef EILSEQ + else if (e == EILSEQ) { return MA_INVALID_DATA; } +#endif +#ifdef ERESTART + else if (e == ERESTART) { return MA_ERROR; } +#endif +#ifdef ESTRPIPE + else if (e == ESTRPIPE) { return MA_ERROR; } +#endif +#ifdef EUSERS + else if (e == EUSERS) { return MA_ERROR; } +#endif +#ifdef ENOTSOCK + else if (e == ENOTSOCK) { return MA_NOT_SOCKET; } +#endif +#ifdef EDESTADDRREQ + else if (e == EDESTADDRREQ) { return MA_NO_ADDRESS; } +#endif +#ifdef EMSGSIZE + else if (e == EMSGSIZE) { return MA_TOO_BIG; } +#endif +#ifdef EPROTOTYPE + else if (e == EPROTOTYPE) { return MA_BAD_PROTOCOL; } +#endif +#ifdef ENOPROTOOPT + else if (e == ENOPROTOOPT) { return MA_PROTOCOL_UNAVAILABLE; } +#endif +#ifdef EPROTONOSUPPORT + else if (e == EPROTONOSUPPORT) { return MA_PROTOCOL_NOT_SUPPORTED; } +#endif +#ifdef ESOCKTNOSUPPORT + else if (e == ESOCKTNOSUPPORT) { return MA_SOCKET_NOT_SUPPORTED; } +#endif +#ifdef EOPNOTSUPP + else if (e == EOPNOTSUPP) { return MA_INVALID_OPERATION; } +#endif +#ifdef EPFNOSUPPORT + else if (e == EPFNOSUPPORT) { return MA_PROTOCOL_FAMILY_NOT_SUPPORTED; } +#endif +#ifdef EAFNOSUPPORT + else if (e == EAFNOSUPPORT) { return MA_ADDRESS_FAMILY_NOT_SUPPORTED; } +#endif +#ifdef EADDRINUSE + else if (e == EADDRINUSE) { return MA_ALREADY_IN_USE; } +#endif +#ifdef EADDRNOTAVAIL + else if (e == EADDRNOTAVAIL) { return MA_ERROR; } +#endif +#ifdef ENETDOWN + else if (e == ENETDOWN) { return MA_NO_NETWORK; } +#endif +#ifdef ENETUNREACH + else if (e == ENETUNREACH) { return MA_NO_NETWORK; } +#endif +#ifdef ENETRESET + else if (e == ENETRESET) { return MA_NO_NETWORK; } +#endif +#ifdef ECONNABORTED + else if (e == ECONNABORTED) { return MA_NO_NETWORK; } +#endif +#ifdef ECONNRESET + else if (e == ECONNRESET) { return MA_CONNECTION_RESET; } +#endif +#ifdef ENOBUFS + else if (e == ENOBUFS) { return MA_NO_SPACE; } +#endif +#ifdef EISCONN + else if (e == EISCONN) { return MA_ALREADY_CONNECTED; } +#endif +#ifdef ENOTCONN + else if (e == ENOTCONN) { return MA_NOT_CONNECTED; } +#endif +#ifdef ESHUTDOWN + else if (e == ESHUTDOWN) { return MA_ERROR; } +#endif +#ifdef ETOOMANYREFS + else if (e == ETOOMANYREFS) { return MA_ERROR; } +#endif +#ifdef ETIMEDOUT + else if (e == ETIMEDOUT) { return MA_TIMEOUT; } +#endif +#ifdef ECONNREFUSED + else if (e == ECONNREFUSED) { return MA_CONNECTION_REFUSED; } +#endif +#ifdef EHOSTDOWN + else if (e == EHOSTDOWN) { return MA_NO_HOST; } +#endif +#ifdef EHOSTUNREACH + else if (e == EHOSTUNREACH) { return MA_NO_HOST; } +#endif +#ifdef EALREADY + else if (e == EALREADY) { return MA_IN_PROGRESS; } +#endif +#ifdef EINPROGRESS + else if (e == EINPROGRESS) { return MA_IN_PROGRESS; } +#endif +#ifdef ESTALE + else if (e == ESTALE) { return MA_INVALID_FILE; } +#endif +#ifdef EUCLEAN + else if (e == EUCLEAN) { return MA_ERROR; } +#endif +#ifdef ENOTNAM + else if (e == ENOTNAM) { return MA_ERROR; } +#endif +#ifdef ENAVAIL + else if (e == ENAVAIL) { return MA_ERROR; } +#endif +#ifdef EISNAM + else if (e == EISNAM) { return MA_ERROR; } +#endif +#ifdef EREMOTEIO + else if (e == EREMOTEIO) { return MA_IO_ERROR; } +#endif +#ifdef EDQUOT + else if (e == EDQUOT) { return MA_NO_SPACE; } +#endif +#ifdef ENOMEDIUM + else if (e == ENOMEDIUM) { return MA_DOES_NOT_EXIST; } +#endif +#ifdef EMEDIUMTYPE + else if (e == EMEDIUMTYPE) { return MA_ERROR; } +#endif +#ifdef ECANCELED + else if (e == ECANCELED) { return MA_CANCELLED; } +#endif +#ifdef ENOKEY + else if (e == ENOKEY) { return MA_ERROR; } +#endif +#ifdef EKEYEXPIRED + else if (e == EKEYEXPIRED) { return MA_ERROR; } +#endif +#ifdef EKEYREVOKED + else if (e == EKEYREVOKED) { return MA_ERROR; } +#endif +#ifdef EKEYREJECTED + else if (e == EKEYREJECTED) { return MA_ERROR; } +#endif +#ifdef EOWNERDEAD + else if (e == EOWNERDEAD) { return MA_ERROR; } +#endif +#ifdef ENOTRECOVERABLE + else if (e == ENOTRECOVERABLE) { return MA_ERROR; } +#endif +#ifdef ERFKILL + else if (e == ERFKILL) { return MA_ERROR; } +#endif +#ifdef EHWPOISON + else if (e == EHWPOISON) { return MA_ERROR; } +#endif + else { + return MA_ERROR; + } +} + +MA_API ma_result ma_fopen(FILE** ppFile, const char* pFilePath, const char* pOpenMode) +{ +#if defined(_MSC_VER) && _MSC_VER >= 1400 + errno_t err; +#endif + + if (ppFile != NULL) { + *ppFile = NULL; /* Safety. */ + } + + if (pFilePath == NULL || pOpenMode == NULL || ppFile == NULL) { + return MA_INVALID_ARGS; + } + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + err = fopen_s(ppFile, pFilePath, pOpenMode); + if (err != 0) { + return ma_result_from_errno(err); + } +#else +#if defined(_WIN32) || defined(__APPLE__) + *ppFile = fopen(pFilePath, pOpenMode); +#else + #if defined(_FILE_OFFSET_BITS) && _FILE_OFFSET_BITS == 64 && defined(_LARGEFILE64_SOURCE) + *ppFile = fopen64(pFilePath, pOpenMode); + #else + *ppFile = fopen(pFilePath, pOpenMode); + #endif +#endif + if (*ppFile == NULL) { + ma_result result = ma_result_from_errno(errno); + if (result == MA_SUCCESS) { + result = MA_ERROR; /* Just a safety check to make sure we never ever return success when pFile == NULL. */ + } + + return result; + } +#endif + + return MA_SUCCESS; +} + + + +/* +_wfopen() isn't always available in all compilation environments. + + * Windows only. + * MSVC seems to support it universally as far back as VC6 from what I can tell (haven't checked further back). + * MinGW-64 (both 32- and 64-bit) seems to support it. + * MinGW wraps it in !defined(__STRICT_ANSI__). + * OpenWatcom wraps it in !defined(_NO_EXT_KEYS). + +This can be reviewed as compatibility issues arise. The preference is to use _wfopen_s() and _wfopen() as opposed to the wcsrtombs() +fallback, so if you notice your compiler not detecting this properly I'm happy to look at adding support. +*/ +#if defined(_WIN32) + #if defined(_MSC_VER) || defined(__MINGW64__) || (!defined(__STRICT_ANSI__) && !defined(_NO_EXT_KEYS)) + #define MA_HAS_WFOPEN + #endif +#endif + +MA_API ma_result ma_wfopen(FILE** ppFile, const wchar_t* pFilePath, const wchar_t* pOpenMode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (ppFile != NULL) { + *ppFile = NULL; /* Safety. */ + } + + if (pFilePath == NULL || pOpenMode == NULL || ppFile == NULL) { + return MA_INVALID_ARGS; + } + +#if defined(MA_HAS_WFOPEN) + { + /* Use _wfopen() on Windows. */ + #if defined(_MSC_VER) && _MSC_VER >= 1400 + errno_t err = _wfopen_s(ppFile, pFilePath, pOpenMode); + if (err != 0) { + return ma_result_from_errno(err); + } + #else + *ppFile = _wfopen(pFilePath, pOpenMode); + if (*ppFile == NULL) { + return ma_result_from_errno(errno); + } + #endif + (void)pAllocationCallbacks; + } +#else + /* + Use fopen() on anything other than Windows. Requires a conversion. This is annoying because fopen() is locale specific. The only real way I can + think of to do this is with wcsrtombs(). Note that wcstombs() is apparently not thread-safe because it uses a static global mbstate_t object for + maintaining state. I've checked this with -std=c89 and it works, but if somebody get's a compiler error I'll look into improving compatibility. + */ + { + mbstate_t mbs; + size_t lenMB; + const wchar_t* pFilePathTemp = pFilePath; + char* pFilePathMB = NULL; + char pOpenModeMB[32] = {0}; + + /* Get the length first. */ + MA_ZERO_OBJECT(&mbs); + lenMB = wcsrtombs(NULL, &pFilePathTemp, 0, &mbs); + if (lenMB == (size_t)-1) { + return ma_result_from_errno(errno); + } + + pFilePathMB = (char*)ma_malloc(lenMB + 1, pAllocationCallbacks); + if (pFilePathMB == NULL) { + return MA_OUT_OF_MEMORY; + } + + pFilePathTemp = pFilePath; + MA_ZERO_OBJECT(&mbs); + wcsrtombs(pFilePathMB, &pFilePathTemp, lenMB + 1, &mbs); + + /* The open mode should always consist of ASCII characters so we should be able to do a trivial conversion. */ + { + size_t i = 0; + for (;;) { + if (pOpenMode[i] == 0) { + pOpenModeMB[i] = '\0'; + break; + } + + pOpenModeMB[i] = (char)pOpenMode[i]; + i += 1; + } + } + + *ppFile = fopen(pFilePathMB, pOpenModeMB); + + ma_free(pFilePathMB, pAllocationCallbacks); + } + + if (*ppFile == NULL) { + return MA_ERROR; + } +#endif + + return MA_SUCCESS; +} + + + +static MA_INLINE void ma_copy_memory_64(void* dst, const void* src, ma_uint64 sizeInBytes) +{ +#if 0xFFFFFFFFFFFFFFFF <= MA_SIZE_MAX + MA_COPY_MEMORY(dst, src, (size_t)sizeInBytes); +#else + while (sizeInBytes > 0) { + ma_uint64 bytesToCopyNow = sizeInBytes; + if (bytesToCopyNow > MA_SIZE_MAX) { + bytesToCopyNow = MA_SIZE_MAX; + } + + MA_COPY_MEMORY(dst, src, (size_t)bytesToCopyNow); /* Safe cast to size_t. */ + + sizeInBytes -= bytesToCopyNow; + dst = ( void*)(( ma_uint8*)dst + bytesToCopyNow); + src = (const void*)((const ma_uint8*)src + bytesToCopyNow); + } +#endif +} + +static MA_INLINE void ma_zero_memory_64(void* dst, ma_uint64 sizeInBytes) +{ +#if 0xFFFFFFFFFFFFFFFF <= MA_SIZE_MAX + MA_ZERO_MEMORY(dst, (size_t)sizeInBytes); +#else + while (sizeInBytes > 0) { + ma_uint64 bytesToZeroNow = sizeInBytes; + if (bytesToZeroNow > MA_SIZE_MAX) { + bytesToZeroNow = MA_SIZE_MAX; + } + + MA_ZERO_MEMORY(dst, (size_t)bytesToZeroNow); /* Safe cast to size_t. */ + + sizeInBytes -= bytesToZeroNow; + dst = (void*)((ma_uint8*)dst + bytesToZeroNow); + } +#endif +} + + +/* Thanks to good old Bit Twiddling Hacks for this one: http://graphics.stanford.edu/~seander/bithacks.html#RoundUpPowerOf2 */ +static MA_INLINE unsigned int ma_next_power_of_2(unsigned int x) +{ + x--; + x |= x >> 1; + x |= x >> 2; + x |= x >> 4; + x |= x >> 8; + x |= x >> 16; + x++; + + return x; +} + +static MA_INLINE unsigned int ma_prev_power_of_2(unsigned int x) +{ + return ma_next_power_of_2(x) >> 1; +} + +static MA_INLINE unsigned int ma_round_to_power_of_2(unsigned int x) +{ + unsigned int prev = ma_prev_power_of_2(x); + unsigned int next = ma_next_power_of_2(x); + if ((next - x) > (x - prev)) { + return prev; + } else { + return next; + } +} + +static MA_INLINE unsigned int ma_count_set_bits(unsigned int x) +{ + unsigned int count = 0; + while (x != 0) { + if (x & 1) { + count += 1; + } + + x = x >> 1; + } + + return count; +} + + + +/************************************************************************************************************************************************************** + +Allocation Callbacks + +**************************************************************************************************************************************************************/ +static void* ma__malloc_default(size_t sz, void* pUserData) +{ + (void)pUserData; + return MA_MALLOC(sz); +} + +static void* ma__realloc_default(void* p, size_t sz, void* pUserData) +{ + (void)pUserData; + return MA_REALLOC(p, sz); +} + +static void ma__free_default(void* p, void* pUserData) +{ + (void)pUserData; + MA_FREE(p); +} + +static ma_allocation_callbacks ma_allocation_callbacks_init_default(void) +{ + ma_allocation_callbacks callbacks; + callbacks.pUserData = NULL; + callbacks.onMalloc = ma__malloc_default; + callbacks.onRealloc = ma__realloc_default; + callbacks.onFree = ma__free_default; + + return callbacks; +} + +static ma_result ma_allocation_callbacks_init_copy(ma_allocation_callbacks* pDst, const ma_allocation_callbacks* pSrc) +{ + if (pDst == NULL) { + return MA_INVALID_ARGS; + } + + if (pSrc == NULL) { + *pDst = ma_allocation_callbacks_init_default(); + } else { + if (pSrc->pUserData == NULL && pSrc->onFree == NULL && pSrc->onMalloc == NULL && pSrc->onRealloc == NULL) { + *pDst = ma_allocation_callbacks_init_default(); + } else { + if (pSrc->onFree == NULL || (pSrc->onMalloc == NULL && pSrc->onRealloc == NULL)) { + return MA_INVALID_ARGS; /* Invalid allocation callbacks. */ + } else { + *pDst = *pSrc; + } + } + } + + return MA_SUCCESS; +} + + + + +/************************************************************************************************************************************************************** + +Logging + +**************************************************************************************************************************************************************/ +MA_API const char* ma_log_level_to_string(ma_uint32 logLevel) +{ + switch (logLevel) + { + case MA_LOG_LEVEL_DEBUG: return "DEBUG"; + case MA_LOG_LEVEL_INFO: return "INFO"; + case MA_LOG_LEVEL_WARNING: return "WARNING"; + case MA_LOG_LEVEL_ERROR: return "ERROR"; + default: return "ERROR"; + } +} + +#if defined(MA_DEBUG_OUTPUT) +#if defined(MA_ANDROID) + #include +#endif + +/* Customize this to use a specific tag in __android_log_print() for debug output messages. */ +#ifndef MA_ANDROID_LOG_TAG +#define MA_ANDROID_LOG_TAG "miniaudio" +#endif + +void ma_log_callback_debug(void* pUserData, ma_uint32 level, const char* pMessage) +{ + (void)pUserData; + + /* Special handling for some platforms. */ + #if defined(MA_ANDROID) + { + /* Android. */ + __android_log_print(ANDROID_LOG_DEBUG, MA_ANDROID_LOG_TAG, "%s: %s", ma_log_level_to_string(level), pMessage); + } + #else + { + /* Everything else. */ + printf("%s: %s", ma_log_level_to_string(level), pMessage); + } + #endif +} +#endif + +MA_API ma_log_callback ma_log_callback_init(ma_log_callback_proc onLog, void* pUserData) +{ + ma_log_callback callback; + + MA_ZERO_OBJECT(&callback); + callback.onLog = onLog; + callback.pUserData = pUserData; + + return callback; +} + + +MA_API ma_result ma_log_init(const ma_allocation_callbacks* pAllocationCallbacks, ma_log* pLog) +{ + if (pLog == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pLog); + ma_allocation_callbacks_init_copy(&pLog->allocationCallbacks, pAllocationCallbacks); + + /* We need a mutex for thread safety. */ + #ifndef MA_NO_THREADING + { + ma_result result = ma_mutex_init(&pLog->lock); + if (result != MA_SUCCESS) { + return result; + } + } + #endif + + /* If we're using debug output, enable it. */ + #if defined(MA_DEBUG_OUTPUT) + { + ma_log_register_callback(pLog, ma_log_callback_init(ma_log_callback_debug, NULL)); /* Doesn't really matter if this fails. */ + } + #endif + + return MA_SUCCESS; +} + +MA_API void ma_log_uninit(ma_log* pLog) +{ + if (pLog == NULL) { + return; + } + +#ifndef MA_NO_THREADING + ma_mutex_uninit(&pLog->lock); +#endif +} + +static void ma_log_lock(ma_log* pLog) +{ +#ifndef MA_NO_THREADING + ma_mutex_lock(&pLog->lock); +#else + (void)pLog; +#endif +} + +static void ma_log_unlock(ma_log* pLog) +{ +#ifndef MA_NO_THREADING + ma_mutex_unlock(&pLog->lock); +#else + (void)pLog; +#endif +} + +MA_API ma_result ma_log_register_callback(ma_log* pLog, ma_log_callback callback) +{ + ma_result result = MA_SUCCESS; + + if (pLog == NULL || callback.onLog == NULL) { + return MA_INVALID_ARGS; + } + + ma_log_lock(pLog); + { + if (pLog->callbackCount == ma_countof(pLog->callbacks)) { + result = MA_OUT_OF_MEMORY; /* Reached the maximum allowed log callbacks. */ + } else { + pLog->callbacks[pLog->callbackCount] = callback; + pLog->callbackCount += 1; + } + } + ma_log_unlock(pLog); + + return result; +} + +MA_API ma_result ma_log_unregister_callback(ma_log* pLog, ma_log_callback callback) +{ + if (pLog == NULL) { + return MA_INVALID_ARGS; + } + + ma_log_lock(pLog); + { + ma_uint32 iLog; + for (iLog = 0; iLog < pLog->callbackCount; ) { + if (pLog->callbacks[iLog].onLog == callback.onLog) { + /* Found. Move everything down a slot. */ + ma_uint32 jLog; + for (jLog = iLog; jLog < pLog->callbackCount-1; jLog += 1) { + pLog->callbacks[jLog] = pLog->callbacks[jLog + 1]; + } + + pLog->callbackCount -= 1; + } else { + /* Not found. */ + iLog += 1; + } + } + } + ma_log_unlock(pLog); + + return MA_SUCCESS; +} + +MA_API ma_result ma_log_post(ma_log* pLog, ma_uint32 level, const char* pMessage) +{ + if (pLog == NULL || pMessage == NULL) { + return MA_INVALID_ARGS; + } + + ma_log_lock(pLog); + { + ma_uint32 iLog; + for (iLog = 0; iLog < pLog->callbackCount; iLog += 1) { + if (pLog->callbacks[iLog].onLog) { + pLog->callbacks[iLog].onLog(pLog->callbacks[iLog].pUserData, level, pMessage); + } + } + } + ma_log_unlock(pLog); + + return MA_SUCCESS; +} + + +/* +We need to emulate _vscprintf() for the VC6 build. This can be more efficient, but since it's only VC6, and it's just a +logging function, I'm happy to keep this simple. In the VC6 build we can implement this in terms of _vsnprintf(). +*/ +#if defined(_MSC_VER) && _MSC_VER < 1900 +static int ma_vscprintf(const ma_allocation_callbacks* pAllocationCallbacks, const char* format, va_list args) +{ +#if _MSC_VER > 1200 + return _vscprintf(format, args); +#else + int result; + char* pTempBuffer = NULL; + size_t tempBufferCap = 1024; + + if (format == NULL) { + errno = EINVAL; + return -1; + } + + for (;;) { + char* pNewTempBuffer = (char*)ma_realloc(pTempBuffer, tempBufferCap, pAllocationCallbacks); + if (pNewTempBuffer == NULL) { + ma_free(pTempBuffer, pAllocationCallbacks); + errno = ENOMEM; + return -1; /* Out of memory. */ + } + + pTempBuffer = pNewTempBuffer; + + result = _vsnprintf(pTempBuffer, tempBufferCap, format, args); + ma_free(pTempBuffer, NULL); + + if (result != -1) { + break; /* Got it. */ + } + + /* Buffer wasn't big enough. Ideally it'd be nice to use an error code to know the reason for sure, but this is reliable enough. */ + tempBufferCap *= 2; + } + + return result; +#endif +} +#endif + +MA_API ma_result ma_log_postv(ma_log* pLog, ma_uint32 level, const char* pFormat, va_list args) +{ + if (pLog == NULL || pFormat == NULL) { + return MA_INVALID_ARGS; + } + + #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || ((!defined(_MSC_VER) || _MSC_VER >= 1900) && !defined(__STRICT_ANSI__) && !defined(_NO_EXT_KEYS)) || (defined(__cplusplus) && __cplusplus >= 201103L) + { + ma_result result; + int length; + char pFormattedMessageStack[1024]; + char* pFormattedMessageHeap = NULL; + + /* First try formatting into our fixed sized stack allocated buffer. If this is too small we'll fallback to a heap allocation. */ + length = vsnprintf(pFormattedMessageStack, sizeof(pFormattedMessageStack), pFormat, args); + if (length < 0) { + return MA_INVALID_OPERATION; /* An error occurred when trying to convert the buffer. */ + } + + if ((size_t)length < sizeof(pFormattedMessageStack)) { + /* The string was written to the stack. */ + result = ma_log_post(pLog, level, pFormattedMessageStack); + } else { + /* The stack buffer was too small, try the heap. */ + pFormattedMessageHeap = (char*)ma_malloc(length + 1, &pLog->allocationCallbacks); + if (pFormattedMessageHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + + length = vsnprintf(pFormattedMessageHeap, length + 1, pFormat, args); + if (length < 0) { + ma_free(pFormattedMessageHeap, &pLog->allocationCallbacks); + return MA_INVALID_OPERATION; + } + + result = ma_log_post(pLog, level, pFormattedMessageHeap); + ma_free(pFormattedMessageHeap, &pLog->allocationCallbacks); + } + + return result; + } + #else + { + /* + Without snprintf() we need to first measure the string and then heap allocate it. I'm only aware of Visual Studio having support for this without snprintf(), so we'll + need to restrict this branch to Visual Studio. For other compilers we need to just not support formatted logging because I don't want the security risk of overflowing + a fixed sized stack allocated buffer. + */ + #if defined(_MSC_VER) && _MSC_VER >= 1200 /* 1200 = VC6 */ + { + ma_result result; + int formattedLen; + char* pFormattedMessage = NULL; + va_list args2; + + #if _MSC_VER >= 1800 + { + va_copy(args2, args); + } + #else + { + args2 = args; + } + #endif + + formattedLen = ma_vscprintf(&pLog->allocationCallbacks, pFormat, args2); + va_end(args2); + + if (formattedLen <= 0) { + return MA_INVALID_OPERATION; + } + + pFormattedMessage = (char*)ma_malloc(formattedLen + 1, &pLog->allocationCallbacks); + if (pFormattedMessage == NULL) { + return MA_OUT_OF_MEMORY; + } + + /* We'll get errors on newer versions of Visual Studio if we try to use vsprintf(). */ + #if _MSC_VER >= 1400 /* 1400 = Visual Studio 2005 */ + { + vsprintf_s(pFormattedMessage, formattedLen + 1, pFormat, args); + } + #else + { + vsprintf(pFormattedMessage, pFormat, args); + } + #endif + + result = ma_log_post(pLog, level, pFormattedMessage); + ma_free(pFormattedMessage, &pLog->allocationCallbacks); + + return result; + } + #else + { + /* Can't do anything because we don't have a safe way of to emulate vsnprintf() without a manual solution. */ + (void)level; + (void)args; + + return MA_INVALID_OPERATION; + } + #endif + } + #endif +} + +MA_API ma_result ma_log_postf(ma_log* pLog, ma_uint32 level, const char* pFormat, ...) +{ + ma_result result; + va_list args; + + if (pLog == NULL || pFormat == NULL) { + return MA_INVALID_ARGS; + } + + va_start(args, pFormat); + { + result = ma_log_postv(pLog, level, pFormat, args); + } + va_end(args); + + return result; +} + + + +static MA_INLINE ma_uint8 ma_clip_u8(ma_int32 x) +{ + return (ma_uint8)(ma_clamp(x, -128, 127) + 128); +} + +static MA_INLINE ma_int16 ma_clip_s16(ma_int32 x) +{ + return (ma_int16)ma_clamp(x, -32768, 32767); +} + +static MA_INLINE ma_int64 ma_clip_s24(ma_int64 x) +{ + return (ma_int64)ma_clamp(x, -8388608, 8388607); +} + +static MA_INLINE ma_int32 ma_clip_s32(ma_int64 x) +{ + /* This dance is to silence warnings with -std=c89. A good compiler should be able to optimize this away. */ + ma_int64 clipMin; + ma_int64 clipMax; + clipMin = -((ma_int64)2147483647 + 1); + clipMax = (ma_int64)2147483647; + + return (ma_int32)ma_clamp(x, clipMin, clipMax); +} + +static MA_INLINE float ma_clip_f32(float x) +{ + if (x < -1) return -1; + if (x > +1) return +1; + return x; +} + + +static MA_INLINE float ma_mix_f32(float x, float y, float a) +{ + return x*(1-a) + y*a; +} +static MA_INLINE float ma_mix_f32_fast(float x, float y, float a) +{ + float r0 = (y - x); + float r1 = r0*a; + return x + r1; + /*return x + (y - x)*a;*/ +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE __m128 ma_mix_f32_fast__sse2(__m128 x, __m128 y, __m128 a) +{ + return _mm_add_ps(x, _mm_mul_ps(_mm_sub_ps(y, x), a)); +} +#endif +#if defined(MA_SUPPORT_AVX2) +static MA_INLINE __m256 ma_mix_f32_fast__avx2(__m256 x, __m256 y, __m256 a) +{ + return _mm256_add_ps(x, _mm256_mul_ps(_mm256_sub_ps(y, x), a)); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE float32x4_t ma_mix_f32_fast__neon(float32x4_t x, float32x4_t y, float32x4_t a) +{ + return vaddq_f32(x, vmulq_f32(vsubq_f32(y, x), a)); +} +#endif + + +static MA_INLINE double ma_mix_f64(double x, double y, double a) +{ + return x*(1-a) + y*a; +} +static MA_INLINE double ma_mix_f64_fast(double x, double y, double a) +{ + return x + (y - x)*a; +} + +static MA_INLINE float ma_scale_to_range_f32(float x, float lo, float hi) +{ + return lo + x*(hi-lo); +} + + +/* +Greatest common factor using Euclid's algorithm iteratively. +*/ +static MA_INLINE ma_uint32 ma_gcf_u32(ma_uint32 a, ma_uint32 b) +{ + for (;;) { + if (b == 0) { + break; + } else { + ma_uint32 t = a; + a = b; + b = t % a; + } + } + + return a; +} + + +static ma_uint32 ma_ffs_32(ma_uint32 x) +{ + ma_uint32 i; + + /* Just a naive implementation just to get things working for now. Will optimize this later. */ + for (i = 0; i < 32; i += 1) { + if ((x & (1U << i)) != 0) { + return i; + } + } + + return i; +} + +static MA_INLINE ma_int16 ma_float_to_fixed_16(float x) +{ + return (ma_int16)(x * (1 << 8)); +} + + + +/* +Random Number Generation + +miniaudio uses the LCG random number generation algorithm. This is good enough for audio. + +Note that miniaudio's global LCG implementation uses global state which is _not_ thread-local. When this is called across +multiple threads, results will be unpredictable. However, it won't crash and results will still be random enough for +miniaudio's purposes. +*/ +#ifndef MA_DEFAULT_LCG_SEED +#define MA_DEFAULT_LCG_SEED 4321 +#endif + +#define MA_LCG_M 2147483647 +#define MA_LCG_A 48271 +#define MA_LCG_C 0 + +static ma_lcg g_maLCG = {MA_DEFAULT_LCG_SEED}; /* Non-zero initial seed. Use ma_seed() to use an explicit seed. */ + +static MA_INLINE void ma_lcg_seed(ma_lcg* pLCG, ma_int32 seed) +{ + MA_ASSERT(pLCG != NULL); + pLCG->state = seed; +} + +static MA_INLINE ma_int32 ma_lcg_rand_s32(ma_lcg* pLCG) +{ + pLCG->state = (MA_LCG_A * pLCG->state + MA_LCG_C) % MA_LCG_M; + return pLCG->state; +} + +static MA_INLINE ma_uint32 ma_lcg_rand_u32(ma_lcg* pLCG) +{ + return (ma_uint32)ma_lcg_rand_s32(pLCG); +} + +static MA_INLINE ma_int16 ma_lcg_rand_s16(ma_lcg* pLCG) +{ + return (ma_int16)(ma_lcg_rand_s32(pLCG) & 0xFFFF); +} + +static MA_INLINE double ma_lcg_rand_f64(ma_lcg* pLCG) +{ + return ma_lcg_rand_s32(pLCG) / (double)0x7FFFFFFF; +} + +static MA_INLINE float ma_lcg_rand_f32(ma_lcg* pLCG) +{ + return (float)ma_lcg_rand_f64(pLCG); +} + +static MA_INLINE float ma_lcg_rand_range_f32(ma_lcg* pLCG, float lo, float hi) +{ + return ma_scale_to_range_f32(ma_lcg_rand_f32(pLCG), lo, hi); +} + +static MA_INLINE ma_int32 ma_lcg_rand_range_s32(ma_lcg* pLCG, ma_int32 lo, ma_int32 hi) +{ + if (lo == hi) { + return lo; + } + + return lo + ma_lcg_rand_u32(pLCG) / (0xFFFFFFFF / (hi - lo + 1) + 1); +} + + + +static MA_INLINE void ma_seed(ma_int32 seed) +{ + ma_lcg_seed(&g_maLCG, seed); +} + +static MA_INLINE ma_int32 ma_rand_s32(void) +{ + return ma_lcg_rand_s32(&g_maLCG); +} + +static MA_INLINE ma_uint32 ma_rand_u32(void) +{ + return ma_lcg_rand_u32(&g_maLCG); +} + +static MA_INLINE double ma_rand_f64(void) +{ + return ma_lcg_rand_f64(&g_maLCG); +} + +static MA_INLINE float ma_rand_f32(void) +{ + return ma_lcg_rand_f32(&g_maLCG); +} + +static MA_INLINE float ma_rand_range_f32(float lo, float hi) +{ + return ma_lcg_rand_range_f32(&g_maLCG, lo, hi); +} + +static MA_INLINE ma_int32 ma_rand_range_s32(ma_int32 lo, ma_int32 hi) +{ + return ma_lcg_rand_range_s32(&g_maLCG, lo, hi); +} + + +static MA_INLINE float ma_dither_f32_rectangle(float ditherMin, float ditherMax) +{ + return ma_rand_range_f32(ditherMin, ditherMax); +} + +static MA_INLINE float ma_dither_f32_triangle(float ditherMin, float ditherMax) +{ + float a = ma_rand_range_f32(ditherMin, 0); + float b = ma_rand_range_f32(0, ditherMax); + return a + b; +} + +static MA_INLINE float ma_dither_f32(ma_dither_mode ditherMode, float ditherMin, float ditherMax) +{ + if (ditherMode == ma_dither_mode_rectangle) { + return ma_dither_f32_rectangle(ditherMin, ditherMax); + } + if (ditherMode == ma_dither_mode_triangle) { + return ma_dither_f32_triangle(ditherMin, ditherMax); + } + + return 0; +} + +static MA_INLINE ma_int32 ma_dither_s32(ma_dither_mode ditherMode, ma_int32 ditherMin, ma_int32 ditherMax) +{ + if (ditherMode == ma_dither_mode_rectangle) { + ma_int32 a = ma_rand_range_s32(ditherMin, ditherMax); + return a; + } + if (ditherMode == ma_dither_mode_triangle) { + ma_int32 a = ma_rand_range_s32(ditherMin, 0); + ma_int32 b = ma_rand_range_s32(0, ditherMax); + return a + b; + } + + return 0; +} + + +/************************************************************************************************************************************************************** + +Atomics + +**************************************************************************************************************************************************************/ +/* c89atomic.h begin */ +#ifndef ma_atomic_h +#if defined(__cplusplus) +extern "C" { +#endif +#if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wlong-long" + #if defined(__clang__) + #pragma GCC diagnostic ignored "-Wc++11-long-long" + #endif +#endif +typedef int ma_atomic_memory_order; +#define MA_ATOMIC_HAS_8 +#define MA_ATOMIC_HAS_16 +#define MA_ATOMIC_HAS_32 +#define MA_ATOMIC_HAS_64 +#if (defined(_MSC_VER) ) || defined(__WATCOMC__) || defined(__DMC__) + #define MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, intrin, ma_atomicType, msvcType) \ + ma_atomicType result; \ + switch (order) \ + { \ + case ma_atomic_memory_order_relaxed: \ + { \ + result = (ma_atomicType)intrin##_nf((volatile msvcType*)dst, (msvcType)src); \ + } break; \ + case ma_atomic_memory_order_consume: \ + case ma_atomic_memory_order_acquire: \ + { \ + result = (ma_atomicType)intrin##_acq((volatile msvcType*)dst, (msvcType)src); \ + } break; \ + case ma_atomic_memory_order_release: \ + { \ + result = (ma_atomicType)intrin##_rel((volatile msvcType*)dst, (msvcType)src); \ + } break; \ + case ma_atomic_memory_order_acq_rel: \ + case ma_atomic_memory_order_seq_cst: \ + default: \ + { \ + result = (ma_atomicType)intrin((volatile msvcType*)dst, (msvcType)src); \ + } break; \ + } \ + return result; + #define MA_ATOMIC_MSVC_ARM_INTRINSIC_COMPARE_EXCHANGE(ptr, expected, desired, order, intrin, ma_atomicType, msvcType) \ + ma_atomicType result; \ + switch (order) \ + { \ + case ma_atomic_memory_order_relaxed: \ + { \ + result = (ma_atomicType)intrin##_nf((volatile msvcType*)ptr, (msvcType)expected, (msvcType)desired); \ + } break; \ + case ma_atomic_memory_order_consume: \ + case ma_atomic_memory_order_acquire: \ + { \ + result = (ma_atomicType)intrin##_acq((volatile msvcType*)ptr, (msvcType)expected, (msvcType)desired); \ + } break; \ + case ma_atomic_memory_order_release: \ + { \ + result = (ma_atomicType)intrin##_rel((volatile msvcType*)ptr, (msvcType)expected, (msvcType)desired); \ + } break; \ + case ma_atomic_memory_order_acq_rel: \ + case ma_atomic_memory_order_seq_cst: \ + default: \ + { \ + result = (ma_atomicType)intrin((volatile msvcType*)ptr, (msvcType)expected, (msvcType)desired); \ + } break; \ + } \ + return result; + #define ma_atomic_memory_order_relaxed 0 + #define ma_atomic_memory_order_consume 1 + #define ma_atomic_memory_order_acquire 2 + #define ma_atomic_memory_order_release 3 + #define ma_atomic_memory_order_acq_rel 4 + #define ma_atomic_memory_order_seq_cst 5 + #if _MSC_VER < 1600 && defined(MA_X86) + #define MA_ATOMIC_MSVC_USE_INLINED_ASSEMBLY + #endif + #if _MSC_VER < 1600 + #undef MA_ATOMIC_HAS_8 + #undef MA_ATOMIC_HAS_16 + #endif + #if !defined(MA_ATOMIC_MSVC_USE_INLINED_ASSEMBLY) + #include + #endif + #if defined(MA_ATOMIC_MSVC_USE_INLINED_ASSEMBLY) + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_compare_and_swap_8(volatile ma_uint8* dst, ma_uint8 expected, ma_uint8 desired) + { + ma_uint8 result = 0; + __asm { + mov ecx, dst + mov al, expected + mov dl, desired + lock cmpxchg [ecx], dl + mov result, al + } + return result; + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_compare_and_swap_16(volatile ma_uint16* dst, ma_uint16 expected, ma_uint16 desired) + { + ma_uint16 result = 0; + __asm { + mov ecx, dst + mov ax, expected + mov dx, desired + lock cmpxchg [ecx], dx + mov result, ax + } + return result; + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_compare_and_swap_32(volatile ma_uint32* dst, ma_uint32 expected, ma_uint32 desired) + { + ma_uint32 result = 0; + __asm { + mov ecx, dst + mov eax, expected + mov edx, desired + lock cmpxchg [ecx], edx + mov result, eax + } + return result; + } + #endif + #if defined(MA_ATOMIC_HAS_64) + static MA_INLINE ma_uint64 __stdcall ma_atomic_compare_and_swap_64(volatile ma_uint64* dst, ma_uint64 expected, ma_uint64 desired) + { + ma_uint32 resultEAX = 0; + ma_uint32 resultEDX = 0; + __asm { + mov esi, dst + mov eax, dword ptr expected + mov edx, dword ptr expected + 4 + mov ebx, dword ptr desired + mov ecx, dword ptr desired + 4 + lock cmpxchg8b qword ptr [esi] + mov resultEAX, eax + mov resultEDX, edx + } + return ((ma_uint64)resultEDX << 32) | resultEAX; + } + #endif + #else + #if defined(MA_ATOMIC_HAS_8) + #define ma_atomic_compare_and_swap_8( dst, expected, desired) (ma_uint8 )_InterlockedCompareExchange8((volatile char*)dst, (char)desired, (char)expected) + #endif + #if defined(MA_ATOMIC_HAS_16) + #define ma_atomic_compare_and_swap_16(dst, expected, desired) (ma_uint16)_InterlockedCompareExchange16((volatile short*)dst, (short)desired, (short)expected) + #endif + #if defined(MA_ATOMIC_HAS_32) + #define ma_atomic_compare_and_swap_32(dst, expected, desired) (ma_uint32)_InterlockedCompareExchange((volatile long*)dst, (long)desired, (long)expected) + #endif + #if defined(MA_ATOMIC_HAS_64) + #define ma_atomic_compare_and_swap_64(dst, expected, desired) (ma_uint64)_InterlockedCompareExchange64((volatile ma_int64*)dst, (ma_int64)desired, (ma_int64)expected) + #endif + #endif + #if defined(MA_ATOMIC_MSVC_USE_INLINED_ASSEMBLY) + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_exchange_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 result = 0; + (void)order; + __asm { + mov ecx, dst + mov al, src + lock xchg [ecx], al + mov result, al + } + return result; + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_exchange_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 result = 0; + (void)order; + __asm { + mov ecx, dst + mov ax, src + lock xchg [ecx], ax + mov result, ax + } + return result; + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_exchange_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 result = 0; + (void)order; + __asm { + mov ecx, dst + mov eax, src + lock xchg [ecx], eax + mov result, eax + } + return result; + } + #endif + #else + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_exchange_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedExchange8, ma_uint8, char); + #else + (void)order; + return (ma_uint8)_InterlockedExchange8((volatile char*)dst, (char)src); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_exchange_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedExchange16, ma_uint16, short); + #else + (void)order; + return (ma_uint16)_InterlockedExchange16((volatile short*)dst, (short)src); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_exchange_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedExchange, ma_uint32, long); + #else + (void)order; + return (ma_uint32)_InterlockedExchange((volatile long*)dst, (long)src); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_64) && defined(MA_64BIT) + static MA_INLINE ma_uint64 __stdcall ma_atomic_exchange_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedExchange64, ma_uint64, long long); + #else + (void)order; + return (ma_uint64)_InterlockedExchange64((volatile long long*)dst, (long long)src); + #endif + } + #else + #endif + #endif + #if defined(MA_ATOMIC_HAS_64) && !defined(MA_64BIT) + static MA_INLINE ma_uint64 __stdcall ma_atomic_exchange_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 oldValue; + do { + oldValue = *dst; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, src) != oldValue); + (void)order; + return oldValue; + } + #endif + #if defined(MA_ATOMIC_MSVC_USE_INLINED_ASSEMBLY) + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_fetch_add_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 result = 0; + (void)order; + __asm { + mov ecx, dst + mov al, src + lock xadd [ecx], al + mov result, al + } + return result; + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_fetch_add_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 result = 0; + (void)order; + __asm { + mov ecx, dst + mov ax, src + lock xadd [ecx], ax + mov result, ax + } + return result; + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_fetch_add_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 result = 0; + (void)order; + __asm { + mov ecx, dst + mov eax, src + lock xadd [ecx], eax + mov result, eax + } + return result; + } + #endif + #else + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_fetch_add_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedExchangeAdd8, ma_uint8, char); + #else + (void)order; + return (ma_uint8)_InterlockedExchangeAdd8((volatile char*)dst, (char)src); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_fetch_add_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedExchangeAdd16, ma_uint16, short); + #else + (void)order; + return (ma_uint16)_InterlockedExchangeAdd16((volatile short*)dst, (short)src); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_fetch_add_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedExchangeAdd, ma_uint32, long); + #else + (void)order; + return (ma_uint32)_InterlockedExchangeAdd((volatile long*)dst, (long)src); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_64) && defined(MA_64BIT) + static MA_INLINE ma_uint64 __stdcall ma_atomic_fetch_add_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedExchangeAdd64, ma_uint64, long long); + #else + (void)order; + return (ma_uint64)_InterlockedExchangeAdd64((volatile long long*)dst, (long long)src); + #endif + } + #else + #endif + #endif + #if defined(MA_ATOMIC_HAS_64) && !defined(MA_64BIT) + static MA_INLINE ma_uint64 __stdcall ma_atomic_fetch_add_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue + src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + #endif + #if defined(MA_ATOMIC_MSVC_USE_INLINED_ASSEMBLY) + static MA_INLINE void __stdcall ma_atomic_thread_fence(ma_atomic_memory_order order) + { + (void)order; + __asm { + lock add [esp], 0 + } + } + #else + #if defined(MA_X64) + #define ma_atomic_thread_fence(order) __faststorefence(), (void)order + #elif defined(MA_ARM64) + #define ma_atomic_thread_fence(order) __dmb(_ARM64_BARRIER_ISH), (void)order + #else + static MA_INLINE void ma_atomic_thread_fence(ma_atomic_memory_order order) + { + volatile ma_uint32 barrier = 0; + ma_atomic_fetch_add_explicit_32(&barrier, 0, order); + } + #endif + #endif + #define ma_atomic_compiler_fence() ma_atomic_thread_fence(ma_atomic_memory_order_seq_cst) + #define ma_atomic_signal_fence(order) ma_atomic_thread_fence(order) + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 ma_atomic_load_explicit_8(volatile const ma_uint8* ptr, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC_COMPARE_EXCHANGE(ptr, 0, 0, order, _InterlockedCompareExchange8, ma_uint8, char); + #else + (void)order; + return ma_atomic_compare_and_swap_8((volatile ma_uint8*)ptr, 0, 0); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 ma_atomic_load_explicit_16(volatile const ma_uint16* ptr, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC_COMPARE_EXCHANGE(ptr, 0, 0, order, _InterlockedCompareExchange16, ma_uint16, short); + #else + (void)order; + return ma_atomic_compare_and_swap_16((volatile ma_uint16*)ptr, 0, 0); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 ma_atomic_load_explicit_32(volatile const ma_uint32* ptr, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC_COMPARE_EXCHANGE(ptr, 0, 0, order, _InterlockedCompareExchange, ma_uint32, long); + #else + (void)order; + return ma_atomic_compare_and_swap_32((volatile ma_uint32*)ptr, 0, 0); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_64) + static MA_INLINE ma_uint64 ma_atomic_load_explicit_64(volatile const ma_uint64* ptr, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC_COMPARE_EXCHANGE(ptr, 0, 0, order, _InterlockedCompareExchange64, ma_uint64, long long); + #else + (void)order; + return ma_atomic_compare_and_swap_64((volatile ma_uint64*)ptr, 0, 0); + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_8) + #define ma_atomic_store_explicit_8( dst, src, order) (void)ma_atomic_exchange_explicit_8 (dst, src, order) + #endif + #if defined(MA_ATOMIC_HAS_16) + #define ma_atomic_store_explicit_16(dst, src, order) (void)ma_atomic_exchange_explicit_16(dst, src, order) + #endif + #if defined(MA_ATOMIC_HAS_32) + #define ma_atomic_store_explicit_32(dst, src, order) (void)ma_atomic_exchange_explicit_32(dst, src, order) + #endif + #if defined(MA_ATOMIC_HAS_64) + #define ma_atomic_store_explicit_64(dst, src, order) (void)ma_atomic_exchange_explicit_64(dst, src, order) + #endif + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_fetch_sub_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 oldValue; + ma_uint8 newValue; + do { + oldValue = *dst; + newValue = (ma_uint8)(oldValue - src); + } while (ma_atomic_compare_and_swap_8(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_fetch_sub_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 oldValue; + ma_uint16 newValue; + do { + oldValue = *dst; + newValue = (ma_uint16)(oldValue - src); + } while (ma_atomic_compare_and_swap_16(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_fetch_sub_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 oldValue; + ma_uint32 newValue; + do { + oldValue = *dst; + newValue = oldValue - src; + } while (ma_atomic_compare_and_swap_32(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + #endif + #if defined(MA_ATOMIC_HAS_64) + static MA_INLINE ma_uint64 __stdcall ma_atomic_fetch_sub_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue - src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + #endif + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_fetch_and_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedAnd8, ma_uint8, char); + #else + ma_uint8 oldValue; + ma_uint8 newValue; + do { + oldValue = *dst; + newValue = (ma_uint8)(oldValue & src); + } while (ma_atomic_compare_and_swap_8(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_fetch_and_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedAnd16, ma_uint16, short); + #else + ma_uint16 oldValue; + ma_uint16 newValue; + do { + oldValue = *dst; + newValue = (ma_uint16)(oldValue & src); + } while (ma_atomic_compare_and_swap_16(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_fetch_and_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedAnd, ma_uint32, long); + #else + ma_uint32 oldValue; + ma_uint32 newValue; + do { + oldValue = *dst; + newValue = oldValue & src; + } while (ma_atomic_compare_and_swap_32(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_64) + static MA_INLINE ma_uint64 __stdcall ma_atomic_fetch_and_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedAnd64, ma_uint64, long long); + #else + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue & src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_fetch_xor_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedXor8, ma_uint8, char); + #else + ma_uint8 oldValue; + ma_uint8 newValue; + do { + oldValue = *dst; + newValue = (ma_uint8)(oldValue ^ src); + } while (ma_atomic_compare_and_swap_8(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_fetch_xor_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedXor16, ma_uint16, short); + #else + ma_uint16 oldValue; + ma_uint16 newValue; + do { + oldValue = *dst; + newValue = (ma_uint16)(oldValue ^ src); + } while (ma_atomic_compare_and_swap_16(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_fetch_xor_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedXor, ma_uint32, long); + #else + ma_uint32 oldValue; + ma_uint32 newValue; + do { + oldValue = *dst; + newValue = oldValue ^ src; + } while (ma_atomic_compare_and_swap_32(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_64) + static MA_INLINE ma_uint64 __stdcall ma_atomic_fetch_xor_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedXor64, ma_uint64, long long); + #else + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue ^ src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_uint8 __stdcall ma_atomic_fetch_or_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedOr8, ma_uint8, char); + #else + ma_uint8 oldValue; + ma_uint8 newValue; + do { + oldValue = *dst; + newValue = (ma_uint8)(oldValue | src); + } while (ma_atomic_compare_and_swap_8(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_uint16 __stdcall ma_atomic_fetch_or_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedOr16, ma_uint16, short); + #else + ma_uint16 oldValue; + ma_uint16 newValue; + do { + oldValue = *dst; + newValue = (ma_uint16)(oldValue | src); + } while (ma_atomic_compare_and_swap_16(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_uint32 __stdcall ma_atomic_fetch_or_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedOr, ma_uint32, long); + #else + ma_uint32 oldValue; + ma_uint32 newValue; + do { + oldValue = *dst; + newValue = oldValue | src; + } while (ma_atomic_compare_and_swap_32(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_64) + static MA_INLINE ma_uint64 __stdcall ma_atomic_fetch_or_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + #if defined(MA_ARM) + MA_ATOMIC_MSVC_ARM_INTRINSIC(dst, src, order, _InterlockedOr64, ma_uint64, long long); + #else + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue | src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + #endif + } + #endif + #if defined(MA_ATOMIC_HAS_8) + #define ma_atomic_test_and_set_explicit_8( dst, order) ma_atomic_exchange_explicit_8 (dst, 1, order) + #endif + #if defined(MA_ATOMIC_HAS_16) + #define ma_atomic_test_and_set_explicit_16(dst, order) ma_atomic_exchange_explicit_16(dst, 1, order) + #endif + #if defined(MA_ATOMIC_HAS_32) + #define ma_atomic_test_and_set_explicit_32(dst, order) ma_atomic_exchange_explicit_32(dst, 1, order) + #endif + #if defined(MA_ATOMIC_HAS_64) + #define ma_atomic_test_and_set_explicit_64(dst, order) ma_atomic_exchange_explicit_64(dst, 1, order) + #endif + #if defined(MA_ATOMIC_HAS_8) + #define ma_atomic_clear_explicit_8( dst, order) ma_atomic_store_explicit_8 (dst, 0, order) + #endif + #if defined(MA_ATOMIC_HAS_16) + #define ma_atomic_clear_explicit_16(dst, order) ma_atomic_store_explicit_16(dst, 0, order) + #endif + #if defined(MA_ATOMIC_HAS_32) + #define ma_atomic_clear_explicit_32(dst, order) ma_atomic_store_explicit_32(dst, 0, order) + #endif + #if defined(MA_ATOMIC_HAS_64) + #define ma_atomic_clear_explicit_64(dst, order) ma_atomic_store_explicit_64(dst, 0, order) + #endif + #if defined(MA_ATOMIC_HAS_8) + typedef ma_uint8 ma_atomic_flag; + #define ma_atomic_flag_test_and_set_explicit(ptr, order) (ma_bool32)ma_atomic_test_and_set_explicit_8(ptr, order) + #define ma_atomic_flag_clear_explicit(ptr, order) ma_atomic_clear_explicit_8(ptr, order) + #define ma_atomic_flag_load_explicit(ptr, order) ma_atomic_load_explicit_8(ptr, order) + #else + typedef ma_uint32 ma_atomic_flag; + #define ma_atomic_flag_test_and_set_explicit(ptr, order) (ma_bool32)ma_atomic_test_and_set_explicit_32(ptr, order) + #define ma_atomic_flag_clear_explicit(ptr, order) ma_atomic_clear_explicit_32(ptr, order) + #define ma_atomic_flag_load_explicit(ptr, order) ma_atomic_load_explicit_32(ptr, order) + #endif +#elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 7))) + #define MA_ATOMIC_HAS_NATIVE_COMPARE_EXCHANGE + #define MA_ATOMIC_HAS_NATIVE_IS_LOCK_FREE + #define ma_atomic_memory_order_relaxed __ATOMIC_RELAXED + #define ma_atomic_memory_order_consume __ATOMIC_CONSUME + #define ma_atomic_memory_order_acquire __ATOMIC_ACQUIRE + #define ma_atomic_memory_order_release __ATOMIC_RELEASE + #define ma_atomic_memory_order_acq_rel __ATOMIC_ACQ_REL + #define ma_atomic_memory_order_seq_cst __ATOMIC_SEQ_CST + #define ma_atomic_compiler_fence() __asm__ __volatile__("":::"memory") + #define ma_atomic_thread_fence(order) __atomic_thread_fence(order) + #define ma_atomic_signal_fence(order) __atomic_signal_fence(order) + #define ma_atomic_is_lock_free_8(ptr) __atomic_is_lock_free(1, ptr) + #define ma_atomic_is_lock_free_16(ptr) __atomic_is_lock_free(2, ptr) + #define ma_atomic_is_lock_free_32(ptr) __atomic_is_lock_free(4, ptr) + #define ma_atomic_is_lock_free_64(ptr) __atomic_is_lock_free(8, ptr) + #define ma_atomic_test_and_set_explicit_8( dst, order) __atomic_exchange_n(dst, 1, order) + #define ma_atomic_test_and_set_explicit_16(dst, order) __atomic_exchange_n(dst, 1, order) + #define ma_atomic_test_and_set_explicit_32(dst, order) __atomic_exchange_n(dst, 1, order) + #define ma_atomic_test_and_set_explicit_64(dst, order) __atomic_exchange_n(dst, 1, order) + #define ma_atomic_clear_explicit_8( dst, order) __atomic_store_n(dst, 0, order) + #define ma_atomic_clear_explicit_16(dst, order) __atomic_store_n(dst, 0, order) + #define ma_atomic_clear_explicit_32(dst, order) __atomic_store_n(dst, 0, order) + #define ma_atomic_clear_explicit_64(dst, order) __atomic_store_n(dst, 0, order) + #define ma_atomic_store_explicit_8( dst, src, order) __atomic_store_n(dst, src, order) + #define ma_atomic_store_explicit_16(dst, src, order) __atomic_store_n(dst, src, order) + #define ma_atomic_store_explicit_32(dst, src, order) __atomic_store_n(dst, src, order) + #define ma_atomic_store_explicit_64(dst, src, order) __atomic_store_n(dst, src, order) + #define ma_atomic_load_explicit_8( dst, order) __atomic_load_n(dst, order) + #define ma_atomic_load_explicit_16(dst, order) __atomic_load_n(dst, order) + #define ma_atomic_load_explicit_32(dst, order) __atomic_load_n(dst, order) + #define ma_atomic_load_explicit_64(dst, order) __atomic_load_n(dst, order) + #define ma_atomic_exchange_explicit_8( dst, src, order) __atomic_exchange_n(dst, src, order) + #define ma_atomic_exchange_explicit_16(dst, src, order) __atomic_exchange_n(dst, src, order) + #define ma_atomic_exchange_explicit_32(dst, src, order) __atomic_exchange_n(dst, src, order) + #define ma_atomic_exchange_explicit_64(dst, src, order) __atomic_exchange_n(dst, src, order) + #define ma_atomic_compare_exchange_strong_explicit_8( dst, expected, desired, successOrder, failureOrder) __atomic_compare_exchange_n(dst, expected, desired, 0, successOrder, failureOrder) + #define ma_atomic_compare_exchange_strong_explicit_16(dst, expected, desired, successOrder, failureOrder) __atomic_compare_exchange_n(dst, expected, desired, 0, successOrder, failureOrder) + #define ma_atomic_compare_exchange_strong_explicit_32(dst, expected, desired, successOrder, failureOrder) __atomic_compare_exchange_n(dst, expected, desired, 0, successOrder, failureOrder) + #define ma_atomic_compare_exchange_strong_explicit_64(dst, expected, desired, successOrder, failureOrder) __atomic_compare_exchange_n(dst, expected, desired, 0, successOrder, failureOrder) + #define ma_atomic_compare_exchange_weak_explicit_8( dst, expected, desired, successOrder, failureOrder) __atomic_compare_exchange_n(dst, expected, desired, 1, successOrder, failureOrder) + #define ma_atomic_compare_exchange_weak_explicit_16(dst, expected, desired, successOrder, failureOrder) __atomic_compare_exchange_n(dst, expected, desired, 1, successOrder, failureOrder) + #define ma_atomic_compare_exchange_weak_explicit_32(dst, expected, desired, successOrder, failureOrder) __atomic_compare_exchange_n(dst, expected, desired, 1, successOrder, failureOrder) + #define ma_atomic_compare_exchange_weak_explicit_64(dst, expected, desired, successOrder, failureOrder) __atomic_compare_exchange_n(dst, expected, desired, 1, successOrder, failureOrder) + #define ma_atomic_fetch_add_explicit_8( dst, src, order) __atomic_fetch_add(dst, src, order) + #define ma_atomic_fetch_add_explicit_16(dst, src, order) __atomic_fetch_add(dst, src, order) + #define ma_atomic_fetch_add_explicit_32(dst, src, order) __atomic_fetch_add(dst, src, order) + #define ma_atomic_fetch_add_explicit_64(dst, src, order) __atomic_fetch_add(dst, src, order) + #define ma_atomic_fetch_sub_explicit_8( dst, src, order) __atomic_fetch_sub(dst, src, order) + #define ma_atomic_fetch_sub_explicit_16(dst, src, order) __atomic_fetch_sub(dst, src, order) + #define ma_atomic_fetch_sub_explicit_32(dst, src, order) __atomic_fetch_sub(dst, src, order) + #define ma_atomic_fetch_sub_explicit_64(dst, src, order) __atomic_fetch_sub(dst, src, order) + #define ma_atomic_fetch_or_explicit_8( dst, src, order) __atomic_fetch_or(dst, src, order) + #define ma_atomic_fetch_or_explicit_16(dst, src, order) __atomic_fetch_or(dst, src, order) + #define ma_atomic_fetch_or_explicit_32(dst, src, order) __atomic_fetch_or(dst, src, order) + #define ma_atomic_fetch_or_explicit_64(dst, src, order) __atomic_fetch_or(dst, src, order) + #define ma_atomic_fetch_xor_explicit_8( dst, src, order) __atomic_fetch_xor(dst, src, order) + #define ma_atomic_fetch_xor_explicit_16(dst, src, order) __atomic_fetch_xor(dst, src, order) + #define ma_atomic_fetch_xor_explicit_32(dst, src, order) __atomic_fetch_xor(dst, src, order) + #define ma_atomic_fetch_xor_explicit_64(dst, src, order) __atomic_fetch_xor(dst, src, order) + #define ma_atomic_fetch_and_explicit_8( dst, src, order) __atomic_fetch_and(dst, src, order) + #define ma_atomic_fetch_and_explicit_16(dst, src, order) __atomic_fetch_and(dst, src, order) + #define ma_atomic_fetch_and_explicit_32(dst, src, order) __atomic_fetch_and(dst, src, order) + #define ma_atomic_fetch_and_explicit_64(dst, src, order) __atomic_fetch_and(dst, src, order) + static MA_INLINE ma_uint8 ma_atomic_compare_and_swap_8(volatile ma_uint8* dst, ma_uint8 expected, ma_uint8 desired) + { + __atomic_compare_exchange_n(dst, &expected, desired, 0, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST); + return expected; + } + static MA_INLINE ma_uint16 ma_atomic_compare_and_swap_16(volatile ma_uint16* dst, ma_uint16 expected, ma_uint16 desired) + { + __atomic_compare_exchange_n(dst, &expected, desired, 0, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST); + return expected; + } + static MA_INLINE ma_uint32 ma_atomic_compare_and_swap_32(volatile ma_uint32* dst, ma_uint32 expected, ma_uint32 desired) + { + __atomic_compare_exchange_n(dst, &expected, desired, 0, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST); + return expected; + } + #if defined(__clang__) + #pragma clang diagnostic push + #if __clang_major__ >= 8 + #pragma clang diagnostic ignored "-Watomic-alignment" + #endif + #endif + static MA_INLINE ma_uint64 ma_atomic_compare_and_swap_64(volatile ma_uint64* dst, ma_uint64 expected, ma_uint64 desired) + { + __atomic_compare_exchange_n(dst, &expected, desired, 0, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST); + return expected; + } + #if defined(__clang__) + #pragma clang diagnostic pop + #endif + typedef ma_uint8 ma_atomic_flag; + #define ma_atomic_flag_test_and_set_explicit(dst, order) (ma_bool32)__atomic_test_and_set(dst, order) + #define ma_atomic_flag_clear_explicit(dst, order) __atomic_clear(dst, order) + #define ma_atomic_flag_load_explicit(ptr, order) ma_atomic_load_explicit_8(ptr, order) +#else + #define ma_atomic_memory_order_relaxed 1 + #define ma_atomic_memory_order_consume 2 + #define ma_atomic_memory_order_acquire 3 + #define ma_atomic_memory_order_release 4 + #define ma_atomic_memory_order_acq_rel 5 + #define ma_atomic_memory_order_seq_cst 6 + #define ma_atomic_compiler_fence() __asm__ __volatile__("":::"memory") + #if defined(__GNUC__) + #define ma_atomic_thread_fence(order) __sync_synchronize(), (void)order + static MA_INLINE ma_uint8 ma_atomic_exchange_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + if (order > ma_atomic_memory_order_acquire) { + __sync_synchronize(); + } + return __sync_lock_test_and_set(dst, src); + } + static MA_INLINE ma_uint16 ma_atomic_exchange_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 oldValue; + do { + oldValue = *dst; + } while (__sync_val_compare_and_swap(dst, oldValue, src) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint32 ma_atomic_exchange_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 oldValue; + do { + oldValue = *dst; + } while (__sync_val_compare_and_swap(dst, oldValue, src) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint64 ma_atomic_exchange_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 oldValue; + do { + oldValue = *dst; + } while (__sync_val_compare_and_swap(dst, oldValue, src) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint8 ma_atomic_fetch_add_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_add(dst, src); + } + static MA_INLINE ma_uint16 ma_atomic_fetch_add_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_add(dst, src); + } + static MA_INLINE ma_uint32 ma_atomic_fetch_add_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_add(dst, src); + } + static MA_INLINE ma_uint64 ma_atomic_fetch_add_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_add(dst, src); + } + static MA_INLINE ma_uint8 ma_atomic_fetch_sub_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_sub(dst, src); + } + static MA_INLINE ma_uint16 ma_atomic_fetch_sub_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_sub(dst, src); + } + static MA_INLINE ma_uint32 ma_atomic_fetch_sub_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_sub(dst, src); + } + static MA_INLINE ma_uint64 ma_atomic_fetch_sub_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_sub(dst, src); + } + static MA_INLINE ma_uint8 ma_atomic_fetch_or_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_or(dst, src); + } + static MA_INLINE ma_uint16 ma_atomic_fetch_or_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_or(dst, src); + } + static MA_INLINE ma_uint32 ma_atomic_fetch_or_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_or(dst, src); + } + static MA_INLINE ma_uint64 ma_atomic_fetch_or_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_or(dst, src); + } + static MA_INLINE ma_uint8 ma_atomic_fetch_xor_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_xor(dst, src); + } + static MA_INLINE ma_uint16 ma_atomic_fetch_xor_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_xor(dst, src); + } + static MA_INLINE ma_uint32 ma_atomic_fetch_xor_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_xor(dst, src); + } + static MA_INLINE ma_uint64 ma_atomic_fetch_xor_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_xor(dst, src); + } + static MA_INLINE ma_uint8 ma_atomic_fetch_and_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_and(dst, src); + } + static MA_INLINE ma_uint16 ma_atomic_fetch_and_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_and(dst, src); + } + static MA_INLINE ma_uint32 ma_atomic_fetch_and_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_and(dst, src); + } + static MA_INLINE ma_uint64 ma_atomic_fetch_and_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + (void)order; + return __sync_fetch_and_and(dst, src); + } + #define ma_atomic_compare_and_swap_8( dst, expected, desired) __sync_val_compare_and_swap(dst, expected, desired) + #define ma_atomic_compare_and_swap_16(dst, expected, desired) __sync_val_compare_and_swap(dst, expected, desired) + #define ma_atomic_compare_and_swap_32(dst, expected, desired) __sync_val_compare_and_swap(dst, expected, desired) + #define ma_atomic_compare_and_swap_64(dst, expected, desired) __sync_val_compare_and_swap(dst, expected, desired) + #else + #if defined(MA_X86) + #define ma_atomic_thread_fence(order) __asm__ __volatile__("lock; addl $0, (%%esp)" ::: "memory", "cc") + #elif defined(MA_X64) + #define ma_atomic_thread_fence(order) __asm__ __volatile__("lock; addq $0, (%%rsp)" ::: "memory", "cc") + #else + #error Unsupported architecture. Please submit a feature request. + #endif + static MA_INLINE ma_uint8 ma_atomic_compare_and_swap_8(volatile ma_uint8* dst, ma_uint8 expected, ma_uint8 desired) + { + ma_uint8 result; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; cmpxchg %3, %0" : "+m"(*dst), "=a"(result) : "a"(expected), "d"(desired) : "cc"); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint16 ma_atomic_compare_and_swap_16(volatile ma_uint16* dst, ma_uint16 expected, ma_uint16 desired) + { + ma_uint16 result; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; cmpxchg %3, %0" : "+m"(*dst), "=a"(result) : "a"(expected), "d"(desired) : "cc"); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint32 ma_atomic_compare_and_swap_32(volatile ma_uint32* dst, ma_uint32 expected, ma_uint32 desired) + { + ma_uint32 result; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; cmpxchg %3, %0" : "+m"(*dst), "=a"(result) : "a"(expected), "d"(desired) : "cc"); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint64 ma_atomic_compare_and_swap_64(volatile ma_uint64* dst, ma_uint64 expected, ma_uint64 desired) + { + volatile ma_uint64 result; + #if defined(MA_X86) + ma_uint32 resultEAX; + ma_uint32 resultEDX; + __asm__ __volatile__("push %%ebx; xchg %5, %%ebx; lock; cmpxchg8b %0; pop %%ebx" : "+m"(*dst), "=a"(resultEAX), "=d"(resultEDX) : "a"(expected & 0xFFFFFFFF), "d"(expected >> 32), "r"(desired & 0xFFFFFFFF), "c"(desired >> 32) : "cc"); + result = ((ma_uint64)resultEDX << 32) | resultEAX; + #elif defined(MA_X64) + __asm__ __volatile__("lock; cmpxchg %3, %0" : "+m"(*dst), "=a"(result) : "a"(expected), "d"(desired) : "cc"); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint8 ma_atomic_exchange_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 result = 0; + (void)order; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; xchg %1, %0" : "+m"(*dst), "=a"(result) : "a"(src)); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint16 ma_atomic_exchange_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 result = 0; + (void)order; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; xchg %1, %0" : "+m"(*dst), "=a"(result) : "a"(src)); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint32 ma_atomic_exchange_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 result; + (void)order; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; xchg %1, %0" : "+m"(*dst), "=a"(result) : "a"(src)); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint64 ma_atomic_exchange_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 result; + (void)order; + #if defined(MA_X86) + do { + result = *dst; + } while (ma_atomic_compare_and_swap_64(dst, result, src) != result); + #elif defined(MA_X64) + __asm__ __volatile__("lock; xchg %1, %0" : "+m"(*dst), "=a"(result) : "a"(src)); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint8 ma_atomic_fetch_add_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 result; + (void)order; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; xadd %1, %0" : "+m"(*dst), "=a"(result) : "a"(src) : "cc"); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint16 ma_atomic_fetch_add_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 result; + (void)order; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; xadd %1, %0" : "+m"(*dst), "=a"(result) : "a"(src) : "cc"); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint32 ma_atomic_fetch_add_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 result; + (void)order; + #if defined(MA_X86) || defined(MA_X64) + __asm__ __volatile__("lock; xadd %1, %0" : "+m"(*dst), "=a"(result) : "a"(src) : "cc"); + #else + #error Unsupported architecture. Please submit a feature request. + #endif + return result; + } + static MA_INLINE ma_uint64 ma_atomic_fetch_add_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + #if defined(MA_X86) + ma_uint64 oldValue; + ma_uint64 newValue; + (void)order; + do { + oldValue = *dst; + newValue = oldValue + src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + return oldValue; + #elif defined(MA_X64) + ma_uint64 result; + (void)order; + __asm__ __volatile__("lock; xadd %1, %0" : "+m"(*dst), "=a"(result) : "a"(src) : "cc"); + return result; + #endif + } + static MA_INLINE ma_uint8 ma_atomic_fetch_sub_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 oldValue; + ma_uint8 newValue; + do { + oldValue = *dst; + newValue = (ma_uint8)(oldValue - src); + } while (ma_atomic_compare_and_swap_8(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint16 ma_atomic_fetch_sub_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 oldValue; + ma_uint16 newValue; + do { + oldValue = *dst; + newValue = (ma_uint16)(oldValue - src); + } while (ma_atomic_compare_and_swap_16(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint32 ma_atomic_fetch_sub_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 oldValue; + ma_uint32 newValue; + do { + oldValue = *dst; + newValue = oldValue - src; + } while (ma_atomic_compare_and_swap_32(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint64 ma_atomic_fetch_sub_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue - src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint8 ma_atomic_fetch_and_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 oldValue; + ma_uint8 newValue; + do { + oldValue = *dst; + newValue = (ma_uint8)(oldValue & src); + } while (ma_atomic_compare_and_swap_8(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint16 ma_atomic_fetch_and_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 oldValue; + ma_uint16 newValue; + do { + oldValue = *dst; + newValue = (ma_uint16)(oldValue & src); + } while (ma_atomic_compare_and_swap_16(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint32 ma_atomic_fetch_and_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 oldValue; + ma_uint32 newValue; + do { + oldValue = *dst; + newValue = oldValue & src; + } while (ma_atomic_compare_and_swap_32(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint64 ma_atomic_fetch_and_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue & src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint8 ma_atomic_fetch_xor_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 oldValue; + ma_uint8 newValue; + do { + oldValue = *dst; + newValue = (ma_uint8)(oldValue ^ src); + } while (ma_atomic_compare_and_swap_8(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint16 ma_atomic_fetch_xor_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 oldValue; + ma_uint16 newValue; + do { + oldValue = *dst; + newValue = (ma_uint16)(oldValue ^ src); + } while (ma_atomic_compare_and_swap_16(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint32 ma_atomic_fetch_xor_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 oldValue; + ma_uint32 newValue; + do { + oldValue = *dst; + newValue = oldValue ^ src; + } while (ma_atomic_compare_and_swap_32(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint64 ma_atomic_fetch_xor_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue ^ src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint8 ma_atomic_fetch_or_explicit_8(volatile ma_uint8* dst, ma_uint8 src, ma_atomic_memory_order order) + { + ma_uint8 oldValue; + ma_uint8 newValue; + do { + oldValue = *dst; + newValue = (ma_uint8)(oldValue | src); + } while (ma_atomic_compare_and_swap_8(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint16 ma_atomic_fetch_or_explicit_16(volatile ma_uint16* dst, ma_uint16 src, ma_atomic_memory_order order) + { + ma_uint16 oldValue; + ma_uint16 newValue; + do { + oldValue = *dst; + newValue = (ma_uint16)(oldValue | src); + } while (ma_atomic_compare_and_swap_16(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint32 ma_atomic_fetch_or_explicit_32(volatile ma_uint32* dst, ma_uint32 src, ma_atomic_memory_order order) + { + ma_uint32 oldValue; + ma_uint32 newValue; + do { + oldValue = *dst; + newValue = oldValue | src; + } while (ma_atomic_compare_and_swap_32(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + static MA_INLINE ma_uint64 ma_atomic_fetch_or_explicit_64(volatile ma_uint64* dst, ma_uint64 src, ma_atomic_memory_order order) + { + ma_uint64 oldValue; + ma_uint64 newValue; + do { + oldValue = *dst; + newValue = oldValue | src; + } while (ma_atomic_compare_and_swap_64(dst, oldValue, newValue) != oldValue); + (void)order; + return oldValue; + } + #endif + #define ma_atomic_signal_fence(order) ma_atomic_thread_fence(order) + static MA_INLINE ma_uint8 ma_atomic_load_explicit_8(volatile const ma_uint8* ptr, ma_atomic_memory_order order) + { + (void)order; + return ma_atomic_compare_and_swap_8((ma_uint8*)ptr, 0, 0); + } + static MA_INLINE ma_uint16 ma_atomic_load_explicit_16(volatile const ma_uint16* ptr, ma_atomic_memory_order order) + { + (void)order; + return ma_atomic_compare_and_swap_16((ma_uint16*)ptr, 0, 0); + } + static MA_INLINE ma_uint32 ma_atomic_load_explicit_32(volatile const ma_uint32* ptr, ma_atomic_memory_order order) + { + (void)order; + return ma_atomic_compare_and_swap_32((ma_uint32*)ptr, 0, 0); + } + static MA_INLINE ma_uint64 ma_atomic_load_explicit_64(volatile const ma_uint64* ptr, ma_atomic_memory_order order) + { + (void)order; + return ma_atomic_compare_and_swap_64((ma_uint64*)ptr, 0, 0); + } + #define ma_atomic_store_explicit_8( dst, src, order) (void)ma_atomic_exchange_explicit_8 (dst, src, order) + #define ma_atomic_store_explicit_16(dst, src, order) (void)ma_atomic_exchange_explicit_16(dst, src, order) + #define ma_atomic_store_explicit_32(dst, src, order) (void)ma_atomic_exchange_explicit_32(dst, src, order) + #define ma_atomic_store_explicit_64(dst, src, order) (void)ma_atomic_exchange_explicit_64(dst, src, order) + #define ma_atomic_test_and_set_explicit_8( dst, order) ma_atomic_exchange_explicit_8 (dst, 1, order) + #define ma_atomic_test_and_set_explicit_16(dst, order) ma_atomic_exchange_explicit_16(dst, 1, order) + #define ma_atomic_test_and_set_explicit_32(dst, order) ma_atomic_exchange_explicit_32(dst, 1, order) + #define ma_atomic_test_and_set_explicit_64(dst, order) ma_atomic_exchange_explicit_64(dst, 1, order) + #define ma_atomic_clear_explicit_8( dst, order) ma_atomic_store_explicit_8 (dst, 0, order) + #define ma_atomic_clear_explicit_16(dst, order) ma_atomic_store_explicit_16(dst, 0, order) + #define ma_atomic_clear_explicit_32(dst, order) ma_atomic_store_explicit_32(dst, 0, order) + #define ma_atomic_clear_explicit_64(dst, order) ma_atomic_store_explicit_64(dst, 0, order) + typedef ma_uint8 ma_atomic_flag; + #define ma_atomic_flag_test_and_set_explicit(ptr, order) (ma_bool32)ma_atomic_test_and_set_explicit_8(ptr, order) + #define ma_atomic_flag_clear_explicit(ptr, order) ma_atomic_clear_explicit_8(ptr, order) + #define ma_atomic_flag_load_explicit(ptr, order) ma_atomic_load_explicit_8(ptr, order) +#endif +#if !defined(MA_ATOMIC_HAS_NATIVE_COMPARE_EXCHANGE) + #if defined(MA_ATOMIC_HAS_8) + static MA_INLINE ma_bool32 ma_atomic_compare_exchange_strong_explicit_8(volatile ma_uint8* dst, ma_uint8* expected, ma_uint8 desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) + { + ma_uint8 expectedValue; + ma_uint8 result; + (void)successOrder; + (void)failureOrder; + expectedValue = ma_atomic_load_explicit_8(expected, ma_atomic_memory_order_seq_cst); + result = ma_atomic_compare_and_swap_8(dst, expectedValue, desired); + if (result == expectedValue) { + return 1; + } else { + ma_atomic_store_explicit_8(expected, result, failureOrder); + return 0; + } + } + #endif + #if defined(MA_ATOMIC_HAS_16) + static MA_INLINE ma_bool32 ma_atomic_compare_exchange_strong_explicit_16(volatile ma_uint16* dst, ma_uint16* expected, ma_uint16 desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) + { + ma_uint16 expectedValue; + ma_uint16 result; + (void)successOrder; + (void)failureOrder; + expectedValue = ma_atomic_load_explicit_16(expected, ma_atomic_memory_order_seq_cst); + result = ma_atomic_compare_and_swap_16(dst, expectedValue, desired); + if (result == expectedValue) { + return 1; + } else { + ma_atomic_store_explicit_16(expected, result, failureOrder); + return 0; + } + } + #endif + #if defined(MA_ATOMIC_HAS_32) + static MA_INLINE ma_bool32 ma_atomic_compare_exchange_strong_explicit_32(volatile ma_uint32* dst, ma_uint32* expected, ma_uint32 desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) + { + ma_uint32 expectedValue; + ma_uint32 result; + (void)successOrder; + (void)failureOrder; + expectedValue = ma_atomic_load_explicit_32(expected, ma_atomic_memory_order_seq_cst); + result = ma_atomic_compare_and_swap_32(dst, expectedValue, desired); + if (result == expectedValue) { + return 1; + } else { + ma_atomic_store_explicit_32(expected, result, failureOrder); + return 0; + } + } + #endif + #if defined(MA_ATOMIC_HAS_64) + static MA_INLINE ma_bool32 ma_atomic_compare_exchange_strong_explicit_64(volatile ma_uint64* dst, volatile ma_uint64* expected, ma_uint64 desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) + { + ma_uint64 expectedValue; + ma_uint64 result; + (void)successOrder; + (void)failureOrder; + expectedValue = ma_atomic_load_explicit_64(expected, ma_atomic_memory_order_seq_cst); + result = ma_atomic_compare_and_swap_64(dst, expectedValue, desired); + if (result == expectedValue) { + return 1; + } else { + ma_atomic_store_explicit_64(expected, result, failureOrder); + return 0; + } + } + #endif + #define ma_atomic_compare_exchange_weak_explicit_8( dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_strong_explicit_8 (dst, expected, desired, successOrder, failureOrder) + #define ma_atomic_compare_exchange_weak_explicit_16(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_strong_explicit_16(dst, expected, desired, successOrder, failureOrder) + #define ma_atomic_compare_exchange_weak_explicit_32(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_strong_explicit_32(dst, expected, desired, successOrder, failureOrder) + #define ma_atomic_compare_exchange_weak_explicit_64(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_strong_explicit_64(dst, expected, desired, successOrder, failureOrder) +#endif +#if !defined(MA_ATOMIC_HAS_NATIVE_IS_LOCK_FREE) + static MA_INLINE ma_bool32 ma_atomic_is_lock_free_8(volatile void* ptr) + { + (void)ptr; + return 1; + } + static MA_INLINE ma_bool32 ma_atomic_is_lock_free_16(volatile void* ptr) + { + (void)ptr; + return 1; + } + static MA_INLINE ma_bool32 ma_atomic_is_lock_free_32(volatile void* ptr) + { + (void)ptr; + return 1; + } + static MA_INLINE ma_bool32 ma_atomic_is_lock_free_64(volatile void* ptr) + { + (void)ptr; + #if defined(MA_64BIT) + return 1; + #else + #if defined(MA_X86) || defined(MA_X64) + return 1; + #else + return 0; + #endif + #endif + } +#endif +#if defined(MA_64BIT) + static MA_INLINE ma_bool32 ma_atomic_is_lock_free_ptr(volatile void** ptr) + { + return ma_atomic_is_lock_free_64((volatile ma_uint64*)ptr); + } + static MA_INLINE void* ma_atomic_load_explicit_ptr(volatile void** ptr, ma_atomic_memory_order order) + { + return (void*)ma_atomic_load_explicit_64((volatile ma_uint64*)ptr, order); + } + static MA_INLINE void ma_atomic_store_explicit_ptr(volatile void** dst, void* src, ma_atomic_memory_order order) + { + ma_atomic_store_explicit_64((volatile ma_uint64*)dst, (ma_uint64)src, order); + } + static MA_INLINE void* ma_atomic_exchange_explicit_ptr(volatile void** dst, void* src, ma_atomic_memory_order order) + { + return (void*)ma_atomic_exchange_explicit_64((volatile ma_uint64*)dst, (ma_uint64)src, order); + } + static MA_INLINE ma_bool32 ma_atomic_compare_exchange_strong_explicit_ptr(volatile void** dst, void** expected, void* desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) + { + return ma_atomic_compare_exchange_strong_explicit_64((volatile ma_uint64*)dst, (ma_uint64*)expected, (ma_uint64)desired, successOrder, failureOrder); + } + static MA_INLINE ma_bool32 ma_atomic_compare_exchange_weak_explicit_ptr(volatile void** dst, void** expected, void* desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) + { + return ma_atomic_compare_exchange_weak_explicit_64((volatile ma_uint64*)dst, (ma_uint64*)expected, (ma_uint64)desired, successOrder, failureOrder); + } + static MA_INLINE void* ma_atomic_compare_and_swap_ptr(volatile void** dst, void* expected, void* desired) + { + return (void*)ma_atomic_compare_and_swap_64((volatile ma_uint64*)dst, (ma_uint64)expected, (ma_uint64)desired); + } +#elif defined(MA_32BIT) + static MA_INLINE ma_bool32 ma_atomic_is_lock_free_ptr(volatile void** ptr) + { + return ma_atomic_is_lock_free_32((volatile ma_uint32*)ptr); + } + static MA_INLINE void* ma_atomic_load_explicit_ptr(volatile void** ptr, ma_atomic_memory_order order) + { + return (void*)ma_atomic_load_explicit_32((volatile ma_uint32*)ptr, order); + } + static MA_INLINE void ma_atomic_store_explicit_ptr(volatile void** dst, void* src, ma_atomic_memory_order order) + { + ma_atomic_store_explicit_32((volatile ma_uint32*)dst, (ma_uint32)src, order); + } + static MA_INLINE void* ma_atomic_exchange_explicit_ptr(volatile void** dst, void* src, ma_atomic_memory_order order) + { + return (void*)ma_atomic_exchange_explicit_32((volatile ma_uint32*)dst, (ma_uint32)src, order); + } + static MA_INLINE ma_bool32 ma_atomic_compare_exchange_strong_explicit_ptr(volatile void** dst, void** expected, void* desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) + { + return ma_atomic_compare_exchange_strong_explicit_32((volatile ma_uint32*)dst, (ma_uint32*)expected, (ma_uint32)desired, successOrder, failureOrder); + } + static MA_INLINE ma_bool32 ma_atomic_compare_exchange_weak_explicit_ptr(volatile void** dst, void** expected, void* desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) + { + return ma_atomic_compare_exchange_weak_explicit_32((volatile ma_uint32*)dst, (ma_uint32*)expected, (ma_uint32)desired, successOrder, failureOrder); + } + static MA_INLINE void* ma_atomic_compare_and_swap_ptr(volatile void** dst, void* expected, void* desired) + { + return (void*)ma_atomic_compare_and_swap_32((volatile ma_uint32*)dst, (ma_uint32)expected, (ma_uint32)desired); + } +#else + #error Unsupported architecture. +#endif +#define ma_atomic_flag_test_and_set(ptr) ma_atomic_flag_test_and_set_explicit(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_flag_clear(ptr) ma_atomic_flag_clear_explicit(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_ptr(dst, src) ma_atomic_store_explicit_ptr((volatile void**)dst, (void*)src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_ptr(ptr) ma_atomic_load_explicit_ptr((volatile void**)ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_ptr(dst, src) ma_atomic_exchange_explicit_ptr((volatile void**)dst, (void*)src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_ptr(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_ptr((volatile void**)dst, (void**)expected, (void*)desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_ptr(dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_ptr((volatile void**)dst, (void**)expected, (void*)desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_test_and_set_8( ptr) ma_atomic_test_and_set_explicit_8( ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_test_and_set_16(ptr) ma_atomic_test_and_set_explicit_16(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_test_and_set_32(ptr) ma_atomic_test_and_set_explicit_32(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_test_and_set_64(ptr) ma_atomic_test_and_set_explicit_64(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_8( ptr) ma_atomic_clear_explicit_8( ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_16(ptr) ma_atomic_clear_explicit_16(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_32(ptr) ma_atomic_clear_explicit_32(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_64(ptr) ma_atomic_clear_explicit_64(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_8( dst, src) ma_atomic_store_explicit_8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_16(dst, src) ma_atomic_store_explicit_16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_32(dst, src) ma_atomic_store_explicit_32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_64(dst, src) ma_atomic_store_explicit_64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_8( ptr) ma_atomic_load_explicit_8( ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_16(ptr) ma_atomic_load_explicit_16(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_32(ptr) ma_atomic_load_explicit_32(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_64(ptr) ma_atomic_load_explicit_64(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_8( dst, src) ma_atomic_exchange_explicit_8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_16(dst, src) ma_atomic_exchange_explicit_16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_32(dst, src) ma_atomic_exchange_explicit_32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_64(dst, src) ma_atomic_exchange_explicit_64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_8( dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_8( dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_16(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_16(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_32(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_32(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_64(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_64(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_8( dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_8( dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_16( dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_16(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_32( dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_32(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_64( dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_64(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_8( dst, src) ma_atomic_fetch_add_explicit_8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_16(dst, src) ma_atomic_fetch_add_explicit_16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_32(dst, src) ma_atomic_fetch_add_explicit_32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_64(dst, src) ma_atomic_fetch_add_explicit_64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_8( dst, src) ma_atomic_fetch_sub_explicit_8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_16(dst, src) ma_atomic_fetch_sub_explicit_16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_32(dst, src) ma_atomic_fetch_sub_explicit_32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_64(dst, src) ma_atomic_fetch_sub_explicit_64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_8( dst, src) ma_atomic_fetch_or_explicit_8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_16(dst, src) ma_atomic_fetch_or_explicit_16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_32(dst, src) ma_atomic_fetch_or_explicit_32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_64(dst, src) ma_atomic_fetch_or_explicit_64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_8( dst, src) ma_atomic_fetch_xor_explicit_8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_16(dst, src) ma_atomic_fetch_xor_explicit_16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_32(dst, src) ma_atomic_fetch_xor_explicit_32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_64(dst, src) ma_atomic_fetch_xor_explicit_64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_8( dst, src) ma_atomic_fetch_and_explicit_8 (dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_16(dst, src) ma_atomic_fetch_and_explicit_16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_32(dst, src) ma_atomic_fetch_and_explicit_32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_64(dst, src) ma_atomic_fetch_and_explicit_64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_test_and_set_explicit_i8( ptr, order) (ma_int8 )ma_atomic_test_and_set_explicit_8( (ma_uint8* )ptr, order) +#define ma_atomic_test_and_set_explicit_i16(ptr, order) (ma_int16)ma_atomic_test_and_set_explicit_16((ma_uint16*)ptr, order) +#define ma_atomic_test_and_set_explicit_i32(ptr, order) (ma_int32)ma_atomic_test_and_set_explicit_32((ma_uint32*)ptr, order) +#define ma_atomic_test_and_set_explicit_i64(ptr, order) (ma_int64)ma_atomic_test_and_set_explicit_64((ma_uint64*)ptr, order) +#define ma_atomic_clear_explicit_i8( ptr, order) ma_atomic_clear_explicit_8( (ma_uint8* )ptr, order) +#define ma_atomic_clear_explicit_i16(ptr, order) ma_atomic_clear_explicit_16((ma_uint16*)ptr, order) +#define ma_atomic_clear_explicit_i32(ptr, order) ma_atomic_clear_explicit_32((ma_uint32*)ptr, order) +#define ma_atomic_clear_explicit_i64(ptr, order) ma_atomic_clear_explicit_64((ma_uint64*)ptr, order) +#define ma_atomic_store_explicit_i8( dst, src, order) ma_atomic_store_explicit_8( (ma_uint8* )dst, (ma_uint8 )src, order) +#define ma_atomic_store_explicit_i16(dst, src, order) ma_atomic_store_explicit_16((ma_uint16*)dst, (ma_uint16)src, order) +#define ma_atomic_store_explicit_i32(dst, src, order) ma_atomic_store_explicit_32((ma_uint32*)dst, (ma_uint32)src, order) +#define ma_atomic_store_explicit_i64(dst, src, order) ma_atomic_store_explicit_64((ma_uint64*)dst, (ma_uint64)src, order) +#define ma_atomic_load_explicit_i8( ptr, order) (ma_int8 )ma_atomic_load_explicit_8( (ma_uint8* )ptr, order) +#define ma_atomic_load_explicit_i16(ptr, order) (ma_int16)ma_atomic_load_explicit_16((ma_uint16*)ptr, order) +#define ma_atomic_load_explicit_i32(ptr, order) (ma_int32)ma_atomic_load_explicit_32((ma_uint32*)ptr, order) +#define ma_atomic_load_explicit_i64(ptr, order) (ma_int64)ma_atomic_load_explicit_64((ma_uint64*)ptr, order) +#define ma_atomic_exchange_explicit_i8( dst, src, order) (ma_int8 )ma_atomic_exchange_explicit_8 ((ma_uint8* )dst, (ma_uint8 )src, order) +#define ma_atomic_exchange_explicit_i16(dst, src, order) (ma_int16)ma_atomic_exchange_explicit_16((ma_uint16*)dst, (ma_uint16)src, order) +#define ma_atomic_exchange_explicit_i32(dst, src, order) (ma_int32)ma_atomic_exchange_explicit_32((ma_uint32*)dst, (ma_uint32)src, order) +#define ma_atomic_exchange_explicit_i64(dst, src, order) (ma_int64)ma_atomic_exchange_explicit_64((ma_uint64*)dst, (ma_uint64)src, order) +#define ma_atomic_compare_exchange_strong_explicit_i8( dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_strong_explicit_8( (ma_uint8* )dst, (ma_uint8* )expected, (ma_uint8 )desired, successOrder, failureOrder) +#define ma_atomic_compare_exchange_strong_explicit_i16(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_strong_explicit_16((ma_uint16*)dst, (ma_uint16*)expected, (ma_uint16)desired, successOrder, failureOrder) +#define ma_atomic_compare_exchange_strong_explicit_i32(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_strong_explicit_32((ma_uint32*)dst, (ma_uint32*)expected, (ma_uint32)desired, successOrder, failureOrder) +#define ma_atomic_compare_exchange_strong_explicit_i64(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_strong_explicit_64((ma_uint64*)dst, (ma_uint64*)expected, (ma_uint64)desired, successOrder, failureOrder) +#define ma_atomic_compare_exchange_weak_explicit_i8( dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_weak_explicit_8( (ma_uint8* )dst, (ma_uint8* )expected, (ma_uint8 )desired, successOrder, failureOrder) +#define ma_atomic_compare_exchange_weak_explicit_i16(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_weak_explicit_16((ma_uint16*)dst, (ma_uint16*)expected, (ma_uint16)desired, successOrder, failureOrder) +#define ma_atomic_compare_exchange_weak_explicit_i32(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_weak_explicit_32((ma_uint32*)dst, (ma_uint32*)expected, (ma_uint32)desired, successOrder, failureOrder) +#define ma_atomic_compare_exchange_weak_explicit_i64(dst, expected, desired, successOrder, failureOrder) ma_atomic_compare_exchange_weak_explicit_64((ma_uint64*)dst, (ma_uint64*)expected, (ma_uint64)desired, successOrder, failureOrder) +#define ma_atomic_fetch_add_explicit_i8( dst, src, order) (ma_int8 )ma_atomic_fetch_add_explicit_8( (ma_uint8* )dst, (ma_uint8 )src, order) +#define ma_atomic_fetch_add_explicit_i16(dst, src, order) (ma_int16)ma_atomic_fetch_add_explicit_16((ma_uint16*)dst, (ma_uint16)src, order) +#define ma_atomic_fetch_add_explicit_i32(dst, src, order) (ma_int32)ma_atomic_fetch_add_explicit_32((ma_uint32*)dst, (ma_uint32)src, order) +#define ma_atomic_fetch_add_explicit_i64(dst, src, order) (ma_int64)ma_atomic_fetch_add_explicit_64((ma_uint64*)dst, (ma_uint64)src, order) +#define ma_atomic_fetch_sub_explicit_i8( dst, src, order) (ma_int8 )ma_atomic_fetch_sub_explicit_8( (ma_uint8* )dst, (ma_uint8 )src, order) +#define ma_atomic_fetch_sub_explicit_i16(dst, src, order) (ma_int16)ma_atomic_fetch_sub_explicit_16((ma_uint16*)dst, (ma_uint16)src, order) +#define ma_atomic_fetch_sub_explicit_i32(dst, src, order) (ma_int32)ma_atomic_fetch_sub_explicit_32((ma_uint32*)dst, (ma_uint32)src, order) +#define ma_atomic_fetch_sub_explicit_i64(dst, src, order) (ma_int64)ma_atomic_fetch_sub_explicit_64((ma_uint64*)dst, (ma_uint64)src, order) +#define ma_atomic_fetch_or_explicit_i8( dst, src, order) (ma_int8 )ma_atomic_fetch_or_explicit_8( (ma_uint8* )dst, (ma_uint8 )src, order) +#define ma_atomic_fetch_or_explicit_i16(dst, src, order) (ma_int16)ma_atomic_fetch_or_explicit_16((ma_uint16*)dst, (ma_uint16)src, order) +#define ma_atomic_fetch_or_explicit_i32(dst, src, order) (ma_int32)ma_atomic_fetch_or_explicit_32((ma_uint32*)dst, (ma_uint32)src, order) +#define ma_atomic_fetch_or_explicit_i64(dst, src, order) (ma_int64)ma_atomic_fetch_or_explicit_64((ma_uint64*)dst, (ma_uint64)src, order) +#define ma_atomic_fetch_xor_explicit_i8( dst, src, order) (ma_int8 )ma_atomic_fetch_xor_explicit_8( (ma_uint8* )dst, (ma_uint8 )src, order) +#define ma_atomic_fetch_xor_explicit_i16(dst, src, order) (ma_int16)ma_atomic_fetch_xor_explicit_16((ma_uint16*)dst, (ma_uint16)src, order) +#define ma_atomic_fetch_xor_explicit_i32(dst, src, order) (ma_int32)ma_atomic_fetch_xor_explicit_32((ma_uint32*)dst, (ma_uint32)src, order) +#define ma_atomic_fetch_xor_explicit_i64(dst, src, order) (ma_int64)ma_atomic_fetch_xor_explicit_64((ma_uint64*)dst, (ma_uint64)src, order) +#define ma_atomic_fetch_and_explicit_i8( dst, src, order) (ma_int8 )ma_atomic_fetch_and_explicit_8( (ma_uint8* )dst, (ma_uint8 )src, order) +#define ma_atomic_fetch_and_explicit_i16(dst, src, order) (ma_int16)ma_atomic_fetch_and_explicit_16((ma_uint16*)dst, (ma_uint16)src, order) +#define ma_atomic_fetch_and_explicit_i32(dst, src, order) (ma_int32)ma_atomic_fetch_and_explicit_32((ma_uint32*)dst, (ma_uint32)src, order) +#define ma_atomic_fetch_and_explicit_i64(dst, src, order) (ma_int64)ma_atomic_fetch_and_explicit_64((ma_uint64*)dst, (ma_uint64)src, order) +#define ma_atomic_test_and_set_i8( ptr) ma_atomic_test_and_set_explicit_i8( ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_test_and_set_i16(ptr) ma_atomic_test_and_set_explicit_i16(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_test_and_set_i32(ptr) ma_atomic_test_and_set_explicit_i32(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_test_and_set_i64(ptr) ma_atomic_test_and_set_explicit_i64(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_i8( ptr) ma_atomic_clear_explicit_i8( ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_i16(ptr) ma_atomic_clear_explicit_i16(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_i32(ptr) ma_atomic_clear_explicit_i32(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_i64(ptr) ma_atomic_clear_explicit_i64(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_i8( dst, src) ma_atomic_store_explicit_i8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_i16(dst, src) ma_atomic_store_explicit_i16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_i32(dst, src) ma_atomic_store_explicit_i32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_i64(dst, src) ma_atomic_store_explicit_i64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_i8( ptr) ma_atomic_load_explicit_i8( ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_i16(ptr) ma_atomic_load_explicit_i16(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_i32(ptr) ma_atomic_load_explicit_i32(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_i64(ptr) ma_atomic_load_explicit_i64(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_i8( dst, src) ma_atomic_exchange_explicit_i8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_i16(dst, src) ma_atomic_exchange_explicit_i16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_i32(dst, src) ma_atomic_exchange_explicit_i32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_i64(dst, src) ma_atomic_exchange_explicit_i64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_i8( dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_i8( dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_i16(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_i16(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_i32(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_i32(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_i64(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_i64(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_i8( dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_i8( dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_i16(dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_i16(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_i32(dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_i32(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_i64(dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_i64(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_i8( dst, src) ma_atomic_fetch_add_explicit_i8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_i16(dst, src) ma_atomic_fetch_add_explicit_i16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_i32(dst, src) ma_atomic_fetch_add_explicit_i32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_i64(dst, src) ma_atomic_fetch_add_explicit_i64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_i8( dst, src) ma_atomic_fetch_sub_explicit_i8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_i16(dst, src) ma_atomic_fetch_sub_explicit_i16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_i32(dst, src) ma_atomic_fetch_sub_explicit_i32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_i64(dst, src) ma_atomic_fetch_sub_explicit_i64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_i8( dst, src) ma_atomic_fetch_or_explicit_i8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_i16(dst, src) ma_atomic_fetch_or_explicit_i16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_i32(dst, src) ma_atomic_fetch_or_explicit_i32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_i64(dst, src) ma_atomic_fetch_or_explicit_i64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_i8( dst, src) ma_atomic_fetch_xor_explicit_i8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_i16(dst, src) ma_atomic_fetch_xor_explicit_i16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_i32(dst, src) ma_atomic_fetch_xor_explicit_i32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_i64(dst, src) ma_atomic_fetch_xor_explicit_i64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_i8( dst, src) ma_atomic_fetch_and_explicit_i8( dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_i16(dst, src) ma_atomic_fetch_and_explicit_i16(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_i32(dst, src) ma_atomic_fetch_and_explicit_i32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_i64(dst, src) ma_atomic_fetch_and_explicit_i64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_and_swap_i8( dst, expected, dedsired) (ma_int8 )ma_atomic_compare_and_swap_8( (ma_uint8* )dst, (ma_uint8 )expected, (ma_uint8 )dedsired) +#define ma_atomic_compare_and_swap_i16(dst, expected, dedsired) (ma_int16)ma_atomic_compare_and_swap_16((ma_uint16*)dst, (ma_uint16)expected, (ma_uint16)dedsired) +#define ma_atomic_compare_and_swap_i32(dst, expected, dedsired) (ma_int32)ma_atomic_compare_and_swap_32((ma_uint32*)dst, (ma_uint32)expected, (ma_uint32)dedsired) +#define ma_atomic_compare_and_swap_i64(dst, expected, dedsired) (ma_int64)ma_atomic_compare_and_swap_64((ma_uint64*)dst, (ma_uint64)expected, (ma_uint64)dedsired) +typedef union +{ + ma_uint32 i; + float f; +} ma_atomic_if32; +typedef union +{ + ma_uint64 i; + double f; +} ma_atomic_if64; +#define ma_atomic_clear_explicit_f32(ptr, order) ma_atomic_clear_explicit_32((ma_uint32*)ptr, order) +#define ma_atomic_clear_explicit_f64(ptr, order) ma_atomic_clear_explicit_64((ma_uint64*)ptr, order) +static MA_INLINE void ma_atomic_store_explicit_f32(volatile float* dst, float src, ma_atomic_memory_order order) +{ + ma_atomic_if32 x; + x.f = src; + ma_atomic_store_explicit_32((volatile ma_uint32*)dst, x.i, order); +} +static MA_INLINE void ma_atomic_store_explicit_f64(volatile double* dst, double src, ma_atomic_memory_order order) +{ + ma_atomic_if64 x; + x.f = src; + ma_atomic_store_explicit_64((volatile ma_uint64*)dst, x.i, order); +} +static MA_INLINE float ma_atomic_load_explicit_f32(volatile const float* ptr, ma_atomic_memory_order order) +{ + ma_atomic_if32 r; + r.i = ma_atomic_load_explicit_32((volatile const ma_uint32*)ptr, order); + return r.f; +} +static MA_INLINE double ma_atomic_load_explicit_f64(volatile const double* ptr, ma_atomic_memory_order order) +{ + ma_atomic_if64 r; + r.i = ma_atomic_load_explicit_64((volatile const ma_uint64*)ptr, order); + return r.f; +} +static MA_INLINE float ma_atomic_exchange_explicit_f32(volatile float* dst, float src, ma_atomic_memory_order order) +{ + ma_atomic_if32 r; + ma_atomic_if32 x; + x.f = src; + r.i = ma_atomic_exchange_explicit_32((volatile ma_uint32*)dst, x.i, order); + return r.f; +} +static MA_INLINE double ma_atomic_exchange_explicit_f64(volatile double* dst, double src, ma_atomic_memory_order order) +{ + ma_atomic_if64 r; + ma_atomic_if64 x; + x.f = src; + r.i = ma_atomic_exchange_explicit_64((volatile ma_uint64*)dst, x.i, order); + return r.f; +} +static MA_INLINE ma_bool32 ma_atomic_compare_exchange_strong_explicit_f32(volatile float* dst, float* expected, float desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) +{ + ma_atomic_if32 d; + d.f = desired; + return ma_atomic_compare_exchange_strong_explicit_32((volatile ma_uint32*)dst, (ma_uint32*)expected, d.i, successOrder, failureOrder); +} +static MA_INLINE ma_bool32 ma_atomic_compare_exchange_strong_explicit_f64(volatile double* dst, double* expected, double desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) +{ + ma_atomic_if64 d; + d.f = desired; + return ma_atomic_compare_exchange_strong_explicit_64((volatile ma_uint64*)dst, (ma_uint64*)expected, d.i, successOrder, failureOrder); +} +static MA_INLINE ma_bool32 ma_atomic_compare_exchange_weak_explicit_f32(volatile float* dst, float* expected, float desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) +{ + ma_atomic_if32 d; + d.f = desired; + return ma_atomic_compare_exchange_weak_explicit_32((volatile ma_uint32*)dst, (ma_uint32*)expected, d.i, successOrder, failureOrder); +} +static MA_INLINE ma_bool32 ma_atomic_compare_exchange_weak_explicit_f64(volatile double* dst, double* expected, double desired, ma_atomic_memory_order successOrder, ma_atomic_memory_order failureOrder) +{ + ma_atomic_if64 d; + d.f = desired; + return ma_atomic_compare_exchange_weak_explicit_64((volatile ma_uint64*)dst, (ma_uint64*)expected, d.i, successOrder, failureOrder); +} +static MA_INLINE float ma_atomic_fetch_add_explicit_f32(volatile float* dst, float src, ma_atomic_memory_order order) +{ + ma_atomic_if32 r; + ma_atomic_if32 x; + x.f = src; + r.i = ma_atomic_fetch_add_explicit_32((volatile ma_uint32*)dst, x.i, order); + return r.f; +} +static MA_INLINE double ma_atomic_fetch_add_explicit_f64(volatile double* dst, double src, ma_atomic_memory_order order) +{ + ma_atomic_if64 r; + ma_atomic_if64 x; + x.f = src; + r.i = ma_atomic_fetch_add_explicit_64((volatile ma_uint64*)dst, x.i, order); + return r.f; +} +static MA_INLINE float ma_atomic_fetch_sub_explicit_f32(volatile float* dst, float src, ma_atomic_memory_order order) +{ + ma_atomic_if32 r; + ma_atomic_if32 x; + x.f = src; + r.i = ma_atomic_fetch_sub_explicit_32((volatile ma_uint32*)dst, x.i, order); + return r.f; +} +static MA_INLINE double ma_atomic_fetch_sub_explicit_f64(volatile double* dst, double src, ma_atomic_memory_order order) +{ + ma_atomic_if64 r; + ma_atomic_if64 x; + x.f = src; + r.i = ma_atomic_fetch_sub_explicit_64((volatile ma_uint64*)dst, x.i, order); + return r.f; +} +static MA_INLINE float ma_atomic_fetch_or_explicit_f32(volatile float* dst, float src, ma_atomic_memory_order order) +{ + ma_atomic_if32 r; + ma_atomic_if32 x; + x.f = src; + r.i = ma_atomic_fetch_or_explicit_32((volatile ma_uint32*)dst, x.i, order); + return r.f; +} +static MA_INLINE double ma_atomic_fetch_or_explicit_f64(volatile double* dst, double src, ma_atomic_memory_order order) +{ + ma_atomic_if64 r; + ma_atomic_if64 x; + x.f = src; + r.i = ma_atomic_fetch_or_explicit_64((volatile ma_uint64*)dst, x.i, order); + return r.f; +} +static MA_INLINE float ma_atomic_fetch_xor_explicit_f32(volatile float* dst, float src, ma_atomic_memory_order order) +{ + ma_atomic_if32 r; + ma_atomic_if32 x; + x.f = src; + r.i = ma_atomic_fetch_xor_explicit_32((volatile ma_uint32*)dst, x.i, order); + return r.f; +} +static MA_INLINE double ma_atomic_fetch_xor_explicit_f64(volatile double* dst, double src, ma_atomic_memory_order order) +{ + ma_atomic_if64 r; + ma_atomic_if64 x; + x.f = src; + r.i = ma_atomic_fetch_xor_explicit_64((volatile ma_uint64*)dst, x.i, order); + return r.f; +} +static MA_INLINE float ma_atomic_fetch_and_explicit_f32(volatile float* dst, float src, ma_atomic_memory_order order) +{ + ma_atomic_if32 r; + ma_atomic_if32 x; + x.f = src; + r.i = ma_atomic_fetch_and_explicit_32((volatile ma_uint32*)dst, x.i, order); + return r.f; +} +static MA_INLINE double ma_atomic_fetch_and_explicit_f64(volatile double* dst, double src, ma_atomic_memory_order order) +{ + ma_atomic_if64 r; + ma_atomic_if64 x; + x.f = src; + r.i = ma_atomic_fetch_and_explicit_64((volatile ma_uint64*)dst, x.i, order); + return r.f; +} +#define ma_atomic_clear_f32(ptr) (float )ma_atomic_clear_explicit_f32(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_clear_f64(ptr) (double)ma_atomic_clear_explicit_f64(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_f32(dst, src) ma_atomic_store_explicit_f32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_store_f64(dst, src) ma_atomic_store_explicit_f64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_f32(ptr) (float )ma_atomic_load_explicit_f32(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_load_f64(ptr) (double)ma_atomic_load_explicit_f64(ptr, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_f32(dst, src) (float )ma_atomic_exchange_explicit_f32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_exchange_f64(dst, src) (double)ma_atomic_exchange_explicit_f64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_f32(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_f32(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_strong_f64(dst, expected, desired) ma_atomic_compare_exchange_strong_explicit_f64(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_f32(dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_f32(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_compare_exchange_weak_f64(dst, expected, desired) ma_atomic_compare_exchange_weak_explicit_f64(dst, expected, desired, ma_atomic_memory_order_seq_cst, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_f32(dst, src) ma_atomic_fetch_add_explicit_f32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_add_f64(dst, src) ma_atomic_fetch_add_explicit_f64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_f32(dst, src) ma_atomic_fetch_sub_explicit_f32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_sub_f64(dst, src) ma_atomic_fetch_sub_explicit_f64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_f32(dst, src) ma_atomic_fetch_or_explicit_f32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_or_f64(dst, src) ma_atomic_fetch_or_explicit_f64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_f32(dst, src) ma_atomic_fetch_xor_explicit_f32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_xor_f64(dst, src) ma_atomic_fetch_xor_explicit_f64(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_f32(dst, src) ma_atomic_fetch_and_explicit_f32(dst, src, ma_atomic_memory_order_seq_cst) +#define ma_atomic_fetch_and_f64(dst, src) ma_atomic_fetch_and_explicit_f64(dst, src, ma_atomic_memory_order_seq_cst) +static MA_INLINE float ma_atomic_compare_and_swap_f32(volatile float* dst, float expected, float desired) +{ + ma_atomic_if32 r; + ma_atomic_if32 e, d; + e.f = expected; + d.f = desired; + r.i = ma_atomic_compare_and_swap_32((volatile ma_uint32*)dst, e.i, d.i); + return r.f; +} +static MA_INLINE double ma_atomic_compare_and_swap_f64(volatile double* dst, double expected, double desired) +{ + ma_atomic_if64 r; + ma_atomic_if64 e, d; + e.f = expected; + d.f = desired; + r.i = ma_atomic_compare_and_swap_64((volatile ma_uint64*)dst, e.i, d.i); + return r.f; +} +typedef ma_atomic_flag ma_atomic_spinlock; +static MA_INLINE void ma_atomic_spinlock_lock(volatile ma_atomic_spinlock* pSpinlock) +{ + for (;;) { + if (ma_atomic_flag_test_and_set_explicit(pSpinlock, ma_atomic_memory_order_acquire) == 0) { + break; + } + while (ma_atomic_flag_load_explicit(pSpinlock, ma_atomic_memory_order_relaxed) == 1) { + } + } +} +static MA_INLINE void ma_atomic_spinlock_unlock(volatile ma_atomic_spinlock* pSpinlock) +{ + ma_atomic_flag_clear_explicit(pSpinlock, ma_atomic_memory_order_release); +} +#if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic pop +#endif +#if defined(__cplusplus) +} +#endif +#endif +/* c89atomic.h end */ + +#define MA_ATOMIC_SAFE_TYPE_IMPL(c89TypeExtension, type) \ + static MA_INLINE ma_##type ma_atomic_##type##_get(ma_atomic_##type* x) \ + { \ + return (ma_##type)ma_atomic_load_##c89TypeExtension(&x->value); \ + } \ + static MA_INLINE void ma_atomic_##type##_set(ma_atomic_##type* x, ma_##type value) \ + { \ + ma_atomic_store_##c89TypeExtension(&x->value, value); \ + } \ + static MA_INLINE ma_##type ma_atomic_##type##_exchange(ma_atomic_##type* x, ma_##type value) \ + { \ + return (ma_##type)ma_atomic_exchange_##c89TypeExtension(&x->value, value); \ + } \ + static MA_INLINE ma_bool32 ma_atomic_##type##_compare_exchange(ma_atomic_##type* x, ma_##type* expected, ma_##type desired) \ + { \ + return ma_atomic_compare_exchange_weak_##c89TypeExtension(&x->value, expected, desired); \ + } \ + static MA_INLINE ma_##type ma_atomic_##type##_fetch_add(ma_atomic_##type* x, ma_##type y) \ + { \ + return (ma_##type)ma_atomic_fetch_add_##c89TypeExtension(&x->value, y); \ + } \ + static MA_INLINE ma_##type ma_atomic_##type##_fetch_sub(ma_atomic_##type* x, ma_##type y) \ + { \ + return (ma_##type)ma_atomic_fetch_sub_##c89TypeExtension(&x->value, y); \ + } \ + static MA_INLINE ma_##type ma_atomic_##type##_fetch_or(ma_atomic_##type* x, ma_##type y) \ + { \ + return (ma_##type)ma_atomic_fetch_or_##c89TypeExtension(&x->value, y); \ + } \ + static MA_INLINE ma_##type ma_atomic_##type##_fetch_xor(ma_atomic_##type* x, ma_##type y) \ + { \ + return (ma_##type)ma_atomic_fetch_xor_##c89TypeExtension(&x->value, y); \ + } \ + static MA_INLINE ma_##type ma_atomic_##type##_fetch_and(ma_atomic_##type* x, ma_##type y) \ + { \ + return (ma_##type)ma_atomic_fetch_and_##c89TypeExtension(&x->value, y); \ + } \ + static MA_INLINE ma_##type ma_atomic_##type##_compare_and_swap(ma_atomic_##type* x, ma_##type expected, ma_##type desired) \ + { \ + return (ma_##type)ma_atomic_compare_and_swap_##c89TypeExtension(&x->value, expected, desired); \ + } \ + +#define MA_ATOMIC_SAFE_TYPE_IMPL_PTR(type) \ + static MA_INLINE ma_##type* ma_atomic_ptr_##type##_get(ma_atomic_ptr_##type* x) \ + { \ + return ma_atomic_load_ptr((void**)&x->value); \ + } \ + static MA_INLINE void ma_atomic_ptr_##type##_set(ma_atomic_ptr_##type* x, ma_##type* value) \ + { \ + ma_atomic_store_ptr((void**)&x->value, (void*)value); \ + } \ + static MA_INLINE ma_##type* ma_atomic_ptr_##type##_exchange(ma_atomic_ptr_##type* x, ma_##type* value) \ + { \ + return ma_atomic_exchange_ptr((void**)&x->value, (void*)value); \ + } \ + static MA_INLINE ma_bool32 ma_atomic_ptr_##type##_compare_exchange(ma_atomic_ptr_##type* x, ma_##type** expected, ma_##type* desired) \ + { \ + return ma_atomic_compare_exchange_weak_ptr((void**)&x->value, (void*)expected, (void*)desired); \ + } \ + static MA_INLINE ma_##type* ma_atomic_ptr_##type##_compare_and_swap(ma_atomic_ptr_##type* x, ma_##type* expected, ma_##type* desired) \ + { \ + return (ma_##type*)ma_atomic_compare_and_swap_ptr((void**)&x->value, (void*)expected, (void*)desired); \ + } \ + +MA_ATOMIC_SAFE_TYPE_IMPL(32, uint32) +MA_ATOMIC_SAFE_TYPE_IMPL(i32, int32) +MA_ATOMIC_SAFE_TYPE_IMPL(64, uint64) +MA_ATOMIC_SAFE_TYPE_IMPL(f32, float) +MA_ATOMIC_SAFE_TYPE_IMPL(32, bool32) + +#if !defined(MA_NO_DEVICE_IO) +MA_ATOMIC_SAFE_TYPE_IMPL(i32, device_state) +#endif + + +MA_API ma_uint64 ma_calculate_frame_count_after_resampling(ma_uint32 sampleRateOut, ma_uint32 sampleRateIn, ma_uint64 frameCountIn) +{ + /* This is based on the calculation in ma_linear_resampler_get_expected_output_frame_count(). */ + ma_uint64 outputFrameCount; + ma_uint64 preliminaryInputFrameCountFromFrac; + ma_uint64 preliminaryInputFrameCount; + + if (sampleRateIn == 0 || sampleRateOut == 0 || frameCountIn == 0) { + return 0; + } + + if (sampleRateOut == sampleRateIn) { + return frameCountIn; + } + + outputFrameCount = (frameCountIn * sampleRateOut) / sampleRateIn; + + preliminaryInputFrameCountFromFrac = (outputFrameCount * (sampleRateIn / sampleRateOut)) / sampleRateOut; + preliminaryInputFrameCount = (outputFrameCount * (sampleRateIn % sampleRateOut)) + preliminaryInputFrameCountFromFrac; + + if (preliminaryInputFrameCount <= frameCountIn) { + outputFrameCount += 1; + } + + return outputFrameCount; +} + +#ifndef MA_DATA_CONVERTER_STACK_BUFFER_SIZE +#define MA_DATA_CONVERTER_STACK_BUFFER_SIZE 4096 +#endif + + + +#if defined(MA_WIN32) +static ma_result ma_result_from_GetLastError(DWORD error) +{ + switch (error) + { + case ERROR_SUCCESS: return MA_SUCCESS; + case ERROR_PATH_NOT_FOUND: return MA_DOES_NOT_EXIST; + case ERROR_TOO_MANY_OPEN_FILES: return MA_TOO_MANY_OPEN_FILES; + case ERROR_NOT_ENOUGH_MEMORY: return MA_OUT_OF_MEMORY; + case ERROR_DISK_FULL: return MA_NO_SPACE; + case ERROR_HANDLE_EOF: return MA_AT_END; + case ERROR_NEGATIVE_SEEK: return MA_BAD_SEEK; + case ERROR_INVALID_PARAMETER: return MA_INVALID_ARGS; + case ERROR_ACCESS_DENIED: return MA_ACCESS_DENIED; + case ERROR_SEM_TIMEOUT: return MA_TIMEOUT; + case ERROR_FILE_NOT_FOUND: return MA_DOES_NOT_EXIST; + default: break; + } + + return MA_ERROR; +} +#endif /* MA_WIN32 */ + + +/******************************************************************************* + +Threading + +*******************************************************************************/ +static MA_INLINE ma_result ma_spinlock_lock_ex(volatile ma_spinlock* pSpinlock, ma_bool32 yield) +{ + if (pSpinlock == NULL) { + return MA_INVALID_ARGS; + } + + for (;;) { + if (ma_atomic_exchange_explicit_32(pSpinlock, 1, ma_atomic_memory_order_acquire) == 0) { + break; + } + + while (ma_atomic_load_explicit_32(pSpinlock, ma_atomic_memory_order_relaxed) == 1) { + if (yield) { + ma_yield(); + } + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_spinlock_lock(volatile ma_spinlock* pSpinlock) +{ + return ma_spinlock_lock_ex(pSpinlock, MA_TRUE); +} + +MA_API ma_result ma_spinlock_lock_noyield(volatile ma_spinlock* pSpinlock) +{ + return ma_spinlock_lock_ex(pSpinlock, MA_FALSE); +} + +MA_API ma_result ma_spinlock_unlock(volatile ma_spinlock* pSpinlock) +{ + if (pSpinlock == NULL) { + return MA_INVALID_ARGS; + } + + ma_atomic_store_explicit_32(pSpinlock, 0, ma_atomic_memory_order_release); + return MA_SUCCESS; +} + + +#ifndef MA_NO_THREADING +#if defined(MA_POSIX) + #define MA_THREADCALL + typedef void* ma_thread_result; +#elif defined(MA_WIN32) + #define MA_THREADCALL WINAPI + typedef unsigned long ma_thread_result; +#endif + +typedef ma_thread_result (MA_THREADCALL * ma_thread_entry_proc)(void* pData); + +#ifdef MA_POSIX +static ma_result ma_thread_create__posix(ma_thread* pThread, ma_thread_priority priority, size_t stackSize, ma_thread_entry_proc entryProc, void* pData) +{ + int result; + pthread_attr_t* pAttr = NULL; + +#if !defined(__EMSCRIPTEN__) && !defined(__3DS__) + /* Try setting the thread priority. It's not critical if anything fails here. */ + pthread_attr_t attr; + if (pthread_attr_init(&attr) == 0) { + int scheduler = -1; + + /* We successfully initialized our attributes object so we can assign the pointer so it's passed into pthread_create(). */ + pAttr = &attr; + + /* We need to set the scheduler policy. Only do this if the OS supports pthread_attr_setschedpolicy() */ + #if !defined(MA_BEOS) + { + if (priority == ma_thread_priority_idle) { + #ifdef SCHED_IDLE + if (pthread_attr_setschedpolicy(&attr, SCHED_IDLE) == 0) { + scheduler = SCHED_IDLE; + } + #endif + } else if (priority == ma_thread_priority_realtime) { + #ifdef SCHED_FIFO + if (pthread_attr_setschedpolicy(&attr, SCHED_FIFO) == 0) { + scheduler = SCHED_FIFO; + } + #endif + #ifdef MA_LINUX + } else { + scheduler = sched_getscheduler(0); + #endif + } + } + #endif + + if (stackSize > 0) { + pthread_attr_setstacksize(&attr, stackSize); + } + + if (scheduler != -1) { + int priorityMin = sched_get_priority_min(scheduler); + int priorityMax = sched_get_priority_max(scheduler); + int priorityStep = (priorityMax - priorityMin) / 7; /* 7 = number of priorities supported by miniaudio. */ + + struct sched_param sched; + if (pthread_attr_getschedparam(&attr, &sched) == 0) { + if (priority == ma_thread_priority_idle) { + sched.sched_priority = priorityMin; + } else if (priority == ma_thread_priority_realtime) { + #if defined(MA_PTHREAD_REALTIME_THREAD_PRIORITY) + { + sched.sched_priority = MA_PTHREAD_REALTIME_THREAD_PRIORITY; + } + #else + { + sched.sched_priority = priorityMax; + } + #endif + } else { + sched.sched_priority += ((int)priority + 5) * priorityStep; /* +5 because the lowest priority is -5. */ + } + + if (sched.sched_priority < priorityMin) { + sched.sched_priority = priorityMin; + } + if (sched.sched_priority > priorityMax) { + sched.sched_priority = priorityMax; + } + + /* I'm not treating a failure of setting the priority as a critical error so not aborting on failure here. */ + if (pthread_attr_setschedparam(&attr, &sched) == 0) { + #if !defined(MA_ANDROID) || (defined(__ANDROID_API__) && __ANDROID_API__ >= 28) + { + pthread_attr_setinheritsched(&attr, PTHREAD_EXPLICIT_SCHED); + } + #endif + } + } + } + } +#else + /* It's the emscripten build. We'll have a few unused parameters. */ + (void)priority; + (void)stackSize; +#endif + + result = pthread_create((pthread_t*)pThread, pAttr, entryProc, pData); + + /* The thread attributes object is no longer required. */ + if (pAttr != NULL) { + pthread_attr_destroy(pAttr); + } + + if (result != 0) { + return ma_result_from_errno(result); + } + + return MA_SUCCESS; +} + +static void ma_thread_wait__posix(ma_thread* pThread) +{ + pthread_join((pthread_t)*pThread, NULL); +} + + +static ma_result ma_mutex_init__posix(ma_mutex* pMutex) +{ + int result; + + if (pMutex == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pMutex); + + result = pthread_mutex_init((pthread_mutex_t*)pMutex, NULL); + if (result != 0) { + return ma_result_from_errno(result); + } + + return MA_SUCCESS; +} + +static void ma_mutex_uninit__posix(ma_mutex* pMutex) +{ + pthread_mutex_destroy((pthread_mutex_t*)pMutex); +} + +static void ma_mutex_lock__posix(ma_mutex* pMutex) +{ + pthread_mutex_lock((pthread_mutex_t*)pMutex); +} + +static void ma_mutex_unlock__posix(ma_mutex* pMutex) +{ + pthread_mutex_unlock((pthread_mutex_t*)pMutex); +} + + +static ma_result ma_event_init__posix(ma_event* pEvent) +{ + int result; + + result = pthread_mutex_init((pthread_mutex_t*)&pEvent->lock, NULL); + if (result != 0) { + return ma_result_from_errno(result); + } + + result = pthread_cond_init((pthread_cond_t*)&pEvent->cond, NULL); + if (result != 0) { + pthread_mutex_destroy((pthread_mutex_t*)&pEvent->lock); + return ma_result_from_errno(result); + } + + pEvent->value = 0; + return MA_SUCCESS; +} + +static void ma_event_uninit__posix(ma_event* pEvent) +{ + pthread_cond_destroy((pthread_cond_t*)&pEvent->cond); + pthread_mutex_destroy((pthread_mutex_t*)&pEvent->lock); +} + +static ma_result ma_event_wait__posix(ma_event* pEvent) +{ + pthread_mutex_lock((pthread_mutex_t*)&pEvent->lock); + { + while (pEvent->value == 0) { + pthread_cond_wait((pthread_cond_t*)&pEvent->cond, (pthread_mutex_t*)&pEvent->lock); + } + pEvent->value = 0; /* Auto-reset. */ + } + pthread_mutex_unlock((pthread_mutex_t*)&pEvent->lock); + + return MA_SUCCESS; +} + +static ma_result ma_event_signal__posix(ma_event* pEvent) +{ + pthread_mutex_lock((pthread_mutex_t*)&pEvent->lock); + { + pEvent->value = 1; + pthread_cond_signal((pthread_cond_t*)&pEvent->cond); + } + pthread_mutex_unlock((pthread_mutex_t*)&pEvent->lock); + + return MA_SUCCESS; +} + + +static ma_result ma_semaphore_init__posix(int initialValue, ma_semaphore* pSemaphore) +{ + int result; + + if (pSemaphore == NULL) { + return MA_INVALID_ARGS; + } + + pSemaphore->value = initialValue; + + result = pthread_mutex_init((pthread_mutex_t*)&pSemaphore->lock, NULL); + if (result != 0) { + return ma_result_from_errno(result); /* Failed to create mutex. */ + } + + result = pthread_cond_init((pthread_cond_t*)&pSemaphore->cond, NULL); + if (result != 0) { + pthread_mutex_destroy((pthread_mutex_t*)&pSemaphore->lock); + return ma_result_from_errno(result); /* Failed to create condition variable. */ + } + + return MA_SUCCESS; +} + +static void ma_semaphore_uninit__posix(ma_semaphore* pSemaphore) +{ + if (pSemaphore == NULL) { + return; + } + + pthread_cond_destroy((pthread_cond_t*)&pSemaphore->cond); + pthread_mutex_destroy((pthread_mutex_t*)&pSemaphore->lock); +} + +static ma_result ma_semaphore_wait__posix(ma_semaphore* pSemaphore) +{ + if (pSemaphore == NULL) { + return MA_INVALID_ARGS; + } + + pthread_mutex_lock((pthread_mutex_t*)&pSemaphore->lock); + { + /* We need to wait on a condition variable before escaping. We can't return from this function until the semaphore has been signaled. */ + while (pSemaphore->value == 0) { + pthread_cond_wait((pthread_cond_t*)&pSemaphore->cond, (pthread_mutex_t*)&pSemaphore->lock); + } + + pSemaphore->value -= 1; + } + pthread_mutex_unlock((pthread_mutex_t*)&pSemaphore->lock); + + return MA_SUCCESS; +} + +static ma_result ma_semaphore_release__posix(ma_semaphore* pSemaphore) +{ + if (pSemaphore == NULL) { + return MA_INVALID_ARGS; + } + + pthread_mutex_lock((pthread_mutex_t*)&pSemaphore->lock); + { + pSemaphore->value += 1; + pthread_cond_signal((pthread_cond_t*)&pSemaphore->cond); + } + pthread_mutex_unlock((pthread_mutex_t*)&pSemaphore->lock); + + return MA_SUCCESS; +} +#elif defined(MA_WIN32) +static int ma_thread_priority_to_win32(ma_thread_priority priority) +{ + switch (priority) { + case ma_thread_priority_idle: return THREAD_PRIORITY_IDLE; + case ma_thread_priority_lowest: return THREAD_PRIORITY_LOWEST; + case ma_thread_priority_low: return THREAD_PRIORITY_BELOW_NORMAL; + case ma_thread_priority_normal: return THREAD_PRIORITY_NORMAL; + case ma_thread_priority_high: return THREAD_PRIORITY_ABOVE_NORMAL; + case ma_thread_priority_highest: return THREAD_PRIORITY_HIGHEST; + case ma_thread_priority_realtime: return THREAD_PRIORITY_TIME_CRITICAL; + default: return THREAD_PRIORITY_NORMAL; + } +} + +static ma_result ma_thread_create__win32(ma_thread* pThread, ma_thread_priority priority, size_t stackSize, ma_thread_entry_proc entryProc, void* pData) +{ + DWORD threadID; /* Not used. Only used for passing into CreateThread() so it doesn't fail on Windows 98. */ + + *pThread = CreateThread(NULL, stackSize, entryProc, pData, 0, &threadID); + if (*pThread == NULL) { + return ma_result_from_GetLastError(GetLastError()); + } + + SetThreadPriority((HANDLE)*pThread, ma_thread_priority_to_win32(priority)); + + return MA_SUCCESS; +} + +static void ma_thread_wait__win32(ma_thread* pThread) +{ + WaitForSingleObject((HANDLE)*pThread, INFINITE); + CloseHandle((HANDLE)*pThread); +} + + +static ma_result ma_mutex_init__win32(ma_mutex* pMutex) +{ + *pMutex = CreateEventA(NULL, FALSE, TRUE, NULL); + if (*pMutex == NULL) { + return ma_result_from_GetLastError(GetLastError()); + } + + return MA_SUCCESS; +} + +static void ma_mutex_uninit__win32(ma_mutex* pMutex) +{ + CloseHandle((HANDLE)*pMutex); +} + +static void ma_mutex_lock__win32(ma_mutex* pMutex) +{ + WaitForSingleObject((HANDLE)*pMutex, INFINITE); +} + +static void ma_mutex_unlock__win32(ma_mutex* pMutex) +{ + SetEvent((HANDLE)*pMutex); +} + + +static ma_result ma_event_init__win32(ma_event* pEvent) +{ + *pEvent = CreateEventA(NULL, FALSE, FALSE, NULL); + if (*pEvent == NULL) { + return ma_result_from_GetLastError(GetLastError()); + } + + return MA_SUCCESS; +} + +static void ma_event_uninit__win32(ma_event* pEvent) +{ + CloseHandle((HANDLE)*pEvent); +} + +static ma_result ma_event_wait__win32(ma_event* pEvent) +{ + DWORD result = WaitForSingleObject((HANDLE)*pEvent, INFINITE); + if (result == WAIT_OBJECT_0) { + return MA_SUCCESS; + } + + if (result == WAIT_TIMEOUT) { + return MA_TIMEOUT; + } + + return ma_result_from_GetLastError(GetLastError()); +} + +static ma_result ma_event_signal__win32(ma_event* pEvent) +{ + BOOL result = SetEvent((HANDLE)*pEvent); + if (result == 0) { + return ma_result_from_GetLastError(GetLastError()); + } + + return MA_SUCCESS; +} + + +static ma_result ma_semaphore_init__win32(int initialValue, ma_semaphore* pSemaphore) +{ + *pSemaphore = CreateSemaphoreW(NULL, (LONG)initialValue, LONG_MAX, NULL); + if (*pSemaphore == NULL) { + return ma_result_from_GetLastError(GetLastError()); + } + + return MA_SUCCESS; +} + +static void ma_semaphore_uninit__win32(ma_semaphore* pSemaphore) +{ + CloseHandle((HANDLE)*pSemaphore); +} + +static ma_result ma_semaphore_wait__win32(ma_semaphore* pSemaphore) +{ + DWORD result = WaitForSingleObject((HANDLE)*pSemaphore, INFINITE); + if (result == WAIT_OBJECT_0) { + return MA_SUCCESS; + } + + if (result == WAIT_TIMEOUT) { + return MA_TIMEOUT; + } + + return ma_result_from_GetLastError(GetLastError()); +} + +static ma_result ma_semaphore_release__win32(ma_semaphore* pSemaphore) +{ + BOOL result = ReleaseSemaphore((HANDLE)*pSemaphore, 1, NULL); + if (result == 0) { + return ma_result_from_GetLastError(GetLastError()); + } + + return MA_SUCCESS; +} +#endif + +typedef struct +{ + ma_thread_entry_proc entryProc; + void* pData; + ma_allocation_callbacks allocationCallbacks; +} ma_thread_proxy_data; + +static ma_thread_result MA_THREADCALL ma_thread_entry_proxy(void* pData) +{ + ma_thread_proxy_data* pProxyData = (ma_thread_proxy_data*)pData; + ma_thread_entry_proc entryProc; + void* pEntryProcData; + ma_thread_result result; + + #if defined(MA_ON_THREAD_ENTRY) + MA_ON_THREAD_ENTRY + #endif + + entryProc = pProxyData->entryProc; + pEntryProcData = pProxyData->pData; + + /* Free the proxy data before getting into the real thread entry proc. */ + ma_free(pProxyData, &pProxyData->allocationCallbacks); + + result = entryProc(pEntryProcData); + + #if defined(MA_ON_THREAD_EXIT) + MA_ON_THREAD_EXIT + #endif + + return result; +} + +static ma_result ma_thread_create(ma_thread* pThread, ma_thread_priority priority, size_t stackSize, ma_thread_entry_proc entryProc, void* pData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_result result; + ma_thread_proxy_data* pProxyData; + + if (pThread == NULL || entryProc == NULL) { + return MA_INVALID_ARGS; + } + + pProxyData = (ma_thread_proxy_data*)ma_malloc(sizeof(*pProxyData), pAllocationCallbacks); /* Will be freed by the proxy entry proc. */ + if (pProxyData == NULL) { + return MA_OUT_OF_MEMORY; + } + +#if defined(MA_THREAD_DEFAULT_STACK_SIZE) + if (stackSize == 0) { + stackSize = MA_THREAD_DEFAULT_STACK_SIZE; + } +#endif + + pProxyData->entryProc = entryProc; + pProxyData->pData = pData; + ma_allocation_callbacks_init_copy(&pProxyData->allocationCallbacks, pAllocationCallbacks); + +#if defined(MA_POSIX) + result = ma_thread_create__posix(pThread, priority, stackSize, ma_thread_entry_proxy, pProxyData); +#elif defined(MA_WIN32) + result = ma_thread_create__win32(pThread, priority, stackSize, ma_thread_entry_proxy, pProxyData); +#endif + + if (result != MA_SUCCESS) { + ma_free(pProxyData, pAllocationCallbacks); + return result; + } + + return MA_SUCCESS; +} + +static void ma_thread_wait(ma_thread* pThread) +{ + if (pThread == NULL) { + return; + } + +#if defined(MA_POSIX) + ma_thread_wait__posix(pThread); +#elif defined(MA_WIN32) + ma_thread_wait__win32(pThread); +#endif +} + + +MA_API ma_result ma_mutex_init(ma_mutex* pMutex) +{ + if (pMutex == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert so the caller is aware of this bug. */ + return MA_INVALID_ARGS; + } + +#if defined(MA_POSIX) + return ma_mutex_init__posix(pMutex); +#elif defined(MA_WIN32) + return ma_mutex_init__win32(pMutex); +#endif +} + +MA_API void ma_mutex_uninit(ma_mutex* pMutex) +{ + if (pMutex == NULL) { + return; + } + +#if defined(MA_POSIX) + ma_mutex_uninit__posix(pMutex); +#elif defined(MA_WIN32) + ma_mutex_uninit__win32(pMutex); +#endif +} + +MA_API void ma_mutex_lock(ma_mutex* pMutex) +{ + if (pMutex == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert so the caller is aware of this bug. */ + return; + } + +#if defined(MA_POSIX) + ma_mutex_lock__posix(pMutex); +#elif defined(MA_WIN32) + ma_mutex_lock__win32(pMutex); +#endif +} + +MA_API void ma_mutex_unlock(ma_mutex* pMutex) +{ + if (pMutex == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert so the caller is aware of this bug. */ + return; + } + +#if defined(MA_POSIX) + ma_mutex_unlock__posix(pMutex); +#elif defined(MA_WIN32) + ma_mutex_unlock__win32(pMutex); +#endif +} + + +MA_API ma_result ma_event_init(ma_event* pEvent) +{ + if (pEvent == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert so the caller is aware of this bug. */ + return MA_INVALID_ARGS; + } + +#if defined(MA_POSIX) + return ma_event_init__posix(pEvent); +#elif defined(MA_WIN32) + return ma_event_init__win32(pEvent); +#endif +} + +#if 0 +static ma_result ma_event_alloc_and_init(ma_event** ppEvent, ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_result result; + ma_event* pEvent; + + if (ppEvent == NULL) { + return MA_INVALID_ARGS; + } + + *ppEvent = NULL; + + pEvent = ma_malloc(sizeof(*pEvent), pAllocationCallbacks); + if (pEvent == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_event_init(pEvent); + if (result != MA_SUCCESS) { + ma_free(pEvent, pAllocationCallbacks); + return result; + } + + *ppEvent = pEvent; + return result; +} +#endif + +MA_API void ma_event_uninit(ma_event* pEvent) +{ + if (pEvent == NULL) { + return; + } + +#if defined(MA_POSIX) + ma_event_uninit__posix(pEvent); +#elif defined(MA_WIN32) + ma_event_uninit__win32(pEvent); +#endif +} + +#if 0 +static void ma_event_uninit_and_free(ma_event* pEvent, ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pEvent == NULL) { + return; + } + + ma_event_uninit(pEvent); + ma_free(pEvent, pAllocationCallbacks); +} +#endif + +MA_API ma_result ma_event_wait(ma_event* pEvent) +{ + if (pEvent == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert to the caller is aware of this bug. */ + return MA_INVALID_ARGS; + } + +#if defined(MA_POSIX) + return ma_event_wait__posix(pEvent); +#elif defined(MA_WIN32) + return ma_event_wait__win32(pEvent); +#endif +} + +MA_API ma_result ma_event_signal(ma_event* pEvent) +{ + if (pEvent == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert to the caller is aware of this bug. */ + return MA_INVALID_ARGS; + } + +#if defined(MA_POSIX) + return ma_event_signal__posix(pEvent); +#elif defined(MA_WIN32) + return ma_event_signal__win32(pEvent); +#endif +} + + +MA_API ma_result ma_semaphore_init(int initialValue, ma_semaphore* pSemaphore) +{ + if (pSemaphore == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert so the caller is aware of this bug. */ + return MA_INVALID_ARGS; + } + +#if defined(MA_POSIX) + return ma_semaphore_init__posix(initialValue, pSemaphore); +#elif defined(MA_WIN32) + return ma_semaphore_init__win32(initialValue, pSemaphore); +#endif +} + +MA_API void ma_semaphore_uninit(ma_semaphore* pSemaphore) +{ + if (pSemaphore == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert so the caller is aware of this bug. */ + return; + } + +#if defined(MA_POSIX) + ma_semaphore_uninit__posix(pSemaphore); +#elif defined(MA_WIN32) + ma_semaphore_uninit__win32(pSemaphore); +#endif +} + +MA_API ma_result ma_semaphore_wait(ma_semaphore* pSemaphore) +{ + if (pSemaphore == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert so the caller is aware of this bug. */ + return MA_INVALID_ARGS; + } + +#if defined(MA_POSIX) + return ma_semaphore_wait__posix(pSemaphore); +#elif defined(MA_WIN32) + return ma_semaphore_wait__win32(pSemaphore); +#endif +} + +MA_API ma_result ma_semaphore_release(ma_semaphore* pSemaphore) +{ + if (pSemaphore == NULL) { + MA_ASSERT(MA_FALSE); /* Fire an assert so the caller is aware of this bug. */ + return MA_INVALID_ARGS; + } + +#if defined(MA_POSIX) + return ma_semaphore_release__posix(pSemaphore); +#elif defined(MA_WIN32) + return ma_semaphore_release__win32(pSemaphore); +#endif +} +#else +/* MA_NO_THREADING is set which means threading is disabled. Threading is required by some API families. If any of these are enabled we need to throw an error. */ +#ifndef MA_NO_DEVICE_IO +#error "MA_NO_THREADING cannot be used without MA_NO_DEVICE_IO"; +#endif +#endif /* MA_NO_THREADING */ + + + +#define MA_FENCE_COUNTER_MAX 0x7FFFFFFF + +MA_API ma_result ma_fence_init(ma_fence* pFence) +{ + if (pFence == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pFence); + pFence->counter = 0; + + #ifndef MA_NO_THREADING + { + ma_result result; + + result = ma_event_init(&pFence->e); + if (result != MA_SUCCESS) { + return result; + } + } + #endif + + return MA_SUCCESS; +} + +MA_API void ma_fence_uninit(ma_fence* pFence) +{ + if (pFence == NULL) { + return; + } + + #ifndef MA_NO_THREADING + { + ma_event_uninit(&pFence->e); + } + #endif + + MA_ZERO_OBJECT(pFence); +} + +MA_API ma_result ma_fence_acquire(ma_fence* pFence) +{ + if (pFence == NULL) { + return MA_INVALID_ARGS; + } + + for (;;) { + ma_uint32 oldCounter = ma_atomic_load_32(&pFence->counter); + ma_uint32 newCounter = oldCounter + 1; + + /* Make sure we're not about to exceed our maximum value. */ + if (newCounter > MA_FENCE_COUNTER_MAX) { + MA_ASSERT(MA_FALSE); + return MA_OUT_OF_RANGE; + } + + if (ma_atomic_compare_exchange_weak_32(&pFence->counter, &oldCounter, newCounter)) { + return MA_SUCCESS; + } else { + if (oldCounter == MA_FENCE_COUNTER_MAX) { + MA_ASSERT(MA_FALSE); + return MA_OUT_OF_RANGE; /* The other thread took the last available slot. Abort. */ + } + } + } + + /* Should never get here. */ + /*return MA_SUCCESS;*/ +} + +MA_API ma_result ma_fence_release(ma_fence* pFence) +{ + if (pFence == NULL) { + return MA_INVALID_ARGS; + } + + for (;;) { + ma_uint32 oldCounter = ma_atomic_load_32(&pFence->counter); + ma_uint32 newCounter = oldCounter - 1; + + if (oldCounter == 0) { + MA_ASSERT(MA_FALSE); + return MA_INVALID_OPERATION; /* Acquire/release mismatch. */ + } + + if (ma_atomic_compare_exchange_weak_32(&pFence->counter, &oldCounter, newCounter)) { + #ifndef MA_NO_THREADING + { + if (newCounter == 0) { + ma_event_signal(&pFence->e); /* <-- ma_fence_wait() will be waiting on this. */ + } + } + #endif + + return MA_SUCCESS; + } else { + if (oldCounter == 0) { + MA_ASSERT(MA_FALSE); + return MA_INVALID_OPERATION; /* Another thread has taken the 0 slot. Acquire/release mismatch. */ + } + } + } + + /* Should never get here. */ + /*return MA_SUCCESS;*/ +} + +MA_API ma_result ma_fence_wait(ma_fence* pFence) +{ + if (pFence == NULL) { + return MA_INVALID_ARGS; + } + + for (;;) { + ma_uint32 counter; + + counter = ma_atomic_load_32(&pFence->counter); + if (counter == 0) { + /* + Counter has hit zero. By the time we get here some other thread may have acquired the + fence again, but that is where the caller needs to take care with how they se the fence. + */ + return MA_SUCCESS; + } + + /* Getting here means the counter is > 0. We'll need to wait for something to happen. */ + #ifndef MA_NO_THREADING + { + ma_result result; + + result = ma_event_wait(&pFence->e); + if (result != MA_SUCCESS) { + return result; + } + } + #endif + } + + /* Should never get here. */ + /*return MA_INVALID_OPERATION;*/ +} + + +MA_API ma_result ma_async_notification_signal(ma_async_notification* pNotification) +{ + ma_async_notification_callbacks* pNotificationCallbacks = (ma_async_notification_callbacks*)pNotification; + + if (pNotification == NULL) { + return MA_INVALID_ARGS; + } + + if (pNotificationCallbacks->onSignal == NULL) { + return MA_NOT_IMPLEMENTED; + } + + pNotificationCallbacks->onSignal(pNotification); + return MA_INVALID_ARGS; +} + + +static void ma_async_notification_poll__on_signal(ma_async_notification* pNotification) +{ + ((ma_async_notification_poll*)pNotification)->signalled = MA_TRUE; +} + +MA_API ma_result ma_async_notification_poll_init(ma_async_notification_poll* pNotificationPoll) +{ + if (pNotificationPoll == NULL) { + return MA_INVALID_ARGS; + } + + pNotificationPoll->cb.onSignal = ma_async_notification_poll__on_signal; + pNotificationPoll->signalled = MA_FALSE; + + return MA_SUCCESS; +} + +MA_API ma_bool32 ma_async_notification_poll_is_signalled(const ma_async_notification_poll* pNotificationPoll) +{ + if (pNotificationPoll == NULL) { + return MA_FALSE; + } + + return pNotificationPoll->signalled; +} + + +static void ma_async_notification_event__on_signal(ma_async_notification* pNotification) +{ + ma_async_notification_event_signal((ma_async_notification_event*)pNotification); +} + +MA_API ma_result ma_async_notification_event_init(ma_async_notification_event* pNotificationEvent) +{ + if (pNotificationEvent == NULL) { + return MA_INVALID_ARGS; + } + + pNotificationEvent->cb.onSignal = ma_async_notification_event__on_signal; + + #ifndef MA_NO_THREADING + { + ma_result result; + + result = ma_event_init(&pNotificationEvent->e); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; + } + #else + { + return MA_NOT_IMPLEMENTED; /* Threading is disabled. */ + } + #endif +} + +MA_API ma_result ma_async_notification_event_uninit(ma_async_notification_event* pNotificationEvent) +{ + if (pNotificationEvent == NULL) { + return MA_INVALID_ARGS; + } + + #ifndef MA_NO_THREADING + { + ma_event_uninit(&pNotificationEvent->e); + return MA_SUCCESS; + } + #else + { + return MA_NOT_IMPLEMENTED; /* Threading is disabled. */ + } + #endif +} + +MA_API ma_result ma_async_notification_event_wait(ma_async_notification_event* pNotificationEvent) +{ + if (pNotificationEvent == NULL) { + return MA_INVALID_ARGS; + } + + #ifndef MA_NO_THREADING + { + return ma_event_wait(&pNotificationEvent->e); + } + #else + { + return MA_NOT_IMPLEMENTED; /* Threading is disabled. */ + } + #endif +} + +MA_API ma_result ma_async_notification_event_signal(ma_async_notification_event* pNotificationEvent) +{ + if (pNotificationEvent == NULL) { + return MA_INVALID_ARGS; + } + + #ifndef MA_NO_THREADING + { + return ma_event_signal(&pNotificationEvent->e); + } + #else + { + return MA_NOT_IMPLEMENTED; /* Threading is disabled. */ + } + #endif +} + + + +/************************************************************************************************************************************************************ + +Job Queue + +************************************************************************************************************************************************************/ +MA_API ma_slot_allocator_config ma_slot_allocator_config_init(ma_uint32 capacity) +{ + ma_slot_allocator_config config; + + MA_ZERO_OBJECT(&config); + config.capacity = capacity; + + return config; +} + + +static MA_INLINE ma_uint32 ma_slot_allocator_calculate_group_capacity(ma_uint32 slotCapacity) +{ + ma_uint32 cap = slotCapacity / 32; + if ((slotCapacity % 32) != 0) { + cap += 1; + } + + return cap; +} + +static MA_INLINE ma_uint32 ma_slot_allocator_group_capacity(const ma_slot_allocator* pAllocator) +{ + return ma_slot_allocator_calculate_group_capacity(pAllocator->capacity); +} + + +typedef struct +{ + size_t sizeInBytes; + size_t groupsOffset; + size_t slotsOffset; +} ma_slot_allocator_heap_layout; + +static ma_result ma_slot_allocator_get_heap_layout(const ma_slot_allocator_config* pConfig, ma_slot_allocator_heap_layout* pHeapLayout) +{ + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->capacity == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* Groups. */ + pHeapLayout->groupsOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(ma_slot_allocator_calculate_group_capacity(pConfig->capacity) * sizeof(ma_slot_allocator_group)); + + /* Slots. */ + pHeapLayout->slotsOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(pConfig->capacity * sizeof(ma_uint32)); + + return MA_SUCCESS; +} + +MA_API ma_result ma_slot_allocator_get_heap_size(const ma_slot_allocator_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_slot_allocator_heap_layout layout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_slot_allocator_get_heap_layout(pConfig, &layout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = layout.sizeInBytes; + + return result; +} + +MA_API ma_result ma_slot_allocator_init_preallocated(const ma_slot_allocator_config* pConfig, void* pHeap, ma_slot_allocator* pAllocator) +{ + ma_result result; + ma_slot_allocator_heap_layout heapLayout; + + if (pAllocator == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pAllocator); + + if (pHeap == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_slot_allocator_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pAllocator->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pAllocator->pGroups = (ma_slot_allocator_group*)ma_offset_ptr(pHeap, heapLayout.groupsOffset); + pAllocator->pSlots = (ma_uint32*)ma_offset_ptr(pHeap, heapLayout.slotsOffset); + pAllocator->capacity = pConfig->capacity; + + return MA_SUCCESS; +} + +MA_API ma_result ma_slot_allocator_init(const ma_slot_allocator_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_slot_allocator* pAllocator) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_slot_allocator_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the size of the heap allocation. */ + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_slot_allocator_init_preallocated(pConfig, pHeap, pAllocator); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pAllocator->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_slot_allocator_uninit(ma_slot_allocator* pAllocator, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocator == NULL) { + return; + } + + if (pAllocator->_ownsHeap) { + ma_free(pAllocator->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_slot_allocator_alloc(ma_slot_allocator* pAllocator, ma_uint64* pSlot) +{ + ma_uint32 iAttempt; + const ma_uint32 maxAttempts = 2; /* The number of iterations to perform until returning MA_OUT_OF_MEMORY if no slots can be found. */ + + if (pAllocator == NULL || pSlot == NULL) { + return MA_INVALID_ARGS; + } + + for (iAttempt = 0; iAttempt < maxAttempts; iAttempt += 1) { + /* We need to acquire a suitable bitfield first. This is a bitfield that's got an available slot within it. */ + ma_uint32 iGroup; + for (iGroup = 0; iGroup < ma_slot_allocator_group_capacity(pAllocator); iGroup += 1) { + /* CAS */ + for (;;) { + ma_uint32 oldBitfield; + ma_uint32 newBitfield; + ma_uint32 bitOffset; + + oldBitfield = ma_atomic_load_32(&pAllocator->pGroups[iGroup].bitfield); /* <-- This copy must happen. The compiler must not optimize this away. */ + + /* Fast check to see if anything is available. */ + if (oldBitfield == 0xFFFFFFFF) { + break; /* No available bits in this bitfield. */ + } + + bitOffset = ma_ffs_32(~oldBitfield); + MA_ASSERT(bitOffset < 32); + + newBitfield = oldBitfield | (1 << bitOffset); + + if (ma_atomic_compare_and_swap_32(&pAllocator->pGroups[iGroup].bitfield, oldBitfield, newBitfield) == oldBitfield) { + ma_uint32 slotIndex; + + /* Increment the counter as soon as possible to have other threads report out-of-memory sooner than later. */ + ma_atomic_fetch_add_32(&pAllocator->count, 1); + + /* The slot index is required for constructing the output value. */ + slotIndex = (iGroup << 5) + bitOffset; /* iGroup << 5 = iGroup * 32 */ + if (slotIndex >= pAllocator->capacity) { + return MA_OUT_OF_MEMORY; + } + + /* Increment the reference count before constructing the output value. */ + pAllocator->pSlots[slotIndex] += 1; + + /* Construct the output value. */ + *pSlot = (((ma_uint64)pAllocator->pSlots[slotIndex] << 32) | slotIndex); + + return MA_SUCCESS; + } + } + } + + /* We weren't able to find a slot. If it's because we've reached our capacity we need to return MA_OUT_OF_MEMORY. Otherwise we need to do another iteration and try again. */ + if (pAllocator->count < pAllocator->capacity) { + ma_yield(); + } else { + return MA_OUT_OF_MEMORY; + } + } + + /* We couldn't find a slot within the maximum number of attempts. */ + return MA_OUT_OF_MEMORY; +} + +MA_API ma_result ma_slot_allocator_free(ma_slot_allocator* pAllocator, ma_uint64 slot) +{ + ma_uint32 iGroup; + ma_uint32 iBit; + + if (pAllocator == NULL) { + return MA_INVALID_ARGS; + } + + iGroup = (ma_uint32)((slot & 0xFFFFFFFF) >> 5); /* slot / 32 */ + iBit = (ma_uint32)((slot & 0xFFFFFFFF) & 31); /* slot % 32 */ + + if (iGroup >= ma_slot_allocator_group_capacity(pAllocator)) { + return MA_INVALID_ARGS; + } + + MA_ASSERT(iBit < 32); /* This must be true due to the logic we used to actually calculate it. */ + + while (ma_atomic_load_32(&pAllocator->count) > 0) { + /* CAS */ + ma_uint32 oldBitfield; + ma_uint32 newBitfield; + + oldBitfield = ma_atomic_load_32(&pAllocator->pGroups[iGroup].bitfield); /* <-- This copy must happen. The compiler must not optimize this away. */ + newBitfield = oldBitfield & ~(1 << iBit); + + /* Debugging for checking for double-frees. */ + #if defined(MA_DEBUG_OUTPUT) + { + if ((oldBitfield & (1 << iBit)) == 0) { + MA_ASSERT(MA_FALSE); /* Double free detected.*/ + } + } + #endif + + if (ma_atomic_compare_and_swap_32(&pAllocator->pGroups[iGroup].bitfield, oldBitfield, newBitfield) == oldBitfield) { + ma_atomic_fetch_sub_32(&pAllocator->count, 1); + return MA_SUCCESS; + } + } + + /* Getting here means there are no allocations available for freeing. */ + return MA_INVALID_OPERATION; +} + + +#define MA_JOB_ID_NONE ~((ma_uint64)0) +#define MA_JOB_SLOT_NONE (ma_uint16)(~0) + +static MA_INLINE ma_uint32 ma_job_extract_refcount(ma_uint64 toc) +{ + return (ma_uint32)(toc >> 32); +} + +static MA_INLINE ma_uint16 ma_job_extract_slot(ma_uint64 toc) +{ + return (ma_uint16)(toc & 0x0000FFFF); +} + +static MA_INLINE ma_uint16 ma_job_extract_code(ma_uint64 toc) +{ + return (ma_uint16)((toc & 0xFFFF0000) >> 16); +} + +static MA_INLINE ma_uint64 ma_job_toc_to_allocation(ma_uint64 toc) +{ + return ((ma_uint64)ma_job_extract_refcount(toc) << 32) | (ma_uint64)ma_job_extract_slot(toc); +} + +static MA_INLINE ma_uint64 ma_job_set_refcount(ma_uint64 toc, ma_uint32 refcount) +{ + /* Clear the reference count first. */ + toc = toc & ~((ma_uint64)0xFFFFFFFF << 32); + toc = toc | ((ma_uint64)refcount << 32); + + return toc; +} + + +MA_API ma_job ma_job_init(ma_uint16 code) +{ + ma_job job; + + MA_ZERO_OBJECT(&job); + job.toc.breakup.code = code; + job.toc.breakup.slot = MA_JOB_SLOT_NONE; /* Temp value. Will be allocated when posted to a queue. */ + job.next = MA_JOB_ID_NONE; + + return job; +} + + +static ma_result ma_job_process__noop(ma_job* pJob); +static ma_result ma_job_process__quit(ma_job* pJob); +static ma_result ma_job_process__custom(ma_job* pJob); +static ma_result ma_job_process__resource_manager__load_data_buffer_node(ma_job* pJob); +static ma_result ma_job_process__resource_manager__free_data_buffer_node(ma_job* pJob); +static ma_result ma_job_process__resource_manager__page_data_buffer_node(ma_job* pJob); +static ma_result ma_job_process__resource_manager__load_data_buffer(ma_job* pJob); +static ma_result ma_job_process__resource_manager__free_data_buffer(ma_job* pJob); +static ma_result ma_job_process__resource_manager__load_data_stream(ma_job* pJob); +static ma_result ma_job_process__resource_manager__free_data_stream(ma_job* pJob); +static ma_result ma_job_process__resource_manager__page_data_stream(ma_job* pJob); +static ma_result ma_job_process__resource_manager__seek_data_stream(ma_job* pJob); + +#if !defined(MA_NO_DEVICE_IO) +static ma_result ma_job_process__device__aaudio_reroute(ma_job* pJob); +#endif + +static ma_job_proc g_jobVTable[MA_JOB_TYPE_COUNT] = +{ + /* Miscellaneous. */ + ma_job_process__quit, /* MA_JOB_TYPE_QUIT */ + ma_job_process__custom, /* MA_JOB_TYPE_CUSTOM */ + + /* Resource Manager. */ + ma_job_process__resource_manager__load_data_buffer_node, /* MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER_NODE */ + ma_job_process__resource_manager__free_data_buffer_node, /* MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER_NODE */ + ma_job_process__resource_manager__page_data_buffer_node, /* MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_BUFFER_NODE */ + ma_job_process__resource_manager__load_data_buffer, /* MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER */ + ma_job_process__resource_manager__free_data_buffer, /* MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER */ + ma_job_process__resource_manager__load_data_stream, /* MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_STREAM */ + ma_job_process__resource_manager__free_data_stream, /* MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_STREAM */ + ma_job_process__resource_manager__page_data_stream, /* MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_STREAM */ + ma_job_process__resource_manager__seek_data_stream, /* MA_JOB_TYPE_RESOURCE_MANAGER_SEEK_DATA_STREAM */ + + /* Device. */ +#if !defined(MA_NO_DEVICE_IO) + ma_job_process__device__aaudio_reroute /* MA_JOB_TYPE_DEVICE_AAUDIO_REROUTE */ +#endif +}; + +MA_API ma_result ma_job_process(ma_job* pJob) +{ + if (pJob == NULL) { + return MA_INVALID_ARGS; + } + + if (pJob->toc.breakup.code >= MA_JOB_TYPE_COUNT) { + return MA_INVALID_OPERATION; + } + + return g_jobVTable[pJob->toc.breakup.code](pJob); +} + +static ma_result ma_job_process__noop(ma_job* pJob) +{ + MA_ASSERT(pJob != NULL); + + /* No-op. */ + (void)pJob; + + return MA_SUCCESS; +} + +static ma_result ma_job_process__quit(ma_job* pJob) +{ + return ma_job_process__noop(pJob); +} + +static ma_result ma_job_process__custom(ma_job* pJob) +{ + MA_ASSERT(pJob != NULL); + + /* No-op if there's no callback. */ + if (pJob->data.custom.proc == NULL) { + return MA_SUCCESS; + } + + return pJob->data.custom.proc(pJob); +} + + + +MA_API ma_job_queue_config ma_job_queue_config_init(ma_uint32 flags, ma_uint32 capacity) +{ + ma_job_queue_config config; + + config.flags = flags; + config.capacity = capacity; + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t allocatorOffset; + size_t jobsOffset; +} ma_job_queue_heap_layout; + +static ma_result ma_job_queue_get_heap_layout(const ma_job_queue_config* pConfig, ma_job_queue_heap_layout* pHeapLayout) +{ + ma_result result; + + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->capacity == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* Allocator. */ + { + ma_slot_allocator_config allocatorConfig; + size_t allocatorHeapSizeInBytes; + + allocatorConfig = ma_slot_allocator_config_init(pConfig->capacity); + result = ma_slot_allocator_get_heap_size(&allocatorConfig, &allocatorHeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->allocatorOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += allocatorHeapSizeInBytes; + } + + /* Jobs. */ + pHeapLayout->jobsOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(pConfig->capacity * sizeof(ma_job)); + + return MA_SUCCESS; +} + +MA_API ma_result ma_job_queue_get_heap_size(const ma_job_queue_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_job_queue_heap_layout layout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_job_queue_get_heap_layout(pConfig, &layout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = layout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_job_queue_init_preallocated(const ma_job_queue_config* pConfig, void* pHeap, ma_job_queue* pQueue) +{ + ma_result result; + ma_job_queue_heap_layout heapLayout; + ma_slot_allocator_config allocatorConfig; + + if (pQueue == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pQueue); + + result = ma_job_queue_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pQueue->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pQueue->flags = pConfig->flags; + pQueue->capacity = pConfig->capacity; + pQueue->pJobs = (ma_job*)ma_offset_ptr(pHeap, heapLayout.jobsOffset); + + allocatorConfig = ma_slot_allocator_config_init(pConfig->capacity); + result = ma_slot_allocator_init_preallocated(&allocatorConfig, ma_offset_ptr(pHeap, heapLayout.allocatorOffset), &pQueue->allocator); + if (result != MA_SUCCESS) { + return result; + } + + /* We need a semaphore if we're running in non-blocking mode. If threading is disabled we need to return an error. */ + if ((pQueue->flags & MA_JOB_QUEUE_FLAG_NON_BLOCKING) == 0) { + #ifndef MA_NO_THREADING + { + ma_semaphore_init(0, &pQueue->sem); + } + #else + { + /* Threading is disabled and we've requested non-blocking mode. */ + return MA_INVALID_OPERATION; + } + #endif + } + + /* + Our queue needs to be initialized with a free standing node. This should always be slot 0. Required for the lock free algorithm. The first job in the queue is + just a dummy item for giving us the first item in the list which is stored in the "next" member. + */ + ma_slot_allocator_alloc(&pQueue->allocator, &pQueue->head); /* Will never fail. */ + pQueue->pJobs[ma_job_extract_slot(pQueue->head)].next = MA_JOB_ID_NONE; + pQueue->tail = pQueue->head; + + return MA_SUCCESS; +} + +MA_API ma_result ma_job_queue_init(const ma_job_queue_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_job_queue* pQueue) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_job_queue_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_job_queue_init_preallocated(pConfig, pHeap, pQueue); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pQueue->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_job_queue_uninit(ma_job_queue* pQueue, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pQueue == NULL) { + return; + } + + /* All we need to do is uninitialize the semaphore. */ + if ((pQueue->flags & MA_JOB_QUEUE_FLAG_NON_BLOCKING) == 0) { + #ifndef MA_NO_THREADING + { + ma_semaphore_uninit(&pQueue->sem); + } + #else + { + MA_ASSERT(MA_FALSE); /* Should never get here. Should have been checked at initialization time. */ + } + #endif + } + + ma_slot_allocator_uninit(&pQueue->allocator, pAllocationCallbacks); + + if (pQueue->_ownsHeap) { + ma_free(pQueue->_pHeap, pAllocationCallbacks); + } +} + +static ma_bool32 ma_job_queue_cas(volatile ma_uint64* dst, ma_uint64 expected, ma_uint64 desired) +{ + /* The new counter is taken from the expected value. */ + return ma_atomic_compare_and_swap_64(dst, expected, ma_job_set_refcount(desired, ma_job_extract_refcount(expected) + 1)) == expected; +} + +MA_API ma_result ma_job_queue_post(ma_job_queue* pQueue, const ma_job* pJob) +{ + /* + Lock free queue implementation based on the paper by Michael and Scott: Nonblocking Algorithms and Preemption-Safe Locking on Multiprogrammed Shared Memory Multiprocessors + */ + ma_result result; + ma_uint64 slot; + ma_uint64 tail; + ma_uint64 next; + + if (pQueue == NULL || pJob == NULL) { + return MA_INVALID_ARGS; + } + + /* We need a new slot. */ + result = ma_slot_allocator_alloc(&pQueue->allocator, &slot); + if (result != MA_SUCCESS) { + return result; /* Probably ran out of slots. If so, MA_OUT_OF_MEMORY will be returned. */ + } + + /* At this point we should have a slot to place the job. */ + MA_ASSERT(ma_job_extract_slot(slot) < pQueue->capacity); + + /* We need to put the job into memory before we do anything. */ + pQueue->pJobs[ma_job_extract_slot(slot)] = *pJob; + pQueue->pJobs[ma_job_extract_slot(slot)].toc.allocation = slot; /* This will overwrite the job code. */ + pQueue->pJobs[ma_job_extract_slot(slot)].toc.breakup.code = pJob->toc.breakup.code; /* The job code needs to be applied again because the line above overwrote it. */ + pQueue->pJobs[ma_job_extract_slot(slot)].next = MA_JOB_ID_NONE; /* Reset for safety. */ + + #ifndef MA_USE_EXPERIMENTAL_LOCK_FREE_JOB_QUEUE + ma_spinlock_lock(&pQueue->lock); + #endif + { + /* The job is stored in memory so now we need to add it to our linked list. We only ever add items to the end of the list. */ + for (;;) { + tail = ma_atomic_load_64(&pQueue->tail); + next = ma_atomic_load_64(&pQueue->pJobs[ma_job_extract_slot(tail)].next); + + if (ma_job_toc_to_allocation(tail) == ma_job_toc_to_allocation(ma_atomic_load_64(&pQueue->tail))) { + if (ma_job_extract_slot(next) == 0xFFFF) { + if (ma_job_queue_cas(&pQueue->pJobs[ma_job_extract_slot(tail)].next, next, slot)) { + break; + } + } else { + ma_job_queue_cas(&pQueue->tail, tail, ma_job_extract_slot(next)); + } + } + } + ma_job_queue_cas(&pQueue->tail, tail, slot); + } + #ifndef MA_USE_EXPERIMENTAL_LOCK_FREE_JOB_QUEUE + ma_spinlock_unlock(&pQueue->lock); + #endif + + + /* Signal the semaphore as the last step if we're using synchronous mode. */ + if ((pQueue->flags & MA_JOB_QUEUE_FLAG_NON_BLOCKING) == 0) { + #ifndef MA_NO_THREADING + { + ma_semaphore_release(&pQueue->sem); + } + #else + { + MA_ASSERT(MA_FALSE); /* Should never get here. Should have been checked at initialization time. */ + } + #endif + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_job_queue_next(ma_job_queue* pQueue, ma_job* pJob) +{ + ma_uint64 head; + ma_uint64 tail; + ma_uint64 next; + + if (pQueue == NULL || pJob == NULL) { + return MA_INVALID_ARGS; + } + + /* If we're running in synchronous mode we'll need to wait on a semaphore. */ + if ((pQueue->flags & MA_JOB_QUEUE_FLAG_NON_BLOCKING) == 0) { + #ifndef MA_NO_THREADING + { + ma_semaphore_wait(&pQueue->sem); + } + #else + { + MA_ASSERT(MA_FALSE); /* Should never get here. Should have been checked at initialization time. */ + } + #endif + } + + #ifndef MA_USE_EXPERIMENTAL_LOCK_FREE_JOB_QUEUE + ma_spinlock_lock(&pQueue->lock); + #endif + { + /* + BUG: In lock-free mode, multiple threads can be in this section of code. The "head" variable in the loop below + is stored. One thread can fall through to the freeing of this item while another is still using "head" for the + retrieval of the "next" variable. + + The slot allocator might need to make use of some reference counting to ensure it's only truly freed when + there are no more references to the item. This must be fixed before removing these locks. + */ + + /* Now we need to remove the root item from the list. */ + for (;;) { + head = ma_atomic_load_64(&pQueue->head); + tail = ma_atomic_load_64(&pQueue->tail); + next = ma_atomic_load_64(&pQueue->pJobs[ma_job_extract_slot(head)].next); + + if (ma_job_toc_to_allocation(head) == ma_job_toc_to_allocation(ma_atomic_load_64(&pQueue->head))) { + if (ma_job_extract_slot(head) == ma_job_extract_slot(tail)) { + if (ma_job_extract_slot(next) == 0xFFFF) { + #ifndef MA_USE_EXPERIMENTAL_LOCK_FREE_JOB_QUEUE + ma_spinlock_unlock(&pQueue->lock); + #endif + return MA_NO_DATA_AVAILABLE; + } + ma_job_queue_cas(&pQueue->tail, tail, ma_job_extract_slot(next)); + } else { + *pJob = pQueue->pJobs[ma_job_extract_slot(next)]; + if (ma_job_queue_cas(&pQueue->head, head, ma_job_extract_slot(next))) { + break; + } + } + } + } + } + #ifndef MA_USE_EXPERIMENTAL_LOCK_FREE_JOB_QUEUE + ma_spinlock_unlock(&pQueue->lock); + #endif + + ma_slot_allocator_free(&pQueue->allocator, head); + + /* + If it's a quit job make sure it's put back on the queue to ensure other threads have an opportunity to detect it and terminate naturally. We + could instead just leave it on the queue, but that would involve fiddling with the lock-free code above and I want to keep that as simple as + possible. + */ + if (pJob->toc.breakup.code == MA_JOB_TYPE_QUIT) { + ma_job_queue_post(pQueue, pJob); + return MA_CANCELLED; /* Return a cancelled status just in case the thread is checking return codes and not properly checking for a quit job. */ + } + + return MA_SUCCESS; +} + + + +/******************************************************************************* + +Dynamic Linking + +*******************************************************************************/ +#ifdef MA_POSIX + /* No need for dlfcn.h if we're not using runtime linking. */ + #ifndef MA_NO_RUNTIME_LINKING + #include + #endif +#endif + +MA_API ma_handle ma_dlopen(ma_log* pLog, const char* filename) +{ +#ifndef MA_NO_RUNTIME_LINKING + ma_handle handle; + + ma_log_postf(pLog, MA_LOG_LEVEL_DEBUG, "Loading library: %s\n", filename); + + #ifdef MA_WIN32 + /* From MSDN: Desktop applications cannot use LoadPackagedLibrary; if a desktop application calls this function it fails with APPMODEL_ERROR_NO_PACKAGE.*/ + #if !defined(MA_WIN32_UWP) || !(defined(WINAPI_FAMILY) && ((defined(WINAPI_FAMILY_PHONE_APP) && WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP))) + handle = (ma_handle)LoadLibraryA(filename); + #else + /* *sigh* It appears there is no ANSI version of LoadPackagedLibrary()... */ + WCHAR filenameW[4096]; + if (MultiByteToWideChar(CP_UTF8, 0, filename, -1, filenameW, sizeof(filenameW)) == 0) { + handle = NULL; + } else { + handle = (ma_handle)LoadPackagedLibrary(filenameW, 0); + } + #endif + #else + handle = (ma_handle)dlopen(filename, RTLD_NOW); + #endif + + /* + I'm not considering failure to load a library an error nor a warning because seamlessly falling through to a lower-priority + backend is a deliberate design choice. Instead I'm logging it as an informational message. + */ + if (handle == NULL) { + ma_log_postf(pLog, MA_LOG_LEVEL_INFO, "Failed to load library: %s\n", filename); + } + + return handle; +#else + /* Runtime linking is disabled. */ + (void)pLog; + (void)filename; + return NULL; +#endif +} + +MA_API void ma_dlclose(ma_log* pLog, ma_handle handle) +{ +#ifndef MA_NO_RUNTIME_LINKING + #ifdef MA_WIN32 + FreeLibrary((HMODULE)handle); + #else + /* Hack for Android bug (see https://github.com/android/ndk/issues/360). Calling dlclose() pre-API 28 may segfault. */ + #if !defined(MA_ANDROID) || (defined(__ANDROID_API__) && __ANDROID_API__ >= 28) + { + dlclose((void*)handle); + } + #else + { + (void)handle; + } + #endif + #endif + + (void)pLog; +#else + /* Runtime linking is disabled. */ + (void)pLog; + (void)handle; +#endif +} + +MA_API ma_proc ma_dlsym(ma_log* pLog, ma_handle handle, const char* symbol) +{ +#ifndef MA_NO_RUNTIME_LINKING + ma_proc proc; + + ma_log_postf(pLog, MA_LOG_LEVEL_DEBUG, "Loading symbol: %s\n", symbol); + +#ifdef _WIN32 + proc = (ma_proc)GetProcAddress((HMODULE)handle, symbol); +#else +#if (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) || defined(__clang__) + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wpedantic" +#endif + proc = (ma_proc)dlsym((void*)handle, symbol); +#if (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) || defined(__clang__) + #pragma GCC diagnostic pop +#endif +#endif + + if (proc == NULL) { + ma_log_postf(pLog, MA_LOG_LEVEL_WARNING, "Failed to load symbol: %s\n", symbol); + } + + (void)pLog; /* It's possible for pContext to be unused. */ + return proc; +#else + /* Runtime linking is disabled. */ + (void)pLog; + (void)handle; + (void)symbol; + return NULL; +#endif +} + + + +/************************************************************************************************************************************************************ +************************************************************************************************************************************************************* + +DEVICE I/O +========== + +************************************************************************************************************************************************************* +************************************************************************************************************************************************************/ + +/* Disable run-time linking on certain backends and platforms. */ +#ifndef MA_NO_RUNTIME_LINKING + #if defined(MA_EMSCRIPTEN) || defined(MA_ORBIS) || defined(MA_PROSPERO) + #define MA_NO_RUNTIME_LINKING + #endif +#endif + +#ifdef MA_APPLE + #include +#endif + +#ifndef MA_NO_DEVICE_IO + +#if defined(MA_APPLE) && (MAC_OS_X_VERSION_MIN_REQUIRED < 101200) + #include /* For mach_absolute_time() */ +#endif + +#ifdef MA_POSIX + #include + #include + + /* No need for dlfcn.h if we're not using runtime linking. */ + #ifndef MA_NO_RUNTIME_LINKING + #include + #endif +#endif + +/* This must be set to at least 26. */ +#ifndef MA_AAUDIO_MIN_ANDROID_SDK_VERSION +#define MA_AAUDIO_MIN_ANDROID_SDK_VERSION 27 +#endif + + +MA_API void ma_device_info_add_native_data_format(ma_device_info* pDeviceInfo, ma_format format, ma_uint32 channels, ma_uint32 sampleRate, ma_uint32 flags) +{ + if (pDeviceInfo == NULL) { + return; + } + + if (pDeviceInfo->nativeDataFormatCount < ma_countof(pDeviceInfo->nativeDataFormats)) { + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].format = format; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].channels = channels; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].sampleRate = sampleRate; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].flags = flags; + pDeviceInfo->nativeDataFormatCount += 1; + } +} + + +typedef struct +{ + ma_backend backend; + const char* pName; +} ma_backend_info; + +static ma_backend_info gBackendInfo[] = /* Indexed by the backend enum. Must be in the order backends are declared in the ma_backend enum. */ +{ + {ma_backend_wasapi, "WASAPI"}, + {ma_backend_dsound, "DirectSound"}, + {ma_backend_winmm, "WinMM"}, + {ma_backend_coreaudio, "Core Audio"}, + {ma_backend_sndio, "sndio"}, + {ma_backend_audio4, "audio(4)"}, + {ma_backend_oss, "OSS"}, + {ma_backend_pulseaudio, "PulseAudio"}, + {ma_backend_alsa, "ALSA"}, + {ma_backend_jack, "JACK"}, + {ma_backend_aaudio, "AAudio"}, + {ma_backend_opensl, "OpenSL|ES"}, + {ma_backend_webaudio, "Web Audio"}, + {ma_backend_custom, "Custom"}, + {ma_backend_null, "Null"} +}; + +MA_API const char* ma_get_backend_name(ma_backend backend) +{ + if (backend < 0 || backend >= (int)ma_countof(gBackendInfo)) { + return "Unknown"; + } + + return gBackendInfo[backend].pName; +} + +MA_API ma_result ma_get_backend_from_name(const char* pBackendName, ma_backend* pBackend) +{ + size_t iBackend; + + if (pBackendName == NULL) { + return MA_INVALID_ARGS; + } + + for (iBackend = 0; iBackend < ma_countof(gBackendInfo); iBackend += 1) { + if (ma_strcmp(pBackendName, gBackendInfo[iBackend].pName) == 0) { + if (pBackend != NULL) { + *pBackend = gBackendInfo[iBackend].backend; + } + + return MA_SUCCESS; + } + } + + /* Getting here means the backend name is unknown. */ + return MA_INVALID_ARGS; +} + +MA_API ma_bool32 ma_is_backend_enabled(ma_backend backend) +{ + /* + This looks a little bit gross, but we want all backends to be included in the switch to avoid warnings on some compilers + about some enums not being handled by the switch statement. + */ + switch (backend) + { + case ma_backend_wasapi: + #if defined(MA_HAS_WASAPI) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_dsound: + #if defined(MA_HAS_DSOUND) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_winmm: + #if defined(MA_HAS_WINMM) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_coreaudio: + #if defined(MA_HAS_COREAUDIO) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_sndio: + #if defined(MA_HAS_SNDIO) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_audio4: + #if defined(MA_HAS_AUDIO4) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_oss: + #if defined(MA_HAS_OSS) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_pulseaudio: + #if defined(MA_HAS_PULSEAUDIO) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_alsa: + #if defined(MA_HAS_ALSA) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_jack: + #if defined(MA_HAS_JACK) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_aaudio: + #if defined(MA_HAS_AAUDIO) + #if defined(MA_ANDROID) + { + return ma_android_sdk_version() >= MA_AAUDIO_MIN_ANDROID_SDK_VERSION; + } + #else + return MA_FALSE; + #endif + #else + return MA_FALSE; + #endif + case ma_backend_opensl: + #if defined(MA_HAS_OPENSL) + #if defined(MA_ANDROID) + { + return ma_android_sdk_version() >= 9; + } + #else + return MA_TRUE; + #endif + #else + return MA_FALSE; + #endif + case ma_backend_webaudio: + #if defined(MA_HAS_WEBAUDIO) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_custom: + #if defined(MA_HAS_CUSTOM) + return MA_TRUE; + #else + return MA_FALSE; + #endif + case ma_backend_null: + #if defined(MA_HAS_NULL) + return MA_TRUE; + #else + return MA_FALSE; + #endif + + default: return MA_FALSE; + } +} + +MA_API ma_result ma_get_enabled_backends(ma_backend* pBackends, size_t backendCap, size_t* pBackendCount) +{ + size_t backendCount; + size_t iBackend; + ma_result result = MA_SUCCESS; + + if (pBackendCount == NULL) { + return MA_INVALID_ARGS; + } + + backendCount = 0; + + for (iBackend = 0; iBackend <= ma_backend_null; iBackend += 1) { + ma_backend backend = (ma_backend)iBackend; + + if (ma_is_backend_enabled(backend)) { + /* The backend is enabled. Try adding it to the list. If there's no room, MA_NO_SPACE needs to be returned. */ + if (backendCount == backendCap) { + result = MA_NO_SPACE; + break; + } else { + pBackends[backendCount] = backend; + backendCount += 1; + } + } + } + + if (pBackendCount != NULL) { + *pBackendCount = backendCount; + } + + return result; +} + +MA_API ma_bool32 ma_is_loopback_supported(ma_backend backend) +{ + switch (backend) + { + case ma_backend_wasapi: return MA_TRUE; + case ma_backend_dsound: return MA_FALSE; + case ma_backend_winmm: return MA_FALSE; + case ma_backend_coreaudio: return MA_FALSE; + case ma_backend_sndio: return MA_FALSE; + case ma_backend_audio4: return MA_FALSE; + case ma_backend_oss: return MA_FALSE; + case ma_backend_pulseaudio: return MA_FALSE; + case ma_backend_alsa: return MA_FALSE; + case ma_backend_jack: return MA_FALSE; + case ma_backend_aaudio: return MA_FALSE; + case ma_backend_opensl: return MA_FALSE; + case ma_backend_webaudio: return MA_FALSE; + case ma_backend_custom: return MA_FALSE; /* <-- Will depend on the implementation of the backend. */ + case ma_backend_null: return MA_FALSE; + default: return MA_FALSE; + } +} + + + +#if defined(MA_WIN32) +/* WASAPI error codes. */ +#define MA_AUDCLNT_E_NOT_INITIALIZED ((HRESULT)0x88890001) +#define MA_AUDCLNT_E_ALREADY_INITIALIZED ((HRESULT)0x88890002) +#define MA_AUDCLNT_E_WRONG_ENDPOINT_TYPE ((HRESULT)0x88890003) +#define MA_AUDCLNT_E_DEVICE_INVALIDATED ((HRESULT)0x88890004) +#define MA_AUDCLNT_E_NOT_STOPPED ((HRESULT)0x88890005) +#define MA_AUDCLNT_E_BUFFER_TOO_LARGE ((HRESULT)0x88890006) +#define MA_AUDCLNT_E_OUT_OF_ORDER ((HRESULT)0x88890007) +#define MA_AUDCLNT_E_UNSUPPORTED_FORMAT ((HRESULT)0x88890008) +#define MA_AUDCLNT_E_INVALID_SIZE ((HRESULT)0x88890009) +#define MA_AUDCLNT_E_DEVICE_IN_USE ((HRESULT)0x8889000A) +#define MA_AUDCLNT_E_BUFFER_OPERATION_PENDING ((HRESULT)0x8889000B) +#define MA_AUDCLNT_E_THREAD_NOT_REGISTERED ((HRESULT)0x8889000C) +#define MA_AUDCLNT_E_NO_SINGLE_PROCESS ((HRESULT)0x8889000D) +#define MA_AUDCLNT_E_EXCLUSIVE_MODE_NOT_ALLOWED ((HRESULT)0x8889000E) +#define MA_AUDCLNT_E_ENDPOINT_CREATE_FAILED ((HRESULT)0x8889000F) +#define MA_AUDCLNT_E_SERVICE_NOT_RUNNING ((HRESULT)0x88890010) +#define MA_AUDCLNT_E_EVENTHANDLE_NOT_EXPECTED ((HRESULT)0x88890011) +#define MA_AUDCLNT_E_EXCLUSIVE_MODE_ONLY ((HRESULT)0x88890012) +#define MA_AUDCLNT_E_BUFDURATION_PERIOD_NOT_EQUAL ((HRESULT)0x88890013) +#define MA_AUDCLNT_E_EVENTHANDLE_NOT_SET ((HRESULT)0x88890014) +#define MA_AUDCLNT_E_INCORRECT_BUFFER_SIZE ((HRESULT)0x88890015) +#define MA_AUDCLNT_E_BUFFER_SIZE_ERROR ((HRESULT)0x88890016) +#define MA_AUDCLNT_E_CPUUSAGE_EXCEEDED ((HRESULT)0x88890017) +#define MA_AUDCLNT_E_BUFFER_ERROR ((HRESULT)0x88890018) +#define MA_AUDCLNT_E_BUFFER_SIZE_NOT_ALIGNED ((HRESULT)0x88890019) +#define MA_AUDCLNT_E_INVALID_DEVICE_PERIOD ((HRESULT)0x88890020) +#define MA_AUDCLNT_E_INVALID_STREAM_FLAG ((HRESULT)0x88890021) +#define MA_AUDCLNT_E_ENDPOINT_OFFLOAD_NOT_CAPABLE ((HRESULT)0x88890022) +#define MA_AUDCLNT_E_OUT_OF_OFFLOAD_RESOURCES ((HRESULT)0x88890023) +#define MA_AUDCLNT_E_OFFLOAD_MODE_ONLY ((HRESULT)0x88890024) +#define MA_AUDCLNT_E_NONOFFLOAD_MODE_ONLY ((HRESULT)0x88890025) +#define MA_AUDCLNT_E_RESOURCES_INVALIDATED ((HRESULT)0x88890026) +#define MA_AUDCLNT_E_RAW_MODE_UNSUPPORTED ((HRESULT)0x88890027) +#define MA_AUDCLNT_E_ENGINE_PERIODICITY_LOCKED ((HRESULT)0x88890028) +#define MA_AUDCLNT_E_ENGINE_FORMAT_LOCKED ((HRESULT)0x88890029) +#define MA_AUDCLNT_E_HEADTRACKING_ENABLED ((HRESULT)0x88890030) +#define MA_AUDCLNT_E_HEADTRACKING_UNSUPPORTED ((HRESULT)0x88890040) +#define MA_AUDCLNT_S_BUFFER_EMPTY ((HRESULT)0x08890001) +#define MA_AUDCLNT_S_THREAD_ALREADY_REGISTERED ((HRESULT)0x08890002) +#define MA_AUDCLNT_S_POSITION_STALLED ((HRESULT)0x08890003) + +#define MA_DS_OK ((HRESULT)0) +#define MA_DS_NO_VIRTUALIZATION ((HRESULT)0x0878000A) +#define MA_DSERR_ALLOCATED ((HRESULT)0x8878000A) +#define MA_DSERR_CONTROLUNAVAIL ((HRESULT)0x8878001E) +#define MA_DSERR_INVALIDPARAM ((HRESULT)0x80070057) /*E_INVALIDARG*/ +#define MA_DSERR_INVALIDCALL ((HRESULT)0x88780032) +#define MA_DSERR_GENERIC ((HRESULT)0x80004005) /*E_FAIL*/ +#define MA_DSERR_PRIOLEVELNEEDED ((HRESULT)0x88780046) +#define MA_DSERR_OUTOFMEMORY ((HRESULT)0x8007000E) /*E_OUTOFMEMORY*/ +#define MA_DSERR_BADFORMAT ((HRESULT)0x88780064) +#define MA_DSERR_UNSUPPORTED ((HRESULT)0x80004001) /*E_NOTIMPL*/ +#define MA_DSERR_NODRIVER ((HRESULT)0x88780078) +#define MA_DSERR_ALREADYINITIALIZED ((HRESULT)0x88780082) +#define MA_DSERR_NOAGGREGATION ((HRESULT)0x80040110) /*CLASS_E_NOAGGREGATION*/ +#define MA_DSERR_BUFFERLOST ((HRESULT)0x88780096) +#define MA_DSERR_OTHERAPPHASPRIO ((HRESULT)0x887800A0) +#define MA_DSERR_UNINITIALIZED ((HRESULT)0x887800AA) +#define MA_DSERR_NOINTERFACE ((HRESULT)0x80004002) /*E_NOINTERFACE*/ +#define MA_DSERR_ACCESSDENIED ((HRESULT)0x80070005) /*E_ACCESSDENIED*/ +#define MA_DSERR_BUFFERTOOSMALL ((HRESULT)0x887800B4) +#define MA_DSERR_DS8_REQUIRED ((HRESULT)0x887800BE) +#define MA_DSERR_SENDLOOP ((HRESULT)0x887800C8) +#define MA_DSERR_BADSENDBUFFERGUID ((HRESULT)0x887800D2) +#define MA_DSERR_OBJECTNOTFOUND ((HRESULT)0x88781161) +#define MA_DSERR_FXUNAVAILABLE ((HRESULT)0x887800DC) + +static ma_result ma_result_from_HRESULT(HRESULT hr) +{ + switch (hr) + { + case NOERROR: return MA_SUCCESS; + /*case S_OK: return MA_SUCCESS;*/ + + case E_POINTER: return MA_INVALID_ARGS; + case E_UNEXPECTED: return MA_ERROR; + case E_NOTIMPL: return MA_NOT_IMPLEMENTED; + case E_OUTOFMEMORY: return MA_OUT_OF_MEMORY; + case E_INVALIDARG: return MA_INVALID_ARGS; + case E_NOINTERFACE: return MA_API_NOT_FOUND; + case E_HANDLE: return MA_INVALID_ARGS; + case E_ABORT: return MA_ERROR; + case E_FAIL: return MA_ERROR; + case E_ACCESSDENIED: return MA_ACCESS_DENIED; + + /* WASAPI */ + case MA_AUDCLNT_E_NOT_INITIALIZED: return MA_DEVICE_NOT_INITIALIZED; + case MA_AUDCLNT_E_ALREADY_INITIALIZED: return MA_DEVICE_ALREADY_INITIALIZED; + case MA_AUDCLNT_E_WRONG_ENDPOINT_TYPE: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_DEVICE_INVALIDATED: return MA_UNAVAILABLE; + case MA_AUDCLNT_E_NOT_STOPPED: return MA_DEVICE_NOT_STOPPED; + case MA_AUDCLNT_E_BUFFER_TOO_LARGE: return MA_TOO_BIG; + case MA_AUDCLNT_E_OUT_OF_ORDER: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_UNSUPPORTED_FORMAT: return MA_FORMAT_NOT_SUPPORTED; + case MA_AUDCLNT_E_INVALID_SIZE: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_DEVICE_IN_USE: return MA_BUSY; + case MA_AUDCLNT_E_BUFFER_OPERATION_PENDING: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_THREAD_NOT_REGISTERED: return MA_DOES_NOT_EXIST; + case MA_AUDCLNT_E_NO_SINGLE_PROCESS: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_EXCLUSIVE_MODE_NOT_ALLOWED: return MA_SHARE_MODE_NOT_SUPPORTED; + case MA_AUDCLNT_E_ENDPOINT_CREATE_FAILED: return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + case MA_AUDCLNT_E_SERVICE_NOT_RUNNING: return MA_NOT_CONNECTED; + case MA_AUDCLNT_E_EVENTHANDLE_NOT_EXPECTED: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_EXCLUSIVE_MODE_ONLY: return MA_SHARE_MODE_NOT_SUPPORTED; + case MA_AUDCLNT_E_BUFDURATION_PERIOD_NOT_EQUAL: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_EVENTHANDLE_NOT_SET: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_INCORRECT_BUFFER_SIZE: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_BUFFER_SIZE_ERROR: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_CPUUSAGE_EXCEEDED: return MA_ERROR; + case MA_AUDCLNT_E_BUFFER_ERROR: return MA_ERROR; + case MA_AUDCLNT_E_BUFFER_SIZE_NOT_ALIGNED: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_INVALID_DEVICE_PERIOD: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_INVALID_STREAM_FLAG: return MA_INVALID_ARGS; + case MA_AUDCLNT_E_ENDPOINT_OFFLOAD_NOT_CAPABLE: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_OUT_OF_OFFLOAD_RESOURCES: return MA_OUT_OF_MEMORY; + case MA_AUDCLNT_E_OFFLOAD_MODE_ONLY: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_NONOFFLOAD_MODE_ONLY: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_RESOURCES_INVALIDATED: return MA_INVALID_DATA; + case MA_AUDCLNT_E_RAW_MODE_UNSUPPORTED: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_ENGINE_PERIODICITY_LOCKED: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_ENGINE_FORMAT_LOCKED: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_HEADTRACKING_ENABLED: return MA_INVALID_OPERATION; + case MA_AUDCLNT_E_HEADTRACKING_UNSUPPORTED: return MA_INVALID_OPERATION; + case MA_AUDCLNT_S_BUFFER_EMPTY: return MA_NO_SPACE; + case MA_AUDCLNT_S_THREAD_ALREADY_REGISTERED: return MA_ALREADY_EXISTS; + case MA_AUDCLNT_S_POSITION_STALLED: return MA_ERROR; + + /* DirectSound */ + /*case MA_DS_OK: return MA_SUCCESS;*/ /* S_OK */ + case MA_DS_NO_VIRTUALIZATION: return MA_SUCCESS; + case MA_DSERR_ALLOCATED: return MA_ALREADY_IN_USE; + case MA_DSERR_CONTROLUNAVAIL: return MA_INVALID_OPERATION; + /*case MA_DSERR_INVALIDPARAM: return MA_INVALID_ARGS;*/ /* E_INVALIDARG */ + case MA_DSERR_INVALIDCALL: return MA_INVALID_OPERATION; + /*case MA_DSERR_GENERIC: return MA_ERROR;*/ /* E_FAIL */ + case MA_DSERR_PRIOLEVELNEEDED: return MA_INVALID_OPERATION; + /*case MA_DSERR_OUTOFMEMORY: return MA_OUT_OF_MEMORY;*/ /* E_OUTOFMEMORY */ + case MA_DSERR_BADFORMAT: return MA_FORMAT_NOT_SUPPORTED; + /*case MA_DSERR_UNSUPPORTED: return MA_NOT_IMPLEMENTED;*/ /* E_NOTIMPL */ + case MA_DSERR_NODRIVER: return MA_FAILED_TO_INIT_BACKEND; + case MA_DSERR_ALREADYINITIALIZED: return MA_DEVICE_ALREADY_INITIALIZED; + case MA_DSERR_NOAGGREGATION: return MA_ERROR; + case MA_DSERR_BUFFERLOST: return MA_UNAVAILABLE; + case MA_DSERR_OTHERAPPHASPRIO: return MA_ACCESS_DENIED; + case MA_DSERR_UNINITIALIZED: return MA_DEVICE_NOT_INITIALIZED; + /*case MA_DSERR_NOINTERFACE: return MA_API_NOT_FOUND;*/ /* E_NOINTERFACE */ + /*case MA_DSERR_ACCESSDENIED: return MA_ACCESS_DENIED;*/ /* E_ACCESSDENIED */ + case MA_DSERR_BUFFERTOOSMALL: return MA_NO_SPACE; + case MA_DSERR_DS8_REQUIRED: return MA_INVALID_OPERATION; + case MA_DSERR_SENDLOOP: return MA_DEADLOCK; + case MA_DSERR_BADSENDBUFFERGUID: return MA_INVALID_ARGS; + case MA_DSERR_OBJECTNOTFOUND: return MA_NO_DEVICE; + case MA_DSERR_FXUNAVAILABLE: return MA_UNAVAILABLE; + + default: return MA_ERROR; + } +} + +/* PROPVARIANT */ +#define MA_VT_LPWSTR 31 +#define MA_VT_BLOB 65 + +#if defined(_MSC_VER) && !defined(__clang__) + #pragma warning(push) + #pragma warning(disable:4201) /* nonstandard extension used: nameless struct/union */ +#elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wpedantic" /* For ISO C99 doesn't support unnamed structs/unions [-Wpedantic] */ + #if defined(__clang__) + #pragma GCC diagnostic ignored "-Wc11-extensions" /* anonymous unions are a C11 extension */ + #endif +#endif +typedef struct +{ + WORD vt; + WORD wReserved1; + WORD wReserved2; + WORD wReserved3; + union + { + struct + { + ULONG cbSize; + BYTE* pBlobData; + } blob; + WCHAR* pwszVal; + char pad[16]; /* Just to ensure the size of the struct matches the official version. */ + }; +} MA_PROPVARIANT; +#if defined(_MSC_VER) && !defined(__clang__) + #pragma warning(pop) +#elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) + #pragma GCC diagnostic pop +#endif + +typedef HRESULT (WINAPI * MA_PFN_CoInitialize)(void* pvReserved); +typedef HRESULT (WINAPI * MA_PFN_CoInitializeEx)(void* pvReserved, DWORD dwCoInit); +typedef void (WINAPI * MA_PFN_CoUninitialize)(void); +typedef HRESULT (WINAPI * MA_PFN_CoCreateInstance)(const IID* rclsid, void* pUnkOuter, DWORD dwClsContext, const IID* riid, void* ppv); +typedef void (WINAPI * MA_PFN_CoTaskMemFree)(void* pv); +typedef HRESULT (WINAPI * MA_PFN_PropVariantClear)(MA_PROPVARIANT *pvar); +typedef int (WINAPI * MA_PFN_StringFromGUID2)(const GUID* const rguid, WCHAR* lpsz, int cchMax); + +typedef HWND (WINAPI * MA_PFN_GetForegroundWindow)(void); +typedef HWND (WINAPI * MA_PFN_GetDesktopWindow)(void); + +#if defined(MA_WIN32_DESKTOP) +/* Microsoft documents these APIs as returning LSTATUS, but the Win32 API shipping with some compilers do not define it. It's just a LONG. */ +typedef LONG (WINAPI * MA_PFN_RegOpenKeyExA)(HKEY hKey, const char* lpSubKey, DWORD ulOptions, DWORD samDesired, HKEY* phkResult); +typedef LONG (WINAPI * MA_PFN_RegCloseKey)(HKEY hKey); +typedef LONG (WINAPI * MA_PFN_RegQueryValueExA)(HKEY hKey, const char* lpValueName, DWORD* lpReserved, DWORD* lpType, BYTE* lpData, DWORD* lpcbData); +#endif /* MA_WIN32_DESKTOP */ + +MA_API size_t ma_strlen_WCHAR(const WCHAR* str) +{ + size_t len = 0; + while (str[len] != '\0') { + len += 1; + } + + return len; +} + +MA_API int ma_strcmp_WCHAR(const WCHAR *s1, const WCHAR *s2) +{ + while (*s1 != '\0' && *s1 == *s2) { + s1 += 1; + s2 += 1; + } + + return *s1 - *s2; +} + +MA_API int ma_strcpy_s_WCHAR(WCHAR* dst, size_t dstCap, const WCHAR* src) +{ + size_t i; + + if (dst == 0) { + return 22; + } + if (dstCap == 0) { + return 34; + } + if (src == 0) { + dst[0] = '\0'; + return 22; + } + + for (i = 0; i < dstCap && src[i] != '\0'; ++i) { + dst[i] = src[i]; + } + + if (i < dstCap) { + dst[i] = '\0'; + return 0; + } + + dst[0] = '\0'; + return 34; +} +#endif /* MA_WIN32 */ + + +#define MA_DEFAULT_PLAYBACK_DEVICE_NAME "Default Playback Device" +#define MA_DEFAULT_CAPTURE_DEVICE_NAME "Default Capture Device" + + + + +/******************************************************************************* + +Timing + +*******************************************************************************/ +#if defined(MA_WIN32) && !defined(MA_POSIX) + static LARGE_INTEGER g_ma_TimerFrequency; /* <-- Initialized to zero since it's static. */ + static void ma_timer_init(ma_timer* pTimer) + { + LARGE_INTEGER counter; + + if (g_ma_TimerFrequency.QuadPart == 0) { + QueryPerformanceFrequency(&g_ma_TimerFrequency); + } + + QueryPerformanceCounter(&counter); + pTimer->counter = counter.QuadPart; + } + + static double ma_timer_get_time_in_seconds(ma_timer* pTimer) + { + LARGE_INTEGER counter; + if (!QueryPerformanceCounter(&counter)) { + return 0; + } + + return (double)(counter.QuadPart - pTimer->counter) / g_ma_TimerFrequency.QuadPart; + } +#elif defined(MA_APPLE) && (MAC_OS_X_VERSION_MIN_REQUIRED < 101200) + static ma_uint64 g_ma_TimerFrequency = 0; + static void ma_timer_init(ma_timer* pTimer) + { + mach_timebase_info_data_t baseTime; + mach_timebase_info(&baseTime); + g_ma_TimerFrequency = (baseTime.denom * 1e9) / baseTime.numer; + + pTimer->counter = mach_absolute_time(); + } + + static double ma_timer_get_time_in_seconds(ma_timer* pTimer) + { + ma_uint64 newTimeCounter = mach_absolute_time(); + ma_uint64 oldTimeCounter = pTimer->counter; + + return (newTimeCounter - oldTimeCounter) / g_ma_TimerFrequency; + } +#elif defined(MA_EMSCRIPTEN) + static MA_INLINE void ma_timer_init(ma_timer* pTimer) + { + pTimer->counterD = emscripten_get_now(); + } + + static MA_INLINE double ma_timer_get_time_in_seconds(ma_timer* pTimer) + { + return (emscripten_get_now() - pTimer->counterD) / 1000; /* Emscripten is in milliseconds. */ + } +#else + #if defined(_POSIX_C_SOURCE) && _POSIX_C_SOURCE >= 199309L + #if defined(CLOCK_MONOTONIC) + #define MA_CLOCK_ID CLOCK_MONOTONIC + #else + #define MA_CLOCK_ID CLOCK_REALTIME + #endif + + static void ma_timer_init(ma_timer* pTimer) + { + struct timespec newTime; + clock_gettime(MA_CLOCK_ID, &newTime); + + pTimer->counter = (newTime.tv_sec * 1000000000) + newTime.tv_nsec; + } + + static double ma_timer_get_time_in_seconds(ma_timer* pTimer) + { + ma_uint64 newTimeCounter; + ma_uint64 oldTimeCounter; + + struct timespec newTime; + clock_gettime(MA_CLOCK_ID, &newTime); + + newTimeCounter = (newTime.tv_sec * 1000000000) + newTime.tv_nsec; + oldTimeCounter = pTimer->counter; + + return (newTimeCounter - oldTimeCounter) / 1000000000.0; + } + #else + static void ma_timer_init(ma_timer* pTimer) + { + struct timeval newTime; + gettimeofday(&newTime, NULL); + + pTimer->counter = (newTime.tv_sec * 1000000) + newTime.tv_usec; + } + + static double ma_timer_get_time_in_seconds(ma_timer* pTimer) + { + ma_uint64 newTimeCounter; + ma_uint64 oldTimeCounter; + + struct timeval newTime; + gettimeofday(&newTime, NULL); + + newTimeCounter = (newTime.tv_sec * 1000000) + newTime.tv_usec; + oldTimeCounter = pTimer->counter; + + return (newTimeCounter - oldTimeCounter) / 1000000.0; + } + #endif +#endif + + + +#if 0 +static ma_uint32 ma_get_closest_standard_sample_rate(ma_uint32 sampleRateIn) +{ + ma_uint32 closestRate = 0; + ma_uint32 closestDiff = 0xFFFFFFFF; + size_t iStandardRate; + + for (iStandardRate = 0; iStandardRate < ma_countof(g_maStandardSampleRatePriorities); ++iStandardRate) { + ma_uint32 standardRate = g_maStandardSampleRatePriorities[iStandardRate]; + ma_uint32 diff; + + if (sampleRateIn > standardRate) { + diff = sampleRateIn - standardRate; + } else { + diff = standardRate - sampleRateIn; + } + + if (diff == 0) { + return standardRate; /* The input sample rate is a standard rate. */ + } + + if (closestDiff > diff) { + closestDiff = diff; + closestRate = standardRate; + } + } + + return closestRate; +} +#endif + + +static MA_INLINE unsigned int ma_device_disable_denormals(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (!pDevice->noDisableDenormals) { + return ma_disable_denormals(); + } else { + return 0; + } +} + +static MA_INLINE void ma_device_restore_denormals(ma_device* pDevice, unsigned int prevState) +{ + MA_ASSERT(pDevice != NULL); + + if (!pDevice->noDisableDenormals) { + ma_restore_denormals(prevState); + } else { + /* Do nothing. */ + (void)prevState; + } +} + +static ma_device_notification ma_device_notification_init(ma_device* pDevice, ma_device_notification_type type) +{ + ma_device_notification notification; + + MA_ZERO_OBJECT(¬ification); + notification.pDevice = pDevice; + notification.type = type; + + return notification; +} + +static void ma_device__on_notification(ma_device_notification notification) +{ + MA_ASSERT(notification.pDevice != NULL); + + if (notification.pDevice->onNotification != NULL) { + notification.pDevice->onNotification(¬ification); + } + + /* TEMP FOR COMPATIBILITY: If it's a stopped notification, fire the onStop callback as well. This is only for backwards compatibility and will be removed. */ + if (notification.pDevice->onStop != NULL && notification.type == ma_device_notification_type_stopped) { + notification.pDevice->onStop(notification.pDevice); + } +} + +static void ma_device__on_notification_started(ma_device* pDevice) +{ + ma_device__on_notification(ma_device_notification_init(pDevice, ma_device_notification_type_started)); +} + +static void ma_device__on_notification_stopped(ma_device* pDevice) +{ + ma_device__on_notification(ma_device_notification_init(pDevice, ma_device_notification_type_stopped)); +} + +/* Not all platforms support reroute notifications. */ +#if !defined(MA_EMSCRIPTEN) +static void ma_device__on_notification_rerouted(ma_device* pDevice) +{ + ma_device__on_notification(ma_device_notification_init(pDevice, ma_device_notification_type_rerouted)); +} +#endif + +#if defined(MA_EMSCRIPTEN) +#ifdef __cplusplus +extern "C" { +#endif +void EMSCRIPTEN_KEEPALIVE ma_device__on_notification_unlocked(ma_device* pDevice) +{ + ma_device__on_notification(ma_device_notification_init(pDevice, ma_device_notification_type_unlocked)); +} +#ifdef __cplusplus +} +#endif +#endif + + +static void ma_device__on_data_inner(ma_device* pDevice, void* pFramesOut, const void* pFramesIn, ma_uint32 frameCount) +{ + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pDevice->onData != NULL); + + if (!pDevice->noPreSilencedOutputBuffer && pFramesOut != NULL) { + ma_silence_pcm_frames(pFramesOut, frameCount, pDevice->playback.format, pDevice->playback.channels); + } + + pDevice->onData(pDevice, pFramesOut, pFramesIn, frameCount); +} + +static void ma_device__on_data(ma_device* pDevice, void* pFramesOut, const void* pFramesIn, ma_uint32 frameCount) +{ + MA_ASSERT(pDevice != NULL); + + /* Don't read more data from the client if we're in the process of stopping. */ + if (ma_device_get_state(pDevice) == ma_device_state_stopping) { + return; + } + + if (pDevice->noFixedSizedCallback) { + /* Fast path. Not using a fixed sized callback. Process directly from the specified buffers. */ + ma_device__on_data_inner(pDevice, pFramesOut, pFramesIn, frameCount); + } else { + /* Slow path. Using a fixed sized callback. Need to use the intermediary buffer. */ + ma_uint32 totalFramesProcessed = 0; + + while (totalFramesProcessed < frameCount) { + ma_uint32 totalFramesRemaining = frameCount - totalFramesProcessed; + ma_uint32 framesToProcessThisIteration = 0; + + if (pFramesIn != NULL) { + /* Capturing. Write to the intermediary buffer. If there's no room, fire the callback to empty it. */ + if (pDevice->capture.intermediaryBufferLen < pDevice->capture.intermediaryBufferCap) { + /* There's some room left in the intermediary buffer. Write to it without firing the callback. */ + framesToProcessThisIteration = totalFramesRemaining; + if (framesToProcessThisIteration > pDevice->capture.intermediaryBufferCap - pDevice->capture.intermediaryBufferLen) { + framesToProcessThisIteration = pDevice->capture.intermediaryBufferCap - pDevice->capture.intermediaryBufferLen; + } + + ma_copy_pcm_frames( + ma_offset_pcm_frames_ptr(pDevice->capture.pIntermediaryBuffer, pDevice->capture.intermediaryBufferLen, pDevice->capture.format, pDevice->capture.channels), + ma_offset_pcm_frames_const_ptr(pFramesIn, totalFramesProcessed, pDevice->capture.format, pDevice->capture.channels), + framesToProcessThisIteration, + pDevice->capture.format, pDevice->capture.channels); + + pDevice->capture.intermediaryBufferLen += framesToProcessThisIteration; + } + + if (pDevice->capture.intermediaryBufferLen == pDevice->capture.intermediaryBufferCap) { + /* No room left in the intermediary buffer. Fire the data callback. */ + if (pDevice->type == ma_device_type_duplex) { + /* We'll do the duplex data callback later after we've processed the playback data. */ + } else { + ma_device__on_data_inner(pDevice, NULL, pDevice->capture.pIntermediaryBuffer, pDevice->capture.intermediaryBufferCap); + + /* The intermediary buffer has just been drained. */ + pDevice->capture.intermediaryBufferLen = 0; + } + } + } + + if (pFramesOut != NULL) { + /* Playing back. Read from the intermediary buffer. If there's nothing in it, fire the callback to fill it. */ + if (pDevice->playback.intermediaryBufferLen > 0) { + /* There's some content in the intermediary buffer. Read from that without firing the callback. */ + if (pDevice->type == ma_device_type_duplex) { + /* The frames processed this iteration for a duplex device will always be based on the capture side. Leave it unmodified. */ + } else { + framesToProcessThisIteration = totalFramesRemaining; + if (framesToProcessThisIteration > pDevice->playback.intermediaryBufferLen) { + framesToProcessThisIteration = pDevice->playback.intermediaryBufferLen; + } + } + + ma_copy_pcm_frames( + ma_offset_pcm_frames_ptr(pFramesOut, totalFramesProcessed, pDevice->playback.format, pDevice->playback.channels), + ma_offset_pcm_frames_ptr(pDevice->playback.pIntermediaryBuffer, pDevice->playback.intermediaryBufferCap - pDevice->playback.intermediaryBufferLen, pDevice->playback.format, pDevice->playback.channels), + framesToProcessThisIteration, + pDevice->playback.format, pDevice->playback.channels); + + pDevice->playback.intermediaryBufferLen -= framesToProcessThisIteration; + } + + if (pDevice->playback.intermediaryBufferLen == 0) { + /* There's nothing in the intermediary buffer. Fire the data callback to fill it. */ + if (pDevice->type == ma_device_type_duplex) { + /* In duplex mode, the data callback will be fired later. Nothing to do here. */ + } else { + ma_device__on_data_inner(pDevice, pDevice->playback.pIntermediaryBuffer, NULL, pDevice->playback.intermediaryBufferCap); + + /* The intermediary buffer has just been filled. */ + pDevice->playback.intermediaryBufferLen = pDevice->playback.intermediaryBufferCap; + } + } + } + + /* If we're in duplex mode we might need to do a refill of the data. */ + if (pDevice->type == ma_device_type_duplex) { + if (pDevice->capture.intermediaryBufferLen == pDevice->capture.intermediaryBufferCap) { + ma_device__on_data_inner(pDevice, pDevice->playback.pIntermediaryBuffer, pDevice->capture.pIntermediaryBuffer, pDevice->capture.intermediaryBufferCap); + + pDevice->playback.intermediaryBufferLen = pDevice->playback.intermediaryBufferCap; /* The playback buffer will have just been filled. */ + pDevice->capture.intermediaryBufferLen = 0; /* The intermediary buffer has just been drained. */ + } + } + + /* Make sure this is only incremented once in the duplex case. */ + totalFramesProcessed += framesToProcessThisIteration; + } + } +} + +static void ma_device__handle_data_callback(ma_device* pDevice, void* pFramesOut, const void* pFramesIn, ma_uint32 frameCount) +{ + float masterVolumeFactor; + + ma_device_get_master_volume(pDevice, &masterVolumeFactor); /* Use ma_device_get_master_volume() to ensure the volume is loaded atomically. */ + + if (pDevice->onData) { + unsigned int prevDenormalState = ma_device_disable_denormals(pDevice); + { + /* Volume control of input makes things a bit awkward because the input buffer is read-only. We'll need to use a temp buffer and loop in this case. */ + if (pFramesIn != NULL && masterVolumeFactor != 1) { + ma_uint8 tempFramesIn[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint32 bpfCapture = ma_get_bytes_per_frame(pDevice->capture.format, pDevice->capture.channels); + ma_uint32 bpfPlayback = ma_get_bytes_per_frame(pDevice->playback.format, pDevice->playback.channels); + ma_uint32 totalFramesProcessed = 0; + while (totalFramesProcessed < frameCount) { + ma_uint32 framesToProcessThisIteration = frameCount - totalFramesProcessed; + if (framesToProcessThisIteration > sizeof(tempFramesIn)/bpfCapture) { + framesToProcessThisIteration = sizeof(tempFramesIn)/bpfCapture; + } + + ma_copy_and_apply_volume_factor_pcm_frames(tempFramesIn, ma_offset_ptr(pFramesIn, totalFramesProcessed*bpfCapture), framesToProcessThisIteration, pDevice->capture.format, pDevice->capture.channels, masterVolumeFactor); + + ma_device__on_data(pDevice, ma_offset_ptr(pFramesOut, totalFramesProcessed*bpfPlayback), tempFramesIn, framesToProcessThisIteration); + + totalFramesProcessed += framesToProcessThisIteration; + } + } else { + ma_device__on_data(pDevice, pFramesOut, pFramesIn, frameCount); + } + + /* Volume control and clipping for playback devices. */ + if (pFramesOut != NULL) { + if (masterVolumeFactor != 1) { + if (pFramesIn == NULL) { /* <-- In full-duplex situations, the volume will have been applied to the input samples before the data callback. Applying it again post-callback will incorrectly compound it. */ + ma_apply_volume_factor_pcm_frames(pFramesOut, frameCount, pDevice->playback.format, pDevice->playback.channels, masterVolumeFactor); + } + } + + if (!pDevice->noClip && pDevice->playback.format == ma_format_f32) { + ma_clip_samples_f32((float*)pFramesOut, (const float*)pFramesOut, frameCount * pDevice->playback.channels); /* Intentionally specifying the same pointer for both input and output for in-place processing. */ + } + } + } + ma_device_restore_denormals(pDevice, prevDenormalState); + } else { + /* No data callback. Just silence the output. */ + if (pFramesOut != NULL) { + ma_silence_pcm_frames(pFramesOut, frameCount, pDevice->playback.format, pDevice->playback.channels); + } + } +} + + + +/* A helper function for reading sample data from the client. */ +static void ma_device__read_frames_from_client(ma_device* pDevice, ma_uint32 frameCount, void* pFramesOut) +{ + MA_ASSERT(pDevice != NULL); + MA_ASSERT(frameCount > 0); + MA_ASSERT(pFramesOut != NULL); + + if (pDevice->playback.converter.isPassthrough) { + ma_device__handle_data_callback(pDevice, pFramesOut, NULL, frameCount); + } else { + ma_result result; + ma_uint64 totalFramesReadOut; + void* pRunningFramesOut; + + totalFramesReadOut = 0; + pRunningFramesOut = pFramesOut; + + /* + We run slightly different logic depending on whether or not we're using a heap-allocated + buffer for caching input data. This will be the case if the data converter does not have + the ability to retrieve the required input frame count for a given output frame count. + */ + if (pDevice->playback.pInputCache != NULL) { + while (totalFramesReadOut < frameCount) { + ma_uint64 framesToReadThisIterationIn; + ma_uint64 framesToReadThisIterationOut; + + /* If there's any data available in the cache, that needs to get processed first. */ + if (pDevice->playback.inputCacheRemaining > 0) { + framesToReadThisIterationOut = (frameCount - totalFramesReadOut); + framesToReadThisIterationIn = framesToReadThisIterationOut; + if (framesToReadThisIterationIn > pDevice->playback.inputCacheRemaining) { + framesToReadThisIterationIn = pDevice->playback.inputCacheRemaining; + } + + result = ma_data_converter_process_pcm_frames(&pDevice->playback.converter, ma_offset_pcm_frames_ptr(pDevice->playback.pInputCache, pDevice->playback.inputCacheConsumed, pDevice->playback.format, pDevice->playback.channels), &framesToReadThisIterationIn, pRunningFramesOut, &framesToReadThisIterationOut); + if (result != MA_SUCCESS) { + break; + } + + pDevice->playback.inputCacheConsumed += framesToReadThisIterationIn; + pDevice->playback.inputCacheRemaining -= framesToReadThisIterationIn; + + totalFramesReadOut += framesToReadThisIterationOut; + pRunningFramesOut = ma_offset_ptr(pRunningFramesOut, framesToReadThisIterationOut * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels)); + + if (framesToReadThisIterationIn == 0 && framesToReadThisIterationOut == 0) { + break; /* We're done. */ + } + } + + /* Getting here means there's no data in the cache and we need to fill it up with data from the client. */ + if (pDevice->playback.inputCacheRemaining == 0) { + ma_device__handle_data_callback(pDevice, pDevice->playback.pInputCache, NULL, (ma_uint32)pDevice->playback.inputCacheCap); + + pDevice->playback.inputCacheConsumed = 0; + pDevice->playback.inputCacheRemaining = pDevice->playback.inputCacheCap; + } + } + } else { + while (totalFramesReadOut < frameCount) { + ma_uint8 pIntermediaryBuffer[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* In client format. */ + ma_uint64 intermediaryBufferCap = sizeof(pIntermediaryBuffer) / ma_get_bytes_per_frame(pDevice->playback.format, pDevice->playback.channels); + ma_uint64 framesToReadThisIterationIn; + ma_uint64 framesReadThisIterationIn; + ma_uint64 framesToReadThisIterationOut; + ma_uint64 framesReadThisIterationOut; + ma_uint64 requiredInputFrameCount; + + framesToReadThisIterationOut = (frameCount - totalFramesReadOut); + framesToReadThisIterationIn = framesToReadThisIterationOut; + if (framesToReadThisIterationIn > intermediaryBufferCap) { + framesToReadThisIterationIn = intermediaryBufferCap; + } + + ma_data_converter_get_required_input_frame_count(&pDevice->playback.converter, framesToReadThisIterationOut, &requiredInputFrameCount); + if (framesToReadThisIterationIn > requiredInputFrameCount) { + framesToReadThisIterationIn = requiredInputFrameCount; + } + + ma_device__handle_data_callback(pDevice, pIntermediaryBuffer, NULL, (ma_uint32)framesToReadThisIterationIn); + + /* + At this point we have our decoded data in input format and now we need to convert to output format. Note that even if we didn't read any + input frames, we still want to try processing frames because there may some output frames generated from cached input data. + */ + framesReadThisIterationIn = framesToReadThisIterationIn; + framesReadThisIterationOut = framesToReadThisIterationOut; + result = ma_data_converter_process_pcm_frames(&pDevice->playback.converter, pIntermediaryBuffer, &framesReadThisIterationIn, pRunningFramesOut, &framesReadThisIterationOut); + if (result != MA_SUCCESS) { + break; + } + + totalFramesReadOut += framesReadThisIterationOut; + pRunningFramesOut = ma_offset_ptr(pRunningFramesOut, framesReadThisIterationOut * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels)); + + if (framesReadThisIterationIn == 0 && framesReadThisIterationOut == 0) { + break; /* We're done. */ + } + } + } + } +} + +/* A helper for sending sample data to the client. */ +static void ma_device__send_frames_to_client(ma_device* pDevice, ma_uint32 frameCountInDeviceFormat, const void* pFramesInDeviceFormat) +{ + MA_ASSERT(pDevice != NULL); + MA_ASSERT(frameCountInDeviceFormat > 0); + MA_ASSERT(pFramesInDeviceFormat != NULL); + + if (pDevice->capture.converter.isPassthrough) { + ma_device__handle_data_callback(pDevice, NULL, pFramesInDeviceFormat, frameCountInDeviceFormat); + } else { + ma_result result; + ma_uint8 pFramesInClientFormat[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint64 framesInClientFormatCap = sizeof(pFramesInClientFormat) / ma_get_bytes_per_frame(pDevice->capture.format, pDevice->capture.channels); + ma_uint64 totalDeviceFramesProcessed = 0; + ma_uint64 totalClientFramesProcessed = 0; + const void* pRunningFramesInDeviceFormat = pFramesInDeviceFormat; + + /* We just keep going until we've exhausted all of our input frames and cannot generate any more output frames. */ + for (;;) { + ma_uint64 deviceFramesProcessedThisIteration; + ma_uint64 clientFramesProcessedThisIteration; + + deviceFramesProcessedThisIteration = (frameCountInDeviceFormat - totalDeviceFramesProcessed); + clientFramesProcessedThisIteration = framesInClientFormatCap; + + result = ma_data_converter_process_pcm_frames(&pDevice->capture.converter, pRunningFramesInDeviceFormat, &deviceFramesProcessedThisIteration, pFramesInClientFormat, &clientFramesProcessedThisIteration); + if (result != MA_SUCCESS) { + break; + } + + if (clientFramesProcessedThisIteration > 0) { + ma_device__handle_data_callback(pDevice, NULL, pFramesInClientFormat, (ma_uint32)clientFramesProcessedThisIteration); /* Safe cast. */ + } + + pRunningFramesInDeviceFormat = ma_offset_ptr(pRunningFramesInDeviceFormat, deviceFramesProcessedThisIteration * ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels)); + totalDeviceFramesProcessed += deviceFramesProcessedThisIteration; + totalClientFramesProcessed += clientFramesProcessedThisIteration; + + /* This is just to silence a warning. I might want to use this variable later so leaving in place for now. */ + (void)totalClientFramesProcessed; + + if (deviceFramesProcessedThisIteration == 0 && clientFramesProcessedThisIteration == 0) { + break; /* We're done. */ + } + } + } +} + +static ma_result ma_device__handle_duplex_callback_capture(ma_device* pDevice, ma_uint32 frameCountInDeviceFormat, const void* pFramesInDeviceFormat, ma_pcm_rb* pRB) +{ + ma_result result; + ma_uint32 totalDeviceFramesProcessed = 0; + const void* pRunningFramesInDeviceFormat = pFramesInDeviceFormat; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(frameCountInDeviceFormat > 0); + MA_ASSERT(pFramesInDeviceFormat != NULL); + MA_ASSERT(pRB != NULL); + + /* Write to the ring buffer. The ring buffer is in the client format which means we need to convert. */ + for (;;) { + ma_uint32 framesToProcessInDeviceFormat = (frameCountInDeviceFormat - totalDeviceFramesProcessed); + ma_uint32 framesToProcessInClientFormat = MA_DATA_CONVERTER_STACK_BUFFER_SIZE / ma_get_bytes_per_frame(pDevice->capture.format, pDevice->capture.channels); + ma_uint64 framesProcessedInDeviceFormat; + ma_uint64 framesProcessedInClientFormat; + void* pFramesInClientFormat; + + result = ma_pcm_rb_acquire_write(pRB, &framesToProcessInClientFormat, &pFramesInClientFormat); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "Failed to acquire capture PCM frames from ring buffer."); + break; + } + + if (framesToProcessInClientFormat == 0) { + if (ma_pcm_rb_pointer_distance(pRB) == (ma_int32)ma_pcm_rb_get_subbuffer_size(pRB)) { + break; /* Overrun. Not enough room in the ring buffer for input frame. Excess frames are dropped. */ + } + } + + /* Convert. */ + framesProcessedInDeviceFormat = framesToProcessInDeviceFormat; + framesProcessedInClientFormat = framesToProcessInClientFormat; + result = ma_data_converter_process_pcm_frames(&pDevice->capture.converter, pRunningFramesInDeviceFormat, &framesProcessedInDeviceFormat, pFramesInClientFormat, &framesProcessedInClientFormat); + if (result != MA_SUCCESS) { + break; + } + + result = ma_pcm_rb_commit_write(pRB, (ma_uint32)framesProcessedInClientFormat); /* Safe cast. */ + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "Failed to commit capture PCM frames to ring buffer."); + break; + } + + pRunningFramesInDeviceFormat = ma_offset_ptr(pRunningFramesInDeviceFormat, framesProcessedInDeviceFormat * ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels)); + totalDeviceFramesProcessed += (ma_uint32)framesProcessedInDeviceFormat; /* Safe cast. */ + + /* We're done when we're unable to process any client nor device frames. */ + if (framesProcessedInClientFormat == 0 && framesProcessedInDeviceFormat == 0) { + break; /* Done. */ + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device__handle_duplex_callback_playback(ma_device* pDevice, ma_uint32 frameCount, void* pFramesInInternalFormat, ma_pcm_rb* pRB) +{ + ma_result result; + ma_uint8 silentInputFrames[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint32 totalFramesReadOut = 0; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(frameCount > 0); + MA_ASSERT(pFramesInInternalFormat != NULL); + MA_ASSERT(pRB != NULL); + MA_ASSERT(pDevice->playback.pInputCache != NULL); + + /* + Sitting in the ring buffer should be captured data from the capture callback in external format. If there's not enough data in there for + the whole frameCount frames we just use silence instead for the input data. + */ + MA_ZERO_MEMORY(silentInputFrames, sizeof(silentInputFrames)); + + while (totalFramesReadOut < frameCount && ma_device_is_started(pDevice)) { + /* + We should have a buffer allocated on the heap. Any playback frames still sitting in there + need to be sent to the internal device before we process any more data from the client. + */ + if (pDevice->playback.inputCacheRemaining > 0) { + ma_uint64 framesConvertedIn = pDevice->playback.inputCacheRemaining; + ma_uint64 framesConvertedOut = (frameCount - totalFramesReadOut); + ma_data_converter_process_pcm_frames(&pDevice->playback.converter, ma_offset_pcm_frames_ptr(pDevice->playback.pInputCache, pDevice->playback.inputCacheConsumed, pDevice->playback.format, pDevice->playback.channels), &framesConvertedIn, pFramesInInternalFormat, &framesConvertedOut); + + pDevice->playback.inputCacheConsumed += framesConvertedIn; + pDevice->playback.inputCacheRemaining -= framesConvertedIn; + + totalFramesReadOut += (ma_uint32)framesConvertedOut; /* Safe cast. */ + pFramesInInternalFormat = ma_offset_ptr(pFramesInInternalFormat, framesConvertedOut * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels)); + } + + /* If there's no more data in the cache we'll need to fill it with some. */ + if (totalFramesReadOut < frameCount && pDevice->playback.inputCacheRemaining == 0) { + ma_uint32 inputFrameCount; + void* pInputFrames; + + inputFrameCount = (ma_uint32)pDevice->playback.inputCacheCap; + result = ma_pcm_rb_acquire_read(pRB, &inputFrameCount, &pInputFrames); + if (result == MA_SUCCESS) { + if (inputFrameCount > 0) { + ma_device__handle_data_callback(pDevice, pDevice->playback.pInputCache, pInputFrames, inputFrameCount); + } else { + if (ma_pcm_rb_pointer_distance(pRB) == 0) { + break; /* Underrun. */ + } + } + } else { + /* No capture data available. Feed in silence. */ + inputFrameCount = (ma_uint32)ma_min(pDevice->playback.inputCacheCap, sizeof(silentInputFrames) / ma_get_bytes_per_frame(pDevice->capture.format, pDevice->capture.channels)); + ma_device__handle_data_callback(pDevice, pDevice->playback.pInputCache, silentInputFrames, inputFrameCount); + } + + pDevice->playback.inputCacheConsumed = 0; + pDevice->playback.inputCacheRemaining = inputFrameCount; + + result = ma_pcm_rb_commit_read(pRB, inputFrameCount); + if (result != MA_SUCCESS) { + return result; /* Should never happen. */ + } + } + } + + return MA_SUCCESS; +} + +/* A helper for changing the state of the device. */ +static MA_INLINE void ma_device__set_state(ma_device* pDevice, ma_device_state newState) +{ + ma_atomic_device_state_set(&pDevice->state, newState); +} + + +#if defined(MA_WIN32) + static GUID MA_GUID_KSDATAFORMAT_SUBTYPE_PCM = {0x00000001, 0x0000, 0x0010, {0x80, 0x00, 0x00, 0xaa, 0x00, 0x38, 0x9b, 0x71}}; + static GUID MA_GUID_KSDATAFORMAT_SUBTYPE_IEEE_FLOAT = {0x00000003, 0x0000, 0x0010, {0x80, 0x00, 0x00, 0xaa, 0x00, 0x38, 0x9b, 0x71}}; + /*static GUID MA_GUID_KSDATAFORMAT_SUBTYPE_ALAW = {0x00000006, 0x0000, 0x0010, {0x80, 0x00, 0x00, 0xaa, 0x00, 0x38, 0x9b, 0x71}};*/ + /*static GUID MA_GUID_KSDATAFORMAT_SUBTYPE_MULAW = {0x00000007, 0x0000, 0x0010, {0x80, 0x00, 0x00, 0xaa, 0x00, 0x38, 0x9b, 0x71}};*/ +#endif + + + +MA_API ma_uint32 ma_get_format_priority_index(ma_format format) /* Lower = better. */ +{ + ma_uint32 i; + for (i = 0; i < ma_countof(g_maFormatPriorities); ++i) { + if (g_maFormatPriorities[i] == format) { + return i; + } + } + + /* Getting here means the format could not be found or is equal to ma_format_unknown. */ + return (ma_uint32)-1; +} + +static ma_result ma_device__post_init_setup(ma_device* pDevice, ma_device_type deviceType); + +static ma_bool32 ma_device_descriptor_is_valid(const ma_device_descriptor* pDeviceDescriptor) +{ + if (pDeviceDescriptor == NULL) { + return MA_FALSE; + } + + if (pDeviceDescriptor->format == ma_format_unknown) { + return MA_FALSE; + } + + if (pDeviceDescriptor->channels == 0 || pDeviceDescriptor->channels > MA_MAX_CHANNELS) { + return MA_FALSE; + } + + if (pDeviceDescriptor->sampleRate == 0) { + return MA_FALSE; + } + + return MA_TRUE; +} + + +static ma_result ma_device_audio_thread__default_read_write(ma_device* pDevice) +{ + ma_result result = MA_SUCCESS; + ma_bool32 exitLoop = MA_FALSE; + ma_uint8 capturedDeviceData[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint8 playbackDeviceData[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint32 capturedDeviceDataCapInFrames = 0; + ma_uint32 playbackDeviceDataCapInFrames = 0; + + MA_ASSERT(pDevice != NULL); + + /* Just some quick validation on the device type and the available callbacks. */ + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex || pDevice->type == ma_device_type_loopback) { + if (pDevice->pContext->callbacks.onDeviceRead == NULL) { + return MA_NOT_IMPLEMENTED; + } + + capturedDeviceDataCapInFrames = sizeof(capturedDeviceData) / ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + if (pDevice->pContext->callbacks.onDeviceWrite == NULL) { + return MA_NOT_IMPLEMENTED; + } + + playbackDeviceDataCapInFrames = sizeof(playbackDeviceData) / ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + } + + /* NOTE: The device was started outside of this function, in the worker thread. */ + + while (ma_device_get_state(pDevice) == ma_device_state_started && !exitLoop) { + switch (pDevice->type) { + case ma_device_type_duplex: + { + /* The process is: onDeviceRead() -> convert -> callback -> convert -> onDeviceWrite() */ + ma_uint32 totalCapturedDeviceFramesProcessed = 0; + ma_uint32 capturedDevicePeriodSizeInFrames = ma_min(pDevice->capture.internalPeriodSizeInFrames, pDevice->playback.internalPeriodSizeInFrames); + + while (totalCapturedDeviceFramesProcessed < capturedDevicePeriodSizeInFrames) { + ma_uint32 capturedDeviceFramesRemaining; + ma_uint32 capturedDeviceFramesProcessed; + ma_uint32 capturedDeviceFramesToProcess; + ma_uint32 capturedDeviceFramesToTryProcessing = capturedDevicePeriodSizeInFrames - totalCapturedDeviceFramesProcessed; + if (capturedDeviceFramesToTryProcessing > capturedDeviceDataCapInFrames) { + capturedDeviceFramesToTryProcessing = capturedDeviceDataCapInFrames; + } + + result = pDevice->pContext->callbacks.onDeviceRead(pDevice, capturedDeviceData, capturedDeviceFramesToTryProcessing, &capturedDeviceFramesToProcess); + if (result != MA_SUCCESS) { + exitLoop = MA_TRUE; + break; + } + + capturedDeviceFramesRemaining = capturedDeviceFramesToProcess; + capturedDeviceFramesProcessed = 0; + + /* At this point we have our captured data in device format and we now need to convert it to client format. */ + for (;;) { + ma_uint8 capturedClientData[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint8 playbackClientData[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint32 capturedClientDataCapInFrames = sizeof(capturedClientData) / ma_get_bytes_per_frame(pDevice->capture.format, pDevice->capture.channels); + ma_uint32 playbackClientDataCapInFrames = sizeof(playbackClientData) / ma_get_bytes_per_frame(pDevice->playback.format, pDevice->playback.channels); + ma_uint64 capturedClientFramesToProcessThisIteration = ma_min(capturedClientDataCapInFrames, playbackClientDataCapInFrames); + ma_uint64 capturedDeviceFramesToProcessThisIteration = capturedDeviceFramesRemaining; + ma_uint8* pRunningCapturedDeviceFrames = ma_offset_ptr(capturedDeviceData, capturedDeviceFramesProcessed * ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels)); + + /* Convert capture data from device format to client format. */ + result = ma_data_converter_process_pcm_frames(&pDevice->capture.converter, pRunningCapturedDeviceFrames, &capturedDeviceFramesToProcessThisIteration, capturedClientData, &capturedClientFramesToProcessThisIteration); + if (result != MA_SUCCESS) { + break; + } + + /* + If we weren't able to generate any output frames it must mean we've exhausted all of our input. The only time this would not be the case is if capturedClientData was too small + which should never be the case when it's of the size MA_DATA_CONVERTER_STACK_BUFFER_SIZE. + */ + if (capturedClientFramesToProcessThisIteration == 0) { + break; + } + + ma_device__handle_data_callback(pDevice, playbackClientData, capturedClientData, (ma_uint32)capturedClientFramesToProcessThisIteration); /* Safe cast .*/ + + capturedDeviceFramesProcessed += (ma_uint32)capturedDeviceFramesToProcessThisIteration; /* Safe cast. */ + capturedDeviceFramesRemaining -= (ma_uint32)capturedDeviceFramesToProcessThisIteration; /* Safe cast. */ + + /* At this point the playbackClientData buffer should be holding data that needs to be written to the device. */ + for (;;) { + ma_uint64 convertedClientFrameCount = capturedClientFramesToProcessThisIteration; + ma_uint64 convertedDeviceFrameCount = playbackDeviceDataCapInFrames; + result = ma_data_converter_process_pcm_frames(&pDevice->playback.converter, playbackClientData, &convertedClientFrameCount, playbackDeviceData, &convertedDeviceFrameCount); + if (result != MA_SUCCESS) { + break; + } + + result = pDevice->pContext->callbacks.onDeviceWrite(pDevice, playbackDeviceData, (ma_uint32)convertedDeviceFrameCount, NULL); /* Safe cast. */ + if (result != MA_SUCCESS) { + exitLoop = MA_TRUE; + break; + } + + capturedClientFramesToProcessThisIteration -= (ma_uint32)convertedClientFrameCount; /* Safe cast. */ + if (capturedClientFramesToProcessThisIteration == 0) { + break; + } + } + + /* In case an error happened from ma_device_write__null()... */ + if (result != MA_SUCCESS) { + exitLoop = MA_TRUE; + break; + } + } + + /* Make sure we don't get stuck in the inner loop. */ + if (capturedDeviceFramesProcessed == 0) { + break; + } + + totalCapturedDeviceFramesProcessed += capturedDeviceFramesProcessed; + } + } break; + + case ma_device_type_capture: + case ma_device_type_loopback: + { + ma_uint32 periodSizeInFrames = pDevice->capture.internalPeriodSizeInFrames; + ma_uint32 framesReadThisPeriod = 0; + while (framesReadThisPeriod < periodSizeInFrames) { + ma_uint32 framesRemainingInPeriod = periodSizeInFrames - framesReadThisPeriod; + ma_uint32 framesProcessed; + ma_uint32 framesToReadThisIteration = framesRemainingInPeriod; + if (framesToReadThisIteration > capturedDeviceDataCapInFrames) { + framesToReadThisIteration = capturedDeviceDataCapInFrames; + } + + result = pDevice->pContext->callbacks.onDeviceRead(pDevice, capturedDeviceData, framesToReadThisIteration, &framesProcessed); + if (result != MA_SUCCESS) { + exitLoop = MA_TRUE; + break; + } + + /* Make sure we don't get stuck in the inner loop. */ + if (framesProcessed == 0) { + break; + } + + ma_device__send_frames_to_client(pDevice, framesProcessed, capturedDeviceData); + + framesReadThisPeriod += framesProcessed; + } + } break; + + case ma_device_type_playback: + { + /* We write in chunks of the period size, but use a stack allocated buffer for the intermediary. */ + ma_uint32 periodSizeInFrames = pDevice->playback.internalPeriodSizeInFrames; + ma_uint32 framesWrittenThisPeriod = 0; + while (framesWrittenThisPeriod < periodSizeInFrames) { + ma_uint32 framesRemainingInPeriod = periodSizeInFrames - framesWrittenThisPeriod; + ma_uint32 framesProcessed; + ma_uint32 framesToWriteThisIteration = framesRemainingInPeriod; + if (framesToWriteThisIteration > playbackDeviceDataCapInFrames) { + framesToWriteThisIteration = playbackDeviceDataCapInFrames; + } + + ma_device__read_frames_from_client(pDevice, framesToWriteThisIteration, playbackDeviceData); + + result = pDevice->pContext->callbacks.onDeviceWrite(pDevice, playbackDeviceData, framesToWriteThisIteration, &framesProcessed); + if (result != MA_SUCCESS) { + exitLoop = MA_TRUE; + break; + } + + /* Make sure we don't get stuck in the inner loop. */ + if (framesProcessed == 0) { + break; + } + + framesWrittenThisPeriod += framesProcessed; + } + } break; + + /* Should never get here. */ + default: break; + } + } + + return result; +} + + + +/******************************************************************************* + +Null Backend + +*******************************************************************************/ +#ifdef MA_HAS_NULL + +#define MA_DEVICE_OP_NONE__NULL 0 +#define MA_DEVICE_OP_START__NULL 1 +#define MA_DEVICE_OP_SUSPEND__NULL 2 +#define MA_DEVICE_OP_KILL__NULL 3 + +static ma_thread_result MA_THREADCALL ma_device_thread__null(void* pData) +{ + ma_device* pDevice = (ma_device*)pData; + MA_ASSERT(pDevice != NULL); + + for (;;) { /* Keep the thread alive until the device is uninitialized. */ + ma_uint32 operation; + + /* Wait for an operation to be requested. */ + ma_event_wait(&pDevice->null_device.operationEvent); + + /* At this point an event should have been triggered. */ + operation = pDevice->null_device.operation; + + /* Starting the device needs to put the thread into a loop. */ + if (operation == MA_DEVICE_OP_START__NULL) { + /* Reset the timer just in case. */ + ma_timer_init(&pDevice->null_device.timer); + + /* Getting here means a suspend or kill operation has been requested. */ + pDevice->null_device.operationResult = MA_SUCCESS; + ma_event_signal(&pDevice->null_device.operationCompletionEvent); + ma_semaphore_release(&pDevice->null_device.operationSemaphore); + continue; + } + + /* Suspending the device means we need to stop the timer and just continue the loop. */ + if (operation == MA_DEVICE_OP_SUSPEND__NULL) { + /* We need to add the current run time to the prior run time, then reset the timer. */ + pDevice->null_device.priorRunTime += ma_timer_get_time_in_seconds(&pDevice->null_device.timer); + ma_timer_init(&pDevice->null_device.timer); + + /* We're done. */ + pDevice->null_device.operationResult = MA_SUCCESS; + ma_event_signal(&pDevice->null_device.operationCompletionEvent); + ma_semaphore_release(&pDevice->null_device.operationSemaphore); + continue; + } + + /* Killing the device means we need to get out of this loop so that this thread can terminate. */ + if (operation == MA_DEVICE_OP_KILL__NULL) { + pDevice->null_device.operationResult = MA_SUCCESS; + ma_event_signal(&pDevice->null_device.operationCompletionEvent); + ma_semaphore_release(&pDevice->null_device.operationSemaphore); + break; + } + + /* Getting a signal on a "none" operation probably means an error. Return invalid operation. */ + if (operation == MA_DEVICE_OP_NONE__NULL) { + MA_ASSERT(MA_FALSE); /* <-- Trigger this in debug mode to ensure developers are aware they're doing something wrong (or there's a bug in a miniaudio). */ + pDevice->null_device.operationResult = MA_INVALID_OPERATION; + ma_event_signal(&pDevice->null_device.operationCompletionEvent); + ma_semaphore_release(&pDevice->null_device.operationSemaphore); + continue; /* Continue the loop. Don't terminate. */ + } + } + + return (ma_thread_result)0; +} + +static ma_result ma_device_do_operation__null(ma_device* pDevice, ma_uint32 operation) +{ + ma_result result; + + /* + TODO: Need to review this and consider just using mutual exclusion. I think the original motivation + for this was to just post the event to a queue and return immediately, but that has since changed + and now this function is synchronous. I think this can be simplified to just use a mutex. + */ + + /* + The first thing to do is wait for an operation slot to become available. We only have a single slot for this, but we could extend this later + to support queuing of operations. + */ + result = ma_semaphore_wait(&pDevice->null_device.operationSemaphore); + if (result != MA_SUCCESS) { + return result; /* Failed to wait for the event. */ + } + + /* + When we get here it means the background thread is not referencing the operation code and it can be changed. After changing this we need to + signal an event to the worker thread to let it know that it can start work. + */ + pDevice->null_device.operation = operation; + + /* Once the operation code has been set, the worker thread can start work. */ + if (ma_event_signal(&pDevice->null_device.operationEvent) != MA_SUCCESS) { + return MA_ERROR; + } + + /* We want everything to be synchronous so we're going to wait for the worker thread to complete it's operation. */ + if (ma_event_wait(&pDevice->null_device.operationCompletionEvent) != MA_SUCCESS) { + return MA_ERROR; + } + + return pDevice->null_device.operationResult; +} + +static ma_uint64 ma_device_get_total_run_time_in_frames__null(ma_device* pDevice) +{ + ma_uint32 internalSampleRate; + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + internalSampleRate = pDevice->capture.internalSampleRate; + } else { + internalSampleRate = pDevice->playback.internalSampleRate; + } + + return (ma_uint64)((pDevice->null_device.priorRunTime + ma_timer_get_time_in_seconds(&pDevice->null_device.timer)) * internalSampleRate); +} + +static ma_result ma_context_enumerate_devices__null(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_bool32 cbResult = MA_TRUE; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + /* Playback. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), "NULL Playback Device", (size_t)-1); + deviceInfo.isDefault = MA_TRUE; /* Only one playback and capture device for the null backend, so might as well mark as default. */ + cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + + /* Capture. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), "NULL Capture Device", (size_t)-1); + deviceInfo.isDefault = MA_TRUE; /* Only one playback and capture device for the null backend, so might as well mark as default. */ + cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + + (void)cbResult; /* Silence a static analysis warning. */ + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info__null(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + MA_ASSERT(pContext != NULL); + + if (pDeviceID != NULL && pDeviceID->nullbackend != 0) { + return MA_NO_DEVICE; /* Don't know the device. */ + } + + /* Name / Description */ + if (deviceType == ma_device_type_playback) { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), "NULL Playback Device", (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), "NULL Capture Device", (size_t)-1); + } + + pDeviceInfo->isDefault = MA_TRUE; /* Only one playback and capture device for the null backend, so might as well mark as default. */ + + /* Support everything on the null backend. */ + pDeviceInfo->nativeDataFormats[0].format = ma_format_unknown; + pDeviceInfo->nativeDataFormats[0].channels = 0; + pDeviceInfo->nativeDataFormats[0].sampleRate = 0; + pDeviceInfo->nativeDataFormats[0].flags = 0; + pDeviceInfo->nativeDataFormatCount = 1; + + (void)pContext; + return MA_SUCCESS; +} + + +static ma_result ma_device_uninit__null(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + /* Keep it clean and wait for the device thread to finish before returning. */ + ma_device_do_operation__null(pDevice, MA_DEVICE_OP_KILL__NULL); + + /* Wait for the thread to finish before continuing. */ + ma_thread_wait(&pDevice->null_device.deviceThread); + + /* At this point the loop in the device thread is as good as terminated so we can uninitialize our events. */ + ma_semaphore_uninit(&pDevice->null_device.operationSemaphore); + ma_event_uninit(&pDevice->null_device.operationCompletionEvent); + ma_event_uninit(&pDevice->null_device.operationEvent); + + return MA_SUCCESS; +} + +static ma_result ma_device_init__null(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + + MA_ZERO_OBJECT(&pDevice->null_device); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + /* The null backend supports everything exactly as we specify it. */ + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + pDescriptorCapture->format = (pDescriptorCapture->format != ma_format_unknown) ? pDescriptorCapture->format : MA_DEFAULT_FORMAT; + pDescriptorCapture->channels = (pDescriptorCapture->channels != 0) ? pDescriptorCapture->channels : MA_DEFAULT_CHANNELS; + pDescriptorCapture->sampleRate = (pDescriptorCapture->sampleRate != 0) ? pDescriptorCapture->sampleRate : MA_DEFAULT_SAMPLE_RATE; + + if (pDescriptorCapture->channelMap[0] == MA_CHANNEL_NONE) { + ma_channel_map_init_standard(ma_standard_channel_map_default, pDescriptorCapture->channelMap, ma_countof(pDescriptorCapture->channelMap), pDescriptorCapture->channels); + } + + pDescriptorCapture->periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptorCapture, pDescriptorCapture->sampleRate, pConfig->performanceProfile); + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + pDescriptorPlayback->format = (pDescriptorPlayback->format != ma_format_unknown) ? pDescriptorPlayback->format : MA_DEFAULT_FORMAT; + pDescriptorPlayback->channels = (pDescriptorPlayback->channels != 0) ? pDescriptorPlayback->channels : MA_DEFAULT_CHANNELS; + pDescriptorPlayback->sampleRate = (pDescriptorPlayback->sampleRate != 0) ? pDescriptorPlayback->sampleRate : MA_DEFAULT_SAMPLE_RATE; + + if (pDescriptorPlayback->channelMap[0] == MA_CHANNEL_NONE) { + ma_channel_map_init_standard(ma_standard_channel_map_default, pDescriptorPlayback->channelMap, ma_countof(pDescriptorCapture->channelMap), pDescriptorPlayback->channels); + } + + pDescriptorPlayback->periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptorPlayback, pDescriptorPlayback->sampleRate, pConfig->performanceProfile); + } + + /* + In order to get timing right, we need to create a thread that does nothing but keeps track of the timer. This timer is started when the + first period is "written" to it, and then stopped in ma_device_stop__null(). + */ + result = ma_event_init(&pDevice->null_device.operationEvent); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_event_init(&pDevice->null_device.operationCompletionEvent); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_semaphore_init(1, &pDevice->null_device.operationSemaphore); /* <-- It's important that the initial value is set to 1. */ + if (result != MA_SUCCESS) { + return result; + } + + result = ma_thread_create(&pDevice->null_device.deviceThread, pDevice->pContext->threadPriority, 0, ma_device_thread__null, pDevice, &pDevice->pContext->allocationCallbacks); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_start__null(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + ma_device_do_operation__null(pDevice, MA_DEVICE_OP_START__NULL); + + ma_atomic_bool32_set(&pDevice->null_device.isStarted, MA_TRUE); + return MA_SUCCESS; +} + +static ma_result ma_device_stop__null(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + ma_device_do_operation__null(pDevice, MA_DEVICE_OP_SUSPEND__NULL); + + ma_atomic_bool32_set(&pDevice->null_device.isStarted, MA_FALSE); + return MA_SUCCESS; +} + +static ma_bool32 ma_device_is_started__null(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + return ma_atomic_bool32_get(&pDevice->null_device.isStarted); +} + +static ma_result ma_device_write__null(ma_device* pDevice, const void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesWritten) +{ + ma_result result = MA_SUCCESS; + ma_uint32 totalPCMFramesProcessed; + ma_bool32 wasStartedOnEntry; + + if (pFramesWritten != NULL) { + *pFramesWritten = 0; + } + + wasStartedOnEntry = ma_device_is_started__null(pDevice); + + /* Keep going until everything has been read. */ + totalPCMFramesProcessed = 0; + while (totalPCMFramesProcessed < frameCount) { + ma_uint64 targetFrame; + + /* If there are any frames remaining in the current period, consume those first. */ + if (pDevice->null_device.currentPeriodFramesRemainingPlayback > 0) { + ma_uint32 framesRemaining = (frameCount - totalPCMFramesProcessed); + ma_uint32 framesToProcess = pDevice->null_device.currentPeriodFramesRemainingPlayback; + if (framesToProcess > framesRemaining) { + framesToProcess = framesRemaining; + } + + /* We don't actually do anything with pPCMFrames, so just mark it as unused to prevent a warning. */ + (void)pPCMFrames; + + pDevice->null_device.currentPeriodFramesRemainingPlayback -= framesToProcess; + totalPCMFramesProcessed += framesToProcess; + } + + /* If we've consumed the current period we'll need to mark it as such an ensure the device is started if it's not already. */ + if (pDevice->null_device.currentPeriodFramesRemainingPlayback == 0) { + pDevice->null_device.currentPeriodFramesRemainingPlayback = 0; + + if (!ma_device_is_started__null(pDevice) && !wasStartedOnEntry) { + result = ma_device_start__null(pDevice); + if (result != MA_SUCCESS) { + break; + } + } + } + + /* If we've consumed the whole buffer we can return now. */ + MA_ASSERT(totalPCMFramesProcessed <= frameCount); + if (totalPCMFramesProcessed == frameCount) { + break; + } + + /* Getting here means we've still got more frames to consume, we but need to wait for it to become available. */ + targetFrame = pDevice->null_device.lastProcessedFramePlayback; + for (;;) { + ma_uint64 currentFrame; + + /* Stop waiting if the device has been stopped. */ + if (!ma_device_is_started__null(pDevice)) { + break; + } + + currentFrame = ma_device_get_total_run_time_in_frames__null(pDevice); + if (currentFrame >= targetFrame) { + break; + } + + /* Getting here means we haven't yet reached the target sample, so continue waiting. */ + ma_sleep(10); + } + + pDevice->null_device.lastProcessedFramePlayback += pDevice->playback.internalPeriodSizeInFrames; + pDevice->null_device.currentPeriodFramesRemainingPlayback = pDevice->playback.internalPeriodSizeInFrames; + } + + if (pFramesWritten != NULL) { + *pFramesWritten = totalPCMFramesProcessed; + } + + return result; +} + +static ma_result ma_device_read__null(ma_device* pDevice, void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesRead) +{ + ma_result result = MA_SUCCESS; + ma_uint32 totalPCMFramesProcessed; + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + /* Keep going until everything has been read. */ + totalPCMFramesProcessed = 0; + while (totalPCMFramesProcessed < frameCount) { + ma_uint64 targetFrame; + + /* If there are any frames remaining in the current period, consume those first. */ + if (pDevice->null_device.currentPeriodFramesRemainingCapture > 0) { + ma_uint32 bpf = ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + ma_uint32 framesRemaining = (frameCount - totalPCMFramesProcessed); + ma_uint32 framesToProcess = pDevice->null_device.currentPeriodFramesRemainingCapture; + if (framesToProcess > framesRemaining) { + framesToProcess = framesRemaining; + } + + /* We need to ensure the output buffer is zeroed. */ + MA_ZERO_MEMORY(ma_offset_ptr(pPCMFrames, totalPCMFramesProcessed*bpf), framesToProcess*bpf); + + pDevice->null_device.currentPeriodFramesRemainingCapture -= framesToProcess; + totalPCMFramesProcessed += framesToProcess; + } + + /* If we've consumed the current period we'll need to mark it as such an ensure the device is started if it's not already. */ + if (pDevice->null_device.currentPeriodFramesRemainingCapture == 0) { + pDevice->null_device.currentPeriodFramesRemainingCapture = 0; + } + + /* If we've consumed the whole buffer we can return now. */ + MA_ASSERT(totalPCMFramesProcessed <= frameCount); + if (totalPCMFramesProcessed == frameCount) { + break; + } + + /* Getting here means we've still got more frames to consume, we but need to wait for it to become available. */ + targetFrame = pDevice->null_device.lastProcessedFrameCapture + pDevice->capture.internalPeriodSizeInFrames; + for (;;) { + ma_uint64 currentFrame; + + /* Stop waiting if the device has been stopped. */ + if (!ma_device_is_started__null(pDevice)) { + break; + } + + currentFrame = ma_device_get_total_run_time_in_frames__null(pDevice); + if (currentFrame >= targetFrame) { + break; + } + + /* Getting here means we haven't yet reached the target sample, so continue waiting. */ + ma_sleep(10); + } + + pDevice->null_device.lastProcessedFrameCapture += pDevice->capture.internalPeriodSizeInFrames; + pDevice->null_device.currentPeriodFramesRemainingCapture = pDevice->capture.internalPeriodSizeInFrames; + } + + if (pFramesRead != NULL) { + *pFramesRead = totalPCMFramesProcessed; + } + + return result; +} + +static ma_result ma_context_uninit__null(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_null); + + (void)pContext; + return MA_SUCCESS; +} + +static ma_result ma_context_init__null(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + MA_ASSERT(pContext != NULL); + + (void)pConfig; + (void)pContext; + + pCallbacks->onContextInit = ma_context_init__null; + pCallbacks->onContextUninit = ma_context_uninit__null; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__null; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__null; + pCallbacks->onDeviceInit = ma_device_init__null; + pCallbacks->onDeviceUninit = ma_device_uninit__null; + pCallbacks->onDeviceStart = ma_device_start__null; + pCallbacks->onDeviceStop = ma_device_stop__null; + pCallbacks->onDeviceRead = ma_device_read__null; + pCallbacks->onDeviceWrite = ma_device_write__null; + pCallbacks->onDeviceDataLoop = NULL; /* Our backend is asynchronous with a blocking read-write API which means we can get miniaudio to deal with the audio thread. */ + + /* The null backend always works. */ + return MA_SUCCESS; +} +#endif + + + +/******************************************************************************* + +WIN32 COMMON + +*******************************************************************************/ +#if defined(MA_WIN32) +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + #define ma_CoInitializeEx(pContext, pvReserved, dwCoInit) ((pContext->win32.CoInitializeEx) ? ((MA_PFN_CoInitializeEx)pContext->win32.CoInitializeEx)(pvReserved, dwCoInit) : ((MA_PFN_CoInitialize)pContext->win32.CoInitialize)(pvReserved)) + #define ma_CoUninitialize(pContext) ((MA_PFN_CoUninitialize)pContext->win32.CoUninitialize)() + #define ma_CoCreateInstance(pContext, rclsid, pUnkOuter, dwClsContext, riid, ppv) ((MA_PFN_CoCreateInstance)pContext->win32.CoCreateInstance)(rclsid, pUnkOuter, dwClsContext, riid, ppv) + #define ma_CoTaskMemFree(pContext, pv) ((MA_PFN_CoTaskMemFree)pContext->win32.CoTaskMemFree)(pv) + #define ma_PropVariantClear(pContext, pvar) ((MA_PFN_PropVariantClear)pContext->win32.PropVariantClear)(pvar) +#else + #define ma_CoInitializeEx(pContext, pvReserved, dwCoInit) CoInitializeEx(pvReserved, dwCoInit) + #define ma_CoUninitialize(pContext) CoUninitialize() + #define ma_CoCreateInstance(pContext, rclsid, pUnkOuter, dwClsContext, riid, ppv) CoCreateInstance(rclsid, pUnkOuter, dwClsContext, riid, ppv) + #define ma_CoTaskMemFree(pContext, pv) CoTaskMemFree(pv) + #define ma_PropVariantClear(pContext, pvar) PropVariantClear(pvar) +#endif + +#if !defined(MAXULONG_PTR) && !defined(__WATCOMC__) +typedef size_t DWORD_PTR; +#endif + +#if !defined(WAVE_FORMAT_1M08) +#define WAVE_FORMAT_1M08 0x00000001 +#define WAVE_FORMAT_1S08 0x00000002 +#define WAVE_FORMAT_1M16 0x00000004 +#define WAVE_FORMAT_1S16 0x00000008 +#define WAVE_FORMAT_2M08 0x00000010 +#define WAVE_FORMAT_2S08 0x00000020 +#define WAVE_FORMAT_2M16 0x00000040 +#define WAVE_FORMAT_2S16 0x00000080 +#define WAVE_FORMAT_4M08 0x00000100 +#define WAVE_FORMAT_4S08 0x00000200 +#define WAVE_FORMAT_4M16 0x00000400 +#define WAVE_FORMAT_4S16 0x00000800 +#endif + +#if !defined(WAVE_FORMAT_44M08) +#define WAVE_FORMAT_44M08 0x00000100 +#define WAVE_FORMAT_44S08 0x00000200 +#define WAVE_FORMAT_44M16 0x00000400 +#define WAVE_FORMAT_44S16 0x00000800 +#define WAVE_FORMAT_48M08 0x00001000 +#define WAVE_FORMAT_48S08 0x00002000 +#define WAVE_FORMAT_48M16 0x00004000 +#define WAVE_FORMAT_48S16 0x00008000 +#define WAVE_FORMAT_96M08 0x00010000 +#define WAVE_FORMAT_96S08 0x00020000 +#define WAVE_FORMAT_96M16 0x00040000 +#define WAVE_FORMAT_96S16 0x00080000 +#endif + +#ifndef SPEAKER_FRONT_LEFT +#define SPEAKER_FRONT_LEFT 0x1 +#define SPEAKER_FRONT_RIGHT 0x2 +#define SPEAKER_FRONT_CENTER 0x4 +#define SPEAKER_LOW_FREQUENCY 0x8 +#define SPEAKER_BACK_LEFT 0x10 +#define SPEAKER_BACK_RIGHT 0x20 +#define SPEAKER_FRONT_LEFT_OF_CENTER 0x40 +#define SPEAKER_FRONT_RIGHT_OF_CENTER 0x80 +#define SPEAKER_BACK_CENTER 0x100 +#define SPEAKER_SIDE_LEFT 0x200 +#define SPEAKER_SIDE_RIGHT 0x400 +#define SPEAKER_TOP_CENTER 0x800 +#define SPEAKER_TOP_FRONT_LEFT 0x1000 +#define SPEAKER_TOP_FRONT_CENTER 0x2000 +#define SPEAKER_TOP_FRONT_RIGHT 0x4000 +#define SPEAKER_TOP_BACK_LEFT 0x8000 +#define SPEAKER_TOP_BACK_CENTER 0x10000 +#define SPEAKER_TOP_BACK_RIGHT 0x20000 +#endif + +/* +Implement our own version of MA_WAVEFORMATEXTENSIBLE so we can avoid a header. Be careful with this +because MA_WAVEFORMATEX has an extra two bytes over standard WAVEFORMATEX due to padding. The +standard version uses tight packing, but for compiler compatibility we're not doing that with ours. +*/ +typedef struct +{ + WORD wFormatTag; + WORD nChannels; + DWORD nSamplesPerSec; + DWORD nAvgBytesPerSec; + WORD nBlockAlign; + WORD wBitsPerSample; + WORD cbSize; +} MA_WAVEFORMATEX; + +typedef struct +{ + WORD wFormatTag; + WORD nChannels; + DWORD nSamplesPerSec; + DWORD nAvgBytesPerSec; + WORD nBlockAlign; + WORD wBitsPerSample; + WORD cbSize; + union + { + WORD wValidBitsPerSample; + WORD wSamplesPerBlock; + WORD wReserved; + } Samples; + DWORD dwChannelMask; + GUID SubFormat; +} MA_WAVEFORMATEXTENSIBLE; + + + +#ifndef WAVE_FORMAT_EXTENSIBLE +#define WAVE_FORMAT_EXTENSIBLE 0xFFFE +#endif + +#ifndef WAVE_FORMAT_PCM +#define WAVE_FORMAT_PCM 1 +#endif + +#ifndef WAVE_FORMAT_IEEE_FLOAT +#define WAVE_FORMAT_IEEE_FLOAT 0x0003 +#endif + +/* Converts an individual Win32-style channel identifier (SPEAKER_FRONT_LEFT, etc.) to miniaudio. */ +static ma_uint8 ma_channel_id_to_ma__win32(DWORD id) +{ + switch (id) + { + case SPEAKER_FRONT_LEFT: return MA_CHANNEL_FRONT_LEFT; + case SPEAKER_FRONT_RIGHT: return MA_CHANNEL_FRONT_RIGHT; + case SPEAKER_FRONT_CENTER: return MA_CHANNEL_FRONT_CENTER; + case SPEAKER_LOW_FREQUENCY: return MA_CHANNEL_LFE; + case SPEAKER_BACK_LEFT: return MA_CHANNEL_BACK_LEFT; + case SPEAKER_BACK_RIGHT: return MA_CHANNEL_BACK_RIGHT; + case SPEAKER_FRONT_LEFT_OF_CENTER: return MA_CHANNEL_FRONT_LEFT_CENTER; + case SPEAKER_FRONT_RIGHT_OF_CENTER: return MA_CHANNEL_FRONT_RIGHT_CENTER; + case SPEAKER_BACK_CENTER: return MA_CHANNEL_BACK_CENTER; + case SPEAKER_SIDE_LEFT: return MA_CHANNEL_SIDE_LEFT; + case SPEAKER_SIDE_RIGHT: return MA_CHANNEL_SIDE_RIGHT; + case SPEAKER_TOP_CENTER: return MA_CHANNEL_TOP_CENTER; + case SPEAKER_TOP_FRONT_LEFT: return MA_CHANNEL_TOP_FRONT_LEFT; + case SPEAKER_TOP_FRONT_CENTER: return MA_CHANNEL_TOP_FRONT_CENTER; + case SPEAKER_TOP_FRONT_RIGHT: return MA_CHANNEL_TOP_FRONT_RIGHT; + case SPEAKER_TOP_BACK_LEFT: return MA_CHANNEL_TOP_BACK_LEFT; + case SPEAKER_TOP_BACK_CENTER: return MA_CHANNEL_TOP_BACK_CENTER; + case SPEAKER_TOP_BACK_RIGHT: return MA_CHANNEL_TOP_BACK_RIGHT; + default: return 0; + } +} + +/* Converts an individual miniaudio channel identifier (MA_CHANNEL_FRONT_LEFT, etc.) to Win32-style. */ +static DWORD ma_channel_id_to_win32(DWORD id) +{ + switch (id) + { + case MA_CHANNEL_MONO: return SPEAKER_FRONT_CENTER; + case MA_CHANNEL_FRONT_LEFT: return SPEAKER_FRONT_LEFT; + case MA_CHANNEL_FRONT_RIGHT: return SPEAKER_FRONT_RIGHT; + case MA_CHANNEL_FRONT_CENTER: return SPEAKER_FRONT_CENTER; + case MA_CHANNEL_LFE: return SPEAKER_LOW_FREQUENCY; + case MA_CHANNEL_BACK_LEFT: return SPEAKER_BACK_LEFT; + case MA_CHANNEL_BACK_RIGHT: return SPEAKER_BACK_RIGHT; + case MA_CHANNEL_FRONT_LEFT_CENTER: return SPEAKER_FRONT_LEFT_OF_CENTER; + case MA_CHANNEL_FRONT_RIGHT_CENTER: return SPEAKER_FRONT_RIGHT_OF_CENTER; + case MA_CHANNEL_BACK_CENTER: return SPEAKER_BACK_CENTER; + case MA_CHANNEL_SIDE_LEFT: return SPEAKER_SIDE_LEFT; + case MA_CHANNEL_SIDE_RIGHT: return SPEAKER_SIDE_RIGHT; + case MA_CHANNEL_TOP_CENTER: return SPEAKER_TOP_CENTER; + case MA_CHANNEL_TOP_FRONT_LEFT: return SPEAKER_TOP_FRONT_LEFT; + case MA_CHANNEL_TOP_FRONT_CENTER: return SPEAKER_TOP_FRONT_CENTER; + case MA_CHANNEL_TOP_FRONT_RIGHT: return SPEAKER_TOP_FRONT_RIGHT; + case MA_CHANNEL_TOP_BACK_LEFT: return SPEAKER_TOP_BACK_LEFT; + case MA_CHANNEL_TOP_BACK_CENTER: return SPEAKER_TOP_BACK_CENTER; + case MA_CHANNEL_TOP_BACK_RIGHT: return SPEAKER_TOP_BACK_RIGHT; + default: return 0; + } +} + +/* Converts a channel mapping to a Win32-style channel mask. */ +static DWORD ma_channel_map_to_channel_mask__win32(const ma_channel* pChannelMap, ma_uint32 channels) +{ + DWORD dwChannelMask = 0; + ma_uint32 iChannel; + + for (iChannel = 0; iChannel < channels; ++iChannel) { + dwChannelMask |= ma_channel_id_to_win32(pChannelMap[iChannel]); + } + + return dwChannelMask; +} + +/* Converts a Win32-style channel mask to a miniaudio channel map. */ +static void ma_channel_mask_to_channel_map__win32(DWORD dwChannelMask, ma_uint32 channels, ma_channel* pChannelMap) +{ + /* If the channel mask is set to 0, just assume a default Win32 channel map. */ + if (dwChannelMask == 0) { + ma_channel_map_init_standard(ma_standard_channel_map_microsoft, pChannelMap, channels, channels); + } else { + if (channels == 1 && (dwChannelMask & SPEAKER_FRONT_CENTER) != 0) { + pChannelMap[0] = MA_CHANNEL_MONO; + } else { + /* Just iterate over each bit. */ + ma_uint32 iChannel = 0; + ma_uint32 iBit; + + for (iBit = 0; iBit < 32 && iChannel < channels; ++iBit) { + DWORD bitValue = (dwChannelMask & (1UL << iBit)); + if (bitValue != 0) { + /* The bit is set. */ + pChannelMap[iChannel] = ma_channel_id_to_ma__win32(bitValue); + iChannel += 1; + } + } + } + } +} + +#ifdef __cplusplus +static ma_bool32 ma_is_guid_equal(const void* a, const void* b) +{ + return IsEqualGUID(*(const GUID*)a, *(const GUID*)b); +} +#else +#define ma_is_guid_equal(a, b) IsEqualGUID((const GUID*)a, (const GUID*)b) +#endif + +static MA_INLINE ma_bool32 ma_is_guid_null(const void* guid) +{ + static GUID nullguid = {0x00000000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}}; + return ma_is_guid_equal(guid, &nullguid); +} + +static ma_format ma_format_from_WAVEFORMATEX(const MA_WAVEFORMATEX* pWF) +{ + MA_ASSERT(pWF != NULL); + + if (pWF->wFormatTag == WAVE_FORMAT_EXTENSIBLE) { + const MA_WAVEFORMATEXTENSIBLE* pWFEX = (const MA_WAVEFORMATEXTENSIBLE*)pWF; + if (ma_is_guid_equal(&pWFEX->SubFormat, &MA_GUID_KSDATAFORMAT_SUBTYPE_PCM)) { + if (pWFEX->Samples.wValidBitsPerSample == 32) { + return ma_format_s32; + } + if (pWFEX->Samples.wValidBitsPerSample == 24) { + if (pWFEX->wBitsPerSample == 32) { + return ma_format_s32; + } + if (pWFEX->wBitsPerSample == 24) { + return ma_format_s24; + } + } + if (pWFEX->Samples.wValidBitsPerSample == 16) { + return ma_format_s16; + } + if (pWFEX->Samples.wValidBitsPerSample == 8) { + return ma_format_u8; + } + } + if (ma_is_guid_equal(&pWFEX->SubFormat, &MA_GUID_KSDATAFORMAT_SUBTYPE_IEEE_FLOAT)) { + if (pWFEX->Samples.wValidBitsPerSample == 32) { + return ma_format_f32; + } + /* + if (pWFEX->Samples.wValidBitsPerSample == 64) { + return ma_format_f64; + } + */ + } + } else { + if (pWF->wFormatTag == WAVE_FORMAT_PCM) { + if (pWF->wBitsPerSample == 32) { + return ma_format_s32; + } + if (pWF->wBitsPerSample == 24) { + return ma_format_s24; + } + if (pWF->wBitsPerSample == 16) { + return ma_format_s16; + } + if (pWF->wBitsPerSample == 8) { + return ma_format_u8; + } + } + if (pWF->wFormatTag == WAVE_FORMAT_IEEE_FLOAT) { + if (pWF->wBitsPerSample == 32) { + return ma_format_f32; + } + if (pWF->wBitsPerSample == 64) { + /*return ma_format_f64;*/ + } + } + } + + return ma_format_unknown; +} +#endif + + +/******************************************************************************* + +WASAPI Backend + +*******************************************************************************/ +#ifdef MA_HAS_WASAPI +#if 0 +#if defined(_MSC_VER) + #pragma warning(push) + #pragma warning(disable:4091) /* 'typedef ': ignored on left of '' when no variable is declared */ +#endif +#include +#include +#if defined(_MSC_VER) + #pragma warning(pop) +#endif +#endif /* 0 */ + +static ma_result ma_device_reroute__wasapi(ma_device* pDevice, ma_device_type deviceType); + +/* Some compilers don't define VerifyVersionInfoW. Need to write this ourselves. */ +#define MA_WIN32_WINNT_VISTA 0x0600 +#define MA_VER_MINORVERSION 0x01 +#define MA_VER_MAJORVERSION 0x02 +#define MA_VER_SERVICEPACKMAJOR 0x20 +#define MA_VER_GREATER_EQUAL 0x03 + +typedef struct { + DWORD dwOSVersionInfoSize; + DWORD dwMajorVersion; + DWORD dwMinorVersion; + DWORD dwBuildNumber; + DWORD dwPlatformId; + WCHAR szCSDVersion[128]; + WORD wServicePackMajor; + WORD wServicePackMinor; + WORD wSuiteMask; + BYTE wProductType; + BYTE wReserved; +} ma_OSVERSIONINFOEXW; + +typedef BOOL (WINAPI * ma_PFNVerifyVersionInfoW) (ma_OSVERSIONINFOEXW* lpVersionInfo, DWORD dwTypeMask, DWORDLONG dwlConditionMask); +typedef ULONGLONG (WINAPI * ma_PFNVerSetConditionMask)(ULONGLONG dwlConditionMask, DWORD dwTypeBitMask, BYTE dwConditionMask); + + +#ifndef PROPERTYKEY_DEFINED +#define PROPERTYKEY_DEFINED +#ifndef __WATCOMC__ +typedef struct +{ + GUID fmtid; + DWORD pid; +} PROPERTYKEY; +#endif +#endif + +/* Some compilers don't define PropVariantInit(). We just do this ourselves since it's just a memset(). */ +static MA_INLINE void ma_PropVariantInit(MA_PROPVARIANT* pProp) +{ + MA_ZERO_OBJECT(pProp); +} + + +static const PROPERTYKEY MA_PKEY_Device_FriendlyName = {{0xA45C254E, 0xDF1C, 0x4EFD, {0x80, 0x20, 0x67, 0xD1, 0x46, 0xA8, 0x50, 0xE0}}, 14}; +static const PROPERTYKEY MA_PKEY_AudioEngine_DeviceFormat = {{0xF19F064D, 0x82C, 0x4E27, {0xBC, 0x73, 0x68, 0x82, 0xA1, 0xBB, 0x8E, 0x4C}}, 0}; + +static const IID MA_IID_IUnknown = {0x00000000, 0x0000, 0x0000, {0xC0, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x46}}; /* 00000000-0000-0000-C000-000000000046 */ +#if !defined(MA_WIN32_DESKTOP) && !defined(MA_WIN32_GDK) +static const IID MA_IID_IAgileObject = {0x94EA2B94, 0xE9CC, 0x49E0, {0xC0, 0xFF, 0xEE, 0x64, 0xCA, 0x8F, 0x5B, 0x90}}; /* 94EA2B94-E9CC-49E0-C0FF-EE64CA8F5B90 */ +#endif + +static const IID MA_IID_IAudioClient = {0x1CB9AD4C, 0xDBFA, 0x4C32, {0xB1, 0x78, 0xC2, 0xF5, 0x68, 0xA7, 0x03, 0xB2}}; /* 1CB9AD4C-DBFA-4C32-B178-C2F568A703B2 = __uuidof(IAudioClient) */ +static const IID MA_IID_IAudioClient2 = {0x726778CD, 0xF60A, 0x4EDA, {0x82, 0xDE, 0xE4, 0x76, 0x10, 0xCD, 0x78, 0xAA}}; /* 726778CD-F60A-4EDA-82DE-E47610CD78AA = __uuidof(IAudioClient2) */ +static const IID MA_IID_IAudioClient3 = {0x7ED4EE07, 0x8E67, 0x4CD4, {0x8C, 0x1A, 0x2B, 0x7A, 0x59, 0x87, 0xAD, 0x42}}; /* 7ED4EE07-8E67-4CD4-8C1A-2B7A5987AD42 = __uuidof(IAudioClient3) */ +static const IID MA_IID_IAudioRenderClient = {0xF294ACFC, 0x3146, 0x4483, {0xA7, 0xBF, 0xAD, 0xDC, 0xA7, 0xC2, 0x60, 0xE2}}; /* F294ACFC-3146-4483-A7BF-ADDCA7C260E2 = __uuidof(IAudioRenderClient) */ +static const IID MA_IID_IAudioCaptureClient = {0xC8ADBD64, 0xE71E, 0x48A0, {0xA4, 0xDE, 0x18, 0x5C, 0x39, 0x5C, 0xD3, 0x17}}; /* C8ADBD64-E71E-48A0-A4DE-185C395CD317 = __uuidof(IAudioCaptureClient) */ +static const IID MA_IID_IMMNotificationClient = {0x7991EEC9, 0x7E89, 0x4D85, {0x83, 0x90, 0x6C, 0x70, 0x3C, 0xEC, 0x60, 0xC0}}; /* 7991EEC9-7E89-4D85-8390-6C703CEC60C0 = __uuidof(IMMNotificationClient) */ +#if !defined(MA_WIN32_DESKTOP) && !defined(MA_WIN32_GDK) +static const IID MA_IID_DEVINTERFACE_AUDIO_RENDER = {0xE6327CAD, 0xDCEC, 0x4949, {0xAE, 0x8A, 0x99, 0x1E, 0x97, 0x6A, 0x79, 0xD2}}; /* E6327CAD-DCEC-4949-AE8A-991E976A79D2 */ +static const IID MA_IID_DEVINTERFACE_AUDIO_CAPTURE = {0x2EEF81BE, 0x33FA, 0x4800, {0x96, 0x70, 0x1C, 0xD4, 0x74, 0x97, 0x2C, 0x3F}}; /* 2EEF81BE-33FA-4800-9670-1CD474972C3F */ +static const IID MA_IID_IActivateAudioInterfaceCompletionHandler = {0x41D949AB, 0x9862, 0x444A, {0x80, 0xF6, 0xC2, 0x61, 0x33, 0x4D, 0xA5, 0xEB}}; /* 41D949AB-9862-444A-80F6-C261334DA5EB */ +#endif + +static const IID MA_CLSID_MMDeviceEnumerator = {0xBCDE0395, 0xE52F, 0x467C, {0x8E, 0x3D, 0xC4, 0x57, 0x92, 0x91, 0x69, 0x2E}}; /* BCDE0395-E52F-467C-8E3D-C4579291692E = __uuidof(MMDeviceEnumerator) */ +static const IID MA_IID_IMMDeviceEnumerator = {0xA95664D2, 0x9614, 0x4F35, {0xA7, 0x46, 0xDE, 0x8D, 0xB6, 0x36, 0x17, 0xE6}}; /* A95664D2-9614-4F35-A746-DE8DB63617E6 = __uuidof(IMMDeviceEnumerator) */ + +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) +#define MA_MM_DEVICE_STATE_ACTIVE 1 +#define MA_MM_DEVICE_STATE_DISABLED 2 +#define MA_MM_DEVICE_STATE_NOTPRESENT 4 +#define MA_MM_DEVICE_STATE_UNPLUGGED 8 + +typedef struct ma_IMMDeviceEnumerator ma_IMMDeviceEnumerator; +typedef struct ma_IMMDeviceCollection ma_IMMDeviceCollection; +typedef struct ma_IMMDevice ma_IMMDevice; +#else +typedef struct ma_IActivateAudioInterfaceCompletionHandler ma_IActivateAudioInterfaceCompletionHandler; +typedef struct ma_IActivateAudioInterfaceAsyncOperation ma_IActivateAudioInterfaceAsyncOperation; +#endif +typedef struct ma_IPropertyStore ma_IPropertyStore; +typedef struct ma_IAudioClient ma_IAudioClient; +typedef struct ma_IAudioClient2 ma_IAudioClient2; +typedef struct ma_IAudioClient3 ma_IAudioClient3; +typedef struct ma_IAudioRenderClient ma_IAudioRenderClient; +typedef struct ma_IAudioCaptureClient ma_IAudioCaptureClient; + +typedef ma_int64 MA_REFERENCE_TIME; + +#define MA_AUDCLNT_STREAMFLAGS_CROSSPROCESS 0x00010000 +#define MA_AUDCLNT_STREAMFLAGS_LOOPBACK 0x00020000 +#define MA_AUDCLNT_STREAMFLAGS_EVENTCALLBACK 0x00040000 +#define MA_AUDCLNT_STREAMFLAGS_NOPERSIST 0x00080000 +#define MA_AUDCLNT_STREAMFLAGS_RATEADJUST 0x00100000 +#define MA_AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY 0x08000000 +#define MA_AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM 0x80000000 +#define MA_AUDCLNT_SESSIONFLAGS_EXPIREWHENUNOWNED 0x10000000 +#define MA_AUDCLNT_SESSIONFLAGS_DISPLAY_HIDE 0x20000000 +#define MA_AUDCLNT_SESSIONFLAGS_DISPLAY_HIDEWHENEXPIRED 0x40000000 + +/* Buffer flags. */ +#define MA_AUDCLNT_BUFFERFLAGS_DATA_DISCONTINUITY 1 +#define MA_AUDCLNT_BUFFERFLAGS_SILENT 2 +#define MA_AUDCLNT_BUFFERFLAGS_TIMESTAMP_ERROR 4 + +typedef enum +{ + ma_eRender = 0, + ma_eCapture = 1, + ma_eAll = 2 +} ma_EDataFlow; + +typedef enum +{ + ma_eConsole = 0, + ma_eMultimedia = 1, + ma_eCommunications = 2 +} ma_ERole; + +typedef enum +{ + MA_AUDCLNT_SHAREMODE_SHARED, + MA_AUDCLNT_SHAREMODE_EXCLUSIVE +} MA_AUDCLNT_SHAREMODE; + +typedef enum +{ + MA_AudioCategory_Other = 0 /* <-- miniaudio is only caring about Other. */ +} MA_AUDIO_STREAM_CATEGORY; + +typedef struct +{ + ma_uint32 cbSize; + BOOL bIsOffload; + MA_AUDIO_STREAM_CATEGORY eCategory; +} ma_AudioClientProperties; + +/* IUnknown */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IUnknown* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IUnknown* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IUnknown* pThis); +} ma_IUnknownVtbl; +struct ma_IUnknown +{ + ma_IUnknownVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IUnknown_QueryInterface(ma_IUnknown* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IUnknown_AddRef(ma_IUnknown* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IUnknown_Release(ma_IUnknown* pThis) { return pThis->lpVtbl->Release(pThis); } + +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + /* IMMNotificationClient */ + typedef struct + { + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IMMNotificationClient* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IMMNotificationClient* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IMMNotificationClient* pThis); + + /* IMMNotificationClient */ + HRESULT (STDMETHODCALLTYPE * OnDeviceStateChanged) (ma_IMMNotificationClient* pThis, const WCHAR* pDeviceID, DWORD dwNewState); + HRESULT (STDMETHODCALLTYPE * OnDeviceAdded) (ma_IMMNotificationClient* pThis, const WCHAR* pDeviceID); + HRESULT (STDMETHODCALLTYPE * OnDeviceRemoved) (ma_IMMNotificationClient* pThis, const WCHAR* pDeviceID); + HRESULT (STDMETHODCALLTYPE * OnDefaultDeviceChanged)(ma_IMMNotificationClient* pThis, ma_EDataFlow dataFlow, ma_ERole role, const WCHAR* pDefaultDeviceID); + HRESULT (STDMETHODCALLTYPE * OnPropertyValueChanged)(ma_IMMNotificationClient* pThis, const WCHAR* pDeviceID, const PROPERTYKEY key); + } ma_IMMNotificationClientVtbl; + + /* IMMDeviceEnumerator */ + typedef struct + { + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IMMDeviceEnumerator* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IMMDeviceEnumerator* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IMMDeviceEnumerator* pThis); + + /* IMMDeviceEnumerator */ + HRESULT (STDMETHODCALLTYPE * EnumAudioEndpoints) (ma_IMMDeviceEnumerator* pThis, ma_EDataFlow dataFlow, DWORD dwStateMask, ma_IMMDeviceCollection** ppDevices); + HRESULT (STDMETHODCALLTYPE * GetDefaultAudioEndpoint) (ma_IMMDeviceEnumerator* pThis, ma_EDataFlow dataFlow, ma_ERole role, ma_IMMDevice** ppEndpoint); + HRESULT (STDMETHODCALLTYPE * GetDevice) (ma_IMMDeviceEnumerator* pThis, const WCHAR* pID, ma_IMMDevice** ppDevice); + HRESULT (STDMETHODCALLTYPE * RegisterEndpointNotificationCallback) (ma_IMMDeviceEnumerator* pThis, ma_IMMNotificationClient* pClient); + HRESULT (STDMETHODCALLTYPE * UnregisterEndpointNotificationCallback)(ma_IMMDeviceEnumerator* pThis, ma_IMMNotificationClient* pClient); + } ma_IMMDeviceEnumeratorVtbl; + struct ma_IMMDeviceEnumerator + { + ma_IMMDeviceEnumeratorVtbl* lpVtbl; + }; + static MA_INLINE HRESULT ma_IMMDeviceEnumerator_QueryInterface(ma_IMMDeviceEnumerator* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } + static MA_INLINE ULONG ma_IMMDeviceEnumerator_AddRef(ma_IMMDeviceEnumerator* pThis) { return pThis->lpVtbl->AddRef(pThis); } + static MA_INLINE ULONG ma_IMMDeviceEnumerator_Release(ma_IMMDeviceEnumerator* pThis) { return pThis->lpVtbl->Release(pThis); } + static MA_INLINE HRESULT ma_IMMDeviceEnumerator_EnumAudioEndpoints(ma_IMMDeviceEnumerator* pThis, ma_EDataFlow dataFlow, DWORD dwStateMask, ma_IMMDeviceCollection** ppDevices) { return pThis->lpVtbl->EnumAudioEndpoints(pThis, dataFlow, dwStateMask, ppDevices); } + static MA_INLINE HRESULT ma_IMMDeviceEnumerator_GetDefaultAudioEndpoint(ma_IMMDeviceEnumerator* pThis, ma_EDataFlow dataFlow, ma_ERole role, ma_IMMDevice** ppEndpoint) { return pThis->lpVtbl->GetDefaultAudioEndpoint(pThis, dataFlow, role, ppEndpoint); } + static MA_INLINE HRESULT ma_IMMDeviceEnumerator_GetDevice(ma_IMMDeviceEnumerator* pThis, const WCHAR* pID, ma_IMMDevice** ppDevice) { return pThis->lpVtbl->GetDevice(pThis, pID, ppDevice); } + static MA_INLINE HRESULT ma_IMMDeviceEnumerator_RegisterEndpointNotificationCallback(ma_IMMDeviceEnumerator* pThis, ma_IMMNotificationClient* pClient) { return pThis->lpVtbl->RegisterEndpointNotificationCallback(pThis, pClient); } + static MA_INLINE HRESULT ma_IMMDeviceEnumerator_UnregisterEndpointNotificationCallback(ma_IMMDeviceEnumerator* pThis, ma_IMMNotificationClient* pClient) { return pThis->lpVtbl->UnregisterEndpointNotificationCallback(pThis, pClient); } + + + /* IMMDeviceCollection */ + typedef struct + { + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IMMDeviceCollection* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IMMDeviceCollection* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IMMDeviceCollection* pThis); + + /* IMMDeviceCollection */ + HRESULT (STDMETHODCALLTYPE * GetCount)(ma_IMMDeviceCollection* pThis, UINT* pDevices); + HRESULT (STDMETHODCALLTYPE * Item) (ma_IMMDeviceCollection* pThis, UINT nDevice, ma_IMMDevice** ppDevice); + } ma_IMMDeviceCollectionVtbl; + struct ma_IMMDeviceCollection + { + ma_IMMDeviceCollectionVtbl* lpVtbl; + }; + static MA_INLINE HRESULT ma_IMMDeviceCollection_QueryInterface(ma_IMMDeviceCollection* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } + static MA_INLINE ULONG ma_IMMDeviceCollection_AddRef(ma_IMMDeviceCollection* pThis) { return pThis->lpVtbl->AddRef(pThis); } + static MA_INLINE ULONG ma_IMMDeviceCollection_Release(ma_IMMDeviceCollection* pThis) { return pThis->lpVtbl->Release(pThis); } + static MA_INLINE HRESULT ma_IMMDeviceCollection_GetCount(ma_IMMDeviceCollection* pThis, UINT* pDevices) { return pThis->lpVtbl->GetCount(pThis, pDevices); } + static MA_INLINE HRESULT ma_IMMDeviceCollection_Item(ma_IMMDeviceCollection* pThis, UINT nDevice, ma_IMMDevice** ppDevice) { return pThis->lpVtbl->Item(pThis, nDevice, ppDevice); } + + + /* IMMDevice */ + typedef struct + { + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IMMDevice* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IMMDevice* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IMMDevice* pThis); + + /* IMMDevice */ + HRESULT (STDMETHODCALLTYPE * Activate) (ma_IMMDevice* pThis, const IID* const iid, DWORD dwClsCtx, MA_PROPVARIANT* pActivationParams, void** ppInterface); + HRESULT (STDMETHODCALLTYPE * OpenPropertyStore)(ma_IMMDevice* pThis, DWORD stgmAccess, ma_IPropertyStore** ppProperties); + HRESULT (STDMETHODCALLTYPE * GetId) (ma_IMMDevice* pThis, WCHAR** pID); + HRESULT (STDMETHODCALLTYPE * GetState) (ma_IMMDevice* pThis, DWORD *pState); + } ma_IMMDeviceVtbl; + struct ma_IMMDevice + { + ma_IMMDeviceVtbl* lpVtbl; + }; + static MA_INLINE HRESULT ma_IMMDevice_QueryInterface(ma_IMMDevice* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } + static MA_INLINE ULONG ma_IMMDevice_AddRef(ma_IMMDevice* pThis) { return pThis->lpVtbl->AddRef(pThis); } + static MA_INLINE ULONG ma_IMMDevice_Release(ma_IMMDevice* pThis) { return pThis->lpVtbl->Release(pThis); } + static MA_INLINE HRESULT ma_IMMDevice_Activate(ma_IMMDevice* pThis, const IID* const iid, DWORD dwClsCtx, MA_PROPVARIANT* pActivationParams, void** ppInterface) { return pThis->lpVtbl->Activate(pThis, iid, dwClsCtx, pActivationParams, ppInterface); } + static MA_INLINE HRESULT ma_IMMDevice_OpenPropertyStore(ma_IMMDevice* pThis, DWORD stgmAccess, ma_IPropertyStore** ppProperties) { return pThis->lpVtbl->OpenPropertyStore(pThis, stgmAccess, ppProperties); } + static MA_INLINE HRESULT ma_IMMDevice_GetId(ma_IMMDevice* pThis, WCHAR** pID) { return pThis->lpVtbl->GetId(pThis, pID); } + static MA_INLINE HRESULT ma_IMMDevice_GetState(ma_IMMDevice* pThis, DWORD *pState) { return pThis->lpVtbl->GetState(pThis, pState); } +#else + /* IActivateAudioInterfaceAsyncOperation */ + typedef struct + { + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IActivateAudioInterfaceAsyncOperation* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IActivateAudioInterfaceAsyncOperation* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IActivateAudioInterfaceAsyncOperation* pThis); + + /* IActivateAudioInterfaceAsyncOperation */ + HRESULT (STDMETHODCALLTYPE * GetActivateResult)(ma_IActivateAudioInterfaceAsyncOperation* pThis, HRESULT *pActivateResult, ma_IUnknown** ppActivatedInterface); + } ma_IActivateAudioInterfaceAsyncOperationVtbl; + struct ma_IActivateAudioInterfaceAsyncOperation + { + ma_IActivateAudioInterfaceAsyncOperationVtbl* lpVtbl; + }; + static MA_INLINE HRESULT ma_IActivateAudioInterfaceAsyncOperation_QueryInterface(ma_IActivateAudioInterfaceAsyncOperation* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } + static MA_INLINE ULONG ma_IActivateAudioInterfaceAsyncOperation_AddRef(ma_IActivateAudioInterfaceAsyncOperation* pThis) { return pThis->lpVtbl->AddRef(pThis); } + static MA_INLINE ULONG ma_IActivateAudioInterfaceAsyncOperation_Release(ma_IActivateAudioInterfaceAsyncOperation* pThis) { return pThis->lpVtbl->Release(pThis); } + static MA_INLINE HRESULT ma_IActivateAudioInterfaceAsyncOperation_GetActivateResult(ma_IActivateAudioInterfaceAsyncOperation* pThis, HRESULT *pActivateResult, ma_IUnknown** ppActivatedInterface) { return pThis->lpVtbl->GetActivateResult(pThis, pActivateResult, ppActivatedInterface); } +#endif + +/* IPropertyStore */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IPropertyStore* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IPropertyStore* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IPropertyStore* pThis); + + /* IPropertyStore */ + HRESULT (STDMETHODCALLTYPE * GetCount)(ma_IPropertyStore* pThis, DWORD* pPropCount); + HRESULT (STDMETHODCALLTYPE * GetAt) (ma_IPropertyStore* pThis, DWORD propIndex, PROPERTYKEY* pPropKey); + HRESULT (STDMETHODCALLTYPE * GetValue)(ma_IPropertyStore* pThis, const PROPERTYKEY* const pKey, MA_PROPVARIANT* pPropVar); + HRESULT (STDMETHODCALLTYPE * SetValue)(ma_IPropertyStore* pThis, const PROPERTYKEY* const pKey, const MA_PROPVARIANT* const pPropVar); + HRESULT (STDMETHODCALLTYPE * Commit) (ma_IPropertyStore* pThis); +} ma_IPropertyStoreVtbl; +struct ma_IPropertyStore +{ + ma_IPropertyStoreVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IPropertyStore_QueryInterface(ma_IPropertyStore* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IPropertyStore_AddRef(ma_IPropertyStore* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IPropertyStore_Release(ma_IPropertyStore* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IPropertyStore_GetCount(ma_IPropertyStore* pThis, DWORD* pPropCount) { return pThis->lpVtbl->GetCount(pThis, pPropCount); } +static MA_INLINE HRESULT ma_IPropertyStore_GetAt(ma_IPropertyStore* pThis, DWORD propIndex, PROPERTYKEY* pPropKey) { return pThis->lpVtbl->GetAt(pThis, propIndex, pPropKey); } +static MA_INLINE HRESULT ma_IPropertyStore_GetValue(ma_IPropertyStore* pThis, const PROPERTYKEY* const pKey, MA_PROPVARIANT* pPropVar) { return pThis->lpVtbl->GetValue(pThis, pKey, pPropVar); } +static MA_INLINE HRESULT ma_IPropertyStore_SetValue(ma_IPropertyStore* pThis, const PROPERTYKEY* const pKey, const MA_PROPVARIANT* const pPropVar) { return pThis->lpVtbl->SetValue(pThis, pKey, pPropVar); } +static MA_INLINE HRESULT ma_IPropertyStore_Commit(ma_IPropertyStore* pThis) { return pThis->lpVtbl->Commit(pThis); } + + +/* IAudioClient */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IAudioClient* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IAudioClient* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IAudioClient* pThis); + + /* IAudioClient */ + HRESULT (STDMETHODCALLTYPE * Initialize) (ma_IAudioClient* pThis, MA_AUDCLNT_SHAREMODE shareMode, DWORD streamFlags, MA_REFERENCE_TIME bufferDuration, MA_REFERENCE_TIME periodicity, const MA_WAVEFORMATEX* pFormat, const GUID* pAudioSessionGuid); + HRESULT (STDMETHODCALLTYPE * GetBufferSize) (ma_IAudioClient* pThis, ma_uint32* pNumBufferFrames); + HRESULT (STDMETHODCALLTYPE * GetStreamLatency) (ma_IAudioClient* pThis, MA_REFERENCE_TIME* pLatency); + HRESULT (STDMETHODCALLTYPE * GetCurrentPadding)(ma_IAudioClient* pThis, ma_uint32* pNumPaddingFrames); + HRESULT (STDMETHODCALLTYPE * IsFormatSupported)(ma_IAudioClient* pThis, MA_AUDCLNT_SHAREMODE shareMode, const MA_WAVEFORMATEX* pFormat, MA_WAVEFORMATEX** ppClosestMatch); + HRESULT (STDMETHODCALLTYPE * GetMixFormat) (ma_IAudioClient* pThis, MA_WAVEFORMATEX** ppDeviceFormat); + HRESULT (STDMETHODCALLTYPE * GetDevicePeriod) (ma_IAudioClient* pThis, MA_REFERENCE_TIME* pDefaultDevicePeriod, MA_REFERENCE_TIME* pMinimumDevicePeriod); + HRESULT (STDMETHODCALLTYPE * Start) (ma_IAudioClient* pThis); + HRESULT (STDMETHODCALLTYPE * Stop) (ma_IAudioClient* pThis); + HRESULT (STDMETHODCALLTYPE * Reset) (ma_IAudioClient* pThis); + HRESULT (STDMETHODCALLTYPE * SetEventHandle) (ma_IAudioClient* pThis, HANDLE eventHandle); + HRESULT (STDMETHODCALLTYPE * GetService) (ma_IAudioClient* pThis, const IID* const riid, void** pp); +} ma_IAudioClientVtbl; +struct ma_IAudioClient +{ + ma_IAudioClientVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IAudioClient_QueryInterface(ma_IAudioClient* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IAudioClient_AddRef(ma_IAudioClient* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IAudioClient_Release(ma_IAudioClient* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IAudioClient_Initialize(ma_IAudioClient* pThis, MA_AUDCLNT_SHAREMODE shareMode, DWORD streamFlags, MA_REFERENCE_TIME bufferDuration, MA_REFERENCE_TIME periodicity, const MA_WAVEFORMATEX* pFormat, const GUID* pAudioSessionGuid) { return pThis->lpVtbl->Initialize(pThis, shareMode, streamFlags, bufferDuration, periodicity, pFormat, pAudioSessionGuid); } +static MA_INLINE HRESULT ma_IAudioClient_GetBufferSize(ma_IAudioClient* pThis, ma_uint32* pNumBufferFrames) { return pThis->lpVtbl->GetBufferSize(pThis, pNumBufferFrames); } +static MA_INLINE HRESULT ma_IAudioClient_GetStreamLatency(ma_IAudioClient* pThis, MA_REFERENCE_TIME* pLatency) { return pThis->lpVtbl->GetStreamLatency(pThis, pLatency); } +static MA_INLINE HRESULT ma_IAudioClient_GetCurrentPadding(ma_IAudioClient* pThis, ma_uint32* pNumPaddingFrames) { return pThis->lpVtbl->GetCurrentPadding(pThis, pNumPaddingFrames); } +static MA_INLINE HRESULT ma_IAudioClient_IsFormatSupported(ma_IAudioClient* pThis, MA_AUDCLNT_SHAREMODE shareMode, const MA_WAVEFORMATEX* pFormat, MA_WAVEFORMATEX** ppClosestMatch) { return pThis->lpVtbl->IsFormatSupported(pThis, shareMode, pFormat, ppClosestMatch); } +static MA_INLINE HRESULT ma_IAudioClient_GetMixFormat(ma_IAudioClient* pThis, MA_WAVEFORMATEX** ppDeviceFormat) { return pThis->lpVtbl->GetMixFormat(pThis, ppDeviceFormat); } +static MA_INLINE HRESULT ma_IAudioClient_GetDevicePeriod(ma_IAudioClient* pThis, MA_REFERENCE_TIME* pDefaultDevicePeriod, MA_REFERENCE_TIME* pMinimumDevicePeriod) { return pThis->lpVtbl->GetDevicePeriod(pThis, pDefaultDevicePeriod, pMinimumDevicePeriod); } +static MA_INLINE HRESULT ma_IAudioClient_Start(ma_IAudioClient* pThis) { return pThis->lpVtbl->Start(pThis); } +static MA_INLINE HRESULT ma_IAudioClient_Stop(ma_IAudioClient* pThis) { return pThis->lpVtbl->Stop(pThis); } +static MA_INLINE HRESULT ma_IAudioClient_Reset(ma_IAudioClient* pThis) { return pThis->lpVtbl->Reset(pThis); } +static MA_INLINE HRESULT ma_IAudioClient_SetEventHandle(ma_IAudioClient* pThis, HANDLE eventHandle) { return pThis->lpVtbl->SetEventHandle(pThis, eventHandle); } +static MA_INLINE HRESULT ma_IAudioClient_GetService(ma_IAudioClient* pThis, const IID* const riid, void** pp) { return pThis->lpVtbl->GetService(pThis, riid, pp); } + +/* IAudioClient2 */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IAudioClient2* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IAudioClient2* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IAudioClient2* pThis); + + /* IAudioClient */ + HRESULT (STDMETHODCALLTYPE * Initialize) (ma_IAudioClient2* pThis, MA_AUDCLNT_SHAREMODE shareMode, DWORD streamFlags, MA_REFERENCE_TIME bufferDuration, MA_REFERENCE_TIME periodicity, const MA_WAVEFORMATEX* pFormat, const GUID* pAudioSessionGuid); + HRESULT (STDMETHODCALLTYPE * GetBufferSize) (ma_IAudioClient2* pThis, ma_uint32* pNumBufferFrames); + HRESULT (STDMETHODCALLTYPE * GetStreamLatency) (ma_IAudioClient2* pThis, MA_REFERENCE_TIME* pLatency); + HRESULT (STDMETHODCALLTYPE * GetCurrentPadding)(ma_IAudioClient2* pThis, ma_uint32* pNumPaddingFrames); + HRESULT (STDMETHODCALLTYPE * IsFormatSupported)(ma_IAudioClient2* pThis, MA_AUDCLNT_SHAREMODE shareMode, const MA_WAVEFORMATEX* pFormat, MA_WAVEFORMATEX** ppClosestMatch); + HRESULT (STDMETHODCALLTYPE * GetMixFormat) (ma_IAudioClient2* pThis, MA_WAVEFORMATEX** ppDeviceFormat); + HRESULT (STDMETHODCALLTYPE * GetDevicePeriod) (ma_IAudioClient2* pThis, MA_REFERENCE_TIME* pDefaultDevicePeriod, MA_REFERENCE_TIME* pMinimumDevicePeriod); + HRESULT (STDMETHODCALLTYPE * Start) (ma_IAudioClient2* pThis); + HRESULT (STDMETHODCALLTYPE * Stop) (ma_IAudioClient2* pThis); + HRESULT (STDMETHODCALLTYPE * Reset) (ma_IAudioClient2* pThis); + HRESULT (STDMETHODCALLTYPE * SetEventHandle) (ma_IAudioClient2* pThis, HANDLE eventHandle); + HRESULT (STDMETHODCALLTYPE * GetService) (ma_IAudioClient2* pThis, const IID* const riid, void** pp); + + /* IAudioClient2 */ + HRESULT (STDMETHODCALLTYPE * IsOffloadCapable) (ma_IAudioClient2* pThis, MA_AUDIO_STREAM_CATEGORY category, BOOL* pOffloadCapable); + HRESULT (STDMETHODCALLTYPE * SetClientProperties)(ma_IAudioClient2* pThis, const ma_AudioClientProperties* pProperties); + HRESULT (STDMETHODCALLTYPE * GetBufferSizeLimits)(ma_IAudioClient2* pThis, const MA_WAVEFORMATEX* pFormat, BOOL eventDriven, MA_REFERENCE_TIME* pMinBufferDuration, MA_REFERENCE_TIME* pMaxBufferDuration); +} ma_IAudioClient2Vtbl; +struct ma_IAudioClient2 +{ + ma_IAudioClient2Vtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IAudioClient2_QueryInterface(ma_IAudioClient2* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IAudioClient2_AddRef(ma_IAudioClient2* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IAudioClient2_Release(ma_IAudioClient2* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IAudioClient2_Initialize(ma_IAudioClient2* pThis, MA_AUDCLNT_SHAREMODE shareMode, DWORD streamFlags, MA_REFERENCE_TIME bufferDuration, MA_REFERENCE_TIME periodicity, const MA_WAVEFORMATEX* pFormat, const GUID* pAudioSessionGuid) { return pThis->lpVtbl->Initialize(pThis, shareMode, streamFlags, bufferDuration, periodicity, pFormat, pAudioSessionGuid); } +static MA_INLINE HRESULT ma_IAudioClient2_GetBufferSize(ma_IAudioClient2* pThis, ma_uint32* pNumBufferFrames) { return pThis->lpVtbl->GetBufferSize(pThis, pNumBufferFrames); } +static MA_INLINE HRESULT ma_IAudioClient2_GetStreamLatency(ma_IAudioClient2* pThis, MA_REFERENCE_TIME* pLatency) { return pThis->lpVtbl->GetStreamLatency(pThis, pLatency); } +static MA_INLINE HRESULT ma_IAudioClient2_GetCurrentPadding(ma_IAudioClient2* pThis, ma_uint32* pNumPaddingFrames) { return pThis->lpVtbl->GetCurrentPadding(pThis, pNumPaddingFrames); } +static MA_INLINE HRESULT ma_IAudioClient2_IsFormatSupported(ma_IAudioClient2* pThis, MA_AUDCLNT_SHAREMODE shareMode, const MA_WAVEFORMATEX* pFormat, MA_WAVEFORMATEX** ppClosestMatch) { return pThis->lpVtbl->IsFormatSupported(pThis, shareMode, pFormat, ppClosestMatch); } +static MA_INLINE HRESULT ma_IAudioClient2_GetMixFormat(ma_IAudioClient2* pThis, MA_WAVEFORMATEX** ppDeviceFormat) { return pThis->lpVtbl->GetMixFormat(pThis, ppDeviceFormat); } +static MA_INLINE HRESULT ma_IAudioClient2_GetDevicePeriod(ma_IAudioClient2* pThis, MA_REFERENCE_TIME* pDefaultDevicePeriod, MA_REFERENCE_TIME* pMinimumDevicePeriod) { return pThis->lpVtbl->GetDevicePeriod(pThis, pDefaultDevicePeriod, pMinimumDevicePeriod); } +static MA_INLINE HRESULT ma_IAudioClient2_Start(ma_IAudioClient2* pThis) { return pThis->lpVtbl->Start(pThis); } +static MA_INLINE HRESULT ma_IAudioClient2_Stop(ma_IAudioClient2* pThis) { return pThis->lpVtbl->Stop(pThis); } +static MA_INLINE HRESULT ma_IAudioClient2_Reset(ma_IAudioClient2* pThis) { return pThis->lpVtbl->Reset(pThis); } +static MA_INLINE HRESULT ma_IAudioClient2_SetEventHandle(ma_IAudioClient2* pThis, HANDLE eventHandle) { return pThis->lpVtbl->SetEventHandle(pThis, eventHandle); } +static MA_INLINE HRESULT ma_IAudioClient2_GetService(ma_IAudioClient2* pThis, const IID* const riid, void** pp) { return pThis->lpVtbl->GetService(pThis, riid, pp); } +static MA_INLINE HRESULT ma_IAudioClient2_IsOffloadCapable(ma_IAudioClient2* pThis, MA_AUDIO_STREAM_CATEGORY category, BOOL* pOffloadCapable) { return pThis->lpVtbl->IsOffloadCapable(pThis, category, pOffloadCapable); } +static MA_INLINE HRESULT ma_IAudioClient2_SetClientProperties(ma_IAudioClient2* pThis, const ma_AudioClientProperties* pProperties) { return pThis->lpVtbl->SetClientProperties(pThis, pProperties); } +static MA_INLINE HRESULT ma_IAudioClient2_GetBufferSizeLimits(ma_IAudioClient2* pThis, const MA_WAVEFORMATEX* pFormat, BOOL eventDriven, MA_REFERENCE_TIME* pMinBufferDuration, MA_REFERENCE_TIME* pMaxBufferDuration) { return pThis->lpVtbl->GetBufferSizeLimits(pThis, pFormat, eventDriven, pMinBufferDuration, pMaxBufferDuration); } + + +/* IAudioClient3 */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IAudioClient3* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IAudioClient3* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IAudioClient3* pThis); + + /* IAudioClient */ + HRESULT (STDMETHODCALLTYPE * Initialize) (ma_IAudioClient3* pThis, MA_AUDCLNT_SHAREMODE shareMode, DWORD streamFlags, MA_REFERENCE_TIME bufferDuration, MA_REFERENCE_TIME periodicity, const MA_WAVEFORMATEX* pFormat, const GUID* pAudioSessionGuid); + HRESULT (STDMETHODCALLTYPE * GetBufferSize) (ma_IAudioClient3* pThis, ma_uint32* pNumBufferFrames); + HRESULT (STDMETHODCALLTYPE * GetStreamLatency) (ma_IAudioClient3* pThis, MA_REFERENCE_TIME* pLatency); + HRESULT (STDMETHODCALLTYPE * GetCurrentPadding)(ma_IAudioClient3* pThis, ma_uint32* pNumPaddingFrames); + HRESULT (STDMETHODCALLTYPE * IsFormatSupported)(ma_IAudioClient3* pThis, MA_AUDCLNT_SHAREMODE shareMode, const MA_WAVEFORMATEX* pFormat, MA_WAVEFORMATEX** ppClosestMatch); + HRESULT (STDMETHODCALLTYPE * GetMixFormat) (ma_IAudioClient3* pThis, MA_WAVEFORMATEX** ppDeviceFormat); + HRESULT (STDMETHODCALLTYPE * GetDevicePeriod) (ma_IAudioClient3* pThis, MA_REFERENCE_TIME* pDefaultDevicePeriod, MA_REFERENCE_TIME* pMinimumDevicePeriod); + HRESULT (STDMETHODCALLTYPE * Start) (ma_IAudioClient3* pThis); + HRESULT (STDMETHODCALLTYPE * Stop) (ma_IAudioClient3* pThis); + HRESULT (STDMETHODCALLTYPE * Reset) (ma_IAudioClient3* pThis); + HRESULT (STDMETHODCALLTYPE * SetEventHandle) (ma_IAudioClient3* pThis, HANDLE eventHandle); + HRESULT (STDMETHODCALLTYPE * GetService) (ma_IAudioClient3* pThis, const IID* const riid, void** pp); + + /* IAudioClient2 */ + HRESULT (STDMETHODCALLTYPE * IsOffloadCapable) (ma_IAudioClient3* pThis, MA_AUDIO_STREAM_CATEGORY category, BOOL* pOffloadCapable); + HRESULT (STDMETHODCALLTYPE * SetClientProperties)(ma_IAudioClient3* pThis, const ma_AudioClientProperties* pProperties); + HRESULT (STDMETHODCALLTYPE * GetBufferSizeLimits)(ma_IAudioClient3* pThis, const MA_WAVEFORMATEX* pFormat, BOOL eventDriven, MA_REFERENCE_TIME* pMinBufferDuration, MA_REFERENCE_TIME* pMaxBufferDuration); + + /* IAudioClient3 */ + HRESULT (STDMETHODCALLTYPE * GetSharedModeEnginePeriod) (ma_IAudioClient3* pThis, const MA_WAVEFORMATEX* pFormat, ma_uint32* pDefaultPeriodInFrames, ma_uint32* pFundamentalPeriodInFrames, ma_uint32* pMinPeriodInFrames, ma_uint32* pMaxPeriodInFrames); + HRESULT (STDMETHODCALLTYPE * GetCurrentSharedModeEnginePeriod)(ma_IAudioClient3* pThis, MA_WAVEFORMATEX** ppFormat, ma_uint32* pCurrentPeriodInFrames); + HRESULT (STDMETHODCALLTYPE * InitializeSharedAudioStream) (ma_IAudioClient3* pThis, DWORD streamFlags, ma_uint32 periodInFrames, const MA_WAVEFORMATEX* pFormat, const GUID* pAudioSessionGuid); +} ma_IAudioClient3Vtbl; +struct ma_IAudioClient3 +{ + ma_IAudioClient3Vtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IAudioClient3_QueryInterface(ma_IAudioClient3* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IAudioClient3_AddRef(ma_IAudioClient3* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IAudioClient3_Release(ma_IAudioClient3* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IAudioClient3_Initialize(ma_IAudioClient3* pThis, MA_AUDCLNT_SHAREMODE shareMode, DWORD streamFlags, MA_REFERENCE_TIME bufferDuration, MA_REFERENCE_TIME periodicity, const MA_WAVEFORMATEX* pFormat, const GUID* pAudioSessionGuid) { return pThis->lpVtbl->Initialize(pThis, shareMode, streamFlags, bufferDuration, periodicity, pFormat, pAudioSessionGuid); } +static MA_INLINE HRESULT ma_IAudioClient3_GetBufferSize(ma_IAudioClient3* pThis, ma_uint32* pNumBufferFrames) { return pThis->lpVtbl->GetBufferSize(pThis, pNumBufferFrames); } +static MA_INLINE HRESULT ma_IAudioClient3_GetStreamLatency(ma_IAudioClient3* pThis, MA_REFERENCE_TIME* pLatency) { return pThis->lpVtbl->GetStreamLatency(pThis, pLatency); } +static MA_INLINE HRESULT ma_IAudioClient3_GetCurrentPadding(ma_IAudioClient3* pThis, ma_uint32* pNumPaddingFrames) { return pThis->lpVtbl->GetCurrentPadding(pThis, pNumPaddingFrames); } +static MA_INLINE HRESULT ma_IAudioClient3_IsFormatSupported(ma_IAudioClient3* pThis, MA_AUDCLNT_SHAREMODE shareMode, const MA_WAVEFORMATEX* pFormat, MA_WAVEFORMATEX** ppClosestMatch) { return pThis->lpVtbl->IsFormatSupported(pThis, shareMode, pFormat, ppClosestMatch); } +static MA_INLINE HRESULT ma_IAudioClient3_GetMixFormat(ma_IAudioClient3* pThis, MA_WAVEFORMATEX** ppDeviceFormat) { return pThis->lpVtbl->GetMixFormat(pThis, ppDeviceFormat); } +static MA_INLINE HRESULT ma_IAudioClient3_GetDevicePeriod(ma_IAudioClient3* pThis, MA_REFERENCE_TIME* pDefaultDevicePeriod, MA_REFERENCE_TIME* pMinimumDevicePeriod) { return pThis->lpVtbl->GetDevicePeriod(pThis, pDefaultDevicePeriod, pMinimumDevicePeriod); } +static MA_INLINE HRESULT ma_IAudioClient3_Start(ma_IAudioClient3* pThis) { return pThis->lpVtbl->Start(pThis); } +static MA_INLINE HRESULT ma_IAudioClient3_Stop(ma_IAudioClient3* pThis) { return pThis->lpVtbl->Stop(pThis); } +static MA_INLINE HRESULT ma_IAudioClient3_Reset(ma_IAudioClient3* pThis) { return pThis->lpVtbl->Reset(pThis); } +static MA_INLINE HRESULT ma_IAudioClient3_SetEventHandle(ma_IAudioClient3* pThis, HANDLE eventHandle) { return pThis->lpVtbl->SetEventHandle(pThis, eventHandle); } +static MA_INLINE HRESULT ma_IAudioClient3_GetService(ma_IAudioClient3* pThis, const IID* const riid, void** pp) { return pThis->lpVtbl->GetService(pThis, riid, pp); } +static MA_INLINE HRESULT ma_IAudioClient3_IsOffloadCapable(ma_IAudioClient3* pThis, MA_AUDIO_STREAM_CATEGORY category, BOOL* pOffloadCapable) { return pThis->lpVtbl->IsOffloadCapable(pThis, category, pOffloadCapable); } +static MA_INLINE HRESULT ma_IAudioClient3_SetClientProperties(ma_IAudioClient3* pThis, const ma_AudioClientProperties* pProperties) { return pThis->lpVtbl->SetClientProperties(pThis, pProperties); } +static MA_INLINE HRESULT ma_IAudioClient3_GetBufferSizeLimits(ma_IAudioClient3* pThis, const MA_WAVEFORMATEX* pFormat, BOOL eventDriven, MA_REFERENCE_TIME* pMinBufferDuration, MA_REFERENCE_TIME* pMaxBufferDuration) { return pThis->lpVtbl->GetBufferSizeLimits(pThis, pFormat, eventDriven, pMinBufferDuration, pMaxBufferDuration); } +static MA_INLINE HRESULT ma_IAudioClient3_GetSharedModeEnginePeriod(ma_IAudioClient3* pThis, const MA_WAVEFORMATEX* pFormat, ma_uint32* pDefaultPeriodInFrames, ma_uint32* pFundamentalPeriodInFrames, ma_uint32* pMinPeriodInFrames, ma_uint32* pMaxPeriodInFrames) { return pThis->lpVtbl->GetSharedModeEnginePeriod(pThis, pFormat, pDefaultPeriodInFrames, pFundamentalPeriodInFrames, pMinPeriodInFrames, pMaxPeriodInFrames); } +static MA_INLINE HRESULT ma_IAudioClient3_GetCurrentSharedModeEnginePeriod(ma_IAudioClient3* pThis, MA_WAVEFORMATEX** ppFormat, ma_uint32* pCurrentPeriodInFrames) { return pThis->lpVtbl->GetCurrentSharedModeEnginePeriod(pThis, ppFormat, pCurrentPeriodInFrames); } +static MA_INLINE HRESULT ma_IAudioClient3_InitializeSharedAudioStream(ma_IAudioClient3* pThis, DWORD streamFlags, ma_uint32 periodInFrames, const MA_WAVEFORMATEX* pFormat, const GUID* pAudioSessionGUID) { return pThis->lpVtbl->InitializeSharedAudioStream(pThis, streamFlags, periodInFrames, pFormat, pAudioSessionGUID); } + + +/* IAudioRenderClient */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IAudioRenderClient* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IAudioRenderClient* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IAudioRenderClient* pThis); + + /* IAudioRenderClient */ + HRESULT (STDMETHODCALLTYPE * GetBuffer) (ma_IAudioRenderClient* pThis, ma_uint32 numFramesRequested, BYTE** ppData); + HRESULT (STDMETHODCALLTYPE * ReleaseBuffer)(ma_IAudioRenderClient* pThis, ma_uint32 numFramesWritten, DWORD dwFlags); +} ma_IAudioRenderClientVtbl; +struct ma_IAudioRenderClient +{ + ma_IAudioRenderClientVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IAudioRenderClient_QueryInterface(ma_IAudioRenderClient* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IAudioRenderClient_AddRef(ma_IAudioRenderClient* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IAudioRenderClient_Release(ma_IAudioRenderClient* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IAudioRenderClient_GetBuffer(ma_IAudioRenderClient* pThis, ma_uint32 numFramesRequested, BYTE** ppData) { return pThis->lpVtbl->GetBuffer(pThis, numFramesRequested, ppData); } +static MA_INLINE HRESULT ma_IAudioRenderClient_ReleaseBuffer(ma_IAudioRenderClient* pThis, ma_uint32 numFramesWritten, DWORD dwFlags) { return pThis->lpVtbl->ReleaseBuffer(pThis, numFramesWritten, dwFlags); } + + +/* IAudioCaptureClient */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IAudioCaptureClient* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IAudioCaptureClient* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IAudioCaptureClient* pThis); + + /* IAudioRenderClient */ + HRESULT (STDMETHODCALLTYPE * GetBuffer) (ma_IAudioCaptureClient* pThis, BYTE** ppData, ma_uint32* pNumFramesToRead, DWORD* pFlags, ma_uint64* pDevicePosition, ma_uint64* pQPCPosition); + HRESULT (STDMETHODCALLTYPE * ReleaseBuffer) (ma_IAudioCaptureClient* pThis, ma_uint32 numFramesRead); + HRESULT (STDMETHODCALLTYPE * GetNextPacketSize)(ma_IAudioCaptureClient* pThis, ma_uint32* pNumFramesInNextPacket); +} ma_IAudioCaptureClientVtbl; +struct ma_IAudioCaptureClient +{ + ma_IAudioCaptureClientVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IAudioCaptureClient_QueryInterface(ma_IAudioCaptureClient* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IAudioCaptureClient_AddRef(ma_IAudioCaptureClient* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IAudioCaptureClient_Release(ma_IAudioCaptureClient* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IAudioCaptureClient_GetBuffer(ma_IAudioCaptureClient* pThis, BYTE** ppData, ma_uint32* pNumFramesToRead, DWORD* pFlags, ma_uint64* pDevicePosition, ma_uint64* pQPCPosition) { return pThis->lpVtbl->GetBuffer(pThis, ppData, pNumFramesToRead, pFlags, pDevicePosition, pQPCPosition); } +static MA_INLINE HRESULT ma_IAudioCaptureClient_ReleaseBuffer(ma_IAudioCaptureClient* pThis, ma_uint32 numFramesRead) { return pThis->lpVtbl->ReleaseBuffer(pThis, numFramesRead); } +static MA_INLINE HRESULT ma_IAudioCaptureClient_GetNextPacketSize(ma_IAudioCaptureClient* pThis, ma_uint32* pNumFramesInNextPacket) { return pThis->lpVtbl->GetNextPacketSize(pThis, pNumFramesInNextPacket); } + +#if defined(MA_WIN32_UWP) +/* mmdevapi Functions */ +typedef HRESULT (WINAPI * MA_PFN_ActivateAudioInterfaceAsync)(const wchar_t* deviceInterfacePath, const IID* riid, MA_PROPVARIANT* activationParams, ma_IActivateAudioInterfaceCompletionHandler* completionHandler, ma_IActivateAudioInterfaceAsyncOperation** activationOperation); +#endif + +/* Avrt Functions */ +typedef HANDLE (WINAPI * MA_PFN_AvSetMmThreadCharacteristicsA)(const char* TaskName, DWORD* TaskIndex); +typedef BOOL (WINAPI * MA_PFN_AvRevertMmThreadCharacteristics)(HANDLE AvrtHandle); + +#if !defined(MA_WIN32_DESKTOP) && !defined(MA_WIN32_GDK) +typedef struct ma_completion_handler_uwp ma_completion_handler_uwp; + +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_completion_handler_uwp* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_completion_handler_uwp* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_completion_handler_uwp* pThis); + + /* IActivateAudioInterfaceCompletionHandler */ + HRESULT (STDMETHODCALLTYPE * ActivateCompleted)(ma_completion_handler_uwp* pThis, ma_IActivateAudioInterfaceAsyncOperation* pActivateOperation); +} ma_completion_handler_uwp_vtbl; +struct ma_completion_handler_uwp +{ + ma_completion_handler_uwp_vtbl* lpVtbl; + MA_ATOMIC(4, ma_uint32) counter; + HANDLE hEvent; +}; + +static HRESULT STDMETHODCALLTYPE ma_completion_handler_uwp_QueryInterface(ma_completion_handler_uwp* pThis, const IID* const riid, void** ppObject) +{ + /* + We need to "implement" IAgileObject which is just an indicator that's used internally by WASAPI for some multithreading management. To + "implement" this, we just make sure we return pThis when the IAgileObject is requested. + */ + if (!ma_is_guid_equal(riid, &MA_IID_IUnknown) && !ma_is_guid_equal(riid, &MA_IID_IActivateAudioInterfaceCompletionHandler) && !ma_is_guid_equal(riid, &MA_IID_IAgileObject)) { + *ppObject = NULL; + return E_NOINTERFACE; + } + + /* Getting here means the IID is IUnknown or IMMNotificationClient. */ + *ppObject = (void*)pThis; + ((ma_completion_handler_uwp_vtbl*)pThis->lpVtbl)->AddRef(pThis); + return S_OK; +} + +static ULONG STDMETHODCALLTYPE ma_completion_handler_uwp_AddRef(ma_completion_handler_uwp* pThis) +{ + return (ULONG)ma_atomic_fetch_add_32(&pThis->counter, 1) + 1; +} + +static ULONG STDMETHODCALLTYPE ma_completion_handler_uwp_Release(ma_completion_handler_uwp* pThis) +{ + ma_uint32 newRefCount = ma_atomic_fetch_sub_32(&pThis->counter, 1) - 1; + if (newRefCount == 0) { + return 0; /* We don't free anything here because we never allocate the object on the heap. */ + } + + return (ULONG)newRefCount; +} + +static HRESULT STDMETHODCALLTYPE ma_completion_handler_uwp_ActivateCompleted(ma_completion_handler_uwp* pThis, ma_IActivateAudioInterfaceAsyncOperation* pActivateOperation) +{ + (void)pActivateOperation; + SetEvent(pThis->hEvent); + return S_OK; +} + + +static ma_completion_handler_uwp_vtbl g_maCompletionHandlerVtblInstance = { + ma_completion_handler_uwp_QueryInterface, + ma_completion_handler_uwp_AddRef, + ma_completion_handler_uwp_Release, + ma_completion_handler_uwp_ActivateCompleted +}; + +static ma_result ma_completion_handler_uwp_init(ma_completion_handler_uwp* pHandler) +{ + MA_ASSERT(pHandler != NULL); + MA_ZERO_OBJECT(pHandler); + + pHandler->lpVtbl = &g_maCompletionHandlerVtblInstance; + pHandler->counter = 1; + pHandler->hEvent = CreateEventA(NULL, FALSE, FALSE, NULL); + if (pHandler->hEvent == NULL) { + return ma_result_from_GetLastError(GetLastError()); + } + + return MA_SUCCESS; +} + +static void ma_completion_handler_uwp_uninit(ma_completion_handler_uwp* pHandler) +{ + if (pHandler->hEvent != NULL) { + CloseHandle(pHandler->hEvent); + } +} + +static void ma_completion_handler_uwp_wait(ma_completion_handler_uwp* pHandler) +{ + WaitForSingleObject((HANDLE)pHandler->hEvent, INFINITE); +} +#endif /* !MA_WIN32_DESKTOP */ + +/* We need a virtual table for our notification client object that's used for detecting changes to the default device. */ +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) +static HRESULT STDMETHODCALLTYPE ma_IMMNotificationClient_QueryInterface(ma_IMMNotificationClient* pThis, const IID* const riid, void** ppObject) +{ + /* + We care about two interfaces - IUnknown and IMMNotificationClient. If the requested IID is something else + we just return E_NOINTERFACE. Otherwise we need to increment the reference counter and return S_OK. + */ + if (!ma_is_guid_equal(riid, &MA_IID_IUnknown) && !ma_is_guid_equal(riid, &MA_IID_IMMNotificationClient)) { + *ppObject = NULL; + return E_NOINTERFACE; + } + + /* Getting here means the IID is IUnknown or IMMNotificationClient. */ + *ppObject = (void*)pThis; + ((ma_IMMNotificationClientVtbl*)pThis->lpVtbl)->AddRef(pThis); + return S_OK; +} + +static ULONG STDMETHODCALLTYPE ma_IMMNotificationClient_AddRef(ma_IMMNotificationClient* pThis) +{ + return (ULONG)ma_atomic_fetch_add_32(&pThis->counter, 1) + 1; +} + +static ULONG STDMETHODCALLTYPE ma_IMMNotificationClient_Release(ma_IMMNotificationClient* pThis) +{ + ma_uint32 newRefCount = ma_atomic_fetch_sub_32(&pThis->counter, 1) - 1; + if (newRefCount == 0) { + return 0; /* We don't free anything here because we never allocate the object on the heap. */ + } + + return (ULONG)newRefCount; +} + +static HRESULT STDMETHODCALLTYPE ma_IMMNotificationClient_OnDeviceStateChanged(ma_IMMNotificationClient* pThis, const WCHAR* pDeviceID, DWORD dwNewState) +{ + ma_bool32 isThisDevice = MA_FALSE; + ma_bool32 isCapture = MA_FALSE; + ma_bool32 isPlayback = MA_FALSE; + +#ifdef MA_DEBUG_OUTPUT + /*ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "IMMNotificationClient_OnDeviceStateChanged(pDeviceID=%S, dwNewState=%u)\n", (pDeviceID != NULL) ? pDeviceID : L"(NULL)", (unsigned int)dwNewState);*/ +#endif + + /* + There have been reports of a hang when a playback device is disconnected. The idea with this code is to explicitly stop the device if we detect + that the device is disabled or has been unplugged. + */ + if (pThis->pDevice->wasapi.allowCaptureAutoStreamRouting && (pThis->pDevice->type == ma_device_type_capture || pThis->pDevice->type == ma_device_type_duplex || pThis->pDevice->type == ma_device_type_loopback)) { + isCapture = MA_TRUE; + if (ma_strcmp_WCHAR(pThis->pDevice->capture.id.wasapi, pDeviceID) == 0) { + isThisDevice = MA_TRUE; + } + } + + if (pThis->pDevice->wasapi.allowPlaybackAutoStreamRouting && (pThis->pDevice->type == ma_device_type_playback || pThis->pDevice->type == ma_device_type_duplex)) { + isPlayback = MA_TRUE; + if (ma_strcmp_WCHAR(pThis->pDevice->playback.id.wasapi, pDeviceID) == 0) { + isThisDevice = MA_TRUE; + } + } + + + /* + If the device ID matches our device we need to mark our device as detached and stop it. When a + device is added in OnDeviceAdded(), we'll restart it. We only mark it as detached if the device + was started at the time of being removed. + */ + if (isThisDevice) { + if ((dwNewState & MA_MM_DEVICE_STATE_ACTIVE) == 0) { + /* + Unplugged or otherwise unavailable. Mark as detached if we were in a playing state. We'll + use this to determine whether or not we need to automatically start the device when it's + plugged back in again. + */ + if (ma_device_get_state(pThis->pDevice) == ma_device_state_started) { + if (isPlayback) { + pThis->pDevice->wasapi.isDetachedPlayback = MA_TRUE; + } + if (isCapture) { + pThis->pDevice->wasapi.isDetachedCapture = MA_TRUE; + } + + ma_device_stop(pThis->pDevice); + } + } + + if ((dwNewState & MA_MM_DEVICE_STATE_ACTIVE) != 0) { + /* The device was activated. If we were detached, we need to start it again. */ + ma_bool8 tryRestartingDevice = MA_FALSE; + + if (isPlayback) { + if (pThis->pDevice->wasapi.isDetachedPlayback) { + pThis->pDevice->wasapi.isDetachedPlayback = MA_FALSE; + ma_device_reroute__wasapi(pThis->pDevice, ma_device_type_playback); + tryRestartingDevice = MA_TRUE; + } + } + + if (isCapture) { + if (pThis->pDevice->wasapi.isDetachedCapture) { + pThis->pDevice->wasapi.isDetachedCapture = MA_FALSE; + ma_device_reroute__wasapi(pThis->pDevice, (pThis->pDevice->type == ma_device_type_loopback) ? ma_device_type_loopback : ma_device_type_capture); + tryRestartingDevice = MA_TRUE; + } + } + + if (tryRestartingDevice) { + if (pThis->pDevice->wasapi.isDetachedPlayback == MA_FALSE && pThis->pDevice->wasapi.isDetachedCapture == MA_FALSE) { + ma_device_start(pThis->pDevice); + } + } + } + } + + return S_OK; +} + +static HRESULT STDMETHODCALLTYPE ma_IMMNotificationClient_OnDeviceAdded(ma_IMMNotificationClient* pThis, const WCHAR* pDeviceID) +{ +#ifdef MA_DEBUG_OUTPUT + /*ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "IMMNotificationClient_OnDeviceAdded(pDeviceID=%S)\n", (pDeviceID != NULL) ? pDeviceID : L"(NULL)");*/ +#endif + + /* We don't need to worry about this event for our purposes. */ + (void)pThis; + (void)pDeviceID; + return S_OK; +} + +static HRESULT STDMETHODCALLTYPE ma_IMMNotificationClient_OnDeviceRemoved(ma_IMMNotificationClient* pThis, const WCHAR* pDeviceID) +{ +#ifdef MA_DEBUG_OUTPUT + /*ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "IMMNotificationClient_OnDeviceRemoved(pDeviceID=%S)\n", (pDeviceID != NULL) ? pDeviceID : L"(NULL)");*/ +#endif + + /* We don't need to worry about this event for our purposes. */ + (void)pThis; + (void)pDeviceID; + return S_OK; +} + +static HRESULT STDMETHODCALLTYPE ma_IMMNotificationClient_OnDefaultDeviceChanged(ma_IMMNotificationClient* pThis, ma_EDataFlow dataFlow, ma_ERole role, const WCHAR* pDefaultDeviceID) +{ +#ifdef MA_DEBUG_OUTPUT + /*ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "IMMNotificationClient_OnDefaultDeviceChanged(dataFlow=%d, role=%d, pDefaultDeviceID=%S)\n", dataFlow, role, (pDefaultDeviceID != NULL) ? pDefaultDeviceID : L"(NULL)");*/ +#endif + + (void)role; + + /* We only care about devices with the same data flow as the current device. */ + if ((pThis->pDevice->type == ma_device_type_playback && dataFlow != ma_eRender) || + (pThis->pDevice->type == ma_device_type_capture && dataFlow != ma_eCapture) || + (pThis->pDevice->type == ma_device_type_loopback && dataFlow != ma_eRender)) { + ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Stream rerouting abandoned because dataFlow does match device type.\n"); + return S_OK; + } + + /* We need to consider dataFlow as ma_eCapture if device is ma_device_type_loopback */ + if (pThis->pDevice->type == ma_device_type_loopback) { + dataFlow = ma_eCapture; + } + + /* Don't do automatic stream routing if we're not allowed. */ + if ((dataFlow == ma_eRender && pThis->pDevice->wasapi.allowPlaybackAutoStreamRouting == MA_FALSE) || + (dataFlow == ma_eCapture && pThis->pDevice->wasapi.allowCaptureAutoStreamRouting == MA_FALSE)) { + ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Stream rerouting abandoned because automatic stream routing has been disabled by the device config.\n"); + return S_OK; + } + + /* + Not currently supporting automatic stream routing in exclusive mode. This is not working correctly on my machine due to + AUDCLNT_E_DEVICE_IN_USE errors when reinitializing the device. If this is a bug in miniaudio, we can try re-enabling this once + it's fixed. + */ + if ((dataFlow == ma_eRender && pThis->pDevice->playback.shareMode == ma_share_mode_exclusive) || + (dataFlow == ma_eCapture && pThis->pDevice->capture.shareMode == ma_share_mode_exclusive)) { + ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Stream rerouting abandoned because the device shared mode is exclusive.\n"); + return S_OK; + } + + + + /* + Second attempt at device rerouting. We're going to retrieve the device's state at the time of + the route change. We're then going to stop the device, reinitialize the device, and then start + it again if the state before stopping was ma_device_state_started. + */ + { + ma_uint32 previousState = ma_device_get_state(pThis->pDevice); + ma_bool8 restartDevice = MA_FALSE; + + if (previousState == ma_device_state_uninitialized || previousState == ma_device_state_starting) { + ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Stream rerouting abandoned because the device is in the process of starting.\n"); + return S_OK; + } + + if (previousState == ma_device_state_started) { + ma_device_stop(pThis->pDevice); + restartDevice = MA_TRUE; + } + + if (pDefaultDeviceID != NULL) { /* <-- The input device ID will be null if there's no other device available. */ + ma_mutex_lock(&pThis->pDevice->wasapi.rerouteLock); + { + if (dataFlow == ma_eRender) { + ma_device_reroute__wasapi(pThis->pDevice, ma_device_type_playback); + + if (pThis->pDevice->wasapi.isDetachedPlayback) { + pThis->pDevice->wasapi.isDetachedPlayback = MA_FALSE; + + if (pThis->pDevice->type == ma_device_type_duplex && pThis->pDevice->wasapi.isDetachedCapture) { + restartDevice = MA_FALSE; /* It's a duplex device and the capture side is detached. We cannot be restarting the device just yet. */ + } + else { + restartDevice = MA_TRUE; /* It's not a duplex device, or the capture side is also attached so we can go ahead and restart the device. */ + } + } + } + else { + ma_device_reroute__wasapi(pThis->pDevice, (pThis->pDevice->type == ma_device_type_loopback) ? ma_device_type_loopback : ma_device_type_capture); + + if (pThis->pDevice->wasapi.isDetachedCapture) { + pThis->pDevice->wasapi.isDetachedCapture = MA_FALSE; + + if (pThis->pDevice->type == ma_device_type_duplex && pThis->pDevice->wasapi.isDetachedPlayback) { + restartDevice = MA_FALSE; /* It's a duplex device and the playback side is detached. We cannot be restarting the device just yet. */ + } + else { + restartDevice = MA_TRUE; /* It's not a duplex device, or the playback side is also attached so we can go ahead and restart the device. */ + } + } + } + } + ma_mutex_unlock(&pThis->pDevice->wasapi.rerouteLock); + + if (restartDevice) { + ma_device_start(pThis->pDevice); + } + } + } + + return S_OK; +} + +static HRESULT STDMETHODCALLTYPE ma_IMMNotificationClient_OnPropertyValueChanged(ma_IMMNotificationClient* pThis, const WCHAR* pDeviceID, const PROPERTYKEY key) +{ +#ifdef MA_DEBUG_OUTPUT + /*ma_log_postf(ma_device_get_log(pThis->pDevice), MA_LOG_LEVEL_DEBUG, "IMMNotificationClient_OnPropertyValueChanged(pDeviceID=%S)\n", (pDeviceID != NULL) ? pDeviceID : L"(NULL)");*/ +#endif + + (void)pThis; + (void)pDeviceID; + (void)key; + return S_OK; +} + +static ma_IMMNotificationClientVtbl g_maNotificationCientVtbl = { + ma_IMMNotificationClient_QueryInterface, + ma_IMMNotificationClient_AddRef, + ma_IMMNotificationClient_Release, + ma_IMMNotificationClient_OnDeviceStateChanged, + ma_IMMNotificationClient_OnDeviceAdded, + ma_IMMNotificationClient_OnDeviceRemoved, + ma_IMMNotificationClient_OnDefaultDeviceChanged, + ma_IMMNotificationClient_OnPropertyValueChanged +}; +#endif /* MA_WIN32_DESKTOP */ + +static const char* ma_to_usage_string__wasapi(ma_wasapi_usage usage) +{ + switch (usage) + { + case ma_wasapi_usage_default: return NULL; + case ma_wasapi_usage_games: return "Games"; + case ma_wasapi_usage_pro_audio: return "Pro Audio"; + default: break; + } + + return NULL; +} + +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) +typedef ma_IMMDevice ma_WASAPIDeviceInterface; +#else +typedef ma_IUnknown ma_WASAPIDeviceInterface; +#endif + + +#define MA_CONTEXT_COMMAND_QUIT__WASAPI 1 +#define MA_CONTEXT_COMMAND_CREATE_IAUDIOCLIENT__WASAPI 2 +#define MA_CONTEXT_COMMAND_RELEASE_IAUDIOCLIENT__WASAPI 3 + +static ma_context_command__wasapi ma_context_init_command__wasapi(int code) +{ + ma_context_command__wasapi cmd; + + MA_ZERO_OBJECT(&cmd); + cmd.code = code; + + return cmd; +} + +static ma_result ma_context_post_command__wasapi(ma_context* pContext, const ma_context_command__wasapi* pCmd) +{ + /* For now we are doing everything synchronously, but I might relax this later if the need arises. */ + ma_result result; + ma_bool32 isUsingLocalEvent = MA_FALSE; + ma_event localEvent; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pCmd != NULL); + + if (pCmd->pEvent == NULL) { + isUsingLocalEvent = MA_TRUE; + + result = ma_event_init(&localEvent); + if (result != MA_SUCCESS) { + return result; /* Failed to create the event for this command. */ + } + } + + /* Here is where we add the command to the list. If there's not enough room we'll spin until there is. */ + ma_mutex_lock(&pContext->wasapi.commandLock); + { + ma_uint32 index; + + /* Spin until we've got some space available. */ + while (pContext->wasapi.commandCount == ma_countof(pContext->wasapi.commands)) { + ma_yield(); + } + + /* Space is now available. Can safely add to the list. */ + index = (pContext->wasapi.commandIndex + pContext->wasapi.commandCount) % ma_countof(pContext->wasapi.commands); + pContext->wasapi.commands[index] = *pCmd; + pContext->wasapi.commands[index].pEvent = &localEvent; + pContext->wasapi.commandCount += 1; + + /* Now that the command has been added, release the semaphore so ma_context_next_command__wasapi() can return. */ + ma_semaphore_release(&pContext->wasapi.commandSem); + } + ma_mutex_unlock(&pContext->wasapi.commandLock); + + if (isUsingLocalEvent) { + ma_event_wait(&localEvent); + ma_event_uninit(&localEvent); + } + + return MA_SUCCESS; +} + +static ma_result ma_context_next_command__wasapi(ma_context* pContext, ma_context_command__wasapi* pCmd) +{ + ma_result result = MA_SUCCESS; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pCmd != NULL); + + result = ma_semaphore_wait(&pContext->wasapi.commandSem); + if (result == MA_SUCCESS) { + ma_mutex_lock(&pContext->wasapi.commandLock); + { + *pCmd = pContext->wasapi.commands[pContext->wasapi.commandIndex]; + pContext->wasapi.commandIndex = (pContext->wasapi.commandIndex + 1) % ma_countof(pContext->wasapi.commands); + pContext->wasapi.commandCount -= 1; + } + ma_mutex_unlock(&pContext->wasapi.commandLock); + } + + return result; +} + +static ma_thread_result MA_THREADCALL ma_context_command_thread__wasapi(void* pUserData) +{ + ma_result result; + ma_context* pContext = (ma_context*)pUserData; + MA_ASSERT(pContext != NULL); + + for (;;) { + ma_context_command__wasapi cmd; + result = ma_context_next_command__wasapi(pContext, &cmd); + if (result != MA_SUCCESS) { + break; + } + + switch (cmd.code) + { + case MA_CONTEXT_COMMAND_QUIT__WASAPI: + { + /* Do nothing. Handled after the switch. */ + } break; + + case MA_CONTEXT_COMMAND_CREATE_IAUDIOCLIENT__WASAPI: + { + if (cmd.data.createAudioClient.deviceType == ma_device_type_playback) { + *cmd.data.createAudioClient.pResult = ma_result_from_HRESULT(ma_IAudioClient_GetService((ma_IAudioClient*)cmd.data.createAudioClient.pAudioClient, &MA_IID_IAudioRenderClient, cmd.data.createAudioClient.ppAudioClientService)); + } else { + *cmd.data.createAudioClient.pResult = ma_result_from_HRESULT(ma_IAudioClient_GetService((ma_IAudioClient*)cmd.data.createAudioClient.pAudioClient, &MA_IID_IAudioCaptureClient, cmd.data.createAudioClient.ppAudioClientService)); + } + } break; + + case MA_CONTEXT_COMMAND_RELEASE_IAUDIOCLIENT__WASAPI: + { + if (cmd.data.releaseAudioClient.deviceType == ma_device_type_playback) { + if (cmd.data.releaseAudioClient.pDevice->wasapi.pAudioClientPlayback != NULL) { + ma_IAudioClient_Release((ma_IAudioClient*)cmd.data.releaseAudioClient.pDevice->wasapi.pAudioClientPlayback); + cmd.data.releaseAudioClient.pDevice->wasapi.pAudioClientPlayback = NULL; + } + } + + if (cmd.data.releaseAudioClient.deviceType == ma_device_type_capture) { + if (cmd.data.releaseAudioClient.pDevice->wasapi.pAudioClientCapture != NULL) { + ma_IAudioClient_Release((ma_IAudioClient*)cmd.data.releaseAudioClient.pDevice->wasapi.pAudioClientCapture); + cmd.data.releaseAudioClient.pDevice->wasapi.pAudioClientCapture = NULL; + } + } + } break; + + default: + { + /* Unknown command. Ignore it, but trigger an assert in debug mode so we're aware of it. */ + MA_ASSERT(MA_FALSE); + } break; + } + + if (cmd.pEvent != NULL) { + ma_event_signal(cmd.pEvent); + } + + if (cmd.code == MA_CONTEXT_COMMAND_QUIT__WASAPI) { + break; /* Received a quit message. Get out of here. */ + } + } + + return (ma_thread_result)0; +} + +static ma_result ma_device_create_IAudioClient_service__wasapi(ma_context* pContext, ma_device_type deviceType, ma_IAudioClient* pAudioClient, void** ppAudioClientService) +{ + ma_result result; + ma_result cmdResult; + ma_context_command__wasapi cmd = ma_context_init_command__wasapi(MA_CONTEXT_COMMAND_CREATE_IAUDIOCLIENT__WASAPI); + cmd.data.createAudioClient.deviceType = deviceType; + cmd.data.createAudioClient.pAudioClient = (void*)pAudioClient; + cmd.data.createAudioClient.ppAudioClientService = ppAudioClientService; + cmd.data.createAudioClient.pResult = &cmdResult; /* Declared locally, but won't be dereferenced after this function returns since execution of the command will wait here. */ + + result = ma_context_post_command__wasapi(pContext, &cmd); /* This will not return until the command has actually been run. */ + if (result != MA_SUCCESS) { + return result; + } + + return *cmd.data.createAudioClient.pResult; +} + +#if 0 /* Not used at the moment, but leaving here for future use. */ +static ma_result ma_device_release_IAudioClient_service__wasapi(ma_device* pDevice, ma_device_type deviceType) +{ + ma_result result; + ma_context_command__wasapi cmd = ma_context_init_command__wasapi(MA_CONTEXT_COMMAND_RELEASE_IAUDIOCLIENT__WASAPI); + cmd.data.releaseAudioClient.pDevice = pDevice; + cmd.data.releaseAudioClient.deviceType = deviceType; + + result = ma_context_post_command__wasapi(pDevice->pContext, &cmd); /* This will not return until the command has actually been run. */ + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} +#endif + + +static void ma_add_native_data_format_to_device_info_from_WAVEFORMATEX(const MA_WAVEFORMATEX* pWF, ma_share_mode shareMode, ma_device_info* pInfo) +{ + MA_ASSERT(pWF != NULL); + MA_ASSERT(pInfo != NULL); + + if (pInfo->nativeDataFormatCount >= ma_countof(pInfo->nativeDataFormats)) { + return; /* Too many data formats. Need to ignore this one. Don't think this should ever happen with WASAPI. */ + } + + pInfo->nativeDataFormats[pInfo->nativeDataFormatCount].format = ma_format_from_WAVEFORMATEX(pWF); + pInfo->nativeDataFormats[pInfo->nativeDataFormatCount].channels = pWF->nChannels; + pInfo->nativeDataFormats[pInfo->nativeDataFormatCount].sampleRate = pWF->nSamplesPerSec; + pInfo->nativeDataFormats[pInfo->nativeDataFormatCount].flags = (shareMode == ma_share_mode_exclusive) ? MA_DATA_FORMAT_FLAG_EXCLUSIVE_MODE : 0; + pInfo->nativeDataFormatCount += 1; +} + +static ma_result ma_context_get_device_info_from_IAudioClient__wasapi(ma_context* pContext, /*ma_IMMDevice**/void* pMMDevice, ma_IAudioClient* pAudioClient, ma_device_info* pInfo) +{ + HRESULT hr; + MA_WAVEFORMATEX* pWF = NULL; + + MA_ASSERT(pAudioClient != NULL); + MA_ASSERT(pInfo != NULL); + + /* Shared Mode. We use GetMixFormat() here. */ + hr = ma_IAudioClient_GetMixFormat((ma_IAudioClient*)pAudioClient, (MA_WAVEFORMATEX**)&pWF); + if (SUCCEEDED(hr)) { + ma_add_native_data_format_to_device_info_from_WAVEFORMATEX(pWF, ma_share_mode_shared, pInfo); + } else { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to retrieve mix format for device info retrieval."); + return ma_result_from_HRESULT(hr); + } + + /* + Exclusive Mode. We repeatedly call IsFormatSupported() here. This is not currently supported on + UWP. Failure to retrieve the exclusive mode format is not considered an error, so from here on + out, MA_SUCCESS is guaranteed to be returned. + */ + #if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + { + ma_IPropertyStore *pProperties; + + /* + The first thing to do is get the format from PKEY_AudioEngine_DeviceFormat. This should give us a channel count we assume is + correct which will simplify our searching. + */ + hr = ma_IMMDevice_OpenPropertyStore((ma_IMMDevice*)pMMDevice, STGM_READ, &pProperties); + if (SUCCEEDED(hr)) { + MA_PROPVARIANT var; + ma_PropVariantInit(&var); + + hr = ma_IPropertyStore_GetValue(pProperties, &MA_PKEY_AudioEngine_DeviceFormat, &var); + if (SUCCEEDED(hr)) { + pWF = (MA_WAVEFORMATEX*)var.blob.pBlobData; + + /* + In my testing, the format returned by PKEY_AudioEngine_DeviceFormat is suitable for exclusive mode so we check this format + first. If this fails, fall back to a search. + */ + hr = ma_IAudioClient_IsFormatSupported((ma_IAudioClient*)pAudioClient, MA_AUDCLNT_SHAREMODE_EXCLUSIVE, pWF, NULL); + if (SUCCEEDED(hr)) { + /* The format returned by PKEY_AudioEngine_DeviceFormat is supported. */ + ma_add_native_data_format_to_device_info_from_WAVEFORMATEX(pWF, ma_share_mode_exclusive, pInfo); + } else { + /* + The format returned by PKEY_AudioEngine_DeviceFormat is not supported, so fall back to a search. We assume the channel + count returned by MA_PKEY_AudioEngine_DeviceFormat is valid and correct. For simplicity we're only returning one format. + */ + ma_uint32 channels = pWF->nChannels; + ma_channel defaultChannelMap[MA_MAX_CHANNELS]; + MA_WAVEFORMATEXTENSIBLE wf; + ma_bool32 found; + ma_uint32 iFormat; + + /* Make sure we don't overflow the channel map. */ + if (channels > MA_MAX_CHANNELS) { + channels = MA_MAX_CHANNELS; + } + + ma_channel_map_init_standard(ma_standard_channel_map_microsoft, defaultChannelMap, ma_countof(defaultChannelMap), channels); + + MA_ZERO_OBJECT(&wf); + wf.cbSize = sizeof(wf); + wf.wFormatTag = WAVE_FORMAT_EXTENSIBLE; + wf.nChannels = (WORD)channels; + wf.dwChannelMask = ma_channel_map_to_channel_mask__win32(defaultChannelMap, channels); + + found = MA_FALSE; + for (iFormat = 0; iFormat < ma_countof(g_maFormatPriorities); ++iFormat) { + ma_format format = g_maFormatPriorities[iFormat]; + ma_uint32 iSampleRate; + + wf.wBitsPerSample = (WORD)(ma_get_bytes_per_sample(format)*8); + wf.nBlockAlign = (WORD)(wf.nChannels * wf.wBitsPerSample / 8); + wf.nAvgBytesPerSec = wf.nBlockAlign * wf.nSamplesPerSec; + wf.Samples.wValidBitsPerSample = /*(format == ma_format_s24_32) ? 24 :*/ wf.wBitsPerSample; + if (format == ma_format_f32) { + wf.SubFormat = MA_GUID_KSDATAFORMAT_SUBTYPE_IEEE_FLOAT; + } else { + wf.SubFormat = MA_GUID_KSDATAFORMAT_SUBTYPE_PCM; + } + + for (iSampleRate = 0; iSampleRate < ma_countof(g_maStandardSampleRatePriorities); ++iSampleRate) { + wf.nSamplesPerSec = g_maStandardSampleRatePriorities[iSampleRate]; + + hr = ma_IAudioClient_IsFormatSupported((ma_IAudioClient*)pAudioClient, MA_AUDCLNT_SHAREMODE_EXCLUSIVE, (MA_WAVEFORMATEX*)&wf, NULL); + if (SUCCEEDED(hr)) { + ma_add_native_data_format_to_device_info_from_WAVEFORMATEX((MA_WAVEFORMATEX*)&wf, ma_share_mode_exclusive, pInfo); + found = MA_TRUE; + break; + } + } + + if (found) { + break; + } + } + + ma_PropVariantClear(pContext, &var); + + if (!found) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_WARNING, "[WASAPI] Failed to find suitable device format for device info retrieval."); + } + } + } else { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_WARNING, "[WASAPI] Failed to retrieve device format for device info retrieval."); + } + + ma_IPropertyStore_Release(pProperties); + } else { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_WARNING, "[WASAPI] Failed to open property store for device info retrieval."); + } + } + #else + { + (void)pMMDevice; /* Unused. */ + } + #endif + + return MA_SUCCESS; +} + +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) +static ma_EDataFlow ma_device_type_to_EDataFlow(ma_device_type deviceType) +{ + if (deviceType == ma_device_type_playback) { + return ma_eRender; + } else if (deviceType == ma_device_type_capture) { + return ma_eCapture; + } else { + MA_ASSERT(MA_FALSE); + return ma_eRender; /* Should never hit this. */ + } +} + +static ma_result ma_context_create_IMMDeviceEnumerator__wasapi(ma_context* pContext, ma_IMMDeviceEnumerator** ppDeviceEnumerator) +{ + HRESULT hr; + ma_IMMDeviceEnumerator* pDeviceEnumerator; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppDeviceEnumerator != NULL); + + *ppDeviceEnumerator = NULL; /* Safety. */ + + hr = ma_CoCreateInstance(pContext, &MA_CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL, &MA_IID_IMMDeviceEnumerator, (void**)&pDeviceEnumerator); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to create device enumerator."); + return ma_result_from_HRESULT(hr); + } + + *ppDeviceEnumerator = pDeviceEnumerator; + + return MA_SUCCESS; +} + +static WCHAR* ma_context_get_default_device_id_from_IMMDeviceEnumerator__wasapi(ma_context* pContext, ma_IMMDeviceEnumerator* pDeviceEnumerator, ma_device_type deviceType) +{ + HRESULT hr; + ma_IMMDevice* pMMDefaultDevice = NULL; + WCHAR* pDefaultDeviceID = NULL; + ma_EDataFlow dataFlow; + ma_ERole role; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pDeviceEnumerator != NULL); + + (void)pContext; + + /* Grab the EDataFlow type from the device type. */ + dataFlow = ma_device_type_to_EDataFlow(deviceType); + + /* The role is always eConsole, but we may make this configurable later. */ + role = ma_eConsole; + + hr = ma_IMMDeviceEnumerator_GetDefaultAudioEndpoint(pDeviceEnumerator, dataFlow, role, &pMMDefaultDevice); + if (FAILED(hr)) { + return NULL; + } + + hr = ma_IMMDevice_GetId(pMMDefaultDevice, &pDefaultDeviceID); + + ma_IMMDevice_Release(pMMDefaultDevice); + pMMDefaultDevice = NULL; + + if (FAILED(hr)) { + return NULL; + } + + return pDefaultDeviceID; +} + +static WCHAR* ma_context_get_default_device_id__wasapi(ma_context* pContext, ma_device_type deviceType) /* Free the returned pointer with ma_CoTaskMemFree() */ +{ + ma_result result; + ma_IMMDeviceEnumerator* pDeviceEnumerator; + WCHAR* pDefaultDeviceID = NULL; + + MA_ASSERT(pContext != NULL); + + result = ma_context_create_IMMDeviceEnumerator__wasapi(pContext, &pDeviceEnumerator); + if (result != MA_SUCCESS) { + return NULL; + } + + pDefaultDeviceID = ma_context_get_default_device_id_from_IMMDeviceEnumerator__wasapi(pContext, pDeviceEnumerator, deviceType); + + ma_IMMDeviceEnumerator_Release(pDeviceEnumerator); + return pDefaultDeviceID; +} + +static ma_result ma_context_get_MMDevice__wasapi(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_IMMDevice** ppMMDevice) +{ + ma_IMMDeviceEnumerator* pDeviceEnumerator; + HRESULT hr; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppMMDevice != NULL); + + /* + This weird COM init/uninit here is a hack to work around a crash when changing devices. What is happening is + WASAPI fires a callback from another thread when the device is changed. It's from that thread where this + function is getting called. What I'm suspecting is that the other thread is not initializing COM which in turn + results in CoCreateInstance() failing. + + The community has reported that this seems to fix the crash. There are future plans to move all WASAPI operation + over to a single thread to make everything safer, but in the meantime while we wait for that to come online I'm + happy enough to use this hack instead. + */ + ma_CoInitializeEx(pContext, NULL, MA_COINIT_VALUE); + { + hr = ma_CoCreateInstance(pContext, &MA_CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL, &MA_IID_IMMDeviceEnumerator, (void**)&pDeviceEnumerator); + } + ma_CoUninitialize(pContext); + + if (FAILED(hr)) { /* <-- This is checking the call above to ma_CoCreateInstance(). */ + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to create IMMDeviceEnumerator.\n"); + return ma_result_from_HRESULT(hr); + } + + if (pDeviceID == NULL) { + hr = ma_IMMDeviceEnumerator_GetDefaultAudioEndpoint(pDeviceEnumerator, (deviceType == ma_device_type_capture) ? ma_eCapture : ma_eRender, ma_eConsole, ppMMDevice); + } else { + hr = ma_IMMDeviceEnumerator_GetDevice(pDeviceEnumerator, pDeviceID->wasapi, ppMMDevice); + } + + ma_IMMDeviceEnumerator_Release(pDeviceEnumerator); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to retrieve IMMDevice.\n"); + return ma_result_from_HRESULT(hr); + } + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_id_from_MMDevice__wasapi(ma_context* pContext, ma_IMMDevice* pMMDevice, ma_device_id* pDeviceID) +{ + WCHAR* pDeviceIDString; + HRESULT hr; + + MA_ASSERT(pDeviceID != NULL); + + hr = ma_IMMDevice_GetId(pMMDevice, &pDeviceIDString); + if (SUCCEEDED(hr)) { + size_t idlen = ma_strlen_WCHAR(pDeviceIDString); + if (idlen+1 > ma_countof(pDeviceID->wasapi)) { + ma_CoTaskMemFree(pContext, pDeviceIDString); + MA_ASSERT(MA_FALSE); /* NOTE: If this is triggered, please report it. It means the format of the ID must have changed and is too long to fit in our fixed sized buffer. */ + return MA_ERROR; + } + + MA_COPY_MEMORY(pDeviceID->wasapi, pDeviceIDString, idlen * sizeof(wchar_t)); + pDeviceID->wasapi[idlen] = '\0'; + + ma_CoTaskMemFree(pContext, pDeviceIDString); + + return MA_SUCCESS; + } + + return MA_ERROR; +} + +static ma_result ma_context_get_device_info_from_MMDevice__wasapi(ma_context* pContext, ma_IMMDevice* pMMDevice, WCHAR* pDefaultDeviceID, ma_bool32 onlySimpleInfo, ma_device_info* pInfo) +{ + ma_result result; + HRESULT hr; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pMMDevice != NULL); + MA_ASSERT(pInfo != NULL); + + /* ID. */ + result = ma_context_get_device_id_from_MMDevice__wasapi(pContext, pMMDevice, &pInfo->id); + if (result == MA_SUCCESS) { + if (pDefaultDeviceID != NULL) { + if (ma_strcmp_WCHAR(pInfo->id.wasapi, pDefaultDeviceID) == 0) { + pInfo->isDefault = MA_TRUE; + } + } + } + + /* Description / Friendly Name */ + { + ma_IPropertyStore *pProperties; + hr = ma_IMMDevice_OpenPropertyStore(pMMDevice, STGM_READ, &pProperties); + if (SUCCEEDED(hr)) { + MA_PROPVARIANT var; + + ma_PropVariantInit(&var); + hr = ma_IPropertyStore_GetValue(pProperties, &MA_PKEY_Device_FriendlyName, &var); + if (SUCCEEDED(hr)) { + WideCharToMultiByte(CP_UTF8, 0, var.pwszVal, -1, pInfo->name, sizeof(pInfo->name), 0, FALSE); + ma_PropVariantClear(pContext, &var); + } + + ma_IPropertyStore_Release(pProperties); + } + } + + /* Format */ + if (!onlySimpleInfo) { + ma_IAudioClient* pAudioClient; + hr = ma_IMMDevice_Activate(pMMDevice, &MA_IID_IAudioClient, CLSCTX_ALL, NULL, (void**)&pAudioClient); + if (SUCCEEDED(hr)) { + result = ma_context_get_device_info_from_IAudioClient__wasapi(pContext, pMMDevice, pAudioClient, pInfo); + + ma_IAudioClient_Release(pAudioClient); + return result; + } else { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to activate audio client for device info retrieval."); + return ma_result_from_HRESULT(hr); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_context_enumerate_devices_by_type__wasapi(ma_context* pContext, ma_IMMDeviceEnumerator* pDeviceEnumerator, ma_device_type deviceType, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_result result = MA_SUCCESS; + UINT deviceCount; + HRESULT hr; + ma_uint32 iDevice; + WCHAR* pDefaultDeviceID = NULL; + ma_IMMDeviceCollection* pDeviceCollection = NULL; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + /* Grab the default device. We use this to know whether or not flag the returned device info as being the default. */ + pDefaultDeviceID = ma_context_get_default_device_id_from_IMMDeviceEnumerator__wasapi(pContext, pDeviceEnumerator, deviceType); + + /* We need to enumerate the devices which returns a device collection. */ + hr = ma_IMMDeviceEnumerator_EnumAudioEndpoints(pDeviceEnumerator, ma_device_type_to_EDataFlow(deviceType), MA_MM_DEVICE_STATE_ACTIVE, &pDeviceCollection); + if (SUCCEEDED(hr)) { + hr = ma_IMMDeviceCollection_GetCount(pDeviceCollection, &deviceCount); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to get device count.\n"); + result = ma_result_from_HRESULT(hr); + goto done; + } + + for (iDevice = 0; iDevice < deviceCount; ++iDevice) { + ma_device_info deviceInfo; + ma_IMMDevice* pMMDevice; + + MA_ZERO_OBJECT(&deviceInfo); + + hr = ma_IMMDeviceCollection_Item(pDeviceCollection, iDevice, &pMMDevice); + if (SUCCEEDED(hr)) { + result = ma_context_get_device_info_from_MMDevice__wasapi(pContext, pMMDevice, pDefaultDeviceID, MA_TRUE, &deviceInfo); /* MA_TRUE = onlySimpleInfo. */ + + ma_IMMDevice_Release(pMMDevice); + if (result == MA_SUCCESS) { + ma_bool32 cbResult = callback(pContext, deviceType, &deviceInfo, pUserData); + if (cbResult == MA_FALSE) { + break; + } + } + } + } + } + +done: + if (pDefaultDeviceID != NULL) { + ma_CoTaskMemFree(pContext, pDefaultDeviceID); + pDefaultDeviceID = NULL; + } + + if (pDeviceCollection != NULL) { + ma_IMMDeviceCollection_Release(pDeviceCollection); + pDeviceCollection = NULL; + } + + return result; +} + +static ma_result ma_context_get_IAudioClient_Desktop__wasapi(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, MA_PROPVARIANT* pActivationParams, ma_IAudioClient** ppAudioClient, ma_IMMDevice** ppMMDevice) +{ + ma_result result; + HRESULT hr; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppAudioClient != NULL); + MA_ASSERT(ppMMDevice != NULL); + + result = ma_context_get_MMDevice__wasapi(pContext, deviceType, pDeviceID, ppMMDevice); + if (result != MA_SUCCESS) { + return result; + } + + hr = ma_IMMDevice_Activate(*ppMMDevice, &MA_IID_IAudioClient, CLSCTX_ALL, pActivationParams, (void**)ppAudioClient); + if (FAILED(hr)) { + return ma_result_from_HRESULT(hr); + } + + return MA_SUCCESS; +} +#else +static ma_result ma_context_get_IAudioClient_UWP__wasapi(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, MA_PROPVARIANT* pActivationParams, ma_IAudioClient** ppAudioClient, ma_IUnknown** ppActivatedInterface) +{ + ma_IActivateAudioInterfaceAsyncOperation *pAsyncOp = NULL; + ma_completion_handler_uwp completionHandler; + IID iid; + WCHAR* iidStr; + HRESULT hr; + ma_result result; + HRESULT activateResult; + ma_IUnknown* pActivatedInterface; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppAudioClient != NULL); + + if (pDeviceID != NULL) { + iidStr = (WCHAR*)pDeviceID->wasapi; + } else { + if (deviceType == ma_device_type_capture) { + iid = MA_IID_DEVINTERFACE_AUDIO_CAPTURE; + } else { + iid = MA_IID_DEVINTERFACE_AUDIO_RENDER; + } + + #if defined(__cplusplus) + hr = StringFromIID(iid, &iidStr); + #else + hr = StringFromIID(&iid, &iidStr); + #endif + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to convert device IID to string for ActivateAudioInterfaceAsync(). Out of memory.\n"); + return ma_result_from_HRESULT(hr); + } + } + + result = ma_completion_handler_uwp_init(&completionHandler); + if (result != MA_SUCCESS) { + ma_CoTaskMemFree(pContext, iidStr); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to create event for waiting for ActivateAudioInterfaceAsync().\n"); + return result; + } + + hr = ((MA_PFN_ActivateAudioInterfaceAsync)pContext->wasapi.ActivateAudioInterfaceAsync)(iidStr, &MA_IID_IAudioClient, pActivationParams, (ma_IActivateAudioInterfaceCompletionHandler*)&completionHandler, (ma_IActivateAudioInterfaceAsyncOperation**)&pAsyncOp); + if (FAILED(hr)) { + ma_completion_handler_uwp_uninit(&completionHandler); + ma_CoTaskMemFree(pContext, iidStr); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] ActivateAudioInterfaceAsync() failed.\n"); + return ma_result_from_HRESULT(hr); + } + + if (pDeviceID == NULL) { + ma_CoTaskMemFree(pContext, iidStr); + } + + /* Wait for the async operation for finish. */ + ma_completion_handler_uwp_wait(&completionHandler); + ma_completion_handler_uwp_uninit(&completionHandler); + + hr = ma_IActivateAudioInterfaceAsyncOperation_GetActivateResult(pAsyncOp, &activateResult, &pActivatedInterface); + ma_IActivateAudioInterfaceAsyncOperation_Release(pAsyncOp); + + if (FAILED(hr) || FAILED(activateResult)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to activate device.\n"); + return FAILED(hr) ? ma_result_from_HRESULT(hr) : ma_result_from_HRESULT(activateResult); + } + + /* Here is where we grab the IAudioClient interface. */ + hr = ma_IUnknown_QueryInterface(pActivatedInterface, &MA_IID_IAudioClient, (void**)ppAudioClient); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to query IAudioClient interface.\n"); + return ma_result_from_HRESULT(hr); + } + + if (ppActivatedInterface) { + *ppActivatedInterface = pActivatedInterface; + } else { + ma_IUnknown_Release(pActivatedInterface); + } + + return MA_SUCCESS; +} +#endif + + +/* https://docs.microsoft.com/en-us/windows/win32/api/audioclientactivationparams/ne-audioclientactivationparams-audioclient_activation_type */ +typedef enum +{ + MA_AUDIOCLIENT_ACTIVATION_TYPE_DEFAULT, + MA_AUDIOCLIENT_ACTIVATION_TYPE_PROCESS_LOOPBACK +} MA_AUDIOCLIENT_ACTIVATION_TYPE; + +/* https://docs.microsoft.com/en-us/windows/win32/api/audioclientactivationparams/ne-audioclientactivationparams-process_loopback_mode */ +typedef enum +{ + MA_PROCESS_LOOPBACK_MODE_INCLUDE_TARGET_PROCESS_TREE, + MA_PROCESS_LOOPBACK_MODE_EXCLUDE_TARGET_PROCESS_TREE +} MA_PROCESS_LOOPBACK_MODE; + +/* https://docs.microsoft.com/en-us/windows/win32/api/audioclientactivationparams/ns-audioclientactivationparams-audioclient_process_loopback_params */ +typedef struct +{ + DWORD TargetProcessId; + MA_PROCESS_LOOPBACK_MODE ProcessLoopbackMode; +} MA_AUDIOCLIENT_PROCESS_LOOPBACK_PARAMS; + +#if defined(_MSC_VER) && !defined(__clang__) + #pragma warning(push) + #pragma warning(disable:4201) /* nonstandard extension used: nameless struct/union */ +#elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) + #pragma GCC diagnostic push + #pragma GCC diagnostic ignored "-Wpedantic" /* For ISO C99 doesn't support unnamed structs/unions [-Wpedantic] */ + #if defined(__clang__) + #pragma GCC diagnostic ignored "-Wc11-extensions" /* anonymous unions are a C11 extension */ + #endif +#endif +/* https://docs.microsoft.com/en-us/windows/win32/api/audioclientactivationparams/ns-audioclientactivationparams-audioclient_activation_params */ +typedef struct +{ + MA_AUDIOCLIENT_ACTIVATION_TYPE ActivationType; + union + { + MA_AUDIOCLIENT_PROCESS_LOOPBACK_PARAMS ProcessLoopbackParams; + }; +} MA_AUDIOCLIENT_ACTIVATION_PARAMS; +#if defined(_MSC_VER) && !defined(__clang__) + #pragma warning(pop) +#elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) + #pragma GCC diagnostic pop +#endif + +#define MA_VIRTUAL_AUDIO_DEVICE_PROCESS_LOOPBACK L"VAD\\Process_Loopback" + +static ma_result ma_context_get_IAudioClient__wasapi(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_uint32 loopbackProcessID, ma_bool32 loopbackProcessExclude, ma_IAudioClient** ppAudioClient, ma_WASAPIDeviceInterface** ppDeviceInterface) +{ + ma_result result; + ma_bool32 usingProcessLoopback = MA_FALSE; + MA_AUDIOCLIENT_ACTIVATION_PARAMS audioclientActivationParams; + MA_PROPVARIANT activationParams; + MA_PROPVARIANT* pActivationParams = NULL; + ma_device_id virtualDeviceID; + + /* Activation parameters specific to loopback mode. Note that process-specific loopback will only work when a default device ID is specified. */ + if (deviceType == ma_device_type_loopback && loopbackProcessID != 0 && pDeviceID == NULL) { + usingProcessLoopback = MA_TRUE; + } + + if (usingProcessLoopback) { + MA_ZERO_OBJECT(&audioclientActivationParams); + audioclientActivationParams.ActivationType = MA_AUDIOCLIENT_ACTIVATION_TYPE_PROCESS_LOOPBACK; + audioclientActivationParams.ProcessLoopbackParams.ProcessLoopbackMode = (loopbackProcessExclude) ? MA_PROCESS_LOOPBACK_MODE_EXCLUDE_TARGET_PROCESS_TREE : MA_PROCESS_LOOPBACK_MODE_INCLUDE_TARGET_PROCESS_TREE; + audioclientActivationParams.ProcessLoopbackParams.TargetProcessId = (DWORD)loopbackProcessID; + + ma_PropVariantInit(&activationParams); + activationParams.vt = MA_VT_BLOB; + activationParams.blob.cbSize = sizeof(audioclientActivationParams); + activationParams.blob.pBlobData = (BYTE*)&audioclientActivationParams; + pActivationParams = &activationParams; + + /* When requesting a specific device ID we need to use a special device ID. */ + MA_COPY_MEMORY(virtualDeviceID.wasapi, MA_VIRTUAL_AUDIO_DEVICE_PROCESS_LOOPBACK, (wcslen(MA_VIRTUAL_AUDIO_DEVICE_PROCESS_LOOPBACK) + 1) * sizeof(wchar_t)); /* +1 for the null terminator. */ + pDeviceID = &virtualDeviceID; + } else { + pActivationParams = NULL; /* No activation parameters required. */ + } + +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + result = ma_context_get_IAudioClient_Desktop__wasapi(pContext, deviceType, pDeviceID, pActivationParams, ppAudioClient, ppDeviceInterface); +#else + result = ma_context_get_IAudioClient_UWP__wasapi(pContext, deviceType, pDeviceID, pActivationParams, ppAudioClient, ppDeviceInterface); +#endif + + /* + If loopback mode was requested with a process ID and initialization failed, it could be because it's + trying to run on an older version of Windows where it's not supported. We need to let the caller + know about this with a log message. + */ + if (result != MA_SUCCESS) { + if (usingProcessLoopback) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Loopback mode requested to %s process ID %u, but initialization failed. Support for this feature begins with Windows 10 Build 20348. Confirm your version of Windows or consider not using process-specific loopback.\n", (loopbackProcessExclude) ? "exclude" : "include", loopbackProcessID); + } + } + + return result; +} + + +static ma_result ma_context_enumerate_devices__wasapi(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + /* Different enumeration for desktop and UWP. */ +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + /* Desktop */ + HRESULT hr; + ma_IMMDeviceEnumerator* pDeviceEnumerator; + + hr = ma_CoCreateInstance(pContext, &MA_CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL, &MA_IID_IMMDeviceEnumerator, (void**)&pDeviceEnumerator); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to create device enumerator."); + return ma_result_from_HRESULT(hr); + } + + ma_context_enumerate_devices_by_type__wasapi(pContext, pDeviceEnumerator, ma_device_type_playback, callback, pUserData); + ma_context_enumerate_devices_by_type__wasapi(pContext, pDeviceEnumerator, ma_device_type_capture, callback, pUserData); + + ma_IMMDeviceEnumerator_Release(pDeviceEnumerator); +#else + /* + UWP + + The MMDevice API is only supported on desktop applications. For now, while I'm still figuring out how to properly enumerate + over devices without using MMDevice, I'm restricting devices to defaults. + + Hint: DeviceInformation::FindAllAsync() with DeviceClass.AudioCapture/AudioRender. https://blogs.windows.com/buildingapps/2014/05/15/real-time-audio-in-windows-store-and-windows-phone-apps/ + */ + if (callback) { + ma_bool32 cbResult = MA_TRUE; + + /* Playback. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + deviceInfo.isDefault = MA_TRUE; + cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + + /* Capture. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + deviceInfo.isDefault = MA_TRUE; + cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + } +#endif + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info__wasapi(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + ma_result result; + ma_IMMDevice* pMMDevice = NULL; + WCHAR* pDefaultDeviceID = NULL; + + result = ma_context_get_MMDevice__wasapi(pContext, deviceType, pDeviceID, &pMMDevice); + if (result != MA_SUCCESS) { + return result; + } + + /* We need the default device ID so we can set the isDefault flag in the device info. */ + pDefaultDeviceID = ma_context_get_default_device_id__wasapi(pContext, deviceType); + + result = ma_context_get_device_info_from_MMDevice__wasapi(pContext, pMMDevice, pDefaultDeviceID, MA_FALSE, pDeviceInfo); /* MA_FALSE = !onlySimpleInfo. */ + + if (pDefaultDeviceID != NULL) { + ma_CoTaskMemFree(pContext, pDefaultDeviceID); + pDefaultDeviceID = NULL; + } + + ma_IMMDevice_Release(pMMDevice); + + return result; +#else + ma_IAudioClient* pAudioClient; + ma_result result; + + /* UWP currently only uses default devices. */ + if (deviceType == ma_device_type_playback) { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } + + result = ma_context_get_IAudioClient_UWP__wasapi(pContext, deviceType, pDeviceID, NULL, &pAudioClient, NULL); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_context_get_device_info_from_IAudioClient__wasapi(pContext, NULL, pAudioClient, pDeviceInfo); + + pDeviceInfo->isDefault = MA_TRUE; /* UWP only supports default devices. */ + + ma_IAudioClient_Release(pAudioClient); + return result; +#endif +} + +static ma_result ma_device_uninit__wasapi(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + #if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + { + if (pDevice->wasapi.pDeviceEnumerator) { + ((ma_IMMDeviceEnumerator*)pDevice->wasapi.pDeviceEnumerator)->lpVtbl->UnregisterEndpointNotificationCallback((ma_IMMDeviceEnumerator*)pDevice->wasapi.pDeviceEnumerator, &pDevice->wasapi.notificationClient); + ma_IMMDeviceEnumerator_Release((ma_IMMDeviceEnumerator*)pDevice->wasapi.pDeviceEnumerator); + } + + ma_mutex_uninit(&pDevice->wasapi.rerouteLock); + } + #endif + + if (pDevice->wasapi.pRenderClient) { + if (pDevice->wasapi.pMappedBufferPlayback != NULL) { + ma_IAudioRenderClient_ReleaseBuffer((ma_IAudioRenderClient*)pDevice->wasapi.pRenderClient, pDevice->wasapi.mappedBufferPlaybackCap, 0); + pDevice->wasapi.pMappedBufferPlayback = NULL; + pDevice->wasapi.mappedBufferPlaybackCap = 0; + pDevice->wasapi.mappedBufferPlaybackLen = 0; + } + + ma_IAudioRenderClient_Release((ma_IAudioRenderClient*)pDevice->wasapi.pRenderClient); + } + if (pDevice->wasapi.pCaptureClient) { + if (pDevice->wasapi.pMappedBufferCapture != NULL) { + ma_IAudioCaptureClient_ReleaseBuffer((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient, pDevice->wasapi.mappedBufferCaptureCap); + pDevice->wasapi.pMappedBufferCapture = NULL; + pDevice->wasapi.mappedBufferCaptureCap = 0; + pDevice->wasapi.mappedBufferCaptureLen = 0; + } + + ma_IAudioCaptureClient_Release((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient); + } + + if (pDevice->wasapi.pAudioClientPlayback) { + ma_IAudioClient_Release((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback); + } + if (pDevice->wasapi.pAudioClientCapture) { + ma_IAudioClient_Release((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture); + } + + if (pDevice->wasapi.hEventPlayback) { + CloseHandle((HANDLE)pDevice->wasapi.hEventPlayback); + } + if (pDevice->wasapi.hEventCapture) { + CloseHandle((HANDLE)pDevice->wasapi.hEventCapture); + } + + return MA_SUCCESS; +} + + +typedef struct +{ + /* Input. */ + ma_format formatIn; + ma_uint32 channelsIn; + ma_uint32 sampleRateIn; + ma_channel channelMapIn[MA_MAX_CHANNELS]; + ma_uint32 periodSizeInFramesIn; + ma_uint32 periodSizeInMillisecondsIn; + ma_uint32 periodsIn; + ma_share_mode shareMode; + ma_performance_profile performanceProfile; + ma_bool32 noAutoConvertSRC; + ma_bool32 noDefaultQualitySRC; + ma_bool32 noHardwareOffloading; + ma_uint32 loopbackProcessID; + ma_bool32 loopbackProcessExclude; + + /* Output. */ + ma_IAudioClient* pAudioClient; + ma_IAudioRenderClient* pRenderClient; + ma_IAudioCaptureClient* pCaptureClient; + ma_format formatOut; + ma_uint32 channelsOut; + ma_uint32 sampleRateOut; + ma_channel channelMapOut[MA_MAX_CHANNELS]; + ma_uint32 periodSizeInFramesOut; + ma_uint32 periodsOut; + ma_bool32 usingAudioClient3; + char deviceName[256]; + ma_device_id id; +} ma_device_init_internal_data__wasapi; + +static ma_result ma_device_init_internal__wasapi(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_init_internal_data__wasapi* pData) +{ + HRESULT hr; + ma_result result = MA_SUCCESS; + const char* errorMsg = ""; + MA_AUDCLNT_SHAREMODE shareMode = MA_AUDCLNT_SHAREMODE_SHARED; + DWORD streamFlags = 0; + MA_REFERENCE_TIME periodDurationInMicroseconds; + ma_bool32 wasInitializedUsingIAudioClient3 = MA_FALSE; + MA_WAVEFORMATEXTENSIBLE wf; + ma_WASAPIDeviceInterface* pDeviceInterface = NULL; + ma_IAudioClient2* pAudioClient2; + ma_uint32 nativeSampleRate; + ma_bool32 usingProcessLoopback = MA_FALSE; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pData != NULL); + + /* This function is only used to initialize one device type: either playback, capture or loopback. Never full-duplex. */ + if (deviceType == ma_device_type_duplex) { + return MA_INVALID_ARGS; + } + + usingProcessLoopback = deviceType == ma_device_type_loopback && pData->loopbackProcessID != 0 && pDeviceID == NULL; + + pData->pAudioClient = NULL; + pData->pRenderClient = NULL; + pData->pCaptureClient = NULL; + + streamFlags = MA_AUDCLNT_STREAMFLAGS_EVENTCALLBACK; + if (!pData->noAutoConvertSRC && pData->sampleRateIn != 0 && pData->shareMode != ma_share_mode_exclusive) { /* <-- Exclusive streams must use the native sample rate. */ + streamFlags |= MA_AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM; + } + if (!pData->noDefaultQualitySRC && pData->sampleRateIn != 0 && (streamFlags & MA_AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM) != 0) { + streamFlags |= MA_AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY; + } + if (deviceType == ma_device_type_loopback) { + streamFlags |= MA_AUDCLNT_STREAMFLAGS_LOOPBACK; + } + + result = ma_context_get_IAudioClient__wasapi(pContext, deviceType, pDeviceID, pData->loopbackProcessID, pData->loopbackProcessExclude, &pData->pAudioClient, &pDeviceInterface); + if (result != MA_SUCCESS) { + goto done; + } + + MA_ZERO_OBJECT(&wf); + + /* Try enabling hardware offloading. */ + if (!pData->noHardwareOffloading) { + hr = ma_IAudioClient_QueryInterface(pData->pAudioClient, &MA_IID_IAudioClient2, (void**)&pAudioClient2); + if (SUCCEEDED(hr)) { + BOOL isHardwareOffloadingSupported = 0; + hr = ma_IAudioClient2_IsOffloadCapable(pAudioClient2, MA_AudioCategory_Other, &isHardwareOffloadingSupported); + if (SUCCEEDED(hr) && isHardwareOffloadingSupported) { + ma_AudioClientProperties clientProperties; + MA_ZERO_OBJECT(&clientProperties); + clientProperties.cbSize = sizeof(clientProperties); + clientProperties.bIsOffload = 1; + clientProperties.eCategory = MA_AudioCategory_Other; + ma_IAudioClient2_SetClientProperties(pAudioClient2, &clientProperties); + } + + pAudioClient2->lpVtbl->Release(pAudioClient2); + } + } + + /* Here is where we try to determine the best format to use with the device. If the client if wanting exclusive mode, first try finding the best format for that. If this fails, fall back to shared mode. */ + result = MA_FORMAT_NOT_SUPPORTED; + if (pData->shareMode == ma_share_mode_exclusive) { + #if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + /* In exclusive mode on desktop we always use the backend's native format. */ + ma_IPropertyStore* pStore = NULL; + hr = ma_IMMDevice_OpenPropertyStore(pDeviceInterface, STGM_READ, &pStore); + if (SUCCEEDED(hr)) { + MA_PROPVARIANT prop; + ma_PropVariantInit(&prop); + hr = ma_IPropertyStore_GetValue(pStore, &MA_PKEY_AudioEngine_DeviceFormat, &prop); + if (SUCCEEDED(hr)) { + MA_WAVEFORMATEX* pActualFormat = (MA_WAVEFORMATEX*)prop.blob.pBlobData; + hr = ma_IAudioClient_IsFormatSupported((ma_IAudioClient*)pData->pAudioClient, MA_AUDCLNT_SHAREMODE_EXCLUSIVE, pActualFormat, NULL); + if (SUCCEEDED(hr)) { + MA_COPY_MEMORY(&wf, pActualFormat, sizeof(MA_WAVEFORMATEXTENSIBLE)); + } + + ma_PropVariantClear(pContext, &prop); + } + + ma_IPropertyStore_Release(pStore); + } + #else + /* + I do not know how to query the device's native format on UWP so for now I'm just disabling support for + exclusive mode. The alternative is to enumerate over different formats and check IsFormatSupported() + until you find one that works. + + TODO: Add support for exclusive mode to UWP. + */ + hr = S_FALSE; + #endif + + if (hr == S_OK) { + shareMode = MA_AUDCLNT_SHAREMODE_EXCLUSIVE; + result = MA_SUCCESS; + } else { + result = MA_SHARE_MODE_NOT_SUPPORTED; + } + } else { + /* In shared mode we are always using the format reported by the operating system. */ + MA_WAVEFORMATEXTENSIBLE* pNativeFormat = NULL; + hr = ma_IAudioClient_GetMixFormat((ma_IAudioClient*)pData->pAudioClient, (MA_WAVEFORMATEX**)&pNativeFormat); + if (hr != S_OK) { + /* When using process-specific loopback, GetMixFormat() seems to always fail. */ + if (usingProcessLoopback) { + wf.wFormatTag = WAVE_FORMAT_IEEE_FLOAT; + wf.nChannels = 2; + wf.nSamplesPerSec = 44100; + wf.wBitsPerSample = 32; + wf.nBlockAlign = wf.nChannels * wf.wBitsPerSample / 8; + wf.nAvgBytesPerSec = wf.nSamplesPerSec * wf.nBlockAlign; + wf.cbSize = sizeof(MA_WAVEFORMATEX); + + result = MA_SUCCESS; + } else { + result = MA_FORMAT_NOT_SUPPORTED; + } + } else { + /* + I've seen cases where cbSize will be set to sizeof(WAVEFORMATEX) even though the structure itself + is given the format tag of WAVE_FORMAT_EXTENSIBLE. If the format tag is WAVE_FORMAT_EXTENSIBLE + want to make sure we copy the whole WAVEFORMATEXTENSIBLE structure. Otherwise we'll have to be + safe and only copy the WAVEFORMATEX part. + */ + if (pNativeFormat->wFormatTag == WAVE_FORMAT_EXTENSIBLE) { + MA_COPY_MEMORY(&wf, pNativeFormat, sizeof(MA_WAVEFORMATEXTENSIBLE)); + } else { + /* I've seen a case where cbSize was set to 0. Assume sizeof(WAVEFORMATEX) in this case. */ + size_t cbSize = pNativeFormat->cbSize; + if (cbSize == 0) { + cbSize = sizeof(MA_WAVEFORMATEX); + } + + /* Make sure we don't copy more than the capacity of `wf`. */ + if (cbSize > sizeof(wf)) { + cbSize = sizeof(wf); + } + + MA_COPY_MEMORY(&wf, pNativeFormat, cbSize); + } + + result = MA_SUCCESS; + } + + ma_CoTaskMemFree(pContext, pNativeFormat); + + shareMode = MA_AUDCLNT_SHAREMODE_SHARED; + } + + /* Return an error if we still haven't found a format. */ + if (result != MA_SUCCESS) { + errorMsg = "[WASAPI] Failed to find best device mix format."; + goto done; + } + + /* + Override the native sample rate with the one requested by the caller, but only if we're not using the default sample rate. We'll use + WASAPI to perform the sample rate conversion. + */ + nativeSampleRate = wf.nSamplesPerSec; + if (streamFlags & MA_AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM) { + wf.nSamplesPerSec = (pData->sampleRateIn != 0) ? pData->sampleRateIn : MA_DEFAULT_SAMPLE_RATE; + wf.nAvgBytesPerSec = wf.nSamplesPerSec * wf.nBlockAlign; + } + + pData->formatOut = ma_format_from_WAVEFORMATEX((MA_WAVEFORMATEX*)&wf); + if (pData->formatOut == ma_format_unknown) { + /* + The format isn't supported. This is almost certainly because the exclusive mode format isn't supported by miniaudio. We need to return MA_SHARE_MODE_NOT_SUPPORTED + in this case so that the caller can detect it and fall back to shared mode if desired. We should never get here if shared mode was requested, but just for + completeness we'll check for it and return MA_FORMAT_NOT_SUPPORTED. + */ + if (shareMode == MA_AUDCLNT_SHAREMODE_EXCLUSIVE) { + result = MA_SHARE_MODE_NOT_SUPPORTED; + } else { + result = MA_FORMAT_NOT_SUPPORTED; + } + + errorMsg = "[WASAPI] Native format not supported."; + goto done; + } + + pData->channelsOut = wf.nChannels; + pData->sampleRateOut = wf.nSamplesPerSec; + + /* + Get the internal channel map based on the channel mask. There is a possibility that GetMixFormat() returns + a WAVEFORMATEX instead of a WAVEFORMATEXTENSIBLE, in which case the channel mask will be undefined. In this + case we'll just use the default channel map. + */ + if (wf.wFormatTag == WAVE_FORMAT_EXTENSIBLE || wf.cbSize >= sizeof(MA_WAVEFORMATEXTENSIBLE)) { + ma_channel_mask_to_channel_map__win32(wf.dwChannelMask, pData->channelsOut, pData->channelMapOut); + } else { + ma_channel_map_init_standard(ma_standard_channel_map_microsoft, pData->channelMapOut, ma_countof(pData->channelMapOut), pData->channelsOut); + } + + /* Period size. */ + pData->periodsOut = (pData->periodsIn != 0) ? pData->periodsIn : MA_DEFAULT_PERIODS; + pData->periodSizeInFramesOut = pData->periodSizeInFramesIn; + if (pData->periodSizeInFramesOut == 0) { + if (pData->periodSizeInMillisecondsIn == 0) { + if (pData->performanceProfile == ma_performance_profile_low_latency) { + pData->periodSizeInFramesOut = ma_calculate_buffer_size_in_frames_from_milliseconds(MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_LOW_LATENCY, wf.nSamplesPerSec); + } else { + pData->periodSizeInFramesOut = ma_calculate_buffer_size_in_frames_from_milliseconds(MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_CONSERVATIVE, wf.nSamplesPerSec); + } + } else { + pData->periodSizeInFramesOut = ma_calculate_buffer_size_in_frames_from_milliseconds(pData->periodSizeInMillisecondsIn, wf.nSamplesPerSec); + } + } + + periodDurationInMicroseconds = ((ma_uint64)pData->periodSizeInFramesOut * 1000 * 1000) / wf.nSamplesPerSec; + + + /* Slightly different initialization for shared and exclusive modes. We try exclusive mode first, and if it fails, fall back to shared mode. */ + if (shareMode == MA_AUDCLNT_SHAREMODE_EXCLUSIVE) { + MA_REFERENCE_TIME bufferDuration = periodDurationInMicroseconds * pData->periodsOut * 10; + + /* + If the periodicity is too small, Initialize() will fail with AUDCLNT_E_INVALID_DEVICE_PERIOD. In this case we should just keep increasing + it and trying it again. + */ + hr = E_FAIL; + for (;;) { + hr = ma_IAudioClient_Initialize((ma_IAudioClient*)pData->pAudioClient, shareMode, streamFlags, bufferDuration, bufferDuration, (MA_WAVEFORMATEX*)&wf, NULL); + if (hr == MA_AUDCLNT_E_INVALID_DEVICE_PERIOD) { + if (bufferDuration > 500*10000) { + break; + } else { + if (bufferDuration == 0) { /* <-- Just a sanity check to prevent an infinite loop. Should never happen, but it makes me feel better. */ + break; + } + + bufferDuration = bufferDuration * 2; + continue; + } + } else { + break; + } + } + + if (hr == MA_AUDCLNT_E_BUFFER_SIZE_NOT_ALIGNED) { + ma_uint32 bufferSizeInFrames; + hr = ma_IAudioClient_GetBufferSize((ma_IAudioClient*)pData->pAudioClient, &bufferSizeInFrames); + if (SUCCEEDED(hr)) { + bufferDuration = (MA_REFERENCE_TIME)((10000.0 * 1000 / wf.nSamplesPerSec * bufferSizeInFrames) + 0.5); + + /* Unfortunately we need to release and re-acquire the audio client according to MSDN. Seems silly - why not just call IAudioClient_Initialize() again?! */ + ma_IAudioClient_Release((ma_IAudioClient*)pData->pAudioClient); + + #if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + hr = ma_IMMDevice_Activate(pDeviceInterface, &MA_IID_IAudioClient, CLSCTX_ALL, NULL, (void**)&pData->pAudioClient); + #else + hr = ma_IUnknown_QueryInterface(pDeviceInterface, &MA_IID_IAudioClient, (void**)&pData->pAudioClient); + #endif + + if (SUCCEEDED(hr)) { + hr = ma_IAudioClient_Initialize((ma_IAudioClient*)pData->pAudioClient, shareMode, streamFlags, bufferDuration, bufferDuration, (MA_WAVEFORMATEX*)&wf, NULL); + } + } + } + + if (FAILED(hr)) { + /* Failed to initialize in exclusive mode. Don't fall back to shared mode - instead tell the client about it. They can reinitialize in shared mode if they want. */ + if (hr == E_ACCESSDENIED) { + errorMsg = "[WASAPI] Failed to initialize device in exclusive mode. Access denied.", result = MA_ACCESS_DENIED; + } else if (hr == MA_AUDCLNT_E_DEVICE_IN_USE) { + errorMsg = "[WASAPI] Failed to initialize device in exclusive mode. Device in use.", result = MA_BUSY; + } else { + errorMsg = "[WASAPI] Failed to initialize device in exclusive mode."; result = ma_result_from_HRESULT(hr); + } + goto done; + } + } + + if (shareMode == MA_AUDCLNT_SHAREMODE_SHARED) { + /* + Low latency shared mode via IAudioClient3. + + NOTE + ==== + Contrary to the documentation on MSDN (https://docs.microsoft.com/en-us/windows/win32/api/audioclient/nf-audioclient-iaudioclient3-initializesharedaudiostream), the + use of AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM and AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY with IAudioClient3_InitializeSharedAudioStream() absolutely does not work. Using + any of these flags will result in HRESULT code 0x88890021. The other problem is that calling IAudioClient3_GetSharedModeEnginePeriod() with a sample rate different to + that returned by IAudioClient_GetMixFormat() also results in an error. I'm therefore disabling low-latency shared mode with AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM. + */ + #ifndef MA_WASAPI_NO_LOW_LATENCY_SHARED_MODE + { + if ((streamFlags & MA_AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM) == 0 || nativeSampleRate == wf.nSamplesPerSec) { + ma_IAudioClient3* pAudioClient3 = NULL; + hr = ma_IAudioClient_QueryInterface(pData->pAudioClient, &MA_IID_IAudioClient3, (void**)&pAudioClient3); + if (SUCCEEDED(hr)) { + ma_uint32 defaultPeriodInFrames; + ma_uint32 fundamentalPeriodInFrames; + ma_uint32 minPeriodInFrames; + ma_uint32 maxPeriodInFrames; + hr = ma_IAudioClient3_GetSharedModeEnginePeriod(pAudioClient3, (MA_WAVEFORMATEX*)&wf, &defaultPeriodInFrames, &fundamentalPeriodInFrames, &minPeriodInFrames, &maxPeriodInFrames); + if (SUCCEEDED(hr)) { + ma_uint32 desiredPeriodInFrames = pData->periodSizeInFramesOut; + ma_uint32 actualPeriodInFrames = desiredPeriodInFrames; + + /* Make sure the period size is a multiple of fundamentalPeriodInFrames. */ + actualPeriodInFrames = actualPeriodInFrames / fundamentalPeriodInFrames; + actualPeriodInFrames = actualPeriodInFrames * fundamentalPeriodInFrames; + + /* The period needs to be clamped between minPeriodInFrames and maxPeriodInFrames. */ + actualPeriodInFrames = ma_clamp(actualPeriodInFrames, minPeriodInFrames, maxPeriodInFrames); + + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "[WASAPI] Trying IAudioClient3_InitializeSharedAudioStream(actualPeriodInFrames=%d)\n", actualPeriodInFrames); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " defaultPeriodInFrames=%d\n", defaultPeriodInFrames); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " fundamentalPeriodInFrames=%d\n", fundamentalPeriodInFrames); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " minPeriodInFrames=%d\n", minPeriodInFrames); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " maxPeriodInFrames=%d\n", maxPeriodInFrames); + + /* If the client requested a largish buffer than we don't actually want to use low latency shared mode because it forces small buffers. */ + if (actualPeriodInFrames >= desiredPeriodInFrames) { + /* + MA_AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM | MA_AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY must not be in the stream flags. If either of these are specified, + IAudioClient3_InitializeSharedAudioStream() will fail. + */ + hr = ma_IAudioClient3_InitializeSharedAudioStream(pAudioClient3, streamFlags & ~(MA_AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM | MA_AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY), actualPeriodInFrames, (MA_WAVEFORMATEX*)&wf, NULL); + if (SUCCEEDED(hr)) { + wasInitializedUsingIAudioClient3 = MA_TRUE; + pData->periodSizeInFramesOut = actualPeriodInFrames; + + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "[WASAPI] Using IAudioClient3\n"); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " periodSizeInFramesOut=%d\n", pData->periodSizeInFramesOut); + } else { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "[WASAPI] IAudioClient3_InitializeSharedAudioStream failed. Falling back to IAudioClient.\n"); + } + } else { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "[WASAPI] Not using IAudioClient3 because the desired period size is larger than the maximum supported by IAudioClient3.\n"); + } + } else { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "[WASAPI] IAudioClient3_GetSharedModeEnginePeriod failed. Falling back to IAudioClient.\n"); + } + + ma_IAudioClient3_Release(pAudioClient3); + pAudioClient3 = NULL; + } + } + } + #else + { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "[WASAPI] Not using IAudioClient3 because MA_WASAPI_NO_LOW_LATENCY_SHARED_MODE is enabled.\n"); + } + #endif + + /* If we don't have an IAudioClient3 then we need to use the normal initialization routine. */ + if (!wasInitializedUsingIAudioClient3) { + MA_REFERENCE_TIME bufferDuration = periodDurationInMicroseconds * pData->periodsOut * 10; /* <-- Multiply by 10 for microseconds to 100-nanoseconds. */ + hr = ma_IAudioClient_Initialize((ma_IAudioClient*)pData->pAudioClient, shareMode, streamFlags, bufferDuration, 0, (const MA_WAVEFORMATEX*)&wf, NULL); + if (FAILED(hr)) { + if (hr == E_ACCESSDENIED) { + errorMsg = "[WASAPI] Failed to initialize device. Access denied.", result = MA_ACCESS_DENIED; + } else if (hr == MA_AUDCLNT_E_DEVICE_IN_USE) { + errorMsg = "[WASAPI] Failed to initialize device. Device in use.", result = MA_BUSY; + } else { + errorMsg = "[WASAPI] Failed to initialize device.", result = ma_result_from_HRESULT(hr); + } + + goto done; + } + } + } + + if (!wasInitializedUsingIAudioClient3) { + ma_uint32 bufferSizeInFrames = 0; + hr = ma_IAudioClient_GetBufferSize((ma_IAudioClient*)pData->pAudioClient, &bufferSizeInFrames); + if (FAILED(hr)) { + errorMsg = "[WASAPI] Failed to get audio client's actual buffer size.", result = ma_result_from_HRESULT(hr); + goto done; + } + + /* + When using process loopback mode, retrieval of the buffer size seems to result in totally + incorrect values. In this case we'll just assume it's the same size as what we requested + when we initialized the client. + */ + if (usingProcessLoopback) { + bufferSizeInFrames = (ma_uint32)((periodDurationInMicroseconds * pData->periodsOut) * pData->sampleRateOut / 1000000); + } + + pData->periodSizeInFramesOut = bufferSizeInFrames / pData->periodsOut; + } + + pData->usingAudioClient3 = wasInitializedUsingIAudioClient3; + + + if (deviceType == ma_device_type_playback) { + result = ma_device_create_IAudioClient_service__wasapi(pContext, deviceType, (ma_IAudioClient*)pData->pAudioClient, (void**)&pData->pRenderClient); + } else { + result = ma_device_create_IAudioClient_service__wasapi(pContext, deviceType, (ma_IAudioClient*)pData->pAudioClient, (void**)&pData->pCaptureClient); + } + + /*if (FAILED(hr)) {*/ + if (result != MA_SUCCESS) { + errorMsg = "[WASAPI] Failed to get audio client service."; + goto done; + } + + + /* Grab the name of the device. */ + #if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + { + ma_IPropertyStore *pProperties; + hr = ma_IMMDevice_OpenPropertyStore(pDeviceInterface, STGM_READ, &pProperties); + if (SUCCEEDED(hr)) { + MA_PROPVARIANT varName; + ma_PropVariantInit(&varName); + hr = ma_IPropertyStore_GetValue(pProperties, &MA_PKEY_Device_FriendlyName, &varName); + if (SUCCEEDED(hr)) { + WideCharToMultiByte(CP_UTF8, 0, varName.pwszVal, -1, pData->deviceName, sizeof(pData->deviceName), 0, FALSE); + ma_PropVariantClear(pContext, &varName); + } + + ma_IPropertyStore_Release(pProperties); + } + } + #endif + + /* + For the WASAPI backend we need to know the actual IDs of the device in order to do automatic + stream routing so that IDs can be compared and we can determine which device has been detached + and whether or not it matches with our ma_device. + */ + #if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + { + /* Desktop */ + ma_context_get_device_id_from_MMDevice__wasapi(pContext, pDeviceInterface, &pData->id); + } + #else + { + /* UWP */ + /* TODO: Implement me. Need to figure out how to get the ID of the default device. */ + } + #endif + +done: + /* Clean up. */ +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + if (pDeviceInterface != NULL) { + ma_IMMDevice_Release(pDeviceInterface); + } +#else + if (pDeviceInterface != NULL) { + ma_IUnknown_Release(pDeviceInterface); + } +#endif + + if (result != MA_SUCCESS) { + if (pData->pRenderClient) { + ma_IAudioRenderClient_Release((ma_IAudioRenderClient*)pData->pRenderClient); + pData->pRenderClient = NULL; + } + if (pData->pCaptureClient) { + ma_IAudioCaptureClient_Release((ma_IAudioCaptureClient*)pData->pCaptureClient); + pData->pCaptureClient = NULL; + } + if (pData->pAudioClient) { + ma_IAudioClient_Release((ma_IAudioClient*)pData->pAudioClient); + pData->pAudioClient = NULL; + } + + if (errorMsg != NULL && errorMsg[0] != '\0') { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "%s\n", errorMsg); + } + + return result; + } else { + return MA_SUCCESS; + } +} + +static ma_result ma_device_reinit__wasapi(ma_device* pDevice, ma_device_type deviceType) +{ + ma_device_init_internal_data__wasapi data; + ma_result result; + + MA_ASSERT(pDevice != NULL); + + /* We only re-initialize the playback or capture device. Never a full-duplex device. */ + if (deviceType == ma_device_type_duplex) { + return MA_INVALID_ARGS; + } + + + /* + Before reinitializing the device we need to free the previous audio clients. + + There's a known memory leak here. We will be calling this from the routing change callback that + is fired by WASAPI. If we attempt to release the IAudioClient we will deadlock. In my opinion + this is a bug. I'm not sure what I need to do to handle this cleanly, but I think we'll probably + need some system where we post an event, but delay the execution of it until the callback has + returned. I'm not sure how to do this reliably, however. I have set up some infrastructure for + a command thread which might be useful for this. + */ + if (deviceType == ma_device_type_capture || deviceType == ma_device_type_loopback) { + if (pDevice->wasapi.pCaptureClient) { + ma_IAudioCaptureClient_Release((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient); + pDevice->wasapi.pCaptureClient = NULL; + } + + if (pDevice->wasapi.pAudioClientCapture) { + /*ma_device_release_IAudioClient_service__wasapi(pDevice, ma_device_type_capture);*/ + pDevice->wasapi.pAudioClientCapture = NULL; + } + } + + if (deviceType == ma_device_type_playback) { + if (pDevice->wasapi.pRenderClient) { + ma_IAudioRenderClient_Release((ma_IAudioRenderClient*)pDevice->wasapi.pRenderClient); + pDevice->wasapi.pRenderClient = NULL; + } + + if (pDevice->wasapi.pAudioClientPlayback) { + /*ma_device_release_IAudioClient_service__wasapi(pDevice, ma_device_type_playback);*/ + pDevice->wasapi.pAudioClientPlayback = NULL; + } + } + + + if (deviceType == ma_device_type_playback) { + data.formatIn = pDevice->playback.format; + data.channelsIn = pDevice->playback.channels; + MA_COPY_MEMORY(data.channelMapIn, pDevice->playback.channelMap, sizeof(pDevice->playback.channelMap)); + data.shareMode = pDevice->playback.shareMode; + } else { + data.formatIn = pDevice->capture.format; + data.channelsIn = pDevice->capture.channels; + MA_COPY_MEMORY(data.channelMapIn, pDevice->capture.channelMap, sizeof(pDevice->capture.channelMap)); + data.shareMode = pDevice->capture.shareMode; + } + + data.sampleRateIn = pDevice->sampleRate; + data.periodSizeInFramesIn = pDevice->wasapi.originalPeriodSizeInFrames; + data.periodSizeInMillisecondsIn = pDevice->wasapi.originalPeriodSizeInMilliseconds; + data.periodsIn = pDevice->wasapi.originalPeriods; + data.performanceProfile = pDevice->wasapi.originalPerformanceProfile; + data.noAutoConvertSRC = pDevice->wasapi.noAutoConvertSRC; + data.noDefaultQualitySRC = pDevice->wasapi.noDefaultQualitySRC; + data.noHardwareOffloading = pDevice->wasapi.noHardwareOffloading; + data.loopbackProcessID = pDevice->wasapi.loopbackProcessID; + data.loopbackProcessExclude = pDevice->wasapi.loopbackProcessExclude; + result = ma_device_init_internal__wasapi(pDevice->pContext, deviceType, NULL, &data); + if (result != MA_SUCCESS) { + return result; + } + + /* At this point we have some new objects ready to go. We need to uninitialize the previous ones and then set the new ones. */ + if (deviceType == ma_device_type_capture || deviceType == ma_device_type_loopback) { + pDevice->wasapi.pAudioClientCapture = data.pAudioClient; + pDevice->wasapi.pCaptureClient = data.pCaptureClient; + + pDevice->capture.internalFormat = data.formatOut; + pDevice->capture.internalChannels = data.channelsOut; + pDevice->capture.internalSampleRate = data.sampleRateOut; + MA_COPY_MEMORY(pDevice->capture.internalChannelMap, data.channelMapOut, sizeof(data.channelMapOut)); + pDevice->capture.internalPeriodSizeInFrames = data.periodSizeInFramesOut; + pDevice->capture.internalPeriods = data.periodsOut; + ma_strcpy_s(pDevice->capture.name, sizeof(pDevice->capture.name), data.deviceName); + + ma_IAudioClient_SetEventHandle((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture, (HANDLE)pDevice->wasapi.hEventCapture); + + pDevice->wasapi.periodSizeInFramesCapture = data.periodSizeInFramesOut; + ma_IAudioClient_GetBufferSize((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture, &pDevice->wasapi.actualBufferSizeInFramesCapture); + + /* We must always have a valid ID. */ + ma_strcpy_s_WCHAR(pDevice->capture.id.wasapi, sizeof(pDevice->capture.id.wasapi), data.id.wasapi); + } + + if (deviceType == ma_device_type_playback) { + pDevice->wasapi.pAudioClientPlayback = data.pAudioClient; + pDevice->wasapi.pRenderClient = data.pRenderClient; + + pDevice->playback.internalFormat = data.formatOut; + pDevice->playback.internalChannels = data.channelsOut; + pDevice->playback.internalSampleRate = data.sampleRateOut; + MA_COPY_MEMORY(pDevice->playback.internalChannelMap, data.channelMapOut, sizeof(data.channelMapOut)); + pDevice->playback.internalPeriodSizeInFrames = data.periodSizeInFramesOut; + pDevice->playback.internalPeriods = data.periodsOut; + ma_strcpy_s(pDevice->playback.name, sizeof(pDevice->playback.name), data.deviceName); + + ma_IAudioClient_SetEventHandle((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback, (HANDLE)pDevice->wasapi.hEventPlayback); + + pDevice->wasapi.periodSizeInFramesPlayback = data.periodSizeInFramesOut; + ma_IAudioClient_GetBufferSize((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback, &pDevice->wasapi.actualBufferSizeInFramesPlayback); + + /* We must always have a valid ID because rerouting will look at it. */ + ma_strcpy_s_WCHAR(pDevice->playback.id.wasapi, sizeof(pDevice->playback.id.wasapi), data.id.wasapi); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_init__wasapi(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + ma_result result = MA_SUCCESS; + +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + HRESULT hr; + ma_IMMDeviceEnumerator* pDeviceEnumerator; +#endif + + MA_ASSERT(pDevice != NULL); + + MA_ZERO_OBJECT(&pDevice->wasapi); + pDevice->wasapi.usage = pConfig->wasapi.usage; + pDevice->wasapi.noAutoConvertSRC = pConfig->wasapi.noAutoConvertSRC; + pDevice->wasapi.noDefaultQualitySRC = pConfig->wasapi.noDefaultQualitySRC; + pDevice->wasapi.noHardwareOffloading = pConfig->wasapi.noHardwareOffloading; + pDevice->wasapi.loopbackProcessID = pConfig->wasapi.loopbackProcessID; + pDevice->wasapi.loopbackProcessExclude = pConfig->wasapi.loopbackProcessExclude; + + /* Exclusive mode is not allowed with loopback. */ + if (pConfig->deviceType == ma_device_type_loopback && pConfig->playback.shareMode == ma_share_mode_exclusive) { + return MA_INVALID_DEVICE_CONFIG; + } + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex || pConfig->deviceType == ma_device_type_loopback) { + ma_device_init_internal_data__wasapi data; + data.formatIn = pDescriptorCapture->format; + data.channelsIn = pDescriptorCapture->channels; + data.sampleRateIn = pDescriptorCapture->sampleRate; + MA_COPY_MEMORY(data.channelMapIn, pDescriptorCapture->channelMap, sizeof(pDescriptorCapture->channelMap)); + data.periodSizeInFramesIn = pDescriptorCapture->periodSizeInFrames; + data.periodSizeInMillisecondsIn = pDescriptorCapture->periodSizeInMilliseconds; + data.periodsIn = pDescriptorCapture->periodCount; + data.shareMode = pDescriptorCapture->shareMode; + data.performanceProfile = pConfig->performanceProfile; + data.noAutoConvertSRC = pConfig->wasapi.noAutoConvertSRC; + data.noDefaultQualitySRC = pConfig->wasapi.noDefaultQualitySRC; + data.noHardwareOffloading = pConfig->wasapi.noHardwareOffloading; + data.loopbackProcessID = pConfig->wasapi.loopbackProcessID; + data.loopbackProcessExclude = pConfig->wasapi.loopbackProcessExclude; + + result = ma_device_init_internal__wasapi(pDevice->pContext, (pConfig->deviceType == ma_device_type_loopback) ? ma_device_type_loopback : ma_device_type_capture, pDescriptorCapture->pDeviceID, &data); + if (result != MA_SUCCESS) { + return result; + } + + pDevice->wasapi.pAudioClientCapture = data.pAudioClient; + pDevice->wasapi.pCaptureClient = data.pCaptureClient; + pDevice->wasapi.originalPeriodSizeInMilliseconds = pDescriptorCapture->periodSizeInMilliseconds; + pDevice->wasapi.originalPeriodSizeInFrames = pDescriptorCapture->periodSizeInFrames; + pDevice->wasapi.originalPeriods = pDescriptorCapture->periodCount; + pDevice->wasapi.originalPerformanceProfile = pConfig->performanceProfile; + + /* + The event for capture needs to be manual reset for the same reason as playback. We keep the initial state set to unsignaled, + however, because we want to block until we actually have something for the first call to ma_device_read(). + */ + pDevice->wasapi.hEventCapture = (ma_handle)CreateEventA(NULL, FALSE, FALSE, NULL); /* Auto reset, unsignaled by default. */ + if (pDevice->wasapi.hEventCapture == NULL) { + result = ma_result_from_GetLastError(GetLastError()); + + if (pDevice->wasapi.pCaptureClient != NULL) { + ma_IAudioCaptureClient_Release((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient); + pDevice->wasapi.pCaptureClient = NULL; + } + if (pDevice->wasapi.pAudioClientCapture != NULL) { + ma_IAudioClient_Release((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture); + pDevice->wasapi.pAudioClientCapture = NULL; + } + + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to create event for capture."); + return result; + } + ma_IAudioClient_SetEventHandle((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture, (HANDLE)pDevice->wasapi.hEventCapture); + + pDevice->wasapi.periodSizeInFramesCapture = data.periodSizeInFramesOut; + ma_IAudioClient_GetBufferSize((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture, &pDevice->wasapi.actualBufferSizeInFramesCapture); + + /* We must always have a valid ID. */ + ma_strcpy_s_WCHAR(pDevice->capture.id.wasapi, sizeof(pDevice->capture.id.wasapi), data.id.wasapi); + + /* The descriptor needs to be updated with actual values. */ + pDescriptorCapture->format = data.formatOut; + pDescriptorCapture->channels = data.channelsOut; + pDescriptorCapture->sampleRate = data.sampleRateOut; + MA_COPY_MEMORY(pDescriptorCapture->channelMap, data.channelMapOut, sizeof(data.channelMapOut)); + pDescriptorCapture->periodSizeInFrames = data.periodSizeInFramesOut; + pDescriptorCapture->periodCount = data.periodsOut; + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_device_init_internal_data__wasapi data; + data.formatIn = pDescriptorPlayback->format; + data.channelsIn = pDescriptorPlayback->channels; + data.sampleRateIn = pDescriptorPlayback->sampleRate; + MA_COPY_MEMORY(data.channelMapIn, pDescriptorPlayback->channelMap, sizeof(pDescriptorPlayback->channelMap)); + data.periodSizeInFramesIn = pDescriptorPlayback->periodSizeInFrames; + data.periodSizeInMillisecondsIn = pDescriptorPlayback->periodSizeInMilliseconds; + data.periodsIn = pDescriptorPlayback->periodCount; + data.shareMode = pDescriptorPlayback->shareMode; + data.performanceProfile = pConfig->performanceProfile; + data.noAutoConvertSRC = pConfig->wasapi.noAutoConvertSRC; + data.noDefaultQualitySRC = pConfig->wasapi.noDefaultQualitySRC; + data.noHardwareOffloading = pConfig->wasapi.noHardwareOffloading; + data.loopbackProcessID = pConfig->wasapi.loopbackProcessID; + data.loopbackProcessExclude = pConfig->wasapi.loopbackProcessExclude; + + result = ma_device_init_internal__wasapi(pDevice->pContext, ma_device_type_playback, pDescriptorPlayback->pDeviceID, &data); + if (result != MA_SUCCESS) { + if (pConfig->deviceType == ma_device_type_duplex) { + if (pDevice->wasapi.pCaptureClient != NULL) { + ma_IAudioCaptureClient_Release((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient); + pDevice->wasapi.pCaptureClient = NULL; + } + if (pDevice->wasapi.pAudioClientCapture != NULL) { + ma_IAudioClient_Release((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture); + pDevice->wasapi.pAudioClientCapture = NULL; + } + + CloseHandle((HANDLE)pDevice->wasapi.hEventCapture); + pDevice->wasapi.hEventCapture = NULL; + } + return result; + } + + pDevice->wasapi.pAudioClientPlayback = data.pAudioClient; + pDevice->wasapi.pRenderClient = data.pRenderClient; + pDevice->wasapi.originalPeriodSizeInMilliseconds = pDescriptorPlayback->periodSizeInMilliseconds; + pDevice->wasapi.originalPeriodSizeInFrames = pDescriptorPlayback->periodSizeInFrames; + pDevice->wasapi.originalPeriods = pDescriptorPlayback->periodCount; + pDevice->wasapi.originalPerformanceProfile = pConfig->performanceProfile; + + /* + The event for playback is needs to be manual reset because we want to explicitly control the fact that it becomes signalled + only after the whole available space has been filled, never before. + + The playback event also needs to be initially set to a signaled state so that the first call to ma_device_write() is able + to get passed WaitForMultipleObjects(). + */ + pDevice->wasapi.hEventPlayback = (ma_handle)CreateEventA(NULL, FALSE, TRUE, NULL); /* Auto reset, signaled by default. */ + if (pDevice->wasapi.hEventPlayback == NULL) { + result = ma_result_from_GetLastError(GetLastError()); + + if (pConfig->deviceType == ma_device_type_duplex) { + if (pDevice->wasapi.pCaptureClient != NULL) { + ma_IAudioCaptureClient_Release((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient); + pDevice->wasapi.pCaptureClient = NULL; + } + if (pDevice->wasapi.pAudioClientCapture != NULL) { + ma_IAudioClient_Release((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture); + pDevice->wasapi.pAudioClientCapture = NULL; + } + + CloseHandle((HANDLE)pDevice->wasapi.hEventCapture); + pDevice->wasapi.hEventCapture = NULL; + } + + if (pDevice->wasapi.pRenderClient != NULL) { + ma_IAudioRenderClient_Release((ma_IAudioRenderClient*)pDevice->wasapi.pRenderClient); + pDevice->wasapi.pRenderClient = NULL; + } + if (pDevice->wasapi.pAudioClientPlayback != NULL) { + ma_IAudioClient_Release((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback); + pDevice->wasapi.pAudioClientPlayback = NULL; + } + + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to create event for playback."); + return result; + } + ma_IAudioClient_SetEventHandle((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback, (HANDLE)pDevice->wasapi.hEventPlayback); + + pDevice->wasapi.periodSizeInFramesPlayback = data.periodSizeInFramesOut; + ma_IAudioClient_GetBufferSize((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback, &pDevice->wasapi.actualBufferSizeInFramesPlayback); + + /* We must always have a valid ID because rerouting will look at it. */ + ma_strcpy_s_WCHAR(pDevice->playback.id.wasapi, sizeof(pDevice->playback.id.wasapi), data.id.wasapi); + + /* The descriptor needs to be updated with actual values. */ + pDescriptorPlayback->format = data.formatOut; + pDescriptorPlayback->channels = data.channelsOut; + pDescriptorPlayback->sampleRate = data.sampleRateOut; + MA_COPY_MEMORY(pDescriptorPlayback->channelMap, data.channelMapOut, sizeof(data.channelMapOut)); + pDescriptorPlayback->periodSizeInFrames = data.periodSizeInFramesOut; + pDescriptorPlayback->periodCount = data.periodsOut; + } + + /* + We need to register a notification client to detect when the device has been disabled, unplugged or re-routed (when the default device changes). When + we are connecting to the default device we want to do automatic stream routing when the device is disabled or unplugged. Otherwise we want to just + stop the device outright and let the application handle it. + */ +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + if (pConfig->wasapi.noAutoStreamRouting == MA_FALSE) { + if ((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex || pConfig->deviceType == ma_device_type_loopback) && pConfig->capture.pDeviceID == NULL) { + pDevice->wasapi.allowCaptureAutoStreamRouting = MA_TRUE; + } + if ((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pConfig->playback.pDeviceID == NULL) { + pDevice->wasapi.allowPlaybackAutoStreamRouting = MA_TRUE; + } + } + + ma_mutex_init(&pDevice->wasapi.rerouteLock); + + hr = ma_CoCreateInstance(pDevice->pContext, &MA_CLSID_MMDeviceEnumerator, NULL, CLSCTX_ALL, &MA_IID_IMMDeviceEnumerator, (void**)&pDeviceEnumerator); + if (FAILED(hr)) { + ma_device_uninit__wasapi(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to create device enumerator."); + return ma_result_from_HRESULT(hr); + } + + pDevice->wasapi.notificationClient.lpVtbl = (void*)&g_maNotificationCientVtbl; + pDevice->wasapi.notificationClient.counter = 1; + pDevice->wasapi.notificationClient.pDevice = pDevice; + + hr = pDeviceEnumerator->lpVtbl->RegisterEndpointNotificationCallback(pDeviceEnumerator, &pDevice->wasapi.notificationClient); + if (SUCCEEDED(hr)) { + pDevice->wasapi.pDeviceEnumerator = (ma_ptr)pDeviceEnumerator; + } else { + /* Not the end of the world if we fail to register the notification callback. We just won't support automatic stream routing. */ + ma_IMMDeviceEnumerator_Release(pDeviceEnumerator); + } +#endif + + ma_atomic_bool32_set(&pDevice->wasapi.isStartedCapture, MA_FALSE); + ma_atomic_bool32_set(&pDevice->wasapi.isStartedPlayback, MA_FALSE); + + return MA_SUCCESS; +} + +static ma_result ma_device__get_available_frames__wasapi(ma_device* pDevice, ma_IAudioClient* pAudioClient, ma_uint32* pFrameCount) +{ + ma_uint32 paddingFramesCount; + HRESULT hr; + ma_share_mode shareMode; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pFrameCount != NULL); + + *pFrameCount = 0; + + if ((ma_ptr)pAudioClient != pDevice->wasapi.pAudioClientPlayback && (ma_ptr)pAudioClient != pDevice->wasapi.pAudioClientCapture) { + return MA_INVALID_OPERATION; + } + + /* + I've had a report that GetCurrentPadding() is returning a frame count of 0 which is preventing + higher level function calls from doing anything because it thinks nothing is available. I have + taken a look at the documentation and it looks like this is unnecessary in exclusive mode. + + From Microsoft's documentation: + + For an exclusive-mode rendering or capture stream that was initialized with the + AUDCLNT_STREAMFLAGS_EVENTCALLBACK flag, the client typically has no use for the padding + value reported by GetCurrentPadding. Instead, the client accesses an entire buffer during + each processing pass. + + Considering this, I'm going to skip GetCurrentPadding() for exclusive mode and just report the + entire buffer. This depends on the caller making sure they wait on the event handler. + */ + shareMode = ((ma_ptr)pAudioClient == pDevice->wasapi.pAudioClientPlayback) ? pDevice->playback.shareMode : pDevice->capture.shareMode; + if (shareMode == ma_share_mode_shared) { + /* Shared mode. */ + hr = ma_IAudioClient_GetCurrentPadding(pAudioClient, &paddingFramesCount); + if (FAILED(hr)) { + return ma_result_from_HRESULT(hr); + } + + if ((ma_ptr)pAudioClient == pDevice->wasapi.pAudioClientPlayback) { + *pFrameCount = pDevice->wasapi.actualBufferSizeInFramesPlayback - paddingFramesCount; + } else { + *pFrameCount = paddingFramesCount; + } + } else { + /* Exclusive mode. */ + if ((ma_ptr)pAudioClient == pDevice->wasapi.pAudioClientPlayback) { + *pFrameCount = pDevice->wasapi.actualBufferSizeInFramesPlayback; + } else { + *pFrameCount = pDevice->wasapi.actualBufferSizeInFramesCapture; + } + } + + return MA_SUCCESS; +} + + +static ma_result ma_device_reroute__wasapi(ma_device* pDevice, ma_device_type deviceType) +{ + ma_result result; + + if (deviceType == ma_device_type_duplex) { + return MA_INVALID_ARGS; + } + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "=== CHANGING DEVICE ===\n"); + + result = ma_device_reinit__wasapi(pDevice, deviceType); + if (result != MA_SUCCESS) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[WASAPI] Reinitializing device after route change failed.\n"); + return result; + } + + ma_device__post_init_setup(pDevice, deviceType); + ma_device__on_notification_rerouted(pDevice); + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "=== DEVICE CHANGED ===\n"); + + return MA_SUCCESS; +} + +static ma_result ma_device_start__wasapi_nolock(ma_device* pDevice) +{ + HRESULT hr; + + if (pDevice->pContext->wasapi.hAvrt) { + const char* pTaskName = ma_to_usage_string__wasapi(pDevice->wasapi.usage); + if (pTaskName) { + DWORD idx = 0; + pDevice->wasapi.hAvrtHandle = (ma_handle)((MA_PFN_AvSetMmThreadCharacteristicsA)pDevice->pContext->wasapi.AvSetMmThreadCharacteristicsA)(pTaskName, &idx); + } + } + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex || pDevice->type == ma_device_type_loopback) { + hr = ma_IAudioClient_Start((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture); + if (FAILED(hr)) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to start internal capture device. HRESULT = %d.", (int)hr); + return ma_result_from_HRESULT(hr); + } + + ma_atomic_bool32_set(&pDevice->wasapi.isStartedCapture, MA_TRUE); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + hr = ma_IAudioClient_Start((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback); + if (FAILED(hr)) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to start internal playback device. HRESULT = %d.", (int)hr); + return ma_result_from_HRESULT(hr); + } + + ma_atomic_bool32_set(&pDevice->wasapi.isStartedPlayback, MA_TRUE); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_start__wasapi(ma_device* pDevice) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + + /* Wait for any rerouting to finish before attempting to start the device. */ + ma_mutex_lock(&pDevice->wasapi.rerouteLock); + { + result = ma_device_start__wasapi_nolock(pDevice); + } + ma_mutex_unlock(&pDevice->wasapi.rerouteLock); + + return result; +} + +static ma_result ma_device_stop__wasapi_nolock(ma_device* pDevice) +{ + ma_result result; + HRESULT hr; + + MA_ASSERT(pDevice != NULL); + + if (pDevice->wasapi.hAvrtHandle) { + ((MA_PFN_AvRevertMmThreadCharacteristics)pDevice->pContext->wasapi.AvRevertMmThreadcharacteristics)((HANDLE)pDevice->wasapi.hAvrtHandle); + pDevice->wasapi.hAvrtHandle = NULL; + } + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex || pDevice->type == ma_device_type_loopback) { + /* If we have a mapped buffer we need to release it. */ + if (pDevice->wasapi.pMappedBufferCapture != NULL) { + ma_IAudioCaptureClient_ReleaseBuffer((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient, pDevice->wasapi.mappedBufferCaptureCap); + pDevice->wasapi.pMappedBufferCapture = NULL; + pDevice->wasapi.mappedBufferCaptureCap = 0; + pDevice->wasapi.mappedBufferCaptureLen = 0; + } + + hr = ma_IAudioClient_Stop((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to stop internal capture device."); + return ma_result_from_HRESULT(hr); + } + + /* The audio client needs to be reset otherwise restarting will fail. */ + hr = ma_IAudioClient_Reset((ma_IAudioClient*)pDevice->wasapi.pAudioClientCapture); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to reset internal capture device."); + return ma_result_from_HRESULT(hr); + } + + ma_atomic_bool32_set(&pDevice->wasapi.isStartedCapture, MA_FALSE); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + if (pDevice->wasapi.pMappedBufferPlayback != NULL) { + ma_silence_pcm_frames( + ma_offset_pcm_frames_ptr(pDevice->wasapi.pMappedBufferPlayback, pDevice->wasapi.mappedBufferPlaybackLen, pDevice->playback.internalFormat, pDevice->playback.internalChannels), + pDevice->wasapi.mappedBufferPlaybackCap - pDevice->wasapi.mappedBufferPlaybackLen, + pDevice->playback.internalFormat, pDevice->playback.internalChannels + ); + ma_IAudioRenderClient_ReleaseBuffer((ma_IAudioRenderClient*)pDevice->wasapi.pRenderClient, pDevice->wasapi.mappedBufferPlaybackCap, 0); + pDevice->wasapi.pMappedBufferPlayback = NULL; + pDevice->wasapi.mappedBufferPlaybackCap = 0; + pDevice->wasapi.mappedBufferPlaybackLen = 0; + } + + /* + The buffer needs to be drained before stopping the device. Not doing this will result in the last few frames not getting output to + the speakers. This is a problem for very short sounds because it'll result in a significant portion of it not getting played. + */ + if (ma_atomic_bool32_get(&pDevice->wasapi.isStartedPlayback)) { + /* We need to make sure we put a timeout here or else we'll risk getting stuck in a deadlock in some cases. */ + DWORD waitTime = (pDevice->wasapi.actualBufferSizeInFramesPlayback * 1000) / pDevice->playback.internalSampleRate; + + if (pDevice->playback.shareMode == ma_share_mode_exclusive) { + WaitForSingleObject((HANDLE)pDevice->wasapi.hEventPlayback, waitTime); + } else { + ma_uint32 prevFramesAvailablePlayback = (ma_uint32)-1; + ma_uint32 framesAvailablePlayback; + for (;;) { + result = ma_device__get_available_frames__wasapi(pDevice, (ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback, &framesAvailablePlayback); + if (result != MA_SUCCESS) { + break; + } + + if (framesAvailablePlayback >= pDevice->wasapi.actualBufferSizeInFramesPlayback) { + break; + } + + /* + Just a safety check to avoid an infinite loop. If this iteration results in a situation where the number of available frames + has not changed, get out of the loop. I don't think this should ever happen, but I think it's nice to have just in case. + */ + if (framesAvailablePlayback == prevFramesAvailablePlayback) { + break; + } + prevFramesAvailablePlayback = framesAvailablePlayback; + + ResetEvent((HANDLE)pDevice->wasapi.hEventPlayback); /* Manual reset. */ + WaitForSingleObject((HANDLE)pDevice->wasapi.hEventPlayback, waitTime); + } + } + } + + hr = ma_IAudioClient_Stop((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to stop internal playback device."); + return ma_result_from_HRESULT(hr); + } + + /* The audio client needs to be reset otherwise restarting will fail. */ + { + ma_int32 retries = 5; + + while ((hr = ma_IAudioClient_Reset((ma_IAudioClient*)pDevice->wasapi.pAudioClientPlayback)) == MA_AUDCLNT_E_BUFFER_OPERATION_PENDING && retries > 0) { + ma_sleep(10); + retries -= 1; + } + } + + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to reset internal playback device."); + return ma_result_from_HRESULT(hr); + } + + ma_atomic_bool32_set(&pDevice->wasapi.isStartedPlayback, MA_FALSE); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__wasapi(ma_device* pDevice) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + + /* Wait for any rerouting to finish before attempting to stop the device. */ + ma_mutex_lock(&pDevice->wasapi.rerouteLock); + { + result = ma_device_stop__wasapi_nolock(pDevice); + } + ma_mutex_unlock(&pDevice->wasapi.rerouteLock); + + return result; +} + + +#ifndef MA_WASAPI_WAIT_TIMEOUT_MILLISECONDS +#define MA_WASAPI_WAIT_TIMEOUT_MILLISECONDS 5000 +#endif + +static ma_result ma_device_read__wasapi(ma_device* pDevice, void* pFrames, ma_uint32 frameCount, ma_uint32* pFramesRead) +{ + ma_result result = MA_SUCCESS; + ma_uint32 totalFramesProcessed = 0; + + /* + When reading, we need to get a buffer and process all of it before releasing it. Because the + frame count (frameCount) can be different to the size of the buffer, we'll need to cache the + pointer to the buffer. + */ + + /* Keep running until we've processed the requested number of frames. */ + while (ma_device_get_state(pDevice) == ma_device_state_started && totalFramesProcessed < frameCount) { + ma_uint32 framesRemaining = frameCount - totalFramesProcessed; + + /* If we have a mapped data buffer, consume that first. */ + if (pDevice->wasapi.pMappedBufferCapture != NULL) { + /* We have a cached data pointer so consume that before grabbing another one from WASAPI. */ + ma_uint32 framesToProcessNow = framesRemaining; + if (framesToProcessNow > pDevice->wasapi.mappedBufferCaptureLen) { + framesToProcessNow = pDevice->wasapi.mappedBufferCaptureLen; + } + + /* Now just copy the data over to the output buffer. */ + ma_copy_pcm_frames( + ma_offset_pcm_frames_ptr(pFrames, totalFramesProcessed, pDevice->capture.internalFormat, pDevice->capture.internalChannels), + ma_offset_pcm_frames_const_ptr(pDevice->wasapi.pMappedBufferCapture, pDevice->wasapi.mappedBufferCaptureCap - pDevice->wasapi.mappedBufferCaptureLen, pDevice->capture.internalFormat, pDevice->capture.internalChannels), + framesToProcessNow, + pDevice->capture.internalFormat, pDevice->capture.internalChannels + ); + + totalFramesProcessed += framesToProcessNow; + pDevice->wasapi.mappedBufferCaptureLen -= framesToProcessNow; + + /* If the data buffer has been fully consumed we need to release it. */ + if (pDevice->wasapi.mappedBufferCaptureLen == 0) { + ma_IAudioCaptureClient_ReleaseBuffer((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient, pDevice->wasapi.mappedBufferCaptureCap); + pDevice->wasapi.pMappedBufferCapture = NULL; + pDevice->wasapi.mappedBufferCaptureCap = 0; + } + } else { + /* We don't have any cached data pointer, so grab another one. */ + HRESULT hr; + DWORD flags = 0; + + /* First just ask WASAPI for a data buffer. If it's not available, we'll wait for more. */ + hr = ma_IAudioCaptureClient_GetBuffer((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient, (BYTE**)&pDevice->wasapi.pMappedBufferCapture, &pDevice->wasapi.mappedBufferCaptureCap, &flags, NULL, NULL); + if (hr == S_OK) { + /* We got a data buffer. Continue to the next loop iteration which will then read from the mapped pointer. */ + pDevice->wasapi.mappedBufferCaptureLen = pDevice->wasapi.mappedBufferCaptureCap; + + /* + There have been reports that indicate that at times the AUDCLNT_BUFFERFLAGS_DATA_DISCONTINUITY is reported for every + call to IAudioCaptureClient_GetBuffer() above which results in spamming of the debug messages below. To partially + work around this, I'm only outputting these messages when MA_DEBUG_OUTPUT is explicitly defined. The better solution + would be to figure out why the flag is always getting reported. + */ + #if defined(MA_DEBUG_OUTPUT) + { + if (flags != 0) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Capture Flags: %ld\n", flags); + + if ((flags & MA_AUDCLNT_BUFFERFLAGS_DATA_DISCONTINUITY) != 0) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Data discontinuity (possible overrun). Attempting recovery. mappedBufferCaptureCap=%d\n", pDevice->wasapi.mappedBufferCaptureCap); + } + } + } + #endif + + /* Overrun detection. */ + if ((flags & MA_AUDCLNT_BUFFERFLAGS_DATA_DISCONTINUITY) != 0) { + /* Glitched. Probably due to an overrun. */ + + /* + If we got an overrun it probably means we're straddling the end of the buffer. In normal capture + mode this is the fault of the client application because they're responsible for ensuring data is + processed fast enough. In duplex mode, however, the processing of audio is tied to the playback + device, so this can possibly be the result of a timing de-sync. + + In capture mode we're not going to do any kind of recovery because the real fix is for the client + application to process faster. In duplex mode, we'll treat this as a desync and reset the buffers + to prevent a never-ending sequence of glitches due to straddling the end of the buffer. + */ + if (pDevice->type == ma_device_type_duplex) { + /* + Experiment: + + If we empty out the *entire* buffer we may end up putting ourselves into an underrun position + which isn't really any better than the overrun we're probably in right now. Instead we'll just + empty out about half. + */ + ma_uint32 i; + ma_uint32 periodCount = (pDevice->wasapi.actualBufferSizeInFramesCapture / pDevice->wasapi.periodSizeInFramesCapture); + ma_uint32 iterationCount = periodCount / 2; + if ((periodCount % 2) > 0) { + iterationCount += 1; + } + + for (i = 0; i < iterationCount; i += 1) { + hr = ma_IAudioCaptureClient_ReleaseBuffer((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient, pDevice->wasapi.mappedBufferCaptureCap); + if (FAILED(hr)) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Data discontinuity recovery: IAudioCaptureClient_ReleaseBuffer() failed with %ld.\n", hr); + break; + } + + flags = 0; + hr = ma_IAudioCaptureClient_GetBuffer((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient, (BYTE**)&pDevice->wasapi.pMappedBufferCapture, &pDevice->wasapi.mappedBufferCaptureCap, &flags, NULL, NULL); + if (hr == MA_AUDCLNT_S_BUFFER_EMPTY || FAILED(hr)) { + /* + The buffer has been completely emptied or an error occurred. In this case we'll need + to reset the state of the mapped buffer which will trigger the next iteration to get + a fresh buffer from WASAPI. + */ + pDevice->wasapi.pMappedBufferCapture = NULL; + pDevice->wasapi.mappedBufferCaptureCap = 0; + pDevice->wasapi.mappedBufferCaptureLen = 0; + + if (hr == MA_AUDCLNT_S_BUFFER_EMPTY) { + if ((flags & MA_AUDCLNT_BUFFERFLAGS_DATA_DISCONTINUITY) != 0) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Data discontinuity recovery: Buffer emptied, and data discontinuity still reported.\n"); + } else { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Data discontinuity recovery: Buffer emptied.\n"); + } + } + + if (FAILED(hr)) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[WASAPI] Data discontinuity recovery: IAudioCaptureClient_GetBuffer() failed with %ld.\n", hr); + } + + break; + } + } + + /* If at this point we have a valid buffer mapped, make sure the buffer length is set appropriately. */ + if (pDevice->wasapi.pMappedBufferCapture != NULL) { + pDevice->wasapi.mappedBufferCaptureLen = pDevice->wasapi.mappedBufferCaptureCap; + } + } + } + + continue; + } else { + if (hr == MA_AUDCLNT_S_BUFFER_EMPTY || hr == MA_AUDCLNT_E_BUFFER_ERROR) { + /* + No data is available. We need to wait for more. There's two situations to consider + here. The first is normal capture mode. If this times out it probably means the + microphone isn't delivering data for whatever reason. In this case we'll just + abort the read and return whatever we were able to get. The other situations is + loopback mode, in which case a timeout probably just means the nothing is playing + through the speakers. + */ + + /* Experiment: Use a shorter timeout for loopback mode. */ + DWORD timeoutInMilliseconds = MA_WASAPI_WAIT_TIMEOUT_MILLISECONDS; + if (pDevice->type == ma_device_type_loopback) { + timeoutInMilliseconds = 10; + } + + if (WaitForSingleObject((HANDLE)pDevice->wasapi.hEventCapture, timeoutInMilliseconds) != WAIT_OBJECT_0) { + if (pDevice->type == ma_device_type_loopback) { + continue; /* Keep waiting in loopback mode. */ + } else { + result = MA_ERROR; + break; /* Wait failed. */ + } + } + + /* At this point we should be able to loop back to the start of the loop and try retrieving a data buffer again. */ + } else { + /* An error occurred and we need to abort. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to retrieve internal buffer from capture device in preparation for reading from the device. HRESULT = %d. Stopping device.\n", (int)hr); + result = ma_result_from_HRESULT(hr); + break; + } + } + } + } + + /* + If we were unable to process the entire requested frame count, but we still have a mapped buffer, + there's a good chance either an error occurred or the device was stopped mid-read. In this case + we'll need to make sure the buffer is released. + */ + if (totalFramesProcessed < frameCount && pDevice->wasapi.pMappedBufferCapture != NULL) { + ma_IAudioCaptureClient_ReleaseBuffer((ma_IAudioCaptureClient*)pDevice->wasapi.pCaptureClient, pDevice->wasapi.mappedBufferCaptureCap); + pDevice->wasapi.pMappedBufferCapture = NULL; + pDevice->wasapi.mappedBufferCaptureCap = 0; + pDevice->wasapi.mappedBufferCaptureLen = 0; + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesProcessed; + } + + return result; +} + +static ma_result ma_device_write__wasapi(ma_device* pDevice, const void* pFrames, ma_uint32 frameCount, ma_uint32* pFramesWritten) +{ + ma_result result = MA_SUCCESS; + ma_uint32 totalFramesProcessed = 0; + + /* Keep writing to the device until it's stopped or we've consumed all of our input. */ + while (ma_device_get_state(pDevice) == ma_device_state_started && totalFramesProcessed < frameCount) { + ma_uint32 framesRemaining = frameCount - totalFramesProcessed; + + /* + We're going to do this in a similar way to capture. We'll first check if the cached data pointer + is valid, and if so, read from that. Otherwise We will call IAudioRenderClient_GetBuffer() with + a requested buffer size equal to our actual period size. If it returns AUDCLNT_E_BUFFER_TOO_LARGE + it means we need to wait for some data to become available. + */ + if (pDevice->wasapi.pMappedBufferPlayback != NULL) { + /* We still have some space available in the mapped data buffer. Write to it. */ + ma_uint32 framesToProcessNow = framesRemaining; + if (framesToProcessNow > (pDevice->wasapi.mappedBufferPlaybackCap - pDevice->wasapi.mappedBufferPlaybackLen)) { + framesToProcessNow = (pDevice->wasapi.mappedBufferPlaybackCap - pDevice->wasapi.mappedBufferPlaybackLen); + } + + /* Now just copy the data over to the output buffer. */ + ma_copy_pcm_frames( + ma_offset_pcm_frames_ptr(pDevice->wasapi.pMappedBufferPlayback, pDevice->wasapi.mappedBufferPlaybackLen, pDevice->playback.internalFormat, pDevice->playback.internalChannels), + ma_offset_pcm_frames_const_ptr(pFrames, totalFramesProcessed, pDevice->playback.internalFormat, pDevice->playback.internalChannels), + framesToProcessNow, + pDevice->playback.internalFormat, pDevice->playback.internalChannels + ); + + totalFramesProcessed += framesToProcessNow; + pDevice->wasapi.mappedBufferPlaybackLen += framesToProcessNow; + + /* If the data buffer has been fully consumed we need to release it. */ + if (pDevice->wasapi.mappedBufferPlaybackLen == pDevice->wasapi.mappedBufferPlaybackCap) { + ma_IAudioRenderClient_ReleaseBuffer((ma_IAudioRenderClient*)pDevice->wasapi.pRenderClient, pDevice->wasapi.mappedBufferPlaybackCap, 0); + pDevice->wasapi.pMappedBufferPlayback = NULL; + pDevice->wasapi.mappedBufferPlaybackCap = 0; + pDevice->wasapi.mappedBufferPlaybackLen = 0; + + /* + In exclusive mode we need to wait here. Exclusive mode is weird because GetBuffer() never + seems to return AUDCLNT_E_BUFFER_TOO_LARGE, which is what we normally use to determine + whether or not we need to wait for more data. + */ + if (pDevice->playback.shareMode == ma_share_mode_exclusive) { + if (WaitForSingleObject((HANDLE)pDevice->wasapi.hEventPlayback, MA_WASAPI_WAIT_TIMEOUT_MILLISECONDS) != WAIT_OBJECT_0) { + result = MA_ERROR; + break; /* Wait failed. Probably timed out. */ + } + } + } + } else { + /* We don't have a mapped data buffer so we'll need to get one. */ + HRESULT hr; + ma_uint32 bufferSizeInFrames; + + /* Special rules for exclusive mode. */ + if (pDevice->playback.shareMode == ma_share_mode_exclusive) { + bufferSizeInFrames = pDevice->wasapi.actualBufferSizeInFramesPlayback; + } else { + bufferSizeInFrames = pDevice->wasapi.periodSizeInFramesPlayback; + } + + hr = ma_IAudioRenderClient_GetBuffer((ma_IAudioRenderClient*)pDevice->wasapi.pRenderClient, bufferSizeInFrames, (BYTE**)&pDevice->wasapi.pMappedBufferPlayback); + if (hr == S_OK) { + /* We have data available. */ + pDevice->wasapi.mappedBufferPlaybackCap = bufferSizeInFrames; + pDevice->wasapi.mappedBufferPlaybackLen = 0; + } else { + if (hr == MA_AUDCLNT_E_BUFFER_TOO_LARGE || hr == MA_AUDCLNT_E_BUFFER_ERROR) { + /* Not enough data available. We need to wait for more. */ + if (WaitForSingleObject((HANDLE)pDevice->wasapi.hEventPlayback, MA_WASAPI_WAIT_TIMEOUT_MILLISECONDS) != WAIT_OBJECT_0) { + result = MA_ERROR; + break; /* Wait failed. Probably timed out. */ + } + } else { + /* Some error occurred. We'll need to abort. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WASAPI] Failed to retrieve internal buffer from playback device in preparation for writing to the device. HRESULT = %d. Stopping device.\n", (int)hr); + result = ma_result_from_HRESULT(hr); + break; + } + } + } + } + + if (pFramesWritten != NULL) { + *pFramesWritten = totalFramesProcessed; + } + + return result; +} + +static ma_result ma_device_data_loop_wakeup__wasapi(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex || pDevice->type == ma_device_type_loopback) { + SetEvent((HANDLE)pDevice->wasapi.hEventCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + SetEvent((HANDLE)pDevice->wasapi.hEventPlayback); + } + + return MA_SUCCESS; +} + + +static ma_result ma_context_uninit__wasapi(ma_context* pContext) +{ + ma_context_command__wasapi cmd = ma_context_init_command__wasapi(MA_CONTEXT_COMMAND_QUIT__WASAPI); + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_wasapi); + + ma_context_post_command__wasapi(pContext, &cmd); + ma_thread_wait(&pContext->wasapi.commandThread); + + if (pContext->wasapi.hAvrt) { + ma_dlclose(ma_context_get_log(pContext), pContext->wasapi.hAvrt); + pContext->wasapi.hAvrt = NULL; + } + + #if defined(MA_WIN32_UWP) + { + if (pContext->wasapi.hMMDevapi) { + ma_dlclose(ma_context_get_log(pContext), pContext->wasapi.hMMDevapi); + pContext->wasapi.hMMDevapi = NULL; + } + } + #endif + + /* Only after the thread has been terminated can we uninitialize the sync objects for the command thread. */ + ma_semaphore_uninit(&pContext->wasapi.commandSem); + ma_mutex_uninit(&pContext->wasapi.commandLock); + + return MA_SUCCESS; +} + +static ma_result ma_context_init__wasapi(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + ma_result result = MA_SUCCESS; + + MA_ASSERT(pContext != NULL); + + (void)pConfig; + +#ifdef MA_WIN32_DESKTOP + /* + WASAPI is only supported in Vista SP1 and newer. The reason for SP1 and not the base version of Vista is that event-driven + exclusive mode does not work until SP1. + + Unfortunately older compilers don't define these functions so we need to dynamically load them in order to avoid a link error. + */ + { + ma_OSVERSIONINFOEXW osvi; + ma_handle kernel32DLL; + ma_PFNVerifyVersionInfoW _VerifyVersionInfoW; + ma_PFNVerSetConditionMask _VerSetConditionMask; + + kernel32DLL = ma_dlopen(ma_context_get_log(pContext), "kernel32.dll"); + if (kernel32DLL == NULL) { + return MA_NO_BACKEND; + } + + _VerifyVersionInfoW = (ma_PFNVerifyVersionInfoW )ma_dlsym(ma_context_get_log(pContext), kernel32DLL, "VerifyVersionInfoW"); + _VerSetConditionMask = (ma_PFNVerSetConditionMask)ma_dlsym(ma_context_get_log(pContext), kernel32DLL, "VerSetConditionMask"); + if (_VerifyVersionInfoW == NULL || _VerSetConditionMask == NULL) { + ma_dlclose(ma_context_get_log(pContext), kernel32DLL); + return MA_NO_BACKEND; + } + + MA_ZERO_OBJECT(&osvi); + osvi.dwOSVersionInfoSize = sizeof(osvi); + osvi.dwMajorVersion = ((MA_WIN32_WINNT_VISTA >> 8) & 0xFF); + osvi.dwMinorVersion = ((MA_WIN32_WINNT_VISTA >> 0) & 0xFF); + osvi.wServicePackMajor = 1; + if (_VerifyVersionInfoW(&osvi, MA_VER_MAJORVERSION | MA_VER_MINORVERSION | MA_VER_SERVICEPACKMAJOR, _VerSetConditionMask(_VerSetConditionMask(_VerSetConditionMask(0, MA_VER_MAJORVERSION, MA_VER_GREATER_EQUAL), MA_VER_MINORVERSION, MA_VER_GREATER_EQUAL), MA_VER_SERVICEPACKMAJOR, MA_VER_GREATER_EQUAL))) { + result = MA_SUCCESS; + } else { + result = MA_NO_BACKEND; + } + + ma_dlclose(ma_context_get_log(pContext), kernel32DLL); + } +#endif + + if (result != MA_SUCCESS) { + return result; + } + + MA_ZERO_OBJECT(&pContext->wasapi); + + + #if defined(MA_WIN32_UWP) + { + /* Link to mmdevapi so we can get access to ActivateAudioInterfaceAsync(). */ + pContext->wasapi.hMMDevapi = ma_dlopen(ma_context_get_log(pContext), "mmdevapi.dll"); + if (pContext->wasapi.hMMDevapi) { + pContext->wasapi.ActivateAudioInterfaceAsync = ma_dlsym(ma_context_get_log(pContext), pContext->wasapi.hMMDevapi, "ActivateAudioInterfaceAsync"); + if (pContext->wasapi.ActivateAudioInterfaceAsync == NULL) { + ma_dlclose(ma_context_get_log(pContext), pContext->wasapi.hMMDevapi); + return MA_NO_BACKEND; /* ActivateAudioInterfaceAsync() could not be loaded. */ + } + } else { + return MA_NO_BACKEND; /* Failed to load mmdevapi.dll which is required for ActivateAudioInterfaceAsync() */ + } + } + #endif + + /* Optionally use the Avrt API to specify the audio thread's latency sensitivity requirements */ + pContext->wasapi.hAvrt = ma_dlopen(ma_context_get_log(pContext), "avrt.dll"); + if (pContext->wasapi.hAvrt) { + pContext->wasapi.AvSetMmThreadCharacteristicsA = ma_dlsym(ma_context_get_log(pContext), pContext->wasapi.hAvrt, "AvSetMmThreadCharacteristicsA"); + pContext->wasapi.AvRevertMmThreadcharacteristics = ma_dlsym(ma_context_get_log(pContext), pContext->wasapi.hAvrt, "AvRevertMmThreadCharacteristics"); + + /* If either function could not be found, disable use of avrt entirely. */ + if (!pContext->wasapi.AvSetMmThreadCharacteristicsA || !pContext->wasapi.AvRevertMmThreadcharacteristics) { + pContext->wasapi.AvSetMmThreadCharacteristicsA = NULL; + pContext->wasapi.AvRevertMmThreadcharacteristics = NULL; + ma_dlclose(ma_context_get_log(pContext), pContext->wasapi.hAvrt); + pContext->wasapi.hAvrt = NULL; + } + } + + + /* + Annoyingly, WASAPI does not allow you to release an IAudioClient object from a different thread + than the one that retrieved it with GetService(). This can result in a deadlock in two + situations: + + 1) When calling ma_device_uninit() from a different thread to ma_device_init(); and + 2) When uninitializing and reinitializing the internal IAudioClient object in response to + automatic stream routing. + + We could define ma_device_uninit() such that it must be called on the same thread as + ma_device_init(). We could also just not release the IAudioClient when performing automatic + stream routing to avoid the deadlock. Neither of these are acceptable solutions in my view so + we're going to have to work around this with a worker thread. This is not ideal, but I can't + think of a better way to do this. + + More information about this can be found here: + + https://docs.microsoft.com/en-us/windows/win32/api/audioclient/nn-audioclient-iaudiorenderclient + + Note this section: + + When releasing an IAudioRenderClient interface instance, the client must call the interface's + Release method from the same thread as the call to IAudioClient::GetService that created the + object. + */ + { + result = ma_mutex_init(&pContext->wasapi.commandLock); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_semaphore_init(0, &pContext->wasapi.commandSem); + if (result != MA_SUCCESS) { + ma_mutex_uninit(&pContext->wasapi.commandLock); + return result; + } + + result = ma_thread_create(&pContext->wasapi.commandThread, ma_thread_priority_normal, 0, ma_context_command_thread__wasapi, pContext, &pContext->allocationCallbacks); + if (result != MA_SUCCESS) { + ma_semaphore_uninit(&pContext->wasapi.commandSem); + ma_mutex_uninit(&pContext->wasapi.commandLock); + return result; + } + } + + + pCallbacks->onContextInit = ma_context_init__wasapi; + pCallbacks->onContextUninit = ma_context_uninit__wasapi; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__wasapi; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__wasapi; + pCallbacks->onDeviceInit = ma_device_init__wasapi; + pCallbacks->onDeviceUninit = ma_device_uninit__wasapi; + pCallbacks->onDeviceStart = ma_device_start__wasapi; + pCallbacks->onDeviceStop = ma_device_stop__wasapi; + pCallbacks->onDeviceRead = ma_device_read__wasapi; + pCallbacks->onDeviceWrite = ma_device_write__wasapi; + pCallbacks->onDeviceDataLoop = NULL; + pCallbacks->onDeviceDataLoopWakeup = ma_device_data_loop_wakeup__wasapi; + + return MA_SUCCESS; +} +#endif + +/****************************************************************************** + +DirectSound Backend + +******************************************************************************/ +#ifdef MA_HAS_DSOUND +/*#include */ + +/*static const GUID MA_GUID_IID_DirectSoundNotify = {0xb0210783, 0x89cd, 0x11d0, {0xaf, 0x08, 0x00, 0xa0, 0xc9, 0x25, 0xcd, 0x16}};*/ + +/* miniaudio only uses priority or exclusive modes. */ +#define MA_DSSCL_NORMAL 1 +#define MA_DSSCL_PRIORITY 2 +#define MA_DSSCL_EXCLUSIVE 3 +#define MA_DSSCL_WRITEPRIMARY 4 + +#define MA_DSCAPS_PRIMARYMONO 0x00000001 +#define MA_DSCAPS_PRIMARYSTEREO 0x00000002 +#define MA_DSCAPS_PRIMARY8BIT 0x00000004 +#define MA_DSCAPS_PRIMARY16BIT 0x00000008 +#define MA_DSCAPS_CONTINUOUSRATE 0x00000010 +#define MA_DSCAPS_EMULDRIVER 0x00000020 +#define MA_DSCAPS_CERTIFIED 0x00000040 +#define MA_DSCAPS_SECONDARYMONO 0x00000100 +#define MA_DSCAPS_SECONDARYSTEREO 0x00000200 +#define MA_DSCAPS_SECONDARY8BIT 0x00000400 +#define MA_DSCAPS_SECONDARY16BIT 0x00000800 + +#define MA_DSBCAPS_PRIMARYBUFFER 0x00000001 +#define MA_DSBCAPS_STATIC 0x00000002 +#define MA_DSBCAPS_LOCHARDWARE 0x00000004 +#define MA_DSBCAPS_LOCSOFTWARE 0x00000008 +#define MA_DSBCAPS_CTRL3D 0x00000010 +#define MA_DSBCAPS_CTRLFREQUENCY 0x00000020 +#define MA_DSBCAPS_CTRLPAN 0x00000040 +#define MA_DSBCAPS_CTRLVOLUME 0x00000080 +#define MA_DSBCAPS_CTRLPOSITIONNOTIFY 0x00000100 +#define MA_DSBCAPS_CTRLFX 0x00000200 +#define MA_DSBCAPS_STICKYFOCUS 0x00004000 +#define MA_DSBCAPS_GLOBALFOCUS 0x00008000 +#define MA_DSBCAPS_GETCURRENTPOSITION2 0x00010000 +#define MA_DSBCAPS_MUTE3DATMAXDISTANCE 0x00020000 +#define MA_DSBCAPS_LOCDEFER 0x00040000 +#define MA_DSBCAPS_TRUEPLAYPOSITION 0x00080000 + +#define MA_DSBPLAY_LOOPING 0x00000001 +#define MA_DSBPLAY_LOCHARDWARE 0x00000002 +#define MA_DSBPLAY_LOCSOFTWARE 0x00000004 +#define MA_DSBPLAY_TERMINATEBY_TIME 0x00000008 +#define MA_DSBPLAY_TERMINATEBY_DISTANCE 0x00000010 +#define MA_DSBPLAY_TERMINATEBY_PRIORITY 0x00000020 + +#define MA_DSBSTATUS_PLAYING 0x00000001 +#define MA_DSBSTATUS_BUFFERLOST 0x00000002 +#define MA_DSBSTATUS_LOOPING 0x00000004 +#define MA_DSBSTATUS_LOCHARDWARE 0x00000008 +#define MA_DSBSTATUS_LOCSOFTWARE 0x00000010 +#define MA_DSBSTATUS_TERMINATED 0x00000020 + +#define MA_DSCBSTART_LOOPING 0x00000001 + +typedef struct +{ + DWORD dwSize; + DWORD dwFlags; + DWORD dwBufferBytes; + DWORD dwReserved; + MA_WAVEFORMATEX* lpwfxFormat; + GUID guid3DAlgorithm; +} MA_DSBUFFERDESC; + +typedef struct +{ + DWORD dwSize; + DWORD dwFlags; + DWORD dwBufferBytes; + DWORD dwReserved; + MA_WAVEFORMATEX* lpwfxFormat; + DWORD dwFXCount; + void* lpDSCFXDesc; /* <-- miniaudio doesn't use this, so set to void*. */ +} MA_DSCBUFFERDESC; + +typedef struct +{ + DWORD dwSize; + DWORD dwFlags; + DWORD dwMinSecondarySampleRate; + DWORD dwMaxSecondarySampleRate; + DWORD dwPrimaryBuffers; + DWORD dwMaxHwMixingAllBuffers; + DWORD dwMaxHwMixingStaticBuffers; + DWORD dwMaxHwMixingStreamingBuffers; + DWORD dwFreeHwMixingAllBuffers; + DWORD dwFreeHwMixingStaticBuffers; + DWORD dwFreeHwMixingStreamingBuffers; + DWORD dwMaxHw3DAllBuffers; + DWORD dwMaxHw3DStaticBuffers; + DWORD dwMaxHw3DStreamingBuffers; + DWORD dwFreeHw3DAllBuffers; + DWORD dwFreeHw3DStaticBuffers; + DWORD dwFreeHw3DStreamingBuffers; + DWORD dwTotalHwMemBytes; + DWORD dwFreeHwMemBytes; + DWORD dwMaxContigFreeHwMemBytes; + DWORD dwUnlockTransferRateHwBuffers; + DWORD dwPlayCpuOverheadSwBuffers; + DWORD dwReserved1; + DWORD dwReserved2; +} MA_DSCAPS; + +typedef struct +{ + DWORD dwSize; + DWORD dwFlags; + DWORD dwBufferBytes; + DWORD dwUnlockTransferRate; + DWORD dwPlayCpuOverhead; +} MA_DSBCAPS; + +typedef struct +{ + DWORD dwSize; + DWORD dwFlags; + DWORD dwFormats; + DWORD dwChannels; +} MA_DSCCAPS; + +typedef struct +{ + DWORD dwSize; + DWORD dwFlags; + DWORD dwBufferBytes; + DWORD dwReserved; +} MA_DSCBCAPS; + +typedef struct +{ + DWORD dwOffset; + HANDLE hEventNotify; +} MA_DSBPOSITIONNOTIFY; + +typedef struct ma_IDirectSound ma_IDirectSound; +typedef struct ma_IDirectSoundBuffer ma_IDirectSoundBuffer; +typedef struct ma_IDirectSoundCapture ma_IDirectSoundCapture; +typedef struct ma_IDirectSoundCaptureBuffer ma_IDirectSoundCaptureBuffer; +typedef struct ma_IDirectSoundNotify ma_IDirectSoundNotify; + + +/* +COM objects. The way these work is that you have a vtable (a list of function pointers, kind of +like how C++ works internally), and then you have a structure with a single member, which is a +pointer to the vtable. The vtable is where the methods of the object are defined. Methods need +to be in a specific order, and parent classes need to have their methods declared first. +*/ + +/* IDirectSound */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IDirectSound* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IDirectSound* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IDirectSound* pThis); + + /* IDirectSound */ + HRESULT (STDMETHODCALLTYPE * CreateSoundBuffer) (ma_IDirectSound* pThis, const MA_DSBUFFERDESC* pDSBufferDesc, ma_IDirectSoundBuffer** ppDSBuffer, void* pUnkOuter); + HRESULT (STDMETHODCALLTYPE * GetCaps) (ma_IDirectSound* pThis, MA_DSCAPS* pDSCaps); + HRESULT (STDMETHODCALLTYPE * DuplicateSoundBuffer)(ma_IDirectSound* pThis, ma_IDirectSoundBuffer* pDSBufferOriginal, ma_IDirectSoundBuffer** ppDSBufferDuplicate); + HRESULT (STDMETHODCALLTYPE * SetCooperativeLevel) (ma_IDirectSound* pThis, HWND hwnd, DWORD dwLevel); + HRESULT (STDMETHODCALLTYPE * Compact) (ma_IDirectSound* pThis); + HRESULT (STDMETHODCALLTYPE * GetSpeakerConfig) (ma_IDirectSound* pThis, DWORD* pSpeakerConfig); + HRESULT (STDMETHODCALLTYPE * SetSpeakerConfig) (ma_IDirectSound* pThis, DWORD dwSpeakerConfig); + HRESULT (STDMETHODCALLTYPE * Initialize) (ma_IDirectSound* pThis, const GUID* pGuidDevice); +} ma_IDirectSoundVtbl; +struct ma_IDirectSound +{ + ma_IDirectSoundVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IDirectSound_QueryInterface(ma_IDirectSound* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IDirectSound_AddRef(ma_IDirectSound* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IDirectSound_Release(ma_IDirectSound* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IDirectSound_CreateSoundBuffer(ma_IDirectSound* pThis, const MA_DSBUFFERDESC* pDSBufferDesc, ma_IDirectSoundBuffer** ppDSBuffer, void* pUnkOuter) { return pThis->lpVtbl->CreateSoundBuffer(pThis, pDSBufferDesc, ppDSBuffer, pUnkOuter); } +static MA_INLINE HRESULT ma_IDirectSound_GetCaps(ma_IDirectSound* pThis, MA_DSCAPS* pDSCaps) { return pThis->lpVtbl->GetCaps(pThis, pDSCaps); } +static MA_INLINE HRESULT ma_IDirectSound_DuplicateSoundBuffer(ma_IDirectSound* pThis, ma_IDirectSoundBuffer* pDSBufferOriginal, ma_IDirectSoundBuffer** ppDSBufferDuplicate) { return pThis->lpVtbl->DuplicateSoundBuffer(pThis, pDSBufferOriginal, ppDSBufferDuplicate); } +static MA_INLINE HRESULT ma_IDirectSound_SetCooperativeLevel(ma_IDirectSound* pThis, HWND hwnd, DWORD dwLevel) { return pThis->lpVtbl->SetCooperativeLevel(pThis, hwnd, dwLevel); } +static MA_INLINE HRESULT ma_IDirectSound_Compact(ma_IDirectSound* pThis) { return pThis->lpVtbl->Compact(pThis); } +static MA_INLINE HRESULT ma_IDirectSound_GetSpeakerConfig(ma_IDirectSound* pThis, DWORD* pSpeakerConfig) { return pThis->lpVtbl->GetSpeakerConfig(pThis, pSpeakerConfig); } +static MA_INLINE HRESULT ma_IDirectSound_SetSpeakerConfig(ma_IDirectSound* pThis, DWORD dwSpeakerConfig) { return pThis->lpVtbl->SetSpeakerConfig(pThis, dwSpeakerConfig); } +static MA_INLINE HRESULT ma_IDirectSound_Initialize(ma_IDirectSound* pThis, const GUID* pGuidDevice) { return pThis->lpVtbl->Initialize(pThis, pGuidDevice); } + + +/* IDirectSoundBuffer */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IDirectSoundBuffer* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IDirectSoundBuffer* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IDirectSoundBuffer* pThis); + + /* IDirectSoundBuffer */ + HRESULT (STDMETHODCALLTYPE * GetCaps) (ma_IDirectSoundBuffer* pThis, MA_DSBCAPS* pDSBufferCaps); + HRESULT (STDMETHODCALLTYPE * GetCurrentPosition)(ma_IDirectSoundBuffer* pThis, DWORD* pCurrentPlayCursor, DWORD* pCurrentWriteCursor); + HRESULT (STDMETHODCALLTYPE * GetFormat) (ma_IDirectSoundBuffer* pThis, MA_WAVEFORMATEX* pFormat, DWORD dwSizeAllocated, DWORD* pSizeWritten); + HRESULT (STDMETHODCALLTYPE * GetVolume) (ma_IDirectSoundBuffer* pThis, LONG* pVolume); + HRESULT (STDMETHODCALLTYPE * GetPan) (ma_IDirectSoundBuffer* pThis, LONG* pPan); + HRESULT (STDMETHODCALLTYPE * GetFrequency) (ma_IDirectSoundBuffer* pThis, DWORD* pFrequency); + HRESULT (STDMETHODCALLTYPE * GetStatus) (ma_IDirectSoundBuffer* pThis, DWORD* pStatus); + HRESULT (STDMETHODCALLTYPE * Initialize) (ma_IDirectSoundBuffer* pThis, ma_IDirectSound* pDirectSound, const MA_DSBUFFERDESC* pDSBufferDesc); + HRESULT (STDMETHODCALLTYPE * Lock) (ma_IDirectSoundBuffer* pThis, DWORD dwOffset, DWORD dwBytes, void** ppAudioPtr1, DWORD* pAudioBytes1, void** ppAudioPtr2, DWORD* pAudioBytes2, DWORD dwFlags); + HRESULT (STDMETHODCALLTYPE * Play) (ma_IDirectSoundBuffer* pThis, DWORD dwReserved1, DWORD dwPriority, DWORD dwFlags); + HRESULT (STDMETHODCALLTYPE * SetCurrentPosition)(ma_IDirectSoundBuffer* pThis, DWORD dwNewPosition); + HRESULT (STDMETHODCALLTYPE * SetFormat) (ma_IDirectSoundBuffer* pThis, const MA_WAVEFORMATEX* pFormat); + HRESULT (STDMETHODCALLTYPE * SetVolume) (ma_IDirectSoundBuffer* pThis, LONG volume); + HRESULT (STDMETHODCALLTYPE * SetPan) (ma_IDirectSoundBuffer* pThis, LONG pan); + HRESULT (STDMETHODCALLTYPE * SetFrequency) (ma_IDirectSoundBuffer* pThis, DWORD dwFrequency); + HRESULT (STDMETHODCALLTYPE * Stop) (ma_IDirectSoundBuffer* pThis); + HRESULT (STDMETHODCALLTYPE * Unlock) (ma_IDirectSoundBuffer* pThis, void* pAudioPtr1, DWORD dwAudioBytes1, void* pAudioPtr2, DWORD dwAudioBytes2); + HRESULT (STDMETHODCALLTYPE * Restore) (ma_IDirectSoundBuffer* pThis); +} ma_IDirectSoundBufferVtbl; +struct ma_IDirectSoundBuffer +{ + ma_IDirectSoundBufferVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IDirectSoundBuffer_QueryInterface(ma_IDirectSoundBuffer* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IDirectSoundBuffer_AddRef(ma_IDirectSoundBuffer* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IDirectSoundBuffer_Release(ma_IDirectSoundBuffer* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_GetCaps(ma_IDirectSoundBuffer* pThis, MA_DSBCAPS* pDSBufferCaps) { return pThis->lpVtbl->GetCaps(pThis, pDSBufferCaps); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_GetCurrentPosition(ma_IDirectSoundBuffer* pThis, DWORD* pCurrentPlayCursor, DWORD* pCurrentWriteCursor) { return pThis->lpVtbl->GetCurrentPosition(pThis, pCurrentPlayCursor, pCurrentWriteCursor); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_GetFormat(ma_IDirectSoundBuffer* pThis, MA_WAVEFORMATEX* pFormat, DWORD dwSizeAllocated, DWORD* pSizeWritten) { return pThis->lpVtbl->GetFormat(pThis, pFormat, dwSizeAllocated, pSizeWritten); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_GetVolume(ma_IDirectSoundBuffer* pThis, LONG* pVolume) { return pThis->lpVtbl->GetVolume(pThis, pVolume); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_GetPan(ma_IDirectSoundBuffer* pThis, LONG* pPan) { return pThis->lpVtbl->GetPan(pThis, pPan); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_GetFrequency(ma_IDirectSoundBuffer* pThis, DWORD* pFrequency) { return pThis->lpVtbl->GetFrequency(pThis, pFrequency); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_GetStatus(ma_IDirectSoundBuffer* pThis, DWORD* pStatus) { return pThis->lpVtbl->GetStatus(pThis, pStatus); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_Initialize(ma_IDirectSoundBuffer* pThis, ma_IDirectSound* pDirectSound, const MA_DSBUFFERDESC* pDSBufferDesc) { return pThis->lpVtbl->Initialize(pThis, pDirectSound, pDSBufferDesc); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_Lock(ma_IDirectSoundBuffer* pThis, DWORD dwOffset, DWORD dwBytes, void** ppAudioPtr1, DWORD* pAudioBytes1, void** ppAudioPtr2, DWORD* pAudioBytes2, DWORD dwFlags) { return pThis->lpVtbl->Lock(pThis, dwOffset, dwBytes, ppAudioPtr1, pAudioBytes1, ppAudioPtr2, pAudioBytes2, dwFlags); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_Play(ma_IDirectSoundBuffer* pThis, DWORD dwReserved1, DWORD dwPriority, DWORD dwFlags) { return pThis->lpVtbl->Play(pThis, dwReserved1, dwPriority, dwFlags); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_SetCurrentPosition(ma_IDirectSoundBuffer* pThis, DWORD dwNewPosition) { return pThis->lpVtbl->SetCurrentPosition(pThis, dwNewPosition); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_SetFormat(ma_IDirectSoundBuffer* pThis, const MA_WAVEFORMATEX* pFormat) { return pThis->lpVtbl->SetFormat(pThis, pFormat); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_SetVolume(ma_IDirectSoundBuffer* pThis, LONG volume) { return pThis->lpVtbl->SetVolume(pThis, volume); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_SetPan(ma_IDirectSoundBuffer* pThis, LONG pan) { return pThis->lpVtbl->SetPan(pThis, pan); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_SetFrequency(ma_IDirectSoundBuffer* pThis, DWORD dwFrequency) { return pThis->lpVtbl->SetFrequency(pThis, dwFrequency); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_Stop(ma_IDirectSoundBuffer* pThis) { return pThis->lpVtbl->Stop(pThis); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_Unlock(ma_IDirectSoundBuffer* pThis, void* pAudioPtr1, DWORD dwAudioBytes1, void* pAudioPtr2, DWORD dwAudioBytes2) { return pThis->lpVtbl->Unlock(pThis, pAudioPtr1, dwAudioBytes1, pAudioPtr2, dwAudioBytes2); } +static MA_INLINE HRESULT ma_IDirectSoundBuffer_Restore(ma_IDirectSoundBuffer* pThis) { return pThis->lpVtbl->Restore(pThis); } + + +/* IDirectSoundCapture */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IDirectSoundCapture* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IDirectSoundCapture* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IDirectSoundCapture* pThis); + + /* IDirectSoundCapture */ + HRESULT (STDMETHODCALLTYPE * CreateCaptureBuffer)(ma_IDirectSoundCapture* pThis, const MA_DSCBUFFERDESC* pDSCBufferDesc, ma_IDirectSoundCaptureBuffer** ppDSCBuffer, void* pUnkOuter); + HRESULT (STDMETHODCALLTYPE * GetCaps) (ma_IDirectSoundCapture* pThis, MA_DSCCAPS* pDSCCaps); + HRESULT (STDMETHODCALLTYPE * Initialize) (ma_IDirectSoundCapture* pThis, const GUID* pGuidDevice); +} ma_IDirectSoundCaptureVtbl; +struct ma_IDirectSoundCapture +{ + ma_IDirectSoundCaptureVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IDirectSoundCapture_QueryInterface (ma_IDirectSoundCapture* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IDirectSoundCapture_AddRef (ma_IDirectSoundCapture* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IDirectSoundCapture_Release (ma_IDirectSoundCapture* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IDirectSoundCapture_CreateCaptureBuffer(ma_IDirectSoundCapture* pThis, const MA_DSCBUFFERDESC* pDSCBufferDesc, ma_IDirectSoundCaptureBuffer** ppDSCBuffer, void* pUnkOuter) { return pThis->lpVtbl->CreateCaptureBuffer(pThis, pDSCBufferDesc, ppDSCBuffer, pUnkOuter); } +static MA_INLINE HRESULT ma_IDirectSoundCapture_GetCaps (ma_IDirectSoundCapture* pThis, MA_DSCCAPS* pDSCCaps) { return pThis->lpVtbl->GetCaps(pThis, pDSCCaps); } +static MA_INLINE HRESULT ma_IDirectSoundCapture_Initialize (ma_IDirectSoundCapture* pThis, const GUID* pGuidDevice) { return pThis->lpVtbl->Initialize(pThis, pGuidDevice); } + + +/* IDirectSoundCaptureBuffer */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IDirectSoundCaptureBuffer* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IDirectSoundCaptureBuffer* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IDirectSoundCaptureBuffer* pThis); + + /* IDirectSoundCaptureBuffer */ + HRESULT (STDMETHODCALLTYPE * GetCaps) (ma_IDirectSoundCaptureBuffer* pThis, MA_DSCBCAPS* pDSCBCaps); + HRESULT (STDMETHODCALLTYPE * GetCurrentPosition)(ma_IDirectSoundCaptureBuffer* pThis, DWORD* pCapturePosition, DWORD* pReadPosition); + HRESULT (STDMETHODCALLTYPE * GetFormat) (ma_IDirectSoundCaptureBuffer* pThis, MA_WAVEFORMATEX* pFormat, DWORD dwSizeAllocated, DWORD* pSizeWritten); + HRESULT (STDMETHODCALLTYPE * GetStatus) (ma_IDirectSoundCaptureBuffer* pThis, DWORD* pStatus); + HRESULT (STDMETHODCALLTYPE * Initialize) (ma_IDirectSoundCaptureBuffer* pThis, ma_IDirectSoundCapture* pDirectSoundCapture, const MA_DSCBUFFERDESC* pDSCBufferDesc); + HRESULT (STDMETHODCALLTYPE * Lock) (ma_IDirectSoundCaptureBuffer* pThis, DWORD dwOffset, DWORD dwBytes, void** ppAudioPtr1, DWORD* pAudioBytes1, void** ppAudioPtr2, DWORD* pAudioBytes2, DWORD dwFlags); + HRESULT (STDMETHODCALLTYPE * Start) (ma_IDirectSoundCaptureBuffer* pThis, DWORD dwFlags); + HRESULT (STDMETHODCALLTYPE * Stop) (ma_IDirectSoundCaptureBuffer* pThis); + HRESULT (STDMETHODCALLTYPE * Unlock) (ma_IDirectSoundCaptureBuffer* pThis, void* pAudioPtr1, DWORD dwAudioBytes1, void* pAudioPtr2, DWORD dwAudioBytes2); +} ma_IDirectSoundCaptureBufferVtbl; +struct ma_IDirectSoundCaptureBuffer +{ + ma_IDirectSoundCaptureBufferVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_QueryInterface(ma_IDirectSoundCaptureBuffer* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IDirectSoundCaptureBuffer_AddRef(ma_IDirectSoundCaptureBuffer* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IDirectSoundCaptureBuffer_Release(ma_IDirectSoundCaptureBuffer* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_GetCaps(ma_IDirectSoundCaptureBuffer* pThis, MA_DSCBCAPS* pDSCBCaps) { return pThis->lpVtbl->GetCaps(pThis, pDSCBCaps); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_GetCurrentPosition(ma_IDirectSoundCaptureBuffer* pThis, DWORD* pCapturePosition, DWORD* pReadPosition) { return pThis->lpVtbl->GetCurrentPosition(pThis, pCapturePosition, pReadPosition); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_GetFormat(ma_IDirectSoundCaptureBuffer* pThis, MA_WAVEFORMATEX* pFormat, DWORD dwSizeAllocated, DWORD* pSizeWritten) { return pThis->lpVtbl->GetFormat(pThis, pFormat, dwSizeAllocated, pSizeWritten); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_GetStatus(ma_IDirectSoundCaptureBuffer* pThis, DWORD* pStatus) { return pThis->lpVtbl->GetStatus(pThis, pStatus); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_Initialize(ma_IDirectSoundCaptureBuffer* pThis, ma_IDirectSoundCapture* pDirectSoundCapture, const MA_DSCBUFFERDESC* pDSCBufferDesc) { return pThis->lpVtbl->Initialize(pThis, pDirectSoundCapture, pDSCBufferDesc); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_Lock(ma_IDirectSoundCaptureBuffer* pThis, DWORD dwOffset, DWORD dwBytes, void** ppAudioPtr1, DWORD* pAudioBytes1, void** ppAudioPtr2, DWORD* pAudioBytes2, DWORD dwFlags) { return pThis->lpVtbl->Lock(pThis, dwOffset, dwBytes, ppAudioPtr1, pAudioBytes1, ppAudioPtr2, pAudioBytes2, dwFlags); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_Start(ma_IDirectSoundCaptureBuffer* pThis, DWORD dwFlags) { return pThis->lpVtbl->Start(pThis, dwFlags); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_Stop(ma_IDirectSoundCaptureBuffer* pThis) { return pThis->lpVtbl->Stop(pThis); } +static MA_INLINE HRESULT ma_IDirectSoundCaptureBuffer_Unlock(ma_IDirectSoundCaptureBuffer* pThis, void* pAudioPtr1, DWORD dwAudioBytes1, void* pAudioPtr2, DWORD dwAudioBytes2) { return pThis->lpVtbl->Unlock(pThis, pAudioPtr1, dwAudioBytes1, pAudioPtr2, dwAudioBytes2); } + + +/* IDirectSoundNotify */ +typedef struct +{ + /* IUnknown */ + HRESULT (STDMETHODCALLTYPE * QueryInterface)(ma_IDirectSoundNotify* pThis, const IID* const riid, void** ppObject); + ULONG (STDMETHODCALLTYPE * AddRef) (ma_IDirectSoundNotify* pThis); + ULONG (STDMETHODCALLTYPE * Release) (ma_IDirectSoundNotify* pThis); + + /* IDirectSoundNotify */ + HRESULT (STDMETHODCALLTYPE * SetNotificationPositions)(ma_IDirectSoundNotify* pThis, DWORD dwPositionNotifies, const MA_DSBPOSITIONNOTIFY* pPositionNotifies); +} ma_IDirectSoundNotifyVtbl; +struct ma_IDirectSoundNotify +{ + ma_IDirectSoundNotifyVtbl* lpVtbl; +}; +static MA_INLINE HRESULT ma_IDirectSoundNotify_QueryInterface(ma_IDirectSoundNotify* pThis, const IID* const riid, void** ppObject) { return pThis->lpVtbl->QueryInterface(pThis, riid, ppObject); } +static MA_INLINE ULONG ma_IDirectSoundNotify_AddRef(ma_IDirectSoundNotify* pThis) { return pThis->lpVtbl->AddRef(pThis); } +static MA_INLINE ULONG ma_IDirectSoundNotify_Release(ma_IDirectSoundNotify* pThis) { return pThis->lpVtbl->Release(pThis); } +static MA_INLINE HRESULT ma_IDirectSoundNotify_SetNotificationPositions(ma_IDirectSoundNotify* pThis, DWORD dwPositionNotifies, const MA_DSBPOSITIONNOTIFY* pPositionNotifies) { return pThis->lpVtbl->SetNotificationPositions(pThis, dwPositionNotifies, pPositionNotifies); } + + +typedef BOOL (CALLBACK * ma_DSEnumCallbackAProc) (GUID* pDeviceGUID, const char* pDeviceDescription, const char* pModule, void* pContext); +typedef HRESULT (WINAPI * ma_DirectSoundCreateProc) (const GUID* pcGuidDevice, ma_IDirectSound** ppDS8, ma_IUnknown* pUnkOuter); +typedef HRESULT (WINAPI * ma_DirectSoundEnumerateAProc) (ma_DSEnumCallbackAProc pDSEnumCallback, void* pContext); +typedef HRESULT (WINAPI * ma_DirectSoundCaptureCreateProc) (const GUID* pcGuidDevice, ma_IDirectSoundCapture** ppDSC8, ma_IUnknown* pUnkOuter); +typedef HRESULT (WINAPI * ma_DirectSoundCaptureEnumerateAProc)(ma_DSEnumCallbackAProc pDSEnumCallback, void* pContext); + +static ma_uint32 ma_get_best_sample_rate_within_range(ma_uint32 sampleRateMin, ma_uint32 sampleRateMax) +{ + /* Normalize the range in case we were given something stupid. */ + if (sampleRateMin < (ma_uint32)ma_standard_sample_rate_min) { + sampleRateMin = (ma_uint32)ma_standard_sample_rate_min; + } + if (sampleRateMax > (ma_uint32)ma_standard_sample_rate_max) { + sampleRateMax = (ma_uint32)ma_standard_sample_rate_max; + } + if (sampleRateMin > sampleRateMax) { + sampleRateMin = sampleRateMax; + } + + if (sampleRateMin == sampleRateMax) { + return sampleRateMax; + } else { + size_t iStandardRate; + for (iStandardRate = 0; iStandardRate < ma_countof(g_maStandardSampleRatePriorities); ++iStandardRate) { + ma_uint32 standardRate = g_maStandardSampleRatePriorities[iStandardRate]; + if (standardRate >= sampleRateMin && standardRate <= sampleRateMax) { + return standardRate; + } + } + } + + /* Should never get here. */ + MA_ASSERT(MA_FALSE); + return 0; +} + +/* +Retrieves the channel count and channel map for the given speaker configuration. If the speaker configuration is unknown, +the channel count and channel map will be left unmodified. +*/ +static void ma_get_channels_from_speaker_config__dsound(DWORD speakerConfig, WORD* pChannelsOut, DWORD* pChannelMapOut) +{ + WORD channels; + DWORD channelMap; + + channels = 0; + if (pChannelsOut != NULL) { + channels = *pChannelsOut; + } + + channelMap = 0; + if (pChannelMapOut != NULL) { + channelMap = *pChannelMapOut; + } + + /* + The speaker configuration is a combination of speaker config and speaker geometry. The lower 8 bits is what we care about. The upper + 16 bits is for the geometry. + */ + switch ((BYTE)(speakerConfig)) { + case 1 /*DSSPEAKER_HEADPHONE*/: channels = 2; channelMap = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT; break; + case 2 /*DSSPEAKER_MONO*/: channels = 1; channelMap = SPEAKER_FRONT_CENTER; break; + case 3 /*DSSPEAKER_QUAD*/: channels = 4; channelMap = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT; break; + case 4 /*DSSPEAKER_STEREO*/: channels = 2; channelMap = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT; break; + case 5 /*DSSPEAKER_SURROUND*/: channels = 4; channelMap = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_BACK_CENTER; break; + case 6 /*DSSPEAKER_5POINT1_BACK*/ /*DSSPEAKER_5POINT1*/: channels = 6; channelMap = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_LOW_FREQUENCY | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT; break; + case 7 /*DSSPEAKER_7POINT1_WIDE*/ /*DSSPEAKER_7POINT1*/: channels = 8; channelMap = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_LOW_FREQUENCY | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT | SPEAKER_FRONT_LEFT_OF_CENTER | SPEAKER_FRONT_RIGHT_OF_CENTER; break; + case 8 /*DSSPEAKER_7POINT1_SURROUND*/: channels = 8; channelMap = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_LOW_FREQUENCY | SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT | SPEAKER_SIDE_LEFT | SPEAKER_SIDE_RIGHT; break; + case 9 /*DSSPEAKER_5POINT1_SURROUND*/: channels = 6; channelMap = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT | SPEAKER_FRONT_CENTER | SPEAKER_LOW_FREQUENCY | SPEAKER_SIDE_LEFT | SPEAKER_SIDE_RIGHT; break; + default: break; + } + + if (pChannelsOut != NULL) { + *pChannelsOut = channels; + } + + if (pChannelMapOut != NULL) { + *pChannelMapOut = channelMap; + } +} + + +static ma_result ma_context_create_IDirectSound__dsound(ma_context* pContext, ma_share_mode shareMode, const ma_device_id* pDeviceID, ma_IDirectSound** ppDirectSound) +{ + ma_IDirectSound* pDirectSound; + HWND hWnd; + HRESULT hr; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppDirectSound != NULL); + + *ppDirectSound = NULL; + pDirectSound = NULL; + + if (FAILED(((ma_DirectSoundCreateProc)pContext->dsound.DirectSoundCreate)((pDeviceID == NULL) ? NULL : (const GUID*)pDeviceID->dsound, &pDirectSound, NULL))) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[DirectSound] DirectSoundCreate() failed for playback device."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + + /* The cooperative level must be set before doing anything else. */ + hWnd = (HWND)pContext->dsound.hWnd; + if (hWnd == 0) { + hWnd = ((MA_PFN_GetForegroundWindow)pContext->win32.GetForegroundWindow)(); + if (hWnd == 0) { + hWnd = ((MA_PFN_GetDesktopWindow)pContext->win32.GetDesktopWindow)(); + } + } + + hr = ma_IDirectSound_SetCooperativeLevel(pDirectSound, hWnd, (shareMode == ma_share_mode_exclusive) ? MA_DSSCL_EXCLUSIVE : MA_DSSCL_PRIORITY); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSound_SetCooperateiveLevel() failed for playback device."); + return ma_result_from_HRESULT(hr); + } + + *ppDirectSound = pDirectSound; + return MA_SUCCESS; +} + +static ma_result ma_context_create_IDirectSoundCapture__dsound(ma_context* pContext, ma_share_mode shareMode, const ma_device_id* pDeviceID, ma_IDirectSoundCapture** ppDirectSoundCapture) +{ + ma_IDirectSoundCapture* pDirectSoundCapture; + HRESULT hr; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppDirectSoundCapture != NULL); + + /* DirectSound does not support exclusive mode for capture. */ + if (shareMode == ma_share_mode_exclusive) { + return MA_SHARE_MODE_NOT_SUPPORTED; + } + + *ppDirectSoundCapture = NULL; + pDirectSoundCapture = NULL; + + hr = ((ma_DirectSoundCaptureCreateProc)pContext->dsound.DirectSoundCaptureCreate)((pDeviceID == NULL) ? NULL : (const GUID*)pDeviceID->dsound, &pDirectSoundCapture, NULL); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[DirectSound] DirectSoundCaptureCreate() failed for capture device."); + return ma_result_from_HRESULT(hr); + } + + *ppDirectSoundCapture = pDirectSoundCapture; + return MA_SUCCESS; +} + +static ma_result ma_context_get_format_info_for_IDirectSoundCapture__dsound(ma_context* pContext, ma_IDirectSoundCapture* pDirectSoundCapture, WORD* pChannels, WORD* pBitsPerSample, DWORD* pSampleRate) +{ + HRESULT hr; + MA_DSCCAPS caps; + WORD bitsPerSample; + DWORD sampleRate; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pDirectSoundCapture != NULL); + + if (pChannels) { + *pChannels = 0; + } + if (pBitsPerSample) { + *pBitsPerSample = 0; + } + if (pSampleRate) { + *pSampleRate = 0; + } + + MA_ZERO_OBJECT(&caps); + caps.dwSize = sizeof(caps); + hr = ma_IDirectSoundCapture_GetCaps(pDirectSoundCapture, &caps); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundCapture_GetCaps() failed for capture device."); + return ma_result_from_HRESULT(hr); + } + + if (pChannels) { + *pChannels = (WORD)caps.dwChannels; + } + + /* The device can support multiple formats. We just go through the different formats in order of priority and pick the first one. This the same type of system as the WinMM backend. */ + bitsPerSample = 16; + sampleRate = 48000; + + if (caps.dwChannels == 1) { + if ((caps.dwFormats & WAVE_FORMAT_48M16) != 0) { + sampleRate = 48000; + } else if ((caps.dwFormats & WAVE_FORMAT_44M16) != 0) { + sampleRate = 44100; + } else if ((caps.dwFormats & WAVE_FORMAT_2M16) != 0) { + sampleRate = 22050; + } else if ((caps.dwFormats & WAVE_FORMAT_1M16) != 0) { + sampleRate = 11025; + } else if ((caps.dwFormats & WAVE_FORMAT_96M16) != 0) { + sampleRate = 96000; + } else { + bitsPerSample = 8; + if ((caps.dwFormats & WAVE_FORMAT_48M08) != 0) { + sampleRate = 48000; + } else if ((caps.dwFormats & WAVE_FORMAT_44M08) != 0) { + sampleRate = 44100; + } else if ((caps.dwFormats & WAVE_FORMAT_2M08) != 0) { + sampleRate = 22050; + } else if ((caps.dwFormats & WAVE_FORMAT_1M08) != 0) { + sampleRate = 11025; + } else if ((caps.dwFormats & WAVE_FORMAT_96M08) != 0) { + sampleRate = 96000; + } else { + bitsPerSample = 16; /* Didn't find it. Just fall back to 16-bit. */ + } + } + } else if (caps.dwChannels == 2) { + if ((caps.dwFormats & WAVE_FORMAT_48S16) != 0) { + sampleRate = 48000; + } else if ((caps.dwFormats & WAVE_FORMAT_44S16) != 0) { + sampleRate = 44100; + } else if ((caps.dwFormats & WAVE_FORMAT_2S16) != 0) { + sampleRate = 22050; + } else if ((caps.dwFormats & WAVE_FORMAT_1S16) != 0) { + sampleRate = 11025; + } else if ((caps.dwFormats & WAVE_FORMAT_96S16) != 0) { + sampleRate = 96000; + } else { + bitsPerSample = 8; + if ((caps.dwFormats & WAVE_FORMAT_48S08) != 0) { + sampleRate = 48000; + } else if ((caps.dwFormats & WAVE_FORMAT_44S08) != 0) { + sampleRate = 44100; + } else if ((caps.dwFormats & WAVE_FORMAT_2S08) != 0) { + sampleRate = 22050; + } else if ((caps.dwFormats & WAVE_FORMAT_1S08) != 0) { + sampleRate = 11025; + } else if ((caps.dwFormats & WAVE_FORMAT_96S08) != 0) { + sampleRate = 96000; + } else { + bitsPerSample = 16; /* Didn't find it. Just fall back to 16-bit. */ + } + } + } + + if (pBitsPerSample) { + *pBitsPerSample = bitsPerSample; + } + if (pSampleRate) { + *pSampleRate = sampleRate; + } + + return MA_SUCCESS; +} + + +typedef struct +{ + ma_context* pContext; + ma_device_type deviceType; + ma_enum_devices_callback_proc callback; + void* pUserData; + ma_bool32 terminated; +} ma_context_enumerate_devices_callback_data__dsound; + +static BOOL CALLBACK ma_context_enumerate_devices_callback__dsound(GUID* lpGuid, const char* lpcstrDescription, const char* lpcstrModule, void* lpContext) +{ + ma_context_enumerate_devices_callback_data__dsound* pData = (ma_context_enumerate_devices_callback_data__dsound*)lpContext; + ma_device_info deviceInfo; + + (void)lpcstrModule; + + MA_ZERO_OBJECT(&deviceInfo); + + /* ID. */ + if (lpGuid != NULL) { + MA_COPY_MEMORY(deviceInfo.id.dsound, lpGuid, 16); + } else { + MA_ZERO_MEMORY(deviceInfo.id.dsound, 16); + deviceInfo.isDefault = MA_TRUE; + } + + /* Name / Description */ + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), lpcstrDescription, (size_t)-1); + + + /* Call the callback function, but make sure we stop enumerating if the callee requested so. */ + MA_ASSERT(pData != NULL); + pData->terminated = (pData->callback(pData->pContext, pData->deviceType, &deviceInfo, pData->pUserData) == MA_FALSE); + if (pData->terminated) { + return FALSE; /* Stop enumeration. */ + } else { + return TRUE; /* Continue enumeration. */ + } +} + +static ma_result ma_context_enumerate_devices__dsound(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_context_enumerate_devices_callback_data__dsound data; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + data.pContext = pContext; + data.callback = callback; + data.pUserData = pUserData; + data.terminated = MA_FALSE; + + /* Playback. */ + if (!data.terminated) { + data.deviceType = ma_device_type_playback; + ((ma_DirectSoundEnumerateAProc)pContext->dsound.DirectSoundEnumerateA)(ma_context_enumerate_devices_callback__dsound, &data); + } + + /* Capture. */ + if (!data.terminated) { + data.deviceType = ma_device_type_capture; + ((ma_DirectSoundCaptureEnumerateAProc)pContext->dsound.DirectSoundCaptureEnumerateA)(ma_context_enumerate_devices_callback__dsound, &data); + } + + return MA_SUCCESS; +} + + +typedef struct +{ + const ma_device_id* pDeviceID; + ma_device_info* pDeviceInfo; + ma_bool32 found; +} ma_context_get_device_info_callback_data__dsound; + +static BOOL CALLBACK ma_context_get_device_info_callback__dsound(GUID* lpGuid, const char* lpcstrDescription, const char* lpcstrModule, void* lpContext) +{ + ma_context_get_device_info_callback_data__dsound* pData = (ma_context_get_device_info_callback_data__dsound*)lpContext; + MA_ASSERT(pData != NULL); + + if ((pData->pDeviceID == NULL || ma_is_guid_null(pData->pDeviceID->dsound)) && (lpGuid == NULL || ma_is_guid_null(lpGuid))) { + /* Default device. */ + ma_strncpy_s(pData->pDeviceInfo->name, sizeof(pData->pDeviceInfo->name), lpcstrDescription, (size_t)-1); + pData->pDeviceInfo->isDefault = MA_TRUE; + pData->found = MA_TRUE; + return FALSE; /* Stop enumeration. */ + } else { + /* Not the default device. */ + if (lpGuid != NULL && pData->pDeviceID != NULL) { + if (memcmp(pData->pDeviceID->dsound, lpGuid, sizeof(pData->pDeviceID->dsound)) == 0) { + ma_strncpy_s(pData->pDeviceInfo->name, sizeof(pData->pDeviceInfo->name), lpcstrDescription, (size_t)-1); + pData->found = MA_TRUE; + return FALSE; /* Stop enumeration. */ + } + } + } + + (void)lpcstrModule; + return TRUE; +} + +static ma_result ma_context_get_device_info__dsound(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + ma_result result; + HRESULT hr; + + if (pDeviceID != NULL) { + ma_context_get_device_info_callback_data__dsound data; + + /* ID. */ + MA_COPY_MEMORY(pDeviceInfo->id.dsound, pDeviceID->dsound, 16); + + /* Name / Description. This is retrieved by enumerating over each device until we find that one that matches the input ID. */ + data.pDeviceID = pDeviceID; + data.pDeviceInfo = pDeviceInfo; + data.found = MA_FALSE; + if (deviceType == ma_device_type_playback) { + ((ma_DirectSoundEnumerateAProc)pContext->dsound.DirectSoundEnumerateA)(ma_context_get_device_info_callback__dsound, &data); + } else { + ((ma_DirectSoundCaptureEnumerateAProc)pContext->dsound.DirectSoundCaptureEnumerateA)(ma_context_get_device_info_callback__dsound, &data); + } + + if (!data.found) { + return MA_NO_DEVICE; + } + } else { + /* I don't think there's a way to get the name of the default device with DirectSound. In this case we just need to use defaults. */ + + /* ID */ + MA_ZERO_MEMORY(pDeviceInfo->id.dsound, 16); + + /* Name / Description */ + if (deviceType == ma_device_type_playback) { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } + + pDeviceInfo->isDefault = MA_TRUE; + } + + /* Retrieving detailed information is slightly different depending on the device type. */ + if (deviceType == ma_device_type_playback) { + /* Playback. */ + ma_IDirectSound* pDirectSound; + MA_DSCAPS caps; + WORD channels; + + result = ma_context_create_IDirectSound__dsound(pContext, ma_share_mode_shared, pDeviceID, &pDirectSound); + if (result != MA_SUCCESS) { + return result; + } + + MA_ZERO_OBJECT(&caps); + caps.dwSize = sizeof(caps); + hr = ma_IDirectSound_GetCaps(pDirectSound, &caps); + if (FAILED(hr)) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSound_GetCaps() failed for playback device."); + return ma_result_from_HRESULT(hr); + } + + + /* Channels. Only a single channel count is reported for DirectSound. */ + if ((caps.dwFlags & MA_DSCAPS_PRIMARYSTEREO) != 0) { + /* It supports at least stereo, but could support more. */ + DWORD speakerConfig; + + channels = 2; + + /* Look at the speaker configuration to get a better idea on the channel count. */ + hr = ma_IDirectSound_GetSpeakerConfig(pDirectSound, &speakerConfig); + if (SUCCEEDED(hr)) { + ma_get_channels_from_speaker_config__dsound(speakerConfig, &channels, NULL); + } + } else { + /* It does not support stereo, which means we are stuck with mono. */ + channels = 1; + } + + + /* + In DirectSound, our native formats are centered around sample rates. All formats are supported, and we're only reporting a single channel + count. However, DirectSound can report a range of supported sample rates. We're only going to include standard rates known by miniaudio + in order to keep the size of this within reason. + */ + if ((caps.dwFlags & MA_DSCAPS_CONTINUOUSRATE) != 0) { + /* Multiple sample rates are supported. We'll report in order of our preferred sample rates. */ + size_t iStandardSampleRate; + for (iStandardSampleRate = 0; iStandardSampleRate < ma_countof(g_maStandardSampleRatePriorities); iStandardSampleRate += 1) { + ma_uint32 sampleRate = g_maStandardSampleRatePriorities[iStandardSampleRate]; + if (sampleRate >= caps.dwMinSecondarySampleRate && sampleRate <= caps.dwMaxSecondarySampleRate) { + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].format = ma_format_unknown; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].channels = channels; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].sampleRate = sampleRate; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].flags = 0; + pDeviceInfo->nativeDataFormatCount += 1; + } + } + } else { + /* Only a single sample rate is supported. */ + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].format = ma_format_unknown; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].channels = channels; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].sampleRate = caps.dwMaxSecondarySampleRate; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].flags = 0; + pDeviceInfo->nativeDataFormatCount += 1; + } + + ma_IDirectSound_Release(pDirectSound); + } else { + /* + Capture. This is a little different to playback due to the say the supported formats are reported. Technically capture + devices can support a number of different formats, but for simplicity and consistency with ma_device_init() I'm just + reporting the best format. + */ + ma_IDirectSoundCapture* pDirectSoundCapture; + WORD channels; + WORD bitsPerSample; + DWORD sampleRate; + + result = ma_context_create_IDirectSoundCapture__dsound(pContext, ma_share_mode_shared, pDeviceID, &pDirectSoundCapture); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_context_get_format_info_for_IDirectSoundCapture__dsound(pContext, pDirectSoundCapture, &channels, &bitsPerSample, &sampleRate); + if (result != MA_SUCCESS) { + ma_IDirectSoundCapture_Release(pDirectSoundCapture); + return result; + } + + ma_IDirectSoundCapture_Release(pDirectSoundCapture); + + /* The format is always an integer format and is based on the bits per sample. */ + if (bitsPerSample == 8) { + pDeviceInfo->nativeDataFormats[0].format = ma_format_u8; + } else if (bitsPerSample == 16) { + pDeviceInfo->nativeDataFormats[0].format = ma_format_s16; + } else if (bitsPerSample == 24) { + pDeviceInfo->nativeDataFormats[0].format = ma_format_s24; + } else if (bitsPerSample == 32) { + pDeviceInfo->nativeDataFormats[0].format = ma_format_s32; + } else { + return MA_FORMAT_NOT_SUPPORTED; + } + + pDeviceInfo->nativeDataFormats[0].channels = channels; + pDeviceInfo->nativeDataFormats[0].sampleRate = sampleRate; + pDeviceInfo->nativeDataFormats[0].flags = 0; + pDeviceInfo->nativeDataFormatCount = 1; + } + + return MA_SUCCESS; +} + + + +static ma_result ma_device_uninit__dsound(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->dsound.pCaptureBuffer != NULL) { + ma_IDirectSoundCaptureBuffer_Release((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer); + } + if (pDevice->dsound.pCapture != NULL) { + ma_IDirectSoundCapture_Release((ma_IDirectSoundCapture*)pDevice->dsound.pCapture); + } + + if (pDevice->dsound.pPlaybackBuffer != NULL) { + ma_IDirectSoundBuffer_Release((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer); + } + if (pDevice->dsound.pPlaybackPrimaryBuffer != NULL) { + ma_IDirectSoundBuffer_Release((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackPrimaryBuffer); + } + if (pDevice->dsound.pPlayback != NULL) { + ma_IDirectSound_Release((ma_IDirectSound*)pDevice->dsound.pPlayback); + } + + return MA_SUCCESS; +} + +static ma_result ma_config_to_WAVEFORMATEXTENSIBLE(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, const ma_channel* pChannelMap, MA_WAVEFORMATEXTENSIBLE* pWF) +{ + GUID subformat; + + if (format == ma_format_unknown) { + format = MA_DEFAULT_FORMAT; + } + + if (channels == 0) { + channels = MA_DEFAULT_CHANNELS; + } + + if (sampleRate == 0) { + sampleRate = MA_DEFAULT_SAMPLE_RATE; + } + + switch (format) + { + case ma_format_u8: + case ma_format_s16: + case ma_format_s24: + /*case ma_format_s24_32:*/ + case ma_format_s32: + { + subformat = MA_GUID_KSDATAFORMAT_SUBTYPE_PCM; + } break; + + case ma_format_f32: + { + subformat = MA_GUID_KSDATAFORMAT_SUBTYPE_IEEE_FLOAT; + } break; + + default: + return MA_FORMAT_NOT_SUPPORTED; + } + + MA_ZERO_OBJECT(pWF); + pWF->cbSize = sizeof(*pWF); + pWF->wFormatTag = WAVE_FORMAT_EXTENSIBLE; + pWF->nChannels = (WORD)channels; + pWF->nSamplesPerSec = (DWORD)sampleRate; + pWF->wBitsPerSample = (WORD)(ma_get_bytes_per_sample(format)*8); + pWF->nBlockAlign = (WORD)(pWF->nChannels * pWF->wBitsPerSample / 8); + pWF->nAvgBytesPerSec = pWF->nBlockAlign * pWF->nSamplesPerSec; + pWF->Samples.wValidBitsPerSample = pWF->wBitsPerSample; + pWF->dwChannelMask = ma_channel_map_to_channel_mask__win32(pChannelMap, channels); + pWF->SubFormat = subformat; + + return MA_SUCCESS; +} + +static ma_uint32 ma_calculate_period_size_in_frames_from_descriptor__dsound(const ma_device_descriptor* pDescriptor, ma_uint32 nativeSampleRate, ma_performance_profile performanceProfile) +{ + /* + DirectSound has a minimum period size of 20ms. In practice, this doesn't seem to be enough for + reliable glitch-free processing so going to use 30ms instead. + */ + ma_uint32 minPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(30, nativeSampleRate); + ma_uint32 periodSizeInFrames; + + periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptor, nativeSampleRate, performanceProfile); + if (periodSizeInFrames < minPeriodSizeInFrames) { + periodSizeInFrames = minPeriodSizeInFrames; + } + + return periodSizeInFrames; +} + +static ma_result ma_device_init__dsound(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + ma_result result; + HRESULT hr; + + MA_ASSERT(pDevice != NULL); + + MA_ZERO_OBJECT(&pDevice->dsound); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + /* + Unfortunately DirectSound uses different APIs and data structures for playback and capture devices. We need to initialize + the capture device first because we'll want to match its buffer size and period count on the playback side if we're using + full-duplex mode. + */ + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + MA_WAVEFORMATEXTENSIBLE wf; + MA_DSCBUFFERDESC descDS; + ma_uint32 periodSizeInFrames; + ma_uint32 periodCount; + char rawdata[1024]; /* <-- Ugly hack to avoid a malloc() due to a crappy DirectSound API. */ + MA_WAVEFORMATEXTENSIBLE* pActualFormat; + + result = ma_config_to_WAVEFORMATEXTENSIBLE(pDescriptorCapture->format, pDescriptorCapture->channels, pDescriptorCapture->sampleRate, pDescriptorCapture->channelMap, &wf); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_context_create_IDirectSoundCapture__dsound(pDevice->pContext, pDescriptorCapture->shareMode, pDescriptorCapture->pDeviceID, (ma_IDirectSoundCapture**)&pDevice->dsound.pCapture); + if (result != MA_SUCCESS) { + ma_device_uninit__dsound(pDevice); + return result; + } + + result = ma_context_get_format_info_for_IDirectSoundCapture__dsound(pDevice->pContext, (ma_IDirectSoundCapture*)pDevice->dsound.pCapture, &wf.nChannels, &wf.wBitsPerSample, &wf.nSamplesPerSec); + if (result != MA_SUCCESS) { + ma_device_uninit__dsound(pDevice); + return result; + } + + wf.nBlockAlign = (WORD)(wf.nChannels * wf.wBitsPerSample / 8); + wf.nAvgBytesPerSec = wf.nBlockAlign * wf.nSamplesPerSec; + wf.Samples.wValidBitsPerSample = wf.wBitsPerSample; + wf.SubFormat = MA_GUID_KSDATAFORMAT_SUBTYPE_PCM; + + /* The size of the buffer must be a clean multiple of the period count. */ + periodSizeInFrames = ma_calculate_period_size_in_frames_from_descriptor__dsound(pDescriptorCapture, wf.nSamplesPerSec, pConfig->performanceProfile); + periodCount = (pDescriptorCapture->periodCount > 0) ? pDescriptorCapture->periodCount : MA_DEFAULT_PERIODS; + + MA_ZERO_OBJECT(&descDS); + descDS.dwSize = sizeof(descDS); + descDS.dwFlags = 0; + descDS.dwBufferBytes = periodSizeInFrames * periodCount * wf.nBlockAlign; + descDS.lpwfxFormat = (MA_WAVEFORMATEX*)&wf; + hr = ma_IDirectSoundCapture_CreateCaptureBuffer((ma_IDirectSoundCapture*)pDevice->dsound.pCapture, &descDS, (ma_IDirectSoundCaptureBuffer**)&pDevice->dsound.pCaptureBuffer, NULL); + if (FAILED(hr)) { + ma_device_uninit__dsound(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundCapture_CreateCaptureBuffer() failed for capture device."); + return ma_result_from_HRESULT(hr); + } + + /* Get the _actual_ properties of the buffer. */ + pActualFormat = (MA_WAVEFORMATEXTENSIBLE*)rawdata; + hr = ma_IDirectSoundCaptureBuffer_GetFormat((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer, (MA_WAVEFORMATEX*)pActualFormat, sizeof(rawdata), NULL); + if (FAILED(hr)) { + ma_device_uninit__dsound(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to retrieve the actual format of the capture device's buffer."); + return ma_result_from_HRESULT(hr); + } + + /* We can now start setting the output data formats. */ + pDescriptorCapture->format = ma_format_from_WAVEFORMATEX((MA_WAVEFORMATEX*)pActualFormat); + pDescriptorCapture->channels = pActualFormat->nChannels; + pDescriptorCapture->sampleRate = pActualFormat->nSamplesPerSec; + + /* Get the native channel map based on the channel mask. */ + if (pActualFormat->wFormatTag == WAVE_FORMAT_EXTENSIBLE) { + ma_channel_mask_to_channel_map__win32(pActualFormat->dwChannelMask, pDescriptorCapture->channels, pDescriptorCapture->channelMap); + } else { + ma_channel_mask_to_channel_map__win32(wf.dwChannelMask, pDescriptorCapture->channels, pDescriptorCapture->channelMap); + } + + /* + After getting the actual format the size of the buffer in frames may have actually changed. However, we want this to be as close to what the + user has asked for as possible, so let's go ahead and release the old capture buffer and create a new one in this case. + */ + if (periodSizeInFrames != (descDS.dwBufferBytes / ma_get_bytes_per_frame(pDescriptorCapture->format, pDescriptorCapture->channels) / periodCount)) { + descDS.dwBufferBytes = periodSizeInFrames * ma_get_bytes_per_frame(pDescriptorCapture->format, pDescriptorCapture->channels) * periodCount; + ma_IDirectSoundCaptureBuffer_Release((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer); + + hr = ma_IDirectSoundCapture_CreateCaptureBuffer((ma_IDirectSoundCapture*)pDevice->dsound.pCapture, &descDS, (ma_IDirectSoundCaptureBuffer**)&pDevice->dsound.pCaptureBuffer, NULL); + if (FAILED(hr)) { + ma_device_uninit__dsound(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Second attempt at IDirectSoundCapture_CreateCaptureBuffer() failed for capture device."); + return ma_result_from_HRESULT(hr); + } + } + + /* DirectSound should give us a buffer exactly the size we asked for. */ + pDescriptorCapture->periodSizeInFrames = periodSizeInFrames; + pDescriptorCapture->periodCount = periodCount; + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + MA_WAVEFORMATEXTENSIBLE wf; + MA_DSBUFFERDESC descDSPrimary; + MA_DSCAPS caps; + char rawdata[1024]; /* <-- Ugly hack to avoid a malloc() due to a crappy DirectSound API. */ + MA_WAVEFORMATEXTENSIBLE* pActualFormat; + ma_uint32 periodSizeInFrames; + ma_uint32 periodCount; + MA_DSBUFFERDESC descDS; + WORD nativeChannelCount; + DWORD nativeChannelMask = 0; + + result = ma_config_to_WAVEFORMATEXTENSIBLE(pDescriptorPlayback->format, pDescriptorPlayback->channels, pDescriptorPlayback->sampleRate, pDescriptorPlayback->channelMap, &wf); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_context_create_IDirectSound__dsound(pDevice->pContext, pDescriptorPlayback->shareMode, pDescriptorPlayback->pDeviceID, (ma_IDirectSound**)&pDevice->dsound.pPlayback); + if (result != MA_SUCCESS) { + ma_device_uninit__dsound(pDevice); + return result; + } + + MA_ZERO_OBJECT(&descDSPrimary); + descDSPrimary.dwSize = sizeof(MA_DSBUFFERDESC); + descDSPrimary.dwFlags = MA_DSBCAPS_PRIMARYBUFFER | MA_DSBCAPS_CTRLVOLUME; + hr = ma_IDirectSound_CreateSoundBuffer((ma_IDirectSound*)pDevice->dsound.pPlayback, &descDSPrimary, (ma_IDirectSoundBuffer**)&pDevice->dsound.pPlaybackPrimaryBuffer, NULL); + if (FAILED(hr)) { + ma_device_uninit__dsound(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSound_CreateSoundBuffer() failed for playback device's primary buffer."); + return ma_result_from_HRESULT(hr); + } + + + /* We may want to make some adjustments to the format if we are using defaults. */ + MA_ZERO_OBJECT(&caps); + caps.dwSize = sizeof(caps); + hr = ma_IDirectSound_GetCaps((ma_IDirectSound*)pDevice->dsound.pPlayback, &caps); + if (FAILED(hr)) { + ma_device_uninit__dsound(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSound_GetCaps() failed for playback device."); + return ma_result_from_HRESULT(hr); + } + + if ((caps.dwFlags & MA_DSCAPS_PRIMARYSTEREO) != 0) { + DWORD speakerConfig; + + /* It supports at least stereo, but could support more. */ + nativeChannelCount = 2; + + /* Look at the speaker configuration to get a better idea on the channel count. */ + if (SUCCEEDED(ma_IDirectSound_GetSpeakerConfig((ma_IDirectSound*)pDevice->dsound.pPlayback, &speakerConfig))) { + ma_get_channels_from_speaker_config__dsound(speakerConfig, &nativeChannelCount, &nativeChannelMask); + } + } else { + /* It does not support stereo, which means we are stuck with mono. */ + nativeChannelCount = 1; + nativeChannelMask = 0x00000001; + } + + if (pDescriptorPlayback->channels == 0) { + wf.nChannels = nativeChannelCount; + wf.dwChannelMask = nativeChannelMask; + } + + if (pDescriptorPlayback->sampleRate == 0) { + /* We base the sample rate on the values returned by GetCaps(). */ + if ((caps.dwFlags & MA_DSCAPS_CONTINUOUSRATE) != 0) { + wf.nSamplesPerSec = ma_get_best_sample_rate_within_range(caps.dwMinSecondarySampleRate, caps.dwMaxSecondarySampleRate); + } else { + wf.nSamplesPerSec = caps.dwMaxSecondarySampleRate; + } + } + + wf.nBlockAlign = (WORD)(wf.nChannels * wf.wBitsPerSample / 8); + wf.nAvgBytesPerSec = wf.nBlockAlign * wf.nSamplesPerSec; + + /* + From MSDN: + + The method succeeds even if the hardware does not support the requested format; DirectSound sets the buffer to the closest + supported format. To determine whether this has happened, an application can call the GetFormat method for the primary buffer + and compare the result with the format that was requested with the SetFormat method. + */ + hr = ma_IDirectSoundBuffer_SetFormat((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackPrimaryBuffer, (MA_WAVEFORMATEX*)&wf); + if (FAILED(hr)) { + /* + If setting of the format failed we'll try again with some fallback settings. On Windows 98 I have + observed that IEEE_FLOAT does not work. We'll therefore enforce PCM. I also had issues where a + sample rate of 48000 did not work correctly. Not sure if it was a driver issue or not, but will + use 44100 for the sample rate. + */ + wf.cbSize = 18; /* NOTE: Don't use sizeof(MA_WAVEFORMATEX) here because it's got an extra 2 bytes due to padding. */ + wf.wFormatTag = WAVE_FORMAT_PCM; + wf.wBitsPerSample = 16; + wf.nChannels = nativeChannelCount; + wf.nSamplesPerSec = 44100; + wf.nBlockAlign = wf.nChannels * (wf.wBitsPerSample / 8); + wf.nAvgBytesPerSec = wf.nSamplesPerSec * wf.nBlockAlign; + + hr = ma_IDirectSoundBuffer_SetFormat((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackPrimaryBuffer, (MA_WAVEFORMATEX*)&wf); + if (FAILED(hr)) { + ma_device_uninit__dsound(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to set format of playback device's primary buffer."); + return ma_result_from_HRESULT(hr); + } + } + + /* Get the _actual_ properties of the buffer. */ + pActualFormat = (MA_WAVEFORMATEXTENSIBLE*)rawdata; + hr = ma_IDirectSoundBuffer_GetFormat((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackPrimaryBuffer, (MA_WAVEFORMATEX*)pActualFormat, sizeof(rawdata), NULL); + if (FAILED(hr)) { + ma_device_uninit__dsound(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to retrieve the actual format of the playback device's primary buffer."); + return ma_result_from_HRESULT(hr); + } + + /* We now have enough information to start setting some output properties. */ + pDescriptorPlayback->format = ma_format_from_WAVEFORMATEX((MA_WAVEFORMATEX*)pActualFormat); + pDescriptorPlayback->channels = pActualFormat->nChannels; + pDescriptorPlayback->sampleRate = pActualFormat->nSamplesPerSec; + + /* Get the internal channel map based on the channel mask. */ + if (pActualFormat->wFormatTag == WAVE_FORMAT_EXTENSIBLE) { + ma_channel_mask_to_channel_map__win32(pActualFormat->dwChannelMask, pDescriptorPlayback->channels, pDescriptorPlayback->channelMap); + } else { + ma_channel_mask_to_channel_map__win32(wf.dwChannelMask, pDescriptorPlayback->channels, pDescriptorPlayback->channelMap); + } + + /* The size of the buffer must be a clean multiple of the period count. */ + periodSizeInFrames = ma_calculate_period_size_in_frames_from_descriptor__dsound(pDescriptorPlayback, pDescriptorPlayback->sampleRate, pConfig->performanceProfile); + periodCount = (pDescriptorPlayback->periodCount > 0) ? pDescriptorPlayback->periodCount : MA_DEFAULT_PERIODS; + + /* + Meaning of dwFlags (from MSDN): + + DSBCAPS_CTRLPOSITIONNOTIFY + The buffer has position notification capability. + + DSBCAPS_GLOBALFOCUS + With this flag set, an application using DirectSound can continue to play its buffers if the user switches focus to + another application, even if the new application uses DirectSound. + + DSBCAPS_GETCURRENTPOSITION2 + In the first version of DirectSound, the play cursor was significantly ahead of the actual playing sound on emulated + sound cards; it was directly behind the write cursor. Now, if the DSBCAPS_GETCURRENTPOSITION2 flag is specified, the + application can get a more accurate play cursor. + */ + MA_ZERO_OBJECT(&descDS); + descDS.dwSize = sizeof(descDS); + descDS.dwFlags = MA_DSBCAPS_CTRLPOSITIONNOTIFY | MA_DSBCAPS_GLOBALFOCUS | MA_DSBCAPS_GETCURRENTPOSITION2; + descDS.dwBufferBytes = periodSizeInFrames * periodCount * ma_get_bytes_per_frame(pDescriptorPlayback->format, pDescriptorPlayback->channels); + descDS.lpwfxFormat = (MA_WAVEFORMATEX*)pActualFormat; + hr = ma_IDirectSound_CreateSoundBuffer((ma_IDirectSound*)pDevice->dsound.pPlayback, &descDS, (ma_IDirectSoundBuffer**)&pDevice->dsound.pPlaybackBuffer, NULL); + if (FAILED(hr)) { + ma_device_uninit__dsound(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSound_CreateSoundBuffer() failed for playback device's secondary buffer."); + return ma_result_from_HRESULT(hr); + } + + /* DirectSound should give us a buffer exactly the size we asked for. */ + pDescriptorPlayback->periodSizeInFrames = periodSizeInFrames; + pDescriptorPlayback->periodCount = periodCount; + } + + return MA_SUCCESS; +} + + +static ma_result ma_device_data_loop__dsound(ma_device* pDevice) +{ + ma_result result = MA_SUCCESS; + ma_uint32 bpfDeviceCapture = ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + ma_uint32 bpfDevicePlayback = ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + HRESULT hr; + DWORD lockOffsetInBytesCapture; + DWORD lockSizeInBytesCapture; + DWORD mappedSizeInBytesCapture; + DWORD mappedDeviceFramesProcessedCapture; + void* pMappedDeviceBufferCapture; + DWORD lockOffsetInBytesPlayback; + DWORD lockSizeInBytesPlayback; + DWORD mappedSizeInBytesPlayback; + void* pMappedDeviceBufferPlayback; + DWORD prevReadCursorInBytesCapture = 0; + DWORD prevPlayCursorInBytesPlayback = 0; + ma_bool32 physicalPlayCursorLoopFlagPlayback = 0; + DWORD virtualWriteCursorInBytesPlayback = 0; + ma_bool32 virtualWriteCursorLoopFlagPlayback = 0; + ma_bool32 isPlaybackDeviceStarted = MA_FALSE; + ma_uint32 framesWrittenToPlaybackDevice = 0; /* For knowing whether or not the playback device needs to be started. */ + ma_uint32 waitTimeInMilliseconds = 1; + DWORD playbackBufferStatus = 0; + + MA_ASSERT(pDevice != NULL); + + /* The first thing to do is start the capture device. The playback device is only started after the first period is written. */ + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + hr = ma_IDirectSoundCaptureBuffer_Start((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer, MA_DSCBSTART_LOOPING); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundCaptureBuffer_Start() failed."); + return ma_result_from_HRESULT(hr); + } + } + + while (ma_device_get_state(pDevice) == ma_device_state_started) { + switch (pDevice->type) + { + case ma_device_type_duplex: + { + DWORD physicalCaptureCursorInBytes; + DWORD physicalReadCursorInBytes; + hr = ma_IDirectSoundCaptureBuffer_GetCurrentPosition((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer, &physicalCaptureCursorInBytes, &physicalReadCursorInBytes); + if (FAILED(hr)) { + return ma_result_from_HRESULT(hr); + } + + /* If nothing is available we just sleep for a bit and return from this iteration. */ + if (physicalReadCursorInBytes == prevReadCursorInBytesCapture) { + ma_sleep(waitTimeInMilliseconds); + continue; /* Nothing is available in the capture buffer. */ + } + + /* + The current position has moved. We need to map all of the captured samples and write them to the playback device, making sure + we don't return until every frame has been copied over. + */ + if (prevReadCursorInBytesCapture < physicalReadCursorInBytes) { + /* The capture position has not looped. This is the simple case. */ + lockOffsetInBytesCapture = prevReadCursorInBytesCapture; + lockSizeInBytesCapture = (physicalReadCursorInBytes - prevReadCursorInBytesCapture); + } else { + /* + The capture position has looped. This is the more complex case. Map to the end of the buffer. If this does not return anything, + do it again from the start. + */ + if (prevReadCursorInBytesCapture < pDevice->capture.internalPeriodSizeInFrames*pDevice->capture.internalPeriods*bpfDeviceCapture) { + /* Lock up to the end of the buffer. */ + lockOffsetInBytesCapture = prevReadCursorInBytesCapture; + lockSizeInBytesCapture = (pDevice->capture.internalPeriodSizeInFrames*pDevice->capture.internalPeriods*bpfDeviceCapture) - prevReadCursorInBytesCapture; + } else { + /* Lock starting from the start of the buffer. */ + lockOffsetInBytesCapture = 0; + lockSizeInBytesCapture = physicalReadCursorInBytes; + } + } + + if (lockSizeInBytesCapture == 0) { + ma_sleep(waitTimeInMilliseconds); + continue; /* Nothing is available in the capture buffer. */ + } + + hr = ma_IDirectSoundCaptureBuffer_Lock((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer, lockOffsetInBytesCapture, lockSizeInBytesCapture, &pMappedDeviceBufferCapture, &mappedSizeInBytesCapture, NULL, NULL, 0); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to map buffer from capture device in preparation for writing to the device."); + return ma_result_from_HRESULT(hr); + } + + + /* At this point we have some input data that we need to output. We do not return until every mapped frame of the input data is written to the playback device. */ + mappedDeviceFramesProcessedCapture = 0; + + for (;;) { /* Keep writing to the playback device. */ + ma_uint8 inputFramesInClientFormat[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint32 inputFramesInClientFormatCap = sizeof(inputFramesInClientFormat) / ma_get_bytes_per_frame(pDevice->capture.format, pDevice->capture.channels); + ma_uint8 outputFramesInClientFormat[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint32 outputFramesInClientFormatCap = sizeof(outputFramesInClientFormat) / ma_get_bytes_per_frame(pDevice->playback.format, pDevice->playback.channels); + ma_uint32 outputFramesInClientFormatCount; + ma_uint32 outputFramesInClientFormatConsumed = 0; + ma_uint64 clientCapturedFramesToProcess = ma_min(inputFramesInClientFormatCap, outputFramesInClientFormatCap); + ma_uint64 deviceCapturedFramesToProcess = (mappedSizeInBytesCapture / bpfDeviceCapture) - mappedDeviceFramesProcessedCapture; + void* pRunningMappedDeviceBufferCapture = ma_offset_ptr(pMappedDeviceBufferCapture, mappedDeviceFramesProcessedCapture * bpfDeviceCapture); + + result = ma_data_converter_process_pcm_frames(&pDevice->capture.converter, pRunningMappedDeviceBufferCapture, &deviceCapturedFramesToProcess, inputFramesInClientFormat, &clientCapturedFramesToProcess); + if (result != MA_SUCCESS) { + break; + } + + outputFramesInClientFormatCount = (ma_uint32)clientCapturedFramesToProcess; + mappedDeviceFramesProcessedCapture += (ma_uint32)deviceCapturedFramesToProcess; + + ma_device__handle_data_callback(pDevice, outputFramesInClientFormat, inputFramesInClientFormat, (ma_uint32)clientCapturedFramesToProcess); + + /* At this point we have input and output data in client format. All we need to do now is convert it to the output device format. This may take a few passes. */ + for (;;) { + ma_uint32 framesWrittenThisIteration; + DWORD physicalPlayCursorInBytes; + DWORD physicalWriteCursorInBytes; + DWORD availableBytesPlayback; + DWORD silentPaddingInBytes = 0; /* <-- Must be initialized to 0. */ + + /* We need the physical play and write cursors. */ + if (FAILED(ma_IDirectSoundBuffer_GetCurrentPosition((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, &physicalPlayCursorInBytes, &physicalWriteCursorInBytes))) { + break; + } + + if (physicalPlayCursorInBytes < prevPlayCursorInBytesPlayback) { + physicalPlayCursorLoopFlagPlayback = !physicalPlayCursorLoopFlagPlayback; + } + prevPlayCursorInBytesPlayback = physicalPlayCursorInBytes; + + /* If there's any bytes available for writing we can do that now. The space between the virtual cursor position and play cursor. */ + if (physicalPlayCursorLoopFlagPlayback == virtualWriteCursorLoopFlagPlayback) { + /* Same loop iteration. The available bytes wraps all the way around from the virtual write cursor to the physical play cursor. */ + if (physicalPlayCursorInBytes <= virtualWriteCursorInBytesPlayback) { + availableBytesPlayback = (pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods*bpfDevicePlayback) - virtualWriteCursorInBytesPlayback; + availableBytesPlayback += physicalPlayCursorInBytes; /* Wrap around. */ + } else { + /* This is an error. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[DirectSound] (Duplex/Playback): Play cursor has moved in front of the write cursor (same loop iteration). physicalPlayCursorInBytes=%ld, virtualWriteCursorInBytes=%ld.\n", physicalPlayCursorInBytes, virtualWriteCursorInBytesPlayback); + availableBytesPlayback = 0; + } + } else { + /* Different loop iterations. The available bytes only goes from the virtual write cursor to the physical play cursor. */ + if (physicalPlayCursorInBytes >= virtualWriteCursorInBytesPlayback) { + availableBytesPlayback = physicalPlayCursorInBytes - virtualWriteCursorInBytesPlayback; + } else { + /* This is an error. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[DirectSound] (Duplex/Playback): Write cursor has moved behind the play cursor (different loop iterations). physicalPlayCursorInBytes=%ld, virtualWriteCursorInBytes=%ld.\n", physicalPlayCursorInBytes, virtualWriteCursorInBytesPlayback); + availableBytesPlayback = 0; + } + } + + /* If there's no room available for writing we need to wait for more. */ + if (availableBytesPlayback == 0) { + /* If we haven't started the device yet, this will never get beyond 0. In this case we need to get the device started. */ + if (!isPlaybackDeviceStarted) { + hr = ma_IDirectSoundBuffer_Play((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, 0, 0, MA_DSBPLAY_LOOPING); + if (FAILED(hr)) { + ma_IDirectSoundCaptureBuffer_Stop((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundBuffer_Play() failed."); + return ma_result_from_HRESULT(hr); + } + isPlaybackDeviceStarted = MA_TRUE; + } else { + ma_sleep(waitTimeInMilliseconds); + continue; + } + } + + + /* Getting here means there room available somewhere. We limit this to either the end of the buffer or the physical play cursor, whichever is closest. */ + lockOffsetInBytesPlayback = virtualWriteCursorInBytesPlayback; + if (physicalPlayCursorLoopFlagPlayback == virtualWriteCursorLoopFlagPlayback) { + /* Same loop iteration. Go up to the end of the buffer. */ + lockSizeInBytesPlayback = (pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods*bpfDevicePlayback) - virtualWriteCursorInBytesPlayback; + } else { + /* Different loop iterations. Go up to the physical play cursor. */ + lockSizeInBytesPlayback = physicalPlayCursorInBytes - virtualWriteCursorInBytesPlayback; + } + + hr = ma_IDirectSoundBuffer_Lock((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, lockOffsetInBytesPlayback, lockSizeInBytesPlayback, &pMappedDeviceBufferPlayback, &mappedSizeInBytesPlayback, NULL, NULL, 0); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to map buffer from playback device in preparation for writing to the device."); + result = ma_result_from_HRESULT(hr); + break; + } + + /* + Experiment: If the playback buffer is being starved, pad it with some silence to get it back in sync. This will cause a glitch, but it may prevent + endless glitching due to it constantly running out of data. + */ + if (isPlaybackDeviceStarted) { + DWORD bytesQueuedForPlayback = (pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods*bpfDevicePlayback) - availableBytesPlayback; + if (bytesQueuedForPlayback < (pDevice->playback.internalPeriodSizeInFrames*bpfDevicePlayback)) { + silentPaddingInBytes = (pDevice->playback.internalPeriodSizeInFrames*2*bpfDevicePlayback) - bytesQueuedForPlayback; + if (silentPaddingInBytes > lockSizeInBytesPlayback) { + silentPaddingInBytes = lockSizeInBytesPlayback; + } + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[DirectSound] (Duplex/Playback) Playback buffer starved. availableBytesPlayback=%ld, silentPaddingInBytes=%ld\n", availableBytesPlayback, silentPaddingInBytes); + } + } + + /* At this point we have a buffer for output. */ + if (silentPaddingInBytes > 0) { + MA_ZERO_MEMORY(pMappedDeviceBufferPlayback, silentPaddingInBytes); + framesWrittenThisIteration = silentPaddingInBytes/bpfDevicePlayback; + } else { + ma_uint64 convertedFrameCountIn = (outputFramesInClientFormatCount - outputFramesInClientFormatConsumed); + ma_uint64 convertedFrameCountOut = mappedSizeInBytesPlayback/bpfDevicePlayback; + void* pConvertedFramesIn = ma_offset_ptr(outputFramesInClientFormat, outputFramesInClientFormatConsumed * bpfDevicePlayback); + void* pConvertedFramesOut = pMappedDeviceBufferPlayback; + + result = ma_data_converter_process_pcm_frames(&pDevice->playback.converter, pConvertedFramesIn, &convertedFrameCountIn, pConvertedFramesOut, &convertedFrameCountOut); + if (result != MA_SUCCESS) { + break; + } + + outputFramesInClientFormatConsumed += (ma_uint32)convertedFrameCountOut; + framesWrittenThisIteration = (ma_uint32)convertedFrameCountOut; + } + + + hr = ma_IDirectSoundBuffer_Unlock((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, pMappedDeviceBufferPlayback, framesWrittenThisIteration*bpfDevicePlayback, NULL, 0); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to unlock internal buffer from playback device after writing to the device."); + result = ma_result_from_HRESULT(hr); + break; + } + + virtualWriteCursorInBytesPlayback += framesWrittenThisIteration*bpfDevicePlayback; + if ((virtualWriteCursorInBytesPlayback/bpfDevicePlayback) == pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods) { + virtualWriteCursorInBytesPlayback = 0; + virtualWriteCursorLoopFlagPlayback = !virtualWriteCursorLoopFlagPlayback; + } + + /* + We may need to start the device. We want two full periods to be written before starting the playback device. Having an extra period adds + a bit of a buffer to prevent the playback buffer from getting starved. + */ + framesWrittenToPlaybackDevice += framesWrittenThisIteration; + if (!isPlaybackDeviceStarted && framesWrittenToPlaybackDevice >= (pDevice->playback.internalPeriodSizeInFrames*2)) { + hr = ma_IDirectSoundBuffer_Play((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, 0, 0, MA_DSBPLAY_LOOPING); + if (FAILED(hr)) { + ma_IDirectSoundCaptureBuffer_Stop((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundBuffer_Play() failed."); + return ma_result_from_HRESULT(hr); + } + isPlaybackDeviceStarted = MA_TRUE; + } + + if (framesWrittenThisIteration < mappedSizeInBytesPlayback/bpfDevicePlayback) { + break; /* We're finished with the output data.*/ + } + } + + if (clientCapturedFramesToProcess == 0) { + break; /* We just consumed every input sample. */ + } + } + + + /* At this point we're done with the mapped portion of the capture buffer. */ + hr = ma_IDirectSoundCaptureBuffer_Unlock((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer, pMappedDeviceBufferCapture, mappedSizeInBytesCapture, NULL, 0); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to unlock internal buffer from capture device after reading from the device."); + return ma_result_from_HRESULT(hr); + } + prevReadCursorInBytesCapture = (lockOffsetInBytesCapture + mappedSizeInBytesCapture); + } break; + + + + case ma_device_type_capture: + { + DWORD physicalCaptureCursorInBytes; + DWORD physicalReadCursorInBytes; + hr = ma_IDirectSoundCaptureBuffer_GetCurrentPosition((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer, &physicalCaptureCursorInBytes, &physicalReadCursorInBytes); + if (FAILED(hr)) { + return MA_ERROR; + } + + /* If the previous capture position is the same as the current position we need to wait a bit longer. */ + if (prevReadCursorInBytesCapture == physicalReadCursorInBytes) { + ma_sleep(waitTimeInMilliseconds); + continue; + } + + /* Getting here means we have capture data available. */ + if (prevReadCursorInBytesCapture < physicalReadCursorInBytes) { + /* The capture position has not looped. This is the simple case. */ + lockOffsetInBytesCapture = prevReadCursorInBytesCapture; + lockSizeInBytesCapture = (physicalReadCursorInBytes - prevReadCursorInBytesCapture); + } else { + /* + The capture position has looped. This is the more complex case. Map to the end of the buffer. If this does not return anything, + do it again from the start. + */ + if (prevReadCursorInBytesCapture < pDevice->capture.internalPeriodSizeInFrames*pDevice->capture.internalPeriods*bpfDeviceCapture) { + /* Lock up to the end of the buffer. */ + lockOffsetInBytesCapture = prevReadCursorInBytesCapture; + lockSizeInBytesCapture = (pDevice->capture.internalPeriodSizeInFrames*pDevice->capture.internalPeriods*bpfDeviceCapture) - prevReadCursorInBytesCapture; + } else { + /* Lock starting from the start of the buffer. */ + lockOffsetInBytesCapture = 0; + lockSizeInBytesCapture = physicalReadCursorInBytes; + } + } + + if (lockSizeInBytesCapture < pDevice->capture.internalPeriodSizeInFrames) { + ma_sleep(waitTimeInMilliseconds); + continue; /* Nothing is available in the capture buffer. */ + } + + hr = ma_IDirectSoundCaptureBuffer_Lock((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer, lockOffsetInBytesCapture, lockSizeInBytesCapture, &pMappedDeviceBufferCapture, &mappedSizeInBytesCapture, NULL, NULL, 0); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to map buffer from capture device in preparation for writing to the device."); + result = ma_result_from_HRESULT(hr); + } + + if (lockSizeInBytesCapture != mappedSizeInBytesCapture) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[DirectSound] (Capture) lockSizeInBytesCapture=%ld != mappedSizeInBytesCapture=%ld\n", lockSizeInBytesCapture, mappedSizeInBytesCapture); + } + + ma_device__send_frames_to_client(pDevice, mappedSizeInBytesCapture/bpfDeviceCapture, pMappedDeviceBufferCapture); + + hr = ma_IDirectSoundCaptureBuffer_Unlock((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer, pMappedDeviceBufferCapture, mappedSizeInBytesCapture, NULL, 0); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to unlock internal buffer from capture device after reading from the device."); + return ma_result_from_HRESULT(hr); + } + prevReadCursorInBytesCapture = lockOffsetInBytesCapture + mappedSizeInBytesCapture; + + if (prevReadCursorInBytesCapture == (pDevice->capture.internalPeriodSizeInFrames*pDevice->capture.internalPeriods*bpfDeviceCapture)) { + prevReadCursorInBytesCapture = 0; + } + } break; + + + + case ma_device_type_playback: + { + DWORD availableBytesPlayback; + DWORD physicalPlayCursorInBytes; + DWORD physicalWriteCursorInBytes; + hr = ma_IDirectSoundBuffer_GetCurrentPosition((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, &physicalPlayCursorInBytes, &physicalWriteCursorInBytes); + if (FAILED(hr)) { + break; + } + + hr = ma_IDirectSoundBuffer_GetStatus((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, &playbackBufferStatus); + if (SUCCEEDED(hr) && (playbackBufferStatus & MA_DSBSTATUS_PLAYING) == 0 && isPlaybackDeviceStarted) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[DirectSound] Attempting to resume audio due to state: %d.", (int)playbackBufferStatus); + hr = ma_IDirectSoundBuffer_Play((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, 0, 0, MA_DSBPLAY_LOOPING); + if (FAILED(hr)) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundBuffer_Play() failed after attempting to resume from state %d.", (int)playbackBufferStatus); + return ma_result_from_HRESULT(hr); + } + + isPlaybackDeviceStarted = MA_TRUE; + ma_sleep(waitTimeInMilliseconds); + continue; + } + + if (physicalPlayCursorInBytes < prevPlayCursorInBytesPlayback) { + physicalPlayCursorLoopFlagPlayback = !physicalPlayCursorLoopFlagPlayback; + } + prevPlayCursorInBytesPlayback = physicalPlayCursorInBytes; + + /* If there's any bytes available for writing we can do that now. The space between the virtual cursor position and play cursor. */ + if (physicalPlayCursorLoopFlagPlayback == virtualWriteCursorLoopFlagPlayback) { + /* Same loop iteration. The available bytes wraps all the way around from the virtual write cursor to the physical play cursor. */ + if (physicalPlayCursorInBytes <= virtualWriteCursorInBytesPlayback) { + availableBytesPlayback = (pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods*bpfDevicePlayback) - virtualWriteCursorInBytesPlayback; + availableBytesPlayback += physicalPlayCursorInBytes; /* Wrap around. */ + } else { + /* This is an error. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[DirectSound] (Playback): Play cursor has moved in front of the write cursor (same loop iterations). physicalPlayCursorInBytes=%ld, virtualWriteCursorInBytes=%ld.\n", physicalPlayCursorInBytes, virtualWriteCursorInBytesPlayback); + availableBytesPlayback = 0; + } + } else { + /* Different loop iterations. The available bytes only goes from the virtual write cursor to the physical play cursor. */ + if (physicalPlayCursorInBytes >= virtualWriteCursorInBytesPlayback) { + availableBytesPlayback = physicalPlayCursorInBytes - virtualWriteCursorInBytesPlayback; + } else { + /* This is an error. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[DirectSound] (Playback): Write cursor has moved behind the play cursor (different loop iterations). physicalPlayCursorInBytes=%ld, virtualWriteCursorInBytes=%ld.\n", physicalPlayCursorInBytes, virtualWriteCursorInBytesPlayback); + availableBytesPlayback = 0; + } + } + + /* If there's no room available for writing we need to wait for more. */ + if (availableBytesPlayback < pDevice->playback.internalPeriodSizeInFrames) { + /* If we haven't started the device yet, this will never get beyond 0. In this case we need to get the device started. */ + if (availableBytesPlayback == 0 && !isPlaybackDeviceStarted) { + hr = ma_IDirectSoundBuffer_Play((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, 0, 0, MA_DSBPLAY_LOOPING); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundBuffer_Play() failed."); + return ma_result_from_HRESULT(hr); + } + isPlaybackDeviceStarted = MA_TRUE; + } else { + ma_sleep(waitTimeInMilliseconds); + continue; + } + } + + /* Getting here means there room available somewhere. We limit this to either the end of the buffer or the physical play cursor, whichever is closest. */ + lockOffsetInBytesPlayback = virtualWriteCursorInBytesPlayback; + if (physicalPlayCursorLoopFlagPlayback == virtualWriteCursorLoopFlagPlayback) { + /* Same loop iteration. Go up to the end of the buffer. */ + lockSizeInBytesPlayback = (pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods*bpfDevicePlayback) - virtualWriteCursorInBytesPlayback; + } else { + /* Different loop iterations. Go up to the physical play cursor. */ + lockSizeInBytesPlayback = physicalPlayCursorInBytes - virtualWriteCursorInBytesPlayback; + } + + hr = ma_IDirectSoundBuffer_Lock((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, lockOffsetInBytesPlayback, lockSizeInBytesPlayback, &pMappedDeviceBufferPlayback, &mappedSizeInBytesPlayback, NULL, NULL, 0); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to map buffer from playback device in preparation for writing to the device."); + result = ma_result_from_HRESULT(hr); + break; + } + + /* At this point we have a buffer for output. */ + ma_device__read_frames_from_client(pDevice, (mappedSizeInBytesPlayback/bpfDevicePlayback), pMappedDeviceBufferPlayback); + + hr = ma_IDirectSoundBuffer_Unlock((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, pMappedDeviceBufferPlayback, mappedSizeInBytesPlayback, NULL, 0); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] Failed to unlock internal buffer from playback device after writing to the device."); + result = ma_result_from_HRESULT(hr); + break; + } + + virtualWriteCursorInBytesPlayback += mappedSizeInBytesPlayback; + if (virtualWriteCursorInBytesPlayback == pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods*bpfDevicePlayback) { + virtualWriteCursorInBytesPlayback = 0; + virtualWriteCursorLoopFlagPlayback = !virtualWriteCursorLoopFlagPlayback; + } + + /* + We may need to start the device. We want two full periods to be written before starting the playback device. Having an extra period adds + a bit of a buffer to prevent the playback buffer from getting starved. + */ + framesWrittenToPlaybackDevice += mappedSizeInBytesPlayback/bpfDevicePlayback; + if (!isPlaybackDeviceStarted && framesWrittenToPlaybackDevice >= pDevice->playback.internalPeriodSizeInFrames) { + hr = ma_IDirectSoundBuffer_Play((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, 0, 0, MA_DSBPLAY_LOOPING); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundBuffer_Play() failed."); + return ma_result_from_HRESULT(hr); + } + isPlaybackDeviceStarted = MA_TRUE; + } + } break; + + + default: return MA_INVALID_ARGS; /* Invalid device type. */ + } + + if (result != MA_SUCCESS) { + return result; + } + } + + /* Getting here means the device is being stopped. */ + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + hr = ma_IDirectSoundCaptureBuffer_Stop((ma_IDirectSoundCaptureBuffer*)pDevice->dsound.pCaptureBuffer); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundCaptureBuffer_Stop() failed."); + return ma_result_from_HRESULT(hr); + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + /* The playback device should be drained before stopping. All we do is wait until the available bytes is equal to the size of the buffer. */ + if (isPlaybackDeviceStarted) { + for (;;) { + DWORD availableBytesPlayback = 0; + DWORD physicalPlayCursorInBytes; + DWORD physicalWriteCursorInBytes; + hr = ma_IDirectSoundBuffer_GetCurrentPosition((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, &physicalPlayCursorInBytes, &physicalWriteCursorInBytes); + if (FAILED(hr)) { + break; + } + + if (physicalPlayCursorInBytes < prevPlayCursorInBytesPlayback) { + physicalPlayCursorLoopFlagPlayback = !physicalPlayCursorLoopFlagPlayback; + } + prevPlayCursorInBytesPlayback = physicalPlayCursorInBytes; + + if (physicalPlayCursorLoopFlagPlayback == virtualWriteCursorLoopFlagPlayback) { + /* Same loop iteration. The available bytes wraps all the way around from the virtual write cursor to the physical play cursor. */ + if (physicalPlayCursorInBytes <= virtualWriteCursorInBytesPlayback) { + availableBytesPlayback = (pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods*bpfDevicePlayback) - virtualWriteCursorInBytesPlayback; + availableBytesPlayback += physicalPlayCursorInBytes; /* Wrap around. */ + } else { + break; + } + } else { + /* Different loop iterations. The available bytes only goes from the virtual write cursor to the physical play cursor. */ + if (physicalPlayCursorInBytes >= virtualWriteCursorInBytesPlayback) { + availableBytesPlayback = physicalPlayCursorInBytes - virtualWriteCursorInBytesPlayback; + } else { + break; + } + } + + if (availableBytesPlayback >= (pDevice->playback.internalPeriodSizeInFrames*pDevice->playback.internalPeriods*bpfDevicePlayback)) { + break; + } + + ma_sleep(waitTimeInMilliseconds); + } + } + + hr = ma_IDirectSoundBuffer_Stop((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer); + if (FAILED(hr)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[DirectSound] IDirectSoundBuffer_Stop() failed."); + return ma_result_from_HRESULT(hr); + } + + ma_IDirectSoundBuffer_SetCurrentPosition((ma_IDirectSoundBuffer*)pDevice->dsound.pPlaybackBuffer, 0); + } + + return MA_SUCCESS; +} + +static ma_result ma_context_uninit__dsound(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_dsound); + + ma_dlclose(ma_context_get_log(pContext), pContext->dsound.hDSoundDLL); + + return MA_SUCCESS; +} + +static ma_result ma_context_init__dsound(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + MA_ASSERT(pContext != NULL); + + (void)pConfig; + + pContext->dsound.hDSoundDLL = ma_dlopen(ma_context_get_log(pContext), "dsound.dll"); + if (pContext->dsound.hDSoundDLL == NULL) { + return MA_API_NOT_FOUND; + } + + pContext->dsound.DirectSoundCreate = ma_dlsym(ma_context_get_log(pContext), pContext->dsound.hDSoundDLL, "DirectSoundCreate"); + pContext->dsound.DirectSoundEnumerateA = ma_dlsym(ma_context_get_log(pContext), pContext->dsound.hDSoundDLL, "DirectSoundEnumerateA"); + pContext->dsound.DirectSoundCaptureCreate = ma_dlsym(ma_context_get_log(pContext), pContext->dsound.hDSoundDLL, "DirectSoundCaptureCreate"); + pContext->dsound.DirectSoundCaptureEnumerateA = ma_dlsym(ma_context_get_log(pContext), pContext->dsound.hDSoundDLL, "DirectSoundCaptureEnumerateA"); + + /* + We need to support all functions or nothing. DirectSound with Windows 95 seems to not work too + well in my testing. For example, it's missing DirectSoundCaptureEnumerateA(). This is a convenient + place to just disable the DirectSound backend for Windows 95. + */ + if (pContext->dsound.DirectSoundCreate == NULL || + pContext->dsound.DirectSoundEnumerateA == NULL || + pContext->dsound.DirectSoundCaptureCreate == NULL || + pContext->dsound.DirectSoundCaptureEnumerateA == NULL) { + return MA_API_NOT_FOUND; + } + + pContext->dsound.hWnd = pConfig->dsound.hWnd; + + pCallbacks->onContextInit = ma_context_init__dsound; + pCallbacks->onContextUninit = ma_context_uninit__dsound; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__dsound; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__dsound; + pCallbacks->onDeviceInit = ma_device_init__dsound; + pCallbacks->onDeviceUninit = ma_device_uninit__dsound; + pCallbacks->onDeviceStart = NULL; /* Not used. Started in onDeviceDataLoop. */ + pCallbacks->onDeviceStop = NULL; /* Not used. Stopped in onDeviceDataLoop. */ + pCallbacks->onDeviceRead = NULL; /* Not used. Data is read directly in onDeviceDataLoop. */ + pCallbacks->onDeviceWrite = NULL; /* Not used. Data is written directly in onDeviceDataLoop. */ + pCallbacks->onDeviceDataLoop = ma_device_data_loop__dsound; + + return MA_SUCCESS; +} +#endif + + + +/****************************************************************************** + +WinMM Backend + +******************************************************************************/ +#ifdef MA_HAS_WINMM + +/* +Some build configurations will exclude the WinMM API. An example is when WIN32_LEAN_AND_MEAN +is defined. We need to define the types and functions we need manually. +*/ +#define MA_MMSYSERR_NOERROR 0 +#define MA_MMSYSERR_ERROR 1 +#define MA_MMSYSERR_BADDEVICEID 2 +#define MA_MMSYSERR_INVALHANDLE 5 +#define MA_MMSYSERR_NOMEM 7 +#define MA_MMSYSERR_INVALFLAG 10 +#define MA_MMSYSERR_INVALPARAM 11 +#define MA_MMSYSERR_HANDLEBUSY 12 + +#define MA_CALLBACK_EVENT 0x00050000 +#define MA_WAVE_ALLOWSYNC 0x0002 + +#define MA_WHDR_DONE 0x00000001 +#define MA_WHDR_PREPARED 0x00000002 +#define MA_WHDR_BEGINLOOP 0x00000004 +#define MA_WHDR_ENDLOOP 0x00000008 +#define MA_WHDR_INQUEUE 0x00000010 + +#define MA_MAXPNAMELEN 32 + +typedef void* MA_HWAVEIN; +typedef void* MA_HWAVEOUT; +typedef UINT MA_MMRESULT; +typedef UINT MA_MMVERSION; + +typedef struct +{ + WORD wMid; + WORD wPid; + MA_MMVERSION vDriverVersion; + CHAR szPname[MA_MAXPNAMELEN]; + DWORD dwFormats; + WORD wChannels; + WORD wReserved1; +} MA_WAVEINCAPSA; + +typedef struct +{ + WORD wMid; + WORD wPid; + MA_MMVERSION vDriverVersion; + CHAR szPname[MA_MAXPNAMELEN]; + DWORD dwFormats; + WORD wChannels; + WORD wReserved1; + DWORD dwSupport; +} MA_WAVEOUTCAPSA; + +typedef struct tagWAVEHDR +{ + char* lpData; + DWORD dwBufferLength; + DWORD dwBytesRecorded; + DWORD_PTR dwUser; + DWORD dwFlags; + DWORD dwLoops; + struct tagWAVEHDR* lpNext; + DWORD_PTR reserved; +} MA_WAVEHDR; + +typedef struct +{ + WORD wMid; + WORD wPid; + MA_MMVERSION vDriverVersion; + CHAR szPname[MA_MAXPNAMELEN]; + DWORD dwFormats; + WORD wChannels; + WORD wReserved1; + DWORD dwSupport; + GUID ManufacturerGuid; + GUID ProductGuid; + GUID NameGuid; +} MA_WAVEOUTCAPS2A; + +typedef struct +{ + WORD wMid; + WORD wPid; + MA_MMVERSION vDriverVersion; + CHAR szPname[MA_MAXPNAMELEN]; + DWORD dwFormats; + WORD wChannels; + WORD wReserved1; + GUID ManufacturerGuid; + GUID ProductGuid; + GUID NameGuid; +} MA_WAVEINCAPS2A; + +typedef UINT (WINAPI * MA_PFN_waveOutGetNumDevs)(void); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveOutGetDevCapsA)(ma_uintptr uDeviceID, MA_WAVEOUTCAPSA* pwoc, UINT cbwoc); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveOutOpen)(MA_HWAVEOUT* phwo, UINT uDeviceID, const MA_WAVEFORMATEX* pwfx, DWORD_PTR dwCallback, DWORD_PTR dwInstance, DWORD fdwOpen); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveOutClose)(MA_HWAVEOUT hwo); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveOutPrepareHeader)(MA_HWAVEOUT hwo, MA_WAVEHDR* pwh, UINT cbwh); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveOutUnprepareHeader)(MA_HWAVEOUT hwo, MA_WAVEHDR* pwh, UINT cbwh); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveOutWrite)(MA_HWAVEOUT hwo, MA_WAVEHDR* pwh, UINT cbwh); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveOutReset)(MA_HWAVEOUT hwo); +typedef UINT (WINAPI * MA_PFN_waveInGetNumDevs)(void); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveInGetDevCapsA)(ma_uintptr uDeviceID, MA_WAVEINCAPSA* pwic, UINT cbwic); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveInOpen)(MA_HWAVEIN* phwi, UINT uDeviceID, const MA_WAVEFORMATEX* pwfx, DWORD_PTR dwCallback, DWORD_PTR dwInstance, DWORD fdwOpen); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveInClose)(MA_HWAVEIN hwi); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveInPrepareHeader)(MA_HWAVEIN hwi, MA_WAVEHDR* pwh, UINT cbwh); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveInUnprepareHeader)(MA_HWAVEIN hwi, MA_WAVEHDR* pwh, UINT cbwh); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveInAddBuffer)(MA_HWAVEIN hwi, MA_WAVEHDR* pwh, UINT cbwh); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveInStart)(MA_HWAVEIN hwi); +typedef MA_MMRESULT (WINAPI * MA_PFN_waveInReset)(MA_HWAVEIN hwi); + +static ma_result ma_result_from_MMRESULT(MA_MMRESULT resultMM) +{ + switch (resultMM) + { + case MA_MMSYSERR_NOERROR: return MA_SUCCESS; + case MA_MMSYSERR_BADDEVICEID: return MA_INVALID_ARGS; + case MA_MMSYSERR_INVALHANDLE: return MA_INVALID_ARGS; + case MA_MMSYSERR_NOMEM: return MA_OUT_OF_MEMORY; + case MA_MMSYSERR_INVALFLAG: return MA_INVALID_ARGS; + case MA_MMSYSERR_INVALPARAM: return MA_INVALID_ARGS; + case MA_MMSYSERR_HANDLEBUSY: return MA_BUSY; + case MA_MMSYSERR_ERROR: return MA_ERROR; + default: return MA_ERROR; + } +} + +static char* ma_find_last_character(char* str, char ch) +{ + char* last; + + if (str == NULL) { + return NULL; + } + + last = NULL; + while (*str != '\0') { + if (*str == ch) { + last = str; + } + + str += 1; + } + + return last; +} + +static ma_uint32 ma_get_period_size_in_bytes(ma_uint32 periodSizeInFrames, ma_format format, ma_uint32 channels) +{ + return periodSizeInFrames * ma_get_bytes_per_frame(format, channels); +} + + +/* +Our own "WAVECAPS" structure that contains generic information shared between WAVEOUTCAPS2 and WAVEINCAPS2 so +we can do things generically and typesafely. Names are being kept the same for consistency. +*/ +typedef struct +{ + CHAR szPname[MA_MAXPNAMELEN]; + DWORD dwFormats; + WORD wChannels; + GUID NameGuid; +} MA_WAVECAPSA; + +static ma_result ma_get_best_info_from_formats_flags__winmm(DWORD dwFormats, WORD channels, WORD* pBitsPerSample, DWORD* pSampleRate) +{ + WORD bitsPerSample = 0; + DWORD sampleRate = 0; + + if (pBitsPerSample) { + *pBitsPerSample = 0; + } + if (pSampleRate) { + *pSampleRate = 0; + } + + if (channels == 1) { + bitsPerSample = 16; + if ((dwFormats & WAVE_FORMAT_48M16) != 0) { + sampleRate = 48000; + } else if ((dwFormats & WAVE_FORMAT_44M16) != 0) { + sampleRate = 44100; + } else if ((dwFormats & WAVE_FORMAT_2M16) != 0) { + sampleRate = 22050; + } else if ((dwFormats & WAVE_FORMAT_1M16) != 0) { + sampleRate = 11025; + } else if ((dwFormats & WAVE_FORMAT_96M16) != 0) { + sampleRate = 96000; + } else { + bitsPerSample = 8; + if ((dwFormats & WAVE_FORMAT_48M08) != 0) { + sampleRate = 48000; + } else if ((dwFormats & WAVE_FORMAT_44M08) != 0) { + sampleRate = 44100; + } else if ((dwFormats & WAVE_FORMAT_2M08) != 0) { + sampleRate = 22050; + } else if ((dwFormats & WAVE_FORMAT_1M08) != 0) { + sampleRate = 11025; + } else if ((dwFormats & WAVE_FORMAT_96M08) != 0) { + sampleRate = 96000; + } else { + return MA_FORMAT_NOT_SUPPORTED; + } + } + } else { + bitsPerSample = 16; + if ((dwFormats & WAVE_FORMAT_48S16) != 0) { + sampleRate = 48000; + } else if ((dwFormats & WAVE_FORMAT_44S16) != 0) { + sampleRate = 44100; + } else if ((dwFormats & WAVE_FORMAT_2S16) != 0) { + sampleRate = 22050; + } else if ((dwFormats & WAVE_FORMAT_1S16) != 0) { + sampleRate = 11025; + } else if ((dwFormats & WAVE_FORMAT_96S16) != 0) { + sampleRate = 96000; + } else { + bitsPerSample = 8; + if ((dwFormats & WAVE_FORMAT_48S08) != 0) { + sampleRate = 48000; + } else if ((dwFormats & WAVE_FORMAT_44S08) != 0) { + sampleRate = 44100; + } else if ((dwFormats & WAVE_FORMAT_2S08) != 0) { + sampleRate = 22050; + } else if ((dwFormats & WAVE_FORMAT_1S08) != 0) { + sampleRate = 11025; + } else if ((dwFormats & WAVE_FORMAT_96S08) != 0) { + sampleRate = 96000; + } else { + return MA_FORMAT_NOT_SUPPORTED; + } + } + } + + if (pBitsPerSample) { + *pBitsPerSample = bitsPerSample; + } + if (pSampleRate) { + *pSampleRate = sampleRate; + } + + return MA_SUCCESS; +} + +static ma_result ma_formats_flags_to_WAVEFORMATEX__winmm(DWORD dwFormats, WORD channels, MA_WAVEFORMATEX* pWF) +{ + ma_result result; + + MA_ASSERT(pWF != NULL); + + MA_ZERO_OBJECT(pWF); + pWF->cbSize = sizeof(*pWF); + pWF->wFormatTag = WAVE_FORMAT_PCM; + pWF->nChannels = (WORD)channels; + if (pWF->nChannels > 2) { + pWF->nChannels = 2; + } + + result = ma_get_best_info_from_formats_flags__winmm(dwFormats, channels, &pWF->wBitsPerSample, &pWF->nSamplesPerSec); + if (result != MA_SUCCESS) { + return result; + } + + pWF->nBlockAlign = (WORD)(pWF->nChannels * pWF->wBitsPerSample / 8); + pWF->nAvgBytesPerSec = pWF->nBlockAlign * pWF->nSamplesPerSec; + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info_from_WAVECAPS(ma_context* pContext, MA_WAVECAPSA* pCaps, ma_device_info* pDeviceInfo) +{ + WORD bitsPerSample; + DWORD sampleRate; + ma_result result; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pCaps != NULL); + MA_ASSERT(pDeviceInfo != NULL); + + /* + Name / Description + + Unfortunately the name specified in WAVE(OUT/IN)CAPS2 is limited to 31 characters. This results in an unprofessional looking + situation where the names of the devices are truncated. To help work around this, we need to look at the name GUID and try + looking in the registry for the full name. If we can't find it there, we need to just fall back to the default name. + */ + + /* Set the default to begin with. */ + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), pCaps->szPname, (size_t)-1); + + /* + Now try the registry. There's a few things to consider here: + - The name GUID can be null, in which we case we just need to stick to the original 31 characters. + - If the name GUID is not present in the registry we'll also need to stick to the original 31 characters. + - I like consistency, so I want the returned device names to be consistent with those returned by WASAPI and DirectSound. The + problem, however is that WASAPI and DirectSound use " ()" format (such as "Speakers (High Definition Audio)"), + but WinMM does not specify the component name. From my admittedly limited testing, I've notice the component name seems to + usually fit within the 31 characters of the fixed sized buffer, so what I'm going to do is parse that string for the component + name, and then concatenate the name from the registry. + */ + if (!ma_is_guid_null(&pCaps->NameGuid)) { + WCHAR guidStrW[256]; + if (((MA_PFN_StringFromGUID2)pContext->win32.StringFromGUID2)(&pCaps->NameGuid, guidStrW, ma_countof(guidStrW)) > 0) { + char guidStr[256]; + char keyStr[1024]; + HKEY hKey; + + WideCharToMultiByte(CP_UTF8, 0, guidStrW, -1, guidStr, sizeof(guidStr), 0, FALSE); + + ma_strcpy_s(keyStr, sizeof(keyStr), "SYSTEM\\CurrentControlSet\\Control\\MediaCategories\\"); + ma_strcat_s(keyStr, sizeof(keyStr), guidStr); + + if (((MA_PFN_RegOpenKeyExA)pContext->win32.RegOpenKeyExA)(HKEY_LOCAL_MACHINE, keyStr, 0, KEY_READ, &hKey) == ERROR_SUCCESS) { + BYTE nameFromReg[512]; + DWORD nameFromRegSize = sizeof(nameFromReg); + LONG resultWin32 = ((MA_PFN_RegQueryValueExA)pContext->win32.RegQueryValueExA)(hKey, "Name", 0, NULL, (BYTE*)nameFromReg, (DWORD*)&nameFromRegSize); + ((MA_PFN_RegCloseKey)pContext->win32.RegCloseKey)(hKey); + + if (resultWin32 == ERROR_SUCCESS) { + /* We have the value from the registry, so now we need to construct the name string. */ + char name[1024]; + if (ma_strcpy_s(name, sizeof(name), pDeviceInfo->name) == 0) { + char* nameBeg = ma_find_last_character(name, '('); + if (nameBeg != NULL) { + size_t leadingLen = (nameBeg - name); + ma_strncpy_s(nameBeg + 1, sizeof(name) - leadingLen, (const char*)nameFromReg, (size_t)-1); + + /* The closing ")", if it can fit. */ + if (leadingLen + nameFromRegSize < sizeof(name)-1) { + ma_strcat_s(name, sizeof(name), ")"); + } + + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), name, (size_t)-1); + } + } + } + } + } + } + + + result = ma_get_best_info_from_formats_flags__winmm(pCaps->dwFormats, pCaps->wChannels, &bitsPerSample, &sampleRate); + if (result != MA_SUCCESS) { + return result; + } + + if (bitsPerSample == 8) { + pDeviceInfo->nativeDataFormats[0].format = ma_format_u8; + } else if (bitsPerSample == 16) { + pDeviceInfo->nativeDataFormats[0].format = ma_format_s16; + } else if (bitsPerSample == 24) { + pDeviceInfo->nativeDataFormats[0].format = ma_format_s24; + } else if (bitsPerSample == 32) { + pDeviceInfo->nativeDataFormats[0].format = ma_format_s32; + } else { + return MA_FORMAT_NOT_SUPPORTED; + } + pDeviceInfo->nativeDataFormats[0].channels = pCaps->wChannels; + pDeviceInfo->nativeDataFormats[0].sampleRate = sampleRate; + pDeviceInfo->nativeDataFormats[0].flags = 0; + pDeviceInfo->nativeDataFormatCount = 1; + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info_from_WAVEOUTCAPS2(ma_context* pContext, MA_WAVEOUTCAPS2A* pCaps, ma_device_info* pDeviceInfo) +{ + MA_WAVECAPSA caps; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pCaps != NULL); + MA_ASSERT(pDeviceInfo != NULL); + + MA_COPY_MEMORY(caps.szPname, pCaps->szPname, sizeof(caps.szPname)); + caps.dwFormats = pCaps->dwFormats; + caps.wChannels = pCaps->wChannels; + caps.NameGuid = pCaps->NameGuid; + return ma_context_get_device_info_from_WAVECAPS(pContext, &caps, pDeviceInfo); +} + +static ma_result ma_context_get_device_info_from_WAVEINCAPS2(ma_context* pContext, MA_WAVEINCAPS2A* pCaps, ma_device_info* pDeviceInfo) +{ + MA_WAVECAPSA caps; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pCaps != NULL); + MA_ASSERT(pDeviceInfo != NULL); + + MA_COPY_MEMORY(caps.szPname, pCaps->szPname, sizeof(caps.szPname)); + caps.dwFormats = pCaps->dwFormats; + caps.wChannels = pCaps->wChannels; + caps.NameGuid = pCaps->NameGuid; + return ma_context_get_device_info_from_WAVECAPS(pContext, &caps, pDeviceInfo); +} + + +static ma_result ma_context_enumerate_devices__winmm(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + UINT playbackDeviceCount; + UINT captureDeviceCount; + UINT iPlaybackDevice; + UINT iCaptureDevice; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + /* Playback. */ + playbackDeviceCount = ((MA_PFN_waveOutGetNumDevs)pContext->winmm.waveOutGetNumDevs)(); + for (iPlaybackDevice = 0; iPlaybackDevice < playbackDeviceCount; ++iPlaybackDevice) { + MA_MMRESULT result; + MA_WAVEOUTCAPS2A caps; + + MA_ZERO_OBJECT(&caps); + + result = ((MA_PFN_waveOutGetDevCapsA)pContext->winmm.waveOutGetDevCapsA)(iPlaybackDevice, (MA_WAVEOUTCAPSA*)&caps, sizeof(caps)); + if (result == MA_MMSYSERR_NOERROR) { + ma_device_info deviceInfo; + + MA_ZERO_OBJECT(&deviceInfo); + deviceInfo.id.winmm = iPlaybackDevice; + + /* The first enumerated device is the default device. */ + if (iPlaybackDevice == 0) { + deviceInfo.isDefault = MA_TRUE; + } + + if (ma_context_get_device_info_from_WAVEOUTCAPS2(pContext, &caps, &deviceInfo) == MA_SUCCESS) { + ma_bool32 cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + if (cbResult == MA_FALSE) { + return MA_SUCCESS; /* Enumeration was stopped. */ + } + } + } + } + + /* Capture. */ + captureDeviceCount = ((MA_PFN_waveInGetNumDevs)pContext->winmm.waveInGetNumDevs)(); + for (iCaptureDevice = 0; iCaptureDevice < captureDeviceCount; ++iCaptureDevice) { + MA_MMRESULT result; + MA_WAVEINCAPS2A caps; + + MA_ZERO_OBJECT(&caps); + + result = ((MA_PFN_waveInGetDevCapsA)pContext->winmm.waveInGetDevCapsA)(iCaptureDevice, (MA_WAVEINCAPSA*)&caps, sizeof(caps)); + if (result == MA_MMSYSERR_NOERROR) { + ma_device_info deviceInfo; + + MA_ZERO_OBJECT(&deviceInfo); + deviceInfo.id.winmm = iCaptureDevice; + + /* The first enumerated device is the default device. */ + if (iCaptureDevice == 0) { + deviceInfo.isDefault = MA_TRUE; + } + + if (ma_context_get_device_info_from_WAVEINCAPS2(pContext, &caps, &deviceInfo) == MA_SUCCESS) { + ma_bool32 cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + if (cbResult == MA_FALSE) { + return MA_SUCCESS; /* Enumeration was stopped. */ + } + } + } + } + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info__winmm(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + UINT winMMDeviceID; + + MA_ASSERT(pContext != NULL); + + winMMDeviceID = 0; + if (pDeviceID != NULL) { + winMMDeviceID = (UINT)pDeviceID->winmm; + } + + pDeviceInfo->id.winmm = winMMDeviceID; + + /* The first ID is the default device. */ + if (winMMDeviceID == 0) { + pDeviceInfo->isDefault = MA_TRUE; + } + + if (deviceType == ma_device_type_playback) { + MA_MMRESULT result; + MA_WAVEOUTCAPS2A caps; + + MA_ZERO_OBJECT(&caps); + + result = ((MA_PFN_waveOutGetDevCapsA)pContext->winmm.waveOutGetDevCapsA)(winMMDeviceID, (MA_WAVEOUTCAPSA*)&caps, sizeof(caps)); + if (result == MA_MMSYSERR_NOERROR) { + return ma_context_get_device_info_from_WAVEOUTCAPS2(pContext, &caps, pDeviceInfo); + } + } else { + MA_MMRESULT result; + MA_WAVEINCAPS2A caps; + + MA_ZERO_OBJECT(&caps); + + result = ((MA_PFN_waveInGetDevCapsA)pContext->winmm.waveInGetDevCapsA)(winMMDeviceID, (MA_WAVEINCAPSA*)&caps, sizeof(caps)); + if (result == MA_MMSYSERR_NOERROR) { + return ma_context_get_device_info_from_WAVEINCAPS2(pContext, &caps, pDeviceInfo); + } + } + + return MA_NO_DEVICE; +} + + +static ma_result ma_device_uninit__winmm(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ((MA_PFN_waveInClose)pDevice->pContext->winmm.waveInClose)((MA_HWAVEIN)pDevice->winmm.hDeviceCapture); + CloseHandle((HANDLE)pDevice->winmm.hEventCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ((MA_PFN_waveOutReset)pDevice->pContext->winmm.waveOutReset)((MA_HWAVEOUT)pDevice->winmm.hDevicePlayback); + ((MA_PFN_waveOutClose)pDevice->pContext->winmm.waveOutClose)((MA_HWAVEOUT)pDevice->winmm.hDevicePlayback); + CloseHandle((HANDLE)pDevice->winmm.hEventPlayback); + } + + ma_free(pDevice->winmm._pHeapData, &pDevice->pContext->allocationCallbacks); + + MA_ZERO_OBJECT(&pDevice->winmm); /* Safety. */ + + return MA_SUCCESS; +} + +static ma_uint32 ma_calculate_period_size_in_frames_from_descriptor__winmm(const ma_device_descriptor* pDescriptor, ma_uint32 nativeSampleRate, ma_performance_profile performanceProfile) +{ + /* WinMM has a minimum period size of 40ms. */ + ma_uint32 minPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(40, nativeSampleRate); + ma_uint32 periodSizeInFrames; + + periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptor, nativeSampleRate, performanceProfile); + if (periodSizeInFrames < minPeriodSizeInFrames) { + periodSizeInFrames = minPeriodSizeInFrames; + } + + return periodSizeInFrames; +} + +static ma_result ma_device_init__winmm(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + const char* errorMsg = ""; + ma_result errorCode = MA_ERROR; + ma_result result = MA_SUCCESS; + ma_uint32 heapSize; + UINT winMMDeviceIDPlayback = 0; + UINT winMMDeviceIDCapture = 0; + + MA_ASSERT(pDevice != NULL); + + MA_ZERO_OBJECT(&pDevice->winmm); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + /* No exclusive mode with WinMM. */ + if (((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pDescriptorPlayback->shareMode == ma_share_mode_exclusive) || + ((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) && pDescriptorCapture->shareMode == ma_share_mode_exclusive)) { + return MA_SHARE_MODE_NOT_SUPPORTED; + } + + if (pDescriptorPlayback->pDeviceID != NULL) { + winMMDeviceIDPlayback = (UINT)pDescriptorPlayback->pDeviceID->winmm; + } + if (pDescriptorCapture->pDeviceID != NULL) { + winMMDeviceIDCapture = (UINT)pDescriptorCapture->pDeviceID->winmm; + } + + /* The capture device needs to be initialized first. */ + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + MA_WAVEINCAPSA caps; + MA_WAVEFORMATEX wf; + MA_MMRESULT resultMM; + + /* We use an event to know when a new fragment needs to be enqueued. */ + pDevice->winmm.hEventCapture = (ma_handle)CreateEventA(NULL, TRUE, TRUE, NULL); + if (pDevice->winmm.hEventCapture == NULL) { + errorMsg = "[WinMM] Failed to create event for fragment enqueuing for the capture device.", errorCode = ma_result_from_GetLastError(GetLastError()); + goto on_error; + } + + /* The format should be based on the device's actual format. */ + if (((MA_PFN_waveInGetDevCapsA)pDevice->pContext->winmm.waveInGetDevCapsA)(winMMDeviceIDCapture, &caps, sizeof(caps)) != MA_MMSYSERR_NOERROR) { + errorMsg = "[WinMM] Failed to retrieve internal device caps.", errorCode = MA_FORMAT_NOT_SUPPORTED; + goto on_error; + } + + result = ma_formats_flags_to_WAVEFORMATEX__winmm(caps.dwFormats, caps.wChannels, &wf); + if (result != MA_SUCCESS) { + errorMsg = "[WinMM] Could not find appropriate format for internal device.", errorCode = result; + goto on_error; + } + + resultMM = ((MA_PFN_waveInOpen)pDevice->pContext->winmm.waveInOpen)((MA_HWAVEIN*)&pDevice->winmm.hDeviceCapture, winMMDeviceIDCapture, &wf, (DWORD_PTR)pDevice->winmm.hEventCapture, (DWORD_PTR)pDevice, MA_CALLBACK_EVENT | MA_WAVE_ALLOWSYNC); + if (resultMM != MA_MMSYSERR_NOERROR) { + errorMsg = "[WinMM] Failed to open capture device.", errorCode = MA_FAILED_TO_OPEN_BACKEND_DEVICE; + goto on_error; + } + + pDescriptorCapture->format = ma_format_from_WAVEFORMATEX(&wf); + pDescriptorCapture->channels = wf.nChannels; + pDescriptorCapture->sampleRate = wf.nSamplesPerSec; + ma_channel_map_init_standard(ma_standard_channel_map_microsoft, pDescriptorCapture->channelMap, ma_countof(pDescriptorCapture->channelMap), pDescriptorCapture->channels); + pDescriptorCapture->periodCount = pDescriptorCapture->periodCount; + pDescriptorCapture->periodSizeInFrames = ma_calculate_period_size_in_frames_from_descriptor__winmm(pDescriptorCapture, pDescriptorCapture->sampleRate, pConfig->performanceProfile); + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + MA_WAVEOUTCAPSA caps; + MA_WAVEFORMATEX wf; + MA_MMRESULT resultMM; + + /* We use an event to know when a new fragment needs to be enqueued. */ + pDevice->winmm.hEventPlayback = (ma_handle)CreateEventA(NULL, TRUE, TRUE, NULL); + if (pDevice->winmm.hEventPlayback == NULL) { + errorMsg = "[WinMM] Failed to create event for fragment enqueuing for the playback device.", errorCode = ma_result_from_GetLastError(GetLastError()); + goto on_error; + } + + /* The format should be based on the device's actual format. */ + if (((MA_PFN_waveOutGetDevCapsA)pDevice->pContext->winmm.waveOutGetDevCapsA)(winMMDeviceIDPlayback, &caps, sizeof(caps)) != MA_MMSYSERR_NOERROR) { + errorMsg = "[WinMM] Failed to retrieve internal device caps.", errorCode = MA_FORMAT_NOT_SUPPORTED; + goto on_error; + } + + result = ma_formats_flags_to_WAVEFORMATEX__winmm(caps.dwFormats, caps.wChannels, &wf); + if (result != MA_SUCCESS) { + errorMsg = "[WinMM] Could not find appropriate format for internal device.", errorCode = result; + goto on_error; + } + + resultMM = ((MA_PFN_waveOutOpen)pDevice->pContext->winmm.waveOutOpen)((MA_HWAVEOUT*)&pDevice->winmm.hDevicePlayback, winMMDeviceIDPlayback, &wf, (DWORD_PTR)pDevice->winmm.hEventPlayback, (DWORD_PTR)pDevice, MA_CALLBACK_EVENT | MA_WAVE_ALLOWSYNC); + if (resultMM != MA_MMSYSERR_NOERROR) { + errorMsg = "[WinMM] Failed to open playback device.", errorCode = MA_FAILED_TO_OPEN_BACKEND_DEVICE; + goto on_error; + } + + pDescriptorPlayback->format = ma_format_from_WAVEFORMATEX(&wf); + pDescriptorPlayback->channels = wf.nChannels; + pDescriptorPlayback->sampleRate = wf.nSamplesPerSec; + ma_channel_map_init_standard(ma_standard_channel_map_microsoft, pDescriptorPlayback->channelMap, ma_countof(pDescriptorPlayback->channelMap), pDescriptorPlayback->channels); + pDescriptorPlayback->periodCount = pDescriptorPlayback->periodCount; + pDescriptorPlayback->periodSizeInFrames = ma_calculate_period_size_in_frames_from_descriptor__winmm(pDescriptorPlayback, pDescriptorPlayback->sampleRate, pConfig->performanceProfile); + } + + /* + The heap allocated data is allocated like so: + + [Capture WAVEHDRs][Playback WAVEHDRs][Capture Intermediary Buffer][Playback Intermediary Buffer] + */ + heapSize = 0; + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + heapSize += sizeof(MA_WAVEHDR)*pDescriptorCapture->periodCount + (pDescriptorCapture->periodSizeInFrames * pDescriptorCapture->periodCount * ma_get_bytes_per_frame(pDescriptorCapture->format, pDescriptorCapture->channels)); + } + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + heapSize += sizeof(MA_WAVEHDR)*pDescriptorPlayback->periodCount + (pDescriptorPlayback->periodSizeInFrames * pDescriptorPlayback->periodCount * ma_get_bytes_per_frame(pDescriptorPlayback->format, pDescriptorPlayback->channels)); + } + + pDevice->winmm._pHeapData = (ma_uint8*)ma_calloc(heapSize, &pDevice->pContext->allocationCallbacks); + if (pDevice->winmm._pHeapData == NULL) { + errorMsg = "[WinMM] Failed to allocate memory for the intermediary buffer.", errorCode = MA_OUT_OF_MEMORY; + goto on_error; + } + + MA_ZERO_MEMORY(pDevice->winmm._pHeapData, heapSize); + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ma_uint32 iPeriod; + + if (pConfig->deviceType == ma_device_type_capture) { + pDevice->winmm.pWAVEHDRCapture = pDevice->winmm._pHeapData; + pDevice->winmm.pIntermediaryBufferCapture = pDevice->winmm._pHeapData + (sizeof(MA_WAVEHDR)*(pDescriptorCapture->periodCount)); + } else { + pDevice->winmm.pWAVEHDRCapture = pDevice->winmm._pHeapData; + pDevice->winmm.pIntermediaryBufferCapture = pDevice->winmm._pHeapData + (sizeof(MA_WAVEHDR)*(pDescriptorCapture->periodCount + pDescriptorPlayback->periodCount)); + } + + /* Prepare headers. */ + for (iPeriod = 0; iPeriod < pDescriptorCapture->periodCount; ++iPeriod) { + ma_uint32 periodSizeInBytes = ma_get_period_size_in_bytes(pDescriptorCapture->periodSizeInFrames, pDescriptorCapture->format, pDescriptorCapture->channels); + + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[iPeriod].lpData = (char*)(pDevice->winmm.pIntermediaryBufferCapture + (periodSizeInBytes*iPeriod)); + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[iPeriod].dwBufferLength = periodSizeInBytes; + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[iPeriod].dwFlags = 0L; + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[iPeriod].dwLoops = 0L; + ((MA_PFN_waveInPrepareHeader)pDevice->pContext->winmm.waveInPrepareHeader)((MA_HWAVEIN)pDevice->winmm.hDeviceCapture, &((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[iPeriod], sizeof(MA_WAVEHDR)); + + /* + The user data of the MA_WAVEHDR structure is a single flag the controls whether or not it is ready for writing. Consider it to be named "isLocked". A value of 0 means + it's unlocked and available for writing. A value of 1 means it's locked. + */ + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[iPeriod].dwUser = 0; + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_uint32 iPeriod; + + if (pConfig->deviceType == ma_device_type_playback) { + pDevice->winmm.pWAVEHDRPlayback = pDevice->winmm._pHeapData; + pDevice->winmm.pIntermediaryBufferPlayback = pDevice->winmm._pHeapData + (sizeof(MA_WAVEHDR)*pDescriptorPlayback->periodCount); + } else { + pDevice->winmm.pWAVEHDRPlayback = pDevice->winmm._pHeapData + (sizeof(MA_WAVEHDR)*(pDescriptorCapture->periodCount)); + pDevice->winmm.pIntermediaryBufferPlayback = pDevice->winmm._pHeapData + (sizeof(MA_WAVEHDR)*(pDescriptorCapture->periodCount + pDescriptorPlayback->periodCount)) + (pDescriptorCapture->periodSizeInFrames*pDescriptorCapture->periodCount*ma_get_bytes_per_frame(pDescriptorCapture->format, pDescriptorCapture->channels)); + } + + /* Prepare headers. */ + for (iPeriod = 0; iPeriod < pDescriptorPlayback->periodCount; ++iPeriod) { + ma_uint32 periodSizeInBytes = ma_get_period_size_in_bytes(pDescriptorPlayback->periodSizeInFrames, pDescriptorPlayback->format, pDescriptorPlayback->channels); + + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback)[iPeriod].lpData = (char*)(pDevice->winmm.pIntermediaryBufferPlayback + (periodSizeInBytes*iPeriod)); + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback)[iPeriod].dwBufferLength = periodSizeInBytes; + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback)[iPeriod].dwFlags = 0L; + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback)[iPeriod].dwLoops = 0L; + ((MA_PFN_waveOutPrepareHeader)pDevice->pContext->winmm.waveOutPrepareHeader)((MA_HWAVEOUT)pDevice->winmm.hDevicePlayback, &((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback)[iPeriod], sizeof(MA_WAVEHDR)); + + /* + The user data of the MA_WAVEHDR structure is a single flag the controls whether or not it is ready for writing. Consider it to be named "isLocked". A value of 0 means + it's unlocked and available for writing. A value of 1 means it's locked. + */ + ((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback)[iPeriod].dwUser = 0; + } + } + + return MA_SUCCESS; + +on_error: + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + if (pDevice->winmm.pWAVEHDRCapture != NULL) { + ma_uint32 iPeriod; + for (iPeriod = 0; iPeriod < pDescriptorCapture->periodCount; ++iPeriod) { + ((MA_PFN_waveInUnprepareHeader)pDevice->pContext->winmm.waveInUnprepareHeader)((MA_HWAVEIN)pDevice->winmm.hDeviceCapture, &((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[iPeriod], sizeof(MA_WAVEHDR)); + } + } + + ((MA_PFN_waveInClose)pDevice->pContext->winmm.waveInClose)((MA_HWAVEIN)pDevice->winmm.hDeviceCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + if (pDevice->winmm.pWAVEHDRCapture != NULL) { + ma_uint32 iPeriod; + for (iPeriod = 0; iPeriod < pDescriptorPlayback->periodCount; ++iPeriod) { + ((MA_PFN_waveOutUnprepareHeader)pDevice->pContext->winmm.waveOutUnprepareHeader)((MA_HWAVEOUT)pDevice->winmm.hDevicePlayback, &((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback)[iPeriod], sizeof(MA_WAVEHDR)); + } + } + + ((MA_PFN_waveOutClose)pDevice->pContext->winmm.waveOutClose)((MA_HWAVEOUT)pDevice->winmm.hDevicePlayback); + } + + ma_free(pDevice->winmm._pHeapData, &pDevice->pContext->allocationCallbacks); + + if (errorMsg != NULL && errorMsg[0] != '\0') { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "%s", errorMsg); + } + + return errorCode; +} + +static ma_result ma_device_start__winmm(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + MA_MMRESULT resultMM; + MA_WAVEHDR* pWAVEHDR; + ma_uint32 iPeriod; + + pWAVEHDR = (MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture; + + /* Make sure the event is reset to a non-signaled state to ensure we don't prematurely return from WaitForSingleObject(). */ + ResetEvent((HANDLE)pDevice->winmm.hEventCapture); + + /* To start the device we attach all of the buffers and then start it. As the buffers are filled with data we will get notifications. */ + for (iPeriod = 0; iPeriod < pDevice->capture.internalPeriods; ++iPeriod) { + resultMM = ((MA_PFN_waveInAddBuffer)pDevice->pContext->winmm.waveInAddBuffer)((MA_HWAVEIN)pDevice->winmm.hDeviceCapture, &((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[iPeriod], sizeof(MA_WAVEHDR)); + if (resultMM != MA_MMSYSERR_NOERROR) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WinMM] Failed to attach input buffers to capture device in preparation for capture."); + return ma_result_from_MMRESULT(resultMM); + } + + /* Make sure all of the buffers start out locked. We don't want to access them until the backend tells us we can. */ + pWAVEHDR[iPeriod].dwUser = 1; /* 1 = locked. */ + } + + /* Capture devices need to be explicitly started, unlike playback devices. */ + resultMM = ((MA_PFN_waveInStart)pDevice->pContext->winmm.waveInStart)((MA_HWAVEIN)pDevice->winmm.hDeviceCapture); + if (resultMM != MA_MMSYSERR_NOERROR) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WinMM] Failed to start backend device."); + return ma_result_from_MMRESULT(resultMM); + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + /* Don't need to do anything for playback. It'll be started automatically in ma_device_start__winmm(). */ + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__winmm(ma_device* pDevice) +{ + MA_MMRESULT resultMM; + + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + if (pDevice->winmm.hDeviceCapture == NULL) { + return MA_INVALID_ARGS; + } + + resultMM = ((MA_PFN_waveInReset)pDevice->pContext->winmm.waveInReset)((MA_HWAVEIN)pDevice->winmm.hDeviceCapture); + if (resultMM != MA_MMSYSERR_NOERROR) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[WinMM] WARNING: Failed to reset capture device."); + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_uint32 iPeriod; + MA_WAVEHDR* pWAVEHDR; + + if (pDevice->winmm.hDevicePlayback == NULL) { + return MA_INVALID_ARGS; + } + + /* We need to drain the device. To do this we just loop over each header and if it's locked just wait for the event. */ + pWAVEHDR = (MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback; + for (iPeriod = 0; iPeriod < pDevice->playback.internalPeriods; iPeriod += 1) { + if (pWAVEHDR[iPeriod].dwUser == 1) { /* 1 = locked. */ + if (WaitForSingleObject((HANDLE)pDevice->winmm.hEventPlayback, INFINITE) != WAIT_OBJECT_0) { + break; /* An error occurred so just abandon ship and stop the device without draining. */ + } + + pWAVEHDR[iPeriod].dwUser = 0; + } + } + + resultMM = ((MA_PFN_waveOutReset)pDevice->pContext->winmm.waveOutReset)((MA_HWAVEOUT)pDevice->winmm.hDevicePlayback); + if (resultMM != MA_MMSYSERR_NOERROR) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[WinMM] WARNING: Failed to reset playback device."); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_write__winmm(ma_device* pDevice, const void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesWritten) +{ + ma_result result = MA_SUCCESS; + MA_MMRESULT resultMM; + ma_uint32 totalFramesWritten; + MA_WAVEHDR* pWAVEHDR; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pPCMFrames != NULL); + + if (pFramesWritten != NULL) { + *pFramesWritten = 0; + } + + pWAVEHDR = (MA_WAVEHDR*)pDevice->winmm.pWAVEHDRPlayback; + + /* Keep processing as much data as possible. */ + totalFramesWritten = 0; + while (totalFramesWritten < frameCount) { + /* If the current header has some space available we need to write part of it. */ + if (pWAVEHDR[pDevice->winmm.iNextHeaderPlayback].dwUser == 0) { /* 0 = unlocked. */ + /* + This header has room in it. We copy as much of it as we can. If we end up fully consuming the buffer we need to + write it out and move on to the next iteration. + */ + ma_uint32 bpf = ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + ma_uint32 framesRemainingInHeader = (pWAVEHDR[pDevice->winmm.iNextHeaderPlayback].dwBufferLength/bpf) - pDevice->winmm.headerFramesConsumedPlayback; + + ma_uint32 framesToCopy = ma_min(framesRemainingInHeader, (frameCount - totalFramesWritten)); + const void* pSrc = ma_offset_ptr(pPCMFrames, totalFramesWritten*bpf); + void* pDst = ma_offset_ptr(pWAVEHDR[pDevice->winmm.iNextHeaderPlayback].lpData, pDevice->winmm.headerFramesConsumedPlayback*bpf); + MA_COPY_MEMORY(pDst, pSrc, framesToCopy*bpf); + + pDevice->winmm.headerFramesConsumedPlayback += framesToCopy; + totalFramesWritten += framesToCopy; + + /* If we've consumed the buffer entirely we need to write it out to the device. */ + if (pDevice->winmm.headerFramesConsumedPlayback == (pWAVEHDR[pDevice->winmm.iNextHeaderPlayback].dwBufferLength/bpf)) { + pWAVEHDR[pDevice->winmm.iNextHeaderPlayback].dwUser = 1; /* 1 = locked. */ + pWAVEHDR[pDevice->winmm.iNextHeaderPlayback].dwFlags &= ~MA_WHDR_DONE; /* <-- Need to make sure the WHDR_DONE flag is unset. */ + + /* Make sure the event is reset to a non-signaled state to ensure we don't prematurely return from WaitForSingleObject(). */ + ResetEvent((HANDLE)pDevice->winmm.hEventPlayback); + + /* The device will be started here. */ + resultMM = ((MA_PFN_waveOutWrite)pDevice->pContext->winmm.waveOutWrite)((MA_HWAVEOUT)pDevice->winmm.hDevicePlayback, &pWAVEHDR[pDevice->winmm.iNextHeaderPlayback], sizeof(MA_WAVEHDR)); + if (resultMM != MA_MMSYSERR_NOERROR) { + result = ma_result_from_MMRESULT(resultMM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WinMM] waveOutWrite() failed."); + break; + } + + /* Make sure we move to the next header. */ + pDevice->winmm.iNextHeaderPlayback = (pDevice->winmm.iNextHeaderPlayback + 1) % pDevice->playback.internalPeriods; + pDevice->winmm.headerFramesConsumedPlayback = 0; + } + + /* If at this point we have consumed the entire input buffer we can return. */ + MA_ASSERT(totalFramesWritten <= frameCount); + if (totalFramesWritten == frameCount) { + break; + } + + /* Getting here means there's more to process. */ + continue; + } + + /* Getting here means there isn't enough room in the buffer and we need to wait for one to become available. */ + if (WaitForSingleObject((HANDLE)pDevice->winmm.hEventPlayback, INFINITE) != WAIT_OBJECT_0) { + result = MA_ERROR; + break; + } + + /* Something happened. If the next buffer has been marked as done we need to reset a bit of state. */ + if ((pWAVEHDR[pDevice->winmm.iNextHeaderPlayback].dwFlags & MA_WHDR_DONE) != 0) { + pWAVEHDR[pDevice->winmm.iNextHeaderPlayback].dwUser = 0; /* 0 = unlocked (make it available for writing). */ + pDevice->winmm.headerFramesConsumedPlayback = 0; + } + + /* If the device has been stopped we need to break. */ + if (ma_device_get_state(pDevice) != ma_device_state_started) { + break; + } + } + + if (pFramesWritten != NULL) { + *pFramesWritten = totalFramesWritten; + } + + return result; +} + +static ma_result ma_device_read__winmm(ma_device* pDevice, void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesRead) +{ + ma_result result = MA_SUCCESS; + MA_MMRESULT resultMM; + ma_uint32 totalFramesRead; + MA_WAVEHDR* pWAVEHDR; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pPCMFrames != NULL); + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + pWAVEHDR = (MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture; + + /* Keep processing as much data as possible. */ + totalFramesRead = 0; + while (totalFramesRead < frameCount) { + /* If the current header has some space available we need to write part of it. */ + if (pWAVEHDR[pDevice->winmm.iNextHeaderCapture].dwUser == 0) { /* 0 = unlocked. */ + /* The buffer is available for reading. If we fully consume it we need to add it back to the buffer. */ + ma_uint32 bpf = ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + ma_uint32 framesRemainingInHeader = (pWAVEHDR[pDevice->winmm.iNextHeaderCapture].dwBufferLength/bpf) - pDevice->winmm.headerFramesConsumedCapture; + + ma_uint32 framesToCopy = ma_min(framesRemainingInHeader, (frameCount - totalFramesRead)); + const void* pSrc = ma_offset_ptr(pWAVEHDR[pDevice->winmm.iNextHeaderCapture].lpData, pDevice->winmm.headerFramesConsumedCapture*bpf); + void* pDst = ma_offset_ptr(pPCMFrames, totalFramesRead*bpf); + MA_COPY_MEMORY(pDst, pSrc, framesToCopy*bpf); + + pDevice->winmm.headerFramesConsumedCapture += framesToCopy; + totalFramesRead += framesToCopy; + + /* If we've consumed the buffer entirely we need to add it back to the device. */ + if (pDevice->winmm.headerFramesConsumedCapture == (pWAVEHDR[pDevice->winmm.iNextHeaderCapture].dwBufferLength/bpf)) { + pWAVEHDR[pDevice->winmm.iNextHeaderCapture].dwUser = 1; /* 1 = locked. */ + pWAVEHDR[pDevice->winmm.iNextHeaderCapture].dwFlags &= ~MA_WHDR_DONE; /* <-- Need to make sure the WHDR_DONE flag is unset. */ + + /* Make sure the event is reset to a non-signaled state to ensure we don't prematurely return from WaitForSingleObject(). */ + ResetEvent((HANDLE)pDevice->winmm.hEventCapture); + + /* The device will be started here. */ + resultMM = ((MA_PFN_waveInAddBuffer)pDevice->pContext->winmm.waveInAddBuffer)((MA_HWAVEIN)pDevice->winmm.hDeviceCapture, &((MA_WAVEHDR*)pDevice->winmm.pWAVEHDRCapture)[pDevice->winmm.iNextHeaderCapture], sizeof(MA_WAVEHDR)); + if (resultMM != MA_MMSYSERR_NOERROR) { + result = ma_result_from_MMRESULT(resultMM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[WinMM] waveInAddBuffer() failed."); + break; + } + + /* Make sure we move to the next header. */ + pDevice->winmm.iNextHeaderCapture = (pDevice->winmm.iNextHeaderCapture + 1) % pDevice->capture.internalPeriods; + pDevice->winmm.headerFramesConsumedCapture = 0; + } + + /* If at this point we have filled the entire input buffer we can return. */ + MA_ASSERT(totalFramesRead <= frameCount); + if (totalFramesRead == frameCount) { + break; + } + + /* Getting here means there's more to process. */ + continue; + } + + /* Getting here means there isn't enough any data left to send to the client which means we need to wait for more. */ + if (WaitForSingleObject((HANDLE)pDevice->winmm.hEventCapture, INFINITE) != WAIT_OBJECT_0) { + result = MA_ERROR; + break; + } + + /* Something happened. If the next buffer has been marked as done we need to reset a bit of state. */ + if ((pWAVEHDR[pDevice->winmm.iNextHeaderCapture].dwFlags & MA_WHDR_DONE) != 0) { + pWAVEHDR[pDevice->winmm.iNextHeaderCapture].dwUser = 0; /* 0 = unlocked (make it available for reading). */ + pDevice->winmm.headerFramesConsumedCapture = 0; + } + + /* If the device has been stopped we need to break. */ + if (ma_device_get_state(pDevice) != ma_device_state_started) { + break; + } + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesRead; + } + + return result; +} + +static ma_result ma_context_uninit__winmm(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_winmm); + + ma_dlclose(ma_context_get_log(pContext), pContext->winmm.hWinMM); + return MA_SUCCESS; +} + +static ma_result ma_context_init__winmm(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + MA_ASSERT(pContext != NULL); + + (void)pConfig; + + pContext->winmm.hWinMM = ma_dlopen(ma_context_get_log(pContext), "winmm.dll"); + if (pContext->winmm.hWinMM == NULL) { + return MA_NO_BACKEND; + } + + pContext->winmm.waveOutGetNumDevs = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveOutGetNumDevs"); + pContext->winmm.waveOutGetDevCapsA = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveOutGetDevCapsA"); + pContext->winmm.waveOutOpen = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveOutOpen"); + pContext->winmm.waveOutClose = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveOutClose"); + pContext->winmm.waveOutPrepareHeader = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveOutPrepareHeader"); + pContext->winmm.waveOutUnprepareHeader = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveOutUnprepareHeader"); + pContext->winmm.waveOutWrite = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveOutWrite"); + pContext->winmm.waveOutReset = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveOutReset"); + pContext->winmm.waveInGetNumDevs = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInGetNumDevs"); + pContext->winmm.waveInGetDevCapsA = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInGetDevCapsA"); + pContext->winmm.waveInOpen = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInOpen"); + pContext->winmm.waveInClose = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInClose"); + pContext->winmm.waveInPrepareHeader = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInPrepareHeader"); + pContext->winmm.waveInUnprepareHeader = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInUnprepareHeader"); + pContext->winmm.waveInAddBuffer = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInAddBuffer"); + pContext->winmm.waveInStart = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInStart"); + pContext->winmm.waveInReset = ma_dlsym(ma_context_get_log(pContext), pContext->winmm.hWinMM, "waveInReset"); + + pCallbacks->onContextInit = ma_context_init__winmm; + pCallbacks->onContextUninit = ma_context_uninit__winmm; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__winmm; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__winmm; + pCallbacks->onDeviceInit = ma_device_init__winmm; + pCallbacks->onDeviceUninit = ma_device_uninit__winmm; + pCallbacks->onDeviceStart = ma_device_start__winmm; + pCallbacks->onDeviceStop = ma_device_stop__winmm; + pCallbacks->onDeviceRead = ma_device_read__winmm; + pCallbacks->onDeviceWrite = ma_device_write__winmm; + pCallbacks->onDeviceDataLoop = NULL; /* This is a blocking read-write API, so this can be NULL since miniaudio will manage the audio thread for us. */ + + return MA_SUCCESS; +} +#endif + + + + +/****************************************************************************** + +ALSA Backend + +******************************************************************************/ +#ifdef MA_HAS_ALSA + +#include /* poll(), struct pollfd */ +#include /* eventfd() */ + +#ifdef MA_NO_RUNTIME_LINKING + +/* asoundlib.h marks some functions with "inline" which isn't always supported. Need to emulate it. */ +#if !defined(__cplusplus) + #if defined(__STRICT_ANSI__) + #if !defined(inline) + #define inline __inline__ __attribute__((always_inline)) + #define MA_INLINE_DEFINED + #endif + #endif +#endif +#include +#if defined(MA_INLINE_DEFINED) + #undef inline + #undef MA_INLINE_DEFINED +#endif + +typedef snd_pcm_uframes_t ma_snd_pcm_uframes_t; +typedef snd_pcm_sframes_t ma_snd_pcm_sframes_t; +typedef snd_pcm_stream_t ma_snd_pcm_stream_t; +typedef snd_pcm_format_t ma_snd_pcm_format_t; +typedef snd_pcm_access_t ma_snd_pcm_access_t; +typedef snd_pcm_t ma_snd_pcm_t; +typedef snd_pcm_hw_params_t ma_snd_pcm_hw_params_t; +typedef snd_pcm_sw_params_t ma_snd_pcm_sw_params_t; +typedef snd_pcm_format_mask_t ma_snd_pcm_format_mask_t; +typedef snd_pcm_info_t ma_snd_pcm_info_t; +typedef snd_pcm_channel_area_t ma_snd_pcm_channel_area_t; +typedef snd_pcm_chmap_t ma_snd_pcm_chmap_t; +typedef snd_pcm_state_t ma_snd_pcm_state_t; + +/* snd_pcm_stream_t */ +#define MA_SND_PCM_STREAM_PLAYBACK SND_PCM_STREAM_PLAYBACK +#define MA_SND_PCM_STREAM_CAPTURE SND_PCM_STREAM_CAPTURE + +/* snd_pcm_format_t */ +#define MA_SND_PCM_FORMAT_UNKNOWN SND_PCM_FORMAT_UNKNOWN +#define MA_SND_PCM_FORMAT_U8 SND_PCM_FORMAT_U8 +#define MA_SND_PCM_FORMAT_S16_LE SND_PCM_FORMAT_S16_LE +#define MA_SND_PCM_FORMAT_S16_BE SND_PCM_FORMAT_S16_BE +#define MA_SND_PCM_FORMAT_S24_LE SND_PCM_FORMAT_S24_LE +#define MA_SND_PCM_FORMAT_S24_BE SND_PCM_FORMAT_S24_BE +#define MA_SND_PCM_FORMAT_S32_LE SND_PCM_FORMAT_S32_LE +#define MA_SND_PCM_FORMAT_S32_BE SND_PCM_FORMAT_S32_BE +#define MA_SND_PCM_FORMAT_FLOAT_LE SND_PCM_FORMAT_FLOAT_LE +#define MA_SND_PCM_FORMAT_FLOAT_BE SND_PCM_FORMAT_FLOAT_BE +#define MA_SND_PCM_FORMAT_FLOAT64_LE SND_PCM_FORMAT_FLOAT64_LE +#define MA_SND_PCM_FORMAT_FLOAT64_BE SND_PCM_FORMAT_FLOAT64_BE +#define MA_SND_PCM_FORMAT_MU_LAW SND_PCM_FORMAT_MU_LAW +#define MA_SND_PCM_FORMAT_A_LAW SND_PCM_FORMAT_A_LAW +#define MA_SND_PCM_FORMAT_S24_3LE SND_PCM_FORMAT_S24_3LE +#define MA_SND_PCM_FORMAT_S24_3BE SND_PCM_FORMAT_S24_3BE + +/* ma_snd_pcm_access_t */ +#define MA_SND_PCM_ACCESS_MMAP_INTERLEAVED SND_PCM_ACCESS_MMAP_INTERLEAVED +#define MA_SND_PCM_ACCESS_MMAP_NONINTERLEAVED SND_PCM_ACCESS_MMAP_NONINTERLEAVED +#define MA_SND_PCM_ACCESS_MMAP_COMPLEX SND_PCM_ACCESS_MMAP_COMPLEX +#define MA_SND_PCM_ACCESS_RW_INTERLEAVED SND_PCM_ACCESS_RW_INTERLEAVED +#define MA_SND_PCM_ACCESS_RW_NONINTERLEAVED SND_PCM_ACCESS_RW_NONINTERLEAVED + +/* Channel positions. */ +#define MA_SND_CHMAP_UNKNOWN SND_CHMAP_UNKNOWN +#define MA_SND_CHMAP_NA SND_CHMAP_NA +#define MA_SND_CHMAP_MONO SND_CHMAP_MONO +#define MA_SND_CHMAP_FL SND_CHMAP_FL +#define MA_SND_CHMAP_FR SND_CHMAP_FR +#define MA_SND_CHMAP_RL SND_CHMAP_RL +#define MA_SND_CHMAP_RR SND_CHMAP_RR +#define MA_SND_CHMAP_FC SND_CHMAP_FC +#define MA_SND_CHMAP_LFE SND_CHMAP_LFE +#define MA_SND_CHMAP_SL SND_CHMAP_SL +#define MA_SND_CHMAP_SR SND_CHMAP_SR +#define MA_SND_CHMAP_RC SND_CHMAP_RC +#define MA_SND_CHMAP_FLC SND_CHMAP_FLC +#define MA_SND_CHMAP_FRC SND_CHMAP_FRC +#define MA_SND_CHMAP_RLC SND_CHMAP_RLC +#define MA_SND_CHMAP_RRC SND_CHMAP_RRC +#define MA_SND_CHMAP_FLW SND_CHMAP_FLW +#define MA_SND_CHMAP_FRW SND_CHMAP_FRW +#define MA_SND_CHMAP_FLH SND_CHMAP_FLH +#define MA_SND_CHMAP_FCH SND_CHMAP_FCH +#define MA_SND_CHMAP_FRH SND_CHMAP_FRH +#define MA_SND_CHMAP_TC SND_CHMAP_TC +#define MA_SND_CHMAP_TFL SND_CHMAP_TFL +#define MA_SND_CHMAP_TFR SND_CHMAP_TFR +#define MA_SND_CHMAP_TFC SND_CHMAP_TFC +#define MA_SND_CHMAP_TRL SND_CHMAP_TRL +#define MA_SND_CHMAP_TRR SND_CHMAP_TRR +#define MA_SND_CHMAP_TRC SND_CHMAP_TRC +#define MA_SND_CHMAP_TFLC SND_CHMAP_TFLC +#define MA_SND_CHMAP_TFRC SND_CHMAP_TFRC +#define MA_SND_CHMAP_TSL SND_CHMAP_TSL +#define MA_SND_CHMAP_TSR SND_CHMAP_TSR +#define MA_SND_CHMAP_LLFE SND_CHMAP_LLFE +#define MA_SND_CHMAP_RLFE SND_CHMAP_RLFE +#define MA_SND_CHMAP_BC SND_CHMAP_BC +#define MA_SND_CHMAP_BLC SND_CHMAP_BLC +#define MA_SND_CHMAP_BRC SND_CHMAP_BRC + +/* Open mode flags. */ +#define MA_SND_PCM_NO_AUTO_RESAMPLE SND_PCM_NO_AUTO_RESAMPLE +#define MA_SND_PCM_NO_AUTO_CHANNELS SND_PCM_NO_AUTO_CHANNELS +#define MA_SND_PCM_NO_AUTO_FORMAT SND_PCM_NO_AUTO_FORMAT +#else +#include /* For EPIPE, etc. */ +typedef unsigned long ma_snd_pcm_uframes_t; +typedef long ma_snd_pcm_sframes_t; +typedef int ma_snd_pcm_stream_t; +typedef int ma_snd_pcm_format_t; +typedef int ma_snd_pcm_access_t; +typedef int ma_snd_pcm_state_t; +typedef struct ma_snd_pcm_t ma_snd_pcm_t; +typedef struct ma_snd_pcm_hw_params_t ma_snd_pcm_hw_params_t; +typedef struct ma_snd_pcm_sw_params_t ma_snd_pcm_sw_params_t; +typedef struct ma_snd_pcm_format_mask_t ma_snd_pcm_format_mask_t; +typedef struct ma_snd_pcm_info_t ma_snd_pcm_info_t; +typedef struct +{ + void* addr; + unsigned int first; + unsigned int step; +} ma_snd_pcm_channel_area_t; +typedef struct +{ + unsigned int channels; + unsigned int pos[1]; +} ma_snd_pcm_chmap_t; + +/* snd_pcm_state_t */ +#define MA_SND_PCM_STATE_OPEN 0 +#define MA_SND_PCM_STATE_SETUP 1 +#define MA_SND_PCM_STATE_PREPARED 2 +#define MA_SND_PCM_STATE_RUNNING 3 +#define MA_SND_PCM_STATE_XRUN 4 +#define MA_SND_PCM_STATE_DRAINING 5 +#define MA_SND_PCM_STATE_PAUSED 6 +#define MA_SND_PCM_STATE_SUSPENDED 7 +#define MA_SND_PCM_STATE_DISCONNECTED 8 + +/* snd_pcm_stream_t */ +#define MA_SND_PCM_STREAM_PLAYBACK 0 +#define MA_SND_PCM_STREAM_CAPTURE 1 + +/* snd_pcm_format_t */ +#define MA_SND_PCM_FORMAT_UNKNOWN -1 +#define MA_SND_PCM_FORMAT_U8 1 +#define MA_SND_PCM_FORMAT_S16_LE 2 +#define MA_SND_PCM_FORMAT_S16_BE 3 +#define MA_SND_PCM_FORMAT_S24_LE 6 +#define MA_SND_PCM_FORMAT_S24_BE 7 +#define MA_SND_PCM_FORMAT_S32_LE 10 +#define MA_SND_PCM_FORMAT_S32_BE 11 +#define MA_SND_PCM_FORMAT_FLOAT_LE 14 +#define MA_SND_PCM_FORMAT_FLOAT_BE 15 +#define MA_SND_PCM_FORMAT_FLOAT64_LE 16 +#define MA_SND_PCM_FORMAT_FLOAT64_BE 17 +#define MA_SND_PCM_FORMAT_MU_LAW 20 +#define MA_SND_PCM_FORMAT_A_LAW 21 +#define MA_SND_PCM_FORMAT_S24_3LE 32 +#define MA_SND_PCM_FORMAT_S24_3BE 33 + +/* snd_pcm_access_t */ +#define MA_SND_PCM_ACCESS_MMAP_INTERLEAVED 0 +#define MA_SND_PCM_ACCESS_MMAP_NONINTERLEAVED 1 +#define MA_SND_PCM_ACCESS_MMAP_COMPLEX 2 +#define MA_SND_PCM_ACCESS_RW_INTERLEAVED 3 +#define MA_SND_PCM_ACCESS_RW_NONINTERLEAVED 4 + +/* Channel positions. */ +#define MA_SND_CHMAP_UNKNOWN 0 +#define MA_SND_CHMAP_NA 1 +#define MA_SND_CHMAP_MONO 2 +#define MA_SND_CHMAP_FL 3 +#define MA_SND_CHMAP_FR 4 +#define MA_SND_CHMAP_RL 5 +#define MA_SND_CHMAP_RR 6 +#define MA_SND_CHMAP_FC 7 +#define MA_SND_CHMAP_LFE 8 +#define MA_SND_CHMAP_SL 9 +#define MA_SND_CHMAP_SR 10 +#define MA_SND_CHMAP_RC 11 +#define MA_SND_CHMAP_FLC 12 +#define MA_SND_CHMAP_FRC 13 +#define MA_SND_CHMAP_RLC 14 +#define MA_SND_CHMAP_RRC 15 +#define MA_SND_CHMAP_FLW 16 +#define MA_SND_CHMAP_FRW 17 +#define MA_SND_CHMAP_FLH 18 +#define MA_SND_CHMAP_FCH 19 +#define MA_SND_CHMAP_FRH 20 +#define MA_SND_CHMAP_TC 21 +#define MA_SND_CHMAP_TFL 22 +#define MA_SND_CHMAP_TFR 23 +#define MA_SND_CHMAP_TFC 24 +#define MA_SND_CHMAP_TRL 25 +#define MA_SND_CHMAP_TRR 26 +#define MA_SND_CHMAP_TRC 27 +#define MA_SND_CHMAP_TFLC 28 +#define MA_SND_CHMAP_TFRC 29 +#define MA_SND_CHMAP_TSL 30 +#define MA_SND_CHMAP_TSR 31 +#define MA_SND_CHMAP_LLFE 32 +#define MA_SND_CHMAP_RLFE 33 +#define MA_SND_CHMAP_BC 34 +#define MA_SND_CHMAP_BLC 35 +#define MA_SND_CHMAP_BRC 36 + +/* Open mode flags. */ +#define MA_SND_PCM_NO_AUTO_RESAMPLE 0x00010000 +#define MA_SND_PCM_NO_AUTO_CHANNELS 0x00020000 +#define MA_SND_PCM_NO_AUTO_FORMAT 0x00040000 +#endif + +typedef int (* ma_snd_pcm_open_proc) (ma_snd_pcm_t **pcm, const char *name, ma_snd_pcm_stream_t stream, int mode); +typedef int (* ma_snd_pcm_close_proc) (ma_snd_pcm_t *pcm); +typedef size_t (* ma_snd_pcm_hw_params_sizeof_proc) (void); +typedef int (* ma_snd_pcm_hw_params_any_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params); +typedef int (* ma_snd_pcm_hw_params_set_format_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, ma_snd_pcm_format_t val); +typedef int (* ma_snd_pcm_hw_params_set_format_first_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, ma_snd_pcm_format_t *format); +typedef void (* ma_snd_pcm_hw_params_get_format_mask_proc) (ma_snd_pcm_hw_params_t *params, ma_snd_pcm_format_mask_t *mask); +typedef int (* ma_snd_pcm_hw_params_set_channels_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int val); +typedef int (* ma_snd_pcm_hw_params_set_channels_near_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int *val); +typedef int (* ma_snd_pcm_hw_params_set_channels_minmax_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int *minimum, unsigned int *maximum); +typedef int (* ma_snd_pcm_hw_params_set_rate_resample_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int val); +typedef int (* ma_snd_pcm_hw_params_set_rate_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int val, int dir); +typedef int (* ma_snd_pcm_hw_params_set_rate_near_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int *val, int *dir); +typedef int (* ma_snd_pcm_hw_params_set_buffer_size_near_proc)(ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, ma_snd_pcm_uframes_t *val); +typedef int (* ma_snd_pcm_hw_params_set_periods_near_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int *val, int *dir); +typedef int (* ma_snd_pcm_hw_params_set_access_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, ma_snd_pcm_access_t _access); +typedef int (* ma_snd_pcm_hw_params_get_format_proc) (const ma_snd_pcm_hw_params_t *params, ma_snd_pcm_format_t *format); +typedef int (* ma_snd_pcm_hw_params_get_channels_proc) (const ma_snd_pcm_hw_params_t *params, unsigned int *val); +typedef int (* ma_snd_pcm_hw_params_get_channels_min_proc) (const ma_snd_pcm_hw_params_t *params, unsigned int *val); +typedef int (* ma_snd_pcm_hw_params_get_channels_max_proc) (const ma_snd_pcm_hw_params_t *params, unsigned int *val); +typedef int (* ma_snd_pcm_hw_params_get_rate_proc) (const ma_snd_pcm_hw_params_t *params, unsigned int *rate, int *dir); +typedef int (* ma_snd_pcm_hw_params_get_rate_min_proc) (const ma_snd_pcm_hw_params_t *params, unsigned int *rate, int *dir); +typedef int (* ma_snd_pcm_hw_params_get_rate_max_proc) (const ma_snd_pcm_hw_params_t *params, unsigned int *rate, int *dir); +typedef int (* ma_snd_pcm_hw_params_get_buffer_size_proc) (const ma_snd_pcm_hw_params_t *params, ma_snd_pcm_uframes_t *val); +typedef int (* ma_snd_pcm_hw_params_get_periods_proc) (const ma_snd_pcm_hw_params_t *params, unsigned int *val, int *dir); +typedef int (* ma_snd_pcm_hw_params_get_access_proc) (const ma_snd_pcm_hw_params_t *params, ma_snd_pcm_access_t *_access); +typedef int (* ma_snd_pcm_hw_params_test_format_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, ma_snd_pcm_format_t val); +typedef int (* ma_snd_pcm_hw_params_test_channels_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int val); +typedef int (* ma_snd_pcm_hw_params_test_rate_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params, unsigned int val, int dir); +typedef int (* ma_snd_pcm_hw_params_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_hw_params_t *params); +typedef size_t (* ma_snd_pcm_sw_params_sizeof_proc) (void); +typedef int (* ma_snd_pcm_sw_params_current_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_sw_params_t *params); +typedef int (* ma_snd_pcm_sw_params_get_boundary_proc) (const ma_snd_pcm_sw_params_t *params, ma_snd_pcm_uframes_t* val); +typedef int (* ma_snd_pcm_sw_params_set_avail_min_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_sw_params_t *params, ma_snd_pcm_uframes_t val); +typedef int (* ma_snd_pcm_sw_params_set_start_threshold_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_sw_params_t *params, ma_snd_pcm_uframes_t val); +typedef int (* ma_snd_pcm_sw_params_set_stop_threshold_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_sw_params_t *params, ma_snd_pcm_uframes_t val); +typedef int (* ma_snd_pcm_sw_params_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_sw_params_t *params); +typedef size_t (* ma_snd_pcm_format_mask_sizeof_proc) (void); +typedef int (* ma_snd_pcm_format_mask_test_proc) (const ma_snd_pcm_format_mask_t *mask, ma_snd_pcm_format_t val); +typedef ma_snd_pcm_chmap_t * (* ma_snd_pcm_get_chmap_proc) (ma_snd_pcm_t *pcm); +typedef ma_snd_pcm_state_t (* ma_snd_pcm_state_proc) (ma_snd_pcm_t *pcm); +typedef int (* ma_snd_pcm_prepare_proc) (ma_snd_pcm_t *pcm); +typedef int (* ma_snd_pcm_start_proc) (ma_snd_pcm_t *pcm); +typedef int (* ma_snd_pcm_drop_proc) (ma_snd_pcm_t *pcm); +typedef int (* ma_snd_pcm_drain_proc) (ma_snd_pcm_t *pcm); +typedef int (* ma_snd_pcm_reset_proc) (ma_snd_pcm_t *pcm); +typedef int (* ma_snd_device_name_hint_proc) (int card, const char *iface, void ***hints); +typedef char * (* ma_snd_device_name_get_hint_proc) (const void *hint, const char *id); +typedef int (* ma_snd_card_get_index_proc) (const char *name); +typedef int (* ma_snd_device_name_free_hint_proc) (void **hints); +typedef int (* ma_snd_pcm_mmap_begin_proc) (ma_snd_pcm_t *pcm, const ma_snd_pcm_channel_area_t **areas, ma_snd_pcm_uframes_t *offset, ma_snd_pcm_uframes_t *frames); +typedef ma_snd_pcm_sframes_t (* ma_snd_pcm_mmap_commit_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_uframes_t offset, ma_snd_pcm_uframes_t frames); +typedef int (* ma_snd_pcm_recover_proc) (ma_snd_pcm_t *pcm, int err, int silent); +typedef ma_snd_pcm_sframes_t (* ma_snd_pcm_readi_proc) (ma_snd_pcm_t *pcm, void *buffer, ma_snd_pcm_uframes_t size); +typedef ma_snd_pcm_sframes_t (* ma_snd_pcm_writei_proc) (ma_snd_pcm_t *pcm, const void *buffer, ma_snd_pcm_uframes_t size); +typedef ma_snd_pcm_sframes_t (* ma_snd_pcm_avail_proc) (ma_snd_pcm_t *pcm); +typedef ma_snd_pcm_sframes_t (* ma_snd_pcm_avail_update_proc) (ma_snd_pcm_t *pcm); +typedef int (* ma_snd_pcm_wait_proc) (ma_snd_pcm_t *pcm, int timeout); +typedef int (* ma_snd_pcm_nonblock_proc) (ma_snd_pcm_t *pcm, int nonblock); +typedef int (* ma_snd_pcm_info_proc) (ma_snd_pcm_t *pcm, ma_snd_pcm_info_t* info); +typedef size_t (* ma_snd_pcm_info_sizeof_proc) (void); +typedef const char* (* ma_snd_pcm_info_get_name_proc) (const ma_snd_pcm_info_t* info); +typedef int (* ma_snd_pcm_poll_descriptors_proc) (ma_snd_pcm_t *pcm, struct pollfd *pfds, unsigned int space); +typedef int (* ma_snd_pcm_poll_descriptors_count_proc) (ma_snd_pcm_t *pcm); +typedef int (* ma_snd_pcm_poll_descriptors_revents_proc) (ma_snd_pcm_t *pcm, struct pollfd *pfds, unsigned int nfds, unsigned short *revents); +typedef int (* ma_snd_config_update_free_global_proc) (void); + +/* This array specifies each of the common devices that can be used for both playback and capture. */ +static const char* g_maCommonDeviceNamesALSA[] = { + "default", + "null", + "pulse", + "jack" +}; + +/* This array allows us to blacklist specific playback devices. */ +static const char* g_maBlacklistedPlaybackDeviceNamesALSA[] = { + "" +}; + +/* This array allows us to blacklist specific capture devices. */ +static const char* g_maBlacklistedCaptureDeviceNamesALSA[] = { + "" +}; + + +static ma_snd_pcm_format_t ma_convert_ma_format_to_alsa_format(ma_format format) +{ + ma_snd_pcm_format_t ALSAFormats[] = { + MA_SND_PCM_FORMAT_UNKNOWN, /* ma_format_unknown */ + MA_SND_PCM_FORMAT_U8, /* ma_format_u8 */ + MA_SND_PCM_FORMAT_S16_LE, /* ma_format_s16 */ + MA_SND_PCM_FORMAT_S24_3LE, /* ma_format_s24 */ + MA_SND_PCM_FORMAT_S32_LE, /* ma_format_s32 */ + MA_SND_PCM_FORMAT_FLOAT_LE /* ma_format_f32 */ + }; + + if (ma_is_big_endian()) { + ALSAFormats[0] = MA_SND_PCM_FORMAT_UNKNOWN; + ALSAFormats[1] = MA_SND_PCM_FORMAT_U8; + ALSAFormats[2] = MA_SND_PCM_FORMAT_S16_BE; + ALSAFormats[3] = MA_SND_PCM_FORMAT_S24_3BE; + ALSAFormats[4] = MA_SND_PCM_FORMAT_S32_BE; + ALSAFormats[5] = MA_SND_PCM_FORMAT_FLOAT_BE; + } + + return ALSAFormats[format]; +} + +static ma_format ma_format_from_alsa(ma_snd_pcm_format_t formatALSA) +{ + if (ma_is_little_endian()) { + switch (formatALSA) { + case MA_SND_PCM_FORMAT_S16_LE: return ma_format_s16; + case MA_SND_PCM_FORMAT_S24_3LE: return ma_format_s24; + case MA_SND_PCM_FORMAT_S32_LE: return ma_format_s32; + case MA_SND_PCM_FORMAT_FLOAT_LE: return ma_format_f32; + default: break; + } + } else { + switch (formatALSA) { + case MA_SND_PCM_FORMAT_S16_BE: return ma_format_s16; + case MA_SND_PCM_FORMAT_S24_3BE: return ma_format_s24; + case MA_SND_PCM_FORMAT_S32_BE: return ma_format_s32; + case MA_SND_PCM_FORMAT_FLOAT_BE: return ma_format_f32; + default: break; + } + } + + /* Endian agnostic. */ + switch (formatALSA) { + case MA_SND_PCM_FORMAT_U8: return ma_format_u8; + default: return ma_format_unknown; + } +} + +static ma_channel ma_convert_alsa_channel_position_to_ma_channel(unsigned int alsaChannelPos) +{ + switch (alsaChannelPos) + { + case MA_SND_CHMAP_MONO: return MA_CHANNEL_MONO; + case MA_SND_CHMAP_FL: return MA_CHANNEL_FRONT_LEFT; + case MA_SND_CHMAP_FR: return MA_CHANNEL_FRONT_RIGHT; + case MA_SND_CHMAP_RL: return MA_CHANNEL_BACK_LEFT; + case MA_SND_CHMAP_RR: return MA_CHANNEL_BACK_RIGHT; + case MA_SND_CHMAP_FC: return MA_CHANNEL_FRONT_CENTER; + case MA_SND_CHMAP_LFE: return MA_CHANNEL_LFE; + case MA_SND_CHMAP_SL: return MA_CHANNEL_SIDE_LEFT; + case MA_SND_CHMAP_SR: return MA_CHANNEL_SIDE_RIGHT; + case MA_SND_CHMAP_RC: return MA_CHANNEL_BACK_CENTER; + case MA_SND_CHMAP_FLC: return MA_CHANNEL_FRONT_LEFT_CENTER; + case MA_SND_CHMAP_FRC: return MA_CHANNEL_FRONT_RIGHT_CENTER; + case MA_SND_CHMAP_RLC: return 0; + case MA_SND_CHMAP_RRC: return 0; + case MA_SND_CHMAP_FLW: return 0; + case MA_SND_CHMAP_FRW: return 0; + case MA_SND_CHMAP_FLH: return 0; + case MA_SND_CHMAP_FCH: return 0; + case MA_SND_CHMAP_FRH: return 0; + case MA_SND_CHMAP_TC: return MA_CHANNEL_TOP_CENTER; + case MA_SND_CHMAP_TFL: return MA_CHANNEL_TOP_FRONT_LEFT; + case MA_SND_CHMAP_TFR: return MA_CHANNEL_TOP_FRONT_RIGHT; + case MA_SND_CHMAP_TFC: return MA_CHANNEL_TOP_FRONT_CENTER; + case MA_SND_CHMAP_TRL: return MA_CHANNEL_TOP_BACK_LEFT; + case MA_SND_CHMAP_TRR: return MA_CHANNEL_TOP_BACK_RIGHT; + case MA_SND_CHMAP_TRC: return MA_CHANNEL_TOP_BACK_CENTER; + default: break; + } + + return 0; +} + +static ma_bool32 ma_is_common_device_name__alsa(const char* name) +{ + size_t iName; + for (iName = 0; iName < ma_countof(g_maCommonDeviceNamesALSA); ++iName) { + if (ma_strcmp(name, g_maCommonDeviceNamesALSA[iName]) == 0) { + return MA_TRUE; + } + } + + return MA_FALSE; +} + + +static ma_bool32 ma_is_playback_device_blacklisted__alsa(const char* name) +{ + size_t iName; + for (iName = 0; iName < ma_countof(g_maBlacklistedPlaybackDeviceNamesALSA); ++iName) { + if (ma_strcmp(name, g_maBlacklistedPlaybackDeviceNamesALSA[iName]) == 0) { + return MA_TRUE; + } + } + + return MA_FALSE; +} + +static ma_bool32 ma_is_capture_device_blacklisted__alsa(const char* name) +{ + size_t iName; + for (iName = 0; iName < ma_countof(g_maBlacklistedCaptureDeviceNamesALSA); ++iName) { + if (ma_strcmp(name, g_maBlacklistedCaptureDeviceNamesALSA[iName]) == 0) { + return MA_TRUE; + } + } + + return MA_FALSE; +} + +static ma_bool32 ma_is_device_blacklisted__alsa(ma_device_type deviceType, const char* name) +{ + if (deviceType == ma_device_type_playback) { + return ma_is_playback_device_blacklisted__alsa(name); + } else { + return ma_is_capture_device_blacklisted__alsa(name); + } +} + + +static const char* ma_find_char(const char* str, char c, int* index) +{ + int i = 0; + for (;;) { + if (str[i] == '\0') { + if (index) *index = -1; + return NULL; + } + + if (str[i] == c) { + if (index) *index = i; + return str + i; + } + + i += 1; + } + + /* Should never get here, but treat it as though the character was not found to make me feel better inside. */ + if (index) *index = -1; + return NULL; +} + +static ma_bool32 ma_is_device_name_in_hw_format__alsa(const char* hwid) +{ + /* This function is just checking whether or not hwid is in "hw:%d,%d" format. */ + + int commaPos; + const char* dev; + int i; + + if (hwid == NULL) { + return MA_FALSE; + } + + if (hwid[0] != 'h' || hwid[1] != 'w' || hwid[2] != ':') { + return MA_FALSE; + } + + hwid += 3; + + dev = ma_find_char(hwid, ',', &commaPos); + if (dev == NULL) { + return MA_FALSE; + } else { + dev += 1; /* Skip past the ",". */ + } + + /* Check if the part between the ":" and the "," contains only numbers. If not, return false. */ + for (i = 0; i < commaPos; ++i) { + if (hwid[i] < '0' || hwid[i] > '9') { + return MA_FALSE; + } + } + + /* Check if everything after the "," is numeric. If not, return false. */ + i = 0; + while (dev[i] != '\0') { + if (dev[i] < '0' || dev[i] > '9') { + return MA_FALSE; + } + i += 1; + } + + return MA_TRUE; +} + +static int ma_convert_device_name_to_hw_format__alsa(ma_context* pContext, char* dst, size_t dstSize, const char* src) /* Returns 0 on success, non-0 on error. */ +{ + /* src should look something like this: "hw:CARD=I82801AAICH,DEV=0" */ + + int colonPos; + int commaPos; + char card[256]; + const char* dev; + int cardIndex; + + if (dst == NULL) { + return -1; + } + if (dstSize < 7) { + return -1; /* Absolute minimum size of the output buffer is 7 bytes. */ + } + + *dst = '\0'; /* Safety. */ + if (src == NULL) { + return -1; + } + + /* If the input name is already in "hw:%d,%d" format, just return that verbatim. */ + if (ma_is_device_name_in_hw_format__alsa(src)) { + return ma_strcpy_s(dst, dstSize, src); + } + + src = ma_find_char(src, ':', &colonPos); + if (src == NULL) { + return -1; /* Couldn't find a colon */ + } + + dev = ma_find_char(src, ',', &commaPos); + if (dev == NULL) { + dev = "0"; + ma_strncpy_s(card, sizeof(card), src+6, (size_t)-1); /* +6 = ":CARD=" */ + } else { + dev = dev + 5; /* +5 = ",DEV=" */ + ma_strncpy_s(card, sizeof(card), src+6, commaPos-6); /* +6 = ":CARD=" */ + } + + cardIndex = ((ma_snd_card_get_index_proc)pContext->alsa.snd_card_get_index)(card); + if (cardIndex < 0) { + return -2; /* Failed to retrieve the card index. */ + } + + + /* Construction. */ + dst[0] = 'h'; dst[1] = 'w'; dst[2] = ':'; + if (ma_itoa_s(cardIndex, dst+3, dstSize-3, 10) != 0) { + return -3; + } + if (ma_strcat_s(dst, dstSize, ",") != 0) { + return -3; + } + if (ma_strcat_s(dst, dstSize, dev) != 0) { + return -3; + } + + return 0; +} + +static ma_bool32 ma_does_id_exist_in_list__alsa(ma_device_id* pUniqueIDs, ma_uint32 count, const char* pHWID) +{ + ma_uint32 i; + + MA_ASSERT(pHWID != NULL); + + for (i = 0; i < count; ++i) { + if (ma_strcmp(pUniqueIDs[i].alsa, pHWID) == 0) { + return MA_TRUE; + } + } + + return MA_FALSE; +} + + +static ma_result ma_context_open_pcm__alsa(ma_context* pContext, ma_share_mode shareMode, ma_device_type deviceType, const ma_device_id* pDeviceID, int openMode, ma_snd_pcm_t** ppPCM) +{ + ma_snd_pcm_t* pPCM; + ma_snd_pcm_stream_t stream; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppPCM != NULL); + + *ppPCM = NULL; + pPCM = NULL; + + stream = (deviceType == ma_device_type_playback) ? MA_SND_PCM_STREAM_PLAYBACK : MA_SND_PCM_STREAM_CAPTURE; + + if (pDeviceID == NULL) { + ma_bool32 isDeviceOpen; + size_t i; + + /* + We're opening the default device. I don't know if trying anything other than "default" is necessary, but it makes + me feel better to try as hard as we can get to get _something_ working. + */ + const char* defaultDeviceNames[] = { + "default", + NULL, + NULL, + NULL, + NULL, + NULL, + NULL + }; + + if (shareMode == ma_share_mode_exclusive) { + defaultDeviceNames[1] = "hw"; + defaultDeviceNames[2] = "hw:0"; + defaultDeviceNames[3] = "hw:0,0"; + } else { + if (deviceType == ma_device_type_playback) { + defaultDeviceNames[1] = "dmix"; + defaultDeviceNames[2] = "dmix:0"; + defaultDeviceNames[3] = "dmix:0,0"; + } else { + defaultDeviceNames[1] = "dsnoop"; + defaultDeviceNames[2] = "dsnoop:0"; + defaultDeviceNames[3] = "dsnoop:0,0"; + } + defaultDeviceNames[4] = "hw"; + defaultDeviceNames[5] = "hw:0"; + defaultDeviceNames[6] = "hw:0,0"; + } + + isDeviceOpen = MA_FALSE; + for (i = 0; i < ma_countof(defaultDeviceNames); ++i) { + if (defaultDeviceNames[i] != NULL && defaultDeviceNames[i][0] != '\0') { + if (((ma_snd_pcm_open_proc)pContext->alsa.snd_pcm_open)(&pPCM, defaultDeviceNames[i], stream, openMode) == 0) { + isDeviceOpen = MA_TRUE; + break; + } + } + } + + if (!isDeviceOpen) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[ALSA] snd_pcm_open() failed when trying to open an appropriate default device."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + } else { + /* + We're trying to open a specific device. There's a few things to consider here: + + miniaudio recognizes a special format of device id that excludes the "hw", "dmix", etc. prefix. It looks like this: ":0,0", ":0,1", etc. When + an ID of this format is specified, it indicates to miniaudio that it can try different combinations of plugins ("hw", "dmix", etc.) until it + finds an appropriate one that works. This comes in very handy when trying to open a device in shared mode ("dmix"), vs exclusive mode ("hw"). + */ + + /* May end up needing to make small adjustments to the ID, so make a copy. */ + ma_device_id deviceID = *pDeviceID; + int resultALSA = -ENODEV; + + if (deviceID.alsa[0] != ':') { + /* The ID is not in ":0,0" format. Use the ID exactly as-is. */ + resultALSA = ((ma_snd_pcm_open_proc)pContext->alsa.snd_pcm_open)(&pPCM, deviceID.alsa, stream, openMode); + } else { + char hwid[256]; + + /* The ID is in ":0,0" format. Try different plugins depending on the shared mode. */ + if (deviceID.alsa[1] == '\0') { + deviceID.alsa[0] = '\0'; /* An ID of ":" should be converted to "". */ + } + + if (shareMode == ma_share_mode_shared) { + if (deviceType == ma_device_type_playback) { + ma_strcpy_s(hwid, sizeof(hwid), "dmix"); + } else { + ma_strcpy_s(hwid, sizeof(hwid), "dsnoop"); + } + + if (ma_strcat_s(hwid, sizeof(hwid), deviceID.alsa) == 0) { + resultALSA = ((ma_snd_pcm_open_proc)pContext->alsa.snd_pcm_open)(&pPCM, hwid, stream, openMode); + } + } + + /* If at this point we still don't have an open device it means we're either preferencing exclusive mode or opening with "dmix"/"dsnoop" failed. */ + if (resultALSA != 0) { + ma_strcpy_s(hwid, sizeof(hwid), "hw"); + if (ma_strcat_s(hwid, sizeof(hwid), deviceID.alsa) == 0) { + resultALSA = ((ma_snd_pcm_open_proc)pContext->alsa.snd_pcm_open)(&pPCM, hwid, stream, openMode); + } + } + } + + if (resultALSA < 0) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[ALSA] snd_pcm_open() failed."); + return ma_result_from_errno(-resultALSA); + } + } + + *ppPCM = pPCM; + return MA_SUCCESS; +} + + +static ma_result ma_context_enumerate_devices__alsa(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + int resultALSA; + ma_bool32 cbResult = MA_TRUE; + char** ppDeviceHints; + ma_device_id* pUniqueIDs = NULL; + ma_uint32 uniqueIDCount = 0; + char** ppNextDeviceHint; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + ma_mutex_lock(&pContext->alsa.internalDeviceEnumLock); + + resultALSA = ((ma_snd_device_name_hint_proc)pContext->alsa.snd_device_name_hint)(-1, "pcm", (void***)&ppDeviceHints); + if (resultALSA < 0) { + ma_mutex_unlock(&pContext->alsa.internalDeviceEnumLock); + return ma_result_from_errno(-resultALSA); + } + + ppNextDeviceHint = ppDeviceHints; + while (*ppNextDeviceHint != NULL) { + char* NAME = ((ma_snd_device_name_get_hint_proc)pContext->alsa.snd_device_name_get_hint)(*ppNextDeviceHint, "NAME"); + char* DESC = ((ma_snd_device_name_get_hint_proc)pContext->alsa.snd_device_name_get_hint)(*ppNextDeviceHint, "DESC"); + char* IOID = ((ma_snd_device_name_get_hint_proc)pContext->alsa.snd_device_name_get_hint)(*ppNextDeviceHint, "IOID"); + ma_device_type deviceType = ma_device_type_playback; + ma_bool32 stopEnumeration = MA_FALSE; + char hwid[sizeof(pUniqueIDs->alsa)]; + ma_device_info deviceInfo; + + if ((IOID == NULL || ma_strcmp(IOID, "Output") == 0)) { + deviceType = ma_device_type_playback; + } + if ((IOID != NULL && ma_strcmp(IOID, "Input" ) == 0)) { + deviceType = ma_device_type_capture; + } + + if (NAME != NULL) { + if (pContext->alsa.useVerboseDeviceEnumeration) { + /* Verbose mode. Use the name exactly as-is. */ + ma_strncpy_s(hwid, sizeof(hwid), NAME, (size_t)-1); + } else { + /* Simplified mode. Use ":%d,%d" format. */ + if (ma_convert_device_name_to_hw_format__alsa(pContext, hwid, sizeof(hwid), NAME) == 0) { + /* + At this point, hwid looks like "hw:0,0". In simplified enumeration mode, we actually want to strip off the + plugin name so it looks like ":0,0". The reason for this is that this special format is detected at device + initialization time and is used as an indicator to try to use the most appropriate plugin depending on the + device type and sharing mode. + */ + char* dst = hwid; + char* src = hwid+2; + while ((*dst++ = *src++)); + } else { + /* Conversion to "hw:%d,%d" failed. Just use the name as-is. */ + ma_strncpy_s(hwid, sizeof(hwid), NAME, (size_t)-1); + } + + if (ma_does_id_exist_in_list__alsa(pUniqueIDs, uniqueIDCount, hwid)) { + goto next_device; /* The device has already been enumerated. Move on to the next one. */ + } else { + /* The device has not yet been enumerated. Make sure it's added to our list so that it's not enumerated again. */ + size_t newCapacity = sizeof(*pUniqueIDs) * (uniqueIDCount + 1); + ma_device_id* pNewUniqueIDs = (ma_device_id*)ma_realloc(pUniqueIDs, newCapacity, &pContext->allocationCallbacks); + if (pNewUniqueIDs == NULL) { + goto next_device; /* Failed to allocate memory. */ + } + + pUniqueIDs = pNewUniqueIDs; + MA_COPY_MEMORY(pUniqueIDs[uniqueIDCount].alsa, hwid, sizeof(hwid)); + uniqueIDCount += 1; + } + } + } else { + MA_ZERO_MEMORY(hwid, sizeof(hwid)); + } + + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.id.alsa, sizeof(deviceInfo.id.alsa), hwid, (size_t)-1); + + /* + There's no good way to determine whether or not a device is the default on Linux. We're just going to do something simple and + just use the name of "default" as the indicator. + */ + if (ma_strcmp(deviceInfo.id.alsa, "default") == 0) { + deviceInfo.isDefault = MA_TRUE; + } + + + /* + DESC is the friendly name. We treat this slightly differently depending on whether or not we are using verbose + device enumeration. In verbose mode we want to take the entire description so that the end-user can distinguish + between the subdevices of each card/dev pair. In simplified mode, however, we only want the first part of the + description. + + The value in DESC seems to be split into two lines, with the first line being the name of the device and the + second line being a description of the device. I don't like having the description be across two lines because + it makes formatting ugly and annoying. I'm therefore deciding to put it all on a single line with the second line + being put into parentheses. In simplified mode I'm just stripping the second line entirely. + */ + if (DESC != NULL) { + int lfPos; + const char* line2 = ma_find_char(DESC, '\n', &lfPos); + if (line2 != NULL) { + line2 += 1; /* Skip past the new-line character. */ + + if (pContext->alsa.useVerboseDeviceEnumeration) { + /* Verbose mode. Put the second line in brackets. */ + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), DESC, lfPos); + ma_strcat_s (deviceInfo.name, sizeof(deviceInfo.name), " ("); + ma_strcat_s (deviceInfo.name, sizeof(deviceInfo.name), line2); + ma_strcat_s (deviceInfo.name, sizeof(deviceInfo.name), ")"); + } else { + /* Simplified mode. Strip the second line entirely. */ + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), DESC, lfPos); + } + } else { + /* There's no second line. Just copy the whole description. */ + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), DESC, (size_t)-1); + } + } + + if (!ma_is_device_blacklisted__alsa(deviceType, NAME)) { + cbResult = callback(pContext, deviceType, &deviceInfo, pUserData); + } + + /* + Some devices are both playback and capture, but they are only enumerated by ALSA once. We need to fire the callback + again for the other device type in this case. We do this for known devices and where the IOID hint is NULL, which + means both Input and Output. + */ + if (cbResult) { + if (ma_is_common_device_name__alsa(NAME) || IOID == NULL) { + if (deviceType == ma_device_type_playback) { + if (!ma_is_capture_device_blacklisted__alsa(NAME)) { + cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + } else { + if (!ma_is_playback_device_blacklisted__alsa(NAME)) { + cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + } + } + } + + if (cbResult == MA_FALSE) { + stopEnumeration = MA_TRUE; + } + + next_device: + free(NAME); + free(DESC); + free(IOID); + ppNextDeviceHint += 1; + + /* We need to stop enumeration if the callback returned false. */ + if (stopEnumeration) { + break; + } + } + + ma_free(pUniqueIDs, &pContext->allocationCallbacks); + ((ma_snd_device_name_free_hint_proc)pContext->alsa.snd_device_name_free_hint)((void**)ppDeviceHints); + + ma_mutex_unlock(&pContext->alsa.internalDeviceEnumLock); + + return MA_SUCCESS; +} + + +typedef struct +{ + ma_device_type deviceType; + const ma_device_id* pDeviceID; + ma_share_mode shareMode; + ma_device_info* pDeviceInfo; + ma_bool32 foundDevice; +} ma_context_get_device_info_enum_callback_data__alsa; + +static ma_bool32 ma_context_get_device_info_enum_callback__alsa(ma_context* pContext, ma_device_type deviceType, const ma_device_info* pDeviceInfo, void* pUserData) +{ + ma_context_get_device_info_enum_callback_data__alsa* pData = (ma_context_get_device_info_enum_callback_data__alsa*)pUserData; + MA_ASSERT(pData != NULL); + + (void)pContext; + + if (pData->pDeviceID == NULL && ma_strcmp(pDeviceInfo->id.alsa, "default") == 0) { + ma_strncpy_s(pData->pDeviceInfo->name, sizeof(pData->pDeviceInfo->name), pDeviceInfo->name, (size_t)-1); + pData->foundDevice = MA_TRUE; + } else { + if (pData->deviceType == deviceType && (pData->pDeviceID != NULL && ma_strcmp(pData->pDeviceID->alsa, pDeviceInfo->id.alsa) == 0)) { + ma_strncpy_s(pData->pDeviceInfo->name, sizeof(pData->pDeviceInfo->name), pDeviceInfo->name, (size_t)-1); + pData->foundDevice = MA_TRUE; + } + } + + /* Keep enumerating until we have found the device. */ + return !pData->foundDevice; +} + +static void ma_context_test_rate_and_add_native_data_format__alsa(ma_context* pContext, ma_snd_pcm_t* pPCM, ma_snd_pcm_hw_params_t* pHWParams, ma_format format, ma_uint32 channels, ma_uint32 sampleRate, ma_uint32 flags, ma_device_info* pDeviceInfo) +{ + MA_ASSERT(pPCM != NULL); + MA_ASSERT(pHWParams != NULL); + MA_ASSERT(pDeviceInfo != NULL); + + if (pDeviceInfo->nativeDataFormatCount < ma_countof(pDeviceInfo->nativeDataFormats) && ((ma_snd_pcm_hw_params_test_rate_proc)pContext->alsa.snd_pcm_hw_params_test_rate)(pPCM, pHWParams, sampleRate, 0) == 0) { + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].format = format; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].channels = channels; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].sampleRate = sampleRate; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].flags = flags; + pDeviceInfo->nativeDataFormatCount += 1; + } +} + +static void ma_context_iterate_rates_and_add_native_data_format__alsa(ma_context* pContext, ma_snd_pcm_t* pPCM, ma_snd_pcm_hw_params_t* pHWParams, ma_format format, ma_uint32 channels, ma_uint32 flags, ma_device_info* pDeviceInfo) +{ + ma_uint32 iSampleRate; + unsigned int minSampleRate; + unsigned int maxSampleRate; + int sampleRateDir; /* Not used. Just passed into snd_pcm_hw_params_get_rate_min/max(). */ + + /* There could be a range. */ + ((ma_snd_pcm_hw_params_get_rate_min_proc)pContext->alsa.snd_pcm_hw_params_get_rate_min)(pHWParams, &minSampleRate, &sampleRateDir); + ((ma_snd_pcm_hw_params_get_rate_max_proc)pContext->alsa.snd_pcm_hw_params_get_rate_max)(pHWParams, &maxSampleRate, &sampleRateDir); + + /* Make sure our sample rates are clamped to sane values. Stupid devices like "pulse" will reports rates like "1" which is ridiculous. */ + minSampleRate = ma_clamp(minSampleRate, (unsigned int)ma_standard_sample_rate_min, (unsigned int)ma_standard_sample_rate_max); + maxSampleRate = ma_clamp(maxSampleRate, (unsigned int)ma_standard_sample_rate_min, (unsigned int)ma_standard_sample_rate_max); + + for (iSampleRate = 0; iSampleRate < ma_countof(g_maStandardSampleRatePriorities); iSampleRate += 1) { + ma_uint32 standardSampleRate = g_maStandardSampleRatePriorities[iSampleRate]; + + if (standardSampleRate >= minSampleRate && standardSampleRate <= maxSampleRate) { + ma_context_test_rate_and_add_native_data_format__alsa(pContext, pPCM, pHWParams, format, channels, standardSampleRate, flags, pDeviceInfo); + } + } + + /* Now make sure our min and max rates are included just in case they aren't in the range of our standard rates. */ + if (!ma_is_standard_sample_rate(minSampleRate)) { + ma_context_test_rate_and_add_native_data_format__alsa(pContext, pPCM, pHWParams, format, channels, minSampleRate, flags, pDeviceInfo); + } + + if (!ma_is_standard_sample_rate(maxSampleRate) && maxSampleRate != minSampleRate) { + ma_context_test_rate_and_add_native_data_format__alsa(pContext, pPCM, pHWParams, format, channels, maxSampleRate, flags, pDeviceInfo); + } +} + +static ma_result ma_context_get_device_info__alsa(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + ma_context_get_device_info_enum_callback_data__alsa data; + ma_result result; + int resultALSA; + ma_snd_pcm_t* pPCM; + ma_snd_pcm_hw_params_t* pHWParams; + ma_uint32 iFormat; + ma_uint32 iChannel; + + MA_ASSERT(pContext != NULL); + + /* We just enumerate to find basic information about the device. */ + data.deviceType = deviceType; + data.pDeviceID = pDeviceID; + data.pDeviceInfo = pDeviceInfo; + data.foundDevice = MA_FALSE; + result = ma_context_enumerate_devices__alsa(pContext, ma_context_get_device_info_enum_callback__alsa, &data); + if (result != MA_SUCCESS) { + return result; + } + + if (!data.foundDevice) { + return MA_NO_DEVICE; + } + + if (ma_strcmp(pDeviceInfo->id.alsa, "default") == 0) { + pDeviceInfo->isDefault = MA_TRUE; + } + + /* For detailed info we need to open the device. */ + result = ma_context_open_pcm__alsa(pContext, ma_share_mode_shared, deviceType, pDeviceID, 0, &pPCM); + if (result != MA_SUCCESS) { + return result; + } + + /* We need to initialize a HW parameters object in order to know what formats are supported. */ + pHWParams = (ma_snd_pcm_hw_params_t*)ma_calloc(((ma_snd_pcm_hw_params_sizeof_proc)pContext->alsa.snd_pcm_hw_params_sizeof)(), &pContext->allocationCallbacks); + if (pHWParams == NULL) { + ((ma_snd_pcm_close_proc)pContext->alsa.snd_pcm_close)(pPCM); + return MA_OUT_OF_MEMORY; + } + + resultALSA = ((ma_snd_pcm_hw_params_any_proc)pContext->alsa.snd_pcm_hw_params_any)(pPCM, pHWParams); + if (resultALSA < 0) { + ma_free(pHWParams, &pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pContext->alsa.snd_pcm_close)(pPCM); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to initialize hardware parameters. snd_pcm_hw_params_any() failed."); + return ma_result_from_errno(-resultALSA); + } + + /* + Some ALSA devices can support many permutations of formats, channels and rates. We only support + a fixed number of permutations which means we need to employ some strategies to ensure the best + combinations are returned. An example is the "pulse" device which can do its own data conversion + in software and as a result can support any combination of format, channels and rate. + + We want to ensure that the first data formats are the best. We have a list of favored sample + formats and sample rates, so these will be the basis of our iteration. + */ + + /* Formats. We just iterate over our standard formats and test them, making sure we reset the configuration space each iteration. */ + for (iFormat = 0; iFormat < ma_countof(g_maFormatPriorities); iFormat += 1) { + ma_format format = g_maFormatPriorities[iFormat]; + + /* + For each format we need to make sure we reset the configuration space so we don't return + channel counts and rates that aren't compatible with a format. + */ + ((ma_snd_pcm_hw_params_any_proc)pContext->alsa.snd_pcm_hw_params_any)(pPCM, pHWParams); + + /* Test the format first. If this fails it means the format is not supported and we can skip it. */ + if (((ma_snd_pcm_hw_params_test_format_proc)pContext->alsa.snd_pcm_hw_params_test_format)(pPCM, pHWParams, ma_convert_ma_format_to_alsa_format(format)) == 0) { + /* The format is supported. */ + unsigned int minChannels; + unsigned int maxChannels; + + /* + The configuration space needs to be restricted to this format so we can get an accurate + picture of which sample rates and channel counts are support with this format. + */ + ((ma_snd_pcm_hw_params_set_format_proc)pContext->alsa.snd_pcm_hw_params_set_format)(pPCM, pHWParams, ma_convert_ma_format_to_alsa_format(format)); + + /* Now we need to check for supported channels. */ + ((ma_snd_pcm_hw_params_get_channels_min_proc)pContext->alsa.snd_pcm_hw_params_get_channels_min)(pHWParams, &minChannels); + ((ma_snd_pcm_hw_params_get_channels_max_proc)pContext->alsa.snd_pcm_hw_params_get_channels_max)(pHWParams, &maxChannels); + + if (minChannels > MA_MAX_CHANNELS) { + continue; /* Too many channels. */ + } + if (maxChannels < MA_MIN_CHANNELS) { + continue; /* Not enough channels. */ + } + + /* + Make sure the channel count is clamped. This is mainly intended for the max channels + because some devices can report an unbound maximum. + */ + minChannels = ma_clamp(minChannels, MA_MIN_CHANNELS, MA_MAX_CHANNELS); + maxChannels = ma_clamp(maxChannels, MA_MIN_CHANNELS, MA_MAX_CHANNELS); + + if (minChannels == MA_MIN_CHANNELS && maxChannels == MA_MAX_CHANNELS) { + /* The device supports all channels. Don't iterate over every single one. Instead just set the channels to 0 which means all channels are supported. */ + ma_context_iterate_rates_and_add_native_data_format__alsa(pContext, pPCM, pHWParams, format, 0, 0, pDeviceInfo); /* Intentionally setting the channel count to 0 as that means all channels are supported. */ + } else { + /* The device only supports a specific set of channels. We need to iterate over all of them. */ + for (iChannel = minChannels; iChannel <= maxChannels; iChannel += 1) { + /* Test the channel before applying it to the configuration space. */ + unsigned int channels = iChannel; + + /* Make sure our channel range is reset before testing again or else we'll always fail the test. */ + ((ma_snd_pcm_hw_params_any_proc)pContext->alsa.snd_pcm_hw_params_any)(pPCM, pHWParams); + ((ma_snd_pcm_hw_params_set_format_proc)pContext->alsa.snd_pcm_hw_params_set_format)(pPCM, pHWParams, ma_convert_ma_format_to_alsa_format(format)); + + if (((ma_snd_pcm_hw_params_test_channels_proc)pContext->alsa.snd_pcm_hw_params_test_channels)(pPCM, pHWParams, channels) == 0) { + /* The channel count is supported. */ + + /* The configuration space now needs to be restricted to the channel count before extracting the sample rate. */ + ((ma_snd_pcm_hw_params_set_channels_proc)pContext->alsa.snd_pcm_hw_params_set_channels)(pPCM, pHWParams, channels); + + /* Only after the configuration space has been restricted to the specific channel count should we iterate over our sample rates. */ + ma_context_iterate_rates_and_add_native_data_format__alsa(pContext, pPCM, pHWParams, format, channels, 0, pDeviceInfo); + } else { + /* The channel count is not supported. Skip. */ + } + } + } + } else { + /* The format is not supported. Skip. */ + } + } + + ma_free(pHWParams, &pContext->allocationCallbacks); + + ((ma_snd_pcm_close_proc)pContext->alsa.snd_pcm_close)(pPCM); + return MA_SUCCESS; +} + +static ma_result ma_device_uninit__alsa(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if ((ma_snd_pcm_t*)pDevice->alsa.pPCMCapture) { + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)((ma_snd_pcm_t*)pDevice->alsa.pPCMCapture); + close(pDevice->alsa.wakeupfdCapture); + ma_free(pDevice->alsa.pPollDescriptorsCapture, &pDevice->pContext->allocationCallbacks); + } + + if ((ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback) { + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)((ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback); + close(pDevice->alsa.wakeupfdPlayback); + ma_free(pDevice->alsa.pPollDescriptorsPlayback, &pDevice->pContext->allocationCallbacks); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_init_by_type__alsa(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptor, ma_device_type deviceType) +{ + ma_result result; + int resultALSA; + ma_snd_pcm_t* pPCM; + ma_bool32 isUsingMMap; + ma_snd_pcm_format_t formatALSA; + ma_format internalFormat; + ma_uint32 internalChannels; + ma_uint32 internalSampleRate; + ma_channel internalChannelMap[MA_MAX_CHANNELS]; + ma_uint32 internalPeriodSizeInFrames; + ma_uint32 internalPeriods; + int openMode; + ma_snd_pcm_hw_params_t* pHWParams; + ma_snd_pcm_sw_params_t* pSWParams; + ma_snd_pcm_uframes_t bufferBoundary; + int pollDescriptorCount; + struct pollfd* pPollDescriptors; + int wakeupfd; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(deviceType != ma_device_type_duplex); /* This function should only be called for playback _or_ capture, never duplex. */ + MA_ASSERT(pDevice != NULL); + + formatALSA = ma_convert_ma_format_to_alsa_format(pDescriptor->format); + + openMode = 0; + if (pConfig->alsa.noAutoResample) { + openMode |= MA_SND_PCM_NO_AUTO_RESAMPLE; + } + if (pConfig->alsa.noAutoChannels) { + openMode |= MA_SND_PCM_NO_AUTO_CHANNELS; + } + if (pConfig->alsa.noAutoFormat) { + openMode |= MA_SND_PCM_NO_AUTO_FORMAT; + } + + result = ma_context_open_pcm__alsa(pDevice->pContext, pDescriptor->shareMode, deviceType, pDescriptor->pDeviceID, openMode, &pPCM); + if (result != MA_SUCCESS) { + return result; + } + + + /* Hardware parameters. */ + pHWParams = (ma_snd_pcm_hw_params_t*)ma_calloc(((ma_snd_pcm_hw_params_sizeof_proc)pDevice->pContext->alsa.snd_pcm_hw_params_sizeof)(), &pDevice->pContext->allocationCallbacks); + if (pHWParams == NULL) { + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to allocate memory for hardware parameters."); + return MA_OUT_OF_MEMORY; + } + + resultALSA = ((ma_snd_pcm_hw_params_any_proc)pDevice->pContext->alsa.snd_pcm_hw_params_any)(pPCM, pHWParams); + if (resultALSA < 0) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to initialize hardware parameters. snd_pcm_hw_params_any() failed."); + return ma_result_from_errno(-resultALSA); + } + + /* MMAP Mode. Try using interleaved MMAP access. If this fails, fall back to standard readi/writei. */ + isUsingMMap = MA_FALSE; +#if 0 /* NOTE: MMAP mode temporarily disabled. */ + if (deviceType != ma_device_type_capture) { /* <-- Disabling MMAP mode for capture devices because I apparently do not have a device that supports it which means I can't test it... Contributions welcome. */ + if (!pConfig->alsa.noMMap) { + if (((ma_snd_pcm_hw_params_set_access_proc)pDevice->pContext->alsa.snd_pcm_hw_params_set_access)(pPCM, pHWParams, MA_SND_PCM_ACCESS_MMAP_INTERLEAVED) == 0) { + pDevice->alsa.isUsingMMap = MA_TRUE; + } + } + } +#endif + + if (!isUsingMMap) { + resultALSA = ((ma_snd_pcm_hw_params_set_access_proc)pDevice->pContext->alsa.snd_pcm_hw_params_set_access)(pPCM, pHWParams, MA_SND_PCM_ACCESS_RW_INTERLEAVED); + if (resultALSA < 0) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to set access mode to neither SND_PCM_ACCESS_MMAP_INTERLEAVED nor SND_PCM_ACCESS_RW_INTERLEAVED. snd_pcm_hw_params_set_access() failed."); + return ma_result_from_errno(-resultALSA); + } + } + + /* + Most important properties first. The documentation for OSS (yes, I know this is ALSA!) recommends format, channels, then sample rate. I can't + find any documentation for ALSA specifically, so I'm going to copy the recommendation for OSS. + */ + + /* Format. */ + { + /* + At this point we should have a list of supported formats, so now we need to find the best one. We first check if the requested format is + supported, and if so, use that one. If it's not supported, we just run though a list of formats and try to find the best one. + */ + if (formatALSA == MA_SND_PCM_FORMAT_UNKNOWN || ((ma_snd_pcm_hw_params_test_format_proc)pDevice->pContext->alsa.snd_pcm_hw_params_test_format)(pPCM, pHWParams, formatALSA) != 0) { + /* We're either requesting the native format or the specified format is not supported. */ + size_t iFormat; + + formatALSA = MA_SND_PCM_FORMAT_UNKNOWN; + for (iFormat = 0; iFormat < ma_countof(g_maFormatPriorities); ++iFormat) { + if (((ma_snd_pcm_hw_params_test_format_proc)pDevice->pContext->alsa.snd_pcm_hw_params_test_format)(pPCM, pHWParams, ma_convert_ma_format_to_alsa_format(g_maFormatPriorities[iFormat])) == 0) { + formatALSA = ma_convert_ma_format_to_alsa_format(g_maFormatPriorities[iFormat]); + break; + } + } + + if (formatALSA == MA_SND_PCM_FORMAT_UNKNOWN) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Format not supported. The device does not support any miniaudio formats."); + return MA_FORMAT_NOT_SUPPORTED; + } + } + + resultALSA = ((ma_snd_pcm_hw_params_set_format_proc)pDevice->pContext->alsa.snd_pcm_hw_params_set_format)(pPCM, pHWParams, formatALSA); + if (resultALSA < 0) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Format not supported. snd_pcm_hw_params_set_format() failed."); + return ma_result_from_errno(-resultALSA); + } + + internalFormat = ma_format_from_alsa(formatALSA); + if (internalFormat == ma_format_unknown) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] The chosen format is not supported by miniaudio."); + return MA_FORMAT_NOT_SUPPORTED; + } + } + + /* Channels. */ + { + unsigned int channels = pDescriptor->channels; + if (channels == 0) { + channels = MA_DEFAULT_CHANNELS; + } + + resultALSA = ((ma_snd_pcm_hw_params_set_channels_near_proc)pDevice->pContext->alsa.snd_pcm_hw_params_set_channels_near)(pPCM, pHWParams, &channels); + if (resultALSA < 0) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to set channel count. snd_pcm_hw_params_set_channels_near() failed."); + return ma_result_from_errno(-resultALSA); + } + + internalChannels = (ma_uint32)channels; + } + + /* Sample Rate */ + { + unsigned int sampleRate; + + /* + It appears there's either a bug in ALSA, a bug in some drivers, or I'm doing something silly; but having resampling enabled causes + problems with some device configurations when used in conjunction with MMAP access mode. To fix this problem we need to disable + resampling. + + To reproduce this problem, open the "plug:dmix" device, and set the sample rate to 44100. Internally, it looks like dmix uses a + sample rate of 48000. The hardware parameters will get set correctly with no errors, but it looks like the 44100 -> 48000 resampling + doesn't work properly - but only with MMAP access mode. You will notice skipping/crackling in the audio, and it'll run at a slightly + faster rate. + + miniaudio has built-in support for sample rate conversion (albeit low quality at the moment), so disabling resampling should be fine + for us. The only problem is that it won't be taking advantage of any kind of hardware-accelerated resampling and it won't be very + good quality until I get a chance to improve the quality of miniaudio's software sample rate conversion. + + I don't currently know if the dmix plugin is the only one with this error. Indeed, this is the only one I've been able to reproduce + this error with. In the future, we may want to restrict the disabling of resampling to only known bad plugins. + */ + ((ma_snd_pcm_hw_params_set_rate_resample_proc)pDevice->pContext->alsa.snd_pcm_hw_params_set_rate_resample)(pPCM, pHWParams, 0); + + sampleRate = pDescriptor->sampleRate; + if (sampleRate == 0) { + sampleRate = MA_DEFAULT_SAMPLE_RATE; + } + + resultALSA = ((ma_snd_pcm_hw_params_set_rate_near_proc)pDevice->pContext->alsa.snd_pcm_hw_params_set_rate_near)(pPCM, pHWParams, &sampleRate, 0); + if (resultALSA < 0) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Sample rate not supported. snd_pcm_hw_params_set_rate_near() failed."); + return ma_result_from_errno(-resultALSA); + } + + internalSampleRate = (ma_uint32)sampleRate; + } + + /* Periods. */ + { + ma_uint32 periods = pDescriptor->periodCount; + + resultALSA = ((ma_snd_pcm_hw_params_set_periods_near_proc)pDevice->pContext->alsa.snd_pcm_hw_params_set_periods_near)(pPCM, pHWParams, &periods, NULL); + if (resultALSA < 0) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to set period count. snd_pcm_hw_params_set_periods_near() failed."); + return ma_result_from_errno(-resultALSA); + } + + internalPeriods = periods; + } + + /* Buffer Size */ + { + ma_snd_pcm_uframes_t actualBufferSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptor, internalSampleRate, pConfig->performanceProfile) * internalPeriods; + + resultALSA = ((ma_snd_pcm_hw_params_set_buffer_size_near_proc)pDevice->pContext->alsa.snd_pcm_hw_params_set_buffer_size_near)(pPCM, pHWParams, &actualBufferSizeInFrames); + if (resultALSA < 0) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to set buffer size for device. snd_pcm_hw_params_set_buffer_size() failed."); + return ma_result_from_errno(-resultALSA); + } + + internalPeriodSizeInFrames = actualBufferSizeInFrames / internalPeriods; + } + + /* Apply hardware parameters. */ + resultALSA = ((ma_snd_pcm_hw_params_proc)pDevice->pContext->alsa.snd_pcm_hw_params)(pPCM, pHWParams); + if (resultALSA < 0) { + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to set hardware parameters. snd_pcm_hw_params() failed."); + return ma_result_from_errno(-resultALSA); + } + + ma_free(pHWParams, &pDevice->pContext->allocationCallbacks); + pHWParams = NULL; + + + /* Software parameters. */ + pSWParams = (ma_snd_pcm_sw_params_t*)ma_calloc(((ma_snd_pcm_sw_params_sizeof_proc)pDevice->pContext->alsa.snd_pcm_sw_params_sizeof)(), &pDevice->pContext->allocationCallbacks); + if (pSWParams == NULL) { + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to allocate memory for software parameters."); + return MA_OUT_OF_MEMORY; + } + + resultALSA = ((ma_snd_pcm_sw_params_current_proc)pDevice->pContext->alsa.snd_pcm_sw_params_current)(pPCM, pSWParams); + if (resultALSA < 0) { + ma_free(pSWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to initialize software parameters. snd_pcm_sw_params_current() failed."); + return ma_result_from_errno(-resultALSA); + } + + resultALSA = ((ma_snd_pcm_sw_params_set_avail_min_proc)pDevice->pContext->alsa.snd_pcm_sw_params_set_avail_min)(pPCM, pSWParams, ma_prev_power_of_2(internalPeriodSizeInFrames)); + if (resultALSA < 0) { + ma_free(pSWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] snd_pcm_sw_params_set_avail_min() failed."); + return ma_result_from_errno(-resultALSA); + } + + resultALSA = ((ma_snd_pcm_sw_params_get_boundary_proc)pDevice->pContext->alsa.snd_pcm_sw_params_get_boundary)(pSWParams, &bufferBoundary); + if (resultALSA < 0) { + bufferBoundary = internalPeriodSizeInFrames * internalPeriods; + } + + if (deviceType == ma_device_type_playback && !isUsingMMap) { /* Only playback devices in writei/readi mode need a start threshold. */ + /* + Subtle detail here with the start threshold. When in playback-only mode (no full-duplex) we can set the start threshold to + the size of a period. But for full-duplex we need to set it such that it is at least two periods. + */ + resultALSA = ((ma_snd_pcm_sw_params_set_start_threshold_proc)pDevice->pContext->alsa.snd_pcm_sw_params_set_start_threshold)(pPCM, pSWParams, internalPeriodSizeInFrames*2); + if (resultALSA < 0) { + ma_free(pSWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to set start threshold for playback device. snd_pcm_sw_params_set_start_threshold() failed."); + return ma_result_from_errno(-resultALSA); + } + + resultALSA = ((ma_snd_pcm_sw_params_set_stop_threshold_proc)pDevice->pContext->alsa.snd_pcm_sw_params_set_stop_threshold)(pPCM, pSWParams, bufferBoundary); + if (resultALSA < 0) { /* Set to boundary to loop instead of stop in the event of an xrun. */ + ma_free(pSWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to set stop threshold for playback device. snd_pcm_sw_params_set_stop_threshold() failed."); + return ma_result_from_errno(-resultALSA); + } + } + + resultALSA = ((ma_snd_pcm_sw_params_proc)pDevice->pContext->alsa.snd_pcm_sw_params)(pPCM, pSWParams); + if (resultALSA < 0) { + ma_free(pSWParams, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to set software parameters. snd_pcm_sw_params() failed."); + return ma_result_from_errno(-resultALSA); + } + + ma_free(pSWParams, &pDevice->pContext->allocationCallbacks); + pSWParams = NULL; + + + /* Grab the internal channel map. For now we're not going to bother trying to change the channel map and instead just do it ourselves. */ + { + ma_snd_pcm_chmap_t* pChmap = NULL; + if (pDevice->pContext->alsa.snd_pcm_get_chmap != NULL) { + pChmap = ((ma_snd_pcm_get_chmap_proc)pDevice->pContext->alsa.snd_pcm_get_chmap)(pPCM); + } + + if (pChmap != NULL) { + ma_uint32 iChannel; + + /* There are cases where the returned channel map can have a different channel count than was returned by snd_pcm_hw_params_set_channels_near(). */ + if (pChmap->channels >= internalChannels) { + /* Drop excess channels. */ + for (iChannel = 0; iChannel < internalChannels; ++iChannel) { + internalChannelMap[iChannel] = ma_convert_alsa_channel_position_to_ma_channel(pChmap->pos[iChannel]); + } + } else { + ma_uint32 i; + + /* + Excess channels use defaults. Do an initial fill with defaults, overwrite the first pChmap->channels, validate to ensure there are no duplicate + channels. If validation fails, fall back to defaults. + */ + ma_bool32 isValid = MA_TRUE; + + /* Fill with defaults. */ + ma_channel_map_init_standard(ma_standard_channel_map_alsa, internalChannelMap, ma_countof(internalChannelMap), internalChannels); + + /* Overwrite first pChmap->channels channels. */ + for (iChannel = 0; iChannel < pChmap->channels; ++iChannel) { + internalChannelMap[iChannel] = ma_convert_alsa_channel_position_to_ma_channel(pChmap->pos[iChannel]); + } + + /* Validate. */ + for (i = 0; i < internalChannels && isValid; ++i) { + ma_uint32 j; + for (j = i+1; j < internalChannels; ++j) { + if (internalChannelMap[i] == internalChannelMap[j]) { + isValid = MA_FALSE; + break; + } + } + } + + /* If our channel map is invalid, fall back to defaults. */ + if (!isValid) { + ma_channel_map_init_standard(ma_standard_channel_map_alsa, internalChannelMap, ma_countof(internalChannelMap), internalChannels); + } + } + + free(pChmap); + pChmap = NULL; + } else { + /* Could not retrieve the channel map. Fall back to a hard-coded assumption. */ + ma_channel_map_init_standard(ma_standard_channel_map_alsa, internalChannelMap, ma_countof(internalChannelMap), internalChannels); + } + } + + + /* + We need to retrieve the poll descriptors so we can use poll() to wait for data to become + available for reading or writing. There's no well defined maximum for this so we're just going + to allocate this on the heap. + */ + pollDescriptorCount = ((ma_snd_pcm_poll_descriptors_count_proc)pDevice->pContext->alsa.snd_pcm_poll_descriptors_count)(pPCM); + if (pollDescriptorCount <= 0) { + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to retrieve poll descriptors count."); + return MA_ERROR; + } + + pPollDescriptors = (struct pollfd*)ma_malloc(sizeof(*pPollDescriptors) * (pollDescriptorCount + 1), &pDevice->pContext->allocationCallbacks); /* +1 because we want room for the wakeup descriptor. */ + if (pPollDescriptors == NULL) { + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to allocate memory for poll descriptors."); + return MA_OUT_OF_MEMORY; + } + + /* + We need an eventfd to wakeup from poll() and avoid a deadlock in situations where the driver + never returns from writei() and readi(). This has been observed with the "pulse" device. + */ + wakeupfd = eventfd(0, 0); + if (wakeupfd < 0) { + ma_free(pPollDescriptors, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to create eventfd for poll wakeup."); + return ma_result_from_errno(errno); + } + + /* We'll place the wakeup fd at the start of the buffer. */ + pPollDescriptors[0].fd = wakeupfd; + pPollDescriptors[0].events = POLLIN; /* We only care about waiting to read from the wakeup file descriptor. */ + pPollDescriptors[0].revents = 0; + + /* We can now extract the PCM poll descriptors which we place after the wakeup descriptor. */ + pollDescriptorCount = ((ma_snd_pcm_poll_descriptors_proc)pDevice->pContext->alsa.snd_pcm_poll_descriptors)(pPCM, pPollDescriptors + 1, pollDescriptorCount); /* +1 because we want to place these descriptors after the wakeup descriptor. */ + if (pollDescriptorCount <= 0) { + close(wakeupfd); + ma_free(pPollDescriptors, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to retrieve poll descriptors."); + return MA_ERROR; + } + + if (deviceType == ma_device_type_capture) { + pDevice->alsa.pollDescriptorCountCapture = pollDescriptorCount; + pDevice->alsa.pPollDescriptorsCapture = pPollDescriptors; + pDevice->alsa.wakeupfdCapture = wakeupfd; + } else { + pDevice->alsa.pollDescriptorCountPlayback = pollDescriptorCount; + pDevice->alsa.pPollDescriptorsPlayback = pPollDescriptors; + pDevice->alsa.wakeupfdPlayback = wakeupfd; + } + + + /* We're done. Prepare the device. */ + resultALSA = ((ma_snd_pcm_prepare_proc)pDevice->pContext->alsa.snd_pcm_prepare)(pPCM); + if (resultALSA < 0) { + close(wakeupfd); + ma_free(pPollDescriptors, &pDevice->pContext->allocationCallbacks); + ((ma_snd_pcm_close_proc)pDevice->pContext->alsa.snd_pcm_close)(pPCM); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to prepare device."); + return ma_result_from_errno(-resultALSA); + } + + + if (deviceType == ma_device_type_capture) { + pDevice->alsa.pPCMCapture = (ma_ptr)pPCM; + pDevice->alsa.isUsingMMapCapture = isUsingMMap; + } else { + pDevice->alsa.pPCMPlayback = (ma_ptr)pPCM; + pDevice->alsa.isUsingMMapPlayback = isUsingMMap; + } + + pDescriptor->format = internalFormat; + pDescriptor->channels = internalChannels; + pDescriptor->sampleRate = internalSampleRate; + ma_channel_map_copy(pDescriptor->channelMap, internalChannelMap, ma_min(internalChannels, MA_MAX_CHANNELS)); + pDescriptor->periodSizeInFrames = internalPeriodSizeInFrames; + pDescriptor->periodCount = internalPeriods; + + return MA_SUCCESS; +} + +static ma_result ma_device_init__alsa(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + MA_ASSERT(pDevice != NULL); + + MA_ZERO_OBJECT(&pDevice->alsa); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ma_result result = ma_device_init_by_type__alsa(pDevice, pConfig, pDescriptorCapture, ma_device_type_capture); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_result result = ma_device_init_by_type__alsa(pDevice, pConfig, pDescriptorPlayback, ma_device_type_playback); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_start__alsa(ma_device* pDevice) +{ + int resultALSA; + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + resultALSA = ((ma_snd_pcm_start_proc)pDevice->pContext->alsa.snd_pcm_start)((ma_snd_pcm_t*)pDevice->alsa.pPCMCapture); + if (resultALSA < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to start capture device."); + return ma_result_from_errno(-resultALSA); + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + /* + When data is written to the device we wait for the device to get ready to receive data with poll(). In my testing + I have observed that poll() can sometimes block forever unless the device is started explicitly with snd_pcm_start() + or some data is written with snd_pcm_writei(). + + To resolve this I've decided to do an explicit start with snd_pcm_start(). The problem with this is that the device + is started without any data in the internal buffer which will result in an immediate underrun. If instead we were + to call into snd_pcm_writei() in an attempt to prevent the underrun, we would run the risk of a weird deadlock + issue as documented inside ma_device_write__alsa(). + */ + resultALSA = ((ma_snd_pcm_start_proc)pDevice->pContext->alsa.snd_pcm_start)((ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback); + if (resultALSA < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to start playback device."); + return ma_result_from_errno(-resultALSA); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__alsa(ma_device* pDevice) +{ + /* + The stop callback will get called on the worker thread after read/write__alsa() has returned. At this point there is + a small chance that our wakeupfd has not been cleared. We'll clear that out now if applicable. + */ + int resultPoll; + int resultRead; + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Dropping capture device...\n"); + ((ma_snd_pcm_drop_proc)pDevice->pContext->alsa.snd_pcm_drop)((ma_snd_pcm_t*)pDevice->alsa.pPCMCapture); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Dropping capture device successful.\n"); + + /* We need to prepare the device again, otherwise we won't be able to restart the device. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Preparing capture device...\n"); + if (((ma_snd_pcm_prepare_proc)pDevice->pContext->alsa.snd_pcm_prepare)((ma_snd_pcm_t*)pDevice->alsa.pPCMCapture) < 0) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Preparing capture device failed.\n"); + } else { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Preparing capture device successful.\n"); + } + + /* Clear the wakeupfd. */ + resultPoll = poll((struct pollfd*)pDevice->alsa.pPollDescriptorsCapture, 1, 0); + if (resultPoll > 0) { + ma_uint64 t; + resultRead = read(((struct pollfd*)pDevice->alsa.pPollDescriptorsCapture)[0].fd, &t, sizeof(t)); + if (resultRead != sizeof(t)) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Failed to read from capture wakeupfd. read() = %d\n", resultRead); + } + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Dropping playback device...\n"); + ((ma_snd_pcm_drop_proc)pDevice->pContext->alsa.snd_pcm_drop)((ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Dropping playback device successful.\n"); + + /* We need to prepare the device again, otherwise we won't be able to restart the device. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Preparing playback device...\n"); + if (((ma_snd_pcm_prepare_proc)pDevice->pContext->alsa.snd_pcm_prepare)((ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback) < 0) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Preparing playback device failed.\n"); + } else { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Preparing playback device successful.\n"); + } + + /* Clear the wakeupfd. */ + resultPoll = poll((struct pollfd*)pDevice->alsa.pPollDescriptorsPlayback, 1, 0); + if (resultPoll > 0) { + ma_uint64 t; + resultRead = read(((struct pollfd*)pDevice->alsa.pPollDescriptorsPlayback)[0].fd, &t, sizeof(t)); + if (resultRead != sizeof(t)) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Failed to read from playback wakeupfd. read() = %d\n", resultRead); + } + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_wait__alsa(ma_device* pDevice, ma_snd_pcm_t* pPCM, struct pollfd* pPollDescriptors, int pollDescriptorCount, short requiredEvent) +{ + for (;;) { + unsigned short revents; + int resultALSA; + int resultPoll = poll(pPollDescriptors, pollDescriptorCount, -1); + if (resultPoll < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[ALSA] poll() failed.\n"); + + /* + There have been reports that poll() is returning an error randomly and that instead of + returning an error, simply trying again will work. I'm experimenting with adopting this + advice. + */ + continue; + /*return ma_result_from_errno(errno);*/ + } + + /* + Before checking the ALSA poll descriptor flag we need to check if the wakeup descriptor + has had it's POLLIN flag set. If so, we need to actually read the data and then exit the + function. The wakeup descriptor will be the first item in the descriptors buffer. + */ + if ((pPollDescriptors[0].revents & POLLIN) != 0) { + ma_uint64 t; + int resultRead = read(pPollDescriptors[0].fd, &t, sizeof(t)); /* <-- Important that we read here so that the next write() does not block. */ + if (resultRead < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] read() failed.\n"); + return ma_result_from_errno(errno); + } + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] POLLIN set for wakeupfd\n"); + return MA_DEVICE_NOT_STARTED; + } + + /* + Getting here means that some data should be able to be read. We need to use ALSA to + translate the revents flags for us. + */ + resultALSA = ((ma_snd_pcm_poll_descriptors_revents_proc)pDevice->pContext->alsa.snd_pcm_poll_descriptors_revents)(pPCM, pPollDescriptors + 1, pollDescriptorCount - 1, &revents); /* +1, -1 to ignore the wakeup descriptor. */ + if (resultALSA < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] snd_pcm_poll_descriptors_revents() failed.\n"); + return ma_result_from_errno(-resultALSA); + } + + if ((revents & POLLERR) != 0) { + ma_snd_pcm_state_t state = ((ma_snd_pcm_state_proc)pDevice->pContext->alsa.snd_pcm_state)(pPCM); + if (state == MA_SND_PCM_STATE_XRUN) { + /* The PCM is in a xrun state. This will be recovered from at a higher level. We can disregard this. */ + } else { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[ALSA] POLLERR detected. status = %d\n", ((ma_snd_pcm_state_proc)pDevice->pContext->alsa.snd_pcm_state)(pPCM)); + } + } + + if ((revents & requiredEvent) == requiredEvent) { + break; /* We're done. Data available for reading or writing. */ + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_wait_read__alsa(ma_device* pDevice) +{ + return ma_device_wait__alsa(pDevice, (ma_snd_pcm_t*)pDevice->alsa.pPCMCapture, (struct pollfd*)pDevice->alsa.pPollDescriptorsCapture, pDevice->alsa.pollDescriptorCountCapture + 1, POLLIN); /* +1 to account for the wakeup descriptor. */ +} + +static ma_result ma_device_wait_write__alsa(ma_device* pDevice) +{ + return ma_device_wait__alsa(pDevice, (ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback, (struct pollfd*)pDevice->alsa.pPollDescriptorsPlayback, pDevice->alsa.pollDescriptorCountPlayback + 1, POLLOUT); /* +1 to account for the wakeup descriptor. */ +} + +static ma_result ma_device_read__alsa(ma_device* pDevice, void* pFramesOut, ma_uint32 frameCount, ma_uint32* pFramesRead) +{ + ma_snd_pcm_sframes_t resultALSA = 0; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pFramesOut != NULL); + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + while (ma_device_get_state(pDevice) == ma_device_state_started) { + ma_result result; + + /* The first thing to do is wait for data to become available for reading. This will return an error code if the device has been stopped. */ + result = ma_device_wait_read__alsa(pDevice); + if (result != MA_SUCCESS) { + return result; + } + + /* Getting here means we should have data available. */ + resultALSA = ((ma_snd_pcm_readi_proc)pDevice->pContext->alsa.snd_pcm_readi)((ma_snd_pcm_t*)pDevice->alsa.pPCMCapture, pFramesOut, frameCount); + if (resultALSA >= 0) { + break; /* Success. */ + } else { + if (resultALSA == -EAGAIN) { + /*ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "EGAIN (read)\n");*/ + continue; /* Try again. */ + } else if (resultALSA == -EPIPE) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "EPIPE (read)\n"); + + /* Overrun. Recover and try again. If this fails we need to return an error. */ + resultALSA = ((ma_snd_pcm_recover_proc)pDevice->pContext->alsa.snd_pcm_recover)((ma_snd_pcm_t*)pDevice->alsa.pPCMCapture, resultALSA, MA_TRUE); + if (resultALSA < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to recover device after overrun."); + return ma_result_from_errno((int)-resultALSA); + } + + resultALSA = ((ma_snd_pcm_start_proc)pDevice->pContext->alsa.snd_pcm_start)((ma_snd_pcm_t*)pDevice->alsa.pPCMCapture); + if (resultALSA < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to start device after underrun."); + return ma_result_from_errno((int)-resultALSA); + } + + continue; /* Try reading again. */ + } + } + } + + if (pFramesRead != NULL) { + *pFramesRead = resultALSA; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_write__alsa(ma_device* pDevice, const void* pFrames, ma_uint32 frameCount, ma_uint32* pFramesWritten) +{ + ma_snd_pcm_sframes_t resultALSA = 0; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pFrames != NULL); + + if (pFramesWritten != NULL) { + *pFramesWritten = 0; + } + + while (ma_device_get_state(pDevice) == ma_device_state_started) { + ma_result result; + + /* The first thing to do is wait for space to become available for writing. This will return an error code if the device has been stopped. */ + result = ma_device_wait_write__alsa(pDevice); + if (result != MA_SUCCESS) { + return result; + } + + resultALSA = ((ma_snd_pcm_writei_proc)pDevice->pContext->alsa.snd_pcm_writei)((ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback, pFrames, frameCount); + if (resultALSA >= 0) { + break; /* Success. */ + } else { + if (resultALSA == -EAGAIN) { + /*ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "EGAIN (write)\n");*/ + continue; /* Try again. */ + } else if (resultALSA == -EPIPE) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "EPIPE (write)\n"); + + /* Underrun. Recover and try again. If this fails we need to return an error. */ + resultALSA = ((ma_snd_pcm_recover_proc)pDevice->pContext->alsa.snd_pcm_recover)((ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback, resultALSA, MA_TRUE); /* MA_TRUE=silent (don't print anything on error). */ + if (resultALSA < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to recover device after underrun."); + return ma_result_from_errno((int)-resultALSA); + } + + /* + In my testing I have had a situation where writei() does not automatically restart the device even though I've set it + up as such in the software parameters. What will happen is writei() will block indefinitely even though the number of + frames is well beyond the auto-start threshold. To work around this I've needed to add an explicit start here. Not sure + if this is me just being stupid and not recovering the device properly, but this definitely feels like something isn't + quite right here. + */ + resultALSA = ((ma_snd_pcm_start_proc)pDevice->pContext->alsa.snd_pcm_start)((ma_snd_pcm_t*)pDevice->alsa.pPCMPlayback); + if (resultALSA < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] Failed to start device after underrun."); + return ma_result_from_errno((int)-resultALSA); + } + + continue; /* Try writing again. */ + } + } + } + + if (pFramesWritten != NULL) { + *pFramesWritten = resultALSA; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_data_loop_wakeup__alsa(ma_device* pDevice) +{ + ma_uint64 t = 1; + int resultWrite = 0; + + MA_ASSERT(pDevice != NULL); + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Waking up...\n"); + + /* Write to an eventfd to trigger a wakeup from poll() and abort any reading or writing. */ + if (pDevice->alsa.pPollDescriptorsCapture != NULL) { + resultWrite = write(pDevice->alsa.wakeupfdCapture, &t, sizeof(t)); + } + if (pDevice->alsa.pPollDescriptorsPlayback != NULL) { + resultWrite = write(pDevice->alsa.wakeupfdPlayback, &t, sizeof(t)); + } + + if (resultWrite < 0) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[ALSA] write() failed.\n"); + return ma_result_from_errno(errno); + } + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[ALSA] Waking up completed successfully.\n"); + + return MA_SUCCESS; +} + +static ma_result ma_context_uninit__alsa(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_alsa); + + /* Clean up memory for memory leak checkers. */ + ((ma_snd_config_update_free_global_proc)pContext->alsa.snd_config_update_free_global)(); + +#ifndef MA_NO_RUNTIME_LINKING + ma_dlclose(ma_context_get_log(pContext), pContext->alsa.asoundSO); +#endif + + ma_mutex_uninit(&pContext->alsa.internalDeviceEnumLock); + + return MA_SUCCESS; +} + +static ma_result ma_context_init__alsa(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + ma_result result; +#ifndef MA_NO_RUNTIME_LINKING + const char* libasoundNames[] = { + "libasound.so.2", + "libasound.so" + }; + size_t i; + + for (i = 0; i < ma_countof(libasoundNames); ++i) { + pContext->alsa.asoundSO = ma_dlopen(ma_context_get_log(pContext), libasoundNames[i]); + if (pContext->alsa.asoundSO != NULL) { + break; + } + } + + if (pContext->alsa.asoundSO == NULL) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "[ALSA] Failed to open shared object.\n"); + return MA_NO_BACKEND; + } + + pContext->alsa.snd_pcm_open = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_open"); + pContext->alsa.snd_pcm_close = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_close"); + pContext->alsa.snd_pcm_hw_params_sizeof = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_sizeof"); + pContext->alsa.snd_pcm_hw_params_any = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_any"); + pContext->alsa.snd_pcm_hw_params_set_format = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_format"); + pContext->alsa.snd_pcm_hw_params_set_format_first = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_format_first"); + pContext->alsa.snd_pcm_hw_params_get_format_mask = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_format_mask"); + pContext->alsa.snd_pcm_hw_params_set_channels = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_channels"); + pContext->alsa.snd_pcm_hw_params_set_channels_near = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_channels_near"); + pContext->alsa.snd_pcm_hw_params_set_channels_minmax = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_channels_minmax"); + pContext->alsa.snd_pcm_hw_params_set_rate_resample = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_rate_resample"); + pContext->alsa.snd_pcm_hw_params_set_rate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_rate"); + pContext->alsa.snd_pcm_hw_params_set_rate_near = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_rate_near"); + pContext->alsa.snd_pcm_hw_params_set_buffer_size_near = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_buffer_size_near"); + pContext->alsa.snd_pcm_hw_params_set_periods_near = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_periods_near"); + pContext->alsa.snd_pcm_hw_params_set_access = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_set_access"); + pContext->alsa.snd_pcm_hw_params_get_format = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_format"); + pContext->alsa.snd_pcm_hw_params_get_channels = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_channels"); + pContext->alsa.snd_pcm_hw_params_get_channels_min = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_channels_min"); + pContext->alsa.snd_pcm_hw_params_get_channels_max = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_channels_max"); + pContext->alsa.snd_pcm_hw_params_get_rate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_rate"); + pContext->alsa.snd_pcm_hw_params_get_rate_min = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_rate_min"); + pContext->alsa.snd_pcm_hw_params_get_rate_max = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_rate_max"); + pContext->alsa.snd_pcm_hw_params_get_buffer_size = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_buffer_size"); + pContext->alsa.snd_pcm_hw_params_get_periods = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_periods"); + pContext->alsa.snd_pcm_hw_params_get_access = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_get_access"); + pContext->alsa.snd_pcm_hw_params_test_format = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_test_format"); + pContext->alsa.snd_pcm_hw_params_test_channels = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_test_channels"); + pContext->alsa.snd_pcm_hw_params_test_rate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params_test_rate"); + pContext->alsa.snd_pcm_hw_params = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_hw_params"); + pContext->alsa.snd_pcm_sw_params_sizeof = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_sw_params_sizeof"); + pContext->alsa.snd_pcm_sw_params_current = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_sw_params_current"); + pContext->alsa.snd_pcm_sw_params_get_boundary = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_sw_params_get_boundary"); + pContext->alsa.snd_pcm_sw_params_set_avail_min = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_sw_params_set_avail_min"); + pContext->alsa.snd_pcm_sw_params_set_start_threshold = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_sw_params_set_start_threshold"); + pContext->alsa.snd_pcm_sw_params_set_stop_threshold = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_sw_params_set_stop_threshold"); + pContext->alsa.snd_pcm_sw_params = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_sw_params"); + pContext->alsa.snd_pcm_format_mask_sizeof = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_format_mask_sizeof"); + pContext->alsa.snd_pcm_format_mask_test = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_format_mask_test"); + pContext->alsa.snd_pcm_get_chmap = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_get_chmap"); + pContext->alsa.snd_pcm_state = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_state"); + pContext->alsa.snd_pcm_prepare = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_prepare"); + pContext->alsa.snd_pcm_start = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_start"); + pContext->alsa.snd_pcm_drop = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_drop"); + pContext->alsa.snd_pcm_drain = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_drain"); + pContext->alsa.snd_pcm_reset = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_reset"); + pContext->alsa.snd_device_name_hint = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_device_name_hint"); + pContext->alsa.snd_device_name_get_hint = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_device_name_get_hint"); + pContext->alsa.snd_card_get_index = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_card_get_index"); + pContext->alsa.snd_device_name_free_hint = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_device_name_free_hint"); + pContext->alsa.snd_pcm_mmap_begin = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_mmap_begin"); + pContext->alsa.snd_pcm_mmap_commit = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_mmap_commit"); + pContext->alsa.snd_pcm_recover = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_recover"); + pContext->alsa.snd_pcm_readi = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_readi"); + pContext->alsa.snd_pcm_writei = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_writei"); + pContext->alsa.snd_pcm_avail = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_avail"); + pContext->alsa.snd_pcm_avail_update = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_avail_update"); + pContext->alsa.snd_pcm_wait = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_wait"); + pContext->alsa.snd_pcm_nonblock = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_nonblock"); + pContext->alsa.snd_pcm_info = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_info"); + pContext->alsa.snd_pcm_info_sizeof = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_info_sizeof"); + pContext->alsa.snd_pcm_info_get_name = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_info_get_name"); + pContext->alsa.snd_pcm_poll_descriptors = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_poll_descriptors"); + pContext->alsa.snd_pcm_poll_descriptors_count = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_poll_descriptors_count"); + pContext->alsa.snd_pcm_poll_descriptors_revents = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_pcm_poll_descriptors_revents"); + pContext->alsa.snd_config_update_free_global = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->alsa.asoundSO, "snd_config_update_free_global"); +#else + /* The system below is just for type safety. */ + ma_snd_pcm_open_proc _snd_pcm_open = snd_pcm_open; + ma_snd_pcm_close_proc _snd_pcm_close = snd_pcm_close; + ma_snd_pcm_hw_params_sizeof_proc _snd_pcm_hw_params_sizeof = snd_pcm_hw_params_sizeof; + ma_snd_pcm_hw_params_any_proc _snd_pcm_hw_params_any = snd_pcm_hw_params_any; + ma_snd_pcm_hw_params_set_format_proc _snd_pcm_hw_params_set_format = snd_pcm_hw_params_set_format; + ma_snd_pcm_hw_params_set_format_first_proc _snd_pcm_hw_params_set_format_first = snd_pcm_hw_params_set_format_first; + ma_snd_pcm_hw_params_get_format_mask_proc _snd_pcm_hw_params_get_format_mask = snd_pcm_hw_params_get_format_mask; + ma_snd_pcm_hw_params_set_channels_proc _snd_pcm_hw_params_set_channels = snd_pcm_hw_params_set_channels; + ma_snd_pcm_hw_params_set_channels_near_proc _snd_pcm_hw_params_set_channels_near = snd_pcm_hw_params_set_channels_near; + ma_snd_pcm_hw_params_set_rate_resample_proc _snd_pcm_hw_params_set_rate_resample = snd_pcm_hw_params_set_rate_resample; + ma_snd_pcm_hw_params_set_rate_near _snd_pcm_hw_params_set_rate = snd_pcm_hw_params_set_rate; + ma_snd_pcm_hw_params_set_rate_near_proc _snd_pcm_hw_params_set_rate_near = snd_pcm_hw_params_set_rate_near; + ma_snd_pcm_hw_params_set_rate_minmax_proc _snd_pcm_hw_params_set_rate_minmax = snd_pcm_hw_params_set_rate_minmax; + ma_snd_pcm_hw_params_set_buffer_size_near_proc _snd_pcm_hw_params_set_buffer_size_near = snd_pcm_hw_params_set_buffer_size_near; + ma_snd_pcm_hw_params_set_periods_near_proc _snd_pcm_hw_params_set_periods_near = snd_pcm_hw_params_set_periods_near; + ma_snd_pcm_hw_params_set_access_proc _snd_pcm_hw_params_set_access = snd_pcm_hw_params_set_access; + ma_snd_pcm_hw_params_get_format_proc _snd_pcm_hw_params_get_format = snd_pcm_hw_params_get_format; + ma_snd_pcm_hw_params_get_channels_proc _snd_pcm_hw_params_get_channels = snd_pcm_hw_params_get_channels; + ma_snd_pcm_hw_params_get_channels_min_proc _snd_pcm_hw_params_get_channels_min = snd_pcm_hw_params_get_channels_min; + ma_snd_pcm_hw_params_get_channels_max_proc _snd_pcm_hw_params_get_channels_max = snd_pcm_hw_params_get_channels_max; + ma_snd_pcm_hw_params_get_rate_proc _snd_pcm_hw_params_get_rate = snd_pcm_hw_params_get_rate; + ma_snd_pcm_hw_params_get_rate_min_proc _snd_pcm_hw_params_get_rate_min = snd_pcm_hw_params_get_rate_min; + ma_snd_pcm_hw_params_get_rate_max_proc _snd_pcm_hw_params_get_rate_max = snd_pcm_hw_params_get_rate_max; + ma_snd_pcm_hw_params_get_buffer_size_proc _snd_pcm_hw_params_get_buffer_size = snd_pcm_hw_params_get_buffer_size; + ma_snd_pcm_hw_params_get_periods_proc _snd_pcm_hw_params_get_periods = snd_pcm_hw_params_get_periods; + ma_snd_pcm_hw_params_get_access_proc _snd_pcm_hw_params_get_access = snd_pcm_hw_params_get_access; + ma_snd_pcm_hw_params_test_format_proc _snd_pcm_hw_params_test_format = snd_pcm_hw_params_test_format; + ma_snd_pcm_hw_params_test_channels_proc _snd_pcm_hw_params_test_channels = snd_pcm_hw_params_test_channels; + ma_snd_pcm_hw_params_test_rate_proc _snd_pcm_hw_params_test_rate = snd_pcm_hw_params_test_rate; + ma_snd_pcm_hw_params_proc _snd_pcm_hw_params = snd_pcm_hw_params; + ma_snd_pcm_sw_params_sizeof_proc _snd_pcm_sw_params_sizeof = snd_pcm_sw_params_sizeof; + ma_snd_pcm_sw_params_current_proc _snd_pcm_sw_params_current = snd_pcm_sw_params_current; + ma_snd_pcm_sw_params_get_boundary_proc _snd_pcm_sw_params_get_boundary = snd_pcm_sw_params_get_boundary; + ma_snd_pcm_sw_params_set_avail_min_proc _snd_pcm_sw_params_set_avail_min = snd_pcm_sw_params_set_avail_min; + ma_snd_pcm_sw_params_set_start_threshold_proc _snd_pcm_sw_params_set_start_threshold = snd_pcm_sw_params_set_start_threshold; + ma_snd_pcm_sw_params_set_stop_threshold_proc _snd_pcm_sw_params_set_stop_threshold = snd_pcm_sw_params_set_stop_threshold; + ma_snd_pcm_sw_params_proc _snd_pcm_sw_params = snd_pcm_sw_params; + ma_snd_pcm_format_mask_sizeof_proc _snd_pcm_format_mask_sizeof = snd_pcm_format_mask_sizeof; + ma_snd_pcm_format_mask_test_proc _snd_pcm_format_mask_test = snd_pcm_format_mask_test; + ma_snd_pcm_get_chmap_proc _snd_pcm_get_chmap = snd_pcm_get_chmap; + ma_snd_pcm_state_proc _snd_pcm_state = snd_pcm_state; + ma_snd_pcm_prepare_proc _snd_pcm_prepare = snd_pcm_prepare; + ma_snd_pcm_start_proc _snd_pcm_start = snd_pcm_start; + ma_snd_pcm_drop_proc _snd_pcm_drop = snd_pcm_drop; + ma_snd_pcm_drain_proc _snd_pcm_drain = snd_pcm_drain; + ma_snd_pcm_reset_proc _snd_pcm_reset = snd_pcm_reset; + ma_snd_device_name_hint_proc _snd_device_name_hint = snd_device_name_hint; + ma_snd_device_name_get_hint_proc _snd_device_name_get_hint = snd_device_name_get_hint; + ma_snd_card_get_index_proc _snd_card_get_index = snd_card_get_index; + ma_snd_device_name_free_hint_proc _snd_device_name_free_hint = snd_device_name_free_hint; + ma_snd_pcm_mmap_begin_proc _snd_pcm_mmap_begin = snd_pcm_mmap_begin; + ma_snd_pcm_mmap_commit_proc _snd_pcm_mmap_commit = snd_pcm_mmap_commit; + ma_snd_pcm_recover_proc _snd_pcm_recover = snd_pcm_recover; + ma_snd_pcm_readi_proc _snd_pcm_readi = snd_pcm_readi; + ma_snd_pcm_writei_proc _snd_pcm_writei = snd_pcm_writei; + ma_snd_pcm_avail_proc _snd_pcm_avail = snd_pcm_avail; + ma_snd_pcm_avail_update_proc _snd_pcm_avail_update = snd_pcm_avail_update; + ma_snd_pcm_wait_proc _snd_pcm_wait = snd_pcm_wait; + ma_snd_pcm_nonblock_proc _snd_pcm_nonblock = snd_pcm_nonblock; + ma_snd_pcm_info_proc _snd_pcm_info = snd_pcm_info; + ma_snd_pcm_info_sizeof_proc _snd_pcm_info_sizeof = snd_pcm_info_sizeof; + ma_snd_pcm_info_get_name_proc _snd_pcm_info_get_name = snd_pcm_info_get_name; + ma_snd_pcm_poll_descriptors _snd_pcm_poll_descriptors = snd_pcm_poll_descriptors; + ma_snd_pcm_poll_descriptors_count _snd_pcm_poll_descriptors_count = snd_pcm_poll_descriptors_count; + ma_snd_pcm_poll_descriptors_revents _snd_pcm_poll_descriptors_revents = snd_pcm_poll_descriptors_revents; + ma_snd_config_update_free_global_proc _snd_config_update_free_global = snd_config_update_free_global; + + pContext->alsa.snd_pcm_open = (ma_proc)_snd_pcm_open; + pContext->alsa.snd_pcm_close = (ma_proc)_snd_pcm_close; + pContext->alsa.snd_pcm_hw_params_sizeof = (ma_proc)_snd_pcm_hw_params_sizeof; + pContext->alsa.snd_pcm_hw_params_any = (ma_proc)_snd_pcm_hw_params_any; + pContext->alsa.snd_pcm_hw_params_set_format = (ma_proc)_snd_pcm_hw_params_set_format; + pContext->alsa.snd_pcm_hw_params_set_format_first = (ma_proc)_snd_pcm_hw_params_set_format_first; + pContext->alsa.snd_pcm_hw_params_get_format_mask = (ma_proc)_snd_pcm_hw_params_get_format_mask; + pContext->alsa.snd_pcm_hw_params_set_channels = (ma_proc)_snd_pcm_hw_params_set_channels; + pContext->alsa.snd_pcm_hw_params_set_channels_near = (ma_proc)_snd_pcm_hw_params_set_channels_near; + pContext->alsa.snd_pcm_hw_params_set_channels_minmax = (ma_proc)_snd_pcm_hw_params_set_channels_minmax; + pContext->alsa.snd_pcm_hw_params_set_rate_resample = (ma_proc)_snd_pcm_hw_params_set_rate_resample; + pContext->alsa.snd_pcm_hw_params_set_rate = (ma_proc)_snd_pcm_hw_params_set_rate; + pContext->alsa.snd_pcm_hw_params_set_rate_near = (ma_proc)_snd_pcm_hw_params_set_rate_near; + pContext->alsa.snd_pcm_hw_params_set_buffer_size_near = (ma_proc)_snd_pcm_hw_params_set_buffer_size_near; + pContext->alsa.snd_pcm_hw_params_set_periods_near = (ma_proc)_snd_pcm_hw_params_set_periods_near; + pContext->alsa.snd_pcm_hw_params_set_access = (ma_proc)_snd_pcm_hw_params_set_access; + pContext->alsa.snd_pcm_hw_params_get_format = (ma_proc)_snd_pcm_hw_params_get_format; + pContext->alsa.snd_pcm_hw_params_get_channels = (ma_proc)_snd_pcm_hw_params_get_channels; + pContext->alsa.snd_pcm_hw_params_get_channels_min = (ma_proc)_snd_pcm_hw_params_get_channels_min; + pContext->alsa.snd_pcm_hw_params_get_channels_max = (ma_proc)_snd_pcm_hw_params_get_channels_max; + pContext->alsa.snd_pcm_hw_params_get_rate = (ma_proc)_snd_pcm_hw_params_get_rate; + pContext->alsa.snd_pcm_hw_params_get_rate_min = (ma_proc)_snd_pcm_hw_params_get_rate_min; + pContext->alsa.snd_pcm_hw_params_get_rate_max = (ma_proc)_snd_pcm_hw_params_get_rate_max; + pContext->alsa.snd_pcm_hw_params_get_buffer_size = (ma_proc)_snd_pcm_hw_params_get_buffer_size; + pContext->alsa.snd_pcm_hw_params_get_periods = (ma_proc)_snd_pcm_hw_params_get_periods; + pContext->alsa.snd_pcm_hw_params_get_access = (ma_proc)_snd_pcm_hw_params_get_access; + pContext->alsa.snd_pcm_hw_params_test_format = (ma_proc)_snd_pcm_hw_params_test_format; + pContext->alsa.snd_pcm_hw_params_test_channels = (ma_proc)_snd_pcm_hw_params_test_channels; + pContext->alsa.snd_pcm_hw_params_test_rate = (ma_proc)_snd_pcm_hw_params_test_rate; + pContext->alsa.snd_pcm_hw_params = (ma_proc)_snd_pcm_hw_params; + pContext->alsa.snd_pcm_sw_params_sizeof = (ma_proc)_snd_pcm_sw_params_sizeof; + pContext->alsa.snd_pcm_sw_params_current = (ma_proc)_snd_pcm_sw_params_current; + pContext->alsa.snd_pcm_sw_params_get_boundary = (ma_proc)_snd_pcm_sw_params_get_boundary; + pContext->alsa.snd_pcm_sw_params_set_avail_min = (ma_proc)_snd_pcm_sw_params_set_avail_min; + pContext->alsa.snd_pcm_sw_params_set_start_threshold = (ma_proc)_snd_pcm_sw_params_set_start_threshold; + pContext->alsa.snd_pcm_sw_params_set_stop_threshold = (ma_proc)_snd_pcm_sw_params_set_stop_threshold; + pContext->alsa.snd_pcm_sw_params = (ma_proc)_snd_pcm_sw_params; + pContext->alsa.snd_pcm_format_mask_sizeof = (ma_proc)_snd_pcm_format_mask_sizeof; + pContext->alsa.snd_pcm_format_mask_test = (ma_proc)_snd_pcm_format_mask_test; + pContext->alsa.snd_pcm_get_chmap = (ma_proc)_snd_pcm_get_chmap; + pContext->alsa.snd_pcm_state = (ma_proc)_snd_pcm_state; + pContext->alsa.snd_pcm_prepare = (ma_proc)_snd_pcm_prepare; + pContext->alsa.snd_pcm_start = (ma_proc)_snd_pcm_start; + pContext->alsa.snd_pcm_drop = (ma_proc)_snd_pcm_drop; + pContext->alsa.snd_pcm_drain = (ma_proc)_snd_pcm_drain; + pContext->alsa.snd_pcm_reset = (ma_proc)_snd_pcm_reset; + pContext->alsa.snd_device_name_hint = (ma_proc)_snd_device_name_hint; + pContext->alsa.snd_device_name_get_hint = (ma_proc)_snd_device_name_get_hint; + pContext->alsa.snd_card_get_index = (ma_proc)_snd_card_get_index; + pContext->alsa.snd_device_name_free_hint = (ma_proc)_snd_device_name_free_hint; + pContext->alsa.snd_pcm_mmap_begin = (ma_proc)_snd_pcm_mmap_begin; + pContext->alsa.snd_pcm_mmap_commit = (ma_proc)_snd_pcm_mmap_commit; + pContext->alsa.snd_pcm_recover = (ma_proc)_snd_pcm_recover; + pContext->alsa.snd_pcm_readi = (ma_proc)_snd_pcm_readi; + pContext->alsa.snd_pcm_writei = (ma_proc)_snd_pcm_writei; + pContext->alsa.snd_pcm_avail = (ma_proc)_snd_pcm_avail; + pContext->alsa.snd_pcm_avail_update = (ma_proc)_snd_pcm_avail_update; + pContext->alsa.snd_pcm_wait = (ma_proc)_snd_pcm_wait; + pContext->alsa.snd_pcm_nonblock = (ma_proc)_snd_pcm_nonblock; + pContext->alsa.snd_pcm_info = (ma_proc)_snd_pcm_info; + pContext->alsa.snd_pcm_info_sizeof = (ma_proc)_snd_pcm_info_sizeof; + pContext->alsa.snd_pcm_info_get_name = (ma_proc)_snd_pcm_info_get_name; + pContext->alsa.snd_pcm_poll_descriptors = (ma_proc)_snd_pcm_poll_descriptors; + pContext->alsa.snd_pcm_poll_descriptors_count = (ma_proc)_snd_pcm_poll_descriptors_count; + pContext->alsa.snd_pcm_poll_descriptors_revents = (ma_proc)_snd_pcm_poll_descriptors_revents; + pContext->alsa.snd_config_update_free_global = (ma_proc)_snd_config_update_free_global; +#endif + + pContext->alsa.useVerboseDeviceEnumeration = pConfig->alsa.useVerboseDeviceEnumeration; + + result = ma_mutex_init(&pContext->alsa.internalDeviceEnumLock); + if (result != MA_SUCCESS) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[ALSA] WARNING: Failed to initialize mutex for internal device enumeration."); + return result; + } + + pCallbacks->onContextInit = ma_context_init__alsa; + pCallbacks->onContextUninit = ma_context_uninit__alsa; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__alsa; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__alsa; + pCallbacks->onDeviceInit = ma_device_init__alsa; + pCallbacks->onDeviceUninit = ma_device_uninit__alsa; + pCallbacks->onDeviceStart = ma_device_start__alsa; + pCallbacks->onDeviceStop = ma_device_stop__alsa; + pCallbacks->onDeviceRead = ma_device_read__alsa; + pCallbacks->onDeviceWrite = ma_device_write__alsa; + pCallbacks->onDeviceDataLoop = NULL; + pCallbacks->onDeviceDataLoopWakeup = ma_device_data_loop_wakeup__alsa; + + return MA_SUCCESS; +} +#endif /* MA_HAS_ALSA */ + + + +/****************************************************************************** + +PulseAudio Backend + +******************************************************************************/ +#ifdef MA_HAS_PULSEAUDIO +/* +The PulseAudio API, along with Apple's Core Audio, is the worst of the mainstream audio APIs. This is a brief description of what's going on +in the PulseAudio backend. I apologize if this gets a bit ranty for your liking - you might want to skip this discussion. + +PulseAudio has something they call the "Simple API", which unfortunately isn't suitable for miniaudio. I've not seen anywhere where it +allows you to enumerate over devices, nor does it seem to support the ability to stop and start streams. Looking at the documentation, it +appears as though the stream is constantly running and you prevent sound from being emitted or captured by simply not calling the read or +write functions. This is not a professional solution as it would be much better to *actually* stop the underlying stream. Perhaps the +simple API has some smarts to do this automatically, but I'm not sure. Another limitation with the simple API is that it seems inefficient +when you want to have multiple streams to a single context. For these reasons, miniaudio is not using the simple API. + +Since we're not using the simple API, we're left with the asynchronous API as our only other option. And boy, is this where it starts to +get fun, and I don't mean that in a good way... + +The problems start with the very name of the API - "asynchronous". Yes, this is an asynchronous oriented API which means your commands +don't immediately take effect. You instead need to issue your commands, and then wait for them to complete. The waiting mechanism is +enabled through the use of a "main loop". In the asynchronous API you cannot get away from the main loop, and the main loop is where almost +all of PulseAudio's problems stem from. + +When you first initialize PulseAudio you need an object referred to as "main loop". You can implement this yourself by defining your own +vtable, but it's much easier to just use one of the built-in main loop implementations. There's two generic implementations called +pa_mainloop and pa_threaded_mainloop, and another implementation specific to GLib called pa_glib_mainloop. We're using pa_threaded_mainloop +because it simplifies management of the worker thread. The idea of the main loop object is pretty self explanatory - you're supposed to use +it to implement a worker thread which runs in a loop. The main loop is where operations are actually executed. + +To initialize the main loop, you just use `pa_threaded_mainloop_new()`. This is the first function you'll call. You can then get a pointer +to the vtable with `pa_threaded_mainloop_get_api()` (the main loop vtable is called `pa_mainloop_api`). Again, you can bypass the threaded +main loop object entirely and just implement `pa_mainloop_api` directly, but there's no need for it unless you're doing something extremely +specialized such as if you want to integrate it into your application's existing main loop infrastructure. + +(EDIT 2021-01-26: miniaudio is no longer using `pa_threaded_mainloop` due to this issue: https://github.com/mackron/miniaudio/issues/262. +It is now using `pa_mainloop` which turns out to be a simpler solution anyway. The rest of this rant still applies, however.) + +Once you have your main loop vtable (the `pa_mainloop_api` object) you can create the PulseAudio context. This is very similar to +miniaudio's context and they map to each other quite well. You have one context to many streams, which is basically the same as miniaudio's +one `ma_context` to many `ma_device`s. Here's where it starts to get annoying, however. When you first create the PulseAudio context, which +is done with `pa_context_new()`, it's not actually connected to anything. When you connect, you call `pa_context_connect()`. However, if +you remember, PulseAudio is an asynchronous API. That means you cannot just assume the context is connected after `pa_context_context()` +has returned. You instead need to wait for it to connect. To do this, you need to either wait for a callback to get fired, which you can +set with `pa_context_set_state_callback()`, or you can continuously poll the context's state. Either way, you need to run this in a loop. +All objects from here out are created from the context, and, I believe, you can't be creating these objects until the context is connected. +This waiting loop is therefore unavoidable. In order for the waiting to ever complete, however, the main loop needs to be running. Before +attempting to connect the context, the main loop needs to be started with `pa_threaded_mainloop_start()`. + +The reason for this asynchronous design is to support cases where you're connecting to a remote server, say through a local network or an +internet connection. However, the *VAST* majority of cases don't involve this at all - they just connect to a local "server" running on the +host machine. The fact that this would be the default rather than making `pa_context_connect()` synchronous tends to boggle the mind. + +Once the context has been created and connected you can start creating a stream. A PulseAudio stream is analogous to miniaudio's device. +The initialization of a stream is fairly standard - you configure some attributes (analogous to miniaudio's device config) and then call +`pa_stream_new()` to actually create it. Here is where we start to get into "operations". When configuring the stream, you can get +information about the source (such as sample format, sample rate, etc.), however it's not synchronous. Instead, a `pa_operation` object +is returned from `pa_context_get_source_info_by_name()` (capture) or `pa_context_get_sink_info_by_name()` (playback). Then, you need to +run a loop (again!) to wait for the operation to complete which you can determine via a callback or polling, just like we did with the +context. Then, as an added bonus, you need to decrement the reference counter of the `pa_operation` object to ensure memory is cleaned up. +All of that just to retrieve basic information about a device! + +Once the basic information about the device has been retrieved, miniaudio can now create the stream with `ma_stream_new()`. Like the +context, this needs to be connected. But we need to be careful here, because we're now about to introduce one of the most horrific design +choices in PulseAudio. + +PulseAudio allows you to specify a callback that is fired when data can be written to or read from a stream. The language is important here +because PulseAudio takes it literally, specifically the "can be". You would think these callbacks would be appropriate as the place for +writing and reading data to and from the stream, and that would be right, except when it's not. When you initialize the stream, you can +set a flag that tells PulseAudio to not start the stream automatically. This is required because miniaudio does not auto-start devices +straight after initialization - you need to call `ma_device_start()` manually. The problem is that even when this flag is specified, +PulseAudio will immediately fire its write or read callback. This is *technically* correct (based on the wording in the documentation) +because indeed, data *can* be written at this point. The problem is that it's not *practical*. It makes sense that the write/read callback +would be where a program will want to write or read data to or from the stream, but when it's called before the application has even +requested that the stream be started, it's just not practical because the program probably isn't ready for any kind of data delivery at +that point (it may still need to load files or whatnot). Instead, this callback should only be fired when the application requests the +stream be started which is how it works with literally *every* other callback-based audio API. Since miniaudio forbids firing of the data +callback until the device has been started (as it should be with *all* callback based APIs), logic needs to be added to ensure miniaudio +doesn't just blindly fire the application-defined data callback from within the PulseAudio callback before the stream has actually been +started. The device state is used for this - if the state is anything other than `ma_device_state_starting` or `ma_device_state_started`, the main data +callback is not fired. + +This, unfortunately, is not the end of the problems with the PulseAudio write callback. Any normal callback based audio API will +continuously fire the callback at regular intervals based on the size of the internal buffer. This will only ever be fired when the device +is running, and will be fired regardless of whether or not the user actually wrote anything to the device/stream. This not the case in +PulseAudio. In PulseAudio, the data callback will *only* be called if you wrote something to it previously. That means, if you don't call +`pa_stream_write()`, the callback will not get fired. On the surface you wouldn't think this would matter because you should be always +writing data, and if you don't have anything to write, just write silence. That's fine until you want to drain the stream. You see, if +you're continuously writing data to the stream, the stream will never get drained! That means in order to drain the stream, you need to +*not* write data to it! But remember, when you don't write data to the stream, the callback won't get fired again! Why is draining +important? Because that's how we've defined stopping to work in miniaudio. In miniaudio, stopping the device requires it to be drained +before returning from ma_device_stop(). So we've stopped the device, which requires us to drain, but draining requires us to *not* write +data to the stream (or else it won't ever complete draining), but not writing to the stream means the callback won't get fired again! + +This becomes a problem when stopping and then restarting the device. When the device is stopped, it's drained, which requires us to *not* +write anything to the stream. But then, since we didn't write anything to it, the write callback will *never* get called again if we just +resume the stream naively. This means that starting the stream requires us to write data to the stream from outside the callback. This +disconnect is something PulseAudio has got seriously wrong - there should only ever be a single source of data delivery, that being the +callback. (I have tried using `pa_stream_flush()` to trigger the write callback to fire, but this just doesn't work for some reason.) + +Once you've created the stream, you need to connect it which involves the whole waiting procedure. This is the same process as the context, +only this time you'll poll for the state with `pa_stream_get_status()`. The starting and stopping of a streaming is referred to as +"corking" in PulseAudio. The analogy is corking a barrel. To start the stream, you uncork it, to stop it you cork it. Personally I think +it's silly - why would you not just call it "starting" and "stopping" like any other normal audio API? Anyway, the act of corking is, you +guessed it, asynchronous. This means you'll need our waiting loop as usual. Again, why this asynchronous design is the default is +absolutely beyond me. Would it really be that hard to just make it run synchronously? + +Teardown is pretty simple (what?!). It's just a matter of calling the relevant `_unref()` function on each object in reverse order that +they were initialized in. + +That's about it from the PulseAudio side. A bit ranty, I know, but they really need to fix that main loop and callback system. They're +embarrassingly unpractical. The main loop thing is an easy fix - have synchronous versions of all APIs. If an application wants these to +run asynchronously, they can execute them in a separate thread themselves. The desire to run these asynchronously is such a niche +requirement - it makes no sense to make it the default. The stream write callback needs to be change, or an alternative provided, that is +constantly fired, regardless of whether or not `pa_stream_write()` has been called, and it needs to take a pointer to a buffer as a +parameter which the program just writes to directly rather than having to call `pa_stream_writable_size()` and `pa_stream_write()`. These +changes alone will change PulseAudio from one of the worst audio APIs to one of the best. +*/ + + +/* +It is assumed pulseaudio.h is available when linking at compile time. When linking at compile time, we use the declarations in the header +to check for type safety. We cannot do this when linking at run time because the header might not be available. +*/ +#ifdef MA_NO_RUNTIME_LINKING + +/* pulseaudio.h marks some functions with "inline" which isn't always supported. Need to emulate it. */ +#if !defined(__cplusplus) + #if defined(__STRICT_ANSI__) + #if !defined(inline) + #define inline __inline__ __attribute__((always_inline)) + #define MA_INLINE_DEFINED + #endif + #endif +#endif +#include +#if defined(MA_INLINE_DEFINED) + #undef inline + #undef MA_INLINE_DEFINED +#endif + +#define MA_PA_OK PA_OK +#define MA_PA_ERR_ACCESS PA_ERR_ACCESS +#define MA_PA_ERR_INVALID PA_ERR_INVALID +#define MA_PA_ERR_NOENTITY PA_ERR_NOENTITY +#define MA_PA_ERR_NOTSUPPORTED PA_ERR_NOTSUPPORTED + +#define MA_PA_CHANNELS_MAX PA_CHANNELS_MAX +#define MA_PA_RATE_MAX PA_RATE_MAX + +typedef pa_context_flags_t ma_pa_context_flags_t; +#define MA_PA_CONTEXT_NOFLAGS PA_CONTEXT_NOFLAGS +#define MA_PA_CONTEXT_NOAUTOSPAWN PA_CONTEXT_NOAUTOSPAWN +#define MA_PA_CONTEXT_NOFAIL PA_CONTEXT_NOFAIL + +typedef pa_stream_flags_t ma_pa_stream_flags_t; +#define MA_PA_STREAM_NOFLAGS PA_STREAM_NOFLAGS +#define MA_PA_STREAM_START_CORKED PA_STREAM_START_CORKED +#define MA_PA_STREAM_INTERPOLATE_TIMING PA_STREAM_INTERPOLATE_TIMING +#define MA_PA_STREAM_NOT_MONOTONIC PA_STREAM_NOT_MONOTONIC +#define MA_PA_STREAM_AUTO_TIMING_UPDATE PA_STREAM_AUTO_TIMING_UPDATE +#define MA_PA_STREAM_NO_REMAP_CHANNELS PA_STREAM_NO_REMAP_CHANNELS +#define MA_PA_STREAM_NO_REMIX_CHANNELS PA_STREAM_NO_REMIX_CHANNELS +#define MA_PA_STREAM_FIX_FORMAT PA_STREAM_FIX_FORMAT +#define MA_PA_STREAM_FIX_RATE PA_STREAM_FIX_RATE +#define MA_PA_STREAM_FIX_CHANNELS PA_STREAM_FIX_CHANNELS +#define MA_PA_STREAM_DONT_MOVE PA_STREAM_DONT_MOVE +#define MA_PA_STREAM_VARIABLE_RATE PA_STREAM_VARIABLE_RATE +#define MA_PA_STREAM_PEAK_DETECT PA_STREAM_PEAK_DETECT +#define MA_PA_STREAM_START_MUTED PA_STREAM_START_MUTED +#define MA_PA_STREAM_ADJUST_LATENCY PA_STREAM_ADJUST_LATENCY +#define MA_PA_STREAM_EARLY_REQUESTS PA_STREAM_EARLY_REQUESTS +#define MA_PA_STREAM_DONT_INHIBIT_AUTO_SUSPEND PA_STREAM_DONT_INHIBIT_AUTO_SUSPEND +#define MA_PA_STREAM_START_UNMUTED PA_STREAM_START_UNMUTED +#define MA_PA_STREAM_FAIL_ON_SUSPEND PA_STREAM_FAIL_ON_SUSPEND +#define MA_PA_STREAM_RELATIVE_VOLUME PA_STREAM_RELATIVE_VOLUME +#define MA_PA_STREAM_PASSTHROUGH PA_STREAM_PASSTHROUGH + +typedef pa_sink_flags_t ma_pa_sink_flags_t; +#define MA_PA_SINK_NOFLAGS PA_SINK_NOFLAGS +#define MA_PA_SINK_HW_VOLUME_CTRL PA_SINK_HW_VOLUME_CTRL +#define MA_PA_SINK_LATENCY PA_SINK_LATENCY +#define MA_PA_SINK_HARDWARE PA_SINK_HARDWARE +#define MA_PA_SINK_NETWORK PA_SINK_NETWORK +#define MA_PA_SINK_HW_MUTE_CTRL PA_SINK_HW_MUTE_CTRL +#define MA_PA_SINK_DECIBEL_VOLUME PA_SINK_DECIBEL_VOLUME +#define MA_PA_SINK_FLAT_VOLUME PA_SINK_FLAT_VOLUME +#define MA_PA_SINK_DYNAMIC_LATENCY PA_SINK_DYNAMIC_LATENCY +#define MA_PA_SINK_SET_FORMATS PA_SINK_SET_FORMATS + +typedef pa_source_flags_t ma_pa_source_flags_t; +#define MA_PA_SOURCE_NOFLAGS PA_SOURCE_NOFLAGS +#define MA_PA_SOURCE_HW_VOLUME_CTRL PA_SOURCE_HW_VOLUME_CTRL +#define MA_PA_SOURCE_LATENCY PA_SOURCE_LATENCY +#define MA_PA_SOURCE_HARDWARE PA_SOURCE_HARDWARE +#define MA_PA_SOURCE_NETWORK PA_SOURCE_NETWORK +#define MA_PA_SOURCE_HW_MUTE_CTRL PA_SOURCE_HW_MUTE_CTRL +#define MA_PA_SOURCE_DECIBEL_VOLUME PA_SOURCE_DECIBEL_VOLUME +#define MA_PA_SOURCE_DYNAMIC_LATENCY PA_SOURCE_DYNAMIC_LATENCY +#define MA_PA_SOURCE_FLAT_VOLUME PA_SOURCE_FLAT_VOLUME + +typedef pa_context_state_t ma_pa_context_state_t; +#define MA_PA_CONTEXT_UNCONNECTED PA_CONTEXT_UNCONNECTED +#define MA_PA_CONTEXT_CONNECTING PA_CONTEXT_CONNECTING +#define MA_PA_CONTEXT_AUTHORIZING PA_CONTEXT_AUTHORIZING +#define MA_PA_CONTEXT_SETTING_NAME PA_CONTEXT_SETTING_NAME +#define MA_PA_CONTEXT_READY PA_CONTEXT_READY +#define MA_PA_CONTEXT_FAILED PA_CONTEXT_FAILED +#define MA_PA_CONTEXT_TERMINATED PA_CONTEXT_TERMINATED + +typedef pa_stream_state_t ma_pa_stream_state_t; +#define MA_PA_STREAM_UNCONNECTED PA_STREAM_UNCONNECTED +#define MA_PA_STREAM_CREATING PA_STREAM_CREATING +#define MA_PA_STREAM_READY PA_STREAM_READY +#define MA_PA_STREAM_FAILED PA_STREAM_FAILED +#define MA_PA_STREAM_TERMINATED PA_STREAM_TERMINATED + +typedef pa_operation_state_t ma_pa_operation_state_t; +#define MA_PA_OPERATION_RUNNING PA_OPERATION_RUNNING +#define MA_PA_OPERATION_DONE PA_OPERATION_DONE +#define MA_PA_OPERATION_CANCELLED PA_OPERATION_CANCELLED + +typedef pa_sink_state_t ma_pa_sink_state_t; +#define MA_PA_SINK_INVALID_STATE PA_SINK_INVALID_STATE +#define MA_PA_SINK_RUNNING PA_SINK_RUNNING +#define MA_PA_SINK_IDLE PA_SINK_IDLE +#define MA_PA_SINK_SUSPENDED PA_SINK_SUSPENDED + +typedef pa_source_state_t ma_pa_source_state_t; +#define MA_PA_SOURCE_INVALID_STATE PA_SOURCE_INVALID_STATE +#define MA_PA_SOURCE_RUNNING PA_SOURCE_RUNNING +#define MA_PA_SOURCE_IDLE PA_SOURCE_IDLE +#define MA_PA_SOURCE_SUSPENDED PA_SOURCE_SUSPENDED + +typedef pa_seek_mode_t ma_pa_seek_mode_t; +#define MA_PA_SEEK_RELATIVE PA_SEEK_RELATIVE +#define MA_PA_SEEK_ABSOLUTE PA_SEEK_ABSOLUTE +#define MA_PA_SEEK_RELATIVE_ON_READ PA_SEEK_RELATIVE_ON_READ +#define MA_PA_SEEK_RELATIVE_END PA_SEEK_RELATIVE_END + +typedef pa_channel_position_t ma_pa_channel_position_t; +#define MA_PA_CHANNEL_POSITION_INVALID PA_CHANNEL_POSITION_INVALID +#define MA_PA_CHANNEL_POSITION_MONO PA_CHANNEL_POSITION_MONO +#define MA_PA_CHANNEL_POSITION_FRONT_LEFT PA_CHANNEL_POSITION_FRONT_LEFT +#define MA_PA_CHANNEL_POSITION_FRONT_RIGHT PA_CHANNEL_POSITION_FRONT_RIGHT +#define MA_PA_CHANNEL_POSITION_FRONT_CENTER PA_CHANNEL_POSITION_FRONT_CENTER +#define MA_PA_CHANNEL_POSITION_REAR_CENTER PA_CHANNEL_POSITION_REAR_CENTER +#define MA_PA_CHANNEL_POSITION_REAR_LEFT PA_CHANNEL_POSITION_REAR_LEFT +#define MA_PA_CHANNEL_POSITION_REAR_RIGHT PA_CHANNEL_POSITION_REAR_RIGHT +#define MA_PA_CHANNEL_POSITION_LFE PA_CHANNEL_POSITION_LFE +#define MA_PA_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER PA_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER +#define MA_PA_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER PA_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER +#define MA_PA_CHANNEL_POSITION_SIDE_LEFT PA_CHANNEL_POSITION_SIDE_LEFT +#define MA_PA_CHANNEL_POSITION_SIDE_RIGHT PA_CHANNEL_POSITION_SIDE_RIGHT +#define MA_PA_CHANNEL_POSITION_AUX0 PA_CHANNEL_POSITION_AUX0 +#define MA_PA_CHANNEL_POSITION_AUX1 PA_CHANNEL_POSITION_AUX1 +#define MA_PA_CHANNEL_POSITION_AUX2 PA_CHANNEL_POSITION_AUX2 +#define MA_PA_CHANNEL_POSITION_AUX3 PA_CHANNEL_POSITION_AUX3 +#define MA_PA_CHANNEL_POSITION_AUX4 PA_CHANNEL_POSITION_AUX4 +#define MA_PA_CHANNEL_POSITION_AUX5 PA_CHANNEL_POSITION_AUX5 +#define MA_PA_CHANNEL_POSITION_AUX6 PA_CHANNEL_POSITION_AUX6 +#define MA_PA_CHANNEL_POSITION_AUX7 PA_CHANNEL_POSITION_AUX7 +#define MA_PA_CHANNEL_POSITION_AUX8 PA_CHANNEL_POSITION_AUX8 +#define MA_PA_CHANNEL_POSITION_AUX9 PA_CHANNEL_POSITION_AUX9 +#define MA_PA_CHANNEL_POSITION_AUX10 PA_CHANNEL_POSITION_AUX10 +#define MA_PA_CHANNEL_POSITION_AUX11 PA_CHANNEL_POSITION_AUX11 +#define MA_PA_CHANNEL_POSITION_AUX12 PA_CHANNEL_POSITION_AUX12 +#define MA_PA_CHANNEL_POSITION_AUX13 PA_CHANNEL_POSITION_AUX13 +#define MA_PA_CHANNEL_POSITION_AUX14 PA_CHANNEL_POSITION_AUX14 +#define MA_PA_CHANNEL_POSITION_AUX15 PA_CHANNEL_POSITION_AUX15 +#define MA_PA_CHANNEL_POSITION_AUX16 PA_CHANNEL_POSITION_AUX16 +#define MA_PA_CHANNEL_POSITION_AUX17 PA_CHANNEL_POSITION_AUX17 +#define MA_PA_CHANNEL_POSITION_AUX18 PA_CHANNEL_POSITION_AUX18 +#define MA_PA_CHANNEL_POSITION_AUX19 PA_CHANNEL_POSITION_AUX19 +#define MA_PA_CHANNEL_POSITION_AUX20 PA_CHANNEL_POSITION_AUX20 +#define MA_PA_CHANNEL_POSITION_AUX21 PA_CHANNEL_POSITION_AUX21 +#define MA_PA_CHANNEL_POSITION_AUX22 PA_CHANNEL_POSITION_AUX22 +#define MA_PA_CHANNEL_POSITION_AUX23 PA_CHANNEL_POSITION_AUX23 +#define MA_PA_CHANNEL_POSITION_AUX24 PA_CHANNEL_POSITION_AUX24 +#define MA_PA_CHANNEL_POSITION_AUX25 PA_CHANNEL_POSITION_AUX25 +#define MA_PA_CHANNEL_POSITION_AUX26 PA_CHANNEL_POSITION_AUX26 +#define MA_PA_CHANNEL_POSITION_AUX27 PA_CHANNEL_POSITION_AUX27 +#define MA_PA_CHANNEL_POSITION_AUX28 PA_CHANNEL_POSITION_AUX28 +#define MA_PA_CHANNEL_POSITION_AUX29 PA_CHANNEL_POSITION_AUX29 +#define MA_PA_CHANNEL_POSITION_AUX30 PA_CHANNEL_POSITION_AUX30 +#define MA_PA_CHANNEL_POSITION_AUX31 PA_CHANNEL_POSITION_AUX31 +#define MA_PA_CHANNEL_POSITION_TOP_CENTER PA_CHANNEL_POSITION_TOP_CENTER +#define MA_PA_CHANNEL_POSITION_TOP_FRONT_LEFT PA_CHANNEL_POSITION_TOP_FRONT_LEFT +#define MA_PA_CHANNEL_POSITION_TOP_FRONT_RIGHT PA_CHANNEL_POSITION_TOP_FRONT_RIGHT +#define MA_PA_CHANNEL_POSITION_TOP_FRONT_CENTER PA_CHANNEL_POSITION_TOP_FRONT_CENTER +#define MA_PA_CHANNEL_POSITION_TOP_REAR_LEFT PA_CHANNEL_POSITION_TOP_REAR_LEFT +#define MA_PA_CHANNEL_POSITION_TOP_REAR_RIGHT PA_CHANNEL_POSITION_TOP_REAR_RIGHT +#define MA_PA_CHANNEL_POSITION_TOP_REAR_CENTER PA_CHANNEL_POSITION_TOP_REAR_CENTER +#define MA_PA_CHANNEL_POSITION_LEFT PA_CHANNEL_POSITION_LEFT +#define MA_PA_CHANNEL_POSITION_RIGHT PA_CHANNEL_POSITION_RIGHT +#define MA_PA_CHANNEL_POSITION_CENTER PA_CHANNEL_POSITION_CENTER +#define MA_PA_CHANNEL_POSITION_SUBWOOFER PA_CHANNEL_POSITION_SUBWOOFER + +typedef pa_channel_map_def_t ma_pa_channel_map_def_t; +#define MA_PA_CHANNEL_MAP_AIFF PA_CHANNEL_MAP_AIFF +#define MA_PA_CHANNEL_MAP_ALSA PA_CHANNEL_MAP_ALSA +#define MA_PA_CHANNEL_MAP_AUX PA_CHANNEL_MAP_AUX +#define MA_PA_CHANNEL_MAP_WAVEEX PA_CHANNEL_MAP_WAVEEX +#define MA_PA_CHANNEL_MAP_OSS PA_CHANNEL_MAP_OSS +#define MA_PA_CHANNEL_MAP_DEFAULT PA_CHANNEL_MAP_DEFAULT + +typedef pa_sample_format_t ma_pa_sample_format_t; +#define MA_PA_SAMPLE_INVALID PA_SAMPLE_INVALID +#define MA_PA_SAMPLE_U8 PA_SAMPLE_U8 +#define MA_PA_SAMPLE_ALAW PA_SAMPLE_ALAW +#define MA_PA_SAMPLE_ULAW PA_SAMPLE_ULAW +#define MA_PA_SAMPLE_S16LE PA_SAMPLE_S16LE +#define MA_PA_SAMPLE_S16BE PA_SAMPLE_S16BE +#define MA_PA_SAMPLE_FLOAT32LE PA_SAMPLE_FLOAT32LE +#define MA_PA_SAMPLE_FLOAT32BE PA_SAMPLE_FLOAT32BE +#define MA_PA_SAMPLE_S32LE PA_SAMPLE_S32LE +#define MA_PA_SAMPLE_S32BE PA_SAMPLE_S32BE +#define MA_PA_SAMPLE_S24LE PA_SAMPLE_S24LE +#define MA_PA_SAMPLE_S24BE PA_SAMPLE_S24BE +#define MA_PA_SAMPLE_S24_32LE PA_SAMPLE_S24_32LE +#define MA_PA_SAMPLE_S24_32BE PA_SAMPLE_S24_32BE + +typedef pa_mainloop ma_pa_mainloop; +typedef pa_threaded_mainloop ma_pa_threaded_mainloop; +typedef pa_mainloop_api ma_pa_mainloop_api; +typedef pa_context ma_pa_context; +typedef pa_operation ma_pa_operation; +typedef pa_stream ma_pa_stream; +typedef pa_spawn_api ma_pa_spawn_api; +typedef pa_buffer_attr ma_pa_buffer_attr; +typedef pa_channel_map ma_pa_channel_map; +typedef pa_cvolume ma_pa_cvolume; +typedef pa_sample_spec ma_pa_sample_spec; +typedef pa_sink_info ma_pa_sink_info; +typedef pa_source_info ma_pa_source_info; + +typedef pa_context_notify_cb_t ma_pa_context_notify_cb_t; +typedef pa_sink_info_cb_t ma_pa_sink_info_cb_t; +typedef pa_source_info_cb_t ma_pa_source_info_cb_t; +typedef pa_stream_success_cb_t ma_pa_stream_success_cb_t; +typedef pa_stream_request_cb_t ma_pa_stream_request_cb_t; +typedef pa_stream_notify_cb_t ma_pa_stream_notify_cb_t; +typedef pa_free_cb_t ma_pa_free_cb_t; +#else +#define MA_PA_OK 0 +#define MA_PA_ERR_ACCESS 1 +#define MA_PA_ERR_INVALID 2 +#define MA_PA_ERR_NOENTITY 5 +#define MA_PA_ERR_NOTSUPPORTED 19 + +#define MA_PA_CHANNELS_MAX 32 +#define MA_PA_RATE_MAX 384000 + +typedef int ma_pa_context_flags_t; +#define MA_PA_CONTEXT_NOFLAGS 0x00000000 +#define MA_PA_CONTEXT_NOAUTOSPAWN 0x00000001 +#define MA_PA_CONTEXT_NOFAIL 0x00000002 + +typedef int ma_pa_stream_flags_t; +#define MA_PA_STREAM_NOFLAGS 0x00000000 +#define MA_PA_STREAM_START_CORKED 0x00000001 +#define MA_PA_STREAM_INTERPOLATE_TIMING 0x00000002 +#define MA_PA_STREAM_NOT_MONOTONIC 0x00000004 +#define MA_PA_STREAM_AUTO_TIMING_UPDATE 0x00000008 +#define MA_PA_STREAM_NO_REMAP_CHANNELS 0x00000010 +#define MA_PA_STREAM_NO_REMIX_CHANNELS 0x00000020 +#define MA_PA_STREAM_FIX_FORMAT 0x00000040 +#define MA_PA_STREAM_FIX_RATE 0x00000080 +#define MA_PA_STREAM_FIX_CHANNELS 0x00000100 +#define MA_PA_STREAM_DONT_MOVE 0x00000200 +#define MA_PA_STREAM_VARIABLE_RATE 0x00000400 +#define MA_PA_STREAM_PEAK_DETECT 0x00000800 +#define MA_PA_STREAM_START_MUTED 0x00001000 +#define MA_PA_STREAM_ADJUST_LATENCY 0x00002000 +#define MA_PA_STREAM_EARLY_REQUESTS 0x00004000 +#define MA_PA_STREAM_DONT_INHIBIT_AUTO_SUSPEND 0x00008000 +#define MA_PA_STREAM_START_UNMUTED 0x00010000 +#define MA_PA_STREAM_FAIL_ON_SUSPEND 0x00020000 +#define MA_PA_STREAM_RELATIVE_VOLUME 0x00040000 +#define MA_PA_STREAM_PASSTHROUGH 0x00080000 + +typedef int ma_pa_sink_flags_t; +#define MA_PA_SINK_NOFLAGS 0x00000000 +#define MA_PA_SINK_HW_VOLUME_CTRL 0x00000001 +#define MA_PA_SINK_LATENCY 0x00000002 +#define MA_PA_SINK_HARDWARE 0x00000004 +#define MA_PA_SINK_NETWORK 0x00000008 +#define MA_PA_SINK_HW_MUTE_CTRL 0x00000010 +#define MA_PA_SINK_DECIBEL_VOLUME 0x00000020 +#define MA_PA_SINK_FLAT_VOLUME 0x00000040 +#define MA_PA_SINK_DYNAMIC_LATENCY 0x00000080 +#define MA_PA_SINK_SET_FORMATS 0x00000100 + +typedef int ma_pa_source_flags_t; +#define MA_PA_SOURCE_NOFLAGS 0x00000000 +#define MA_PA_SOURCE_HW_VOLUME_CTRL 0x00000001 +#define MA_PA_SOURCE_LATENCY 0x00000002 +#define MA_PA_SOURCE_HARDWARE 0x00000004 +#define MA_PA_SOURCE_NETWORK 0x00000008 +#define MA_PA_SOURCE_HW_MUTE_CTRL 0x00000010 +#define MA_PA_SOURCE_DECIBEL_VOLUME 0x00000020 +#define MA_PA_SOURCE_DYNAMIC_LATENCY 0x00000040 +#define MA_PA_SOURCE_FLAT_VOLUME 0x00000080 + +typedef int ma_pa_context_state_t; +#define MA_PA_CONTEXT_UNCONNECTED 0 +#define MA_PA_CONTEXT_CONNECTING 1 +#define MA_PA_CONTEXT_AUTHORIZING 2 +#define MA_PA_CONTEXT_SETTING_NAME 3 +#define MA_PA_CONTEXT_READY 4 +#define MA_PA_CONTEXT_FAILED 5 +#define MA_PA_CONTEXT_TERMINATED 6 + +typedef int ma_pa_stream_state_t; +#define MA_PA_STREAM_UNCONNECTED 0 +#define MA_PA_STREAM_CREATING 1 +#define MA_PA_STREAM_READY 2 +#define MA_PA_STREAM_FAILED 3 +#define MA_PA_STREAM_TERMINATED 4 + +typedef int ma_pa_operation_state_t; +#define MA_PA_OPERATION_RUNNING 0 +#define MA_PA_OPERATION_DONE 1 +#define MA_PA_OPERATION_CANCELLED 2 + +typedef int ma_pa_sink_state_t; +#define MA_PA_SINK_INVALID_STATE -1 +#define MA_PA_SINK_RUNNING 0 +#define MA_PA_SINK_IDLE 1 +#define MA_PA_SINK_SUSPENDED 2 + +typedef int ma_pa_source_state_t; +#define MA_PA_SOURCE_INVALID_STATE -1 +#define MA_PA_SOURCE_RUNNING 0 +#define MA_PA_SOURCE_IDLE 1 +#define MA_PA_SOURCE_SUSPENDED 2 + +typedef int ma_pa_seek_mode_t; +#define MA_PA_SEEK_RELATIVE 0 +#define MA_PA_SEEK_ABSOLUTE 1 +#define MA_PA_SEEK_RELATIVE_ON_READ 2 +#define MA_PA_SEEK_RELATIVE_END 3 + +typedef int ma_pa_channel_position_t; +#define MA_PA_CHANNEL_POSITION_INVALID -1 +#define MA_PA_CHANNEL_POSITION_MONO 0 +#define MA_PA_CHANNEL_POSITION_FRONT_LEFT 1 +#define MA_PA_CHANNEL_POSITION_FRONT_RIGHT 2 +#define MA_PA_CHANNEL_POSITION_FRONT_CENTER 3 +#define MA_PA_CHANNEL_POSITION_REAR_CENTER 4 +#define MA_PA_CHANNEL_POSITION_REAR_LEFT 5 +#define MA_PA_CHANNEL_POSITION_REAR_RIGHT 6 +#define MA_PA_CHANNEL_POSITION_LFE 7 +#define MA_PA_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER 8 +#define MA_PA_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER 9 +#define MA_PA_CHANNEL_POSITION_SIDE_LEFT 10 +#define MA_PA_CHANNEL_POSITION_SIDE_RIGHT 11 +#define MA_PA_CHANNEL_POSITION_AUX0 12 +#define MA_PA_CHANNEL_POSITION_AUX1 13 +#define MA_PA_CHANNEL_POSITION_AUX2 14 +#define MA_PA_CHANNEL_POSITION_AUX3 15 +#define MA_PA_CHANNEL_POSITION_AUX4 16 +#define MA_PA_CHANNEL_POSITION_AUX5 17 +#define MA_PA_CHANNEL_POSITION_AUX6 18 +#define MA_PA_CHANNEL_POSITION_AUX7 19 +#define MA_PA_CHANNEL_POSITION_AUX8 20 +#define MA_PA_CHANNEL_POSITION_AUX9 21 +#define MA_PA_CHANNEL_POSITION_AUX10 22 +#define MA_PA_CHANNEL_POSITION_AUX11 23 +#define MA_PA_CHANNEL_POSITION_AUX12 24 +#define MA_PA_CHANNEL_POSITION_AUX13 25 +#define MA_PA_CHANNEL_POSITION_AUX14 26 +#define MA_PA_CHANNEL_POSITION_AUX15 27 +#define MA_PA_CHANNEL_POSITION_AUX16 28 +#define MA_PA_CHANNEL_POSITION_AUX17 29 +#define MA_PA_CHANNEL_POSITION_AUX18 30 +#define MA_PA_CHANNEL_POSITION_AUX19 31 +#define MA_PA_CHANNEL_POSITION_AUX20 32 +#define MA_PA_CHANNEL_POSITION_AUX21 33 +#define MA_PA_CHANNEL_POSITION_AUX22 34 +#define MA_PA_CHANNEL_POSITION_AUX23 35 +#define MA_PA_CHANNEL_POSITION_AUX24 36 +#define MA_PA_CHANNEL_POSITION_AUX25 37 +#define MA_PA_CHANNEL_POSITION_AUX26 38 +#define MA_PA_CHANNEL_POSITION_AUX27 39 +#define MA_PA_CHANNEL_POSITION_AUX28 40 +#define MA_PA_CHANNEL_POSITION_AUX29 41 +#define MA_PA_CHANNEL_POSITION_AUX30 42 +#define MA_PA_CHANNEL_POSITION_AUX31 43 +#define MA_PA_CHANNEL_POSITION_TOP_CENTER 44 +#define MA_PA_CHANNEL_POSITION_TOP_FRONT_LEFT 45 +#define MA_PA_CHANNEL_POSITION_TOP_FRONT_RIGHT 46 +#define MA_PA_CHANNEL_POSITION_TOP_FRONT_CENTER 47 +#define MA_PA_CHANNEL_POSITION_TOP_REAR_LEFT 48 +#define MA_PA_CHANNEL_POSITION_TOP_REAR_RIGHT 49 +#define MA_PA_CHANNEL_POSITION_TOP_REAR_CENTER 50 +#define MA_PA_CHANNEL_POSITION_LEFT MA_PA_CHANNEL_POSITION_FRONT_LEFT +#define MA_PA_CHANNEL_POSITION_RIGHT MA_PA_CHANNEL_POSITION_FRONT_RIGHT +#define MA_PA_CHANNEL_POSITION_CENTER MA_PA_CHANNEL_POSITION_FRONT_CENTER +#define MA_PA_CHANNEL_POSITION_SUBWOOFER MA_PA_CHANNEL_POSITION_LFE + +typedef int ma_pa_channel_map_def_t; +#define MA_PA_CHANNEL_MAP_AIFF 0 +#define MA_PA_CHANNEL_MAP_ALSA 1 +#define MA_PA_CHANNEL_MAP_AUX 2 +#define MA_PA_CHANNEL_MAP_WAVEEX 3 +#define MA_PA_CHANNEL_MAP_OSS 4 +#define MA_PA_CHANNEL_MAP_DEFAULT MA_PA_CHANNEL_MAP_AIFF + +typedef int ma_pa_sample_format_t; +#define MA_PA_SAMPLE_INVALID -1 +#define MA_PA_SAMPLE_U8 0 +#define MA_PA_SAMPLE_ALAW 1 +#define MA_PA_SAMPLE_ULAW 2 +#define MA_PA_SAMPLE_S16LE 3 +#define MA_PA_SAMPLE_S16BE 4 +#define MA_PA_SAMPLE_FLOAT32LE 5 +#define MA_PA_SAMPLE_FLOAT32BE 6 +#define MA_PA_SAMPLE_S32LE 7 +#define MA_PA_SAMPLE_S32BE 8 +#define MA_PA_SAMPLE_S24LE 9 +#define MA_PA_SAMPLE_S24BE 10 +#define MA_PA_SAMPLE_S24_32LE 11 +#define MA_PA_SAMPLE_S24_32BE 12 + +typedef struct ma_pa_mainloop ma_pa_mainloop; +typedef struct ma_pa_threaded_mainloop ma_pa_threaded_mainloop; +typedef struct ma_pa_mainloop_api ma_pa_mainloop_api; +typedef struct ma_pa_context ma_pa_context; +typedef struct ma_pa_operation ma_pa_operation; +typedef struct ma_pa_stream ma_pa_stream; +typedef struct ma_pa_spawn_api ma_pa_spawn_api; + +typedef struct +{ + ma_uint32 maxlength; + ma_uint32 tlength; + ma_uint32 prebuf; + ma_uint32 minreq; + ma_uint32 fragsize; +} ma_pa_buffer_attr; + +typedef struct +{ + ma_uint8 channels; + ma_pa_channel_position_t map[MA_PA_CHANNELS_MAX]; +} ma_pa_channel_map; + +typedef struct +{ + ma_uint8 channels; + ma_uint32 values[MA_PA_CHANNELS_MAX]; +} ma_pa_cvolume; + +typedef struct +{ + ma_pa_sample_format_t format; + ma_uint32 rate; + ma_uint8 channels; +} ma_pa_sample_spec; + +typedef struct +{ + const char* name; + ma_uint32 index; + const char* description; + ma_pa_sample_spec sample_spec; + ma_pa_channel_map channel_map; + ma_uint32 owner_module; + ma_pa_cvolume volume; + int mute; + ma_uint32 monitor_source; + const char* monitor_source_name; + ma_uint64 latency; + const char* driver; + ma_pa_sink_flags_t flags; + void* proplist; + ma_uint64 configured_latency; + ma_uint32 base_volume; + ma_pa_sink_state_t state; + ma_uint32 n_volume_steps; + ma_uint32 card; + ma_uint32 n_ports; + void** ports; + void* active_port; + ma_uint8 n_formats; + void** formats; +} ma_pa_sink_info; + +typedef struct +{ + const char *name; + ma_uint32 index; + const char *description; + ma_pa_sample_spec sample_spec; + ma_pa_channel_map channel_map; + ma_uint32 owner_module; + ma_pa_cvolume volume; + int mute; + ma_uint32 monitor_of_sink; + const char *monitor_of_sink_name; + ma_uint64 latency; + const char *driver; + ma_pa_source_flags_t flags; + void* proplist; + ma_uint64 configured_latency; + ma_uint32 base_volume; + ma_pa_source_state_t state; + ma_uint32 n_volume_steps; + ma_uint32 card; + ma_uint32 n_ports; + void** ports; + void* active_port; + ma_uint8 n_formats; + void** formats; +} ma_pa_source_info; + +typedef void (* ma_pa_context_notify_cb_t)(ma_pa_context* c, void* userdata); +typedef void (* ma_pa_sink_info_cb_t) (ma_pa_context* c, const ma_pa_sink_info* i, int eol, void* userdata); +typedef void (* ma_pa_source_info_cb_t) (ma_pa_context* c, const ma_pa_source_info* i, int eol, void* userdata); +typedef void (* ma_pa_stream_success_cb_t)(ma_pa_stream* s, int success, void* userdata); +typedef void (* ma_pa_stream_request_cb_t)(ma_pa_stream* s, size_t nbytes, void* userdata); +typedef void (* ma_pa_stream_notify_cb_t) (ma_pa_stream* s, void* userdata); +typedef void (* ma_pa_free_cb_t) (void* p); +#endif + + +typedef ma_pa_mainloop* (* ma_pa_mainloop_new_proc) (void); +typedef void (* ma_pa_mainloop_free_proc) (ma_pa_mainloop* m); +typedef void (* ma_pa_mainloop_quit_proc) (ma_pa_mainloop* m, int retval); +typedef ma_pa_mainloop_api* (* ma_pa_mainloop_get_api_proc) (ma_pa_mainloop* m); +typedef int (* ma_pa_mainloop_iterate_proc) (ma_pa_mainloop* m, int block, int* retval); +typedef void (* ma_pa_mainloop_wakeup_proc) (ma_pa_mainloop* m); +typedef ma_pa_threaded_mainloop* (* ma_pa_threaded_mainloop_new_proc) (void); +typedef void (* ma_pa_threaded_mainloop_free_proc) (ma_pa_threaded_mainloop* m); +typedef int (* ma_pa_threaded_mainloop_start_proc) (ma_pa_threaded_mainloop* m); +typedef void (* ma_pa_threaded_mainloop_stop_proc) (ma_pa_threaded_mainloop* m); +typedef void (* ma_pa_threaded_mainloop_lock_proc) (ma_pa_threaded_mainloop* m); +typedef void (* ma_pa_threaded_mainloop_unlock_proc) (ma_pa_threaded_mainloop* m); +typedef void (* ma_pa_threaded_mainloop_wait_proc) (ma_pa_threaded_mainloop* m); +typedef void (* ma_pa_threaded_mainloop_signal_proc) (ma_pa_threaded_mainloop* m, int wait_for_accept); +typedef void (* ma_pa_threaded_mainloop_accept_proc) (ma_pa_threaded_mainloop* m); +typedef int (* ma_pa_threaded_mainloop_get_retval_proc) (ma_pa_threaded_mainloop* m); +typedef ma_pa_mainloop_api* (* ma_pa_threaded_mainloop_get_api_proc) (ma_pa_threaded_mainloop* m); +typedef int (* ma_pa_threaded_mainloop_in_thread_proc) (ma_pa_threaded_mainloop* m); +typedef void (* ma_pa_threaded_mainloop_set_name_proc) (ma_pa_threaded_mainloop* m, const char* name); +typedef ma_pa_context* (* ma_pa_context_new_proc) (ma_pa_mainloop_api* mainloop, const char* name); +typedef void (* ma_pa_context_unref_proc) (ma_pa_context* c); +typedef int (* ma_pa_context_connect_proc) (ma_pa_context* c, const char* server, ma_pa_context_flags_t flags, const ma_pa_spawn_api* api); +typedef void (* ma_pa_context_disconnect_proc) (ma_pa_context* c); +typedef void (* ma_pa_context_set_state_callback_proc) (ma_pa_context* c, ma_pa_context_notify_cb_t cb, void* userdata); +typedef ma_pa_context_state_t (* ma_pa_context_get_state_proc) (ma_pa_context* c); +typedef ma_pa_operation* (* ma_pa_context_get_sink_info_list_proc) (ma_pa_context* c, ma_pa_sink_info_cb_t cb, void* userdata); +typedef ma_pa_operation* (* ma_pa_context_get_source_info_list_proc) (ma_pa_context* c, ma_pa_source_info_cb_t cb, void* userdata); +typedef ma_pa_operation* (* ma_pa_context_get_sink_info_by_name_proc) (ma_pa_context* c, const char* name, ma_pa_sink_info_cb_t cb, void* userdata); +typedef ma_pa_operation* (* ma_pa_context_get_source_info_by_name_proc)(ma_pa_context* c, const char* name, ma_pa_source_info_cb_t cb, void* userdata); +typedef void (* ma_pa_operation_unref_proc) (ma_pa_operation* o); +typedef ma_pa_operation_state_t (* ma_pa_operation_get_state_proc) (ma_pa_operation* o); +typedef ma_pa_channel_map* (* ma_pa_channel_map_init_extend_proc) (ma_pa_channel_map* m, unsigned channels, ma_pa_channel_map_def_t def); +typedef int (* ma_pa_channel_map_valid_proc) (const ma_pa_channel_map* m); +typedef int (* ma_pa_channel_map_compatible_proc) (const ma_pa_channel_map* m, const ma_pa_sample_spec* ss); +typedef ma_pa_stream* (* ma_pa_stream_new_proc) (ma_pa_context* c, const char* name, const ma_pa_sample_spec* ss, const ma_pa_channel_map* map); +typedef void (* ma_pa_stream_unref_proc) (ma_pa_stream* s); +typedef int (* ma_pa_stream_connect_playback_proc) (ma_pa_stream* s, const char* dev, const ma_pa_buffer_attr* attr, ma_pa_stream_flags_t flags, const ma_pa_cvolume* volume, ma_pa_stream* sync_stream); +typedef int (* ma_pa_stream_connect_record_proc) (ma_pa_stream* s, const char* dev, const ma_pa_buffer_attr* attr, ma_pa_stream_flags_t flags); +typedef int (* ma_pa_stream_disconnect_proc) (ma_pa_stream* s); +typedef ma_pa_stream_state_t (* ma_pa_stream_get_state_proc) (ma_pa_stream* s); +typedef const ma_pa_sample_spec* (* ma_pa_stream_get_sample_spec_proc) (ma_pa_stream* s); +typedef const ma_pa_channel_map* (* ma_pa_stream_get_channel_map_proc) (ma_pa_stream* s); +typedef const ma_pa_buffer_attr* (* ma_pa_stream_get_buffer_attr_proc) (ma_pa_stream* s); +typedef ma_pa_operation* (* ma_pa_stream_set_buffer_attr_proc) (ma_pa_stream* s, const ma_pa_buffer_attr* attr, ma_pa_stream_success_cb_t cb, void* userdata); +typedef const char* (* ma_pa_stream_get_device_name_proc) (ma_pa_stream* s); +typedef void (* ma_pa_stream_set_write_callback_proc) (ma_pa_stream* s, ma_pa_stream_request_cb_t cb, void* userdata); +typedef void (* ma_pa_stream_set_read_callback_proc) (ma_pa_stream* s, ma_pa_stream_request_cb_t cb, void* userdata); +typedef void (* ma_pa_stream_set_suspended_callback_proc) (ma_pa_stream* s, ma_pa_stream_notify_cb_t cb, void* userdata); +typedef void (* ma_pa_stream_set_moved_callback_proc) (ma_pa_stream* s, ma_pa_stream_notify_cb_t cb, void* userdata); +typedef int (* ma_pa_stream_is_suspended_proc) (const ma_pa_stream* s); +typedef ma_pa_operation* (* ma_pa_stream_flush_proc) (ma_pa_stream* s, ma_pa_stream_success_cb_t cb, void* userdata); +typedef ma_pa_operation* (* ma_pa_stream_drain_proc) (ma_pa_stream* s, ma_pa_stream_success_cb_t cb, void* userdata); +typedef int (* ma_pa_stream_is_corked_proc) (ma_pa_stream* s); +typedef ma_pa_operation* (* ma_pa_stream_cork_proc) (ma_pa_stream* s, int b, ma_pa_stream_success_cb_t cb, void* userdata); +typedef ma_pa_operation* (* ma_pa_stream_trigger_proc) (ma_pa_stream* s, ma_pa_stream_success_cb_t cb, void* userdata); +typedef int (* ma_pa_stream_begin_write_proc) (ma_pa_stream* s, void** data, size_t* nbytes); +typedef int (* ma_pa_stream_write_proc) (ma_pa_stream* s, const void* data, size_t nbytes, ma_pa_free_cb_t free_cb, int64_t offset, ma_pa_seek_mode_t seek); +typedef int (* ma_pa_stream_peek_proc) (ma_pa_stream* s, const void** data, size_t* nbytes); +typedef int (* ma_pa_stream_drop_proc) (ma_pa_stream* s); +typedef size_t (* ma_pa_stream_writable_size_proc) (ma_pa_stream* s); +typedef size_t (* ma_pa_stream_readable_size_proc) (ma_pa_stream* s); + +typedef struct +{ + ma_uint32 count; + ma_uint32 capacity; + ma_device_info* pInfo; +} ma_pulse_device_enum_data; + +static ma_result ma_result_from_pulse(int result) +{ + if (result < 0) { + return MA_ERROR; + } + + switch (result) { + case MA_PA_OK: return MA_SUCCESS; + case MA_PA_ERR_ACCESS: return MA_ACCESS_DENIED; + case MA_PA_ERR_INVALID: return MA_INVALID_ARGS; + case MA_PA_ERR_NOENTITY: return MA_NO_DEVICE; + default: return MA_ERROR; + } +} + +#if 0 +static ma_pa_sample_format_t ma_format_to_pulse(ma_format format) +{ + if (ma_is_little_endian()) { + switch (format) { + case ma_format_s16: return MA_PA_SAMPLE_S16LE; + case ma_format_s24: return MA_PA_SAMPLE_S24LE; + case ma_format_s32: return MA_PA_SAMPLE_S32LE; + case ma_format_f32: return MA_PA_SAMPLE_FLOAT32LE; + default: break; + } + } else { + switch (format) { + case ma_format_s16: return MA_PA_SAMPLE_S16BE; + case ma_format_s24: return MA_PA_SAMPLE_S24BE; + case ma_format_s32: return MA_PA_SAMPLE_S32BE; + case ma_format_f32: return MA_PA_SAMPLE_FLOAT32BE; + default: break; + } + } + + /* Endian agnostic. */ + switch (format) { + case ma_format_u8: return MA_PA_SAMPLE_U8; + default: return MA_PA_SAMPLE_INVALID; + } +} +#endif + +static ma_format ma_format_from_pulse(ma_pa_sample_format_t format) +{ + if (ma_is_little_endian()) { + switch (format) { + case MA_PA_SAMPLE_S16LE: return ma_format_s16; + case MA_PA_SAMPLE_S24LE: return ma_format_s24; + case MA_PA_SAMPLE_S32LE: return ma_format_s32; + case MA_PA_SAMPLE_FLOAT32LE: return ma_format_f32; + default: break; + } + } else { + switch (format) { + case MA_PA_SAMPLE_S16BE: return ma_format_s16; + case MA_PA_SAMPLE_S24BE: return ma_format_s24; + case MA_PA_SAMPLE_S32BE: return ma_format_s32; + case MA_PA_SAMPLE_FLOAT32BE: return ma_format_f32; + default: break; + } + } + + /* Endian agnostic. */ + switch (format) { + case MA_PA_SAMPLE_U8: return ma_format_u8; + default: return ma_format_unknown; + } +} + +static ma_channel ma_channel_position_from_pulse(ma_pa_channel_position_t position) +{ + switch (position) + { + case MA_PA_CHANNEL_POSITION_INVALID: return MA_CHANNEL_NONE; + case MA_PA_CHANNEL_POSITION_MONO: return MA_CHANNEL_MONO; + case MA_PA_CHANNEL_POSITION_FRONT_LEFT: return MA_CHANNEL_FRONT_LEFT; + case MA_PA_CHANNEL_POSITION_FRONT_RIGHT: return MA_CHANNEL_FRONT_RIGHT; + case MA_PA_CHANNEL_POSITION_FRONT_CENTER: return MA_CHANNEL_FRONT_CENTER; + case MA_PA_CHANNEL_POSITION_REAR_CENTER: return MA_CHANNEL_BACK_CENTER; + case MA_PA_CHANNEL_POSITION_REAR_LEFT: return MA_CHANNEL_BACK_LEFT; + case MA_PA_CHANNEL_POSITION_REAR_RIGHT: return MA_CHANNEL_BACK_RIGHT; + case MA_PA_CHANNEL_POSITION_LFE: return MA_CHANNEL_LFE; + case MA_PA_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER: return MA_CHANNEL_FRONT_LEFT_CENTER; + case MA_PA_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER: return MA_CHANNEL_FRONT_RIGHT_CENTER; + case MA_PA_CHANNEL_POSITION_SIDE_LEFT: return MA_CHANNEL_SIDE_LEFT; + case MA_PA_CHANNEL_POSITION_SIDE_RIGHT: return MA_CHANNEL_SIDE_RIGHT; + case MA_PA_CHANNEL_POSITION_AUX0: return MA_CHANNEL_AUX_0; + case MA_PA_CHANNEL_POSITION_AUX1: return MA_CHANNEL_AUX_1; + case MA_PA_CHANNEL_POSITION_AUX2: return MA_CHANNEL_AUX_2; + case MA_PA_CHANNEL_POSITION_AUX3: return MA_CHANNEL_AUX_3; + case MA_PA_CHANNEL_POSITION_AUX4: return MA_CHANNEL_AUX_4; + case MA_PA_CHANNEL_POSITION_AUX5: return MA_CHANNEL_AUX_5; + case MA_PA_CHANNEL_POSITION_AUX6: return MA_CHANNEL_AUX_6; + case MA_PA_CHANNEL_POSITION_AUX7: return MA_CHANNEL_AUX_7; + case MA_PA_CHANNEL_POSITION_AUX8: return MA_CHANNEL_AUX_8; + case MA_PA_CHANNEL_POSITION_AUX9: return MA_CHANNEL_AUX_9; + case MA_PA_CHANNEL_POSITION_AUX10: return MA_CHANNEL_AUX_10; + case MA_PA_CHANNEL_POSITION_AUX11: return MA_CHANNEL_AUX_11; + case MA_PA_CHANNEL_POSITION_AUX12: return MA_CHANNEL_AUX_12; + case MA_PA_CHANNEL_POSITION_AUX13: return MA_CHANNEL_AUX_13; + case MA_PA_CHANNEL_POSITION_AUX14: return MA_CHANNEL_AUX_14; + case MA_PA_CHANNEL_POSITION_AUX15: return MA_CHANNEL_AUX_15; + case MA_PA_CHANNEL_POSITION_AUX16: return MA_CHANNEL_AUX_16; + case MA_PA_CHANNEL_POSITION_AUX17: return MA_CHANNEL_AUX_17; + case MA_PA_CHANNEL_POSITION_AUX18: return MA_CHANNEL_AUX_18; + case MA_PA_CHANNEL_POSITION_AUX19: return MA_CHANNEL_AUX_19; + case MA_PA_CHANNEL_POSITION_AUX20: return MA_CHANNEL_AUX_20; + case MA_PA_CHANNEL_POSITION_AUX21: return MA_CHANNEL_AUX_21; + case MA_PA_CHANNEL_POSITION_AUX22: return MA_CHANNEL_AUX_22; + case MA_PA_CHANNEL_POSITION_AUX23: return MA_CHANNEL_AUX_23; + case MA_PA_CHANNEL_POSITION_AUX24: return MA_CHANNEL_AUX_24; + case MA_PA_CHANNEL_POSITION_AUX25: return MA_CHANNEL_AUX_25; + case MA_PA_CHANNEL_POSITION_AUX26: return MA_CHANNEL_AUX_26; + case MA_PA_CHANNEL_POSITION_AUX27: return MA_CHANNEL_AUX_27; + case MA_PA_CHANNEL_POSITION_AUX28: return MA_CHANNEL_AUX_28; + case MA_PA_CHANNEL_POSITION_AUX29: return MA_CHANNEL_AUX_29; + case MA_PA_CHANNEL_POSITION_AUX30: return MA_CHANNEL_AUX_30; + case MA_PA_CHANNEL_POSITION_AUX31: return MA_CHANNEL_AUX_31; + case MA_PA_CHANNEL_POSITION_TOP_CENTER: return MA_CHANNEL_TOP_CENTER; + case MA_PA_CHANNEL_POSITION_TOP_FRONT_LEFT: return MA_CHANNEL_TOP_FRONT_LEFT; + case MA_PA_CHANNEL_POSITION_TOP_FRONT_RIGHT: return MA_CHANNEL_TOP_FRONT_RIGHT; + case MA_PA_CHANNEL_POSITION_TOP_FRONT_CENTER: return MA_CHANNEL_TOP_FRONT_CENTER; + case MA_PA_CHANNEL_POSITION_TOP_REAR_LEFT: return MA_CHANNEL_TOP_BACK_LEFT; + case MA_PA_CHANNEL_POSITION_TOP_REAR_RIGHT: return MA_CHANNEL_TOP_BACK_RIGHT; + case MA_PA_CHANNEL_POSITION_TOP_REAR_CENTER: return MA_CHANNEL_TOP_BACK_CENTER; + default: return MA_CHANNEL_NONE; + } +} + +#if 0 +static ma_pa_channel_position_t ma_channel_position_to_pulse(ma_channel position) +{ + switch (position) + { + case MA_CHANNEL_NONE: return MA_PA_CHANNEL_POSITION_INVALID; + case MA_CHANNEL_FRONT_LEFT: return MA_PA_CHANNEL_POSITION_FRONT_LEFT; + case MA_CHANNEL_FRONT_RIGHT: return MA_PA_CHANNEL_POSITION_FRONT_RIGHT; + case MA_CHANNEL_FRONT_CENTER: return MA_PA_CHANNEL_POSITION_FRONT_CENTER; + case MA_CHANNEL_LFE: return MA_PA_CHANNEL_POSITION_LFE; + case MA_CHANNEL_BACK_LEFT: return MA_PA_CHANNEL_POSITION_REAR_LEFT; + case MA_CHANNEL_BACK_RIGHT: return MA_PA_CHANNEL_POSITION_REAR_RIGHT; + case MA_CHANNEL_FRONT_LEFT_CENTER: return MA_PA_CHANNEL_POSITION_FRONT_LEFT_OF_CENTER; + case MA_CHANNEL_FRONT_RIGHT_CENTER: return MA_PA_CHANNEL_POSITION_FRONT_RIGHT_OF_CENTER; + case MA_CHANNEL_BACK_CENTER: return MA_PA_CHANNEL_POSITION_REAR_CENTER; + case MA_CHANNEL_SIDE_LEFT: return MA_PA_CHANNEL_POSITION_SIDE_LEFT; + case MA_CHANNEL_SIDE_RIGHT: return MA_PA_CHANNEL_POSITION_SIDE_RIGHT; + case MA_CHANNEL_TOP_CENTER: return MA_PA_CHANNEL_POSITION_TOP_CENTER; + case MA_CHANNEL_TOP_FRONT_LEFT: return MA_PA_CHANNEL_POSITION_TOP_FRONT_LEFT; + case MA_CHANNEL_TOP_FRONT_CENTER: return MA_PA_CHANNEL_POSITION_TOP_FRONT_CENTER; + case MA_CHANNEL_TOP_FRONT_RIGHT: return MA_PA_CHANNEL_POSITION_TOP_FRONT_RIGHT; + case MA_CHANNEL_TOP_BACK_LEFT: return MA_PA_CHANNEL_POSITION_TOP_REAR_LEFT; + case MA_CHANNEL_TOP_BACK_CENTER: return MA_PA_CHANNEL_POSITION_TOP_REAR_CENTER; + case MA_CHANNEL_TOP_BACK_RIGHT: return MA_PA_CHANNEL_POSITION_TOP_REAR_RIGHT; + case MA_CHANNEL_19: return MA_PA_CHANNEL_POSITION_AUX18; + case MA_CHANNEL_20: return MA_PA_CHANNEL_POSITION_AUX19; + case MA_CHANNEL_21: return MA_PA_CHANNEL_POSITION_AUX20; + case MA_CHANNEL_22: return MA_PA_CHANNEL_POSITION_AUX21; + case MA_CHANNEL_23: return MA_PA_CHANNEL_POSITION_AUX22; + case MA_CHANNEL_24: return MA_PA_CHANNEL_POSITION_AUX23; + case MA_CHANNEL_25: return MA_PA_CHANNEL_POSITION_AUX24; + case MA_CHANNEL_26: return MA_PA_CHANNEL_POSITION_AUX25; + case MA_CHANNEL_27: return MA_PA_CHANNEL_POSITION_AUX26; + case MA_CHANNEL_28: return MA_PA_CHANNEL_POSITION_AUX27; + case MA_CHANNEL_29: return MA_PA_CHANNEL_POSITION_AUX28; + case MA_CHANNEL_30: return MA_PA_CHANNEL_POSITION_AUX29; + case MA_CHANNEL_31: return MA_PA_CHANNEL_POSITION_AUX30; + case MA_CHANNEL_32: return MA_PA_CHANNEL_POSITION_AUX31; + default: return (ma_pa_channel_position_t)position; + } +} +#endif + +static ma_result ma_wait_for_operation__pulse(ma_context* pContext, ma_ptr pMainLoop, ma_pa_operation* pOP) +{ + int resultPA; + ma_pa_operation_state_t state; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pOP != NULL); + + for (;;) { + state = ((ma_pa_operation_get_state_proc)pContext->pulse.pa_operation_get_state)(pOP); + if (state != MA_PA_OPERATION_RUNNING) { + break; /* Done. */ + } + + resultPA = ((ma_pa_mainloop_iterate_proc)pContext->pulse.pa_mainloop_iterate)((ma_pa_mainloop*)pMainLoop, 1, NULL); + if (resultPA < 0) { + return ma_result_from_pulse(resultPA); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_wait_for_operation_and_unref__pulse(ma_context* pContext, ma_ptr pMainLoop, ma_pa_operation* pOP) +{ + ma_result result; + + if (pOP == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_wait_for_operation__pulse(pContext, pMainLoop, pOP); + ((ma_pa_operation_unref_proc)pContext->pulse.pa_operation_unref)(pOP); + + return result; +} + +static ma_result ma_wait_for_pa_context_to_connect__pulse(ma_context* pContext, ma_ptr pMainLoop, ma_ptr pPulseContext) +{ + int resultPA; + ma_pa_context_state_t state; + + for (;;) { + state = ((ma_pa_context_get_state_proc)pContext->pulse.pa_context_get_state)((ma_pa_context*)pPulseContext); + if (state == MA_PA_CONTEXT_READY) { + break; /* Done. */ + } + + if (state == MA_PA_CONTEXT_FAILED || state == MA_PA_CONTEXT_TERMINATED) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[PulseAudio] An error occurred while connecting the PulseAudio context."); + return MA_ERROR; + } + + resultPA = ((ma_pa_mainloop_iterate_proc)pContext->pulse.pa_mainloop_iterate)((ma_pa_mainloop*)pMainLoop, 1, NULL); + if (resultPA < 0) { + return ma_result_from_pulse(resultPA); + } + } + + /* Should never get here. */ + return MA_SUCCESS; +} + +static ma_result ma_wait_for_pa_stream_to_connect__pulse(ma_context* pContext, ma_ptr pMainLoop, ma_ptr pStream) +{ + int resultPA; + ma_pa_stream_state_t state; + + for (;;) { + state = ((ma_pa_stream_get_state_proc)pContext->pulse.pa_stream_get_state)((ma_pa_stream*)pStream); + if (state == MA_PA_STREAM_READY) { + break; /* Done. */ + } + + if (state == MA_PA_STREAM_FAILED || state == MA_PA_STREAM_TERMINATED) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[PulseAudio] An error occurred while connecting the PulseAudio stream."); + return MA_ERROR; + } + + resultPA = ((ma_pa_mainloop_iterate_proc)pContext->pulse.pa_mainloop_iterate)((ma_pa_mainloop*)pMainLoop, 1, NULL); + if (resultPA < 0) { + return ma_result_from_pulse(resultPA); + } + } + + return MA_SUCCESS; +} + + +static ma_result ma_init_pa_mainloop_and_pa_context__pulse(ma_context* pContext, const char* pApplicationName, const char* pServerName, ma_bool32 tryAutoSpawn, ma_ptr* ppMainLoop, ma_ptr* ppPulseContext) +{ + ma_result result; + ma_ptr pMainLoop; + ma_ptr pPulseContext; + + MA_ASSERT(ppMainLoop != NULL); + MA_ASSERT(ppPulseContext != NULL); + + /* The PulseAudio context maps well to miniaudio's notion of a context. The pa_context object will be initialized as part of the ma_context. */ + pMainLoop = ((ma_pa_mainloop_new_proc)pContext->pulse.pa_mainloop_new)(); + if (pMainLoop == NULL) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to create mainloop."); + return MA_FAILED_TO_INIT_BACKEND; + } + + pPulseContext = ((ma_pa_context_new_proc)pContext->pulse.pa_context_new)(((ma_pa_mainloop_get_api_proc)pContext->pulse.pa_mainloop_get_api)((ma_pa_mainloop*)pMainLoop), pApplicationName); + if (pPulseContext == NULL) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to create PulseAudio context."); + ((ma_pa_mainloop_free_proc)pContext->pulse.pa_mainloop_free)((ma_pa_mainloop*)(pMainLoop)); + return MA_FAILED_TO_INIT_BACKEND; + } + + /* Now we need to connect to the context. Everything is asynchronous so we need to wait for it to connect before returning. */ + result = ma_result_from_pulse(((ma_pa_context_connect_proc)pContext->pulse.pa_context_connect)((ma_pa_context*)pPulseContext, pServerName, (tryAutoSpawn) ? 0 : MA_PA_CONTEXT_NOAUTOSPAWN, NULL)); + if (result != MA_SUCCESS) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to connect PulseAudio context."); + ((ma_pa_mainloop_free_proc)pContext->pulse.pa_mainloop_free)((ma_pa_mainloop*)(pMainLoop)); + return result; + } + + /* Since ma_context_init() runs synchronously we need to wait for the PulseAudio context to connect before we return. */ + result = ma_wait_for_pa_context_to_connect__pulse(pContext, pMainLoop, pPulseContext); + if (result != MA_SUCCESS) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[PulseAudio] Waiting for connection failed."); + ((ma_pa_mainloop_free_proc)pContext->pulse.pa_mainloop_free)((ma_pa_mainloop*)(pMainLoop)); + return result; + } + + *ppMainLoop = pMainLoop; + *ppPulseContext = pPulseContext; + + return MA_SUCCESS; +} + + +static void ma_device_sink_info_callback(ma_pa_context* pPulseContext, const ma_pa_sink_info* pInfo, int endOfList, void* pUserData) +{ + ma_pa_sink_info* pInfoOut; + + if (endOfList > 0) { + return; + } + + /* + There has been a report that indicates that pInfo can be null which results + in a null pointer dereference below. We'll check for this for safety. + */ + if (pInfo == NULL) { + return; + } + + pInfoOut = (ma_pa_sink_info*)pUserData; + MA_ASSERT(pInfoOut != NULL); + + *pInfoOut = *pInfo; + + (void)pPulseContext; /* Unused. */ +} + +static void ma_device_source_info_callback(ma_pa_context* pPulseContext, const ma_pa_source_info* pInfo, int endOfList, void* pUserData) +{ + ma_pa_source_info* pInfoOut; + + if (endOfList > 0) { + return; + } + + /* + There has been a report that indicates that pInfo can be null which results + in a null pointer dereference below. We'll check for this for safety. + */ + if (pInfo == NULL) { + return; + } + + pInfoOut = (ma_pa_source_info*)pUserData; + MA_ASSERT(pInfoOut != NULL); + + *pInfoOut = *pInfo; + + (void)pPulseContext; /* Unused. */ +} + +#if 0 +static void ma_device_sink_name_callback(ma_pa_context* pPulseContext, const ma_pa_sink_info* pInfo, int endOfList, void* pUserData) +{ + ma_device* pDevice; + + if (endOfList > 0) { + return; + } + + pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + ma_strncpy_s(pDevice->playback.name, sizeof(pDevice->playback.name), pInfo->description, (size_t)-1); + + (void)pPulseContext; /* Unused. */ +} + +static void ma_device_source_name_callback(ma_pa_context* pPulseContext, const ma_pa_source_info* pInfo, int endOfList, void* pUserData) +{ + ma_device* pDevice; + + if (endOfList > 0) { + return; + } + + pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + ma_strncpy_s(pDevice->capture.name, sizeof(pDevice->capture.name), pInfo->description, (size_t)-1); + + (void)pPulseContext; /* Unused. */ +} +#endif + +static ma_result ma_context_get_sink_info__pulse(ma_context* pContext, const char* pDeviceName, ma_pa_sink_info* pSinkInfo) +{ + ma_pa_operation* pOP; + + pOP = ((ma_pa_context_get_sink_info_by_name_proc)pContext->pulse.pa_context_get_sink_info_by_name)((ma_pa_context*)pContext->pulse.pPulseContext, pDeviceName, ma_device_sink_info_callback, pSinkInfo); + if (pOP == NULL) { + return MA_ERROR; + } + + return ma_wait_for_operation_and_unref__pulse(pContext, pContext->pulse.pMainLoop, pOP); +} + +static ma_result ma_context_get_source_info__pulse(ma_context* pContext, const char* pDeviceName, ma_pa_source_info* pSourceInfo) +{ + ma_pa_operation* pOP; + + pOP = ((ma_pa_context_get_source_info_by_name_proc)pContext->pulse.pa_context_get_source_info_by_name)((ma_pa_context*)pContext->pulse.pPulseContext, pDeviceName, ma_device_source_info_callback, pSourceInfo); + if (pOP == NULL) { + return MA_ERROR; + } + + return ma_wait_for_operation_and_unref__pulse(pContext, pContext->pulse.pMainLoop, pOP); +} + +static ma_result ma_context_get_default_device_index__pulse(ma_context* pContext, ma_device_type deviceType, ma_uint32* pIndex) +{ + ma_result result; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pIndex != NULL); + + if (pIndex != NULL) { + *pIndex = (ma_uint32)-1; + } + + if (deviceType == ma_device_type_playback) { + ma_pa_sink_info sinkInfo; + result = ma_context_get_sink_info__pulse(pContext, NULL, &sinkInfo); + if (result != MA_SUCCESS) { + return result; + } + + if (pIndex != NULL) { + *pIndex = sinkInfo.index; + } + } + + if (deviceType == ma_device_type_capture) { + ma_pa_source_info sourceInfo; + result = ma_context_get_source_info__pulse(pContext, NULL, &sourceInfo); + if (result != MA_SUCCESS) { + return result; + } + + if (pIndex != NULL) { + *pIndex = sourceInfo.index; + } + } + + return MA_SUCCESS; +} + + +typedef struct +{ + ma_context* pContext; + ma_enum_devices_callback_proc callback; + void* pUserData; + ma_bool32 isTerminated; + ma_uint32 defaultDeviceIndexPlayback; + ma_uint32 defaultDeviceIndexCapture; +} ma_context_enumerate_devices_callback_data__pulse; + +static void ma_context_enumerate_devices_sink_callback__pulse(ma_pa_context* pPulseContext, const ma_pa_sink_info* pSinkInfo, int endOfList, void* pUserData) +{ + ma_context_enumerate_devices_callback_data__pulse* pData = (ma_context_enumerate_devices_callback_data__pulse*)pUserData; + ma_device_info deviceInfo; + + MA_ASSERT(pData != NULL); + + if (endOfList || pData->isTerminated) { + return; + } + + MA_ZERO_OBJECT(&deviceInfo); + + /* The name from PulseAudio is the ID for miniaudio. */ + if (pSinkInfo->name != NULL) { + ma_strncpy_s(deviceInfo.id.pulse, sizeof(deviceInfo.id.pulse), pSinkInfo->name, (size_t)-1); + } + + /* The description from PulseAudio is the name for miniaudio. */ + if (pSinkInfo->description != NULL) { + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), pSinkInfo->description, (size_t)-1); + } + + if (pSinkInfo->index == pData->defaultDeviceIndexPlayback) { + deviceInfo.isDefault = MA_TRUE; + } + + pData->isTerminated = !pData->callback(pData->pContext, ma_device_type_playback, &deviceInfo, pData->pUserData); + + (void)pPulseContext; /* Unused. */ +} + +static void ma_context_enumerate_devices_source_callback__pulse(ma_pa_context* pPulseContext, const ma_pa_source_info* pSourceInfo, int endOfList, void* pUserData) +{ + ma_context_enumerate_devices_callback_data__pulse* pData = (ma_context_enumerate_devices_callback_data__pulse*)pUserData; + ma_device_info deviceInfo; + + MA_ASSERT(pData != NULL); + + if (endOfList || pData->isTerminated) { + return; + } + + MA_ZERO_OBJECT(&deviceInfo); + + /* The name from PulseAudio is the ID for miniaudio. */ + if (pSourceInfo->name != NULL) { + ma_strncpy_s(deviceInfo.id.pulse, sizeof(deviceInfo.id.pulse), pSourceInfo->name, (size_t)-1); + } + + /* The description from PulseAudio is the name for miniaudio. */ + if (pSourceInfo->description != NULL) { + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), pSourceInfo->description, (size_t)-1); + } + + if (pSourceInfo->index == pData->defaultDeviceIndexCapture) { + deviceInfo.isDefault = MA_TRUE; + } + + pData->isTerminated = !pData->callback(pData->pContext, ma_device_type_capture, &deviceInfo, pData->pUserData); + + (void)pPulseContext; /* Unused. */ +} + +static ma_result ma_context_enumerate_devices__pulse(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_result result = MA_SUCCESS; + ma_context_enumerate_devices_callback_data__pulse callbackData; + ma_pa_operation* pOP = NULL; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + callbackData.pContext = pContext; + callbackData.callback = callback; + callbackData.pUserData = pUserData; + callbackData.isTerminated = MA_FALSE; + callbackData.defaultDeviceIndexPlayback = (ma_uint32)-1; + callbackData.defaultDeviceIndexCapture = (ma_uint32)-1; + + /* We need to get the index of the default devices. */ + ma_context_get_default_device_index__pulse(pContext, ma_device_type_playback, &callbackData.defaultDeviceIndexPlayback); + ma_context_get_default_device_index__pulse(pContext, ma_device_type_capture, &callbackData.defaultDeviceIndexCapture); + + /* Playback. */ + if (!callbackData.isTerminated) { + pOP = ((ma_pa_context_get_sink_info_list_proc)pContext->pulse.pa_context_get_sink_info_list)((ma_pa_context*)(pContext->pulse.pPulseContext), ma_context_enumerate_devices_sink_callback__pulse, &callbackData); + if (pOP == NULL) { + result = MA_ERROR; + goto done; + } + + result = ma_wait_for_operation__pulse(pContext, pContext->pulse.pMainLoop, pOP); + ((ma_pa_operation_unref_proc)pContext->pulse.pa_operation_unref)(pOP); + + if (result != MA_SUCCESS) { + goto done; + } + } + + + /* Capture. */ + if (!callbackData.isTerminated) { + pOP = ((ma_pa_context_get_source_info_list_proc)pContext->pulse.pa_context_get_source_info_list)((ma_pa_context*)(pContext->pulse.pPulseContext), ma_context_enumerate_devices_source_callback__pulse, &callbackData); + if (pOP == NULL) { + result = MA_ERROR; + goto done; + } + + result = ma_wait_for_operation__pulse(pContext, pContext->pulse.pMainLoop, pOP); + ((ma_pa_operation_unref_proc)pContext->pulse.pa_operation_unref)(pOP); + + if (result != MA_SUCCESS) { + goto done; + } + } + +done: + return result; +} + + +typedef struct +{ + ma_device_info* pDeviceInfo; + ma_uint32 defaultDeviceIndex; + ma_bool32 foundDevice; +} ma_context_get_device_info_callback_data__pulse; + +static void ma_context_get_device_info_sink_callback__pulse(ma_pa_context* pPulseContext, const ma_pa_sink_info* pInfo, int endOfList, void* pUserData) +{ + ma_context_get_device_info_callback_data__pulse* pData = (ma_context_get_device_info_callback_data__pulse*)pUserData; + + if (endOfList > 0) { + return; + } + + MA_ASSERT(pData != NULL); + pData->foundDevice = MA_TRUE; + + if (pInfo->name != NULL) { + ma_strncpy_s(pData->pDeviceInfo->id.pulse, sizeof(pData->pDeviceInfo->id.pulse), pInfo->name, (size_t)-1); + } + + if (pInfo->description != NULL) { + ma_strncpy_s(pData->pDeviceInfo->name, sizeof(pData->pDeviceInfo->name), pInfo->description, (size_t)-1); + } + + /* + We're just reporting a single data format here. I think technically PulseAudio might support + all formats, but I don't trust that PulseAudio will do *anything* right, so I'm just going to + report the "native" device format. + */ + pData->pDeviceInfo->nativeDataFormats[0].format = ma_format_from_pulse(pInfo->sample_spec.format); + pData->pDeviceInfo->nativeDataFormats[0].channels = pInfo->sample_spec.channels; + pData->pDeviceInfo->nativeDataFormats[0].sampleRate = pInfo->sample_spec.rate; + pData->pDeviceInfo->nativeDataFormats[0].flags = 0; + pData->pDeviceInfo->nativeDataFormatCount = 1; + + if (pData->defaultDeviceIndex == pInfo->index) { + pData->pDeviceInfo->isDefault = MA_TRUE; + } + + (void)pPulseContext; /* Unused. */ +} + +static void ma_context_get_device_info_source_callback__pulse(ma_pa_context* pPulseContext, const ma_pa_source_info* pInfo, int endOfList, void* pUserData) +{ + ma_context_get_device_info_callback_data__pulse* pData = (ma_context_get_device_info_callback_data__pulse*)pUserData; + + if (endOfList > 0) { + return; + } + + MA_ASSERT(pData != NULL); + pData->foundDevice = MA_TRUE; + + if (pInfo->name != NULL) { + ma_strncpy_s(pData->pDeviceInfo->id.pulse, sizeof(pData->pDeviceInfo->id.pulse), pInfo->name, (size_t)-1); + } + + if (pInfo->description != NULL) { + ma_strncpy_s(pData->pDeviceInfo->name, sizeof(pData->pDeviceInfo->name), pInfo->description, (size_t)-1); + } + + /* + We're just reporting a single data format here. I think technically PulseAudio might support + all formats, but I don't trust that PulseAudio will do *anything* right, so I'm just going to + report the "native" device format. + */ + pData->pDeviceInfo->nativeDataFormats[0].format = ma_format_from_pulse(pInfo->sample_spec.format); + pData->pDeviceInfo->nativeDataFormats[0].channels = pInfo->sample_spec.channels; + pData->pDeviceInfo->nativeDataFormats[0].sampleRate = pInfo->sample_spec.rate; + pData->pDeviceInfo->nativeDataFormats[0].flags = 0; + pData->pDeviceInfo->nativeDataFormatCount = 1; + + if (pData->defaultDeviceIndex == pInfo->index) { + pData->pDeviceInfo->isDefault = MA_TRUE; + } + + (void)pPulseContext; /* Unused. */ +} + +static ma_result ma_context_get_device_info__pulse(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + ma_result result = MA_SUCCESS; + ma_context_get_device_info_callback_data__pulse callbackData; + ma_pa_operation* pOP = NULL; + const char* pDeviceName = NULL; + + MA_ASSERT(pContext != NULL); + + callbackData.pDeviceInfo = pDeviceInfo; + callbackData.foundDevice = MA_FALSE; + + if (pDeviceID != NULL) { + pDeviceName = pDeviceID->pulse; + } else { + pDeviceName = NULL; + } + + result = ma_context_get_default_device_index__pulse(pContext, deviceType, &callbackData.defaultDeviceIndex); + + if (deviceType == ma_device_type_playback) { + pOP = ((ma_pa_context_get_sink_info_by_name_proc)pContext->pulse.pa_context_get_sink_info_by_name)((ma_pa_context*)(pContext->pulse.pPulseContext), pDeviceName, ma_context_get_device_info_sink_callback__pulse, &callbackData); + } else { + pOP = ((ma_pa_context_get_source_info_by_name_proc)pContext->pulse.pa_context_get_source_info_by_name)((ma_pa_context*)(pContext->pulse.pPulseContext), pDeviceName, ma_context_get_device_info_source_callback__pulse, &callbackData); + } + + if (pOP != NULL) { + ma_wait_for_operation_and_unref__pulse(pContext, pContext->pulse.pMainLoop, pOP); + } else { + result = MA_ERROR; + goto done; + } + + if (!callbackData.foundDevice) { + result = MA_NO_DEVICE; + goto done; + } + +done: + return result; +} + +static ma_result ma_device_uninit__pulse(ma_device* pDevice) +{ + ma_context* pContext; + + MA_ASSERT(pDevice != NULL); + + pContext = pDevice->pContext; + MA_ASSERT(pContext != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ((ma_pa_stream_disconnect_proc)pContext->pulse.pa_stream_disconnect)((ma_pa_stream*)pDevice->pulse.pStreamCapture); + ((ma_pa_stream_unref_proc)pContext->pulse.pa_stream_unref)((ma_pa_stream*)pDevice->pulse.pStreamCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ((ma_pa_stream_disconnect_proc)pContext->pulse.pa_stream_disconnect)((ma_pa_stream*)pDevice->pulse.pStreamPlayback); + ((ma_pa_stream_unref_proc)pContext->pulse.pa_stream_unref)((ma_pa_stream*)pDevice->pulse.pStreamPlayback); + } + + if (pDevice->type == ma_device_type_duplex) { + ma_duplex_rb_uninit(&pDevice->duplexRB); + } + + ((ma_pa_context_disconnect_proc)pContext->pulse.pa_context_disconnect)((ma_pa_context*)pDevice->pulse.pPulseContext); + ((ma_pa_context_unref_proc)pContext->pulse.pa_context_unref)((ma_pa_context*)pDevice->pulse.pPulseContext); + ((ma_pa_mainloop_free_proc)pContext->pulse.pa_mainloop_free)((ma_pa_mainloop*)pDevice->pulse.pMainLoop); + + return MA_SUCCESS; +} + +static ma_pa_buffer_attr ma_device__pa_buffer_attr_new(ma_uint32 periodSizeInFrames, ma_uint32 periods, const ma_pa_sample_spec* ss) +{ + ma_pa_buffer_attr attr; + attr.maxlength = periodSizeInFrames * periods * ma_get_bytes_per_frame(ma_format_from_pulse(ss->format), ss->channels); + attr.tlength = attr.maxlength / periods; + attr.prebuf = (ma_uint32)-1; + attr.minreq = (ma_uint32)-1; + attr.fragsize = attr.maxlength / periods; + + return attr; +} + +static ma_pa_stream* ma_device__pa_stream_new__pulse(ma_device* pDevice, const char* pStreamName, const ma_pa_sample_spec* ss, const ma_pa_channel_map* cmap) +{ + static ma_atomic_uint32 g_StreamCounter = { 0 }; + char actualStreamName[256]; + + if (pStreamName != NULL) { + ma_strncpy_s(actualStreamName, sizeof(actualStreamName), pStreamName, (size_t)-1); + } else { + const char* pBaseName = "miniaudio:"; + size_t baseNameLen = strlen(pBaseName); + ma_strcpy_s(actualStreamName, sizeof(actualStreamName), pBaseName); + ma_itoa_s((int)ma_atomic_uint32_get(&g_StreamCounter), actualStreamName + baseNameLen, sizeof(actualStreamName)-baseNameLen, 10); + } + ma_atomic_uint32_fetch_add(&g_StreamCounter, 1); + + return ((ma_pa_stream_new_proc)pDevice->pContext->pulse.pa_stream_new)((ma_pa_context*)pDevice->pulse.pPulseContext, actualStreamName, ss, cmap); +} + + +static void ma_device_on_read__pulse(ma_pa_stream* pStream, size_t byteCount, void* pUserData) +{ + ma_device* pDevice = (ma_device*)pUserData; + ma_uint32 bpf; + ma_uint32 deviceState; + ma_uint64 frameCount; + ma_uint64 framesProcessed; + + MA_ASSERT(pDevice != NULL); + + /* + Don't do anything if the device isn't initialized yet. Yes, this can happen because PulseAudio + can fire this callback before the stream has even started. Ridiculous. + */ + deviceState = ma_device_get_state(pDevice); + if (deviceState != ma_device_state_starting && deviceState != ma_device_state_started) { + return; + } + + bpf = ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + MA_ASSERT(bpf > 0); + + frameCount = byteCount / bpf; + framesProcessed = 0; + + while (ma_device_get_state(pDevice) == ma_device_state_started && framesProcessed < frameCount) { + const void* pMappedPCMFrames; + size_t bytesMapped; + ma_uint64 framesMapped; + + int pulseResult = ((ma_pa_stream_peek_proc)pDevice->pContext->pulse.pa_stream_peek)(pStream, &pMappedPCMFrames, &bytesMapped); + if (pulseResult < 0) { + break; /* Failed to map. Abort. */ + } + + framesMapped = bytesMapped / bpf; + if (framesMapped > 0) { + if (pMappedPCMFrames != NULL) { + ma_device_handle_backend_data_callback(pDevice, NULL, pMappedPCMFrames, framesMapped); + } else { + /* It's a hole. */ + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[PulseAudio] ma_device_on_read__pulse: Hole.\n"); + } + + pulseResult = ((ma_pa_stream_drop_proc)pDevice->pContext->pulse.pa_stream_drop)(pStream); + if (pulseResult < 0) { + break; /* Failed to drop the buffer. */ + } + + framesProcessed += framesMapped; + + } else { + /* Nothing was mapped. Just abort. */ + break; + } + } +} + +static ma_result ma_device_write_to_stream__pulse(ma_device* pDevice, ma_pa_stream* pStream, ma_uint64* pFramesProcessed) +{ + ma_result result = MA_SUCCESS; + ma_uint64 framesProcessed = 0; + size_t bytesMapped; + ma_uint32 bpf; + ma_uint32 deviceState; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pStream != NULL); + + bpf = ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + MA_ASSERT(bpf > 0); + + deviceState = ma_device_get_state(pDevice); + + bytesMapped = ((ma_pa_stream_writable_size_proc)pDevice->pContext->pulse.pa_stream_writable_size)(pStream); + if (bytesMapped != (size_t)-1) { + if (bytesMapped > 0) { + ma_uint64 framesMapped; + void* pMappedPCMFrames; + int pulseResult = ((ma_pa_stream_begin_write_proc)pDevice->pContext->pulse.pa_stream_begin_write)(pStream, &pMappedPCMFrames, &bytesMapped); + if (pulseResult < 0) { + result = ma_result_from_pulse(pulseResult); + goto done; + } + + framesMapped = bytesMapped / bpf; + + if (deviceState == ma_device_state_started || deviceState == ma_device_state_starting) { /* Check for starting state just in case this is being used to do the initial fill. */ + ma_device_handle_backend_data_callback(pDevice, pMappedPCMFrames, NULL, framesMapped); + } else { + /* Device is not started. Write silence. */ + ma_silence_pcm_frames(pMappedPCMFrames, framesMapped, pDevice->playback.format, pDevice->playback.channels); + } + + pulseResult = ((ma_pa_stream_write_proc)pDevice->pContext->pulse.pa_stream_write)(pStream, pMappedPCMFrames, bytesMapped, NULL, 0, MA_PA_SEEK_RELATIVE); + if (pulseResult < 0) { + result = ma_result_from_pulse(pulseResult); + goto done; /* Failed to write data to stream. */ + } + + framesProcessed += framesMapped; + } else { + result = MA_SUCCESS; /* No data available for writing. */ + goto done; + } + } else { + result = MA_ERROR; /* Failed to retrieve the writable size. Abort. */ + goto done; + } + +done: + if (pFramesProcessed != NULL) { + *pFramesProcessed = framesProcessed; + } + + return result; +} + +static void ma_device_on_write__pulse(ma_pa_stream* pStream, size_t byteCount, void* pUserData) +{ + ma_device* pDevice = (ma_device*)pUserData; + ma_uint32 bpf; + ma_uint64 frameCount; + ma_uint64 framesProcessed; + ma_uint32 deviceState; + ma_result result; + + MA_ASSERT(pDevice != NULL); + + /* + Don't do anything if the device isn't initialized yet. Yes, this can happen because PulseAudio + can fire this callback before the stream has even started. Ridiculous. + */ + deviceState = ma_device_get_state(pDevice); + if (deviceState != ma_device_state_starting && deviceState != ma_device_state_started) { + return; + } + + bpf = ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + MA_ASSERT(bpf > 0); + + frameCount = byteCount / bpf; + framesProcessed = 0; + + while (framesProcessed < frameCount) { + ma_uint64 framesProcessedThisIteration; + + /* Don't keep trying to process frames if the device isn't started. */ + deviceState = ma_device_get_state(pDevice); + if (deviceState != ma_device_state_starting && deviceState != ma_device_state_started) { + break; + } + + result = ma_device_write_to_stream__pulse(pDevice, pStream, &framesProcessedThisIteration); + if (result != MA_SUCCESS) { + break; + } + + framesProcessed += framesProcessedThisIteration; + } +} + +static void ma_device_on_suspended__pulse(ma_pa_stream* pStream, void* pUserData) +{ + ma_device* pDevice = (ma_device*)pUserData; + int suspended; + + (void)pStream; + + suspended = ((ma_pa_stream_is_suspended_proc)pDevice->pContext->pulse.pa_stream_is_suspended)(pStream); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[Pulse] Device suspended state changed. pa_stream_is_suspended() returned %d.\n", suspended); + + if (suspended < 0) { + return; + } + + if (suspended == 1) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[Pulse] Device suspended state changed. Suspended.\n"); + ma_device__on_notification_stopped(pDevice); + } else { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "[Pulse] Device suspended state changed. Resumed.\n"); + ma_device__on_notification_started(pDevice); + } +} + +static void ma_device_on_rerouted__pulse(ma_pa_stream* pStream, void* pUserData) +{ + ma_device* pDevice = (ma_device*)pUserData; + + (void)pStream; + (void)pUserData; + + ma_device__on_notification_rerouted(pDevice); +} + +static ma_uint32 ma_calculate_period_size_in_frames_from_descriptor__pulse(const ma_device_descriptor* pDescriptor, ma_uint32 nativeSampleRate, ma_performance_profile performanceProfile) +{ + /* + There have been reports from users where buffers of < ~20ms result glitches when running through + PipeWire. To work around this we're going to have to use a different default buffer size. + */ + const ma_uint32 defaultPeriodSizeInMilliseconds_LowLatency = 25; + const ma_uint32 defaultPeriodSizeInMilliseconds_Conservative = MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_CONSERVATIVE; + + MA_ASSERT(nativeSampleRate != 0); + + if (pDescriptor->periodSizeInFrames == 0) { + if (pDescriptor->periodSizeInMilliseconds == 0) { + if (performanceProfile == ma_performance_profile_low_latency) { + return ma_calculate_buffer_size_in_frames_from_milliseconds(defaultPeriodSizeInMilliseconds_LowLatency, nativeSampleRate); + } else { + return ma_calculate_buffer_size_in_frames_from_milliseconds(defaultPeriodSizeInMilliseconds_Conservative, nativeSampleRate); + } + } else { + return ma_calculate_buffer_size_in_frames_from_milliseconds(pDescriptor->periodSizeInMilliseconds, nativeSampleRate); + } + } else { + return pDescriptor->periodSizeInFrames; + } +} + +static ma_result ma_device_init__pulse(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + /* + Notes for PulseAudio: + + - When both the period size in frames and milliseconds are 0, we default to miniaudio's + default buffer sizes rather than leaving it up to PulseAudio because I don't trust + PulseAudio to give us any kind of reasonable latency by default. + + - Do not ever, *ever* forget to use MA_PA_STREAM_ADJUST_LATENCY. If you don't specify this + flag, capture mode will just not work properly until you open another PulseAudio app. + */ + + ma_result result = MA_SUCCESS; + int error = 0; + const char* devPlayback = NULL; + const char* devCapture = NULL; + ma_format format = ma_format_unknown; + ma_uint32 channels = 0; + ma_uint32 sampleRate = 0; + ma_pa_sink_info sinkInfo; + ma_pa_source_info sourceInfo; + ma_pa_sample_spec ss; + ma_pa_channel_map cmap; + ma_pa_buffer_attr attr; + const ma_pa_sample_spec* pActualSS = NULL; + const ma_pa_buffer_attr* pActualAttr = NULL; + const ma_pa_channel_map* pActualChannelMap = NULL; + ma_uint32 iChannel; + ma_pa_stream_flags_t streamFlags; + + MA_ASSERT(pDevice != NULL); + MA_ZERO_OBJECT(&pDevice->pulse); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + /* No exclusive mode with the PulseAudio backend. */ + if (((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pConfig->playback.shareMode == ma_share_mode_exclusive) || + ((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) && pConfig->capture.shareMode == ma_share_mode_exclusive)) { + return MA_SHARE_MODE_NOT_SUPPORTED; + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + if (pDescriptorPlayback->pDeviceID != NULL) { + devPlayback = pDescriptorPlayback->pDeviceID->pulse; + } + + format = pDescriptorPlayback->format; + channels = pDescriptorPlayback->channels; + sampleRate = pDescriptorPlayback->sampleRate; + } + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + if (pDescriptorCapture->pDeviceID != NULL) { + devCapture = pDescriptorCapture->pDeviceID->pulse; + } + + format = pDescriptorCapture->format; + channels = pDescriptorCapture->channels; + sampleRate = pDescriptorCapture->sampleRate; + } + + + + result = ma_init_pa_mainloop_and_pa_context__pulse(pDevice->pContext, pDevice->pContext->pulse.pApplicationName, pDevice->pContext->pulse.pServerName, MA_FALSE, &pDevice->pulse.pMainLoop, &pDevice->pulse.pPulseContext); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to initialize PA mainloop and context for device.\n"); + return result; + } + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + result = ma_context_get_source_info__pulse(pDevice->pContext, devCapture, &sourceInfo); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to retrieve source info for capture device."); + goto on_error0; + } + + ss = sourceInfo.sample_spec; + cmap = sourceInfo.channel_map; + + /* Use the requested channel count if we have one. */ + if (pDescriptorCapture->channels != 0) { + ss.channels = pDescriptorCapture->channels; + } + + /* Use a default channel map. */ + ((ma_pa_channel_map_init_extend_proc)pDevice->pContext->pulse.pa_channel_map_init_extend)(&cmap, ss.channels, pConfig->pulse.channelMap); + + /* Use the requested sample rate if one was specified. */ + if (pDescriptorCapture->sampleRate != 0) { + ss.rate = pDescriptorCapture->sampleRate; + } + streamFlags = MA_PA_STREAM_START_CORKED | MA_PA_STREAM_ADJUST_LATENCY; + + if (ma_format_from_pulse(ss.format) == ma_format_unknown) { + if (ma_is_little_endian()) { + ss.format = MA_PA_SAMPLE_FLOAT32LE; + } else { + ss.format = MA_PA_SAMPLE_FLOAT32BE; + } + streamFlags |= MA_PA_STREAM_FIX_FORMAT; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] sample_spec.format not supported by miniaudio. Defaulting to PA_SAMPLE_FLOAT32.\n"); + } + if (ss.rate == 0) { + ss.rate = MA_DEFAULT_SAMPLE_RATE; + streamFlags |= MA_PA_STREAM_FIX_RATE; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] sample_spec.rate = 0. Defaulting to %d.\n", ss.rate); + } + if (ss.channels == 0) { + ss.channels = MA_DEFAULT_CHANNELS; + streamFlags |= MA_PA_STREAM_FIX_CHANNELS; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] sample_spec.channels = 0. Defaulting to %d.\n", ss.channels); + } + + /* We now have enough information to calculate our actual period size in frames. */ + pDescriptorCapture->periodSizeInFrames = ma_calculate_period_size_in_frames_from_descriptor__pulse(pDescriptorCapture, ss.rate, pConfig->performanceProfile); + + attr = ma_device__pa_buffer_attr_new(pDescriptorCapture->periodSizeInFrames, pDescriptorCapture->periodCount, &ss); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] Capture attr: maxlength=%d, tlength=%d, prebuf=%d, minreq=%d, fragsize=%d; periodSizeInFrames=%d\n", attr.maxlength, attr.tlength, attr.prebuf, attr.minreq, attr.fragsize, pDescriptorCapture->periodSizeInFrames); + + pDevice->pulse.pStreamCapture = ma_device__pa_stream_new__pulse(pDevice, pConfig->pulse.pStreamNameCapture, &ss, &cmap); + if (pDevice->pulse.pStreamCapture == NULL) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to create PulseAudio capture stream.\n"); + result = MA_ERROR; + goto on_error0; + } + + + /* The callback needs to be set before connecting the stream. */ + ((ma_pa_stream_set_read_callback_proc)pDevice->pContext->pulse.pa_stream_set_read_callback)((ma_pa_stream*)pDevice->pulse.pStreamCapture, ma_device_on_read__pulse, pDevice); + + /* State callback for checking when the device has been corked. */ + ((ma_pa_stream_set_suspended_callback_proc)pDevice->pContext->pulse.pa_stream_set_suspended_callback)((ma_pa_stream*)pDevice->pulse.pStreamCapture, ma_device_on_suspended__pulse, pDevice); + + /* Rerouting notification. */ + ((ma_pa_stream_set_moved_callback_proc)pDevice->pContext->pulse.pa_stream_set_moved_callback)((ma_pa_stream*)pDevice->pulse.pStreamCapture, ma_device_on_rerouted__pulse, pDevice); + + + /* Connect after we've got all of our internal state set up. */ + if (devCapture != NULL) { + streamFlags |= MA_PA_STREAM_DONT_MOVE; + } + + error = ((ma_pa_stream_connect_record_proc)pDevice->pContext->pulse.pa_stream_connect_record)((ma_pa_stream*)pDevice->pulse.pStreamCapture, devCapture, &attr, streamFlags); + if (error != MA_PA_OK) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to connect PulseAudio capture stream."); + result = ma_result_from_pulse(error); + goto on_error1; + } + + result = ma_wait_for_pa_stream_to_connect__pulse(pDevice->pContext, pDevice->pulse.pMainLoop, (ma_pa_stream*)pDevice->pulse.pStreamCapture); + if (result != MA_SUCCESS) { + goto on_error2; + } + + + /* Internal format. */ + pActualSS = ((ma_pa_stream_get_sample_spec_proc)pDevice->pContext->pulse.pa_stream_get_sample_spec)((ma_pa_stream*)pDevice->pulse.pStreamCapture); + if (pActualSS != NULL) { + ss = *pActualSS; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] Capture sample spec: format=%s, channels=%d, rate=%d\n", ma_get_format_name(ma_format_from_pulse(ss.format)), ss.channels, ss.rate); + } else { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] Failed to retrieve capture sample spec.\n"); + } + + pDescriptorCapture->format = ma_format_from_pulse(ss.format); + pDescriptorCapture->channels = ss.channels; + pDescriptorCapture->sampleRate = ss.rate; + + if (pDescriptorCapture->format == ma_format_unknown || pDescriptorCapture->channels == 0 || pDescriptorCapture->sampleRate == 0) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Capture sample spec is invalid. Device unusable by miniaudio. format=%s, channels=%d, sampleRate=%d.\n", ma_get_format_name(pDescriptorCapture->format), pDescriptorCapture->channels, pDescriptorCapture->sampleRate); + result = MA_ERROR; + goto on_error4; + } + + + /* Internal channel map. */ + pActualChannelMap = ((ma_pa_stream_get_channel_map_proc)pDevice->pContext->pulse.pa_stream_get_channel_map)((ma_pa_stream*)pDevice->pulse.pStreamCapture); + if (pActualChannelMap == NULL) { + pActualChannelMap = &cmap; /* Fallback just in case. */ + } + + /* + Bug in PipeWire. There have been reports that PipeWire is returning AUX channels when reporting + the channel map. To somewhat workaround this, I'm hacking in a hard coded channel map for mono + and stereo. In this case it should be safe to assume mono = MONO and stereo = LEFT/RIGHT. For + all other channel counts we need to just put up with whatever PipeWire reports and hope it gets + fixed sooner than later. I might remove this hack later. + */ + if (pDescriptorCapture->channels > 2) { + for (iChannel = 0; iChannel < pDescriptorCapture->channels; iChannel += 1) { + pDescriptorCapture->channelMap[iChannel] = ma_channel_position_from_pulse(pActualChannelMap->map[iChannel]); + } + } else { + /* Hack for mono and stereo. */ + if (pDescriptorCapture->channels == 1) { + pDescriptorCapture->channelMap[0] = MA_CHANNEL_MONO; + } else if (pDescriptorCapture->channels == 2) { + pDescriptorCapture->channelMap[0] = MA_CHANNEL_FRONT_LEFT; + pDescriptorCapture->channelMap[1] = MA_CHANNEL_FRONT_RIGHT; + } else { + MA_ASSERT(MA_FALSE); /* Should never hit this. */ + } + } + + + /* Buffer. */ + pActualAttr = ((ma_pa_stream_get_buffer_attr_proc)pDevice->pContext->pulse.pa_stream_get_buffer_attr)((ma_pa_stream*)pDevice->pulse.pStreamCapture); + if (pActualAttr != NULL) { + attr = *pActualAttr; + } + + if (attr.fragsize > 0) { + pDescriptorCapture->periodCount = ma_max(attr.maxlength / attr.fragsize, 1); + } else { + pDescriptorCapture->periodCount = 1; + } + + pDescriptorCapture->periodSizeInFrames = attr.maxlength / ma_get_bytes_per_frame(pDescriptorCapture->format, pDescriptorCapture->channels) / pDescriptorCapture->periodCount; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] Capture actual attr: maxlength=%d, tlength=%d, prebuf=%d, minreq=%d, fragsize=%d; periodSizeInFrames=%d\n", attr.maxlength, attr.tlength, attr.prebuf, attr.minreq, attr.fragsize, pDescriptorCapture->periodSizeInFrames); + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + result = ma_context_get_sink_info__pulse(pDevice->pContext, devPlayback, &sinkInfo); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to retrieve sink info for playback device.\n"); + goto on_error2; + } + + ss = sinkInfo.sample_spec; + cmap = sinkInfo.channel_map; + + /* Use the requested channel count if we have one. */ + if (pDescriptorPlayback->channels != 0) { + ss.channels = pDescriptorPlayback->channels; + } + + /* Use a default channel map. */ + ((ma_pa_channel_map_init_extend_proc)pDevice->pContext->pulse.pa_channel_map_init_extend)(&cmap, ss.channels, pConfig->pulse.channelMap); + + + /* Use the requested sample rate if one was specified. */ + if (pDescriptorPlayback->sampleRate != 0) { + ss.rate = pDescriptorPlayback->sampleRate; + } + + streamFlags = MA_PA_STREAM_START_CORKED | MA_PA_STREAM_ADJUST_LATENCY; + if (ma_format_from_pulse(ss.format) == ma_format_unknown) { + if (ma_is_little_endian()) { + ss.format = MA_PA_SAMPLE_FLOAT32LE; + } else { + ss.format = MA_PA_SAMPLE_FLOAT32BE; + } + streamFlags |= MA_PA_STREAM_FIX_FORMAT; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] sample_spec.format not supported by miniaudio. Defaulting to PA_SAMPLE_FLOAT32.\n"); + } + if (ss.rate == 0) { + ss.rate = MA_DEFAULT_SAMPLE_RATE; + streamFlags |= MA_PA_STREAM_FIX_RATE; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] sample_spec.rate = 0. Defaulting to %d.\n", ss.rate); + } + if (ss.channels == 0) { + ss.channels = MA_DEFAULT_CHANNELS; + streamFlags |= MA_PA_STREAM_FIX_CHANNELS; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] sample_spec.channels = 0. Defaulting to %d.\n", ss.channels); + } + + /* We now have enough information to calculate the actual buffer size in frames. */ + pDescriptorPlayback->periodSizeInFrames = ma_calculate_period_size_in_frames_from_descriptor__pulse(pDescriptorPlayback, ss.rate, pConfig->performanceProfile); + + attr = ma_device__pa_buffer_attr_new(pDescriptorPlayback->periodSizeInFrames, pDescriptorPlayback->periodCount, &ss); + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] Playback attr: maxlength=%d, tlength=%d, prebuf=%d, minreq=%d, fragsize=%d; periodSizeInFrames=%d\n", attr.maxlength, attr.tlength, attr.prebuf, attr.minreq, attr.fragsize, pDescriptorPlayback->periodSizeInFrames); + + pDevice->pulse.pStreamPlayback = ma_device__pa_stream_new__pulse(pDevice, pConfig->pulse.pStreamNamePlayback, &ss, &cmap); + if (pDevice->pulse.pStreamPlayback == NULL) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to create PulseAudio playback stream.\n"); + result = MA_ERROR; + goto on_error2; + } + + + /* + Note that this callback will be fired as soon as the stream is connected, even though it's started as corked. The callback needs to handle a + device state of ma_device_state_uninitialized. + */ + ((ma_pa_stream_set_write_callback_proc)pDevice->pContext->pulse.pa_stream_set_write_callback)((ma_pa_stream*)pDevice->pulse.pStreamPlayback, ma_device_on_write__pulse, pDevice); + + /* State callback for checking when the device has been corked. */ + ((ma_pa_stream_set_suspended_callback_proc)pDevice->pContext->pulse.pa_stream_set_suspended_callback)((ma_pa_stream*)pDevice->pulse.pStreamPlayback, ma_device_on_suspended__pulse, pDevice); + + /* Rerouting notification. */ + ((ma_pa_stream_set_moved_callback_proc)pDevice->pContext->pulse.pa_stream_set_moved_callback)((ma_pa_stream*)pDevice->pulse.pStreamPlayback, ma_device_on_rerouted__pulse, pDevice); + + + /* Connect after we've got all of our internal state set up. */ + if (devPlayback != NULL) { + streamFlags |= MA_PA_STREAM_DONT_MOVE; + } + + error = ((ma_pa_stream_connect_playback_proc)pDevice->pContext->pulse.pa_stream_connect_playback)((ma_pa_stream*)pDevice->pulse.pStreamPlayback, devPlayback, &attr, streamFlags, NULL, NULL); + if (error != MA_PA_OK) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to connect PulseAudio playback stream."); + result = ma_result_from_pulse(error); + goto on_error3; + } + + result = ma_wait_for_pa_stream_to_connect__pulse(pDevice->pContext, pDevice->pulse.pMainLoop, (ma_pa_stream*)pDevice->pulse.pStreamPlayback); + if (result != MA_SUCCESS) { + goto on_error3; + } + + + /* Internal format. */ + pActualSS = ((ma_pa_stream_get_sample_spec_proc)pDevice->pContext->pulse.pa_stream_get_sample_spec)((ma_pa_stream*)pDevice->pulse.pStreamPlayback); + if (pActualSS != NULL) { + ss = *pActualSS; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] Playback sample spec: format=%s, channels=%d, rate=%d\n", ma_get_format_name(ma_format_from_pulse(ss.format)), ss.channels, ss.rate); + } else { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] Failed to retrieve playback sample spec.\n"); + } + + pDescriptorPlayback->format = ma_format_from_pulse(ss.format); + pDescriptorPlayback->channels = ss.channels; + pDescriptorPlayback->sampleRate = ss.rate; + + if (pDescriptorPlayback->format == ma_format_unknown || pDescriptorPlayback->channels == 0 || pDescriptorPlayback->sampleRate == 0) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Playback sample spec is invalid. Device unusable by miniaudio. format=%s, channels=%d, sampleRate=%d.\n", ma_get_format_name(pDescriptorPlayback->format), pDescriptorPlayback->channels, pDescriptorPlayback->sampleRate); + result = MA_ERROR; + goto on_error4; + } + + + /* Internal channel map. */ + pActualChannelMap = ((ma_pa_stream_get_channel_map_proc)pDevice->pContext->pulse.pa_stream_get_channel_map)((ma_pa_stream*)pDevice->pulse.pStreamPlayback); + if (pActualChannelMap == NULL) { + pActualChannelMap = &cmap; /* Fallback just in case. */ + } + + /* + Bug in PipeWire. There have been reports that PipeWire is returning AUX channels when reporting + the channel map. To somewhat workaround this, I'm hacking in a hard coded channel map for mono + and stereo. In this case it should be safe to assume mono = MONO and stereo = LEFT/RIGHT. For + all other channel counts we need to just put up with whatever PipeWire reports and hope it gets + fixed sooner than later. I might remove this hack later. + */ + if (pDescriptorPlayback->channels > 2) { + for (iChannel = 0; iChannel < pDescriptorPlayback->channels; iChannel += 1) { + pDescriptorPlayback->channelMap[iChannel] = ma_channel_position_from_pulse(pActualChannelMap->map[iChannel]); + } + } else { + /* Hack for mono and stereo. */ + if (pDescriptorPlayback->channels == 1) { + pDescriptorPlayback->channelMap[0] = MA_CHANNEL_MONO; + } else if (pDescriptorPlayback->channels == 2) { + pDescriptorPlayback->channelMap[0] = MA_CHANNEL_FRONT_LEFT; + pDescriptorPlayback->channelMap[1] = MA_CHANNEL_FRONT_RIGHT; + } else { + MA_ASSERT(MA_FALSE); /* Should never hit this. */ + } + } + + + /* Buffer. */ + pActualAttr = ((ma_pa_stream_get_buffer_attr_proc)pDevice->pContext->pulse.pa_stream_get_buffer_attr)((ma_pa_stream*)pDevice->pulse.pStreamPlayback); + if (pActualAttr != NULL) { + attr = *pActualAttr; + } + + if (attr.tlength > 0) { + pDescriptorPlayback->periodCount = ma_max(attr.maxlength / attr.tlength, 1); + } else { + pDescriptorPlayback->periodCount = 1; + } + + pDescriptorPlayback->periodSizeInFrames = attr.maxlength / ma_get_bytes_per_frame(pDescriptorPlayback->format, pDescriptorPlayback->channels) / pDescriptorPlayback->periodCount; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[PulseAudio] Playback actual attr: maxlength=%d, tlength=%d, prebuf=%d, minreq=%d, fragsize=%d; internalPeriodSizeInFrames=%d\n", attr.maxlength, attr.tlength, attr.prebuf, attr.minreq, attr.fragsize, pDescriptorPlayback->periodSizeInFrames); + } + + + /* + We need a ring buffer for handling duplex mode. We can use the main duplex ring buffer in the main + part of the ma_device struct. We cannot, however, depend on ma_device_init() initializing this for + us later on because that will only do it if it's a fully asynchronous backend - i.e. the + onDeviceDataLoop callback is NULL, which is not the case for PulseAudio. + */ + if (pConfig->deviceType == ma_device_type_duplex) { + ma_format rbFormat = (format != ma_format_unknown) ? format : pDescriptorCapture->format; + ma_uint32 rbChannels = (channels > 0) ? channels : pDescriptorCapture->channels; + ma_uint32 rbSampleRate = (sampleRate > 0) ? sampleRate : pDescriptorCapture->sampleRate; + + result = ma_duplex_rb_init(rbFormat, rbChannels, rbSampleRate, pDescriptorCapture->sampleRate, pDescriptorCapture->periodSizeInFrames, &pDevice->pContext->allocationCallbacks, &pDevice->duplexRB); + if (result != MA_SUCCESS) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to initialize ring buffer. %s.\n", ma_result_description(result)); + goto on_error4; + } + } + + return MA_SUCCESS; + + +on_error4: + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ((ma_pa_stream_disconnect_proc)pDevice->pContext->pulse.pa_stream_disconnect)((ma_pa_stream*)pDevice->pulse.pStreamPlayback); + } +on_error3: + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ((ma_pa_stream_unref_proc)pDevice->pContext->pulse.pa_stream_unref)((ma_pa_stream*)pDevice->pulse.pStreamPlayback); + } +on_error2: + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ((ma_pa_stream_disconnect_proc)pDevice->pContext->pulse.pa_stream_disconnect)((ma_pa_stream*)pDevice->pulse.pStreamCapture); + } +on_error1: + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ((ma_pa_stream_unref_proc)pDevice->pContext->pulse.pa_stream_unref)((ma_pa_stream*)pDevice->pulse.pStreamCapture); + } +on_error0: + return result; +} + + +static void ma_pulse_operation_complete_callback(ma_pa_stream* pStream, int success, void* pUserData) +{ + ma_bool32* pIsSuccessful = (ma_bool32*)pUserData; + MA_ASSERT(pIsSuccessful != NULL); + + *pIsSuccessful = (ma_bool32)success; + + (void)pStream; /* Unused. */ +} + +static ma_result ma_device__cork_stream__pulse(ma_device* pDevice, ma_device_type deviceType, int cork) +{ + ma_context* pContext = pDevice->pContext; + ma_bool32 wasSuccessful; + ma_pa_stream* pStream; + ma_pa_operation* pOP; + ma_result result; + + /* This should not be called with a duplex device type. */ + if (deviceType == ma_device_type_duplex) { + return MA_INVALID_ARGS; + } + + wasSuccessful = MA_FALSE; + + pStream = (ma_pa_stream*)((deviceType == ma_device_type_capture) ? pDevice->pulse.pStreamCapture : pDevice->pulse.pStreamPlayback); + MA_ASSERT(pStream != NULL); + + pOP = ((ma_pa_stream_cork_proc)pContext->pulse.pa_stream_cork)(pStream, cork, ma_pulse_operation_complete_callback, &wasSuccessful); + if (pOP == NULL) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to cork PulseAudio stream."); + return MA_ERROR; + } + + result = ma_wait_for_operation_and_unref__pulse(pDevice->pContext, pDevice->pulse.pMainLoop, pOP); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] An error occurred while waiting for the PulseAudio stream to cork."); + return result; + } + + if (!wasSuccessful) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[PulseAudio] Failed to %s PulseAudio stream.", (cork) ? "stop" : "start"); + return MA_ERROR; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_start__pulse(ma_device* pDevice) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + result = ma_device__cork_stream__pulse(pDevice, ma_device_type_capture, 0); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + /* + We need to fill some data before uncorking. Not doing this will result in the write callback + never getting fired. We're not going to abort if writing fails because I still want the device + to get uncorked. + */ + ma_device_write_to_stream__pulse(pDevice, (ma_pa_stream*)(pDevice->pulse.pStreamPlayback), NULL); /* No need to check the result here. Always want to fall through an uncork.*/ + + result = ma_device__cork_stream__pulse(pDevice, ma_device_type_playback, 0); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__pulse(ma_device* pDevice) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + result = ma_device__cork_stream__pulse(pDevice, ma_device_type_capture, 1); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + /* + Ideally we would drain the device here, but there's been cases where PulseAudio seems to be + broken on some systems to the point where no audio processing seems to happen. When this + happens, draining never completes and we get stuck here. For now I'm disabling draining of + the device so we don't just freeze the application. + */ + #if 0 + ma_pa_operation* pOP = ((ma_pa_stream_drain_proc)pDevice->pContext->pulse.pa_stream_drain)((ma_pa_stream*)pDevice->pulse.pStreamPlayback, ma_pulse_operation_complete_callback, &wasSuccessful); + ma_wait_for_operation_and_unref__pulse(pDevice->pContext, pDevice->pulse.pMainLoop, pOP); + #endif + + result = ma_device__cork_stream__pulse(pDevice, ma_device_type_playback, 1); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_data_loop__pulse(ma_device* pDevice) +{ + int resultPA; + + MA_ASSERT(pDevice != NULL); + + /* NOTE: Don't start the device here. It'll be done at a higher level. */ + + /* + All data is handled through callbacks. All we need to do is iterate over the main loop and let + the callbacks deal with it. + */ + while (ma_device_get_state(pDevice) == ma_device_state_started) { + resultPA = ((ma_pa_mainloop_iterate_proc)pDevice->pContext->pulse.pa_mainloop_iterate)((ma_pa_mainloop*)pDevice->pulse.pMainLoop, 1, NULL); + if (resultPA < 0) { + break; + } + } + + /* NOTE: Don't stop the device here. It'll be done at a higher level. */ + return MA_SUCCESS; +} + +static ma_result ma_device_data_loop_wakeup__pulse(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + ((ma_pa_mainloop_wakeup_proc)pDevice->pContext->pulse.pa_mainloop_wakeup)((ma_pa_mainloop*)pDevice->pulse.pMainLoop); + + return MA_SUCCESS; +} + +static ma_result ma_context_uninit__pulse(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_pulseaudio); + + ((ma_pa_context_disconnect_proc)pContext->pulse.pa_context_disconnect)((ma_pa_context*)pContext->pulse.pPulseContext); + ((ma_pa_context_unref_proc)pContext->pulse.pa_context_unref)((ma_pa_context*)pContext->pulse.pPulseContext); + ((ma_pa_mainloop_free_proc)pContext->pulse.pa_mainloop_free)((ma_pa_mainloop*)pContext->pulse.pMainLoop); + + ma_free(pContext->pulse.pServerName, &pContext->allocationCallbacks); + ma_free(pContext->pulse.pApplicationName, &pContext->allocationCallbacks); + +#ifndef MA_NO_RUNTIME_LINKING + ma_dlclose(ma_context_get_log(pContext), pContext->pulse.pulseSO); +#endif + + return MA_SUCCESS; +} + +static ma_result ma_context_init__pulse(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + ma_result result; +#ifndef MA_NO_RUNTIME_LINKING + const char* libpulseNames[] = { + "libpulse.so", + "libpulse.so.0" + }; + size_t i; + + for (i = 0; i < ma_countof(libpulseNames); ++i) { + pContext->pulse.pulseSO = ma_dlopen(ma_context_get_log(pContext), libpulseNames[i]); + if (pContext->pulse.pulseSO != NULL) { + break; + } + } + + if (pContext->pulse.pulseSO == NULL) { + return MA_NO_BACKEND; + } + + pContext->pulse.pa_mainloop_new = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_mainloop_new"); + pContext->pulse.pa_mainloop_free = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_mainloop_free"); + pContext->pulse.pa_mainloop_quit = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_mainloop_quit"); + pContext->pulse.pa_mainloop_get_api = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_mainloop_get_api"); + pContext->pulse.pa_mainloop_iterate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_mainloop_iterate"); + pContext->pulse.pa_mainloop_wakeup = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_mainloop_wakeup"); + pContext->pulse.pa_threaded_mainloop_new = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_new"); + pContext->pulse.pa_threaded_mainloop_free = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_free"); + pContext->pulse.pa_threaded_mainloop_start = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_start"); + pContext->pulse.pa_threaded_mainloop_stop = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_stop"); + pContext->pulse.pa_threaded_mainloop_lock = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_lock"); + pContext->pulse.pa_threaded_mainloop_unlock = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_unlock"); + pContext->pulse.pa_threaded_mainloop_wait = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_wait"); + pContext->pulse.pa_threaded_mainloop_signal = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_signal"); + pContext->pulse.pa_threaded_mainloop_accept = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_accept"); + pContext->pulse.pa_threaded_mainloop_get_retval = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_get_retval"); + pContext->pulse.pa_threaded_mainloop_get_api = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_get_api"); + pContext->pulse.pa_threaded_mainloop_in_thread = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_in_thread"); + pContext->pulse.pa_threaded_mainloop_set_name = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_threaded_mainloop_set_name"); + pContext->pulse.pa_context_new = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_new"); + pContext->pulse.pa_context_unref = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_unref"); + pContext->pulse.pa_context_connect = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_connect"); + pContext->pulse.pa_context_disconnect = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_disconnect"); + pContext->pulse.pa_context_set_state_callback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_set_state_callback"); + pContext->pulse.pa_context_get_state = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_get_state"); + pContext->pulse.pa_context_get_sink_info_list = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_get_sink_info_list"); + pContext->pulse.pa_context_get_source_info_list = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_get_source_info_list"); + pContext->pulse.pa_context_get_sink_info_by_name = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_get_sink_info_by_name"); + pContext->pulse.pa_context_get_source_info_by_name = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_context_get_source_info_by_name"); + pContext->pulse.pa_operation_unref = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_operation_unref"); + pContext->pulse.pa_operation_get_state = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_operation_get_state"); + pContext->pulse.pa_channel_map_init_extend = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_channel_map_init_extend"); + pContext->pulse.pa_channel_map_valid = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_channel_map_valid"); + pContext->pulse.pa_channel_map_compatible = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_channel_map_compatible"); + pContext->pulse.pa_stream_new = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_new"); + pContext->pulse.pa_stream_unref = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_unref"); + pContext->pulse.pa_stream_connect_playback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_connect_playback"); + pContext->pulse.pa_stream_connect_record = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_connect_record"); + pContext->pulse.pa_stream_disconnect = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_disconnect"); + pContext->pulse.pa_stream_get_state = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_get_state"); + pContext->pulse.pa_stream_get_sample_spec = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_get_sample_spec"); + pContext->pulse.pa_stream_get_channel_map = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_get_channel_map"); + pContext->pulse.pa_stream_get_buffer_attr = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_get_buffer_attr"); + pContext->pulse.pa_stream_set_buffer_attr = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_set_buffer_attr"); + pContext->pulse.pa_stream_get_device_name = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_get_device_name"); + pContext->pulse.pa_stream_set_write_callback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_set_write_callback"); + pContext->pulse.pa_stream_set_read_callback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_set_read_callback"); + pContext->pulse.pa_stream_set_suspended_callback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_set_suspended_callback"); + pContext->pulse.pa_stream_set_moved_callback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_set_moved_callback"); + pContext->pulse.pa_stream_is_suspended = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_is_suspended"); + pContext->pulse.pa_stream_flush = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_flush"); + pContext->pulse.pa_stream_drain = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_drain"); + pContext->pulse.pa_stream_is_corked = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_is_corked"); + pContext->pulse.pa_stream_cork = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_cork"); + pContext->pulse.pa_stream_trigger = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_trigger"); + pContext->pulse.pa_stream_begin_write = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_begin_write"); + pContext->pulse.pa_stream_write = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_write"); + pContext->pulse.pa_stream_peek = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_peek"); + pContext->pulse.pa_stream_drop = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_drop"); + pContext->pulse.pa_stream_writable_size = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_writable_size"); + pContext->pulse.pa_stream_readable_size = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->pulse.pulseSO, "pa_stream_readable_size"); +#else + /* This strange assignment system is just for type safety. */ + ma_pa_mainloop_new_proc _pa_mainloop_new = pa_mainloop_new; + ma_pa_mainloop_free_proc _pa_mainloop_free = pa_mainloop_free; + ma_pa_mainloop_quit_proc _pa_mainloop_quit = pa_mainloop_quit; + ma_pa_mainloop_get_api_proc _pa_mainloop_get_api = pa_mainloop_get_api; + ma_pa_mainloop_iterate_proc _pa_mainloop_iterate = pa_mainloop_iterate; + ma_pa_mainloop_wakeup_proc _pa_mainloop_wakeup = pa_mainloop_wakeup; + ma_pa_threaded_mainloop_new_proc _pa_threaded_mainloop_new = pa_threaded_mainloop_new; + ma_pa_threaded_mainloop_free_proc _pa_threaded_mainloop_free = pa_threaded_mainloop_free; + ma_pa_threaded_mainloop_start_proc _pa_threaded_mainloop_start = pa_threaded_mainloop_start; + ma_pa_threaded_mainloop_stop_proc _pa_threaded_mainloop_stop = pa_threaded_mainloop_stop; + ma_pa_threaded_mainloop_lock_proc _pa_threaded_mainloop_lock = pa_threaded_mainloop_lock; + ma_pa_threaded_mainloop_unlock_proc _pa_threaded_mainloop_unlock = pa_threaded_mainloop_unlock; + ma_pa_threaded_mainloop_wait_proc _pa_threaded_mainloop_wait = pa_threaded_mainloop_wait; + ma_pa_threaded_mainloop_signal_proc _pa_threaded_mainloop_signal = pa_threaded_mainloop_signal; + ma_pa_threaded_mainloop_accept_proc _pa_threaded_mainloop_accept = pa_threaded_mainloop_accept; + ma_pa_threaded_mainloop_get_retval_proc _pa_threaded_mainloop_get_retval = pa_threaded_mainloop_get_retval; + ma_pa_threaded_mainloop_get_api_proc _pa_threaded_mainloop_get_api = pa_threaded_mainloop_get_api; + ma_pa_threaded_mainloop_in_thread_proc _pa_threaded_mainloop_in_thread = pa_threaded_mainloop_in_thread; + ma_pa_threaded_mainloop_set_name_proc _pa_threaded_mainloop_set_name = pa_threaded_mainloop_set_name; + ma_pa_context_new_proc _pa_context_new = pa_context_new; + ma_pa_context_unref_proc _pa_context_unref = pa_context_unref; + ma_pa_context_connect_proc _pa_context_connect = pa_context_connect; + ma_pa_context_disconnect_proc _pa_context_disconnect = pa_context_disconnect; + ma_pa_context_set_state_callback_proc _pa_context_set_state_callback = pa_context_set_state_callback; + ma_pa_context_get_state_proc _pa_context_get_state = pa_context_get_state; + ma_pa_context_get_sink_info_list_proc _pa_context_get_sink_info_list = pa_context_get_sink_info_list; + ma_pa_context_get_source_info_list_proc _pa_context_get_source_info_list = pa_context_get_source_info_list; + ma_pa_context_get_sink_info_by_name_proc _pa_context_get_sink_info_by_name = pa_context_get_sink_info_by_name; + ma_pa_context_get_source_info_by_name_proc _pa_context_get_source_info_by_name= pa_context_get_source_info_by_name; + ma_pa_operation_unref_proc _pa_operation_unref = pa_operation_unref; + ma_pa_operation_get_state_proc _pa_operation_get_state = pa_operation_get_state; + ma_pa_channel_map_init_extend_proc _pa_channel_map_init_extend = pa_channel_map_init_extend; + ma_pa_channel_map_valid_proc _pa_channel_map_valid = pa_channel_map_valid; + ma_pa_channel_map_compatible_proc _pa_channel_map_compatible = pa_channel_map_compatible; + ma_pa_stream_new_proc _pa_stream_new = pa_stream_new; + ma_pa_stream_unref_proc _pa_stream_unref = pa_stream_unref; + ma_pa_stream_connect_playback_proc _pa_stream_connect_playback = pa_stream_connect_playback; + ma_pa_stream_connect_record_proc _pa_stream_connect_record = pa_stream_connect_record; + ma_pa_stream_disconnect_proc _pa_stream_disconnect = pa_stream_disconnect; + ma_pa_stream_get_state_proc _pa_stream_get_state = pa_stream_get_state; + ma_pa_stream_get_sample_spec_proc _pa_stream_get_sample_spec = pa_stream_get_sample_spec; + ma_pa_stream_get_channel_map_proc _pa_stream_get_channel_map = pa_stream_get_channel_map; + ma_pa_stream_get_buffer_attr_proc _pa_stream_get_buffer_attr = pa_stream_get_buffer_attr; + ma_pa_stream_set_buffer_attr_proc _pa_stream_set_buffer_attr = pa_stream_set_buffer_attr; + ma_pa_stream_get_device_name_proc _pa_stream_get_device_name = pa_stream_get_device_name; + ma_pa_stream_set_write_callback_proc _pa_stream_set_write_callback = pa_stream_set_write_callback; + ma_pa_stream_set_read_callback_proc _pa_stream_set_read_callback = pa_stream_set_read_callback; + ma_pa_stream_set_suspended_callback_proc _pa_stream_set_suspended_callback = pa_stream_set_suspended_callback; + ma_pa_stream_set_moved_callback_proc _pa_stream_set_moved_callback = pa_stream_set_moved_callback; + ma_pa_stream_is_suspended_proc _pa_stream_is_suspended = pa_stream_is_suspended; + ma_pa_stream_flush_proc _pa_stream_flush = pa_stream_flush; + ma_pa_stream_drain_proc _pa_stream_drain = pa_stream_drain; + ma_pa_stream_is_corked_proc _pa_stream_is_corked = pa_stream_is_corked; + ma_pa_stream_cork_proc _pa_stream_cork = pa_stream_cork; + ma_pa_stream_trigger_proc _pa_stream_trigger = pa_stream_trigger; + ma_pa_stream_begin_write_proc _pa_stream_begin_write = pa_stream_begin_write; + ma_pa_stream_write_proc _pa_stream_write = pa_stream_write; + ma_pa_stream_peek_proc _pa_stream_peek = pa_stream_peek; + ma_pa_stream_drop_proc _pa_stream_drop = pa_stream_drop; + ma_pa_stream_writable_size_proc _pa_stream_writable_size = pa_stream_writable_size; + ma_pa_stream_readable_size_proc _pa_stream_readable_size = pa_stream_readable_size; + + pContext->pulse.pa_mainloop_new = (ma_proc)_pa_mainloop_new; + pContext->pulse.pa_mainloop_free = (ma_proc)_pa_mainloop_free; + pContext->pulse.pa_mainloop_quit = (ma_proc)_pa_mainloop_quit; + pContext->pulse.pa_mainloop_get_api = (ma_proc)_pa_mainloop_get_api; + pContext->pulse.pa_mainloop_iterate = (ma_proc)_pa_mainloop_iterate; + pContext->pulse.pa_mainloop_wakeup = (ma_proc)_pa_mainloop_wakeup; + pContext->pulse.pa_threaded_mainloop_new = (ma_proc)_pa_threaded_mainloop_new; + pContext->pulse.pa_threaded_mainloop_free = (ma_proc)_pa_threaded_mainloop_free; + pContext->pulse.pa_threaded_mainloop_start = (ma_proc)_pa_threaded_mainloop_start; + pContext->pulse.pa_threaded_mainloop_stop = (ma_proc)_pa_threaded_mainloop_stop; + pContext->pulse.pa_threaded_mainloop_lock = (ma_proc)_pa_threaded_mainloop_lock; + pContext->pulse.pa_threaded_mainloop_unlock = (ma_proc)_pa_threaded_mainloop_unlock; + pContext->pulse.pa_threaded_mainloop_wait = (ma_proc)_pa_threaded_mainloop_wait; + pContext->pulse.pa_threaded_mainloop_signal = (ma_proc)_pa_threaded_mainloop_signal; + pContext->pulse.pa_threaded_mainloop_accept = (ma_proc)_pa_threaded_mainloop_accept; + pContext->pulse.pa_threaded_mainloop_get_retval = (ma_proc)_pa_threaded_mainloop_get_retval; + pContext->pulse.pa_threaded_mainloop_get_api = (ma_proc)_pa_threaded_mainloop_get_api; + pContext->pulse.pa_threaded_mainloop_in_thread = (ma_proc)_pa_threaded_mainloop_in_thread; + pContext->pulse.pa_threaded_mainloop_set_name = (ma_proc)_pa_threaded_mainloop_set_name; + pContext->pulse.pa_context_new = (ma_proc)_pa_context_new; + pContext->pulse.pa_context_unref = (ma_proc)_pa_context_unref; + pContext->pulse.pa_context_connect = (ma_proc)_pa_context_connect; + pContext->pulse.pa_context_disconnect = (ma_proc)_pa_context_disconnect; + pContext->pulse.pa_context_set_state_callback = (ma_proc)_pa_context_set_state_callback; + pContext->pulse.pa_context_get_state = (ma_proc)_pa_context_get_state; + pContext->pulse.pa_context_get_sink_info_list = (ma_proc)_pa_context_get_sink_info_list; + pContext->pulse.pa_context_get_source_info_list = (ma_proc)_pa_context_get_source_info_list; + pContext->pulse.pa_context_get_sink_info_by_name = (ma_proc)_pa_context_get_sink_info_by_name; + pContext->pulse.pa_context_get_source_info_by_name = (ma_proc)_pa_context_get_source_info_by_name; + pContext->pulse.pa_operation_unref = (ma_proc)_pa_operation_unref; + pContext->pulse.pa_operation_get_state = (ma_proc)_pa_operation_get_state; + pContext->pulse.pa_channel_map_init_extend = (ma_proc)_pa_channel_map_init_extend; + pContext->pulse.pa_channel_map_valid = (ma_proc)_pa_channel_map_valid; + pContext->pulse.pa_channel_map_compatible = (ma_proc)_pa_channel_map_compatible; + pContext->pulse.pa_stream_new = (ma_proc)_pa_stream_new; + pContext->pulse.pa_stream_unref = (ma_proc)_pa_stream_unref; + pContext->pulse.pa_stream_connect_playback = (ma_proc)_pa_stream_connect_playback; + pContext->pulse.pa_stream_connect_record = (ma_proc)_pa_stream_connect_record; + pContext->pulse.pa_stream_disconnect = (ma_proc)_pa_stream_disconnect; + pContext->pulse.pa_stream_get_state = (ma_proc)_pa_stream_get_state; + pContext->pulse.pa_stream_get_sample_spec = (ma_proc)_pa_stream_get_sample_spec; + pContext->pulse.pa_stream_get_channel_map = (ma_proc)_pa_stream_get_channel_map; + pContext->pulse.pa_stream_get_buffer_attr = (ma_proc)_pa_stream_get_buffer_attr; + pContext->pulse.pa_stream_set_buffer_attr = (ma_proc)_pa_stream_set_buffer_attr; + pContext->pulse.pa_stream_get_device_name = (ma_proc)_pa_stream_get_device_name; + pContext->pulse.pa_stream_set_write_callback = (ma_proc)_pa_stream_set_write_callback; + pContext->pulse.pa_stream_set_read_callback = (ma_proc)_pa_stream_set_read_callback; + pContext->pulse.pa_stream_set_suspended_callback = (ma_proc)_pa_stream_set_suspended_callback; + pContext->pulse.pa_stream_set_moved_callback = (ma_proc)_pa_stream_set_moved_callback; + pContext->pulse.pa_stream_is_suspended = (ma_proc)_pa_stream_is_suspended; + pContext->pulse.pa_stream_flush = (ma_proc)_pa_stream_flush; + pContext->pulse.pa_stream_drain = (ma_proc)_pa_stream_drain; + pContext->pulse.pa_stream_is_corked = (ma_proc)_pa_stream_is_corked; + pContext->pulse.pa_stream_cork = (ma_proc)_pa_stream_cork; + pContext->pulse.pa_stream_trigger = (ma_proc)_pa_stream_trigger; + pContext->pulse.pa_stream_begin_write = (ma_proc)_pa_stream_begin_write; + pContext->pulse.pa_stream_write = (ma_proc)_pa_stream_write; + pContext->pulse.pa_stream_peek = (ma_proc)_pa_stream_peek; + pContext->pulse.pa_stream_drop = (ma_proc)_pa_stream_drop; + pContext->pulse.pa_stream_writable_size = (ma_proc)_pa_stream_writable_size; + pContext->pulse.pa_stream_readable_size = (ma_proc)_pa_stream_readable_size; +#endif + + /* We need to make a copy of the application and server names so we can pass them to the pa_context of each device. */ + pContext->pulse.pApplicationName = ma_copy_string(pConfig->pulse.pApplicationName, &pContext->allocationCallbacks); + if (pContext->pulse.pApplicationName == NULL && pConfig->pulse.pApplicationName != NULL) { + return MA_OUT_OF_MEMORY; + } + + pContext->pulse.pServerName = ma_copy_string(pConfig->pulse.pServerName, &pContext->allocationCallbacks); + if (pContext->pulse.pServerName == NULL && pConfig->pulse.pServerName != NULL) { + ma_free(pContext->pulse.pApplicationName, &pContext->allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + + result = ma_init_pa_mainloop_and_pa_context__pulse(pContext, pConfig->pulse.pApplicationName, pConfig->pulse.pServerName, pConfig->pulse.tryAutoSpawn, &pContext->pulse.pMainLoop, &pContext->pulse.pPulseContext); + if (result != MA_SUCCESS) { + ma_free(pContext->pulse.pServerName, &pContext->allocationCallbacks); + ma_free(pContext->pulse.pApplicationName, &pContext->allocationCallbacks); + #ifndef MA_NO_RUNTIME_LINKING + ma_dlclose(ma_context_get_log(pContext), pContext->pulse.pulseSO); + #endif + return result; + } + + /* With pa_mainloop we run a synchronous backend, but we implement our own main loop. */ + pCallbacks->onContextInit = ma_context_init__pulse; + pCallbacks->onContextUninit = ma_context_uninit__pulse; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__pulse; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__pulse; + pCallbacks->onDeviceInit = ma_device_init__pulse; + pCallbacks->onDeviceUninit = ma_device_uninit__pulse; + pCallbacks->onDeviceStart = ma_device_start__pulse; + pCallbacks->onDeviceStop = ma_device_stop__pulse; + pCallbacks->onDeviceRead = NULL; /* Not used because we're implementing onDeviceDataLoop. */ + pCallbacks->onDeviceWrite = NULL; /* Not used because we're implementing onDeviceDataLoop. */ + pCallbacks->onDeviceDataLoop = ma_device_data_loop__pulse; + pCallbacks->onDeviceDataLoopWakeup = ma_device_data_loop_wakeup__pulse; + + return MA_SUCCESS; +} +#endif + + +/****************************************************************************** + +JACK Backend + +******************************************************************************/ +#ifdef MA_HAS_JACK + +/* It is assumed jack.h is available when compile-time linking is being used. */ +#ifdef MA_NO_RUNTIME_LINKING +#include + +typedef jack_nframes_t ma_jack_nframes_t; +typedef jack_options_t ma_jack_options_t; +typedef jack_status_t ma_jack_status_t; +typedef jack_client_t ma_jack_client_t; +typedef jack_port_t ma_jack_port_t; +typedef JackProcessCallback ma_JackProcessCallback; +typedef JackBufferSizeCallback ma_JackBufferSizeCallback; +typedef JackShutdownCallback ma_JackShutdownCallback; +#define MA_JACK_DEFAULT_AUDIO_TYPE JACK_DEFAULT_AUDIO_TYPE +#define ma_JackNoStartServer JackNoStartServer +#define ma_JackPortIsInput JackPortIsInput +#define ma_JackPortIsOutput JackPortIsOutput +#define ma_JackPortIsPhysical JackPortIsPhysical +#else +typedef ma_uint32 ma_jack_nframes_t; +typedef int ma_jack_options_t; +typedef int ma_jack_status_t; +typedef struct ma_jack_client_t ma_jack_client_t; +typedef struct ma_jack_port_t ma_jack_port_t; +typedef int (* ma_JackProcessCallback) (ma_jack_nframes_t nframes, void* arg); +typedef int (* ma_JackBufferSizeCallback)(ma_jack_nframes_t nframes, void* arg); +typedef void (* ma_JackShutdownCallback) (void* arg); +#define MA_JACK_DEFAULT_AUDIO_TYPE "32 bit float mono audio" +#define ma_JackNoStartServer 1 +#define ma_JackPortIsInput 1 +#define ma_JackPortIsOutput 2 +#define ma_JackPortIsPhysical 4 +#endif + +typedef ma_jack_client_t* (* ma_jack_client_open_proc) (const char* client_name, ma_jack_options_t options, ma_jack_status_t* status, ...); +typedef int (* ma_jack_client_close_proc) (ma_jack_client_t* client); +typedef int (* ma_jack_client_name_size_proc) (void); +typedef int (* ma_jack_set_process_callback_proc) (ma_jack_client_t* client, ma_JackProcessCallback process_callback, void* arg); +typedef int (* ma_jack_set_buffer_size_callback_proc)(ma_jack_client_t* client, ma_JackBufferSizeCallback bufsize_callback, void* arg); +typedef void (* ma_jack_on_shutdown_proc) (ma_jack_client_t* client, ma_JackShutdownCallback function, void* arg); +typedef ma_jack_nframes_t (* ma_jack_get_sample_rate_proc) (ma_jack_client_t* client); +typedef ma_jack_nframes_t (* ma_jack_get_buffer_size_proc) (ma_jack_client_t* client); +typedef const char** (* ma_jack_get_ports_proc) (ma_jack_client_t* client, const char* port_name_pattern, const char* type_name_pattern, unsigned long flags); +typedef int (* ma_jack_activate_proc) (ma_jack_client_t* client); +typedef int (* ma_jack_deactivate_proc) (ma_jack_client_t* client); +typedef int (* ma_jack_connect_proc) (ma_jack_client_t* client, const char* source_port, const char* destination_port); +typedef ma_jack_port_t* (* ma_jack_port_register_proc) (ma_jack_client_t* client, const char* port_name, const char* port_type, unsigned long flags, unsigned long buffer_size); +typedef const char* (* ma_jack_port_name_proc) (const ma_jack_port_t* port); +typedef void* (* ma_jack_port_get_buffer_proc) (ma_jack_port_t* port, ma_jack_nframes_t nframes); +typedef void (* ma_jack_free_proc) (void* ptr); + +static ma_result ma_context_open_client__jack(ma_context* pContext, ma_jack_client_t** ppClient) +{ + size_t maxClientNameSize; + char clientName[256]; + ma_jack_status_t status; + ma_jack_client_t* pClient; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppClient != NULL); + + if (ppClient) { + *ppClient = NULL; + } + + maxClientNameSize = ((ma_jack_client_name_size_proc)pContext->jack.jack_client_name_size)(); /* Includes null terminator. */ + ma_strncpy_s(clientName, ma_min(sizeof(clientName), maxClientNameSize), (pContext->jack.pClientName != NULL) ? pContext->jack.pClientName : "miniaudio", (size_t)-1); + + pClient = ((ma_jack_client_open_proc)pContext->jack.jack_client_open)(clientName, (pContext->jack.tryStartServer) ? 0 : ma_JackNoStartServer, &status, NULL); + if (pClient == NULL) { + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + + if (ppClient) { + *ppClient = pClient; + } + + return MA_SUCCESS; +} + + +static ma_result ma_context_enumerate_devices__jack(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_bool32 cbResult = MA_TRUE; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + /* Playback. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + deviceInfo.isDefault = MA_TRUE; /* JACK only uses default devices. */ + cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + + /* Capture. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + deviceInfo.isDefault = MA_TRUE; /* JACK only uses default devices. */ + cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + + (void)cbResult; /* For silencing a static analysis warning. */ + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info__jack(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + ma_jack_client_t* pClient; + ma_result result; + const char** ppPorts; + + MA_ASSERT(pContext != NULL); + + if (pDeviceID != NULL && pDeviceID->jack != 0) { + return MA_NO_DEVICE; /* Don't know the device. */ + } + + /* Name / Description */ + if (deviceType == ma_device_type_playback) { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } + + /* Jack only uses default devices. */ + pDeviceInfo->isDefault = MA_TRUE; + + /* Jack only supports f32 and has a specific channel count and sample rate. */ + pDeviceInfo->nativeDataFormats[0].format = ma_format_f32; + + /* The channel count and sample rate can only be determined by opening the device. */ + result = ma_context_open_client__jack(pContext, &pClient); + if (result != MA_SUCCESS) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[JACK] Failed to open client."); + return result; + } + + pDeviceInfo->nativeDataFormats[0].sampleRate = ((ma_jack_get_sample_rate_proc)pContext->jack.jack_get_sample_rate)((ma_jack_client_t*)pClient); + pDeviceInfo->nativeDataFormats[0].channels = 0; + + ppPorts = ((ma_jack_get_ports_proc)pContext->jack.jack_get_ports)((ma_jack_client_t*)pClient, NULL, MA_JACK_DEFAULT_AUDIO_TYPE, ma_JackPortIsPhysical | ((deviceType == ma_device_type_playback) ? ma_JackPortIsInput : ma_JackPortIsOutput)); + if (ppPorts == NULL) { + ((ma_jack_client_close_proc)pContext->jack.jack_client_close)((ma_jack_client_t*)pClient); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[JACK] Failed to query physical ports."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + + while (ppPorts[pDeviceInfo->nativeDataFormats[0].channels] != NULL) { + pDeviceInfo->nativeDataFormats[0].channels += 1; + } + + pDeviceInfo->nativeDataFormats[0].flags = 0; + pDeviceInfo->nativeDataFormatCount = 1; + + ((ma_jack_free_proc)pContext->jack.jack_free)((void*)ppPorts); + ((ma_jack_client_close_proc)pContext->jack.jack_client_close)((ma_jack_client_t*)pClient); + + (void)pContext; + return MA_SUCCESS; +} + + +static ma_result ma_device_uninit__jack(ma_device* pDevice) +{ + ma_context* pContext; + + MA_ASSERT(pDevice != NULL); + + pContext = pDevice->pContext; + MA_ASSERT(pContext != NULL); + + if (pDevice->jack.pClient != NULL) { + ((ma_jack_client_close_proc)pContext->jack.jack_client_close)((ma_jack_client_t*)pDevice->jack.pClient); + } + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ma_free(pDevice->jack.pIntermediaryBufferCapture, &pDevice->pContext->allocationCallbacks); + ma_free(pDevice->jack.ppPortsCapture, &pDevice->pContext->allocationCallbacks); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_free(pDevice->jack.pIntermediaryBufferPlayback, &pDevice->pContext->allocationCallbacks); + ma_free(pDevice->jack.ppPortsPlayback, &pDevice->pContext->allocationCallbacks); + } + + return MA_SUCCESS; +} + +static void ma_device__jack_shutdown_callback(void* pUserData) +{ + /* JACK died. Stop the device. */ + ma_device* pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + ma_device_stop(pDevice); +} + +static int ma_device__jack_buffer_size_callback(ma_jack_nframes_t frameCount, void* pUserData) +{ + ma_device* pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + size_t newBufferSize = frameCount * (pDevice->capture.internalChannels * ma_get_bytes_per_sample(pDevice->capture.internalFormat)); + float* pNewBuffer = (float*)ma_calloc(newBufferSize, &pDevice->pContext->allocationCallbacks); + if (pNewBuffer == NULL) { + return MA_OUT_OF_MEMORY; + } + + ma_free(pDevice->jack.pIntermediaryBufferCapture, &pDevice->pContext->allocationCallbacks); + + pDevice->jack.pIntermediaryBufferCapture = pNewBuffer; + pDevice->playback.internalPeriodSizeInFrames = frameCount; + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + size_t newBufferSize = frameCount * (pDevice->playback.internalChannels * ma_get_bytes_per_sample(pDevice->playback.internalFormat)); + float* pNewBuffer = (float*)ma_calloc(newBufferSize, &pDevice->pContext->allocationCallbacks); + if (pNewBuffer == NULL) { + return MA_OUT_OF_MEMORY; + } + + ma_free(pDevice->jack.pIntermediaryBufferPlayback, &pDevice->pContext->allocationCallbacks); + + pDevice->jack.pIntermediaryBufferPlayback = pNewBuffer; + pDevice->playback.internalPeriodSizeInFrames = frameCount; + } + + return 0; +} + +static int ma_device__jack_process_callback(ma_jack_nframes_t frameCount, void* pUserData) +{ + ma_device* pDevice; + ma_context* pContext; + ma_uint32 iChannel; + + pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + pContext = pDevice->pContext; + MA_ASSERT(pContext != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + /* Channels need to be interleaved. */ + for (iChannel = 0; iChannel < pDevice->capture.internalChannels; ++iChannel) { + const float* pSrc = (const float*)((ma_jack_port_get_buffer_proc)pContext->jack.jack_port_get_buffer)((ma_jack_port_t*)pDevice->jack.ppPortsCapture[iChannel], frameCount); + if (pSrc != NULL) { + float* pDst = pDevice->jack.pIntermediaryBufferCapture + iChannel; + ma_jack_nframes_t iFrame; + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + *pDst = *pSrc; + + pDst += pDevice->capture.internalChannels; + pSrc += 1; + } + } + } + + ma_device_handle_backend_data_callback(pDevice, NULL, pDevice->jack.pIntermediaryBufferCapture, frameCount); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_device_handle_backend_data_callback(pDevice, pDevice->jack.pIntermediaryBufferPlayback, NULL, frameCount); + + /* Channels need to be deinterleaved. */ + for (iChannel = 0; iChannel < pDevice->playback.internalChannels; ++iChannel) { + float* pDst = (float*)((ma_jack_port_get_buffer_proc)pContext->jack.jack_port_get_buffer)((ma_jack_port_t*)pDevice->jack.ppPortsPlayback[iChannel], frameCount); + if (pDst != NULL) { + const float* pSrc = pDevice->jack.pIntermediaryBufferPlayback + iChannel; + ma_jack_nframes_t iFrame; + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + *pDst = *pSrc; + + pDst += 1; + pSrc += pDevice->playback.internalChannels; + } + } + } + } + + return 0; +} + +static ma_result ma_device_init__jack(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + ma_result result; + ma_uint32 periodSizeInFrames; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDevice != NULL); + + if (pConfig->deviceType == ma_device_type_loopback) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Loopback mode not supported."); + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + /* Only supporting default devices with JACK. */ + if (((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pDescriptorPlayback->pDeviceID != NULL && pDescriptorPlayback->pDeviceID->jack != 0) || + ((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) && pDescriptorCapture->pDeviceID != NULL && pDescriptorCapture->pDeviceID->jack != 0)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Only default devices are supported."); + return MA_NO_DEVICE; + } + + /* No exclusive mode with the JACK backend. */ + if (((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pDescriptorPlayback->shareMode == ma_share_mode_exclusive) || + ((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) && pDescriptorCapture->shareMode == ma_share_mode_exclusive)) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Exclusive mode not supported."); + return MA_SHARE_MODE_NOT_SUPPORTED; + } + + /* Open the client. */ + result = ma_context_open_client__jack(pDevice->pContext, (ma_jack_client_t**)&pDevice->jack.pClient); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to open client."); + return result; + } + + /* Callbacks. */ + if (((ma_jack_set_process_callback_proc)pDevice->pContext->jack.jack_set_process_callback)((ma_jack_client_t*)pDevice->jack.pClient, ma_device__jack_process_callback, pDevice) != 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to set process callback."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + if (((ma_jack_set_buffer_size_callback_proc)pDevice->pContext->jack.jack_set_buffer_size_callback)((ma_jack_client_t*)pDevice->jack.pClient, ma_device__jack_buffer_size_callback, pDevice) != 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to set buffer size callback."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + + ((ma_jack_on_shutdown_proc)pDevice->pContext->jack.jack_on_shutdown)((ma_jack_client_t*)pDevice->jack.pClient, ma_device__jack_shutdown_callback, pDevice); + + + /* The buffer size in frames can change. */ + periodSizeInFrames = ((ma_jack_get_buffer_size_proc)pDevice->pContext->jack.jack_get_buffer_size)((ma_jack_client_t*)pDevice->jack.pClient); + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ma_uint32 iPort; + const char** ppPorts; + + pDescriptorCapture->format = ma_format_f32; + pDescriptorCapture->channels = 0; + pDescriptorCapture->sampleRate = ((ma_jack_get_sample_rate_proc)pDevice->pContext->jack.jack_get_sample_rate)((ma_jack_client_t*)pDevice->jack.pClient); + ma_channel_map_init_standard(ma_standard_channel_map_alsa, pDescriptorCapture->channelMap, ma_countof(pDescriptorCapture->channelMap), pDescriptorCapture->channels); + + ppPorts = ((ma_jack_get_ports_proc)pDevice->pContext->jack.jack_get_ports)((ma_jack_client_t*)pDevice->jack.pClient, NULL, MA_JACK_DEFAULT_AUDIO_TYPE, ma_JackPortIsPhysical | ma_JackPortIsOutput); + if (ppPorts == NULL) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to query physical ports."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + + /* Need to count the number of ports first so we can allocate some memory. */ + while (ppPorts[pDescriptorCapture->channels] != NULL) { + pDescriptorCapture->channels += 1; + } + + pDevice->jack.ppPortsCapture = (ma_ptr*)ma_malloc(sizeof(*pDevice->jack.ppPortsCapture) * pDescriptorCapture->channels, &pDevice->pContext->allocationCallbacks); + if (pDevice->jack.ppPortsCapture == NULL) { + return MA_OUT_OF_MEMORY; + } + + for (iPort = 0; iPort < pDescriptorCapture->channels; iPort += 1) { + char name[64]; + ma_strcpy_s(name, sizeof(name), "capture"); + ma_itoa_s((int)iPort, name+7, sizeof(name)-7, 10); /* 7 = length of "capture" */ + + pDevice->jack.ppPortsCapture[iPort] = ((ma_jack_port_register_proc)pDevice->pContext->jack.jack_port_register)((ma_jack_client_t*)pDevice->jack.pClient, name, MA_JACK_DEFAULT_AUDIO_TYPE, ma_JackPortIsInput, 0); + if (pDevice->jack.ppPortsCapture[iPort] == NULL) { + ((ma_jack_free_proc)pDevice->pContext->jack.jack_free)((void*)ppPorts); + ma_device_uninit__jack(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to register ports."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + } + + ((ma_jack_free_proc)pDevice->pContext->jack.jack_free)((void*)ppPorts); + + pDescriptorCapture->periodSizeInFrames = periodSizeInFrames; + pDescriptorCapture->periodCount = 1; /* There's no notion of a period in JACK. Just set to 1. */ + + pDevice->jack.pIntermediaryBufferCapture = (float*)ma_calloc(pDescriptorCapture->periodSizeInFrames * ma_get_bytes_per_frame(pDescriptorCapture->format, pDescriptorCapture->channels), &pDevice->pContext->allocationCallbacks); + if (pDevice->jack.pIntermediaryBufferCapture == NULL) { + ma_device_uninit__jack(pDevice); + return MA_OUT_OF_MEMORY; + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_uint32 iPort; + const char** ppPorts; + + pDescriptorPlayback->format = ma_format_f32; + pDescriptorPlayback->channels = 0; + pDescriptorPlayback->sampleRate = ((ma_jack_get_sample_rate_proc)pDevice->pContext->jack.jack_get_sample_rate)((ma_jack_client_t*)pDevice->jack.pClient); + ma_channel_map_init_standard(ma_standard_channel_map_alsa, pDescriptorPlayback->channelMap, ma_countof(pDescriptorPlayback->channelMap), pDescriptorPlayback->channels); + + ppPorts = ((ma_jack_get_ports_proc)pDevice->pContext->jack.jack_get_ports)((ma_jack_client_t*)pDevice->jack.pClient, NULL, MA_JACK_DEFAULT_AUDIO_TYPE, ma_JackPortIsPhysical | ma_JackPortIsInput); + if (ppPorts == NULL) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to query physical ports."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + + /* Need to count the number of ports first so we can allocate some memory. */ + while (ppPorts[pDescriptorPlayback->channels] != NULL) { + pDescriptorPlayback->channels += 1; + } + + pDevice->jack.ppPortsPlayback = (ma_ptr*)ma_malloc(sizeof(*pDevice->jack.ppPortsPlayback) * pDescriptorPlayback->channels, &pDevice->pContext->allocationCallbacks); + if (pDevice->jack.ppPortsPlayback == NULL) { + ma_free(pDevice->jack.ppPortsCapture, &pDevice->pContext->allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + + for (iPort = 0; iPort < pDescriptorPlayback->channels; iPort += 1) { + char name[64]; + ma_strcpy_s(name, sizeof(name), "playback"); + ma_itoa_s((int)iPort, name+8, sizeof(name)-8, 10); /* 8 = length of "playback" */ + + pDevice->jack.ppPortsPlayback[iPort] = ((ma_jack_port_register_proc)pDevice->pContext->jack.jack_port_register)((ma_jack_client_t*)pDevice->jack.pClient, name, MA_JACK_DEFAULT_AUDIO_TYPE, ma_JackPortIsOutput, 0); + if (pDevice->jack.ppPortsPlayback[iPort] == NULL) { + ((ma_jack_free_proc)pDevice->pContext->jack.jack_free)((void*)ppPorts); + ma_device_uninit__jack(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to register ports."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + } + + ((ma_jack_free_proc)pDevice->pContext->jack.jack_free)((void*)ppPorts); + + pDescriptorPlayback->periodSizeInFrames = periodSizeInFrames; + pDescriptorPlayback->periodCount = 1; /* There's no notion of a period in JACK. Just set to 1. */ + + pDevice->jack.pIntermediaryBufferPlayback = (float*)ma_calloc(pDescriptorPlayback->periodSizeInFrames * ma_get_bytes_per_frame(pDescriptorPlayback->format, pDescriptorPlayback->channels), &pDevice->pContext->allocationCallbacks); + if (pDevice->jack.pIntermediaryBufferPlayback == NULL) { + ma_device_uninit__jack(pDevice); + return MA_OUT_OF_MEMORY; + } + } + + return MA_SUCCESS; +} + + +static ma_result ma_device_start__jack(ma_device* pDevice) +{ + ma_context* pContext = pDevice->pContext; + int resultJACK; + size_t i; + + resultJACK = ((ma_jack_activate_proc)pContext->jack.jack_activate)((ma_jack_client_t*)pDevice->jack.pClient); + if (resultJACK != 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to activate the JACK client."); + return MA_FAILED_TO_START_BACKEND_DEVICE; + } + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + const char** ppServerPorts = ((ma_jack_get_ports_proc)pContext->jack.jack_get_ports)((ma_jack_client_t*)pDevice->jack.pClient, NULL, MA_JACK_DEFAULT_AUDIO_TYPE, ma_JackPortIsPhysical | ma_JackPortIsOutput); + if (ppServerPorts == NULL) { + ((ma_jack_deactivate_proc)pContext->jack.jack_deactivate)((ma_jack_client_t*)pDevice->jack.pClient); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to retrieve physical ports."); + return MA_ERROR; + } + + for (i = 0; ppServerPorts[i] != NULL; ++i) { + const char* pServerPort = ppServerPorts[i]; + const char* pClientPort = ((ma_jack_port_name_proc)pContext->jack.jack_port_name)((ma_jack_port_t*)pDevice->jack.ppPortsCapture[i]); + + resultJACK = ((ma_jack_connect_proc)pContext->jack.jack_connect)((ma_jack_client_t*)pDevice->jack.pClient, pServerPort, pClientPort); + if (resultJACK != 0) { + ((ma_jack_free_proc)pContext->jack.jack_free)((void*)ppServerPorts); + ((ma_jack_deactivate_proc)pContext->jack.jack_deactivate)((ma_jack_client_t*)pDevice->jack.pClient); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to connect ports."); + return MA_ERROR; + } + } + + ((ma_jack_free_proc)pContext->jack.jack_free)((void*)ppServerPorts); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + const char** ppServerPorts = ((ma_jack_get_ports_proc)pContext->jack.jack_get_ports)((ma_jack_client_t*)pDevice->jack.pClient, NULL, MA_JACK_DEFAULT_AUDIO_TYPE, ma_JackPortIsPhysical | ma_JackPortIsInput); + if (ppServerPorts == NULL) { + ((ma_jack_deactivate_proc)pContext->jack.jack_deactivate)((ma_jack_client_t*)pDevice->jack.pClient); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to retrieve physical ports."); + return MA_ERROR; + } + + for (i = 0; ppServerPorts[i] != NULL; ++i) { + const char* pServerPort = ppServerPorts[i]; + const char* pClientPort = ((ma_jack_port_name_proc)pContext->jack.jack_port_name)((ma_jack_port_t*)pDevice->jack.ppPortsPlayback[i]); + + resultJACK = ((ma_jack_connect_proc)pContext->jack.jack_connect)((ma_jack_client_t*)pDevice->jack.pClient, pClientPort, pServerPort); + if (resultJACK != 0) { + ((ma_jack_free_proc)pContext->jack.jack_free)((void*)ppServerPorts); + ((ma_jack_deactivate_proc)pContext->jack.jack_deactivate)((ma_jack_client_t*)pDevice->jack.pClient); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] Failed to connect ports."); + return MA_ERROR; + } + } + + ((ma_jack_free_proc)pContext->jack.jack_free)((void*)ppServerPorts); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__jack(ma_device* pDevice) +{ + ma_context* pContext = pDevice->pContext; + + if (((ma_jack_deactivate_proc)pContext->jack.jack_deactivate)((ma_jack_client_t*)pDevice->jack.pClient) != 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[JACK] An error occurred when deactivating the JACK client."); + return MA_ERROR; + } + + ma_device__on_notification_stopped(pDevice); + + return MA_SUCCESS; +} + + +static ma_result ma_context_uninit__jack(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_jack); + + ma_free(pContext->jack.pClientName, &pContext->allocationCallbacks); + pContext->jack.pClientName = NULL; + +#ifndef MA_NO_RUNTIME_LINKING + ma_dlclose(ma_context_get_log(pContext), pContext->jack.jackSO); +#endif + + return MA_SUCCESS; +} + +static ma_result ma_context_init__jack(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ +#ifndef MA_NO_RUNTIME_LINKING + const char* libjackNames[] = { +#if defined(MA_WIN32) + "libjack.dll", + "libjack64.dll" +#endif +#if defined(MA_UNIX) + "libjack.so", + "libjack.so.0" +#endif + }; + size_t i; + + for (i = 0; i < ma_countof(libjackNames); ++i) { + pContext->jack.jackSO = ma_dlopen(ma_context_get_log(pContext), libjackNames[i]); + if (pContext->jack.jackSO != NULL) { + break; + } + } + + if (pContext->jack.jackSO == NULL) { + return MA_NO_BACKEND; + } + + pContext->jack.jack_client_open = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_client_open"); + pContext->jack.jack_client_close = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_client_close"); + pContext->jack.jack_client_name_size = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_client_name_size"); + pContext->jack.jack_set_process_callback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_set_process_callback"); + pContext->jack.jack_set_buffer_size_callback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_set_buffer_size_callback"); + pContext->jack.jack_on_shutdown = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_on_shutdown"); + pContext->jack.jack_get_sample_rate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_get_sample_rate"); + pContext->jack.jack_get_buffer_size = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_get_buffer_size"); + pContext->jack.jack_get_ports = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_get_ports"); + pContext->jack.jack_activate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_activate"); + pContext->jack.jack_deactivate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_deactivate"); + pContext->jack.jack_connect = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_connect"); + pContext->jack.jack_port_register = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_port_register"); + pContext->jack.jack_port_name = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_port_name"); + pContext->jack.jack_port_get_buffer = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_port_get_buffer"); + pContext->jack.jack_free = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->jack.jackSO, "jack_free"); +#else + /* + This strange assignment system is here just to ensure type safety of miniaudio's function pointer + types. If anything differs slightly the compiler should throw a warning. + */ + ma_jack_client_open_proc _jack_client_open = jack_client_open; + ma_jack_client_close_proc _jack_client_close = jack_client_close; + ma_jack_client_name_size_proc _jack_client_name_size = jack_client_name_size; + ma_jack_set_process_callback_proc _jack_set_process_callback = jack_set_process_callback; + ma_jack_set_buffer_size_callback_proc _jack_set_buffer_size_callback = jack_set_buffer_size_callback; + ma_jack_on_shutdown_proc _jack_on_shutdown = jack_on_shutdown; + ma_jack_get_sample_rate_proc _jack_get_sample_rate = jack_get_sample_rate; + ma_jack_get_buffer_size_proc _jack_get_buffer_size = jack_get_buffer_size; + ma_jack_get_ports_proc _jack_get_ports = jack_get_ports; + ma_jack_activate_proc _jack_activate = jack_activate; + ma_jack_deactivate_proc _jack_deactivate = jack_deactivate; + ma_jack_connect_proc _jack_connect = jack_connect; + ma_jack_port_register_proc _jack_port_register = jack_port_register; + ma_jack_port_name_proc _jack_port_name = jack_port_name; + ma_jack_port_get_buffer_proc _jack_port_get_buffer = jack_port_get_buffer; + ma_jack_free_proc _jack_free = jack_free; + + pContext->jack.jack_client_open = (ma_proc)_jack_client_open; + pContext->jack.jack_client_close = (ma_proc)_jack_client_close; + pContext->jack.jack_client_name_size = (ma_proc)_jack_client_name_size; + pContext->jack.jack_set_process_callback = (ma_proc)_jack_set_process_callback; + pContext->jack.jack_set_buffer_size_callback = (ma_proc)_jack_set_buffer_size_callback; + pContext->jack.jack_on_shutdown = (ma_proc)_jack_on_shutdown; + pContext->jack.jack_get_sample_rate = (ma_proc)_jack_get_sample_rate; + pContext->jack.jack_get_buffer_size = (ma_proc)_jack_get_buffer_size; + pContext->jack.jack_get_ports = (ma_proc)_jack_get_ports; + pContext->jack.jack_activate = (ma_proc)_jack_activate; + pContext->jack.jack_deactivate = (ma_proc)_jack_deactivate; + pContext->jack.jack_connect = (ma_proc)_jack_connect; + pContext->jack.jack_port_register = (ma_proc)_jack_port_register; + pContext->jack.jack_port_name = (ma_proc)_jack_port_name; + pContext->jack.jack_port_get_buffer = (ma_proc)_jack_port_get_buffer; + pContext->jack.jack_free = (ma_proc)_jack_free; +#endif + + if (pConfig->jack.pClientName != NULL) { + pContext->jack.pClientName = ma_copy_string(pConfig->jack.pClientName, &pContext->allocationCallbacks); + } + pContext->jack.tryStartServer = pConfig->jack.tryStartServer; + + /* + Getting here means the JACK library is installed, but it doesn't necessarily mean it's usable. We need to quickly test this by connecting + a temporary client. + */ + { + ma_jack_client_t* pDummyClient; + ma_result result = ma_context_open_client__jack(pContext, &pDummyClient); + if (result != MA_SUCCESS) { + ma_free(pContext->jack.pClientName, &pContext->allocationCallbacks); + #ifndef MA_NO_RUNTIME_LINKING + ma_dlclose(ma_context_get_log(pContext), pContext->jack.jackSO); + #endif + return MA_NO_BACKEND; + } + + ((ma_jack_client_close_proc)pContext->jack.jack_client_close)((ma_jack_client_t*)pDummyClient); + } + + + pCallbacks->onContextInit = ma_context_init__jack; + pCallbacks->onContextUninit = ma_context_uninit__jack; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__jack; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__jack; + pCallbacks->onDeviceInit = ma_device_init__jack; + pCallbacks->onDeviceUninit = ma_device_uninit__jack; + pCallbacks->onDeviceStart = ma_device_start__jack; + pCallbacks->onDeviceStop = ma_device_stop__jack; + pCallbacks->onDeviceRead = NULL; /* Not used because JACK is asynchronous. */ + pCallbacks->onDeviceWrite = NULL; /* Not used because JACK is asynchronous. */ + pCallbacks->onDeviceDataLoop = NULL; /* Not used because JACK is asynchronous. */ + + return MA_SUCCESS; +} +#endif /* MA_HAS_JACK */ + + + +/****************************************************************************** + +Core Audio Backend + +References +========== +- Technical Note TN2091: Device input using the HAL Output Audio Unit + https://developer.apple.com/library/archive/technotes/tn2091/_index.html + +******************************************************************************/ +#ifdef MA_HAS_COREAUDIO +#include + +#if defined(TARGET_OS_IPHONE) && TARGET_OS_IPHONE == 1 + #define MA_APPLE_MOBILE + #if defined(TARGET_OS_TV) && TARGET_OS_TV == 1 + #define MA_APPLE_TV + #endif + #if defined(TARGET_OS_WATCH) && TARGET_OS_WATCH == 1 + #define MA_APPLE_WATCH + #endif + #if __has_feature(objc_arc) + #define MA_BRIDGE_TRANSFER __bridge_transfer + #define MA_BRIDGE_RETAINED __bridge_retained + #else + #define MA_BRIDGE_TRANSFER + #define MA_BRIDGE_RETAINED + #endif +#else + #define MA_APPLE_DESKTOP +#endif + +#if defined(MA_APPLE_DESKTOP) +#include +#else +#include +#endif + +#include + +/* CoreFoundation */ +typedef Boolean (* ma_CFStringGetCString_proc)(CFStringRef theString, char* buffer, CFIndex bufferSize, CFStringEncoding encoding); +typedef void (* ma_CFRelease_proc)(CFTypeRef cf); + +/* CoreAudio */ +#if defined(MA_APPLE_DESKTOP) +typedef OSStatus (* ma_AudioObjectGetPropertyData_proc)(AudioObjectID inObjectID, const AudioObjectPropertyAddress* inAddress, UInt32 inQualifierDataSize, const void* inQualifierData, UInt32* ioDataSize, void* outData); +typedef OSStatus (* ma_AudioObjectGetPropertyDataSize_proc)(AudioObjectID inObjectID, const AudioObjectPropertyAddress* inAddress, UInt32 inQualifierDataSize, const void* inQualifierData, UInt32* outDataSize); +typedef OSStatus (* ma_AudioObjectSetPropertyData_proc)(AudioObjectID inObjectID, const AudioObjectPropertyAddress* inAddress, UInt32 inQualifierDataSize, const void* inQualifierData, UInt32 inDataSize, const void* inData); +typedef OSStatus (* ma_AudioObjectAddPropertyListener_proc)(AudioObjectID inObjectID, const AudioObjectPropertyAddress* inAddress, AudioObjectPropertyListenerProc inListener, void* inClientData); +typedef OSStatus (* ma_AudioObjectRemovePropertyListener_proc)(AudioObjectID inObjectID, const AudioObjectPropertyAddress* inAddress, AudioObjectPropertyListenerProc inListener, void* inClientData); +#endif + +/* AudioToolbox */ +typedef AudioComponent (* ma_AudioComponentFindNext_proc)(AudioComponent inComponent, const AudioComponentDescription* inDesc); +typedef OSStatus (* ma_AudioComponentInstanceDispose_proc)(AudioComponentInstance inInstance); +typedef OSStatus (* ma_AudioComponentInstanceNew_proc)(AudioComponent inComponent, AudioComponentInstance* outInstance); +typedef OSStatus (* ma_AudioOutputUnitStart_proc)(AudioUnit inUnit); +typedef OSStatus (* ma_AudioOutputUnitStop_proc)(AudioUnit inUnit); +typedef OSStatus (* ma_AudioUnitAddPropertyListener_proc)(AudioUnit inUnit, AudioUnitPropertyID inID, AudioUnitPropertyListenerProc inProc, void* inProcUserData); +typedef OSStatus (* ma_AudioUnitGetPropertyInfo_proc)(AudioUnit inUnit, AudioUnitPropertyID inID, AudioUnitScope inScope, AudioUnitElement inElement, UInt32* outDataSize, Boolean* outWriteable); +typedef OSStatus (* ma_AudioUnitGetProperty_proc)(AudioUnit inUnit, AudioUnitPropertyID inID, AudioUnitScope inScope, AudioUnitElement inElement, void* outData, UInt32* ioDataSize); +typedef OSStatus (* ma_AudioUnitSetProperty_proc)(AudioUnit inUnit, AudioUnitPropertyID inID, AudioUnitScope inScope, AudioUnitElement inElement, const void* inData, UInt32 inDataSize); +typedef OSStatus (* ma_AudioUnitInitialize_proc)(AudioUnit inUnit); +typedef OSStatus (* ma_AudioUnitRender_proc)(AudioUnit inUnit, AudioUnitRenderActionFlags* ioActionFlags, const AudioTimeStamp* inTimeStamp, UInt32 inOutputBusNumber, UInt32 inNumberFrames, AudioBufferList* ioData); + + +#define MA_COREAUDIO_OUTPUT_BUS 0 +#define MA_COREAUDIO_INPUT_BUS 1 + +#if defined(MA_APPLE_DESKTOP) +static ma_result ma_device_reinit_internal__coreaudio(ma_device* pDevice, ma_device_type deviceType, ma_bool32 disposePreviousAudioUnit); +#endif + +/* +Core Audio + +So far, Core Audio has been the worst backend to work with due to being both unintuitive and having almost no documentation +apart from comments in the headers (which admittedly are quite good). For my own purposes, and for anybody out there whose +needing to figure out how this darn thing works, I'm going to outline a few things here. + +Since miniaudio is a fairly low-level API, one of the things it needs is control over specific devices, and it needs to be +able to identify whether or not it can be used as playback and/or capture. The AudioObject API is the only one I've seen +that supports this level of detail. There was some public domain sample code I stumbled across that used the AudioComponent +and AudioUnit APIs, but I couldn't see anything that gave low-level control over device selection and capabilities (the +distinction between playback and capture in particular). Therefore, miniaudio is using the AudioObject API. + +Most (all?) functions in the AudioObject API take a AudioObjectID as its input. This is the device identifier. When +retrieving global information, such as the device list, you use kAudioObjectSystemObject. When retrieving device-specific +data, you pass in the ID for that device. In order to retrieve device-specific IDs you need to enumerate over each of the +devices. This is done using the AudioObjectGetPropertyDataSize() and AudioObjectGetPropertyData() APIs which seem to be +the central APIs for retrieving information about the system and specific devices. + +To use the AudioObjectGetPropertyData() API you need to use the notion of a property address. A property address is a +structure with three variables and is used to identify which property you are getting or setting. The first is the "selector" +which is basically the specific property that you're wanting to retrieve or set. The second is the "scope", which is +typically set to kAudioObjectPropertyScopeGlobal, kAudioObjectPropertyScopeInput for input-specific properties and +kAudioObjectPropertyScopeOutput for output-specific properties. The last is the "element" which is always set to +kAudioObjectPropertyElementMain in miniaudio's case. I don't know of any cases where this would be set to anything different. + +Back to the earlier issue of device retrieval, you first use the AudioObjectGetPropertyDataSize() API to retrieve the size +of the raw data which is just a list of AudioDeviceID's. You use the kAudioObjectSystemObject AudioObjectID, and a property +address with the kAudioHardwarePropertyDevices selector and the kAudioObjectPropertyScopeGlobal scope. Once you have the +size, allocate a block of memory of that size and then call AudioObjectGetPropertyData(). The data is just a list of +AudioDeviceID's so just do "dataSize/sizeof(AudioDeviceID)" to know the device count. +*/ + +#if defined(MA_APPLE_MOBILE) +static void ma_device__on_notification_interruption_began(ma_device* pDevice) +{ + ma_device__on_notification(ma_device_notification_init(pDevice, ma_device_notification_type_interruption_began)); +} + +static void ma_device__on_notification_interruption_ended(ma_device* pDevice) +{ + ma_device__on_notification(ma_device_notification_init(pDevice, ma_device_notification_type_interruption_ended)); +} +#endif + +static ma_result ma_result_from_OSStatus(OSStatus status) +{ + switch (status) + { + case noErr: return MA_SUCCESS; + #if defined(MA_APPLE_DESKTOP) + case kAudioHardwareNotRunningError: return MA_DEVICE_NOT_STARTED; + case kAudioHardwareUnspecifiedError: return MA_ERROR; + case kAudioHardwareUnknownPropertyError: return MA_INVALID_ARGS; + case kAudioHardwareBadPropertySizeError: return MA_INVALID_OPERATION; + case kAudioHardwareIllegalOperationError: return MA_INVALID_OPERATION; + case kAudioHardwareBadObjectError: return MA_INVALID_ARGS; + case kAudioHardwareBadDeviceError: return MA_INVALID_ARGS; + case kAudioHardwareBadStreamError: return MA_INVALID_ARGS; + case kAudioHardwareUnsupportedOperationError: return MA_INVALID_OPERATION; + case kAudioDeviceUnsupportedFormatError: return MA_FORMAT_NOT_SUPPORTED; + case kAudioDevicePermissionsError: return MA_ACCESS_DENIED; + #endif + default: return MA_ERROR; + } +} + +#if 0 +static ma_channel ma_channel_from_AudioChannelBitmap(AudioChannelBitmap bit) +{ + switch (bit) + { + case kAudioChannelBit_Left: return MA_CHANNEL_LEFT; + case kAudioChannelBit_Right: return MA_CHANNEL_RIGHT; + case kAudioChannelBit_Center: return MA_CHANNEL_FRONT_CENTER; + case kAudioChannelBit_LFEScreen: return MA_CHANNEL_LFE; + case kAudioChannelBit_LeftSurround: return MA_CHANNEL_BACK_LEFT; + case kAudioChannelBit_RightSurround: return MA_CHANNEL_BACK_RIGHT; + case kAudioChannelBit_LeftCenter: return MA_CHANNEL_FRONT_LEFT_CENTER; + case kAudioChannelBit_RightCenter: return MA_CHANNEL_FRONT_RIGHT_CENTER; + case kAudioChannelBit_CenterSurround: return MA_CHANNEL_BACK_CENTER; + case kAudioChannelBit_LeftSurroundDirect: return MA_CHANNEL_SIDE_LEFT; + case kAudioChannelBit_RightSurroundDirect: return MA_CHANNEL_SIDE_RIGHT; + case kAudioChannelBit_TopCenterSurround: return MA_CHANNEL_TOP_CENTER; + case kAudioChannelBit_VerticalHeightLeft: return MA_CHANNEL_TOP_FRONT_LEFT; + case kAudioChannelBit_VerticalHeightCenter: return MA_CHANNEL_TOP_FRONT_CENTER; + case kAudioChannelBit_VerticalHeightRight: return MA_CHANNEL_TOP_FRONT_RIGHT; + case kAudioChannelBit_TopBackLeft: return MA_CHANNEL_TOP_BACK_LEFT; + case kAudioChannelBit_TopBackCenter: return MA_CHANNEL_TOP_BACK_CENTER; + case kAudioChannelBit_TopBackRight: return MA_CHANNEL_TOP_BACK_RIGHT; + default: return MA_CHANNEL_NONE; + } +} +#endif + +static ma_result ma_format_from_AudioStreamBasicDescription(const AudioStreamBasicDescription* pDescription, ma_format* pFormatOut) +{ + MA_ASSERT(pDescription != NULL); + MA_ASSERT(pFormatOut != NULL); + + *pFormatOut = ma_format_unknown; /* Safety. */ + + /* There's a few things miniaudio doesn't support. */ + if (pDescription->mFormatID != kAudioFormatLinearPCM) { + return MA_FORMAT_NOT_SUPPORTED; + } + + /* We don't support any non-packed formats that are aligned high. */ + if ((pDescription->mFormatFlags & kLinearPCMFormatFlagIsAlignedHigh) != 0) { + return MA_FORMAT_NOT_SUPPORTED; + } + + /* Only supporting native-endian. */ + if ((ma_is_little_endian() && (pDescription->mFormatFlags & kAudioFormatFlagIsBigEndian) != 0) || (ma_is_big_endian() && (pDescription->mFormatFlags & kAudioFormatFlagIsBigEndian) == 0)) { + return MA_FORMAT_NOT_SUPPORTED; + } + + /* We are not currently supporting non-interleaved formats (this will be added in a future version of miniaudio). */ + /*if ((pDescription->mFormatFlags & kAudioFormatFlagIsNonInterleaved) != 0) { + return MA_FORMAT_NOT_SUPPORTED; + }*/ + + if ((pDescription->mFormatFlags & kLinearPCMFormatFlagIsFloat) != 0) { + if (pDescription->mBitsPerChannel == 32) { + *pFormatOut = ma_format_f32; + return MA_SUCCESS; + } + } else { + if ((pDescription->mFormatFlags & kLinearPCMFormatFlagIsSignedInteger) != 0) { + if (pDescription->mBitsPerChannel == 16) { + *pFormatOut = ma_format_s16; + return MA_SUCCESS; + } else if (pDescription->mBitsPerChannel == 24) { + if (pDescription->mBytesPerFrame == (pDescription->mBitsPerChannel/8 * pDescription->mChannelsPerFrame)) { + *pFormatOut = ma_format_s24; + return MA_SUCCESS; + } else { + if (pDescription->mBytesPerFrame/pDescription->mChannelsPerFrame == sizeof(ma_int32)) { + /* TODO: Implement ma_format_s24_32. */ + /**pFormatOut = ma_format_s24_32;*/ + /*return MA_SUCCESS;*/ + return MA_FORMAT_NOT_SUPPORTED; + } + } + } else if (pDescription->mBitsPerChannel == 32) { + *pFormatOut = ma_format_s32; + return MA_SUCCESS; + } + } else { + if (pDescription->mBitsPerChannel == 8) { + *pFormatOut = ma_format_u8; + return MA_SUCCESS; + } + } + } + + /* Getting here means the format is not supported. */ + return MA_FORMAT_NOT_SUPPORTED; +} + +#if defined(MA_APPLE_DESKTOP) +static ma_channel ma_channel_from_AudioChannelLabel(AudioChannelLabel label) +{ + switch (label) + { + case kAudioChannelLabel_Unknown: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Unused: return MA_CHANNEL_NONE; + case kAudioChannelLabel_UseCoordinates: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Left: return MA_CHANNEL_LEFT; + case kAudioChannelLabel_Right: return MA_CHANNEL_RIGHT; + case kAudioChannelLabel_Center: return MA_CHANNEL_FRONT_CENTER; + case kAudioChannelLabel_LFEScreen: return MA_CHANNEL_LFE; + case kAudioChannelLabel_LeftSurround: return MA_CHANNEL_BACK_LEFT; + case kAudioChannelLabel_RightSurround: return MA_CHANNEL_BACK_RIGHT; + case kAudioChannelLabel_LeftCenter: return MA_CHANNEL_FRONT_LEFT_CENTER; + case kAudioChannelLabel_RightCenter: return MA_CHANNEL_FRONT_RIGHT_CENTER; + case kAudioChannelLabel_CenterSurround: return MA_CHANNEL_BACK_CENTER; + case kAudioChannelLabel_LeftSurroundDirect: return MA_CHANNEL_SIDE_LEFT; + case kAudioChannelLabel_RightSurroundDirect: return MA_CHANNEL_SIDE_RIGHT; + case kAudioChannelLabel_TopCenterSurround: return MA_CHANNEL_TOP_CENTER; + case kAudioChannelLabel_VerticalHeightLeft: return MA_CHANNEL_TOP_FRONT_LEFT; + case kAudioChannelLabel_VerticalHeightCenter: return MA_CHANNEL_TOP_FRONT_CENTER; + case kAudioChannelLabel_VerticalHeightRight: return MA_CHANNEL_TOP_FRONT_RIGHT; + case kAudioChannelLabel_TopBackLeft: return MA_CHANNEL_TOP_BACK_LEFT; + case kAudioChannelLabel_TopBackCenter: return MA_CHANNEL_TOP_BACK_CENTER; + case kAudioChannelLabel_TopBackRight: return MA_CHANNEL_TOP_BACK_RIGHT; + case kAudioChannelLabel_RearSurroundLeft: return MA_CHANNEL_BACK_LEFT; + case kAudioChannelLabel_RearSurroundRight: return MA_CHANNEL_BACK_RIGHT; + case kAudioChannelLabel_LeftWide: return MA_CHANNEL_SIDE_LEFT; + case kAudioChannelLabel_RightWide: return MA_CHANNEL_SIDE_RIGHT; + case kAudioChannelLabel_LFE2: return MA_CHANNEL_LFE; + case kAudioChannelLabel_LeftTotal: return MA_CHANNEL_LEFT; + case kAudioChannelLabel_RightTotal: return MA_CHANNEL_RIGHT; + case kAudioChannelLabel_HearingImpaired: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Narration: return MA_CHANNEL_MONO; + case kAudioChannelLabel_Mono: return MA_CHANNEL_MONO; + case kAudioChannelLabel_DialogCentricMix: return MA_CHANNEL_MONO; + case kAudioChannelLabel_CenterSurroundDirect: return MA_CHANNEL_BACK_CENTER; + case kAudioChannelLabel_Haptic: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Ambisonic_W: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Ambisonic_X: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Ambisonic_Y: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Ambisonic_Z: return MA_CHANNEL_NONE; + case kAudioChannelLabel_MS_Mid: return MA_CHANNEL_LEFT; + case kAudioChannelLabel_MS_Side: return MA_CHANNEL_RIGHT; + case kAudioChannelLabel_XY_X: return MA_CHANNEL_LEFT; + case kAudioChannelLabel_XY_Y: return MA_CHANNEL_RIGHT; + case kAudioChannelLabel_HeadphonesLeft: return MA_CHANNEL_LEFT; + case kAudioChannelLabel_HeadphonesRight: return MA_CHANNEL_RIGHT; + case kAudioChannelLabel_ClickTrack: return MA_CHANNEL_NONE; + case kAudioChannelLabel_ForeignLanguage: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Discrete: return MA_CHANNEL_NONE; + case kAudioChannelLabel_Discrete_0: return MA_CHANNEL_AUX_0; + case kAudioChannelLabel_Discrete_1: return MA_CHANNEL_AUX_1; + case kAudioChannelLabel_Discrete_2: return MA_CHANNEL_AUX_2; + case kAudioChannelLabel_Discrete_3: return MA_CHANNEL_AUX_3; + case kAudioChannelLabel_Discrete_4: return MA_CHANNEL_AUX_4; + case kAudioChannelLabel_Discrete_5: return MA_CHANNEL_AUX_5; + case kAudioChannelLabel_Discrete_6: return MA_CHANNEL_AUX_6; + case kAudioChannelLabel_Discrete_7: return MA_CHANNEL_AUX_7; + case kAudioChannelLabel_Discrete_8: return MA_CHANNEL_AUX_8; + case kAudioChannelLabel_Discrete_9: return MA_CHANNEL_AUX_9; + case kAudioChannelLabel_Discrete_10: return MA_CHANNEL_AUX_10; + case kAudioChannelLabel_Discrete_11: return MA_CHANNEL_AUX_11; + case kAudioChannelLabel_Discrete_12: return MA_CHANNEL_AUX_12; + case kAudioChannelLabel_Discrete_13: return MA_CHANNEL_AUX_13; + case kAudioChannelLabel_Discrete_14: return MA_CHANNEL_AUX_14; + case kAudioChannelLabel_Discrete_15: return MA_CHANNEL_AUX_15; + case kAudioChannelLabel_Discrete_65535: return MA_CHANNEL_NONE; + + #if 0 /* Introduced in a later version of macOS. */ + case kAudioChannelLabel_HOA_ACN: return MA_CHANNEL_NONE; + case kAudioChannelLabel_HOA_ACN_0: return MA_CHANNEL_AUX_0; + case kAudioChannelLabel_HOA_ACN_1: return MA_CHANNEL_AUX_1; + case kAudioChannelLabel_HOA_ACN_2: return MA_CHANNEL_AUX_2; + case kAudioChannelLabel_HOA_ACN_3: return MA_CHANNEL_AUX_3; + case kAudioChannelLabel_HOA_ACN_4: return MA_CHANNEL_AUX_4; + case kAudioChannelLabel_HOA_ACN_5: return MA_CHANNEL_AUX_5; + case kAudioChannelLabel_HOA_ACN_6: return MA_CHANNEL_AUX_6; + case kAudioChannelLabel_HOA_ACN_7: return MA_CHANNEL_AUX_7; + case kAudioChannelLabel_HOA_ACN_8: return MA_CHANNEL_AUX_8; + case kAudioChannelLabel_HOA_ACN_9: return MA_CHANNEL_AUX_9; + case kAudioChannelLabel_HOA_ACN_10: return MA_CHANNEL_AUX_10; + case kAudioChannelLabel_HOA_ACN_11: return MA_CHANNEL_AUX_11; + case kAudioChannelLabel_HOA_ACN_12: return MA_CHANNEL_AUX_12; + case kAudioChannelLabel_HOA_ACN_13: return MA_CHANNEL_AUX_13; + case kAudioChannelLabel_HOA_ACN_14: return MA_CHANNEL_AUX_14; + case kAudioChannelLabel_HOA_ACN_15: return MA_CHANNEL_AUX_15; + case kAudioChannelLabel_HOA_ACN_65024: return MA_CHANNEL_NONE; + #endif + + default: return MA_CHANNEL_NONE; + } +} + +static ma_result ma_get_channel_map_from_AudioChannelLayout(AudioChannelLayout* pChannelLayout, ma_channel* pChannelMap, size_t channelMapCap) +{ + MA_ASSERT(pChannelLayout != NULL); + + if (pChannelLayout->mChannelLayoutTag == kAudioChannelLayoutTag_UseChannelDescriptions) { + UInt32 iChannel; + for (iChannel = 0; iChannel < pChannelLayout->mNumberChannelDescriptions && iChannel < channelMapCap; ++iChannel) { + pChannelMap[iChannel] = ma_channel_from_AudioChannelLabel(pChannelLayout->mChannelDescriptions[iChannel].mChannelLabel); + } + } else +#if 0 + if (pChannelLayout->mChannelLayoutTag == kAudioChannelLayoutTag_UseChannelBitmap) { + /* This is the same kind of system that's used by Windows audio APIs. */ + UInt32 iChannel = 0; + UInt32 iBit; + AudioChannelBitmap bitmap = pChannelLayout->mChannelBitmap; + for (iBit = 0; iBit < 32 && iChannel < channelMapCap; ++iBit) { + AudioChannelBitmap bit = bitmap & (1 << iBit); + if (bit != 0) { + pChannelMap[iChannel++] = ma_channel_from_AudioChannelBit(bit); + } + } + } else +#endif + { + /* + Need to use the tag to determine the channel map. For now I'm just assuming a default channel map, but later on this should + be updated to determine the mapping based on the tag. + */ + UInt32 channelCount; + + /* Our channel map retrieval APIs below take 32-bit integers, so we'll want to clamp the channel map capacity. */ + if (channelMapCap > 0xFFFFFFFF) { + channelMapCap = 0xFFFFFFFF; + } + + channelCount = ma_min(AudioChannelLayoutTag_GetNumberOfChannels(pChannelLayout->mChannelLayoutTag), (UInt32)channelMapCap); + + switch (pChannelLayout->mChannelLayoutTag) + { + case kAudioChannelLayoutTag_Mono: + case kAudioChannelLayoutTag_Stereo: + case kAudioChannelLayoutTag_StereoHeadphones: + case kAudioChannelLayoutTag_MatrixStereo: + case kAudioChannelLayoutTag_MidSide: + case kAudioChannelLayoutTag_XY: + case kAudioChannelLayoutTag_Binaural: + case kAudioChannelLayoutTag_Ambisonic_B_Format: + { + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, channelCount); + } break; + + case kAudioChannelLayoutTag_Octagonal: + { + pChannelMap[7] = MA_CHANNEL_SIDE_RIGHT; + pChannelMap[6] = MA_CHANNEL_SIDE_LEFT; + } MA_FALLTHROUGH; /* Intentional fallthrough. */ + case kAudioChannelLayoutTag_Hexagonal: + { + pChannelMap[5] = MA_CHANNEL_BACK_CENTER; + } MA_FALLTHROUGH; /* Intentional fallthrough. */ + case kAudioChannelLayoutTag_Pentagonal: + { + pChannelMap[4] = MA_CHANNEL_FRONT_CENTER; + } MA_FALLTHROUGH; /* Intentional fallthrough. */ + case kAudioChannelLayoutTag_Quadraphonic: + { + pChannelMap[3] = MA_CHANNEL_BACK_RIGHT; + pChannelMap[2] = MA_CHANNEL_BACK_LEFT; + pChannelMap[1] = MA_CHANNEL_RIGHT; + pChannelMap[0] = MA_CHANNEL_LEFT; + } break; + + /* TODO: Add support for more tags here. */ + + default: + { + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, channelCount); + } break; + } + } + + return MA_SUCCESS; +} + +#if (defined(MAC_OS_VERSION_12_0) && MAC_OS_X_VERSION_MAX_ALLOWED >= MAC_OS_VERSION_12_0) || \ + (defined(__IPHONE_15_0) && __IPHONE_OS_VERSION_MAX_ALLOWED >= __IPHONE_15_0) +#define AUDIO_OBJECT_PROPERTY_ELEMENT kAudioObjectPropertyElementMain +#else +/* kAudioObjectPropertyElementMaster is deprecated. */ +#define AUDIO_OBJECT_PROPERTY_ELEMENT kAudioObjectPropertyElementMaster +#endif + +/* kAudioDevicePropertyScope* were renamed to kAudioObjectPropertyScope* in 10.8. */ +#if !defined(MAC_OS_X_VERSION_10_8) || (MAC_OS_X_VERSION_MIN_REQUIRED < MAC_OS_X_VERSION_10_8) +#define kAudioObjectPropertyScopeInput kAudioDevicePropertyScopeInput +#define kAudioObjectPropertyScopeOutput kAudioDevicePropertyScopeOutput +#endif + +static ma_result ma_get_device_object_ids__coreaudio(ma_context* pContext, UInt32* pDeviceCount, AudioObjectID** ppDeviceObjectIDs) /* NOTE: Free the returned buffer with ma_free(). */ +{ + AudioObjectPropertyAddress propAddressDevices; + UInt32 deviceObjectsDataSize; + OSStatus status; + AudioObjectID* pDeviceObjectIDs; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pDeviceCount != NULL); + MA_ASSERT(ppDeviceObjectIDs != NULL); + + /* Safety. */ + *pDeviceCount = 0; + *ppDeviceObjectIDs = NULL; + + propAddressDevices.mSelector = kAudioHardwarePropertyDevices; + propAddressDevices.mScope = kAudioObjectPropertyScopeGlobal; + propAddressDevices.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + status = ((ma_AudioObjectGetPropertyDataSize_proc)pContext->coreaudio.AudioObjectGetPropertyDataSize)(kAudioObjectSystemObject, &propAddressDevices, 0, NULL, &deviceObjectsDataSize); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + pDeviceObjectIDs = (AudioObjectID*)ma_malloc(deviceObjectsDataSize, &pContext->allocationCallbacks); + if (pDeviceObjectIDs == NULL) { + return MA_OUT_OF_MEMORY; + } + + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(kAudioObjectSystemObject, &propAddressDevices, 0, NULL, &deviceObjectsDataSize, pDeviceObjectIDs); + if (status != noErr) { + ma_free(pDeviceObjectIDs, &pContext->allocationCallbacks); + return ma_result_from_OSStatus(status); + } + + *pDeviceCount = deviceObjectsDataSize / sizeof(AudioObjectID); + *ppDeviceObjectIDs = pDeviceObjectIDs; + + return MA_SUCCESS; +} + +static ma_result ma_get_AudioObject_uid_as_CFStringRef(ma_context* pContext, AudioObjectID objectID, CFStringRef* pUID) +{ + AudioObjectPropertyAddress propAddress; + UInt32 dataSize; + OSStatus status; + + MA_ASSERT(pContext != NULL); + + propAddress.mSelector = kAudioDevicePropertyDeviceUID; + propAddress.mScope = kAudioObjectPropertyScopeGlobal; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + dataSize = sizeof(*pUID); + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(objectID, &propAddress, 0, NULL, &dataSize, pUID); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + return MA_SUCCESS; +} + +static ma_result ma_get_AudioObject_uid(ma_context* pContext, AudioObjectID objectID, size_t bufferSize, char* bufferOut) +{ + CFStringRef uid; + ma_result result; + + MA_ASSERT(pContext != NULL); + + result = ma_get_AudioObject_uid_as_CFStringRef(pContext, objectID, &uid); + if (result != MA_SUCCESS) { + return result; + } + + if (!((ma_CFStringGetCString_proc)pContext->coreaudio.CFStringGetCString)(uid, bufferOut, bufferSize, kCFStringEncodingUTF8)) { + return MA_ERROR; + } + + ((ma_CFRelease_proc)pContext->coreaudio.CFRelease)(uid); + return MA_SUCCESS; +} + +static ma_result ma_get_AudioObject_name(ma_context* pContext, AudioObjectID objectID, size_t bufferSize, char* bufferOut) +{ + AudioObjectPropertyAddress propAddress; + CFStringRef deviceName = NULL; + UInt32 dataSize; + OSStatus status; + + MA_ASSERT(pContext != NULL); + + propAddress.mSelector = kAudioDevicePropertyDeviceNameCFString; + propAddress.mScope = kAudioObjectPropertyScopeGlobal; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + dataSize = sizeof(deviceName); + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(objectID, &propAddress, 0, NULL, &dataSize, &deviceName); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + if (!((ma_CFStringGetCString_proc)pContext->coreaudio.CFStringGetCString)(deviceName, bufferOut, bufferSize, kCFStringEncodingUTF8)) { + return MA_ERROR; + } + + ((ma_CFRelease_proc)pContext->coreaudio.CFRelease)(deviceName); + return MA_SUCCESS; +} + +static ma_bool32 ma_does_AudioObject_support_scope(ma_context* pContext, AudioObjectID deviceObjectID, AudioObjectPropertyScope scope) +{ + AudioObjectPropertyAddress propAddress; + UInt32 dataSize; + OSStatus status; + AudioBufferList* pBufferList; + ma_bool32 isSupported; + + MA_ASSERT(pContext != NULL); + + /* To know whether or not a device is an input device we need ot look at the stream configuration. If it has an output channel it's a playback device. */ + propAddress.mSelector = kAudioDevicePropertyStreamConfiguration; + propAddress.mScope = scope; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + status = ((ma_AudioObjectGetPropertyDataSize_proc)pContext->coreaudio.AudioObjectGetPropertyDataSize)(deviceObjectID, &propAddress, 0, NULL, &dataSize); + if (status != noErr) { + return MA_FALSE; + } + + pBufferList = (AudioBufferList*)ma_malloc(dataSize, &pContext->allocationCallbacks); + if (pBufferList == NULL) { + return MA_FALSE; /* Out of memory. */ + } + + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(deviceObjectID, &propAddress, 0, NULL, &dataSize, pBufferList); + if (status != noErr) { + ma_free(pBufferList, &pContext->allocationCallbacks); + return MA_FALSE; + } + + isSupported = MA_FALSE; + if (pBufferList->mNumberBuffers > 0) { + isSupported = MA_TRUE; + } + + ma_free(pBufferList, &pContext->allocationCallbacks); + return isSupported; +} + +static ma_bool32 ma_does_AudioObject_support_playback(ma_context* pContext, AudioObjectID deviceObjectID) +{ + return ma_does_AudioObject_support_scope(pContext, deviceObjectID, kAudioObjectPropertyScopeOutput); +} + +static ma_bool32 ma_does_AudioObject_support_capture(ma_context* pContext, AudioObjectID deviceObjectID) +{ + return ma_does_AudioObject_support_scope(pContext, deviceObjectID, kAudioObjectPropertyScopeInput); +} + + +static ma_result ma_get_AudioObject_stream_descriptions(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, UInt32* pDescriptionCount, AudioStreamRangedDescription** ppDescriptions) /* NOTE: Free the returned pointer with ma_free(). */ +{ + AudioObjectPropertyAddress propAddress; + UInt32 dataSize; + OSStatus status; + AudioStreamRangedDescription* pDescriptions; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pDescriptionCount != NULL); + MA_ASSERT(ppDescriptions != NULL); + + /* + TODO: Experiment with kAudioStreamPropertyAvailablePhysicalFormats instead of (or in addition to) kAudioStreamPropertyAvailableVirtualFormats. My + MacBook Pro uses s24/32 format, however, which miniaudio does not currently support. + */ + propAddress.mSelector = kAudioStreamPropertyAvailableVirtualFormats; /*kAudioStreamPropertyAvailablePhysicalFormats;*/ + propAddress.mScope = (deviceType == ma_device_type_playback) ? kAudioObjectPropertyScopeOutput : kAudioObjectPropertyScopeInput; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + status = ((ma_AudioObjectGetPropertyDataSize_proc)pContext->coreaudio.AudioObjectGetPropertyDataSize)(deviceObjectID, &propAddress, 0, NULL, &dataSize); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + pDescriptions = (AudioStreamRangedDescription*)ma_malloc(dataSize, &pContext->allocationCallbacks); + if (pDescriptions == NULL) { + return MA_OUT_OF_MEMORY; + } + + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(deviceObjectID, &propAddress, 0, NULL, &dataSize, pDescriptions); + if (status != noErr) { + ma_free(pDescriptions, &pContext->allocationCallbacks); + return ma_result_from_OSStatus(status); + } + + *pDescriptionCount = dataSize / sizeof(*pDescriptions); + *ppDescriptions = pDescriptions; + return MA_SUCCESS; +} + + +static ma_result ma_get_AudioObject_channel_layout(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, AudioChannelLayout** ppChannelLayout) /* NOTE: Free the returned pointer with ma_free(). */ +{ + AudioObjectPropertyAddress propAddress; + UInt32 dataSize; + OSStatus status; + AudioChannelLayout* pChannelLayout; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(ppChannelLayout != NULL); + + *ppChannelLayout = NULL; /* Safety. */ + + propAddress.mSelector = kAudioDevicePropertyPreferredChannelLayout; + propAddress.mScope = (deviceType == ma_device_type_playback) ? kAudioObjectPropertyScopeOutput : kAudioObjectPropertyScopeInput; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + status = ((ma_AudioObjectGetPropertyDataSize_proc)pContext->coreaudio.AudioObjectGetPropertyDataSize)(deviceObjectID, &propAddress, 0, NULL, &dataSize); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + pChannelLayout = (AudioChannelLayout*)ma_malloc(dataSize, &pContext->allocationCallbacks); + if (pChannelLayout == NULL) { + return MA_OUT_OF_MEMORY; + } + + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(deviceObjectID, &propAddress, 0, NULL, &dataSize, pChannelLayout); + if (status != noErr) { + ma_free(pChannelLayout, &pContext->allocationCallbacks); + return ma_result_from_OSStatus(status); + } + + *ppChannelLayout = pChannelLayout; + return MA_SUCCESS; +} + +static ma_result ma_get_AudioObject_channel_count(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, ma_uint32* pChannelCount) +{ + AudioChannelLayout* pChannelLayout; + ma_result result; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pChannelCount != NULL); + + *pChannelCount = 0; /* Safety. */ + + result = ma_get_AudioObject_channel_layout(pContext, deviceObjectID, deviceType, &pChannelLayout); + if (result != MA_SUCCESS) { + return result; + } + + if (pChannelLayout->mChannelLayoutTag == kAudioChannelLayoutTag_UseChannelDescriptions) { + *pChannelCount = pChannelLayout->mNumberChannelDescriptions; + } else if (pChannelLayout->mChannelLayoutTag == kAudioChannelLayoutTag_UseChannelBitmap) { + *pChannelCount = ma_count_set_bits(pChannelLayout->mChannelBitmap); + } else { + *pChannelCount = AudioChannelLayoutTag_GetNumberOfChannels(pChannelLayout->mChannelLayoutTag); + } + + ma_free(pChannelLayout, &pContext->allocationCallbacks); + return MA_SUCCESS; +} + +#if 0 +static ma_result ma_get_AudioObject_channel_map(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, ma_channel* pChannelMap, size_t channelMapCap) +{ + AudioChannelLayout* pChannelLayout; + ma_result result; + + MA_ASSERT(pContext != NULL); + + result = ma_get_AudioObject_channel_layout(pContext, deviceObjectID, deviceType, &pChannelLayout); + if (result != MA_SUCCESS) { + return result; /* Rather than always failing here, would it be more robust to simply assume a default? */ + } + + result = ma_get_channel_map_from_AudioChannelLayout(pChannelLayout, pChannelMap, channelMapCap); + if (result != MA_SUCCESS) { + ma_free(pChannelLayout, &pContext->allocationCallbacks); + return result; + } + + ma_free(pChannelLayout, &pContext->allocationCallbacks); + return result; +} +#endif + +static ma_result ma_get_AudioObject_sample_rates(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, UInt32* pSampleRateRangesCount, AudioValueRange** ppSampleRateRanges) /* NOTE: Free the returned pointer with ma_free(). */ +{ + AudioObjectPropertyAddress propAddress; + UInt32 dataSize; + OSStatus status; + AudioValueRange* pSampleRateRanges; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pSampleRateRangesCount != NULL); + MA_ASSERT(ppSampleRateRanges != NULL); + + /* Safety. */ + *pSampleRateRangesCount = 0; + *ppSampleRateRanges = NULL; + + propAddress.mSelector = kAudioDevicePropertyAvailableNominalSampleRates; + propAddress.mScope = (deviceType == ma_device_type_playback) ? kAudioObjectPropertyScopeOutput : kAudioObjectPropertyScopeInput; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + status = ((ma_AudioObjectGetPropertyDataSize_proc)pContext->coreaudio.AudioObjectGetPropertyDataSize)(deviceObjectID, &propAddress, 0, NULL, &dataSize); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + pSampleRateRanges = (AudioValueRange*)ma_malloc(dataSize, &pContext->allocationCallbacks); + if (pSampleRateRanges == NULL) { + return MA_OUT_OF_MEMORY; + } + + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(deviceObjectID, &propAddress, 0, NULL, &dataSize, pSampleRateRanges); + if (status != noErr) { + ma_free(pSampleRateRanges, &pContext->allocationCallbacks); + return ma_result_from_OSStatus(status); + } + + *pSampleRateRangesCount = dataSize / sizeof(*pSampleRateRanges); + *ppSampleRateRanges = pSampleRateRanges; + return MA_SUCCESS; +} + +#if 0 +static ma_result ma_get_AudioObject_get_closest_sample_rate(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, ma_uint32 sampleRateIn, ma_uint32* pSampleRateOut) +{ + UInt32 sampleRateRangeCount; + AudioValueRange* pSampleRateRanges; + ma_result result; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pSampleRateOut != NULL); + + *pSampleRateOut = 0; /* Safety. */ + + result = ma_get_AudioObject_sample_rates(pContext, deviceObjectID, deviceType, &sampleRateRangeCount, &pSampleRateRanges); + if (result != MA_SUCCESS) { + return result; + } + + if (sampleRateRangeCount == 0) { + ma_free(pSampleRateRanges, &pContext->allocationCallbacks); + return MA_ERROR; /* Should never hit this case should we? */ + } + + if (sampleRateIn == 0) { + /* Search in order of miniaudio's preferred priority. */ + UInt32 iMALSampleRate; + for (iMALSampleRate = 0; iMALSampleRate < ma_countof(g_maStandardSampleRatePriorities); ++iMALSampleRate) { + ma_uint32 malSampleRate = g_maStandardSampleRatePriorities[iMALSampleRate]; + UInt32 iCASampleRate; + for (iCASampleRate = 0; iCASampleRate < sampleRateRangeCount; ++iCASampleRate) { + AudioValueRange caSampleRate = pSampleRateRanges[iCASampleRate]; + if (caSampleRate.mMinimum <= malSampleRate && caSampleRate.mMaximum >= malSampleRate) { + *pSampleRateOut = malSampleRate; + ma_free(pSampleRateRanges, &pContext->allocationCallbacks); + return MA_SUCCESS; + } + } + } + + /* + If we get here it means none of miniaudio's standard sample rates matched any of the supported sample rates from the device. In this + case we just fall back to the first one reported by Core Audio. + */ + MA_ASSERT(sampleRateRangeCount > 0); + + *pSampleRateOut = pSampleRateRanges[0].mMinimum; + ma_free(pSampleRateRanges, &pContext->allocationCallbacks); + return MA_SUCCESS; + } else { + /* Find the closest match to this sample rate. */ + UInt32 currentAbsoluteDifference = INT32_MAX; + UInt32 iCurrentClosestRange = (UInt32)-1; + UInt32 iRange; + for (iRange = 0; iRange < sampleRateRangeCount; ++iRange) { + if (pSampleRateRanges[iRange].mMinimum <= sampleRateIn && pSampleRateRanges[iRange].mMaximum >= sampleRateIn) { + *pSampleRateOut = sampleRateIn; + ma_free(pSampleRateRanges, &pContext->allocationCallbacks); + return MA_SUCCESS; + } else { + UInt32 absoluteDifference; + if (pSampleRateRanges[iRange].mMinimum > sampleRateIn) { + absoluteDifference = pSampleRateRanges[iRange].mMinimum - sampleRateIn; + } else { + absoluteDifference = sampleRateIn - pSampleRateRanges[iRange].mMaximum; + } + + if (currentAbsoluteDifference > absoluteDifference) { + currentAbsoluteDifference = absoluteDifference; + iCurrentClosestRange = iRange; + } + } + } + + MA_ASSERT(iCurrentClosestRange != (UInt32)-1); + + *pSampleRateOut = pSampleRateRanges[iCurrentClosestRange].mMinimum; + ma_free(pSampleRateRanges, &pContext->allocationCallbacks); + return MA_SUCCESS; + } + + /* Should never get here, but it would mean we weren't able to find any suitable sample rates. */ + /*ma_free(pSampleRateRanges, &pContext->allocationCallbacks);*/ + /*return MA_ERROR;*/ +} +#endif + +static ma_result ma_get_AudioObject_closest_buffer_size_in_frames(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, ma_uint32 bufferSizeInFramesIn, ma_uint32* pBufferSizeInFramesOut) +{ + AudioObjectPropertyAddress propAddress; + AudioValueRange bufferSizeRange; + UInt32 dataSize; + OSStatus status; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pBufferSizeInFramesOut != NULL); + + *pBufferSizeInFramesOut = 0; /* Safety. */ + + propAddress.mSelector = kAudioDevicePropertyBufferFrameSizeRange; + propAddress.mScope = (deviceType == ma_device_type_playback) ? kAudioObjectPropertyScopeOutput : kAudioObjectPropertyScopeInput; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + dataSize = sizeof(bufferSizeRange); + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(deviceObjectID, &propAddress, 0, NULL, &dataSize, &bufferSizeRange); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + /* This is just a clamp. */ + if (bufferSizeInFramesIn < bufferSizeRange.mMinimum) { + *pBufferSizeInFramesOut = (ma_uint32)bufferSizeRange.mMinimum; + } else if (bufferSizeInFramesIn > bufferSizeRange.mMaximum) { + *pBufferSizeInFramesOut = (ma_uint32)bufferSizeRange.mMaximum; + } else { + *pBufferSizeInFramesOut = bufferSizeInFramesIn; + } + + return MA_SUCCESS; +} + +static ma_result ma_set_AudioObject_buffer_size_in_frames(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, ma_uint32* pPeriodSizeInOut) +{ + ma_result result; + ma_uint32 chosenBufferSizeInFrames; + AudioObjectPropertyAddress propAddress; + UInt32 dataSize; + OSStatus status; + + MA_ASSERT(pContext != NULL); + + result = ma_get_AudioObject_closest_buffer_size_in_frames(pContext, deviceObjectID, deviceType, *pPeriodSizeInOut, &chosenBufferSizeInFrames); + if (result != MA_SUCCESS) { + return result; + } + + /* Try setting the size of the buffer... If this fails we just use whatever is currently set. */ + propAddress.mSelector = kAudioDevicePropertyBufferFrameSize; + propAddress.mScope = (deviceType == ma_device_type_playback) ? kAudioObjectPropertyScopeOutput : kAudioObjectPropertyScopeInput; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + ((ma_AudioObjectSetPropertyData_proc)pContext->coreaudio.AudioObjectSetPropertyData)(deviceObjectID, &propAddress, 0, NULL, sizeof(chosenBufferSizeInFrames), &chosenBufferSizeInFrames); + + /* Get the actual size of the buffer. */ + dataSize = sizeof(*pPeriodSizeInOut); + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(deviceObjectID, &propAddress, 0, NULL, &dataSize, &chosenBufferSizeInFrames); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + *pPeriodSizeInOut = chosenBufferSizeInFrames; + return MA_SUCCESS; +} + +static ma_result ma_find_default_AudioObjectID(ma_context* pContext, ma_device_type deviceType, AudioObjectID* pDeviceObjectID) +{ + AudioObjectPropertyAddress propAddressDefaultDevice; + UInt32 defaultDeviceObjectIDSize = sizeof(AudioObjectID); + AudioObjectID defaultDeviceObjectID; + OSStatus status; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pDeviceObjectID != NULL); + + /* Safety. */ + *pDeviceObjectID = 0; + + propAddressDefaultDevice.mScope = kAudioObjectPropertyScopeGlobal; + propAddressDefaultDevice.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + if (deviceType == ma_device_type_playback) { + propAddressDefaultDevice.mSelector = kAudioHardwarePropertyDefaultOutputDevice; + } else { + propAddressDefaultDevice.mSelector = kAudioHardwarePropertyDefaultInputDevice; + } + + defaultDeviceObjectIDSize = sizeof(AudioObjectID); + status = ((ma_AudioObjectGetPropertyData_proc)pContext->coreaudio.AudioObjectGetPropertyData)(kAudioObjectSystemObject, &propAddressDefaultDevice, 0, NULL, &defaultDeviceObjectIDSize, &defaultDeviceObjectID); + if (status == noErr) { + *pDeviceObjectID = defaultDeviceObjectID; + return MA_SUCCESS; + } + + /* If we get here it means we couldn't find the device. */ + return MA_NO_DEVICE; +} + +static ma_result ma_find_AudioObjectID(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, AudioObjectID* pDeviceObjectID) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pDeviceObjectID != NULL); + + /* Safety. */ + *pDeviceObjectID = 0; + + if (pDeviceID == NULL) { + /* Default device. */ + return ma_find_default_AudioObjectID(pContext, deviceType, pDeviceObjectID); + } else { + /* Explicit device. */ + UInt32 deviceCount; + AudioObjectID* pDeviceObjectIDs; + ma_result result; + UInt32 iDevice; + + result = ma_get_device_object_ids__coreaudio(pContext, &deviceCount, &pDeviceObjectIDs); + if (result != MA_SUCCESS) { + return result; + } + + for (iDevice = 0; iDevice < deviceCount; ++iDevice) { + AudioObjectID deviceObjectID = pDeviceObjectIDs[iDevice]; + + char uid[256]; + if (ma_get_AudioObject_uid(pContext, deviceObjectID, sizeof(uid), uid) != MA_SUCCESS) { + continue; + } + + if (deviceType == ma_device_type_playback) { + if (ma_does_AudioObject_support_playback(pContext, deviceObjectID)) { + if (strcmp(uid, pDeviceID->coreaudio) == 0) { + *pDeviceObjectID = deviceObjectID; + ma_free(pDeviceObjectIDs, &pContext->allocationCallbacks); + return MA_SUCCESS; + } + } + } else { + if (ma_does_AudioObject_support_capture(pContext, deviceObjectID)) { + if (strcmp(uid, pDeviceID->coreaudio) == 0) { + *pDeviceObjectID = deviceObjectID; + ma_free(pDeviceObjectIDs, &pContext->allocationCallbacks); + return MA_SUCCESS; + } + } + } + } + + ma_free(pDeviceObjectIDs, &pContext->allocationCallbacks); + } + + /* If we get here it means we couldn't find the device. */ + return MA_NO_DEVICE; +} + + +static ma_result ma_find_best_format__coreaudio(ma_context* pContext, AudioObjectID deviceObjectID, ma_device_type deviceType, ma_format format, ma_uint32 channels, ma_uint32 sampleRate, const AudioStreamBasicDescription* pOrigFormat, AudioStreamBasicDescription* pFormat) +{ + UInt32 deviceFormatDescriptionCount; + AudioStreamRangedDescription* pDeviceFormatDescriptions; + ma_result result; + ma_uint32 desiredSampleRate; + ma_uint32 desiredChannelCount; + ma_format desiredFormat; + AudioStreamBasicDescription bestDeviceFormatSoFar; + ma_bool32 hasSupportedFormat; + UInt32 iFormat; + + result = ma_get_AudioObject_stream_descriptions(pContext, deviceObjectID, deviceType, &deviceFormatDescriptionCount, &pDeviceFormatDescriptions); + if (result != MA_SUCCESS) { + return result; + } + + desiredSampleRate = sampleRate; + if (desiredSampleRate == 0) { + desiredSampleRate = (ma_uint32)pOrigFormat->mSampleRate; + } + + desiredChannelCount = channels; + if (desiredChannelCount == 0) { + desiredChannelCount = pOrigFormat->mChannelsPerFrame; + } + + desiredFormat = format; + if (desiredFormat == ma_format_unknown) { + result = ma_format_from_AudioStreamBasicDescription(pOrigFormat, &desiredFormat); + if (result != MA_SUCCESS || desiredFormat == ma_format_unknown) { + desiredFormat = g_maFormatPriorities[0]; + } + } + + /* + If we get here it means we don't have an exact match to what the client is asking for. We'll need to find the closest one. The next + loop will check for formats that have the same sample rate to what we're asking for. If there is, we prefer that one in all cases. + */ + MA_ZERO_OBJECT(&bestDeviceFormatSoFar); + + hasSupportedFormat = MA_FALSE; + for (iFormat = 0; iFormat < deviceFormatDescriptionCount; ++iFormat) { + ma_format formatFromDescription; + ma_result formatResult = ma_format_from_AudioStreamBasicDescription(&pDeviceFormatDescriptions[iFormat].mFormat, &formatFromDescription); + if (formatResult == MA_SUCCESS && formatFromDescription != ma_format_unknown) { + hasSupportedFormat = MA_TRUE; + bestDeviceFormatSoFar = pDeviceFormatDescriptions[iFormat].mFormat; + break; + } + } + + if (!hasSupportedFormat) { + ma_free(pDeviceFormatDescriptions, &pContext->allocationCallbacks); + return MA_FORMAT_NOT_SUPPORTED; + } + + + for (iFormat = 0; iFormat < deviceFormatDescriptionCount; ++iFormat) { + AudioStreamBasicDescription thisDeviceFormat = pDeviceFormatDescriptions[iFormat].mFormat; + ma_format thisSampleFormat; + ma_result formatResult; + ma_format bestSampleFormatSoFar; + + /* If the format is not supported by miniaudio we need to skip this one entirely. */ + formatResult = ma_format_from_AudioStreamBasicDescription(&pDeviceFormatDescriptions[iFormat].mFormat, &thisSampleFormat); + if (formatResult != MA_SUCCESS || thisSampleFormat == ma_format_unknown) { + continue; /* The format is not supported by miniaudio. Skip. */ + } + + ma_format_from_AudioStreamBasicDescription(&bestDeviceFormatSoFar, &bestSampleFormatSoFar); + + /* Getting here means the format is supported by miniaudio which makes this format a candidate. */ + if (thisDeviceFormat.mSampleRate != desiredSampleRate) { + /* + The sample rate does not match, but this format could still be usable, although it's a very low priority. If the best format + so far has an equal sample rate we can just ignore this one. + */ + if (bestDeviceFormatSoFar.mSampleRate == desiredSampleRate) { + continue; /* The best sample rate so far has the same sample rate as what we requested which means it's still the best so far. Skip this format. */ + } else { + /* In this case, neither the best format so far nor this one have the same sample rate. Check the channel count next. */ + if (thisDeviceFormat.mChannelsPerFrame != desiredChannelCount) { + /* This format has a different sample rate _and_ a different channel count. */ + if (bestDeviceFormatSoFar.mChannelsPerFrame == desiredChannelCount) { + continue; /* No change to the best format. */ + } else { + /* + Both this format and the best so far have different sample rates and different channel counts. Whichever has the + best format is the new best. + */ + if (ma_get_format_priority_index(thisSampleFormat) < ma_get_format_priority_index(bestSampleFormatSoFar)) { + bestDeviceFormatSoFar = thisDeviceFormat; + continue; + } else { + continue; /* No change to the best format. */ + } + } + } else { + /* This format has a different sample rate but the desired channel count. */ + if (bestDeviceFormatSoFar.mChannelsPerFrame == desiredChannelCount) { + /* Both this format and the best so far have the desired channel count. Whichever has the best format is the new best. */ + if (ma_get_format_priority_index(thisSampleFormat) < ma_get_format_priority_index(bestSampleFormatSoFar)) { + bestDeviceFormatSoFar = thisDeviceFormat; + continue; + } else { + continue; /* No change to the best format for now. */ + } + } else { + /* This format has the desired channel count, but the best so far does not. We have a new best. */ + bestDeviceFormatSoFar = thisDeviceFormat; + continue; + } + } + } + } else { + /* + The sample rates match which makes this format a very high priority contender. If the best format so far has a different + sample rate it needs to be replaced with this one. + */ + if (bestDeviceFormatSoFar.mSampleRate != desiredSampleRate) { + bestDeviceFormatSoFar = thisDeviceFormat; + continue; + } else { + /* In this case both this format and the best format so far have the same sample rate. Check the channel count next. */ + if (thisDeviceFormat.mChannelsPerFrame == desiredChannelCount) { + /* + In this case this format has the same channel count as what the client is requesting. If the best format so far has + a different count, this one becomes the new best. + */ + if (bestDeviceFormatSoFar.mChannelsPerFrame != desiredChannelCount) { + bestDeviceFormatSoFar = thisDeviceFormat; + continue; + } else { + /* In this case both this format and the best so far have the ideal sample rate and channel count. Check the format. */ + if (thisSampleFormat == desiredFormat) { + bestDeviceFormatSoFar = thisDeviceFormat; + break; /* Found the exact match. */ + } else { + /* The formats are different. The new best format is the one with the highest priority format according to miniaudio. */ + if (ma_get_format_priority_index(thisSampleFormat) < ma_get_format_priority_index(bestSampleFormatSoFar)) { + bestDeviceFormatSoFar = thisDeviceFormat; + continue; + } else { + continue; /* No change to the best format for now. */ + } + } + } + } else { + /* + In this case the channel count is different to what the client has requested. If the best so far has the same channel + count as the requested count then it remains the best. + */ + if (bestDeviceFormatSoFar.mChannelsPerFrame == desiredChannelCount) { + continue; + } else { + /* + This is the case where both have the same sample rate (good) but different channel counts. Right now both have about + the same priority, but we need to compare the format now. + */ + if (thisSampleFormat == bestSampleFormatSoFar) { + if (ma_get_format_priority_index(thisSampleFormat) < ma_get_format_priority_index(bestSampleFormatSoFar)) { + bestDeviceFormatSoFar = thisDeviceFormat; + continue; + } else { + continue; /* No change to the best format for now. */ + } + } + } + } + } + } + } + + *pFormat = bestDeviceFormatSoFar; + + ma_free(pDeviceFormatDescriptions, &pContext->allocationCallbacks); + return MA_SUCCESS; +} + +static ma_result ma_get_AudioUnit_channel_map(ma_context* pContext, AudioUnit audioUnit, ma_device_type deviceType, ma_channel* pChannelMap, size_t channelMapCap) +{ + AudioUnitScope deviceScope; + AudioUnitElement deviceBus; + UInt32 channelLayoutSize; + OSStatus status; + AudioChannelLayout* pChannelLayout; + ma_result result; + + MA_ASSERT(pContext != NULL); + + if (deviceType == ma_device_type_playback) { + deviceScope = kAudioUnitScope_Input; + deviceBus = MA_COREAUDIO_OUTPUT_BUS; + } else { + deviceScope = kAudioUnitScope_Output; + deviceBus = MA_COREAUDIO_INPUT_BUS; + } + + status = ((ma_AudioUnitGetPropertyInfo_proc)pContext->coreaudio.AudioUnitGetPropertyInfo)(audioUnit, kAudioUnitProperty_AudioChannelLayout, deviceScope, deviceBus, &channelLayoutSize, NULL); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + pChannelLayout = (AudioChannelLayout*)ma_malloc(channelLayoutSize, &pContext->allocationCallbacks); + if (pChannelLayout == NULL) { + return MA_OUT_OF_MEMORY; + } + + status = ((ma_AudioUnitGetProperty_proc)pContext->coreaudio.AudioUnitGetProperty)(audioUnit, kAudioUnitProperty_AudioChannelLayout, deviceScope, deviceBus, pChannelLayout, &channelLayoutSize); + if (status != noErr) { + ma_free(pChannelLayout, &pContext->allocationCallbacks); + return ma_result_from_OSStatus(status); + } + + result = ma_get_channel_map_from_AudioChannelLayout(pChannelLayout, pChannelMap, channelMapCap); + if (result != MA_SUCCESS) { + ma_free(pChannelLayout, &pContext->allocationCallbacks); + return result; + } + + ma_free(pChannelLayout, &pContext->allocationCallbacks); + return MA_SUCCESS; +} +#endif /* MA_APPLE_DESKTOP */ + + +#if !defined(MA_APPLE_DESKTOP) +static void ma_AVAudioSessionPortDescription_to_device_info(AVAudioSessionPortDescription* pPortDesc, ma_device_info* pInfo) +{ + MA_ZERO_OBJECT(pInfo); + ma_strncpy_s(pInfo->name, sizeof(pInfo->name), [pPortDesc.portName UTF8String], (size_t)-1); + ma_strncpy_s(pInfo->id.coreaudio, sizeof(pInfo->id.coreaudio), [pPortDesc.UID UTF8String], (size_t)-1); +} +#endif + +static ma_result ma_context_enumerate_devices__coreaudio(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ +#if defined(MA_APPLE_DESKTOP) + UInt32 deviceCount; + AudioObjectID* pDeviceObjectIDs; + AudioObjectID defaultDeviceObjectIDPlayback; + AudioObjectID defaultDeviceObjectIDCapture; + ma_result result; + UInt32 iDevice; + + ma_find_default_AudioObjectID(pContext, ma_device_type_playback, &defaultDeviceObjectIDPlayback); /* OK if this fails. */ + ma_find_default_AudioObjectID(pContext, ma_device_type_capture, &defaultDeviceObjectIDCapture); /* OK if this fails. */ + + result = ma_get_device_object_ids__coreaudio(pContext, &deviceCount, &pDeviceObjectIDs); + if (result != MA_SUCCESS) { + return result; + } + + for (iDevice = 0; iDevice < deviceCount; ++iDevice) { + AudioObjectID deviceObjectID = pDeviceObjectIDs[iDevice]; + ma_device_info info; + + MA_ZERO_OBJECT(&info); + if (ma_get_AudioObject_uid(pContext, deviceObjectID, sizeof(info.id.coreaudio), info.id.coreaudio) != MA_SUCCESS) { + continue; + } + if (ma_get_AudioObject_name(pContext, deviceObjectID, sizeof(info.name), info.name) != MA_SUCCESS) { + continue; + } + + if (ma_does_AudioObject_support_playback(pContext, deviceObjectID)) { + if (deviceObjectID == defaultDeviceObjectIDPlayback) { + info.isDefault = MA_TRUE; + } + + if (!callback(pContext, ma_device_type_playback, &info, pUserData)) { + break; + } + } + if (ma_does_AudioObject_support_capture(pContext, deviceObjectID)) { + if (deviceObjectID == defaultDeviceObjectIDCapture) { + info.isDefault = MA_TRUE; + } + + if (!callback(pContext, ma_device_type_capture, &info, pUserData)) { + break; + } + } + } + + ma_free(pDeviceObjectIDs, &pContext->allocationCallbacks); +#else + ma_device_info info; + NSArray *pInputs = [[[AVAudioSession sharedInstance] currentRoute] inputs]; + NSArray *pOutputs = [[[AVAudioSession sharedInstance] currentRoute] outputs]; + + for (AVAudioSessionPortDescription* pPortDesc in pOutputs) { + ma_AVAudioSessionPortDescription_to_device_info(pPortDesc, &info); + if (!callback(pContext, ma_device_type_playback, &info, pUserData)) { + return MA_SUCCESS; + } + } + + for (AVAudioSessionPortDescription* pPortDesc in pInputs) { + ma_AVAudioSessionPortDescription_to_device_info(pPortDesc, &info); + if (!callback(pContext, ma_device_type_capture, &info, pUserData)) { + return MA_SUCCESS; + } + } +#endif + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info__coreaudio(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + ma_result result; + + MA_ASSERT(pContext != NULL); + +#if defined(MA_APPLE_DESKTOP) + /* Desktop */ + { + AudioObjectID deviceObjectID; + AudioObjectID defaultDeviceObjectID; + UInt32 streamDescriptionCount; + AudioStreamRangedDescription* pStreamDescriptions; + UInt32 iStreamDescription; + UInt32 sampleRateRangeCount; + AudioValueRange* pSampleRateRanges; + + ma_find_default_AudioObjectID(pContext, deviceType, &defaultDeviceObjectID); /* OK if this fails. */ + + result = ma_find_AudioObjectID(pContext, deviceType, pDeviceID, &deviceObjectID); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_get_AudioObject_uid(pContext, deviceObjectID, sizeof(pDeviceInfo->id.coreaudio), pDeviceInfo->id.coreaudio); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_get_AudioObject_name(pContext, deviceObjectID, sizeof(pDeviceInfo->name), pDeviceInfo->name); + if (result != MA_SUCCESS) { + return result; + } + + if (deviceObjectID == defaultDeviceObjectID) { + pDeviceInfo->isDefault = MA_TRUE; + } + + /* + There could be a large number of permutations here. Fortunately there is only a single channel count + being reported which reduces this quite a bit. For sample rates we're only reporting those that are + one of miniaudio's recognized "standard" rates. If there are still more formats than can fit into + our fixed sized array we'll just need to truncate them. This is unlikely and will probably only happen + if some driver performs software data conversion and therefore reports every possible format and + sample rate. + */ + pDeviceInfo->nativeDataFormatCount = 0; + + /* Formats. */ + { + ma_format uniqueFormats[ma_format_count]; + ma_uint32 uniqueFormatCount = 0; + ma_uint32 channels; + + /* Channels. */ + result = ma_get_AudioObject_channel_count(pContext, deviceObjectID, deviceType, &channels); + if (result != MA_SUCCESS) { + return result; + } + + /* Formats. */ + result = ma_get_AudioObject_stream_descriptions(pContext, deviceObjectID, deviceType, &streamDescriptionCount, &pStreamDescriptions); + if (result != MA_SUCCESS) { + return result; + } + + for (iStreamDescription = 0; iStreamDescription < streamDescriptionCount; ++iStreamDescription) { + ma_format format; + ma_bool32 hasFormatBeenHandled = MA_FALSE; + ma_uint32 iOutputFormat; + ma_uint32 iSampleRate; + + result = ma_format_from_AudioStreamBasicDescription(&pStreamDescriptions[iStreamDescription].mFormat, &format); + if (result != MA_SUCCESS) { + continue; + } + + MA_ASSERT(format != ma_format_unknown); + + /* Make sure the format isn't already in the output list. */ + for (iOutputFormat = 0; iOutputFormat < uniqueFormatCount; ++iOutputFormat) { + if (uniqueFormats[iOutputFormat] == format) { + hasFormatBeenHandled = MA_TRUE; + break; + } + } + + /* If we've already handled this format just skip it. */ + if (hasFormatBeenHandled) { + continue; + } + + uniqueFormats[uniqueFormatCount] = format; + uniqueFormatCount += 1; + + /* Sample Rates */ + result = ma_get_AudioObject_sample_rates(pContext, deviceObjectID, deviceType, &sampleRateRangeCount, &pSampleRateRanges); + if (result != MA_SUCCESS) { + return result; + } + + /* + Annoyingly Core Audio reports a sample rate range. We just get all the standard rates that are + between this range. + */ + for (iSampleRate = 0; iSampleRate < sampleRateRangeCount; ++iSampleRate) { + ma_uint32 iStandardSampleRate; + for (iStandardSampleRate = 0; iStandardSampleRate < ma_countof(g_maStandardSampleRatePriorities); iStandardSampleRate += 1) { + ma_uint32 standardSampleRate = g_maStandardSampleRatePriorities[iStandardSampleRate]; + if (standardSampleRate >= pSampleRateRanges[iSampleRate].mMinimum && standardSampleRate <= pSampleRateRanges[iSampleRate].mMaximum) { + /* We have a new data format. Add it to the list. */ + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].format = format; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].channels = channels; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].sampleRate = standardSampleRate; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].flags = 0; + pDeviceInfo->nativeDataFormatCount += 1; + + if (pDeviceInfo->nativeDataFormatCount >= ma_countof(pDeviceInfo->nativeDataFormats)) { + break; /* No more room for any more formats. */ + } + } + } + } + + ma_free(pSampleRateRanges, &pContext->allocationCallbacks); + + if (pDeviceInfo->nativeDataFormatCount >= ma_countof(pDeviceInfo->nativeDataFormats)) { + break; /* No more room for any more formats. */ + } + } + + ma_free(pStreamDescriptions, &pContext->allocationCallbacks); + } + } +#else + /* Mobile */ + { + AudioComponentDescription desc; + AudioComponent component; + AudioUnit audioUnit; + OSStatus status; + AudioUnitScope formatScope; + AudioUnitElement formatElement; + AudioStreamBasicDescription bestFormat; + UInt32 propSize; + + /* We want to ensure we use a consistent device name to device enumeration. */ + if (pDeviceID != NULL && pDeviceID->coreaudio[0] != '\0') { + ma_bool32 found = MA_FALSE; + if (deviceType == ma_device_type_playback) { + NSArray *pOutputs = [[[AVAudioSession sharedInstance] currentRoute] outputs]; + for (AVAudioSessionPortDescription* pPortDesc in pOutputs) { + if (strcmp(pDeviceID->coreaudio, [pPortDesc.UID UTF8String]) == 0) { + ma_AVAudioSessionPortDescription_to_device_info(pPortDesc, pDeviceInfo); + found = MA_TRUE; + break; + } + } + } else { + NSArray *pInputs = [[[AVAudioSession sharedInstance] currentRoute] inputs]; + for (AVAudioSessionPortDescription* pPortDesc in pInputs) { + if (strcmp(pDeviceID->coreaudio, [pPortDesc.UID UTF8String]) == 0) { + ma_AVAudioSessionPortDescription_to_device_info(pPortDesc, pDeviceInfo); + found = MA_TRUE; + break; + } + } + } + + if (!found) { + return MA_DOES_NOT_EXIST; + } + } else { + if (deviceType == ma_device_type_playback) { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } + } + + + /* + Retrieving device information is more annoying on mobile than desktop. For simplicity I'm locking this down to whatever format is + reported on a temporary I/O unit. The problem, however, is that this doesn't return a value for the sample rate which we need to + retrieve from the AVAudioSession shared instance. + */ + desc.componentType = kAudioUnitType_Output; + desc.componentSubType = kAudioUnitSubType_RemoteIO; + desc.componentManufacturer = kAudioUnitManufacturer_Apple; + desc.componentFlags = 0; + desc.componentFlagsMask = 0; + + component = ((ma_AudioComponentFindNext_proc)pContext->coreaudio.AudioComponentFindNext)(NULL, &desc); + if (component == NULL) { + return MA_FAILED_TO_INIT_BACKEND; + } + + status = ((ma_AudioComponentInstanceNew_proc)pContext->coreaudio.AudioComponentInstanceNew)(component, &audioUnit); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + formatScope = (deviceType == ma_device_type_playback) ? kAudioUnitScope_Input : kAudioUnitScope_Output; + formatElement = (deviceType == ma_device_type_playback) ? MA_COREAUDIO_OUTPUT_BUS : MA_COREAUDIO_INPUT_BUS; + + propSize = sizeof(bestFormat); + status = ((ma_AudioUnitGetProperty_proc)pContext->coreaudio.AudioUnitGetProperty)(audioUnit, kAudioUnitProperty_StreamFormat, formatScope, formatElement, &bestFormat, &propSize); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(audioUnit); + return ma_result_from_OSStatus(status); + } + + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(audioUnit); + audioUnit = NULL; + + /* Only a single format is being reported for iOS. */ + pDeviceInfo->nativeDataFormatCount = 1; + + result = ma_format_from_AudioStreamBasicDescription(&bestFormat, &pDeviceInfo->nativeDataFormats[0].format); + if (result != MA_SUCCESS) { + return result; + } + + pDeviceInfo->nativeDataFormats[0].channels = bestFormat.mChannelsPerFrame; + + /* + It looks like Apple are wanting to push the whole AVAudioSession thing. Thus, we need to use that to determine device settings. To do + this we just get the shared instance and inspect. + */ + @autoreleasepool { + AVAudioSession* pAudioSession = [AVAudioSession sharedInstance]; + MA_ASSERT(pAudioSession != NULL); + + pDeviceInfo->nativeDataFormats[0].sampleRate = (ma_uint32)pAudioSession.sampleRate; + } + } +#endif + + (void)pDeviceInfo; /* Unused. */ + return MA_SUCCESS; +} + +static AudioBufferList* ma_allocate_AudioBufferList__coreaudio(ma_uint32 sizeInFrames, ma_format format, ma_uint32 channels, ma_stream_layout layout, const ma_allocation_callbacks* pAllocationCallbacks) +{ + AudioBufferList* pBufferList; + UInt32 audioBufferSizeInBytes; + size_t allocationSize; + + MA_ASSERT(sizeInFrames > 0); + MA_ASSERT(format != ma_format_unknown); + MA_ASSERT(channels > 0); + + allocationSize = sizeof(AudioBufferList) - sizeof(AudioBuffer); /* Subtract sizeof(AudioBuffer) because that part is dynamically sized. */ + if (layout == ma_stream_layout_interleaved) { + /* Interleaved case. This is the simple case because we just have one buffer. */ + allocationSize += sizeof(AudioBuffer) * 1; + } else { + /* Non-interleaved case. This is the more complex case because there's more than one buffer. */ + allocationSize += sizeof(AudioBuffer) * channels; + } + + allocationSize += sizeInFrames * ma_get_bytes_per_frame(format, channels); + + pBufferList = (AudioBufferList*)ma_malloc(allocationSize, pAllocationCallbacks); + if (pBufferList == NULL) { + return NULL; + } + + audioBufferSizeInBytes = (UInt32)(sizeInFrames * ma_get_bytes_per_sample(format)); + + if (layout == ma_stream_layout_interleaved) { + pBufferList->mNumberBuffers = 1; + pBufferList->mBuffers[0].mNumberChannels = channels; + pBufferList->mBuffers[0].mDataByteSize = audioBufferSizeInBytes * channels; + pBufferList->mBuffers[0].mData = (ma_uint8*)pBufferList + sizeof(AudioBufferList); + } else { + ma_uint32 iBuffer; + pBufferList->mNumberBuffers = channels; + for (iBuffer = 0; iBuffer < pBufferList->mNumberBuffers; ++iBuffer) { + pBufferList->mBuffers[iBuffer].mNumberChannels = 1; + pBufferList->mBuffers[iBuffer].mDataByteSize = audioBufferSizeInBytes; + pBufferList->mBuffers[iBuffer].mData = (ma_uint8*)pBufferList + ((sizeof(AudioBufferList) - sizeof(AudioBuffer)) + (sizeof(AudioBuffer) * channels)) + (audioBufferSizeInBytes * iBuffer); + } + } + + return pBufferList; +} + +static ma_result ma_device_realloc_AudioBufferList__coreaudio(ma_device* pDevice, ma_uint32 sizeInFrames, ma_format format, ma_uint32 channels, ma_stream_layout layout) +{ + MA_ASSERT(pDevice != NULL); + MA_ASSERT(format != ma_format_unknown); + MA_ASSERT(channels > 0); + + /* Only resize the buffer if necessary. */ + if (pDevice->coreaudio.audioBufferCapInFrames < sizeInFrames) { + AudioBufferList* pNewAudioBufferList; + + pNewAudioBufferList = ma_allocate_AudioBufferList__coreaudio(sizeInFrames, format, channels, layout, &pDevice->pContext->allocationCallbacks); + if (pNewAudioBufferList == NULL) { + return MA_OUT_OF_MEMORY; + } + + /* At this point we'll have a new AudioBufferList and we can free the old one. */ + ma_free(pDevice->coreaudio.pAudioBufferList, &pDevice->pContext->allocationCallbacks); + pDevice->coreaudio.pAudioBufferList = pNewAudioBufferList; + pDevice->coreaudio.audioBufferCapInFrames = sizeInFrames; + } + + /* Getting here means the capacity of the audio is fine. */ + return MA_SUCCESS; +} + + +static OSStatus ma_on_output__coreaudio(void* pUserData, AudioUnitRenderActionFlags* pActionFlags, const AudioTimeStamp* pTimeStamp, UInt32 busNumber, UInt32 frameCount, AudioBufferList* pBufferList) +{ + ma_device* pDevice = (ma_device*)pUserData; + ma_stream_layout layout; + + MA_ASSERT(pDevice != NULL); + + /*ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "INFO: Output Callback: busNumber=%d, frameCount=%d, mNumberBuffers=%d\n", (int)busNumber, (int)frameCount, (int)pBufferList->mNumberBuffers);*/ + + /* We need to check whether or not we are outputting interleaved or non-interleaved samples. The way we do this is slightly different for each type. */ + layout = ma_stream_layout_interleaved; + if (pBufferList->mBuffers[0].mNumberChannels != pDevice->playback.internalChannels) { + layout = ma_stream_layout_deinterleaved; + } + + if (layout == ma_stream_layout_interleaved) { + /* For now we can assume everything is interleaved. */ + UInt32 iBuffer; + for (iBuffer = 0; iBuffer < pBufferList->mNumberBuffers; ++iBuffer) { + if (pBufferList->mBuffers[iBuffer].mNumberChannels == pDevice->playback.internalChannels) { + ma_uint32 frameCountForThisBuffer = pBufferList->mBuffers[iBuffer].mDataByteSize / ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + if (frameCountForThisBuffer > 0) { + ma_device_handle_backend_data_callback(pDevice, pBufferList->mBuffers[iBuffer].mData, NULL, frameCountForThisBuffer); + } + + /*a_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, " frameCount=%d, mNumberChannels=%d, mDataByteSize=%d\n", (int)frameCount, (int)pBufferList->mBuffers[iBuffer].mNumberChannels, (int)pBufferList->mBuffers[iBuffer].mDataByteSize);*/ + } else { + /* + This case is where the number of channels in the output buffer do not match our internal channels. It could mean that it's + not interleaved, in which case we can't handle right now since miniaudio does not yet support non-interleaved streams. We just + output silence here. + */ + MA_ZERO_MEMORY(pBufferList->mBuffers[iBuffer].mData, pBufferList->mBuffers[iBuffer].mDataByteSize); + /*ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, " WARNING: Outputting silence. frameCount=%d, mNumberChannels=%d, mDataByteSize=%d\n", (int)frameCount, (int)pBufferList->mBuffers[iBuffer].mNumberChannels, (int)pBufferList->mBuffers[iBuffer].mDataByteSize);*/ + } + } + } else { + /* This is the deinterleaved case. We need to update each buffer in groups of internalChannels. This assumes each buffer is the same size. */ + MA_ASSERT(pDevice->playback.internalChannels <= MA_MAX_CHANNELS); /* This should have been validated at initialization time. */ + + /* + For safety we'll check that the internal channels is a multiple of the buffer count. If it's not it means something + very strange has happened and we're not going to support it. + */ + if ((pBufferList->mNumberBuffers % pDevice->playback.internalChannels) == 0) { + ma_uint8 tempBuffer[4096]; + UInt32 iBuffer; + + for (iBuffer = 0; iBuffer < pBufferList->mNumberBuffers; iBuffer += pDevice->playback.internalChannels) { + ma_uint32 frameCountPerBuffer = pBufferList->mBuffers[iBuffer].mDataByteSize / ma_get_bytes_per_sample(pDevice->playback.internalFormat); + ma_uint32 framesRemaining = frameCountPerBuffer; + + while (framesRemaining > 0) { + void* ppDeinterleavedBuffers[MA_MAX_CHANNELS]; + ma_uint32 iChannel; + ma_uint32 framesToRead = sizeof(tempBuffer) / ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + if (framesToRead > framesRemaining) { + framesToRead = framesRemaining; + } + + ma_device_handle_backend_data_callback(pDevice, tempBuffer, NULL, framesToRead); + + for (iChannel = 0; iChannel < pDevice->playback.internalChannels; ++iChannel) { + ppDeinterleavedBuffers[iChannel] = (void*)ma_offset_ptr(pBufferList->mBuffers[iBuffer+iChannel].mData, (frameCountPerBuffer - framesRemaining) * ma_get_bytes_per_sample(pDevice->playback.internalFormat)); + } + + ma_deinterleave_pcm_frames(pDevice->playback.internalFormat, pDevice->playback.internalChannels, framesToRead, tempBuffer, ppDeinterleavedBuffers); + + framesRemaining -= framesToRead; + } + } + } + } + + (void)pActionFlags; + (void)pTimeStamp; + (void)busNumber; + (void)frameCount; + + return noErr; +} + +static OSStatus ma_on_input__coreaudio(void* pUserData, AudioUnitRenderActionFlags* pActionFlags, const AudioTimeStamp* pTimeStamp, UInt32 busNumber, UInt32 frameCount, AudioBufferList* pUnusedBufferList) +{ + ma_device* pDevice = (ma_device*)pUserData; + AudioBufferList* pRenderedBufferList; + ma_result result; + ma_stream_layout layout; + ma_uint32 iBuffer; + OSStatus status; + + MA_ASSERT(pDevice != NULL); + + pRenderedBufferList = (AudioBufferList*)pDevice->coreaudio.pAudioBufferList; + MA_ASSERT(pRenderedBufferList); + + /* We need to check whether or not we are outputting interleaved or non-interleaved samples. The way we do this is slightly different for each type. */ + layout = ma_stream_layout_interleaved; + if (pRenderedBufferList->mBuffers[0].mNumberChannels != pDevice->capture.internalChannels) { + layout = ma_stream_layout_deinterleaved; + } + + /*ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "INFO: Input Callback: busNumber=%d, frameCount=%d, mNumberBuffers=%d\n", (int)busNumber, (int)frameCount, (int)pRenderedBufferList->mNumberBuffers);*/ + + /* + There has been a situation reported where frame count passed into this function is greater than the capacity of + our capture buffer. There doesn't seem to be a reliable way to determine what the maximum frame count will be, + so we need to instead resort to dynamically reallocating our buffer to ensure it's large enough to capture the + number of frames requested by this callback. + */ + result = ma_device_realloc_AudioBufferList__coreaudio(pDevice, frameCount, pDevice->capture.internalFormat, pDevice->capture.internalChannels, layout); + if (result != MA_SUCCESS) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "Failed to allocate AudioBufferList for capture.\n"); + return noErr; + } + + pRenderedBufferList = (AudioBufferList*)pDevice->coreaudio.pAudioBufferList; + MA_ASSERT(pRenderedBufferList); + + /* + When you call AudioUnitRender(), Core Audio tries to be helpful by setting the mDataByteSize to the number of bytes + that were actually rendered. The problem with this is that the next call can fail with -50 due to the size no longer + being set to the capacity of the buffer, but instead the size in bytes of the previous render. This will cause a + problem when a future call to this callback specifies a larger number of frames. + + To work around this we need to explicitly set the size of each buffer to their respective size in bytes. + */ + for (iBuffer = 0; iBuffer < pRenderedBufferList->mNumberBuffers; ++iBuffer) { + pRenderedBufferList->mBuffers[iBuffer].mDataByteSize = pDevice->coreaudio.audioBufferCapInFrames * ma_get_bytes_per_sample(pDevice->capture.internalFormat) * pRenderedBufferList->mBuffers[iBuffer].mNumberChannels; + /*printf("DEBUG: nDataByteSize = %d\n", (int)pRenderedBufferList->mBuffers[iBuffer].mDataByteSize);*/ + } + + status = ((ma_AudioUnitRender_proc)pDevice->pContext->coreaudio.AudioUnitRender)((AudioUnit)pDevice->coreaudio.audioUnitCapture, pActionFlags, pTimeStamp, busNumber, frameCount, pRenderedBufferList); + if (status != noErr) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, "ERROR: AudioUnitRender() failed with %d.\n", (int)status); + return status; + } + + if (layout == ma_stream_layout_interleaved) { + for (iBuffer = 0; iBuffer < pRenderedBufferList->mNumberBuffers; ++iBuffer) { + if (pRenderedBufferList->mBuffers[iBuffer].mNumberChannels == pDevice->capture.internalChannels) { + ma_device_handle_backend_data_callback(pDevice, NULL, pRenderedBufferList->mBuffers[iBuffer].mData, frameCount); + /*ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, " mDataByteSize=%d.\n", (int)pRenderedBufferList->mBuffers[iBuffer].mDataByteSize);*/ + } else { + /* + This case is where the number of channels in the output buffer do not match our internal channels. It could mean that it's + not interleaved, in which case we can't handle right now since miniaudio does not yet support non-interleaved streams. + */ + ma_uint8 silentBuffer[4096]; + ma_uint32 framesRemaining; + + MA_ZERO_MEMORY(silentBuffer, sizeof(silentBuffer)); + + framesRemaining = frameCount; + while (framesRemaining > 0) { + ma_uint32 framesToSend = sizeof(silentBuffer) / ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + if (framesToSend > framesRemaining) { + framesToSend = framesRemaining; + } + + ma_device_handle_backend_data_callback(pDevice, NULL, silentBuffer, framesToSend); + + framesRemaining -= framesToSend; + } + + /*ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_DEBUG, " WARNING: Outputting silence. frameCount=%d, mNumberChannels=%d, mDataByteSize=%d\n", (int)frameCount, (int)pRenderedBufferList->mBuffers[iBuffer].mNumberChannels, (int)pRenderedBufferList->mBuffers[iBuffer].mDataByteSize);*/ + } + } + } else { + /* This is the deinterleaved case. We need to interleave the audio data before sending it to the client. This assumes each buffer is the same size. */ + MA_ASSERT(pDevice->capture.internalChannels <= MA_MAX_CHANNELS); /* This should have been validated at initialization time. */ + + /* + For safety we'll check that the internal channels is a multiple of the buffer count. If it's not it means something + very strange has happened and we're not going to support it. + */ + if ((pRenderedBufferList->mNumberBuffers % pDevice->capture.internalChannels) == 0) { + ma_uint8 tempBuffer[4096]; + for (iBuffer = 0; iBuffer < pRenderedBufferList->mNumberBuffers; iBuffer += pDevice->capture.internalChannels) { + ma_uint32 framesRemaining = frameCount; + while (framesRemaining > 0) { + void* ppDeinterleavedBuffers[MA_MAX_CHANNELS]; + ma_uint32 iChannel; + ma_uint32 framesToSend = sizeof(tempBuffer) / ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + if (framesToSend > framesRemaining) { + framesToSend = framesRemaining; + } + + for (iChannel = 0; iChannel < pDevice->capture.internalChannels; ++iChannel) { + ppDeinterleavedBuffers[iChannel] = (void*)ma_offset_ptr(pRenderedBufferList->mBuffers[iBuffer+iChannel].mData, (frameCount - framesRemaining) * ma_get_bytes_per_sample(pDevice->capture.internalFormat)); + } + + ma_interleave_pcm_frames(pDevice->capture.internalFormat, pDevice->capture.internalChannels, framesToSend, (const void**)ppDeinterleavedBuffers, tempBuffer); + ma_device_handle_backend_data_callback(pDevice, NULL, tempBuffer, framesToSend); + + framesRemaining -= framesToSend; + } + } + } + } + + (void)pActionFlags; + (void)pTimeStamp; + (void)busNumber; + (void)frameCount; + (void)pUnusedBufferList; + + return noErr; +} + +static void on_start_stop__coreaudio(void* pUserData, AudioUnit audioUnit, AudioUnitPropertyID propertyID, AudioUnitScope scope, AudioUnitElement element) +{ + ma_device* pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + /* Don't do anything if it looks like we're just reinitializing due to a device switch. */ + if (((audioUnit == pDevice->coreaudio.audioUnitPlayback) && pDevice->coreaudio.isSwitchingPlaybackDevice) || + ((audioUnit == pDevice->coreaudio.audioUnitCapture) && pDevice->coreaudio.isSwitchingCaptureDevice)) { + return; + } + + /* + There's been a report of a deadlock here when triggered by ma_device_uninit(). It looks like + AudioUnitGetProprty (called below) and AudioComponentInstanceDispose (called in ma_device_uninit) + can try waiting on the same lock. I'm going to try working around this by not calling any Core + Audio APIs in the callback when the device has been stopped or uninitialized. + */ + if (ma_device_get_state(pDevice) == ma_device_state_uninitialized || ma_device_get_state(pDevice) == ma_device_state_stopping || ma_device_get_state(pDevice) == ma_device_state_stopped) { + ma_device__on_notification_stopped(pDevice); + } else { + UInt32 isRunning; + UInt32 isRunningSize = sizeof(isRunning); + OSStatus status = ((ma_AudioUnitGetProperty_proc)pDevice->pContext->coreaudio.AudioUnitGetProperty)(audioUnit, kAudioOutputUnitProperty_IsRunning, scope, element, &isRunning, &isRunningSize); + if (status != noErr) { + goto done; /* Don't really know what to do in this case... just ignore it, I suppose... */ + } + + if (!isRunning) { + /* + The stop event is a bit annoying in Core Audio because it will be called when we automatically switch the default device. Some scenarios to consider: + + 1) When the device is unplugged, this will be called _before_ the default device change notification. + 2) When the device is changed via the default device change notification, this will be called _after_ the switch. + + For case #1, we just check if there's a new default device available. If so, we just ignore the stop event. For case #2 we check a flag. + */ + if (((audioUnit == pDevice->coreaudio.audioUnitPlayback) && pDevice->coreaudio.isDefaultPlaybackDevice) || + ((audioUnit == pDevice->coreaudio.audioUnitCapture) && pDevice->coreaudio.isDefaultCaptureDevice)) { + /* + It looks like the device is switching through an external event, such as the user unplugging the device or changing the default device + via the operating system's sound settings. If we're re-initializing the device, we just terminate because we want the stopping of the + device to be seamless to the client (we don't want them receiving the stopped event and thinking that the device has stopped when it + hasn't!). + */ + if (((audioUnit == pDevice->coreaudio.audioUnitPlayback) && pDevice->coreaudio.isSwitchingPlaybackDevice) || + ((audioUnit == pDevice->coreaudio.audioUnitCapture) && pDevice->coreaudio.isSwitchingCaptureDevice)) { + goto done; + } + + /* + Getting here means the device is not reinitializing which means it may have been unplugged. From what I can see, it looks like Core Audio + will try switching to the new default device seamlessly. We need to somehow find a way to determine whether or not Core Audio will most + likely be successful in switching to the new device. + + TODO: Try to predict if Core Audio will switch devices. If not, the stopped callback needs to be posted. + */ + goto done; + } + + /* Getting here means we need to stop the device. */ + ma_device__on_notification_stopped(pDevice); + } + } + + (void)propertyID; /* Unused. */ + +done: + /* Always signal the stop event. It's possible for the "else" case to get hit which can happen during an interruption. */ + ma_event_signal(&pDevice->coreaudio.stopEvent); +} + +#if defined(MA_APPLE_DESKTOP) +static ma_spinlock g_DeviceTrackingInitLock_CoreAudio = 0; /* A spinlock for mutal exclusion of the init/uninit of the global tracking data. Initialization to 0 is what we need. */ +static ma_uint32 g_DeviceTrackingInitCounter_CoreAudio = 0; +static ma_mutex g_DeviceTrackingMutex_CoreAudio; +static ma_device** g_ppTrackedDevices_CoreAudio = NULL; +static ma_uint32 g_TrackedDeviceCap_CoreAudio = 0; +static ma_uint32 g_TrackedDeviceCount_CoreAudio = 0; + +static OSStatus ma_default_device_changed__coreaudio(AudioObjectID objectID, UInt32 addressCount, const AudioObjectPropertyAddress* pAddresses, void* pUserData) +{ + ma_device_type deviceType; + + /* Not sure if I really need to check this, but it makes me feel better. */ + if (addressCount == 0) { + return noErr; + } + + if (pAddresses[0].mSelector == kAudioHardwarePropertyDefaultOutputDevice) { + deviceType = ma_device_type_playback; + } else if (pAddresses[0].mSelector == kAudioHardwarePropertyDefaultInputDevice) { + deviceType = ma_device_type_capture; + } else { + return noErr; /* Should never hit this. */ + } + + ma_mutex_lock(&g_DeviceTrackingMutex_CoreAudio); + { + ma_uint32 iDevice; + for (iDevice = 0; iDevice < g_TrackedDeviceCount_CoreAudio; iDevice += 1) { + ma_result reinitResult; + ma_device* pDevice; + + pDevice = g_ppTrackedDevices_CoreAudio[iDevice]; + if (pDevice->type == deviceType || pDevice->type == ma_device_type_duplex) { + if (deviceType == ma_device_type_playback) { + pDevice->coreaudio.isSwitchingPlaybackDevice = MA_TRUE; + reinitResult = ma_device_reinit_internal__coreaudio(pDevice, deviceType, MA_TRUE); + pDevice->coreaudio.isSwitchingPlaybackDevice = MA_FALSE; + } else { + pDevice->coreaudio.isSwitchingCaptureDevice = MA_TRUE; + reinitResult = ma_device_reinit_internal__coreaudio(pDevice, deviceType, MA_TRUE); + pDevice->coreaudio.isSwitchingCaptureDevice = MA_FALSE; + } + + if (reinitResult == MA_SUCCESS) { + ma_device__post_init_setup(pDevice, deviceType); + + /* Restart the device if required. If this fails we need to stop the device entirely. */ + if (ma_device_get_state(pDevice) == ma_device_state_started) { + OSStatus status; + if (deviceType == ma_device_type_playback) { + status = ((ma_AudioOutputUnitStart_proc)pDevice->pContext->coreaudio.AudioOutputUnitStart)((AudioUnit)pDevice->coreaudio.audioUnitPlayback); + if (status != noErr) { + if (pDevice->type == ma_device_type_duplex) { + ((ma_AudioOutputUnitStop_proc)pDevice->pContext->coreaudio.AudioOutputUnitStop)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + } + ma_device__set_state(pDevice, ma_device_state_stopped); + } + } else if (deviceType == ma_device_type_capture) { + status = ((ma_AudioOutputUnitStart_proc)pDevice->pContext->coreaudio.AudioOutputUnitStart)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + if (status != noErr) { + if (pDevice->type == ma_device_type_duplex) { + ((ma_AudioOutputUnitStop_proc)pDevice->pContext->coreaudio.AudioOutputUnitStop)((AudioUnit)pDevice->coreaudio.audioUnitPlayback); + } + ma_device__set_state(pDevice, ma_device_state_stopped); + } + } + } + + ma_device__on_notification_rerouted(pDevice); + } + } + } + } + ma_mutex_unlock(&g_DeviceTrackingMutex_CoreAudio); + + /* Unused parameters. */ + (void)objectID; + (void)pUserData; + + return noErr; +} + +static ma_result ma_context__init_device_tracking__coreaudio(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + + ma_spinlock_lock(&g_DeviceTrackingInitLock_CoreAudio); + { + /* Don't do anything if we've already initialized device tracking. */ + if (g_DeviceTrackingInitCounter_CoreAudio == 0) { + AudioObjectPropertyAddress propAddress; + propAddress.mScope = kAudioObjectPropertyScopeGlobal; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + ma_mutex_init(&g_DeviceTrackingMutex_CoreAudio); + + propAddress.mSelector = kAudioHardwarePropertyDefaultInputDevice; + ((ma_AudioObjectAddPropertyListener_proc)pContext->coreaudio.AudioObjectAddPropertyListener)(kAudioObjectSystemObject, &propAddress, &ma_default_device_changed__coreaudio, NULL); + + propAddress.mSelector = kAudioHardwarePropertyDefaultOutputDevice; + ((ma_AudioObjectAddPropertyListener_proc)pContext->coreaudio.AudioObjectAddPropertyListener)(kAudioObjectSystemObject, &propAddress, &ma_default_device_changed__coreaudio, NULL); + + } + g_DeviceTrackingInitCounter_CoreAudio += 1; + } + ma_spinlock_unlock(&g_DeviceTrackingInitLock_CoreAudio); + + return MA_SUCCESS; +} + +static ma_result ma_context__uninit_device_tracking__coreaudio(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + + ma_spinlock_lock(&g_DeviceTrackingInitLock_CoreAudio); + { + if (g_DeviceTrackingInitCounter_CoreAudio > 0) + g_DeviceTrackingInitCounter_CoreAudio -= 1; + + if (g_DeviceTrackingInitCounter_CoreAudio == 0) { + AudioObjectPropertyAddress propAddress; + propAddress.mScope = kAudioObjectPropertyScopeGlobal; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + propAddress.mSelector = kAudioHardwarePropertyDefaultInputDevice; + ((ma_AudioObjectRemovePropertyListener_proc)pContext->coreaudio.AudioObjectRemovePropertyListener)(kAudioObjectSystemObject, &propAddress, &ma_default_device_changed__coreaudio, NULL); + + propAddress.mSelector = kAudioHardwarePropertyDefaultOutputDevice; + ((ma_AudioObjectRemovePropertyListener_proc)pContext->coreaudio.AudioObjectRemovePropertyListener)(kAudioObjectSystemObject, &propAddress, &ma_default_device_changed__coreaudio, NULL); + + /* At this point there should be no tracked devices. If not there's an error somewhere. */ + if (g_ppTrackedDevices_CoreAudio != NULL) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_WARNING, "You have uninitialized all contexts while an associated device is still active."); + ma_spinlock_unlock(&g_DeviceTrackingInitLock_CoreAudio); + return MA_INVALID_OPERATION; + } + + ma_mutex_uninit(&g_DeviceTrackingMutex_CoreAudio); + } + } + ma_spinlock_unlock(&g_DeviceTrackingInitLock_CoreAudio); + + return MA_SUCCESS; +} + +static ma_result ma_device__track__coreaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + ma_mutex_lock(&g_DeviceTrackingMutex_CoreAudio); + { + /* Allocate memory if required. */ + if (g_TrackedDeviceCap_CoreAudio <= g_TrackedDeviceCount_CoreAudio) { + ma_uint32 newCap; + ma_device** ppNewDevices; + + newCap = g_TrackedDeviceCap_CoreAudio * 2; + if (newCap == 0) { + newCap = 1; + } + + ppNewDevices = (ma_device**)ma_realloc(g_ppTrackedDevices_CoreAudio, sizeof(*g_ppTrackedDevices_CoreAudio)*newCap, &pDevice->pContext->allocationCallbacks); + if (ppNewDevices == NULL) { + ma_mutex_unlock(&g_DeviceTrackingMutex_CoreAudio); + return MA_OUT_OF_MEMORY; + } + + g_ppTrackedDevices_CoreAudio = ppNewDevices; + g_TrackedDeviceCap_CoreAudio = newCap; + } + + g_ppTrackedDevices_CoreAudio[g_TrackedDeviceCount_CoreAudio] = pDevice; + g_TrackedDeviceCount_CoreAudio += 1; + } + ma_mutex_unlock(&g_DeviceTrackingMutex_CoreAudio); + + return MA_SUCCESS; +} + +static ma_result ma_device__untrack__coreaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + ma_mutex_lock(&g_DeviceTrackingMutex_CoreAudio); + { + ma_uint32 iDevice; + for (iDevice = 0; iDevice < g_TrackedDeviceCount_CoreAudio; iDevice += 1) { + if (g_ppTrackedDevices_CoreAudio[iDevice] == pDevice) { + /* We've found the device. We now need to remove it from the list. */ + ma_uint32 jDevice; + for (jDevice = iDevice; jDevice < g_TrackedDeviceCount_CoreAudio-1; jDevice += 1) { + g_ppTrackedDevices_CoreAudio[jDevice] = g_ppTrackedDevices_CoreAudio[jDevice+1]; + } + + g_TrackedDeviceCount_CoreAudio -= 1; + + /* If there's nothing else in the list we need to free memory. */ + if (g_TrackedDeviceCount_CoreAudio == 0) { + ma_free(g_ppTrackedDevices_CoreAudio, &pDevice->pContext->allocationCallbacks); + g_ppTrackedDevices_CoreAudio = NULL; + g_TrackedDeviceCap_CoreAudio = 0; + } + + break; + } + } + } + ma_mutex_unlock(&g_DeviceTrackingMutex_CoreAudio); + + return MA_SUCCESS; +} +#endif + +#if defined(MA_APPLE_MOBILE) +@interface ma_ios_notification_handler:NSObject { + ma_device* m_pDevice; +} +@end + +@implementation ma_ios_notification_handler +-(id)init:(ma_device*)pDevice +{ + self = [super init]; + m_pDevice = pDevice; + + /* For route changes. */ + [[NSNotificationCenter defaultCenter] addObserver:self selector:@selector(handle_route_change:) name:AVAudioSessionRouteChangeNotification object:[AVAudioSession sharedInstance]]; + + /* For interruptions. */ + [[NSNotificationCenter defaultCenter] addObserver:self selector:@selector(handle_interruption:) name:AVAudioSessionInterruptionNotification object:[AVAudioSession sharedInstance]]; + + return self; +} + +-(void)dealloc +{ + [self remove_handler]; + + #if defined(__has_feature) + #if !__has_feature(objc_arc) + [super dealloc]; + #endif + #endif +} + +-(void)remove_handler +{ + [[NSNotificationCenter defaultCenter] removeObserver:self name:AVAudioSessionRouteChangeNotification object:nil]; + [[NSNotificationCenter defaultCenter] removeObserver:self name:AVAudioSessionInterruptionNotification object:nil]; +} + +-(void)handle_interruption:(NSNotification*)pNotification +{ + NSInteger type = [[[pNotification userInfo] objectForKey:AVAudioSessionInterruptionTypeKey] integerValue]; + switch (type) + { + case AVAudioSessionInterruptionTypeBegan: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Interruption: AVAudioSessionInterruptionTypeBegan\n"); + + /* + Core Audio will have stopped the internal device automatically, but we need explicitly + stop it at a higher level to ensure miniaudio-specific state is updated for consistency. + */ + ma_device_stop(m_pDevice); + + /* + Fire the notification after the device has been stopped to ensure it's in the correct + state when the notification handler is invoked. + */ + ma_device__on_notification_interruption_began(m_pDevice); + } break; + + case AVAudioSessionInterruptionTypeEnded: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Interruption: AVAudioSessionInterruptionTypeEnded\n"); + ma_device__on_notification_interruption_ended(m_pDevice); + } break; + } +} + +-(void)handle_route_change:(NSNotification*)pNotification +{ + AVAudioSession* pSession = [AVAudioSession sharedInstance]; + + NSInteger reason = [[[pNotification userInfo] objectForKey:AVAudioSessionRouteChangeReasonKey] integerValue]; + switch (reason) + { + case AVAudioSessionRouteChangeReasonOldDeviceUnavailable: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Route Changed: AVAudioSessionRouteChangeReasonOldDeviceUnavailable\n"); + } break; + + case AVAudioSessionRouteChangeReasonNewDeviceAvailable: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Route Changed: AVAudioSessionRouteChangeReasonNewDeviceAvailable\n"); + } break; + + case AVAudioSessionRouteChangeReasonNoSuitableRouteForCategory: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Route Changed: AVAudioSessionRouteChangeReasonNoSuitableRouteForCategory\n"); + } break; + + case AVAudioSessionRouteChangeReasonWakeFromSleep: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Route Changed: AVAudioSessionRouteChangeReasonWakeFromSleep\n"); + } break; + + case AVAudioSessionRouteChangeReasonOverride: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Route Changed: AVAudioSessionRouteChangeReasonOverride\n"); + } break; + + case AVAudioSessionRouteChangeReasonCategoryChange: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Route Changed: AVAudioSessionRouteChangeReasonCategoryChange\n"); + } break; + + case AVAudioSessionRouteChangeReasonUnknown: + default: + { + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_INFO, "[Core Audio] Route Changed: AVAudioSessionRouteChangeReasonUnknown\n"); + } break; + } + + ma_log_postf(ma_device_get_log(m_pDevice), MA_LOG_LEVEL_DEBUG, "[Core Audio] Changing Route. inputNumberChannels=%d; outputNumberOfChannels=%d\n", (int)pSession.inputNumberOfChannels, (int)pSession.outputNumberOfChannels); + + /* Let the application know about the route change. */ + ma_device__on_notification_rerouted(m_pDevice); +} +@end +#endif + +static ma_result ma_device_uninit__coreaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + MA_ASSERT(ma_device_get_state(pDevice) == ma_device_state_uninitialized); + +#if defined(MA_APPLE_DESKTOP) + /* + Make sure we're no longer tracking the device. It doesn't matter if we call this for a non-default device because it'll + just gracefully ignore it. + */ + ma_device__untrack__coreaudio(pDevice); +#endif +#if defined(MA_APPLE_MOBILE) + if (pDevice->coreaudio.pNotificationHandler != NULL) { + ma_ios_notification_handler* pNotificationHandler = (MA_BRIDGE_TRANSFER ma_ios_notification_handler*)pDevice->coreaudio.pNotificationHandler; + [pNotificationHandler remove_handler]; + } +#endif + + if (pDevice->coreaudio.audioUnitCapture != NULL) { + ((ma_AudioComponentInstanceDispose_proc)pDevice->pContext->coreaudio.AudioComponentInstanceDispose)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + } + if (pDevice->coreaudio.audioUnitPlayback != NULL) { + ((ma_AudioComponentInstanceDispose_proc)pDevice->pContext->coreaudio.AudioComponentInstanceDispose)((AudioUnit)pDevice->coreaudio.audioUnitPlayback); + } + + if (pDevice->coreaudio.pAudioBufferList) { + ma_free(pDevice->coreaudio.pAudioBufferList, &pDevice->pContext->allocationCallbacks); + } + + return MA_SUCCESS; +} + +typedef struct +{ + ma_bool32 allowNominalSampleRateChange; + + /* Input. */ + ma_format formatIn; + ma_uint32 channelsIn; + ma_uint32 sampleRateIn; + ma_channel channelMapIn[MA_MAX_CHANNELS]; + ma_uint32 periodSizeInFramesIn; + ma_uint32 periodSizeInMillisecondsIn; + ma_uint32 periodsIn; + ma_share_mode shareMode; + ma_performance_profile performanceProfile; + ma_bool32 registerStopEvent; + + /* Output. */ +#if defined(MA_APPLE_DESKTOP) + AudioObjectID deviceObjectID; +#endif + AudioComponent component; + AudioUnit audioUnit; + AudioBufferList* pAudioBufferList; /* Only used for input devices. */ + ma_format formatOut; + ma_uint32 channelsOut; + ma_uint32 sampleRateOut; + ma_channel channelMapOut[MA_MAX_CHANNELS]; + ma_uint32 periodSizeInFramesOut; + ma_uint32 periodsOut; + char deviceName[256]; +} ma_device_init_internal_data__coreaudio; + +static ma_result ma_device_init_internal__coreaudio(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_init_internal_data__coreaudio* pData, void* pDevice_DoNotReference) /* <-- pDevice is typed as void* intentionally so as to avoid accidentally referencing it. */ +{ + ma_result result = MA_SUCCESS; + OSStatus status; + UInt32 enableIOFlag; + AudioStreamBasicDescription bestFormat; + ma_uint32 actualPeriodSizeInFrames; + AURenderCallbackStruct callbackInfo; +#if defined(MA_APPLE_DESKTOP) + AudioObjectID deviceObjectID; +#endif + + /* This API should only be used for a single device type: playback or capture. No full-duplex mode. */ + if (deviceType == ma_device_type_duplex) { + return MA_INVALID_ARGS; + } + + MA_ASSERT(pContext != NULL); + MA_ASSERT(deviceType == ma_device_type_playback || deviceType == ma_device_type_capture); + +#if defined(MA_APPLE_DESKTOP) + pData->deviceObjectID = 0; +#endif + pData->component = NULL; + pData->audioUnit = NULL; + pData->pAudioBufferList = NULL; + +#if defined(MA_APPLE_DESKTOP) + result = ma_find_AudioObjectID(pContext, deviceType, pDeviceID, &deviceObjectID); + if (result != MA_SUCCESS) { + return result; + } + + pData->deviceObjectID = deviceObjectID; +#endif + + /* Core audio doesn't really use the notion of a period so we can leave this unmodified, but not too over the top. */ + pData->periodsOut = pData->periodsIn; + if (pData->periodsOut == 0) { + pData->periodsOut = MA_DEFAULT_PERIODS; + } + if (pData->periodsOut > 16) { + pData->periodsOut = 16; + } + + + /* Audio unit. */ + status = ((ma_AudioComponentInstanceNew_proc)pContext->coreaudio.AudioComponentInstanceNew)((AudioComponent)pContext->coreaudio.component, (AudioUnit*)&pData->audioUnit); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + + + /* The input/output buses need to be explicitly enabled and disabled. We set the flag based on the output unit first, then we just swap it for input. */ + enableIOFlag = 1; + if (deviceType == ma_device_type_capture) { + enableIOFlag = 0; + } + + status = ((ma_AudioUnitSetProperty_proc)pContext->coreaudio.AudioUnitSetProperty)(pData->audioUnit, kAudioOutputUnitProperty_EnableIO, kAudioUnitScope_Output, MA_COREAUDIO_OUTPUT_BUS, &enableIOFlag, sizeof(enableIOFlag)); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + + enableIOFlag = (enableIOFlag == 0) ? 1 : 0; + status = ((ma_AudioUnitSetProperty_proc)pContext->coreaudio.AudioUnitSetProperty)(pData->audioUnit, kAudioOutputUnitProperty_EnableIO, kAudioUnitScope_Input, MA_COREAUDIO_INPUT_BUS, &enableIOFlag, sizeof(enableIOFlag)); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + + + /* Set the device to use with this audio unit. This is only used on desktop since we are using defaults on mobile. */ +#if defined(MA_APPLE_DESKTOP) + status = ((ma_AudioUnitSetProperty_proc)pContext->coreaudio.AudioUnitSetProperty)(pData->audioUnit, kAudioOutputUnitProperty_CurrentDevice, kAudioUnitScope_Global, 0, &deviceObjectID, sizeof(deviceObjectID)); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(result); + } +#else + /* + For some reason it looks like Apple is only allowing selection of the input device. There does not appear to be any way to change + the default output route. I have no idea why this is like this, but for now we'll only be able to configure capture devices. + */ + if (pDeviceID != NULL) { + if (deviceType == ma_device_type_capture) { + ma_bool32 found = MA_FALSE; + NSArray *pInputs = [[[AVAudioSession sharedInstance] currentRoute] inputs]; + for (AVAudioSessionPortDescription* pPortDesc in pInputs) { + if (strcmp(pDeviceID->coreaudio, [pPortDesc.UID UTF8String]) == 0) { + [[AVAudioSession sharedInstance] setPreferredInput:pPortDesc error:nil]; + found = MA_TRUE; + break; + } + } + + if (found == MA_FALSE) { + return MA_DOES_NOT_EXIST; + } + } + } +#endif + + /* + Format. This is the hardest part of initialization because there's a few variables to take into account. + 1) The format must be supported by the device. + 2) The format must be supported miniaudio. + 3) There's a priority that miniaudio prefers. + + Ideally we would like to use a format that's as close to the hardware as possible so we can get as close to a passthrough as possible. The + most important property is the sample rate. miniaudio can do format conversion for any sample rate and channel count, but cannot do the same + for the sample data format. If the sample data format is not supported by miniaudio it must be ignored completely. + + On mobile platforms this is a bit different. We just force the use of whatever the audio unit's current format is set to. + */ + { + AudioStreamBasicDescription origFormat; + UInt32 origFormatSize = sizeof(origFormat); + AudioUnitScope formatScope = (deviceType == ma_device_type_playback) ? kAudioUnitScope_Input : kAudioUnitScope_Output; + AudioUnitElement formatElement = (deviceType == ma_device_type_playback) ? MA_COREAUDIO_OUTPUT_BUS : MA_COREAUDIO_INPUT_BUS; + + if (deviceType == ma_device_type_playback) { + status = ((ma_AudioUnitGetProperty_proc)pContext->coreaudio.AudioUnitGetProperty)(pData->audioUnit, kAudioUnitProperty_StreamFormat, kAudioUnitScope_Output, MA_COREAUDIO_OUTPUT_BUS, &origFormat, &origFormatSize); + } else { + status = ((ma_AudioUnitGetProperty_proc)pContext->coreaudio.AudioUnitGetProperty)(pData->audioUnit, kAudioUnitProperty_StreamFormat, kAudioUnitScope_Input, MA_COREAUDIO_INPUT_BUS, &origFormat, &origFormatSize); + } + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + + #if defined(MA_APPLE_DESKTOP) + result = ma_find_best_format__coreaudio(pContext, deviceObjectID, deviceType, pData->formatIn, pData->channelsIn, pData->sampleRateIn, &origFormat, &bestFormat); + if (result != MA_SUCCESS) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return result; + } + + /* + Technical Note TN2091: Device input using the HAL Output Audio Unit + https://developer.apple.com/library/archive/technotes/tn2091/_index.html + + This documentation says the following: + + The internal AudioConverter can handle any *simple* conversion. Typically, this means that a client can specify ANY + variant of the PCM formats. Consequently, the device's sample rate should match the desired sample rate. If sample rate + conversion is needed, it can be accomplished by buffering the input and converting the data on a separate thread with + another AudioConverter. + + The important part here is the mention that it can handle *simple* conversions, which does *not* include sample rate. We + therefore want to ensure the sample rate stays consistent. This document is specifically for input, but I'm going to play it + safe and apply the same rule to output as well. + + I have tried going against the documentation by setting the sample rate anyway, but this just results in AudioUnitRender() + returning a result code of -10863. I have also tried changing the format directly on the input scope on the input bus, but + this just results in `ca_require: IsStreamFormatWritable(inScope, inElement) NotWritable` when trying to set the format. + + Something that does seem to work, however, has been setting the nominal sample rate on the device object. The problem with + this, however, is that it actually changes the sample rate at the operating system level and not just the application. This + could be intrusive to the user, however, so I don't think it's wise to make this the default. Instead I'm making this a + configuration option. When the `coreaudio.allowNominalSampleRateChange` config option is set to true, changing the sample + rate will be allowed. Otherwise it'll be fixed to the current sample rate. To check the system-defined sample rate, run + the Audio MIDI Setup program that comes installed on macOS and observe how the sample rate changes as the sample rate is + changed by miniaudio. + */ + if (pData->allowNominalSampleRateChange) { + AudioValueRange sampleRateRange; + AudioObjectPropertyAddress propAddress; + + sampleRateRange.mMinimum = bestFormat.mSampleRate; + sampleRateRange.mMaximum = bestFormat.mSampleRate; + + propAddress.mSelector = kAudioDevicePropertyNominalSampleRate; + propAddress.mScope = (deviceType == ma_device_type_playback) ? kAudioObjectPropertyScopeOutput : kAudioObjectPropertyScopeInput; + propAddress.mElement = AUDIO_OBJECT_PROPERTY_ELEMENT; + + status = ((ma_AudioObjectSetPropertyData_proc)pContext->coreaudio.AudioObjectSetPropertyData)(deviceObjectID, &propAddress, 0, NULL, sizeof(sampleRateRange), &sampleRateRange); + if (status != noErr) { + bestFormat.mSampleRate = origFormat.mSampleRate; + } + } else { + bestFormat.mSampleRate = origFormat.mSampleRate; + } + + status = ((ma_AudioUnitSetProperty_proc)pContext->coreaudio.AudioUnitSetProperty)(pData->audioUnit, kAudioUnitProperty_StreamFormat, formatScope, formatElement, &bestFormat, sizeof(bestFormat)); + if (status != noErr) { + /* We failed to set the format, so fall back to the current format of the audio unit. */ + bestFormat = origFormat; + } + #else + bestFormat = origFormat; + + /* + Sample rate is a little different here because for some reason kAudioUnitProperty_StreamFormat returns 0... Oh well. We need to instead try + setting the sample rate to what the user has requested and then just see the results of it. Need to use some Objective-C here for this since + it depends on Apple's AVAudioSession API. To do this we just get the shared AVAudioSession instance and then set it. Note that from what I + can tell, it looks like the sample rate is shared between playback and capture for everything. + */ + @autoreleasepool { + AVAudioSession* pAudioSession = [AVAudioSession sharedInstance]; + MA_ASSERT(pAudioSession != NULL); + + [pAudioSession setPreferredSampleRate:(double)pData->sampleRateIn error:nil]; + bestFormat.mSampleRate = pAudioSession.sampleRate; + + /* + I've had a report that the channel count returned by AudioUnitGetProperty above is inconsistent with + AVAudioSession outputNumberOfChannels. I'm going to try using the AVAudioSession values instead. + + UPDATE 20/02/2025: + When testing on the simulator with an iPhone 15 and iOS 17 I get an error when initializing the audio + unit if set the input channels to pAudioSession.inputNumberOfChannels. What is happening is the channel + count returned from AudioUnitGetProperty() above is set to 2, but pAudioSession is reporting a channel + count of 1. When this happens, the call to AudioUnitSetProprty() below just down below will succeed, but + AudioUnitInitialize() further down will fail. The only solution I have come up with is to not set the + channel count to pAudioSession.inputNumberOfChannels. + */ + if (deviceType == ma_device_type_playback) { + bestFormat.mChannelsPerFrame = (UInt32)pAudioSession.outputNumberOfChannels; + } + + #if 0 + if (deviceType == ma_device_type_capture) { + /*printf("DEBUG: bestFormat.mChannelsPerFrame = %d; pAudioSession.inputNumberOfChannels = %d\n", (int)bestFormat.mChannelsPerFrame, (int)pAudioSession.inputNumberOfChannels);*/ + bestFormat.mChannelsPerFrame = (UInt32)pAudioSession.inputNumberOfChannels; + } + #endif + } + + + status = ((ma_AudioUnitSetProperty_proc)pContext->coreaudio.AudioUnitSetProperty)(pData->audioUnit, kAudioUnitProperty_StreamFormat, formatScope, formatElement, &bestFormat, sizeof(bestFormat)); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + #endif + + result = ma_format_from_AudioStreamBasicDescription(&bestFormat, &pData->formatOut); + if (result != MA_SUCCESS) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return result; + } + + if (pData->formatOut == ma_format_unknown) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return MA_FORMAT_NOT_SUPPORTED; + } + + pData->channelsOut = bestFormat.mChannelsPerFrame; + pData->sampleRateOut = (ma_uint32)bestFormat.mSampleRate; + } + + /* Clamp the channel count for safety. */ + if (pData->channelsOut > MA_MAX_CHANNELS) { + pData->channelsOut = MA_MAX_CHANNELS; + } + + /* + Internal channel map. This is weird in my testing. If I use the AudioObject to get the + channel map, the channel descriptions are set to "Unknown" for some reason. To work around + this it looks like retrieving it from the AudioUnit will work. However, and this is where + it gets weird, it doesn't seem to work with capture devices, nor at all on iOS... Therefore + I'm going to fall back to a default assumption in these cases. + */ +#if defined(MA_APPLE_DESKTOP) + result = ma_get_AudioUnit_channel_map(pContext, pData->audioUnit, deviceType, pData->channelMapOut, pData->channelsOut); + if (result != MA_SUCCESS) { + #if 0 + /* Try falling back to the channel map from the AudioObject. */ + result = ma_get_AudioObject_channel_map(pContext, deviceObjectID, deviceType, pData->channelMapOut, pData->channelsOut); + if (result != MA_SUCCESS) { + return result; + } + #else + /* Fall back to default assumptions. */ + ma_channel_map_init_standard(ma_standard_channel_map_default, pData->channelMapOut, ma_countof(pData->channelMapOut), pData->channelsOut); + #endif + } +#else + /* TODO: Figure out how to get the channel map using AVAudioSession. */ + ma_channel_map_init_standard(ma_standard_channel_map_default, pData->channelMapOut, ma_countof(pData->channelMapOut), pData->channelsOut); +#endif + + + /* Buffer size. Not allowing this to be configurable on iOS. */ + if (pData->periodSizeInFramesIn == 0) { + if (pData->periodSizeInMillisecondsIn == 0) { + if (pData->performanceProfile == ma_performance_profile_low_latency) { + actualPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_LOW_LATENCY, pData->sampleRateOut); + } else { + actualPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_CONSERVATIVE, pData->sampleRateOut); + } + } else { + actualPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(pData->periodSizeInMillisecondsIn, pData->sampleRateOut); + } + } else { + actualPeriodSizeInFrames = pData->periodSizeInFramesIn; + } + +#if defined(MA_APPLE_DESKTOP) + result = ma_set_AudioObject_buffer_size_in_frames(pContext, deviceObjectID, deviceType, &actualPeriodSizeInFrames); + if (result != MA_SUCCESS) { + return result; + } +#else + /* + On iOS, the size of the IO buffer needs to be specified in seconds and is a floating point + number. I don't trust any potential truncation errors due to converting from float to integer + so I'm going to explicitly set the actual period size to the next power of 2. + */ + @autoreleasepool { + AVAudioSession* pAudioSession = [AVAudioSession sharedInstance]; + MA_ASSERT(pAudioSession != NULL); + + [pAudioSession setPreferredIOBufferDuration:((float)actualPeriodSizeInFrames / pAudioSession.sampleRate) error:nil]; + actualPeriodSizeInFrames = ma_next_power_of_2((ma_uint32)(pAudioSession.IOBufferDuration * pAudioSession.sampleRate)); + } +#endif + + + /* + During testing I discovered that the buffer size can be too big. You'll get an error like this: + + kAudioUnitErr_TooManyFramesToProcess : inFramesToProcess=4096, mMaxFramesPerSlice=512 + + Note how inFramesToProcess is smaller than mMaxFramesPerSlice. To fix, we need to set kAudioUnitProperty_MaximumFramesPerSlice to that + of the size of our buffer, or do it the other way around and set our buffer size to the kAudioUnitProperty_MaximumFramesPerSlice. + */ + status = ((ma_AudioUnitSetProperty_proc)pContext->coreaudio.AudioUnitSetProperty)(pData->audioUnit, kAudioUnitProperty_MaximumFramesPerSlice, kAudioUnitScope_Global, 0, &actualPeriodSizeInFrames, sizeof(actualPeriodSizeInFrames)); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + + pData->periodSizeInFramesOut = (ma_uint32)actualPeriodSizeInFrames; + + /* We need a buffer list if this is an input device. We render into this in the input callback. */ + if (deviceType == ma_device_type_capture) { + ma_bool32 isInterleaved = (bestFormat.mFormatFlags & kAudioFormatFlagIsNonInterleaved) == 0; + AudioBufferList* pBufferList; + + pBufferList = ma_allocate_AudioBufferList__coreaudio(pData->periodSizeInFramesOut, pData->formatOut, pData->channelsOut, (isInterleaved) ? ma_stream_layout_interleaved : ma_stream_layout_deinterleaved, &pContext->allocationCallbacks); + if (pBufferList == NULL) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return MA_OUT_OF_MEMORY; + } + + pData->pAudioBufferList = pBufferList; + } + + /* Callbacks. */ + callbackInfo.inputProcRefCon = pDevice_DoNotReference; + if (deviceType == ma_device_type_playback) { + callbackInfo.inputProc = ma_on_output__coreaudio; + status = ((ma_AudioUnitSetProperty_proc)pContext->coreaudio.AudioUnitSetProperty)(pData->audioUnit, kAudioUnitProperty_SetRenderCallback, kAudioUnitScope_Global, 0, &callbackInfo, sizeof(callbackInfo)); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + } else { + callbackInfo.inputProc = ma_on_input__coreaudio; + status = ((ma_AudioUnitSetProperty_proc)pContext->coreaudio.AudioUnitSetProperty)(pData->audioUnit, kAudioOutputUnitProperty_SetInputCallback, kAudioUnitScope_Global, 0, &callbackInfo, sizeof(callbackInfo)); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + } + + /* We need to listen for stop events. */ + if (pData->registerStopEvent) { + status = ((ma_AudioUnitAddPropertyListener_proc)pContext->coreaudio.AudioUnitAddPropertyListener)(pData->audioUnit, kAudioOutputUnitProperty_IsRunning, on_start_stop__coreaudio, pDevice_DoNotReference); + if (status != noErr) { + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + } + + /* Initialize the audio unit. */ + status = ((ma_AudioUnitInitialize_proc)pContext->coreaudio.AudioUnitInitialize)(pData->audioUnit); + if (status != noErr) { + ma_free(pData->pAudioBufferList, &pContext->allocationCallbacks); + pData->pAudioBufferList = NULL; + ((ma_AudioComponentInstanceDispose_proc)pContext->coreaudio.AudioComponentInstanceDispose)(pData->audioUnit); + return ma_result_from_OSStatus(status); + } + + /* Grab the name. */ +#if defined(MA_APPLE_DESKTOP) + ma_get_AudioObject_name(pContext, deviceObjectID, sizeof(pData->deviceName), pData->deviceName); +#else + if (deviceType == ma_device_type_playback) { + ma_strcpy_s(pData->deviceName, sizeof(pData->deviceName), MA_DEFAULT_PLAYBACK_DEVICE_NAME); + } else { + ma_strcpy_s(pData->deviceName, sizeof(pData->deviceName), MA_DEFAULT_CAPTURE_DEVICE_NAME); + } +#endif + + return result; +} + +#if defined(MA_APPLE_DESKTOP) +static ma_result ma_device_reinit_internal__coreaudio(ma_device* pDevice, ma_device_type deviceType, ma_bool32 disposePreviousAudioUnit) +{ + ma_device_init_internal_data__coreaudio data; + ma_result result; + + /* This should only be called for playback or capture, not duplex. */ + if (deviceType == ma_device_type_duplex) { + return MA_INVALID_ARGS; + } + + data.allowNominalSampleRateChange = MA_FALSE; /* Don't change the nominal sample rate when switching devices. */ + + if (deviceType == ma_device_type_capture) { + data.formatIn = pDevice->capture.format; + data.channelsIn = pDevice->capture.channels; + data.sampleRateIn = pDevice->sampleRate; + MA_COPY_MEMORY(data.channelMapIn, pDevice->capture.channelMap, sizeof(pDevice->capture.channelMap)); + data.shareMode = pDevice->capture.shareMode; + data.performanceProfile = pDevice->coreaudio.originalPerformanceProfile; + data.registerStopEvent = MA_TRUE; + + if (disposePreviousAudioUnit) { + ((ma_AudioOutputUnitStop_proc)pDevice->pContext->coreaudio.AudioOutputUnitStop)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + ((ma_AudioComponentInstanceDispose_proc)pDevice->pContext->coreaudio.AudioComponentInstanceDispose)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + } + if (pDevice->coreaudio.pAudioBufferList) { + ma_free(pDevice->coreaudio.pAudioBufferList, &pDevice->pContext->allocationCallbacks); + } + } else if (deviceType == ma_device_type_playback) { + data.formatIn = pDevice->playback.format; + data.channelsIn = pDevice->playback.channels; + data.sampleRateIn = pDevice->sampleRate; + MA_COPY_MEMORY(data.channelMapIn, pDevice->playback.channelMap, sizeof(pDevice->playback.channelMap)); + data.shareMode = pDevice->playback.shareMode; + data.performanceProfile = pDevice->coreaudio.originalPerformanceProfile; + data.registerStopEvent = (pDevice->type != ma_device_type_duplex); + + if (disposePreviousAudioUnit) { + ((ma_AudioOutputUnitStop_proc)pDevice->pContext->coreaudio.AudioOutputUnitStop)((AudioUnit)pDevice->coreaudio.audioUnitPlayback); + ((ma_AudioComponentInstanceDispose_proc)pDevice->pContext->coreaudio.AudioComponentInstanceDispose)((AudioUnit)pDevice->coreaudio.audioUnitPlayback); + } + } + data.periodSizeInFramesIn = pDevice->coreaudio.originalPeriodSizeInFrames; + data.periodSizeInMillisecondsIn = pDevice->coreaudio.originalPeriodSizeInMilliseconds; + data.periodsIn = pDevice->coreaudio.originalPeriods; + + /* Need at least 3 periods for duplex. */ + if (data.periodsIn < 3 && pDevice->type == ma_device_type_duplex) { + data.periodsIn = 3; + } + + result = ma_device_init_internal__coreaudio(pDevice->pContext, deviceType, NULL, &data, (void*)pDevice); + if (result != MA_SUCCESS) { + return result; + } + + if (deviceType == ma_device_type_capture) { + #if defined(MA_APPLE_DESKTOP) + pDevice->coreaudio.deviceObjectIDCapture = (ma_uint32)data.deviceObjectID; + ma_get_AudioObject_uid(pDevice->pContext, pDevice->coreaudio.deviceObjectIDCapture, sizeof(pDevice->capture.id.coreaudio), pDevice->capture.id.coreaudio); + #endif + pDevice->coreaudio.audioUnitCapture = (ma_ptr)data.audioUnit; + pDevice->coreaudio.pAudioBufferList = (ma_ptr)data.pAudioBufferList; + pDevice->coreaudio.audioBufferCapInFrames = data.periodSizeInFramesOut; + + pDevice->capture.internalFormat = data.formatOut; + pDevice->capture.internalChannels = data.channelsOut; + pDevice->capture.internalSampleRate = data.sampleRateOut; + MA_COPY_MEMORY(pDevice->capture.internalChannelMap, data.channelMapOut, sizeof(data.channelMapOut)); + pDevice->capture.internalPeriodSizeInFrames = data.periodSizeInFramesOut; + pDevice->capture.internalPeriods = data.periodsOut; + } else if (deviceType == ma_device_type_playback) { + #if defined(MA_APPLE_DESKTOP) + pDevice->coreaudio.deviceObjectIDPlayback = (ma_uint32)data.deviceObjectID; + ma_get_AudioObject_uid(pDevice->pContext, pDevice->coreaudio.deviceObjectIDPlayback, sizeof(pDevice->playback.id.coreaudio), pDevice->playback.id.coreaudio); + #endif + pDevice->coreaudio.audioUnitPlayback = (ma_ptr)data.audioUnit; + + pDevice->playback.internalFormat = data.formatOut; + pDevice->playback.internalChannels = data.channelsOut; + pDevice->playback.internalSampleRate = data.sampleRateOut; + MA_COPY_MEMORY(pDevice->playback.internalChannelMap, data.channelMapOut, sizeof(data.channelMapOut)); + pDevice->playback.internalPeriodSizeInFrames = data.periodSizeInFramesOut; + pDevice->playback.internalPeriods = data.periodsOut; + } + + return MA_SUCCESS; +} +#endif /* MA_APPLE_DESKTOP */ + +static ma_result ma_device_init__coreaudio(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pConfig != NULL); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + /* No exclusive mode with the Core Audio backend for now. */ + if (((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) && pDescriptorCapture->shareMode == ma_share_mode_exclusive) || + ((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pDescriptorPlayback->shareMode == ma_share_mode_exclusive)) { + return MA_SHARE_MODE_NOT_SUPPORTED; + } + + /* Capture needs to be initialized first. */ + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ma_device_init_internal_data__coreaudio data; + data.allowNominalSampleRateChange = pConfig->coreaudio.allowNominalSampleRateChange; + data.formatIn = pDescriptorCapture->format; + data.channelsIn = pDescriptorCapture->channels; + data.sampleRateIn = pDescriptorCapture->sampleRate; + MA_COPY_MEMORY(data.channelMapIn, pDescriptorCapture->channelMap, sizeof(pDescriptorCapture->channelMap)); + data.periodSizeInFramesIn = pDescriptorCapture->periodSizeInFrames; + data.periodSizeInMillisecondsIn = pDescriptorCapture->periodSizeInMilliseconds; + data.periodsIn = pDescriptorCapture->periodCount; + data.shareMode = pDescriptorCapture->shareMode; + data.performanceProfile = pConfig->performanceProfile; + data.registerStopEvent = MA_TRUE; + + /* Need at least 3 periods for duplex. */ + if (data.periodsIn < 3 && pConfig->deviceType == ma_device_type_duplex) { + data.periodsIn = 3; + } + + result = ma_device_init_internal__coreaudio(pDevice->pContext, ma_device_type_capture, pDescriptorCapture->pDeviceID, &data, (void*)pDevice); + if (result != MA_SUCCESS) { + return result; + } + + pDevice->coreaudio.isDefaultCaptureDevice = (pConfig->capture.pDeviceID == NULL); + #if defined(MA_APPLE_DESKTOP) + pDevice->coreaudio.deviceObjectIDCapture = (ma_uint32)data.deviceObjectID; + #endif + pDevice->coreaudio.audioUnitCapture = (ma_ptr)data.audioUnit; + pDevice->coreaudio.pAudioBufferList = (ma_ptr)data.pAudioBufferList; + pDevice->coreaudio.audioBufferCapInFrames = data.periodSizeInFramesOut; + pDevice->coreaudio.originalPeriodSizeInFrames = pDescriptorCapture->periodSizeInFrames; + pDevice->coreaudio.originalPeriodSizeInMilliseconds = pDescriptorCapture->periodSizeInMilliseconds; + pDevice->coreaudio.originalPeriods = pDescriptorCapture->periodCount; + pDevice->coreaudio.originalPerformanceProfile = pConfig->performanceProfile; + + pDescriptorCapture->format = data.formatOut; + pDescriptorCapture->channels = data.channelsOut; + pDescriptorCapture->sampleRate = data.sampleRateOut; + MA_COPY_MEMORY(pDescriptorCapture->channelMap, data.channelMapOut, sizeof(data.channelMapOut)); + pDescriptorCapture->periodSizeInFrames = data.periodSizeInFramesOut; + pDescriptorCapture->periodCount = data.periodsOut; + + #if defined(MA_APPLE_DESKTOP) + ma_get_AudioObject_uid(pDevice->pContext, pDevice->coreaudio.deviceObjectIDCapture, sizeof(pDevice->capture.id.coreaudio), pDevice->capture.id.coreaudio); + + /* + If we are using the default device we'll need to listen for changes to the system's default device so we can seamlessly + switch the device in the background. + */ + if (pConfig->capture.pDeviceID == NULL) { + ma_device__track__coreaudio(pDevice); + } + #endif + } + + /* Playback. */ + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_device_init_internal_data__coreaudio data; + data.allowNominalSampleRateChange = pConfig->coreaudio.allowNominalSampleRateChange; + data.formatIn = pDescriptorPlayback->format; + data.channelsIn = pDescriptorPlayback->channels; + data.sampleRateIn = pDescriptorPlayback->sampleRate; + MA_COPY_MEMORY(data.channelMapIn, pDescriptorPlayback->channelMap, sizeof(pDescriptorPlayback->channelMap)); + data.shareMode = pDescriptorPlayback->shareMode; + data.performanceProfile = pConfig->performanceProfile; + + /* In full-duplex mode we want the playback buffer to be the same size as the capture buffer. */ + if (pConfig->deviceType == ma_device_type_duplex) { + data.periodSizeInFramesIn = pDescriptorCapture->periodSizeInFrames; + data.periodsIn = pDescriptorCapture->periodCount; + data.registerStopEvent = MA_FALSE; + } else { + data.periodSizeInFramesIn = pDescriptorPlayback->periodSizeInFrames; + data.periodSizeInMillisecondsIn = pDescriptorPlayback->periodSizeInMilliseconds; + data.periodsIn = pDescriptorPlayback->periodCount; + data.registerStopEvent = MA_TRUE; + } + + result = ma_device_init_internal__coreaudio(pDevice->pContext, ma_device_type_playback, pDescriptorPlayback->pDeviceID, &data, (void*)pDevice); + if (result != MA_SUCCESS) { + if (pConfig->deviceType == ma_device_type_duplex) { + ((ma_AudioComponentInstanceDispose_proc)pDevice->pContext->coreaudio.AudioComponentInstanceDispose)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + if (pDevice->coreaudio.pAudioBufferList) { + ma_free(pDevice->coreaudio.pAudioBufferList, &pDevice->pContext->allocationCallbacks); + } + } + return result; + } + + pDevice->coreaudio.isDefaultPlaybackDevice = (pConfig->playback.pDeviceID == NULL); + #if defined(MA_APPLE_DESKTOP) + pDevice->coreaudio.deviceObjectIDPlayback = (ma_uint32)data.deviceObjectID; + #endif + pDevice->coreaudio.audioUnitPlayback = (ma_ptr)data.audioUnit; + pDevice->coreaudio.originalPeriodSizeInFrames = pDescriptorPlayback->periodSizeInFrames; + pDevice->coreaudio.originalPeriodSizeInMilliseconds = pDescriptorPlayback->periodSizeInMilliseconds; + pDevice->coreaudio.originalPeriods = pDescriptorPlayback->periodCount; + pDevice->coreaudio.originalPerformanceProfile = pConfig->performanceProfile; + + pDescriptorPlayback->format = data.formatOut; + pDescriptorPlayback->channels = data.channelsOut; + pDescriptorPlayback->sampleRate = data.sampleRateOut; + MA_COPY_MEMORY(pDescriptorPlayback->channelMap, data.channelMapOut, sizeof(data.channelMapOut)); + pDescriptorPlayback->periodSizeInFrames = data.periodSizeInFramesOut; + pDescriptorPlayback->periodCount = data.periodsOut; + + #if defined(MA_APPLE_DESKTOP) + ma_get_AudioObject_uid(pDevice->pContext, pDevice->coreaudio.deviceObjectIDPlayback, sizeof(pDevice->playback.id.coreaudio), pDevice->playback.id.coreaudio); + + /* + If we are using the default device we'll need to listen for changes to the system's default device so we can seamlessly + switch the device in the background. + */ + if (pDescriptorPlayback->pDeviceID == NULL && (pConfig->deviceType != ma_device_type_duplex || pDescriptorCapture->pDeviceID != NULL)) { + ma_device__track__coreaudio(pDevice); + } + #endif + } + + + + /* + When stopping the device, a callback is called on another thread. We need to wait for this callback + before returning from ma_device_stop(). This event is used for this. + */ + ma_event_init(&pDevice->coreaudio.stopEvent); + + /* + We need to detect when a route has changed so we can update the data conversion pipeline accordingly. This is done + differently on non-Desktop Apple platforms. + */ +#if defined(MA_APPLE_MOBILE) + pDevice->coreaudio.pNotificationHandler = (MA_BRIDGE_RETAINED void*)[[ma_ios_notification_handler alloc] init:pDevice]; +#endif + + return MA_SUCCESS; +} + + +static ma_result ma_device_start__coreaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + OSStatus status = ((ma_AudioOutputUnitStart_proc)pDevice->pContext->coreaudio.AudioOutputUnitStart)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + OSStatus status = ((ma_AudioOutputUnitStart_proc)pDevice->pContext->coreaudio.AudioOutputUnitStart)((AudioUnit)pDevice->coreaudio.audioUnitPlayback); + if (status != noErr) { + if (pDevice->type == ma_device_type_duplex) { + ((ma_AudioOutputUnitStop_proc)pDevice->pContext->coreaudio.AudioOutputUnitStop)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + } + return ma_result_from_OSStatus(status); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__coreaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + /* It's not clear from the documentation whether or not AudioOutputUnitStop() actually drains the device or not. */ + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + OSStatus status = ((ma_AudioOutputUnitStop_proc)pDevice->pContext->coreaudio.AudioOutputUnitStop)((AudioUnit)pDevice->coreaudio.audioUnitCapture); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + OSStatus status = ((ma_AudioOutputUnitStop_proc)pDevice->pContext->coreaudio.AudioOutputUnitStop)((AudioUnit)pDevice->coreaudio.audioUnitPlayback); + if (status != noErr) { + return ma_result_from_OSStatus(status); + } + } + + /* We need to wait for the callback to finish before returning. */ + ma_event_wait(&pDevice->coreaudio.stopEvent); + return MA_SUCCESS; +} + + +static ma_result ma_context_uninit__coreaudio(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_coreaudio); + +#if defined(MA_APPLE_MOBILE) + if (!pContext->coreaudio.noAudioSessionDeactivate) { + if (![[AVAudioSession sharedInstance] setActive:false error:nil]) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "Failed to deactivate audio session."); + return MA_FAILED_TO_INIT_BACKEND; + } + } +#endif + +#if !defined(MA_NO_RUNTIME_LINKING) && !defined(MA_APPLE_MOBILE) + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit); + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio); + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreFoundation); +#endif + +#if !defined(MA_APPLE_MOBILE) + ma_context__uninit_device_tracking__coreaudio(pContext); +#endif + + (void)pContext; + return MA_SUCCESS; +} + +#if defined(MA_APPLE_MOBILE) && defined(__IPHONE_12_0) +static AVAudioSessionCategory ma_to_AVAudioSessionCategory(ma_ios_session_category category) +{ + /* The "default" and "none" categories are treated different and should not be used as an input into this function. */ + MA_ASSERT(category != ma_ios_session_category_default); + MA_ASSERT(category != ma_ios_session_category_none); + + switch (category) { + case ma_ios_session_category_ambient: return AVAudioSessionCategoryAmbient; + case ma_ios_session_category_solo_ambient: return AVAudioSessionCategorySoloAmbient; + case ma_ios_session_category_playback: return AVAudioSessionCategoryPlayback; + case ma_ios_session_category_record: return AVAudioSessionCategoryRecord; + case ma_ios_session_category_play_and_record: return AVAudioSessionCategoryPlayAndRecord; + case ma_ios_session_category_multi_route: return AVAudioSessionCategoryMultiRoute; + case ma_ios_session_category_none: return AVAudioSessionCategoryAmbient; + case ma_ios_session_category_default: return AVAudioSessionCategoryAmbient; + default: return AVAudioSessionCategoryAmbient; + } +} +#endif + +static ma_result ma_context_init__coreaudio(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ +#if !defined(MA_APPLE_MOBILE) + ma_result result; +#endif + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pContext != NULL); + +#if defined(MA_APPLE_MOBILE) + @autoreleasepool { + AVAudioSession* pAudioSession = [AVAudioSession sharedInstance]; + AVAudioSessionCategoryOptions options = pConfig->coreaudio.sessionCategoryOptions; + + MA_ASSERT(pAudioSession != NULL); + + if (pConfig->coreaudio.sessionCategory == ma_ios_session_category_default) { + /* + I'm going to use trial and error to determine our default session category. First we'll try PlayAndRecord. If that fails + we'll try Playback and if that fails we'll try record. If all of these fail we'll just not set the category. + */ + #if !defined(MA_APPLE_TV) && !defined(MA_APPLE_WATCH) + options |= AVAudioSessionCategoryOptionDefaultToSpeaker; + #endif + + if ([pAudioSession setCategory: AVAudioSessionCategoryPlayAndRecord withOptions:options error:nil]) { + /* Using PlayAndRecord */ + } else if ([pAudioSession setCategory: AVAudioSessionCategoryPlayback withOptions:options error:nil]) { + /* Using Playback */ + } else if ([pAudioSession setCategory: AVAudioSessionCategoryRecord withOptions:options error:nil]) { + /* Using Record */ + } else { + /* Leave as default? */ + } + } else { + if (pConfig->coreaudio.sessionCategory != ma_ios_session_category_none) { + #if defined(__IPHONE_12_0) + if (![pAudioSession setCategory: ma_to_AVAudioSessionCategory(pConfig->coreaudio.sessionCategory) withOptions:options error:nil]) { + return MA_INVALID_OPERATION; /* Failed to set session category. */ + } + #else + /* Ignore the session category on version 11 and older, but post a warning. */ + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_WARNING, "Session category only supported in iOS 12 and newer."); + #endif + } + } + + if (!pConfig->coreaudio.noAudioSessionActivate) { + if (![pAudioSession setActive:true error:nil]) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "Failed to activate audio session."); + return MA_FAILED_TO_INIT_BACKEND; + } + } + } +#endif + +#if !defined(MA_NO_RUNTIME_LINKING) && !defined(MA_APPLE_MOBILE) + pContext->coreaudio.hCoreFoundation = ma_dlopen(ma_context_get_log(pContext), "/System/Library/Frameworks/CoreFoundation.framework/CoreFoundation"); + if (pContext->coreaudio.hCoreFoundation == NULL) { + return MA_API_NOT_FOUND; + } + + pContext->coreaudio.CFStringGetCString = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hCoreFoundation, "CFStringGetCString"); + pContext->coreaudio.CFRelease = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hCoreFoundation, "CFRelease"); + + + pContext->coreaudio.hCoreAudio = ma_dlopen(ma_context_get_log(pContext), "/System/Library/Frameworks/CoreAudio.framework/CoreAudio"); + if (pContext->coreaudio.hCoreAudio == NULL) { + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreFoundation); + return MA_API_NOT_FOUND; + } + + pContext->coreaudio.AudioObjectGetPropertyData = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio, "AudioObjectGetPropertyData"); + pContext->coreaudio.AudioObjectGetPropertyDataSize = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio, "AudioObjectGetPropertyDataSize"); + pContext->coreaudio.AudioObjectSetPropertyData = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio, "AudioObjectSetPropertyData"); + pContext->coreaudio.AudioObjectAddPropertyListener = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio, "AudioObjectAddPropertyListener"); + pContext->coreaudio.AudioObjectRemovePropertyListener = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio, "AudioObjectRemovePropertyListener"); + + /* + It looks like Apple has moved some APIs from AudioUnit into AudioToolbox on more recent versions of macOS. They are still + defined in AudioUnit, but just in case they decide to remove them from there entirely I'm going to implement a fallback. + The way it'll work is that it'll first try AudioUnit, and if the required symbols are not present there we'll fall back to + AudioToolbox. + */ + pContext->coreaudio.hAudioUnit = ma_dlopen(ma_context_get_log(pContext), "/System/Library/Frameworks/AudioUnit.framework/AudioUnit"); + if (pContext->coreaudio.hAudioUnit == NULL) { + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio); + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreFoundation); + return MA_API_NOT_FOUND; + } + + if (ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioComponentFindNext") == NULL) { + /* Couldn't find the required symbols in AudioUnit, so fall back to AudioToolbox. */ + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit); + pContext->coreaudio.hAudioUnit = ma_dlopen(ma_context_get_log(pContext), "/System/Library/Frameworks/AudioToolbox.framework/AudioToolbox"); + if (pContext->coreaudio.hAudioUnit == NULL) { + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio); + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreFoundation); + return MA_API_NOT_FOUND; + } + } + + pContext->coreaudio.AudioComponentFindNext = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioComponentFindNext"); + pContext->coreaudio.AudioComponentInstanceDispose = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioComponentInstanceDispose"); + pContext->coreaudio.AudioComponentInstanceNew = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioComponentInstanceNew"); + pContext->coreaudio.AudioOutputUnitStart = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioOutputUnitStart"); + pContext->coreaudio.AudioOutputUnitStop = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioOutputUnitStop"); + pContext->coreaudio.AudioUnitAddPropertyListener = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioUnitAddPropertyListener"); + pContext->coreaudio.AudioUnitGetPropertyInfo = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioUnitGetPropertyInfo"); + pContext->coreaudio.AudioUnitGetProperty = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioUnitGetProperty"); + pContext->coreaudio.AudioUnitSetProperty = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioUnitSetProperty"); + pContext->coreaudio.AudioUnitInitialize = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioUnitInitialize"); + pContext->coreaudio.AudioUnitRender = ma_dlsym(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit, "AudioUnitRender"); +#else + pContext->coreaudio.CFStringGetCString = (ma_proc)CFStringGetCString; + pContext->coreaudio.CFRelease = (ma_proc)CFRelease; + + #if defined(MA_APPLE_DESKTOP) + pContext->coreaudio.AudioObjectGetPropertyData = (ma_proc)AudioObjectGetPropertyData; + pContext->coreaudio.AudioObjectGetPropertyDataSize = (ma_proc)AudioObjectGetPropertyDataSize; + pContext->coreaudio.AudioObjectSetPropertyData = (ma_proc)AudioObjectSetPropertyData; + pContext->coreaudio.AudioObjectAddPropertyListener = (ma_proc)AudioObjectAddPropertyListener; + pContext->coreaudio.AudioObjectRemovePropertyListener = (ma_proc)AudioObjectRemovePropertyListener; + #endif + + pContext->coreaudio.AudioComponentFindNext = (ma_proc)AudioComponentFindNext; + pContext->coreaudio.AudioComponentInstanceDispose = (ma_proc)AudioComponentInstanceDispose; + pContext->coreaudio.AudioComponentInstanceNew = (ma_proc)AudioComponentInstanceNew; + pContext->coreaudio.AudioOutputUnitStart = (ma_proc)AudioOutputUnitStart; + pContext->coreaudio.AudioOutputUnitStop = (ma_proc)AudioOutputUnitStop; + pContext->coreaudio.AudioUnitAddPropertyListener = (ma_proc)AudioUnitAddPropertyListener; + pContext->coreaudio.AudioUnitGetPropertyInfo = (ma_proc)AudioUnitGetPropertyInfo; + pContext->coreaudio.AudioUnitGetProperty = (ma_proc)AudioUnitGetProperty; + pContext->coreaudio.AudioUnitSetProperty = (ma_proc)AudioUnitSetProperty; + pContext->coreaudio.AudioUnitInitialize = (ma_proc)AudioUnitInitialize; + pContext->coreaudio.AudioUnitRender = (ma_proc)AudioUnitRender; +#endif + + /* Audio component. */ + { + AudioComponentDescription desc; + desc.componentType = kAudioUnitType_Output; + #if defined(MA_APPLE_DESKTOP) + desc.componentSubType = kAudioUnitSubType_HALOutput; + #else + desc.componentSubType = kAudioUnitSubType_RemoteIO; + #endif + desc.componentManufacturer = kAudioUnitManufacturer_Apple; + desc.componentFlags = 0; + desc.componentFlagsMask = 0; + + pContext->coreaudio.component = ((ma_AudioComponentFindNext_proc)pContext->coreaudio.AudioComponentFindNext)(NULL, &desc); + if (pContext->coreaudio.component == NULL) { + #if !defined(MA_NO_RUNTIME_LINKING) && !defined(MA_APPLE_MOBILE) + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit); + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio); + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreFoundation); + #endif + return MA_FAILED_TO_INIT_BACKEND; + } + } + +#if !defined(MA_APPLE_MOBILE) + result = ma_context__init_device_tracking__coreaudio(pContext); + if (result != MA_SUCCESS) { + #if !defined(MA_NO_RUNTIME_LINKING) && !defined(MA_APPLE_MOBILE) + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hAudioUnit); + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreAudio); + ma_dlclose(ma_context_get_log(pContext), pContext->coreaudio.hCoreFoundation); + #endif + return result; + } +#endif + + pContext->coreaudio.noAudioSessionDeactivate = pConfig->coreaudio.noAudioSessionDeactivate; + + pCallbacks->onContextInit = ma_context_init__coreaudio; + pCallbacks->onContextUninit = ma_context_uninit__coreaudio; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__coreaudio; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__coreaudio; + pCallbacks->onDeviceInit = ma_device_init__coreaudio; + pCallbacks->onDeviceUninit = ma_device_uninit__coreaudio; + pCallbacks->onDeviceStart = ma_device_start__coreaudio; + pCallbacks->onDeviceStop = ma_device_stop__coreaudio; + pCallbacks->onDeviceRead = NULL; + pCallbacks->onDeviceWrite = NULL; + pCallbacks->onDeviceDataLoop = NULL; + + return MA_SUCCESS; +} +#endif /* MA_HAS_COREAUDIO */ + + + +/****************************************************************************** + +sndio Backend + +******************************************************************************/ +#ifdef MA_HAS_SNDIO +#include + +/* +Only supporting OpenBSD. This did not work very well at all on FreeBSD when I tried it. Not sure if this is due +to miniaudio's implementation or if it's some kind of system configuration issue, but basically the default device +just doesn't emit any sound, or at times you'll hear tiny pieces. I will consider enabling this when there's +demand for it or if I can get it tested and debugged more thoroughly. +*/ +#if 0 +#if defined(__NetBSD__) || defined(__OpenBSD__) +#include +#endif +#if defined(__FreeBSD__) || defined(__DragonFly__) +#include +#endif +#endif + +#define MA_SIO_DEVANY "default" +#define MA_SIO_PLAY 1 +#define MA_SIO_REC 2 +#define MA_SIO_NENC 8 +#define MA_SIO_NCHAN 8 +#define MA_SIO_NRATE 16 +#define MA_SIO_NCONF 4 + +struct ma_sio_hdl; /* <-- Opaque */ + +struct ma_sio_par +{ + unsigned int bits; + unsigned int bps; + unsigned int sig; + unsigned int le; + unsigned int msb; + unsigned int rchan; + unsigned int pchan; + unsigned int rate; + unsigned int bufsz; + unsigned int xrun; + unsigned int round; + unsigned int appbufsz; + int __pad[3]; + unsigned int __magic; +}; + +struct ma_sio_enc +{ + unsigned int bits; + unsigned int bps; + unsigned int sig; + unsigned int le; + unsigned int msb; +}; + +struct ma_sio_conf +{ + unsigned int enc; + unsigned int rchan; + unsigned int pchan; + unsigned int rate; +}; + +struct ma_sio_cap +{ + struct ma_sio_enc enc[MA_SIO_NENC]; + unsigned int rchan[MA_SIO_NCHAN]; + unsigned int pchan[MA_SIO_NCHAN]; + unsigned int rate[MA_SIO_NRATE]; + int __pad[7]; + unsigned int nconf; + struct ma_sio_conf confs[MA_SIO_NCONF]; +}; + +typedef struct ma_sio_hdl* (* ma_sio_open_proc) (const char*, unsigned int, int); +typedef void (* ma_sio_close_proc) (struct ma_sio_hdl*); +typedef int (* ma_sio_setpar_proc) (struct ma_sio_hdl*, struct ma_sio_par*); +typedef int (* ma_sio_getpar_proc) (struct ma_sio_hdl*, struct ma_sio_par*); +typedef int (* ma_sio_getcap_proc) (struct ma_sio_hdl*, struct ma_sio_cap*); +typedef size_t (* ma_sio_write_proc) (struct ma_sio_hdl*, const void*, size_t); +typedef size_t (* ma_sio_read_proc) (struct ma_sio_hdl*, void*, size_t); +typedef int (* ma_sio_start_proc) (struct ma_sio_hdl*); +typedef int (* ma_sio_stop_proc) (struct ma_sio_hdl*); +typedef int (* ma_sio_initpar_proc)(struct ma_sio_par*); + +static ma_uint32 ma_get_standard_sample_rate_priority_index__sndio(ma_uint32 sampleRate) /* Lower = higher priority */ +{ + ma_uint32 i; + for (i = 0; i < ma_countof(g_maStandardSampleRatePriorities); ++i) { + if (g_maStandardSampleRatePriorities[i] == sampleRate) { + return i; + } + } + + return (ma_uint32)-1; +} + +static ma_format ma_format_from_sio_enc__sndio(unsigned int bits, unsigned int bps, unsigned int sig, unsigned int le, unsigned int msb) +{ + /* We only support native-endian right now. */ + if ((ma_is_little_endian() && le == 0) || (ma_is_big_endian() && le == 1)) { + return ma_format_unknown; + } + + if (bits == 8 && bps == 1 && sig == 0) { + return ma_format_u8; + } + if (bits == 16 && bps == 2 && sig == 1) { + return ma_format_s16; + } + if (bits == 24 && bps == 3 && sig == 1) { + return ma_format_s24; + } + if (bits == 24 && bps == 4 && sig == 1 && msb == 0) { + /*return ma_format_s24_32;*/ + } + if (bits == 32 && bps == 4 && sig == 1) { + return ma_format_s32; + } + + return ma_format_unknown; +} + +static ma_format ma_find_best_format_from_sio_cap__sndio(struct ma_sio_cap* caps) +{ + ma_format bestFormat; + unsigned int iConfig; + + MA_ASSERT(caps != NULL); + + bestFormat = ma_format_unknown; + for (iConfig = 0; iConfig < caps->nconf; iConfig += 1) { + unsigned int iEncoding; + for (iEncoding = 0; iEncoding < MA_SIO_NENC; iEncoding += 1) { + unsigned int bits; + unsigned int bps; + unsigned int sig; + unsigned int le; + unsigned int msb; + ma_format format; + + if ((caps->confs[iConfig].enc & (1UL << iEncoding)) == 0) { + continue; + } + + bits = caps->enc[iEncoding].bits; + bps = caps->enc[iEncoding].bps; + sig = caps->enc[iEncoding].sig; + le = caps->enc[iEncoding].le; + msb = caps->enc[iEncoding].msb; + format = ma_format_from_sio_enc__sndio(bits, bps, sig, le, msb); + if (format == ma_format_unknown) { + continue; /* Format not supported. */ + } + + if (bestFormat == ma_format_unknown) { + bestFormat = format; + } else { + if (ma_get_format_priority_index(bestFormat) > ma_get_format_priority_index(format)) { /* <-- Lower = better. */ + bestFormat = format; + } + } + } + } + + return bestFormat; +} + +static ma_uint32 ma_find_best_channels_from_sio_cap__sndio(struct ma_sio_cap* caps, ma_device_type deviceType, ma_format requiredFormat) +{ + ma_uint32 maxChannels; + unsigned int iConfig; + + MA_ASSERT(caps != NULL); + MA_ASSERT(requiredFormat != ma_format_unknown); + + /* Just pick whatever configuration has the most channels. */ + maxChannels = 0; + for (iConfig = 0; iConfig < caps->nconf; iConfig += 1) { + /* The encoding should be of requiredFormat. */ + unsigned int iEncoding; + for (iEncoding = 0; iEncoding < MA_SIO_NENC; iEncoding += 1) { + unsigned int iChannel; + unsigned int bits; + unsigned int bps; + unsigned int sig; + unsigned int le; + unsigned int msb; + ma_format format; + + if ((caps->confs[iConfig].enc & (1UL << iEncoding)) == 0) { + continue; + } + + bits = caps->enc[iEncoding].bits; + bps = caps->enc[iEncoding].bps; + sig = caps->enc[iEncoding].sig; + le = caps->enc[iEncoding].le; + msb = caps->enc[iEncoding].msb; + format = ma_format_from_sio_enc__sndio(bits, bps, sig, le, msb); + if (format != requiredFormat) { + continue; + } + + /* Getting here means the format is supported. Iterate over each channel count and grab the biggest one. */ + for (iChannel = 0; iChannel < MA_SIO_NCHAN; iChannel += 1) { + unsigned int chan = 0; + unsigned int channels; + + if (deviceType == ma_device_type_playback) { + chan = caps->confs[iConfig].pchan; + } else { + chan = caps->confs[iConfig].rchan; + } + + if ((chan & (1UL << iChannel)) == 0) { + continue; + } + + if (deviceType == ma_device_type_playback) { + channels = caps->pchan[iChannel]; + } else { + channels = caps->rchan[iChannel]; + } + + if (maxChannels < channels) { + maxChannels = channels; + } + } + } + } + + return maxChannels; +} + +static ma_uint32 ma_find_best_sample_rate_from_sio_cap__sndio(struct ma_sio_cap* caps, ma_device_type deviceType, ma_format requiredFormat, ma_uint32 requiredChannels) +{ + ma_uint32 firstSampleRate; + ma_uint32 bestSampleRate; + unsigned int iConfig; + + MA_ASSERT(caps != NULL); + MA_ASSERT(requiredFormat != ma_format_unknown); + MA_ASSERT(requiredChannels > 0); + MA_ASSERT(requiredChannels <= MA_MAX_CHANNELS); + + firstSampleRate = 0; /* <-- If the device does not support a standard rate we'll fall back to the first one that's found. */ + bestSampleRate = 0; + + for (iConfig = 0; iConfig < caps->nconf; iConfig += 1) { + /* The encoding should be of requiredFormat. */ + unsigned int iEncoding; + for (iEncoding = 0; iEncoding < MA_SIO_NENC; iEncoding += 1) { + unsigned int iChannel; + unsigned int bits; + unsigned int bps; + unsigned int sig; + unsigned int le; + unsigned int msb; + ma_format format; + + if ((caps->confs[iConfig].enc & (1UL << iEncoding)) == 0) { + continue; + } + + bits = caps->enc[iEncoding].bits; + bps = caps->enc[iEncoding].bps; + sig = caps->enc[iEncoding].sig; + le = caps->enc[iEncoding].le; + msb = caps->enc[iEncoding].msb; + format = ma_format_from_sio_enc__sndio(bits, bps, sig, le, msb); + if (format != requiredFormat) { + continue; + } + + /* Getting here means the format is supported. Iterate over each channel count and grab the biggest one. */ + for (iChannel = 0; iChannel < MA_SIO_NCHAN; iChannel += 1) { + unsigned int chan = 0; + unsigned int channels; + unsigned int iRate; + + if (deviceType == ma_device_type_playback) { + chan = caps->confs[iConfig].pchan; + } else { + chan = caps->confs[iConfig].rchan; + } + + if ((chan & (1UL << iChannel)) == 0) { + continue; + } + + if (deviceType == ma_device_type_playback) { + channels = caps->pchan[iChannel]; + } else { + channels = caps->rchan[iChannel]; + } + + if (channels != requiredChannels) { + continue; + } + + /* Getting here means we have found a compatible encoding/channel pair. */ + for (iRate = 0; iRate < MA_SIO_NRATE; iRate += 1) { + ma_uint32 rate = (ma_uint32)caps->rate[iRate]; + ma_uint32 ratePriority; + + if (firstSampleRate == 0) { + firstSampleRate = rate; + } + + /* Disregard this rate if it's not a standard one. */ + ratePriority = ma_get_standard_sample_rate_priority_index__sndio(rate); + if (ratePriority == (ma_uint32)-1) { + continue; + } + + if (ma_get_standard_sample_rate_priority_index__sndio(bestSampleRate) > ratePriority) { /* Lower = better. */ + bestSampleRate = rate; + } + } + } + } + } + + /* If a standard sample rate was not found just fall back to the first one that was iterated. */ + if (bestSampleRate == 0) { + bestSampleRate = firstSampleRate; + } + + return bestSampleRate; +} + + +static ma_result ma_context_enumerate_devices__sndio(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_bool32 isTerminating = MA_FALSE; + struct ma_sio_hdl* handle; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + /* sndio doesn't seem to have a good device enumeration API, so I'm therefore only enumerating over default devices for now. */ + + /* Playback. */ + if (!isTerminating) { + handle = ((ma_sio_open_proc)pContext->sndio.sio_open)(MA_SIO_DEVANY, MA_SIO_PLAY, 0); + if (handle != NULL) { + /* Supports playback. */ + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strcpy_s(deviceInfo.id.sndio, sizeof(deviceInfo.id.sndio), MA_SIO_DEVANY); + ma_strcpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_PLAYBACK_DEVICE_NAME); + + isTerminating = !callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + + ((ma_sio_close_proc)pContext->sndio.sio_close)(handle); + } + } + + /* Capture. */ + if (!isTerminating) { + handle = ((ma_sio_open_proc)pContext->sndio.sio_open)(MA_SIO_DEVANY, MA_SIO_REC, 0); + if (handle != NULL) { + /* Supports capture. */ + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strcpy_s(deviceInfo.id.sndio, sizeof(deviceInfo.id.sndio), "default"); + ma_strcpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_CAPTURE_DEVICE_NAME); + + isTerminating = !callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + + ((ma_sio_close_proc)pContext->sndio.sio_close)(handle); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info__sndio(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + char devid[256]; + struct ma_sio_hdl* handle; + struct ma_sio_cap caps; + unsigned int iConfig; + + MA_ASSERT(pContext != NULL); + + /* We need to open the device before we can get information about it. */ + if (pDeviceID == NULL) { + ma_strcpy_s(devid, sizeof(devid), MA_SIO_DEVANY); + ma_strcpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), (deviceType == ma_device_type_playback) ? MA_DEFAULT_PLAYBACK_DEVICE_NAME : MA_DEFAULT_CAPTURE_DEVICE_NAME); + } else { + ma_strcpy_s(devid, sizeof(devid), pDeviceID->sndio); + ma_strcpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), devid); + } + + handle = ((ma_sio_open_proc)pContext->sndio.sio_open)(devid, (deviceType == ma_device_type_playback) ? MA_SIO_PLAY : MA_SIO_REC, 0); + if (handle == NULL) { + return MA_NO_DEVICE; + } + + if (((ma_sio_getcap_proc)pContext->sndio.sio_getcap)(handle, &caps) == 0) { + return MA_ERROR; + } + + pDeviceInfo->nativeDataFormatCount = 0; + + for (iConfig = 0; iConfig < caps.nconf; iConfig += 1) { + /* + The main thing we care about is that the encoding is supported by miniaudio. If it is, we want to give + preference to some formats over others. + */ + unsigned int iEncoding; + unsigned int iChannel; + unsigned int iRate; + + for (iEncoding = 0; iEncoding < MA_SIO_NENC; iEncoding += 1) { + unsigned int bits; + unsigned int bps; + unsigned int sig; + unsigned int le; + unsigned int msb; + ma_format format; + + if ((caps.confs[iConfig].enc & (1UL << iEncoding)) == 0) { + continue; + } + + bits = caps.enc[iEncoding].bits; + bps = caps.enc[iEncoding].bps; + sig = caps.enc[iEncoding].sig; + le = caps.enc[iEncoding].le; + msb = caps.enc[iEncoding].msb; + format = ma_format_from_sio_enc__sndio(bits, bps, sig, le, msb); + if (format == ma_format_unknown) { + continue; /* Format not supported. */ + } + + + /* Channels. */ + for (iChannel = 0; iChannel < MA_SIO_NCHAN; iChannel += 1) { + unsigned int chan = 0; + unsigned int channels; + + if (deviceType == ma_device_type_playback) { + chan = caps.confs[iConfig].pchan; + } else { + chan = caps.confs[iConfig].rchan; + } + + if ((chan & (1UL << iChannel)) == 0) { + continue; + } + + if (deviceType == ma_device_type_playback) { + channels = caps.pchan[iChannel]; + } else { + channels = caps.rchan[iChannel]; + } + + + /* Sample Rates. */ + for (iRate = 0; iRate < MA_SIO_NRATE; iRate += 1) { + if ((caps.confs[iConfig].rate & (1UL << iRate)) != 0) { + ma_device_info_add_native_data_format(pDeviceInfo, format, channels, caps.rate[iRate], 0); + } + } + } + } + } + + ((ma_sio_close_proc)pContext->sndio.sio_close)(handle); + return MA_SUCCESS; +} + +static ma_result ma_device_uninit__sndio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ((ma_sio_close_proc)pDevice->pContext->sndio.sio_close)((struct ma_sio_hdl*)pDevice->sndio.handleCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ((ma_sio_close_proc)pDevice->pContext->sndio.sio_close)((struct ma_sio_hdl*)pDevice->sndio.handlePlayback); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_init_handle__sndio(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptor, ma_device_type deviceType) +{ + const char* pDeviceName; + ma_ptr handle; + int openFlags = 0; + struct ma_sio_cap caps; + struct ma_sio_par par; + const ma_device_id* pDeviceID; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_format internalFormat; + ma_uint32 internalChannels; + ma_uint32 internalSampleRate; + ma_uint32 internalPeriodSizeInFrames; + ma_uint32 internalPeriods; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(deviceType != ma_device_type_duplex); + MA_ASSERT(pDevice != NULL); + + if (deviceType == ma_device_type_capture) { + openFlags = MA_SIO_REC; + } else { + openFlags = MA_SIO_PLAY; + } + + pDeviceID = pDescriptor->pDeviceID; + format = pDescriptor->format; + channels = pDescriptor->channels; + sampleRate = pDescriptor->sampleRate; + + pDeviceName = MA_SIO_DEVANY; + if (pDeviceID != NULL) { + pDeviceName = pDeviceID->sndio; + } + + handle = (ma_ptr)((ma_sio_open_proc)pDevice->pContext->sndio.sio_open)(pDeviceName, openFlags, 0); + if (handle == NULL) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[sndio] Failed to open device."); + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + + /* We need to retrieve the device caps to determine the most appropriate format to use. */ + if (((ma_sio_getcap_proc)pDevice->pContext->sndio.sio_getcap)((struct ma_sio_hdl*)handle, &caps) == 0) { + ((ma_sio_close_proc)pDevice->pContext->sndio.sio_close)((struct ma_sio_hdl*)handle); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[sndio] Failed to retrieve device caps."); + return MA_ERROR; + } + + /* + Note: sndio reports a huge range of available channels. This is inconvenient for us because there's no real + way, as far as I can tell, to get the _actual_ channel count of the device. I'm therefore restricting this + to the requested channels, regardless of whether or not the default channel count is requested. + + For hardware devices, I'm suspecting only a single channel count will be reported and we can safely use the + value returned by ma_find_best_channels_from_sio_cap__sndio(). + */ + if (deviceType == ma_device_type_capture) { + if (format == ma_format_unknown) { + format = ma_find_best_format_from_sio_cap__sndio(&caps); + } + + if (channels == 0) { + if (strlen(pDeviceName) > strlen("rsnd/") && strncmp(pDeviceName, "rsnd/", strlen("rsnd/")) == 0) { + channels = ma_find_best_channels_from_sio_cap__sndio(&caps, deviceType, format); + } else { + channels = MA_DEFAULT_CHANNELS; + } + } + } else { + if (format == ma_format_unknown) { + format = ma_find_best_format_from_sio_cap__sndio(&caps); + } + + if (channels == 0) { + if (strlen(pDeviceName) > strlen("rsnd/") && strncmp(pDeviceName, "rsnd/", strlen("rsnd/")) == 0) { + channels = ma_find_best_channels_from_sio_cap__sndio(&caps, deviceType, format); + } else { + channels = MA_DEFAULT_CHANNELS; + } + } + } + + if (sampleRate == 0) { + sampleRate = ma_find_best_sample_rate_from_sio_cap__sndio(&caps, pConfig->deviceType, format, channels); + } + + + ((ma_sio_initpar_proc)pDevice->pContext->sndio.sio_initpar)(&par); + par.msb = 0; + par.le = ma_is_little_endian(); + + switch (format) { + case ma_format_u8: + { + par.bits = 8; + par.bps = 1; + par.sig = 0; + } break; + + case ma_format_s24: + { + par.bits = 24; + par.bps = 3; + par.sig = 1; + } break; + + case ma_format_s32: + { + par.bits = 32; + par.bps = 4; + par.sig = 1; + } break; + + case ma_format_s16: + case ma_format_f32: + case ma_format_unknown: + default: + { + par.bits = 16; + par.bps = 2; + par.sig = 1; + } break; + } + + if (deviceType == ma_device_type_capture) { + par.rchan = channels; + } else { + par.pchan = channels; + } + + par.rate = sampleRate; + + internalPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptor, par.rate, pConfig->performanceProfile); + + par.round = internalPeriodSizeInFrames; + par.appbufsz = par.round * pDescriptor->periodCount; + + if (((ma_sio_setpar_proc)pDevice->pContext->sndio.sio_setpar)((struct ma_sio_hdl*)handle, &par) == 0) { + ((ma_sio_close_proc)pDevice->pContext->sndio.sio_close)((struct ma_sio_hdl*)handle); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[sndio] Failed to set buffer size."); + return MA_ERROR; + } + + if (((ma_sio_getpar_proc)pDevice->pContext->sndio.sio_getpar)((struct ma_sio_hdl*)handle, &par) == 0) { + ((ma_sio_close_proc)pDevice->pContext->sndio.sio_close)((struct ma_sio_hdl*)handle); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[sndio] Failed to retrieve buffer size."); + return MA_ERROR; + } + + internalFormat = ma_format_from_sio_enc__sndio(par.bits, par.bps, par.sig, par.le, par.msb); + internalChannels = (deviceType == ma_device_type_capture) ? par.rchan : par.pchan; + internalSampleRate = par.rate; + internalPeriods = par.appbufsz / par.round; + internalPeriodSizeInFrames = par.round; + + if (deviceType == ma_device_type_capture) { + pDevice->sndio.handleCapture = handle; + } else { + pDevice->sndio.handlePlayback = handle; + } + + pDescriptor->format = internalFormat; + pDescriptor->channels = internalChannels; + pDescriptor->sampleRate = internalSampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_sndio, pDescriptor->channelMap, ma_countof(pDescriptor->channelMap), internalChannels); + pDescriptor->periodSizeInFrames = internalPeriodSizeInFrames; + pDescriptor->periodCount = internalPeriods; + + return MA_SUCCESS; +} + +static ma_result ma_device_init__sndio(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + MA_ASSERT(pDevice != NULL); + + MA_ZERO_OBJECT(&pDevice->sndio); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ma_result result = ma_device_init_handle__sndio(pDevice, pConfig, pDescriptorCapture, ma_device_type_capture); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_result result = ma_device_init_handle__sndio(pDevice, pConfig, pDescriptorPlayback, ma_device_type_playback); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_start__sndio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ((ma_sio_start_proc)pDevice->pContext->sndio.sio_start)((struct ma_sio_hdl*)pDevice->sndio.handleCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ((ma_sio_start_proc)pDevice->pContext->sndio.sio_start)((struct ma_sio_hdl*)pDevice->sndio.handlePlayback); /* <-- Doesn't actually playback until data is written. */ + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__sndio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + /* + From the documentation: + + The sio_stop() function puts the audio subsystem in the same state as before sio_start() is called. It stops recording, drains the play buffer and then + stops playback. If samples to play are queued but playback hasn't started yet then playback is forced immediately; playback will actually stop once the + buffer is drained. In no case are samples in the play buffer discarded. + + Therefore, sio_stop() performs all of the necessary draining for us. + */ + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ((ma_sio_stop_proc)pDevice->pContext->sndio.sio_stop)((struct ma_sio_hdl*)pDevice->sndio.handleCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ((ma_sio_stop_proc)pDevice->pContext->sndio.sio_stop)((struct ma_sio_hdl*)pDevice->sndio.handlePlayback); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_write__sndio(ma_device* pDevice, const void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesWritten) +{ + int result; + + if (pFramesWritten != NULL) { + *pFramesWritten = 0; + } + + result = ((ma_sio_write_proc)pDevice->pContext->sndio.sio_write)((struct ma_sio_hdl*)pDevice->sndio.handlePlayback, pPCMFrames, frameCount * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels)); + if (result == 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[sndio] Failed to send data from the client to the device."); + return MA_IO_ERROR; + } + + if (pFramesWritten != NULL) { + *pFramesWritten = frameCount; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_read__sndio(ma_device* pDevice, void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesRead) +{ + int result; + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + result = ((ma_sio_read_proc)pDevice->pContext->sndio.sio_read)((struct ma_sio_hdl*)pDevice->sndio.handleCapture, pPCMFrames, frameCount * ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels)); + if (result == 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[sndio] Failed to read data from the device to be sent to the device."); + return MA_IO_ERROR; + } + + if (pFramesRead != NULL) { + *pFramesRead = frameCount; + } + + return MA_SUCCESS; +} + +static ma_result ma_context_uninit__sndio(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_sndio); + + (void)pContext; + return MA_SUCCESS; +} + +static ma_result ma_context_init__sndio(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ +#ifndef MA_NO_RUNTIME_LINKING + const char* libsndioNames[] = { + "libsndio.so" + }; + size_t i; + + for (i = 0; i < ma_countof(libsndioNames); ++i) { + pContext->sndio.sndioSO = ma_dlopen(ma_context_get_log(pContext), libsndioNames[i]); + if (pContext->sndio.sndioSO != NULL) { + break; + } + } + + if (pContext->sndio.sndioSO == NULL) { + return MA_NO_BACKEND; + } + + pContext->sndio.sio_open = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_open"); + pContext->sndio.sio_close = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_close"); + pContext->sndio.sio_setpar = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_setpar"); + pContext->sndio.sio_getpar = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_getpar"); + pContext->sndio.sio_getcap = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_getcap"); + pContext->sndio.sio_write = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_write"); + pContext->sndio.sio_read = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_read"); + pContext->sndio.sio_start = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_start"); + pContext->sndio.sio_stop = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_stop"); + pContext->sndio.sio_initpar = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->sndio.sndioSO, "sio_initpar"); +#else + pContext->sndio.sio_open = sio_open; + pContext->sndio.sio_close = sio_close; + pContext->sndio.sio_setpar = sio_setpar; + pContext->sndio.sio_getpar = sio_getpar; + pContext->sndio.sio_getcap = sio_getcap; + pContext->sndio.sio_write = sio_write; + pContext->sndio.sio_read = sio_read; + pContext->sndio.sio_start = sio_start; + pContext->sndio.sio_stop = sio_stop; + pContext->sndio.sio_initpar = sio_initpar; +#endif + + pCallbacks->onContextInit = ma_context_init__sndio; + pCallbacks->onContextUninit = ma_context_uninit__sndio; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__sndio; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__sndio; + pCallbacks->onDeviceInit = ma_device_init__sndio; + pCallbacks->onDeviceUninit = ma_device_uninit__sndio; + pCallbacks->onDeviceStart = ma_device_start__sndio; + pCallbacks->onDeviceStop = ma_device_stop__sndio; + pCallbacks->onDeviceRead = ma_device_read__sndio; + pCallbacks->onDeviceWrite = ma_device_write__sndio; + pCallbacks->onDeviceDataLoop = NULL; + + (void)pConfig; + return MA_SUCCESS; +} +#endif /* MA_HAS_SNDIO */ + + + +/****************************************************************************** + +audio(4) Backend + +******************************************************************************/ +#ifdef MA_HAS_AUDIO4 +#include +#include +#include +#include +#include +#include +#include + +#ifdef __NetBSD__ +#include +#endif + +#if defined(__OpenBSD__) + #include + #if defined(OpenBSD) && OpenBSD >= 201709 + #define MA_AUDIO4_USE_NEW_API + #endif +#endif + +static void ma_construct_device_id__audio4(char* id, size_t idSize, const char* base, int deviceIndex) +{ + size_t baseLen; + + MA_ASSERT(id != NULL); + MA_ASSERT(idSize > 0); + MA_ASSERT(deviceIndex >= 0); + + baseLen = strlen(base); + MA_ASSERT(idSize > baseLen); + + ma_strcpy_s(id, idSize, base); + ma_itoa_s(deviceIndex, id+baseLen, idSize-baseLen, 10); +} + +static ma_result ma_extract_device_index_from_id__audio4(const char* id, const char* base, int* pIndexOut) +{ + size_t idLen; + size_t baseLen; + const char* deviceIndexStr; + + MA_ASSERT(id != NULL); + MA_ASSERT(base != NULL); + MA_ASSERT(pIndexOut != NULL); + + idLen = strlen(id); + baseLen = strlen(base); + if (idLen <= baseLen) { + return MA_ERROR; /* Doesn't look like the id starts with the base. */ + } + + if (strncmp(id, base, baseLen) != 0) { + return MA_ERROR; /* ID does not begin with base. */ + } + + deviceIndexStr = id + baseLen; + if (deviceIndexStr[0] == '\0') { + return MA_ERROR; /* No index specified in the ID. */ + } + + if (pIndexOut) { + *pIndexOut = atoi(deviceIndexStr); + } + + return MA_SUCCESS; +} + + +#if !defined(MA_AUDIO4_USE_NEW_API) /* Old API */ +static ma_format ma_format_from_encoding__audio4(unsigned int encoding, unsigned int precision) +{ + if (precision == 8 && (encoding == AUDIO_ENCODING_ULINEAR || encoding == AUDIO_ENCODING_ULINEAR || encoding == AUDIO_ENCODING_ULINEAR_LE || encoding == AUDIO_ENCODING_ULINEAR_BE)) { + return ma_format_u8; + } else { + if (ma_is_little_endian() && encoding == AUDIO_ENCODING_SLINEAR_LE) { + if (precision == 16) { + return ma_format_s16; + } else if (precision == 24) { + return ma_format_s24; + } else if (precision == 32) { + return ma_format_s32; + } + } else if (ma_is_big_endian() && encoding == AUDIO_ENCODING_SLINEAR_BE) { + if (precision == 16) { + return ma_format_s16; + } else if (precision == 24) { + return ma_format_s24; + } else if (precision == 32) { + return ma_format_s32; + } + } + } + + return ma_format_unknown; /* Encoding not supported. */ +} + +static void ma_encoding_from_format__audio4(ma_format format, unsigned int* pEncoding, unsigned int* pPrecision) +{ + MA_ASSERT(pEncoding != NULL); + MA_ASSERT(pPrecision != NULL); + + switch (format) + { + case ma_format_u8: + { + *pEncoding = AUDIO_ENCODING_ULINEAR; + *pPrecision = 8; + } break; + + case ma_format_s24: + { + *pEncoding = (ma_is_little_endian()) ? AUDIO_ENCODING_SLINEAR_LE : AUDIO_ENCODING_SLINEAR_BE; + *pPrecision = 24; + } break; + + case ma_format_s32: + { + *pEncoding = (ma_is_little_endian()) ? AUDIO_ENCODING_SLINEAR_LE : AUDIO_ENCODING_SLINEAR_BE; + *pPrecision = 32; + } break; + + case ma_format_s16: + case ma_format_f32: + case ma_format_unknown: + default: + { + *pEncoding = (ma_is_little_endian()) ? AUDIO_ENCODING_SLINEAR_LE : AUDIO_ENCODING_SLINEAR_BE; + *pPrecision = 16; + } break; + } +} + +static ma_format ma_format_from_prinfo__audio4(struct audio_prinfo* prinfo) +{ + return ma_format_from_encoding__audio4(prinfo->encoding, prinfo->precision); +} + +static ma_format ma_best_format_from_fd__audio4(int fd, ma_format preferredFormat) +{ + audio_encoding_t encoding; + ma_uint32 iFormat; + int counter = 0; + + /* First check to see if the preferred format is supported. */ + if (preferredFormat != ma_format_unknown) { + counter = 0; + for (;;) { + MA_ZERO_OBJECT(&encoding); + encoding.index = counter; + if (ioctl(fd, AUDIO_GETENC, &encoding) < 0) { + break; + } + + if (preferredFormat == ma_format_from_encoding__audio4(encoding.encoding, encoding.precision)) { + return preferredFormat; /* Found the preferred format. */ + } + + /* Getting here means this encoding does not match our preferred format so we need to more on to the next encoding. */ + counter += 1; + } + } + + /* Getting here means our preferred format is not supported, so fall back to our standard priorities. */ + for (iFormat = 0; iFormat < ma_countof(g_maFormatPriorities); iFormat += 1) { + ma_format format = g_maFormatPriorities[iFormat]; + + counter = 0; + for (;;) { + MA_ZERO_OBJECT(&encoding); + encoding.index = counter; + if (ioctl(fd, AUDIO_GETENC, &encoding) < 0) { + break; + } + + if (format == ma_format_from_encoding__audio4(encoding.encoding, encoding.precision)) { + return format; /* Found a workable format. */ + } + + /* Getting here means this encoding does not match our preferred format so we need to more on to the next encoding. */ + counter += 1; + } + } + + /* Getting here means not appropriate format was found. */ + return ma_format_unknown; +} +#else +static ma_format ma_format_from_swpar__audio4(struct audio_swpar* par) +{ + if (par->bits == 8 && par->bps == 1 && par->sig == 0) { + return ma_format_u8; + } + if (par->bits == 16 && par->bps == 2 && par->sig == 1 && par->le == ma_is_little_endian()) { + return ma_format_s16; + } + if (par->bits == 24 && par->bps == 3 && par->sig == 1 && par->le == ma_is_little_endian()) { + return ma_format_s24; + } + if (par->bits == 32 && par->bps == 4 && par->sig == 1 && par->le == ma_is_little_endian()) { + return ma_format_f32; + } + + /* Format not supported. */ + return ma_format_unknown; +} +#endif + +static ma_result ma_context_get_device_info_from_fd__audio4(ma_context* pContext, ma_device_type deviceType, int fd, ma_device_info* pDeviceInfo) +{ + audio_device_t fdDevice; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(fd >= 0); + MA_ASSERT(pDeviceInfo != NULL); + + (void)pContext; + (void)deviceType; + + if (ioctl(fd, AUDIO_GETDEV, &fdDevice) < 0) { + return MA_ERROR; /* Failed to retrieve device info. */ + } + + /* Name. */ + ma_strcpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), fdDevice.name); + + #if !defined(MA_AUDIO4_USE_NEW_API) + { + audio_info_t fdInfo; + int counter = 0; + ma_uint32 channels; + ma_uint32 sampleRate; + +#if defined(__NetBSD__) && (__NetBSD_Version__ >= 900000000) + if (ioctl(fd, AUDIO_GETFORMAT, &fdInfo) < 0) { + return MA_ERROR; + } +#else + if (ioctl(fd, AUDIO_GETINFO, &fdInfo) < 0) { + return MA_ERROR; + } +#endif + + if (deviceType == ma_device_type_playback) { + channels = fdInfo.play.channels; + sampleRate = fdInfo.play.sample_rate; + } else { + channels = fdInfo.record.channels; + sampleRate = fdInfo.record.sample_rate; + } + + /* Supported formats. We get this by looking at the encodings. */ + pDeviceInfo->nativeDataFormatCount = 0; + for (;;) { + audio_encoding_t encoding; + ma_format format; + + MA_ZERO_OBJECT(&encoding); + encoding.index = counter; + if (ioctl(fd, AUDIO_GETENC, &encoding) < 0) { + break; + } + + format = ma_format_from_encoding__audio4(encoding.encoding, encoding.precision); + if (format != ma_format_unknown) { + ma_device_info_add_native_data_format(pDeviceInfo, format, channels, sampleRate, 0); + } + + counter += 1; + } + } + #else + { + struct audio_swpar fdPar; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + + if (ioctl(fd, AUDIO_GETPAR, &fdPar) < 0) { + return MA_ERROR; + } + + format = ma_format_from_swpar__audio4(&fdPar); + if (format == ma_format_unknown) { + return MA_FORMAT_NOT_SUPPORTED; + } + + if (deviceType == ma_device_type_playback) { + channels = fdPar.pchan; + } else { + channels = fdPar.rchan; + } + + sampleRate = fdPar.rate; + + pDeviceInfo->nativeDataFormatCount = 0; + ma_device_info_add_native_data_format(pDeviceInfo, format, channels, sampleRate, 0); + } + #endif + + return MA_SUCCESS; +} + +static ma_result ma_context_enumerate_devices__audio4(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + const int maxDevices = 64; + char devpath[256]; + int iDevice; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + /* + Every device will be named "/dev/audioN", with a "/dev/audioctlN" equivalent. We use the "/dev/audioctlN" + version here since we can open it even when another process has control of the "/dev/audioN" device. + */ + for (iDevice = 0; iDevice < maxDevices; ++iDevice) { + struct stat st; + int fd; + ma_bool32 isTerminating = MA_FALSE; + + ma_strcpy_s(devpath, sizeof(devpath), "/dev/audioctl"); + ma_itoa_s(iDevice, devpath+strlen(devpath), sizeof(devpath)-strlen(devpath), 10); + + if (stat(devpath, &st) < 0) { + break; + } + + /* The device exists, but we need to check if it's usable as playback and/or capture. */ + + /* Playback. */ + if (!isTerminating) { + fd = open(devpath, O_RDONLY, 0); + if (fd >= 0) { + /* Supports playback. */ + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_construct_device_id__audio4(deviceInfo.id.audio4, sizeof(deviceInfo.id.audio4), "/dev/audio", iDevice); + if (ma_context_get_device_info_from_fd__audio4(pContext, ma_device_type_playback, fd, &deviceInfo) == MA_SUCCESS) { + isTerminating = !callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + + close(fd); + } + } + + /* Capture. */ + if (!isTerminating) { + fd = open(devpath, O_WRONLY, 0); + if (fd >= 0) { + /* Supports capture. */ + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_construct_device_id__audio4(deviceInfo.id.audio4, sizeof(deviceInfo.id.audio4), "/dev/audio", iDevice); + if (ma_context_get_device_info_from_fd__audio4(pContext, ma_device_type_capture, fd, &deviceInfo) == MA_SUCCESS) { + isTerminating = !callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + + close(fd); + } + } + + if (isTerminating) { + break; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info__audio4(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + int fd = -1; + int deviceIndex = -1; + char ctlid[256]; + ma_result result; + + MA_ASSERT(pContext != NULL); + + /* + We need to open the "/dev/audioctlN" device to get the info. To do this we need to extract the number + from the device ID which will be in "/dev/audioN" format. + */ + if (pDeviceID == NULL) { + /* Default device. */ + ma_strcpy_s(ctlid, sizeof(ctlid), "/dev/audioctl"); + } else { + /* Specific device. We need to convert from "/dev/audioN" to "/dev/audioctlN". */ + result = ma_extract_device_index_from_id__audio4(pDeviceID->audio4, "/dev/audio", &deviceIndex); + if (result != MA_SUCCESS) { + return result; + } + + ma_construct_device_id__audio4(ctlid, sizeof(ctlid), "/dev/audioctl", deviceIndex); + } + + fd = open(ctlid, (deviceType == ma_device_type_playback) ? O_WRONLY : O_RDONLY, 0); + if (fd == -1) { + return MA_NO_DEVICE; + } + + if (deviceIndex == -1) { + ma_strcpy_s(pDeviceInfo->id.audio4, sizeof(pDeviceInfo->id.audio4), "/dev/audio"); + } else { + ma_construct_device_id__audio4(pDeviceInfo->id.audio4, sizeof(pDeviceInfo->id.audio4), "/dev/audio", deviceIndex); + } + + result = ma_context_get_device_info_from_fd__audio4(pContext, deviceType, fd, pDeviceInfo); + + close(fd); + return result; +} + +static ma_result ma_device_uninit__audio4(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + close(pDevice->audio4.fdCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + close(pDevice->audio4.fdPlayback); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_init_fd__audio4(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptor, ma_device_type deviceType) +{ + const char* pDefaultDeviceNames[] = { + "/dev/audio", + "/dev/audio0" + }; + const char* pDefaultDeviceCtlNames[] = { + "/dev/audioctl", + "/dev/audioctl0" + }; + int fd; + int fdFlags = 0; + size_t iDefaultDevice = (size_t)-1; + ma_format internalFormat; + ma_uint32 internalChannels; + ma_uint32 internalSampleRate; + ma_uint32 internalPeriodSizeInFrames; + ma_uint32 internalPeriods; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(deviceType != ma_device_type_duplex); + MA_ASSERT(pDevice != NULL); + + /* The first thing to do is open the file. */ + if (deviceType == ma_device_type_capture) { + fdFlags = O_RDONLY; + } else { + fdFlags = O_WRONLY; + } + /*fdFlags |= O_NONBLOCK;*/ + + /* Find the index of the default device as a start. We'll use this index later. Set it to (size_t)-1 otherwise. */ + if (pDescriptor->pDeviceID == NULL) { + /* Default device. */ + for (iDefaultDevice = 0; iDefaultDevice < ma_countof(pDefaultDeviceNames); ++iDefaultDevice) { + fd = open(pDefaultDeviceNames[iDefaultDevice], fdFlags, 0); + if (fd != -1) { + break; + } + } + } else { + /* Specific device. */ + fd = open(pDescriptor->pDeviceID->audio4, fdFlags, 0); + + for (iDefaultDevice = 0; iDefaultDevice < ma_countof(pDefaultDeviceNames); iDefaultDevice += 1) { + if (ma_strcmp(pDefaultDeviceNames[iDefaultDevice], pDescriptor->pDeviceID->audio4) == 0) { + break; + } + } + + if (iDefaultDevice == ma_countof(pDefaultDeviceNames)) { + iDefaultDevice = (size_t)-1; + } + } + + if (fd == -1) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to open device."); + return ma_result_from_errno(errno); + } + + #if !defined(MA_AUDIO4_USE_NEW_API) /* Old API */ + { + audio_info_t fdInfo; + int fdInfoResult = -1; + + /* + The documentation is a little bit unclear to me as to how it handles formats. It says the + following: + + Regardless of formats supported by underlying driver, the audio driver accepts the + following formats. + + By then the next sentence says this: + + `encoding` and `precision` are one of the values obtained by AUDIO_GETENC. + + It sounds like a direct contradiction to me. I'm going to play this safe any only use the + best sample format returned by AUDIO_GETENC. If the requested format is supported we'll + use that, but otherwise we'll just use our standard format priorities to pick an + appropriate one. + */ + AUDIO_INITINFO(&fdInfo); + + /* + Get the default format from the audioctl file if we're asking for a default device. If we + retrieve it from /dev/audio it'll default to mono 8000Hz. + */ + if (iDefaultDevice != (size_t)-1) { + /* We're using a default device. Get the info from the /dev/audioctl file instead of /dev/audio. */ + int fdctl = open(pDefaultDeviceCtlNames[iDefaultDevice], fdFlags, 0); + if (fdctl != -1) { +#if defined(__NetBSD__) && (__NetBSD_Version__ >= 900000000) + fdInfoResult = ioctl(fdctl, AUDIO_GETFORMAT, &fdInfo); +#else + fdInfoResult = ioctl(fdctl, AUDIO_GETINFO, &fdInfo); +#endif + close(fdctl); + } + } + + if (fdInfoResult == -1) { + /* We still don't have the default device info so just retrieve it from the main audio device. */ + if (ioctl(fd, AUDIO_GETINFO, &fdInfo) < 0) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] AUDIO_GETINFO failed."); + return ma_result_from_errno(errno); + } + } + + /* We get the driver to do as much of the data conversion as possible. */ + if (deviceType == ma_device_type_capture) { + fdInfo.mode = AUMODE_RECORD; + ma_encoding_from_format__audio4(ma_best_format_from_fd__audio4(fd, pDescriptor->format), &fdInfo.record.encoding, &fdInfo.record.precision); + + if (pDescriptor->channels != 0) { + fdInfo.record.channels = ma_clamp(pDescriptor->channels, 1, 12); /* From the documentation: `channels` ranges from 1 to 12. */ + } + + if (pDescriptor->sampleRate != 0) { + fdInfo.record.sample_rate = ma_clamp(pDescriptor->sampleRate, 1000, 192000); /* From the documentation: `frequency` ranges from 1000Hz to 192000Hz. (They mean `sample_rate` instead of `frequency`.) */ + } + } else { + fdInfo.mode = AUMODE_PLAY; + ma_encoding_from_format__audio4(ma_best_format_from_fd__audio4(fd, pDescriptor->format), &fdInfo.play.encoding, &fdInfo.play.precision); + + if (pDescriptor->channels != 0) { + fdInfo.play.channels = ma_clamp(pDescriptor->channels, 1, 12); /* From the documentation: `channels` ranges from 1 to 12. */ + } + + if (pDescriptor->sampleRate != 0) { + fdInfo.play.sample_rate = ma_clamp(pDescriptor->sampleRate, 1000, 192000); /* From the documentation: `frequency` ranges from 1000Hz to 192000Hz. (They mean `sample_rate` instead of `frequency`.) */ + } + } + + if (ioctl(fd, AUDIO_SETINFO, &fdInfo) < 0) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to set device format. AUDIO_SETINFO failed."); + return ma_result_from_errno(errno); + } + + if (ioctl(fd, AUDIO_GETINFO, &fdInfo) < 0) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] AUDIO_GETINFO failed."); + return ma_result_from_errno(errno); + } + + if (deviceType == ma_device_type_capture) { + internalFormat = ma_format_from_prinfo__audio4(&fdInfo.record); + internalChannels = fdInfo.record.channels; + internalSampleRate = fdInfo.record.sample_rate; + } else { + internalFormat = ma_format_from_prinfo__audio4(&fdInfo.play); + internalChannels = fdInfo.play.channels; + internalSampleRate = fdInfo.play.sample_rate; + } + + if (internalFormat == ma_format_unknown) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] The device's internal device format is not supported by miniaudio. The device is unusable."); + return MA_FORMAT_NOT_SUPPORTED; + } + + /* Buffer. */ + { + ma_uint32 internalPeriodSizeInBytes; + + internalPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptor, internalSampleRate, pConfig->performanceProfile); + + internalPeriodSizeInBytes = internalPeriodSizeInFrames * ma_get_bytes_per_frame(internalFormat, internalChannels); + if (internalPeriodSizeInBytes < 16) { + internalPeriodSizeInBytes = 16; + } + + internalPeriods = pDescriptor->periodCount; + if (internalPeriods < 2) { + internalPeriods = 2; + } + + /* What miniaudio calls a period, audio4 calls a block. */ + AUDIO_INITINFO(&fdInfo); + fdInfo.hiwat = internalPeriods; + fdInfo.lowat = internalPeriods-1; + fdInfo.blocksize = internalPeriodSizeInBytes; + if (ioctl(fd, AUDIO_SETINFO, &fdInfo) < 0) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to set internal buffer size. AUDIO_SETINFO failed."); + return ma_result_from_errno(errno); + } + + internalPeriods = fdInfo.hiwat; + internalPeriodSizeInFrames = fdInfo.blocksize / ma_get_bytes_per_frame(internalFormat, internalChannels); + } + } + #else + { + struct audio_swpar fdPar; + + /* We need to retrieve the format of the device so we can know the channel count and sample rate. Then we can calculate the buffer size. */ + if (ioctl(fd, AUDIO_GETPAR, &fdPar) < 0) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to retrieve initial device parameters."); + return ma_result_from_errno(errno); + } + + internalFormat = ma_format_from_swpar__audio4(&fdPar); + internalChannels = (deviceType == ma_device_type_capture) ? fdPar.rchan : fdPar.pchan; + internalSampleRate = fdPar.rate; + + if (internalFormat == ma_format_unknown) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] The device's internal device format is not supported by miniaudio. The device is unusable."); + return MA_FORMAT_NOT_SUPPORTED; + } + + /* Buffer. */ + { + ma_uint32 internalPeriodSizeInBytes; + + internalPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptor, internalSampleRate, pConfig->performanceProfile); + + /* What miniaudio calls a period, audio4 calls a block. */ + internalPeriodSizeInBytes = internalPeriodSizeInFrames * ma_get_bytes_per_frame(internalFormat, internalChannels); + if (internalPeriodSizeInBytes < 16) { + internalPeriodSizeInBytes = 16; + } + + fdPar.nblks = pDescriptor->periodCount; + fdPar.round = internalPeriodSizeInBytes; + + if (ioctl(fd, AUDIO_SETPAR, &fdPar) < 0) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to set device parameters."); + return ma_result_from_errno(errno); + } + + if (ioctl(fd, AUDIO_GETPAR, &fdPar) < 0) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to retrieve actual device parameters."); + return ma_result_from_errno(errno); + } + } + + internalFormat = ma_format_from_swpar__audio4(&fdPar); + internalChannels = (deviceType == ma_device_type_capture) ? fdPar.rchan : fdPar.pchan; + internalSampleRate = fdPar.rate; + internalPeriods = fdPar.nblks; + internalPeriodSizeInFrames = fdPar.round / ma_get_bytes_per_frame(internalFormat, internalChannels); + } + #endif + + if (internalFormat == ma_format_unknown) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] The device's internal device format is not supported by miniaudio. The device is unusable."); + return MA_FORMAT_NOT_SUPPORTED; + } + + if (deviceType == ma_device_type_capture) { + pDevice->audio4.fdCapture = fd; + } else { + pDevice->audio4.fdPlayback = fd; + } + + pDescriptor->format = internalFormat; + pDescriptor->channels = internalChannels; + pDescriptor->sampleRate = internalSampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_sound4, pDescriptor->channelMap, ma_countof(pDescriptor->channelMap), internalChannels); + pDescriptor->periodSizeInFrames = internalPeriodSizeInFrames; + pDescriptor->periodCount = internalPeriods; + + return MA_SUCCESS; +} + +static ma_result ma_device_init__audio4(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + MA_ASSERT(pDevice != NULL); + + MA_ZERO_OBJECT(&pDevice->audio4); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + pDevice->audio4.fdCapture = -1; + pDevice->audio4.fdPlayback = -1; + + /* + The version of the operating system dictates whether or not the device is exclusive or shared. NetBSD + introduced in-kernel mixing which means it's shared. All other BSD flavours are exclusive as far as + I'm aware. + */ +#if defined(__NetBSD_Version__) && __NetBSD_Version__ >= 800000000 + /* NetBSD 8.0+ */ + if (((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pDescriptorPlayback->shareMode == ma_share_mode_exclusive) || + ((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) && pDescriptorCapture->shareMode == ma_share_mode_exclusive)) { + return MA_SHARE_MODE_NOT_SUPPORTED; + } +#else + /* All other flavors. */ +#endif + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ma_result result = ma_device_init_fd__audio4(pDevice, pConfig, pDescriptorCapture, ma_device_type_capture); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_result result = ma_device_init_fd__audio4(pDevice, pConfig, pDescriptorPlayback, ma_device_type_playback); + if (result != MA_SUCCESS) { + if (pConfig->deviceType == ma_device_type_duplex) { + close(pDevice->audio4.fdCapture); + } + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_start__audio4(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + if (pDevice->audio4.fdCapture == -1) { + return MA_INVALID_ARGS; + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + if (pDevice->audio4.fdPlayback == -1) { + return MA_INVALID_ARGS; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop_fd__audio4(ma_device* pDevice, int fd) +{ + if (fd == -1) { + return MA_INVALID_ARGS; + } + +#if !defined(MA_AUDIO4_USE_NEW_API) + if (ioctl(fd, AUDIO_FLUSH, 0) < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to stop device. AUDIO_FLUSH failed."); + return ma_result_from_errno(errno); + } +#else + if (ioctl(fd, AUDIO_STOP, 0) < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to stop device. AUDIO_STOP failed."); + return ma_result_from_errno(errno); + } +#endif + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__audio4(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ma_result result; + + result = ma_device_stop_fd__audio4(pDevice, pDevice->audio4.fdCapture); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_result result; + + /* Drain the device first. If this fails we'll just need to flush without draining. Unfortunately draining isn't available on newer version of OpenBSD. */ + #if !defined(MA_AUDIO4_USE_NEW_API) + ioctl(pDevice->audio4.fdPlayback, AUDIO_DRAIN, 0); + #endif + + /* Here is where the device is stopped immediately. */ + result = ma_device_stop_fd__audio4(pDevice, pDevice->audio4.fdPlayback); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_write__audio4(ma_device* pDevice, const void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesWritten) +{ + int result; + + if (pFramesWritten != NULL) { + *pFramesWritten = 0; + } + + result = write(pDevice->audio4.fdPlayback, pPCMFrames, frameCount * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels)); + if (result < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to write data to the device."); + return ma_result_from_errno(errno); + } + + if (pFramesWritten != NULL) { + *pFramesWritten = (ma_uint32)result / ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_read__audio4(ma_device* pDevice, void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesRead) +{ + int result; + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + result = read(pDevice->audio4.fdCapture, pPCMFrames, frameCount * ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels)); + if (result < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[audio4] Failed to read data from the device."); + return ma_result_from_errno(errno); + } + + if (pFramesRead != NULL) { + *pFramesRead = (ma_uint32)result / ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + } + + return MA_SUCCESS; +} + +static ma_result ma_context_uninit__audio4(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_audio4); + + (void)pContext; + return MA_SUCCESS; +} + +static ma_result ma_context_init__audio4(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + MA_ASSERT(pContext != NULL); + + (void)pConfig; + + pCallbacks->onContextInit = ma_context_init__audio4; + pCallbacks->onContextUninit = ma_context_uninit__audio4; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__audio4; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__audio4; + pCallbacks->onDeviceInit = ma_device_init__audio4; + pCallbacks->onDeviceUninit = ma_device_uninit__audio4; + pCallbacks->onDeviceStart = ma_device_start__audio4; + pCallbacks->onDeviceStop = ma_device_stop__audio4; + pCallbacks->onDeviceRead = ma_device_read__audio4; + pCallbacks->onDeviceWrite = ma_device_write__audio4; + pCallbacks->onDeviceDataLoop = NULL; + + return MA_SUCCESS; +} +#endif /* MA_HAS_AUDIO4 */ + + +/****************************************************************************** + +OSS Backend + +******************************************************************************/ +#ifdef MA_HAS_OSS +#include +#include +#include +#include + +#ifndef SNDCTL_DSP_HALT +#define SNDCTL_DSP_HALT SNDCTL_DSP_RESET +#endif + +#define MA_OSS_DEFAULT_DEVICE_NAME "/dev/dsp" + +static int ma_open_temp_device__oss() +{ + /* The OSS sample code uses "/dev/mixer" as the device for getting system properties so I'm going to do the same. */ + int fd = open("/dev/mixer", O_RDONLY, 0); + if (fd >= 0) { + return fd; + } + + return -1; +} + +static ma_result ma_context_open_device__oss(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_share_mode shareMode, int* pfd) +{ + const char* deviceName; + int flags; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pfd != NULL); + (void)pContext; + + *pfd = -1; + + /* This function should only be called for playback or capture, not duplex. */ + if (deviceType == ma_device_type_duplex) { + return MA_INVALID_ARGS; + } + + deviceName = MA_OSS_DEFAULT_DEVICE_NAME; + if (pDeviceID != NULL) { + deviceName = pDeviceID->oss; + } + + flags = (deviceType == ma_device_type_playback) ? O_WRONLY : O_RDONLY; + if (shareMode == ma_share_mode_exclusive) { + flags |= O_EXCL; + } + + *pfd = open(deviceName, flags, 0); + if (*pfd == -1) { + return ma_result_from_errno(errno); + } + + return MA_SUCCESS; +} + +static ma_result ma_context_enumerate_devices__oss(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + int fd; + oss_sysinfo si; + int result; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + fd = ma_open_temp_device__oss(); + if (fd == -1) { + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[OSS] Failed to open a temporary device for retrieving system information used for device enumeration."); + return MA_NO_BACKEND; + } + + result = ioctl(fd, SNDCTL_SYSINFO, &si); + if (result != -1) { + int iAudioDevice; + for (iAudioDevice = 0; iAudioDevice < si.numaudios; ++iAudioDevice) { + oss_audioinfo ai; + ai.dev = iAudioDevice; + result = ioctl(fd, SNDCTL_AUDIOINFO, &ai); + if (result != -1) { + if (ai.devnode[0] != '\0') { /* <-- Can be blank, according to documentation. */ + ma_device_info deviceInfo; + ma_bool32 isTerminating = MA_FALSE; + + MA_ZERO_OBJECT(&deviceInfo); + + /* ID */ + ma_strncpy_s(deviceInfo.id.oss, sizeof(deviceInfo.id.oss), ai.devnode, (size_t)-1); + + /* + The human readable device name should be in the "ai.handle" variable, but it can + sometimes be empty in which case we just fall back to "ai.name" which is less user + friendly, but usually has a value. + */ + if (ai.handle[0] != '\0') { + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), ai.handle, (size_t)-1); + } else { + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), ai.name, (size_t)-1); + } + + /* The device can be both playback and capture. */ + if (!isTerminating && (ai.caps & PCM_CAP_OUTPUT) != 0) { + isTerminating = !callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + if (!isTerminating && (ai.caps & PCM_CAP_INPUT) != 0) { + isTerminating = !callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + + if (isTerminating) { + break; + } + } + } + } + } else { + close(fd); + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[OSS] Failed to retrieve system information for device enumeration."); + return MA_NO_BACKEND; + } + + close(fd); + return MA_SUCCESS; +} + +static void ma_context_add_native_data_format__oss(ma_context* pContext, oss_audioinfo* pAudioInfo, ma_format format, ma_device_info* pDeviceInfo) +{ + unsigned int minChannels; + unsigned int maxChannels; + unsigned int iRate; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pAudioInfo != NULL); + MA_ASSERT(pDeviceInfo != NULL); + + /* If we support all channels we just report 0. */ + minChannels = ma_clamp(pAudioInfo->min_channels, MA_MIN_CHANNELS, MA_MAX_CHANNELS); + maxChannels = ma_clamp(pAudioInfo->max_channels, MA_MIN_CHANNELS, MA_MAX_CHANNELS); + + /* + OSS has this annoying thing where sample rates can be reported in two ways. We prefer explicitness, + which OSS has in the form of nrates/rates, however there are times where nrates can be 0, in which + case we'll need to use min_rate and max_rate and report only standard rates. + */ + if (pAudioInfo->nrates > 0) { + for (iRate = 0; iRate < pAudioInfo->nrates; iRate += 1) { + unsigned int rate = pAudioInfo->rates[iRate]; + + if (minChannels == MA_MIN_CHANNELS && maxChannels == MA_MAX_CHANNELS) { + ma_device_info_add_native_data_format(pDeviceInfo, format, 0, rate, 0); /* Set the channel count to 0 to indicate that all channel counts are supported. */ + } else { + unsigned int iChannel; + for (iChannel = minChannels; iChannel <= maxChannels; iChannel += 1) { + ma_device_info_add_native_data_format(pDeviceInfo, format, iChannel, rate, 0); + } + } + } + } else { + for (iRate = 0; iRate < ma_countof(g_maStandardSampleRatePriorities); iRate += 1) { + ma_uint32 standardRate = g_maStandardSampleRatePriorities[iRate]; + + if (standardRate >= (ma_uint32)pAudioInfo->min_rate && standardRate <= (ma_uint32)pAudioInfo->max_rate) { + if (minChannels == MA_MIN_CHANNELS && maxChannels == MA_MAX_CHANNELS) { + ma_device_info_add_native_data_format(pDeviceInfo, format, 0, standardRate, 0); /* Set the channel count to 0 to indicate that all channel counts are supported. */ + } else { + unsigned int iChannel; + for (iChannel = minChannels; iChannel <= maxChannels; iChannel += 1) { + ma_device_info_add_native_data_format(pDeviceInfo, format, iChannel, standardRate, 0); + } + } + } + } + } +} + +static ma_result ma_context_get_device_info__oss(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + ma_bool32 foundDevice; + int fdTemp; + oss_sysinfo si; + int result; + + MA_ASSERT(pContext != NULL); + + /* Handle the default device a little differently. */ + if (pDeviceID == NULL) { + if (deviceType == ma_device_type_playback) { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } + + return MA_SUCCESS; + } + + + /* If we get here it means we are _not_ using the default device. */ + foundDevice = MA_FALSE; + + fdTemp = ma_open_temp_device__oss(); + if (fdTemp == -1) { + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[OSS] Failed to open a temporary device for retrieving system information used for device enumeration."); + return MA_NO_BACKEND; + } + + result = ioctl(fdTemp, SNDCTL_SYSINFO, &si); + if (result != -1) { + int iAudioDevice; + for (iAudioDevice = 0; iAudioDevice < si.numaudios; ++iAudioDevice) { + oss_audioinfo ai; + ai.dev = iAudioDevice; + result = ioctl(fdTemp, SNDCTL_AUDIOINFO, &ai); + if (result != -1) { + if (ma_strcmp(ai.devnode, pDeviceID->oss) == 0) { + /* It has the same name, so now just confirm the type. */ + if ((deviceType == ma_device_type_playback && ((ai.caps & PCM_CAP_OUTPUT) != 0)) || + (deviceType == ma_device_type_capture && ((ai.caps & PCM_CAP_INPUT) != 0))) { + unsigned int formatMask; + + /* ID */ + ma_strncpy_s(pDeviceInfo->id.oss, sizeof(pDeviceInfo->id.oss), ai.devnode, (size_t)-1); + + /* + The human readable device name should be in the "ai.handle" variable, but it can + sometimes be empty in which case we just fall back to "ai.name" which is less user + friendly, but usually has a value. + */ + if (ai.handle[0] != '\0') { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), ai.handle, (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), ai.name, (size_t)-1); + } + + + pDeviceInfo->nativeDataFormatCount = 0; + + if (deviceType == ma_device_type_playback) { + formatMask = ai.oformats; + } else { + formatMask = ai.iformats; + } + + if (((formatMask & AFMT_S16_LE) != 0 && ma_is_little_endian()) || (AFMT_S16_BE && ma_is_big_endian())) { + ma_context_add_native_data_format__oss(pContext, &ai, ma_format_s16, pDeviceInfo); + } + if (((formatMask & AFMT_S32_LE) != 0 && ma_is_little_endian()) || (AFMT_S32_BE && ma_is_big_endian())) { + ma_context_add_native_data_format__oss(pContext, &ai, ma_format_s32, pDeviceInfo); + } + if ((formatMask & AFMT_U8) != 0) { + ma_context_add_native_data_format__oss(pContext, &ai, ma_format_u8, pDeviceInfo); + } + + foundDevice = MA_TRUE; + break; + } + } + } + } + } else { + close(fdTemp); + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[OSS] Failed to retrieve system information for device enumeration."); + return MA_NO_BACKEND; + } + + + close(fdTemp); + + if (!foundDevice) { + return MA_NO_DEVICE; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_uninit__oss(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + close(pDevice->oss.fdCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + close(pDevice->oss.fdPlayback); + } + + return MA_SUCCESS; +} + +static int ma_format_to_oss(ma_format format) +{ + int ossFormat = AFMT_U8; + switch (format) { + case ma_format_s16: ossFormat = (ma_is_little_endian()) ? AFMT_S16_LE : AFMT_S16_BE; break; + case ma_format_s24: ossFormat = (ma_is_little_endian()) ? AFMT_S32_LE : AFMT_S32_BE; break; + case ma_format_s32: ossFormat = (ma_is_little_endian()) ? AFMT_S32_LE : AFMT_S32_BE; break; + case ma_format_f32: ossFormat = (ma_is_little_endian()) ? AFMT_S16_LE : AFMT_S16_BE; break; + case ma_format_u8: + default: ossFormat = AFMT_U8; break; + } + + return ossFormat; +} + +static ma_format ma_format_from_oss(int ossFormat) +{ + if (ossFormat == AFMT_U8) { + return ma_format_u8; + } else { + if (ma_is_little_endian()) { + switch (ossFormat) { + case AFMT_S16_LE: return ma_format_s16; + case AFMT_S32_LE: return ma_format_s32; + default: return ma_format_unknown; + } + } else { + switch (ossFormat) { + case AFMT_S16_BE: return ma_format_s16; + case AFMT_S32_BE: return ma_format_s32; + default: return ma_format_unknown; + } + } + } + + return ma_format_unknown; +} + +static ma_result ma_device_init_fd__oss(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptor, ma_device_type deviceType) +{ + ma_result result; + int ossResult; + int fd; + const ma_device_id* pDeviceID = NULL; + ma_share_mode shareMode; + int ossFormat; + int ossChannels; + int ossSampleRate; + int ossFragment; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pConfig != NULL); + MA_ASSERT(deviceType != ma_device_type_duplex); + + pDeviceID = pDescriptor->pDeviceID; + shareMode = pDescriptor->shareMode; + ossFormat = ma_format_to_oss((pDescriptor->format != ma_format_unknown) ? pDescriptor->format : ma_format_s16); /* Use s16 by default because OSS doesn't like floating point. */ + ossChannels = (int)(pDescriptor->channels > 0) ? pDescriptor->channels : MA_DEFAULT_CHANNELS; + ossSampleRate = (int)(pDescriptor->sampleRate > 0) ? pDescriptor->sampleRate : MA_DEFAULT_SAMPLE_RATE; + + result = ma_context_open_device__oss(pDevice->pContext, deviceType, pDeviceID, shareMode, &fd); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to open device."); + return result; + } + + /* + The OSS documentation is very clear about the order we should be initializing the device's properties: + 1) Format + 2) Channels + 3) Sample rate. + */ + + /* Format. */ + ossResult = ioctl(fd, SNDCTL_DSP_SETFMT, &ossFormat); + if (ossResult == -1) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to set format."); + return ma_result_from_errno(errno); + } + + /* Channels. */ + ossResult = ioctl(fd, SNDCTL_DSP_CHANNELS, &ossChannels); + if (ossResult == -1) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to set channel count."); + return ma_result_from_errno(errno); + } + + /* Sample Rate. */ + ossResult = ioctl(fd, SNDCTL_DSP_SPEED, &ossSampleRate); + if (ossResult == -1) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to set sample rate."); + return ma_result_from_errno(errno); + } + + /* + Buffer. + + The documentation says that the fragment settings should be set as soon as possible, but I'm not sure if + it should be done before or after format/channels/rate. + + OSS wants the fragment size in bytes and a power of 2. When setting, we specify the power, not the actual + value. + */ + { + ma_uint32 periodSizeInFrames; + ma_uint32 periodSizeInBytes; + ma_uint32 ossFragmentSizePower; + + periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptor, (ma_uint32)ossSampleRate, pConfig->performanceProfile); + + periodSizeInBytes = ma_round_to_power_of_2(periodSizeInFrames * ma_get_bytes_per_frame(ma_format_from_oss(ossFormat), ossChannels)); + if (periodSizeInBytes < 16) { + periodSizeInBytes = 16; + } + + ossFragmentSizePower = 4; + periodSizeInBytes >>= 4; + while (periodSizeInBytes >>= 1) { + ossFragmentSizePower += 1; + } + + ossFragment = (int)((pConfig->periods << 16) | ossFragmentSizePower); + ossResult = ioctl(fd, SNDCTL_DSP_SETFRAGMENT, &ossFragment); + if (ossResult == -1) { + close(fd); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to set fragment size and period count."); + return ma_result_from_errno(errno); + } + } + + /* Internal settings. */ + if (deviceType == ma_device_type_capture) { + pDevice->oss.fdCapture = fd; + } else { + pDevice->oss.fdPlayback = fd; + } + + pDescriptor->format = ma_format_from_oss(ossFormat); + pDescriptor->channels = ossChannels; + pDescriptor->sampleRate = ossSampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_sound4, pDescriptor->channelMap, ma_countof(pDescriptor->channelMap), pDescriptor->channels); + pDescriptor->periodCount = (ma_uint32)(ossFragment >> 16); + pDescriptor->periodSizeInFrames = (ma_uint32)(1 << (ossFragment & 0xFFFF)) / ma_get_bytes_per_frame(pDescriptor->format, pDescriptor->channels); + + if (pDescriptor->format == ma_format_unknown) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] The device's internal format is not supported by miniaudio."); + return MA_FORMAT_NOT_SUPPORTED; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_init__oss(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pConfig != NULL); + + MA_ZERO_OBJECT(&pDevice->oss); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ma_result result = ma_device_init_fd__oss(pDevice, pConfig, pDescriptorCapture, ma_device_type_capture); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to open device."); + return result; + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_result result = ma_device_init_fd__oss(pDevice, pConfig, pDescriptorPlayback, ma_device_type_playback); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to open device."); + return result; + } + } + + return MA_SUCCESS; +} + +/* +Note on Starting and Stopping +============================= +In the past I was using SNDCTL_DSP_HALT to stop the device, however this results in issues when +trying to resume the device again. If we use SNDCTL_DSP_HALT, the next write() or read() will +fail. Instead what we need to do is just not write or read to and from the device when the +device is not running. + +As a result, both the start and stop functions for OSS are just empty stubs. The starting and +stopping logic is handled by ma_device_write__oss() and ma_device_read__oss(). These will check +the device state, and if the device is stopped they will simply not do any kind of processing. + +The downside to this technique is that I've noticed a fairly lengthy delay in stopping the +device, up to a second. This is on a virtual machine, and as such might just be due to the +virtual drivers, but I'm not fully sure. I am not sure how to work around this problem so for +the moment that's just how it's going to have to be. + +When starting the device, OSS will automatically start it when write() or read() is called. +*/ +static ma_result ma_device_start__oss(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + /* The device is automatically started with reading and writing. */ + (void)pDevice; + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__oss(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + /* See note above on why this is empty. */ + (void)pDevice; + + return MA_SUCCESS; +} + +static ma_result ma_device_write__oss(ma_device* pDevice, const void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesWritten) +{ + int resultOSS; + ma_uint32 deviceState; + + if (pFramesWritten != NULL) { + *pFramesWritten = 0; + } + + /* Don't do any processing if the device is stopped. */ + deviceState = ma_device_get_state(pDevice); + if (deviceState != ma_device_state_started && deviceState != ma_device_state_starting) { + return MA_SUCCESS; + } + + resultOSS = write(pDevice->oss.fdPlayback, pPCMFrames, frameCount * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels)); + if (resultOSS < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to send data from the client to the device."); + return ma_result_from_errno(errno); + } + + if (pFramesWritten != NULL) { + *pFramesWritten = (ma_uint32)resultOSS / ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + } + + return MA_SUCCESS; +} + +static ma_result ma_device_read__oss(ma_device* pDevice, void* pPCMFrames, ma_uint32 frameCount, ma_uint32* pFramesRead) +{ + int resultOSS; + ma_uint32 deviceState; + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + /* Don't do any processing if the device is stopped. */ + deviceState = ma_device_get_state(pDevice); + if (deviceState != ma_device_state_started && deviceState != ma_device_state_starting) { + return MA_SUCCESS; + } + + resultOSS = read(pDevice->oss.fdCapture, pPCMFrames, frameCount * ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels)); + if (resultOSS < 0) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OSS] Failed to read data from the device to be sent to the client."); + return ma_result_from_errno(errno); + } + + if (pFramesRead != NULL) { + *pFramesRead = (ma_uint32)resultOSS / ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + } + + return MA_SUCCESS; +} + +static ma_result ma_context_uninit__oss(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_oss); + + (void)pContext; + return MA_SUCCESS; +} + +static ma_result ma_context_init__oss(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + int fd; + int ossVersion; + int result; + + MA_ASSERT(pContext != NULL); + + (void)pConfig; + + /* Try opening a temporary device first so we can get version information. This is closed at the end. */ + fd = ma_open_temp_device__oss(); + if (fd == -1) { + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[OSS] Failed to open temporary device for retrieving system properties."); /* Looks liks OSS isn't installed, or there are no available devices. */ + return MA_NO_BACKEND; + } + + /* Grab the OSS version. */ + ossVersion = 0; + result = ioctl(fd, OSS_GETVERSION, &ossVersion); + if (result == -1) { + close(fd); + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_ERROR, "[OSS] Failed to retrieve OSS version."); + return MA_NO_BACKEND; + } + + /* The file handle to temp device is no longer needed. Close ASAP. */ + close(fd); + + pContext->oss.versionMajor = ((ossVersion & 0xFF0000) >> 16); + pContext->oss.versionMinor = ((ossVersion & 0x00FF00) >> 8); + + pCallbacks->onContextInit = ma_context_init__oss; + pCallbacks->onContextUninit = ma_context_uninit__oss; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__oss; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__oss; + pCallbacks->onDeviceInit = ma_device_init__oss; + pCallbacks->onDeviceUninit = ma_device_uninit__oss; + pCallbacks->onDeviceStart = ma_device_start__oss; + pCallbacks->onDeviceStop = ma_device_stop__oss; + pCallbacks->onDeviceRead = ma_device_read__oss; + pCallbacks->onDeviceWrite = ma_device_write__oss; + pCallbacks->onDeviceDataLoop = NULL; + + return MA_SUCCESS; +} +#endif /* MA_HAS_OSS */ + + + + + +/****************************************************************************** + +AAudio Backend + +******************************************************************************/ +#ifdef MA_HAS_AAUDIO + +#ifdef MA_NO_RUNTIME_LINKING + #include +#endif + +typedef int32_t ma_aaudio_result_t; +typedef int32_t ma_aaudio_direction_t; +typedef int32_t ma_aaudio_sharing_mode_t; +typedef int32_t ma_aaudio_format_t; +typedef int32_t ma_aaudio_stream_state_t; +typedef int32_t ma_aaudio_performance_mode_t; +typedef int32_t ma_aaudio_usage_t; +typedef int32_t ma_aaudio_content_type_t; +typedef int32_t ma_aaudio_input_preset_t; +typedef int32_t ma_aaudio_allowed_capture_policy_t; +typedef int32_t ma_aaudio_data_callback_result_t; +typedef struct ma_AAudioStreamBuilder_t* ma_AAudioStreamBuilder; +typedef struct ma_AAudioStream_t* ma_AAudioStream; + +#define MA_AAUDIO_UNSPECIFIED 0 + +/* Result codes. miniaudio only cares about the success code. */ +#define MA_AAUDIO_OK 0 + +/* Directions. */ +#define MA_AAUDIO_DIRECTION_OUTPUT 0 +#define MA_AAUDIO_DIRECTION_INPUT 1 + +/* Sharing modes. */ +#define MA_AAUDIO_SHARING_MODE_EXCLUSIVE 0 +#define MA_AAUDIO_SHARING_MODE_SHARED 1 + +/* Formats. */ +#define MA_AAUDIO_FORMAT_PCM_I16 1 +#define MA_AAUDIO_FORMAT_PCM_FLOAT 2 + +/* Stream states. */ +#define MA_AAUDIO_STREAM_STATE_UNINITIALIZED 0 +#define MA_AAUDIO_STREAM_STATE_UNKNOWN 1 +#define MA_AAUDIO_STREAM_STATE_OPEN 2 +#define MA_AAUDIO_STREAM_STATE_STARTING 3 +#define MA_AAUDIO_STREAM_STATE_STARTED 4 +#define MA_AAUDIO_STREAM_STATE_PAUSING 5 +#define MA_AAUDIO_STREAM_STATE_PAUSED 6 +#define MA_AAUDIO_STREAM_STATE_FLUSHING 7 +#define MA_AAUDIO_STREAM_STATE_FLUSHED 8 +#define MA_AAUDIO_STREAM_STATE_STOPPING 9 +#define MA_AAUDIO_STREAM_STATE_STOPPED 10 +#define MA_AAUDIO_STREAM_STATE_CLOSING 11 +#define MA_AAUDIO_STREAM_STATE_CLOSED 12 +#define MA_AAUDIO_STREAM_STATE_DISCONNECTED 13 + +/* Performance modes. */ +#define MA_AAUDIO_PERFORMANCE_MODE_NONE 10 +#define MA_AAUDIO_PERFORMANCE_MODE_POWER_SAVING 11 +#define MA_AAUDIO_PERFORMANCE_MODE_LOW_LATENCY 12 + +/* Usage types. */ +#define MA_AAUDIO_USAGE_MEDIA 1 +#define MA_AAUDIO_USAGE_VOICE_COMMUNICATION 2 +#define MA_AAUDIO_USAGE_VOICE_COMMUNICATION_SIGNALLING 3 +#define MA_AAUDIO_USAGE_ALARM 4 +#define MA_AAUDIO_USAGE_NOTIFICATION 5 +#define MA_AAUDIO_USAGE_NOTIFICATION_RINGTONE 6 +#define MA_AAUDIO_USAGE_NOTIFICATION_EVENT 10 +#define MA_AAUDIO_USAGE_ASSISTANCE_ACCESSIBILITY 11 +#define MA_AAUDIO_USAGE_ASSISTANCE_NAVIGATION_GUIDANCE 12 +#define MA_AAUDIO_USAGE_ASSISTANCE_SONIFICATION 13 +#define MA_AAUDIO_USAGE_GAME 14 +#define MA_AAUDIO_USAGE_ASSISTANT 16 +#define MA_AAUDIO_SYSTEM_USAGE_EMERGENCY 1000 +#define MA_AAUDIO_SYSTEM_USAGE_SAFETY 1001 +#define MA_AAUDIO_SYSTEM_USAGE_VEHICLE_STATUS 1002 +#define MA_AAUDIO_SYSTEM_USAGE_ANNOUNCEMENT 1003 + +/* Content types. */ +#define MA_AAUDIO_CONTENT_TYPE_SPEECH 1 +#define MA_AAUDIO_CONTENT_TYPE_MUSIC 2 +#define MA_AAUDIO_CONTENT_TYPE_MOVIE 3 +#define MA_AAUDIO_CONTENT_TYPE_SONIFICATION 4 + +/* Input presets. */ +#define MA_AAUDIO_INPUT_PRESET_GENERIC 1 +#define MA_AAUDIO_INPUT_PRESET_CAMCORDER 5 +#define MA_AAUDIO_INPUT_PRESET_VOICE_RECOGNITION 6 +#define MA_AAUDIO_INPUT_PRESET_VOICE_COMMUNICATION 7 +#define MA_AAUDIO_INPUT_PRESET_UNPROCESSED 9 +#define MA_AAUDIO_INPUT_PRESET_VOICE_PERFORMANCE 10 + +/* Allowed Capture Policies */ +#define MA_AAUDIO_ALLOW_CAPTURE_BY_ALL 1 +#define MA_AAUDIO_ALLOW_CAPTURE_BY_SYSTEM 2 +#define MA_AAUDIO_ALLOW_CAPTURE_BY_NONE 3 + +/* Callback results. */ +#define MA_AAUDIO_CALLBACK_RESULT_CONTINUE 0 +#define MA_AAUDIO_CALLBACK_RESULT_STOP 1 + + +typedef ma_aaudio_data_callback_result_t (* ma_AAudioStream_dataCallback) (ma_AAudioStream* pStream, void* pUserData, void* pAudioData, int32_t numFrames); +typedef void (* ma_AAudioStream_errorCallback)(ma_AAudioStream *pStream, void *pUserData, ma_aaudio_result_t error); + +typedef ma_aaudio_result_t (* MA_PFN_AAudio_createStreamBuilder) (ma_AAudioStreamBuilder** ppBuilder); +typedef ma_aaudio_result_t (* MA_PFN_AAudioStreamBuilder_delete) (ma_AAudioStreamBuilder* pBuilder); +typedef void (* MA_PFN_AAudioStreamBuilder_setDeviceId) (ma_AAudioStreamBuilder* pBuilder, int32_t deviceId); +typedef void (* MA_PFN_AAudioStreamBuilder_setDirection) (ma_AAudioStreamBuilder* pBuilder, ma_aaudio_direction_t direction); +typedef void (* MA_PFN_AAudioStreamBuilder_setSharingMode) (ma_AAudioStreamBuilder* pBuilder, ma_aaudio_sharing_mode_t sharingMode); +typedef void (* MA_PFN_AAudioStreamBuilder_setFormat) (ma_AAudioStreamBuilder* pBuilder, ma_aaudio_format_t format); +typedef void (* MA_PFN_AAudioStreamBuilder_setChannelCount) (ma_AAudioStreamBuilder* pBuilder, int32_t channelCount); +typedef void (* MA_PFN_AAudioStreamBuilder_setSampleRate) (ma_AAudioStreamBuilder* pBuilder, int32_t sampleRate); +typedef void (* MA_PFN_AAudioStreamBuilder_setBufferCapacityInFrames)(ma_AAudioStreamBuilder* pBuilder, int32_t numFrames); +typedef void (* MA_PFN_AAudioStreamBuilder_setFramesPerDataCallback) (ma_AAudioStreamBuilder* pBuilder, int32_t numFrames); +typedef void (* MA_PFN_AAudioStreamBuilder_setDataCallback) (ma_AAudioStreamBuilder* pBuilder, ma_AAudioStream_dataCallback callback, void* pUserData); +typedef void (* MA_PFN_AAudioStreamBuilder_setErrorCallback) (ma_AAudioStreamBuilder* pBuilder, ma_AAudioStream_errorCallback callback, void* pUserData); +typedef void (* MA_PFN_AAudioStreamBuilder_setPerformanceMode) (ma_AAudioStreamBuilder* pBuilder, ma_aaudio_performance_mode_t mode); +typedef void (* MA_PFN_AAudioStreamBuilder_setUsage) (ma_AAudioStreamBuilder* pBuilder, ma_aaudio_usage_t contentType); +typedef void (* MA_PFN_AAudioStreamBuilder_setContentType) (ma_AAudioStreamBuilder* pBuilder, ma_aaudio_content_type_t contentType); +typedef void (* MA_PFN_AAudioStreamBuilder_setInputPreset) (ma_AAudioStreamBuilder* pBuilder, ma_aaudio_input_preset_t inputPreset); +typedef void (* MA_PFN_AAudioStreamBuilder_setAllowedCapturePolicy) (ma_AAudioStreamBuilder* pBuilder, ma_aaudio_allowed_capture_policy_t policy); +typedef ma_aaudio_result_t (* MA_PFN_AAudioStreamBuilder_openStream) (ma_AAudioStreamBuilder* pBuilder, ma_AAudioStream** ppStream); +typedef ma_aaudio_result_t (* MA_PFN_AAudioStream_close) (ma_AAudioStream* pStream); +typedef ma_aaudio_stream_state_t (* MA_PFN_AAudioStream_getState) (ma_AAudioStream* pStream); +typedef ma_aaudio_result_t (* MA_PFN_AAudioStream_waitForStateChange) (ma_AAudioStream* pStream, ma_aaudio_stream_state_t inputState, ma_aaudio_stream_state_t* pNextState, int64_t timeoutInNanoseconds); +typedef ma_aaudio_format_t (* MA_PFN_AAudioStream_getFormat) (ma_AAudioStream* pStream); +typedef int32_t (* MA_PFN_AAudioStream_getChannelCount) (ma_AAudioStream* pStream); +typedef int32_t (* MA_PFN_AAudioStream_getSampleRate) (ma_AAudioStream* pStream); +typedef int32_t (* MA_PFN_AAudioStream_getBufferCapacityInFrames) (ma_AAudioStream* pStream); +typedef int32_t (* MA_PFN_AAudioStream_getFramesPerDataCallback) (ma_AAudioStream* pStream); +typedef int32_t (* MA_PFN_AAudioStream_getFramesPerBurst) (ma_AAudioStream* pStream); +typedef ma_aaudio_result_t (* MA_PFN_AAudioStream_requestStart) (ma_AAudioStream* pStream); +typedef ma_aaudio_result_t (* MA_PFN_AAudioStream_requestStop) (ma_AAudioStream* pStream); + +static ma_result ma_result_from_aaudio(ma_aaudio_result_t resultAA) +{ + switch (resultAA) + { + case MA_AAUDIO_OK: return MA_SUCCESS; + default: break; + } + + return MA_ERROR; +} + +static ma_aaudio_usage_t ma_to_usage__aaudio(ma_aaudio_usage usage) +{ + switch (usage) { + case ma_aaudio_usage_media: return MA_AAUDIO_USAGE_MEDIA; + case ma_aaudio_usage_voice_communication: return MA_AAUDIO_USAGE_VOICE_COMMUNICATION; + case ma_aaudio_usage_voice_communication_signalling: return MA_AAUDIO_USAGE_VOICE_COMMUNICATION_SIGNALLING; + case ma_aaudio_usage_alarm: return MA_AAUDIO_USAGE_ALARM; + case ma_aaudio_usage_notification: return MA_AAUDIO_USAGE_NOTIFICATION; + case ma_aaudio_usage_notification_ringtone: return MA_AAUDIO_USAGE_NOTIFICATION_RINGTONE; + case ma_aaudio_usage_notification_event: return MA_AAUDIO_USAGE_NOTIFICATION_EVENT; + case ma_aaudio_usage_assistance_accessibility: return MA_AAUDIO_USAGE_ASSISTANCE_ACCESSIBILITY; + case ma_aaudio_usage_assistance_navigation_guidance: return MA_AAUDIO_USAGE_ASSISTANCE_NAVIGATION_GUIDANCE; + case ma_aaudio_usage_assistance_sonification: return MA_AAUDIO_USAGE_ASSISTANCE_SONIFICATION; + case ma_aaudio_usage_game: return MA_AAUDIO_USAGE_GAME; + case ma_aaudio_usage_assitant: return MA_AAUDIO_USAGE_ASSISTANT; + case ma_aaudio_usage_emergency: return MA_AAUDIO_SYSTEM_USAGE_EMERGENCY; + case ma_aaudio_usage_safety: return MA_AAUDIO_SYSTEM_USAGE_SAFETY; + case ma_aaudio_usage_vehicle_status: return MA_AAUDIO_SYSTEM_USAGE_VEHICLE_STATUS; + case ma_aaudio_usage_announcement: return MA_AAUDIO_SYSTEM_USAGE_ANNOUNCEMENT; + default: break; + } + + return MA_AAUDIO_USAGE_MEDIA; +} + +static ma_aaudio_content_type_t ma_to_content_type__aaudio(ma_aaudio_content_type contentType) +{ + switch (contentType) { + case ma_aaudio_content_type_speech: return MA_AAUDIO_CONTENT_TYPE_SPEECH; + case ma_aaudio_content_type_music: return MA_AAUDIO_CONTENT_TYPE_MUSIC; + case ma_aaudio_content_type_movie: return MA_AAUDIO_CONTENT_TYPE_MOVIE; + case ma_aaudio_content_type_sonification: return MA_AAUDIO_CONTENT_TYPE_SONIFICATION; + default: break; + } + + return MA_AAUDIO_CONTENT_TYPE_SPEECH; +} + +static ma_aaudio_input_preset_t ma_to_input_preset__aaudio(ma_aaudio_input_preset inputPreset) +{ + switch (inputPreset) { + case ma_aaudio_input_preset_generic: return MA_AAUDIO_INPUT_PRESET_GENERIC; + case ma_aaudio_input_preset_camcorder: return MA_AAUDIO_INPUT_PRESET_CAMCORDER; + case ma_aaudio_input_preset_voice_recognition: return MA_AAUDIO_INPUT_PRESET_VOICE_RECOGNITION; + case ma_aaudio_input_preset_voice_communication: return MA_AAUDIO_INPUT_PRESET_VOICE_COMMUNICATION; + case ma_aaudio_input_preset_unprocessed: return MA_AAUDIO_INPUT_PRESET_UNPROCESSED; + case ma_aaudio_input_preset_voice_performance: return MA_AAUDIO_INPUT_PRESET_VOICE_PERFORMANCE; + default: break; + } + + return MA_AAUDIO_INPUT_PRESET_GENERIC; +} + +static ma_aaudio_allowed_capture_policy_t ma_to_allowed_capture_policy__aaudio(ma_aaudio_allowed_capture_policy allowedCapturePolicy) +{ + switch (allowedCapturePolicy) { + case ma_aaudio_allow_capture_by_all: return MA_AAUDIO_ALLOW_CAPTURE_BY_ALL; + case ma_aaudio_allow_capture_by_system: return MA_AAUDIO_ALLOW_CAPTURE_BY_SYSTEM; + case ma_aaudio_allow_capture_by_none: return MA_AAUDIO_ALLOW_CAPTURE_BY_NONE; + default: break; + } + + return MA_AAUDIO_ALLOW_CAPTURE_BY_ALL; +} + +static void ma_stream_error_callback__aaudio(ma_AAudioStream* pStream, void* pUserData, ma_aaudio_result_t error) +{ + ma_result result; + ma_job job; + ma_device* pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + (void)error; + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[AAudio] ERROR CALLBACK: error=%d, AAudioStream_getState()=%d\n", error, ((MA_PFN_AAudioStream_getState)pDevice->pContext->aaudio.AAudioStream_getState)(pStream)); + /* + When we get an error, we'll assume that the stream is in an erroneous state and needs to be restarted. From the documentation, + we cannot do this from the error callback. Therefore we are going to use an event thread for the AAudio backend to do this + cleanly and safely. + */ + job = ma_job_init(MA_JOB_TYPE_DEVICE_AAUDIO_REROUTE); + job.data.device.aaudio.reroute.pDevice = pDevice; + + if (pStream == pDevice->aaudio.pStreamCapture) { + job.data.device.aaudio.reroute.deviceType = ma_device_type_capture; + } + else { + job.data.device.aaudio.reroute.deviceType = ma_device_type_playback; + } + + result = ma_device_job_thread_post(&pDevice->pContext->aaudio.jobThread, &job); + if (result != MA_SUCCESS) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[AAudio] Device Disconnected. Failed to post job for rerouting.\n"); + return; + } +} + +static ma_aaudio_data_callback_result_t ma_stream_data_callback_capture__aaudio(ma_AAudioStream* pStream, void* pUserData, void* pAudioData, int32_t frameCount) +{ + ma_device* pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + if (frameCount > 0) { + ma_device_handle_backend_data_callback(pDevice, NULL, pAudioData, (ma_uint32)frameCount); + } + + (void)pStream; + return MA_AAUDIO_CALLBACK_RESULT_CONTINUE; +} + +static ma_aaudio_data_callback_result_t ma_stream_data_callback_playback__aaudio(ma_AAudioStream* pStream, void* pUserData, void* pAudioData, int32_t frameCount) +{ + ma_device* pDevice = (ma_device*)pUserData; + MA_ASSERT(pDevice != NULL); + + /* + I've had a report that AAudio can sometimes post a frame count of 0. We need to check for that here + so we don't get any errors at a deeper level. I'm doing the same with the capture side for safety, + though I've not yet had any reports about that one. + */ + if (frameCount > 0) { + ma_device_handle_backend_data_callback(pDevice, pAudioData, NULL, (ma_uint32)frameCount); + } + + (void)pStream; + return MA_AAUDIO_CALLBACK_RESULT_CONTINUE; +} + +static ma_result ma_create_and_configure_AAudioStreamBuilder__aaudio(ma_context* pContext, const ma_device_id* pDeviceID, ma_device_type deviceType, ma_share_mode shareMode, const ma_device_descriptor* pDescriptor, const ma_device_config* pConfig, ma_device* pDevice, ma_AAudioStreamBuilder** ppBuilder) +{ + ma_AAudioStreamBuilder* pBuilder; + ma_aaudio_result_t resultAA; + + /* Safety. */ + *ppBuilder = NULL; + + resultAA = ((MA_PFN_AAudio_createStreamBuilder)pContext->aaudio.AAudio_createStreamBuilder)(&pBuilder); + if (resultAA != MA_AAUDIO_OK) { + return ma_result_from_aaudio(resultAA); + } + + if (pDeviceID != NULL) { + ((MA_PFN_AAudioStreamBuilder_setDeviceId)pContext->aaudio.AAudioStreamBuilder_setDeviceId)(pBuilder, pDeviceID->aaudio); + } + + ((MA_PFN_AAudioStreamBuilder_setDirection)pContext->aaudio.AAudioStreamBuilder_setDirection)(pBuilder, (deviceType == ma_device_type_playback) ? MA_AAUDIO_DIRECTION_OUTPUT : MA_AAUDIO_DIRECTION_INPUT); + ((MA_PFN_AAudioStreamBuilder_setSharingMode)pContext->aaudio.AAudioStreamBuilder_setSharingMode)(pBuilder, (shareMode == ma_share_mode_shared) ? MA_AAUDIO_SHARING_MODE_SHARED : MA_AAUDIO_SHARING_MODE_EXCLUSIVE); + + + /* If we have a device descriptor make sure we configure the stream builder to take our requested parameters. */ + if (pDescriptor != NULL) { + MA_ASSERT(pConfig != NULL); /* We must have a device config if we also have a descriptor. The config is required for AAudio specific configuration options. */ + + if (pDescriptor->sampleRate != 0) { + ((MA_PFN_AAudioStreamBuilder_setSampleRate)pContext->aaudio.AAudioStreamBuilder_setSampleRate)(pBuilder, pDescriptor->sampleRate); + } + + if (pDescriptor->channels != 0) { + ((MA_PFN_AAudioStreamBuilder_setChannelCount)pContext->aaudio.AAudioStreamBuilder_setChannelCount)(pBuilder, pDescriptor->channels); + } + + if (pDescriptor->format != ma_format_unknown) { + ((MA_PFN_AAudioStreamBuilder_setFormat)pContext->aaudio.AAudioStreamBuilder_setFormat)(pBuilder, (pDescriptor->format == ma_format_s16) ? MA_AAUDIO_FORMAT_PCM_I16 : MA_AAUDIO_FORMAT_PCM_FLOAT); + } + + + /* + There have been reports where setting the frames per data callback results in an error. + In particular, re-routing may inadvertently switch from low-latency mode, resulting in a less stable + stream from the legacy path (AudioStreamLegacy). To address this, we simply don't set the value. It + can still be set if it's explicitly requested via the aaudio.allowSetBufferCapacity variable in the + device config. + */ + if ((!pConfig->aaudio.enableCompatibilityWorkarounds || ma_android_sdk_version() > 30) && pConfig->aaudio.allowSetBufferCapacity) { + /* + AAudio is annoying when it comes to its buffer calculation stuff because it doesn't let you + retrieve the actual sample rate until after you've opened the stream. But you need to configure + the buffer capacity before you open the stream... :/ + + To solve, we're just going to assume MA_DEFAULT_SAMPLE_RATE (48000) and move on. + */ + ma_uint32 bufferCapacityInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptor, pDescriptor->sampleRate, pConfig->performanceProfile) * pDescriptor->periodCount; + + ((MA_PFN_AAudioStreamBuilder_setBufferCapacityInFrames)pContext->aaudio.AAudioStreamBuilder_setBufferCapacityInFrames)(pBuilder, bufferCapacityInFrames); + ((MA_PFN_AAudioStreamBuilder_setFramesPerDataCallback)pContext->aaudio.AAudioStreamBuilder_setFramesPerDataCallback)(pBuilder, bufferCapacityInFrames / pDescriptor->periodCount); + } + + if (deviceType == ma_device_type_capture) { + if (pConfig->aaudio.inputPreset != ma_aaudio_input_preset_default && pContext->aaudio.AAudioStreamBuilder_setInputPreset != NULL) { + ((MA_PFN_AAudioStreamBuilder_setInputPreset)pContext->aaudio.AAudioStreamBuilder_setInputPreset)(pBuilder, ma_to_input_preset__aaudio(pConfig->aaudio.inputPreset)); + } + + ((MA_PFN_AAudioStreamBuilder_setDataCallback)pContext->aaudio.AAudioStreamBuilder_setDataCallback)(pBuilder, ma_stream_data_callback_capture__aaudio, (void*)pDevice); + } else { + if (pConfig->aaudio.usage != ma_aaudio_usage_default && pContext->aaudio.AAudioStreamBuilder_setUsage != NULL) { + ((MA_PFN_AAudioStreamBuilder_setUsage)pContext->aaudio.AAudioStreamBuilder_setUsage)(pBuilder, ma_to_usage__aaudio(pConfig->aaudio.usage)); + } + + if (pConfig->aaudio.contentType != ma_aaudio_content_type_default && pContext->aaudio.AAudioStreamBuilder_setContentType != NULL) { + ((MA_PFN_AAudioStreamBuilder_setContentType)pContext->aaudio.AAudioStreamBuilder_setContentType)(pBuilder, ma_to_content_type__aaudio(pConfig->aaudio.contentType)); + } + + if (pConfig->aaudio.allowedCapturePolicy != ma_aaudio_allow_capture_default && pContext->aaudio.AAudioStreamBuilder_setAllowedCapturePolicy != NULL) { + ((MA_PFN_AAudioStreamBuilder_setAllowedCapturePolicy)pContext->aaudio.AAudioStreamBuilder_setAllowedCapturePolicy)(pBuilder, ma_to_allowed_capture_policy__aaudio(pConfig->aaudio.allowedCapturePolicy)); + } + + ((MA_PFN_AAudioStreamBuilder_setDataCallback)pContext->aaudio.AAudioStreamBuilder_setDataCallback)(pBuilder, ma_stream_data_callback_playback__aaudio, (void*)pDevice); + } + + /* + If we set AAUDIO_PERFORMANCE_MODE_LOW_LATENCY, we allow for MMAP (non-legacy path). + Since there's a mapping between miniaudio's performance profiles and AAudio's performance modes, let's use it. + Beware though, with a conservative performance profile, AAudio will indeed take the legacy path. + */ + ((MA_PFN_AAudioStreamBuilder_setPerformanceMode)pContext->aaudio.AAudioStreamBuilder_setPerformanceMode)(pBuilder, (pConfig->performanceProfile == ma_performance_profile_low_latency) ? MA_AAUDIO_PERFORMANCE_MODE_LOW_LATENCY : MA_AAUDIO_PERFORMANCE_MODE_NONE); + + /* We need to set an error callback to detect device changes. */ + if (pDevice != NULL) { /* <-- pDevice should never be null if pDescriptor is not null, which is always the case if we hit this branch. Check anyway for safety. */ + ((MA_PFN_AAudioStreamBuilder_setErrorCallback)pContext->aaudio.AAudioStreamBuilder_setErrorCallback)(pBuilder, ma_stream_error_callback__aaudio, (void*)pDevice); + } + } + + *ppBuilder = pBuilder; + + return MA_SUCCESS; +} + +static ma_result ma_open_stream_and_close_builder__aaudio(ma_context* pContext, ma_AAudioStreamBuilder* pBuilder, ma_AAudioStream** ppStream) +{ + ma_result result; + + result = ma_result_from_aaudio(((MA_PFN_AAudioStreamBuilder_openStream)pContext->aaudio.AAudioStreamBuilder_openStream)(pBuilder, ppStream)); + ((MA_PFN_AAudioStreamBuilder_delete)pContext->aaudio.AAudioStreamBuilder_delete)(pBuilder); + + return result; +} + +static ma_result ma_open_stream_basic__aaudio(ma_context* pContext, const ma_device_id* pDeviceID, ma_device_type deviceType, ma_share_mode shareMode, ma_AAudioStream** ppStream) +{ + ma_result result; + ma_AAudioStreamBuilder* pBuilder; + + *ppStream = NULL; + + result = ma_create_and_configure_AAudioStreamBuilder__aaudio(pContext, pDeviceID, deviceType, shareMode, NULL, NULL, NULL, &pBuilder); + if (result != MA_SUCCESS) { + return result; + } + + /* Let's give AAudio a hint to avoid the legacy path (AudioStreamLegacy). */ + ((MA_PFN_AAudioStreamBuilder_setPerformanceMode)pContext->aaudio.AAudioStreamBuilder_setPerformanceMode)(pBuilder, MA_AAUDIO_PERFORMANCE_MODE_LOW_LATENCY); + + return ma_open_stream_and_close_builder__aaudio(pContext, pBuilder, ppStream); +} + +static ma_result ma_open_stream__aaudio(ma_device* pDevice, const ma_device_config* pConfig, ma_device_type deviceType, const ma_device_descriptor* pDescriptor, ma_AAudioStream** ppStream) +{ + ma_result result; + ma_AAudioStreamBuilder* pBuilder; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pDescriptor != NULL); + MA_ASSERT(deviceType != ma_device_type_duplex); /* This function should not be called for a full-duplex device type. */ + + *ppStream = NULL; + + result = ma_create_and_configure_AAudioStreamBuilder__aaudio(pDevice->pContext, pDescriptor->pDeviceID, deviceType, pDescriptor->shareMode, pDescriptor, pConfig, pDevice, &pBuilder); + if (result != MA_SUCCESS) { + return result; + } + + return ma_open_stream_and_close_builder__aaudio(pDevice->pContext, pBuilder, ppStream); +} + +static ma_result ma_close_stream__aaudio(ma_context* pContext, ma_AAudioStream* pStream) +{ + if (pStream == NULL) { + return MA_INVALID_ARGS; + } + + return ma_result_from_aaudio(((MA_PFN_AAudioStream_close)pContext->aaudio.AAudioStream_close)(pStream)); +} + +static ma_bool32 ma_has_default_device__aaudio(ma_context* pContext, ma_device_type deviceType) +{ + /* The only way to know this is to try creating a stream. */ + ma_AAudioStream* pStream; + ma_result result = ma_open_stream_basic__aaudio(pContext, NULL, deviceType, ma_share_mode_shared, &pStream); + if (result != MA_SUCCESS) { + return MA_FALSE; + } + + ma_close_stream__aaudio(pContext, pStream); + return MA_TRUE; +} + +static ma_result ma_wait_for_simple_state_transition__aaudio(ma_context* pContext, ma_AAudioStream* pStream, ma_aaudio_stream_state_t oldState, ma_aaudio_stream_state_t newState) +{ + ma_aaudio_stream_state_t actualNewState; + ma_aaudio_result_t resultAA = ((MA_PFN_AAudioStream_waitForStateChange)pContext->aaudio.AAudioStream_waitForStateChange)(pStream, oldState, &actualNewState, 5000000000); /* 5 second timeout. */ + if (resultAA != MA_AAUDIO_OK) { + return ma_result_from_aaudio(resultAA); + } + + if (newState != actualNewState) { + return MA_ERROR; /* Failed to transition into the expected state. */ + } + + return MA_SUCCESS; +} + + +static ma_result ma_context_enumerate_devices__aaudio(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_bool32 cbResult = MA_TRUE; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + /* Unfortunately AAudio does not have an enumeration API. Therefore I'm only going to report default devices, but only if it can instantiate a stream. */ + + /* Playback. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + deviceInfo.id.aaudio = MA_AAUDIO_UNSPECIFIED; + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + + if (ma_has_default_device__aaudio(pContext, ma_device_type_playback)) { + cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + } + + /* Capture. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + deviceInfo.id.aaudio = MA_AAUDIO_UNSPECIFIED; + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + + if (ma_has_default_device__aaudio(pContext, ma_device_type_capture)) { + cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + } + + return MA_SUCCESS; +} + +static void ma_context_add_native_data_format_from_AAudioStream_ex__aaudio(ma_context* pContext, ma_AAudioStream* pStream, ma_format format, ma_uint32 flags, ma_device_info* pDeviceInfo) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pStream != NULL); + MA_ASSERT(pDeviceInfo != NULL); + + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].format = format; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].channels = ((MA_PFN_AAudioStream_getChannelCount)pContext->aaudio.AAudioStream_getChannelCount)(pStream); + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].sampleRate = ((MA_PFN_AAudioStream_getSampleRate)pContext->aaudio.AAudioStream_getSampleRate)(pStream); + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].flags = flags; + pDeviceInfo->nativeDataFormatCount += 1; +} + +static void ma_context_add_native_data_format_from_AAudioStream__aaudio(ma_context* pContext, ma_AAudioStream* pStream, ma_uint32 flags, ma_device_info* pDeviceInfo) +{ + /* AAudio supports s16 and f32. */ + ma_context_add_native_data_format_from_AAudioStream_ex__aaudio(pContext, pStream, ma_format_f32, flags, pDeviceInfo); + ma_context_add_native_data_format_from_AAudioStream_ex__aaudio(pContext, pStream, ma_format_s16, flags, pDeviceInfo); +} + +static ma_result ma_context_get_device_info__aaudio(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + ma_AAudioStream* pStream; + ma_result result; + + MA_ASSERT(pContext != NULL); + + /* ID */ + if (pDeviceID != NULL) { + pDeviceInfo->id.aaudio = pDeviceID->aaudio; + } else { + pDeviceInfo->id.aaudio = MA_AAUDIO_UNSPECIFIED; + } + + /* Name */ + if (deviceType == ma_device_type_playback) { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } + + + pDeviceInfo->nativeDataFormatCount = 0; + + /* We'll need to open the device to get accurate sample rate and channel count information. */ + result = ma_open_stream_basic__aaudio(pContext, pDeviceID, deviceType, ma_share_mode_shared, &pStream); + if (result != MA_SUCCESS) { + return result; + } + + ma_context_add_native_data_format_from_AAudioStream__aaudio(pContext, pStream, 0, pDeviceInfo); + + ma_close_stream__aaudio(pContext, pStream); + pStream = NULL; + + return MA_SUCCESS; +} + +static ma_result ma_close_streams__aaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + /* When re-routing, streams may have been closed and never re-opened. Hence the extra checks below. */ + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ma_close_stream__aaudio(pDevice->pContext, (ma_AAudioStream*)pDevice->aaudio.pStreamCapture); + pDevice->aaudio.pStreamCapture = NULL; + } + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_close_stream__aaudio(pDevice->pContext, (ma_AAudioStream*)pDevice->aaudio.pStreamPlayback); + pDevice->aaudio.pStreamPlayback = NULL; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_uninit__aaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + /* Wait for any rerouting to finish before attempting to close the streams. */ + ma_mutex_lock(&pDevice->aaudio.rerouteLock); + { + ma_close_streams__aaudio(pDevice); + } + ma_mutex_unlock(&pDevice->aaudio.rerouteLock); + + /* Destroy re-routing lock. */ + ma_mutex_uninit(&pDevice->aaudio.rerouteLock); + + return MA_SUCCESS; +} + +static ma_result ma_device_init_by_type__aaudio(ma_device* pDevice, const ma_device_config* pConfig, ma_device_type deviceType, ma_device_descriptor* pDescriptor, ma_AAudioStream** ppStream) +{ + ma_result result; + int32_t bufferCapacityInFrames; + int32_t framesPerDataCallback; + ma_AAudioStream* pStream; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDescriptor != NULL); + + *ppStream = NULL; /* Safety. */ + + /* First step is to open the stream. From there we'll be able to extract the internal configuration. */ + result = ma_open_stream__aaudio(pDevice, pConfig, deviceType, pDescriptor, &pStream); + if (result != MA_SUCCESS) { + return result; /* Failed to open the AAudio stream. */ + } + + /* Now extract the internal configuration. */ + pDescriptor->format = (((MA_PFN_AAudioStream_getFormat)pDevice->pContext->aaudio.AAudioStream_getFormat)(pStream) == MA_AAUDIO_FORMAT_PCM_I16) ? ma_format_s16 : ma_format_f32; + pDescriptor->channels = ((MA_PFN_AAudioStream_getChannelCount)pDevice->pContext->aaudio.AAudioStream_getChannelCount)(pStream); + pDescriptor->sampleRate = ((MA_PFN_AAudioStream_getSampleRate)pDevice->pContext->aaudio.AAudioStream_getSampleRate)(pStream); + + /* For the channel map we need to be sure we don't overflow any buffers. */ + if (pDescriptor->channels <= MA_MAX_CHANNELS) { + ma_channel_map_init_standard(ma_standard_channel_map_default, pDescriptor->channelMap, ma_countof(pDescriptor->channelMap), pDescriptor->channels); /* <-- Cannot find info on channel order, so assuming a default. */ + } else { + ma_channel_map_init_blank(pDescriptor->channelMap, MA_MAX_CHANNELS); /* Too many channels. Use a blank channel map. */ + } + + bufferCapacityInFrames = ((MA_PFN_AAudioStream_getBufferCapacityInFrames)pDevice->pContext->aaudio.AAudioStream_getBufferCapacityInFrames)(pStream); + framesPerDataCallback = ((MA_PFN_AAudioStream_getFramesPerDataCallback)pDevice->pContext->aaudio.AAudioStream_getFramesPerDataCallback)(pStream); + + if (framesPerDataCallback > 0) { + pDescriptor->periodSizeInFrames = framesPerDataCallback; + pDescriptor->periodCount = bufferCapacityInFrames / framesPerDataCallback; + } else { + pDescriptor->periodSizeInFrames = bufferCapacityInFrames; + pDescriptor->periodCount = 1; + } + + *ppStream = pStream; + + return MA_SUCCESS; +} + +static ma_result ma_device_init_streams__aaudio(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + pDevice->aaudio.usage = pConfig->aaudio.usage; + pDevice->aaudio.contentType = pConfig->aaudio.contentType; + pDevice->aaudio.inputPreset = pConfig->aaudio.inputPreset; + pDevice->aaudio.allowedCapturePolicy = pConfig->aaudio.allowedCapturePolicy; + pDevice->aaudio.noAutoStartAfterReroute = pConfig->aaudio.noAutoStartAfterReroute; + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + result = ma_device_init_by_type__aaudio(pDevice, pConfig, ma_device_type_capture, pDescriptorCapture, (ma_AAudioStream**)&pDevice->aaudio.pStreamCapture); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + result = ma_device_init_by_type__aaudio(pDevice, pConfig, ma_device_type_playback, pDescriptorPlayback, (ma_AAudioStream**)&pDevice->aaudio.pStreamPlayback); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_init__aaudio(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + + result = ma_device_init_streams__aaudio(pDevice, pConfig, pDescriptorPlayback, pDescriptorCapture); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_mutex_init(&pDevice->aaudio.rerouteLock); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_start_stream__aaudio(ma_device* pDevice, ma_AAudioStream* pStream) +{ + ma_aaudio_result_t resultAA; + ma_aaudio_stream_state_t currentState; + + MA_ASSERT(pDevice != NULL); + + if (pStream == NULL) { + return MA_INVALID_ARGS; + } + + resultAA = ((MA_PFN_AAudioStream_requestStart)pDevice->pContext->aaudio.AAudioStream_requestStart)(pStream); + if (resultAA != MA_AAUDIO_OK) { + return ma_result_from_aaudio(resultAA); + } + + /* Do we actually need to wait for the device to transition into its started state? */ + + /* The device should be in either a starting or started state. If it's not set to started we need to wait for it to transition. It should go from starting to started. */ + currentState = ((MA_PFN_AAudioStream_getState)pDevice->pContext->aaudio.AAudioStream_getState)(pStream); + if (currentState != MA_AAUDIO_STREAM_STATE_STARTED) { + ma_result result; + + if (currentState != MA_AAUDIO_STREAM_STATE_STARTING) { + return MA_ERROR; /* Expecting the stream to be a starting or started state. */ + } + + result = ma_wait_for_simple_state_transition__aaudio(pDevice->pContext, pStream, currentState, MA_AAUDIO_STREAM_STATE_STARTED); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop_stream__aaudio(ma_device* pDevice, ma_AAudioStream* pStream) +{ + ma_aaudio_result_t resultAA; + ma_aaudio_stream_state_t currentState; + + MA_ASSERT(pDevice != NULL); + + if (pStream == NULL) { + return MA_INVALID_ARGS; + } + + /* + From the AAudio documentation: + + The stream will stop after all of the data currently buffered has been played. + + This maps with miniaudio's requirement that device's be drained which means we don't need to implement any draining logic. + */ + currentState = ((MA_PFN_AAudioStream_getState)pDevice->pContext->aaudio.AAudioStream_getState)(pStream); + if (currentState == MA_AAUDIO_STREAM_STATE_DISCONNECTED) { + return MA_SUCCESS; /* The device is disconnected. Don't try stopping it. */ + } + + resultAA = ((MA_PFN_AAudioStream_requestStop)pDevice->pContext->aaudio.AAudioStream_requestStop)(pStream); + if (resultAA != MA_AAUDIO_OK) { + return ma_result_from_aaudio(resultAA); + } + + /* The device should be in either a stopping or stopped state. If it's not set to started we need to wait for it to transition. It should go from stopping to stopped. */ + currentState = ((MA_PFN_AAudioStream_getState)pDevice->pContext->aaudio.AAudioStream_getState)(pStream); + if (currentState != MA_AAUDIO_STREAM_STATE_STOPPED) { + ma_result result; + + if (currentState != MA_AAUDIO_STREAM_STATE_STOPPING) { + return MA_ERROR; /* Expecting the stream to be a stopping or stopped state. */ + } + + result = ma_wait_for_simple_state_transition__aaudio(pDevice->pContext, pStream, currentState, MA_AAUDIO_STREAM_STATE_STOPPED); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_start__aaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ma_result result = ma_device_start_stream__aaudio(pDevice, (ma_AAudioStream*)pDevice->aaudio.pStreamCapture); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_result result = ma_device_start_stream__aaudio(pDevice, (ma_AAudioStream*)pDevice->aaudio.pStreamPlayback); + if (result != MA_SUCCESS) { + if (pDevice->type == ma_device_type_duplex) { + ma_device_stop_stream__aaudio(pDevice, (ma_AAudioStream*)pDevice->aaudio.pStreamCapture); + } + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__aaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ma_result result = ma_device_stop_stream__aaudio(pDevice, (ma_AAudioStream*)pDevice->aaudio.pStreamCapture); + if (result != MA_SUCCESS) { + return result; + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_result result = ma_device_stop_stream__aaudio(pDevice, (ma_AAudioStream*)pDevice->aaudio.pStreamPlayback); + if (result != MA_SUCCESS) { + return result; + } + } + + ma_device__on_notification_stopped(pDevice); + + return MA_SUCCESS; +} + +static ma_result ma_device_reinit__aaudio(ma_device* pDevice, ma_device_type deviceType) +{ + ma_result result; + int32_t retries = 0; + + MA_ASSERT(pDevice != NULL); + + /* + TODO: Stop retrying if main thread is about to uninit device. + */ + ma_mutex_lock(&pDevice->aaudio.rerouteLock); + { +error_disconnected: + /* The first thing to do is close the streams. */ + ma_close_streams__aaudio(pDevice); + + /* Now we need to reinitialize each streams. The hardest part with this is just filling output the config and descriptors. */ + ma_device_config deviceConfig; + ma_device_descriptor descriptorPlayback; + ma_device_descriptor descriptorCapture; + + deviceConfig = ma_device_config_init(deviceType); + deviceConfig.playback.pDeviceID = NULL; /* Only doing rerouting with default devices. */ + deviceConfig.playback.shareMode = pDevice->playback.shareMode; + deviceConfig.playback.format = pDevice->playback.format; + deviceConfig.playback.channels = pDevice->playback.channels; + deviceConfig.capture.pDeviceID = NULL; /* Only doing rerouting with default devices. */ + deviceConfig.capture.shareMode = pDevice->capture.shareMode; + deviceConfig.capture.format = pDevice->capture.format; + deviceConfig.capture.channels = pDevice->capture.channels; + deviceConfig.sampleRate = pDevice->sampleRate; + deviceConfig.aaudio.usage = pDevice->aaudio.usage; + deviceConfig.aaudio.contentType = pDevice->aaudio.contentType; + deviceConfig.aaudio.inputPreset = pDevice->aaudio.inputPreset; + deviceConfig.aaudio.allowedCapturePolicy = pDevice->aaudio.allowedCapturePolicy; + deviceConfig.aaudio.noAutoStartAfterReroute = pDevice->aaudio.noAutoStartAfterReroute; + deviceConfig.periods = 1; + + /* Try to get an accurate period size. */ + if (deviceType == ma_device_type_playback || deviceType == ma_device_type_duplex) { + deviceConfig.periodSizeInFrames = pDevice->playback.internalPeriodSizeInFrames; + } else { + deviceConfig.periodSizeInFrames = pDevice->capture.internalPeriodSizeInFrames; + } + + if (deviceType == ma_device_type_capture || deviceType == ma_device_type_duplex || deviceType == ma_device_type_loopback) { + descriptorCapture.pDeviceID = deviceConfig.capture.pDeviceID; + descriptorCapture.shareMode = deviceConfig.capture.shareMode; + descriptorCapture.format = deviceConfig.capture.format; + descriptorCapture.channels = deviceConfig.capture.channels; + descriptorCapture.sampleRate = deviceConfig.sampleRate; + descriptorCapture.periodSizeInFrames = deviceConfig.periodSizeInFrames; + descriptorCapture.periodCount = deviceConfig.periods; + } + + if (deviceType == ma_device_type_playback || deviceType == ma_device_type_duplex) { + descriptorPlayback.pDeviceID = deviceConfig.playback.pDeviceID; + descriptorPlayback.shareMode = deviceConfig.playback.shareMode; + descriptorPlayback.format = deviceConfig.playback.format; + descriptorPlayback.channels = deviceConfig.playback.channels; + descriptorPlayback.sampleRate = deviceConfig.sampleRate; + descriptorPlayback.periodSizeInFrames = deviceConfig.periodSizeInFrames; + descriptorPlayback.periodCount = deviceConfig.periods; + } + + result = ma_device_init_streams__aaudio(pDevice, &deviceConfig, &descriptorPlayback, &descriptorCapture); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[AAudio] Failed to create stream after route change."); + goto done; + } + + result = ma_device_post_init(pDevice, deviceType, &descriptorPlayback, &descriptorCapture); + if (result != MA_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_WARNING, "[AAudio] Failed to initialize device after route change."); + ma_close_streams__aaudio(pDevice); + goto done; + } + + /* We'll only ever do this in response to a reroute. */ + ma_device__on_notification_rerouted(pDevice); + + /* If the device is started, start the streams. Maybe make this configurable? */ + if (ma_device_get_state(pDevice) == ma_device_state_started) { + if (pDevice->aaudio.noAutoStartAfterReroute == MA_FALSE) { + result = ma_device_start__aaudio(pDevice); + if (result != MA_SUCCESS) { + /* We got disconnected! Retry a few times, until we find a connected device! */ + retries += 1; + if (retries <= 3) { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[AAudio] Failed to start stream after route change, retrying(%d)", retries); + goto error_disconnected; + } + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[AAudio] Failed to start stream after route change."); + goto done; + } + } else { + ma_device_stop(pDevice); /* Do a full device stop so we set internal state correctly. */ + } + } + + result = MA_SUCCESS; + } +done: + /* Re-routing done */ + ma_mutex_unlock(&pDevice->aaudio.rerouteLock); + + return result; +} + +static ma_result ma_device_get_info__aaudio(ma_device* pDevice, ma_device_type type, ma_device_info* pDeviceInfo) +{ + ma_AAudioStream* pStream = NULL; + + MA_ASSERT(pDevice != NULL); + MA_ASSERT(type != ma_device_type_duplex); + MA_ASSERT(pDeviceInfo != NULL); + + if (type == ma_device_type_capture) { + pStream = (ma_AAudioStream*)pDevice->aaudio.pStreamCapture; + pDeviceInfo->id.aaudio = pDevice->capture.id.aaudio; + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); /* Only supporting default devices. */ + } + if (type == ma_device_type_playback) { + pStream = (ma_AAudioStream*)pDevice->aaudio.pStreamPlayback; + pDeviceInfo->id.aaudio = pDevice->playback.id.aaudio; + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); /* Only supporting default devices. */ + } + + /* Safety. Should never happen. */ + if (pStream == NULL) { + return MA_INVALID_OPERATION; + } + + pDeviceInfo->nativeDataFormatCount = 0; + ma_context_add_native_data_format_from_AAudioStream__aaudio(pDevice->pContext, pStream, 0, pDeviceInfo); + + return MA_SUCCESS; +} + + +static ma_result ma_context_uninit__aaudio(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_aaudio); + + ma_device_job_thread_uninit(&pContext->aaudio.jobThread, &pContext->allocationCallbacks); + + ma_dlclose(ma_context_get_log(pContext), pContext->aaudio.hAAudio); + pContext->aaudio.hAAudio = NULL; + + return MA_SUCCESS; +} + +static ma_result ma_context_init__aaudio(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ +#if !defined(MA_NO_RUNTIME_LINKING) + size_t i; + const char* libNames[] = { + "libaaudio.so" + }; + + for (i = 0; i < ma_countof(libNames); ++i) { + pContext->aaudio.hAAudio = ma_dlopen(ma_context_get_log(pContext), libNames[i]); + if (pContext->aaudio.hAAudio != NULL) { + break; + } + } + + if (pContext->aaudio.hAAudio == NULL) { + return MA_FAILED_TO_INIT_BACKEND; + } + + pContext->aaudio.AAudio_createStreamBuilder = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudio_createStreamBuilder"); + pContext->aaudio.AAudioStreamBuilder_delete = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_delete"); + pContext->aaudio.AAudioStreamBuilder_setDeviceId = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setDeviceId"); + pContext->aaudio.AAudioStreamBuilder_setDirection = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setDirection"); + pContext->aaudio.AAudioStreamBuilder_setSharingMode = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setSharingMode"); + pContext->aaudio.AAudioStreamBuilder_setFormat = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setFormat"); + pContext->aaudio.AAudioStreamBuilder_setChannelCount = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setChannelCount"); + pContext->aaudio.AAudioStreamBuilder_setSampleRate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setSampleRate"); + pContext->aaudio.AAudioStreamBuilder_setBufferCapacityInFrames = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setBufferCapacityInFrames"); + pContext->aaudio.AAudioStreamBuilder_setFramesPerDataCallback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setFramesPerDataCallback"); + pContext->aaudio.AAudioStreamBuilder_setDataCallback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setDataCallback"); + pContext->aaudio.AAudioStreamBuilder_setErrorCallback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setErrorCallback"); + pContext->aaudio.AAudioStreamBuilder_setPerformanceMode = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setPerformanceMode"); + pContext->aaudio.AAudioStreamBuilder_setUsage = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setUsage"); + pContext->aaudio.AAudioStreamBuilder_setContentType = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setContentType"); + pContext->aaudio.AAudioStreamBuilder_setInputPreset = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setInputPreset"); + pContext->aaudio.AAudioStreamBuilder_setAllowedCapturePolicy = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_setAllowedCapturePolicy"); + pContext->aaudio.AAudioStreamBuilder_openStream = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStreamBuilder_openStream"); + pContext->aaudio.AAudioStream_close = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_close"); + pContext->aaudio.AAudioStream_getState = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_getState"); + pContext->aaudio.AAudioStream_waitForStateChange = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_waitForStateChange"); + pContext->aaudio.AAudioStream_getFormat = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_getFormat"); + pContext->aaudio.AAudioStream_getChannelCount = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_getChannelCount"); + pContext->aaudio.AAudioStream_getSampleRate = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_getSampleRate"); + pContext->aaudio.AAudioStream_getBufferCapacityInFrames = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_getBufferCapacityInFrames"); + pContext->aaudio.AAudioStream_getFramesPerDataCallback = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_getFramesPerDataCallback"); + pContext->aaudio.AAudioStream_getFramesPerBurst = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_getFramesPerBurst"); + pContext->aaudio.AAudioStream_requestStart = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_requestStart"); + pContext->aaudio.AAudioStream_requestStop = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->aaudio.hAAudio, "AAudioStream_requestStop"); +#else + pContext->aaudio.AAudio_createStreamBuilder = (ma_proc)AAudio_createStreamBuilder; + pContext->aaudio.AAudioStreamBuilder_delete = (ma_proc)AAudioStreamBuilder_delete; + pContext->aaudio.AAudioStreamBuilder_setDeviceId = (ma_proc)AAudioStreamBuilder_setDeviceId; + pContext->aaudio.AAudioStreamBuilder_setDirection = (ma_proc)AAudioStreamBuilder_setDirection; + pContext->aaudio.AAudioStreamBuilder_setSharingMode = (ma_proc)AAudioStreamBuilder_setSharingMode; + pContext->aaudio.AAudioStreamBuilder_setFormat = (ma_proc)AAudioStreamBuilder_setFormat; + pContext->aaudio.AAudioStreamBuilder_setChannelCount = (ma_proc)AAudioStreamBuilder_setChannelCount; + pContext->aaudio.AAudioStreamBuilder_setSampleRate = (ma_proc)AAudioStreamBuilder_setSampleRate; + pContext->aaudio.AAudioStreamBuilder_setBufferCapacityInFrames = (ma_proc)AAudioStreamBuilder_setBufferCapacityInFrames; + pContext->aaudio.AAudioStreamBuilder_setFramesPerDataCallback = (ma_proc)AAudioStreamBuilder_setFramesPerDataCallback; + pContext->aaudio.AAudioStreamBuilder_setDataCallback = (ma_proc)AAudioStreamBuilder_setDataCallback; + pContext->aaudio.AAudioStreamBuilder_setErrorCallback = (ma_proc)AAudioStreamBuilder_setErrorCallback; + pContext->aaudio.AAudioStreamBuilder_setPerformanceMode = (ma_proc)AAudioStreamBuilder_setPerformanceMode; + pContext->aaudio.AAudioStreamBuilder_setUsage = (ma_proc)AAudioStreamBuilder_setUsage; + pContext->aaudio.AAudioStreamBuilder_setContentType = (ma_proc)AAudioStreamBuilder_setContentType; + pContext->aaudio.AAudioStreamBuilder_setInputPreset = (ma_proc)AAudioStreamBuilder_setInputPreset; + #if defined(__ANDROID_API__) && __ANDROID_API__ >= 29 + pContext->aaudio.AAudioStreamBuilder_setAllowedCapturePolicy = (ma_proc)AAudioStreamBuilder_setAllowedCapturePolicy; + #endif + pContext->aaudio.AAudioStreamBuilder_openStream = (ma_proc)AAudioStreamBuilder_openStream; + pContext->aaudio.AAudioStream_close = (ma_proc)AAudioStream_close; + pContext->aaudio.AAudioStream_getState = (ma_proc)AAudioStream_getState; + pContext->aaudio.AAudioStream_waitForStateChange = (ma_proc)AAudioStream_waitForStateChange; + pContext->aaudio.AAudioStream_getFormat = (ma_proc)AAudioStream_getFormat; + pContext->aaudio.AAudioStream_getChannelCount = (ma_proc)AAudioStream_getChannelCount; + pContext->aaudio.AAudioStream_getSampleRate = (ma_proc)AAudioStream_getSampleRate; + pContext->aaudio.AAudioStream_getBufferCapacityInFrames = (ma_proc)AAudioStream_getBufferCapacityInFrames; + pContext->aaudio.AAudioStream_getFramesPerDataCallback = (ma_proc)AAudioStream_getFramesPerDataCallback; + pContext->aaudio.AAudioStream_getFramesPerBurst = (ma_proc)AAudioStream_getFramesPerBurst; + pContext->aaudio.AAudioStream_requestStart = (ma_proc)AAudioStream_requestStart; + pContext->aaudio.AAudioStream_requestStop = (ma_proc)AAudioStream_requestStop; +#endif + + pCallbacks->onContextInit = ma_context_init__aaudio; + pCallbacks->onContextUninit = ma_context_uninit__aaudio; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__aaudio; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__aaudio; + pCallbacks->onDeviceInit = ma_device_init__aaudio; + pCallbacks->onDeviceUninit = ma_device_uninit__aaudio; + pCallbacks->onDeviceStart = ma_device_start__aaudio; + pCallbacks->onDeviceStop = ma_device_stop__aaudio; + pCallbacks->onDeviceRead = NULL; /* Not used because AAudio is asynchronous. */ + pCallbacks->onDeviceWrite = NULL; /* Not used because AAudio is asynchronous. */ + pCallbacks->onDeviceDataLoop = NULL; /* Not used because AAudio is asynchronous. */ + pCallbacks->onDeviceGetInfo = ma_device_get_info__aaudio; + + + /* We need a job thread so we can deal with rerouting. */ + { + ma_result result; + ma_device_job_thread_config jobThreadConfig; + + jobThreadConfig = ma_device_job_thread_config_init(); + + result = ma_device_job_thread_init(&jobThreadConfig, &pContext->allocationCallbacks, &pContext->aaudio.jobThread); + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->aaudio.hAAudio); + pContext->aaudio.hAAudio = NULL; + return result; + } + } + + + (void)pConfig; + return MA_SUCCESS; +} + +static ma_result ma_job_process__device__aaudio_reroute(ma_job* pJob) +{ + ma_result result; + ma_device* pDevice; + + MA_ASSERT(pJob != NULL); + + pDevice = (ma_device*)pJob->data.device.aaudio.reroute.pDevice; + MA_ASSERT(pDevice != NULL); + + /* Here is where we need to reroute the device. To do this we need to uninitialize the stream and reinitialize it. */ + result = ma_device_reinit__aaudio(pDevice, (ma_device_type)pJob->data.device.aaudio.reroute.deviceType); + if (result != MA_SUCCESS) { + /* + Getting here means we failed to reroute the device. The best thing I can think of here is to + just stop the device. + */ + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[AAudio] Stopping device due to reroute failure."); + ma_device_stop(pDevice); + return result; + } + + return MA_SUCCESS; +} +#else +/* Getting here means there is no AAudio backend so we need a no-op job implementation. */ +static ma_result ma_job_process__device__aaudio_reroute(ma_job* pJob) +{ + return ma_job_process__noop(pJob); +} +#endif /* AAudio */ + + +/****************************************************************************** + +OpenSL|ES Backend + +******************************************************************************/ +#ifdef MA_HAS_OPENSL +#include +#ifdef MA_ANDROID +#include +#endif + +typedef SLresult (SLAPIENTRY * ma_slCreateEngine_proc)(SLObjectItf* pEngine, SLuint32 numOptions, SLEngineOption* pEngineOptions, SLuint32 numInterfaces, SLInterfaceID* pInterfaceIds, SLboolean* pInterfaceRequired); + +/* OpenSL|ES has one-per-application objects :( */ +static SLObjectItf g_maEngineObjectSL = NULL; +static SLEngineItf g_maEngineSL = NULL; +static ma_uint32 g_maOpenSLInitCounter = 0; +static ma_spinlock g_maOpenSLSpinlock = 0; /* For init/uninit. */ + +#define MA_OPENSL_OBJ(p) (*((SLObjectItf)(p))) +#define MA_OPENSL_OUTPUTMIX(p) (*((SLOutputMixItf)(p))) +#define MA_OPENSL_PLAY(p) (*((SLPlayItf)(p))) +#define MA_OPENSL_RECORD(p) (*((SLRecordItf)(p))) + +#ifdef MA_ANDROID +#define MA_OPENSL_BUFFERQUEUE(p) (*((SLAndroidSimpleBufferQueueItf)(p))) +#else +#define MA_OPENSL_BUFFERQUEUE(p) (*((SLBufferQueueItf)(p))) +#endif + +static ma_result ma_result_from_OpenSL(SLuint32 result) +{ + switch (result) + { + case SL_RESULT_SUCCESS: return MA_SUCCESS; + case SL_RESULT_PRECONDITIONS_VIOLATED: return MA_ERROR; + case SL_RESULT_PARAMETER_INVALID: return MA_INVALID_ARGS; + case SL_RESULT_MEMORY_FAILURE: return MA_OUT_OF_MEMORY; + case SL_RESULT_RESOURCE_ERROR: return MA_INVALID_DATA; + case SL_RESULT_RESOURCE_LOST: return MA_ERROR; + case SL_RESULT_IO_ERROR: return MA_IO_ERROR; + case SL_RESULT_BUFFER_INSUFFICIENT: return MA_NO_SPACE; + case SL_RESULT_CONTENT_CORRUPTED: return MA_INVALID_DATA; + case SL_RESULT_CONTENT_UNSUPPORTED: return MA_FORMAT_NOT_SUPPORTED; + case SL_RESULT_CONTENT_NOT_FOUND: return MA_ERROR; + case SL_RESULT_PERMISSION_DENIED: return MA_ACCESS_DENIED; + case SL_RESULT_FEATURE_UNSUPPORTED: return MA_NOT_IMPLEMENTED; + case SL_RESULT_INTERNAL_ERROR: return MA_ERROR; + case SL_RESULT_UNKNOWN_ERROR: return MA_ERROR; + case SL_RESULT_OPERATION_ABORTED: return MA_ERROR; + case SL_RESULT_CONTROL_LOST: return MA_ERROR; + default: return MA_ERROR; + } +} + +/* Converts an individual OpenSL-style channel identifier (SL_SPEAKER_FRONT_LEFT, etc.) to miniaudio. */ +static ma_uint8 ma_channel_id_to_ma__opensl(SLuint32 id) +{ + switch (id) + { + case SL_SPEAKER_FRONT_LEFT: return MA_CHANNEL_FRONT_LEFT; + case SL_SPEAKER_FRONT_RIGHT: return MA_CHANNEL_FRONT_RIGHT; + case SL_SPEAKER_FRONT_CENTER: return MA_CHANNEL_FRONT_CENTER; + case SL_SPEAKER_LOW_FREQUENCY: return MA_CHANNEL_LFE; + case SL_SPEAKER_BACK_LEFT: return MA_CHANNEL_BACK_LEFT; + case SL_SPEAKER_BACK_RIGHT: return MA_CHANNEL_BACK_RIGHT; + case SL_SPEAKER_FRONT_LEFT_OF_CENTER: return MA_CHANNEL_FRONT_LEFT_CENTER; + case SL_SPEAKER_FRONT_RIGHT_OF_CENTER: return MA_CHANNEL_FRONT_RIGHT_CENTER; + case SL_SPEAKER_BACK_CENTER: return MA_CHANNEL_BACK_CENTER; + case SL_SPEAKER_SIDE_LEFT: return MA_CHANNEL_SIDE_LEFT; + case SL_SPEAKER_SIDE_RIGHT: return MA_CHANNEL_SIDE_RIGHT; + case SL_SPEAKER_TOP_CENTER: return MA_CHANNEL_TOP_CENTER; + case SL_SPEAKER_TOP_FRONT_LEFT: return MA_CHANNEL_TOP_FRONT_LEFT; + case SL_SPEAKER_TOP_FRONT_CENTER: return MA_CHANNEL_TOP_FRONT_CENTER; + case SL_SPEAKER_TOP_FRONT_RIGHT: return MA_CHANNEL_TOP_FRONT_RIGHT; + case SL_SPEAKER_TOP_BACK_LEFT: return MA_CHANNEL_TOP_BACK_LEFT; + case SL_SPEAKER_TOP_BACK_CENTER: return MA_CHANNEL_TOP_BACK_CENTER; + case SL_SPEAKER_TOP_BACK_RIGHT: return MA_CHANNEL_TOP_BACK_RIGHT; + default: return 0; + } +} + +/* Converts an individual miniaudio channel identifier (MA_CHANNEL_FRONT_LEFT, etc.) to OpenSL-style. */ +static SLuint32 ma_channel_id_to_opensl(ma_uint8 id) +{ + switch (id) + { + case MA_CHANNEL_MONO: return SL_SPEAKER_FRONT_CENTER; + case MA_CHANNEL_FRONT_LEFT: return SL_SPEAKER_FRONT_LEFT; + case MA_CHANNEL_FRONT_RIGHT: return SL_SPEAKER_FRONT_RIGHT; + case MA_CHANNEL_FRONT_CENTER: return SL_SPEAKER_FRONT_CENTER; + case MA_CHANNEL_LFE: return SL_SPEAKER_LOW_FREQUENCY; + case MA_CHANNEL_BACK_LEFT: return SL_SPEAKER_BACK_LEFT; + case MA_CHANNEL_BACK_RIGHT: return SL_SPEAKER_BACK_RIGHT; + case MA_CHANNEL_FRONT_LEFT_CENTER: return SL_SPEAKER_FRONT_LEFT_OF_CENTER; + case MA_CHANNEL_FRONT_RIGHT_CENTER: return SL_SPEAKER_FRONT_RIGHT_OF_CENTER; + case MA_CHANNEL_BACK_CENTER: return SL_SPEAKER_BACK_CENTER; + case MA_CHANNEL_SIDE_LEFT: return SL_SPEAKER_SIDE_LEFT; + case MA_CHANNEL_SIDE_RIGHT: return SL_SPEAKER_SIDE_RIGHT; + case MA_CHANNEL_TOP_CENTER: return SL_SPEAKER_TOP_CENTER; + case MA_CHANNEL_TOP_FRONT_LEFT: return SL_SPEAKER_TOP_FRONT_LEFT; + case MA_CHANNEL_TOP_FRONT_CENTER: return SL_SPEAKER_TOP_FRONT_CENTER; + case MA_CHANNEL_TOP_FRONT_RIGHT: return SL_SPEAKER_TOP_FRONT_RIGHT; + case MA_CHANNEL_TOP_BACK_LEFT: return SL_SPEAKER_TOP_BACK_LEFT; + case MA_CHANNEL_TOP_BACK_CENTER: return SL_SPEAKER_TOP_BACK_CENTER; + case MA_CHANNEL_TOP_BACK_RIGHT: return SL_SPEAKER_TOP_BACK_RIGHT; + default: return 0; + } +} + +/* Converts a channel mapping to an OpenSL-style channel mask. */ +static SLuint32 ma_channel_map_to_channel_mask__opensl(const ma_channel* pChannelMap, ma_uint32 channels) +{ + SLuint32 channelMask = 0; + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; ++iChannel) { + channelMask |= ma_channel_id_to_opensl(pChannelMap[iChannel]); + } + + return channelMask; +} + +/* Converts an OpenSL-style channel mask to a miniaudio channel map. */ +static void ma_channel_mask_to_channel_map__opensl(SLuint32 channelMask, ma_uint32 channels, ma_channel* pChannelMap) +{ + if (channels == 1 && channelMask == 0) { + pChannelMap[0] = MA_CHANNEL_MONO; + } else if (channels == 2 && channelMask == 0) { + pChannelMap[0] = MA_CHANNEL_FRONT_LEFT; + pChannelMap[1] = MA_CHANNEL_FRONT_RIGHT; + } else { + if (channels == 1 && (channelMask & SL_SPEAKER_FRONT_CENTER) != 0) { + pChannelMap[0] = MA_CHANNEL_MONO; + } else { + /* Just iterate over each bit. */ + ma_uint32 iChannel = 0; + ma_uint32 iBit; + for (iBit = 0; iBit < 32 && iChannel < channels; ++iBit) { + SLuint32 bitValue = (channelMask & (1UL << iBit)); + if (bitValue != 0) { + /* The bit is set. */ + pChannelMap[iChannel] = ma_channel_id_to_ma__opensl(bitValue); + iChannel += 1; + } + } + } + } +} + +static SLuint32 ma_round_to_standard_sample_rate__opensl(SLuint32 samplesPerSec) +{ + if (samplesPerSec <= SL_SAMPLINGRATE_8) { + return SL_SAMPLINGRATE_8; + } + if (samplesPerSec <= SL_SAMPLINGRATE_11_025) { + return SL_SAMPLINGRATE_11_025; + } + if (samplesPerSec <= SL_SAMPLINGRATE_12) { + return SL_SAMPLINGRATE_12; + } + if (samplesPerSec <= SL_SAMPLINGRATE_16) { + return SL_SAMPLINGRATE_16; + } + if (samplesPerSec <= SL_SAMPLINGRATE_22_05) { + return SL_SAMPLINGRATE_22_05; + } + if (samplesPerSec <= SL_SAMPLINGRATE_24) { + return SL_SAMPLINGRATE_24; + } + if (samplesPerSec <= SL_SAMPLINGRATE_32) { + return SL_SAMPLINGRATE_32; + } + if (samplesPerSec <= SL_SAMPLINGRATE_44_1) { + return SL_SAMPLINGRATE_44_1; + } + if (samplesPerSec <= SL_SAMPLINGRATE_48) { + return SL_SAMPLINGRATE_48; + } + + /* Android doesn't support more than 48000. */ +#ifndef MA_ANDROID + if (samplesPerSec <= SL_SAMPLINGRATE_64) { + return SL_SAMPLINGRATE_64; + } + if (samplesPerSec <= SL_SAMPLINGRATE_88_2) { + return SL_SAMPLINGRATE_88_2; + } + if (samplesPerSec <= SL_SAMPLINGRATE_96) { + return SL_SAMPLINGRATE_96; + } + if (samplesPerSec <= SL_SAMPLINGRATE_192) { + return SL_SAMPLINGRATE_192; + } +#endif + + return SL_SAMPLINGRATE_16; +} + + +static SLint32 ma_to_stream_type__opensl(ma_opensl_stream_type streamType) +{ + switch (streamType) { + case ma_opensl_stream_type_voice: return SL_ANDROID_STREAM_VOICE; + case ma_opensl_stream_type_system: return SL_ANDROID_STREAM_SYSTEM; + case ma_opensl_stream_type_ring: return SL_ANDROID_STREAM_RING; + case ma_opensl_stream_type_media: return SL_ANDROID_STREAM_MEDIA; + case ma_opensl_stream_type_alarm: return SL_ANDROID_STREAM_ALARM; + case ma_opensl_stream_type_notification: return SL_ANDROID_STREAM_NOTIFICATION; + default: break; + } + + return SL_ANDROID_STREAM_VOICE; +} + +static SLint32 ma_to_recording_preset__opensl(ma_opensl_recording_preset recordingPreset) +{ + switch (recordingPreset) { + case ma_opensl_recording_preset_generic: return SL_ANDROID_RECORDING_PRESET_GENERIC; + case ma_opensl_recording_preset_camcorder: return SL_ANDROID_RECORDING_PRESET_CAMCORDER; + case ma_opensl_recording_preset_voice_recognition: return SL_ANDROID_RECORDING_PRESET_VOICE_RECOGNITION; + case ma_opensl_recording_preset_voice_communication: return SL_ANDROID_RECORDING_PRESET_VOICE_COMMUNICATION; + case ma_opensl_recording_preset_voice_unprocessed: return SL_ANDROID_RECORDING_PRESET_UNPROCESSED; + default: break; + } + + return SL_ANDROID_RECORDING_PRESET_NONE; +} + + +static ma_result ma_context_enumerate_devices__opensl(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_bool32 cbResult; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + MA_ASSERT(g_maOpenSLInitCounter > 0); /* <-- If you trigger this it means you've either not initialized the context, or you've uninitialized it and then attempted to enumerate devices. */ + if (g_maOpenSLInitCounter == 0) { + return MA_INVALID_OPERATION; + } + + /* + TODO: Test Me. + + This is currently untested, so for now we are just returning default devices. + */ +#if 0 && !defined(MA_ANDROID) + ma_bool32 isTerminated = MA_FALSE; + + SLuint32 pDeviceIDs[128]; + SLint32 deviceCount = sizeof(pDeviceIDs) / sizeof(pDeviceIDs[0]); + + SLAudioIODeviceCapabilitiesItf deviceCaps; + SLresult resultSL = (*g_maEngineObjectSL)->GetInterface(g_maEngineObjectSL, (SLInterfaceID)pContext->opensl.SL_IID_AUDIOIODEVICECAPABILITIES, &deviceCaps); + if (resultSL != SL_RESULT_SUCCESS) { + /* The interface may not be supported so just report a default device. */ + goto return_default_device; + } + + /* Playback */ + if (!isTerminated) { + resultSL = (*deviceCaps)->GetAvailableAudioOutputs(deviceCaps, &deviceCount, pDeviceIDs); + if (resultSL != SL_RESULT_SUCCESS) { + return ma_result_from_OpenSL(resultSL); + } + + for (SLint32 iDevice = 0; iDevice < deviceCount; ++iDevice) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + deviceInfo.id.opensl = pDeviceIDs[iDevice]; + + SLAudioOutputDescriptor desc; + resultSL = (*deviceCaps)->QueryAudioOutputCapabilities(deviceCaps, deviceInfo.id.opensl, &desc); + if (resultSL == SL_RESULT_SUCCESS) { + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), (const char*)desc.pDeviceName, (size_t)-1); + + ma_bool32 cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + if (cbResult == MA_FALSE) { + isTerminated = MA_TRUE; + break; + } + } + } + } + + /* Capture */ + if (!isTerminated) { + resultSL = (*deviceCaps)->GetAvailableAudioInputs(deviceCaps, &deviceCount, pDeviceIDs); + if (resultSL != SL_RESULT_SUCCESS) { + return ma_result_from_OpenSL(resultSL); + } + + for (SLint32 iDevice = 0; iDevice < deviceCount; ++iDevice) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + deviceInfo.id.opensl = pDeviceIDs[iDevice]; + + SLAudioInputDescriptor desc; + resultSL = (*deviceCaps)->QueryAudioInputCapabilities(deviceCaps, deviceInfo.id.opensl, &desc); + if (resultSL == SL_RESULT_SUCCESS) { + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), (const char*)desc.deviceName, (size_t)-1); + + ma_bool32 cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + if (cbResult == MA_FALSE) { + isTerminated = MA_TRUE; + break; + } + } + } + } + + return MA_SUCCESS; +#else + goto return_default_device; +#endif + +return_default_device:; + cbResult = MA_TRUE; + + /* Playback. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + deviceInfo.id.opensl = SL_DEFAULTDEVICEID_AUDIOOUTPUT; + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + + /* Capture. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + deviceInfo.id.opensl = SL_DEFAULTDEVICEID_AUDIOINPUT; + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + + return MA_SUCCESS; +} + +static void ma_context_add_data_format_ex__opensl(ma_context* pContext, ma_format format, ma_uint32 channels, ma_uint32 sampleRate, ma_device_info* pDeviceInfo) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pDeviceInfo != NULL); + + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].format = format; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].channels = channels; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].sampleRate = sampleRate; + pDeviceInfo->nativeDataFormats[pDeviceInfo->nativeDataFormatCount].flags = 0; + pDeviceInfo->nativeDataFormatCount += 1; +} + +static void ma_context_add_data_format__opensl(ma_context* pContext, ma_format format, ma_device_info* pDeviceInfo) +{ + ma_uint32 minChannels = 1; + ma_uint32 maxChannels = 2; + ma_uint32 minSampleRate = (ma_uint32)ma_standard_sample_rate_8000; + ma_uint32 maxSampleRate = (ma_uint32)ma_standard_sample_rate_48000; + ma_uint32 iChannel; + ma_uint32 iSampleRate; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(pDeviceInfo != NULL); + + /* + Each sample format can support mono and stereo, and we'll support a small subset of standard + rates (up to 48000). A better solution would be to somehow find a native sample rate. + */ + for (iChannel = minChannels; iChannel < maxChannels; iChannel += 1) { + for (iSampleRate = 0; iSampleRate < ma_countof(g_maStandardSampleRatePriorities); iSampleRate += 1) { + ma_uint32 standardSampleRate = g_maStandardSampleRatePriorities[iSampleRate]; + if (standardSampleRate >= minSampleRate && standardSampleRate <= maxSampleRate) { + ma_context_add_data_format_ex__opensl(pContext, format, iChannel, standardSampleRate, pDeviceInfo); + } + } + } +} + +static ma_result ma_context_get_device_info__opensl(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + MA_ASSERT(pContext != NULL); + + MA_ASSERT(g_maOpenSLInitCounter > 0); /* <-- If you trigger this it means you've either not initialized the context, or you've uninitialized it and then attempted to get device info. */ + if (g_maOpenSLInitCounter == 0) { + return MA_INVALID_OPERATION; + } + + /* + TODO: Test Me. + + This is currently untested, so for now we are just returning default devices. + */ +#if 0 && !defined(MA_ANDROID) + SLAudioIODeviceCapabilitiesItf deviceCaps; + SLresult resultSL = (*g_maEngineObjectSL)->GetInterface(g_maEngineObjectSL, (SLInterfaceID)pContext->opensl.SL_IID_AUDIOIODEVICECAPABILITIES, &deviceCaps); + if (resultSL != SL_RESULT_SUCCESS) { + /* The interface may not be supported so just report a default device. */ + goto return_default_device; + } + + if (deviceType == ma_device_type_playback) { + SLAudioOutputDescriptor desc; + resultSL = (*deviceCaps)->QueryAudioOutputCapabilities(deviceCaps, pDeviceID->opensl, &desc); + if (resultSL != SL_RESULT_SUCCESS) { + return ma_result_from_OpenSL(resultSL); + } + + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), (const char*)desc.pDeviceName, (size_t)-1); + } else { + SLAudioInputDescriptor desc; + resultSL = (*deviceCaps)->QueryAudioInputCapabilities(deviceCaps, pDeviceID->opensl, &desc); + if (resultSL != SL_RESULT_SUCCESS) { + return ma_result_from_OpenSL(resultSL); + } + + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), (const char*)desc.deviceName, (size_t)-1); + } + + goto return_detailed_info; +#else + goto return_default_device; +#endif + +return_default_device: + if (pDeviceID != NULL) { + if ((deviceType == ma_device_type_playback && pDeviceID->opensl != SL_DEFAULTDEVICEID_AUDIOOUTPUT) || + (deviceType == ma_device_type_capture && pDeviceID->opensl != SL_DEFAULTDEVICEID_AUDIOINPUT)) { + return MA_NO_DEVICE; /* Don't know the device. */ + } + } + + /* ID and Name / Description */ + if (deviceType == ma_device_type_playback) { + pDeviceInfo->id.opensl = SL_DEFAULTDEVICEID_AUDIOOUTPUT; + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + pDeviceInfo->id.opensl = SL_DEFAULTDEVICEID_AUDIOINPUT; + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } + + pDeviceInfo->isDefault = MA_TRUE; + + goto return_detailed_info; + + +return_detailed_info: + + /* + For now we're just outputting a set of values that are supported by the API but not necessarily supported + by the device natively. Later on we should work on this so that it more closely reflects the device's + actual native format. + */ + pDeviceInfo->nativeDataFormatCount = 0; +#if defined(MA_ANDROID) && __ANDROID_API__ >= 21 + ma_context_add_data_format__opensl(pContext, ma_format_f32, pDeviceInfo); +#endif + ma_context_add_data_format__opensl(pContext, ma_format_s16, pDeviceInfo); + ma_context_add_data_format__opensl(pContext, ma_format_u8, pDeviceInfo); + + return MA_SUCCESS; +} + + +#ifdef MA_ANDROID +/*void ma_buffer_queue_callback_capture__opensl_android(SLAndroidSimpleBufferQueueItf pBufferQueue, SLuint32 eventFlags, const void* pBuffer, SLuint32 bufferSize, SLuint32 dataUsed, void* pContext)*/ +static void ma_buffer_queue_callback_capture__opensl_android(SLAndroidSimpleBufferQueueItf pBufferQueue, void* pUserData) +{ + ma_device* pDevice = (ma_device*)pUserData; + size_t periodSizeInBytes; + ma_uint8* pBuffer; + SLresult resultSL; + + MA_ASSERT(pDevice != NULL); + + (void)pBufferQueue; + + /* + For now, don't do anything unless the buffer was fully processed. From what I can tell, it looks like + OpenSL|ES 1.1 improves on buffer queues to the point that we could much more intelligently handle this, + but unfortunately it looks like Android is only supporting OpenSL|ES 1.0.1 for now :( + */ + + /* Don't do anything if the device is not started. */ + if (ma_device_get_state(pDevice) != ma_device_state_started) { + return; + } + + /* Don't do anything if the device is being drained. */ + if (pDevice->opensl.isDrainingCapture) { + return; + } + + periodSizeInBytes = pDevice->capture.internalPeriodSizeInFrames * ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + pBuffer = pDevice->opensl.pBufferCapture + (pDevice->opensl.currentBufferIndexCapture * periodSizeInBytes); + + ma_device_handle_backend_data_callback(pDevice, NULL, pBuffer, pDevice->capture.internalPeriodSizeInFrames); + + resultSL = MA_OPENSL_BUFFERQUEUE(pDevice->opensl.pBufferQueueCapture)->Enqueue((SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueueCapture, pBuffer, periodSizeInBytes); + if (resultSL != SL_RESULT_SUCCESS) { + return; + } + + pDevice->opensl.currentBufferIndexCapture = (pDevice->opensl.currentBufferIndexCapture + 1) % pDevice->capture.internalPeriods; +} + +static void ma_buffer_queue_callback_playback__opensl_android(SLAndroidSimpleBufferQueueItf pBufferQueue, void* pUserData) +{ + ma_device* pDevice = (ma_device*)pUserData; + size_t periodSizeInBytes; + ma_uint8* pBuffer; + SLresult resultSL; + + MA_ASSERT(pDevice != NULL); + + (void)pBufferQueue; + + /* Don't do anything if the device is not started. */ + if (ma_device_get_state(pDevice) != ma_device_state_started) { + return; + } + + /* Don't do anything if the device is being drained. */ + if (pDevice->opensl.isDrainingPlayback) { + return; + } + + periodSizeInBytes = pDevice->playback.internalPeriodSizeInFrames * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + pBuffer = pDevice->opensl.pBufferPlayback + (pDevice->opensl.currentBufferIndexPlayback * periodSizeInBytes); + + ma_device_handle_backend_data_callback(pDevice, pBuffer, NULL, pDevice->playback.internalPeriodSizeInFrames); + + resultSL = MA_OPENSL_BUFFERQUEUE(pDevice->opensl.pBufferQueuePlayback)->Enqueue((SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueuePlayback, pBuffer, periodSizeInBytes); + if (resultSL != SL_RESULT_SUCCESS) { + return; + } + + pDevice->opensl.currentBufferIndexPlayback = (pDevice->opensl.currentBufferIndexPlayback + 1) % pDevice->playback.internalPeriods; +} +#endif + +static ma_result ma_device_uninit__opensl(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + MA_ASSERT(g_maOpenSLInitCounter > 0); /* <-- If you trigger this it means you've either not initialized the context, or you've uninitialized it before uninitializing the device. */ + if (g_maOpenSLInitCounter == 0) { + return MA_INVALID_OPERATION; + } + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + if (pDevice->opensl.pAudioRecorderObj) { + MA_OPENSL_OBJ(pDevice->opensl.pAudioRecorderObj)->Destroy((SLObjectItf)pDevice->opensl.pAudioRecorderObj); + } + + ma_free(pDevice->opensl.pBufferCapture, &pDevice->pContext->allocationCallbacks); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + if (pDevice->opensl.pAudioPlayerObj) { + MA_OPENSL_OBJ(pDevice->opensl.pAudioPlayerObj)->Destroy((SLObjectItf)pDevice->opensl.pAudioPlayerObj); + } + if (pDevice->opensl.pOutputMixObj) { + MA_OPENSL_OBJ(pDevice->opensl.pOutputMixObj)->Destroy((SLObjectItf)pDevice->opensl.pOutputMixObj); + } + + ma_free(pDevice->opensl.pBufferPlayback, &pDevice->pContext->allocationCallbacks); + } + + return MA_SUCCESS; +} + +#if defined(MA_ANDROID) && __ANDROID_API__ >= 21 +typedef SLAndroidDataFormat_PCM_EX ma_SLDataFormat_PCM; +#else +typedef SLDataFormat_PCM ma_SLDataFormat_PCM; +#endif + +static ma_result ma_SLDataFormat_PCM_init__opensl(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, const ma_channel* channelMap, ma_SLDataFormat_PCM* pDataFormat) +{ + /* We need to convert our format/channels/rate so that they aren't set to default. */ + if (format == ma_format_unknown) { + format = MA_DEFAULT_FORMAT; + } + if (channels == 0) { + channels = MA_DEFAULT_CHANNELS; + } + if (sampleRate == 0) { + sampleRate = MA_DEFAULT_SAMPLE_RATE; + } + +#if defined(MA_ANDROID) && __ANDROID_API__ >= 21 + if (format == ma_format_f32) { + pDataFormat->formatType = SL_ANDROID_DATAFORMAT_PCM_EX; + pDataFormat->representation = SL_ANDROID_PCM_REPRESENTATION_FLOAT; + } else { + pDataFormat->formatType = SL_DATAFORMAT_PCM; + } +#else + pDataFormat->formatType = SL_DATAFORMAT_PCM; +#endif + + pDataFormat->numChannels = channels; + ((SLDataFormat_PCM*)pDataFormat)->samplesPerSec = ma_round_to_standard_sample_rate__opensl(sampleRate * 1000); /* In millihertz. Annoyingly, the sample rate variable is named differently between SLAndroidDataFormat_PCM_EX and SLDataFormat_PCM */ + pDataFormat->bitsPerSample = ma_get_bytes_per_sample(format) * 8; + pDataFormat->channelMask = ma_channel_map_to_channel_mask__opensl(channelMap, channels); + pDataFormat->endianness = (ma_is_little_endian()) ? SL_BYTEORDER_LITTLEENDIAN : SL_BYTEORDER_BIGENDIAN; + + /* + Android has a few restrictions on the format as documented here: https://developer.android.com/ndk/guides/audio/opensl-for-android.html + - Only mono and stereo is supported. + - Only u8 and s16 formats are supported. + - Maximum sample rate of 48000. + */ +#ifdef MA_ANDROID + if (pDataFormat->numChannels > 2) { + pDataFormat->numChannels = 2; + } +#if __ANDROID_API__ >= 21 + if (pDataFormat->formatType == SL_ANDROID_DATAFORMAT_PCM_EX) { + /* It's floating point. */ + MA_ASSERT(pDataFormat->representation == SL_ANDROID_PCM_REPRESENTATION_FLOAT); + if (pDataFormat->bitsPerSample > 32) { + pDataFormat->bitsPerSample = 32; + } + } else { + if (pDataFormat->bitsPerSample > 16) { + pDataFormat->bitsPerSample = 16; + } + } +#else + if (pDataFormat->bitsPerSample > 16) { + pDataFormat->bitsPerSample = 16; + } +#endif + if (((SLDataFormat_PCM*)pDataFormat)->samplesPerSec > SL_SAMPLINGRATE_48) { + ((SLDataFormat_PCM*)pDataFormat)->samplesPerSec = SL_SAMPLINGRATE_48; + } +#endif + + pDataFormat->containerSize = pDataFormat->bitsPerSample; /* Always tightly packed for now. */ + + return MA_SUCCESS; +} + +static ma_result ma_deconstruct_SLDataFormat_PCM__opensl(ma_SLDataFormat_PCM* pDataFormat, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + ma_bool32 isFloatingPoint = MA_FALSE; +#if defined(MA_ANDROID) && __ANDROID_API__ >= 21 + if (pDataFormat->formatType == SL_ANDROID_DATAFORMAT_PCM_EX) { + MA_ASSERT(pDataFormat->representation == SL_ANDROID_PCM_REPRESENTATION_FLOAT); + isFloatingPoint = MA_TRUE; + } +#endif + if (isFloatingPoint) { + if (pDataFormat->bitsPerSample == 32) { + *pFormat = ma_format_f32; + } + } else { + if (pDataFormat->bitsPerSample == 8) { + *pFormat = ma_format_u8; + } else if (pDataFormat->bitsPerSample == 16) { + *pFormat = ma_format_s16; + } else if (pDataFormat->bitsPerSample == 24) { + *pFormat = ma_format_s24; + } else if (pDataFormat->bitsPerSample == 32) { + *pFormat = ma_format_s32; + } + } + + *pChannels = pDataFormat->numChannels; + *pSampleRate = ((SLDataFormat_PCM*)pDataFormat)->samplesPerSec / 1000; + ma_channel_mask_to_channel_map__opensl(pDataFormat->channelMask, ma_min(pDataFormat->numChannels, channelMapCap), pChannelMap); + + return MA_SUCCESS; +} + +static ma_result ma_device_init__opensl(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ +#ifdef MA_ANDROID + SLDataLocator_AndroidSimpleBufferQueue queue; + SLresult resultSL; + size_t bufferSizeInBytes; + SLInterfaceID itfIDs[2]; + const SLboolean itfIDsRequired[] = { + SL_BOOLEAN_TRUE, /* SL_IID_ANDROIDSIMPLEBUFFERQUEUE */ + SL_BOOLEAN_FALSE /* SL_IID_ANDROIDCONFIGURATION */ + }; +#endif + + MA_ASSERT(g_maOpenSLInitCounter > 0); /* <-- If you trigger this it means you've either not initialized the context, or you've uninitialized it and then attempted to initialize a new device. */ + if (g_maOpenSLInitCounter == 0) { + return MA_INVALID_OPERATION; + } + + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + /* + For now, only supporting Android implementations of OpenSL|ES since that's the only one I've + been able to test with and I currently depend on Android-specific extensions (simple buffer + queues). + */ +#ifdef MA_ANDROID + itfIDs[0] = (SLInterfaceID)pDevice->pContext->opensl.SL_IID_ANDROIDSIMPLEBUFFERQUEUE; + itfIDs[1] = (SLInterfaceID)pDevice->pContext->opensl.SL_IID_ANDROIDCONFIGURATION; + + /* No exclusive mode with OpenSL|ES. */ + if (((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pDescriptorPlayback->shareMode == ma_share_mode_exclusive) || + ((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) && pDescriptorCapture->shareMode == ma_share_mode_exclusive)) { + return MA_SHARE_MODE_NOT_SUPPORTED; + } + + /* Now we can start initializing the device properly. */ + MA_ASSERT(pDevice != NULL); + MA_ZERO_OBJECT(&pDevice->opensl); + + queue.locatorType = SL_DATALOCATOR_ANDROIDSIMPLEBUFFERQUEUE; + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + ma_SLDataFormat_PCM pcm; + SLDataLocator_IODevice locatorDevice; + SLDataSource source; + SLDataSink sink; + SLAndroidConfigurationItf pRecorderConfig; + + ma_SLDataFormat_PCM_init__opensl(pDescriptorCapture->format, pDescriptorCapture->channels, pDescriptorCapture->sampleRate, pDescriptorCapture->channelMap, &pcm); + + locatorDevice.locatorType = SL_DATALOCATOR_IODEVICE; + locatorDevice.deviceType = SL_IODEVICE_AUDIOINPUT; + locatorDevice.deviceID = SL_DEFAULTDEVICEID_AUDIOINPUT; /* Must always use the default device with Android. */ + locatorDevice.device = NULL; + + source.pLocator = &locatorDevice; + source.pFormat = NULL; + + queue.numBuffers = pDescriptorCapture->periodCount; + + sink.pLocator = &queue; + sink.pFormat = (SLDataFormat_PCM*)&pcm; + + resultSL = (*g_maEngineSL)->CreateAudioRecorder(g_maEngineSL, (SLObjectItf*)&pDevice->opensl.pAudioRecorderObj, &source, &sink, ma_countof(itfIDs), itfIDs, itfIDsRequired); + if (resultSL == SL_RESULT_CONTENT_UNSUPPORTED || resultSL == SL_RESULT_PARAMETER_INVALID) { + /* Unsupported format. Fall back to something safer and try again. If this fails, just abort. */ + pcm.formatType = SL_DATAFORMAT_PCM; + pcm.numChannels = 1; + ((SLDataFormat_PCM*)&pcm)->samplesPerSec = SL_SAMPLINGRATE_16; /* The name of the sample rate variable is different between SLAndroidDataFormat_PCM_EX and SLDataFormat_PCM. */ + pcm.bitsPerSample = 16; + pcm.containerSize = pcm.bitsPerSample; /* Always tightly packed for now. */ + pcm.channelMask = 0; + resultSL = (*g_maEngineSL)->CreateAudioRecorder(g_maEngineSL, (SLObjectItf*)&pDevice->opensl.pAudioRecorderObj, &source, &sink, ma_countof(itfIDs), itfIDs, itfIDsRequired); + } + + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to create audio recorder."); + return ma_result_from_OpenSL(resultSL); + } + + + /* Set the recording preset before realizing the player. */ + if (pConfig->opensl.recordingPreset != ma_opensl_recording_preset_default) { + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pAudioRecorderObj)->GetInterface((SLObjectItf)pDevice->opensl.pAudioRecorderObj, (SLInterfaceID)pDevice->pContext->opensl.SL_IID_ANDROIDCONFIGURATION, &pRecorderConfig); + if (resultSL == SL_RESULT_SUCCESS) { + SLint32 recordingPreset = ma_to_recording_preset__opensl(pConfig->opensl.recordingPreset); + resultSL = (*pRecorderConfig)->SetConfiguration(pRecorderConfig, SL_ANDROID_KEY_RECORDING_PRESET, &recordingPreset, sizeof(SLint32)); + if (resultSL != SL_RESULT_SUCCESS) { + /* Failed to set the configuration. Just keep going. */ + } + } + } + + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pAudioRecorderObj)->Realize((SLObjectItf)pDevice->opensl.pAudioRecorderObj, SL_BOOLEAN_FALSE); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to realize audio recorder."); + return ma_result_from_OpenSL(resultSL); + } + + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pAudioRecorderObj)->GetInterface((SLObjectItf)pDevice->opensl.pAudioRecorderObj, (SLInterfaceID)pDevice->pContext->opensl.SL_IID_RECORD, &pDevice->opensl.pAudioRecorder); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to retrieve SL_IID_RECORD interface."); + return ma_result_from_OpenSL(resultSL); + } + + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pAudioRecorderObj)->GetInterface((SLObjectItf)pDevice->opensl.pAudioRecorderObj, (SLInterfaceID)pDevice->pContext->opensl.SL_IID_ANDROIDSIMPLEBUFFERQUEUE, &pDevice->opensl.pBufferQueueCapture); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to retrieve SL_IID_ANDROIDSIMPLEBUFFERQUEUE interface."); + return ma_result_from_OpenSL(resultSL); + } + + resultSL = MA_OPENSL_BUFFERQUEUE(pDevice->opensl.pBufferQueueCapture)->RegisterCallback((SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueueCapture, ma_buffer_queue_callback_capture__opensl_android, pDevice); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to register buffer queue callback."); + return ma_result_from_OpenSL(resultSL); + } + + /* The internal format is determined by the "pcm" object. */ + ma_deconstruct_SLDataFormat_PCM__opensl(&pcm, &pDescriptorCapture->format, &pDescriptorCapture->channels, &pDescriptorCapture->sampleRate, pDescriptorCapture->channelMap, ma_countof(pDescriptorCapture->channelMap)); + + /* Buffer. */ + pDescriptorCapture->periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptorCapture, pDescriptorCapture->sampleRate, pConfig->performanceProfile); + pDevice->opensl.currentBufferIndexCapture = 0; + + bufferSizeInBytes = pDescriptorCapture->periodSizeInFrames * ma_get_bytes_per_frame(pDescriptorCapture->format, pDescriptorCapture->channels) * pDescriptorCapture->periodCount; + pDevice->opensl.pBufferCapture = (ma_uint8*)ma_calloc(bufferSizeInBytes, &pDevice->pContext->allocationCallbacks); + if (pDevice->opensl.pBufferCapture == NULL) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to allocate memory for data buffer."); + return MA_OUT_OF_MEMORY; + } + MA_ZERO_MEMORY(pDevice->opensl.pBufferCapture, bufferSizeInBytes); + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_SLDataFormat_PCM pcm; + SLDataSource source; + SLDataLocator_OutputMix outmixLocator; + SLDataSink sink; + SLAndroidConfigurationItf pPlayerConfig; + + ma_SLDataFormat_PCM_init__opensl(pDescriptorPlayback->format, pDescriptorPlayback->channels, pDescriptorPlayback->sampleRate, pDescriptorPlayback->channelMap, &pcm); + + resultSL = (*g_maEngineSL)->CreateOutputMix(g_maEngineSL, (SLObjectItf*)&pDevice->opensl.pOutputMixObj, 0, NULL, NULL); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to create output mix."); + return ma_result_from_OpenSL(resultSL); + } + + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pOutputMixObj)->Realize((SLObjectItf)pDevice->opensl.pOutputMixObj, SL_BOOLEAN_FALSE); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to realize output mix object."); + return ma_result_from_OpenSL(resultSL); + } + + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pOutputMixObj)->GetInterface((SLObjectItf)pDevice->opensl.pOutputMixObj, (SLInterfaceID)pDevice->pContext->opensl.SL_IID_OUTPUTMIX, &pDevice->opensl.pOutputMix); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to retrieve SL_IID_OUTPUTMIX interface."); + return ma_result_from_OpenSL(resultSL); + } + + /* Set the output device. */ + if (pDescriptorPlayback->pDeviceID != NULL) { + SLuint32 deviceID_OpenSL = pDescriptorPlayback->pDeviceID->opensl; + MA_OPENSL_OUTPUTMIX(pDevice->opensl.pOutputMix)->ReRoute((SLOutputMixItf)pDevice->opensl.pOutputMix, 1, &deviceID_OpenSL); + } + + queue.numBuffers = pDescriptorPlayback->periodCount; + + source.pLocator = &queue; + source.pFormat = (SLDataFormat_PCM*)&pcm; + + outmixLocator.locatorType = SL_DATALOCATOR_OUTPUTMIX; + outmixLocator.outputMix = (SLObjectItf)pDevice->opensl.pOutputMixObj; + + sink.pLocator = &outmixLocator; + sink.pFormat = NULL; + + resultSL = (*g_maEngineSL)->CreateAudioPlayer(g_maEngineSL, (SLObjectItf*)&pDevice->opensl.pAudioPlayerObj, &source, &sink, ma_countof(itfIDs), itfIDs, itfIDsRequired); + if (resultSL == SL_RESULT_CONTENT_UNSUPPORTED || resultSL == SL_RESULT_PARAMETER_INVALID) { + /* Unsupported format. Fall back to something safer and try again. If this fails, just abort. */ + pcm.formatType = SL_DATAFORMAT_PCM; + pcm.numChannels = 2; + ((SLDataFormat_PCM*)&pcm)->samplesPerSec = SL_SAMPLINGRATE_16; + pcm.bitsPerSample = 16; + pcm.containerSize = pcm.bitsPerSample; /* Always tightly packed for now. */ + pcm.channelMask = SL_SPEAKER_FRONT_LEFT | SL_SPEAKER_FRONT_RIGHT; + resultSL = (*g_maEngineSL)->CreateAudioPlayer(g_maEngineSL, (SLObjectItf*)&pDevice->opensl.pAudioPlayerObj, &source, &sink, ma_countof(itfIDs), itfIDs, itfIDsRequired); + } + + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to create audio player."); + return ma_result_from_OpenSL(resultSL); + } + + + /* Set the stream type before realizing the player. */ + if (pConfig->opensl.streamType != ma_opensl_stream_type_default) { + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pAudioPlayerObj)->GetInterface((SLObjectItf)pDevice->opensl.pAudioPlayerObj, (SLInterfaceID)pDevice->pContext->opensl.SL_IID_ANDROIDCONFIGURATION, &pPlayerConfig); + if (resultSL == SL_RESULT_SUCCESS) { + SLint32 streamType = ma_to_stream_type__opensl(pConfig->opensl.streamType); + resultSL = (*pPlayerConfig)->SetConfiguration(pPlayerConfig, SL_ANDROID_KEY_STREAM_TYPE, &streamType, sizeof(SLint32)); + if (resultSL != SL_RESULT_SUCCESS) { + /* Failed to set the configuration. Just keep going. */ + } + } + } + + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pAudioPlayerObj)->Realize((SLObjectItf)pDevice->opensl.pAudioPlayerObj, SL_BOOLEAN_FALSE); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to realize audio player."); + return ma_result_from_OpenSL(resultSL); + } + + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pAudioPlayerObj)->GetInterface((SLObjectItf)pDevice->opensl.pAudioPlayerObj, (SLInterfaceID)pDevice->pContext->opensl.SL_IID_PLAY, &pDevice->opensl.pAudioPlayer); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to retrieve SL_IID_PLAY interface."); + return ma_result_from_OpenSL(resultSL); + } + + resultSL = MA_OPENSL_OBJ(pDevice->opensl.pAudioPlayerObj)->GetInterface((SLObjectItf)pDevice->opensl.pAudioPlayerObj, (SLInterfaceID)pDevice->pContext->opensl.SL_IID_ANDROIDSIMPLEBUFFERQUEUE, &pDevice->opensl.pBufferQueuePlayback); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to retrieve SL_IID_ANDROIDSIMPLEBUFFERQUEUE interface."); + return ma_result_from_OpenSL(resultSL); + } + + resultSL = MA_OPENSL_BUFFERQUEUE(pDevice->opensl.pBufferQueuePlayback)->RegisterCallback((SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueuePlayback, ma_buffer_queue_callback_playback__opensl_android, pDevice); + if (resultSL != SL_RESULT_SUCCESS) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to register buffer queue callback."); + return ma_result_from_OpenSL(resultSL); + } + + /* The internal format is determined by the "pcm" object. */ + ma_deconstruct_SLDataFormat_PCM__opensl(&pcm, &pDescriptorPlayback->format, &pDescriptorPlayback->channels, &pDescriptorPlayback->sampleRate, pDescriptorPlayback->channelMap, ma_countof(pDescriptorPlayback->channelMap)); + + /* Buffer. */ + pDescriptorPlayback->periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_descriptor(pDescriptorPlayback, pDescriptorPlayback->sampleRate, pConfig->performanceProfile); + pDevice->opensl.currentBufferIndexPlayback = 0; + + bufferSizeInBytes = pDescriptorPlayback->periodSizeInFrames * ma_get_bytes_per_frame(pDescriptorPlayback->format, pDescriptorPlayback->channels) * pDescriptorPlayback->periodCount; + pDevice->opensl.pBufferPlayback = (ma_uint8*)ma_calloc(bufferSizeInBytes, &pDevice->pContext->allocationCallbacks); + if (pDevice->opensl.pBufferPlayback == NULL) { + ma_device_uninit__opensl(pDevice); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to allocate memory for data buffer."); + return MA_OUT_OF_MEMORY; + } + MA_ZERO_MEMORY(pDevice->opensl.pBufferPlayback, bufferSizeInBytes); + } + + return MA_SUCCESS; +#else + return MA_NO_BACKEND; /* Non-Android implementations are not supported. */ +#endif +} + +static ma_result ma_device_start__opensl(ma_device* pDevice) +{ + SLresult resultSL; + size_t periodSizeInBytes; + ma_uint32 iPeriod; + + MA_ASSERT(pDevice != NULL); + + MA_ASSERT(g_maOpenSLInitCounter > 0); /* <-- If you trigger this it means you've either not initialized the context, or you've uninitialized it and then attempted to start the device. */ + if (g_maOpenSLInitCounter == 0) { + return MA_INVALID_OPERATION; + } + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + resultSL = MA_OPENSL_RECORD(pDevice->opensl.pAudioRecorder)->SetRecordState((SLRecordItf)pDevice->opensl.pAudioRecorder, SL_RECORDSTATE_RECORDING); + if (resultSL != SL_RESULT_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to start internal capture device."); + return ma_result_from_OpenSL(resultSL); + } + + periodSizeInBytes = pDevice->capture.internalPeriodSizeInFrames * ma_get_bytes_per_frame(pDevice->capture.internalFormat, pDevice->capture.internalChannels); + for (iPeriod = 0; iPeriod < pDevice->capture.internalPeriods; ++iPeriod) { + resultSL = MA_OPENSL_BUFFERQUEUE(pDevice->opensl.pBufferQueueCapture)->Enqueue((SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueueCapture, pDevice->opensl.pBufferCapture + (periodSizeInBytes * iPeriod), periodSizeInBytes); + if (resultSL != SL_RESULT_SUCCESS) { + MA_OPENSL_RECORD(pDevice->opensl.pAudioRecorder)->SetRecordState((SLRecordItf)pDevice->opensl.pAudioRecorder, SL_RECORDSTATE_STOPPED); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to enqueue buffer for capture device."); + return ma_result_from_OpenSL(resultSL); + } + } + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + resultSL = MA_OPENSL_PLAY(pDevice->opensl.pAudioPlayer)->SetPlayState((SLPlayItf)pDevice->opensl.pAudioPlayer, SL_PLAYSTATE_PLAYING); + if (resultSL != SL_RESULT_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to start internal playback device."); + return ma_result_from_OpenSL(resultSL); + } + + /* In playback mode (no duplex) we need to load some initial buffers. In duplex mode we need to enqueue silent buffers. */ + if (pDevice->type == ma_device_type_duplex) { + MA_ZERO_MEMORY(pDevice->opensl.pBufferPlayback, pDevice->playback.internalPeriodSizeInFrames * pDevice->playback.internalPeriods * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels)); + } else { + ma_device__read_frames_from_client(pDevice, pDevice->playback.internalPeriodSizeInFrames * pDevice->playback.internalPeriods, pDevice->opensl.pBufferPlayback); + } + + periodSizeInBytes = pDevice->playback.internalPeriodSizeInFrames * ma_get_bytes_per_frame(pDevice->playback.internalFormat, pDevice->playback.internalChannels); + for (iPeriod = 0; iPeriod < pDevice->playback.internalPeriods; ++iPeriod) { + resultSL = MA_OPENSL_BUFFERQUEUE(pDevice->opensl.pBufferQueuePlayback)->Enqueue((SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueuePlayback, pDevice->opensl.pBufferPlayback + (periodSizeInBytes * iPeriod), periodSizeInBytes); + if (resultSL != SL_RESULT_SUCCESS) { + MA_OPENSL_PLAY(pDevice->opensl.pAudioPlayer)->SetPlayState((SLPlayItf)pDevice->opensl.pAudioPlayer, SL_PLAYSTATE_STOPPED); + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to enqueue buffer for playback device."); + return ma_result_from_OpenSL(resultSL); + } + } + } + + return MA_SUCCESS; +} + +static ma_result ma_device_drain__opensl(ma_device* pDevice, ma_device_type deviceType) +{ + SLAndroidSimpleBufferQueueItf pBufferQueue; + + MA_ASSERT(deviceType == ma_device_type_capture || deviceType == ma_device_type_playback); + + if (pDevice->type == ma_device_type_capture) { + pBufferQueue = (SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueueCapture; + pDevice->opensl.isDrainingCapture = MA_TRUE; + } else { + pBufferQueue = (SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueuePlayback; + pDevice->opensl.isDrainingPlayback = MA_TRUE; + } + + for (;;) { + SLAndroidSimpleBufferQueueState state; + + MA_OPENSL_BUFFERQUEUE(pBufferQueue)->GetState(pBufferQueue, &state); + if (state.count == 0) { + break; + } + + ma_sleep(10); + } + + if (pDevice->type == ma_device_type_capture) { + pDevice->opensl.isDrainingCapture = MA_FALSE; + } else { + pDevice->opensl.isDrainingPlayback = MA_FALSE; + } + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__opensl(ma_device* pDevice) +{ + SLresult resultSL; + + MA_ASSERT(pDevice != NULL); + + MA_ASSERT(g_maOpenSLInitCounter > 0); /* <-- If you trigger this it means you've either not initialized the context, or you've uninitialized it before stopping/uninitializing the device. */ + if (g_maOpenSLInitCounter == 0) { + return MA_INVALID_OPERATION; + } + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex) { + ma_device_drain__opensl(pDevice, ma_device_type_capture); + + resultSL = MA_OPENSL_RECORD(pDevice->opensl.pAudioRecorder)->SetRecordState((SLRecordItf)pDevice->opensl.pAudioRecorder, SL_RECORDSTATE_STOPPED); + if (resultSL != SL_RESULT_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to stop internal capture device."); + return ma_result_from_OpenSL(resultSL); + } + + MA_OPENSL_BUFFERQUEUE(pDevice->opensl.pBufferQueueCapture)->Clear((SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueueCapture); + } + + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_device_drain__opensl(pDevice, ma_device_type_playback); + + resultSL = MA_OPENSL_PLAY(pDevice->opensl.pAudioPlayer)->SetPlayState((SLPlayItf)pDevice->opensl.pAudioPlayer, SL_PLAYSTATE_STOPPED); + if (resultSL != SL_RESULT_SUCCESS) { + ma_log_post(ma_device_get_log(pDevice), MA_LOG_LEVEL_ERROR, "[OpenSL] Failed to stop internal playback device."); + return ma_result_from_OpenSL(resultSL); + } + + MA_OPENSL_BUFFERQUEUE(pDevice->opensl.pBufferQueuePlayback)->Clear((SLAndroidSimpleBufferQueueItf)pDevice->opensl.pBufferQueuePlayback); + } + + /* Make sure the client is aware that the device has stopped. There may be an OpenSL|ES callback for this, but I haven't found it. */ + ma_device__on_notification_stopped(pDevice); + + return MA_SUCCESS; +} + + +static ma_result ma_context_uninit__opensl(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_opensl); + (void)pContext; + + /* Uninit global data. */ + ma_spinlock_lock(&g_maOpenSLSpinlock); + { + MA_ASSERT(g_maOpenSLInitCounter > 0); /* If you've triggered this, it means you have ma_context_init/uninit mismatch. Each successful call to ma_context_init() must be matched up with a call to ma_context_uninit(). */ + + g_maOpenSLInitCounter -= 1; + if (g_maOpenSLInitCounter == 0) { + (*g_maEngineObjectSL)->Destroy(g_maEngineObjectSL); + } + } + ma_spinlock_unlock(&g_maOpenSLSpinlock); + + return MA_SUCCESS; +} + +static ma_result ma_dlsym_SLInterfaceID__opensl(ma_context* pContext, const char* pName, ma_handle* pHandle) +{ + /* We need to return an error if the symbol cannot be found. This is important because there have been reports that some symbols do not exist. */ + ma_handle* p = (ma_handle*)ma_dlsym(ma_context_get_log(pContext), pContext->opensl.libOpenSLES, pName); + if (p == NULL) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_INFO, "[OpenSL] Cannot find symbol %s", pName); + return MA_NO_BACKEND; + } + + *pHandle = *p; + return MA_SUCCESS; +} + +static ma_result ma_context_init_engine_nolock__opensl(ma_context* pContext) +{ + g_maOpenSLInitCounter += 1; + if (g_maOpenSLInitCounter == 1) { + SLresult resultSL; + + resultSL = ((ma_slCreateEngine_proc)pContext->opensl.slCreateEngine)(&g_maEngineObjectSL, 0, NULL, 0, NULL, NULL); + if (resultSL != SL_RESULT_SUCCESS) { + g_maOpenSLInitCounter -= 1; + return ma_result_from_OpenSL(resultSL); + } + + (*g_maEngineObjectSL)->Realize(g_maEngineObjectSL, SL_BOOLEAN_FALSE); + + resultSL = (*g_maEngineObjectSL)->GetInterface(g_maEngineObjectSL, (SLInterfaceID)pContext->opensl.SL_IID_ENGINE, &g_maEngineSL); + if (resultSL != SL_RESULT_SUCCESS) { + (*g_maEngineObjectSL)->Destroy(g_maEngineObjectSL); + g_maOpenSLInitCounter -= 1; + return ma_result_from_OpenSL(resultSL); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_context_init__opensl(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + ma_result result; + +#if !defined(MA_NO_RUNTIME_LINKING) + size_t i; + const char* libOpenSLESNames[] = { + "libOpenSLES.so" + }; +#endif + + MA_ASSERT(pContext != NULL); + + (void)pConfig; + +#if !defined(MA_NO_RUNTIME_LINKING) + /* + Dynamically link against libOpenSLES.so. I have now had multiple reports that SL_IID_ANDROIDSIMPLEBUFFERQUEUE cannot be found. One + report was happening at compile time and another at runtime. To try working around this, I'm going to link to libOpenSLES at runtime + and extract the symbols rather than reference them directly. This should, hopefully, fix these issues as the compiler won't see any + references to the symbols and will hopefully skip the checks. + */ + for (i = 0; i < ma_countof(libOpenSLESNames); i += 1) { + pContext->opensl.libOpenSLES = ma_dlopen(ma_context_get_log(pContext), libOpenSLESNames[i]); + if (pContext->opensl.libOpenSLES != NULL) { + break; + } + } + + if (pContext->opensl.libOpenSLES == NULL) { + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_INFO, "[OpenSL] Could not find libOpenSLES.so"); + return MA_NO_BACKEND; + } + + result = ma_dlsym_SLInterfaceID__opensl(pContext, "SL_IID_ENGINE", &pContext->opensl.SL_IID_ENGINE); + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + return result; + } + + result = ma_dlsym_SLInterfaceID__opensl(pContext, "SL_IID_AUDIOIODEVICECAPABILITIES", &pContext->opensl.SL_IID_AUDIOIODEVICECAPABILITIES); + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + return result; + } + + result = ma_dlsym_SLInterfaceID__opensl(pContext, "SL_IID_ANDROIDSIMPLEBUFFERQUEUE", &pContext->opensl.SL_IID_ANDROIDSIMPLEBUFFERQUEUE); + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + return result; + } + + result = ma_dlsym_SLInterfaceID__opensl(pContext, "SL_IID_RECORD", &pContext->opensl.SL_IID_RECORD); + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + return result; + } + + result = ma_dlsym_SLInterfaceID__opensl(pContext, "SL_IID_PLAY", &pContext->opensl.SL_IID_PLAY); + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + return result; + } + + result = ma_dlsym_SLInterfaceID__opensl(pContext, "SL_IID_OUTPUTMIX", &pContext->opensl.SL_IID_OUTPUTMIX); + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + return result; + } + + result = ma_dlsym_SLInterfaceID__opensl(pContext, "SL_IID_ANDROIDCONFIGURATION", &pContext->opensl.SL_IID_ANDROIDCONFIGURATION); + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + return result; + } + + pContext->opensl.slCreateEngine = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->opensl.libOpenSLES, "slCreateEngine"); + if (pContext->opensl.slCreateEngine == NULL) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_INFO, "[OpenSL] Cannot find symbol slCreateEngine."); + return MA_NO_BACKEND; + } +#else + pContext->opensl.SL_IID_ENGINE = (ma_handle)SL_IID_ENGINE; + pContext->opensl.SL_IID_AUDIOIODEVICECAPABILITIES = (ma_handle)SL_IID_AUDIOIODEVICECAPABILITIES; + pContext->opensl.SL_IID_ANDROIDSIMPLEBUFFERQUEUE = (ma_handle)SL_IID_ANDROIDSIMPLEBUFFERQUEUE; + pContext->opensl.SL_IID_RECORD = (ma_handle)SL_IID_RECORD; + pContext->opensl.SL_IID_PLAY = (ma_handle)SL_IID_PLAY; + pContext->opensl.SL_IID_OUTPUTMIX = (ma_handle)SL_IID_OUTPUTMIX; + pContext->opensl.SL_IID_ANDROIDCONFIGURATION = (ma_handle)SL_IID_ANDROIDCONFIGURATION; + pContext->opensl.slCreateEngine = (ma_proc)slCreateEngine; +#endif + + + /* Initialize global data first if applicable. */ + ma_spinlock_lock(&g_maOpenSLSpinlock); + { + result = ma_context_init_engine_nolock__opensl(pContext); + } + ma_spinlock_unlock(&g_maOpenSLSpinlock); + + if (result != MA_SUCCESS) { + ma_dlclose(ma_context_get_log(pContext), pContext->opensl.libOpenSLES); + ma_log_post(ma_context_get_log(pContext), MA_LOG_LEVEL_INFO, "[OpenSL] Failed to initialize OpenSL engine."); + return result; + } + + pCallbacks->onContextInit = ma_context_init__opensl; + pCallbacks->onContextUninit = ma_context_uninit__opensl; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__opensl; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__opensl; + pCallbacks->onDeviceInit = ma_device_init__opensl; + pCallbacks->onDeviceUninit = ma_device_uninit__opensl; + pCallbacks->onDeviceStart = ma_device_start__opensl; + pCallbacks->onDeviceStop = ma_device_stop__opensl; + pCallbacks->onDeviceRead = NULL; /* Not needed because OpenSL|ES is asynchronous. */ + pCallbacks->onDeviceWrite = NULL; /* Not needed because OpenSL|ES is asynchronous. */ + pCallbacks->onDeviceDataLoop = NULL; /* Not needed because OpenSL|ES is asynchronous. */ + + return MA_SUCCESS; +} +#endif /* OpenSL|ES */ + + +/****************************************************************************** + +Web Audio Backend + +******************************************************************************/ +#ifdef MA_HAS_WEBAUDIO +#include + +#if (__EMSCRIPTEN_major__ > 3) || (__EMSCRIPTEN_major__ == 3 && (__EMSCRIPTEN_minor__ > 1 || (__EMSCRIPTEN_minor__ == 1 && __EMSCRIPTEN_tiny__ >= 32))) + #include + #define MA_SUPPORT_AUDIO_WORKLETS + + #if (__EMSCRIPTEN_major__ > 3) || (__EMSCRIPTEN_major__ == 3 && (__EMSCRIPTEN_minor__ > 1 || (__EMSCRIPTEN_minor__ == 1 && __EMSCRIPTEN_tiny__ >= 70))) + #define MA_SUPPORT_AUDIO_WORKLETS_VARIABLE_BUFFER_SIZE + #endif +#endif + +/* +TODO: Version 0.12: Swap this logic around so that AudioWorklets are used by default. Add MA_NO_AUDIO_WORKLETS. +*/ +#if defined(MA_ENABLE_AUDIO_WORKLETS) && defined(MA_SUPPORT_AUDIO_WORKLETS) + #define MA_USE_AUDIO_WORKLETS +#endif + +/* The thread stack size must be a multiple of 16. */ +#ifndef MA_AUDIO_WORKLETS_THREAD_STACK_SIZE +#define MA_AUDIO_WORKLETS_THREAD_STACK_SIZE 131072 +#endif + +#if defined(MA_USE_AUDIO_WORKLETS) +#define MA_WEBAUDIO_LATENCY_HINT_BALANCED "balanced" +#define MA_WEBAUDIO_LATENCY_HINT_INTERACTIVE "interactive" +#define MA_WEBAUDIO_LATENCY_HINT_PLAYBACK "playback" +#endif + +static ma_bool32 ma_is_capture_supported__webaudio() +{ + return EM_ASM_INT({ + return (navigator.mediaDevices !== undefined && navigator.mediaDevices.getUserMedia !== undefined); + }, 0) != 0; /* Must pass in a dummy argument for C99 compatibility. */ +} + +#ifdef __cplusplus +extern "C" { +#endif +void* EMSCRIPTEN_KEEPALIVE ma_malloc_emscripten(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_malloc(sz, pAllocationCallbacks); +} + +void EMSCRIPTEN_KEEPALIVE ma_free_emscripten(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_free(p, pAllocationCallbacks); +} + +void EMSCRIPTEN_KEEPALIVE ma_device_process_pcm_frames_capture__webaudio(ma_device* pDevice, int frameCount, float* pFrames) +{ + ma_device_handle_backend_data_callback(pDevice, NULL, pFrames, (ma_uint32)frameCount); +} + +void EMSCRIPTEN_KEEPALIVE ma_device_process_pcm_frames_playback__webaudio(ma_device* pDevice, int frameCount, float* pFrames) +{ + ma_device_handle_backend_data_callback(pDevice, pFrames, NULL, (ma_uint32)frameCount); +} +#ifdef __cplusplus +} +#endif + +static ma_result ma_context_enumerate_devices__webaudio(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_bool32 cbResult = MA_TRUE; + + MA_ASSERT(pContext != NULL); + MA_ASSERT(callback != NULL); + + /* Only supporting default devices for now. */ + + /* Playback. */ + if (cbResult) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + deviceInfo.isDefault = MA_TRUE; /* Only supporting default devices. */ + cbResult = callback(pContext, ma_device_type_playback, &deviceInfo, pUserData); + } + + /* Capture. */ + if (cbResult) { + if (ma_is_capture_supported__webaudio()) { + ma_device_info deviceInfo; + MA_ZERO_OBJECT(&deviceInfo); + ma_strncpy_s(deviceInfo.name, sizeof(deviceInfo.name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + deviceInfo.isDefault = MA_TRUE; /* Only supporting default devices. */ + cbResult = callback(pContext, ma_device_type_capture, &deviceInfo, pUserData); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_context_get_device_info__webaudio(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + MA_ASSERT(pContext != NULL); + + if (deviceType == ma_device_type_capture && !ma_is_capture_supported__webaudio()) { + return MA_NO_DEVICE; + } + + MA_ZERO_MEMORY(pDeviceInfo->id.webaudio, sizeof(pDeviceInfo->id.webaudio)); + + /* Only supporting default devices for now. */ + (void)pDeviceID; + if (deviceType == ma_device_type_playback) { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDeviceInfo->name, sizeof(pDeviceInfo->name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } + + /* Only supporting default devices. */ + pDeviceInfo->isDefault = MA_TRUE; + + /* Web Audio can support any number of channels and sample rates. It only supports f32 formats, however. */ + pDeviceInfo->nativeDataFormats[0].flags = 0; + pDeviceInfo->nativeDataFormats[0].format = ma_format_unknown; + pDeviceInfo->nativeDataFormats[0].channels = 0; /* All channels are supported. */ + pDeviceInfo->nativeDataFormats[0].sampleRate = EM_ASM_INT({ + try { + var temp = new (window.AudioContext || window.webkitAudioContext)(); + var sampleRate = temp.sampleRate; + temp.close(); + return sampleRate; + } catch(e) { + return 0; + } + }, 0); /* Must pass in a dummy argument for C99 compatibility. */ + + if (pDeviceInfo->nativeDataFormats[0].sampleRate == 0) { + return MA_NO_DEVICE; + } + + pDeviceInfo->nativeDataFormatCount = 1; + + return MA_SUCCESS; +} + +static ma_result ma_device_uninit__webaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + #if defined(MA_USE_AUDIO_WORKLETS) + { + EM_ASM({ + var device = window.miniaudio.get_device_by_index($0); + + if (device.streamNode !== undefined) { + device.streamNode.disconnect(); + device.streamNode = undefined; + } + + device.pDevice = undefined; + }, pDevice->webaudio.deviceIndex); + + emscripten_destroy_web_audio_node(pDevice->webaudio.audioWorklet); + emscripten_destroy_audio_context(pDevice->webaudio.audioContext); + ma_free(pDevice->webaudio.pStackBuffer, &pDevice->pContext->allocationCallbacks); + } + #else + { + EM_ASM({ + var device = window.miniaudio.get_device_by_index($0); + + /* Make sure all nodes are disconnected and marked for collection. */ + if (device.scriptNode !== undefined) { + device.scriptNode.onaudioprocess = function(e) {}; /* We want to reset the callback to ensure it doesn't get called after AudioContext.close() has returned. Shouldn't happen since we're disconnecting, but just to be safe... */ + device.scriptNode.disconnect(); + device.scriptNode = undefined; + } + + if (device.streamNode !== undefined) { + device.streamNode.disconnect(); + device.streamNode = undefined; + } + + /* + Stop the device. I think there is a chance the callback could get fired after calling this, hence why we want + to clear the callback before closing. + */ + device.webaudio.close(); + device.webaudio = undefined; + device.pDevice = undefined; + }, pDevice->webaudio.deviceIndex); + } + #endif + + /* Clean up the device on the JS side. */ + EM_ASM({ + window.miniaudio.untrack_device_by_index($0); + }, pDevice->webaudio.deviceIndex); + + ma_free(pDevice->webaudio.pIntermediaryBuffer, &pDevice->pContext->allocationCallbacks); + + return MA_SUCCESS; +} + +#if !defined(MA_USE_AUDIO_WORKLETS) +static ma_uint32 ma_calculate_period_size_in_frames_from_descriptor__webaudio(const ma_device_descriptor* pDescriptor, ma_uint32 nativeSampleRate, ma_performance_profile performanceProfile) +{ + /* + There have been reports of the default buffer size being too small on some browsers. If we're using + the default buffer size, we'll make sure the period size is bigger than our standard defaults. + */ + ma_uint32 periodSizeInFrames; + + if (nativeSampleRate == 0) { + nativeSampleRate = MA_DEFAULT_SAMPLE_RATE; + } + + if (pDescriptor->periodSizeInFrames == 0) { + if (pDescriptor->periodSizeInMilliseconds == 0) { + if (performanceProfile == ma_performance_profile_low_latency) { + periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(33, nativeSampleRate); /* 1 frame @ 30 FPS */ + } else { + periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(333, nativeSampleRate); + } + } else { + periodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(pDescriptor->periodSizeInMilliseconds, nativeSampleRate); + } + } else { + periodSizeInFrames = pDescriptor->periodSizeInFrames; + } + + /* The size of the buffer must be a power of 2 and between 256 and 16384. */ + if (periodSizeInFrames < 256) { + periodSizeInFrames = 256; + } else if (periodSizeInFrames > 16384) { + periodSizeInFrames = 16384; + } else { + periodSizeInFrames = ma_next_power_of_2(periodSizeInFrames); + } + + return periodSizeInFrames; +} +#endif + + +#if defined(MA_USE_AUDIO_WORKLETS) +typedef struct +{ + ma_device* pDevice; + const ma_device_config* pConfig; + ma_device_descriptor* pDescriptorPlayback; + ma_device_descriptor* pDescriptorCapture; +} ma_audio_worklet_thread_initialized_data; + +static EM_BOOL ma_audio_worklet_process_callback__webaudio(int inputCount, const AudioSampleFrame* pInputs, int outputCount, AudioSampleFrame* pOutputs, int paramCount, const AudioParamFrame* pParams, void* pUserData) +{ + ma_device* pDevice = (ma_device*)pUserData; + ma_uint32 frameCount; + + (void)paramCount; + (void)pParams; + + /* + The Emscripten documentation says that it'll always be 128 frames being passed in. Hard coding it like that feels + like a very bad idea to me. Even if it's hard coded in the backend, the API and documentation should always refer + to variables instead of a hard coded number. In any case, will follow along for the time being. + + Unfortunately the audio data is not interleaved so we'll need to convert it before we give the data to miniaudio + for further processing. + */ + if (pDevice->type == ma_device_type_playback) { + frameCount = pDevice->playback.internalPeriodSizeInFrames; + } else { + frameCount = pDevice->capture.internalPeriodSizeInFrames; + } + + if (ma_device_get_state(pDevice) != ma_device_state_started) { + /* Fill the output buffer with zero to avoid a noise sound */ + for (int i = 0; i < outputCount; i += 1) { + MA_ZERO_MEMORY(pOutputs[i].data, pOutputs[i].numberOfChannels * frameCount * sizeof(float)); + } + + return EM_TRUE; + } + + if (inputCount > 0) { + /* Input data needs to be interleaved before we hand it to the client. */ + for (ma_uint32 iChannel = 0; iChannel < pDevice->capture.internalChannels; iChannel += 1) { + for (ma_uint32 iFrame = 0; iFrame < frameCount; iFrame += 1) { + pDevice->webaudio.pIntermediaryBuffer[iFrame*pDevice->capture.internalChannels + iChannel] = pInputs[0].data[frameCount*iChannel + iFrame]; + } + } + + ma_device_process_pcm_frames_capture__webaudio(pDevice, frameCount, pDevice->webaudio.pIntermediaryBuffer); + } + + if (outputCount > 0) { + /* If it's a capture-only device, we'll need to output silence. */ + if (pDevice->type == ma_device_type_capture) { + MA_ZERO_MEMORY(pOutputs[0].data, frameCount * pDevice->playback.internalChannels * sizeof(float)); + } else { + ma_device_process_pcm_frames_playback__webaudio(pDevice, frameCount, pDevice->webaudio.pIntermediaryBuffer); + + /* We've read the data from the client. Now we need to deinterleave the buffer and output to the output buffer. */ + for (ma_uint32 iChannel = 0; iChannel < pDevice->playback.internalChannels; iChannel += 1) { + for (ma_uint32 iFrame = 0; iFrame < frameCount; iFrame += 1) { + pOutputs[0].data[frameCount*iChannel + iFrame] = pDevice->webaudio.pIntermediaryBuffer[iFrame*pDevice->playback.internalChannels + iChannel]; + } + } + } + } + + return EM_TRUE; +} + + +static void ma_audio_worklet_processor_created__webaudio(EMSCRIPTEN_WEBAUDIO_T audioContext, EM_BOOL success, void* pUserData) +{ + ma_audio_worklet_thread_initialized_data* pParameters = (ma_audio_worklet_thread_initialized_data*)pUserData; + EmscriptenAudioWorkletNodeCreateOptions audioWorkletOptions; + int channels = 0; + size_t intermediaryBufferSizeInFrames; + int sampleRate; + + if (success == EM_FALSE) { + pParameters->pDevice->webaudio.initResult = MA_ERROR; + ma_free(pParameters, &pParameters->pDevice->pContext->allocationCallbacks); + return; + } + + /* The next step is to initialize the audio worklet node. */ + MA_ZERO_OBJECT(&audioWorkletOptions); + + /* + The way channel counts work with Web Audio is confusing. As far as I can tell, there's no way to know the channel + count from MediaStreamAudioSourceNode (what we use for capture)? The only way to have control is to configure an + output channel count on the capture side. This is slightly confusing for capture mode because intuitively you + wouldn't actually connect an output to an input-only node, but this is what we'll have to do in order to have + proper control over the channel count. In the capture case, we'll have to output silence to its output node. + */ + if (pParameters->pConfig->deviceType == ma_device_type_capture) { + channels = (int)((pParameters->pDescriptorCapture->channels > 0) ? pParameters->pDescriptorCapture->channels : MA_DEFAULT_CHANNELS); + audioWorkletOptions.numberOfInputs = 1; + } else { + channels = (int)((pParameters->pDescriptorPlayback->channels > 0) ? pParameters->pDescriptorPlayback->channels : MA_DEFAULT_CHANNELS); + + if (pParameters->pConfig->deviceType == ma_device_type_duplex) { + audioWorkletOptions.numberOfInputs = 1; + } else { + audioWorkletOptions.numberOfInputs = 0; + } + } + + audioWorkletOptions.numberOfOutputs = 1; + audioWorkletOptions.outputChannelCounts = &channels; + + + /* + Now that we know the channel count to use we can allocate the intermediary buffer. The + intermediary buffer is used for interleaving and deinterleaving. + */ + #if defined(MA_SUPPORT_AUDIO_WORKLETS_VARIABLE_BUFFER_SIZE) + { + intermediaryBufferSizeInFrames = (size_t)emscripten_audio_context_quantum_size(audioContext); + } + #else + { + intermediaryBufferSizeInFrames = 128; + } + #endif + + pParameters->pDevice->webaudio.pIntermediaryBuffer = (float*)ma_malloc(intermediaryBufferSizeInFrames * (ma_uint32)channels * sizeof(float), &pParameters->pDevice->pContext->allocationCallbacks); + if (pParameters->pDevice->webaudio.pIntermediaryBuffer == NULL) { + pParameters->pDevice->webaudio.initResult = MA_OUT_OF_MEMORY; + ma_free(pParameters, &pParameters->pDevice->pContext->allocationCallbacks); + return; + } + + pParameters->pDevice->webaudio.audioWorklet = emscripten_create_wasm_audio_worklet_node(audioContext, "miniaudio", &audioWorkletOptions, &ma_audio_worklet_process_callback__webaudio, pParameters->pDevice); + + /* With the audio worklet initialized we can now attach it to the graph. */ + if (pParameters->pConfig->deviceType == ma_device_type_capture || pParameters->pConfig->deviceType == ma_device_type_duplex) { + ma_result attachmentResult = (ma_result)EM_ASM_INT({ + var getUserMediaResult = 0; + var audioWorklet = emscriptenGetAudioObject($0); + var audioContext = emscriptenGetAudioObject($1); + + navigator.mediaDevices.getUserMedia({audio:true, video:false}) + .then(function(stream) { + audioContext.streamNode = audioContext.createMediaStreamSource(stream); + audioContext.streamNode.connect(audioWorklet); + audioWorklet.connect(audioContext.destination); + getUserMediaResult = 0; /* 0 = MA_SUCCESS */ + }) + .catch(function(error) { + console.log("navigator.mediaDevices.getUserMedia Failed: " + error); + getUserMediaResult = -1; /* -1 = MA_ERROR */ + }); + + return getUserMediaResult; + }, pParameters->pDevice->webaudio.audioWorklet, audioContext); + + if (attachmentResult != MA_SUCCESS) { + ma_log_postf(ma_device_get_log(pParameters->pDevice), MA_LOG_LEVEL_ERROR, "Web Audio: Failed to connect capture node."); + emscripten_destroy_web_audio_node(pParameters->pDevice->webaudio.audioWorklet); + pParameters->pDevice->webaudio.initResult = attachmentResult; + ma_free(pParameters, &pParameters->pDevice->pContext->allocationCallbacks); + return; + } + } + + /* If it's playback only we can now attach the worklet node to the graph. This has already been done for the duplex case. */ + if (pParameters->pConfig->deviceType == ma_device_type_playback) { + ma_result attachmentResult = (ma_result)EM_ASM_INT({ + var audioWorklet = emscriptenGetAudioObject($0); + var audioContext = emscriptenGetAudioObject($1); + audioWorklet.connect(audioContext.destination); + return 0; /* 0 = MA_SUCCESS */ + }, pParameters->pDevice->webaudio.audioWorklet, audioContext); + + if (attachmentResult != MA_SUCCESS) { + ma_log_postf(ma_device_get_log(pParameters->pDevice), MA_LOG_LEVEL_ERROR, "Web Audio: Failed to connect playback node."); + pParameters->pDevice->webaudio.initResult = attachmentResult; + ma_free(pParameters, &pParameters->pDevice->pContext->allocationCallbacks); + return; + } + } + + /* We need to update the descriptors so that they reflect the internal data format. Both capture and playback should be the same. */ + sampleRate = EM_ASM_INT({ return emscriptenGetAudioObject($0).sampleRate; }, audioContext); + + if (pParameters->pDescriptorCapture != NULL) { + pParameters->pDescriptorCapture->format = ma_format_f32; + pParameters->pDescriptorCapture->channels = (ma_uint32)channels; + pParameters->pDescriptorCapture->sampleRate = (ma_uint32)sampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_webaudio, pParameters->pDescriptorCapture->channelMap, ma_countof(pParameters->pDescriptorCapture->channelMap), pParameters->pDescriptorCapture->channels); + pParameters->pDescriptorCapture->periodSizeInFrames = intermediaryBufferSizeInFrames; + pParameters->pDescriptorCapture->periodCount = 1; + } + + if (pParameters->pDescriptorPlayback != NULL) { + pParameters->pDescriptorPlayback->format = ma_format_f32; + pParameters->pDescriptorPlayback->channels = (ma_uint32)channels; + pParameters->pDescriptorPlayback->sampleRate = (ma_uint32)sampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_webaudio, pParameters->pDescriptorPlayback->channelMap, ma_countof(pParameters->pDescriptorPlayback->channelMap), pParameters->pDescriptorPlayback->channels); + pParameters->pDescriptorPlayback->periodSizeInFrames = intermediaryBufferSizeInFrames; + pParameters->pDescriptorPlayback->periodCount = 1; + } + + /* At this point we're done and we can return. */ + ma_log_postf(ma_device_get_log(pParameters->pDevice), MA_LOG_LEVEL_DEBUG, "AudioWorklets: Created worklet node: %d\n", pParameters->pDevice->webaudio.audioWorklet); + pParameters->pDevice->webaudio.initResult = MA_SUCCESS; + ma_free(pParameters, &pParameters->pDevice->pContext->allocationCallbacks); +} + +static void ma_audio_worklet_thread_initialized__webaudio(EMSCRIPTEN_WEBAUDIO_T audioContext, EM_BOOL success, void* pUserData) +{ + ma_audio_worklet_thread_initialized_data* pParameters = (ma_audio_worklet_thread_initialized_data*)pUserData; + WebAudioWorkletProcessorCreateOptions workletProcessorOptions; + + MA_ASSERT(pParameters != NULL); + + if (success == EM_FALSE) { + pParameters->pDevice->webaudio.initResult = MA_ERROR; + return; + } + + MA_ZERO_OBJECT(&workletProcessorOptions); + workletProcessorOptions.name = "miniaudio"; /* I'm not entirely sure what to call this. Does this need to be globally unique, or does it need only be unique for a given AudioContext? */ + + emscripten_create_wasm_audio_worklet_processor_async(audioContext, &workletProcessorOptions, ma_audio_worklet_processor_created__webaudio, pParameters); +} +#endif + +static ma_result ma_device_init__webaudio(ma_device* pDevice, const ma_device_config* pConfig, ma_device_descriptor* pDescriptorPlayback, ma_device_descriptor* pDescriptorCapture) +{ + if (pConfig->deviceType == ma_device_type_loopback) { + return MA_DEVICE_TYPE_NOT_SUPPORTED; + } + + /* No exclusive mode with Web Audio. */ + if (((pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) && pDescriptorPlayback->shareMode == ma_share_mode_exclusive) || + ((pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) && pDescriptorCapture->shareMode == ma_share_mode_exclusive)) { + return MA_SHARE_MODE_NOT_SUPPORTED; + } + + /* + With AudioWorklets we'll have just a single AudioContext. I'm not sure why I'm not doing this for ScriptProcessorNode so + it might be worthwhile to look into that as well. + */ + #if defined(MA_USE_AUDIO_WORKLETS) + { + EmscriptenWebAudioCreateAttributes audioContextAttributes; + ma_audio_worklet_thread_initialized_data* pInitParameters; + void* pStackBuffer; + + if (pConfig->performanceProfile == ma_performance_profile_conservative) { + audioContextAttributes.latencyHint = MA_WEBAUDIO_LATENCY_HINT_PLAYBACK; + } else { + audioContextAttributes.latencyHint = MA_WEBAUDIO_LATENCY_HINT_INTERACTIVE; + } + + /* + In my testing, Firefox does not seem to capture audio data properly if the sample rate is set + to anything other than 48K. This does not seem to be the case for other browsers. For this reason, + if the device type is anything other than playback, we'll leave the sample rate as-is and let the + browser pick the appropriate rate for us. + */ + if (pConfig->deviceType == ma_device_type_playback) { + audioContextAttributes.sampleRate = pDescriptorPlayback->sampleRate; + } else { + audioContextAttributes.sampleRate = 0; + } + + /* It's not clear if this can return an error. None of the tests in the Emscripten repository check for this, so neither am I for now. */ + pDevice->webaudio.audioContext = emscripten_create_audio_context(&audioContextAttributes); + + /* + With the context created we can now create the worklet. We can only have a single worklet per audio + context which means we'll need to craft this appropriately to handle duplex devices correctly. + */ + + /* + We now need to create a worker thread. This is a bit weird because we need to allocate our + own buffer for the thread's stack. The stack needs to be aligned to 16 bytes. I'm going to + allocate this on the heap to keep it simple. + */ + pStackBuffer = ma_aligned_malloc(MA_AUDIO_WORKLETS_THREAD_STACK_SIZE, 16, &pDevice->pContext->allocationCallbacks); + if (pStackBuffer == NULL) { + emscripten_destroy_audio_context(pDevice->webaudio.audioContext); + return MA_OUT_OF_MEMORY; + } + + /* Our thread initialization parameters need to be allocated on the heap so they don't go out of scope. */ + pInitParameters = (ma_audio_worklet_thread_initialized_data*)ma_malloc(sizeof(*pInitParameters), &pDevice->pContext->allocationCallbacks); + if (pInitParameters == NULL) { + ma_free(pStackBuffer, &pDevice->pContext->allocationCallbacks); + emscripten_destroy_audio_context(pDevice->webaudio.audioContext); + return MA_OUT_OF_MEMORY; + } + + pInitParameters->pDevice = pDevice; + pInitParameters->pConfig = pConfig; + pInitParameters->pDescriptorPlayback = pDescriptorPlayback; + pInitParameters->pDescriptorCapture = pDescriptorCapture; + + /* + We need to flag the device as not yet initialized so we can wait on it later. Unfortunately all of + the Emscripten WebAudio stuff is asynchronous. + */ + pDevice->webaudio.initResult = MA_BUSY; + { + emscripten_start_wasm_audio_worklet_thread_async(pDevice->webaudio.audioContext, pStackBuffer, MA_AUDIO_WORKLETS_THREAD_STACK_SIZE, ma_audio_worklet_thread_initialized__webaudio, pInitParameters); + } + while (pDevice->webaudio.initResult == MA_BUSY) { emscripten_sleep(1); } /* We must wait for initialization to complete. We're just spinning here. The emscripten_sleep() call is why we need to build with `-sASYNCIFY`. */ + + /* Initialization is now complete. Descriptors were updated when the worklet was initialized. */ + if (pDevice->webaudio.initResult != MA_SUCCESS) { + ma_free(pStackBuffer, &pDevice->pContext->allocationCallbacks); + emscripten_destroy_audio_context(pDevice->webaudio.audioContext); + return pDevice->webaudio.initResult; + } + + /* We need to add an entry to the miniaudio.devices list on the JS side so we can do some JS/C interop. */ + pDevice->webaudio.deviceIndex = EM_ASM_INT({ + return window.miniaudio.track_device({ + webaudio: emscriptenGetAudioObject($0), + state: 1, /* 1 = ma_device_state_stopped */ + pDevice: $1 + }); + }, pDevice->webaudio.audioContext, pDevice); + + return MA_SUCCESS; + } + #else + { + /* ScriptProcessorNode. This path requires us to do almost everything in JS, but we'll do as much as we can in C. */ + ma_uint32 deviceIndex; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_uint32 periodSizeInFrames; + + /* The channel count will depend on the device type. If it's a capture, use its, otherwise use the playback side. */ + if (pConfig->deviceType == ma_device_type_capture) { + channels = (pDescriptorCapture->channels > 0) ? pDescriptorCapture->channels : MA_DEFAULT_CHANNELS; + } else { + channels = (pDescriptorPlayback->channels > 0) ? pDescriptorPlayback->channels : MA_DEFAULT_CHANNELS; + } + + /* + When testing in Firefox, I've seen it where capture mode fails if the sample rate is changed to anything other than it's + native rate. For this reason we're leaving the sample rate untouched for capture devices. + */ + if (pConfig->deviceType == ma_device_type_playback) { + sampleRate = pDescriptorPlayback->sampleRate; + } else { + sampleRate = 0; /* Let the browser decide when capturing. */ + } + + /* The period size needs to be a power of 2. */ + if (pConfig->deviceType == ma_device_type_capture) { + periodSizeInFrames = ma_calculate_period_size_in_frames_from_descriptor__webaudio(pDescriptorCapture, sampleRate, pConfig->performanceProfile); + } else { + periodSizeInFrames = ma_calculate_period_size_in_frames_from_descriptor__webaudio(pDescriptorPlayback, sampleRate, pConfig->performanceProfile); + } + + /* We need an intermediary buffer for doing interleaving and deinterleaving. */ + pDevice->webaudio.pIntermediaryBuffer = (float*)ma_malloc(periodSizeInFrames * channels * sizeof(float), &pDevice->pContext->allocationCallbacks); + if (pDevice->webaudio.pIntermediaryBuffer == NULL) { + return MA_OUT_OF_MEMORY; + } + + deviceIndex = EM_ASM_INT({ + var deviceType = $0; + var channels = $1; + var sampleRate = $2; + var bufferSize = $3; + var pIntermediaryBuffer = $4; + var pDevice = $5; + + if (typeof(window.miniaudio) === 'undefined') { + return -1; /* Context not initialized. */ + } + + var device = {}; + + /* First thing we need is an AudioContext. */ + var audioContextOptions = {}; + if (deviceType == window.miniaudio.device_type.playback && sampleRate != 0) { + audioContextOptions.sampleRate = sampleRate; + } + + device.webaudio = new (window.AudioContext || window.webkitAudioContext)(audioContextOptions); + device.webaudio.suspend(); /* The AudioContext must be created in a suspended state. */ + device.state = window.miniaudio.device_state.stopped; + + /* + We need to create a ScriptProcessorNode. The channel situation is the same as the AudioWorklet path in that we + need to specify an output and configure the channel count there. + */ + var channelCountIn = 0; + var channelCountOut = channels; + if (deviceType != window.miniaudio.device_type.playback) { + channelCountIn = channels; + } + + device.scriptNode = device.webaudio.createScriptProcessor(bufferSize, channelCountIn, channelCountOut); + + /* The node processing callback. */ + device.scriptNode.onaudioprocess = function(e) { + if (device.intermediaryBufferView == null || device.intermediaryBufferView.length == 0) { + device.intermediaryBufferView = new Float32Array(HEAPF32.buffer, pIntermediaryBuffer, bufferSize * channels); + } + + /* Do the capture side first. */ + if (deviceType == window.miniaudio.device_type.capture || deviceType == window.miniaudio.device_type.duplex) { + /* The data must be interleaved before being processed miniaudio. */ + for (var iChannel = 0; iChannel < channels; iChannel += 1) { + var inputBuffer = e.inputBuffer.getChannelData(iChannel); + var intermediaryBuffer = device.intermediaryBufferView; + + for (var iFrame = 0; iFrame < bufferSize; iFrame += 1) { + intermediaryBuffer[iFrame*channels + iChannel] = inputBuffer[iFrame]; + } + } + + _ma_device_process_pcm_frames_capture__webaudio(pDevice, bufferSize, pIntermediaryBuffer); + } + + if (deviceType == window.miniaudio.device_type.playback || deviceType == window.miniaudio.device_type.duplex) { + _ma_device_process_pcm_frames_playback__webaudio(pDevice, bufferSize, pIntermediaryBuffer); + + for (var iChannel = 0; iChannel < e.outputBuffer.numberOfChannels; ++iChannel) { + var outputBuffer = e.outputBuffer.getChannelData(iChannel); + var intermediaryBuffer = device.intermediaryBufferView; + + for (var iFrame = 0; iFrame < bufferSize; iFrame += 1) { + outputBuffer[iFrame] = intermediaryBuffer[iFrame*channels + iChannel]; + } + } + } else { + /* It's a capture-only device. Make sure the output is silenced. */ + for (var iChannel = 0; iChannel < e.outputBuffer.numberOfChannels; ++iChannel) { + e.outputBuffer.getChannelData(iChannel).fill(0.0); + } + } + }; + + /* Now we need to connect our node to the graph. */ + if (deviceType == window.miniaudio.device_type.capture || deviceType == window.miniaudio.device_type.duplex) { + navigator.mediaDevices.getUserMedia({audio:true, video:false}) + .then(function(stream) { + device.streamNode = device.webaudio.createMediaStreamSource(stream); + device.streamNode.connect(device.scriptNode); + device.scriptNode.connect(device.webaudio.destination); + }) + .catch(function(error) { + console.log("Failed to get user media: " + error); + }); + } + + if (deviceType == window.miniaudio.device_type.playback) { + device.scriptNode.connect(device.webaudio.destination); + } + + device.pDevice = pDevice; + + return window.miniaudio.track_device(device); + }, pConfig->deviceType, channels, sampleRate, periodSizeInFrames, pDevice->webaudio.pIntermediaryBuffer, pDevice); + + if (deviceIndex < 0) { + return MA_FAILED_TO_OPEN_BACKEND_DEVICE; + } + + pDevice->webaudio.deviceIndex = deviceIndex; + + /* Grab the sample rate from the audio context directly. */ + sampleRate = (ma_uint32)EM_ASM_INT({ return window.miniaudio.get_device_by_index($0).webaudio.sampleRate; }, deviceIndex); + + if (pDescriptorCapture != NULL) { + pDescriptorCapture->format = ma_format_f32; + pDescriptorCapture->channels = channels; + pDescriptorCapture->sampleRate = sampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_webaudio, pDescriptorCapture->channelMap, ma_countof(pDescriptorCapture->channelMap), pDescriptorCapture->channels); + pDescriptorCapture->periodSizeInFrames = periodSizeInFrames; + pDescriptorCapture->periodCount = 1; + } + + if (pDescriptorPlayback != NULL) { + pDescriptorPlayback->format = ma_format_f32; + pDescriptorPlayback->channels = channels; + pDescriptorPlayback->sampleRate = sampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_webaudio, pDescriptorPlayback->channelMap, ma_countof(pDescriptorPlayback->channelMap), pDescriptorPlayback->channels); + pDescriptorPlayback->periodSizeInFrames = periodSizeInFrames; + pDescriptorPlayback->periodCount = 1; + } + + return MA_SUCCESS; + } + #endif +} + +static ma_result ma_device_start__webaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + EM_ASM({ + var device = window.miniaudio.get_device_by_index($0); + device.webaudio.resume(); + device.state = window.miniaudio.device_state.started; + }, pDevice->webaudio.deviceIndex); + + return MA_SUCCESS; +} + +static ma_result ma_device_stop__webaudio(ma_device* pDevice) +{ + MA_ASSERT(pDevice != NULL); + + /* + From the WebAudio API documentation for AudioContext.suspend(): + + Suspends the progression of AudioContext's currentTime, allows any current context processing blocks that are already processed to be played to the + destination, and then allows the system to release its claim on audio hardware. + + I read this to mean that "any current context processing blocks" are processed by suspend() - i.e. They they are drained. We therefore shouldn't need to + do any kind of explicit draining. + */ + EM_ASM({ + var device = window.miniaudio.get_device_by_index($0); + device.webaudio.suspend(); + device.state = window.miniaudio.device_state.stopped; + }, pDevice->webaudio.deviceIndex); + + ma_device__on_notification_stopped(pDevice); + + return MA_SUCCESS; +} + +static ma_result ma_context_uninit__webaudio(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + MA_ASSERT(pContext->backend == ma_backend_webaudio); + + (void)pContext; /* Unused. */ + + /* Remove the global miniaudio object from window if there are no more references to it. */ + EM_ASM({ + if (typeof(window.miniaudio) !== 'undefined') { + miniaudio.unlock_event_types.map(function(event_type) { + document.removeEventListener(event_type, miniaudio.unlock, true); + }); + + window.miniaudio.referenceCount -= 1; + if (window.miniaudio.referenceCount === 0) { + delete window.miniaudio; + } + } + }); + + return MA_SUCCESS; +} + +static ma_result ma_context_init__webaudio(ma_context* pContext, const ma_context_config* pConfig, ma_backend_callbacks* pCallbacks) +{ + int resultFromJS; + + MA_ASSERT(pContext != NULL); + + (void)pConfig; /* Unused. */ + + /* Here is where our global JavaScript object is initialized. */ + resultFromJS = EM_ASM_INT({ + if (typeof window === 'undefined' || (window.AudioContext || window.webkitAudioContext) === undefined) { + return 0; /* Web Audio not supported. */ + } + + if (typeof(window.miniaudio) === 'undefined') { + window.miniaudio = { + referenceCount: 0 + }; + + /* Device types. */ + window.miniaudio.device_type = {}; + window.miniaudio.device_type.playback = $0; + window.miniaudio.device_type.capture = $1; + window.miniaudio.device_type.duplex = $2; + + /* Device states. */ + window.miniaudio.device_state = {}; + window.miniaudio.device_state.stopped = $3; + window.miniaudio.device_state.started = $4; + + /* Device cache for mapping devices to indexes for JavaScript/C interop. */ + let miniaudio = window.miniaudio; + miniaudio.devices = []; + + miniaudio.track_device = function(device) { + /* Try inserting into a free slot first. */ + for (var iDevice = 0; iDevice < miniaudio.devices.length; ++iDevice) { + if (miniaudio.devices[iDevice] == null) { + miniaudio.devices[iDevice] = device; + return iDevice; + } + } + + /* Getting here means there is no empty slots in the array so we just push to the end. */ + miniaudio.devices.push(device); + return miniaudio.devices.length - 1; + }; + + miniaudio.untrack_device_by_index = function(deviceIndex) { + /* We just set the device's slot to null. The slot will get reused in the next call to ma_track_device. */ + miniaudio.devices[deviceIndex] = null; + + /* Trim the array if possible. */ + while (miniaudio.devices.length > 0) { + if (miniaudio.devices[miniaudio.devices.length-1] == null) { + miniaudio.devices.pop(); + } else { + break; + } + } + }; + + miniaudio.untrack_device = function(device) { + for (var iDevice = 0; iDevice < miniaudio.devices.length; ++iDevice) { + if (miniaudio.devices[iDevice] == device) { + return miniaudio.untrack_device_by_index(iDevice); + } + } + }; + + miniaudio.get_device_by_index = function(deviceIndex) { + return miniaudio.devices[deviceIndex]; + }; + + miniaudio.unlock_event_types = (function(){ + return ['touchend', 'click']; + })(); + + miniaudio.unlock = function() { + for(var i = 0; i < miniaudio.devices.length; ++i) { + var device = miniaudio.devices[i]; + if (device != null && + device.webaudio != null && + device.state === miniaudio.device_state.started) { + + device.webaudio.resume().then(() => { + _ma_device__on_notification_unlocked(device.pDevice); + }, + (error) => {console.error("Failed to resume audiocontext", error); + }); + } + } + miniaudio.unlock_event_types.map(function(event_type) { + document.removeEventListener(event_type, miniaudio.unlock, true); + }); + }; + + miniaudio.unlock_event_types.map(function(event_type) { + document.addEventListener(event_type, miniaudio.unlock, true); + }); + } + + window.miniaudio.referenceCount += 1; + + return 1; + }, ma_device_type_playback, ma_device_type_capture, ma_device_type_duplex, ma_device_state_stopped, ma_device_state_started); + + if (resultFromJS != 1) { + return MA_FAILED_TO_INIT_BACKEND; + } + + pCallbacks->onContextInit = ma_context_init__webaudio; + pCallbacks->onContextUninit = ma_context_uninit__webaudio; + pCallbacks->onContextEnumerateDevices = ma_context_enumerate_devices__webaudio; + pCallbacks->onContextGetDeviceInfo = ma_context_get_device_info__webaudio; + pCallbacks->onDeviceInit = ma_device_init__webaudio; + pCallbacks->onDeviceUninit = ma_device_uninit__webaudio; + pCallbacks->onDeviceStart = ma_device_start__webaudio; + pCallbacks->onDeviceStop = ma_device_stop__webaudio; + pCallbacks->onDeviceRead = NULL; /* Not needed because WebAudio is asynchronous. */ + pCallbacks->onDeviceWrite = NULL; /* Not needed because WebAudio is asynchronous. */ + pCallbacks->onDeviceDataLoop = NULL; /* Not needed because WebAudio is asynchronous. */ + + return MA_SUCCESS; +} +#endif /* MA_HAS_WEBAUDIO */ + + + +static ma_bool32 ma__is_channel_map_valid(const ma_channel* pChannelMap, ma_uint32 channels) +{ + /* A blank channel map should be allowed, in which case it should use an appropriate default which will depend on context. */ + if (pChannelMap != NULL && pChannelMap[0] != MA_CHANNEL_NONE) { + ma_uint32 iChannel; + + if (channels == 0 || channels > MA_MAX_CHANNELS) { + return MA_FALSE; /* Channel count out of range. */ + } + + /* A channel cannot be present in the channel map more than once. */ + for (iChannel = 0; iChannel < channels; ++iChannel) { + ma_uint32 jChannel; + for (jChannel = iChannel + 1; jChannel < channels; ++jChannel) { + if (pChannelMap[iChannel] == pChannelMap[jChannel]) { + return MA_FALSE; + } + } + } + } + + return MA_TRUE; +} + + +static ma_bool32 ma_context_is_backend_asynchronous(ma_context* pContext) +{ + MA_ASSERT(pContext != NULL); + + if (pContext->callbacks.onDeviceRead == NULL && pContext->callbacks.onDeviceWrite == NULL) { + if (pContext->callbacks.onDeviceDataLoop == NULL) { + return MA_TRUE; + } else { + return MA_FALSE; + } + } else { + return MA_FALSE; + } +} + + +static ma_result ma_device__post_init_setup(ma_device* pDevice, ma_device_type deviceType) +{ + ma_result result; + + MA_ASSERT(pDevice != NULL); + + if (deviceType == ma_device_type_capture || deviceType == ma_device_type_duplex || deviceType == ma_device_type_loopback) { + if (pDevice->capture.format == ma_format_unknown) { + pDevice->capture.format = pDevice->capture.internalFormat; + } + if (pDevice->capture.channels == 0) { + pDevice->capture.channels = pDevice->capture.internalChannels; + } + if (pDevice->capture.channelMap[0] == MA_CHANNEL_NONE) { + MA_ASSERT(pDevice->capture.channels <= MA_MAX_CHANNELS); + if (pDevice->capture.internalChannels == pDevice->capture.channels) { + ma_channel_map_copy(pDevice->capture.channelMap, pDevice->capture.internalChannelMap, pDevice->capture.channels); + } else { + if (pDevice->capture.channelMixMode == ma_channel_mix_mode_simple) { + ma_channel_map_init_blank(pDevice->capture.channelMap, pDevice->capture.channels); + } else { + ma_channel_map_init_standard(ma_standard_channel_map_default, pDevice->capture.channelMap, ma_countof(pDevice->capture.channelMap), pDevice->capture.channels); + } + } + } + } + + if (deviceType == ma_device_type_playback || deviceType == ma_device_type_duplex) { + if (pDevice->playback.format == ma_format_unknown) { + pDevice->playback.format = pDevice->playback.internalFormat; + } + if (pDevice->playback.channels == 0) { + pDevice->playback.channels = pDevice->playback.internalChannels; + } + if (pDevice->playback.channelMap[0] == MA_CHANNEL_NONE) { + MA_ASSERT(pDevice->playback.channels <= MA_MAX_CHANNELS); + if (pDevice->playback.internalChannels == pDevice->playback.channels) { + ma_channel_map_copy(pDevice->playback.channelMap, pDevice->playback.internalChannelMap, pDevice->playback.channels); + } else { + if (pDevice->playback.channelMixMode == ma_channel_mix_mode_simple) { + ma_channel_map_init_blank(pDevice->playback.channelMap, pDevice->playback.channels); + } else { + ma_channel_map_init_standard(ma_standard_channel_map_default, pDevice->playback.channelMap, ma_countof(pDevice->playback.channelMap), pDevice->playback.channels); + } + } + } + } + + if (pDevice->sampleRate == 0) { + if (deviceType == ma_device_type_capture || deviceType == ma_device_type_duplex || deviceType == ma_device_type_loopback) { + pDevice->sampleRate = pDevice->capture.internalSampleRate; + } else { + pDevice->sampleRate = pDevice->playback.internalSampleRate; + } + } + + /* Data converters. */ + if (deviceType == ma_device_type_capture || deviceType == ma_device_type_duplex || deviceType == ma_device_type_loopback) { + /* Converting from internal device format to client format. */ + ma_data_converter_config converterConfig = ma_data_converter_config_init_default(); + converterConfig.formatIn = pDevice->capture.internalFormat; + converterConfig.channelsIn = pDevice->capture.internalChannels; + converterConfig.sampleRateIn = pDevice->capture.internalSampleRate; + converterConfig.pChannelMapIn = pDevice->capture.internalChannelMap; + converterConfig.formatOut = pDevice->capture.format; + converterConfig.channelsOut = pDevice->capture.channels; + converterConfig.sampleRateOut = pDevice->sampleRate; + converterConfig.pChannelMapOut = pDevice->capture.channelMap; + converterConfig.channelMixMode = pDevice->capture.channelMixMode; + converterConfig.calculateLFEFromSpatialChannels = pDevice->capture.calculateLFEFromSpatialChannels; + converterConfig.allowDynamicSampleRate = MA_FALSE; + converterConfig.resampling.algorithm = pDevice->resampling.algorithm; + converterConfig.resampling.linear.lpfOrder = pDevice->resampling.linear.lpfOrder; + converterConfig.resampling.pBackendVTable = pDevice->resampling.pBackendVTable; + converterConfig.resampling.pBackendUserData = pDevice->resampling.pBackendUserData; + + /* Make sure the old converter is uninitialized first. */ + if (ma_device_get_state(pDevice) != ma_device_state_uninitialized) { + ma_data_converter_uninit(&pDevice->capture.converter, &pDevice->pContext->allocationCallbacks); + } + + result = ma_data_converter_init(&converterConfig, &pDevice->pContext->allocationCallbacks, &pDevice->capture.converter); + if (result != MA_SUCCESS) { + return result; + } + } + + if (deviceType == ma_device_type_playback || deviceType == ma_device_type_duplex) { + /* Converting from client format to device format. */ + ma_data_converter_config converterConfig = ma_data_converter_config_init_default(); + converterConfig.formatIn = pDevice->playback.format; + converterConfig.channelsIn = pDevice->playback.channels; + converterConfig.sampleRateIn = pDevice->sampleRate; + converterConfig.pChannelMapIn = pDevice->playback.channelMap; + converterConfig.formatOut = pDevice->playback.internalFormat; + converterConfig.channelsOut = pDevice->playback.internalChannels; + converterConfig.sampleRateOut = pDevice->playback.internalSampleRate; + converterConfig.pChannelMapOut = pDevice->playback.internalChannelMap; + converterConfig.channelMixMode = pDevice->playback.channelMixMode; + converterConfig.calculateLFEFromSpatialChannels = pDevice->playback.calculateLFEFromSpatialChannels; + converterConfig.allowDynamicSampleRate = MA_FALSE; + converterConfig.resampling.algorithm = pDevice->resampling.algorithm; + converterConfig.resampling.linear.lpfOrder = pDevice->resampling.linear.lpfOrder; + converterConfig.resampling.pBackendVTable = pDevice->resampling.pBackendVTable; + converterConfig.resampling.pBackendUserData = pDevice->resampling.pBackendUserData; + + /* Make sure the old converter is uninitialized first. */ + if (ma_device_get_state(pDevice) != ma_device_state_uninitialized) { + ma_data_converter_uninit(&pDevice->playback.converter, &pDevice->pContext->allocationCallbacks); + } + + result = ma_data_converter_init(&converterConfig, &pDevice->pContext->allocationCallbacks, &pDevice->playback.converter); + if (result != MA_SUCCESS) { + return result; + } + } + + + /* + If the device is doing playback (ma_device_type_playback or ma_device_type_duplex), there's + a couple of situations where we'll need a heap allocated cache. + + The first is a duplex device for backends that use a callback for data delivery. The reason + this is needed is that the input stage needs to have a buffer to place the input data while it + waits for the playback stage, after which the miniaudio data callback will get fired. This is + not needed for backends that use a blocking API because miniaudio manages temporary buffers on + the stack to achieve this. + + The other situation is when the data converter does not have the ability to query the number + of input frames that are required in order to process a given number of output frames. When + performing data conversion, it's useful if miniaudio know exactly how many frames it needs + from the client in order to generate a given number of output frames. This way, only exactly + the number of frames are needed to be read from the client which means no cache is necessary. + On the other hand, if miniaudio doesn't know how many frames to read, it is forced to read + in fixed sized chunks and then cache any residual unused input frames, those of which will be + processed at a later stage. + */ + if (deviceType == ma_device_type_playback || deviceType == ma_device_type_duplex) { + ma_uint64 unused; + + pDevice->playback.inputCacheConsumed = 0; + pDevice->playback.inputCacheRemaining = 0; + + if (pDevice->type == ma_device_type_duplex || /* Duplex. backend may decide to use ma_device_handle_backend_data_callback() which will require this cache. */ + ma_data_converter_get_required_input_frame_count(&pDevice->playback.converter, 1, &unused) != MA_SUCCESS) /* Data conversion required input frame calculation not supported. */ + { + /* We need a heap allocated cache. We want to size this based on the period size. */ + void* pNewInputCache; + ma_uint64 newInputCacheCap; + ma_uint64 newInputCacheSizeInBytes; + + newInputCacheCap = ma_calculate_frame_count_after_resampling(pDevice->playback.internalSampleRate, pDevice->sampleRate, pDevice->playback.internalPeriodSizeInFrames); + + newInputCacheSizeInBytes = newInputCacheCap * ma_get_bytes_per_frame(pDevice->playback.format, pDevice->playback.channels); + if (newInputCacheSizeInBytes > MA_SIZE_MAX) { + ma_free(pDevice->playback.pInputCache, &pDevice->pContext->allocationCallbacks); + pDevice->playback.pInputCache = NULL; + pDevice->playback.inputCacheCap = 0; + return MA_OUT_OF_MEMORY; /* Allocation too big. Should never hit this, but makes the cast below safer for 32-bit builds. */ + } + + pNewInputCache = ma_realloc(pDevice->playback.pInputCache, (size_t)newInputCacheSizeInBytes, &pDevice->pContext->allocationCallbacks); + if (pNewInputCache == NULL) { + ma_free(pDevice->playback.pInputCache, &pDevice->pContext->allocationCallbacks); + pDevice->playback.pInputCache = NULL; + pDevice->playback.inputCacheCap = 0; + return MA_OUT_OF_MEMORY; + } + + pDevice->playback.pInputCache = pNewInputCache; + pDevice->playback.inputCacheCap = newInputCacheCap; + } else { + /* Heap allocation not required. Make sure we clear out the old cache just in case this function was called in response to a route change. */ + ma_free(pDevice->playback.pInputCache, &pDevice->pContext->allocationCallbacks); + pDevice->playback.pInputCache = NULL; + pDevice->playback.inputCacheCap = 0; + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_device_post_init(ma_device* pDevice, ma_device_type deviceType, const ma_device_descriptor* pDescriptorPlayback, const ma_device_descriptor* pDescriptorCapture) +{ + ma_result result; + + if (pDevice == NULL) { + return MA_INVALID_ARGS; + } + + /* Capture. */ + if (deviceType == ma_device_type_capture || deviceType == ma_device_type_duplex || deviceType == ma_device_type_loopback) { + if (ma_device_descriptor_is_valid(pDescriptorCapture) == MA_FALSE) { + return MA_INVALID_ARGS; + } + + pDevice->capture.internalFormat = pDescriptorCapture->format; + pDevice->capture.internalChannels = pDescriptorCapture->channels; + pDevice->capture.internalSampleRate = pDescriptorCapture->sampleRate; + MA_COPY_MEMORY(pDevice->capture.internalChannelMap, pDescriptorCapture->channelMap, sizeof(pDescriptorCapture->channelMap)); + pDevice->capture.internalPeriodSizeInFrames = pDescriptorCapture->periodSizeInFrames; + pDevice->capture.internalPeriods = pDescriptorCapture->periodCount; + + if (pDevice->capture.internalPeriodSizeInFrames == 0) { + pDevice->capture.internalPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(pDescriptorCapture->periodSizeInMilliseconds, pDescriptorCapture->sampleRate); + } + } + + /* Playback. */ + if (deviceType == ma_device_type_playback || deviceType == ma_device_type_duplex) { + if (ma_device_descriptor_is_valid(pDescriptorPlayback) == MA_FALSE) { + return MA_INVALID_ARGS; + } + + pDevice->playback.internalFormat = pDescriptorPlayback->format; + pDevice->playback.internalChannels = pDescriptorPlayback->channels; + pDevice->playback.internalSampleRate = pDescriptorPlayback->sampleRate; + MA_COPY_MEMORY(pDevice->playback.internalChannelMap, pDescriptorPlayback->channelMap, sizeof(pDescriptorPlayback->channelMap)); + pDevice->playback.internalPeriodSizeInFrames = pDescriptorPlayback->periodSizeInFrames; + pDevice->playback.internalPeriods = pDescriptorPlayback->periodCount; + + if (pDevice->playback.internalPeriodSizeInFrames == 0) { + pDevice->playback.internalPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(pDescriptorPlayback->periodSizeInMilliseconds, pDescriptorPlayback->sampleRate); + } + } + + /* + The name of the device can be retrieved from device info. This may be temporary and replaced with a `ma_device_get_info(pDevice, deviceType)` instead. + For loopback devices, we need to retrieve the name of the playback device. + */ + { + ma_device_info deviceInfo; + + if (deviceType == ma_device_type_capture || deviceType == ma_device_type_duplex || deviceType == ma_device_type_loopback) { + result = ma_device_get_info(pDevice, ma_device_type_capture, &deviceInfo); + if (result == MA_SUCCESS) { + ma_strncpy_s(pDevice->capture.name, sizeof(pDevice->capture.name), deviceInfo.name, (size_t)-1); + } else { + /* We failed to retrieve the device info. Fall back to a default name. */ + if (pDescriptorCapture->pDeviceID == NULL) { + ma_strncpy_s(pDevice->capture.name, sizeof(pDevice->capture.name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDevice->capture.name, sizeof(pDevice->capture.name), "Capture Device", (size_t)-1); + } + } + } + + if (deviceType == ma_device_type_playback || deviceType == ma_device_type_duplex) { + result = ma_device_get_info(pDevice, ma_device_type_playback, &deviceInfo); + if (result == MA_SUCCESS) { + ma_strncpy_s(pDevice->playback.name, sizeof(pDevice->playback.name), deviceInfo.name, (size_t)-1); + } else { + /* We failed to retrieve the device info. Fall back to a default name. */ + if (pDescriptorPlayback->pDeviceID == NULL) { + ma_strncpy_s(pDevice->playback.name, sizeof(pDevice->playback.name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDevice->playback.name, sizeof(pDevice->playback.name), "Playback Device", (size_t)-1); + } + } + } + } + + /* Update data conversion. */ + return ma_device__post_init_setup(pDevice, deviceType); /* TODO: Should probably rename ma_device__post_init_setup() to something better. */ +} + + +static ma_thread_result MA_THREADCALL ma_worker_thread(void* pData) +{ + ma_device* pDevice = (ma_device*)pData; +#ifdef MA_WIN32 + HRESULT CoInitializeResult; +#endif + + MA_ASSERT(pDevice != NULL); + +#ifdef MA_WIN32 + CoInitializeResult = ma_CoInitializeEx(pDevice->pContext, NULL, MA_COINIT_VALUE); +#endif + + /* + When the device is being initialized its initial state is set to ma_device_state_uninitialized. Before returning from + ma_device_init(), the state needs to be set to something valid. In miniaudio the device's default state immediately + after initialization is stopped, so therefore we need to mark the device as such. miniaudio will wait on the worker + thread to signal an event to know when the worker thread is ready for action. + */ + ma_device__set_state(pDevice, ma_device_state_stopped); + ma_event_signal(&pDevice->stopEvent); + + for (;;) { /* <-- This loop just keeps the thread alive. The main audio loop is inside. */ + ma_result startResult; + ma_result stopResult; /* <-- This will store the result from onDeviceStop(). If it returns an error, we don't fire the stopped notification callback. */ + + /* We wait on an event to know when something has requested that the device be started and the main loop entered. */ + ma_event_wait(&pDevice->wakeupEvent); + + /* Default result code. */ + pDevice->workResult = MA_SUCCESS; + + /* If the reason for the wake up is that we are terminating, just break from the loop. */ + if (ma_device_get_state(pDevice) == ma_device_state_uninitialized) { + break; + } + + /* + Getting to this point means the device is wanting to get started. The function that has requested that the device + be started will be waiting on an event (pDevice->startEvent) which means we need to make sure we signal the event + in both the success and error case. It's important that the state of the device is set _before_ signaling the event. + */ + MA_ASSERT(ma_device_get_state(pDevice) == ma_device_state_starting); + + /* If the device has a start callback, start it now. */ + if (pDevice->pContext->callbacks.onDeviceStart != NULL) { + startResult = pDevice->pContext->callbacks.onDeviceStart(pDevice); + } else { + startResult = MA_SUCCESS; + } + + /* + If starting was not successful we'll need to loop back to the start and wait for something + to happen (pDevice->wakeupEvent). + */ + if (startResult != MA_SUCCESS) { + pDevice->workResult = startResult; + ma_event_signal(&pDevice->startEvent); /* <-- Always signal the start event so ma_device_start() can return as it'll be waiting on it. */ + continue; + } + + /* Make sure the state is set appropriately. */ + ma_device__set_state(pDevice, ma_device_state_started); /* <-- Set this before signaling the event so that the state is always guaranteed to be good after ma_device_start() has returned. */ + ma_event_signal(&pDevice->startEvent); + + ma_device__on_notification_started(pDevice); + + if (pDevice->pContext->callbacks.onDeviceDataLoop != NULL) { + pDevice->pContext->callbacks.onDeviceDataLoop(pDevice); + } else { + /* The backend is not using a custom main loop implementation, so now fall back to the blocking read-write implementation. */ + ma_device_audio_thread__default_read_write(pDevice); + } + + /* Getting here means we have broken from the main loop which happens the application has requested that device be stopped. */ + if (pDevice->pContext->callbacks.onDeviceStop != NULL) { + stopResult = pDevice->pContext->callbacks.onDeviceStop(pDevice); + } else { + stopResult = MA_SUCCESS; /* No stop callback with the backend. Just assume successful. */ + } + + /* + After the device has stopped, make sure an event is posted. Don't post a stopped event if + stopping failed. This can happen on some backends when the underlying stream has been + stopped due to the device being physically unplugged or disabled via an OS setting. + */ + if (stopResult == MA_SUCCESS) { + ma_device__on_notification_stopped(pDevice); + } + + /* If we stopped because the device has been uninitialized, abort now. */ + if (ma_device_get_state(pDevice) == ma_device_state_uninitialized) { + break; + } + + /* A function somewhere is waiting for the device to have stopped for real so we need to signal an event to allow it to continue. */ + ma_device__set_state(pDevice, ma_device_state_stopped); + ma_event_signal(&pDevice->stopEvent); + } + +#ifdef MA_WIN32 + if (CoInitializeResult == S_OK) { + ma_CoUninitialize(pDevice->pContext); + } +#endif + + return (ma_thread_result)0; +} + + +/* Helper for determining whether or not the given device is initialized. */ +static ma_bool32 ma_device__is_initialized(ma_device* pDevice) +{ + if (pDevice == NULL) { + return MA_FALSE; + } + + return ma_device_get_state(pDevice) != ma_device_state_uninitialized; +} + + +#ifdef MA_WIN32 +static ma_result ma_context_uninit_backend_apis__win32(ma_context* pContext) +{ + /* For some reason UWP complains when CoUninitialize() is called. I'm just not going to call it on UWP. */ +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + if (pContext->win32.CoInitializeResult == S_OK) { + ma_CoUninitialize(pContext); + } + + #if defined(MA_WIN32_DESKTOP) + ma_dlclose(ma_context_get_log(pContext), pContext->win32.hUser32DLL); + ma_dlclose(ma_context_get_log(pContext), pContext->win32.hAdvapi32DLL); + #endif + + ma_dlclose(ma_context_get_log(pContext), pContext->win32.hOle32DLL); +#else + (void)pContext; +#endif + + return MA_SUCCESS; +} + +static ma_result ma_context_init_backend_apis__win32(ma_context* pContext) +{ +#if defined(MA_WIN32_DESKTOP) || defined(MA_WIN32_GDK) + #if defined(MA_WIN32_DESKTOP) + /* User32.dll */ + pContext->win32.hUser32DLL = ma_dlopen(ma_context_get_log(pContext), "user32.dll"); + if (pContext->win32.hUser32DLL == NULL) { + return MA_FAILED_TO_INIT_BACKEND; + } + + pContext->win32.GetForegroundWindow = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hUser32DLL, "GetForegroundWindow"); + pContext->win32.GetDesktopWindow = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hUser32DLL, "GetDesktopWindow"); + + + /* Advapi32.dll */ + pContext->win32.hAdvapi32DLL = ma_dlopen(ma_context_get_log(pContext), "advapi32.dll"); + if (pContext->win32.hAdvapi32DLL == NULL) { + return MA_FAILED_TO_INIT_BACKEND; + } + + pContext->win32.RegOpenKeyExA = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hAdvapi32DLL, "RegOpenKeyExA"); + pContext->win32.RegCloseKey = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hAdvapi32DLL, "RegCloseKey"); + pContext->win32.RegQueryValueExA = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hAdvapi32DLL, "RegQueryValueExA"); + #endif + + /* Ole32.dll */ + pContext->win32.hOle32DLL = ma_dlopen(ma_context_get_log(pContext), "ole32.dll"); + if (pContext->win32.hOle32DLL == NULL) { + return MA_FAILED_TO_INIT_BACKEND; + } + + pContext->win32.CoInitialize = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hOle32DLL, "CoInitialize"); + pContext->win32.CoInitializeEx = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hOle32DLL, "CoInitializeEx"); + pContext->win32.CoUninitialize = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hOle32DLL, "CoUninitialize"); + pContext->win32.CoCreateInstance = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hOle32DLL, "CoCreateInstance"); + pContext->win32.CoTaskMemFree = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hOle32DLL, "CoTaskMemFree"); + pContext->win32.PropVariantClear = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hOle32DLL, "PropVariantClear"); + pContext->win32.StringFromGUID2 = (ma_proc)ma_dlsym(ma_context_get_log(pContext), pContext->win32.hOle32DLL, "StringFromGUID2"); +#else + (void)pContext; /* Unused. */ +#endif + + pContext->win32.CoInitializeResult = ma_CoInitializeEx(pContext, NULL, MA_COINIT_VALUE); + return MA_SUCCESS; +} +#else +static ma_result ma_context_uninit_backend_apis__nix(ma_context* pContext) +{ + (void)pContext; + + return MA_SUCCESS; +} + +static ma_result ma_context_init_backend_apis__nix(ma_context* pContext) +{ + (void)pContext; + + return MA_SUCCESS; +} +#endif + +static ma_result ma_context_init_backend_apis(ma_context* pContext) +{ + ma_result result; +#ifdef MA_WIN32 + result = ma_context_init_backend_apis__win32(pContext); +#else + result = ma_context_init_backend_apis__nix(pContext); +#endif + + return result; +} + +static ma_result ma_context_uninit_backend_apis(ma_context* pContext) +{ + ma_result result; +#ifdef MA_WIN32 + result = ma_context_uninit_backend_apis__win32(pContext); +#else + result = ma_context_uninit_backend_apis__nix(pContext); +#endif + + return result; +} + + +/* The default capacity doesn't need to be too big. */ +#ifndef MA_DEFAULT_DEVICE_JOB_QUEUE_CAPACITY +#define MA_DEFAULT_DEVICE_JOB_QUEUE_CAPACITY 32 +#endif + +MA_API ma_device_job_thread_config ma_device_job_thread_config_init(void) +{ + ma_device_job_thread_config config; + + MA_ZERO_OBJECT(&config); + config.noThread = MA_FALSE; + config.jobQueueCapacity = MA_DEFAULT_DEVICE_JOB_QUEUE_CAPACITY; + config.jobQueueFlags = 0; + + return config; +} + + +static ma_thread_result MA_THREADCALL ma_device_job_thread_entry(void* pUserData) +{ + ma_device_job_thread* pJobThread = (ma_device_job_thread*)pUserData; + MA_ASSERT(pJobThread != NULL); + + for (;;) { + ma_result result; + ma_job job; + + result = ma_device_job_thread_next(pJobThread, &job); + if (result != MA_SUCCESS) { + break; + } + + if (job.toc.breakup.code == MA_JOB_TYPE_QUIT) { + break; + } + + ma_job_process(&job); + } + + return (ma_thread_result)0; +} + +MA_API ma_result ma_device_job_thread_init(const ma_device_job_thread_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_device_job_thread* pJobThread) +{ + ma_result result; + ma_job_queue_config jobQueueConfig; + + if (pJobThread == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pJobThread); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + + /* Initialize the job queue before the thread to ensure it's in a valid state. */ + jobQueueConfig = ma_job_queue_config_init(pConfig->jobQueueFlags, pConfig->jobQueueCapacity); + + result = ma_job_queue_init(&jobQueueConfig, pAllocationCallbacks, &pJobThread->jobQueue); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize job queue. */ + } + + + /* The thread needs to be initialized after the job queue to ensure the thread doesn't try to access it prematurely. */ + if (pConfig->noThread == MA_FALSE) { + result = ma_thread_create(&pJobThread->thread, ma_thread_priority_normal, 0, ma_device_job_thread_entry, pJobThread, pAllocationCallbacks); + if (result != MA_SUCCESS) { + ma_job_queue_uninit(&pJobThread->jobQueue, pAllocationCallbacks); + return result; /* Failed to create the job thread. */ + } + + pJobThread->_hasThread = MA_TRUE; + } else { + pJobThread->_hasThread = MA_FALSE; + } + + + return MA_SUCCESS; +} + +MA_API void ma_device_job_thread_uninit(ma_device_job_thread* pJobThread, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pJobThread == NULL) { + return; + } + + /* The first thing to do is post a quit message to the job queue. If we're using a thread we'll need to wait for it. */ + { + ma_job job = ma_job_init(MA_JOB_TYPE_QUIT); + ma_device_job_thread_post(pJobThread, &job); + } + + /* Wait for the thread to terminate naturally. */ + if (pJobThread->_hasThread) { + ma_thread_wait(&pJobThread->thread); + } + + /* At this point the thread should be terminated so we can safely uninitialize the job queue. */ + ma_job_queue_uninit(&pJobThread->jobQueue, pAllocationCallbacks); +} + +MA_API ma_result ma_device_job_thread_post(ma_device_job_thread* pJobThread, const ma_job* pJob) +{ + if (pJobThread == NULL || pJob == NULL) { + return MA_INVALID_ARGS; + } + + return ma_job_queue_post(&pJobThread->jobQueue, pJob); +} + +MA_API ma_result ma_device_job_thread_next(ma_device_job_thread* pJobThread, ma_job* pJob) +{ + if (pJob == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pJob); + + if (pJobThread == NULL) { + return MA_INVALID_ARGS; + } + + return ma_job_queue_next(&pJobThread->jobQueue, pJob); +} + + +MA_API ma_bool32 ma_device_id_equal(const ma_device_id* pA, const ma_device_id* pB) +{ + size_t i; + + if (pA == NULL || pB == NULL) { + return MA_FALSE; + } + + for (i = 0; i < sizeof(ma_device_id); i += 1) { + if (((const char*)pA)[i] != ((const char*)pB)[i]) { + return MA_FALSE; + } + } + + return MA_TRUE; +} + + + +MA_API ma_context_config ma_context_config_init(void) +{ + ma_context_config config; + MA_ZERO_OBJECT(&config); + + return config; +} + +MA_API ma_result ma_context_init(const ma_backend backends[], ma_uint32 backendCount, const ma_context_config* pConfig, ma_context* pContext) +{ + ma_result result; + ma_context_config defaultConfig; + ma_backend defaultBackends[ma_backend_null+1]; + ma_uint32 iBackend; + ma_backend* pBackendsToIterate; + ma_uint32 backendsToIterateCount; + + if (pContext == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pContext); + + /* Always make sure the config is set first to ensure properties are available as soon as possible. */ + if (pConfig == NULL) { + defaultConfig = ma_context_config_init(); + pConfig = &defaultConfig; + } + + /* Allocation callbacks need to come first because they'll be passed around to other areas. */ + result = ma_allocation_callbacks_init_copy(&pContext->allocationCallbacks, &pConfig->allocationCallbacks); + if (result != MA_SUCCESS) { + return result; + } + + /* Get a lot set up first so we can start logging ASAP. */ + if (pConfig->pLog != NULL) { + pContext->pLog = pConfig->pLog; + } else { + result = ma_log_init(&pContext->allocationCallbacks, &pContext->log); + if (result == MA_SUCCESS) { + pContext->pLog = &pContext->log; + } else { + pContext->pLog = NULL; /* Logging is not available. */ + } + } + + pContext->threadPriority = pConfig->threadPriority; + pContext->threadStackSize = pConfig->threadStackSize; + pContext->pUserData = pConfig->pUserData; + + /* Backend APIs need to be initialized first. This is where external libraries will be loaded and linked. */ + result = ma_context_init_backend_apis(pContext); + if (result != MA_SUCCESS) { + return result; + } + + for (iBackend = 0; iBackend <= ma_backend_null; ++iBackend) { + defaultBackends[iBackend] = (ma_backend)iBackend; + } + + pBackendsToIterate = (ma_backend*)backends; + backendsToIterateCount = backendCount; + if (pBackendsToIterate == NULL) { + pBackendsToIterate = (ma_backend*)defaultBackends; + backendsToIterateCount = ma_countof(defaultBackends); + } + + MA_ASSERT(pBackendsToIterate != NULL); + + for (iBackend = 0; iBackend < backendsToIterateCount; iBackend += 1) { + ma_backend backend = pBackendsToIterate[iBackend]; + + /* Make sure all callbacks are reset so we don't accidentally drag in any from previously failed initialization attempts. */ + MA_ZERO_OBJECT(&pContext->callbacks); + + /* These backends are using the new callback system. */ + switch (backend) { + #ifdef MA_HAS_WASAPI + case ma_backend_wasapi: + { + pContext->callbacks.onContextInit = ma_context_init__wasapi; + } break; + #endif + #ifdef MA_HAS_DSOUND + case ma_backend_dsound: + { + pContext->callbacks.onContextInit = ma_context_init__dsound; + } break; + #endif + #ifdef MA_HAS_WINMM + case ma_backend_winmm: + { + pContext->callbacks.onContextInit = ma_context_init__winmm; + } break; + #endif + #ifdef MA_HAS_COREAUDIO + case ma_backend_coreaudio: + { + pContext->callbacks.onContextInit = ma_context_init__coreaudio; + } break; + #endif + #ifdef MA_HAS_SNDIO + case ma_backend_sndio: + { + pContext->callbacks.onContextInit = ma_context_init__sndio; + } break; + #endif + #ifdef MA_HAS_AUDIO4 + case ma_backend_audio4: + { + pContext->callbacks.onContextInit = ma_context_init__audio4; + } break; + #endif + #ifdef MA_HAS_OSS + case ma_backend_oss: + { + pContext->callbacks.onContextInit = ma_context_init__oss; + } break; + #endif + #ifdef MA_HAS_PULSEAUDIO + case ma_backend_pulseaudio: + { + pContext->callbacks.onContextInit = ma_context_init__pulse; + } break; + #endif + #ifdef MA_HAS_ALSA + case ma_backend_alsa: + { + pContext->callbacks.onContextInit = ma_context_init__alsa; + } break; + #endif + #ifdef MA_HAS_JACK + case ma_backend_jack: + { + pContext->callbacks.onContextInit = ma_context_init__jack; + } break; + #endif + #ifdef MA_HAS_AAUDIO + case ma_backend_aaudio: + { + if (ma_is_backend_enabled(backend)) { + pContext->callbacks.onContextInit = ma_context_init__aaudio; + } + } break; + #endif + #ifdef MA_HAS_OPENSL + case ma_backend_opensl: + { + if (ma_is_backend_enabled(backend)) { + pContext->callbacks.onContextInit = ma_context_init__opensl; + } + } break; + #endif + #ifdef MA_HAS_WEBAUDIO + case ma_backend_webaudio: + { + pContext->callbacks.onContextInit = ma_context_init__webaudio; + } break; + #endif + #ifdef MA_HAS_CUSTOM + case ma_backend_custom: + { + /* Slightly different logic for custom backends. Custom backends can optionally set all of their callbacks in the config. */ + pContext->callbacks = pConfig->custom; + } break; + #endif + #ifdef MA_HAS_NULL + case ma_backend_null: + { + pContext->callbacks.onContextInit = ma_context_init__null; + } break; + #endif + + default: break; + } + + if (pContext->callbacks.onContextInit != NULL) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "Attempting to initialize %s backend...\n", ma_get_backend_name(backend)); + result = pContext->callbacks.onContextInit(pContext, pConfig, &pContext->callbacks); + } else { + /* Getting here means the onContextInit callback is not set which means the backend is not enabled. Special case for the custom backend. */ + if (backend != ma_backend_custom) { + result = MA_BACKEND_NOT_ENABLED; + } else { + #if !defined(MA_HAS_CUSTOM) + result = MA_BACKEND_NOT_ENABLED; + #else + result = MA_NO_BACKEND; + #endif + } + } + + /* If this iteration was successful, return. */ + if (result == MA_SUCCESS) { + result = ma_mutex_init(&pContext->deviceEnumLock); + if (result != MA_SUCCESS) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_WARNING, "Failed to initialize mutex for device enumeration. ma_context_get_devices() is not thread safe.\n"); + } + + result = ma_mutex_init(&pContext->deviceInfoLock); + if (result != MA_SUCCESS) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_WARNING, "Failed to initialize mutex for device info retrieval. ma_context_get_device_info() is not thread safe.\n"); + } + + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "System Architecture:\n"); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " Endian: %s\n", ma_is_little_endian() ? "LE" : "BE"); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " SSE2: %s\n", ma_has_sse2() ? "YES" : "NO"); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " AVX2: %s\n", ma_has_avx2() ? "YES" : "NO"); + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, " NEON: %s\n", ma_has_neon() ? "YES" : "NO"); + + pContext->backend = backend; + return result; + } else { + if (result == MA_BACKEND_NOT_ENABLED) { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "%s backend is disabled.\n", ma_get_backend_name(backend)); + } else { + ma_log_postf(ma_context_get_log(pContext), MA_LOG_LEVEL_DEBUG, "Failed to initialize %s backend.\n", ma_get_backend_name(backend)); + } + } + } + + /* If we get here it means an error occurred. */ + MA_ZERO_OBJECT(pContext); /* Safety. */ + return MA_NO_BACKEND; +} + +MA_API ma_result ma_context_uninit(ma_context* pContext) +{ + if (pContext == NULL) { + return MA_INVALID_ARGS; + } + + if (pContext->callbacks.onContextUninit != NULL) { + pContext->callbacks.onContextUninit(pContext); + } + + ma_mutex_uninit(&pContext->deviceEnumLock); + ma_mutex_uninit(&pContext->deviceInfoLock); + ma_free(pContext->pDeviceInfos, &pContext->allocationCallbacks); + ma_context_uninit_backend_apis(pContext); + + if (pContext->pLog == &pContext->log) { + ma_log_uninit(&pContext->log); + } + + return MA_SUCCESS; +} + +MA_API size_t ma_context_sizeof(void) +{ + return sizeof(ma_context); +} + + +MA_API ma_log* ma_context_get_log(ma_context* pContext) +{ + if (pContext == NULL) { + return NULL; + } + + return pContext->pLog; +} + + +MA_API ma_result ma_context_enumerate_devices(ma_context* pContext, ma_enum_devices_callback_proc callback, void* pUserData) +{ + ma_result result; + + if (pContext == NULL || callback == NULL) { + return MA_INVALID_ARGS; + } + + if (pContext->callbacks.onContextEnumerateDevices == NULL) { + return MA_INVALID_OPERATION; + } + + ma_mutex_lock(&pContext->deviceEnumLock); + { + result = pContext->callbacks.onContextEnumerateDevices(pContext, callback, pUserData); + } + ma_mutex_unlock(&pContext->deviceEnumLock); + + return result; +} + + +static ma_bool32 ma_context_get_devices__enum_callback(ma_context* pContext, ma_device_type deviceType, const ma_device_info* pInfo, void* pUserData) +{ + /* + We need to insert the device info into our main internal buffer. Where it goes depends on the device type. If it's a capture device + it's just appended to the end. If it's a playback device it's inserted just before the first capture device. + */ + + /* + First make sure we have room. Since the number of devices we add to the list is usually relatively small I've decided to use a + simple fixed size increment for buffer expansion. + */ + const ma_uint32 bufferExpansionCount = 2; + const ma_uint32 totalDeviceInfoCount = pContext->playbackDeviceInfoCount + pContext->captureDeviceInfoCount; + + if (totalDeviceInfoCount >= pContext->deviceInfoCapacity) { + ma_uint32 newCapacity = pContext->deviceInfoCapacity + bufferExpansionCount; + ma_device_info* pNewInfos = (ma_device_info*)ma_realloc(pContext->pDeviceInfos, sizeof(*pContext->pDeviceInfos)*newCapacity, &pContext->allocationCallbacks); + if (pNewInfos == NULL) { + return MA_FALSE; /* Out of memory. */ + } + + pContext->pDeviceInfos = pNewInfos; + pContext->deviceInfoCapacity = newCapacity; + } + + if (deviceType == ma_device_type_playback) { + /* Playback. Insert just before the first capture device. */ + + /* The first thing to do is move all of the capture devices down a slot. */ + ma_uint32 iFirstCaptureDevice = pContext->playbackDeviceInfoCount; + size_t iCaptureDevice; + for (iCaptureDevice = totalDeviceInfoCount; iCaptureDevice > iFirstCaptureDevice; --iCaptureDevice) { + pContext->pDeviceInfos[iCaptureDevice] = pContext->pDeviceInfos[iCaptureDevice-1]; + } + + /* Now just insert where the first capture device was before moving it down a slot. */ + pContext->pDeviceInfos[iFirstCaptureDevice] = *pInfo; + pContext->playbackDeviceInfoCount += 1; + } else { + /* Capture. Insert at the end. */ + pContext->pDeviceInfos[totalDeviceInfoCount] = *pInfo; + pContext->captureDeviceInfoCount += 1; + } + + (void)pUserData; + return MA_TRUE; +} + +MA_API ma_result ma_context_get_devices(ma_context* pContext, ma_device_info** ppPlaybackDeviceInfos, ma_uint32* pPlaybackDeviceCount, ma_device_info** ppCaptureDeviceInfos, ma_uint32* pCaptureDeviceCount) +{ + ma_result result; + + /* Safety. */ + if (ppPlaybackDeviceInfos != NULL) *ppPlaybackDeviceInfos = NULL; + if (pPlaybackDeviceCount != NULL) *pPlaybackDeviceCount = 0; + if (ppCaptureDeviceInfos != NULL) *ppCaptureDeviceInfos = NULL; + if (pCaptureDeviceCount != NULL) *pCaptureDeviceCount = 0; + + if (pContext == NULL) { + return MA_INVALID_ARGS; + } + + if (pContext->callbacks.onContextEnumerateDevices == NULL) { + return MA_INVALID_OPERATION; + } + + /* Note that we don't use ma_context_enumerate_devices() here because we want to do locking at a higher level. */ + ma_mutex_lock(&pContext->deviceEnumLock); + { + /* Reset everything first. */ + pContext->playbackDeviceInfoCount = 0; + pContext->captureDeviceInfoCount = 0; + + /* Now enumerate over available devices. */ + result = pContext->callbacks.onContextEnumerateDevices(pContext, ma_context_get_devices__enum_callback, NULL); + if (result == MA_SUCCESS) { + /* Playback devices. */ + if (ppPlaybackDeviceInfos != NULL) { + *ppPlaybackDeviceInfos = pContext->pDeviceInfos; + } + if (pPlaybackDeviceCount != NULL) { + *pPlaybackDeviceCount = pContext->playbackDeviceInfoCount; + } + + /* Capture devices. */ + if (ppCaptureDeviceInfos != NULL) { + *ppCaptureDeviceInfos = pContext->pDeviceInfos; + /* Capture devices come after playback devices. */ + if (pContext->playbackDeviceInfoCount > 0) { + /* Conditional, because NULL+0 is undefined behavior. */ + *ppCaptureDeviceInfos += pContext->playbackDeviceInfoCount; + } + } + if (pCaptureDeviceCount != NULL) { + *pCaptureDeviceCount = pContext->captureDeviceInfoCount; + } + } + } + ma_mutex_unlock(&pContext->deviceEnumLock); + + return result; +} + +MA_API ma_result ma_context_get_device_info(ma_context* pContext, ma_device_type deviceType, const ma_device_id* pDeviceID, ma_device_info* pDeviceInfo) +{ + ma_result result; + ma_device_info deviceInfo; + + /* NOTE: Do not clear pDeviceInfo on entry. The reason is the pDeviceID may actually point to pDeviceInfo->id which will break things. */ + if (pContext == NULL || pDeviceInfo == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(&deviceInfo); + + /* Help the backend out by copying over the device ID if we have one. */ + if (pDeviceID != NULL) { + MA_COPY_MEMORY(&deviceInfo.id, pDeviceID, sizeof(*pDeviceID)); + } + + if (pContext->callbacks.onContextGetDeviceInfo == NULL) { + return MA_INVALID_OPERATION; + } + + ma_mutex_lock(&pContext->deviceInfoLock); + { + result = pContext->callbacks.onContextGetDeviceInfo(pContext, deviceType, pDeviceID, &deviceInfo); + } + ma_mutex_unlock(&pContext->deviceInfoLock); + + *pDeviceInfo = deviceInfo; + return result; +} + +MA_API ma_bool32 ma_context_is_loopback_supported(ma_context* pContext) +{ + if (pContext == NULL) { + return MA_FALSE; + } + + return ma_is_loopback_supported(pContext->backend); +} + + +MA_API ma_device_config ma_device_config_init(ma_device_type deviceType) +{ + ma_device_config config; + MA_ZERO_OBJECT(&config); + config.deviceType = deviceType; + config.resampling = ma_resampler_config_init(ma_format_unknown, 0, 0, 0, ma_resample_algorithm_linear); /* Format/channels/rate don't matter here. */ + + return config; +} + +MA_API ma_result ma_device_init(ma_context* pContext, const ma_device_config* pConfig, ma_device* pDevice) +{ + ma_result result; + ma_device_descriptor descriptorPlayback; + ma_device_descriptor descriptorCapture; + + /* The context can be null, in which case we self-manage it. */ + if (pContext == NULL) { + return ma_device_init_ex(NULL, 0, NULL, pConfig, pDevice); + } + + if (pDevice == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDevice); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + /* Check that we have our callbacks defined. */ + if (pContext->callbacks.onDeviceInit == NULL) { + return MA_INVALID_OPERATION; + } + + /* Basic config validation. */ + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex) { + if (pConfig->capture.channels > MA_MAX_CHANNELS) { + return MA_INVALID_ARGS; + } + + if (!ma__is_channel_map_valid(pConfig->capture.pChannelMap, pConfig->capture.channels)) { + return MA_INVALID_ARGS; + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex || pConfig->deviceType == ma_device_type_loopback) { + if (pConfig->playback.channels > MA_MAX_CHANNELS) { + return MA_INVALID_ARGS; + } + + if (!ma__is_channel_map_valid(pConfig->playback.pChannelMap, pConfig->playback.channels)) { + return MA_INVALID_ARGS; + } + } + + pDevice->pContext = pContext; + + /* Set the user data and log callback ASAP to ensure it is available for the entire initialization process. */ + pDevice->pUserData = pConfig->pUserData; + pDevice->onData = pConfig->dataCallback; + pDevice->onNotification = pConfig->notificationCallback; + pDevice->onStop = pConfig->stopCallback; + + if (pConfig->playback.pDeviceID != NULL) { + MA_COPY_MEMORY(&pDevice->playback.id, pConfig->playback.pDeviceID, sizeof(pDevice->playback.id)); + pDevice->playback.pID = &pDevice->playback.id; + } else { + pDevice->playback.pID = NULL; + } + + if (pConfig->capture.pDeviceID != NULL) { + MA_COPY_MEMORY(&pDevice->capture.id, pConfig->capture.pDeviceID, sizeof(pDevice->capture.id)); + pDevice->capture.pID = &pDevice->capture.id; + } else { + pDevice->capture.pID = NULL; + } + + pDevice->noPreSilencedOutputBuffer = pConfig->noPreSilencedOutputBuffer; + pDevice->noClip = pConfig->noClip; + pDevice->noDisableDenormals = pConfig->noDisableDenormals; + pDevice->noFixedSizedCallback = pConfig->noFixedSizedCallback; + ma_atomic_float_set(&pDevice->masterVolumeFactor, 1); + + pDevice->type = pConfig->deviceType; + pDevice->sampleRate = pConfig->sampleRate; + pDevice->resampling.algorithm = pConfig->resampling.algorithm; + pDevice->resampling.linear.lpfOrder = pConfig->resampling.linear.lpfOrder; + pDevice->resampling.pBackendVTable = pConfig->resampling.pBackendVTable; + pDevice->resampling.pBackendUserData = pConfig->resampling.pBackendUserData; + + pDevice->capture.shareMode = pConfig->capture.shareMode; + pDevice->capture.format = pConfig->capture.format; + pDevice->capture.channels = pConfig->capture.channels; + ma_channel_map_copy_or_default(pDevice->capture.channelMap, ma_countof(pDevice->capture.channelMap), pConfig->capture.pChannelMap, pConfig->capture.channels); + pDevice->capture.channelMixMode = pConfig->capture.channelMixMode; + pDevice->capture.calculateLFEFromSpatialChannels = pConfig->capture.calculateLFEFromSpatialChannels; + + pDevice->playback.shareMode = pConfig->playback.shareMode; + pDevice->playback.format = pConfig->playback.format; + pDevice->playback.channels = pConfig->playback.channels; + ma_channel_map_copy_or_default(pDevice->playback.channelMap, ma_countof(pDevice->playback.channelMap), pConfig->playback.pChannelMap, pConfig->playback.channels); + pDevice->playback.channelMixMode = pConfig->playback.channelMixMode; + pDevice->playback.calculateLFEFromSpatialChannels = pConfig->playback.calculateLFEFromSpatialChannels; + + result = ma_mutex_init(&pDevice->startStopLock); + if (result != MA_SUCCESS) { + return result; + } + + /* + When the device is started, the worker thread is the one that does the actual startup of the backend device. We + use a semaphore to wait for the background thread to finish the work. The same applies for stopping the device. + + Each of these semaphores is released internally by the worker thread when the work is completed. The start + semaphore is also used to wake up the worker thread. + */ + result = ma_event_init(&pDevice->wakeupEvent); + if (result != MA_SUCCESS) { + ma_mutex_uninit(&pDevice->startStopLock); + return result; + } + + result = ma_event_init(&pDevice->startEvent); + if (result != MA_SUCCESS) { + ma_event_uninit(&pDevice->wakeupEvent); + ma_mutex_uninit(&pDevice->startStopLock); + return result; + } + + result = ma_event_init(&pDevice->stopEvent); + if (result != MA_SUCCESS) { + ma_event_uninit(&pDevice->startEvent); + ma_event_uninit(&pDevice->wakeupEvent); + ma_mutex_uninit(&pDevice->startStopLock); + return result; + } + + + MA_ZERO_OBJECT(&descriptorPlayback); + descriptorPlayback.pDeviceID = pConfig->playback.pDeviceID; + descriptorPlayback.shareMode = pConfig->playback.shareMode; + descriptorPlayback.format = pConfig->playback.format; + descriptorPlayback.channels = pConfig->playback.channels; + descriptorPlayback.sampleRate = pConfig->sampleRate; + ma_channel_map_copy_or_default(descriptorPlayback.channelMap, ma_countof(descriptorPlayback.channelMap), pConfig->playback.pChannelMap, pConfig->playback.channels); + descriptorPlayback.periodSizeInFrames = pConfig->periodSizeInFrames; + descriptorPlayback.periodSizeInMilliseconds = pConfig->periodSizeInMilliseconds; + descriptorPlayback.periodCount = pConfig->periods; + + if (descriptorPlayback.periodCount == 0) { + descriptorPlayback.periodCount = MA_DEFAULT_PERIODS; + } + + + MA_ZERO_OBJECT(&descriptorCapture); + descriptorCapture.pDeviceID = pConfig->capture.pDeviceID; + descriptorCapture.shareMode = pConfig->capture.shareMode; + descriptorCapture.format = pConfig->capture.format; + descriptorCapture.channels = pConfig->capture.channels; + descriptorCapture.sampleRate = pConfig->sampleRate; + ma_channel_map_copy_or_default(descriptorCapture.channelMap, ma_countof(descriptorCapture.channelMap), pConfig->capture.pChannelMap, pConfig->capture.channels); + descriptorCapture.periodSizeInFrames = pConfig->periodSizeInFrames; + descriptorCapture.periodSizeInMilliseconds = pConfig->periodSizeInMilliseconds; + descriptorCapture.periodCount = pConfig->periods; + + if (descriptorCapture.periodCount == 0) { + descriptorCapture.periodCount = MA_DEFAULT_PERIODS; + } + + + result = pContext->callbacks.onDeviceInit(pDevice, pConfig, &descriptorPlayback, &descriptorCapture); + if (result != MA_SUCCESS) { + ma_event_uninit(&pDevice->startEvent); + ma_event_uninit(&pDevice->wakeupEvent); + ma_mutex_uninit(&pDevice->startStopLock); + return result; + } + +#if 0 + /* + On output the descriptors will contain the *actual* data format of the device. We need this to know how to convert the data between + the requested format and the internal format. + */ + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex || pConfig->deviceType == ma_device_type_loopback) { + if (!ma_device_descriptor_is_valid(&descriptorCapture)) { + ma_device_uninit(pDevice); + return MA_INVALID_ARGS; + } + + pDevice->capture.internalFormat = descriptorCapture.format; + pDevice->capture.internalChannels = descriptorCapture.channels; + pDevice->capture.internalSampleRate = descriptorCapture.sampleRate; + ma_channel_map_copy(pDevice->capture.internalChannelMap, descriptorCapture.channelMap, descriptorCapture.channels); + pDevice->capture.internalPeriodSizeInFrames = descriptorCapture.periodSizeInFrames; + pDevice->capture.internalPeriods = descriptorCapture.periodCount; + + if (pDevice->capture.internalPeriodSizeInFrames == 0) { + pDevice->capture.internalPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(descriptorCapture.periodSizeInMilliseconds, descriptorCapture.sampleRate); + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + if (!ma_device_descriptor_is_valid(&descriptorPlayback)) { + ma_device_uninit(pDevice); + return MA_INVALID_ARGS; + } + + pDevice->playback.internalFormat = descriptorPlayback.format; + pDevice->playback.internalChannels = descriptorPlayback.channels; + pDevice->playback.internalSampleRate = descriptorPlayback.sampleRate; + ma_channel_map_copy(pDevice->playback.internalChannelMap, descriptorPlayback.channelMap, descriptorPlayback.channels); + pDevice->playback.internalPeriodSizeInFrames = descriptorPlayback.periodSizeInFrames; + pDevice->playback.internalPeriods = descriptorPlayback.periodCount; + + if (pDevice->playback.internalPeriodSizeInFrames == 0) { + pDevice->playback.internalPeriodSizeInFrames = ma_calculate_buffer_size_in_frames_from_milliseconds(descriptorPlayback.periodSizeInMilliseconds, descriptorPlayback.sampleRate); + } + } + + + /* + The name of the device can be retrieved from device info. This may be temporary and replaced with a `ma_device_get_info(pDevice, deviceType)` instead. + For loopback devices, we need to retrieve the name of the playback device. + */ + { + ma_device_info deviceInfo; + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex || pConfig->deviceType == ma_device_type_loopback) { + result = ma_device_get_info(pDevice, (pConfig->deviceType == ma_device_type_loopback) ? ma_device_type_playback : ma_device_type_capture, &deviceInfo); + if (result == MA_SUCCESS) { + ma_strncpy_s(pDevice->capture.name, sizeof(pDevice->capture.name), deviceInfo.name, (size_t)-1); + } else { + /* We failed to retrieve the device info. Fall back to a default name. */ + if (descriptorCapture.pDeviceID == NULL) { + ma_strncpy_s(pDevice->capture.name, sizeof(pDevice->capture.name), MA_DEFAULT_CAPTURE_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDevice->capture.name, sizeof(pDevice->capture.name), "Capture Device", (size_t)-1); + } + } + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + result = ma_device_get_info(pDevice, ma_device_type_playback, &deviceInfo); + if (result == MA_SUCCESS) { + ma_strncpy_s(pDevice->playback.name, sizeof(pDevice->playback.name), deviceInfo.name, (size_t)-1); + } else { + /* We failed to retrieve the device info. Fall back to a default name. */ + if (descriptorPlayback.pDeviceID == NULL) { + ma_strncpy_s(pDevice->playback.name, sizeof(pDevice->playback.name), MA_DEFAULT_PLAYBACK_DEVICE_NAME, (size_t)-1); + } else { + ma_strncpy_s(pDevice->playback.name, sizeof(pDevice->playback.name), "Playback Device", (size_t)-1); + } + } + } + } + + + ma_device__post_init_setup(pDevice, pConfig->deviceType); +#endif + + result = ma_device_post_init(pDevice, pConfig->deviceType, &descriptorPlayback, &descriptorCapture); + if (result != MA_SUCCESS) { + ma_device_uninit(pDevice); + return result; + } + + + /* + If we're using fixed sized callbacks we'll need to make use of an intermediary buffer. Needs to + be done after post_init_setup() because we'll need access to the sample rate. + */ + if (pConfig->noFixedSizedCallback == MA_FALSE) { + /* We're using a fixed sized data callback so we'll need an intermediary buffer. */ + ma_uint32 intermediaryBufferCap = pConfig->periodSizeInFrames; + if (intermediaryBufferCap == 0) { + intermediaryBufferCap = ma_calculate_buffer_size_in_frames_from_milliseconds(pConfig->periodSizeInMilliseconds, pDevice->sampleRate); + } + + if (pConfig->deviceType == ma_device_type_capture || pConfig->deviceType == ma_device_type_duplex || pConfig->deviceType == ma_device_type_loopback) { + ma_uint32 intermediaryBufferSizeInBytes; + + pDevice->capture.intermediaryBufferLen = 0; + pDevice->capture.intermediaryBufferCap = intermediaryBufferCap; + if (pDevice->capture.intermediaryBufferCap == 0) { + pDevice->capture.intermediaryBufferCap = pDevice->capture.internalPeriodSizeInFrames; + } + + intermediaryBufferSizeInBytes = pDevice->capture.intermediaryBufferCap * ma_get_bytes_per_frame(pDevice->capture.format, pDevice->capture.channels); + + pDevice->capture.pIntermediaryBuffer = ma_malloc((size_t)intermediaryBufferSizeInBytes, &pContext->allocationCallbacks); + if (pDevice->capture.pIntermediaryBuffer == NULL) { + ma_device_uninit(pDevice); + return MA_OUT_OF_MEMORY; + } + + /* Silence the buffer for safety. */ + ma_silence_pcm_frames(pDevice->capture.pIntermediaryBuffer, pDevice->capture.intermediaryBufferCap, pDevice->capture.format, pDevice->capture.channels); + pDevice->capture.intermediaryBufferLen = pDevice->capture.intermediaryBufferCap; + } + + if (pConfig->deviceType == ma_device_type_playback || pConfig->deviceType == ma_device_type_duplex) { + ma_uint64 intermediaryBufferSizeInBytes; + + pDevice->playback.intermediaryBufferLen = 0; + if (pConfig->deviceType == ma_device_type_duplex) { + pDevice->playback.intermediaryBufferCap = pDevice->capture.intermediaryBufferCap; /* In duplex mode, make sure the intermediary buffer is always the same size as the capture side. */ + } else { + pDevice->playback.intermediaryBufferCap = intermediaryBufferCap; + if (pDevice->playback.intermediaryBufferCap == 0) { + pDevice->playback.intermediaryBufferCap = pDevice->playback.internalPeriodSizeInFrames; + } + } + + intermediaryBufferSizeInBytes = pDevice->playback.intermediaryBufferCap * ma_get_bytes_per_frame(pDevice->playback.format, pDevice->playback.channels); + + pDevice->playback.pIntermediaryBuffer = ma_malloc((size_t)intermediaryBufferSizeInBytes, &pContext->allocationCallbacks); + if (pDevice->playback.pIntermediaryBuffer == NULL) { + ma_device_uninit(pDevice); + return MA_OUT_OF_MEMORY; + } + + /* Silence the buffer for safety. */ + ma_silence_pcm_frames(pDevice->playback.pIntermediaryBuffer, pDevice->playback.intermediaryBufferCap, pDevice->playback.format, pDevice->playback.channels); + pDevice->playback.intermediaryBufferLen = 0; + } + } else { + /* Not using a fixed sized data callback so no need for an intermediary buffer. */ + } + + + /* Some backends don't require the worker thread. */ + if (!ma_context_is_backend_asynchronous(pContext)) { + /* The worker thread. */ + result = ma_thread_create(&pDevice->thread, pContext->threadPriority, pContext->threadStackSize, ma_worker_thread, pDevice, &pContext->allocationCallbacks); + if (result != MA_SUCCESS) { + ma_device_uninit(pDevice); + return result; + } + + /* Wait for the worker thread to put the device into its stopped state for real. */ + ma_event_wait(&pDevice->stopEvent); + MA_ASSERT(ma_device_get_state(pDevice) == ma_device_state_stopped); + } else { + /* + If the backend is asynchronous and the device is duplex, we'll need an intermediary ring buffer. Note that this needs to be done + after ma_device__post_init_setup(). + */ + if (ma_context_is_backend_asynchronous(pContext)) { + if (pConfig->deviceType == ma_device_type_duplex) { + result = ma_duplex_rb_init(pDevice->capture.format, pDevice->capture.channels, pDevice->sampleRate, pDevice->capture.internalSampleRate, pDevice->capture.internalPeriodSizeInFrames, &pDevice->pContext->allocationCallbacks, &pDevice->duplexRB); + if (result != MA_SUCCESS) { + ma_device_uninit(pDevice); + return result; + } + } + } + + ma_device__set_state(pDevice, ma_device_state_stopped); + } + + /* Log device information. */ + { + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, "[%s]\n", ma_get_backend_name(pDevice->pContext->backend)); + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex || pDevice->type == ma_device_type_loopback) { + char name[MA_MAX_DEVICE_NAME_LENGTH + 1]; + ma_device_get_name(pDevice, ma_device_type_capture, name, sizeof(name), NULL); + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " %s (%s)\n", name, "Capture"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Format: %s -> %s\n", ma_get_format_name(pDevice->capture.internalFormat), ma_get_format_name(pDevice->capture.format)); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Channels: %d -> %d\n", pDevice->capture.internalChannels, pDevice->capture.channels); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Sample Rate: %d -> %d\n", pDevice->capture.internalSampleRate, pDevice->sampleRate); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Buffer Size: %d*%d (%d)\n", pDevice->capture.internalPeriodSizeInFrames, pDevice->capture.internalPeriods, (pDevice->capture.internalPeriodSizeInFrames * pDevice->capture.internalPeriods)); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Conversion:\n"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Pre Format Conversion: %s\n", pDevice->capture.converter.hasPreFormatConversion ? "YES" : "NO"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Post Format Conversion: %s\n", pDevice->capture.converter.hasPostFormatConversion ? "YES" : "NO"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Channel Routing: %s\n", pDevice->capture.converter.hasChannelConverter ? "YES" : "NO"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Resampling: %s\n", pDevice->capture.converter.hasResampler ? "YES" : "NO"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Passthrough: %s\n", pDevice->capture.converter.isPassthrough ? "YES" : "NO"); + { + char channelMapStr[1024]; + ma_channel_map_to_string(pDevice->capture.internalChannelMap, pDevice->capture.internalChannels, channelMapStr, sizeof(channelMapStr)); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Channel Map In: {%s}\n", channelMapStr); + + ma_channel_map_to_string(pDevice->capture.channelMap, pDevice->capture.channels, channelMapStr, sizeof(channelMapStr)); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Channel Map Out: {%s}\n", channelMapStr); + } + } + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + char name[MA_MAX_DEVICE_NAME_LENGTH + 1]; + ma_device_get_name(pDevice, ma_device_type_playback, name, sizeof(name), NULL); + + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " %s (%s)\n", name, "Playback"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Format: %s -> %s\n", ma_get_format_name(pDevice->playback.format), ma_get_format_name(pDevice->playback.internalFormat)); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Channels: %d -> %d\n", pDevice->playback.channels, pDevice->playback.internalChannels); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Sample Rate: %d -> %d\n", pDevice->sampleRate, pDevice->playback.internalSampleRate); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Buffer Size: %d*%d (%d)\n", pDevice->playback.internalPeriodSizeInFrames, pDevice->playback.internalPeriods, (pDevice->playback.internalPeriodSizeInFrames * pDevice->playback.internalPeriods)); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Conversion:\n"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Pre Format Conversion: %s\n", pDevice->playback.converter.hasPreFormatConversion ? "YES" : "NO"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Post Format Conversion: %s\n", pDevice->playback.converter.hasPostFormatConversion ? "YES" : "NO"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Channel Routing: %s\n", pDevice->playback.converter.hasChannelConverter ? "YES" : "NO"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Resampling: %s\n", pDevice->playback.converter.hasResampler ? "YES" : "NO"); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Passthrough: %s\n", pDevice->playback.converter.isPassthrough ? "YES" : "NO"); + { + char channelMapStr[1024]; + ma_channel_map_to_string(pDevice->playback.channelMap, pDevice->playback.channels, channelMapStr, sizeof(channelMapStr)); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Channel Map In: {%s}\n", channelMapStr); + + ma_channel_map_to_string(pDevice->playback.internalChannelMap, pDevice->playback.internalChannels, channelMapStr, sizeof(channelMapStr)); + ma_log_postf(ma_device_get_log(pDevice), MA_LOG_LEVEL_INFO, " Channel Map Out: {%s}\n", channelMapStr); + } + } + } + + MA_ASSERT(ma_device_get_state(pDevice) == ma_device_state_stopped); + return MA_SUCCESS; +} + +MA_API ma_result ma_device_init_ex(const ma_backend backends[], ma_uint32 backendCount, const ma_context_config* pContextConfig, const ma_device_config* pConfig, ma_device* pDevice) +{ + ma_result result; + ma_context* pContext; + ma_backend defaultBackends[ma_backend_null+1]; + ma_uint32 iBackend; + ma_backend* pBackendsToIterate; + ma_uint32 backendsToIterateCount; + ma_allocation_callbacks allocationCallbacks; + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pContextConfig != NULL) { + result = ma_allocation_callbacks_init_copy(&allocationCallbacks, &pContextConfig->allocationCallbacks); + if (result != MA_SUCCESS) { + return result; + } + } else { + allocationCallbacks = ma_allocation_callbacks_init_default(); + } + + pContext = (ma_context*)ma_malloc(sizeof(*pContext), &allocationCallbacks); + if (pContext == NULL) { + return MA_OUT_OF_MEMORY; + } + + for (iBackend = 0; iBackend <= ma_backend_null; ++iBackend) { + defaultBackends[iBackend] = (ma_backend)iBackend; + } + + pBackendsToIterate = (ma_backend*)backends; + backendsToIterateCount = backendCount; + if (pBackendsToIterate == NULL) { + pBackendsToIterate = (ma_backend*)defaultBackends; + backendsToIterateCount = ma_countof(defaultBackends); + } + + result = MA_NO_BACKEND; + + for (iBackend = 0; iBackend < backendsToIterateCount; ++iBackend) { + /* + This is a hack for iOS. If the context config is null, there's a good chance the + `ma_device_init(NULL, &deviceConfig, pDevice);` pattern is being used. In this + case, set the session category based on the device type. + */ + #if defined(MA_APPLE_MOBILE) + ma_context_config contextConfig; + + if (pContextConfig == NULL) { + contextConfig = ma_context_config_init(); + switch (pConfig->deviceType) { + case ma_device_type_duplex: { + contextConfig.coreaudio.sessionCategory = ma_ios_session_category_play_and_record; + } break; + case ma_device_type_capture: { + contextConfig.coreaudio.sessionCategory = ma_ios_session_category_record; + } break; + case ma_device_type_playback: + default: { + contextConfig.coreaudio.sessionCategory = ma_ios_session_category_playback; + } break; + } + + pContextConfig = &contextConfig; + } + #endif + + result = ma_context_init(&pBackendsToIterate[iBackend], 1, pContextConfig, pContext); + if (result == MA_SUCCESS) { + result = ma_device_init(pContext, pConfig, pDevice); + if (result == MA_SUCCESS) { + break; /* Success. */ + } else { + ma_context_uninit(pContext); /* Failure. */ + } + } + } + + if (result != MA_SUCCESS) { + ma_free(pContext, &allocationCallbacks); + return result; + } + + pDevice->isOwnerOfContext = MA_TRUE; + return result; +} + +MA_API void ma_device_uninit(ma_device* pDevice) +{ + if (!ma_device__is_initialized(pDevice)) { + return; + } + + /* + It's possible for the miniaudio side of the device and the backend to not be in sync due to + system-level situations such as the computer being put into sleep mode and the backend not + notifying miniaudio of the fact the device has stopped. It's possible for this to result in a + deadlock due to miniaudio thinking the device is in a running state, when in fact it's not + running at all. For this reason I am no longer explicitly stopping the device. I don't think + this should affect anyone in practice since uninitializing the backend will naturally stop the + device anyway. + */ + #if 0 + { + /* Make sure the device is stopped first. The backends will probably handle this naturally, but I like to do it explicitly for my own sanity. */ + if (ma_device_is_started(pDevice)) { + ma_device_stop(pDevice); + } + } + #endif + + /* Putting the device into an uninitialized state will make the worker thread return. */ + ma_device__set_state(pDevice, ma_device_state_uninitialized); + + /* Wake up the worker thread and wait for it to properly terminate. */ + if (!ma_context_is_backend_asynchronous(pDevice->pContext)) { + ma_event_signal(&pDevice->wakeupEvent); + ma_thread_wait(&pDevice->thread); + } + + if (pDevice->pContext->callbacks.onDeviceUninit != NULL) { + pDevice->pContext->callbacks.onDeviceUninit(pDevice); + } + + + ma_event_uninit(&pDevice->stopEvent); + ma_event_uninit(&pDevice->startEvent); + ma_event_uninit(&pDevice->wakeupEvent); + ma_mutex_uninit(&pDevice->startStopLock); + + if (ma_context_is_backend_asynchronous(pDevice->pContext)) { + if (pDevice->type == ma_device_type_duplex) { + ma_duplex_rb_uninit(&pDevice->duplexRB); + } + } + + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_duplex || pDevice->type == ma_device_type_loopback) { + ma_data_converter_uninit(&pDevice->capture.converter, &pDevice->pContext->allocationCallbacks); + } + if (pDevice->type == ma_device_type_playback || pDevice->type == ma_device_type_duplex) { + ma_data_converter_uninit(&pDevice->playback.converter, &pDevice->pContext->allocationCallbacks); + } + + if (pDevice->playback.pInputCache != NULL) { + ma_free(pDevice->playback.pInputCache, &pDevice->pContext->allocationCallbacks); + } + + if (pDevice->capture.pIntermediaryBuffer != NULL) { + ma_free(pDevice->capture.pIntermediaryBuffer, &pDevice->pContext->allocationCallbacks); + } + if (pDevice->playback.pIntermediaryBuffer != NULL) { + ma_free(pDevice->playback.pIntermediaryBuffer, &pDevice->pContext->allocationCallbacks); + } + + if (pDevice->isOwnerOfContext) { + ma_allocation_callbacks allocationCallbacks = pDevice->pContext->allocationCallbacks; + + ma_context_uninit(pDevice->pContext); + ma_free(pDevice->pContext, &allocationCallbacks); + } + + MA_ZERO_OBJECT(pDevice); +} + +MA_API ma_context* ma_device_get_context(ma_device* pDevice) +{ + if (pDevice == NULL) { + return NULL; + } + + return pDevice->pContext; +} + +MA_API ma_log* ma_device_get_log(ma_device* pDevice) +{ + return ma_context_get_log(ma_device_get_context(pDevice)); +} + +MA_API ma_result ma_device_get_info(ma_device* pDevice, ma_device_type type, ma_device_info* pDeviceInfo) +{ + if (pDeviceInfo == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDeviceInfo); + + if (pDevice == NULL) { + return MA_INVALID_ARGS; + } + + /* If the onDeviceGetInfo() callback is set, use that. Otherwise we'll fall back to ma_context_get_device_info(). */ + if (pDevice->pContext->callbacks.onDeviceGetInfo != NULL) { + return pDevice->pContext->callbacks.onDeviceGetInfo(pDevice, type, pDeviceInfo); + } + + /* Getting here means onDeviceGetInfo is not implemented so we need to fall back to an alternative. */ + if (type == ma_device_type_playback) { + return ma_context_get_device_info(pDevice->pContext, type, pDevice->playback.pID, pDeviceInfo); + } else { + /* + Here we're getting the capture side, which is the branch we'll be entering for a loopback + device, since loopback is capturing. However, if the device is using the default device ID, + it won't get the correct information because it'll think we're asking for the default + capture device, where in fact for loopback we want the default *playback* device. We'll do + a bit of a hack here to make sure we get the correct info. + */ + if (pDevice->type == ma_device_type_loopback && pDevice->capture.pID == NULL) { + type = ma_device_type_playback; + } + + return ma_context_get_device_info(pDevice->pContext, type, pDevice->capture.pID, pDeviceInfo); + } +} + +MA_API ma_result ma_device_get_name(ma_device* pDevice, ma_device_type type, char* pName, size_t nameCap, size_t* pLengthNotIncludingNullTerminator) +{ + ma_result result; + ma_device_info deviceInfo; + + if (pLengthNotIncludingNullTerminator != NULL) { + *pLengthNotIncludingNullTerminator = 0; + } + + if (pName != NULL && nameCap > 0) { + pName[0] = '\0'; + } + + result = ma_device_get_info(pDevice, type, &deviceInfo); + if (result != MA_SUCCESS) { + return result; + } + + if (pName != NULL) { + ma_strncpy_s(pName, nameCap, deviceInfo.name, (size_t)-1); + + /* + For safety, make sure the length is based on the truncated output string rather than the + source. Otherwise the caller might assume the output buffer contains more content than it + actually does. + */ + if (pLengthNotIncludingNullTerminator != NULL) { + *pLengthNotIncludingNullTerminator = strlen(pName); + } + } else { + /* Name not specified. Just report the length of the source string. */ + if (pLengthNotIncludingNullTerminator != NULL) { + *pLengthNotIncludingNullTerminator = strlen(deviceInfo.name); + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_device_start(ma_device* pDevice) +{ + ma_result result; + + if (pDevice == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_device_get_state(pDevice) == ma_device_state_uninitialized) { + return MA_INVALID_OPERATION; /* Not initialized. */ + } + + if (ma_device_get_state(pDevice) == ma_device_state_started) { + return MA_SUCCESS; /* Already started. */ + } + + ma_mutex_lock(&pDevice->startStopLock); + { + /* + We need to check again if the device is in a started state because it's possible for one thread to have started the device + while another was waiting on the mutex. + */ + if (ma_device_get_state(pDevice) == ma_device_state_started) { + ma_mutex_unlock(&pDevice->startStopLock); + return MA_SUCCESS; /* Already started. */ + } + + /* Starting and stopping are wrapped in a mutex which means we can assert that the device is in a stopped or paused state. */ + MA_ASSERT(ma_device_get_state(pDevice) == ma_device_state_stopped); + + ma_device__set_state(pDevice, ma_device_state_starting); + + /* Asynchronous backends need to be handled differently. */ + if (ma_context_is_backend_asynchronous(pDevice->pContext)) { + if (pDevice->pContext->callbacks.onDeviceStart != NULL) { + result = pDevice->pContext->callbacks.onDeviceStart(pDevice); + } else { + result = MA_INVALID_OPERATION; + } + + if (result == MA_SUCCESS) { + ma_device__set_state(pDevice, ma_device_state_started); + ma_device__on_notification_started(pDevice); + } + } else { + /* + Synchronous backends are started by signaling an event that's being waited on in the worker thread. We first wake up the + thread and then wait for the start event. + */ + ma_event_signal(&pDevice->wakeupEvent); + + /* + Wait for the worker thread to finish starting the device. Note that the worker thread will be the one who puts the device + into the started state. Don't call ma_device__set_state() here. + */ + ma_event_wait(&pDevice->startEvent); + result = pDevice->workResult; + } + + /* We changed the state from stopped to started, so if we failed, make sure we put the state back to stopped. */ + if (result != MA_SUCCESS) { + ma_device__set_state(pDevice, ma_device_state_stopped); + } + } + ma_mutex_unlock(&pDevice->startStopLock); + + return result; +} + +MA_API ma_result ma_device_stop(ma_device* pDevice) +{ + ma_result result; + + if (pDevice == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_device_get_state(pDevice) == ma_device_state_uninitialized) { + return MA_INVALID_OPERATION; /* Not initialized. */ + } + + if (ma_device_get_state(pDevice) == ma_device_state_stopped) { + return MA_SUCCESS; /* Already stopped. */ + } + + ma_mutex_lock(&pDevice->startStopLock); + { + /* + We need to check again if the device is in a stopped state because it's possible for one thread to have stopped the device + while another was waiting on the mutex. + */ + if (ma_device_get_state(pDevice) == ma_device_state_stopped) { + ma_mutex_unlock(&pDevice->startStopLock); + return MA_SUCCESS; /* Already stopped. */ + } + + /* Starting and stopping are wrapped in a mutex which means we can assert that the device is in a started or paused state. */ + MA_ASSERT(ma_device_get_state(pDevice) == ma_device_state_started); + + ma_device__set_state(pDevice, ma_device_state_stopping); + + /* Asynchronous backends need to be handled differently. */ + if (ma_context_is_backend_asynchronous(pDevice->pContext)) { + /* Asynchronous backends must have a stop operation. */ + if (pDevice->pContext->callbacks.onDeviceStop != NULL) { + result = pDevice->pContext->callbacks.onDeviceStop(pDevice); + } else { + result = MA_INVALID_OPERATION; + } + + ma_device__set_state(pDevice, ma_device_state_stopped); + } else { + /* + Synchronous backends. The stop callback is always called from the worker thread. Do not call the stop callback here. If + the backend is implementing its own audio thread loop we'll need to wake it up if required. Note that we need to make + sure the state of the device is *not* playing right now, which it shouldn't be since we set it above. This is super + important though, so I'm asserting it here as well for extra safety in case we accidentally change something later. + */ + MA_ASSERT(ma_device_get_state(pDevice) != ma_device_state_started); + + if (pDevice->pContext->callbacks.onDeviceDataLoopWakeup != NULL) { + pDevice->pContext->callbacks.onDeviceDataLoopWakeup(pDevice); + } + + /* + We need to wait for the worker thread to become available for work before returning. Note that the worker thread will be + the one who puts the device into the stopped state. Don't call ma_device__set_state() here. + */ + ma_event_wait(&pDevice->stopEvent); + result = MA_SUCCESS; + } + + /* + This is a safety measure to ensure the internal buffer has been cleared so any leftover + does not get played the next time the device starts. Ideally this should be drained by + the backend first. + */ + pDevice->playback.intermediaryBufferLen = 0; + pDevice->playback.inputCacheConsumed = 0; + pDevice->playback.inputCacheRemaining = 0; + } + ma_mutex_unlock(&pDevice->startStopLock); + + return result; +} + +MA_API ma_bool32 ma_device_is_started(const ma_device* pDevice) +{ + return ma_device_get_state(pDevice) == ma_device_state_started; +} + +MA_API ma_device_state ma_device_get_state(const ma_device* pDevice) +{ + if (pDevice == NULL) { + return ma_device_state_uninitialized; + } + + return ma_atomic_device_state_get((ma_atomic_device_state*)&pDevice->state); /* Naughty cast to get rid of a const warning. */ +} + +MA_API ma_result ma_device_set_master_volume(ma_device* pDevice, float volume) +{ + if (pDevice == NULL) { + return MA_INVALID_ARGS; + } + + if (volume < 0.0f) { + return MA_INVALID_ARGS; + } + + ma_atomic_float_set(&pDevice->masterVolumeFactor, volume); + + return MA_SUCCESS; +} + +MA_API ma_result ma_device_get_master_volume(ma_device* pDevice, float* pVolume) +{ + if (pVolume == NULL) { + return MA_INVALID_ARGS; + } + + if (pDevice == NULL) { + *pVolume = 0; + return MA_INVALID_ARGS; + } + + *pVolume = ma_atomic_float_get(&pDevice->masterVolumeFactor); + + return MA_SUCCESS; +} + +MA_API ma_result ma_device_set_master_volume_db(ma_device* pDevice, float gainDB) +{ + if (gainDB > 0) { + return MA_INVALID_ARGS; + } + + return ma_device_set_master_volume(pDevice, ma_volume_db_to_linear(gainDB)); +} + +MA_API ma_result ma_device_get_master_volume_db(ma_device* pDevice, float* pGainDB) +{ + float factor; + ma_result result; + + if (pGainDB == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_device_get_master_volume(pDevice, &factor); + if (result != MA_SUCCESS) { + *pGainDB = 0; + return result; + } + + *pGainDB = ma_volume_linear_to_db(factor); + + return MA_SUCCESS; +} + + +MA_API ma_result ma_device_handle_backend_data_callback(ma_device* pDevice, void* pOutput, const void* pInput, ma_uint32 frameCount) +{ + if (pDevice == NULL) { + return MA_INVALID_ARGS; + } + + if (pOutput == NULL && pInput == NULL) { + return MA_INVALID_ARGS; + } + + /* + There is an assert deeper in the code that checks that frameCount > 0. Since this is a public facing + API we'll need to check for that here. I've had reports that AAudio can sometimes post a frame count + of 0. + */ + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pDevice->type == ma_device_type_duplex) { + if (pInput != NULL) { + ma_device__handle_duplex_callback_capture(pDevice, frameCount, pInput, &pDevice->duplexRB.rb); + } + + if (pOutput != NULL) { + ma_device__handle_duplex_callback_playback(pDevice, frameCount, pOutput, &pDevice->duplexRB.rb); + } + } else { + if (pDevice->type == ma_device_type_capture || pDevice->type == ma_device_type_loopback) { + if (pInput == NULL) { + return MA_INVALID_ARGS; + } + + ma_device__send_frames_to_client(pDevice, frameCount, pInput); + } + + if (pDevice->type == ma_device_type_playback) { + if (pOutput == NULL) { + return MA_INVALID_ARGS; + } + + ma_device__read_frames_from_client(pDevice, frameCount, pOutput); + } + } + + return MA_SUCCESS; +} + +MA_API ma_uint32 ma_calculate_buffer_size_in_frames_from_descriptor(const ma_device_descriptor* pDescriptor, ma_uint32 nativeSampleRate, ma_performance_profile performanceProfile) +{ + if (pDescriptor == NULL) { + return 0; + } + + /* + We must have a non-0 native sample rate, but some backends don't allow retrieval of this at the + time when the size of the buffer needs to be determined. In this case we need to just take a best + guess and move on. We'll try using the sample rate in pDescriptor first. If that's not set we'll + just fall back to MA_DEFAULT_SAMPLE_RATE. + */ + if (nativeSampleRate == 0) { + nativeSampleRate = pDescriptor->sampleRate; + } + if (nativeSampleRate == 0) { + nativeSampleRate = MA_DEFAULT_SAMPLE_RATE; + } + + MA_ASSERT(nativeSampleRate != 0); + + if (pDescriptor->periodSizeInFrames == 0) { + if (pDescriptor->periodSizeInMilliseconds == 0) { + if (performanceProfile == ma_performance_profile_low_latency) { + return ma_calculate_buffer_size_in_frames_from_milliseconds(MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_LOW_LATENCY, nativeSampleRate); + } else { + return ma_calculate_buffer_size_in_frames_from_milliseconds(MA_DEFAULT_PERIOD_SIZE_IN_MILLISECONDS_CONSERVATIVE, nativeSampleRate); + } + } else { + return ma_calculate_buffer_size_in_frames_from_milliseconds(pDescriptor->periodSizeInMilliseconds, nativeSampleRate); + } + } else { + return pDescriptor->periodSizeInFrames; + } +} +#endif /* MA_NO_DEVICE_IO */ + + +MA_API ma_uint32 ma_calculate_buffer_size_in_milliseconds_from_frames(ma_uint32 bufferSizeInFrames, ma_uint32 sampleRate) +{ + /* Prevent a division by zero. */ + if (sampleRate == 0) { + return 0; + } + + return (bufferSizeInFrames*1000 + (sampleRate - 1)) / sampleRate; +} + +MA_API ma_uint32 ma_calculate_buffer_size_in_frames_from_milliseconds(ma_uint32 bufferSizeInMilliseconds, ma_uint32 sampleRate) +{ + /* Prevent a division by zero. */ + if (sampleRate == 0) { + return 0; + } + + return bufferSizeInMilliseconds*sampleRate / 1000; +} + +MA_API void ma_copy_pcm_frames(void* dst, const void* src, ma_uint64 frameCount, ma_format format, ma_uint32 channels) +{ + if (dst == src) { + return; /* No-op. */ + } + + ma_copy_memory_64(dst, src, frameCount * ma_get_bytes_per_frame(format, channels)); +} + +MA_API void ma_silence_pcm_frames(void* p, ma_uint64 frameCount, ma_format format, ma_uint32 channels) +{ + if (format == ma_format_u8) { + ma_uint64 sampleCount = frameCount * channels; + ma_uint64 iSample; + for (iSample = 0; iSample < sampleCount; iSample += 1) { + ((ma_uint8*)p)[iSample] = 128; + } + } else { + ma_zero_memory_64(p, frameCount * ma_get_bytes_per_frame(format, channels)); + } +} + +MA_API void* ma_offset_pcm_frames_ptr(void* p, ma_uint64 offsetInFrames, ma_format format, ma_uint32 channels) +{ + return ma_offset_ptr(p, offsetInFrames * ma_get_bytes_per_frame(format, channels)); +} + +MA_API const void* ma_offset_pcm_frames_const_ptr(const void* p, ma_uint64 offsetInFrames, ma_format format, ma_uint32 channels) +{ + return ma_offset_ptr(p, offsetInFrames * ma_get_bytes_per_frame(format, channels)); +} + + +MA_API void ma_clip_samples_u8(ma_uint8* pDst, const ma_int16* pSrc, ma_uint64 count) +{ + ma_uint64 iSample; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + for (iSample = 0; iSample < count; iSample += 1) { + pDst[iSample] = ma_clip_u8(pSrc[iSample]); + } +} + +MA_API void ma_clip_samples_s16(ma_int16* pDst, const ma_int32* pSrc, ma_uint64 count) +{ + ma_uint64 iSample; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + for (iSample = 0; iSample < count; iSample += 1) { + pDst[iSample] = ma_clip_s16(pSrc[iSample]); + } +} + +MA_API void ma_clip_samples_s24(ma_uint8* pDst, const ma_int64* pSrc, ma_uint64 count) +{ + ma_uint64 iSample; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + for (iSample = 0; iSample < count; iSample += 1) { + ma_int64 s = ma_clip_s24(pSrc[iSample]); + pDst[iSample*3 + 0] = (ma_uint8)((s & 0x000000FF) >> 0); + pDst[iSample*3 + 1] = (ma_uint8)((s & 0x0000FF00) >> 8); + pDst[iSample*3 + 2] = (ma_uint8)((s & 0x00FF0000) >> 16); + } +} + +MA_API void ma_clip_samples_s32(ma_int32* pDst, const ma_int64* pSrc, ma_uint64 count) +{ + ma_uint64 iSample; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + for (iSample = 0; iSample < count; iSample += 1) { + pDst[iSample] = ma_clip_s32(pSrc[iSample]); + } +} + +MA_API void ma_clip_samples_f32(float* pDst, const float* pSrc, ma_uint64 count) +{ + ma_uint64 iSample; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + for (iSample = 0; iSample < count; iSample += 1) { + pDst[iSample] = ma_clip_f32(pSrc[iSample]); + } +} + +MA_API void ma_clip_pcm_frames(void* pDst, const void* pSrc, ma_uint64 frameCount, ma_format format, ma_uint32 channels) +{ + ma_uint64 sampleCount; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + sampleCount = frameCount * channels; + + switch (format) { + case ma_format_u8: ma_clip_samples_u8( (ma_uint8*)pDst, (const ma_int16*)pSrc, sampleCount); break; + case ma_format_s16: ma_clip_samples_s16((ma_int16*)pDst, (const ma_int32*)pSrc, sampleCount); break; + case ma_format_s24: ma_clip_samples_s24((ma_uint8*)pDst, (const ma_int64*)pSrc, sampleCount); break; + case ma_format_s32: ma_clip_samples_s32((ma_int32*)pDst, (const ma_int64*)pSrc, sampleCount); break; + case ma_format_f32: ma_clip_samples_f32(( float*)pDst, (const float*)pSrc, sampleCount); break; + + /* Do nothing if we don't know the format. We're including these here to silence a compiler warning about enums not being handled by the switch. */ + case ma_format_unknown: + case ma_format_count: + break; + } +} + + +MA_API void ma_copy_and_apply_volume_factor_u8(ma_uint8* pSamplesOut, const ma_uint8* pSamplesIn, ma_uint64 sampleCount, float factor) +{ + ma_uint64 iSample; + + if (pSamplesOut == NULL || pSamplesIn == NULL) { + return; + } + + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamplesOut[iSample] = (ma_uint8)(pSamplesIn[iSample] * factor); + } +} + +MA_API void ma_copy_and_apply_volume_factor_s16(ma_int16* pSamplesOut, const ma_int16* pSamplesIn, ma_uint64 sampleCount, float factor) +{ + ma_uint64 iSample; + + if (pSamplesOut == NULL || pSamplesIn == NULL) { + return; + } + + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamplesOut[iSample] = (ma_int16)(pSamplesIn[iSample] * factor); + } +} + +MA_API void ma_copy_and_apply_volume_factor_s24(void* pSamplesOut, const void* pSamplesIn, ma_uint64 sampleCount, float factor) +{ + ma_uint64 iSample; + ma_uint8* pSamplesOut8; + ma_uint8* pSamplesIn8; + + if (pSamplesOut == NULL || pSamplesIn == NULL) { + return; + } + + pSamplesOut8 = (ma_uint8*)pSamplesOut; + pSamplesIn8 = (ma_uint8*)pSamplesIn; + + for (iSample = 0; iSample < sampleCount; iSample += 1) { + ma_int32 sampleS32; + + sampleS32 = (ma_int32)(((ma_uint32)(pSamplesIn8[iSample*3+0]) << 8) | ((ma_uint32)(pSamplesIn8[iSample*3+1]) << 16) | ((ma_uint32)(pSamplesIn8[iSample*3+2])) << 24); + sampleS32 = (ma_int32)(sampleS32 * factor); + + pSamplesOut8[iSample*3+0] = (ma_uint8)(((ma_uint32)sampleS32 & 0x0000FF00) >> 8); + pSamplesOut8[iSample*3+1] = (ma_uint8)(((ma_uint32)sampleS32 & 0x00FF0000) >> 16); + pSamplesOut8[iSample*3+2] = (ma_uint8)(((ma_uint32)sampleS32 & 0xFF000000) >> 24); + } +} + +MA_API void ma_copy_and_apply_volume_factor_s32(ma_int32* pSamplesOut, const ma_int32* pSamplesIn, ma_uint64 sampleCount, float factor) +{ + ma_uint64 iSample; + + if (pSamplesOut == NULL || pSamplesIn == NULL) { + return; + } + + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamplesOut[iSample] = (ma_int32)(pSamplesIn[iSample] * factor); + } +} + +MA_API void ma_copy_and_apply_volume_factor_f32(float* pSamplesOut, const float* pSamplesIn, ma_uint64 sampleCount, float factor) +{ + ma_uint64 iSample; + + if (pSamplesOut == NULL || pSamplesIn == NULL) { + return; + } + + if (factor == 1) { + if (pSamplesOut == pSamplesIn) { + /* In place. No-op. */ + } else { + /* Just a copy. */ + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamplesOut[iSample] = pSamplesIn[iSample]; + } + } + } else { + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamplesOut[iSample] = pSamplesIn[iSample] * factor; + } + } +} + +MA_API void ma_apply_volume_factor_u8(ma_uint8* pSamples, ma_uint64 sampleCount, float factor) +{ + ma_copy_and_apply_volume_factor_u8(pSamples, pSamples, sampleCount, factor); +} + +MA_API void ma_apply_volume_factor_s16(ma_int16* pSamples, ma_uint64 sampleCount, float factor) +{ + ma_copy_and_apply_volume_factor_s16(pSamples, pSamples, sampleCount, factor); +} + +MA_API void ma_apply_volume_factor_s24(void* pSamples, ma_uint64 sampleCount, float factor) +{ + ma_copy_and_apply_volume_factor_s24(pSamples, pSamples, sampleCount, factor); +} + +MA_API void ma_apply_volume_factor_s32(ma_int32* pSamples, ma_uint64 sampleCount, float factor) +{ + ma_copy_and_apply_volume_factor_s32(pSamples, pSamples, sampleCount, factor); +} + +MA_API void ma_apply_volume_factor_f32(float* pSamples, ma_uint64 sampleCount, float factor) +{ + ma_copy_and_apply_volume_factor_f32(pSamples, pSamples, sampleCount, factor); +} + +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_u8(ma_uint8* pFramesOut, const ma_uint8* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_u8(pFramesOut, pFramesIn, frameCount*channels, factor); +} + +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_s16(ma_int16* pFramesOut, const ma_int16* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_s16(pFramesOut, pFramesIn, frameCount*channels, factor); +} + +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_s24(void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_s24(pFramesOut, pFramesIn, frameCount*channels, factor); +} + +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_s32(ma_int32* pFramesOut, const ma_int32* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_s32(pFramesOut, pFramesIn, frameCount*channels, factor); +} + +MA_API void ma_copy_and_apply_volume_factor_pcm_frames_f32(float* pFramesOut, const float* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_f32(pFramesOut, pFramesIn, frameCount*channels, factor); +} + +MA_API void ma_copy_and_apply_volume_factor_pcm_frames(void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount, ma_format format, ma_uint32 channels, float factor) +{ + switch (format) + { + case ma_format_u8: ma_copy_and_apply_volume_factor_pcm_frames_u8 ((ma_uint8*)pFramesOut, (const ma_uint8*)pFramesIn, frameCount, channels, factor); return; + case ma_format_s16: ma_copy_and_apply_volume_factor_pcm_frames_s16((ma_int16*)pFramesOut, (const ma_int16*)pFramesIn, frameCount, channels, factor); return; + case ma_format_s24: ma_copy_and_apply_volume_factor_pcm_frames_s24( pFramesOut, pFramesIn, frameCount, channels, factor); return; + case ma_format_s32: ma_copy_and_apply_volume_factor_pcm_frames_s32((ma_int32*)pFramesOut, (const ma_int32*)pFramesIn, frameCount, channels, factor); return; + case ma_format_f32: ma_copy_and_apply_volume_factor_pcm_frames_f32( (float*)pFramesOut, (const float*)pFramesIn, frameCount, channels, factor); return; + default: return; /* Do nothing. */ + } +} + +MA_API void ma_apply_volume_factor_pcm_frames_u8(ma_uint8* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_pcm_frames_u8(pFrames, pFrames, frameCount, channels, factor); +} + +MA_API void ma_apply_volume_factor_pcm_frames_s16(ma_int16* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_pcm_frames_s16(pFrames, pFrames, frameCount, channels, factor); +} + +MA_API void ma_apply_volume_factor_pcm_frames_s24(void* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_pcm_frames_s24(pFrames, pFrames, frameCount, channels, factor); +} + +MA_API void ma_apply_volume_factor_pcm_frames_s32(ma_int32* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_pcm_frames_s32(pFrames, pFrames, frameCount, channels, factor); +} + +MA_API void ma_apply_volume_factor_pcm_frames_f32(float* pFrames, ma_uint64 frameCount, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_pcm_frames_f32(pFrames, pFrames, frameCount, channels, factor); +} + +MA_API void ma_apply_volume_factor_pcm_frames(void* pFramesOut, ma_uint64 frameCount, ma_format format, ma_uint32 channels, float factor) +{ + ma_copy_and_apply_volume_factor_pcm_frames(pFramesOut, pFramesOut, frameCount, format, channels, factor); +} + + +MA_API void ma_copy_and_apply_volume_factor_per_channel_f32(float* pFramesOut, const float* pFramesIn, ma_uint64 frameCount, ma_uint32 channels, float* pChannelGains) +{ + ma_uint64 iFrame; + + if (channels == 2) { + /* TODO: Do an optimized implementation for stereo and mono. Can do a SIMD optimized implementation as well. */ + } + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOut[iFrame * channels + iChannel] = pFramesIn[iFrame * channels + iChannel] * pChannelGains[iChannel]; + } + } +} + + + +static MA_INLINE ma_int16 ma_apply_volume_unclipped_u8(ma_int16 x, ma_int16 volume) +{ + return (ma_int16)(((ma_int32)x * (ma_int32)volume) >> 8); +} + +static MA_INLINE ma_int32 ma_apply_volume_unclipped_s16(ma_int32 x, ma_int16 volume) +{ + return (ma_int32)((x * volume) >> 8); +} + +static MA_INLINE ma_int64 ma_apply_volume_unclipped_s24(ma_int64 x, ma_int16 volume) +{ + return (ma_int64)((x * volume) >> 8); +} + +static MA_INLINE ma_int64 ma_apply_volume_unclipped_s32(ma_int64 x, ma_int16 volume) +{ + return (ma_int64)((x * volume) >> 8); +} + +static MA_INLINE float ma_apply_volume_unclipped_f32(float x, float volume) +{ + return x * volume; +} + + +MA_API void ma_copy_and_apply_volume_and_clip_samples_u8(ma_uint8* pDst, const ma_int16* pSrc, ma_uint64 count, float volume) +{ + ma_uint64 iSample; + ma_int16 volumeFixed; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + volumeFixed = ma_float_to_fixed_16(volume); + + for (iSample = 0; iSample < count; iSample += 1) { + pDst[iSample] = ma_clip_u8(ma_apply_volume_unclipped_u8(pSrc[iSample], volumeFixed)); + } +} + +MA_API void ma_copy_and_apply_volume_and_clip_samples_s16(ma_int16* pDst, const ma_int32* pSrc, ma_uint64 count, float volume) +{ + ma_uint64 iSample; + ma_int16 volumeFixed; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + volumeFixed = ma_float_to_fixed_16(volume); + + for (iSample = 0; iSample < count; iSample += 1) { + pDst[iSample] = ma_clip_s16(ma_apply_volume_unclipped_s16(pSrc[iSample], volumeFixed)); + } +} + +MA_API void ma_copy_and_apply_volume_and_clip_samples_s24(ma_uint8* pDst, const ma_int64* pSrc, ma_uint64 count, float volume) +{ + ma_uint64 iSample; + ma_int16 volumeFixed; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + volumeFixed = ma_float_to_fixed_16(volume); + + for (iSample = 0; iSample < count; iSample += 1) { + ma_int64 s = ma_clip_s24(ma_apply_volume_unclipped_s24(pSrc[iSample], volumeFixed)); + pDst[iSample*3 + 0] = (ma_uint8)((s & 0x000000FF) >> 0); + pDst[iSample*3 + 1] = (ma_uint8)((s & 0x0000FF00) >> 8); + pDst[iSample*3 + 2] = (ma_uint8)((s & 0x00FF0000) >> 16); + } +} + +MA_API void ma_copy_and_apply_volume_and_clip_samples_s32(ma_int32* pDst, const ma_int64* pSrc, ma_uint64 count, float volume) +{ + ma_uint64 iSample; + ma_int16 volumeFixed; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + volumeFixed = ma_float_to_fixed_16(volume); + + for (iSample = 0; iSample < count; iSample += 1) { + pDst[iSample] = ma_clip_s32(ma_apply_volume_unclipped_s32(pSrc[iSample], volumeFixed)); + } +} + +MA_API void ma_copy_and_apply_volume_and_clip_samples_f32(float* pDst, const float* pSrc, ma_uint64 count, float volume) +{ + ma_uint64 iSample; + + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + /* For the f32 case we need to make sure this supports in-place processing where the input and output buffers are the same. */ + + for (iSample = 0; iSample < count; iSample += 1) { + pDst[iSample] = ma_clip_f32(ma_apply_volume_unclipped_f32(pSrc[iSample], volume)); + } +} + +MA_API void ma_copy_and_apply_volume_and_clip_pcm_frames(void* pDst, const void* pSrc, ma_uint64 frameCount, ma_format format, ma_uint32 channels, float volume) +{ + MA_ASSERT(pDst != NULL); + MA_ASSERT(pSrc != NULL); + + if (volume == 1) { + ma_clip_pcm_frames(pDst, pSrc, frameCount, format, channels); /* Optimized case for volume = 1. */ + } else if (volume == 0) { + ma_silence_pcm_frames(pDst, frameCount, format, channels); /* Optimized case for volume = 0. */ + } else { + ma_uint64 sampleCount = frameCount * channels; + + switch (format) { + case ma_format_u8: ma_copy_and_apply_volume_and_clip_samples_u8( (ma_uint8*)pDst, (const ma_int16*)pSrc, sampleCount, volume); break; + case ma_format_s16: ma_copy_and_apply_volume_and_clip_samples_s16((ma_int16*)pDst, (const ma_int32*)pSrc, sampleCount, volume); break; + case ma_format_s24: ma_copy_and_apply_volume_and_clip_samples_s24((ma_uint8*)pDst, (const ma_int64*)pSrc, sampleCount, volume); break; + case ma_format_s32: ma_copy_and_apply_volume_and_clip_samples_s32((ma_int32*)pDst, (const ma_int64*)pSrc, sampleCount, volume); break; + case ma_format_f32: ma_copy_and_apply_volume_and_clip_samples_f32(( float*)pDst, (const float*)pSrc, sampleCount, volume); break; + + /* Do nothing if we don't know the format. We're including these here to silence a compiler warning about enums not being handled by the switch. */ + case ma_format_unknown: + case ma_format_count: + break; + } + } +} + + + +MA_API float ma_volume_linear_to_db(float factor) +{ + return 20*ma_log10f(factor); +} + +MA_API float ma_volume_db_to_linear(float gain) +{ + return ma_powf(10, gain/20.0f); +} + + +MA_API ma_result ma_mix_pcm_frames_f32(float* pDst, const float* pSrc, ma_uint64 frameCount, ma_uint32 channels, float volume) +{ + ma_uint64 iSample; + ma_uint64 sampleCount; + + if (pDst == NULL || pSrc == NULL || channels == 0) { + return MA_INVALID_ARGS; + } + + if (volume == 0) { + return MA_SUCCESS; /* No changes if the volume is 0. */ + } + + sampleCount = frameCount * channels; + + if (volume == 1) { + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pDst[iSample] += pSrc[iSample]; + } + } else { + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pDst[iSample] += ma_apply_volume_unclipped_f32(pSrc[iSample], volume); + } + } + + return MA_SUCCESS; +} + + + +/************************************************************************************************************************************************************** + +Format Conversion + +**************************************************************************************************************************************************************/ + +static MA_INLINE ma_int16 ma_pcm_sample_f32_to_s16(float x) +{ + return (ma_int16)(x * 32767.0f); +} + +static MA_INLINE ma_int16 ma_pcm_sample_u8_to_s16_no_scale(ma_uint8 x) +{ + return (ma_int16)((ma_int16)x - 128); +} + +static MA_INLINE ma_int64 ma_pcm_sample_s24_to_s32_no_scale(const ma_uint8* x) +{ + return (ma_int64)(((ma_uint64)x[0] << 40) | ((ma_uint64)x[1] << 48) | ((ma_uint64)x[2] << 56)) >> 40; /* Make sure the sign bits are maintained. */ +} + +static MA_INLINE void ma_pcm_sample_s32_to_s24_no_scale(ma_int64 x, ma_uint8* s24) +{ + s24[0] = (ma_uint8)((x & 0x000000FF) >> 0); + s24[1] = (ma_uint8)((x & 0x0000FF00) >> 8); + s24[2] = (ma_uint8)((x & 0x00FF0000) >> 16); +} + + +/* u8 */ +MA_API void ma_pcm_u8_to_u8(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + (void)ditherMode; + ma_copy_memory_64(dst, src, count * sizeof(ma_uint8)); +} + + +static MA_INLINE void ma_pcm_u8_to_s16__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_int16* dst_s16 = (ma_int16*)dst; + const ma_uint8* src_u8 = (const ma_uint8*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int16 x = src_u8[i]; + x = (ma_int16)(x - 128); + x = (ma_int16)(x << 8); + dst_s16[i] = x; + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_u8_to_s16__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s16__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_u8_to_s16__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s16__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_u8_to_s16__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s16__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_u8_to_s16(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_u8_to_s16__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_u8_to_s16__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_u8_to_s16__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_u8_to_s16__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_u8_to_s24__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint8* dst_s24 = (ma_uint8*)dst; + const ma_uint8* src_u8 = (const ma_uint8*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int16 x = src_u8[i]; + x = (ma_int16)(x - 128); + + dst_s24[i*3+0] = 0; + dst_s24[i*3+1] = 0; + dst_s24[i*3+2] = (ma_uint8)((ma_int8)x); + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_u8_to_s24__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s24__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_u8_to_s24__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s24__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_u8_to_s24__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s24__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_u8_to_s24(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_u8_to_s24__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_u8_to_s24__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_u8_to_s24__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_u8_to_s24__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_u8_to_s32__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_int32* dst_s32 = (ma_int32*)dst; + const ma_uint8* src_u8 = (const ma_uint8*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int32 x = src_u8[i]; + x = x - 128; + x = x << 24; + dst_s32[i] = x; + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_u8_to_s32__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s32__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_u8_to_s32__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s32__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_u8_to_s32__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_s32__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_u8_to_s32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_u8_to_s32__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_u8_to_s32__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_u8_to_s32__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_u8_to_s32__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_u8_to_f32__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + float* dst_f32 = (float*)dst; + const ma_uint8* src_u8 = (const ma_uint8*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + float x = (float)src_u8[i]; + x = x * 0.00784313725490196078f; /* 0..255 to 0..2 */ + x = x - 1; /* 0..2 to -1..1 */ + + dst_f32[i] = x; + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_u8_to_f32__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_f32__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_u8_to_f32__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_f32__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_u8_to_f32__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_u8_to_f32__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_u8_to_f32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_u8_to_f32__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_u8_to_f32__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_u8_to_f32__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_u8_to_f32__optimized(dst, src, count, ditherMode); + } +#endif +} + + +#ifdef MA_USE_REFERENCE_CONVERSION_APIS +static MA_INLINE void ma_pcm_interleave_u8__reference(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_uint8* dst_u8 = (ma_uint8*)dst; + const ma_uint8** src_u8 = (const ma_uint8**)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_u8[iFrame*channels + iChannel] = src_u8[iChannel][iFrame]; + } + } +} +#else +static MA_INLINE void ma_pcm_interleave_u8__optimized(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_uint8* dst_u8 = (ma_uint8*)dst; + const ma_uint8** src_u8 = (const ma_uint8**)src; + + if (channels == 1) { + ma_copy_memory_64(dst, src[0], frameCount * sizeof(ma_uint8)); + } else if (channels == 2) { + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + dst_u8[iFrame*2 + 0] = src_u8[0][iFrame]; + dst_u8[iFrame*2 + 1] = src_u8[1][iFrame]; + } + } else { + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_u8[iFrame*channels + iChannel] = src_u8[iChannel][iFrame]; + } + } + } +} +#endif + +MA_API void ma_pcm_interleave_u8(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_interleave_u8__reference(dst, src, frameCount, channels); +#else + ma_pcm_interleave_u8__optimized(dst, src, frameCount, channels); +#endif +} + + +static MA_INLINE void ma_pcm_deinterleave_u8__reference(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_uint8** dst_u8 = (ma_uint8**)dst; + const ma_uint8* src_u8 = (const ma_uint8*)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_u8[iChannel][iFrame] = src_u8[iFrame*channels + iChannel]; + } + } +} + +static MA_INLINE void ma_pcm_deinterleave_u8__optimized(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_deinterleave_u8__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_deinterleave_u8(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_deinterleave_u8__reference(dst, src, frameCount, channels); +#else + ma_pcm_deinterleave_u8__optimized(dst, src, frameCount, channels); +#endif +} + + +/* s16 */ +static MA_INLINE void ma_pcm_s16_to_u8__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint8* dst_u8 = (ma_uint8*)dst; + const ma_int16* src_s16 = (const ma_int16*)src; + + if (ditherMode == ma_dither_mode_none) { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int16 x = src_s16[i]; + x = (ma_int16)(x >> 8); + x = (ma_int16)(x + 128); + dst_u8[i] = (ma_uint8)x; + } + } else { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int16 x = src_s16[i]; + + /* Dither. Don't overflow. */ + ma_int32 dither = ma_dither_s32(ditherMode, -0x80, 0x7F); + if ((x + dither) <= 0x7FFF) { + x = (ma_int16)(x + dither); + } else { + x = 0x7FFF; + } + + x = (ma_int16)(x >> 8); + x = (ma_int16)(x + 128); + dst_u8[i] = (ma_uint8)x; + } + } +} + +static MA_INLINE void ma_pcm_s16_to_u8__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_u8__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s16_to_u8__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_u8__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s16_to_u8__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_u8__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s16_to_u8(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s16_to_u8__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s16_to_u8__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s16_to_u8__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s16_to_u8__optimized(dst, src, count, ditherMode); + } +#endif +} + + +MA_API void ma_pcm_s16_to_s16(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + (void)ditherMode; + ma_copy_memory_64(dst, src, count * sizeof(ma_int16)); +} + + +static MA_INLINE void ma_pcm_s16_to_s24__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint8* dst_s24 = (ma_uint8*)dst; + const ma_int16* src_s16 = (const ma_int16*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + dst_s24[i*3+0] = 0; + dst_s24[i*3+1] = (ma_uint8)(src_s16[i] & 0xFF); + dst_s24[i*3+2] = (ma_uint8)(src_s16[i] >> 8); + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_s16_to_s24__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_s24__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s16_to_s24__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_s24__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s16_to_s24__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_s24__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s16_to_s24(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s16_to_s24__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s16_to_s24__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s16_to_s24__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s16_to_s24__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_s16_to_s32__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_int32* dst_s32 = (ma_int32*)dst; + const ma_int16* src_s16 = (const ma_int16*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + dst_s32[i] = src_s16[i] << 16; + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_s16_to_s32__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_s32__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s16_to_s32__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_s32__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s16_to_s32__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_s32__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s16_to_s32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s16_to_s32__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s16_to_s32__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s16_to_s32__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s16_to_s32__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_s16_to_f32__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + float* dst_f32 = (float*)dst; + const ma_int16* src_s16 = (const ma_int16*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + float x = (float)src_s16[i]; + +#if 0 + /* The accurate way. */ + x = x + 32768.0f; /* -32768..32767 to 0..65535 */ + x = x * 0.00003051804379339284f; /* 0..65535 to 0..2 */ + x = x - 1; /* 0..2 to -1..1 */ +#else + /* The fast way. */ + x = x * 0.000030517578125f; /* -32768..32767 to -1..0.999969482421875 */ +#endif + + dst_f32[i] = x; + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_s16_to_f32__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_f32__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s16_to_f32__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_f32__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s16_to_f32__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s16_to_f32__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s16_to_f32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s16_to_f32__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s16_to_f32__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s16_to_f32__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s16_to_f32__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_interleave_s16__reference(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_int16* dst_s16 = (ma_int16*)dst; + const ma_int16** src_s16 = (const ma_int16**)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_s16[iFrame*channels + iChannel] = src_s16[iChannel][iFrame]; + } + } +} + +static MA_INLINE void ma_pcm_interleave_s16__optimized(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_interleave_s16__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_interleave_s16(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_interleave_s16__reference(dst, src, frameCount, channels); +#else + ma_pcm_interleave_s16__optimized(dst, src, frameCount, channels); +#endif +} + + +static MA_INLINE void ma_pcm_deinterleave_s16__reference(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_int16** dst_s16 = (ma_int16**)dst; + const ma_int16* src_s16 = (const ma_int16*)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_s16[iChannel][iFrame] = src_s16[iFrame*channels + iChannel]; + } + } +} + +static MA_INLINE void ma_pcm_deinterleave_s16__optimized(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_deinterleave_s16__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_deinterleave_s16(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_deinterleave_s16__reference(dst, src, frameCount, channels); +#else + ma_pcm_deinterleave_s16__optimized(dst, src, frameCount, channels); +#endif +} + + +/* s24 */ +static MA_INLINE void ma_pcm_s24_to_u8__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint8* dst_u8 = (ma_uint8*)dst; + const ma_uint8* src_s24 = (const ma_uint8*)src; + + if (ditherMode == ma_dither_mode_none) { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + dst_u8[i] = (ma_uint8)((ma_int8)src_s24[i*3 + 2] + 128); + } + } else { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int32 x = (ma_int32)(((ma_uint32)(src_s24[i*3+0]) << 8) | ((ma_uint32)(src_s24[i*3+1]) << 16) | ((ma_uint32)(src_s24[i*3+2])) << 24); + + /* Dither. Don't overflow. */ + ma_int32 dither = ma_dither_s32(ditherMode, -0x800000, 0x7FFFFF); + if ((ma_int64)x + dither <= 0x7FFFFFFF) { + x = x + dither; + } else { + x = 0x7FFFFFFF; + } + + x = x >> 24; + x = x + 128; + dst_u8[i] = (ma_uint8)x; + } + } +} + +static MA_INLINE void ma_pcm_s24_to_u8__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_u8__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s24_to_u8__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_u8__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s24_to_u8__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_u8__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s24_to_u8(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s24_to_u8__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s24_to_u8__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s24_to_u8__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s24_to_u8__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_s24_to_s16__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_int16* dst_s16 = (ma_int16*)dst; + const ma_uint8* src_s24 = (const ma_uint8*)src; + + if (ditherMode == ma_dither_mode_none) { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_uint16 dst_lo = ((ma_uint16)src_s24[i*3 + 1]); + ma_uint16 dst_hi = (ma_uint16)((ma_uint16)src_s24[i*3 + 2] << 8); + dst_s16[i] = (ma_int16)(dst_lo | dst_hi); + } + } else { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int32 x = (ma_int32)(((ma_uint32)(src_s24[i*3+0]) << 8) | ((ma_uint32)(src_s24[i*3+1]) << 16) | ((ma_uint32)(src_s24[i*3+2])) << 24); + + /* Dither. Don't overflow. */ + ma_int32 dither = ma_dither_s32(ditherMode, -0x8000, 0x7FFF); + if ((ma_int64)x + dither <= 0x7FFFFFFF) { + x = x + dither; + } else { + x = 0x7FFFFFFF; + } + + x = x >> 16; + dst_s16[i] = (ma_int16)x; + } + } +} + +static MA_INLINE void ma_pcm_s24_to_s16__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_s16__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s24_to_s16__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_s16__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s24_to_s16__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_s16__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s24_to_s16(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s24_to_s16__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s24_to_s16__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s24_to_s16__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s24_to_s16__optimized(dst, src, count, ditherMode); + } +#endif +} + + +MA_API void ma_pcm_s24_to_s24(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + (void)ditherMode; + + ma_copy_memory_64(dst, src, count * 3); +} + + +static MA_INLINE void ma_pcm_s24_to_s32__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_int32* dst_s32 = (ma_int32*)dst; + const ma_uint8* src_s24 = (const ma_uint8*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + dst_s32[i] = (ma_int32)(((ma_uint32)(src_s24[i*3+0]) << 8) | ((ma_uint32)(src_s24[i*3+1]) << 16) | ((ma_uint32)(src_s24[i*3+2])) << 24); + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_s24_to_s32__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_s32__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s24_to_s32__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_s32__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s24_to_s32__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_s32__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s24_to_s32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s24_to_s32__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s24_to_s32__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s24_to_s32__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s24_to_s32__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_s24_to_f32__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + float* dst_f32 = (float*)dst; + const ma_uint8* src_s24 = (const ma_uint8*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + float x = (float)(((ma_int32)(((ma_uint32)(src_s24[i*3+0]) << 8) | ((ma_uint32)(src_s24[i*3+1]) << 16) | ((ma_uint32)(src_s24[i*3+2])) << 24)) >> 8); + +#if 0 + /* The accurate way. */ + x = x + 8388608.0f; /* -8388608..8388607 to 0..16777215 */ + x = x * 0.00000011920929665621f; /* 0..16777215 to 0..2 */ + x = x - 1; /* 0..2 to -1..1 */ +#else + /* The fast way. */ + x = x * 0.00000011920928955078125f; /* -8388608..8388607 to -1..0.999969482421875 */ +#endif + + dst_f32[i] = x; + } + + (void)ditherMode; +} + +static MA_INLINE void ma_pcm_s24_to_f32__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_f32__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s24_to_f32__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_f32__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s24_to_f32__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s24_to_f32__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s24_to_f32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s24_to_f32__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s24_to_f32__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s24_to_f32__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s24_to_f32__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_interleave_s24__reference(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_uint8* dst8 = (ma_uint8*)dst; + const ma_uint8** src8 = (const ma_uint8**)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst8[iFrame*3*channels + iChannel*3 + 0] = src8[iChannel][iFrame*3 + 0]; + dst8[iFrame*3*channels + iChannel*3 + 1] = src8[iChannel][iFrame*3 + 1]; + dst8[iFrame*3*channels + iChannel*3 + 2] = src8[iChannel][iFrame*3 + 2]; + } + } +} + +static MA_INLINE void ma_pcm_interleave_s24__optimized(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_interleave_s24__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_interleave_s24(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_interleave_s24__reference(dst, src, frameCount, channels); +#else + ma_pcm_interleave_s24__optimized(dst, src, frameCount, channels); +#endif +} + + +static MA_INLINE void ma_pcm_deinterleave_s24__reference(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_uint8** dst8 = (ma_uint8**)dst; + const ma_uint8* src8 = (const ma_uint8*)src; + + ma_uint32 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst8[iChannel][iFrame*3 + 0] = src8[iFrame*3*channels + iChannel*3 + 0]; + dst8[iChannel][iFrame*3 + 1] = src8[iFrame*3*channels + iChannel*3 + 1]; + dst8[iChannel][iFrame*3 + 2] = src8[iFrame*3*channels + iChannel*3 + 2]; + } + } +} + +static MA_INLINE void ma_pcm_deinterleave_s24__optimized(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_deinterleave_s24__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_deinterleave_s24(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_deinterleave_s24__reference(dst, src, frameCount, channels); +#else + ma_pcm_deinterleave_s24__optimized(dst, src, frameCount, channels); +#endif +} + + + +/* s32 */ +static MA_INLINE void ma_pcm_s32_to_u8__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint8* dst_u8 = (ma_uint8*)dst; + const ma_int32* src_s32 = (const ma_int32*)src; + + if (ditherMode == ma_dither_mode_none) { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int32 x = src_s32[i]; + x = x >> 24; + x = x + 128; + dst_u8[i] = (ma_uint8)x; + } + } else { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int32 x = src_s32[i]; + + /* Dither. Don't overflow. */ + ma_int32 dither = ma_dither_s32(ditherMode, -0x800000, 0x7FFFFF); + if ((ma_int64)x + dither <= 0x7FFFFFFF) { + x = x + dither; + } else { + x = 0x7FFFFFFF; + } + + x = x >> 24; + x = x + 128; + dst_u8[i] = (ma_uint8)x; + } + } +} + +static MA_INLINE void ma_pcm_s32_to_u8__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_u8__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s32_to_u8__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_u8__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s32_to_u8__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_u8__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s32_to_u8(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s32_to_u8__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s32_to_u8__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s32_to_u8__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s32_to_u8__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_s32_to_s16__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_int16* dst_s16 = (ma_int16*)dst; + const ma_int32* src_s32 = (const ma_int32*)src; + + if (ditherMode == ma_dither_mode_none) { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int32 x = src_s32[i]; + x = x >> 16; + dst_s16[i] = (ma_int16)x; + } + } else { + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int32 x = src_s32[i]; + + /* Dither. Don't overflow. */ + ma_int32 dither = ma_dither_s32(ditherMode, -0x8000, 0x7FFF); + if ((ma_int64)x + dither <= 0x7FFFFFFF) { + x = x + dither; + } else { + x = 0x7FFFFFFF; + } + + x = x >> 16; + dst_s16[i] = (ma_int16)x; + } + } +} + +static MA_INLINE void ma_pcm_s32_to_s16__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_s16__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s32_to_s16__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_s16__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s32_to_s16__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_s16__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s32_to_s16(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s32_to_s16__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s32_to_s16__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s32_to_s16__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s32_to_s16__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_s32_to_s24__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint8* dst_s24 = (ma_uint8*)dst; + const ma_int32* src_s32 = (const ma_int32*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_uint32 x = (ma_uint32)src_s32[i]; + dst_s24[i*3+0] = (ma_uint8)((x & 0x0000FF00) >> 8); + dst_s24[i*3+1] = (ma_uint8)((x & 0x00FF0000) >> 16); + dst_s24[i*3+2] = (ma_uint8)((x & 0xFF000000) >> 24); + } + + (void)ditherMode; /* No dithering for s32 -> s24. */ +} + +static MA_INLINE void ma_pcm_s32_to_s24__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_s24__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s32_to_s24__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_s24__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s32_to_s24__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_s24__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s32_to_s24(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s32_to_s24__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s32_to_s24__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s32_to_s24__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s32_to_s24__optimized(dst, src, count, ditherMode); + } +#endif +} + + +MA_API void ma_pcm_s32_to_s32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + (void)ditherMode; + + ma_copy_memory_64(dst, src, count * sizeof(ma_int32)); +} + + +static MA_INLINE void ma_pcm_s32_to_f32__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + float* dst_f32 = (float*)dst; + const ma_int32* src_s32 = (const ma_int32*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + double x = src_s32[i]; + +#if 0 + x = x + 2147483648.0; + x = x * 0.0000000004656612873077392578125; + x = x - 1; +#else + x = x / 2147483648.0; +#endif + + dst_f32[i] = (float)x; + } + + (void)ditherMode; /* No dithering for s32 -> f32. */ +} + +static MA_INLINE void ma_pcm_s32_to_f32__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_f32__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_s32_to_f32__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_f32__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_s32_to_f32__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_s32_to_f32__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_s32_to_f32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_s32_to_f32__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_s32_to_f32__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_s32_to_f32__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_s32_to_f32__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_interleave_s32__reference(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_int32* dst_s32 = (ma_int32*)dst; + const ma_int32** src_s32 = (const ma_int32**)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_s32[iFrame*channels + iChannel] = src_s32[iChannel][iFrame]; + } + } +} + +static MA_INLINE void ma_pcm_interleave_s32__optimized(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_interleave_s32__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_interleave_s32(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_interleave_s32__reference(dst, src, frameCount, channels); +#else + ma_pcm_interleave_s32__optimized(dst, src, frameCount, channels); +#endif +} + + +static MA_INLINE void ma_pcm_deinterleave_s32__reference(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_int32** dst_s32 = (ma_int32**)dst; + const ma_int32* src_s32 = (const ma_int32*)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_s32[iChannel][iFrame] = src_s32[iFrame*channels + iChannel]; + } + } +} + +static MA_INLINE void ma_pcm_deinterleave_s32__optimized(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_deinterleave_s32__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_deinterleave_s32(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_deinterleave_s32__reference(dst, src, frameCount, channels); +#else + ma_pcm_deinterleave_s32__optimized(dst, src, frameCount, channels); +#endif +} + + +/* f32 */ +static MA_INLINE void ma_pcm_f32_to_u8__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint64 i; + + ma_uint8* dst_u8 = (ma_uint8*)dst; + const float* src_f32 = (const float*)src; + + float ditherMin = 0; + float ditherMax = 0; + if (ditherMode != ma_dither_mode_none) { + ditherMin = 1.0f / -128; + ditherMax = 1.0f / 127; + } + + for (i = 0; i < count; i += 1) { + float x = src_f32[i]; + x = x + ma_dither_f32(ditherMode, ditherMin, ditherMax); + x = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); /* clip */ + x = x + 1; /* -1..1 to 0..2 */ + x = x * 127.5f; /* 0..2 to 0..255 */ + + dst_u8[i] = (ma_uint8)x; + } +} + +static MA_INLINE void ma_pcm_f32_to_u8__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_u8__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_f32_to_u8__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_u8__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_f32_to_u8__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_u8__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_f32_to_u8(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_f32_to_u8__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_f32_to_u8__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_f32_to_u8__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_f32_to_u8__optimized(dst, src, count, ditherMode); + } +#endif +} + +#ifdef MA_USE_REFERENCE_CONVERSION_APIS +static MA_INLINE void ma_pcm_f32_to_s16__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint64 i; + + ma_int16* dst_s16 = (ma_int16*)dst; + const float* src_f32 = (const float*)src; + + float ditherMin = 0; + float ditherMax = 0; + if (ditherMode != ma_dither_mode_none) { + ditherMin = 1.0f / -32768; + ditherMax = 1.0f / 32767; + } + + for (i = 0; i < count; i += 1) { + float x = src_f32[i]; + x = x + ma_dither_f32(ditherMode, ditherMin, ditherMax); + x = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); /* clip */ + +#if 0 + /* The accurate way. */ + x = x + 1; /* -1..1 to 0..2 */ + x = x * 32767.5f; /* 0..2 to 0..65535 */ + x = x - 32768.0f; /* 0...65535 to -32768..32767 */ +#else + /* The fast way. */ + x = x * 32767.0f; /* -1..1 to -32767..32767 */ +#endif + + dst_s16[i] = (ma_int16)x; + } +} +#else +static MA_INLINE void ma_pcm_f32_to_s16__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint64 i; + ma_uint64 i4; + ma_uint64 count4; + + ma_int16* dst_s16 = (ma_int16*)dst; + const float* src_f32 = (const float*)src; + + float ditherMin = 0; + float ditherMax = 0; + if (ditherMode != ma_dither_mode_none) { + ditherMin = 1.0f / -32768; + ditherMax = 1.0f / 32767; + } + + /* Unrolled. */ + i = 0; + count4 = count >> 2; + for (i4 = 0; i4 < count4; i4 += 1) { + float d0 = ma_dither_f32(ditherMode, ditherMin, ditherMax); + float d1 = ma_dither_f32(ditherMode, ditherMin, ditherMax); + float d2 = ma_dither_f32(ditherMode, ditherMin, ditherMax); + float d3 = ma_dither_f32(ditherMode, ditherMin, ditherMax); + + float x0 = src_f32[i+0]; + float x1 = src_f32[i+1]; + float x2 = src_f32[i+2]; + float x3 = src_f32[i+3]; + + x0 = x0 + d0; + x1 = x1 + d1; + x2 = x2 + d2; + x3 = x3 + d3; + + x0 = ((x0 < -1) ? -1 : ((x0 > 1) ? 1 : x0)); + x1 = ((x1 < -1) ? -1 : ((x1 > 1) ? 1 : x1)); + x2 = ((x2 < -1) ? -1 : ((x2 > 1) ? 1 : x2)); + x3 = ((x3 < -1) ? -1 : ((x3 > 1) ? 1 : x3)); + + x0 = x0 * 32767.0f; + x1 = x1 * 32767.0f; + x2 = x2 * 32767.0f; + x3 = x3 * 32767.0f; + + dst_s16[i+0] = (ma_int16)x0; + dst_s16[i+1] = (ma_int16)x1; + dst_s16[i+2] = (ma_int16)x2; + dst_s16[i+3] = (ma_int16)x3; + + i += 4; + } + + /* Leftover. */ + for (; i < count; i += 1) { + float x = src_f32[i]; + x = x + ma_dither_f32(ditherMode, ditherMin, ditherMax); + x = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); /* clip */ + x = x * 32767.0f; /* -1..1 to -32767..32767 */ + + dst_s16[i] = (ma_int16)x; + } +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_f32_to_s16__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint64 i; + ma_uint64 i8; + ma_uint64 count8; + ma_int16* dst_s16; + const float* src_f32; + float ditherMin; + float ditherMax; + + /* Both the input and output buffers need to be aligned to 16 bytes. */ + if ((((ma_uintptr)dst & 15) != 0) || (((ma_uintptr)src & 15) != 0)) { + ma_pcm_f32_to_s16__optimized(dst, src, count, ditherMode); + return; + } + + dst_s16 = (ma_int16*)dst; + src_f32 = (const float*)src; + + ditherMin = 0; + ditherMax = 0; + if (ditherMode != ma_dither_mode_none) { + ditherMin = 1.0f / -32768; + ditherMax = 1.0f / 32767; + } + + i = 0; + + /* SSE2. SSE allows us to output 8 s16's at a time which means our loop is unrolled 8 times. */ + count8 = count >> 3; + for (i8 = 0; i8 < count8; i8 += 1) { + __m128 d0; + __m128 d1; + __m128 x0; + __m128 x1; + + if (ditherMode == ma_dither_mode_none) { + d0 = _mm_set1_ps(0); + d1 = _mm_set1_ps(0); + } else if (ditherMode == ma_dither_mode_rectangle) { + d0 = _mm_set_ps( + ma_dither_f32_rectangle(ditherMin, ditherMax), + ma_dither_f32_rectangle(ditherMin, ditherMax), + ma_dither_f32_rectangle(ditherMin, ditherMax), + ma_dither_f32_rectangle(ditherMin, ditherMax) + ); + d1 = _mm_set_ps( + ma_dither_f32_rectangle(ditherMin, ditherMax), + ma_dither_f32_rectangle(ditherMin, ditherMax), + ma_dither_f32_rectangle(ditherMin, ditherMax), + ma_dither_f32_rectangle(ditherMin, ditherMax) + ); + } else { + d0 = _mm_set_ps( + ma_dither_f32_triangle(ditherMin, ditherMax), + ma_dither_f32_triangle(ditherMin, ditherMax), + ma_dither_f32_triangle(ditherMin, ditherMax), + ma_dither_f32_triangle(ditherMin, ditherMax) + ); + d1 = _mm_set_ps( + ma_dither_f32_triangle(ditherMin, ditherMax), + ma_dither_f32_triangle(ditherMin, ditherMax), + ma_dither_f32_triangle(ditherMin, ditherMax), + ma_dither_f32_triangle(ditherMin, ditherMax) + ); + } + + x0 = *((__m128*)(src_f32 + i) + 0); + x1 = *((__m128*)(src_f32 + i) + 1); + + x0 = _mm_add_ps(x0, d0); + x1 = _mm_add_ps(x1, d1); + + x0 = _mm_mul_ps(x0, _mm_set1_ps(32767.0f)); + x1 = _mm_mul_ps(x1, _mm_set1_ps(32767.0f)); + + _mm_stream_si128(((__m128i*)(dst_s16 + i)), _mm_packs_epi32(_mm_cvttps_epi32(x0), _mm_cvttps_epi32(x1))); + + i += 8; + } + + + /* Leftover. */ + for (; i < count; i += 1) { + float x = src_f32[i]; + x = x + ma_dither_f32(ditherMode, ditherMin, ditherMax); + x = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); /* clip */ + x = x * 32767.0f; /* -1..1 to -32767..32767 */ + + dst_s16[i] = (ma_int16)x; + } +} +#endif /* SSE2 */ + +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_f32_to_s16__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint64 i; + ma_uint64 i8; + ma_uint64 count8; + ma_int16* dst_s16; + const float* src_f32; + float ditherMin; + float ditherMax; + + if (!ma_has_neon()) { + ma_pcm_f32_to_s16__optimized(dst, src, count, ditherMode); + return; + } + + /* Both the input and output buffers need to be aligned to 16 bytes. */ + if ((((ma_uintptr)dst & 15) != 0) || (((ma_uintptr)src & 15) != 0)) { + ma_pcm_f32_to_s16__optimized(dst, src, count, ditherMode); + return; + } + + dst_s16 = (ma_int16*)dst; + src_f32 = (const float*)src; + + ditherMin = 0; + ditherMax = 0; + if (ditherMode != ma_dither_mode_none) { + ditherMin = 1.0f / -32768; + ditherMax = 1.0f / 32767; + } + + i = 0; + + /* NEON. NEON allows us to output 8 s16's at a time which means our loop is unrolled 8 times. */ + count8 = count >> 3; + for (i8 = 0; i8 < count8; i8 += 1) { + float32x4_t d0; + float32x4_t d1; + float32x4_t x0; + float32x4_t x1; + int32x4_t i0; + int32x4_t i1; + + if (ditherMode == ma_dither_mode_none) { + d0 = vmovq_n_f32(0); + d1 = vmovq_n_f32(0); + } else if (ditherMode == ma_dither_mode_rectangle) { + float d0v[4]; + float d1v[4]; + + d0v[0] = ma_dither_f32_rectangle(ditherMin, ditherMax); + d0v[1] = ma_dither_f32_rectangle(ditherMin, ditherMax); + d0v[2] = ma_dither_f32_rectangle(ditherMin, ditherMax); + d0v[3] = ma_dither_f32_rectangle(ditherMin, ditherMax); + d0 = vld1q_f32(d0v); + + d1v[0] = ma_dither_f32_rectangle(ditherMin, ditherMax); + d1v[1] = ma_dither_f32_rectangle(ditherMin, ditherMax); + d1v[2] = ma_dither_f32_rectangle(ditherMin, ditherMax); + d1v[3] = ma_dither_f32_rectangle(ditherMin, ditherMax); + d1 = vld1q_f32(d1v); + } else { + float d0v[4]; + float d1v[4]; + + d0v[0] = ma_dither_f32_triangle(ditherMin, ditherMax); + d0v[1] = ma_dither_f32_triangle(ditherMin, ditherMax); + d0v[2] = ma_dither_f32_triangle(ditherMin, ditherMax); + d0v[3] = ma_dither_f32_triangle(ditherMin, ditherMax); + d0 = vld1q_f32(d0v); + + d1v[0] = ma_dither_f32_triangle(ditherMin, ditherMax); + d1v[1] = ma_dither_f32_triangle(ditherMin, ditherMax); + d1v[2] = ma_dither_f32_triangle(ditherMin, ditherMax); + d1v[3] = ma_dither_f32_triangle(ditherMin, ditherMax); + d1 = vld1q_f32(d1v); + } + + x0 = *((float32x4_t*)(src_f32 + i) + 0); + x1 = *((float32x4_t*)(src_f32 + i) + 1); + + x0 = vaddq_f32(x0, d0); + x1 = vaddq_f32(x1, d1); + + x0 = vmulq_n_f32(x0, 32767.0f); + x1 = vmulq_n_f32(x1, 32767.0f); + + i0 = vcvtq_s32_f32(x0); + i1 = vcvtq_s32_f32(x1); + *((int16x8_t*)(dst_s16 + i)) = vcombine_s16(vqmovn_s32(i0), vqmovn_s32(i1)); + + i += 8; + } + + + /* Leftover. */ + for (; i < count; i += 1) { + float x = src_f32[i]; + x = x + ma_dither_f32(ditherMode, ditherMin, ditherMax); + x = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); /* clip */ + x = x * 32767.0f; /* -1..1 to -32767..32767 */ + + dst_s16[i] = (ma_int16)x; + } +} +#endif /* Neon */ +#endif /* MA_USE_REFERENCE_CONVERSION_APIS */ + +MA_API void ma_pcm_f32_to_s16(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_f32_to_s16__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_f32_to_s16__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_f32_to_s16__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_f32_to_s16__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_f32_to_s24__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_uint8* dst_s24 = (ma_uint8*)dst; + const float* src_f32 = (const float*)src; + + ma_uint64 i; + for (i = 0; i < count; i += 1) { + ma_int32 r; + float x = src_f32[i]; + x = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); /* clip */ + +#if 0 + /* The accurate way. */ + x = x + 1; /* -1..1 to 0..2 */ + x = x * 8388607.5f; /* 0..2 to 0..16777215 */ + x = x - 8388608.0f; /* 0..16777215 to -8388608..8388607 */ +#else + /* The fast way. */ + x = x * 8388607.0f; /* -1..1 to -8388607..8388607 */ +#endif + + r = (ma_int32)x; + dst_s24[(i*3)+0] = (ma_uint8)((r & 0x0000FF) >> 0); + dst_s24[(i*3)+1] = (ma_uint8)((r & 0x00FF00) >> 8); + dst_s24[(i*3)+2] = (ma_uint8)((r & 0xFF0000) >> 16); + } + + (void)ditherMode; /* No dithering for f32 -> s24. */ +} + +static MA_INLINE void ma_pcm_f32_to_s24__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_s24__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_f32_to_s24__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_s24__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_f32_to_s24__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_s24__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_f32_to_s24(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_f32_to_s24__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_f32_to_s24__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_f32_to_s24__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_f32_to_s24__optimized(dst, src, count, ditherMode); + } +#endif +} + + +static MA_INLINE void ma_pcm_f32_to_s32__reference(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_int32* dst_s32 = (ma_int32*)dst; + const float* src_f32 = (const float*)src; + + ma_uint32 i; + for (i = 0; i < count; i += 1) { + double x = src_f32[i]; + x = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); /* clip */ + +#if 0 + /* The accurate way. */ + x = x + 1; /* -1..1 to 0..2 */ + x = x * 2147483647.5; /* 0..2 to 0..4294967295 */ + x = x - 2147483648.0; /* 0...4294967295 to -2147483648..2147483647 */ +#else + /* The fast way. */ + x = x * 2147483647.0; /* -1..1 to -2147483647..2147483647 */ +#endif + + dst_s32[i] = (ma_int32)x; + } + + (void)ditherMode; /* No dithering for f32 -> s32. */ +} + +static MA_INLINE void ma_pcm_f32_to_s32__optimized(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_s32__reference(dst, src, count, ditherMode); +} + +#if defined(MA_SUPPORT_SSE2) +static MA_INLINE void ma_pcm_f32_to_s32__sse2(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_s32__optimized(dst, src, count, ditherMode); +} +#endif +#if defined(MA_SUPPORT_NEON) +static MA_INLINE void ma_pcm_f32_to_s32__neon(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + ma_pcm_f32_to_s32__optimized(dst, src, count, ditherMode); +} +#endif + +MA_API void ma_pcm_f32_to_s32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_f32_to_s32__reference(dst, src, count, ditherMode); +#else + # if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_pcm_f32_to_s32__sse2(dst, src, count, ditherMode); + } else + #elif defined(MA_SUPPORT_NEON) + if (ma_has_neon()) { + ma_pcm_f32_to_s32__neon(dst, src, count, ditherMode); + } else + #endif + { + ma_pcm_f32_to_s32__optimized(dst, src, count, ditherMode); + } +#endif +} + + +MA_API void ma_pcm_f32_to_f32(void* dst, const void* src, ma_uint64 count, ma_dither_mode ditherMode) +{ + (void)ditherMode; + + ma_copy_memory_64(dst, src, count * sizeof(float)); +} + + +static void ma_pcm_interleave_f32__reference(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + float* dst_f32 = (float*)dst; + const float** src_f32 = (const float**)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_f32[iFrame*channels + iChannel] = src_f32[iChannel][iFrame]; + } + } +} + +static void ma_pcm_interleave_f32__optimized(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_interleave_f32__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_interleave_f32(void* dst, const void** src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_interleave_f32__reference(dst, src, frameCount, channels); +#else + ma_pcm_interleave_f32__optimized(dst, src, frameCount, channels); +#endif +} + + +static void ma_pcm_deinterleave_f32__reference(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + float** dst_f32 = (float**)dst; + const float* src_f32 = (const float*)src; + + ma_uint64 iFrame; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; iChannel += 1) { + dst_f32[iChannel][iFrame] = src_f32[iFrame*channels + iChannel]; + } + } +} + +static void ma_pcm_deinterleave_f32__optimized(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ + ma_pcm_deinterleave_f32__reference(dst, src, frameCount, channels); +} + +MA_API void ma_pcm_deinterleave_f32(void** dst, const void* src, ma_uint64 frameCount, ma_uint32 channels) +{ +#ifdef MA_USE_REFERENCE_CONVERSION_APIS + ma_pcm_deinterleave_f32__reference(dst, src, frameCount, channels); +#else + ma_pcm_deinterleave_f32__optimized(dst, src, frameCount, channels); +#endif +} + + +MA_API void ma_pcm_convert(void* pOut, ma_format formatOut, const void* pIn, ma_format formatIn, ma_uint64 sampleCount, ma_dither_mode ditherMode) +{ + if (formatOut == formatIn) { + ma_copy_memory_64(pOut, pIn, sampleCount * ma_get_bytes_per_sample(formatOut)); + return; + } + + switch (formatIn) + { + case ma_format_u8: + { + switch (formatOut) + { + case ma_format_s16: ma_pcm_u8_to_s16(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s24: ma_pcm_u8_to_s24(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s32: ma_pcm_u8_to_s32(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_f32: ma_pcm_u8_to_f32(pOut, pIn, sampleCount, ditherMode); return; + default: break; + } + } break; + + case ma_format_s16: + { + switch (formatOut) + { + case ma_format_u8: ma_pcm_s16_to_u8( pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s24: ma_pcm_s16_to_s24(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s32: ma_pcm_s16_to_s32(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_f32: ma_pcm_s16_to_f32(pOut, pIn, sampleCount, ditherMode); return; + default: break; + } + } break; + + case ma_format_s24: + { + switch (formatOut) + { + case ma_format_u8: ma_pcm_s24_to_u8( pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s16: ma_pcm_s24_to_s16(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s32: ma_pcm_s24_to_s32(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_f32: ma_pcm_s24_to_f32(pOut, pIn, sampleCount, ditherMode); return; + default: break; + } + } break; + + case ma_format_s32: + { + switch (formatOut) + { + case ma_format_u8: ma_pcm_s32_to_u8( pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s16: ma_pcm_s32_to_s16(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s24: ma_pcm_s32_to_s24(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_f32: ma_pcm_s32_to_f32(pOut, pIn, sampleCount, ditherMode); return; + default: break; + } + } break; + + case ma_format_f32: + { + switch (formatOut) + { + case ma_format_u8: ma_pcm_f32_to_u8( pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s16: ma_pcm_f32_to_s16(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s24: ma_pcm_f32_to_s24(pOut, pIn, sampleCount, ditherMode); return; + case ma_format_s32: ma_pcm_f32_to_s32(pOut, pIn, sampleCount, ditherMode); return; + default: break; + } + } break; + + default: break; + } +} + +MA_API void ma_convert_pcm_frames_format(void* pOut, ma_format formatOut, const void* pIn, ma_format formatIn, ma_uint64 frameCount, ma_uint32 channels, ma_dither_mode ditherMode) +{ + ma_pcm_convert(pOut, formatOut, pIn, formatIn, frameCount * channels, ditherMode); +} + +MA_API void ma_deinterleave_pcm_frames(ma_format format, ma_uint32 channels, ma_uint64 frameCount, const void* pInterleavedPCMFrames, void** ppDeinterleavedPCMFrames) +{ + if (pInterleavedPCMFrames == NULL || ppDeinterleavedPCMFrames == NULL) { + return; /* Invalid args. */ + } + + /* For efficiency we do this per format. */ + switch (format) { + case ma_format_s16: + { + const ma_int16* pSrcS16 = (const ma_int16*)pInterleavedPCMFrames; + ma_uint64 iPCMFrame; + for (iPCMFrame = 0; iPCMFrame < frameCount; ++iPCMFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; ++iChannel) { + ma_int16* pDstS16 = (ma_int16*)ppDeinterleavedPCMFrames[iChannel]; + pDstS16[iPCMFrame] = pSrcS16[iPCMFrame*channels+iChannel]; + } + } + } break; + + case ma_format_f32: + { + const float* pSrcF32 = (const float*)pInterleavedPCMFrames; + ma_uint64 iPCMFrame; + for (iPCMFrame = 0; iPCMFrame < frameCount; ++iPCMFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; ++iChannel) { + float* pDstF32 = (float*)ppDeinterleavedPCMFrames[iChannel]; + pDstF32[iPCMFrame] = pSrcF32[iPCMFrame*channels+iChannel]; + } + } + } break; + + default: + { + ma_uint32 sampleSizeInBytes = ma_get_bytes_per_sample(format); + ma_uint64 iPCMFrame; + for (iPCMFrame = 0; iPCMFrame < frameCount; ++iPCMFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; ++iChannel) { + void* pDst = ma_offset_ptr(ppDeinterleavedPCMFrames[iChannel], iPCMFrame*sampleSizeInBytes); + const void* pSrc = ma_offset_ptr(pInterleavedPCMFrames, (iPCMFrame*channels+iChannel)*sampleSizeInBytes); + memcpy(pDst, pSrc, sampleSizeInBytes); + } + } + } break; + } +} + +MA_API void ma_interleave_pcm_frames(ma_format format, ma_uint32 channels, ma_uint64 frameCount, const void** ppDeinterleavedPCMFrames, void* pInterleavedPCMFrames) +{ + switch (format) + { + case ma_format_s16: + { + ma_int16* pDstS16 = (ma_int16*)pInterleavedPCMFrames; + ma_uint64 iPCMFrame; + for (iPCMFrame = 0; iPCMFrame < frameCount; ++iPCMFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; ++iChannel) { + const ma_int16* pSrcS16 = (const ma_int16*)ppDeinterleavedPCMFrames[iChannel]; + pDstS16[iPCMFrame*channels+iChannel] = pSrcS16[iPCMFrame]; + } + } + } break; + + case ma_format_f32: + { + float* pDstF32 = (float*)pInterleavedPCMFrames; + ma_uint64 iPCMFrame; + for (iPCMFrame = 0; iPCMFrame < frameCount; ++iPCMFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; ++iChannel) { + const float* pSrcF32 = (const float*)ppDeinterleavedPCMFrames[iChannel]; + pDstF32[iPCMFrame*channels+iChannel] = pSrcF32[iPCMFrame]; + } + } + } break; + + default: + { + ma_uint32 sampleSizeInBytes = ma_get_bytes_per_sample(format); + ma_uint64 iPCMFrame; + for (iPCMFrame = 0; iPCMFrame < frameCount; ++iPCMFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; ++iChannel) { + void* pDst = ma_offset_ptr(pInterleavedPCMFrames, (iPCMFrame*channels+iChannel)*sampleSizeInBytes); + const void* pSrc = ma_offset_ptr(ppDeinterleavedPCMFrames[iChannel], iPCMFrame*sampleSizeInBytes); + memcpy(pDst, pSrc, sampleSizeInBytes); + } + } + } break; + } +} + + +/************************************************************************************************************************************************************** + +Biquad Filter + +**************************************************************************************************************************************************************/ +#ifndef MA_BIQUAD_FIXED_POINT_SHIFT +#define MA_BIQUAD_FIXED_POINT_SHIFT 14 +#endif + +static ma_int32 ma_biquad_float_to_fp(double x) +{ + return (ma_int32)(x * (1 << MA_BIQUAD_FIXED_POINT_SHIFT)); +} + +MA_API ma_biquad_config ma_biquad_config_init(ma_format format, ma_uint32 channels, double b0, double b1, double b2, double a0, double a1, double a2) +{ + ma_biquad_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.b0 = b0; + config.b1 = b1; + config.b2 = b2; + config.a0 = a0; + config.a1 = a1; + config.a2 = a2; + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t r1Offset; + size_t r2Offset; +} ma_biquad_heap_layout; + +static ma_result ma_biquad_get_heap_layout(const ma_biquad_config* pConfig, ma_biquad_heap_layout* pHeapLayout) +{ + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channels == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* R0 */ + pHeapLayout->r1Offset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(ma_biquad_coefficient) * pConfig->channels; + + /* R1 */ + pHeapLayout->r2Offset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(ma_biquad_coefficient) * pConfig->channels; + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +MA_API ma_result ma_biquad_get_heap_size(const ma_biquad_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_biquad_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_biquad_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_biquad_init_preallocated(const ma_biquad_config* pConfig, void* pHeap, ma_biquad* pBQ) +{ + ma_result result; + ma_biquad_heap_layout heapLayout; + + if (pBQ == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pBQ); + + result = ma_biquad_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pBQ->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pBQ->pR1 = (ma_biquad_coefficient*)ma_offset_ptr(pHeap, heapLayout.r1Offset); + pBQ->pR2 = (ma_biquad_coefficient*)ma_offset_ptr(pHeap, heapLayout.r2Offset); + + return ma_biquad_reinit(pConfig, pBQ); +} + +MA_API ma_result ma_biquad_init(const ma_biquad_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_biquad* pBQ) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_biquad_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_biquad_init_preallocated(pConfig, pHeap, pBQ); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pBQ->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_biquad_uninit(ma_biquad* pBQ, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pBQ == NULL) { + return; + } + + if (pBQ->_ownsHeap) { + ma_free(pBQ->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_biquad_reinit(const ma_biquad_config* pConfig, ma_biquad* pBQ) +{ + if (pBQ == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->a0 == 0) { + return MA_INVALID_ARGS; /* Division by zero. */ + } + + /* Only supporting f32 and s16. */ + if (pConfig->format != ma_format_f32 && pConfig->format != ma_format_s16) { + return MA_INVALID_ARGS; + } + + /* The format cannot be changed after initialization. */ + if (pBQ->format != ma_format_unknown && pBQ->format != pConfig->format) { + return MA_INVALID_OPERATION; + } + + /* The channel count cannot be changed after initialization. */ + if (pBQ->channels != 0 && pBQ->channels != pConfig->channels) { + return MA_INVALID_OPERATION; + } + + + pBQ->format = pConfig->format; + pBQ->channels = pConfig->channels; + + /* Normalize. */ + if (pConfig->format == ma_format_f32) { + pBQ->b0.f32 = (float)(pConfig->b0 / pConfig->a0); + pBQ->b1.f32 = (float)(pConfig->b1 / pConfig->a0); + pBQ->b2.f32 = (float)(pConfig->b2 / pConfig->a0); + pBQ->a1.f32 = (float)(pConfig->a1 / pConfig->a0); + pBQ->a2.f32 = (float)(pConfig->a2 / pConfig->a0); + } else { + pBQ->b0.s32 = ma_biquad_float_to_fp(pConfig->b0 / pConfig->a0); + pBQ->b1.s32 = ma_biquad_float_to_fp(pConfig->b1 / pConfig->a0); + pBQ->b2.s32 = ma_biquad_float_to_fp(pConfig->b2 / pConfig->a0); + pBQ->a1.s32 = ma_biquad_float_to_fp(pConfig->a1 / pConfig->a0); + pBQ->a2.s32 = ma_biquad_float_to_fp(pConfig->a2 / pConfig->a0); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_biquad_clear_cache(ma_biquad* pBQ) +{ + if (pBQ == NULL) { + return MA_INVALID_ARGS; + } + + if (pBQ->format == ma_format_f32) { + pBQ->pR1->f32 = 0; + pBQ->pR2->f32 = 0; + } else { + pBQ->pR1->s32 = 0; + pBQ->pR2->s32 = 0; + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_biquad_process_pcm_frame_f32__direct_form_2_transposed(ma_biquad* pBQ, float* pY, const float* pX) +{ + ma_uint32 c; + const ma_uint32 channels = pBQ->channels; + const float b0 = pBQ->b0.f32; + const float b1 = pBQ->b1.f32; + const float b2 = pBQ->b2.f32; + const float a1 = pBQ->a1.f32; + const float a2 = pBQ->a2.f32; + + MA_ASSUME(channels > 0); + for (c = 0; c < channels; c += 1) { + float r1 = pBQ->pR1[c].f32; + float r2 = pBQ->pR2[c].f32; + float x = pX[c]; + float y; + + y = b0*x + r1; + r1 = b1*x - a1*y + r2; + r2 = b2*x - a2*y; + + pY[c] = y; + pBQ->pR1[c].f32 = r1; + pBQ->pR2[c].f32 = r2; + } +} + +static MA_INLINE void ma_biquad_process_pcm_frame_f32(ma_biquad* pBQ, float* pY, const float* pX) +{ + ma_biquad_process_pcm_frame_f32__direct_form_2_transposed(pBQ, pY, pX); +} + +static MA_INLINE void ma_biquad_process_pcm_frame_s16__direct_form_2_transposed(ma_biquad* pBQ, ma_int16* pY, const ma_int16* pX) +{ + ma_uint32 c; + const ma_uint32 channels = pBQ->channels; + const ma_int32 b0 = pBQ->b0.s32; + const ma_int32 b1 = pBQ->b1.s32; + const ma_int32 b2 = pBQ->b2.s32; + const ma_int32 a1 = pBQ->a1.s32; + const ma_int32 a2 = pBQ->a2.s32; + + MA_ASSUME(channels > 0); + for (c = 0; c < channels; c += 1) { + ma_int32 r1 = pBQ->pR1[c].s32; + ma_int32 r2 = pBQ->pR2[c].s32; + ma_int32 x = pX[c]; + ma_int32 y; + + y = (b0*x + r1) >> MA_BIQUAD_FIXED_POINT_SHIFT; + r1 = (b1*x - a1*y + r2); + r2 = (b2*x - a2*y); + + pY[c] = (ma_int16)ma_clamp(y, -32768, 32767); + pBQ->pR1[c].s32 = r1; + pBQ->pR2[c].s32 = r2; + } +} + +static MA_INLINE void ma_biquad_process_pcm_frame_s16(ma_biquad* pBQ, ma_int16* pY, const ma_int16* pX) +{ + ma_biquad_process_pcm_frame_s16__direct_form_2_transposed(pBQ, pY, pX); +} + +MA_API ma_result ma_biquad_process_pcm_frames(ma_biquad* pBQ, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_uint32 n; + + if (pBQ == NULL || pFramesOut == NULL || pFramesIn == NULL) { + return MA_INVALID_ARGS; + } + + /* Note that the logic below needs to support in-place filtering. That is, it must support the case where pFramesOut and pFramesIn are the same. */ + + if (pBQ->format == ma_format_f32) { + /* */ float* pY = ( float*)pFramesOut; + const float* pX = (const float*)pFramesIn; + + for (n = 0; n < frameCount; n += 1) { + ma_biquad_process_pcm_frame_f32__direct_form_2_transposed(pBQ, pY, pX); + pY += pBQ->channels; + pX += pBQ->channels; + } + } else if (pBQ->format == ma_format_s16) { + /* */ ma_int16* pY = ( ma_int16*)pFramesOut; + const ma_int16* pX = (const ma_int16*)pFramesIn; + + for (n = 0; n < frameCount; n += 1) { + ma_biquad_process_pcm_frame_s16__direct_form_2_transposed(pBQ, pY, pX); + pY += pBQ->channels; + pX += pBQ->channels; + } + } else { + MA_ASSERT(MA_FALSE); + return MA_INVALID_ARGS; /* Format not supported. Should never hit this because it's checked in ma_biquad_init() and ma_biquad_reinit(). */ + } + + return MA_SUCCESS; +} + +MA_API ma_uint32 ma_biquad_get_latency(const ma_biquad* pBQ) +{ + if (pBQ == NULL) { + return 0; + } + + return 2; +} + + +/************************************************************************************************************************************************************** + +Low-Pass Filter + +**************************************************************************************************************************************************************/ +MA_API ma_lpf1_config ma_lpf1_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency) +{ + ma_lpf1_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.cutoffFrequency = cutoffFrequency; + config.q = 0.5; + + return config; +} + +MA_API ma_lpf2_config ma_lpf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, double q) +{ + ma_lpf2_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.cutoffFrequency = cutoffFrequency; + config.q = q; + + /* Q cannot be 0 or else it'll result in a division by 0. In this case just default to 0.707107. */ + if (config.q == 0) { + config.q = 0.707107; + } + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t r1Offset; +} ma_lpf1_heap_layout; + +static ma_result ma_lpf1_get_heap_layout(const ma_lpf1_config* pConfig, ma_lpf1_heap_layout* pHeapLayout) +{ + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channels == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* R1 */ + pHeapLayout->r1Offset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(ma_biquad_coefficient) * pConfig->channels; + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +MA_API ma_result ma_lpf1_get_heap_size(const ma_lpf1_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_lpf1_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_lpf1_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_lpf1_init_preallocated(const ma_lpf1_config* pConfig, void* pHeap, ma_lpf1* pLPF) +{ + ma_result result; + ma_lpf1_heap_layout heapLayout; + + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pLPF); + + result = ma_lpf1_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pLPF->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pLPF->pR1 = (ma_biquad_coefficient*)ma_offset_ptr(pHeap, heapLayout.r1Offset); + + return ma_lpf1_reinit(pConfig, pLPF); +} + +MA_API ma_result ma_lpf1_init(const ma_lpf1_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_lpf1* pLPF) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_lpf1_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_lpf1_init_preallocated(pConfig, pHeap, pLPF); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pLPF->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_lpf1_uninit(ma_lpf1* pLPF, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pLPF == NULL) { + return; + } + + if (pLPF->_ownsHeap) { + ma_free(pLPF->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_lpf1_reinit(const ma_lpf1_config* pConfig, ma_lpf1* pLPF) +{ + double a; + + if (pLPF == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + /* Only supporting f32 and s16. */ + if (pConfig->format != ma_format_f32 && pConfig->format != ma_format_s16) { + return MA_INVALID_ARGS; + } + + /* The format cannot be changed after initialization. */ + if (pLPF->format != ma_format_unknown && pLPF->format != pConfig->format) { + return MA_INVALID_OPERATION; + } + + /* The channel count cannot be changed after initialization. */ + if (pLPF->channels != 0 && pLPF->channels != pConfig->channels) { + return MA_INVALID_OPERATION; + } + + pLPF->format = pConfig->format; + pLPF->channels = pConfig->channels; + + a = ma_expd(-2 * MA_PI_D * pConfig->cutoffFrequency / pConfig->sampleRate); + if (pConfig->format == ma_format_f32) { + pLPF->a.f32 = (float)a; + } else { + pLPF->a.s32 = ma_biquad_float_to_fp(a); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_lpf1_clear_cache(ma_lpf1* pLPF) +{ + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + if (pLPF->format == ma_format_f32) { + pLPF->a.f32 = 0; + } else { + pLPF->a.s32 = 0; + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_lpf1_process_pcm_frame_f32(ma_lpf1* pLPF, float* pY, const float* pX) +{ + ma_uint32 c; + const ma_uint32 channels = pLPF->channels; + const float a = pLPF->a.f32; + const float b = 1 - a; + + MA_ASSUME(channels > 0); + for (c = 0; c < channels; c += 1) { + float r1 = pLPF->pR1[c].f32; + float x = pX[c]; + float y; + + y = b*x + a*r1; + + pY[c] = y; + pLPF->pR1[c].f32 = y; + } +} + +static MA_INLINE void ma_lpf1_process_pcm_frame_s16(ma_lpf1* pLPF, ma_int16* pY, const ma_int16* pX) +{ + ma_uint32 c; + const ma_uint32 channels = pLPF->channels; + const ma_int32 a = pLPF->a.s32; + const ma_int32 b = ((1 << MA_BIQUAD_FIXED_POINT_SHIFT) - a); + + MA_ASSUME(channels > 0); + for (c = 0; c < channels; c += 1) { + ma_int32 r1 = pLPF->pR1[c].s32; + ma_int32 x = pX[c]; + ma_int32 y; + + y = (b*x + a*r1) >> MA_BIQUAD_FIXED_POINT_SHIFT; + + pY[c] = (ma_int16)y; + pLPF->pR1[c].s32 = (ma_int32)y; + } +} + +MA_API ma_result ma_lpf1_process_pcm_frames(ma_lpf1* pLPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_uint32 n; + + if (pLPF == NULL || pFramesOut == NULL || pFramesIn == NULL) { + return MA_INVALID_ARGS; + } + + /* Note that the logic below needs to support in-place filtering. That is, it must support the case where pFramesOut and pFramesIn are the same. */ + + if (pLPF->format == ma_format_f32) { + /* */ float* pY = ( float*)pFramesOut; + const float* pX = (const float*)pFramesIn; + + for (n = 0; n < frameCount; n += 1) { + ma_lpf1_process_pcm_frame_f32(pLPF, pY, pX); + pY += pLPF->channels; + pX += pLPF->channels; + } + } else if (pLPF->format == ma_format_s16) { + /* */ ma_int16* pY = ( ma_int16*)pFramesOut; + const ma_int16* pX = (const ma_int16*)pFramesIn; + + for (n = 0; n < frameCount; n += 1) { + ma_lpf1_process_pcm_frame_s16(pLPF, pY, pX); + pY += pLPF->channels; + pX += pLPF->channels; + } + } else { + MA_ASSERT(MA_FALSE); + return MA_INVALID_ARGS; /* Format not supported. Should never hit this because it's checked in ma_biquad_init() and ma_biquad_reinit(). */ + } + + return MA_SUCCESS; +} + +MA_API ma_uint32 ma_lpf1_get_latency(const ma_lpf1* pLPF) +{ + if (pLPF == NULL) { + return 0; + } + + return 1; +} + + +static MA_INLINE ma_biquad_config ma_lpf2__get_biquad_config(const ma_lpf2_config* pConfig) +{ + ma_biquad_config bqConfig; + double q; + double w; + double s; + double c; + double a; + + MA_ASSERT(pConfig != NULL); + + q = pConfig->q; + w = 2 * MA_PI_D * pConfig->cutoffFrequency / pConfig->sampleRate; + s = ma_sind(w); + c = ma_cosd(w); + a = s / (2*q); + + bqConfig.b0 = (1 - c) / 2; + bqConfig.b1 = 1 - c; + bqConfig.b2 = (1 - c) / 2; + bqConfig.a0 = 1 + a; + bqConfig.a1 = -2 * c; + bqConfig.a2 = 1 - a; + + bqConfig.format = pConfig->format; + bqConfig.channels = pConfig->channels; + + return bqConfig; +} + +MA_API ma_result ma_lpf2_get_heap_size(const ma_lpf2_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_biquad_config bqConfig; + bqConfig = ma_lpf2__get_biquad_config(pConfig); + + return ma_biquad_get_heap_size(&bqConfig, pHeapSizeInBytes); +} + +MA_API ma_result ma_lpf2_init_preallocated(const ma_lpf2_config* pConfig, void* pHeap, ma_lpf2* pLPF) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pLPF); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_lpf2__get_biquad_config(pConfig); + result = ma_biquad_init_preallocated(&bqConfig, pHeap, &pLPF->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_lpf2_init(const ma_lpf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_lpf2* pLPF) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_lpf2_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_lpf2_init_preallocated(pConfig, pHeap, pLPF); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pLPF->bq._ownsHeap = MA_TRUE; /* <-- This will cause the biquad to take ownership of the heap and free it when it's uninitialized. */ + return MA_SUCCESS; +} + +MA_API void ma_lpf2_uninit(ma_lpf2* pLPF, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pLPF == NULL) { + return; + } + + ma_biquad_uninit(&pLPF->bq, pAllocationCallbacks); /* <-- This will free the heap allocation. */ +} + +MA_API ma_result ma_lpf2_reinit(const ma_lpf2_config* pConfig, ma_lpf2* pLPF) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pLPF == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_lpf2__get_biquad_config(pConfig); + result = ma_biquad_reinit(&bqConfig, &pLPF->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_lpf2_clear_cache(ma_lpf2* pLPF) +{ + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + ma_biquad_clear_cache(&pLPF->bq); + + return MA_SUCCESS; +} + +static MA_INLINE void ma_lpf2_process_pcm_frame_s16(ma_lpf2* pLPF, ma_int16* pFrameOut, const ma_int16* pFrameIn) +{ + ma_biquad_process_pcm_frame_s16(&pLPF->bq, pFrameOut, pFrameIn); +} + +static MA_INLINE void ma_lpf2_process_pcm_frame_f32(ma_lpf2* pLPF, float* pFrameOut, const float* pFrameIn) +{ + ma_biquad_process_pcm_frame_f32(&pLPF->bq, pFrameOut, pFrameIn); +} + +MA_API ma_result ma_lpf2_process_pcm_frames(ma_lpf2* pLPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + return ma_biquad_process_pcm_frames(&pLPF->bq, pFramesOut, pFramesIn, frameCount); +} + +MA_API ma_uint32 ma_lpf2_get_latency(const ma_lpf2* pLPF) +{ + if (pLPF == NULL) { + return 0; + } + + return ma_biquad_get_latency(&pLPF->bq); +} + + +MA_API ma_lpf_config ma_lpf_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order) +{ + ma_lpf_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.cutoffFrequency = cutoffFrequency; + config.order = ma_min(order, MA_MAX_FILTER_ORDER); + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t lpf1Offset; + size_t lpf2Offset; /* Offset of the first second order filter. Subsequent filters will come straight after, and will each have the same heap size. */ +} ma_lpf_heap_layout; + +static void ma_lpf_calculate_sub_lpf_counts(ma_uint32 order, ma_uint32* pLPF1Count, ma_uint32* pLPF2Count) +{ + MA_ASSERT(pLPF1Count != NULL); + MA_ASSERT(pLPF2Count != NULL); + + *pLPF1Count = order % 2; + *pLPF2Count = order / 2; +} + +static ma_result ma_lpf_get_heap_layout(const ma_lpf_config* pConfig, ma_lpf_heap_layout* pHeapLayout) +{ + ma_result result; + ma_uint32 lpf1Count; + ma_uint32 lpf2Count; + ma_uint32 ilpf1; + ma_uint32 ilpf2; + + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channels == 0) { + return MA_INVALID_ARGS; + } + + if (pConfig->order > MA_MAX_FILTER_ORDER) { + return MA_INVALID_ARGS; + } + + ma_lpf_calculate_sub_lpf_counts(pConfig->order, &lpf1Count, &lpf2Count); + + pHeapLayout->sizeInBytes = 0; + + /* LPF 1 */ + pHeapLayout->lpf1Offset = pHeapLayout->sizeInBytes; + for (ilpf1 = 0; ilpf1 < lpf1Count; ilpf1 += 1) { + size_t lpf1HeapSizeInBytes; + ma_lpf1_config lpf1Config = ma_lpf1_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency); + + result = ma_lpf1_get_heap_size(&lpf1Config, &lpf1HeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes += sizeof(ma_lpf1) + lpf1HeapSizeInBytes; + } + + /* LPF 2*/ + pHeapLayout->lpf2Offset = pHeapLayout->sizeInBytes; + for (ilpf2 = 0; ilpf2 < lpf2Count; ilpf2 += 1) { + size_t lpf2HeapSizeInBytes; + ma_lpf2_config lpf2Config = ma_lpf2_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency, 0.707107); /* <-- The "q" parameter does not matter for the purpose of calculating the heap size. */ + + result = ma_lpf2_get_heap_size(&lpf2Config, &lpf2HeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes += sizeof(ma_lpf2) + lpf2HeapSizeInBytes; + } + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +static ma_result ma_lpf_reinit__internal(const ma_lpf_config* pConfig, void* pHeap, ma_lpf* pLPF, ma_bool32 isNew) +{ + ma_result result; + ma_uint32 lpf1Count; + ma_uint32 lpf2Count; + ma_uint32 ilpf1; + ma_uint32 ilpf2; + ma_lpf_heap_layout heapLayout; /* Only used if isNew is true. */ + + if (pLPF == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + /* Only supporting f32 and s16. */ + if (pConfig->format != ma_format_f32 && pConfig->format != ma_format_s16) { + return MA_INVALID_ARGS; + } + + /* The format cannot be changed after initialization. */ + if (pLPF->format != ma_format_unknown && pLPF->format != pConfig->format) { + return MA_INVALID_OPERATION; + } + + /* The channel count cannot be changed after initialization. */ + if (pLPF->channels != 0 && pLPF->channels != pConfig->channels) { + return MA_INVALID_OPERATION; + } + + if (pConfig->order > MA_MAX_FILTER_ORDER) { + return MA_INVALID_ARGS; + } + + ma_lpf_calculate_sub_lpf_counts(pConfig->order, &lpf1Count, &lpf2Count); + + /* The filter order can't change between reinits. */ + if (!isNew) { + if (pLPF->lpf1Count != lpf1Count || pLPF->lpf2Count != lpf2Count) { + return MA_INVALID_OPERATION; + } + } + + if (isNew) { + result = ma_lpf_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pLPF->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pLPF->pLPF1 = (ma_lpf1*)ma_offset_ptr(pHeap, heapLayout.lpf1Offset); + pLPF->pLPF2 = (ma_lpf2*)ma_offset_ptr(pHeap, heapLayout.lpf2Offset); + } else { + MA_ZERO_OBJECT(&heapLayout); /* To silence a compiler warning. */ + } + + for (ilpf1 = 0; ilpf1 < lpf1Count; ilpf1 += 1) { + ma_lpf1_config lpf1Config = ma_lpf1_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency); + + if (isNew) { + size_t lpf1HeapSizeInBytes; + + result = ma_lpf1_get_heap_size(&lpf1Config, &lpf1HeapSizeInBytes); + if (result == MA_SUCCESS) { + result = ma_lpf1_init_preallocated(&lpf1Config, ma_offset_ptr(pHeap, heapLayout.lpf1Offset + (sizeof(ma_lpf1) * lpf1Count) + (ilpf1 * lpf1HeapSizeInBytes)), &pLPF->pLPF1[ilpf1]); + } + } else { + result = ma_lpf1_reinit(&lpf1Config, &pLPF->pLPF1[ilpf1]); + } + + if (result != MA_SUCCESS) { + ma_uint32 jlpf1; + + for (jlpf1 = 0; jlpf1 < ilpf1; jlpf1 += 1) { + ma_lpf1_uninit(&pLPF->pLPF1[jlpf1], NULL); /* No need for allocation callbacks here since we used a preallocated heap allocation. */ + } + + return result; + } + } + + for (ilpf2 = 0; ilpf2 < lpf2Count; ilpf2 += 1) { + ma_lpf2_config lpf2Config; + double q; + double a; + + /* Tempting to use 0.707107, but won't result in a Butterworth filter if the order is > 2. */ + if (lpf1Count == 1) { + a = (1 + ilpf2*1) * (MA_PI_D/(pConfig->order*1)); /* Odd order. */ + } else { + a = (1 + ilpf2*2) * (MA_PI_D/(pConfig->order*2)); /* Even order. */ + } + q = 1 / (2*ma_cosd(a)); + + lpf2Config = ma_lpf2_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency, q); + + if (isNew) { + size_t lpf2HeapSizeInBytes; + + result = ma_lpf2_get_heap_size(&lpf2Config, &lpf2HeapSizeInBytes); + if (result == MA_SUCCESS) { + result = ma_lpf2_init_preallocated(&lpf2Config, ma_offset_ptr(pHeap, heapLayout.lpf2Offset + (sizeof(ma_lpf2) * lpf2Count) + (ilpf2 * lpf2HeapSizeInBytes)), &pLPF->pLPF2[ilpf2]); + } + } else { + result = ma_lpf2_reinit(&lpf2Config, &pLPF->pLPF2[ilpf2]); + } + + if (result != MA_SUCCESS) { + ma_uint32 jlpf1; + ma_uint32 jlpf2; + + for (jlpf1 = 0; jlpf1 < lpf1Count; jlpf1 += 1) { + ma_lpf1_uninit(&pLPF->pLPF1[jlpf1], NULL); /* No need for allocation callbacks here since we used a preallocated heap allocation. */ + } + + for (jlpf2 = 0; jlpf2 < ilpf2; jlpf2 += 1) { + ma_lpf2_uninit(&pLPF->pLPF2[jlpf2], NULL); /* No need for allocation callbacks here since we used a preallocated heap allocation. */ + } + + return result; + } + } + + pLPF->lpf1Count = lpf1Count; + pLPF->lpf2Count = lpf2Count; + pLPF->format = pConfig->format; + pLPF->channels = pConfig->channels; + pLPF->sampleRate = pConfig->sampleRate; + + return MA_SUCCESS; +} + +MA_API ma_result ma_lpf_get_heap_size(const ma_lpf_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_lpf_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_lpf_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return result; +} + +MA_API ma_result ma_lpf_init_preallocated(const ma_lpf_config* pConfig, void* pHeap, ma_lpf* pLPF) +{ + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pLPF); + + return ma_lpf_reinit__internal(pConfig, pHeap, pLPF, /*isNew*/MA_TRUE); +} + +MA_API ma_result ma_lpf_init(const ma_lpf_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_lpf* pLPF) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_lpf_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_lpf_init_preallocated(pConfig, pHeap, pLPF); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pLPF->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_lpf_uninit(ma_lpf* pLPF, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_uint32 ilpf1; + ma_uint32 ilpf2; + + if (pLPF == NULL) { + return; + } + + for (ilpf1 = 0; ilpf1 < pLPF->lpf1Count; ilpf1 += 1) { + ma_lpf1_uninit(&pLPF->pLPF1[ilpf1], pAllocationCallbacks); + } + + for (ilpf2 = 0; ilpf2 < pLPF->lpf2Count; ilpf2 += 1) { + ma_lpf2_uninit(&pLPF->pLPF2[ilpf2], pAllocationCallbacks); + } + + if (pLPF->_ownsHeap) { + ma_free(pLPF->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_lpf_reinit(const ma_lpf_config* pConfig, ma_lpf* pLPF) +{ + return ma_lpf_reinit__internal(pConfig, NULL, pLPF, /*isNew*/MA_FALSE); +} + +MA_API ma_result ma_lpf_clear_cache(ma_lpf* pLPF) +{ + ma_uint32 ilpf1; + ma_uint32 ilpf2; + + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + for (ilpf1 = 0; ilpf1 < pLPF->lpf1Count; ilpf1 += 1) { + ma_lpf1_clear_cache(&pLPF->pLPF1[ilpf1]); + } + + for (ilpf2 = 0; ilpf2 < pLPF->lpf2Count; ilpf2 += 1) { + ma_lpf2_clear_cache(&pLPF->pLPF2[ilpf2]); + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_lpf_process_pcm_frame_f32(ma_lpf* pLPF, float* pY, const void* pX) +{ + ma_uint32 ilpf1; + ma_uint32 ilpf2; + + MA_ASSERT(pLPF->format == ma_format_f32); + + MA_MOVE_MEMORY(pY, pX, ma_get_bytes_per_frame(pLPF->format, pLPF->channels)); + + for (ilpf1 = 0; ilpf1 < pLPF->lpf1Count; ilpf1 += 1) { + ma_lpf1_process_pcm_frame_f32(&pLPF->pLPF1[ilpf1], pY, pY); + } + + for (ilpf2 = 0; ilpf2 < pLPF->lpf2Count; ilpf2 += 1) { + ma_lpf2_process_pcm_frame_f32(&pLPF->pLPF2[ilpf2], pY, pY); + } +} + +static MA_INLINE void ma_lpf_process_pcm_frame_s16(ma_lpf* pLPF, ma_int16* pY, const ma_int16* pX) +{ + ma_uint32 ilpf1; + ma_uint32 ilpf2; + + MA_ASSERT(pLPF->format == ma_format_s16); + + MA_MOVE_MEMORY(pY, pX, ma_get_bytes_per_frame(pLPF->format, pLPF->channels)); + + for (ilpf1 = 0; ilpf1 < pLPF->lpf1Count; ilpf1 += 1) { + ma_lpf1_process_pcm_frame_s16(&pLPF->pLPF1[ilpf1], pY, pY); + } + + for (ilpf2 = 0; ilpf2 < pLPF->lpf2Count; ilpf2 += 1) { + ma_lpf2_process_pcm_frame_s16(&pLPF->pLPF2[ilpf2], pY, pY); + } +} + +MA_API ma_result ma_lpf_process_pcm_frames(ma_lpf* pLPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_result result; + ma_uint32 ilpf1; + ma_uint32 ilpf2; + + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + /* Faster path for in-place. */ + if (pFramesOut == pFramesIn) { + for (ilpf1 = 0; ilpf1 < pLPF->lpf1Count; ilpf1 += 1) { + result = ma_lpf1_process_pcm_frames(&pLPF->pLPF1[ilpf1], pFramesOut, pFramesOut, frameCount); + if (result != MA_SUCCESS) { + return result; + } + } + + for (ilpf2 = 0; ilpf2 < pLPF->lpf2Count; ilpf2 += 1) { + result = ma_lpf2_process_pcm_frames(&pLPF->pLPF2[ilpf2], pFramesOut, pFramesOut, frameCount); + if (result != MA_SUCCESS) { + return result; + } + } + } + + /* Slightly slower path for copying. */ + if (pFramesOut != pFramesIn) { + ma_uint32 iFrame; + + /* */ if (pLPF->format == ma_format_f32) { + /* */ float* pFramesOutF32 = ( float*)pFramesOut; + const float* pFramesInF32 = (const float*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_lpf_process_pcm_frame_f32(pLPF, pFramesOutF32, pFramesInF32); + pFramesOutF32 += pLPF->channels; + pFramesInF32 += pLPF->channels; + } + } else if (pLPF->format == ma_format_s16) { + /* */ ma_int16* pFramesOutS16 = ( ma_int16*)pFramesOut; + const ma_int16* pFramesInS16 = (const ma_int16*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_lpf_process_pcm_frame_s16(pLPF, pFramesOutS16, pFramesInS16); + pFramesOutS16 += pLPF->channels; + pFramesInS16 += pLPF->channels; + } + } else { + MA_ASSERT(MA_FALSE); + return MA_INVALID_OPERATION; /* Should never hit this. */ + } + } + + return MA_SUCCESS; +} + +MA_API ma_uint32 ma_lpf_get_latency(const ma_lpf* pLPF) +{ + if (pLPF == NULL) { + return 0; + } + + return pLPF->lpf2Count*2 + pLPF->lpf1Count; +} + + +/************************************************************************************************************************************************************** + +High-Pass Filtering + +**************************************************************************************************************************************************************/ +MA_API ma_hpf1_config ma_hpf1_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency) +{ + ma_hpf1_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.cutoffFrequency = cutoffFrequency; + + return config; +} + +MA_API ma_hpf2_config ma_hpf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, double q) +{ + ma_hpf2_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.cutoffFrequency = cutoffFrequency; + config.q = q; + + /* Q cannot be 0 or else it'll result in a division by 0. In this case just default to 0.707107. */ + if (config.q == 0) { + config.q = 0.707107; + } + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t r1Offset; +} ma_hpf1_heap_layout; + +static ma_result ma_hpf1_get_heap_layout(const ma_hpf1_config* pConfig, ma_hpf1_heap_layout* pHeapLayout) +{ + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channels == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* R1 */ + pHeapLayout->r1Offset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(ma_biquad_coefficient) * pConfig->channels; + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +MA_API ma_result ma_hpf1_get_heap_size(const ma_hpf1_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_hpf1_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_hpf1_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_hpf1_init_preallocated(const ma_hpf1_config* pConfig, void* pHeap, ma_hpf1* pLPF) +{ + ma_result result; + ma_hpf1_heap_layout heapLayout; + + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pLPF); + + result = ma_hpf1_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pLPF->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pLPF->pR1 = (ma_biquad_coefficient*)ma_offset_ptr(pHeap, heapLayout.r1Offset); + + return ma_hpf1_reinit(pConfig, pLPF); +} + +MA_API ma_result ma_hpf1_init(const ma_hpf1_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hpf1* pLPF) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_hpf1_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_hpf1_init_preallocated(pConfig, pHeap, pLPF); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pLPF->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_hpf1_uninit(ma_hpf1* pHPF, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pHPF == NULL) { + return; + } + + if (pHPF->_ownsHeap) { + ma_free(pHPF->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_hpf1_reinit(const ma_hpf1_config* pConfig, ma_hpf1* pHPF) +{ + double a; + + if (pHPF == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + /* Only supporting f32 and s16. */ + if (pConfig->format != ma_format_f32 && pConfig->format != ma_format_s16) { + return MA_INVALID_ARGS; + } + + /* The format cannot be changed after initialization. */ + if (pHPF->format != ma_format_unknown && pHPF->format != pConfig->format) { + return MA_INVALID_OPERATION; + } + + /* The channel count cannot be changed after initialization. */ + if (pHPF->channels != 0 && pHPF->channels != pConfig->channels) { + return MA_INVALID_OPERATION; + } + + pHPF->format = pConfig->format; + pHPF->channels = pConfig->channels; + + a = ma_expd(-2 * MA_PI_D * pConfig->cutoffFrequency / pConfig->sampleRate); + if (pConfig->format == ma_format_f32) { + pHPF->a.f32 = (float)a; + } else { + pHPF->a.s32 = ma_biquad_float_to_fp(a); + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_hpf1_process_pcm_frame_f32(ma_hpf1* pHPF, float* pY, const float* pX) +{ + ma_uint32 c; + const ma_uint32 channels = pHPF->channels; + const float a = 1 - pHPF->a.f32; + const float b = 1 - a; + + MA_ASSUME(channels > 0); + for (c = 0; c < channels; c += 1) { + float r1 = pHPF->pR1[c].f32; + float x = pX[c]; + float y; + + y = b*x - a*r1; + + pY[c] = y; + pHPF->pR1[c].f32 = y; + } +} + +static MA_INLINE void ma_hpf1_process_pcm_frame_s16(ma_hpf1* pHPF, ma_int16* pY, const ma_int16* pX) +{ + ma_uint32 c; + const ma_uint32 channels = pHPF->channels; + const ma_int32 a = ((1 << MA_BIQUAD_FIXED_POINT_SHIFT) - pHPF->a.s32); + const ma_int32 b = ((1 << MA_BIQUAD_FIXED_POINT_SHIFT) - a); + + MA_ASSUME(channels > 0); + for (c = 0; c < channels; c += 1) { + ma_int32 r1 = pHPF->pR1[c].s32; + ma_int32 x = pX[c]; + ma_int32 y; + + y = (b*x - a*r1) >> MA_BIQUAD_FIXED_POINT_SHIFT; + + pY[c] = (ma_int16)y; + pHPF->pR1[c].s32 = (ma_int32)y; + } +} + +MA_API ma_result ma_hpf1_process_pcm_frames(ma_hpf1* pHPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_uint32 n; + + if (pHPF == NULL || pFramesOut == NULL || pFramesIn == NULL) { + return MA_INVALID_ARGS; + } + + /* Note that the logic below needs to support in-place filtering. That is, it must support the case where pFramesOut and pFramesIn are the same. */ + + if (pHPF->format == ma_format_f32) { + /* */ float* pY = ( float*)pFramesOut; + const float* pX = (const float*)pFramesIn; + + for (n = 0; n < frameCount; n += 1) { + ma_hpf1_process_pcm_frame_f32(pHPF, pY, pX); + pY += pHPF->channels; + pX += pHPF->channels; + } + } else if (pHPF->format == ma_format_s16) { + /* */ ma_int16* pY = ( ma_int16*)pFramesOut; + const ma_int16* pX = (const ma_int16*)pFramesIn; + + for (n = 0; n < frameCount; n += 1) { + ma_hpf1_process_pcm_frame_s16(pHPF, pY, pX); + pY += pHPF->channels; + pX += pHPF->channels; + } + } else { + MA_ASSERT(MA_FALSE); + return MA_INVALID_ARGS; /* Format not supported. Should never hit this because it's checked in ma_biquad_init() and ma_biquad_reinit(). */ + } + + return MA_SUCCESS; +} + +MA_API ma_uint32 ma_hpf1_get_latency(const ma_hpf1* pHPF) +{ + if (pHPF == NULL) { + return 0; + } + + return 1; +} + + +static MA_INLINE ma_biquad_config ma_hpf2__get_biquad_config(const ma_hpf2_config* pConfig) +{ + ma_biquad_config bqConfig; + double q; + double w; + double s; + double c; + double a; + + MA_ASSERT(pConfig != NULL); + + q = pConfig->q; + w = 2 * MA_PI_D * pConfig->cutoffFrequency / pConfig->sampleRate; + s = ma_sind(w); + c = ma_cosd(w); + a = s / (2*q); + + bqConfig.b0 = (1 + c) / 2; + bqConfig.b1 = -(1 + c); + bqConfig.b2 = (1 + c) / 2; + bqConfig.a0 = 1 + a; + bqConfig.a1 = -2 * c; + bqConfig.a2 = 1 - a; + + bqConfig.format = pConfig->format; + bqConfig.channels = pConfig->channels; + + return bqConfig; +} + +MA_API ma_result ma_hpf2_get_heap_size(const ma_hpf2_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_biquad_config bqConfig; + bqConfig = ma_hpf2__get_biquad_config(pConfig); + + return ma_biquad_get_heap_size(&bqConfig, pHeapSizeInBytes); +} + +MA_API ma_result ma_hpf2_init_preallocated(const ma_hpf2_config* pConfig, void* pHeap, ma_hpf2* pHPF) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pHPF == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pHPF); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_hpf2__get_biquad_config(pConfig); + result = ma_biquad_init_preallocated(&bqConfig, pHeap, &pHPF->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_hpf2_init(const ma_hpf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hpf2* pHPF) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_hpf2_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_hpf2_init_preallocated(pConfig, pHeap, pHPF); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pHPF->bq._ownsHeap = MA_TRUE; /* <-- This will cause the biquad to take ownership of the heap and free it when it's uninitialized. */ + return MA_SUCCESS; +} + +MA_API void ma_hpf2_uninit(ma_hpf2* pHPF, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pHPF == NULL) { + return; + } + + ma_biquad_uninit(&pHPF->bq, pAllocationCallbacks); /* <-- This will free the heap allocation. */ +} + +MA_API ma_result ma_hpf2_reinit(const ma_hpf2_config* pConfig, ma_hpf2* pHPF) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pHPF == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_hpf2__get_biquad_config(pConfig); + result = ma_biquad_reinit(&bqConfig, &pHPF->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_hpf2_process_pcm_frame_s16(ma_hpf2* pHPF, ma_int16* pFrameOut, const ma_int16* pFrameIn) +{ + ma_biquad_process_pcm_frame_s16(&pHPF->bq, pFrameOut, pFrameIn); +} + +static MA_INLINE void ma_hpf2_process_pcm_frame_f32(ma_hpf2* pHPF, float* pFrameOut, const float* pFrameIn) +{ + ma_biquad_process_pcm_frame_f32(&pHPF->bq, pFrameOut, pFrameIn); +} + +MA_API ma_result ma_hpf2_process_pcm_frames(ma_hpf2* pHPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pHPF == NULL) { + return MA_INVALID_ARGS; + } + + return ma_biquad_process_pcm_frames(&pHPF->bq, pFramesOut, pFramesIn, frameCount); +} + +MA_API ma_uint32 ma_hpf2_get_latency(const ma_hpf2* pHPF) +{ + if (pHPF == NULL) { + return 0; + } + + return ma_biquad_get_latency(&pHPF->bq); +} + + +MA_API ma_hpf_config ma_hpf_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order) +{ + ma_hpf_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.cutoffFrequency = cutoffFrequency; + config.order = ma_min(order, MA_MAX_FILTER_ORDER); + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t hpf1Offset; + size_t hpf2Offset; /* Offset of the first second order filter. Subsequent filters will come straight after, and will each have the same heap size. */ +} ma_hpf_heap_layout; + +static void ma_hpf_calculate_sub_hpf_counts(ma_uint32 order, ma_uint32* pHPF1Count, ma_uint32* pHPF2Count) +{ + MA_ASSERT(pHPF1Count != NULL); + MA_ASSERT(pHPF2Count != NULL); + + *pHPF1Count = order % 2; + *pHPF2Count = order / 2; +} + +static ma_result ma_hpf_get_heap_layout(const ma_hpf_config* pConfig, ma_hpf_heap_layout* pHeapLayout) +{ + ma_result result; + ma_uint32 hpf1Count; + ma_uint32 hpf2Count; + ma_uint32 ihpf1; + ma_uint32 ihpf2; + + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channels == 0) { + return MA_INVALID_ARGS; + } + + if (pConfig->order > MA_MAX_FILTER_ORDER) { + return MA_INVALID_ARGS; + } + + ma_hpf_calculate_sub_hpf_counts(pConfig->order, &hpf1Count, &hpf2Count); + + pHeapLayout->sizeInBytes = 0; + + /* HPF 1 */ + pHeapLayout->hpf1Offset = pHeapLayout->sizeInBytes; + for (ihpf1 = 0; ihpf1 < hpf1Count; ihpf1 += 1) { + size_t hpf1HeapSizeInBytes; + ma_hpf1_config hpf1Config = ma_hpf1_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency); + + result = ma_hpf1_get_heap_size(&hpf1Config, &hpf1HeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes += sizeof(ma_hpf1) + hpf1HeapSizeInBytes; + } + + /* HPF 2*/ + pHeapLayout->hpf2Offset = pHeapLayout->sizeInBytes; + for (ihpf2 = 0; ihpf2 < hpf2Count; ihpf2 += 1) { + size_t hpf2HeapSizeInBytes; + ma_hpf2_config hpf2Config = ma_hpf2_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency, 0.707107); /* <-- The "q" parameter does not matter for the purpose of calculating the heap size. */ + + result = ma_hpf2_get_heap_size(&hpf2Config, &hpf2HeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes += sizeof(ma_hpf2) + hpf2HeapSizeInBytes; + } + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +static ma_result ma_hpf_reinit__internal(const ma_hpf_config* pConfig, void* pHeap, ma_hpf* pHPF, ma_bool32 isNew) +{ + ma_result result; + ma_uint32 hpf1Count; + ma_uint32 hpf2Count; + ma_uint32 ihpf1; + ma_uint32 ihpf2; + ma_hpf_heap_layout heapLayout; /* Only used if isNew is true. */ + + if (pHPF == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + /* Only supporting f32 and s16. */ + if (pConfig->format != ma_format_f32 && pConfig->format != ma_format_s16) { + return MA_INVALID_ARGS; + } + + /* The format cannot be changed after initialization. */ + if (pHPF->format != ma_format_unknown && pHPF->format != pConfig->format) { + return MA_INVALID_OPERATION; + } + + /* The channel count cannot be changed after initialization. */ + if (pHPF->channels != 0 && pHPF->channels != pConfig->channels) { + return MA_INVALID_OPERATION; + } + + if (pConfig->order > MA_MAX_FILTER_ORDER) { + return MA_INVALID_ARGS; + } + + ma_hpf_calculate_sub_hpf_counts(pConfig->order, &hpf1Count, &hpf2Count); + + /* The filter order can't change between reinits. */ + if (!isNew) { + if (pHPF->hpf1Count != hpf1Count || pHPF->hpf2Count != hpf2Count) { + return MA_INVALID_OPERATION; + } + } + + if (isNew) { + result = ma_hpf_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pHPF->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pHPF->pHPF1 = (ma_hpf1*)ma_offset_ptr(pHeap, heapLayout.hpf1Offset); + pHPF->pHPF2 = (ma_hpf2*)ma_offset_ptr(pHeap, heapLayout.hpf2Offset); + } else { + MA_ZERO_OBJECT(&heapLayout); /* To silence a compiler warning. */ + } + + for (ihpf1 = 0; ihpf1 < hpf1Count; ihpf1 += 1) { + ma_hpf1_config hpf1Config = ma_hpf1_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency); + + if (isNew) { + size_t hpf1HeapSizeInBytes; + + result = ma_hpf1_get_heap_size(&hpf1Config, &hpf1HeapSizeInBytes); + if (result == MA_SUCCESS) { + result = ma_hpf1_init_preallocated(&hpf1Config, ma_offset_ptr(pHeap, heapLayout.hpf1Offset + (sizeof(ma_hpf1) * hpf1Count) + (ihpf1 * hpf1HeapSizeInBytes)), &pHPF->pHPF1[ihpf1]); + } + } else { + result = ma_hpf1_reinit(&hpf1Config, &pHPF->pHPF1[ihpf1]); + } + + if (result != MA_SUCCESS) { + ma_uint32 jhpf1; + + for (jhpf1 = 0; jhpf1 < ihpf1; jhpf1 += 1) { + ma_hpf1_uninit(&pHPF->pHPF1[jhpf1], NULL); /* No need for allocation callbacks here since we used a preallocated heap allocation. */ + } + + return result; + } + } + + for (ihpf2 = 0; ihpf2 < hpf2Count; ihpf2 += 1) { + ma_hpf2_config hpf2Config; + double q; + double a; + + /* Tempting to use 0.707107, but won't result in a Butterworth filter if the order is > 2. */ + if (hpf1Count == 1) { + a = (1 + ihpf2*1) * (MA_PI_D/(pConfig->order*1)); /* Odd order. */ + } else { + a = (1 + ihpf2*2) * (MA_PI_D/(pConfig->order*2)); /* Even order. */ + } + q = 1 / (2*ma_cosd(a)); + + hpf2Config = ma_hpf2_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency, q); + + if (isNew) { + size_t hpf2HeapSizeInBytes; + + result = ma_hpf2_get_heap_size(&hpf2Config, &hpf2HeapSizeInBytes); + if (result == MA_SUCCESS) { + result = ma_hpf2_init_preallocated(&hpf2Config, ma_offset_ptr(pHeap, heapLayout.hpf2Offset + (sizeof(ma_hpf2) * hpf2Count) + (ihpf2 * hpf2HeapSizeInBytes)), &pHPF->pHPF2[ihpf2]); + } + } else { + result = ma_hpf2_reinit(&hpf2Config, &pHPF->pHPF2[ihpf2]); + } + + if (result != MA_SUCCESS) { + ma_uint32 jhpf1; + ma_uint32 jhpf2; + + for (jhpf1 = 0; jhpf1 < hpf1Count; jhpf1 += 1) { + ma_hpf1_uninit(&pHPF->pHPF1[jhpf1], NULL); /* No need for allocation callbacks here since we used a preallocated heap allocation. */ + } + + for (jhpf2 = 0; jhpf2 < ihpf2; jhpf2 += 1) { + ma_hpf2_uninit(&pHPF->pHPF2[jhpf2], NULL); /* No need for allocation callbacks here since we used a preallocated heap allocation. */ + } + + return result; + } + } + + pHPF->hpf1Count = hpf1Count; + pHPF->hpf2Count = hpf2Count; + pHPF->format = pConfig->format; + pHPF->channels = pConfig->channels; + pHPF->sampleRate = pConfig->sampleRate; + + return MA_SUCCESS; +} + +MA_API ma_result ma_hpf_get_heap_size(const ma_hpf_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_hpf_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_hpf_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return result; +} + +MA_API ma_result ma_hpf_init_preallocated(const ma_hpf_config* pConfig, void* pHeap, ma_hpf* pLPF) +{ + if (pLPF == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pLPF); + + return ma_hpf_reinit__internal(pConfig, pHeap, pLPF, /*isNew*/MA_TRUE); +} + +MA_API ma_result ma_hpf_init(const ma_hpf_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hpf* pHPF) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_hpf_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_hpf_init_preallocated(pConfig, pHeap, pHPF); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pHPF->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_hpf_uninit(ma_hpf* pHPF, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_uint32 ihpf1; + ma_uint32 ihpf2; + + if (pHPF == NULL) { + return; + } + + for (ihpf1 = 0; ihpf1 < pHPF->hpf1Count; ihpf1 += 1) { + ma_hpf1_uninit(&pHPF->pHPF1[ihpf1], pAllocationCallbacks); + } + + for (ihpf2 = 0; ihpf2 < pHPF->hpf2Count; ihpf2 += 1) { + ma_hpf2_uninit(&pHPF->pHPF2[ihpf2], pAllocationCallbacks); + } + + if (pHPF->_ownsHeap) { + ma_free(pHPF->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_hpf_reinit(const ma_hpf_config* pConfig, ma_hpf* pHPF) +{ + return ma_hpf_reinit__internal(pConfig, NULL, pHPF, /*isNew*/MA_FALSE); +} + +MA_API ma_result ma_hpf_process_pcm_frames(ma_hpf* pHPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_result result; + ma_uint32 ihpf1; + ma_uint32 ihpf2; + + if (pHPF == NULL) { + return MA_INVALID_ARGS; + } + + /* Faster path for in-place. */ + if (pFramesOut == pFramesIn) { + for (ihpf1 = 0; ihpf1 < pHPF->hpf1Count; ihpf1 += 1) { + result = ma_hpf1_process_pcm_frames(&pHPF->pHPF1[ihpf1], pFramesOut, pFramesOut, frameCount); + if (result != MA_SUCCESS) { + return result; + } + } + + for (ihpf2 = 0; ihpf2 < pHPF->hpf2Count; ihpf2 += 1) { + result = ma_hpf2_process_pcm_frames(&pHPF->pHPF2[ihpf2], pFramesOut, pFramesOut, frameCount); + if (result != MA_SUCCESS) { + return result; + } + } + } + + /* Slightly slower path for copying. */ + if (pFramesOut != pFramesIn) { + ma_uint32 iFrame; + + /* */ if (pHPF->format == ma_format_f32) { + /* */ float* pFramesOutF32 = ( float*)pFramesOut; + const float* pFramesInF32 = (const float*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + MA_COPY_MEMORY(pFramesOutF32, pFramesInF32, ma_get_bytes_per_frame(pHPF->format, pHPF->channels)); + + for (ihpf1 = 0; ihpf1 < pHPF->hpf1Count; ihpf1 += 1) { + ma_hpf1_process_pcm_frame_f32(&pHPF->pHPF1[ihpf1], pFramesOutF32, pFramesOutF32); + } + + for (ihpf2 = 0; ihpf2 < pHPF->hpf2Count; ihpf2 += 1) { + ma_hpf2_process_pcm_frame_f32(&pHPF->pHPF2[ihpf2], pFramesOutF32, pFramesOutF32); + } + + pFramesOutF32 += pHPF->channels; + pFramesInF32 += pHPF->channels; + } + } else if (pHPF->format == ma_format_s16) { + /* */ ma_int16* pFramesOutS16 = ( ma_int16*)pFramesOut; + const ma_int16* pFramesInS16 = (const ma_int16*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + MA_COPY_MEMORY(pFramesOutS16, pFramesInS16, ma_get_bytes_per_frame(pHPF->format, pHPF->channels)); + + for (ihpf1 = 0; ihpf1 < pHPF->hpf1Count; ihpf1 += 1) { + ma_hpf1_process_pcm_frame_s16(&pHPF->pHPF1[ihpf1], pFramesOutS16, pFramesOutS16); + } + + for (ihpf2 = 0; ihpf2 < pHPF->hpf2Count; ihpf2 += 1) { + ma_hpf2_process_pcm_frame_s16(&pHPF->pHPF2[ihpf2], pFramesOutS16, pFramesOutS16); + } + + pFramesOutS16 += pHPF->channels; + pFramesInS16 += pHPF->channels; + } + } else { + MA_ASSERT(MA_FALSE); + return MA_INVALID_OPERATION; /* Should never hit this. */ + } + } + + return MA_SUCCESS; +} + +MA_API ma_uint32 ma_hpf_get_latency(const ma_hpf* pHPF) +{ + if (pHPF == NULL) { + return 0; + } + + return pHPF->hpf2Count*2 + pHPF->hpf1Count; +} + + +/************************************************************************************************************************************************************** + +Band-Pass Filtering + +**************************************************************************************************************************************************************/ +MA_API ma_bpf2_config ma_bpf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, double q) +{ + ma_bpf2_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.cutoffFrequency = cutoffFrequency; + config.q = q; + + /* Q cannot be 0 or else it'll result in a division by 0. In this case just default to 0.707107. */ + if (config.q == 0) { + config.q = 0.707107; + } + + return config; +} + + +static MA_INLINE ma_biquad_config ma_bpf2__get_biquad_config(const ma_bpf2_config* pConfig) +{ + ma_biquad_config bqConfig; + double q; + double w; + double s; + double c; + double a; + + MA_ASSERT(pConfig != NULL); + + q = pConfig->q; + w = 2 * MA_PI_D * pConfig->cutoffFrequency / pConfig->sampleRate; + s = ma_sind(w); + c = ma_cosd(w); + a = s / (2*q); + + bqConfig.b0 = q * a; + bqConfig.b1 = 0; + bqConfig.b2 = -q * a; + bqConfig.a0 = 1 + a; + bqConfig.a1 = -2 * c; + bqConfig.a2 = 1 - a; + + bqConfig.format = pConfig->format; + bqConfig.channels = pConfig->channels; + + return bqConfig; +} + +MA_API ma_result ma_bpf2_get_heap_size(const ma_bpf2_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_biquad_config bqConfig; + bqConfig = ma_bpf2__get_biquad_config(pConfig); + + return ma_biquad_get_heap_size(&bqConfig, pHeapSizeInBytes); +} + +MA_API ma_result ma_bpf2_init_preallocated(const ma_bpf2_config* pConfig, void* pHeap, ma_bpf2* pBPF) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pBPF == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pBPF); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_bpf2__get_biquad_config(pConfig); + result = ma_biquad_init_preallocated(&bqConfig, pHeap, &pBPF->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_bpf2_init(const ma_bpf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_bpf2* pBPF) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_bpf2_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_bpf2_init_preallocated(pConfig, pHeap, pBPF); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pBPF->bq._ownsHeap = MA_TRUE; /* <-- This will cause the biquad to take ownership of the heap and free it when it's uninitialized. */ + return MA_SUCCESS; +} + +MA_API void ma_bpf2_uninit(ma_bpf2* pBPF, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pBPF == NULL) { + return; + } + + ma_biquad_uninit(&pBPF->bq, pAllocationCallbacks); /* <-- This will free the heap allocation. */ +} + +MA_API ma_result ma_bpf2_reinit(const ma_bpf2_config* pConfig, ma_bpf2* pBPF) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pBPF == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_bpf2__get_biquad_config(pConfig); + result = ma_biquad_reinit(&bqConfig, &pBPF->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_bpf2_process_pcm_frame_s16(ma_bpf2* pBPF, ma_int16* pFrameOut, const ma_int16* pFrameIn) +{ + ma_biquad_process_pcm_frame_s16(&pBPF->bq, pFrameOut, pFrameIn); +} + +static MA_INLINE void ma_bpf2_process_pcm_frame_f32(ma_bpf2* pBPF, float* pFrameOut, const float* pFrameIn) +{ + ma_biquad_process_pcm_frame_f32(&pBPF->bq, pFrameOut, pFrameIn); +} + +MA_API ma_result ma_bpf2_process_pcm_frames(ma_bpf2* pBPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pBPF == NULL) { + return MA_INVALID_ARGS; + } + + return ma_biquad_process_pcm_frames(&pBPF->bq, pFramesOut, pFramesIn, frameCount); +} + +MA_API ma_uint32 ma_bpf2_get_latency(const ma_bpf2* pBPF) +{ + if (pBPF == NULL) { + return 0; + } + + return ma_biquad_get_latency(&pBPF->bq); +} + + +MA_API ma_bpf_config ma_bpf_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order) +{ + ma_bpf_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.cutoffFrequency = cutoffFrequency; + config.order = ma_min(order, MA_MAX_FILTER_ORDER); + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t bpf2Offset; +} ma_bpf_heap_layout; + +static ma_result ma_bpf_get_heap_layout(const ma_bpf_config* pConfig, ma_bpf_heap_layout* pHeapLayout) +{ + ma_result result; + ma_uint32 bpf2Count; + ma_uint32 ibpf2; + + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->order > MA_MAX_FILTER_ORDER) { + return MA_INVALID_ARGS; + } + + /* We must have an even number of order. */ + if ((pConfig->order & 0x1) != 0) { + return MA_INVALID_ARGS; + } + + bpf2Count = pConfig->order / 2; + + pHeapLayout->sizeInBytes = 0; + + /* BPF 2 */ + pHeapLayout->bpf2Offset = pHeapLayout->sizeInBytes; + for (ibpf2 = 0; ibpf2 < bpf2Count; ibpf2 += 1) { + size_t bpf2HeapSizeInBytes; + ma_bpf2_config bpf2Config = ma_bpf2_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency, 0.707107); /* <-- The "q" parameter does not matter for the purpose of calculating the heap size. */ + + result = ma_bpf2_get_heap_size(&bpf2Config, &bpf2HeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes += sizeof(ma_bpf2) + bpf2HeapSizeInBytes; + } + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +static ma_result ma_bpf_reinit__internal(const ma_bpf_config* pConfig, void* pHeap, ma_bpf* pBPF, ma_bool32 isNew) +{ + ma_result result; + ma_uint32 bpf2Count; + ma_uint32 ibpf2; + ma_bpf_heap_layout heapLayout; /* Only used if isNew is true. */ + + if (pBPF == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + /* Only supporting f32 and s16. */ + if (pConfig->format != ma_format_f32 && pConfig->format != ma_format_s16) { + return MA_INVALID_ARGS; + } + + /* The format cannot be changed after initialization. */ + if (pBPF->format != ma_format_unknown && pBPF->format != pConfig->format) { + return MA_INVALID_OPERATION; + } + + /* The channel count cannot be changed after initialization. */ + if (pBPF->channels != 0 && pBPF->channels != pConfig->channels) { + return MA_INVALID_OPERATION; + } + + if (pConfig->order > MA_MAX_FILTER_ORDER) { + return MA_INVALID_ARGS; + } + + /* We must have an even number of order. */ + if ((pConfig->order & 0x1) != 0) { + return MA_INVALID_ARGS; + } + + bpf2Count = pConfig->order / 2; + + /* The filter order can't change between reinits. */ + if (!isNew) { + if (pBPF->bpf2Count != bpf2Count) { + return MA_INVALID_OPERATION; + } + } + + if (isNew) { + result = ma_bpf_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pBPF->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pBPF->pBPF2 = (ma_bpf2*)ma_offset_ptr(pHeap, heapLayout.bpf2Offset); + } else { + MA_ZERO_OBJECT(&heapLayout); + } + + for (ibpf2 = 0; ibpf2 < bpf2Count; ibpf2 += 1) { + ma_bpf2_config bpf2Config; + double q; + + /* TODO: Calculate Q to make this a proper Butterworth filter. */ + q = 0.707107; + + bpf2Config = ma_bpf2_config_init(pConfig->format, pConfig->channels, pConfig->sampleRate, pConfig->cutoffFrequency, q); + + if (isNew) { + size_t bpf2HeapSizeInBytes; + + result = ma_bpf2_get_heap_size(&bpf2Config, &bpf2HeapSizeInBytes); + if (result == MA_SUCCESS) { + result = ma_bpf2_init_preallocated(&bpf2Config, ma_offset_ptr(pHeap, heapLayout.bpf2Offset + (sizeof(ma_bpf2) * bpf2Count) + (ibpf2 * bpf2HeapSizeInBytes)), &pBPF->pBPF2[ibpf2]); + } + } else { + result = ma_bpf2_reinit(&bpf2Config, &pBPF->pBPF2[ibpf2]); + } + + if (result != MA_SUCCESS) { + return result; + } + } + + pBPF->bpf2Count = bpf2Count; + pBPF->format = pConfig->format; + pBPF->channels = pConfig->channels; + + return MA_SUCCESS; +} + + +MA_API ma_result ma_bpf_get_heap_size(const ma_bpf_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_bpf_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_bpf_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_bpf_init_preallocated(const ma_bpf_config* pConfig, void* pHeap, ma_bpf* pBPF) +{ + if (pBPF == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pBPF); + + return ma_bpf_reinit__internal(pConfig, pHeap, pBPF, /*isNew*/MA_TRUE); +} + +MA_API ma_result ma_bpf_init(const ma_bpf_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_bpf* pBPF) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_bpf_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_bpf_init_preallocated(pConfig, pHeap, pBPF); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pBPF->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_bpf_uninit(ma_bpf* pBPF, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_uint32 ibpf2; + + if (pBPF == NULL) { + return; + } + + for (ibpf2 = 0; ibpf2 < pBPF->bpf2Count; ibpf2 += 1) { + ma_bpf2_uninit(&pBPF->pBPF2[ibpf2], pAllocationCallbacks); + } + + if (pBPF->_ownsHeap) { + ma_free(pBPF->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_bpf_reinit(const ma_bpf_config* pConfig, ma_bpf* pBPF) +{ + return ma_bpf_reinit__internal(pConfig, NULL, pBPF, /*isNew*/MA_FALSE); +} + +MA_API ma_result ma_bpf_process_pcm_frames(ma_bpf* pBPF, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_result result; + ma_uint32 ibpf2; + + if (pBPF == NULL) { + return MA_INVALID_ARGS; + } + + /* Faster path for in-place. */ + if (pFramesOut == pFramesIn) { + for (ibpf2 = 0; ibpf2 < pBPF->bpf2Count; ibpf2 += 1) { + result = ma_bpf2_process_pcm_frames(&pBPF->pBPF2[ibpf2], pFramesOut, pFramesOut, frameCount); + if (result != MA_SUCCESS) { + return result; + } + } + } + + /* Slightly slower path for copying. */ + if (pFramesOut != pFramesIn) { + ma_uint32 iFrame; + + /* */ if (pBPF->format == ma_format_f32) { + /* */ float* pFramesOutF32 = ( float*)pFramesOut; + const float* pFramesInF32 = (const float*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + MA_COPY_MEMORY(pFramesOutF32, pFramesInF32, ma_get_bytes_per_frame(pBPF->format, pBPF->channels)); + + for (ibpf2 = 0; ibpf2 < pBPF->bpf2Count; ibpf2 += 1) { + ma_bpf2_process_pcm_frame_f32(&pBPF->pBPF2[ibpf2], pFramesOutF32, pFramesOutF32); + } + + pFramesOutF32 += pBPF->channels; + pFramesInF32 += pBPF->channels; + } + } else if (pBPF->format == ma_format_s16) { + /* */ ma_int16* pFramesOutS16 = ( ma_int16*)pFramesOut; + const ma_int16* pFramesInS16 = (const ma_int16*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + MA_COPY_MEMORY(pFramesOutS16, pFramesInS16, ma_get_bytes_per_frame(pBPF->format, pBPF->channels)); + + for (ibpf2 = 0; ibpf2 < pBPF->bpf2Count; ibpf2 += 1) { + ma_bpf2_process_pcm_frame_s16(&pBPF->pBPF2[ibpf2], pFramesOutS16, pFramesOutS16); + } + + pFramesOutS16 += pBPF->channels; + pFramesInS16 += pBPF->channels; + } + } else { + MA_ASSERT(MA_FALSE); + return MA_INVALID_OPERATION; /* Should never hit this. */ + } + } + + return MA_SUCCESS; +} + +MA_API ma_uint32 ma_bpf_get_latency(const ma_bpf* pBPF) +{ + if (pBPF == NULL) { + return 0; + } + + return pBPF->bpf2Count*2; +} + + +/************************************************************************************************************************************************************** + +Notching Filter + +**************************************************************************************************************************************************************/ +MA_API ma_notch2_config ma_notch2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double q, double frequency) +{ + ma_notch2_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.q = q; + config.frequency = frequency; + + if (config.q == 0) { + config.q = 0.707107; + } + + return config; +} + + +static MA_INLINE ma_biquad_config ma_notch2__get_biquad_config(const ma_notch2_config* pConfig) +{ + ma_biquad_config bqConfig; + double q; + double w; + double s; + double c; + double a; + + MA_ASSERT(pConfig != NULL); + + q = pConfig->q; + w = 2 * MA_PI_D * pConfig->frequency / pConfig->sampleRate; + s = ma_sind(w); + c = ma_cosd(w); + a = s / (2*q); + + bqConfig.b0 = 1; + bqConfig.b1 = -2 * c; + bqConfig.b2 = 1; + bqConfig.a0 = 1 + a; + bqConfig.a1 = -2 * c; + bqConfig.a2 = 1 - a; + + bqConfig.format = pConfig->format; + bqConfig.channels = pConfig->channels; + + return bqConfig; +} + +MA_API ma_result ma_notch2_get_heap_size(const ma_notch2_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_biquad_config bqConfig; + bqConfig = ma_notch2__get_biquad_config(pConfig); + + return ma_biquad_get_heap_size(&bqConfig, pHeapSizeInBytes); +} + +MA_API ma_result ma_notch2_init_preallocated(const ma_notch2_config* pConfig, void* pHeap, ma_notch2* pFilter) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pFilter == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pFilter); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_notch2__get_biquad_config(pConfig); + result = ma_biquad_init_preallocated(&bqConfig, pHeap, &pFilter->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_notch2_init(const ma_notch2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_notch2* pFilter) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_notch2_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_notch2_init_preallocated(pConfig, pHeap, pFilter); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pFilter->bq._ownsHeap = MA_TRUE; /* <-- This will cause the biquad to take ownership of the heap and free it when it's uninitialized. */ + return MA_SUCCESS; +} + +MA_API void ma_notch2_uninit(ma_notch2* pFilter, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFilter == NULL) { + return; + } + + ma_biquad_uninit(&pFilter->bq, pAllocationCallbacks); /* <-- This will free the heap allocation. */ +} + +MA_API ma_result ma_notch2_reinit(const ma_notch2_config* pConfig, ma_notch2* pFilter) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pFilter == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_notch2__get_biquad_config(pConfig); + result = ma_biquad_reinit(&bqConfig, &pFilter->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_notch2_process_pcm_frame_s16(ma_notch2* pFilter, ma_int16* pFrameOut, const ma_int16* pFrameIn) +{ + ma_biquad_process_pcm_frame_s16(&pFilter->bq, pFrameOut, pFrameIn); +} + +static MA_INLINE void ma_notch2_process_pcm_frame_f32(ma_notch2* pFilter, float* pFrameOut, const float* pFrameIn) +{ + ma_biquad_process_pcm_frame_f32(&pFilter->bq, pFrameOut, pFrameIn); +} + +MA_API ma_result ma_notch2_process_pcm_frames(ma_notch2* pFilter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pFilter == NULL) { + return MA_INVALID_ARGS; + } + + return ma_biquad_process_pcm_frames(&pFilter->bq, pFramesOut, pFramesIn, frameCount); +} + +MA_API ma_uint32 ma_notch2_get_latency(const ma_notch2* pFilter) +{ + if (pFilter == NULL) { + return 0; + } + + return ma_biquad_get_latency(&pFilter->bq); +} + + + +/************************************************************************************************************************************************************** + +Peaking EQ Filter + +**************************************************************************************************************************************************************/ +MA_API ma_peak2_config ma_peak2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double q, double frequency) +{ + ma_peak2_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.gainDB = gainDB; + config.q = q; + config.frequency = frequency; + + if (config.q == 0) { + config.q = 0.707107; + } + + return config; +} + + +static MA_INLINE ma_biquad_config ma_peak2__get_biquad_config(const ma_peak2_config* pConfig) +{ + ma_biquad_config bqConfig; + double q; + double w; + double s; + double c; + double a; + double A; + + MA_ASSERT(pConfig != NULL); + + q = pConfig->q; + w = 2 * MA_PI_D * pConfig->frequency / pConfig->sampleRate; + s = ma_sind(w); + c = ma_cosd(w); + a = s / (2*q); + A = ma_powd(10, (pConfig->gainDB / 40)); + + bqConfig.b0 = 1 + (a * A); + bqConfig.b1 = -2 * c; + bqConfig.b2 = 1 - (a * A); + bqConfig.a0 = 1 + (a / A); + bqConfig.a1 = -2 * c; + bqConfig.a2 = 1 - (a / A); + + bqConfig.format = pConfig->format; + bqConfig.channels = pConfig->channels; + + return bqConfig; +} + +MA_API ma_result ma_peak2_get_heap_size(const ma_peak2_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_biquad_config bqConfig; + bqConfig = ma_peak2__get_biquad_config(pConfig); + + return ma_biquad_get_heap_size(&bqConfig, pHeapSizeInBytes); +} + +MA_API ma_result ma_peak2_init_preallocated(const ma_peak2_config* pConfig, void* pHeap, ma_peak2* pFilter) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pFilter == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pFilter); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_peak2__get_biquad_config(pConfig); + result = ma_biquad_init_preallocated(&bqConfig, pHeap, &pFilter->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_peak2_init(const ma_peak2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_peak2* pFilter) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_peak2_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_peak2_init_preallocated(pConfig, pHeap, pFilter); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pFilter->bq._ownsHeap = MA_TRUE; /* <-- This will cause the biquad to take ownership of the heap and free it when it's uninitialized. */ + return MA_SUCCESS; +} + +MA_API void ma_peak2_uninit(ma_peak2* pFilter, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFilter == NULL) { + return; + } + + ma_biquad_uninit(&pFilter->bq, pAllocationCallbacks); /* <-- This will free the heap allocation. */ +} + +MA_API ma_result ma_peak2_reinit(const ma_peak2_config* pConfig, ma_peak2* pFilter) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pFilter == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_peak2__get_biquad_config(pConfig); + result = ma_biquad_reinit(&bqConfig, &pFilter->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_peak2_process_pcm_frame_s16(ma_peak2* pFilter, ma_int16* pFrameOut, const ma_int16* pFrameIn) +{ + ma_biquad_process_pcm_frame_s16(&pFilter->bq, pFrameOut, pFrameIn); +} + +static MA_INLINE void ma_peak2_process_pcm_frame_f32(ma_peak2* pFilter, float* pFrameOut, const float* pFrameIn) +{ + ma_biquad_process_pcm_frame_f32(&pFilter->bq, pFrameOut, pFrameIn); +} + +MA_API ma_result ma_peak2_process_pcm_frames(ma_peak2* pFilter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pFilter == NULL) { + return MA_INVALID_ARGS; + } + + return ma_biquad_process_pcm_frames(&pFilter->bq, pFramesOut, pFramesIn, frameCount); +} + +MA_API ma_uint32 ma_peak2_get_latency(const ma_peak2* pFilter) +{ + if (pFilter == NULL) { + return 0; + } + + return ma_biquad_get_latency(&pFilter->bq); +} + + +/************************************************************************************************************************************************************** + +Low Shelf Filter + +**************************************************************************************************************************************************************/ +MA_API ma_loshelf2_config ma_loshelf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double shelfSlope, double frequency) +{ + ma_loshelf2_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.gainDB = gainDB; + config.shelfSlope = shelfSlope; + config.frequency = frequency; + + return config; +} + + +static MA_INLINE ma_biquad_config ma_loshelf2__get_biquad_config(const ma_loshelf2_config* pConfig) +{ + ma_biquad_config bqConfig; + double w; + double s; + double c; + double A; + double S; + double a; + double sqrtA; + + MA_ASSERT(pConfig != NULL); + + w = 2 * MA_PI_D * pConfig->frequency / pConfig->sampleRate; + s = ma_sind(w); + c = ma_cosd(w); + A = ma_powd(10, (pConfig->gainDB / 40)); + S = pConfig->shelfSlope; + a = s/2 * ma_sqrtd((A + 1/A) * (1/S - 1) + 2); + sqrtA = 2*ma_sqrtd(A)*a; + + bqConfig.b0 = A * ((A + 1) - (A - 1)*c + sqrtA); + bqConfig.b1 = 2 * A * ((A - 1) - (A + 1)*c); + bqConfig.b2 = A * ((A + 1) - (A - 1)*c - sqrtA); + bqConfig.a0 = (A + 1) + (A - 1)*c + sqrtA; + bqConfig.a1 = -2 * ((A - 1) + (A + 1)*c); + bqConfig.a2 = (A + 1) + (A - 1)*c - sqrtA; + + bqConfig.format = pConfig->format; + bqConfig.channels = pConfig->channels; + + return bqConfig; +} + +MA_API ma_result ma_loshelf2_get_heap_size(const ma_loshelf2_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_biquad_config bqConfig; + bqConfig = ma_loshelf2__get_biquad_config(pConfig); + + return ma_biquad_get_heap_size(&bqConfig, pHeapSizeInBytes); +} + +MA_API ma_result ma_loshelf2_init_preallocated(const ma_loshelf2_config* pConfig, void* pHeap, ma_loshelf2* pFilter) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pFilter == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pFilter); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_loshelf2__get_biquad_config(pConfig); + result = ma_biquad_init_preallocated(&bqConfig, pHeap, &pFilter->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_loshelf2_init(const ma_loshelf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_loshelf2* pFilter) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_loshelf2_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_loshelf2_init_preallocated(pConfig, pHeap, pFilter); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pFilter->bq._ownsHeap = MA_TRUE; /* <-- This will cause the biquad to take ownership of the heap and free it when it's uninitialized. */ + return MA_SUCCESS; +} + +MA_API void ma_loshelf2_uninit(ma_loshelf2* pFilter, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFilter == NULL) { + return; + } + + ma_biquad_uninit(&pFilter->bq, pAllocationCallbacks); /* <-- This will free the heap allocation. */ +} + +MA_API ma_result ma_loshelf2_reinit(const ma_loshelf2_config* pConfig, ma_loshelf2* pFilter) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pFilter == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_loshelf2__get_biquad_config(pConfig); + result = ma_biquad_reinit(&bqConfig, &pFilter->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_loshelf2_process_pcm_frame_s16(ma_loshelf2* pFilter, ma_int16* pFrameOut, const ma_int16* pFrameIn) +{ + ma_biquad_process_pcm_frame_s16(&pFilter->bq, pFrameOut, pFrameIn); +} + +static MA_INLINE void ma_loshelf2_process_pcm_frame_f32(ma_loshelf2* pFilter, float* pFrameOut, const float* pFrameIn) +{ + ma_biquad_process_pcm_frame_f32(&pFilter->bq, pFrameOut, pFrameIn); +} + +MA_API ma_result ma_loshelf2_process_pcm_frames(ma_loshelf2* pFilter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pFilter == NULL) { + return MA_INVALID_ARGS; + } + + return ma_biquad_process_pcm_frames(&pFilter->bq, pFramesOut, pFramesIn, frameCount); +} + +MA_API ma_uint32 ma_loshelf2_get_latency(const ma_loshelf2* pFilter) +{ + if (pFilter == NULL) { + return 0; + } + + return ma_biquad_get_latency(&pFilter->bq); +} + + +/************************************************************************************************************************************************************** + +High Shelf Filter + +**************************************************************************************************************************************************************/ +MA_API ma_hishelf2_config ma_hishelf2_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double shelfSlope, double frequency) +{ + ma_hishelf2_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.gainDB = gainDB; + config.shelfSlope = shelfSlope; + config.frequency = frequency; + + return config; +} + + +static MA_INLINE ma_biquad_config ma_hishelf2__get_biquad_config(const ma_hishelf2_config* pConfig) +{ + ma_biquad_config bqConfig; + double w; + double s; + double c; + double A; + double S; + double a; + double sqrtA; + + MA_ASSERT(pConfig != NULL); + + w = 2 * MA_PI_D * pConfig->frequency / pConfig->sampleRate; + s = ma_sind(w); + c = ma_cosd(w); + A = ma_powd(10, (pConfig->gainDB / 40)); + S = pConfig->shelfSlope; + a = s/2 * ma_sqrtd((A + 1/A) * (1/S - 1) + 2); + sqrtA = 2*ma_sqrtd(A)*a; + + bqConfig.b0 = A * ((A + 1) + (A - 1)*c + sqrtA); + bqConfig.b1 = -2 * A * ((A - 1) + (A + 1)*c); + bqConfig.b2 = A * ((A + 1) + (A - 1)*c - sqrtA); + bqConfig.a0 = (A + 1) - (A - 1)*c + sqrtA; + bqConfig.a1 = 2 * ((A - 1) - (A + 1)*c); + bqConfig.a2 = (A + 1) - (A - 1)*c - sqrtA; + + bqConfig.format = pConfig->format; + bqConfig.channels = pConfig->channels; + + return bqConfig; +} + +MA_API ma_result ma_hishelf2_get_heap_size(const ma_hishelf2_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_biquad_config bqConfig; + bqConfig = ma_hishelf2__get_biquad_config(pConfig); + + return ma_biquad_get_heap_size(&bqConfig, pHeapSizeInBytes); +} + +MA_API ma_result ma_hishelf2_init_preallocated(const ma_hishelf2_config* pConfig, void* pHeap, ma_hishelf2* pFilter) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pFilter == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pFilter); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_hishelf2__get_biquad_config(pConfig); + result = ma_biquad_init_preallocated(&bqConfig, pHeap, &pFilter->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_hishelf2_init(const ma_hishelf2_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hishelf2* pFilter) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_hishelf2_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_hishelf2_init_preallocated(pConfig, pHeap, pFilter); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pFilter->bq._ownsHeap = MA_TRUE; /* <-- This will cause the biquad to take ownership of the heap and free it when it's uninitialized. */ + return MA_SUCCESS; +} + +MA_API void ma_hishelf2_uninit(ma_hishelf2* pFilter, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFilter == NULL) { + return; + } + + ma_biquad_uninit(&pFilter->bq, pAllocationCallbacks); /* <-- This will free the heap allocation. */ +} + +MA_API ma_result ma_hishelf2_reinit(const ma_hishelf2_config* pConfig, ma_hishelf2* pFilter) +{ + ma_result result; + ma_biquad_config bqConfig; + + if (pFilter == NULL || pConfig == NULL) { + return MA_INVALID_ARGS; + } + + bqConfig = ma_hishelf2__get_biquad_config(pConfig); + result = ma_biquad_reinit(&bqConfig, &pFilter->bq); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +static MA_INLINE void ma_hishelf2_process_pcm_frame_s16(ma_hishelf2* pFilter, ma_int16* pFrameOut, const ma_int16* pFrameIn) +{ + ma_biquad_process_pcm_frame_s16(&pFilter->bq, pFrameOut, pFrameIn); +} + +static MA_INLINE void ma_hishelf2_process_pcm_frame_f32(ma_hishelf2* pFilter, float* pFrameOut, const float* pFrameIn) +{ + ma_biquad_process_pcm_frame_f32(&pFilter->bq, pFrameOut, pFrameIn); +} + +MA_API ma_result ma_hishelf2_process_pcm_frames(ma_hishelf2* pFilter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pFilter == NULL) { + return MA_INVALID_ARGS; + } + + return ma_biquad_process_pcm_frames(&pFilter->bq, pFramesOut, pFramesIn, frameCount); +} + +MA_API ma_uint32 ma_hishelf2_get_latency(const ma_hishelf2* pFilter) +{ + if (pFilter == NULL) { + return 0; + } + + return ma_biquad_get_latency(&pFilter->bq); +} + + + +/* +Delay +*/ +MA_API ma_delay_config ma_delay_config_init(ma_uint32 channels, ma_uint32 sampleRate, ma_uint32 delayInFrames, float decay) +{ + ma_delay_config config; + + MA_ZERO_OBJECT(&config); + config.channels = channels; + config.sampleRate = sampleRate; + config.delayInFrames = delayInFrames; + config.delayStart = (decay == 0) ? MA_TRUE : MA_FALSE; /* Delay the start if it looks like we're not configuring an echo. */ + config.wet = 1; + config.dry = 1; + config.decay = decay; + + return config; +} + + +MA_API ma_result ma_delay_init(const ma_delay_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_delay* pDelay) +{ + if (pDelay == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDelay); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->decay < 0 || pConfig->decay > 1) { + return MA_INVALID_ARGS; + } + + pDelay->config = *pConfig; + pDelay->bufferSizeInFrames = pConfig->delayInFrames; + pDelay->cursor = 0; + + pDelay->pBuffer = (float*)ma_malloc((size_t)(pDelay->bufferSizeInFrames * ma_get_bytes_per_frame(ma_format_f32, pConfig->channels)), pAllocationCallbacks); + if (pDelay->pBuffer == NULL) { + return MA_OUT_OF_MEMORY; + } + + ma_silence_pcm_frames(pDelay->pBuffer, pDelay->bufferSizeInFrames, ma_format_f32, pConfig->channels); + + return MA_SUCCESS; +} + +MA_API void ma_delay_uninit(ma_delay* pDelay, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pDelay == NULL) { + return; + } + + ma_free(pDelay->pBuffer, pAllocationCallbacks); +} + +MA_API ma_result ma_delay_process_pcm_frames(ma_delay* pDelay, void* pFramesOut, const void* pFramesIn, ma_uint32 frameCount) +{ + ma_uint32 iFrame; + ma_uint32 iChannel; + float* pFramesOutF32 = (float*)pFramesOut; + const float* pFramesInF32 = (const float*)pFramesIn; + + if (pDelay == NULL || pFramesOut == NULL || pFramesIn == NULL) { + return MA_INVALID_ARGS; + } + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < pDelay->config.channels; iChannel += 1) { + ma_uint32 iBuffer = (pDelay->cursor * pDelay->config.channels) + iChannel; + + if (pDelay->config.delayStart) { + /* Delayed start. */ + + /* Read */ + pFramesOutF32[iChannel] = pDelay->pBuffer[iBuffer] * pDelay->config.wet; + + /* Feedback */ + pDelay->pBuffer[iBuffer] = (pDelay->pBuffer[iBuffer] * pDelay->config.decay) + (pFramesInF32[iChannel] * pDelay->config.dry); + } else { + /* Immediate start */ + + /* Feedback */ + pDelay->pBuffer[iBuffer] = (pDelay->pBuffer[iBuffer] * pDelay->config.decay) + (pFramesInF32[iChannel] * pDelay->config.dry); + + /* Read */ + pFramesOutF32[iChannel] = pDelay->pBuffer[iBuffer] * pDelay->config.wet; + } + } + + pDelay->cursor = (pDelay->cursor + 1) % pDelay->bufferSizeInFrames; + + pFramesOutF32 += pDelay->config.channels; + pFramesInF32 += pDelay->config.channels; + } + + return MA_SUCCESS; +} + +MA_API void ma_delay_set_wet(ma_delay* pDelay, float value) +{ + if (pDelay == NULL) { + return; + } + + pDelay->config.wet = value; +} + +MA_API float ma_delay_get_wet(const ma_delay* pDelay) +{ + if (pDelay == NULL) { + return 0; + } + + return pDelay->config.wet; +} + +MA_API void ma_delay_set_dry(ma_delay* pDelay, float value) +{ + if (pDelay == NULL) { + return; + } + + pDelay->config.dry = value; +} + +MA_API float ma_delay_get_dry(const ma_delay* pDelay) +{ + if (pDelay == NULL) { + return 0; + } + + return pDelay->config.dry; +} + +MA_API void ma_delay_set_decay(ma_delay* pDelay, float value) +{ + if (pDelay == NULL) { + return; + } + + pDelay->config.decay = value; +} + +MA_API float ma_delay_get_decay(const ma_delay* pDelay) +{ + if (pDelay == NULL) { + return 0; + } + + return pDelay->config.decay; +} + + +MA_API ma_gainer_config ma_gainer_config_init(ma_uint32 channels, ma_uint32 smoothTimeInFrames) +{ + ma_gainer_config config; + + MA_ZERO_OBJECT(&config); + config.channels = channels; + config.smoothTimeInFrames = smoothTimeInFrames; + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t oldGainsOffset; + size_t newGainsOffset; +} ma_gainer_heap_layout; + +static ma_result ma_gainer_get_heap_layout(const ma_gainer_config* pConfig, ma_gainer_heap_layout* pHeapLayout) +{ + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channels == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* Old gains. */ + pHeapLayout->oldGainsOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(float) * pConfig->channels; + + /* New gains. */ + pHeapLayout->newGainsOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(float) * pConfig->channels; + + /* Alignment. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + + +MA_API ma_result ma_gainer_get_heap_size(const ma_gainer_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_gainer_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_gainer_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + + +MA_API ma_result ma_gainer_init_preallocated(const ma_gainer_config* pConfig, void* pHeap, ma_gainer* pGainer) +{ + ma_result result; + ma_gainer_heap_layout heapLayout; + ma_uint32 iChannel; + + if (pGainer == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pGainer); + + if (pConfig == NULL || pHeap == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_gainer_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pGainer->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pGainer->pOldGains = (float*)ma_offset_ptr(pHeap, heapLayout.oldGainsOffset); + pGainer->pNewGains = (float*)ma_offset_ptr(pHeap, heapLayout.newGainsOffset); + pGainer->masterVolume = 1; + + pGainer->config = *pConfig; + pGainer->t = (ma_uint32)-1; /* No interpolation by default. */ + + for (iChannel = 0; iChannel < pConfig->channels; iChannel += 1) { + pGainer->pOldGains[iChannel] = 1; + pGainer->pNewGains[iChannel] = 1; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_gainer_init(const ma_gainer_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_gainer* pGainer) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_gainer_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the size of the heap allocation. */ + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_gainer_init_preallocated(pConfig, pHeap, pGainer); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pGainer->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_gainer_uninit(ma_gainer* pGainer, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pGainer == NULL) { + return; + } + + if (pGainer->_ownsHeap) { + ma_free(pGainer->_pHeap, pAllocationCallbacks); + } +} + +static float ma_gainer_calculate_current_gain(const ma_gainer* pGainer, ma_uint32 channel) +{ + float a = (float)pGainer->t / pGainer->config.smoothTimeInFrames; + return ma_mix_f32_fast(pGainer->pOldGains[channel], pGainer->pNewGains[channel], a); +} + +static /*__attribute__((noinline))*/ ma_result ma_gainer_process_pcm_frames_internal(ma_gainer * pGainer, void* MA_RESTRICT pFramesOut, const void* MA_RESTRICT pFramesIn, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint32 iChannel; + ma_uint64 interpolatedFrameCount; + + MA_ASSERT(pGainer != NULL); + + /* + We don't necessarily need to apply a linear interpolation for the entire frameCount frames. When + linear interpolation is not needed we can do a simple volume adjustment which will be more + efficient than a lerp with an alpha value of 1. + + To do this, all we need to do is determine how many frames need to have a lerp applied. Then we + just process that number of frames with linear interpolation. After that we run on an optimized + path which just applies the new gains without a lerp. + */ + if (pGainer->t >= pGainer->config.smoothTimeInFrames) { + interpolatedFrameCount = 0; + } else { + interpolatedFrameCount = pGainer->t - pGainer->config.smoothTimeInFrames; + if (interpolatedFrameCount > frameCount) { + interpolatedFrameCount = frameCount; + } + } + + /* + Start off with our interpolated frames. When we do this, we'll adjust frameCount and our pointers + so that the fast path can work naturally without consideration of the interpolated path. + */ + if (interpolatedFrameCount > 0) { + /* We can allow the input and output buffers to be null in which case we'll just update the internal timer. */ + if (pFramesOut != NULL && pFramesIn != NULL) { + /* + All we're really doing here is moving the old gains towards the new gains. We don't want to + be modifying the gains inside the ma_gainer object because that will break things. Instead + we can make a copy here on the stack. For extreme channel counts we can fall back to a slower + implementation which just uses a standard lerp. + */ + float* pFramesOutF32 = (float*)pFramesOut; + const float* pFramesInF32 = (const float*)pFramesIn; + float a = (float)pGainer->t / pGainer->config.smoothTimeInFrames; + float d = 1.0f / pGainer->config.smoothTimeInFrames; + + if (pGainer->config.channels <= 32) { + float pRunningGain[32]; + float pRunningGainDelta[32]; /* Could this be heap-allocated as part of the ma_gainer object? */ + + /* Initialize the running gain. */ + for (iChannel = 0; iChannel < pGainer->config.channels; iChannel += 1) { + float t = (pGainer->pNewGains[iChannel] - pGainer->pOldGains[iChannel]) * pGainer->masterVolume; + pRunningGainDelta[iChannel] = t * d; + pRunningGain[iChannel] = (pGainer->pOldGains[iChannel] * pGainer->masterVolume) + (t * a); + } + + iFrame = 0; + + /* Optimized paths for common channel counts. This is mostly just experimenting with some SIMD ideas. It's not necessarily final. */ + if (pGainer->config.channels == 2) { + #if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + ma_uint64 unrolledLoopCount = interpolatedFrameCount >> 1; + + /* Expand some arrays so we can have a clean SIMD loop below. */ + __m128 runningGainDelta0 = _mm_set_ps(pRunningGainDelta[1], pRunningGainDelta[0], pRunningGainDelta[1], pRunningGainDelta[0]); + __m128 runningGain0 = _mm_set_ps(pRunningGain[1] + pRunningGainDelta[1], pRunningGain[0] + pRunningGainDelta[0], pRunningGain[1], pRunningGain[0]); + + for (; iFrame < unrolledLoopCount; iFrame += 1) { + _mm_storeu_ps(&pFramesOutF32[iFrame*4 + 0], _mm_mul_ps(_mm_loadu_ps(&pFramesInF32[iFrame*4 + 0]), runningGain0)); + runningGain0 = _mm_add_ps(runningGain0, runningGainDelta0); + } + + iFrame = unrolledLoopCount << 1; + } else + #endif + { + /* + Two different scalar implementations here. Clang (and I assume GCC) will vectorize + both of these, but the bottom version results in a nicer vectorization with less + instructions emitted. The problem, however, is that the bottom version runs slower + when compiled with MSVC. The top version will be partially vectorized by MSVC. + */ + #if defined(_MSC_VER) && !defined(__clang__) + ma_uint64 unrolledLoopCount = interpolatedFrameCount >> 1; + + /* Expand some arrays so we can have a clean 4x SIMD operation in the loop. */ + pRunningGainDelta[2] = pRunningGainDelta[0]; + pRunningGainDelta[3] = pRunningGainDelta[1]; + pRunningGain[2] = pRunningGain[0] + pRunningGainDelta[0]; + pRunningGain[3] = pRunningGain[1] + pRunningGainDelta[1]; + + for (; iFrame < unrolledLoopCount; iFrame += 1) { + pFramesOutF32[iFrame*4 + 0] = pFramesInF32[iFrame*4 + 0] * pRunningGain[0]; + pFramesOutF32[iFrame*4 + 1] = pFramesInF32[iFrame*4 + 1] * pRunningGain[1]; + pFramesOutF32[iFrame*4 + 2] = pFramesInF32[iFrame*4 + 2] * pRunningGain[2]; + pFramesOutF32[iFrame*4 + 3] = pFramesInF32[iFrame*4 + 3] * pRunningGain[3]; + + /* Move the running gain forward towards the new gain. */ + pRunningGain[0] += pRunningGainDelta[0]; + pRunningGain[1] += pRunningGainDelta[1]; + pRunningGain[2] += pRunningGainDelta[2]; + pRunningGain[3] += pRunningGainDelta[3]; + } + + iFrame = unrolledLoopCount << 1; + #else + for (; iFrame < interpolatedFrameCount; iFrame += 1) { + for (iChannel = 0; iChannel < 2; iChannel += 1) { + pFramesOutF32[iFrame*2 + iChannel] = pFramesInF32[iFrame*2 + iChannel] * pRunningGain[iChannel]; + } + + for (iChannel = 0; iChannel < 2; iChannel += 1) { + pRunningGain[iChannel] += pRunningGainDelta[iChannel]; + } + } + #endif + } + } else if (pGainer->config.channels == 6) { + #if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + /* + For 6 channels things are a bit more complicated because 6 isn't cleanly divisible by 4. We need to do 2 frames + at a time, meaning we'll be doing 12 samples in a group. Like the stereo case we'll need to expand some arrays + so we can do clean 4x SIMD operations. + */ + ma_uint64 unrolledLoopCount = interpolatedFrameCount >> 1; + + /* Expand some arrays so we can have a clean SIMD loop below. */ + __m128 runningGainDelta0 = _mm_set_ps(pRunningGainDelta[3], pRunningGainDelta[2], pRunningGainDelta[1], pRunningGainDelta[0]); + __m128 runningGainDelta1 = _mm_set_ps(pRunningGainDelta[1], pRunningGainDelta[0], pRunningGainDelta[5], pRunningGainDelta[4]); + __m128 runningGainDelta2 = _mm_set_ps(pRunningGainDelta[5], pRunningGainDelta[4], pRunningGainDelta[3], pRunningGainDelta[2]); + + __m128 runningGain0 = _mm_set_ps(pRunningGain[3], pRunningGain[2], pRunningGain[1], pRunningGain[0]); + __m128 runningGain1 = _mm_set_ps(pRunningGain[1] + pRunningGainDelta[1], pRunningGain[0] + pRunningGainDelta[0], pRunningGain[5], pRunningGain[4]); + __m128 runningGain2 = _mm_set_ps(pRunningGain[5] + pRunningGainDelta[5], pRunningGain[4] + pRunningGainDelta[4], pRunningGain[3] + pRunningGainDelta[3], pRunningGain[2] + pRunningGainDelta[2]); + + for (; iFrame < unrolledLoopCount; iFrame += 1) { + _mm_storeu_ps(&pFramesOutF32[iFrame*12 + 0], _mm_mul_ps(_mm_loadu_ps(&pFramesInF32[iFrame*12 + 0]), runningGain0)); + _mm_storeu_ps(&pFramesOutF32[iFrame*12 + 4], _mm_mul_ps(_mm_loadu_ps(&pFramesInF32[iFrame*12 + 4]), runningGain1)); + _mm_storeu_ps(&pFramesOutF32[iFrame*12 + 8], _mm_mul_ps(_mm_loadu_ps(&pFramesInF32[iFrame*12 + 8]), runningGain2)); + + runningGain0 = _mm_add_ps(runningGain0, runningGainDelta0); + runningGain1 = _mm_add_ps(runningGain1, runningGainDelta1); + runningGain2 = _mm_add_ps(runningGain2, runningGainDelta2); + } + + iFrame = unrolledLoopCount << 1; + } else + #endif + { + for (; iFrame < interpolatedFrameCount; iFrame += 1) { + for (iChannel = 0; iChannel < 6; iChannel += 1) { + pFramesOutF32[iFrame*6 + iChannel] = pFramesInF32[iFrame*6 + iChannel] * pRunningGain[iChannel]; + } + + /* Move the running gain forward towards the new gain. */ + for (iChannel = 0; iChannel < 6; iChannel += 1) { + pRunningGain[iChannel] += pRunningGainDelta[iChannel]; + } + } + } + } else if (pGainer->config.channels == 8) { + /* For 8 channels we can just go over frame by frame and do all eight channels as 2 separate 4x SIMD operations. */ + #if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + __m128 runningGainDelta0 = _mm_loadu_ps(&pRunningGainDelta[0]); + __m128 runningGainDelta1 = _mm_loadu_ps(&pRunningGainDelta[4]); + __m128 runningGain0 = _mm_loadu_ps(&pRunningGain[0]); + __m128 runningGain1 = _mm_loadu_ps(&pRunningGain[4]); + + for (; iFrame < interpolatedFrameCount; iFrame += 1) { + _mm_storeu_ps(&pFramesOutF32[iFrame*8 + 0], _mm_mul_ps(_mm_loadu_ps(&pFramesInF32[iFrame*8 + 0]), runningGain0)); + _mm_storeu_ps(&pFramesOutF32[iFrame*8 + 4], _mm_mul_ps(_mm_loadu_ps(&pFramesInF32[iFrame*8 + 4]), runningGain1)); + + runningGain0 = _mm_add_ps(runningGain0, runningGainDelta0); + runningGain1 = _mm_add_ps(runningGain1, runningGainDelta1); + } + } else + #endif + { + /* This is crafted so that it auto-vectorizes when compiled with Clang. */ + for (; iFrame < interpolatedFrameCount; iFrame += 1) { + for (iChannel = 0; iChannel < 8; iChannel += 1) { + pFramesOutF32[iFrame*8 + iChannel] = pFramesInF32[iFrame*8 + iChannel] * pRunningGain[iChannel]; + } + + /* Move the running gain forward towards the new gain. */ + for (iChannel = 0; iChannel < 8; iChannel += 1) { + pRunningGain[iChannel] += pRunningGainDelta[iChannel]; + } + } + } + } + + for (; iFrame < interpolatedFrameCount; iFrame += 1) { + for (iChannel = 0; iChannel < pGainer->config.channels; iChannel += 1) { + pFramesOutF32[iFrame*pGainer->config.channels + iChannel] = pFramesInF32[iFrame*pGainer->config.channels + iChannel] * pRunningGain[iChannel]; + pRunningGain[iChannel] += pRunningGainDelta[iChannel]; + } + } + } else { + /* Slower path for extreme channel counts where we can't fit enough on the stack. We could also move this to the heap as part of the ma_gainer object which might even be better since it'll only be updated when the gains actually change. */ + for (iFrame = 0; iFrame < interpolatedFrameCount; iFrame += 1) { + for (iChannel = 0; iChannel < pGainer->config.channels; iChannel += 1) { + pFramesOutF32[iFrame*pGainer->config.channels + iChannel] = pFramesInF32[iFrame*pGainer->config.channels + iChannel] * ma_mix_f32_fast(pGainer->pOldGains[iChannel], pGainer->pNewGains[iChannel], a) * pGainer->masterVolume; + } + + a += d; + } + } + } + + /* Make sure the timer is updated. */ + pGainer->t = (ma_uint32)ma_min(pGainer->t + interpolatedFrameCount, pGainer->config.smoothTimeInFrames); + + /* Adjust our arguments so the next part can work normally. */ + frameCount -= interpolatedFrameCount; + pFramesOut = ma_offset_ptr(pFramesOut, interpolatedFrameCount * sizeof(float)); + pFramesIn = ma_offset_ptr(pFramesIn, interpolatedFrameCount * sizeof(float)); + } + + /* All we need to do here is apply the new gains using an optimized path. */ + if (pFramesOut != NULL && pFramesIn != NULL) { + if (pGainer->config.channels <= 32) { + float gains[32]; + for (iChannel = 0; iChannel < pGainer->config.channels; iChannel += 1) { + gains[iChannel] = pGainer->pNewGains[iChannel] * pGainer->masterVolume; + } + + ma_copy_and_apply_volume_factor_per_channel_f32((float*)pFramesOut, (const float*)pFramesIn, frameCount, pGainer->config.channels, gains); + } else { + /* Slow path. Too many channels to fit on the stack. Need to apply a master volume as a separate path. */ + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < pGainer->config.channels; iChannel += 1) { + ((float*)pFramesOut)[iFrame*pGainer->config.channels + iChannel] = ((const float*)pFramesIn)[iFrame*pGainer->config.channels + iChannel] * pGainer->pNewGains[iChannel] * pGainer->masterVolume; + } + } + } + } + + /* Now that some frames have been processed we need to make sure future changes to the gain are interpolated. */ + if (pGainer->t == (ma_uint32)-1) { + pGainer->t = (ma_uint32)ma_min(pGainer->config.smoothTimeInFrames, frameCount); + } + +#if 0 + if (pGainer->t >= pGainer->config.smoothTimeInFrames) { + /* Fast path. No gain calculation required. */ + ma_copy_and_apply_volume_factor_per_channel_f32(pFramesOutF32, pFramesInF32, frameCount, pGainer->config.channels, pGainer->pNewGains); + ma_apply_volume_factor_f32(pFramesOutF32, frameCount * pGainer->config.channels, pGainer->masterVolume); + + /* Now that some frames have been processed we need to make sure future changes to the gain are interpolated. */ + if (pGainer->t == (ma_uint32)-1) { + pGainer->t = pGainer->config.smoothTimeInFrames; + } + } else { + /* Slow path. Need to interpolate the gain for each channel individually. */ + + /* We can allow the input and output buffers to be null in which case we'll just update the internal timer. */ + if (pFramesOut != NULL && pFramesIn != NULL) { + float a = (float)pGainer->t / pGainer->config.smoothTimeInFrames; + float d = 1.0f / pGainer->config.smoothTimeInFrames; + ma_uint32 channelCount = pGainer->config.channels; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channelCount; iChannel += 1) { + pFramesOutF32[iChannel] = pFramesInF32[iChannel] * ma_mix_f32_fast(pGainer->pOldGains[iChannel], pGainer->pNewGains[iChannel], a) * pGainer->masterVolume; + } + + pFramesOutF32 += channelCount; + pFramesInF32 += channelCount; + + a += d; + if (a > 1) { + a = 1; + } + } + } + + pGainer->t = (ma_uint32)ma_min(pGainer->t + frameCount, pGainer->config.smoothTimeInFrames); + + #if 0 /* Reference implementation. */ + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + /* We can allow the input and output buffers to be null in which case we'll just update the internal timer. */ + if (pFramesOut != NULL && pFramesIn != NULL) { + for (iChannel = 0; iChannel < pGainer->config.channels; iChannel += 1) { + pFramesOutF32[iFrame * pGainer->config.channels + iChannel] = pFramesInF32[iFrame * pGainer->config.channels + iChannel] * ma_gainer_calculate_current_gain(pGainer, iChannel) * pGainer->masterVolume; + } + } + + /* Move interpolation time forward, but don't go beyond our smoothing time. */ + pGainer->t = ma_min(pGainer->t + 1, pGainer->config.smoothTimeInFrames); + } + #endif + } +#endif + + return MA_SUCCESS; +} + +MA_API ma_result ma_gainer_process_pcm_frames(ma_gainer* pGainer, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pGainer == NULL) { + return MA_INVALID_ARGS; + } + + /* + ma_gainer_process_pcm_frames_internal() marks pFramesOut and pFramesIn with MA_RESTRICT which + helps with auto-vectorization. + */ + return ma_gainer_process_pcm_frames_internal(pGainer, pFramesOut, pFramesIn, frameCount); +} + +static void ma_gainer_set_gain_by_index(ma_gainer* pGainer, float newGain, ma_uint32 iChannel) +{ + pGainer->pOldGains[iChannel] = ma_gainer_calculate_current_gain(pGainer, iChannel); + pGainer->pNewGains[iChannel] = newGain; +} + +static void ma_gainer_reset_smoothing_time(ma_gainer* pGainer) +{ + if (pGainer->t == (ma_uint32)-1) { + pGainer->t = pGainer->config.smoothTimeInFrames; /* No smoothing required for initial gains setting. */ + } else { + pGainer->t = 0; + } +} + +MA_API ma_result ma_gainer_set_gain(ma_gainer* pGainer, float newGain) +{ + ma_uint32 iChannel; + + if (pGainer == NULL) { + return MA_INVALID_ARGS; + } + + for (iChannel = 0; iChannel < pGainer->config.channels; iChannel += 1) { + ma_gainer_set_gain_by_index(pGainer, newGain, iChannel); + } + + /* The smoothing time needs to be reset to ensure we always interpolate by the configured smoothing time, but only if it's not the first setting. */ + ma_gainer_reset_smoothing_time(pGainer); + + return MA_SUCCESS; +} + +MA_API ma_result ma_gainer_set_gains(ma_gainer* pGainer, float* pNewGains) +{ + ma_uint32 iChannel; + + if (pGainer == NULL || pNewGains == NULL) { + return MA_INVALID_ARGS; + } + + for (iChannel = 0; iChannel < pGainer->config.channels; iChannel += 1) { + ma_gainer_set_gain_by_index(pGainer, pNewGains[iChannel], iChannel); + } + + /* The smoothing time needs to be reset to ensure we always interpolate by the configured smoothing time, but only if it's not the first setting. */ + ma_gainer_reset_smoothing_time(pGainer); + + return MA_SUCCESS; +} + +MA_API ma_result ma_gainer_set_master_volume(ma_gainer* pGainer, float volume) +{ + if (pGainer == NULL) { + return MA_INVALID_ARGS; + } + + pGainer->masterVolume = volume; + + return MA_SUCCESS; +} + +MA_API ma_result ma_gainer_get_master_volume(const ma_gainer* pGainer, float* pVolume) +{ + if (pGainer == NULL || pVolume == NULL) { + return MA_INVALID_ARGS; + } + + *pVolume = pGainer->masterVolume; + + return MA_SUCCESS; +} + + +MA_API ma_panner_config ma_panner_config_init(ma_format format, ma_uint32 channels) +{ + ma_panner_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.mode = ma_pan_mode_balance; /* Set to balancing mode by default because it's consistent with other audio engines and most likely what the caller is expecting. */ + config.pan = 0; + + return config; +} + + +MA_API ma_result ma_panner_init(const ma_panner_config* pConfig, ma_panner* pPanner) +{ + if (pPanner == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pPanner); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + pPanner->format = pConfig->format; + pPanner->channels = pConfig->channels; + pPanner->mode = pConfig->mode; + pPanner->pan = pConfig->pan; + + return MA_SUCCESS; +} + +static void ma_stereo_balance_pcm_frames_f32(float* pFramesOut, const float* pFramesIn, ma_uint64 frameCount, float pan) +{ + ma_uint64 iFrame; + + if (pan > 0) { + float factor = 1.0f - pan; + if (pFramesOut == pFramesIn) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + pFramesOut[iFrame*2 + 0] = pFramesIn[iFrame*2 + 0] * factor; + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + pFramesOut[iFrame*2 + 0] = pFramesIn[iFrame*2 + 0] * factor; + pFramesOut[iFrame*2 + 1] = pFramesIn[iFrame*2 + 1]; + } + } + } else { + float factor = 1.0f + pan; + if (pFramesOut == pFramesIn) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + pFramesOut[iFrame*2 + 1] = pFramesIn[iFrame*2 + 1] * factor; + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + pFramesOut[iFrame*2 + 0] = pFramesIn[iFrame*2 + 0]; + pFramesOut[iFrame*2 + 1] = pFramesIn[iFrame*2 + 1] * factor; + } + } + } +} + +static void ma_stereo_balance_pcm_frames(void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount, ma_format format, float pan) +{ + if (pan == 0) { + /* Fast path. No panning required. */ + if (pFramesOut == pFramesIn) { + /* No-op */ + } else { + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, format, 2); + } + + return; + } + + switch (format) { + case ma_format_f32: ma_stereo_balance_pcm_frames_f32((float*)pFramesOut, (float*)pFramesIn, frameCount, pan); break; + + /* Unknown format. Just copy. */ + default: + { + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, format, 2); + } break; + } +} + + +static void ma_stereo_pan_pcm_frames_f32(float* pFramesOut, const float* pFramesIn, ma_uint64 frameCount, float pan) +{ + ma_uint64 iFrame; + + if (pan > 0) { + float factorL0 = 1.0f - pan; + float factorL1 = 0.0f + pan; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float sample0 = (pFramesIn[iFrame*2 + 0] * factorL0); + float sample1 = (pFramesIn[iFrame*2 + 0] * factorL1) + pFramesIn[iFrame*2 + 1]; + + pFramesOut[iFrame*2 + 0] = sample0; + pFramesOut[iFrame*2 + 1] = sample1; + } + } else { + float factorR0 = 0.0f - pan; + float factorR1 = 1.0f + pan; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float sample0 = pFramesIn[iFrame*2 + 0] + (pFramesIn[iFrame*2 + 1] * factorR0); + float sample1 = (pFramesIn[iFrame*2 + 1] * factorR1); + + pFramesOut[iFrame*2 + 0] = sample0; + pFramesOut[iFrame*2 + 1] = sample1; + } + } +} + +static void ma_stereo_pan_pcm_frames(void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount, ma_format format, float pan) +{ + if (pan == 0) { + /* Fast path. No panning required. */ + if (pFramesOut == pFramesIn) { + /* No-op */ + } else { + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, format, 2); + } + + return; + } + + switch (format) { + case ma_format_f32: ma_stereo_pan_pcm_frames_f32((float*)pFramesOut, (float*)pFramesIn, frameCount, pan); break; + + /* Unknown format. Just copy. */ + default: + { + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, format, 2); + } break; + } +} + +MA_API ma_result ma_panner_process_pcm_frames(ma_panner* pPanner, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pPanner == NULL || pFramesOut == NULL || pFramesIn == NULL) { + return MA_INVALID_ARGS; + } + + if (pPanner->channels == 2) { + /* Stereo case. For now assume channel 0 is left and channel right is 1, but should probably add support for a channel map. */ + if (pPanner->mode == ma_pan_mode_balance) { + ma_stereo_balance_pcm_frames(pFramesOut, pFramesIn, frameCount, pPanner->format, pPanner->pan); + } else { + ma_stereo_pan_pcm_frames(pFramesOut, pFramesIn, frameCount, pPanner->format, pPanner->pan); + } + } else { + if (pPanner->channels == 1) { + /* Panning has no effect on mono streams. */ + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, pPanner->format, pPanner->channels); + } else { + /* For now we're not going to support non-stereo set ups. Not sure how I want to handle this case just yet. */ + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, pPanner->format, pPanner->channels); + } + } + + return MA_SUCCESS; +} + +MA_API void ma_panner_set_mode(ma_panner* pPanner, ma_pan_mode mode) +{ + if (pPanner == NULL) { + return; + } + + pPanner->mode = mode; +} + +MA_API ma_pan_mode ma_panner_get_mode(const ma_panner* pPanner) +{ + if (pPanner == NULL) { + return ma_pan_mode_balance; + } + + return pPanner->mode; +} + +MA_API void ma_panner_set_pan(ma_panner* pPanner, float pan) +{ + if (pPanner == NULL) { + return; + } + + pPanner->pan = ma_clamp(pan, -1.0f, 1.0f); +} + +MA_API float ma_panner_get_pan(const ma_panner* pPanner) +{ + if (pPanner == NULL) { + return 0; + } + + return pPanner->pan; +} + + + + +MA_API ma_fader_config ma_fader_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate) +{ + ma_fader_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + + return config; +} + + +MA_API ma_result ma_fader_init(const ma_fader_config* pConfig, ma_fader* pFader) +{ + if (pFader == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pFader); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + /* Only f32 is supported for now. */ + if (pConfig->format != ma_format_f32) { + return MA_INVALID_ARGS; + } + + pFader->config = *pConfig; + pFader->volumeBeg = 1; + pFader->volumeEnd = 1; + pFader->lengthInFrames = 0; + pFader->cursorInFrames = 0; + + return MA_SUCCESS; +} + +MA_API ma_result ma_fader_process_pcm_frames(ma_fader* pFader, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pFader == NULL) { + return MA_INVALID_ARGS; + } + + /* If the cursor is still negative we need to just copy the absolute number of those frames, but no more than frameCount. */ + if (pFader->cursorInFrames < 0) { + ma_uint64 absCursorInFrames = (ma_uint64)0 - pFader->cursorInFrames; + if (absCursorInFrames > frameCount) { + absCursorInFrames = frameCount; + } + + ma_copy_pcm_frames(pFramesOut, pFramesIn, absCursorInFrames, pFader->config.format, pFader->config.channels); + + pFader->cursorInFrames += absCursorInFrames; + frameCount -= absCursorInFrames; + pFramesOut = ma_offset_ptr(pFramesOut, ma_get_bytes_per_frame(pFader->config.format, pFader->config.channels)*absCursorInFrames); + pFramesIn = ma_offset_ptr(pFramesIn, ma_get_bytes_per_frame(pFader->config.format, pFader->config.channels)*absCursorInFrames); + } + + if (pFader->cursorInFrames >= 0) { + /* + For now we need to clamp frameCount so that the cursor never overflows 32-bits. This is required for + the conversion to a float which we use for the linear interpolation. This might be changed later. + */ + if (frameCount + pFader->cursorInFrames > UINT_MAX) { + frameCount = UINT_MAX - pFader->cursorInFrames; + } + + /* Optimized path if volumeBeg and volumeEnd are equal. */ + if (pFader->volumeBeg == pFader->volumeEnd) { + if (pFader->volumeBeg == 1) { + /* Straight copy. */ + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, pFader->config.format, pFader->config.channels); + } else { + /* Copy with volume. */ + ma_copy_and_apply_volume_and_clip_pcm_frames(pFramesOut, pFramesIn, frameCount, pFader->config.format, pFader->config.channels, pFader->volumeBeg); + } + } else { + /* Slower path. Volumes are different, so may need to do an interpolation. */ + if ((ma_uint64)pFader->cursorInFrames >= pFader->lengthInFrames) { + /* Fast path. We've gone past the end of the fade period so just apply the end volume to all samples. */ + ma_copy_and_apply_volume_and_clip_pcm_frames(pFramesOut, pFramesIn, frameCount, pFader->config.format, pFader->config.channels, pFader->volumeEnd); + } else { + /* Slow path. This is where we do the actual fading. */ + ma_uint64 iFrame; + ma_uint32 iChannel; + + /* For now we only support f32. Support for other formats might be added later. */ + if (pFader->config.format == ma_format_f32) { + const float* pFramesInF32 = (const float*)pFramesIn; + /* */ float* pFramesOutF32 = ( float*)pFramesOut; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float a = (ma_uint32)ma_min(pFader->cursorInFrames + iFrame, pFader->lengthInFrames) / (float)((ma_uint32)pFader->lengthInFrames); /* Safe cast due to the frameCount clamp at the top of this function. */ + float volume = ma_mix_f32_fast(pFader->volumeBeg, pFader->volumeEnd, a); + + for (iChannel = 0; iChannel < pFader->config.channels; iChannel += 1) { + pFramesOutF32[iFrame*pFader->config.channels + iChannel] = pFramesInF32[iFrame*pFader->config.channels + iChannel] * volume; + } + } + } else { + return MA_NOT_IMPLEMENTED; + } + } + } + } + + pFader->cursorInFrames += frameCount; + + return MA_SUCCESS; +} + +MA_API void ma_fader_get_data_format(const ma_fader* pFader, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate) +{ + if (pFader == NULL) { + return; + } + + if (pFormat != NULL) { + *pFormat = pFader->config.format; + } + + if (pChannels != NULL) { + *pChannels = pFader->config.channels; + } + + if (pSampleRate != NULL) { + *pSampleRate = pFader->config.sampleRate; + } +} + +MA_API void ma_fader_set_fade(ma_fader* pFader, float volumeBeg, float volumeEnd, ma_uint64 lengthInFrames) +{ + ma_fader_set_fade_ex(pFader, volumeBeg, volumeEnd, lengthInFrames, 0); +} + +MA_API void ma_fader_set_fade_ex(ma_fader* pFader, float volumeBeg, float volumeEnd, ma_uint64 lengthInFrames, ma_int64 startOffsetInFrames) +{ + if (pFader == NULL) { + return; + } + + /* If the volume is negative, use current volume. */ + if (volumeBeg < 0) { + volumeBeg = ma_fader_get_current_volume(pFader); + } + + /* + The length needs to be clamped to 32-bits due to how we convert it to a float for linear + interpolation reasons. I might change this requirement later, but for now it's not important. + */ + if (lengthInFrames > UINT_MAX) { + lengthInFrames = UINT_MAX; + } + + /* The start offset needs to be clamped to ensure it doesn't overflow a signed number. */ + if (startOffsetInFrames > INT_MAX) { + startOffsetInFrames = INT_MAX; + } + + pFader->volumeBeg = volumeBeg; + pFader->volumeEnd = volumeEnd; + pFader->lengthInFrames = lengthInFrames; + pFader->cursorInFrames = -startOffsetInFrames; +} + +MA_API float ma_fader_get_current_volume(const ma_fader* pFader) +{ + if (pFader == NULL) { + return 0.0f; + } + + /* Any frames prior to the start of the fade period will be at unfaded volume. */ + if (pFader->cursorInFrames < 0) { + return 1.0f; + } + + /* The current volume depends on the position of the cursor. */ + if (pFader->cursorInFrames == 0) { + return pFader->volumeBeg; + } else if ((ma_uint64)pFader->cursorInFrames >= pFader->lengthInFrames) { /* Safe case because the < 0 case was checked above. */ + return pFader->volumeEnd; + } else { + /* The cursor is somewhere inside the fading period. We can figure this out with a simple linear interpolation between volumeBeg and volumeEnd based on our cursor position. */ + return ma_mix_f32_fast(pFader->volumeBeg, pFader->volumeEnd, (ma_uint32)pFader->cursorInFrames / (float)((ma_uint32)pFader->lengthInFrames)); /* Safe cast to uint32 because we clamp it in ma_fader_process_pcm_frames(). */ + } +} + + + + + +MA_API ma_vec3f ma_vec3f_init_3f(float x, float y, float z) +{ + ma_vec3f v; + + v.x = x; + v.y = y; + v.z = z; + + return v; +} + +MA_API ma_vec3f ma_vec3f_sub(ma_vec3f a, ma_vec3f b) +{ + return ma_vec3f_init_3f( + a.x - b.x, + a.y - b.y, + a.z - b.z + ); +} + +MA_API ma_vec3f ma_vec3f_neg(ma_vec3f a) +{ + return ma_vec3f_init_3f( + -a.x, + -a.y, + -a.z + ); +} + +MA_API float ma_vec3f_dot(ma_vec3f a, ma_vec3f b) +{ + return a.x*b.x + a.y*b.y + a.z*b.z; +} + +MA_API float ma_vec3f_len2(ma_vec3f v) +{ + return ma_vec3f_dot(v, v); +} + +MA_API float ma_vec3f_len(ma_vec3f v) +{ + return (float)ma_sqrtd(ma_vec3f_len2(v)); +} + + + +MA_API float ma_vec3f_dist(ma_vec3f a, ma_vec3f b) +{ + return ma_vec3f_len(ma_vec3f_sub(a, b)); +} + +MA_API ma_vec3f ma_vec3f_normalize(ma_vec3f v) +{ + float invLen; + float len2 = ma_vec3f_len2(v); + if (len2 == 0) { + return ma_vec3f_init_3f(0, 0, 0); + } + + invLen = ma_rsqrtf(len2); + v.x *= invLen; + v.y *= invLen; + v.z *= invLen; + + return v; +} + +MA_API ma_vec3f ma_vec3f_cross(ma_vec3f a, ma_vec3f b) +{ + return ma_vec3f_init_3f( + a.y*b.z - a.z*b.y, + a.z*b.x - a.x*b.z, + a.x*b.y - a.y*b.x + ); +} + + +MA_API void ma_atomic_vec3f_init(ma_atomic_vec3f* v, ma_vec3f value) +{ + v->v = value; + v->lock = 0; /* Important this is initialized to 0. */ +} + +MA_API void ma_atomic_vec3f_set(ma_atomic_vec3f* v, ma_vec3f value) +{ + ma_spinlock_lock(&v->lock); + { + v->v = value; + } + ma_spinlock_unlock(&v->lock); +} + +MA_API ma_vec3f ma_atomic_vec3f_get(ma_atomic_vec3f* v) +{ + ma_vec3f r; + + ma_spinlock_lock(&v->lock); + { + r = v->v; + } + ma_spinlock_unlock(&v->lock); + + return r; +} + + + +static void ma_channel_map_apply_f32(float* pFramesOut, const ma_channel* pChannelMapOut, ma_uint32 channelsOut, const float* pFramesIn, const ma_channel* pChannelMapIn, ma_uint32 channelsIn, ma_uint64 frameCount, ma_channel_mix_mode mode, ma_mono_expansion_mode monoExpansionMode); +static ma_bool32 ma_is_spatial_channel_position(ma_channel channelPosition); + + +#ifndef MA_DEFAULT_SPEED_OF_SOUND +#define MA_DEFAULT_SPEED_OF_SOUND 343.3f +#endif + +/* +These vectors represent the direction that speakers are facing from the center point. They're used +for panning in the spatializer. Must be normalized. +*/ +static ma_vec3f g_maChannelDirections[MA_CHANNEL_POSITION_COUNT] = { + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_NONE */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_MONO */ + {-0.7071f, 0.0f, -0.7071f }, /* MA_CHANNEL_FRONT_LEFT */ + {+0.7071f, 0.0f, -0.7071f }, /* MA_CHANNEL_FRONT_RIGHT */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_FRONT_CENTER */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_LFE */ + {-0.7071f, 0.0f, +0.7071f }, /* MA_CHANNEL_BACK_LEFT */ + {+0.7071f, 0.0f, +0.7071f }, /* MA_CHANNEL_BACK_RIGHT */ + {-0.3162f, 0.0f, -0.9487f }, /* MA_CHANNEL_FRONT_LEFT_CENTER */ + {+0.3162f, 0.0f, -0.9487f }, /* MA_CHANNEL_FRONT_RIGHT_CENTER */ + { 0.0f, 0.0f, +1.0f }, /* MA_CHANNEL_BACK_CENTER */ + {-1.0f, 0.0f, 0.0f }, /* MA_CHANNEL_SIDE_LEFT */ + {+1.0f, 0.0f, 0.0f }, /* MA_CHANNEL_SIDE_RIGHT */ + { 0.0f, +1.0f, 0.0f }, /* MA_CHANNEL_TOP_CENTER */ + {-0.5774f, +0.5774f, -0.5774f }, /* MA_CHANNEL_TOP_FRONT_LEFT */ + { 0.0f, +0.7071f, -0.7071f }, /* MA_CHANNEL_TOP_FRONT_CENTER */ + {+0.5774f, +0.5774f, -0.5774f }, /* MA_CHANNEL_TOP_FRONT_RIGHT */ + {-0.5774f, +0.5774f, +0.5774f }, /* MA_CHANNEL_TOP_BACK_LEFT */ + { 0.0f, +0.7071f, +0.7071f }, /* MA_CHANNEL_TOP_BACK_CENTER */ + {+0.5774f, +0.5774f, +0.5774f }, /* MA_CHANNEL_TOP_BACK_RIGHT */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_0 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_1 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_2 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_3 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_4 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_5 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_6 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_7 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_8 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_9 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_10 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_11 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_12 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_13 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_14 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_15 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_16 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_17 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_18 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_19 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_20 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_21 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_22 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_23 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_24 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_25 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_26 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_27 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_28 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_29 */ + { 0.0f, 0.0f, -1.0f }, /* MA_CHANNEL_AUX_30 */ + { 0.0f, 0.0f, -1.0f } /* MA_CHANNEL_AUX_31 */ +}; + +static ma_vec3f ma_get_channel_direction(ma_channel channel) +{ + if (channel >= MA_CHANNEL_POSITION_COUNT) { + return ma_vec3f_init_3f(0, 0, -1); + } else { + return g_maChannelDirections[channel]; + } +} + + + +static float ma_attenuation_inverse(float distance, float minDistance, float maxDistance, float rolloff) +{ + if (minDistance >= maxDistance) { + return 1; /* To avoid division by zero. Do not attenuate. */ + } + + return minDistance / (minDistance + rolloff * (ma_clamp(distance, minDistance, maxDistance) - minDistance)); +} + +static float ma_attenuation_linear(float distance, float minDistance, float maxDistance, float rolloff) +{ + if (minDistance >= maxDistance) { + return 1; /* To avoid division by zero. Do not attenuate. */ + } + + return 1 - rolloff * (ma_clamp(distance, minDistance, maxDistance) - minDistance) / (maxDistance - minDistance); +} + +static float ma_attenuation_exponential(float distance, float minDistance, float maxDistance, float rolloff) +{ + if (minDistance >= maxDistance) { + return 1; /* To avoid division by zero. Do not attenuate. */ + } + + return (float)ma_powd(ma_clamp(distance, minDistance, maxDistance) / minDistance, -rolloff); +} + + +/* +Doppler Effect calculation taken from the OpenAL spec, with two main differences: + + 1) The source to listener vector will have already been calculated at an earlier step so we can + just use that directly. We need only the position of the source relative to the origin. + + 2) We don't scale by a frequency because we actually just want the ratio which we'll plug straight + into the resampler directly. +*/ +static float ma_doppler_pitch(ma_vec3f relativePosition, ma_vec3f sourceVelocity, ma_vec3f listenVelocity, float speedOfSound, float dopplerFactor) +{ + float len; + float vls; + float vss; + + len = ma_vec3f_len(relativePosition); + + /* + There's a case where the position of the source will be right on top of the listener in which + case the length will be 0 and we'll end up with a division by zero. We can just return a ratio + of 1.0 in this case. This is not considered in the OpenAL spec, but is necessary. + */ + if (len == 0) { + return 1.0; + } + + vls = ma_vec3f_dot(relativePosition, listenVelocity) / len; + vss = ma_vec3f_dot(relativePosition, sourceVelocity) / len; + + vls = ma_min(vls, speedOfSound / dopplerFactor); + vss = ma_min(vss, speedOfSound / dopplerFactor); + + return (speedOfSound - dopplerFactor*vls) / (speedOfSound - dopplerFactor*vss); +} + + +static void ma_get_default_channel_map_for_spatializer(ma_channel* pChannelMap, size_t channelMapCap, ma_uint32 channelCount) +{ + /* + Special case for stereo. Want to default the left and right speakers to side left and side + right so that they're facing directly down the X axis rather than slightly forward. Not + doing this will result in sounds being quieter when behind the listener. This might + actually be good for some scenarios, but I don't think it's an appropriate default because + it can be a bit unexpected. + */ + if (channelCount == 2) { + pChannelMap[0] = MA_CHANNEL_SIDE_LEFT; + pChannelMap[1] = MA_CHANNEL_SIDE_RIGHT; + } else { + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, channelCount); + } +} + + +MA_API ma_spatializer_listener_config ma_spatializer_listener_config_init(ma_uint32 channelsOut) +{ + ma_spatializer_listener_config config; + + MA_ZERO_OBJECT(&config); + config.channelsOut = channelsOut; + config.pChannelMapOut = NULL; + config.handedness = ma_handedness_right; + config.worldUp = ma_vec3f_init_3f(0, 1, 0); + config.coneInnerAngleInRadians = 6.283185f; /* 360 degrees. */ + config.coneOuterAngleInRadians = 6.283185f; /* 360 degrees. */ + config.coneOuterGain = 0; + config.speedOfSound = 343.3f; /* Same as OpenAL. Used for doppler effect. */ + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t channelMapOutOffset; +} ma_spatializer_listener_heap_layout; + +static ma_result ma_spatializer_listener_get_heap_layout(const ma_spatializer_listener_config* pConfig, ma_spatializer_listener_heap_layout* pHeapLayout) +{ + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channelsOut == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* Channel map. We always need this, even for passthroughs. */ + pHeapLayout->channelMapOutOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(sizeof(*pConfig->pChannelMapOut) * pConfig->channelsOut); + + return MA_SUCCESS; +} + + +MA_API ma_result ma_spatializer_listener_get_heap_size(const ma_spatializer_listener_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_spatializer_listener_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_spatializer_listener_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_spatializer_listener_init_preallocated(const ma_spatializer_listener_config* pConfig, void* pHeap, ma_spatializer_listener* pListener) +{ + ma_result result; + ma_spatializer_listener_heap_layout heapLayout; + + if (pListener == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pListener); + + result = ma_spatializer_listener_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pListener->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pListener->config = *pConfig; + ma_atomic_vec3f_init(&pListener->position, ma_vec3f_init_3f(0, 0, 0)); + ma_atomic_vec3f_init(&pListener->direction, ma_vec3f_init_3f(0, 0, -1)); + ma_atomic_vec3f_init(&pListener->velocity, ma_vec3f_init_3f(0, 0, 0)); + pListener->isEnabled = MA_TRUE; + + /* Swap the forward direction if we're left handed (it was initialized based on right handed). */ + if (pListener->config.handedness == ma_handedness_left) { + ma_vec3f negDir = ma_vec3f_neg(ma_spatializer_listener_get_direction(pListener)); + ma_spatializer_listener_set_direction(pListener, negDir.x, negDir.y, negDir.z); + } + + + /* We must always have a valid channel map. */ + pListener->config.pChannelMapOut = (ma_channel*)ma_offset_ptr(pHeap, heapLayout.channelMapOutOffset); + + /* Use a slightly different default channel map for stereo. */ + if (pConfig->pChannelMapOut == NULL) { + ma_get_default_channel_map_for_spatializer(pListener->config.pChannelMapOut, pConfig->channelsOut, pConfig->channelsOut); + } else { + ma_channel_map_copy_or_default(pListener->config.pChannelMapOut, pConfig->channelsOut, pConfig->pChannelMapOut, pConfig->channelsOut); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_spatializer_listener_init(const ma_spatializer_listener_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_spatializer_listener* pListener) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_spatializer_listener_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_spatializer_listener_init_preallocated(pConfig, pHeap, pListener); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pListener->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_spatializer_listener_uninit(ma_spatializer_listener* pListener, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pListener == NULL) { + return; + } + + if (pListener->_ownsHeap) { + ma_free(pListener->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_channel* ma_spatializer_listener_get_channel_map(ma_spatializer_listener* pListener) +{ + if (pListener == NULL) { + return NULL; + } + + return pListener->config.pChannelMapOut; +} + +MA_API void ma_spatializer_listener_set_cone(ma_spatializer_listener* pListener, float innerAngleInRadians, float outerAngleInRadians, float outerGain) +{ + if (pListener == NULL) { + return; + } + + pListener->config.coneInnerAngleInRadians = innerAngleInRadians; + pListener->config.coneOuterAngleInRadians = outerAngleInRadians; + pListener->config.coneOuterGain = outerGain; +} + +MA_API void ma_spatializer_listener_get_cone(const ma_spatializer_listener* pListener, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain) +{ + if (pListener == NULL) { + return; + } + + if (pInnerAngleInRadians != NULL) { + *pInnerAngleInRadians = pListener->config.coneInnerAngleInRadians; + } + + if (pOuterAngleInRadians != NULL) { + *pOuterAngleInRadians = pListener->config.coneOuterAngleInRadians; + } + + if (pOuterGain != NULL) { + *pOuterGain = pListener->config.coneOuterGain; + } +} + +MA_API void ma_spatializer_listener_set_position(ma_spatializer_listener* pListener, float x, float y, float z) +{ + if (pListener == NULL) { + return; + } + + ma_atomic_vec3f_set(&pListener->position, ma_vec3f_init_3f(x, y, z)); +} + +MA_API ma_vec3f ma_spatializer_listener_get_position(const ma_spatializer_listener* pListener) +{ + if (pListener == NULL) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_atomic_vec3f_get((ma_atomic_vec3f*)&pListener->position); /* Naughty const-cast. It's just for atomically loading the vec3 which should be safe. */ +} + +MA_API void ma_spatializer_listener_set_direction(ma_spatializer_listener* pListener, float x, float y, float z) +{ + if (pListener == NULL) { + return; + } + + ma_atomic_vec3f_set(&pListener->direction, ma_vec3f_init_3f(x, y, z)); +} + +MA_API ma_vec3f ma_spatializer_listener_get_direction(const ma_spatializer_listener* pListener) +{ + if (pListener == NULL) { + return ma_vec3f_init_3f(0, 0, -1); + } + + return ma_atomic_vec3f_get((ma_atomic_vec3f*)&pListener->direction); /* Naughty const-cast. It's just for atomically loading the vec3 which should be safe. */ +} + +MA_API void ma_spatializer_listener_set_velocity(ma_spatializer_listener* pListener, float x, float y, float z) +{ + if (pListener == NULL) { + return; + } + + ma_atomic_vec3f_set(&pListener->velocity, ma_vec3f_init_3f(x, y, z)); +} + +MA_API ma_vec3f ma_spatializer_listener_get_velocity(const ma_spatializer_listener* pListener) +{ + if (pListener == NULL) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_atomic_vec3f_get((ma_atomic_vec3f*)&pListener->velocity); /* Naughty const-cast. It's just for atomically loading the vec3 which should be safe. */ +} + +MA_API void ma_spatializer_listener_set_speed_of_sound(ma_spatializer_listener* pListener, float speedOfSound) +{ + if (pListener == NULL) { + return; + } + + pListener->config.speedOfSound = speedOfSound; +} + +MA_API float ma_spatializer_listener_get_speed_of_sound(const ma_spatializer_listener* pListener) +{ + if (pListener == NULL) { + return 0; + } + + return pListener->config.speedOfSound; +} + +MA_API void ma_spatializer_listener_set_world_up(ma_spatializer_listener* pListener, float x, float y, float z) +{ + if (pListener == NULL) { + return; + } + + pListener->config.worldUp = ma_vec3f_init_3f(x, y, z); +} + +MA_API ma_vec3f ma_spatializer_listener_get_world_up(const ma_spatializer_listener* pListener) +{ + if (pListener == NULL) { + return ma_vec3f_init_3f(0, 1, 0); + } + + return pListener->config.worldUp; +} + +MA_API void ma_spatializer_listener_set_enabled(ma_spatializer_listener* pListener, ma_bool32 isEnabled) +{ + if (pListener == NULL) { + return; + } + + pListener->isEnabled = isEnabled; +} + +MA_API ma_bool32 ma_spatializer_listener_is_enabled(const ma_spatializer_listener* pListener) +{ + if (pListener == NULL) { + return MA_FALSE; + } + + return pListener->isEnabled; +} + + + + +MA_API ma_spatializer_config ma_spatializer_config_init(ma_uint32 channelsIn, ma_uint32 channelsOut) +{ + ma_spatializer_config config; + + MA_ZERO_OBJECT(&config); + config.channelsIn = channelsIn; + config.channelsOut = channelsOut; + config.pChannelMapIn = NULL; + config.attenuationModel = ma_attenuation_model_inverse; + config.positioning = ma_positioning_absolute; + config.handedness = ma_handedness_right; + config.minGain = 0; + config.maxGain = 1; + config.minDistance = 1; + config.maxDistance = MA_FLT_MAX; + config.rolloff = 1; + config.coneInnerAngleInRadians = 6.283185f; /* 360 degrees. */ + config.coneOuterAngleInRadians = 6.283185f; /* 360 degrees. */ + config.coneOuterGain = 0.0f; + config.dopplerFactor = 1; + config.directionalAttenuationFactor = 1; + config.minSpatializationChannelGain = 0.2f; + config.gainSmoothTimeInFrames = 360; /* 7.5ms @ 48K. */ + + return config; +} + + +static ma_gainer_config ma_spatializer_gainer_config_init(const ma_spatializer_config* pConfig) +{ + MA_ASSERT(pConfig != NULL); + return ma_gainer_config_init(pConfig->channelsOut, pConfig->gainSmoothTimeInFrames); +} + +static ma_result ma_spatializer_validate_config(const ma_spatializer_config* pConfig) +{ + MA_ASSERT(pConfig != NULL); + + if (pConfig->channelsIn == 0 || pConfig->channelsOut == 0) { + return MA_INVALID_ARGS; + } + + return MA_SUCCESS; +} + +typedef struct +{ + size_t sizeInBytes; + size_t channelMapInOffset; + size_t newChannelGainsOffset; + size_t gainerOffset; +} ma_spatializer_heap_layout; + +static ma_result ma_spatializer_get_heap_layout(const ma_spatializer_config* pConfig, ma_spatializer_heap_layout* pHeapLayout) +{ + ma_result result; + + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_spatializer_validate_config(pConfig); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes = 0; + + /* Channel map. */ + pHeapLayout->channelMapInOffset = MA_SIZE_MAX; /* <-- MA_SIZE_MAX indicates no allocation necessary. */ + if (pConfig->pChannelMapIn != NULL) { + pHeapLayout->channelMapInOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(sizeof(*pConfig->pChannelMapIn) * pConfig->channelsIn); + } + + /* New channel gains for output. */ + pHeapLayout->newChannelGainsOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(sizeof(float) * pConfig->channelsOut); + + /* Gainer. */ + { + size_t gainerHeapSizeInBytes; + ma_gainer_config gainerConfig; + + gainerConfig = ma_spatializer_gainer_config_init(pConfig); + + result = ma_gainer_get_heap_size(&gainerConfig, &gainerHeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->gainerOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(gainerHeapSizeInBytes); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_spatializer_get_heap_size(const ma_spatializer_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_spatializer_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; /* Safety. */ + + result = ma_spatializer_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + + +MA_API ma_result ma_spatializer_init_preallocated(const ma_spatializer_config* pConfig, void* pHeap, ma_spatializer* pSpatializer) +{ + ma_result result; + ma_spatializer_heap_layout heapLayout; + ma_gainer_config gainerConfig; + + if (pSpatializer == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pSpatializer); + + if (pConfig == NULL || pHeap == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_spatializer_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pSpatializer->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pSpatializer->channelsIn = pConfig->channelsIn; + pSpatializer->channelsOut = pConfig->channelsOut; + pSpatializer->attenuationModel = pConfig->attenuationModel; + pSpatializer->positioning = pConfig->positioning; + pSpatializer->handedness = pConfig->handedness; + pSpatializer->minGain = pConfig->minGain; + pSpatializer->maxGain = pConfig->maxGain; + pSpatializer->minDistance = pConfig->minDistance; + pSpatializer->maxDistance = pConfig->maxDistance; + pSpatializer->rolloff = pConfig->rolloff; + pSpatializer->coneInnerAngleInRadians = pConfig->coneInnerAngleInRadians; + pSpatializer->coneOuterAngleInRadians = pConfig->coneOuterAngleInRadians; + pSpatializer->coneOuterGain = pConfig->coneOuterGain; + pSpatializer->dopplerFactor = pConfig->dopplerFactor; + pSpatializer->minSpatializationChannelGain = pConfig->minSpatializationChannelGain; + pSpatializer->directionalAttenuationFactor = pConfig->directionalAttenuationFactor; + pSpatializer->gainSmoothTimeInFrames = pConfig->gainSmoothTimeInFrames; + ma_atomic_vec3f_init(&pSpatializer->position, ma_vec3f_init_3f(0, 0, 0)); + ma_atomic_vec3f_init(&pSpatializer->direction, ma_vec3f_init_3f(0, 0, -1)); + ma_atomic_vec3f_init(&pSpatializer->velocity, ma_vec3f_init_3f(0, 0, 0)); + pSpatializer->dopplerPitch = 1; + + /* Swap the forward direction if we're left handed (it was initialized based on right handed). */ + if (pSpatializer->handedness == ma_handedness_left) { + ma_vec3f negDir = ma_vec3f_neg(ma_spatializer_get_direction(pSpatializer)); + ma_spatializer_set_direction(pSpatializer, negDir.x, negDir.y, negDir.z); + } + + /* Channel map. This will be on the heap. */ + if (pConfig->pChannelMapIn != NULL) { + pSpatializer->pChannelMapIn = (ma_channel*)ma_offset_ptr(pHeap, heapLayout.channelMapInOffset); + ma_channel_map_copy_or_default(pSpatializer->pChannelMapIn, pSpatializer->channelsIn, pConfig->pChannelMapIn, pSpatializer->channelsIn); + } + + /* New channel gains for output channels. */ + pSpatializer->pNewChannelGainsOut = (float*)ma_offset_ptr(pHeap, heapLayout.newChannelGainsOffset); + + /* Gainer. */ + gainerConfig = ma_spatializer_gainer_config_init(pConfig); + + result = ma_gainer_init_preallocated(&gainerConfig, ma_offset_ptr(pHeap, heapLayout.gainerOffset), &pSpatializer->gainer); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the gainer. */ + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_spatializer_init(const ma_spatializer_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_spatializer* pSpatializer) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + /* We'll need a heap allocation to retrieve the size. */ + result = ma_spatializer_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_spatializer_init_preallocated(pConfig, pHeap, pSpatializer); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pSpatializer->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_spatializer_uninit(ma_spatializer* pSpatializer, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pSpatializer == NULL) { + return; + } + + ma_gainer_uninit(&pSpatializer->gainer, pAllocationCallbacks); + + if (pSpatializer->_ownsHeap) { + ma_free(pSpatializer->_pHeap, pAllocationCallbacks); + } +} + +static float ma_calculate_angular_gain(ma_vec3f dirA, ma_vec3f dirB, float coneInnerAngleInRadians, float coneOuterAngleInRadians, float coneOuterGain) +{ + /* + Angular attenuation. + + Unlike distance gain, the math for this is not specified by the OpenAL spec so we'll just go ahead and figure + this out for ourselves at the expense of possibly being inconsistent with other implementations. + + To do cone attenuation, I'm just using the same math that we'd use to implement a basic spotlight in OpenGL. We + just need to get the direction from the source to the listener and then do a dot product against that and the + direction of the spotlight. Then we just compare that dot product against the cosine of the inner and outer + angles. If the dot product is greater than the outer angle, we just use coneOuterGain. If it's less than + the inner angle, we just use a gain of 1. Otherwise we linearly interpolate between 1 and coneOuterGain. + */ + if (coneInnerAngleInRadians < 6.283185f) { + float angularGain = 1; + float cutoffInner = (float)ma_cosd(coneInnerAngleInRadians*0.5f); + float cutoffOuter = (float)ma_cosd(coneOuterAngleInRadians*0.5f); + float d; + + d = ma_vec3f_dot(dirA, dirB); + + if (d > cutoffInner) { + /* It's inside the inner angle. */ + angularGain = 1; + } else { + /* It's outside the inner angle. */ + if (d > cutoffOuter) { + /* It's between the inner and outer angle. We need to linearly interpolate between 1 and coneOuterGain. */ + angularGain = ma_mix_f32(coneOuterGain, 1, (d - cutoffOuter) / (cutoffInner - cutoffOuter)); + } else { + /* It's outside the outer angle. */ + angularGain = coneOuterGain; + } + } + + /*printf("d = %f; cutoffInner = %f; cutoffOuter = %f; angularGain = %f\n", d, cutoffInner, cutoffOuter, angularGain);*/ + return angularGain; + } else { + /* Inner angle is 360 degrees so no need to do any attenuation. */ + return 1; + } +} + +MA_API ma_result ma_spatializer_process_pcm_frames(ma_spatializer* pSpatializer, ma_spatializer_listener* pListener, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_channel* pChannelMapIn = pSpatializer->pChannelMapIn; + ma_channel* pChannelMapOut = pListener->config.pChannelMapOut; + + if (pSpatializer == NULL) { + return MA_INVALID_ARGS; + } + + /* If we're not spatializing we need to run an optimized path. */ + if (ma_atomic_load_i32(&pSpatializer->attenuationModel) == ma_attenuation_model_none) { + if (ma_spatializer_listener_is_enabled(pListener)) { + /* No attenuation is required, but we'll need to do some channel conversion. */ + if (pSpatializer->channelsIn == pSpatializer->channelsOut) { + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, ma_format_f32, pSpatializer->channelsIn); + } else { + ma_channel_map_apply_f32((float*)pFramesOut, pChannelMapOut, pSpatializer->channelsOut, (const float*)pFramesIn, pChannelMapIn, pSpatializer->channelsIn, frameCount, ma_channel_mix_mode_rectangular, ma_mono_expansion_mode_default); /* Safe casts to float* because f32 is the only supported format. */ + } + } else { + /* The listener is disabled. Output silence. */ + ma_silence_pcm_frames(pFramesOut, frameCount, ma_format_f32, pSpatializer->channelsOut); + } + + /* + We're not doing attenuation so don't bother with doppler for now. I'm not sure if this is + the correct thinking so might need to review this later. + */ + pSpatializer->dopplerPitch = 1; + } else { + /* + Let's first determine which listener the sound is closest to. Need to keep in mind that we + might not have a world or any listeners, in which case we just spatializer based on the + listener being positioned at the origin (0, 0, 0). + */ + ma_vec3f relativePosNormalized; + ma_vec3f relativePos; /* The position relative to the listener. */ + ma_vec3f relativeDir; /* The direction of the sound, relative to the listener. */ + ma_vec3f listenerVel; /* The velocity of the listener. For doppler pitch calculation. */ + float speedOfSound; + float distance = 0; + float gain = 1; + ma_uint32 iChannel; + const ma_uint32 channelsOut = pSpatializer->channelsOut; + const ma_uint32 channelsIn = pSpatializer->channelsIn; + float minDistance = ma_spatializer_get_min_distance(pSpatializer); + float maxDistance = ma_spatializer_get_max_distance(pSpatializer); + float rolloff = ma_spatializer_get_rolloff(pSpatializer); + float dopplerFactor = ma_spatializer_get_doppler_factor(pSpatializer); + + /* + We'll need the listener velocity for doppler pitch calculations. The speed of sound is + defined by the listener, so we'll grab that here too. + */ + if (pListener != NULL) { + listenerVel = ma_spatializer_listener_get_velocity(pListener); + speedOfSound = pListener->config.speedOfSound; + } else { + listenerVel = ma_vec3f_init_3f(0, 0, 0); + speedOfSound = MA_DEFAULT_SPEED_OF_SOUND; + } + + if (pListener == NULL || ma_spatializer_get_positioning(pSpatializer) == ma_positioning_relative) { + /* There's no listener or we're using relative positioning. */ + relativePos = ma_spatializer_get_position(pSpatializer); + relativeDir = ma_spatializer_get_direction(pSpatializer); + } else { + /* + We've found a listener and we're using absolute positioning. We need to transform the + sound's position and direction so that it's relative to listener. Later on we'll use + this for determining the factors to apply to each channel to apply the panning effect. + */ + ma_spatializer_get_relative_position_and_direction(pSpatializer, pListener, &relativePos, &relativeDir); + } + + distance = ma_vec3f_len(relativePos); + + /* We've gathered the data, so now we can apply some spatialization. */ + switch (ma_spatializer_get_attenuation_model(pSpatializer)) { + case ma_attenuation_model_inverse: + { + gain = ma_attenuation_inverse(distance, minDistance, maxDistance, rolloff); + } break; + case ma_attenuation_model_linear: + { + gain = ma_attenuation_linear(distance, minDistance, maxDistance, rolloff); + } break; + case ma_attenuation_model_exponential: + { + gain = ma_attenuation_exponential(distance, minDistance, maxDistance, rolloff); + } break; + case ma_attenuation_model_none: + default: + { + gain = 1; + } break; + } + + /* Normalize the position. */ + if (distance > 0.001f) { + float distanceInv = 1/distance; + relativePosNormalized = relativePos; + relativePosNormalized.x *= distanceInv; + relativePosNormalized.y *= distanceInv; + relativePosNormalized.z *= distanceInv; + } else { + distance = 0; + relativePosNormalized = ma_vec3f_init_3f(0, 0, 0); + } + + /* + Angular attenuation. + + Unlike distance gain, the math for this is not specified by the OpenAL spec so we'll just go ahead and figure + this out for ourselves at the expense of possibly being inconsistent with other implementations. + + To do cone attenuation, I'm just using the same math that we'd use to implement a basic spotlight in OpenGL. We + just need to get the direction from the source to the listener and then do a dot product against that and the + direction of the spotlight. Then we just compare that dot product against the cosine of the inner and outer + angles. If the dot product is greater than the outer angle, we just use coneOuterGain. If it's less than + the inner angle, we just use a gain of 1. Otherwise we linearly interpolate between 1 and coneOuterGain. + */ + if (distance > 0) { + /* Source angular gain. */ + float spatializerConeInnerAngle; + float spatializerConeOuterAngle; + float spatializerConeOuterGain; + ma_spatializer_get_cone(pSpatializer, &spatializerConeInnerAngle, &spatializerConeOuterAngle, &spatializerConeOuterGain); + + gain *= ma_calculate_angular_gain(relativeDir, ma_vec3f_neg(relativePosNormalized), spatializerConeInnerAngle, spatializerConeOuterAngle, spatializerConeOuterGain); + + /* + We're supporting angular gain on the listener as well for those who want to reduce the volume of sounds that + are positioned behind the listener. On default settings, this will have no effect. + */ + if (pListener != NULL && pListener->config.coneInnerAngleInRadians < 6.283185f) { + ma_vec3f listenerDirection; + float listenerInnerAngle; + float listenerOuterAngle; + float listenerOuterGain; + + if (pListener->config.handedness == ma_handedness_right) { + listenerDirection = ma_vec3f_init_3f(0, 0, -1); + } else { + listenerDirection = ma_vec3f_init_3f(0, 0, +1); + } + + listenerInnerAngle = pListener->config.coneInnerAngleInRadians; + listenerOuterAngle = pListener->config.coneOuterAngleInRadians; + listenerOuterGain = pListener->config.coneOuterGain; + + gain *= ma_calculate_angular_gain(listenerDirection, relativePosNormalized, listenerInnerAngle, listenerOuterAngle, listenerOuterGain); + } + } else { + /* The sound is right on top of the listener. Don't do any angular attenuation. */ + } + + + /* Clamp the gain. */ + gain = ma_clamp(gain, ma_spatializer_get_min_gain(pSpatializer), ma_spatializer_get_max_gain(pSpatializer)); + + /* + The gain needs to be applied per-channel here. The spatialization code below will be changing the per-channel + gains which will then eventually be passed into the gainer which will deal with smoothing the gain transitions + to avoid harsh changes in gain. + */ + for (iChannel = 0; iChannel < channelsOut; iChannel += 1) { + pSpatializer->pNewChannelGainsOut[iChannel] = gain; + } + + /* + Convert to our output channel count. If the listener is disabled we just output silence here. We cannot ignore + the whole section of code here because we need to update some internal spatialization state. + */ + if (ma_spatializer_listener_is_enabled(pListener)) { + ma_channel_map_apply_f32((float*)pFramesOut, pChannelMapOut, channelsOut, (const float*)pFramesIn, pChannelMapIn, channelsIn, frameCount, ma_channel_mix_mode_rectangular, ma_mono_expansion_mode_default); + } else { + ma_silence_pcm_frames(pFramesOut, frameCount, ma_format_f32, pSpatializer->channelsOut); + } + + + /* + Panning. This is where we'll apply the gain and convert to the output channel count. We have an optimized path for + when we're converting to a mono stream. In that case we don't really need to do any panning - we just apply the + gain to the final output. + */ + /*printf("distance=%f; gain=%f\n", distance, gain);*/ + + /* We must have a valid channel map here to ensure we spatialize properly. */ + MA_ASSERT(pChannelMapOut != NULL); + + /* + We're not converting to mono so we'll want to apply some panning. This is where the feeling of something being + to the left, right, infront or behind the listener is calculated. I'm just using a basic model here. Note that + the code below is not based on any specific algorithm. I'm just implementing this off the top of my head and + seeing how it goes. There might be better ways to do this. + + To determine the direction of the sound relative to a speaker I'm using dot products. Each speaker is given a + direction. For example, the left channel in a stereo system will be -1 on the X axis and the right channel will + be +1 on the X axis. A dot product is performed against the direction vector of the channel and the normalized + position of the sound. + */ + + /* + Calculate our per-channel gains. We do this based on the normalized relative position of the sound and it's + relation to the direction of the channel. + */ + if (distance > 0) { + ma_vec3f unitPos = relativePos; + float distanceInv = 1/distance; + unitPos.x *= distanceInv; + unitPos.y *= distanceInv; + unitPos.z *= distanceInv; + + for (iChannel = 0; iChannel < channelsOut; iChannel += 1) { + ma_channel channelOut; + float d; + float dMin; + + channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannel); + if (ma_is_spatial_channel_position(channelOut)) { + d = ma_mix_f32_fast(1, ma_vec3f_dot(unitPos, ma_get_channel_direction(channelOut)), ma_spatializer_get_directional_attenuation_factor(pSpatializer)); + } else { + d = 1; /* It's not a spatial channel so there's no real notion of direction. */ + } + + /* + In my testing, if the panning effect is too aggressive it makes spatialization feel uncomfortable. + The "dMin" variable below is used to control the aggressiveness of the panning effect. When set to + 0, panning will be most extreme and any sounds that are positioned on the opposite side of the + speaker will be completely silent from that speaker. Not only does this feel uncomfortable, it + doesn't even remotely represent the real world at all because sounds that come from your right side + are still clearly audible from your left side. Setting "dMin" to 1 will result in no panning at + all, which is also not ideal. By setting it to something greater than 0, the spatialization effect + becomes much less dramatic and a lot more bearable. + + Summary: 0 = more extreme panning; 1 = no panning. + */ + dMin = pSpatializer->minSpatializationChannelGain; + + /* + At this point, "d" will be positive if the sound is on the same side as the channel and negative if + it's on the opposite side. It will be in the range of -1..1. There's two ways I can think of to + calculate a panning value. The first is to simply convert it to 0..1, however this has a problem + which I'm not entirely happy with. Considering a stereo system, when a sound is positioned right + in front of the listener it'll result in each speaker getting a gain of 0.5. I don't know if I like + the idea of having a scaling factor of 0.5 being applied to a sound when it's sitting right in front + of the listener. I would intuitively expect that to be played at full volume, or close to it. + + The second idea I think of is to only apply a reduction in gain when the sound is on the opposite + side of the speaker. That is, reduce the gain only when the dot product is negative. The problem + with this is that there will not be any attenuation as the sound sweeps around the 180 degrees + where the dot product is positive. The idea with this option is that you leave the gain at 1 when + the sound is being played on the same side as the speaker and then you just reduce the volume when + the sound is on the other side. + + The summarize, I think the first option should give a better sense of spatialization, but the second + option is better for preserving the sound's power. + + UPDATE: In my testing, I find the first option to sound better. You can feel the sense of space a + bit better, but you can also hear the reduction in volume when it's right in front. + */ + #if 1 + { + /* + Scale the dot product from -1..1 to 0..1. Will result in a sound directly in front losing power + by being played at 0.5 gain. + */ + d = (d + 1) * 0.5f; /* -1..1 to 0..1 */ + d = ma_max(d, dMin); + pSpatializer->pNewChannelGainsOut[iChannel] *= d; + } + #else + { + /* + Only reduce the volume of the sound if it's on the opposite side. This path keeps the volume more + consistent, but comes at the expense of a worse sense of space and positioning. + */ + if (d < 0) { + d += 1; /* Move into the positive range. */ + d = ma_max(d, dMin); + channelGainsOut[iChannel] *= d; + } + } + #endif + } + } else { + /* Assume the sound is right on top of us. Don't do any panning. */ + } + + /* Now we need to apply the volume to each channel. This needs to run through the gainer to ensure we get a smooth volume transition. */ + ma_gainer_set_gains(&pSpatializer->gainer, pSpatializer->pNewChannelGainsOut); + ma_gainer_process_pcm_frames(&pSpatializer->gainer, pFramesOut, pFramesOut, frameCount); + + /* + Before leaving we'll want to update our doppler pitch so that the caller can apply some + pitch shifting if they desire. Note that we need to negate the relative position here + because the doppler calculation needs to be source-to-listener, but ours is listener-to- + source. + */ + if (dopplerFactor > 0) { + pSpatializer->dopplerPitch = ma_doppler_pitch(ma_vec3f_sub(ma_spatializer_listener_get_position(pListener), ma_spatializer_get_position(pSpatializer)), ma_spatializer_get_velocity(pSpatializer), listenerVel, speedOfSound, dopplerFactor); + } else { + pSpatializer->dopplerPitch = 1; + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_spatializer_set_master_volume(ma_spatializer* pSpatializer, float volume) +{ + if (pSpatializer == NULL) { + return MA_INVALID_ARGS; + } + + return ma_gainer_set_master_volume(&pSpatializer->gainer, volume); +} + +MA_API ma_result ma_spatializer_get_master_volume(const ma_spatializer* pSpatializer, float* pVolume) +{ + if (pSpatializer == NULL) { + return MA_INVALID_ARGS; + } + + return ma_gainer_get_master_volume(&pSpatializer->gainer, pVolume); +} + +MA_API ma_uint32 ma_spatializer_get_input_channels(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 0; + } + + return pSpatializer->channelsIn; +} + +MA_API ma_uint32 ma_spatializer_get_output_channels(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 0; + } + + return pSpatializer->channelsOut; +} + +MA_API void ma_spatializer_set_attenuation_model(ma_spatializer* pSpatializer, ma_attenuation_model attenuationModel) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_i32(&pSpatializer->attenuationModel, attenuationModel); +} + +MA_API ma_attenuation_model ma_spatializer_get_attenuation_model(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return ma_attenuation_model_none; + } + + return (ma_attenuation_model)ma_atomic_load_i32(&pSpatializer->attenuationModel); +} + +MA_API void ma_spatializer_set_positioning(ma_spatializer* pSpatializer, ma_positioning positioning) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_i32(&pSpatializer->positioning, positioning); +} + +MA_API ma_positioning ma_spatializer_get_positioning(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return ma_positioning_absolute; + } + + return (ma_positioning)ma_atomic_load_i32(&pSpatializer->positioning); +} + +MA_API void ma_spatializer_set_rolloff(ma_spatializer* pSpatializer, float rolloff) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_f32(&pSpatializer->rolloff, rolloff); +} + +MA_API float ma_spatializer_get_rolloff(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 0; + } + + return ma_atomic_load_f32(&pSpatializer->rolloff); +} + +MA_API void ma_spatializer_set_min_gain(ma_spatializer* pSpatializer, float minGain) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_f32(&pSpatializer->minGain, minGain); +} + +MA_API float ma_spatializer_get_min_gain(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 0; + } + + return ma_atomic_load_f32(&pSpatializer->minGain); +} + +MA_API void ma_spatializer_set_max_gain(ma_spatializer* pSpatializer, float maxGain) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_f32(&pSpatializer->maxGain, maxGain); +} + +MA_API float ma_spatializer_get_max_gain(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 0; + } + + return ma_atomic_load_f32(&pSpatializer->maxGain); +} + +MA_API void ma_spatializer_set_min_distance(ma_spatializer* pSpatializer, float minDistance) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_f32(&pSpatializer->minDistance, minDistance); +} + +MA_API float ma_spatializer_get_min_distance(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 0; + } + + return ma_atomic_load_f32(&pSpatializer->minDistance); +} + +MA_API void ma_spatializer_set_max_distance(ma_spatializer* pSpatializer, float maxDistance) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_f32(&pSpatializer->maxDistance, maxDistance); +} + +MA_API float ma_spatializer_get_max_distance(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 0; + } + + return ma_atomic_load_f32(&pSpatializer->maxDistance); +} + +MA_API void ma_spatializer_set_cone(ma_spatializer* pSpatializer, float innerAngleInRadians, float outerAngleInRadians, float outerGain) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_f32(&pSpatializer->coneInnerAngleInRadians, innerAngleInRadians); + ma_atomic_exchange_f32(&pSpatializer->coneOuterAngleInRadians, outerAngleInRadians); + ma_atomic_exchange_f32(&pSpatializer->coneOuterGain, outerGain); +} + +MA_API void ma_spatializer_get_cone(const ma_spatializer* pSpatializer, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain) +{ + if (pSpatializer == NULL) { + return; + } + + if (pInnerAngleInRadians != NULL) { + *pInnerAngleInRadians = ma_atomic_load_f32(&pSpatializer->coneInnerAngleInRadians); + } + + if (pOuterAngleInRadians != NULL) { + *pOuterAngleInRadians = ma_atomic_load_f32(&pSpatializer->coneOuterAngleInRadians); + } + + if (pOuterGain != NULL) { + *pOuterGain = ma_atomic_load_f32(&pSpatializer->coneOuterGain); + } +} + +MA_API void ma_spatializer_set_doppler_factor(ma_spatializer* pSpatializer, float dopplerFactor) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_f32(&pSpatializer->dopplerFactor, dopplerFactor); +} + +MA_API float ma_spatializer_get_doppler_factor(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 1; + } + + return ma_atomic_load_f32(&pSpatializer->dopplerFactor); +} + +MA_API void ma_spatializer_set_directional_attenuation_factor(ma_spatializer* pSpatializer, float directionalAttenuationFactor) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_exchange_f32(&pSpatializer->directionalAttenuationFactor, directionalAttenuationFactor); +} + +MA_API float ma_spatializer_get_directional_attenuation_factor(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return 1; + } + + return ma_atomic_load_f32(&pSpatializer->directionalAttenuationFactor); +} + +MA_API void ma_spatializer_set_position(ma_spatializer* pSpatializer, float x, float y, float z) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_vec3f_set(&pSpatializer->position, ma_vec3f_init_3f(x, y, z)); +} + +MA_API ma_vec3f ma_spatializer_get_position(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_atomic_vec3f_get((ma_atomic_vec3f*)&pSpatializer->position); /* Naughty const-cast. It's just for atomically loading the vec3 which should be safe. */ +} + +MA_API void ma_spatializer_set_direction(ma_spatializer* pSpatializer, float x, float y, float z) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_vec3f_set(&pSpatializer->direction, ma_vec3f_init_3f(x, y, z)); +} + +MA_API ma_vec3f ma_spatializer_get_direction(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return ma_vec3f_init_3f(0, 0, -1); + } + + return ma_atomic_vec3f_get((ma_atomic_vec3f*)&pSpatializer->direction); /* Naughty const-cast. It's just for atomically loading the vec3 which should be safe. */ +} + +MA_API void ma_spatializer_set_velocity(ma_spatializer* pSpatializer, float x, float y, float z) +{ + if (pSpatializer == NULL) { + return; + } + + ma_atomic_vec3f_set(&pSpatializer->velocity, ma_vec3f_init_3f(x, y, z)); +} + +MA_API ma_vec3f ma_spatializer_get_velocity(const ma_spatializer* pSpatializer) +{ + if (pSpatializer == NULL) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_atomic_vec3f_get((ma_atomic_vec3f*)&pSpatializer->velocity); /* Naughty const-cast. It's just for atomically loading the vec3 which should be safe. */ +} + +MA_API void ma_spatializer_get_relative_position_and_direction(const ma_spatializer* pSpatializer, const ma_spatializer_listener* pListener, ma_vec3f* pRelativePos, ma_vec3f* pRelativeDir) +{ + if (pRelativePos != NULL) { + pRelativePos->x = 0; + pRelativePos->y = 0; + pRelativePos->z = 0; + } + + if (pRelativeDir != NULL) { + pRelativeDir->x = 0; + pRelativeDir->y = 0; + pRelativeDir->z = -1; + } + + if (pSpatializer == NULL) { + return; + } + + if (pListener == NULL || ma_spatializer_get_positioning(pSpatializer) == ma_positioning_relative) { + /* There's no listener or we're using relative positioning. */ + if (pRelativePos != NULL) { + *pRelativePos = ma_spatializer_get_position(pSpatializer); + } + if (pRelativeDir != NULL) { + *pRelativeDir = ma_spatializer_get_direction(pSpatializer); + } + } else { + ma_vec3f spatializerPosition; + ma_vec3f spatializerDirection; + ma_vec3f listenerPosition; + ma_vec3f listenerDirection; + ma_vec3f v; + ma_vec3f axisX; + ma_vec3f axisY; + ma_vec3f axisZ; + float m[4][4]; + + spatializerPosition = ma_spatializer_get_position(pSpatializer); + spatializerDirection = ma_spatializer_get_direction(pSpatializer); + listenerPosition = ma_spatializer_listener_get_position(pListener); + listenerDirection = ma_spatializer_listener_get_direction(pListener); + + /* + We need to calculate the right vector from our forward and up vectors. This is done with + a cross product. + */ + axisZ = ma_vec3f_normalize(listenerDirection); /* Normalization required here because we can't trust the caller. */ + axisX = ma_vec3f_normalize(ma_vec3f_cross(axisZ, pListener->config.worldUp)); /* Normalization required here because the world up vector may not be perpendicular with the forward vector. */ + + /* + The calculation of axisX above can result in a zero-length vector if the listener is + looking straight up on the Y axis. We'll need to fall back to a +X in this case so that + the calculations below don't fall apart. This is where a quaternion based listener and + sound orientation would come in handy. + */ + if (ma_vec3f_len2(axisX) == 0) { + axisX = ma_vec3f_init_3f(1, 0, 0); + } + + axisY = ma_vec3f_cross(axisX, axisZ); /* No normalization is required here because axisX and axisZ are unit length and perpendicular. */ + + /* + We need to swap the X axis if we're left handed because otherwise the cross product above + will have resulted in it pointing in the wrong direction (right handed was assumed in the + cross products above). + */ + if (pListener->config.handedness == ma_handedness_left) { + axisX = ma_vec3f_neg(axisX); + } + + /* Lookat. */ + m[0][0] = axisX.x; m[1][0] = axisX.y; m[2][0] = axisX.z; m[3][0] = -ma_vec3f_dot(axisX, listenerPosition); + m[0][1] = axisY.x; m[1][1] = axisY.y; m[2][1] = axisY.z; m[3][1] = -ma_vec3f_dot(axisY, listenerPosition); + m[0][2] = -axisZ.x; m[1][2] = -axisZ.y; m[2][2] = -axisZ.z; m[3][2] = -ma_vec3f_dot(ma_vec3f_neg(axisZ), listenerPosition); + m[0][3] = 0; m[1][3] = 0; m[2][3] = 0; m[3][3] = 1; + + /* + Multiply the lookat matrix by the spatializer position to transform it to listener + space. This allows calculations to work based on the sound being relative to the + origin which makes things simpler. + */ + if (pRelativePos != NULL) { + v = spatializerPosition; + pRelativePos->x = m[0][0] * v.x + m[1][0] * v.y + m[2][0] * v.z + m[3][0] * 1; + pRelativePos->y = m[0][1] * v.x + m[1][1] * v.y + m[2][1] * v.z + m[3][1] * 1; + pRelativePos->z = m[0][2] * v.x + m[1][2] * v.y + m[2][2] * v.z + m[3][2] * 1; + } + + /* + The direction of the sound needs to also be transformed so that it's relative to the + rotation of the listener. + */ + if (pRelativeDir != NULL) { + v = spatializerDirection; + pRelativeDir->x = m[0][0] * v.x + m[1][0] * v.y + m[2][0] * v.z; + pRelativeDir->y = m[0][1] * v.x + m[1][1] * v.y + m[2][1] * v.z; + pRelativeDir->z = m[0][2] * v.x + m[1][2] * v.y + m[2][2] * v.z; + } + } +} + + + + +/************************************************************************************************************************************************************** + +Resampling + +**************************************************************************************************************************************************************/ +MA_API ma_linear_resampler_config ma_linear_resampler_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut) +{ + ma_linear_resampler_config config; + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRateIn = sampleRateIn; + config.sampleRateOut = sampleRateOut; + config.lpfOrder = ma_min(MA_DEFAULT_RESAMPLER_LPF_ORDER, MA_MAX_FILTER_ORDER); + config.lpfNyquistFactor = 1; + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t x0Offset; + size_t x1Offset; + size_t lpfOffset; +} ma_linear_resampler_heap_layout; + + +static void ma_linear_resampler_adjust_timer_for_new_rate(ma_linear_resampler* pResampler, ma_uint32 oldSampleRateOut, ma_uint32 newSampleRateOut) +{ + /* + So what's happening here? Basically we need to adjust the fractional component of the time advance based on the new rate. The old time advance will + be based on the old sample rate, but we are needing to adjust it to that it's based on the new sample rate. + */ + ma_uint32 oldRateTimeWhole = pResampler->inTimeFrac / oldSampleRateOut; /* <-- This should almost never be anything other than 0, but leaving it here to make this more general and robust just in case. */ + ma_uint32 oldRateTimeFract = pResampler->inTimeFrac % oldSampleRateOut; + + pResampler->inTimeFrac = + (oldRateTimeWhole * newSampleRateOut) + + ((oldRateTimeFract * newSampleRateOut) / oldSampleRateOut); + + /* Make sure the fractional part is less than the output sample rate. */ + pResampler->inTimeInt += pResampler->inTimeFrac / pResampler->config.sampleRateOut; + pResampler->inTimeFrac = pResampler->inTimeFrac % pResampler->config.sampleRateOut; +} + +static ma_result ma_linear_resampler_set_rate_internal(ma_linear_resampler* pResampler, void* pHeap, ma_linear_resampler_heap_layout* pHeapLayout, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut, ma_bool32 isResamplerAlreadyInitialized) +{ + ma_result result; + ma_uint32 gcf; + ma_uint32 lpfSampleRate; + double lpfCutoffFrequency; + ma_lpf_config lpfConfig; + ma_uint32 oldSampleRateOut; /* Required for adjusting time advance down the bottom. */ + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (sampleRateIn == 0 || sampleRateOut == 0) { + return MA_INVALID_ARGS; + } + + oldSampleRateOut = pResampler->config.sampleRateOut; + + pResampler->config.sampleRateIn = sampleRateIn; + pResampler->config.sampleRateOut = sampleRateOut; + + /* Simplify the sample rate. */ + gcf = ma_gcf_u32(pResampler->config.sampleRateIn, pResampler->config.sampleRateOut); + pResampler->config.sampleRateIn /= gcf; + pResampler->config.sampleRateOut /= gcf; + + /* Always initialize the low-pass filter, even when the order is 0. */ + if (pResampler->config.lpfOrder > MA_MAX_FILTER_ORDER) { + return MA_INVALID_ARGS; + } + + lpfSampleRate = (ma_uint32)(ma_max(pResampler->config.sampleRateIn, pResampler->config.sampleRateOut)); + lpfCutoffFrequency = ( double)(ma_min(pResampler->config.sampleRateIn, pResampler->config.sampleRateOut) * 0.5 * pResampler->config.lpfNyquistFactor); + + lpfConfig = ma_lpf_config_init(pResampler->config.format, pResampler->config.channels, lpfSampleRate, lpfCutoffFrequency, pResampler->config.lpfOrder); + + /* + If the resampler is already initialized we don't want to do a fresh initialization of the low-pass filter because it will result in the cached frames + getting cleared. Instead we re-initialize the filter which will maintain any cached frames. + */ + if (isResamplerAlreadyInitialized) { + result = ma_lpf_reinit(&lpfConfig, &pResampler->lpf); + } else { + result = ma_lpf_init_preallocated(&lpfConfig, ma_offset_ptr(pHeap, pHeapLayout->lpfOffset), &pResampler->lpf); + } + + if (result != MA_SUCCESS) { + return result; + } + + + pResampler->inAdvanceInt = pResampler->config.sampleRateIn / pResampler->config.sampleRateOut; + pResampler->inAdvanceFrac = pResampler->config.sampleRateIn % pResampler->config.sampleRateOut; + + /* Our timer was based on the old rate. We need to adjust it so that it's based on the new rate. */ + ma_linear_resampler_adjust_timer_for_new_rate(pResampler, oldSampleRateOut, pResampler->config.sampleRateOut); + + return MA_SUCCESS; +} + +static ma_result ma_linear_resampler_get_heap_layout(const ma_linear_resampler_config* pConfig, ma_linear_resampler_heap_layout* pHeapLayout) +{ + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->format != ma_format_f32 && pConfig->format != ma_format_s16) { + return MA_INVALID_ARGS; + } + + if (pConfig->channels == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* x0 */ + pHeapLayout->x0Offset = pHeapLayout->sizeInBytes; + if (pConfig->format == ma_format_f32) { + pHeapLayout->sizeInBytes += sizeof(float) * pConfig->channels; + } else { + pHeapLayout->sizeInBytes += sizeof(ma_int16) * pConfig->channels; + } + + /* x1 */ + pHeapLayout->x1Offset = pHeapLayout->sizeInBytes; + if (pConfig->format == ma_format_f32) { + pHeapLayout->sizeInBytes += sizeof(float) * pConfig->channels; + } else { + pHeapLayout->sizeInBytes += sizeof(ma_int16) * pConfig->channels; + } + + /* LPF */ + pHeapLayout->lpfOffset = ma_align_64(pHeapLayout->sizeInBytes); + { + ma_result result; + size_t lpfHeapSizeInBytes; + ma_lpf_config lpfConfig = ma_lpf_config_init(pConfig->format, pConfig->channels, 1, 1, pConfig->lpfOrder); /* Sample rate and cutoff frequency do not matter. */ + + result = ma_lpf_get_heap_size(&lpfConfig, &lpfHeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes += lpfHeapSizeInBytes; + } + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +MA_API ma_result ma_linear_resampler_get_heap_size(const ma_linear_resampler_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_linear_resampler_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_linear_resampler_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_linear_resampler_init_preallocated(const ma_linear_resampler_config* pConfig, void* pHeap, ma_linear_resampler* pResampler) +{ + ma_result result; + ma_linear_resampler_heap_layout heapLayout; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pResampler); + + result = ma_linear_resampler_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pResampler->config = *pConfig; + + pResampler->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + if (pConfig->format == ma_format_f32) { + pResampler->x0.f32 = (float*)ma_offset_ptr(pHeap, heapLayout.x0Offset); + pResampler->x1.f32 = (float*)ma_offset_ptr(pHeap, heapLayout.x1Offset); + } else { + pResampler->x0.s16 = (ma_int16*)ma_offset_ptr(pHeap, heapLayout.x0Offset); + pResampler->x1.s16 = (ma_int16*)ma_offset_ptr(pHeap, heapLayout.x1Offset); + } + + /* Setting the rate will set up the filter and time advances for us. */ + result = ma_linear_resampler_set_rate_internal(pResampler, pHeap, &heapLayout, pConfig->sampleRateIn, pConfig->sampleRateOut, /* isResamplerAlreadyInitialized = */ MA_FALSE); + if (result != MA_SUCCESS) { + return result; + } + + pResampler->inTimeInt = 1; /* Set this to one to force an input sample to always be loaded for the first output frame. */ + pResampler->inTimeFrac = 0; + + return MA_SUCCESS; +} + +MA_API ma_result ma_linear_resampler_init(const ma_linear_resampler_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_linear_resampler* pResampler) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_linear_resampler_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_linear_resampler_init_preallocated(pConfig, pHeap, pResampler); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pResampler->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_linear_resampler_uninit(ma_linear_resampler* pResampler, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pResampler == NULL) { + return; + } + + ma_lpf_uninit(&pResampler->lpf, pAllocationCallbacks); + + if (pResampler->_ownsHeap) { + ma_free(pResampler->_pHeap, pAllocationCallbacks); + } +} + +static MA_INLINE ma_int16 ma_linear_resampler_mix_s16(ma_int16 x, ma_int16 y, ma_int32 a, const ma_int32 shift) +{ + ma_int32 b; + ma_int32 c; + ma_int32 r; + + MA_ASSERT(a <= (1<> shift); +} + +static void ma_linear_resampler_interpolate_frame_s16(ma_linear_resampler* pResampler, ma_int16* MA_RESTRICT pFrameOut) +{ + ma_uint32 c; + ma_uint32 a; + const ma_uint32 channels = pResampler->config.channels; + const ma_uint32 shift = 12; + + MA_ASSERT(pResampler != NULL); + MA_ASSERT(pFrameOut != NULL); + + a = (pResampler->inTimeFrac << shift) / pResampler->config.sampleRateOut; + + MA_ASSUME(channels > 0); + for (c = 0; c < channels; c += 1) { + ma_int16 s = ma_linear_resampler_mix_s16(pResampler->x0.s16[c], pResampler->x1.s16[c], a, shift); + pFrameOut[c] = s; + } +} + + +static void ma_linear_resampler_interpolate_frame_f32(ma_linear_resampler* pResampler, float* MA_RESTRICT pFrameOut) +{ + ma_uint32 c; + float a; + const ma_uint32 channels = pResampler->config.channels; + + MA_ASSERT(pResampler != NULL); + MA_ASSERT(pFrameOut != NULL); + + a = (float)pResampler->inTimeFrac / pResampler->config.sampleRateOut; + + MA_ASSUME(channels > 0); + for (c = 0; c < channels; c += 1) { + float s = ma_mix_f32_fast(pResampler->x0.f32[c], pResampler->x1.f32[c], a); + pFrameOut[c] = s; + } +} + +static ma_result ma_linear_resampler_process_pcm_frames_s16_downsample(ma_linear_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + const ma_int16* pFramesInS16; + /* */ ma_int16* pFramesOutS16; + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 framesProcessedIn; + ma_uint64 framesProcessedOut; + + MA_ASSERT(pResampler != NULL); + MA_ASSERT(pFrameCountIn != NULL); + MA_ASSERT(pFrameCountOut != NULL); + + pFramesInS16 = (const ma_int16*)pFramesIn; + pFramesOutS16 = ( ma_int16*)pFramesOut; + frameCountIn = *pFrameCountIn; + frameCountOut = *pFrameCountOut; + framesProcessedIn = 0; + framesProcessedOut = 0; + + while (framesProcessedOut < frameCountOut) { + /* Before interpolating we need to load the buffers. When doing this we need to ensure we run every input sample through the filter. */ + while (pResampler->inTimeInt > 0 && frameCountIn > framesProcessedIn) { + ma_uint32 iChannel; + + if (pFramesInS16 != NULL) { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.s16[iChannel] = pResampler->x1.s16[iChannel]; + pResampler->x1.s16[iChannel] = pFramesInS16[iChannel]; + } + pFramesInS16 += pResampler->config.channels; + } else { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.s16[iChannel] = pResampler->x1.s16[iChannel]; + pResampler->x1.s16[iChannel] = 0; + } + } + + /* Filter. Do not apply filtering if sample rates are the same or else you'll get dangerous glitching. */ + if (pResampler->config.sampleRateIn != pResampler->config.sampleRateOut) { + ma_lpf_process_pcm_frame_s16(&pResampler->lpf, pResampler->x1.s16, pResampler->x1.s16); + } + + framesProcessedIn += 1; + pResampler->inTimeInt -= 1; + } + + if (pResampler->inTimeInt > 0) { + break; /* Ran out of input data. */ + } + + /* Getting here means the frames have been loaded and filtered and we can generate the next output frame. */ + if (pFramesOutS16 != NULL) { + MA_ASSERT(pResampler->inTimeInt == 0); + ma_linear_resampler_interpolate_frame_s16(pResampler, pFramesOutS16); + + pFramesOutS16 += pResampler->config.channels; + } + + framesProcessedOut += 1; + + /* Advance time forward. */ + pResampler->inTimeInt += pResampler->inAdvanceInt; + pResampler->inTimeFrac += pResampler->inAdvanceFrac; + if (pResampler->inTimeFrac >= pResampler->config.sampleRateOut) { + pResampler->inTimeFrac -= pResampler->config.sampleRateOut; + pResampler->inTimeInt += 1; + } + } + + *pFrameCountIn = framesProcessedIn; + *pFrameCountOut = framesProcessedOut; + + return MA_SUCCESS; +} + +static ma_result ma_linear_resampler_process_pcm_frames_s16_upsample(ma_linear_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + const ma_int16* pFramesInS16; + /* */ ma_int16* pFramesOutS16; + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 framesProcessedIn; + ma_uint64 framesProcessedOut; + + MA_ASSERT(pResampler != NULL); + MA_ASSERT(pFrameCountIn != NULL); + MA_ASSERT(pFrameCountOut != NULL); + + pFramesInS16 = (const ma_int16*)pFramesIn; + pFramesOutS16 = ( ma_int16*)pFramesOut; + frameCountIn = *pFrameCountIn; + frameCountOut = *pFrameCountOut; + framesProcessedIn = 0; + framesProcessedOut = 0; + + while (framesProcessedOut < frameCountOut) { + /* Before interpolating we need to load the buffers. */ + while (pResampler->inTimeInt > 0 && frameCountIn > framesProcessedIn) { + ma_uint32 iChannel; + + if (pFramesInS16 != NULL) { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.s16[iChannel] = pResampler->x1.s16[iChannel]; + pResampler->x1.s16[iChannel] = pFramesInS16[iChannel]; + } + pFramesInS16 += pResampler->config.channels; + } else { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.s16[iChannel] = pResampler->x1.s16[iChannel]; + pResampler->x1.s16[iChannel] = 0; + } + } + + framesProcessedIn += 1; + pResampler->inTimeInt -= 1; + } + + if (pResampler->inTimeInt > 0) { + break; /* Ran out of input data. */ + } + + /* Getting here means the frames have been loaded and we can generate the next output frame. */ + if (pFramesOutS16 != NULL) { + MA_ASSERT(pResampler->inTimeInt == 0); + ma_linear_resampler_interpolate_frame_s16(pResampler, pFramesOutS16); + + /* Filter. Do not apply filtering if sample rates are the same or else you'll get dangerous glitching. */ + if (pResampler->config.sampleRateIn != pResampler->config.sampleRateOut) { + ma_lpf_process_pcm_frame_s16(&pResampler->lpf, pFramesOutS16, pFramesOutS16); + } + + pFramesOutS16 += pResampler->config.channels; + } + + framesProcessedOut += 1; + + /* Advance time forward. */ + pResampler->inTimeInt += pResampler->inAdvanceInt; + pResampler->inTimeFrac += pResampler->inAdvanceFrac; + if (pResampler->inTimeFrac >= pResampler->config.sampleRateOut) { + pResampler->inTimeFrac -= pResampler->config.sampleRateOut; + pResampler->inTimeInt += 1; + } + } + + *pFrameCountIn = framesProcessedIn; + *pFrameCountOut = framesProcessedOut; + + return MA_SUCCESS; +} + +static ma_result ma_linear_resampler_process_pcm_frames_s16(ma_linear_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + MA_ASSERT(pResampler != NULL); + + if (pResampler->config.sampleRateIn > pResampler->config.sampleRateOut) { + return ma_linear_resampler_process_pcm_frames_s16_downsample(pResampler, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + } else { + return ma_linear_resampler_process_pcm_frames_s16_upsample(pResampler, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + } +} + + +static ma_result ma_linear_resampler_process_pcm_frames_f32_downsample(ma_linear_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + const float* pFramesInF32; + /* */ float* pFramesOutF32; + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 framesProcessedIn; + ma_uint64 framesProcessedOut; + + MA_ASSERT(pResampler != NULL); + MA_ASSERT(pFrameCountIn != NULL); + MA_ASSERT(pFrameCountOut != NULL); + + pFramesInF32 = (const float*)pFramesIn; + pFramesOutF32 = ( float*)pFramesOut; + frameCountIn = *pFrameCountIn; + frameCountOut = *pFrameCountOut; + framesProcessedIn = 0; + framesProcessedOut = 0; + + while (framesProcessedOut < frameCountOut) { + /* Before interpolating we need to load the buffers. When doing this we need to ensure we run every input sample through the filter. */ + while (pResampler->inTimeInt > 0 && frameCountIn > framesProcessedIn) { + ma_uint32 iChannel; + + if (pFramesInF32 != NULL) { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.f32[iChannel] = pResampler->x1.f32[iChannel]; + pResampler->x1.f32[iChannel] = pFramesInF32[iChannel]; + } + pFramesInF32 += pResampler->config.channels; + } else { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.f32[iChannel] = pResampler->x1.f32[iChannel]; + pResampler->x1.f32[iChannel] = 0; + } + } + + /* Filter. Do not apply filtering if sample rates are the same or else you'll get dangerous glitching. */ + if (pResampler->config.sampleRateIn != pResampler->config.sampleRateOut) { + ma_lpf_process_pcm_frame_f32(&pResampler->lpf, pResampler->x1.f32, pResampler->x1.f32); + } + + framesProcessedIn += 1; + pResampler->inTimeInt -= 1; + } + + if (pResampler->inTimeInt > 0) { + break; /* Ran out of input data. */ + } + + /* Getting here means the frames have been loaded and filtered and we can generate the next output frame. */ + if (pFramesOutF32 != NULL) { + MA_ASSERT(pResampler->inTimeInt == 0); + ma_linear_resampler_interpolate_frame_f32(pResampler, pFramesOutF32); + + pFramesOutF32 += pResampler->config.channels; + } + + framesProcessedOut += 1; + + /* Advance time forward. */ + pResampler->inTimeInt += pResampler->inAdvanceInt; + pResampler->inTimeFrac += pResampler->inAdvanceFrac; + if (pResampler->inTimeFrac >= pResampler->config.sampleRateOut) { + pResampler->inTimeFrac -= pResampler->config.sampleRateOut; + pResampler->inTimeInt += 1; + } + } + + *pFrameCountIn = framesProcessedIn; + *pFrameCountOut = framesProcessedOut; + + return MA_SUCCESS; +} + +static ma_result ma_linear_resampler_process_pcm_frames_f32_upsample(ma_linear_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + const float* pFramesInF32; + /* */ float* pFramesOutF32; + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 framesProcessedIn; + ma_uint64 framesProcessedOut; + + MA_ASSERT(pResampler != NULL); + MA_ASSERT(pFrameCountIn != NULL); + MA_ASSERT(pFrameCountOut != NULL); + + pFramesInF32 = (const float*)pFramesIn; + pFramesOutF32 = ( float*)pFramesOut; + frameCountIn = *pFrameCountIn; + frameCountOut = *pFrameCountOut; + framesProcessedIn = 0; + framesProcessedOut = 0; + + while (framesProcessedOut < frameCountOut) { + /* Before interpolating we need to load the buffers. */ + while (pResampler->inTimeInt > 0 && frameCountIn > framesProcessedIn) { + ma_uint32 iChannel; + + if (pFramesInF32 != NULL) { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.f32[iChannel] = pResampler->x1.f32[iChannel]; + pResampler->x1.f32[iChannel] = pFramesInF32[iChannel]; + } + pFramesInF32 += pResampler->config.channels; + } else { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.f32[iChannel] = pResampler->x1.f32[iChannel]; + pResampler->x1.f32[iChannel] = 0; + } + } + + framesProcessedIn += 1; + pResampler->inTimeInt -= 1; + } + + if (pResampler->inTimeInt > 0) { + break; /* Ran out of input data. */ + } + + /* Getting here means the frames have been loaded and we can generate the next output frame. */ + if (pFramesOutF32 != NULL) { + MA_ASSERT(pResampler->inTimeInt == 0); + ma_linear_resampler_interpolate_frame_f32(pResampler, pFramesOutF32); + + /* Filter. Do not apply filtering if sample rates are the same or else you'll get dangerous glitching. */ + if (pResampler->config.sampleRateIn != pResampler->config.sampleRateOut) { + ma_lpf_process_pcm_frame_f32(&pResampler->lpf, pFramesOutF32, pFramesOutF32); + } + + pFramesOutF32 += pResampler->config.channels; + } + + framesProcessedOut += 1; + + /* Advance time forward. */ + pResampler->inTimeInt += pResampler->inAdvanceInt; + pResampler->inTimeFrac += pResampler->inAdvanceFrac; + if (pResampler->inTimeFrac >= pResampler->config.sampleRateOut) { + pResampler->inTimeFrac -= pResampler->config.sampleRateOut; + pResampler->inTimeInt += 1; + } + } + + *pFrameCountIn = framesProcessedIn; + *pFrameCountOut = framesProcessedOut; + + return MA_SUCCESS; +} + +static ma_result ma_linear_resampler_process_pcm_frames_f32(ma_linear_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + MA_ASSERT(pResampler != NULL); + + if (pResampler->config.sampleRateIn > pResampler->config.sampleRateOut) { + return ma_linear_resampler_process_pcm_frames_f32_downsample(pResampler, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + } else { + return ma_linear_resampler_process_pcm_frames_f32_upsample(pResampler, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + } +} + + +MA_API ma_result ma_linear_resampler_process_pcm_frames(ma_linear_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + /* */ if (pResampler->config.format == ma_format_s16) { + return ma_linear_resampler_process_pcm_frames_s16(pResampler, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + } else if (pResampler->config.format == ma_format_f32) { + return ma_linear_resampler_process_pcm_frames_f32(pResampler, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + } else { + /* Should never get here. Getting here means the format is not supported and you didn't check the return value of ma_linear_resampler_init(). */ + MA_ASSERT(MA_FALSE); + return MA_INVALID_ARGS; + } +} + + +MA_API ma_result ma_linear_resampler_set_rate(ma_linear_resampler* pResampler, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut) +{ + return ma_linear_resampler_set_rate_internal(pResampler, NULL, NULL, sampleRateIn, sampleRateOut, /* isResamplerAlreadyInitialized = */ MA_TRUE); +} + +MA_API ma_result ma_linear_resampler_set_rate_ratio(ma_linear_resampler* pResampler, float ratioInOut) +{ + ma_uint32 n; + ma_uint32 d; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (ratioInOut <= 0) { + return MA_INVALID_ARGS; + } + + d = 1000000; + n = (ma_uint32)(ratioInOut * d); + + if (n == 0) { + return MA_INVALID_ARGS; /* Ratio too small. */ + } + + MA_ASSERT(n != 0); + + return ma_linear_resampler_set_rate(pResampler, n, d); +} + +MA_API ma_uint64 ma_linear_resampler_get_input_latency(const ma_linear_resampler* pResampler) +{ + if (pResampler == NULL) { + return 0; + } + + return 1 + ma_lpf_get_latency(&pResampler->lpf); +} + +MA_API ma_uint64 ma_linear_resampler_get_output_latency(const ma_linear_resampler* pResampler) +{ + if (pResampler == NULL) { + return 0; + } + + return ma_linear_resampler_get_input_latency(pResampler) * pResampler->config.sampleRateOut / pResampler->config.sampleRateIn; +} + +MA_API ma_result ma_linear_resampler_get_required_input_frame_count(const ma_linear_resampler* pResampler, ma_uint64 outputFrameCount, ma_uint64* pInputFrameCount) +{ + ma_uint64 inputFrameCount; + + if (pInputFrameCount == NULL) { + return MA_INVALID_ARGS; + } + + *pInputFrameCount = 0; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (outputFrameCount == 0) { + return MA_SUCCESS; + } + + /* Any whole input frames are consumed before the first output frame is generated. */ + inputFrameCount = pResampler->inTimeInt; + outputFrameCount -= 1; + + /* The rest of the output frames can be calculated in constant time. */ + inputFrameCount += outputFrameCount * pResampler->inAdvanceInt; + inputFrameCount += (pResampler->inTimeFrac + (outputFrameCount * pResampler->inAdvanceFrac)) / pResampler->config.sampleRateOut; + + *pInputFrameCount = inputFrameCount; + + return MA_SUCCESS; +} + +MA_API ma_result ma_linear_resampler_get_expected_output_frame_count(const ma_linear_resampler* pResampler, ma_uint64 inputFrameCount, ma_uint64* pOutputFrameCount) +{ + ma_uint64 outputFrameCount; + ma_uint64 preliminaryInputFrameCountFromFrac; + ma_uint64 preliminaryInputFrameCount; + + if (pOutputFrameCount == NULL) { + return MA_INVALID_ARGS; + } + + *pOutputFrameCount = 0; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + /* + The first step is to get a preliminary output frame count. This will either be exactly equal to what we need, or less by 1. We need to + determine how many input frames will be consumed by this value. If it's greater than our original input frame count it means we won't + be able to generate an extra frame because we will have run out of input data. Otherwise we will have enough input for the generation + of an extra output frame. This add-by-one logic is necessary due to how the data loading logic works when processing frames. + */ + outputFrameCount = (inputFrameCount * pResampler->config.sampleRateOut) / pResampler->config.sampleRateIn; + + /* + We need to determine how many *whole* input frames will have been processed to generate our preliminary output frame count. This is + used in the logic below to determine whether or not we need to add an extra output frame. + */ + preliminaryInputFrameCountFromFrac = (pResampler->inTimeFrac + outputFrameCount*pResampler->inAdvanceFrac) / pResampler->config.sampleRateOut; + preliminaryInputFrameCount = (pResampler->inTimeInt + outputFrameCount*pResampler->inAdvanceInt ) + preliminaryInputFrameCountFromFrac; + + /* + If the total number of *whole* input frames that would be required to generate our preliminary output frame count is greater than + the amount of whole input frames we have available as input we need to *not* add an extra output frame as there won't be enough data + to actually process. Otherwise we need to add the extra output frame. + */ + if (preliminaryInputFrameCount <= inputFrameCount) { + outputFrameCount += 1; + } + + *pOutputFrameCount = outputFrameCount; + + return MA_SUCCESS; +} + +MA_API ma_result ma_linear_resampler_reset(ma_linear_resampler* pResampler) +{ + ma_uint32 iChannel; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + /* Timers need to be cleared back to zero. */ + pResampler->inTimeInt = 1; /* Set this to one to force an input sample to always be loaded for the first output frame. */ + pResampler->inTimeFrac = 0; + + /* Cached samples need to be cleared. */ + if (pResampler->config.format == ma_format_f32) { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.f32[iChannel] = 0; + pResampler->x1.f32[iChannel] = 0; + } + } else { + for (iChannel = 0; iChannel < pResampler->config.channels; iChannel += 1) { + pResampler->x0.s16[iChannel] = 0; + pResampler->x1.s16[iChannel] = 0; + } + } + + /* The low pass filter needs to have its cache reset. */ + ma_lpf_clear_cache(&pResampler->lpf); + + return MA_SUCCESS; +} + + + +/* Linear resampler backend vtable. */ +static ma_linear_resampler_config ma_resampling_backend_get_config__linear(const ma_resampler_config* pConfig) +{ + ma_linear_resampler_config linearConfig; + + linearConfig = ma_linear_resampler_config_init(pConfig->format, pConfig->channels, pConfig->sampleRateIn, pConfig->sampleRateOut); + linearConfig.lpfOrder = pConfig->linear.lpfOrder; + + return linearConfig; +} + +static ma_result ma_resampling_backend_get_heap_size__linear(void* pUserData, const ma_resampler_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_linear_resampler_config linearConfig; + + (void)pUserData; + + linearConfig = ma_resampling_backend_get_config__linear(pConfig); + + return ma_linear_resampler_get_heap_size(&linearConfig, pHeapSizeInBytes); +} + +static ma_result ma_resampling_backend_init__linear(void* pUserData, const ma_resampler_config* pConfig, void* pHeap, ma_resampling_backend** ppBackend) +{ + ma_resampler* pResampler = (ma_resampler*)pUserData; + ma_result result; + ma_linear_resampler_config linearConfig; + + (void)pUserData; + + linearConfig = ma_resampling_backend_get_config__linear(pConfig); + + result = ma_linear_resampler_init_preallocated(&linearConfig, pHeap, &pResampler->state.linear); + if (result != MA_SUCCESS) { + return result; + } + + *ppBackend = &pResampler->state.linear; + + return MA_SUCCESS; +} + +static void ma_resampling_backend_uninit__linear(void* pUserData, ma_resampling_backend* pBackend, const ma_allocation_callbacks* pAllocationCallbacks) +{ + (void)pUserData; + + ma_linear_resampler_uninit((ma_linear_resampler*)pBackend, pAllocationCallbacks); +} + +static ma_result ma_resampling_backend_process__linear(void* pUserData, ma_resampling_backend* pBackend, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + (void)pUserData; + + return ma_linear_resampler_process_pcm_frames((ma_linear_resampler*)pBackend, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); +} + +static ma_result ma_resampling_backend_set_rate__linear(void* pUserData, ma_resampling_backend* pBackend, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut) +{ + (void)pUserData; + + return ma_linear_resampler_set_rate((ma_linear_resampler*)pBackend, sampleRateIn, sampleRateOut); +} + +static ma_uint64 ma_resampling_backend_get_input_latency__linear(void* pUserData, const ma_resampling_backend* pBackend) +{ + (void)pUserData; + + return ma_linear_resampler_get_input_latency((const ma_linear_resampler*)pBackend); +} + +static ma_uint64 ma_resampling_backend_get_output_latency__linear(void* pUserData, const ma_resampling_backend* pBackend) +{ + (void)pUserData; + + return ma_linear_resampler_get_output_latency((const ma_linear_resampler*)pBackend); +} + +static ma_result ma_resampling_backend_get_required_input_frame_count__linear(void* pUserData, const ma_resampling_backend* pBackend, ma_uint64 outputFrameCount, ma_uint64* pInputFrameCount) +{ + (void)pUserData; + + return ma_linear_resampler_get_required_input_frame_count((const ma_linear_resampler*)pBackend, outputFrameCount, pInputFrameCount); +} + +static ma_result ma_resampling_backend_get_expected_output_frame_count__linear(void* pUserData, const ma_resampling_backend* pBackend, ma_uint64 inputFrameCount, ma_uint64* pOutputFrameCount) +{ + (void)pUserData; + + return ma_linear_resampler_get_expected_output_frame_count((const ma_linear_resampler*)pBackend, inputFrameCount, pOutputFrameCount); +} + +static ma_result ma_resampling_backend_reset__linear(void* pUserData, ma_resampling_backend* pBackend) +{ + (void)pUserData; + + return ma_linear_resampler_reset((ma_linear_resampler*)pBackend); +} + +static ma_resampling_backend_vtable g_ma_linear_resampler_vtable = +{ + ma_resampling_backend_get_heap_size__linear, + ma_resampling_backend_init__linear, + ma_resampling_backend_uninit__linear, + ma_resampling_backend_process__linear, + ma_resampling_backend_set_rate__linear, + ma_resampling_backend_get_input_latency__linear, + ma_resampling_backend_get_output_latency__linear, + ma_resampling_backend_get_required_input_frame_count__linear, + ma_resampling_backend_get_expected_output_frame_count__linear, + ma_resampling_backend_reset__linear +}; + + + +MA_API ma_resampler_config ma_resampler_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut, ma_resample_algorithm algorithm) +{ + ma_resampler_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRateIn = sampleRateIn; + config.sampleRateOut = sampleRateOut; + config.algorithm = algorithm; + + /* Linear. */ + config.linear.lpfOrder = ma_min(MA_DEFAULT_RESAMPLER_LPF_ORDER, MA_MAX_FILTER_ORDER); + + return config; +} + +static ma_result ma_resampler_get_vtable(const ma_resampler_config* pConfig, ma_resampler* pResampler, ma_resampling_backend_vtable** ppVTable, void** ppUserData) +{ + MA_ASSERT(pConfig != NULL); + MA_ASSERT(ppVTable != NULL); + MA_ASSERT(ppUserData != NULL); + + /* Safety. */ + *ppVTable = NULL; + *ppUserData = NULL; + + switch (pConfig->algorithm) + { + case ma_resample_algorithm_linear: + { + *ppVTable = &g_ma_linear_resampler_vtable; + *ppUserData = pResampler; + } break; + + case ma_resample_algorithm_custom: + { + *ppVTable = pConfig->pBackendVTable; + *ppUserData = pConfig->pBackendUserData; + } break; + + default: return MA_INVALID_ARGS; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_resampler_get_heap_size(const ma_resampler_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_resampling_backend_vtable* pVTable; + void* pVTableUserData; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_resampler_get_vtable(pConfig, NULL, &pVTable, &pVTableUserData); + if (result != MA_SUCCESS) { + return result; + } + + if (pVTable == NULL || pVTable->onGetHeapSize == NULL) { + return MA_NOT_IMPLEMENTED; + } + + result = pVTable->onGetHeapSize(pVTableUserData, pConfig, pHeapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_resampler_init_preallocated(const ma_resampler_config* pConfig, void* pHeap, ma_resampler* pResampler) +{ + ma_result result; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pResampler); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + pResampler->_pHeap = pHeap; + pResampler->format = pConfig->format; + pResampler->channels = pConfig->channels; + pResampler->sampleRateIn = pConfig->sampleRateIn; + pResampler->sampleRateOut = pConfig->sampleRateOut; + + result = ma_resampler_get_vtable(pConfig, pResampler, &pResampler->pBackendVTable, &pResampler->pBackendUserData); + if (result != MA_SUCCESS) { + return result; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onInit == NULL) { + return MA_NOT_IMPLEMENTED; /* onInit not implemented. */ + } + + result = pResampler->pBackendVTable->onInit(pResampler->pBackendUserData, pConfig, pHeap, &pResampler->pBackend); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_resampler_init(const ma_resampler_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_resampler* pResampler) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_resampler_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_resampler_init_preallocated(pConfig, pHeap, pResampler); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pResampler->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_resampler_uninit(ma_resampler* pResampler, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pResampler == NULL) { + return; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onUninit == NULL) { + return; + } + + pResampler->pBackendVTable->onUninit(pResampler->pBackendUserData, pResampler->pBackend, pAllocationCallbacks); + + if (pResampler->_ownsHeap) { + ma_free(pResampler->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_resampler_process_pcm_frames(ma_resampler* pResampler, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (pFrameCountOut == NULL && pFrameCountIn == NULL) { + return MA_INVALID_ARGS; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onProcess == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pResampler->pBackendVTable->onProcess(pResampler->pBackendUserData, pResampler->pBackend, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); +} + +MA_API ma_result ma_resampler_set_rate(ma_resampler* pResampler, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut) +{ + ma_result result; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (sampleRateIn == 0 || sampleRateOut == 0) { + return MA_INVALID_ARGS; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onSetRate == NULL) { + return MA_NOT_IMPLEMENTED; + } + + result = pResampler->pBackendVTable->onSetRate(pResampler->pBackendUserData, pResampler->pBackend, sampleRateIn, sampleRateOut); + if (result != MA_SUCCESS) { + return result; + } + + pResampler->sampleRateIn = sampleRateIn; + pResampler->sampleRateOut = sampleRateOut; + + return MA_SUCCESS; +} + +MA_API ma_result ma_resampler_set_rate_ratio(ma_resampler* pResampler, float ratio) +{ + ma_uint32 n; + ma_uint32 d; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (ratio <= 0) { + return MA_INVALID_ARGS; + } + + d = 1000; + n = (ma_uint32)(ratio * d); + + if (n == 0) { + return MA_INVALID_ARGS; /* Ratio too small. */ + } + + MA_ASSERT(n != 0); + + return ma_resampler_set_rate(pResampler, n, d); +} + +MA_API ma_uint64 ma_resampler_get_input_latency(const ma_resampler* pResampler) +{ + if (pResampler == NULL) { + return 0; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onGetInputLatency == NULL) { + return 0; + } + + return pResampler->pBackendVTable->onGetInputLatency(pResampler->pBackendUserData, pResampler->pBackend); +} + +MA_API ma_uint64 ma_resampler_get_output_latency(const ma_resampler* pResampler) +{ + if (pResampler == NULL) { + return 0; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onGetOutputLatency == NULL) { + return 0; + } + + return pResampler->pBackendVTable->onGetOutputLatency(pResampler->pBackendUserData, pResampler->pBackend); +} + +MA_API ma_result ma_resampler_get_required_input_frame_count(const ma_resampler* pResampler, ma_uint64 outputFrameCount, ma_uint64* pInputFrameCount) +{ + if (pInputFrameCount == NULL) { + return MA_INVALID_ARGS; + } + + *pInputFrameCount = 0; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onGetRequiredInputFrameCount == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pResampler->pBackendVTable->onGetRequiredInputFrameCount(pResampler->pBackendUserData, pResampler->pBackend, outputFrameCount, pInputFrameCount); +} + +MA_API ma_result ma_resampler_get_expected_output_frame_count(const ma_resampler* pResampler, ma_uint64 inputFrameCount, ma_uint64* pOutputFrameCount) +{ + if (pOutputFrameCount == NULL) { + return MA_INVALID_ARGS; + } + + *pOutputFrameCount = 0; + + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onGetExpectedOutputFrameCount == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pResampler->pBackendVTable->onGetExpectedOutputFrameCount(pResampler->pBackendUserData, pResampler->pBackend, inputFrameCount, pOutputFrameCount); +} + +MA_API ma_result ma_resampler_reset(ma_resampler* pResampler) +{ + if (pResampler == NULL) { + return MA_INVALID_ARGS; + } + + if (pResampler->pBackendVTable == NULL || pResampler->pBackendVTable->onReset == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pResampler->pBackendVTable->onReset(pResampler->pBackendUserData, pResampler->pBackend); +} + +/************************************************************************************************************************************************************** + +Channel Conversion + +**************************************************************************************************************************************************************/ +#ifndef MA_CHANNEL_CONVERTER_FIXED_POINT_SHIFT +#define MA_CHANNEL_CONVERTER_FIXED_POINT_SHIFT 12 +#endif + +#define MA_PLANE_LEFT 0 +#define MA_PLANE_RIGHT 1 +#define MA_PLANE_FRONT 2 +#define MA_PLANE_BACK 3 +#define MA_PLANE_BOTTOM 4 +#define MA_PLANE_TOP 5 + +static float g_maChannelPlaneRatios[MA_CHANNEL_POSITION_COUNT][6] = { + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_NONE */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_MONO */ + { 0.5f, 0.0f, 0.5f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_FRONT_LEFT */ + { 0.0f, 0.5f, 0.5f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_FRONT_RIGHT */ + { 0.0f, 0.0f, 1.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_FRONT_CENTER */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_LFE */ + { 0.5f, 0.0f, 0.0f, 0.5f, 0.0f, 0.0f}, /* MA_CHANNEL_BACK_LEFT */ + { 0.0f, 0.5f, 0.0f, 0.5f, 0.0f, 0.0f}, /* MA_CHANNEL_BACK_RIGHT */ + { 0.25f, 0.0f, 0.75f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_FRONT_LEFT_CENTER */ + { 0.0f, 0.25f, 0.75f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_FRONT_RIGHT_CENTER */ + { 0.0f, 0.0f, 0.0f, 1.0f, 0.0f, 0.0f}, /* MA_CHANNEL_BACK_CENTER */ + { 1.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_SIDE_LEFT */ + { 0.0f, 1.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_SIDE_RIGHT */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f}, /* MA_CHANNEL_TOP_CENTER */ + { 0.33f, 0.0f, 0.33f, 0.0f, 0.0f, 0.34f}, /* MA_CHANNEL_TOP_FRONT_LEFT */ + { 0.0f, 0.0f, 0.5f, 0.0f, 0.0f, 0.5f}, /* MA_CHANNEL_TOP_FRONT_CENTER */ + { 0.0f, 0.33f, 0.33f, 0.0f, 0.0f, 0.34f}, /* MA_CHANNEL_TOP_FRONT_RIGHT */ + { 0.33f, 0.0f, 0.0f, 0.33f, 0.0f, 0.34f}, /* MA_CHANNEL_TOP_BACK_LEFT */ + { 0.0f, 0.0f, 0.0f, 0.5f, 0.0f, 0.5f}, /* MA_CHANNEL_TOP_BACK_CENTER */ + { 0.0f, 0.33f, 0.0f, 0.33f, 0.0f, 0.34f}, /* MA_CHANNEL_TOP_BACK_RIGHT */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_0 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_1 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_2 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_3 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_4 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_5 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_6 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_7 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_8 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_9 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_10 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_11 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_12 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_13 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_14 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_15 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_16 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_17 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_18 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_19 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_20 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_21 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_22 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_23 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_24 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_25 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_26 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_27 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_28 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_29 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_30 */ + { 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f}, /* MA_CHANNEL_AUX_31 */ +}; + +static float ma_calculate_channel_position_rectangular_weight(ma_channel channelPositionA, ma_channel channelPositionB) +{ + /* + Imagine the following simplified example: You have a single input speaker which is the front/left speaker which you want to convert to + the following output configuration: + + - front/left + - side/left + - back/left + + The front/left output is easy - it the same speaker position so it receives the full contribution of the front/left input. The amount + of contribution to apply to the side/left and back/left speakers, however, is a bit more complicated. + + Imagine the front/left speaker as emitting audio from two planes - the front plane and the left plane. You can think of the front/left + speaker emitting half of its total volume from the front, and the other half from the left. Since part of its volume is being emitted + from the left side, and the side/left and back/left channels also emit audio from the left plane, one would expect that they would + receive some amount of contribution from front/left speaker. The amount of contribution depends on how many planes are shared between + the two speakers. Note that in the examples below I've added a top/front/left speaker as an example just to show how the math works + across 3 spatial dimensions. + + The first thing to do is figure out how each speaker's volume is spread over each of plane: + - front/left: 2 planes (front and left) = 1/2 = half its total volume on each plane + - side/left: 1 plane (left only) = 1/1 = entire volume from left plane + - back/left: 2 planes (back and left) = 1/2 = half its total volume on each plane + - top/front/left: 3 planes (top, front and left) = 1/3 = one third its total volume on each plane + + The amount of volume each channel contributes to each of its planes is what controls how much it is willing to given and take to other + channels on the same plane. The volume that is willing to the given by one channel is multiplied by the volume that is willing to be + taken by the other to produce the final contribution. + */ + + /* Contribution = Sum(Volume to Give * Volume to Take) */ + float contribution = + g_maChannelPlaneRatios[channelPositionA][0] * g_maChannelPlaneRatios[channelPositionB][0] + + g_maChannelPlaneRatios[channelPositionA][1] * g_maChannelPlaneRatios[channelPositionB][1] + + g_maChannelPlaneRatios[channelPositionA][2] * g_maChannelPlaneRatios[channelPositionB][2] + + g_maChannelPlaneRatios[channelPositionA][3] * g_maChannelPlaneRatios[channelPositionB][3] + + g_maChannelPlaneRatios[channelPositionA][4] * g_maChannelPlaneRatios[channelPositionB][4] + + g_maChannelPlaneRatios[channelPositionA][5] * g_maChannelPlaneRatios[channelPositionB][5]; + + return contribution; +} + +MA_API ma_channel_converter_config ma_channel_converter_config_init(ma_format format, ma_uint32 channelsIn, const ma_channel* pChannelMapIn, ma_uint32 channelsOut, const ma_channel* pChannelMapOut, ma_channel_mix_mode mixingMode) +{ + ma_channel_converter_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channelsIn = channelsIn; + config.channelsOut = channelsOut; + config.pChannelMapIn = pChannelMapIn; + config.pChannelMapOut = pChannelMapOut; + config.mixingMode = mixingMode; + + return config; +} + +static ma_int32 ma_channel_converter_float_to_fixed(float x) +{ + return (ma_int32)(x * (1< 0); + + for (iChannel = 0; iChannel < channels; ++iChannel) { + if (ma_is_spatial_channel_position(ma_channel_map_get_channel(pChannelMap, channels, iChannel))) { + spatialChannelCount++; + } + } + + return spatialChannelCount; +} + +static ma_bool32 ma_is_spatial_channel_position(ma_channel channelPosition) +{ + int i; + + if (channelPosition == MA_CHANNEL_NONE || channelPosition == MA_CHANNEL_MONO || channelPosition == MA_CHANNEL_LFE) { + return MA_FALSE; + } + + if (channelPosition >= MA_CHANNEL_AUX_0 && channelPosition <= MA_CHANNEL_AUX_31) { + return MA_FALSE; + } + + for (i = 0; i < 6; ++i) { /* Each side of a cube. */ + if (g_maChannelPlaneRatios[channelPosition][i] != 0) { + return MA_TRUE; + } + } + + return MA_FALSE; +} + + +static ma_bool32 ma_channel_map_is_passthrough(const ma_channel* pChannelMapIn, ma_uint32 channelsIn, const ma_channel* pChannelMapOut, ma_uint32 channelsOut) +{ + if (channelsOut == channelsIn) { + return ma_channel_map_is_equal(pChannelMapOut, pChannelMapIn, channelsOut); + } else { + return MA_FALSE; /* Channel counts differ, so cannot be a passthrough. */ + } +} + +static ma_channel_conversion_path ma_channel_map_get_conversion_path(const ma_channel* pChannelMapIn, ma_uint32 channelsIn, const ma_channel* pChannelMapOut, ma_uint32 channelsOut, ma_channel_mix_mode mode) +{ + if (ma_channel_map_is_passthrough(pChannelMapIn, channelsIn, pChannelMapOut, channelsOut)) { + return ma_channel_conversion_path_passthrough; + } + + if (channelsOut == 1 && (pChannelMapOut == NULL || pChannelMapOut[0] == MA_CHANNEL_MONO)) { + return ma_channel_conversion_path_mono_out; + } + + if (channelsIn == 1 && (pChannelMapIn == NULL || pChannelMapIn[0] == MA_CHANNEL_MONO)) { + return ma_channel_conversion_path_mono_in; + } + + if (mode == ma_channel_mix_mode_custom_weights) { + return ma_channel_conversion_path_weights; + } + + /* + We can use a simple shuffle if both channel maps have the same channel count and all channel + positions are present in both. + */ + if (channelsIn == channelsOut) { + ma_uint32 iChannelIn; + ma_bool32 areAllChannelPositionsPresent = MA_TRUE; + for (iChannelIn = 0; iChannelIn < channelsIn; ++iChannelIn) { + ma_bool32 isInputChannelPositionInOutput = ma_channel_map_contains_channel_position(channelsOut, pChannelMapOut, ma_channel_map_get_channel(pChannelMapIn, channelsIn, iChannelIn)); + if (!isInputChannelPositionInOutput) { + areAllChannelPositionsPresent = MA_FALSE; + break; + } + } + + if (areAllChannelPositionsPresent) { + return ma_channel_conversion_path_shuffle; + } + } + + /* Getting here means we'll need to use weights. */ + return ma_channel_conversion_path_weights; +} + + +static ma_result ma_channel_map_build_shuffle_table(const ma_channel* pChannelMapIn, ma_uint32 channelCountIn, const ma_channel* pChannelMapOut, ma_uint32 channelCountOut, ma_uint8* pShuffleTable) +{ + ma_uint32 iChannelIn; + ma_uint32 iChannelOut; + + if (pShuffleTable == NULL || channelCountIn == 0 || channelCountOut == 0) { + return MA_INVALID_ARGS; + } + + /* + When building the shuffle table we just do a 1:1 mapping based on the first occurrence of a channel. If the + input channel has more than one occurrence of a channel position, the second one will be ignored. + */ + for (iChannelOut = 0; iChannelOut < channelCountOut; iChannelOut += 1) { + ma_channel channelOut; + + /* Default to MA_CHANNEL_INDEX_NULL so that if a mapping is not found it'll be set appropriately. */ + pShuffleTable[iChannelOut] = MA_CHANNEL_INDEX_NULL; + + channelOut = ma_channel_map_get_channel(pChannelMapOut, channelCountOut, iChannelOut); + for (iChannelIn = 0; iChannelIn < channelCountIn; iChannelIn += 1) { + ma_channel channelIn; + + channelIn = ma_channel_map_get_channel(pChannelMapIn, channelCountIn, iChannelIn); + if (channelOut == channelIn) { + pShuffleTable[iChannelOut] = (ma_uint8)iChannelIn; + break; + } + + /* + Getting here means the channels don't exactly match, but we are going to support some + relaxed matching for practicality. If, for example, there are two stereo channel maps, + but one uses front left/right and the other uses side left/right, it makes logical + sense to just map these. The way we'll do it is we'll check if there is a logical + corresponding mapping, and if so, apply it, but we will *not* break from the loop, + thereby giving the loop a chance to find an exact match later which will take priority. + */ + switch (channelOut) + { + /* Left channels. */ + case MA_CHANNEL_FRONT_LEFT: + case MA_CHANNEL_SIDE_LEFT: + { + switch (channelIn) { + case MA_CHANNEL_FRONT_LEFT: + case MA_CHANNEL_SIDE_LEFT: + { + pShuffleTable[iChannelOut] = (ma_uint8)iChannelIn; + } break; + } + } break; + + /* Right channels. */ + case MA_CHANNEL_FRONT_RIGHT: + case MA_CHANNEL_SIDE_RIGHT: + { + switch (channelIn) { + case MA_CHANNEL_FRONT_RIGHT: + case MA_CHANNEL_SIDE_RIGHT: + { + pShuffleTable[iChannelOut] = (ma_uint8)iChannelIn; + } break; + } + } break; + + default: break; + } + } + } + + return MA_SUCCESS; +} + + +static void ma_channel_map_apply_shuffle_table_u8(ma_uint8* pFramesOut, ma_uint32 channelsOut, const ma_uint8* pFramesIn, ma_uint32 channelsIn, ma_uint64 frameCount, const ma_uint8* pShuffleTable) +{ + ma_uint64 iFrame; + ma_uint32 iChannelOut; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_uint8 iChannelIn = pShuffleTable[iChannelOut]; + if (iChannelIn < channelsIn) { /* For safety, and to deal with MA_CHANNEL_INDEX_NULL. */ + pFramesOut[iChannelOut] = pFramesIn[iChannelIn]; + } else { + pFramesOut[iChannelOut] = 0; + } + } + + pFramesOut += channelsOut; + pFramesIn += channelsIn; + } +} + +static void ma_channel_map_apply_shuffle_table_s16(ma_int16* pFramesOut, ma_uint32 channelsOut, const ma_int16* pFramesIn, ma_uint32 channelsIn, ma_uint64 frameCount, const ma_uint8* pShuffleTable) +{ + ma_uint64 iFrame; + ma_uint32 iChannelOut; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_uint8 iChannelIn = pShuffleTable[iChannelOut]; + if (iChannelIn < channelsIn) { /* For safety, and to deal with MA_CHANNEL_INDEX_NULL. */ + pFramesOut[iChannelOut] = pFramesIn[iChannelIn]; + } else { + pFramesOut[iChannelOut] = 0; + } + } + + pFramesOut += channelsOut; + pFramesIn += channelsIn; + } +} + +static void ma_channel_map_apply_shuffle_table_s24(ma_uint8* pFramesOut, ma_uint32 channelsOut, const ma_uint8* pFramesIn, ma_uint32 channelsIn, ma_uint64 frameCount, const ma_uint8* pShuffleTable) +{ + ma_uint64 iFrame; + ma_uint32 iChannelOut; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_uint8 iChannelIn = pShuffleTable[iChannelOut]; + if (iChannelIn < channelsIn) { /* For safety, and to deal with MA_CHANNEL_INDEX_NULL. */ + pFramesOut[iChannelOut*3 + 0] = pFramesIn[iChannelIn*3 + 0]; + pFramesOut[iChannelOut*3 + 1] = pFramesIn[iChannelIn*3 + 1]; + pFramesOut[iChannelOut*3 + 2] = pFramesIn[iChannelIn*3 + 2]; + } else { + pFramesOut[iChannelOut*3 + 0] = 0; + } pFramesOut[iChannelOut*3 + 1] = 0; + } pFramesOut[iChannelOut*3 + 2] = 0; + + pFramesOut += channelsOut*3; + pFramesIn += channelsIn*3; + } +} + +static void ma_channel_map_apply_shuffle_table_s32(ma_int32* pFramesOut, ma_uint32 channelsOut, const ma_int32* pFramesIn, ma_uint32 channelsIn, ma_uint64 frameCount, const ma_uint8* pShuffleTable) +{ + ma_uint64 iFrame; + ma_uint32 iChannelOut; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_uint8 iChannelIn = pShuffleTable[iChannelOut]; + if (iChannelIn < channelsIn) { /* For safety, and to deal with MA_CHANNEL_INDEX_NULL. */ + pFramesOut[iChannelOut] = pFramesIn[iChannelIn]; + } else { + pFramesOut[iChannelOut] = 0; + } + } + + pFramesOut += channelsOut; + pFramesIn += channelsIn; + } +} + +static void ma_channel_map_apply_shuffle_table_f32(float* pFramesOut, ma_uint32 channelsOut, const float* pFramesIn, ma_uint32 channelsIn, ma_uint64 frameCount, const ma_uint8* pShuffleTable) +{ + ma_uint64 iFrame; + ma_uint32 iChannelOut; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_uint8 iChannelIn = pShuffleTable[iChannelOut]; + if (iChannelIn < channelsIn) { /* For safety, and to deal with MA_CHANNEL_INDEX_NULL. */ + pFramesOut[iChannelOut] = pFramesIn[iChannelIn]; + } else { + pFramesOut[iChannelOut] = 0; + } + } + + pFramesOut += channelsOut; + pFramesIn += channelsIn; + } +} + +static ma_result ma_channel_map_apply_shuffle_table(void* pFramesOut, ma_uint32 channelsOut, const void* pFramesIn, ma_uint32 channelsIn, ma_uint64 frameCount, const ma_uint8* pShuffleTable, ma_format format) +{ + if (pFramesOut == NULL || pFramesIn == NULL || channelsOut == 0 || pShuffleTable == NULL) { + return MA_INVALID_ARGS; + } + + switch (format) + { + case ma_format_u8: + { + ma_channel_map_apply_shuffle_table_u8((ma_uint8*)pFramesOut, channelsOut, (const ma_uint8*)pFramesIn, channelsIn, frameCount, pShuffleTable); + } break; + + case ma_format_s16: + { + ma_channel_map_apply_shuffle_table_s16((ma_int16*)pFramesOut, channelsOut, (const ma_int16*)pFramesIn, channelsIn, frameCount, pShuffleTable); + } break; + + case ma_format_s24: + { + ma_channel_map_apply_shuffle_table_s24((ma_uint8*)pFramesOut, channelsOut, (const ma_uint8*)pFramesIn, channelsIn, frameCount, pShuffleTable); + } break; + + case ma_format_s32: + { + ma_channel_map_apply_shuffle_table_s32((ma_int32*)pFramesOut, channelsOut, (const ma_int32*)pFramesIn, channelsIn, frameCount, pShuffleTable); + } break; + + case ma_format_f32: + { + ma_channel_map_apply_shuffle_table_f32((float*)pFramesOut, channelsOut, (const float*)pFramesIn, channelsIn, frameCount, pShuffleTable); + } break; + + default: return MA_INVALID_ARGS; /* Unknown format. */ + } + + return MA_SUCCESS; +} + +static ma_result ma_channel_map_apply_mono_out_f32(float* pFramesOut, const float* pFramesIn, const ma_channel* pChannelMapIn, ma_uint32 channelsIn, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint32 iChannelIn; + ma_uint32 accumulationCount; + + if (pFramesOut == NULL || pFramesIn == NULL || channelsIn == 0) { + return MA_INVALID_ARGS; + } + + /* In this case the output stream needs to be the average of all channels, ignoring NONE. */ + + /* A quick pre-processing step to get the accumulation counter since we're ignoring NONE channels. */ + accumulationCount = 0; + for (iChannelIn = 0; iChannelIn < channelsIn; iChannelIn += 1) { + if (ma_channel_map_get_channel(pChannelMapIn, channelsIn, iChannelIn) != MA_CHANNEL_NONE) { + accumulationCount += 1; + } + } + + if (accumulationCount > 0) { /* <-- Prevent a division by zero. */ + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float accumulation = 0; + + for (iChannelIn = 0; iChannelIn < channelsIn; iChannelIn += 1) { + ma_channel channelIn = ma_channel_map_get_channel(pChannelMapIn, channelsIn, iChannelIn); + if (channelIn != MA_CHANNEL_NONE) { + accumulation += pFramesIn[iChannelIn]; + } + } + + pFramesOut[0] = accumulation / accumulationCount; + pFramesOut += 1; + pFramesIn += channelsIn; + } + } else { + ma_silence_pcm_frames(pFramesOut, frameCount, ma_format_f32, 1); + } + + return MA_SUCCESS; +} + +static ma_result ma_channel_map_apply_mono_in_f32(float* MA_RESTRICT pFramesOut, const ma_channel* pChannelMapOut, ma_uint32 channelsOut, const float* MA_RESTRICT pFramesIn, ma_uint64 frameCount, ma_mono_expansion_mode monoExpansionMode) +{ + ma_uint64 iFrame; + ma_uint32 iChannelOut; + + if (pFramesOut == NULL || channelsOut == 0 || pFramesIn == NULL) { + return MA_INVALID_ARGS; + } + + /* Note that the MA_CHANNEL_NONE channel must be ignored in all cases. */ + switch (monoExpansionMode) + { + case ma_mono_expansion_mode_average: + { + float weight; + ma_uint32 validChannelCount = 0; + + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_channel channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + if (channelOut != MA_CHANNEL_NONE) { + validChannelCount += 1; + } + } + + weight = 1.0f / validChannelCount; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_channel channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + if (channelOut != MA_CHANNEL_NONE) { + pFramesOut[iChannelOut] = pFramesIn[0] * weight; + } + } + + pFramesOut += channelsOut; + pFramesIn += 1; + } + } break; + + case ma_mono_expansion_mode_stereo_only: + { + if (channelsOut >= 2) { + ma_uint32 iChannelLeft = (ma_uint32)-1; + ma_uint32 iChannelRight = (ma_uint32)-1; + + /* + We first need to find our stereo channels. We prefer front-left and front-right, but + if they're not available, we'll also try side-left and side-right. If neither are + available we'll fall through to the default case below. + */ + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_channel channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + if (channelOut == MA_CHANNEL_SIDE_LEFT) { + iChannelLeft = iChannelOut; + } + if (channelOut == MA_CHANNEL_SIDE_RIGHT) { + iChannelRight = iChannelOut; + } + } + + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_channel channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + if (channelOut == MA_CHANNEL_FRONT_LEFT) { + iChannelLeft = iChannelOut; + } + if (channelOut == MA_CHANNEL_FRONT_RIGHT) { + iChannelRight = iChannelOut; + } + } + + + if (iChannelLeft != (ma_uint32)-1 && iChannelRight != (ma_uint32)-1) { + /* We found our stereo channels so we can duplicate the signal across those channels. */ + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_channel channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + if (channelOut != MA_CHANNEL_NONE) { + if (iChannelOut == iChannelLeft || iChannelOut == iChannelRight) { + pFramesOut[iChannelOut] = pFramesIn[0]; + } else { + pFramesOut[iChannelOut] = 0.0f; + } + } + } + + pFramesOut += channelsOut; + pFramesIn += 1; + } + + break; /* Get out of the switch. */ + } else { + /* Fallthrough. Does not have left and right channels. */ + goto default_handler; + } + } else { + /* Fallthrough. Does not have stereo channels. */ + goto default_handler; + } + }; /* Fallthrough. See comments above. */ + + case ma_mono_expansion_mode_duplicate: + default: + { + default_handler: + { + if (channelsOut <= MA_MAX_CHANNELS) { + ma_bool32 hasEmptyChannel = MA_FALSE; + ma_channel channelPositions[MA_MAX_CHANNELS]; + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + channelPositions[iChannelOut] = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + if (channelPositions[iChannelOut] == MA_CHANNEL_NONE) { + hasEmptyChannel = MA_TRUE; + } + } + + if (hasEmptyChannel == MA_FALSE) { + /* + Faster path when there's no MA_CHANNEL_NONE channel positions. This should hopefully + help the compiler with auto-vectorization.m + */ + if (channelsOut == 2) { + #if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + /* We want to do two frames in each iteration. */ + ma_uint64 unrolledFrameCount = frameCount >> 1; + + for (iFrame = 0; iFrame < unrolledFrameCount; iFrame += 1) { + __m128 in0 = _mm_set1_ps(pFramesIn[iFrame*2 + 0]); + __m128 in1 = _mm_set1_ps(pFramesIn[iFrame*2 + 1]); + _mm_storeu_ps(&pFramesOut[iFrame*4 + 0], _mm_shuffle_ps(in0, in1, _MM_SHUFFLE(0, 0, 0, 0))); + } + + /* Tail. */ + iFrame = unrolledFrameCount << 1; + goto generic_on_fastpath; + } else + #endif + { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < 2; iChannelOut += 1) { + pFramesOut[iFrame*2 + iChannelOut] = pFramesIn[iFrame]; + } + } + } + } else if (channelsOut == 6) { + #if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + /* We want to do two frames in each iteration so we can have a multiple of 4 samples. */ + ma_uint64 unrolledFrameCount = frameCount >> 1; + + for (iFrame = 0; iFrame < unrolledFrameCount; iFrame += 1) { + __m128 in0 = _mm_set1_ps(pFramesIn[iFrame*2 + 0]); + __m128 in1 = _mm_set1_ps(pFramesIn[iFrame*2 + 1]); + + _mm_storeu_ps(&pFramesOut[iFrame*12 + 0], in0); + _mm_storeu_ps(&pFramesOut[iFrame*12 + 4], _mm_shuffle_ps(in0, in1, _MM_SHUFFLE(0, 0, 0, 0))); + _mm_storeu_ps(&pFramesOut[iFrame*12 + 8], in1); + } + + /* Tail. */ + iFrame = unrolledFrameCount << 1; + goto generic_on_fastpath; + } else + #endif + { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < 6; iChannelOut += 1) { + pFramesOut[iFrame*6 + iChannelOut] = pFramesIn[iFrame]; + } + } + } + } else if (channelsOut == 8) { + #if defined(MA_SUPPORT_SSE2) + if (ma_has_sse2()) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + __m128 in = _mm_set1_ps(pFramesIn[iFrame]); + _mm_storeu_ps(&pFramesOut[iFrame*8 + 0], in); + _mm_storeu_ps(&pFramesOut[iFrame*8 + 4], in); + } + } else + #endif + { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < 8; iChannelOut += 1) { + pFramesOut[iFrame*8 + iChannelOut] = pFramesIn[iFrame]; + } + } + } + } else { + iFrame = 0; + + #if defined(MA_SUPPORT_SSE2) /* For silencing a warning with non-x86 builds. */ + generic_on_fastpath: + #endif + { + for (; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + pFramesOut[iFrame*channelsOut + iChannelOut] = pFramesIn[iFrame]; + } + } + } + } + } else { + /* Slow path. Need to handle MA_CHANNEL_NONE. */ + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + if (channelPositions[iChannelOut] != MA_CHANNEL_NONE) { + pFramesOut[iFrame*channelsOut + iChannelOut] = pFramesIn[iFrame]; + } + } + } + } + } else { + /* Slow path. Too many channels to store on the stack. */ + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_channel channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + if (channelOut != MA_CHANNEL_NONE) { + pFramesOut[iFrame*channelsOut + iChannelOut] = pFramesIn[iFrame]; + } + } + } + } + } + } break; + } + + return MA_SUCCESS; +} + +static void ma_channel_map_apply_f32(float* pFramesOut, const ma_channel* pChannelMapOut, ma_uint32 channelsOut, const float* pFramesIn, const ma_channel* pChannelMapIn, ma_uint32 channelsIn, ma_uint64 frameCount, ma_channel_mix_mode mode, ma_mono_expansion_mode monoExpansionMode) +{ + ma_channel_conversion_path conversionPath = ma_channel_map_get_conversion_path(pChannelMapIn, channelsIn, pChannelMapOut, channelsOut, mode); + + /* Optimized Path: Passthrough */ + if (conversionPath == ma_channel_conversion_path_passthrough) { + ma_copy_pcm_frames(pFramesOut, pFramesIn, frameCount, ma_format_f32, channelsOut); + return; + } + + /* Special Path: Mono Output. */ + if (conversionPath == ma_channel_conversion_path_mono_out) { + ma_channel_map_apply_mono_out_f32(pFramesOut, pFramesIn, pChannelMapIn, channelsIn, frameCount); + return; + } + + /* Special Path: Mono Input. */ + if (conversionPath == ma_channel_conversion_path_mono_in) { + ma_channel_map_apply_mono_in_f32(pFramesOut, pChannelMapOut, channelsOut, pFramesIn, frameCount, monoExpansionMode); + return; + } + + /* Getting here means we aren't running on an optimized conversion path. */ + if (channelsOut <= MA_MAX_CHANNELS) { + ma_result result; + + if (mode == ma_channel_mix_mode_simple) { + ma_channel shuffleTable[MA_MAX_CHANNELS]; + + result = ma_channel_map_build_shuffle_table(pChannelMapIn, channelsIn, pChannelMapOut, channelsOut, shuffleTable); + if (result != MA_SUCCESS) { + return; + } + + result = ma_channel_map_apply_shuffle_table(pFramesOut, channelsOut, pFramesIn, channelsIn, frameCount, shuffleTable, ma_format_f32); + if (result != MA_SUCCESS) { + return; + } + } else { + ma_uint32 iFrame; + ma_uint32 iChannelOut; + ma_uint32 iChannelIn; + float weights[32][32]; /* Do not use MA_MAX_CHANNELS here! */ + + /* + If we have a small enough number of channels, pre-compute the weights. Otherwise we'll just need to + fall back to a slower path because otherwise we'll run out of stack space. + */ + if (channelsIn <= ma_countof(weights) && channelsOut <= ma_countof(weights)) { + /* Pre-compute weights. */ + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + ma_channel channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + for (iChannelIn = 0; iChannelIn < channelsIn; iChannelIn += 1) { + ma_channel channelIn = ma_channel_map_get_channel(pChannelMapIn, channelsIn, iChannelIn); + weights[iChannelOut][iChannelIn] = ma_calculate_channel_position_rectangular_weight(channelOut, channelIn); + } + } + + iFrame = 0; + + /* Experiment: Try an optimized unroll for some specific cases to see how it improves performance. RESULT: Good gains. */ + if (channelsOut == 8) { + /* Experiment 2: Expand the inner loop to see what kind of different it makes. RESULT: Small, but worthwhile gain. */ + if (channelsIn == 2) { + for (; iFrame < frameCount; iFrame += 1) { + float accumulation[8] = { 0, 0, 0, 0, 0, 0, 0, 0 }; + + accumulation[0] += pFramesIn[iFrame*2 + 0] * weights[0][0]; + accumulation[1] += pFramesIn[iFrame*2 + 0] * weights[1][0]; + accumulation[2] += pFramesIn[iFrame*2 + 0] * weights[2][0]; + accumulation[3] += pFramesIn[iFrame*2 + 0] * weights[3][0]; + accumulation[4] += pFramesIn[iFrame*2 + 0] * weights[4][0]; + accumulation[5] += pFramesIn[iFrame*2 + 0] * weights[5][0]; + accumulation[6] += pFramesIn[iFrame*2 + 0] * weights[6][0]; + accumulation[7] += pFramesIn[iFrame*2 + 0] * weights[7][0]; + + accumulation[0] += pFramesIn[iFrame*2 + 1] * weights[0][1]; + accumulation[1] += pFramesIn[iFrame*2 + 1] * weights[1][1]; + accumulation[2] += pFramesIn[iFrame*2 + 1] * weights[2][1]; + accumulation[3] += pFramesIn[iFrame*2 + 1] * weights[3][1]; + accumulation[4] += pFramesIn[iFrame*2 + 1] * weights[4][1]; + accumulation[5] += pFramesIn[iFrame*2 + 1] * weights[5][1]; + accumulation[6] += pFramesIn[iFrame*2 + 1] * weights[6][1]; + accumulation[7] += pFramesIn[iFrame*2 + 1] * weights[7][1]; + + pFramesOut[iFrame*8 + 0] = accumulation[0]; + pFramesOut[iFrame*8 + 1] = accumulation[1]; + pFramesOut[iFrame*8 + 2] = accumulation[2]; + pFramesOut[iFrame*8 + 3] = accumulation[3]; + pFramesOut[iFrame*8 + 4] = accumulation[4]; + pFramesOut[iFrame*8 + 5] = accumulation[5]; + pFramesOut[iFrame*8 + 6] = accumulation[6]; + pFramesOut[iFrame*8 + 7] = accumulation[7]; + } + } else { + /* When outputting to 8 channels, we can do everything in groups of two 4x SIMD operations. */ + for (; iFrame < frameCount; iFrame += 1) { + float accumulation[8] = { 0, 0, 0, 0, 0, 0, 0, 0 }; + + for (iChannelIn = 0; iChannelIn < channelsIn; iChannelIn += 1) { + accumulation[0] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[0][iChannelIn]; + accumulation[1] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[1][iChannelIn]; + accumulation[2] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[2][iChannelIn]; + accumulation[3] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[3][iChannelIn]; + accumulation[4] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[4][iChannelIn]; + accumulation[5] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[5][iChannelIn]; + accumulation[6] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[6][iChannelIn]; + accumulation[7] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[7][iChannelIn]; + } + + pFramesOut[iFrame*8 + 0] = accumulation[0]; + pFramesOut[iFrame*8 + 1] = accumulation[1]; + pFramesOut[iFrame*8 + 2] = accumulation[2]; + pFramesOut[iFrame*8 + 3] = accumulation[3]; + pFramesOut[iFrame*8 + 4] = accumulation[4]; + pFramesOut[iFrame*8 + 5] = accumulation[5]; + pFramesOut[iFrame*8 + 6] = accumulation[6]; + pFramesOut[iFrame*8 + 7] = accumulation[7]; + } + } + } else if (channelsOut == 6) { + /* + When outputting to 6 channels we unfortunately don't have a nice multiple of 4 to do 4x SIMD operations. Instead we'll + expand our weights and do two frames at a time. + */ + for (; iFrame < frameCount; iFrame += 1) { + float accumulation[12] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; + + for (iChannelIn = 0; iChannelIn < channelsIn; iChannelIn += 1) { + accumulation[0] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[0][iChannelIn]; + accumulation[1] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[1][iChannelIn]; + accumulation[2] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[2][iChannelIn]; + accumulation[3] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[3][iChannelIn]; + accumulation[4] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[4][iChannelIn]; + accumulation[5] += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[5][iChannelIn]; + } + + pFramesOut[iFrame*6 + 0] = accumulation[0]; + pFramesOut[iFrame*6 + 1] = accumulation[1]; + pFramesOut[iFrame*6 + 2] = accumulation[2]; + pFramesOut[iFrame*6 + 3] = accumulation[3]; + pFramesOut[iFrame*6 + 4] = accumulation[4]; + pFramesOut[iFrame*6 + 5] = accumulation[5]; + } + } + + /* Leftover frames. */ + for (; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + float accumulation = 0; + + for (iChannelIn = 0; iChannelIn < channelsIn; iChannelIn += 1) { + accumulation += pFramesIn[iFrame*channelsIn + iChannelIn] * weights[iChannelOut][iChannelIn]; + } + + pFramesOut[iFrame*channelsOut + iChannelOut] = accumulation; + } + } + } else { + /* Cannot pre-compute weights because not enough room in stack-allocated buffer. */ + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelOut = 0; iChannelOut < channelsOut; iChannelOut += 1) { + float accumulation = 0; + ma_channel channelOut = ma_channel_map_get_channel(pChannelMapOut, channelsOut, iChannelOut); + + for (iChannelIn = 0; iChannelIn < channelsIn; iChannelIn += 1) { + ma_channel channelIn = ma_channel_map_get_channel(pChannelMapIn, channelsIn, iChannelIn); + accumulation += pFramesIn[iFrame*channelsIn + iChannelIn] * ma_calculate_channel_position_rectangular_weight(channelOut, channelIn); + } + + pFramesOut[iFrame*channelsOut + iChannelOut] = accumulation; + } + } + } + } + } else { + /* Fall back to silence. If you hit this, what are you doing with so many channels?! */ + ma_silence_pcm_frames(pFramesOut, frameCount, ma_format_f32, channelsOut); + } +} + + +typedef struct +{ + size_t sizeInBytes; + size_t channelMapInOffset; + size_t channelMapOutOffset; + size_t shuffleTableOffset; + size_t weightsOffset; +} ma_channel_converter_heap_layout; + +static ma_channel_conversion_path ma_channel_converter_config_get_conversion_path(const ma_channel_converter_config* pConfig) +{ + return ma_channel_map_get_conversion_path(pConfig->pChannelMapIn, pConfig->channelsIn, pConfig->pChannelMapOut, pConfig->channelsOut, pConfig->mixingMode); +} + +static ma_result ma_channel_converter_get_heap_layout(const ma_channel_converter_config* pConfig, ma_channel_converter_heap_layout* pHeapLayout) +{ + ma_channel_conversion_path conversionPath; + + MA_ASSERT(pHeapLayout != NULL); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channelsIn == 0 || pConfig->channelsOut == 0) { + return MA_INVALID_ARGS; + } + + if (!ma_channel_map_is_valid(pConfig->pChannelMapIn, pConfig->channelsIn)) { + return MA_INVALID_ARGS; + } + + if (!ma_channel_map_is_valid(pConfig->pChannelMapOut, pConfig->channelsOut)) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* Input channel map. Only need to allocate this if we have an input channel map (otherwise default channel map is assumed). */ + pHeapLayout->channelMapInOffset = pHeapLayout->sizeInBytes; + if (pConfig->pChannelMapIn != NULL) { + pHeapLayout->sizeInBytes += sizeof(ma_channel) * pConfig->channelsIn; + } + + /* Output channel map. Only need to allocate this if we have an output channel map (otherwise default channel map is assumed). */ + pHeapLayout->channelMapOutOffset = pHeapLayout->sizeInBytes; + if (pConfig->pChannelMapOut != NULL) { + pHeapLayout->sizeInBytes += sizeof(ma_channel) * pConfig->channelsOut; + } + + /* Alignment for the next section. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + /* Whether or not we use weights of a shuffle table depends on the channel map themselves and the algorithm we've chosen. */ + conversionPath = ma_channel_converter_config_get_conversion_path(pConfig); + + /* Shuffle table */ + pHeapLayout->shuffleTableOffset = pHeapLayout->sizeInBytes; + if (conversionPath == ma_channel_conversion_path_shuffle) { + pHeapLayout->sizeInBytes += sizeof(ma_uint8) * pConfig->channelsOut; + } + + /* Weights */ + pHeapLayout->weightsOffset = pHeapLayout->sizeInBytes; + if (conversionPath == ma_channel_conversion_path_weights) { + pHeapLayout->sizeInBytes += sizeof(float*) * pConfig->channelsIn; + pHeapLayout->sizeInBytes += sizeof(float ) * pConfig->channelsIn * pConfig->channelsOut; + } + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +MA_API ma_result ma_channel_converter_get_heap_size(const ma_channel_converter_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_channel_converter_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_channel_converter_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_channel_converter_init_preallocated(const ma_channel_converter_config* pConfig, void* pHeap, ma_channel_converter* pConverter) +{ + ma_result result; + ma_channel_converter_heap_layout heapLayout; + + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pConverter); + + result = ma_channel_converter_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pConverter->_pHeap = pHeap; + MA_ZERO_MEMORY(pConverter->_pHeap, heapLayout.sizeInBytes); + + pConverter->format = pConfig->format; + pConverter->channelsIn = pConfig->channelsIn; + pConverter->channelsOut = pConfig->channelsOut; + pConverter->mixingMode = pConfig->mixingMode; + + if (pConfig->pChannelMapIn != NULL) { + pConverter->pChannelMapIn = (ma_channel*)ma_offset_ptr(pHeap, heapLayout.channelMapInOffset); + ma_channel_map_copy_or_default(pConverter->pChannelMapIn, pConfig->channelsIn, pConfig->pChannelMapIn, pConfig->channelsIn); + } else { + pConverter->pChannelMapIn = NULL; /* Use default channel map. */ + } + + if (pConfig->pChannelMapOut != NULL) { + pConverter->pChannelMapOut = (ma_channel*)ma_offset_ptr(pHeap, heapLayout.channelMapOutOffset); + ma_channel_map_copy_or_default(pConverter->pChannelMapOut, pConfig->channelsOut, pConfig->pChannelMapOut, pConfig->channelsOut); + } else { + pConverter->pChannelMapOut = NULL; /* Use default channel map. */ + } + + pConverter->conversionPath = ma_channel_converter_config_get_conversion_path(pConfig); + + if (pConverter->conversionPath == ma_channel_conversion_path_shuffle) { + pConverter->pShuffleTable = (ma_uint8*)ma_offset_ptr(pHeap, heapLayout.shuffleTableOffset); + ma_channel_map_build_shuffle_table(pConverter->pChannelMapIn, pConverter->channelsIn, pConverter->pChannelMapOut, pConverter->channelsOut, pConverter->pShuffleTable); + } + + if (pConverter->conversionPath == ma_channel_conversion_path_weights) { + ma_uint32 iChannelIn; + ma_uint32 iChannelOut; + + if (pConverter->format == ma_format_f32) { + pConverter->weights.f32 = (float** )ma_offset_ptr(pHeap, heapLayout.weightsOffset); + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; iChannelIn += 1) { + pConverter->weights.f32[iChannelIn] = (float*)ma_offset_ptr(pHeap, heapLayout.weightsOffset + ((sizeof(float*) * pConverter->channelsIn) + (sizeof(float) * pConverter->channelsOut * iChannelIn))); + } + } else { + pConverter->weights.s16 = (ma_int32**)ma_offset_ptr(pHeap, heapLayout.weightsOffset); + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; iChannelIn += 1) { + pConverter->weights.s16[iChannelIn] = (ma_int32*)ma_offset_ptr(pHeap, heapLayout.weightsOffset + ((sizeof(ma_int32*) * pConverter->channelsIn) + (sizeof(ma_int32) * pConverter->channelsOut * iChannelIn))); + } + } + + /* Silence our weights by default. */ + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; iChannelIn += 1) { + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; iChannelOut += 1) { + if (pConverter->format == ma_format_f32) { + pConverter->weights.f32[iChannelIn][iChannelOut] = 0.0f; + } else { + pConverter->weights.s16[iChannelIn][iChannelOut] = 0; + } + } + } + + /* + We now need to fill out our weights table. This is determined by the mixing mode. + */ + + /* In all cases we need to make sure all channels that are present in both channel maps have a 1:1 mapping. */ + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + ma_channel channelPosIn = ma_channel_map_get_channel(pConverter->pChannelMapIn, pConverter->channelsIn, iChannelIn); + + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; ++iChannelOut) { + ma_channel channelPosOut = ma_channel_map_get_channel(pConverter->pChannelMapOut, pConverter->channelsOut, iChannelOut); + + if (channelPosIn == channelPosOut) { + float weight = 1; + + if (pConverter->format == ma_format_f32) { + pConverter->weights.f32[iChannelIn][iChannelOut] = weight; + } else { + pConverter->weights.s16[iChannelIn][iChannelOut] = ma_channel_converter_float_to_fixed(weight); + } + } + } + } + + switch (pConverter->mixingMode) + { + case ma_channel_mix_mode_custom_weights: + { + if (pConfig->ppWeights == NULL) { + return MA_INVALID_ARGS; /* Config specified a custom weights mixing mode, but no custom weights have been specified. */ + } + + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; iChannelIn += 1) { + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; iChannelOut += 1) { + float weight = pConfig->ppWeights[iChannelIn][iChannelOut]; + + if (pConverter->format == ma_format_f32) { + pConverter->weights.f32[iChannelIn][iChannelOut] = weight; + } else { + pConverter->weights.s16[iChannelIn][iChannelOut] = ma_channel_converter_float_to_fixed(weight); + } + } + } + } break; + + case ma_channel_mix_mode_simple: + { + /* + In simple mode, only set weights for channels that have exactly matching types, leave the rest at + zero. The 1:1 mappings have already been covered before this switch statement. + */ + } break; + + case ma_channel_mix_mode_rectangular: + default: + { + /* Unmapped input channels. */ + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + ma_channel channelPosIn = ma_channel_map_get_channel(pConverter->pChannelMapIn, pConverter->channelsIn, iChannelIn); + + if (ma_is_spatial_channel_position(channelPosIn)) { + if (!ma_channel_map_contains_channel_position(pConverter->channelsOut, pConverter->pChannelMapOut, channelPosIn)) { + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; ++iChannelOut) { + ma_channel channelPosOut = ma_channel_map_get_channel(pConverter->pChannelMapOut, pConverter->channelsOut, iChannelOut); + + if (ma_is_spatial_channel_position(channelPosOut)) { + float weight = 0; + if (pConverter->mixingMode == ma_channel_mix_mode_rectangular) { + weight = ma_calculate_channel_position_rectangular_weight(channelPosIn, channelPosOut); + } + + /* Only apply the weight if we haven't already got some contribution from the respective channels. */ + if (pConverter->format == ma_format_f32) { + if (pConverter->weights.f32[iChannelIn][iChannelOut] == 0) { + pConverter->weights.f32[iChannelIn][iChannelOut] = weight; + } + } else { + if (pConverter->weights.s16[iChannelIn][iChannelOut] == 0) { + pConverter->weights.s16[iChannelIn][iChannelOut] = ma_channel_converter_float_to_fixed(weight); + } + } + } + } + } + } + } + + /* Unmapped output channels. */ + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; ++iChannelOut) { + ma_channel channelPosOut = ma_channel_map_get_channel(pConverter->pChannelMapOut, pConverter->channelsOut, iChannelOut); + + if (ma_is_spatial_channel_position(channelPosOut)) { + if (!ma_channel_map_contains_channel_position(pConverter->channelsIn, pConverter->pChannelMapIn, channelPosOut)) { + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + ma_channel channelPosIn = ma_channel_map_get_channel(pConverter->pChannelMapIn, pConverter->channelsIn, iChannelIn); + + if (ma_is_spatial_channel_position(channelPosIn)) { + float weight = 0; + if (pConverter->mixingMode == ma_channel_mix_mode_rectangular) { + weight = ma_calculate_channel_position_rectangular_weight(channelPosIn, channelPosOut); + } + + /* Only apply the weight if we haven't already got some contribution from the respective channels. */ + if (pConverter->format == ma_format_f32) { + if (pConverter->weights.f32[iChannelIn][iChannelOut] == 0) { + pConverter->weights.f32[iChannelIn][iChannelOut] = weight; + } + } else { + if (pConverter->weights.s16[iChannelIn][iChannelOut] == 0) { + pConverter->weights.s16[iChannelIn][iChannelOut] = ma_channel_converter_float_to_fixed(weight); + } + } + } + } + } + } + } + + /* If LFE is in the output channel map but was not present in the input channel map, configure its weight now */ + if (pConfig->calculateLFEFromSpatialChannels) { + if (!ma_channel_map_contains_channel_position(pConverter->channelsIn, pConverter->pChannelMapIn, MA_CHANNEL_LFE)) { + ma_uint32 spatialChannelCount = ma_channel_map_get_spatial_channel_count(pConverter->pChannelMapIn, pConverter->channelsIn); + ma_uint32 iChannelOutLFE; + + if (spatialChannelCount > 0 && ma_channel_map_find_channel_position(pConverter->channelsOut, pConverter->pChannelMapOut, MA_CHANNEL_LFE, &iChannelOutLFE)) { + const float weightForLFE = 1.0f / spatialChannelCount; + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + const ma_channel channelPosIn = ma_channel_map_get_channel(pConverter->pChannelMapIn, pConverter->channelsIn, iChannelIn); + if (ma_is_spatial_channel_position(channelPosIn)) { + if (pConverter->format == ma_format_f32) { + if (pConverter->weights.f32[iChannelIn][iChannelOutLFE] == 0) { + pConverter->weights.f32[iChannelIn][iChannelOutLFE] = weightForLFE; + } + } else { + if (pConverter->weights.s16[iChannelIn][iChannelOutLFE] == 0) { + pConverter->weights.s16[iChannelIn][iChannelOutLFE] = ma_channel_converter_float_to_fixed(weightForLFE); + } + } + } + } + } + } + } + } break; + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_channel_converter_init(const ma_channel_converter_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_channel_converter* pConverter) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_channel_converter_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_channel_converter_init_preallocated(pConfig, pHeap, pConverter); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pConverter->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_channel_converter_uninit(ma_channel_converter* pConverter, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pConverter == NULL) { + return; + } + + if (pConverter->_ownsHeap) { + ma_free(pConverter->_pHeap, pAllocationCallbacks); + } +} + +static ma_result ma_channel_converter_process_pcm_frames__passthrough(ma_channel_converter* pConverter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + MA_ASSERT(pConverter != NULL); + MA_ASSERT(pFramesOut != NULL); + MA_ASSERT(pFramesIn != NULL); + + ma_copy_memory_64(pFramesOut, pFramesIn, frameCount * ma_get_bytes_per_frame(pConverter->format, pConverter->channelsOut)); + return MA_SUCCESS; +} + +static ma_result ma_channel_converter_process_pcm_frames__shuffle(ma_channel_converter* pConverter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + MA_ASSERT(pConverter != NULL); + MA_ASSERT(pFramesOut != NULL); + MA_ASSERT(pFramesIn != NULL); + MA_ASSERT(pConverter->channelsIn == pConverter->channelsOut); + + return ma_channel_map_apply_shuffle_table(pFramesOut, pConverter->channelsOut, pFramesIn, pConverter->channelsIn, frameCount, pConverter->pShuffleTable, pConverter->format); +} + +static ma_result ma_channel_converter_process_pcm_frames__mono_in(ma_channel_converter* pConverter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + + MA_ASSERT(pConverter != NULL); + MA_ASSERT(pFramesOut != NULL); + MA_ASSERT(pFramesIn != NULL); + MA_ASSERT(pConverter->channelsIn == 1); + + switch (pConverter->format) + { + case ma_format_u8: + { + /* */ ma_uint8* pFramesOutU8 = ( ma_uint8*)pFramesOut; + const ma_uint8* pFramesInU8 = (const ma_uint8*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < pConverter->channelsOut; iChannel += 1) { + pFramesOutU8[iFrame*pConverter->channelsOut + iChannel] = pFramesInU8[iFrame]; + } + } + } break; + + case ma_format_s16: + { + /* */ ma_int16* pFramesOutS16 = ( ma_int16*)pFramesOut; + const ma_int16* pFramesInS16 = (const ma_int16*)pFramesIn; + + if (pConverter->channelsOut == 2) { + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + pFramesOutS16[iFrame*2 + 0] = pFramesInS16[iFrame]; + pFramesOutS16[iFrame*2 + 1] = pFramesInS16[iFrame]; + } + } else { + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < pConverter->channelsOut; iChannel += 1) { + pFramesOutS16[iFrame*pConverter->channelsOut + iChannel] = pFramesInS16[iFrame]; + } + } + } + } break; + + case ma_format_s24: + { + /* */ ma_uint8* pFramesOutS24 = ( ma_uint8*)pFramesOut; + const ma_uint8* pFramesInS24 = (const ma_uint8*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < pConverter->channelsOut; iChannel += 1) { + ma_uint64 iSampleOut = iFrame*pConverter->channelsOut + iChannel; + ma_uint64 iSampleIn = iFrame; + pFramesOutS24[iSampleOut*3 + 0] = pFramesInS24[iSampleIn*3 + 0]; + pFramesOutS24[iSampleOut*3 + 1] = pFramesInS24[iSampleIn*3 + 1]; + pFramesOutS24[iSampleOut*3 + 2] = pFramesInS24[iSampleIn*3 + 2]; + } + } + } break; + + case ma_format_s32: + { + /* */ ma_int32* pFramesOutS32 = ( ma_int32*)pFramesOut; + const ma_int32* pFramesInS32 = (const ma_int32*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < pConverter->channelsOut; iChannel += 1) { + pFramesOutS32[iFrame*pConverter->channelsOut + iChannel] = pFramesInS32[iFrame]; + } + } + } break; + + case ma_format_f32: + { + /* */ float* pFramesOutF32 = ( float*)pFramesOut; + const float* pFramesInF32 = (const float*)pFramesIn; + + if (pConverter->channelsOut == 2) { + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + pFramesOutF32[iFrame*2 + 0] = pFramesInF32[iFrame]; + pFramesOutF32[iFrame*2 + 1] = pFramesInF32[iFrame]; + } + } else { + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < pConverter->channelsOut; iChannel += 1) { + pFramesOutF32[iFrame*pConverter->channelsOut + iChannel] = pFramesInF32[iFrame]; + } + } + } + } break; + + default: return MA_INVALID_OPERATION; /* Unknown format. */ + } + + return MA_SUCCESS; +} + +static ma_result ma_channel_converter_process_pcm_frames__mono_out(ma_channel_converter* pConverter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint32 iChannel; + + MA_ASSERT(pConverter != NULL); + MA_ASSERT(pFramesOut != NULL); + MA_ASSERT(pFramesIn != NULL); + MA_ASSERT(pConverter->channelsOut == 1); + + switch (pConverter->format) + { + case ma_format_u8: + { + /* */ ma_uint8* pFramesOutU8 = ( ma_uint8*)pFramesOut; + const ma_uint8* pFramesInU8 = (const ma_uint8*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_int32 t = 0; + for (iChannel = 0; iChannel < pConverter->channelsIn; iChannel += 1) { + t += ma_pcm_sample_u8_to_s16_no_scale(pFramesInU8[iFrame*pConverter->channelsIn + iChannel]); + } + + pFramesOutU8[iFrame] = ma_clip_u8(t / pConverter->channelsOut); + } + } break; + + case ma_format_s16: + { + /* */ ma_int16* pFramesOutS16 = ( ma_int16*)pFramesOut; + const ma_int16* pFramesInS16 = (const ma_int16*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_int32 t = 0; + for (iChannel = 0; iChannel < pConverter->channelsIn; iChannel += 1) { + t += pFramesInS16[iFrame*pConverter->channelsIn + iChannel]; + } + + pFramesOutS16[iFrame] = (ma_int16)(t / pConverter->channelsIn); + } + } break; + + case ma_format_s24: + { + /* */ ma_uint8* pFramesOutS24 = ( ma_uint8*)pFramesOut; + const ma_uint8* pFramesInS24 = (const ma_uint8*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_int64 t = 0; + for (iChannel = 0; iChannel < pConverter->channelsIn; iChannel += 1) { + t += ma_pcm_sample_s24_to_s32_no_scale(&pFramesInS24[(iFrame*pConverter->channelsIn + iChannel)*3]); + } + + ma_pcm_sample_s32_to_s24_no_scale(t / pConverter->channelsIn, &pFramesOutS24[iFrame*3]); + } + } break; + + case ma_format_s32: + { + /* */ ma_int32* pFramesOutS32 = ( ma_int32*)pFramesOut; + const ma_int32* pFramesInS32 = (const ma_int32*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + ma_int64 t = 0; + for (iChannel = 0; iChannel < pConverter->channelsIn; iChannel += 1) { + t += pFramesInS32[iFrame*pConverter->channelsIn + iChannel]; + } + + pFramesOutS32[iFrame] = (ma_int32)(t / pConverter->channelsIn); + } + } break; + + case ma_format_f32: + { + /* */ float* pFramesOutF32 = ( float*)pFramesOut; + const float* pFramesInF32 = (const float*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; ++iFrame) { + float t = 0; + for (iChannel = 0; iChannel < pConverter->channelsIn; iChannel += 1) { + t += pFramesInF32[iFrame*pConverter->channelsIn + iChannel]; + } + + pFramesOutF32[iFrame] = t / pConverter->channelsIn; + } + } break; + + default: return MA_INVALID_OPERATION; /* Unknown format. */ + } + + return MA_SUCCESS; +} + +static ma_result ma_channel_converter_process_pcm_frames__weights(ma_channel_converter* pConverter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + ma_uint32 iFrame; + ma_uint32 iChannelIn; + ma_uint32 iChannelOut; + + MA_ASSERT(pConverter != NULL); + MA_ASSERT(pFramesOut != NULL); + MA_ASSERT(pFramesIn != NULL); + + /* This is the more complicated case. Each of the output channels is accumulated with 0 or more input channels. */ + + /* Clear. */ + ma_zero_memory_64(pFramesOut, frameCount * ma_get_bytes_per_frame(pConverter->format, pConverter->channelsOut)); + + /* Accumulate. */ + switch (pConverter->format) + { + case ma_format_u8: + { + /* */ ma_uint8* pFramesOutU8 = ( ma_uint8*)pFramesOut; + const ma_uint8* pFramesInU8 = (const ma_uint8*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; ++iChannelOut) { + ma_int16 u8_O = ma_pcm_sample_u8_to_s16_no_scale(pFramesOutU8[iFrame*pConverter->channelsOut + iChannelOut]); + ma_int16 u8_I = ma_pcm_sample_u8_to_s16_no_scale(pFramesInU8 [iFrame*pConverter->channelsIn + iChannelIn ]); + ma_int32 s = (ma_int32)ma_clamp(u8_O + ((u8_I * pConverter->weights.s16[iChannelIn][iChannelOut]) >> MA_CHANNEL_CONVERTER_FIXED_POINT_SHIFT), -128, 127); + pFramesOutU8[iFrame*pConverter->channelsOut + iChannelOut] = ma_clip_u8((ma_int16)s); + } + } + } + } break; + + case ma_format_s16: + { + /* */ ma_int16* pFramesOutS16 = ( ma_int16*)pFramesOut; + const ma_int16* pFramesInS16 = (const ma_int16*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; ++iChannelOut) { + ma_int32 s = pFramesOutS16[iFrame*pConverter->channelsOut + iChannelOut]; + s += (pFramesInS16[iFrame*pConverter->channelsIn + iChannelIn] * pConverter->weights.s16[iChannelIn][iChannelOut]) >> MA_CHANNEL_CONVERTER_FIXED_POINT_SHIFT; + + pFramesOutS16[iFrame*pConverter->channelsOut + iChannelOut] = (ma_int16)ma_clamp(s, -32768, 32767); + } + } + } + } break; + + case ma_format_s24: + { + /* */ ma_uint8* pFramesOutS24 = ( ma_uint8*)pFramesOut; + const ma_uint8* pFramesInS24 = (const ma_uint8*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; ++iChannelOut) { + ma_int64 s24_O = ma_pcm_sample_s24_to_s32_no_scale(&pFramesOutS24[(iFrame*pConverter->channelsOut + iChannelOut)*3]); + ma_int64 s24_I = ma_pcm_sample_s24_to_s32_no_scale(&pFramesInS24 [(iFrame*pConverter->channelsIn + iChannelIn )*3]); + ma_int64 s24 = (ma_int32)ma_clamp(s24_O + ((s24_I * pConverter->weights.s16[iChannelIn][iChannelOut]) >> MA_CHANNEL_CONVERTER_FIXED_POINT_SHIFT), -8388608, 8388607); + ma_pcm_sample_s32_to_s24_no_scale(s24, &pFramesOutS24[(iFrame*pConverter->channelsOut + iChannelOut)*3]); + } + } + } + } break; + + case ma_format_s32: + { + /* */ ma_int32* pFramesOutS32 = ( ma_int32*)pFramesOut; + const ma_int32* pFramesInS32 = (const ma_int32*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; ++iChannelOut) { + ma_int64 s = pFramesOutS32[iFrame*pConverter->channelsOut + iChannelOut]; + s += ((ma_int64)pFramesInS32[iFrame*pConverter->channelsIn + iChannelIn] * pConverter->weights.s16[iChannelIn][iChannelOut]) >> MA_CHANNEL_CONVERTER_FIXED_POINT_SHIFT; + + pFramesOutS32[iFrame*pConverter->channelsOut + iChannelOut] = ma_clip_s32(s); + } + } + } + } break; + + case ma_format_f32: + { + /* */ float* pFramesOutF32 = ( float*)pFramesOut; + const float* pFramesInF32 = (const float*)pFramesIn; + + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannelIn = 0; iChannelIn < pConverter->channelsIn; ++iChannelIn) { + for (iChannelOut = 0; iChannelOut < pConverter->channelsOut; ++iChannelOut) { + pFramesOutF32[iFrame*pConverter->channelsOut + iChannelOut] += pFramesInF32[iFrame*pConverter->channelsIn + iChannelIn] * pConverter->weights.f32[iChannelIn][iChannelOut]; + } + } + } + } break; + + default: return MA_INVALID_OPERATION; /* Unknown format. */ + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_channel_converter_process_pcm_frames(ma_channel_converter* pConverter, void* pFramesOut, const void* pFramesIn, ma_uint64 frameCount) +{ + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + if (pFramesOut == NULL) { + return MA_INVALID_ARGS; + } + + if (pFramesIn == NULL) { + ma_zero_memory_64(pFramesOut, frameCount * ma_get_bytes_per_frame(pConverter->format, pConverter->channelsOut)); + return MA_SUCCESS; + } + + switch (pConverter->conversionPath) + { + case ma_channel_conversion_path_passthrough: return ma_channel_converter_process_pcm_frames__passthrough(pConverter, pFramesOut, pFramesIn, frameCount); + case ma_channel_conversion_path_mono_out: return ma_channel_converter_process_pcm_frames__mono_out(pConverter, pFramesOut, pFramesIn, frameCount); + case ma_channel_conversion_path_mono_in: return ma_channel_converter_process_pcm_frames__mono_in(pConverter, pFramesOut, pFramesIn, frameCount); + case ma_channel_conversion_path_shuffle: return ma_channel_converter_process_pcm_frames__shuffle(pConverter, pFramesOut, pFramesIn, frameCount); + case ma_channel_conversion_path_weights: + default: + { + return ma_channel_converter_process_pcm_frames__weights(pConverter, pFramesOut, pFramesIn, frameCount); + } + } +} + +MA_API ma_result ma_channel_converter_get_input_channel_map(const ma_channel_converter* pConverter, ma_channel* pChannelMap, size_t channelMapCap) +{ + if (pConverter == NULL || pChannelMap == NULL) { + return MA_INVALID_ARGS; + } + + ma_channel_map_copy_or_default(pChannelMap, channelMapCap, pConverter->pChannelMapIn, pConverter->channelsIn); + + return MA_SUCCESS; +} + +MA_API ma_result ma_channel_converter_get_output_channel_map(const ma_channel_converter* pConverter, ma_channel* pChannelMap, size_t channelMapCap) +{ + if (pConverter == NULL || pChannelMap == NULL) { + return MA_INVALID_ARGS; + } + + ma_channel_map_copy_or_default(pChannelMap, channelMapCap, pConverter->pChannelMapOut, pConverter->channelsOut); + + return MA_SUCCESS; +} + + +/************************************************************************************************************************************************************** + +Data Conversion + +**************************************************************************************************************************************************************/ +MA_API ma_data_converter_config ma_data_converter_config_init_default(void) +{ + ma_data_converter_config config; + MA_ZERO_OBJECT(&config); + + config.ditherMode = ma_dither_mode_none; + config.resampling.algorithm = ma_resample_algorithm_linear; + config.allowDynamicSampleRate = MA_FALSE; /* Disable dynamic sample rates by default because dynamic rate adjustments should be quite rare and it allows an optimization for cases when the in and out sample rates are the same. */ + + /* Linear resampling defaults. */ + config.resampling.linear.lpfOrder = 1; + + return config; +} + +MA_API ma_data_converter_config ma_data_converter_config_init(ma_format formatIn, ma_format formatOut, ma_uint32 channelsIn, ma_uint32 channelsOut, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut) +{ + ma_data_converter_config config = ma_data_converter_config_init_default(); + config.formatIn = formatIn; + config.formatOut = formatOut; + config.channelsIn = channelsIn; + config.channelsOut = channelsOut; + config.sampleRateIn = sampleRateIn; + config.sampleRateOut = sampleRateOut; + + return config; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t channelConverterOffset; + size_t resamplerOffset; +} ma_data_converter_heap_layout; + +static ma_bool32 ma_data_converter_config_is_resampler_required(const ma_data_converter_config* pConfig) +{ + MA_ASSERT(pConfig != NULL); + + return pConfig->allowDynamicSampleRate || pConfig->sampleRateIn != pConfig->sampleRateOut; +} + +static ma_format ma_data_converter_config_get_mid_format(const ma_data_converter_config* pConfig) +{ + MA_ASSERT(pConfig != NULL); + + /* + We want to avoid as much data conversion as possible. The channel converter and linear + resampler both support s16 and f32 natively. We need to decide on the format to use for this + stage. We call this the mid format because it's used in the middle stage of the conversion + pipeline. If the output format is either s16 or f32 we use that one. If that is not the case it + will do the same thing for the input format. If it's neither we just use f32. If we are using a + custom resampling backend, we can only guarantee that f32 will be supported so we'll be forced + to use that if resampling is required. + */ + if (ma_data_converter_config_is_resampler_required(pConfig) && pConfig->resampling.algorithm != ma_resample_algorithm_linear) { + return ma_format_f32; /* <-- Force f32 since that is the only one we can guarantee will be supported by the resampler. */ + } else { + /* */ if (pConfig->formatOut == ma_format_s16 || pConfig->formatOut == ma_format_f32) { + return pConfig->formatOut; + } else if (pConfig->formatIn == ma_format_s16 || pConfig->formatIn == ma_format_f32) { + return pConfig->formatIn; + } else { + return ma_format_f32; + } + } +} + +static ma_channel_converter_config ma_channel_converter_config_init_from_data_converter_config(const ma_data_converter_config* pConfig) +{ + ma_channel_converter_config channelConverterConfig; + + MA_ASSERT(pConfig != NULL); + + channelConverterConfig = ma_channel_converter_config_init(ma_data_converter_config_get_mid_format(pConfig), pConfig->channelsIn, pConfig->pChannelMapIn, pConfig->channelsOut, pConfig->pChannelMapOut, pConfig->channelMixMode); + channelConverterConfig.ppWeights = pConfig->ppChannelWeights; + channelConverterConfig.calculateLFEFromSpatialChannels = pConfig->calculateLFEFromSpatialChannels; + + return channelConverterConfig; +} + +static ma_resampler_config ma_resampler_config_init_from_data_converter_config(const ma_data_converter_config* pConfig) +{ + ma_resampler_config resamplerConfig; + ma_uint32 resamplerChannels; + + MA_ASSERT(pConfig != NULL); + + /* The resampler is the most expensive part of the conversion process, so we need to do it at the stage where the channel count is at it's lowest. */ + if (pConfig->channelsIn < pConfig->channelsOut) { + resamplerChannels = pConfig->channelsIn; + } else { + resamplerChannels = pConfig->channelsOut; + } + + resamplerConfig = ma_resampler_config_init(ma_data_converter_config_get_mid_format(pConfig), resamplerChannels, pConfig->sampleRateIn, pConfig->sampleRateOut, pConfig->resampling.algorithm); + resamplerConfig.linear = pConfig->resampling.linear; + resamplerConfig.pBackendVTable = pConfig->resampling.pBackendVTable; + resamplerConfig.pBackendUserData = pConfig->resampling.pBackendUserData; + + return resamplerConfig; +} + +static ma_result ma_data_converter_get_heap_layout(const ma_data_converter_config* pConfig, ma_data_converter_heap_layout* pHeapLayout) +{ + ma_result result; + + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channelsIn == 0 || pConfig->channelsOut == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* Channel converter. */ + pHeapLayout->channelConverterOffset = pHeapLayout->sizeInBytes; + { + size_t heapSizeInBytes; + ma_channel_converter_config channelConverterConfig = ma_channel_converter_config_init_from_data_converter_config(pConfig); + + result = ma_channel_converter_get_heap_size(&channelConverterConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes += heapSizeInBytes; + } + + /* Resampler. */ + pHeapLayout->resamplerOffset = pHeapLayout->sizeInBytes; + if (ma_data_converter_config_is_resampler_required(pConfig)) { + size_t heapSizeInBytes; + ma_resampler_config resamplerConfig = ma_resampler_config_init_from_data_converter_config(pConfig); + + result = ma_resampler_get_heap_size(&resamplerConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes += heapSizeInBytes; + } + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_converter_get_heap_size(const ma_data_converter_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_data_converter_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_data_converter_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_converter_init_preallocated(const ma_data_converter_config* pConfig, void* pHeap, ma_data_converter* pConverter) +{ + ma_result result; + ma_data_converter_heap_layout heapLayout; + ma_format midFormat; + ma_bool32 isResamplingRequired; + + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pConverter); + + result = ma_data_converter_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pConverter->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pConverter->formatIn = pConfig->formatIn; + pConverter->formatOut = pConfig->formatOut; + pConverter->channelsIn = pConfig->channelsIn; + pConverter->channelsOut = pConfig->channelsOut; + pConverter->sampleRateIn = pConfig->sampleRateIn; + pConverter->sampleRateOut = pConfig->sampleRateOut; + pConverter->ditherMode = pConfig->ditherMode; + + /* + Determine if resampling is required. We need to do this so we can determine an appropriate + mid format to use. If resampling is required, the mid format must be ma_format_f32 since + that is the only one that is guaranteed to supported by custom resampling backends. + */ + isResamplingRequired = ma_data_converter_config_is_resampler_required(pConfig); + midFormat = ma_data_converter_config_get_mid_format(pConfig); + + + /* Channel converter. We always initialize this, but we check if it configures itself as a passthrough to determine whether or not it's needed. */ + { + ma_channel_converter_config channelConverterConfig = ma_channel_converter_config_init_from_data_converter_config(pConfig); + + result = ma_channel_converter_init_preallocated(&channelConverterConfig, ma_offset_ptr(pHeap, heapLayout.channelConverterOffset), &pConverter->channelConverter); + if (result != MA_SUCCESS) { + return result; + } + + /* If the channel converter is not a passthrough we need to enable it. Otherwise we can skip it. */ + if (pConverter->channelConverter.conversionPath != ma_channel_conversion_path_passthrough) { + pConverter->hasChannelConverter = MA_TRUE; + } + } + + + /* Resampler. */ + if (isResamplingRequired) { + ma_resampler_config resamplerConfig = ma_resampler_config_init_from_data_converter_config(pConfig); + + result = ma_resampler_init_preallocated(&resamplerConfig, ma_offset_ptr(pHeap, heapLayout.resamplerOffset), &pConverter->resampler); + if (result != MA_SUCCESS) { + return result; + } + + pConverter->hasResampler = MA_TRUE; + } + + + /* We can simplify pre- and post-format conversion if we have neither channel conversion nor resampling. */ + if (pConverter->hasChannelConverter == MA_FALSE && pConverter->hasResampler == MA_FALSE) { + /* We have neither channel conversion nor resampling so we'll only need one of pre- or post-format conversion, or none if the input and output formats are the same. */ + if (pConverter->formatIn == pConverter->formatOut) { + /* The formats are the same so we can just pass through. */ + pConverter->hasPreFormatConversion = MA_FALSE; + pConverter->hasPostFormatConversion = MA_FALSE; + } else { + /* The formats are different so we need to do either pre- or post-format conversion. It doesn't matter which. */ + pConverter->hasPreFormatConversion = MA_FALSE; + pConverter->hasPostFormatConversion = MA_TRUE; + } + } else { + /* We have a channel converter and/or resampler so we'll need channel conversion based on the mid format. */ + if (pConverter->formatIn != midFormat) { + pConverter->hasPreFormatConversion = MA_TRUE; + } + if (pConverter->formatOut != midFormat) { + pConverter->hasPostFormatConversion = MA_TRUE; + } + } + + /* We can enable passthrough optimizations if applicable. Note that we'll only be able to do this if the sample rate is static. */ + if (pConverter->hasPreFormatConversion == MA_FALSE && + pConverter->hasPostFormatConversion == MA_FALSE && + pConverter->hasChannelConverter == MA_FALSE && + pConverter->hasResampler == MA_FALSE) { + pConverter->isPassthrough = MA_TRUE; + } + + + /* We now need to determine our execution path. */ + if (pConverter->isPassthrough) { + pConverter->executionPath = ma_data_converter_execution_path_passthrough; + } else { + if (pConverter->channelsIn < pConverter->channelsOut) { + /* Do resampling first, if necessary. */ + MA_ASSERT(pConverter->hasChannelConverter == MA_TRUE); + + if (pConverter->hasResampler) { + pConverter->executionPath = ma_data_converter_execution_path_resample_first; + } else { + pConverter->executionPath = ma_data_converter_execution_path_channels_only; + } + } else { + /* Do channel conversion first, if necessary. */ + if (pConverter->hasChannelConverter) { + if (pConverter->hasResampler) { + pConverter->executionPath = ma_data_converter_execution_path_channels_first; + } else { + pConverter->executionPath = ma_data_converter_execution_path_channels_only; + } + } else { + /* Channel routing not required. */ + if (pConverter->hasResampler) { + pConverter->executionPath = ma_data_converter_execution_path_resample_only; + } else { + pConverter->executionPath = ma_data_converter_execution_path_format_only; + } + } + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_converter_init(const ma_data_converter_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_converter* pConverter) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_data_converter_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_data_converter_init_preallocated(pConfig, pHeap, pConverter); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pConverter->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_data_converter_uninit(ma_data_converter* pConverter, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pConverter == NULL) { + return; + } + + if (pConverter->hasResampler) { + ma_resampler_uninit(&pConverter->resampler, pAllocationCallbacks); + } + + ma_channel_converter_uninit(&pConverter->channelConverter, pAllocationCallbacks); + + if (pConverter->_ownsHeap) { + ma_free(pConverter->_pHeap, pAllocationCallbacks); + } +} + +static ma_result ma_data_converter_process_pcm_frames__passthrough(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 frameCount; + + MA_ASSERT(pConverter != NULL); + + frameCountIn = 0; + if (pFrameCountIn != NULL) { + frameCountIn = *pFrameCountIn; + } + + frameCountOut = 0; + if (pFrameCountOut != NULL) { + frameCountOut = *pFrameCountOut; + } + + frameCount = ma_min(frameCountIn, frameCountOut); + + if (pFramesOut != NULL) { + if (pFramesIn != NULL) { + ma_copy_memory_64(pFramesOut, pFramesIn, frameCount * ma_get_bytes_per_frame(pConverter->formatOut, pConverter->channelsOut)); + } else { + ma_zero_memory_64(pFramesOut, frameCount * ma_get_bytes_per_frame(pConverter->formatOut, pConverter->channelsOut)); + } + } + + if (pFrameCountIn != NULL) { + *pFrameCountIn = frameCount; + } + if (pFrameCountOut != NULL) { + *pFrameCountOut = frameCount; + } + + return MA_SUCCESS; +} + +static ma_result ma_data_converter_process_pcm_frames__format_only(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 frameCount; + + MA_ASSERT(pConverter != NULL); + + frameCountIn = 0; + if (pFrameCountIn != NULL) { + frameCountIn = *pFrameCountIn; + } + + frameCountOut = 0; + if (pFrameCountOut != NULL) { + frameCountOut = *pFrameCountOut; + } + + frameCount = ma_min(frameCountIn, frameCountOut); + + if (pFramesOut != NULL) { + if (pFramesIn != NULL) { + ma_convert_pcm_frames_format(pFramesOut, pConverter->formatOut, pFramesIn, pConverter->formatIn, frameCount, pConverter->channelsIn, pConverter->ditherMode); + } else { + ma_zero_memory_64(pFramesOut, frameCount * ma_get_bytes_per_frame(pConverter->formatOut, pConverter->channelsOut)); + } + } + + if (pFrameCountIn != NULL) { + *pFrameCountIn = frameCount; + } + if (pFrameCountOut != NULL) { + *pFrameCountOut = frameCount; + } + + return MA_SUCCESS; +} + + +static ma_result ma_data_converter_process_pcm_frames__resample_with_format_conversion(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + ma_result result = MA_SUCCESS; + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 framesProcessedIn; + ma_uint64 framesProcessedOut; + + MA_ASSERT(pConverter != NULL); + + frameCountIn = 0; + if (pFrameCountIn != NULL) { + frameCountIn = *pFrameCountIn; + } + + frameCountOut = 0; + if (pFrameCountOut != NULL) { + frameCountOut = *pFrameCountOut; + } + + framesProcessedIn = 0; + framesProcessedOut = 0; + + while (framesProcessedOut < frameCountOut) { + ma_uint8 pTempBufferOut[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + const ma_uint32 tempBufferOutCap = sizeof(pTempBufferOut) / ma_get_bytes_per_frame(pConverter->resampler.format, pConverter->resampler.channels); + const void* pFramesInThisIteration; + /* */ void* pFramesOutThisIteration; + ma_uint64 frameCountInThisIteration; + ma_uint64 frameCountOutThisIteration; + + if (pFramesIn != NULL) { + pFramesInThisIteration = ma_offset_ptr(pFramesIn, framesProcessedIn * ma_get_bytes_per_frame(pConverter->formatIn, pConverter->channelsIn)); + } else { + pFramesInThisIteration = NULL; + } + + if (pFramesOut != NULL) { + pFramesOutThisIteration = ma_offset_ptr(pFramesOut, framesProcessedOut * ma_get_bytes_per_frame(pConverter->formatOut, pConverter->channelsOut)); + } else { + pFramesOutThisIteration = NULL; + } + + /* Do a pre format conversion if necessary. */ + if (pConverter->hasPreFormatConversion) { + ma_uint8 pTempBufferIn[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + const ma_uint32 tempBufferInCap = sizeof(pTempBufferIn) / ma_get_bytes_per_frame(pConverter->resampler.format, pConverter->resampler.channels); + + frameCountInThisIteration = (frameCountIn - framesProcessedIn); + if (frameCountInThisIteration > tempBufferInCap) { + frameCountInThisIteration = tempBufferInCap; + } + + if (pConverter->hasPostFormatConversion) { + if (frameCountInThisIteration > tempBufferOutCap) { + frameCountInThisIteration = tempBufferOutCap; + } + } + + if (pFramesInThisIteration != NULL) { + ma_convert_pcm_frames_format(pTempBufferIn, pConverter->resampler.format, pFramesInThisIteration, pConverter->formatIn, frameCountInThisIteration, pConverter->channelsIn, pConverter->ditherMode); + } else { + MA_ZERO_MEMORY(pTempBufferIn, sizeof(pTempBufferIn)); + } + + frameCountOutThisIteration = (frameCountOut - framesProcessedOut); + + if (pConverter->hasPostFormatConversion) { + /* Both input and output conversion required. Output to the temp buffer. */ + if (frameCountOutThisIteration > tempBufferOutCap) { + frameCountOutThisIteration = tempBufferOutCap; + } + + result = ma_resampler_process_pcm_frames(&pConverter->resampler, pTempBufferIn, &frameCountInThisIteration, pTempBufferOut, &frameCountOutThisIteration); + } else { + /* Only pre-format required. Output straight to the output buffer. */ + result = ma_resampler_process_pcm_frames(&pConverter->resampler, pTempBufferIn, &frameCountInThisIteration, pFramesOutThisIteration, &frameCountOutThisIteration); + } + + if (result != MA_SUCCESS) { + break; + } + } else { + /* No pre-format required. Just read straight from the input buffer. */ + MA_ASSERT(pConverter->hasPostFormatConversion == MA_TRUE); + + frameCountInThisIteration = (frameCountIn - framesProcessedIn); + frameCountOutThisIteration = (frameCountOut - framesProcessedOut); + if (frameCountOutThisIteration > tempBufferOutCap) { + frameCountOutThisIteration = tempBufferOutCap; + } + + result = ma_resampler_process_pcm_frames(&pConverter->resampler, pFramesInThisIteration, &frameCountInThisIteration, pTempBufferOut, &frameCountOutThisIteration); + if (result != MA_SUCCESS) { + break; + } + } + + /* If we are doing a post format conversion we need to do that now. */ + if (pConverter->hasPostFormatConversion) { + if (pFramesOutThisIteration != NULL) { + ma_convert_pcm_frames_format(pFramesOutThisIteration, pConverter->formatOut, pTempBufferOut, pConverter->resampler.format, frameCountOutThisIteration, pConverter->resampler.channels, pConverter->ditherMode); + } + } + + framesProcessedIn += frameCountInThisIteration; + framesProcessedOut += frameCountOutThisIteration; + + MA_ASSERT(framesProcessedIn <= frameCountIn); + MA_ASSERT(framesProcessedOut <= frameCountOut); + + if (frameCountOutThisIteration == 0) { + break; /* Consumed all of our input data. */ + } + } + + if (pFrameCountIn != NULL) { + *pFrameCountIn = framesProcessedIn; + } + if (pFrameCountOut != NULL) { + *pFrameCountOut = framesProcessedOut; + } + + return result; +} + +static ma_result ma_data_converter_process_pcm_frames__resample_only(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + MA_ASSERT(pConverter != NULL); + + if (pConverter->hasPreFormatConversion == MA_FALSE && pConverter->hasPostFormatConversion == MA_FALSE) { + /* Neither pre- nor post-format required. This is simple case where only resampling is required. */ + return ma_resampler_process_pcm_frames(&pConverter->resampler, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + } else { + /* Format conversion required. */ + return ma_data_converter_process_pcm_frames__resample_with_format_conversion(pConverter, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + } +} + +static ma_result ma_data_converter_process_pcm_frames__channels_only(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + ma_result result; + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 frameCount; + + MA_ASSERT(pConverter != NULL); + + frameCountIn = 0; + if (pFrameCountIn != NULL) { + frameCountIn = *pFrameCountIn; + } + + frameCountOut = 0; + if (pFrameCountOut != NULL) { + frameCountOut = *pFrameCountOut; + } + + frameCount = ma_min(frameCountIn, frameCountOut); + + if (pConverter->hasPreFormatConversion == MA_FALSE && pConverter->hasPostFormatConversion == MA_FALSE) { + /* No format conversion required. */ + result = ma_channel_converter_process_pcm_frames(&pConverter->channelConverter, pFramesOut, pFramesIn, frameCount); + if (result != MA_SUCCESS) { + return result; + } + } else { + /* Format conversion required. */ + ma_uint64 framesProcessed = 0; + + while (framesProcessed < frameCount) { + ma_uint8 pTempBufferOut[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + const ma_uint32 tempBufferOutCap = sizeof(pTempBufferOut) / ma_get_bytes_per_frame(pConverter->channelConverter.format, pConverter->channelConverter.channelsOut); + const void* pFramesInThisIteration; + /* */ void* pFramesOutThisIteration; + ma_uint64 frameCountThisIteration; + + if (pFramesIn != NULL) { + pFramesInThisIteration = ma_offset_ptr(pFramesIn, framesProcessed * ma_get_bytes_per_frame(pConverter->formatIn, pConverter->channelsIn)); + } else { + pFramesInThisIteration = NULL; + } + + if (pFramesOut != NULL) { + pFramesOutThisIteration = ma_offset_ptr(pFramesOut, framesProcessed * ma_get_bytes_per_frame(pConverter->formatOut, pConverter->channelsOut)); + } else { + pFramesOutThisIteration = NULL; + } + + /* Do a pre format conversion if necessary. */ + if (pConverter->hasPreFormatConversion) { + ma_uint8 pTempBufferIn[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + const ma_uint32 tempBufferInCap = sizeof(pTempBufferIn) / ma_get_bytes_per_frame(pConverter->channelConverter.format, pConverter->channelConverter.channelsIn); + + frameCountThisIteration = (frameCount - framesProcessed); + if (frameCountThisIteration > tempBufferInCap) { + frameCountThisIteration = tempBufferInCap; + } + + if (pConverter->hasPostFormatConversion) { + if (frameCountThisIteration > tempBufferOutCap) { + frameCountThisIteration = tempBufferOutCap; + } + } + + if (pFramesInThisIteration != NULL) { + ma_convert_pcm_frames_format(pTempBufferIn, pConverter->channelConverter.format, pFramesInThisIteration, pConverter->formatIn, frameCountThisIteration, pConverter->channelsIn, pConverter->ditherMode); + } else { + MA_ZERO_MEMORY(pTempBufferIn, sizeof(pTempBufferIn)); + } + + if (pConverter->hasPostFormatConversion) { + /* Both input and output conversion required. Output to the temp buffer. */ + result = ma_channel_converter_process_pcm_frames(&pConverter->channelConverter, pTempBufferOut, pTempBufferIn, frameCountThisIteration); + } else { + /* Only pre-format required. Output straight to the output buffer. */ + result = ma_channel_converter_process_pcm_frames(&pConverter->channelConverter, pFramesOutThisIteration, pTempBufferIn, frameCountThisIteration); + } + + if (result != MA_SUCCESS) { + break; + } + } else { + /* No pre-format required. Just read straight from the input buffer. */ + MA_ASSERT(pConverter->hasPostFormatConversion == MA_TRUE); + + frameCountThisIteration = (frameCount - framesProcessed); + if (frameCountThisIteration > tempBufferOutCap) { + frameCountThisIteration = tempBufferOutCap; + } + + result = ma_channel_converter_process_pcm_frames(&pConverter->channelConverter, pTempBufferOut, pFramesInThisIteration, frameCountThisIteration); + if (result != MA_SUCCESS) { + break; + } + } + + /* If we are doing a post format conversion we need to do that now. */ + if (pConverter->hasPostFormatConversion) { + if (pFramesOutThisIteration != NULL) { + ma_convert_pcm_frames_format(pFramesOutThisIteration, pConverter->formatOut, pTempBufferOut, pConverter->channelConverter.format, frameCountThisIteration, pConverter->channelConverter.channelsOut, pConverter->ditherMode); + } + } + + framesProcessed += frameCountThisIteration; + } + } + + if (pFrameCountIn != NULL) { + *pFrameCountIn = frameCount; + } + if (pFrameCountOut != NULL) { + *pFrameCountOut = frameCount; + } + + return MA_SUCCESS; +} + +static ma_result ma_data_converter_process_pcm_frames__resample_first(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + ma_result result; + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 framesProcessedIn; + ma_uint64 framesProcessedOut; + ma_uint8 pTempBufferIn[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* In resampler format. */ + ma_uint64 tempBufferInCap; + ma_uint8 pTempBufferMid[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* In resampler format, channel converter input format. */ + ma_uint64 tempBufferMidCap; + ma_uint8 pTempBufferOut[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* In channel converter output format. */ + ma_uint64 tempBufferOutCap; + + MA_ASSERT(pConverter != NULL); + MA_ASSERT(pConverter->resampler.format == pConverter->channelConverter.format); + MA_ASSERT(pConverter->resampler.channels == pConverter->channelConverter.channelsIn); + MA_ASSERT(pConverter->resampler.channels < pConverter->channelConverter.channelsOut); + + frameCountIn = 0; + if (pFrameCountIn != NULL) { + frameCountIn = *pFrameCountIn; + } + + frameCountOut = 0; + if (pFrameCountOut != NULL) { + frameCountOut = *pFrameCountOut; + } + + framesProcessedIn = 0; + framesProcessedOut = 0; + + tempBufferInCap = sizeof(pTempBufferIn) / ma_get_bytes_per_frame(pConverter->resampler.format, pConverter->resampler.channels); + tempBufferMidCap = sizeof(pTempBufferIn) / ma_get_bytes_per_frame(pConverter->resampler.format, pConverter->resampler.channels); + tempBufferOutCap = sizeof(pTempBufferOut) / ma_get_bytes_per_frame(pConverter->channelConverter.format, pConverter->channelConverter.channelsOut); + + while (framesProcessedOut < frameCountOut) { + ma_uint64 frameCountInThisIteration; + ma_uint64 frameCountOutThisIteration; + const void* pRunningFramesIn = NULL; + void* pRunningFramesOut = NULL; + const void* pResampleBufferIn; + void* pChannelsBufferOut; + + if (pFramesIn != NULL) { + pRunningFramesIn = ma_offset_ptr(pFramesIn, framesProcessedIn * ma_get_bytes_per_frame(pConverter->formatIn, pConverter->channelsIn)); + } + if (pFramesOut != NULL) { + pRunningFramesOut = ma_offset_ptr(pFramesOut, framesProcessedOut * ma_get_bytes_per_frame(pConverter->formatOut, pConverter->channelsOut)); + } + + /* Run input data through the resampler and output it to the temporary buffer. */ + frameCountInThisIteration = (frameCountIn - framesProcessedIn); + + if (pConverter->hasPreFormatConversion) { + if (frameCountInThisIteration > tempBufferInCap) { + frameCountInThisIteration = tempBufferInCap; + } + } + + frameCountOutThisIteration = (frameCountOut - framesProcessedOut); + if (frameCountOutThisIteration > tempBufferMidCap) { + frameCountOutThisIteration = tempBufferMidCap; + } + + /* We can't read more frames than can fit in the output buffer. */ + if (pConverter->hasPostFormatConversion) { + if (frameCountOutThisIteration > tempBufferOutCap) { + frameCountOutThisIteration = tempBufferOutCap; + } + } + + /* We need to ensure we don't try to process too many input frames that we run out of room in the output buffer. If this happens we'll end up glitching. */ + + /* + We need to try to predict how many input frames will be required for the resampler. If the + resampler can tell us, we'll use that. Otherwise we'll need to make a best guess. The further + off we are from this, the more wasted format conversions we'll end up doing. + */ + #if 1 + { + ma_uint64 requiredInputFrameCount; + + result = ma_resampler_get_required_input_frame_count(&pConverter->resampler, frameCountOutThisIteration, &requiredInputFrameCount); + if (result != MA_SUCCESS) { + /* Fall back to a best guess. */ + requiredInputFrameCount = (frameCountOutThisIteration * pConverter->resampler.sampleRateIn) / pConverter->resampler.sampleRateOut; + } + + if (frameCountInThisIteration > requiredInputFrameCount) { + frameCountInThisIteration = requiredInputFrameCount; + } + } + #endif + + if (pConverter->hasPreFormatConversion) { + if (pFramesIn != NULL) { + ma_convert_pcm_frames_format(pTempBufferIn, pConverter->resampler.format, pRunningFramesIn, pConverter->formatIn, frameCountInThisIteration, pConverter->channelsIn, pConverter->ditherMode); + pResampleBufferIn = pTempBufferIn; + } else { + pResampleBufferIn = NULL; + } + } else { + pResampleBufferIn = pRunningFramesIn; + } + + result = ma_resampler_process_pcm_frames(&pConverter->resampler, pResampleBufferIn, &frameCountInThisIteration, pTempBufferMid, &frameCountOutThisIteration); + if (result != MA_SUCCESS) { + return result; + } + + + /* + The input data has been resampled so now we need to run it through the channel converter. The input data is always contained in pTempBufferMid. We only need to do + this part if we have an output buffer. + */ + if (pFramesOut != NULL) { + if (pConverter->hasPostFormatConversion) { + pChannelsBufferOut = pTempBufferOut; + } else { + pChannelsBufferOut = pRunningFramesOut; + } + + result = ma_channel_converter_process_pcm_frames(&pConverter->channelConverter, pChannelsBufferOut, pTempBufferMid, frameCountOutThisIteration); + if (result != MA_SUCCESS) { + return result; + } + + /* Finally we do post format conversion. */ + if (pConverter->hasPostFormatConversion) { + ma_convert_pcm_frames_format(pRunningFramesOut, pConverter->formatOut, pChannelsBufferOut, pConverter->channelConverter.format, frameCountOutThisIteration, pConverter->channelConverter.channelsOut, pConverter->ditherMode); + } + } + + + framesProcessedIn += frameCountInThisIteration; + framesProcessedOut += frameCountOutThisIteration; + + MA_ASSERT(framesProcessedIn <= frameCountIn); + MA_ASSERT(framesProcessedOut <= frameCountOut); + + if (frameCountOutThisIteration == 0) { + break; /* Consumed all of our input data. */ + } + } + + if (pFrameCountIn != NULL) { + *pFrameCountIn = framesProcessedIn; + } + if (pFrameCountOut != NULL) { + *pFrameCountOut = framesProcessedOut; + } + + return MA_SUCCESS; +} + +static ma_result ma_data_converter_process_pcm_frames__channels_first(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + ma_result result; + ma_uint64 frameCountIn; + ma_uint64 frameCountOut; + ma_uint64 framesProcessedIn; + ma_uint64 framesProcessedOut; + ma_uint8 pTempBufferIn[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* In resampler format. */ + ma_uint64 tempBufferInCap; + ma_uint8 pTempBufferMid[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* In resampler format, channel converter input format. */ + ma_uint64 tempBufferMidCap; + ma_uint8 pTempBufferOut[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* In channel converter output format. */ + ma_uint64 tempBufferOutCap; + + MA_ASSERT(pConverter != NULL); + MA_ASSERT(pConverter->resampler.format == pConverter->channelConverter.format); + MA_ASSERT(pConverter->resampler.channels == pConverter->channelConverter.channelsOut); + MA_ASSERT(pConverter->resampler.channels <= pConverter->channelConverter.channelsIn); + + frameCountIn = 0; + if (pFrameCountIn != NULL) { + frameCountIn = *pFrameCountIn; + } + + frameCountOut = 0; + if (pFrameCountOut != NULL) { + frameCountOut = *pFrameCountOut; + } + + framesProcessedIn = 0; + framesProcessedOut = 0; + + tempBufferInCap = sizeof(pTempBufferIn) / ma_get_bytes_per_frame(pConverter->channelConverter.format, pConverter->channelConverter.channelsIn); + tempBufferMidCap = sizeof(pTempBufferIn) / ma_get_bytes_per_frame(pConverter->channelConverter.format, pConverter->channelConverter.channelsOut); + tempBufferOutCap = sizeof(pTempBufferOut) / ma_get_bytes_per_frame(pConverter->resampler.format, pConverter->resampler.channels); + + while (framesProcessedOut < frameCountOut) { + ma_uint64 frameCountInThisIteration; + ma_uint64 frameCountOutThisIteration; + const void* pRunningFramesIn = NULL; + void* pRunningFramesOut = NULL; + const void* pChannelsBufferIn; + void* pResampleBufferOut; + + if (pFramesIn != NULL) { + pRunningFramesIn = ma_offset_ptr(pFramesIn, framesProcessedIn * ma_get_bytes_per_frame(pConverter->formatIn, pConverter->channelsIn)); + } + if (pFramesOut != NULL) { + pRunningFramesOut = ma_offset_ptr(pFramesOut, framesProcessedOut * ma_get_bytes_per_frame(pConverter->formatOut, pConverter->channelsOut)); + } + + /* + Before doing any processing we need to determine how many frames we should try processing + this iteration, for both input and output. The resampler requires us to perform format and + channel conversion before passing any data into it. If we get our input count wrong, we'll + end up performing redundant pre-processing. This isn't the end of the world, but it does + result in some inefficiencies proportionate to how far our estimates are off. + + If the resampler has a means to calculate exactly how much we'll need, we'll use that. + Otherwise we'll make a best guess. In order to do this, we'll need to calculate the output + frame count first. + */ + frameCountOutThisIteration = (frameCountOut - framesProcessedOut); + if (frameCountOutThisIteration > tempBufferMidCap) { + frameCountOutThisIteration = tempBufferMidCap; + } + + if (pConverter->hasPostFormatConversion) { + if (frameCountOutThisIteration > tempBufferOutCap) { + frameCountOutThisIteration = tempBufferOutCap; + } + } + + /* Now that we have the output frame count we can determine the input frame count. */ + frameCountInThisIteration = (frameCountIn - framesProcessedIn); + if (pConverter->hasPreFormatConversion) { + if (frameCountInThisIteration > tempBufferInCap) { + frameCountInThisIteration = tempBufferInCap; + } + } + + if (frameCountInThisIteration > tempBufferMidCap) { + frameCountInThisIteration = tempBufferMidCap; + } + + #if 1 + { + ma_uint64 requiredInputFrameCount; + + result = ma_resampler_get_required_input_frame_count(&pConverter->resampler, frameCountOutThisIteration, &requiredInputFrameCount); + if (result != MA_SUCCESS) { + /* Fall back to a best guess. */ + requiredInputFrameCount = (frameCountOutThisIteration * pConverter->resampler.sampleRateIn) / pConverter->resampler.sampleRateOut; + } + + if (frameCountInThisIteration > requiredInputFrameCount) { + frameCountInThisIteration = requiredInputFrameCount; + } + } + #endif + + + /* Pre format conversion. */ + if (pConverter->hasPreFormatConversion) { + if (pRunningFramesIn != NULL) { + ma_convert_pcm_frames_format(pTempBufferIn, pConverter->channelConverter.format, pRunningFramesIn, pConverter->formatIn, frameCountInThisIteration, pConverter->channelsIn, pConverter->ditherMode); + pChannelsBufferIn = pTempBufferIn; + } else { + pChannelsBufferIn = NULL; + } + } else { + pChannelsBufferIn = pRunningFramesIn; + } + + + /* Channel conversion. */ + result = ma_channel_converter_process_pcm_frames(&pConverter->channelConverter, pTempBufferMid, pChannelsBufferIn, frameCountInThisIteration); + if (result != MA_SUCCESS) { + return result; + } + + + /* Resampling. */ + if (pConverter->hasPostFormatConversion) { + pResampleBufferOut = pTempBufferOut; + } else { + pResampleBufferOut = pRunningFramesOut; + } + + result = ma_resampler_process_pcm_frames(&pConverter->resampler, pTempBufferMid, &frameCountInThisIteration, pResampleBufferOut, &frameCountOutThisIteration); + if (result != MA_SUCCESS) { + return result; + } + + + /* Post format conversion. */ + if (pConverter->hasPostFormatConversion) { + if (pRunningFramesOut != NULL) { + ma_convert_pcm_frames_format(pRunningFramesOut, pConverter->formatOut, pResampleBufferOut, pConverter->resampler.format, frameCountOutThisIteration, pConverter->channelsOut, pConverter->ditherMode); + } + } + + + framesProcessedIn += frameCountInThisIteration; + framesProcessedOut += frameCountOutThisIteration; + + MA_ASSERT(framesProcessedIn <= frameCountIn); + MA_ASSERT(framesProcessedOut <= frameCountOut); + + if (frameCountOutThisIteration == 0) { + break; /* Consumed all of our input data. */ + } + } + + if (pFrameCountIn != NULL) { + *pFrameCountIn = framesProcessedIn; + } + if (pFrameCountOut != NULL) { + *pFrameCountOut = framesProcessedOut; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_converter_process_pcm_frames(ma_data_converter* pConverter, const void* pFramesIn, ma_uint64* pFrameCountIn, void* pFramesOut, ma_uint64* pFrameCountOut) +{ + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + switch (pConverter->executionPath) + { + case ma_data_converter_execution_path_passthrough: return ma_data_converter_process_pcm_frames__passthrough(pConverter, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + case ma_data_converter_execution_path_format_only: return ma_data_converter_process_pcm_frames__format_only(pConverter, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + case ma_data_converter_execution_path_channels_only: return ma_data_converter_process_pcm_frames__channels_only(pConverter, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + case ma_data_converter_execution_path_resample_only: return ma_data_converter_process_pcm_frames__resample_only(pConverter, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + case ma_data_converter_execution_path_resample_first: return ma_data_converter_process_pcm_frames__resample_first(pConverter, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + case ma_data_converter_execution_path_channels_first: return ma_data_converter_process_pcm_frames__channels_first(pConverter, pFramesIn, pFrameCountIn, pFramesOut, pFrameCountOut); + default: return MA_INVALID_OPERATION; /* Should never hit this. */ + } +} + +MA_API ma_result ma_data_converter_set_rate(ma_data_converter* pConverter, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut) +{ + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + if (pConverter->hasResampler == MA_FALSE) { + return MA_INVALID_OPERATION; /* Dynamic resampling not enabled. */ + } + + return ma_resampler_set_rate(&pConverter->resampler, sampleRateIn, sampleRateOut); +} + +MA_API ma_result ma_data_converter_set_rate_ratio(ma_data_converter* pConverter, float ratioInOut) +{ + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + if (pConverter->hasResampler == MA_FALSE) { + return MA_INVALID_OPERATION; /* Dynamic resampling not enabled. */ + } + + return ma_resampler_set_rate_ratio(&pConverter->resampler, ratioInOut); +} + +MA_API ma_uint64 ma_data_converter_get_input_latency(const ma_data_converter* pConverter) +{ + if (pConverter == NULL) { + return 0; + } + + if (pConverter->hasResampler) { + return ma_resampler_get_input_latency(&pConverter->resampler); + } + + return 0; /* No latency without a resampler. */ +} + +MA_API ma_uint64 ma_data_converter_get_output_latency(const ma_data_converter* pConverter) +{ + if (pConverter == NULL) { + return 0; + } + + if (pConverter->hasResampler) { + return ma_resampler_get_output_latency(&pConverter->resampler); + } + + return 0; /* No latency without a resampler. */ +} + +MA_API ma_result ma_data_converter_get_required_input_frame_count(const ma_data_converter* pConverter, ma_uint64 outputFrameCount, ma_uint64* pInputFrameCount) +{ + if (pInputFrameCount == NULL) { + return MA_INVALID_ARGS; + } + + *pInputFrameCount = 0; + + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + if (pConverter->hasResampler) { + return ma_resampler_get_required_input_frame_count(&pConverter->resampler, outputFrameCount, pInputFrameCount); + } else { + *pInputFrameCount = outputFrameCount; /* 1:1 */ + return MA_SUCCESS; + } +} + +MA_API ma_result ma_data_converter_get_expected_output_frame_count(const ma_data_converter* pConverter, ma_uint64 inputFrameCount, ma_uint64* pOutputFrameCount) +{ + if (pOutputFrameCount == NULL) { + return MA_INVALID_ARGS; + } + + *pOutputFrameCount = 0; + + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + if (pConverter->hasResampler) { + return ma_resampler_get_expected_output_frame_count(&pConverter->resampler, inputFrameCount, pOutputFrameCount); + } else { + *pOutputFrameCount = inputFrameCount; /* 1:1 */ + return MA_SUCCESS; + } +} + +MA_API ma_result ma_data_converter_get_input_channel_map(const ma_data_converter* pConverter, ma_channel* pChannelMap, size_t channelMapCap) +{ + if (pConverter == NULL || pChannelMap == NULL) { + return MA_INVALID_ARGS; + } + + if (pConverter->hasChannelConverter) { + ma_channel_converter_get_output_channel_map(&pConverter->channelConverter, pChannelMap, channelMapCap); + } else { + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pConverter->channelsOut); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_converter_get_output_channel_map(const ma_data_converter* pConverter, ma_channel* pChannelMap, size_t channelMapCap) +{ + if (pConverter == NULL || pChannelMap == NULL) { + return MA_INVALID_ARGS; + } + + if (pConverter->hasChannelConverter) { + ma_channel_converter_get_input_channel_map(&pConverter->channelConverter, pChannelMap, channelMapCap); + } else { + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pConverter->channelsIn); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_converter_reset(ma_data_converter* pConverter) +{ + if (pConverter == NULL) { + return MA_INVALID_ARGS; + } + + /* There's nothing to do if we're not resampling. */ + if (pConverter->hasResampler == MA_FALSE) { + return MA_SUCCESS; + } + + return ma_resampler_reset(&pConverter->resampler); +} + + + +/************************************************************************************************************************************************************** + +Channel Maps + +**************************************************************************************************************************************************************/ +static ma_channel ma_channel_map_init_standard_channel(ma_standard_channel_map standardChannelMap, ma_uint32 channelCount, ma_uint32 channelIndex); + +MA_API ma_channel ma_channel_map_get_channel(const ma_channel* pChannelMap, ma_uint32 channelCount, ma_uint32 channelIndex) +{ + if (pChannelMap == NULL) { + return ma_channel_map_init_standard_channel(ma_standard_channel_map_default, channelCount, channelIndex); + } else { + if (channelIndex >= channelCount) { + return MA_CHANNEL_NONE; + } + + return pChannelMap[channelIndex]; + } +} + +MA_API void ma_channel_map_init_blank(ma_channel* pChannelMap, ma_uint32 channels) +{ + if (pChannelMap == NULL) { + return; + } + + MA_ZERO_MEMORY(pChannelMap, sizeof(*pChannelMap) * channels); +} + + +static ma_channel ma_channel_map_init_standard_channel_microsoft(ma_uint32 channelCount, ma_uint32 channelIndex) +{ + if (channelCount == 0 || channelIndex >= channelCount) { + return MA_CHANNEL_NONE; + } + + /* This is the Microsoft channel map. Based off the speaker configurations mentioned here: https://docs.microsoft.com/en-us/windows-hardware/drivers/ddi/content/ksmedia/ns-ksmedia-ksaudio_channel_config */ + switch (channelCount) + { + case 0: return MA_CHANNEL_NONE; + + case 1: + { + return MA_CHANNEL_MONO; + } break; + + case 2: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + } + } break; + + case 3: /* No defined, but best guess. */ + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + } + } break; + + case 4: + { + switch (channelIndex) { + #ifndef MA_USE_QUAD_MICROSOFT_CHANNEL_MAP + /* Surround. Using the Surround profile has the advantage of the 3rd channel (MA_CHANNEL_FRONT_CENTER) mapping nicely with higher channel counts. */ + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_BACK_CENTER; + #else + /* Quad. */ + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + #endif + } + } break; + + case 5: /* Not defined, but best guess. */ + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_BACK_LEFT; + case 4: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 6: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_LFE; + case 4: return MA_CHANNEL_SIDE_LEFT; + case 5: return MA_CHANNEL_SIDE_RIGHT; + } + } break; + + case 7: /* Not defined, but best guess. */ + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_LFE; + case 4: return MA_CHANNEL_BACK_CENTER; + case 5: return MA_CHANNEL_SIDE_LEFT; + case 6: return MA_CHANNEL_SIDE_RIGHT; + } + } break; + + case 8: + default: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_LFE; + case 4: return MA_CHANNEL_BACK_LEFT; + case 5: return MA_CHANNEL_BACK_RIGHT; + case 6: return MA_CHANNEL_SIDE_LEFT; + case 7: return MA_CHANNEL_SIDE_RIGHT; + } + } break; + } + + if (channelCount > 8) { + if (channelIndex < 32) { /* We have 32 AUX channels. */ + return (ma_channel)(MA_CHANNEL_AUX_0 + (channelIndex - 8)); + } + } + + /* Getting here means we don't know how to map the channel position so just return MA_CHANNEL_NONE. */ + return MA_CHANNEL_NONE; +} + +static ma_channel ma_channel_map_init_standard_channel_alsa(ma_uint32 channelCount, ma_uint32 channelIndex) +{ + switch (channelCount) + { + case 0: return MA_CHANNEL_NONE; + + case 1: + { + return MA_CHANNEL_MONO; + } break; + + case 2: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + } + } break; + + case 3: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + } + } break; + + case 4: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 5: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + case 4: return MA_CHANNEL_FRONT_CENTER; + } + } break; + + case 6: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + case 4: return MA_CHANNEL_FRONT_CENTER; + case 5: return MA_CHANNEL_LFE; + } + } break; + + case 7: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + case 4: return MA_CHANNEL_FRONT_CENTER; + case 5: return MA_CHANNEL_LFE; + case 6: return MA_CHANNEL_BACK_CENTER; + } + } break; + + case 8: + default: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + case 4: return MA_CHANNEL_FRONT_CENTER; + case 5: return MA_CHANNEL_LFE; + case 6: return MA_CHANNEL_SIDE_LEFT; + case 7: return MA_CHANNEL_SIDE_RIGHT; + } + } break; + } + + if (channelCount > 8) { + if (channelIndex < 32) { /* We have 32 AUX channels. */ + return (ma_channel)(MA_CHANNEL_AUX_0 + (channelIndex - 8)); + } + } + + /* Getting here means we don't know how to map the channel position so just return MA_CHANNEL_NONE. */ + return MA_CHANNEL_NONE; +} + +static ma_channel ma_channel_map_init_standard_channel_rfc3551(ma_uint32 channelCount, ma_uint32 channelIndex) +{ + switch (channelCount) + { + case 0: return MA_CHANNEL_NONE; + + case 1: + { + return MA_CHANNEL_MONO; + } break; + + case 2: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + } + } break; + + case 3: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + } + } break; + + case 4: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 3: return MA_CHANNEL_BACK_CENTER; + } + } break; + + case 5: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_BACK_LEFT; + case 4: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 6: + default: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_SIDE_LEFT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_FRONT_RIGHT; + case 4: return MA_CHANNEL_SIDE_RIGHT; + case 5: return MA_CHANNEL_BACK_CENTER; + } + } break; + } + + if (channelCount > 6) { + if (channelIndex < 32) { /* We have 32 AUX channels. */ + return (ma_channel)(MA_CHANNEL_AUX_0 + (channelIndex - 6)); + } + } + + /* Getting here means we don't know how to map the channel position so just return MA_CHANNEL_NONE. */ + return MA_CHANNEL_NONE; +} + +static ma_channel ma_channel_map_init_standard_channel_flac(ma_uint32 channelCount, ma_uint32 channelIndex) +{ + switch (channelCount) + { + case 0: return MA_CHANNEL_NONE; + + case 1: + { + return MA_CHANNEL_MONO; + } break; + + case 2: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + } + } break; + + case 3: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + } + } break; + + case 4: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 5: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_BACK_LEFT; + case 4: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 6: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_LFE; + case 4: return MA_CHANNEL_BACK_LEFT; + case 5: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 7: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_LFE; + case 4: return MA_CHANNEL_BACK_CENTER; + case 5: return MA_CHANNEL_SIDE_LEFT; + case 6: return MA_CHANNEL_SIDE_RIGHT; + } + } break; + + case 8: + default: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_LFE; + case 4: return MA_CHANNEL_BACK_LEFT; + case 5: return MA_CHANNEL_BACK_RIGHT; + case 6: return MA_CHANNEL_SIDE_LEFT; + case 7: return MA_CHANNEL_SIDE_RIGHT; + } + } break; + } + + if (channelCount > 8) { + if (channelIndex < 32) { /* We have 32 AUX channels. */ + return (ma_channel)(MA_CHANNEL_AUX_0 + (channelIndex - 8)); + } + } + + /* Getting here means we don't know how to map the channel position so just return MA_CHANNEL_NONE. */ + return MA_CHANNEL_NONE; +} + +static ma_channel ma_channel_map_init_standard_channel_vorbis(ma_uint32 channelCount, ma_uint32 channelIndex) +{ + switch (channelCount) + { + case 0: return MA_CHANNEL_NONE; + + case 1: + { + return MA_CHANNEL_MONO; + } break; + + case 2: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + } + } break; + + case 3: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_CENTER; + case 2: return MA_CHANNEL_FRONT_RIGHT; + } + } break; + + case 4: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 5: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_CENTER; + case 2: return MA_CHANNEL_FRONT_RIGHT; + case 3: return MA_CHANNEL_BACK_LEFT; + case 4: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 6: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_CENTER; + case 2: return MA_CHANNEL_FRONT_RIGHT; + case 3: return MA_CHANNEL_BACK_LEFT; + case 4: return MA_CHANNEL_BACK_RIGHT; + case 5: return MA_CHANNEL_LFE; + } + } break; + + case 7: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_CENTER; + case 2: return MA_CHANNEL_FRONT_RIGHT; + case 3: return MA_CHANNEL_SIDE_LEFT; + case 4: return MA_CHANNEL_SIDE_RIGHT; + case 5: return MA_CHANNEL_BACK_CENTER; + case 6: return MA_CHANNEL_LFE; + } + } break; + + case 8: + default: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_CENTER; + case 2: return MA_CHANNEL_FRONT_RIGHT; + case 3: return MA_CHANNEL_SIDE_LEFT; + case 4: return MA_CHANNEL_SIDE_RIGHT; + case 5: return MA_CHANNEL_BACK_LEFT; + case 6: return MA_CHANNEL_BACK_RIGHT; + case 7: return MA_CHANNEL_LFE; + } + } break; + } + + if (channelCount > 8) { + if (channelIndex < 32) { /* We have 32 AUX channels. */ + return (ma_channel)(MA_CHANNEL_AUX_0 + (channelIndex - 8)); + } + } + + /* Getting here means we don't know how to map the channel position so just return MA_CHANNEL_NONE. */ + return MA_CHANNEL_NONE; +} + +static ma_channel ma_channel_map_init_standard_channel_sound4(ma_uint32 channelCount, ma_uint32 channelIndex) +{ + switch (channelCount) + { + case 0: return MA_CHANNEL_NONE; + + case 1: + { + return MA_CHANNEL_MONO; + } break; + + case 2: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + } + } break; + + case 3: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + } + } break; + + case 4: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 5: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + case 3: return MA_CHANNEL_BACK_LEFT; + case 4: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 6: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_CENTER; + case 2: return MA_CHANNEL_FRONT_RIGHT; + case 3: return MA_CHANNEL_BACK_LEFT; + case 4: return MA_CHANNEL_BACK_RIGHT; + case 5: return MA_CHANNEL_LFE; + } + } break; + + case 7: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_CENTER; + case 2: return MA_CHANNEL_FRONT_RIGHT; + case 3: return MA_CHANNEL_SIDE_LEFT; + case 4: return MA_CHANNEL_SIDE_RIGHT; + case 5: return MA_CHANNEL_BACK_CENTER; + case 6: return MA_CHANNEL_LFE; + } + } break; + + case 8: + default: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_CENTER; + case 2: return MA_CHANNEL_FRONT_RIGHT; + case 3: return MA_CHANNEL_SIDE_LEFT; + case 4: return MA_CHANNEL_SIDE_RIGHT; + case 5: return MA_CHANNEL_BACK_LEFT; + case 6: return MA_CHANNEL_BACK_RIGHT; + case 7: return MA_CHANNEL_LFE; + } + } break; + } + + if (channelCount > 8) { + if (channelIndex < 32) { /* We have 32 AUX channels. */ + return (ma_channel)(MA_CHANNEL_AUX_0 + (channelIndex - 8)); + } + } + + /* Getting here means we don't know how to map the channel position so just return MA_CHANNEL_NONE. */ + return MA_CHANNEL_NONE; +} + +static ma_channel ma_channel_map_init_standard_channel_sndio(ma_uint32 channelCount, ma_uint32 channelIndex) +{ + switch (channelCount) + { + case 0: return MA_CHANNEL_NONE; + + case 1: + { + return MA_CHANNEL_MONO; + } break; + + case 2: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + } + } break; + + case 3: /* No defined, but best guess. */ + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_FRONT_CENTER; + } + } break; + + case 4: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + } + } break; + + case 5: /* Not defined, but best guess. */ + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + case 4: return MA_CHANNEL_FRONT_CENTER; + } + } break; + + case 6: + default: + { + switch (channelIndex) { + case 0: return MA_CHANNEL_FRONT_LEFT; + case 1: return MA_CHANNEL_FRONT_RIGHT; + case 2: return MA_CHANNEL_BACK_LEFT; + case 3: return MA_CHANNEL_BACK_RIGHT; + case 4: return MA_CHANNEL_FRONT_CENTER; + case 5: return MA_CHANNEL_LFE; + } + } break; + } + + if (channelCount > 6) { + if (channelIndex < 32) { /* We have 32 AUX channels. */ + return (ma_channel)(MA_CHANNEL_AUX_0 + (channelIndex - 6)); + } + } + + /* Getting here means we don't know how to map the channel position so just return MA_CHANNEL_NONE. */ + return MA_CHANNEL_NONE; +} + + +static ma_channel ma_channel_map_init_standard_channel(ma_standard_channel_map standardChannelMap, ma_uint32 channelCount, ma_uint32 channelIndex) +{ + if (channelCount == 0 || channelIndex >= channelCount) { + return MA_CHANNEL_NONE; + } + + switch (standardChannelMap) + { + case ma_standard_channel_map_alsa: + { + return ma_channel_map_init_standard_channel_alsa(channelCount, channelIndex); + } break; + + case ma_standard_channel_map_rfc3551: + { + return ma_channel_map_init_standard_channel_rfc3551(channelCount, channelIndex); + } break; + + case ma_standard_channel_map_flac: + { + return ma_channel_map_init_standard_channel_flac(channelCount, channelIndex); + } break; + + case ma_standard_channel_map_vorbis: + { + return ma_channel_map_init_standard_channel_vorbis(channelCount, channelIndex); + } break; + + case ma_standard_channel_map_sound4: + { + return ma_channel_map_init_standard_channel_sound4(channelCount, channelIndex); + } break; + + case ma_standard_channel_map_sndio: + { + return ma_channel_map_init_standard_channel_sndio(channelCount, channelIndex); + } break; + + case ma_standard_channel_map_microsoft: /* Also default. */ + /*case ma_standard_channel_map_default;*/ + default: + { + return ma_channel_map_init_standard_channel_microsoft(channelCount, channelIndex); + } break; + } +} + +MA_API void ma_channel_map_init_standard(ma_standard_channel_map standardChannelMap, ma_channel* pChannelMap, size_t channelMapCap, ma_uint32 channels) +{ + ma_uint32 iChannel; + + if (pChannelMap == NULL || channelMapCap == 0 || channels == 0) { + return; + } + + for (iChannel = 0; iChannel < channels; iChannel += 1) { + if (channelMapCap == 0) { + break; /* Ran out of room. */ + } + + pChannelMap[0] = ma_channel_map_init_standard_channel(standardChannelMap, channels, iChannel); + pChannelMap += 1; + channelMapCap -= 1; + } +} + +MA_API void ma_channel_map_copy(ma_channel* pOut, const ma_channel* pIn, ma_uint32 channels) +{ + if (pOut != NULL && pIn != NULL && channels > 0) { + MA_COPY_MEMORY(pOut, pIn, sizeof(*pOut) * channels); + } +} + +MA_API void ma_channel_map_copy_or_default(ma_channel* pOut, size_t channelMapCapOut, const ma_channel* pIn, ma_uint32 channels) +{ + if (pOut == NULL || channels == 0) { + return; + } + + if (pIn != NULL) { + ma_channel_map_copy(pOut, pIn, channels); + } else { + ma_channel_map_init_standard(ma_standard_channel_map_default, pOut, channelMapCapOut, channels); + } +} + +MA_API ma_bool32 ma_channel_map_is_valid(const ma_channel* pChannelMap, ma_uint32 channels) +{ + /* A channel count of 0 is invalid. */ + if (channels == 0) { + return MA_FALSE; + } + + /* It does not make sense to have a mono channel when there is more than 1 channel. */ + if (channels > 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < channels; ++iChannel) { + if (ma_channel_map_get_channel(pChannelMap, channels, iChannel) == MA_CHANNEL_MONO) { + return MA_FALSE; + } + } + } + + return MA_TRUE; +} + +MA_API ma_bool32 ma_channel_map_is_equal(const ma_channel* pChannelMapA, const ma_channel* pChannelMapB, ma_uint32 channels) +{ + ma_uint32 iChannel; + + if (pChannelMapA == pChannelMapB) { + return MA_TRUE; + } + + for (iChannel = 0; iChannel < channels; ++iChannel) { + if (ma_channel_map_get_channel(pChannelMapA, channels, iChannel) != ma_channel_map_get_channel(pChannelMapB, channels, iChannel)) { + return MA_FALSE; + } + } + + return MA_TRUE; +} + +MA_API ma_bool32 ma_channel_map_is_blank(const ma_channel* pChannelMap, ma_uint32 channels) +{ + ma_uint32 iChannel; + + /* A null channel map is equivalent to the default channel map. */ + if (pChannelMap == NULL) { + return MA_FALSE; + } + + for (iChannel = 0; iChannel < channels; ++iChannel) { + if (pChannelMap[iChannel] != MA_CHANNEL_NONE) { + return MA_FALSE; + } + } + + return MA_TRUE; +} + +MA_API ma_bool32 ma_channel_map_contains_channel_position(ma_uint32 channels, const ma_channel* pChannelMap, ma_channel channelPosition) +{ + return ma_channel_map_find_channel_position(channels, pChannelMap, channelPosition, NULL); +} + +MA_API ma_bool32 ma_channel_map_find_channel_position(ma_uint32 channels, const ma_channel* pChannelMap, ma_channel channelPosition, ma_uint32* pChannelIndex) +{ + ma_uint32 iChannel; + + if (pChannelIndex != NULL) { + *pChannelIndex = (ma_uint32)-1; + } + + for (iChannel = 0; iChannel < channels; ++iChannel) { + if (ma_channel_map_get_channel(pChannelMap, channels, iChannel) == channelPosition) { + if (pChannelIndex != NULL) { + *pChannelIndex = iChannel; + } + + return MA_TRUE; + } + } + + /* Getting here means the channel position was not found. */ + return MA_FALSE; +} + +MA_API size_t ma_channel_map_to_string(const ma_channel* pChannelMap, ma_uint32 channels, char* pBufferOut, size_t bufferCap) +{ + size_t len; + ma_uint32 iChannel; + + len = 0; + + for (iChannel = 0; iChannel < channels; iChannel += 1) { + const char* pChannelStr = ma_channel_position_to_string(ma_channel_map_get_channel(pChannelMap, channels, iChannel)); + size_t channelStrLen = strlen(pChannelStr); + + /* Append the string if necessary. */ + if (pBufferOut != NULL && bufferCap > len + channelStrLen) { + MA_COPY_MEMORY(pBufferOut + len, pChannelStr, channelStrLen); + } + len += channelStrLen; + + /* Append a space if it's not the last item. */ + if (iChannel+1 < channels) { + if (pBufferOut != NULL && bufferCap > len + 1) { + pBufferOut[len] = ' '; + } + len += 1; + } + } + + /* Null terminate. Don't increment the length here. */ + if (pBufferOut != NULL && bufferCap > len + 1) { + pBufferOut[len] = '\0'; + } + + return len; +} + +MA_API const char* ma_channel_position_to_string(ma_channel channel) +{ + switch (channel) + { + case MA_CHANNEL_NONE : return "CHANNEL_NONE"; + case MA_CHANNEL_MONO : return "CHANNEL_MONO"; + case MA_CHANNEL_FRONT_LEFT : return "CHANNEL_FRONT_LEFT"; + case MA_CHANNEL_FRONT_RIGHT : return "CHANNEL_FRONT_RIGHT"; + case MA_CHANNEL_FRONT_CENTER : return "CHANNEL_FRONT_CENTER"; + case MA_CHANNEL_LFE : return "CHANNEL_LFE"; + case MA_CHANNEL_BACK_LEFT : return "CHANNEL_BACK_LEFT"; + case MA_CHANNEL_BACK_RIGHT : return "CHANNEL_BACK_RIGHT"; + case MA_CHANNEL_FRONT_LEFT_CENTER : return "CHANNEL_FRONT_LEFT_CENTER"; + case MA_CHANNEL_FRONT_RIGHT_CENTER: return "CHANNEL_FRONT_RIGHT_CENTER"; + case MA_CHANNEL_BACK_CENTER : return "CHANNEL_BACK_CENTER"; + case MA_CHANNEL_SIDE_LEFT : return "CHANNEL_SIDE_LEFT"; + case MA_CHANNEL_SIDE_RIGHT : return "CHANNEL_SIDE_RIGHT"; + case MA_CHANNEL_TOP_CENTER : return "CHANNEL_TOP_CENTER"; + case MA_CHANNEL_TOP_FRONT_LEFT : return "CHANNEL_TOP_FRONT_LEFT"; + case MA_CHANNEL_TOP_FRONT_CENTER : return "CHANNEL_TOP_FRONT_CENTER"; + case MA_CHANNEL_TOP_FRONT_RIGHT : return "CHANNEL_TOP_FRONT_RIGHT"; + case MA_CHANNEL_TOP_BACK_LEFT : return "CHANNEL_TOP_BACK_LEFT"; + case MA_CHANNEL_TOP_BACK_CENTER : return "CHANNEL_TOP_BACK_CENTER"; + case MA_CHANNEL_TOP_BACK_RIGHT : return "CHANNEL_TOP_BACK_RIGHT"; + case MA_CHANNEL_AUX_0 : return "CHANNEL_AUX_0"; + case MA_CHANNEL_AUX_1 : return "CHANNEL_AUX_1"; + case MA_CHANNEL_AUX_2 : return "CHANNEL_AUX_2"; + case MA_CHANNEL_AUX_3 : return "CHANNEL_AUX_3"; + case MA_CHANNEL_AUX_4 : return "CHANNEL_AUX_4"; + case MA_CHANNEL_AUX_5 : return "CHANNEL_AUX_5"; + case MA_CHANNEL_AUX_6 : return "CHANNEL_AUX_6"; + case MA_CHANNEL_AUX_7 : return "CHANNEL_AUX_7"; + case MA_CHANNEL_AUX_8 : return "CHANNEL_AUX_8"; + case MA_CHANNEL_AUX_9 : return "CHANNEL_AUX_9"; + case MA_CHANNEL_AUX_10 : return "CHANNEL_AUX_10"; + case MA_CHANNEL_AUX_11 : return "CHANNEL_AUX_11"; + case MA_CHANNEL_AUX_12 : return "CHANNEL_AUX_12"; + case MA_CHANNEL_AUX_13 : return "CHANNEL_AUX_13"; + case MA_CHANNEL_AUX_14 : return "CHANNEL_AUX_14"; + case MA_CHANNEL_AUX_15 : return "CHANNEL_AUX_15"; + case MA_CHANNEL_AUX_16 : return "CHANNEL_AUX_16"; + case MA_CHANNEL_AUX_17 : return "CHANNEL_AUX_17"; + case MA_CHANNEL_AUX_18 : return "CHANNEL_AUX_18"; + case MA_CHANNEL_AUX_19 : return "CHANNEL_AUX_19"; + case MA_CHANNEL_AUX_20 : return "CHANNEL_AUX_20"; + case MA_CHANNEL_AUX_21 : return "CHANNEL_AUX_21"; + case MA_CHANNEL_AUX_22 : return "CHANNEL_AUX_22"; + case MA_CHANNEL_AUX_23 : return "CHANNEL_AUX_23"; + case MA_CHANNEL_AUX_24 : return "CHANNEL_AUX_24"; + case MA_CHANNEL_AUX_25 : return "CHANNEL_AUX_25"; + case MA_CHANNEL_AUX_26 : return "CHANNEL_AUX_26"; + case MA_CHANNEL_AUX_27 : return "CHANNEL_AUX_27"; + case MA_CHANNEL_AUX_28 : return "CHANNEL_AUX_28"; + case MA_CHANNEL_AUX_29 : return "CHANNEL_AUX_29"; + case MA_CHANNEL_AUX_30 : return "CHANNEL_AUX_30"; + case MA_CHANNEL_AUX_31 : return "CHANNEL_AUX_31"; + default: break; + } + + return "UNKNOWN"; +} + + + +/************************************************************************************************************************************************************** + +Conversion Helpers + +**************************************************************************************************************************************************************/ +MA_API ma_uint64 ma_convert_frames(void* pOut, ma_uint64 frameCountOut, ma_format formatOut, ma_uint32 channelsOut, ma_uint32 sampleRateOut, const void* pIn, ma_uint64 frameCountIn, ma_format formatIn, ma_uint32 channelsIn, ma_uint32 sampleRateIn) +{ + ma_data_converter_config config; + + config = ma_data_converter_config_init(formatIn, formatOut, channelsIn, channelsOut, sampleRateIn, sampleRateOut); + config.resampling.linear.lpfOrder = ma_min(MA_DEFAULT_RESAMPLER_LPF_ORDER, MA_MAX_FILTER_ORDER); + + return ma_convert_frames_ex(pOut, frameCountOut, pIn, frameCountIn, &config); +} + +MA_API ma_uint64 ma_convert_frames_ex(void* pOut, ma_uint64 frameCountOut, const void* pIn, ma_uint64 frameCountIn, const ma_data_converter_config* pConfig) +{ + ma_result result; + ma_data_converter converter; + + if (frameCountIn == 0 || pConfig == NULL) { + return 0; + } + + result = ma_data_converter_init(pConfig, NULL, &converter); + if (result != MA_SUCCESS) { + return 0; /* Failed to initialize the data converter. */ + } + + if (pOut == NULL) { + result = ma_data_converter_get_expected_output_frame_count(&converter, frameCountIn, &frameCountOut); + if (result != MA_SUCCESS) { + if (result == MA_NOT_IMPLEMENTED) { + /* No way to calculate the number of frames, so we'll need to brute force it and loop. */ + frameCountOut = 0; + + while (frameCountIn > 0) { + ma_uint64 framesProcessedIn = frameCountIn; + ma_uint64 framesProcessedOut = 0xFFFFFFFF; + + result = ma_data_converter_process_pcm_frames(&converter, pIn, &framesProcessedIn, NULL, &framesProcessedOut); + if (result != MA_SUCCESS) { + break; + } + + frameCountIn -= framesProcessedIn; + } + } + } + } else { + result = ma_data_converter_process_pcm_frames(&converter, pIn, &frameCountIn, pOut, &frameCountOut); + if (result != MA_SUCCESS) { + frameCountOut = 0; + } + } + + ma_data_converter_uninit(&converter, NULL); + return frameCountOut; +} + + +/************************************************************************************************************************************************************** + +Ring Buffer + +**************************************************************************************************************************************************************/ +static MA_INLINE ma_uint32 ma_rb__extract_offset_in_bytes(ma_uint32 encodedOffset) +{ + return encodedOffset & 0x7FFFFFFF; +} + +static MA_INLINE ma_uint32 ma_rb__extract_offset_loop_flag(ma_uint32 encodedOffset) +{ + return encodedOffset & 0x80000000; +} + +static MA_INLINE void* ma_rb__get_read_ptr(ma_rb* pRB) +{ + MA_ASSERT(pRB != NULL); + return ma_offset_ptr(pRB->pBuffer, ma_rb__extract_offset_in_bytes(ma_atomic_load_32(&pRB->encodedReadOffset))); +} + +static MA_INLINE void* ma_rb__get_write_ptr(ma_rb* pRB) +{ + MA_ASSERT(pRB != NULL); + return ma_offset_ptr(pRB->pBuffer, ma_rb__extract_offset_in_bytes(ma_atomic_load_32(&pRB->encodedWriteOffset))); +} + +static MA_INLINE ma_uint32 ma_rb__construct_offset(ma_uint32 offsetInBytes, ma_uint32 offsetLoopFlag) +{ + return offsetLoopFlag | offsetInBytes; +} + +static MA_INLINE void ma_rb__deconstruct_offset(ma_uint32 encodedOffset, ma_uint32* pOffsetInBytes, ma_uint32* pOffsetLoopFlag) +{ + MA_ASSERT(pOffsetInBytes != NULL); + MA_ASSERT(pOffsetLoopFlag != NULL); + + *pOffsetInBytes = ma_rb__extract_offset_in_bytes(encodedOffset); + *pOffsetLoopFlag = ma_rb__extract_offset_loop_flag(encodedOffset); +} + + +MA_API ma_result ma_rb_init_ex(size_t subbufferSizeInBytes, size_t subbufferCount, size_t subbufferStrideInBytes, void* pOptionalPreallocatedBuffer, const ma_allocation_callbacks* pAllocationCallbacks, ma_rb* pRB) +{ + ma_result result; + const ma_uint32 maxSubBufferSize = 0x7FFFFFFF - (MA_SIMD_ALIGNMENT-1); + + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + if (subbufferSizeInBytes == 0 || subbufferCount == 0) { + return MA_INVALID_ARGS; + } + + if (subbufferSizeInBytes > maxSubBufferSize) { + return MA_INVALID_ARGS; /* Maximum buffer size is ~2GB. The most significant bit is a flag for use internally. */ + } + + + MA_ZERO_OBJECT(pRB); + + result = ma_allocation_callbacks_init_copy(&pRB->allocationCallbacks, pAllocationCallbacks); + if (result != MA_SUCCESS) { + return result; + } + + pRB->subbufferSizeInBytes = (ma_uint32)subbufferSizeInBytes; + pRB->subbufferCount = (ma_uint32)subbufferCount; + + if (pOptionalPreallocatedBuffer != NULL) { + pRB->subbufferStrideInBytes = (ma_uint32)subbufferStrideInBytes; + pRB->pBuffer = pOptionalPreallocatedBuffer; + } else { + size_t bufferSizeInBytes; + + /* + Here is where we allocate our own buffer. We always want to align this to MA_SIMD_ALIGNMENT for future SIMD optimization opportunity. To do this + we need to make sure the stride is a multiple of MA_SIMD_ALIGNMENT. + */ + pRB->subbufferStrideInBytes = (pRB->subbufferSizeInBytes + (MA_SIMD_ALIGNMENT-1)) & ~MA_SIMD_ALIGNMENT; + + bufferSizeInBytes = (size_t)pRB->subbufferCount*pRB->subbufferStrideInBytes; + pRB->pBuffer = ma_aligned_malloc(bufferSizeInBytes, MA_SIMD_ALIGNMENT, &pRB->allocationCallbacks); + if (pRB->pBuffer == NULL) { + return MA_OUT_OF_MEMORY; + } + + MA_ZERO_MEMORY(pRB->pBuffer, bufferSizeInBytes); + pRB->ownsBuffer = MA_TRUE; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_rb_init(size_t bufferSizeInBytes, void* pOptionalPreallocatedBuffer, const ma_allocation_callbacks* pAllocationCallbacks, ma_rb* pRB) +{ + return ma_rb_init_ex(bufferSizeInBytes, 1, 0, pOptionalPreallocatedBuffer, pAllocationCallbacks, pRB); +} + +MA_API void ma_rb_uninit(ma_rb* pRB) +{ + if (pRB == NULL) { + return; + } + + if (pRB->ownsBuffer) { + ma_aligned_free(pRB->pBuffer, &pRB->allocationCallbacks); + } +} + +MA_API void ma_rb_reset(ma_rb* pRB) +{ + if (pRB == NULL) { + return; + } + + ma_atomic_exchange_32(&pRB->encodedReadOffset, 0); + ma_atomic_exchange_32(&pRB->encodedWriteOffset, 0); +} + +MA_API ma_result ma_rb_acquire_read(ma_rb* pRB, size_t* pSizeInBytes, void** ppBufferOut) +{ + ma_uint32 writeOffset; + ma_uint32 writeOffsetInBytes; + ma_uint32 writeOffsetLoopFlag; + ma_uint32 readOffset; + ma_uint32 readOffsetInBytes; + ma_uint32 readOffsetLoopFlag; + size_t bytesAvailable; + size_t bytesRequested; + + if (pRB == NULL || pSizeInBytes == NULL || ppBufferOut == NULL) { + return MA_INVALID_ARGS; + } + + /* The returned buffer should never move ahead of the write pointer. */ + writeOffset = ma_atomic_load_32(&pRB->encodedWriteOffset); + ma_rb__deconstruct_offset(writeOffset, &writeOffsetInBytes, &writeOffsetLoopFlag); + + readOffset = ma_atomic_load_32(&pRB->encodedReadOffset); + ma_rb__deconstruct_offset(readOffset, &readOffsetInBytes, &readOffsetLoopFlag); + + /* + The number of bytes available depends on whether or not the read and write pointers are on the same loop iteration. If so, we + can only read up to the write pointer. If not, we can only read up to the end of the buffer. + */ + if (readOffsetLoopFlag == writeOffsetLoopFlag) { + bytesAvailable = writeOffsetInBytes - readOffsetInBytes; + } else { + bytesAvailable = pRB->subbufferSizeInBytes - readOffsetInBytes; + } + + bytesRequested = *pSizeInBytes; + if (bytesRequested > bytesAvailable) { + bytesRequested = bytesAvailable; + } + + *pSizeInBytes = bytesRequested; + (*ppBufferOut) = ma_rb__get_read_ptr(pRB); + + return MA_SUCCESS; +} + +MA_API ma_result ma_rb_commit_read(ma_rb* pRB, size_t sizeInBytes) +{ + ma_uint32 readOffset; + ma_uint32 readOffsetInBytes; + ma_uint32 readOffsetLoopFlag; + ma_uint32 newReadOffsetInBytes; + ma_uint32 newReadOffsetLoopFlag; + + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + readOffset = ma_atomic_load_32(&pRB->encodedReadOffset); + ma_rb__deconstruct_offset(readOffset, &readOffsetInBytes, &readOffsetLoopFlag); + + /* Check that sizeInBytes is correct. It should never go beyond the end of the buffer. */ + newReadOffsetInBytes = (ma_uint32)(readOffsetInBytes + sizeInBytes); + if (newReadOffsetInBytes > pRB->subbufferSizeInBytes) { + return MA_INVALID_ARGS; /* <-- sizeInBytes will cause the read offset to overflow. */ + } + + /* Move the read pointer back to the start if necessary. */ + newReadOffsetLoopFlag = readOffsetLoopFlag; + if (newReadOffsetInBytes == pRB->subbufferSizeInBytes) { + newReadOffsetInBytes = 0; + newReadOffsetLoopFlag ^= 0x80000000; + } + + ma_atomic_exchange_32(&pRB->encodedReadOffset, ma_rb__construct_offset(newReadOffsetInBytes, newReadOffsetLoopFlag)); + + return MA_SUCCESS; +} + +MA_API ma_result ma_rb_acquire_write(ma_rb* pRB, size_t* pSizeInBytes, void** ppBufferOut) +{ + ma_uint32 readOffset; + ma_uint32 readOffsetInBytes; + ma_uint32 readOffsetLoopFlag; + ma_uint32 writeOffset; + ma_uint32 writeOffsetInBytes; + ma_uint32 writeOffsetLoopFlag; + size_t bytesAvailable; + size_t bytesRequested; + + if (pRB == NULL || pSizeInBytes == NULL || ppBufferOut == NULL) { + return MA_INVALID_ARGS; + } + + /* The returned buffer should never overtake the read buffer. */ + readOffset = ma_atomic_load_32(&pRB->encodedReadOffset); + ma_rb__deconstruct_offset(readOffset, &readOffsetInBytes, &readOffsetLoopFlag); + + writeOffset = ma_atomic_load_32(&pRB->encodedWriteOffset); + ma_rb__deconstruct_offset(writeOffset, &writeOffsetInBytes, &writeOffsetLoopFlag); + + /* + In the case of writing, if the write pointer and the read pointer are on the same loop iteration we can only + write up to the end of the buffer. Otherwise we can only write up to the read pointer. The write pointer should + never overtake the read pointer. + */ + if (writeOffsetLoopFlag == readOffsetLoopFlag) { + bytesAvailable = pRB->subbufferSizeInBytes - writeOffsetInBytes; + } else { + bytesAvailable = readOffsetInBytes - writeOffsetInBytes; + } + + bytesRequested = *pSizeInBytes; + if (bytesRequested > bytesAvailable) { + bytesRequested = bytesAvailable; + } + + *pSizeInBytes = bytesRequested; + *ppBufferOut = ma_rb__get_write_ptr(pRB); + + /* Clear the buffer if desired. */ + if (pRB->clearOnWriteAcquire) { + MA_ZERO_MEMORY(*ppBufferOut, *pSizeInBytes); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_rb_commit_write(ma_rb* pRB, size_t sizeInBytes) +{ + ma_uint32 writeOffset; + ma_uint32 writeOffsetInBytes; + ma_uint32 writeOffsetLoopFlag; + ma_uint32 newWriteOffsetInBytes; + ma_uint32 newWriteOffsetLoopFlag; + + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + writeOffset = ma_atomic_load_32(&pRB->encodedWriteOffset); + ma_rb__deconstruct_offset(writeOffset, &writeOffsetInBytes, &writeOffsetLoopFlag); + + /* Check that sizeInBytes is correct. It should never go beyond the end of the buffer. */ + newWriteOffsetInBytes = (ma_uint32)(writeOffsetInBytes + sizeInBytes); + if (newWriteOffsetInBytes > pRB->subbufferSizeInBytes) { + return MA_INVALID_ARGS; /* <-- sizeInBytes will cause the read offset to overflow. */ + } + + /* Move the read pointer back to the start if necessary. */ + newWriteOffsetLoopFlag = writeOffsetLoopFlag; + if (newWriteOffsetInBytes == pRB->subbufferSizeInBytes) { + newWriteOffsetInBytes = 0; + newWriteOffsetLoopFlag ^= 0x80000000; + } + + ma_atomic_exchange_32(&pRB->encodedWriteOffset, ma_rb__construct_offset(newWriteOffsetInBytes, newWriteOffsetLoopFlag)); + + return MA_SUCCESS; +} + +MA_API ma_result ma_rb_seek_read(ma_rb* pRB, size_t offsetInBytes) +{ + ma_uint32 readOffset; + ma_uint32 readOffsetInBytes; + ma_uint32 readOffsetLoopFlag; + ma_uint32 writeOffset; + ma_uint32 writeOffsetInBytes; + ma_uint32 writeOffsetLoopFlag; + ma_uint32 newReadOffsetInBytes; + ma_uint32 newReadOffsetLoopFlag; + + if (pRB == NULL || offsetInBytes > pRB->subbufferSizeInBytes) { + return MA_INVALID_ARGS; + } + + readOffset = ma_atomic_load_32(&pRB->encodedReadOffset); + ma_rb__deconstruct_offset(readOffset, &readOffsetInBytes, &readOffsetLoopFlag); + + writeOffset = ma_atomic_load_32(&pRB->encodedWriteOffset); + ma_rb__deconstruct_offset(writeOffset, &writeOffsetInBytes, &writeOffsetLoopFlag); + + newReadOffsetLoopFlag = readOffsetLoopFlag; + + /* We cannot go past the write buffer. */ + if (readOffsetLoopFlag == writeOffsetLoopFlag) { + if ((readOffsetInBytes + offsetInBytes) > writeOffsetInBytes) { + newReadOffsetInBytes = writeOffsetInBytes; + } else { + newReadOffsetInBytes = (ma_uint32)(readOffsetInBytes + offsetInBytes); + } + } else { + /* May end up looping. */ + if ((readOffsetInBytes + offsetInBytes) >= pRB->subbufferSizeInBytes) { + newReadOffsetInBytes = (ma_uint32)(readOffsetInBytes + offsetInBytes) - pRB->subbufferSizeInBytes; + newReadOffsetLoopFlag ^= 0x80000000; /* <-- Looped. */ + } else { + newReadOffsetInBytes = (ma_uint32)(readOffsetInBytes + offsetInBytes); + } + } + + ma_atomic_exchange_32(&pRB->encodedReadOffset, ma_rb__construct_offset(newReadOffsetInBytes, newReadOffsetLoopFlag)); + return MA_SUCCESS; +} + +MA_API ma_result ma_rb_seek_write(ma_rb* pRB, size_t offsetInBytes) +{ + ma_uint32 readOffset; + ma_uint32 readOffsetInBytes; + ma_uint32 readOffsetLoopFlag; + ma_uint32 writeOffset; + ma_uint32 writeOffsetInBytes; + ma_uint32 writeOffsetLoopFlag; + ma_uint32 newWriteOffsetInBytes; + ma_uint32 newWriteOffsetLoopFlag; + + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + readOffset = ma_atomic_load_32(&pRB->encodedReadOffset); + ma_rb__deconstruct_offset(readOffset, &readOffsetInBytes, &readOffsetLoopFlag); + + writeOffset = ma_atomic_load_32(&pRB->encodedWriteOffset); + ma_rb__deconstruct_offset(writeOffset, &writeOffsetInBytes, &writeOffsetLoopFlag); + + newWriteOffsetLoopFlag = writeOffsetLoopFlag; + + /* We cannot go past the write buffer. */ + if (readOffsetLoopFlag == writeOffsetLoopFlag) { + /* May end up looping. */ + if ((writeOffsetInBytes + offsetInBytes) >= pRB->subbufferSizeInBytes) { + newWriteOffsetInBytes = (ma_uint32)(writeOffsetInBytes + offsetInBytes) - pRB->subbufferSizeInBytes; + newWriteOffsetLoopFlag ^= 0x80000000; /* <-- Looped. */ + } else { + newWriteOffsetInBytes = (ma_uint32)(writeOffsetInBytes + offsetInBytes); + } + } else { + if ((writeOffsetInBytes + offsetInBytes) > readOffsetInBytes) { + newWriteOffsetInBytes = readOffsetInBytes; + } else { + newWriteOffsetInBytes = (ma_uint32)(writeOffsetInBytes + offsetInBytes); + } + } + + ma_atomic_exchange_32(&pRB->encodedWriteOffset, ma_rb__construct_offset(newWriteOffsetInBytes, newWriteOffsetLoopFlag)); + return MA_SUCCESS; +} + +MA_API ma_int32 ma_rb_pointer_distance(ma_rb* pRB) +{ + ma_uint32 readOffset; + ma_uint32 readOffsetInBytes; + ma_uint32 readOffsetLoopFlag; + ma_uint32 writeOffset; + ma_uint32 writeOffsetInBytes; + ma_uint32 writeOffsetLoopFlag; + + if (pRB == NULL) { + return 0; + } + + readOffset = ma_atomic_load_32(&pRB->encodedReadOffset); + ma_rb__deconstruct_offset(readOffset, &readOffsetInBytes, &readOffsetLoopFlag); + + writeOffset = ma_atomic_load_32(&pRB->encodedWriteOffset); + ma_rb__deconstruct_offset(writeOffset, &writeOffsetInBytes, &writeOffsetLoopFlag); + + if (readOffsetLoopFlag == writeOffsetLoopFlag) { + return writeOffsetInBytes - readOffsetInBytes; + } else { + return writeOffsetInBytes + (pRB->subbufferSizeInBytes - readOffsetInBytes); + } +} + +MA_API ma_uint32 ma_rb_available_read(ma_rb* pRB) +{ + ma_int32 dist; + + if (pRB == NULL) { + return 0; + } + + dist = ma_rb_pointer_distance(pRB); + if (dist < 0) { + return 0; + } + + return dist; +} + +MA_API ma_uint32 ma_rb_available_write(ma_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return (ma_uint32)(ma_rb_get_subbuffer_size(pRB) - ma_rb_pointer_distance(pRB)); +} + +MA_API size_t ma_rb_get_subbuffer_size(ma_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return pRB->subbufferSizeInBytes; +} + +MA_API size_t ma_rb_get_subbuffer_stride(ma_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + if (pRB->subbufferStrideInBytes == 0) { + return (size_t)pRB->subbufferSizeInBytes; + } + + return (size_t)pRB->subbufferStrideInBytes; +} + +MA_API size_t ma_rb_get_subbuffer_offset(ma_rb* pRB, size_t subbufferIndex) +{ + if (pRB == NULL) { + return 0; + } + + return subbufferIndex * ma_rb_get_subbuffer_stride(pRB); +} + +MA_API void* ma_rb_get_subbuffer_ptr(ma_rb* pRB, size_t subbufferIndex, void* pBuffer) +{ + if (pRB == NULL) { + return NULL; + } + + return ma_offset_ptr(pBuffer, ma_rb_get_subbuffer_offset(pRB, subbufferIndex)); +} + + + +static ma_result ma_pcm_rb_data_source__on_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + /* Since there's no notion of an end, we don't ever want to return MA_AT_END here. But it is possible to return 0. */ + ma_pcm_rb* pRB = (ma_pcm_rb*)pDataSource; + ma_result result; + ma_uint64 totalFramesRead; + + MA_ASSERT(pRB != NULL); + + /* We need to run this in a loop since the ring buffer itself may loop. */ + totalFramesRead = 0; + while (totalFramesRead < frameCount) { + void* pMappedBuffer; + ma_uint32 mappedFrameCount; + ma_uint64 framesToRead = frameCount - totalFramesRead; + if (framesToRead > 0xFFFFFFFF) { + framesToRead = 0xFFFFFFFF; + } + + mappedFrameCount = (ma_uint32)framesToRead; + result = ma_pcm_rb_acquire_read(pRB, &mappedFrameCount, &pMappedBuffer); + if (result != MA_SUCCESS) { + break; + } + + if (mappedFrameCount == 0) { + break; /* <-- End of ring buffer. */ + } + + ma_copy_pcm_frames(ma_offset_pcm_frames_ptr(pFramesOut, totalFramesRead, pRB->format, pRB->channels), pMappedBuffer, mappedFrameCount, pRB->format, pRB->channels); + + result = ma_pcm_rb_commit_read(pRB, mappedFrameCount); + if (result != MA_SUCCESS) { + break; + } + + totalFramesRead += mappedFrameCount; + } + + /* + There is no notion of an "end" in a ring buffer. If we didn't have enough data to fill the requested frame + count we'll need to pad with silence. If we don't do this, totalFramesRead might equal 0 which will result + in the data source layer at a higher level translating this to MA_AT_END which is incorrect for a ring buffer. + */ + if (totalFramesRead < frameCount) { + ma_silence_pcm_frames(ma_offset_pcm_frames_ptr(pFramesOut, totalFramesRead, pRB->format, pRB->channels), (frameCount - totalFramesRead), pRB->format, pRB->channels); + totalFramesRead = frameCount; + } + + *pFramesRead = totalFramesRead; + return MA_SUCCESS; +} + +static ma_result ma_pcm_rb_data_source__on_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + ma_pcm_rb* pRB = (ma_pcm_rb*)pDataSource; + MA_ASSERT(pRB != NULL); + + if (pFormat != NULL) { + *pFormat = pRB->format; + } + + if (pChannels != NULL) { + *pChannels = pRB->channels; + } + + if (pSampleRate != NULL) { + *pSampleRate = pRB->sampleRate; + } + + /* Just assume the default channel map. */ + if (pChannelMap != NULL) { + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pRB->channels); + } + + return MA_SUCCESS; +} + +static ma_data_source_vtable ma_gRBDataSourceVTable = +{ + ma_pcm_rb_data_source__on_read, + NULL, /* onSeek */ + ma_pcm_rb_data_source__on_get_data_format, + NULL, /* onGetCursor */ + NULL, /* onGetLength */ + NULL, /* onSetLooping */ + 0 +}; + +static MA_INLINE ma_uint32 ma_pcm_rb_get_bpf(ma_pcm_rb* pRB) +{ + MA_ASSERT(pRB != NULL); + + return ma_get_bytes_per_frame(pRB->format, pRB->channels); +} + +MA_API ma_result ma_pcm_rb_init_ex(ma_format format, ma_uint32 channels, ma_uint32 subbufferSizeInFrames, ma_uint32 subbufferCount, ma_uint32 subbufferStrideInFrames, void* pOptionalPreallocatedBuffer, const ma_allocation_callbacks* pAllocationCallbacks, ma_pcm_rb* pRB) +{ + ma_uint32 bpf; + ma_result result; + + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pRB); + + bpf = ma_get_bytes_per_frame(format, channels); + if (bpf == 0) { + return MA_INVALID_ARGS; + } + + result = ma_rb_init_ex(subbufferSizeInFrames*bpf, subbufferCount, subbufferStrideInFrames*bpf, pOptionalPreallocatedBuffer, pAllocationCallbacks, &pRB->rb); + if (result != MA_SUCCESS) { + return result; + } + + pRB->format = format; + pRB->channels = channels; + pRB->sampleRate = 0; /* The sample rate is not passed in as a parameter. */ + + /* The PCM ring buffer is a data source. We need to get that set up as well. */ + { + ma_data_source_config dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &ma_gRBDataSourceVTable; + + result = ma_data_source_init(&dataSourceConfig, &pRB->ds); + if (result != MA_SUCCESS) { + ma_rb_uninit(&pRB->rb); + return result; + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_pcm_rb_init(ma_format format, ma_uint32 channels, ma_uint32 bufferSizeInFrames, void* pOptionalPreallocatedBuffer, const ma_allocation_callbacks* pAllocationCallbacks, ma_pcm_rb* pRB) +{ + return ma_pcm_rb_init_ex(format, channels, bufferSizeInFrames, 1, 0, pOptionalPreallocatedBuffer, pAllocationCallbacks, pRB); +} + +MA_API void ma_pcm_rb_uninit(ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return; + } + + ma_data_source_uninit(&pRB->ds); + ma_rb_uninit(&pRB->rb); +} + +MA_API void ma_pcm_rb_reset(ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return; + } + + ma_rb_reset(&pRB->rb); +} + +MA_API ma_result ma_pcm_rb_acquire_read(ma_pcm_rb* pRB, ma_uint32* pSizeInFrames, void** ppBufferOut) +{ + size_t sizeInBytes; + ma_result result; + + if (pRB == NULL || pSizeInFrames == NULL) { + return MA_INVALID_ARGS; + } + + sizeInBytes = *pSizeInFrames * ma_pcm_rb_get_bpf(pRB); + + result = ma_rb_acquire_read(&pRB->rb, &sizeInBytes, ppBufferOut); + if (result != MA_SUCCESS) { + return result; + } + + *pSizeInFrames = (ma_uint32)(sizeInBytes / (size_t)ma_pcm_rb_get_bpf(pRB)); + return MA_SUCCESS; +} + +MA_API ma_result ma_pcm_rb_commit_read(ma_pcm_rb* pRB, ma_uint32 sizeInFrames) +{ + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + return ma_rb_commit_read(&pRB->rb, sizeInFrames * ma_pcm_rb_get_bpf(pRB)); +} + +MA_API ma_result ma_pcm_rb_acquire_write(ma_pcm_rb* pRB, ma_uint32* pSizeInFrames, void** ppBufferOut) +{ + size_t sizeInBytes; + ma_result result; + + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + sizeInBytes = *pSizeInFrames * ma_pcm_rb_get_bpf(pRB); + + result = ma_rb_acquire_write(&pRB->rb, &sizeInBytes, ppBufferOut); + if (result != MA_SUCCESS) { + return result; + } + + *pSizeInFrames = (ma_uint32)(sizeInBytes / ma_pcm_rb_get_bpf(pRB)); + return MA_SUCCESS; +} + +MA_API ma_result ma_pcm_rb_commit_write(ma_pcm_rb* pRB, ma_uint32 sizeInFrames) +{ + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + return ma_rb_commit_write(&pRB->rb, sizeInFrames * ma_pcm_rb_get_bpf(pRB)); +} + +MA_API ma_result ma_pcm_rb_seek_read(ma_pcm_rb* pRB, ma_uint32 offsetInFrames) +{ + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + return ma_rb_seek_read(&pRB->rb, offsetInFrames * ma_pcm_rb_get_bpf(pRB)); +} + +MA_API ma_result ma_pcm_rb_seek_write(ma_pcm_rb* pRB, ma_uint32 offsetInFrames) +{ + if (pRB == NULL) { + return MA_INVALID_ARGS; + } + + return ma_rb_seek_write(&pRB->rb, offsetInFrames * ma_pcm_rb_get_bpf(pRB)); +} + +MA_API ma_int32 ma_pcm_rb_pointer_distance(ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return ma_rb_pointer_distance(&pRB->rb) / ma_pcm_rb_get_bpf(pRB); +} + +MA_API ma_uint32 ma_pcm_rb_available_read(ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return ma_rb_available_read(&pRB->rb) / ma_pcm_rb_get_bpf(pRB); +} + +MA_API ma_uint32 ma_pcm_rb_available_write(ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return ma_rb_available_write(&pRB->rb) / ma_pcm_rb_get_bpf(pRB); +} + +MA_API ma_uint32 ma_pcm_rb_get_subbuffer_size(ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return (ma_uint32)(ma_rb_get_subbuffer_size(&pRB->rb) / ma_pcm_rb_get_bpf(pRB)); +} + +MA_API ma_uint32 ma_pcm_rb_get_subbuffer_stride(ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return (ma_uint32)(ma_rb_get_subbuffer_stride(&pRB->rb) / ma_pcm_rb_get_bpf(pRB)); +} + +MA_API ma_uint32 ma_pcm_rb_get_subbuffer_offset(ma_pcm_rb* pRB, ma_uint32 subbufferIndex) +{ + if (pRB == NULL) { + return 0; + } + + return (ma_uint32)(ma_rb_get_subbuffer_offset(&pRB->rb, subbufferIndex) / ma_pcm_rb_get_bpf(pRB)); +} + +MA_API void* ma_pcm_rb_get_subbuffer_ptr(ma_pcm_rb* pRB, ma_uint32 subbufferIndex, void* pBuffer) +{ + if (pRB == NULL) { + return NULL; + } + + return ma_rb_get_subbuffer_ptr(&pRB->rb, subbufferIndex, pBuffer); +} + +MA_API ma_format ma_pcm_rb_get_format(const ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return ma_format_unknown; + } + + return pRB->format; +} + +MA_API ma_uint32 ma_pcm_rb_get_channels(const ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return pRB->channels; +} + +MA_API ma_uint32 ma_pcm_rb_get_sample_rate(const ma_pcm_rb* pRB) +{ + if (pRB == NULL) { + return 0; + } + + return pRB->sampleRate; +} + +MA_API void ma_pcm_rb_set_sample_rate(ma_pcm_rb* pRB, ma_uint32 sampleRate) +{ + if (pRB == NULL) { + return; + } + + pRB->sampleRate = sampleRate; +} + + + +MA_API ma_result ma_duplex_rb_init(ma_format captureFormat, ma_uint32 captureChannels, ma_uint32 sampleRate, ma_uint32 captureInternalSampleRate, ma_uint32 captureInternalPeriodSizeInFrames, const ma_allocation_callbacks* pAllocationCallbacks, ma_duplex_rb* pRB) +{ + ma_result result; + ma_uint32 sizeInFrames; + + sizeInFrames = (ma_uint32)ma_calculate_frame_count_after_resampling(sampleRate, captureInternalSampleRate, captureInternalPeriodSizeInFrames * 5); + if (sizeInFrames == 0) { + return MA_INVALID_ARGS; + } + + result = ma_pcm_rb_init(captureFormat, captureChannels, sizeInFrames, NULL, pAllocationCallbacks, &pRB->rb); + if (result != MA_SUCCESS) { + return result; + } + + /* Seek forward a bit so we have a bit of a buffer in case of desyncs. */ + ma_pcm_rb_seek_write((ma_pcm_rb*)pRB, captureInternalPeriodSizeInFrames * 2); + + return MA_SUCCESS; +} + +MA_API ma_result ma_duplex_rb_uninit(ma_duplex_rb* pRB) +{ + ma_pcm_rb_uninit((ma_pcm_rb*)pRB); + return MA_SUCCESS; +} + + + +/************************************************************************************************************************************************************** + +Miscellaneous Helpers + +**************************************************************************************************************************************************************/ +MA_API const char* ma_result_description(ma_result result) +{ + switch (result) + { + case MA_SUCCESS: return "No error"; + case MA_ERROR: return "Unknown error"; + case MA_INVALID_ARGS: return "Invalid argument"; + case MA_INVALID_OPERATION: return "Invalid operation"; + case MA_OUT_OF_MEMORY: return "Out of memory"; + case MA_OUT_OF_RANGE: return "Out of range"; + case MA_ACCESS_DENIED: return "Permission denied"; + case MA_DOES_NOT_EXIST: return "Resource does not exist"; + case MA_ALREADY_EXISTS: return "Resource already exists"; + case MA_TOO_MANY_OPEN_FILES: return "Too many open files"; + case MA_INVALID_FILE: return "Invalid file"; + case MA_TOO_BIG: return "Too large"; + case MA_PATH_TOO_LONG: return "Path too long"; + case MA_NAME_TOO_LONG: return "Name too long"; + case MA_NOT_DIRECTORY: return "Not a directory"; + case MA_IS_DIRECTORY: return "Is a directory"; + case MA_DIRECTORY_NOT_EMPTY: return "Directory not empty"; + case MA_AT_END: return "At end"; + case MA_NO_SPACE: return "No space available"; + case MA_BUSY: return "Device or resource busy"; + case MA_IO_ERROR: return "Input/output error"; + case MA_INTERRUPT: return "Interrupted"; + case MA_UNAVAILABLE: return "Resource unavailable"; + case MA_ALREADY_IN_USE: return "Resource already in use"; + case MA_BAD_ADDRESS: return "Bad address"; + case MA_BAD_SEEK: return "Illegal seek"; + case MA_BAD_PIPE: return "Broken pipe"; + case MA_DEADLOCK: return "Deadlock"; + case MA_TOO_MANY_LINKS: return "Too many links"; + case MA_NOT_IMPLEMENTED: return "Not implemented"; + case MA_NO_MESSAGE: return "No message of desired type"; + case MA_BAD_MESSAGE: return "Invalid message"; + case MA_NO_DATA_AVAILABLE: return "No data available"; + case MA_INVALID_DATA: return "Invalid data"; + case MA_TIMEOUT: return "Timeout"; + case MA_NO_NETWORK: return "Network unavailable"; + case MA_NOT_UNIQUE: return "Not unique"; + case MA_NOT_SOCKET: return "Socket operation on non-socket"; + case MA_NO_ADDRESS: return "Destination address required"; + case MA_BAD_PROTOCOL: return "Protocol wrong type for socket"; + case MA_PROTOCOL_UNAVAILABLE: return "Protocol not available"; + case MA_PROTOCOL_NOT_SUPPORTED: return "Protocol not supported"; + case MA_PROTOCOL_FAMILY_NOT_SUPPORTED: return "Protocol family not supported"; + case MA_ADDRESS_FAMILY_NOT_SUPPORTED: return "Address family not supported"; + case MA_SOCKET_NOT_SUPPORTED: return "Socket type not supported"; + case MA_CONNECTION_RESET: return "Connection reset"; + case MA_ALREADY_CONNECTED: return "Already connected"; + case MA_NOT_CONNECTED: return "Not connected"; + case MA_CONNECTION_REFUSED: return "Connection refused"; + case MA_NO_HOST: return "No host"; + case MA_IN_PROGRESS: return "Operation in progress"; + case MA_CANCELLED: return "Operation cancelled"; + case MA_MEMORY_ALREADY_MAPPED: return "Memory already mapped"; + + case MA_FORMAT_NOT_SUPPORTED: return "Format not supported"; + case MA_DEVICE_TYPE_NOT_SUPPORTED: return "Device type not supported"; + case MA_SHARE_MODE_NOT_SUPPORTED: return "Share mode not supported"; + case MA_NO_BACKEND: return "No backend"; + case MA_NO_DEVICE: return "No device"; + case MA_API_NOT_FOUND: return "API not found"; + case MA_INVALID_DEVICE_CONFIG: return "Invalid device config"; + + case MA_DEVICE_NOT_INITIALIZED: return "Device not initialized"; + case MA_DEVICE_NOT_STARTED: return "Device not started"; + + case MA_FAILED_TO_INIT_BACKEND: return "Failed to initialize backend"; + case MA_FAILED_TO_OPEN_BACKEND_DEVICE: return "Failed to open backend device"; + case MA_FAILED_TO_START_BACKEND_DEVICE: return "Failed to start backend device"; + case MA_FAILED_TO_STOP_BACKEND_DEVICE: return "Failed to stop backend device"; + + default: return "Unknown error"; + } +} + +MA_API void* ma_malloc(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks != NULL) { + if (pAllocationCallbacks->onMalloc != NULL) { + return pAllocationCallbacks->onMalloc(sz, pAllocationCallbacks->pUserData); + } else { + return NULL; /* Do not fall back to the default implementation. */ + } + } else { + return ma__malloc_default(sz, NULL); + } +} + +MA_API void* ma_calloc(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks) +{ + void* p = ma_malloc(sz, pAllocationCallbacks); + if (p != NULL) { + MA_ZERO_MEMORY(p, sz); + } + + return p; +} + +MA_API void* ma_realloc(void* p, size_t sz, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks != NULL) { + if (pAllocationCallbacks->onRealloc != NULL) { + return pAllocationCallbacks->onRealloc(p, sz, pAllocationCallbacks->pUserData); + } else { + return NULL; /* Do not fall back to the default implementation. */ + } + } else { + return ma__realloc_default(p, sz, NULL); + } +} + +MA_API void ma_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (p == NULL) { + return; + } + + if (pAllocationCallbacks != NULL) { + if (pAllocationCallbacks->onFree != NULL) { + pAllocationCallbacks->onFree(p, pAllocationCallbacks->pUserData); + } else { + return; /* Do no fall back to the default implementation. */ + } + } else { + ma__free_default(p, NULL); + } +} + +MA_API void* ma_aligned_malloc(size_t sz, size_t alignment, const ma_allocation_callbacks* pAllocationCallbacks) +{ + size_t extraBytes; + void* pUnaligned; + void* pAligned; + + if (alignment == 0) { + return 0; + } + + extraBytes = alignment-1 + sizeof(void*); + + pUnaligned = ma_malloc(sz + extraBytes, pAllocationCallbacks); + if (pUnaligned == NULL) { + return NULL; + } + + pAligned = (void*)(((ma_uintptr)pUnaligned + extraBytes) & ~((ma_uintptr)(alignment-1))); + ((void**)pAligned)[-1] = pUnaligned; + + return pAligned; +} + +MA_API void ma_aligned_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_free(((void**)p)[-1], pAllocationCallbacks); +} + +MA_API const char* ma_get_format_name(ma_format format) +{ + switch (format) + { + case ma_format_unknown: return "Unknown"; + case ma_format_u8: return "8-bit Unsigned Integer"; + case ma_format_s16: return "16-bit Signed Integer"; + case ma_format_s24: return "24-bit Signed Integer (Tightly Packed)"; + case ma_format_s32: return "32-bit Signed Integer"; + case ma_format_f32: return "32-bit IEEE Floating Point"; + default: return "Invalid"; + } +} + +MA_API void ma_blend_f32(float* pOut, float* pInA, float* pInB, float factor, ma_uint32 channels) +{ + ma_uint32 i; + for (i = 0; i < channels; ++i) { + pOut[i] = ma_mix_f32(pInA[i], pInB[i], factor); + } +} + + +MA_API ma_uint32 ma_get_bytes_per_sample(ma_format format) +{ + ma_uint32 sizes[] = { + 0, /* unknown */ + 1, /* u8 */ + 2, /* s16 */ + 3, /* s24 */ + 4, /* s32 */ + 4, /* f32 */ + }; + return sizes[format]; +} + + + +#define MA_DATA_SOURCE_DEFAULT_RANGE_BEG 0 +#define MA_DATA_SOURCE_DEFAULT_RANGE_END ~((ma_uint64)0) +#define MA_DATA_SOURCE_DEFAULT_LOOP_POINT_BEG 0 +#define MA_DATA_SOURCE_DEFAULT_LOOP_POINT_END ~((ma_uint64)0) + +MA_API ma_data_source_config ma_data_source_config_init(void) +{ + ma_data_source_config config; + + MA_ZERO_OBJECT(&config); + + return config; +} + + +MA_API ma_result ma_data_source_init(const ma_data_source_config* pConfig, ma_data_source* pDataSource) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDataSourceBase); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->vtable == NULL) { + return MA_INVALID_ARGS; + } + + pDataSourceBase->vtable = pConfig->vtable; + pDataSourceBase->rangeBegInFrames = MA_DATA_SOURCE_DEFAULT_RANGE_BEG; + pDataSourceBase->rangeEndInFrames = MA_DATA_SOURCE_DEFAULT_RANGE_END; + pDataSourceBase->loopBegInFrames = MA_DATA_SOURCE_DEFAULT_LOOP_POINT_BEG; + pDataSourceBase->loopEndInFrames = MA_DATA_SOURCE_DEFAULT_LOOP_POINT_END; + pDataSourceBase->pCurrent = pDataSource; /* Always read from ourself by default. */ + pDataSourceBase->pNext = NULL; + pDataSourceBase->onGetNext = NULL; + + return MA_SUCCESS; +} + +MA_API void ma_data_source_uninit(ma_data_source* pDataSource) +{ + if (pDataSource == NULL) { + return; + } + + /* + This is placeholder in case we need this later. Data sources need to call this in their + uninitialization routine to ensure things work later on if something is added here. + */ +} + +static ma_result ma_data_source_resolve_current(ma_data_source* pDataSource, ma_data_source** ppCurrentDataSource) +{ + ma_data_source_base* pCurrentDataSource = (ma_data_source_base*)pDataSource; + + MA_ASSERT(pDataSource != NULL); + MA_ASSERT(ppCurrentDataSource != NULL); + + if (pCurrentDataSource->pCurrent == NULL) { + /* + The current data source is NULL. If we're using this in the context of a chain we need to return NULL + here so that we don't end up looping. Otherwise we just return the data source itself. + */ + if (pCurrentDataSource->pNext != NULL || pCurrentDataSource->onGetNext != NULL) { + pCurrentDataSource = NULL; + } else { + pCurrentDataSource = (ma_data_source_base*)pDataSource; /* Not being used in a chain. Make sure we just always read from the data source itself at all times. */ + } + } else { + pCurrentDataSource = (ma_data_source_base*)pCurrentDataSource->pCurrent; + } + + *ppCurrentDataSource = pCurrentDataSource; + + return MA_SUCCESS; +} + +static ma_result ma_data_source_read_pcm_frames_from_backend(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + MA_ASSERT(pDataSourceBase != NULL); + MA_ASSERT(pDataSourceBase->vtable != NULL); + MA_ASSERT(pDataSourceBase->vtable->onRead != NULL); + MA_ASSERT(pFramesRead != NULL); + + if (pFramesOut != NULL) { + return pDataSourceBase->vtable->onRead(pDataSourceBase, pFramesOut, frameCount, pFramesRead); + } else { + /* + No output buffer. Probably seeking forward. Read and discard. Can probably optimize this in terms of + onSeek and onGetCursor, but need to keep in mind that the data source may not implement these functions. + */ + ma_result result; + ma_uint64 framesRead; + ma_format format; + ma_uint32 channels; + ma_uint64 discardBufferCapInFrames; + ma_uint8 pDiscardBuffer[4096]; + + result = ma_data_source_get_data_format(pDataSource, &format, &channels, NULL, NULL, 0); + if (result != MA_SUCCESS) { + return result; + } + + discardBufferCapInFrames = sizeof(pDiscardBuffer) / ma_get_bytes_per_frame(format, channels); + + framesRead = 0; + while (framesRead < frameCount) { + ma_uint64 framesReadThisIteration = 0; + ma_uint64 framesToRead = frameCount - framesRead; + if (framesToRead > discardBufferCapInFrames) { + framesToRead = discardBufferCapInFrames; + } + + result = pDataSourceBase->vtable->onRead(pDataSourceBase, pDiscardBuffer, framesToRead, &framesReadThisIteration); + if (result != MA_SUCCESS) { + return result; + } + + framesRead += framesReadThisIteration; + } + + *pFramesRead = framesRead; + + return MA_SUCCESS; + } +} + +static ma_result ma_data_source_read_pcm_frames_within_range(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + ma_result result; + ma_uint64 framesRead = 0; + ma_bool32 loop = ma_data_source_is_looping(pDataSource); + + if (pDataSourceBase == NULL) { + return MA_AT_END; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + MA_ASSERT(pDataSourceBase->vtable != NULL); + + if ((pDataSourceBase->vtable->flags & MA_DATA_SOURCE_SELF_MANAGED_RANGE_AND_LOOP_POINT) != 0 || (pDataSourceBase->rangeEndInFrames == ~((ma_uint64)0) && (pDataSourceBase->loopEndInFrames == ~((ma_uint64)0) || loop == MA_FALSE))) { + /* Either the data source is self-managing the range, or no range is set - just read like normal. The data source itself will tell us when the end is reached. */ + result = ma_data_source_read_pcm_frames_from_backend(pDataSource, pFramesOut, frameCount, &framesRead); + } else { + /* Need to clamp to within the range. */ + ma_uint64 relativeCursor; + ma_uint64 absoluteCursor; + + result = ma_data_source_get_cursor_in_pcm_frames(pDataSourceBase, &relativeCursor); + if (result != MA_SUCCESS) { + /* Failed to retrieve the cursor. Cannot read within a range or loop points. Just read like normal - this may happen for things like noise data sources where it doesn't really matter. */ + result = ma_data_source_read_pcm_frames_from_backend(pDataSource, pFramesOut, frameCount, &framesRead); + } else { + ma_uint64 rangeBeg; + ma_uint64 rangeEnd; + + /* We have the cursor. We need to make sure we don't read beyond our range. */ + rangeBeg = pDataSourceBase->rangeBegInFrames; + rangeEnd = pDataSourceBase->rangeEndInFrames; + + absoluteCursor = rangeBeg + relativeCursor; + + /* If looping, make sure we're within range. */ + if (loop) { + if (pDataSourceBase->loopEndInFrames != ~((ma_uint64)0)) { + rangeEnd = ma_min(rangeEnd, pDataSourceBase->rangeBegInFrames + pDataSourceBase->loopEndInFrames); + } + } + + if (frameCount > (rangeEnd - absoluteCursor) && rangeEnd != ~((ma_uint64)0)) { + frameCount = (rangeEnd - absoluteCursor); + } + + /* + If the cursor is sitting on the end of the range the frame count will be set to 0 which can + result in MA_INVALID_ARGS. In this case, we don't want to try reading, but instead return + MA_AT_END so the higher level function can know about it. + */ + if (frameCount > 0) { + result = ma_data_source_read_pcm_frames_from_backend(pDataSource, pFramesOut, frameCount, &framesRead); + } else { + result = MA_AT_END; /* The cursor is sitting on the end of the range which means we're at the end. */ + } + } + } + + if (pFramesRead != NULL) { + *pFramesRead = framesRead; + } + + /* We need to make sure MA_AT_END is returned if we hit the end of the range. */ + if (result == MA_SUCCESS && framesRead == 0) { + result = MA_AT_END; + } + + return result; +} + +MA_API ma_result ma_data_source_read_pcm_frames(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_result result = MA_SUCCESS; + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + ma_data_source_base* pCurrentDataSource; + void* pRunningFramesOut = pFramesOut; + ma_uint64 totalFramesProcessed = 0; + ma_format format; + ma_uint32 channels; + ma_uint32 emptyLoopCounter = 0; /* Keeps track of how many times 0 frames have been read. For infinite loop detection of sounds with no audio data. */ + ma_bool32 loop; + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pDataSourceBase == NULL) { + return MA_INVALID_ARGS; + } + + loop = ma_data_source_is_looping(pDataSource); + + /* + We need to know the data format so we can advance the output buffer as we read frames. If this + fails, chaining will not work and we'll just read as much as we can from the current source. + */ + if (ma_data_source_get_data_format(pDataSource, &format, &channels, NULL, NULL, 0) != MA_SUCCESS) { + result = ma_data_source_resolve_current(pDataSource, (ma_data_source**)&pCurrentDataSource); + if (result != MA_SUCCESS) { + return result; + } + + return ma_data_source_read_pcm_frames_within_range(pCurrentDataSource, pFramesOut, frameCount, pFramesRead); + } + + /* + Looping is a bit of a special case. When the `loop` argument is true, chaining will not work and + only the current data source will be read from. + */ + + /* Keep reading until we've read as many frames as possible. */ + while (totalFramesProcessed < frameCount) { + ma_uint64 framesProcessed; + ma_uint64 framesRemaining = frameCount - totalFramesProcessed; + + /* We need to resolve the data source that we'll actually be reading from. */ + result = ma_data_source_resolve_current(pDataSource, (ma_data_source**)&pCurrentDataSource); + if (result != MA_SUCCESS) { + break; + } + + if (pCurrentDataSource == NULL) { + break; + } + + result = ma_data_source_read_pcm_frames_within_range(pCurrentDataSource, pRunningFramesOut, framesRemaining, &framesProcessed); + totalFramesProcessed += framesProcessed; + + /* + If we encountered an error from the read callback, make sure it's propagated to the caller. The caller may need to know whether or not MA_BUSY is returned which is + not necessarily considered an error. + */ + if (result != MA_SUCCESS && result != MA_AT_END) { + break; + } + + /* + We can determine if we've reached the end by checking if ma_data_source_read_pcm_frames_within_range() returned + MA_AT_END. To loop back to the start, all we need to do is seek back to the first frame. + */ + if (result == MA_AT_END) { + /* + The result needs to be reset back to MA_SUCCESS (from MA_AT_END) so that we don't + accidentally return MA_AT_END when data has been read in prior loop iterations. at the + end of this function, the result will be checked for MA_SUCCESS, and if the total + number of frames processed is 0, will be explicitly set to MA_AT_END. + */ + result = MA_SUCCESS; + + /* + We reached the end. If we're looping, we just loop back to the start of the current + data source. If we're not looping we need to check if we have another in the chain, and + if so, switch to it. + */ + if (loop) { + if (framesProcessed == 0) { + emptyLoopCounter += 1; + if (emptyLoopCounter > 1) { + break; /* Infinite loop detected. Get out. */ + } + } else { + emptyLoopCounter = 0; + } + + result = ma_data_source_seek_to_pcm_frame(pCurrentDataSource, pCurrentDataSource->loopBegInFrames); + if (result != MA_SUCCESS) { + break; /* Failed to loop. Abort. */ + } + + /* Don't return MA_AT_END for looping sounds. */ + result = MA_SUCCESS; + } else { + if (pCurrentDataSource->pNext != NULL) { + pDataSourceBase->pCurrent = pCurrentDataSource->pNext; + } else if (pCurrentDataSource->onGetNext != NULL) { + pDataSourceBase->pCurrent = pCurrentDataSource->onGetNext(pCurrentDataSource); + if (pDataSourceBase->pCurrent == NULL) { + break; /* Our callback did not return a next data source. We're done. */ + } + } else { + /* Reached the end of the chain. We're done. */ + break; + } + + /* The next data source needs to be rewound to ensure data is read in looping scenarios. */ + result = ma_data_source_seek_to_pcm_frame(pDataSourceBase->pCurrent, 0); + if (result != MA_SUCCESS) { + break; + } + } + } + + if (pRunningFramesOut != NULL) { + pRunningFramesOut = ma_offset_ptr(pRunningFramesOut, framesProcessed * ma_get_bytes_per_frame(format, channels)); + } + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesProcessed; + } + + MA_ASSERT(!(result == MA_AT_END && totalFramesProcessed > 0)); /* We should never be returning MA_AT_END if we read some data. */ + + if (result == MA_SUCCESS && totalFramesProcessed == 0) { + result = MA_AT_END; + } + + return result; +} + +MA_API ma_result ma_data_source_seek_pcm_frames(ma_data_source* pDataSource, ma_uint64 frameCount, ma_uint64* pFramesSeeked) +{ + return ma_data_source_read_pcm_frames(pDataSource, NULL, frameCount, pFramesSeeked); +} + +MA_API ma_result ma_data_source_seek_to_pcm_frame(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + if (pDataSourceBase == NULL) { + return MA_INVALID_ARGS; + } + + if (pDataSourceBase->vtable->onSeek == NULL) { + return MA_NOT_IMPLEMENTED; + } + + if (frameIndex > pDataSourceBase->rangeEndInFrames) { + return MA_INVALID_OPERATION; /* Trying to seek too far forward. */ + } + + MA_ASSERT(pDataSourceBase->vtable != NULL); + + return pDataSourceBase->vtable->onSeek(pDataSource, pDataSourceBase->rangeBegInFrames + frameIndex); +} + +MA_API ma_result ma_data_source_seek_seconds(ma_data_source* pDataSource, float secondCount, float* pSecondsSeeked) +{ + ma_uint64 frameCount; + ma_uint64 framesSeeked = 0; + ma_uint32 sampleRate; + ma_result result; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_data_source_get_data_format(pDataSource, NULL, NULL, &sampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; + } + + /* We need PCM frames instead of seconds */ + frameCount = (ma_uint64)(secondCount * sampleRate); + + result = ma_data_source_seek_pcm_frames(pDataSource, frameCount, &framesSeeked); + + /* VC6 doesn't support division between unsigned 64-bit integer and floating point number. Signed integer needed. This shouldn't affect anything in practice */ + *pSecondsSeeked = (ma_int64)framesSeeked / (float)sampleRate; + return result; +} + +MA_API ma_result ma_data_source_seek_to_second(ma_data_source* pDataSource, float seekPointInSeconds) +{ + ma_uint64 frameIndex; + ma_uint32 sampleRate; + ma_result result; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_data_source_get_data_format(pDataSource, NULL, NULL, &sampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; + } + + /* We need PCM frames instead of seconds */ + frameIndex = (ma_uint64)(seekPointInSeconds * sampleRate); + + return ma_data_source_seek_to_pcm_frame(pDataSource, frameIndex); +} + +MA_API ma_result ma_data_source_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + ma_result result; + ma_format format; + ma_uint32 channels; + ma_uint32 sampleRate; + + /* Initialize to defaults for safety just in case the data source does not implement this callback. */ + if (pFormat != NULL) { + *pFormat = ma_format_unknown; + } + if (pChannels != NULL) { + *pChannels = 0; + } + if (pSampleRate != NULL) { + *pSampleRate = 0; + } + if (pChannelMap != NULL) { + MA_ZERO_MEMORY(pChannelMap, sizeof(*pChannelMap) * channelMapCap); + } + + if (pDataSourceBase == NULL) { + return MA_INVALID_ARGS; + } + + MA_ASSERT(pDataSourceBase->vtable != NULL); + + if (pDataSourceBase->vtable->onGetDataFormat == NULL) { + return MA_NOT_IMPLEMENTED; + } + + result = pDataSourceBase->vtable->onGetDataFormat(pDataSource, &format, &channels, &sampleRate, pChannelMap, channelMapCap); + if (result != MA_SUCCESS) { + return result; + } + + if (pFormat != NULL) { + *pFormat = format; + } + if (pChannels != NULL) { + *pChannels = channels; + } + if (pSampleRate != NULL) { + *pSampleRate = sampleRate; + } + + /* Channel map was passed in directly to the callback. This is safe due to the channelMapCap parameter. */ + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_source_get_cursor_in_pcm_frames(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + ma_result result; + ma_uint64 cursor; + + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; + + if (pDataSourceBase == NULL) { + return MA_SUCCESS; + } + + MA_ASSERT(pDataSourceBase->vtable != NULL); + + if (pDataSourceBase->vtable->onGetCursor == NULL) { + return MA_NOT_IMPLEMENTED; + } + + result = pDataSourceBase->vtable->onGetCursor(pDataSourceBase, &cursor); + if (result != MA_SUCCESS) { + return result; + } + + /* The cursor needs to be made relative to the start of the range. */ + if (cursor < pDataSourceBase->rangeBegInFrames) { /* Safety check so we don't return some huge number. */ + *pCursor = 0; + } else { + *pCursor = cursor - pDataSourceBase->rangeBegInFrames; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_source_get_length_in_pcm_frames(ma_data_source* pDataSource, ma_uint64* pLength) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; + + if (pDataSourceBase == NULL) { + return MA_INVALID_ARGS; + } + + MA_ASSERT(pDataSourceBase->vtable != NULL); + + /* + If we have a range defined we'll use that to determine the length. This is one of rare times + where we'll actually trust the caller. If they've set the range, I think it's mostly safe to + assume they've set it based on some higher level knowledge of the structure of the sound bank. + */ + if (pDataSourceBase->rangeEndInFrames != ~((ma_uint64)0)) { + *pLength = pDataSourceBase->rangeEndInFrames - pDataSourceBase->rangeBegInFrames; + return MA_SUCCESS; + } + + /* + Getting here means a range is not defined so we'll need to get the data source itself to tell + us the length. + */ + if (pDataSourceBase->vtable->onGetLength == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pDataSourceBase->vtable->onGetLength(pDataSource, pLength); +} + +MA_API ma_result ma_data_source_get_cursor_in_seconds(ma_data_source* pDataSource, float* pCursor) +{ + ma_result result; + ma_uint64 cursorInPCMFrames; + ma_uint32 sampleRate; + + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; + + result = ma_data_source_get_cursor_in_pcm_frames(pDataSource, &cursorInPCMFrames); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_data_source_get_data_format(pDataSource, NULL, NULL, &sampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; + } + + /* VC6 does not support division of unsigned 64-bit integers with floating point numbers. Need to use a signed number. This shouldn't effect anything in practice. */ + *pCursor = (ma_int64)cursorInPCMFrames / (float)sampleRate; + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_source_get_length_in_seconds(ma_data_source* pDataSource, float* pLength) +{ + ma_result result; + ma_uint64 lengthInPCMFrames; + ma_uint32 sampleRate; + + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; + + result = ma_data_source_get_length_in_pcm_frames(pDataSource, &lengthInPCMFrames); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_data_source_get_data_format(pDataSource, NULL, NULL, &sampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; + } + + /* VC6 does not support division of unsigned 64-bit integers with floating point numbers. Need to use a signed number. This shouldn't effect anything in practice. */ + *pLength = (ma_int64)lengthInPCMFrames / (float)sampleRate; + + return MA_SUCCESS; +} + +MA_API ma_result ma_data_source_set_looping(ma_data_source* pDataSource, ma_bool32 isLooping) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + ma_atomic_exchange_32(&pDataSourceBase->isLooping, isLooping); + + MA_ASSERT(pDataSourceBase->vtable != NULL); + + /* If there's no callback for this just treat it as a successful no-op. */ + if (pDataSourceBase->vtable->onSetLooping == NULL) { + return MA_SUCCESS; + } + + return pDataSourceBase->vtable->onSetLooping(pDataSource, isLooping); +} + +MA_API ma_bool32 ma_data_source_is_looping(const ma_data_source* pDataSource) +{ + const ma_data_source_base* pDataSourceBase = (const ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return MA_FALSE; + } + + return ma_atomic_load_32(&pDataSourceBase->isLooping); +} + +MA_API ma_result ma_data_source_set_range_in_pcm_frames(ma_data_source* pDataSource, ma_uint64 rangeBegInFrames, ma_uint64 rangeEndInFrames) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + ma_result result; + ma_uint64 relativeCursor; + ma_uint64 absoluteCursor; + ma_bool32 doSeekAdjustment = MA_FALSE; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if (rangeEndInFrames < rangeBegInFrames) { + return MA_INVALID_ARGS; /* The end of the range must come after the beginning. */ + } + + /* + We may need to adjust the position of the cursor to ensure it's clamped to the range. Grab it now + so we can calculate its absolute position before we change the range. + */ + result = ma_data_source_get_cursor_in_pcm_frames(pDataSource, &relativeCursor); + if (result == MA_SUCCESS) { + doSeekAdjustment = MA_TRUE; + absoluteCursor = relativeCursor + pDataSourceBase->rangeBegInFrames; + } else { + /* + We couldn't get the position of the cursor. It probably means the data source has no notion + of a cursor. We'll just leave it at position 0. Don't treat this as an error. + */ + doSeekAdjustment = MA_FALSE; + relativeCursor = 0; + absoluteCursor = 0; + } + + pDataSourceBase->rangeBegInFrames = rangeBegInFrames; + pDataSourceBase->rangeEndInFrames = rangeEndInFrames; + + /* + The commented out logic below was intended to maintain loop points in response to a change in the + range. However, this is not useful because it results in the sound breaking when you move the range + outside of the old loop points. I'm simplifying this by simply resetting the loop points. The + caller is expected to update their loop points if they change the range. + + In practice this should be mostly a non-issue because the majority of the time the range will be + set once right after initialization. + */ + pDataSourceBase->loopBegInFrames = 0; + pDataSourceBase->loopEndInFrames = ~((ma_uint64)0); + + + /* + Seek to within range. Note that our seek positions here are relative to the new range. We don't want + to do this if we failed to retrieve the cursor earlier on because it probably means the data source + has no notion of a cursor. In practice the seek would probably fail (which we silently ignore), but + I'm just not even going to attempt it. + */ + if (doSeekAdjustment) { + if (absoluteCursor < rangeBegInFrames) { + ma_data_source_seek_to_pcm_frame(pDataSource, 0); + } else if (absoluteCursor > rangeEndInFrames) { + ma_data_source_seek_to_pcm_frame(pDataSource, rangeEndInFrames - rangeBegInFrames); + } + } + + return MA_SUCCESS; +} + +MA_API void ma_data_source_get_range_in_pcm_frames(const ma_data_source* pDataSource, ma_uint64* pRangeBegInFrames, ma_uint64* pRangeEndInFrames) +{ + const ma_data_source_base* pDataSourceBase = (const ma_data_source_base*)pDataSource; + + if (pRangeBegInFrames != NULL) { + *pRangeBegInFrames = 0; + } + if (pRangeEndInFrames != NULL) { + *pRangeEndInFrames = 0; + } + + if (pDataSource == NULL) { + return; + } + + if (pRangeBegInFrames != NULL) { + *pRangeBegInFrames = pDataSourceBase->rangeBegInFrames; + } + + if (pRangeEndInFrames != NULL) { + *pRangeEndInFrames = pDataSourceBase->rangeEndInFrames; + } +} + +MA_API ma_result ma_data_source_set_loop_point_in_pcm_frames(ma_data_source* pDataSource, ma_uint64 loopBegInFrames, ma_uint64 loopEndInFrames) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if (loopEndInFrames < loopBegInFrames) { + return MA_INVALID_ARGS; /* The end of the loop point must come after the beginning. */ + } + + if (loopEndInFrames > pDataSourceBase->rangeEndInFrames && loopEndInFrames != ~((ma_uint64)0)) { + return MA_INVALID_ARGS; /* The end of the loop point must not go beyond the range. */ + } + + pDataSourceBase->loopBegInFrames = loopBegInFrames; + pDataSourceBase->loopEndInFrames = loopEndInFrames; + + /* The end cannot exceed the range. */ + if (pDataSourceBase->loopEndInFrames > (pDataSourceBase->rangeEndInFrames - pDataSourceBase->rangeBegInFrames) && pDataSourceBase->loopEndInFrames != ~((ma_uint64)0)) { + pDataSourceBase->loopEndInFrames = (pDataSourceBase->rangeEndInFrames - pDataSourceBase->rangeBegInFrames); + } + + return MA_SUCCESS; +} + +MA_API void ma_data_source_get_loop_point_in_pcm_frames(const ma_data_source* pDataSource, ma_uint64* pLoopBegInFrames, ma_uint64* pLoopEndInFrames) +{ + const ma_data_source_base* pDataSourceBase = (const ma_data_source_base*)pDataSource; + + if (pLoopBegInFrames != NULL) { + *pLoopBegInFrames = 0; + } + if (pLoopEndInFrames != NULL) { + *pLoopEndInFrames = 0; + } + + if (pDataSource == NULL) { + return; + } + + if (pLoopBegInFrames != NULL) { + *pLoopBegInFrames = pDataSourceBase->loopBegInFrames; + } + + if (pLoopEndInFrames != NULL) { + *pLoopEndInFrames = pDataSourceBase->loopEndInFrames; + } +} + +MA_API ma_result ma_data_source_set_current(ma_data_source* pDataSource, ma_data_source* pCurrentDataSource) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + pDataSourceBase->pCurrent = pCurrentDataSource; + + return MA_SUCCESS; +} + +MA_API ma_data_source* ma_data_source_get_current(const ma_data_source* pDataSource) +{ + const ma_data_source_base* pDataSourceBase = (const ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return NULL; + } + + return pDataSourceBase->pCurrent; +} + +MA_API ma_result ma_data_source_set_next(ma_data_source* pDataSource, ma_data_source* pNextDataSource) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + pDataSourceBase->pNext = pNextDataSource; + + return MA_SUCCESS; +} + +MA_API ma_data_source* ma_data_source_get_next(const ma_data_source* pDataSource) +{ + const ma_data_source_base* pDataSourceBase = (const ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return NULL; + } + + return pDataSourceBase->pNext; +} + +MA_API ma_result ma_data_source_set_next_callback(ma_data_source* pDataSource, ma_data_source_get_next_proc onGetNext) +{ + ma_data_source_base* pDataSourceBase = (ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + pDataSourceBase->onGetNext = onGetNext; + + return MA_SUCCESS; +} + +MA_API ma_data_source_get_next_proc ma_data_source_get_next_callback(const ma_data_source* pDataSource) +{ + const ma_data_source_base* pDataSourceBase = (const ma_data_source_base*)pDataSource; + + if (pDataSource == NULL) { + return NULL; + } + + return pDataSourceBase->onGetNext; +} + + +static ma_result ma_audio_buffer_ref__data_source_on_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_audio_buffer_ref* pAudioBufferRef = (ma_audio_buffer_ref*)pDataSource; + ma_uint64 framesRead = ma_audio_buffer_ref_read_pcm_frames(pAudioBufferRef, pFramesOut, frameCount, MA_FALSE); + + if (pFramesRead != NULL) { + *pFramesRead = framesRead; + } + + if (framesRead < frameCount || framesRead == 0) { + return MA_AT_END; + } + + return MA_SUCCESS; +} + +static ma_result ma_audio_buffer_ref__data_source_on_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_audio_buffer_ref_seek_to_pcm_frame((ma_audio_buffer_ref*)pDataSource, frameIndex); +} + +static ma_result ma_audio_buffer_ref__data_source_on_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + ma_audio_buffer_ref* pAudioBufferRef = (ma_audio_buffer_ref*)pDataSource; + + *pFormat = pAudioBufferRef->format; + *pChannels = pAudioBufferRef->channels; + *pSampleRate = pAudioBufferRef->sampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pAudioBufferRef->channels); + + return MA_SUCCESS; +} + +static ma_result ma_audio_buffer_ref__data_source_on_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + ma_audio_buffer_ref* pAudioBufferRef = (ma_audio_buffer_ref*)pDataSource; + + *pCursor = pAudioBufferRef->cursor; + + return MA_SUCCESS; +} + +static ma_result ma_audio_buffer_ref__data_source_on_get_length(ma_data_source* pDataSource, ma_uint64* pLength) +{ + ma_audio_buffer_ref* pAudioBufferRef = (ma_audio_buffer_ref*)pDataSource; + + *pLength = pAudioBufferRef->sizeInFrames; + + return MA_SUCCESS; +} + +static ma_data_source_vtable g_ma_audio_buffer_ref_data_source_vtable = +{ + ma_audio_buffer_ref__data_source_on_read, + ma_audio_buffer_ref__data_source_on_seek, + ma_audio_buffer_ref__data_source_on_get_data_format, + ma_audio_buffer_ref__data_source_on_get_cursor, + ma_audio_buffer_ref__data_source_on_get_length, + NULL, /* onSetLooping */ + 0 +}; + +MA_API ma_result ma_audio_buffer_ref_init(ma_format format, ma_uint32 channels, const void* pData, ma_uint64 sizeInFrames, ma_audio_buffer_ref* pAudioBufferRef) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + + if (pAudioBufferRef == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pAudioBufferRef); + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_audio_buffer_ref_data_source_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pAudioBufferRef->ds); + if (result != MA_SUCCESS) { + return result; + } + + pAudioBufferRef->format = format; + pAudioBufferRef->channels = channels; + pAudioBufferRef->sampleRate = 0; /* TODO: Version 0.12. Set this to sampleRate. */ + pAudioBufferRef->cursor = 0; + pAudioBufferRef->sizeInFrames = sizeInFrames; + pAudioBufferRef->pData = pData; + + return MA_SUCCESS; +} + +MA_API void ma_audio_buffer_ref_uninit(ma_audio_buffer_ref* pAudioBufferRef) +{ + if (pAudioBufferRef == NULL) { + return; + } + + ma_data_source_uninit(&pAudioBufferRef->ds); +} + +MA_API ma_result ma_audio_buffer_ref_set_data(ma_audio_buffer_ref* pAudioBufferRef, const void* pData, ma_uint64 sizeInFrames) +{ + if (pAudioBufferRef == NULL) { + return MA_INVALID_ARGS; + } + + pAudioBufferRef->cursor = 0; + pAudioBufferRef->sizeInFrames = sizeInFrames; + pAudioBufferRef->pData = pData; + + return MA_SUCCESS; +} + +MA_API ma_uint64 ma_audio_buffer_ref_read_pcm_frames(ma_audio_buffer_ref* pAudioBufferRef, void* pFramesOut, ma_uint64 frameCount, ma_bool32 loop) +{ + ma_uint64 totalFramesRead = 0; + + if (pAudioBufferRef == NULL) { + return 0; + } + + if (frameCount == 0) { + return 0; + } + + while (totalFramesRead < frameCount) { + ma_uint64 framesAvailable = pAudioBufferRef->sizeInFrames - pAudioBufferRef->cursor; + ma_uint64 framesRemaining = frameCount - totalFramesRead; + ma_uint64 framesToRead; + + framesToRead = framesRemaining; + if (framesToRead > framesAvailable) { + framesToRead = framesAvailable; + } + + if (pFramesOut != NULL) { + ma_copy_pcm_frames(ma_offset_ptr(pFramesOut, totalFramesRead * ma_get_bytes_per_frame(pAudioBufferRef->format, pAudioBufferRef->channels)), ma_offset_ptr(pAudioBufferRef->pData, pAudioBufferRef->cursor * ma_get_bytes_per_frame(pAudioBufferRef->format, pAudioBufferRef->channels)), framesToRead, pAudioBufferRef->format, pAudioBufferRef->channels); + } + + totalFramesRead += framesToRead; + + pAudioBufferRef->cursor += framesToRead; + if (pAudioBufferRef->cursor == pAudioBufferRef->sizeInFrames) { + if (loop) { + pAudioBufferRef->cursor = 0; + } else { + break; /* We've reached the end and we're not looping. Done. */ + } + } + + MA_ASSERT(pAudioBufferRef->cursor < pAudioBufferRef->sizeInFrames); + } + + return totalFramesRead; +} + +MA_API ma_result ma_audio_buffer_ref_seek_to_pcm_frame(ma_audio_buffer_ref* pAudioBufferRef, ma_uint64 frameIndex) +{ + if (pAudioBufferRef == NULL) { + return MA_INVALID_ARGS; + } + + if (frameIndex > pAudioBufferRef->sizeInFrames) { + return MA_INVALID_ARGS; + } + + pAudioBufferRef->cursor = (size_t)frameIndex; + + return MA_SUCCESS; +} + +MA_API ma_result ma_audio_buffer_ref_map(ma_audio_buffer_ref* pAudioBufferRef, void** ppFramesOut, ma_uint64* pFrameCount) +{ + ma_uint64 framesAvailable; + ma_uint64 frameCount = 0; + + if (ppFramesOut != NULL) { + *ppFramesOut = NULL; /* Safety. */ + } + + if (pFrameCount != NULL) { + frameCount = *pFrameCount; + *pFrameCount = 0; /* Safety. */ + } + + if (pAudioBufferRef == NULL || ppFramesOut == NULL || pFrameCount == NULL) { + return MA_INVALID_ARGS; + } + + framesAvailable = pAudioBufferRef->sizeInFrames - pAudioBufferRef->cursor; + if (frameCount > framesAvailable) { + frameCount = framesAvailable; + } + + *ppFramesOut = ma_offset_ptr(pAudioBufferRef->pData, pAudioBufferRef->cursor * ma_get_bytes_per_frame(pAudioBufferRef->format, pAudioBufferRef->channels)); + *pFrameCount = frameCount; + + return MA_SUCCESS; +} + +MA_API ma_result ma_audio_buffer_ref_unmap(ma_audio_buffer_ref* pAudioBufferRef, ma_uint64 frameCount) +{ + ma_uint64 framesAvailable; + + if (pAudioBufferRef == NULL) { + return MA_INVALID_ARGS; + } + + framesAvailable = pAudioBufferRef->sizeInFrames - pAudioBufferRef->cursor; + if (frameCount > framesAvailable) { + return MA_INVALID_ARGS; /* The frame count was too big. This should never happen in an unmapping. Need to make sure the caller is aware of this. */ + } + + pAudioBufferRef->cursor += frameCount; + + if (pAudioBufferRef->cursor == pAudioBufferRef->sizeInFrames) { + return MA_AT_END; /* Successful. Need to tell the caller that the end has been reached so that it can loop if desired. */ + } else { + return MA_SUCCESS; + } +} + +MA_API ma_bool32 ma_audio_buffer_ref_at_end(const ma_audio_buffer_ref* pAudioBufferRef) +{ + if (pAudioBufferRef == NULL) { + return MA_FALSE; + } + + return pAudioBufferRef->cursor == pAudioBufferRef->sizeInFrames; +} + +MA_API ma_result ma_audio_buffer_ref_get_cursor_in_pcm_frames(const ma_audio_buffer_ref* pAudioBufferRef, ma_uint64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; + + if (pAudioBufferRef == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = pAudioBufferRef->cursor; + + return MA_SUCCESS; +} + +MA_API ma_result ma_audio_buffer_ref_get_length_in_pcm_frames(const ma_audio_buffer_ref* pAudioBufferRef, ma_uint64* pLength) +{ + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; + + if (pAudioBufferRef == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = pAudioBufferRef->sizeInFrames; + + return MA_SUCCESS; +} + +MA_API ma_result ma_audio_buffer_ref_get_available_frames(const ma_audio_buffer_ref* pAudioBufferRef, ma_uint64* pAvailableFrames) +{ + if (pAvailableFrames == NULL) { + return MA_INVALID_ARGS; + } + + *pAvailableFrames = 0; + + if (pAudioBufferRef == NULL) { + return MA_INVALID_ARGS; + } + + if (pAudioBufferRef->sizeInFrames <= pAudioBufferRef->cursor) { + *pAvailableFrames = 0; + } else { + *pAvailableFrames = pAudioBufferRef->sizeInFrames - pAudioBufferRef->cursor; + } + + return MA_SUCCESS; +} + + + + +MA_API ma_audio_buffer_config ma_audio_buffer_config_init(ma_format format, ma_uint32 channels, ma_uint64 sizeInFrames, const void* pData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_audio_buffer_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = 0; /* TODO: Version 0.12. Set this to sampleRate. */ + config.sizeInFrames = sizeInFrames; + config.pData = pData; + ma_allocation_callbacks_init_copy(&config.allocationCallbacks, pAllocationCallbacks); + + return config; +} + +static ma_result ma_audio_buffer_init_ex(const ma_audio_buffer_config* pConfig, ma_bool32 doCopy, ma_audio_buffer* pAudioBuffer) +{ + ma_result result; + + if (pAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_MEMORY(pAudioBuffer, sizeof(*pAudioBuffer) - sizeof(pAudioBuffer->_pExtraData)); /* Safety. Don't overwrite the extra data. */ + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->sizeInFrames == 0) { + return MA_INVALID_ARGS; /* Not allowing buffer sizes of 0 frames. */ + } + + result = ma_audio_buffer_ref_init(pConfig->format, pConfig->channels, NULL, 0, &pAudioBuffer->ref); + if (result != MA_SUCCESS) { + return result; + } + + /* TODO: Version 0.12. Set this in ma_audio_buffer_ref_init() instead of here. */ + pAudioBuffer->ref.sampleRate = pConfig->sampleRate; + + ma_allocation_callbacks_init_copy(&pAudioBuffer->allocationCallbacks, &pConfig->allocationCallbacks); + + if (doCopy) { + ma_uint64 allocationSizeInBytes; + void* pData; + + allocationSizeInBytes = pConfig->sizeInFrames * ma_get_bytes_per_frame(pConfig->format, pConfig->channels); + if (allocationSizeInBytes > MA_SIZE_MAX) { + return MA_OUT_OF_MEMORY; /* Too big. */ + } + + pData = ma_malloc((size_t)allocationSizeInBytes, &pAudioBuffer->allocationCallbacks); /* Safe cast to size_t. */ + if (pData == NULL) { + return MA_OUT_OF_MEMORY; + } + + if (pConfig->pData != NULL) { + ma_copy_pcm_frames(pData, pConfig->pData, pConfig->sizeInFrames, pConfig->format, pConfig->channels); + } else { + ma_silence_pcm_frames(pData, pConfig->sizeInFrames, pConfig->format, pConfig->channels); + } + + ma_audio_buffer_ref_set_data(&pAudioBuffer->ref, pData, pConfig->sizeInFrames); + pAudioBuffer->ownsData = MA_TRUE; + } else { + ma_audio_buffer_ref_set_data(&pAudioBuffer->ref, pConfig->pData, pConfig->sizeInFrames); + pAudioBuffer->ownsData = MA_FALSE; + } + + return MA_SUCCESS; +} + +static void ma_audio_buffer_uninit_ex(ma_audio_buffer* pAudioBuffer, ma_bool32 doFree) +{ + if (pAudioBuffer == NULL) { + return; + } + + if (pAudioBuffer->ownsData && pAudioBuffer->ref.pData != &pAudioBuffer->_pExtraData[0]) { + ma_free((void*)pAudioBuffer->ref.pData, &pAudioBuffer->allocationCallbacks); /* Naugty const cast, but OK in this case since we've guarded it with the ownsData check. */ + } + + if (doFree) { + ma_free(pAudioBuffer, &pAudioBuffer->allocationCallbacks); + } + + ma_audio_buffer_ref_uninit(&pAudioBuffer->ref); +} + +MA_API ma_result ma_audio_buffer_init(const ma_audio_buffer_config* pConfig, ma_audio_buffer* pAudioBuffer) +{ + return ma_audio_buffer_init_ex(pConfig, MA_FALSE, pAudioBuffer); +} + +MA_API ma_result ma_audio_buffer_init_copy(const ma_audio_buffer_config* pConfig, ma_audio_buffer* pAudioBuffer) +{ + return ma_audio_buffer_init_ex(pConfig, MA_TRUE, pAudioBuffer); +} + +MA_API ma_result ma_audio_buffer_alloc_and_init(const ma_audio_buffer_config* pConfig, ma_audio_buffer** ppAudioBuffer) +{ + ma_result result; + ma_audio_buffer* pAudioBuffer; + ma_audio_buffer_config innerConfig; /* We'll be making some changes to the config, so need to make a copy. */ + ma_uint64 allocationSizeInBytes; + + if (ppAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + *ppAudioBuffer = NULL; /* Safety. */ + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + innerConfig = *pConfig; + ma_allocation_callbacks_init_copy(&innerConfig.allocationCallbacks, &pConfig->allocationCallbacks); + + allocationSizeInBytes = sizeof(*pAudioBuffer) - sizeof(pAudioBuffer->_pExtraData) + (pConfig->sizeInFrames * ma_get_bytes_per_frame(pConfig->format, pConfig->channels)); + if (allocationSizeInBytes > MA_SIZE_MAX) { + return MA_OUT_OF_MEMORY; /* Too big. */ + } + + pAudioBuffer = (ma_audio_buffer*)ma_malloc((size_t)allocationSizeInBytes, &innerConfig.allocationCallbacks); /* Safe cast to size_t. */ + if (pAudioBuffer == NULL) { + return MA_OUT_OF_MEMORY; + } + + if (pConfig->pData != NULL) { + ma_copy_pcm_frames(&pAudioBuffer->_pExtraData[0], pConfig->pData, pConfig->sizeInFrames, pConfig->format, pConfig->channels); + } else { + ma_silence_pcm_frames(&pAudioBuffer->_pExtraData[0], pConfig->sizeInFrames, pConfig->format, pConfig->channels); + } + + innerConfig.pData = &pAudioBuffer->_pExtraData[0]; + + result = ma_audio_buffer_init_ex(&innerConfig, MA_FALSE, pAudioBuffer); + if (result != MA_SUCCESS) { + ma_free(pAudioBuffer, &innerConfig.allocationCallbacks); + return result; + } + + *ppAudioBuffer = pAudioBuffer; + + return MA_SUCCESS; +} + +MA_API void ma_audio_buffer_uninit(ma_audio_buffer* pAudioBuffer) +{ + ma_audio_buffer_uninit_ex(pAudioBuffer, MA_FALSE); +} + +MA_API void ma_audio_buffer_uninit_and_free(ma_audio_buffer* pAudioBuffer) +{ + ma_audio_buffer_uninit_ex(pAudioBuffer, MA_TRUE); +} + +MA_API ma_uint64 ma_audio_buffer_read_pcm_frames(ma_audio_buffer* pAudioBuffer, void* pFramesOut, ma_uint64 frameCount, ma_bool32 loop) +{ + if (pAudioBuffer == NULL) { + return 0; + } + + return ma_audio_buffer_ref_read_pcm_frames(&pAudioBuffer->ref, pFramesOut, frameCount, loop); +} + +MA_API ma_result ma_audio_buffer_seek_to_pcm_frame(ma_audio_buffer* pAudioBuffer, ma_uint64 frameIndex) +{ + if (pAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + return ma_audio_buffer_ref_seek_to_pcm_frame(&pAudioBuffer->ref, frameIndex); +} + +MA_API ma_result ma_audio_buffer_map(ma_audio_buffer* pAudioBuffer, void** ppFramesOut, ma_uint64* pFrameCount) +{ + if (ppFramesOut != NULL) { + *ppFramesOut = NULL; /* Safety. */ + } + + if (pAudioBuffer == NULL) { + if (pFrameCount != NULL) { + *pFrameCount = 0; + } + + return MA_INVALID_ARGS; + } + + return ma_audio_buffer_ref_map(&pAudioBuffer->ref, ppFramesOut, pFrameCount); +} + +MA_API ma_result ma_audio_buffer_unmap(ma_audio_buffer* pAudioBuffer, ma_uint64 frameCount) +{ + if (pAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + return ma_audio_buffer_ref_unmap(&pAudioBuffer->ref, frameCount); +} + +MA_API ma_bool32 ma_audio_buffer_at_end(const ma_audio_buffer* pAudioBuffer) +{ + if (pAudioBuffer == NULL) { + return MA_FALSE; + } + + return ma_audio_buffer_ref_at_end(&pAudioBuffer->ref); +} + +MA_API ma_result ma_audio_buffer_get_cursor_in_pcm_frames(const ma_audio_buffer* pAudioBuffer, ma_uint64* pCursor) +{ + if (pAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + return ma_audio_buffer_ref_get_cursor_in_pcm_frames(&pAudioBuffer->ref, pCursor); +} + +MA_API ma_result ma_audio_buffer_get_length_in_pcm_frames(const ma_audio_buffer* pAudioBuffer, ma_uint64* pLength) +{ + if (pAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + return ma_audio_buffer_ref_get_length_in_pcm_frames(&pAudioBuffer->ref, pLength); +} + +MA_API ma_result ma_audio_buffer_get_available_frames(const ma_audio_buffer* pAudioBuffer, ma_uint64* pAvailableFrames) +{ + if (pAvailableFrames == NULL) { + return MA_INVALID_ARGS; + } + + *pAvailableFrames = 0; + + if (pAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + return ma_audio_buffer_ref_get_available_frames(&pAudioBuffer->ref, pAvailableFrames); +} + + + + + +MA_API ma_result ma_paged_audio_buffer_data_init(ma_format format, ma_uint32 channels, ma_paged_audio_buffer_data* pData) +{ + if (pData == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pData); + + pData->format = format; + pData->channels = channels; + pData->pTail = &pData->head; + + return MA_SUCCESS; +} + +MA_API void ma_paged_audio_buffer_data_uninit(ma_paged_audio_buffer_data* pData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_paged_audio_buffer_page* pPage; + + if (pData == NULL) { + return; + } + + /* All pages need to be freed. */ + pPage = (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(&pData->head.pNext); + while (pPage != NULL) { + ma_paged_audio_buffer_page* pNext = (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(&pPage->pNext); + + ma_free(pPage, pAllocationCallbacks); + pPage = pNext; + } +} + +MA_API ma_paged_audio_buffer_page* ma_paged_audio_buffer_data_get_head(ma_paged_audio_buffer_data* pData) +{ + if (pData == NULL) { + return NULL; + } + + return &pData->head; +} + +MA_API ma_paged_audio_buffer_page* ma_paged_audio_buffer_data_get_tail(ma_paged_audio_buffer_data* pData) +{ + if (pData == NULL) { + return NULL; + } + + return pData->pTail; +} + +MA_API ma_result ma_paged_audio_buffer_data_get_length_in_pcm_frames(ma_paged_audio_buffer_data* pData, ma_uint64* pLength) +{ + ma_paged_audio_buffer_page* pPage; + + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; + + if (pData == NULL) { + return MA_INVALID_ARGS; + } + + /* Calculate the length from the linked list. */ + for (pPage = (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(&pData->head.pNext); pPage != NULL; pPage = (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(&pPage->pNext)) { + *pLength += pPage->sizeInFrames; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_paged_audio_buffer_data_allocate_page(ma_paged_audio_buffer_data* pData, ma_uint64 pageSizeInFrames, const void* pInitialData, const ma_allocation_callbacks* pAllocationCallbacks, ma_paged_audio_buffer_page** ppPage) +{ + ma_paged_audio_buffer_page* pPage; + ma_uint64 allocationSize; + + if (ppPage == NULL) { + return MA_INVALID_ARGS; + } + + *ppPage = NULL; + + if (pData == NULL) { + return MA_INVALID_ARGS; + } + + allocationSize = sizeof(*pPage) + (pageSizeInFrames * ma_get_bytes_per_frame(pData->format, pData->channels)); + if (allocationSize > MA_SIZE_MAX) { + return MA_OUT_OF_MEMORY; /* Too big. */ + } + + pPage = (ma_paged_audio_buffer_page*)ma_malloc((size_t)allocationSize, pAllocationCallbacks); /* Safe cast to size_t. */ + if (pPage == NULL) { + return MA_OUT_OF_MEMORY; + } + + pPage->pNext = NULL; + pPage->sizeInFrames = pageSizeInFrames; + + if (pInitialData != NULL) { + ma_copy_pcm_frames(pPage->pAudioData, pInitialData, pageSizeInFrames, pData->format, pData->channels); + } + + *ppPage = pPage; + + return MA_SUCCESS; +} + +MA_API ma_result ma_paged_audio_buffer_data_free_page(ma_paged_audio_buffer_data* pData, ma_paged_audio_buffer_page* pPage, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pData == NULL || pPage == NULL) { + return MA_INVALID_ARGS; + } + + /* It's assumed the page is not attached to the list. */ + ma_free(pPage, pAllocationCallbacks); + + return MA_SUCCESS; +} + +MA_API ma_result ma_paged_audio_buffer_data_append_page(ma_paged_audio_buffer_data* pData, ma_paged_audio_buffer_page* pPage) +{ + if (pData == NULL || pPage == NULL) { + return MA_INVALID_ARGS; + } + + /* This function assumes the page has been filled with audio data by this point. As soon as we append, the page will be available for reading. */ + + /* First thing to do is update the tail. */ + for (;;) { + ma_paged_audio_buffer_page* pOldTail = (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(&pData->pTail); + ma_paged_audio_buffer_page* pNewTail = pPage; + + if (ma_atomic_compare_exchange_weak_ptr((volatile void**)&pData->pTail, (void**)&pOldTail, pNewTail)) { + /* Here is where we append the page to the list. After this, the page is attached to the list and ready to be read from. */ + ma_atomic_exchange_ptr(&pOldTail->pNext, pPage); + break; /* Done. */ + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_paged_audio_buffer_data_allocate_and_append_page(ma_paged_audio_buffer_data* pData, ma_uint32 pageSizeInFrames, const void* pInitialData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_result result; + ma_paged_audio_buffer_page* pPage; + + result = ma_paged_audio_buffer_data_allocate_page(pData, pageSizeInFrames, pInitialData, pAllocationCallbacks, &pPage); + if (result != MA_SUCCESS) { + return result; + } + + return ma_paged_audio_buffer_data_append_page(pData, pPage); /* <-- Should never fail. */ +} + + +MA_API ma_paged_audio_buffer_config ma_paged_audio_buffer_config_init(ma_paged_audio_buffer_data* pData) +{ + ma_paged_audio_buffer_config config; + + MA_ZERO_OBJECT(&config); + config.pData = pData; + + return config; +} + + +static ma_result ma_paged_audio_buffer__data_source_on_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_paged_audio_buffer_read_pcm_frames((ma_paged_audio_buffer*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_paged_audio_buffer__data_source_on_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_paged_audio_buffer_seek_to_pcm_frame((ma_paged_audio_buffer*)pDataSource, frameIndex); +} + +static ma_result ma_paged_audio_buffer__data_source_on_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + ma_paged_audio_buffer* pPagedAudioBuffer = (ma_paged_audio_buffer*)pDataSource; + + *pFormat = pPagedAudioBuffer->pData->format; + *pChannels = pPagedAudioBuffer->pData->channels; + *pSampleRate = 0; /* There is no notion of a sample rate with audio buffers. */ + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pPagedAudioBuffer->pData->channels); + + return MA_SUCCESS; +} + +static ma_result ma_paged_audio_buffer__data_source_on_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + return ma_paged_audio_buffer_get_cursor_in_pcm_frames((ma_paged_audio_buffer*)pDataSource, pCursor); +} + +static ma_result ma_paged_audio_buffer__data_source_on_get_length(ma_data_source* pDataSource, ma_uint64* pLength) +{ + return ma_paged_audio_buffer_get_length_in_pcm_frames((ma_paged_audio_buffer*)pDataSource, pLength); +} + +static ma_data_source_vtable g_ma_paged_audio_buffer_data_source_vtable = +{ + ma_paged_audio_buffer__data_source_on_read, + ma_paged_audio_buffer__data_source_on_seek, + ma_paged_audio_buffer__data_source_on_get_data_format, + ma_paged_audio_buffer__data_source_on_get_cursor, + ma_paged_audio_buffer__data_source_on_get_length, + NULL, /* onSetLooping */ + 0 +}; + +MA_API ma_result ma_paged_audio_buffer_init(const ma_paged_audio_buffer_config* pConfig, ma_paged_audio_buffer* pPagedAudioBuffer) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + + if (pPagedAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pPagedAudioBuffer); + + /* A config is required for the format and channel count. */ + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->pData == NULL) { + return MA_INVALID_ARGS; /* No underlying data specified. */ + } + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_paged_audio_buffer_data_source_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pPagedAudioBuffer->ds); + if (result != MA_SUCCESS) { + return result; + } + + pPagedAudioBuffer->pData = pConfig->pData; + pPagedAudioBuffer->pCurrent = ma_paged_audio_buffer_data_get_head(pConfig->pData); + pPagedAudioBuffer->relativeCursor = 0; + pPagedAudioBuffer->absoluteCursor = 0; + + return MA_SUCCESS; +} + +MA_API void ma_paged_audio_buffer_uninit(ma_paged_audio_buffer* pPagedAudioBuffer) +{ + if (pPagedAudioBuffer == NULL) { + return; + } + + /* Nothing to do. The data needs to be deleted separately. */ +} + +MA_API ma_result ma_paged_audio_buffer_read_pcm_frames(ma_paged_audio_buffer* pPagedAudioBuffer, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_result result = MA_SUCCESS; + ma_uint64 totalFramesRead = 0; + ma_format format; + ma_uint32 channels; + + if (pPagedAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + format = pPagedAudioBuffer->pData->format; + channels = pPagedAudioBuffer->pData->channels; + + while (totalFramesRead < frameCount) { + /* Read from the current page. The buffer should never be in a state where this is NULL. */ + ma_uint64 framesRemainingInCurrentPage; + ma_uint64 framesRemainingToRead = frameCount - totalFramesRead; + ma_uint64 framesToReadThisIteration; + + MA_ASSERT(pPagedAudioBuffer->pCurrent != NULL); + + framesRemainingInCurrentPage = pPagedAudioBuffer->pCurrent->sizeInFrames - pPagedAudioBuffer->relativeCursor; + + framesToReadThisIteration = ma_min(framesRemainingInCurrentPage, framesRemainingToRead); + ma_copy_pcm_frames(ma_offset_pcm_frames_ptr(pFramesOut, totalFramesRead, format, channels), ma_offset_pcm_frames_ptr(pPagedAudioBuffer->pCurrent->pAudioData, pPagedAudioBuffer->relativeCursor, format, channels), framesToReadThisIteration, format, channels); + totalFramesRead += framesToReadThisIteration; + + pPagedAudioBuffer->absoluteCursor += framesToReadThisIteration; + pPagedAudioBuffer->relativeCursor += framesToReadThisIteration; + + /* Move to the next page if necessary. If there's no more pages, we need to return MA_AT_END. */ + MA_ASSERT(pPagedAudioBuffer->relativeCursor <= pPagedAudioBuffer->pCurrent->sizeInFrames); + + if (pPagedAudioBuffer->relativeCursor == pPagedAudioBuffer->pCurrent->sizeInFrames) { + /* We reached the end of the page. Need to move to the next. If there's no more pages, we're done. */ + ma_paged_audio_buffer_page* pNext = (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(&pPagedAudioBuffer->pCurrent->pNext); + if (pNext == NULL) { + result = MA_AT_END; + break; /* We've reached the end. */ + } else { + pPagedAudioBuffer->pCurrent = pNext; + pPagedAudioBuffer->relativeCursor = 0; + } + } + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesRead; + } + + return result; +} + +MA_API ma_result ma_paged_audio_buffer_seek_to_pcm_frame(ma_paged_audio_buffer* pPagedAudioBuffer, ma_uint64 frameIndex) +{ + if (pPagedAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + if (frameIndex == pPagedAudioBuffer->absoluteCursor) { + return MA_SUCCESS; /* Nothing to do. */ + } + + if (frameIndex < pPagedAudioBuffer->absoluteCursor) { + /* Moving backwards. Need to move the cursor back to the start, and then move forward. */ + pPagedAudioBuffer->pCurrent = ma_paged_audio_buffer_data_get_head(pPagedAudioBuffer->pData); + pPagedAudioBuffer->absoluteCursor = 0; + pPagedAudioBuffer->relativeCursor = 0; + + /* Fall through to the forward seeking section below. */ + } + + if (frameIndex > pPagedAudioBuffer->absoluteCursor) { + /* Moving forward. */ + ma_paged_audio_buffer_page* pPage; + ma_uint64 runningCursor = 0; + + for (pPage = (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(&ma_paged_audio_buffer_data_get_head(pPagedAudioBuffer->pData)->pNext); pPage != NULL; pPage = (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(&pPage->pNext)) { + ma_uint64 pageRangeBeg = runningCursor; + ma_uint64 pageRangeEnd = pageRangeBeg + pPage->sizeInFrames; + + if (frameIndex >= pageRangeBeg) { + if (frameIndex < pageRangeEnd || (frameIndex == pageRangeEnd && pPage == (ma_paged_audio_buffer_page*)ma_atomic_load_ptr(ma_paged_audio_buffer_data_get_tail(pPagedAudioBuffer->pData)))) { /* A small edge case - allow seeking to the very end of the buffer. */ + /* We found the page. */ + pPagedAudioBuffer->pCurrent = pPage; + pPagedAudioBuffer->absoluteCursor = frameIndex; + pPagedAudioBuffer->relativeCursor = frameIndex - pageRangeBeg; + return MA_SUCCESS; + } + } + + runningCursor = pageRangeEnd; + } + + /* Getting here means we tried seeking too far forward. Don't change any state. */ + return MA_BAD_SEEK; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_paged_audio_buffer_get_cursor_in_pcm_frames(ma_paged_audio_buffer* pPagedAudioBuffer, ma_uint64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; /* Safety. */ + + if (pPagedAudioBuffer == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = pPagedAudioBuffer->absoluteCursor; + + return MA_SUCCESS; +} + +MA_API ma_result ma_paged_audio_buffer_get_length_in_pcm_frames(ma_paged_audio_buffer* pPagedAudioBuffer, ma_uint64* pLength) +{ + return ma_paged_audio_buffer_data_get_length_in_pcm_frames(pPagedAudioBuffer->pData, pLength); +} + + + +/************************************************************************************************************************************************************** + +VFS + +**************************************************************************************************************************************************************/ +MA_API ma_result ma_vfs_open(ma_vfs* pVFS, const char* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + ma_vfs_callbacks* pCallbacks = (ma_vfs_callbacks*)pVFS; + + if (pFile == NULL) { + return MA_INVALID_ARGS; + } + + *pFile = NULL; + + if (pVFS == NULL || pFilePath == NULL || openMode == 0) { + return MA_INVALID_ARGS; + } + + if (pCallbacks->onOpen == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pCallbacks->onOpen(pVFS, pFilePath, openMode, pFile); +} + +MA_API ma_result ma_vfs_open_w(ma_vfs* pVFS, const wchar_t* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + ma_vfs_callbacks* pCallbacks = (ma_vfs_callbacks*)pVFS; + + if (pFile == NULL) { + return MA_INVALID_ARGS; + } + + *pFile = NULL; + + if (pVFS == NULL || pFilePath == NULL || openMode == 0) { + return MA_INVALID_ARGS; + } + + if (pCallbacks->onOpenW == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pCallbacks->onOpenW(pVFS, pFilePath, openMode, pFile); +} + +MA_API ma_result ma_vfs_close(ma_vfs* pVFS, ma_vfs_file file) +{ + ma_vfs_callbacks* pCallbacks = (ma_vfs_callbacks*)pVFS; + + if (pVFS == NULL || file == NULL) { + return MA_INVALID_ARGS; + } + + if (pCallbacks->onClose == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pCallbacks->onClose(pVFS, file); +} + +MA_API ma_result ma_vfs_read(ma_vfs* pVFS, ma_vfs_file file, void* pDst, size_t sizeInBytes, size_t* pBytesRead) +{ + ma_vfs_callbacks* pCallbacks = (ma_vfs_callbacks*)pVFS; + ma_result result; + size_t bytesRead = 0; + + if (pBytesRead != NULL) { + *pBytesRead = 0; + } + + if (pVFS == NULL || file == NULL || pDst == NULL) { + return MA_INVALID_ARGS; + } + + if (pCallbacks->onRead == NULL) { + return MA_NOT_IMPLEMENTED; + } + + result = pCallbacks->onRead(pVFS, file, pDst, sizeInBytes, &bytesRead); + + if (pBytesRead != NULL) { + *pBytesRead = bytesRead; + } + + if (result == MA_SUCCESS && bytesRead == 0 && sizeInBytes > 0) { + result = MA_AT_END; + } + + return result; +} + +MA_API ma_result ma_vfs_write(ma_vfs* pVFS, ma_vfs_file file, const void* pSrc, size_t sizeInBytes, size_t* pBytesWritten) +{ + ma_vfs_callbacks* pCallbacks = (ma_vfs_callbacks*)pVFS; + + if (pBytesWritten != NULL) { + *pBytesWritten = 0; + } + + if (pVFS == NULL || file == NULL || pSrc == NULL) { + return MA_INVALID_ARGS; + } + + if (pCallbacks->onWrite == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pCallbacks->onWrite(pVFS, file, pSrc, sizeInBytes, pBytesWritten); +} + +MA_API ma_result ma_vfs_seek(ma_vfs* pVFS, ma_vfs_file file, ma_int64 offset, ma_seek_origin origin) +{ + ma_vfs_callbacks* pCallbacks = (ma_vfs_callbacks*)pVFS; + + if (pVFS == NULL || file == NULL) { + return MA_INVALID_ARGS; + } + + if (pCallbacks->onSeek == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pCallbacks->onSeek(pVFS, file, offset, origin); +} + +MA_API ma_result ma_vfs_tell(ma_vfs* pVFS, ma_vfs_file file, ma_int64* pCursor) +{ + ma_vfs_callbacks* pCallbacks = (ma_vfs_callbacks*)pVFS; + + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; + + if (pVFS == NULL || file == NULL) { + return MA_INVALID_ARGS; + } + + if (pCallbacks->onTell == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pCallbacks->onTell(pVFS, file, pCursor); +} + +MA_API ma_result ma_vfs_info(ma_vfs* pVFS, ma_vfs_file file, ma_file_info* pInfo) +{ + ma_vfs_callbacks* pCallbacks = (ma_vfs_callbacks*)pVFS; + + if (pInfo == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pInfo); + + if (pVFS == NULL || file == NULL) { + return MA_INVALID_ARGS; + } + + if (pCallbacks->onInfo == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pCallbacks->onInfo(pVFS, file, pInfo); +} + + +#if !defined(MA_USE_WIN32_FILEIO) && (defined(MA_WIN32) && defined(MA_WIN32_DESKTOP) && !defined(MA_NO_WIN32_FILEIO) && !defined(MA_POSIX)) + #define MA_USE_WIN32_FILEIO +#endif + +#if defined(MA_USE_WIN32_FILEIO) +/* +We need to dynamically load SetFilePointer or SetFilePointerEx because older versions of Windows do +not have the Ex version. We therefore need to do some dynamic branching depending on what's available. + +We load these when we load our first file from the default VFS. It's left open for the life of the +program and is left to the OS to uninitialize when the program terminates. +*/ +typedef DWORD (__stdcall * ma_SetFilePointer_proc)(HANDLE hFile, LONG lDistanceToMove, LONG* lpDistanceToMoveHigh, DWORD dwMoveMethod); +typedef BOOL (__stdcall * ma_SetFilePointerEx_proc)(HANDLE hFile, LARGE_INTEGER liDistanceToMove, LARGE_INTEGER* lpNewFilePointer, DWORD dwMoveMethod); + +static ma_handle hKernel32DLL = NULL; +static ma_SetFilePointer_proc ma_SetFilePointer = NULL; +static ma_SetFilePointerEx_proc ma_SetFilePointerEx = NULL; + +static void ma_win32_fileio_init(void) +{ + if (hKernel32DLL == NULL) { + hKernel32DLL = ma_dlopen(NULL, "kernel32.dll"); + if (hKernel32DLL != NULL) { + ma_SetFilePointer = (ma_SetFilePointer_proc) ma_dlsym(NULL, hKernel32DLL, "SetFilePointer"); + ma_SetFilePointerEx = (ma_SetFilePointerEx_proc)ma_dlsym(NULL, hKernel32DLL, "SetFilePointerEx"); + } + } +} + +static void ma_default_vfs__get_open_settings_win32(ma_uint32 openMode, DWORD* pDesiredAccess, DWORD* pShareMode, DWORD* pCreationDisposition) +{ + *pDesiredAccess = 0; + if ((openMode & MA_OPEN_MODE_READ) != 0) { + *pDesiredAccess |= GENERIC_READ; + } + if ((openMode & MA_OPEN_MODE_WRITE) != 0) { + *pDesiredAccess |= GENERIC_WRITE; + } + + *pShareMode = 0; + if ((openMode & MA_OPEN_MODE_READ) != 0) { + *pShareMode |= FILE_SHARE_READ; + } + + if ((openMode & MA_OPEN_MODE_WRITE) != 0) { + *pCreationDisposition = CREATE_ALWAYS; /* Opening in write mode. Truncate. */ + } else { + *pCreationDisposition = OPEN_EXISTING; /* Opening in read mode. File must exist. */ + } +} + +static ma_result ma_default_vfs_open__win32(ma_vfs* pVFS, const char* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + HANDLE hFile; + DWORD dwDesiredAccess; + DWORD dwShareMode; + DWORD dwCreationDisposition; + + (void)pVFS; + + /* Load some Win32 symbols dynamically so we can dynamically check for the existence of SetFilePointerEx. */ + ma_win32_fileio_init(); + + ma_default_vfs__get_open_settings_win32(openMode, &dwDesiredAccess, &dwShareMode, &dwCreationDisposition); + + hFile = CreateFileA(pFilePath, dwDesiredAccess, dwShareMode, NULL, dwCreationDisposition, FILE_ATTRIBUTE_NORMAL, NULL); + if (hFile == INVALID_HANDLE_VALUE) { + return ma_result_from_GetLastError(GetLastError()); + } + + *pFile = hFile; + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_open_w__win32(ma_vfs* pVFS, const wchar_t* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + HANDLE hFile; + DWORD dwDesiredAccess; + DWORD dwShareMode; + DWORD dwCreationDisposition; + + (void)pVFS; + + /* Load some Win32 symbols dynamically so we can dynamically check for the existence of SetFilePointerEx. */ + ma_win32_fileio_init(); + + ma_default_vfs__get_open_settings_win32(openMode, &dwDesiredAccess, &dwShareMode, &dwCreationDisposition); + + hFile = CreateFileW(pFilePath, dwDesiredAccess, dwShareMode, NULL, dwCreationDisposition, FILE_ATTRIBUTE_NORMAL, NULL); + if (hFile == INVALID_HANDLE_VALUE) { + return ma_result_from_GetLastError(GetLastError()); + } + + *pFile = hFile; + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_close__win32(ma_vfs* pVFS, ma_vfs_file file) +{ + (void)pVFS; + + if (CloseHandle((HANDLE)file) == 0) { + return ma_result_from_GetLastError(GetLastError()); + } + + return MA_SUCCESS; +} + + +static ma_result ma_default_vfs_read__win32(ma_vfs* pVFS, ma_vfs_file file, void* pDst, size_t sizeInBytes, size_t* pBytesRead) +{ + ma_result result = MA_SUCCESS; + size_t totalBytesRead; + + (void)pVFS; + + totalBytesRead = 0; + while (totalBytesRead < sizeInBytes) { + size_t bytesRemaining; + DWORD bytesToRead; + DWORD bytesRead; + BOOL readResult; + + bytesRemaining = sizeInBytes - totalBytesRead; + if (bytesRemaining >= 0xFFFFFFFF) { + bytesToRead = 0xFFFFFFFF; + } else { + bytesToRead = (DWORD)bytesRemaining; + } + + readResult = ReadFile((HANDLE)file, ma_offset_ptr(pDst, totalBytesRead), bytesToRead, &bytesRead, NULL); + if (readResult == 1 && bytesRead == 0) { + result = MA_AT_END; + break; /* EOF */ + } + + totalBytesRead += bytesRead; + + if (bytesRead < bytesToRead) { + break; /* EOF */ + } + + if (readResult == 0) { + result = ma_result_from_GetLastError(GetLastError()); + break; + } + } + + if (pBytesRead != NULL) { + *pBytesRead = totalBytesRead; + } + + return result; +} + +static ma_result ma_default_vfs_write__win32(ma_vfs* pVFS, ma_vfs_file file, const void* pSrc, size_t sizeInBytes, size_t* pBytesWritten) +{ + ma_result result = MA_SUCCESS; + size_t totalBytesWritten; + + (void)pVFS; + + totalBytesWritten = 0; + while (totalBytesWritten < sizeInBytes) { + size_t bytesRemaining; + DWORD bytesToWrite; + DWORD bytesWritten; + BOOL writeResult; + + bytesRemaining = sizeInBytes - totalBytesWritten; + if (bytesRemaining >= 0xFFFFFFFF) { + bytesToWrite = 0xFFFFFFFF; + } else { + bytesToWrite = (DWORD)bytesRemaining; + } + + writeResult = WriteFile((HANDLE)file, ma_offset_ptr(pSrc, totalBytesWritten), bytesToWrite, &bytesWritten, NULL); + totalBytesWritten += bytesWritten; + + if (writeResult == 0) { + result = ma_result_from_GetLastError(GetLastError()); + break; + } + } + + if (pBytesWritten != NULL) { + *pBytesWritten = totalBytesWritten; + } + + return result; +} + + +static ma_result ma_default_vfs_seek__win32(ma_vfs* pVFS, ma_vfs_file file, ma_int64 offset, ma_seek_origin origin) +{ + LARGE_INTEGER liDistanceToMove; + DWORD dwMoveMethod; + BOOL result; + + (void)pVFS; + + liDistanceToMove.QuadPart = offset; + + /* */ if (origin == ma_seek_origin_current) { + dwMoveMethod = FILE_CURRENT; + } else if (origin == ma_seek_origin_end) { + dwMoveMethod = FILE_END; + } else { + dwMoveMethod = FILE_BEGIN; + } + + if (ma_SetFilePointerEx != NULL) { + result = ma_SetFilePointerEx((HANDLE)file, liDistanceToMove, NULL, dwMoveMethod); + } else if (ma_SetFilePointer != NULL) { + /* No SetFilePointerEx() so restrict to 31 bits. */ + if (offset > 0x7FFFFFFF) { + return MA_OUT_OF_RANGE; + } + + result = ma_SetFilePointer((HANDLE)file, (LONG)liDistanceToMove.QuadPart, NULL, dwMoveMethod); + } else { + return MA_NOT_IMPLEMENTED; + } + + if (result == 0) { + return ma_result_from_GetLastError(GetLastError()); + } + + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_tell__win32(ma_vfs* pVFS, ma_vfs_file file, ma_int64* pCursor) +{ + LARGE_INTEGER liZero; + LARGE_INTEGER liTell; + BOOL result; + + (void)pVFS; + + liZero.QuadPart = 0; + + if (ma_SetFilePointerEx != NULL) { + result = ma_SetFilePointerEx((HANDLE)file, liZero, &liTell, FILE_CURRENT); + } else if (ma_SetFilePointer != NULL) { + LONG tell; + + result = ma_SetFilePointer((HANDLE)file, (LONG)liZero.QuadPart, &tell, FILE_CURRENT); + liTell.QuadPart = tell; + } else { + return MA_NOT_IMPLEMENTED; + } + + if (result == 0) { + return ma_result_from_GetLastError(GetLastError()); + } + + if (pCursor != NULL) { + *pCursor = liTell.QuadPart; + } + + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_info__win32(ma_vfs* pVFS, ma_vfs_file file, ma_file_info* pInfo) +{ + BY_HANDLE_FILE_INFORMATION fi; + BOOL result; + + (void)pVFS; + + result = GetFileInformationByHandle((HANDLE)file, &fi); + if (result == 0) { + return ma_result_from_GetLastError(GetLastError()); + } + + pInfo->sizeInBytes = ((ma_uint64)fi.nFileSizeHigh << 32) | ((ma_uint64)fi.nFileSizeLow); + + return MA_SUCCESS; +} +#else +static ma_result ma_default_vfs_open__stdio(ma_vfs* pVFS, const char* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + ma_result result; + FILE* pFileStd; + const char* pOpenModeStr; + + MA_ASSERT(pFilePath != NULL); + MA_ASSERT(openMode != 0); + MA_ASSERT(pFile != NULL); + + (void)pVFS; + + if ((openMode & MA_OPEN_MODE_READ) != 0) { + if ((openMode & MA_OPEN_MODE_WRITE) != 0) { + pOpenModeStr = "r+"; + } else { + pOpenModeStr = "rb"; + } + } else { + pOpenModeStr = "wb"; + } + + result = ma_fopen(&pFileStd, pFilePath, pOpenModeStr); + if (result != MA_SUCCESS) { + return result; + } + + *pFile = pFileStd; + + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_open_w__stdio(ma_vfs* pVFS, const wchar_t* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + ma_result result; + FILE* pFileStd; + const wchar_t* pOpenModeStr; + + MA_ASSERT(pFilePath != NULL); + MA_ASSERT(openMode != 0); + MA_ASSERT(pFile != NULL); + + (void)pVFS; + + if ((openMode & MA_OPEN_MODE_READ) != 0) { + if ((openMode & MA_OPEN_MODE_WRITE) != 0) { + pOpenModeStr = L"r+"; + } else { + pOpenModeStr = L"rb"; + } + } else { + pOpenModeStr = L"wb"; + } + + result = ma_wfopen(&pFileStd, pFilePath, pOpenModeStr, (pVFS != NULL) ? &((ma_default_vfs*)pVFS)->allocationCallbacks : NULL); + if (result != MA_SUCCESS) { + return result; + } + + *pFile = pFileStd; + + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_close__stdio(ma_vfs* pVFS, ma_vfs_file file) +{ + MA_ASSERT(file != NULL); + + (void)pVFS; + + fclose((FILE*)file); + + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_read__stdio(ma_vfs* pVFS, ma_vfs_file file, void* pDst, size_t sizeInBytes, size_t* pBytesRead) +{ + size_t result; + + MA_ASSERT(file != NULL); + MA_ASSERT(pDst != NULL); + + (void)pVFS; + + result = fread(pDst, 1, sizeInBytes, (FILE*)file); + + if (pBytesRead != NULL) { + *pBytesRead = result; + } + + if (result != sizeInBytes) { + if (result == 0 && feof((FILE*)file)) { + return MA_AT_END; + } else { + return ma_result_from_errno(ferror((FILE*)file)); + } + } + + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_write__stdio(ma_vfs* pVFS, ma_vfs_file file, const void* pSrc, size_t sizeInBytes, size_t* pBytesWritten) +{ + size_t result; + + MA_ASSERT(file != NULL); + MA_ASSERT(pSrc != NULL); + + (void)pVFS; + + result = fwrite(pSrc, 1, sizeInBytes, (FILE*)file); + + if (pBytesWritten != NULL) { + *pBytesWritten = result; + } + + if (result != sizeInBytes) { + return ma_result_from_errno(ferror((FILE*)file)); + } + + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_seek__stdio(ma_vfs* pVFS, ma_vfs_file file, ma_int64 offset, ma_seek_origin origin) +{ + int result; + int whence; + + MA_ASSERT(file != NULL); + + (void)pVFS; + + if (origin == ma_seek_origin_start) { + whence = SEEK_SET; + } else if (origin == ma_seek_origin_end) { + whence = SEEK_END; + } else { + whence = SEEK_CUR; + } + +#if defined(_WIN32) + #if defined(_MSC_VER) && _MSC_VER > 1200 + result = _fseeki64((FILE*)file, offset, whence); + #else + /* No _fseeki64() so restrict to 31 bits. */ + if (offset > 0x7FFFFFFF) { + return MA_OUT_OF_RANGE; + } + + result = fseek((FILE*)file, (int)offset, whence); + #endif +#else + result = fseek((FILE*)file, (long int)offset, whence); +#endif + if (result != 0) { + return MA_ERROR; + } + + return MA_SUCCESS; +} + +static ma_result ma_default_vfs_tell__stdio(ma_vfs* pVFS, ma_vfs_file file, ma_int64* pCursor) +{ + ma_int64 result; + + MA_ASSERT(file != NULL); + MA_ASSERT(pCursor != NULL); + + (void)pVFS; + +#if defined(_WIN32) + #if defined(_MSC_VER) && _MSC_VER > 1200 + result = _ftelli64((FILE*)file); + #else + result = ftell((FILE*)file); + #endif +#else + result = ftell((FILE*)file); +#endif + + *pCursor = result; + + return MA_SUCCESS; +} + +#if !defined(_MSC_VER) && !((defined(_POSIX_C_SOURCE) && _POSIX_C_SOURCE >= 1) || defined(_XOPEN_SOURCE) || defined(_POSIX_SOURCE)) && !defined(MA_BSD) +int fileno(FILE *stream); +#endif + +static ma_result ma_default_vfs_info__stdio(ma_vfs* pVFS, ma_vfs_file file, ma_file_info* pInfo) +{ + int fd; + struct stat info; + + MA_ASSERT(file != NULL); + MA_ASSERT(pInfo != NULL); + + (void)pVFS; + +#if defined(_MSC_VER) + fd = _fileno((FILE*)file); +#else + fd = fileno((FILE*)file); +#endif + + if (fstat(fd, &info) != 0) { + return ma_result_from_errno(errno); + } + + pInfo->sizeInBytes = info.st_size; + + return MA_SUCCESS; +} +#endif + + +static ma_result ma_default_vfs_open(ma_vfs* pVFS, const char* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + if (pFile == NULL) { + return MA_INVALID_ARGS; + } + + *pFile = NULL; + + if (pFilePath == NULL || openMode == 0) { + return MA_INVALID_ARGS; + } + +#if defined(MA_USE_WIN32_FILEIO) + return ma_default_vfs_open__win32(pVFS, pFilePath, openMode, pFile); +#else + return ma_default_vfs_open__stdio(pVFS, pFilePath, openMode, pFile); +#endif +} + +static ma_result ma_default_vfs_open_w(ma_vfs* pVFS, const wchar_t* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + if (pFile == NULL) { + return MA_INVALID_ARGS; + } + + *pFile = NULL; + + if (pFilePath == NULL || openMode == 0) { + return MA_INVALID_ARGS; + } + +#if defined(MA_USE_WIN32_FILEIO) + return ma_default_vfs_open_w__win32(pVFS, pFilePath, openMode, pFile); +#else + return ma_default_vfs_open_w__stdio(pVFS, pFilePath, openMode, pFile); +#endif +} + +static ma_result ma_default_vfs_close(ma_vfs* pVFS, ma_vfs_file file) +{ + if (file == NULL) { + return MA_INVALID_ARGS; + } + +#if defined(MA_USE_WIN32_FILEIO) + return ma_default_vfs_close__win32(pVFS, file); +#else + return ma_default_vfs_close__stdio(pVFS, file); +#endif +} + +static ma_result ma_default_vfs_read(ma_vfs* pVFS, ma_vfs_file file, void* pDst, size_t sizeInBytes, size_t* pBytesRead) +{ + if (pBytesRead != NULL) { + *pBytesRead = 0; + } + + if (file == NULL || pDst == NULL) { + return MA_INVALID_ARGS; + } + +#if defined(MA_USE_WIN32_FILEIO) + return ma_default_vfs_read__win32(pVFS, file, pDst, sizeInBytes, pBytesRead); +#else + return ma_default_vfs_read__stdio(pVFS, file, pDst, sizeInBytes, pBytesRead); +#endif +} + +static ma_result ma_default_vfs_write(ma_vfs* pVFS, ma_vfs_file file, const void* pSrc, size_t sizeInBytes, size_t* pBytesWritten) +{ + if (pBytesWritten != NULL) { + *pBytesWritten = 0; + } + + if (file == NULL || pSrc == NULL) { + return MA_INVALID_ARGS; + } + +#if defined(MA_USE_WIN32_FILEIO) + return ma_default_vfs_write__win32(pVFS, file, pSrc, sizeInBytes, pBytesWritten); +#else + return ma_default_vfs_write__stdio(pVFS, file, pSrc, sizeInBytes, pBytesWritten); +#endif +} + +static ma_result ma_default_vfs_seek(ma_vfs* pVFS, ma_vfs_file file, ma_int64 offset, ma_seek_origin origin) +{ + if (file == NULL) { + return MA_INVALID_ARGS; + } + +#if defined(MA_USE_WIN32_FILEIO) + return ma_default_vfs_seek__win32(pVFS, file, offset, origin); +#else + return ma_default_vfs_seek__stdio(pVFS, file, offset, origin); +#endif +} + +static ma_result ma_default_vfs_tell(ma_vfs* pVFS, ma_vfs_file file, ma_int64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; + + if (file == NULL) { + return MA_INVALID_ARGS; + } + +#if defined(MA_USE_WIN32_FILEIO) + return ma_default_vfs_tell__win32(pVFS, file, pCursor); +#else + return ma_default_vfs_tell__stdio(pVFS, file, pCursor); +#endif +} + +static ma_result ma_default_vfs_info(ma_vfs* pVFS, ma_vfs_file file, ma_file_info* pInfo) +{ + if (pInfo == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pInfo); + + if (file == NULL) { + return MA_INVALID_ARGS; + } + +#if defined(MA_USE_WIN32_FILEIO) + return ma_default_vfs_info__win32(pVFS, file, pInfo); +#else + return ma_default_vfs_info__stdio(pVFS, file, pInfo); +#endif +} + + +MA_API ma_result ma_default_vfs_init(ma_default_vfs* pVFS, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pVFS == NULL) { + return MA_INVALID_ARGS; + } + + pVFS->cb.onOpen = ma_default_vfs_open; + pVFS->cb.onOpenW = ma_default_vfs_open_w; + pVFS->cb.onClose = ma_default_vfs_close; + pVFS->cb.onRead = ma_default_vfs_read; + pVFS->cb.onWrite = ma_default_vfs_write; + pVFS->cb.onSeek = ma_default_vfs_seek; + pVFS->cb.onTell = ma_default_vfs_tell; + pVFS->cb.onInfo = ma_default_vfs_info; + ma_allocation_callbacks_init_copy(&pVFS->allocationCallbacks, pAllocationCallbacks); + + return MA_SUCCESS; +} + + +MA_API ma_result ma_vfs_or_default_open(ma_vfs* pVFS, const char* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + if (pVFS != NULL) { + return ma_vfs_open(pVFS, pFilePath, openMode, pFile); + } else { + return ma_default_vfs_open(pVFS, pFilePath, openMode, pFile); + } +} + +MA_API ma_result ma_vfs_or_default_open_w(ma_vfs* pVFS, const wchar_t* pFilePath, ma_uint32 openMode, ma_vfs_file* pFile) +{ + if (pVFS != NULL) { + return ma_vfs_open_w(pVFS, pFilePath, openMode, pFile); + } else { + return ma_default_vfs_open_w(pVFS, pFilePath, openMode, pFile); + } +} + +MA_API ma_result ma_vfs_or_default_close(ma_vfs* pVFS, ma_vfs_file file) +{ + if (pVFS != NULL) { + return ma_vfs_close(pVFS, file); + } else { + return ma_default_vfs_close(pVFS, file); + } +} + +MA_API ma_result ma_vfs_or_default_read(ma_vfs* pVFS, ma_vfs_file file, void* pDst, size_t sizeInBytes, size_t* pBytesRead) +{ + if (pVFS != NULL) { + return ma_vfs_read(pVFS, file, pDst, sizeInBytes, pBytesRead); + } else { + return ma_default_vfs_read(pVFS, file, pDst, sizeInBytes, pBytesRead); + } +} + +MA_API ma_result ma_vfs_or_default_write(ma_vfs* pVFS, ma_vfs_file file, const void* pSrc, size_t sizeInBytes, size_t* pBytesWritten) +{ + if (pVFS != NULL) { + return ma_vfs_write(pVFS, file, pSrc, sizeInBytes, pBytesWritten); + } else { + return ma_default_vfs_write(pVFS, file, pSrc, sizeInBytes, pBytesWritten); + } +} + +MA_API ma_result ma_vfs_or_default_seek(ma_vfs* pVFS, ma_vfs_file file, ma_int64 offset, ma_seek_origin origin) +{ + if (pVFS != NULL) { + return ma_vfs_seek(pVFS, file, offset, origin); + } else { + return ma_default_vfs_seek(pVFS, file, offset, origin); + } +} + +MA_API ma_result ma_vfs_or_default_tell(ma_vfs* pVFS, ma_vfs_file file, ma_int64* pCursor) +{ + if (pVFS != NULL) { + return ma_vfs_tell(pVFS, file, pCursor); + } else { + return ma_default_vfs_tell(pVFS, file, pCursor); + } +} + +MA_API ma_result ma_vfs_or_default_info(ma_vfs* pVFS, ma_vfs_file file, ma_file_info* pInfo) +{ + if (pVFS != NULL) { + return ma_vfs_info(pVFS, file, pInfo); + } else { + return ma_default_vfs_info(pVFS, file, pInfo); + } +} + + + +static ma_result ma_vfs_open_and_read_file_ex(ma_vfs* pVFS, const char* pFilePath, const wchar_t* pFilePathW, void** ppData, size_t* pSize, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_result result; + ma_vfs_file file; + ma_file_info info; + void* pData; + size_t bytesRead; + + if (ppData != NULL) { + *ppData = NULL; + } + if (pSize != NULL) { + *pSize = 0; + } + + if (ppData == NULL) { + return MA_INVALID_ARGS; + } + + if (pFilePath != NULL) { + result = ma_vfs_or_default_open(pVFS, pFilePath, MA_OPEN_MODE_READ, &file); + } else { + result = ma_vfs_or_default_open_w(pVFS, pFilePathW, MA_OPEN_MODE_READ, &file); + } + if (result != MA_SUCCESS) { + return result; + } + + result = ma_vfs_or_default_info(pVFS, file, &info); + if (result != MA_SUCCESS) { + ma_vfs_or_default_close(pVFS, file); + return result; + } + + if (info.sizeInBytes > MA_SIZE_MAX) { + ma_vfs_or_default_close(pVFS, file); + return MA_TOO_BIG; + } + + pData = ma_malloc((size_t)info.sizeInBytes, pAllocationCallbacks); /* Safe cast. */ + if (pData == NULL) { + ma_vfs_or_default_close(pVFS, file); + return result; + } + + result = ma_vfs_or_default_read(pVFS, file, pData, (size_t)info.sizeInBytes, &bytesRead); /* Safe cast. */ + ma_vfs_or_default_close(pVFS, file); + + if (result != MA_SUCCESS) { + ma_free(pData, pAllocationCallbacks); + return result; + } + + if (pSize != NULL) { + *pSize = bytesRead; + } + + MA_ASSERT(ppData != NULL); + *ppData = pData; + + return MA_SUCCESS; +} + +MA_API ma_result ma_vfs_open_and_read_file(ma_vfs* pVFS, const char* pFilePath, void** ppData, size_t* pSize, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_vfs_open_and_read_file_ex(pVFS, pFilePath, NULL, ppData, pSize, pAllocationCallbacks); +} + +MA_API ma_result ma_vfs_open_and_read_file_w(ma_vfs* pVFS, const wchar_t* pFilePath, void** ppData, size_t* pSize, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_vfs_open_and_read_file_ex(pVFS, NULL, pFilePath, ppData, pSize, pAllocationCallbacks); +} + + + +/************************************************************************************************************************************************************** + +Decoding and Encoding Headers. These are auto-generated from a tool. + +**************************************************************************************************************************************************************/ +#if !defined(MA_NO_WAV) && (!defined(MA_NO_DECODING) || !defined(MA_NO_ENCODING)) +/* dr_wav_h begin */ +#ifndef ma_dr_wav_h +#define ma_dr_wav_h +#ifdef __cplusplus +extern "C" { +#endif +#define MA_DR_WAV_STRINGIFY(x) #x +#define MA_DR_WAV_XSTRINGIFY(x) MA_DR_WAV_STRINGIFY(x) +#define MA_DR_WAV_VERSION_MAJOR 0 +#define MA_DR_WAV_VERSION_MINOR 13 +#define MA_DR_WAV_VERSION_REVISION 18 +#define MA_DR_WAV_VERSION_STRING MA_DR_WAV_XSTRINGIFY(MA_DR_WAV_VERSION_MAJOR) "." MA_DR_WAV_XSTRINGIFY(MA_DR_WAV_VERSION_MINOR) "." MA_DR_WAV_XSTRINGIFY(MA_DR_WAV_VERSION_REVISION) +#include +#define MA_DR_WAVE_FORMAT_PCM 0x1 +#define MA_DR_WAVE_FORMAT_ADPCM 0x2 +#define MA_DR_WAVE_FORMAT_IEEE_FLOAT 0x3 +#define MA_DR_WAVE_FORMAT_ALAW 0x6 +#define MA_DR_WAVE_FORMAT_MULAW 0x7 +#define MA_DR_WAVE_FORMAT_DVI_ADPCM 0x11 +#define MA_DR_WAVE_FORMAT_EXTENSIBLE 0xFFFE +#define MA_DR_WAV_SEQUENTIAL 0x00000001 +#define MA_DR_WAV_WITH_METADATA 0x00000002 +MA_API void ma_dr_wav_version(ma_uint32* pMajor, ma_uint32* pMinor, ma_uint32* pRevision); +MA_API const char* ma_dr_wav_version_string(void); +typedef enum +{ + ma_dr_wav_seek_origin_start, + ma_dr_wav_seek_origin_current +} ma_dr_wav_seek_origin; +typedef enum +{ + ma_dr_wav_container_riff, + ma_dr_wav_container_rifx, + ma_dr_wav_container_w64, + ma_dr_wav_container_rf64, + ma_dr_wav_container_aiff +} ma_dr_wav_container; +typedef struct +{ + union + { + ma_uint8 fourcc[4]; + ma_uint8 guid[16]; + } id; + ma_uint64 sizeInBytes; + unsigned int paddingSize; +} ma_dr_wav_chunk_header; +typedef struct +{ + ma_uint16 formatTag; + ma_uint16 channels; + ma_uint32 sampleRate; + ma_uint32 avgBytesPerSec; + ma_uint16 blockAlign; + ma_uint16 bitsPerSample; + ma_uint16 extendedSize; + ma_uint16 validBitsPerSample; + ma_uint32 channelMask; + ma_uint8 subFormat[16]; +} ma_dr_wav_fmt; +MA_API ma_uint16 ma_dr_wav_fmt_get_format(const ma_dr_wav_fmt* pFMT); +typedef size_t (* ma_dr_wav_read_proc)(void* pUserData, void* pBufferOut, size_t bytesToRead); +typedef size_t (* ma_dr_wav_write_proc)(void* pUserData, const void* pData, size_t bytesToWrite); +typedef ma_bool32 (* ma_dr_wav_seek_proc)(void* pUserData, int offset, ma_dr_wav_seek_origin origin); +typedef ma_uint64 (* ma_dr_wav_chunk_proc)(void* pChunkUserData, ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pReadSeekUserData, const ma_dr_wav_chunk_header* pChunkHeader, ma_dr_wav_container container, const ma_dr_wav_fmt* pFMT); +typedef struct +{ + const ma_uint8* data; + size_t dataSize; + size_t currentReadPos; +} ma_dr_wav__memory_stream; +typedef struct +{ + void** ppData; + size_t* pDataSize; + size_t dataSize; + size_t dataCapacity; + size_t currentWritePos; +} ma_dr_wav__memory_stream_write; +typedef struct +{ + ma_dr_wav_container container; + ma_uint32 format; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_uint32 bitsPerSample; +} ma_dr_wav_data_format; +typedef enum +{ + ma_dr_wav_metadata_type_none = 0, + ma_dr_wav_metadata_type_unknown = 1 << 0, + ma_dr_wav_metadata_type_smpl = 1 << 1, + ma_dr_wav_metadata_type_inst = 1 << 2, + ma_dr_wav_metadata_type_cue = 1 << 3, + ma_dr_wav_metadata_type_acid = 1 << 4, + ma_dr_wav_metadata_type_bext = 1 << 5, + ma_dr_wav_metadata_type_list_label = 1 << 6, + ma_dr_wav_metadata_type_list_note = 1 << 7, + ma_dr_wav_metadata_type_list_labelled_cue_region = 1 << 8, + ma_dr_wav_metadata_type_list_info_software = 1 << 9, + ma_dr_wav_metadata_type_list_info_copyright = 1 << 10, + ma_dr_wav_metadata_type_list_info_title = 1 << 11, + ma_dr_wav_metadata_type_list_info_artist = 1 << 12, + ma_dr_wav_metadata_type_list_info_comment = 1 << 13, + ma_dr_wav_metadata_type_list_info_date = 1 << 14, + ma_dr_wav_metadata_type_list_info_genre = 1 << 15, + ma_dr_wav_metadata_type_list_info_album = 1 << 16, + ma_dr_wav_metadata_type_list_info_tracknumber = 1 << 17, + ma_dr_wav_metadata_type_list_all_info_strings = ma_dr_wav_metadata_type_list_info_software + | ma_dr_wav_metadata_type_list_info_copyright + | ma_dr_wav_metadata_type_list_info_title + | ma_dr_wav_metadata_type_list_info_artist + | ma_dr_wav_metadata_type_list_info_comment + | ma_dr_wav_metadata_type_list_info_date + | ma_dr_wav_metadata_type_list_info_genre + | ma_dr_wav_metadata_type_list_info_album + | ma_dr_wav_metadata_type_list_info_tracknumber, + ma_dr_wav_metadata_type_list_all_adtl = ma_dr_wav_metadata_type_list_label + | ma_dr_wav_metadata_type_list_note + | ma_dr_wav_metadata_type_list_labelled_cue_region, + ma_dr_wav_metadata_type_all = -2, + ma_dr_wav_metadata_type_all_including_unknown = -1 +} ma_dr_wav_metadata_type; +typedef enum +{ + ma_dr_wav_smpl_loop_type_forward = 0, + ma_dr_wav_smpl_loop_type_pingpong = 1, + ma_dr_wav_smpl_loop_type_backward = 2 +} ma_dr_wav_smpl_loop_type; +typedef struct +{ + ma_uint32 cuePointId; + ma_uint32 type; + ma_uint32 firstSampleByteOffset; + ma_uint32 lastSampleByteOffset; + ma_uint32 sampleFraction; + ma_uint32 playCount; +} ma_dr_wav_smpl_loop; +typedef struct +{ + ma_uint32 manufacturerId; + ma_uint32 productId; + ma_uint32 samplePeriodNanoseconds; + ma_uint32 midiUnityNote; + ma_uint32 midiPitchFraction; + ma_uint32 smpteFormat; + ma_uint32 smpteOffset; + ma_uint32 sampleLoopCount; + ma_uint32 samplerSpecificDataSizeInBytes; + ma_dr_wav_smpl_loop* pLoops; + ma_uint8* pSamplerSpecificData; +} ma_dr_wav_smpl; +typedef struct +{ + ma_int8 midiUnityNote; + ma_int8 fineTuneCents; + ma_int8 gainDecibels; + ma_int8 lowNote; + ma_int8 highNote; + ma_int8 lowVelocity; + ma_int8 highVelocity; +} ma_dr_wav_inst; +typedef struct +{ + ma_uint32 id; + ma_uint32 playOrderPosition; + ma_uint8 dataChunkId[4]; + ma_uint32 chunkStart; + ma_uint32 blockStart; + ma_uint32 sampleByteOffset; +} ma_dr_wav_cue_point; +typedef struct +{ + ma_uint32 cuePointCount; + ma_dr_wav_cue_point *pCuePoints; +} ma_dr_wav_cue; +typedef enum +{ + ma_dr_wav_acid_flag_one_shot = 1, + ma_dr_wav_acid_flag_root_note_set = 2, + ma_dr_wav_acid_flag_stretch = 4, + ma_dr_wav_acid_flag_disk_based = 8, + ma_dr_wav_acid_flag_acidizer = 16 +} ma_dr_wav_acid_flag; +typedef struct +{ + ma_uint32 flags; + ma_uint16 midiUnityNote; + ma_uint16 reserved1; + float reserved2; + ma_uint32 numBeats; + ma_uint16 meterDenominator; + ma_uint16 meterNumerator; + float tempo; +} ma_dr_wav_acid; +typedef struct +{ + ma_uint32 cuePointId; + ma_uint32 stringLength; + char* pString; +} ma_dr_wav_list_label_or_note; +typedef struct +{ + char* pDescription; + char* pOriginatorName; + char* pOriginatorReference; + char pOriginationDate[10]; + char pOriginationTime[8]; + ma_uint64 timeReference; + ma_uint16 version; + char* pCodingHistory; + ma_uint32 codingHistorySize; + ma_uint8* pUMID; + ma_uint16 loudnessValue; + ma_uint16 loudnessRange; + ma_uint16 maxTruePeakLevel; + ma_uint16 maxMomentaryLoudness; + ma_uint16 maxShortTermLoudness; +} ma_dr_wav_bext; +typedef struct +{ + ma_uint32 stringLength; + char* pString; +} ma_dr_wav_list_info_text; +typedef struct +{ + ma_uint32 cuePointId; + ma_uint32 sampleLength; + ma_uint8 purposeId[4]; + ma_uint16 country; + ma_uint16 language; + ma_uint16 dialect; + ma_uint16 codePage; + ma_uint32 stringLength; + char* pString; +} ma_dr_wav_list_labelled_cue_region; +typedef enum +{ + ma_dr_wav_metadata_location_invalid, + ma_dr_wav_metadata_location_top_level, + ma_dr_wav_metadata_location_inside_info_list, + ma_dr_wav_metadata_location_inside_adtl_list +} ma_dr_wav_metadata_location; +typedef struct +{ + ma_uint8 id[4]; + ma_dr_wav_metadata_location chunkLocation; + ma_uint32 dataSizeInBytes; + ma_uint8* pData; +} ma_dr_wav_unknown_metadata; +typedef struct +{ + ma_dr_wav_metadata_type type; + union + { + ma_dr_wav_cue cue; + ma_dr_wav_smpl smpl; + ma_dr_wav_acid acid; + ma_dr_wav_inst inst; + ma_dr_wav_bext bext; + ma_dr_wav_list_label_or_note labelOrNote; + ma_dr_wav_list_labelled_cue_region labelledCueRegion; + ma_dr_wav_list_info_text infoText; + ma_dr_wav_unknown_metadata unknown; + } data; +} ma_dr_wav_metadata; +typedef struct +{ + ma_dr_wav_read_proc onRead; + ma_dr_wav_write_proc onWrite; + ma_dr_wav_seek_proc onSeek; + void* pUserData; + ma_allocation_callbacks allocationCallbacks; + ma_dr_wav_container container; + ma_dr_wav_fmt fmt; + ma_uint32 sampleRate; + ma_uint16 channels; + ma_uint16 bitsPerSample; + ma_uint16 translatedFormatTag; + ma_uint64 totalPCMFrameCount; + ma_uint64 dataChunkDataSize; + ma_uint64 dataChunkDataPos; + ma_uint64 bytesRemaining; + ma_uint64 readCursorInPCMFrames; + ma_uint64 dataChunkDataSizeTargetWrite; + ma_bool32 isSequentialWrite; + ma_dr_wav_metadata* pMetadata; + ma_uint32 metadataCount; + ma_dr_wav__memory_stream memoryStream; + ma_dr_wav__memory_stream_write memoryStreamWrite; + struct + { + ma_uint32 bytesRemainingInBlock; + ma_uint16 predictor[2]; + ma_int32 delta[2]; + ma_int32 cachedFrames[4]; + ma_uint32 cachedFrameCount; + ma_int32 prevFrames[2][2]; + } msadpcm; + struct + { + ma_uint32 bytesRemainingInBlock; + ma_int32 predictor[2]; + ma_int32 stepIndex[2]; + ma_int32 cachedFrames[16]; + ma_uint32 cachedFrameCount; + } ima; + struct + { + ma_bool8 isLE; + ma_bool8 isUnsigned; + } aiff; +} ma_dr_wav; +MA_API ma_bool32 ma_dr_wav_init(ma_dr_wav* pWav, ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_ex(ma_dr_wav* pWav, ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, ma_dr_wav_chunk_proc onChunk, void* pReadSeekUserData, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_with_metadata(ma_dr_wav* pWav, ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_write(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_dr_wav_write_proc onWrite, ma_dr_wav_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_write_sequential(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, ma_dr_wav_write_proc onWrite, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_write_sequential_pcm_frames(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_uint64 totalPCMFrameCount, ma_dr_wav_write_proc onWrite, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_write_with_metadata(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_dr_wav_write_proc onWrite, ma_dr_wav_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks, ma_dr_wav_metadata* pMetadata, ma_uint32 metadataCount); +MA_API ma_uint64 ma_dr_wav_target_write_size_bytes(const ma_dr_wav_data_format* pFormat, ma_uint64 totalFrameCount, ma_dr_wav_metadata* pMetadata, ma_uint32 metadataCount); +MA_API ma_dr_wav_metadata* ma_dr_wav_take_ownership_of_metadata(ma_dr_wav* pWav); +MA_API ma_result ma_dr_wav_uninit(ma_dr_wav* pWav); +MA_API size_t ma_dr_wav_read_raw(ma_dr_wav* pWav, size_t bytesToRead, void* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames(ma_dr_wav* pWav, ma_uint64 framesToRead, void* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_le(ma_dr_wav* pWav, ma_uint64 framesToRead, void* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_be(ma_dr_wav* pWav, ma_uint64 framesToRead, void* pBufferOut); +MA_API ma_bool32 ma_dr_wav_seek_to_pcm_frame(ma_dr_wav* pWav, ma_uint64 targetFrameIndex); +MA_API ma_result ma_dr_wav_get_cursor_in_pcm_frames(ma_dr_wav* pWav, ma_uint64* pCursor); +MA_API ma_result ma_dr_wav_get_length_in_pcm_frames(ma_dr_wav* pWav, ma_uint64* pLength); +MA_API size_t ma_dr_wav_write_raw(ma_dr_wav* pWav, size_t bytesToWrite, const void* pData); +MA_API ma_uint64 ma_dr_wav_write_pcm_frames(ma_dr_wav* pWav, ma_uint64 framesToWrite, const void* pData); +MA_API ma_uint64 ma_dr_wav_write_pcm_frames_le(ma_dr_wav* pWav, ma_uint64 framesToWrite, const void* pData); +MA_API ma_uint64 ma_dr_wav_write_pcm_frames_be(ma_dr_wav* pWav, ma_uint64 framesToWrite, const void* pData); +#ifndef MA_DR_WAV_NO_CONVERSION_API +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s16(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s16le(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s16be(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut); +MA_API void ma_dr_wav_u8_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_s24_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_s32_to_s16(ma_int16* pOut, const ma_int32* pIn, size_t sampleCount); +MA_API void ma_dr_wav_f32_to_s16(ma_int16* pOut, const float* pIn, size_t sampleCount); +MA_API void ma_dr_wav_f64_to_s16(ma_int16* pOut, const double* pIn, size_t sampleCount); +MA_API void ma_dr_wav_alaw_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_mulaw_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_f32(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_f32le(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_f32be(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut); +MA_API void ma_dr_wav_u8_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_s16_to_f32(float* pOut, const ma_int16* pIn, size_t sampleCount); +MA_API void ma_dr_wav_s24_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_s32_to_f32(float* pOut, const ma_int32* pIn, size_t sampleCount); +MA_API void ma_dr_wav_f64_to_f32(float* pOut, const double* pIn, size_t sampleCount); +MA_API void ma_dr_wav_alaw_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_mulaw_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s32(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s32le(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut); +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s32be(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut); +MA_API void ma_dr_wav_u8_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_s16_to_s32(ma_int32* pOut, const ma_int16* pIn, size_t sampleCount); +MA_API void ma_dr_wav_s24_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_f32_to_s32(ma_int32* pOut, const float* pIn, size_t sampleCount); +MA_API void ma_dr_wav_f64_to_s32(ma_int32* pOut, const double* pIn, size_t sampleCount); +MA_API void ma_dr_wav_alaw_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t sampleCount); +MA_API void ma_dr_wav_mulaw_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t sampleCount); +#endif +#ifndef MA_DR_WAV_NO_STDIO +MA_API ma_bool32 ma_dr_wav_init_file(ma_dr_wav* pWav, const char* filename, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_ex(ma_dr_wav* pWav, const char* filename, ma_dr_wav_chunk_proc onChunk, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_w(ma_dr_wav* pWav, const wchar_t* filename, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_ex_w(ma_dr_wav* pWav, const wchar_t* filename, ma_dr_wav_chunk_proc onChunk, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_with_metadata(ma_dr_wav* pWav, const char* filename, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_with_metadata_w(ma_dr_wav* pWav, const wchar_t* filename, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_write(ma_dr_wav* pWav, const char* filename, const ma_dr_wav_data_format* pFormat, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_write_sequential(ma_dr_wav* pWav, const char* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_write_sequential_pcm_frames(ma_dr_wav* pWav, const char* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_write_w(ma_dr_wav* pWav, const wchar_t* filename, const ma_dr_wav_data_format* pFormat, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_write_sequential_w(ma_dr_wav* pWav, const wchar_t* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_file_write_sequential_pcm_frames_w(ma_dr_wav* pWav, const wchar_t* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +#endif +MA_API ma_bool32 ma_dr_wav_init_memory(ma_dr_wav* pWav, const void* data, size_t dataSize, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_memory_ex(ma_dr_wav* pWav, const void* data, size_t dataSize, ma_dr_wav_chunk_proc onChunk, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_memory_with_metadata(ma_dr_wav* pWav, const void* data, size_t dataSize, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_memory_write(ma_dr_wav* pWav, void** ppData, size_t* pDataSize, const ma_dr_wav_data_format* pFormat, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_memory_write_sequential(ma_dr_wav* pWav, void** ppData, size_t* pDataSize, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_wav_init_memory_write_sequential_pcm_frames(ma_dr_wav* pWav, void** ppData, size_t* pDataSize, const ma_dr_wav_data_format* pFormat, ma_uint64 totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +#ifndef MA_DR_WAV_NO_CONVERSION_API +MA_API ma_int16* ma_dr_wav_open_and_read_pcm_frames_s16(ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API float* ma_dr_wav_open_and_read_pcm_frames_f32(ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int32* ma_dr_wav_open_and_read_pcm_frames_s32(ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +#ifndef MA_DR_WAV_NO_STDIO +MA_API ma_int16* ma_dr_wav_open_file_and_read_pcm_frames_s16(const char* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API float* ma_dr_wav_open_file_and_read_pcm_frames_f32(const char* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int32* ma_dr_wav_open_file_and_read_pcm_frames_s32(const char* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int16* ma_dr_wav_open_file_and_read_pcm_frames_s16_w(const wchar_t* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API float* ma_dr_wav_open_file_and_read_pcm_frames_f32_w(const wchar_t* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int32* ma_dr_wav_open_file_and_read_pcm_frames_s32_w(const wchar_t* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +#endif +MA_API ma_int16* ma_dr_wav_open_memory_and_read_pcm_frames_s16(const void* data, size_t dataSize, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API float* ma_dr_wav_open_memory_and_read_pcm_frames_f32(const void* data, size_t dataSize, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int32* ma_dr_wav_open_memory_and_read_pcm_frames_s32(const void* data, size_t dataSize, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks); +#endif +MA_API void ma_dr_wav_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_uint16 ma_dr_wav_bytes_to_u16(const ma_uint8* data); +MA_API ma_int16 ma_dr_wav_bytes_to_s16(const ma_uint8* data); +MA_API ma_uint32 ma_dr_wav_bytes_to_u32(const ma_uint8* data); +MA_API ma_int32 ma_dr_wav_bytes_to_s32(const ma_uint8* data); +MA_API ma_uint64 ma_dr_wav_bytes_to_u64(const ma_uint8* data); +MA_API ma_int64 ma_dr_wav_bytes_to_s64(const ma_uint8* data); +MA_API float ma_dr_wav_bytes_to_f32(const ma_uint8* data); +MA_API ma_bool32 ma_dr_wav_guid_equal(const ma_uint8 a[16], const ma_uint8 b[16]); +MA_API ma_bool32 ma_dr_wav_fourcc_equal(const ma_uint8* a, const char* b); +#ifdef __cplusplus +} +#endif +#endif +/* dr_wav_h end */ +#endif /* MA_NO_WAV */ + +#if !defined(MA_NO_FLAC) && !defined(MA_NO_DECODING) +/* dr_flac_h begin */ +#ifndef ma_dr_flac_h +#define ma_dr_flac_h +#ifdef __cplusplus +extern "C" { +#endif +#define MA_DR_FLAC_STRINGIFY(x) #x +#define MA_DR_FLAC_XSTRINGIFY(x) MA_DR_FLAC_STRINGIFY(x) +#define MA_DR_FLAC_VERSION_MAJOR 0 +#define MA_DR_FLAC_VERSION_MINOR 12 +#define MA_DR_FLAC_VERSION_REVISION 43 +#define MA_DR_FLAC_VERSION_STRING MA_DR_FLAC_XSTRINGIFY(MA_DR_FLAC_VERSION_MAJOR) "." MA_DR_FLAC_XSTRINGIFY(MA_DR_FLAC_VERSION_MINOR) "." MA_DR_FLAC_XSTRINGIFY(MA_DR_FLAC_VERSION_REVISION) +#include +#if defined(_MSC_VER) && _MSC_VER >= 1700 + #define MA_DR_FLAC_DEPRECATED __declspec(deprecated) +#elif (defined(__GNUC__) && __GNUC__ >= 4) + #define MA_DR_FLAC_DEPRECATED __attribute__((deprecated)) +#elif defined(__has_feature) + #if __has_feature(attribute_deprecated) + #define MA_DR_FLAC_DEPRECATED __attribute__((deprecated)) + #else + #define MA_DR_FLAC_DEPRECATED + #endif +#else + #define MA_DR_FLAC_DEPRECATED +#endif +MA_API void ma_dr_flac_version(ma_uint32* pMajor, ma_uint32* pMinor, ma_uint32* pRevision); +MA_API const char* ma_dr_flac_version_string(void); +#ifndef MA_DR_FLAC_BUFFER_SIZE +#define MA_DR_FLAC_BUFFER_SIZE 4096 +#endif +#ifdef MA_64BIT +typedef ma_uint64 ma_dr_flac_cache_t; +#else +typedef ma_uint32 ma_dr_flac_cache_t; +#endif +#define MA_DR_FLAC_METADATA_BLOCK_TYPE_STREAMINFO 0 +#define MA_DR_FLAC_METADATA_BLOCK_TYPE_PADDING 1 +#define MA_DR_FLAC_METADATA_BLOCK_TYPE_APPLICATION 2 +#define MA_DR_FLAC_METADATA_BLOCK_TYPE_SEEKTABLE 3 +#define MA_DR_FLAC_METADATA_BLOCK_TYPE_VORBIS_COMMENT 4 +#define MA_DR_FLAC_METADATA_BLOCK_TYPE_CUESHEET 5 +#define MA_DR_FLAC_METADATA_BLOCK_TYPE_PICTURE 6 +#define MA_DR_FLAC_METADATA_BLOCK_TYPE_INVALID 127 +#define MA_DR_FLAC_PICTURE_TYPE_OTHER 0 +#define MA_DR_FLAC_PICTURE_TYPE_FILE_ICON 1 +#define MA_DR_FLAC_PICTURE_TYPE_OTHER_FILE_ICON 2 +#define MA_DR_FLAC_PICTURE_TYPE_COVER_FRONT 3 +#define MA_DR_FLAC_PICTURE_TYPE_COVER_BACK 4 +#define MA_DR_FLAC_PICTURE_TYPE_LEAFLET_PAGE 5 +#define MA_DR_FLAC_PICTURE_TYPE_MEDIA 6 +#define MA_DR_FLAC_PICTURE_TYPE_LEAD_ARTIST 7 +#define MA_DR_FLAC_PICTURE_TYPE_ARTIST 8 +#define MA_DR_FLAC_PICTURE_TYPE_CONDUCTOR 9 +#define MA_DR_FLAC_PICTURE_TYPE_BAND 10 +#define MA_DR_FLAC_PICTURE_TYPE_COMPOSER 11 +#define MA_DR_FLAC_PICTURE_TYPE_LYRICIST 12 +#define MA_DR_FLAC_PICTURE_TYPE_RECORDING_LOCATION 13 +#define MA_DR_FLAC_PICTURE_TYPE_DURING_RECORDING 14 +#define MA_DR_FLAC_PICTURE_TYPE_DURING_PERFORMANCE 15 +#define MA_DR_FLAC_PICTURE_TYPE_SCREEN_CAPTURE 16 +#define MA_DR_FLAC_PICTURE_TYPE_BRIGHT_COLORED_FISH 17 +#define MA_DR_FLAC_PICTURE_TYPE_ILLUSTRATION 18 +#define MA_DR_FLAC_PICTURE_TYPE_BAND_LOGOTYPE 19 +#define MA_DR_FLAC_PICTURE_TYPE_PUBLISHER_LOGOTYPE 20 +typedef enum +{ + ma_dr_flac_container_native, + ma_dr_flac_container_ogg, + ma_dr_flac_container_unknown +} ma_dr_flac_container; +typedef enum +{ + ma_dr_flac_seek_origin_start, + ma_dr_flac_seek_origin_current +} ma_dr_flac_seek_origin; +typedef struct +{ + ma_uint64 firstPCMFrame; + ma_uint64 flacFrameOffset; + ma_uint16 pcmFrameCount; +} ma_dr_flac_seekpoint; +typedef struct +{ + ma_uint16 minBlockSizeInPCMFrames; + ma_uint16 maxBlockSizeInPCMFrames; + ma_uint32 minFrameSizeInPCMFrames; + ma_uint32 maxFrameSizeInPCMFrames; + ma_uint32 sampleRate; + ma_uint8 channels; + ma_uint8 bitsPerSample; + ma_uint64 totalPCMFrameCount; + ma_uint8 md5[16]; +} ma_dr_flac_streaminfo; +typedef struct +{ + ma_uint32 type; + const void* pRawData; + ma_uint32 rawDataSize; + union + { + ma_dr_flac_streaminfo streaminfo; + struct + { + int unused; + } padding; + struct + { + ma_uint32 id; + const void* pData; + ma_uint32 dataSize; + } application; + struct + { + ma_uint32 seekpointCount; + const ma_dr_flac_seekpoint* pSeekpoints; + } seektable; + struct + { + ma_uint32 vendorLength; + const char* vendor; + ma_uint32 commentCount; + const void* pComments; + } vorbis_comment; + struct + { + char catalog[128]; + ma_uint64 leadInSampleCount; + ma_bool32 isCD; + ma_uint8 trackCount; + const void* pTrackData; + } cuesheet; + struct + { + ma_uint32 type; + ma_uint32 mimeLength; + const char* mime; + ma_uint32 descriptionLength; + const char* description; + ma_uint32 width; + ma_uint32 height; + ma_uint32 colorDepth; + ma_uint32 indexColorCount; + ma_uint32 pictureDataSize; + const ma_uint8* pPictureData; + } picture; + } data; +} ma_dr_flac_metadata; +typedef size_t (* ma_dr_flac_read_proc)(void* pUserData, void* pBufferOut, size_t bytesToRead); +typedef ma_bool32 (* ma_dr_flac_seek_proc)(void* pUserData, int offset, ma_dr_flac_seek_origin origin); +typedef void (* ma_dr_flac_meta_proc)(void* pUserData, ma_dr_flac_metadata* pMetadata); +typedef struct +{ + const ma_uint8* data; + size_t dataSize; + size_t currentReadPos; +} ma_dr_flac__memory_stream; +typedef struct +{ + ma_dr_flac_read_proc onRead; + ma_dr_flac_seek_proc onSeek; + void* pUserData; + size_t unalignedByteCount; + ma_dr_flac_cache_t unalignedCache; + ma_uint32 nextL2Line; + ma_uint32 consumedBits; + ma_dr_flac_cache_t cacheL2[MA_DR_FLAC_BUFFER_SIZE/sizeof(ma_dr_flac_cache_t)]; + ma_dr_flac_cache_t cache; + ma_uint16 crc16; + ma_dr_flac_cache_t crc16Cache; + ma_uint32 crc16CacheIgnoredBytes; +} ma_dr_flac_bs; +typedef struct +{ + ma_uint8 subframeType; + ma_uint8 wastedBitsPerSample; + ma_uint8 lpcOrder; + ma_int32* pSamplesS32; +} ma_dr_flac_subframe; +typedef struct +{ + ma_uint64 pcmFrameNumber; + ma_uint32 flacFrameNumber; + ma_uint32 sampleRate; + ma_uint16 blockSizeInPCMFrames; + ma_uint8 channelAssignment; + ma_uint8 bitsPerSample; + ma_uint8 crc8; +} ma_dr_flac_frame_header; +typedef struct +{ + ma_dr_flac_frame_header header; + ma_uint32 pcmFramesRemaining; + ma_dr_flac_subframe subframes[8]; +} ma_dr_flac_frame; +typedef struct +{ + ma_dr_flac_meta_proc onMeta; + void* pUserDataMD; + ma_allocation_callbacks allocationCallbacks; + ma_uint32 sampleRate; + ma_uint8 channels; + ma_uint8 bitsPerSample; + ma_uint16 maxBlockSizeInPCMFrames; + ma_uint64 totalPCMFrameCount; + ma_dr_flac_container container; + ma_uint32 seekpointCount; + ma_dr_flac_frame currentFLACFrame; + ma_uint64 currentPCMFrame; + ma_uint64 firstFLACFramePosInBytes; + ma_dr_flac__memory_stream memoryStream; + ma_int32* pDecodedSamples; + ma_dr_flac_seekpoint* pSeekpoints; + void* _oggbs; + ma_bool32 _noSeekTableSeek : 1; + ma_bool32 _noBinarySearchSeek : 1; + ma_bool32 _noBruteForceSeek : 1; + ma_dr_flac_bs bs; + ma_uint8 pExtraData[1]; +} ma_dr_flac; +MA_API ma_dr_flac* ma_dr_flac_open(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_dr_flac* ma_dr_flac_open_relaxed(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_container container, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_dr_flac* ma_dr_flac_open_with_metadata(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_dr_flac* ma_dr_flac_open_with_metadata_relaxed(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, ma_dr_flac_container container, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API void ma_dr_flac_close(ma_dr_flac* pFlac); +MA_API ma_uint64 ma_dr_flac_read_pcm_frames_s32(ma_dr_flac* pFlac, ma_uint64 framesToRead, ma_int32* pBufferOut); +MA_API ma_uint64 ma_dr_flac_read_pcm_frames_s16(ma_dr_flac* pFlac, ma_uint64 framesToRead, ma_int16* pBufferOut); +MA_API ma_uint64 ma_dr_flac_read_pcm_frames_f32(ma_dr_flac* pFlac, ma_uint64 framesToRead, float* pBufferOut); +MA_API ma_bool32 ma_dr_flac_seek_to_pcm_frame(ma_dr_flac* pFlac, ma_uint64 pcmFrameIndex); +#ifndef MA_DR_FLAC_NO_STDIO +MA_API ma_dr_flac* ma_dr_flac_open_file(const char* pFileName, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_dr_flac* ma_dr_flac_open_file_w(const wchar_t* pFileName, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_dr_flac* ma_dr_flac_open_file_with_metadata(const char* pFileName, ma_dr_flac_meta_proc onMeta, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_dr_flac* ma_dr_flac_open_file_with_metadata_w(const wchar_t* pFileName, ma_dr_flac_meta_proc onMeta, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +#endif +MA_API ma_dr_flac* ma_dr_flac_open_memory(const void* pData, size_t dataSize, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_dr_flac* ma_dr_flac_open_memory_with_metadata(const void* pData, size_t dataSize, ma_dr_flac_meta_proc onMeta, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int32* ma_dr_flac_open_and_read_pcm_frames_s32(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, void* pUserData, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int16* ma_dr_flac_open_and_read_pcm_frames_s16(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, void* pUserData, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API float* ma_dr_flac_open_and_read_pcm_frames_f32(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, void* pUserData, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +#ifndef MA_DR_FLAC_NO_STDIO +MA_API ma_int32* ma_dr_flac_open_file_and_read_pcm_frames_s32(const char* filename, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int16* ma_dr_flac_open_file_and_read_pcm_frames_s16(const char* filename, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API float* ma_dr_flac_open_file_and_read_pcm_frames_f32(const char* filename, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +#endif +MA_API ma_int32* ma_dr_flac_open_memory_and_read_pcm_frames_s32(const void* data, size_t dataSize, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int16* ma_dr_flac_open_memory_and_read_pcm_frames_s16(const void* data, size_t dataSize, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API float* ma_dr_flac_open_memory_and_read_pcm_frames_f32(const void* data, size_t dataSize, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API void ma_dr_flac_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks); +typedef struct +{ + ma_uint32 countRemaining; + const char* pRunningData; +} ma_dr_flac_vorbis_comment_iterator; +MA_API void ma_dr_flac_init_vorbis_comment_iterator(ma_dr_flac_vorbis_comment_iterator* pIter, ma_uint32 commentCount, const void* pComments); +MA_API const char* ma_dr_flac_next_vorbis_comment(ma_dr_flac_vorbis_comment_iterator* pIter, ma_uint32* pCommentLengthOut); +typedef struct +{ + ma_uint32 countRemaining; + const char* pRunningData; +} ma_dr_flac_cuesheet_track_iterator; +typedef struct +{ + ma_uint64 offset; + ma_uint8 index; + ma_uint8 reserved[3]; +} ma_dr_flac_cuesheet_track_index; +typedef struct +{ + ma_uint64 offset; + ma_uint8 trackNumber; + char ISRC[12]; + ma_bool8 isAudio; + ma_bool8 preEmphasis; + ma_uint8 indexCount; + const ma_dr_flac_cuesheet_track_index* pIndexPoints; +} ma_dr_flac_cuesheet_track; +MA_API void ma_dr_flac_init_cuesheet_track_iterator(ma_dr_flac_cuesheet_track_iterator* pIter, ma_uint32 trackCount, const void* pTrackData); +MA_API ma_bool32 ma_dr_flac_next_cuesheet_track(ma_dr_flac_cuesheet_track_iterator* pIter, ma_dr_flac_cuesheet_track* pCuesheetTrack); +#ifdef __cplusplus +} +#endif +#endif +/* dr_flac_h end */ +#endif /* MA_NO_FLAC */ + +#if !defined(MA_NO_MP3) && !defined(MA_NO_DECODING) +/* dr_mp3_h begin */ +#ifndef ma_dr_mp3_h +#define ma_dr_mp3_h +#ifdef __cplusplus +extern "C" { +#endif +#define MA_DR_MP3_STRINGIFY(x) #x +#define MA_DR_MP3_XSTRINGIFY(x) MA_DR_MP3_STRINGIFY(x) +#define MA_DR_MP3_VERSION_MAJOR 0 +#define MA_DR_MP3_VERSION_MINOR 6 +#define MA_DR_MP3_VERSION_REVISION 40 +#define MA_DR_MP3_VERSION_STRING MA_DR_MP3_XSTRINGIFY(MA_DR_MP3_VERSION_MAJOR) "." MA_DR_MP3_XSTRINGIFY(MA_DR_MP3_VERSION_MINOR) "." MA_DR_MP3_XSTRINGIFY(MA_DR_MP3_VERSION_REVISION) +#include +#define MA_DR_MP3_MAX_PCM_FRAMES_PER_MP3_FRAME 1152 +#define MA_DR_MP3_MAX_SAMPLES_PER_FRAME (MA_DR_MP3_MAX_PCM_FRAMES_PER_MP3_FRAME*2) +MA_API void ma_dr_mp3_version(ma_uint32* pMajor, ma_uint32* pMinor, ma_uint32* pRevision); +MA_API const char* ma_dr_mp3_version_string(void); +typedef struct +{ + int frame_bytes, channels, hz, layer, bitrate_kbps; +} ma_dr_mp3dec_frame_info; +typedef struct +{ + float mdct_overlap[2][9*32], qmf_state[15*2*32]; + int reserv, free_format_bytes; + ma_uint8 header[4], reserv_buf[511]; +} ma_dr_mp3dec; +MA_API void ma_dr_mp3dec_init(ma_dr_mp3dec *dec); +MA_API int ma_dr_mp3dec_decode_frame(ma_dr_mp3dec *dec, const ma_uint8 *mp3, int mp3_bytes, void *pcm, ma_dr_mp3dec_frame_info *info); +MA_API void ma_dr_mp3dec_f32_to_s16(const float *in, ma_int16 *out, size_t num_samples); +typedef enum +{ + ma_dr_mp3_seek_origin_start, + ma_dr_mp3_seek_origin_current +} ma_dr_mp3_seek_origin; +typedef struct +{ + ma_uint64 seekPosInBytes; + ma_uint64 pcmFrameIndex; + ma_uint16 mp3FramesToDiscard; + ma_uint16 pcmFramesToDiscard; +} ma_dr_mp3_seek_point; +typedef size_t (* ma_dr_mp3_read_proc)(void* pUserData, void* pBufferOut, size_t bytesToRead); +typedef ma_bool32 (* ma_dr_mp3_seek_proc)(void* pUserData, int offset, ma_dr_mp3_seek_origin origin); +typedef struct +{ + ma_uint32 channels; + ma_uint32 sampleRate; +} ma_dr_mp3_config; +typedef struct +{ + ma_dr_mp3dec decoder; + ma_uint32 channels; + ma_uint32 sampleRate; + ma_dr_mp3_read_proc onRead; + ma_dr_mp3_seek_proc onSeek; + void* pUserData; + ma_allocation_callbacks allocationCallbacks; + ma_uint32 mp3FrameChannels; + ma_uint32 mp3FrameSampleRate; + ma_uint32 pcmFramesConsumedInMP3Frame; + ma_uint32 pcmFramesRemainingInMP3Frame; + ma_uint8 pcmFrames[sizeof(float)*MA_DR_MP3_MAX_SAMPLES_PER_FRAME]; + ma_uint64 currentPCMFrame; + ma_uint64 streamCursor; + ma_dr_mp3_seek_point* pSeekPoints; + ma_uint32 seekPointCount; + size_t dataSize; + size_t dataCapacity; + size_t dataConsumed; + ma_uint8* pData; + ma_bool32 atEnd : 1; + struct + { + const ma_uint8* pData; + size_t dataSize; + size_t currentReadPos; + } memory; +} ma_dr_mp3; +MA_API ma_bool32 ma_dr_mp3_init(ma_dr_mp3* pMP3, ma_dr_mp3_read_proc onRead, ma_dr_mp3_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_mp3_init_memory(ma_dr_mp3* pMP3, const void* pData, size_t dataSize, const ma_allocation_callbacks* pAllocationCallbacks); +#ifndef MA_DR_MP3_NO_STDIO +MA_API ma_bool32 ma_dr_mp3_init_file(ma_dr_mp3* pMP3, const char* pFilePath, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_bool32 ma_dr_mp3_init_file_w(ma_dr_mp3* pMP3, const wchar_t* pFilePath, const ma_allocation_callbacks* pAllocationCallbacks); +#endif +MA_API void ma_dr_mp3_uninit(ma_dr_mp3* pMP3); +MA_API ma_uint64 ma_dr_mp3_read_pcm_frames_f32(ma_dr_mp3* pMP3, ma_uint64 framesToRead, float* pBufferOut); +MA_API ma_uint64 ma_dr_mp3_read_pcm_frames_s16(ma_dr_mp3* pMP3, ma_uint64 framesToRead, ma_int16* pBufferOut); +MA_API ma_bool32 ma_dr_mp3_seek_to_pcm_frame(ma_dr_mp3* pMP3, ma_uint64 frameIndex); +MA_API ma_uint64 ma_dr_mp3_get_pcm_frame_count(ma_dr_mp3* pMP3); +MA_API ma_uint64 ma_dr_mp3_get_mp3_frame_count(ma_dr_mp3* pMP3); +MA_API ma_bool32 ma_dr_mp3_get_mp3_and_pcm_frame_count(ma_dr_mp3* pMP3, ma_uint64* pMP3FrameCount, ma_uint64* pPCMFrameCount); +MA_API ma_bool32 ma_dr_mp3_calculate_seek_points(ma_dr_mp3* pMP3, ma_uint32* pSeekPointCount, ma_dr_mp3_seek_point* pSeekPoints); +MA_API ma_bool32 ma_dr_mp3_bind_seek_table(ma_dr_mp3* pMP3, ma_uint32 seekPointCount, ma_dr_mp3_seek_point* pSeekPoints); +MA_API float* ma_dr_mp3_open_and_read_pcm_frames_f32(ma_dr_mp3_read_proc onRead, ma_dr_mp3_seek_proc onSeek, void* pUserData, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int16* ma_dr_mp3_open_and_read_pcm_frames_s16(ma_dr_mp3_read_proc onRead, ma_dr_mp3_seek_proc onSeek, void* pUserData, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API float* ma_dr_mp3_open_memory_and_read_pcm_frames_f32(const void* pData, size_t dataSize, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int16* ma_dr_mp3_open_memory_and_read_pcm_frames_s16(const void* pData, size_t dataSize, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +#ifndef MA_DR_MP3_NO_STDIO +MA_API float* ma_dr_mp3_open_file_and_read_pcm_frames_f32(const char* filePath, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_int16* ma_dr_mp3_open_file_and_read_pcm_frames_s16(const char* filePath, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks); +#endif +MA_API void* ma_dr_mp3_malloc(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API void ma_dr_mp3_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks); +#ifdef __cplusplus +} +#endif +#endif +/* dr_mp3_h end */ +#endif /* MA_NO_MP3 */ + + +/************************************************************************************************************************************************************** + +Decoding + +**************************************************************************************************************************************************************/ +#ifndef MA_NO_DECODING + +static ma_result ma_decoder_read_bytes(ma_decoder* pDecoder, void* pBufferOut, size_t bytesToRead, size_t* pBytesRead) +{ + MA_ASSERT(pDecoder != NULL); + + return pDecoder->onRead(pDecoder, pBufferOut, bytesToRead, pBytesRead); +} + +static ma_result ma_decoder_seek_bytes(ma_decoder* pDecoder, ma_int64 byteOffset, ma_seek_origin origin) +{ + MA_ASSERT(pDecoder != NULL); + + return pDecoder->onSeek(pDecoder, byteOffset, origin); +} + +static ma_result ma_decoder_tell_bytes(ma_decoder* pDecoder, ma_int64* pCursor) +{ + MA_ASSERT(pDecoder != NULL); + + if (pDecoder->onTell == NULL) { + return MA_NOT_IMPLEMENTED; + } + + return pDecoder->onTell(pDecoder, pCursor); +} + + +MA_API ma_decoding_backend_config ma_decoding_backend_config_init(ma_format preferredFormat, ma_uint32 seekPointCount) +{ + ma_decoding_backend_config config; + + MA_ZERO_OBJECT(&config); + config.preferredFormat = preferredFormat; + config.seekPointCount = seekPointCount; + + return config; +} + + +MA_API ma_decoder_config ma_decoder_config_init(ma_format outputFormat, ma_uint32 outputChannels, ma_uint32 outputSampleRate) +{ + ma_decoder_config config; + MA_ZERO_OBJECT(&config); + config.format = outputFormat; + config.channels = outputChannels; + config.sampleRate = outputSampleRate; + config.resampling = ma_resampler_config_init(ma_format_unknown, 0, 0, 0, ma_resample_algorithm_linear); /* Format/channels/rate doesn't matter here. */ + config.encodingFormat = ma_encoding_format_unknown; + + /* Note that we are intentionally leaving the channel map empty here which will cause the default channel map to be used. */ + + return config; +} + +MA_API ma_decoder_config ma_decoder_config_init_default(void) +{ + return ma_decoder_config_init(ma_format_unknown, 0, 0); +} + +MA_API ma_decoder_config ma_decoder_config_init_copy(const ma_decoder_config* pConfig) +{ + ma_decoder_config config; + if (pConfig != NULL) { + config = *pConfig; + } else { + MA_ZERO_OBJECT(&config); + } + + return config; +} + +static ma_result ma_decoder__init_data_converter(ma_decoder* pDecoder, const ma_decoder_config* pConfig) +{ + ma_result result; + ma_data_converter_config converterConfig; + ma_format internalFormat; + ma_uint32 internalChannels; + ma_uint32 internalSampleRate; + ma_channel internalChannelMap[MA_MAX_CHANNELS]; + + MA_ASSERT(pDecoder != NULL); + MA_ASSERT(pConfig != NULL); + + result = ma_data_source_get_data_format(pDecoder->pBackend, &internalFormat, &internalChannels, &internalSampleRate, internalChannelMap, ma_countof(internalChannelMap)); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the internal data format. */ + } + + + /* Make sure we're not asking for too many channels. */ + if (pConfig->channels > MA_MAX_CHANNELS) { + return MA_INVALID_ARGS; + } + + /* The internal channels should have already been validated at a higher level, but we'll do it again explicitly here for safety. */ + if (internalChannels > MA_MAX_CHANNELS) { + return MA_INVALID_ARGS; + } + + + /* Output format. */ + if (pConfig->format == ma_format_unknown) { + pDecoder->outputFormat = internalFormat; + } else { + pDecoder->outputFormat = pConfig->format; + } + + if (pConfig->channels == 0) { + pDecoder->outputChannels = internalChannels; + } else { + pDecoder->outputChannels = pConfig->channels; + } + + if (pConfig->sampleRate == 0) { + pDecoder->outputSampleRate = internalSampleRate; + } else { + pDecoder->outputSampleRate = pConfig->sampleRate; + } + + converterConfig = ma_data_converter_config_init( + internalFormat, pDecoder->outputFormat, + internalChannels, pDecoder->outputChannels, + internalSampleRate, pDecoder->outputSampleRate + ); + converterConfig.pChannelMapIn = internalChannelMap; + converterConfig.pChannelMapOut = pConfig->pChannelMap; + converterConfig.channelMixMode = pConfig->channelMixMode; + converterConfig.ditherMode = pConfig->ditherMode; + converterConfig.allowDynamicSampleRate = MA_FALSE; /* Never allow dynamic sample rate conversion. Setting this to true will disable passthrough optimizations. */ + converterConfig.resampling = pConfig->resampling; + + result = ma_data_converter_init(&converterConfig, &pDecoder->allocationCallbacks, &pDecoder->converter); + if (result != MA_SUCCESS) { + return result; + } + + /* + Now that we have the decoder we need to determine whether or not we need a heap-allocated cache. We'll + need this if the data converter does not support calculation of the required input frame count. To + determine support for this we'll just run a test. + */ + { + ma_uint64 unused; + + result = ma_data_converter_get_required_input_frame_count(&pDecoder->converter, 1, &unused); + if (result != MA_SUCCESS) { + /* + We were unable to calculate the required input frame count which means we'll need to use + a heap-allocated cache. + */ + ma_uint64 inputCacheCapSizeInBytes; + + pDecoder->inputCacheCap = MA_DATA_CONVERTER_STACK_BUFFER_SIZE / ma_get_bytes_per_frame(internalFormat, internalChannels); + + /* Not strictly necessary, but keeping here for safety in case we change the default value of pDecoder->inputCacheCap. */ + inputCacheCapSizeInBytes = pDecoder->inputCacheCap * ma_get_bytes_per_frame(internalFormat, internalChannels); + if (inputCacheCapSizeInBytes > MA_SIZE_MAX) { + ma_data_converter_uninit(&pDecoder->converter, &pDecoder->allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + + pDecoder->pInputCache = ma_malloc((size_t)inputCacheCapSizeInBytes, &pDecoder->allocationCallbacks); /* Safe cast to size_t. */ + if (pDecoder->pInputCache == NULL) { + ma_data_converter_uninit(&pDecoder->converter, &pDecoder->allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + } + } + + return MA_SUCCESS; +} + + + +static ma_result ma_decoder_internal_on_read__custom(void* pUserData, void* pBufferOut, size_t bytesToRead, size_t* pBytesRead) +{ + ma_decoder* pDecoder = (ma_decoder*)pUserData; + MA_ASSERT(pDecoder != NULL); + + return ma_decoder_read_bytes(pDecoder, pBufferOut, bytesToRead, pBytesRead); +} + +static ma_result ma_decoder_internal_on_seek__custom(void* pUserData, ma_int64 offset, ma_seek_origin origin) +{ + ma_decoder* pDecoder = (ma_decoder*)pUserData; + MA_ASSERT(pDecoder != NULL); + + return ma_decoder_seek_bytes(pDecoder, offset, origin); +} + +static ma_result ma_decoder_internal_on_tell__custom(void* pUserData, ma_int64* pCursor) +{ + ma_decoder* pDecoder = (ma_decoder*)pUserData; + MA_ASSERT(pDecoder != NULL); + + return ma_decoder_tell_bytes(pDecoder, pCursor); +} + + +static ma_result ma_decoder_init_from_vtable__internal(const ma_decoding_backend_vtable* pVTable, void* pVTableUserData, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoding_backend_config backendConfig; + ma_data_source* pBackend; + + MA_ASSERT(pVTable != NULL); + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + if (pVTable->onInit == NULL) { + return MA_NOT_IMPLEMENTED; + } + + backendConfig = ma_decoding_backend_config_init(pConfig->format, pConfig->seekPointCount); + + result = pVTable->onInit(pVTableUserData, ma_decoder_internal_on_read__custom, ma_decoder_internal_on_seek__custom, ma_decoder_internal_on_tell__custom, pDecoder, &backendConfig, &pDecoder->allocationCallbacks, &pBackend); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the backend from this vtable. */ + } + + /* Getting here means we were able to initialize the backend so we can now initialize the decoder. */ + pDecoder->pBackend = pBackend; + pDecoder->pBackendVTable = pVTable; + pDecoder->pBackendUserData = pConfig->pCustomBackendUserData; + + return MA_SUCCESS; +} + +static ma_result ma_decoder_init_from_file__internal(const ma_decoding_backend_vtable* pVTable, void* pVTableUserData, const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoding_backend_config backendConfig; + ma_data_source* pBackend; + + MA_ASSERT(pVTable != NULL); + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + if (pVTable->onInitFile == NULL) { + return MA_NOT_IMPLEMENTED; + } + + backendConfig = ma_decoding_backend_config_init(pConfig->format, pConfig->seekPointCount); + + result = pVTable->onInitFile(pVTableUserData, pFilePath, &backendConfig, &pDecoder->allocationCallbacks, &pBackend); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the backend from this vtable. */ + } + + /* Getting here means we were able to initialize the backend so we can now initialize the decoder. */ + pDecoder->pBackend = pBackend; + pDecoder->pBackendVTable = pVTable; + pDecoder->pBackendUserData = pConfig->pCustomBackendUserData; + + return MA_SUCCESS; +} + +static ma_result ma_decoder_init_from_file_w__internal(const ma_decoding_backend_vtable* pVTable, void* pVTableUserData, const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoding_backend_config backendConfig; + ma_data_source* pBackend; + + MA_ASSERT(pVTable != NULL); + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + if (pVTable->onInitFileW == NULL) { + return MA_NOT_IMPLEMENTED; + } + + backendConfig = ma_decoding_backend_config_init(pConfig->format, pConfig->seekPointCount); + + result = pVTable->onInitFileW(pVTableUserData, pFilePath, &backendConfig, &pDecoder->allocationCallbacks, &pBackend); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the backend from this vtable. */ + } + + /* Getting here means we were able to initialize the backend so we can now initialize the decoder. */ + pDecoder->pBackend = pBackend; + pDecoder->pBackendVTable = pVTable; + pDecoder->pBackendUserData = pConfig->pCustomBackendUserData; + + return MA_SUCCESS; +} + +static ma_result ma_decoder_init_from_memory__internal(const ma_decoding_backend_vtable* pVTable, void* pVTableUserData, const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoding_backend_config backendConfig; + ma_data_source* pBackend; + + MA_ASSERT(pVTable != NULL); + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + if (pVTable->onInitMemory == NULL) { + return MA_NOT_IMPLEMENTED; + } + + backendConfig = ma_decoding_backend_config_init(pConfig->format, pConfig->seekPointCount); + + result = pVTable->onInitMemory(pVTableUserData, pData, dataSize, &backendConfig, &pDecoder->allocationCallbacks, &pBackend); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the backend from this vtable. */ + } + + /* Getting here means we were able to initialize the backend so we can now initialize the decoder. */ + pDecoder->pBackend = pBackend; + pDecoder->pBackendVTable = pVTable; + pDecoder->pBackendUserData = pConfig->pCustomBackendUserData; + + return MA_SUCCESS; +} + + + +static ma_result ma_decoder_init_custom__internal(const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result = MA_NO_BACKEND; + size_t ivtable; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + if (pConfig->ppCustomBackendVTables == NULL) { + return MA_NO_BACKEND; + } + + /* The order each backend is listed is what defines the priority. */ + for (ivtable = 0; ivtable < pConfig->customBackendCount; ivtable += 1) { + const ma_decoding_backend_vtable* pVTable = pConfig->ppCustomBackendVTables[ivtable]; + if (pVTable != NULL) { + result = ma_decoder_init_from_vtable__internal(pVTable, pConfig->pCustomBackendUserData, pConfig, pDecoder); + if (result == MA_SUCCESS) { + return MA_SUCCESS; + } else { + /* Initialization failed. Move on to the next one, but seek back to the start first so the next vtable starts from the first byte of the file. */ + result = ma_decoder_seek_bytes(pDecoder, 0, ma_seek_origin_start); + if (result != MA_SUCCESS) { + return result; /* Failed to seek back to the start. */ + } + } + } else { + /* No vtable. */ + } + } + + /* Getting here means we couldn't find a backend. */ + return MA_NO_BACKEND; +} + +static ma_result ma_decoder_init_custom_from_file__internal(const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result = MA_NO_BACKEND; + size_t ivtable; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + if (pConfig->ppCustomBackendVTables == NULL) { + return MA_NO_BACKEND; + } + + /* The order each backend is listed is what defines the priority. */ + for (ivtable = 0; ivtable < pConfig->customBackendCount; ivtable += 1) { + const ma_decoding_backend_vtable* pVTable = pConfig->ppCustomBackendVTables[ivtable]; + if (pVTable != NULL) { + result = ma_decoder_init_from_file__internal(pVTable, pConfig->pCustomBackendUserData, pFilePath, pConfig, pDecoder); + if (result == MA_SUCCESS) { + return MA_SUCCESS; + } + } else { + /* No vtable. */ + } + } + + /* Getting here means we couldn't find a backend. */ + return MA_NO_BACKEND; +} + +static ma_result ma_decoder_init_custom_from_file_w__internal(const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result = MA_NO_BACKEND; + size_t ivtable; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + if (pConfig->ppCustomBackendVTables == NULL) { + return MA_NO_BACKEND; + } + + /* The order each backend is listed is what defines the priority. */ + for (ivtable = 0; ivtable < pConfig->customBackendCount; ivtable += 1) { + const ma_decoding_backend_vtable* pVTable = pConfig->ppCustomBackendVTables[ivtable]; + if (pVTable != NULL) { + result = ma_decoder_init_from_file_w__internal(pVTable, pConfig->pCustomBackendUserData, pFilePath, pConfig, pDecoder); + if (result == MA_SUCCESS) { + return MA_SUCCESS; + } + } else { + /* No vtable. */ + } + } + + /* Getting here means we couldn't find a backend. */ + return MA_NO_BACKEND; +} + +static ma_result ma_decoder_init_custom_from_memory__internal(const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result = MA_NO_BACKEND; + size_t ivtable; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + if (pConfig->ppCustomBackendVTables == NULL) { + return MA_NO_BACKEND; + } + + /* The order each backend is listed is what defines the priority. */ + for (ivtable = 0; ivtable < pConfig->customBackendCount; ivtable += 1) { + const ma_decoding_backend_vtable* pVTable = pConfig->ppCustomBackendVTables[ivtable]; + if (pVTable != NULL) { + result = ma_decoder_init_from_memory__internal(pVTable, pConfig->pCustomBackendUserData, pData, dataSize, pConfig, pDecoder); + if (result == MA_SUCCESS) { + return MA_SUCCESS; + } + } else { + /* No vtable. */ + } + } + + /* Getting here means we couldn't find a backend. */ + return MA_NO_BACKEND; +} + + +/* WAV */ +#ifdef ma_dr_wav_h +#define MA_HAS_WAV + +typedef struct +{ + ma_data_source_base ds; + ma_read_proc onRead; + ma_seek_proc onSeek; + ma_tell_proc onTell; + void* pReadSeekTellUserData; + ma_format format; /* Can be f32, s16 or s32. */ +#if !defined(MA_NO_WAV) + ma_dr_wav dr; +#endif +} ma_wav; + +MA_API ma_result ma_wav_init(ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_wav* pWav); +MA_API ma_result ma_wav_init_file(const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_wav* pWav); +MA_API ma_result ma_wav_init_file_w(const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_wav* pWav); +MA_API ma_result ma_wav_init_memory(const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_wav* pWav); +MA_API void ma_wav_uninit(ma_wav* pWav, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_wav_read_pcm_frames(ma_wav* pWav, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_wav_seek_to_pcm_frame(ma_wav* pWav, ma_uint64 frameIndex); +MA_API ma_result ma_wav_get_data_format(ma_wav* pWav, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_wav_get_cursor_in_pcm_frames(ma_wav* pWav, ma_uint64* pCursor); +MA_API ma_result ma_wav_get_length_in_pcm_frames(ma_wav* pWav, ma_uint64* pLength); + + +static ma_result ma_wav_ds_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_wav_read_pcm_frames((ma_wav*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_wav_ds_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_wav_seek_to_pcm_frame((ma_wav*)pDataSource, frameIndex); +} + +static ma_result ma_wav_ds_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + return ma_wav_get_data_format((ma_wav*)pDataSource, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); +} + +static ma_result ma_wav_ds_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + return ma_wav_get_cursor_in_pcm_frames((ma_wav*)pDataSource, pCursor); +} + +static ma_result ma_wav_ds_get_length(ma_data_source* pDataSource, ma_uint64* pLength) +{ + return ma_wav_get_length_in_pcm_frames((ma_wav*)pDataSource, pLength); +} + +static ma_data_source_vtable g_ma_wav_ds_vtable = +{ + ma_wav_ds_read, + ma_wav_ds_seek, + ma_wav_ds_get_data_format, + ma_wav_ds_get_cursor, + ma_wav_ds_get_length, + NULL, /* onSetLooping */ + 0 +}; + + +#if !defined(MA_NO_WAV) +static size_t ma_wav_dr_callback__read(void* pUserData, void* pBufferOut, size_t bytesToRead) +{ + ma_wav* pWav = (ma_wav*)pUserData; + ma_result result; + size_t bytesRead; + + MA_ASSERT(pWav != NULL); + + result = pWav->onRead(pWav->pReadSeekTellUserData, pBufferOut, bytesToRead, &bytesRead); + (void)result; + + return bytesRead; +} + +static ma_bool32 ma_wav_dr_callback__seek(void* pUserData, int offset, ma_dr_wav_seek_origin origin) +{ + ma_wav* pWav = (ma_wav*)pUserData; + ma_result result; + ma_seek_origin maSeekOrigin; + + MA_ASSERT(pWav != NULL); + + maSeekOrigin = ma_seek_origin_start; + if (origin == ma_dr_wav_seek_origin_current) { + maSeekOrigin = ma_seek_origin_current; + } + + result = pWav->onSeek(pWav->pReadSeekTellUserData, offset, maSeekOrigin); + if (result != MA_SUCCESS) { + return MA_FALSE; + } + + return MA_TRUE; +} +#endif + +static ma_result ma_wav_init_internal(const ma_decoding_backend_config* pConfig, ma_wav* pWav) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + + if (pWav == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pWav); + pWav->format = ma_format_unknown; /* Use closest match to source file by default. */ + + if (pConfig != NULL && (pConfig->preferredFormat == ma_format_f32 || pConfig->preferredFormat == ma_format_s16 || pConfig->preferredFormat == ma_format_s32)) { + pWav->format = pConfig->preferredFormat; + } else { + /* Getting here means something other than f32 and s16 was specified. Just leave this unset to use the default format. */ + } + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_wav_ds_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pWav->ds); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the base data source. */ + } + + return MA_SUCCESS; +} + +static ma_result ma_wav_post_init(ma_wav* pWav) +{ + /* + If an explicit format was not specified, try picking the closest match based on the internal + format. The format needs to be supported by miniaudio. + */ + if (pWav->format == ma_format_unknown) { + switch (pWav->dr.translatedFormatTag) + { + case MA_DR_WAVE_FORMAT_PCM: + { + if (pWav->dr.bitsPerSample == 8) { + pWav->format = ma_format_u8; + } else if (pWav->dr.bitsPerSample == 16) { + pWav->format = ma_format_s16; + } else if (pWav->dr.bitsPerSample == 24) { + pWav->format = ma_format_s24; + } else if (pWav->dr.bitsPerSample == 32) { + pWav->format = ma_format_s32; + } + } break; + + case MA_DR_WAVE_FORMAT_IEEE_FLOAT: + { + if (pWav->dr.bitsPerSample == 32) { + pWav->format = ma_format_f32; + } + } break; + + default: break; + } + + /* Fall back to f32 if we couldn't find anything. */ + if (pWav->format == ma_format_unknown) { + pWav->format = ma_format_f32; + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_wav_init(ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_wav* pWav) +{ + ma_result result; + + result = ma_wav_init_internal(pConfig, pWav); + if (result != MA_SUCCESS) { + return result; + } + + if (onRead == NULL || onSeek == NULL) { + return MA_INVALID_ARGS; /* onRead and onSeek are mandatory. */ + } + + pWav->onRead = onRead; + pWav->onSeek = onSeek; + pWav->onTell = onTell; + pWav->pReadSeekTellUserData = pReadSeekTellUserData; + + #if !defined(MA_NO_WAV) + { + ma_bool32 wavResult; + + wavResult = ma_dr_wav_init(&pWav->dr, ma_wav_dr_callback__read, ma_wav_dr_callback__seek, pWav, pAllocationCallbacks); + if (wavResult != MA_TRUE) { + return MA_INVALID_FILE; + } + + ma_wav_post_init(pWav); + + return MA_SUCCESS; + } + #else + { + /* wav is disabled. */ + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_wav_init_file(const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_wav* pWav) +{ + ma_result result; + + result = ma_wav_init_internal(pConfig, pWav); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_WAV) + { + ma_bool32 wavResult; + + wavResult = ma_dr_wav_init_file(&pWav->dr, pFilePath, pAllocationCallbacks); + if (wavResult != MA_TRUE) { + return MA_INVALID_FILE; + } + + ma_wav_post_init(pWav); + + return MA_SUCCESS; + } + #else + { + /* wav is disabled. */ + (void)pFilePath; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_wav_init_file_w(const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_wav* pWav) +{ + ma_result result; + + result = ma_wav_init_internal(pConfig, pWav); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_WAV) + { + ma_bool32 wavResult; + + wavResult = ma_dr_wav_init_file_w(&pWav->dr, pFilePath, pAllocationCallbacks); + if (wavResult != MA_TRUE) { + return MA_INVALID_FILE; + } + + ma_wav_post_init(pWav); + + return MA_SUCCESS; + } + #else + { + /* wav is disabled. */ + (void)pFilePath; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_wav_init_memory(const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_wav* pWav) +{ + ma_result result; + + result = ma_wav_init_internal(pConfig, pWav); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_WAV) + { + ma_bool32 wavResult; + + wavResult = ma_dr_wav_init_memory(&pWav->dr, pData, dataSize, pAllocationCallbacks); + if (wavResult != MA_TRUE) { + return MA_INVALID_FILE; + } + + ma_wav_post_init(pWav); + + return MA_SUCCESS; + } + #else + { + /* wav is disabled. */ + (void)pData; + (void)dataSize; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API void ma_wav_uninit(ma_wav* pWav, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pWav == NULL) { + return; + } + + (void)pAllocationCallbacks; + + #if !defined(MA_NO_WAV) + { + ma_dr_wav_uninit(&pWav->dr); + } + #else + { + /* wav is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + } + #endif + + ma_data_source_uninit(&pWav->ds); +} + +MA_API ma_result ma_wav_read_pcm_frames(ma_wav* pWav, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pWav == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_WAV) + { + /* We always use floating point format. */ + ma_result result = MA_SUCCESS; /* Must be initialized to MA_SUCCESS. */ + ma_uint64 totalFramesRead = 0; + ma_format format; + + ma_wav_get_data_format(pWav, &format, NULL, NULL, NULL, 0); + + switch (format) + { + case ma_format_f32: + { + totalFramesRead = ma_dr_wav_read_pcm_frames_f32(&pWav->dr, frameCount, (float*)pFramesOut); + } break; + + case ma_format_s16: + { + totalFramesRead = ma_dr_wav_read_pcm_frames_s16(&pWav->dr, frameCount, (ma_int16*)pFramesOut); + } break; + + case ma_format_s32: + { + totalFramesRead = ma_dr_wav_read_pcm_frames_s32(&pWav->dr, frameCount, (ma_int32*)pFramesOut); + } break; + + /* Fallback to a raw read. */ + case ma_format_unknown: return MA_INVALID_OPERATION; /* <-- this should never be hit because initialization would just fall back to a supported format. */ + default: + { + totalFramesRead = ma_dr_wav_read_pcm_frames(&pWav->dr, frameCount, pFramesOut); + } break; + } + + /* In the future we'll update ma_dr_wav to return MA_AT_END for us. */ + if (totalFramesRead == 0) { + result = MA_AT_END; + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesRead; + } + + if (result == MA_SUCCESS && totalFramesRead == 0) { + result = MA_AT_END; + } + + return result; + } + #else + { + /* wav is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + + (void)pFramesOut; + (void)frameCount; + (void)pFramesRead; + + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_wav_seek_to_pcm_frame(ma_wav* pWav, ma_uint64 frameIndex) +{ + if (pWav == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_WAV) + { + ma_bool32 wavResult; + + wavResult = ma_dr_wav_seek_to_pcm_frame(&pWav->dr, frameIndex); + if (wavResult != MA_TRUE) { + return MA_ERROR; + } + + return MA_SUCCESS; + } + #else + { + /* wav is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + + (void)frameIndex; + + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_wav_get_data_format(ma_wav* pWav, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + /* Defaults for safety. */ + if (pFormat != NULL) { + *pFormat = ma_format_unknown; + } + if (pChannels != NULL) { + *pChannels = 0; + } + if (pSampleRate != NULL) { + *pSampleRate = 0; + } + if (pChannelMap != NULL) { + MA_ZERO_MEMORY(pChannelMap, sizeof(*pChannelMap) * channelMapCap); + } + + if (pWav == NULL) { + return MA_INVALID_OPERATION; + } + + if (pFormat != NULL) { + *pFormat = pWav->format; + } + + #if !defined(MA_NO_WAV) + { + if (pChannels != NULL) { + *pChannels = pWav->dr.channels; + } + + if (pSampleRate != NULL) { + *pSampleRate = pWav->dr.sampleRate; + } + + if (pChannelMap != NULL) { + ma_channel_map_init_standard(ma_standard_channel_map_microsoft, pChannelMap, channelMapCap, pWav->dr.channels); + } + + return MA_SUCCESS; + } + #else + { + /* wav is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_wav_get_cursor_in_pcm_frames(ma_wav* pWav, ma_uint64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; /* Safety. */ + + if (pWav == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_WAV) + { + ma_result wavResult = ma_dr_wav_get_cursor_in_pcm_frames(&pWav->dr, pCursor); + if (wavResult != MA_SUCCESS) { + return (ma_result)wavResult; /* ma_dr_wav result codes map to miniaudio's. */ + } + + return MA_SUCCESS; + } + #else + { + /* wav is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_wav_get_length_in_pcm_frames(ma_wav* pWav, ma_uint64* pLength) +{ + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; /* Safety. */ + + if (pWav == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_WAV) + { + ma_result wavResult = ma_dr_wav_get_length_in_pcm_frames(&pWav->dr, pLength); + if (wavResult != MA_SUCCESS) { + return (ma_result)wavResult; /* ma_dr_wav result codes map to miniaudio's. */ + } + + return MA_SUCCESS; + } + #else + { + /* wav is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + + +static ma_result ma_decoding_backend_init__wav(void* pUserData, ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_wav* pWav; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pWav = (ma_wav*)ma_malloc(sizeof(*pWav), pAllocationCallbacks); + if (pWav == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_wav_init(onRead, onSeek, onTell, pReadSeekTellUserData, pConfig, pAllocationCallbacks, pWav); + if (result != MA_SUCCESS) { + ma_free(pWav, pAllocationCallbacks); + return result; + } + + *ppBackend = pWav; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_file__wav(void* pUserData, const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_wav* pWav; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pWav = (ma_wav*)ma_malloc(sizeof(*pWav), pAllocationCallbacks); + if (pWav == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_wav_init_file(pFilePath, pConfig, pAllocationCallbacks, pWav); + if (result != MA_SUCCESS) { + ma_free(pWav, pAllocationCallbacks); + return result; + } + + *ppBackend = pWav; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_file_w__wav(void* pUserData, const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_wav* pWav; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pWav = (ma_wav*)ma_malloc(sizeof(*pWav), pAllocationCallbacks); + if (pWav == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_wav_init_file_w(pFilePath, pConfig, pAllocationCallbacks, pWav); + if (result != MA_SUCCESS) { + ma_free(pWav, pAllocationCallbacks); + return result; + } + + *ppBackend = pWav; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_memory__wav(void* pUserData, const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_wav* pWav; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pWav = (ma_wav*)ma_malloc(sizeof(*pWav), pAllocationCallbacks); + if (pWav == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_wav_init_memory(pData, dataSize, pConfig, pAllocationCallbacks, pWav); + if (result != MA_SUCCESS) { + ma_free(pWav, pAllocationCallbacks); + return result; + } + + *ppBackend = pWav; + + return MA_SUCCESS; +} + +static void ma_decoding_backend_uninit__wav(void* pUserData, ma_data_source* pBackend, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_wav* pWav = (ma_wav*)pBackend; + + (void)pUserData; + + ma_wav_uninit(pWav, pAllocationCallbacks); + ma_free(pWav, pAllocationCallbacks); +} + +static ma_decoding_backend_vtable g_ma_decoding_backend_vtable_wav = +{ + ma_decoding_backend_init__wav, + ma_decoding_backend_init_file__wav, + ma_decoding_backend_init_file_w__wav, + ma_decoding_backend_init_memory__wav, + ma_decoding_backend_uninit__wav +}; + +static ma_result ma_decoder_init_wav__internal(const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_vtable__internal(&g_ma_decoding_backend_vtable_wav, NULL, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_wav_from_file__internal(const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_file__internal(&g_ma_decoding_backend_vtable_wav, NULL, pFilePath, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_wav_from_file_w__internal(const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_file_w__internal(&g_ma_decoding_backend_vtable_wav, NULL, pFilePath, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_wav_from_memory__internal(const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_memory__internal(&g_ma_decoding_backend_vtable_wav, NULL, pData, dataSize, pConfig, pDecoder); +} +#endif /* ma_dr_wav_h */ + +/* FLAC */ +#ifdef ma_dr_flac_h +#define MA_HAS_FLAC + +typedef struct +{ + ma_data_source_base ds; + ma_read_proc onRead; + ma_seek_proc onSeek; + ma_tell_proc onTell; + void* pReadSeekTellUserData; + ma_format format; /* Can be f32, s16 or s32. */ +#if !defined(MA_NO_FLAC) + ma_dr_flac* dr; +#endif +} ma_flac; + +MA_API ma_result ma_flac_init(ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_flac* pFlac); +MA_API ma_result ma_flac_init_file(const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_flac* pFlac); +MA_API ma_result ma_flac_init_file_w(const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_flac* pFlac); +MA_API ma_result ma_flac_init_memory(const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_flac* pFlac); +MA_API void ma_flac_uninit(ma_flac* pFlac, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_flac_read_pcm_frames(ma_flac* pFlac, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_flac_seek_to_pcm_frame(ma_flac* pFlac, ma_uint64 frameIndex); +MA_API ma_result ma_flac_get_data_format(ma_flac* pFlac, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_flac_get_cursor_in_pcm_frames(ma_flac* pFlac, ma_uint64* pCursor); +MA_API ma_result ma_flac_get_length_in_pcm_frames(ma_flac* pFlac, ma_uint64* pLength); + + +static ma_result ma_flac_ds_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_flac_read_pcm_frames((ma_flac*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_flac_ds_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_flac_seek_to_pcm_frame((ma_flac*)pDataSource, frameIndex); +} + +static ma_result ma_flac_ds_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + return ma_flac_get_data_format((ma_flac*)pDataSource, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); +} + +static ma_result ma_flac_ds_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + return ma_flac_get_cursor_in_pcm_frames((ma_flac*)pDataSource, pCursor); +} + +static ma_result ma_flac_ds_get_length(ma_data_source* pDataSource, ma_uint64* pLength) +{ + return ma_flac_get_length_in_pcm_frames((ma_flac*)pDataSource, pLength); +} + +static ma_data_source_vtable g_ma_flac_ds_vtable = +{ + ma_flac_ds_read, + ma_flac_ds_seek, + ma_flac_ds_get_data_format, + ma_flac_ds_get_cursor, + ma_flac_ds_get_length, + NULL, /* onSetLooping */ + 0 +}; + + +#if !defined(MA_NO_FLAC) +static size_t ma_flac_dr_callback__read(void* pUserData, void* pBufferOut, size_t bytesToRead) +{ + ma_flac* pFlac = (ma_flac*)pUserData; + ma_result result; + size_t bytesRead; + + MA_ASSERT(pFlac != NULL); + + result = pFlac->onRead(pFlac->pReadSeekTellUserData, pBufferOut, bytesToRead, &bytesRead); + (void)result; + + return bytesRead; +} + +static ma_bool32 ma_flac_dr_callback__seek(void* pUserData, int offset, ma_dr_flac_seek_origin origin) +{ + ma_flac* pFlac = (ma_flac*)pUserData; + ma_result result; + ma_seek_origin maSeekOrigin; + + MA_ASSERT(pFlac != NULL); + + maSeekOrigin = ma_seek_origin_start; + if (origin == ma_dr_flac_seek_origin_current) { + maSeekOrigin = ma_seek_origin_current; + } + + result = pFlac->onSeek(pFlac->pReadSeekTellUserData, offset, maSeekOrigin); + if (result != MA_SUCCESS) { + return MA_FALSE; + } + + return MA_TRUE; +} +#endif + +static ma_result ma_flac_init_internal(const ma_decoding_backend_config* pConfig, ma_flac* pFlac) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + + if (pFlac == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pFlac); + pFlac->format = ma_format_f32; /* f32 by default. */ + + if (pConfig != NULL && (pConfig->preferredFormat == ma_format_f32 || pConfig->preferredFormat == ma_format_s16 || pConfig->preferredFormat == ma_format_s32)) { + pFlac->format = pConfig->preferredFormat; + } else { + /* Getting here means something other than f32 and s16 was specified. Just leave this unset to use the default format. */ + } + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_flac_ds_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pFlac->ds); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the base data source. */ + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_flac_init(ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_flac* pFlac) +{ + ma_result result; + + result = ma_flac_init_internal(pConfig, pFlac); + if (result != MA_SUCCESS) { + return result; + } + + if (onRead == NULL || onSeek == NULL) { + return MA_INVALID_ARGS; /* onRead and onSeek are mandatory. */ + } + + pFlac->onRead = onRead; + pFlac->onSeek = onSeek; + pFlac->onTell = onTell; + pFlac->pReadSeekTellUserData = pReadSeekTellUserData; + + #if !defined(MA_NO_FLAC) + { + pFlac->dr = ma_dr_flac_open(ma_flac_dr_callback__read, ma_flac_dr_callback__seek, pFlac, pAllocationCallbacks); + if (pFlac->dr == NULL) { + return MA_INVALID_FILE; + } + + return MA_SUCCESS; + } + #else + { + /* flac is disabled. */ + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_flac_init_file(const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_flac* pFlac) +{ + ma_result result; + + result = ma_flac_init_internal(pConfig, pFlac); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_FLAC) + { + pFlac->dr = ma_dr_flac_open_file(pFilePath, pAllocationCallbacks); + if (pFlac->dr == NULL) { + return MA_INVALID_FILE; + } + + return MA_SUCCESS; + } + #else + { + /* flac is disabled. */ + (void)pFilePath; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_flac_init_file_w(const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_flac* pFlac) +{ + ma_result result; + + result = ma_flac_init_internal(pConfig, pFlac); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_FLAC) + { + pFlac->dr = ma_dr_flac_open_file_w(pFilePath, pAllocationCallbacks); + if (pFlac->dr == NULL) { + return MA_INVALID_FILE; + } + + return MA_SUCCESS; + } + #else + { + /* flac is disabled. */ + (void)pFilePath; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_flac_init_memory(const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_flac* pFlac) +{ + ma_result result; + + result = ma_flac_init_internal(pConfig, pFlac); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_FLAC) + { + pFlac->dr = ma_dr_flac_open_memory(pData, dataSize, pAllocationCallbacks); + if (pFlac->dr == NULL) { + return MA_INVALID_FILE; + } + + return MA_SUCCESS; + } + #else + { + /* flac is disabled. */ + (void)pData; + (void)dataSize; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API void ma_flac_uninit(ma_flac* pFlac, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFlac == NULL) { + return; + } + + (void)pAllocationCallbacks; + + #if !defined(MA_NO_FLAC) + { + ma_dr_flac_close(pFlac->dr); + } + #else + { + /* flac is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + } + #endif + + ma_data_source_uninit(&pFlac->ds); +} + +MA_API ma_result ma_flac_read_pcm_frames(ma_flac* pFlac, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pFlac == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_FLAC) + { + /* We always use floating point format. */ + ma_result result = MA_SUCCESS; /* Must be initialized to MA_SUCCESS. */ + ma_uint64 totalFramesRead = 0; + ma_format format; + + ma_flac_get_data_format(pFlac, &format, NULL, NULL, NULL, 0); + + switch (format) + { + case ma_format_f32: + { + totalFramesRead = ma_dr_flac_read_pcm_frames_f32(pFlac->dr, frameCount, (float*)pFramesOut); + } break; + + case ma_format_s16: + { + totalFramesRead = ma_dr_flac_read_pcm_frames_s16(pFlac->dr, frameCount, (ma_int16*)pFramesOut); + } break; + + case ma_format_s32: + { + totalFramesRead = ma_dr_flac_read_pcm_frames_s32(pFlac->dr, frameCount, (ma_int32*)pFramesOut); + } break; + + case ma_format_u8: + case ma_format_s24: + case ma_format_unknown: + default: + { + return MA_INVALID_OPERATION; + }; + } + + /* In the future we'll update ma_dr_flac to return MA_AT_END for us. */ + if (totalFramesRead == 0) { + result = MA_AT_END; + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesRead; + } + + if (result == MA_SUCCESS && totalFramesRead == 0) { + result = MA_AT_END; + } + + return result; + } + #else + { + /* flac is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + + (void)pFramesOut; + (void)frameCount; + (void)pFramesRead; + + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_flac_seek_to_pcm_frame(ma_flac* pFlac, ma_uint64 frameIndex) +{ + if (pFlac == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_FLAC) + { + ma_bool32 flacResult; + + flacResult = ma_dr_flac_seek_to_pcm_frame(pFlac->dr, frameIndex); + if (flacResult != MA_TRUE) { + return MA_ERROR; + } + + return MA_SUCCESS; + } + #else + { + /* flac is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + + (void)frameIndex; + + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_flac_get_data_format(ma_flac* pFlac, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + /* Defaults for safety. */ + if (pFormat != NULL) { + *pFormat = ma_format_unknown; + } + if (pChannels != NULL) { + *pChannels = 0; + } + if (pSampleRate != NULL) { + *pSampleRate = 0; + } + if (pChannelMap != NULL) { + MA_ZERO_MEMORY(pChannelMap, sizeof(*pChannelMap) * channelMapCap); + } + + if (pFlac == NULL) { + return MA_INVALID_OPERATION; + } + + if (pFormat != NULL) { + *pFormat = pFlac->format; + } + + #if !defined(MA_NO_FLAC) + { + if (pChannels != NULL) { + *pChannels = pFlac->dr->channels; + } + + if (pSampleRate != NULL) { + *pSampleRate = pFlac->dr->sampleRate; + } + + if (pChannelMap != NULL) { + ma_channel_map_init_standard(ma_standard_channel_map_microsoft, pChannelMap, channelMapCap, pFlac->dr->channels); + } + + return MA_SUCCESS; + } + #else + { + /* flac is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_flac_get_cursor_in_pcm_frames(ma_flac* pFlac, ma_uint64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; /* Safety. */ + + if (pFlac == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_FLAC) + { + *pCursor = pFlac->dr->currentPCMFrame; + + return MA_SUCCESS; + } + #else + { + /* flac is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_flac_get_length_in_pcm_frames(ma_flac* pFlac, ma_uint64* pLength) +{ + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; /* Safety. */ + + if (pFlac == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_FLAC) + { + *pLength = pFlac->dr->totalPCMFrameCount; + + return MA_SUCCESS; + } + #else + { + /* flac is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + + +static ma_result ma_decoding_backend_init__flac(void* pUserData, ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_flac* pFlac; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pFlac = (ma_flac*)ma_malloc(sizeof(*pFlac), pAllocationCallbacks); + if (pFlac == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_flac_init(onRead, onSeek, onTell, pReadSeekTellUserData, pConfig, pAllocationCallbacks, pFlac); + if (result != MA_SUCCESS) { + ma_free(pFlac, pAllocationCallbacks); + return result; + } + + *ppBackend = pFlac; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_file__flac(void* pUserData, const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_flac* pFlac; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pFlac = (ma_flac*)ma_malloc(sizeof(*pFlac), pAllocationCallbacks); + if (pFlac == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_flac_init_file(pFilePath, pConfig, pAllocationCallbacks, pFlac); + if (result != MA_SUCCESS) { + ma_free(pFlac, pAllocationCallbacks); + return result; + } + + *ppBackend = pFlac; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_file_w__flac(void* pUserData, const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_flac* pFlac; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pFlac = (ma_flac*)ma_malloc(sizeof(*pFlac), pAllocationCallbacks); + if (pFlac == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_flac_init_file_w(pFilePath, pConfig, pAllocationCallbacks, pFlac); + if (result != MA_SUCCESS) { + ma_free(pFlac, pAllocationCallbacks); + return result; + } + + *ppBackend = pFlac; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_memory__flac(void* pUserData, const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_flac* pFlac; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pFlac = (ma_flac*)ma_malloc(sizeof(*pFlac), pAllocationCallbacks); + if (pFlac == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_flac_init_memory(pData, dataSize, pConfig, pAllocationCallbacks, pFlac); + if (result != MA_SUCCESS) { + ma_free(pFlac, pAllocationCallbacks); + return result; + } + + *ppBackend = pFlac; + + return MA_SUCCESS; +} + +static void ma_decoding_backend_uninit__flac(void* pUserData, ma_data_source* pBackend, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_flac* pFlac = (ma_flac*)pBackend; + + (void)pUserData; + + ma_flac_uninit(pFlac, pAllocationCallbacks); + ma_free(pFlac, pAllocationCallbacks); +} + +static ma_decoding_backend_vtable g_ma_decoding_backend_vtable_flac = +{ + ma_decoding_backend_init__flac, + ma_decoding_backend_init_file__flac, + ma_decoding_backend_init_file_w__flac, + ma_decoding_backend_init_memory__flac, + ma_decoding_backend_uninit__flac +}; + +static ma_result ma_decoder_init_flac__internal(const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_vtable__internal(&g_ma_decoding_backend_vtable_flac, NULL, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_flac_from_file__internal(const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_file__internal(&g_ma_decoding_backend_vtable_flac, NULL, pFilePath, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_flac_from_file_w__internal(const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_file_w__internal(&g_ma_decoding_backend_vtable_flac, NULL, pFilePath, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_flac_from_memory__internal(const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_memory__internal(&g_ma_decoding_backend_vtable_flac, NULL, pData, dataSize, pConfig, pDecoder); +} +#endif /* ma_dr_flac_h */ + +/* MP3 */ +#ifdef ma_dr_mp3_h +#define MA_HAS_MP3 + +typedef struct +{ + ma_data_source_base ds; + ma_read_proc onRead; + ma_seek_proc onSeek; + ma_tell_proc onTell; + void* pReadSeekTellUserData; + ma_format format; /* Can be f32 or s16. */ +#if !defined(MA_NO_MP3) + ma_dr_mp3 dr; + ma_uint32 seekPointCount; + ma_dr_mp3_seek_point* pSeekPoints; /* Only used if seek table generation is used. */ +#endif +} ma_mp3; + +MA_API ma_result ma_mp3_init(ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_mp3* pMP3); +MA_API ma_result ma_mp3_init_file(const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_mp3* pMP3); +MA_API ma_result ma_mp3_init_file_w(const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_mp3* pMP3); +MA_API ma_result ma_mp3_init_memory(const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_mp3* pMP3); +MA_API void ma_mp3_uninit(ma_mp3* pMP3, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_mp3_read_pcm_frames(ma_mp3* pMP3, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_mp3_seek_to_pcm_frame(ma_mp3* pMP3, ma_uint64 frameIndex); +MA_API ma_result ma_mp3_get_data_format(ma_mp3* pMP3, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_mp3_get_cursor_in_pcm_frames(ma_mp3* pMP3, ma_uint64* pCursor); +MA_API ma_result ma_mp3_get_length_in_pcm_frames(ma_mp3* pMP3, ma_uint64* pLength); + + +static ma_result ma_mp3_ds_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_mp3_read_pcm_frames((ma_mp3*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_mp3_ds_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_mp3_seek_to_pcm_frame((ma_mp3*)pDataSource, frameIndex); +} + +static ma_result ma_mp3_ds_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + return ma_mp3_get_data_format((ma_mp3*)pDataSource, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); +} + +static ma_result ma_mp3_ds_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + return ma_mp3_get_cursor_in_pcm_frames((ma_mp3*)pDataSource, pCursor); +} + +static ma_result ma_mp3_ds_get_length(ma_data_source* pDataSource, ma_uint64* pLength) +{ + return ma_mp3_get_length_in_pcm_frames((ma_mp3*)pDataSource, pLength); +} + +static ma_data_source_vtable g_ma_mp3_ds_vtable = +{ + ma_mp3_ds_read, + ma_mp3_ds_seek, + ma_mp3_ds_get_data_format, + ma_mp3_ds_get_cursor, + ma_mp3_ds_get_length, + NULL, /* onSetLooping */ + 0 +}; + + +#if !defined(MA_NO_MP3) +static size_t ma_mp3_dr_callback__read(void* pUserData, void* pBufferOut, size_t bytesToRead) +{ + ma_mp3* pMP3 = (ma_mp3*)pUserData; + ma_result result; + size_t bytesRead; + + MA_ASSERT(pMP3 != NULL); + + result = pMP3->onRead(pMP3->pReadSeekTellUserData, pBufferOut, bytesToRead, &bytesRead); + (void)result; + + return bytesRead; +} + +static ma_bool32 ma_mp3_dr_callback__seek(void* pUserData, int offset, ma_dr_mp3_seek_origin origin) +{ + ma_mp3* pMP3 = (ma_mp3*)pUserData; + ma_result result; + ma_seek_origin maSeekOrigin; + + MA_ASSERT(pMP3 != NULL); + + maSeekOrigin = ma_seek_origin_start; + if (origin == ma_dr_mp3_seek_origin_current) { + maSeekOrigin = ma_seek_origin_current; + } + + result = pMP3->onSeek(pMP3->pReadSeekTellUserData, offset, maSeekOrigin); + if (result != MA_SUCCESS) { + return MA_FALSE; + } + + return MA_TRUE; +} +#endif + +static ma_result ma_mp3_init_internal(const ma_decoding_backend_config* pConfig, ma_mp3* pMP3) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + + if (pMP3 == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pMP3); + pMP3->format = ma_format_f32; /* f32 by default. */ + + if (pConfig != NULL && (pConfig->preferredFormat == ma_format_f32 || pConfig->preferredFormat == ma_format_s16)) { + pMP3->format = pConfig->preferredFormat; + } else { + /* Getting here means something other than f32 and s16 was specified. Just leave this unset to use the default format. */ + } + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_mp3_ds_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pMP3->ds); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the base data source. */ + } + + return MA_SUCCESS; +} + +static ma_result ma_mp3_generate_seek_table(ma_mp3* pMP3, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_bool32 mp3Result; + ma_uint32 seekPointCount = 0; + ma_dr_mp3_seek_point* pSeekPoints = NULL; + + MA_ASSERT(pMP3 != NULL); + MA_ASSERT(pConfig != NULL); + + seekPointCount = pConfig->seekPointCount; + if (seekPointCount > 0) { + pSeekPoints = (ma_dr_mp3_seek_point*)ma_malloc(sizeof(*pMP3->pSeekPoints) * seekPointCount, pAllocationCallbacks); + if (pSeekPoints == NULL) { + return MA_OUT_OF_MEMORY; + } + } + + mp3Result = ma_dr_mp3_calculate_seek_points(&pMP3->dr, &seekPointCount, pSeekPoints); + if (mp3Result != MA_TRUE) { + ma_free(pSeekPoints, pAllocationCallbacks); + return MA_ERROR; + } + + mp3Result = ma_dr_mp3_bind_seek_table(&pMP3->dr, seekPointCount, pSeekPoints); + if (mp3Result != MA_TRUE) { + ma_free(pSeekPoints, pAllocationCallbacks); + return MA_ERROR; + } + + pMP3->seekPointCount = seekPointCount; + pMP3->pSeekPoints = pSeekPoints; + + return MA_SUCCESS; +} + +static ma_result ma_mp3_post_init(ma_mp3* pMP3, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_result result; + + result = ma_mp3_generate_seek_table(pMP3, pConfig, pAllocationCallbacks); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_mp3_init(ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_mp3* pMP3) +{ + ma_result result; + + result = ma_mp3_init_internal(pConfig, pMP3); + if (result != MA_SUCCESS) { + return result; + } + + if (onRead == NULL || onSeek == NULL) { + return MA_INVALID_ARGS; /* onRead and onSeek are mandatory. */ + } + + pMP3->onRead = onRead; + pMP3->onSeek = onSeek; + pMP3->onTell = onTell; + pMP3->pReadSeekTellUserData = pReadSeekTellUserData; + + #if !defined(MA_NO_MP3) + { + ma_bool32 mp3Result; + + mp3Result = ma_dr_mp3_init(&pMP3->dr, ma_mp3_dr_callback__read, ma_mp3_dr_callback__seek, pMP3, pAllocationCallbacks); + if (mp3Result != MA_TRUE) { + return MA_INVALID_FILE; + } + + ma_mp3_post_init(pMP3, pConfig, pAllocationCallbacks); + + return MA_SUCCESS; + } + #else + { + /* mp3 is disabled. */ + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_mp3_init_file(const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_mp3* pMP3) +{ + ma_result result; + + result = ma_mp3_init_internal(pConfig, pMP3); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_MP3) + { + ma_bool32 mp3Result; + + mp3Result = ma_dr_mp3_init_file(&pMP3->dr, pFilePath, pAllocationCallbacks); + if (mp3Result != MA_TRUE) { + return MA_INVALID_FILE; + } + + ma_mp3_post_init(pMP3, pConfig, pAllocationCallbacks); + + return MA_SUCCESS; + } + #else + { + /* mp3 is disabled. */ + (void)pFilePath; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_mp3_init_file_w(const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_mp3* pMP3) +{ + ma_result result; + + result = ma_mp3_init_internal(pConfig, pMP3); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_MP3) + { + ma_bool32 mp3Result; + + mp3Result = ma_dr_mp3_init_file_w(&pMP3->dr, pFilePath, pAllocationCallbacks); + if (mp3Result != MA_TRUE) { + return MA_INVALID_FILE; + } + + ma_mp3_post_init(pMP3, pConfig, pAllocationCallbacks); + + return MA_SUCCESS; + } + #else + { + /* mp3 is disabled. */ + (void)pFilePath; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_mp3_init_memory(const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_mp3* pMP3) +{ + ma_result result; + + result = ma_mp3_init_internal(pConfig, pMP3); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_MP3) + { + ma_bool32 mp3Result; + + mp3Result = ma_dr_mp3_init_memory(&pMP3->dr, pData, dataSize, pAllocationCallbacks); + if (mp3Result != MA_TRUE) { + return MA_INVALID_FILE; + } + + ma_mp3_post_init(pMP3, pConfig, pAllocationCallbacks); + + return MA_SUCCESS; + } + #else + { + /* mp3 is disabled. */ + (void)pData; + (void)dataSize; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API void ma_mp3_uninit(ma_mp3* pMP3, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pMP3 == NULL) { + return; + } + + #if !defined(MA_NO_MP3) + { + ma_dr_mp3_uninit(&pMP3->dr); + } + #else + { + /* mp3 is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + } + #endif + + /* Seek points need to be freed after the MP3 decoder has been uninitialized to ensure they're no longer being referenced. */ + ma_free(pMP3->pSeekPoints, pAllocationCallbacks); + + ma_data_source_uninit(&pMP3->ds); +} + +MA_API ma_result ma_mp3_read_pcm_frames(ma_mp3* pMP3, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pMP3 == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_MP3) + { + /* We always use floating point format. */ + ma_result result = MA_SUCCESS; /* Must be initialized to MA_SUCCESS. */ + ma_uint64 totalFramesRead = 0; + ma_format format; + + ma_mp3_get_data_format(pMP3, &format, NULL, NULL, NULL, 0); + + switch (format) + { + case ma_format_f32: + { + totalFramesRead = ma_dr_mp3_read_pcm_frames_f32(&pMP3->dr, frameCount, (float*)pFramesOut); + } break; + + case ma_format_s16: + { + totalFramesRead = ma_dr_mp3_read_pcm_frames_s16(&pMP3->dr, frameCount, (ma_int16*)pFramesOut); + } break; + + case ma_format_u8: + case ma_format_s24: + case ma_format_s32: + case ma_format_unknown: + default: + { + return MA_INVALID_OPERATION; + }; + } + + /* In the future we'll update ma_dr_mp3 to return MA_AT_END for us. */ + if (totalFramesRead == 0) { + result = MA_AT_END; + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesRead; + } + + return result; + } + #else + { + /* mp3 is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + + (void)pFramesOut; + (void)frameCount; + (void)pFramesRead; + + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_mp3_seek_to_pcm_frame(ma_mp3* pMP3, ma_uint64 frameIndex) +{ + if (pMP3 == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_MP3) + { + ma_bool32 mp3Result; + + mp3Result = ma_dr_mp3_seek_to_pcm_frame(&pMP3->dr, frameIndex); + if (mp3Result != MA_TRUE) { + return MA_ERROR; + } + + return MA_SUCCESS; + } + #else + { + /* mp3 is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + + (void)frameIndex; + + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_mp3_get_data_format(ma_mp3* pMP3, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + /* Defaults for safety. */ + if (pFormat != NULL) { + *pFormat = ma_format_unknown; + } + if (pChannels != NULL) { + *pChannels = 0; + } + if (pSampleRate != NULL) { + *pSampleRate = 0; + } + if (pChannelMap != NULL) { + MA_ZERO_MEMORY(pChannelMap, sizeof(*pChannelMap) * channelMapCap); + } + + if (pMP3 == NULL) { + return MA_INVALID_OPERATION; + } + + if (pFormat != NULL) { + *pFormat = pMP3->format; + } + + #if !defined(MA_NO_MP3) + { + if (pChannels != NULL) { + *pChannels = pMP3->dr.channels; + } + + if (pSampleRate != NULL) { + *pSampleRate = pMP3->dr.sampleRate; + } + + if (pChannelMap != NULL) { + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pMP3->dr.channels); + } + + return MA_SUCCESS; + } + #else + { + /* mp3 is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_mp3_get_cursor_in_pcm_frames(ma_mp3* pMP3, ma_uint64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; /* Safety. */ + + if (pMP3 == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_MP3) + { + *pCursor = pMP3->dr.currentPCMFrame; + + return MA_SUCCESS; + } + #else + { + /* mp3 is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_mp3_get_length_in_pcm_frames(ma_mp3* pMP3, ma_uint64* pLength) +{ + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; /* Safety. */ + + if (pMP3 == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_MP3) + { + *pLength = ma_dr_mp3_get_pcm_frame_count(&pMP3->dr); + + return MA_SUCCESS; + } + #else + { + /* mp3 is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + + +static ma_result ma_decoding_backend_init__mp3(void* pUserData, ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_mp3* pMP3; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pMP3 = (ma_mp3*)ma_malloc(sizeof(*pMP3), pAllocationCallbacks); + if (pMP3 == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_mp3_init(onRead, onSeek, onTell, pReadSeekTellUserData, pConfig, pAllocationCallbacks, pMP3); + if (result != MA_SUCCESS) { + ma_free(pMP3, pAllocationCallbacks); + return result; + } + + *ppBackend = pMP3; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_file__mp3(void* pUserData, const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_mp3* pMP3; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pMP3 = (ma_mp3*)ma_malloc(sizeof(*pMP3), pAllocationCallbacks); + if (pMP3 == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_mp3_init_file(pFilePath, pConfig, pAllocationCallbacks, pMP3); + if (result != MA_SUCCESS) { + ma_free(pMP3, pAllocationCallbacks); + return result; + } + + *ppBackend = pMP3; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_file_w__mp3(void* pUserData, const wchar_t* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_mp3* pMP3; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pMP3 = (ma_mp3*)ma_malloc(sizeof(*pMP3), pAllocationCallbacks); + if (pMP3 == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_mp3_init_file_w(pFilePath, pConfig, pAllocationCallbacks, pMP3); + if (result != MA_SUCCESS) { + ma_free(pMP3, pAllocationCallbacks); + return result; + } + + *ppBackend = pMP3; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_memory__mp3(void* pUserData, const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_mp3* pMP3; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pMP3 = (ma_mp3*)ma_malloc(sizeof(*pMP3), pAllocationCallbacks); + if (pMP3 == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_mp3_init_memory(pData, dataSize, pConfig, pAllocationCallbacks, pMP3); + if (result != MA_SUCCESS) { + ma_free(pMP3, pAllocationCallbacks); + return result; + } + + *ppBackend = pMP3; + + return MA_SUCCESS; +} + +static void ma_decoding_backend_uninit__mp3(void* pUserData, ma_data_source* pBackend, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_mp3* pMP3 = (ma_mp3*)pBackend; + + (void)pUserData; + + ma_mp3_uninit(pMP3, pAllocationCallbacks); + ma_free(pMP3, pAllocationCallbacks); +} + +static ma_decoding_backend_vtable g_ma_decoding_backend_vtable_mp3 = +{ + ma_decoding_backend_init__mp3, + ma_decoding_backend_init_file__mp3, + ma_decoding_backend_init_file_w__mp3, + ma_decoding_backend_init_memory__mp3, + ma_decoding_backend_uninit__mp3 +}; + +static ma_result ma_decoder_init_mp3__internal(const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_vtable__internal(&g_ma_decoding_backend_vtable_mp3, NULL, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_mp3_from_file__internal(const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_file__internal(&g_ma_decoding_backend_vtable_mp3, NULL, pFilePath, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_mp3_from_file_w__internal(const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_file_w__internal(&g_ma_decoding_backend_vtable_mp3, NULL, pFilePath, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_mp3_from_memory__internal(const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_memory__internal(&g_ma_decoding_backend_vtable_mp3, NULL, pData, dataSize, pConfig, pDecoder); +} +#endif /* ma_dr_mp3_h */ + +/* Vorbis */ +#ifdef STB_VORBIS_INCLUDE_STB_VORBIS_H +#define MA_HAS_VORBIS + +/* The size in bytes of each chunk of data to read from the Vorbis stream. */ +#define MA_VORBIS_DATA_CHUNK_SIZE 4096 + +typedef struct +{ + ma_data_source_base ds; + ma_read_proc onRead; + ma_seek_proc onSeek; + ma_tell_proc onTell; + void* pReadSeekTellUserData; + ma_allocation_callbacks allocationCallbacks; /* Store the allocation callbacks within the structure because we may need to dynamically expand a buffer in ma_stbvorbis_read_pcm_frames() when using push mode. */ + ma_format format; /* Only f32 is allowed with stb_vorbis. */ + ma_uint32 channels; + ma_uint32 sampleRate; + ma_uint64 cursor; +#if !defined(MA_NO_VORBIS) + stb_vorbis* stb; + ma_bool32 usingPushMode; + struct + { + ma_uint8* pData; + size_t dataSize; + size_t dataCapacity; + size_t audioStartOffsetInBytes; + ma_uint32 framesConsumed; /* The number of frames consumed in ppPacketData. */ + ma_uint32 framesRemaining; /* The number of frames remaining in ppPacketData. */ + float** ppPacketData; + } push; +#endif +} ma_stbvorbis; + +MA_API ma_result ma_stbvorbis_init(ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_stbvorbis* pVorbis); +MA_API ma_result ma_stbvorbis_init_file(const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_stbvorbis* pVorbis); +MA_API ma_result ma_stbvorbis_init_memory(const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_stbvorbis* pVorbis); +MA_API void ma_stbvorbis_uninit(ma_stbvorbis* pVorbis, const ma_allocation_callbacks* pAllocationCallbacks); +MA_API ma_result ma_stbvorbis_read_pcm_frames(ma_stbvorbis* pVorbis, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead); +MA_API ma_result ma_stbvorbis_seek_to_pcm_frame(ma_stbvorbis* pVorbis, ma_uint64 frameIndex); +MA_API ma_result ma_stbvorbis_get_data_format(ma_stbvorbis* pVorbis, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap); +MA_API ma_result ma_stbvorbis_get_cursor_in_pcm_frames(ma_stbvorbis* pVorbis, ma_uint64* pCursor); +MA_API ma_result ma_stbvorbis_get_length_in_pcm_frames(ma_stbvorbis* pVorbis, ma_uint64* pLength); + + +static ma_result ma_stbvorbis_ds_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_stbvorbis_read_pcm_frames((ma_stbvorbis*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_stbvorbis_ds_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_stbvorbis_seek_to_pcm_frame((ma_stbvorbis*)pDataSource, frameIndex); +} + +static ma_result ma_stbvorbis_ds_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + return ma_stbvorbis_get_data_format((ma_stbvorbis*)pDataSource, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); +} + +static ma_result ma_stbvorbis_ds_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + return ma_stbvorbis_get_cursor_in_pcm_frames((ma_stbvorbis*)pDataSource, pCursor); +} + +static ma_result ma_stbvorbis_ds_get_length(ma_data_source* pDataSource, ma_uint64* pLength) +{ + return ma_stbvorbis_get_length_in_pcm_frames((ma_stbvorbis*)pDataSource, pLength); +} + +static ma_data_source_vtable g_ma_stbvorbis_ds_vtable = +{ + ma_stbvorbis_ds_read, + ma_stbvorbis_ds_seek, + ma_stbvorbis_ds_get_data_format, + ma_stbvorbis_ds_get_cursor, + ma_stbvorbis_ds_get_length, + NULL, /* onSetLooping */ + 0 +}; + + +static ma_result ma_stbvorbis_init_internal(const ma_decoding_backend_config* pConfig, ma_stbvorbis* pVorbis) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + + (void)pConfig; + + if (pVorbis == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pVorbis); + pVorbis->format = ma_format_f32; /* Only supporting f32. */ + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_stbvorbis_ds_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pVorbis->ds); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the base data source. */ + } + + return MA_SUCCESS; +} + +#if !defined(MA_NO_VORBIS) +static ma_result ma_stbvorbis_post_init(ma_stbvorbis* pVorbis) +{ + stb_vorbis_info info; + + MA_ASSERT(pVorbis != NULL); + + info = stb_vorbis_get_info(pVorbis->stb); + + pVorbis->channels = info.channels; + pVorbis->sampleRate = info.sample_rate; + + return MA_SUCCESS; +} + +static ma_result ma_stbvorbis_init_internal_decoder_push(ma_stbvorbis* pVorbis) +{ + ma_result result; + stb_vorbis* stb; + size_t dataSize = 0; + size_t dataCapacity = 0; + ma_uint8* pData = NULL; /* <-- Must be initialized to NULL. */ + + for (;;) { + int vorbisError; + int consumedDataSize; /* <-- Fill by stb_vorbis_open_pushdata(). */ + size_t bytesRead; + ma_uint8* pNewData; + + /* Allocate memory for the new chunk. */ + dataCapacity += MA_VORBIS_DATA_CHUNK_SIZE; + pNewData = (ma_uint8*)ma_realloc(pData, dataCapacity, &pVorbis->allocationCallbacks); + if (pNewData == NULL) { + ma_free(pData, &pVorbis->allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + + pData = pNewData; + + /* Read in the next chunk. */ + result = pVorbis->onRead(pVorbis->pReadSeekTellUserData, ma_offset_ptr(pData, dataSize), (dataCapacity - dataSize), &bytesRead); + dataSize += bytesRead; + + if (result != MA_SUCCESS) { + ma_free(pData, &pVorbis->allocationCallbacks); + return result; + } + + /* We have a maximum of 31 bits with stb_vorbis. */ + if (dataSize > INT_MAX) { + ma_free(pData, &pVorbis->allocationCallbacks); + return MA_TOO_BIG; + } + + stb = stb_vorbis_open_pushdata(pData, (int)dataSize, &consumedDataSize, &vorbisError, NULL); + if (stb != NULL) { + /* + Successfully opened the Vorbis decoder. We might have some leftover unprocessed + data so we'll need to move that down to the front. + */ + dataSize -= (size_t)consumedDataSize; /* Consume the data. */ + MA_MOVE_MEMORY(pData, ma_offset_ptr(pData, consumedDataSize), dataSize); + + /* + We need to track the start point so we can seek back to the start of the audio + data when seeking. + */ + pVorbis->push.audioStartOffsetInBytes = consumedDataSize; + + break; + } else { + /* Failed to open the decoder. */ + if (vorbisError == VORBIS_need_more_data) { + continue; + } else { + ma_free(pData, &pVorbis->allocationCallbacks); + return MA_ERROR; /* Failed to open the stb_vorbis decoder. */ + } + } + } + + MA_ASSERT(stb != NULL); + pVorbis->stb = stb; + pVorbis->push.pData = pData; + pVorbis->push.dataSize = dataSize; + pVorbis->push.dataCapacity = dataCapacity; + + return MA_SUCCESS; +} +#endif + +MA_API ma_result ma_stbvorbis_init(ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_stbvorbis* pVorbis) +{ + ma_result result; + + result = ma_stbvorbis_init_internal(pConfig, pVorbis); + if (result != MA_SUCCESS) { + return result; + } + + if (onRead == NULL || onSeek == NULL) { + return MA_INVALID_ARGS; /* onRead and onSeek are mandatory. */ + } + + pVorbis->onRead = onRead; + pVorbis->onSeek = onSeek; + pVorbis->onTell = onTell; + pVorbis->pReadSeekTellUserData = pReadSeekTellUserData; + ma_allocation_callbacks_init_copy(&pVorbis->allocationCallbacks, pAllocationCallbacks); + + #if !defined(MA_NO_VORBIS) + { + /* + stb_vorbis lacks a callback based API for its pulling API which means we're stuck with the + pushing API. In order for us to be able to successfully initialize the decoder we need to + supply it with enough data. We need to keep loading data until we have enough. + */ + result = ma_stbvorbis_init_internal_decoder_push(pVorbis); + if (result != MA_SUCCESS) { + return result; + } + + pVorbis->usingPushMode = MA_TRUE; + + result = ma_stbvorbis_post_init(pVorbis); + if (result != MA_SUCCESS) { + stb_vorbis_close(pVorbis->stb); + ma_free(pVorbis->push.pData, pAllocationCallbacks); + return result; + } + + return MA_SUCCESS; + } + #else + { + /* vorbis is disabled. */ + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_stbvorbis_init_file(const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_stbvorbis* pVorbis) +{ + ma_result result; + + result = ma_stbvorbis_init_internal(pConfig, pVorbis); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_VORBIS) + { + (void)pAllocationCallbacks; /* Don't know how to make use of this with stb_vorbis. */ + + /* We can use stb_vorbis' pull mode for file based streams. */ + pVorbis->stb = stb_vorbis_open_filename(pFilePath, NULL, NULL); + if (pVorbis->stb == NULL) { + return MA_INVALID_FILE; + } + + pVorbis->usingPushMode = MA_FALSE; + + result = ma_stbvorbis_post_init(pVorbis); + if (result != MA_SUCCESS) { + stb_vorbis_close(pVorbis->stb); + return result; + } + + return MA_SUCCESS; + } + #else + { + /* vorbis is disabled. */ + (void)pFilePath; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_stbvorbis_init_memory(const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_stbvorbis* pVorbis) +{ + ma_result result; + + result = ma_stbvorbis_init_internal(pConfig, pVorbis); + if (result != MA_SUCCESS) { + return result; + } + + #if !defined(MA_NO_VORBIS) + { + (void)pAllocationCallbacks; + + /* stb_vorbis uses an int as its size specifier, restricting it to 32-bit even on 64-bit systems. *sigh*. */ + if (dataSize > INT_MAX) { + return MA_TOO_BIG; + } + + pVorbis->stb = stb_vorbis_open_memory((const unsigned char*)pData, (int)dataSize, NULL, NULL); + if (pVorbis->stb == NULL) { + return MA_INVALID_FILE; + } + + pVorbis->usingPushMode = MA_FALSE; + + result = ma_stbvorbis_post_init(pVorbis); + if (result != MA_SUCCESS) { + stb_vorbis_close(pVorbis->stb); + return result; + } + + return MA_SUCCESS; + } + #else + { + /* vorbis is disabled. */ + (void)pData; + (void)dataSize; + (void)pAllocationCallbacks; + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API void ma_stbvorbis_uninit(ma_stbvorbis* pVorbis, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pVorbis == NULL) { + return; + } + + #if !defined(MA_NO_VORBIS) + { + stb_vorbis_close(pVorbis->stb); + + /* We'll have to clear some memory if we're using push mode. */ + if (pVorbis->usingPushMode) { + ma_free(pVorbis->push.pData, pAllocationCallbacks); + } + } + #else + { + /* vorbis is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + } + #endif + + ma_data_source_uninit(&pVorbis->ds); +} + +MA_API ma_result ma_stbvorbis_read_pcm_frames(ma_stbvorbis* pVorbis, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pVorbis == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_VORBIS) + { + /* We always use floating point format. */ + ma_result result = MA_SUCCESS; /* Must be initialized to MA_SUCCESS. */ + ma_uint64 totalFramesRead = 0; + ma_format format; + ma_uint32 channels; + + ma_stbvorbis_get_data_format(pVorbis, &format, &channels, NULL, NULL, 0); + + if (format == ma_format_f32) { + /* We read differently depending on whether or not we're using push mode. */ + if (pVorbis->usingPushMode) { + /* Push mode. This is the complex case. */ + float* pFramesOutF32 = (float*)pFramesOut; + + while (totalFramesRead < frameCount) { + /* The first thing to do is read from any already-cached frames. */ + ma_uint32 framesToReadFromCache = (ma_uint32)ma_min(pVorbis->push.framesRemaining, (frameCount - totalFramesRead)); /* Safe cast because pVorbis->framesRemaining is 32-bit. */ + + /* The output pointer can be null in which case we just treat it as a seek. */ + if (pFramesOut != NULL) { + ma_uint64 iFrame; + for (iFrame = 0; iFrame < framesToReadFromCache; iFrame += 1) { + ma_uint32 iChannel; + for (iChannel = 0; iChannel < pVorbis->channels; iChannel += 1) { + pFramesOutF32[iChannel] = pVorbis->push.ppPacketData[iChannel][pVorbis->push.framesConsumed + iFrame]; + } + + pFramesOutF32 += pVorbis->channels; + } + } + + /* Update pointers and counters. */ + pVorbis->push.framesConsumed += framesToReadFromCache; + pVorbis->push.framesRemaining -= framesToReadFromCache; + totalFramesRead += framesToReadFromCache; + + /* Don't bother reading any more frames right now if we've just finished loading. */ + if (totalFramesRead == frameCount) { + break; + } + + MA_ASSERT(pVorbis->push.framesRemaining == 0); + + /* Getting here means we've run out of cached frames. We'll need to load some more. */ + for (;;) { + int samplesRead = 0; + int consumedDataSize; + + /* We need to case dataSize to an int, so make sure we can do it safely. */ + if (pVorbis->push.dataSize > INT_MAX) { + break; /* Too big. */ + } + + consumedDataSize = stb_vorbis_decode_frame_pushdata(pVorbis->stb, pVorbis->push.pData, (int)pVorbis->push.dataSize, NULL, &pVorbis->push.ppPacketData, &samplesRead); + if (consumedDataSize != 0) { + /* Successfully decoded a Vorbis frame. Consume the data. */ + pVorbis->push.dataSize -= (size_t)consumedDataSize; + MA_MOVE_MEMORY(pVorbis->push.pData, ma_offset_ptr(pVorbis->push.pData, consumedDataSize), pVorbis->push.dataSize); + + pVorbis->push.framesConsumed = 0; + pVorbis->push.framesRemaining = samplesRead; + + break; + } else { + /* Not enough data. Read more. */ + size_t bytesRead; + + /* Expand the data buffer if necessary. */ + if (pVorbis->push.dataCapacity == pVorbis->push.dataSize) { + size_t newCap = pVorbis->push.dataCapacity + MA_VORBIS_DATA_CHUNK_SIZE; + ma_uint8* pNewData; + + pNewData = (ma_uint8*)ma_realloc(pVorbis->push.pData, newCap, &pVorbis->allocationCallbacks); + if (pNewData == NULL) { + result = MA_OUT_OF_MEMORY; + break; + } + + pVorbis->push.pData = pNewData; + pVorbis->push.dataCapacity = newCap; + } + + /* We should have enough room to load some data. */ + result = pVorbis->onRead(pVorbis->pReadSeekTellUserData, ma_offset_ptr(pVorbis->push.pData, pVorbis->push.dataSize), (pVorbis->push.dataCapacity - pVorbis->push.dataSize), &bytesRead); + pVorbis->push.dataSize += bytesRead; + + if (result != MA_SUCCESS) { + break; /* Failed to read any data. Get out. */ + } + } + } + + /* If we don't have a success code at this point it means we've encountered an error or the end of the file has been reached (probably the latter). */ + if (result != MA_SUCCESS) { + break; + } + } + } else { + /* Pull mode. This is the simple case, but we still need to run in a loop because stb_vorbis loves using 32-bit instead of 64-bit. */ + while (totalFramesRead < frameCount) { + ma_uint64 framesRemaining = (frameCount - totalFramesRead); + int framesRead; + + if (framesRemaining > INT_MAX) { + framesRemaining = INT_MAX; + } + + framesRead = stb_vorbis_get_samples_float_interleaved(pVorbis->stb, channels, (float*)ma_offset_pcm_frames_ptr(pFramesOut, totalFramesRead, format, channels), (int)framesRemaining * channels); /* Safe cast. */ + totalFramesRead += framesRead; + + if (framesRead < (int)framesRemaining) { + break; /* Nothing left to read. Get out. */ + } + } + } + } else { + result = MA_INVALID_ARGS; + } + + pVorbis->cursor += totalFramesRead; + + if (totalFramesRead == 0) { + result = MA_AT_END; + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesRead; + } + + if (result == MA_SUCCESS && totalFramesRead == 0) { + result = MA_AT_END; + } + + return result; + } + #else + { + /* vorbis is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + + (void)pFramesOut; + (void)frameCount; + (void)pFramesRead; + + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_stbvorbis_seek_to_pcm_frame(ma_stbvorbis* pVorbis, ma_uint64 frameIndex) +{ + if (pVorbis == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_VORBIS) + { + /* Different seeking methods depending on whether or not we're using push mode. */ + if (pVorbis->usingPushMode) { + /* Push mode. This is the complex case. */ + ma_result result; + float buffer[4096]; + + /* If we're seeking backwards, we need to seek back to the start and then brute-force forward. */ + if (frameIndex < pVorbis->cursor) { + if (frameIndex > 0x7FFFFFFF) { + return MA_INVALID_ARGS; /* Trying to seek beyond the 32-bit maximum of stb_vorbis. */ + } + + /* + This is wildly inefficient due to me having trouble getting sample exact seeking working + robustly with stb_vorbis_flush_pushdata(). The only way I can think to make this work + perfectly is to reinitialize the decoder. Note that we only enter this path when seeking + backwards. This will hopefully be removed once we get our own Vorbis decoder implemented. + */ + stb_vorbis_close(pVorbis->stb); + ma_free(pVorbis->push.pData, &pVorbis->allocationCallbacks); + + MA_ZERO_OBJECT(&pVorbis->push); + + /* Seek to the start of the file. */ + result = pVorbis->onSeek(pVorbis->pReadSeekTellUserData, 0, ma_seek_origin_start); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_stbvorbis_init_internal_decoder_push(pVorbis); + if (result != MA_SUCCESS) { + return result; + } + + /* At this point we should be sitting on the first frame. */ + pVorbis->cursor = 0; + } + + /* We're just brute-forcing this for now. */ + while (pVorbis->cursor < frameIndex) { + ma_uint64 framesRead; + ma_uint64 framesToRead = ma_countof(buffer)/pVorbis->channels; + if (framesToRead > (frameIndex - pVorbis->cursor)) { + framesToRead = (frameIndex - pVorbis->cursor); + } + + result = ma_stbvorbis_read_pcm_frames(pVorbis, buffer, framesToRead, &framesRead); + if (result != MA_SUCCESS) { + return result; + } + } + } else { + /* Pull mode. This is the simple case. */ + int vorbisResult; + + if (frameIndex > UINT_MAX) { + return MA_INVALID_ARGS; /* Trying to seek beyond the 32-bit maximum of stb_vorbis. */ + } + + vorbisResult = stb_vorbis_seek(pVorbis->stb, (unsigned int)frameIndex); /* Safe cast. */ + if (vorbisResult == 0) { + return MA_ERROR; /* See failed. */ + } + + pVorbis->cursor = frameIndex; + } + + return MA_SUCCESS; + } + #else + { + /* vorbis is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + + (void)frameIndex; + + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_stbvorbis_get_data_format(ma_stbvorbis* pVorbis, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + /* Defaults for safety. */ + if (pFormat != NULL) { + *pFormat = ma_format_unknown; + } + if (pChannels != NULL) { + *pChannels = 0; + } + if (pSampleRate != NULL) { + *pSampleRate = 0; + } + if (pChannelMap != NULL) { + MA_ZERO_MEMORY(pChannelMap, sizeof(*pChannelMap) * channelMapCap); + } + + if (pVorbis == NULL) { + return MA_INVALID_OPERATION; + } + + if (pFormat != NULL) { + *pFormat = pVorbis->format; + } + + #if !defined(MA_NO_VORBIS) + { + if (pChannels != NULL) { + *pChannels = pVorbis->channels; + } + + if (pSampleRate != NULL) { + *pSampleRate = pVorbis->sampleRate; + } + + if (pChannelMap != NULL) { + ma_channel_map_init_standard(ma_standard_channel_map_vorbis, pChannelMap, channelMapCap, pVorbis->channels); + } + + return MA_SUCCESS; + } + #else + { + /* vorbis is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_stbvorbis_get_cursor_in_pcm_frames(ma_stbvorbis* pVorbis, ma_uint64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; /* Safety. */ + + if (pVorbis == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_VORBIS) + { + *pCursor = pVorbis->cursor; + + return MA_SUCCESS; + } + #else + { + /* vorbis is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + +MA_API ma_result ma_stbvorbis_get_length_in_pcm_frames(ma_stbvorbis* pVorbis, ma_uint64* pLength) +{ + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; /* Safety. */ + + if (pVorbis == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_VORBIS) + { + if (pVorbis->usingPushMode) { + *pLength = 0; /* I don't know of a good way to determine this reliably with stb_vorbis and push mode. */ + } else { + *pLength = stb_vorbis_stream_length_in_samples(pVorbis->stb); + } + + return MA_SUCCESS; + } + #else + { + /* vorbis is disabled. Should never hit this since initialization would have failed. */ + MA_ASSERT(MA_FALSE); + return MA_NOT_IMPLEMENTED; + } + #endif +} + + +static ma_result ma_decoding_backend_init__stbvorbis(void* pUserData, ma_read_proc onRead, ma_seek_proc onSeek, ma_tell_proc onTell, void* pReadSeekTellUserData, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_stbvorbis* pVorbis; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pVorbis = (ma_stbvorbis*)ma_malloc(sizeof(*pVorbis), pAllocationCallbacks); + if (pVorbis == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_stbvorbis_init(onRead, onSeek, onTell, pReadSeekTellUserData, pConfig, pAllocationCallbacks, pVorbis); + if (result != MA_SUCCESS) { + ma_free(pVorbis, pAllocationCallbacks); + return result; + } + + *ppBackend = pVorbis; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_file__stbvorbis(void* pUserData, const char* pFilePath, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_stbvorbis* pVorbis; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pVorbis = (ma_stbvorbis*)ma_malloc(sizeof(*pVorbis), pAllocationCallbacks); + if (pVorbis == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_stbvorbis_init_file(pFilePath, pConfig, pAllocationCallbacks, pVorbis); + if (result != MA_SUCCESS) { + ma_free(pVorbis, pAllocationCallbacks); + return result; + } + + *ppBackend = pVorbis; + + return MA_SUCCESS; +} + +static ma_result ma_decoding_backend_init_memory__stbvorbis(void* pUserData, const void* pData, size_t dataSize, const ma_decoding_backend_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source** ppBackend) +{ + ma_result result; + ma_stbvorbis* pVorbis; + + (void)pUserData; /* For now not using pUserData, but once we start storing the vorbis decoder state within the ma_decoder structure this will be set to the decoder so we can avoid a malloc. */ + + /* For now we're just allocating the decoder backend on the heap. */ + pVorbis = (ma_stbvorbis*)ma_malloc(sizeof(*pVorbis), pAllocationCallbacks); + if (pVorbis == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_stbvorbis_init_memory(pData, dataSize, pConfig, pAllocationCallbacks, pVorbis); + if (result != MA_SUCCESS) { + ma_free(pVorbis, pAllocationCallbacks); + return result; + } + + *ppBackend = pVorbis; + + return MA_SUCCESS; +} + +static void ma_decoding_backend_uninit__stbvorbis(void* pUserData, ma_data_source* pBackend, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_stbvorbis* pVorbis = (ma_stbvorbis*)pBackend; + + (void)pUserData; + + ma_stbvorbis_uninit(pVorbis, pAllocationCallbacks); + ma_free(pVorbis, pAllocationCallbacks); +} + +static ma_decoding_backend_vtable g_ma_decoding_backend_vtable_stbvorbis = +{ + ma_decoding_backend_init__stbvorbis, + ma_decoding_backend_init_file__stbvorbis, + NULL, /* onInitFileW() */ + ma_decoding_backend_init_memory__stbvorbis, + ma_decoding_backend_uninit__stbvorbis +}; + +static ma_result ma_decoder_init_vorbis__internal(const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_vtable__internal(&g_ma_decoding_backend_vtable_stbvorbis, NULL, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_vorbis_from_file__internal(const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_file__internal(&g_ma_decoding_backend_vtable_stbvorbis, NULL, pFilePath, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_vorbis_from_file_w__internal(const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_file_w__internal(&g_ma_decoding_backend_vtable_stbvorbis, NULL, pFilePath, pConfig, pDecoder); +} + +static ma_result ma_decoder_init_vorbis_from_memory__internal(const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + return ma_decoder_init_from_memory__internal(&g_ma_decoding_backend_vtable_stbvorbis, NULL, pData, dataSize, pConfig, pDecoder); +} +#endif /* STB_VORBIS_INCLUDE_STB_VORBIS_H */ + + + +static ma_result ma_decoder__init_allocation_callbacks(const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + MA_ASSERT(pDecoder != NULL); + + if (pConfig != NULL) { + return ma_allocation_callbacks_init_copy(&pDecoder->allocationCallbacks, &pConfig->allocationCallbacks); + } else { + pDecoder->allocationCallbacks = ma_allocation_callbacks_init_default(); + return MA_SUCCESS; + } +} + +static ma_result ma_decoder__data_source_on_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_decoder_read_pcm_frames((ma_decoder*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_decoder__data_source_on_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_decoder_seek_to_pcm_frame((ma_decoder*)pDataSource, frameIndex); +} + +static ma_result ma_decoder__data_source_on_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + return ma_decoder_get_data_format((ma_decoder*)pDataSource, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); +} + +static ma_result ma_decoder__data_source_on_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + return ma_decoder_get_cursor_in_pcm_frames((ma_decoder*)pDataSource, pCursor); +} + +static ma_result ma_decoder__data_source_on_get_length(ma_data_source* pDataSource, ma_uint64* pLength) +{ + return ma_decoder_get_length_in_pcm_frames((ma_decoder*)pDataSource, pLength); +} + +static ma_data_source_vtable g_ma_decoder_data_source_vtable = +{ + ma_decoder__data_source_on_read, + ma_decoder__data_source_on_seek, + ma_decoder__data_source_on_get_data_format, + ma_decoder__data_source_on_get_cursor, + ma_decoder__data_source_on_get_length, + NULL, /* onSetLooping */ + 0 +}; + +static ma_result ma_decoder__preinit(ma_decoder_read_proc onRead, ma_decoder_seek_proc onSeek, ma_decoder_tell_proc onTell, void* pUserData, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + + MA_ASSERT(pConfig != NULL); + + if (pDecoder == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDecoder); + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_decoder_data_source_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pDecoder->ds); + if (result != MA_SUCCESS) { + return result; + } + + pDecoder->onRead = onRead; + pDecoder->onSeek = onSeek; + pDecoder->onTell = onTell; + pDecoder->pUserData = pUserData; + + result = ma_decoder__init_allocation_callbacks(pConfig, pDecoder); + if (result != MA_SUCCESS) { + ma_data_source_uninit(&pDecoder->ds); + return result; + } + + return MA_SUCCESS; +} + +static ma_result ma_decoder__postinit(const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + + result = ma_decoder__init_data_converter(pDecoder, pConfig); + + /* If we failed post initialization we need to uninitialize the decoder before returning to prevent a memory leak. */ + if (result != MA_SUCCESS) { + ma_decoder_uninit(pDecoder); + return result; + } + + return result; +} + + +static ma_result ma_decoder_init__internal(ma_decoder_read_proc onRead, ma_decoder_seek_proc onSeek, void* pUserData, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result = MA_NO_BACKEND; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pDecoder != NULL); + + /* Silence some warnings in the case that we don't have any decoder backends enabled. */ + (void)onRead; + (void)onSeek; + (void)pUserData; + + + /* If we've specified a specific encoding type, try that first. */ + if (pConfig->encodingFormat != ma_encoding_format_unknown) { + #ifdef MA_HAS_WAV + if (pConfig->encodingFormat == ma_encoding_format_wav) { + result = ma_decoder_init_wav__internal(pConfig, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (pConfig->encodingFormat == ma_encoding_format_flac) { + result = ma_decoder_init_flac__internal(pConfig, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (pConfig->encodingFormat == ma_encoding_format_mp3) { + result = ma_decoder_init_mp3__internal(pConfig, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (pConfig->encodingFormat == ma_encoding_format_vorbis) { + result = ma_decoder_init_vorbis__internal(pConfig, pDecoder); + } + #endif + + /* If we weren't able to initialize the decoder, seek back to the start to give the next attempts a clean start. */ + if (result != MA_SUCCESS) { + onSeek(pDecoder, 0, ma_seek_origin_start); + } + } + + if (result != MA_SUCCESS) { + /* Getting here means we couldn't load a specific decoding backend based on the encoding format. */ + + /* + We use trial and error to open a decoder. We prioritize custom decoders so that if they + implement the same encoding format they take priority over the built-in decoders. + */ + if (result != MA_SUCCESS) { + result = ma_decoder_init_custom__internal(pConfig, pDecoder); + if (result != MA_SUCCESS) { + onSeek(pDecoder, 0, ma_seek_origin_start); + } + } + + /* + If we get to this point and we still haven't found a decoder, and the caller has requested a + specific encoding format, there's no hope for it. Abort. + */ + if (pConfig->encodingFormat != ma_encoding_format_unknown) { + return MA_NO_BACKEND; + } + + #ifdef MA_HAS_WAV + if (result != MA_SUCCESS) { + result = ma_decoder_init_wav__internal(pConfig, pDecoder); + if (result != MA_SUCCESS) { + onSeek(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + #ifdef MA_HAS_FLAC + if (result != MA_SUCCESS) { + result = ma_decoder_init_flac__internal(pConfig, pDecoder); + if (result != MA_SUCCESS) { + onSeek(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + #ifdef MA_HAS_MP3 + if (result != MA_SUCCESS) { + result = ma_decoder_init_mp3__internal(pConfig, pDecoder); + if (result != MA_SUCCESS) { + onSeek(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + #ifdef MA_HAS_VORBIS + if (result != MA_SUCCESS) { + result = ma_decoder_init_vorbis__internal(pConfig, pDecoder); + if (result != MA_SUCCESS) { + onSeek(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + } + + if (result != MA_SUCCESS) { + return result; + } + + return ma_decoder__postinit(pConfig, pDecoder); +} + +MA_API ma_result ma_decoder_init(ma_decoder_read_proc onRead, ma_decoder_seek_proc onSeek, void* pUserData, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_decoder_config config; + ma_result result; + + config = ma_decoder_config_init_copy(pConfig); + + result = ma_decoder__preinit(onRead, onSeek, NULL, pUserData, &config, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + return ma_decoder_init__internal(onRead, onSeek, pUserData, &config, pDecoder); +} + + +static ma_result ma_decoder__on_read_memory(ma_decoder* pDecoder, void* pBufferOut, size_t bytesToRead, size_t* pBytesRead) +{ + size_t bytesRemaining; + + MA_ASSERT(pDecoder->data.memory.dataSize >= pDecoder->data.memory.currentReadPos); + + if (pBytesRead != NULL) { + *pBytesRead = 0; + } + + bytesRemaining = pDecoder->data.memory.dataSize - pDecoder->data.memory.currentReadPos; + if (bytesToRead > bytesRemaining) { + bytesToRead = bytesRemaining; + } + + if (bytesRemaining == 0) { + return MA_AT_END; + } + + if (bytesToRead > 0) { + MA_COPY_MEMORY(pBufferOut, pDecoder->data.memory.pData + pDecoder->data.memory.currentReadPos, bytesToRead); + pDecoder->data.memory.currentReadPos += bytesToRead; + } + + if (pBytesRead != NULL) { + *pBytesRead = bytesToRead; + } + + return MA_SUCCESS; +} + +static ma_result ma_decoder__on_seek_memory(ma_decoder* pDecoder, ma_int64 byteOffset, ma_seek_origin origin) +{ + if (byteOffset > 0 && (ma_uint64)byteOffset > MA_SIZE_MAX) { + return MA_BAD_SEEK; + } + + if (origin == ma_seek_origin_current) { + if (byteOffset > 0) { + if (pDecoder->data.memory.currentReadPos + byteOffset > pDecoder->data.memory.dataSize) { + byteOffset = (ma_int64)(pDecoder->data.memory.dataSize - pDecoder->data.memory.currentReadPos); /* Trying to seek too far forward. */ + } + + pDecoder->data.memory.currentReadPos += (size_t)byteOffset; + } else { + if (pDecoder->data.memory.currentReadPos < (size_t)-byteOffset) { + byteOffset = -(ma_int64)pDecoder->data.memory.currentReadPos; /* Trying to seek too far backwards. */ + } + + pDecoder->data.memory.currentReadPos -= (size_t)-byteOffset; + } + } else { + if (origin == ma_seek_origin_end) { + if (byteOffset < 0) { + byteOffset = -byteOffset; + } + + if (byteOffset > (ma_int64)pDecoder->data.memory.dataSize) { + pDecoder->data.memory.currentReadPos = 0; /* Trying to seek too far back. */ + } else { + pDecoder->data.memory.currentReadPos = pDecoder->data.memory.dataSize - (size_t)byteOffset; + } + } else { + if ((size_t)byteOffset <= pDecoder->data.memory.dataSize) { + pDecoder->data.memory.currentReadPos = (size_t)byteOffset; + } else { + pDecoder->data.memory.currentReadPos = pDecoder->data.memory.dataSize; /* Trying to seek too far forward. */ + } + } + } + + return MA_SUCCESS; +} + +static ma_result ma_decoder__on_tell_memory(ma_decoder* pDecoder, ma_int64* pCursor) +{ + MA_ASSERT(pDecoder != NULL); + MA_ASSERT(pCursor != NULL); + + *pCursor = (ma_int64)pDecoder->data.memory.currentReadPos; + + return MA_SUCCESS; +} + +static ma_result ma_decoder__preinit_memory_wrapper(const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result = ma_decoder__preinit(ma_decoder__on_read_memory, ma_decoder__on_seek_memory, ma_decoder__on_tell_memory, NULL, pConfig, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + if (pData == NULL || dataSize == 0) { + return MA_INVALID_ARGS; + } + + pDecoder->data.memory.pData = (const ma_uint8*)pData; + pDecoder->data.memory.dataSize = dataSize; + pDecoder->data.memory.currentReadPos = 0; + + (void)pConfig; + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_init_memory(const void* pData, size_t dataSize, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoder_config config; + + config = ma_decoder_config_init_copy(pConfig); + + result = ma_decoder__preinit(NULL, NULL, NULL, NULL, &config, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + if (pData == NULL || dataSize == 0) { + return MA_INVALID_ARGS; + } + + /* If the backend has support for loading from a file path we'll want to use that. If that all fails we'll fall back to the VFS path. */ + result = MA_NO_BACKEND; + + if (config.encodingFormat != ma_encoding_format_unknown) { + #ifdef MA_HAS_WAV + if (config.encodingFormat == ma_encoding_format_wav) { + result = ma_decoder_init_wav_from_memory__internal(pData, dataSize, &config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (config.encodingFormat == ma_encoding_format_flac) { + result = ma_decoder_init_flac_from_memory__internal(pData, dataSize, &config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (config.encodingFormat == ma_encoding_format_mp3) { + result = ma_decoder_init_mp3_from_memory__internal(pData, dataSize, &config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (config.encodingFormat == ma_encoding_format_vorbis) { + result = ma_decoder_init_vorbis_from_memory__internal(pData, dataSize, &config, pDecoder); + } + #endif + } + + if (result != MA_SUCCESS) { + /* Getting here means we weren't able to initialize a decoder of a specific encoding format. */ + + /* + We use trial and error to open a decoder. We prioritize custom decoders so that if they + implement the same encoding format they take priority over the built-in decoders. + */ + result = ma_decoder_init_custom_from_memory__internal(pData, dataSize, &config, pDecoder); + + /* + If we get to this point and we still haven't found a decoder, and the caller has requested a + specific encoding format, there's no hope for it. Abort. + */ + if (result != MA_SUCCESS && config.encodingFormat != ma_encoding_format_unknown) { + return MA_NO_BACKEND; + } + + /* Use trial and error for stock decoders. */ + if (result != MA_SUCCESS) { + #ifdef MA_HAS_WAV + if (result != MA_SUCCESS) { + result = ma_decoder_init_wav_from_memory__internal(pData, dataSize, &config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (result != MA_SUCCESS) { + result = ma_decoder_init_flac_from_memory__internal(pData, dataSize, &config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (result != MA_SUCCESS) { + result = ma_decoder_init_mp3_from_memory__internal(pData, dataSize, &config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (result != MA_SUCCESS) { + result = ma_decoder_init_vorbis_from_memory__internal(pData, dataSize, &config, pDecoder); + } + #endif + } + } + + /* + If at this point we still haven't successfully initialized the decoder it most likely means + the backend doesn't have an implementation for loading from a file path. We'll try using + miniaudio's built-in file IO for loading file. + */ + if (result == MA_SUCCESS) { + /* Initialization was successful. Finish up. */ + result = ma_decoder__postinit(&config, pDecoder); + if (result != MA_SUCCESS) { + /* + The backend was initialized successfully, but for some reason post-initialization failed. This is most likely + due to an out of memory error. We're going to abort with an error here and not try to recover. + */ + if (pDecoder->pBackendVTable != NULL && pDecoder->pBackendVTable->onUninit != NULL) { + pDecoder->pBackendVTable->onUninit(pDecoder->pBackendUserData, &pDecoder->pBackend, &pDecoder->allocationCallbacks); + } + + return result; + } + } else { + /* Probably no implementation for loading from a block of memory. Use miniaudio's abstraction instead. */ + result = ma_decoder__preinit_memory_wrapper(pData, dataSize, &config, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_decoder_init__internal(ma_decoder__on_read_memory, ma_decoder__on_seek_memory, NULL, &config, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + + +#if defined(MA_HAS_WAV) || \ + defined(MA_HAS_MP3) || \ + defined(MA_HAS_FLAC) || \ + defined(MA_HAS_VORBIS) +#define MA_HAS_PATH_API +#endif + +#if defined(MA_HAS_PATH_API) +static const char* ma_path_file_name(const char* path) +{ + const char* fileName; + + if (path == NULL) { + return NULL; + } + + fileName = path; + + /* We just loop through the path until we find the last slash. */ + while (path[0] != '\0') { + if (path[0] == '/' || path[0] == '\\') { + fileName = path; + } + + path += 1; + } + + /* At this point the file name is sitting on a slash, so just move forward. */ + while (fileName[0] != '\0' && (fileName[0] == '/' || fileName[0] == '\\')) { + fileName += 1; + } + + return fileName; +} + +static const wchar_t* ma_path_file_name_w(const wchar_t* path) +{ + const wchar_t* fileName; + + if (path == NULL) { + return NULL; + } + + fileName = path; + + /* We just loop through the path until we find the last slash. */ + while (path[0] != '\0') { + if (path[0] == '/' || path[0] == '\\') { + fileName = path; + } + + path += 1; + } + + /* At this point the file name is sitting on a slash, so just move forward. */ + while (fileName[0] != '\0' && (fileName[0] == '/' || fileName[0] == '\\')) { + fileName += 1; + } + + return fileName; +} + + +static const char* ma_path_extension(const char* path) +{ + const char* extension; + const char* lastOccurance; + + if (path == NULL) { + path = ""; + } + + extension = ma_path_file_name(path); + lastOccurance = NULL; + + /* Just find the last '.' and return. */ + while (extension[0] != '\0') { + if (extension[0] == '.') { + extension += 1; + lastOccurance = extension; + } + + extension += 1; + } + + return (lastOccurance != NULL) ? lastOccurance : extension; +} + +static const wchar_t* ma_path_extension_w(const wchar_t* path) +{ + const wchar_t* extension; + const wchar_t* lastOccurance; + + if (path == NULL) { + path = L""; + } + + extension = ma_path_file_name_w(path); + lastOccurance = NULL; + + /* Just find the last '.' and return. */ + while (extension[0] != '\0') { + if (extension[0] == '.') { + extension += 1; + lastOccurance = extension; + } + + extension += 1; + } + + return (lastOccurance != NULL) ? lastOccurance : extension; +} + + +static ma_bool32 ma_path_extension_equal(const char* path, const char* extension) +{ + const char* ext1; + const char* ext2; + + if (path == NULL || extension == NULL) { + return MA_FALSE; + } + + ext1 = extension; + ext2 = ma_path_extension(path); + +#if defined(_MSC_VER) || defined(__DMC__) + return _stricmp(ext1, ext2) == 0; +#else + return strcasecmp(ext1, ext2) == 0; +#endif +} + +static ma_bool32 ma_path_extension_equal_w(const wchar_t* path, const wchar_t* extension) +{ + const wchar_t* ext1; + const wchar_t* ext2; + + if (path == NULL || extension == NULL) { + return MA_FALSE; + } + + ext1 = extension; + ext2 = ma_path_extension_w(path); + +#if defined(_MSC_VER) || defined(__WATCOMC__) || defined(__DMC__) + return _wcsicmp(ext1, ext2) == 0; +#else + /* + I'm not aware of a wide character version of strcasecmp(). I'm therefore converting the extensions to multibyte strings and comparing those. This + isn't the most efficient way to do it, but it should work OK. + */ + { + char ext1MB[4096]; + char ext2MB[4096]; + const wchar_t* pext1 = ext1; + const wchar_t* pext2 = ext2; + mbstate_t mbs1; + mbstate_t mbs2; + + MA_ZERO_OBJECT(&mbs1); + MA_ZERO_OBJECT(&mbs2); + + if (wcsrtombs(ext1MB, &pext1, sizeof(ext1MB), &mbs1) == (size_t)-1) { + return MA_FALSE; + } + if (wcsrtombs(ext2MB, &pext2, sizeof(ext2MB), &mbs2) == (size_t)-1) { + return MA_FALSE; + } + + return strcasecmp(ext1MB, ext2MB) == 0; + } +#endif +} +#endif /* MA_HAS_PATH_API */ + + + +static ma_result ma_decoder__on_read_vfs(ma_decoder* pDecoder, void* pBufferOut, size_t bytesToRead, size_t* pBytesRead) +{ + MA_ASSERT(pDecoder != NULL); + MA_ASSERT(pBufferOut != NULL); + + return ma_vfs_or_default_read(pDecoder->data.vfs.pVFS, pDecoder->data.vfs.file, pBufferOut, bytesToRead, pBytesRead); +} + +static ma_result ma_decoder__on_seek_vfs(ma_decoder* pDecoder, ma_int64 offset, ma_seek_origin origin) +{ + MA_ASSERT(pDecoder != NULL); + + return ma_vfs_or_default_seek(pDecoder->data.vfs.pVFS, pDecoder->data.vfs.file, offset, origin); +} + +static ma_result ma_decoder__on_tell_vfs(ma_decoder* pDecoder, ma_int64* pCursor) +{ + MA_ASSERT(pDecoder != NULL); + + return ma_vfs_or_default_tell(pDecoder->data.vfs.pVFS, pDecoder->data.vfs.file, pCursor); +} + +static ma_result ma_decoder__preinit_vfs(ma_vfs* pVFS, const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_vfs_file file; + + result = ma_decoder__preinit(ma_decoder__on_read_vfs, ma_decoder__on_seek_vfs, ma_decoder__on_tell_vfs, NULL, pConfig, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + if (pFilePath == NULL || pFilePath[0] == '\0') { + return MA_INVALID_ARGS; + } + + result = ma_vfs_or_default_open(pVFS, pFilePath, MA_OPEN_MODE_READ, &file); + if (result != MA_SUCCESS) { + return result; + } + + pDecoder->data.vfs.pVFS = pVFS; + pDecoder->data.vfs.file = file; + + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_init_vfs(ma_vfs* pVFS, const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoder_config config; + + config = ma_decoder_config_init_copy(pConfig); + result = ma_decoder__preinit_vfs(pVFS, pFilePath, &config, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + result = MA_NO_BACKEND; + + if (config.encodingFormat != ma_encoding_format_unknown) { + #ifdef MA_HAS_WAV + if (config.encodingFormat == ma_encoding_format_wav) { + result = ma_decoder_init_wav__internal(&config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (config.encodingFormat == ma_encoding_format_flac) { + result = ma_decoder_init_flac__internal(&config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (config.encodingFormat == ma_encoding_format_mp3) { + result = ma_decoder_init_mp3__internal(&config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (config.encodingFormat == ma_encoding_format_vorbis) { + result = ma_decoder_init_vorbis__internal(&config, pDecoder); + } + #endif + + /* Make sure we seek back to the start if we didn't initialize a decoder successfully so the next attempts have a fresh start. */ + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + + if (result != MA_SUCCESS) { + /* Getting here means we weren't able to initialize a decoder of a specific encoding format. */ + + /* + We use trial and error to open a decoder. We prioritize custom decoders so that if they + implement the same encoding format they take priority over the built-in decoders. + */ + if (result != MA_SUCCESS) { + result = ma_decoder_init_custom__internal(&config, pDecoder); + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + + /* + If we get to this point and we still haven't found a decoder, and the caller has requested a + specific encoding format, there's no hope for it. Abort. + */ + if (config.encodingFormat != ma_encoding_format_unknown) { + return MA_NO_BACKEND; + } + + #ifdef MA_HAS_WAV + if (result != MA_SUCCESS && ma_path_extension_equal(pFilePath, "wav")) { + result = ma_decoder_init_wav__internal(&config, pDecoder); + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + #ifdef MA_HAS_FLAC + if (result != MA_SUCCESS && ma_path_extension_equal(pFilePath, "flac")) { + result = ma_decoder_init_flac__internal(&config, pDecoder); + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + #ifdef MA_HAS_MP3 + if (result != MA_SUCCESS && ma_path_extension_equal(pFilePath, "mp3")) { + result = ma_decoder_init_mp3__internal(&config, pDecoder); + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + } + + /* If we still haven't got a result just use trial and error. Otherwise we can finish up. */ + if (result != MA_SUCCESS) { + result = ma_decoder_init__internal(ma_decoder__on_read_vfs, ma_decoder__on_seek_vfs, NULL, &config, pDecoder); + } else { + result = ma_decoder__postinit(&config, pDecoder); + } + + if (result != MA_SUCCESS) { + if (pDecoder->data.vfs.file != NULL) { /* <-- Will be reset to NULL if ma_decoder_uninit() is called in one of the steps above which allows us to avoid a double close of the file. */ + ma_vfs_or_default_close(pVFS, pDecoder->data.vfs.file); + } + + return result; + } + + return MA_SUCCESS; +} + + +static ma_result ma_decoder__preinit_vfs_w(ma_vfs* pVFS, const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_vfs_file file; + + result = ma_decoder__preinit(ma_decoder__on_read_vfs, ma_decoder__on_seek_vfs, ma_decoder__on_tell_vfs, NULL, pConfig, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + if (pFilePath == NULL || pFilePath[0] == '\0') { + return MA_INVALID_ARGS; + } + + result = ma_vfs_or_default_open_w(pVFS, pFilePath, MA_OPEN_MODE_READ, &file); + if (result != MA_SUCCESS) { + return result; + } + + pDecoder->data.vfs.pVFS = pVFS; + pDecoder->data.vfs.file = file; + + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_init_vfs_w(ma_vfs* pVFS, const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoder_config config; + + config = ma_decoder_config_init_copy(pConfig); + result = ma_decoder__preinit_vfs_w(pVFS, pFilePath, &config, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + result = MA_NO_BACKEND; + + if (config.encodingFormat != ma_encoding_format_unknown) { + #ifdef MA_HAS_WAV + if (config.encodingFormat == ma_encoding_format_wav) { + result = ma_decoder_init_wav__internal(&config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (config.encodingFormat == ma_encoding_format_flac) { + result = ma_decoder_init_flac__internal(&config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (config.encodingFormat == ma_encoding_format_mp3) { + result = ma_decoder_init_mp3__internal(&config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (config.encodingFormat == ma_encoding_format_vorbis) { + result = ma_decoder_init_vorbis__internal(&config, pDecoder); + } + #endif + + /* Make sure we seek back to the start if we didn't initialize a decoder successfully so the next attempts have a fresh start. */ + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + + if (result != MA_SUCCESS) { + /* Getting here means we weren't able to initialize a decoder of a specific encoding format. */ + + /* + We use trial and error to open a decoder. We prioritize custom decoders so that if they + implement the same encoding format they take priority over the built-in decoders. + */ + if (result != MA_SUCCESS) { + result = ma_decoder_init_custom__internal(&config, pDecoder); + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + + /* + If we get to this point and we still haven't found a decoder, and the caller has requested a + specific encoding format, there's no hope for it. Abort. + */ + if (config.encodingFormat != ma_encoding_format_unknown) { + return MA_NO_BACKEND; + } + + #ifdef MA_HAS_WAV + if (result != MA_SUCCESS && ma_path_extension_equal_w(pFilePath, L"wav")) { + result = ma_decoder_init_wav__internal(&config, pDecoder); + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + #ifdef MA_HAS_FLAC + if (result != MA_SUCCESS && ma_path_extension_equal_w(pFilePath, L"flac")) { + result = ma_decoder_init_flac__internal(&config, pDecoder); + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + #ifdef MA_HAS_MP3 + if (result != MA_SUCCESS && ma_path_extension_equal_w(pFilePath, L"mp3")) { + result = ma_decoder_init_mp3__internal(&config, pDecoder); + if (result != MA_SUCCESS) { + ma_decoder__on_seek_vfs(pDecoder, 0, ma_seek_origin_start); + } + } + #endif + } + + /* If we still haven't got a result just use trial and error. Otherwise we can finish up. */ + if (result != MA_SUCCESS) { + result = ma_decoder_init__internal(ma_decoder__on_read_vfs, ma_decoder__on_seek_vfs, NULL, &config, pDecoder); + } else { + result = ma_decoder__postinit(&config, pDecoder); + } + + if (result != MA_SUCCESS) { + ma_vfs_or_default_close(pVFS, pDecoder->data.vfs.file); + return result; + } + + return MA_SUCCESS; +} + + +static ma_result ma_decoder__preinit_file(const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + + result = ma_decoder__preinit(NULL, NULL, NULL, NULL, pConfig, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + if (pFilePath == NULL || pFilePath[0] == '\0') { + return MA_INVALID_ARGS; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_init_file(const char* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoder_config config; + + config = ma_decoder_config_init_copy(pConfig); + result = ma_decoder__preinit_file(pFilePath, &config, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + /* If the backend has support for loading from a file path we'll want to use that. If that all fails we'll fall back to the VFS path. */ + result = MA_NO_BACKEND; + + if (config.encodingFormat != ma_encoding_format_unknown) { + #ifdef MA_HAS_WAV + if (config.encodingFormat == ma_encoding_format_wav) { + result = ma_decoder_init_wav_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (config.encodingFormat == ma_encoding_format_flac) { + result = ma_decoder_init_flac_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (config.encodingFormat == ma_encoding_format_mp3) { + result = ma_decoder_init_mp3_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (config.encodingFormat == ma_encoding_format_vorbis) { + result = ma_decoder_init_vorbis_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + } + + if (result != MA_SUCCESS) { + /* Getting here means we weren't able to initialize a decoder of a specific encoding format. */ + + /* + We use trial and error to open a decoder. We prioritize custom decoders so that if they + implement the same encoding format they take priority over the built-in decoders. + */ + result = ma_decoder_init_custom_from_file__internal(pFilePath, &config, pDecoder); + + /* + If we get to this point and we still haven't found a decoder, and the caller has requested a + specific encoding format, there's no hope for it. Abort. + */ + if (result != MA_SUCCESS && config.encodingFormat != ma_encoding_format_unknown) { + return MA_NO_BACKEND; + } + + /* First try loading based on the file extension so we don't waste time opening and closing files. */ + #ifdef MA_HAS_WAV + if (result != MA_SUCCESS && ma_path_extension_equal(pFilePath, "wav")) { + result = ma_decoder_init_wav_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (result != MA_SUCCESS && ma_path_extension_equal(pFilePath, "flac")) { + result = ma_decoder_init_flac_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (result != MA_SUCCESS && ma_path_extension_equal(pFilePath, "mp3")) { + result = ma_decoder_init_mp3_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (result != MA_SUCCESS && ma_path_extension_equal(pFilePath, "ogg")) { + result = ma_decoder_init_vorbis_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + + /* + If we still haven't got a result just use trial and error. Custom decoders have already been attempted, so here we + need only iterate over our stock decoders. + */ + if (result != MA_SUCCESS) { + #ifdef MA_HAS_WAV + if (result != MA_SUCCESS) { + result = ma_decoder_init_wav_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (result != MA_SUCCESS) { + result = ma_decoder_init_flac_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (result != MA_SUCCESS) { + result = ma_decoder_init_mp3_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (result != MA_SUCCESS) { + result = ma_decoder_init_vorbis_from_file__internal(pFilePath, &config, pDecoder); + } + #endif + } + } + + /* + If at this point we still haven't successfully initialized the decoder it most likely means + the backend doesn't have an implementation for loading from a file path. We'll try using + miniaudio's built-in file IO for loading file. + */ + if (result == MA_SUCCESS) { + /* Initialization was successful. Finish up. */ + result = ma_decoder__postinit(&config, pDecoder); + if (result != MA_SUCCESS) { + /* + The backend was initialized successfully, but for some reason post-initialization failed. This is most likely + due to an out of memory error. We're going to abort with an error here and not try to recover. + */ + if (pDecoder->pBackendVTable != NULL && pDecoder->pBackendVTable->onUninit != NULL) { + pDecoder->pBackendVTable->onUninit(pDecoder->pBackendUserData, &pDecoder->pBackend, &pDecoder->allocationCallbacks); + } + + return result; + } + } else { + /* Probably no implementation for loading from a file path. Use miniaudio's file IO instead. */ + result = ma_decoder_init_vfs(NULL, pFilePath, pConfig, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +static ma_result ma_decoder__preinit_file_w(const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + + result = ma_decoder__preinit(NULL, NULL, NULL, NULL, pConfig, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + if (pFilePath == NULL || pFilePath[0] == '\0') { + return MA_INVALID_ARGS; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_init_file_w(const wchar_t* pFilePath, const ma_decoder_config* pConfig, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoder_config config; + + config = ma_decoder_config_init_copy(pConfig); + result = ma_decoder__preinit_file_w(pFilePath, &config, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + + /* If the backend has support for loading from a file path we'll want to use that. If that all fails we'll fall back to the VFS path. */ + result = MA_NO_BACKEND; + + if (config.encodingFormat != ma_encoding_format_unknown) { + #ifdef MA_HAS_WAV + if (config.encodingFormat == ma_encoding_format_wav) { + result = ma_decoder_init_wav_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (config.encodingFormat == ma_encoding_format_flac) { + result = ma_decoder_init_flac_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (config.encodingFormat == ma_encoding_format_mp3) { + result = ma_decoder_init_mp3_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (config.encodingFormat == ma_encoding_format_vorbis) { + result = ma_decoder_init_vorbis_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + } + + if (result != MA_SUCCESS) { + /* Getting here means we weren't able to initialize a decoder of a specific encoding format. */ + + /* + We use trial and error to open a decoder. We prioritize custom decoders so that if they + implement the same encoding format they take priority over the built-in decoders. + */ + result = ma_decoder_init_custom_from_file_w__internal(pFilePath, &config, pDecoder); + + /* + If we get to this point and we still haven't found a decoder, and the caller has requested a + specific encoding format, there's no hope for it. Abort. + */ + if (result != MA_SUCCESS && config.encodingFormat != ma_encoding_format_unknown) { + return MA_NO_BACKEND; + } + + /* First try loading based on the file extension so we don't waste time opening and closing files. */ + #ifdef MA_HAS_WAV + if (result != MA_SUCCESS && ma_path_extension_equal_w(pFilePath, L"wav")) { + result = ma_decoder_init_wav_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (result != MA_SUCCESS && ma_path_extension_equal_w(pFilePath, L"flac")) { + result = ma_decoder_init_flac_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (result != MA_SUCCESS && ma_path_extension_equal_w(pFilePath, L"mp3")) { + result = ma_decoder_init_mp3_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (result != MA_SUCCESS && ma_path_extension_equal_w(pFilePath, L"ogg")) { + result = ma_decoder_init_vorbis_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + + /* + If we still haven't got a result just use trial and error. Custom decoders have already been attempted, so here we + need only iterate over our stock decoders. + */ + if (result != MA_SUCCESS) { + #ifdef MA_HAS_WAV + if (result != MA_SUCCESS) { + result = ma_decoder_init_wav_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_FLAC + if (result != MA_SUCCESS) { + result = ma_decoder_init_flac_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_MP3 + if (result != MA_SUCCESS) { + result = ma_decoder_init_mp3_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + #ifdef MA_HAS_VORBIS + if (result != MA_SUCCESS) { + result = ma_decoder_init_vorbis_from_file_w__internal(pFilePath, &config, pDecoder); + } + #endif + } + } + + /* + If at this point we still haven't successfully initialized the decoder it most likely means + the backend doesn't have an implementation for loading from a file path. We'll try using + miniaudio's built-in file IO for loading file. + */ + if (result == MA_SUCCESS) { + /* Initialization was successful. Finish up. */ + result = ma_decoder__postinit(&config, pDecoder); + if (result != MA_SUCCESS) { + /* + The backend was initialized successfully, but for some reason post-initialization failed. This is most likely + due to an out of memory error. We're going to abort with an error here and not try to recover. + */ + if (pDecoder->pBackendVTable != NULL && pDecoder->pBackendVTable->onUninit != NULL) { + pDecoder->pBackendVTable->onUninit(pDecoder->pBackendUserData, &pDecoder->pBackend, &pDecoder->allocationCallbacks); + } + + return result; + } + } else { + /* Probably no implementation for loading from a file path. Use miniaudio's file IO instead. */ + result = ma_decoder_init_vfs_w(NULL, pFilePath, pConfig, pDecoder); + if (result != MA_SUCCESS) { + return result; + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_uninit(ma_decoder* pDecoder) +{ + if (pDecoder == NULL) { + return MA_INVALID_ARGS; + } + + if (pDecoder->pBackend != NULL) { + if (pDecoder->pBackendVTable != NULL && pDecoder->pBackendVTable->onUninit != NULL) { + pDecoder->pBackendVTable->onUninit(pDecoder->pBackendUserData, pDecoder->pBackend, &pDecoder->allocationCallbacks); + } + } + + if (pDecoder->onRead == ma_decoder__on_read_vfs) { + ma_vfs_or_default_close(pDecoder->data.vfs.pVFS, pDecoder->data.vfs.file); + pDecoder->data.vfs.file = NULL; + } + + ma_data_converter_uninit(&pDecoder->converter, &pDecoder->allocationCallbacks); + ma_data_source_uninit(&pDecoder->ds); + + if (pDecoder->pInputCache != NULL) { + ma_free(pDecoder->pInputCache, &pDecoder->allocationCallbacks); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_read_pcm_frames(ma_decoder* pDecoder, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_result result = MA_SUCCESS; + ma_uint64 totalFramesReadOut; + void* pRunningFramesOut; + + if (pFramesRead != NULL) { + *pFramesRead = 0; /* Safety. */ + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pDecoder == NULL) { + return MA_INVALID_ARGS; + } + + if (pDecoder->pBackend == NULL) { + return MA_INVALID_OPERATION; + } + + /* Fast path. */ + if (pDecoder->converter.isPassthrough) { + result = ma_data_source_read_pcm_frames(pDecoder->pBackend, pFramesOut, frameCount, &totalFramesReadOut); + } else { + /* + Getting here means we need to do data conversion. If we're seeking forward and are _not_ doing resampling we can run this in a fast path. If we're doing resampling we + need to run through each sample because we need to ensure its internal cache is updated. + */ + if (pFramesOut == NULL && pDecoder->converter.hasResampler == MA_FALSE) { + result = ma_data_source_read_pcm_frames(pDecoder->pBackend, NULL, frameCount, &totalFramesReadOut); + } else { + /* Slow path. Need to run everything through the data converter. */ + ma_format internalFormat; + ma_uint32 internalChannels; + + totalFramesReadOut = 0; + pRunningFramesOut = pFramesOut; + + result = ma_data_source_get_data_format(pDecoder->pBackend, &internalFormat, &internalChannels, NULL, NULL, 0); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the internal format and channel count. */ + } + + /* + We run a different path depending on whether or not we are using a heap-allocated + intermediary buffer or not. If the data converter does not support the calculation of + the required number of input frames, we'll use the heap-allocated path. Otherwise we'll + use the stack-allocated path. + */ + if (pDecoder->pInputCache != NULL) { + /* We don't have a way of determining the required number of input frames, so need to persistently store input data in a cache. */ + while (totalFramesReadOut < frameCount) { + ma_uint64 framesToReadThisIterationIn; + ma_uint64 framesToReadThisIterationOut; + + /* If there's any data available in the cache, that needs to get processed first. */ + if (pDecoder->inputCacheRemaining > 0) { + framesToReadThisIterationOut = (frameCount - totalFramesReadOut); + framesToReadThisIterationIn = framesToReadThisIterationOut; + if (framesToReadThisIterationIn > pDecoder->inputCacheRemaining) { + framesToReadThisIterationIn = pDecoder->inputCacheRemaining; + } + + result = ma_data_converter_process_pcm_frames(&pDecoder->converter, ma_offset_pcm_frames_ptr(pDecoder->pInputCache, pDecoder->inputCacheConsumed, internalFormat, internalChannels), &framesToReadThisIterationIn, pRunningFramesOut, &framesToReadThisIterationOut); + if (result != MA_SUCCESS) { + break; + } + + pDecoder->inputCacheConsumed += framesToReadThisIterationIn; + pDecoder->inputCacheRemaining -= framesToReadThisIterationIn; + + totalFramesReadOut += framesToReadThisIterationOut; + + if (pRunningFramesOut != NULL) { + pRunningFramesOut = ma_offset_ptr(pRunningFramesOut, framesToReadThisIterationOut * ma_get_bytes_per_frame(pDecoder->outputFormat, pDecoder->outputChannels)); + } + + if (framesToReadThisIterationIn == 0 && framesToReadThisIterationOut == 0) { + break; /* We're done. */ + } + } + + /* Getting here means there's no data in the cache and we need to fill it up from the data source. */ + if (pDecoder->inputCacheRemaining == 0) { + pDecoder->inputCacheConsumed = 0; + + result = ma_data_source_read_pcm_frames(pDecoder->pBackend, pDecoder->pInputCache, pDecoder->inputCacheCap, &pDecoder->inputCacheRemaining); + if (result != MA_SUCCESS) { + break; + } + } + } + } else { + /* We have a way of determining the required number of input frames so just use the stack. */ + while (totalFramesReadOut < frameCount) { + ma_uint8 pIntermediaryBuffer[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* In internal format. */ + ma_uint64 intermediaryBufferCap = sizeof(pIntermediaryBuffer) / ma_get_bytes_per_frame(internalFormat, internalChannels); + ma_uint64 framesToReadThisIterationIn; + ma_uint64 framesReadThisIterationIn; + ma_uint64 framesToReadThisIterationOut; + ma_uint64 framesReadThisIterationOut; + ma_uint64 requiredInputFrameCount; + + framesToReadThisIterationOut = (frameCount - totalFramesReadOut); + framesToReadThisIterationIn = framesToReadThisIterationOut; + if (framesToReadThisIterationIn > intermediaryBufferCap) { + framesToReadThisIterationIn = intermediaryBufferCap; + } + + ma_data_converter_get_required_input_frame_count(&pDecoder->converter, framesToReadThisIterationOut, &requiredInputFrameCount); + if (framesToReadThisIterationIn > requiredInputFrameCount) { + framesToReadThisIterationIn = requiredInputFrameCount; + } + + if (requiredInputFrameCount > 0) { + result = ma_data_source_read_pcm_frames(pDecoder->pBackend, pIntermediaryBuffer, framesToReadThisIterationIn, &framesReadThisIterationIn); + + /* + Note here that even if we've reached the end, we don't want to abort because there might be more output frames needing to be + generated from cached input data, which might happen if resampling is being performed. + */ + if (result != MA_SUCCESS && result != MA_AT_END) { + break; + } + } else { + framesReadThisIterationIn = 0; + pIntermediaryBuffer[0] = 0; /* <-- This is just to silence a static analysis warning. */ + } + + /* + At this point we have our decoded data in input format and now we need to convert to output format. Note that even if we didn't read any + input frames, we still want to try processing frames because there may some output frames generated from cached input data. + */ + framesReadThisIterationOut = framesToReadThisIterationOut; + result = ma_data_converter_process_pcm_frames(&pDecoder->converter, pIntermediaryBuffer, &framesReadThisIterationIn, pRunningFramesOut, &framesReadThisIterationOut); + if (result != MA_SUCCESS) { + break; + } + + totalFramesReadOut += framesReadThisIterationOut; + + if (pRunningFramesOut != NULL) { + pRunningFramesOut = ma_offset_ptr(pRunningFramesOut, framesReadThisIterationOut * ma_get_bytes_per_frame(pDecoder->outputFormat, pDecoder->outputChannels)); + } + + if (framesReadThisIterationIn == 0 && framesReadThisIterationOut == 0) { + break; /* We're done. */ + } + } + } + } + } + + pDecoder->readPointerInPCMFrames += totalFramesReadOut; + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesReadOut; + } + + if (result == MA_SUCCESS && totalFramesReadOut == 0) { + result = MA_AT_END; + } + + return result; +} + +MA_API ma_result ma_decoder_seek_to_pcm_frame(ma_decoder* pDecoder, ma_uint64 frameIndex) +{ + if (pDecoder == NULL) { + return MA_INVALID_ARGS; + } + + if (pDecoder->pBackend != NULL) { + ma_result result; + ma_uint64 internalFrameIndex; + ma_uint32 internalSampleRate; + ma_uint64 currentFrameIndex; + + result = ma_data_source_get_data_format(pDecoder->pBackend, NULL, NULL, &internalSampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the internal sample rate. */ + } + + if (internalSampleRate == pDecoder->outputSampleRate) { + internalFrameIndex = frameIndex; + } else { + internalFrameIndex = ma_calculate_frame_count_after_resampling(internalSampleRate, pDecoder->outputSampleRate, frameIndex); + } + + /* Only seek if we're requesting a different frame to what we're currently sitting on. */ + ma_data_source_get_cursor_in_pcm_frames(pDecoder->pBackend, ¤tFrameIndex); + if (currentFrameIndex != internalFrameIndex) { + result = ma_data_source_seek_to_pcm_frame(pDecoder->pBackend, internalFrameIndex); + if (result == MA_SUCCESS) { + pDecoder->readPointerInPCMFrames = frameIndex; + } + + /* Reset the data converter so that any cached data in the resampler is cleared. */ + ma_data_converter_reset(&pDecoder->converter); + } + + return result; + } + + /* Should never get here, but if we do it means onSeekToPCMFrame was not set by the backend. */ + return MA_INVALID_ARGS; +} + +MA_API ma_result ma_decoder_get_data_format(ma_decoder* pDecoder, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + if (pDecoder == NULL) { + return MA_INVALID_ARGS; + } + + if (pFormat != NULL) { + *pFormat = pDecoder->outputFormat; + } + + if (pChannels != NULL) { + *pChannels = pDecoder->outputChannels; + } + + if (pSampleRate != NULL) { + *pSampleRate = pDecoder->outputSampleRate; + } + + if (pChannelMap != NULL) { + ma_data_converter_get_output_channel_map(&pDecoder->converter, pChannelMap, channelMapCap); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_get_cursor_in_pcm_frames(ma_decoder* pDecoder, ma_uint64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; + + if (pDecoder == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = pDecoder->readPointerInPCMFrames; + + return MA_SUCCESS; +} + +MA_API ma_result ma_decoder_get_length_in_pcm_frames(ma_decoder* pDecoder, ma_uint64* pLength) +{ + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; + + if (pDecoder == NULL) { + return MA_INVALID_ARGS; + } + + if (pDecoder->pBackend != NULL) { + ma_result result; + ma_uint64 internalLengthInPCMFrames; + ma_uint32 internalSampleRate; + + result = ma_data_source_get_length_in_pcm_frames(pDecoder->pBackend, &internalLengthInPCMFrames); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the internal length. */ + } + + result = ma_data_source_get_data_format(pDecoder->pBackend, NULL, NULL, &internalSampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the internal sample rate. */ + } + + if (internalSampleRate == pDecoder->outputSampleRate) { + *pLength = internalLengthInPCMFrames; + } else { + *pLength = ma_calculate_frame_count_after_resampling(pDecoder->outputSampleRate, internalSampleRate, internalLengthInPCMFrames); + } + + return MA_SUCCESS; + } else { + return MA_NO_BACKEND; + } +} + +MA_API ma_result ma_decoder_get_available_frames(ma_decoder* pDecoder, ma_uint64* pAvailableFrames) +{ + ma_result result; + ma_uint64 totalFrameCount; + + if (pAvailableFrames == NULL) { + return MA_INVALID_ARGS; + } + + *pAvailableFrames = 0; + + if (pDecoder == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_decoder_get_length_in_pcm_frames(pDecoder, &totalFrameCount); + if (result != MA_SUCCESS) { + return result; + } + + if (totalFrameCount <= pDecoder->readPointerInPCMFrames) { + *pAvailableFrames = 0; + } else { + *pAvailableFrames = totalFrameCount - pDecoder->readPointerInPCMFrames; + } + + return MA_SUCCESS; +} + + +static ma_result ma_decoder__full_decode_and_uninit(ma_decoder* pDecoder, ma_decoder_config* pConfigOut, ma_uint64* pFrameCountOut, void** ppPCMFramesOut) +{ + ma_result result; + ma_uint64 totalFrameCount; + ma_uint64 bpf; + ma_uint64 dataCapInFrames; + void* pPCMFramesOut; + + MA_ASSERT(pDecoder != NULL); + + totalFrameCount = 0; + bpf = ma_get_bytes_per_frame(pDecoder->outputFormat, pDecoder->outputChannels); + + /* The frame count is unknown until we try reading. Thus, we just run in a loop. */ + dataCapInFrames = 0; + pPCMFramesOut = NULL; + for (;;) { + ma_uint64 frameCountToTryReading; + ma_uint64 framesJustRead; + + /* Make room if there's not enough. */ + if (totalFrameCount == dataCapInFrames) { + void* pNewPCMFramesOut; + ma_uint64 newDataCapInFrames = dataCapInFrames*2; + if (newDataCapInFrames == 0) { + newDataCapInFrames = 4096; + } + + if ((newDataCapInFrames * bpf) > MA_SIZE_MAX) { + ma_free(pPCMFramesOut, &pDecoder->allocationCallbacks); + return MA_TOO_BIG; + } + + pNewPCMFramesOut = (void*)ma_realloc(pPCMFramesOut, (size_t)(newDataCapInFrames * bpf), &pDecoder->allocationCallbacks); + if (pNewPCMFramesOut == NULL) { + ma_free(pPCMFramesOut, &pDecoder->allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + + dataCapInFrames = newDataCapInFrames; + pPCMFramesOut = pNewPCMFramesOut; + } + + frameCountToTryReading = dataCapInFrames - totalFrameCount; + MA_ASSERT(frameCountToTryReading > 0); + + result = ma_decoder_read_pcm_frames(pDecoder, (ma_uint8*)pPCMFramesOut + (totalFrameCount * bpf), frameCountToTryReading, &framesJustRead); + totalFrameCount += framesJustRead; + + if (result != MA_SUCCESS) { + break; + } + + if (framesJustRead < frameCountToTryReading) { + break; + } + } + + + if (pConfigOut != NULL) { + pConfigOut->format = pDecoder->outputFormat; + pConfigOut->channels = pDecoder->outputChannels; + pConfigOut->sampleRate = pDecoder->outputSampleRate; + } + + if (ppPCMFramesOut != NULL) { + *ppPCMFramesOut = pPCMFramesOut; + } else { + ma_free(pPCMFramesOut, &pDecoder->allocationCallbacks); + } + + if (pFrameCountOut != NULL) { + *pFrameCountOut = totalFrameCount; + } + + ma_decoder_uninit(pDecoder); + return MA_SUCCESS; +} + +MA_API ma_result ma_decode_from_vfs(ma_vfs* pVFS, const char* pFilePath, ma_decoder_config* pConfig, ma_uint64* pFrameCountOut, void** ppPCMFramesOut) +{ + ma_result result; + ma_decoder_config config; + ma_decoder decoder; + + if (pFrameCountOut != NULL) { + *pFrameCountOut = 0; + } + if (ppPCMFramesOut != NULL) { + *ppPCMFramesOut = NULL; + } + + config = ma_decoder_config_init_copy(pConfig); + + result = ma_decoder_init_vfs(pVFS, pFilePath, &config, &decoder); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_decoder__full_decode_and_uninit(&decoder, pConfig, pFrameCountOut, ppPCMFramesOut); + + return result; +} + +MA_API ma_result ma_decode_file(const char* pFilePath, ma_decoder_config* pConfig, ma_uint64* pFrameCountOut, void** ppPCMFramesOut) +{ + return ma_decode_from_vfs(NULL, pFilePath, pConfig, pFrameCountOut, ppPCMFramesOut); +} + +MA_API ma_result ma_decode_memory(const void* pData, size_t dataSize, ma_decoder_config* pConfig, ma_uint64* pFrameCountOut, void** ppPCMFramesOut) +{ + ma_decoder_config config; + ma_decoder decoder; + ma_result result; + + if (pFrameCountOut != NULL) { + *pFrameCountOut = 0; + } + if (ppPCMFramesOut != NULL) { + *ppPCMFramesOut = NULL; + } + + if (pData == NULL || dataSize == 0) { + return MA_INVALID_ARGS; + } + + config = ma_decoder_config_init_copy(pConfig); + + result = ma_decoder_init_memory(pData, dataSize, &config, &decoder); + if (result != MA_SUCCESS) { + return result; + } + + return ma_decoder__full_decode_and_uninit(&decoder, pConfig, pFrameCountOut, ppPCMFramesOut); +} +#endif /* MA_NO_DECODING */ + + +#ifndef MA_NO_ENCODING + +#if defined(MA_HAS_WAV) +static size_t ma_encoder__internal_on_write_wav(void* pUserData, const void* pData, size_t bytesToWrite) +{ + ma_encoder* pEncoder = (ma_encoder*)pUserData; + size_t bytesWritten = 0; + + MA_ASSERT(pEncoder != NULL); + + pEncoder->onWrite(pEncoder, pData, bytesToWrite, &bytesWritten); + return bytesWritten; +} + +static ma_bool32 ma_encoder__internal_on_seek_wav(void* pUserData, int offset, ma_dr_wav_seek_origin origin) +{ + ma_encoder* pEncoder = (ma_encoder*)pUserData; + ma_result result; + + MA_ASSERT(pEncoder != NULL); + + result = pEncoder->onSeek(pEncoder, offset, (origin == ma_dr_wav_seek_origin_start) ? ma_seek_origin_start : ma_seek_origin_current); + if (result != MA_SUCCESS) { + return MA_FALSE; + } else { + return MA_TRUE; + } +} + +static ma_result ma_encoder__on_init_wav(ma_encoder* pEncoder) +{ + ma_dr_wav_data_format wavFormat; + ma_allocation_callbacks allocationCallbacks; + ma_dr_wav* pWav; + + MA_ASSERT(pEncoder != NULL); + + pWav = (ma_dr_wav*)ma_malloc(sizeof(*pWav), &pEncoder->config.allocationCallbacks); + if (pWav == NULL) { + return MA_OUT_OF_MEMORY; + } + + wavFormat.container = ma_dr_wav_container_riff; + wavFormat.channels = pEncoder->config.channels; + wavFormat.sampleRate = pEncoder->config.sampleRate; + wavFormat.bitsPerSample = ma_get_bytes_per_sample(pEncoder->config.format) * 8; + if (pEncoder->config.format == ma_format_f32) { + wavFormat.format = MA_DR_WAVE_FORMAT_IEEE_FLOAT; + } else { + wavFormat.format = MA_DR_WAVE_FORMAT_PCM; + } + + allocationCallbacks.pUserData = pEncoder->config.allocationCallbacks.pUserData; + allocationCallbacks.onMalloc = pEncoder->config.allocationCallbacks.onMalloc; + allocationCallbacks.onRealloc = pEncoder->config.allocationCallbacks.onRealloc; + allocationCallbacks.onFree = pEncoder->config.allocationCallbacks.onFree; + + if (!ma_dr_wav_init_write(pWav, &wavFormat, ma_encoder__internal_on_write_wav, ma_encoder__internal_on_seek_wav, pEncoder, &allocationCallbacks)) { + return MA_ERROR; + } + + pEncoder->pInternalEncoder = pWav; + + return MA_SUCCESS; +} + +static void ma_encoder__on_uninit_wav(ma_encoder* pEncoder) +{ + ma_dr_wav* pWav; + + MA_ASSERT(pEncoder != NULL); + + pWav = (ma_dr_wav*)pEncoder->pInternalEncoder; + MA_ASSERT(pWav != NULL); + + ma_dr_wav_uninit(pWav); + ma_free(pWav, &pEncoder->config.allocationCallbacks); +} + +static ma_result ma_encoder__on_write_pcm_frames_wav(ma_encoder* pEncoder, const void* pFramesIn, ma_uint64 frameCount, ma_uint64* pFramesWritten) +{ + ma_dr_wav* pWav; + ma_uint64 framesWritten; + + MA_ASSERT(pEncoder != NULL); + + pWav = (ma_dr_wav*)pEncoder->pInternalEncoder; + MA_ASSERT(pWav != NULL); + + framesWritten = ma_dr_wav_write_pcm_frames(pWav, frameCount, pFramesIn); + + if (pFramesWritten != NULL) { + *pFramesWritten = framesWritten; + } + + return MA_SUCCESS; +} +#endif + +MA_API ma_encoder_config ma_encoder_config_init(ma_encoding_format encodingFormat, ma_format format, ma_uint32 channels, ma_uint32 sampleRate) +{ + ma_encoder_config config; + + MA_ZERO_OBJECT(&config); + config.encodingFormat = encodingFormat; + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + + return config; +} + +MA_API ma_result ma_encoder_preinit(const ma_encoder_config* pConfig, ma_encoder* pEncoder) +{ + ma_result result; + + if (pEncoder == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pEncoder); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->format == ma_format_unknown || pConfig->channels == 0 || pConfig->sampleRate == 0) { + return MA_INVALID_ARGS; + } + + pEncoder->config = *pConfig; + + result = ma_allocation_callbacks_init_copy(&pEncoder->config.allocationCallbacks, &pConfig->allocationCallbacks); + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_encoder_init__internal(ma_encoder_write_proc onWrite, ma_encoder_seek_proc onSeek, void* pUserData, ma_encoder* pEncoder) +{ + ma_result result = MA_SUCCESS; + + /* This assumes ma_encoder_preinit() has been called prior. */ + MA_ASSERT(pEncoder != NULL); + + if (onWrite == NULL || onSeek == NULL) { + return MA_INVALID_ARGS; + } + + pEncoder->onWrite = onWrite; + pEncoder->onSeek = onSeek; + pEncoder->pUserData = pUserData; + + switch (pEncoder->config.encodingFormat) + { + case ma_encoding_format_wav: + { + #if defined(MA_HAS_WAV) + pEncoder->onInit = ma_encoder__on_init_wav; + pEncoder->onUninit = ma_encoder__on_uninit_wav; + pEncoder->onWritePCMFrames = ma_encoder__on_write_pcm_frames_wav; + #else + result = MA_NO_BACKEND; + #endif + } break; + + default: + { + result = MA_INVALID_ARGS; + } break; + } + + /* Getting here means we should have our backend callbacks set up. */ + if (result == MA_SUCCESS) { + result = pEncoder->onInit(pEncoder); + } + + return result; +} + +static ma_result ma_encoder__on_write_vfs(ma_encoder* pEncoder, const void* pBufferIn, size_t bytesToWrite, size_t* pBytesWritten) +{ + return ma_vfs_or_default_write(pEncoder->data.vfs.pVFS, pEncoder->data.vfs.file, pBufferIn, bytesToWrite, pBytesWritten); +} + +static ma_result ma_encoder__on_seek_vfs(ma_encoder* pEncoder, ma_int64 offset, ma_seek_origin origin) +{ + return ma_vfs_or_default_seek(pEncoder->data.vfs.pVFS, pEncoder->data.vfs.file, offset, origin); +} + +MA_API ma_result ma_encoder_init_vfs(ma_vfs* pVFS, const char* pFilePath, const ma_encoder_config* pConfig, ma_encoder* pEncoder) +{ + ma_result result; + ma_vfs_file file; + + result = ma_encoder_preinit(pConfig, pEncoder); + if (result != MA_SUCCESS) { + return result; + } + + /* Now open the file. If this fails we don't need to uninitialize the encoder. */ + result = ma_vfs_or_default_open(pVFS, pFilePath, MA_OPEN_MODE_WRITE, &file); + if (result != MA_SUCCESS) { + return result; + } + + pEncoder->data.vfs.pVFS = pVFS; + pEncoder->data.vfs.file = file; + + result = ma_encoder_init__internal(ma_encoder__on_write_vfs, ma_encoder__on_seek_vfs, NULL, pEncoder); + if (result != MA_SUCCESS) { + ma_vfs_or_default_close(pVFS, file); + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_encoder_init_vfs_w(ma_vfs* pVFS, const wchar_t* pFilePath, const ma_encoder_config* pConfig, ma_encoder* pEncoder) +{ + ma_result result; + ma_vfs_file file; + + result = ma_encoder_preinit(pConfig, pEncoder); + if (result != MA_SUCCESS) { + return result; + } + + /* Now open the file. If this fails we don't need to uninitialize the encoder. */ + result = ma_vfs_or_default_open_w(pVFS, pFilePath, MA_OPEN_MODE_WRITE, &file); + if (result != MA_SUCCESS) { + return result; + } + + pEncoder->data.vfs.pVFS = pVFS; + pEncoder->data.vfs.file = file; + + result = ma_encoder_init__internal(ma_encoder__on_write_vfs, ma_encoder__on_seek_vfs, NULL, pEncoder); + if (result != MA_SUCCESS) { + ma_vfs_or_default_close(pVFS, file); + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_encoder_init_file(const char* pFilePath, const ma_encoder_config* pConfig, ma_encoder* pEncoder) +{ + return ma_encoder_init_vfs(NULL, pFilePath, pConfig, pEncoder); +} + +MA_API ma_result ma_encoder_init_file_w(const wchar_t* pFilePath, const ma_encoder_config* pConfig, ma_encoder* pEncoder) +{ + return ma_encoder_init_vfs_w(NULL, pFilePath, pConfig, pEncoder); +} + +MA_API ma_result ma_encoder_init(ma_encoder_write_proc onWrite, ma_encoder_seek_proc onSeek, void* pUserData, const ma_encoder_config* pConfig, ma_encoder* pEncoder) +{ + ma_result result; + + result = ma_encoder_preinit(pConfig, pEncoder); + if (result != MA_SUCCESS) { + return result; + } + + return ma_encoder_init__internal(onWrite, onSeek, pUserData, pEncoder); +} + + +MA_API void ma_encoder_uninit(ma_encoder* pEncoder) +{ + if (pEncoder == NULL) { + return; + } + + if (pEncoder->onUninit) { + pEncoder->onUninit(pEncoder); + } + + /* If we have a file handle, close it. */ + if (pEncoder->onWrite == ma_encoder__on_write_vfs) { + ma_vfs_or_default_close(pEncoder->data.vfs.pVFS, pEncoder->data.vfs.file); + pEncoder->data.vfs.file = NULL; + } +} + + +MA_API ma_result ma_encoder_write_pcm_frames(ma_encoder* pEncoder, const void* pFramesIn, ma_uint64 frameCount, ma_uint64* pFramesWritten) +{ + if (pFramesWritten != NULL) { + *pFramesWritten = 0; + } + + if (pEncoder == NULL || pFramesIn == NULL) { + return MA_INVALID_ARGS; + } + + return pEncoder->onWritePCMFrames(pEncoder, pFramesIn, frameCount, pFramesWritten); +} +#endif /* MA_NO_ENCODING */ + + + +/************************************************************************************************************************************************************** + +Generation + +**************************************************************************************************************************************************************/ +#ifndef MA_NO_GENERATION +MA_API ma_waveform_config ma_waveform_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, ma_waveform_type type, double amplitude, double frequency) +{ + ma_waveform_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.type = type; + config.amplitude = amplitude; + config.frequency = frequency; + + return config; +} + +static ma_result ma_waveform__data_source_on_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_waveform_read_pcm_frames((ma_waveform*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_waveform__data_source_on_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_waveform_seek_to_pcm_frame((ma_waveform*)pDataSource, frameIndex); +} + +static ma_result ma_waveform__data_source_on_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + ma_waveform* pWaveform = (ma_waveform*)pDataSource; + + *pFormat = pWaveform->config.format; + *pChannels = pWaveform->config.channels; + *pSampleRate = pWaveform->config.sampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pWaveform->config.channels); + + return MA_SUCCESS; +} + +static ma_result ma_waveform__data_source_on_get_cursor(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + ma_waveform* pWaveform = (ma_waveform*)pDataSource; + + *pCursor = (ma_uint64)(pWaveform->time / pWaveform->advance); + + return MA_SUCCESS; +} + +static double ma_waveform__calculate_advance(ma_uint32 sampleRate, double frequency) +{ + return (1.0 / (sampleRate / frequency)); +} + +static void ma_waveform__update_advance(ma_waveform* pWaveform) +{ + pWaveform->advance = ma_waveform__calculate_advance(pWaveform->config.sampleRate, pWaveform->config.frequency); +} + +static ma_data_source_vtable g_ma_waveform_data_source_vtable = +{ + ma_waveform__data_source_on_read, + ma_waveform__data_source_on_seek, + ma_waveform__data_source_on_get_data_format, + ma_waveform__data_source_on_get_cursor, + NULL, /* onGetLength. There's no notion of a length in waveforms. */ + NULL, /* onSetLooping */ + 0 +}; + +MA_API ma_result ma_waveform_init(const ma_waveform_config* pConfig, ma_waveform* pWaveform) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pWaveform); + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_waveform_data_source_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pWaveform->ds); + if (result != MA_SUCCESS) { + return result; + } + + pWaveform->config = *pConfig; + pWaveform->advance = ma_waveform__calculate_advance(pWaveform->config.sampleRate, pWaveform->config.frequency); + pWaveform->time = 0; + + return MA_SUCCESS; +} + +MA_API void ma_waveform_uninit(ma_waveform* pWaveform) +{ + if (pWaveform == NULL) { + return; + } + + ma_data_source_uninit(&pWaveform->ds); +} + +MA_API ma_result ma_waveform_set_amplitude(ma_waveform* pWaveform, double amplitude) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->config.amplitude = amplitude; + return MA_SUCCESS; +} + +MA_API ma_result ma_waveform_set_frequency(ma_waveform* pWaveform, double frequency) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->config.frequency = frequency; + ma_waveform__update_advance(pWaveform); + + return MA_SUCCESS; +} + +MA_API ma_result ma_waveform_set_type(ma_waveform* pWaveform, ma_waveform_type type) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->config.type = type; + return MA_SUCCESS; +} + +MA_API ma_result ma_waveform_set_sample_rate(ma_waveform* pWaveform, ma_uint32 sampleRate) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->config.sampleRate = sampleRate; + ma_waveform__update_advance(pWaveform); + + return MA_SUCCESS; +} + +static float ma_waveform_sine_f32(double time, double amplitude) +{ + return (float)(ma_sind(MA_TAU_D * time) * amplitude); +} + +static ma_int16 ma_waveform_sine_s16(double time, double amplitude) +{ + return ma_pcm_sample_f32_to_s16(ma_waveform_sine_f32(time, amplitude)); +} + +static float ma_waveform_square_f32(double time, double dutyCycle, double amplitude) +{ + double f = time - (ma_int64)time; + double r; + + if (f < dutyCycle) { + r = amplitude; + } else { + r = -amplitude; + } + + return (float)r; +} + +static ma_int16 ma_waveform_square_s16(double time, double dutyCycle, double amplitude) +{ + return ma_pcm_sample_f32_to_s16(ma_waveform_square_f32(time, dutyCycle, amplitude)); +} + +static float ma_waveform_triangle_f32(double time, double amplitude) +{ + double f = time - (ma_int64)time; + double r; + + r = 2 * ma_abs(2 * (f - 0.5)) - 1; + + return (float)(r * amplitude); +} + +static ma_int16 ma_waveform_triangle_s16(double time, double amplitude) +{ + return ma_pcm_sample_f32_to_s16(ma_waveform_triangle_f32(time, amplitude)); +} + +static float ma_waveform_sawtooth_f32(double time, double amplitude) +{ + double f = time - (ma_int64)time; + double r; + + r = 2 * (f - 0.5); + + return (float)(r * amplitude); +} + +static ma_int16 ma_waveform_sawtooth_s16(double time, double amplitude) +{ + return ma_pcm_sample_f32_to_s16(ma_waveform_sawtooth_f32(time, amplitude)); +} + +static void ma_waveform_read_pcm_frames__sine(ma_waveform* pWaveform, void* pFramesOut, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint64 iChannel; + ma_uint32 bps = ma_get_bytes_per_sample(pWaveform->config.format); + ma_uint32 bpf = bps * pWaveform->config.channels; + + MA_ASSERT(pWaveform != NULL); + MA_ASSERT(pFramesOut != NULL); + + if (pWaveform->config.format == ma_format_f32) { + float* pFramesOutF32 = (float*)pFramesOut; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_waveform_sine_f32(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + pFramesOutF32[iFrame*pWaveform->config.channels + iChannel] = s; + } + } + } else if (pWaveform->config.format == ma_format_s16) { + ma_int16* pFramesOutS16 = (ma_int16*)pFramesOut; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_int16 s = ma_waveform_sine_s16(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + pFramesOutS16[iFrame*pWaveform->config.channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_waveform_sine_f32(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pWaveform->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } +} + +static void ma_waveform_read_pcm_frames__square(ma_waveform* pWaveform, double dutyCycle, void* pFramesOut, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint64 iChannel; + ma_uint32 bps = ma_get_bytes_per_sample(pWaveform->config.format); + ma_uint32 bpf = bps * pWaveform->config.channels; + + MA_ASSERT(pWaveform != NULL); + MA_ASSERT(pFramesOut != NULL); + + if (pWaveform->config.format == ma_format_f32) { + float* pFramesOutF32 = (float*)pFramesOut; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_waveform_square_f32(pWaveform->time, dutyCycle, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + pFramesOutF32[iFrame*pWaveform->config.channels + iChannel] = s; + } + } + } else if (pWaveform->config.format == ma_format_s16) { + ma_int16* pFramesOutS16 = (ma_int16*)pFramesOut; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_int16 s = ma_waveform_square_s16(pWaveform->time, dutyCycle, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + pFramesOutS16[iFrame*pWaveform->config.channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_waveform_square_f32(pWaveform->time, dutyCycle, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pWaveform->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } +} + +static void ma_waveform_read_pcm_frames__triangle(ma_waveform* pWaveform, void* pFramesOut, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint64 iChannel; + ma_uint32 bps = ma_get_bytes_per_sample(pWaveform->config.format); + ma_uint32 bpf = bps * pWaveform->config.channels; + + MA_ASSERT(pWaveform != NULL); + MA_ASSERT(pFramesOut != NULL); + + if (pWaveform->config.format == ma_format_f32) { + float* pFramesOutF32 = (float*)pFramesOut; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_waveform_triangle_f32(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + pFramesOutF32[iFrame*pWaveform->config.channels + iChannel] = s; + } + } + } else if (pWaveform->config.format == ma_format_s16) { + ma_int16* pFramesOutS16 = (ma_int16*)pFramesOut; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_int16 s = ma_waveform_triangle_s16(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + pFramesOutS16[iFrame*pWaveform->config.channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_waveform_triangle_f32(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pWaveform->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } +} + +static void ma_waveform_read_pcm_frames__sawtooth(ma_waveform* pWaveform, void* pFramesOut, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint64 iChannel; + ma_uint32 bps = ma_get_bytes_per_sample(pWaveform->config.format); + ma_uint32 bpf = bps * pWaveform->config.channels; + + MA_ASSERT(pWaveform != NULL); + MA_ASSERT(pFramesOut != NULL); + + if (pWaveform->config.format == ma_format_f32) { + float* pFramesOutF32 = (float*)pFramesOut; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_waveform_sawtooth_f32(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + pFramesOutF32[iFrame*pWaveform->config.channels + iChannel] = s; + } + } + } else if (pWaveform->config.format == ma_format_s16) { + ma_int16* pFramesOutS16 = (ma_int16*)pFramesOut; + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_int16 s = ma_waveform_sawtooth_s16(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + pFramesOutS16[iFrame*pWaveform->config.channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_waveform_sawtooth_f32(pWaveform->time, pWaveform->config.amplitude); + pWaveform->time += pWaveform->advance; + + for (iChannel = 0; iChannel < pWaveform->config.channels; iChannel += 1) { + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pWaveform->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } +} + +MA_API ma_result ma_waveform_read_pcm_frames(ma_waveform* pWaveform, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + if (pFramesOut != NULL) { + switch (pWaveform->config.type) + { + case ma_waveform_type_sine: + { + ma_waveform_read_pcm_frames__sine(pWaveform, pFramesOut, frameCount); + } break; + + case ma_waveform_type_square: + { + ma_waveform_read_pcm_frames__square(pWaveform, 0.5, pFramesOut, frameCount); + } break; + + case ma_waveform_type_triangle: + { + ma_waveform_read_pcm_frames__triangle(pWaveform, pFramesOut, frameCount); + } break; + + case ma_waveform_type_sawtooth: + { + ma_waveform_read_pcm_frames__sawtooth(pWaveform, pFramesOut, frameCount); + } break; + + default: return MA_INVALID_OPERATION; /* Unknown waveform type. */ + } + } else { + pWaveform->time += pWaveform->advance * (ma_int64)frameCount; /* Cast to int64 required for VC6. Won't affect anything in practice. */ + } + + if (pFramesRead != NULL) { + *pFramesRead = frameCount; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_waveform_seek_to_pcm_frame(ma_waveform* pWaveform, ma_uint64 frameIndex) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->time = pWaveform->advance * (ma_int64)frameIndex; /* Casting for VC6. Won't be an issue in practice. */ + + return MA_SUCCESS; +} + +MA_API ma_pulsewave_config ma_pulsewave_config_init(ma_format format, ma_uint32 channels, ma_uint32 sampleRate, double dutyCycle, double amplitude, double frequency) +{ + ma_pulsewave_config config; + + MA_ZERO_OBJECT(&config); + config.format = format; + config.channels = channels; + config.sampleRate = sampleRate; + config.dutyCycle = dutyCycle; + config.amplitude = amplitude; + config.frequency = frequency; + + return config; +} + +MA_API ma_result ma_pulsewave_init(const ma_pulsewave_config* pConfig, ma_pulsewave* pWaveform) +{ + ma_result result; + ma_waveform_config config; + + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pWaveform); + + config = ma_waveform_config_init( + pConfig->format, + pConfig->channels, + pConfig->sampleRate, + ma_waveform_type_square, + pConfig->amplitude, + pConfig->frequency + ); + + result = ma_waveform_init(&config, &pWaveform->waveform); + ma_pulsewave_set_duty_cycle(pWaveform, pConfig->dutyCycle); + + return result; +} + +MA_API void ma_pulsewave_uninit(ma_pulsewave* pWaveform) +{ + if (pWaveform == NULL) { + return; + } + + ma_waveform_uninit(&pWaveform->waveform); +} + +MA_API ma_result ma_pulsewave_read_pcm_frames(ma_pulsewave* pWaveform, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + if (pFramesOut != NULL) { + ma_waveform_read_pcm_frames__square(&pWaveform->waveform, pWaveform->config.dutyCycle, pFramesOut, frameCount); + } else { + pWaveform->waveform.time += pWaveform->waveform.advance * (ma_int64)frameCount; /* Cast to int64 required for VC6. Won't affect anything in practice. */ + } + + if (pFramesRead != NULL) { + *pFramesRead = frameCount; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_pulsewave_seek_to_pcm_frame(ma_pulsewave* pWaveform, ma_uint64 frameIndex) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + ma_waveform_seek_to_pcm_frame(&pWaveform->waveform, frameIndex); + + return MA_SUCCESS; +} + +MA_API ma_result ma_pulsewave_set_amplitude(ma_pulsewave* pWaveform, double amplitude) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->config.amplitude = amplitude; + ma_waveform_set_amplitude(&pWaveform->waveform, amplitude); + + return MA_SUCCESS; +} + +MA_API ma_result ma_pulsewave_set_frequency(ma_pulsewave* pWaveform, double frequency) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->config.frequency = frequency; + ma_waveform_set_frequency(&pWaveform->waveform, frequency); + + return MA_SUCCESS; +} + +MA_API ma_result ma_pulsewave_set_sample_rate(ma_pulsewave* pWaveform, ma_uint32 sampleRate) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->config.sampleRate = sampleRate; + ma_waveform_set_sample_rate(&pWaveform->waveform, sampleRate); + + return MA_SUCCESS; +} + +MA_API ma_result ma_pulsewave_set_duty_cycle(ma_pulsewave* pWaveform, double dutyCycle) +{ + if (pWaveform == NULL) { + return MA_INVALID_ARGS; + } + + pWaveform->config.dutyCycle = dutyCycle; + + return MA_SUCCESS; +} + + + +MA_API ma_noise_config ma_noise_config_init(ma_format format, ma_uint32 channels, ma_noise_type type, ma_int32 seed, double amplitude) +{ + ma_noise_config config; + MA_ZERO_OBJECT(&config); + + config.format = format; + config.channels = channels; + config.type = type; + config.seed = seed; + config.amplitude = amplitude; + + if (config.seed == 0) { + config.seed = MA_DEFAULT_LCG_SEED; + } + + return config; +} + + +static ma_result ma_noise__data_source_on_read(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_noise_read_pcm_frames((ma_noise*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_noise__data_source_on_seek(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + /* No-op. Just pretend to be successful. */ + (void)pDataSource; + (void)frameIndex; + return MA_SUCCESS; +} + +static ma_result ma_noise__data_source_on_get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + ma_noise* pNoise = (ma_noise*)pDataSource; + + *pFormat = pNoise->config.format; + *pChannels = pNoise->config.channels; + *pSampleRate = 0; /* There is no notion of sample rate with noise generation. */ + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pNoise->config.channels); + + return MA_SUCCESS; +} + +static ma_data_source_vtable g_ma_noise_data_source_vtable = +{ + ma_noise__data_source_on_read, + ma_noise__data_source_on_seek, /* No-op for noise. */ + ma_noise__data_source_on_get_data_format, + NULL, /* onGetCursor. No notion of a cursor for noise. */ + NULL, /* onGetLength. No notion of a length for noise. */ + NULL, /* onSetLooping */ + 0 +}; + + +#ifndef MA_PINK_NOISE_BIN_SIZE +#define MA_PINK_NOISE_BIN_SIZE 16 +#endif + +typedef struct +{ + size_t sizeInBytes; + struct + { + size_t binOffset; + size_t accumulationOffset; + size_t counterOffset; + } pink; + struct + { + size_t accumulationOffset; + } brownian; +} ma_noise_heap_layout; + +static ma_result ma_noise_get_heap_layout(const ma_noise_config* pConfig, ma_noise_heap_layout* pHeapLayout) +{ + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->channels == 0) { + return MA_INVALID_ARGS; + } + + pHeapLayout->sizeInBytes = 0; + + /* Pink. */ + if (pConfig->type == ma_noise_type_pink) { + /* bin */ + pHeapLayout->pink.binOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(double*) * pConfig->channels; + pHeapLayout->sizeInBytes += sizeof(double ) * pConfig->channels * MA_PINK_NOISE_BIN_SIZE; + + /* accumulation */ + pHeapLayout->pink.accumulationOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(double) * pConfig->channels; + + /* counter */ + pHeapLayout->pink.counterOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(ma_uint32) * pConfig->channels; + } + + /* Brownian. */ + if (pConfig->type == ma_noise_type_brownian) { + /* accumulation */ + pHeapLayout->brownian.accumulationOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += sizeof(double) * pConfig->channels; + } + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +MA_API ma_result ma_noise_get_heap_size(const ma_noise_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_noise_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_noise_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_noise_init_preallocated(const ma_noise_config* pConfig, void* pHeap, ma_noise* pNoise) +{ + ma_result result; + ma_noise_heap_layout heapLayout; + ma_data_source_config dataSourceConfig; + ma_uint32 iChannel; + + if (pNoise == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNoise); + + result = ma_noise_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pNoise->_pHeap = pHeap; + MA_ZERO_MEMORY(pNoise->_pHeap, heapLayout.sizeInBytes); + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_noise_data_source_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pNoise->ds); + if (result != MA_SUCCESS) { + return result; + } + + pNoise->config = *pConfig; + ma_lcg_seed(&pNoise->lcg, pConfig->seed); + + if (pNoise->config.type == ma_noise_type_pink) { + pNoise->state.pink.bin = (double** )ma_offset_ptr(pHeap, heapLayout.pink.binOffset); + pNoise->state.pink.accumulation = (double* )ma_offset_ptr(pHeap, heapLayout.pink.accumulationOffset); + pNoise->state.pink.counter = (ma_uint32*)ma_offset_ptr(pHeap, heapLayout.pink.counterOffset); + + for (iChannel = 0; iChannel < pConfig->channels; iChannel += 1) { + pNoise->state.pink.bin[iChannel] = (double*)ma_offset_ptr(pHeap, heapLayout.pink.binOffset + (sizeof(double*) * pConfig->channels) + (sizeof(double) * MA_PINK_NOISE_BIN_SIZE * iChannel)); + pNoise->state.pink.accumulation[iChannel] = 0; + pNoise->state.pink.counter[iChannel] = 1; + } + } + + if (pNoise->config.type == ma_noise_type_brownian) { + pNoise->state.brownian.accumulation = (double*)ma_offset_ptr(pHeap, heapLayout.brownian.accumulationOffset); + + for (iChannel = 0; iChannel < pConfig->channels; iChannel += 1) { + pNoise->state.brownian.accumulation[iChannel] = 0; + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_noise_init(const ma_noise_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_noise* pNoise) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_noise_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_noise_init_preallocated(pConfig, pHeap, pNoise); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pNoise->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_noise_uninit(ma_noise* pNoise, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pNoise == NULL) { + return; + } + + ma_data_source_uninit(&pNoise->ds); + + if (pNoise->_ownsHeap) { + ma_free(pNoise->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_result ma_noise_set_amplitude(ma_noise* pNoise, double amplitude) +{ + if (pNoise == NULL) { + return MA_INVALID_ARGS; + } + + pNoise->config.amplitude = amplitude; + return MA_SUCCESS; +} + +MA_API ma_result ma_noise_set_seed(ma_noise* pNoise, ma_int32 seed) +{ + if (pNoise == NULL) { + return MA_INVALID_ARGS; + } + + pNoise->lcg.state = seed; + return MA_SUCCESS; +} + + +MA_API ma_result ma_noise_set_type(ma_noise* pNoise, ma_noise_type type) +{ + if (pNoise == NULL) { + return MA_INVALID_ARGS; + } + + /* + This function should never have been implemented in the first place. Changing the type dynamically is not + supported. Instead you need to uninitialize and reinitialize a fresh `ma_noise` object. This function + will be removed in version 0.12. + */ + MA_ASSERT(MA_FALSE); + (void)type; + + return MA_INVALID_OPERATION; +} + +static MA_INLINE float ma_noise_f32_white(ma_noise* pNoise) +{ + return (float)(ma_lcg_rand_f64(&pNoise->lcg) * pNoise->config.amplitude); +} + +static MA_INLINE ma_int16 ma_noise_s16_white(ma_noise* pNoise) +{ + return ma_pcm_sample_f32_to_s16(ma_noise_f32_white(pNoise)); +} + +static MA_INLINE ma_uint64 ma_noise_read_pcm_frames__white(ma_noise* pNoise, void* pFramesOut, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint32 iChannel; + const ma_uint32 channels = pNoise->config.channels; + MA_ASSUME(channels > 0); + + if (pNoise->config.format == ma_format_f32) { + float* pFramesOutF32 = (float*)pFramesOut; + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_noise_f32_white(pNoise); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutF32[iFrame*channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutF32[iFrame*channels + iChannel] = ma_noise_f32_white(pNoise); + } + } + } + } else if (pNoise->config.format == ma_format_s16) { + ma_int16* pFramesOutS16 = (ma_int16*)pFramesOut; + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_int16 s = ma_noise_s16_white(pNoise); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutS16[iFrame*channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutS16[iFrame*channels + iChannel] = ma_noise_s16_white(pNoise); + } + } + } + } else { + const ma_uint32 bps = ma_get_bytes_per_sample(pNoise->config.format); + const ma_uint32 bpf = bps * channels; + + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_noise_f32_white(pNoise); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pNoise->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + float s = ma_noise_f32_white(pNoise); + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pNoise->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } + } + + return frameCount; +} + + +static MA_INLINE unsigned int ma_tzcnt32(unsigned int x) +{ + unsigned int n; + + /* Special case for odd numbers since they should happen about half the time. */ + if (x & 0x1) { + return 0; + } + + if (x == 0) { + return sizeof(x) << 3; + } + + n = 1; + if ((x & 0x0000FFFF) == 0) { x >>= 16; n += 16; } + if ((x & 0x000000FF) == 0) { x >>= 8; n += 8; } + if ((x & 0x0000000F) == 0) { x >>= 4; n += 4; } + if ((x & 0x00000003) == 0) { x >>= 2; n += 2; } + n -= x & 0x00000001; + + return n; +} + +/* +Pink noise generation based on Tonic (public domain) with modifications. https://github.com/TonicAudio/Tonic/blob/master/src/Tonic/Noise.h + +This is basically _the_ reference for pink noise from what I've found: http://www.firstpr.com.au/dsp/pink-noise/ +*/ +static MA_INLINE float ma_noise_f32_pink(ma_noise* pNoise, ma_uint32 iChannel) +{ + double result; + double binPrev; + double binNext; + unsigned int ibin; + + ibin = ma_tzcnt32(pNoise->state.pink.counter[iChannel]) & (MA_PINK_NOISE_BIN_SIZE - 1); + + binPrev = pNoise->state.pink.bin[iChannel][ibin]; + binNext = ma_lcg_rand_f64(&pNoise->lcg); + pNoise->state.pink.bin[iChannel][ibin] = binNext; + + pNoise->state.pink.accumulation[iChannel] += (binNext - binPrev); + pNoise->state.pink.counter[iChannel] += 1; + + result = (ma_lcg_rand_f64(&pNoise->lcg) + pNoise->state.pink.accumulation[iChannel]); + result /= 10; + + return (float)(result * pNoise->config.amplitude); +} + +static MA_INLINE ma_int16 ma_noise_s16_pink(ma_noise* pNoise, ma_uint32 iChannel) +{ + return ma_pcm_sample_f32_to_s16(ma_noise_f32_pink(pNoise, iChannel)); +} + +static MA_INLINE ma_uint64 ma_noise_read_pcm_frames__pink(ma_noise* pNoise, void* pFramesOut, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint32 iChannel; + const ma_uint32 channels = pNoise->config.channels; + MA_ASSUME(channels > 0); + + if (pNoise->config.format == ma_format_f32) { + float* pFramesOutF32 = (float*)pFramesOut; + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_noise_f32_pink(pNoise, 0); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutF32[iFrame*channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutF32[iFrame*channels + iChannel] = ma_noise_f32_pink(pNoise, iChannel); + } + } + } + } else if (pNoise->config.format == ma_format_s16) { + ma_int16* pFramesOutS16 = (ma_int16*)pFramesOut; + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_int16 s = ma_noise_s16_pink(pNoise, 0); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutS16[iFrame*channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutS16[iFrame*channels + iChannel] = ma_noise_s16_pink(pNoise, iChannel); + } + } + } + } else { + const ma_uint32 bps = ma_get_bytes_per_sample(pNoise->config.format); + const ma_uint32 bpf = bps * channels; + + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_noise_f32_pink(pNoise, 0); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pNoise->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + float s = ma_noise_f32_pink(pNoise, iChannel); + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pNoise->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } + } + + return frameCount; +} + + +static MA_INLINE float ma_noise_f32_brownian(ma_noise* pNoise, ma_uint32 iChannel) +{ + double result; + + result = (ma_lcg_rand_f64(&pNoise->lcg) + pNoise->state.brownian.accumulation[iChannel]); + result /= 1.005; /* Don't escape the -1..1 range on average. */ + + pNoise->state.brownian.accumulation[iChannel] = result; + result /= 20; + + return (float)(result * pNoise->config.amplitude); +} + +static MA_INLINE ma_int16 ma_noise_s16_brownian(ma_noise* pNoise, ma_uint32 iChannel) +{ + return ma_pcm_sample_f32_to_s16(ma_noise_f32_brownian(pNoise, iChannel)); +} + +static MA_INLINE ma_uint64 ma_noise_read_pcm_frames__brownian(ma_noise* pNoise, void* pFramesOut, ma_uint64 frameCount) +{ + ma_uint64 iFrame; + ma_uint32 iChannel; + const ma_uint32 channels = pNoise->config.channels; + MA_ASSUME(channels > 0); + + if (pNoise->config.format == ma_format_f32) { + float* pFramesOutF32 = (float*)pFramesOut; + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_noise_f32_brownian(pNoise, 0); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutF32[iFrame*channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutF32[iFrame*channels + iChannel] = ma_noise_f32_brownian(pNoise, iChannel); + } + } + } + } else if (pNoise->config.format == ma_format_s16) { + ma_int16* pFramesOutS16 = (ma_int16*)pFramesOut; + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + ma_int16 s = ma_noise_s16_brownian(pNoise, 0); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutS16[iFrame*channels + iChannel] = s; + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + pFramesOutS16[iFrame*channels + iChannel] = ma_noise_s16_brownian(pNoise, iChannel); + } + } + } + } else { + const ma_uint32 bps = ma_get_bytes_per_sample(pNoise->config.format); + const ma_uint32 bpf = bps * channels; + + if (pNoise->config.duplicateChannels) { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + float s = ma_noise_f32_brownian(pNoise, 0); + for (iChannel = 0; iChannel < channels; iChannel += 1) { + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pNoise->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } else { + for (iFrame = 0; iFrame < frameCount; iFrame += 1) { + for (iChannel = 0; iChannel < channels; iChannel += 1) { + float s = ma_noise_f32_brownian(pNoise, iChannel); + ma_pcm_convert(ma_offset_ptr(pFramesOut, iFrame*bpf + iChannel*bps), pNoise->config.format, &s, ma_format_f32, 1, ma_dither_mode_none); + } + } + } + } + + return frameCount; +} + +MA_API ma_result ma_noise_read_pcm_frames(ma_noise* pNoise, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_uint64 framesRead = 0; + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + if (pNoise == NULL) { + return MA_INVALID_ARGS; + } + + /* The output buffer is allowed to be NULL. Since we aren't tracking cursors or anything we can just do nothing and pretend to be successful. */ + if (pFramesOut == NULL) { + framesRead = frameCount; + } else { + switch (pNoise->config.type) { + case ma_noise_type_white: framesRead = ma_noise_read_pcm_frames__white (pNoise, pFramesOut, frameCount); break; + case ma_noise_type_pink: framesRead = ma_noise_read_pcm_frames__pink (pNoise, pFramesOut, frameCount); break; + case ma_noise_type_brownian: framesRead = ma_noise_read_pcm_frames__brownian(pNoise, pFramesOut, frameCount); break; + default: return MA_INVALID_OPERATION; /* Unknown noise type. */ + } + } + + if (pFramesRead != NULL) { + *pFramesRead = framesRead; + } + + return MA_SUCCESS; +} +#endif /* MA_NO_GENERATION */ + + + +#ifndef MA_NO_RESOURCE_MANAGER +#ifndef MA_RESOURCE_MANAGER_PAGE_SIZE_IN_MILLISECONDS +#define MA_RESOURCE_MANAGER_PAGE_SIZE_IN_MILLISECONDS 1000 +#endif + +#ifndef MA_JOB_TYPE_RESOURCE_MANAGER_QUEUE_CAPACITY +#define MA_JOB_TYPE_RESOURCE_MANAGER_QUEUE_CAPACITY 1024 +#endif + +MA_API ma_resource_manager_pipeline_notifications ma_resource_manager_pipeline_notifications_init(void) +{ + ma_resource_manager_pipeline_notifications notifications; + + MA_ZERO_OBJECT(¬ifications); + + return notifications; +} + +static void ma_resource_manager_pipeline_notifications_signal_all_notifications(const ma_resource_manager_pipeline_notifications* pPipelineNotifications) +{ + if (pPipelineNotifications == NULL) { + return; + } + + if (pPipelineNotifications->init.pNotification) { ma_async_notification_signal(pPipelineNotifications->init.pNotification); } + if (pPipelineNotifications->done.pNotification) { ma_async_notification_signal(pPipelineNotifications->done.pNotification); } +} + +static void ma_resource_manager_pipeline_notifications_acquire_all_fences(const ma_resource_manager_pipeline_notifications* pPipelineNotifications) +{ + if (pPipelineNotifications == NULL) { + return; + } + + if (pPipelineNotifications->init.pFence != NULL) { ma_fence_acquire(pPipelineNotifications->init.pFence); } + if (pPipelineNotifications->done.pFence != NULL) { ma_fence_acquire(pPipelineNotifications->done.pFence); } +} + +static void ma_resource_manager_pipeline_notifications_release_all_fences(const ma_resource_manager_pipeline_notifications* pPipelineNotifications) +{ + if (pPipelineNotifications == NULL) { + return; + } + + if (pPipelineNotifications->init.pFence != NULL) { ma_fence_release(pPipelineNotifications->init.pFence); } + if (pPipelineNotifications->done.pFence != NULL) { ma_fence_release(pPipelineNotifications->done.pFence); } +} + + + +#ifndef MA_DEFAULT_HASH_SEED +#define MA_DEFAULT_HASH_SEED 42 +#endif + +/* MurmurHash3. Based on code from https://github.com/PeterScott/murmur3/blob/master/murmur3.c (public domain). */ +#if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic push + #if __GNUC__ >= 7 + #pragma GCC diagnostic ignored "-Wimplicit-fallthrough" + #endif +#endif + +static MA_INLINE ma_uint32 ma_rotl32(ma_uint32 x, ma_int8 r) +{ + return (x << r) | (x >> (32 - r)); +} + +static MA_INLINE ma_uint32 ma_hash_getblock(const ma_uint32* blocks, int i) +{ + ma_uint32 block; + + /* Try silencing a sanitization warning about unaligned access by doing a memcpy() instead of assignment. */ + MA_COPY_MEMORY(&block, ma_offset_ptr(blocks, i * sizeof(block)), sizeof(block)); + + if (ma_is_little_endian()) { + return block; + } else { + return ma_swap_endian_uint32(block); + } +} + +static MA_INLINE ma_uint32 ma_hash_fmix32(ma_uint32 h) +{ + h ^= h >> 16; + h *= 0x85ebca6b; + h ^= h >> 13; + h *= 0xc2b2ae35; + h ^= h >> 16; + + return h; +} + +static ma_uint32 ma_hash_32(const void* key, int len, ma_uint32 seed) +{ + const ma_uint8* data = (const ma_uint8*)key; + const ma_uint32* blocks; + const ma_uint8* tail; + const int nblocks = len / 4; + ma_uint32 h1 = seed; + ma_uint32 c1 = 0xcc9e2d51; + ma_uint32 c2 = 0x1b873593; + ma_uint32 k1; + int i; + + blocks = (const ma_uint32 *)(data + nblocks*4); + + for(i = -nblocks; i; i++) { + k1 = ma_hash_getblock(blocks,i); + + k1 *= c1; + k1 = ma_rotl32(k1, 15); + k1 *= c2; + + h1 ^= k1; + h1 = ma_rotl32(h1, 13); + h1 = h1*5 + 0xe6546b64; + } + + + tail = (const ma_uint8*)(data + nblocks*4); + + k1 = 0; + switch(len & 3) { + case 3: k1 ^= tail[2] << 16; + case 2: k1 ^= tail[1] << 8; + case 1: k1 ^= tail[0]; + k1 *= c1; k1 = ma_rotl32(k1, 15); k1 *= c2; h1 ^= k1; + }; + + + h1 ^= len; + h1 = ma_hash_fmix32(h1); + + return h1; +} + +#if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic push +#endif +/* End MurmurHash3 */ + +static ma_uint32 ma_hash_string_32(const char* str) +{ + return ma_hash_32(str, (int)strlen(str), MA_DEFAULT_HASH_SEED); +} + +static ma_uint32 ma_hash_string_w_32(const wchar_t* str) +{ + return ma_hash_32(str, (int)wcslen(str) * sizeof(*str), MA_DEFAULT_HASH_SEED); +} + + + + +/* +Basic BST Functions +*/ +static ma_result ma_resource_manager_data_buffer_node_search(ma_resource_manager* pResourceManager, ma_uint32 hashedName32, ma_resource_manager_data_buffer_node** ppDataBufferNode) +{ + ma_resource_manager_data_buffer_node* pCurrentNode; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(ppDataBufferNode != NULL); + + pCurrentNode = pResourceManager->pRootDataBufferNode; + while (pCurrentNode != NULL) { + if (hashedName32 == pCurrentNode->hashedName32) { + break; /* Found. */ + } else if (hashedName32 < pCurrentNode->hashedName32) { + pCurrentNode = pCurrentNode->pChildLo; + } else { + pCurrentNode = pCurrentNode->pChildHi; + } + } + + *ppDataBufferNode = pCurrentNode; + + if (pCurrentNode == NULL) { + return MA_DOES_NOT_EXIST; + } else { + return MA_SUCCESS; + } +} + +static ma_result ma_resource_manager_data_buffer_node_insert_point(ma_resource_manager* pResourceManager, ma_uint32 hashedName32, ma_resource_manager_data_buffer_node** ppInsertPoint) +{ + ma_result result = MA_SUCCESS; + ma_resource_manager_data_buffer_node* pCurrentNode; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(ppInsertPoint != NULL); + + *ppInsertPoint = NULL; + + if (pResourceManager->pRootDataBufferNode == NULL) { + return MA_SUCCESS; /* No items. */ + } + + /* We need to find the node that will become the parent of the new node. If a node is found that already has the same hashed name we need to return MA_ALREADY_EXISTS. */ + pCurrentNode = pResourceManager->pRootDataBufferNode; + while (pCurrentNode != NULL) { + if (hashedName32 == pCurrentNode->hashedName32) { + result = MA_ALREADY_EXISTS; + break; + } else { + if (hashedName32 < pCurrentNode->hashedName32) { + if (pCurrentNode->pChildLo == NULL) { + result = MA_SUCCESS; + break; + } else { + pCurrentNode = pCurrentNode->pChildLo; + } + } else { + if (pCurrentNode->pChildHi == NULL) { + result = MA_SUCCESS; + break; + } else { + pCurrentNode = pCurrentNode->pChildHi; + } + } + } + } + + *ppInsertPoint = pCurrentNode; + return result; +} + +static ma_result ma_resource_manager_data_buffer_node_insert_at(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode, ma_resource_manager_data_buffer_node* pInsertPoint) +{ + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + + /* The key must have been set before calling this function. */ + MA_ASSERT(pDataBufferNode->hashedName32 != 0); + + if (pInsertPoint == NULL) { + /* It's the first node. */ + pResourceManager->pRootDataBufferNode = pDataBufferNode; + } else { + /* It's not the first node. It needs to be inserted. */ + if (pDataBufferNode->hashedName32 < pInsertPoint->hashedName32) { + MA_ASSERT(pInsertPoint->pChildLo == NULL); + pInsertPoint->pChildLo = pDataBufferNode; + } else { + MA_ASSERT(pInsertPoint->pChildHi == NULL); + pInsertPoint->pChildHi = pDataBufferNode; + } + } + + pDataBufferNode->pParent = pInsertPoint; + + return MA_SUCCESS; +} + +#if 0 /* Unused for now. */ +static ma_result ma_resource_manager_data_buffer_node_insert(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + ma_result result; + ma_resource_manager_data_buffer_node* pInsertPoint; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + + result = ma_resource_manager_data_buffer_node_insert_point(pResourceManager, pDataBufferNode->hashedName32, &pInsertPoint); + if (result != MA_SUCCESS) { + return MA_INVALID_ARGS; + } + + return ma_resource_manager_data_buffer_node_insert_at(pResourceManager, pDataBufferNode, pInsertPoint); +} +#endif + +static MA_INLINE ma_resource_manager_data_buffer_node* ma_resource_manager_data_buffer_node_find_min(ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + ma_resource_manager_data_buffer_node* pCurrentNode; + + MA_ASSERT(pDataBufferNode != NULL); + + pCurrentNode = pDataBufferNode; + while (pCurrentNode->pChildLo != NULL) { + pCurrentNode = pCurrentNode->pChildLo; + } + + return pCurrentNode; +} + +static MA_INLINE ma_resource_manager_data_buffer_node* ma_resource_manager_data_buffer_node_find_max(ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + ma_resource_manager_data_buffer_node* pCurrentNode; + + MA_ASSERT(pDataBufferNode != NULL); + + pCurrentNode = pDataBufferNode; + while (pCurrentNode->pChildHi != NULL) { + pCurrentNode = pCurrentNode->pChildHi; + } + + return pCurrentNode; +} + +static MA_INLINE ma_resource_manager_data_buffer_node* ma_resource_manager_data_buffer_node_find_inorder_successor(ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + MA_ASSERT(pDataBufferNode != NULL); + MA_ASSERT(pDataBufferNode->pChildHi != NULL); + + return ma_resource_manager_data_buffer_node_find_min(pDataBufferNode->pChildHi); +} + +static MA_INLINE ma_resource_manager_data_buffer_node* ma_resource_manager_data_buffer_node_find_inorder_predecessor(ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + MA_ASSERT(pDataBufferNode != NULL); + MA_ASSERT(pDataBufferNode->pChildLo != NULL); + + return ma_resource_manager_data_buffer_node_find_max(pDataBufferNode->pChildLo); +} + +static ma_result ma_resource_manager_data_buffer_node_remove(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + + if (pDataBufferNode->pChildLo == NULL) { + if (pDataBufferNode->pChildHi == NULL) { + /* Simple case - deleting a buffer with no children. */ + if (pDataBufferNode->pParent == NULL) { + MA_ASSERT(pResourceManager->pRootDataBufferNode == pDataBufferNode); /* There is only a single buffer in the tree which should be equal to the root node. */ + pResourceManager->pRootDataBufferNode = NULL; + } else { + if (pDataBufferNode->pParent->pChildLo == pDataBufferNode) { + pDataBufferNode->pParent->pChildLo = NULL; + } else { + pDataBufferNode->pParent->pChildHi = NULL; + } + } + } else { + /* Node has one child - pChildHi != NULL. */ + pDataBufferNode->pChildHi->pParent = pDataBufferNode->pParent; + + if (pDataBufferNode->pParent == NULL) { + MA_ASSERT(pResourceManager->pRootDataBufferNode == pDataBufferNode); + pResourceManager->pRootDataBufferNode = pDataBufferNode->pChildHi; + } else { + if (pDataBufferNode->pParent->pChildLo == pDataBufferNode) { + pDataBufferNode->pParent->pChildLo = pDataBufferNode->pChildHi; + } else { + pDataBufferNode->pParent->pChildHi = pDataBufferNode->pChildHi; + } + } + } + } else { + if (pDataBufferNode->pChildHi == NULL) { + /* Node has one child - pChildLo != NULL. */ + pDataBufferNode->pChildLo->pParent = pDataBufferNode->pParent; + + if (pDataBufferNode->pParent == NULL) { + MA_ASSERT(pResourceManager->pRootDataBufferNode == pDataBufferNode); + pResourceManager->pRootDataBufferNode = pDataBufferNode->pChildLo; + } else { + if (pDataBufferNode->pParent->pChildLo == pDataBufferNode) { + pDataBufferNode->pParent->pChildLo = pDataBufferNode->pChildLo; + } else { + pDataBufferNode->pParent->pChildHi = pDataBufferNode->pChildLo; + } + } + } else { + /* Complex case - deleting a node with two children. */ + ma_resource_manager_data_buffer_node* pReplacementDataBufferNode; + + /* For now we are just going to use the in-order successor as the replacement, but we may want to try to keep this balanced by switching between the two. */ + pReplacementDataBufferNode = ma_resource_manager_data_buffer_node_find_inorder_successor(pDataBufferNode); + MA_ASSERT(pReplacementDataBufferNode != NULL); + + /* + Now that we have our replacement node we can make the change. The simple way to do this would be to just exchange the values, and then remove the replacement + node, however we track specific nodes via pointers which means we can't just swap out the values. We need to instead just change the pointers around. The + replacement node should have at most 1 child. Therefore, we can detach it in terms of our simpler cases above. What we're essentially doing is detaching the + replacement node and reinserting it into the same position as the deleted node. + */ + MA_ASSERT(pReplacementDataBufferNode->pParent != NULL); /* The replacement node should never be the root which means it should always have a parent. */ + MA_ASSERT(pReplacementDataBufferNode->pChildLo == NULL); /* Because we used in-order successor. This would be pChildHi == NULL if we used in-order predecessor. */ + + if (pReplacementDataBufferNode->pChildHi == NULL) { + if (pReplacementDataBufferNode->pParent->pChildLo == pReplacementDataBufferNode) { + pReplacementDataBufferNode->pParent->pChildLo = NULL; + } else { + pReplacementDataBufferNode->pParent->pChildHi = NULL; + } + } else { + pReplacementDataBufferNode->pChildHi->pParent = pReplacementDataBufferNode->pParent; + if (pReplacementDataBufferNode->pParent->pChildLo == pReplacementDataBufferNode) { + pReplacementDataBufferNode->pParent->pChildLo = pReplacementDataBufferNode->pChildHi; + } else { + pReplacementDataBufferNode->pParent->pChildHi = pReplacementDataBufferNode->pChildHi; + } + } + + + /* The replacement node has essentially been detached from the binary tree, so now we need to replace the old data buffer with it. The first thing to update is the parent */ + if (pDataBufferNode->pParent != NULL) { + if (pDataBufferNode->pParent->pChildLo == pDataBufferNode) { + pDataBufferNode->pParent->pChildLo = pReplacementDataBufferNode; + } else { + pDataBufferNode->pParent->pChildHi = pReplacementDataBufferNode; + } + } + + /* Now need to update the replacement node's pointers. */ + pReplacementDataBufferNode->pParent = pDataBufferNode->pParent; + pReplacementDataBufferNode->pChildLo = pDataBufferNode->pChildLo; + pReplacementDataBufferNode->pChildHi = pDataBufferNode->pChildHi; + + /* Now the children of the replacement node need to have their parent pointers updated. */ + if (pReplacementDataBufferNode->pChildLo != NULL) { + pReplacementDataBufferNode->pChildLo->pParent = pReplacementDataBufferNode; + } + if (pReplacementDataBufferNode->pChildHi != NULL) { + pReplacementDataBufferNode->pChildHi->pParent = pReplacementDataBufferNode; + } + + /* Now the root node needs to be updated. */ + if (pResourceManager->pRootDataBufferNode == pDataBufferNode) { + pResourceManager->pRootDataBufferNode = pReplacementDataBufferNode; + } + } + } + + return MA_SUCCESS; +} + +#if 0 /* Unused for now. */ +static ma_result ma_resource_manager_data_buffer_node_remove_by_key(ma_resource_manager* pResourceManager, ma_uint32 hashedName32) +{ + ma_result result; + ma_resource_manager_data_buffer_node* pDataBufferNode; + + result = ma_resource_manager_data_buffer_search(pResourceManager, hashedName32, &pDataBufferNode); + if (result != MA_SUCCESS) { + return result; /* Could not find the data buffer. */ + } + + return ma_resource_manager_data_buffer_remove(pResourceManager, pDataBufferNode); +} +#endif + +static ma_resource_manager_data_supply_type ma_resource_manager_data_buffer_node_get_data_supply_type(ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + return (ma_resource_manager_data_supply_type)ma_atomic_load_i32(&pDataBufferNode->data.type); +} + +static void ma_resource_manager_data_buffer_node_set_data_supply_type(ma_resource_manager_data_buffer_node* pDataBufferNode, ma_resource_manager_data_supply_type supplyType) +{ + ma_atomic_exchange_i32(&pDataBufferNode->data.type, supplyType); +} + +static ma_result ma_resource_manager_data_buffer_node_increment_ref(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode, ma_uint32* pNewRefCount) +{ + ma_uint32 refCount; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + + (void)pResourceManager; + + refCount = ma_atomic_fetch_add_32(&pDataBufferNode->refCount, 1) + 1; + + if (pNewRefCount != NULL) { + *pNewRefCount = refCount; + } + + return MA_SUCCESS; +} + +static ma_result ma_resource_manager_data_buffer_node_decrement_ref(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode, ma_uint32* pNewRefCount) +{ + ma_uint32 refCount; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + + (void)pResourceManager; + + refCount = ma_atomic_fetch_sub_32(&pDataBufferNode->refCount, 1) - 1; + + if (pNewRefCount != NULL) { + *pNewRefCount = refCount; + } + + return MA_SUCCESS; +} + +static void ma_resource_manager_data_buffer_node_free(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + + if (pDataBufferNode->isDataOwnedByResourceManager) { + if (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBufferNode) == ma_resource_manager_data_supply_type_encoded) { + ma_free((void*)pDataBufferNode->data.backend.encoded.pData, &pResourceManager->config.allocationCallbacks); + pDataBufferNode->data.backend.encoded.pData = NULL; + pDataBufferNode->data.backend.encoded.sizeInBytes = 0; + } else if (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBufferNode) == ma_resource_manager_data_supply_type_decoded) { + ma_free((void*)pDataBufferNode->data.backend.decoded.pData, &pResourceManager->config.allocationCallbacks); + pDataBufferNode->data.backend.decoded.pData = NULL; + pDataBufferNode->data.backend.decoded.totalFrameCount = 0; + } else if (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBufferNode) == ma_resource_manager_data_supply_type_decoded_paged) { + ma_paged_audio_buffer_data_uninit(&pDataBufferNode->data.backend.decodedPaged.data, &pResourceManager->config.allocationCallbacks); + } else { + /* Should never hit this if the node was successfully initialized. */ + MA_ASSERT(pDataBufferNode->result != MA_SUCCESS); + } + } + + /* The data buffer itself needs to be freed. */ + ma_free(pDataBufferNode, &pResourceManager->config.allocationCallbacks); +} + +static ma_result ma_resource_manager_data_buffer_node_result(const ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + MA_ASSERT(pDataBufferNode != NULL); + + return (ma_result)ma_atomic_load_i32((ma_result*)&pDataBufferNode->result); /* Need a naughty const-cast here. */ +} + + +static ma_bool32 ma_resource_manager_is_threading_enabled(const ma_resource_manager* pResourceManager) +{ + MA_ASSERT(pResourceManager != NULL); + + return (pResourceManager->config.flags & MA_RESOURCE_MANAGER_FLAG_NO_THREADING) == 0; +} + + +typedef struct +{ + union + { + ma_async_notification_event e; + ma_async_notification_poll p; + } backend; /* Must be the first member. */ + ma_resource_manager* pResourceManager; +} ma_resource_manager_inline_notification; + +static ma_result ma_resource_manager_inline_notification_init(ma_resource_manager* pResourceManager, ma_resource_manager_inline_notification* pNotification) +{ + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pNotification != NULL); + + pNotification->pResourceManager = pResourceManager; + + if (ma_resource_manager_is_threading_enabled(pResourceManager)) { + return ma_async_notification_event_init(&pNotification->backend.e); + } else { + return ma_async_notification_poll_init(&pNotification->backend.p); + } +} + +static void ma_resource_manager_inline_notification_uninit(ma_resource_manager_inline_notification* pNotification) +{ + MA_ASSERT(pNotification != NULL); + + if (ma_resource_manager_is_threading_enabled(pNotification->pResourceManager)) { + ma_async_notification_event_uninit(&pNotification->backend.e); + } else { + /* No need to uninitialize a polling notification. */ + } +} + +static void ma_resource_manager_inline_notification_wait(ma_resource_manager_inline_notification* pNotification) +{ + MA_ASSERT(pNotification != NULL); + + if (ma_resource_manager_is_threading_enabled(pNotification->pResourceManager)) { + ma_async_notification_event_wait(&pNotification->backend.e); + } else { + while (ma_async_notification_poll_is_signalled(&pNotification->backend.p) == MA_FALSE) { + ma_result result = ma_resource_manager_process_next_job(pNotification->pResourceManager); + if (result == MA_NO_DATA_AVAILABLE || result == MA_CANCELLED) { + break; + } + } + } +} + +static void ma_resource_manager_inline_notification_wait_and_uninit(ma_resource_manager_inline_notification* pNotification) +{ + ma_resource_manager_inline_notification_wait(pNotification); + ma_resource_manager_inline_notification_uninit(pNotification); +} + + +static void ma_resource_manager_data_buffer_bst_lock(ma_resource_manager* pResourceManager) +{ + MA_ASSERT(pResourceManager != NULL); + + if (ma_resource_manager_is_threading_enabled(pResourceManager)) { + #ifndef MA_NO_THREADING + { + ma_mutex_lock(&pResourceManager->dataBufferBSTLock); + } + #else + { + MA_ASSERT(MA_FALSE); /* Should never hit this. */ + } + #endif + } else { + /* Threading not enabled. Do nothing. */ + } +} + +static void ma_resource_manager_data_buffer_bst_unlock(ma_resource_manager* pResourceManager) +{ + MA_ASSERT(pResourceManager != NULL); + + if (ma_resource_manager_is_threading_enabled(pResourceManager)) { + #ifndef MA_NO_THREADING + { + ma_mutex_unlock(&pResourceManager->dataBufferBSTLock); + } + #else + { + MA_ASSERT(MA_FALSE); /* Should never hit this. */ + } + #endif + } else { + /* Threading not enabled. Do nothing. */ + } +} + +#ifndef MA_NO_THREADING +static ma_thread_result MA_THREADCALL ma_resource_manager_job_thread(void* pUserData) +{ + ma_resource_manager* pResourceManager = (ma_resource_manager*)pUserData; + MA_ASSERT(pResourceManager != NULL); + + for (;;) { + ma_result result; + ma_job job; + + result = ma_resource_manager_next_job(pResourceManager, &job); + if (result != MA_SUCCESS) { + break; + } + + /* Terminate if we got a quit message. */ + if (job.toc.breakup.code == MA_JOB_TYPE_QUIT) { + break; + } + + ma_job_process(&job); + } + + return (ma_thread_result)0; +} +#endif + +MA_API ma_resource_manager_config ma_resource_manager_config_init(void) +{ + ma_resource_manager_config config; + + MA_ZERO_OBJECT(&config); + config.decodedFormat = ma_format_unknown; + config.decodedChannels = 0; + config.decodedSampleRate = 0; + config.jobThreadCount = 1; /* A single miniaudio-managed job thread by default. */ + config.jobQueueCapacity = MA_JOB_TYPE_RESOURCE_MANAGER_QUEUE_CAPACITY; + + /* Flags. */ + config.flags = 0; + #ifdef MA_NO_THREADING + { + /* Threading is disabled at compile time so disable threading at runtime as well by default. */ + config.flags |= MA_RESOURCE_MANAGER_FLAG_NO_THREADING; + config.jobThreadCount = 0; + } + #endif + + return config; +} + + +MA_API ma_result ma_resource_manager_init(const ma_resource_manager_config* pConfig, ma_resource_manager* pResourceManager) +{ + ma_result result; + ma_job_queue_config jobQueueConfig; + + if (pResourceManager == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pResourceManager); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + #ifndef MA_NO_THREADING + { + if (pConfig->jobThreadCount > ma_countof(pResourceManager->jobThreads)) { + return MA_INVALID_ARGS; /* Requesting too many job threads. */ + } + } + #endif + + pResourceManager->config = *pConfig; + ma_allocation_callbacks_init_copy(&pResourceManager->config.allocationCallbacks, &pConfig->allocationCallbacks); + + /* Get the log set up early so we can start using it as soon as possible. */ + if (pResourceManager->config.pLog == NULL) { + result = ma_log_init(&pResourceManager->config.allocationCallbacks, &pResourceManager->log); + if (result == MA_SUCCESS) { + pResourceManager->config.pLog = &pResourceManager->log; + } else { + pResourceManager->config.pLog = NULL; /* Logging is unavailable. */ + } + } + + if (pResourceManager->config.pVFS == NULL) { + result = ma_default_vfs_init(&pResourceManager->defaultVFS, &pResourceManager->config.allocationCallbacks); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the default file system. */ + } + + pResourceManager->config.pVFS = &pResourceManager->defaultVFS; + } + + /* If threading has been disabled at compile time, enforce it at run time as well. */ + #ifdef MA_NO_THREADING + { + pResourceManager->config.flags |= MA_RESOURCE_MANAGER_FLAG_NO_THREADING; + } + #endif + + /* We need to force MA_RESOURCE_MANAGER_FLAG_NON_BLOCKING if MA_RESOURCE_MANAGER_FLAG_NO_THREADING is set. */ + if ((pResourceManager->config.flags & MA_RESOURCE_MANAGER_FLAG_NO_THREADING) != 0) { + pResourceManager->config.flags |= MA_RESOURCE_MANAGER_FLAG_NON_BLOCKING; + + /* We cannot allow job threads when MA_RESOURCE_MANAGER_FLAG_NO_THREADING has been set. This is an invalid use case. */ + if (pResourceManager->config.jobThreadCount > 0) { + return MA_INVALID_ARGS; + } + } + + /* Job queue. */ + jobQueueConfig.capacity = pResourceManager->config.jobQueueCapacity; + jobQueueConfig.flags = 0; + if ((pResourceManager->config.flags & MA_RESOURCE_MANAGER_FLAG_NON_BLOCKING) != 0) { + if (pResourceManager->config.jobThreadCount > 0) { + return MA_INVALID_ARGS; /* Non-blocking mode is only valid for self-managed job threads. */ + } + + jobQueueConfig.flags |= MA_JOB_QUEUE_FLAG_NON_BLOCKING; + } + + result = ma_job_queue_init(&jobQueueConfig, &pResourceManager->config.allocationCallbacks, &pResourceManager->jobQueue); + if (result != MA_SUCCESS) { + return result; + } + + + /* Custom decoding backends. */ + if (pConfig->ppCustomDecodingBackendVTables != NULL && pConfig->customDecodingBackendCount > 0) { + size_t sizeInBytes = sizeof(*pResourceManager->config.ppCustomDecodingBackendVTables) * pConfig->customDecodingBackendCount; + ma_decoding_backend_vtable** ppCustomDecodingBackendVTables; + + ppCustomDecodingBackendVTables = (ma_decoding_backend_vtable**)ma_malloc(sizeInBytes, &pResourceManager->config.allocationCallbacks); + if (pResourceManager->config.ppCustomDecodingBackendVTables == NULL) { + ma_job_queue_uninit(&pResourceManager->jobQueue, &pResourceManager->config.allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + + MA_COPY_MEMORY(ppCustomDecodingBackendVTables, pConfig->ppCustomDecodingBackendVTables, sizeInBytes); + + pResourceManager->config.ppCustomDecodingBackendVTables = ppCustomDecodingBackendVTables; + pResourceManager->config.customDecodingBackendCount = pConfig->customDecodingBackendCount; + pResourceManager->config.pCustomDecodingBackendUserData = pConfig->pCustomDecodingBackendUserData; + } + + + + /* Here is where we initialize our threading stuff. We don't do this if we don't support threading. */ + if (ma_resource_manager_is_threading_enabled(pResourceManager)) { + #ifndef MA_NO_THREADING + { + ma_uint32 iJobThread; + + /* Data buffer lock. */ + result = ma_mutex_init(&pResourceManager->dataBufferBSTLock); + if (result != MA_SUCCESS) { + ma_job_queue_uninit(&pResourceManager->jobQueue, &pResourceManager->config.allocationCallbacks); + return result; + } + + /* Create the job threads last to ensure the threads has access to valid data. */ + for (iJobThread = 0; iJobThread < pResourceManager->config.jobThreadCount; iJobThread += 1) { + result = ma_thread_create(&pResourceManager->jobThreads[iJobThread], ma_thread_priority_normal, pResourceManager->config.jobThreadStackSize, ma_resource_manager_job_thread, pResourceManager, &pResourceManager->config.allocationCallbacks); + if (result != MA_SUCCESS) { + ma_mutex_uninit(&pResourceManager->dataBufferBSTLock); + ma_job_queue_uninit(&pResourceManager->jobQueue, &pResourceManager->config.allocationCallbacks); + return result; + } + } + } + #else + { + /* Threading is disabled at compile time. We should never get here because validation checks should have already been performed. */ + MA_ASSERT(MA_FALSE); + } + #endif + } + + return MA_SUCCESS; +} + + +static void ma_resource_manager_delete_all_data_buffer_nodes(ma_resource_manager* pResourceManager) +{ + MA_ASSERT(pResourceManager); + + /* If everything was done properly, there shouldn't be any active data buffers. */ + while (pResourceManager->pRootDataBufferNode != NULL) { + ma_resource_manager_data_buffer_node* pDataBufferNode = pResourceManager->pRootDataBufferNode; + ma_resource_manager_data_buffer_node_remove(pResourceManager, pDataBufferNode); + + /* The data buffer has been removed from the BST, so now we need to free its data. */ + ma_resource_manager_data_buffer_node_free(pResourceManager, pDataBufferNode); + } +} + +MA_API void ma_resource_manager_uninit(ma_resource_manager* pResourceManager) +{ + if (pResourceManager == NULL) { + return; + } + + /* + Job threads need to be killed first. To do this we need to post a quit message to the message queue and then wait for the thread. The quit message will never be removed from the + queue which means it will never not be returned after being encountered for the first time which means all threads will eventually receive it. + */ + ma_resource_manager_post_job_quit(pResourceManager); + + /* Wait for every job to finish before continuing to ensure nothing is sill trying to access any of our objects below. */ + if (ma_resource_manager_is_threading_enabled(pResourceManager)) { + #ifndef MA_NO_THREADING + { + ma_uint32 iJobThread; + + for (iJobThread = 0; iJobThread < pResourceManager->config.jobThreadCount; iJobThread += 1) { + ma_thread_wait(&pResourceManager->jobThreads[iJobThread]); + } + } + #else + { + MA_ASSERT(MA_FALSE); /* Should never hit this. */ + } + #endif + } + + /* At this point the thread should have returned and no other thread should be accessing our data. We can now delete all data buffers. */ + ma_resource_manager_delete_all_data_buffer_nodes(pResourceManager); + + /* The job queue is no longer needed. */ + ma_job_queue_uninit(&pResourceManager->jobQueue, &pResourceManager->config.allocationCallbacks); + + /* We're no longer doing anything with data buffers so the lock can now be uninitialized. */ + if (ma_resource_manager_is_threading_enabled(pResourceManager)) { + #ifndef MA_NO_THREADING + { + ma_mutex_uninit(&pResourceManager->dataBufferBSTLock); + } + #else + { + MA_ASSERT(MA_FALSE); /* Should never hit this. */ + } + #endif + } + + ma_free((ma_decoding_backend_vtable**)pResourceManager->config.ppCustomDecodingBackendVTables, &pResourceManager->config.allocationCallbacks); /* <-- Naughty const-cast, but this is safe. */ + + if (pResourceManager->config.pLog == &pResourceManager->log) { + ma_log_uninit(&pResourceManager->log); + } +} + +MA_API ma_log* ma_resource_manager_get_log(ma_resource_manager* pResourceManager) +{ + if (pResourceManager == NULL) { + return NULL; + } + + return pResourceManager->config.pLog; +} + + + +MA_API ma_resource_manager_data_source_config ma_resource_manager_data_source_config_init(void) +{ + ma_resource_manager_data_source_config config; + + MA_ZERO_OBJECT(&config); + config.rangeBegInPCMFrames = MA_DATA_SOURCE_DEFAULT_RANGE_BEG; + config.rangeEndInPCMFrames = MA_DATA_SOURCE_DEFAULT_RANGE_END; + config.loopPointBegInPCMFrames = MA_DATA_SOURCE_DEFAULT_LOOP_POINT_BEG; + config.loopPointEndInPCMFrames = MA_DATA_SOURCE_DEFAULT_LOOP_POINT_END; + config.isLooping = MA_FALSE; + + return config; +} + + +static ma_decoder_config ma_resource_manager__init_decoder_config(ma_resource_manager* pResourceManager) +{ + ma_decoder_config config; + + config = ma_decoder_config_init(pResourceManager->config.decodedFormat, pResourceManager->config.decodedChannels, pResourceManager->config.decodedSampleRate); + config.allocationCallbacks = pResourceManager->config.allocationCallbacks; + config.ppCustomBackendVTables = pResourceManager->config.ppCustomDecodingBackendVTables; + config.customBackendCount = pResourceManager->config.customDecodingBackendCount; + config.pCustomBackendUserData = pResourceManager->config.pCustomDecodingBackendUserData; + + return config; +} + +static ma_result ma_resource_manager__init_decoder(ma_resource_manager* pResourceManager, const char* pFilePath, const wchar_t* pFilePathW, ma_decoder* pDecoder) +{ + ma_result result; + ma_decoder_config config; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pFilePath != NULL || pFilePathW != NULL); + MA_ASSERT(pDecoder != NULL); + + config = ma_resource_manager__init_decoder_config(pResourceManager); + + if (pFilePath != NULL) { + result = ma_decoder_init_vfs(pResourceManager->config.pVFS, pFilePath, &config, pDecoder); + if (result != MA_SUCCESS) { + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_WARNING, "Failed to load file \"%s\". %s.\n", pFilePath, ma_result_description(result)); + return result; + } + } else { + result = ma_decoder_init_vfs_w(pResourceManager->config.pVFS, pFilePathW, &config, pDecoder); + if (result != MA_SUCCESS) { + #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || defined(_MSC_VER) + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_WARNING, "Failed to load file \"%ls\". %s.\n", pFilePathW, ma_result_description(result)); + #endif + return result; + } + } + + return MA_SUCCESS; +} + +static ma_bool32 ma_resource_manager_data_buffer_has_connector(ma_resource_manager_data_buffer* pDataBuffer) +{ + return ma_atomic_bool32_get(&pDataBuffer->isConnectorInitialized); +} + +static ma_data_source* ma_resource_manager_data_buffer_get_connector(ma_resource_manager_data_buffer* pDataBuffer) +{ + if (ma_resource_manager_data_buffer_has_connector(pDataBuffer) == MA_FALSE) { + return NULL; /* Connector not yet initialized. */ + } + + switch (pDataBuffer->pNode->data.type) + { + case ma_resource_manager_data_supply_type_encoded: return &pDataBuffer->connector.decoder; + case ma_resource_manager_data_supply_type_decoded: return &pDataBuffer->connector.buffer; + case ma_resource_manager_data_supply_type_decoded_paged: return &pDataBuffer->connector.pagedBuffer; + + case ma_resource_manager_data_supply_type_unknown: + default: + { + ma_log_postf(ma_resource_manager_get_log(pDataBuffer->pResourceManager), MA_LOG_LEVEL_ERROR, "Failed to retrieve data buffer connector. Unknown data supply type.\n"); + return NULL; + }; + }; +} + +static ma_result ma_resource_manager_data_buffer_init_connector(ma_resource_manager_data_buffer* pDataBuffer, const ma_resource_manager_data_source_config* pConfig, ma_async_notification* pInitNotification, ma_fence* pInitFence) +{ + ma_result result; + + MA_ASSERT(pDataBuffer != NULL); + MA_ASSERT(pConfig != NULL); + MA_ASSERT(ma_resource_manager_data_buffer_has_connector(pDataBuffer) == MA_FALSE); + + /* The underlying data buffer must be initialized before we'll be able to know how to initialize the backend. */ + result = ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode); + if (result != MA_SUCCESS && result != MA_BUSY) { + return result; /* The data buffer is in an erroneous state. */ + } + + /* + We need to initialize either a ma_decoder or an ma_audio_buffer depending on whether or not the backing data is encoded or decoded. These act as the + "instance" to the data and are used to form the connection between underlying data buffer and the data source. If the data buffer is decoded, we can use + an ma_audio_buffer. This enables us to use memory mapping when mixing which saves us a bit of data movement overhead. + */ + switch (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode)) + { + case ma_resource_manager_data_supply_type_encoded: /* Connector is a decoder. */ + { + ma_decoder_config config; + config = ma_resource_manager__init_decoder_config(pDataBuffer->pResourceManager); + result = ma_decoder_init_memory(pDataBuffer->pNode->data.backend.encoded.pData, pDataBuffer->pNode->data.backend.encoded.sizeInBytes, &config, &pDataBuffer->connector.decoder); + } break; + + case ma_resource_manager_data_supply_type_decoded: /* Connector is an audio buffer. */ + { + ma_audio_buffer_config config; + config = ma_audio_buffer_config_init(pDataBuffer->pNode->data.backend.decoded.format, pDataBuffer->pNode->data.backend.decoded.channels, pDataBuffer->pNode->data.backend.decoded.totalFrameCount, pDataBuffer->pNode->data.backend.decoded.pData, NULL); + result = ma_audio_buffer_init(&config, &pDataBuffer->connector.buffer); + } break; + + case ma_resource_manager_data_supply_type_decoded_paged: /* Connector is a paged audio buffer. */ + { + ma_paged_audio_buffer_config config; + config = ma_paged_audio_buffer_config_init(&pDataBuffer->pNode->data.backend.decodedPaged.data); + result = ma_paged_audio_buffer_init(&config, &pDataBuffer->connector.pagedBuffer); + } break; + + case ma_resource_manager_data_supply_type_unknown: + default: + { + /* Unknown data supply type. Should never happen. Need to post an error here. */ + return MA_INVALID_ARGS; + }; + } + + /* + Initialization of the connector is when we can fire the init notification. This will give the application access to + the format/channels/rate of the data source. + */ + if (result == MA_SUCCESS) { + /* + The resource manager supports the ability to set the range and loop settings via a config at + initialization time. This results in an case where the ranges could be set explicitly via + ma_data_source_set_*() before we get to this point here. If this happens, we'll end up + hitting a case where we just override those settings which results in what feels like a bug. + + To address this we only change the relevant properties if they're not equal to defaults. If + they're equal to defaults there's no need to change them anyway. If they're *not* set to the + default values, we can assume the user has set the range and loop settings via the config. If + they're doing their own calls to ma_data_source_set_*() in addition to setting them via the + config, that's entirely on the caller and any synchronization issue becomes their problem. + */ + if (pConfig->rangeBegInPCMFrames != MA_DATA_SOURCE_DEFAULT_RANGE_BEG || pConfig->rangeEndInPCMFrames != MA_DATA_SOURCE_DEFAULT_RANGE_END) { + ma_data_source_set_range_in_pcm_frames(pDataBuffer, pConfig->rangeBegInPCMFrames, pConfig->rangeEndInPCMFrames); + } + + if (pConfig->loopPointBegInPCMFrames != MA_DATA_SOURCE_DEFAULT_LOOP_POINT_BEG || pConfig->loopPointEndInPCMFrames != MA_DATA_SOURCE_DEFAULT_LOOP_POINT_END) { + ma_data_source_set_loop_point_in_pcm_frames(pDataBuffer, pConfig->loopPointBegInPCMFrames, pConfig->loopPointEndInPCMFrames); + } + + if (pConfig->isLooping != MA_FALSE) { + ma_data_source_set_looping(pDataBuffer, pConfig->isLooping); + } + + ma_atomic_bool32_set(&pDataBuffer->isConnectorInitialized, MA_TRUE); + + if (pInitNotification != NULL) { + ma_async_notification_signal(pInitNotification); + } + + if (pInitFence != NULL) { + ma_fence_release(pInitFence); + } + } + + /* At this point the backend should be initialized. We do *not* want to set pDataSource->result here - that needs to be done at a higher level to ensure it's done as the last step. */ + return result; +} + +static ma_result ma_resource_manager_data_buffer_uninit_connector(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer* pDataBuffer) +{ + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBuffer != NULL); + + (void)pResourceManager; + + switch (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode)) + { + case ma_resource_manager_data_supply_type_encoded: /* Connector is a decoder. */ + { + ma_decoder_uninit(&pDataBuffer->connector.decoder); + } break; + + case ma_resource_manager_data_supply_type_decoded: /* Connector is an audio buffer. */ + { + ma_audio_buffer_uninit(&pDataBuffer->connector.buffer); + } break; + + case ma_resource_manager_data_supply_type_decoded_paged: /* Connector is a paged audio buffer. */ + { + ma_paged_audio_buffer_uninit(&pDataBuffer->connector.pagedBuffer); + } break; + + case ma_resource_manager_data_supply_type_unknown: + default: + { + /* Unknown data supply type. Should never happen. Need to post an error here. */ + return MA_INVALID_ARGS; + }; + } + + return MA_SUCCESS; +} + +static ma_uint32 ma_resource_manager_data_buffer_node_next_execution_order(ma_resource_manager_data_buffer_node* pDataBufferNode) +{ + MA_ASSERT(pDataBufferNode != NULL); + return ma_atomic_fetch_add_32(&pDataBufferNode->executionCounter, 1); +} + +static ma_result ma_resource_manager_data_buffer_node_init_supply_encoded(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode, const char* pFilePath, const wchar_t* pFilePathW) +{ + ma_result result; + size_t dataSizeInBytes; + void* pData; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + MA_ASSERT(pFilePath != NULL || pFilePathW != NULL); + + result = ma_vfs_open_and_read_file_ex(pResourceManager->config.pVFS, pFilePath, pFilePathW, &pData, &dataSizeInBytes, &pResourceManager->config.allocationCallbacks); + if (result != MA_SUCCESS) { + if (pFilePath != NULL) { + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_WARNING, "Failed to load file \"%s\". %s.\n", pFilePath, ma_result_description(result)); + } else { + #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || defined(_MSC_VER) + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_WARNING, "Failed to load file \"%ls\". %s.\n", pFilePathW, ma_result_description(result)); + #endif + } + + return result; + } + + pDataBufferNode->data.backend.encoded.pData = pData; + pDataBufferNode->data.backend.encoded.sizeInBytes = dataSizeInBytes; + ma_resource_manager_data_buffer_node_set_data_supply_type(pDataBufferNode, ma_resource_manager_data_supply_type_encoded); /* <-- Must be set last. */ + + return MA_SUCCESS; +} + +static ma_result ma_resource_manager_data_buffer_node_init_supply_decoded(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode, const char* pFilePath, const wchar_t* pFilePathW, ma_uint32 flags, ma_decoder** ppDecoder) +{ + ma_result result = MA_SUCCESS; + ma_decoder* pDecoder; + ma_uint64 totalFrameCount; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + MA_ASSERT(ppDecoder != NULL); + MA_ASSERT(pFilePath != NULL || pFilePathW != NULL); + + *ppDecoder = NULL; /* For safety. */ + + pDecoder = (ma_decoder*)ma_malloc(sizeof(*pDecoder), &pResourceManager->config.allocationCallbacks); + if (pDecoder == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_resource_manager__init_decoder(pResourceManager, pFilePath, pFilePathW, pDecoder); + if (result != MA_SUCCESS) { + ma_free(pDecoder, &pResourceManager->config.allocationCallbacks); + return result; + } + + /* + At this point we have the decoder and we now need to initialize the data supply. This will + be either a decoded buffer, or a decoded paged buffer. A regular buffer is just one big heap + allocated buffer, whereas a paged buffer is a linked list of paged-sized buffers. The latter + is used when the length of a sound is unknown until a full decode has been performed. + */ + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_UNKNOWN_LENGTH) == 0) { + result = ma_decoder_get_length_in_pcm_frames(pDecoder, &totalFrameCount); + if (result != MA_SUCCESS) { + return result; + } + } else { + totalFrameCount = 0; + } + + if (totalFrameCount > 0) { + /* It's a known length. The data supply is a regular decoded buffer. */ + ma_uint64 dataSizeInBytes; + void* pData; + + dataSizeInBytes = totalFrameCount * ma_get_bytes_per_frame(pDecoder->outputFormat, pDecoder->outputChannels); + if (dataSizeInBytes > MA_SIZE_MAX) { + ma_decoder_uninit(pDecoder); + ma_free(pDecoder, &pResourceManager->config.allocationCallbacks); + return MA_TOO_BIG; + } + + pData = ma_malloc((size_t)dataSizeInBytes, &pResourceManager->config.allocationCallbacks); + if (pData == NULL) { + ma_decoder_uninit(pDecoder); + ma_free(pDecoder, &pResourceManager->config.allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + + /* The buffer needs to be initialized to silence in case the caller reads from it. */ + ma_silence_pcm_frames(pData, totalFrameCount, pDecoder->outputFormat, pDecoder->outputChannels); + + /* Data has been allocated and the data supply can now be initialized. */ + pDataBufferNode->data.backend.decoded.pData = pData; + pDataBufferNode->data.backend.decoded.totalFrameCount = totalFrameCount; + pDataBufferNode->data.backend.decoded.format = pDecoder->outputFormat; + pDataBufferNode->data.backend.decoded.channels = pDecoder->outputChannels; + pDataBufferNode->data.backend.decoded.sampleRate = pDecoder->outputSampleRate; + pDataBufferNode->data.backend.decoded.decodedFrameCount = 0; + ma_resource_manager_data_buffer_node_set_data_supply_type(pDataBufferNode, ma_resource_manager_data_supply_type_decoded); /* <-- Must be set last. */ + } else { + /* + It's an unknown length. The data supply is a paged decoded buffer. Setting this up is + actually easier than the non-paged decoded buffer because we just need to initialize + a ma_paged_audio_buffer object. + */ + result = ma_paged_audio_buffer_data_init(pDecoder->outputFormat, pDecoder->outputChannels, &pDataBufferNode->data.backend.decodedPaged.data); + if (result != MA_SUCCESS) { + ma_decoder_uninit(pDecoder); + ma_free(pDecoder, &pResourceManager->config.allocationCallbacks); + return result; + } + + pDataBufferNode->data.backend.decodedPaged.sampleRate = pDecoder->outputSampleRate; + pDataBufferNode->data.backend.decodedPaged.decodedFrameCount = 0; + ma_resource_manager_data_buffer_node_set_data_supply_type(pDataBufferNode, ma_resource_manager_data_supply_type_decoded_paged); /* <-- Must be set last. */ + } + + *ppDecoder = pDecoder; + + return MA_SUCCESS; +} + +static ma_result ma_resource_manager_data_buffer_node_decode_next_page(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode, ma_decoder* pDecoder) +{ + ma_result result = MA_SUCCESS; + ma_uint64 pageSizeInFrames; + ma_uint64 framesToTryReading; + ma_uint64 framesRead; + + MA_ASSERT(pResourceManager != NULL); + MA_ASSERT(pDataBufferNode != NULL); + MA_ASSERT(pDecoder != NULL); + + /* We need to know the size of a page in frames to know how many frames to decode. */ + pageSizeInFrames = MA_RESOURCE_MANAGER_PAGE_SIZE_IN_MILLISECONDS * (pDecoder->outputSampleRate/1000); + framesToTryReading = pageSizeInFrames; + + /* + Here is where we do the decoding of the next page. We'll run a slightly different path depending + on whether or not we're using a flat or paged buffer because the allocation of the page differs + between the two. For a flat buffer it's an offset to an already-allocated buffer. For a paged + buffer, we need to allocate a new page and attach it to the linked list. + */ + switch (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBufferNode)) + { + case ma_resource_manager_data_supply_type_decoded: + { + /* The destination buffer is an offset to the existing buffer. Don't read more than we originally retrieved when we first initialized the decoder. */ + void* pDst; + ma_uint64 framesRemaining = pDataBufferNode->data.backend.decoded.totalFrameCount - pDataBufferNode->data.backend.decoded.decodedFrameCount; + if (framesToTryReading > framesRemaining) { + framesToTryReading = framesRemaining; + } + + if (framesToTryReading > 0) { + pDst = ma_offset_ptr( + pDataBufferNode->data.backend.decoded.pData, + pDataBufferNode->data.backend.decoded.decodedFrameCount * ma_get_bytes_per_frame(pDataBufferNode->data.backend.decoded.format, pDataBufferNode->data.backend.decoded.channels) + ); + MA_ASSERT(pDst != NULL); + + result = ma_decoder_read_pcm_frames(pDecoder, pDst, framesToTryReading, &framesRead); + if (framesRead > 0) { + pDataBufferNode->data.backend.decoded.decodedFrameCount += framesRead; + } + } else { + framesRead = 0; + } + } break; + + case ma_resource_manager_data_supply_type_decoded_paged: + { + /* The destination buffer is a freshly allocated page. */ + ma_paged_audio_buffer_page* pPage; + + result = ma_paged_audio_buffer_data_allocate_page(&pDataBufferNode->data.backend.decodedPaged.data, framesToTryReading, NULL, &pResourceManager->config.allocationCallbacks, &pPage); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_decoder_read_pcm_frames(pDecoder, pPage->pAudioData, framesToTryReading, &framesRead); + if (result == MA_SUCCESS && framesRead > 0) { + pPage->sizeInFrames = framesRead; + + result = ma_paged_audio_buffer_data_append_page(&pDataBufferNode->data.backend.decodedPaged.data, pPage); + if (result == MA_SUCCESS) { + pDataBufferNode->data.backend.decodedPaged.decodedFrameCount += framesRead; + } else { + /* Failed to append the page. Just abort and set the status to MA_AT_END. */ + ma_paged_audio_buffer_data_free_page(&pDataBufferNode->data.backend.decodedPaged.data, pPage, &pResourceManager->config.allocationCallbacks); + result = MA_AT_END; + } + } else { + /* No frames were read. Free the page and just set the status to MA_AT_END. */ + ma_paged_audio_buffer_data_free_page(&pDataBufferNode->data.backend.decodedPaged.data, pPage, &pResourceManager->config.allocationCallbacks); + result = MA_AT_END; + } + } break; + + case ma_resource_manager_data_supply_type_encoded: + case ma_resource_manager_data_supply_type_unknown: + default: + { + /* Unexpected data supply type. */ + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_ERROR, "Unexpected data supply type (%d) when decoding page.", ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBufferNode)); + return MA_ERROR; + }; + } + + if (result == MA_SUCCESS && framesRead == 0) { + result = MA_AT_END; + } + + return result; +} + +static ma_result ma_resource_manager_data_buffer_node_acquire_critical_section(ma_resource_manager* pResourceManager, const char* pFilePath, const wchar_t* pFilePathW, ma_uint32 hashedName32, ma_uint32 flags, const ma_resource_manager_data_supply* pExistingData, ma_fence* pInitFence, ma_fence* pDoneFence, ma_resource_manager_inline_notification* pInitNotification, ma_resource_manager_data_buffer_node** ppDataBufferNode) +{ + ma_result result = MA_SUCCESS; + ma_resource_manager_data_buffer_node* pDataBufferNode = NULL; + ma_resource_manager_data_buffer_node* pInsertPoint; + + if (ppDataBufferNode != NULL) { + *ppDataBufferNode = NULL; + } + + result = ma_resource_manager_data_buffer_node_insert_point(pResourceManager, hashedName32, &pInsertPoint); + if (result == MA_ALREADY_EXISTS) { + /* The node already exists. We just need to increment the reference count. */ + pDataBufferNode = pInsertPoint; + + result = ma_resource_manager_data_buffer_node_increment_ref(pResourceManager, pDataBufferNode, NULL); + if (result != MA_SUCCESS) { + return result; /* Should never happen. Failed to increment the reference count. */ + } + + result = MA_ALREADY_EXISTS; + goto done; + } else { + /* + The node does not already exist. We need to post a LOAD_DATA_BUFFER_NODE job here. This + needs to be done inside the critical section to ensure an uninitialization of the node + does not occur before initialization on another thread. + */ + pDataBufferNode = (ma_resource_manager_data_buffer_node*)ma_malloc(sizeof(*pDataBufferNode), &pResourceManager->config.allocationCallbacks); + if (pDataBufferNode == NULL) { + return MA_OUT_OF_MEMORY; + } + + MA_ZERO_OBJECT(pDataBufferNode); + pDataBufferNode->hashedName32 = hashedName32; + pDataBufferNode->refCount = 1; /* Always set to 1 by default (this is our first reference). */ + + if (pExistingData == NULL) { + pDataBufferNode->data.type = ma_resource_manager_data_supply_type_unknown; /* <-- We won't know this until we start decoding. */ + pDataBufferNode->result = MA_BUSY; /* Must be set to MA_BUSY before we leave the critical section, so might as well do it now. */ + pDataBufferNode->isDataOwnedByResourceManager = MA_TRUE; + } else { + pDataBufferNode->data = *pExistingData; + pDataBufferNode->result = MA_SUCCESS; /* Not loading asynchronously, so just set the status */ + pDataBufferNode->isDataOwnedByResourceManager = MA_FALSE; + } + + result = ma_resource_manager_data_buffer_node_insert_at(pResourceManager, pDataBufferNode, pInsertPoint); + if (result != MA_SUCCESS) { + ma_free(pDataBufferNode, &pResourceManager->config.allocationCallbacks); + return result; /* Should never happen. Failed to insert the data buffer into the BST. */ + } + + /* + Here is where we'll post the job, but only if we're loading asynchronously. If we're + loading synchronously we'll defer loading to a later stage, outside of the critical + section. + */ + if (pDataBufferNode->isDataOwnedByResourceManager && (flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC) != 0) { + /* Loading asynchronously. Post the job. */ + ma_job job; + char* pFilePathCopy = NULL; + wchar_t* pFilePathWCopy = NULL; + + /* We need a copy of the file path. We should probably make this more efficient, but for now we'll do a transient memory allocation. */ + if (pFilePath != NULL) { + pFilePathCopy = ma_copy_string(pFilePath, &pResourceManager->config.allocationCallbacks); + } else { + pFilePathWCopy = ma_copy_string_w(pFilePathW, &pResourceManager->config.allocationCallbacks); + } + + if (pFilePathCopy == NULL && pFilePathWCopy == NULL) { + ma_resource_manager_data_buffer_node_remove(pResourceManager, pDataBufferNode); + ma_free(pDataBufferNode, &pResourceManager->config.allocationCallbacks); + return MA_OUT_OF_MEMORY; + } + + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + ma_resource_manager_inline_notification_init(pResourceManager, pInitNotification); + } + + /* Acquire init and done fences before posting the job. These will be unacquired by the job thread. */ + if (pInitFence != NULL) { ma_fence_acquire(pInitFence); } + if (pDoneFence != NULL) { ma_fence_acquire(pDoneFence); } + + /* We now have everything we need to post the job to the job thread. */ + job = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER_NODE); + job.order = ma_resource_manager_data_buffer_node_next_execution_order(pDataBufferNode); + job.data.resourceManager.loadDataBufferNode.pResourceManager = pResourceManager; + job.data.resourceManager.loadDataBufferNode.pDataBufferNode = pDataBufferNode; + job.data.resourceManager.loadDataBufferNode.pFilePath = pFilePathCopy; + job.data.resourceManager.loadDataBufferNode.pFilePathW = pFilePathWCopy; + job.data.resourceManager.loadDataBufferNode.flags = flags; + job.data.resourceManager.loadDataBufferNode.pInitNotification = ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) ? pInitNotification : NULL; + job.data.resourceManager.loadDataBufferNode.pDoneNotification = NULL; + job.data.resourceManager.loadDataBufferNode.pInitFence = pInitFence; + job.data.resourceManager.loadDataBufferNode.pDoneFence = pDoneFence; + + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + result = ma_job_process(&job); + } else { + result = ma_resource_manager_post_job(pResourceManager, &job); + } + + if (result != MA_SUCCESS) { + /* Failed to post job. Probably ran out of memory. */ + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_ERROR, "Failed to post MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER_NODE job. %s.\n", ma_result_description(result)); + + /* + Fences were acquired before posting the job, but since the job was not able to + be posted, we need to make sure we release them so nothing gets stuck waiting. + */ + if (pInitFence != NULL) { ma_fence_release(pInitFence); } + if (pDoneFence != NULL) { ma_fence_release(pDoneFence); } + + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + ma_resource_manager_inline_notification_uninit(pInitNotification); + } else { + /* These will have been freed by the job thread, but with WAIT_INIT they will already have happened since the job has already been handled. */ + ma_free(pFilePathCopy, &pResourceManager->config.allocationCallbacks); + ma_free(pFilePathWCopy, &pResourceManager->config.allocationCallbacks); + } + + ma_resource_manager_data_buffer_node_remove(pResourceManager, pDataBufferNode); + ma_free(pDataBufferNode, &pResourceManager->config.allocationCallbacks); + + return result; + } + } + } + +done: + if (ppDataBufferNode != NULL) { + *ppDataBufferNode = pDataBufferNode; + } + + return result; +} + +static ma_result ma_resource_manager_data_buffer_node_acquire(ma_resource_manager* pResourceManager, const char* pFilePath, const wchar_t* pFilePathW, ma_uint32 hashedName32, ma_uint32 flags, const ma_resource_manager_data_supply* pExistingData, ma_fence* pInitFence, ma_fence* pDoneFence, ma_resource_manager_data_buffer_node** ppDataBufferNode) +{ + ma_result result = MA_SUCCESS; + ma_bool32 nodeAlreadyExists = MA_FALSE; + ma_resource_manager_data_buffer_node* pDataBufferNode = NULL; + ma_resource_manager_inline_notification initNotification; /* Used when the WAIT_INIT flag is set. */ + + if (ppDataBufferNode != NULL) { + *ppDataBufferNode = NULL; /* Safety. */ + } + + if (pResourceManager == NULL || (pFilePath == NULL && pFilePathW == NULL && hashedName32 == 0)) { + return MA_INVALID_ARGS; + } + + /* If we're specifying existing data, it must be valid. */ + if (pExistingData != NULL && pExistingData->type == ma_resource_manager_data_supply_type_unknown) { + return MA_INVALID_ARGS; + } + + /* If we don't support threading, remove the ASYNC flag to make the rest of this a bit simpler. */ + if (ma_resource_manager_is_threading_enabled(pResourceManager) == MA_FALSE) { + flags &= ~MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC; + } + + if (hashedName32 == 0) { + if (pFilePath != NULL) { + hashedName32 = ma_hash_string_32(pFilePath); + } else { + hashedName32 = ma_hash_string_w_32(pFilePathW); + } + } + + /* + Here is where we either increment the node's reference count or allocate a new one and add it + to the BST. When allocating a new node, we need to make sure the LOAD_DATA_BUFFER_NODE job is + posted inside the critical section just in case the caller immediately uninitializes the node + as this will ensure the FREE_DATA_BUFFER_NODE job is given an execution order such that the + node is not uninitialized before initialization. + */ + ma_resource_manager_data_buffer_bst_lock(pResourceManager); + { + result = ma_resource_manager_data_buffer_node_acquire_critical_section(pResourceManager, pFilePath, pFilePathW, hashedName32, flags, pExistingData, pInitFence, pDoneFence, &initNotification, &pDataBufferNode); + } + ma_resource_manager_data_buffer_bst_unlock(pResourceManager); + + if (result == MA_ALREADY_EXISTS) { + nodeAlreadyExists = MA_TRUE; + result = MA_SUCCESS; + } else { + if (result != MA_SUCCESS) { + return result; + } + } + + /* + If we're loading synchronously, we'll need to load everything now. When loading asynchronously, + a job will have been posted inside the BST critical section so that an uninitialization can be + allocated an appropriate execution order thereby preventing it from being uninitialized before + the node is initialized by the decoding thread(s). + */ + if (nodeAlreadyExists == MA_FALSE) { /* Don't need to try loading anything if the node already exists. */ + if (pFilePath == NULL && pFilePathW == NULL) { + /* + If this path is hit, it means a buffer is being copied (i.e. initialized from only the + hashed name), but that node has been freed in the meantime, probably from some other + thread. This is an invalid operation. + */ + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_WARNING, "Cloning data buffer node failed because the source node was released. The source node must remain valid until the cloning has completed.\n"); + result = MA_INVALID_OPERATION; + goto done; + } + + if (pDataBufferNode->isDataOwnedByResourceManager) { + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC) == 0) { + /* Loading synchronously. Load the sound in it's entirety here. */ + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_DECODE) == 0) { + /* No decoding. This is the simple case - just store the file contents in memory. */ + result = ma_resource_manager_data_buffer_node_init_supply_encoded(pResourceManager, pDataBufferNode, pFilePath, pFilePathW); + if (result != MA_SUCCESS) { + goto done; + } + } else { + /* Decoding. We do this the same way as we do when loading asynchronously. */ + ma_decoder* pDecoder; + result = ma_resource_manager_data_buffer_node_init_supply_decoded(pResourceManager, pDataBufferNode, pFilePath, pFilePathW, flags, &pDecoder); + if (result != MA_SUCCESS) { + goto done; + } + + /* We have the decoder, now decode page by page just like we do when loading asynchronously. */ + for (;;) { + /* Decode next page. */ + result = ma_resource_manager_data_buffer_node_decode_next_page(pResourceManager, pDataBufferNode, pDecoder); + if (result != MA_SUCCESS) { + break; /* Will return MA_AT_END when the last page has been decoded. */ + } + } + + /* Reaching the end needs to be considered successful. */ + if (result == MA_AT_END) { + result = MA_SUCCESS; + } + + /* + At this point the data buffer is either fully decoded or some error occurred. Either + way, the decoder is no longer necessary. + */ + ma_decoder_uninit(pDecoder); + ma_free(pDecoder, &pResourceManager->config.allocationCallbacks); + } + + /* Getting here means we were successful. Make sure the status of the node is updated accordingly. */ + ma_atomic_exchange_i32(&pDataBufferNode->result, result); + } else { + /* Loading asynchronously. We may need to wait for initialization. */ + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + ma_resource_manager_inline_notification_wait(&initNotification); + } + } + } else { + /* The data is not managed by the resource manager so there's nothing else to do. */ + MA_ASSERT(pExistingData != NULL); + } + } + +done: + /* If we failed to initialize the data buffer we need to free it. */ + if (result != MA_SUCCESS) { + if (nodeAlreadyExists == MA_FALSE) { + ma_resource_manager_data_buffer_node_remove(pResourceManager, pDataBufferNode); + ma_free(pDataBufferNode, &pResourceManager->config.allocationCallbacks); + } + } + + /* + The init notification needs to be uninitialized. This will be used if the node does not already + exist, and we've specified ASYNC | WAIT_INIT. + */ + if (nodeAlreadyExists == MA_FALSE && pDataBufferNode->isDataOwnedByResourceManager && (flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC) != 0) { + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + ma_resource_manager_inline_notification_uninit(&initNotification); + } + } + + if (ppDataBufferNode != NULL) { + *ppDataBufferNode = pDataBufferNode; + } + + return result; +} + +static ma_result ma_resource_manager_data_buffer_node_unacquire(ma_resource_manager* pResourceManager, ma_resource_manager_data_buffer_node* pDataBufferNode, const char* pName, const wchar_t* pNameW) +{ + ma_result result = MA_SUCCESS; + ma_uint32 refCount = 0xFFFFFFFF; /* The new reference count of the node after decrementing. Initialize to non-0 to be safe we don't fall into the freeing path. */ + ma_uint32 hashedName32 = 0; + + if (pResourceManager == NULL) { + return MA_INVALID_ARGS; + } + + if (pDataBufferNode == NULL) { + if (pName == NULL && pNameW == NULL) { + return MA_INVALID_ARGS; + } + + if (pName != NULL) { + hashedName32 = ma_hash_string_32(pName); + } else { + hashedName32 = ma_hash_string_w_32(pNameW); + } + } + + /* + The first thing to do is decrement the reference counter of the node. Then, if the reference + count is zero, we need to free the node. If the node is still in the process of loading, we'll + need to post a job to the job queue to free the node. Otherwise we'll just do it here. + */ + ma_resource_manager_data_buffer_bst_lock(pResourceManager); + { + /* Might need to find the node. Must be done inside the critical section. */ + if (pDataBufferNode == NULL) { + result = ma_resource_manager_data_buffer_node_search(pResourceManager, hashedName32, &pDataBufferNode); + if (result != MA_SUCCESS) { + goto stage2; /* Couldn't find the node. */ + } + } + + result = ma_resource_manager_data_buffer_node_decrement_ref(pResourceManager, pDataBufferNode, &refCount); + if (result != MA_SUCCESS) { + goto stage2; /* Should never happen. */ + } + + if (refCount == 0) { + result = ma_resource_manager_data_buffer_node_remove(pResourceManager, pDataBufferNode); + if (result != MA_SUCCESS) { + goto stage2; /* An error occurred when trying to remove the data buffer. This should never happen. */ + } + } + } + ma_resource_manager_data_buffer_bst_unlock(pResourceManager); + +stage2: + if (result != MA_SUCCESS) { + return result; + } + + /* + Here is where we need to free the node. We don't want to do this inside the critical section + above because we want to keep that as small as possible for multi-threaded efficiency. + */ + if (refCount == 0) { + if (ma_resource_manager_data_buffer_node_result(pDataBufferNode) == MA_BUSY) { + /* The sound is still loading. We need to delay the freeing of the node to a safe time. */ + ma_job job; + + /* We need to mark the node as unavailable for the sake of the resource manager worker threads. */ + ma_atomic_exchange_i32(&pDataBufferNode->result, MA_UNAVAILABLE); + + job = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER_NODE); + job.order = ma_resource_manager_data_buffer_node_next_execution_order(pDataBufferNode); + job.data.resourceManager.freeDataBufferNode.pResourceManager = pResourceManager; + job.data.resourceManager.freeDataBufferNode.pDataBufferNode = pDataBufferNode; + + result = ma_resource_manager_post_job(pResourceManager, &job); + if (result != MA_SUCCESS) { + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_ERROR, "Failed to post MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER_NODE job. %s.\n", ma_result_description(result)); + return result; + } + + /* If we don't support threading, process the job queue here. */ + if (ma_resource_manager_is_threading_enabled(pResourceManager) == MA_FALSE) { + while (ma_resource_manager_data_buffer_node_result(pDataBufferNode) == MA_BUSY) { + result = ma_resource_manager_process_next_job(pResourceManager); + if (result == MA_NO_DATA_AVAILABLE || result == MA_CANCELLED) { + result = MA_SUCCESS; + break; + } + } + } else { + /* Threading is enabled. The job queue will deal with the rest of the cleanup from here. */ + } + } else { + /* The sound isn't loading so we can just free the node here. */ + ma_resource_manager_data_buffer_node_free(pResourceManager, pDataBufferNode); + } + } + + return result; +} + + + +static ma_uint32 ma_resource_manager_data_buffer_next_execution_order(ma_resource_manager_data_buffer* pDataBuffer) +{ + MA_ASSERT(pDataBuffer != NULL); + return ma_atomic_fetch_add_32(&pDataBuffer->executionCounter, 1); +} + +static ma_result ma_resource_manager_data_buffer_cb__read_pcm_frames(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_resource_manager_data_buffer_read_pcm_frames((ma_resource_manager_data_buffer*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_resource_manager_data_buffer_cb__seek_to_pcm_frame(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_resource_manager_data_buffer_seek_to_pcm_frame((ma_resource_manager_data_buffer*)pDataSource, frameIndex); +} + +static ma_result ma_resource_manager_data_buffer_cb__get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + return ma_resource_manager_data_buffer_get_data_format((ma_resource_manager_data_buffer*)pDataSource, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); +} + +static ma_result ma_resource_manager_data_buffer_cb__get_cursor_in_pcm_frames(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + return ma_resource_manager_data_buffer_get_cursor_in_pcm_frames((ma_resource_manager_data_buffer*)pDataSource, pCursor); +} + +static ma_result ma_resource_manager_data_buffer_cb__get_length_in_pcm_frames(ma_data_source* pDataSource, ma_uint64* pLength) +{ + return ma_resource_manager_data_buffer_get_length_in_pcm_frames((ma_resource_manager_data_buffer*)pDataSource, pLength); +} + +static ma_result ma_resource_manager_data_buffer_cb__set_looping(ma_data_source* pDataSource, ma_bool32 isLooping) +{ + ma_resource_manager_data_buffer* pDataBuffer = (ma_resource_manager_data_buffer*)pDataSource; + MA_ASSERT(pDataBuffer != NULL); + + ma_atomic_exchange_32(&pDataBuffer->isLooping, isLooping); + + /* The looping state needs to be set on the connector as well or else looping won't work when we read audio data. */ + ma_data_source_set_looping(ma_resource_manager_data_buffer_get_connector(pDataBuffer), isLooping); + + return MA_SUCCESS; +} + +static ma_data_source_vtable g_ma_resource_manager_data_buffer_vtable = +{ + ma_resource_manager_data_buffer_cb__read_pcm_frames, + ma_resource_manager_data_buffer_cb__seek_to_pcm_frame, + ma_resource_manager_data_buffer_cb__get_data_format, + ma_resource_manager_data_buffer_cb__get_cursor_in_pcm_frames, + ma_resource_manager_data_buffer_cb__get_length_in_pcm_frames, + ma_resource_manager_data_buffer_cb__set_looping, + 0 +}; + +static ma_result ma_resource_manager_data_buffer_init_ex_internal(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source_config* pConfig, ma_uint32 hashedName32, ma_resource_manager_data_buffer* pDataBuffer) +{ + ma_result result = MA_SUCCESS; + ma_resource_manager_data_buffer_node* pDataBufferNode; + ma_data_source_config dataSourceConfig; + ma_bool32 async; + ma_uint32 flags; + ma_resource_manager_pipeline_notifications notifications; + + if (pDataBuffer == NULL) { + if (pConfig != NULL && pConfig->pNotifications != NULL) { + ma_resource_manager_pipeline_notifications_signal_all_notifications(pConfig->pNotifications); + } + + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDataBuffer); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->pNotifications != NULL) { + notifications = *pConfig->pNotifications; /* From here on out we should be referencing `notifications` instead of `pNotifications`. Set this to NULL to catch errors at testing time. */ + } else { + MA_ZERO_OBJECT(¬ifications); + } + + /* For safety, always remove the ASYNC flag if threading is disabled on the resource manager. */ + flags = pConfig->flags; + if (ma_resource_manager_is_threading_enabled(pResourceManager) == MA_FALSE) { + flags &= ~MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC; + } + + if (pConfig->isLooping) { + flags |= MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING; + } + + async = (flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC) != 0; + + /* + Fences need to be acquired before doing anything. These must be acquired and released outside of + the node to ensure there's no holes where ma_fence_wait() could prematurely return before the + data buffer has completed initialization. + + When loading asynchronously, the node acquisition routine below will acquire the fences on this + thread and then release them on the async thread when the operation is complete. + + These fences are always released at the "done" tag at the end of this function. They'll be + acquired a second if loading asynchronously. This double acquisition system is just done to + simplify code maintenance. + */ + ma_resource_manager_pipeline_notifications_acquire_all_fences(¬ifications); + { + /* We first need to acquire a node. If ASYNC is not set, this will not return until the entire sound has been loaded. */ + result = ma_resource_manager_data_buffer_node_acquire(pResourceManager, pConfig->pFilePath, pConfig->pFilePathW, hashedName32, flags, NULL, notifications.init.pFence, notifications.done.pFence, &pDataBufferNode); + if (result != MA_SUCCESS) { + ma_resource_manager_pipeline_notifications_signal_all_notifications(¬ifications); + goto done; + } + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_resource_manager_data_buffer_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pDataBuffer->ds); + if (result != MA_SUCCESS) { + ma_resource_manager_data_buffer_node_unacquire(pResourceManager, pDataBufferNode, NULL, NULL); + ma_resource_manager_pipeline_notifications_signal_all_notifications(¬ifications); + goto done; + } + + pDataBuffer->pResourceManager = pResourceManager; + pDataBuffer->pNode = pDataBufferNode; + pDataBuffer->flags = flags; + pDataBuffer->result = MA_BUSY; /* Always default to MA_BUSY for safety. It'll be overwritten when loading completes or an error occurs. */ + + /* If we're loading asynchronously we need to post a job to the job queue to initialize the connector. */ + if (async == MA_FALSE || ma_resource_manager_data_buffer_node_result(pDataBufferNode) == MA_SUCCESS) { + /* Loading synchronously or the data has already been fully loaded. We can just initialize the connector from here without a job. */ + result = ma_resource_manager_data_buffer_init_connector(pDataBuffer, pConfig, NULL, NULL); + ma_atomic_exchange_i32(&pDataBuffer->result, result); + + ma_resource_manager_pipeline_notifications_signal_all_notifications(¬ifications); + goto done; + } else { + /* The node's data supply isn't initialized yet. The caller has requested that we load asynchronously so we need to post a job to do this. */ + ma_job job; + ma_resource_manager_inline_notification initNotification; /* Used when the WAIT_INIT flag is set. */ + + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + ma_resource_manager_inline_notification_init(pResourceManager, &initNotification); + } + + /* + The status of the data buffer needs to be set to MA_BUSY before posting the job so that the + worker thread is aware of its busy state. If the LOAD_DATA_BUFFER job sees a status other + than MA_BUSY, it'll assume an error and fall through to an early exit. + */ + ma_atomic_exchange_i32(&pDataBuffer->result, MA_BUSY); + + /* Acquire fences a second time. These will be released by the async thread. */ + ma_resource_manager_pipeline_notifications_acquire_all_fences(¬ifications); + + job = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER); + job.order = ma_resource_manager_data_buffer_next_execution_order(pDataBuffer); + job.data.resourceManager.loadDataBuffer.pDataBuffer = pDataBuffer; + job.data.resourceManager.loadDataBuffer.pInitNotification = ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) ? &initNotification : notifications.init.pNotification; + job.data.resourceManager.loadDataBuffer.pDoneNotification = notifications.done.pNotification; + job.data.resourceManager.loadDataBuffer.pInitFence = notifications.init.pFence; + job.data.resourceManager.loadDataBuffer.pDoneFence = notifications.done.pFence; + job.data.resourceManager.loadDataBuffer.rangeBegInPCMFrames = pConfig->rangeBegInPCMFrames; + job.data.resourceManager.loadDataBuffer.rangeEndInPCMFrames = pConfig->rangeEndInPCMFrames; + job.data.resourceManager.loadDataBuffer.loopPointBegInPCMFrames = pConfig->loopPointBegInPCMFrames; + job.data.resourceManager.loadDataBuffer.loopPointEndInPCMFrames = pConfig->loopPointEndInPCMFrames; + job.data.resourceManager.loadDataBuffer.isLooping = (flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING) != 0; + + /* If we need to wait for initialization to complete we can just process the job in place. */ + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + result = ma_job_process(&job); + } else { + result = ma_resource_manager_post_job(pResourceManager, &job); + } + + if (result != MA_SUCCESS) { + /* We failed to post the job. Most likely there isn't enough room in the queue's buffer. */ + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_ERROR, "Failed to post MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_BUFFER job. %s.\n", ma_result_description(result)); + ma_atomic_exchange_i32(&pDataBuffer->result, result); + + /* Release the fences after the result has been set on the data buffer. */ + ma_resource_manager_pipeline_notifications_release_all_fences(¬ifications); + } else { + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + ma_resource_manager_inline_notification_wait(&initNotification); + + if (notifications.init.pNotification != NULL) { + ma_async_notification_signal(notifications.init.pNotification); + } + + /* NOTE: Do not release the init fence here. It will have been done by the job. */ + + /* Make sure we return an error if initialization failed on the async thread. */ + result = ma_resource_manager_data_buffer_result(pDataBuffer); + if (result == MA_BUSY) { + result = MA_SUCCESS; + } + } + } + + if ((flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + ma_resource_manager_inline_notification_uninit(&initNotification); + } + } + + if (result != MA_SUCCESS) { + ma_resource_manager_data_buffer_node_unacquire(pResourceManager, pDataBufferNode, NULL, NULL); + goto done; + } + } +done: + if (result == MA_SUCCESS) { + if (pConfig->initialSeekPointInPCMFrames > 0) { + ma_resource_manager_data_buffer_seek_to_pcm_frame(pDataBuffer, pConfig->initialSeekPointInPCMFrames); + } + } + + ma_resource_manager_pipeline_notifications_release_all_fences(¬ifications); + + return result; +} + +MA_API ma_result ma_resource_manager_data_buffer_init_ex(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source_config* pConfig, ma_resource_manager_data_buffer* pDataBuffer) +{ + return ma_resource_manager_data_buffer_init_ex_internal(pResourceManager, pConfig, 0, pDataBuffer); +} + +MA_API ma_result ma_resource_manager_data_buffer_init(ma_resource_manager* pResourceManager, const char* pFilePath, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_buffer* pDataBuffer) +{ + ma_resource_manager_data_source_config config; + + config = ma_resource_manager_data_source_config_init(); + config.pFilePath = pFilePath; + config.flags = flags; + config.pNotifications = pNotifications; + + return ma_resource_manager_data_buffer_init_ex(pResourceManager, &config, pDataBuffer); +} + +MA_API ma_result ma_resource_manager_data_buffer_init_w(ma_resource_manager* pResourceManager, const wchar_t* pFilePath, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_buffer* pDataBuffer) +{ + ma_resource_manager_data_source_config config; + + config = ma_resource_manager_data_source_config_init(); + config.pFilePathW = pFilePath; + config.flags = flags; + config.pNotifications = pNotifications; + + return ma_resource_manager_data_buffer_init_ex(pResourceManager, &config, pDataBuffer); +} + +MA_API ma_result ma_resource_manager_data_buffer_init_copy(ma_resource_manager* pResourceManager, const ma_resource_manager_data_buffer* pExistingDataBuffer, ma_resource_manager_data_buffer* pDataBuffer) +{ + ma_resource_manager_data_source_config config; + + if (pExistingDataBuffer == NULL) { + return MA_INVALID_ARGS; + } + + MA_ASSERT(pExistingDataBuffer->pNode != NULL); /* <-- If you've triggered this, you've passed in an invalid existing data buffer. */ + + config = ma_resource_manager_data_source_config_init(); + config.flags = pExistingDataBuffer->flags; + + return ma_resource_manager_data_buffer_init_ex_internal(pResourceManager, &config, pExistingDataBuffer->pNode->hashedName32, pDataBuffer); +} + +static ma_result ma_resource_manager_data_buffer_uninit_internal(ma_resource_manager_data_buffer* pDataBuffer) +{ + MA_ASSERT(pDataBuffer != NULL); + + /* The connector should be uninitialized first. */ + ma_resource_manager_data_buffer_uninit_connector(pDataBuffer->pResourceManager, pDataBuffer); + + /* With the connector uninitialized we can unacquire the node. */ + ma_resource_manager_data_buffer_node_unacquire(pDataBuffer->pResourceManager, pDataBuffer->pNode, NULL, NULL); + + /* The base data source needs to be uninitialized as well. */ + ma_data_source_uninit(&pDataBuffer->ds); + + return MA_SUCCESS; +} + +MA_API ma_result ma_resource_manager_data_buffer_uninit(ma_resource_manager_data_buffer* pDataBuffer) +{ + ma_result result; + + if (pDataBuffer == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_resource_manager_data_buffer_result(pDataBuffer) == MA_SUCCESS) { + /* The data buffer can be deleted synchronously. */ + return ma_resource_manager_data_buffer_uninit_internal(pDataBuffer); + } else { + /* + The data buffer needs to be deleted asynchronously because it's still loading. With the status set to MA_UNAVAILABLE, no more pages will + be loaded and the uninitialization should happen fairly quickly. Since the caller owns the data buffer, we need to wait for this event + to get processed before returning. + */ + ma_resource_manager_inline_notification notification; + ma_job job; + + /* + We need to mark the node as unavailable so we don't try reading from it anymore, but also to + let the loading thread know that it needs to abort it's loading procedure. + */ + ma_atomic_exchange_i32(&pDataBuffer->result, MA_UNAVAILABLE); + + result = ma_resource_manager_inline_notification_init(pDataBuffer->pResourceManager, ¬ification); + if (result != MA_SUCCESS) { + return result; /* Failed to create the notification. This should rarely, if ever, happen. */ + } + + job = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER); + job.order = ma_resource_manager_data_buffer_next_execution_order(pDataBuffer); + job.data.resourceManager.freeDataBuffer.pDataBuffer = pDataBuffer; + job.data.resourceManager.freeDataBuffer.pDoneNotification = ¬ification; + job.data.resourceManager.freeDataBuffer.pDoneFence = NULL; + + result = ma_resource_manager_post_job(pDataBuffer->pResourceManager, &job); + if (result != MA_SUCCESS) { + ma_resource_manager_inline_notification_uninit(¬ification); + return result; + } + + ma_resource_manager_inline_notification_wait_and_uninit(¬ification); + } + + return result; +} + +MA_API ma_result ma_resource_manager_data_buffer_read_pcm_frames(ma_resource_manager_data_buffer* pDataBuffer, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_result result = MA_SUCCESS; + ma_uint64 framesRead = 0; + ma_bool32 isDecodedBufferBusy = MA_FALSE; + + /* Safety. */ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + /* + We cannot be using the data buffer after it's been uninitialized. If you trigger this assert it means you're trying to read from the data buffer after + it's been uninitialized or is in the process of uninitializing. + */ + MA_ASSERT(ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode) != MA_UNAVAILABLE); + + /* If the node is not initialized we need to abort with a busy code. */ + if (ma_resource_manager_data_buffer_has_connector(pDataBuffer) == MA_FALSE) { + return MA_BUSY; /* Still loading. */ + } + + /* + If we've got a seek scheduled we'll want to do that before reading. However, for paged buffers, there's + a chance that the sound hasn't yet been decoded up to the seek point will result in the seek failing. If + this happens, we need to keep the seek scheduled and return MA_BUSY. + */ + if (pDataBuffer->seekToCursorOnNextRead) { + pDataBuffer->seekToCursorOnNextRead = MA_FALSE; + + result = ma_data_source_seek_to_pcm_frame(ma_resource_manager_data_buffer_get_connector(pDataBuffer), pDataBuffer->seekTargetInPCMFrames); + if (result != MA_SUCCESS) { + if (result == MA_BAD_SEEK && ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode) == ma_resource_manager_data_supply_type_decoded_paged) { + pDataBuffer->seekToCursorOnNextRead = MA_TRUE; /* Keep the seek scheduled. We just haven't loaded enough data yet to do the seek properly. */ + return MA_BUSY; + } + + return result; + } + } + + /* + For decoded buffers (not paged) we need to check beforehand how many frames we have available. We cannot + exceed this amount. We'll read as much as we can, and then return MA_BUSY. + */ + if (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode) == ma_resource_manager_data_supply_type_decoded) { + ma_uint64 availableFrames; + + isDecodedBufferBusy = (ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode) == MA_BUSY); + + if (ma_resource_manager_data_buffer_get_available_frames(pDataBuffer, &availableFrames) == MA_SUCCESS) { + /* Don't try reading more than the available frame count if the data buffer node is still loading. */ + if (isDecodedBufferBusy) { + if (frameCount > availableFrames) { + frameCount = availableFrames; + + /* + If there's no frames available we want to set the status to MA_AT_END. The logic below + will check if the node is busy, and if so, change it to MA_BUSY. The reason we do this + is because we don't want to call `ma_data_source_read_pcm_frames()` if the frame count + is 0 because that'll result in a situation where it's possible MA_AT_END won't get + returned. + */ + if (frameCount == 0) { + result = MA_AT_END; + } + } else { + isDecodedBufferBusy = MA_FALSE; /* We have enough frames available in the buffer to avoid a MA_BUSY status. */ + } + } else { + /* + Getting here means the buffer has been fully loaded. We can just pass the frame count straight + into ma_data_source_read_pcm_frames() below and let ma_data_source handle it. + */ + } + } + } + + /* Don't attempt to read anything if we've got no frames available. */ + if (frameCount > 0) { + result = ma_data_source_read_pcm_frames(ma_resource_manager_data_buffer_get_connector(pDataBuffer), pFramesOut, frameCount, &framesRead); + } + + /* + If we returned MA_AT_END, but the node is still loading, we don't want to return that code or else the caller will interpret the sound + as at the end and terminate decoding. + */ + if (result == MA_AT_END) { + if (ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode) == MA_BUSY) { + result = MA_BUSY; + } + } + + if (isDecodedBufferBusy) { + result = MA_BUSY; + } + + if (pFramesRead != NULL) { + *pFramesRead = framesRead; + } + + if (result == MA_SUCCESS && framesRead == 0) { + result = MA_AT_END; + } + + return result; +} + +MA_API ma_result ma_resource_manager_data_buffer_seek_to_pcm_frame(ma_resource_manager_data_buffer* pDataBuffer, ma_uint64 frameIndex) +{ + ma_result result; + + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode) != MA_UNAVAILABLE); + + /* If we haven't yet got a connector we need to abort. */ + if (ma_resource_manager_data_buffer_has_connector(pDataBuffer) == MA_FALSE) { + pDataBuffer->seekTargetInPCMFrames = frameIndex; + pDataBuffer->seekToCursorOnNextRead = MA_TRUE; + return MA_BUSY; /* Still loading. */ + } + + result = ma_data_source_seek_to_pcm_frame(ma_resource_manager_data_buffer_get_connector(pDataBuffer), frameIndex); + if (result != MA_SUCCESS) { + return result; + } + + pDataBuffer->seekTargetInPCMFrames = ~(ma_uint64)0; /* <-- For identification purposes. */ + pDataBuffer->seekToCursorOnNextRead = MA_FALSE; + + return MA_SUCCESS; +} + +MA_API ma_result ma_resource_manager_data_buffer_get_data_format(ma_resource_manager_data_buffer* pDataBuffer, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode) != MA_UNAVAILABLE); + + switch (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode)) + { + case ma_resource_manager_data_supply_type_encoded: + { + return ma_data_source_get_data_format(&pDataBuffer->connector.decoder, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); + }; + + case ma_resource_manager_data_supply_type_decoded: + { + *pFormat = pDataBuffer->pNode->data.backend.decoded.format; + *pChannels = pDataBuffer->pNode->data.backend.decoded.channels; + *pSampleRate = pDataBuffer->pNode->data.backend.decoded.sampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pDataBuffer->pNode->data.backend.decoded.channels); + return MA_SUCCESS; + }; + + case ma_resource_manager_data_supply_type_decoded_paged: + { + *pFormat = pDataBuffer->pNode->data.backend.decodedPaged.data.format; + *pChannels = pDataBuffer->pNode->data.backend.decodedPaged.data.channels; + *pSampleRate = pDataBuffer->pNode->data.backend.decodedPaged.sampleRate; + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, pDataBuffer->pNode->data.backend.decoded.channels); + return MA_SUCCESS; + }; + + case ma_resource_manager_data_supply_type_unknown: + { + return MA_BUSY; /* Still loading. */ + }; + + default: + { + /* Unknown supply type. Should never hit this. */ + return MA_INVALID_ARGS; + } + } +} + +MA_API ma_result ma_resource_manager_data_buffer_get_cursor_in_pcm_frames(ma_resource_manager_data_buffer* pDataBuffer, ma_uint64* pCursor) +{ + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode) != MA_UNAVAILABLE); + + if (pDataBuffer == NULL || pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; + + switch (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode)) + { + case ma_resource_manager_data_supply_type_encoded: + { + return ma_decoder_get_cursor_in_pcm_frames(&pDataBuffer->connector.decoder, pCursor); + }; + + case ma_resource_manager_data_supply_type_decoded: + { + return ma_audio_buffer_get_cursor_in_pcm_frames(&pDataBuffer->connector.buffer, pCursor); + }; + + case ma_resource_manager_data_supply_type_decoded_paged: + { + return ma_paged_audio_buffer_get_cursor_in_pcm_frames(&pDataBuffer->connector.pagedBuffer, pCursor); + }; + + case ma_resource_manager_data_supply_type_unknown: + { + return MA_BUSY; + }; + + default: + { + return MA_INVALID_ARGS; + } + } +} + +MA_API ma_result ma_resource_manager_data_buffer_get_length_in_pcm_frames(ma_resource_manager_data_buffer* pDataBuffer, ma_uint64* pLength) +{ + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode) != MA_UNAVAILABLE); + + if (pDataBuffer == NULL || pLength == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode) == ma_resource_manager_data_supply_type_unknown) { + return MA_BUSY; /* Still loading. */ + } + + return ma_data_source_get_length_in_pcm_frames(ma_resource_manager_data_buffer_get_connector(pDataBuffer), pLength); +} + +MA_API ma_result ma_resource_manager_data_buffer_result(const ma_resource_manager_data_buffer* pDataBuffer) +{ + if (pDataBuffer == NULL) { + return MA_INVALID_ARGS; + } + + return (ma_result)ma_atomic_load_i32((ma_result*)&pDataBuffer->result); /* Need a naughty const-cast here. */ +} + +MA_API ma_result ma_resource_manager_data_buffer_set_looping(ma_resource_manager_data_buffer* pDataBuffer, ma_bool32 isLooping) +{ + return ma_data_source_set_looping(pDataBuffer, isLooping); +} + +MA_API ma_bool32 ma_resource_manager_data_buffer_is_looping(const ma_resource_manager_data_buffer* pDataBuffer) +{ + return ma_data_source_is_looping(pDataBuffer); +} + +MA_API ma_result ma_resource_manager_data_buffer_get_available_frames(ma_resource_manager_data_buffer* pDataBuffer, ma_uint64* pAvailableFrames) +{ + if (pAvailableFrames == NULL) { + return MA_INVALID_ARGS; + } + + *pAvailableFrames = 0; + + if (pDataBuffer == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode) == ma_resource_manager_data_supply_type_unknown) { + if (ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode) == MA_BUSY) { + return MA_BUSY; + } else { + return MA_INVALID_OPERATION; /* No connector. */ + } + } + + switch (ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode)) + { + case ma_resource_manager_data_supply_type_encoded: + { + return ma_decoder_get_available_frames(&pDataBuffer->connector.decoder, pAvailableFrames); + }; + + case ma_resource_manager_data_supply_type_decoded: + { + return ma_audio_buffer_get_available_frames(&pDataBuffer->connector.buffer, pAvailableFrames); + }; + + case ma_resource_manager_data_supply_type_decoded_paged: + { + ma_uint64 cursor; + ma_paged_audio_buffer_get_cursor_in_pcm_frames(&pDataBuffer->connector.pagedBuffer, &cursor); + + if (pDataBuffer->pNode->data.backend.decodedPaged.decodedFrameCount > cursor) { + *pAvailableFrames = pDataBuffer->pNode->data.backend.decodedPaged.decodedFrameCount - cursor; + } else { + *pAvailableFrames = 0; + } + + return MA_SUCCESS; + }; + + case ma_resource_manager_data_supply_type_unknown: + default: + { + /* Unknown supply type. Should never hit this. */ + return MA_INVALID_ARGS; + } + } +} + +MA_API ma_result ma_resource_manager_register_file(ma_resource_manager* pResourceManager, const char* pFilePath, ma_uint32 flags) +{ + return ma_resource_manager_data_buffer_node_acquire(pResourceManager, pFilePath, NULL, 0, flags, NULL, NULL, NULL, NULL); +} + +MA_API ma_result ma_resource_manager_register_file_w(ma_resource_manager* pResourceManager, const wchar_t* pFilePath, ma_uint32 flags) +{ + return ma_resource_manager_data_buffer_node_acquire(pResourceManager, NULL, pFilePath, 0, flags, NULL, NULL, NULL, NULL); +} + + +static ma_result ma_resource_manager_register_data(ma_resource_manager* pResourceManager, const char* pName, const wchar_t* pNameW, ma_resource_manager_data_supply* pExistingData) +{ + return ma_resource_manager_data_buffer_node_acquire(pResourceManager, pName, pNameW, 0, 0, pExistingData, NULL, NULL, NULL); +} + +static ma_result ma_resource_manager_register_decoded_data_internal(ma_resource_manager* pResourceManager, const char* pName, const wchar_t* pNameW, const void* pData, ma_uint64 frameCount, ma_format format, ma_uint32 channels, ma_uint32 sampleRate) +{ + ma_resource_manager_data_supply data; + data.type = ma_resource_manager_data_supply_type_decoded; + data.backend.decoded.pData = pData; + data.backend.decoded.totalFrameCount = frameCount; + data.backend.decoded.format = format; + data.backend.decoded.channels = channels; + data.backend.decoded.sampleRate = sampleRate; + + return ma_resource_manager_register_data(pResourceManager, pName, pNameW, &data); +} + +MA_API ma_result ma_resource_manager_register_decoded_data(ma_resource_manager* pResourceManager, const char* pName, const void* pData, ma_uint64 frameCount, ma_format format, ma_uint32 channels, ma_uint32 sampleRate) +{ + return ma_resource_manager_register_decoded_data_internal(pResourceManager, pName, NULL, pData, frameCount, format, channels, sampleRate); +} + +MA_API ma_result ma_resource_manager_register_decoded_data_w(ma_resource_manager* pResourceManager, const wchar_t* pName, const void* pData, ma_uint64 frameCount, ma_format format, ma_uint32 channels, ma_uint32 sampleRate) +{ + return ma_resource_manager_register_decoded_data_internal(pResourceManager, NULL, pName, pData, frameCount, format, channels, sampleRate); +} + + +static ma_result ma_resource_manager_register_encoded_data_internal(ma_resource_manager* pResourceManager, const char* pName, const wchar_t* pNameW, const void* pData, size_t sizeInBytes) +{ + ma_resource_manager_data_supply data; + data.type = ma_resource_manager_data_supply_type_encoded; + data.backend.encoded.pData = pData; + data.backend.encoded.sizeInBytes = sizeInBytes; + + return ma_resource_manager_register_data(pResourceManager, pName, pNameW, &data); +} + +MA_API ma_result ma_resource_manager_register_encoded_data(ma_resource_manager* pResourceManager, const char* pName, const void* pData, size_t sizeInBytes) +{ + return ma_resource_manager_register_encoded_data_internal(pResourceManager, pName, NULL, pData, sizeInBytes); +} + +MA_API ma_result ma_resource_manager_register_encoded_data_w(ma_resource_manager* pResourceManager, const wchar_t* pName, const void* pData, size_t sizeInBytes) +{ + return ma_resource_manager_register_encoded_data_internal(pResourceManager, NULL, pName, pData, sizeInBytes); +} + + +MA_API ma_result ma_resource_manager_unregister_file(ma_resource_manager* pResourceManager, const char* pFilePath) +{ + return ma_resource_manager_unregister_data(pResourceManager, pFilePath); +} + +MA_API ma_result ma_resource_manager_unregister_file_w(ma_resource_manager* pResourceManager, const wchar_t* pFilePath) +{ + return ma_resource_manager_unregister_data_w(pResourceManager, pFilePath); +} + +MA_API ma_result ma_resource_manager_unregister_data(ma_resource_manager* pResourceManager, const char* pName) +{ + return ma_resource_manager_data_buffer_node_unacquire(pResourceManager, NULL, pName, NULL); +} + +MA_API ma_result ma_resource_manager_unregister_data_w(ma_resource_manager* pResourceManager, const wchar_t* pName) +{ + return ma_resource_manager_data_buffer_node_unacquire(pResourceManager, NULL, NULL, pName); +} + + +static ma_uint32 ma_resource_manager_data_stream_next_execution_order(ma_resource_manager_data_stream* pDataStream) +{ + MA_ASSERT(pDataStream != NULL); + return ma_atomic_fetch_add_32(&pDataStream->executionCounter, 1); +} + +static ma_bool32 ma_resource_manager_data_stream_is_decoder_at_end(const ma_resource_manager_data_stream* pDataStream) +{ + MA_ASSERT(pDataStream != NULL); + return ma_atomic_load_32((ma_bool32*)&pDataStream->isDecoderAtEnd); +} + +static ma_uint32 ma_resource_manager_data_stream_seek_counter(const ma_resource_manager_data_stream* pDataStream) +{ + MA_ASSERT(pDataStream != NULL); + return ma_atomic_load_32((ma_uint32*)&pDataStream->seekCounter); +} + + +static ma_result ma_resource_manager_data_stream_cb__read_pcm_frames(ma_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + return ma_resource_manager_data_stream_read_pcm_frames((ma_resource_manager_data_stream*)pDataSource, pFramesOut, frameCount, pFramesRead); +} + +static ma_result ma_resource_manager_data_stream_cb__seek_to_pcm_frame(ma_data_source* pDataSource, ma_uint64 frameIndex) +{ + return ma_resource_manager_data_stream_seek_to_pcm_frame((ma_resource_manager_data_stream*)pDataSource, frameIndex); +} + +static ma_result ma_resource_manager_data_stream_cb__get_data_format(ma_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + return ma_resource_manager_data_stream_get_data_format((ma_resource_manager_data_stream*)pDataSource, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); +} + +static ma_result ma_resource_manager_data_stream_cb__get_cursor_in_pcm_frames(ma_data_source* pDataSource, ma_uint64* pCursor) +{ + return ma_resource_manager_data_stream_get_cursor_in_pcm_frames((ma_resource_manager_data_stream*)pDataSource, pCursor); +} + +static ma_result ma_resource_manager_data_stream_cb__get_length_in_pcm_frames(ma_data_source* pDataSource, ma_uint64* pLength) +{ + return ma_resource_manager_data_stream_get_length_in_pcm_frames((ma_resource_manager_data_stream*)pDataSource, pLength); +} + +static ma_result ma_resource_manager_data_stream_cb__set_looping(ma_data_source* pDataSource, ma_bool32 isLooping) +{ + ma_resource_manager_data_stream* pDataStream = (ma_resource_manager_data_stream*)pDataSource; + MA_ASSERT(pDataStream != NULL); + + ma_atomic_exchange_32(&pDataStream->isLooping, isLooping); + + return MA_SUCCESS; +} + +static ma_data_source_vtable g_ma_resource_manager_data_stream_vtable = +{ + ma_resource_manager_data_stream_cb__read_pcm_frames, + ma_resource_manager_data_stream_cb__seek_to_pcm_frame, + ma_resource_manager_data_stream_cb__get_data_format, + ma_resource_manager_data_stream_cb__get_cursor_in_pcm_frames, + ma_resource_manager_data_stream_cb__get_length_in_pcm_frames, + ma_resource_manager_data_stream_cb__set_looping, + 0 /*MA_DATA_SOURCE_SELF_MANAGED_RANGE_AND_LOOP_POINT*/ +}; + +static void ma_resource_manager_data_stream_set_absolute_cursor(ma_resource_manager_data_stream* pDataStream, ma_uint64 absoluteCursor) +{ + /* Loop if possible. */ + if (absoluteCursor > pDataStream->totalLengthInPCMFrames && pDataStream->totalLengthInPCMFrames > 0) { + absoluteCursor = absoluteCursor % pDataStream->totalLengthInPCMFrames; + } + + ma_atomic_exchange_64(&pDataStream->absoluteCursor, absoluteCursor); +} + +MA_API ma_result ma_resource_manager_data_stream_init_ex(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source_config* pConfig, ma_resource_manager_data_stream* pDataStream) +{ + ma_result result; + ma_data_source_config dataSourceConfig; + char* pFilePathCopy = NULL; + wchar_t* pFilePathWCopy = NULL; + ma_job job; + ma_bool32 waitBeforeReturning = MA_FALSE; + ma_resource_manager_inline_notification waitNotification; + ma_resource_manager_pipeline_notifications notifications; + ma_uint32 flags; + + if (pDataStream == NULL) { + if (pConfig != NULL && pConfig->pNotifications != NULL) { + ma_resource_manager_pipeline_notifications_signal_all_notifications(pConfig->pNotifications); + } + + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDataStream); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->pNotifications != NULL) { + notifications = *pConfig->pNotifications; /* From here on out, `notifications` should be used instead of `pNotifications`. Setting this to NULL to catch any errors at testing time. */ + } else { + MA_ZERO_OBJECT(¬ifications); + } + + dataSourceConfig = ma_data_source_config_init(); + dataSourceConfig.vtable = &g_ma_resource_manager_data_stream_vtable; + + result = ma_data_source_init(&dataSourceConfig, &pDataStream->ds); + if (result != MA_SUCCESS) { + ma_resource_manager_pipeline_notifications_signal_all_notifications(¬ifications); + return result; + } + + flags = pConfig->flags; + if (pConfig->isLooping) { + flags |= MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING; + } + + pDataStream->pResourceManager = pResourceManager; + pDataStream->flags = pConfig->flags; + pDataStream->result = MA_BUSY; + + ma_data_source_set_range_in_pcm_frames(pDataStream, pConfig->rangeBegInPCMFrames, pConfig->rangeEndInPCMFrames); + ma_data_source_set_loop_point_in_pcm_frames(pDataStream, pConfig->loopPointBegInPCMFrames, pConfig->loopPointEndInPCMFrames); + ma_data_source_set_looping(pDataStream, (flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING) != 0); + + if (pResourceManager == NULL || (pConfig->pFilePath == NULL && pConfig->pFilePathW == NULL)) { + ma_resource_manager_pipeline_notifications_signal_all_notifications(¬ifications); + return MA_INVALID_ARGS; + } + + /* We want all access to the VFS and the internal decoder to happen on the job thread just to keep things easier to manage for the VFS. */ + + /* We need a copy of the file path. We should probably make this more efficient, but for now we'll do a transient memory allocation. */ + if (pConfig->pFilePath != NULL) { + pFilePathCopy = ma_copy_string(pConfig->pFilePath, &pResourceManager->config.allocationCallbacks); + } else { + pFilePathWCopy = ma_copy_string_w(pConfig->pFilePathW, &pResourceManager->config.allocationCallbacks); + } + + if (pFilePathCopy == NULL && pFilePathWCopy == NULL) { + ma_resource_manager_pipeline_notifications_signal_all_notifications(¬ifications); + return MA_OUT_OF_MEMORY; + } + + /* + We need to check for the presence of MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC. If it's not set, we need to wait before returning. Otherwise we + can return immediately. Likewise, we'll also check for MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT and do the same. + */ + if ((pConfig->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_ASYNC) == 0 || (pConfig->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT) != 0) { + waitBeforeReturning = MA_TRUE; + ma_resource_manager_inline_notification_init(pResourceManager, &waitNotification); + } + + ma_resource_manager_pipeline_notifications_acquire_all_fences(¬ifications); + + /* Set the absolute cursor to our initial seek position so retrieval of the cursor returns a good value. */ + ma_resource_manager_data_stream_set_absolute_cursor(pDataStream, pConfig->initialSeekPointInPCMFrames); + + /* We now have everything we need to post the job. This is the last thing we need to do from here. The rest will be done by the job thread. */ + job = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_LOAD_DATA_STREAM); + job.order = ma_resource_manager_data_stream_next_execution_order(pDataStream); + job.data.resourceManager.loadDataStream.pDataStream = pDataStream; + job.data.resourceManager.loadDataStream.pFilePath = pFilePathCopy; + job.data.resourceManager.loadDataStream.pFilePathW = pFilePathWCopy; + job.data.resourceManager.loadDataStream.initialSeekPoint = pConfig->initialSeekPointInPCMFrames; + job.data.resourceManager.loadDataStream.pInitNotification = (waitBeforeReturning == MA_TRUE) ? &waitNotification : notifications.init.pNotification; + job.data.resourceManager.loadDataStream.pInitFence = notifications.init.pFence; + result = ma_resource_manager_post_job(pResourceManager, &job); + if (result != MA_SUCCESS) { + ma_resource_manager_pipeline_notifications_signal_all_notifications(¬ifications); + ma_resource_manager_pipeline_notifications_release_all_fences(¬ifications); + + if (waitBeforeReturning) { + ma_resource_manager_inline_notification_uninit(&waitNotification); + } + + ma_free(pFilePathCopy, &pResourceManager->config.allocationCallbacks); + ma_free(pFilePathWCopy, &pResourceManager->config.allocationCallbacks); + return result; + } + + /* Wait if needed. */ + if (waitBeforeReturning) { + ma_resource_manager_inline_notification_wait_and_uninit(&waitNotification); + + if (notifications.init.pNotification != NULL) { + ma_async_notification_signal(notifications.init.pNotification); + } + + /* + If there was an error during initialization make sure we return that result here. We don't want to do this + if we're not waiting because it will most likely be in a busy state. + */ + if (pDataStream->result != MA_SUCCESS) { + return pDataStream->result; + } + + /* NOTE: Do not release pInitFence here. That will be done by the job. */ + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_resource_manager_data_stream_init(ma_resource_manager* pResourceManager, const char* pFilePath, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_stream* pDataStream) +{ + ma_resource_manager_data_source_config config; + + config = ma_resource_manager_data_source_config_init(); + config.pFilePath = pFilePath; + config.flags = flags; + config.pNotifications = pNotifications; + + return ma_resource_manager_data_stream_init_ex(pResourceManager, &config, pDataStream); +} + +MA_API ma_result ma_resource_manager_data_stream_init_w(ma_resource_manager* pResourceManager, const wchar_t* pFilePath, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_stream* pDataStream) +{ + ma_resource_manager_data_source_config config; + + config = ma_resource_manager_data_source_config_init(); + config.pFilePathW = pFilePath; + config.flags = flags; + config.pNotifications = pNotifications; + + return ma_resource_manager_data_stream_init_ex(pResourceManager, &config, pDataStream); +} + +MA_API ma_result ma_resource_manager_data_stream_uninit(ma_resource_manager_data_stream* pDataStream) +{ + ma_resource_manager_inline_notification freeEvent; + ma_job job; + + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + /* The first thing to do is set the result to unavailable. This will prevent future page decoding. */ + ma_atomic_exchange_i32(&pDataStream->result, MA_UNAVAILABLE); + + /* + We need to post a job to ensure we're not in the middle or decoding or anything. Because the object is owned by the caller, we'll need + to wait for it to complete before returning which means we need an event. + */ + ma_resource_manager_inline_notification_init(pDataStream->pResourceManager, &freeEvent); + + job = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_STREAM); + job.order = ma_resource_manager_data_stream_next_execution_order(pDataStream); + job.data.resourceManager.freeDataStream.pDataStream = pDataStream; + job.data.resourceManager.freeDataStream.pDoneNotification = &freeEvent; + job.data.resourceManager.freeDataStream.pDoneFence = NULL; + ma_resource_manager_post_job(pDataStream->pResourceManager, &job); + + /* We need to wait for the job to finish processing before we return. */ + ma_resource_manager_inline_notification_wait_and_uninit(&freeEvent); + + return MA_SUCCESS; +} + + +static ma_uint32 ma_resource_manager_data_stream_get_page_size_in_frames(ma_resource_manager_data_stream* pDataStream) +{ + MA_ASSERT(pDataStream != NULL); + MA_ASSERT(pDataStream->isDecoderInitialized == MA_TRUE); + + return MA_RESOURCE_MANAGER_PAGE_SIZE_IN_MILLISECONDS * (pDataStream->decoder.outputSampleRate/1000); +} + +static void* ma_resource_manager_data_stream_get_page_data_pointer(ma_resource_manager_data_stream* pDataStream, ma_uint32 pageIndex, ma_uint32 relativeCursor) +{ + MA_ASSERT(pDataStream != NULL); + MA_ASSERT(pDataStream->isDecoderInitialized == MA_TRUE); + MA_ASSERT(pageIndex == 0 || pageIndex == 1); + + return ma_offset_ptr(pDataStream->pPageData, ((ma_resource_manager_data_stream_get_page_size_in_frames(pDataStream) * pageIndex) + relativeCursor) * ma_get_bytes_per_frame(pDataStream->decoder.outputFormat, pDataStream->decoder.outputChannels)); +} + +static void ma_resource_manager_data_stream_fill_page(ma_resource_manager_data_stream* pDataStream, ma_uint32 pageIndex) +{ + ma_result result = MA_SUCCESS; + ma_uint64 pageSizeInFrames; + ma_uint64 totalFramesReadForThisPage = 0; + void* pPageData = ma_resource_manager_data_stream_get_page_data_pointer(pDataStream, pageIndex, 0); + + pageSizeInFrames = ma_resource_manager_data_stream_get_page_size_in_frames(pDataStream); + + /* The decoder needs to inherit the stream's looping and range state. */ + { + ma_uint64 rangeBeg; + ma_uint64 rangeEnd; + ma_uint64 loopPointBeg; + ma_uint64 loopPointEnd; + + ma_data_source_set_looping(&pDataStream->decoder, ma_resource_manager_data_stream_is_looping(pDataStream)); + + ma_data_source_get_range_in_pcm_frames(pDataStream, &rangeBeg, &rangeEnd); + ma_data_source_set_range_in_pcm_frames(&pDataStream->decoder, rangeBeg, rangeEnd); + + ma_data_source_get_loop_point_in_pcm_frames(pDataStream, &loopPointBeg, &loopPointEnd); + ma_data_source_set_loop_point_in_pcm_frames(&pDataStream->decoder, loopPointBeg, loopPointEnd); + } + + /* Just read straight from the decoder. It will deal with ranges and looping for us. */ + result = ma_data_source_read_pcm_frames(&pDataStream->decoder, pPageData, pageSizeInFrames, &totalFramesReadForThisPage); + if (result == MA_AT_END || totalFramesReadForThisPage < pageSizeInFrames) { + ma_atomic_exchange_32(&pDataStream->isDecoderAtEnd, MA_TRUE); + } + + ma_atomic_exchange_32(&pDataStream->pageFrameCount[pageIndex], (ma_uint32)totalFramesReadForThisPage); + ma_atomic_exchange_32(&pDataStream->isPageValid[pageIndex], MA_TRUE); +} + +static void ma_resource_manager_data_stream_fill_pages(ma_resource_manager_data_stream* pDataStream) +{ + ma_uint32 iPage; + + MA_ASSERT(pDataStream != NULL); + + for (iPage = 0; iPage < 2; iPage += 1) { + ma_resource_manager_data_stream_fill_page(pDataStream, iPage); + } +} + + +static ma_result ma_resource_manager_data_stream_map(ma_resource_manager_data_stream* pDataStream, void** ppFramesOut, ma_uint64* pFrameCount) +{ + ma_uint64 framesAvailable; + ma_uint64 frameCount = 0; + + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_stream_result(pDataStream) != MA_UNAVAILABLE); + + if (pFrameCount != NULL) { + frameCount = *pFrameCount; + *pFrameCount = 0; + } + if (ppFramesOut != NULL) { + *ppFramesOut = NULL; + } + + if (pDataStream == NULL || ppFramesOut == NULL || pFrameCount == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_resource_manager_data_stream_result(pDataStream) != MA_SUCCESS) { + return MA_INVALID_OPERATION; + } + + /* Don't attempt to read while we're in the middle of seeking. Tell the caller that we're busy. */ + if (ma_resource_manager_data_stream_seek_counter(pDataStream) > 0) { + return MA_BUSY; + } + + /* If the page we're on is invalid it means we've caught up to the job thread. */ + if (ma_atomic_load_32(&pDataStream->isPageValid[pDataStream->currentPageIndex]) == MA_FALSE) { + framesAvailable = 0; + } else { + /* + The page we're on is valid so we must have some frames available. We need to make sure that we don't overflow into the next page, even if it's valid. The reason is + that the unmap process will only post an update for one page at a time. Keeping mapping tied to page boundaries makes this simpler. + */ + ma_uint32 currentPageFrameCount = ma_atomic_load_32(&pDataStream->pageFrameCount[pDataStream->currentPageIndex]); + MA_ASSERT(currentPageFrameCount >= pDataStream->relativeCursor); + + framesAvailable = currentPageFrameCount - pDataStream->relativeCursor; + } + + /* If there's no frames available and the result is set to MA_AT_END we need to return MA_AT_END. */ + if (framesAvailable == 0) { + if (ma_resource_manager_data_stream_is_decoder_at_end(pDataStream)) { + return MA_AT_END; + } else { + return MA_BUSY; /* There are no frames available, but we're not marked as EOF so we might have caught up to the job thread. Need to return MA_BUSY and wait for more data. */ + } + } + + MA_ASSERT(framesAvailable > 0); + + if (frameCount > framesAvailable) { + frameCount = framesAvailable; + } + + *ppFramesOut = ma_resource_manager_data_stream_get_page_data_pointer(pDataStream, pDataStream->currentPageIndex, pDataStream->relativeCursor); + *pFrameCount = frameCount; + + return MA_SUCCESS; +} + +static ma_result ma_resource_manager_data_stream_unmap(ma_resource_manager_data_stream* pDataStream, ma_uint64 frameCount) +{ + ma_uint32 newRelativeCursor; + ma_uint32 pageSizeInFrames; + ma_job job; + + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_stream_result(pDataStream) != MA_UNAVAILABLE); + + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_resource_manager_data_stream_result(pDataStream) != MA_SUCCESS) { + return MA_INVALID_OPERATION; + } + + /* The frame count should always fit inside a 32-bit integer. */ + if (frameCount > 0xFFFFFFFF) { + return MA_INVALID_ARGS; + } + + pageSizeInFrames = ma_resource_manager_data_stream_get_page_size_in_frames(pDataStream); + + /* The absolute cursor needs to be updated for ma_resource_manager_data_stream_get_cursor_in_pcm_frames(). */ + ma_resource_manager_data_stream_set_absolute_cursor(pDataStream, ma_atomic_load_64(&pDataStream->absoluteCursor) + frameCount); + + /* Here is where we need to check if we need to load a new page, and if so, post a job to load it. */ + newRelativeCursor = pDataStream->relativeCursor + (ma_uint32)frameCount; + + /* If the new cursor has flowed over to the next page we need to mark the old one as invalid and post an event for it. */ + if (newRelativeCursor >= pageSizeInFrames) { + newRelativeCursor -= pageSizeInFrames; + + /* Here is where we post the job start decoding. */ + job = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_STREAM); + job.order = ma_resource_manager_data_stream_next_execution_order(pDataStream); + job.data.resourceManager.pageDataStream.pDataStream = pDataStream; + job.data.resourceManager.pageDataStream.pageIndex = pDataStream->currentPageIndex; + + /* The page needs to be marked as invalid so that the public API doesn't try reading from it. */ + ma_atomic_exchange_32(&pDataStream->isPageValid[pDataStream->currentPageIndex], MA_FALSE); + + /* Before posting the job we need to make sure we set some state. */ + pDataStream->relativeCursor = newRelativeCursor; + pDataStream->currentPageIndex = (pDataStream->currentPageIndex + 1) & 0x01; + return ma_resource_manager_post_job(pDataStream->pResourceManager, &job); + } else { + /* We haven't moved into a new page so we can just move the cursor forward. */ + pDataStream->relativeCursor = newRelativeCursor; + return MA_SUCCESS; + } +} + + +MA_API ma_result ma_resource_manager_data_stream_read_pcm_frames(ma_resource_manager_data_stream* pDataStream, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_result result = MA_SUCCESS; + ma_uint64 totalFramesProcessed; + ma_format format; + ma_uint32 channels; + + /* Safety. */ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (frameCount == 0) { + return MA_INVALID_ARGS; + } + + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_stream_result(pDataStream) != MA_UNAVAILABLE); + + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_resource_manager_data_stream_result(pDataStream) != MA_SUCCESS) { + return MA_INVALID_OPERATION; + } + + /* Don't attempt to read while we're in the middle of seeking. Tell the caller that we're busy. */ + if (ma_resource_manager_data_stream_seek_counter(pDataStream) > 0) { + return MA_BUSY; + } + + ma_resource_manager_data_stream_get_data_format(pDataStream, &format, &channels, NULL, NULL, 0); + + /* Reading is implemented in terms of map/unmap. We need to run this in a loop because mapping is clamped against page boundaries. */ + totalFramesProcessed = 0; + while (totalFramesProcessed < frameCount) { + void* pMappedFrames; + ma_uint64 mappedFrameCount; + + mappedFrameCount = frameCount - totalFramesProcessed; + result = ma_resource_manager_data_stream_map(pDataStream, &pMappedFrames, &mappedFrameCount); + if (result != MA_SUCCESS) { + break; + } + + /* Copy the mapped data to the output buffer if we have one. It's allowed for pFramesOut to be NULL in which case a relative forward seek is performed. */ + if (pFramesOut != NULL) { + ma_copy_pcm_frames(ma_offset_pcm_frames_ptr(pFramesOut, totalFramesProcessed, format, channels), pMappedFrames, mappedFrameCount, format, channels); + } + + totalFramesProcessed += mappedFrameCount; + + result = ma_resource_manager_data_stream_unmap(pDataStream, mappedFrameCount); + if (result != MA_SUCCESS) { + break; /* This is really bad - will only get an error here if we failed to post a job to the queue for loading the next page. */ + } + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesProcessed; + } + + if (result == MA_SUCCESS && totalFramesProcessed == 0) { + result = MA_AT_END; + } + + return result; +} + +MA_API ma_result ma_resource_manager_data_stream_seek_to_pcm_frame(ma_resource_manager_data_stream* pDataStream, ma_uint64 frameIndex) +{ + ma_job job; + ma_result streamResult; + + streamResult = ma_resource_manager_data_stream_result(pDataStream); + + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(streamResult != MA_UNAVAILABLE); + + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + if (streamResult != MA_SUCCESS && streamResult != MA_BUSY) { + return MA_INVALID_OPERATION; + } + + /* If we're not already seeking and we're sitting on the same frame, just make this a no-op. */ + if (ma_atomic_load_32(&pDataStream->seekCounter) == 0) { + if (ma_atomic_load_64(&pDataStream->absoluteCursor) == frameIndex) { + return MA_SUCCESS; + } + } + + + /* Increment the seek counter first to indicate to read_paged_pcm_frames() and map_paged_pcm_frames() that we are in the middle of a seek and MA_BUSY should be returned. */ + ma_atomic_fetch_add_32(&pDataStream->seekCounter, 1); + + /* Update the absolute cursor so that ma_resource_manager_data_stream_get_cursor_in_pcm_frames() returns the new position. */ + ma_resource_manager_data_stream_set_absolute_cursor(pDataStream, frameIndex); + + /* + We need to clear our currently loaded pages so that the stream starts playback from the new seek point as soon as possible. These are for the purpose of the public + API and will be ignored by the seek job. The seek job will operate on the assumption that both pages have been marked as invalid and the cursor is at the start of + the first page. + */ + pDataStream->relativeCursor = 0; + pDataStream->currentPageIndex = 0; + ma_atomic_exchange_32(&pDataStream->isPageValid[0], MA_FALSE); + ma_atomic_exchange_32(&pDataStream->isPageValid[1], MA_FALSE); + + /* Make sure the data stream is not marked as at the end or else if we seek in response to hitting the end, we won't be able to read any more data. */ + ma_atomic_exchange_32(&pDataStream->isDecoderAtEnd, MA_FALSE); + + /* + The public API is not allowed to touch the internal decoder so we need to use a job to perform the seek. When seeking, the job thread will assume both pages + are invalid and any content contained within them will be discarded and replaced with newly decoded data. + */ + job = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_SEEK_DATA_STREAM); + job.order = ma_resource_manager_data_stream_next_execution_order(pDataStream); + job.data.resourceManager.seekDataStream.pDataStream = pDataStream; + job.data.resourceManager.seekDataStream.frameIndex = frameIndex; + return ma_resource_manager_post_job(pDataStream->pResourceManager, &job); +} + +MA_API ma_result ma_resource_manager_data_stream_get_data_format(ma_resource_manager_data_stream* pDataStream, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_stream_result(pDataStream) != MA_UNAVAILABLE); + + if (pFormat != NULL) { + *pFormat = ma_format_unknown; + } + + if (pChannels != NULL) { + *pChannels = 0; + } + + if (pSampleRate != NULL) { + *pSampleRate = 0; + } + + if (pChannelMap != NULL) { + MA_ZERO_MEMORY(pChannelMap, sizeof(*pChannelMap) * channelMapCap); + } + + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + if (ma_resource_manager_data_stream_result(pDataStream) != MA_SUCCESS) { + return MA_INVALID_OPERATION; + } + + /* + We're being a little bit naughty here and accessing the internal decoder from the public API. The output data format is constant, and we've defined this function + such that the application is responsible for ensuring it's not called while uninitializing so it should be safe. + */ + return ma_data_source_get_data_format(&pDataStream->decoder, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); +} + +MA_API ma_result ma_resource_manager_data_stream_get_cursor_in_pcm_frames(ma_resource_manager_data_stream* pDataStream, ma_uint64* pCursor) +{ + ma_result result; + + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + + *pCursor = 0; + + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(ma_resource_manager_data_stream_result(pDataStream) != MA_UNAVAILABLE); + + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + /* + If the stream is in an erroneous state we need to return an invalid operation. We can allow + this to be called when the data stream is in a busy state because the caller may have asked + for an initial seek position and it's convenient to return that as the cursor position. + */ + result = ma_resource_manager_data_stream_result(pDataStream); + if (result != MA_SUCCESS && result != MA_BUSY) { + return MA_INVALID_OPERATION; + } + + *pCursor = ma_atomic_load_64(&pDataStream->absoluteCursor); + + return MA_SUCCESS; +} + +MA_API ma_result ma_resource_manager_data_stream_get_length_in_pcm_frames(ma_resource_manager_data_stream* pDataStream, ma_uint64* pLength) +{ + ma_result streamResult; + + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + + *pLength = 0; + + streamResult = ma_resource_manager_data_stream_result(pDataStream); + + /* We cannot be using the data source after it's been uninitialized. */ + MA_ASSERT(streamResult != MA_UNAVAILABLE); + + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + if (streamResult != MA_SUCCESS) { + return streamResult; + } + + /* + We most definitely do not want to be calling ma_decoder_get_length_in_pcm_frames() directly. Instead we want to use a cached value that we + calculated when we initialized it on the job thread. + */ + *pLength = pDataStream->totalLengthInPCMFrames; + if (*pLength == 0) { + return MA_NOT_IMPLEMENTED; /* Some decoders may not have a known length. */ + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_resource_manager_data_stream_result(const ma_resource_manager_data_stream* pDataStream) +{ + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + return (ma_result)ma_atomic_load_i32(&pDataStream->result); +} + +MA_API ma_result ma_resource_manager_data_stream_set_looping(ma_resource_manager_data_stream* pDataStream, ma_bool32 isLooping) +{ + return ma_data_source_set_looping(pDataStream, isLooping); +} + +MA_API ma_bool32 ma_resource_manager_data_stream_is_looping(const ma_resource_manager_data_stream* pDataStream) +{ + if (pDataStream == NULL) { + return MA_FALSE; + } + + return ma_atomic_load_32((ma_bool32*)&pDataStream->isLooping); /* Naughty const-cast. Value won't change from here in practice (maybe from another thread). */ +} + +MA_API ma_result ma_resource_manager_data_stream_get_available_frames(ma_resource_manager_data_stream* pDataStream, ma_uint64* pAvailableFrames) +{ + ma_uint32 pageIndex0; + ma_uint32 pageIndex1; + ma_uint32 relativeCursor; + ma_uint64 availableFrames; + + if (pAvailableFrames == NULL) { + return MA_INVALID_ARGS; + } + + *pAvailableFrames = 0; + + if (pDataStream == NULL) { + return MA_INVALID_ARGS; + } + + pageIndex0 = pDataStream->currentPageIndex; + pageIndex1 = (pDataStream->currentPageIndex + 1) & 0x01; + relativeCursor = pDataStream->relativeCursor; + + availableFrames = 0; + if (ma_atomic_load_32(&pDataStream->isPageValid[pageIndex0])) { + availableFrames += ma_atomic_load_32(&pDataStream->pageFrameCount[pageIndex0]) - relativeCursor; + if (ma_atomic_load_32(&pDataStream->isPageValid[pageIndex1])) { + availableFrames += ma_atomic_load_32(&pDataStream->pageFrameCount[pageIndex1]); + } + } + + *pAvailableFrames = availableFrames; + return MA_SUCCESS; +} + + +static ma_result ma_resource_manager_data_source_preinit(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source_config* pConfig, ma_resource_manager_data_source* pDataSource) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDataSource); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pResourceManager == NULL) { + return MA_INVALID_ARGS; + } + + pDataSource->flags = pConfig->flags; + if (pConfig->isLooping) { + pDataSource->flags |= MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_resource_manager_data_source_init_ex(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source_config* pConfig, ma_resource_manager_data_source* pDataSource) +{ + ma_result result; + + result = ma_resource_manager_data_source_preinit(pResourceManager, pConfig, pDataSource); + if (result != MA_SUCCESS) { + return result; + } + + /* The data source itself is just a data stream or a data buffer. */ + if ((pConfig->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_init_ex(pResourceManager, pConfig, &pDataSource->backend.stream); + } else { + return ma_resource_manager_data_buffer_init_ex(pResourceManager, pConfig, &pDataSource->backend.buffer); + } +} + +MA_API ma_result ma_resource_manager_data_source_init(ma_resource_manager* pResourceManager, const char* pName, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_source* pDataSource) +{ + ma_resource_manager_data_source_config config; + + config = ma_resource_manager_data_source_config_init(); + config.pFilePath = pName; + config.flags = flags; + config.pNotifications = pNotifications; + + return ma_resource_manager_data_source_init_ex(pResourceManager, &config, pDataSource); +} + +MA_API ma_result ma_resource_manager_data_source_init_w(ma_resource_manager* pResourceManager, const wchar_t* pName, ma_uint32 flags, const ma_resource_manager_pipeline_notifications* pNotifications, ma_resource_manager_data_source* pDataSource) +{ + ma_resource_manager_data_source_config config; + + config = ma_resource_manager_data_source_config_init(); + config.pFilePathW = pName; + config.flags = flags; + config.pNotifications = pNotifications; + + return ma_resource_manager_data_source_init_ex(pResourceManager, &config, pDataSource); +} + +MA_API ma_result ma_resource_manager_data_source_init_copy(ma_resource_manager* pResourceManager, const ma_resource_manager_data_source* pExistingDataSource, ma_resource_manager_data_source* pDataSource) +{ + ma_result result; + ma_resource_manager_data_source_config config; + + if (pExistingDataSource == NULL) { + return MA_INVALID_ARGS; + } + + config = ma_resource_manager_data_source_config_init(); + config.flags = pExistingDataSource->flags; + + result = ma_resource_manager_data_source_preinit(pResourceManager, &config, pDataSource); + if (result != MA_SUCCESS) { + return result; + } + + /* Copying can only be done from data buffers. Streams cannot be copied. */ + if ((pExistingDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return MA_INVALID_OPERATION; + } + + return ma_resource_manager_data_buffer_init_copy(pResourceManager, &pExistingDataSource->backend.buffer, &pDataSource->backend.buffer); +} + +MA_API ma_result ma_resource_manager_data_source_uninit(ma_resource_manager_data_source* pDataSource) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + /* All we need to is uninitialize the underlying data buffer or data stream. */ + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_uninit(&pDataSource->backend.stream); + } else { + return ma_resource_manager_data_buffer_uninit(&pDataSource->backend.buffer); + } +} + +MA_API ma_result ma_resource_manager_data_source_read_pcm_frames(ma_resource_manager_data_source* pDataSource, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + /* Safety. */ + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_read_pcm_frames(&pDataSource->backend.stream, pFramesOut, frameCount, pFramesRead); + } else { + return ma_resource_manager_data_buffer_read_pcm_frames(&pDataSource->backend.buffer, pFramesOut, frameCount, pFramesRead); + } +} + +MA_API ma_result ma_resource_manager_data_source_seek_to_pcm_frame(ma_resource_manager_data_source* pDataSource, ma_uint64 frameIndex) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_seek_to_pcm_frame(&pDataSource->backend.stream, frameIndex); + } else { + return ma_resource_manager_data_buffer_seek_to_pcm_frame(&pDataSource->backend.buffer, frameIndex); + } +} + +MA_API ma_result ma_resource_manager_data_source_map(ma_resource_manager_data_source* pDataSource, void** ppFramesOut, ma_uint64* pFrameCount) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_map(&pDataSource->backend.stream, ppFramesOut, pFrameCount); + } else { + return MA_NOT_IMPLEMENTED; /* Mapping not supported with data buffers. */ + } +} + +MA_API ma_result ma_resource_manager_data_source_unmap(ma_resource_manager_data_source* pDataSource, ma_uint64 frameCount) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_unmap(&pDataSource->backend.stream, frameCount); + } else { + return MA_NOT_IMPLEMENTED; /* Mapping not supported with data buffers. */ + } +} + +MA_API ma_result ma_resource_manager_data_source_get_data_format(ma_resource_manager_data_source* pDataSource, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_get_data_format(&pDataSource->backend.stream, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); + } else { + return ma_resource_manager_data_buffer_get_data_format(&pDataSource->backend.buffer, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); + } +} + +MA_API ma_result ma_resource_manager_data_source_get_cursor_in_pcm_frames(ma_resource_manager_data_source* pDataSource, ma_uint64* pCursor) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_get_cursor_in_pcm_frames(&pDataSource->backend.stream, pCursor); + } else { + return ma_resource_manager_data_buffer_get_cursor_in_pcm_frames(&pDataSource->backend.buffer, pCursor); + } +} + +MA_API ma_result ma_resource_manager_data_source_get_length_in_pcm_frames(ma_resource_manager_data_source* pDataSource, ma_uint64* pLength) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_get_length_in_pcm_frames(&pDataSource->backend.stream, pLength); + } else { + return ma_resource_manager_data_buffer_get_length_in_pcm_frames(&pDataSource->backend.buffer, pLength); + } +} + +MA_API ma_result ma_resource_manager_data_source_result(const ma_resource_manager_data_source* pDataSource) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_result(&pDataSource->backend.stream); + } else { + return ma_resource_manager_data_buffer_result(&pDataSource->backend.buffer); + } +} + +MA_API ma_result ma_resource_manager_data_source_set_looping(ma_resource_manager_data_source* pDataSource, ma_bool32 isLooping) +{ + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_set_looping(&pDataSource->backend.stream, isLooping); + } else { + return ma_resource_manager_data_buffer_set_looping(&pDataSource->backend.buffer, isLooping); + } +} + +MA_API ma_bool32 ma_resource_manager_data_source_is_looping(const ma_resource_manager_data_source* pDataSource) +{ + if (pDataSource == NULL) { + return MA_FALSE; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_is_looping(&pDataSource->backend.stream); + } else { + return ma_resource_manager_data_buffer_is_looping(&pDataSource->backend.buffer); + } +} + +MA_API ma_result ma_resource_manager_data_source_get_available_frames(ma_resource_manager_data_source* pDataSource, ma_uint64* pAvailableFrames) +{ + if (pAvailableFrames == NULL) { + return MA_INVALID_ARGS; + } + + *pAvailableFrames = 0; + + if (pDataSource == NULL) { + return MA_INVALID_ARGS; + } + + if ((pDataSource->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_STREAM) != 0) { + return ma_resource_manager_data_stream_get_available_frames(&pDataSource->backend.stream, pAvailableFrames); + } else { + return ma_resource_manager_data_buffer_get_available_frames(&pDataSource->backend.buffer, pAvailableFrames); + } +} + + +MA_API ma_result ma_resource_manager_post_job(ma_resource_manager* pResourceManager, const ma_job* pJob) +{ + if (pResourceManager == NULL) { + return MA_INVALID_ARGS; + } + + return ma_job_queue_post(&pResourceManager->jobQueue, pJob); +} + +MA_API ma_result ma_resource_manager_post_job_quit(ma_resource_manager* pResourceManager) +{ + ma_job job = ma_job_init(MA_JOB_TYPE_QUIT); + return ma_resource_manager_post_job(pResourceManager, &job); +} + +MA_API ma_result ma_resource_manager_next_job(ma_resource_manager* pResourceManager, ma_job* pJob) +{ + if (pResourceManager == NULL) { + return MA_INVALID_ARGS; + } + + return ma_job_queue_next(&pResourceManager->jobQueue, pJob); +} + + +static ma_result ma_job_process__resource_manager__load_data_buffer_node(ma_job* pJob) +{ + ma_result result = MA_SUCCESS; + ma_resource_manager* pResourceManager; + ma_resource_manager_data_buffer_node* pDataBufferNode; + + MA_ASSERT(pJob != NULL); + + pResourceManager = (ma_resource_manager*)pJob->data.resourceManager.loadDataBufferNode.pResourceManager; + MA_ASSERT(pResourceManager != NULL); + + pDataBufferNode = (ma_resource_manager_data_buffer_node*)pJob->data.resourceManager.loadDataBufferNode.pDataBufferNode; + MA_ASSERT(pDataBufferNode != NULL); + MA_ASSERT(pDataBufferNode->isDataOwnedByResourceManager == MA_TRUE); /* The data should always be owned by the resource manager. */ + + /* The data buffer is not getting deleted, but we may be getting executed out of order. If so, we need to push the job back onto the queue and return. */ + if (pJob->order != ma_atomic_load_32(&pDataBufferNode->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Attempting to execute out of order. Probably interleaved with a MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER job. */ + } + + /* First thing we need to do is check whether or not the data buffer is getting deleted. If so we just abort. */ + if (ma_resource_manager_data_buffer_node_result(pDataBufferNode) != MA_BUSY) { + result = ma_resource_manager_data_buffer_node_result(pDataBufferNode); /* The data buffer may be getting deleted before it's even been loaded. */ + goto done; + } + + /* + We're ready to start loading. Essentially what we're doing here is initializing the data supply + of the node. Once this is complete, data buffers can have their connectors initialized which + will allow then to have audio data read from them. + + Note that when the data supply type has been moved away from "unknown", that is when other threads + will determine that the node is available for data delivery and the data buffer connectors can be + initialized. Therefore, it's important that it is set after the data supply has been initialized. + */ + if ((pJob->data.resourceManager.loadDataBufferNode.flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_DECODE) != 0) { + /* + Decoding. This is the complex case because we're not going to be doing the entire decoding + process here. Instead it's going to be split of multiple jobs and loaded in pages. The + reason for this is to evenly distribute decoding time across multiple sounds, rather than + having one huge sound hog all the available processing resources. + + The first thing we do is initialize a decoder. This is allocated on the heap and is passed + around to the paging jobs. When the last paging job has completed it's processing, it'll + free the decoder for us. + + This job does not do any actual decoding. It instead just posts a PAGE_DATA_BUFFER_NODE job + which is where the actual decoding work will be done. However, once this job is complete, + the node will be in a state where data buffer connectors can be initialized. + */ + ma_decoder* pDecoder; /* <-- Free'd on the last page decode. */ + ma_job pageDataBufferNodeJob; + + /* Allocate the decoder by initializing a decoded data supply. */ + result = ma_resource_manager_data_buffer_node_init_supply_decoded(pResourceManager, pDataBufferNode, pJob->data.resourceManager.loadDataBufferNode.pFilePath, pJob->data.resourceManager.loadDataBufferNode.pFilePathW, pJob->data.resourceManager.loadDataBufferNode.flags, &pDecoder); + + /* + Don't ever propagate an MA_BUSY result code or else the resource manager will think the + node is just busy decoding rather than in an error state. This should never happen, but + including this logic for safety just in case. + */ + if (result == MA_BUSY) { + result = MA_ERROR; + } + + if (result != MA_SUCCESS) { + if (pJob->data.resourceManager.loadDataBufferNode.pFilePath != NULL) { + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_WARNING, "Failed to initialize data supply for \"%s\". %s.\n", pJob->data.resourceManager.loadDataBufferNode.pFilePath, ma_result_description(result)); + } else { + #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || defined(_MSC_VER) + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_WARNING, "Failed to initialize data supply for \"%ls\", %s.\n", pJob->data.resourceManager.loadDataBufferNode.pFilePathW, ma_result_description(result)); + #endif + } + + goto done; + } + + /* + At this point the node's data supply is initialized and other threads can start initializing + their data buffer connectors. However, no data will actually be available until we start to + actually decode it. To do this, we need to post a paging job which is where the decoding + work is done. + + Note that if an error occurred at an earlier point, this section will have been skipped. + */ + pageDataBufferNodeJob = ma_job_init(MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_BUFFER_NODE); + pageDataBufferNodeJob.order = ma_resource_manager_data_buffer_node_next_execution_order(pDataBufferNode); + pageDataBufferNodeJob.data.resourceManager.pageDataBufferNode.pResourceManager = pResourceManager; + pageDataBufferNodeJob.data.resourceManager.pageDataBufferNode.pDataBufferNode = pDataBufferNode; + pageDataBufferNodeJob.data.resourceManager.pageDataBufferNode.pDecoder = pDecoder; + pageDataBufferNodeJob.data.resourceManager.pageDataBufferNode.pDoneNotification = pJob->data.resourceManager.loadDataBufferNode.pDoneNotification; + pageDataBufferNodeJob.data.resourceManager.pageDataBufferNode.pDoneFence = pJob->data.resourceManager.loadDataBufferNode.pDoneFence; + + /* The job has been set up so it can now be posted. */ + result = ma_resource_manager_post_job(pResourceManager, &pageDataBufferNodeJob); + + /* + When we get here, we want to make sure the result code is set to MA_BUSY. The reason for + this is that the result will be copied over to the node's internal result variable. In + this case, since the decoding is still in-progress, we need to make sure the result code + is set to MA_BUSY. + */ + if (result != MA_SUCCESS) { + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_ERROR, "Failed to post MA_JOB_TYPE_RESOURCE_MANAGER_PAGE_DATA_BUFFER_NODE job. %s\n", ma_result_description(result)); + ma_decoder_uninit(pDecoder); + ma_free(pDecoder, &pResourceManager->config.allocationCallbacks); + } else { + result = MA_BUSY; + } + } else { + /* No decoding. This is the simple case. We need only read the file content into memory and we're done. */ + result = ma_resource_manager_data_buffer_node_init_supply_encoded(pResourceManager, pDataBufferNode, pJob->data.resourceManager.loadDataBufferNode.pFilePath, pJob->data.resourceManager.loadDataBufferNode.pFilePathW); + } + + +done: + /* File paths are no longer needed. */ + ma_free(pJob->data.resourceManager.loadDataBufferNode.pFilePath, &pResourceManager->config.allocationCallbacks); + ma_free(pJob->data.resourceManager.loadDataBufferNode.pFilePathW, &pResourceManager->config.allocationCallbacks); + + /* + We need to set the result to at the very end to ensure no other threads try reading the data before we've fully initialized the object. Other threads + are going to be inspecting this variable to determine whether or not they're ready to read data. We can only change the result if it's set to MA_BUSY + because otherwise we may be changing away from an error code which would be bad. An example is if the application creates a data buffer, but then + immediately deletes it before we've got to this point. In this case, pDataBuffer->result will be MA_UNAVAILABLE, and setting it to MA_SUCCESS or any + other error code would cause the buffer to look like it's in a state that it's not. + */ + ma_atomic_compare_and_swap_i32(&pDataBufferNode->result, MA_BUSY, result); + + /* At this point initialization is complete and we can signal the notification if any. */ + if (pJob->data.resourceManager.loadDataBufferNode.pInitNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.loadDataBufferNode.pInitNotification); + } + if (pJob->data.resourceManager.loadDataBufferNode.pInitFence != NULL) { + ma_fence_release(pJob->data.resourceManager.loadDataBufferNode.pInitFence); + } + + /* If we have a success result it means we've fully loaded the buffer. This will happen in the non-decoding case. */ + if (result != MA_BUSY) { + if (pJob->data.resourceManager.loadDataBufferNode.pDoneNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.loadDataBufferNode.pDoneNotification); + } + if (pJob->data.resourceManager.loadDataBufferNode.pDoneFence != NULL) { + ma_fence_release(pJob->data.resourceManager.loadDataBufferNode.pDoneFence); + } + } + + /* Increment the node's execution pointer so that the next jobs can be processed. This is how we keep decoding of pages in-order. */ + ma_atomic_fetch_add_32(&pDataBufferNode->executionPointer, 1); + + /* A busy result should be considered successful from the point of view of the job system. */ + if (result == MA_BUSY) { + result = MA_SUCCESS; + } + + return result; +} + +static ma_result ma_job_process__resource_manager__free_data_buffer_node(ma_job* pJob) +{ + ma_resource_manager* pResourceManager; + ma_resource_manager_data_buffer_node* pDataBufferNode; + + MA_ASSERT(pJob != NULL); + + pResourceManager = (ma_resource_manager*)pJob->data.resourceManager.freeDataBufferNode.pResourceManager; + MA_ASSERT(pResourceManager != NULL); + + pDataBufferNode = (ma_resource_manager_data_buffer_node*)pJob->data.resourceManager.freeDataBufferNode.pDataBufferNode; + MA_ASSERT(pDataBufferNode != NULL); + + if (pJob->order != ma_atomic_load_32(&pDataBufferNode->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Out of order. */ + } + + ma_resource_manager_data_buffer_node_free(pResourceManager, pDataBufferNode); + + /* The event needs to be signalled last. */ + if (pJob->data.resourceManager.freeDataBufferNode.pDoneNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.freeDataBufferNode.pDoneNotification); + } + + if (pJob->data.resourceManager.freeDataBufferNode.pDoneFence != NULL) { + ma_fence_release(pJob->data.resourceManager.freeDataBufferNode.pDoneFence); + } + + ma_atomic_fetch_add_32(&pDataBufferNode->executionPointer, 1); + return MA_SUCCESS; +} + +static ma_result ma_job_process__resource_manager__page_data_buffer_node(ma_job* pJob) +{ + ma_result result = MA_SUCCESS; + ma_resource_manager* pResourceManager; + ma_resource_manager_data_buffer_node* pDataBufferNode; + + MA_ASSERT(pJob != NULL); + + pResourceManager = (ma_resource_manager*)pJob->data.resourceManager.pageDataBufferNode.pResourceManager; + MA_ASSERT(pResourceManager != NULL); + + pDataBufferNode = (ma_resource_manager_data_buffer_node*)pJob->data.resourceManager.pageDataBufferNode.pDataBufferNode; + MA_ASSERT(pDataBufferNode != NULL); + + if (pJob->order != ma_atomic_load_32(&pDataBufferNode->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Out of order. */ + } + + /* Don't do any more decoding if the data buffer has started the uninitialization process. */ + result = ma_resource_manager_data_buffer_node_result(pDataBufferNode); + if (result != MA_BUSY) { + goto done; + } + + /* We're ready to decode the next page. */ + result = ma_resource_manager_data_buffer_node_decode_next_page(pResourceManager, pDataBufferNode, (ma_decoder*)pJob->data.resourceManager.pageDataBufferNode.pDecoder); + + /* + If we have a success code by this point, we want to post another job. We're going to set the + result back to MA_BUSY to make it clear that there's still more to load. + */ + if (result == MA_SUCCESS) { + ma_job newJob; + newJob = *pJob; /* Everything is the same as the input job, except the execution order. */ + newJob.order = ma_resource_manager_data_buffer_node_next_execution_order(pDataBufferNode); /* We need a fresh execution order. */ + + result = ma_resource_manager_post_job(pResourceManager, &newJob); + + /* Since the sound isn't yet fully decoded we want the status to be set to busy. */ + if (result == MA_SUCCESS) { + result = MA_BUSY; + } + } + +done: + /* If there's still more to decode the result will be set to MA_BUSY. Otherwise we can free the decoder. */ + if (result != MA_BUSY) { + ma_decoder_uninit((ma_decoder*)pJob->data.resourceManager.pageDataBufferNode.pDecoder); + ma_free(pJob->data.resourceManager.pageDataBufferNode.pDecoder, &pResourceManager->config.allocationCallbacks); + } + + /* If we reached the end we need to treat it as successful. */ + if (result == MA_AT_END) { + result = MA_SUCCESS; + } + + /* Make sure we set the result of node in case some error occurred. */ + ma_atomic_compare_and_swap_i32(&pDataBufferNode->result, MA_BUSY, result); + + /* Signal the notification after setting the result in case the notification callback wants to inspect the result code. */ + if (result != MA_BUSY) { + if (pJob->data.resourceManager.pageDataBufferNode.pDoneNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.pageDataBufferNode.pDoneNotification); + } + + if (pJob->data.resourceManager.pageDataBufferNode.pDoneFence != NULL) { + ma_fence_release(pJob->data.resourceManager.pageDataBufferNode.pDoneFence); + } + } + + ma_atomic_fetch_add_32(&pDataBufferNode->executionPointer, 1); + return result; +} + + +static ma_result ma_job_process__resource_manager__load_data_buffer(ma_job* pJob) +{ + ma_result result = MA_SUCCESS; + ma_resource_manager* pResourceManager; + ma_resource_manager_data_buffer* pDataBuffer; + ma_resource_manager_data_supply_type dataSupplyType = ma_resource_manager_data_supply_type_unknown; + ma_bool32 isConnectorInitialized = MA_FALSE; + + /* + All we're doing here is checking if the node has finished loading. If not, we just re-post the job + and keep waiting. Otherwise we increment the execution counter and set the buffer's result code. + */ + MA_ASSERT(pJob != NULL); + + pDataBuffer = (ma_resource_manager_data_buffer*)pJob->data.resourceManager.loadDataBuffer.pDataBuffer; + MA_ASSERT(pDataBuffer != NULL); + + pResourceManager = pDataBuffer->pResourceManager; + + if (pJob->order != ma_atomic_load_32(&pDataBuffer->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Attempting to execute out of order. Probably interleaved with a MA_JOB_TYPE_RESOURCE_MANAGER_FREE_DATA_BUFFER job. */ + } + + /* + First thing we need to do is check whether or not the data buffer is getting deleted. If so we + just abort, but making sure we increment the execution pointer. + */ + result = ma_resource_manager_data_buffer_result(pDataBuffer); + if (result != MA_BUSY) { + goto done; /* <-- This will ensure the execution pointer is incremented. */ + } else { + result = MA_SUCCESS; /* <-- Make sure this is reset. */ + (void)result; /* <-- This is to suppress a static analysis diagnostic about "result" not being used. But for safety when I do future maintenance I don't want to delete that assignment. */ + } + + /* Try initializing the connector if we haven't already. */ + isConnectorInitialized = ma_resource_manager_data_buffer_has_connector(pDataBuffer); + if (isConnectorInitialized == MA_FALSE) { + dataSupplyType = ma_resource_manager_data_buffer_node_get_data_supply_type(pDataBuffer->pNode); + + if (dataSupplyType != ma_resource_manager_data_supply_type_unknown) { + /* We can now initialize the connector. If this fails, we need to abort. It's very rare for this to fail. */ + ma_resource_manager_data_source_config dataSourceConfig; /* For setting initial looping state and range. */ + dataSourceConfig = ma_resource_manager_data_source_config_init(); + dataSourceConfig.rangeBegInPCMFrames = pJob->data.resourceManager.loadDataBuffer.rangeBegInPCMFrames; + dataSourceConfig.rangeEndInPCMFrames = pJob->data.resourceManager.loadDataBuffer.rangeEndInPCMFrames; + dataSourceConfig.loopPointBegInPCMFrames = pJob->data.resourceManager.loadDataBuffer.loopPointBegInPCMFrames; + dataSourceConfig.loopPointEndInPCMFrames = pJob->data.resourceManager.loadDataBuffer.loopPointEndInPCMFrames; + dataSourceConfig.isLooping = pJob->data.resourceManager.loadDataBuffer.isLooping; + + result = ma_resource_manager_data_buffer_init_connector(pDataBuffer, &dataSourceConfig, pJob->data.resourceManager.loadDataBuffer.pInitNotification, pJob->data.resourceManager.loadDataBuffer.pInitFence); + if (result != MA_SUCCESS) { + ma_log_postf(ma_resource_manager_get_log(pResourceManager), MA_LOG_LEVEL_ERROR, "Failed to initialize connector for data buffer. %s.\n", ma_result_description(result)); + goto done; + } + } else { + /* Don't have a known data supply type. Most likely the data buffer node is still loading, but it could be that an error occurred. */ + } + } else { + /* The connector is already initialized. Nothing to do here. */ + } + + /* + If the data node is still loading, we need to repost the job and *not* increment the execution + pointer (i.e. we need to not fall through to the "done" label). + + There is a hole between here and the where the data connector is initialized where the data + buffer node may have finished initializing. We need to check for this by checking the result of + the data buffer node and whether or not we had an unknown data supply type at the time of + trying to initialize the data connector. + */ + result = ma_resource_manager_data_buffer_node_result(pDataBuffer->pNode); + if (result == MA_BUSY || (result == MA_SUCCESS && isConnectorInitialized == MA_FALSE && dataSupplyType == ma_resource_manager_data_supply_type_unknown)) { + return ma_resource_manager_post_job(pResourceManager, pJob); + } + +done: + /* Only move away from a busy code so that we don't trash any existing error codes. */ + ma_atomic_compare_and_swap_i32(&pDataBuffer->result, MA_BUSY, result); + + /* Only signal the other threads after the result has been set just for cleanliness sake. */ + if (pJob->data.resourceManager.loadDataBuffer.pDoneNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.loadDataBuffer.pDoneNotification); + } + if (pJob->data.resourceManager.loadDataBuffer.pDoneFence != NULL) { + ma_fence_release(pJob->data.resourceManager.loadDataBuffer.pDoneFence); + } + + /* + If at this point the data buffer has not had it's connector initialized, it means the + notification event was never signalled which means we need to signal it here. + */ + if (ma_resource_manager_data_buffer_has_connector(pDataBuffer) == MA_FALSE && result != MA_SUCCESS) { + if (pJob->data.resourceManager.loadDataBuffer.pInitNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.loadDataBuffer.pInitNotification); + } + if (pJob->data.resourceManager.loadDataBuffer.pInitFence != NULL) { + ma_fence_release(pJob->data.resourceManager.loadDataBuffer.pInitFence); + } + } + + ma_atomic_fetch_add_32(&pDataBuffer->executionPointer, 1); + return result; +} + +static ma_result ma_job_process__resource_manager__free_data_buffer(ma_job* pJob) +{ + ma_resource_manager* pResourceManager; + ma_resource_manager_data_buffer* pDataBuffer; + + MA_ASSERT(pJob != NULL); + + pDataBuffer = (ma_resource_manager_data_buffer*)pJob->data.resourceManager.freeDataBuffer.pDataBuffer; + MA_ASSERT(pDataBuffer != NULL); + + pResourceManager = pDataBuffer->pResourceManager; + + if (pJob->order != ma_atomic_load_32(&pDataBuffer->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Out of order. */ + } + + ma_resource_manager_data_buffer_uninit_internal(pDataBuffer); + + /* The event needs to be signalled last. */ + if (pJob->data.resourceManager.freeDataBuffer.pDoneNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.freeDataBuffer.pDoneNotification); + } + + if (pJob->data.resourceManager.freeDataBuffer.pDoneFence != NULL) { + ma_fence_release(pJob->data.resourceManager.freeDataBuffer.pDoneFence); + } + + ma_atomic_fetch_add_32(&pDataBuffer->executionPointer, 1); + return MA_SUCCESS; +} + +static ma_result ma_job_process__resource_manager__load_data_stream(ma_job* pJob) +{ + ma_result result = MA_SUCCESS; + ma_decoder_config decoderConfig; + ma_uint32 pageBufferSizeInBytes; + ma_resource_manager* pResourceManager; + ma_resource_manager_data_stream* pDataStream; + + MA_ASSERT(pJob != NULL); + + pDataStream = (ma_resource_manager_data_stream*)pJob->data.resourceManager.loadDataStream.pDataStream; + MA_ASSERT(pDataStream != NULL); + + pResourceManager = pDataStream->pResourceManager; + + if (pJob->order != ma_atomic_load_32(&pDataStream->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Out of order. */ + } + + if (ma_resource_manager_data_stream_result(pDataStream) != MA_BUSY) { + result = MA_INVALID_OPERATION; /* Most likely the data stream is being uninitialized. */ + goto done; + } + + /* We need to initialize the decoder first so we can determine the size of the pages. */ + decoderConfig = ma_resource_manager__init_decoder_config(pResourceManager); + + if (pJob->data.resourceManager.loadDataStream.pFilePath != NULL) { + result = ma_decoder_init_vfs(pResourceManager->config.pVFS, pJob->data.resourceManager.loadDataStream.pFilePath, &decoderConfig, &pDataStream->decoder); + } else { + result = ma_decoder_init_vfs_w(pResourceManager->config.pVFS, pJob->data.resourceManager.loadDataStream.pFilePathW, &decoderConfig, &pDataStream->decoder); + } + if (result != MA_SUCCESS) { + goto done; + } + + /* Retrieve the total length of the file before marking the decoder as loaded. */ + if ((pDataStream->flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_UNKNOWN_LENGTH) == 0) { + result = ma_decoder_get_length_in_pcm_frames(&pDataStream->decoder, &pDataStream->totalLengthInPCMFrames); + if (result != MA_SUCCESS) { + goto done; /* Failed to retrieve the length. */ + } + } else { + pDataStream->totalLengthInPCMFrames = 0; + } + + /* + Only mark the decoder as initialized when the length of the decoder has been retrieved because that can possibly require a scan over the whole file + and we don't want to have another thread trying to access the decoder while it's scanning. + */ + pDataStream->isDecoderInitialized = MA_TRUE; + + /* We have the decoder so we can now initialize our page buffer. */ + pageBufferSizeInBytes = ma_resource_manager_data_stream_get_page_size_in_frames(pDataStream) * 2 * ma_get_bytes_per_frame(pDataStream->decoder.outputFormat, pDataStream->decoder.outputChannels); + + pDataStream->pPageData = ma_malloc(pageBufferSizeInBytes, &pResourceManager->config.allocationCallbacks); + if (pDataStream->pPageData == NULL) { + ma_decoder_uninit(&pDataStream->decoder); + result = MA_OUT_OF_MEMORY; + goto done; + } + + /* Seek to our initial seek point before filling the initial pages. */ + ma_decoder_seek_to_pcm_frame(&pDataStream->decoder, pJob->data.resourceManager.loadDataStream.initialSeekPoint); + + /* We have our decoder and our page buffer, so now we need to fill our pages. */ + ma_resource_manager_data_stream_fill_pages(pDataStream); + + /* And now we're done. We want to make sure the result is MA_SUCCESS. */ + result = MA_SUCCESS; + +done: + ma_free(pJob->data.resourceManager.loadDataStream.pFilePath, &pResourceManager->config.allocationCallbacks); + ma_free(pJob->data.resourceManager.loadDataStream.pFilePathW, &pResourceManager->config.allocationCallbacks); + + /* We can only change the status away from MA_BUSY. If it's set to anything else it means an error has occurred somewhere or the uninitialization process has started (most likely). */ + ma_atomic_compare_and_swap_i32(&pDataStream->result, MA_BUSY, result); + + /* Only signal the other threads after the result has been set just for cleanliness sake. */ + if (pJob->data.resourceManager.loadDataStream.pInitNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.loadDataStream.pInitNotification); + } + if (pJob->data.resourceManager.loadDataStream.pInitFence != NULL) { + ma_fence_release(pJob->data.resourceManager.loadDataStream.pInitFence); + } + + ma_atomic_fetch_add_32(&pDataStream->executionPointer, 1); + return result; +} + +static ma_result ma_job_process__resource_manager__free_data_stream(ma_job* pJob) +{ + ma_resource_manager* pResourceManager; + ma_resource_manager_data_stream* pDataStream; + + MA_ASSERT(pJob != NULL); + + pDataStream = (ma_resource_manager_data_stream*)pJob->data.resourceManager.freeDataStream.pDataStream; + MA_ASSERT(pDataStream != NULL); + + pResourceManager = pDataStream->pResourceManager; + + if (pJob->order != ma_atomic_load_32(&pDataStream->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Out of order. */ + } + + /* If our status is not MA_UNAVAILABLE we have a bug somewhere. */ + MA_ASSERT(ma_resource_manager_data_stream_result(pDataStream) == MA_UNAVAILABLE); + + if (pDataStream->isDecoderInitialized) { + ma_decoder_uninit(&pDataStream->decoder); + } + + if (pDataStream->pPageData != NULL) { + ma_free(pDataStream->pPageData, &pResourceManager->config.allocationCallbacks); + pDataStream->pPageData = NULL; /* Just in case... */ + } + + ma_data_source_uninit(&pDataStream->ds); + + /* The event needs to be signalled last. */ + if (pJob->data.resourceManager.freeDataStream.pDoneNotification != NULL) { + ma_async_notification_signal(pJob->data.resourceManager.freeDataStream.pDoneNotification); + } + if (pJob->data.resourceManager.freeDataStream.pDoneFence != NULL) { + ma_fence_release(pJob->data.resourceManager.freeDataStream.pDoneFence); + } + + /*ma_atomic_fetch_add_32(&pDataStream->executionPointer, 1);*/ + return MA_SUCCESS; +} + +static ma_result ma_job_process__resource_manager__page_data_stream(ma_job* pJob) +{ + ma_result result = MA_SUCCESS; + ma_resource_manager* pResourceManager; + ma_resource_manager_data_stream* pDataStream; + + MA_ASSERT(pJob != NULL); + + pDataStream = (ma_resource_manager_data_stream*)pJob->data.resourceManager.pageDataStream.pDataStream; + MA_ASSERT(pDataStream != NULL); + + pResourceManager = pDataStream->pResourceManager; + + if (pJob->order != ma_atomic_load_32(&pDataStream->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Out of order. */ + } + + /* For streams, the status should be MA_SUCCESS. */ + if (ma_resource_manager_data_stream_result(pDataStream) != MA_SUCCESS) { + result = MA_INVALID_OPERATION; + goto done; + } + + ma_resource_manager_data_stream_fill_page(pDataStream, pJob->data.resourceManager.pageDataStream.pageIndex); + +done: + ma_atomic_fetch_add_32(&pDataStream->executionPointer, 1); + return result; +} + +static ma_result ma_job_process__resource_manager__seek_data_stream(ma_job* pJob) +{ + ma_result result = MA_SUCCESS; + ma_resource_manager* pResourceManager; + ma_resource_manager_data_stream* pDataStream; + + MA_ASSERT(pJob != NULL); + + pDataStream = (ma_resource_manager_data_stream*)pJob->data.resourceManager.seekDataStream.pDataStream; + MA_ASSERT(pDataStream != NULL); + + pResourceManager = pDataStream->pResourceManager; + + if (pJob->order != ma_atomic_load_32(&pDataStream->executionPointer)) { + return ma_resource_manager_post_job(pResourceManager, pJob); /* Out of order. */ + } + + /* For streams the status should be MA_SUCCESS for this to do anything. */ + if (ma_resource_manager_data_stream_result(pDataStream) != MA_SUCCESS || pDataStream->isDecoderInitialized == MA_FALSE) { + result = MA_INVALID_OPERATION; + goto done; + } + + /* + With seeking we just assume both pages are invalid and the relative frame cursor at position 0. This is basically exactly the same as loading, except + instead of initializing the decoder, we seek to a frame. + */ + ma_decoder_seek_to_pcm_frame(&pDataStream->decoder, pJob->data.resourceManager.seekDataStream.frameIndex); + + /* After seeking we'll need to reload the pages. */ + ma_resource_manager_data_stream_fill_pages(pDataStream); + + /* We need to let the public API know that we're done seeking. */ + ma_atomic_fetch_sub_32(&pDataStream->seekCounter, 1); + +done: + ma_atomic_fetch_add_32(&pDataStream->executionPointer, 1); + return result; +} + +MA_API ma_result ma_resource_manager_process_job(ma_resource_manager* pResourceManager, ma_job* pJob) +{ + if (pResourceManager == NULL || pJob == NULL) { + return MA_INVALID_ARGS; + } + + return ma_job_process(pJob); +} + +MA_API ma_result ma_resource_manager_process_next_job(ma_resource_manager* pResourceManager) +{ + ma_result result; + ma_job job; + + if (pResourceManager == NULL) { + return MA_INVALID_ARGS; + } + + /* This will return MA_CANCELLED if the next job is a quit job. */ + result = ma_resource_manager_next_job(pResourceManager, &job); + if (result != MA_SUCCESS) { + return result; + } + + return ma_job_process(&job); +} +#else +/* We'll get here if the resource manager is being excluded from the build. We need to define the job processing callbacks as no-ops. */ +static ma_result ma_job_process__resource_manager__load_data_buffer_node(ma_job* pJob) { return ma_job_process__noop(pJob); } +static ma_result ma_job_process__resource_manager__free_data_buffer_node(ma_job* pJob) { return ma_job_process__noop(pJob); } +static ma_result ma_job_process__resource_manager__page_data_buffer_node(ma_job* pJob) { return ma_job_process__noop(pJob); } +static ma_result ma_job_process__resource_manager__load_data_buffer(ma_job* pJob) { return ma_job_process__noop(pJob); } +static ma_result ma_job_process__resource_manager__free_data_buffer(ma_job* pJob) { return ma_job_process__noop(pJob); } +static ma_result ma_job_process__resource_manager__load_data_stream(ma_job* pJob) { return ma_job_process__noop(pJob); } +static ma_result ma_job_process__resource_manager__free_data_stream(ma_job* pJob) { return ma_job_process__noop(pJob); } +static ma_result ma_job_process__resource_manager__page_data_stream(ma_job* pJob) { return ma_job_process__noop(pJob); } +static ma_result ma_job_process__resource_manager__seek_data_stream(ma_job* pJob) { return ma_job_process__noop(pJob); } +#endif /* MA_NO_RESOURCE_MANAGER */ + + +#ifndef MA_NO_NODE_GRAPH + +static ma_stack* ma_stack_init(size_t sizeInBytes, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_stack* pStack; + + if (sizeInBytes == 0) { + return NULL; + } + + pStack = (ma_stack*)ma_malloc(sizeof(*pStack) - sizeof(pStack->_data) + sizeInBytes, pAllocationCallbacks); + if (pStack == NULL) { + return NULL; + } + + pStack->offset = 0; + pStack->sizeInBytes = sizeInBytes; + + return pStack; +} + +static void ma_stack_uninit(ma_stack* pStack, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pStack == NULL) { + return; + } + + ma_free(pStack, pAllocationCallbacks); +} + +static void* ma_stack_alloc(ma_stack* pStack, size_t sz) +{ + /* The size of the allocation is stored in the memory directly before the pointer. This needs to include padding to keep it aligned to ma_uintptr */ + void* p = (void*)((char*)pStack->_data + pStack->offset); + size_t* pSize = (size_t*)p; + + sz = (sz + (sizeof(ma_uintptr) - 1)) & ~(sizeof(ma_uintptr) - 1); /* Padding. */ + if (pStack->offset + sz + sizeof(size_t) > pStack->sizeInBytes) { + return NULL; /* Out of memory. */ + } + + pStack->offset += sz + sizeof(size_t); + + *pSize = sz; + return (void*)((char*)p + sizeof(size_t)); +} + +static void ma_stack_free(ma_stack* pStack, void* p) +{ + size_t* pSize; + + if (p == NULL) { + return; + } + + pSize = (size_t*)p - 1; + pStack->offset -= *pSize + sizeof(size_t); +} + + + +/* 10ms @ 48K = 480. Must never exceed 65535. */ +#ifndef MA_DEFAULT_NODE_CACHE_CAP_IN_FRAMES_PER_BUS +#define MA_DEFAULT_NODE_CACHE_CAP_IN_FRAMES_PER_BUS 480 +#endif + +#ifndef MA_DEFAULT_PREMIX_STACK_SIZE_PER_CHANNEL +#define MA_DEFAULT_PREMIX_STACK_SIZE_PER_CHANNEL 524288 +#endif + +static ma_result ma_node_read_pcm_frames(ma_node* pNode, ma_uint32 outputBusIndex, float* pFramesOut, ma_uint32 frameCount, ma_uint32* pFramesRead, ma_uint64 globalTime); + +MA_API void ma_debug_fill_pcm_frames_with_sine_wave(float* pFramesOut, ma_uint32 frameCount, ma_format format, ma_uint32 channels, ma_uint32 sampleRate) +{ + #ifndef MA_NO_GENERATION + { + ma_waveform_config waveformConfig; + ma_waveform waveform; + + waveformConfig = ma_waveform_config_init(format, channels, sampleRate, ma_waveform_type_sine, 1.0, 400); + ma_waveform_init(&waveformConfig, &waveform); + ma_waveform_read_pcm_frames(&waveform, pFramesOut, frameCount, NULL); + } + #else + { + (void)pFramesOut; + (void)frameCount; + (void)format; + (void)channels; + (void)sampleRate; + #if defined(MA_DEBUG_OUTPUT) + { + #if _MSC_VER + #pragma message ("ma_debug_fill_pcm_frames_with_sine_wave() will do nothing because MA_NO_GENERATION is enabled.") + #endif + } + #endif + } + #endif +} + + + +MA_API ma_node_graph_config ma_node_graph_config_init(ma_uint32 channels) +{ + ma_node_graph_config config; + + MA_ZERO_OBJECT(&config); + config.channels = channels; + config.processingSizeInFrames = 0; + + return config; +} + + +static void ma_node_graph_set_is_reading(ma_node_graph* pNodeGraph, ma_bool32 isReading) +{ + MA_ASSERT(pNodeGraph != NULL); + ma_atomic_exchange_32(&pNodeGraph->isReading, isReading); +} + +#if 0 +static ma_bool32 ma_node_graph_is_reading(ma_node_graph* pNodeGraph) +{ + MA_ASSERT(pNodeGraph != NULL); + return ma_atomic_load_32(&pNodeGraph->isReading); +} +#endif + + +static void ma_node_graph_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_node_graph* pNodeGraph = (ma_node_graph*)pNode; + ma_uint64 framesRead; + + ma_node_graph_read_pcm_frames(pNodeGraph, ppFramesOut[0], *pFrameCountOut, &framesRead); + + *pFrameCountOut = (ma_uint32)framesRead; /* Safe cast. */ + + (void)ppFramesIn; + (void)pFrameCountIn; +} + +static ma_node_vtable g_node_graph_node_vtable = +{ + ma_node_graph_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 0, /* 0 input buses. */ + 1, /* 1 output bus. */ + 0 /* Flags. */ +}; + +static void ma_node_graph_endpoint_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + MA_ASSERT(pNode != NULL); + MA_ASSERT(ma_node_get_input_bus_count(pNode) == 1); + MA_ASSERT(ma_node_get_output_bus_count(pNode) == 1); + + /* Input channel count needs to be the same as the output channel count. */ + MA_ASSERT(ma_node_get_input_channels(pNode, 0) == ma_node_get_output_channels(pNode, 0)); + + /* We don't need to do anything here because it's a passthrough. */ + (void)pNode; + (void)ppFramesIn; + (void)pFrameCountIn; + (void)ppFramesOut; + (void)pFrameCountOut; + +#if 0 + /* The data has already been mixed. We just need to move it to the output buffer. */ + if (ppFramesIn != NULL) { + ma_copy_pcm_frames(ppFramesOut[0], ppFramesIn[0], *pFrameCountOut, ma_format_f32, ma_node_get_output_channels(pNode, 0)); + } +#endif +} + +static ma_node_vtable g_node_graph_endpoint_vtable = +{ + ma_node_graph_endpoint_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* 1 input bus. */ + 1, /* 1 output bus. */ + MA_NODE_FLAG_PASSTHROUGH /* Flags. The endpoint is a passthrough. */ +}; + +MA_API ma_result ma_node_graph_init(const ma_node_graph_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_node_graph* pNodeGraph) +{ + ma_result result; + ma_node_config baseConfig; + ma_node_config endpointConfig; + + if (pNodeGraph == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNodeGraph); + pNodeGraph->processingSizeInFrames = pConfig->processingSizeInFrames; + + /* Base node so we can use the node graph as a node into another graph. */ + baseConfig = ma_node_config_init(); + baseConfig.vtable = &g_node_graph_node_vtable; + baseConfig.pOutputChannels = &pConfig->channels; + + result = ma_node_init(pNodeGraph, &baseConfig, pAllocationCallbacks, &pNodeGraph->base); + if (result != MA_SUCCESS) { + return result; + } + + + /* Endpoint. */ + endpointConfig = ma_node_config_init(); + endpointConfig.vtable = &g_node_graph_endpoint_vtable; + endpointConfig.pInputChannels = &pConfig->channels; + endpointConfig.pOutputChannels = &pConfig->channels; + + result = ma_node_init(pNodeGraph, &endpointConfig, pAllocationCallbacks, &pNodeGraph->endpoint); + if (result != MA_SUCCESS) { + ma_node_uninit(&pNodeGraph->base, pAllocationCallbacks); + return result; + } + + + /* Processing cache. */ + if (pConfig->processingSizeInFrames > 0) { + pNodeGraph->pProcessingCache = (float*)ma_malloc(pConfig->processingSizeInFrames * pConfig->channels * sizeof(float), pAllocationCallbacks); + if (pNodeGraph->pProcessingCache == NULL) { + ma_node_uninit(&pNodeGraph->endpoint, pAllocationCallbacks); + ma_node_uninit(&pNodeGraph->base, pAllocationCallbacks); + return MA_OUT_OF_MEMORY; + } + } + + + /* + We need a pre-mix stack. The size of this stack is configurable via the config. The default value depends on the channel count. + */ + { + size_t preMixStackSizeInBytes = pConfig->preMixStackSizeInBytes; + if (preMixStackSizeInBytes == 0) { + preMixStackSizeInBytes = pConfig->channels * MA_DEFAULT_PREMIX_STACK_SIZE_PER_CHANNEL; + } + + pNodeGraph->pPreMixStack = ma_stack_init(preMixStackSizeInBytes, pAllocationCallbacks); + if (pNodeGraph->pPreMixStack == NULL) { + ma_node_uninit(&pNodeGraph->endpoint, pAllocationCallbacks); + ma_node_uninit(&pNodeGraph->base, pAllocationCallbacks); + if (pNodeGraph->pProcessingCache != NULL) { + ma_free(pNodeGraph->pProcessingCache, pAllocationCallbacks); + } + + return MA_OUT_OF_MEMORY; + } + } + + + return MA_SUCCESS; +} + +MA_API void ma_node_graph_uninit(ma_node_graph* pNodeGraph, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pNodeGraph == NULL) { + return; + } + + ma_node_uninit(&pNodeGraph->endpoint, pAllocationCallbacks); + ma_node_uninit(&pNodeGraph->base, pAllocationCallbacks); + + if (pNodeGraph->pProcessingCache != NULL) { + ma_free(pNodeGraph->pProcessingCache, pAllocationCallbacks); + pNodeGraph->pProcessingCache = NULL; + } + + if (pNodeGraph->pPreMixStack != NULL) { + ma_stack_uninit(pNodeGraph->pPreMixStack, pAllocationCallbacks); + pNodeGraph->pPreMixStack = NULL; + } +} + +MA_API ma_node* ma_node_graph_get_endpoint(ma_node_graph* pNodeGraph) +{ + if (pNodeGraph == NULL) { + return NULL; + } + + return &pNodeGraph->endpoint; +} + +MA_API ma_result ma_node_graph_read_pcm_frames(ma_node_graph* pNodeGraph, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_result result = MA_SUCCESS; + ma_uint64 totalFramesRead; + ma_uint32 channels; + + if (pFramesRead != NULL) { + *pFramesRead = 0; /* Safety. */ + } + + if (pNodeGraph == NULL) { + return MA_INVALID_ARGS; + } + + channels = ma_node_get_output_channels(&pNodeGraph->endpoint, 0); + + + /* We'll be nice and try to do a full read of all frameCount frames. */ + totalFramesRead = 0; + while (totalFramesRead < frameCount) { + ma_uint32 framesJustRead; + ma_uint64 framesToRead; + float* pRunningFramesOut; + + framesToRead = frameCount - totalFramesRead; + if (framesToRead > 0xFFFFFFFF) { + framesToRead = 0xFFFFFFFF; + } + + pRunningFramesOut = (float*)ma_offset_pcm_frames_ptr(pFramesOut, totalFramesRead, ma_format_f32, channels); + + /* If there's anything in the cache, consume that first. */ + if (pNodeGraph->processingCacheFramesRemaining > 0) { + ma_uint32 framesToReadFromCache; + + framesToReadFromCache = (ma_uint32)framesToRead; + if (framesToReadFromCache > pNodeGraph->processingCacheFramesRemaining) { + framesToReadFromCache = pNodeGraph->processingCacheFramesRemaining; + } + + MA_COPY_MEMORY(pRunningFramesOut, pNodeGraph->pProcessingCache, framesToReadFromCache * channels * sizeof(float)); + MA_MOVE_MEMORY(pNodeGraph->pProcessingCache, pNodeGraph->pProcessingCache + (framesToReadFromCache * channels), (pNodeGraph->processingCacheFramesRemaining - framesToReadFromCache) * channels * sizeof(float)); + pNodeGraph->processingCacheFramesRemaining -= framesToReadFromCache; + + totalFramesRead += framesToReadFromCache; + continue; + } else { + /* + If processingSizeInFrames is non-zero, we need to make sure we always read in chunks of that size. If the frame count is less than + that, we need to read into the cache and then continue on. + */ + float* pReadDst = pRunningFramesOut; + + if (pNodeGraph->processingSizeInFrames > 0) { + if (framesToRead < pNodeGraph->processingSizeInFrames) { + pReadDst = pNodeGraph->pProcessingCache; /* We need to read into the cache because otherwise we'll overflow the output buffer. */ + } + + framesToRead = pNodeGraph->processingSizeInFrames; + } + + ma_node_graph_set_is_reading(pNodeGraph, MA_TRUE); + { + result = ma_node_read_pcm_frames(&pNodeGraph->endpoint, 0, pReadDst, (ma_uint32)framesToRead, &framesJustRead, ma_node_get_time(&pNodeGraph->endpoint)); + } + ma_node_graph_set_is_reading(pNodeGraph, MA_FALSE); + + /* + Do not increment the total frames read counter if we read into the cache. We use this to determine how many frames have + been written to the final output buffer. + */ + if (pReadDst == pNodeGraph->pProcessingCache) { + /* We read into the cache. */ + pNodeGraph->processingCacheFramesRemaining = framesJustRead; + } else { + /* We read straight into the output buffer. */ + totalFramesRead += framesJustRead; + } + + if (result != MA_SUCCESS) { + break; + } + + /* Abort if we weren't able to read any frames or else we risk getting stuck in a loop. */ + if (framesJustRead == 0) { + break; + } + } + } + + /* Let's go ahead and silence any leftover frames just for some added safety to ensure the caller doesn't try emitting garbage out of the speakers. */ + if (totalFramesRead < frameCount) { + ma_silence_pcm_frames(ma_offset_pcm_frames_ptr(pFramesOut, totalFramesRead, ma_format_f32, channels), (frameCount - totalFramesRead), ma_format_f32, channels); + } + + if (pFramesRead != NULL) { + *pFramesRead = totalFramesRead; + } + + return result; +} + +MA_API ma_uint32 ma_node_graph_get_channels(const ma_node_graph* pNodeGraph) +{ + if (pNodeGraph == NULL) { + return 0; + } + + return ma_node_get_output_channels(&pNodeGraph->endpoint, 0); +} + +MA_API ma_uint64 ma_node_graph_get_time(const ma_node_graph* pNodeGraph) +{ + if (pNodeGraph == NULL) { + return 0; + } + + return ma_node_get_time(&pNodeGraph->endpoint); /* Global time is just the local time of the endpoint. */ +} + +MA_API ma_result ma_node_graph_set_time(ma_node_graph* pNodeGraph, ma_uint64 globalTime) +{ + if (pNodeGraph == NULL) { + return MA_INVALID_ARGS; + } + + return ma_node_set_time(&pNodeGraph->endpoint, globalTime); /* Global time is just the local time of the endpoint. */ +} + + +#define MA_NODE_OUTPUT_BUS_FLAG_HAS_READ 0x01 /* Whether or not this bus ready to read more data. Only used on nodes with multiple output buses. */ + +static ma_result ma_node_output_bus_init(ma_node* pNode, ma_uint32 outputBusIndex, ma_uint32 channels, ma_node_output_bus* pOutputBus) +{ + MA_ASSERT(pOutputBus != NULL); + MA_ASSERT(outputBusIndex < MA_MAX_NODE_BUS_COUNT); + MA_ASSERT(outputBusIndex < ma_node_get_output_bus_count(pNode)); + MA_ASSERT(channels < 256); + + MA_ZERO_OBJECT(pOutputBus); + + if (channels == 0) { + return MA_INVALID_ARGS; + } + + pOutputBus->pNode = pNode; + pOutputBus->outputBusIndex = (ma_uint8)outputBusIndex; + pOutputBus->channels = (ma_uint8)channels; + pOutputBus->flags = MA_NODE_OUTPUT_BUS_FLAG_HAS_READ; /* <-- Important that this flag is set by default. */ + pOutputBus->volume = 1; + + return MA_SUCCESS; +} + +static void ma_node_output_bus_lock(ma_node_output_bus* pOutputBus) +{ + ma_spinlock_lock(&pOutputBus->lock); +} + +static void ma_node_output_bus_unlock(ma_node_output_bus* pOutputBus) +{ + ma_spinlock_unlock(&pOutputBus->lock); +} + + +static ma_uint32 ma_node_output_bus_get_channels(const ma_node_output_bus* pOutputBus) +{ + return pOutputBus->channels; +} + + +static void ma_node_output_bus_set_has_read(ma_node_output_bus* pOutputBus, ma_bool32 hasRead) +{ + if (hasRead) { + ma_atomic_fetch_or_32(&pOutputBus->flags, MA_NODE_OUTPUT_BUS_FLAG_HAS_READ); + } else { + ma_atomic_fetch_and_32(&pOutputBus->flags, (ma_uint32)~MA_NODE_OUTPUT_BUS_FLAG_HAS_READ); + } +} + +static ma_bool32 ma_node_output_bus_has_read(ma_node_output_bus* pOutputBus) +{ + return (ma_atomic_load_32(&pOutputBus->flags) & MA_NODE_OUTPUT_BUS_FLAG_HAS_READ) != 0; +} + + +static void ma_node_output_bus_set_is_attached(ma_node_output_bus* pOutputBus, ma_bool32 isAttached) +{ + ma_atomic_exchange_32(&pOutputBus->isAttached, isAttached); +} + +static ma_bool32 ma_node_output_bus_is_attached(ma_node_output_bus* pOutputBus) +{ + return ma_atomic_load_32(&pOutputBus->isAttached); +} + + +static ma_result ma_node_output_bus_set_volume(ma_node_output_bus* pOutputBus, float volume) +{ + MA_ASSERT(pOutputBus != NULL); + + if (volume < 0.0f) { + volume = 0.0f; + } + + ma_atomic_exchange_f32(&pOutputBus->volume, volume); + + return MA_SUCCESS; +} + +static float ma_node_output_bus_get_volume(const ma_node_output_bus* pOutputBus) +{ + return ma_atomic_load_f32((float*)&pOutputBus->volume); +} + + +static ma_result ma_node_input_bus_init(ma_uint32 channels, ma_node_input_bus* pInputBus) +{ + MA_ASSERT(pInputBus != NULL); + MA_ASSERT(channels < 256); + + MA_ZERO_OBJECT(pInputBus); + + if (channels == 0) { + return MA_INVALID_ARGS; + } + + pInputBus->channels = (ma_uint8)channels; + + return MA_SUCCESS; +} + +static void ma_node_input_bus_lock(ma_node_input_bus* pInputBus) +{ + MA_ASSERT(pInputBus != NULL); + + ma_spinlock_lock(&pInputBus->lock); +} + +static void ma_node_input_bus_unlock(ma_node_input_bus* pInputBus) +{ + MA_ASSERT(pInputBus != NULL); + + ma_spinlock_unlock(&pInputBus->lock); +} + + +static void ma_node_input_bus_next_begin(ma_node_input_bus* pInputBus) +{ + ma_atomic_fetch_add_32(&pInputBus->nextCounter, 1); +} + +static void ma_node_input_bus_next_end(ma_node_input_bus* pInputBus) +{ + ma_atomic_fetch_sub_32(&pInputBus->nextCounter, 1); +} + +static ma_uint32 ma_node_input_bus_get_next_counter(ma_node_input_bus* pInputBus) +{ + return ma_atomic_load_32(&pInputBus->nextCounter); +} + + +static ma_uint32 ma_node_input_bus_get_channels(const ma_node_input_bus* pInputBus) +{ + return pInputBus->channels; +} + + +static void ma_node_input_bus_detach__no_output_bus_lock(ma_node_input_bus* pInputBus, ma_node_output_bus* pOutputBus) +{ + MA_ASSERT(pInputBus != NULL); + MA_ASSERT(pOutputBus != NULL); + + /* + Mark the output bus as detached first. This will prevent future iterations on the audio thread + from iterating this output bus. + */ + ma_node_output_bus_set_is_attached(pOutputBus, MA_FALSE); + + /* + We cannot use the output bus lock here since it'll be getting used at a higher level, but we do + still need to use the input bus lock since we'll be updating pointers on two different output + buses. The same rules apply here as the attaching case. Although we're using a lock here, we're + *not* using a lock when iterating over the list in the audio thread. We therefore need to craft + this in a way such that the iteration on the audio thread doesn't break. + + The first thing to do is swap out the "next" pointer of the previous output bus with the + new "next" output bus. This is the operation that matters for iteration on the audio thread. + After that, the previous pointer on the new "next" pointer needs to be updated, after which + point the linked list will be in a good state. + */ + ma_node_input_bus_lock(pInputBus); + { + ma_node_output_bus* pOldPrev = (ma_node_output_bus*)ma_atomic_load_ptr(&pOutputBus->pPrev); + ma_node_output_bus* pOldNext = (ma_node_output_bus*)ma_atomic_load_ptr(&pOutputBus->pNext); + + if (pOldPrev != NULL) { + ma_atomic_exchange_ptr(&pOldPrev->pNext, pOldNext); /* <-- This is where the output bus is detached from the list. */ + } + if (pOldNext != NULL) { + ma_atomic_exchange_ptr(&pOldNext->pPrev, pOldPrev); /* <-- This is required for detachment. */ + } + } + ma_node_input_bus_unlock(pInputBus); + + /* At this point the output bus is detached and the linked list is completely unaware of it. Reset some data for safety. */ + ma_atomic_exchange_ptr(&pOutputBus->pNext, NULL); /* Using atomic exchanges here, mainly for the benefit of analysis tools which don't always recognize spinlocks. */ + ma_atomic_exchange_ptr(&pOutputBus->pPrev, NULL); /* As above. */ + pOutputBus->pInputNode = NULL; + pOutputBus->inputNodeInputBusIndex = 0; + + + /* + For thread-safety reasons, we don't want to be returning from this straight away. We need to + wait for the audio thread to finish with the output bus. There's two things we need to wait + for. The first is the part that selects the next output bus in the list, and the other is the + part that reads from the output bus. Basically all we're doing is waiting for the input bus + to stop referencing the output bus. + + We're doing this part last because we want the section above to run while the audio thread + is finishing up with the output bus, just for efficiency reasons. We marked the output bus as + detached right at the top of this function which is going to prevent the audio thread from + iterating the output bus again. + */ + + /* Part 1: Wait for the current iteration to complete. */ + while (ma_node_input_bus_get_next_counter(pInputBus) > 0) { + ma_yield(); + } + + /* Part 2: Wait for any reads to complete. */ + while (ma_atomic_load_32(&pOutputBus->refCount) > 0) { + ma_yield(); + } + + /* + At this point we're done detaching and we can be guaranteed that the audio thread is not going + to attempt to reference this output bus again (until attached again). + */ +} + +#if 0 /* Not used at the moment, but leaving here in case I need it later. */ +static void ma_node_input_bus_detach(ma_node_input_bus* pInputBus, ma_node_output_bus* pOutputBus) +{ + MA_ASSERT(pInputBus != NULL); + MA_ASSERT(pOutputBus != NULL); + + ma_node_output_bus_lock(pOutputBus); + { + ma_node_input_bus_detach__no_output_bus_lock(pInputBus, pOutputBus); + } + ma_node_output_bus_unlock(pOutputBus); +} +#endif + +static void ma_node_input_bus_attach(ma_node_input_bus* pInputBus, ma_node_output_bus* pOutputBus, ma_node* pNewInputNode, ma_uint32 inputNodeInputBusIndex) +{ + MA_ASSERT(pInputBus != NULL); + MA_ASSERT(pOutputBus != NULL); + + ma_node_output_bus_lock(pOutputBus); + { + ma_node_output_bus* pOldInputNode = (ma_node_output_bus*)ma_atomic_load_ptr(&pOutputBus->pInputNode); + + /* Detach from any existing attachment first if necessary. */ + if (pOldInputNode != NULL) { + ma_node_input_bus_detach__no_output_bus_lock(pInputBus, pOutputBus); + } + + /* + At this point we can be sure the output bus is not attached to anything. The linked list in the + old input bus has been updated so that pOutputBus will not get iterated again. + */ + pOutputBus->pInputNode = pNewInputNode; /* No need for an atomic assignment here because modification of this variable always happens within a lock. */ + pOutputBus->inputNodeInputBusIndex = (ma_uint8)inputNodeInputBusIndex; + + /* + Now we need to attach the output bus to the linked list. This involves updating two pointers on + two different output buses so I'm going to go ahead and keep this simple and just use a lock. + There are ways to do this without a lock, but it's just too hard to maintain for its value. + + Although we're locking here, it's important to remember that we're *not* locking when iterating + and reading audio data since that'll be running on the audio thread. As a result we need to be + careful how we craft this so that we don't break iteration. What we're going to do is always + attach the new item so that it becomes the first item in the list. That way, as we're iterating + we won't break any links in the list and iteration will continue safely. The detaching case will + also be crafted in a way as to not break list iteration. It's important to remember to use + atomic exchanges here since no locking is happening on the audio thread during iteration. + */ + ma_node_input_bus_lock(pInputBus); + { + ma_node_output_bus* pNewPrev = &pInputBus->head; + ma_node_output_bus* pNewNext = (ma_node_output_bus*)ma_atomic_load_ptr(&pInputBus->head.pNext); + + /* Update the local output bus. */ + ma_atomic_exchange_ptr(&pOutputBus->pPrev, pNewPrev); + ma_atomic_exchange_ptr(&pOutputBus->pNext, pNewNext); + + /* Update the other output buses to point back to the local output bus. */ + ma_atomic_exchange_ptr(&pInputBus->head.pNext, pOutputBus); /* <-- This is where the output bus is actually attached to the input bus. */ + + /* Do the previous pointer last. This is only used for detachment. */ + if (pNewNext != NULL) { + ma_atomic_exchange_ptr(&pNewNext->pPrev, pOutputBus); + } + } + ma_node_input_bus_unlock(pInputBus); + + /* + Mark the node as attached last. This is used to controlling whether or the output bus will be + iterated on the audio thread. Mainly required for detachment purposes. + */ + ma_node_output_bus_set_is_attached(pOutputBus, MA_TRUE); + } + ma_node_output_bus_unlock(pOutputBus); +} + +static ma_node_output_bus* ma_node_input_bus_next(ma_node_input_bus* pInputBus, ma_node_output_bus* pOutputBus) +{ + ma_node_output_bus* pNext; + + MA_ASSERT(pInputBus != NULL); + + if (pOutputBus == NULL) { + return NULL; + } + + ma_node_input_bus_next_begin(pInputBus); + { + pNext = pOutputBus; + for (;;) { + pNext = (ma_node_output_bus*)ma_atomic_load_ptr(&pNext->pNext); + if (pNext == NULL) { + break; /* Reached the end. */ + } + + if (ma_node_output_bus_is_attached(pNext) == MA_FALSE) { + continue; /* The node is not attached. Keep checking. */ + } + + /* The next node has been selected. */ + break; + } + + /* We need to increment the reference count of the selected node. */ + if (pNext != NULL) { + ma_atomic_fetch_add_32(&pNext->refCount, 1); + } + + /* The previous node is no longer being referenced. */ + ma_atomic_fetch_sub_32(&pOutputBus->refCount, 1); + } + ma_node_input_bus_next_end(pInputBus); + + return pNext; +} + +static ma_node_output_bus* ma_node_input_bus_first(ma_node_input_bus* pInputBus) +{ + return ma_node_input_bus_next(pInputBus, &pInputBus->head); +} + + + +static ma_result ma_node_input_bus_read_pcm_frames(ma_node* pInputNode, ma_node_input_bus* pInputBus, float* pFramesOut, ma_uint32 frameCount, ma_uint32* pFramesRead, ma_uint64 globalTime) +{ + ma_result result = MA_SUCCESS; + ma_node_output_bus* pOutputBus; + ma_node_output_bus* pFirst; + ma_uint32 inputChannels; + ma_bool32 doesOutputBufferHaveContent = MA_FALSE; + + /* + This will be called from the audio thread which means we can't be doing any locking. Basically, + this function will not perform any locking, whereas attaching and detaching will, but crafted in + such a way that we don't need to perform any locking here. The important thing to remember is + to always iterate in a forward direction. + + In order to process any data we need to first read from all input buses. That's where this + function comes in. This iterates over each of the attachments and accumulates/mixes them. We + also convert the channels to the nodes output channel count before mixing. We want to do this + channel conversion so that the caller of this function can invoke the processing callback + without having to do it themselves. + + When we iterate over each of the attachments on the input bus, we need to read as much data as + we can from each of them so that we don't end up with holes between each of the attachments. To + do this, we need to read from each attachment in a loop and read as many frames as we can, up + to `frameCount`. + */ + MA_ASSERT(pInputNode != NULL); + MA_ASSERT(pFramesRead != NULL); /* pFramesRead is critical and must always be specified. On input it's undefined and on output it'll be set to the number of frames actually read. */ + + *pFramesRead = 0; /* Safety. */ + + inputChannels = ma_node_input_bus_get_channels(pInputBus); + + /* + We need to be careful with how we call ma_node_input_bus_first() and ma_node_input_bus_next(). They + are both critical to our lock-free thread-safety system. We can only call ma_node_input_bus_first() + once per iteration, however we have an optimization to checks whether or not it's the first item in + the list. We therefore need to store a pointer to the first item rather than repeatedly calling + ma_node_input_bus_first(). It's safe to keep hold of this pointer, so long as we don't dereference it + after calling ma_node_input_bus_next(), which we won't be. + */ + pFirst = ma_node_input_bus_first(pInputBus); + if (pFirst == NULL) { + return MA_SUCCESS; /* No attachments. Read nothing. */ + } + + for (pOutputBus = pFirst; pOutputBus != NULL; pOutputBus = ma_node_input_bus_next(pInputBus, pOutputBus)) { + ma_uint32 framesProcessed = 0; + ma_bool32 isSilentOutput = MA_FALSE; + + MA_ASSERT(pOutputBus->pNode != NULL); + MA_ASSERT(((ma_node_base*)pOutputBus->pNode)->vtable != NULL); + + isSilentOutput = (((ma_node_base*)pOutputBus->pNode)->vtable->flags & MA_NODE_FLAG_SILENT_OUTPUT) != 0; + + if (pFramesOut != NULL) { + /* Read. */ + while (framesProcessed < frameCount) { + float* pRunningFramesOut; + ma_uint32 framesToRead; + ma_uint32 framesJustRead = 0; + + framesToRead = frameCount - framesProcessed; + pRunningFramesOut = ma_offset_pcm_frames_ptr_f32(pFramesOut, framesProcessed, inputChannels); + + if (doesOutputBufferHaveContent == MA_FALSE) { + /* Fast path. First attachment. We just read straight into the output buffer (no mixing required). */ + result = ma_node_read_pcm_frames(pOutputBus->pNode, pOutputBus->outputBusIndex, pRunningFramesOut, framesToRead, &framesJustRead, globalTime + framesProcessed); + } else { + /* Slow path. Not the first attachment. Mixing required. */ + ma_uint32 preMixBufferCapInFrames = ((ma_node_base*)pInputNode)->cachedDataCapInFramesPerBus; + float* pPreMixBuffer = (float*)ma_stack_alloc(((ma_node_base*)pInputNode)->pNodeGraph->pPreMixStack, preMixBufferCapInFrames * inputChannels * sizeof(float)); + + if (pPreMixBuffer == NULL) { + /* + If you're hitting this assert it means you've got an unusually deep chain of nodes, you've got an excessively large processing + size, or you have a combination of both, and as a result have run out of stack space. You can increase this using the + preMixStackSizeInBytes variable in ma_node_graph_config. If you're using ma_engine, you can do it via the preMixStackSizeInBytes + variable in ma_engine_config. It defaults to 512KB per output channel. + */ + MA_ASSERT(MA_FALSE); + } else { + if (framesToRead > preMixBufferCapInFrames) { + framesToRead = preMixBufferCapInFrames; + } + + result = ma_node_read_pcm_frames(pOutputBus->pNode, pOutputBus->outputBusIndex, pPreMixBuffer, framesToRead, &framesJustRead, globalTime + framesProcessed); + if (result == MA_SUCCESS || result == MA_AT_END) { + if (isSilentOutput == MA_FALSE) { /* Don't mix if the node outputs silence. */ + ma_mix_pcm_frames_f32(pRunningFramesOut, pPreMixBuffer, framesJustRead, inputChannels, /*volume*/1); + } + } + + /* The pre-mix buffer is no longer required. */ + ma_stack_free(((ma_node_base*)pInputNode)->pNodeGraph->pPreMixStack, pPreMixBuffer); + pPreMixBuffer = NULL; + } + } + + framesProcessed += framesJustRead; + + /* If we reached the end or otherwise failed to read any data we need to finish up with this output node. */ + if (result != MA_SUCCESS) { + break; + } + + /* If we didn't read anything, abort so we don't get stuck in a loop. */ + if (framesJustRead == 0) { + break; + } + } + + /* If it's the first attachment we didn't do any mixing. Any leftover samples need to be silenced. */ + if (pOutputBus == pFirst && framesProcessed < frameCount) { + ma_silence_pcm_frames(ma_offset_pcm_frames_ptr(pFramesOut, framesProcessed, ma_format_f32, inputChannels), (frameCount - framesProcessed), ma_format_f32, inputChannels); + } + + if (isSilentOutput == MA_FALSE) { + doesOutputBufferHaveContent = MA_TRUE; + } + } else { + /* Seek. */ + ma_node_read_pcm_frames(pOutputBus->pNode, pOutputBus->outputBusIndex, NULL, frameCount, &framesProcessed, globalTime); + } + } + + /* If we didn't output anything, output silence. */ + if (doesOutputBufferHaveContent == MA_FALSE && pFramesOut != NULL) { + ma_silence_pcm_frames(pFramesOut, frameCount, ma_format_f32, inputChannels); + } + + /* In this path we always "process" the entire amount. */ + *pFramesRead = frameCount; + + return result; +} + + +MA_API ma_node_config ma_node_config_init(void) +{ + ma_node_config config; + + MA_ZERO_OBJECT(&config); + config.initialState = ma_node_state_started; /* Nodes are started by default. */ + config.inputBusCount = MA_NODE_BUS_COUNT_UNKNOWN; + config.outputBusCount = MA_NODE_BUS_COUNT_UNKNOWN; + + return config; +} + +static ma_uint16 ma_node_config_get_cache_size_in_frames(const ma_node_config* pConfig, const ma_node_graph* pNodeGraph) +{ + ma_uint32 cacheSizeInFrames; + + (void)pConfig; + + if (pNodeGraph->processingSizeInFrames > 0) { + cacheSizeInFrames = pNodeGraph->processingSizeInFrames; + } else { + cacheSizeInFrames = MA_DEFAULT_NODE_CACHE_CAP_IN_FRAMES_PER_BUS; + } + + if (cacheSizeInFrames > 0xFFFF) { + cacheSizeInFrames = 0xFFFF; + } + + return (ma_uint16)cacheSizeInFrames; +} + + + +static ma_result ma_node_detach_full(ma_node* pNode); + +static float* ma_node_get_cached_input_ptr(ma_node* pNode, ma_uint32 inputBusIndex) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + ma_uint32 iInputBus; + float* pBasePtr; + + MA_ASSERT(pNodeBase != NULL); + + /* Input data is stored at the front of the buffer. */ + pBasePtr = pNodeBase->pCachedData; + for (iInputBus = 0; iInputBus < inputBusIndex; iInputBus += 1) { + pBasePtr += pNodeBase->cachedDataCapInFramesPerBus * ma_node_input_bus_get_channels(&pNodeBase->pInputBuses[iInputBus]); + } + + return pBasePtr; +} + +static float* ma_node_get_cached_output_ptr(ma_node* pNode, ma_uint32 outputBusIndex) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + ma_uint32 iInputBus; + ma_uint32 iOutputBus; + float* pBasePtr; + + MA_ASSERT(pNodeBase != NULL); + + /* Cached output data starts after the input data. */ + pBasePtr = pNodeBase->pCachedData; + for (iInputBus = 0; iInputBus < ma_node_get_input_bus_count(pNodeBase); iInputBus += 1) { + pBasePtr += pNodeBase->cachedDataCapInFramesPerBus * ma_node_input_bus_get_channels(&pNodeBase->pInputBuses[iInputBus]); + } + + for (iOutputBus = 0; iOutputBus < outputBusIndex; iOutputBus += 1) { + pBasePtr += pNodeBase->cachedDataCapInFramesPerBus * ma_node_output_bus_get_channels(&pNodeBase->pOutputBuses[iOutputBus]); + } + + return pBasePtr; +} + + +typedef struct +{ + size_t sizeInBytes; + size_t inputBusOffset; + size_t outputBusOffset; + size_t cachedDataOffset; + ma_uint32 inputBusCount; /* So it doesn't have to be calculated twice. */ + ma_uint32 outputBusCount; /* So it doesn't have to be calculated twice. */ +} ma_node_heap_layout; + +static ma_result ma_node_translate_bus_counts(const ma_node_config* pConfig, ma_uint32* pInputBusCount, ma_uint32* pOutputBusCount) +{ + ma_uint32 inputBusCount; + ma_uint32 outputBusCount; + + MA_ASSERT(pConfig != NULL); + MA_ASSERT(pInputBusCount != NULL); + MA_ASSERT(pOutputBusCount != NULL); + + /* Bus counts are determined by the vtable, unless they're set to `MA_NODE_BUS_COUNT_UNKNWON`, in which case they're taken from the config. */ + if (pConfig->vtable->inputBusCount == MA_NODE_BUS_COUNT_UNKNOWN) { + inputBusCount = pConfig->inputBusCount; + } else { + inputBusCount = pConfig->vtable->inputBusCount; + + if (pConfig->inputBusCount != MA_NODE_BUS_COUNT_UNKNOWN && pConfig->inputBusCount != pConfig->vtable->inputBusCount) { + return MA_INVALID_ARGS; /* Invalid configuration. You must not specify a conflicting bus count between the node's config and the vtable. */ + } + } + + if (pConfig->vtable->outputBusCount == MA_NODE_BUS_COUNT_UNKNOWN) { + outputBusCount = pConfig->outputBusCount; + } else { + outputBusCount = pConfig->vtable->outputBusCount; + + if (pConfig->outputBusCount != MA_NODE_BUS_COUNT_UNKNOWN && pConfig->outputBusCount != pConfig->vtable->outputBusCount) { + return MA_INVALID_ARGS; /* Invalid configuration. You must not specify a conflicting bus count between the node's config and the vtable. */ + } + } + + /* Bus counts must be within limits. */ + if (inputBusCount > MA_MAX_NODE_BUS_COUNT || outputBusCount > MA_MAX_NODE_BUS_COUNT) { + return MA_INVALID_ARGS; + } + + + /* We must have channel counts for each bus. */ + if ((inputBusCount > 0 && pConfig->pInputChannels == NULL) || (outputBusCount > 0 && pConfig->pOutputChannels == NULL)) { + return MA_INVALID_ARGS; /* You must specify channel counts for each input and output bus. */ + } + + + /* Some special rules for passthrough nodes. */ + if ((pConfig->vtable->flags & MA_NODE_FLAG_PASSTHROUGH) != 0) { + if ((pConfig->vtable->inputBusCount != 0 && pConfig->vtable->inputBusCount != 1) || pConfig->vtable->outputBusCount != 1) { + return MA_INVALID_ARGS; /* Passthrough nodes must have exactly 1 output bus and either 0 or 1 input bus. */ + } + + if (pConfig->pInputChannels[0] != pConfig->pOutputChannels[0]) { + return MA_INVALID_ARGS; /* Passthrough nodes must have the same number of channels between input and output nodes. */ + } + } + + + *pInputBusCount = inputBusCount; + *pOutputBusCount = outputBusCount; + + return MA_SUCCESS; +} + +static ma_result ma_node_get_heap_layout(ma_node_graph* pNodeGraph, const ma_node_config* pConfig, ma_node_heap_layout* pHeapLayout) +{ + ma_result result; + ma_uint32 inputBusCount; + ma_uint32 outputBusCount; + + MA_ASSERT(pHeapLayout != NULL); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL || pConfig->vtable == NULL || pConfig->vtable->onProcess == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_node_translate_bus_counts(pConfig, &inputBusCount, &outputBusCount); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->sizeInBytes = 0; + + /* Input buses. */ + if (inputBusCount > MA_MAX_NODE_LOCAL_BUS_COUNT) { + pHeapLayout->inputBusOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(sizeof(ma_node_input_bus) * inputBusCount); + } else { + pHeapLayout->inputBusOffset = MA_SIZE_MAX; /* MA_SIZE_MAX indicates that no heap allocation is required for the input bus. */ + } + + /* Output buses. */ + if (outputBusCount > MA_MAX_NODE_LOCAL_BUS_COUNT) { + pHeapLayout->outputBusOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(sizeof(ma_node_output_bus) * outputBusCount); + } else { + pHeapLayout->outputBusOffset = MA_SIZE_MAX; + } + + /* + Cached audio data. + + We need to allocate memory for caching both input and output data. We have an optimization + where no caching is necessary for specific conditions: + + - The node has 0 inputs and 1 output. + + When a node meets the above conditions, no cache is allocated. + + The size choice for this buffer is a little bit finicky. We don't want to be too wasteful by + allocating too much, but at the same time we want it be large enough so that enough frames can + be processed for each call to ma_node_read_pcm_frames() so that it keeps things efficient. For + now I'm going with 10ms @ 48K which is 480 frames per bus. This is configurable at compile + time. It might also be worth investigating whether or not this can be configured at run time. + */ + if (inputBusCount == 0 && outputBusCount == 1) { + /* Fast path. No cache needed. */ + pHeapLayout->cachedDataOffset = MA_SIZE_MAX; + } else { + /* Slow path. Cache needed. */ + size_t cachedDataSizeInBytes = 0; + ma_uint32 cacheCapInFrames; + ma_uint32 iBus; + + /* The capacity of the cache is based on our callback processing size. */ + cacheCapInFrames = ma_node_config_get_cache_size_in_frames(pConfig, pNodeGraph); + + for (iBus = 0; iBus < inputBusCount; iBus += 1) { + cachedDataSizeInBytes += cacheCapInFrames * ma_get_bytes_per_frame(ma_format_f32, pConfig->pInputChannels[iBus]); + } + + for (iBus = 0; iBus < outputBusCount; iBus += 1) { + cachedDataSizeInBytes += cacheCapInFrames * ma_get_bytes_per_frame(ma_format_f32, pConfig->pOutputChannels[iBus]); + } + + pHeapLayout->cachedDataOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(cachedDataSizeInBytes); + } + + + /* + Not technically part of the heap, but we can output the input and output bus counts so we can + avoid a redundant call to ma_node_translate_bus_counts(). + */ + pHeapLayout->inputBusCount = inputBusCount; + pHeapLayout->outputBusCount = outputBusCount; + + /* Make sure allocation size is aligned. */ + pHeapLayout->sizeInBytes = ma_align_64(pHeapLayout->sizeInBytes); + + return MA_SUCCESS; +} + +MA_API ma_result ma_node_get_heap_size(ma_node_graph* pNodeGraph, const ma_node_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_node_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_node_get_heap_layout(pNodeGraph, pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_node_init_preallocated(ma_node_graph* pNodeGraph, const ma_node_config* pConfig, void* pHeap, ma_node* pNode) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + ma_result result; + ma_node_heap_layout heapLayout; + ma_uint32 iInputBus; + ma_uint32 iOutputBus; + + if (pNodeBase == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNodeBase); + + result = ma_node_get_heap_layout(pNodeGraph, pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + pNodeBase->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pNodeBase->pNodeGraph = pNodeGraph; + pNodeBase->vtable = pConfig->vtable; + pNodeBase->state = pConfig->initialState; + pNodeBase->stateTimes[ma_node_state_started] = 0; + pNodeBase->stateTimes[ma_node_state_stopped] = (ma_uint64)(ma_int64)-1; /* Weird casting for VC6 compatibility. */ + pNodeBase->inputBusCount = heapLayout.inputBusCount; + pNodeBase->outputBusCount = heapLayout.outputBusCount; + + if (heapLayout.inputBusOffset != MA_SIZE_MAX) { + pNodeBase->pInputBuses = (ma_node_input_bus*)ma_offset_ptr(pHeap, heapLayout.inputBusOffset); + } else { + pNodeBase->pInputBuses = pNodeBase->_inputBuses; + } + + if (heapLayout.outputBusOffset != MA_SIZE_MAX) { + pNodeBase->pOutputBuses = (ma_node_output_bus*)ma_offset_ptr(pHeap, heapLayout.outputBusOffset); + } else { + pNodeBase->pOutputBuses = pNodeBase->_outputBuses; + } + + if (heapLayout.cachedDataOffset != MA_SIZE_MAX) { + pNodeBase->pCachedData = (float*)ma_offset_ptr(pHeap, heapLayout.cachedDataOffset); + pNodeBase->cachedDataCapInFramesPerBus = ma_node_config_get_cache_size_in_frames(pConfig, pNodeGraph); + } else { + pNodeBase->pCachedData = NULL; + } + + + /* We need to run an initialization step for each input and output bus. */ + for (iInputBus = 0; iInputBus < ma_node_get_input_bus_count(pNodeBase); iInputBus += 1) { + result = ma_node_input_bus_init(pConfig->pInputChannels[iInputBus], &pNodeBase->pInputBuses[iInputBus]); + if (result != MA_SUCCESS) { + return result; + } + } + + for (iOutputBus = 0; iOutputBus < ma_node_get_output_bus_count(pNodeBase); iOutputBus += 1) { + result = ma_node_output_bus_init(pNodeBase, iOutputBus, pConfig->pOutputChannels[iOutputBus], &pNodeBase->pOutputBuses[iOutputBus]); + if (result != MA_SUCCESS) { + return result; + } + } + + + /* The cached data needs to be initialized to silence (or a sine wave tone if we're debugging). */ + if (pNodeBase->pCachedData != NULL) { + ma_uint32 iBus; + + #if 1 /* Toggle this between 0 and 1 to turn debugging on or off. 1 = fill with a sine wave for debugging; 0 = fill with silence. */ + /* For safety we'll go ahead and default the buffer to silence. */ + for (iBus = 0; iBus < ma_node_get_input_bus_count(pNodeBase); iBus += 1) { + ma_silence_pcm_frames(ma_node_get_cached_input_ptr(pNode, iBus), pNodeBase->cachedDataCapInFramesPerBus, ma_format_f32, ma_node_input_bus_get_channels(&pNodeBase->pInputBuses[iBus])); + } + for (iBus = 0; iBus < ma_node_get_output_bus_count(pNodeBase); iBus += 1) { + ma_silence_pcm_frames(ma_node_get_cached_output_ptr(pNode, iBus), pNodeBase->cachedDataCapInFramesPerBus, ma_format_f32, ma_node_output_bus_get_channels(&pNodeBase->pOutputBuses[iBus])); + } + #else + /* For debugging. Default to a sine wave. */ + for (iBus = 0; iBus < ma_node_get_input_bus_count(pNodeBase); iBus += 1) { + ma_debug_fill_pcm_frames_with_sine_wave(ma_node_get_cached_input_ptr(pNode, iBus), pNodeBase->cachedDataCapInFramesPerBus, ma_format_f32, ma_node_input_bus_get_channels(&pNodeBase->pInputBuses[iBus]), 48000); + } + for (iBus = 0; iBus < ma_node_get_output_bus_count(pNodeBase); iBus += 1) { + ma_debug_fill_pcm_frames_with_sine_wave(ma_node_get_cached_output_ptr(pNode, iBus), pNodeBase->cachedDataCapInFramesPerBus, ma_format_f32, ma_node_output_bus_get_channels(&pNodeBase->pOutputBuses[iBus]), 48000); + } + #endif + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_node_init(ma_node_graph* pNodeGraph, const ma_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_node* pNode) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_node_get_heap_size(pNodeGraph, pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_node_init_preallocated(pNodeGraph, pConfig, pHeap, pNode); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + ((ma_node_base*)pNode)->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_node_uninit(ma_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + + if (pNodeBase == NULL) { + return; + } + + /* + The first thing we need to do is fully detach the node. This will detach all inputs and + outputs. We need to do this first because it will sever the connection with the node graph and + allow us to complete uninitialization without needing to worry about thread-safety with the + audio thread. The detachment process will wait for any local processing of the node to finish. + */ + ma_node_detach_full(pNode); + + /* + At this point the node should be completely unreferenced by the node graph and we can finish up + the uninitialization process without needing to worry about thread-safety. + */ + if (pNodeBase->_ownsHeap) { + ma_free(pNodeBase->_pHeap, pAllocationCallbacks); + } +} + +MA_API ma_node_graph* ma_node_get_node_graph(const ma_node* pNode) +{ + if (pNode == NULL) { + return NULL; + } + + return ((const ma_node_base*)pNode)->pNodeGraph; +} + +MA_API ma_uint32 ma_node_get_input_bus_count(const ma_node* pNode) +{ + if (pNode == NULL) { + return 0; + } + + return ((ma_node_base*)pNode)->inputBusCount; +} + +MA_API ma_uint32 ma_node_get_output_bus_count(const ma_node* pNode) +{ + if (pNode == NULL) { + return 0; + } + + return ((ma_node_base*)pNode)->outputBusCount; +} + + +MA_API ma_uint32 ma_node_get_input_channels(const ma_node* pNode, ma_uint32 inputBusIndex) +{ + const ma_node_base* pNodeBase = (const ma_node_base*)pNode; + + if (pNode == NULL) { + return 0; + } + + if (inputBusIndex >= ma_node_get_input_bus_count(pNode)) { + return 0; /* Invalid bus index. */ + } + + return ma_node_input_bus_get_channels(&pNodeBase->pInputBuses[inputBusIndex]); +} + +MA_API ma_uint32 ma_node_get_output_channels(const ma_node* pNode, ma_uint32 outputBusIndex) +{ + const ma_node_base* pNodeBase = (const ma_node_base*)pNode; + + if (pNode == NULL) { + return 0; + } + + if (outputBusIndex >= ma_node_get_output_bus_count(pNode)) { + return 0; /* Invalid bus index. */ + } + + return ma_node_output_bus_get_channels(&pNodeBase->pOutputBuses[outputBusIndex]); +} + + +static ma_result ma_node_detach_full(ma_node* pNode) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + ma_uint32 iInputBus; + + if (pNodeBase == NULL) { + return MA_INVALID_ARGS; + } + + /* + Make sure the node is completely detached first. This will not return until the output bus is + guaranteed to no longer be referenced by the audio thread. + */ + ma_node_detach_all_output_buses(pNode); + + /* + At this point all output buses will have been detached from the graph and we can be guaranteed + that none of its input nodes will be getting processed by the graph. We can detach these + without needing to worry about the audio thread touching them. + */ + for (iInputBus = 0; iInputBus < ma_node_get_input_bus_count(pNode); iInputBus += 1) { + ma_node_input_bus* pInputBus; + ma_node_output_bus* pOutputBus; + + pInputBus = &pNodeBase->pInputBuses[iInputBus]; + + /* + This is important. We cannot be using ma_node_input_bus_first() or ma_node_input_bus_next(). Those + functions are specifically for the audio thread. We'll instead just manually iterate using standard + linked list logic. We don't need to worry about the audio thread referencing these because the step + above severed the connection to the graph. + */ + for (pOutputBus = (ma_node_output_bus*)ma_atomic_load_ptr(&pInputBus->head.pNext); pOutputBus != NULL; pOutputBus = (ma_node_output_bus*)ma_atomic_load_ptr(&pInputBus->head.pNext)) { + ma_node_detach_output_bus(pOutputBus->pNode, pOutputBus->outputBusIndex); /* This won't do any waiting in practice and should be efficient. */ + } + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_node_detach_output_bus(ma_node* pNode, ma_uint32 outputBusIndex) +{ + ma_result result = MA_SUCCESS; + ma_node_base* pNodeBase = (ma_node_base*)pNode; + ma_node_base* pInputNodeBase; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + if (outputBusIndex >= ma_node_get_output_bus_count(pNode)) { + return MA_INVALID_ARGS; /* Invalid output bus index. */ + } + + /* We need to lock the output bus because we need to inspect the input node and grab its input bus. */ + ma_node_output_bus_lock(&pNodeBase->pOutputBuses[outputBusIndex]); + { + pInputNodeBase = (ma_node_base*)pNodeBase->pOutputBuses[outputBusIndex].pInputNode; + if (pInputNodeBase != NULL) { + ma_node_input_bus_detach__no_output_bus_lock(&pInputNodeBase->pInputBuses[pNodeBase->pOutputBuses[outputBusIndex].inputNodeInputBusIndex], &pNodeBase->pOutputBuses[outputBusIndex]); + } + } + ma_node_output_bus_unlock(&pNodeBase->pOutputBuses[outputBusIndex]); + + return result; +} + +MA_API ma_result ma_node_detach_all_output_buses(ma_node* pNode) +{ + ma_uint32 iOutputBus; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + for (iOutputBus = 0; iOutputBus < ma_node_get_output_bus_count(pNode); iOutputBus += 1) { + ma_node_detach_output_bus(pNode, iOutputBus); + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_node_attach_output_bus(ma_node* pNode, ma_uint32 outputBusIndex, ma_node* pOtherNode, ma_uint32 otherNodeInputBusIndex) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + ma_node_base* pOtherNodeBase = (ma_node_base*)pOtherNode; + + if (pNodeBase == NULL || pOtherNodeBase == NULL) { + return MA_INVALID_ARGS; + } + + if (pNodeBase == pOtherNodeBase) { + return MA_INVALID_OPERATION; /* Cannot attach a node to itself. */ + } + + if (outputBusIndex >= ma_node_get_output_bus_count(pNode) || otherNodeInputBusIndex >= ma_node_get_input_bus_count(pOtherNode)) { + return MA_INVALID_OPERATION; /* Invalid bus index. */ + } + + /* The output channel count of the output node must be the same as the input channel count of the input node. */ + if (ma_node_get_output_channels(pNode, outputBusIndex) != ma_node_get_input_channels(pOtherNode, otherNodeInputBusIndex)) { + return MA_INVALID_OPERATION; /* Channel count is incompatible. */ + } + + /* This will deal with detaching if the output bus is already attached to something. */ + ma_node_input_bus_attach(&pOtherNodeBase->pInputBuses[otherNodeInputBusIndex], &pNodeBase->pOutputBuses[outputBusIndex], pOtherNode, otherNodeInputBusIndex); + + return MA_SUCCESS; +} + +MA_API ma_result ma_node_set_output_bus_volume(ma_node* pNode, ma_uint32 outputBusIndex, float volume) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + + if (pNodeBase == NULL) { + return MA_INVALID_ARGS; + } + + if (outputBusIndex >= ma_node_get_output_bus_count(pNode)) { + return MA_INVALID_ARGS; /* Invalid bus index. */ + } + + return ma_node_output_bus_set_volume(&pNodeBase->pOutputBuses[outputBusIndex], volume); +} + +MA_API float ma_node_get_output_bus_volume(const ma_node* pNode, ma_uint32 outputBusIndex) +{ + const ma_node_base* pNodeBase = (const ma_node_base*)pNode; + + if (pNodeBase == NULL) { + return 0; + } + + if (outputBusIndex >= ma_node_get_output_bus_count(pNode)) { + return 0; /* Invalid bus index. */ + } + + return ma_node_output_bus_get_volume(&pNodeBase->pOutputBuses[outputBusIndex]); +} + +MA_API ma_result ma_node_set_state(ma_node* pNode, ma_node_state state) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + + if (pNodeBase == NULL) { + return MA_INVALID_ARGS; + } + + ma_atomic_exchange_i32(&pNodeBase->state, state); + + return MA_SUCCESS; +} + +MA_API ma_node_state ma_node_get_state(const ma_node* pNode) +{ + const ma_node_base* pNodeBase = (const ma_node_base*)pNode; + + if (pNodeBase == NULL) { + return ma_node_state_stopped; + } + + return (ma_node_state)ma_atomic_load_i32(&pNodeBase->state); +} + +MA_API ma_result ma_node_set_state_time(ma_node* pNode, ma_node_state state, ma_uint64 globalTime) +{ + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + /* Validation check for safety since we'll be using this as an index into stateTimes[]. */ + if (state != ma_node_state_started && state != ma_node_state_stopped) { + return MA_INVALID_ARGS; + } + + ma_atomic_exchange_64(&((ma_node_base*)pNode)->stateTimes[state], globalTime); + + return MA_SUCCESS; +} + +MA_API ma_uint64 ma_node_get_state_time(const ma_node* pNode, ma_node_state state) +{ + if (pNode == NULL) { + return 0; + } + + /* Validation check for safety since we'll be using this as an index into stateTimes[]. */ + if (state != ma_node_state_started && state != ma_node_state_stopped) { + return 0; + } + + return ma_atomic_load_64(&((ma_node_base*)pNode)->stateTimes[state]); +} + +MA_API ma_node_state ma_node_get_state_by_time(const ma_node* pNode, ma_uint64 globalTime) +{ + if (pNode == NULL) { + return ma_node_state_stopped; + } + + return ma_node_get_state_by_time_range(pNode, globalTime, globalTime); +} + +MA_API ma_node_state ma_node_get_state_by_time_range(const ma_node* pNode, ma_uint64 globalTimeBeg, ma_uint64 globalTimeEnd) +{ + ma_node_state state; + + if (pNode == NULL) { + return ma_node_state_stopped; + } + + state = ma_node_get_state(pNode); + + /* An explicitly stopped node is always stopped. */ + if (state == ma_node_state_stopped) { + return ma_node_state_stopped; + } + + /* + Getting here means the node is marked as started, but it may still not be truly started due to + its start time not having been reached yet. Also, the stop time may have also been reached in + which case it'll be considered stopped. + */ + if (ma_node_get_state_time(pNode, ma_node_state_started) > globalTimeBeg) { + return ma_node_state_stopped; /* Start time has not yet been reached. */ + } + + if (ma_node_get_state_time(pNode, ma_node_state_stopped) <= globalTimeEnd) { + return ma_node_state_stopped; /* Stop time has been reached. */ + } + + /* Getting here means the node is marked as started and is within its start/stop times. */ + return ma_node_state_started; +} + +MA_API ma_uint64 ma_node_get_time(const ma_node* pNode) +{ + if (pNode == NULL) { + return 0; + } + + return ma_atomic_load_64(&((ma_node_base*)pNode)->localTime); +} + +MA_API ma_result ma_node_set_time(ma_node* pNode, ma_uint64 localTime) +{ + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + ma_atomic_exchange_64(&((ma_node_base*)pNode)->localTime, localTime); + + return MA_SUCCESS; +} + + + +static void ma_node_process_pcm_frames_internal(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + + MA_ASSERT(pNode != NULL); + + if (pNodeBase->vtable->onProcess) { + pNodeBase->vtable->onProcess(pNode, ppFramesIn, pFrameCountIn, ppFramesOut, pFrameCountOut); + } +} + +static ma_result ma_node_read_pcm_frames(ma_node* pNode, ma_uint32 outputBusIndex, float* pFramesOut, ma_uint32 frameCount, ma_uint32* pFramesRead, ma_uint64 globalTime) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + ma_result result = MA_SUCCESS; + ma_uint32 iInputBus; + ma_uint32 iOutputBus; + ma_uint32 inputBusCount; + ma_uint32 outputBusCount; + ma_uint32 totalFramesRead = 0; + float* ppFramesIn[MA_MAX_NODE_BUS_COUNT]; + float* ppFramesOut[MA_MAX_NODE_BUS_COUNT]; + ma_uint64 globalTimeBeg; + ma_uint64 globalTimeEnd; + ma_uint64 startTime; + ma_uint64 stopTime; + ma_uint32 timeOffsetBeg; + ma_uint32 timeOffsetEnd; + ma_uint32 frameCountIn; + ma_uint32 frameCountOut; + + /* + pFramesRead is mandatory. It must be used to determine how many frames were read. It's normal and + expected that the number of frames read may be different to that requested. Therefore, the caller + must look at this value to correctly determine how many frames were read. + */ + MA_ASSERT(pFramesRead != NULL); /* <-- If you've triggered this assert, you're using this function wrong. You *must* use this variable and inspect it after the call returns. */ + if (pFramesRead == NULL) { + return MA_INVALID_ARGS; + } + + *pFramesRead = 0; /* Safety. */ + + if (pNodeBase == NULL) { + return MA_INVALID_ARGS; + } + + if (outputBusIndex >= ma_node_get_output_bus_count(pNodeBase)) { + return MA_INVALID_ARGS; /* Invalid output bus index. */ + } + + /* Don't do anything if we're in a stopped state. */ + if (ma_node_get_state_by_time_range(pNode, globalTime, globalTime + frameCount) != ma_node_state_started) { + return MA_SUCCESS; /* We're in a stopped state. This is not an error - we just need to not read anything. */ + } + + + globalTimeBeg = globalTime; + globalTimeEnd = globalTime + frameCount; + startTime = ma_node_get_state_time(pNode, ma_node_state_started); + stopTime = ma_node_get_state_time(pNode, ma_node_state_stopped); + + /* + At this point we know that we are inside our start/stop times. However, we may need to adjust + our frame count and output pointer to accommodate since we could be straddling the time period + that this function is getting called for. + + It's possible (and likely) that the start time does not line up with the output buffer. We + therefore need to offset it by a number of frames to accommodate. The same thing applies for + the stop time. + */ + timeOffsetBeg = (globalTimeBeg < startTime) ? (ma_uint32)(globalTimeEnd - startTime) : 0; + timeOffsetEnd = (globalTimeEnd > stopTime) ? (ma_uint32)(globalTimeEnd - stopTime) : 0; + + /* Trim based on the start offset. We need to silence the start of the buffer. */ + if (timeOffsetBeg > 0) { + ma_silence_pcm_frames(pFramesOut, timeOffsetBeg, ma_format_f32, ma_node_get_output_channels(pNode, outputBusIndex)); + pFramesOut += timeOffsetBeg * ma_node_get_output_channels(pNode, outputBusIndex); + frameCount -= timeOffsetBeg; + } + + /* Trim based on the end offset. We don't need to silence the tail section because we'll just have a reduced value written to pFramesRead. */ + if (timeOffsetEnd > 0) { + frameCount -= timeOffsetEnd; + } + + + /* We run on different paths depending on the bus counts. */ + inputBusCount = ma_node_get_input_bus_count(pNode); + outputBusCount = ma_node_get_output_bus_count(pNode); + + /* + Run a simplified path when there are no inputs and one output. In this case there's nothing to + actually read and we can go straight to output. This is a very common scenario because the vast + majority of data source nodes will use this setup so this optimization I think is worthwhile. + */ + if (inputBusCount == 0 && outputBusCount == 1) { + /* Fast path. No need to read from input and no need for any caching. */ + frameCountIn = 0; + frameCountOut = frameCount; /* Just read as much as we can. The callback will return what was actually read. */ + + ppFramesOut[0] = pFramesOut; + + /* + If it's a passthrough we won't be expecting the callback to output anything, so we'll + need to pre-silence the output buffer. + */ + if ((pNodeBase->vtable->flags & MA_NODE_FLAG_PASSTHROUGH) != 0) { + ma_silence_pcm_frames(pFramesOut, frameCount, ma_format_f32, ma_node_get_output_channels(pNode, outputBusIndex)); + } + + ma_node_process_pcm_frames_internal(pNode, NULL, &frameCountIn, ppFramesOut, &frameCountOut); + totalFramesRead = frameCountOut; + } else { + /* Slow path. Need to read input data. */ + if ((pNodeBase->vtable->flags & MA_NODE_FLAG_PASSTHROUGH) != 0) { + /* + Fast path. We're running a passthrough. We need to read directly into the output buffer, but + still fire the callback so that event handling and trigger nodes can do their thing. Since + it's a passthrough there's no need for any kind of caching logic. + */ + MA_ASSERT(outputBusCount == inputBusCount); + MA_ASSERT(outputBusCount == 1); + MA_ASSERT(outputBusIndex == 0); + + /* We just read directly from input bus to output buffer, and then afterwards fire the callback. */ + ppFramesOut[0] = pFramesOut; + ppFramesIn[0] = ppFramesOut[0]; + + result = ma_node_input_bus_read_pcm_frames(pNodeBase, &pNodeBase->pInputBuses[0], ppFramesIn[0], frameCount, &totalFramesRead, globalTime); + if (result == MA_SUCCESS) { + /* Even though it's a passthrough, we still need to fire the callback. */ + frameCountIn = totalFramesRead; + frameCountOut = totalFramesRead; + + if (totalFramesRead > 0) { + ma_node_process_pcm_frames_internal(pNode, (const float**)ppFramesIn, &frameCountIn, ppFramesOut, &frameCountOut); /* From GCC: expected 'const float **' but argument is of type 'float **'. Shouldn't this be implicit? Explicit cast to silence the warning. */ + } + + /* + A passthrough should never have modified the input and output frame counts. If you're + triggering these asserts you need to fix your processing callback. + */ + MA_ASSERT(frameCountIn == totalFramesRead); + MA_ASSERT(frameCountOut == totalFramesRead); + } + } else { + /* Slow path. Need to do caching. */ + ma_uint32 framesToProcessIn; + ma_uint32 framesToProcessOut; + ma_bool32 consumeNullInput = MA_FALSE; + + /* + We use frameCount as a basis for the number of frames to read since that's what's being + requested, however we still need to clamp it to whatever can fit in the cache. + + This will also be used as the basis for determining how many input frames to read. This is + not ideal because it can result in too many input frames being read which introduces latency. + To solve this, nodes can implement an optional callback called onGetRequiredInputFrameCount + which is used as hint to miniaudio as to how many input frames it needs to read at a time. This + callback is completely optional, and if it's not set, miniaudio will assume `frameCount`. + + This function will be called multiple times for each period of time, once for each output node. + We cannot read from each input node each time this function is called. Instead we need to check + whether or not this is first output bus to be read from for this time period, and if so, read + from our input data. + + To determine whether or not we're ready to read data, we check a flag. There will be one flag + for each output. When the flag is set, it means data has been read previously and that we're + ready to advance time forward for our input nodes by reading fresh data. + */ + framesToProcessOut = frameCount; + if (framesToProcessOut > pNodeBase->cachedDataCapInFramesPerBus) { + framesToProcessOut = pNodeBase->cachedDataCapInFramesPerBus; + } + + framesToProcessIn = frameCount; + if (pNodeBase->vtable->onGetRequiredInputFrameCount) { + pNodeBase->vtable->onGetRequiredInputFrameCount(pNode, framesToProcessOut, &framesToProcessIn); /* <-- It does not matter if this fails. */ + } + if (framesToProcessIn > pNodeBase->cachedDataCapInFramesPerBus) { + framesToProcessIn = pNodeBase->cachedDataCapInFramesPerBus; + } + + + MA_ASSERT(framesToProcessIn <= 0xFFFF); + MA_ASSERT(framesToProcessOut <= 0xFFFF); + + if (ma_node_output_bus_has_read(&pNodeBase->pOutputBuses[outputBusIndex])) { + /* Getting here means we need to do another round of processing. */ + pNodeBase->cachedFrameCountOut = 0; + + for (;;) { + frameCountOut = 0; + + /* + We need to prepare our output frame pointers for processing. In the same iteration we need + to mark every output bus as unread so that future calls to this function for different buses + for the current time period don't pull in data when they should instead be reading from cache. + */ + for (iOutputBus = 0; iOutputBus < outputBusCount; iOutputBus += 1) { + ma_node_output_bus_set_has_read(&pNodeBase->pOutputBuses[iOutputBus], MA_FALSE); /* <-- This is what tells the next calls to this function for other output buses for this time period to read from cache instead of pulling in more data. */ + ppFramesOut[iOutputBus] = ma_node_get_cached_output_ptr(pNode, iOutputBus); + } + + /* We only need to read from input buses if there isn't already some data in the cache. */ + if (pNodeBase->cachedFrameCountIn == 0) { + ma_uint32 maxFramesReadIn = 0; + + /* Here is where we pull in data from the input buses. This is what will trigger an advance in time. */ + for (iInputBus = 0; iInputBus < inputBusCount; iInputBus += 1) { + ma_uint32 framesRead; + + /* The first thing to do is get the offset within our bulk allocation to store this input data. */ + ppFramesIn[iInputBus] = ma_node_get_cached_input_ptr(pNode, iInputBus); + + /* Once we've determined our destination pointer we can read. Note that we must inspect the number of frames read and fill any leftovers with silence for safety. */ + result = ma_node_input_bus_read_pcm_frames(pNodeBase, &pNodeBase->pInputBuses[iInputBus], ppFramesIn[iInputBus], framesToProcessIn, &framesRead, globalTime); + if (result != MA_SUCCESS) { + /* It doesn't really matter if we fail because we'll just fill with silence. */ + framesRead = 0; /* Just for safety, but I don't think it's really needed. */ + } + + /* TODO: Minor optimization opportunity here. If no frames were read and the buffer is already filled with silence, no need to re-silence it. */ + /* Any leftover frames need to silenced for safety. */ + if (framesRead < framesToProcessIn) { + ma_silence_pcm_frames(ppFramesIn[iInputBus] + (framesRead * ma_node_get_input_channels(pNodeBase, iInputBus)), (framesToProcessIn - framesRead), ma_format_f32, ma_node_get_input_channels(pNodeBase, iInputBus)); + } + + maxFramesReadIn = ma_max(maxFramesReadIn, framesRead); + } + + /* This was a fresh load of input data so reset our consumption counter. */ + pNodeBase->consumedFrameCountIn = 0; + + /* + We don't want to keep processing if there's nothing to process, so set the number of cached + input frames to the maximum number we read from each attachment (the lesser will be padded + with silence). If we didn't read anything, this will be set to 0 and the entire buffer will + have been assigned to silence. This being equal to 0 is an important property for us because + it allows us to detect when NULL can be passed into the processing callback for the input + buffer for the purpose of continuous processing. + */ + pNodeBase->cachedFrameCountIn = (ma_uint16)maxFramesReadIn; + } else { + /* We don't need to read anything, but we do need to prepare our input frame pointers. */ + for (iInputBus = 0; iInputBus < inputBusCount; iInputBus += 1) { + ppFramesIn[iInputBus] = ma_node_get_cached_input_ptr(pNode, iInputBus) + (pNodeBase->consumedFrameCountIn * ma_node_get_input_channels(pNodeBase, iInputBus)); + } + } + + /* + At this point we have our input data so now we need to do some processing. Sneaky little + optimization here - we can set the pointer to the output buffer for this output bus so + that the final copy into the output buffer is done directly by onProcess(). + */ + if (pFramesOut != NULL) { + ppFramesOut[outputBusIndex] = ma_offset_pcm_frames_ptr_f32(pFramesOut, pNodeBase->cachedFrameCountOut, ma_node_get_output_channels(pNode, outputBusIndex)); + } + + + /* Give the processing function the entire capacity of the output buffer. */ + frameCountOut = (framesToProcessOut - pNodeBase->cachedFrameCountOut); + + /* + We need to treat nodes with continuous processing a little differently. For these ones, + we always want to fire the callback with the requested number of frames, regardless of + pNodeBase->cachedFrameCountIn, which could be 0. Also, we want to check if we can pass + in NULL for the input buffer to the callback. + */ + if ((pNodeBase->vtable->flags & MA_NODE_FLAG_CONTINUOUS_PROCESSING) != 0) { + /* We're using continuous processing. Make sure we specify the whole frame count at all times. */ + frameCountIn = framesToProcessIn; /* Give the processing function as much input data as we've got in the buffer, including any silenced padding from short reads. */ + + if ((pNodeBase->vtable->flags & MA_NODE_FLAG_ALLOW_NULL_INPUT) != 0 && pNodeBase->consumedFrameCountIn == 0 && pNodeBase->cachedFrameCountIn == 0) { + consumeNullInput = MA_TRUE; + } else { + consumeNullInput = MA_FALSE; + } + + /* + Since we're using continuous processing we're always passing in a full frame count + regardless of how much input data was read. If this is greater than what we read as + input, we'll end up with an underflow. We instead need to make sure our cached frame + count is set to the number of frames we'll be passing to the data callback. Not + doing this will result in an underflow when we "consume" the cached data later on. + + Note that this check needs to be done after the "consumeNullInput" check above because + we use the property of cachedFrameCountIn being 0 to determine whether or not we + should be passing in a null pointer to the processing callback for when the node is + configured with MA_NODE_FLAG_ALLOW_NULL_INPUT. + */ + if (pNodeBase->cachedFrameCountIn < (ma_uint16)frameCountIn) { + pNodeBase->cachedFrameCountIn = (ma_uint16)frameCountIn; + } + } else { + frameCountIn = pNodeBase->cachedFrameCountIn; /* Give the processing function as much valid input data as we've got. */ + consumeNullInput = MA_FALSE; + } + + /* + Process data slightly differently depending on whether or not we're consuming NULL + input (checked just above). + */ + if (consumeNullInput) { + ma_node_process_pcm_frames_internal(pNode, NULL, &frameCountIn, ppFramesOut, &frameCountOut); + } else { + /* + We want to skip processing if there's no input data, but we can only do that safely if + we know that there is no chance of any output frames being produced. If continuous + processing is being used, this won't be a problem because the input frame count will + always be non-0. However, if continuous processing is *not* enabled and input and output + data is processed at different rates, we still need to process that last input frame + because there could be a few excess output frames needing to be produced from cached + data. The `MA_NODE_FLAG_DIFFERENT_PROCESSING_RATES` flag is used as the indicator for + determining whether or not we need to process the node even when there are no input + frames available right now. + */ + if (frameCountIn > 0 || (pNodeBase->vtable->flags & MA_NODE_FLAG_DIFFERENT_PROCESSING_RATES) != 0) { + ma_node_process_pcm_frames_internal(pNode, (const float**)ppFramesIn, &frameCountIn, ppFramesOut, &frameCountOut); /* From GCC: expected 'const float **' but argument is of type 'float **'. Shouldn't this be implicit? Explicit cast to silence the warning. */ + } else { + frameCountOut = 0; /* No data was processed. */ + } + } + + /* + Thanks to our sneaky optimization above we don't need to do any data copying directly into + the output buffer - the onProcess() callback just did that for us. We do, however, need to + apply the number of input and output frames that were processed. Note that due to continuous + processing above, we need to do explicit checks here. If we just consumed a NULL input + buffer it means that no actual input data was processed from the internal buffers and we + don't want to be modifying any counters. + */ + if (consumeNullInput == MA_FALSE) { + pNodeBase->consumedFrameCountIn += (ma_uint16)frameCountIn; + pNodeBase->cachedFrameCountIn -= (ma_uint16)frameCountIn; + } + + /* The cached output frame count is always equal to what we just read. */ + pNodeBase->cachedFrameCountOut += (ma_uint16)frameCountOut; + + /* If we couldn't process any data, we're done. The loop needs to be terminated here or else we'll get stuck in a loop. */ + if (pNodeBase->cachedFrameCountOut == framesToProcessOut || (frameCountOut == 0 && frameCountIn == 0)) { + break; + } + } + } else { + /* + We're not needing to read anything from the input buffer so just read directly from our + already-processed data. + */ + if (pFramesOut != NULL) { + ma_copy_pcm_frames(pFramesOut, ma_node_get_cached_output_ptr(pNodeBase, outputBusIndex), pNodeBase->cachedFrameCountOut, ma_format_f32, ma_node_get_output_channels(pNodeBase, outputBusIndex)); + } + } + + /* The number of frames read is always equal to the number of cached output frames. */ + totalFramesRead = pNodeBase->cachedFrameCountOut; + + /* Now that we've read the data, make sure our read flag is set. */ + ma_node_output_bus_set_has_read(&pNodeBase->pOutputBuses[outputBusIndex], MA_TRUE); + } + } + + /* Apply volume, if necessary. */ + ma_apply_volume_factor_f32(pFramesOut, totalFramesRead * ma_node_get_output_channels(pNodeBase, outputBusIndex), ma_node_output_bus_get_volume(&pNodeBase->pOutputBuses[outputBusIndex])); + + /* Advance our local time forward. */ + ma_atomic_fetch_add_64(&pNodeBase->localTime, (ma_uint64)totalFramesRead); + + *pFramesRead = totalFramesRead + timeOffsetBeg; /* Must include the silenced section at the start of the buffer. */ + return result; +} + + + + +/* Data source node. */ +MA_API ma_data_source_node_config ma_data_source_node_config_init(ma_data_source* pDataSource) +{ + ma_data_source_node_config config; + + MA_ZERO_OBJECT(&config); + config.nodeConfig = ma_node_config_init(); + config.pDataSource = pDataSource; + + return config; +} + + +static void ma_data_source_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_data_source_node* pDataSourceNode = (ma_data_source_node*)pNode; + ma_format format; + ma_uint32 channels; + ma_uint32 frameCount; + ma_uint64 framesRead = 0; + + MA_ASSERT(pDataSourceNode != NULL); + MA_ASSERT(pDataSourceNode->pDataSource != NULL); + MA_ASSERT(ma_node_get_input_bus_count(pDataSourceNode) == 0); + MA_ASSERT(ma_node_get_output_bus_count(pDataSourceNode) == 1); + + /* We don't want to read from ppFramesIn at all. Instead we read from the data source. */ + (void)ppFramesIn; + (void)pFrameCountIn; + + frameCount = *pFrameCountOut; + + /* miniaudio should never be calling this with a frame count of zero. */ + MA_ASSERT(frameCount > 0); + + if (ma_data_source_get_data_format(pDataSourceNode->pDataSource, &format, &channels, NULL, NULL, 0) == MA_SUCCESS) { /* <-- Don't care about sample rate here. */ + /* The node graph system requires samples be in floating point format. This is checked in ma_data_source_node_init(). */ + MA_ASSERT(format == ma_format_f32); + (void)format; /* Just to silence some static analysis tools. */ + + ma_data_source_read_pcm_frames(pDataSourceNode->pDataSource, ppFramesOut[0], frameCount, &framesRead); + } + + *pFrameCountOut = (ma_uint32)framesRead; +} + +static ma_node_vtable g_ma_data_source_node_vtable = +{ + ma_data_source_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 0, /* 0 input buses. */ + 1, /* 1 output bus. */ + 0 +}; + +MA_API ma_result ma_data_source_node_init(ma_node_graph* pNodeGraph, const ma_data_source_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_data_source_node* pDataSourceNode) +{ + ma_result result; + ma_format format; /* For validating the format, which must be ma_format_f32. */ + ma_uint32 channels; /* For specifying the channel count of the output bus. */ + ma_node_config baseConfig; + + if (pDataSourceNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDataSourceNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_data_source_get_data_format(pConfig->pDataSource, &format, &channels, NULL, NULL, 0); /* Don't care about sample rate. This will check pDataSource for NULL. */ + if (result != MA_SUCCESS) { + return result; + } + + MA_ASSERT(format == ma_format_f32); /* <-- If you've triggered this it means your data source is not outputting floating-point samples. You must configure your data source to use ma_format_f32. */ + if (format != ma_format_f32) { + return MA_INVALID_ARGS; /* Invalid format. */ + } + + /* The channel count is defined by the data source. If the caller has manually changed the channels we just ignore it. */ + baseConfig = pConfig->nodeConfig; + baseConfig.vtable = &g_ma_data_source_node_vtable; /* Explicitly set the vtable here to prevent callers from setting it incorrectly. */ + + /* + The channel count is defined by the data source. It is invalid for the caller to manually set + the channel counts in the config. `ma_data_source_node_config_init()` will have defaulted the + channel count pointer to NULL which is how it must remain. If you trigger any of these asserts + it means you're explicitly setting the channel count. Instead, configure the output channel + count of your data source to be the necessary channel count. + */ + if (baseConfig.pOutputChannels != NULL) { + return MA_INVALID_ARGS; + } + + baseConfig.pOutputChannels = &channels; + + result = ma_node_init(pNodeGraph, &baseConfig, pAllocationCallbacks, &pDataSourceNode->base); + if (result != MA_SUCCESS) { + return result; + } + + pDataSourceNode->pDataSource = pConfig->pDataSource; + + return MA_SUCCESS; +} + +MA_API void ma_data_source_node_uninit(ma_data_source_node* pDataSourceNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_node_uninit(&pDataSourceNode->base, pAllocationCallbacks); +} + +MA_API ma_result ma_data_source_node_set_looping(ma_data_source_node* pDataSourceNode, ma_bool32 isLooping) +{ + if (pDataSourceNode == NULL) { + return MA_INVALID_ARGS; + } + + return ma_data_source_set_looping(pDataSourceNode->pDataSource, isLooping); +} + +MA_API ma_bool32 ma_data_source_node_is_looping(ma_data_source_node* pDataSourceNode) +{ + if (pDataSourceNode == NULL) { + return MA_FALSE; + } + + return ma_data_source_is_looping(pDataSourceNode->pDataSource); +} + + + +/* Splitter Node. */ +MA_API ma_splitter_node_config ma_splitter_node_config_init(ma_uint32 channels) +{ + ma_splitter_node_config config; + + MA_ZERO_OBJECT(&config); + config.nodeConfig = ma_node_config_init(); + config.channels = channels; + config.outputBusCount = 2; + + return config; +} + + +static void ma_splitter_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_node_base* pNodeBase = (ma_node_base*)pNode; + ma_uint32 iOutputBus; + ma_uint32 channels; + + MA_ASSERT(pNodeBase != NULL); + MA_ASSERT(ma_node_get_input_bus_count(pNodeBase) == 1); + + /* We don't need to consider the input frame count - it'll be the same as the output frame count and we process everything. */ + (void)pFrameCountIn; + + /* NOTE: This assumes the same number of channels for all inputs and outputs. This was checked in ma_splitter_node_init(). */ + channels = ma_node_get_input_channels(pNodeBase, 0); + + /* Splitting is just copying the first input bus and copying it over to each output bus. */ + for (iOutputBus = 0; iOutputBus < ma_node_get_output_bus_count(pNodeBase); iOutputBus += 1) { + ma_copy_pcm_frames(ppFramesOut[iOutputBus], ppFramesIn[0], *pFrameCountOut, ma_format_f32, channels); + } +} + +static ma_node_vtable g_ma_splitter_node_vtable = +{ + ma_splitter_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* 1 input bus. */ + MA_NODE_BUS_COUNT_UNKNOWN, /* The output bus count is specified on a per-node basis. */ + 0 +}; + +MA_API ma_result ma_splitter_node_init(ma_node_graph* pNodeGraph, const ma_splitter_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_splitter_node* pSplitterNode) +{ + ma_result result; + ma_node_config baseConfig; + ma_uint32 pInputChannels[1]; + ma_uint32 pOutputChannels[MA_MAX_NODE_BUS_COUNT]; + ma_uint32 iOutputBus; + + if (pSplitterNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pSplitterNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->outputBusCount > MA_MAX_NODE_BUS_COUNT) { + return MA_INVALID_ARGS; /* Too many output buses. */ + } + + /* Splitters require the same number of channels between inputs and outputs. */ + pInputChannels[0] = pConfig->channels; + for (iOutputBus = 0; iOutputBus < pConfig->outputBusCount; iOutputBus += 1) { + pOutputChannels[iOutputBus] = pConfig->channels; + } + + baseConfig = pConfig->nodeConfig; + baseConfig.vtable = &g_ma_splitter_node_vtable; + baseConfig.pInputChannels = pInputChannels; + baseConfig.pOutputChannels = pOutputChannels; + baseConfig.outputBusCount = pConfig->outputBusCount; + + result = ma_node_init(pNodeGraph, &baseConfig, pAllocationCallbacks, &pSplitterNode->base); + if (result != MA_SUCCESS) { + return result; /* Failed to initialize the base node. */ + } + + return MA_SUCCESS; +} + +MA_API void ma_splitter_node_uninit(ma_splitter_node* pSplitterNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_node_uninit(pSplitterNode, pAllocationCallbacks); +} + + +/* +Biquad Node +*/ +MA_API ma_biquad_node_config ma_biquad_node_config_init(ma_uint32 channels, float b0, float b1, float b2, float a0, float a1, float a2) +{ + ma_biquad_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.biquad = ma_biquad_config_init(ma_format_f32, channels, b0, b1, b2, a0, a1, a2); + + return config; +} + +static void ma_biquad_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_biquad_node* pLPFNode = (ma_biquad_node*)pNode; + + MA_ASSERT(pNode != NULL); + (void)pFrameCountIn; + + ma_biquad_process_pcm_frames(&pLPFNode->biquad, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_biquad_node_vtable = +{ + ma_biquad_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* One input. */ + 1, /* One output. */ + 0 /* Default flags. */ +}; + +MA_API ma_result ma_biquad_node_init(ma_node_graph* pNodeGraph, const ma_biquad_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_biquad_node* pNode) +{ + ma_result result; + ma_node_config baseNodeConfig; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->biquad.format != ma_format_f32) { + return MA_INVALID_ARGS; /* The format must be f32. */ + } + + result = ma_biquad_init(&pConfig->biquad, pAllocationCallbacks, &pNode->biquad); + if (result != MA_SUCCESS) { + return result; + } + + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_biquad_node_vtable; + baseNodeConfig.pInputChannels = &pConfig->biquad.channels; + baseNodeConfig.pOutputChannels = &pConfig->biquad.channels; + + result = ma_node_init(pNodeGraph, &baseNodeConfig, pAllocationCallbacks, pNode); + if (result != MA_SUCCESS) { + return result; + } + + return result; +} + +MA_API ma_result ma_biquad_node_reinit(const ma_biquad_config* pConfig, ma_biquad_node* pNode) +{ + ma_biquad_node* pLPFNode = (ma_biquad_node*)pNode; + + MA_ASSERT(pNode != NULL); + + return ma_biquad_reinit(pConfig, &pLPFNode->biquad); +} + +MA_API void ma_biquad_node_uninit(ma_biquad_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_biquad_node* pLPFNode = (ma_biquad_node*)pNode; + + if (pNode == NULL) { + return; + } + + ma_node_uninit(pNode, pAllocationCallbacks); + ma_biquad_uninit(&pLPFNode->biquad, pAllocationCallbacks); +} + + + +/* +Low Pass Filter Node +*/ +MA_API ma_lpf_node_config ma_lpf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order) +{ + ma_lpf_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.lpf = ma_lpf_config_init(ma_format_f32, channels, sampleRate, cutoffFrequency, order); + + return config; +} + +static void ma_lpf_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_lpf_node* pLPFNode = (ma_lpf_node*)pNode; + + MA_ASSERT(pNode != NULL); + (void)pFrameCountIn; + + ma_lpf_process_pcm_frames(&pLPFNode->lpf, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_lpf_node_vtable = +{ + ma_lpf_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* One input. */ + 1, /* One output. */ + 0 /* Default flags. */ +}; + +MA_API ma_result ma_lpf_node_init(ma_node_graph* pNodeGraph, const ma_lpf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_lpf_node* pNode) +{ + ma_result result; + ma_node_config baseNodeConfig; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->lpf.format != ma_format_f32) { + return MA_INVALID_ARGS; /* The format must be f32. */ + } + + result = ma_lpf_init(&pConfig->lpf, pAllocationCallbacks, &pNode->lpf); + if (result != MA_SUCCESS) { + return result; + } + + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_lpf_node_vtable; + baseNodeConfig.pInputChannels = &pConfig->lpf.channels; + baseNodeConfig.pOutputChannels = &pConfig->lpf.channels; + + result = ma_node_init(pNodeGraph, &baseNodeConfig, pAllocationCallbacks, pNode); + if (result != MA_SUCCESS) { + return result; + } + + return result; +} + +MA_API ma_result ma_lpf_node_reinit(const ma_lpf_config* pConfig, ma_lpf_node* pNode) +{ + ma_lpf_node* pLPFNode = (ma_lpf_node*)pNode; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + return ma_lpf_reinit(pConfig, &pLPFNode->lpf); +} + +MA_API void ma_lpf_node_uninit(ma_lpf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_lpf_node* pLPFNode = (ma_lpf_node*)pNode; + + if (pNode == NULL) { + return; + } + + ma_node_uninit(pNode, pAllocationCallbacks); + ma_lpf_uninit(&pLPFNode->lpf, pAllocationCallbacks); +} + + + +/* +High Pass Filter Node +*/ +MA_API ma_hpf_node_config ma_hpf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order) +{ + ma_hpf_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.hpf = ma_hpf_config_init(ma_format_f32, channels, sampleRate, cutoffFrequency, order); + + return config; +} + +static void ma_hpf_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_hpf_node* pHPFNode = (ma_hpf_node*)pNode; + + MA_ASSERT(pNode != NULL); + (void)pFrameCountIn; + + ma_hpf_process_pcm_frames(&pHPFNode->hpf, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_hpf_node_vtable = +{ + ma_hpf_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* One input. */ + 1, /* One output. */ + 0 /* Default flags. */ +}; + +MA_API ma_result ma_hpf_node_init(ma_node_graph* pNodeGraph, const ma_hpf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hpf_node* pNode) +{ + ma_result result; + ma_node_config baseNodeConfig; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->hpf.format != ma_format_f32) { + return MA_INVALID_ARGS; /* The format must be f32. */ + } + + result = ma_hpf_init(&pConfig->hpf, pAllocationCallbacks, &pNode->hpf); + if (result != MA_SUCCESS) { + return result; + } + + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_hpf_node_vtable; + baseNodeConfig.pInputChannels = &pConfig->hpf.channels; + baseNodeConfig.pOutputChannels = &pConfig->hpf.channels; + + result = ma_node_init(pNodeGraph, &baseNodeConfig, pAllocationCallbacks, pNode); + if (result != MA_SUCCESS) { + return result; + } + + return result; +} + +MA_API ma_result ma_hpf_node_reinit(const ma_hpf_config* pConfig, ma_hpf_node* pNode) +{ + ma_hpf_node* pHPFNode = (ma_hpf_node*)pNode; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + return ma_hpf_reinit(pConfig, &pHPFNode->hpf); +} + +MA_API void ma_hpf_node_uninit(ma_hpf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_hpf_node* pHPFNode = (ma_hpf_node*)pNode; + + if (pNode == NULL) { + return; + } + + ma_node_uninit(pNode, pAllocationCallbacks); + ma_hpf_uninit(&pHPFNode->hpf, pAllocationCallbacks); +} + + + + +/* +Band Pass Filter Node +*/ +MA_API ma_bpf_node_config ma_bpf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double cutoffFrequency, ma_uint32 order) +{ + ma_bpf_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.bpf = ma_bpf_config_init(ma_format_f32, channels, sampleRate, cutoffFrequency, order); + + return config; +} + +static void ma_bpf_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_bpf_node* pBPFNode = (ma_bpf_node*)pNode; + + MA_ASSERT(pNode != NULL); + (void)pFrameCountIn; + + ma_bpf_process_pcm_frames(&pBPFNode->bpf, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_bpf_node_vtable = +{ + ma_bpf_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* One input. */ + 1, /* One output. */ + 0 /* Default flags. */ +}; + +MA_API ma_result ma_bpf_node_init(ma_node_graph* pNodeGraph, const ma_bpf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_bpf_node* pNode) +{ + ma_result result; + ma_node_config baseNodeConfig; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->bpf.format != ma_format_f32) { + return MA_INVALID_ARGS; /* The format must be f32. */ + } + + result = ma_bpf_init(&pConfig->bpf, pAllocationCallbacks, &pNode->bpf); + if (result != MA_SUCCESS) { + return result; + } + + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_bpf_node_vtable; + baseNodeConfig.pInputChannels = &pConfig->bpf.channels; + baseNodeConfig.pOutputChannels = &pConfig->bpf.channels; + + result = ma_node_init(pNodeGraph, &baseNodeConfig, pAllocationCallbacks, pNode); + if (result != MA_SUCCESS) { + return result; + } + + return result; +} + +MA_API ma_result ma_bpf_node_reinit(const ma_bpf_config* pConfig, ma_bpf_node* pNode) +{ + ma_bpf_node* pBPFNode = (ma_bpf_node*)pNode; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + return ma_bpf_reinit(pConfig, &pBPFNode->bpf); +} + +MA_API void ma_bpf_node_uninit(ma_bpf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_bpf_node* pBPFNode = (ma_bpf_node*)pNode; + + if (pNode == NULL) { + return; + } + + ma_node_uninit(pNode, pAllocationCallbacks); + ma_bpf_uninit(&pBPFNode->bpf, pAllocationCallbacks); +} + + + +/* +Notching Filter Node +*/ +MA_API ma_notch_node_config ma_notch_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double q, double frequency) +{ + ma_notch_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.notch = ma_notch2_config_init(ma_format_f32, channels, sampleRate, q, frequency); + + return config; +} + +static void ma_notch_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_notch_node* pBPFNode = (ma_notch_node*)pNode; + + MA_ASSERT(pNode != NULL); + (void)pFrameCountIn; + + ma_notch2_process_pcm_frames(&pBPFNode->notch, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_notch_node_vtable = +{ + ma_notch_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* One input. */ + 1, /* One output. */ + 0 /* Default flags. */ +}; + +MA_API ma_result ma_notch_node_init(ma_node_graph* pNodeGraph, const ma_notch_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_notch_node* pNode) +{ + ma_result result; + ma_node_config baseNodeConfig; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->notch.format != ma_format_f32) { + return MA_INVALID_ARGS; /* The format must be f32. */ + } + + result = ma_notch2_init(&pConfig->notch, pAllocationCallbacks, &pNode->notch); + if (result != MA_SUCCESS) { + return result; + } + + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_notch_node_vtable; + baseNodeConfig.pInputChannels = &pConfig->notch.channels; + baseNodeConfig.pOutputChannels = &pConfig->notch.channels; + + result = ma_node_init(pNodeGraph, &baseNodeConfig, pAllocationCallbacks, pNode); + if (result != MA_SUCCESS) { + return result; + } + + return result; +} + +MA_API ma_result ma_notch_node_reinit(const ma_notch_config* pConfig, ma_notch_node* pNode) +{ + ma_notch_node* pNotchNode = (ma_notch_node*)pNode; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + return ma_notch2_reinit(pConfig, &pNotchNode->notch); +} + +MA_API void ma_notch_node_uninit(ma_notch_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_notch_node* pNotchNode = (ma_notch_node*)pNode; + + if (pNode == NULL) { + return; + } + + ma_node_uninit(pNode, pAllocationCallbacks); + ma_notch2_uninit(&pNotchNode->notch, pAllocationCallbacks); +} + + + +/* +Peaking Filter Node +*/ +MA_API ma_peak_node_config ma_peak_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double q, double frequency) +{ + ma_peak_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.peak = ma_peak2_config_init(ma_format_f32, channels, sampleRate, gainDB, q, frequency); + + return config; +} + +static void ma_peak_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_peak_node* pBPFNode = (ma_peak_node*)pNode; + + MA_ASSERT(pNode != NULL); + (void)pFrameCountIn; + + ma_peak2_process_pcm_frames(&pBPFNode->peak, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_peak_node_vtable = +{ + ma_peak_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* One input. */ + 1, /* One output. */ + 0 /* Default flags. */ +}; + +MA_API ma_result ma_peak_node_init(ma_node_graph* pNodeGraph, const ma_peak_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_peak_node* pNode) +{ + ma_result result; + ma_node_config baseNodeConfig; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->peak.format != ma_format_f32) { + return MA_INVALID_ARGS; /* The format must be f32. */ + } + + result = ma_peak2_init(&pConfig->peak, pAllocationCallbacks, &pNode->peak); + if (result != MA_SUCCESS) { + ma_node_uninit(pNode, pAllocationCallbacks); + return result; + } + + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_peak_node_vtable; + baseNodeConfig.pInputChannels = &pConfig->peak.channels; + baseNodeConfig.pOutputChannels = &pConfig->peak.channels; + + result = ma_node_init(pNodeGraph, &baseNodeConfig, pAllocationCallbacks, pNode); + if (result != MA_SUCCESS) { + return result; + } + + return result; +} + +MA_API ma_result ma_peak_node_reinit(const ma_peak_config* pConfig, ma_peak_node* pNode) +{ + ma_peak_node* pPeakNode = (ma_peak_node*)pNode; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + return ma_peak2_reinit(pConfig, &pPeakNode->peak); +} + +MA_API void ma_peak_node_uninit(ma_peak_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_peak_node* pPeakNode = (ma_peak_node*)pNode; + + if (pNode == NULL) { + return; + } + + ma_node_uninit(pNode, pAllocationCallbacks); + ma_peak2_uninit(&pPeakNode->peak, pAllocationCallbacks); +} + + + +/* +Low Shelf Filter Node +*/ +MA_API ma_loshelf_node_config ma_loshelf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double q, double frequency) +{ + ma_loshelf_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.loshelf = ma_loshelf2_config_init(ma_format_f32, channels, sampleRate, gainDB, q, frequency); + + return config; +} + +static void ma_loshelf_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_loshelf_node* pBPFNode = (ma_loshelf_node*)pNode; + + MA_ASSERT(pNode != NULL); + (void)pFrameCountIn; + + ma_loshelf2_process_pcm_frames(&pBPFNode->loshelf, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_loshelf_node_vtable = +{ + ma_loshelf_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* One input. */ + 1, /* One output. */ + 0 /* Default flags. */ +}; + +MA_API ma_result ma_loshelf_node_init(ma_node_graph* pNodeGraph, const ma_loshelf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_loshelf_node* pNode) +{ + ma_result result; + ma_node_config baseNodeConfig; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->loshelf.format != ma_format_f32) { + return MA_INVALID_ARGS; /* The format must be f32. */ + } + + result = ma_loshelf2_init(&pConfig->loshelf, pAllocationCallbacks, &pNode->loshelf); + if (result != MA_SUCCESS) { + return result; + } + + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_loshelf_node_vtable; + baseNodeConfig.pInputChannels = &pConfig->loshelf.channels; + baseNodeConfig.pOutputChannels = &pConfig->loshelf.channels; + + result = ma_node_init(pNodeGraph, &baseNodeConfig, pAllocationCallbacks, pNode); + if (result != MA_SUCCESS) { + return result; + } + + return result; +} + +MA_API ma_result ma_loshelf_node_reinit(const ma_loshelf_config* pConfig, ma_loshelf_node* pNode) +{ + ma_loshelf_node* pLoshelfNode = (ma_loshelf_node*)pNode; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + return ma_loshelf2_reinit(pConfig, &pLoshelfNode->loshelf); +} + +MA_API void ma_loshelf_node_uninit(ma_loshelf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_loshelf_node* pLoshelfNode = (ma_loshelf_node*)pNode; + + if (pNode == NULL) { + return; + } + + ma_node_uninit(pNode, pAllocationCallbacks); + ma_loshelf2_uninit(&pLoshelfNode->loshelf, pAllocationCallbacks); +} + + + +/* +High Shelf Filter Node +*/ +MA_API ma_hishelf_node_config ma_hishelf_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, double gainDB, double q, double frequency) +{ + ma_hishelf_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.hishelf = ma_hishelf2_config_init(ma_format_f32, channels, sampleRate, gainDB, q, frequency); + + return config; +} + +static void ma_hishelf_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_hishelf_node* pBPFNode = (ma_hishelf_node*)pNode; + + MA_ASSERT(pNode != NULL); + (void)pFrameCountIn; + + ma_hishelf2_process_pcm_frames(&pBPFNode->hishelf, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_hishelf_node_vtable = +{ + ma_hishelf_node_process_pcm_frames, + NULL, /* onGetRequiredInputFrameCount */ + 1, /* One input. */ + 1, /* One output. */ + 0 /* Default flags. */ +}; + +MA_API ma_result ma_hishelf_node_init(ma_node_graph* pNodeGraph, const ma_hishelf_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_hishelf_node* pNode) +{ + ma_result result; + ma_node_config baseNodeConfig; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pNode); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->hishelf.format != ma_format_f32) { + return MA_INVALID_ARGS; /* The format must be f32. */ + } + + result = ma_hishelf2_init(&pConfig->hishelf, pAllocationCallbacks, &pNode->hishelf); + if (result != MA_SUCCESS) { + return result; + } + + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_hishelf_node_vtable; + baseNodeConfig.pInputChannels = &pConfig->hishelf.channels; + baseNodeConfig.pOutputChannels = &pConfig->hishelf.channels; + + result = ma_node_init(pNodeGraph, &baseNodeConfig, pAllocationCallbacks, pNode); + if (result != MA_SUCCESS) { + return result; + } + + return result; +} + +MA_API ma_result ma_hishelf_node_reinit(const ma_hishelf_config* pConfig, ma_hishelf_node* pNode) +{ + ma_hishelf_node* pHishelfNode = (ma_hishelf_node*)pNode; + + if (pNode == NULL) { + return MA_INVALID_ARGS; + } + + return ma_hishelf2_reinit(pConfig, &pHishelfNode->hishelf); +} + +MA_API void ma_hishelf_node_uninit(ma_hishelf_node* pNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_hishelf_node* pHishelfNode = (ma_hishelf_node*)pNode; + + if (pNode == NULL) { + return; + } + + ma_node_uninit(pNode, pAllocationCallbacks); + ma_hishelf2_uninit(&pHishelfNode->hishelf, pAllocationCallbacks); +} + + + + +MA_API ma_delay_node_config ma_delay_node_config_init(ma_uint32 channels, ma_uint32 sampleRate, ma_uint32 delayInFrames, float decay) +{ + ma_delay_node_config config; + + config.nodeConfig = ma_node_config_init(); + config.delay = ma_delay_config_init(channels, sampleRate, delayInFrames, decay); + + return config; +} + + +static void ma_delay_node_process_pcm_frames(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_delay_node* pDelayNode = (ma_delay_node*)pNode; + + (void)pFrameCountIn; + + ma_delay_process_pcm_frames(&pDelayNode->delay, ppFramesOut[0], ppFramesIn[0], *pFrameCountOut); +} + +static ma_node_vtable g_ma_delay_node_vtable = +{ + ma_delay_node_process_pcm_frames, + NULL, + 1, /* 1 input channels. */ + 1, /* 1 output channel. */ + MA_NODE_FLAG_CONTINUOUS_PROCESSING /* Delay requires continuous processing to ensure the tail get's processed. */ +}; + +MA_API ma_result ma_delay_node_init(ma_node_graph* pNodeGraph, const ma_delay_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_delay_node* pDelayNode) +{ + ma_result result; + ma_node_config baseConfig; + + if (pDelayNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pDelayNode); + + result = ma_delay_init(&pConfig->delay, pAllocationCallbacks, &pDelayNode->delay); + if (result != MA_SUCCESS) { + return result; + } + + baseConfig = pConfig->nodeConfig; + baseConfig.vtable = &g_ma_delay_node_vtable; + baseConfig.pInputChannels = &pConfig->delay.channels; + baseConfig.pOutputChannels = &pConfig->delay.channels; + + result = ma_node_init(pNodeGraph, &baseConfig, pAllocationCallbacks, &pDelayNode->baseNode); + if (result != MA_SUCCESS) { + ma_delay_uninit(&pDelayNode->delay, pAllocationCallbacks); + return result; + } + + return result; +} + +MA_API void ma_delay_node_uninit(ma_delay_node* pDelayNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pDelayNode == NULL) { + return; + } + + /* The base node is always uninitialized first. */ + ma_node_uninit(pDelayNode, pAllocationCallbacks); + ma_delay_uninit(&pDelayNode->delay, pAllocationCallbacks); +} + +MA_API void ma_delay_node_set_wet(ma_delay_node* pDelayNode, float value) +{ + if (pDelayNode == NULL) { + return; + } + + ma_delay_set_wet(&pDelayNode->delay, value); +} + +MA_API float ma_delay_node_get_wet(const ma_delay_node* pDelayNode) +{ + if (pDelayNode == NULL) { + return 0; + } + + return ma_delay_get_wet(&pDelayNode->delay); +} + +MA_API void ma_delay_node_set_dry(ma_delay_node* pDelayNode, float value) +{ + if (pDelayNode == NULL) { + return; + } + + ma_delay_set_dry(&pDelayNode->delay, value); +} + +MA_API float ma_delay_node_get_dry(const ma_delay_node* pDelayNode) +{ + if (pDelayNode == NULL) { + return 0; + } + + return ma_delay_get_dry(&pDelayNode->delay); +} + +MA_API void ma_delay_node_set_decay(ma_delay_node* pDelayNode, float value) +{ + if (pDelayNode == NULL) { + return; + } + + ma_delay_set_decay(&pDelayNode->delay, value); +} + +MA_API float ma_delay_node_get_decay(const ma_delay_node* pDelayNode) +{ + if (pDelayNode == NULL) { + return 0; + } + + return ma_delay_get_decay(&pDelayNode->delay); +} +#endif /* MA_NO_NODE_GRAPH */ + + +/* SECTION: miniaudio_engine.c */ +#if !defined(MA_NO_ENGINE) && !defined(MA_NO_NODE_GRAPH) +/************************************************************************************************************************************************************** + +Engine + +**************************************************************************************************************************************************************/ +#define MA_SEEK_TARGET_NONE (~(ma_uint64)0) + + +static void ma_sound_set_at_end(ma_sound* pSound, ma_bool32 atEnd) +{ + MA_ASSERT(pSound != NULL); + ma_atomic_exchange_32(&pSound->atEnd, atEnd); + + /* Fire any callbacks or events. */ + if (atEnd) { + if (pSound->endCallback != NULL) { + pSound->endCallback(pSound->pEndCallbackUserData, pSound); + } + } +} + +static ma_bool32 ma_sound_get_at_end(const ma_sound* pSound) +{ + MA_ASSERT(pSound != NULL); + return ma_atomic_load_32(&pSound->atEnd); +} + + +MA_API ma_engine_node_config ma_engine_node_config_init(ma_engine* pEngine, ma_engine_node_type type, ma_uint32 flags) +{ + ma_engine_node_config config; + + MA_ZERO_OBJECT(&config); + config.pEngine = pEngine; + config.type = type; + config.isPitchDisabled = (flags & MA_SOUND_FLAG_NO_PITCH) != 0; + config.isSpatializationDisabled = (flags & MA_SOUND_FLAG_NO_SPATIALIZATION) != 0; + config.monoExpansionMode = pEngine->monoExpansionMode; + + return config; +} + + +static void ma_engine_node_update_pitch_if_required(ma_engine_node* pEngineNode) +{ + ma_bool32 isUpdateRequired = MA_FALSE; + float newPitch; + + MA_ASSERT(pEngineNode != NULL); + + newPitch = ma_atomic_load_explicit_f32(&pEngineNode->pitch, ma_atomic_memory_order_acquire); + + if (pEngineNode->oldPitch != newPitch) { + pEngineNode->oldPitch = newPitch; + isUpdateRequired = MA_TRUE; + } + + if (pEngineNode->oldDopplerPitch != pEngineNode->spatializer.dopplerPitch) { + pEngineNode->oldDopplerPitch = pEngineNode->spatializer.dopplerPitch; + isUpdateRequired = MA_TRUE; + } + + if (isUpdateRequired) { + float basePitch = (float)pEngineNode->sampleRate / ma_engine_get_sample_rate(pEngineNode->pEngine); + ma_linear_resampler_set_rate_ratio(&pEngineNode->resampler, basePitch * pEngineNode->oldPitch * pEngineNode->oldDopplerPitch); + } +} + +static ma_bool32 ma_engine_node_is_pitching_enabled(const ma_engine_node* pEngineNode) +{ + MA_ASSERT(pEngineNode != NULL); + + /* Don't try to be clever by skipping resampling in the pitch=1 case or else you'll glitch when moving away from 1. */ + return !ma_atomic_load_explicit_32(&pEngineNode->isPitchDisabled, ma_atomic_memory_order_acquire); +} + +static ma_bool32 ma_engine_node_is_spatialization_enabled(const ma_engine_node* pEngineNode) +{ + MA_ASSERT(pEngineNode != NULL); + + return !ma_atomic_load_explicit_32(&pEngineNode->isSpatializationDisabled, ma_atomic_memory_order_acquire); +} + +static ma_uint64 ma_engine_node_get_required_input_frame_count(const ma_engine_node* pEngineNode, ma_uint64 outputFrameCount) +{ + ma_uint64 inputFrameCount = 0; + + if (ma_engine_node_is_pitching_enabled(pEngineNode)) { + ma_result result = ma_linear_resampler_get_required_input_frame_count(&pEngineNode->resampler, outputFrameCount, &inputFrameCount); + if (result != MA_SUCCESS) { + inputFrameCount = 0; + } + } else { + inputFrameCount = outputFrameCount; /* No resampling, so 1:1. */ + } + + return inputFrameCount; +} + +static ma_result ma_engine_node_set_volume(ma_engine_node* pEngineNode, float volume) +{ + if (pEngineNode == NULL) { + return MA_INVALID_ARGS; + } + + ma_atomic_float_set(&pEngineNode->volume, volume); + + /* If we're not smoothing we should bypass the volume gainer entirely. */ + if (pEngineNode->volumeSmoothTimeInPCMFrames == 0) { + /* We should always have an active spatializer because it can be enabled and disabled dynamically. We can just use that for holding our volume. */ + ma_spatializer_set_master_volume(&pEngineNode->spatializer, volume); + } else { + /* We're using volume smoothing, so apply the master volume to the gainer. */ + ma_gainer_set_gain(&pEngineNode->volumeGainer, volume); + } + + return MA_SUCCESS; +} + +static ma_result ma_engine_node_get_volume(const ma_engine_node* pEngineNode, float* pVolume) +{ + if (pVolume == NULL) { + return MA_INVALID_ARGS; + } + + *pVolume = 0.0f; + + if (pEngineNode == NULL) { + return MA_INVALID_ARGS; + } + + *pVolume = ma_atomic_float_get((ma_atomic_float*)&pEngineNode->volume); + + return MA_SUCCESS; +} + + +static void ma_engine_node_process_pcm_frames__general(ma_engine_node* pEngineNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + ma_uint32 frameCountIn; + ma_uint32 frameCountOut; + ma_uint32 totalFramesProcessedIn; + ma_uint32 totalFramesProcessedOut; + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_bool32 isPitchingEnabled; + ma_bool32 isFadingEnabled; + ma_bool32 isSpatializationEnabled; + ma_bool32 isPanningEnabled; + ma_bool32 isVolumeSmoothingEnabled; + + frameCountIn = *pFrameCountIn; + frameCountOut = *pFrameCountOut; + + channelsIn = ma_spatializer_get_input_channels(&pEngineNode->spatializer); + channelsOut = ma_spatializer_get_output_channels(&pEngineNode->spatializer); + + totalFramesProcessedIn = 0; + totalFramesProcessedOut = 0; + + /* Update the fader if applicable. */ + { + ma_uint64 fadeLengthInFrames = ma_atomic_uint64_get(&pEngineNode->fadeSettings.fadeLengthInFrames); + if (fadeLengthInFrames != ~(ma_uint64)0) { + float fadeVolumeBeg = ma_atomic_float_get(&pEngineNode->fadeSettings.volumeBeg); + float fadeVolumeEnd = ma_atomic_float_get(&pEngineNode->fadeSettings.volumeEnd); + ma_int64 fadeStartOffsetInFrames = (ma_int64)ma_atomic_uint64_get(&pEngineNode->fadeSettings.absoluteGlobalTimeInFrames); + if (fadeStartOffsetInFrames == (ma_int64)(~(ma_uint64)0)) { + fadeStartOffsetInFrames = 0; + } else { + fadeStartOffsetInFrames -= ma_engine_get_time_in_pcm_frames(pEngineNode->pEngine); + } + + ma_fader_set_fade_ex(&pEngineNode->fader, fadeVolumeBeg, fadeVolumeEnd, fadeLengthInFrames, fadeStartOffsetInFrames); + + /* Reset the fade length so we don't erroneously apply it again. */ + ma_atomic_uint64_set(&pEngineNode->fadeSettings.fadeLengthInFrames, ~(ma_uint64)0); + } + } + + isPitchingEnabled = ma_engine_node_is_pitching_enabled(pEngineNode); + isFadingEnabled = pEngineNode->fader.volumeBeg != 1 || pEngineNode->fader.volumeEnd != 1; + isSpatializationEnabled = ma_engine_node_is_spatialization_enabled(pEngineNode); + isPanningEnabled = pEngineNode->panner.pan != 0 && channelsOut != 1; + isVolumeSmoothingEnabled = pEngineNode->volumeSmoothTimeInPCMFrames > 0; + + /* Keep going while we've still got data available for processing. */ + while (totalFramesProcessedOut < frameCountOut) { + /* + We need to process in a specific order. We always do resampling first because it's likely + we're going to be increasing the channel count after spatialization. Also, I want to do + fading based on the output sample rate. + + We'll first read into a buffer from the resampler. Then we'll do all processing that + operates on the on the input channel count. We'll then get the spatializer to output to + the output buffer and then do all effects from that point directly in the output buffer + in-place. + + Note that we're always running the resampler if pitching is enabled, even when the pitch + is 1. If we try to be clever and skip resampling when the pitch is 1, we'll get a glitch + when we move away from 1, back to 1, and then away from 1 again. We'll want to implement + any pitch=1 optimizations in the resampler itself. + + There's a small optimization here that we'll utilize since it might be a fairly common + case. When the input and output channel counts are the same, we'll read straight into the + output buffer from the resampler and do everything in-place. + */ + const float* pRunningFramesIn; + float* pRunningFramesOut; + float* pWorkingBuffer; /* This is the buffer that we'll be processing frames in. This is in input channels. */ + float temp[MA_DATA_CONVERTER_STACK_BUFFER_SIZE / sizeof(float)]; + ma_uint32 tempCapInFrames = ma_countof(temp) / channelsIn; + ma_uint32 framesAvailableIn; + ma_uint32 framesAvailableOut; + ma_uint32 framesJustProcessedIn; + ma_uint32 framesJustProcessedOut; + ma_bool32 isWorkingBufferValid = MA_FALSE; + + framesAvailableIn = frameCountIn - totalFramesProcessedIn; + framesAvailableOut = frameCountOut - totalFramesProcessedOut; + + pRunningFramesIn = ma_offset_pcm_frames_const_ptr_f32(ppFramesIn[0], totalFramesProcessedIn, channelsIn); + pRunningFramesOut = ma_offset_pcm_frames_ptr_f32(ppFramesOut[0], totalFramesProcessedOut, channelsOut); + + if (channelsIn == channelsOut) { + /* Fast path. Channel counts are the same. No need for an intermediary input buffer. */ + pWorkingBuffer = pRunningFramesOut; + } else { + /* Slow path. Channel counts are different. Need to use an intermediary input buffer. */ + pWorkingBuffer = temp; + if (framesAvailableOut > tempCapInFrames) { + framesAvailableOut = tempCapInFrames; + } + } + + /* First is resampler. */ + if (isPitchingEnabled) { + ma_uint64 resampleFrameCountIn = framesAvailableIn; + ma_uint64 resampleFrameCountOut = framesAvailableOut; + + ma_linear_resampler_process_pcm_frames(&pEngineNode->resampler, pRunningFramesIn, &resampleFrameCountIn, pWorkingBuffer, &resampleFrameCountOut); + isWorkingBufferValid = MA_TRUE; + + framesJustProcessedIn = (ma_uint32)resampleFrameCountIn; + framesJustProcessedOut = (ma_uint32)resampleFrameCountOut; + } else { + framesJustProcessedIn = ma_min(framesAvailableIn, framesAvailableOut); + framesJustProcessedOut = framesJustProcessedIn; /* When no resampling is being performed, the number of output frames is the same as input frames. */ + } + + /* Fading. */ + if (isFadingEnabled) { + if (isWorkingBufferValid) { + ma_fader_process_pcm_frames(&pEngineNode->fader, pWorkingBuffer, pWorkingBuffer, framesJustProcessedOut); /* In-place processing. */ + } else { + ma_fader_process_pcm_frames(&pEngineNode->fader, pWorkingBuffer, pRunningFramesIn, framesJustProcessedOut); + isWorkingBufferValid = MA_TRUE; + } + } + + /* + If we're using smoothing, we won't be applying volume via the spatializer, but instead from a ma_gainer. In this case + we'll want to apply our volume now. + */ + if (isVolumeSmoothingEnabled) { + if (isWorkingBufferValid) { + ma_gainer_process_pcm_frames(&pEngineNode->volumeGainer, pWorkingBuffer, pWorkingBuffer, framesJustProcessedOut); + } else { + ma_gainer_process_pcm_frames(&pEngineNode->volumeGainer, pWorkingBuffer, pRunningFramesIn, framesJustProcessedOut); + isWorkingBufferValid = MA_TRUE; + } + } + + /* + If at this point we still haven't actually done anything with the working buffer we need + to just read straight from the input buffer. + */ + if (isWorkingBufferValid == MA_FALSE) { + pWorkingBuffer = (float*)pRunningFramesIn; /* Naughty const cast, but it's safe at this point because we won't ever be writing to it from this point out. */ + } + + /* Spatialization. */ + if (isSpatializationEnabled) { + ma_uint32 iListener; + + /* + When determining the listener to use, we first check to see if the sound is pinned to a + specific listener. If so, we use that. Otherwise we just use the closest listener. + */ + if (pEngineNode->pinnedListenerIndex != MA_LISTENER_INDEX_CLOSEST && pEngineNode->pinnedListenerIndex < ma_engine_get_listener_count(pEngineNode->pEngine)) { + iListener = pEngineNode->pinnedListenerIndex; + } else { + ma_vec3f spatializerPosition = ma_spatializer_get_position(&pEngineNode->spatializer); + iListener = ma_engine_find_closest_listener(pEngineNode->pEngine, spatializerPosition.x, spatializerPosition.y, spatializerPosition.z); + } + + ma_spatializer_process_pcm_frames(&pEngineNode->spatializer, &pEngineNode->pEngine->listeners[iListener], pRunningFramesOut, pWorkingBuffer, framesJustProcessedOut); + } else { + /* No spatialization, but we still need to do channel conversion and master volume. */ + float volume; + ma_engine_node_get_volume(pEngineNode, &volume); /* Should never fail. */ + + if (channelsIn == channelsOut) { + /* No channel conversion required. Just copy straight to the output buffer. */ + if (isVolumeSmoothingEnabled) { + /* Volume has already been applied. Just copy straight to the output buffer. */ + ma_copy_pcm_frames(pRunningFramesOut, pWorkingBuffer, framesJustProcessedOut * channelsOut, ma_format_f32, channelsOut); + } else { + /* Volume has not been applied yet. Copy and apply volume in the same pass. */ + ma_copy_and_apply_volume_factor_f32(pRunningFramesOut, pWorkingBuffer, framesJustProcessedOut * channelsOut, volume); + } + } else { + /* Channel conversion required. TODO: Add support for channel maps here. */ + ma_channel_map_apply_f32(pRunningFramesOut, NULL, channelsOut, pWorkingBuffer, NULL, channelsIn, framesJustProcessedOut, ma_channel_mix_mode_simple, pEngineNode->monoExpansionMode); + + /* If we're using smoothing, the volume will have already been applied. */ + if (!isVolumeSmoothingEnabled) { + ma_apply_volume_factor_f32(pRunningFramesOut, framesJustProcessedOut * channelsOut, volume); + } + } + } + + /* At this point we can guarantee that the output buffer contains valid data. We can process everything in place now. */ + + /* Panning. */ + if (isPanningEnabled) { + ma_panner_process_pcm_frames(&pEngineNode->panner, pRunningFramesOut, pRunningFramesOut, framesJustProcessedOut); /* In-place processing. */ + } + + /* We're done for this chunk. */ + totalFramesProcessedIn += framesJustProcessedIn; + totalFramesProcessedOut += framesJustProcessedOut; + + /* If we didn't process any output frames this iteration it means we've either run out of input data, or run out of room in the output buffer. */ + if (framesJustProcessedOut == 0) { + break; + } + } + + /* At this point we're done processing. */ + *pFrameCountIn = totalFramesProcessedIn; + *pFrameCountOut = totalFramesProcessedOut; +} + +static void ma_engine_node_process_pcm_frames__sound(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + /* For sounds, we need to first read from the data source. Then we need to apply the engine effects (pan, pitch, fades, etc.). */ + ma_result result = MA_SUCCESS; + ma_sound* pSound = (ma_sound*)pNode; + ma_uint32 frameCount = *pFrameCountOut; + ma_uint32 totalFramesRead = 0; + ma_format dataSourceFormat; + ma_uint32 dataSourceChannels; + ma_uint8 temp[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; + ma_uint32 tempCapInFrames; + ma_uint64 seekTarget; + + /* This is a data source node which means no input buses. */ + (void)ppFramesIn; + (void)pFrameCountIn; + + /* If we're marked at the end we need to stop the sound and do nothing. */ + if (ma_sound_at_end(pSound)) { + ma_sound_stop(pSound); + *pFrameCountOut = 0; + return; + } + + /* If we're seeking, do so now before reading. */ + seekTarget = ma_atomic_load_64(&pSound->seekTarget); + if (seekTarget != MA_SEEK_TARGET_NONE) { + ma_data_source_seek_to_pcm_frame(pSound->pDataSource, seekTarget); + + /* Any time-dependant effects need to have their times updated. */ + ma_node_set_time(pSound, seekTarget); + + ma_atomic_exchange_64(&pSound->seekTarget, MA_SEEK_TARGET_NONE); + } + + /* + We want to update the pitch once. For sounds, this can be either at the start or at the end. If + we don't force this to only ever be updating once, we could end up in a situation where + retrieving the required input frame count ends up being different to what we actually retrieve. + What could happen is that the required input frame count is calculated, the pitch is update, + and then this processing function is called resulting in a different number of input frames + being processed. Do not call this in ma_engine_node_process_pcm_frames__general() or else + you'll hit the aforementioned bug. + */ + ma_engine_node_update_pitch_if_required(&pSound->engineNode); + + /* + For the convenience of the caller, we're doing to allow data sources to use non-floating-point formats and channel counts that differ + from the main engine. + */ + result = ma_data_source_get_data_format(pSound->pDataSource, &dataSourceFormat, &dataSourceChannels, NULL, NULL, 0); + if (result == MA_SUCCESS) { + tempCapInFrames = sizeof(temp) / ma_get_bytes_per_frame(dataSourceFormat, dataSourceChannels); + + /* Keep reading until we've read as much as was requested or we reach the end of the data source. */ + while (totalFramesRead < frameCount) { + ma_uint32 framesRemaining = frameCount - totalFramesRead; + ma_uint32 framesToRead; + ma_uint64 framesJustRead; + ma_uint32 frameCountIn; + ma_uint32 frameCountOut; + const float* pRunningFramesIn; + float* pRunningFramesOut; + + /* + The first thing we need to do is read into the temporary buffer. We can calculate exactly + how many input frames we'll need after resampling. + */ + framesToRead = (ma_uint32)ma_engine_node_get_required_input_frame_count(&pSound->engineNode, framesRemaining); + if (framesToRead > tempCapInFrames) { + framesToRead = tempCapInFrames; + } + + result = ma_data_source_read_pcm_frames(pSound->pDataSource, temp, framesToRead, &framesJustRead); + + /* If we reached the end of the sound we'll want to mark it as at the end and stop it. This should never be returned for looping sounds. */ + if (result == MA_AT_END) { + ma_sound_set_at_end(pSound, MA_TRUE); /* This will be set to false in ma_sound_start(). */ + } + + pRunningFramesOut = ma_offset_pcm_frames_ptr_f32(ppFramesOut[0], totalFramesRead, ma_node_get_output_channels(pNode, 0)); + + frameCountIn = (ma_uint32)framesJustRead; + frameCountOut = framesRemaining; + + /* Convert if necessary. */ + if (dataSourceFormat == ma_format_f32) { + /* Fast path. No data conversion necessary. */ + pRunningFramesIn = (float*)temp; + ma_engine_node_process_pcm_frames__general(&pSound->engineNode, &pRunningFramesIn, &frameCountIn, &pRunningFramesOut, &frameCountOut); + } else { + /* Slow path. Need to do sample format conversion to f32. If we give the f32 buffer the same count as the first temp buffer, we're guaranteed it'll be large enough. */ + float tempf32[MA_DATA_CONVERTER_STACK_BUFFER_SIZE]; /* Do not do `MA_DATA_CONVERTER_STACK_BUFFER_SIZE/sizeof(float)` here like we've done in other places. */ + ma_convert_pcm_frames_format(tempf32, ma_format_f32, temp, dataSourceFormat, framesJustRead, dataSourceChannels, ma_dither_mode_none); + + /* Now that we have our samples in f32 format we can process like normal. */ + pRunningFramesIn = tempf32; + ma_engine_node_process_pcm_frames__general(&pSound->engineNode, &pRunningFramesIn, &frameCountIn, &pRunningFramesOut, &frameCountOut); + } + + /* We should have processed all of our input frames since we calculated the required number of input frames at the top. */ + MA_ASSERT(frameCountIn == framesJustRead); + totalFramesRead += (ma_uint32)frameCountOut; /* Safe cast. */ + + if (result != MA_SUCCESS || ma_sound_at_end(pSound)) { + break; /* Might have reached the end. */ + } + } + } + + *pFrameCountOut = totalFramesRead; +} + +static void ma_engine_node_process_pcm_frames__group(ma_node* pNode, const float** ppFramesIn, ma_uint32* pFrameCountIn, float** ppFramesOut, ma_uint32* pFrameCountOut) +{ + /* + Make sure the pitch is updated before trying to read anything. It's important that this is done + only once and not in ma_engine_node_process_pcm_frames__general(). The reason for this is that + ma_engine_node_process_pcm_frames__general() will call ma_engine_node_get_required_input_frame_count(), + and if another thread modifies the pitch just after that call it can result in a glitch due to + the input rate changing. + */ + ma_engine_node_update_pitch_if_required((ma_engine_node*)pNode); + + /* For groups, the input data has already been read and we just need to apply the effect. */ + ma_engine_node_process_pcm_frames__general((ma_engine_node*)pNode, ppFramesIn, pFrameCountIn, ppFramesOut, pFrameCountOut); +} + +static ma_result ma_engine_node_get_required_input_frame_count__group(ma_node* pNode, ma_uint32 outputFrameCount, ma_uint32* pInputFrameCount) +{ + ma_uint64 inputFrameCount; + + MA_ASSERT(pInputFrameCount != NULL); + + /* Our pitch will affect this calculation. We need to update it. */ + ma_engine_node_update_pitch_if_required((ma_engine_node*)pNode); + + inputFrameCount = ma_engine_node_get_required_input_frame_count((ma_engine_node*)pNode, outputFrameCount); + if (inputFrameCount > 0xFFFFFFFF) { + inputFrameCount = 0xFFFFFFFF; /* Will never happen because miniaudio will only ever process in relatively small chunks. */ + } + + *pInputFrameCount = (ma_uint32)inputFrameCount; + + return MA_SUCCESS; +} + + +static ma_node_vtable g_ma_engine_node_vtable__sound = +{ + ma_engine_node_process_pcm_frames__sound, + NULL, /* onGetRequiredInputFrameCount */ + 0, /* Sounds are data source nodes which means they have zero inputs (their input is drawn from the data source itself). */ + 1, /* Sounds have one output bus. */ + 0 /* Default flags. */ +}; + +static ma_node_vtable g_ma_engine_node_vtable__group = +{ + ma_engine_node_process_pcm_frames__group, + ma_engine_node_get_required_input_frame_count__group, + 1, /* Groups have one input bus. */ + 1, /* Groups have one output bus. */ + MA_NODE_FLAG_DIFFERENT_PROCESSING_RATES /* The engine node does resampling so should let miniaudio know about it. */ +}; + + + +static ma_node_config ma_engine_node_base_node_config_init(const ma_engine_node_config* pConfig) +{ + ma_node_config baseNodeConfig; + + if (pConfig->type == ma_engine_node_type_sound) { + /* Sound. */ + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_engine_node_vtable__sound; + baseNodeConfig.initialState = ma_node_state_stopped; /* Sounds are stopped by default. */ + } else { + /* Group. */ + baseNodeConfig = ma_node_config_init(); + baseNodeConfig.vtable = &g_ma_engine_node_vtable__group; + baseNodeConfig.initialState = ma_node_state_started; /* Groups are started by default. */ + } + + return baseNodeConfig; +} + +static ma_spatializer_config ma_engine_node_spatializer_config_init(const ma_node_config* pBaseNodeConfig) +{ + return ma_spatializer_config_init(pBaseNodeConfig->pInputChannels[0], pBaseNodeConfig->pOutputChannels[0]); +} + +typedef struct +{ + size_t sizeInBytes; + size_t baseNodeOffset; + size_t resamplerOffset; + size_t spatializerOffset; + size_t gainerOffset; +} ma_engine_node_heap_layout; + +static ma_result ma_engine_node_get_heap_layout(const ma_engine_node_config* pConfig, ma_engine_node_heap_layout* pHeapLayout) +{ + ma_result result; + size_t tempHeapSize; + ma_node_config baseNodeConfig; + ma_linear_resampler_config resamplerConfig; + ma_spatializer_config spatializerConfig; + ma_gainer_config gainerConfig; + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_channel defaultStereoChannelMap[2] = {MA_CHANNEL_SIDE_LEFT, MA_CHANNEL_SIDE_RIGHT}; /* <-- Consistent with the default channel map of a stereo listener. Means channel conversion can run on a fast path. */ + + MA_ASSERT(pHeapLayout); + + MA_ZERO_OBJECT(pHeapLayout); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + if (pConfig->pEngine == NULL) { + return MA_INVALID_ARGS; /* An engine must be specified. */ + } + + pHeapLayout->sizeInBytes = 0; + + channelsIn = (pConfig->channelsIn != 0) ? pConfig->channelsIn : ma_engine_get_channels(pConfig->pEngine); + channelsOut = (pConfig->channelsOut != 0) ? pConfig->channelsOut : ma_engine_get_channels(pConfig->pEngine); + + + /* Base node. */ + baseNodeConfig = ma_engine_node_base_node_config_init(pConfig); + baseNodeConfig.pInputChannels = &channelsIn; + baseNodeConfig.pOutputChannels = &channelsOut; + + result = ma_node_get_heap_size(ma_engine_get_node_graph(pConfig->pEngine), &baseNodeConfig, &tempHeapSize); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the size of the heap for the base node. */ + } + + pHeapLayout->baseNodeOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(tempHeapSize); + + + /* Resmapler. */ + resamplerConfig = ma_linear_resampler_config_init(ma_format_f32, channelsIn, 1, 1); /* Input and output sample rates don't affect the calculation of the heap size. */ + resamplerConfig.lpfOrder = 0; + + result = ma_linear_resampler_get_heap_size(&resamplerConfig, &tempHeapSize); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the size of the heap for the resampler. */ + } + + pHeapLayout->resamplerOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(tempHeapSize); + + + /* Spatializer. */ + spatializerConfig = ma_engine_node_spatializer_config_init(&baseNodeConfig); + + if (spatializerConfig.channelsIn == 2) { + spatializerConfig.pChannelMapIn = defaultStereoChannelMap; + } + + result = ma_spatializer_get_heap_size(&spatializerConfig, &tempHeapSize); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the size of the heap for the spatializer. */ + } + + pHeapLayout->spatializerOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(tempHeapSize); + + + /* Gainer. Will not be used if we are not using smoothing. */ + if (pConfig->volumeSmoothTimeInPCMFrames > 0) { + gainerConfig = ma_gainer_config_init(channelsIn, pConfig->volumeSmoothTimeInPCMFrames); + + result = ma_gainer_get_heap_size(&gainerConfig, &tempHeapSize); + if (result != MA_SUCCESS) { + return result; + } + + pHeapLayout->gainerOffset = pHeapLayout->sizeInBytes; + pHeapLayout->sizeInBytes += ma_align_64(tempHeapSize); + } + + + return MA_SUCCESS; +} + +MA_API ma_result ma_engine_node_get_heap_size(const ma_engine_node_config* pConfig, size_t* pHeapSizeInBytes) +{ + ma_result result; + ma_engine_node_heap_layout heapLayout; + + if (pHeapSizeInBytes == NULL) { + return MA_INVALID_ARGS; + } + + *pHeapSizeInBytes = 0; + + result = ma_engine_node_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + *pHeapSizeInBytes = heapLayout.sizeInBytes; + + return MA_SUCCESS; +} + +MA_API ma_result ma_engine_node_init_preallocated(const ma_engine_node_config* pConfig, void* pHeap, ma_engine_node* pEngineNode) +{ + ma_result result; + ma_engine_node_heap_layout heapLayout; + ma_node_config baseNodeConfig; + ma_linear_resampler_config resamplerConfig; + ma_fader_config faderConfig; + ma_spatializer_config spatializerConfig; + ma_panner_config pannerConfig; + ma_gainer_config gainerConfig; + ma_uint32 channelsIn; + ma_uint32 channelsOut; + ma_channel defaultStereoChannelMap[2] = {MA_CHANNEL_SIDE_LEFT, MA_CHANNEL_SIDE_RIGHT}; /* <-- Consistent with the default channel map of a stereo listener. Means channel conversion can run on a fast path. */ + + if (pEngineNode == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pEngineNode); + + result = ma_engine_node_get_heap_layout(pConfig, &heapLayout); + if (result != MA_SUCCESS) { + return result; + } + + if (pConfig->pinnedListenerIndex != MA_LISTENER_INDEX_CLOSEST && pConfig->pinnedListenerIndex >= ma_engine_get_listener_count(pConfig->pEngine)) { + return MA_INVALID_ARGS; /* Invalid listener. */ + } + + pEngineNode->_pHeap = pHeap; + MA_ZERO_MEMORY(pHeap, heapLayout.sizeInBytes); + + pEngineNode->pEngine = pConfig->pEngine; + pEngineNode->sampleRate = (pConfig->sampleRate > 0) ? pConfig->sampleRate : ma_engine_get_sample_rate(pEngineNode->pEngine); + pEngineNode->volumeSmoothTimeInPCMFrames = pConfig->volumeSmoothTimeInPCMFrames; + pEngineNode->monoExpansionMode = pConfig->monoExpansionMode; + ma_atomic_float_set(&pEngineNode->volume, 1); + pEngineNode->pitch = 1; + pEngineNode->oldPitch = 1; + pEngineNode->oldDopplerPitch = 1; + pEngineNode->isPitchDisabled = pConfig->isPitchDisabled; + pEngineNode->isSpatializationDisabled = pConfig->isSpatializationDisabled; + pEngineNode->pinnedListenerIndex = pConfig->pinnedListenerIndex; + ma_atomic_float_set(&pEngineNode->fadeSettings.volumeBeg, 1); + ma_atomic_float_set(&pEngineNode->fadeSettings.volumeEnd, 1); + ma_atomic_uint64_set(&pEngineNode->fadeSettings.fadeLengthInFrames, (~(ma_uint64)0)); + ma_atomic_uint64_set(&pEngineNode->fadeSettings.absoluteGlobalTimeInFrames, (~(ma_uint64)0)); /* <-- Indicates that the fade should start immediately. */ + + channelsIn = (pConfig->channelsIn != 0) ? pConfig->channelsIn : ma_engine_get_channels(pConfig->pEngine); + channelsOut = (pConfig->channelsOut != 0) ? pConfig->channelsOut : ma_engine_get_channels(pConfig->pEngine); + + /* + If the sample rate of the sound is different to the engine, make sure pitching is enabled so that the resampler + is activated. Not doing this will result in the sound not being resampled if MA_SOUND_FLAG_NO_PITCH is used. + */ + if (pEngineNode->sampleRate != ma_engine_get_sample_rate(pEngineNode->pEngine)) { + pEngineNode->isPitchDisabled = MA_FALSE; + } + + + /* Base node. */ + baseNodeConfig = ma_engine_node_base_node_config_init(pConfig); + baseNodeConfig.pInputChannels = &channelsIn; + baseNodeConfig.pOutputChannels = &channelsOut; + + result = ma_node_init_preallocated(&pConfig->pEngine->nodeGraph, &baseNodeConfig, ma_offset_ptr(pHeap, heapLayout.baseNodeOffset), &pEngineNode->baseNode); + if (result != MA_SUCCESS) { + goto error0; + } + + + /* + We can now initialize the effects we need in order to implement the engine node. There's a + defined order of operations here, mainly centered around when we convert our channels from the + data source's native channel count to the engine's channel count. As a rule, we want to do as + much computation as possible before spatialization because there's a chance that will increase + the channel count, thereby increasing the amount of work needing to be done to process. + */ + + /* We'll always do resampling first. */ + resamplerConfig = ma_linear_resampler_config_init(ma_format_f32, baseNodeConfig.pInputChannels[0], pEngineNode->sampleRate, ma_engine_get_sample_rate(pEngineNode->pEngine)); + resamplerConfig.lpfOrder = 0; /* <-- Need to disable low-pass filtering for pitch shifting for now because there's cases where the biquads are becoming unstable. Need to figure out a better fix for this. */ + + result = ma_linear_resampler_init_preallocated(&resamplerConfig, ma_offset_ptr(pHeap, heapLayout.resamplerOffset), &pEngineNode->resampler); + if (result != MA_SUCCESS) { + goto error1; + } + + + /* After resampling will come the fader. */ + faderConfig = ma_fader_config_init(ma_format_f32, baseNodeConfig.pInputChannels[0], ma_engine_get_sample_rate(pEngineNode->pEngine)); + + result = ma_fader_init(&faderConfig, &pEngineNode->fader); + if (result != MA_SUCCESS) { + goto error2; + } + + + /* + Spatialization comes next. We spatialize based on the node's output channel count. It's up the caller to + ensure channels counts link up correctly in the node graph. + */ + spatializerConfig = ma_engine_node_spatializer_config_init(&baseNodeConfig); + spatializerConfig.gainSmoothTimeInFrames = pEngineNode->pEngine->gainSmoothTimeInFrames; + + if (spatializerConfig.channelsIn == 2) { + spatializerConfig.pChannelMapIn = defaultStereoChannelMap; + } + + result = ma_spatializer_init_preallocated(&spatializerConfig, ma_offset_ptr(pHeap, heapLayout.spatializerOffset), &pEngineNode->spatializer); + if (result != MA_SUCCESS) { + goto error2; + } + + + /* + After spatialization comes panning. We need to do this after spatialization because otherwise we wouldn't + be able to pan mono sounds. + */ + pannerConfig = ma_panner_config_init(ma_format_f32, baseNodeConfig.pOutputChannels[0]); + + result = ma_panner_init(&pannerConfig, &pEngineNode->panner); + if (result != MA_SUCCESS) { + goto error3; + } + + + /* We'll need a gainer for smoothing out volume changes if we have a non-zero smooth time. We apply this before converting to the output channel count. */ + if (pConfig->volumeSmoothTimeInPCMFrames > 0) { + gainerConfig = ma_gainer_config_init(channelsIn, pConfig->volumeSmoothTimeInPCMFrames); + + result = ma_gainer_init_preallocated(&gainerConfig, ma_offset_ptr(pHeap, heapLayout.gainerOffset), &pEngineNode->volumeGainer); + if (result != MA_SUCCESS) { + goto error3; + } + } + + + return MA_SUCCESS; + + /* No need for allocation callbacks here because we use a preallocated heap. */ +error3: ma_spatializer_uninit(&pEngineNode->spatializer, NULL); +error2: ma_linear_resampler_uninit(&pEngineNode->resampler, NULL); +error1: ma_node_uninit(&pEngineNode->baseNode, NULL); +error0: return result; +} + +MA_API ma_result ma_engine_node_init(const ma_engine_node_config* pConfig, const ma_allocation_callbacks* pAllocationCallbacks, ma_engine_node* pEngineNode) +{ + ma_result result; + size_t heapSizeInBytes; + void* pHeap; + + result = ma_engine_node_get_heap_size(pConfig, &heapSizeInBytes); + if (result != MA_SUCCESS) { + return result; + } + + if (heapSizeInBytes > 0) { + pHeap = ma_malloc(heapSizeInBytes, pAllocationCallbacks); + if (pHeap == NULL) { + return MA_OUT_OF_MEMORY; + } + } else { + pHeap = NULL; + } + + result = ma_engine_node_init_preallocated(pConfig, pHeap, pEngineNode); + if (result != MA_SUCCESS) { + ma_free(pHeap, pAllocationCallbacks); + return result; + } + + pEngineNode->_ownsHeap = MA_TRUE; + return MA_SUCCESS; +} + +MA_API void ma_engine_node_uninit(ma_engine_node* pEngineNode, const ma_allocation_callbacks* pAllocationCallbacks) +{ + /* + The base node always needs to be uninitialized first to ensure it's detached from the graph completely before we + destroy anything that might be in the middle of being used by the processing function. + */ + ma_node_uninit(&pEngineNode->baseNode, pAllocationCallbacks); + + /* Now that the node has been uninitialized we can safely uninitialize the rest. */ + if (pEngineNode->volumeSmoothTimeInPCMFrames > 0) { + ma_gainer_uninit(&pEngineNode->volumeGainer, pAllocationCallbacks); + } + + ma_spatializer_uninit(&pEngineNode->spatializer, pAllocationCallbacks); + ma_linear_resampler_uninit(&pEngineNode->resampler, pAllocationCallbacks); + + /* Free the heap last. */ + if (pEngineNode->_ownsHeap) { + ma_free(pEngineNode->_pHeap, pAllocationCallbacks); + } +} + + +MA_API ma_sound_config ma_sound_config_init(void) +{ + return ma_sound_config_init_2(NULL); +} + +MA_API ma_sound_config ma_sound_config_init_2(ma_engine* pEngine) +{ + ma_sound_config config; + + MA_ZERO_OBJECT(&config); + + if (pEngine != NULL) { + config.monoExpansionMode = pEngine->monoExpansionMode; + } else { + config.monoExpansionMode = ma_mono_expansion_mode_default; + } + + config.rangeEndInPCMFrames = ~((ma_uint64)0); + config.loopPointEndInPCMFrames = ~((ma_uint64)0); + + return config; +} + +MA_API ma_sound_group_config ma_sound_group_config_init(void) +{ + return ma_sound_group_config_init_2(NULL); +} + +MA_API ma_sound_group_config ma_sound_group_config_init_2(ma_engine* pEngine) +{ + ma_sound_group_config config; + + MA_ZERO_OBJECT(&config); + + if (pEngine != NULL) { + config.monoExpansionMode = pEngine->monoExpansionMode; + } else { + config.monoExpansionMode = ma_mono_expansion_mode_default; + } + + return config; +} + + +MA_API ma_engine_config ma_engine_config_init(void) +{ + ma_engine_config config; + + MA_ZERO_OBJECT(&config); + config.listenerCount = 1; /* Always want at least one listener. */ + config.monoExpansionMode = ma_mono_expansion_mode_default; + + return config; +} + + +#if !defined(MA_NO_DEVICE_IO) +static void ma_engine_data_callback_internal(ma_device* pDevice, void* pFramesOut, const void* pFramesIn, ma_uint32 frameCount) +{ + ma_engine* pEngine = (ma_engine*)pDevice->pUserData; + + (void)pFramesIn; + + /* + Experiment: Try processing a resource manager job if we're on the Emscripten build. + + This serves two purposes: + + 1) It ensures jobs are actually processed at some point since we cannot guarantee that the + caller is doing the right thing and calling ma_resource_manager_process_next_job(); and + + 2) It's an attempt at working around an issue where processing jobs on the Emscripten main + loop doesn't work as well as it should. When trying to load sounds without the `DECODE` + flag or with the `ASYNC` flag, the sound data is just not able to be loaded in time + before the callback is processed. I think it's got something to do with the single- + threaded nature of Web, but I'm not entirely sure. + */ + #if !defined(MA_NO_RESOURCE_MANAGER) && defined(MA_EMSCRIPTEN) + { + if (pEngine->pResourceManager != NULL) { + if ((pEngine->pResourceManager->config.flags & MA_RESOURCE_MANAGER_FLAG_NO_THREADING) != 0) { + ma_resource_manager_process_next_job(pEngine->pResourceManager); + } + } + } + #endif + + ma_engine_read_pcm_frames(pEngine, pFramesOut, frameCount, NULL); +} + +static ma_uint32 ma_device__get_processing_size_in_frames(ma_device* pDevice) +{ + /* + The processing size is the period size. The device can have a fixed sized processing size, or + it can be decided by the backend in which case it can be variable. + */ + if (pDevice->playback.intermediaryBufferCap > 0) { + /* Using a fixed sized processing callback. */ + return pDevice->playback.intermediaryBufferCap; + } else { + /* Not using a fixed sized processing callback. Need to estimate the processing size based on the backend. */ + return pDevice->playback.internalPeriodSizeInFrames; + } +} +#endif + +MA_API ma_result ma_engine_init(const ma_engine_config* pConfig, ma_engine* pEngine) +{ + ma_result result; + ma_node_graph_config nodeGraphConfig; + ma_engine_config engineConfig; + ma_spatializer_listener_config listenerConfig; + ma_uint32 iListener; + + if (pEngine == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pEngine); + + /* The config is allowed to be NULL in which case we use defaults for everything. */ + if (pConfig != NULL) { + engineConfig = *pConfig; + } else { + engineConfig = ma_engine_config_init(); + } + + pEngine->monoExpansionMode = engineConfig.monoExpansionMode; + pEngine->defaultVolumeSmoothTimeInPCMFrames = engineConfig.defaultVolumeSmoothTimeInPCMFrames; + pEngine->onProcess = engineConfig.onProcess; + pEngine->pProcessUserData = engineConfig.pProcessUserData; + ma_allocation_callbacks_init_copy(&pEngine->allocationCallbacks, &engineConfig.allocationCallbacks); + + #if !defined(MA_NO_RESOURCE_MANAGER) + { + pEngine->pResourceManager = engineConfig.pResourceManager; + } + #endif + + #if !defined(MA_NO_DEVICE_IO) + { + pEngine->pDevice = engineConfig.pDevice; + + /* If we don't have a device, we need one. */ + if (pEngine->pDevice == NULL && engineConfig.noDevice == MA_FALSE) { + ma_device_config deviceConfig; + + pEngine->pDevice = (ma_device*)ma_malloc(sizeof(*pEngine->pDevice), &pEngine->allocationCallbacks); + if (pEngine->pDevice == NULL) { + return MA_OUT_OF_MEMORY; + } + + deviceConfig = ma_device_config_init(ma_device_type_playback); + deviceConfig.playback.pDeviceID = engineConfig.pPlaybackDeviceID; + deviceConfig.playback.format = ma_format_f32; + deviceConfig.playback.channels = engineConfig.channels; + deviceConfig.sampleRate = engineConfig.sampleRate; + deviceConfig.dataCallback = (engineConfig.dataCallback != NULL) ? engineConfig.dataCallback : ma_engine_data_callback_internal; + deviceConfig.pUserData = pEngine; + deviceConfig.notificationCallback = engineConfig.notificationCallback; + deviceConfig.periodSizeInFrames = engineConfig.periodSizeInFrames; + deviceConfig.periodSizeInMilliseconds = engineConfig.periodSizeInMilliseconds; + deviceConfig.noPreSilencedOutputBuffer = MA_TRUE; /* We'll always be outputting to every frame in the callback so there's no need for a pre-silenced buffer. */ + deviceConfig.noClip = MA_TRUE; /* The engine will do clipping itself. */ + + if (engineConfig.pContext == NULL) { + ma_context_config contextConfig = ma_context_config_init(); + contextConfig.allocationCallbacks = pEngine->allocationCallbacks; + contextConfig.pLog = engineConfig.pLog; + + /* If the engine config does not specify a log, use the resource manager's if we have one. */ + #ifndef MA_NO_RESOURCE_MANAGER + { + if (contextConfig.pLog == NULL && engineConfig.pResourceManager != NULL) { + contextConfig.pLog = ma_resource_manager_get_log(engineConfig.pResourceManager); + } + } + #endif + + result = ma_device_init_ex(NULL, 0, &contextConfig, &deviceConfig, pEngine->pDevice); + } else { + result = ma_device_init(engineConfig.pContext, &deviceConfig, pEngine->pDevice); + } + + if (result != MA_SUCCESS) { + ma_free(pEngine->pDevice, &pEngine->allocationCallbacks); + pEngine->pDevice = NULL; + return result; + } + + pEngine->ownsDevice = MA_TRUE; + } + + /* Update the channel count and sample rate of the engine config so we can reference it below. */ + if (pEngine->pDevice != NULL) { + engineConfig.channels = pEngine->pDevice->playback.channels; + engineConfig.sampleRate = pEngine->pDevice->sampleRate; + + /* + The processing size used by the engine is determined by engineConfig.periodSizeInFrames. We want + to make this equal to what the device is using for it's period size. If we don't do that, it's + possible that the node graph will split it's processing into multiple passes which can introduce + glitching. + */ + engineConfig.periodSizeInFrames = ma_device__get_processing_size_in_frames(pEngine->pDevice); + } + } + #endif + + if (engineConfig.channels == 0 || engineConfig.sampleRate == 0) { + return MA_INVALID_ARGS; + } + + pEngine->sampleRate = engineConfig.sampleRate; + + /* The engine always uses either the log that was passed into the config, or the context's log is available. */ + if (engineConfig.pLog != NULL) { + pEngine->pLog = engineConfig.pLog; + } else { + #if !defined(MA_NO_DEVICE_IO) + { + pEngine->pLog = ma_device_get_log(pEngine->pDevice); + } + #else + { + pEngine->pLog = NULL; + } + #endif + } + + + /* The engine is a node graph. This needs to be initialized after we have the device so we can determine the channel count. */ + nodeGraphConfig = ma_node_graph_config_init(engineConfig.channels); + nodeGraphConfig.processingSizeInFrames = engineConfig.periodSizeInFrames; + nodeGraphConfig.preMixStackSizeInBytes = engineConfig.preMixStackSizeInBytes; + + result = ma_node_graph_init(&nodeGraphConfig, &pEngine->allocationCallbacks, &pEngine->nodeGraph); + if (result != MA_SUCCESS) { + goto on_error_1; + } + + + /* We need at least one listener. */ + if (engineConfig.listenerCount == 0) { + engineConfig.listenerCount = 1; + } + + if (engineConfig.listenerCount > MA_ENGINE_MAX_LISTENERS) { + result = MA_INVALID_ARGS; /* Too many listeners. */ + goto on_error_1; + } + + for (iListener = 0; iListener < engineConfig.listenerCount; iListener += 1) { + listenerConfig = ma_spatializer_listener_config_init(ma_node_graph_get_channels(&pEngine->nodeGraph)); + + /* + If we're using a device, use the device's channel map for the listener. Otherwise just use + miniaudio's default channel map. + */ + #if !defined(MA_NO_DEVICE_IO) + { + if (pEngine->pDevice != NULL) { + /* + Temporarily disabled. There is a subtle bug here where front-left and front-right + will be used by the device's channel map, but this is not what we want to use for + spatialization. Instead we want to use side-left and side-right. I need to figure + out a better solution for this. For now, disabling the use of device channel maps. + */ + /*listenerConfig.pChannelMapOut = pEngine->pDevice->playback.channelMap;*/ + } + } + #endif + + result = ma_spatializer_listener_init(&listenerConfig, &pEngine->allocationCallbacks, &pEngine->listeners[iListener]); /* TODO: Change this to a pre-allocated heap. */ + if (result != MA_SUCCESS) { + goto on_error_2; + } + + pEngine->listenerCount += 1; + } + + + /* Gain smoothing for spatialized sounds. */ + pEngine->gainSmoothTimeInFrames = engineConfig.gainSmoothTimeInFrames; + if (pEngine->gainSmoothTimeInFrames == 0) { + ma_uint32 gainSmoothTimeInMilliseconds = engineConfig.gainSmoothTimeInMilliseconds; + if (gainSmoothTimeInMilliseconds == 0) { + gainSmoothTimeInMilliseconds = 8; + } + + pEngine->gainSmoothTimeInFrames = (gainSmoothTimeInMilliseconds * ma_engine_get_sample_rate(pEngine)) / 1000; /* 8ms by default. */ + } + + + /* We need a resource manager. */ + #ifndef MA_NO_RESOURCE_MANAGER + { + if (pEngine->pResourceManager == NULL) { + ma_resource_manager_config resourceManagerConfig; + + pEngine->pResourceManager = (ma_resource_manager*)ma_malloc(sizeof(*pEngine->pResourceManager), &pEngine->allocationCallbacks); + if (pEngine->pResourceManager == NULL) { + result = MA_OUT_OF_MEMORY; + goto on_error_2; + } + + resourceManagerConfig = ma_resource_manager_config_init(); + resourceManagerConfig.pLog = pEngine->pLog; /* Always use the engine's log for internally-managed resource managers. */ + resourceManagerConfig.decodedFormat = ma_format_f32; + resourceManagerConfig.decodedChannels = 0; /* Leave the decoded channel count as 0 so we can get good spatialization. */ + resourceManagerConfig.decodedSampleRate = ma_engine_get_sample_rate(pEngine); + ma_allocation_callbacks_init_copy(&resourceManagerConfig.allocationCallbacks, &pEngine->allocationCallbacks); + resourceManagerConfig.pVFS = engineConfig.pResourceManagerVFS; + + /* The Emscripten build cannot use threads unless it's targeting pthreads. */ + #if defined(MA_EMSCRIPTEN) && !defined(__EMSCRIPTEN_PTHREADS__) + { + resourceManagerConfig.jobThreadCount = 0; + resourceManagerConfig.flags |= MA_RESOURCE_MANAGER_FLAG_NO_THREADING; + } + #endif + + result = ma_resource_manager_init(&resourceManagerConfig, pEngine->pResourceManager); + if (result != MA_SUCCESS) { + goto on_error_3; + } + + pEngine->ownsResourceManager = MA_TRUE; + } + } + #endif + + /* Setup some stuff for inlined sounds. That is sounds played with ma_engine_play_sound(). */ + pEngine->inlinedSoundLock = 0; + pEngine->pInlinedSoundHead = NULL; + + /* Start the engine if required. This should always be the last step. */ + #if !defined(MA_NO_DEVICE_IO) + { + if (engineConfig.noAutoStart == MA_FALSE && pEngine->pDevice != NULL) { + result = ma_engine_start(pEngine); + if (result != MA_SUCCESS) { + goto on_error_4; /* Failed to start the engine. */ + } + } + } + #endif + + return MA_SUCCESS; + +#if !defined(MA_NO_DEVICE_IO) +on_error_4: +#endif +#if !defined(MA_NO_RESOURCE_MANAGER) +on_error_3: + if (pEngine->ownsResourceManager) { + ma_free(pEngine->pResourceManager, &pEngine->allocationCallbacks); + } +#endif /* MA_NO_RESOURCE_MANAGER */ +on_error_2: + for (iListener = 0; iListener < pEngine->listenerCount; iListener += 1) { + ma_spatializer_listener_uninit(&pEngine->listeners[iListener], &pEngine->allocationCallbacks); + } + + ma_node_graph_uninit(&pEngine->nodeGraph, &pEngine->allocationCallbacks); +on_error_1: + #if !defined(MA_NO_DEVICE_IO) + { + if (pEngine->ownsDevice) { + ma_device_uninit(pEngine->pDevice); + ma_free(pEngine->pDevice, &pEngine->allocationCallbacks); + } + } + #endif + + return result; +} + +MA_API void ma_engine_uninit(ma_engine* pEngine) +{ + ma_uint32 iListener; + + if (pEngine == NULL) { + return; + } + + /* The device must be uninitialized before the node graph to ensure the audio thread doesn't try accessing it. */ + #if !defined(MA_NO_DEVICE_IO) + { + if (pEngine->ownsDevice) { + ma_device_uninit(pEngine->pDevice); + ma_free(pEngine->pDevice, &pEngine->allocationCallbacks); + } else { + if (pEngine->pDevice != NULL) { + ma_device_stop(pEngine->pDevice); + } + } + } + #endif + + /* + All inlined sounds need to be deleted. I'm going to use a lock here just to future proof in case + I want to do some kind of garbage collection later on. + */ + ma_spinlock_lock(&pEngine->inlinedSoundLock); + { + for (;;) { + ma_sound_inlined* pSoundToDelete = pEngine->pInlinedSoundHead; + if (pSoundToDelete == NULL) { + break; /* Done. */ + } + + pEngine->pInlinedSoundHead = pSoundToDelete->pNext; + + ma_sound_uninit(&pSoundToDelete->sound); + ma_free(pSoundToDelete, &pEngine->allocationCallbacks); + } + } + ma_spinlock_unlock(&pEngine->inlinedSoundLock); + + for (iListener = 0; iListener < pEngine->listenerCount; iListener += 1) { + ma_spatializer_listener_uninit(&pEngine->listeners[iListener], &pEngine->allocationCallbacks); + } + + /* Make sure the node graph is uninitialized after the audio thread has been shutdown to prevent accessing of the node graph after being uninitialized. */ + ma_node_graph_uninit(&pEngine->nodeGraph, &pEngine->allocationCallbacks); + + /* Uninitialize the resource manager last to ensure we don't have a thread still trying to access it. */ +#ifndef MA_NO_RESOURCE_MANAGER + if (pEngine->ownsResourceManager) { + ma_resource_manager_uninit(pEngine->pResourceManager); + ma_free(pEngine->pResourceManager, &pEngine->allocationCallbacks); + } +#endif +} + +MA_API ma_result ma_engine_read_pcm_frames(ma_engine* pEngine, void* pFramesOut, ma_uint64 frameCount, ma_uint64* pFramesRead) +{ + ma_result result; + ma_uint64 framesRead = 0; + + if (pFramesRead != NULL) { + *pFramesRead = 0; + } + + result = ma_node_graph_read_pcm_frames(&pEngine->nodeGraph, pFramesOut, frameCount, &framesRead); + if (result != MA_SUCCESS) { + return result; + } + + if (pFramesRead != NULL) { + *pFramesRead = framesRead; + } + + if (pEngine->onProcess) { + pEngine->onProcess(pEngine->pProcessUserData, (float*)pFramesOut, framesRead); /* Safe cast to float* because the engine always works on floating point samples. */ + } + + return MA_SUCCESS; +} + +MA_API ma_node_graph* ma_engine_get_node_graph(ma_engine* pEngine) +{ + if (pEngine == NULL) { + return NULL; + } + + return &pEngine->nodeGraph; +} + +#if !defined(MA_NO_RESOURCE_MANAGER) +MA_API ma_resource_manager* ma_engine_get_resource_manager(ma_engine* pEngine) +{ + if (pEngine == NULL) { + return NULL; + } + + #if !defined(MA_NO_RESOURCE_MANAGER) + { + return pEngine->pResourceManager; + } + #else + { + return NULL; + } + #endif +} +#endif + +MA_API ma_device* ma_engine_get_device(ma_engine* pEngine) +{ + if (pEngine == NULL) { + return NULL; + } + + #if !defined(MA_NO_DEVICE_IO) + { + return pEngine->pDevice; + } + #else + { + return NULL; + } + #endif +} + +MA_API ma_log* ma_engine_get_log(ma_engine* pEngine) +{ + if (pEngine == NULL) { + return NULL; + } + + if (pEngine->pLog != NULL) { + return pEngine->pLog; + } else { + #if !defined(MA_NO_DEVICE_IO) + { + return ma_device_get_log(ma_engine_get_device(pEngine)); + } + #else + { + return NULL; + } + #endif + } +} + +MA_API ma_node* ma_engine_get_endpoint(ma_engine* pEngine) +{ + return ma_node_graph_get_endpoint(&pEngine->nodeGraph); +} + +MA_API ma_uint64 ma_engine_get_time_in_pcm_frames(const ma_engine* pEngine) +{ + return ma_node_graph_get_time(&pEngine->nodeGraph); +} + +MA_API ma_uint64 ma_engine_get_time_in_milliseconds(const ma_engine* pEngine) +{ + return ma_engine_get_time_in_pcm_frames(pEngine) * 1000 / ma_engine_get_sample_rate(pEngine); +} + +MA_API ma_result ma_engine_set_time_in_pcm_frames(ma_engine* pEngine, ma_uint64 globalTime) +{ + return ma_node_graph_set_time(&pEngine->nodeGraph, globalTime); +} + +MA_API ma_result ma_engine_set_time_in_milliseconds(ma_engine* pEngine, ma_uint64 globalTime) +{ + return ma_engine_set_time_in_pcm_frames(pEngine, globalTime * ma_engine_get_sample_rate(pEngine) / 1000); +} + +MA_API ma_uint64 ma_engine_get_time(const ma_engine* pEngine) +{ + return ma_engine_get_time_in_pcm_frames(pEngine); +} + +MA_API ma_result ma_engine_set_time(ma_engine* pEngine, ma_uint64 globalTime) +{ + return ma_engine_set_time_in_pcm_frames(pEngine, globalTime); +} + +MA_API ma_uint32 ma_engine_get_channels(const ma_engine* pEngine) +{ + return ma_node_graph_get_channels(&pEngine->nodeGraph); +} + +MA_API ma_uint32 ma_engine_get_sample_rate(const ma_engine* pEngine) +{ + if (pEngine == NULL) { + return 0; + } + + return pEngine->sampleRate; +} + + +MA_API ma_result ma_engine_start(ma_engine* pEngine) +{ + ma_result result; + + if (pEngine == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_DEVICE_IO) + { + if (pEngine->pDevice != NULL) { + result = ma_device_start(pEngine->pDevice); + } else { + result = MA_INVALID_OPERATION; /* The engine is running without a device which means there's no real notion of "starting" the engine. */ + } + } + #else + { + result = MA_INVALID_OPERATION; /* Device IO is disabled, so there's no real notion of "starting" the engine. */ + } + #endif + + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_engine_stop(ma_engine* pEngine) +{ + ma_result result; + + if (pEngine == NULL) { + return MA_INVALID_ARGS; + } + + #if !defined(MA_NO_DEVICE_IO) + { + if (pEngine->pDevice != NULL) { + result = ma_device_stop(pEngine->pDevice); + } else { + result = MA_INVALID_OPERATION; /* The engine is running without a device which means there's no real notion of "stopping" the engine. */ + } + } + #else + { + result = MA_INVALID_OPERATION; /* Device IO is disabled, so there's no real notion of "stopping" the engine. */ + } + #endif + + if (result != MA_SUCCESS) { + return result; + } + + return MA_SUCCESS; +} + +MA_API ma_result ma_engine_set_volume(ma_engine* pEngine, float volume) +{ + if (pEngine == NULL) { + return MA_INVALID_ARGS; + } + + return ma_node_set_output_bus_volume(ma_node_graph_get_endpoint(&pEngine->nodeGraph), 0, volume); +} + +MA_API float ma_engine_get_volume(ma_engine* pEngine) +{ + if (pEngine == NULL) { + return 0; + } + + return ma_node_get_output_bus_volume(ma_node_graph_get_endpoint(&pEngine->nodeGraph), 0); +} + +MA_API ma_result ma_engine_set_gain_db(ma_engine* pEngine, float gainDB) +{ + return ma_engine_set_volume(pEngine, ma_volume_db_to_linear(gainDB)); +} + +MA_API float ma_engine_get_gain_db(ma_engine* pEngine) +{ + return ma_volume_linear_to_db(ma_engine_get_volume(pEngine)); +} + + +MA_API ma_uint32 ma_engine_get_listener_count(const ma_engine* pEngine) +{ + if (pEngine == NULL) { + return 0; + } + + return pEngine->listenerCount; +} + +MA_API ma_uint32 ma_engine_find_closest_listener(const ma_engine* pEngine, float absolutePosX, float absolutePosY, float absolutePosZ) +{ + ma_uint32 iListener; + ma_uint32 iListenerClosest; + float closestLen2 = MA_FLT_MAX; + + if (pEngine == NULL || pEngine->listenerCount == 1) { + return 0; + } + + iListenerClosest = 0; + for (iListener = 0; iListener < pEngine->listenerCount; iListener += 1) { + if (ma_engine_listener_is_enabled(pEngine, iListener)) { + float len2 = ma_vec3f_len2(ma_vec3f_sub(ma_spatializer_listener_get_position(&pEngine->listeners[iListener]), ma_vec3f_init_3f(absolutePosX, absolutePosY, absolutePosZ))); + if (closestLen2 > len2) { + closestLen2 = len2; + iListenerClosest = iListener; + } + } + } + + MA_ASSERT(iListenerClosest < 255); + return iListenerClosest; +} + +MA_API void ma_engine_listener_set_position(ma_engine* pEngine, ma_uint32 listenerIndex, float x, float y, float z) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return; + } + + ma_spatializer_listener_set_position(&pEngine->listeners[listenerIndex], x, y, z); +} + +MA_API ma_vec3f ma_engine_listener_get_position(const ma_engine* pEngine, ma_uint32 listenerIndex) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_spatializer_listener_get_position(&pEngine->listeners[listenerIndex]); +} + +MA_API void ma_engine_listener_set_direction(ma_engine* pEngine, ma_uint32 listenerIndex, float x, float y, float z) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return; + } + + ma_spatializer_listener_set_direction(&pEngine->listeners[listenerIndex], x, y, z); +} + +MA_API ma_vec3f ma_engine_listener_get_direction(const ma_engine* pEngine, ma_uint32 listenerIndex) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return ma_vec3f_init_3f(0, 0, -1); + } + + return ma_spatializer_listener_get_direction(&pEngine->listeners[listenerIndex]); +} + +MA_API void ma_engine_listener_set_velocity(ma_engine* pEngine, ma_uint32 listenerIndex, float x, float y, float z) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return; + } + + ma_spatializer_listener_set_velocity(&pEngine->listeners[listenerIndex], x, y, z); +} + +MA_API ma_vec3f ma_engine_listener_get_velocity(const ma_engine* pEngine, ma_uint32 listenerIndex) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_spatializer_listener_get_velocity(&pEngine->listeners[listenerIndex]); +} + +MA_API void ma_engine_listener_set_cone(ma_engine* pEngine, ma_uint32 listenerIndex, float innerAngleInRadians, float outerAngleInRadians, float outerGain) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return; + } + + ma_spatializer_listener_set_cone(&pEngine->listeners[listenerIndex], innerAngleInRadians, outerAngleInRadians, outerGain); +} + +MA_API void ma_engine_listener_get_cone(const ma_engine* pEngine, ma_uint32 listenerIndex, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain) +{ + if (pInnerAngleInRadians != NULL) { + *pInnerAngleInRadians = 0; + } + + if (pOuterAngleInRadians != NULL) { + *pOuterAngleInRadians = 0; + } + + if (pOuterGain != NULL) { + *pOuterGain = 0; + } + + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return; + } + + ma_spatializer_listener_get_cone(&pEngine->listeners[listenerIndex], pInnerAngleInRadians, pOuterAngleInRadians, pOuterGain); +} + +MA_API void ma_engine_listener_set_world_up(ma_engine* pEngine, ma_uint32 listenerIndex, float x, float y, float z) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return; + } + + ma_spatializer_listener_set_world_up(&pEngine->listeners[listenerIndex], x, y, z); +} + +MA_API ma_vec3f ma_engine_listener_get_world_up(const ma_engine* pEngine, ma_uint32 listenerIndex) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return ma_vec3f_init_3f(0, 1, 0); + } + + return ma_spatializer_listener_get_world_up(&pEngine->listeners[listenerIndex]); +} + +MA_API void ma_engine_listener_set_enabled(ma_engine* pEngine, ma_uint32 listenerIndex, ma_bool32 isEnabled) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return; + } + + ma_spatializer_listener_set_enabled(&pEngine->listeners[listenerIndex], isEnabled); +} + +MA_API ma_bool32 ma_engine_listener_is_enabled(const ma_engine* pEngine, ma_uint32 listenerIndex) +{ + if (pEngine == NULL || listenerIndex >= pEngine->listenerCount) { + return MA_FALSE; + } + + return ma_spatializer_listener_is_enabled(&pEngine->listeners[listenerIndex]); +} + + +#ifndef MA_NO_RESOURCE_MANAGER +MA_API ma_result ma_engine_play_sound_ex(ma_engine* pEngine, const char* pFilePath, ma_node* pNode, ma_uint32 nodeInputBusIndex) +{ + ma_result result = MA_SUCCESS; + ma_sound_inlined* pSound = NULL; + ma_sound_inlined* pNextSound = NULL; + + if (pEngine == NULL || pFilePath == NULL) { + return MA_INVALID_ARGS; + } + + /* Attach to the endpoint node if nothing is specified. */ + if (pNode == NULL) { + pNode = ma_node_graph_get_endpoint(&pEngine->nodeGraph); + nodeInputBusIndex = 0; + } + + /* + We want to check if we can recycle an already-allocated inlined sound. Since this is just a + helper I'm not *too* concerned about performance here and I'm happy to use a lock to keep + the implementation simple. Maybe this can be optimized later if there's enough demand, but + if this function is being used it probably means the caller doesn't really care too much. + + What we do is check the atEnd flag. When this is true, we can recycle the sound. Otherwise + we just keep iterating. If we reach the end without finding a sound to recycle we just + allocate a new one. This doesn't scale well for a massive number of sounds being played + simultaneously as we don't ever actually free the sound objects. Some kind of garbage + collection routine might be valuable for this which I'll think about. + */ + ma_spinlock_lock(&pEngine->inlinedSoundLock); + { + ma_uint32 soundFlags = 0; + + for (pNextSound = pEngine->pInlinedSoundHead; pNextSound != NULL; pNextSound = pNextSound->pNext) { + if (ma_sound_at_end(&pNextSound->sound)) { + /* + The sound is at the end which means it's available for recycling. All we need to do + is uninitialize it and reinitialize it. All we're doing is recycling memory. + */ + pSound = pNextSound; + ma_atomic_fetch_sub_32(&pEngine->inlinedSoundCount, 1); + break; + } + } + + if (pSound != NULL) { + /* + We actually want to detach the sound from the list here. The reason is because we want the sound + to be in a consistent state at the non-recycled case to simplify the logic below. + */ + if (pEngine->pInlinedSoundHead == pSound) { + pEngine->pInlinedSoundHead = pSound->pNext; + } + + if (pSound->pPrev != NULL) { + pSound->pPrev->pNext = pSound->pNext; + } + if (pSound->pNext != NULL) { + pSound->pNext->pPrev = pSound->pPrev; + } + + /* Now the previous sound needs to be uninitialized. */ + ma_sound_uninit(&pNextSound->sound); + } else { + /* No sound available for recycling. Allocate one now. */ + pSound = (ma_sound_inlined*)ma_malloc(sizeof(*pSound), &pEngine->allocationCallbacks); + } + + if (pSound != NULL) { /* Safety check for the allocation above. */ + /* + At this point we should have memory allocated for the inlined sound. We just need + to initialize it like a normal sound now. + */ + soundFlags |= MA_SOUND_FLAG_ASYNC; /* For inlined sounds we don't want to be sitting around waiting for stuff to load so force an async load. */ + soundFlags |= MA_SOUND_FLAG_NO_DEFAULT_ATTACHMENT; /* We want specific control over where the sound is attached in the graph. We'll attach it manually just before playing the sound. */ + soundFlags |= MA_SOUND_FLAG_NO_PITCH; /* Pitching isn't usable with inlined sounds, so disable it to save on speed. */ + soundFlags |= MA_SOUND_FLAG_NO_SPATIALIZATION; /* Not currently doing spatialization with inlined sounds, but this might actually change later. For now disable spatialization. Will be removed if we ever add support for spatialization here. */ + + result = ma_sound_init_from_file(pEngine, pFilePath, soundFlags, NULL, NULL, &pSound->sound); + if (result == MA_SUCCESS) { + /* Now attach the sound to the graph. */ + result = ma_node_attach_output_bus(pSound, 0, pNode, nodeInputBusIndex); + if (result == MA_SUCCESS) { + /* At this point the sound should be loaded and we can go ahead and add it to the list. The new item becomes the new head. */ + pSound->pNext = pEngine->pInlinedSoundHead; + pSound->pPrev = NULL; + + pEngine->pInlinedSoundHead = pSound; /* <-- This is what attaches the sound to the list. */ + if (pSound->pNext != NULL) { + pSound->pNext->pPrev = pSound; + } + } else { + ma_free(pSound, &pEngine->allocationCallbacks); + } + } else { + ma_free(pSound, &pEngine->allocationCallbacks); + } + } else { + result = MA_OUT_OF_MEMORY; + } + } + ma_spinlock_unlock(&pEngine->inlinedSoundLock); + + if (result != MA_SUCCESS) { + return result; + } + + /* Finally we can start playing the sound. */ + result = ma_sound_start(&pSound->sound); + if (result != MA_SUCCESS) { + /* Failed to start the sound. We need to mark it for recycling and return an error. */ + ma_atomic_exchange_32(&pSound->sound.atEnd, MA_TRUE); + return result; + } + + ma_atomic_fetch_add_32(&pEngine->inlinedSoundCount, 1); + return result; +} + +MA_API ma_result ma_engine_play_sound(ma_engine* pEngine, const char* pFilePath, ma_sound_group* pGroup) +{ + return ma_engine_play_sound_ex(pEngine, pFilePath, pGroup, 0); +} +#endif + + +static ma_result ma_sound_preinit(ma_engine* pEngine, ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pSound); + pSound->seekTarget = MA_SEEK_TARGET_NONE; + + if (pEngine == NULL) { + return MA_INVALID_ARGS; + } + + return MA_SUCCESS; +} + +static ma_result ma_sound_init_from_data_source_internal(ma_engine* pEngine, const ma_sound_config* pConfig, ma_sound* pSound) +{ + ma_result result; + ma_engine_node_config engineNodeConfig; + ma_engine_node_type type; /* Will be set to ma_engine_node_type_group if no data source is specified. */ + + /* Do not clear pSound to zero here - that's done at a higher level with ma_sound_preinit(). */ + MA_ASSERT(pEngine != NULL); + MA_ASSERT(pSound != NULL); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + pSound->pDataSource = pConfig->pDataSource; + + if (pConfig->pDataSource != NULL) { + type = ma_engine_node_type_sound; + } else { + type = ma_engine_node_type_group; + } + + /* + Sounds are engine nodes. Before we can initialize this we need to determine the channel count. + If we can't do this we need to abort. It's up to the caller to ensure they're using a data + source that provides this information upfront. + */ + engineNodeConfig = ma_engine_node_config_init(pEngine, type, pConfig->flags); + engineNodeConfig.channelsIn = pConfig->channelsIn; + engineNodeConfig.channelsOut = pConfig->channelsOut; + engineNodeConfig.volumeSmoothTimeInPCMFrames = pConfig->volumeSmoothTimeInPCMFrames; + engineNodeConfig.monoExpansionMode = pConfig->monoExpansionMode; + + if (engineNodeConfig.volumeSmoothTimeInPCMFrames == 0) { + engineNodeConfig.volumeSmoothTimeInPCMFrames = pEngine->defaultVolumeSmoothTimeInPCMFrames; + } + + /* If we're loading from a data source the input channel count needs to be the data source's native channel count. */ + if (pConfig->pDataSource != NULL) { + result = ma_data_source_get_data_format(pConfig->pDataSource, NULL, &engineNodeConfig.channelsIn, &engineNodeConfig.sampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; /* Failed to retrieve the channel count. */ + } + + if (engineNodeConfig.channelsIn == 0) { + return MA_INVALID_OPERATION; /* Invalid channel count. */ + } + + if (engineNodeConfig.channelsOut == MA_SOUND_SOURCE_CHANNEL_COUNT) { + engineNodeConfig.channelsOut = engineNodeConfig.channelsIn; + } + } + + + /* Getting here means we should have a valid channel count and we can initialize the engine node. */ + result = ma_engine_node_init(&engineNodeConfig, &pEngine->allocationCallbacks, &pSound->engineNode); + if (result != MA_SUCCESS) { + return result; + } + + /* If no attachment is specified, attach the sound straight to the endpoint. */ + if (pConfig->pInitialAttachment == NULL) { + /* No group. Attach straight to the endpoint by default, unless the caller has requested that it not. */ + if ((pConfig->flags & MA_SOUND_FLAG_NO_DEFAULT_ATTACHMENT) == 0) { + result = ma_node_attach_output_bus(pSound, 0, ma_node_graph_get_endpoint(&pEngine->nodeGraph), 0); + } + } else { + /* An attachment is specified. Attach to it by default. The sound has only a single output bus, and the config will specify which input bus to attach to. */ + result = ma_node_attach_output_bus(pSound, 0, pConfig->pInitialAttachment, pConfig->initialAttachmentInputBusIndex); + } + + if (result != MA_SUCCESS) { + ma_engine_node_uninit(&pSound->engineNode, &pEngine->allocationCallbacks); + return result; + } + + + /* Apply initial range and looping state to the data source if applicable. */ + if (pConfig->rangeBegInPCMFrames != 0 || pConfig->rangeEndInPCMFrames != ~((ma_uint64)0)) { + ma_data_source_set_range_in_pcm_frames(ma_sound_get_data_source(pSound), pConfig->rangeBegInPCMFrames, pConfig->rangeEndInPCMFrames); + } + + if (pConfig->loopPointBegInPCMFrames != 0 || pConfig->loopPointEndInPCMFrames != ~((ma_uint64)0)) { + ma_data_source_set_range_in_pcm_frames(ma_sound_get_data_source(pSound), pConfig->loopPointBegInPCMFrames, pConfig->loopPointEndInPCMFrames); + } + + ma_sound_set_looping(pSound, pConfig->isLooping || ((pConfig->flags & MA_SOUND_FLAG_LOOPING) != 0)); + + return MA_SUCCESS; +} + +#ifndef MA_NO_RESOURCE_MANAGER +MA_API ma_result ma_sound_init_from_file_internal(ma_engine* pEngine, const ma_sound_config* pConfig, ma_sound* pSound) +{ + ma_result result = MA_SUCCESS; + ma_uint32 flags; + ma_sound_config config; + ma_resource_manager_pipeline_notifications notifications; + + /* + The engine requires knowledge of the channel count of the underlying data source before it can + initialize the sound. Therefore, we need to make the resource manager wait until initialization + of the underlying data source to be initialized so we can get access to the channel count. To + do this, the MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT is forced. + + Because we're initializing the data source before the sound, there's a chance the notification + will get triggered before this function returns. This is OK, so long as the caller is aware of + it and can avoid accessing the sound from within the notification. + */ + flags = pConfig->flags | MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_WAIT_INIT; + if (pConfig->isLooping) { + flags |= MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING; + } + + pSound->pResourceManagerDataSource = (ma_resource_manager_data_source*)ma_malloc(sizeof(*pSound->pResourceManagerDataSource), &pEngine->allocationCallbacks); + if (pSound->pResourceManagerDataSource == NULL) { + return MA_OUT_OF_MEMORY; + } + + /* Removed in 0.12. Set pDoneFence on the notifications. */ + notifications = pConfig->initNotifications; + if (pConfig->pDoneFence != NULL && notifications.done.pFence == NULL) { + notifications.done.pFence = pConfig->pDoneFence; + } + + /* + We must wrap everything around the fence if one was specified. This ensures ma_fence_wait() does + not return prematurely before the sound has finished initializing. + */ + if (notifications.done.pFence) { ma_fence_acquire(notifications.done.pFence); } + { + ma_resource_manager_data_source_config resourceManagerDataSourceConfig = ma_resource_manager_data_source_config_init(); + resourceManagerDataSourceConfig.pFilePath = pConfig->pFilePath; + resourceManagerDataSourceConfig.pFilePathW = pConfig->pFilePathW; + resourceManagerDataSourceConfig.flags = flags; + resourceManagerDataSourceConfig.pNotifications = ¬ifications; + resourceManagerDataSourceConfig.initialSeekPointInPCMFrames = pConfig->initialSeekPointInPCMFrames; + resourceManagerDataSourceConfig.rangeBegInPCMFrames = pConfig->rangeBegInPCMFrames; + resourceManagerDataSourceConfig.rangeEndInPCMFrames = pConfig->rangeEndInPCMFrames; + resourceManagerDataSourceConfig.loopPointBegInPCMFrames = pConfig->loopPointBegInPCMFrames; + resourceManagerDataSourceConfig.loopPointEndInPCMFrames = pConfig->loopPointEndInPCMFrames; + resourceManagerDataSourceConfig.isLooping = (flags & MA_RESOURCE_MANAGER_DATA_SOURCE_FLAG_LOOPING) != 0; + + result = ma_resource_manager_data_source_init_ex(pEngine->pResourceManager, &resourceManagerDataSourceConfig, pSound->pResourceManagerDataSource); + if (result != MA_SUCCESS) { + goto done; + } + + pSound->ownsDataSource = MA_TRUE; /* <-- Important. Not setting this will result in the resource manager data source never getting uninitialized. */ + + /* We need to use a slightly customized version of the config so we'll need to make a copy. */ + config = *pConfig; + config.pFilePath = NULL; + config.pFilePathW = NULL; + config.pDataSource = pSound->pResourceManagerDataSource; + + result = ma_sound_init_from_data_source_internal(pEngine, &config, pSound); + if (result != MA_SUCCESS) { + ma_resource_manager_data_source_uninit(pSound->pResourceManagerDataSource); + ma_free(pSound->pResourceManagerDataSource, &pEngine->allocationCallbacks); + MA_ZERO_OBJECT(pSound); + goto done; + } + } +done: + if (notifications.done.pFence) { ma_fence_release(notifications.done.pFence); } + return result; +} + +MA_API ma_result ma_sound_init_from_file(ma_engine* pEngine, const char* pFilePath, ma_uint32 flags, ma_sound_group* pGroup, ma_fence* pDoneFence, ma_sound* pSound) +{ + ma_sound_config config; + + if (pFilePath == NULL) { + return MA_INVALID_ARGS; + } + + config = ma_sound_config_init_2(pEngine); + config.pFilePath = pFilePath; + config.flags = flags; + config.pInitialAttachment = pGroup; + config.pDoneFence = pDoneFence; + + return ma_sound_init_ex(pEngine, &config, pSound); +} + +MA_API ma_result ma_sound_init_from_file_w(ma_engine* pEngine, const wchar_t* pFilePath, ma_uint32 flags, ma_sound_group* pGroup, ma_fence* pDoneFence, ma_sound* pSound) +{ + ma_sound_config config; + + if (pFilePath == NULL) { + return MA_INVALID_ARGS; + } + + config = ma_sound_config_init_2(pEngine); + config.pFilePathW = pFilePath; + config.flags = flags; + config.pInitialAttachment = pGroup; + config.pDoneFence = pDoneFence; + + return ma_sound_init_ex(pEngine, &config, pSound); +} + +MA_API ma_result ma_sound_init_copy(ma_engine* pEngine, const ma_sound* pExistingSound, ma_uint32 flags, ma_sound_group* pGroup, ma_sound* pSound) +{ + ma_result result; + ma_sound_config config; + + result = ma_sound_preinit(pEngine, pSound); + if (result != MA_SUCCESS) { + return result; + } + + if (pExistingSound == NULL) { + return MA_INVALID_ARGS; + } + + /* Cloning only works for data buffers (not streams) that are loaded from the resource manager. */ + if (pExistingSound->pResourceManagerDataSource == NULL) { + return MA_INVALID_OPERATION; + } + + /* + We need to make a clone of the data source. If the data source is not a data buffer (i.e. a stream) + this will fail. + */ + pSound->pResourceManagerDataSource = (ma_resource_manager_data_source*)ma_malloc(sizeof(*pSound->pResourceManagerDataSource), &pEngine->allocationCallbacks); + if (pSound->pResourceManagerDataSource == NULL) { + return MA_OUT_OF_MEMORY; + } + + result = ma_resource_manager_data_source_init_copy(pEngine->pResourceManager, pExistingSound->pResourceManagerDataSource, pSound->pResourceManagerDataSource); + if (result != MA_SUCCESS) { + ma_free(pSound->pResourceManagerDataSource, &pEngine->allocationCallbacks); + return result; + } + + config = ma_sound_config_init_2(pEngine); + config.pDataSource = pSound->pResourceManagerDataSource; + config.flags = flags; + config.pInitialAttachment = pGroup; + config.monoExpansionMode = pExistingSound->engineNode.monoExpansionMode; + config.volumeSmoothTimeInPCMFrames = pExistingSound->engineNode.volumeSmoothTimeInPCMFrames; + + result = ma_sound_init_from_data_source_internal(pEngine, &config, pSound); + if (result != MA_SUCCESS) { + ma_resource_manager_data_source_uninit(pSound->pResourceManagerDataSource); + ma_free(pSound->pResourceManagerDataSource, &pEngine->allocationCallbacks); + MA_ZERO_OBJECT(pSound); + return result; + } + + /* Make sure the sound is marked as the owner of the data source or else it will never get uninitialized. */ + pSound->ownsDataSource = MA_TRUE; + + return MA_SUCCESS; +} +#endif + +MA_API ma_result ma_sound_init_from_data_source(ma_engine* pEngine, ma_data_source* pDataSource, ma_uint32 flags, ma_sound_group* pGroup, ma_sound* pSound) +{ + ma_sound_config config = ma_sound_config_init_2(pEngine); + config.pDataSource = pDataSource; + config.flags = flags; + config.pInitialAttachment = pGroup; + return ma_sound_init_ex(pEngine, &config, pSound); +} + +MA_API ma_result ma_sound_init_ex(ma_engine* pEngine, const ma_sound_config* pConfig, ma_sound* pSound) +{ + ma_result result; + + result = ma_sound_preinit(pEngine, pSound); + if (result != MA_SUCCESS) { + return result; + } + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + pSound->endCallback = pConfig->endCallback; + pSound->pEndCallbackUserData = pConfig->pEndCallbackUserData; + + /* We need to load the sound differently depending on whether or not we're loading from a file. */ +#ifndef MA_NO_RESOURCE_MANAGER + if (pConfig->pFilePath != NULL || pConfig->pFilePathW != NULL) { + return ma_sound_init_from_file_internal(pEngine, pConfig, pSound); + } else +#endif + { + /* + Getting here means we're not loading from a file. We may be loading from an already-initialized + data source, or none at all. If we aren't specifying any data source, we'll be initializing + the equivalent to a group. ma_data_source_init_from_data_source_internal() will deal with this + for us, so no special treatment required here. + */ + return ma_sound_init_from_data_source_internal(pEngine, pConfig, pSound); + } +} + +MA_API void ma_sound_uninit(ma_sound* pSound) +{ + if (pSound == NULL) { + return; + } + + /* + Always uninitialize the node first. This ensures it's detached from the graph and does not return until it has done + so which makes thread safety beyond this point trivial. + */ + ma_engine_node_uninit(&pSound->engineNode, &pSound->engineNode.pEngine->allocationCallbacks); + + /* Once the sound is detached from the group we can guarantee that it won't be referenced by the mixer thread which means it's safe for us to destroy the data source. */ +#ifndef MA_NO_RESOURCE_MANAGER + if (pSound->ownsDataSource) { + ma_resource_manager_data_source_uninit(pSound->pResourceManagerDataSource); + ma_free(pSound->pResourceManagerDataSource, &pSound->engineNode.pEngine->allocationCallbacks); + pSound->pDataSource = NULL; + } +#else + MA_ASSERT(pSound->ownsDataSource == MA_FALSE); +#endif +} + +MA_API ma_engine* ma_sound_get_engine(const ma_sound* pSound) +{ + if (pSound == NULL) { + return NULL; + } + + return pSound->engineNode.pEngine; +} + +MA_API ma_data_source* ma_sound_get_data_source(const ma_sound* pSound) +{ + if (pSound == NULL) { + return NULL; + } + + return pSound->pDataSource; +} + +MA_API ma_result ma_sound_start(ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* If the sound is already playing, do nothing. */ + if (ma_sound_is_playing(pSound)) { + return MA_SUCCESS; + } + + /* If the sound is at the end it means we want to start from the start again. */ + if (ma_sound_at_end(pSound)) { + ma_result result = ma_data_source_seek_to_pcm_frame(pSound->pDataSource, 0); + if (result != MA_SUCCESS && result != MA_NOT_IMPLEMENTED) { + return result; /* Failed to seek back to the start. */ + } + + /* Make sure we clear the end indicator. */ + ma_atomic_exchange_32(&pSound->atEnd, MA_FALSE); + } + + /* Make sure the sound is started. If there's a start delay, the sound won't actually start until the start time is reached. */ + ma_node_set_state(pSound, ma_node_state_started); + + return MA_SUCCESS; +} + +MA_API ma_result ma_sound_stop(ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* This will stop the sound immediately. Use ma_sound_set_stop_time() to stop the sound at a specific time. */ + ma_node_set_state(pSound, ma_node_state_stopped); + + return MA_SUCCESS; +} + +MA_API ma_result ma_sound_stop_with_fade_in_pcm_frames(ma_sound* pSound, ma_uint64 fadeLengthInFrames) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* Stopping with a fade out requires us to schedule the stop into the future by the fade length. */ + ma_sound_set_stop_time_with_fade_in_pcm_frames(pSound, ma_engine_get_time_in_pcm_frames(ma_sound_get_engine(pSound)) + fadeLengthInFrames, fadeLengthInFrames); + + return MA_SUCCESS; +} + +MA_API ma_result ma_sound_stop_with_fade_in_milliseconds(ma_sound* pSound, ma_uint64 fadeLengthInMilliseconds) +{ + ma_uint64 sampleRate; + + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + sampleRate = ma_engine_get_sample_rate(ma_sound_get_engine(pSound)); + + return ma_sound_stop_with_fade_in_pcm_frames(pSound, (fadeLengthInMilliseconds * sampleRate) / 1000); +} + +MA_API void ma_sound_set_volume(ma_sound* pSound, float volume) +{ + if (pSound == NULL) { + return; + } + + ma_engine_node_set_volume(&pSound->engineNode, volume); +} + +MA_API float ma_sound_get_volume(const ma_sound* pSound) +{ + float volume = 0; + + if (pSound == NULL) { + return 0; + } + + ma_engine_node_get_volume(&pSound->engineNode, &volume); + + return volume; +} + +MA_API void ma_sound_set_pan(ma_sound* pSound, float pan) +{ + if (pSound == NULL) { + return; + } + + ma_panner_set_pan(&pSound->engineNode.panner, pan); +} + +MA_API float ma_sound_get_pan(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_panner_get_pan(&pSound->engineNode.panner); +} + +MA_API void ma_sound_set_pan_mode(ma_sound* pSound, ma_pan_mode panMode) +{ + if (pSound == NULL) { + return; + } + + ma_panner_set_mode(&pSound->engineNode.panner, panMode); +} + +MA_API ma_pan_mode ma_sound_get_pan_mode(const ma_sound* pSound) +{ + if (pSound == NULL) { + return ma_pan_mode_balance; + } + + return ma_panner_get_mode(&pSound->engineNode.panner); +} + +MA_API void ma_sound_set_pitch(ma_sound* pSound, float pitch) +{ + if (pSound == NULL) { + return; + } + + if (pitch <= 0) { + return; + } + + ma_atomic_exchange_explicit_f32(&pSound->engineNode.pitch, pitch, ma_atomic_memory_order_release); +} + +MA_API float ma_sound_get_pitch(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_atomic_load_f32(&pSound->engineNode.pitch); /* Naughty const-cast for this. */ +} + +MA_API void ma_sound_set_spatialization_enabled(ma_sound* pSound, ma_bool32 enabled) +{ + if (pSound == NULL) { + return; + } + + ma_atomic_exchange_explicit_32(&pSound->engineNode.isSpatializationDisabled, !enabled, ma_atomic_memory_order_release); +} + +MA_API ma_bool32 ma_sound_is_spatialization_enabled(const ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_FALSE; + } + + return ma_engine_node_is_spatialization_enabled(&pSound->engineNode); +} + +MA_API void ma_sound_set_pinned_listener_index(ma_sound* pSound, ma_uint32 listenerIndex) +{ + if (pSound == NULL || listenerIndex >= ma_engine_get_listener_count(ma_sound_get_engine(pSound))) { + return; + } + + ma_atomic_exchange_explicit_32(&pSound->engineNode.pinnedListenerIndex, listenerIndex, ma_atomic_memory_order_release); +} + +MA_API ma_uint32 ma_sound_get_pinned_listener_index(const ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_LISTENER_INDEX_CLOSEST; + } + + return ma_atomic_load_explicit_32(&pSound->engineNode.pinnedListenerIndex, ma_atomic_memory_order_acquire); +} + +MA_API ma_uint32 ma_sound_get_listener_index(const ma_sound* pSound) +{ + ma_uint32 listenerIndex; + + if (pSound == NULL) { + return 0; + } + + listenerIndex = ma_sound_get_pinned_listener_index(pSound); + if (listenerIndex == MA_LISTENER_INDEX_CLOSEST) { + ma_vec3f position = ma_sound_get_position(pSound); + return ma_engine_find_closest_listener(ma_sound_get_engine(pSound), position.x, position.y, position.z); + } + + return listenerIndex; +} + +MA_API ma_vec3f ma_sound_get_direction_to_listener(const ma_sound* pSound) +{ + ma_vec3f relativePos; + ma_engine* pEngine; + + if (pSound == NULL) { + return ma_vec3f_init_3f(0, 0, -1); + } + + pEngine = ma_sound_get_engine(pSound); + if (pEngine == NULL) { + return ma_vec3f_init_3f(0, 0, -1); + } + + ma_spatializer_get_relative_position_and_direction(&pSound->engineNode.spatializer, &pEngine->listeners[ma_sound_get_listener_index(pSound)], &relativePos, NULL); + + return ma_vec3f_normalize(ma_vec3f_neg(relativePos)); +} + +MA_API void ma_sound_set_position(ma_sound* pSound, float x, float y, float z) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_position(&pSound->engineNode.spatializer, x, y, z); +} + +MA_API ma_vec3f ma_sound_get_position(const ma_sound* pSound) +{ + if (pSound == NULL) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_spatializer_get_position(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_direction(ma_sound* pSound, float x, float y, float z) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_direction(&pSound->engineNode.spatializer, x, y, z); +} + +MA_API ma_vec3f ma_sound_get_direction(const ma_sound* pSound) +{ + if (pSound == NULL) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_spatializer_get_direction(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_velocity(ma_sound* pSound, float x, float y, float z) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_velocity(&pSound->engineNode.spatializer, x, y, z); +} + +MA_API ma_vec3f ma_sound_get_velocity(const ma_sound* pSound) +{ + if (pSound == NULL) { + return ma_vec3f_init_3f(0, 0, 0); + } + + return ma_spatializer_get_velocity(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_attenuation_model(ma_sound* pSound, ma_attenuation_model attenuationModel) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_attenuation_model(&pSound->engineNode.spatializer, attenuationModel); +} + +MA_API ma_attenuation_model ma_sound_get_attenuation_model(const ma_sound* pSound) +{ + if (pSound == NULL) { + return ma_attenuation_model_none; + } + + return ma_spatializer_get_attenuation_model(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_positioning(ma_sound* pSound, ma_positioning positioning) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_positioning(&pSound->engineNode.spatializer, positioning); +} + +MA_API ma_positioning ma_sound_get_positioning(const ma_sound* pSound) +{ + if (pSound == NULL) { + return ma_positioning_absolute; + } + + return ma_spatializer_get_positioning(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_rolloff(ma_sound* pSound, float rolloff) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_rolloff(&pSound->engineNode.spatializer, rolloff); +} + +MA_API float ma_sound_get_rolloff(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_spatializer_get_rolloff(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_min_gain(ma_sound* pSound, float minGain) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_min_gain(&pSound->engineNode.spatializer, minGain); +} + +MA_API float ma_sound_get_min_gain(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_spatializer_get_min_gain(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_max_gain(ma_sound* pSound, float maxGain) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_max_gain(&pSound->engineNode.spatializer, maxGain); +} + +MA_API float ma_sound_get_max_gain(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_spatializer_get_max_gain(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_min_distance(ma_sound* pSound, float minDistance) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_min_distance(&pSound->engineNode.spatializer, minDistance); +} + +MA_API float ma_sound_get_min_distance(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_spatializer_get_min_distance(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_max_distance(ma_sound* pSound, float maxDistance) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_max_distance(&pSound->engineNode.spatializer, maxDistance); +} + +MA_API float ma_sound_get_max_distance(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_spatializer_get_max_distance(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_cone(ma_sound* pSound, float innerAngleInRadians, float outerAngleInRadians, float outerGain) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_cone(&pSound->engineNode.spatializer, innerAngleInRadians, outerAngleInRadians, outerGain); +} + +MA_API void ma_sound_get_cone(const ma_sound* pSound, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain) +{ + if (pInnerAngleInRadians != NULL) { + *pInnerAngleInRadians = 0; + } + + if (pOuterAngleInRadians != NULL) { + *pOuterAngleInRadians = 0; + } + + if (pOuterGain != NULL) { + *pOuterGain = 0; + } + + if (pSound == NULL) { + return; + } + + ma_spatializer_get_cone(&pSound->engineNode.spatializer, pInnerAngleInRadians, pOuterAngleInRadians, pOuterGain); +} + +MA_API void ma_sound_set_doppler_factor(ma_sound* pSound, float dopplerFactor) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_doppler_factor(&pSound->engineNode.spatializer, dopplerFactor); +} + +MA_API float ma_sound_get_doppler_factor(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_spatializer_get_doppler_factor(&pSound->engineNode.spatializer); +} + +MA_API void ma_sound_set_directional_attenuation_factor(ma_sound* pSound, float directionalAttenuationFactor) +{ + if (pSound == NULL) { + return; + } + + ma_spatializer_set_directional_attenuation_factor(&pSound->engineNode.spatializer, directionalAttenuationFactor); +} + +MA_API float ma_sound_get_directional_attenuation_factor(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 1; + } + + return ma_spatializer_get_directional_attenuation_factor(&pSound->engineNode.spatializer); +} + + +MA_API void ma_sound_set_fade_in_pcm_frames(ma_sound* pSound, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInFrames) +{ + if (pSound == NULL) { + return; + } + + ma_sound_set_fade_start_in_pcm_frames(pSound, volumeBeg, volumeEnd, fadeLengthInFrames, (~(ma_uint64)0)); +} + +MA_API void ma_sound_set_fade_in_milliseconds(ma_sound* pSound, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInMilliseconds) +{ + if (pSound == NULL) { + return; + } + + ma_sound_set_fade_in_pcm_frames(pSound, volumeBeg, volumeEnd, (fadeLengthInMilliseconds * pSound->engineNode.fader.config.sampleRate) / 1000); +} + +MA_API void ma_sound_set_fade_start_in_pcm_frames(ma_sound* pSound, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInFrames, ma_uint64 absoluteGlobalTimeInFrames) +{ + if (pSound == NULL) { + return; + } + + /* + We don't want to update the fader at this point because we need to use the engine's current time + to derive the fader's start offset. The timer is being updated on the audio thread so in order to + do this as accurately as possible we'll need to defer this to the audio thread. + */ + ma_atomic_float_set(&pSound->engineNode.fadeSettings.volumeBeg, volumeBeg); + ma_atomic_float_set(&pSound->engineNode.fadeSettings.volumeEnd, volumeEnd); + ma_atomic_uint64_set(&pSound->engineNode.fadeSettings.fadeLengthInFrames, fadeLengthInFrames); + ma_atomic_uint64_set(&pSound->engineNode.fadeSettings.absoluteGlobalTimeInFrames, absoluteGlobalTimeInFrames); +} + +MA_API void ma_sound_set_fade_start_in_milliseconds(ma_sound* pSound, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInMilliseconds, ma_uint64 absoluteGlobalTimeInMilliseconds) +{ + ma_uint32 sampleRate; + + if (pSound == NULL) { + return; + } + + sampleRate = ma_engine_get_sample_rate(ma_sound_get_engine(pSound)); + + ma_sound_set_fade_start_in_pcm_frames(pSound, volumeBeg, volumeEnd, (fadeLengthInMilliseconds * sampleRate) / 1000, (absoluteGlobalTimeInMilliseconds * sampleRate) / 1000); +} + +MA_API float ma_sound_get_current_fade_volume(const ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + return ma_fader_get_current_volume(&pSound->engineNode.fader); +} + +MA_API void ma_sound_set_start_time_in_pcm_frames(ma_sound* pSound, ma_uint64 absoluteGlobalTimeInFrames) +{ + if (pSound == NULL) { + return; + } + + ma_node_set_state_time(pSound, ma_node_state_started, absoluteGlobalTimeInFrames); +} + +MA_API void ma_sound_set_start_time_in_milliseconds(ma_sound* pSound, ma_uint64 absoluteGlobalTimeInMilliseconds) +{ + if (pSound == NULL) { + return; + } + + ma_sound_set_start_time_in_pcm_frames(pSound, absoluteGlobalTimeInMilliseconds * ma_engine_get_sample_rate(ma_sound_get_engine(pSound)) / 1000); +} + +MA_API void ma_sound_set_stop_time_in_pcm_frames(ma_sound* pSound, ma_uint64 absoluteGlobalTimeInFrames) +{ + if (pSound == NULL) { + return; + } + + ma_sound_set_stop_time_with_fade_in_pcm_frames(pSound, absoluteGlobalTimeInFrames, 0); +} + +MA_API void ma_sound_set_stop_time_in_milliseconds(ma_sound* pSound, ma_uint64 absoluteGlobalTimeInMilliseconds) +{ + if (pSound == NULL) { + return; + } + + ma_sound_set_stop_time_in_pcm_frames(pSound, absoluteGlobalTimeInMilliseconds * ma_engine_get_sample_rate(ma_sound_get_engine(pSound)) / 1000); +} + +MA_API void ma_sound_set_stop_time_with_fade_in_pcm_frames(ma_sound* pSound, ma_uint64 stopAbsoluteGlobalTimeInFrames, ma_uint64 fadeLengthInFrames) +{ + if (pSound == NULL) { + return; + } + + if (fadeLengthInFrames > 0) { + if (fadeLengthInFrames > stopAbsoluteGlobalTimeInFrames) { + fadeLengthInFrames = stopAbsoluteGlobalTimeInFrames; + } + + ma_sound_set_fade_start_in_pcm_frames(pSound, -1, 0, fadeLengthInFrames, stopAbsoluteGlobalTimeInFrames - fadeLengthInFrames); + } + + ma_node_set_state_time(pSound, ma_node_state_stopped, stopAbsoluteGlobalTimeInFrames); +} + +MA_API void ma_sound_set_stop_time_with_fade_in_milliseconds(ma_sound* pSound, ma_uint64 stopAbsoluteGlobalTimeInMilliseconds, ma_uint64 fadeLengthInMilliseconds) +{ + ma_uint32 sampleRate; + + if (pSound == NULL) { + return; + } + + sampleRate = ma_engine_get_sample_rate(ma_sound_get_engine(pSound)); + + ma_sound_set_stop_time_with_fade_in_pcm_frames(pSound, (stopAbsoluteGlobalTimeInMilliseconds * sampleRate) / 1000, (fadeLengthInMilliseconds * sampleRate) / 1000); +} + +MA_API ma_bool32 ma_sound_is_playing(const ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_FALSE; + } + + return ma_node_get_state_by_time(pSound, ma_engine_get_time_in_pcm_frames(ma_sound_get_engine(pSound))) == ma_node_state_started; +} + +MA_API ma_uint64 ma_sound_get_time_in_pcm_frames(const ma_sound* pSound) +{ + if (pSound == NULL) { + return 0; + } + + return ma_node_get_time(pSound); +} + +MA_API ma_uint64 ma_sound_get_time_in_milliseconds(const ma_sound* pSound) +{ + return ma_sound_get_time_in_pcm_frames(pSound) * 1000 / ma_engine_get_sample_rate(ma_sound_get_engine(pSound)); +} + +MA_API void ma_sound_set_looping(ma_sound* pSound, ma_bool32 isLooping) +{ + if (pSound == NULL) { + return; + } + + /* Looping is only a valid concept if the sound is backed by a data source. */ + if (pSound->pDataSource == NULL) { + return; + } + + /* The looping state needs to be applied to the data source in order for any looping to actually happen. */ + ma_data_source_set_looping(pSound->pDataSource, isLooping); +} + +MA_API ma_bool32 ma_sound_is_looping(const ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_FALSE; + } + + /* There is no notion of looping for sounds that are not backed by a data source. */ + if (pSound->pDataSource == NULL) { + return MA_FALSE; + } + + return ma_data_source_is_looping(pSound->pDataSource); +} + +MA_API ma_bool32 ma_sound_at_end(const ma_sound* pSound) +{ + if (pSound == NULL) { + return MA_FALSE; + } + + /* There is no notion of an end of a sound if it's not backed by a data source. */ + if (pSound->pDataSource == NULL) { + return MA_FALSE; + } + + return ma_sound_get_at_end(pSound); +} + +MA_API ma_result ma_sound_seek_to_pcm_frame(ma_sound* pSound, ma_uint64 frameIndex) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* Seeking is only valid for sounds that are backed by a data source. */ + if (pSound->pDataSource == NULL) { + return MA_INVALID_OPERATION; + } + + /* We can't be seeking while reading at the same time. We just set the seek target and get the mixing thread to do the actual seek. */ + ma_atomic_exchange_64(&pSound->seekTarget, frameIndex); + + return MA_SUCCESS; +} + +MA_API ma_result ma_sound_seek_to_second(ma_sound* pSound, float seekPointInSeconds) +{ + ma_uint64 frameIndex; + ma_uint32 sampleRate; + ma_result result; + + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + result = ma_sound_get_data_format(pSound, NULL, NULL, &sampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; + } + + /* We need PCM frames. We need to convert first */ + frameIndex = (ma_uint64)(seekPointInSeconds * sampleRate); + + return ma_sound_seek_to_pcm_frame(pSound, frameIndex); +} + +MA_API ma_result ma_sound_get_data_format(ma_sound* pSound, ma_format* pFormat, ma_uint32* pChannels, ma_uint32* pSampleRate, ma_channel* pChannelMap, size_t channelMapCap) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* The data format is retrieved directly from the data source if the sound is backed by one. Otherwise we pull it from the node. */ + if (pSound->pDataSource == NULL) { + ma_uint32 channels; + + if (pFormat != NULL) { + *pFormat = ma_format_f32; + } + + channels = ma_node_get_input_channels(&pSound->engineNode, 0); + if (pChannels != NULL) { + *pChannels = channels; + } + + if (pSampleRate != NULL) { + *pSampleRate = pSound->engineNode.resampler.config.sampleRateIn; + } + + if (pChannelMap != NULL) { + ma_channel_map_init_standard(ma_standard_channel_map_default, pChannelMap, channelMapCap, channels); + } + + return MA_SUCCESS; + } else { + return ma_data_source_get_data_format(pSound->pDataSource, pFormat, pChannels, pSampleRate, pChannelMap, channelMapCap); + } +} + +MA_API ma_result ma_sound_get_cursor_in_pcm_frames(ma_sound* pSound, ma_uint64* pCursor) +{ + ma_uint64 seekTarget; + + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* The notion of a cursor is only valid for sounds that are backed by a data source. */ + if (pSound->pDataSource == NULL) { + return MA_INVALID_OPERATION; + } + + seekTarget = ma_atomic_load_64(&pSound->seekTarget); + if (seekTarget != MA_SEEK_TARGET_NONE) { + *pCursor = seekTarget; + return MA_SUCCESS; + } else { + return ma_data_source_get_cursor_in_pcm_frames(pSound->pDataSource, pCursor); + } +} + +MA_API ma_result ma_sound_get_length_in_pcm_frames(ma_sound* pSound, ma_uint64* pLength) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* The notion of a sound length is only valid for sounds that are backed by a data source. */ + if (pSound->pDataSource == NULL) { + return MA_INVALID_OPERATION; + } + + return ma_data_source_get_length_in_pcm_frames(pSound->pDataSource, pLength); +} + +MA_API ma_result ma_sound_get_cursor_in_seconds(ma_sound* pSound, float* pCursor) +{ + ma_result result; + ma_uint64 cursorInPCMFrames; + ma_uint32 sampleRate; + + if (pCursor != NULL) { + *pCursor = 0; + } + + result = ma_sound_get_cursor_in_pcm_frames(pSound, &cursorInPCMFrames); + if (result != MA_SUCCESS) { + return result; + } + + result = ma_sound_get_data_format(pSound, NULL, NULL, &sampleRate, NULL, 0); + if (result != MA_SUCCESS) { + return result; + } + + /* VC6 does not support division of unsigned 64-bit integers with floating point numbers. Need to use a signed number. This shouldn't effect anything in practice. */ + *pCursor = (ma_int64)cursorInPCMFrames / (float)sampleRate; + + return MA_SUCCESS; +} + +MA_API ma_result ma_sound_get_length_in_seconds(ma_sound* pSound, float* pLength) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* The notion of a sound length is only valid for sounds that are backed by a data source. */ + if (pSound->pDataSource == NULL) { + return MA_INVALID_OPERATION; + } + + return ma_data_source_get_length_in_seconds(pSound->pDataSource, pLength); +} + +MA_API ma_result ma_sound_set_end_callback(ma_sound* pSound, ma_sound_end_proc callback, void* pUserData) +{ + if (pSound == NULL) { + return MA_INVALID_ARGS; + } + + /* The notion of an end is only valid for sounds that are backed by a data source. */ + if (pSound->pDataSource == NULL) { + return MA_INVALID_OPERATION; + } + + pSound->endCallback = callback; + pSound->pEndCallbackUserData = pUserData; + + return MA_SUCCESS; +} + + +MA_API ma_result ma_sound_group_init(ma_engine* pEngine, ma_uint32 flags, ma_sound_group* pParentGroup, ma_sound_group* pGroup) +{ + ma_sound_group_config config = ma_sound_group_config_init_2(pEngine); + config.flags = flags; + config.pInitialAttachment = pParentGroup; + return ma_sound_group_init_ex(pEngine, &config, pGroup); +} + +MA_API ma_result ma_sound_group_init_ex(ma_engine* pEngine, const ma_sound_group_config* pConfig, ma_sound_group* pGroup) +{ + ma_sound_config soundConfig; + + if (pGroup == NULL) { + return MA_INVALID_ARGS; + } + + MA_ZERO_OBJECT(pGroup); + + if (pConfig == NULL) { + return MA_INVALID_ARGS; + } + + /* A sound group is just a sound without a data source. */ + soundConfig = *pConfig; + soundConfig.pFilePath = NULL; + soundConfig.pFilePathW = NULL; + soundConfig.pDataSource = NULL; + + /* + Groups need to have spatialization disabled by default because I think it'll be pretty rare + that programs will want to spatialize groups (but not unheard of). Certainly it feels like + disabling this by default feels like the right option. Spatialization can be enabled with a + call to ma_sound_group_set_spatialization_enabled(). + */ + soundConfig.flags |= MA_SOUND_FLAG_NO_SPATIALIZATION; + + return ma_sound_init_ex(pEngine, &soundConfig, pGroup); +} + +MA_API void ma_sound_group_uninit(ma_sound_group* pGroup) +{ + ma_sound_uninit(pGroup); +} + +MA_API ma_engine* ma_sound_group_get_engine(const ma_sound_group* pGroup) +{ + return ma_sound_get_engine(pGroup); +} + +MA_API ma_result ma_sound_group_start(ma_sound_group* pGroup) +{ + return ma_sound_start(pGroup); +} + +MA_API ma_result ma_sound_group_stop(ma_sound_group* pGroup) +{ + return ma_sound_stop(pGroup); +} + +MA_API void ma_sound_group_set_volume(ma_sound_group* pGroup, float volume) +{ + ma_sound_set_volume(pGroup, volume); +} + +MA_API float ma_sound_group_get_volume(const ma_sound_group* pGroup) +{ + return ma_sound_get_volume(pGroup); +} + +MA_API void ma_sound_group_set_pan(ma_sound_group* pGroup, float pan) +{ + ma_sound_set_pan(pGroup, pan); +} + +MA_API float ma_sound_group_get_pan(const ma_sound_group* pGroup) +{ + return ma_sound_get_pan(pGroup); +} + +MA_API void ma_sound_group_set_pan_mode(ma_sound_group* pGroup, ma_pan_mode panMode) +{ + ma_sound_set_pan_mode(pGroup, panMode); +} + +MA_API ma_pan_mode ma_sound_group_get_pan_mode(const ma_sound_group* pGroup) +{ + return ma_sound_get_pan_mode(pGroup); +} + +MA_API void ma_sound_group_set_pitch(ma_sound_group* pGroup, float pitch) +{ + ma_sound_set_pitch(pGroup, pitch); +} + +MA_API float ma_sound_group_get_pitch(const ma_sound_group* pGroup) +{ + return ma_sound_get_pitch(pGroup); +} + +MA_API void ma_sound_group_set_spatialization_enabled(ma_sound_group* pGroup, ma_bool32 enabled) +{ + ma_sound_set_spatialization_enabled(pGroup, enabled); +} + +MA_API ma_bool32 ma_sound_group_is_spatialization_enabled(const ma_sound_group* pGroup) +{ + return ma_sound_is_spatialization_enabled(pGroup); +} + +MA_API void ma_sound_group_set_pinned_listener_index(ma_sound_group* pGroup, ma_uint32 listenerIndex) +{ + ma_sound_set_pinned_listener_index(pGroup, listenerIndex); +} + +MA_API ma_uint32 ma_sound_group_get_pinned_listener_index(const ma_sound_group* pGroup) +{ + return ma_sound_get_pinned_listener_index(pGroup); +} + +MA_API ma_uint32 ma_sound_group_get_listener_index(const ma_sound_group* pGroup) +{ + return ma_sound_get_listener_index(pGroup); +} + +MA_API ma_vec3f ma_sound_group_get_direction_to_listener(const ma_sound_group* pGroup) +{ + return ma_sound_get_direction_to_listener(pGroup); +} + +MA_API void ma_sound_group_set_position(ma_sound_group* pGroup, float x, float y, float z) +{ + ma_sound_set_position(pGroup, x, y, z); +} + +MA_API ma_vec3f ma_sound_group_get_position(const ma_sound_group* pGroup) +{ + return ma_sound_get_position(pGroup); +} + +MA_API void ma_sound_group_set_direction(ma_sound_group* pGroup, float x, float y, float z) +{ + ma_sound_set_direction(pGroup, x, y, z); +} + +MA_API ma_vec3f ma_sound_group_get_direction(const ma_sound_group* pGroup) +{ + return ma_sound_get_direction(pGroup); +} + +MA_API void ma_sound_group_set_velocity(ma_sound_group* pGroup, float x, float y, float z) +{ + ma_sound_set_velocity(pGroup, x, y, z); +} + +MA_API ma_vec3f ma_sound_group_get_velocity(const ma_sound_group* pGroup) +{ + return ma_sound_get_velocity(pGroup); +} + +MA_API void ma_sound_group_set_attenuation_model(ma_sound_group* pGroup, ma_attenuation_model attenuationModel) +{ + ma_sound_set_attenuation_model(pGroup, attenuationModel); +} + +MA_API ma_attenuation_model ma_sound_group_get_attenuation_model(const ma_sound_group* pGroup) +{ + return ma_sound_get_attenuation_model(pGroup); +} + +MA_API void ma_sound_group_set_positioning(ma_sound_group* pGroup, ma_positioning positioning) +{ + ma_sound_set_positioning(pGroup, positioning); +} + +MA_API ma_positioning ma_sound_group_get_positioning(const ma_sound_group* pGroup) +{ + return ma_sound_get_positioning(pGroup); +} + +MA_API void ma_sound_group_set_rolloff(ma_sound_group* pGroup, float rolloff) +{ + ma_sound_set_rolloff(pGroup, rolloff); +} + +MA_API float ma_sound_group_get_rolloff(const ma_sound_group* pGroup) +{ + return ma_sound_get_rolloff(pGroup); +} + +MA_API void ma_sound_group_set_min_gain(ma_sound_group* pGroup, float minGain) +{ + ma_sound_set_min_gain(pGroup, minGain); +} + +MA_API float ma_sound_group_get_min_gain(const ma_sound_group* pGroup) +{ + return ma_sound_get_min_gain(pGroup); +} + +MA_API void ma_sound_group_set_max_gain(ma_sound_group* pGroup, float maxGain) +{ + ma_sound_set_max_gain(pGroup, maxGain); +} + +MA_API float ma_sound_group_get_max_gain(const ma_sound_group* pGroup) +{ + return ma_sound_get_max_gain(pGroup); +} + +MA_API void ma_sound_group_set_min_distance(ma_sound_group* pGroup, float minDistance) +{ + ma_sound_set_min_distance(pGroup, minDistance); +} + +MA_API float ma_sound_group_get_min_distance(const ma_sound_group* pGroup) +{ + return ma_sound_get_min_distance(pGroup); +} + +MA_API void ma_sound_group_set_max_distance(ma_sound_group* pGroup, float maxDistance) +{ + ma_sound_set_max_distance(pGroup, maxDistance); +} + +MA_API float ma_sound_group_get_max_distance(const ma_sound_group* pGroup) +{ + return ma_sound_get_max_distance(pGroup); +} + +MA_API void ma_sound_group_set_cone(ma_sound_group* pGroup, float innerAngleInRadians, float outerAngleInRadians, float outerGain) +{ + ma_sound_set_cone(pGroup, innerAngleInRadians, outerAngleInRadians, outerGain); +} + +MA_API void ma_sound_group_get_cone(const ma_sound_group* pGroup, float* pInnerAngleInRadians, float* pOuterAngleInRadians, float* pOuterGain) +{ + ma_sound_get_cone(pGroup, pInnerAngleInRadians, pOuterAngleInRadians, pOuterGain); +} + +MA_API void ma_sound_group_set_doppler_factor(ma_sound_group* pGroup, float dopplerFactor) +{ + ma_sound_set_doppler_factor(pGroup, dopplerFactor); +} + +MA_API float ma_sound_group_get_doppler_factor(const ma_sound_group* pGroup) +{ + return ma_sound_get_doppler_factor(pGroup); +} + +MA_API void ma_sound_group_set_directional_attenuation_factor(ma_sound_group* pGroup, float directionalAttenuationFactor) +{ + ma_sound_set_directional_attenuation_factor(pGroup, directionalAttenuationFactor); +} + +MA_API float ma_sound_group_get_directional_attenuation_factor(const ma_sound_group* pGroup) +{ + return ma_sound_get_directional_attenuation_factor(pGroup); +} + +MA_API void ma_sound_group_set_fade_in_pcm_frames(ma_sound_group* pGroup, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInFrames) +{ + ma_sound_set_fade_in_pcm_frames(pGroup, volumeBeg, volumeEnd, fadeLengthInFrames); +} + +MA_API void ma_sound_group_set_fade_in_milliseconds(ma_sound_group* pGroup, float volumeBeg, float volumeEnd, ma_uint64 fadeLengthInMilliseconds) +{ + ma_sound_set_fade_in_milliseconds(pGroup, volumeBeg, volumeEnd, fadeLengthInMilliseconds); +} + +MA_API float ma_sound_group_get_current_fade_volume(ma_sound_group* pGroup) +{ + return ma_sound_get_current_fade_volume(pGroup); +} + +MA_API void ma_sound_group_set_start_time_in_pcm_frames(ma_sound_group* pGroup, ma_uint64 absoluteGlobalTimeInFrames) +{ + ma_sound_set_start_time_in_pcm_frames(pGroup, absoluteGlobalTimeInFrames); +} + +MA_API void ma_sound_group_set_start_time_in_milliseconds(ma_sound_group* pGroup, ma_uint64 absoluteGlobalTimeInMilliseconds) +{ + ma_sound_set_start_time_in_milliseconds(pGroup, absoluteGlobalTimeInMilliseconds); +} + +MA_API void ma_sound_group_set_stop_time_in_pcm_frames(ma_sound_group* pGroup, ma_uint64 absoluteGlobalTimeInFrames) +{ + ma_sound_set_stop_time_in_pcm_frames(pGroup, absoluteGlobalTimeInFrames); +} + +MA_API void ma_sound_group_set_stop_time_in_milliseconds(ma_sound_group* pGroup, ma_uint64 absoluteGlobalTimeInMilliseconds) +{ + ma_sound_set_stop_time_in_milliseconds(pGroup, absoluteGlobalTimeInMilliseconds); +} + +MA_API ma_bool32 ma_sound_group_is_playing(const ma_sound_group* pGroup) +{ + return ma_sound_is_playing(pGroup); +} + +MA_API ma_uint64 ma_sound_group_get_time_in_pcm_frames(const ma_sound_group* pGroup) +{ + return ma_sound_get_time_in_pcm_frames(pGroup); +} +#endif /* MA_NO_ENGINE */ +/* END SECTION: miniaudio_engine.c */ + + + +/************************************************************************************************************************************************************** +*************************************************************************************************************************************************************** + +Auto Generated +============== +All code below is auto-generated from a tool. This mostly consists of decoding backend implementations such as ma_dr_wav, ma_dr_flac, etc. If you find a bug in the +code below please report the bug to the respective repository for the relevant project (probably dr_libs). + +*************************************************************************************************************************************************************** +**************************************************************************************************************************************************************/ +#if !defined(MA_NO_WAV) && (!defined(MA_NO_DECODING) || !defined(MA_NO_ENCODING)) +#if !defined(MA_DR_WAV_IMPLEMENTATION) +/* dr_wav_c begin */ +#ifndef ma_dr_wav_c +#define ma_dr_wav_c +#ifdef __MRC__ +#pragma options opt off +#endif +#include +#include +#include +#ifndef MA_DR_WAV_NO_STDIO +#include +#ifndef MA_DR_WAV_NO_WCHAR +#include +#endif +#endif +#ifndef MA_DR_WAV_ASSERT +#include +#define MA_DR_WAV_ASSERT(expression) assert(expression) +#endif +#ifndef MA_DR_WAV_MALLOC +#define MA_DR_WAV_MALLOC(sz) malloc((sz)) +#endif +#ifndef MA_DR_WAV_REALLOC +#define MA_DR_WAV_REALLOC(p, sz) realloc((p), (sz)) +#endif +#ifndef MA_DR_WAV_FREE +#define MA_DR_WAV_FREE(p) free((p)) +#endif +#ifndef MA_DR_WAV_COPY_MEMORY +#define MA_DR_WAV_COPY_MEMORY(dst, src, sz) memcpy((dst), (src), (sz)) +#endif +#ifndef MA_DR_WAV_ZERO_MEMORY +#define MA_DR_WAV_ZERO_MEMORY(p, sz) memset((p), 0, (sz)) +#endif +#ifndef MA_DR_WAV_ZERO_OBJECT +#define MA_DR_WAV_ZERO_OBJECT(p) MA_DR_WAV_ZERO_MEMORY((p), sizeof(*p)) +#endif +#define ma_dr_wav_countof(x) (sizeof(x) / sizeof(x[0])) +#define ma_dr_wav_align(x, a) ((((x) + (a) - 1) / (a)) * (a)) +#define ma_dr_wav_min(a, b) (((a) < (b)) ? (a) : (b)) +#define ma_dr_wav_max(a, b) (((a) > (b)) ? (a) : (b)) +#define ma_dr_wav_clamp(x, lo, hi) (ma_dr_wav_max((lo), ma_dr_wav_min((hi), (x)))) +#define ma_dr_wav_offset_ptr(p, offset) (((ma_uint8*)(p)) + (offset)) +#define MA_DR_WAV_MAX_SIMD_VECTOR_SIZE 32 +#define MA_DR_WAV_INT64_MIN ((ma_int64) ((ma_uint64)0x80000000 << 32)) +#define MA_DR_WAV_INT64_MAX ((ma_int64)(((ma_uint64)0x7FFFFFFF << 32) | 0xFFFFFFFF)) +#if defined(_MSC_VER) && _MSC_VER >= 1400 + #define MA_DR_WAV_HAS_BYTESWAP16_INTRINSIC + #define MA_DR_WAV_HAS_BYTESWAP32_INTRINSIC + #define MA_DR_WAV_HAS_BYTESWAP64_INTRINSIC +#elif defined(__clang__) + #if defined(__has_builtin) + #if __has_builtin(__builtin_bswap16) + #define MA_DR_WAV_HAS_BYTESWAP16_INTRINSIC + #endif + #if __has_builtin(__builtin_bswap32) + #define MA_DR_WAV_HAS_BYTESWAP32_INTRINSIC + #endif + #if __has_builtin(__builtin_bswap64) + #define MA_DR_WAV_HAS_BYTESWAP64_INTRINSIC + #endif + #endif +#elif defined(__GNUC__) + #if ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 3)) + #define MA_DR_WAV_HAS_BYTESWAP32_INTRINSIC + #define MA_DR_WAV_HAS_BYTESWAP64_INTRINSIC + #endif + #if ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)) + #define MA_DR_WAV_HAS_BYTESWAP16_INTRINSIC + #endif +#endif +MA_API void ma_dr_wav_version(ma_uint32* pMajor, ma_uint32* pMinor, ma_uint32* pRevision) +{ + if (pMajor) { + *pMajor = MA_DR_WAV_VERSION_MAJOR; + } + if (pMinor) { + *pMinor = MA_DR_WAV_VERSION_MINOR; + } + if (pRevision) { + *pRevision = MA_DR_WAV_VERSION_REVISION; + } +} +MA_API const char* ma_dr_wav_version_string(void) +{ + return MA_DR_WAV_VERSION_STRING; +} +#ifndef MA_DR_WAV_MAX_SAMPLE_RATE +#define MA_DR_WAV_MAX_SAMPLE_RATE 384000 +#endif +#ifndef MA_DR_WAV_MAX_CHANNELS +#define MA_DR_WAV_MAX_CHANNELS 256 +#endif +#ifndef MA_DR_WAV_MAX_BITS_PER_SAMPLE +#define MA_DR_WAV_MAX_BITS_PER_SAMPLE 64 +#endif +static const ma_uint8 ma_dr_wavGUID_W64_RIFF[16] = {0x72,0x69,0x66,0x66, 0x2E,0x91, 0xCF,0x11, 0xA5,0xD6, 0x28,0xDB,0x04,0xC1,0x00,0x00}; +static const ma_uint8 ma_dr_wavGUID_W64_WAVE[16] = {0x77,0x61,0x76,0x65, 0xF3,0xAC, 0xD3,0x11, 0x8C,0xD1, 0x00,0xC0,0x4F,0x8E,0xDB,0x8A}; +static const ma_uint8 ma_dr_wavGUID_W64_FMT [16] = {0x66,0x6D,0x74,0x20, 0xF3,0xAC, 0xD3,0x11, 0x8C,0xD1, 0x00,0xC0,0x4F,0x8E,0xDB,0x8A}; +static const ma_uint8 ma_dr_wavGUID_W64_FACT[16] = {0x66,0x61,0x63,0x74, 0xF3,0xAC, 0xD3,0x11, 0x8C,0xD1, 0x00,0xC0,0x4F,0x8E,0xDB,0x8A}; +static const ma_uint8 ma_dr_wavGUID_W64_DATA[16] = {0x64,0x61,0x74,0x61, 0xF3,0xAC, 0xD3,0x11, 0x8C,0xD1, 0x00,0xC0,0x4F,0x8E,0xDB,0x8A}; +static MA_INLINE int ma_dr_wav__is_little_endian(void) +{ +#if defined(MA_X86) || defined(MA_X64) + return MA_TRUE; +#elif defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && __BYTE_ORDER == __LITTLE_ENDIAN + return MA_TRUE; +#else + int n = 1; + return (*(char*)&n) == 1; +#endif +} +static MA_INLINE void ma_dr_wav_bytes_to_guid(const ma_uint8* data, ma_uint8* guid) +{ + int i; + for (i = 0; i < 16; ++i) { + guid[i] = data[i]; + } +} +static MA_INLINE ma_uint16 ma_dr_wav__bswap16(ma_uint16 n) +{ +#ifdef MA_DR_WAV_HAS_BYTESWAP16_INTRINSIC + #if defined(_MSC_VER) + return _byteswap_ushort(n); + #elif defined(__GNUC__) || defined(__clang__) + return __builtin_bswap16(n); + #else + #error "This compiler does not support the byte swap intrinsic." + #endif +#else + return ((n & 0xFF00) >> 8) | + ((n & 0x00FF) << 8); +#endif +} +static MA_INLINE ma_uint32 ma_dr_wav__bswap32(ma_uint32 n) +{ +#ifdef MA_DR_WAV_HAS_BYTESWAP32_INTRINSIC + #if defined(_MSC_VER) + return _byteswap_ulong(n); + #elif defined(__GNUC__) || defined(__clang__) + #if defined(MA_ARM) && (defined(__ARM_ARCH) && __ARM_ARCH >= 6) && !defined(MA_64BIT) + ma_uint32 r; + __asm__ __volatile__ ( + #if defined(MA_64BIT) + "rev %w[out], %w[in]" : [out]"=r"(r) : [in]"r"(n) + #else + "rev %[out], %[in]" : [out]"=r"(r) : [in]"r"(n) + #endif + ); + return r; + #else + return __builtin_bswap32(n); + #endif + #else + #error "This compiler does not support the byte swap intrinsic." + #endif +#else + return ((n & 0xFF000000) >> 24) | + ((n & 0x00FF0000) >> 8) | + ((n & 0x0000FF00) << 8) | + ((n & 0x000000FF) << 24); +#endif +} +static MA_INLINE ma_uint64 ma_dr_wav__bswap64(ma_uint64 n) +{ +#ifdef MA_DR_WAV_HAS_BYTESWAP64_INTRINSIC + #if defined(_MSC_VER) + return _byteswap_uint64(n); + #elif defined(__GNUC__) || defined(__clang__) + return __builtin_bswap64(n); + #else + #error "This compiler does not support the byte swap intrinsic." + #endif +#else + return ((n & ((ma_uint64)0xFF000000 << 32)) >> 56) | + ((n & ((ma_uint64)0x00FF0000 << 32)) >> 40) | + ((n & ((ma_uint64)0x0000FF00 << 32)) >> 24) | + ((n & ((ma_uint64)0x000000FF << 32)) >> 8) | + ((n & ((ma_uint64)0xFF000000 )) << 8) | + ((n & ((ma_uint64)0x00FF0000 )) << 24) | + ((n & ((ma_uint64)0x0000FF00 )) << 40) | + ((n & ((ma_uint64)0x000000FF )) << 56); +#endif +} +static MA_INLINE ma_int16 ma_dr_wav__bswap_s16(ma_int16 n) +{ + return (ma_int16)ma_dr_wav__bswap16((ma_uint16)n); +} +static MA_INLINE void ma_dr_wav__bswap_samples_s16(ma_int16* pSamples, ma_uint64 sampleCount) +{ + ma_uint64 iSample; + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamples[iSample] = ma_dr_wav__bswap_s16(pSamples[iSample]); + } +} +static MA_INLINE void ma_dr_wav__bswap_s24(ma_uint8* p) +{ + ma_uint8 t; + t = p[0]; + p[0] = p[2]; + p[2] = t; +} +static MA_INLINE void ma_dr_wav__bswap_samples_s24(ma_uint8* pSamples, ma_uint64 sampleCount) +{ + ma_uint64 iSample; + for (iSample = 0; iSample < sampleCount; iSample += 1) { + ma_uint8* pSample = pSamples + (iSample*3); + ma_dr_wav__bswap_s24(pSample); + } +} +static MA_INLINE ma_int32 ma_dr_wav__bswap_s32(ma_int32 n) +{ + return (ma_int32)ma_dr_wav__bswap32((ma_uint32)n); +} +static MA_INLINE void ma_dr_wav__bswap_samples_s32(ma_int32* pSamples, ma_uint64 sampleCount) +{ + ma_uint64 iSample; + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamples[iSample] = ma_dr_wav__bswap_s32(pSamples[iSample]); + } +} +static MA_INLINE ma_int64 ma_dr_wav__bswap_s64(ma_int64 n) +{ + return (ma_int64)ma_dr_wav__bswap64((ma_uint64)n); +} +static MA_INLINE void ma_dr_wav__bswap_samples_s64(ma_int64* pSamples, ma_uint64 sampleCount) +{ + ma_uint64 iSample; + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamples[iSample] = ma_dr_wav__bswap_s64(pSamples[iSample]); + } +} +static MA_INLINE float ma_dr_wav__bswap_f32(float n) +{ + union { + ma_uint32 i; + float f; + } x; + x.f = n; + x.i = ma_dr_wav__bswap32(x.i); + return x.f; +} +static MA_INLINE void ma_dr_wav__bswap_samples_f32(float* pSamples, ma_uint64 sampleCount) +{ + ma_uint64 iSample; + for (iSample = 0; iSample < sampleCount; iSample += 1) { + pSamples[iSample] = ma_dr_wav__bswap_f32(pSamples[iSample]); + } +} +static MA_INLINE void ma_dr_wav__bswap_samples(void* pSamples, ma_uint64 sampleCount, ma_uint32 bytesPerSample) +{ + switch (bytesPerSample) + { + case 1: + { + } break; + case 2: + { + ma_dr_wav__bswap_samples_s16((ma_int16*)pSamples, sampleCount); + } break; + case 3: + { + ma_dr_wav__bswap_samples_s24((ma_uint8*)pSamples, sampleCount); + } break; + case 4: + { + ma_dr_wav__bswap_samples_s32((ma_int32*)pSamples, sampleCount); + } break; + case 8: + { + ma_dr_wav__bswap_samples_s64((ma_int64*)pSamples, sampleCount); + } break; + default: + { + MA_DR_WAV_ASSERT(MA_FALSE); + } break; + } +} +MA_PRIVATE MA_INLINE ma_bool32 ma_dr_wav_is_container_be(ma_dr_wav_container container) +{ + if (container == ma_dr_wav_container_rifx || container == ma_dr_wav_container_aiff) { + return MA_TRUE; + } else { + return MA_FALSE; + } +} +MA_PRIVATE MA_INLINE ma_uint16 ma_dr_wav_bytes_to_u16_le(const ma_uint8* data) +{ + return ((ma_uint16)data[0] << 0) | ((ma_uint16)data[1] << 8); +} +MA_PRIVATE MA_INLINE ma_uint16 ma_dr_wav_bytes_to_u16_be(const ma_uint8* data) +{ + return ((ma_uint16)data[1] << 0) | ((ma_uint16)data[0] << 8); +} +MA_PRIVATE MA_INLINE ma_uint16 ma_dr_wav_bytes_to_u16_ex(const ma_uint8* data, ma_dr_wav_container container) +{ + if (ma_dr_wav_is_container_be(container)) { + return ma_dr_wav_bytes_to_u16_be(data); + } else { + return ma_dr_wav_bytes_to_u16_le(data); + } +} +MA_PRIVATE MA_INLINE ma_uint32 ma_dr_wav_bytes_to_u32_le(const ma_uint8* data) +{ + return ((ma_uint32)data[0] << 0) | ((ma_uint32)data[1] << 8) | ((ma_uint32)data[2] << 16) | ((ma_uint32)data[3] << 24); +} +MA_PRIVATE MA_INLINE ma_uint32 ma_dr_wav_bytes_to_u32_be(const ma_uint8* data) +{ + return ((ma_uint32)data[3] << 0) | ((ma_uint32)data[2] << 8) | ((ma_uint32)data[1] << 16) | ((ma_uint32)data[0] << 24); +} +MA_PRIVATE MA_INLINE ma_uint32 ma_dr_wav_bytes_to_u32_ex(const ma_uint8* data, ma_dr_wav_container container) +{ + if (ma_dr_wav_is_container_be(container)) { + return ma_dr_wav_bytes_to_u32_be(data); + } else { + return ma_dr_wav_bytes_to_u32_le(data); + } +} +MA_PRIVATE ma_int64 ma_dr_wav_aiff_extented_to_s64(const ma_uint8* data) +{ + ma_uint32 exponent = ((ma_uint32)data[0] << 8) | data[1]; + ma_uint64 hi = ((ma_uint64)data[2] << 24) | ((ma_uint64)data[3] << 16) | ((ma_uint64)data[4] << 8) | ((ma_uint64)data[5] << 0); + ma_uint64 lo = ((ma_uint64)data[6] << 24) | ((ma_uint64)data[7] << 16) | ((ma_uint64)data[8] << 8) | ((ma_uint64)data[9] << 0); + ma_uint64 significand = (hi << 32) | lo; + int sign = exponent >> 15; + exponent &= 0x7FFF; + if (exponent == 0 && significand == 0) { + return 0; + } else if (exponent == 0x7FFF) { + return sign ? MA_DR_WAV_INT64_MIN : MA_DR_WAV_INT64_MAX; + } + exponent -= 16383; + if (exponent > 63) { + return sign ? MA_DR_WAV_INT64_MIN : MA_DR_WAV_INT64_MAX; + } else if (exponent < 1) { + return 0; + } + significand >>= (63 - exponent); + if (sign) { + return -(ma_int64)significand; + } else { + return (ma_int64)significand; + } +} +MA_PRIVATE void* ma_dr_wav__malloc_default(size_t sz, void* pUserData) +{ + (void)pUserData; + return MA_DR_WAV_MALLOC(sz); +} +MA_PRIVATE void* ma_dr_wav__realloc_default(void* p, size_t sz, void* pUserData) +{ + (void)pUserData; + return MA_DR_WAV_REALLOC(p, sz); +} +MA_PRIVATE void ma_dr_wav__free_default(void* p, void* pUserData) +{ + (void)pUserData; + MA_DR_WAV_FREE(p); +} +MA_PRIVATE void* ma_dr_wav__malloc_from_callbacks(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks == NULL) { + return NULL; + } + if (pAllocationCallbacks->onMalloc != NULL) { + return pAllocationCallbacks->onMalloc(sz, pAllocationCallbacks->pUserData); + } + if (pAllocationCallbacks->onRealloc != NULL) { + return pAllocationCallbacks->onRealloc(NULL, sz, pAllocationCallbacks->pUserData); + } + return NULL; +} +MA_PRIVATE void* ma_dr_wav__realloc_from_callbacks(void* p, size_t szNew, size_t szOld, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks == NULL) { + return NULL; + } + if (pAllocationCallbacks->onRealloc != NULL) { + return pAllocationCallbacks->onRealloc(p, szNew, pAllocationCallbacks->pUserData); + } + if (pAllocationCallbacks->onMalloc != NULL && pAllocationCallbacks->onFree != NULL) { + void* p2; + p2 = pAllocationCallbacks->onMalloc(szNew, pAllocationCallbacks->pUserData); + if (p2 == NULL) { + return NULL; + } + if (p != NULL) { + MA_DR_WAV_COPY_MEMORY(p2, p, szOld); + pAllocationCallbacks->onFree(p, pAllocationCallbacks->pUserData); + } + return p2; + } + return NULL; +} +MA_PRIVATE void ma_dr_wav__free_from_callbacks(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (p == NULL || pAllocationCallbacks == NULL) { + return; + } + if (pAllocationCallbacks->onFree != NULL) { + pAllocationCallbacks->onFree(p, pAllocationCallbacks->pUserData); + } +} +MA_PRIVATE ma_allocation_callbacks ma_dr_wav_copy_allocation_callbacks_or_defaults(const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks != NULL) { + return *pAllocationCallbacks; + } else { + ma_allocation_callbacks allocationCallbacks; + allocationCallbacks.pUserData = NULL; + allocationCallbacks.onMalloc = ma_dr_wav__malloc_default; + allocationCallbacks.onRealloc = ma_dr_wav__realloc_default; + allocationCallbacks.onFree = ma_dr_wav__free_default; + return allocationCallbacks; + } +} +static MA_INLINE ma_bool32 ma_dr_wav__is_compressed_format_tag(ma_uint16 formatTag) +{ + return + formatTag == MA_DR_WAVE_FORMAT_ADPCM || + formatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM; +} +MA_PRIVATE unsigned int ma_dr_wav__chunk_padding_size_riff(ma_uint64 chunkSize) +{ + return (unsigned int)(chunkSize % 2); +} +MA_PRIVATE unsigned int ma_dr_wav__chunk_padding_size_w64(ma_uint64 chunkSize) +{ + return (unsigned int)(chunkSize % 8); +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s16__msadpcm(ma_dr_wav* pWav, ma_uint64 samplesToRead, ma_int16* pBufferOut); +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s16__ima(ma_dr_wav* pWav, ma_uint64 samplesToRead, ma_int16* pBufferOut); +MA_PRIVATE ma_bool32 ma_dr_wav_init_write__internal(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount); +MA_PRIVATE ma_result ma_dr_wav__read_chunk_header(ma_dr_wav_read_proc onRead, void* pUserData, ma_dr_wav_container container, ma_uint64* pRunningBytesReadOut, ma_dr_wav_chunk_header* pHeaderOut) +{ + if (container == ma_dr_wav_container_riff || container == ma_dr_wav_container_rifx || container == ma_dr_wav_container_rf64 || container == ma_dr_wav_container_aiff) { + ma_uint8 sizeInBytes[4]; + if (onRead(pUserData, pHeaderOut->id.fourcc, 4) != 4) { + return MA_AT_END; + } + if (onRead(pUserData, sizeInBytes, 4) != 4) { + return MA_INVALID_FILE; + } + pHeaderOut->sizeInBytes = ma_dr_wav_bytes_to_u32_ex(sizeInBytes, container); + pHeaderOut->paddingSize = ma_dr_wav__chunk_padding_size_riff(pHeaderOut->sizeInBytes); + *pRunningBytesReadOut += 8; + } else if (container == ma_dr_wav_container_w64) { + ma_uint8 sizeInBytes[8]; + if (onRead(pUserData, pHeaderOut->id.guid, 16) != 16) { + return MA_AT_END; + } + if (onRead(pUserData, sizeInBytes, 8) != 8) { + return MA_INVALID_FILE; + } + pHeaderOut->sizeInBytes = ma_dr_wav_bytes_to_u64(sizeInBytes) - 24; + pHeaderOut->paddingSize = ma_dr_wav__chunk_padding_size_w64(pHeaderOut->sizeInBytes); + *pRunningBytesReadOut += 24; + } else { + return MA_INVALID_FILE; + } + return MA_SUCCESS; +} +MA_PRIVATE ma_bool32 ma_dr_wav__seek_forward(ma_dr_wav_seek_proc onSeek, ma_uint64 offset, void* pUserData) +{ + ma_uint64 bytesRemainingToSeek = offset; + while (bytesRemainingToSeek > 0) { + if (bytesRemainingToSeek > 0x7FFFFFFF) { + if (!onSeek(pUserData, 0x7FFFFFFF, ma_dr_wav_seek_origin_current)) { + return MA_FALSE; + } + bytesRemainingToSeek -= 0x7FFFFFFF; + } else { + if (!onSeek(pUserData, (int)bytesRemainingToSeek, ma_dr_wav_seek_origin_current)) { + return MA_FALSE; + } + bytesRemainingToSeek = 0; + } + } + return MA_TRUE; +} +MA_PRIVATE ma_bool32 ma_dr_wav__seek_from_start(ma_dr_wav_seek_proc onSeek, ma_uint64 offset, void* pUserData) +{ + if (offset <= 0x7FFFFFFF) { + return onSeek(pUserData, (int)offset, ma_dr_wav_seek_origin_start); + } + if (!onSeek(pUserData, 0x7FFFFFFF, ma_dr_wav_seek_origin_start)) { + return MA_FALSE; + } + offset -= 0x7FFFFFFF; + for (;;) { + if (offset <= 0x7FFFFFFF) { + return onSeek(pUserData, (int)offset, ma_dr_wav_seek_origin_current); + } + if (!onSeek(pUserData, 0x7FFFFFFF, ma_dr_wav_seek_origin_current)) { + return MA_FALSE; + } + offset -= 0x7FFFFFFF; + } +} +MA_PRIVATE size_t ma_dr_wav__on_read(ma_dr_wav_read_proc onRead, void* pUserData, void* pBufferOut, size_t bytesToRead, ma_uint64* pCursor) +{ + size_t bytesRead; + MA_DR_WAV_ASSERT(onRead != NULL); + MA_DR_WAV_ASSERT(pCursor != NULL); + bytesRead = onRead(pUserData, pBufferOut, bytesToRead); + *pCursor += bytesRead; + return bytesRead; +} +#if 0 +MA_PRIVATE ma_bool32 ma_dr_wav__on_seek(ma_dr_wav_seek_proc onSeek, void* pUserData, int offset, ma_dr_wav_seek_origin origin, ma_uint64* pCursor) +{ + MA_DR_WAV_ASSERT(onSeek != NULL); + MA_DR_WAV_ASSERT(pCursor != NULL); + if (!onSeek(pUserData, offset, origin)) { + return MA_FALSE; + } + if (origin == ma_dr_wav_seek_origin_start) { + *pCursor = offset; + } else { + *pCursor += offset; + } + return MA_TRUE; +} +#endif +#define MA_DR_WAV_SMPL_BYTES 36 +#define MA_DR_WAV_SMPL_LOOP_BYTES 24 +#define MA_DR_WAV_INST_BYTES 7 +#define MA_DR_WAV_ACID_BYTES 24 +#define MA_DR_WAV_CUE_BYTES 4 +#define MA_DR_WAV_BEXT_BYTES 602 +#define MA_DR_WAV_BEXT_DESCRIPTION_BYTES 256 +#define MA_DR_WAV_BEXT_ORIGINATOR_NAME_BYTES 32 +#define MA_DR_WAV_BEXT_ORIGINATOR_REF_BYTES 32 +#define MA_DR_WAV_BEXT_RESERVED_BYTES 180 +#define MA_DR_WAV_BEXT_UMID_BYTES 64 +#define MA_DR_WAV_CUE_POINT_BYTES 24 +#define MA_DR_WAV_LIST_LABEL_OR_NOTE_BYTES 4 +#define MA_DR_WAV_LIST_LABELLED_TEXT_BYTES 20 +#define MA_DR_WAV_METADATA_ALIGNMENT 8 +typedef enum +{ + ma_dr_wav__metadata_parser_stage_count, + ma_dr_wav__metadata_parser_stage_read +} ma_dr_wav__metadata_parser_stage; +typedef struct +{ + ma_dr_wav_read_proc onRead; + ma_dr_wav_seek_proc onSeek; + void *pReadSeekUserData; + ma_dr_wav__metadata_parser_stage stage; + ma_dr_wav_metadata *pMetadata; + ma_uint32 metadataCount; + ma_uint8 *pData; + ma_uint8 *pDataCursor; + ma_uint64 metadataCursor; + ma_uint64 extraCapacity; +} ma_dr_wav__metadata_parser; +MA_PRIVATE size_t ma_dr_wav__metadata_memory_capacity(ma_dr_wav__metadata_parser* pParser) +{ + ma_uint64 cap = sizeof(ma_dr_wav_metadata) * (ma_uint64)pParser->metadataCount + pParser->extraCapacity; + if (cap > MA_SIZE_MAX) { + return 0; + } + return (size_t)cap; +} +MA_PRIVATE ma_uint8* ma_dr_wav__metadata_get_memory(ma_dr_wav__metadata_parser* pParser, size_t size, size_t align) +{ + ma_uint8* pResult; + if (align) { + ma_uintptr modulo = (ma_uintptr)pParser->pDataCursor % align; + if (modulo != 0) { + pParser->pDataCursor += align - modulo; + } + } + pResult = pParser->pDataCursor; + MA_DR_WAV_ASSERT((pResult + size) <= (pParser->pData + ma_dr_wav__metadata_memory_capacity(pParser))); + pParser->pDataCursor += size; + return pResult; +} +MA_PRIVATE void ma_dr_wav__metadata_request_extra_memory_for_stage_2(ma_dr_wav__metadata_parser* pParser, size_t bytes, size_t align) +{ + size_t extra = bytes + (align ? (align - 1) : 0); + pParser->extraCapacity += extra; +} +MA_PRIVATE ma_result ma_dr_wav__metadata_alloc(ma_dr_wav__metadata_parser* pParser, ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pParser->extraCapacity != 0 || pParser->metadataCount != 0) { + pAllocationCallbacks->onFree(pParser->pData, pAllocationCallbacks->pUserData); + pParser->pData = (ma_uint8*)pAllocationCallbacks->onMalloc(ma_dr_wav__metadata_memory_capacity(pParser), pAllocationCallbacks->pUserData); + pParser->pDataCursor = pParser->pData; + if (pParser->pData == NULL) { + return MA_OUT_OF_MEMORY; + } + pParser->pMetadata = (ma_dr_wav_metadata*)ma_dr_wav__metadata_get_memory(pParser, sizeof(ma_dr_wav_metadata) * pParser->metadataCount, 1); + pParser->metadataCursor = 0; + } + return MA_SUCCESS; +} +MA_PRIVATE size_t ma_dr_wav__metadata_parser_read(ma_dr_wav__metadata_parser* pParser, void* pBufferOut, size_t bytesToRead, ma_uint64* pCursor) +{ + if (pCursor != NULL) { + return ma_dr_wav__on_read(pParser->onRead, pParser->pReadSeekUserData, pBufferOut, bytesToRead, pCursor); + } else { + return pParser->onRead(pParser->pReadSeekUserData, pBufferOut, bytesToRead); + } +} +MA_PRIVATE ma_uint64 ma_dr_wav__read_smpl_to_metadata_obj(ma_dr_wav__metadata_parser* pParser, const ma_dr_wav_chunk_header* pChunkHeader, ma_dr_wav_metadata* pMetadata) +{ + ma_uint8 smplHeaderData[MA_DR_WAV_SMPL_BYTES]; + ma_uint64 totalBytesRead = 0; + size_t bytesJustRead; + if (pMetadata == NULL) { + return 0; + } + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, smplHeaderData, sizeof(smplHeaderData), &totalBytesRead); + MA_DR_WAV_ASSERT(pParser->stage == ma_dr_wav__metadata_parser_stage_read); + MA_DR_WAV_ASSERT(pChunkHeader != NULL); + if (pMetadata != NULL && bytesJustRead == sizeof(smplHeaderData)) { + ma_uint32 iSampleLoop; + pMetadata->type = ma_dr_wav_metadata_type_smpl; + pMetadata->data.smpl.manufacturerId = ma_dr_wav_bytes_to_u32(smplHeaderData + 0); + pMetadata->data.smpl.productId = ma_dr_wav_bytes_to_u32(smplHeaderData + 4); + pMetadata->data.smpl.samplePeriodNanoseconds = ma_dr_wav_bytes_to_u32(smplHeaderData + 8); + pMetadata->data.smpl.midiUnityNote = ma_dr_wav_bytes_to_u32(smplHeaderData + 12); + pMetadata->data.smpl.midiPitchFraction = ma_dr_wav_bytes_to_u32(smplHeaderData + 16); + pMetadata->data.smpl.smpteFormat = ma_dr_wav_bytes_to_u32(smplHeaderData + 20); + pMetadata->data.smpl.smpteOffset = ma_dr_wav_bytes_to_u32(smplHeaderData + 24); + pMetadata->data.smpl.sampleLoopCount = ma_dr_wav_bytes_to_u32(smplHeaderData + 28); + pMetadata->data.smpl.samplerSpecificDataSizeInBytes = ma_dr_wav_bytes_to_u32(smplHeaderData + 32); + if (pMetadata->data.smpl.sampleLoopCount == (pChunkHeader->sizeInBytes - MA_DR_WAV_SMPL_BYTES) / MA_DR_WAV_SMPL_LOOP_BYTES) { + pMetadata->data.smpl.pLoops = (ma_dr_wav_smpl_loop*)ma_dr_wav__metadata_get_memory(pParser, sizeof(ma_dr_wav_smpl_loop) * pMetadata->data.smpl.sampleLoopCount, MA_DR_WAV_METADATA_ALIGNMENT); + for (iSampleLoop = 0; iSampleLoop < pMetadata->data.smpl.sampleLoopCount; ++iSampleLoop) { + ma_uint8 smplLoopData[MA_DR_WAV_SMPL_LOOP_BYTES]; + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, smplLoopData, sizeof(smplLoopData), &totalBytesRead); + if (bytesJustRead == sizeof(smplLoopData)) { + pMetadata->data.smpl.pLoops[iSampleLoop].cuePointId = ma_dr_wav_bytes_to_u32(smplLoopData + 0); + pMetadata->data.smpl.pLoops[iSampleLoop].type = ma_dr_wav_bytes_to_u32(smplLoopData + 4); + pMetadata->data.smpl.pLoops[iSampleLoop].firstSampleByteOffset = ma_dr_wav_bytes_to_u32(smplLoopData + 8); + pMetadata->data.smpl.pLoops[iSampleLoop].lastSampleByteOffset = ma_dr_wav_bytes_to_u32(smplLoopData + 12); + pMetadata->data.smpl.pLoops[iSampleLoop].sampleFraction = ma_dr_wav_bytes_to_u32(smplLoopData + 16); + pMetadata->data.smpl.pLoops[iSampleLoop].playCount = ma_dr_wav_bytes_to_u32(smplLoopData + 20); + } else { + break; + } + } + if (pMetadata->data.smpl.samplerSpecificDataSizeInBytes > 0) { + pMetadata->data.smpl.pSamplerSpecificData = ma_dr_wav__metadata_get_memory(pParser, pMetadata->data.smpl.samplerSpecificDataSizeInBytes, 1); + MA_DR_WAV_ASSERT(pMetadata->data.smpl.pSamplerSpecificData != NULL); + ma_dr_wav__metadata_parser_read(pParser, pMetadata->data.smpl.pSamplerSpecificData, pMetadata->data.smpl.samplerSpecificDataSizeInBytes, &totalBytesRead); + } + } + } + return totalBytesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav__read_cue_to_metadata_obj(ma_dr_wav__metadata_parser* pParser, const ma_dr_wav_chunk_header* pChunkHeader, ma_dr_wav_metadata* pMetadata) +{ + ma_uint8 cueHeaderSectionData[MA_DR_WAV_CUE_BYTES]; + ma_uint64 totalBytesRead = 0; + size_t bytesJustRead; + if (pMetadata == NULL) { + return 0; + } + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, cueHeaderSectionData, sizeof(cueHeaderSectionData), &totalBytesRead); + MA_DR_WAV_ASSERT(pParser->stage == ma_dr_wav__metadata_parser_stage_read); + if (bytesJustRead == sizeof(cueHeaderSectionData)) { + pMetadata->type = ma_dr_wav_metadata_type_cue; + pMetadata->data.cue.cuePointCount = ma_dr_wav_bytes_to_u32(cueHeaderSectionData); + if (pMetadata->data.cue.cuePointCount == (pChunkHeader->sizeInBytes - MA_DR_WAV_CUE_BYTES) / MA_DR_WAV_CUE_POINT_BYTES) { + pMetadata->data.cue.pCuePoints = (ma_dr_wav_cue_point*)ma_dr_wav__metadata_get_memory(pParser, sizeof(ma_dr_wav_cue_point) * pMetadata->data.cue.cuePointCount, MA_DR_WAV_METADATA_ALIGNMENT); + MA_DR_WAV_ASSERT(pMetadata->data.cue.pCuePoints != NULL); + if (pMetadata->data.cue.cuePointCount > 0) { + ma_uint32 iCuePoint; + for (iCuePoint = 0; iCuePoint < pMetadata->data.cue.cuePointCount; ++iCuePoint) { + ma_uint8 cuePointData[MA_DR_WAV_CUE_POINT_BYTES]; + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, cuePointData, sizeof(cuePointData), &totalBytesRead); + if (bytesJustRead == sizeof(cuePointData)) { + pMetadata->data.cue.pCuePoints[iCuePoint].id = ma_dr_wav_bytes_to_u32(cuePointData + 0); + pMetadata->data.cue.pCuePoints[iCuePoint].playOrderPosition = ma_dr_wav_bytes_to_u32(cuePointData + 4); + pMetadata->data.cue.pCuePoints[iCuePoint].dataChunkId[0] = cuePointData[8]; + pMetadata->data.cue.pCuePoints[iCuePoint].dataChunkId[1] = cuePointData[9]; + pMetadata->data.cue.pCuePoints[iCuePoint].dataChunkId[2] = cuePointData[10]; + pMetadata->data.cue.pCuePoints[iCuePoint].dataChunkId[3] = cuePointData[11]; + pMetadata->data.cue.pCuePoints[iCuePoint].chunkStart = ma_dr_wav_bytes_to_u32(cuePointData + 12); + pMetadata->data.cue.pCuePoints[iCuePoint].blockStart = ma_dr_wav_bytes_to_u32(cuePointData + 16); + pMetadata->data.cue.pCuePoints[iCuePoint].sampleByteOffset = ma_dr_wav_bytes_to_u32(cuePointData + 20); + } else { + break; + } + } + } + } + } + return totalBytesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav__read_inst_to_metadata_obj(ma_dr_wav__metadata_parser* pParser, ma_dr_wav_metadata* pMetadata) +{ + ma_uint8 instData[MA_DR_WAV_INST_BYTES]; + ma_uint64 bytesRead; + if (pMetadata == NULL) { + return 0; + } + bytesRead = ma_dr_wav__metadata_parser_read(pParser, instData, sizeof(instData), NULL); + MA_DR_WAV_ASSERT(pParser->stage == ma_dr_wav__metadata_parser_stage_read); + if (bytesRead == sizeof(instData)) { + pMetadata->type = ma_dr_wav_metadata_type_inst; + pMetadata->data.inst.midiUnityNote = (ma_int8)instData[0]; + pMetadata->data.inst.fineTuneCents = (ma_int8)instData[1]; + pMetadata->data.inst.gainDecibels = (ma_int8)instData[2]; + pMetadata->data.inst.lowNote = (ma_int8)instData[3]; + pMetadata->data.inst.highNote = (ma_int8)instData[4]; + pMetadata->data.inst.lowVelocity = (ma_int8)instData[5]; + pMetadata->data.inst.highVelocity = (ma_int8)instData[6]; + } + return bytesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav__read_acid_to_metadata_obj(ma_dr_wav__metadata_parser* pParser, ma_dr_wav_metadata* pMetadata) +{ + ma_uint8 acidData[MA_DR_WAV_ACID_BYTES]; + ma_uint64 bytesRead; + if (pMetadata == NULL) { + return 0; + } + bytesRead = ma_dr_wav__metadata_parser_read(pParser, acidData, sizeof(acidData), NULL); + MA_DR_WAV_ASSERT(pParser->stage == ma_dr_wav__metadata_parser_stage_read); + if (bytesRead == sizeof(acidData)) { + pMetadata->type = ma_dr_wav_metadata_type_acid; + pMetadata->data.acid.flags = ma_dr_wav_bytes_to_u32(acidData + 0); + pMetadata->data.acid.midiUnityNote = ma_dr_wav_bytes_to_u16(acidData + 4); + pMetadata->data.acid.reserved1 = ma_dr_wav_bytes_to_u16(acidData + 6); + pMetadata->data.acid.reserved2 = ma_dr_wav_bytes_to_f32(acidData + 8); + pMetadata->data.acid.numBeats = ma_dr_wav_bytes_to_u32(acidData + 12); + pMetadata->data.acid.meterDenominator = ma_dr_wav_bytes_to_u16(acidData + 16); + pMetadata->data.acid.meterNumerator = ma_dr_wav_bytes_to_u16(acidData + 18); + pMetadata->data.acid.tempo = ma_dr_wav_bytes_to_f32(acidData + 20); + } + return bytesRead; +} +MA_PRIVATE size_t ma_dr_wav__strlen(const char* str) +{ + size_t result = 0; + while (*str++) { + result += 1; + } + return result; +} +MA_PRIVATE size_t ma_dr_wav__strlen_clamped(const char* str, size_t maxToRead) +{ + size_t result = 0; + while (*str++ && result < maxToRead) { + result += 1; + } + return result; +} +MA_PRIVATE char* ma_dr_wav__metadata_copy_string(ma_dr_wav__metadata_parser* pParser, const char* str, size_t maxToRead) +{ + size_t len = ma_dr_wav__strlen_clamped(str, maxToRead); + if (len) { + char* result = (char*)ma_dr_wav__metadata_get_memory(pParser, len + 1, 1); + MA_DR_WAV_ASSERT(result != NULL); + MA_DR_WAV_COPY_MEMORY(result, str, len); + result[len] = '\0'; + return result; + } else { + return NULL; + } +} +typedef struct +{ + const void* pBuffer; + size_t sizeInBytes; + size_t cursor; +} ma_dr_wav_buffer_reader; +MA_PRIVATE ma_result ma_dr_wav_buffer_reader_init(const void* pBuffer, size_t sizeInBytes, ma_dr_wav_buffer_reader* pReader) +{ + MA_DR_WAV_ASSERT(pBuffer != NULL); + MA_DR_WAV_ASSERT(pReader != NULL); + MA_DR_WAV_ZERO_OBJECT(pReader); + pReader->pBuffer = pBuffer; + pReader->sizeInBytes = sizeInBytes; + pReader->cursor = 0; + return MA_SUCCESS; +} +MA_PRIVATE const void* ma_dr_wav_buffer_reader_ptr(const ma_dr_wav_buffer_reader* pReader) +{ + MA_DR_WAV_ASSERT(pReader != NULL); + return ma_dr_wav_offset_ptr(pReader->pBuffer, pReader->cursor); +} +MA_PRIVATE ma_result ma_dr_wav_buffer_reader_seek(ma_dr_wav_buffer_reader* pReader, size_t bytesToSeek) +{ + MA_DR_WAV_ASSERT(pReader != NULL); + if (pReader->cursor + bytesToSeek > pReader->sizeInBytes) { + return MA_BAD_SEEK; + } + pReader->cursor += bytesToSeek; + return MA_SUCCESS; +} +MA_PRIVATE ma_result ma_dr_wav_buffer_reader_read(ma_dr_wav_buffer_reader* pReader, void* pDst, size_t bytesToRead, size_t* pBytesRead) +{ + ma_result result = MA_SUCCESS; + size_t bytesRemaining; + MA_DR_WAV_ASSERT(pReader != NULL); + if (pBytesRead != NULL) { + *pBytesRead = 0; + } + bytesRemaining = (pReader->sizeInBytes - pReader->cursor); + if (bytesToRead > bytesRemaining) { + bytesToRead = bytesRemaining; + } + if (pDst == NULL) { + result = ma_dr_wav_buffer_reader_seek(pReader, bytesToRead); + } else { + MA_DR_WAV_COPY_MEMORY(pDst, ma_dr_wav_buffer_reader_ptr(pReader), bytesToRead); + pReader->cursor += bytesToRead; + } + MA_DR_WAV_ASSERT(pReader->cursor <= pReader->sizeInBytes); + if (result == MA_SUCCESS) { + if (pBytesRead != NULL) { + *pBytesRead = bytesToRead; + } + } + return MA_SUCCESS; +} +MA_PRIVATE ma_result ma_dr_wav_buffer_reader_read_u16(ma_dr_wav_buffer_reader* pReader, ma_uint16* pDst) +{ + ma_result result; + size_t bytesRead; + ma_uint8 data[2]; + MA_DR_WAV_ASSERT(pReader != NULL); + MA_DR_WAV_ASSERT(pDst != NULL); + *pDst = 0; + result = ma_dr_wav_buffer_reader_read(pReader, data, sizeof(*pDst), &bytesRead); + if (result != MA_SUCCESS || bytesRead != sizeof(*pDst)) { + return result; + } + *pDst = ma_dr_wav_bytes_to_u16(data); + return MA_SUCCESS; +} +MA_PRIVATE ma_result ma_dr_wav_buffer_reader_read_u32(ma_dr_wav_buffer_reader* pReader, ma_uint32* pDst) +{ + ma_result result; + size_t bytesRead; + ma_uint8 data[4]; + MA_DR_WAV_ASSERT(pReader != NULL); + MA_DR_WAV_ASSERT(pDst != NULL); + *pDst = 0; + result = ma_dr_wav_buffer_reader_read(pReader, data, sizeof(*pDst), &bytesRead); + if (result != MA_SUCCESS || bytesRead != sizeof(*pDst)) { + return result; + } + *pDst = ma_dr_wav_bytes_to_u32(data); + return MA_SUCCESS; +} +MA_PRIVATE ma_uint64 ma_dr_wav__read_bext_to_metadata_obj(ma_dr_wav__metadata_parser* pParser, ma_dr_wav_metadata* pMetadata, ma_uint64 chunkSize) +{ + ma_uint8 bextData[MA_DR_WAV_BEXT_BYTES]; + size_t bytesRead = ma_dr_wav__metadata_parser_read(pParser, bextData, sizeof(bextData), NULL); + MA_DR_WAV_ASSERT(pParser->stage == ma_dr_wav__metadata_parser_stage_read); + if (bytesRead == sizeof(bextData)) { + ma_dr_wav_buffer_reader reader; + ma_uint32 timeReferenceLow; + ma_uint32 timeReferenceHigh; + size_t extraBytes; + pMetadata->type = ma_dr_wav_metadata_type_bext; + if (ma_dr_wav_buffer_reader_init(bextData, bytesRead, &reader) == MA_SUCCESS) { + pMetadata->data.bext.pDescription = ma_dr_wav__metadata_copy_string(pParser, (const char*)ma_dr_wav_buffer_reader_ptr(&reader), MA_DR_WAV_BEXT_DESCRIPTION_BYTES); + ma_dr_wav_buffer_reader_seek(&reader, MA_DR_WAV_BEXT_DESCRIPTION_BYTES); + pMetadata->data.bext.pOriginatorName = ma_dr_wav__metadata_copy_string(pParser, (const char*)ma_dr_wav_buffer_reader_ptr(&reader), MA_DR_WAV_BEXT_ORIGINATOR_NAME_BYTES); + ma_dr_wav_buffer_reader_seek(&reader, MA_DR_WAV_BEXT_ORIGINATOR_NAME_BYTES); + pMetadata->data.bext.pOriginatorReference = ma_dr_wav__metadata_copy_string(pParser, (const char*)ma_dr_wav_buffer_reader_ptr(&reader), MA_DR_WAV_BEXT_ORIGINATOR_REF_BYTES); + ma_dr_wav_buffer_reader_seek(&reader, MA_DR_WAV_BEXT_ORIGINATOR_REF_BYTES); + ma_dr_wav_buffer_reader_read(&reader, pMetadata->data.bext.pOriginationDate, sizeof(pMetadata->data.bext.pOriginationDate), NULL); + ma_dr_wav_buffer_reader_read(&reader, pMetadata->data.bext.pOriginationTime, sizeof(pMetadata->data.bext.pOriginationTime), NULL); + ma_dr_wav_buffer_reader_read_u32(&reader, &timeReferenceLow); + ma_dr_wav_buffer_reader_read_u32(&reader, &timeReferenceHigh); + pMetadata->data.bext.timeReference = ((ma_uint64)timeReferenceHigh << 32) + timeReferenceLow; + ma_dr_wav_buffer_reader_read_u16(&reader, &pMetadata->data.bext.version); + pMetadata->data.bext.pUMID = ma_dr_wav__metadata_get_memory(pParser, MA_DR_WAV_BEXT_UMID_BYTES, 1); + ma_dr_wav_buffer_reader_read(&reader, pMetadata->data.bext.pUMID, MA_DR_WAV_BEXT_UMID_BYTES, NULL); + ma_dr_wav_buffer_reader_read_u16(&reader, &pMetadata->data.bext.loudnessValue); + ma_dr_wav_buffer_reader_read_u16(&reader, &pMetadata->data.bext.loudnessRange); + ma_dr_wav_buffer_reader_read_u16(&reader, &pMetadata->data.bext.maxTruePeakLevel); + ma_dr_wav_buffer_reader_read_u16(&reader, &pMetadata->data.bext.maxMomentaryLoudness); + ma_dr_wav_buffer_reader_read_u16(&reader, &pMetadata->data.bext.maxShortTermLoudness); + MA_DR_WAV_ASSERT((ma_dr_wav_offset_ptr(ma_dr_wav_buffer_reader_ptr(&reader), MA_DR_WAV_BEXT_RESERVED_BYTES)) == (bextData + MA_DR_WAV_BEXT_BYTES)); + extraBytes = (size_t)(chunkSize - MA_DR_WAV_BEXT_BYTES); + if (extraBytes > 0) { + pMetadata->data.bext.pCodingHistory = (char*)ma_dr_wav__metadata_get_memory(pParser, extraBytes + 1, 1); + MA_DR_WAV_ASSERT(pMetadata->data.bext.pCodingHistory != NULL); + bytesRead += ma_dr_wav__metadata_parser_read(pParser, pMetadata->data.bext.pCodingHistory, extraBytes, NULL); + pMetadata->data.bext.codingHistorySize = (ma_uint32)ma_dr_wav__strlen(pMetadata->data.bext.pCodingHistory); + } else { + pMetadata->data.bext.pCodingHistory = NULL; + pMetadata->data.bext.codingHistorySize = 0; + } + } + } + return bytesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav__read_list_label_or_note_to_metadata_obj(ma_dr_wav__metadata_parser* pParser, ma_dr_wav_metadata* pMetadata, ma_uint64 chunkSize, ma_dr_wav_metadata_type type) +{ + ma_uint8 cueIDBuffer[MA_DR_WAV_LIST_LABEL_OR_NOTE_BYTES]; + ma_uint64 totalBytesRead = 0; + size_t bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, cueIDBuffer, sizeof(cueIDBuffer), &totalBytesRead); + MA_DR_WAV_ASSERT(pParser->stage == ma_dr_wav__metadata_parser_stage_read); + if (bytesJustRead == sizeof(cueIDBuffer)) { + ma_uint32 sizeIncludingNullTerminator; + pMetadata->type = type; + pMetadata->data.labelOrNote.cuePointId = ma_dr_wav_bytes_to_u32(cueIDBuffer); + sizeIncludingNullTerminator = (ma_uint32)chunkSize - MA_DR_WAV_LIST_LABEL_OR_NOTE_BYTES; + if (sizeIncludingNullTerminator > 0) { + pMetadata->data.labelOrNote.stringLength = sizeIncludingNullTerminator - 1; + pMetadata->data.labelOrNote.pString = (char*)ma_dr_wav__metadata_get_memory(pParser, sizeIncludingNullTerminator, 1); + MA_DR_WAV_ASSERT(pMetadata->data.labelOrNote.pString != NULL); + ma_dr_wav__metadata_parser_read(pParser, pMetadata->data.labelOrNote.pString, sizeIncludingNullTerminator, &totalBytesRead); + } else { + pMetadata->data.labelOrNote.stringLength = 0; + pMetadata->data.labelOrNote.pString = NULL; + } + } + return totalBytesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav__read_list_labelled_cue_region_to_metadata_obj(ma_dr_wav__metadata_parser* pParser, ma_dr_wav_metadata* pMetadata, ma_uint64 chunkSize) +{ + ma_uint8 buffer[MA_DR_WAV_LIST_LABELLED_TEXT_BYTES]; + ma_uint64 totalBytesRead = 0; + size_t bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, buffer, sizeof(buffer), &totalBytesRead); + MA_DR_WAV_ASSERT(pParser->stage == ma_dr_wav__metadata_parser_stage_read); + if (bytesJustRead == sizeof(buffer)) { + ma_uint32 sizeIncludingNullTerminator; + pMetadata->type = ma_dr_wav_metadata_type_list_labelled_cue_region; + pMetadata->data.labelledCueRegion.cuePointId = ma_dr_wav_bytes_to_u32(buffer + 0); + pMetadata->data.labelledCueRegion.sampleLength = ma_dr_wav_bytes_to_u32(buffer + 4); + pMetadata->data.labelledCueRegion.purposeId[0] = buffer[8]; + pMetadata->data.labelledCueRegion.purposeId[1] = buffer[9]; + pMetadata->data.labelledCueRegion.purposeId[2] = buffer[10]; + pMetadata->data.labelledCueRegion.purposeId[3] = buffer[11]; + pMetadata->data.labelledCueRegion.country = ma_dr_wav_bytes_to_u16(buffer + 12); + pMetadata->data.labelledCueRegion.language = ma_dr_wav_bytes_to_u16(buffer + 14); + pMetadata->data.labelledCueRegion.dialect = ma_dr_wav_bytes_to_u16(buffer + 16); + pMetadata->data.labelledCueRegion.codePage = ma_dr_wav_bytes_to_u16(buffer + 18); + sizeIncludingNullTerminator = (ma_uint32)chunkSize - MA_DR_WAV_LIST_LABELLED_TEXT_BYTES; + if (sizeIncludingNullTerminator > 0) { + pMetadata->data.labelledCueRegion.stringLength = sizeIncludingNullTerminator - 1; + pMetadata->data.labelledCueRegion.pString = (char*)ma_dr_wav__metadata_get_memory(pParser, sizeIncludingNullTerminator, 1); + MA_DR_WAV_ASSERT(pMetadata->data.labelledCueRegion.pString != NULL); + ma_dr_wav__metadata_parser_read(pParser, pMetadata->data.labelledCueRegion.pString, sizeIncludingNullTerminator, &totalBytesRead); + } else { + pMetadata->data.labelledCueRegion.stringLength = 0; + pMetadata->data.labelledCueRegion.pString = NULL; + } + } + return totalBytesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav__metadata_process_info_text_chunk(ma_dr_wav__metadata_parser* pParser, ma_uint64 chunkSize, ma_dr_wav_metadata_type type) +{ + ma_uint64 bytesRead = 0; + ma_uint32 stringSizeWithNullTerminator = (ma_uint32)chunkSize; + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + pParser->metadataCount += 1; + ma_dr_wav__metadata_request_extra_memory_for_stage_2(pParser, stringSizeWithNullTerminator, 1); + } else { + ma_dr_wav_metadata* pMetadata = &pParser->pMetadata[pParser->metadataCursor]; + pMetadata->type = type; + if (stringSizeWithNullTerminator > 0) { + pMetadata->data.infoText.stringLength = stringSizeWithNullTerminator - 1; + pMetadata->data.infoText.pString = (char*)ma_dr_wav__metadata_get_memory(pParser, stringSizeWithNullTerminator, 1); + MA_DR_WAV_ASSERT(pMetadata->data.infoText.pString != NULL); + bytesRead = ma_dr_wav__metadata_parser_read(pParser, pMetadata->data.infoText.pString, (size_t)stringSizeWithNullTerminator, NULL); + if (bytesRead == chunkSize) { + pParser->metadataCursor += 1; + } else { + } + } else { + pMetadata->data.infoText.stringLength = 0; + pMetadata->data.infoText.pString = NULL; + pParser->metadataCursor += 1; + } + } + return bytesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav__metadata_process_unknown_chunk(ma_dr_wav__metadata_parser* pParser, const ma_uint8* pChunkId, ma_uint64 chunkSize, ma_dr_wav_metadata_location location) +{ + ma_uint64 bytesRead = 0; + if (location == ma_dr_wav_metadata_location_invalid) { + return 0; + } + if (ma_dr_wav_fourcc_equal(pChunkId, "data") || ma_dr_wav_fourcc_equal(pChunkId, "fmt ") || ma_dr_wav_fourcc_equal(pChunkId, "fact")) { + return 0; + } + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + pParser->metadataCount += 1; + ma_dr_wav__metadata_request_extra_memory_for_stage_2(pParser, (size_t)chunkSize, 1); + } else { + ma_dr_wav_metadata* pMetadata = &pParser->pMetadata[pParser->metadataCursor]; + pMetadata->type = ma_dr_wav_metadata_type_unknown; + pMetadata->data.unknown.chunkLocation = location; + pMetadata->data.unknown.id[0] = pChunkId[0]; + pMetadata->data.unknown.id[1] = pChunkId[1]; + pMetadata->data.unknown.id[2] = pChunkId[2]; + pMetadata->data.unknown.id[3] = pChunkId[3]; + pMetadata->data.unknown.dataSizeInBytes = (ma_uint32)chunkSize; + pMetadata->data.unknown.pData = (ma_uint8 *)ma_dr_wav__metadata_get_memory(pParser, (size_t)chunkSize, 1); + MA_DR_WAV_ASSERT(pMetadata->data.unknown.pData != NULL); + bytesRead = ma_dr_wav__metadata_parser_read(pParser, pMetadata->data.unknown.pData, pMetadata->data.unknown.dataSizeInBytes, NULL); + if (bytesRead == pMetadata->data.unknown.dataSizeInBytes) { + pParser->metadataCursor += 1; + } else { + } + } + return bytesRead; +} +MA_PRIVATE ma_bool32 ma_dr_wav__chunk_matches(ma_dr_wav_metadata_type allowedMetadataTypes, const ma_uint8* pChunkID, ma_dr_wav_metadata_type type, const char* pID) +{ + return (allowedMetadataTypes & type) && ma_dr_wav_fourcc_equal(pChunkID, pID); +} +MA_PRIVATE ma_uint64 ma_dr_wav__metadata_process_chunk(ma_dr_wav__metadata_parser* pParser, const ma_dr_wav_chunk_header* pChunkHeader, ma_dr_wav_metadata_type allowedMetadataTypes) +{ + const ma_uint8 *pChunkID = pChunkHeader->id.fourcc; + ma_uint64 bytesRead = 0; + if (ma_dr_wav__chunk_matches(allowedMetadataTypes, pChunkID, ma_dr_wav_metadata_type_smpl, "smpl")) { + if (pChunkHeader->sizeInBytes >= MA_DR_WAV_SMPL_BYTES) { + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + ma_uint8 buffer[4]; + size_t bytesJustRead; + if (!pParser->onSeek(pParser->pReadSeekUserData, 28, ma_dr_wav_seek_origin_current)) { + return bytesRead; + } + bytesRead += 28; + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, buffer, sizeof(buffer), &bytesRead); + if (bytesJustRead == sizeof(buffer)) { + ma_uint32 loopCount = ma_dr_wav_bytes_to_u32(buffer); + ma_uint64 calculatedLoopCount; + calculatedLoopCount = (pChunkHeader->sizeInBytes - MA_DR_WAV_SMPL_BYTES) / MA_DR_WAV_SMPL_LOOP_BYTES; + if (calculatedLoopCount == loopCount) { + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, buffer, sizeof(buffer), &bytesRead); + if (bytesJustRead == sizeof(buffer)) { + ma_uint32 samplerSpecificDataSizeInBytes = ma_dr_wav_bytes_to_u32(buffer); + pParser->metadataCount += 1; + ma_dr_wav__metadata_request_extra_memory_for_stage_2(pParser, sizeof(ma_dr_wav_smpl_loop) * loopCount, MA_DR_WAV_METADATA_ALIGNMENT); + ma_dr_wav__metadata_request_extra_memory_for_stage_2(pParser, samplerSpecificDataSizeInBytes, 1); + } + } else { + } + } + } else { + bytesRead = ma_dr_wav__read_smpl_to_metadata_obj(pParser, pChunkHeader, &pParser->pMetadata[pParser->metadataCursor]); + if (bytesRead == pChunkHeader->sizeInBytes) { + pParser->metadataCursor += 1; + } else { + } + } + } else { + } + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, pChunkID, ma_dr_wav_metadata_type_inst, "inst")) { + if (pChunkHeader->sizeInBytes == MA_DR_WAV_INST_BYTES) { + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + pParser->metadataCount += 1; + } else { + bytesRead = ma_dr_wav__read_inst_to_metadata_obj(pParser, &pParser->pMetadata[pParser->metadataCursor]); + if (bytesRead == pChunkHeader->sizeInBytes) { + pParser->metadataCursor += 1; + } else { + } + } + } else { + } + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, pChunkID, ma_dr_wav_metadata_type_acid, "acid")) { + if (pChunkHeader->sizeInBytes == MA_DR_WAV_ACID_BYTES) { + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + pParser->metadataCount += 1; + } else { + bytesRead = ma_dr_wav__read_acid_to_metadata_obj(pParser, &pParser->pMetadata[pParser->metadataCursor]); + if (bytesRead == pChunkHeader->sizeInBytes) { + pParser->metadataCursor += 1; + } else { + } + } + } else { + } + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, pChunkID, ma_dr_wav_metadata_type_cue, "cue ")) { + if (pChunkHeader->sizeInBytes >= MA_DR_WAV_CUE_BYTES) { + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + size_t cueCount; + pParser->metadataCount += 1; + cueCount = (size_t)(pChunkHeader->sizeInBytes - MA_DR_WAV_CUE_BYTES) / MA_DR_WAV_CUE_POINT_BYTES; + ma_dr_wav__metadata_request_extra_memory_for_stage_2(pParser, sizeof(ma_dr_wav_cue_point) * cueCount, MA_DR_WAV_METADATA_ALIGNMENT); + } else { + bytesRead = ma_dr_wav__read_cue_to_metadata_obj(pParser, pChunkHeader, &pParser->pMetadata[pParser->metadataCursor]); + if (bytesRead == pChunkHeader->sizeInBytes) { + pParser->metadataCursor += 1; + } else { + } + } + } else { + } + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, pChunkID, ma_dr_wav_metadata_type_bext, "bext")) { + if (pChunkHeader->sizeInBytes >= MA_DR_WAV_BEXT_BYTES) { + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + char buffer[MA_DR_WAV_BEXT_DESCRIPTION_BYTES + 1]; + size_t allocSizeNeeded = MA_DR_WAV_BEXT_UMID_BYTES; + size_t bytesJustRead; + buffer[MA_DR_WAV_BEXT_DESCRIPTION_BYTES] = '\0'; + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, buffer, MA_DR_WAV_BEXT_DESCRIPTION_BYTES, &bytesRead); + if (bytesJustRead != MA_DR_WAV_BEXT_DESCRIPTION_BYTES) { + return bytesRead; + } + allocSizeNeeded += ma_dr_wav__strlen(buffer) + 1; + buffer[MA_DR_WAV_BEXT_ORIGINATOR_NAME_BYTES] = '\0'; + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, buffer, MA_DR_WAV_BEXT_ORIGINATOR_NAME_BYTES, &bytesRead); + if (bytesJustRead != MA_DR_WAV_BEXT_ORIGINATOR_NAME_BYTES) { + return bytesRead; + } + allocSizeNeeded += ma_dr_wav__strlen(buffer) + 1; + buffer[MA_DR_WAV_BEXT_ORIGINATOR_REF_BYTES] = '\0'; + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, buffer, MA_DR_WAV_BEXT_ORIGINATOR_REF_BYTES, &bytesRead); + if (bytesJustRead != MA_DR_WAV_BEXT_ORIGINATOR_REF_BYTES) { + return bytesRead; + } + allocSizeNeeded += ma_dr_wav__strlen(buffer) + 1; + allocSizeNeeded += (size_t)pChunkHeader->sizeInBytes - MA_DR_WAV_BEXT_BYTES; + ma_dr_wav__metadata_request_extra_memory_for_stage_2(pParser, allocSizeNeeded, 1); + pParser->metadataCount += 1; + } else { + bytesRead = ma_dr_wav__read_bext_to_metadata_obj(pParser, &pParser->pMetadata[pParser->metadataCursor], pChunkHeader->sizeInBytes); + if (bytesRead == pChunkHeader->sizeInBytes) { + pParser->metadataCursor += 1; + } else { + } + } + } else { + } + } else if (ma_dr_wav_fourcc_equal(pChunkID, "LIST") || ma_dr_wav_fourcc_equal(pChunkID, "list")) { + ma_dr_wav_metadata_location listType = ma_dr_wav_metadata_location_invalid; + while (bytesRead < pChunkHeader->sizeInBytes) { + ma_uint8 subchunkId[4]; + ma_uint8 subchunkSizeBuffer[4]; + ma_uint64 subchunkDataSize; + ma_uint64 subchunkBytesRead = 0; + ma_uint64 bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, subchunkId, sizeof(subchunkId), &bytesRead); + if (bytesJustRead != sizeof(subchunkId)) { + break; + } + if (ma_dr_wav_fourcc_equal(subchunkId, "adtl")) { + listType = ma_dr_wav_metadata_location_inside_adtl_list; + continue; + } else if (ma_dr_wav_fourcc_equal(subchunkId, "INFO")) { + listType = ma_dr_wav_metadata_location_inside_info_list; + continue; + } + bytesJustRead = ma_dr_wav__metadata_parser_read(pParser, subchunkSizeBuffer, sizeof(subchunkSizeBuffer), &bytesRead); + if (bytesJustRead != sizeof(subchunkSizeBuffer)) { + break; + } + subchunkDataSize = ma_dr_wav_bytes_to_u32(subchunkSizeBuffer); + if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_label, "labl") || ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_note, "note")) { + if (subchunkDataSize >= MA_DR_WAV_LIST_LABEL_OR_NOTE_BYTES) { + ma_uint64 stringSizeWithNullTerm = subchunkDataSize - MA_DR_WAV_LIST_LABEL_OR_NOTE_BYTES; + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + pParser->metadataCount += 1; + ma_dr_wav__metadata_request_extra_memory_for_stage_2(pParser, (size_t)stringSizeWithNullTerm, 1); + } else { + subchunkBytesRead = ma_dr_wav__read_list_label_or_note_to_metadata_obj(pParser, &pParser->pMetadata[pParser->metadataCursor], subchunkDataSize, ma_dr_wav_fourcc_equal(subchunkId, "labl") ? ma_dr_wav_metadata_type_list_label : ma_dr_wav_metadata_type_list_note); + if (subchunkBytesRead == subchunkDataSize) { + pParser->metadataCursor += 1; + } else { + } + } + } else { + } + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_labelled_cue_region, "ltxt")) { + if (subchunkDataSize >= MA_DR_WAV_LIST_LABELLED_TEXT_BYTES) { + ma_uint64 stringSizeWithNullTerminator = subchunkDataSize - MA_DR_WAV_LIST_LABELLED_TEXT_BYTES; + if (pParser->stage == ma_dr_wav__metadata_parser_stage_count) { + pParser->metadataCount += 1; + ma_dr_wav__metadata_request_extra_memory_for_stage_2(pParser, (size_t)stringSizeWithNullTerminator, 1); + } else { + subchunkBytesRead = ma_dr_wav__read_list_labelled_cue_region_to_metadata_obj(pParser, &pParser->pMetadata[pParser->metadataCursor], subchunkDataSize); + if (subchunkBytesRead == subchunkDataSize) { + pParser->metadataCursor += 1; + } else { + } + } + } else { + } + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_software, "ISFT")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_software); + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_copyright, "ICOP")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_copyright); + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_title, "INAM")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_title); + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_artist, "IART")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_artist); + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_comment, "ICMT")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_comment); + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_date, "ICRD")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_date); + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_genre, "IGNR")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_genre); + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_album, "IPRD")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_album); + } else if (ma_dr_wav__chunk_matches(allowedMetadataTypes, subchunkId, ma_dr_wav_metadata_type_list_info_tracknumber, "ITRK")) { + subchunkBytesRead = ma_dr_wav__metadata_process_info_text_chunk(pParser, subchunkDataSize, ma_dr_wav_metadata_type_list_info_tracknumber); + } else if ((allowedMetadataTypes & ma_dr_wav_metadata_type_unknown) != 0) { + subchunkBytesRead = ma_dr_wav__metadata_process_unknown_chunk(pParser, subchunkId, subchunkDataSize, listType); + } + bytesRead += subchunkBytesRead; + MA_DR_WAV_ASSERT(subchunkBytesRead <= subchunkDataSize); + if (subchunkBytesRead < subchunkDataSize) { + ma_uint64 bytesToSeek = subchunkDataSize - subchunkBytesRead; + if (!pParser->onSeek(pParser->pReadSeekUserData, (int)bytesToSeek, ma_dr_wav_seek_origin_current)) { + break; + } + bytesRead += bytesToSeek; + } + if ((subchunkDataSize % 2) == 1) { + if (!pParser->onSeek(pParser->pReadSeekUserData, 1, ma_dr_wav_seek_origin_current)) { + break; + } + bytesRead += 1; + } + } + } else if ((allowedMetadataTypes & ma_dr_wav_metadata_type_unknown) != 0) { + bytesRead = ma_dr_wav__metadata_process_unknown_chunk(pParser, pChunkID, pChunkHeader->sizeInBytes, ma_dr_wav_metadata_location_top_level); + } + return bytesRead; +} +MA_PRIVATE ma_uint32 ma_dr_wav_get_bytes_per_pcm_frame(ma_dr_wav* pWav) +{ + ma_uint32 bytesPerFrame; + if ((pWav->bitsPerSample & 0x7) == 0) { + bytesPerFrame = (pWav->bitsPerSample * pWav->fmt.channels) >> 3; + } else { + bytesPerFrame = pWav->fmt.blockAlign; + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ALAW || pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_MULAW) { + if (bytesPerFrame != pWav->fmt.channels) { + return 0; + } + } + return bytesPerFrame; +} +MA_API ma_uint16 ma_dr_wav_fmt_get_format(const ma_dr_wav_fmt* pFMT) +{ + if (pFMT == NULL) { + return 0; + } + if (pFMT->formatTag != MA_DR_WAVE_FORMAT_EXTENSIBLE) { + return pFMT->formatTag; + } else { + return ma_dr_wav_bytes_to_u16(pFMT->subFormat); + } +} +MA_PRIVATE ma_bool32 ma_dr_wav_preinit(ma_dr_wav* pWav, ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pReadSeekUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pWav == NULL || onRead == NULL || onSeek == NULL) { + return MA_FALSE; + } + MA_DR_WAV_ZERO_MEMORY(pWav, sizeof(*pWav)); + pWav->onRead = onRead; + pWav->onSeek = onSeek; + pWav->pUserData = pReadSeekUserData; + pWav->allocationCallbacks = ma_dr_wav_copy_allocation_callbacks_or_defaults(pAllocationCallbacks); + if (pWav->allocationCallbacks.onFree == NULL || (pWav->allocationCallbacks.onMalloc == NULL && pWav->allocationCallbacks.onRealloc == NULL)) { + return MA_FALSE; + } + return MA_TRUE; +} +MA_PRIVATE ma_bool32 ma_dr_wav_init__internal(ma_dr_wav* pWav, ma_dr_wav_chunk_proc onChunk, void* pChunkUserData, ma_uint32 flags) +{ + ma_result result; + ma_uint64 cursor; + ma_bool32 sequential; + ma_uint8 riff[4]; + ma_dr_wav_fmt fmt; + unsigned short translatedFormatTag; + ma_uint64 dataChunkSize = 0; + ma_uint64 sampleCountFromFactChunk = 0; + ma_uint64 metadataStartPos; + ma_dr_wav__metadata_parser metadataParser; + ma_bool8 isProcessingMetadata = MA_FALSE; + ma_bool8 foundChunk_fmt = MA_FALSE; + ma_bool8 foundChunk_data = MA_FALSE; + ma_bool8 isAIFCFormType = MA_FALSE; + ma_uint64 aiffFrameCount = 0; + cursor = 0; + sequential = (flags & MA_DR_WAV_SEQUENTIAL) != 0; + MA_DR_WAV_ZERO_OBJECT(&fmt); + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, riff, sizeof(riff), &cursor) != sizeof(riff)) { + return MA_FALSE; + } + if (ma_dr_wav_fourcc_equal(riff, "RIFF")) { + pWav->container = ma_dr_wav_container_riff; + } else if (ma_dr_wav_fourcc_equal(riff, "RIFX")) { + pWav->container = ma_dr_wav_container_rifx; + } else if (ma_dr_wav_fourcc_equal(riff, "riff")) { + int i; + ma_uint8 riff2[12]; + pWav->container = ma_dr_wav_container_w64; + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, riff2, sizeof(riff2), &cursor) != sizeof(riff2)) { + return MA_FALSE; + } + for (i = 0; i < 12; ++i) { + if (riff2[i] != ma_dr_wavGUID_W64_RIFF[i+4]) { + return MA_FALSE; + } + } + } else if (ma_dr_wav_fourcc_equal(riff, "RF64")) { + pWav->container = ma_dr_wav_container_rf64; + } else if (ma_dr_wav_fourcc_equal(riff, "FORM")) { + pWav->container = ma_dr_wav_container_aiff; + } else { + return MA_FALSE; + } + if (pWav->container == ma_dr_wav_container_riff || pWav->container == ma_dr_wav_container_rifx || pWav->container == ma_dr_wav_container_rf64) { + ma_uint8 chunkSizeBytes[4]; + ma_uint8 wave[4]; + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, chunkSizeBytes, sizeof(chunkSizeBytes), &cursor) != sizeof(chunkSizeBytes)) { + return MA_FALSE; + } + if (pWav->container == ma_dr_wav_container_riff || pWav->container == ma_dr_wav_container_rifx) { + if (ma_dr_wav_bytes_to_u32_ex(chunkSizeBytes, pWav->container) < 36) { + } + } else if (pWav->container == ma_dr_wav_container_rf64) { + if (ma_dr_wav_bytes_to_u32_le(chunkSizeBytes) != 0xFFFFFFFF) { + return MA_FALSE; + } + } else { + return MA_FALSE; + } + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, wave, sizeof(wave), &cursor) != sizeof(wave)) { + return MA_FALSE; + } + if (!ma_dr_wav_fourcc_equal(wave, "WAVE")) { + return MA_FALSE; + } + } else if (pWav->container == ma_dr_wav_container_w64) { + ma_uint8 chunkSizeBytes[8]; + ma_uint8 wave[16]; + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, chunkSizeBytes, sizeof(chunkSizeBytes), &cursor) != sizeof(chunkSizeBytes)) { + return MA_FALSE; + } + if (ma_dr_wav_bytes_to_u64(chunkSizeBytes) < 80) { + return MA_FALSE; + } + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, wave, sizeof(wave), &cursor) != sizeof(wave)) { + return MA_FALSE; + } + if (!ma_dr_wav_guid_equal(wave, ma_dr_wavGUID_W64_WAVE)) { + return MA_FALSE; + } + } else if (pWav->container == ma_dr_wav_container_aiff) { + ma_uint8 chunkSizeBytes[4]; + ma_uint8 aiff[4]; + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, chunkSizeBytes, sizeof(chunkSizeBytes), &cursor) != sizeof(chunkSizeBytes)) { + return MA_FALSE; + } + if (ma_dr_wav_bytes_to_u32_be(chunkSizeBytes) < 18) { + return MA_FALSE; + } + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, aiff, sizeof(aiff), &cursor) != sizeof(aiff)) { + return MA_FALSE; + } + if (ma_dr_wav_fourcc_equal(aiff, "AIFF")) { + isAIFCFormType = MA_FALSE; + } else if (ma_dr_wav_fourcc_equal(aiff, "AIFC")) { + isAIFCFormType = MA_TRUE; + } else { + return MA_FALSE; + } + } else { + return MA_FALSE; + } + if (pWav->container == ma_dr_wav_container_rf64) { + ma_uint8 sizeBytes[8]; + ma_uint64 bytesRemainingInChunk; + ma_dr_wav_chunk_header header; + result = ma_dr_wav__read_chunk_header(pWav->onRead, pWav->pUserData, pWav->container, &cursor, &header); + if (result != MA_SUCCESS) { + return MA_FALSE; + } + if (!ma_dr_wav_fourcc_equal(header.id.fourcc, "ds64")) { + return MA_FALSE; + } + bytesRemainingInChunk = header.sizeInBytes + header.paddingSize; + if (!ma_dr_wav__seek_forward(pWav->onSeek, 8, pWav->pUserData)) { + return MA_FALSE; + } + bytesRemainingInChunk -= 8; + cursor += 8; + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, sizeBytes, sizeof(sizeBytes), &cursor) != sizeof(sizeBytes)) { + return MA_FALSE; + } + bytesRemainingInChunk -= 8; + dataChunkSize = ma_dr_wav_bytes_to_u64(sizeBytes); + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, sizeBytes, sizeof(sizeBytes), &cursor) != sizeof(sizeBytes)) { + return MA_FALSE; + } + bytesRemainingInChunk -= 8; + sampleCountFromFactChunk = ma_dr_wav_bytes_to_u64(sizeBytes); + if (!ma_dr_wav__seek_forward(pWav->onSeek, bytesRemainingInChunk, pWav->pUserData)) { + return MA_FALSE; + } + cursor += bytesRemainingInChunk; + } + metadataStartPos = cursor; + isProcessingMetadata = !sequential && ((flags & MA_DR_WAV_WITH_METADATA) != 0); + if (pWav->container != ma_dr_wav_container_riff && pWav->container != ma_dr_wav_container_rf64) { + isProcessingMetadata = MA_FALSE; + } + MA_DR_WAV_ZERO_MEMORY(&metadataParser, sizeof(metadataParser)); + if (isProcessingMetadata) { + metadataParser.onRead = pWav->onRead; + metadataParser.onSeek = pWav->onSeek; + metadataParser.pReadSeekUserData = pWav->pUserData; + metadataParser.stage = ma_dr_wav__metadata_parser_stage_count; + } + for (;;) { + ma_dr_wav_chunk_header header; + ma_uint64 chunkSize; + result = ma_dr_wav__read_chunk_header(pWav->onRead, pWav->pUserData, pWav->container, &cursor, &header); + if (result != MA_SUCCESS) { + break; + } + chunkSize = header.sizeInBytes; + if (!sequential && onChunk != NULL) { + ma_uint64 callbackBytesRead = onChunk(pChunkUserData, pWav->onRead, pWav->onSeek, pWav->pUserData, &header, pWav->container, &fmt); + if (callbackBytesRead > 0) { + if (ma_dr_wav__seek_from_start(pWav->onSeek, cursor, pWav->pUserData) == MA_FALSE) { + return MA_FALSE; + } + } + } + if (((pWav->container == ma_dr_wav_container_riff || pWav->container == ma_dr_wav_container_rifx || pWav->container == ma_dr_wav_container_rf64) && ma_dr_wav_fourcc_equal(header.id.fourcc, "fmt ")) || + ((pWav->container == ma_dr_wav_container_w64) && ma_dr_wav_guid_equal(header.id.guid, ma_dr_wavGUID_W64_FMT))) { + ma_uint8 fmtData[16]; + foundChunk_fmt = MA_TRUE; + if (pWav->onRead(pWav->pUserData, fmtData, sizeof(fmtData)) != sizeof(fmtData)) { + return MA_FALSE; + } + cursor += sizeof(fmtData); + fmt.formatTag = ma_dr_wav_bytes_to_u16_ex(fmtData + 0, pWav->container); + fmt.channels = ma_dr_wav_bytes_to_u16_ex(fmtData + 2, pWav->container); + fmt.sampleRate = ma_dr_wav_bytes_to_u32_ex(fmtData + 4, pWav->container); + fmt.avgBytesPerSec = ma_dr_wav_bytes_to_u32_ex(fmtData + 8, pWav->container); + fmt.blockAlign = ma_dr_wav_bytes_to_u16_ex(fmtData + 12, pWav->container); + fmt.bitsPerSample = ma_dr_wav_bytes_to_u16_ex(fmtData + 14, pWav->container); + fmt.extendedSize = 0; + fmt.validBitsPerSample = 0; + fmt.channelMask = 0; + MA_DR_WAV_ZERO_MEMORY(fmt.subFormat, sizeof(fmt.subFormat)); + if (header.sizeInBytes > 16) { + ma_uint8 fmt_cbSize[2]; + int bytesReadSoFar = 0; + if (pWav->onRead(pWav->pUserData, fmt_cbSize, sizeof(fmt_cbSize)) != sizeof(fmt_cbSize)) { + return MA_FALSE; + } + cursor += sizeof(fmt_cbSize); + bytesReadSoFar = 18; + fmt.extendedSize = ma_dr_wav_bytes_to_u16_ex(fmt_cbSize, pWav->container); + if (fmt.extendedSize > 0) { + if (fmt.formatTag == MA_DR_WAVE_FORMAT_EXTENSIBLE) { + if (fmt.extendedSize != 22) { + return MA_FALSE; + } + } + if (fmt.formatTag == MA_DR_WAVE_FORMAT_EXTENSIBLE) { + ma_uint8 fmtext[22]; + if (pWav->onRead(pWav->pUserData, fmtext, fmt.extendedSize) != fmt.extendedSize) { + return MA_FALSE; + } + fmt.validBitsPerSample = ma_dr_wav_bytes_to_u16_ex(fmtext + 0, pWav->container); + fmt.channelMask = ma_dr_wav_bytes_to_u32_ex(fmtext + 2, pWav->container); + ma_dr_wav_bytes_to_guid(fmtext + 6, fmt.subFormat); + } else { + if (pWav->onSeek(pWav->pUserData, fmt.extendedSize, ma_dr_wav_seek_origin_current) == MA_FALSE) { + return MA_FALSE; + } + } + cursor += fmt.extendedSize; + bytesReadSoFar += fmt.extendedSize; + } + if (pWav->onSeek(pWav->pUserData, (int)(header.sizeInBytes - bytesReadSoFar), ma_dr_wav_seek_origin_current) == MA_FALSE) { + return MA_FALSE; + } + cursor += (header.sizeInBytes - bytesReadSoFar); + } + if (header.paddingSize > 0) { + if (ma_dr_wav__seek_forward(pWav->onSeek, header.paddingSize, pWav->pUserData) == MA_FALSE) { + break; + } + cursor += header.paddingSize; + } + continue; + } + if (((pWav->container == ma_dr_wav_container_riff || pWav->container == ma_dr_wav_container_rifx || pWav->container == ma_dr_wav_container_rf64) && ma_dr_wav_fourcc_equal(header.id.fourcc, "data")) || + ((pWav->container == ma_dr_wav_container_w64) && ma_dr_wav_guid_equal(header.id.guid, ma_dr_wavGUID_W64_DATA))) { + foundChunk_data = MA_TRUE; + pWav->dataChunkDataPos = cursor; + if (pWav->container != ma_dr_wav_container_rf64) { + dataChunkSize = chunkSize; + } + if (sequential || !isProcessingMetadata) { + break; + } else { + chunkSize += header.paddingSize; + if (ma_dr_wav__seek_forward(pWav->onSeek, chunkSize, pWav->pUserData) == MA_FALSE) { + break; + } + cursor += chunkSize; + continue; + } + } + if (((pWav->container == ma_dr_wav_container_riff || pWav->container == ma_dr_wav_container_rifx || pWav->container == ma_dr_wav_container_rf64) && ma_dr_wav_fourcc_equal(header.id.fourcc, "fact")) || + ((pWav->container == ma_dr_wav_container_w64) && ma_dr_wav_guid_equal(header.id.guid, ma_dr_wavGUID_W64_FACT))) { + if (pWav->container == ma_dr_wav_container_riff || pWav->container == ma_dr_wav_container_rifx) { + ma_uint8 sampleCount[4]; + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, &sampleCount, 4, &cursor) != 4) { + return MA_FALSE; + } + chunkSize -= 4; + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM) { + sampleCountFromFactChunk = ma_dr_wav_bytes_to_u32_ex(sampleCount, pWav->container); + } else { + sampleCountFromFactChunk = 0; + } + } else if (pWav->container == ma_dr_wav_container_w64) { + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, &sampleCountFromFactChunk, 8, &cursor) != 8) { + return MA_FALSE; + } + chunkSize -= 8; + } else if (pWav->container == ma_dr_wav_container_rf64) { + } + chunkSize += header.paddingSize; + if (ma_dr_wav__seek_forward(pWav->onSeek, chunkSize, pWav->pUserData) == MA_FALSE) { + break; + } + cursor += chunkSize; + continue; + } + if (pWav->container == ma_dr_wav_container_aiff && ma_dr_wav_fourcc_equal(header.id.fourcc, "COMM")) { + ma_uint8 commData[24]; + ma_uint32 commDataBytesToRead; + ma_uint16 channels; + ma_uint32 frameCount; + ma_uint16 sampleSizeInBits; + ma_int64 sampleRate; + ma_uint16 compressionFormat; + foundChunk_fmt = MA_TRUE; + if (isAIFCFormType) { + commDataBytesToRead = 24; + if (header.sizeInBytes < commDataBytesToRead) { + return MA_FALSE; + } + } else { + commDataBytesToRead = 18; + if (header.sizeInBytes != commDataBytesToRead) { + return MA_FALSE; + } + } + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, commData, commDataBytesToRead, &cursor) != commDataBytesToRead) { + return MA_FALSE; + } + channels = ma_dr_wav_bytes_to_u16_ex (commData + 0, pWav->container); + frameCount = ma_dr_wav_bytes_to_u32_ex (commData + 2, pWav->container); + sampleSizeInBits = ma_dr_wav_bytes_to_u16_ex (commData + 6, pWav->container); + sampleRate = ma_dr_wav_aiff_extented_to_s64(commData + 8); + if (sampleRate < 0 || sampleRate > 0xFFFFFFFF) { + return MA_FALSE; + } + if (isAIFCFormType) { + const ma_uint8* type = commData + 18; + if (ma_dr_wav_fourcc_equal(type, "NONE")) { + compressionFormat = MA_DR_WAVE_FORMAT_PCM; + } else if (ma_dr_wav_fourcc_equal(type, "raw ")) { + compressionFormat = MA_DR_WAVE_FORMAT_PCM; + if (sampleSizeInBits == 8) { + pWav->aiff.isUnsigned = MA_TRUE; + } + } else if (ma_dr_wav_fourcc_equal(type, "sowt")) { + compressionFormat = MA_DR_WAVE_FORMAT_PCM; + pWav->aiff.isLE = MA_TRUE; + } else if (ma_dr_wav_fourcc_equal(type, "fl32") || ma_dr_wav_fourcc_equal(type, "fl64") || ma_dr_wav_fourcc_equal(type, "FL32") || ma_dr_wav_fourcc_equal(type, "FL64")) { + compressionFormat = MA_DR_WAVE_FORMAT_IEEE_FLOAT; + } else if (ma_dr_wav_fourcc_equal(type, "alaw") || ma_dr_wav_fourcc_equal(type, "ALAW")) { + compressionFormat = MA_DR_WAVE_FORMAT_ALAW; + } else if (ma_dr_wav_fourcc_equal(type, "ulaw") || ma_dr_wav_fourcc_equal(type, "ULAW")) { + compressionFormat = MA_DR_WAVE_FORMAT_MULAW; + } else if (ma_dr_wav_fourcc_equal(type, "ima4")) { + compressionFormat = MA_DR_WAVE_FORMAT_DVI_ADPCM; + sampleSizeInBits = 4; + (void)compressionFormat; + (void)sampleSizeInBits; + return MA_FALSE; + } else { + return MA_FALSE; + } + } else { + compressionFormat = MA_DR_WAVE_FORMAT_PCM; + } + aiffFrameCount = frameCount; + fmt.formatTag = compressionFormat; + fmt.channels = channels; + fmt.sampleRate = (ma_uint32)sampleRate; + fmt.bitsPerSample = sampleSizeInBits; + fmt.blockAlign = (ma_uint16)(fmt.channels * fmt.bitsPerSample / 8); + fmt.avgBytesPerSec = fmt.blockAlign * fmt.sampleRate; + if (fmt.blockAlign == 0 && compressionFormat == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + fmt.blockAlign = 34 * fmt.channels; + } + if (compressionFormat == MA_DR_WAVE_FORMAT_ALAW || compressionFormat == MA_DR_WAVE_FORMAT_MULAW) { + if (fmt.bitsPerSample > 8) { + fmt.bitsPerSample = 8; + fmt.blockAlign = fmt.channels; + } + } + fmt.bitsPerSample += (fmt.bitsPerSample & 7); + if (isAIFCFormType) { + if (ma_dr_wav__seek_forward(pWav->onSeek, (chunkSize - commDataBytesToRead), pWav->pUserData) == MA_FALSE) { + return MA_FALSE; + } + cursor += (chunkSize - commDataBytesToRead); + } + continue; + } + if (pWav->container == ma_dr_wav_container_aiff && ma_dr_wav_fourcc_equal(header.id.fourcc, "SSND")) { + ma_uint8 offsetAndBlockSizeData[8]; + ma_uint32 offset; + foundChunk_data = MA_TRUE; + if (ma_dr_wav__on_read(pWav->onRead, pWav->pUserData, offsetAndBlockSizeData, sizeof(offsetAndBlockSizeData), &cursor) != sizeof(offsetAndBlockSizeData)) { + return MA_FALSE; + } + offset = ma_dr_wav_bytes_to_u32_ex(offsetAndBlockSizeData + 0, pWav->container); + if (ma_dr_wav__seek_forward(pWav->onSeek, offset, pWav->pUserData) == MA_FALSE) { + return MA_FALSE; + } + cursor += offset; + pWav->dataChunkDataPos = cursor; + dataChunkSize = chunkSize; + if (sequential || !isProcessingMetadata) { + break; + } else { + if (ma_dr_wav__seek_forward(pWav->onSeek, chunkSize, pWav->pUserData) == MA_FALSE) { + break; + } + cursor += chunkSize; + continue; + } + } + if (isProcessingMetadata) { + ma_dr_wav__metadata_process_chunk(&metadataParser, &header, ma_dr_wav_metadata_type_all_including_unknown); + if (ma_dr_wav__seek_from_start(pWav->onSeek, cursor, pWav->pUserData) == MA_FALSE) { + break; + } + } + chunkSize += header.paddingSize; + if (ma_dr_wav__seek_forward(pWav->onSeek, chunkSize, pWav->pUserData) == MA_FALSE) { + break; + } + cursor += chunkSize; + } + if (!foundChunk_fmt || !foundChunk_data) { + return MA_FALSE; + } + if ((fmt.sampleRate == 0 || fmt.sampleRate > MA_DR_WAV_MAX_SAMPLE_RATE ) || + (fmt.channels == 0 || fmt.channels > MA_DR_WAV_MAX_CHANNELS ) || + (fmt.bitsPerSample == 0 || fmt.bitsPerSample > MA_DR_WAV_MAX_BITS_PER_SAMPLE) || + fmt.blockAlign == 0) { + return MA_FALSE; + } + translatedFormatTag = fmt.formatTag; + if (translatedFormatTag == MA_DR_WAVE_FORMAT_EXTENSIBLE) { + translatedFormatTag = ma_dr_wav_bytes_to_u16_ex(fmt.subFormat + 0, pWav->container); + } + if (!sequential) { + if (!ma_dr_wav__seek_from_start(pWav->onSeek, pWav->dataChunkDataPos, pWav->pUserData)) { + return MA_FALSE; + } + cursor = pWav->dataChunkDataPos; + } + if (isProcessingMetadata && metadataParser.metadataCount > 0) { + if (ma_dr_wav__seek_from_start(pWav->onSeek, metadataStartPos, pWav->pUserData) == MA_FALSE) { + return MA_FALSE; + } + result = ma_dr_wav__metadata_alloc(&metadataParser, &pWav->allocationCallbacks); + if (result != MA_SUCCESS) { + return MA_FALSE; + } + metadataParser.stage = ma_dr_wav__metadata_parser_stage_read; + for (;;) { + ma_dr_wav_chunk_header header; + ma_uint64 metadataBytesRead; + result = ma_dr_wav__read_chunk_header(pWav->onRead, pWav->pUserData, pWav->container, &cursor, &header); + if (result != MA_SUCCESS) { + break; + } + metadataBytesRead = ma_dr_wav__metadata_process_chunk(&metadataParser, &header, ma_dr_wav_metadata_type_all_including_unknown); + if (ma_dr_wav__seek_forward(pWav->onSeek, (header.sizeInBytes + header.paddingSize) - metadataBytesRead, pWav->pUserData) == MA_FALSE) { + ma_dr_wav_free(metadataParser.pMetadata, &pWav->allocationCallbacks); + return MA_FALSE; + } + } + pWav->pMetadata = metadataParser.pMetadata; + pWav->metadataCount = metadataParser.metadataCount; + } + if (dataChunkSize == 0xFFFFFFFF && (pWav->container == ma_dr_wav_container_riff || pWav->container == ma_dr_wav_container_rifx) && pWav->isSequentialWrite == MA_FALSE) { + dataChunkSize = 0; + for (;;) { + ma_uint8 temp[4096]; + size_t bytesRead = pWav->onRead(pWav->pUserData, temp, sizeof(temp)); + dataChunkSize += bytesRead; + if (bytesRead < sizeof(temp)) { + break; + } + } + } + if (ma_dr_wav__seek_from_start(pWav->onSeek, pWav->dataChunkDataPos, pWav->pUserData) == MA_FALSE) { + ma_dr_wav_free(pWav->pMetadata, &pWav->allocationCallbacks); + return MA_FALSE; + } + pWav->fmt = fmt; + pWav->sampleRate = fmt.sampleRate; + pWav->channels = fmt.channels; + pWav->bitsPerSample = fmt.bitsPerSample; + pWav->bytesRemaining = dataChunkSize; + pWav->translatedFormatTag = translatedFormatTag; + pWav->dataChunkDataSize = dataChunkSize; + if (sampleCountFromFactChunk != 0) { + pWav->totalPCMFrameCount = sampleCountFromFactChunk; + } else if (aiffFrameCount != 0) { + pWav->totalPCMFrameCount = aiffFrameCount; + } else { + ma_uint32 bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + ma_dr_wav_free(pWav->pMetadata, &pWav->allocationCallbacks); + return MA_FALSE; + } + pWav->totalPCMFrameCount = dataChunkSize / bytesPerFrame; + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM) { + ma_uint64 totalBlockHeaderSizeInBytes; + ma_uint64 blockCount = dataChunkSize / fmt.blockAlign; + if ((blockCount * fmt.blockAlign) < dataChunkSize) { + blockCount += 1; + } + totalBlockHeaderSizeInBytes = blockCount * (6*fmt.channels); + pWav->totalPCMFrameCount = ((dataChunkSize - totalBlockHeaderSizeInBytes) * 2) / fmt.channels; + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + ma_uint64 totalBlockHeaderSizeInBytes; + ma_uint64 blockCount = dataChunkSize / fmt.blockAlign; + if ((blockCount * fmt.blockAlign) < dataChunkSize) { + blockCount += 1; + } + totalBlockHeaderSizeInBytes = blockCount * (4*fmt.channels); + pWav->totalPCMFrameCount = ((dataChunkSize - totalBlockHeaderSizeInBytes) * 2) / fmt.channels; + pWav->totalPCMFrameCount += blockCount; + } + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM || pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + if (pWav->channels > 2) { + ma_dr_wav_free(pWav->pMetadata, &pWav->allocationCallbacks); + return MA_FALSE; + } + } + if (ma_dr_wav_get_bytes_per_pcm_frame(pWav) == 0) { + ma_dr_wav_free(pWav->pMetadata, &pWav->allocationCallbacks); + return MA_FALSE; + } +#ifdef MA_DR_WAV_LIBSNDFILE_COMPAT + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM) { + ma_uint64 blockCount = dataChunkSize / fmt.blockAlign; + pWav->totalPCMFrameCount = (((blockCount * (fmt.blockAlign - (6*pWav->channels))) * 2)) / fmt.channels; + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + ma_uint64 blockCount = dataChunkSize / fmt.blockAlign; + pWav->totalPCMFrameCount = (((blockCount * (fmt.blockAlign - (4*pWav->channels))) * 2) + (blockCount * pWav->channels)) / fmt.channels; + } +#endif + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_wav_init(ma_dr_wav* pWav, ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_ex(pWav, onRead, onSeek, NULL, pUserData, NULL, 0, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_ex(ma_dr_wav* pWav, ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, ma_dr_wav_chunk_proc onChunk, void* pReadSeekUserData, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (!ma_dr_wav_preinit(pWav, onRead, onSeek, pReadSeekUserData, pAllocationCallbacks)) { + return MA_FALSE; + } + return ma_dr_wav_init__internal(pWav, onChunk, pChunkUserData, flags); +} +MA_API ma_bool32 ma_dr_wav_init_with_metadata(ma_dr_wav* pWav, ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (!ma_dr_wav_preinit(pWav, onRead, onSeek, pUserData, pAllocationCallbacks)) { + return MA_FALSE; + } + return ma_dr_wav_init__internal(pWav, NULL, NULL, flags | MA_DR_WAV_WITH_METADATA); +} +MA_API ma_dr_wav_metadata* ma_dr_wav_take_ownership_of_metadata(ma_dr_wav* pWav) +{ + ma_dr_wav_metadata *result = pWav->pMetadata; + pWav->pMetadata = NULL; + pWav->metadataCount = 0; + return result; +} +MA_PRIVATE size_t ma_dr_wav__write(ma_dr_wav* pWav, const void* pData, size_t dataSize) +{ + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(pWav->onWrite != NULL); + return pWav->onWrite(pWav->pUserData, pData, dataSize); +} +MA_PRIVATE size_t ma_dr_wav__write_byte(ma_dr_wav* pWav, ma_uint8 byte) +{ + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(pWav->onWrite != NULL); + return pWav->onWrite(pWav->pUserData, &byte, 1); +} +MA_PRIVATE size_t ma_dr_wav__write_u16ne_to_le(ma_dr_wav* pWav, ma_uint16 value) +{ + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(pWav->onWrite != NULL); + if (!ma_dr_wav__is_little_endian()) { + value = ma_dr_wav__bswap16(value); + } + return ma_dr_wav__write(pWav, &value, 2); +} +MA_PRIVATE size_t ma_dr_wav__write_u32ne_to_le(ma_dr_wav* pWav, ma_uint32 value) +{ + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(pWav->onWrite != NULL); + if (!ma_dr_wav__is_little_endian()) { + value = ma_dr_wav__bswap32(value); + } + return ma_dr_wav__write(pWav, &value, 4); +} +MA_PRIVATE size_t ma_dr_wav__write_u64ne_to_le(ma_dr_wav* pWav, ma_uint64 value) +{ + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(pWav->onWrite != NULL); + if (!ma_dr_wav__is_little_endian()) { + value = ma_dr_wav__bswap64(value); + } + return ma_dr_wav__write(pWav, &value, 8); +} +MA_PRIVATE size_t ma_dr_wav__write_f32ne_to_le(ma_dr_wav* pWav, float value) +{ + union { + ma_uint32 u32; + float f32; + } u; + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(pWav->onWrite != NULL); + u.f32 = value; + if (!ma_dr_wav__is_little_endian()) { + u.u32 = ma_dr_wav__bswap32(u.u32); + } + return ma_dr_wav__write(pWav, &u.u32, 4); +} +MA_PRIVATE size_t ma_dr_wav__write_or_count(ma_dr_wav* pWav, const void* pData, size_t dataSize) +{ + if (pWav == NULL) { + return dataSize; + } + return ma_dr_wav__write(pWav, pData, dataSize); +} +MA_PRIVATE size_t ma_dr_wav__write_or_count_byte(ma_dr_wav* pWav, ma_uint8 byte) +{ + if (pWav == NULL) { + return 1; + } + return ma_dr_wav__write_byte(pWav, byte); +} +MA_PRIVATE size_t ma_dr_wav__write_or_count_u16ne_to_le(ma_dr_wav* pWav, ma_uint16 value) +{ + if (pWav == NULL) { + return 2; + } + return ma_dr_wav__write_u16ne_to_le(pWav, value); +} +MA_PRIVATE size_t ma_dr_wav__write_or_count_u32ne_to_le(ma_dr_wav* pWav, ma_uint32 value) +{ + if (pWav == NULL) { + return 4; + } + return ma_dr_wav__write_u32ne_to_le(pWav, value); +} +#if 0 +MA_PRIVATE size_t ma_dr_wav__write_or_count_u64ne_to_le(ma_dr_wav* pWav, ma_uint64 value) +{ + if (pWav == NULL) { + return 8; + } + return ma_dr_wav__write_u64ne_to_le(pWav, value); +} +#endif +MA_PRIVATE size_t ma_dr_wav__write_or_count_f32ne_to_le(ma_dr_wav* pWav, float value) +{ + if (pWav == NULL) { + return 4; + } + return ma_dr_wav__write_f32ne_to_le(pWav, value); +} +MA_PRIVATE size_t ma_dr_wav__write_or_count_string_to_fixed_size_buf(ma_dr_wav* pWav, char* str, size_t bufFixedSize) +{ + size_t len; + if (pWav == NULL) { + return bufFixedSize; + } + len = ma_dr_wav__strlen_clamped(str, bufFixedSize); + ma_dr_wav__write_or_count(pWav, str, len); + if (len < bufFixedSize) { + size_t i; + for (i = 0; i < bufFixedSize - len; ++i) { + ma_dr_wav__write_byte(pWav, 0); + } + } + return bufFixedSize; +} +MA_PRIVATE size_t ma_dr_wav__write_or_count_metadata(ma_dr_wav* pWav, ma_dr_wav_metadata* pMetadatas, ma_uint32 metadataCount) +{ + size_t bytesWritten = 0; + ma_bool32 hasListAdtl = MA_FALSE; + ma_bool32 hasListInfo = MA_FALSE; + ma_uint32 iMetadata; + if (pMetadatas == NULL || metadataCount == 0) { + return 0; + } + for (iMetadata = 0; iMetadata < metadataCount; ++iMetadata) { + ma_dr_wav_metadata* pMetadata = &pMetadatas[iMetadata]; + ma_uint32 chunkSize = 0; + if ((pMetadata->type & ma_dr_wav_metadata_type_list_all_info_strings) || (pMetadata->type == ma_dr_wav_metadata_type_unknown && pMetadata->data.unknown.chunkLocation == ma_dr_wav_metadata_location_inside_info_list)) { + hasListInfo = MA_TRUE; + } + if ((pMetadata->type & ma_dr_wav_metadata_type_list_all_adtl) || (pMetadata->type == ma_dr_wav_metadata_type_unknown && pMetadata->data.unknown.chunkLocation == ma_dr_wav_metadata_location_inside_adtl_list)) { + hasListAdtl = MA_TRUE; + } + switch (pMetadata->type) { + case ma_dr_wav_metadata_type_smpl: + { + ma_uint32 iLoop; + chunkSize = MA_DR_WAV_SMPL_BYTES + MA_DR_WAV_SMPL_LOOP_BYTES * pMetadata->data.smpl.sampleLoopCount + pMetadata->data.smpl.samplerSpecificDataSizeInBytes; + bytesWritten += ma_dr_wav__write_or_count(pWav, "smpl", 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, chunkSize); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.manufacturerId); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.productId); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.samplePeriodNanoseconds); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.midiUnityNote); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.midiPitchFraction); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.smpteFormat); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.smpteOffset); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.sampleLoopCount); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.samplerSpecificDataSizeInBytes); + for (iLoop = 0; iLoop < pMetadata->data.smpl.sampleLoopCount; ++iLoop) { + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.pLoops[iLoop].cuePointId); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.pLoops[iLoop].type); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.pLoops[iLoop].firstSampleByteOffset); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.pLoops[iLoop].lastSampleByteOffset); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.pLoops[iLoop].sampleFraction); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.smpl.pLoops[iLoop].playCount); + } + if (pMetadata->data.smpl.samplerSpecificDataSizeInBytes > 0) { + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.smpl.pSamplerSpecificData, pMetadata->data.smpl.samplerSpecificDataSizeInBytes); + } + } break; + case ma_dr_wav_metadata_type_inst: + { + chunkSize = MA_DR_WAV_INST_BYTES; + bytesWritten += ma_dr_wav__write_or_count(pWav, "inst", 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, chunkSize); + bytesWritten += ma_dr_wav__write_or_count(pWav, &pMetadata->data.inst.midiUnityNote, 1); + bytesWritten += ma_dr_wav__write_or_count(pWav, &pMetadata->data.inst.fineTuneCents, 1); + bytesWritten += ma_dr_wav__write_or_count(pWav, &pMetadata->data.inst.gainDecibels, 1); + bytesWritten += ma_dr_wav__write_or_count(pWav, &pMetadata->data.inst.lowNote, 1); + bytesWritten += ma_dr_wav__write_or_count(pWav, &pMetadata->data.inst.highNote, 1); + bytesWritten += ma_dr_wav__write_or_count(pWav, &pMetadata->data.inst.lowVelocity, 1); + bytesWritten += ma_dr_wav__write_or_count(pWav, &pMetadata->data.inst.highVelocity, 1); + } break; + case ma_dr_wav_metadata_type_cue: + { + ma_uint32 iCuePoint; + chunkSize = MA_DR_WAV_CUE_BYTES + MA_DR_WAV_CUE_POINT_BYTES * pMetadata->data.cue.cuePointCount; + bytesWritten += ma_dr_wav__write_or_count(pWav, "cue ", 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, chunkSize); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.cue.cuePointCount); + for (iCuePoint = 0; iCuePoint < pMetadata->data.cue.cuePointCount; ++iCuePoint) { + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.cue.pCuePoints[iCuePoint].id); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.cue.pCuePoints[iCuePoint].playOrderPosition); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.cue.pCuePoints[iCuePoint].dataChunkId, 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.cue.pCuePoints[iCuePoint].chunkStart); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.cue.pCuePoints[iCuePoint].blockStart); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.cue.pCuePoints[iCuePoint].sampleByteOffset); + } + } break; + case ma_dr_wav_metadata_type_acid: + { + chunkSize = MA_DR_WAV_ACID_BYTES; + bytesWritten += ma_dr_wav__write_or_count(pWav, "acid", 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, chunkSize); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.acid.flags); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.acid.midiUnityNote); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.acid.reserved1); + bytesWritten += ma_dr_wav__write_or_count_f32ne_to_le(pWav, pMetadata->data.acid.reserved2); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.acid.numBeats); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.acid.meterDenominator); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.acid.meterNumerator); + bytesWritten += ma_dr_wav__write_or_count_f32ne_to_le(pWav, pMetadata->data.acid.tempo); + } break; + case ma_dr_wav_metadata_type_bext: + { + char reservedBuf[MA_DR_WAV_BEXT_RESERVED_BYTES]; + ma_uint32 timeReferenceLow; + ma_uint32 timeReferenceHigh; + chunkSize = MA_DR_WAV_BEXT_BYTES + pMetadata->data.bext.codingHistorySize; + bytesWritten += ma_dr_wav__write_or_count(pWav, "bext", 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, chunkSize); + bytesWritten += ma_dr_wav__write_or_count_string_to_fixed_size_buf(pWav, pMetadata->data.bext.pDescription, MA_DR_WAV_BEXT_DESCRIPTION_BYTES); + bytesWritten += ma_dr_wav__write_or_count_string_to_fixed_size_buf(pWav, pMetadata->data.bext.pOriginatorName, MA_DR_WAV_BEXT_ORIGINATOR_NAME_BYTES); + bytesWritten += ma_dr_wav__write_or_count_string_to_fixed_size_buf(pWav, pMetadata->data.bext.pOriginatorReference, MA_DR_WAV_BEXT_ORIGINATOR_REF_BYTES); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.bext.pOriginationDate, sizeof(pMetadata->data.bext.pOriginationDate)); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.bext.pOriginationTime, sizeof(pMetadata->data.bext.pOriginationTime)); + timeReferenceLow = (ma_uint32)(pMetadata->data.bext.timeReference & 0xFFFFFFFF); + timeReferenceHigh = (ma_uint32)(pMetadata->data.bext.timeReference >> 32); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, timeReferenceLow); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, timeReferenceHigh); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.bext.version); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.bext.pUMID, MA_DR_WAV_BEXT_UMID_BYTES); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.bext.loudnessValue); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.bext.loudnessRange); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.bext.maxTruePeakLevel); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.bext.maxMomentaryLoudness); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.bext.maxShortTermLoudness); + MA_DR_WAV_ZERO_MEMORY(reservedBuf, sizeof(reservedBuf)); + bytesWritten += ma_dr_wav__write_or_count(pWav, reservedBuf, sizeof(reservedBuf)); + if (pMetadata->data.bext.codingHistorySize > 0) { + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.bext.pCodingHistory, pMetadata->data.bext.codingHistorySize); + } + } break; + case ma_dr_wav_metadata_type_unknown: + { + if (pMetadata->data.unknown.chunkLocation == ma_dr_wav_metadata_location_top_level) { + chunkSize = pMetadata->data.unknown.dataSizeInBytes; + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.unknown.id, 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, chunkSize); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.unknown.pData, pMetadata->data.unknown.dataSizeInBytes); + } + } break; + default: break; + } + if ((chunkSize % 2) != 0) { + bytesWritten += ma_dr_wav__write_or_count_byte(pWav, 0); + } + } + if (hasListInfo) { + ma_uint32 chunkSize = 4; + for (iMetadata = 0; iMetadata < metadataCount; ++iMetadata) { + ma_dr_wav_metadata* pMetadata = &pMetadatas[iMetadata]; + if ((pMetadata->type & ma_dr_wav_metadata_type_list_all_info_strings)) { + chunkSize += 8; + chunkSize += pMetadata->data.infoText.stringLength + 1; + } else if (pMetadata->type == ma_dr_wav_metadata_type_unknown && pMetadata->data.unknown.chunkLocation == ma_dr_wav_metadata_location_inside_info_list) { + chunkSize += 8; + chunkSize += pMetadata->data.unknown.dataSizeInBytes; + } + if ((chunkSize % 2) != 0) { + chunkSize += 1; + } + } + bytesWritten += ma_dr_wav__write_or_count(pWav, "LIST", 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, chunkSize); + bytesWritten += ma_dr_wav__write_or_count(pWav, "INFO", 4); + for (iMetadata = 0; iMetadata < metadataCount; ++iMetadata) { + ma_dr_wav_metadata* pMetadata = &pMetadatas[iMetadata]; + ma_uint32 subchunkSize = 0; + if (pMetadata->type & ma_dr_wav_metadata_type_list_all_info_strings) { + const char* pID = NULL; + switch (pMetadata->type) { + case ma_dr_wav_metadata_type_list_info_software: pID = "ISFT"; break; + case ma_dr_wav_metadata_type_list_info_copyright: pID = "ICOP"; break; + case ma_dr_wav_metadata_type_list_info_title: pID = "INAM"; break; + case ma_dr_wav_metadata_type_list_info_artist: pID = "IART"; break; + case ma_dr_wav_metadata_type_list_info_comment: pID = "ICMT"; break; + case ma_dr_wav_metadata_type_list_info_date: pID = "ICRD"; break; + case ma_dr_wav_metadata_type_list_info_genre: pID = "IGNR"; break; + case ma_dr_wav_metadata_type_list_info_album: pID = "IPRD"; break; + case ma_dr_wav_metadata_type_list_info_tracknumber: pID = "ITRK"; break; + default: break; + } + MA_DR_WAV_ASSERT(pID != NULL); + if (pMetadata->data.infoText.stringLength) { + subchunkSize = pMetadata->data.infoText.stringLength + 1; + bytesWritten += ma_dr_wav__write_or_count(pWav, pID, 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, subchunkSize); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.infoText.pString, pMetadata->data.infoText.stringLength); + bytesWritten += ma_dr_wav__write_or_count_byte(pWav, '\0'); + } + } else if (pMetadata->type == ma_dr_wav_metadata_type_unknown && pMetadata->data.unknown.chunkLocation == ma_dr_wav_metadata_location_inside_info_list) { + if (pMetadata->data.unknown.dataSizeInBytes) { + subchunkSize = pMetadata->data.unknown.dataSizeInBytes; + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.unknown.id, 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.unknown.dataSizeInBytes); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.unknown.pData, subchunkSize); + } + } + if ((subchunkSize % 2) != 0) { + bytesWritten += ma_dr_wav__write_or_count_byte(pWav, 0); + } + } + } + if (hasListAdtl) { + ma_uint32 chunkSize = 4; + for (iMetadata = 0; iMetadata < metadataCount; ++iMetadata) { + ma_dr_wav_metadata* pMetadata = &pMetadatas[iMetadata]; + switch (pMetadata->type) + { + case ma_dr_wav_metadata_type_list_label: + case ma_dr_wav_metadata_type_list_note: + { + chunkSize += 8; + chunkSize += MA_DR_WAV_LIST_LABEL_OR_NOTE_BYTES; + if (pMetadata->data.labelOrNote.stringLength > 0) { + chunkSize += pMetadata->data.labelOrNote.stringLength + 1; + } + } break; + case ma_dr_wav_metadata_type_list_labelled_cue_region: + { + chunkSize += 8; + chunkSize += MA_DR_WAV_LIST_LABELLED_TEXT_BYTES; + if (pMetadata->data.labelledCueRegion.stringLength > 0) { + chunkSize += pMetadata->data.labelledCueRegion.stringLength + 1; + } + } break; + case ma_dr_wav_metadata_type_unknown: + { + if (pMetadata->data.unknown.chunkLocation == ma_dr_wav_metadata_location_inside_adtl_list) { + chunkSize += 8; + chunkSize += pMetadata->data.unknown.dataSizeInBytes; + } + } break; + default: break; + } + if ((chunkSize % 2) != 0) { + chunkSize += 1; + } + } + bytesWritten += ma_dr_wav__write_or_count(pWav, "LIST", 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, chunkSize); + bytesWritten += ma_dr_wav__write_or_count(pWav, "adtl", 4); + for (iMetadata = 0; iMetadata < metadataCount; ++iMetadata) { + ma_dr_wav_metadata* pMetadata = &pMetadatas[iMetadata]; + ma_uint32 subchunkSize = 0; + switch (pMetadata->type) + { + case ma_dr_wav_metadata_type_list_label: + case ma_dr_wav_metadata_type_list_note: + { + if (pMetadata->data.labelOrNote.stringLength > 0) { + const char *pID = NULL; + if (pMetadata->type == ma_dr_wav_metadata_type_list_label) { + pID = "labl"; + } + else if (pMetadata->type == ma_dr_wav_metadata_type_list_note) { + pID = "note"; + } + MA_DR_WAV_ASSERT(pID != NULL); + MA_DR_WAV_ASSERT(pMetadata->data.labelOrNote.pString != NULL); + subchunkSize = MA_DR_WAV_LIST_LABEL_OR_NOTE_BYTES; + bytesWritten += ma_dr_wav__write_or_count(pWav, pID, 4); + subchunkSize += pMetadata->data.labelOrNote.stringLength + 1; + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, subchunkSize); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.labelOrNote.cuePointId); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.labelOrNote.pString, pMetadata->data.labelOrNote.stringLength); + bytesWritten += ma_dr_wav__write_or_count_byte(pWav, '\0'); + } + } break; + case ma_dr_wav_metadata_type_list_labelled_cue_region: + { + subchunkSize = MA_DR_WAV_LIST_LABELLED_TEXT_BYTES; + bytesWritten += ma_dr_wav__write_or_count(pWav, "ltxt", 4); + if (pMetadata->data.labelledCueRegion.stringLength > 0) { + subchunkSize += pMetadata->data.labelledCueRegion.stringLength + 1; + } + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, subchunkSize); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.labelledCueRegion.cuePointId); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, pMetadata->data.labelledCueRegion.sampleLength); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.labelledCueRegion.purposeId, 4); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.labelledCueRegion.country); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.labelledCueRegion.language); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.labelledCueRegion.dialect); + bytesWritten += ma_dr_wav__write_or_count_u16ne_to_le(pWav, pMetadata->data.labelledCueRegion.codePage); + if (pMetadata->data.labelledCueRegion.stringLength > 0) { + MA_DR_WAV_ASSERT(pMetadata->data.labelledCueRegion.pString != NULL); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.labelledCueRegion.pString, pMetadata->data.labelledCueRegion.stringLength); + bytesWritten += ma_dr_wav__write_or_count_byte(pWav, '\0'); + } + } break; + case ma_dr_wav_metadata_type_unknown: + { + if (pMetadata->data.unknown.chunkLocation == ma_dr_wav_metadata_location_inside_adtl_list) { + subchunkSize = pMetadata->data.unknown.dataSizeInBytes; + MA_DR_WAV_ASSERT(pMetadata->data.unknown.pData != NULL); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.unknown.id, 4); + bytesWritten += ma_dr_wav__write_or_count_u32ne_to_le(pWav, subchunkSize); + bytesWritten += ma_dr_wav__write_or_count(pWav, pMetadata->data.unknown.pData, subchunkSize); + } + } break; + default: break; + } + if ((subchunkSize % 2) != 0) { + bytesWritten += ma_dr_wav__write_or_count_byte(pWav, 0); + } + } + } + MA_DR_WAV_ASSERT((bytesWritten % 2) == 0); + return bytesWritten; +} +MA_PRIVATE ma_uint32 ma_dr_wav__riff_chunk_size_riff(ma_uint64 dataChunkSize, ma_dr_wav_metadata* pMetadata, ma_uint32 metadataCount) +{ + ma_uint64 chunkSize = 4 + 24 + (ma_uint64)ma_dr_wav__write_or_count_metadata(NULL, pMetadata, metadataCount) + 8 + dataChunkSize + ma_dr_wav__chunk_padding_size_riff(dataChunkSize); + if (chunkSize > 0xFFFFFFFFUL) { + chunkSize = 0xFFFFFFFFUL; + } + return (ma_uint32)chunkSize; +} +MA_PRIVATE ma_uint32 ma_dr_wav__data_chunk_size_riff(ma_uint64 dataChunkSize) +{ + if (dataChunkSize <= 0xFFFFFFFFUL) { + return (ma_uint32)dataChunkSize; + } else { + return 0xFFFFFFFFUL; + } +} +MA_PRIVATE ma_uint64 ma_dr_wav__riff_chunk_size_w64(ma_uint64 dataChunkSize) +{ + ma_uint64 dataSubchunkPaddingSize = ma_dr_wav__chunk_padding_size_w64(dataChunkSize); + return 80 + 24 + dataChunkSize + dataSubchunkPaddingSize; +} +MA_PRIVATE ma_uint64 ma_dr_wav__data_chunk_size_w64(ma_uint64 dataChunkSize) +{ + return 24 + dataChunkSize; +} +MA_PRIVATE ma_uint64 ma_dr_wav__riff_chunk_size_rf64(ma_uint64 dataChunkSize, ma_dr_wav_metadata *metadata, ma_uint32 numMetadata) +{ + ma_uint64 chunkSize = 4 + 36 + 24 + (ma_uint64)ma_dr_wav__write_or_count_metadata(NULL, metadata, numMetadata) + 8 + dataChunkSize + ma_dr_wav__chunk_padding_size_riff(dataChunkSize); + if (chunkSize > 0xFFFFFFFFUL) { + chunkSize = 0xFFFFFFFFUL; + } + return chunkSize; +} +MA_PRIVATE ma_uint64 ma_dr_wav__data_chunk_size_rf64(ma_uint64 dataChunkSize) +{ + return dataChunkSize; +} +MA_PRIVATE ma_bool32 ma_dr_wav_preinit_write(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_bool32 isSequential, ma_dr_wav_write_proc onWrite, ma_dr_wav_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pWav == NULL || onWrite == NULL) { + return MA_FALSE; + } + if (!isSequential && onSeek == NULL) { + return MA_FALSE; + } + if (pFormat->format == MA_DR_WAVE_FORMAT_EXTENSIBLE) { + return MA_FALSE; + } + if (pFormat->format == MA_DR_WAVE_FORMAT_ADPCM || pFormat->format == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + return MA_FALSE; + } + MA_DR_WAV_ZERO_MEMORY(pWav, sizeof(*pWav)); + pWav->onWrite = onWrite; + pWav->onSeek = onSeek; + pWav->pUserData = pUserData; + pWav->allocationCallbacks = ma_dr_wav_copy_allocation_callbacks_or_defaults(pAllocationCallbacks); + if (pWav->allocationCallbacks.onFree == NULL || (pWav->allocationCallbacks.onMalloc == NULL && pWav->allocationCallbacks.onRealloc == NULL)) { + return MA_FALSE; + } + pWav->fmt.formatTag = (ma_uint16)pFormat->format; + pWav->fmt.channels = (ma_uint16)pFormat->channels; + pWav->fmt.sampleRate = pFormat->sampleRate; + pWav->fmt.avgBytesPerSec = (ma_uint32)((pFormat->bitsPerSample * pFormat->sampleRate * pFormat->channels) / 8); + pWav->fmt.blockAlign = (ma_uint16)((pFormat->channels * pFormat->bitsPerSample) / 8); + pWav->fmt.bitsPerSample = (ma_uint16)pFormat->bitsPerSample; + pWav->fmt.extendedSize = 0; + pWav->isSequentialWrite = isSequential; + return MA_TRUE; +} +MA_PRIVATE ma_bool32 ma_dr_wav_init_write__internal(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount) +{ + size_t runningPos = 0; + ma_uint64 initialDataChunkSize = 0; + ma_uint64 chunkSizeFMT; + if (pWav->isSequentialWrite) { + initialDataChunkSize = (totalSampleCount * pWav->fmt.bitsPerSample) / 8; + if (pFormat->container == ma_dr_wav_container_riff) { + if (initialDataChunkSize > (0xFFFFFFFFUL - 36)) { + return MA_FALSE; + } + } + } + pWav->dataChunkDataSizeTargetWrite = initialDataChunkSize; + if (pFormat->container == ma_dr_wav_container_riff) { + ma_uint32 chunkSizeRIFF = 28 + (ma_uint32)initialDataChunkSize; + runningPos += ma_dr_wav__write(pWav, "RIFF", 4); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, chunkSizeRIFF); + runningPos += ma_dr_wav__write(pWav, "WAVE", 4); + } else if (pFormat->container == ma_dr_wav_container_w64) { + ma_uint64 chunkSizeRIFF = 80 + 24 + initialDataChunkSize; + runningPos += ma_dr_wav__write(pWav, ma_dr_wavGUID_W64_RIFF, 16); + runningPos += ma_dr_wav__write_u64ne_to_le(pWav, chunkSizeRIFF); + runningPos += ma_dr_wav__write(pWav, ma_dr_wavGUID_W64_WAVE, 16); + } else if (pFormat->container == ma_dr_wav_container_rf64) { + runningPos += ma_dr_wav__write(pWav, "RF64", 4); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, 0xFFFFFFFF); + runningPos += ma_dr_wav__write(pWav, "WAVE", 4); + } else { + return MA_FALSE; + } + if (pFormat->container == ma_dr_wav_container_rf64) { + ma_uint32 initialds64ChunkSize = 28; + ma_uint64 initialRiffChunkSize = 8 + initialds64ChunkSize + initialDataChunkSize; + runningPos += ma_dr_wav__write(pWav, "ds64", 4); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, initialds64ChunkSize); + runningPos += ma_dr_wav__write_u64ne_to_le(pWav, initialRiffChunkSize); + runningPos += ma_dr_wav__write_u64ne_to_le(pWav, initialDataChunkSize); + runningPos += ma_dr_wav__write_u64ne_to_le(pWav, totalSampleCount); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, 0); + } + if (pFormat->container == ma_dr_wav_container_riff || pFormat->container == ma_dr_wav_container_rf64) { + chunkSizeFMT = 16; + runningPos += ma_dr_wav__write(pWav, "fmt ", 4); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, (ma_uint32)chunkSizeFMT); + } else if (pFormat->container == ma_dr_wav_container_w64) { + chunkSizeFMT = 40; + runningPos += ma_dr_wav__write(pWav, ma_dr_wavGUID_W64_FMT, 16); + runningPos += ma_dr_wav__write_u64ne_to_le(pWav, chunkSizeFMT); + } + runningPos += ma_dr_wav__write_u16ne_to_le(pWav, pWav->fmt.formatTag); + runningPos += ma_dr_wav__write_u16ne_to_le(pWav, pWav->fmt.channels); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, pWav->fmt.sampleRate); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, pWav->fmt.avgBytesPerSec); + runningPos += ma_dr_wav__write_u16ne_to_le(pWav, pWav->fmt.blockAlign); + runningPos += ma_dr_wav__write_u16ne_to_le(pWav, pWav->fmt.bitsPerSample); + if (!pWav->isSequentialWrite && pWav->pMetadata != NULL && pWav->metadataCount > 0 && (pFormat->container == ma_dr_wav_container_riff || pFormat->container == ma_dr_wav_container_rf64)) { + runningPos += ma_dr_wav__write_or_count_metadata(pWav, pWav->pMetadata, pWav->metadataCount); + } + pWav->dataChunkDataPos = runningPos; + if (pFormat->container == ma_dr_wav_container_riff) { + ma_uint32 chunkSizeDATA = (ma_uint32)initialDataChunkSize; + runningPos += ma_dr_wav__write(pWav, "data", 4); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, chunkSizeDATA); + } else if (pFormat->container == ma_dr_wav_container_w64) { + ma_uint64 chunkSizeDATA = 24 + initialDataChunkSize; + runningPos += ma_dr_wav__write(pWav, ma_dr_wavGUID_W64_DATA, 16); + runningPos += ma_dr_wav__write_u64ne_to_le(pWav, chunkSizeDATA); + } else if (pFormat->container == ma_dr_wav_container_rf64) { + runningPos += ma_dr_wav__write(pWav, "data", 4); + runningPos += ma_dr_wav__write_u32ne_to_le(pWav, 0xFFFFFFFF); + } + pWav->container = pFormat->container; + pWav->channels = (ma_uint16)pFormat->channels; + pWav->sampleRate = pFormat->sampleRate; + pWav->bitsPerSample = (ma_uint16)pFormat->bitsPerSample; + pWav->translatedFormatTag = (ma_uint16)pFormat->format; + pWav->dataChunkDataPos = runningPos; + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_wav_init_write(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_dr_wav_write_proc onWrite, ma_dr_wav_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (!ma_dr_wav_preinit_write(pWav, pFormat, MA_FALSE, onWrite, onSeek, pUserData, pAllocationCallbacks)) { + return MA_FALSE; + } + return ma_dr_wav_init_write__internal(pWav, pFormat, 0); +} +MA_API ma_bool32 ma_dr_wav_init_write_sequential(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, ma_dr_wav_write_proc onWrite, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (!ma_dr_wav_preinit_write(pWav, pFormat, MA_TRUE, onWrite, NULL, pUserData, pAllocationCallbacks)) { + return MA_FALSE; + } + return ma_dr_wav_init_write__internal(pWav, pFormat, totalSampleCount); +} +MA_API ma_bool32 ma_dr_wav_init_write_sequential_pcm_frames(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_uint64 totalPCMFrameCount, ma_dr_wav_write_proc onWrite, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFormat == NULL) { + return MA_FALSE; + } + return ma_dr_wav_init_write_sequential(pWav, pFormat, totalPCMFrameCount*pFormat->channels, onWrite, pUserData, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_write_with_metadata(ma_dr_wav* pWav, const ma_dr_wav_data_format* pFormat, ma_dr_wav_write_proc onWrite, ma_dr_wav_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks, ma_dr_wav_metadata* pMetadata, ma_uint32 metadataCount) +{ + if (!ma_dr_wav_preinit_write(pWav, pFormat, MA_FALSE, onWrite, onSeek, pUserData, pAllocationCallbacks)) { + return MA_FALSE; + } + pWav->pMetadata = pMetadata; + pWav->metadataCount = metadataCount; + return ma_dr_wav_init_write__internal(pWav, pFormat, 0); +} +MA_API ma_uint64 ma_dr_wav_target_write_size_bytes(const ma_dr_wav_data_format* pFormat, ma_uint64 totalFrameCount, ma_dr_wav_metadata* pMetadata, ma_uint32 metadataCount) +{ + ma_uint64 targetDataSizeBytes = (ma_uint64)((ma_int64)totalFrameCount * pFormat->channels * pFormat->bitsPerSample/8.0); + ma_uint64 riffChunkSizeBytes; + ma_uint64 fileSizeBytes = 0; + if (pFormat->container == ma_dr_wav_container_riff) { + riffChunkSizeBytes = ma_dr_wav__riff_chunk_size_riff(targetDataSizeBytes, pMetadata, metadataCount); + fileSizeBytes = (8 + riffChunkSizeBytes); + } else if (pFormat->container == ma_dr_wav_container_w64) { + riffChunkSizeBytes = ma_dr_wav__riff_chunk_size_w64(targetDataSizeBytes); + fileSizeBytes = riffChunkSizeBytes; + } else if (pFormat->container == ma_dr_wav_container_rf64) { + riffChunkSizeBytes = ma_dr_wav__riff_chunk_size_rf64(targetDataSizeBytes, pMetadata, metadataCount); + fileSizeBytes = (8 + riffChunkSizeBytes); + } + return fileSizeBytes; +} +#ifndef MA_DR_WAV_NO_STDIO +MA_PRIVATE size_t ma_dr_wav__on_read_stdio(void* pUserData, void* pBufferOut, size_t bytesToRead) +{ + return fread(pBufferOut, 1, bytesToRead, (FILE*)pUserData); +} +MA_PRIVATE size_t ma_dr_wav__on_write_stdio(void* pUserData, const void* pData, size_t bytesToWrite) +{ + return fwrite(pData, 1, bytesToWrite, (FILE*)pUserData); +} +MA_PRIVATE ma_bool32 ma_dr_wav__on_seek_stdio(void* pUserData, int offset, ma_dr_wav_seek_origin origin) +{ + return fseek((FILE*)pUserData, offset, (origin == ma_dr_wav_seek_origin_current) ? SEEK_CUR : SEEK_SET) == 0; +} +MA_API ma_bool32 ma_dr_wav_init_file(ma_dr_wav* pWav, const char* filename, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_file_ex(pWav, filename, NULL, NULL, 0, pAllocationCallbacks); +} +MA_PRIVATE ma_bool32 ma_dr_wav_init_file__internal_FILE(ma_dr_wav* pWav, FILE* pFile, ma_dr_wav_chunk_proc onChunk, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_bool32 result; + result = ma_dr_wav_preinit(pWav, ma_dr_wav__on_read_stdio, ma_dr_wav__on_seek_stdio, (void*)pFile, pAllocationCallbacks); + if (result != MA_TRUE) { + fclose(pFile); + return result; + } + result = ma_dr_wav_init__internal(pWav, onChunk, pChunkUserData, flags); + if (result != MA_TRUE) { + fclose(pFile); + return result; + } + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_wav_init_file_ex(ma_dr_wav* pWav, const char* filename, ma_dr_wav_chunk_proc onChunk, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + FILE* pFile; + if (ma_fopen(&pFile, filename, "rb") != MA_SUCCESS) { + return MA_FALSE; + } + return ma_dr_wav_init_file__internal_FILE(pWav, pFile, onChunk, pChunkUserData, flags, pAllocationCallbacks); +} +#ifndef MA_DR_WAV_NO_WCHAR +MA_API ma_bool32 ma_dr_wav_init_file_w(ma_dr_wav* pWav, const wchar_t* filename, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_file_ex_w(pWav, filename, NULL, NULL, 0, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_file_ex_w(ma_dr_wav* pWav, const wchar_t* filename, ma_dr_wav_chunk_proc onChunk, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + FILE* pFile; + if (ma_wfopen(&pFile, filename, L"rb", pAllocationCallbacks) != MA_SUCCESS) { + return MA_FALSE; + } + return ma_dr_wav_init_file__internal_FILE(pWav, pFile, onChunk, pChunkUserData, flags, pAllocationCallbacks); +} +#endif +MA_API ma_bool32 ma_dr_wav_init_file_with_metadata(ma_dr_wav* pWav, const char* filename, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + FILE* pFile; + if (ma_fopen(&pFile, filename, "rb") != MA_SUCCESS) { + return MA_FALSE; + } + return ma_dr_wav_init_file__internal_FILE(pWav, pFile, NULL, NULL, flags | MA_DR_WAV_WITH_METADATA, pAllocationCallbacks); +} +#ifndef MA_DR_WAV_NO_WCHAR +MA_API ma_bool32 ma_dr_wav_init_file_with_metadata_w(ma_dr_wav* pWav, const wchar_t* filename, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + FILE* pFile; + if (ma_wfopen(&pFile, filename, L"rb", pAllocationCallbacks) != MA_SUCCESS) { + return MA_FALSE; + } + return ma_dr_wav_init_file__internal_FILE(pWav, pFile, NULL, NULL, flags | MA_DR_WAV_WITH_METADATA, pAllocationCallbacks); +} +#endif +MA_PRIVATE ma_bool32 ma_dr_wav_init_file_write__internal_FILE(ma_dr_wav* pWav, FILE* pFile, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, ma_bool32 isSequential, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_bool32 result; + result = ma_dr_wav_preinit_write(pWav, pFormat, isSequential, ma_dr_wav__on_write_stdio, ma_dr_wav__on_seek_stdio, (void*)pFile, pAllocationCallbacks); + if (result != MA_TRUE) { + fclose(pFile); + return result; + } + result = ma_dr_wav_init_write__internal(pWav, pFormat, totalSampleCount); + if (result != MA_TRUE) { + fclose(pFile); + return result; + } + return MA_TRUE; +} +MA_PRIVATE ma_bool32 ma_dr_wav_init_file_write__internal(ma_dr_wav* pWav, const char* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, ma_bool32 isSequential, const ma_allocation_callbacks* pAllocationCallbacks) +{ + FILE* pFile; + if (ma_fopen(&pFile, filename, "wb") != MA_SUCCESS) { + return MA_FALSE; + } + return ma_dr_wav_init_file_write__internal_FILE(pWav, pFile, pFormat, totalSampleCount, isSequential, pAllocationCallbacks); +} +#ifndef MA_DR_WAV_NO_WCHAR +MA_PRIVATE ma_bool32 ma_dr_wav_init_file_write_w__internal(ma_dr_wav* pWav, const wchar_t* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, ma_bool32 isSequential, const ma_allocation_callbacks* pAllocationCallbacks) +{ + FILE* pFile; + if (ma_wfopen(&pFile, filename, L"wb", pAllocationCallbacks) != MA_SUCCESS) { + return MA_FALSE; + } + return ma_dr_wav_init_file_write__internal_FILE(pWav, pFile, pFormat, totalSampleCount, isSequential, pAllocationCallbacks); +} +#endif +MA_API ma_bool32 ma_dr_wav_init_file_write(ma_dr_wav* pWav, const char* filename, const ma_dr_wav_data_format* pFormat, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_file_write__internal(pWav, filename, pFormat, 0, MA_FALSE, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_file_write_sequential(ma_dr_wav* pWav, const char* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_file_write__internal(pWav, filename, pFormat, totalSampleCount, MA_TRUE, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_file_write_sequential_pcm_frames(ma_dr_wav* pWav, const char* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFormat == NULL) { + return MA_FALSE; + } + return ma_dr_wav_init_file_write_sequential(pWav, filename, pFormat, totalPCMFrameCount*pFormat->channels, pAllocationCallbacks); +} +#ifndef MA_DR_WAV_NO_WCHAR +MA_API ma_bool32 ma_dr_wav_init_file_write_w(ma_dr_wav* pWav, const wchar_t* filename, const ma_dr_wav_data_format* pFormat, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_file_write_w__internal(pWav, filename, pFormat, 0, MA_FALSE, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_file_write_sequential_w(ma_dr_wav* pWav, const wchar_t* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_file_write_w__internal(pWav, filename, pFormat, totalSampleCount, MA_TRUE, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_file_write_sequential_pcm_frames_w(ma_dr_wav* pWav, const wchar_t* filename, const ma_dr_wav_data_format* pFormat, ma_uint64 totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFormat == NULL) { + return MA_FALSE; + } + return ma_dr_wav_init_file_write_sequential_w(pWav, filename, pFormat, totalPCMFrameCount*pFormat->channels, pAllocationCallbacks); +} +#endif +#endif +MA_PRIVATE size_t ma_dr_wav__on_read_memory(void* pUserData, void* pBufferOut, size_t bytesToRead) +{ + ma_dr_wav* pWav = (ma_dr_wav*)pUserData; + size_t bytesRemaining; + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(pWav->memoryStream.dataSize >= pWav->memoryStream.currentReadPos); + bytesRemaining = pWav->memoryStream.dataSize - pWav->memoryStream.currentReadPos; + if (bytesToRead > bytesRemaining) { + bytesToRead = bytesRemaining; + } + if (bytesToRead > 0) { + MA_DR_WAV_COPY_MEMORY(pBufferOut, pWav->memoryStream.data + pWav->memoryStream.currentReadPos, bytesToRead); + pWav->memoryStream.currentReadPos += bytesToRead; + } + return bytesToRead; +} +MA_PRIVATE ma_bool32 ma_dr_wav__on_seek_memory(void* pUserData, int offset, ma_dr_wav_seek_origin origin) +{ + ma_dr_wav* pWav = (ma_dr_wav*)pUserData; + MA_DR_WAV_ASSERT(pWav != NULL); + if (origin == ma_dr_wav_seek_origin_current) { + if (offset > 0) { + if (pWav->memoryStream.currentReadPos + offset > pWav->memoryStream.dataSize) { + return MA_FALSE; + } + } else { + if (pWav->memoryStream.currentReadPos < (size_t)-offset) { + return MA_FALSE; + } + } + pWav->memoryStream.currentReadPos += offset; + } else { + if ((ma_uint32)offset <= pWav->memoryStream.dataSize) { + pWav->memoryStream.currentReadPos = offset; + } else { + return MA_FALSE; + } + } + return MA_TRUE; +} +MA_PRIVATE size_t ma_dr_wav__on_write_memory(void* pUserData, const void* pDataIn, size_t bytesToWrite) +{ + ma_dr_wav* pWav = (ma_dr_wav*)pUserData; + size_t bytesRemaining; + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(pWav->memoryStreamWrite.dataCapacity >= pWav->memoryStreamWrite.currentWritePos); + bytesRemaining = pWav->memoryStreamWrite.dataCapacity - pWav->memoryStreamWrite.currentWritePos; + if (bytesRemaining < bytesToWrite) { + void* pNewData; + size_t newDataCapacity = (pWav->memoryStreamWrite.dataCapacity == 0) ? 256 : pWav->memoryStreamWrite.dataCapacity * 2; + if ((newDataCapacity - pWav->memoryStreamWrite.currentWritePos) < bytesToWrite) { + newDataCapacity = pWav->memoryStreamWrite.currentWritePos + bytesToWrite; + } + pNewData = ma_dr_wav__realloc_from_callbacks(*pWav->memoryStreamWrite.ppData, newDataCapacity, pWav->memoryStreamWrite.dataCapacity, &pWav->allocationCallbacks); + if (pNewData == NULL) { + return 0; + } + *pWav->memoryStreamWrite.ppData = pNewData; + pWav->memoryStreamWrite.dataCapacity = newDataCapacity; + } + MA_DR_WAV_COPY_MEMORY(((ma_uint8*)(*pWav->memoryStreamWrite.ppData)) + pWav->memoryStreamWrite.currentWritePos, pDataIn, bytesToWrite); + pWav->memoryStreamWrite.currentWritePos += bytesToWrite; + if (pWav->memoryStreamWrite.dataSize < pWav->memoryStreamWrite.currentWritePos) { + pWav->memoryStreamWrite.dataSize = pWav->memoryStreamWrite.currentWritePos; + } + *pWav->memoryStreamWrite.pDataSize = pWav->memoryStreamWrite.dataSize; + return bytesToWrite; +} +MA_PRIVATE ma_bool32 ma_dr_wav__on_seek_memory_write(void* pUserData, int offset, ma_dr_wav_seek_origin origin) +{ + ma_dr_wav* pWav = (ma_dr_wav*)pUserData; + MA_DR_WAV_ASSERT(pWav != NULL); + if (origin == ma_dr_wav_seek_origin_current) { + if (offset > 0) { + if (pWav->memoryStreamWrite.currentWritePos + offset > pWav->memoryStreamWrite.dataSize) { + offset = (int)(pWav->memoryStreamWrite.dataSize - pWav->memoryStreamWrite.currentWritePos); + } + } else { + if (pWav->memoryStreamWrite.currentWritePos < (size_t)-offset) { + offset = -(int)pWav->memoryStreamWrite.currentWritePos; + } + } + pWav->memoryStreamWrite.currentWritePos += offset; + } else { + if ((ma_uint32)offset <= pWav->memoryStreamWrite.dataSize) { + pWav->memoryStreamWrite.currentWritePos = offset; + } else { + pWav->memoryStreamWrite.currentWritePos = pWav->memoryStreamWrite.dataSize; + } + } + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_wav_init_memory(ma_dr_wav* pWav, const void* data, size_t dataSize, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_memory_ex(pWav, data, dataSize, NULL, NULL, 0, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_memory_ex(ma_dr_wav* pWav, const void* data, size_t dataSize, ma_dr_wav_chunk_proc onChunk, void* pChunkUserData, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (data == NULL || dataSize == 0) { + return MA_FALSE; + } + if (!ma_dr_wav_preinit(pWav, ma_dr_wav__on_read_memory, ma_dr_wav__on_seek_memory, pWav, pAllocationCallbacks)) { + return MA_FALSE; + } + pWav->memoryStream.data = (const ma_uint8*)data; + pWav->memoryStream.dataSize = dataSize; + pWav->memoryStream.currentReadPos = 0; + return ma_dr_wav_init__internal(pWav, onChunk, pChunkUserData, flags); +} +MA_API ma_bool32 ma_dr_wav_init_memory_with_metadata(ma_dr_wav* pWav, const void* data, size_t dataSize, ma_uint32 flags, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (data == NULL || dataSize == 0) { + return MA_FALSE; + } + if (!ma_dr_wav_preinit(pWav, ma_dr_wav__on_read_memory, ma_dr_wav__on_seek_memory, pWav, pAllocationCallbacks)) { + return MA_FALSE; + } + pWav->memoryStream.data = (const ma_uint8*)data; + pWav->memoryStream.dataSize = dataSize; + pWav->memoryStream.currentReadPos = 0; + return ma_dr_wav_init__internal(pWav, NULL, NULL, flags | MA_DR_WAV_WITH_METADATA); +} +MA_PRIVATE ma_bool32 ma_dr_wav_init_memory_write__internal(ma_dr_wav* pWav, void** ppData, size_t* pDataSize, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, ma_bool32 isSequential, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (ppData == NULL || pDataSize == NULL) { + return MA_FALSE; + } + *ppData = NULL; + *pDataSize = 0; + if (!ma_dr_wav_preinit_write(pWav, pFormat, isSequential, ma_dr_wav__on_write_memory, ma_dr_wav__on_seek_memory_write, pWav, pAllocationCallbacks)) { + return MA_FALSE; + } + pWav->memoryStreamWrite.ppData = ppData; + pWav->memoryStreamWrite.pDataSize = pDataSize; + pWav->memoryStreamWrite.dataSize = 0; + pWav->memoryStreamWrite.dataCapacity = 0; + pWav->memoryStreamWrite.currentWritePos = 0; + return ma_dr_wav_init_write__internal(pWav, pFormat, totalSampleCount); +} +MA_API ma_bool32 ma_dr_wav_init_memory_write(ma_dr_wav* pWav, void** ppData, size_t* pDataSize, const ma_dr_wav_data_format* pFormat, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_memory_write__internal(pWav, ppData, pDataSize, pFormat, 0, MA_FALSE, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_memory_write_sequential(ma_dr_wav* pWav, void** ppData, size_t* pDataSize, const ma_dr_wav_data_format* pFormat, ma_uint64 totalSampleCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_wav_init_memory_write__internal(pWav, ppData, pDataSize, pFormat, totalSampleCount, MA_TRUE, pAllocationCallbacks); +} +MA_API ma_bool32 ma_dr_wav_init_memory_write_sequential_pcm_frames(ma_dr_wav* pWav, void** ppData, size_t* pDataSize, const ma_dr_wav_data_format* pFormat, ma_uint64 totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pFormat == NULL) { + return MA_FALSE; + } + return ma_dr_wav_init_memory_write_sequential(pWav, ppData, pDataSize, pFormat, totalPCMFrameCount*pFormat->channels, pAllocationCallbacks); +} +MA_API ma_result ma_dr_wav_uninit(ma_dr_wav* pWav) +{ + ma_result result = MA_SUCCESS; + if (pWav == NULL) { + return MA_INVALID_ARGS; + } + if (pWav->onWrite != NULL) { + ma_uint32 paddingSize = 0; + if (pWav->container == ma_dr_wav_container_riff || pWav->container == ma_dr_wav_container_rf64) { + paddingSize = ma_dr_wav__chunk_padding_size_riff(pWav->dataChunkDataSize); + } else { + paddingSize = ma_dr_wav__chunk_padding_size_w64(pWav->dataChunkDataSize); + } + if (paddingSize > 0) { + ma_uint64 paddingData = 0; + ma_dr_wav__write(pWav, &paddingData, paddingSize); + } + if (pWav->onSeek && !pWav->isSequentialWrite) { + if (pWav->container == ma_dr_wav_container_riff) { + if (pWav->onSeek(pWav->pUserData, 4, ma_dr_wav_seek_origin_start)) { + ma_uint32 riffChunkSize = ma_dr_wav__riff_chunk_size_riff(pWav->dataChunkDataSize, pWav->pMetadata, pWav->metadataCount); + ma_dr_wav__write_u32ne_to_le(pWav, riffChunkSize); + } + if (pWav->onSeek(pWav->pUserData, (int)pWav->dataChunkDataPos - 4, ma_dr_wav_seek_origin_start)) { + ma_uint32 dataChunkSize = ma_dr_wav__data_chunk_size_riff(pWav->dataChunkDataSize); + ma_dr_wav__write_u32ne_to_le(pWav, dataChunkSize); + } + } else if (pWav->container == ma_dr_wav_container_w64) { + if (pWav->onSeek(pWav->pUserData, 16, ma_dr_wav_seek_origin_start)) { + ma_uint64 riffChunkSize = ma_dr_wav__riff_chunk_size_w64(pWav->dataChunkDataSize); + ma_dr_wav__write_u64ne_to_le(pWav, riffChunkSize); + } + if (pWav->onSeek(pWav->pUserData, (int)pWav->dataChunkDataPos - 8, ma_dr_wav_seek_origin_start)) { + ma_uint64 dataChunkSize = ma_dr_wav__data_chunk_size_w64(pWav->dataChunkDataSize); + ma_dr_wav__write_u64ne_to_le(pWav, dataChunkSize); + } + } else if (pWav->container == ma_dr_wav_container_rf64) { + int ds64BodyPos = 12 + 8; + if (pWav->onSeek(pWav->pUserData, ds64BodyPos + 0, ma_dr_wav_seek_origin_start)) { + ma_uint64 riffChunkSize = ma_dr_wav__riff_chunk_size_rf64(pWav->dataChunkDataSize, pWav->pMetadata, pWav->metadataCount); + ma_dr_wav__write_u64ne_to_le(pWav, riffChunkSize); + } + if (pWav->onSeek(pWav->pUserData, ds64BodyPos + 8, ma_dr_wav_seek_origin_start)) { + ma_uint64 dataChunkSize = ma_dr_wav__data_chunk_size_rf64(pWav->dataChunkDataSize); + ma_dr_wav__write_u64ne_to_le(pWav, dataChunkSize); + } + } + } + if (pWav->isSequentialWrite) { + if (pWav->dataChunkDataSize != pWav->dataChunkDataSizeTargetWrite) { + result = MA_INVALID_FILE; + } + } + } else { + ma_dr_wav_free(pWav->pMetadata, &pWav->allocationCallbacks); + } +#ifndef MA_DR_WAV_NO_STDIO + if (pWav->onRead == ma_dr_wav__on_read_stdio || pWav->onWrite == ma_dr_wav__on_write_stdio) { + fclose((FILE*)pWav->pUserData); + } +#endif + return result; +} +MA_API size_t ma_dr_wav_read_raw(ma_dr_wav* pWav, size_t bytesToRead, void* pBufferOut) +{ + size_t bytesRead; + ma_uint32 bytesPerFrame; + if (pWav == NULL || bytesToRead == 0) { + return 0; + } + if (bytesToRead > pWav->bytesRemaining) { + bytesToRead = (size_t)pWav->bytesRemaining; + } + if (bytesToRead == 0) { + return 0; + } + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + if (pBufferOut != NULL) { + bytesRead = pWav->onRead(pWav->pUserData, pBufferOut, bytesToRead); + } else { + bytesRead = 0; + while (bytesRead < bytesToRead) { + size_t bytesToSeek = (bytesToRead - bytesRead); + if (bytesToSeek > 0x7FFFFFFF) { + bytesToSeek = 0x7FFFFFFF; + } + if (pWav->onSeek(pWav->pUserData, (int)bytesToSeek, ma_dr_wav_seek_origin_current) == MA_FALSE) { + break; + } + bytesRead += bytesToSeek; + } + while (bytesRead < bytesToRead) { + ma_uint8 buffer[4096]; + size_t bytesSeeked; + size_t bytesToSeek = (bytesToRead - bytesRead); + if (bytesToSeek > sizeof(buffer)) { + bytesToSeek = sizeof(buffer); + } + bytesSeeked = pWav->onRead(pWav->pUserData, buffer, bytesToSeek); + bytesRead += bytesSeeked; + if (bytesSeeked < bytesToSeek) { + break; + } + } + } + pWav->readCursorInPCMFrames += bytesRead / bytesPerFrame; + pWav->bytesRemaining -= bytesRead; + return bytesRead; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_le(ma_dr_wav* pWav, ma_uint64 framesToRead, void* pBufferOut) +{ + ma_uint32 bytesPerFrame; + ma_uint64 bytesToRead; + ma_uint64 framesRemainingInFile; + if (pWav == NULL || framesToRead == 0) { + return 0; + } + if (ma_dr_wav__is_compressed_format_tag(pWav->translatedFormatTag)) { + return 0; + } + framesRemainingInFile = pWav->totalPCMFrameCount - pWav->readCursorInPCMFrames; + if (framesToRead > framesRemainingInFile) { + framesToRead = framesRemainingInFile; + } + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesToRead = framesToRead * bytesPerFrame; + if (bytesToRead > MA_SIZE_MAX) { + bytesToRead = (MA_SIZE_MAX / bytesPerFrame) * bytesPerFrame; + } + if (bytesToRead == 0) { + return 0; + } + return ma_dr_wav_read_raw(pWav, (size_t)bytesToRead, pBufferOut) / bytesPerFrame; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_be(ma_dr_wav* pWav, ma_uint64 framesToRead, void* pBufferOut) +{ + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_le(pWav, framesToRead, pBufferOut); + if (pBufferOut != NULL) { + ma_uint32 bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + ma_dr_wav__bswap_samples(pBufferOut, framesRead*pWav->channels, bytesPerFrame/pWav->channels); + } + return framesRead; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames(ma_dr_wav* pWav, ma_uint64 framesToRead, void* pBufferOut) +{ + ma_uint64 framesRead = 0; + if (ma_dr_wav_is_container_be(pWav->container)) { + if (pWav->container != ma_dr_wav_container_aiff || pWav->aiff.isLE == MA_FALSE) { + if (ma_dr_wav__is_little_endian()) { + framesRead = ma_dr_wav_read_pcm_frames_be(pWav, framesToRead, pBufferOut); + } else { + framesRead = ma_dr_wav_read_pcm_frames_le(pWav, framesToRead, pBufferOut); + } + goto post_process; + } + } + if (ma_dr_wav__is_little_endian()) { + framesRead = ma_dr_wav_read_pcm_frames_le(pWav, framesToRead, pBufferOut); + } else { + framesRead = ma_dr_wav_read_pcm_frames_be(pWav, framesToRead, pBufferOut); + } + post_process: + { + if (pWav->container == ma_dr_wav_container_aiff && pWav->bitsPerSample == 8 && pWav->aiff.isUnsigned == MA_FALSE) { + if (pBufferOut != NULL) { + ma_uint64 iSample; + for (iSample = 0; iSample < framesRead * pWav->channels; iSample += 1) { + ((ma_uint8*)pBufferOut)[iSample] += 128; + } + } + } + } + return framesRead; +} +MA_PRIVATE ma_bool32 ma_dr_wav_seek_to_first_pcm_frame(ma_dr_wav* pWav) +{ + if (pWav->onWrite != NULL) { + return MA_FALSE; + } + if (!pWav->onSeek(pWav->pUserData, (int)pWav->dataChunkDataPos, ma_dr_wav_seek_origin_start)) { + return MA_FALSE; + } + if (ma_dr_wav__is_compressed_format_tag(pWav->translatedFormatTag)) { + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM) { + MA_DR_WAV_ZERO_OBJECT(&pWav->msadpcm); + } else if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + MA_DR_WAV_ZERO_OBJECT(&pWav->ima); + } else { + MA_DR_WAV_ASSERT(MA_FALSE); + } + } + pWav->readCursorInPCMFrames = 0; + pWav->bytesRemaining = pWav->dataChunkDataSize; + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_wav_seek_to_pcm_frame(ma_dr_wav* pWav, ma_uint64 targetFrameIndex) +{ + if (pWav == NULL || pWav->onSeek == NULL) { + return MA_FALSE; + } + if (pWav->onWrite != NULL) { + return MA_FALSE; + } + if (pWav->totalPCMFrameCount == 0) { + return MA_TRUE; + } + if (targetFrameIndex > pWav->totalPCMFrameCount) { + targetFrameIndex = pWav->totalPCMFrameCount; + } + if (ma_dr_wav__is_compressed_format_tag(pWav->translatedFormatTag)) { + if (targetFrameIndex < pWav->readCursorInPCMFrames) { + if (!ma_dr_wav_seek_to_first_pcm_frame(pWav)) { + return MA_FALSE; + } + } + if (targetFrameIndex > pWav->readCursorInPCMFrames) { + ma_uint64 offsetInFrames = targetFrameIndex - pWav->readCursorInPCMFrames; + ma_int16 devnull[2048]; + while (offsetInFrames > 0) { + ma_uint64 framesRead = 0; + ma_uint64 framesToRead = offsetInFrames; + if (framesToRead > ma_dr_wav_countof(devnull)/pWav->channels) { + framesToRead = ma_dr_wav_countof(devnull)/pWav->channels; + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM) { + framesRead = ma_dr_wav_read_pcm_frames_s16__msadpcm(pWav, framesToRead, devnull); + } else if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + framesRead = ma_dr_wav_read_pcm_frames_s16__ima(pWav, framesToRead, devnull); + } else { + MA_DR_WAV_ASSERT(MA_FALSE); + } + if (framesRead != framesToRead) { + return MA_FALSE; + } + offsetInFrames -= framesRead; + } + } + } else { + ma_uint64 totalSizeInBytes; + ma_uint64 currentBytePos; + ma_uint64 targetBytePos; + ma_uint64 offset; + ma_uint32 bytesPerFrame; + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return MA_FALSE; + } + totalSizeInBytes = pWav->totalPCMFrameCount * bytesPerFrame; + currentBytePos = totalSizeInBytes - pWav->bytesRemaining; + targetBytePos = targetFrameIndex * bytesPerFrame; + if (currentBytePos < targetBytePos) { + offset = (targetBytePos - currentBytePos); + } else { + if (!ma_dr_wav_seek_to_first_pcm_frame(pWav)) { + return MA_FALSE; + } + offset = targetBytePos; + } + while (offset > 0) { + int offset32 = ((offset > INT_MAX) ? INT_MAX : (int)offset); + if (!pWav->onSeek(pWav->pUserData, offset32, ma_dr_wav_seek_origin_current)) { + return MA_FALSE; + } + pWav->readCursorInPCMFrames += offset32 / bytesPerFrame; + pWav->bytesRemaining -= offset32; + offset -= offset32; + } + } + return MA_TRUE; +} +MA_API ma_result ma_dr_wav_get_cursor_in_pcm_frames(ma_dr_wav* pWav, ma_uint64* pCursor) +{ + if (pCursor == NULL) { + return MA_INVALID_ARGS; + } + *pCursor = 0; + if (pWav == NULL) { + return MA_INVALID_ARGS; + } + *pCursor = pWav->readCursorInPCMFrames; + return MA_SUCCESS; +} +MA_API ma_result ma_dr_wav_get_length_in_pcm_frames(ma_dr_wav* pWav, ma_uint64* pLength) +{ + if (pLength == NULL) { + return MA_INVALID_ARGS; + } + *pLength = 0; + if (pWav == NULL) { + return MA_INVALID_ARGS; + } + *pLength = pWav->totalPCMFrameCount; + return MA_SUCCESS; +} +MA_API size_t ma_dr_wav_write_raw(ma_dr_wav* pWav, size_t bytesToWrite, const void* pData) +{ + size_t bytesWritten; + if (pWav == NULL || bytesToWrite == 0 || pData == NULL) { + return 0; + } + bytesWritten = pWav->onWrite(pWav->pUserData, pData, bytesToWrite); + pWav->dataChunkDataSize += bytesWritten; + return bytesWritten; +} +MA_API ma_uint64 ma_dr_wav_write_pcm_frames_le(ma_dr_wav* pWav, ma_uint64 framesToWrite, const void* pData) +{ + ma_uint64 bytesToWrite; + ma_uint64 bytesWritten; + const ma_uint8* pRunningData; + if (pWav == NULL || framesToWrite == 0 || pData == NULL) { + return 0; + } + bytesToWrite = ((framesToWrite * pWav->channels * pWav->bitsPerSample) / 8); + if (bytesToWrite > MA_SIZE_MAX) { + return 0; + } + bytesWritten = 0; + pRunningData = (const ma_uint8*)pData; + while (bytesToWrite > 0) { + size_t bytesJustWritten; + ma_uint64 bytesToWriteThisIteration; + bytesToWriteThisIteration = bytesToWrite; + MA_DR_WAV_ASSERT(bytesToWriteThisIteration <= MA_SIZE_MAX); + bytesJustWritten = ma_dr_wav_write_raw(pWav, (size_t)bytesToWriteThisIteration, pRunningData); + if (bytesJustWritten == 0) { + break; + } + bytesToWrite -= bytesJustWritten; + bytesWritten += bytesJustWritten; + pRunningData += bytesJustWritten; + } + return (bytesWritten * 8) / pWav->bitsPerSample / pWav->channels; +} +MA_API ma_uint64 ma_dr_wav_write_pcm_frames_be(ma_dr_wav* pWav, ma_uint64 framesToWrite, const void* pData) +{ + ma_uint64 bytesToWrite; + ma_uint64 bytesWritten; + ma_uint32 bytesPerSample; + const ma_uint8* pRunningData; + if (pWav == NULL || framesToWrite == 0 || pData == NULL) { + return 0; + } + bytesToWrite = ((framesToWrite * pWav->channels * pWav->bitsPerSample) / 8); + if (bytesToWrite > MA_SIZE_MAX) { + return 0; + } + bytesWritten = 0; + pRunningData = (const ma_uint8*)pData; + bytesPerSample = ma_dr_wav_get_bytes_per_pcm_frame(pWav) / pWav->channels; + if (bytesPerSample == 0) { + return 0; + } + while (bytesToWrite > 0) { + ma_uint8 temp[4096]; + ma_uint32 sampleCount; + size_t bytesJustWritten; + ma_uint64 bytesToWriteThisIteration; + bytesToWriteThisIteration = bytesToWrite; + MA_DR_WAV_ASSERT(bytesToWriteThisIteration <= MA_SIZE_MAX); + sampleCount = sizeof(temp)/bytesPerSample; + if (bytesToWriteThisIteration > ((ma_uint64)sampleCount)*bytesPerSample) { + bytesToWriteThisIteration = ((ma_uint64)sampleCount)*bytesPerSample; + } + MA_DR_WAV_COPY_MEMORY(temp, pRunningData, (size_t)bytesToWriteThisIteration); + ma_dr_wav__bswap_samples(temp, sampleCount, bytesPerSample); + bytesJustWritten = ma_dr_wav_write_raw(pWav, (size_t)bytesToWriteThisIteration, temp); + if (bytesJustWritten == 0) { + break; + } + bytesToWrite -= bytesJustWritten; + bytesWritten += bytesJustWritten; + pRunningData += bytesJustWritten; + } + return (bytesWritten * 8) / pWav->bitsPerSample / pWav->channels; +} +MA_API ma_uint64 ma_dr_wav_write_pcm_frames(ma_dr_wav* pWav, ma_uint64 framesToWrite, const void* pData) +{ + if (ma_dr_wav__is_little_endian()) { + return ma_dr_wav_write_pcm_frames_le(pWav, framesToWrite, pData); + } else { + return ma_dr_wav_write_pcm_frames_be(pWav, framesToWrite, pData); + } +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s16__msadpcm(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 totalFramesRead = 0; + static ma_int32 adaptationTable[] = { + 230, 230, 230, 230, 307, 409, 512, 614, + 768, 614, 512, 409, 307, 230, 230, 230 + }; + static ma_int32 coeff1Table[] = { 256, 512, 0, 192, 240, 460, 392 }; + static ma_int32 coeff2Table[] = { 0, -256, 0, 64, 0, -208, -232 }; + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(framesToRead > 0); + while (pWav->readCursorInPCMFrames < pWav->totalPCMFrameCount) { + MA_DR_WAV_ASSERT(framesToRead > 0); + if (pWav->msadpcm.cachedFrameCount == 0 && pWav->msadpcm.bytesRemainingInBlock == 0) { + if (pWav->channels == 1) { + ma_uint8 header[7]; + if (pWav->onRead(pWav->pUserData, header, sizeof(header)) != sizeof(header)) { + return totalFramesRead; + } + pWav->msadpcm.bytesRemainingInBlock = pWav->fmt.blockAlign - sizeof(header); + pWav->msadpcm.predictor[0] = header[0]; + pWav->msadpcm.delta[0] = ma_dr_wav_bytes_to_s16(header + 1); + pWav->msadpcm.prevFrames[0][1] = (ma_int32)ma_dr_wav_bytes_to_s16(header + 3); + pWav->msadpcm.prevFrames[0][0] = (ma_int32)ma_dr_wav_bytes_to_s16(header + 5); + pWav->msadpcm.cachedFrames[2] = pWav->msadpcm.prevFrames[0][0]; + pWav->msadpcm.cachedFrames[3] = pWav->msadpcm.prevFrames[0][1]; + pWav->msadpcm.cachedFrameCount = 2; + if (pWav->msadpcm.predictor[0] >= ma_dr_wav_countof(coeff1Table)) { + return totalFramesRead; + } + } else { + ma_uint8 header[14]; + if (pWav->onRead(pWav->pUserData, header, sizeof(header)) != sizeof(header)) { + return totalFramesRead; + } + pWav->msadpcm.bytesRemainingInBlock = pWav->fmt.blockAlign - sizeof(header); + pWav->msadpcm.predictor[0] = header[0]; + pWav->msadpcm.predictor[1] = header[1]; + pWav->msadpcm.delta[0] = ma_dr_wav_bytes_to_s16(header + 2); + pWav->msadpcm.delta[1] = ma_dr_wav_bytes_to_s16(header + 4); + pWav->msadpcm.prevFrames[0][1] = (ma_int32)ma_dr_wav_bytes_to_s16(header + 6); + pWav->msadpcm.prevFrames[1][1] = (ma_int32)ma_dr_wav_bytes_to_s16(header + 8); + pWav->msadpcm.prevFrames[0][0] = (ma_int32)ma_dr_wav_bytes_to_s16(header + 10); + pWav->msadpcm.prevFrames[1][0] = (ma_int32)ma_dr_wav_bytes_to_s16(header + 12); + pWav->msadpcm.cachedFrames[0] = pWav->msadpcm.prevFrames[0][0]; + pWav->msadpcm.cachedFrames[1] = pWav->msadpcm.prevFrames[1][0]; + pWav->msadpcm.cachedFrames[2] = pWav->msadpcm.prevFrames[0][1]; + pWav->msadpcm.cachedFrames[3] = pWav->msadpcm.prevFrames[1][1]; + pWav->msadpcm.cachedFrameCount = 2; + if (pWav->msadpcm.predictor[0] >= ma_dr_wav_countof(coeff1Table) || pWav->msadpcm.predictor[1] >= ma_dr_wav_countof(coeff2Table)) { + return totalFramesRead; + } + } + } + while (framesToRead > 0 && pWav->msadpcm.cachedFrameCount > 0 && pWav->readCursorInPCMFrames < pWav->totalPCMFrameCount) { + if (pBufferOut != NULL) { + ma_uint32 iSample = 0; + for (iSample = 0; iSample < pWav->channels; iSample += 1) { + pBufferOut[iSample] = (ma_int16)pWav->msadpcm.cachedFrames[(ma_dr_wav_countof(pWav->msadpcm.cachedFrames) - (pWav->msadpcm.cachedFrameCount*pWav->channels)) + iSample]; + } + pBufferOut += pWav->channels; + } + framesToRead -= 1; + totalFramesRead += 1; + pWav->readCursorInPCMFrames += 1; + pWav->msadpcm.cachedFrameCount -= 1; + } + if (framesToRead == 0) { + break; + } + if (pWav->msadpcm.cachedFrameCount == 0) { + if (pWav->msadpcm.bytesRemainingInBlock == 0) { + continue; + } else { + ma_uint8 nibbles; + ma_int32 nibble0; + ma_int32 nibble1; + if (pWav->onRead(pWav->pUserData, &nibbles, 1) != 1) { + return totalFramesRead; + } + pWav->msadpcm.bytesRemainingInBlock -= 1; + nibble0 = ((nibbles & 0xF0) >> 4); if ((nibbles & 0x80)) { nibble0 |= 0xFFFFFFF0UL; } + nibble1 = ((nibbles & 0x0F) >> 0); if ((nibbles & 0x08)) { nibble1 |= 0xFFFFFFF0UL; } + if (pWav->channels == 1) { + ma_int32 newSample0; + ma_int32 newSample1; + newSample0 = ((pWav->msadpcm.prevFrames[0][1] * coeff1Table[pWav->msadpcm.predictor[0]]) + (pWav->msadpcm.prevFrames[0][0] * coeff2Table[pWav->msadpcm.predictor[0]])) >> 8; + newSample0 += nibble0 * pWav->msadpcm.delta[0]; + newSample0 = ma_dr_wav_clamp(newSample0, -32768, 32767); + pWav->msadpcm.delta[0] = (adaptationTable[((nibbles & 0xF0) >> 4)] * pWav->msadpcm.delta[0]) >> 8; + if (pWav->msadpcm.delta[0] < 16) { + pWav->msadpcm.delta[0] = 16; + } + pWav->msadpcm.prevFrames[0][0] = pWav->msadpcm.prevFrames[0][1]; + pWav->msadpcm.prevFrames[0][1] = newSample0; + newSample1 = ((pWav->msadpcm.prevFrames[0][1] * coeff1Table[pWav->msadpcm.predictor[0]]) + (pWav->msadpcm.prevFrames[0][0] * coeff2Table[pWav->msadpcm.predictor[0]])) >> 8; + newSample1 += nibble1 * pWav->msadpcm.delta[0]; + newSample1 = ma_dr_wav_clamp(newSample1, -32768, 32767); + pWav->msadpcm.delta[0] = (adaptationTable[((nibbles & 0x0F) >> 0)] * pWav->msadpcm.delta[0]) >> 8; + if (pWav->msadpcm.delta[0] < 16) { + pWav->msadpcm.delta[0] = 16; + } + pWav->msadpcm.prevFrames[0][0] = pWav->msadpcm.prevFrames[0][1]; + pWav->msadpcm.prevFrames[0][1] = newSample1; + pWav->msadpcm.cachedFrames[2] = newSample0; + pWav->msadpcm.cachedFrames[3] = newSample1; + pWav->msadpcm.cachedFrameCount = 2; + } else { + ma_int32 newSample0; + ma_int32 newSample1; + newSample0 = ((pWav->msadpcm.prevFrames[0][1] * coeff1Table[pWav->msadpcm.predictor[0]]) + (pWav->msadpcm.prevFrames[0][0] * coeff2Table[pWav->msadpcm.predictor[0]])) >> 8; + newSample0 += nibble0 * pWav->msadpcm.delta[0]; + newSample0 = ma_dr_wav_clamp(newSample0, -32768, 32767); + pWav->msadpcm.delta[0] = (adaptationTable[((nibbles & 0xF0) >> 4)] * pWav->msadpcm.delta[0]) >> 8; + if (pWav->msadpcm.delta[0] < 16) { + pWav->msadpcm.delta[0] = 16; + } + pWav->msadpcm.prevFrames[0][0] = pWav->msadpcm.prevFrames[0][1]; + pWav->msadpcm.prevFrames[0][1] = newSample0; + newSample1 = ((pWav->msadpcm.prevFrames[1][1] * coeff1Table[pWav->msadpcm.predictor[1]]) + (pWav->msadpcm.prevFrames[1][0] * coeff2Table[pWav->msadpcm.predictor[1]])) >> 8; + newSample1 += nibble1 * pWav->msadpcm.delta[1]; + newSample1 = ma_dr_wav_clamp(newSample1, -32768, 32767); + pWav->msadpcm.delta[1] = (adaptationTable[((nibbles & 0x0F) >> 0)] * pWav->msadpcm.delta[1]) >> 8; + if (pWav->msadpcm.delta[1] < 16) { + pWav->msadpcm.delta[1] = 16; + } + pWav->msadpcm.prevFrames[1][0] = pWav->msadpcm.prevFrames[1][1]; + pWav->msadpcm.prevFrames[1][1] = newSample1; + pWav->msadpcm.cachedFrames[2] = newSample0; + pWav->msadpcm.cachedFrames[3] = newSample1; + pWav->msadpcm.cachedFrameCount = 1; + } + } + } + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s16__ima(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 totalFramesRead = 0; + ma_uint32 iChannel; + static ma_int32 indexTable[16] = { + -1, -1, -1, -1, 2, 4, 6, 8, + -1, -1, -1, -1, 2, 4, 6, 8 + }; + static ma_int32 stepTable[89] = { + 7, 8, 9, 10, 11, 12, 13, 14, 16, 17, + 19, 21, 23, 25, 28, 31, 34, 37, 41, 45, + 50, 55, 60, 66, 73, 80, 88, 97, 107, 118, + 130, 143, 157, 173, 190, 209, 230, 253, 279, 307, + 337, 371, 408, 449, 494, 544, 598, 658, 724, 796, + 876, 963, 1060, 1166, 1282, 1411, 1552, 1707, 1878, 2066, + 2272, 2499, 2749, 3024, 3327, 3660, 4026, 4428, 4871, 5358, + 5894, 6484, 7132, 7845, 8630, 9493, 10442, 11487, 12635, 13899, + 15289, 16818, 18500, 20350, 22385, 24623, 27086, 29794, 32767 + }; + MA_DR_WAV_ASSERT(pWav != NULL); + MA_DR_WAV_ASSERT(framesToRead > 0); + while (pWav->readCursorInPCMFrames < pWav->totalPCMFrameCount) { + MA_DR_WAV_ASSERT(framesToRead > 0); + if (pWav->ima.cachedFrameCount == 0 && pWav->ima.bytesRemainingInBlock == 0) { + if (pWav->channels == 1) { + ma_uint8 header[4]; + if (pWav->onRead(pWav->pUserData, header, sizeof(header)) != sizeof(header)) { + return totalFramesRead; + } + pWav->ima.bytesRemainingInBlock = pWav->fmt.blockAlign - sizeof(header); + if (header[2] >= ma_dr_wav_countof(stepTable)) { + pWav->onSeek(pWav->pUserData, pWav->ima.bytesRemainingInBlock, ma_dr_wav_seek_origin_current); + pWav->ima.bytesRemainingInBlock = 0; + return totalFramesRead; + } + pWav->ima.predictor[0] = (ma_int16)ma_dr_wav_bytes_to_u16(header + 0); + pWav->ima.stepIndex[0] = ma_dr_wav_clamp(header[2], 0, (ma_int32)ma_dr_wav_countof(stepTable)-1); + pWav->ima.cachedFrames[ma_dr_wav_countof(pWav->ima.cachedFrames) - 1] = pWav->ima.predictor[0]; + pWav->ima.cachedFrameCount = 1; + } else { + ma_uint8 header[8]; + if (pWav->onRead(pWav->pUserData, header, sizeof(header)) != sizeof(header)) { + return totalFramesRead; + } + pWav->ima.bytesRemainingInBlock = pWav->fmt.blockAlign - sizeof(header); + if (header[2] >= ma_dr_wav_countof(stepTable) || header[6] >= ma_dr_wav_countof(stepTable)) { + pWav->onSeek(pWav->pUserData, pWav->ima.bytesRemainingInBlock, ma_dr_wav_seek_origin_current); + pWav->ima.bytesRemainingInBlock = 0; + return totalFramesRead; + } + pWav->ima.predictor[0] = ma_dr_wav_bytes_to_s16(header + 0); + pWav->ima.stepIndex[0] = ma_dr_wav_clamp(header[2], 0, (ma_int32)ma_dr_wav_countof(stepTable)-1); + pWav->ima.predictor[1] = ma_dr_wav_bytes_to_s16(header + 4); + pWav->ima.stepIndex[1] = ma_dr_wav_clamp(header[6], 0, (ma_int32)ma_dr_wav_countof(stepTable)-1); + pWav->ima.cachedFrames[ma_dr_wav_countof(pWav->ima.cachedFrames) - 2] = pWav->ima.predictor[0]; + pWav->ima.cachedFrames[ma_dr_wav_countof(pWav->ima.cachedFrames) - 1] = pWav->ima.predictor[1]; + pWav->ima.cachedFrameCount = 1; + } + } + while (framesToRead > 0 && pWav->ima.cachedFrameCount > 0 && pWav->readCursorInPCMFrames < pWav->totalPCMFrameCount) { + if (pBufferOut != NULL) { + ma_uint32 iSample; + for (iSample = 0; iSample < pWav->channels; iSample += 1) { + pBufferOut[iSample] = (ma_int16)pWav->ima.cachedFrames[(ma_dr_wav_countof(pWav->ima.cachedFrames) - (pWav->ima.cachedFrameCount*pWav->channels)) + iSample]; + } + pBufferOut += pWav->channels; + } + framesToRead -= 1; + totalFramesRead += 1; + pWav->readCursorInPCMFrames += 1; + pWav->ima.cachedFrameCount -= 1; + } + if (framesToRead == 0) { + break; + } + if (pWav->ima.cachedFrameCount == 0) { + if (pWav->ima.bytesRemainingInBlock == 0) { + continue; + } else { + pWav->ima.cachedFrameCount = 8; + for (iChannel = 0; iChannel < pWav->channels; ++iChannel) { + ma_uint32 iByte; + ma_uint8 nibbles[4]; + if (pWav->onRead(pWav->pUserData, &nibbles, 4) != 4) { + pWav->ima.cachedFrameCount = 0; + return totalFramesRead; + } + pWav->ima.bytesRemainingInBlock -= 4; + for (iByte = 0; iByte < 4; ++iByte) { + ma_uint8 nibble0 = ((nibbles[iByte] & 0x0F) >> 0); + ma_uint8 nibble1 = ((nibbles[iByte] & 0xF0) >> 4); + ma_int32 step = stepTable[pWav->ima.stepIndex[iChannel]]; + ma_int32 predictor = pWav->ima.predictor[iChannel]; + ma_int32 diff = step >> 3; + if (nibble0 & 1) diff += step >> 2; + if (nibble0 & 2) diff += step >> 1; + if (nibble0 & 4) diff += step; + if (nibble0 & 8) diff = -diff; + predictor = ma_dr_wav_clamp(predictor + diff, -32768, 32767); + pWav->ima.predictor[iChannel] = predictor; + pWav->ima.stepIndex[iChannel] = ma_dr_wav_clamp(pWav->ima.stepIndex[iChannel] + indexTable[nibble0], 0, (ma_int32)ma_dr_wav_countof(stepTable)-1); + pWav->ima.cachedFrames[(ma_dr_wav_countof(pWav->ima.cachedFrames) - (pWav->ima.cachedFrameCount*pWav->channels)) + (iByte*2+0)*pWav->channels + iChannel] = predictor; + step = stepTable[pWav->ima.stepIndex[iChannel]]; + predictor = pWav->ima.predictor[iChannel]; + diff = step >> 3; + if (nibble1 & 1) diff += step >> 2; + if (nibble1 & 2) diff += step >> 1; + if (nibble1 & 4) diff += step; + if (nibble1 & 8) diff = -diff; + predictor = ma_dr_wav_clamp(predictor + diff, -32768, 32767); + pWav->ima.predictor[iChannel] = predictor; + pWav->ima.stepIndex[iChannel] = ma_dr_wav_clamp(pWav->ima.stepIndex[iChannel] + indexTable[nibble1], 0, (ma_int32)ma_dr_wav_countof(stepTable)-1); + pWav->ima.cachedFrames[(ma_dr_wav_countof(pWav->ima.cachedFrames) - (pWav->ima.cachedFrameCount*pWav->channels)) + (iByte*2+1)*pWav->channels + iChannel] = predictor; + } + } + } + } + } + return totalFramesRead; +} +#ifndef MA_DR_WAV_NO_CONVERSION_API +static unsigned short g_ma_dr_wavAlawTable[256] = { + 0xEA80, 0xEB80, 0xE880, 0xE980, 0xEE80, 0xEF80, 0xEC80, 0xED80, 0xE280, 0xE380, 0xE080, 0xE180, 0xE680, 0xE780, 0xE480, 0xE580, + 0xF540, 0xF5C0, 0xF440, 0xF4C0, 0xF740, 0xF7C0, 0xF640, 0xF6C0, 0xF140, 0xF1C0, 0xF040, 0xF0C0, 0xF340, 0xF3C0, 0xF240, 0xF2C0, + 0xAA00, 0xAE00, 0xA200, 0xA600, 0xBA00, 0xBE00, 0xB200, 0xB600, 0x8A00, 0x8E00, 0x8200, 0x8600, 0x9A00, 0x9E00, 0x9200, 0x9600, + 0xD500, 0xD700, 0xD100, 0xD300, 0xDD00, 0xDF00, 0xD900, 0xDB00, 0xC500, 0xC700, 0xC100, 0xC300, 0xCD00, 0xCF00, 0xC900, 0xCB00, + 0xFEA8, 0xFEB8, 0xFE88, 0xFE98, 0xFEE8, 0xFEF8, 0xFEC8, 0xFED8, 0xFE28, 0xFE38, 0xFE08, 0xFE18, 0xFE68, 0xFE78, 0xFE48, 0xFE58, + 0xFFA8, 0xFFB8, 0xFF88, 0xFF98, 0xFFE8, 0xFFF8, 0xFFC8, 0xFFD8, 0xFF28, 0xFF38, 0xFF08, 0xFF18, 0xFF68, 0xFF78, 0xFF48, 0xFF58, + 0xFAA0, 0xFAE0, 0xFA20, 0xFA60, 0xFBA0, 0xFBE0, 0xFB20, 0xFB60, 0xF8A0, 0xF8E0, 0xF820, 0xF860, 0xF9A0, 0xF9E0, 0xF920, 0xF960, + 0xFD50, 0xFD70, 0xFD10, 0xFD30, 0xFDD0, 0xFDF0, 0xFD90, 0xFDB0, 0xFC50, 0xFC70, 0xFC10, 0xFC30, 0xFCD0, 0xFCF0, 0xFC90, 0xFCB0, + 0x1580, 0x1480, 0x1780, 0x1680, 0x1180, 0x1080, 0x1380, 0x1280, 0x1D80, 0x1C80, 0x1F80, 0x1E80, 0x1980, 0x1880, 0x1B80, 0x1A80, + 0x0AC0, 0x0A40, 0x0BC0, 0x0B40, 0x08C0, 0x0840, 0x09C0, 0x0940, 0x0EC0, 0x0E40, 0x0FC0, 0x0F40, 0x0CC0, 0x0C40, 0x0DC0, 0x0D40, + 0x5600, 0x5200, 0x5E00, 0x5A00, 0x4600, 0x4200, 0x4E00, 0x4A00, 0x7600, 0x7200, 0x7E00, 0x7A00, 0x6600, 0x6200, 0x6E00, 0x6A00, + 0x2B00, 0x2900, 0x2F00, 0x2D00, 0x2300, 0x2100, 0x2700, 0x2500, 0x3B00, 0x3900, 0x3F00, 0x3D00, 0x3300, 0x3100, 0x3700, 0x3500, + 0x0158, 0x0148, 0x0178, 0x0168, 0x0118, 0x0108, 0x0138, 0x0128, 0x01D8, 0x01C8, 0x01F8, 0x01E8, 0x0198, 0x0188, 0x01B8, 0x01A8, + 0x0058, 0x0048, 0x0078, 0x0068, 0x0018, 0x0008, 0x0038, 0x0028, 0x00D8, 0x00C8, 0x00F8, 0x00E8, 0x0098, 0x0088, 0x00B8, 0x00A8, + 0x0560, 0x0520, 0x05E0, 0x05A0, 0x0460, 0x0420, 0x04E0, 0x04A0, 0x0760, 0x0720, 0x07E0, 0x07A0, 0x0660, 0x0620, 0x06E0, 0x06A0, + 0x02B0, 0x0290, 0x02F0, 0x02D0, 0x0230, 0x0210, 0x0270, 0x0250, 0x03B0, 0x0390, 0x03F0, 0x03D0, 0x0330, 0x0310, 0x0370, 0x0350 +}; +static unsigned short g_ma_dr_wavMulawTable[256] = { + 0x8284, 0x8684, 0x8A84, 0x8E84, 0x9284, 0x9684, 0x9A84, 0x9E84, 0xA284, 0xA684, 0xAA84, 0xAE84, 0xB284, 0xB684, 0xBA84, 0xBE84, + 0xC184, 0xC384, 0xC584, 0xC784, 0xC984, 0xCB84, 0xCD84, 0xCF84, 0xD184, 0xD384, 0xD584, 0xD784, 0xD984, 0xDB84, 0xDD84, 0xDF84, + 0xE104, 0xE204, 0xE304, 0xE404, 0xE504, 0xE604, 0xE704, 0xE804, 0xE904, 0xEA04, 0xEB04, 0xEC04, 0xED04, 0xEE04, 0xEF04, 0xF004, + 0xF0C4, 0xF144, 0xF1C4, 0xF244, 0xF2C4, 0xF344, 0xF3C4, 0xF444, 0xF4C4, 0xF544, 0xF5C4, 0xF644, 0xF6C4, 0xF744, 0xF7C4, 0xF844, + 0xF8A4, 0xF8E4, 0xF924, 0xF964, 0xF9A4, 0xF9E4, 0xFA24, 0xFA64, 0xFAA4, 0xFAE4, 0xFB24, 0xFB64, 0xFBA4, 0xFBE4, 0xFC24, 0xFC64, + 0xFC94, 0xFCB4, 0xFCD4, 0xFCF4, 0xFD14, 0xFD34, 0xFD54, 0xFD74, 0xFD94, 0xFDB4, 0xFDD4, 0xFDF4, 0xFE14, 0xFE34, 0xFE54, 0xFE74, + 0xFE8C, 0xFE9C, 0xFEAC, 0xFEBC, 0xFECC, 0xFEDC, 0xFEEC, 0xFEFC, 0xFF0C, 0xFF1C, 0xFF2C, 0xFF3C, 0xFF4C, 0xFF5C, 0xFF6C, 0xFF7C, + 0xFF88, 0xFF90, 0xFF98, 0xFFA0, 0xFFA8, 0xFFB0, 0xFFB8, 0xFFC0, 0xFFC8, 0xFFD0, 0xFFD8, 0xFFE0, 0xFFE8, 0xFFF0, 0xFFF8, 0x0000, + 0x7D7C, 0x797C, 0x757C, 0x717C, 0x6D7C, 0x697C, 0x657C, 0x617C, 0x5D7C, 0x597C, 0x557C, 0x517C, 0x4D7C, 0x497C, 0x457C, 0x417C, + 0x3E7C, 0x3C7C, 0x3A7C, 0x387C, 0x367C, 0x347C, 0x327C, 0x307C, 0x2E7C, 0x2C7C, 0x2A7C, 0x287C, 0x267C, 0x247C, 0x227C, 0x207C, + 0x1EFC, 0x1DFC, 0x1CFC, 0x1BFC, 0x1AFC, 0x19FC, 0x18FC, 0x17FC, 0x16FC, 0x15FC, 0x14FC, 0x13FC, 0x12FC, 0x11FC, 0x10FC, 0x0FFC, + 0x0F3C, 0x0EBC, 0x0E3C, 0x0DBC, 0x0D3C, 0x0CBC, 0x0C3C, 0x0BBC, 0x0B3C, 0x0ABC, 0x0A3C, 0x09BC, 0x093C, 0x08BC, 0x083C, 0x07BC, + 0x075C, 0x071C, 0x06DC, 0x069C, 0x065C, 0x061C, 0x05DC, 0x059C, 0x055C, 0x051C, 0x04DC, 0x049C, 0x045C, 0x041C, 0x03DC, 0x039C, + 0x036C, 0x034C, 0x032C, 0x030C, 0x02EC, 0x02CC, 0x02AC, 0x028C, 0x026C, 0x024C, 0x022C, 0x020C, 0x01EC, 0x01CC, 0x01AC, 0x018C, + 0x0174, 0x0164, 0x0154, 0x0144, 0x0134, 0x0124, 0x0114, 0x0104, 0x00F4, 0x00E4, 0x00D4, 0x00C4, 0x00B4, 0x00A4, 0x0094, 0x0084, + 0x0078, 0x0070, 0x0068, 0x0060, 0x0058, 0x0050, 0x0048, 0x0040, 0x0038, 0x0030, 0x0028, 0x0020, 0x0018, 0x0010, 0x0008, 0x0000 +}; +static MA_INLINE ma_int16 ma_dr_wav__alaw_to_s16(ma_uint8 sampleIn) +{ + return (short)g_ma_dr_wavAlawTable[sampleIn]; +} +static MA_INLINE ma_int16 ma_dr_wav__mulaw_to_s16(ma_uint8 sampleIn) +{ + return (short)g_ma_dr_wavMulawTable[sampleIn]; +} +MA_PRIVATE void ma_dr_wav__pcm_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t totalSampleCount, unsigned int bytesPerSample) +{ + size_t i; + if (bytesPerSample == 1) { + ma_dr_wav_u8_to_s16(pOut, pIn, totalSampleCount); + return; + } + if (bytesPerSample == 2) { + for (i = 0; i < totalSampleCount; ++i) { + *pOut++ = ((const ma_int16*)pIn)[i]; + } + return; + } + if (bytesPerSample == 3) { + ma_dr_wav_s24_to_s16(pOut, pIn, totalSampleCount); + return; + } + if (bytesPerSample == 4) { + ma_dr_wav_s32_to_s16(pOut, (const ma_int32*)pIn, totalSampleCount); + return; + } + if (bytesPerSample > 8) { + MA_DR_WAV_ZERO_MEMORY(pOut, totalSampleCount * sizeof(*pOut)); + return; + } + for (i = 0; i < totalSampleCount; ++i) { + ma_uint64 sample = 0; + unsigned int shift = (8 - bytesPerSample) * 8; + unsigned int j; + for (j = 0; j < bytesPerSample; j += 1) { + MA_DR_WAV_ASSERT(j < 8); + sample |= (ma_uint64)(pIn[j]) << shift; + shift += 8; + } + pIn += j; + *pOut++ = (ma_int16)((ma_int64)sample >> 48); + } +} +MA_PRIVATE void ma_dr_wav__ieee_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t totalSampleCount, unsigned int bytesPerSample) +{ + if (bytesPerSample == 4) { + ma_dr_wav_f32_to_s16(pOut, (const float*)pIn, totalSampleCount); + return; + } else if (bytesPerSample == 8) { + ma_dr_wav_f64_to_s16(pOut, (const double*)pIn, totalSampleCount); + return; + } else { + MA_DR_WAV_ZERO_MEMORY(pOut, totalSampleCount * sizeof(*pOut)); + return; + } +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s16__pcm(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + if ((pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_PCM && pWav->bitsPerSample == 16) || pBufferOut == NULL) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, pBufferOut); + } + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav__pcm_to_s16(pBufferOut, sampleData, (size_t)samplesRead, bytesPerSample); + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s16__ieee(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + if (pBufferOut == NULL) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, NULL); + } + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav__ieee_to_s16(pBufferOut, sampleData, (size_t)samplesRead, bytesPerSample); + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s16__alaw(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + if (pBufferOut == NULL) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, NULL); + } + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav_alaw_to_s16(pBufferOut, sampleData, (size_t)samplesRead); + #ifdef MA_DR_WAV_LIBSNDFILE_COMPAT + { + if (pWav->container == ma_dr_wav_container_aiff) { + ma_uint64 iSample; + for (iSample = 0; iSample < samplesRead; iSample += 1) { + pBufferOut[iSample] = -pBufferOut[iSample]; + } + } + } + #endif + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s16__mulaw(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + if (pBufferOut == NULL) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, NULL); + } + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav_mulaw_to_s16(pBufferOut, sampleData, (size_t)samplesRead); + #ifdef MA_DR_WAV_LIBSNDFILE_COMPAT + { + if (pWav->container == ma_dr_wav_container_aiff) { + ma_uint64 iSample; + for (iSample = 0; iSample < samplesRead; iSample += 1) { + pBufferOut[iSample] = -pBufferOut[iSample]; + } + } + } + #endif + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s16(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + if (pWav == NULL || framesToRead == 0) { + return 0; + } + if (pBufferOut == NULL) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, NULL); + } + if (framesToRead * pWav->channels * sizeof(ma_int16) > MA_SIZE_MAX) { + framesToRead = MA_SIZE_MAX / sizeof(ma_int16) / pWav->channels; + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_PCM) { + return ma_dr_wav_read_pcm_frames_s16__pcm(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_IEEE_FLOAT) { + return ma_dr_wav_read_pcm_frames_s16__ieee(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ALAW) { + return ma_dr_wav_read_pcm_frames_s16__alaw(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_MULAW) { + return ma_dr_wav_read_pcm_frames_s16__mulaw(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM) { + return ma_dr_wav_read_pcm_frames_s16__msadpcm(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + return ma_dr_wav_read_pcm_frames_s16__ima(pWav, framesToRead, pBufferOut); + } + return 0; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s16le(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_s16(pWav, framesToRead, pBufferOut); + if (pBufferOut != NULL && ma_dr_wav__is_little_endian() == MA_FALSE) { + ma_dr_wav__bswap_samples_s16(pBufferOut, framesRead*pWav->channels); + } + return framesRead; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s16be(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_s16(pWav, framesToRead, pBufferOut); + if (pBufferOut != NULL && ma_dr_wav__is_little_endian() == MA_TRUE) { + ma_dr_wav__bswap_samples_s16(pBufferOut, framesRead*pWav->channels); + } + return framesRead; +} +MA_API void ma_dr_wav_u8_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + int r; + size_t i; + for (i = 0; i < sampleCount; ++i) { + int x = pIn[i]; + r = x << 8; + r = r - 32768; + pOut[i] = (short)r; + } +} +MA_API void ma_dr_wav_s24_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + int r; + size_t i; + for (i = 0; i < sampleCount; ++i) { + int x = ((int)(((unsigned int)(((const ma_uint8*)pIn)[i*3+0]) << 8) | ((unsigned int)(((const ma_uint8*)pIn)[i*3+1]) << 16) | ((unsigned int)(((const ma_uint8*)pIn)[i*3+2])) << 24)) >> 8; + r = x >> 8; + pOut[i] = (short)r; + } +} +MA_API void ma_dr_wav_s32_to_s16(ma_int16* pOut, const ma_int32* pIn, size_t sampleCount) +{ + int r; + size_t i; + for (i = 0; i < sampleCount; ++i) { + int x = pIn[i]; + r = x >> 16; + pOut[i] = (short)r; + } +} +MA_API void ma_dr_wav_f32_to_s16(ma_int16* pOut, const float* pIn, size_t sampleCount) +{ + int r; + size_t i; + for (i = 0; i < sampleCount; ++i) { + float x = pIn[i]; + float c; + c = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); + c = c + 1; + r = (int)(c * 32767.5f); + r = r - 32768; + pOut[i] = (short)r; + } +} +MA_API void ma_dr_wav_f64_to_s16(ma_int16* pOut, const double* pIn, size_t sampleCount) +{ + int r; + size_t i; + for (i = 0; i < sampleCount; ++i) { + double x = pIn[i]; + double c; + c = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); + c = c + 1; + r = (int)(c * 32767.5); + r = r - 32768; + pOut[i] = (short)r; + } +} +MA_API void ma_dr_wav_alaw_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + for (i = 0; i < sampleCount; ++i) { + pOut[i] = ma_dr_wav__alaw_to_s16(pIn[i]); + } +} +MA_API void ma_dr_wav_mulaw_to_s16(ma_int16* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + for (i = 0; i < sampleCount; ++i) { + pOut[i] = ma_dr_wav__mulaw_to_s16(pIn[i]); + } +} +MA_PRIVATE void ma_dr_wav__pcm_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount, unsigned int bytesPerSample) +{ + unsigned int i; + if (bytesPerSample == 1) { + ma_dr_wav_u8_to_f32(pOut, pIn, sampleCount); + return; + } + if (bytesPerSample == 2) { + ma_dr_wav_s16_to_f32(pOut, (const ma_int16*)pIn, sampleCount); + return; + } + if (bytesPerSample == 3) { + ma_dr_wav_s24_to_f32(pOut, pIn, sampleCount); + return; + } + if (bytesPerSample == 4) { + ma_dr_wav_s32_to_f32(pOut, (const ma_int32*)pIn, sampleCount); + return; + } + if (bytesPerSample > 8) { + MA_DR_WAV_ZERO_MEMORY(pOut, sampleCount * sizeof(*pOut)); + return; + } + for (i = 0; i < sampleCount; ++i) { + ma_uint64 sample = 0; + unsigned int shift = (8 - bytesPerSample) * 8; + unsigned int j; + for (j = 0; j < bytesPerSample; j += 1) { + MA_DR_WAV_ASSERT(j < 8); + sample |= (ma_uint64)(pIn[j]) << shift; + shift += 8; + } + pIn += j; + *pOut++ = (float)((ma_int64)sample / 9223372036854775807.0); + } +} +MA_PRIVATE void ma_dr_wav__ieee_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount, unsigned int bytesPerSample) +{ + if (bytesPerSample == 4) { + unsigned int i; + for (i = 0; i < sampleCount; ++i) { + *pOut++ = ((const float*)pIn)[i]; + } + return; + } else if (bytesPerSample == 8) { + ma_dr_wav_f64_to_f32(pOut, (const double*)pIn, sampleCount); + return; + } else { + MA_DR_WAV_ZERO_MEMORY(pOut, sampleCount * sizeof(*pOut)); + return; + } +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_f32__pcm(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav__pcm_to_f32(pBufferOut, sampleData, (size_t)samplesRead, bytesPerSample); + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_f32__msadpcm_ima(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_int16 samples16[2048]; + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, ma_dr_wav_countof(samples16)/pWav->channels); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_s16(pWav, framesToReadThisIteration, samples16); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + ma_dr_wav_s16_to_f32(pBufferOut, samples16, (size_t)(framesRead*pWav->channels)); + pBufferOut += framesRead*pWav->channels; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_f32__ieee(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_IEEE_FLOAT && pWav->bitsPerSample == 32) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, pBufferOut); + } + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav__ieee_to_f32(pBufferOut, sampleData, (size_t)samplesRead, bytesPerSample); + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_f32__alaw(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav_alaw_to_f32(pBufferOut, sampleData, (size_t)samplesRead); + #ifdef MA_DR_WAV_LIBSNDFILE_COMPAT + { + if (pWav->container == ma_dr_wav_container_aiff) { + ma_uint64 iSample; + for (iSample = 0; iSample < samplesRead; iSample += 1) { + pBufferOut[iSample] = -pBufferOut[iSample]; + } + } + } + #endif + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_f32__mulaw(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav_mulaw_to_f32(pBufferOut, sampleData, (size_t)samplesRead); + #ifdef MA_DR_WAV_LIBSNDFILE_COMPAT + { + if (pWav->container == ma_dr_wav_container_aiff) { + ma_uint64 iSample; + for (iSample = 0; iSample < samplesRead; iSample += 1) { + pBufferOut[iSample] = -pBufferOut[iSample]; + } + } + } + #endif + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_f32(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut) +{ + if (pWav == NULL || framesToRead == 0) { + return 0; + } + if (pBufferOut == NULL) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, NULL); + } + if (framesToRead * pWav->channels * sizeof(float) > MA_SIZE_MAX) { + framesToRead = MA_SIZE_MAX / sizeof(float) / pWav->channels; + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_PCM) { + return ma_dr_wav_read_pcm_frames_f32__pcm(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM || pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + return ma_dr_wav_read_pcm_frames_f32__msadpcm_ima(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_IEEE_FLOAT) { + return ma_dr_wav_read_pcm_frames_f32__ieee(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ALAW) { + return ma_dr_wav_read_pcm_frames_f32__alaw(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_MULAW) { + return ma_dr_wav_read_pcm_frames_f32__mulaw(pWav, framesToRead, pBufferOut); + } + return 0; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_f32le(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut) +{ + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_f32(pWav, framesToRead, pBufferOut); + if (pBufferOut != NULL && ma_dr_wav__is_little_endian() == MA_FALSE) { + ma_dr_wav__bswap_samples_f32(pBufferOut, framesRead*pWav->channels); + } + return framesRead; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_f32be(ma_dr_wav* pWav, ma_uint64 framesToRead, float* pBufferOut) +{ + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_f32(pWav, framesToRead, pBufferOut); + if (pBufferOut != NULL && ma_dr_wav__is_little_endian() == MA_TRUE) { + ma_dr_wav__bswap_samples_f32(pBufferOut, framesRead*pWav->channels); + } + return framesRead; +} +MA_API void ma_dr_wav_u8_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } +#ifdef MA_DR_WAV_LIBSNDFILE_COMPAT + for (i = 0; i < sampleCount; ++i) { + *pOut++ = (pIn[i] / 256.0f) * 2 - 1; + } +#else + for (i = 0; i < sampleCount; ++i) { + float x = pIn[i]; + x = x * 0.00784313725490196078f; + x = x - 1; + *pOut++ = x; + } +#endif +} +MA_API void ma_dr_wav_s16_to_f32(float* pOut, const ma_int16* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = pIn[i] * 0.000030517578125f; + } +} +MA_API void ma_dr_wav_s24_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + double x; + ma_uint32 a = ((ma_uint32)(pIn[i*3+0]) << 8); + ma_uint32 b = ((ma_uint32)(pIn[i*3+1]) << 16); + ma_uint32 c = ((ma_uint32)(pIn[i*3+2]) << 24); + x = (double)((ma_int32)(a | b | c) >> 8); + *pOut++ = (float)(x * 0.00000011920928955078125); + } +} +MA_API void ma_dr_wav_s32_to_f32(float* pOut, const ma_int32* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = (float)(pIn[i] / 2147483648.0); + } +} +MA_API void ma_dr_wav_f64_to_f32(float* pOut, const double* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = (float)pIn[i]; + } +} +MA_API void ma_dr_wav_alaw_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = ma_dr_wav__alaw_to_s16(pIn[i]) / 32768.0f; + } +} +MA_API void ma_dr_wav_mulaw_to_f32(float* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = ma_dr_wav__mulaw_to_s16(pIn[i]) / 32768.0f; + } +} +MA_PRIVATE void ma_dr_wav__pcm_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t totalSampleCount, unsigned int bytesPerSample) +{ + unsigned int i; + if (bytesPerSample == 1) { + ma_dr_wav_u8_to_s32(pOut, pIn, totalSampleCount); + return; + } + if (bytesPerSample == 2) { + ma_dr_wav_s16_to_s32(pOut, (const ma_int16*)pIn, totalSampleCount); + return; + } + if (bytesPerSample == 3) { + ma_dr_wav_s24_to_s32(pOut, pIn, totalSampleCount); + return; + } + if (bytesPerSample == 4) { + for (i = 0; i < totalSampleCount; ++i) { + *pOut++ = ((const ma_int32*)pIn)[i]; + } + return; + } + if (bytesPerSample > 8) { + MA_DR_WAV_ZERO_MEMORY(pOut, totalSampleCount * sizeof(*pOut)); + return; + } + for (i = 0; i < totalSampleCount; ++i) { + ma_uint64 sample = 0; + unsigned int shift = (8 - bytesPerSample) * 8; + unsigned int j; + for (j = 0; j < bytesPerSample; j += 1) { + MA_DR_WAV_ASSERT(j < 8); + sample |= (ma_uint64)(pIn[j]) << shift; + shift += 8; + } + pIn += j; + *pOut++ = (ma_int32)((ma_int64)sample >> 32); + } +} +MA_PRIVATE void ma_dr_wav__ieee_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t totalSampleCount, unsigned int bytesPerSample) +{ + if (bytesPerSample == 4) { + ma_dr_wav_f32_to_s32(pOut, (const float*)pIn, totalSampleCount); + return; + } else if (bytesPerSample == 8) { + ma_dr_wav_f64_to_s32(pOut, (const double*)pIn, totalSampleCount); + return; + } else { + MA_DR_WAV_ZERO_MEMORY(pOut, totalSampleCount * sizeof(*pOut)); + return; + } +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s32__pcm(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_PCM && pWav->bitsPerSample == 32) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, pBufferOut); + } + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav__pcm_to_s32(pBufferOut, sampleData, (size_t)samplesRead, bytesPerSample); + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s32__msadpcm_ima(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + ma_uint64 totalFramesRead = 0; + ma_int16 samples16[2048]; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, ma_dr_wav_countof(samples16)/pWav->channels); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_s16(pWav, framesToReadThisIteration, samples16); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + ma_dr_wav_s16_to_s32(pBufferOut, samples16, (size_t)(framesRead*pWav->channels)); + pBufferOut += framesRead*pWav->channels; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s32__ieee(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav__ieee_to_s32(pBufferOut, sampleData, (size_t)samplesRead, bytesPerSample); + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s32__alaw(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav_alaw_to_s32(pBufferOut, sampleData, (size_t)samplesRead); + #ifdef MA_DR_WAV_LIBSNDFILE_COMPAT + { + if (pWav->container == ma_dr_wav_container_aiff) { + ma_uint64 iSample; + for (iSample = 0; iSample < samplesRead; iSample += 1) { + pBufferOut[iSample] = -pBufferOut[iSample]; + } + } + } + #endif + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_PRIVATE ma_uint64 ma_dr_wav_read_pcm_frames_s32__mulaw(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + ma_uint64 totalFramesRead; + ma_uint8 sampleData[4096] = {0}; + ma_uint32 bytesPerFrame; + ma_uint32 bytesPerSample; + ma_uint64 samplesRead; + bytesPerFrame = ma_dr_wav_get_bytes_per_pcm_frame(pWav); + if (bytesPerFrame == 0) { + return 0; + } + bytesPerSample = bytesPerFrame / pWav->channels; + if (bytesPerSample == 0 || (bytesPerFrame % pWav->channels) != 0) { + return 0; + } + totalFramesRead = 0; + while (framesToRead > 0) { + ma_uint64 framesToReadThisIteration = ma_dr_wav_min(framesToRead, sizeof(sampleData)/bytesPerFrame); + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames(pWav, framesToReadThisIteration, sampleData); + if (framesRead == 0) { + break; + } + MA_DR_WAV_ASSERT(framesRead <= framesToReadThisIteration); + samplesRead = framesRead * pWav->channels; + if ((samplesRead * bytesPerSample) > sizeof(sampleData)) { + MA_DR_WAV_ASSERT(MA_FALSE); + break; + } + ma_dr_wav_mulaw_to_s32(pBufferOut, sampleData, (size_t)samplesRead); + #ifdef MA_DR_WAV_LIBSNDFILE_COMPAT + { + if (pWav->container == ma_dr_wav_container_aiff) { + ma_uint64 iSample; + for (iSample = 0; iSample < samplesRead; iSample += 1) { + pBufferOut[iSample] = -pBufferOut[iSample]; + } + } + } + #endif + pBufferOut += samplesRead; + framesToRead -= framesRead; + totalFramesRead += framesRead; + } + return totalFramesRead; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s32(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + if (pWav == NULL || framesToRead == 0) { + return 0; + } + if (pBufferOut == NULL) { + return ma_dr_wav_read_pcm_frames(pWav, framesToRead, NULL); + } + if (framesToRead * pWav->channels * sizeof(ma_int32) > MA_SIZE_MAX) { + framesToRead = MA_SIZE_MAX / sizeof(ma_int32) / pWav->channels; + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_PCM) { + return ma_dr_wav_read_pcm_frames_s32__pcm(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ADPCM || pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_DVI_ADPCM) { + return ma_dr_wav_read_pcm_frames_s32__msadpcm_ima(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_IEEE_FLOAT) { + return ma_dr_wav_read_pcm_frames_s32__ieee(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_ALAW) { + return ma_dr_wav_read_pcm_frames_s32__alaw(pWav, framesToRead, pBufferOut); + } + if (pWav->translatedFormatTag == MA_DR_WAVE_FORMAT_MULAW) { + return ma_dr_wav_read_pcm_frames_s32__mulaw(pWav, framesToRead, pBufferOut); + } + return 0; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s32le(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_s32(pWav, framesToRead, pBufferOut); + if (pBufferOut != NULL && ma_dr_wav__is_little_endian() == MA_FALSE) { + ma_dr_wav__bswap_samples_s32(pBufferOut, framesRead*pWav->channels); + } + return framesRead; +} +MA_API ma_uint64 ma_dr_wav_read_pcm_frames_s32be(ma_dr_wav* pWav, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + ma_uint64 framesRead = ma_dr_wav_read_pcm_frames_s32(pWav, framesToRead, pBufferOut); + if (pBufferOut != NULL && ma_dr_wav__is_little_endian() == MA_TRUE) { + ma_dr_wav__bswap_samples_s32(pBufferOut, framesRead*pWav->channels); + } + return framesRead; +} +MA_API void ma_dr_wav_u8_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = ((int)pIn[i] - 128) << 24; + } +} +MA_API void ma_dr_wav_s16_to_s32(ma_int32* pOut, const ma_int16* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = pIn[i] << 16; + } +} +MA_API void ma_dr_wav_s24_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + unsigned int s0 = pIn[i*3 + 0]; + unsigned int s1 = pIn[i*3 + 1]; + unsigned int s2 = pIn[i*3 + 2]; + ma_int32 sample32 = (ma_int32)((s0 << 8) | (s1 << 16) | (s2 << 24)); + *pOut++ = sample32; + } +} +MA_API void ma_dr_wav_f32_to_s32(ma_int32* pOut, const float* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = (ma_int32)(2147483648.0f * pIn[i]); + } +} +MA_API void ma_dr_wav_f64_to_s32(ma_int32* pOut, const double* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = (ma_int32)(2147483648.0 * pIn[i]); + } +} +MA_API void ma_dr_wav_alaw_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i = 0; i < sampleCount; ++i) { + *pOut++ = ((ma_int32)ma_dr_wav__alaw_to_s16(pIn[i])) << 16; + } +} +MA_API void ma_dr_wav_mulaw_to_s32(ma_int32* pOut, const ma_uint8* pIn, size_t sampleCount) +{ + size_t i; + if (pOut == NULL || pIn == NULL) { + return; + } + for (i= 0; i < sampleCount; ++i) { + *pOut++ = ((ma_int32)ma_dr_wav__mulaw_to_s16(pIn[i])) << 16; + } +} +MA_PRIVATE ma_int16* ma_dr_wav__read_pcm_frames_and_close_s16(ma_dr_wav* pWav, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalFrameCount) +{ + ma_uint64 sampleDataSize; + ma_int16* pSampleData; + ma_uint64 framesRead; + MA_DR_WAV_ASSERT(pWav != NULL); + sampleDataSize = pWav->totalPCMFrameCount * pWav->channels * sizeof(ma_int16); + if (sampleDataSize > MA_SIZE_MAX) { + ma_dr_wav_uninit(pWav); + return NULL; + } + pSampleData = (ma_int16*)ma_dr_wav__malloc_from_callbacks((size_t)sampleDataSize, &pWav->allocationCallbacks); + if (pSampleData == NULL) { + ma_dr_wav_uninit(pWav); + return NULL; + } + framesRead = ma_dr_wav_read_pcm_frames_s16(pWav, (size_t)pWav->totalPCMFrameCount, pSampleData); + if (framesRead != pWav->totalPCMFrameCount) { + ma_dr_wav__free_from_callbacks(pSampleData, &pWav->allocationCallbacks); + ma_dr_wav_uninit(pWav); + return NULL; + } + ma_dr_wav_uninit(pWav); + if (sampleRate) { + *sampleRate = pWav->sampleRate; + } + if (channels) { + *channels = pWav->channels; + } + if (totalFrameCount) { + *totalFrameCount = pWav->totalPCMFrameCount; + } + return pSampleData; +} +MA_PRIVATE float* ma_dr_wav__read_pcm_frames_and_close_f32(ma_dr_wav* pWav, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalFrameCount) +{ + ma_uint64 sampleDataSize; + float* pSampleData; + ma_uint64 framesRead; + MA_DR_WAV_ASSERT(pWav != NULL); + sampleDataSize = pWav->totalPCMFrameCount * pWav->channels * sizeof(float); + if (sampleDataSize > MA_SIZE_MAX) { + ma_dr_wav_uninit(pWav); + return NULL; + } + pSampleData = (float*)ma_dr_wav__malloc_from_callbacks((size_t)sampleDataSize, &pWav->allocationCallbacks); + if (pSampleData == NULL) { + ma_dr_wav_uninit(pWav); + return NULL; + } + framesRead = ma_dr_wav_read_pcm_frames_f32(pWav, (size_t)pWav->totalPCMFrameCount, pSampleData); + if (framesRead != pWav->totalPCMFrameCount) { + ma_dr_wav__free_from_callbacks(pSampleData, &pWav->allocationCallbacks); + ma_dr_wav_uninit(pWav); + return NULL; + } + ma_dr_wav_uninit(pWav); + if (sampleRate) { + *sampleRate = pWav->sampleRate; + } + if (channels) { + *channels = pWav->channels; + } + if (totalFrameCount) { + *totalFrameCount = pWav->totalPCMFrameCount; + } + return pSampleData; +} +MA_PRIVATE ma_int32* ma_dr_wav__read_pcm_frames_and_close_s32(ma_dr_wav* pWav, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalFrameCount) +{ + ma_uint64 sampleDataSize; + ma_int32* pSampleData; + ma_uint64 framesRead; + MA_DR_WAV_ASSERT(pWav != NULL); + sampleDataSize = pWav->totalPCMFrameCount * pWav->channels * sizeof(ma_int32); + if (sampleDataSize > MA_SIZE_MAX) { + ma_dr_wav_uninit(pWav); + return NULL; + } + pSampleData = (ma_int32*)ma_dr_wav__malloc_from_callbacks((size_t)sampleDataSize, &pWav->allocationCallbacks); + if (pSampleData == NULL) { + ma_dr_wav_uninit(pWav); + return NULL; + } + framesRead = ma_dr_wav_read_pcm_frames_s32(pWav, (size_t)pWav->totalPCMFrameCount, pSampleData); + if (framesRead != pWav->totalPCMFrameCount) { + ma_dr_wav__free_from_callbacks(pSampleData, &pWav->allocationCallbacks); + ma_dr_wav_uninit(pWav); + return NULL; + } + ma_dr_wav_uninit(pWav); + if (sampleRate) { + *sampleRate = pWav->sampleRate; + } + if (channels) { + *channels = pWav->channels; + } + if (totalFrameCount) { + *totalFrameCount = pWav->totalPCMFrameCount; + } + return pSampleData; +} +MA_API ma_int16* ma_dr_wav_open_and_read_pcm_frames_s16(ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init(&wav, onRead, onSeek, pUserData, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_s16(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +MA_API float* ma_dr_wav_open_and_read_pcm_frames_f32(ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init(&wav, onRead, onSeek, pUserData, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_f32(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +MA_API ma_int32* ma_dr_wav_open_and_read_pcm_frames_s32(ma_dr_wav_read_proc onRead, ma_dr_wav_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init(&wav, onRead, onSeek, pUserData, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_s32(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +#ifndef MA_DR_WAV_NO_STDIO +MA_API ma_int16* ma_dr_wav_open_file_and_read_pcm_frames_s16(const char* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_file(&wav, filename, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_s16(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +MA_API float* ma_dr_wav_open_file_and_read_pcm_frames_f32(const char* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_file(&wav, filename, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_f32(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +MA_API ma_int32* ma_dr_wav_open_file_and_read_pcm_frames_s32(const char* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_file(&wav, filename, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_s32(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +#ifndef MA_DR_WAV_NO_WCHAR +MA_API ma_int16* ma_dr_wav_open_file_and_read_pcm_frames_s16_w(const wchar_t* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (channelsOut) { + *channelsOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_file_w(&wav, filename, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_s16(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +MA_API float* ma_dr_wav_open_file_and_read_pcm_frames_f32_w(const wchar_t* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (channelsOut) { + *channelsOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_file_w(&wav, filename, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_f32(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +MA_API ma_int32* ma_dr_wav_open_file_and_read_pcm_frames_s32_w(const wchar_t* filename, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (channelsOut) { + *channelsOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_file_w(&wav, filename, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_s32(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +#endif +#endif +MA_API ma_int16* ma_dr_wav_open_memory_and_read_pcm_frames_s16(const void* data, size_t dataSize, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_memory(&wav, data, dataSize, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_s16(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +MA_API float* ma_dr_wav_open_memory_and_read_pcm_frames_f32(const void* data, size_t dataSize, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_memory(&wav, data, dataSize, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_f32(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +MA_API ma_int32* ma_dr_wav_open_memory_and_read_pcm_frames_s32(const void* data, size_t dataSize, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_wav wav; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalFrameCountOut) { + *totalFrameCountOut = 0; + } + if (!ma_dr_wav_init_memory(&wav, data, dataSize, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_wav__read_pcm_frames_and_close_s32(&wav, channelsOut, sampleRateOut, totalFrameCountOut); +} +#endif +MA_API void ma_dr_wav_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks != NULL) { + ma_dr_wav__free_from_callbacks(p, pAllocationCallbacks); + } else { + ma_dr_wav__free_default(p, NULL); + } +} +MA_API ma_uint16 ma_dr_wav_bytes_to_u16(const ma_uint8* data) +{ + return ((ma_uint16)data[0] << 0) | ((ma_uint16)data[1] << 8); +} +MA_API ma_int16 ma_dr_wav_bytes_to_s16(const ma_uint8* data) +{ + return (ma_int16)ma_dr_wav_bytes_to_u16(data); +} +MA_API ma_uint32 ma_dr_wav_bytes_to_u32(const ma_uint8* data) +{ + return ma_dr_wav_bytes_to_u32_le(data); +} +MA_API float ma_dr_wav_bytes_to_f32(const ma_uint8* data) +{ + union { + ma_uint32 u32; + float f32; + } value; + value.u32 = ma_dr_wav_bytes_to_u32(data); + return value.f32; +} +MA_API ma_int32 ma_dr_wav_bytes_to_s32(const ma_uint8* data) +{ + return (ma_int32)ma_dr_wav_bytes_to_u32(data); +} +MA_API ma_uint64 ma_dr_wav_bytes_to_u64(const ma_uint8* data) +{ + return + ((ma_uint64)data[0] << 0) | ((ma_uint64)data[1] << 8) | ((ma_uint64)data[2] << 16) | ((ma_uint64)data[3] << 24) | + ((ma_uint64)data[4] << 32) | ((ma_uint64)data[5] << 40) | ((ma_uint64)data[6] << 48) | ((ma_uint64)data[7] << 56); +} +MA_API ma_int64 ma_dr_wav_bytes_to_s64(const ma_uint8* data) +{ + return (ma_int64)ma_dr_wav_bytes_to_u64(data); +} +MA_API ma_bool32 ma_dr_wav_guid_equal(const ma_uint8 a[16], const ma_uint8 b[16]) +{ + int i; + for (i = 0; i < 16; i += 1) { + if (a[i] != b[i]) { + return MA_FALSE; + } + } + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_wav_fourcc_equal(const ma_uint8* a, const char* b) +{ + return + a[0] == b[0] && + a[1] == b[1] && + a[2] == b[2] && + a[3] == b[3]; +} +#ifdef __MRC__ +#pragma options opt reset +#endif +#endif +/* dr_wav_c end */ +#endif /* MA_DR_WAV_IMPLEMENTATION */ +#endif /* MA_NO_WAV */ + +#if !defined(MA_NO_FLAC) && !defined(MA_NO_DECODING) +#if !defined(MA_DR_FLAC_IMPLEMENTATION) +/* dr_flac_c begin */ +#ifndef ma_dr_flac_c +#define ma_dr_flac_c +#if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic push + #if __GNUC__ >= 7 + #pragma GCC diagnostic ignored "-Wimplicit-fallthrough" + #endif +#endif +#ifdef __linux__ + #ifndef _BSD_SOURCE + #define _BSD_SOURCE + #endif + #ifndef _DEFAULT_SOURCE + #define _DEFAULT_SOURCE + #endif + #ifndef __USE_BSD + #define __USE_BSD + #endif + #include +#endif +#include +#include +#if !defined(MA_DR_FLAC_NO_SIMD) + #if defined(MA_X64) || defined(MA_X86) + #if defined(_MSC_VER) && !defined(__clang__) + #if _MSC_VER >= 1400 && !defined(MA_DR_FLAC_NO_SSE2) + #define MA_DR_FLAC_SUPPORT_SSE2 + #endif + #if _MSC_VER >= 1600 && !defined(MA_DR_FLAC_NO_SSE41) + #define MA_DR_FLAC_SUPPORT_SSE41 + #endif + #elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 3))) + #if defined(__SSE2__) && !defined(MA_DR_FLAC_NO_SSE2) + #define MA_DR_FLAC_SUPPORT_SSE2 + #endif + #if defined(__SSE4_1__) && !defined(MA_DR_FLAC_NO_SSE41) + #define MA_DR_FLAC_SUPPORT_SSE41 + #endif + #endif + #if !defined(__GNUC__) && !defined(__clang__) && defined(__has_include) + #if !defined(MA_DR_FLAC_SUPPORT_SSE2) && !defined(MA_DR_FLAC_NO_SSE2) && __has_include() + #define MA_DR_FLAC_SUPPORT_SSE2 + #endif + #if !defined(MA_DR_FLAC_SUPPORT_SSE41) && !defined(MA_DR_FLAC_NO_SSE41) && __has_include() + #define MA_DR_FLAC_SUPPORT_SSE41 + #endif + #endif + #if defined(MA_DR_FLAC_SUPPORT_SSE41) + #include + #elif defined(MA_DR_FLAC_SUPPORT_SSE2) + #include + #endif + #endif + #if defined(MA_ARM) + #if !defined(MA_DR_FLAC_NO_NEON) && (defined(__ARM_NEON) || defined(__aarch64__) || defined(_M_ARM64)) + #define MA_DR_FLAC_SUPPORT_NEON + #include + #endif + #endif +#endif +#if !defined(MA_DR_FLAC_NO_SIMD) && (defined(MA_X86) || defined(MA_X64)) + #if defined(_MSC_VER) && !defined(__clang__) + #if _MSC_VER >= 1400 + #include + static void ma_dr_flac__cpuid(int info[4], int fid) + { + __cpuid(info, fid); + } + #else + #define MA_DR_FLAC_NO_CPUID + #endif + #else + #if defined(__GNUC__) || defined(__clang__) + static void ma_dr_flac__cpuid(int info[4], int fid) + { + #if defined(MA_X86) && defined(__PIC__) + __asm__ __volatile__ ( + "xchg{l} {%%}ebx, %k1;" + "cpuid;" + "xchg{l} {%%}ebx, %k1;" + : "=a"(info[0]), "=&r"(info[1]), "=c"(info[2]), "=d"(info[3]) : "a"(fid), "c"(0) + ); + #else + __asm__ __volatile__ ( + "cpuid" : "=a"(info[0]), "=b"(info[1]), "=c"(info[2]), "=d"(info[3]) : "a"(fid), "c"(0) + ); + #endif + } + #else + #define MA_DR_FLAC_NO_CPUID + #endif + #endif +#else + #define MA_DR_FLAC_NO_CPUID +#endif +static MA_INLINE ma_bool32 ma_dr_flac_has_sse2(void) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + #if (defined(MA_X64) || defined(MA_X86)) && !defined(MA_DR_FLAC_NO_SSE2) + #if defined(MA_X64) + return MA_TRUE; + #elif (defined(_M_IX86_FP) && _M_IX86_FP == 2) || defined(__SSE2__) + return MA_TRUE; + #else + #if defined(MA_DR_FLAC_NO_CPUID) + return MA_FALSE; + #else + int info[4]; + ma_dr_flac__cpuid(info, 1); + return (info[3] & (1 << 26)) != 0; + #endif + #endif + #else + return MA_FALSE; + #endif +#else + return MA_FALSE; +#endif +} +static MA_INLINE ma_bool32 ma_dr_flac_has_sse41(void) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE41) + #if (defined(MA_X64) || defined(MA_X86)) && !defined(MA_DR_FLAC_NO_SSE41) + #if defined(__SSE4_1__) || defined(__AVX__) + return MA_TRUE; + #else + #if defined(MA_DR_FLAC_NO_CPUID) + return MA_FALSE; + #else + int info[4]; + ma_dr_flac__cpuid(info, 1); + return (info[2] & (1 << 19)) != 0; + #endif + #endif + #else + return MA_FALSE; + #endif +#else + return MA_FALSE; +#endif +} +#if defined(_MSC_VER) && _MSC_VER >= 1500 && (defined(MA_X86) || defined(MA_X64)) && !defined(__clang__) + #define MA_DR_FLAC_HAS_LZCNT_INTRINSIC +#elif (defined(__GNUC__) && ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 7))) + #define MA_DR_FLAC_HAS_LZCNT_INTRINSIC +#elif defined(__clang__) + #if defined(__has_builtin) + #if __has_builtin(__builtin_clzll) || __has_builtin(__builtin_clzl) + #define MA_DR_FLAC_HAS_LZCNT_INTRINSIC + #endif + #endif +#endif +#if defined(_MSC_VER) && _MSC_VER >= 1400 && !defined(__clang__) + #define MA_DR_FLAC_HAS_BYTESWAP16_INTRINSIC + #define MA_DR_FLAC_HAS_BYTESWAP32_INTRINSIC + #define MA_DR_FLAC_HAS_BYTESWAP64_INTRINSIC +#elif defined(__clang__) + #if defined(__has_builtin) + #if __has_builtin(__builtin_bswap16) + #define MA_DR_FLAC_HAS_BYTESWAP16_INTRINSIC + #endif + #if __has_builtin(__builtin_bswap32) + #define MA_DR_FLAC_HAS_BYTESWAP32_INTRINSIC + #endif + #if __has_builtin(__builtin_bswap64) + #define MA_DR_FLAC_HAS_BYTESWAP64_INTRINSIC + #endif + #endif +#elif defined(__GNUC__) + #if ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 3)) + #define MA_DR_FLAC_HAS_BYTESWAP32_INTRINSIC + #define MA_DR_FLAC_HAS_BYTESWAP64_INTRINSIC + #endif + #if ((__GNUC__ > 4) || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)) + #define MA_DR_FLAC_HAS_BYTESWAP16_INTRINSIC + #endif +#elif defined(__WATCOMC__) && defined(__386__) + #define MA_DR_FLAC_HAS_BYTESWAP16_INTRINSIC + #define MA_DR_FLAC_HAS_BYTESWAP32_INTRINSIC + #define MA_DR_FLAC_HAS_BYTESWAP64_INTRINSIC + extern __inline ma_uint16 _watcom_bswap16(ma_uint16); + extern __inline ma_uint32 _watcom_bswap32(ma_uint32); + extern __inline ma_uint64 _watcom_bswap64(ma_uint64); +#pragma aux _watcom_bswap16 = \ + "xchg al, ah" \ + parm [ax] \ + value [ax] \ + modify nomemory; +#pragma aux _watcom_bswap32 = \ + "bswap eax" \ + parm [eax] \ + value [eax] \ + modify nomemory; +#pragma aux _watcom_bswap64 = \ + "bswap eax" \ + "bswap edx" \ + "xchg eax,edx" \ + parm [eax edx] \ + value [eax edx] \ + modify nomemory; +#endif +#ifndef MA_DR_FLAC_ASSERT +#include +#define MA_DR_FLAC_ASSERT(expression) assert(expression) +#endif +#ifndef MA_DR_FLAC_MALLOC +#define MA_DR_FLAC_MALLOC(sz) malloc((sz)) +#endif +#ifndef MA_DR_FLAC_REALLOC +#define MA_DR_FLAC_REALLOC(p, sz) realloc((p), (sz)) +#endif +#ifndef MA_DR_FLAC_FREE +#define MA_DR_FLAC_FREE(p) free((p)) +#endif +#ifndef MA_DR_FLAC_COPY_MEMORY +#define MA_DR_FLAC_COPY_MEMORY(dst, src, sz) memcpy((dst), (src), (sz)) +#endif +#ifndef MA_DR_FLAC_ZERO_MEMORY +#define MA_DR_FLAC_ZERO_MEMORY(p, sz) memset((p), 0, (sz)) +#endif +#ifndef MA_DR_FLAC_ZERO_OBJECT +#define MA_DR_FLAC_ZERO_OBJECT(p) MA_DR_FLAC_ZERO_MEMORY((p), sizeof(*(p))) +#endif +#define MA_DR_FLAC_MAX_SIMD_VECTOR_SIZE 64 +#define MA_DR_FLAC_SUBFRAME_CONSTANT 0 +#define MA_DR_FLAC_SUBFRAME_VERBATIM 1 +#define MA_DR_FLAC_SUBFRAME_FIXED 8 +#define MA_DR_FLAC_SUBFRAME_LPC 32 +#define MA_DR_FLAC_SUBFRAME_RESERVED 255 +#define MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE 0 +#define MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE2 1 +#define MA_DR_FLAC_CHANNEL_ASSIGNMENT_INDEPENDENT 0 +#define MA_DR_FLAC_CHANNEL_ASSIGNMENT_LEFT_SIDE 8 +#define MA_DR_FLAC_CHANNEL_ASSIGNMENT_RIGHT_SIDE 9 +#define MA_DR_FLAC_CHANNEL_ASSIGNMENT_MID_SIDE 10 +#define MA_DR_FLAC_SEEKPOINT_SIZE_IN_BYTES 18 +#define MA_DR_FLAC_CUESHEET_TRACK_SIZE_IN_BYTES 36 +#define MA_DR_FLAC_CUESHEET_TRACK_INDEX_SIZE_IN_BYTES 12 +#define ma_dr_flac_align(x, a) ((((x) + (a) - 1) / (a)) * (a)) +MA_API void ma_dr_flac_version(ma_uint32* pMajor, ma_uint32* pMinor, ma_uint32* pRevision) +{ + if (pMajor) { + *pMajor = MA_DR_FLAC_VERSION_MAJOR; + } + if (pMinor) { + *pMinor = MA_DR_FLAC_VERSION_MINOR; + } + if (pRevision) { + *pRevision = MA_DR_FLAC_VERSION_REVISION; + } +} +MA_API const char* ma_dr_flac_version_string(void) +{ + return MA_DR_FLAC_VERSION_STRING; +} +#if defined(__has_feature) + #if __has_feature(thread_sanitizer) + #define MA_DR_FLAC_NO_THREAD_SANITIZE __attribute__((no_sanitize("thread"))) + #else + #define MA_DR_FLAC_NO_THREAD_SANITIZE + #endif +#else + #define MA_DR_FLAC_NO_THREAD_SANITIZE +#endif +#if defined(MA_DR_FLAC_HAS_LZCNT_INTRINSIC) +static ma_bool32 ma_dr_flac__gIsLZCNTSupported = MA_FALSE; +#endif +#ifndef MA_DR_FLAC_NO_CPUID +static ma_bool32 ma_dr_flac__gIsSSE2Supported = MA_FALSE; +static ma_bool32 ma_dr_flac__gIsSSE41Supported = MA_FALSE; +MA_DR_FLAC_NO_THREAD_SANITIZE static void ma_dr_flac__init_cpu_caps(void) +{ + static ma_bool32 isCPUCapsInitialized = MA_FALSE; + if (!isCPUCapsInitialized) { +#if defined(MA_DR_FLAC_HAS_LZCNT_INTRINSIC) + int info[4] = {0}; + ma_dr_flac__cpuid(info, 0x80000001); + ma_dr_flac__gIsLZCNTSupported = (info[2] & (1 << 5)) != 0; +#endif + ma_dr_flac__gIsSSE2Supported = ma_dr_flac_has_sse2(); + ma_dr_flac__gIsSSE41Supported = ma_dr_flac_has_sse41(); + isCPUCapsInitialized = MA_TRUE; + } +} +#else +static ma_bool32 ma_dr_flac__gIsNEONSupported = MA_FALSE; +static MA_INLINE ma_bool32 ma_dr_flac__has_neon(void) +{ +#if defined(MA_DR_FLAC_SUPPORT_NEON) + #if defined(MA_ARM) && !defined(MA_DR_FLAC_NO_NEON) + #if (defined(__ARM_NEON) || defined(__aarch64__) || defined(_M_ARM64)) + return MA_TRUE; + #else + return MA_FALSE; + #endif + #else + return MA_FALSE; + #endif +#else + return MA_FALSE; +#endif +} +MA_DR_FLAC_NO_THREAD_SANITIZE static void ma_dr_flac__init_cpu_caps(void) +{ + ma_dr_flac__gIsNEONSupported = ma_dr_flac__has_neon(); +#if defined(MA_DR_FLAC_HAS_LZCNT_INTRINSIC) && defined(MA_ARM) && (defined(__ARM_ARCH) && __ARM_ARCH >= 5) + ma_dr_flac__gIsLZCNTSupported = MA_TRUE; +#endif +} +#endif +static MA_INLINE ma_bool32 ma_dr_flac__is_little_endian(void) +{ +#if defined(MA_X86) || defined(MA_X64) + return MA_TRUE; +#elif defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && __BYTE_ORDER == __LITTLE_ENDIAN + return MA_TRUE; +#else + int n = 1; + return (*(char*)&n) == 1; +#endif +} +static MA_INLINE ma_uint16 ma_dr_flac__swap_endian_uint16(ma_uint16 n) +{ +#ifdef MA_DR_FLAC_HAS_BYTESWAP16_INTRINSIC + #if defined(_MSC_VER) && !defined(__clang__) + return _byteswap_ushort(n); + #elif defined(__GNUC__) || defined(__clang__) + return __builtin_bswap16(n); + #elif defined(__WATCOMC__) && defined(__386__) + return _watcom_bswap16(n); + #else + #error "This compiler does not support the byte swap intrinsic." + #endif +#else + return ((n & 0xFF00) >> 8) | + ((n & 0x00FF) << 8); +#endif +} +static MA_INLINE ma_uint32 ma_dr_flac__swap_endian_uint32(ma_uint32 n) +{ +#ifdef MA_DR_FLAC_HAS_BYTESWAP32_INTRINSIC + #if defined(_MSC_VER) && !defined(__clang__) + return _byteswap_ulong(n); + #elif defined(__GNUC__) || defined(__clang__) + #if defined(MA_ARM) && (defined(__ARM_ARCH) && __ARM_ARCH >= 6) && !defined(__ARM_ARCH_6M__) && !defined(MA_64BIT) + ma_uint32 r; + __asm__ __volatile__ ( + #if defined(MA_64BIT) + "rev %w[out], %w[in]" : [out]"=r"(r) : [in]"r"(n) + #else + "rev %[out], %[in]" : [out]"=r"(r) : [in]"r"(n) + #endif + ); + return r; + #else + return __builtin_bswap32(n); + #endif + #elif defined(__WATCOMC__) && defined(__386__) + return _watcom_bswap32(n); + #else + #error "This compiler does not support the byte swap intrinsic." + #endif +#else + return ((n & 0xFF000000) >> 24) | + ((n & 0x00FF0000) >> 8) | + ((n & 0x0000FF00) << 8) | + ((n & 0x000000FF) << 24); +#endif +} +static MA_INLINE ma_uint64 ma_dr_flac__swap_endian_uint64(ma_uint64 n) +{ +#ifdef MA_DR_FLAC_HAS_BYTESWAP64_INTRINSIC + #if defined(_MSC_VER) && !defined(__clang__) + return _byteswap_uint64(n); + #elif defined(__GNUC__) || defined(__clang__) + return __builtin_bswap64(n); + #elif defined(__WATCOMC__) && defined(__386__) + return _watcom_bswap64(n); + #else + #error "This compiler does not support the byte swap intrinsic." + #endif +#else + return ((n & ((ma_uint64)0xFF000000 << 32)) >> 56) | + ((n & ((ma_uint64)0x00FF0000 << 32)) >> 40) | + ((n & ((ma_uint64)0x0000FF00 << 32)) >> 24) | + ((n & ((ma_uint64)0x000000FF << 32)) >> 8) | + ((n & ((ma_uint64)0xFF000000 )) << 8) | + ((n & ((ma_uint64)0x00FF0000 )) << 24) | + ((n & ((ma_uint64)0x0000FF00 )) << 40) | + ((n & ((ma_uint64)0x000000FF )) << 56); +#endif +} +static MA_INLINE ma_uint16 ma_dr_flac__be2host_16(ma_uint16 n) +{ + if (ma_dr_flac__is_little_endian()) { + return ma_dr_flac__swap_endian_uint16(n); + } + return n; +} +static MA_INLINE ma_uint32 ma_dr_flac__be2host_32(ma_uint32 n) +{ + if (ma_dr_flac__is_little_endian()) { + return ma_dr_flac__swap_endian_uint32(n); + } + return n; +} +static MA_INLINE ma_uint32 ma_dr_flac__be2host_32_ptr_unaligned(const void* pData) +{ + const ma_uint8* pNum = (ma_uint8*)pData; + return *(pNum) << 24 | *(pNum+1) << 16 | *(pNum+2) << 8 | *(pNum+3); +} +static MA_INLINE ma_uint64 ma_dr_flac__be2host_64(ma_uint64 n) +{ + if (ma_dr_flac__is_little_endian()) { + return ma_dr_flac__swap_endian_uint64(n); + } + return n; +} +static MA_INLINE ma_uint32 ma_dr_flac__le2host_32(ma_uint32 n) +{ + if (!ma_dr_flac__is_little_endian()) { + return ma_dr_flac__swap_endian_uint32(n); + } + return n; +} +static MA_INLINE ma_uint32 ma_dr_flac__le2host_32_ptr_unaligned(const void* pData) +{ + const ma_uint8* pNum = (ma_uint8*)pData; + return *pNum | *(pNum+1) << 8 | *(pNum+2) << 16 | *(pNum+3) << 24; +} +static MA_INLINE ma_uint32 ma_dr_flac__unsynchsafe_32(ma_uint32 n) +{ + ma_uint32 result = 0; + result |= (n & 0x7F000000) >> 3; + result |= (n & 0x007F0000) >> 2; + result |= (n & 0x00007F00) >> 1; + result |= (n & 0x0000007F) >> 0; + return result; +} +static ma_uint8 ma_dr_flac__crc8_table[] = { + 0x00, 0x07, 0x0E, 0x09, 0x1C, 0x1B, 0x12, 0x15, 0x38, 0x3F, 0x36, 0x31, 0x24, 0x23, 0x2A, 0x2D, + 0x70, 0x77, 0x7E, 0x79, 0x6C, 0x6B, 0x62, 0x65, 0x48, 0x4F, 0x46, 0x41, 0x54, 0x53, 0x5A, 0x5D, + 0xE0, 0xE7, 0xEE, 0xE9, 0xFC, 0xFB, 0xF2, 0xF5, 0xD8, 0xDF, 0xD6, 0xD1, 0xC4, 0xC3, 0xCA, 0xCD, + 0x90, 0x97, 0x9E, 0x99, 0x8C, 0x8B, 0x82, 0x85, 0xA8, 0xAF, 0xA6, 0xA1, 0xB4, 0xB3, 0xBA, 0xBD, + 0xC7, 0xC0, 0xC9, 0xCE, 0xDB, 0xDC, 0xD5, 0xD2, 0xFF, 0xF8, 0xF1, 0xF6, 0xE3, 0xE4, 0xED, 0xEA, + 0xB7, 0xB0, 0xB9, 0xBE, 0xAB, 0xAC, 0xA5, 0xA2, 0x8F, 0x88, 0x81, 0x86, 0x93, 0x94, 0x9D, 0x9A, + 0x27, 0x20, 0x29, 0x2E, 0x3B, 0x3C, 0x35, 0x32, 0x1F, 0x18, 0x11, 0x16, 0x03, 0x04, 0x0D, 0x0A, + 0x57, 0x50, 0x59, 0x5E, 0x4B, 0x4C, 0x45, 0x42, 0x6F, 0x68, 0x61, 0x66, 0x73, 0x74, 0x7D, 0x7A, + 0x89, 0x8E, 0x87, 0x80, 0x95, 0x92, 0x9B, 0x9C, 0xB1, 0xB6, 0xBF, 0xB8, 0xAD, 0xAA, 0xA3, 0xA4, + 0xF9, 0xFE, 0xF7, 0xF0, 0xE5, 0xE2, 0xEB, 0xEC, 0xC1, 0xC6, 0xCF, 0xC8, 0xDD, 0xDA, 0xD3, 0xD4, + 0x69, 0x6E, 0x67, 0x60, 0x75, 0x72, 0x7B, 0x7C, 0x51, 0x56, 0x5F, 0x58, 0x4D, 0x4A, 0x43, 0x44, + 0x19, 0x1E, 0x17, 0x10, 0x05, 0x02, 0x0B, 0x0C, 0x21, 0x26, 0x2F, 0x28, 0x3D, 0x3A, 0x33, 0x34, + 0x4E, 0x49, 0x40, 0x47, 0x52, 0x55, 0x5C, 0x5B, 0x76, 0x71, 0x78, 0x7F, 0x6A, 0x6D, 0x64, 0x63, + 0x3E, 0x39, 0x30, 0x37, 0x22, 0x25, 0x2C, 0x2B, 0x06, 0x01, 0x08, 0x0F, 0x1A, 0x1D, 0x14, 0x13, + 0xAE, 0xA9, 0xA0, 0xA7, 0xB2, 0xB5, 0xBC, 0xBB, 0x96, 0x91, 0x98, 0x9F, 0x8A, 0x8D, 0x84, 0x83, + 0xDE, 0xD9, 0xD0, 0xD7, 0xC2, 0xC5, 0xCC, 0xCB, 0xE6, 0xE1, 0xE8, 0xEF, 0xFA, 0xFD, 0xF4, 0xF3 +}; +static ma_uint16 ma_dr_flac__crc16_table[] = { + 0x0000, 0x8005, 0x800F, 0x000A, 0x801B, 0x001E, 0x0014, 0x8011, + 0x8033, 0x0036, 0x003C, 0x8039, 0x0028, 0x802D, 0x8027, 0x0022, + 0x8063, 0x0066, 0x006C, 0x8069, 0x0078, 0x807D, 0x8077, 0x0072, + 0x0050, 0x8055, 0x805F, 0x005A, 0x804B, 0x004E, 0x0044, 0x8041, + 0x80C3, 0x00C6, 0x00CC, 0x80C9, 0x00D8, 0x80DD, 0x80D7, 0x00D2, + 0x00F0, 0x80F5, 0x80FF, 0x00FA, 0x80EB, 0x00EE, 0x00E4, 0x80E1, + 0x00A0, 0x80A5, 0x80AF, 0x00AA, 0x80BB, 0x00BE, 0x00B4, 0x80B1, + 0x8093, 0x0096, 0x009C, 0x8099, 0x0088, 0x808D, 0x8087, 0x0082, + 0x8183, 0x0186, 0x018C, 0x8189, 0x0198, 0x819D, 0x8197, 0x0192, + 0x01B0, 0x81B5, 0x81BF, 0x01BA, 0x81AB, 0x01AE, 0x01A4, 0x81A1, + 0x01E0, 0x81E5, 0x81EF, 0x01EA, 0x81FB, 0x01FE, 0x01F4, 0x81F1, + 0x81D3, 0x01D6, 0x01DC, 0x81D9, 0x01C8, 0x81CD, 0x81C7, 0x01C2, + 0x0140, 0x8145, 0x814F, 0x014A, 0x815B, 0x015E, 0x0154, 0x8151, + 0x8173, 0x0176, 0x017C, 0x8179, 0x0168, 0x816D, 0x8167, 0x0162, + 0x8123, 0x0126, 0x012C, 0x8129, 0x0138, 0x813D, 0x8137, 0x0132, + 0x0110, 0x8115, 0x811F, 0x011A, 0x810B, 0x010E, 0x0104, 0x8101, + 0x8303, 0x0306, 0x030C, 0x8309, 0x0318, 0x831D, 0x8317, 0x0312, + 0x0330, 0x8335, 0x833F, 0x033A, 0x832B, 0x032E, 0x0324, 0x8321, + 0x0360, 0x8365, 0x836F, 0x036A, 0x837B, 0x037E, 0x0374, 0x8371, + 0x8353, 0x0356, 0x035C, 0x8359, 0x0348, 0x834D, 0x8347, 0x0342, + 0x03C0, 0x83C5, 0x83CF, 0x03CA, 0x83DB, 0x03DE, 0x03D4, 0x83D1, + 0x83F3, 0x03F6, 0x03FC, 0x83F9, 0x03E8, 0x83ED, 0x83E7, 0x03E2, + 0x83A3, 0x03A6, 0x03AC, 0x83A9, 0x03B8, 0x83BD, 0x83B7, 0x03B2, + 0x0390, 0x8395, 0x839F, 0x039A, 0x838B, 0x038E, 0x0384, 0x8381, + 0x0280, 0x8285, 0x828F, 0x028A, 0x829B, 0x029E, 0x0294, 0x8291, + 0x82B3, 0x02B6, 0x02BC, 0x82B9, 0x02A8, 0x82AD, 0x82A7, 0x02A2, + 0x82E3, 0x02E6, 0x02EC, 0x82E9, 0x02F8, 0x82FD, 0x82F7, 0x02F2, + 0x02D0, 0x82D5, 0x82DF, 0x02DA, 0x82CB, 0x02CE, 0x02C4, 0x82C1, + 0x8243, 0x0246, 0x024C, 0x8249, 0x0258, 0x825D, 0x8257, 0x0252, + 0x0270, 0x8275, 0x827F, 0x027A, 0x826B, 0x026E, 0x0264, 0x8261, + 0x0220, 0x8225, 0x822F, 0x022A, 0x823B, 0x023E, 0x0234, 0x8231, + 0x8213, 0x0216, 0x021C, 0x8219, 0x0208, 0x820D, 0x8207, 0x0202 +}; +static MA_INLINE ma_uint8 ma_dr_flac_crc8_byte(ma_uint8 crc, ma_uint8 data) +{ + return ma_dr_flac__crc8_table[crc ^ data]; +} +static MA_INLINE ma_uint8 ma_dr_flac_crc8(ma_uint8 crc, ma_uint32 data, ma_uint32 count) +{ +#ifdef MA_DR_FLAC_NO_CRC + (void)crc; + (void)data; + (void)count; + return 0; +#else +#if 0 + ma_uint8 p = 0x07; + for (int i = count-1; i >= 0; --i) { + ma_uint8 bit = (data & (1 << i)) >> i; + if (crc & 0x80) { + crc = ((crc << 1) | bit) ^ p; + } else { + crc = ((crc << 1) | bit); + } + } + return crc; +#else + ma_uint32 wholeBytes; + ma_uint32 leftoverBits; + ma_uint64 leftoverDataMask; + static ma_uint64 leftoverDataMaskTable[8] = { + 0x00, 0x01, 0x03, 0x07, 0x0F, 0x1F, 0x3F, 0x7F + }; + MA_DR_FLAC_ASSERT(count <= 32); + wholeBytes = count >> 3; + leftoverBits = count - (wholeBytes*8); + leftoverDataMask = leftoverDataMaskTable[leftoverBits]; + switch (wholeBytes) { + case 4: crc = ma_dr_flac_crc8_byte(crc, (ma_uint8)((data & (0xFF000000UL << leftoverBits)) >> (24 + leftoverBits))); + case 3: crc = ma_dr_flac_crc8_byte(crc, (ma_uint8)((data & (0x00FF0000UL << leftoverBits)) >> (16 + leftoverBits))); + case 2: crc = ma_dr_flac_crc8_byte(crc, (ma_uint8)((data & (0x0000FF00UL << leftoverBits)) >> ( 8 + leftoverBits))); + case 1: crc = ma_dr_flac_crc8_byte(crc, (ma_uint8)((data & (0x000000FFUL << leftoverBits)) >> ( 0 + leftoverBits))); + case 0: if (leftoverBits > 0) crc = (ma_uint8)((crc << leftoverBits) ^ ma_dr_flac__crc8_table[(crc >> (8 - leftoverBits)) ^ (data & leftoverDataMask)]); + } + return crc; +#endif +#endif +} +static MA_INLINE ma_uint16 ma_dr_flac_crc16_byte(ma_uint16 crc, ma_uint8 data) +{ + return (crc << 8) ^ ma_dr_flac__crc16_table[(ma_uint8)(crc >> 8) ^ data]; +} +static MA_INLINE ma_uint16 ma_dr_flac_crc16_cache(ma_uint16 crc, ma_dr_flac_cache_t data) +{ +#ifdef MA_64BIT + crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 56) & 0xFF)); + crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 48) & 0xFF)); + crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 40) & 0xFF)); + crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 32) & 0xFF)); +#endif + crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 24) & 0xFF)); + crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 16) & 0xFF)); + crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 8) & 0xFF)); + crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 0) & 0xFF)); + return crc; +} +static MA_INLINE ma_uint16 ma_dr_flac_crc16_bytes(ma_uint16 crc, ma_dr_flac_cache_t data, ma_uint32 byteCount) +{ + switch (byteCount) + { +#ifdef MA_64BIT + case 8: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 56) & 0xFF)); + case 7: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 48) & 0xFF)); + case 6: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 40) & 0xFF)); + case 5: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 32) & 0xFF)); +#endif + case 4: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 24) & 0xFF)); + case 3: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 16) & 0xFF)); + case 2: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 8) & 0xFF)); + case 1: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data >> 0) & 0xFF)); + } + return crc; +} +#if 0 +static MA_INLINE ma_uint16 ma_dr_flac_crc16__32bit(ma_uint16 crc, ma_uint32 data, ma_uint32 count) +{ +#ifdef MA_DR_FLAC_NO_CRC + (void)crc; + (void)data; + (void)count; + return 0; +#else +#if 0 + ma_uint16 p = 0x8005; + for (int i = count-1; i >= 0; --i) { + ma_uint16 bit = (data & (1ULL << i)) >> i; + if (r & 0x8000) { + r = ((r << 1) | bit) ^ p; + } else { + r = ((r << 1) | bit); + } + } + return crc; +#else + ma_uint32 wholeBytes; + ma_uint32 leftoverBits; + ma_uint64 leftoverDataMask; + static ma_uint64 leftoverDataMaskTable[8] = { + 0x00, 0x01, 0x03, 0x07, 0x0F, 0x1F, 0x3F, 0x7F + }; + MA_DR_FLAC_ASSERT(count <= 64); + wholeBytes = count >> 3; + leftoverBits = count & 7; + leftoverDataMask = leftoverDataMaskTable[leftoverBits]; + switch (wholeBytes) { + default: + case 4: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (0xFF000000UL << leftoverBits)) >> (24 + leftoverBits))); + case 3: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (0x00FF0000UL << leftoverBits)) >> (16 + leftoverBits))); + case 2: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (0x0000FF00UL << leftoverBits)) >> ( 8 + leftoverBits))); + case 1: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (0x000000FFUL << leftoverBits)) >> ( 0 + leftoverBits))); + case 0: if (leftoverBits > 0) crc = (crc << leftoverBits) ^ ma_dr_flac__crc16_table[(crc >> (16 - leftoverBits)) ^ (data & leftoverDataMask)]; + } + return crc; +#endif +#endif +} +static MA_INLINE ma_uint16 ma_dr_flac_crc16__64bit(ma_uint16 crc, ma_uint64 data, ma_uint32 count) +{ +#ifdef MA_DR_FLAC_NO_CRC + (void)crc; + (void)data; + (void)count; + return 0; +#else + ma_uint32 wholeBytes; + ma_uint32 leftoverBits; + ma_uint64 leftoverDataMask; + static ma_uint64 leftoverDataMaskTable[8] = { + 0x00, 0x01, 0x03, 0x07, 0x0F, 0x1F, 0x3F, 0x7F + }; + MA_DR_FLAC_ASSERT(count <= 64); + wholeBytes = count >> 3; + leftoverBits = count & 7; + leftoverDataMask = leftoverDataMaskTable[leftoverBits]; + switch (wholeBytes) { + default: + case 8: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (((ma_uint64)0xFF000000 << 32) << leftoverBits)) >> (56 + leftoverBits))); + case 7: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (((ma_uint64)0x00FF0000 << 32) << leftoverBits)) >> (48 + leftoverBits))); + case 6: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (((ma_uint64)0x0000FF00 << 32) << leftoverBits)) >> (40 + leftoverBits))); + case 5: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (((ma_uint64)0x000000FF << 32) << leftoverBits)) >> (32 + leftoverBits))); + case 4: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (((ma_uint64)0xFF000000 ) << leftoverBits)) >> (24 + leftoverBits))); + case 3: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (((ma_uint64)0x00FF0000 ) << leftoverBits)) >> (16 + leftoverBits))); + case 2: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (((ma_uint64)0x0000FF00 ) << leftoverBits)) >> ( 8 + leftoverBits))); + case 1: crc = ma_dr_flac_crc16_byte(crc, (ma_uint8)((data & (((ma_uint64)0x000000FF ) << leftoverBits)) >> ( 0 + leftoverBits))); + case 0: if (leftoverBits > 0) crc = (crc << leftoverBits) ^ ma_dr_flac__crc16_table[(crc >> (16 - leftoverBits)) ^ (data & leftoverDataMask)]; + } + return crc; +#endif +} +static MA_INLINE ma_uint16 ma_dr_flac_crc16(ma_uint16 crc, ma_dr_flac_cache_t data, ma_uint32 count) +{ +#ifdef MA_64BIT + return ma_dr_flac_crc16__64bit(crc, data, count); +#else + return ma_dr_flac_crc16__32bit(crc, data, count); +#endif +} +#endif +#ifdef MA_64BIT +#define ma_dr_flac__be2host__cache_line ma_dr_flac__be2host_64 +#else +#define ma_dr_flac__be2host__cache_line ma_dr_flac__be2host_32 +#endif +#define MA_DR_FLAC_CACHE_L1_SIZE_BYTES(bs) (sizeof((bs)->cache)) +#define MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs) (sizeof((bs)->cache)*8) +#define MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs) (MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs) - (bs)->consumedBits) +#define MA_DR_FLAC_CACHE_L1_SELECTION_MASK(_bitCount) (~((~(ma_dr_flac_cache_t)0) >> (_bitCount))) +#define MA_DR_FLAC_CACHE_L1_SELECTION_SHIFT(bs, _bitCount) (MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs) - (_bitCount)) +#define MA_DR_FLAC_CACHE_L1_SELECT(bs, _bitCount) (((bs)->cache) & MA_DR_FLAC_CACHE_L1_SELECTION_MASK(_bitCount)) +#define MA_DR_FLAC_CACHE_L1_SELECT_AND_SHIFT(bs, _bitCount) (MA_DR_FLAC_CACHE_L1_SELECT((bs), (_bitCount)) >> MA_DR_FLAC_CACHE_L1_SELECTION_SHIFT((bs), (_bitCount))) +#define MA_DR_FLAC_CACHE_L1_SELECT_AND_SHIFT_SAFE(bs, _bitCount)(MA_DR_FLAC_CACHE_L1_SELECT((bs), (_bitCount)) >> (MA_DR_FLAC_CACHE_L1_SELECTION_SHIFT((bs), (_bitCount)) & (MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs)-1))) +#define MA_DR_FLAC_CACHE_L2_SIZE_BYTES(bs) (sizeof((bs)->cacheL2)) +#define MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs) (MA_DR_FLAC_CACHE_L2_SIZE_BYTES(bs) / sizeof((bs)->cacheL2[0])) +#define MA_DR_FLAC_CACHE_L2_LINES_REMAINING(bs) (MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs) - (bs)->nextL2Line) +#ifndef MA_DR_FLAC_NO_CRC +static MA_INLINE void ma_dr_flac__reset_crc16(ma_dr_flac_bs* bs) +{ + bs->crc16 = 0; + bs->crc16CacheIgnoredBytes = bs->consumedBits >> 3; +} +static MA_INLINE void ma_dr_flac__update_crc16(ma_dr_flac_bs* bs) +{ + if (bs->crc16CacheIgnoredBytes == 0) { + bs->crc16 = ma_dr_flac_crc16_cache(bs->crc16, bs->crc16Cache); + } else { + bs->crc16 = ma_dr_flac_crc16_bytes(bs->crc16, bs->crc16Cache, MA_DR_FLAC_CACHE_L1_SIZE_BYTES(bs) - bs->crc16CacheIgnoredBytes); + bs->crc16CacheIgnoredBytes = 0; + } +} +static MA_INLINE ma_uint16 ma_dr_flac__flush_crc16(ma_dr_flac_bs* bs) +{ + MA_DR_FLAC_ASSERT((MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs) & 7) == 0); + if (MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs) == 0) { + ma_dr_flac__update_crc16(bs); + } else { + bs->crc16 = ma_dr_flac_crc16_bytes(bs->crc16, bs->crc16Cache >> MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs), (bs->consumedBits >> 3) - bs->crc16CacheIgnoredBytes); + bs->crc16CacheIgnoredBytes = bs->consumedBits >> 3; + } + return bs->crc16; +} +#endif +static MA_INLINE ma_bool32 ma_dr_flac__reload_l1_cache_from_l2(ma_dr_flac_bs* bs) +{ + size_t bytesRead; + size_t alignedL1LineCount; + if (bs->nextL2Line < MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs)) { + bs->cache = bs->cacheL2[bs->nextL2Line++]; + return MA_TRUE; + } + if (bs->unalignedByteCount > 0) { + return MA_FALSE; + } + bytesRead = bs->onRead(bs->pUserData, bs->cacheL2, MA_DR_FLAC_CACHE_L2_SIZE_BYTES(bs)); + bs->nextL2Line = 0; + if (bytesRead == MA_DR_FLAC_CACHE_L2_SIZE_BYTES(bs)) { + bs->cache = bs->cacheL2[bs->nextL2Line++]; + return MA_TRUE; + } + alignedL1LineCount = bytesRead / MA_DR_FLAC_CACHE_L1_SIZE_BYTES(bs); + bs->unalignedByteCount = bytesRead - (alignedL1LineCount * MA_DR_FLAC_CACHE_L1_SIZE_BYTES(bs)); + if (bs->unalignedByteCount > 0) { + bs->unalignedCache = bs->cacheL2[alignedL1LineCount]; + } + if (alignedL1LineCount > 0) { + size_t offset = MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs) - alignedL1LineCount; + size_t i; + for (i = alignedL1LineCount; i > 0; --i) { + bs->cacheL2[i-1 + offset] = bs->cacheL2[i-1]; + } + bs->nextL2Line = (ma_uint32)offset; + bs->cache = bs->cacheL2[bs->nextL2Line++]; + return MA_TRUE; + } else { + bs->nextL2Line = MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs); + return MA_FALSE; + } +} +static ma_bool32 ma_dr_flac__reload_cache(ma_dr_flac_bs* bs) +{ + size_t bytesRead; +#ifndef MA_DR_FLAC_NO_CRC + ma_dr_flac__update_crc16(bs); +#endif + if (ma_dr_flac__reload_l1_cache_from_l2(bs)) { + bs->cache = ma_dr_flac__be2host__cache_line(bs->cache); + bs->consumedBits = 0; +#ifndef MA_DR_FLAC_NO_CRC + bs->crc16Cache = bs->cache; +#endif + return MA_TRUE; + } + bytesRead = bs->unalignedByteCount; + if (bytesRead == 0) { + bs->consumedBits = MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs); + return MA_FALSE; + } + MA_DR_FLAC_ASSERT(bytesRead < MA_DR_FLAC_CACHE_L1_SIZE_BYTES(bs)); + bs->consumedBits = (ma_uint32)(MA_DR_FLAC_CACHE_L1_SIZE_BYTES(bs) - bytesRead) * 8; + bs->cache = ma_dr_flac__be2host__cache_line(bs->unalignedCache); + bs->cache &= MA_DR_FLAC_CACHE_L1_SELECTION_MASK(MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)); + bs->unalignedByteCount = 0; +#ifndef MA_DR_FLAC_NO_CRC + bs->crc16Cache = bs->cache >> bs->consumedBits; + bs->crc16CacheIgnoredBytes = bs->consumedBits >> 3; +#endif + return MA_TRUE; +} +static void ma_dr_flac__reset_cache(ma_dr_flac_bs* bs) +{ + bs->nextL2Line = MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs); + bs->consumedBits = MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs); + bs->cache = 0; + bs->unalignedByteCount = 0; + bs->unalignedCache = 0; +#ifndef MA_DR_FLAC_NO_CRC + bs->crc16Cache = 0; + bs->crc16CacheIgnoredBytes = 0; +#endif +} +static MA_INLINE ma_bool32 ma_dr_flac__read_uint32(ma_dr_flac_bs* bs, unsigned int bitCount, ma_uint32* pResultOut) +{ + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pResultOut != NULL); + MA_DR_FLAC_ASSERT(bitCount > 0); + MA_DR_FLAC_ASSERT(bitCount <= 32); + if (bs->consumedBits == MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs)) { + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + } + if (bitCount <= MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)) { +#ifdef MA_64BIT + *pResultOut = (ma_uint32)MA_DR_FLAC_CACHE_L1_SELECT_AND_SHIFT(bs, bitCount); + bs->consumedBits += bitCount; + bs->cache <<= bitCount; +#else + if (bitCount < MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs)) { + *pResultOut = (ma_uint32)MA_DR_FLAC_CACHE_L1_SELECT_AND_SHIFT(bs, bitCount); + bs->consumedBits += bitCount; + bs->cache <<= bitCount; + } else { + *pResultOut = (ma_uint32)bs->cache; + bs->consumedBits = MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs); + bs->cache = 0; + } +#endif + return MA_TRUE; + } else { + ma_uint32 bitCountHi = MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs); + ma_uint32 bitCountLo = bitCount - bitCountHi; + ma_uint32 resultHi; + MA_DR_FLAC_ASSERT(bitCountHi > 0); + MA_DR_FLAC_ASSERT(bitCountHi < 32); + resultHi = (ma_uint32)MA_DR_FLAC_CACHE_L1_SELECT_AND_SHIFT(bs, bitCountHi); + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + if (bitCountLo > MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)) { + return MA_FALSE; + } + *pResultOut = (resultHi << bitCountLo) | (ma_uint32)MA_DR_FLAC_CACHE_L1_SELECT_AND_SHIFT(bs, bitCountLo); + bs->consumedBits += bitCountLo; + bs->cache <<= bitCountLo; + return MA_TRUE; + } +} +static ma_bool32 ma_dr_flac__read_int32(ma_dr_flac_bs* bs, unsigned int bitCount, ma_int32* pResult) +{ + ma_uint32 result; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pResult != NULL); + MA_DR_FLAC_ASSERT(bitCount > 0); + MA_DR_FLAC_ASSERT(bitCount <= 32); + if (!ma_dr_flac__read_uint32(bs, bitCount, &result)) { + return MA_FALSE; + } + if (bitCount < 32) { + ma_uint32 signbit; + signbit = ((result >> (bitCount-1)) & 0x01); + result |= (~signbit + 1) << bitCount; + } + *pResult = (ma_int32)result; + return MA_TRUE; +} +#ifdef MA_64BIT +static ma_bool32 ma_dr_flac__read_uint64(ma_dr_flac_bs* bs, unsigned int bitCount, ma_uint64* pResultOut) +{ + ma_uint32 resultHi; + ma_uint32 resultLo; + MA_DR_FLAC_ASSERT(bitCount <= 64); + MA_DR_FLAC_ASSERT(bitCount > 32); + if (!ma_dr_flac__read_uint32(bs, bitCount - 32, &resultHi)) { + return MA_FALSE; + } + if (!ma_dr_flac__read_uint32(bs, 32, &resultLo)) { + return MA_FALSE; + } + *pResultOut = (((ma_uint64)resultHi) << 32) | ((ma_uint64)resultLo); + return MA_TRUE; +} +#endif +#if 0 +static ma_bool32 ma_dr_flac__read_int64(ma_dr_flac_bs* bs, unsigned int bitCount, ma_int64* pResultOut) +{ + ma_uint64 result; + ma_uint64 signbit; + MA_DR_FLAC_ASSERT(bitCount <= 64); + if (!ma_dr_flac__read_uint64(bs, bitCount, &result)) { + return MA_FALSE; + } + signbit = ((result >> (bitCount-1)) & 0x01); + result |= (~signbit + 1) << bitCount; + *pResultOut = (ma_int64)result; + return MA_TRUE; +} +#endif +static ma_bool32 ma_dr_flac__read_uint16(ma_dr_flac_bs* bs, unsigned int bitCount, ma_uint16* pResult) +{ + ma_uint32 result; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pResult != NULL); + MA_DR_FLAC_ASSERT(bitCount > 0); + MA_DR_FLAC_ASSERT(bitCount <= 16); + if (!ma_dr_flac__read_uint32(bs, bitCount, &result)) { + return MA_FALSE; + } + *pResult = (ma_uint16)result; + return MA_TRUE; +} +#if 0 +static ma_bool32 ma_dr_flac__read_int16(ma_dr_flac_bs* bs, unsigned int bitCount, ma_int16* pResult) +{ + ma_int32 result; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pResult != NULL); + MA_DR_FLAC_ASSERT(bitCount > 0); + MA_DR_FLAC_ASSERT(bitCount <= 16); + if (!ma_dr_flac__read_int32(bs, bitCount, &result)) { + return MA_FALSE; + } + *pResult = (ma_int16)result; + return MA_TRUE; +} +#endif +static ma_bool32 ma_dr_flac__read_uint8(ma_dr_flac_bs* bs, unsigned int bitCount, ma_uint8* pResult) +{ + ma_uint32 result; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pResult != NULL); + MA_DR_FLAC_ASSERT(bitCount > 0); + MA_DR_FLAC_ASSERT(bitCount <= 8); + if (!ma_dr_flac__read_uint32(bs, bitCount, &result)) { + return MA_FALSE; + } + *pResult = (ma_uint8)result; + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__read_int8(ma_dr_flac_bs* bs, unsigned int bitCount, ma_int8* pResult) +{ + ma_int32 result; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pResult != NULL); + MA_DR_FLAC_ASSERT(bitCount > 0); + MA_DR_FLAC_ASSERT(bitCount <= 8); + if (!ma_dr_flac__read_int32(bs, bitCount, &result)) { + return MA_FALSE; + } + *pResult = (ma_int8)result; + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__seek_bits(ma_dr_flac_bs* bs, size_t bitsToSeek) +{ + if (bitsToSeek <= MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)) { + bs->consumedBits += (ma_uint32)bitsToSeek; + bs->cache <<= bitsToSeek; + return MA_TRUE; + } else { + bitsToSeek -= MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs); + bs->consumedBits += MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs); + bs->cache = 0; +#ifdef MA_64BIT + while (bitsToSeek >= MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs)) { + ma_uint64 bin; + if (!ma_dr_flac__read_uint64(bs, MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs), &bin)) { + return MA_FALSE; + } + bitsToSeek -= MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs); + } +#else + while (bitsToSeek >= MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs)) { + ma_uint32 bin; + if (!ma_dr_flac__read_uint32(bs, MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs), &bin)) { + return MA_FALSE; + } + bitsToSeek -= MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs); + } +#endif + while (bitsToSeek >= 8) { + ma_uint8 bin; + if (!ma_dr_flac__read_uint8(bs, 8, &bin)) { + return MA_FALSE; + } + bitsToSeek -= 8; + } + if (bitsToSeek > 0) { + ma_uint8 bin; + if (!ma_dr_flac__read_uint8(bs, (ma_uint32)bitsToSeek, &bin)) { + return MA_FALSE; + } + bitsToSeek = 0; + } + MA_DR_FLAC_ASSERT(bitsToSeek == 0); + return MA_TRUE; + } +} +static ma_bool32 ma_dr_flac__find_and_seek_to_next_sync_code(ma_dr_flac_bs* bs) +{ + MA_DR_FLAC_ASSERT(bs != NULL); + if (!ma_dr_flac__seek_bits(bs, MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs) & 7)) { + return MA_FALSE; + } + for (;;) { + ma_uint8 hi; +#ifndef MA_DR_FLAC_NO_CRC + ma_dr_flac__reset_crc16(bs); +#endif + if (!ma_dr_flac__read_uint8(bs, 8, &hi)) { + return MA_FALSE; + } + if (hi == 0xFF) { + ma_uint8 lo; + if (!ma_dr_flac__read_uint8(bs, 6, &lo)) { + return MA_FALSE; + } + if (lo == 0x3E) { + return MA_TRUE; + } else { + if (!ma_dr_flac__seek_bits(bs, MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs) & 7)) { + return MA_FALSE; + } + } + } + } +} +#if defined(MA_DR_FLAC_HAS_LZCNT_INTRINSIC) +#define MA_DR_FLAC_IMPLEMENT_CLZ_LZCNT +#endif +#if defined(_MSC_VER) && _MSC_VER >= 1400 && (defined(MA_X64) || defined(MA_X86)) && !defined(__clang__) +#define MA_DR_FLAC_IMPLEMENT_CLZ_MSVC +#endif +#if defined(__WATCOMC__) && defined(__386__) +#define MA_DR_FLAC_IMPLEMENT_CLZ_WATCOM +#endif +#ifdef __MRC__ +#include +#define MA_DR_FLAC_IMPLEMENT_CLZ_MRC +#endif +static MA_INLINE ma_uint32 ma_dr_flac__clz_software(ma_dr_flac_cache_t x) +{ + ma_uint32 n; + static ma_uint32 clz_table_4[] = { + 0, + 4, + 3, 3, + 2, 2, 2, 2, + 1, 1, 1, 1, 1, 1, 1, 1 + }; + if (x == 0) { + return sizeof(x)*8; + } + n = clz_table_4[x >> (sizeof(x)*8 - 4)]; + if (n == 0) { +#ifdef MA_64BIT + if ((x & ((ma_uint64)0xFFFFFFFF << 32)) == 0) { n = 32; x <<= 32; } + if ((x & ((ma_uint64)0xFFFF0000 << 32)) == 0) { n += 16; x <<= 16; } + if ((x & ((ma_uint64)0xFF000000 << 32)) == 0) { n += 8; x <<= 8; } + if ((x & ((ma_uint64)0xF0000000 << 32)) == 0) { n += 4; x <<= 4; } +#else + if ((x & 0xFFFF0000) == 0) { n = 16; x <<= 16; } + if ((x & 0xFF000000) == 0) { n += 8; x <<= 8; } + if ((x & 0xF0000000) == 0) { n += 4; x <<= 4; } +#endif + n += clz_table_4[x >> (sizeof(x)*8 - 4)]; + } + return n - 1; +} +#ifdef MA_DR_FLAC_IMPLEMENT_CLZ_LZCNT +static MA_INLINE ma_bool32 ma_dr_flac__is_lzcnt_supported(void) +{ +#if defined(MA_DR_FLAC_HAS_LZCNT_INTRINSIC) && defined(MA_ARM) && (defined(__ARM_ARCH) && __ARM_ARCH >= 5) + return MA_TRUE; +#elif defined(__MRC__) + return MA_TRUE; +#else + #ifdef MA_DR_FLAC_HAS_LZCNT_INTRINSIC + return ma_dr_flac__gIsLZCNTSupported; + #else + return MA_FALSE; + #endif +#endif +} +static MA_INLINE ma_uint32 ma_dr_flac__clz_lzcnt(ma_dr_flac_cache_t x) +{ +#if defined(_MSC_VER) + #ifdef MA_64BIT + return (ma_uint32)__lzcnt64(x); + #else + return (ma_uint32)__lzcnt(x); + #endif +#else + #if defined(__GNUC__) || defined(__clang__) + #if defined(MA_X64) + { + ma_uint64 r; + __asm__ __volatile__ ( + "lzcnt{ %1, %0| %0, %1}" : "=r"(r) : "r"(x) : "cc" + ); + return (ma_uint32)r; + } + #elif defined(MA_X86) + { + ma_uint32 r; + __asm__ __volatile__ ( + "lzcnt{l %1, %0| %0, %1}" : "=r"(r) : "r"(x) : "cc" + ); + return r; + } + #elif defined(MA_ARM) && (defined(__ARM_ARCH) && __ARM_ARCH >= 5) && !defined(__ARM_ARCH_6M__) && !defined(MA_64BIT) + { + unsigned int r; + __asm__ __volatile__ ( + #if defined(MA_64BIT) + "clz %w[out], %w[in]" : [out]"=r"(r) : [in]"r"(x) + #else + "clz %[out], %[in]" : [out]"=r"(r) : [in]"r"(x) + #endif + ); + return r; + } + #else + if (x == 0) { + return sizeof(x)*8; + } + #ifdef MA_64BIT + return (ma_uint32)__builtin_clzll((ma_uint64)x); + #else + return (ma_uint32)__builtin_clzl((ma_uint32)x); + #endif + #endif + #else + #error "This compiler does not support the lzcnt intrinsic." + #endif +#endif +} +#endif +#ifdef MA_DR_FLAC_IMPLEMENT_CLZ_MSVC +#include +static MA_INLINE ma_uint32 ma_dr_flac__clz_msvc(ma_dr_flac_cache_t x) +{ + ma_uint32 n; + if (x == 0) { + return sizeof(x)*8; + } +#ifdef MA_64BIT + _BitScanReverse64((unsigned long*)&n, x); +#else + _BitScanReverse((unsigned long*)&n, x); +#endif + return sizeof(x)*8 - n - 1; +} +#endif +#ifdef MA_DR_FLAC_IMPLEMENT_CLZ_WATCOM +static __inline ma_uint32 ma_dr_flac__clz_watcom (ma_uint32); +#ifdef MA_DR_FLAC_IMPLEMENT_CLZ_WATCOM_LZCNT +#pragma aux ma_dr_flac__clz_watcom_lzcnt = \ + "db 0F3h, 0Fh, 0BDh, 0C0h" \ + parm [eax] \ + value [eax] \ + modify nomemory; +#else +#pragma aux ma_dr_flac__clz_watcom = \ + "bsr eax, eax" \ + "xor eax, 31" \ + parm [eax] nomemory \ + value [eax] \ + modify exact [eax] nomemory; +#endif +#endif +static MA_INLINE ma_uint32 ma_dr_flac__clz(ma_dr_flac_cache_t x) +{ +#ifdef MA_DR_FLAC_IMPLEMENT_CLZ_LZCNT + if (ma_dr_flac__is_lzcnt_supported()) { + return ma_dr_flac__clz_lzcnt(x); + } else +#endif + { +#ifdef MA_DR_FLAC_IMPLEMENT_CLZ_MSVC + return ma_dr_flac__clz_msvc(x); +#elif defined(MA_DR_FLAC_IMPLEMENT_CLZ_WATCOM_LZCNT) + return ma_dr_flac__clz_watcom_lzcnt(x); +#elif defined(MA_DR_FLAC_IMPLEMENT_CLZ_WATCOM) + return (x == 0) ? sizeof(x)*8 : ma_dr_flac__clz_watcom(x); +#elif defined(__MRC__) + return __cntlzw(x); +#else + return ma_dr_flac__clz_software(x); +#endif + } +} +static MA_INLINE ma_bool32 ma_dr_flac__seek_past_next_set_bit(ma_dr_flac_bs* bs, unsigned int* pOffsetOut) +{ + ma_uint32 zeroCounter = 0; + ma_uint32 setBitOffsetPlus1; + while (bs->cache == 0) { + zeroCounter += (ma_uint32)MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs); + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + } + if (bs->cache == 1) { + *pOffsetOut = zeroCounter + (ma_uint32)MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs) - 1; + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + return MA_TRUE; + } + setBitOffsetPlus1 = ma_dr_flac__clz(bs->cache); + setBitOffsetPlus1 += 1; + if (setBitOffsetPlus1 > MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)) { + return MA_FALSE; + } + bs->consumedBits += setBitOffsetPlus1; + bs->cache <<= setBitOffsetPlus1; + *pOffsetOut = zeroCounter + setBitOffsetPlus1 - 1; + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__seek_to_byte(ma_dr_flac_bs* bs, ma_uint64 offsetFromStart) +{ + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(offsetFromStart > 0); + if (offsetFromStart > 0x7FFFFFFF) { + ma_uint64 bytesRemaining = offsetFromStart; + if (!bs->onSeek(bs->pUserData, 0x7FFFFFFF, ma_dr_flac_seek_origin_start)) { + return MA_FALSE; + } + bytesRemaining -= 0x7FFFFFFF; + while (bytesRemaining > 0x7FFFFFFF) { + if (!bs->onSeek(bs->pUserData, 0x7FFFFFFF, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + bytesRemaining -= 0x7FFFFFFF; + } + if (bytesRemaining > 0) { + if (!bs->onSeek(bs->pUserData, (int)bytesRemaining, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + } + } else { + if (!bs->onSeek(bs->pUserData, (int)offsetFromStart, ma_dr_flac_seek_origin_start)) { + return MA_FALSE; + } + } + ma_dr_flac__reset_cache(bs); + return MA_TRUE; +} +static ma_result ma_dr_flac__read_utf8_coded_number(ma_dr_flac_bs* bs, ma_uint64* pNumberOut, ma_uint8* pCRCOut) +{ + ma_uint8 crc; + ma_uint64 result; + ma_uint8 utf8[7] = {0}; + int byteCount; + int i; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pNumberOut != NULL); + MA_DR_FLAC_ASSERT(pCRCOut != NULL); + crc = *pCRCOut; + if (!ma_dr_flac__read_uint8(bs, 8, utf8)) { + *pNumberOut = 0; + return MA_AT_END; + } + crc = ma_dr_flac_crc8(crc, utf8[0], 8); + if ((utf8[0] & 0x80) == 0) { + *pNumberOut = utf8[0]; + *pCRCOut = crc; + return MA_SUCCESS; + } + if ((utf8[0] & 0xE0) == 0xC0) { + byteCount = 2; + } else if ((utf8[0] & 0xF0) == 0xE0) { + byteCount = 3; + } else if ((utf8[0] & 0xF8) == 0xF0) { + byteCount = 4; + } else if ((utf8[0] & 0xFC) == 0xF8) { + byteCount = 5; + } else if ((utf8[0] & 0xFE) == 0xFC) { + byteCount = 6; + } else if ((utf8[0] & 0xFF) == 0xFE) { + byteCount = 7; + } else { + *pNumberOut = 0; + return MA_CRC_MISMATCH; + } + MA_DR_FLAC_ASSERT(byteCount > 1); + result = (ma_uint64)(utf8[0] & (0xFF >> (byteCount + 1))); + for (i = 1; i < byteCount; ++i) { + if (!ma_dr_flac__read_uint8(bs, 8, utf8 + i)) { + *pNumberOut = 0; + return MA_AT_END; + } + crc = ma_dr_flac_crc8(crc, utf8[i], 8); + result = (result << 6) | (utf8[i] & 0x3F); + } + *pNumberOut = result; + *pCRCOut = crc; + return MA_SUCCESS; +} +static MA_INLINE ma_uint32 ma_dr_flac__ilog2_u32(ma_uint32 x) +{ +#if 1 + ma_uint32 result = 0; + while (x > 0) { + result += 1; + x >>= 1; + } + return result; +#endif +} +static MA_INLINE ma_bool32 ma_dr_flac__use_64_bit_prediction(ma_uint32 bitsPerSample, ma_uint32 order, ma_uint32 precision) +{ + return bitsPerSample + precision + ma_dr_flac__ilog2_u32(order) > 32; +} +#if defined(__clang__) +__attribute__((no_sanitize("signed-integer-overflow"))) +#endif +static MA_INLINE ma_int32 ma_dr_flac__calculate_prediction_32(ma_uint32 order, ma_int32 shift, const ma_int32* coefficients, ma_int32* pDecodedSamples) +{ + ma_int32 prediction = 0; + MA_DR_FLAC_ASSERT(order <= 32); + switch (order) + { + case 32: prediction += coefficients[31] * pDecodedSamples[-32]; + case 31: prediction += coefficients[30] * pDecodedSamples[-31]; + case 30: prediction += coefficients[29] * pDecodedSamples[-30]; + case 29: prediction += coefficients[28] * pDecodedSamples[-29]; + case 28: prediction += coefficients[27] * pDecodedSamples[-28]; + case 27: prediction += coefficients[26] * pDecodedSamples[-27]; + case 26: prediction += coefficients[25] * pDecodedSamples[-26]; + case 25: prediction += coefficients[24] * pDecodedSamples[-25]; + case 24: prediction += coefficients[23] * pDecodedSamples[-24]; + case 23: prediction += coefficients[22] * pDecodedSamples[-23]; + case 22: prediction += coefficients[21] * pDecodedSamples[-22]; + case 21: prediction += coefficients[20] * pDecodedSamples[-21]; + case 20: prediction += coefficients[19] * pDecodedSamples[-20]; + case 19: prediction += coefficients[18] * pDecodedSamples[-19]; + case 18: prediction += coefficients[17] * pDecodedSamples[-18]; + case 17: prediction += coefficients[16] * pDecodedSamples[-17]; + case 16: prediction += coefficients[15] * pDecodedSamples[-16]; + case 15: prediction += coefficients[14] * pDecodedSamples[-15]; + case 14: prediction += coefficients[13] * pDecodedSamples[-14]; + case 13: prediction += coefficients[12] * pDecodedSamples[-13]; + case 12: prediction += coefficients[11] * pDecodedSamples[-12]; + case 11: prediction += coefficients[10] * pDecodedSamples[-11]; + case 10: prediction += coefficients[ 9] * pDecodedSamples[-10]; + case 9: prediction += coefficients[ 8] * pDecodedSamples[- 9]; + case 8: prediction += coefficients[ 7] * pDecodedSamples[- 8]; + case 7: prediction += coefficients[ 6] * pDecodedSamples[- 7]; + case 6: prediction += coefficients[ 5] * pDecodedSamples[- 6]; + case 5: prediction += coefficients[ 4] * pDecodedSamples[- 5]; + case 4: prediction += coefficients[ 3] * pDecodedSamples[- 4]; + case 3: prediction += coefficients[ 2] * pDecodedSamples[- 3]; + case 2: prediction += coefficients[ 1] * pDecodedSamples[- 2]; + case 1: prediction += coefficients[ 0] * pDecodedSamples[- 1]; + } + return (ma_int32)(prediction >> shift); +} +static MA_INLINE ma_int32 ma_dr_flac__calculate_prediction_64(ma_uint32 order, ma_int32 shift, const ma_int32* coefficients, ma_int32* pDecodedSamples) +{ + ma_int64 prediction; + MA_DR_FLAC_ASSERT(order <= 32); +#ifndef MA_64BIT + if (order == 8) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + prediction += coefficients[4] * (ma_int64)pDecodedSamples[-5]; + prediction += coefficients[5] * (ma_int64)pDecodedSamples[-6]; + prediction += coefficients[6] * (ma_int64)pDecodedSamples[-7]; + prediction += coefficients[7] * (ma_int64)pDecodedSamples[-8]; + } + else if (order == 7) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + prediction += coefficients[4] * (ma_int64)pDecodedSamples[-5]; + prediction += coefficients[5] * (ma_int64)pDecodedSamples[-6]; + prediction += coefficients[6] * (ma_int64)pDecodedSamples[-7]; + } + else if (order == 3) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + } + else if (order == 6) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + prediction += coefficients[4] * (ma_int64)pDecodedSamples[-5]; + prediction += coefficients[5] * (ma_int64)pDecodedSamples[-6]; + } + else if (order == 5) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + prediction += coefficients[4] * (ma_int64)pDecodedSamples[-5]; + } + else if (order == 4) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + } + else if (order == 12) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + prediction += coefficients[4] * (ma_int64)pDecodedSamples[-5]; + prediction += coefficients[5] * (ma_int64)pDecodedSamples[-6]; + prediction += coefficients[6] * (ma_int64)pDecodedSamples[-7]; + prediction += coefficients[7] * (ma_int64)pDecodedSamples[-8]; + prediction += coefficients[8] * (ma_int64)pDecodedSamples[-9]; + prediction += coefficients[9] * (ma_int64)pDecodedSamples[-10]; + prediction += coefficients[10] * (ma_int64)pDecodedSamples[-11]; + prediction += coefficients[11] * (ma_int64)pDecodedSamples[-12]; + } + else if (order == 2) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + } + else if (order == 1) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + } + else if (order == 10) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + prediction += coefficients[4] * (ma_int64)pDecodedSamples[-5]; + prediction += coefficients[5] * (ma_int64)pDecodedSamples[-6]; + prediction += coefficients[6] * (ma_int64)pDecodedSamples[-7]; + prediction += coefficients[7] * (ma_int64)pDecodedSamples[-8]; + prediction += coefficients[8] * (ma_int64)pDecodedSamples[-9]; + prediction += coefficients[9] * (ma_int64)pDecodedSamples[-10]; + } + else if (order == 9) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + prediction += coefficients[4] * (ma_int64)pDecodedSamples[-5]; + prediction += coefficients[5] * (ma_int64)pDecodedSamples[-6]; + prediction += coefficients[6] * (ma_int64)pDecodedSamples[-7]; + prediction += coefficients[7] * (ma_int64)pDecodedSamples[-8]; + prediction += coefficients[8] * (ma_int64)pDecodedSamples[-9]; + } + else if (order == 11) + { + prediction = coefficients[0] * (ma_int64)pDecodedSamples[-1]; + prediction += coefficients[1] * (ma_int64)pDecodedSamples[-2]; + prediction += coefficients[2] * (ma_int64)pDecodedSamples[-3]; + prediction += coefficients[3] * (ma_int64)pDecodedSamples[-4]; + prediction += coefficients[4] * (ma_int64)pDecodedSamples[-5]; + prediction += coefficients[5] * (ma_int64)pDecodedSamples[-6]; + prediction += coefficients[6] * (ma_int64)pDecodedSamples[-7]; + prediction += coefficients[7] * (ma_int64)pDecodedSamples[-8]; + prediction += coefficients[8] * (ma_int64)pDecodedSamples[-9]; + prediction += coefficients[9] * (ma_int64)pDecodedSamples[-10]; + prediction += coefficients[10] * (ma_int64)pDecodedSamples[-11]; + } + else + { + int j; + prediction = 0; + for (j = 0; j < (int)order; ++j) { + prediction += coefficients[j] * (ma_int64)pDecodedSamples[-j-1]; + } + } +#endif +#ifdef MA_64BIT + prediction = 0; + switch (order) + { + case 32: prediction += coefficients[31] * (ma_int64)pDecodedSamples[-32]; + case 31: prediction += coefficients[30] * (ma_int64)pDecodedSamples[-31]; + case 30: prediction += coefficients[29] * (ma_int64)pDecodedSamples[-30]; + case 29: prediction += coefficients[28] * (ma_int64)pDecodedSamples[-29]; + case 28: prediction += coefficients[27] * (ma_int64)pDecodedSamples[-28]; + case 27: prediction += coefficients[26] * (ma_int64)pDecodedSamples[-27]; + case 26: prediction += coefficients[25] * (ma_int64)pDecodedSamples[-26]; + case 25: prediction += coefficients[24] * (ma_int64)pDecodedSamples[-25]; + case 24: prediction += coefficients[23] * (ma_int64)pDecodedSamples[-24]; + case 23: prediction += coefficients[22] * (ma_int64)pDecodedSamples[-23]; + case 22: prediction += coefficients[21] * (ma_int64)pDecodedSamples[-22]; + case 21: prediction += coefficients[20] * (ma_int64)pDecodedSamples[-21]; + case 20: prediction += coefficients[19] * (ma_int64)pDecodedSamples[-20]; + case 19: prediction += coefficients[18] * (ma_int64)pDecodedSamples[-19]; + case 18: prediction += coefficients[17] * (ma_int64)pDecodedSamples[-18]; + case 17: prediction += coefficients[16] * (ma_int64)pDecodedSamples[-17]; + case 16: prediction += coefficients[15] * (ma_int64)pDecodedSamples[-16]; + case 15: prediction += coefficients[14] * (ma_int64)pDecodedSamples[-15]; + case 14: prediction += coefficients[13] * (ma_int64)pDecodedSamples[-14]; + case 13: prediction += coefficients[12] * (ma_int64)pDecodedSamples[-13]; + case 12: prediction += coefficients[11] * (ma_int64)pDecodedSamples[-12]; + case 11: prediction += coefficients[10] * (ma_int64)pDecodedSamples[-11]; + case 10: prediction += coefficients[ 9] * (ma_int64)pDecodedSamples[-10]; + case 9: prediction += coefficients[ 8] * (ma_int64)pDecodedSamples[- 9]; + case 8: prediction += coefficients[ 7] * (ma_int64)pDecodedSamples[- 8]; + case 7: prediction += coefficients[ 6] * (ma_int64)pDecodedSamples[- 7]; + case 6: prediction += coefficients[ 5] * (ma_int64)pDecodedSamples[- 6]; + case 5: prediction += coefficients[ 4] * (ma_int64)pDecodedSamples[- 5]; + case 4: prediction += coefficients[ 3] * (ma_int64)pDecodedSamples[- 4]; + case 3: prediction += coefficients[ 2] * (ma_int64)pDecodedSamples[- 3]; + case 2: prediction += coefficients[ 1] * (ma_int64)pDecodedSamples[- 2]; + case 1: prediction += coefficients[ 0] * (ma_int64)pDecodedSamples[- 1]; + } +#endif + return (ma_int32)(prediction >> shift); +} +#if 0 +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__reference(ma_dr_flac_bs* bs, ma_uint32 bitsPerSample, ma_uint32 count, ma_uint8 riceParam, ma_uint32 lpcOrder, ma_int32 lpcShift, ma_uint32 lpcPrecision, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + ma_uint32 i; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pSamplesOut != NULL); + for (i = 0; i < count; ++i) { + ma_uint32 zeroCounter = 0; + for (;;) { + ma_uint8 bit; + if (!ma_dr_flac__read_uint8(bs, 1, &bit)) { + return MA_FALSE; + } + if (bit == 0) { + zeroCounter += 1; + } else { + break; + } + } + ma_uint32 decodedRice; + if (riceParam > 0) { + if (!ma_dr_flac__read_uint32(bs, riceParam, &decodedRice)) { + return MA_FALSE; + } + } else { + decodedRice = 0; + } + decodedRice |= (zeroCounter << riceParam); + if ((decodedRice & 0x01)) { + decodedRice = ~(decodedRice >> 1); + } else { + decodedRice = (decodedRice >> 1); + } + if (ma_dr_flac__use_64_bit_prediction(bitsPerSample, lpcOrder, lpcPrecision)) { + pSamplesOut[i] = decodedRice + ma_dr_flac__calculate_prediction_64(lpcOrder, lpcShift, coefficients, pSamplesOut + i); + } else { + pSamplesOut[i] = decodedRice + ma_dr_flac__calculate_prediction_32(lpcOrder, lpcShift, coefficients, pSamplesOut + i); + } + } + return MA_TRUE; +} +#endif +#if 0 +static ma_bool32 ma_dr_flac__read_rice_parts__reference(ma_dr_flac_bs* bs, ma_uint8 riceParam, ma_uint32* pZeroCounterOut, ma_uint32* pRiceParamPartOut) +{ + ma_uint32 zeroCounter = 0; + ma_uint32 decodedRice; + for (;;) { + ma_uint8 bit; + if (!ma_dr_flac__read_uint8(bs, 1, &bit)) { + return MA_FALSE; + } + if (bit == 0) { + zeroCounter += 1; + } else { + break; + } + } + if (riceParam > 0) { + if (!ma_dr_flac__read_uint32(bs, riceParam, &decodedRice)) { + return MA_FALSE; + } + } else { + decodedRice = 0; + } + *pZeroCounterOut = zeroCounter; + *pRiceParamPartOut = decodedRice; + return MA_TRUE; +} +#endif +#if 0 +static MA_INLINE ma_bool32 ma_dr_flac__read_rice_parts(ma_dr_flac_bs* bs, ma_uint8 riceParam, ma_uint32* pZeroCounterOut, ma_uint32* pRiceParamPartOut) +{ + ma_dr_flac_cache_t riceParamMask; + ma_uint32 zeroCounter; + ma_uint32 setBitOffsetPlus1; + ma_uint32 riceParamPart; + ma_uint32 riceLength; + MA_DR_FLAC_ASSERT(riceParam > 0); + riceParamMask = MA_DR_FLAC_CACHE_L1_SELECTION_MASK(riceParam); + zeroCounter = 0; + while (bs->cache == 0) { + zeroCounter += (ma_uint32)MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs); + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + } + setBitOffsetPlus1 = ma_dr_flac__clz(bs->cache); + zeroCounter += setBitOffsetPlus1; + setBitOffsetPlus1 += 1; + riceLength = setBitOffsetPlus1 + riceParam; + if (riceLength < MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)) { + riceParamPart = (ma_uint32)((bs->cache & (riceParamMask >> setBitOffsetPlus1)) >> MA_DR_FLAC_CACHE_L1_SELECTION_SHIFT(bs, riceLength)); + bs->consumedBits += riceLength; + bs->cache <<= riceLength; + } else { + ma_uint32 bitCountLo; + ma_dr_flac_cache_t resultHi; + bs->consumedBits += riceLength; + bs->cache <<= setBitOffsetPlus1 & (MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs)-1); + bitCountLo = bs->consumedBits - MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs); + resultHi = MA_DR_FLAC_CACHE_L1_SELECT_AND_SHIFT(bs, riceParam); + if (bs->nextL2Line < MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs)) { +#ifndef MA_DR_FLAC_NO_CRC + ma_dr_flac__update_crc16(bs); +#endif + bs->cache = ma_dr_flac__be2host__cache_line(bs->cacheL2[bs->nextL2Line++]); + bs->consumedBits = 0; +#ifndef MA_DR_FLAC_NO_CRC + bs->crc16Cache = bs->cache; +#endif + } else { + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + if (bitCountLo > MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)) { + return MA_FALSE; + } + } + riceParamPart = (ma_uint32)(resultHi | MA_DR_FLAC_CACHE_L1_SELECT_AND_SHIFT_SAFE(bs, bitCountLo)); + bs->consumedBits += bitCountLo; + bs->cache <<= bitCountLo; + } + pZeroCounterOut[0] = zeroCounter; + pRiceParamPartOut[0] = riceParamPart; + return MA_TRUE; +} +#endif +static MA_INLINE ma_bool32 ma_dr_flac__read_rice_parts_x1(ma_dr_flac_bs* bs, ma_uint8 riceParam, ma_uint32* pZeroCounterOut, ma_uint32* pRiceParamPartOut) +{ + ma_uint32 riceParamPlus1 = riceParam + 1; + ma_uint32 riceParamPlus1Shift = MA_DR_FLAC_CACHE_L1_SELECTION_SHIFT(bs, riceParamPlus1); + ma_uint32 riceParamPlus1MaxConsumedBits = MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs) - riceParamPlus1; + ma_dr_flac_cache_t bs_cache = bs->cache; + ma_uint32 bs_consumedBits = bs->consumedBits; + ma_uint32 lzcount = ma_dr_flac__clz(bs_cache); + if (lzcount < sizeof(bs_cache)*8) { + pZeroCounterOut[0] = lzcount; + extract_rice_param_part: + bs_cache <<= lzcount; + bs_consumedBits += lzcount; + if (bs_consumedBits <= riceParamPlus1MaxConsumedBits) { + pRiceParamPartOut[0] = (ma_uint32)(bs_cache >> riceParamPlus1Shift); + bs_cache <<= riceParamPlus1; + bs_consumedBits += riceParamPlus1; + } else { + ma_uint32 riceParamPartHi; + ma_uint32 riceParamPartLo; + ma_uint32 riceParamPartLoBitCount; + riceParamPartHi = (ma_uint32)(bs_cache >> riceParamPlus1Shift); + riceParamPartLoBitCount = bs_consumedBits - riceParamPlus1MaxConsumedBits; + MA_DR_FLAC_ASSERT(riceParamPartLoBitCount > 0 && riceParamPartLoBitCount < 32); + if (bs->nextL2Line < MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs)) { + #ifndef MA_DR_FLAC_NO_CRC + ma_dr_flac__update_crc16(bs); + #endif + bs_cache = ma_dr_flac__be2host__cache_line(bs->cacheL2[bs->nextL2Line++]); + bs_consumedBits = riceParamPartLoBitCount; + #ifndef MA_DR_FLAC_NO_CRC + bs->crc16Cache = bs_cache; + #endif + } else { + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + if (riceParamPartLoBitCount > MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)) { + return MA_FALSE; + } + bs_cache = bs->cache; + bs_consumedBits = bs->consumedBits + riceParamPartLoBitCount; + } + riceParamPartLo = (ma_uint32)(bs_cache >> (MA_DR_FLAC_CACHE_L1_SELECTION_SHIFT(bs, riceParamPartLoBitCount))); + pRiceParamPartOut[0] = riceParamPartHi | riceParamPartLo; + bs_cache <<= riceParamPartLoBitCount; + } + } else { + ma_uint32 zeroCounter = (ma_uint32)(MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs) - bs_consumedBits); + for (;;) { + if (bs->nextL2Line < MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs)) { + #ifndef MA_DR_FLAC_NO_CRC + ma_dr_flac__update_crc16(bs); + #endif + bs_cache = ma_dr_flac__be2host__cache_line(bs->cacheL2[bs->nextL2Line++]); + bs_consumedBits = 0; + #ifndef MA_DR_FLAC_NO_CRC + bs->crc16Cache = bs_cache; + #endif + } else { + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + bs_cache = bs->cache; + bs_consumedBits = bs->consumedBits; + } + lzcount = ma_dr_flac__clz(bs_cache); + zeroCounter += lzcount; + if (lzcount < sizeof(bs_cache)*8) { + break; + } + } + pZeroCounterOut[0] = zeroCounter; + goto extract_rice_param_part; + } + bs->cache = bs_cache; + bs->consumedBits = bs_consumedBits; + return MA_TRUE; +} +static MA_INLINE ma_bool32 ma_dr_flac__seek_rice_parts(ma_dr_flac_bs* bs, ma_uint8 riceParam) +{ + ma_uint32 riceParamPlus1 = riceParam + 1; + ma_uint32 riceParamPlus1MaxConsumedBits = MA_DR_FLAC_CACHE_L1_SIZE_BITS(bs) - riceParamPlus1; + ma_dr_flac_cache_t bs_cache = bs->cache; + ma_uint32 bs_consumedBits = bs->consumedBits; + ma_uint32 lzcount = ma_dr_flac__clz(bs_cache); + if (lzcount < sizeof(bs_cache)*8) { + extract_rice_param_part: + bs_cache <<= lzcount; + bs_consumedBits += lzcount; + if (bs_consumedBits <= riceParamPlus1MaxConsumedBits) { + bs_cache <<= riceParamPlus1; + bs_consumedBits += riceParamPlus1; + } else { + ma_uint32 riceParamPartLoBitCount = bs_consumedBits - riceParamPlus1MaxConsumedBits; + MA_DR_FLAC_ASSERT(riceParamPartLoBitCount > 0 && riceParamPartLoBitCount < 32); + if (bs->nextL2Line < MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs)) { + #ifndef MA_DR_FLAC_NO_CRC + ma_dr_flac__update_crc16(bs); + #endif + bs_cache = ma_dr_flac__be2host__cache_line(bs->cacheL2[bs->nextL2Line++]); + bs_consumedBits = riceParamPartLoBitCount; + #ifndef MA_DR_FLAC_NO_CRC + bs->crc16Cache = bs_cache; + #endif + } else { + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + if (riceParamPartLoBitCount > MA_DR_FLAC_CACHE_L1_BITS_REMAINING(bs)) { + return MA_FALSE; + } + bs_cache = bs->cache; + bs_consumedBits = bs->consumedBits + riceParamPartLoBitCount; + } + bs_cache <<= riceParamPartLoBitCount; + } + } else { + for (;;) { + if (bs->nextL2Line < MA_DR_FLAC_CACHE_L2_LINE_COUNT(bs)) { + #ifndef MA_DR_FLAC_NO_CRC + ma_dr_flac__update_crc16(bs); + #endif + bs_cache = ma_dr_flac__be2host__cache_line(bs->cacheL2[bs->nextL2Line++]); + bs_consumedBits = 0; + #ifndef MA_DR_FLAC_NO_CRC + bs->crc16Cache = bs_cache; + #endif + } else { + if (!ma_dr_flac__reload_cache(bs)) { + return MA_FALSE; + } + bs_cache = bs->cache; + bs_consumedBits = bs->consumedBits; + } + lzcount = ma_dr_flac__clz(bs_cache); + if (lzcount < sizeof(bs_cache)*8) { + break; + } + } + goto extract_rice_param_part; + } + bs->cache = bs_cache; + bs->consumedBits = bs_consumedBits; + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__scalar_zeroorder(ma_dr_flac_bs* bs, ma_uint32 bitsPerSample, ma_uint32 count, ma_uint8 riceParam, ma_uint32 order, ma_int32 shift, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + ma_uint32 t[2] = {0x00000000, 0xFFFFFFFF}; + ma_uint32 zeroCountPart0; + ma_uint32 riceParamPart0; + ma_uint32 riceParamMask; + ma_uint32 i; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pSamplesOut != NULL); + (void)bitsPerSample; + (void)order; + (void)shift; + (void)coefficients; + riceParamMask = (ma_uint32)~((~0UL) << riceParam); + i = 0; + while (i < count) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart0, &riceParamPart0)) { + return MA_FALSE; + } + riceParamPart0 &= riceParamMask; + riceParamPart0 |= (zeroCountPart0 << riceParam); + riceParamPart0 = (riceParamPart0 >> 1) ^ t[riceParamPart0 & 0x01]; + pSamplesOut[i] = riceParamPart0; + i += 1; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__scalar(ma_dr_flac_bs* bs, ma_uint32 bitsPerSample, ma_uint32 count, ma_uint8 riceParam, ma_uint32 lpcOrder, ma_int32 lpcShift, ma_uint32 lpcPrecision, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + ma_uint32 t[2] = {0x00000000, 0xFFFFFFFF}; + ma_uint32 zeroCountPart0 = 0; + ma_uint32 zeroCountPart1 = 0; + ma_uint32 zeroCountPart2 = 0; + ma_uint32 zeroCountPart3 = 0; + ma_uint32 riceParamPart0 = 0; + ma_uint32 riceParamPart1 = 0; + ma_uint32 riceParamPart2 = 0; + ma_uint32 riceParamPart3 = 0; + ma_uint32 riceParamMask; + const ma_int32* pSamplesOutEnd; + ma_uint32 i; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pSamplesOut != NULL); + if (lpcOrder == 0) { + return ma_dr_flac__decode_samples_with_residual__rice__scalar_zeroorder(bs, bitsPerSample, count, riceParam, lpcOrder, lpcShift, coefficients, pSamplesOut); + } + riceParamMask = (ma_uint32)~((~0UL) << riceParam); + pSamplesOutEnd = pSamplesOut + (count & ~3); + if (ma_dr_flac__use_64_bit_prediction(bitsPerSample, lpcOrder, lpcPrecision)) { + while (pSamplesOut < pSamplesOutEnd) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart0, &riceParamPart0) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart1, &riceParamPart1) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart2, &riceParamPart2) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart3, &riceParamPart3)) { + return MA_FALSE; + } + riceParamPart0 &= riceParamMask; + riceParamPart1 &= riceParamMask; + riceParamPart2 &= riceParamMask; + riceParamPart3 &= riceParamMask; + riceParamPart0 |= (zeroCountPart0 << riceParam); + riceParamPart1 |= (zeroCountPart1 << riceParam); + riceParamPart2 |= (zeroCountPart2 << riceParam); + riceParamPart3 |= (zeroCountPart3 << riceParam); + riceParamPart0 = (riceParamPart0 >> 1) ^ t[riceParamPart0 & 0x01]; + riceParamPart1 = (riceParamPart1 >> 1) ^ t[riceParamPart1 & 0x01]; + riceParamPart2 = (riceParamPart2 >> 1) ^ t[riceParamPart2 & 0x01]; + riceParamPart3 = (riceParamPart3 >> 1) ^ t[riceParamPart3 & 0x01]; + pSamplesOut[0] = riceParamPart0 + ma_dr_flac__calculate_prediction_64(lpcOrder, lpcShift, coefficients, pSamplesOut + 0); + pSamplesOut[1] = riceParamPart1 + ma_dr_flac__calculate_prediction_64(lpcOrder, lpcShift, coefficients, pSamplesOut + 1); + pSamplesOut[2] = riceParamPart2 + ma_dr_flac__calculate_prediction_64(lpcOrder, lpcShift, coefficients, pSamplesOut + 2); + pSamplesOut[3] = riceParamPart3 + ma_dr_flac__calculate_prediction_64(lpcOrder, lpcShift, coefficients, pSamplesOut + 3); + pSamplesOut += 4; + } + } else { + while (pSamplesOut < pSamplesOutEnd) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart0, &riceParamPart0) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart1, &riceParamPart1) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart2, &riceParamPart2) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart3, &riceParamPart3)) { + return MA_FALSE; + } + riceParamPart0 &= riceParamMask; + riceParamPart1 &= riceParamMask; + riceParamPart2 &= riceParamMask; + riceParamPart3 &= riceParamMask; + riceParamPart0 |= (zeroCountPart0 << riceParam); + riceParamPart1 |= (zeroCountPart1 << riceParam); + riceParamPart2 |= (zeroCountPart2 << riceParam); + riceParamPart3 |= (zeroCountPart3 << riceParam); + riceParamPart0 = (riceParamPart0 >> 1) ^ t[riceParamPart0 & 0x01]; + riceParamPart1 = (riceParamPart1 >> 1) ^ t[riceParamPart1 & 0x01]; + riceParamPart2 = (riceParamPart2 >> 1) ^ t[riceParamPart2 & 0x01]; + riceParamPart3 = (riceParamPart3 >> 1) ^ t[riceParamPart3 & 0x01]; + pSamplesOut[0] = riceParamPart0 + ma_dr_flac__calculate_prediction_32(lpcOrder, lpcShift, coefficients, pSamplesOut + 0); + pSamplesOut[1] = riceParamPart1 + ma_dr_flac__calculate_prediction_32(lpcOrder, lpcShift, coefficients, pSamplesOut + 1); + pSamplesOut[2] = riceParamPart2 + ma_dr_flac__calculate_prediction_32(lpcOrder, lpcShift, coefficients, pSamplesOut + 2); + pSamplesOut[3] = riceParamPart3 + ma_dr_flac__calculate_prediction_32(lpcOrder, lpcShift, coefficients, pSamplesOut + 3); + pSamplesOut += 4; + } + } + i = (count & ~3); + while (i < count) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountPart0, &riceParamPart0)) { + return MA_FALSE; + } + riceParamPart0 &= riceParamMask; + riceParamPart0 |= (zeroCountPart0 << riceParam); + riceParamPart0 = (riceParamPart0 >> 1) ^ t[riceParamPart0 & 0x01]; + if (ma_dr_flac__use_64_bit_prediction(bitsPerSample, lpcOrder, lpcPrecision)) { + pSamplesOut[0] = riceParamPart0 + ma_dr_flac__calculate_prediction_64(lpcOrder, lpcShift, coefficients, pSamplesOut + 0); + } else { + pSamplesOut[0] = riceParamPart0 + ma_dr_flac__calculate_prediction_32(lpcOrder, lpcShift, coefficients, pSamplesOut + 0); + } + i += 1; + pSamplesOut += 1; + } + return MA_TRUE; +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE __m128i ma_dr_flac__mm_packs_interleaved_epi32(__m128i a, __m128i b) +{ + __m128i r; + r = _mm_packs_epi32(a, b); + r = _mm_shuffle_epi32(r, _MM_SHUFFLE(3, 1, 2, 0)); + r = _mm_shufflehi_epi16(r, _MM_SHUFFLE(3, 1, 2, 0)); + r = _mm_shufflelo_epi16(r, _MM_SHUFFLE(3, 1, 2, 0)); + return r; +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_SSE41) +static MA_INLINE __m128i ma_dr_flac__mm_not_si128(__m128i a) +{ + return _mm_xor_si128(a, _mm_cmpeq_epi32(_mm_setzero_si128(), _mm_setzero_si128())); +} +static MA_INLINE __m128i ma_dr_flac__mm_hadd_epi32(__m128i x) +{ + __m128i x64 = _mm_add_epi32(x, _mm_shuffle_epi32(x, _MM_SHUFFLE(1, 0, 3, 2))); + __m128i x32 = _mm_shufflelo_epi16(x64, _MM_SHUFFLE(1, 0, 3, 2)); + return _mm_add_epi32(x64, x32); +} +static MA_INLINE __m128i ma_dr_flac__mm_hadd_epi64(__m128i x) +{ + return _mm_add_epi64(x, _mm_shuffle_epi32(x, _MM_SHUFFLE(1, 0, 3, 2))); +} +static MA_INLINE __m128i ma_dr_flac__mm_srai_epi64(__m128i x, int count) +{ + __m128i lo = _mm_srli_epi64(x, count); + __m128i hi = _mm_srai_epi32(x, count); + hi = _mm_and_si128(hi, _mm_set_epi32(0xFFFFFFFF, 0, 0xFFFFFFFF, 0)); + return _mm_or_si128(lo, hi); +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__sse41_32(ma_dr_flac_bs* bs, ma_uint32 count, ma_uint8 riceParam, ma_uint32 order, ma_int32 shift, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + int i; + ma_uint32 riceParamMask; + ma_int32* pDecodedSamples = pSamplesOut; + ma_int32* pDecodedSamplesEnd = pSamplesOut + (count & ~3); + ma_uint32 zeroCountParts0 = 0; + ma_uint32 zeroCountParts1 = 0; + ma_uint32 zeroCountParts2 = 0; + ma_uint32 zeroCountParts3 = 0; + ma_uint32 riceParamParts0 = 0; + ma_uint32 riceParamParts1 = 0; + ma_uint32 riceParamParts2 = 0; + ma_uint32 riceParamParts3 = 0; + __m128i coefficients128_0; + __m128i coefficients128_4; + __m128i coefficients128_8; + __m128i samples128_0; + __m128i samples128_4; + __m128i samples128_8; + __m128i riceParamMask128; + const ma_uint32 t[2] = {0x00000000, 0xFFFFFFFF}; + riceParamMask = (ma_uint32)~((~0UL) << riceParam); + riceParamMask128 = _mm_set1_epi32(riceParamMask); + coefficients128_0 = _mm_setzero_si128(); + coefficients128_4 = _mm_setzero_si128(); + coefficients128_8 = _mm_setzero_si128(); + samples128_0 = _mm_setzero_si128(); + samples128_4 = _mm_setzero_si128(); + samples128_8 = _mm_setzero_si128(); +#if 1 + { + int runningOrder = order; + if (runningOrder >= 4) { + coefficients128_0 = _mm_loadu_si128((const __m128i*)(coefficients + 0)); + samples128_0 = _mm_loadu_si128((const __m128i*)(pSamplesOut - 4)); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: coefficients128_0 = _mm_set_epi32(0, coefficients[2], coefficients[1], coefficients[0]); samples128_0 = _mm_set_epi32(pSamplesOut[-1], pSamplesOut[-2], pSamplesOut[-3], 0); break; + case 2: coefficients128_0 = _mm_set_epi32(0, 0, coefficients[1], coefficients[0]); samples128_0 = _mm_set_epi32(pSamplesOut[-1], pSamplesOut[-2], 0, 0); break; + case 1: coefficients128_0 = _mm_set_epi32(0, 0, 0, coefficients[0]); samples128_0 = _mm_set_epi32(pSamplesOut[-1], 0, 0, 0); break; + } + runningOrder = 0; + } + if (runningOrder >= 4) { + coefficients128_4 = _mm_loadu_si128((const __m128i*)(coefficients + 4)); + samples128_4 = _mm_loadu_si128((const __m128i*)(pSamplesOut - 8)); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: coefficients128_4 = _mm_set_epi32(0, coefficients[6], coefficients[5], coefficients[4]); samples128_4 = _mm_set_epi32(pSamplesOut[-5], pSamplesOut[-6], pSamplesOut[-7], 0); break; + case 2: coefficients128_4 = _mm_set_epi32(0, 0, coefficients[5], coefficients[4]); samples128_4 = _mm_set_epi32(pSamplesOut[-5], pSamplesOut[-6], 0, 0); break; + case 1: coefficients128_4 = _mm_set_epi32(0, 0, 0, coefficients[4]); samples128_4 = _mm_set_epi32(pSamplesOut[-5], 0, 0, 0); break; + } + runningOrder = 0; + } + if (runningOrder == 4) { + coefficients128_8 = _mm_loadu_si128((const __m128i*)(coefficients + 8)); + samples128_8 = _mm_loadu_si128((const __m128i*)(pSamplesOut - 12)); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: coefficients128_8 = _mm_set_epi32(0, coefficients[10], coefficients[9], coefficients[8]); samples128_8 = _mm_set_epi32(pSamplesOut[-9], pSamplesOut[-10], pSamplesOut[-11], 0); break; + case 2: coefficients128_8 = _mm_set_epi32(0, 0, coefficients[9], coefficients[8]); samples128_8 = _mm_set_epi32(pSamplesOut[-9], pSamplesOut[-10], 0, 0); break; + case 1: coefficients128_8 = _mm_set_epi32(0, 0, 0, coefficients[8]); samples128_8 = _mm_set_epi32(pSamplesOut[-9], 0, 0, 0); break; + } + runningOrder = 0; + } + coefficients128_0 = _mm_shuffle_epi32(coefficients128_0, _MM_SHUFFLE(0, 1, 2, 3)); + coefficients128_4 = _mm_shuffle_epi32(coefficients128_4, _MM_SHUFFLE(0, 1, 2, 3)); + coefficients128_8 = _mm_shuffle_epi32(coefficients128_8, _MM_SHUFFLE(0, 1, 2, 3)); + } +#else + switch (order) + { + case 12: ((ma_int32*)&coefficients128_8)[0] = coefficients[11]; ((ma_int32*)&samples128_8)[0] = pDecodedSamples[-12]; + case 11: ((ma_int32*)&coefficients128_8)[1] = coefficients[10]; ((ma_int32*)&samples128_8)[1] = pDecodedSamples[-11]; + case 10: ((ma_int32*)&coefficients128_8)[2] = coefficients[ 9]; ((ma_int32*)&samples128_8)[2] = pDecodedSamples[-10]; + case 9: ((ma_int32*)&coefficients128_8)[3] = coefficients[ 8]; ((ma_int32*)&samples128_8)[3] = pDecodedSamples[- 9]; + case 8: ((ma_int32*)&coefficients128_4)[0] = coefficients[ 7]; ((ma_int32*)&samples128_4)[0] = pDecodedSamples[- 8]; + case 7: ((ma_int32*)&coefficients128_4)[1] = coefficients[ 6]; ((ma_int32*)&samples128_4)[1] = pDecodedSamples[- 7]; + case 6: ((ma_int32*)&coefficients128_4)[2] = coefficients[ 5]; ((ma_int32*)&samples128_4)[2] = pDecodedSamples[- 6]; + case 5: ((ma_int32*)&coefficients128_4)[3] = coefficients[ 4]; ((ma_int32*)&samples128_4)[3] = pDecodedSamples[- 5]; + case 4: ((ma_int32*)&coefficients128_0)[0] = coefficients[ 3]; ((ma_int32*)&samples128_0)[0] = pDecodedSamples[- 4]; + case 3: ((ma_int32*)&coefficients128_0)[1] = coefficients[ 2]; ((ma_int32*)&samples128_0)[1] = pDecodedSamples[- 3]; + case 2: ((ma_int32*)&coefficients128_0)[2] = coefficients[ 1]; ((ma_int32*)&samples128_0)[2] = pDecodedSamples[- 2]; + case 1: ((ma_int32*)&coefficients128_0)[3] = coefficients[ 0]; ((ma_int32*)&samples128_0)[3] = pDecodedSamples[- 1]; + } +#endif + while (pDecodedSamples < pDecodedSamplesEnd) { + __m128i prediction128; + __m128i zeroCountPart128; + __m128i riceParamPart128; + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts0, &riceParamParts0) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts1, &riceParamParts1) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts2, &riceParamParts2) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts3, &riceParamParts3)) { + return MA_FALSE; + } + zeroCountPart128 = _mm_set_epi32(zeroCountParts3, zeroCountParts2, zeroCountParts1, zeroCountParts0); + riceParamPart128 = _mm_set_epi32(riceParamParts3, riceParamParts2, riceParamParts1, riceParamParts0); + riceParamPart128 = _mm_and_si128(riceParamPart128, riceParamMask128); + riceParamPart128 = _mm_or_si128(riceParamPart128, _mm_slli_epi32(zeroCountPart128, riceParam)); + riceParamPart128 = _mm_xor_si128(_mm_srli_epi32(riceParamPart128, 1), _mm_add_epi32(ma_dr_flac__mm_not_si128(_mm_and_si128(riceParamPart128, _mm_set1_epi32(0x01))), _mm_set1_epi32(0x01))); + if (order <= 4) { + for (i = 0; i < 4; i += 1) { + prediction128 = _mm_mullo_epi32(coefficients128_0, samples128_0); + prediction128 = ma_dr_flac__mm_hadd_epi32(prediction128); + prediction128 = _mm_srai_epi32(prediction128, shift); + prediction128 = _mm_add_epi32(riceParamPart128, prediction128); + samples128_0 = _mm_alignr_epi8(prediction128, samples128_0, 4); + riceParamPart128 = _mm_alignr_epi8(_mm_setzero_si128(), riceParamPart128, 4); + } + } else if (order <= 8) { + for (i = 0; i < 4; i += 1) { + prediction128 = _mm_mullo_epi32(coefficients128_4, samples128_4); + prediction128 = _mm_add_epi32(prediction128, _mm_mullo_epi32(coefficients128_0, samples128_0)); + prediction128 = ma_dr_flac__mm_hadd_epi32(prediction128); + prediction128 = _mm_srai_epi32(prediction128, shift); + prediction128 = _mm_add_epi32(riceParamPart128, prediction128); + samples128_4 = _mm_alignr_epi8(samples128_0, samples128_4, 4); + samples128_0 = _mm_alignr_epi8(prediction128, samples128_0, 4); + riceParamPart128 = _mm_alignr_epi8(_mm_setzero_si128(), riceParamPart128, 4); + } + } else { + for (i = 0; i < 4; i += 1) { + prediction128 = _mm_mullo_epi32(coefficients128_8, samples128_8); + prediction128 = _mm_add_epi32(prediction128, _mm_mullo_epi32(coefficients128_4, samples128_4)); + prediction128 = _mm_add_epi32(prediction128, _mm_mullo_epi32(coefficients128_0, samples128_0)); + prediction128 = ma_dr_flac__mm_hadd_epi32(prediction128); + prediction128 = _mm_srai_epi32(prediction128, shift); + prediction128 = _mm_add_epi32(riceParamPart128, prediction128); + samples128_8 = _mm_alignr_epi8(samples128_4, samples128_8, 4); + samples128_4 = _mm_alignr_epi8(samples128_0, samples128_4, 4); + samples128_0 = _mm_alignr_epi8(prediction128, samples128_0, 4); + riceParamPart128 = _mm_alignr_epi8(_mm_setzero_si128(), riceParamPart128, 4); + } + } + _mm_storeu_si128((__m128i*)pDecodedSamples, samples128_0); + pDecodedSamples += 4; + } + i = (count & ~3); + while (i < (int)count) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts0, &riceParamParts0)) { + return MA_FALSE; + } + riceParamParts0 &= riceParamMask; + riceParamParts0 |= (zeroCountParts0 << riceParam); + riceParamParts0 = (riceParamParts0 >> 1) ^ t[riceParamParts0 & 0x01]; + pDecodedSamples[0] = riceParamParts0 + ma_dr_flac__calculate_prediction_32(order, shift, coefficients, pDecodedSamples); + i += 1; + pDecodedSamples += 1; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__sse41_64(ma_dr_flac_bs* bs, ma_uint32 count, ma_uint8 riceParam, ma_uint32 order, ma_int32 shift, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + int i; + ma_uint32 riceParamMask; + ma_int32* pDecodedSamples = pSamplesOut; + ma_int32* pDecodedSamplesEnd = pSamplesOut + (count & ~3); + ma_uint32 zeroCountParts0 = 0; + ma_uint32 zeroCountParts1 = 0; + ma_uint32 zeroCountParts2 = 0; + ma_uint32 zeroCountParts3 = 0; + ma_uint32 riceParamParts0 = 0; + ma_uint32 riceParamParts1 = 0; + ma_uint32 riceParamParts2 = 0; + ma_uint32 riceParamParts3 = 0; + __m128i coefficients128_0; + __m128i coefficients128_4; + __m128i coefficients128_8; + __m128i samples128_0; + __m128i samples128_4; + __m128i samples128_8; + __m128i prediction128; + __m128i riceParamMask128; + const ma_uint32 t[2] = {0x00000000, 0xFFFFFFFF}; + MA_DR_FLAC_ASSERT(order <= 12); + riceParamMask = (ma_uint32)~((~0UL) << riceParam); + riceParamMask128 = _mm_set1_epi32(riceParamMask); + prediction128 = _mm_setzero_si128(); + coefficients128_0 = _mm_setzero_si128(); + coefficients128_4 = _mm_setzero_si128(); + coefficients128_8 = _mm_setzero_si128(); + samples128_0 = _mm_setzero_si128(); + samples128_4 = _mm_setzero_si128(); + samples128_8 = _mm_setzero_si128(); +#if 1 + { + int runningOrder = order; + if (runningOrder >= 4) { + coefficients128_0 = _mm_loadu_si128((const __m128i*)(coefficients + 0)); + samples128_0 = _mm_loadu_si128((const __m128i*)(pSamplesOut - 4)); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: coefficients128_0 = _mm_set_epi32(0, coefficients[2], coefficients[1], coefficients[0]); samples128_0 = _mm_set_epi32(pSamplesOut[-1], pSamplesOut[-2], pSamplesOut[-3], 0); break; + case 2: coefficients128_0 = _mm_set_epi32(0, 0, coefficients[1], coefficients[0]); samples128_0 = _mm_set_epi32(pSamplesOut[-1], pSamplesOut[-2], 0, 0); break; + case 1: coefficients128_0 = _mm_set_epi32(0, 0, 0, coefficients[0]); samples128_0 = _mm_set_epi32(pSamplesOut[-1], 0, 0, 0); break; + } + runningOrder = 0; + } + if (runningOrder >= 4) { + coefficients128_4 = _mm_loadu_si128((const __m128i*)(coefficients + 4)); + samples128_4 = _mm_loadu_si128((const __m128i*)(pSamplesOut - 8)); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: coefficients128_4 = _mm_set_epi32(0, coefficients[6], coefficients[5], coefficients[4]); samples128_4 = _mm_set_epi32(pSamplesOut[-5], pSamplesOut[-6], pSamplesOut[-7], 0); break; + case 2: coefficients128_4 = _mm_set_epi32(0, 0, coefficients[5], coefficients[4]); samples128_4 = _mm_set_epi32(pSamplesOut[-5], pSamplesOut[-6], 0, 0); break; + case 1: coefficients128_4 = _mm_set_epi32(0, 0, 0, coefficients[4]); samples128_4 = _mm_set_epi32(pSamplesOut[-5], 0, 0, 0); break; + } + runningOrder = 0; + } + if (runningOrder == 4) { + coefficients128_8 = _mm_loadu_si128((const __m128i*)(coefficients + 8)); + samples128_8 = _mm_loadu_si128((const __m128i*)(pSamplesOut - 12)); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: coefficients128_8 = _mm_set_epi32(0, coefficients[10], coefficients[9], coefficients[8]); samples128_8 = _mm_set_epi32(pSamplesOut[-9], pSamplesOut[-10], pSamplesOut[-11], 0); break; + case 2: coefficients128_8 = _mm_set_epi32(0, 0, coefficients[9], coefficients[8]); samples128_8 = _mm_set_epi32(pSamplesOut[-9], pSamplesOut[-10], 0, 0); break; + case 1: coefficients128_8 = _mm_set_epi32(0, 0, 0, coefficients[8]); samples128_8 = _mm_set_epi32(pSamplesOut[-9], 0, 0, 0); break; + } + runningOrder = 0; + } + coefficients128_0 = _mm_shuffle_epi32(coefficients128_0, _MM_SHUFFLE(0, 1, 2, 3)); + coefficients128_4 = _mm_shuffle_epi32(coefficients128_4, _MM_SHUFFLE(0, 1, 2, 3)); + coefficients128_8 = _mm_shuffle_epi32(coefficients128_8, _MM_SHUFFLE(0, 1, 2, 3)); + } +#else + switch (order) + { + case 12: ((ma_int32*)&coefficients128_8)[0] = coefficients[11]; ((ma_int32*)&samples128_8)[0] = pDecodedSamples[-12]; + case 11: ((ma_int32*)&coefficients128_8)[1] = coefficients[10]; ((ma_int32*)&samples128_8)[1] = pDecodedSamples[-11]; + case 10: ((ma_int32*)&coefficients128_8)[2] = coefficients[ 9]; ((ma_int32*)&samples128_8)[2] = pDecodedSamples[-10]; + case 9: ((ma_int32*)&coefficients128_8)[3] = coefficients[ 8]; ((ma_int32*)&samples128_8)[3] = pDecodedSamples[- 9]; + case 8: ((ma_int32*)&coefficients128_4)[0] = coefficients[ 7]; ((ma_int32*)&samples128_4)[0] = pDecodedSamples[- 8]; + case 7: ((ma_int32*)&coefficients128_4)[1] = coefficients[ 6]; ((ma_int32*)&samples128_4)[1] = pDecodedSamples[- 7]; + case 6: ((ma_int32*)&coefficients128_4)[2] = coefficients[ 5]; ((ma_int32*)&samples128_4)[2] = pDecodedSamples[- 6]; + case 5: ((ma_int32*)&coefficients128_4)[3] = coefficients[ 4]; ((ma_int32*)&samples128_4)[3] = pDecodedSamples[- 5]; + case 4: ((ma_int32*)&coefficients128_0)[0] = coefficients[ 3]; ((ma_int32*)&samples128_0)[0] = pDecodedSamples[- 4]; + case 3: ((ma_int32*)&coefficients128_0)[1] = coefficients[ 2]; ((ma_int32*)&samples128_0)[1] = pDecodedSamples[- 3]; + case 2: ((ma_int32*)&coefficients128_0)[2] = coefficients[ 1]; ((ma_int32*)&samples128_0)[2] = pDecodedSamples[- 2]; + case 1: ((ma_int32*)&coefficients128_0)[3] = coefficients[ 0]; ((ma_int32*)&samples128_0)[3] = pDecodedSamples[- 1]; + } +#endif + while (pDecodedSamples < pDecodedSamplesEnd) { + __m128i zeroCountPart128; + __m128i riceParamPart128; + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts0, &riceParamParts0) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts1, &riceParamParts1) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts2, &riceParamParts2) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts3, &riceParamParts3)) { + return MA_FALSE; + } + zeroCountPart128 = _mm_set_epi32(zeroCountParts3, zeroCountParts2, zeroCountParts1, zeroCountParts0); + riceParamPart128 = _mm_set_epi32(riceParamParts3, riceParamParts2, riceParamParts1, riceParamParts0); + riceParamPart128 = _mm_and_si128(riceParamPart128, riceParamMask128); + riceParamPart128 = _mm_or_si128(riceParamPart128, _mm_slli_epi32(zeroCountPart128, riceParam)); + riceParamPart128 = _mm_xor_si128(_mm_srli_epi32(riceParamPart128, 1), _mm_add_epi32(ma_dr_flac__mm_not_si128(_mm_and_si128(riceParamPart128, _mm_set1_epi32(1))), _mm_set1_epi32(1))); + for (i = 0; i < 4; i += 1) { + prediction128 = _mm_xor_si128(prediction128, prediction128); + switch (order) + { + case 12: + case 11: prediction128 = _mm_add_epi64(prediction128, _mm_mul_epi32(_mm_shuffle_epi32(coefficients128_8, _MM_SHUFFLE(1, 1, 0, 0)), _mm_shuffle_epi32(samples128_8, _MM_SHUFFLE(1, 1, 0, 0)))); + case 10: + case 9: prediction128 = _mm_add_epi64(prediction128, _mm_mul_epi32(_mm_shuffle_epi32(coefficients128_8, _MM_SHUFFLE(3, 3, 2, 2)), _mm_shuffle_epi32(samples128_8, _MM_SHUFFLE(3, 3, 2, 2)))); + case 8: + case 7: prediction128 = _mm_add_epi64(prediction128, _mm_mul_epi32(_mm_shuffle_epi32(coefficients128_4, _MM_SHUFFLE(1, 1, 0, 0)), _mm_shuffle_epi32(samples128_4, _MM_SHUFFLE(1, 1, 0, 0)))); + case 6: + case 5: prediction128 = _mm_add_epi64(prediction128, _mm_mul_epi32(_mm_shuffle_epi32(coefficients128_4, _MM_SHUFFLE(3, 3, 2, 2)), _mm_shuffle_epi32(samples128_4, _MM_SHUFFLE(3, 3, 2, 2)))); + case 4: + case 3: prediction128 = _mm_add_epi64(prediction128, _mm_mul_epi32(_mm_shuffle_epi32(coefficients128_0, _MM_SHUFFLE(1, 1, 0, 0)), _mm_shuffle_epi32(samples128_0, _MM_SHUFFLE(1, 1, 0, 0)))); + case 2: + case 1: prediction128 = _mm_add_epi64(prediction128, _mm_mul_epi32(_mm_shuffle_epi32(coefficients128_0, _MM_SHUFFLE(3, 3, 2, 2)), _mm_shuffle_epi32(samples128_0, _MM_SHUFFLE(3, 3, 2, 2)))); + } + prediction128 = ma_dr_flac__mm_hadd_epi64(prediction128); + prediction128 = ma_dr_flac__mm_srai_epi64(prediction128, shift); + prediction128 = _mm_add_epi32(riceParamPart128, prediction128); + samples128_8 = _mm_alignr_epi8(samples128_4, samples128_8, 4); + samples128_4 = _mm_alignr_epi8(samples128_0, samples128_4, 4); + samples128_0 = _mm_alignr_epi8(prediction128, samples128_0, 4); + riceParamPart128 = _mm_alignr_epi8(_mm_setzero_si128(), riceParamPart128, 4); + } + _mm_storeu_si128((__m128i*)pDecodedSamples, samples128_0); + pDecodedSamples += 4; + } + i = (count & ~3); + while (i < (int)count) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts0, &riceParamParts0)) { + return MA_FALSE; + } + riceParamParts0 &= riceParamMask; + riceParamParts0 |= (zeroCountParts0 << riceParam); + riceParamParts0 = (riceParamParts0 >> 1) ^ t[riceParamParts0 & 0x01]; + pDecodedSamples[0] = riceParamParts0 + ma_dr_flac__calculate_prediction_64(order, shift, coefficients, pDecodedSamples); + i += 1; + pDecodedSamples += 1; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__sse41(ma_dr_flac_bs* bs, ma_uint32 bitsPerSample, ma_uint32 count, ma_uint8 riceParam, ma_uint32 lpcOrder, ma_int32 lpcShift, ma_uint32 lpcPrecision, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pSamplesOut != NULL); + if (lpcOrder > 0 && lpcOrder <= 12) { + if (ma_dr_flac__use_64_bit_prediction(bitsPerSample, lpcOrder, lpcPrecision)) { + return ma_dr_flac__decode_samples_with_residual__rice__sse41_64(bs, count, riceParam, lpcOrder, lpcShift, coefficients, pSamplesOut); + } else { + return ma_dr_flac__decode_samples_with_residual__rice__sse41_32(bs, count, riceParam, lpcOrder, lpcShift, coefficients, pSamplesOut); + } + } else { + return ma_dr_flac__decode_samples_with_residual__rice__scalar(bs, bitsPerSample, count, riceParam, lpcOrder, lpcShift, lpcPrecision, coefficients, pSamplesOut); + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac__vst2q_s32(ma_int32* p, int32x4x2_t x) +{ + vst1q_s32(p+0, x.val[0]); + vst1q_s32(p+4, x.val[1]); +} +static MA_INLINE void ma_dr_flac__vst2q_u32(ma_uint32* p, uint32x4x2_t x) +{ + vst1q_u32(p+0, x.val[0]); + vst1q_u32(p+4, x.val[1]); +} +static MA_INLINE void ma_dr_flac__vst2q_f32(float* p, float32x4x2_t x) +{ + vst1q_f32(p+0, x.val[0]); + vst1q_f32(p+4, x.val[1]); +} +static MA_INLINE void ma_dr_flac__vst2q_s16(ma_int16* p, int16x4x2_t x) +{ + vst1q_s16(p, vcombine_s16(x.val[0], x.val[1])); +} +static MA_INLINE void ma_dr_flac__vst2q_u16(ma_uint16* p, uint16x4x2_t x) +{ + vst1q_u16(p, vcombine_u16(x.val[0], x.val[1])); +} +static MA_INLINE int32x4_t ma_dr_flac__vdupq_n_s32x4(ma_int32 x3, ma_int32 x2, ma_int32 x1, ma_int32 x0) +{ + ma_int32 x[4]; + x[3] = x3; + x[2] = x2; + x[1] = x1; + x[0] = x0; + return vld1q_s32(x); +} +static MA_INLINE int32x4_t ma_dr_flac__valignrq_s32_1(int32x4_t a, int32x4_t b) +{ + return vextq_s32(b, a, 1); +} +static MA_INLINE uint32x4_t ma_dr_flac__valignrq_u32_1(uint32x4_t a, uint32x4_t b) +{ + return vextq_u32(b, a, 1); +} +static MA_INLINE int32x2_t ma_dr_flac__vhaddq_s32(int32x4_t x) +{ + int32x2_t r = vadd_s32(vget_high_s32(x), vget_low_s32(x)); + return vpadd_s32(r, r); +} +static MA_INLINE int64x1_t ma_dr_flac__vhaddq_s64(int64x2_t x) +{ + return vadd_s64(vget_high_s64(x), vget_low_s64(x)); +} +static MA_INLINE int32x4_t ma_dr_flac__vrevq_s32(int32x4_t x) +{ + return vrev64q_s32(vcombine_s32(vget_high_s32(x), vget_low_s32(x))); +} +static MA_INLINE int32x4_t ma_dr_flac__vnotq_s32(int32x4_t x) +{ + return veorq_s32(x, vdupq_n_s32(0xFFFFFFFF)); +} +static MA_INLINE uint32x4_t ma_dr_flac__vnotq_u32(uint32x4_t x) +{ + return veorq_u32(x, vdupq_n_u32(0xFFFFFFFF)); +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__neon_32(ma_dr_flac_bs* bs, ma_uint32 count, ma_uint8 riceParam, ma_uint32 order, ma_int32 shift, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + int i; + ma_uint32 riceParamMask; + ma_int32* pDecodedSamples = pSamplesOut; + ma_int32* pDecodedSamplesEnd = pSamplesOut + (count & ~3); + ma_uint32 zeroCountParts[4]; + ma_uint32 riceParamParts[4]; + int32x4_t coefficients128_0; + int32x4_t coefficients128_4; + int32x4_t coefficients128_8; + int32x4_t samples128_0; + int32x4_t samples128_4; + int32x4_t samples128_8; + uint32x4_t riceParamMask128; + int32x4_t riceParam128; + int32x2_t shift64; + uint32x4_t one128; + const ma_uint32 t[2] = {0x00000000, 0xFFFFFFFF}; + riceParamMask = (ma_uint32)~((~0UL) << riceParam); + riceParamMask128 = vdupq_n_u32(riceParamMask); + riceParam128 = vdupq_n_s32(riceParam); + shift64 = vdup_n_s32(-shift); + one128 = vdupq_n_u32(1); + { + int runningOrder = order; + ma_int32 tempC[4] = {0, 0, 0, 0}; + ma_int32 tempS[4] = {0, 0, 0, 0}; + if (runningOrder >= 4) { + coefficients128_0 = vld1q_s32(coefficients + 0); + samples128_0 = vld1q_s32(pSamplesOut - 4); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: tempC[2] = coefficients[2]; tempS[1] = pSamplesOut[-3]; + case 2: tempC[1] = coefficients[1]; tempS[2] = pSamplesOut[-2]; + case 1: tempC[0] = coefficients[0]; tempS[3] = pSamplesOut[-1]; + } + coefficients128_0 = vld1q_s32(tempC); + samples128_0 = vld1q_s32(tempS); + runningOrder = 0; + } + if (runningOrder >= 4) { + coefficients128_4 = vld1q_s32(coefficients + 4); + samples128_4 = vld1q_s32(pSamplesOut - 8); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: tempC[2] = coefficients[6]; tempS[1] = pSamplesOut[-7]; + case 2: tempC[1] = coefficients[5]; tempS[2] = pSamplesOut[-6]; + case 1: tempC[0] = coefficients[4]; tempS[3] = pSamplesOut[-5]; + } + coefficients128_4 = vld1q_s32(tempC); + samples128_4 = vld1q_s32(tempS); + runningOrder = 0; + } + if (runningOrder == 4) { + coefficients128_8 = vld1q_s32(coefficients + 8); + samples128_8 = vld1q_s32(pSamplesOut - 12); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: tempC[2] = coefficients[10]; tempS[1] = pSamplesOut[-11]; + case 2: tempC[1] = coefficients[ 9]; tempS[2] = pSamplesOut[-10]; + case 1: tempC[0] = coefficients[ 8]; tempS[3] = pSamplesOut[- 9]; + } + coefficients128_8 = vld1q_s32(tempC); + samples128_8 = vld1q_s32(tempS); + runningOrder = 0; + } + coefficients128_0 = ma_dr_flac__vrevq_s32(coefficients128_0); + coefficients128_4 = ma_dr_flac__vrevq_s32(coefficients128_4); + coefficients128_8 = ma_dr_flac__vrevq_s32(coefficients128_8); + } + while (pDecodedSamples < pDecodedSamplesEnd) { + int32x4_t prediction128; + int32x2_t prediction64; + uint32x4_t zeroCountPart128; + uint32x4_t riceParamPart128; + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[0], &riceParamParts[0]) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[1], &riceParamParts[1]) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[2], &riceParamParts[2]) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[3], &riceParamParts[3])) { + return MA_FALSE; + } + zeroCountPart128 = vld1q_u32(zeroCountParts); + riceParamPart128 = vld1q_u32(riceParamParts); + riceParamPart128 = vandq_u32(riceParamPart128, riceParamMask128); + riceParamPart128 = vorrq_u32(riceParamPart128, vshlq_u32(zeroCountPart128, riceParam128)); + riceParamPart128 = veorq_u32(vshrq_n_u32(riceParamPart128, 1), vaddq_u32(ma_dr_flac__vnotq_u32(vandq_u32(riceParamPart128, one128)), one128)); + if (order <= 4) { + for (i = 0; i < 4; i += 1) { + prediction128 = vmulq_s32(coefficients128_0, samples128_0); + prediction64 = ma_dr_flac__vhaddq_s32(prediction128); + prediction64 = vshl_s32(prediction64, shift64); + prediction64 = vadd_s32(prediction64, vget_low_s32(vreinterpretq_s32_u32(riceParamPart128))); + samples128_0 = ma_dr_flac__valignrq_s32_1(vcombine_s32(prediction64, vdup_n_s32(0)), samples128_0); + riceParamPart128 = ma_dr_flac__valignrq_u32_1(vdupq_n_u32(0), riceParamPart128); + } + } else if (order <= 8) { + for (i = 0; i < 4; i += 1) { + prediction128 = vmulq_s32(coefficients128_4, samples128_4); + prediction128 = vmlaq_s32(prediction128, coefficients128_0, samples128_0); + prediction64 = ma_dr_flac__vhaddq_s32(prediction128); + prediction64 = vshl_s32(prediction64, shift64); + prediction64 = vadd_s32(prediction64, vget_low_s32(vreinterpretq_s32_u32(riceParamPart128))); + samples128_4 = ma_dr_flac__valignrq_s32_1(samples128_0, samples128_4); + samples128_0 = ma_dr_flac__valignrq_s32_1(vcombine_s32(prediction64, vdup_n_s32(0)), samples128_0); + riceParamPart128 = ma_dr_flac__valignrq_u32_1(vdupq_n_u32(0), riceParamPart128); + } + } else { + for (i = 0; i < 4; i += 1) { + prediction128 = vmulq_s32(coefficients128_8, samples128_8); + prediction128 = vmlaq_s32(prediction128, coefficients128_4, samples128_4); + prediction128 = vmlaq_s32(prediction128, coefficients128_0, samples128_0); + prediction64 = ma_dr_flac__vhaddq_s32(prediction128); + prediction64 = vshl_s32(prediction64, shift64); + prediction64 = vadd_s32(prediction64, vget_low_s32(vreinterpretq_s32_u32(riceParamPart128))); + samples128_8 = ma_dr_flac__valignrq_s32_1(samples128_4, samples128_8); + samples128_4 = ma_dr_flac__valignrq_s32_1(samples128_0, samples128_4); + samples128_0 = ma_dr_flac__valignrq_s32_1(vcombine_s32(prediction64, vdup_n_s32(0)), samples128_0); + riceParamPart128 = ma_dr_flac__valignrq_u32_1(vdupq_n_u32(0), riceParamPart128); + } + } + vst1q_s32(pDecodedSamples, samples128_0); + pDecodedSamples += 4; + } + i = (count & ~3); + while (i < (int)count) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[0], &riceParamParts[0])) { + return MA_FALSE; + } + riceParamParts[0] &= riceParamMask; + riceParamParts[0] |= (zeroCountParts[0] << riceParam); + riceParamParts[0] = (riceParamParts[0] >> 1) ^ t[riceParamParts[0] & 0x01]; + pDecodedSamples[0] = riceParamParts[0] + ma_dr_flac__calculate_prediction_32(order, shift, coefficients, pDecodedSamples); + i += 1; + pDecodedSamples += 1; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__neon_64(ma_dr_flac_bs* bs, ma_uint32 count, ma_uint8 riceParam, ma_uint32 order, ma_int32 shift, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + int i; + ma_uint32 riceParamMask; + ma_int32* pDecodedSamples = pSamplesOut; + ma_int32* pDecodedSamplesEnd = pSamplesOut + (count & ~3); + ma_uint32 zeroCountParts[4]; + ma_uint32 riceParamParts[4]; + int32x4_t coefficients128_0; + int32x4_t coefficients128_4; + int32x4_t coefficients128_8; + int32x4_t samples128_0; + int32x4_t samples128_4; + int32x4_t samples128_8; + uint32x4_t riceParamMask128; + int32x4_t riceParam128; + int64x1_t shift64; + uint32x4_t one128; + int64x2_t prediction128 = { 0 }; + uint32x4_t zeroCountPart128; + uint32x4_t riceParamPart128; + const ma_uint32 t[2] = {0x00000000, 0xFFFFFFFF}; + riceParamMask = (ma_uint32)~((~0UL) << riceParam); + riceParamMask128 = vdupq_n_u32(riceParamMask); + riceParam128 = vdupq_n_s32(riceParam); + shift64 = vdup_n_s64(-shift); + one128 = vdupq_n_u32(1); + { + int runningOrder = order; + ma_int32 tempC[4] = {0, 0, 0, 0}; + ma_int32 tempS[4] = {0, 0, 0, 0}; + if (runningOrder >= 4) { + coefficients128_0 = vld1q_s32(coefficients + 0); + samples128_0 = vld1q_s32(pSamplesOut - 4); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: tempC[2] = coefficients[2]; tempS[1] = pSamplesOut[-3]; + case 2: tempC[1] = coefficients[1]; tempS[2] = pSamplesOut[-2]; + case 1: tempC[0] = coefficients[0]; tempS[3] = pSamplesOut[-1]; + } + coefficients128_0 = vld1q_s32(tempC); + samples128_0 = vld1q_s32(tempS); + runningOrder = 0; + } + if (runningOrder >= 4) { + coefficients128_4 = vld1q_s32(coefficients + 4); + samples128_4 = vld1q_s32(pSamplesOut - 8); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: tempC[2] = coefficients[6]; tempS[1] = pSamplesOut[-7]; + case 2: tempC[1] = coefficients[5]; tempS[2] = pSamplesOut[-6]; + case 1: tempC[0] = coefficients[4]; tempS[3] = pSamplesOut[-5]; + } + coefficients128_4 = vld1q_s32(tempC); + samples128_4 = vld1q_s32(tempS); + runningOrder = 0; + } + if (runningOrder == 4) { + coefficients128_8 = vld1q_s32(coefficients + 8); + samples128_8 = vld1q_s32(pSamplesOut - 12); + runningOrder -= 4; + } else { + switch (runningOrder) { + case 3: tempC[2] = coefficients[10]; tempS[1] = pSamplesOut[-11]; + case 2: tempC[1] = coefficients[ 9]; tempS[2] = pSamplesOut[-10]; + case 1: tempC[0] = coefficients[ 8]; tempS[3] = pSamplesOut[- 9]; + } + coefficients128_8 = vld1q_s32(tempC); + samples128_8 = vld1q_s32(tempS); + runningOrder = 0; + } + coefficients128_0 = ma_dr_flac__vrevq_s32(coefficients128_0); + coefficients128_4 = ma_dr_flac__vrevq_s32(coefficients128_4); + coefficients128_8 = ma_dr_flac__vrevq_s32(coefficients128_8); + } + while (pDecodedSamples < pDecodedSamplesEnd) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[0], &riceParamParts[0]) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[1], &riceParamParts[1]) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[2], &riceParamParts[2]) || + !ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[3], &riceParamParts[3])) { + return MA_FALSE; + } + zeroCountPart128 = vld1q_u32(zeroCountParts); + riceParamPart128 = vld1q_u32(riceParamParts); + riceParamPart128 = vandq_u32(riceParamPart128, riceParamMask128); + riceParamPart128 = vorrq_u32(riceParamPart128, vshlq_u32(zeroCountPart128, riceParam128)); + riceParamPart128 = veorq_u32(vshrq_n_u32(riceParamPart128, 1), vaddq_u32(ma_dr_flac__vnotq_u32(vandq_u32(riceParamPart128, one128)), one128)); + for (i = 0; i < 4; i += 1) { + int64x1_t prediction64; + prediction128 = veorq_s64(prediction128, prediction128); + switch (order) + { + case 12: + case 11: prediction128 = vaddq_s64(prediction128, vmull_s32(vget_low_s32(coefficients128_8), vget_low_s32(samples128_8))); + case 10: + case 9: prediction128 = vaddq_s64(prediction128, vmull_s32(vget_high_s32(coefficients128_8), vget_high_s32(samples128_8))); + case 8: + case 7: prediction128 = vaddq_s64(prediction128, vmull_s32(vget_low_s32(coefficients128_4), vget_low_s32(samples128_4))); + case 6: + case 5: prediction128 = vaddq_s64(prediction128, vmull_s32(vget_high_s32(coefficients128_4), vget_high_s32(samples128_4))); + case 4: + case 3: prediction128 = vaddq_s64(prediction128, vmull_s32(vget_low_s32(coefficients128_0), vget_low_s32(samples128_0))); + case 2: + case 1: prediction128 = vaddq_s64(prediction128, vmull_s32(vget_high_s32(coefficients128_0), vget_high_s32(samples128_0))); + } + prediction64 = ma_dr_flac__vhaddq_s64(prediction128); + prediction64 = vshl_s64(prediction64, shift64); + prediction64 = vadd_s64(prediction64, vdup_n_s64(vgetq_lane_u32(riceParamPart128, 0))); + samples128_8 = ma_dr_flac__valignrq_s32_1(samples128_4, samples128_8); + samples128_4 = ma_dr_flac__valignrq_s32_1(samples128_0, samples128_4); + samples128_0 = ma_dr_flac__valignrq_s32_1(vcombine_s32(vreinterpret_s32_s64(prediction64), vdup_n_s32(0)), samples128_0); + riceParamPart128 = ma_dr_flac__valignrq_u32_1(vdupq_n_u32(0), riceParamPart128); + } + vst1q_s32(pDecodedSamples, samples128_0); + pDecodedSamples += 4; + } + i = (count & ~3); + while (i < (int)count) { + if (!ma_dr_flac__read_rice_parts_x1(bs, riceParam, &zeroCountParts[0], &riceParamParts[0])) { + return MA_FALSE; + } + riceParamParts[0] &= riceParamMask; + riceParamParts[0] |= (zeroCountParts[0] << riceParam); + riceParamParts[0] = (riceParamParts[0] >> 1) ^ t[riceParamParts[0] & 0x01]; + pDecodedSamples[0] = riceParamParts[0] + ma_dr_flac__calculate_prediction_64(order, shift, coefficients, pDecodedSamples); + i += 1; + pDecodedSamples += 1; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice__neon(ma_dr_flac_bs* bs, ma_uint32 bitsPerSample, ma_uint32 count, ma_uint8 riceParam, ma_uint32 lpcOrder, ma_int32 lpcShift, ma_uint32 lpcPrecision, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(pSamplesOut != NULL); + if (lpcOrder > 0 && lpcOrder <= 12) { + if (ma_dr_flac__use_64_bit_prediction(bitsPerSample, lpcOrder, lpcPrecision)) { + return ma_dr_flac__decode_samples_with_residual__rice__neon_64(bs, count, riceParam, lpcOrder, lpcShift, coefficients, pSamplesOut); + } else { + return ma_dr_flac__decode_samples_with_residual__rice__neon_32(bs, count, riceParam, lpcOrder, lpcShift, coefficients, pSamplesOut); + } + } else { + return ma_dr_flac__decode_samples_with_residual__rice__scalar(bs, bitsPerSample, count, riceParam, lpcOrder, lpcShift, lpcPrecision, coefficients, pSamplesOut); + } +} +#endif +static ma_bool32 ma_dr_flac__decode_samples_with_residual__rice(ma_dr_flac_bs* bs, ma_uint32 bitsPerSample, ma_uint32 count, ma_uint8 riceParam, ma_uint32 lpcOrder, ma_int32 lpcShift, ma_uint32 lpcPrecision, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE41) + if (ma_dr_flac__gIsSSE41Supported) { + return ma_dr_flac__decode_samples_with_residual__rice__sse41(bs, bitsPerSample, count, riceParam, lpcOrder, lpcShift, lpcPrecision, coefficients, pSamplesOut); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported) { + return ma_dr_flac__decode_samples_with_residual__rice__neon(bs, bitsPerSample, count, riceParam, lpcOrder, lpcShift, lpcPrecision, coefficients, pSamplesOut); + } else +#endif + { + #if 0 + return ma_dr_flac__decode_samples_with_residual__rice__reference(bs, bitsPerSample, count, riceParam, lpcOrder, lpcShift, lpcPrecision, coefficients, pSamplesOut); + #else + return ma_dr_flac__decode_samples_with_residual__rice__scalar(bs, bitsPerSample, count, riceParam, lpcOrder, lpcShift, lpcPrecision, coefficients, pSamplesOut); + #endif + } +} +static ma_bool32 ma_dr_flac__read_and_seek_residual__rice(ma_dr_flac_bs* bs, ma_uint32 count, ma_uint8 riceParam) +{ + ma_uint32 i; + MA_DR_FLAC_ASSERT(bs != NULL); + for (i = 0; i < count; ++i) { + if (!ma_dr_flac__seek_rice_parts(bs, riceParam)) { + return MA_FALSE; + } + } + return MA_TRUE; +} +#if defined(__clang__) +__attribute__((no_sanitize("signed-integer-overflow"))) +#endif +static ma_bool32 ma_dr_flac__decode_samples_with_residual__unencoded(ma_dr_flac_bs* bs, ma_uint32 bitsPerSample, ma_uint32 count, ma_uint8 unencodedBitsPerSample, ma_uint32 lpcOrder, ma_int32 lpcShift, ma_uint32 lpcPrecision, const ma_int32* coefficients, ma_int32* pSamplesOut) +{ + ma_uint32 i; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(unencodedBitsPerSample <= 31); + MA_DR_FLAC_ASSERT(pSamplesOut != NULL); + for (i = 0; i < count; ++i) { + if (unencodedBitsPerSample > 0) { + if (!ma_dr_flac__read_int32(bs, unencodedBitsPerSample, pSamplesOut + i)) { + return MA_FALSE; + } + } else { + pSamplesOut[i] = 0; + } + if (ma_dr_flac__use_64_bit_prediction(bitsPerSample, lpcOrder, lpcPrecision)) { + pSamplesOut[i] += ma_dr_flac__calculate_prediction_64(lpcOrder, lpcShift, coefficients, pSamplesOut + i); + } else { + pSamplesOut[i] += ma_dr_flac__calculate_prediction_32(lpcOrder, lpcShift, coefficients, pSamplesOut + i); + } + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples_with_residual(ma_dr_flac_bs* bs, ma_uint32 bitsPerSample, ma_uint32 blockSize, ma_uint32 lpcOrder, ma_int32 lpcShift, ma_uint32 lpcPrecision, const ma_int32* coefficients, ma_int32* pDecodedSamples) +{ + ma_uint8 residualMethod; + ma_uint8 partitionOrder; + ma_uint32 samplesInPartition; + ma_uint32 partitionsRemaining; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(blockSize != 0); + MA_DR_FLAC_ASSERT(pDecodedSamples != NULL); + if (!ma_dr_flac__read_uint8(bs, 2, &residualMethod)) { + return MA_FALSE; + } + if (residualMethod != MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE && residualMethod != MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE2) { + return MA_FALSE; + } + pDecodedSamples += lpcOrder; + if (!ma_dr_flac__read_uint8(bs, 4, &partitionOrder)) { + return MA_FALSE; + } + if (partitionOrder > 8) { + return MA_FALSE; + } + if ((blockSize / (1 << partitionOrder)) < lpcOrder) { + return MA_FALSE; + } + samplesInPartition = (blockSize / (1 << partitionOrder)) - lpcOrder; + partitionsRemaining = (1 << partitionOrder); + for (;;) { + ma_uint8 riceParam = 0; + if (residualMethod == MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE) { + if (!ma_dr_flac__read_uint8(bs, 4, &riceParam)) { + return MA_FALSE; + } + if (riceParam == 15) { + riceParam = 0xFF; + } + } else if (residualMethod == MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE2) { + if (!ma_dr_flac__read_uint8(bs, 5, &riceParam)) { + return MA_FALSE; + } + if (riceParam == 31) { + riceParam = 0xFF; + } + } + if (riceParam != 0xFF) { + if (!ma_dr_flac__decode_samples_with_residual__rice(bs, bitsPerSample, samplesInPartition, riceParam, lpcOrder, lpcShift, lpcPrecision, coefficients, pDecodedSamples)) { + return MA_FALSE; + } + } else { + ma_uint8 unencodedBitsPerSample = 0; + if (!ma_dr_flac__read_uint8(bs, 5, &unencodedBitsPerSample)) { + return MA_FALSE; + } + if (!ma_dr_flac__decode_samples_with_residual__unencoded(bs, bitsPerSample, samplesInPartition, unencodedBitsPerSample, lpcOrder, lpcShift, lpcPrecision, coefficients, pDecodedSamples)) { + return MA_FALSE; + } + } + pDecodedSamples += samplesInPartition; + if (partitionsRemaining == 1) { + break; + } + partitionsRemaining -= 1; + if (partitionOrder != 0) { + samplesInPartition = blockSize / (1 << partitionOrder); + } + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__read_and_seek_residual(ma_dr_flac_bs* bs, ma_uint32 blockSize, ma_uint32 order) +{ + ma_uint8 residualMethod; + ma_uint8 partitionOrder; + ma_uint32 samplesInPartition; + ma_uint32 partitionsRemaining; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(blockSize != 0); + if (!ma_dr_flac__read_uint8(bs, 2, &residualMethod)) { + return MA_FALSE; + } + if (residualMethod != MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE && residualMethod != MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE2) { + return MA_FALSE; + } + if (!ma_dr_flac__read_uint8(bs, 4, &partitionOrder)) { + return MA_FALSE; + } + if (partitionOrder > 8) { + return MA_FALSE; + } + if ((blockSize / (1 << partitionOrder)) <= order) { + return MA_FALSE; + } + samplesInPartition = (blockSize / (1 << partitionOrder)) - order; + partitionsRemaining = (1 << partitionOrder); + for (;;) + { + ma_uint8 riceParam = 0; + if (residualMethod == MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE) { + if (!ma_dr_flac__read_uint8(bs, 4, &riceParam)) { + return MA_FALSE; + } + if (riceParam == 15) { + riceParam = 0xFF; + } + } else if (residualMethod == MA_DR_FLAC_RESIDUAL_CODING_METHOD_PARTITIONED_RICE2) { + if (!ma_dr_flac__read_uint8(bs, 5, &riceParam)) { + return MA_FALSE; + } + if (riceParam == 31) { + riceParam = 0xFF; + } + } + if (riceParam != 0xFF) { + if (!ma_dr_flac__read_and_seek_residual__rice(bs, samplesInPartition, riceParam)) { + return MA_FALSE; + } + } else { + ma_uint8 unencodedBitsPerSample = 0; + if (!ma_dr_flac__read_uint8(bs, 5, &unencodedBitsPerSample)) { + return MA_FALSE; + } + if (!ma_dr_flac__seek_bits(bs, unencodedBitsPerSample * samplesInPartition)) { + return MA_FALSE; + } + } + if (partitionsRemaining == 1) { + break; + } + partitionsRemaining -= 1; + samplesInPartition = blockSize / (1 << partitionOrder); + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples__constant(ma_dr_flac_bs* bs, ma_uint32 blockSize, ma_uint32 subframeBitsPerSample, ma_int32* pDecodedSamples) +{ + ma_uint32 i; + ma_int32 sample; + if (!ma_dr_flac__read_int32(bs, subframeBitsPerSample, &sample)) { + return MA_FALSE; + } + for (i = 0; i < blockSize; ++i) { + pDecodedSamples[i] = sample; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples__verbatim(ma_dr_flac_bs* bs, ma_uint32 blockSize, ma_uint32 subframeBitsPerSample, ma_int32* pDecodedSamples) +{ + ma_uint32 i; + for (i = 0; i < blockSize; ++i) { + ma_int32 sample; + if (!ma_dr_flac__read_int32(bs, subframeBitsPerSample, &sample)) { + return MA_FALSE; + } + pDecodedSamples[i] = sample; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples__fixed(ma_dr_flac_bs* bs, ma_uint32 blockSize, ma_uint32 subframeBitsPerSample, ma_uint8 lpcOrder, ma_int32* pDecodedSamples) +{ + ma_uint32 i; + static ma_int32 lpcCoefficientsTable[5][4] = { + {0, 0, 0, 0}, + {1, 0, 0, 0}, + {2, -1, 0, 0}, + {3, -3, 1, 0}, + {4, -6, 4, -1} + }; + for (i = 0; i < lpcOrder; ++i) { + ma_int32 sample; + if (!ma_dr_flac__read_int32(bs, subframeBitsPerSample, &sample)) { + return MA_FALSE; + } + pDecodedSamples[i] = sample; + } + if (!ma_dr_flac__decode_samples_with_residual(bs, subframeBitsPerSample, blockSize, lpcOrder, 0, 4, lpcCoefficientsTable[lpcOrder], pDecodedSamples)) { + return MA_FALSE; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_samples__lpc(ma_dr_flac_bs* bs, ma_uint32 blockSize, ma_uint32 bitsPerSample, ma_uint8 lpcOrder, ma_int32* pDecodedSamples) +{ + ma_uint8 i; + ma_uint8 lpcPrecision; + ma_int8 lpcShift; + ma_int32 coefficients[32]; + for (i = 0; i < lpcOrder; ++i) { + ma_int32 sample; + if (!ma_dr_flac__read_int32(bs, bitsPerSample, &sample)) { + return MA_FALSE; + } + pDecodedSamples[i] = sample; + } + if (!ma_dr_flac__read_uint8(bs, 4, &lpcPrecision)) { + return MA_FALSE; + } + if (lpcPrecision == 15) { + return MA_FALSE; + } + lpcPrecision += 1; + if (!ma_dr_flac__read_int8(bs, 5, &lpcShift)) { + return MA_FALSE; + } + if (lpcShift < 0) { + return MA_FALSE; + } + MA_DR_FLAC_ZERO_MEMORY(coefficients, sizeof(coefficients)); + for (i = 0; i < lpcOrder; ++i) { + if (!ma_dr_flac__read_int32(bs, lpcPrecision, coefficients + i)) { + return MA_FALSE; + } + } + if (!ma_dr_flac__decode_samples_with_residual(bs, bitsPerSample, blockSize, lpcOrder, lpcShift, lpcPrecision, coefficients, pDecodedSamples)) { + return MA_FALSE; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__read_next_flac_frame_header(ma_dr_flac_bs* bs, ma_uint8 streaminfoBitsPerSample, ma_dr_flac_frame_header* header) +{ + const ma_uint32 sampleRateTable[12] = {0, 88200, 176400, 192000, 8000, 16000, 22050, 24000, 32000, 44100, 48000, 96000}; + const ma_uint8 bitsPerSampleTable[8] = {0, 8, 12, (ma_uint8)-1, 16, 20, 24, (ma_uint8)-1}; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(header != NULL); + for (;;) { + ma_uint8 crc8 = 0xCE; + ma_uint8 reserved = 0; + ma_uint8 blockingStrategy = 0; + ma_uint8 blockSize = 0; + ma_uint8 sampleRate = 0; + ma_uint8 channelAssignment = 0; + ma_uint8 bitsPerSample = 0; + ma_bool32 isVariableBlockSize; + if (!ma_dr_flac__find_and_seek_to_next_sync_code(bs)) { + return MA_FALSE; + } + if (!ma_dr_flac__read_uint8(bs, 1, &reserved)) { + return MA_FALSE; + } + if (reserved == 1) { + continue; + } + crc8 = ma_dr_flac_crc8(crc8, reserved, 1); + if (!ma_dr_flac__read_uint8(bs, 1, &blockingStrategy)) { + return MA_FALSE; + } + crc8 = ma_dr_flac_crc8(crc8, blockingStrategy, 1); + if (!ma_dr_flac__read_uint8(bs, 4, &blockSize)) { + return MA_FALSE; + } + if (blockSize == 0) { + continue; + } + crc8 = ma_dr_flac_crc8(crc8, blockSize, 4); + if (!ma_dr_flac__read_uint8(bs, 4, &sampleRate)) { + return MA_FALSE; + } + crc8 = ma_dr_flac_crc8(crc8, sampleRate, 4); + if (!ma_dr_flac__read_uint8(bs, 4, &channelAssignment)) { + return MA_FALSE; + } + if (channelAssignment > 10) { + continue; + } + crc8 = ma_dr_flac_crc8(crc8, channelAssignment, 4); + if (!ma_dr_flac__read_uint8(bs, 3, &bitsPerSample)) { + return MA_FALSE; + } + if (bitsPerSample == 3 || bitsPerSample == 7) { + continue; + } + crc8 = ma_dr_flac_crc8(crc8, bitsPerSample, 3); + if (!ma_dr_flac__read_uint8(bs, 1, &reserved)) { + return MA_FALSE; + } + if (reserved == 1) { + continue; + } + crc8 = ma_dr_flac_crc8(crc8, reserved, 1); + isVariableBlockSize = blockingStrategy == 1; + if (isVariableBlockSize) { + ma_uint64 pcmFrameNumber; + ma_result result = ma_dr_flac__read_utf8_coded_number(bs, &pcmFrameNumber, &crc8); + if (result != MA_SUCCESS) { + if (result == MA_AT_END) { + return MA_FALSE; + } else { + continue; + } + } + header->flacFrameNumber = 0; + header->pcmFrameNumber = pcmFrameNumber; + } else { + ma_uint64 flacFrameNumber = 0; + ma_result result = ma_dr_flac__read_utf8_coded_number(bs, &flacFrameNumber, &crc8); + if (result != MA_SUCCESS) { + if (result == MA_AT_END) { + return MA_FALSE; + } else { + continue; + } + } + header->flacFrameNumber = (ma_uint32)flacFrameNumber; + header->pcmFrameNumber = 0; + } + MA_DR_FLAC_ASSERT(blockSize > 0); + if (blockSize == 1) { + header->blockSizeInPCMFrames = 192; + } else if (blockSize <= 5) { + MA_DR_FLAC_ASSERT(blockSize >= 2); + header->blockSizeInPCMFrames = 576 * (1 << (blockSize - 2)); + } else if (blockSize == 6) { + if (!ma_dr_flac__read_uint16(bs, 8, &header->blockSizeInPCMFrames)) { + return MA_FALSE; + } + crc8 = ma_dr_flac_crc8(crc8, header->blockSizeInPCMFrames, 8); + header->blockSizeInPCMFrames += 1; + } else if (blockSize == 7) { + if (!ma_dr_flac__read_uint16(bs, 16, &header->blockSizeInPCMFrames)) { + return MA_FALSE; + } + crc8 = ma_dr_flac_crc8(crc8, header->blockSizeInPCMFrames, 16); + if (header->blockSizeInPCMFrames == 0xFFFF) { + return MA_FALSE; + } + header->blockSizeInPCMFrames += 1; + } else { + MA_DR_FLAC_ASSERT(blockSize >= 8); + header->blockSizeInPCMFrames = 256 * (1 << (blockSize - 8)); + } + if (sampleRate <= 11) { + header->sampleRate = sampleRateTable[sampleRate]; + } else if (sampleRate == 12) { + if (!ma_dr_flac__read_uint32(bs, 8, &header->sampleRate)) { + return MA_FALSE; + } + crc8 = ma_dr_flac_crc8(crc8, header->sampleRate, 8); + header->sampleRate *= 1000; + } else if (sampleRate == 13) { + if (!ma_dr_flac__read_uint32(bs, 16, &header->sampleRate)) { + return MA_FALSE; + } + crc8 = ma_dr_flac_crc8(crc8, header->sampleRate, 16); + } else if (sampleRate == 14) { + if (!ma_dr_flac__read_uint32(bs, 16, &header->sampleRate)) { + return MA_FALSE; + } + crc8 = ma_dr_flac_crc8(crc8, header->sampleRate, 16); + header->sampleRate *= 10; + } else { + continue; + } + header->channelAssignment = channelAssignment; + header->bitsPerSample = bitsPerSampleTable[bitsPerSample]; + if (header->bitsPerSample == 0) { + header->bitsPerSample = streaminfoBitsPerSample; + } + if (header->bitsPerSample != streaminfoBitsPerSample) { + return MA_FALSE; + } + if (!ma_dr_flac__read_uint8(bs, 8, &header->crc8)) { + return MA_FALSE; + } +#ifndef MA_DR_FLAC_NO_CRC + if (header->crc8 != crc8) { + continue; + } +#endif + return MA_TRUE; + } +} +static ma_bool32 ma_dr_flac__read_subframe_header(ma_dr_flac_bs* bs, ma_dr_flac_subframe* pSubframe) +{ + ma_uint8 header; + int type; + if (!ma_dr_flac__read_uint8(bs, 8, &header)) { + return MA_FALSE; + } + if ((header & 0x80) != 0) { + return MA_FALSE; + } + pSubframe->lpcOrder = 0; + type = (header & 0x7E) >> 1; + if (type == 0) { + pSubframe->subframeType = MA_DR_FLAC_SUBFRAME_CONSTANT; + } else if (type == 1) { + pSubframe->subframeType = MA_DR_FLAC_SUBFRAME_VERBATIM; + } else { + if ((type & 0x20) != 0) { + pSubframe->subframeType = MA_DR_FLAC_SUBFRAME_LPC; + pSubframe->lpcOrder = (ma_uint8)(type & 0x1F) + 1; + } else if ((type & 0x08) != 0) { + pSubframe->subframeType = MA_DR_FLAC_SUBFRAME_FIXED; + pSubframe->lpcOrder = (ma_uint8)(type & 0x07); + if (pSubframe->lpcOrder > 4) { + pSubframe->subframeType = MA_DR_FLAC_SUBFRAME_RESERVED; + pSubframe->lpcOrder = 0; + } + } else { + pSubframe->subframeType = MA_DR_FLAC_SUBFRAME_RESERVED; + } + } + if (pSubframe->subframeType == MA_DR_FLAC_SUBFRAME_RESERVED) { + return MA_FALSE; + } + pSubframe->wastedBitsPerSample = 0; + if ((header & 0x01) == 1) { + unsigned int wastedBitsPerSample; + if (!ma_dr_flac__seek_past_next_set_bit(bs, &wastedBitsPerSample)) { + return MA_FALSE; + } + pSubframe->wastedBitsPerSample = (ma_uint8)wastedBitsPerSample + 1; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_subframe(ma_dr_flac_bs* bs, ma_dr_flac_frame* frame, int subframeIndex, ma_int32* pDecodedSamplesOut) +{ + ma_dr_flac_subframe* pSubframe; + ma_uint32 subframeBitsPerSample; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(frame != NULL); + pSubframe = frame->subframes + subframeIndex; + if (!ma_dr_flac__read_subframe_header(bs, pSubframe)) { + return MA_FALSE; + } + subframeBitsPerSample = frame->header.bitsPerSample; + if ((frame->header.channelAssignment == MA_DR_FLAC_CHANNEL_ASSIGNMENT_LEFT_SIDE || frame->header.channelAssignment == MA_DR_FLAC_CHANNEL_ASSIGNMENT_MID_SIDE) && subframeIndex == 1) { + subframeBitsPerSample += 1; + } else if (frame->header.channelAssignment == MA_DR_FLAC_CHANNEL_ASSIGNMENT_RIGHT_SIDE && subframeIndex == 0) { + subframeBitsPerSample += 1; + } + if (subframeBitsPerSample > 32) { + return MA_FALSE; + } + if (pSubframe->wastedBitsPerSample >= subframeBitsPerSample) { + return MA_FALSE; + } + subframeBitsPerSample -= pSubframe->wastedBitsPerSample; + pSubframe->pSamplesS32 = pDecodedSamplesOut; + if (frame->header.blockSizeInPCMFrames < pSubframe->lpcOrder) { + return MA_FALSE; + } + switch (pSubframe->subframeType) + { + case MA_DR_FLAC_SUBFRAME_CONSTANT: + { + ma_dr_flac__decode_samples__constant(bs, frame->header.blockSizeInPCMFrames, subframeBitsPerSample, pSubframe->pSamplesS32); + } break; + case MA_DR_FLAC_SUBFRAME_VERBATIM: + { + ma_dr_flac__decode_samples__verbatim(bs, frame->header.blockSizeInPCMFrames, subframeBitsPerSample, pSubframe->pSamplesS32); + } break; + case MA_DR_FLAC_SUBFRAME_FIXED: + { + ma_dr_flac__decode_samples__fixed(bs, frame->header.blockSizeInPCMFrames, subframeBitsPerSample, pSubframe->lpcOrder, pSubframe->pSamplesS32); + } break; + case MA_DR_FLAC_SUBFRAME_LPC: + { + ma_dr_flac__decode_samples__lpc(bs, frame->header.blockSizeInPCMFrames, subframeBitsPerSample, pSubframe->lpcOrder, pSubframe->pSamplesS32); + } break; + default: return MA_FALSE; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__seek_subframe(ma_dr_flac_bs* bs, ma_dr_flac_frame* frame, int subframeIndex) +{ + ma_dr_flac_subframe* pSubframe; + ma_uint32 subframeBitsPerSample; + MA_DR_FLAC_ASSERT(bs != NULL); + MA_DR_FLAC_ASSERT(frame != NULL); + pSubframe = frame->subframes + subframeIndex; + if (!ma_dr_flac__read_subframe_header(bs, pSubframe)) { + return MA_FALSE; + } + subframeBitsPerSample = frame->header.bitsPerSample; + if ((frame->header.channelAssignment == MA_DR_FLAC_CHANNEL_ASSIGNMENT_LEFT_SIDE || frame->header.channelAssignment == MA_DR_FLAC_CHANNEL_ASSIGNMENT_MID_SIDE) && subframeIndex == 1) { + subframeBitsPerSample += 1; + } else if (frame->header.channelAssignment == MA_DR_FLAC_CHANNEL_ASSIGNMENT_RIGHT_SIDE && subframeIndex == 0) { + subframeBitsPerSample += 1; + } + if (pSubframe->wastedBitsPerSample >= subframeBitsPerSample) { + return MA_FALSE; + } + subframeBitsPerSample -= pSubframe->wastedBitsPerSample; + pSubframe->pSamplesS32 = NULL; + switch (pSubframe->subframeType) + { + case MA_DR_FLAC_SUBFRAME_CONSTANT: + { + if (!ma_dr_flac__seek_bits(bs, subframeBitsPerSample)) { + return MA_FALSE; + } + } break; + case MA_DR_FLAC_SUBFRAME_VERBATIM: + { + unsigned int bitsToSeek = frame->header.blockSizeInPCMFrames * subframeBitsPerSample; + if (!ma_dr_flac__seek_bits(bs, bitsToSeek)) { + return MA_FALSE; + } + } break; + case MA_DR_FLAC_SUBFRAME_FIXED: + { + unsigned int bitsToSeek = pSubframe->lpcOrder * subframeBitsPerSample; + if (!ma_dr_flac__seek_bits(bs, bitsToSeek)) { + return MA_FALSE; + } + if (!ma_dr_flac__read_and_seek_residual(bs, frame->header.blockSizeInPCMFrames, pSubframe->lpcOrder)) { + return MA_FALSE; + } + } break; + case MA_DR_FLAC_SUBFRAME_LPC: + { + ma_uint8 lpcPrecision; + unsigned int bitsToSeek = pSubframe->lpcOrder * subframeBitsPerSample; + if (!ma_dr_flac__seek_bits(bs, bitsToSeek)) { + return MA_FALSE; + } + if (!ma_dr_flac__read_uint8(bs, 4, &lpcPrecision)) { + return MA_FALSE; + } + if (lpcPrecision == 15) { + return MA_FALSE; + } + lpcPrecision += 1; + bitsToSeek = (pSubframe->lpcOrder * lpcPrecision) + 5; + if (!ma_dr_flac__seek_bits(bs, bitsToSeek)) { + return MA_FALSE; + } + if (!ma_dr_flac__read_and_seek_residual(bs, frame->header.blockSizeInPCMFrames, pSubframe->lpcOrder)) { + return MA_FALSE; + } + } break; + default: return MA_FALSE; + } + return MA_TRUE; +} +static MA_INLINE ma_uint8 ma_dr_flac__get_channel_count_from_channel_assignment(ma_int8 channelAssignment) +{ + ma_uint8 lookup[] = {1, 2, 3, 4, 5, 6, 7, 8, 2, 2, 2}; + MA_DR_FLAC_ASSERT(channelAssignment <= 10); + return lookup[channelAssignment]; +} +static ma_result ma_dr_flac__decode_flac_frame(ma_dr_flac* pFlac) +{ + int channelCount; + int i; + ma_uint8 paddingSizeInBits; + ma_uint16 desiredCRC16; +#ifndef MA_DR_FLAC_NO_CRC + ma_uint16 actualCRC16; +#endif + MA_DR_FLAC_ZERO_MEMORY(pFlac->currentFLACFrame.subframes, sizeof(pFlac->currentFLACFrame.subframes)); + if (pFlac->currentFLACFrame.header.blockSizeInPCMFrames > pFlac->maxBlockSizeInPCMFrames) { + return MA_ERROR; + } + channelCount = ma_dr_flac__get_channel_count_from_channel_assignment(pFlac->currentFLACFrame.header.channelAssignment); + if (channelCount != (int)pFlac->channels) { + return MA_ERROR; + } + for (i = 0; i < channelCount; ++i) { + if (!ma_dr_flac__decode_subframe(&pFlac->bs, &pFlac->currentFLACFrame, i, pFlac->pDecodedSamples + (pFlac->currentFLACFrame.header.blockSizeInPCMFrames * i))) { + return MA_ERROR; + } + } + paddingSizeInBits = (ma_uint8)(MA_DR_FLAC_CACHE_L1_BITS_REMAINING(&pFlac->bs) & 7); + if (paddingSizeInBits > 0) { + ma_uint8 padding = 0; + if (!ma_dr_flac__read_uint8(&pFlac->bs, paddingSizeInBits, &padding)) { + return MA_AT_END; + } + } +#ifndef MA_DR_FLAC_NO_CRC + actualCRC16 = ma_dr_flac__flush_crc16(&pFlac->bs); +#endif + if (!ma_dr_flac__read_uint16(&pFlac->bs, 16, &desiredCRC16)) { + return MA_AT_END; + } +#ifndef MA_DR_FLAC_NO_CRC + if (actualCRC16 != desiredCRC16) { + return MA_CRC_MISMATCH; + } +#endif + pFlac->currentFLACFrame.pcmFramesRemaining = pFlac->currentFLACFrame.header.blockSizeInPCMFrames; + return MA_SUCCESS; +} +static ma_result ma_dr_flac__seek_flac_frame(ma_dr_flac* pFlac) +{ + int channelCount; + int i; + ma_uint16 desiredCRC16; +#ifndef MA_DR_FLAC_NO_CRC + ma_uint16 actualCRC16; +#endif + channelCount = ma_dr_flac__get_channel_count_from_channel_assignment(pFlac->currentFLACFrame.header.channelAssignment); + for (i = 0; i < channelCount; ++i) { + if (!ma_dr_flac__seek_subframe(&pFlac->bs, &pFlac->currentFLACFrame, i)) { + return MA_ERROR; + } + } + if (!ma_dr_flac__seek_bits(&pFlac->bs, MA_DR_FLAC_CACHE_L1_BITS_REMAINING(&pFlac->bs) & 7)) { + return MA_ERROR; + } +#ifndef MA_DR_FLAC_NO_CRC + actualCRC16 = ma_dr_flac__flush_crc16(&pFlac->bs); +#endif + if (!ma_dr_flac__read_uint16(&pFlac->bs, 16, &desiredCRC16)) { + return MA_AT_END; + } +#ifndef MA_DR_FLAC_NO_CRC + if (actualCRC16 != desiredCRC16) { + return MA_CRC_MISMATCH; + } +#endif + return MA_SUCCESS; +} +static ma_bool32 ma_dr_flac__read_and_decode_next_flac_frame(ma_dr_flac* pFlac) +{ + MA_DR_FLAC_ASSERT(pFlac != NULL); + for (;;) { + ma_result result; + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + return MA_FALSE; + } + result = ma_dr_flac__decode_flac_frame(pFlac); + if (result != MA_SUCCESS) { + if (result == MA_CRC_MISMATCH) { + continue; + } else { + return MA_FALSE; + } + } + return MA_TRUE; + } +} +static void ma_dr_flac__get_pcm_frame_range_of_current_flac_frame(ma_dr_flac* pFlac, ma_uint64* pFirstPCMFrame, ma_uint64* pLastPCMFrame) +{ + ma_uint64 firstPCMFrame; + ma_uint64 lastPCMFrame; + MA_DR_FLAC_ASSERT(pFlac != NULL); + firstPCMFrame = pFlac->currentFLACFrame.header.pcmFrameNumber; + if (firstPCMFrame == 0) { + firstPCMFrame = ((ma_uint64)pFlac->currentFLACFrame.header.flacFrameNumber) * pFlac->maxBlockSizeInPCMFrames; + } + lastPCMFrame = firstPCMFrame + pFlac->currentFLACFrame.header.blockSizeInPCMFrames; + if (lastPCMFrame > 0) { + lastPCMFrame -= 1; + } + if (pFirstPCMFrame) { + *pFirstPCMFrame = firstPCMFrame; + } + if (pLastPCMFrame) { + *pLastPCMFrame = lastPCMFrame; + } +} +static ma_bool32 ma_dr_flac__seek_to_first_frame(ma_dr_flac* pFlac) +{ + ma_bool32 result; + MA_DR_FLAC_ASSERT(pFlac != NULL); + result = ma_dr_flac__seek_to_byte(&pFlac->bs, pFlac->firstFLACFramePosInBytes); + MA_DR_FLAC_ZERO_MEMORY(&pFlac->currentFLACFrame, sizeof(pFlac->currentFLACFrame)); + pFlac->currentPCMFrame = 0; + return result; +} +static MA_INLINE ma_result ma_dr_flac__seek_to_next_flac_frame(ma_dr_flac* pFlac) +{ + MA_DR_FLAC_ASSERT(pFlac != NULL); + return ma_dr_flac__seek_flac_frame(pFlac); +} +static ma_uint64 ma_dr_flac__seek_forward_by_pcm_frames(ma_dr_flac* pFlac, ma_uint64 pcmFramesToSeek) +{ + ma_uint64 pcmFramesRead = 0; + while (pcmFramesToSeek > 0) { + if (pFlac->currentFLACFrame.pcmFramesRemaining == 0) { + if (!ma_dr_flac__read_and_decode_next_flac_frame(pFlac)) { + break; + } + } else { + if (pFlac->currentFLACFrame.pcmFramesRemaining > pcmFramesToSeek) { + pcmFramesRead += pcmFramesToSeek; + pFlac->currentFLACFrame.pcmFramesRemaining -= (ma_uint32)pcmFramesToSeek; + pcmFramesToSeek = 0; + } else { + pcmFramesRead += pFlac->currentFLACFrame.pcmFramesRemaining; + pcmFramesToSeek -= pFlac->currentFLACFrame.pcmFramesRemaining; + pFlac->currentFLACFrame.pcmFramesRemaining = 0; + } + } + } + pFlac->currentPCMFrame += pcmFramesRead; + return pcmFramesRead; +} +static ma_bool32 ma_dr_flac__seek_to_pcm_frame__brute_force(ma_dr_flac* pFlac, ma_uint64 pcmFrameIndex) +{ + ma_bool32 isMidFrame = MA_FALSE; + ma_uint64 runningPCMFrameCount; + MA_DR_FLAC_ASSERT(pFlac != NULL); + if (pcmFrameIndex >= pFlac->currentPCMFrame) { + runningPCMFrameCount = pFlac->currentPCMFrame; + if (pFlac->currentPCMFrame == 0 && pFlac->currentFLACFrame.pcmFramesRemaining == 0) { + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + return MA_FALSE; + } + } else { + isMidFrame = MA_TRUE; + } + } else { + runningPCMFrameCount = 0; + if (!ma_dr_flac__seek_to_first_frame(pFlac)) { + return MA_FALSE; + } + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + return MA_FALSE; + } + } + for (;;) { + ma_uint64 pcmFrameCountInThisFLACFrame; + ma_uint64 firstPCMFrameInFLACFrame = 0; + ma_uint64 lastPCMFrameInFLACFrame = 0; + ma_dr_flac__get_pcm_frame_range_of_current_flac_frame(pFlac, &firstPCMFrameInFLACFrame, &lastPCMFrameInFLACFrame); + pcmFrameCountInThisFLACFrame = (lastPCMFrameInFLACFrame - firstPCMFrameInFLACFrame) + 1; + if (pcmFrameIndex < (runningPCMFrameCount + pcmFrameCountInThisFLACFrame)) { + ma_uint64 pcmFramesToDecode = pcmFrameIndex - runningPCMFrameCount; + if (!isMidFrame) { + ma_result result = ma_dr_flac__decode_flac_frame(pFlac); + if (result == MA_SUCCESS) { + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, pcmFramesToDecode) == pcmFramesToDecode; + } else { + if (result == MA_CRC_MISMATCH) { + goto next_iteration; + } else { + return MA_FALSE; + } + } + } else { + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, pcmFramesToDecode) == pcmFramesToDecode; + } + } else { + if (!isMidFrame) { + ma_result result = ma_dr_flac__seek_to_next_flac_frame(pFlac); + if (result == MA_SUCCESS) { + runningPCMFrameCount += pcmFrameCountInThisFLACFrame; + } else { + if (result == MA_CRC_MISMATCH) { + goto next_iteration; + } else { + return MA_FALSE; + } + } + } else { + runningPCMFrameCount += pFlac->currentFLACFrame.pcmFramesRemaining; + pFlac->currentFLACFrame.pcmFramesRemaining = 0; + isMidFrame = MA_FALSE; + } + if (pcmFrameIndex == pFlac->totalPCMFrameCount && runningPCMFrameCount == pFlac->totalPCMFrameCount) { + return MA_TRUE; + } + } + next_iteration: + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + return MA_FALSE; + } + } +} +#if !defined(MA_DR_FLAC_NO_CRC) +#define MA_DR_FLAC_BINARY_SEARCH_APPROX_COMPRESSION_RATIO 0.6f +static ma_bool32 ma_dr_flac__seek_to_approximate_flac_frame_to_byte(ma_dr_flac* pFlac, ma_uint64 targetByte, ma_uint64 rangeLo, ma_uint64 rangeHi, ma_uint64* pLastSuccessfulSeekOffset) +{ + MA_DR_FLAC_ASSERT(pFlac != NULL); + MA_DR_FLAC_ASSERT(pLastSuccessfulSeekOffset != NULL); + MA_DR_FLAC_ASSERT(targetByte >= rangeLo); + MA_DR_FLAC_ASSERT(targetByte <= rangeHi); + *pLastSuccessfulSeekOffset = pFlac->firstFLACFramePosInBytes; + for (;;) { + ma_uint64 lastTargetByte = targetByte; + if (!ma_dr_flac__seek_to_byte(&pFlac->bs, targetByte)) { + if (targetByte == 0) { + ma_dr_flac__seek_to_first_frame(pFlac); + return MA_FALSE; + } + targetByte = rangeLo + ((rangeHi - rangeLo)/2); + rangeHi = targetByte; + } else { + MA_DR_FLAC_ZERO_MEMORY(&pFlac->currentFLACFrame, sizeof(pFlac->currentFLACFrame)); +#if 1 + if (!ma_dr_flac__read_and_decode_next_flac_frame(pFlac)) { + targetByte = rangeLo + ((rangeHi - rangeLo)/2); + rangeHi = targetByte; + } else { + break; + } +#else + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + targetByte = rangeLo + ((rangeHi - rangeLo)/2); + rangeHi = targetByte; + } else { + break; + } +#endif + } + if(targetByte == lastTargetByte) { + return MA_FALSE; + } + } + ma_dr_flac__get_pcm_frame_range_of_current_flac_frame(pFlac, &pFlac->currentPCMFrame, NULL); + MA_DR_FLAC_ASSERT(targetByte <= rangeHi); + *pLastSuccessfulSeekOffset = targetByte; + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__decode_flac_frame_and_seek_forward_by_pcm_frames(ma_dr_flac* pFlac, ma_uint64 offset) +{ +#if 0 + if (ma_dr_flac__decode_flac_frame(pFlac) != MA_SUCCESS) { + if (ma_dr_flac__read_and_decode_next_flac_frame(pFlac) == MA_FALSE) { + return MA_FALSE; + } + } +#endif + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, offset) == offset; +} +static ma_bool32 ma_dr_flac__seek_to_pcm_frame__binary_search_internal(ma_dr_flac* pFlac, ma_uint64 pcmFrameIndex, ma_uint64 byteRangeLo, ma_uint64 byteRangeHi) +{ + ma_uint64 targetByte; + ma_uint64 pcmRangeLo = pFlac->totalPCMFrameCount; + ma_uint64 pcmRangeHi = 0; + ma_uint64 lastSuccessfulSeekOffset = (ma_uint64)-1; + ma_uint64 closestSeekOffsetBeforeTargetPCMFrame = byteRangeLo; + ma_uint32 seekForwardThreshold = (pFlac->maxBlockSizeInPCMFrames != 0) ? pFlac->maxBlockSizeInPCMFrames*2 : 4096; + targetByte = byteRangeLo + (ma_uint64)(((ma_int64)((pcmFrameIndex - pFlac->currentPCMFrame) * pFlac->channels * pFlac->bitsPerSample)/8.0f) * MA_DR_FLAC_BINARY_SEARCH_APPROX_COMPRESSION_RATIO); + if (targetByte > byteRangeHi) { + targetByte = byteRangeHi; + } + for (;;) { + if (ma_dr_flac__seek_to_approximate_flac_frame_to_byte(pFlac, targetByte, byteRangeLo, byteRangeHi, &lastSuccessfulSeekOffset)) { + ma_uint64 newPCMRangeLo; + ma_uint64 newPCMRangeHi; + ma_dr_flac__get_pcm_frame_range_of_current_flac_frame(pFlac, &newPCMRangeLo, &newPCMRangeHi); + if (pcmRangeLo == newPCMRangeLo) { + if (!ma_dr_flac__seek_to_approximate_flac_frame_to_byte(pFlac, closestSeekOffsetBeforeTargetPCMFrame, closestSeekOffsetBeforeTargetPCMFrame, byteRangeHi, &lastSuccessfulSeekOffset)) { + break; + } + if (ma_dr_flac__decode_flac_frame_and_seek_forward_by_pcm_frames(pFlac, pcmFrameIndex - pFlac->currentPCMFrame)) { + return MA_TRUE; + } else { + break; + } + } + pcmRangeLo = newPCMRangeLo; + pcmRangeHi = newPCMRangeHi; + if (pcmRangeLo <= pcmFrameIndex && pcmRangeHi >= pcmFrameIndex) { + if (ma_dr_flac__decode_flac_frame_and_seek_forward_by_pcm_frames(pFlac, pcmFrameIndex - pFlac->currentPCMFrame) ) { + return MA_TRUE; + } else { + break; + } + } else { + const float approxCompressionRatio = (ma_int64)(lastSuccessfulSeekOffset - pFlac->firstFLACFramePosInBytes) / ((ma_int64)(pcmRangeLo * pFlac->channels * pFlac->bitsPerSample)/8.0f); + if (pcmRangeLo > pcmFrameIndex) { + byteRangeHi = lastSuccessfulSeekOffset; + if (byteRangeLo > byteRangeHi) { + byteRangeLo = byteRangeHi; + } + targetByte = byteRangeLo + ((byteRangeHi - byteRangeLo) / 2); + if (targetByte < byteRangeLo) { + targetByte = byteRangeLo; + } + } else { + if ((pcmFrameIndex - pcmRangeLo) < seekForwardThreshold) { + if (ma_dr_flac__decode_flac_frame_and_seek_forward_by_pcm_frames(pFlac, pcmFrameIndex - pFlac->currentPCMFrame)) { + return MA_TRUE; + } else { + break; + } + } else { + byteRangeLo = lastSuccessfulSeekOffset; + if (byteRangeHi < byteRangeLo) { + byteRangeHi = byteRangeLo; + } + targetByte = lastSuccessfulSeekOffset + (ma_uint64)(((ma_int64)((pcmFrameIndex-pcmRangeLo) * pFlac->channels * pFlac->bitsPerSample)/8.0f) * approxCompressionRatio); + if (targetByte > byteRangeHi) { + targetByte = byteRangeHi; + } + if (closestSeekOffsetBeforeTargetPCMFrame < lastSuccessfulSeekOffset) { + closestSeekOffsetBeforeTargetPCMFrame = lastSuccessfulSeekOffset; + } + } + } + } + } else { + break; + } + } + ma_dr_flac__seek_to_first_frame(pFlac); + return MA_FALSE; +} +static ma_bool32 ma_dr_flac__seek_to_pcm_frame__binary_search(ma_dr_flac* pFlac, ma_uint64 pcmFrameIndex) +{ + ma_uint64 byteRangeLo; + ma_uint64 byteRangeHi; + ma_uint32 seekForwardThreshold = (pFlac->maxBlockSizeInPCMFrames != 0) ? pFlac->maxBlockSizeInPCMFrames*2 : 4096; + if (ma_dr_flac__seek_to_first_frame(pFlac) == MA_FALSE) { + return MA_FALSE; + } + if (pcmFrameIndex < seekForwardThreshold) { + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, pcmFrameIndex) == pcmFrameIndex; + } + byteRangeLo = pFlac->firstFLACFramePosInBytes; + byteRangeHi = pFlac->firstFLACFramePosInBytes + (ma_uint64)((ma_int64)(pFlac->totalPCMFrameCount * pFlac->channels * pFlac->bitsPerSample)/8.0f); + return ma_dr_flac__seek_to_pcm_frame__binary_search_internal(pFlac, pcmFrameIndex, byteRangeLo, byteRangeHi); +} +#endif +static ma_bool32 ma_dr_flac__seek_to_pcm_frame__seek_table(ma_dr_flac* pFlac, ma_uint64 pcmFrameIndex) +{ + ma_uint32 iClosestSeekpoint = 0; + ma_bool32 isMidFrame = MA_FALSE; + ma_uint64 runningPCMFrameCount; + ma_uint32 iSeekpoint; + MA_DR_FLAC_ASSERT(pFlac != NULL); + if (pFlac->pSeekpoints == NULL || pFlac->seekpointCount == 0) { + return MA_FALSE; + } + if (pFlac->pSeekpoints[0].firstPCMFrame > pcmFrameIndex) { + return MA_FALSE; + } + for (iSeekpoint = 0; iSeekpoint < pFlac->seekpointCount; ++iSeekpoint) { + if (pFlac->pSeekpoints[iSeekpoint].firstPCMFrame >= pcmFrameIndex) { + break; + } + iClosestSeekpoint = iSeekpoint; + } + if (pFlac->pSeekpoints[iClosestSeekpoint].pcmFrameCount == 0 || pFlac->pSeekpoints[iClosestSeekpoint].pcmFrameCount > pFlac->maxBlockSizeInPCMFrames) { + return MA_FALSE; + } + if (pFlac->pSeekpoints[iClosestSeekpoint].firstPCMFrame > pFlac->totalPCMFrameCount && pFlac->totalPCMFrameCount > 0) { + return MA_FALSE; + } +#if !defined(MA_DR_FLAC_NO_CRC) + if (pFlac->totalPCMFrameCount > 0) { + ma_uint64 byteRangeLo; + ma_uint64 byteRangeHi; + byteRangeHi = pFlac->firstFLACFramePosInBytes + (ma_uint64)((ma_int64)(pFlac->totalPCMFrameCount * pFlac->channels * pFlac->bitsPerSample)/8.0f); + byteRangeLo = pFlac->firstFLACFramePosInBytes + pFlac->pSeekpoints[iClosestSeekpoint].flacFrameOffset; + if (iClosestSeekpoint < pFlac->seekpointCount-1) { + ma_uint32 iNextSeekpoint = iClosestSeekpoint + 1; + if (pFlac->pSeekpoints[iClosestSeekpoint].flacFrameOffset >= pFlac->pSeekpoints[iNextSeekpoint].flacFrameOffset || pFlac->pSeekpoints[iNextSeekpoint].pcmFrameCount == 0) { + return MA_FALSE; + } + if (pFlac->pSeekpoints[iNextSeekpoint].firstPCMFrame != (((ma_uint64)0xFFFFFFFF << 32) | 0xFFFFFFFF)) { + byteRangeHi = pFlac->firstFLACFramePosInBytes + pFlac->pSeekpoints[iNextSeekpoint].flacFrameOffset - 1; + } + } + if (ma_dr_flac__seek_to_byte(&pFlac->bs, pFlac->firstFLACFramePosInBytes + pFlac->pSeekpoints[iClosestSeekpoint].flacFrameOffset)) { + if (ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + ma_dr_flac__get_pcm_frame_range_of_current_flac_frame(pFlac, &pFlac->currentPCMFrame, NULL); + if (ma_dr_flac__seek_to_pcm_frame__binary_search_internal(pFlac, pcmFrameIndex, byteRangeLo, byteRangeHi)) { + return MA_TRUE; + } + } + } + } +#endif + if (pcmFrameIndex >= pFlac->currentPCMFrame && pFlac->pSeekpoints[iClosestSeekpoint].firstPCMFrame <= pFlac->currentPCMFrame) { + runningPCMFrameCount = pFlac->currentPCMFrame; + if (pFlac->currentPCMFrame == 0 && pFlac->currentFLACFrame.pcmFramesRemaining == 0) { + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + return MA_FALSE; + } + } else { + isMidFrame = MA_TRUE; + } + } else { + runningPCMFrameCount = pFlac->pSeekpoints[iClosestSeekpoint].firstPCMFrame; + if (!ma_dr_flac__seek_to_byte(&pFlac->bs, pFlac->firstFLACFramePosInBytes + pFlac->pSeekpoints[iClosestSeekpoint].flacFrameOffset)) { + return MA_FALSE; + } + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + return MA_FALSE; + } + } + for (;;) { + ma_uint64 pcmFrameCountInThisFLACFrame; + ma_uint64 firstPCMFrameInFLACFrame = 0; + ma_uint64 lastPCMFrameInFLACFrame = 0; + ma_dr_flac__get_pcm_frame_range_of_current_flac_frame(pFlac, &firstPCMFrameInFLACFrame, &lastPCMFrameInFLACFrame); + pcmFrameCountInThisFLACFrame = (lastPCMFrameInFLACFrame - firstPCMFrameInFLACFrame) + 1; + if (pcmFrameIndex < (runningPCMFrameCount + pcmFrameCountInThisFLACFrame)) { + ma_uint64 pcmFramesToDecode = pcmFrameIndex - runningPCMFrameCount; + if (!isMidFrame) { + ma_result result = ma_dr_flac__decode_flac_frame(pFlac); + if (result == MA_SUCCESS) { + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, pcmFramesToDecode) == pcmFramesToDecode; + } else { + if (result == MA_CRC_MISMATCH) { + goto next_iteration; + } else { + return MA_FALSE; + } + } + } else { + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, pcmFramesToDecode) == pcmFramesToDecode; + } + } else { + if (!isMidFrame) { + ma_result result = ma_dr_flac__seek_to_next_flac_frame(pFlac); + if (result == MA_SUCCESS) { + runningPCMFrameCount += pcmFrameCountInThisFLACFrame; + } else { + if (result == MA_CRC_MISMATCH) { + goto next_iteration; + } else { + return MA_FALSE; + } + } + } else { + runningPCMFrameCount += pFlac->currentFLACFrame.pcmFramesRemaining; + pFlac->currentFLACFrame.pcmFramesRemaining = 0; + isMidFrame = MA_FALSE; + } + if (pcmFrameIndex == pFlac->totalPCMFrameCount && runningPCMFrameCount == pFlac->totalPCMFrameCount) { + return MA_TRUE; + } + } + next_iteration: + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + return MA_FALSE; + } + } +} +#ifndef MA_DR_FLAC_NO_OGG +typedef struct +{ + ma_uint8 capturePattern[4]; + ma_uint8 structureVersion; + ma_uint8 headerType; + ma_uint64 granulePosition; + ma_uint32 serialNumber; + ma_uint32 sequenceNumber; + ma_uint32 checksum; + ma_uint8 segmentCount; + ma_uint8 segmentTable[255]; +} ma_dr_flac_ogg_page_header; +#endif +typedef struct +{ + ma_dr_flac_read_proc onRead; + ma_dr_flac_seek_proc onSeek; + ma_dr_flac_meta_proc onMeta; + ma_dr_flac_container container; + void* pUserData; + void* pUserDataMD; + ma_uint32 sampleRate; + ma_uint8 channels; + ma_uint8 bitsPerSample; + ma_uint64 totalPCMFrameCount; + ma_uint16 maxBlockSizeInPCMFrames; + ma_uint64 runningFilePos; + ma_bool32 hasStreamInfoBlock; + ma_bool32 hasMetadataBlocks; + ma_dr_flac_bs bs; + ma_dr_flac_frame_header firstFrameHeader; +#ifndef MA_DR_FLAC_NO_OGG + ma_uint32 oggSerial; + ma_uint64 oggFirstBytePos; + ma_dr_flac_ogg_page_header oggBosHeader; +#endif +} ma_dr_flac_init_info; +static MA_INLINE void ma_dr_flac__decode_block_header(ma_uint32 blockHeader, ma_uint8* isLastBlock, ma_uint8* blockType, ma_uint32* blockSize) +{ + blockHeader = ma_dr_flac__be2host_32(blockHeader); + *isLastBlock = (ma_uint8)((blockHeader & 0x80000000UL) >> 31); + *blockType = (ma_uint8)((blockHeader & 0x7F000000UL) >> 24); + *blockSize = (blockHeader & 0x00FFFFFFUL); +} +static MA_INLINE ma_bool32 ma_dr_flac__read_and_decode_block_header(ma_dr_flac_read_proc onRead, void* pUserData, ma_uint8* isLastBlock, ma_uint8* blockType, ma_uint32* blockSize) +{ + ma_uint32 blockHeader; + *blockSize = 0; + if (onRead(pUserData, &blockHeader, 4) != 4) { + return MA_FALSE; + } + ma_dr_flac__decode_block_header(blockHeader, isLastBlock, blockType, blockSize); + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__read_streaminfo(ma_dr_flac_read_proc onRead, void* pUserData, ma_dr_flac_streaminfo* pStreamInfo) +{ + ma_uint32 blockSizes; + ma_uint64 frameSizes = 0; + ma_uint64 importantProps; + ma_uint8 md5[16]; + if (onRead(pUserData, &blockSizes, 4) != 4) { + return MA_FALSE; + } + if (onRead(pUserData, &frameSizes, 6) != 6) { + return MA_FALSE; + } + if (onRead(pUserData, &importantProps, 8) != 8) { + return MA_FALSE; + } + if (onRead(pUserData, md5, sizeof(md5)) != sizeof(md5)) { + return MA_FALSE; + } + blockSizes = ma_dr_flac__be2host_32(blockSizes); + frameSizes = ma_dr_flac__be2host_64(frameSizes); + importantProps = ma_dr_flac__be2host_64(importantProps); + pStreamInfo->minBlockSizeInPCMFrames = (ma_uint16)((blockSizes & 0xFFFF0000) >> 16); + pStreamInfo->maxBlockSizeInPCMFrames = (ma_uint16) (blockSizes & 0x0000FFFF); + pStreamInfo->minFrameSizeInPCMFrames = (ma_uint32)((frameSizes & (((ma_uint64)0x00FFFFFF << 16) << 24)) >> 40); + pStreamInfo->maxFrameSizeInPCMFrames = (ma_uint32)((frameSizes & (((ma_uint64)0x00FFFFFF << 16) << 0)) >> 16); + pStreamInfo->sampleRate = (ma_uint32)((importantProps & (((ma_uint64)0x000FFFFF << 16) << 28)) >> 44); + pStreamInfo->channels = (ma_uint8 )((importantProps & (((ma_uint64)0x0000000E << 16) << 24)) >> 41) + 1; + pStreamInfo->bitsPerSample = (ma_uint8 )((importantProps & (((ma_uint64)0x0000001F << 16) << 20)) >> 36) + 1; + pStreamInfo->totalPCMFrameCount = ((importantProps & ((((ma_uint64)0x0000000F << 16) << 16) | 0xFFFFFFFF))); + MA_DR_FLAC_COPY_MEMORY(pStreamInfo->md5, md5, sizeof(md5)); + return MA_TRUE; +} +static void* ma_dr_flac__malloc_default(size_t sz, void* pUserData) +{ + (void)pUserData; + return MA_DR_FLAC_MALLOC(sz); +} +static void* ma_dr_flac__realloc_default(void* p, size_t sz, void* pUserData) +{ + (void)pUserData; + return MA_DR_FLAC_REALLOC(p, sz); +} +static void ma_dr_flac__free_default(void* p, void* pUserData) +{ + (void)pUserData; + MA_DR_FLAC_FREE(p); +} +static void* ma_dr_flac__malloc_from_callbacks(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks == NULL) { + return NULL; + } + if (pAllocationCallbacks->onMalloc != NULL) { + return pAllocationCallbacks->onMalloc(sz, pAllocationCallbacks->pUserData); + } + if (pAllocationCallbacks->onRealloc != NULL) { + return pAllocationCallbacks->onRealloc(NULL, sz, pAllocationCallbacks->pUserData); + } + return NULL; +} +static void* ma_dr_flac__realloc_from_callbacks(void* p, size_t szNew, size_t szOld, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks == NULL) { + return NULL; + } + if (pAllocationCallbacks->onRealloc != NULL) { + return pAllocationCallbacks->onRealloc(p, szNew, pAllocationCallbacks->pUserData); + } + if (pAllocationCallbacks->onMalloc != NULL && pAllocationCallbacks->onFree != NULL) { + void* p2; + p2 = pAllocationCallbacks->onMalloc(szNew, pAllocationCallbacks->pUserData); + if (p2 == NULL) { + return NULL; + } + if (p != NULL) { + MA_DR_FLAC_COPY_MEMORY(p2, p, szOld); + pAllocationCallbacks->onFree(p, pAllocationCallbacks->pUserData); + } + return p2; + } + return NULL; +} +static void ma_dr_flac__free_from_callbacks(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (p == NULL || pAllocationCallbacks == NULL) { + return; + } + if (pAllocationCallbacks->onFree != NULL) { + pAllocationCallbacks->onFree(p, pAllocationCallbacks->pUserData); + } +} +static ma_bool32 ma_dr_flac__read_and_decode_metadata(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, void* pUserData, void* pUserDataMD, ma_uint64* pFirstFramePos, ma_uint64* pSeektablePos, ma_uint32* pSeekpointCount, ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_uint64 runningFilePos = 42; + ma_uint64 seektablePos = 0; + ma_uint32 seektableSize = 0; + for (;;) { + ma_dr_flac_metadata metadata; + ma_uint8 isLastBlock = 0; + ma_uint8 blockType = 0; + ma_uint32 blockSize; + if (ma_dr_flac__read_and_decode_block_header(onRead, pUserData, &isLastBlock, &blockType, &blockSize) == MA_FALSE) { + return MA_FALSE; + } + runningFilePos += 4; + metadata.type = blockType; + metadata.pRawData = NULL; + metadata.rawDataSize = 0; + switch (blockType) + { + case MA_DR_FLAC_METADATA_BLOCK_TYPE_APPLICATION: + { + if (blockSize < 4) { + return MA_FALSE; + } + if (onMeta) { + void* pRawData = ma_dr_flac__malloc_from_callbacks(blockSize, pAllocationCallbacks); + if (pRawData == NULL) { + return MA_FALSE; + } + if (onRead(pUserData, pRawData, blockSize) != blockSize) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.pRawData = pRawData; + metadata.rawDataSize = blockSize; + metadata.data.application.id = ma_dr_flac__be2host_32(*(ma_uint32*)pRawData); + metadata.data.application.pData = (const void*)((ma_uint8*)pRawData + sizeof(ma_uint32)); + metadata.data.application.dataSize = blockSize - sizeof(ma_uint32); + onMeta(pUserDataMD, &metadata); + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + } + } break; + case MA_DR_FLAC_METADATA_BLOCK_TYPE_SEEKTABLE: + { + seektablePos = runningFilePos; + seektableSize = blockSize; + if (onMeta) { + ma_uint32 seekpointCount; + ma_uint32 iSeekpoint; + void* pRawData; + seekpointCount = blockSize/MA_DR_FLAC_SEEKPOINT_SIZE_IN_BYTES; + pRawData = ma_dr_flac__malloc_from_callbacks(seekpointCount * sizeof(ma_dr_flac_seekpoint), pAllocationCallbacks); + if (pRawData == NULL) { + return MA_FALSE; + } + for (iSeekpoint = 0; iSeekpoint < seekpointCount; ++iSeekpoint) { + ma_dr_flac_seekpoint* pSeekpoint = (ma_dr_flac_seekpoint*)pRawData + iSeekpoint; + if (onRead(pUserData, pSeekpoint, MA_DR_FLAC_SEEKPOINT_SIZE_IN_BYTES) != MA_DR_FLAC_SEEKPOINT_SIZE_IN_BYTES) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + pSeekpoint->firstPCMFrame = ma_dr_flac__be2host_64(pSeekpoint->firstPCMFrame); + pSeekpoint->flacFrameOffset = ma_dr_flac__be2host_64(pSeekpoint->flacFrameOffset); + pSeekpoint->pcmFrameCount = ma_dr_flac__be2host_16(pSeekpoint->pcmFrameCount); + } + metadata.pRawData = pRawData; + metadata.rawDataSize = blockSize; + metadata.data.seektable.seekpointCount = seekpointCount; + metadata.data.seektable.pSeekpoints = (const ma_dr_flac_seekpoint*)pRawData; + onMeta(pUserDataMD, &metadata); + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + } + } break; + case MA_DR_FLAC_METADATA_BLOCK_TYPE_VORBIS_COMMENT: + { + if (blockSize < 8) { + return MA_FALSE; + } + if (onMeta) { + void* pRawData; + const char* pRunningData; + const char* pRunningDataEnd; + ma_uint32 i; + pRawData = ma_dr_flac__malloc_from_callbacks(blockSize, pAllocationCallbacks); + if (pRawData == NULL) { + return MA_FALSE; + } + if (onRead(pUserData, pRawData, blockSize) != blockSize) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.pRawData = pRawData; + metadata.rawDataSize = blockSize; + pRunningData = (const char*)pRawData; + pRunningDataEnd = (const char*)pRawData + blockSize; + metadata.data.vorbis_comment.vendorLength = ma_dr_flac__le2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + if ((pRunningDataEnd - pRunningData) - 4 < (ma_int64)metadata.data.vorbis_comment.vendorLength) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.data.vorbis_comment.vendor = pRunningData; pRunningData += metadata.data.vorbis_comment.vendorLength; + metadata.data.vorbis_comment.commentCount = ma_dr_flac__le2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + if ((pRunningDataEnd - pRunningData) / sizeof(ma_uint32) < metadata.data.vorbis_comment.commentCount) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.data.vorbis_comment.pComments = pRunningData; + for (i = 0; i < metadata.data.vorbis_comment.commentCount; ++i) { + ma_uint32 commentLength; + if (pRunningDataEnd - pRunningData < 4) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + commentLength = ma_dr_flac__le2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + if (pRunningDataEnd - pRunningData < (ma_int64)commentLength) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + pRunningData += commentLength; + } + onMeta(pUserDataMD, &metadata); + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + } + } break; + case MA_DR_FLAC_METADATA_BLOCK_TYPE_CUESHEET: + { + if (blockSize < 396) { + return MA_FALSE; + } + if (onMeta) { + void* pRawData; + const char* pRunningData; + const char* pRunningDataEnd; + size_t bufferSize; + ma_uint8 iTrack; + ma_uint8 iIndex; + void* pTrackData; + pRawData = ma_dr_flac__malloc_from_callbacks(blockSize, pAllocationCallbacks); + if (pRawData == NULL) { + return MA_FALSE; + } + if (onRead(pUserData, pRawData, blockSize) != blockSize) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.pRawData = pRawData; + metadata.rawDataSize = blockSize; + pRunningData = (const char*)pRawData; + pRunningDataEnd = (const char*)pRawData + blockSize; + MA_DR_FLAC_COPY_MEMORY(metadata.data.cuesheet.catalog, pRunningData, 128); pRunningData += 128; + metadata.data.cuesheet.leadInSampleCount = ma_dr_flac__be2host_64(*(const ma_uint64*)pRunningData); pRunningData += 8; + metadata.data.cuesheet.isCD = (pRunningData[0] & 0x80) != 0; pRunningData += 259; + metadata.data.cuesheet.trackCount = pRunningData[0]; pRunningData += 1; + metadata.data.cuesheet.pTrackData = NULL; + { + const char* pRunningDataSaved = pRunningData; + bufferSize = metadata.data.cuesheet.trackCount * MA_DR_FLAC_CUESHEET_TRACK_SIZE_IN_BYTES; + for (iTrack = 0; iTrack < metadata.data.cuesheet.trackCount; ++iTrack) { + ma_uint8 indexCount; + ma_uint32 indexPointSize; + if (pRunningDataEnd - pRunningData < MA_DR_FLAC_CUESHEET_TRACK_SIZE_IN_BYTES) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + pRunningData += 35; + indexCount = pRunningData[0]; + pRunningData += 1; + bufferSize += indexCount * sizeof(ma_dr_flac_cuesheet_track_index); + indexPointSize = indexCount * MA_DR_FLAC_CUESHEET_TRACK_INDEX_SIZE_IN_BYTES; + if (pRunningDataEnd - pRunningData < (ma_int64)indexPointSize) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + pRunningData += indexPointSize; + } + pRunningData = pRunningDataSaved; + } + { + char* pRunningTrackData; + pTrackData = ma_dr_flac__malloc_from_callbacks(bufferSize, pAllocationCallbacks); + if (pTrackData == NULL) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + pRunningTrackData = (char*)pTrackData; + for (iTrack = 0; iTrack < metadata.data.cuesheet.trackCount; ++iTrack) { + ma_uint8 indexCount; + MA_DR_FLAC_COPY_MEMORY(pRunningTrackData, pRunningData, MA_DR_FLAC_CUESHEET_TRACK_SIZE_IN_BYTES); + pRunningData += MA_DR_FLAC_CUESHEET_TRACK_SIZE_IN_BYTES-1; + pRunningTrackData += MA_DR_FLAC_CUESHEET_TRACK_SIZE_IN_BYTES-1; + indexCount = pRunningData[0]; + pRunningData += 1; + pRunningTrackData += 1; + for (iIndex = 0; iIndex < indexCount; ++iIndex) { + ma_dr_flac_cuesheet_track_index* pTrackIndex = (ma_dr_flac_cuesheet_track_index*)pRunningTrackData; + MA_DR_FLAC_COPY_MEMORY(pRunningTrackData, pRunningData, MA_DR_FLAC_CUESHEET_TRACK_INDEX_SIZE_IN_BYTES); + pRunningData += MA_DR_FLAC_CUESHEET_TRACK_INDEX_SIZE_IN_BYTES; + pRunningTrackData += sizeof(ma_dr_flac_cuesheet_track_index); + pTrackIndex->offset = ma_dr_flac__be2host_64(pTrackIndex->offset); + } + } + metadata.data.cuesheet.pTrackData = pTrackData; + } + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + pRawData = NULL; + onMeta(pUserDataMD, &metadata); + ma_dr_flac__free_from_callbacks(pTrackData, pAllocationCallbacks); + pTrackData = NULL; + } + } break; + case MA_DR_FLAC_METADATA_BLOCK_TYPE_PICTURE: + { + if (blockSize < 32) { + return MA_FALSE; + } + if (onMeta) { + void* pRawData; + const char* pRunningData; + const char* pRunningDataEnd; + pRawData = ma_dr_flac__malloc_from_callbacks(blockSize, pAllocationCallbacks); + if (pRawData == NULL) { + return MA_FALSE; + } + if (onRead(pUserData, pRawData, blockSize) != blockSize) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.pRawData = pRawData; + metadata.rawDataSize = blockSize; + pRunningData = (const char*)pRawData; + pRunningDataEnd = (const char*)pRawData + blockSize; + metadata.data.picture.type = ma_dr_flac__be2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + metadata.data.picture.mimeLength = ma_dr_flac__be2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + if ((pRunningDataEnd - pRunningData) - 24 < (ma_int64)metadata.data.picture.mimeLength) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.data.picture.mime = pRunningData; pRunningData += metadata.data.picture.mimeLength; + metadata.data.picture.descriptionLength = ma_dr_flac__be2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + if ((pRunningDataEnd - pRunningData) - 20 < (ma_int64)metadata.data.picture.descriptionLength) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.data.picture.description = pRunningData; pRunningData += metadata.data.picture.descriptionLength; + metadata.data.picture.width = ma_dr_flac__be2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + metadata.data.picture.height = ma_dr_flac__be2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + metadata.data.picture.colorDepth = ma_dr_flac__be2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + metadata.data.picture.indexColorCount = ma_dr_flac__be2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + metadata.data.picture.pictureDataSize = ma_dr_flac__be2host_32_ptr_unaligned(pRunningData); pRunningData += 4; + metadata.data.picture.pPictureData = (const ma_uint8*)pRunningData; + if (pRunningDataEnd - pRunningData < (ma_int64)metadata.data.picture.pictureDataSize) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + onMeta(pUserDataMD, &metadata); + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + } + } break; + case MA_DR_FLAC_METADATA_BLOCK_TYPE_PADDING: + { + if (onMeta) { + metadata.data.padding.unused = 0; + if (!onSeek(pUserData, blockSize, ma_dr_flac_seek_origin_current)) { + isLastBlock = MA_TRUE; + } else { + onMeta(pUserDataMD, &metadata); + } + } + } break; + case MA_DR_FLAC_METADATA_BLOCK_TYPE_INVALID: + { + if (onMeta) { + if (!onSeek(pUserData, blockSize, ma_dr_flac_seek_origin_current)) { + isLastBlock = MA_TRUE; + } + } + } break; + default: + { + if (onMeta) { + void* pRawData = ma_dr_flac__malloc_from_callbacks(blockSize, pAllocationCallbacks); + if (pRawData == NULL) { + return MA_FALSE; + } + if (onRead(pUserData, pRawData, blockSize) != blockSize) { + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + return MA_FALSE; + } + metadata.pRawData = pRawData; + metadata.rawDataSize = blockSize; + onMeta(pUserDataMD, &metadata); + ma_dr_flac__free_from_callbacks(pRawData, pAllocationCallbacks); + } + } break; + } + if (onMeta == NULL && blockSize > 0) { + if (!onSeek(pUserData, blockSize, ma_dr_flac_seek_origin_current)) { + isLastBlock = MA_TRUE; + } + } + runningFilePos += blockSize; + if (isLastBlock) { + break; + } + } + *pSeektablePos = seektablePos; + *pSeekpointCount = seektableSize / MA_DR_FLAC_SEEKPOINT_SIZE_IN_BYTES; + *pFirstFramePos = runningFilePos; + return MA_TRUE; +} +static ma_bool32 ma_dr_flac__init_private__native(ma_dr_flac_init_info* pInit, ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, void* pUserData, void* pUserDataMD, ma_bool32 relaxed) +{ + ma_uint8 isLastBlock; + ma_uint8 blockType; + ma_uint32 blockSize; + (void)onSeek; + pInit->container = ma_dr_flac_container_native; + if (!ma_dr_flac__read_and_decode_block_header(onRead, pUserData, &isLastBlock, &blockType, &blockSize)) { + return MA_FALSE; + } + if (blockType != MA_DR_FLAC_METADATA_BLOCK_TYPE_STREAMINFO || blockSize != 34) { + if (!relaxed) { + return MA_FALSE; + } else { + pInit->hasStreamInfoBlock = MA_FALSE; + pInit->hasMetadataBlocks = MA_FALSE; + if (!ma_dr_flac__read_next_flac_frame_header(&pInit->bs, 0, &pInit->firstFrameHeader)) { + return MA_FALSE; + } + if (pInit->firstFrameHeader.bitsPerSample == 0) { + return MA_FALSE; + } + pInit->sampleRate = pInit->firstFrameHeader.sampleRate; + pInit->channels = ma_dr_flac__get_channel_count_from_channel_assignment(pInit->firstFrameHeader.channelAssignment); + pInit->bitsPerSample = pInit->firstFrameHeader.bitsPerSample; + pInit->maxBlockSizeInPCMFrames = 65535; + return MA_TRUE; + } + } else { + ma_dr_flac_streaminfo streaminfo; + if (!ma_dr_flac__read_streaminfo(onRead, pUserData, &streaminfo)) { + return MA_FALSE; + } + pInit->hasStreamInfoBlock = MA_TRUE; + pInit->sampleRate = streaminfo.sampleRate; + pInit->channels = streaminfo.channels; + pInit->bitsPerSample = streaminfo.bitsPerSample; + pInit->totalPCMFrameCount = streaminfo.totalPCMFrameCount; + pInit->maxBlockSizeInPCMFrames = streaminfo.maxBlockSizeInPCMFrames; + pInit->hasMetadataBlocks = !isLastBlock; + if (onMeta) { + ma_dr_flac_metadata metadata; + metadata.type = MA_DR_FLAC_METADATA_BLOCK_TYPE_STREAMINFO; + metadata.pRawData = NULL; + metadata.rawDataSize = 0; + metadata.data.streaminfo = streaminfo; + onMeta(pUserDataMD, &metadata); + } + return MA_TRUE; + } +} +#ifndef MA_DR_FLAC_NO_OGG +#define MA_DR_FLAC_OGG_MAX_PAGE_SIZE 65307 +#define MA_DR_FLAC_OGG_CAPTURE_PATTERN_CRC32 1605413199 +typedef enum +{ + ma_dr_flac_ogg_recover_on_crc_mismatch, + ma_dr_flac_ogg_fail_on_crc_mismatch +} ma_dr_flac_ogg_crc_mismatch_recovery; +#ifndef MA_DR_FLAC_NO_CRC +static ma_uint32 ma_dr_flac__crc32_table[] = { + 0x00000000L, 0x04C11DB7L, 0x09823B6EL, 0x0D4326D9L, + 0x130476DCL, 0x17C56B6BL, 0x1A864DB2L, 0x1E475005L, + 0x2608EDB8L, 0x22C9F00FL, 0x2F8AD6D6L, 0x2B4BCB61L, + 0x350C9B64L, 0x31CD86D3L, 0x3C8EA00AL, 0x384FBDBDL, + 0x4C11DB70L, 0x48D0C6C7L, 0x4593E01EL, 0x4152FDA9L, + 0x5F15ADACL, 0x5BD4B01BL, 0x569796C2L, 0x52568B75L, + 0x6A1936C8L, 0x6ED82B7FL, 0x639B0DA6L, 0x675A1011L, + 0x791D4014L, 0x7DDC5DA3L, 0x709F7B7AL, 0x745E66CDL, + 0x9823B6E0L, 0x9CE2AB57L, 0x91A18D8EL, 0x95609039L, + 0x8B27C03CL, 0x8FE6DD8BL, 0x82A5FB52L, 0x8664E6E5L, + 0xBE2B5B58L, 0xBAEA46EFL, 0xB7A96036L, 0xB3687D81L, + 0xAD2F2D84L, 0xA9EE3033L, 0xA4AD16EAL, 0xA06C0B5DL, + 0xD4326D90L, 0xD0F37027L, 0xDDB056FEL, 0xD9714B49L, + 0xC7361B4CL, 0xC3F706FBL, 0xCEB42022L, 0xCA753D95L, + 0xF23A8028L, 0xF6FB9D9FL, 0xFBB8BB46L, 0xFF79A6F1L, + 0xE13EF6F4L, 0xE5FFEB43L, 0xE8BCCD9AL, 0xEC7DD02DL, + 0x34867077L, 0x30476DC0L, 0x3D044B19L, 0x39C556AEL, + 0x278206ABL, 0x23431B1CL, 0x2E003DC5L, 0x2AC12072L, + 0x128E9DCFL, 0x164F8078L, 0x1B0CA6A1L, 0x1FCDBB16L, + 0x018AEB13L, 0x054BF6A4L, 0x0808D07DL, 0x0CC9CDCAL, + 0x7897AB07L, 0x7C56B6B0L, 0x71159069L, 0x75D48DDEL, + 0x6B93DDDBL, 0x6F52C06CL, 0x6211E6B5L, 0x66D0FB02L, + 0x5E9F46BFL, 0x5A5E5B08L, 0x571D7DD1L, 0x53DC6066L, + 0x4D9B3063L, 0x495A2DD4L, 0x44190B0DL, 0x40D816BAL, + 0xACA5C697L, 0xA864DB20L, 0xA527FDF9L, 0xA1E6E04EL, + 0xBFA1B04BL, 0xBB60ADFCL, 0xB6238B25L, 0xB2E29692L, + 0x8AAD2B2FL, 0x8E6C3698L, 0x832F1041L, 0x87EE0DF6L, + 0x99A95DF3L, 0x9D684044L, 0x902B669DL, 0x94EA7B2AL, + 0xE0B41DE7L, 0xE4750050L, 0xE9362689L, 0xEDF73B3EL, + 0xF3B06B3BL, 0xF771768CL, 0xFA325055L, 0xFEF34DE2L, + 0xC6BCF05FL, 0xC27DEDE8L, 0xCF3ECB31L, 0xCBFFD686L, + 0xD5B88683L, 0xD1799B34L, 0xDC3ABDEDL, 0xD8FBA05AL, + 0x690CE0EEL, 0x6DCDFD59L, 0x608EDB80L, 0x644FC637L, + 0x7A089632L, 0x7EC98B85L, 0x738AAD5CL, 0x774BB0EBL, + 0x4F040D56L, 0x4BC510E1L, 0x46863638L, 0x42472B8FL, + 0x5C007B8AL, 0x58C1663DL, 0x558240E4L, 0x51435D53L, + 0x251D3B9EL, 0x21DC2629L, 0x2C9F00F0L, 0x285E1D47L, + 0x36194D42L, 0x32D850F5L, 0x3F9B762CL, 0x3B5A6B9BL, + 0x0315D626L, 0x07D4CB91L, 0x0A97ED48L, 0x0E56F0FFL, + 0x1011A0FAL, 0x14D0BD4DL, 0x19939B94L, 0x1D528623L, + 0xF12F560EL, 0xF5EE4BB9L, 0xF8AD6D60L, 0xFC6C70D7L, + 0xE22B20D2L, 0xE6EA3D65L, 0xEBA91BBCL, 0xEF68060BL, + 0xD727BBB6L, 0xD3E6A601L, 0xDEA580D8L, 0xDA649D6FL, + 0xC423CD6AL, 0xC0E2D0DDL, 0xCDA1F604L, 0xC960EBB3L, + 0xBD3E8D7EL, 0xB9FF90C9L, 0xB4BCB610L, 0xB07DABA7L, + 0xAE3AFBA2L, 0xAAFBE615L, 0xA7B8C0CCL, 0xA379DD7BL, + 0x9B3660C6L, 0x9FF77D71L, 0x92B45BA8L, 0x9675461FL, + 0x8832161AL, 0x8CF30BADL, 0x81B02D74L, 0x857130C3L, + 0x5D8A9099L, 0x594B8D2EL, 0x5408ABF7L, 0x50C9B640L, + 0x4E8EE645L, 0x4A4FFBF2L, 0x470CDD2BL, 0x43CDC09CL, + 0x7B827D21L, 0x7F436096L, 0x7200464FL, 0x76C15BF8L, + 0x68860BFDL, 0x6C47164AL, 0x61043093L, 0x65C52D24L, + 0x119B4BE9L, 0x155A565EL, 0x18197087L, 0x1CD86D30L, + 0x029F3D35L, 0x065E2082L, 0x0B1D065BL, 0x0FDC1BECL, + 0x3793A651L, 0x3352BBE6L, 0x3E119D3FL, 0x3AD08088L, + 0x2497D08DL, 0x2056CD3AL, 0x2D15EBE3L, 0x29D4F654L, + 0xC5A92679L, 0xC1683BCEL, 0xCC2B1D17L, 0xC8EA00A0L, + 0xD6AD50A5L, 0xD26C4D12L, 0xDF2F6BCBL, 0xDBEE767CL, + 0xE3A1CBC1L, 0xE760D676L, 0xEA23F0AFL, 0xEEE2ED18L, + 0xF0A5BD1DL, 0xF464A0AAL, 0xF9278673L, 0xFDE69BC4L, + 0x89B8FD09L, 0x8D79E0BEL, 0x803AC667L, 0x84FBDBD0L, + 0x9ABC8BD5L, 0x9E7D9662L, 0x933EB0BBL, 0x97FFAD0CL, + 0xAFB010B1L, 0xAB710D06L, 0xA6322BDFL, 0xA2F33668L, + 0xBCB4666DL, 0xB8757BDAL, 0xB5365D03L, 0xB1F740B4L +}; +#endif +static MA_INLINE ma_uint32 ma_dr_flac_crc32_byte(ma_uint32 crc32, ma_uint8 data) +{ +#ifndef MA_DR_FLAC_NO_CRC + return (crc32 << 8) ^ ma_dr_flac__crc32_table[(ma_uint8)((crc32 >> 24) & 0xFF) ^ data]; +#else + (void)data; + return crc32; +#endif +} +#if 0 +static MA_INLINE ma_uint32 ma_dr_flac_crc32_uint32(ma_uint32 crc32, ma_uint32 data) +{ + crc32 = ma_dr_flac_crc32_byte(crc32, (ma_uint8)((data >> 24) & 0xFF)); + crc32 = ma_dr_flac_crc32_byte(crc32, (ma_uint8)((data >> 16) & 0xFF)); + crc32 = ma_dr_flac_crc32_byte(crc32, (ma_uint8)((data >> 8) & 0xFF)); + crc32 = ma_dr_flac_crc32_byte(crc32, (ma_uint8)((data >> 0) & 0xFF)); + return crc32; +} +static MA_INLINE ma_uint32 ma_dr_flac_crc32_uint64(ma_uint32 crc32, ma_uint64 data) +{ + crc32 = ma_dr_flac_crc32_uint32(crc32, (ma_uint32)((data >> 32) & 0xFFFFFFFF)); + crc32 = ma_dr_flac_crc32_uint32(crc32, (ma_uint32)((data >> 0) & 0xFFFFFFFF)); + return crc32; +} +#endif +static MA_INLINE ma_uint32 ma_dr_flac_crc32_buffer(ma_uint32 crc32, ma_uint8* pData, ma_uint32 dataSize) +{ + ma_uint32 i; + for (i = 0; i < dataSize; ++i) { + crc32 = ma_dr_flac_crc32_byte(crc32, pData[i]); + } + return crc32; +} +static MA_INLINE ma_bool32 ma_dr_flac_ogg__is_capture_pattern(ma_uint8 pattern[4]) +{ + return pattern[0] == 'O' && pattern[1] == 'g' && pattern[2] == 'g' && pattern[3] == 'S'; +} +static MA_INLINE ma_uint32 ma_dr_flac_ogg__get_page_header_size(ma_dr_flac_ogg_page_header* pHeader) +{ + return 27 + pHeader->segmentCount; +} +static MA_INLINE ma_uint32 ma_dr_flac_ogg__get_page_body_size(ma_dr_flac_ogg_page_header* pHeader) +{ + ma_uint32 pageBodySize = 0; + int i; + for (i = 0; i < pHeader->segmentCount; ++i) { + pageBodySize += pHeader->segmentTable[i]; + } + return pageBodySize; +} +static ma_result ma_dr_flac_ogg__read_page_header_after_capture_pattern(ma_dr_flac_read_proc onRead, void* pUserData, ma_dr_flac_ogg_page_header* pHeader, ma_uint32* pBytesRead, ma_uint32* pCRC32) +{ + ma_uint8 data[23]; + ma_uint32 i; + MA_DR_FLAC_ASSERT(*pCRC32 == MA_DR_FLAC_OGG_CAPTURE_PATTERN_CRC32); + if (onRead(pUserData, data, 23) != 23) { + return MA_AT_END; + } + *pBytesRead += 23; + pHeader->capturePattern[0] = 'O'; + pHeader->capturePattern[1] = 'g'; + pHeader->capturePattern[2] = 'g'; + pHeader->capturePattern[3] = 'S'; + pHeader->structureVersion = data[0]; + pHeader->headerType = data[1]; + MA_DR_FLAC_COPY_MEMORY(&pHeader->granulePosition, &data[ 2], 8); + MA_DR_FLAC_COPY_MEMORY(&pHeader->serialNumber, &data[10], 4); + MA_DR_FLAC_COPY_MEMORY(&pHeader->sequenceNumber, &data[14], 4); + MA_DR_FLAC_COPY_MEMORY(&pHeader->checksum, &data[18], 4); + pHeader->segmentCount = data[22]; + data[18] = 0; + data[19] = 0; + data[20] = 0; + data[21] = 0; + for (i = 0; i < 23; ++i) { + *pCRC32 = ma_dr_flac_crc32_byte(*pCRC32, data[i]); + } + if (onRead(pUserData, pHeader->segmentTable, pHeader->segmentCount) != pHeader->segmentCount) { + return MA_AT_END; + } + *pBytesRead += pHeader->segmentCount; + for (i = 0; i < pHeader->segmentCount; ++i) { + *pCRC32 = ma_dr_flac_crc32_byte(*pCRC32, pHeader->segmentTable[i]); + } + return MA_SUCCESS; +} +static ma_result ma_dr_flac_ogg__read_page_header(ma_dr_flac_read_proc onRead, void* pUserData, ma_dr_flac_ogg_page_header* pHeader, ma_uint32* pBytesRead, ma_uint32* pCRC32) +{ + ma_uint8 id[4]; + *pBytesRead = 0; + if (onRead(pUserData, id, 4) != 4) { + return MA_AT_END; + } + *pBytesRead += 4; + for (;;) { + if (ma_dr_flac_ogg__is_capture_pattern(id)) { + ma_result result; + *pCRC32 = MA_DR_FLAC_OGG_CAPTURE_PATTERN_CRC32; + result = ma_dr_flac_ogg__read_page_header_after_capture_pattern(onRead, pUserData, pHeader, pBytesRead, pCRC32); + if (result == MA_SUCCESS) { + return MA_SUCCESS; + } else { + if (result == MA_CRC_MISMATCH) { + continue; + } else { + return result; + } + } + } else { + id[0] = id[1]; + id[1] = id[2]; + id[2] = id[3]; + if (onRead(pUserData, &id[3], 1) != 1) { + return MA_AT_END; + } + *pBytesRead += 1; + } + } +} +typedef struct +{ + ma_dr_flac_read_proc onRead; + ma_dr_flac_seek_proc onSeek; + void* pUserData; + ma_uint64 currentBytePos; + ma_uint64 firstBytePos; + ma_uint32 serialNumber; + ma_dr_flac_ogg_page_header bosPageHeader; + ma_dr_flac_ogg_page_header currentPageHeader; + ma_uint32 bytesRemainingInPage; + ma_uint32 pageDataSize; + ma_uint8 pageData[MA_DR_FLAC_OGG_MAX_PAGE_SIZE]; +} ma_dr_flac_oggbs; +static size_t ma_dr_flac_oggbs__read_physical(ma_dr_flac_oggbs* oggbs, void* bufferOut, size_t bytesToRead) +{ + size_t bytesActuallyRead = oggbs->onRead(oggbs->pUserData, bufferOut, bytesToRead); + oggbs->currentBytePos += bytesActuallyRead; + return bytesActuallyRead; +} +static ma_bool32 ma_dr_flac_oggbs__seek_physical(ma_dr_flac_oggbs* oggbs, ma_uint64 offset, ma_dr_flac_seek_origin origin) +{ + if (origin == ma_dr_flac_seek_origin_start) { + if (offset <= 0x7FFFFFFF) { + if (!oggbs->onSeek(oggbs->pUserData, (int)offset, ma_dr_flac_seek_origin_start)) { + return MA_FALSE; + } + oggbs->currentBytePos = offset; + return MA_TRUE; + } else { + if (!oggbs->onSeek(oggbs->pUserData, 0x7FFFFFFF, ma_dr_flac_seek_origin_start)) { + return MA_FALSE; + } + oggbs->currentBytePos = offset; + return ma_dr_flac_oggbs__seek_physical(oggbs, offset - 0x7FFFFFFF, ma_dr_flac_seek_origin_current); + } + } else { + while (offset > 0x7FFFFFFF) { + if (!oggbs->onSeek(oggbs->pUserData, 0x7FFFFFFF, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + oggbs->currentBytePos += 0x7FFFFFFF; + offset -= 0x7FFFFFFF; + } + if (!oggbs->onSeek(oggbs->pUserData, (int)offset, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + oggbs->currentBytePos += offset; + return MA_TRUE; + } +} +static ma_bool32 ma_dr_flac_oggbs__goto_next_page(ma_dr_flac_oggbs* oggbs, ma_dr_flac_ogg_crc_mismatch_recovery recoveryMethod) +{ + ma_dr_flac_ogg_page_header header; + for (;;) { + ma_uint32 crc32 = 0; + ma_uint32 bytesRead; + ma_uint32 pageBodySize; +#ifndef MA_DR_FLAC_NO_CRC + ma_uint32 actualCRC32; +#endif + if (ma_dr_flac_ogg__read_page_header(oggbs->onRead, oggbs->pUserData, &header, &bytesRead, &crc32) != MA_SUCCESS) { + return MA_FALSE; + } + oggbs->currentBytePos += bytesRead; + pageBodySize = ma_dr_flac_ogg__get_page_body_size(&header); + if (pageBodySize > MA_DR_FLAC_OGG_MAX_PAGE_SIZE) { + continue; + } + if (header.serialNumber != oggbs->serialNumber) { + if (pageBodySize > 0 && !ma_dr_flac_oggbs__seek_physical(oggbs, pageBodySize, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + continue; + } + if (ma_dr_flac_oggbs__read_physical(oggbs, oggbs->pageData, pageBodySize) != pageBodySize) { + return MA_FALSE; + } + oggbs->pageDataSize = pageBodySize; +#ifndef MA_DR_FLAC_NO_CRC + actualCRC32 = ma_dr_flac_crc32_buffer(crc32, oggbs->pageData, oggbs->pageDataSize); + if (actualCRC32 != header.checksum) { + if (recoveryMethod == ma_dr_flac_ogg_recover_on_crc_mismatch) { + continue; + } else { + ma_dr_flac_oggbs__goto_next_page(oggbs, ma_dr_flac_ogg_recover_on_crc_mismatch); + return MA_FALSE; + } + } +#else + (void)recoveryMethod; +#endif + oggbs->currentPageHeader = header; + oggbs->bytesRemainingInPage = pageBodySize; + return MA_TRUE; + } +} +#if 0 +static ma_uint8 ma_dr_flac_oggbs__get_current_segment_index(ma_dr_flac_oggbs* oggbs, ma_uint8* pBytesRemainingInSeg) +{ + ma_uint32 bytesConsumedInPage = ma_dr_flac_ogg__get_page_body_size(&oggbs->currentPageHeader) - oggbs->bytesRemainingInPage; + ma_uint8 iSeg = 0; + ma_uint32 iByte = 0; + while (iByte < bytesConsumedInPage) { + ma_uint8 segmentSize = oggbs->currentPageHeader.segmentTable[iSeg]; + if (iByte + segmentSize > bytesConsumedInPage) { + break; + } else { + iSeg += 1; + iByte += segmentSize; + } + } + *pBytesRemainingInSeg = oggbs->currentPageHeader.segmentTable[iSeg] - (ma_uint8)(bytesConsumedInPage - iByte); + return iSeg; +} +static ma_bool32 ma_dr_flac_oggbs__seek_to_next_packet(ma_dr_flac_oggbs* oggbs) +{ + for (;;) { + ma_bool32 atEndOfPage = MA_FALSE; + ma_uint8 bytesRemainingInSeg; + ma_uint8 iFirstSeg = ma_dr_flac_oggbs__get_current_segment_index(oggbs, &bytesRemainingInSeg); + ma_uint32 bytesToEndOfPacketOrPage = bytesRemainingInSeg; + for (ma_uint8 iSeg = iFirstSeg; iSeg < oggbs->currentPageHeader.segmentCount; ++iSeg) { + ma_uint8 segmentSize = oggbs->currentPageHeader.segmentTable[iSeg]; + if (segmentSize < 255) { + if (iSeg == oggbs->currentPageHeader.segmentCount-1) { + atEndOfPage = MA_TRUE; + } + break; + } + bytesToEndOfPacketOrPage += segmentSize; + } + ma_dr_flac_oggbs__seek_physical(oggbs, bytesToEndOfPacketOrPage, ma_dr_flac_seek_origin_current); + oggbs->bytesRemainingInPage -= bytesToEndOfPacketOrPage; + if (atEndOfPage) { + if (!ma_dr_flac_oggbs__goto_next_page(oggbs)) { + return MA_FALSE; + } + if ((oggbs->currentPageHeader.headerType & 0x01) == 0) { + return MA_TRUE; + } + } else { + return MA_TRUE; + } + } +} +static ma_bool32 ma_dr_flac_oggbs__seek_to_next_frame(ma_dr_flac_oggbs* oggbs) +{ + return ma_dr_flac_oggbs__seek_to_next_packet(oggbs); +} +#endif +static size_t ma_dr_flac__on_read_ogg(void* pUserData, void* bufferOut, size_t bytesToRead) +{ + ma_dr_flac_oggbs* oggbs = (ma_dr_flac_oggbs*)pUserData; + ma_uint8* pRunningBufferOut = (ma_uint8*)bufferOut; + size_t bytesRead = 0; + MA_DR_FLAC_ASSERT(oggbs != NULL); + MA_DR_FLAC_ASSERT(pRunningBufferOut != NULL); + while (bytesRead < bytesToRead) { + size_t bytesRemainingToRead = bytesToRead - bytesRead; + if (oggbs->bytesRemainingInPage >= bytesRemainingToRead) { + MA_DR_FLAC_COPY_MEMORY(pRunningBufferOut, oggbs->pageData + (oggbs->pageDataSize - oggbs->bytesRemainingInPage), bytesRemainingToRead); + bytesRead += bytesRemainingToRead; + oggbs->bytesRemainingInPage -= (ma_uint32)bytesRemainingToRead; + break; + } + if (oggbs->bytesRemainingInPage > 0) { + MA_DR_FLAC_COPY_MEMORY(pRunningBufferOut, oggbs->pageData + (oggbs->pageDataSize - oggbs->bytesRemainingInPage), oggbs->bytesRemainingInPage); + bytesRead += oggbs->bytesRemainingInPage; + pRunningBufferOut += oggbs->bytesRemainingInPage; + oggbs->bytesRemainingInPage = 0; + } + MA_DR_FLAC_ASSERT(bytesRemainingToRead > 0); + if (!ma_dr_flac_oggbs__goto_next_page(oggbs, ma_dr_flac_ogg_recover_on_crc_mismatch)) { + break; + } + } + return bytesRead; +} +static ma_bool32 ma_dr_flac__on_seek_ogg(void* pUserData, int offset, ma_dr_flac_seek_origin origin) +{ + ma_dr_flac_oggbs* oggbs = (ma_dr_flac_oggbs*)pUserData; + int bytesSeeked = 0; + MA_DR_FLAC_ASSERT(oggbs != NULL); + MA_DR_FLAC_ASSERT(offset >= 0); + if (origin == ma_dr_flac_seek_origin_start) { + if (!ma_dr_flac_oggbs__seek_physical(oggbs, (int)oggbs->firstBytePos, ma_dr_flac_seek_origin_start)) { + return MA_FALSE; + } + if (!ma_dr_flac_oggbs__goto_next_page(oggbs, ma_dr_flac_ogg_fail_on_crc_mismatch)) { + return MA_FALSE; + } + return ma_dr_flac__on_seek_ogg(pUserData, offset, ma_dr_flac_seek_origin_current); + } + MA_DR_FLAC_ASSERT(origin == ma_dr_flac_seek_origin_current); + while (bytesSeeked < offset) { + int bytesRemainingToSeek = offset - bytesSeeked; + MA_DR_FLAC_ASSERT(bytesRemainingToSeek >= 0); + if (oggbs->bytesRemainingInPage >= (size_t)bytesRemainingToSeek) { + bytesSeeked += bytesRemainingToSeek; + (void)bytesSeeked; + oggbs->bytesRemainingInPage -= bytesRemainingToSeek; + break; + } + if (oggbs->bytesRemainingInPage > 0) { + bytesSeeked += (int)oggbs->bytesRemainingInPage; + oggbs->bytesRemainingInPage = 0; + } + MA_DR_FLAC_ASSERT(bytesRemainingToSeek > 0); + if (!ma_dr_flac_oggbs__goto_next_page(oggbs, ma_dr_flac_ogg_fail_on_crc_mismatch)) { + return MA_FALSE; + } + } + return MA_TRUE; +} +static ma_bool32 ma_dr_flac_ogg__seek_to_pcm_frame(ma_dr_flac* pFlac, ma_uint64 pcmFrameIndex) +{ + ma_dr_flac_oggbs* oggbs = (ma_dr_flac_oggbs*)pFlac->_oggbs; + ma_uint64 originalBytePos; + ma_uint64 runningGranulePosition; + ma_uint64 runningFrameBytePos; + ma_uint64 runningPCMFrameCount; + MA_DR_FLAC_ASSERT(oggbs != NULL); + originalBytePos = oggbs->currentBytePos; + if (!ma_dr_flac__seek_to_byte(&pFlac->bs, pFlac->firstFLACFramePosInBytes)) { + return MA_FALSE; + } + oggbs->bytesRemainingInPage = 0; + runningGranulePosition = 0; + for (;;) { + if (!ma_dr_flac_oggbs__goto_next_page(oggbs, ma_dr_flac_ogg_recover_on_crc_mismatch)) { + ma_dr_flac_oggbs__seek_physical(oggbs, originalBytePos, ma_dr_flac_seek_origin_start); + return MA_FALSE; + } + runningFrameBytePos = oggbs->currentBytePos - ma_dr_flac_ogg__get_page_header_size(&oggbs->currentPageHeader) - oggbs->pageDataSize; + if (oggbs->currentPageHeader.granulePosition >= pcmFrameIndex) { + break; + } + if ((oggbs->currentPageHeader.headerType & 0x01) == 0) { + if (oggbs->currentPageHeader.segmentTable[0] >= 2) { + ma_uint8 firstBytesInPage[2]; + firstBytesInPage[0] = oggbs->pageData[0]; + firstBytesInPage[1] = oggbs->pageData[1]; + if ((firstBytesInPage[0] == 0xFF) && (firstBytesInPage[1] & 0xFC) == 0xF8) { + runningGranulePosition = oggbs->currentPageHeader.granulePosition; + } + continue; + } + } + } + if (!ma_dr_flac_oggbs__seek_physical(oggbs, runningFrameBytePos, ma_dr_flac_seek_origin_start)) { + return MA_FALSE; + } + if (!ma_dr_flac_oggbs__goto_next_page(oggbs, ma_dr_flac_ogg_recover_on_crc_mismatch)) { + return MA_FALSE; + } + runningPCMFrameCount = runningGranulePosition; + for (;;) { + ma_uint64 firstPCMFrameInFLACFrame = 0; + ma_uint64 lastPCMFrameInFLACFrame = 0; + ma_uint64 pcmFrameCountInThisFrame; + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + return MA_FALSE; + } + ma_dr_flac__get_pcm_frame_range_of_current_flac_frame(pFlac, &firstPCMFrameInFLACFrame, &lastPCMFrameInFLACFrame); + pcmFrameCountInThisFrame = (lastPCMFrameInFLACFrame - firstPCMFrameInFLACFrame) + 1; + if (pcmFrameIndex == pFlac->totalPCMFrameCount && (runningPCMFrameCount + pcmFrameCountInThisFrame) == pFlac->totalPCMFrameCount) { + ma_result result = ma_dr_flac__decode_flac_frame(pFlac); + if (result == MA_SUCCESS) { + pFlac->currentPCMFrame = pcmFrameIndex; + pFlac->currentFLACFrame.pcmFramesRemaining = 0; + return MA_TRUE; + } else { + return MA_FALSE; + } + } + if (pcmFrameIndex < (runningPCMFrameCount + pcmFrameCountInThisFrame)) { + ma_result result = ma_dr_flac__decode_flac_frame(pFlac); + if (result == MA_SUCCESS) { + ma_uint64 pcmFramesToDecode = (size_t)(pcmFrameIndex - runningPCMFrameCount); + if (pcmFramesToDecode == 0) { + return MA_TRUE; + } + pFlac->currentPCMFrame = runningPCMFrameCount; + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, pcmFramesToDecode) == pcmFramesToDecode; + } else { + if (result == MA_CRC_MISMATCH) { + continue; + } else { + return MA_FALSE; + } + } + } else { + ma_result result = ma_dr_flac__seek_to_next_flac_frame(pFlac); + if (result == MA_SUCCESS) { + runningPCMFrameCount += pcmFrameCountInThisFrame; + } else { + if (result == MA_CRC_MISMATCH) { + continue; + } else { + return MA_FALSE; + } + } + } + } +} +static ma_bool32 ma_dr_flac__init_private__ogg(ma_dr_flac_init_info* pInit, ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, void* pUserData, void* pUserDataMD, ma_bool32 relaxed) +{ + ma_dr_flac_ogg_page_header header; + ma_uint32 crc32 = MA_DR_FLAC_OGG_CAPTURE_PATTERN_CRC32; + ma_uint32 bytesRead = 0; + (void)relaxed; + pInit->container = ma_dr_flac_container_ogg; + pInit->oggFirstBytePos = 0; + if (ma_dr_flac_ogg__read_page_header_after_capture_pattern(onRead, pUserData, &header, &bytesRead, &crc32) != MA_SUCCESS) { + return MA_FALSE; + } + pInit->runningFilePos += bytesRead; + for (;;) { + int pageBodySize; + if ((header.headerType & 0x02) == 0) { + return MA_FALSE; + } + pageBodySize = ma_dr_flac_ogg__get_page_body_size(&header); + if (pageBodySize == 51) { + ma_uint32 bytesRemainingInPage = pageBodySize; + ma_uint8 packetType; + if (onRead(pUserData, &packetType, 1) != 1) { + return MA_FALSE; + } + bytesRemainingInPage -= 1; + if (packetType == 0x7F) { + ma_uint8 sig[4]; + if (onRead(pUserData, sig, 4) != 4) { + return MA_FALSE; + } + bytesRemainingInPage -= 4; + if (sig[0] == 'F' && sig[1] == 'L' && sig[2] == 'A' && sig[3] == 'C') { + ma_uint8 mappingVersion[2]; + if (onRead(pUserData, mappingVersion, 2) != 2) { + return MA_FALSE; + } + if (mappingVersion[0] != 1) { + return MA_FALSE; + } + if (!onSeek(pUserData, 2, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + if (onRead(pUserData, sig, 4) != 4) { + return MA_FALSE; + } + if (sig[0] == 'f' && sig[1] == 'L' && sig[2] == 'a' && sig[3] == 'C') { + ma_dr_flac_streaminfo streaminfo; + ma_uint8 isLastBlock; + ma_uint8 blockType; + ma_uint32 blockSize; + if (!ma_dr_flac__read_and_decode_block_header(onRead, pUserData, &isLastBlock, &blockType, &blockSize)) { + return MA_FALSE; + } + if (blockType != MA_DR_FLAC_METADATA_BLOCK_TYPE_STREAMINFO || blockSize != 34) { + return MA_FALSE; + } + if (ma_dr_flac__read_streaminfo(onRead, pUserData, &streaminfo)) { + pInit->hasStreamInfoBlock = MA_TRUE; + pInit->sampleRate = streaminfo.sampleRate; + pInit->channels = streaminfo.channels; + pInit->bitsPerSample = streaminfo.bitsPerSample; + pInit->totalPCMFrameCount = streaminfo.totalPCMFrameCount; + pInit->maxBlockSizeInPCMFrames = streaminfo.maxBlockSizeInPCMFrames; + pInit->hasMetadataBlocks = !isLastBlock; + if (onMeta) { + ma_dr_flac_metadata metadata; + metadata.type = MA_DR_FLAC_METADATA_BLOCK_TYPE_STREAMINFO; + metadata.pRawData = NULL; + metadata.rawDataSize = 0; + metadata.data.streaminfo = streaminfo; + onMeta(pUserDataMD, &metadata); + } + pInit->runningFilePos += pageBodySize; + pInit->oggFirstBytePos = pInit->runningFilePos - 79; + pInit->oggSerial = header.serialNumber; + pInit->oggBosHeader = header; + break; + } else { + return MA_FALSE; + } + } else { + return MA_FALSE; + } + } else { + if (!onSeek(pUserData, bytesRemainingInPage, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + } + } else { + if (!onSeek(pUserData, bytesRemainingInPage, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + } + } else { + if (!onSeek(pUserData, pageBodySize, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + } + pInit->runningFilePos += pageBodySize; + if (ma_dr_flac_ogg__read_page_header(onRead, pUserData, &header, &bytesRead, &crc32) != MA_SUCCESS) { + return MA_FALSE; + } + pInit->runningFilePos += bytesRead; + } + pInit->hasMetadataBlocks = MA_TRUE; + return MA_TRUE; +} +#endif +static ma_bool32 ma_dr_flac__init_private(ma_dr_flac_init_info* pInit, ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, ma_dr_flac_container container, void* pUserData, void* pUserDataMD) +{ + ma_bool32 relaxed; + ma_uint8 id[4]; + if (pInit == NULL || onRead == NULL || onSeek == NULL) { + return MA_FALSE; + } + MA_DR_FLAC_ZERO_MEMORY(pInit, sizeof(*pInit)); + pInit->onRead = onRead; + pInit->onSeek = onSeek; + pInit->onMeta = onMeta; + pInit->container = container; + pInit->pUserData = pUserData; + pInit->pUserDataMD = pUserDataMD; + pInit->bs.onRead = onRead; + pInit->bs.onSeek = onSeek; + pInit->bs.pUserData = pUserData; + ma_dr_flac__reset_cache(&pInit->bs); + relaxed = container != ma_dr_flac_container_unknown; + for (;;) { + if (onRead(pUserData, id, 4) != 4) { + return MA_FALSE; + } + pInit->runningFilePos += 4; + if (id[0] == 'I' && id[1] == 'D' && id[2] == '3') { + ma_uint8 header[6]; + ma_uint8 flags; + ma_uint32 headerSize; + if (onRead(pUserData, header, 6) != 6) { + return MA_FALSE; + } + pInit->runningFilePos += 6; + flags = header[1]; + MA_DR_FLAC_COPY_MEMORY(&headerSize, header+2, 4); + headerSize = ma_dr_flac__unsynchsafe_32(ma_dr_flac__be2host_32(headerSize)); + if (flags & 0x10) { + headerSize += 10; + } + if (!onSeek(pUserData, headerSize, ma_dr_flac_seek_origin_current)) { + return MA_FALSE; + } + pInit->runningFilePos += headerSize; + } else { + break; + } + } + if (id[0] == 'f' && id[1] == 'L' && id[2] == 'a' && id[3] == 'C') { + return ma_dr_flac__init_private__native(pInit, onRead, onSeek, onMeta, pUserData, pUserDataMD, relaxed); + } +#ifndef MA_DR_FLAC_NO_OGG + if (id[0] == 'O' && id[1] == 'g' && id[2] == 'g' && id[3] == 'S') { + return ma_dr_flac__init_private__ogg(pInit, onRead, onSeek, onMeta, pUserData, pUserDataMD, relaxed); + } +#endif + if (relaxed) { + if (container == ma_dr_flac_container_native) { + return ma_dr_flac__init_private__native(pInit, onRead, onSeek, onMeta, pUserData, pUserDataMD, relaxed); + } +#ifndef MA_DR_FLAC_NO_OGG + if (container == ma_dr_flac_container_ogg) { + return ma_dr_flac__init_private__ogg(pInit, onRead, onSeek, onMeta, pUserData, pUserDataMD, relaxed); + } +#endif + } + return MA_FALSE; +} +static void ma_dr_flac__init_from_info(ma_dr_flac* pFlac, const ma_dr_flac_init_info* pInit) +{ + MA_DR_FLAC_ASSERT(pFlac != NULL); + MA_DR_FLAC_ASSERT(pInit != NULL); + MA_DR_FLAC_ZERO_MEMORY(pFlac, sizeof(*pFlac)); + pFlac->bs = pInit->bs; + pFlac->onMeta = pInit->onMeta; + pFlac->pUserDataMD = pInit->pUserDataMD; + pFlac->maxBlockSizeInPCMFrames = pInit->maxBlockSizeInPCMFrames; + pFlac->sampleRate = pInit->sampleRate; + pFlac->channels = (ma_uint8)pInit->channels; + pFlac->bitsPerSample = (ma_uint8)pInit->bitsPerSample; + pFlac->totalPCMFrameCount = pInit->totalPCMFrameCount; + pFlac->container = pInit->container; +} +static ma_dr_flac* ma_dr_flac_open_with_metadata_private(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, ma_dr_flac_container container, void* pUserData, void* pUserDataMD, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac_init_info init; + ma_uint32 allocationSize; + ma_uint32 wholeSIMDVectorCountPerChannel; + ma_uint32 decodedSamplesAllocationSize; +#ifndef MA_DR_FLAC_NO_OGG + ma_dr_flac_oggbs* pOggbs = NULL; +#endif + ma_uint64 firstFramePos; + ma_uint64 seektablePos; + ma_uint32 seekpointCount; + ma_allocation_callbacks allocationCallbacks; + ma_dr_flac* pFlac; + ma_dr_flac__init_cpu_caps(); + if (!ma_dr_flac__init_private(&init, onRead, onSeek, onMeta, container, pUserData, pUserDataMD)) { + return NULL; + } + if (pAllocationCallbacks != NULL) { + allocationCallbacks = *pAllocationCallbacks; + if (allocationCallbacks.onFree == NULL || (allocationCallbacks.onMalloc == NULL && allocationCallbacks.onRealloc == NULL)) { + return NULL; + } + } else { + allocationCallbacks.pUserData = NULL; + allocationCallbacks.onMalloc = ma_dr_flac__malloc_default; + allocationCallbacks.onRealloc = ma_dr_flac__realloc_default; + allocationCallbacks.onFree = ma_dr_flac__free_default; + } + allocationSize = sizeof(ma_dr_flac); + if ((init.maxBlockSizeInPCMFrames % (MA_DR_FLAC_MAX_SIMD_VECTOR_SIZE / sizeof(ma_int32))) == 0) { + wholeSIMDVectorCountPerChannel = (init.maxBlockSizeInPCMFrames / (MA_DR_FLAC_MAX_SIMD_VECTOR_SIZE / sizeof(ma_int32))); + } else { + wholeSIMDVectorCountPerChannel = (init.maxBlockSizeInPCMFrames / (MA_DR_FLAC_MAX_SIMD_VECTOR_SIZE / sizeof(ma_int32))) + 1; + } + decodedSamplesAllocationSize = wholeSIMDVectorCountPerChannel * MA_DR_FLAC_MAX_SIMD_VECTOR_SIZE * init.channels; + allocationSize += decodedSamplesAllocationSize; + allocationSize += MA_DR_FLAC_MAX_SIMD_VECTOR_SIZE; +#ifndef MA_DR_FLAC_NO_OGG + if (init.container == ma_dr_flac_container_ogg) { + allocationSize += sizeof(ma_dr_flac_oggbs); + pOggbs = (ma_dr_flac_oggbs*)ma_dr_flac__malloc_from_callbacks(sizeof(*pOggbs), &allocationCallbacks); + if (pOggbs == NULL) { + return NULL; + } + MA_DR_FLAC_ZERO_MEMORY(pOggbs, sizeof(*pOggbs)); + pOggbs->onRead = onRead; + pOggbs->onSeek = onSeek; + pOggbs->pUserData = pUserData; + pOggbs->currentBytePos = init.oggFirstBytePos; + pOggbs->firstBytePos = init.oggFirstBytePos; + pOggbs->serialNumber = init.oggSerial; + pOggbs->bosPageHeader = init.oggBosHeader; + pOggbs->bytesRemainingInPage = 0; + } +#endif + firstFramePos = 42; + seektablePos = 0; + seekpointCount = 0; + if (init.hasMetadataBlocks) { + ma_dr_flac_read_proc onReadOverride = onRead; + ma_dr_flac_seek_proc onSeekOverride = onSeek; + void* pUserDataOverride = pUserData; +#ifndef MA_DR_FLAC_NO_OGG + if (init.container == ma_dr_flac_container_ogg) { + onReadOverride = ma_dr_flac__on_read_ogg; + onSeekOverride = ma_dr_flac__on_seek_ogg; + pUserDataOverride = (void*)pOggbs; + } +#endif + if (!ma_dr_flac__read_and_decode_metadata(onReadOverride, onSeekOverride, onMeta, pUserDataOverride, pUserDataMD, &firstFramePos, &seektablePos, &seekpointCount, &allocationCallbacks)) { + #ifndef MA_DR_FLAC_NO_OGG + ma_dr_flac__free_from_callbacks(pOggbs, &allocationCallbacks); + #endif + return NULL; + } + allocationSize += seekpointCount * sizeof(ma_dr_flac_seekpoint); + } + pFlac = (ma_dr_flac*)ma_dr_flac__malloc_from_callbacks(allocationSize, &allocationCallbacks); + if (pFlac == NULL) { + #ifndef MA_DR_FLAC_NO_OGG + ma_dr_flac__free_from_callbacks(pOggbs, &allocationCallbacks); + #endif + return NULL; + } + ma_dr_flac__init_from_info(pFlac, &init); + pFlac->allocationCallbacks = allocationCallbacks; + pFlac->pDecodedSamples = (ma_int32*)ma_dr_flac_align((size_t)pFlac->pExtraData, MA_DR_FLAC_MAX_SIMD_VECTOR_SIZE); +#ifndef MA_DR_FLAC_NO_OGG + if (init.container == ma_dr_flac_container_ogg) { + ma_dr_flac_oggbs* pInternalOggbs = (ma_dr_flac_oggbs*)((ma_uint8*)pFlac->pDecodedSamples + decodedSamplesAllocationSize + (seekpointCount * sizeof(ma_dr_flac_seekpoint))); + MA_DR_FLAC_COPY_MEMORY(pInternalOggbs, pOggbs, sizeof(*pOggbs)); + ma_dr_flac__free_from_callbacks(pOggbs, &allocationCallbacks); + pOggbs = NULL; + pFlac->bs.onRead = ma_dr_flac__on_read_ogg; + pFlac->bs.onSeek = ma_dr_flac__on_seek_ogg; + pFlac->bs.pUserData = (void*)pInternalOggbs; + pFlac->_oggbs = (void*)pInternalOggbs; + } +#endif + pFlac->firstFLACFramePosInBytes = firstFramePos; +#ifndef MA_DR_FLAC_NO_OGG + if (init.container == ma_dr_flac_container_ogg) + { + pFlac->pSeekpoints = NULL; + pFlac->seekpointCount = 0; + } + else +#endif + { + if (seektablePos != 0) { + pFlac->seekpointCount = seekpointCount; + pFlac->pSeekpoints = (ma_dr_flac_seekpoint*)((ma_uint8*)pFlac->pDecodedSamples + decodedSamplesAllocationSize); + MA_DR_FLAC_ASSERT(pFlac->bs.onSeek != NULL); + MA_DR_FLAC_ASSERT(pFlac->bs.onRead != NULL); + if (pFlac->bs.onSeek(pFlac->bs.pUserData, (int)seektablePos, ma_dr_flac_seek_origin_start)) { + ma_uint32 iSeekpoint; + for (iSeekpoint = 0; iSeekpoint < seekpointCount; iSeekpoint += 1) { + if (pFlac->bs.onRead(pFlac->bs.pUserData, pFlac->pSeekpoints + iSeekpoint, MA_DR_FLAC_SEEKPOINT_SIZE_IN_BYTES) == MA_DR_FLAC_SEEKPOINT_SIZE_IN_BYTES) { + pFlac->pSeekpoints[iSeekpoint].firstPCMFrame = ma_dr_flac__be2host_64(pFlac->pSeekpoints[iSeekpoint].firstPCMFrame); + pFlac->pSeekpoints[iSeekpoint].flacFrameOffset = ma_dr_flac__be2host_64(pFlac->pSeekpoints[iSeekpoint].flacFrameOffset); + pFlac->pSeekpoints[iSeekpoint].pcmFrameCount = ma_dr_flac__be2host_16(pFlac->pSeekpoints[iSeekpoint].pcmFrameCount); + } else { + pFlac->pSeekpoints = NULL; + pFlac->seekpointCount = 0; + break; + } + } + if (!pFlac->bs.onSeek(pFlac->bs.pUserData, (int)pFlac->firstFLACFramePosInBytes, ma_dr_flac_seek_origin_start)) { + ma_dr_flac__free_from_callbacks(pFlac, &allocationCallbacks); + return NULL; + } + } else { + pFlac->pSeekpoints = NULL; + pFlac->seekpointCount = 0; + } + } + } + if (!init.hasStreamInfoBlock) { + pFlac->currentFLACFrame.header = init.firstFrameHeader; + for (;;) { + ma_result result = ma_dr_flac__decode_flac_frame(pFlac); + if (result == MA_SUCCESS) { + break; + } else { + if (result == MA_CRC_MISMATCH) { + if (!ma_dr_flac__read_next_flac_frame_header(&pFlac->bs, pFlac->bitsPerSample, &pFlac->currentFLACFrame.header)) { + ma_dr_flac__free_from_callbacks(pFlac, &allocationCallbacks); + return NULL; + } + continue; + } else { + ma_dr_flac__free_from_callbacks(pFlac, &allocationCallbacks); + return NULL; + } + } + } + } + return pFlac; +} +#ifndef MA_DR_FLAC_NO_STDIO +#include +#ifndef MA_DR_FLAC_NO_WCHAR +#include +#endif +static size_t ma_dr_flac__on_read_stdio(void* pUserData, void* bufferOut, size_t bytesToRead) +{ + return fread(bufferOut, 1, bytesToRead, (FILE*)pUserData); +} +static ma_bool32 ma_dr_flac__on_seek_stdio(void* pUserData, int offset, ma_dr_flac_seek_origin origin) +{ + MA_DR_FLAC_ASSERT(offset >= 0); + return fseek((FILE*)pUserData, offset, (origin == ma_dr_flac_seek_origin_current) ? SEEK_CUR : SEEK_SET) == 0; +} +MA_API ma_dr_flac* ma_dr_flac_open_file(const char* pFileName, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + FILE* pFile; + if (ma_fopen(&pFile, pFileName, "rb") != MA_SUCCESS) { + return NULL; + } + pFlac = ma_dr_flac_open(ma_dr_flac__on_read_stdio, ma_dr_flac__on_seek_stdio, (void*)pFile, pAllocationCallbacks); + if (pFlac == NULL) { + fclose(pFile); + return NULL; + } + return pFlac; +} +#ifndef MA_DR_FLAC_NO_WCHAR +MA_API ma_dr_flac* ma_dr_flac_open_file_w(const wchar_t* pFileName, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + FILE* pFile; + if (ma_wfopen(&pFile, pFileName, L"rb", pAllocationCallbacks) != MA_SUCCESS) { + return NULL; + } + pFlac = ma_dr_flac_open(ma_dr_flac__on_read_stdio, ma_dr_flac__on_seek_stdio, (void*)pFile, pAllocationCallbacks); + if (pFlac == NULL) { + fclose(pFile); + return NULL; + } + return pFlac; +} +#endif +MA_API ma_dr_flac* ma_dr_flac_open_file_with_metadata(const char* pFileName, ma_dr_flac_meta_proc onMeta, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + FILE* pFile; + if (ma_fopen(&pFile, pFileName, "rb") != MA_SUCCESS) { + return NULL; + } + pFlac = ma_dr_flac_open_with_metadata_private(ma_dr_flac__on_read_stdio, ma_dr_flac__on_seek_stdio, onMeta, ma_dr_flac_container_unknown, (void*)pFile, pUserData, pAllocationCallbacks); + if (pFlac == NULL) { + fclose(pFile); + return pFlac; + } + return pFlac; +} +#ifndef MA_DR_FLAC_NO_WCHAR +MA_API ma_dr_flac* ma_dr_flac_open_file_with_metadata_w(const wchar_t* pFileName, ma_dr_flac_meta_proc onMeta, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + FILE* pFile; + if (ma_wfopen(&pFile, pFileName, L"rb", pAllocationCallbacks) != MA_SUCCESS) { + return NULL; + } + pFlac = ma_dr_flac_open_with_metadata_private(ma_dr_flac__on_read_stdio, ma_dr_flac__on_seek_stdio, onMeta, ma_dr_flac_container_unknown, (void*)pFile, pUserData, pAllocationCallbacks); + if (pFlac == NULL) { + fclose(pFile); + return pFlac; + } + return pFlac; +} +#endif +#endif +static size_t ma_dr_flac__on_read_memory(void* pUserData, void* bufferOut, size_t bytesToRead) +{ + ma_dr_flac__memory_stream* memoryStream = (ma_dr_flac__memory_stream*)pUserData; + size_t bytesRemaining; + MA_DR_FLAC_ASSERT(memoryStream != NULL); + MA_DR_FLAC_ASSERT(memoryStream->dataSize >= memoryStream->currentReadPos); + bytesRemaining = memoryStream->dataSize - memoryStream->currentReadPos; + if (bytesToRead > bytesRemaining) { + bytesToRead = bytesRemaining; + } + if (bytesToRead > 0) { + MA_DR_FLAC_COPY_MEMORY(bufferOut, memoryStream->data + memoryStream->currentReadPos, bytesToRead); + memoryStream->currentReadPos += bytesToRead; + } + return bytesToRead; +} +static ma_bool32 ma_dr_flac__on_seek_memory(void* pUserData, int offset, ma_dr_flac_seek_origin origin) +{ + ma_dr_flac__memory_stream* memoryStream = (ma_dr_flac__memory_stream*)pUserData; + MA_DR_FLAC_ASSERT(memoryStream != NULL); + MA_DR_FLAC_ASSERT(offset >= 0); + if (offset > (ma_int64)memoryStream->dataSize) { + return MA_FALSE; + } + if (origin == ma_dr_flac_seek_origin_current) { + if (memoryStream->currentReadPos + offset <= memoryStream->dataSize) { + memoryStream->currentReadPos += offset; + } else { + return MA_FALSE; + } + } else { + if ((ma_uint32)offset <= memoryStream->dataSize) { + memoryStream->currentReadPos = offset; + } else { + return MA_FALSE; + } + } + return MA_TRUE; +} +MA_API ma_dr_flac* ma_dr_flac_open_memory(const void* pData, size_t dataSize, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac__memory_stream memoryStream; + ma_dr_flac* pFlac; + memoryStream.data = (const ma_uint8*)pData; + memoryStream.dataSize = dataSize; + memoryStream.currentReadPos = 0; + pFlac = ma_dr_flac_open(ma_dr_flac__on_read_memory, ma_dr_flac__on_seek_memory, &memoryStream, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + pFlac->memoryStream = memoryStream; +#ifndef MA_DR_FLAC_NO_OGG + if (pFlac->container == ma_dr_flac_container_ogg) + { + ma_dr_flac_oggbs* oggbs = (ma_dr_flac_oggbs*)pFlac->_oggbs; + oggbs->pUserData = &pFlac->memoryStream; + } + else +#endif + { + pFlac->bs.pUserData = &pFlac->memoryStream; + } + return pFlac; +} +MA_API ma_dr_flac* ma_dr_flac_open_memory_with_metadata(const void* pData, size_t dataSize, ma_dr_flac_meta_proc onMeta, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac__memory_stream memoryStream; + ma_dr_flac* pFlac; + memoryStream.data = (const ma_uint8*)pData; + memoryStream.dataSize = dataSize; + memoryStream.currentReadPos = 0; + pFlac = ma_dr_flac_open_with_metadata_private(ma_dr_flac__on_read_memory, ma_dr_flac__on_seek_memory, onMeta, ma_dr_flac_container_unknown, &memoryStream, pUserData, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + pFlac->memoryStream = memoryStream; +#ifndef MA_DR_FLAC_NO_OGG + if (pFlac->container == ma_dr_flac_container_ogg) + { + ma_dr_flac_oggbs* oggbs = (ma_dr_flac_oggbs*)pFlac->_oggbs; + oggbs->pUserData = &pFlac->memoryStream; + } + else +#endif + { + pFlac->bs.pUserData = &pFlac->memoryStream; + } + return pFlac; +} +MA_API ma_dr_flac* ma_dr_flac_open(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_flac_open_with_metadata_private(onRead, onSeek, NULL, ma_dr_flac_container_unknown, pUserData, pUserData, pAllocationCallbacks); +} +MA_API ma_dr_flac* ma_dr_flac_open_relaxed(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_container container, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_flac_open_with_metadata_private(onRead, onSeek, NULL, container, pUserData, pUserData, pAllocationCallbacks); +} +MA_API ma_dr_flac* ma_dr_flac_open_with_metadata(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_flac_open_with_metadata_private(onRead, onSeek, onMeta, ma_dr_flac_container_unknown, pUserData, pUserData, pAllocationCallbacks); +} +MA_API ma_dr_flac* ma_dr_flac_open_with_metadata_relaxed(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, ma_dr_flac_meta_proc onMeta, ma_dr_flac_container container, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + return ma_dr_flac_open_with_metadata_private(onRead, onSeek, onMeta, container, pUserData, pUserData, pAllocationCallbacks); +} +MA_API void ma_dr_flac_close(ma_dr_flac* pFlac) +{ + if (pFlac == NULL) { + return; + } +#ifndef MA_DR_FLAC_NO_STDIO + if (pFlac->bs.onRead == ma_dr_flac__on_read_stdio) { + fclose((FILE*)pFlac->bs.pUserData); + } +#ifndef MA_DR_FLAC_NO_OGG + if (pFlac->container == ma_dr_flac_container_ogg) { + ma_dr_flac_oggbs* oggbs = (ma_dr_flac_oggbs*)pFlac->_oggbs; + MA_DR_FLAC_ASSERT(pFlac->bs.onRead == ma_dr_flac__on_read_ogg); + if (oggbs->onRead == ma_dr_flac__on_read_stdio) { + fclose((FILE*)oggbs->pUserData); + } + } +#endif +#endif + ma_dr_flac__free_from_callbacks(pFlac, &pFlac->allocationCallbacks); +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_left_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + for (i = 0; i < frameCount; ++i) { + ma_uint32 left = (ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + ma_uint32 side = (ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + ma_uint32 right = left - side; + pOutputSamples[i*2+0] = (ma_int32)left; + pOutputSamples[i*2+1] = (ma_int32)right; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_left_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 left0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 left1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 left2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 left3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << shift1; + ma_uint32 right0 = left0 - side0; + ma_uint32 right1 = left1 - side1; + ma_uint32 right2 = left2 - side2; + ma_uint32 right3 = left3 - side3; + pOutputSamples[i*8+0] = (ma_int32)left0; + pOutputSamples[i*8+1] = (ma_int32)right0; + pOutputSamples[i*8+2] = (ma_int32)left1; + pOutputSamples[i*8+3] = (ma_int32)right1; + pOutputSamples[i*8+4] = (ma_int32)left2; + pOutputSamples[i*8+5] = (ma_int32)right2; + pOutputSamples[i*8+6] = (ma_int32)left3; + pOutputSamples[i*8+7] = (ma_int32)right3; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + pOutputSamples[i*2+0] = (ma_int32)left; + pOutputSamples[i*2+1] = (ma_int32)right; + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_left_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + for (i = 0; i < frameCount4; ++i) { + __m128i left = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + __m128i side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + __m128i right = _mm_sub_epi32(left, side); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 0), _mm_unpacklo_epi32(left, right)); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 4), _mm_unpackhi_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + pOutputSamples[i*2+0] = (ma_int32)left; + pOutputSamples[i*2+1] = (ma_int32)right; + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_left_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + int32x4_t shift0_4; + int32x4_t shift1_4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + shift0_4 = vdupq_n_s32(shift0); + shift1_4 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t left; + uint32x4_t side; + uint32x4_t right; + left = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift0_4); + side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift1_4); + right = vsubq_u32(left, side); + ma_dr_flac__vst2q_u32((ma_uint32*)pOutputSamples + i*8, vzipq_u32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + pOutputSamples[i*2+0] = (ma_int32)left; + pOutputSamples[i*2+1] = (ma_int32)right; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_left_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s32__decode_left_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s32__decode_left_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_s32__decode_left_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_s32__decode_left_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_right_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + for (i = 0; i < frameCount; ++i) { + ma_uint32 side = (ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + ma_uint32 right = (ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + ma_uint32 left = right + side; + pOutputSamples[i*2+0] = (ma_int32)left; + pOutputSamples[i*2+1] = (ma_int32)right; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_right_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 side0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 side1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 side2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 side3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 right0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 right1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 right2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 right3 = pInputSamples1U32[i*4+3] << shift1; + ma_uint32 left0 = right0 + side0; + ma_uint32 left1 = right1 + side1; + ma_uint32 left2 = right2 + side2; + ma_uint32 left3 = right3 + side3; + pOutputSamples[i*8+0] = (ma_int32)left0; + pOutputSamples[i*8+1] = (ma_int32)right0; + pOutputSamples[i*8+2] = (ma_int32)left1; + pOutputSamples[i*8+3] = (ma_int32)right1; + pOutputSamples[i*8+4] = (ma_int32)left2; + pOutputSamples[i*8+5] = (ma_int32)right2; + pOutputSamples[i*8+6] = (ma_int32)left3; + pOutputSamples[i*8+7] = (ma_int32)right3; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + pOutputSamples[i*2+0] = (ma_int32)left; + pOutputSamples[i*2+1] = (ma_int32)right; + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_right_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + for (i = 0; i < frameCount4; ++i) { + __m128i side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + __m128i right = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + __m128i left = _mm_add_epi32(right, side); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 0), _mm_unpacklo_epi32(left, right)); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 4), _mm_unpackhi_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + pOutputSamples[i*2+0] = (ma_int32)left; + pOutputSamples[i*2+1] = (ma_int32)right; + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_right_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + int32x4_t shift0_4; + int32x4_t shift1_4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + shift0_4 = vdupq_n_s32(shift0); + shift1_4 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t side; + uint32x4_t right; + uint32x4_t left; + side = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift0_4); + right = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift1_4); + left = vaddq_u32(right, side); + ma_dr_flac__vst2q_u32((ma_uint32*)pOutputSamples + i*8, vzipq_u32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + pOutputSamples[i*2+0] = (ma_int32)left; + pOutputSamples[i*2+1] = (ma_int32)right; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_right_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s32__decode_right_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s32__decode_right_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_s32__decode_right_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_s32__decode_right_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_mid_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + for (ma_uint64 i = 0; i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)((ma_uint32)((ma_int32)(mid + side) >> 1) << unusedBitsPerSample); + pOutputSamples[i*2+1] = (ma_int32)((ma_uint32)((ma_int32)(mid - side) >> 1) << unusedBitsPerSample); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_mid_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_int32 shift = unusedBitsPerSample; + if (shift > 0) { + shift -= 1; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 temp0L; + ma_uint32 temp1L; + ma_uint32 temp2L; + ma_uint32 temp3L; + ma_uint32 temp0R; + ma_uint32 temp1R; + ma_uint32 temp2R; + ma_uint32 temp3R; + ma_uint32 mid0 = pInputSamples0U32[i*4+0] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid1 = pInputSamples0U32[i*4+1] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid2 = pInputSamples0U32[i*4+2] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid3 = pInputSamples0U32[i*4+3] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid0 = (mid0 << 1) | (side0 & 0x01); + mid1 = (mid1 << 1) | (side1 & 0x01); + mid2 = (mid2 << 1) | (side2 & 0x01); + mid3 = (mid3 << 1) | (side3 & 0x01); + temp0L = (mid0 + side0) << shift; + temp1L = (mid1 + side1) << shift; + temp2L = (mid2 + side2) << shift; + temp3L = (mid3 + side3) << shift; + temp0R = (mid0 - side0) << shift; + temp1R = (mid1 - side1) << shift; + temp2R = (mid2 - side2) << shift; + temp3R = (mid3 - side3) << shift; + pOutputSamples[i*8+0] = (ma_int32)temp0L; + pOutputSamples[i*8+1] = (ma_int32)temp0R; + pOutputSamples[i*8+2] = (ma_int32)temp1L; + pOutputSamples[i*8+3] = (ma_int32)temp1R; + pOutputSamples[i*8+4] = (ma_int32)temp2L; + pOutputSamples[i*8+5] = (ma_int32)temp2R; + pOutputSamples[i*8+6] = (ma_int32)temp3L; + pOutputSamples[i*8+7] = (ma_int32)temp3R; + } + } else { + for (i = 0; i < frameCount4; ++i) { + ma_uint32 temp0L; + ma_uint32 temp1L; + ma_uint32 temp2L; + ma_uint32 temp3L; + ma_uint32 temp0R; + ma_uint32 temp1R; + ma_uint32 temp2R; + ma_uint32 temp3R; + ma_uint32 mid0 = pInputSamples0U32[i*4+0] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid1 = pInputSamples0U32[i*4+1] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid2 = pInputSamples0U32[i*4+2] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid3 = pInputSamples0U32[i*4+3] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid0 = (mid0 << 1) | (side0 & 0x01); + mid1 = (mid1 << 1) | (side1 & 0x01); + mid2 = (mid2 << 1) | (side2 & 0x01); + mid3 = (mid3 << 1) | (side3 & 0x01); + temp0L = (ma_uint32)((ma_int32)(mid0 + side0) >> 1); + temp1L = (ma_uint32)((ma_int32)(mid1 + side1) >> 1); + temp2L = (ma_uint32)((ma_int32)(mid2 + side2) >> 1); + temp3L = (ma_uint32)((ma_int32)(mid3 + side3) >> 1); + temp0R = (ma_uint32)((ma_int32)(mid0 - side0) >> 1); + temp1R = (ma_uint32)((ma_int32)(mid1 - side1) >> 1); + temp2R = (ma_uint32)((ma_int32)(mid2 - side2) >> 1); + temp3R = (ma_uint32)((ma_int32)(mid3 - side3) >> 1); + pOutputSamples[i*8+0] = (ma_int32)temp0L; + pOutputSamples[i*8+1] = (ma_int32)temp0R; + pOutputSamples[i*8+2] = (ma_int32)temp1L; + pOutputSamples[i*8+3] = (ma_int32)temp1R; + pOutputSamples[i*8+4] = (ma_int32)temp2L; + pOutputSamples[i*8+5] = (ma_int32)temp2R; + pOutputSamples[i*8+6] = (ma_int32)temp3L; + pOutputSamples[i*8+7] = (ma_int32)temp3R; + } + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)((ma_uint32)((ma_int32)(mid + side) >> 1) << unusedBitsPerSample); + pOutputSamples[i*2+1] = (ma_int32)((ma_uint32)((ma_int32)(mid - side) >> 1) << unusedBitsPerSample); + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_mid_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_int32 shift = unusedBitsPerSample; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + if (shift == 0) { + for (i = 0; i < frameCount4; ++i) { + __m128i mid; + __m128i side; + __m128i left; + __m128i right; + mid = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + mid = _mm_or_si128(_mm_slli_epi32(mid, 1), _mm_and_si128(side, _mm_set1_epi32(0x01))); + left = _mm_srai_epi32(_mm_add_epi32(mid, side), 1); + right = _mm_srai_epi32(_mm_sub_epi32(mid, side), 1); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 0), _mm_unpacklo_epi32(left, right)); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 4), _mm_unpackhi_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)(mid + side) >> 1; + pOutputSamples[i*2+1] = (ma_int32)(mid - side) >> 1; + } + } else { + shift -= 1; + for (i = 0; i < frameCount4; ++i) { + __m128i mid; + __m128i side; + __m128i left; + __m128i right; + mid = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + mid = _mm_or_si128(_mm_slli_epi32(mid, 1), _mm_and_si128(side, _mm_set1_epi32(0x01))); + left = _mm_slli_epi32(_mm_add_epi32(mid, side), shift); + right = _mm_slli_epi32(_mm_sub_epi32(mid, side), shift); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 0), _mm_unpacklo_epi32(left, right)); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 4), _mm_unpackhi_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)((mid + side) << shift); + pOutputSamples[i*2+1] = (ma_int32)((mid - side) << shift); + } + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_mid_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_int32 shift = unusedBitsPerSample; + int32x4_t wbpsShift0_4; + int32x4_t wbpsShift1_4; + uint32x4_t one4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + wbpsShift0_4 = vdupq_n_s32(pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + wbpsShift1_4 = vdupq_n_s32(pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + one4 = vdupq_n_u32(1); + if (shift == 0) { + for (i = 0; i < frameCount4; ++i) { + uint32x4_t mid; + uint32x4_t side; + int32x4_t left; + int32x4_t right; + mid = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), wbpsShift0_4); + side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), wbpsShift1_4); + mid = vorrq_u32(vshlq_n_u32(mid, 1), vandq_u32(side, one4)); + left = vshrq_n_s32(vreinterpretq_s32_u32(vaddq_u32(mid, side)), 1); + right = vshrq_n_s32(vreinterpretq_s32_u32(vsubq_u32(mid, side)), 1); + ma_dr_flac__vst2q_s32(pOutputSamples + i*8, vzipq_s32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)(mid + side) >> 1; + pOutputSamples[i*2+1] = (ma_int32)(mid - side) >> 1; + } + } else { + int32x4_t shift4; + shift -= 1; + shift4 = vdupq_n_s32(shift); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t mid; + uint32x4_t side; + int32x4_t left; + int32x4_t right; + mid = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), wbpsShift0_4); + side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), wbpsShift1_4); + mid = vorrq_u32(vshlq_n_u32(mid, 1), vandq_u32(side, one4)); + left = vreinterpretq_s32_u32(vshlq_u32(vaddq_u32(mid, side), shift4)); + right = vreinterpretq_s32_u32(vshlq_u32(vsubq_u32(mid, side), shift4)); + ma_dr_flac__vst2q_s32(pOutputSamples + i*8, vzipq_s32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)((mid + side) << shift); + pOutputSamples[i*2+1] = (ma_int32)((mid - side) << shift); + } + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_mid_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s32__decode_mid_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s32__decode_mid_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_s32__decode_mid_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_s32__decode_mid_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + for (ma_uint64 i = 0; i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int32)((ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample)); + pOutputSamples[i*2+1] = (ma_int32)((ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample)); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 tempL0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 tempL1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 tempL2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 tempL3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 tempR0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 tempR1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 tempR2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 tempR3 = pInputSamples1U32[i*4+3] << shift1; + pOutputSamples[i*8+0] = (ma_int32)tempL0; + pOutputSamples[i*8+1] = (ma_int32)tempR0; + pOutputSamples[i*8+2] = (ma_int32)tempL1; + pOutputSamples[i*8+3] = (ma_int32)tempR1; + pOutputSamples[i*8+4] = (ma_int32)tempL2; + pOutputSamples[i*8+5] = (ma_int32)tempR2; + pOutputSamples[i*8+6] = (ma_int32)tempL3; + pOutputSamples[i*8+7] = (ma_int32)tempR3; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int32)(pInputSamples0U32[i] << shift0); + pOutputSamples[i*2+1] = (ma_int32)(pInputSamples1U32[i] << shift1); + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + for (i = 0; i < frameCount4; ++i) { + __m128i left = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + __m128i right = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 0), _mm_unpacklo_epi32(left, right)); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8 + 4), _mm_unpackhi_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int32)(pInputSamples0U32[i] << shift0); + pOutputSamples[i*2+1] = (ma_int32)(pInputSamples1U32[i] << shift1); + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + int32x4_t shift4_0 = vdupq_n_s32(shift0); + int32x4_t shift4_1 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + int32x4_t left; + int32x4_t right; + left = vreinterpretq_s32_u32(vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift4_0)); + right = vreinterpretq_s32_u32(vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift4_1)); + ma_dr_flac__vst2q_s32(pOutputSamples + i*8, vzipq_s32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int32)(pInputSamples0U32[i] << shift0); + pOutputSamples[i*2+1] = (ma_int32)(pInputSamples1U32[i] << shift1); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int32* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +MA_API ma_uint64 ma_dr_flac_read_pcm_frames_s32(ma_dr_flac* pFlac, ma_uint64 framesToRead, ma_int32* pBufferOut) +{ + ma_uint64 framesRead; + ma_uint32 unusedBitsPerSample; + if (pFlac == NULL || framesToRead == 0) { + return 0; + } + if (pBufferOut == NULL) { + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, framesToRead); + } + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 32); + unusedBitsPerSample = 32 - pFlac->bitsPerSample; + framesRead = 0; + while (framesToRead > 0) { + if (pFlac->currentFLACFrame.pcmFramesRemaining == 0) { + if (!ma_dr_flac__read_and_decode_next_flac_frame(pFlac)) { + break; + } + } else { + unsigned int channelCount = ma_dr_flac__get_channel_count_from_channel_assignment(pFlac->currentFLACFrame.header.channelAssignment); + ma_uint64 iFirstPCMFrame = pFlac->currentFLACFrame.header.blockSizeInPCMFrames - pFlac->currentFLACFrame.pcmFramesRemaining; + ma_uint64 frameCountThisIteration = framesToRead; + if (frameCountThisIteration > pFlac->currentFLACFrame.pcmFramesRemaining) { + frameCountThisIteration = pFlac->currentFLACFrame.pcmFramesRemaining; + } + if (channelCount == 2) { + const ma_int32* pDecodedSamples0 = pFlac->currentFLACFrame.subframes[0].pSamplesS32 + iFirstPCMFrame; + const ma_int32* pDecodedSamples1 = pFlac->currentFLACFrame.subframes[1].pSamplesS32 + iFirstPCMFrame; + switch (pFlac->currentFLACFrame.header.channelAssignment) + { + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_LEFT_SIDE: + { + ma_dr_flac_read_pcm_frames_s32__decode_left_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_RIGHT_SIDE: + { + ma_dr_flac_read_pcm_frames_s32__decode_right_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_MID_SIDE: + { + ma_dr_flac_read_pcm_frames_s32__decode_mid_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_INDEPENDENT: + default: + { + ma_dr_flac_read_pcm_frames_s32__decode_independent_stereo(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + } + } else { + ma_uint64 i; + for (i = 0; i < frameCountThisIteration; ++i) { + unsigned int j; + for (j = 0; j < channelCount; ++j) { + pBufferOut[(i*channelCount)+j] = (ma_int32)((ma_uint32)(pFlac->currentFLACFrame.subframes[j].pSamplesS32[iFirstPCMFrame + i]) << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[j].wastedBitsPerSample)); + } + } + } + framesRead += frameCountThisIteration; + pBufferOut += frameCountThisIteration * channelCount; + framesToRead -= frameCountThisIteration; + pFlac->currentPCMFrame += frameCountThisIteration; + pFlac->currentFLACFrame.pcmFramesRemaining -= (ma_uint32)frameCountThisIteration; + } + } + return framesRead; +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_left_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + for (i = 0; i < frameCount; ++i) { + ma_uint32 left = (ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + ma_uint32 side = (ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + ma_uint32 right = left - side; + left >>= 16; + right >>= 16; + pOutputSamples[i*2+0] = (ma_int16)left; + pOutputSamples[i*2+1] = (ma_int16)right; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_left_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 left0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 left1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 left2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 left3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << shift1; + ma_uint32 right0 = left0 - side0; + ma_uint32 right1 = left1 - side1; + ma_uint32 right2 = left2 - side2; + ma_uint32 right3 = left3 - side3; + left0 >>= 16; + left1 >>= 16; + left2 >>= 16; + left3 >>= 16; + right0 >>= 16; + right1 >>= 16; + right2 >>= 16; + right3 >>= 16; + pOutputSamples[i*8+0] = (ma_int16)left0; + pOutputSamples[i*8+1] = (ma_int16)right0; + pOutputSamples[i*8+2] = (ma_int16)left1; + pOutputSamples[i*8+3] = (ma_int16)right1; + pOutputSamples[i*8+4] = (ma_int16)left2; + pOutputSamples[i*8+5] = (ma_int16)right2; + pOutputSamples[i*8+6] = (ma_int16)left3; + pOutputSamples[i*8+7] = (ma_int16)right3; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + left >>= 16; + right >>= 16; + pOutputSamples[i*2+0] = (ma_int16)left; + pOutputSamples[i*2+1] = (ma_int16)right; + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_left_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + for (i = 0; i < frameCount4; ++i) { + __m128i left = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + __m128i side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + __m128i right = _mm_sub_epi32(left, side); + left = _mm_srai_epi32(left, 16); + right = _mm_srai_epi32(right, 16); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8), ma_dr_flac__mm_packs_interleaved_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + left >>= 16; + right >>= 16; + pOutputSamples[i*2+0] = (ma_int16)left; + pOutputSamples[i*2+1] = (ma_int16)right; + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_left_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + int32x4_t shift0_4; + int32x4_t shift1_4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + shift0_4 = vdupq_n_s32(shift0); + shift1_4 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t left; + uint32x4_t side; + uint32x4_t right; + left = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift0_4); + side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift1_4); + right = vsubq_u32(left, side); + left = vshrq_n_u32(left, 16); + right = vshrq_n_u32(right, 16); + ma_dr_flac__vst2q_u16((ma_uint16*)pOutputSamples + i*8, vzip_u16(vmovn_u32(left), vmovn_u32(right))); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + left >>= 16; + right >>= 16; + pOutputSamples[i*2+0] = (ma_int16)left; + pOutputSamples[i*2+1] = (ma_int16)right; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_left_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s16__decode_left_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s16__decode_left_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_s16__decode_left_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_s16__decode_left_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_right_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + for (i = 0; i < frameCount; ++i) { + ma_uint32 side = (ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + ma_uint32 right = (ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + ma_uint32 left = right + side; + left >>= 16; + right >>= 16; + pOutputSamples[i*2+0] = (ma_int16)left; + pOutputSamples[i*2+1] = (ma_int16)right; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_right_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 side0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 side1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 side2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 side3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 right0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 right1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 right2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 right3 = pInputSamples1U32[i*4+3] << shift1; + ma_uint32 left0 = right0 + side0; + ma_uint32 left1 = right1 + side1; + ma_uint32 left2 = right2 + side2; + ma_uint32 left3 = right3 + side3; + left0 >>= 16; + left1 >>= 16; + left2 >>= 16; + left3 >>= 16; + right0 >>= 16; + right1 >>= 16; + right2 >>= 16; + right3 >>= 16; + pOutputSamples[i*8+0] = (ma_int16)left0; + pOutputSamples[i*8+1] = (ma_int16)right0; + pOutputSamples[i*8+2] = (ma_int16)left1; + pOutputSamples[i*8+3] = (ma_int16)right1; + pOutputSamples[i*8+4] = (ma_int16)left2; + pOutputSamples[i*8+5] = (ma_int16)right2; + pOutputSamples[i*8+6] = (ma_int16)left3; + pOutputSamples[i*8+7] = (ma_int16)right3; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + left >>= 16; + right >>= 16; + pOutputSamples[i*2+0] = (ma_int16)left; + pOutputSamples[i*2+1] = (ma_int16)right; + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_right_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + for (i = 0; i < frameCount4; ++i) { + __m128i side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + __m128i right = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + __m128i left = _mm_add_epi32(right, side); + left = _mm_srai_epi32(left, 16); + right = _mm_srai_epi32(right, 16); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8), ma_dr_flac__mm_packs_interleaved_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + left >>= 16; + right >>= 16; + pOutputSamples[i*2+0] = (ma_int16)left; + pOutputSamples[i*2+1] = (ma_int16)right; + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_right_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + int32x4_t shift0_4; + int32x4_t shift1_4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + shift0_4 = vdupq_n_s32(shift0); + shift1_4 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t side; + uint32x4_t right; + uint32x4_t left; + side = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift0_4); + right = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift1_4); + left = vaddq_u32(right, side); + left = vshrq_n_u32(left, 16); + right = vshrq_n_u32(right, 16); + ma_dr_flac__vst2q_u16((ma_uint16*)pOutputSamples + i*8, vzip_u16(vmovn_u32(left), vmovn_u32(right))); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + left >>= 16; + right >>= 16; + pOutputSamples[i*2+0] = (ma_int16)left; + pOutputSamples[i*2+1] = (ma_int16)right; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_right_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s16__decode_right_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s16__decode_right_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_s16__decode_right_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_s16__decode_right_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_mid_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + for (ma_uint64 i = 0; i < frameCount; ++i) { + ma_uint32 mid = (ma_uint32)pInputSamples0[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = (ma_uint32)pInputSamples1[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int16)(((ma_uint32)((ma_int32)(mid + side) >> 1) << unusedBitsPerSample) >> 16); + pOutputSamples[i*2+1] = (ma_int16)(((ma_uint32)((ma_int32)(mid - side) >> 1) << unusedBitsPerSample) >> 16); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_mid_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift = unusedBitsPerSample; + if (shift > 0) { + shift -= 1; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 temp0L; + ma_uint32 temp1L; + ma_uint32 temp2L; + ma_uint32 temp3L; + ma_uint32 temp0R; + ma_uint32 temp1R; + ma_uint32 temp2R; + ma_uint32 temp3R; + ma_uint32 mid0 = pInputSamples0U32[i*4+0] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid1 = pInputSamples0U32[i*4+1] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid2 = pInputSamples0U32[i*4+2] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid3 = pInputSamples0U32[i*4+3] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid0 = (mid0 << 1) | (side0 & 0x01); + mid1 = (mid1 << 1) | (side1 & 0x01); + mid2 = (mid2 << 1) | (side2 & 0x01); + mid3 = (mid3 << 1) | (side3 & 0x01); + temp0L = (mid0 + side0) << shift; + temp1L = (mid1 + side1) << shift; + temp2L = (mid2 + side2) << shift; + temp3L = (mid3 + side3) << shift; + temp0R = (mid0 - side0) << shift; + temp1R = (mid1 - side1) << shift; + temp2R = (mid2 - side2) << shift; + temp3R = (mid3 - side3) << shift; + temp0L >>= 16; + temp1L >>= 16; + temp2L >>= 16; + temp3L >>= 16; + temp0R >>= 16; + temp1R >>= 16; + temp2R >>= 16; + temp3R >>= 16; + pOutputSamples[i*8+0] = (ma_int16)temp0L; + pOutputSamples[i*8+1] = (ma_int16)temp0R; + pOutputSamples[i*8+2] = (ma_int16)temp1L; + pOutputSamples[i*8+3] = (ma_int16)temp1R; + pOutputSamples[i*8+4] = (ma_int16)temp2L; + pOutputSamples[i*8+5] = (ma_int16)temp2R; + pOutputSamples[i*8+6] = (ma_int16)temp3L; + pOutputSamples[i*8+7] = (ma_int16)temp3R; + } + } else { + for (i = 0; i < frameCount4; ++i) { + ma_uint32 temp0L; + ma_uint32 temp1L; + ma_uint32 temp2L; + ma_uint32 temp3L; + ma_uint32 temp0R; + ma_uint32 temp1R; + ma_uint32 temp2R; + ma_uint32 temp3R; + ma_uint32 mid0 = pInputSamples0U32[i*4+0] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid1 = pInputSamples0U32[i*4+1] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid2 = pInputSamples0U32[i*4+2] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid3 = pInputSamples0U32[i*4+3] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid0 = (mid0 << 1) | (side0 & 0x01); + mid1 = (mid1 << 1) | (side1 & 0x01); + mid2 = (mid2 << 1) | (side2 & 0x01); + mid3 = (mid3 << 1) | (side3 & 0x01); + temp0L = ((ma_int32)(mid0 + side0) >> 1); + temp1L = ((ma_int32)(mid1 + side1) >> 1); + temp2L = ((ma_int32)(mid2 + side2) >> 1); + temp3L = ((ma_int32)(mid3 + side3) >> 1); + temp0R = ((ma_int32)(mid0 - side0) >> 1); + temp1R = ((ma_int32)(mid1 - side1) >> 1); + temp2R = ((ma_int32)(mid2 - side2) >> 1); + temp3R = ((ma_int32)(mid3 - side3) >> 1); + temp0L >>= 16; + temp1L >>= 16; + temp2L >>= 16; + temp3L >>= 16; + temp0R >>= 16; + temp1R >>= 16; + temp2R >>= 16; + temp3R >>= 16; + pOutputSamples[i*8+0] = (ma_int16)temp0L; + pOutputSamples[i*8+1] = (ma_int16)temp0R; + pOutputSamples[i*8+2] = (ma_int16)temp1L; + pOutputSamples[i*8+3] = (ma_int16)temp1R; + pOutputSamples[i*8+4] = (ma_int16)temp2L; + pOutputSamples[i*8+5] = (ma_int16)temp2R; + pOutputSamples[i*8+6] = (ma_int16)temp3L; + pOutputSamples[i*8+7] = (ma_int16)temp3R; + } + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int16)(((ma_uint32)((ma_int32)(mid + side) >> 1) << unusedBitsPerSample) >> 16); + pOutputSamples[i*2+1] = (ma_int16)(((ma_uint32)((ma_int32)(mid - side) >> 1) << unusedBitsPerSample) >> 16); + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_mid_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift = unusedBitsPerSample; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + if (shift == 0) { + for (i = 0; i < frameCount4; ++i) { + __m128i mid; + __m128i side; + __m128i left; + __m128i right; + mid = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + mid = _mm_or_si128(_mm_slli_epi32(mid, 1), _mm_and_si128(side, _mm_set1_epi32(0x01))); + left = _mm_srai_epi32(_mm_add_epi32(mid, side), 1); + right = _mm_srai_epi32(_mm_sub_epi32(mid, side), 1); + left = _mm_srai_epi32(left, 16); + right = _mm_srai_epi32(right, 16); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8), ma_dr_flac__mm_packs_interleaved_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int16)(((ma_int32)(mid + side) >> 1) >> 16); + pOutputSamples[i*2+1] = (ma_int16)(((ma_int32)(mid - side) >> 1) >> 16); + } + } else { + shift -= 1; + for (i = 0; i < frameCount4; ++i) { + __m128i mid; + __m128i side; + __m128i left; + __m128i right; + mid = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + mid = _mm_or_si128(_mm_slli_epi32(mid, 1), _mm_and_si128(side, _mm_set1_epi32(0x01))); + left = _mm_slli_epi32(_mm_add_epi32(mid, side), shift); + right = _mm_slli_epi32(_mm_sub_epi32(mid, side), shift); + left = _mm_srai_epi32(left, 16); + right = _mm_srai_epi32(right, 16); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8), ma_dr_flac__mm_packs_interleaved_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int16)(((mid + side) << shift) >> 16); + pOutputSamples[i*2+1] = (ma_int16)(((mid - side) << shift) >> 16); + } + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_mid_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift = unusedBitsPerSample; + int32x4_t wbpsShift0_4; + int32x4_t wbpsShift1_4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + wbpsShift0_4 = vdupq_n_s32(pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + wbpsShift1_4 = vdupq_n_s32(pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + if (shift == 0) { + for (i = 0; i < frameCount4; ++i) { + uint32x4_t mid; + uint32x4_t side; + int32x4_t left; + int32x4_t right; + mid = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), wbpsShift0_4); + side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), wbpsShift1_4); + mid = vorrq_u32(vshlq_n_u32(mid, 1), vandq_u32(side, vdupq_n_u32(1))); + left = vshrq_n_s32(vreinterpretq_s32_u32(vaddq_u32(mid, side)), 1); + right = vshrq_n_s32(vreinterpretq_s32_u32(vsubq_u32(mid, side)), 1); + left = vshrq_n_s32(left, 16); + right = vshrq_n_s32(right, 16); + ma_dr_flac__vst2q_s16(pOutputSamples + i*8, vzip_s16(vmovn_s32(left), vmovn_s32(right))); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int16)(((ma_int32)(mid + side) >> 1) >> 16); + pOutputSamples[i*2+1] = (ma_int16)(((ma_int32)(mid - side) >> 1) >> 16); + } + } else { + int32x4_t shift4; + shift -= 1; + shift4 = vdupq_n_s32(shift); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t mid; + uint32x4_t side; + int32x4_t left; + int32x4_t right; + mid = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), wbpsShift0_4); + side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), wbpsShift1_4); + mid = vorrq_u32(vshlq_n_u32(mid, 1), vandq_u32(side, vdupq_n_u32(1))); + left = vreinterpretq_s32_u32(vshlq_u32(vaddq_u32(mid, side), shift4)); + right = vreinterpretq_s32_u32(vshlq_u32(vsubq_u32(mid, side), shift4)); + left = vshrq_n_s32(left, 16); + right = vshrq_n_s32(right, 16); + ma_dr_flac__vst2q_s16(pOutputSamples + i*8, vzip_s16(vmovn_s32(left), vmovn_s32(right))); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int16)(((mid + side) << shift) >> 16); + pOutputSamples[i*2+1] = (ma_int16)(((mid - side) << shift) >> 16); + } + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_mid_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s16__decode_mid_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s16__decode_mid_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_s16__decode_mid_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_s16__decode_mid_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + for (ma_uint64 i = 0; i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int16)((ma_int32)((ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample)) >> 16); + pOutputSamples[i*2+1] = (ma_int16)((ma_int32)((ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample)) >> 16); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 tempL0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 tempL1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 tempL2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 tempL3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 tempR0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 tempR1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 tempR2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 tempR3 = pInputSamples1U32[i*4+3] << shift1; + tempL0 >>= 16; + tempL1 >>= 16; + tempL2 >>= 16; + tempL3 >>= 16; + tempR0 >>= 16; + tempR1 >>= 16; + tempR2 >>= 16; + tempR3 >>= 16; + pOutputSamples[i*8+0] = (ma_int16)tempL0; + pOutputSamples[i*8+1] = (ma_int16)tempR0; + pOutputSamples[i*8+2] = (ma_int16)tempL1; + pOutputSamples[i*8+3] = (ma_int16)tempR1; + pOutputSamples[i*8+4] = (ma_int16)tempL2; + pOutputSamples[i*8+5] = (ma_int16)tempR2; + pOutputSamples[i*8+6] = (ma_int16)tempL3; + pOutputSamples[i*8+7] = (ma_int16)tempR3; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int16)((pInputSamples0U32[i] << shift0) >> 16); + pOutputSamples[i*2+1] = (ma_int16)((pInputSamples1U32[i] << shift1) >> 16); + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + for (i = 0; i < frameCount4; ++i) { + __m128i left = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + __m128i right = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + left = _mm_srai_epi32(left, 16); + right = _mm_srai_epi32(right, 16); + _mm_storeu_si128((__m128i*)(pOutputSamples + i*8), ma_dr_flac__mm_packs_interleaved_epi32(left, right)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int16)((pInputSamples0U32[i] << shift0) >> 16); + pOutputSamples[i*2+1] = (ma_int16)((pInputSamples1U32[i] << shift1) >> 16); + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + int32x4_t shift0_4 = vdupq_n_s32(shift0); + int32x4_t shift1_4 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + int32x4_t left; + int32x4_t right; + left = vreinterpretq_s32_u32(vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift0_4)); + right = vreinterpretq_s32_u32(vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift1_4)); + left = vshrq_n_s32(left, 16); + right = vshrq_n_s32(right, 16); + ma_dr_flac__vst2q_s16(pOutputSamples + i*8, vzip_s16(vmovn_s32(left), vmovn_s32(right))); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int16)((pInputSamples0U32[i] << shift0) >> 16); + pOutputSamples[i*2+1] = (ma_int16)((pInputSamples1U32[i] << shift1) >> 16); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, ma_int16* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +MA_API ma_uint64 ma_dr_flac_read_pcm_frames_s16(ma_dr_flac* pFlac, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + ma_uint64 framesRead; + ma_uint32 unusedBitsPerSample; + if (pFlac == NULL || framesToRead == 0) { + return 0; + } + if (pBufferOut == NULL) { + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, framesToRead); + } + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 32); + unusedBitsPerSample = 32 - pFlac->bitsPerSample; + framesRead = 0; + while (framesToRead > 0) { + if (pFlac->currentFLACFrame.pcmFramesRemaining == 0) { + if (!ma_dr_flac__read_and_decode_next_flac_frame(pFlac)) { + break; + } + } else { + unsigned int channelCount = ma_dr_flac__get_channel_count_from_channel_assignment(pFlac->currentFLACFrame.header.channelAssignment); + ma_uint64 iFirstPCMFrame = pFlac->currentFLACFrame.header.blockSizeInPCMFrames - pFlac->currentFLACFrame.pcmFramesRemaining; + ma_uint64 frameCountThisIteration = framesToRead; + if (frameCountThisIteration > pFlac->currentFLACFrame.pcmFramesRemaining) { + frameCountThisIteration = pFlac->currentFLACFrame.pcmFramesRemaining; + } + if (channelCount == 2) { + const ma_int32* pDecodedSamples0 = pFlac->currentFLACFrame.subframes[0].pSamplesS32 + iFirstPCMFrame; + const ma_int32* pDecodedSamples1 = pFlac->currentFLACFrame.subframes[1].pSamplesS32 + iFirstPCMFrame; + switch (pFlac->currentFLACFrame.header.channelAssignment) + { + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_LEFT_SIDE: + { + ma_dr_flac_read_pcm_frames_s16__decode_left_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_RIGHT_SIDE: + { + ma_dr_flac_read_pcm_frames_s16__decode_right_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_MID_SIDE: + { + ma_dr_flac_read_pcm_frames_s16__decode_mid_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_INDEPENDENT: + default: + { + ma_dr_flac_read_pcm_frames_s16__decode_independent_stereo(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + } + } else { + ma_uint64 i; + for (i = 0; i < frameCountThisIteration; ++i) { + unsigned int j; + for (j = 0; j < channelCount; ++j) { + ma_int32 sampleS32 = (ma_int32)((ma_uint32)(pFlac->currentFLACFrame.subframes[j].pSamplesS32[iFirstPCMFrame + i]) << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[j].wastedBitsPerSample)); + pBufferOut[(i*channelCount)+j] = (ma_int16)(sampleS32 >> 16); + } + } + } + framesRead += frameCountThisIteration; + pBufferOut += frameCountThisIteration * channelCount; + framesToRead -= frameCountThisIteration; + pFlac->currentPCMFrame += frameCountThisIteration; + pFlac->currentFLACFrame.pcmFramesRemaining -= (ma_uint32)frameCountThisIteration; + } + } + return framesRead; +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_left_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + for (i = 0; i < frameCount; ++i) { + ma_uint32 left = (ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + ma_uint32 side = (ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + ma_uint32 right = left - side; + pOutputSamples[i*2+0] = (float)((ma_int32)left / 2147483648.0); + pOutputSamples[i*2+1] = (float)((ma_int32)right / 2147483648.0); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_left_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + float factor = 1 / 2147483648.0; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 left0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 left1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 left2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 left3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << shift1; + ma_uint32 right0 = left0 - side0; + ma_uint32 right1 = left1 - side1; + ma_uint32 right2 = left2 - side2; + ma_uint32 right3 = left3 - side3; + pOutputSamples[i*8+0] = (ma_int32)left0 * factor; + pOutputSamples[i*8+1] = (ma_int32)right0 * factor; + pOutputSamples[i*8+2] = (ma_int32)left1 * factor; + pOutputSamples[i*8+3] = (ma_int32)right1 * factor; + pOutputSamples[i*8+4] = (ma_int32)left2 * factor; + pOutputSamples[i*8+5] = (ma_int32)right2 * factor; + pOutputSamples[i*8+6] = (ma_int32)left3 * factor; + pOutputSamples[i*8+7] = (ma_int32)right3 * factor; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + pOutputSamples[i*2+0] = (ma_int32)left * factor; + pOutputSamples[i*2+1] = (ma_int32)right * factor; + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_left_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample) - 8; + ma_uint32 shift1 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample) - 8; + __m128 factor; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + factor = _mm_set1_ps(1.0f / 8388608.0f); + for (i = 0; i < frameCount4; ++i) { + __m128i left = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + __m128i side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + __m128i right = _mm_sub_epi32(left, side); + __m128 leftf = _mm_mul_ps(_mm_cvtepi32_ps(left), factor); + __m128 rightf = _mm_mul_ps(_mm_cvtepi32_ps(right), factor); + _mm_storeu_ps(pOutputSamples + i*8 + 0, _mm_unpacklo_ps(leftf, rightf)); + _mm_storeu_ps(pOutputSamples + i*8 + 4, _mm_unpackhi_ps(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + pOutputSamples[i*2+0] = (ma_int32)left / 8388608.0f; + pOutputSamples[i*2+1] = (ma_int32)right / 8388608.0f; + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_left_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample) - 8; + ma_uint32 shift1 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample) - 8; + float32x4_t factor4; + int32x4_t shift0_4; + int32x4_t shift1_4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + factor4 = vdupq_n_f32(1.0f / 8388608.0f); + shift0_4 = vdupq_n_s32(shift0); + shift1_4 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t left; + uint32x4_t side; + uint32x4_t right; + float32x4_t leftf; + float32x4_t rightf; + left = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift0_4); + side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift1_4); + right = vsubq_u32(left, side); + leftf = vmulq_f32(vcvtq_f32_s32(vreinterpretq_s32_u32(left)), factor4); + rightf = vmulq_f32(vcvtq_f32_s32(vreinterpretq_s32_u32(right)), factor4); + ma_dr_flac__vst2q_f32(pOutputSamples + i*8, vzipq_f32(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 left = pInputSamples0U32[i] << shift0; + ma_uint32 side = pInputSamples1U32[i] << shift1; + ma_uint32 right = left - side; + pOutputSamples[i*2+0] = (ma_int32)left / 8388608.0f; + pOutputSamples[i*2+1] = (ma_int32)right / 8388608.0f; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_left_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_f32__decode_left_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_f32__decode_left_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_f32__decode_left_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_f32__decode_left_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_right_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + for (i = 0; i < frameCount; ++i) { + ma_uint32 side = (ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + ma_uint32 right = (ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + ma_uint32 left = right + side; + pOutputSamples[i*2+0] = (float)((ma_int32)left / 2147483648.0); + pOutputSamples[i*2+1] = (float)((ma_int32)right / 2147483648.0); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_right_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + float factor = 1 / 2147483648.0; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 side0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 side1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 side2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 side3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 right0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 right1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 right2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 right3 = pInputSamples1U32[i*4+3] << shift1; + ma_uint32 left0 = right0 + side0; + ma_uint32 left1 = right1 + side1; + ma_uint32 left2 = right2 + side2; + ma_uint32 left3 = right3 + side3; + pOutputSamples[i*8+0] = (ma_int32)left0 * factor; + pOutputSamples[i*8+1] = (ma_int32)right0 * factor; + pOutputSamples[i*8+2] = (ma_int32)left1 * factor; + pOutputSamples[i*8+3] = (ma_int32)right1 * factor; + pOutputSamples[i*8+4] = (ma_int32)left2 * factor; + pOutputSamples[i*8+5] = (ma_int32)right2 * factor; + pOutputSamples[i*8+6] = (ma_int32)left3 * factor; + pOutputSamples[i*8+7] = (ma_int32)right3 * factor; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + pOutputSamples[i*2+0] = (ma_int32)left * factor; + pOutputSamples[i*2+1] = (ma_int32)right * factor; + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_right_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample) - 8; + ma_uint32 shift1 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample) - 8; + __m128 factor; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + factor = _mm_set1_ps(1.0f / 8388608.0f); + for (i = 0; i < frameCount4; ++i) { + __m128i side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + __m128i right = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + __m128i left = _mm_add_epi32(right, side); + __m128 leftf = _mm_mul_ps(_mm_cvtepi32_ps(left), factor); + __m128 rightf = _mm_mul_ps(_mm_cvtepi32_ps(right), factor); + _mm_storeu_ps(pOutputSamples + i*8 + 0, _mm_unpacklo_ps(leftf, rightf)); + _mm_storeu_ps(pOutputSamples + i*8 + 4, _mm_unpackhi_ps(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + pOutputSamples[i*2+0] = (ma_int32)left / 8388608.0f; + pOutputSamples[i*2+1] = (ma_int32)right / 8388608.0f; + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_right_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample) - 8; + ma_uint32 shift1 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample) - 8; + float32x4_t factor4; + int32x4_t shift0_4; + int32x4_t shift1_4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + factor4 = vdupq_n_f32(1.0f / 8388608.0f); + shift0_4 = vdupq_n_s32(shift0); + shift1_4 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t side; + uint32x4_t right; + uint32x4_t left; + float32x4_t leftf; + float32x4_t rightf; + side = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift0_4); + right = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift1_4); + left = vaddq_u32(right, side); + leftf = vmulq_f32(vcvtq_f32_s32(vreinterpretq_s32_u32(left)), factor4); + rightf = vmulq_f32(vcvtq_f32_s32(vreinterpretq_s32_u32(right)), factor4); + ma_dr_flac__vst2q_f32(pOutputSamples + i*8, vzipq_f32(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 side = pInputSamples0U32[i] << shift0; + ma_uint32 right = pInputSamples1U32[i] << shift1; + ma_uint32 left = right + side; + pOutputSamples[i*2+0] = (ma_int32)left / 8388608.0f; + pOutputSamples[i*2+1] = (ma_int32)right / 8388608.0f; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_right_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_f32__decode_right_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_f32__decode_right_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_f32__decode_right_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_f32__decode_right_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_mid_side__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + for (ma_uint64 i = 0; i < frameCount; ++i) { + ma_uint32 mid = (ma_uint32)pInputSamples0[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = (ma_uint32)pInputSamples1[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (float)((((ma_int32)(mid + side) >> 1) << (unusedBitsPerSample)) / 2147483648.0); + pOutputSamples[i*2+1] = (float)((((ma_int32)(mid - side) >> 1) << (unusedBitsPerSample)) / 2147483648.0); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_mid_side__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift = unusedBitsPerSample; + float factor = 1 / 2147483648.0; + if (shift > 0) { + shift -= 1; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 temp0L; + ma_uint32 temp1L; + ma_uint32 temp2L; + ma_uint32 temp3L; + ma_uint32 temp0R; + ma_uint32 temp1R; + ma_uint32 temp2R; + ma_uint32 temp3R; + ma_uint32 mid0 = pInputSamples0U32[i*4+0] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid1 = pInputSamples0U32[i*4+1] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid2 = pInputSamples0U32[i*4+2] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid3 = pInputSamples0U32[i*4+3] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid0 = (mid0 << 1) | (side0 & 0x01); + mid1 = (mid1 << 1) | (side1 & 0x01); + mid2 = (mid2 << 1) | (side2 & 0x01); + mid3 = (mid3 << 1) | (side3 & 0x01); + temp0L = (mid0 + side0) << shift; + temp1L = (mid1 + side1) << shift; + temp2L = (mid2 + side2) << shift; + temp3L = (mid3 + side3) << shift; + temp0R = (mid0 - side0) << shift; + temp1R = (mid1 - side1) << shift; + temp2R = (mid2 - side2) << shift; + temp3R = (mid3 - side3) << shift; + pOutputSamples[i*8+0] = (ma_int32)temp0L * factor; + pOutputSamples[i*8+1] = (ma_int32)temp0R * factor; + pOutputSamples[i*8+2] = (ma_int32)temp1L * factor; + pOutputSamples[i*8+3] = (ma_int32)temp1R * factor; + pOutputSamples[i*8+4] = (ma_int32)temp2L * factor; + pOutputSamples[i*8+5] = (ma_int32)temp2R * factor; + pOutputSamples[i*8+6] = (ma_int32)temp3L * factor; + pOutputSamples[i*8+7] = (ma_int32)temp3R * factor; + } + } else { + for (i = 0; i < frameCount4; ++i) { + ma_uint32 temp0L; + ma_uint32 temp1L; + ma_uint32 temp2L; + ma_uint32 temp3L; + ma_uint32 temp0R; + ma_uint32 temp1R; + ma_uint32 temp2R; + ma_uint32 temp3R; + ma_uint32 mid0 = pInputSamples0U32[i*4+0] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid1 = pInputSamples0U32[i*4+1] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid2 = pInputSamples0U32[i*4+2] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 mid3 = pInputSamples0U32[i*4+3] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side0 = pInputSamples1U32[i*4+0] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side1 = pInputSamples1U32[i*4+1] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side2 = pInputSamples1U32[i*4+2] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + ma_uint32 side3 = pInputSamples1U32[i*4+3] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid0 = (mid0 << 1) | (side0 & 0x01); + mid1 = (mid1 << 1) | (side1 & 0x01); + mid2 = (mid2 << 1) | (side2 & 0x01); + mid3 = (mid3 << 1) | (side3 & 0x01); + temp0L = (ma_uint32)((ma_int32)(mid0 + side0) >> 1); + temp1L = (ma_uint32)((ma_int32)(mid1 + side1) >> 1); + temp2L = (ma_uint32)((ma_int32)(mid2 + side2) >> 1); + temp3L = (ma_uint32)((ma_int32)(mid3 + side3) >> 1); + temp0R = (ma_uint32)((ma_int32)(mid0 - side0) >> 1); + temp1R = (ma_uint32)((ma_int32)(mid1 - side1) >> 1); + temp2R = (ma_uint32)((ma_int32)(mid2 - side2) >> 1); + temp3R = (ma_uint32)((ma_int32)(mid3 - side3) >> 1); + pOutputSamples[i*8+0] = (ma_int32)temp0L * factor; + pOutputSamples[i*8+1] = (ma_int32)temp0R * factor; + pOutputSamples[i*8+2] = (ma_int32)temp1L * factor; + pOutputSamples[i*8+3] = (ma_int32)temp1R * factor; + pOutputSamples[i*8+4] = (ma_int32)temp2L * factor; + pOutputSamples[i*8+5] = (ma_int32)temp2R * factor; + pOutputSamples[i*8+6] = (ma_int32)temp3L * factor; + pOutputSamples[i*8+7] = (ma_int32)temp3R * factor; + } + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)((ma_uint32)((ma_int32)(mid + side) >> 1) << unusedBitsPerSample) * factor; + pOutputSamples[i*2+1] = (ma_int32)((ma_uint32)((ma_int32)(mid - side) >> 1) << unusedBitsPerSample) * factor; + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_mid_side__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift = unusedBitsPerSample - 8; + float factor; + __m128 factor128; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + factor = 1.0f / 8388608.0f; + factor128 = _mm_set1_ps(factor); + if (shift == 0) { + for (i = 0; i < frameCount4; ++i) { + __m128i mid; + __m128i side; + __m128i tempL; + __m128i tempR; + __m128 leftf; + __m128 rightf; + mid = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + mid = _mm_or_si128(_mm_slli_epi32(mid, 1), _mm_and_si128(side, _mm_set1_epi32(0x01))); + tempL = _mm_srai_epi32(_mm_add_epi32(mid, side), 1); + tempR = _mm_srai_epi32(_mm_sub_epi32(mid, side), 1); + leftf = _mm_mul_ps(_mm_cvtepi32_ps(tempL), factor128); + rightf = _mm_mul_ps(_mm_cvtepi32_ps(tempR), factor128); + _mm_storeu_ps(pOutputSamples + i*8 + 0, _mm_unpacklo_ps(leftf, rightf)); + _mm_storeu_ps(pOutputSamples + i*8 + 4, _mm_unpackhi_ps(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = ((ma_int32)(mid + side) >> 1) * factor; + pOutputSamples[i*2+1] = ((ma_int32)(mid - side) >> 1) * factor; + } + } else { + shift -= 1; + for (i = 0; i < frameCount4; ++i) { + __m128i mid; + __m128i side; + __m128i tempL; + __m128i tempR; + __m128 leftf; + __m128 rightf; + mid = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + side = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + mid = _mm_or_si128(_mm_slli_epi32(mid, 1), _mm_and_si128(side, _mm_set1_epi32(0x01))); + tempL = _mm_slli_epi32(_mm_add_epi32(mid, side), shift); + tempR = _mm_slli_epi32(_mm_sub_epi32(mid, side), shift); + leftf = _mm_mul_ps(_mm_cvtepi32_ps(tempL), factor128); + rightf = _mm_mul_ps(_mm_cvtepi32_ps(tempR), factor128); + _mm_storeu_ps(pOutputSamples + i*8 + 0, _mm_unpacklo_ps(leftf, rightf)); + _mm_storeu_ps(pOutputSamples + i*8 + 4, _mm_unpackhi_ps(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)((mid + side) << shift) * factor; + pOutputSamples[i*2+1] = (ma_int32)((mid - side) << shift) * factor; + } + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_mid_side__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift = unusedBitsPerSample - 8; + float factor; + float32x4_t factor4; + int32x4_t shift4; + int32x4_t wbps0_4; + int32x4_t wbps1_4; + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 24); + factor = 1.0f / 8388608.0f; + factor4 = vdupq_n_f32(factor); + wbps0_4 = vdupq_n_s32(pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample); + wbps1_4 = vdupq_n_s32(pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample); + if (shift == 0) { + for (i = 0; i < frameCount4; ++i) { + int32x4_t lefti; + int32x4_t righti; + float32x4_t leftf; + float32x4_t rightf; + uint32x4_t mid = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), wbps0_4); + uint32x4_t side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), wbps1_4); + mid = vorrq_u32(vshlq_n_u32(mid, 1), vandq_u32(side, vdupq_n_u32(1))); + lefti = vshrq_n_s32(vreinterpretq_s32_u32(vaddq_u32(mid, side)), 1); + righti = vshrq_n_s32(vreinterpretq_s32_u32(vsubq_u32(mid, side)), 1); + leftf = vmulq_f32(vcvtq_f32_s32(lefti), factor4); + rightf = vmulq_f32(vcvtq_f32_s32(righti), factor4); + ma_dr_flac__vst2q_f32(pOutputSamples + i*8, vzipq_f32(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = ((ma_int32)(mid + side) >> 1) * factor; + pOutputSamples[i*2+1] = ((ma_int32)(mid - side) >> 1) * factor; + } + } else { + shift -= 1; + shift4 = vdupq_n_s32(shift); + for (i = 0; i < frameCount4; ++i) { + uint32x4_t mid; + uint32x4_t side; + int32x4_t lefti; + int32x4_t righti; + float32x4_t leftf; + float32x4_t rightf; + mid = vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), wbps0_4); + side = vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), wbps1_4); + mid = vorrq_u32(vshlq_n_u32(mid, 1), vandq_u32(side, vdupq_n_u32(1))); + lefti = vreinterpretq_s32_u32(vshlq_u32(vaddq_u32(mid, side), shift4)); + righti = vreinterpretq_s32_u32(vshlq_u32(vsubq_u32(mid, side), shift4)); + leftf = vmulq_f32(vcvtq_f32_s32(lefti), factor4); + rightf = vmulq_f32(vcvtq_f32_s32(righti), factor4); + ma_dr_flac__vst2q_f32(pOutputSamples + i*8, vzipq_f32(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + ma_uint32 mid = pInputSamples0U32[i] << pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 side = pInputSamples1U32[i] << pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + mid = (mid << 1) | (side & 0x01); + pOutputSamples[i*2+0] = (ma_int32)((mid + side) << shift) * factor; + pOutputSamples[i*2+1] = (ma_int32)((mid - side) << shift) * factor; + } + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_mid_side(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_f32__decode_mid_side__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_f32__decode_mid_side__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_f32__decode_mid_side__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_f32__decode_mid_side__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +#if 0 +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo__reference(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + for (ma_uint64 i = 0; i < frameCount; ++i) { + pOutputSamples[i*2+0] = (float)((ma_int32)((ma_uint32)pInputSamples0[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample)) / 2147483648.0); + pOutputSamples[i*2+1] = (float)((ma_int32)((ma_uint32)pInputSamples1[i] << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample)) / 2147483648.0); + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo__scalar(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample; + ma_uint32 shift1 = unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample; + float factor = 1 / 2147483648.0; + for (i = 0; i < frameCount4; ++i) { + ma_uint32 tempL0 = pInputSamples0U32[i*4+0] << shift0; + ma_uint32 tempL1 = pInputSamples0U32[i*4+1] << shift0; + ma_uint32 tempL2 = pInputSamples0U32[i*4+2] << shift0; + ma_uint32 tempL3 = pInputSamples0U32[i*4+3] << shift0; + ma_uint32 tempR0 = pInputSamples1U32[i*4+0] << shift1; + ma_uint32 tempR1 = pInputSamples1U32[i*4+1] << shift1; + ma_uint32 tempR2 = pInputSamples1U32[i*4+2] << shift1; + ma_uint32 tempR3 = pInputSamples1U32[i*4+3] << shift1; + pOutputSamples[i*8+0] = (ma_int32)tempL0 * factor; + pOutputSamples[i*8+1] = (ma_int32)tempR0 * factor; + pOutputSamples[i*8+2] = (ma_int32)tempL1 * factor; + pOutputSamples[i*8+3] = (ma_int32)tempR1 * factor; + pOutputSamples[i*8+4] = (ma_int32)tempL2 * factor; + pOutputSamples[i*8+5] = (ma_int32)tempR2 * factor; + pOutputSamples[i*8+6] = (ma_int32)tempL3 * factor; + pOutputSamples[i*8+7] = (ma_int32)tempR3 * factor; + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int32)(pInputSamples0U32[i] << shift0) * factor; + pOutputSamples[i*2+1] = (ma_int32)(pInputSamples1U32[i] << shift1) * factor; + } +} +#if defined(MA_DR_FLAC_SUPPORT_SSE2) +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo__sse2(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample) - 8; + ma_uint32 shift1 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample) - 8; + float factor = 1.0f / 8388608.0f; + __m128 factor128 = _mm_set1_ps(factor); + for (i = 0; i < frameCount4; ++i) { + __m128i lefti; + __m128i righti; + __m128 leftf; + __m128 rightf; + lefti = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples0 + i), shift0); + righti = _mm_slli_epi32(_mm_loadu_si128((const __m128i*)pInputSamples1 + i), shift1); + leftf = _mm_mul_ps(_mm_cvtepi32_ps(lefti), factor128); + rightf = _mm_mul_ps(_mm_cvtepi32_ps(righti), factor128); + _mm_storeu_ps(pOutputSamples + i*8 + 0, _mm_unpacklo_ps(leftf, rightf)); + _mm_storeu_ps(pOutputSamples + i*8 + 4, _mm_unpackhi_ps(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int32)(pInputSamples0U32[i] << shift0) * factor; + pOutputSamples[i*2+1] = (ma_int32)(pInputSamples1U32[i] << shift1) * factor; + } +} +#endif +#if defined(MA_DR_FLAC_SUPPORT_NEON) +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo__neon(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ + ma_uint64 i; + ma_uint64 frameCount4 = frameCount >> 2; + const ma_uint32* pInputSamples0U32 = (const ma_uint32*)pInputSamples0; + const ma_uint32* pInputSamples1U32 = (const ma_uint32*)pInputSamples1; + ma_uint32 shift0 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[0].wastedBitsPerSample) - 8; + ma_uint32 shift1 = (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[1].wastedBitsPerSample) - 8; + float factor = 1.0f / 8388608.0f; + float32x4_t factor4 = vdupq_n_f32(factor); + int32x4_t shift0_4 = vdupq_n_s32(shift0); + int32x4_t shift1_4 = vdupq_n_s32(shift1); + for (i = 0; i < frameCount4; ++i) { + int32x4_t lefti; + int32x4_t righti; + float32x4_t leftf; + float32x4_t rightf; + lefti = vreinterpretq_s32_u32(vshlq_u32(vld1q_u32(pInputSamples0U32 + i*4), shift0_4)); + righti = vreinterpretq_s32_u32(vshlq_u32(vld1q_u32(pInputSamples1U32 + i*4), shift1_4)); + leftf = vmulq_f32(vcvtq_f32_s32(lefti), factor4); + rightf = vmulq_f32(vcvtq_f32_s32(righti), factor4); + ma_dr_flac__vst2q_f32(pOutputSamples + i*8, vzipq_f32(leftf, rightf)); + } + for (i = (frameCount4 << 2); i < frameCount; ++i) { + pOutputSamples[i*2+0] = (ma_int32)(pInputSamples0U32[i] << shift0) * factor; + pOutputSamples[i*2+1] = (ma_int32)(pInputSamples1U32[i] << shift1) * factor; + } +} +#endif +static MA_INLINE void ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo(ma_dr_flac* pFlac, ma_uint64 frameCount, ma_uint32 unusedBitsPerSample, const ma_int32* pInputSamples0, const ma_int32* pInputSamples1, float* pOutputSamples) +{ +#if defined(MA_DR_FLAC_SUPPORT_SSE2) + if (ma_dr_flac__gIsSSE2Supported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo__sse2(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#elif defined(MA_DR_FLAC_SUPPORT_NEON) + if (ma_dr_flac__gIsNEONSupported && pFlac->bitsPerSample <= 24) { + ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo__neon(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); + } else +#endif + { +#if 0 + ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo__reference(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#else + ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo__scalar(pFlac, frameCount, unusedBitsPerSample, pInputSamples0, pInputSamples1, pOutputSamples); +#endif + } +} +MA_API ma_uint64 ma_dr_flac_read_pcm_frames_f32(ma_dr_flac* pFlac, ma_uint64 framesToRead, float* pBufferOut) +{ + ma_uint64 framesRead; + ma_uint32 unusedBitsPerSample; + if (pFlac == NULL || framesToRead == 0) { + return 0; + } + if (pBufferOut == NULL) { + return ma_dr_flac__seek_forward_by_pcm_frames(pFlac, framesToRead); + } + MA_DR_FLAC_ASSERT(pFlac->bitsPerSample <= 32); + unusedBitsPerSample = 32 - pFlac->bitsPerSample; + framesRead = 0; + while (framesToRead > 0) { + if (pFlac->currentFLACFrame.pcmFramesRemaining == 0) { + if (!ma_dr_flac__read_and_decode_next_flac_frame(pFlac)) { + break; + } + } else { + unsigned int channelCount = ma_dr_flac__get_channel_count_from_channel_assignment(pFlac->currentFLACFrame.header.channelAssignment); + ma_uint64 iFirstPCMFrame = pFlac->currentFLACFrame.header.blockSizeInPCMFrames - pFlac->currentFLACFrame.pcmFramesRemaining; + ma_uint64 frameCountThisIteration = framesToRead; + if (frameCountThisIteration > pFlac->currentFLACFrame.pcmFramesRemaining) { + frameCountThisIteration = pFlac->currentFLACFrame.pcmFramesRemaining; + } + if (channelCount == 2) { + const ma_int32* pDecodedSamples0 = pFlac->currentFLACFrame.subframes[0].pSamplesS32 + iFirstPCMFrame; + const ma_int32* pDecodedSamples1 = pFlac->currentFLACFrame.subframes[1].pSamplesS32 + iFirstPCMFrame; + switch (pFlac->currentFLACFrame.header.channelAssignment) + { + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_LEFT_SIDE: + { + ma_dr_flac_read_pcm_frames_f32__decode_left_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_RIGHT_SIDE: + { + ma_dr_flac_read_pcm_frames_f32__decode_right_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_MID_SIDE: + { + ma_dr_flac_read_pcm_frames_f32__decode_mid_side(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + case MA_DR_FLAC_CHANNEL_ASSIGNMENT_INDEPENDENT: + default: + { + ma_dr_flac_read_pcm_frames_f32__decode_independent_stereo(pFlac, frameCountThisIteration, unusedBitsPerSample, pDecodedSamples0, pDecodedSamples1, pBufferOut); + } break; + } + } else { + ma_uint64 i; + for (i = 0; i < frameCountThisIteration; ++i) { + unsigned int j; + for (j = 0; j < channelCount; ++j) { + ma_int32 sampleS32 = (ma_int32)((ma_uint32)(pFlac->currentFLACFrame.subframes[j].pSamplesS32[iFirstPCMFrame + i]) << (unusedBitsPerSample + pFlac->currentFLACFrame.subframes[j].wastedBitsPerSample)); + pBufferOut[(i*channelCount)+j] = (float)(sampleS32 / 2147483648.0); + } + } + } + framesRead += frameCountThisIteration; + pBufferOut += frameCountThisIteration * channelCount; + framesToRead -= frameCountThisIteration; + pFlac->currentPCMFrame += frameCountThisIteration; + pFlac->currentFLACFrame.pcmFramesRemaining -= (unsigned int)frameCountThisIteration; + } + } + return framesRead; +} +MA_API ma_bool32 ma_dr_flac_seek_to_pcm_frame(ma_dr_flac* pFlac, ma_uint64 pcmFrameIndex) +{ + if (pFlac == NULL) { + return MA_FALSE; + } + if (pFlac->currentPCMFrame == pcmFrameIndex) { + return MA_TRUE; + } + if (pFlac->firstFLACFramePosInBytes == 0) { + return MA_FALSE; + } + if (pcmFrameIndex == 0) { + pFlac->currentPCMFrame = 0; + return ma_dr_flac__seek_to_first_frame(pFlac); + } else { + ma_bool32 wasSuccessful = MA_FALSE; + ma_uint64 originalPCMFrame = pFlac->currentPCMFrame; + if (pcmFrameIndex > pFlac->totalPCMFrameCount) { + pcmFrameIndex = pFlac->totalPCMFrameCount; + } + if (pcmFrameIndex > pFlac->currentPCMFrame) { + ma_uint32 offset = (ma_uint32)(pcmFrameIndex - pFlac->currentPCMFrame); + if (pFlac->currentFLACFrame.pcmFramesRemaining > offset) { + pFlac->currentFLACFrame.pcmFramesRemaining -= offset; + pFlac->currentPCMFrame = pcmFrameIndex; + return MA_TRUE; + } + } else { + ma_uint32 offsetAbs = (ma_uint32)(pFlac->currentPCMFrame - pcmFrameIndex); + ma_uint32 currentFLACFramePCMFrameCount = pFlac->currentFLACFrame.header.blockSizeInPCMFrames; + ma_uint32 currentFLACFramePCMFramesConsumed = currentFLACFramePCMFrameCount - pFlac->currentFLACFrame.pcmFramesRemaining; + if (currentFLACFramePCMFramesConsumed > offsetAbs) { + pFlac->currentFLACFrame.pcmFramesRemaining += offsetAbs; + pFlac->currentPCMFrame = pcmFrameIndex; + return MA_TRUE; + } + } +#ifndef MA_DR_FLAC_NO_OGG + if (pFlac->container == ma_dr_flac_container_ogg) + { + wasSuccessful = ma_dr_flac_ogg__seek_to_pcm_frame(pFlac, pcmFrameIndex); + } + else +#endif + { + if (!pFlac->_noSeekTableSeek) { + wasSuccessful = ma_dr_flac__seek_to_pcm_frame__seek_table(pFlac, pcmFrameIndex); + } +#if !defined(MA_DR_FLAC_NO_CRC) + if (!wasSuccessful && !pFlac->_noBinarySearchSeek && pFlac->totalPCMFrameCount > 0) { + wasSuccessful = ma_dr_flac__seek_to_pcm_frame__binary_search(pFlac, pcmFrameIndex); + } +#endif + if (!wasSuccessful && !pFlac->_noBruteForceSeek) { + wasSuccessful = ma_dr_flac__seek_to_pcm_frame__brute_force(pFlac, pcmFrameIndex); + } + } + if (wasSuccessful) { + pFlac->currentPCMFrame = pcmFrameIndex; + } else { + if (ma_dr_flac_seek_to_pcm_frame(pFlac, originalPCMFrame) == MA_FALSE) { + ma_dr_flac_seek_to_pcm_frame(pFlac, 0); + } + } + return wasSuccessful; + } +} +#define MA_DR_FLAC_DEFINE_FULL_READ_AND_CLOSE(extension, type) \ +static type* ma_dr_flac__full_read_and_close_ ## extension (ma_dr_flac* pFlac, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalPCMFrameCountOut)\ +{ \ + type* pSampleData = NULL; \ + ma_uint64 totalPCMFrameCount; \ + \ + MA_DR_FLAC_ASSERT(pFlac != NULL); \ + \ + totalPCMFrameCount = pFlac->totalPCMFrameCount; \ + \ + if (totalPCMFrameCount == 0) { \ + type buffer[4096]; \ + ma_uint64 pcmFramesRead; \ + size_t sampleDataBufferSize = sizeof(buffer); \ + \ + pSampleData = (type*)ma_dr_flac__malloc_from_callbacks(sampleDataBufferSize, &pFlac->allocationCallbacks); \ + if (pSampleData == NULL) { \ + goto on_error; \ + } \ + \ + while ((pcmFramesRead = (ma_uint64)ma_dr_flac_read_pcm_frames_##extension(pFlac, sizeof(buffer)/sizeof(buffer[0])/pFlac->channels, buffer)) > 0) { \ + if (((totalPCMFrameCount + pcmFramesRead) * pFlac->channels * sizeof(type)) > sampleDataBufferSize) { \ + type* pNewSampleData; \ + size_t newSampleDataBufferSize; \ + \ + newSampleDataBufferSize = sampleDataBufferSize * 2; \ + pNewSampleData = (type*)ma_dr_flac__realloc_from_callbacks(pSampleData, newSampleDataBufferSize, sampleDataBufferSize, &pFlac->allocationCallbacks); \ + if (pNewSampleData == NULL) { \ + ma_dr_flac__free_from_callbacks(pSampleData, &pFlac->allocationCallbacks); \ + goto on_error; \ + } \ + \ + sampleDataBufferSize = newSampleDataBufferSize; \ + pSampleData = pNewSampleData; \ + } \ + \ + MA_DR_FLAC_COPY_MEMORY(pSampleData + (totalPCMFrameCount*pFlac->channels), buffer, (size_t)(pcmFramesRead*pFlac->channels*sizeof(type))); \ + totalPCMFrameCount += pcmFramesRead; \ + } \ + \ + \ + MA_DR_FLAC_ZERO_MEMORY(pSampleData + (totalPCMFrameCount*pFlac->channels), (size_t)(sampleDataBufferSize - totalPCMFrameCount*pFlac->channels*sizeof(type))); \ + } else { \ + ma_uint64 dataSize = totalPCMFrameCount*pFlac->channels*sizeof(type); \ + if (dataSize > (ma_uint64)MA_SIZE_MAX) { \ + goto on_error; \ + } \ + \ + pSampleData = (type*)ma_dr_flac__malloc_from_callbacks((size_t)dataSize, &pFlac->allocationCallbacks); \ + if (pSampleData == NULL) { \ + goto on_error; \ + } \ + \ + totalPCMFrameCount = ma_dr_flac_read_pcm_frames_##extension(pFlac, pFlac->totalPCMFrameCount, pSampleData); \ + } \ + \ + if (sampleRateOut) *sampleRateOut = pFlac->sampleRate; \ + if (channelsOut) *channelsOut = pFlac->channels; \ + if (totalPCMFrameCountOut) *totalPCMFrameCountOut = totalPCMFrameCount; \ + \ + ma_dr_flac_close(pFlac); \ + return pSampleData; \ + \ +on_error: \ + ma_dr_flac_close(pFlac); \ + return NULL; \ +} +MA_DR_FLAC_DEFINE_FULL_READ_AND_CLOSE(s32, ma_int32) +MA_DR_FLAC_DEFINE_FULL_READ_AND_CLOSE(s16, ma_int16) +MA_DR_FLAC_DEFINE_FULL_READ_AND_CLOSE(f32, float) +MA_API ma_int32* ma_dr_flac_open_and_read_pcm_frames_s32(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalPCMFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalPCMFrameCountOut) { + *totalPCMFrameCountOut = 0; + } + pFlac = ma_dr_flac_open(onRead, onSeek, pUserData, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_s32(pFlac, channelsOut, sampleRateOut, totalPCMFrameCountOut); +} +MA_API ma_int16* ma_dr_flac_open_and_read_pcm_frames_s16(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalPCMFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalPCMFrameCountOut) { + *totalPCMFrameCountOut = 0; + } + pFlac = ma_dr_flac_open(onRead, onSeek, pUserData, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_s16(pFlac, channelsOut, sampleRateOut, totalPCMFrameCountOut); +} +MA_API float* ma_dr_flac_open_and_read_pcm_frames_f32(ma_dr_flac_read_proc onRead, ma_dr_flac_seek_proc onSeek, void* pUserData, unsigned int* channelsOut, unsigned int* sampleRateOut, ma_uint64* totalPCMFrameCountOut, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (channelsOut) { + *channelsOut = 0; + } + if (sampleRateOut) { + *sampleRateOut = 0; + } + if (totalPCMFrameCountOut) { + *totalPCMFrameCountOut = 0; + } + pFlac = ma_dr_flac_open(onRead, onSeek, pUserData, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_f32(pFlac, channelsOut, sampleRateOut, totalPCMFrameCountOut); +} +#ifndef MA_DR_FLAC_NO_STDIO +MA_API ma_int32* ma_dr_flac_open_file_and_read_pcm_frames_s32(const char* filename, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (sampleRate) { + *sampleRate = 0; + } + if (channels) { + *channels = 0; + } + if (totalPCMFrameCount) { + *totalPCMFrameCount = 0; + } + pFlac = ma_dr_flac_open_file(filename, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_s32(pFlac, channels, sampleRate, totalPCMFrameCount); +} +MA_API ma_int16* ma_dr_flac_open_file_and_read_pcm_frames_s16(const char* filename, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (sampleRate) { + *sampleRate = 0; + } + if (channels) { + *channels = 0; + } + if (totalPCMFrameCount) { + *totalPCMFrameCount = 0; + } + pFlac = ma_dr_flac_open_file(filename, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_s16(pFlac, channels, sampleRate, totalPCMFrameCount); +} +MA_API float* ma_dr_flac_open_file_and_read_pcm_frames_f32(const char* filename, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (sampleRate) { + *sampleRate = 0; + } + if (channels) { + *channels = 0; + } + if (totalPCMFrameCount) { + *totalPCMFrameCount = 0; + } + pFlac = ma_dr_flac_open_file(filename, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_f32(pFlac, channels, sampleRate, totalPCMFrameCount); +} +#endif +MA_API ma_int32* ma_dr_flac_open_memory_and_read_pcm_frames_s32(const void* data, size_t dataSize, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (sampleRate) { + *sampleRate = 0; + } + if (channels) { + *channels = 0; + } + if (totalPCMFrameCount) { + *totalPCMFrameCount = 0; + } + pFlac = ma_dr_flac_open_memory(data, dataSize, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_s32(pFlac, channels, sampleRate, totalPCMFrameCount); +} +MA_API ma_int16* ma_dr_flac_open_memory_and_read_pcm_frames_s16(const void* data, size_t dataSize, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (sampleRate) { + *sampleRate = 0; + } + if (channels) { + *channels = 0; + } + if (totalPCMFrameCount) { + *totalPCMFrameCount = 0; + } + pFlac = ma_dr_flac_open_memory(data, dataSize, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_s16(pFlac, channels, sampleRate, totalPCMFrameCount); +} +MA_API float* ma_dr_flac_open_memory_and_read_pcm_frames_f32(const void* data, size_t dataSize, unsigned int* channels, unsigned int* sampleRate, ma_uint64* totalPCMFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_flac* pFlac; + if (sampleRate) { + *sampleRate = 0; + } + if (channels) { + *channels = 0; + } + if (totalPCMFrameCount) { + *totalPCMFrameCount = 0; + } + pFlac = ma_dr_flac_open_memory(data, dataSize, pAllocationCallbacks); + if (pFlac == NULL) { + return NULL; + } + return ma_dr_flac__full_read_and_close_f32(pFlac, channels, sampleRate, totalPCMFrameCount); +} +MA_API void ma_dr_flac_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks != NULL) { + ma_dr_flac__free_from_callbacks(p, pAllocationCallbacks); + } else { + ma_dr_flac__free_default(p, NULL); + } +} +MA_API void ma_dr_flac_init_vorbis_comment_iterator(ma_dr_flac_vorbis_comment_iterator* pIter, ma_uint32 commentCount, const void* pComments) +{ + if (pIter == NULL) { + return; + } + pIter->countRemaining = commentCount; + pIter->pRunningData = (const char*)pComments; +} +MA_API const char* ma_dr_flac_next_vorbis_comment(ma_dr_flac_vorbis_comment_iterator* pIter, ma_uint32* pCommentLengthOut) +{ + ma_int32 length; + const char* pComment; + if (pCommentLengthOut) { + *pCommentLengthOut = 0; + } + if (pIter == NULL || pIter->countRemaining == 0 || pIter->pRunningData == NULL) { + return NULL; + } + length = ma_dr_flac__le2host_32_ptr_unaligned(pIter->pRunningData); + pIter->pRunningData += 4; + pComment = pIter->pRunningData; + pIter->pRunningData += length; + pIter->countRemaining -= 1; + if (pCommentLengthOut) { + *pCommentLengthOut = length; + } + return pComment; +} +MA_API void ma_dr_flac_init_cuesheet_track_iterator(ma_dr_flac_cuesheet_track_iterator* pIter, ma_uint32 trackCount, const void* pTrackData) +{ + if (pIter == NULL) { + return; + } + pIter->countRemaining = trackCount; + pIter->pRunningData = (const char*)pTrackData; +} +MA_API ma_bool32 ma_dr_flac_next_cuesheet_track(ma_dr_flac_cuesheet_track_iterator* pIter, ma_dr_flac_cuesheet_track* pCuesheetTrack) +{ + ma_dr_flac_cuesheet_track cuesheetTrack; + const char* pRunningData; + ma_uint64 offsetHi; + ma_uint64 offsetLo; + if (pIter == NULL || pIter->countRemaining == 0 || pIter->pRunningData == NULL) { + return MA_FALSE; + } + pRunningData = pIter->pRunningData; + offsetHi = ma_dr_flac__be2host_32(*(const ma_uint32*)pRunningData); pRunningData += 4; + offsetLo = ma_dr_flac__be2host_32(*(const ma_uint32*)pRunningData); pRunningData += 4; + cuesheetTrack.offset = offsetLo | (offsetHi << 32); + cuesheetTrack.trackNumber = pRunningData[0]; pRunningData += 1; + MA_DR_FLAC_COPY_MEMORY(cuesheetTrack.ISRC, pRunningData, sizeof(cuesheetTrack.ISRC)); pRunningData += 12; + cuesheetTrack.isAudio = (pRunningData[0] & 0x80) != 0; + cuesheetTrack.preEmphasis = (pRunningData[0] & 0x40) != 0; pRunningData += 14; + cuesheetTrack.indexCount = pRunningData[0]; pRunningData += 1; + cuesheetTrack.pIndexPoints = (const ma_dr_flac_cuesheet_track_index*)pRunningData; pRunningData += cuesheetTrack.indexCount * sizeof(ma_dr_flac_cuesheet_track_index); + pIter->pRunningData = pRunningData; + pIter->countRemaining -= 1; + if (pCuesheetTrack) { + *pCuesheetTrack = cuesheetTrack; + } + return MA_TRUE; +} +#if defined(__clang__) || (defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6))) + #pragma GCC diagnostic pop +#endif +#endif +/* dr_flac_c end */ +#endif /* MA_DR_FLAC_IMPLEMENTATION */ +#endif /* MA_NO_FLAC */ + +#if !defined(MA_NO_MP3) && !defined(MA_NO_DECODING) +#if !defined(MA_DR_MP3_IMPLEMENTATION) +/* dr_mp3_c begin */ +#ifndef ma_dr_mp3_c +#define ma_dr_mp3_c +#include +#include +#include +MA_API void ma_dr_mp3_version(ma_uint32* pMajor, ma_uint32* pMinor, ma_uint32* pRevision) +{ + if (pMajor) { + *pMajor = MA_DR_MP3_VERSION_MAJOR; + } + if (pMinor) { + *pMinor = MA_DR_MP3_VERSION_MINOR; + } + if (pRevision) { + *pRevision = MA_DR_MP3_VERSION_REVISION; + } +} +MA_API const char* ma_dr_mp3_version_string(void) +{ + return MA_DR_MP3_VERSION_STRING; +} +#if defined(__TINYC__) +#define MA_DR_MP3_NO_SIMD +#endif +#define MA_DR_MP3_OFFSET_PTR(p, offset) ((void*)((ma_uint8*)(p) + (offset))) +#define MA_DR_MP3_MAX_FREE_FORMAT_FRAME_SIZE 2304 +#ifndef MA_DR_MP3_MAX_FRAME_SYNC_MATCHES +#define MA_DR_MP3_MAX_FRAME_SYNC_MATCHES 10 +#endif +#define MA_DR_MP3_MAX_L3_FRAME_PAYLOAD_BYTES MA_DR_MP3_MAX_FREE_FORMAT_FRAME_SIZE +#define MA_DR_MP3_MAX_BITRESERVOIR_BYTES 511 +#define MA_DR_MP3_SHORT_BLOCK_TYPE 2 +#define MA_DR_MP3_STOP_BLOCK_TYPE 3 +#define MA_DR_MP3_MODE_MONO 3 +#define MA_DR_MP3_MODE_JOINT_STEREO 1 +#define MA_DR_MP3_HDR_SIZE 4 +#define MA_DR_MP3_HDR_IS_MONO(h) (((h[3]) & 0xC0) == 0xC0) +#define MA_DR_MP3_HDR_IS_MS_STEREO(h) (((h[3]) & 0xE0) == 0x60) +#define MA_DR_MP3_HDR_IS_FREE_FORMAT(h) (((h[2]) & 0xF0) == 0) +#define MA_DR_MP3_HDR_IS_CRC(h) (!((h[1]) & 1)) +#define MA_DR_MP3_HDR_TEST_PADDING(h) ((h[2]) & 0x2) +#define MA_DR_MP3_HDR_TEST_MPEG1(h) ((h[1]) & 0x8) +#define MA_DR_MP3_HDR_TEST_NOT_MPEG25(h) ((h[1]) & 0x10) +#define MA_DR_MP3_HDR_TEST_I_STEREO(h) ((h[3]) & 0x10) +#define MA_DR_MP3_HDR_TEST_MS_STEREO(h) ((h[3]) & 0x20) +#define MA_DR_MP3_HDR_GET_STEREO_MODE(h) (((h[3]) >> 6) & 3) +#define MA_DR_MP3_HDR_GET_STEREO_MODE_EXT(h) (((h[3]) >> 4) & 3) +#define MA_DR_MP3_HDR_GET_LAYER(h) (((h[1]) >> 1) & 3) +#define MA_DR_MP3_HDR_GET_BITRATE(h) ((h[2]) >> 4) +#define MA_DR_MP3_HDR_GET_SAMPLE_RATE(h) (((h[2]) >> 2) & 3) +#define MA_DR_MP3_HDR_GET_MY_SAMPLE_RATE(h) (MA_DR_MP3_HDR_GET_SAMPLE_RATE(h) + (((h[1] >> 3) & 1) + ((h[1] >> 4) & 1))*3) +#define MA_DR_MP3_HDR_IS_FRAME_576(h) ((h[1] & 14) == 2) +#define MA_DR_MP3_HDR_IS_LAYER_1(h) ((h[1] & 6) == 6) +#define MA_DR_MP3_BITS_DEQUANTIZER_OUT -1 +#define MA_DR_MP3_MAX_SCF (255 + MA_DR_MP3_BITS_DEQUANTIZER_OUT*4 - 210) +#define MA_DR_MP3_MAX_SCFI ((MA_DR_MP3_MAX_SCF + 3) & ~3) +#define MA_DR_MP3_MIN(a, b) ((a) > (b) ? (b) : (a)) +#define MA_DR_MP3_MAX(a, b) ((a) < (b) ? (b) : (a)) +#if !defined(MA_DR_MP3_NO_SIMD) +#if !defined(MA_DR_MP3_ONLY_SIMD) && (defined(_M_X64) || defined(__x86_64__) || defined(__aarch64__) || defined(_M_ARM64) || defined(_M_ARM64EC)) +#define MA_DR_MP3_ONLY_SIMD +#endif +#if ((defined(_MSC_VER) && _MSC_VER >= 1400) && defined(_M_X64)) || ((defined(__i386) || defined(_M_IX86) || defined(__i386__) || defined(__x86_64__)) && ((defined(_M_IX86_FP) && _M_IX86_FP == 2) || defined(__SSE2__))) +#if defined(_MSC_VER) +#include +#endif +#include +#define MA_DR_MP3_HAVE_SSE 1 +#define MA_DR_MP3_HAVE_SIMD 1 +#define MA_DR_MP3_VSTORE _mm_storeu_ps +#define MA_DR_MP3_VLD _mm_loadu_ps +#define MA_DR_MP3_VSET _mm_set1_ps +#define MA_DR_MP3_VADD _mm_add_ps +#define MA_DR_MP3_VSUB _mm_sub_ps +#define MA_DR_MP3_VMUL _mm_mul_ps +#define MA_DR_MP3_VMAC(a, x, y) _mm_add_ps(a, _mm_mul_ps(x, y)) +#define MA_DR_MP3_VMSB(a, x, y) _mm_sub_ps(a, _mm_mul_ps(x, y)) +#define MA_DR_MP3_VMUL_S(x, s) _mm_mul_ps(x, _mm_set1_ps(s)) +#define MA_DR_MP3_VREV(x) _mm_shuffle_ps(x, x, _MM_SHUFFLE(0, 1, 2, 3)) +typedef __m128 ma_dr_mp3_f4; +#if defined(_MSC_VER) || defined(MA_DR_MP3_ONLY_SIMD) +#define ma_dr_mp3_cpuid __cpuid +#else +static __inline__ __attribute__((always_inline)) void ma_dr_mp3_cpuid(int CPUInfo[], const int InfoType) +{ +#if defined(__PIC__) + __asm__ __volatile__( +#if defined(__x86_64__) + "push %%rbx\n" + "cpuid\n" + "xchgl %%ebx, %1\n" + "pop %%rbx\n" +#else + "xchgl %%ebx, %1\n" + "cpuid\n" + "xchgl %%ebx, %1\n" +#endif + : "=a" (CPUInfo[0]), "=r" (CPUInfo[1]), "=c" (CPUInfo[2]), "=d" (CPUInfo[3]) + : "a" (InfoType)); +#else + __asm__ __volatile__( + "cpuid" + : "=a" (CPUInfo[0]), "=b" (CPUInfo[1]), "=c" (CPUInfo[2]), "=d" (CPUInfo[3]) + : "a" (InfoType)); +#endif +} +#endif +static int ma_dr_mp3_have_simd(void) +{ +#ifdef MA_DR_MP3_ONLY_SIMD + return 1; +#else + static int g_have_simd; + int CPUInfo[4]; +#ifdef MINIMP3_TEST + static int g_counter; + if (g_counter++ > 100) + return 0; +#endif + if (g_have_simd) + goto end; + ma_dr_mp3_cpuid(CPUInfo, 0); + if (CPUInfo[0] > 0) + { + ma_dr_mp3_cpuid(CPUInfo, 1); + g_have_simd = (CPUInfo[3] & (1 << 26)) + 1; + return g_have_simd - 1; + } +end: + return g_have_simd - 1; +#endif +} +#elif defined(__ARM_NEON) || defined(__aarch64__) || defined(_M_ARM64) || defined(_M_ARM64EC) +#include +#define MA_DR_MP3_HAVE_SSE 0 +#define MA_DR_MP3_HAVE_SIMD 1 +#define MA_DR_MP3_VSTORE vst1q_f32 +#define MA_DR_MP3_VLD vld1q_f32 +#define MA_DR_MP3_VSET vmovq_n_f32 +#define MA_DR_MP3_VADD vaddq_f32 +#define MA_DR_MP3_VSUB vsubq_f32 +#define MA_DR_MP3_VMUL vmulq_f32 +#define MA_DR_MP3_VMAC(a, x, y) vmlaq_f32(a, x, y) +#define MA_DR_MP3_VMSB(a, x, y) vmlsq_f32(a, x, y) +#define MA_DR_MP3_VMUL_S(x, s) vmulq_f32(x, vmovq_n_f32(s)) +#define MA_DR_MP3_VREV(x) vcombine_f32(vget_high_f32(vrev64q_f32(x)), vget_low_f32(vrev64q_f32(x))) +typedef float32x4_t ma_dr_mp3_f4; +static int ma_dr_mp3_have_simd(void) +{ + return 1; +} +#else +#define MA_DR_MP3_HAVE_SSE 0 +#define MA_DR_MP3_HAVE_SIMD 0 +#ifdef MA_DR_MP3_ONLY_SIMD +#error MA_DR_MP3_ONLY_SIMD used, but SSE/NEON not enabled +#endif +#endif +#else +#define MA_DR_MP3_HAVE_SIMD 0 +#endif +#if defined(__ARM_ARCH) && (__ARM_ARCH >= 6) && !defined(__aarch64__) && !defined(_M_ARM64) && !defined(_M_ARM64EC) && !defined(__ARM_ARCH_6M__) +#define MA_DR_MP3_HAVE_ARMV6 1 +static __inline__ __attribute__((always_inline)) ma_int32 ma_dr_mp3_clip_int16_arm(ma_int32 a) +{ + ma_int32 x = 0; + __asm__ ("ssat %0, #16, %1" : "=r"(x) : "r"(a)); + return x; +} +#else +#define MA_DR_MP3_HAVE_ARMV6 0 +#endif +#ifndef MA_DR_MP3_ASSERT +#include +#define MA_DR_MP3_ASSERT(expression) assert(expression) +#endif +#ifndef MA_DR_MP3_COPY_MEMORY +#define MA_DR_MP3_COPY_MEMORY(dst, src, sz) memcpy((dst), (src), (sz)) +#endif +#ifndef MA_DR_MP3_MOVE_MEMORY +#define MA_DR_MP3_MOVE_MEMORY(dst, src, sz) memmove((dst), (src), (sz)) +#endif +#ifndef MA_DR_MP3_ZERO_MEMORY +#define MA_DR_MP3_ZERO_MEMORY(p, sz) memset((p), 0, (sz)) +#endif +#define MA_DR_MP3_ZERO_OBJECT(p) MA_DR_MP3_ZERO_MEMORY((p), sizeof(*(p))) +#ifndef MA_DR_MP3_MALLOC +#define MA_DR_MP3_MALLOC(sz) malloc((sz)) +#endif +#ifndef MA_DR_MP3_REALLOC +#define MA_DR_MP3_REALLOC(p, sz) realloc((p), (sz)) +#endif +#ifndef MA_DR_MP3_FREE +#define MA_DR_MP3_FREE(p) free((p)) +#endif +typedef struct +{ + const ma_uint8 *buf; + int pos, limit; +} ma_dr_mp3_bs; +typedef struct +{ + float scf[3*64]; + ma_uint8 total_bands, stereo_bands, bitalloc[64], scfcod[64]; +} ma_dr_mp3_L12_scale_info; +typedef struct +{ + ma_uint8 tab_offset, code_tab_width, band_count; +} ma_dr_mp3_L12_subband_alloc; +typedef struct +{ + const ma_uint8 *sfbtab; + ma_uint16 part_23_length, big_values, scalefac_compress; + ma_uint8 global_gain, block_type, mixed_block_flag, n_long_sfb, n_short_sfb; + ma_uint8 table_select[3], region_count[3], subblock_gain[3]; + ma_uint8 preflag, scalefac_scale, count1_table, scfsi; +} ma_dr_mp3_L3_gr_info; +typedef struct +{ + ma_dr_mp3_bs bs; + ma_uint8 maindata[MA_DR_MP3_MAX_BITRESERVOIR_BYTES + MA_DR_MP3_MAX_L3_FRAME_PAYLOAD_BYTES]; + ma_dr_mp3_L3_gr_info gr_info[4]; + float grbuf[2][576], scf[40], syn[18 + 15][2*32]; + ma_uint8 ist_pos[2][39]; +} ma_dr_mp3dec_scratch; +static void ma_dr_mp3_bs_init(ma_dr_mp3_bs *bs, const ma_uint8 *data, int bytes) +{ + bs->buf = data; + bs->pos = 0; + bs->limit = bytes*8; +} +static ma_uint32 ma_dr_mp3_bs_get_bits(ma_dr_mp3_bs *bs, int n) +{ + ma_uint32 next, cache = 0, s = bs->pos & 7; + int shl = n + s; + const ma_uint8 *p = bs->buf + (bs->pos >> 3); + if ((bs->pos += n) > bs->limit) + return 0; + next = *p++ & (255 >> s); + while ((shl -= 8) > 0) + { + cache |= next << shl; + next = *p++; + } + return cache | (next >> -shl); +} +static int ma_dr_mp3_hdr_valid(const ma_uint8 *h) +{ + return h[0] == 0xff && + ((h[1] & 0xF0) == 0xf0 || (h[1] & 0xFE) == 0xe2) && + (MA_DR_MP3_HDR_GET_LAYER(h) != 0) && + (MA_DR_MP3_HDR_GET_BITRATE(h) != 15) && + (MA_DR_MP3_HDR_GET_SAMPLE_RATE(h) != 3); +} +static int ma_dr_mp3_hdr_compare(const ma_uint8 *h1, const ma_uint8 *h2) +{ + return ma_dr_mp3_hdr_valid(h2) && + ((h1[1] ^ h2[1]) & 0xFE) == 0 && + ((h1[2] ^ h2[2]) & 0x0C) == 0 && + !(MA_DR_MP3_HDR_IS_FREE_FORMAT(h1) ^ MA_DR_MP3_HDR_IS_FREE_FORMAT(h2)); +} +static unsigned ma_dr_mp3_hdr_bitrate_kbps(const ma_uint8 *h) +{ + static const ma_uint8 halfrate[2][3][15] = { + { { 0,4,8,12,16,20,24,28,32,40,48,56,64,72,80 }, { 0,4,8,12,16,20,24,28,32,40,48,56,64,72,80 }, { 0,16,24,28,32,40,48,56,64,72,80,88,96,112,128 } }, + { { 0,16,20,24,28,32,40,48,56,64,80,96,112,128,160 }, { 0,16,24,28,32,40,48,56,64,80,96,112,128,160,192 }, { 0,16,32,48,64,80,96,112,128,144,160,176,192,208,224 } }, + }; + return 2*halfrate[!!MA_DR_MP3_HDR_TEST_MPEG1(h)][MA_DR_MP3_HDR_GET_LAYER(h) - 1][MA_DR_MP3_HDR_GET_BITRATE(h)]; +} +static unsigned ma_dr_mp3_hdr_sample_rate_hz(const ma_uint8 *h) +{ + static const unsigned g_hz[3] = { 44100, 48000, 32000 }; + return g_hz[MA_DR_MP3_HDR_GET_SAMPLE_RATE(h)] >> (int)!MA_DR_MP3_HDR_TEST_MPEG1(h) >> (int)!MA_DR_MP3_HDR_TEST_NOT_MPEG25(h); +} +static unsigned ma_dr_mp3_hdr_frame_samples(const ma_uint8 *h) +{ + return MA_DR_MP3_HDR_IS_LAYER_1(h) ? 384 : (1152 >> (int)MA_DR_MP3_HDR_IS_FRAME_576(h)); +} +static int ma_dr_mp3_hdr_frame_bytes(const ma_uint8 *h, int free_format_size) +{ + int frame_bytes = ma_dr_mp3_hdr_frame_samples(h)*ma_dr_mp3_hdr_bitrate_kbps(h)*125/ma_dr_mp3_hdr_sample_rate_hz(h); + if (MA_DR_MP3_HDR_IS_LAYER_1(h)) + { + frame_bytes &= ~3; + } + return frame_bytes ? frame_bytes : free_format_size; +} +static int ma_dr_mp3_hdr_padding(const ma_uint8 *h) +{ + return MA_DR_MP3_HDR_TEST_PADDING(h) ? (MA_DR_MP3_HDR_IS_LAYER_1(h) ? 4 : 1) : 0; +} +#ifndef MA_DR_MP3_ONLY_MP3 +static const ma_dr_mp3_L12_subband_alloc *ma_dr_mp3_L12_subband_alloc_table(const ma_uint8 *hdr, ma_dr_mp3_L12_scale_info *sci) +{ + const ma_dr_mp3_L12_subband_alloc *alloc; + int mode = MA_DR_MP3_HDR_GET_STEREO_MODE(hdr); + int nbands, stereo_bands = (mode == MA_DR_MP3_MODE_MONO) ? 0 : (mode == MA_DR_MP3_MODE_JOINT_STEREO) ? (MA_DR_MP3_HDR_GET_STEREO_MODE_EXT(hdr) << 2) + 4 : 32; + if (MA_DR_MP3_HDR_IS_LAYER_1(hdr)) + { + static const ma_dr_mp3_L12_subband_alloc g_alloc_L1[] = { { 76, 4, 32 } }; + alloc = g_alloc_L1; + nbands = 32; + } else if (!MA_DR_MP3_HDR_TEST_MPEG1(hdr)) + { + static const ma_dr_mp3_L12_subband_alloc g_alloc_L2M2[] = { { 60, 4, 4 }, { 44, 3, 7 }, { 44, 2, 19 } }; + alloc = g_alloc_L2M2; + nbands = 30; + } else + { + static const ma_dr_mp3_L12_subband_alloc g_alloc_L2M1[] = { { 0, 4, 3 }, { 16, 4, 8 }, { 32, 3, 12 }, { 40, 2, 7 } }; + int sample_rate_idx = MA_DR_MP3_HDR_GET_SAMPLE_RATE(hdr); + unsigned kbps = ma_dr_mp3_hdr_bitrate_kbps(hdr) >> (int)(mode != MA_DR_MP3_MODE_MONO); + if (!kbps) + { + kbps = 192; + } + alloc = g_alloc_L2M1; + nbands = 27; + if (kbps < 56) + { + static const ma_dr_mp3_L12_subband_alloc g_alloc_L2M1_lowrate[] = { { 44, 4, 2 }, { 44, 3, 10 } }; + alloc = g_alloc_L2M1_lowrate; + nbands = sample_rate_idx == 2 ? 12 : 8; + } else if (kbps >= 96 && sample_rate_idx != 1) + { + nbands = 30; + } + } + sci->total_bands = (ma_uint8)nbands; + sci->stereo_bands = (ma_uint8)MA_DR_MP3_MIN(stereo_bands, nbands); + return alloc; +} +static void ma_dr_mp3_L12_read_scalefactors(ma_dr_mp3_bs *bs, ma_uint8 *pba, ma_uint8 *scfcod, int bands, float *scf) +{ + static const float g_deq_L12[18*3] = { +#define MA_DR_MP3_DQ(x) 9.53674316e-07f/x, 7.56931807e-07f/x, 6.00777173e-07f/x + MA_DR_MP3_DQ(3),MA_DR_MP3_DQ(7),MA_DR_MP3_DQ(15),MA_DR_MP3_DQ(31),MA_DR_MP3_DQ(63),MA_DR_MP3_DQ(127),MA_DR_MP3_DQ(255),MA_DR_MP3_DQ(511),MA_DR_MP3_DQ(1023),MA_DR_MP3_DQ(2047),MA_DR_MP3_DQ(4095),MA_DR_MP3_DQ(8191),MA_DR_MP3_DQ(16383),MA_DR_MP3_DQ(32767),MA_DR_MP3_DQ(65535),MA_DR_MP3_DQ(3),MA_DR_MP3_DQ(5),MA_DR_MP3_DQ(9) + }; + int i, m; + for (i = 0; i < bands; i++) + { + float s = 0; + int ba = *pba++; + int mask = ba ? 4 + ((19 >> scfcod[i]) & 3) : 0; + for (m = 4; m; m >>= 1) + { + if (mask & m) + { + int b = ma_dr_mp3_bs_get_bits(bs, 6); + s = g_deq_L12[ba*3 - 6 + b % 3]*(int)(1 << 21 >> b/3); + } + *scf++ = s; + } + } +} +static void ma_dr_mp3_L12_read_scale_info(const ma_uint8 *hdr, ma_dr_mp3_bs *bs, ma_dr_mp3_L12_scale_info *sci) +{ + static const ma_uint8 g_bitalloc_code_tab[] = { + 0,17, 3, 4, 5,6,7, 8,9,10,11,12,13,14,15,16, + 0,17,18, 3,19,4,5, 6,7, 8, 9,10,11,12,13,16, + 0,17,18, 3,19,4,5,16, + 0,17,18,16, + 0,17,18,19, 4,5,6, 7,8, 9,10,11,12,13,14,15, + 0,17,18, 3,19,4,5, 6,7, 8, 9,10,11,12,13,14, + 0, 2, 3, 4, 5,6,7, 8,9,10,11,12,13,14,15,16 + }; + const ma_dr_mp3_L12_subband_alloc *subband_alloc = ma_dr_mp3_L12_subband_alloc_table(hdr, sci); + int i, k = 0, ba_bits = 0; + const ma_uint8 *ba_code_tab = g_bitalloc_code_tab; + for (i = 0; i < sci->total_bands; i++) + { + ma_uint8 ba; + if (i == k) + { + k += subband_alloc->band_count; + ba_bits = subband_alloc->code_tab_width; + ba_code_tab = g_bitalloc_code_tab + subband_alloc->tab_offset; + subband_alloc++; + } + ba = ba_code_tab[ma_dr_mp3_bs_get_bits(bs, ba_bits)]; + sci->bitalloc[2*i] = ba; + if (i < sci->stereo_bands) + { + ba = ba_code_tab[ma_dr_mp3_bs_get_bits(bs, ba_bits)]; + } + sci->bitalloc[2*i + 1] = sci->stereo_bands ? ba : 0; + } + for (i = 0; i < 2*sci->total_bands; i++) + { + sci->scfcod[i] = (ma_uint8)(sci->bitalloc[i] ? MA_DR_MP3_HDR_IS_LAYER_1(hdr) ? 2 : ma_dr_mp3_bs_get_bits(bs, 2) : 6); + } + ma_dr_mp3_L12_read_scalefactors(bs, sci->bitalloc, sci->scfcod, sci->total_bands*2, sci->scf); + for (i = sci->stereo_bands; i < sci->total_bands; i++) + { + sci->bitalloc[2*i + 1] = 0; + } +} +static int ma_dr_mp3_L12_dequantize_granule(float *grbuf, ma_dr_mp3_bs *bs, ma_dr_mp3_L12_scale_info *sci, int group_size) +{ + int i, j, k, choff = 576; + for (j = 0; j < 4; j++) + { + float *dst = grbuf + group_size*j; + for (i = 0; i < 2*sci->total_bands; i++) + { + int ba = sci->bitalloc[i]; + if (ba != 0) + { + if (ba < 17) + { + int half = (1 << (ba - 1)) - 1; + for (k = 0; k < group_size; k++) + { + dst[k] = (float)((int)ma_dr_mp3_bs_get_bits(bs, ba) - half); + } + } else + { + unsigned mod = (2 << (ba - 17)) + 1; + unsigned code = ma_dr_mp3_bs_get_bits(bs, mod + 2 - (mod >> 3)); + for (k = 0; k < group_size; k++, code /= mod) + { + dst[k] = (float)((int)(code % mod - mod/2)); + } + } + } + dst += choff; + choff = 18 - choff; + } + } + return group_size*4; +} +static void ma_dr_mp3_L12_apply_scf_384(ma_dr_mp3_L12_scale_info *sci, const float *scf, float *dst) +{ + int i, k; + MA_DR_MP3_COPY_MEMORY(dst + 576 + sci->stereo_bands*18, dst + sci->stereo_bands*18, (sci->total_bands - sci->stereo_bands)*18*sizeof(float)); + for (i = 0; i < sci->total_bands; i++, dst += 18, scf += 6) + { + for (k = 0; k < 12; k++) + { + dst[k + 0] *= scf[0]; + dst[k + 576] *= scf[3]; + } + } +} +#endif +static int ma_dr_mp3_L3_read_side_info(ma_dr_mp3_bs *bs, ma_dr_mp3_L3_gr_info *gr, const ma_uint8 *hdr) +{ + static const ma_uint8 g_scf_long[8][23] = { + { 6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54,0 }, + { 12,12,12,12,12,12,16,20,24,28,32,40,48,56,64,76,90,2,2,2,2,2,0 }, + { 6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54,0 }, + { 6,6,6,6,6,6,8,10,12,14,16,18,22,26,32,38,46,54,62,70,76,36,0 }, + { 6,6,6,6,6,6,8,10,12,14,16,20,24,28,32,38,46,52,60,68,58,54,0 }, + { 4,4,4,4,4,4,6,6,8,8,10,12,16,20,24,28,34,42,50,54,76,158,0 }, + { 4,4,4,4,4,4,6,6,6,8,10,12,16,18,22,28,34,40,46,54,54,192,0 }, + { 4,4,4,4,4,4,6,6,8,10,12,16,20,24,30,38,46,56,68,84,102,26,0 } + }; + static const ma_uint8 g_scf_short[8][40] = { + { 4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,30,30,30,40,40,40,18,18,18,0 }, + { 8,8,8,8,8,8,8,8,8,12,12,12,16,16,16,20,20,20,24,24,24,28,28,28,36,36,36,2,2,2,2,2,2,2,2,2,26,26,26,0 }, + { 4,4,4,4,4,4,4,4,4,6,6,6,6,6,6,8,8,8,10,10,10,14,14,14,18,18,18,26,26,26,32,32,32,42,42,42,18,18,18,0 }, + { 4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,32,32,32,44,44,44,12,12,12,0 }, + { 4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,30,30,30,40,40,40,18,18,18,0 }, + { 4,4,4,4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,22,22,22,30,30,30,56,56,56,0 }, + { 4,4,4,4,4,4,4,4,4,4,4,4,6,6,6,6,6,6,10,10,10,12,12,12,14,14,14,16,16,16,20,20,20,26,26,26,66,66,66,0 }, + { 4,4,4,4,4,4,4,4,4,4,4,4,6,6,6,8,8,8,12,12,12,16,16,16,20,20,20,26,26,26,34,34,34,42,42,42,12,12,12,0 } + }; + static const ma_uint8 g_scf_mixed[8][40] = { + { 6,6,6,6,6,6,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,30,30,30,40,40,40,18,18,18,0 }, + { 12,12,12,4,4,4,8,8,8,12,12,12,16,16,16,20,20,20,24,24,24,28,28,28,36,36,36,2,2,2,2,2,2,2,2,2,26,26,26,0 }, + { 6,6,6,6,6,6,6,6,6,6,6,6,8,8,8,10,10,10,14,14,14,18,18,18,26,26,26,32,32,32,42,42,42,18,18,18,0 }, + { 6,6,6,6,6,6,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,32,32,32,44,44,44,12,12,12,0 }, + { 6,6,6,6,6,6,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,24,24,24,30,30,30,40,40,40,18,18,18,0 }, + { 4,4,4,4,4,4,6,6,4,4,4,6,6,6,8,8,8,10,10,10,12,12,12,14,14,14,18,18,18,22,22,22,30,30,30,56,56,56,0 }, + { 4,4,4,4,4,4,6,6,4,4,4,6,6,6,6,6,6,10,10,10,12,12,12,14,14,14,16,16,16,20,20,20,26,26,26,66,66,66,0 }, + { 4,4,4,4,4,4,6,6,4,4,4,6,6,6,8,8,8,12,12,12,16,16,16,20,20,20,26,26,26,34,34,34,42,42,42,12,12,12,0 } + }; + unsigned tables, scfsi = 0; + int main_data_begin, part_23_sum = 0; + int gr_count = MA_DR_MP3_HDR_IS_MONO(hdr) ? 1 : 2; + int sr_idx = MA_DR_MP3_HDR_GET_MY_SAMPLE_RATE(hdr); sr_idx -= (sr_idx != 0); + if (MA_DR_MP3_HDR_TEST_MPEG1(hdr)) + { + gr_count *= 2; + main_data_begin = ma_dr_mp3_bs_get_bits(bs, 9); + scfsi = ma_dr_mp3_bs_get_bits(bs, 7 + gr_count); + } else + { + main_data_begin = ma_dr_mp3_bs_get_bits(bs, 8 + gr_count) >> gr_count; + } + do + { + if (MA_DR_MP3_HDR_IS_MONO(hdr)) + { + scfsi <<= 4; + } + gr->part_23_length = (ma_uint16)ma_dr_mp3_bs_get_bits(bs, 12); + part_23_sum += gr->part_23_length; + gr->big_values = (ma_uint16)ma_dr_mp3_bs_get_bits(bs, 9); + if (gr->big_values > 288) + { + return -1; + } + gr->global_gain = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 8); + gr->scalefac_compress = (ma_uint16)ma_dr_mp3_bs_get_bits(bs, MA_DR_MP3_HDR_TEST_MPEG1(hdr) ? 4 : 9); + gr->sfbtab = g_scf_long[sr_idx]; + gr->n_long_sfb = 22; + gr->n_short_sfb = 0; + if (ma_dr_mp3_bs_get_bits(bs, 1)) + { + gr->block_type = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 2); + if (!gr->block_type) + { + return -1; + } + gr->mixed_block_flag = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 1); + gr->region_count[0] = 7; + gr->region_count[1] = 255; + if (gr->block_type == MA_DR_MP3_SHORT_BLOCK_TYPE) + { + scfsi &= 0x0F0F; + if (!gr->mixed_block_flag) + { + gr->region_count[0] = 8; + gr->sfbtab = g_scf_short[sr_idx]; + gr->n_long_sfb = 0; + gr->n_short_sfb = 39; + } else + { + gr->sfbtab = g_scf_mixed[sr_idx]; + gr->n_long_sfb = MA_DR_MP3_HDR_TEST_MPEG1(hdr) ? 8 : 6; + gr->n_short_sfb = 30; + } + } + tables = ma_dr_mp3_bs_get_bits(bs, 10); + tables <<= 5; + gr->subblock_gain[0] = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 3); + gr->subblock_gain[1] = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 3); + gr->subblock_gain[2] = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 3); + } else + { + gr->block_type = 0; + gr->mixed_block_flag = 0; + tables = ma_dr_mp3_bs_get_bits(bs, 15); + gr->region_count[0] = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 4); + gr->region_count[1] = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 3); + gr->region_count[2] = 255; + } + gr->table_select[0] = (ma_uint8)(tables >> 10); + gr->table_select[1] = (ma_uint8)((tables >> 5) & 31); + gr->table_select[2] = (ma_uint8)((tables) & 31); + gr->preflag = (ma_uint8)(MA_DR_MP3_HDR_TEST_MPEG1(hdr) ? ma_dr_mp3_bs_get_bits(bs, 1) : (gr->scalefac_compress >= 500)); + gr->scalefac_scale = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 1); + gr->count1_table = (ma_uint8)ma_dr_mp3_bs_get_bits(bs, 1); + gr->scfsi = (ma_uint8)((scfsi >> 12) & 15); + scfsi <<= 4; + gr++; + } while(--gr_count); + if (part_23_sum + bs->pos > bs->limit + main_data_begin*8) + { + return -1; + } + return main_data_begin; +} +static void ma_dr_mp3_L3_read_scalefactors(ma_uint8 *scf, ma_uint8 *ist_pos, const ma_uint8 *scf_size, const ma_uint8 *scf_count, ma_dr_mp3_bs *bitbuf, int scfsi) +{ + int i, k; + for (i = 0; i < 4 && scf_count[i]; i++, scfsi *= 2) + { + int cnt = scf_count[i]; + if (scfsi & 8) + { + MA_DR_MP3_COPY_MEMORY(scf, ist_pos, cnt); + } else + { + int bits = scf_size[i]; + if (!bits) + { + MA_DR_MP3_ZERO_MEMORY(scf, cnt); + MA_DR_MP3_ZERO_MEMORY(ist_pos, cnt); + } else + { + int max_scf = (scfsi < 0) ? (1 << bits) - 1 : -1; + for (k = 0; k < cnt; k++) + { + int s = ma_dr_mp3_bs_get_bits(bitbuf, bits); + ist_pos[k] = (ma_uint8)(s == max_scf ? -1 : s); + scf[k] = (ma_uint8)s; + } + } + } + ist_pos += cnt; + scf += cnt; + } + scf[0] = scf[1] = scf[2] = 0; +} +static float ma_dr_mp3_L3_ldexp_q2(float y, int exp_q2) +{ + static const float g_expfrac[4] = { 9.31322575e-10f,7.83145814e-10f,6.58544508e-10f,5.53767716e-10f }; + int e; + do + { + e = MA_DR_MP3_MIN(30*4, exp_q2); + y *= g_expfrac[e & 3]*(1 << 30 >> (e >> 2)); + } while ((exp_q2 -= e) > 0); + return y; +} +static void ma_dr_mp3_L3_decode_scalefactors(const ma_uint8 *hdr, ma_uint8 *ist_pos, ma_dr_mp3_bs *bs, const ma_dr_mp3_L3_gr_info *gr, float *scf, int ch) +{ + static const ma_uint8 g_scf_partitions[3][28] = { + { 6,5,5, 5,6,5,5,5,6,5, 7,3,11,10,0,0, 7, 7, 7,0, 6, 6,6,3, 8, 8,5,0 }, + { 8,9,6,12,6,9,9,9,6,9,12,6,15,18,0,0, 6,15,12,0, 6,12,9,6, 6,18,9,0 }, + { 9,9,6,12,9,9,9,9,9,9,12,6,18,18,0,0,12,12,12,0,12, 9,9,6,15,12,9,0 } + }; + const ma_uint8 *scf_partition = g_scf_partitions[!!gr->n_short_sfb + !gr->n_long_sfb]; + ma_uint8 scf_size[4], iscf[40]; + int i, scf_shift = gr->scalefac_scale + 1, gain_exp, scfsi = gr->scfsi; + float gain; + if (MA_DR_MP3_HDR_TEST_MPEG1(hdr)) + { + static const ma_uint8 g_scfc_decode[16] = { 0,1,2,3, 12,5,6,7, 9,10,11,13, 14,15,18,19 }; + int part = g_scfc_decode[gr->scalefac_compress]; + scf_size[1] = scf_size[0] = (ma_uint8)(part >> 2); + scf_size[3] = scf_size[2] = (ma_uint8)(part & 3); + } else + { + static const ma_uint8 g_mod[6*4] = { 5,5,4,4,5,5,4,1,4,3,1,1,5,6,6,1,4,4,4,1,4,3,1,1 }; + int k, modprod, sfc, ist = MA_DR_MP3_HDR_TEST_I_STEREO(hdr) && ch; + sfc = gr->scalefac_compress >> ist; + for (k = ist*3*4; sfc >= 0; sfc -= modprod, k += 4) + { + for (modprod = 1, i = 3; i >= 0; i--) + { + scf_size[i] = (ma_uint8)(sfc / modprod % g_mod[k + i]); + modprod *= g_mod[k + i]; + } + } + scf_partition += k; + scfsi = -16; + } + ma_dr_mp3_L3_read_scalefactors(iscf, ist_pos, scf_size, scf_partition, bs, scfsi); + if (gr->n_short_sfb) + { + int sh = 3 - scf_shift; + for (i = 0; i < gr->n_short_sfb; i += 3) + { + iscf[gr->n_long_sfb + i + 0] = (ma_uint8)(iscf[gr->n_long_sfb + i + 0] + (gr->subblock_gain[0] << sh)); + iscf[gr->n_long_sfb + i + 1] = (ma_uint8)(iscf[gr->n_long_sfb + i + 1] + (gr->subblock_gain[1] << sh)); + iscf[gr->n_long_sfb + i + 2] = (ma_uint8)(iscf[gr->n_long_sfb + i + 2] + (gr->subblock_gain[2] << sh)); + } + } else if (gr->preflag) + { + static const ma_uint8 g_preamp[10] = { 1,1,1,1,2,2,3,3,3,2 }; + for (i = 0; i < 10; i++) + { + iscf[11 + i] = (ma_uint8)(iscf[11 + i] + g_preamp[i]); + } + } + gain_exp = gr->global_gain + MA_DR_MP3_BITS_DEQUANTIZER_OUT*4 - 210 - (MA_DR_MP3_HDR_IS_MS_STEREO(hdr) ? 2 : 0); + gain = ma_dr_mp3_L3_ldexp_q2(1 << (MA_DR_MP3_MAX_SCFI/4), MA_DR_MP3_MAX_SCFI - gain_exp); + for (i = 0; i < (int)(gr->n_long_sfb + gr->n_short_sfb); i++) + { + scf[i] = ma_dr_mp3_L3_ldexp_q2(gain, iscf[i] << scf_shift); + } +} +static const float g_ma_dr_mp3_pow43[129 + 16] = { + 0,-1,-2.519842f,-4.326749f,-6.349604f,-8.549880f,-10.902724f,-13.390518f,-16.000000f,-18.720754f,-21.544347f,-24.463781f,-27.473142f,-30.567351f,-33.741992f,-36.993181f, + 0,1,2.519842f,4.326749f,6.349604f,8.549880f,10.902724f,13.390518f,16.000000f,18.720754f,21.544347f,24.463781f,27.473142f,30.567351f,33.741992f,36.993181f,40.317474f,43.711787f,47.173345f,50.699631f,54.288352f,57.937408f,61.644865f,65.408941f,69.227979f,73.100443f,77.024898f,81.000000f,85.024491f,89.097188f,93.216975f,97.382800f,101.593667f,105.848633f,110.146801f,114.487321f,118.869381f,123.292209f,127.755065f,132.257246f,136.798076f,141.376907f,145.993119f,150.646117f,155.335327f,160.060199f,164.820202f,169.614826f,174.443577f,179.305980f,184.201575f,189.129918f,194.090580f,199.083145f,204.107210f,209.162385f,214.248292f,219.364564f,224.510845f,229.686789f,234.892058f,240.126328f,245.389280f,250.680604f,256.000000f,261.347174f,266.721841f,272.123723f,277.552547f,283.008049f,288.489971f,293.998060f,299.532071f,305.091761f,310.676898f,316.287249f,321.922592f,327.582707f,333.267377f,338.976394f,344.709550f,350.466646f,356.247482f,362.051866f,367.879608f,373.730522f,379.604427f,385.501143f,391.420496f,397.362314f,403.326427f,409.312672f,415.320884f,421.350905f,427.402579f,433.475750f,439.570269f,445.685987f,451.822757f,457.980436f,464.158883f,470.357960f,476.577530f,482.817459f,489.077615f,495.357868f,501.658090f,507.978156f,514.317941f,520.677324f,527.056184f,533.454404f,539.871867f,546.308458f,552.764065f,559.238575f,565.731879f,572.243870f,578.774440f,585.323483f,591.890898f,598.476581f,605.080431f,611.702349f,618.342238f,625.000000f,631.675540f,638.368763f,645.079578f +}; +static float ma_dr_mp3_L3_pow_43(int x) +{ + float frac; + int sign, mult = 256; + if (x < 129) + { + return g_ma_dr_mp3_pow43[16 + x]; + } + if (x < 1024) + { + mult = 16; + x <<= 3; + } + sign = 2*x & 64; + frac = (float)((x & 63) - sign) / ((x & ~63) + sign); + return g_ma_dr_mp3_pow43[16 + ((x + sign) >> 6)]*(1.f + frac*((4.f/3) + frac*(2.f/9)))*mult; +} +static void ma_dr_mp3_L3_huffman(float *dst, ma_dr_mp3_bs *bs, const ma_dr_mp3_L3_gr_info *gr_info, const float *scf, int layer3gr_limit) +{ + static const ma_int16 tabs[] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 785,785,785,785,784,784,784,784,513,513,513,513,513,513,513,513,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256, + -255,1313,1298,1282,785,785,785,785,784,784,784,784,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,290,288, + -255,1313,1298,1282,769,769,769,769,529,529,529,529,529,529,529,529,528,528,528,528,528,528,528,528,512,512,512,512,512,512,512,512,290,288, + -253,-318,-351,-367,785,785,785,785,784,784,784,784,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,819,818,547,547,275,275,275,275,561,560,515,546,289,274,288,258, + -254,-287,1329,1299,1314,1312,1057,1057,1042,1042,1026,1026,784,784,784,784,529,529,529,529,529,529,529,529,769,769,769,769,768,768,768,768,563,560,306,306,291,259, + -252,-413,-477,-542,1298,-575,1041,1041,784,784,784,784,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,-383,-399,1107,1092,1106,1061,849,849,789,789,1104,1091,773,773,1076,1075,341,340,325,309,834,804,577,577,532,532,516,516,832,818,803,816,561,561,531,531,515,546,289,289,288,258, + -252,-429,-493,-559,1057,1057,1042,1042,529,529,529,529,529,529,529,529,784,784,784,784,769,769,769,769,512,512,512,512,512,512,512,512,-382,1077,-415,1106,1061,1104,849,849,789,789,1091,1076,1029,1075,834,834,597,581,340,340,339,324,804,833,532,532,832,772,818,803,817,787,816,771,290,290,290,290,288,258, + -253,-349,-414,-447,-463,1329,1299,-479,1314,1312,1057,1057,1042,1042,1026,1026,785,785,785,785,784,784,784,784,769,769,769,769,768,768,768,768,-319,851,821,-335,836,850,805,849,341,340,325,336,533,533,579,579,564,564,773,832,578,548,563,516,321,276,306,291,304,259, + -251,-572,-733,-830,-863,-879,1041,1041,784,784,784,784,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,-511,-527,-543,1396,1351,1381,1366,1395,1335,1380,-559,1334,1138,1138,1063,1063,1350,1392,1031,1031,1062,1062,1364,1363,1120,1120,1333,1348,881,881,881,881,375,374,359,373,343,358,341,325,791,791,1123,1122,-703,1105,1045,-719,865,865,790,790,774,774,1104,1029,338,293,323,308,-799,-815,833,788,772,818,803,816,322,292,307,320,561,531,515,546,289,274,288,258, + -251,-525,-605,-685,-765,-831,-846,1298,1057,1057,1312,1282,785,785,785,785,784,784,784,784,769,769,769,769,512,512,512,512,512,512,512,512,1399,1398,1383,1367,1382,1396,1351,-511,1381,1366,1139,1139,1079,1079,1124,1124,1364,1349,1363,1333,882,882,882,882,807,807,807,807,1094,1094,1136,1136,373,341,535,535,881,775,867,822,774,-591,324,338,-671,849,550,550,866,864,609,609,293,336,534,534,789,835,773,-751,834,804,308,307,833,788,832,772,562,562,547,547,305,275,560,515,290,290, + -252,-397,-477,-557,-622,-653,-719,-735,-750,1329,1299,1314,1057,1057,1042,1042,1312,1282,1024,1024,785,785,785,785,784,784,784,784,769,769,769,769,-383,1127,1141,1111,1126,1140,1095,1110,869,869,883,883,1079,1109,882,882,375,374,807,868,838,881,791,-463,867,822,368,263,852,837,836,-543,610,610,550,550,352,336,534,534,865,774,851,821,850,805,593,533,579,564,773,832,578,578,548,548,577,577,307,276,306,291,516,560,259,259, + -250,-2107,-2507,-2764,-2909,-2974,-3007,-3023,1041,1041,1040,1040,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,-767,-1052,-1213,-1277,-1358,-1405,-1469,-1535,-1550,-1582,-1614,-1647,-1662,-1694,-1726,-1759,-1774,-1807,-1822,-1854,-1886,1565,-1919,-1935,-1951,-1967,1731,1730,1580,1717,-1983,1729,1564,-1999,1548,-2015,-2031,1715,1595,-2047,1714,-2063,1610,-2079,1609,-2095,1323,1323,1457,1457,1307,1307,1712,1547,1641,1700,1699,1594,1685,1625,1442,1442,1322,1322,-780,-973,-910,1279,1278,1277,1262,1276,1261,1275,1215,1260,1229,-959,974,974,989,989,-943,735,478,478,495,463,506,414,-1039,1003,958,1017,927,942,987,957,431,476,1272,1167,1228,-1183,1256,-1199,895,895,941,941,1242,1227,1212,1135,1014,1014,490,489,503,487,910,1013,985,925,863,894,970,955,1012,847,-1343,831,755,755,984,909,428,366,754,559,-1391,752,486,457,924,997,698,698,983,893,740,740,908,877,739,739,667,667,953,938,497,287,271,271,683,606,590,712,726,574,302,302,738,736,481,286,526,725,605,711,636,724,696,651,589,681,666,710,364,467,573,695,466,466,301,465,379,379,709,604,665,679,316,316,634,633,436,436,464,269,424,394,452,332,438,363,347,408,393,448,331,422,362,407,392,421,346,406,391,376,375,359,1441,1306,-2367,1290,-2383,1337,-2399,-2415,1426,1321,-2431,1411,1336,-2447,-2463,-2479,1169,1169,1049,1049,1424,1289,1412,1352,1319,-2495,1154,1154,1064,1064,1153,1153,416,390,360,404,403,389,344,374,373,343,358,372,327,357,342,311,356,326,1395,1394,1137,1137,1047,1047,1365,1392,1287,1379,1334,1364,1349,1378,1318,1363,792,792,792,792,1152,1152,1032,1032,1121,1121,1046,1046,1120,1120,1030,1030,-2895,1106,1061,1104,849,849,789,789,1091,1076,1029,1090,1060,1075,833,833,309,324,532,532,832,772,818,803,561,561,531,560,515,546,289,274,288,258, + -250,-1179,-1579,-1836,-1996,-2124,-2253,-2333,-2413,-2477,-2542,-2574,-2607,-2622,-2655,1314,1313,1298,1312,1282,785,785,785,785,1040,1040,1025,1025,768,768,768,768,-766,-798,-830,-862,-895,-911,-927,-943,-959,-975,-991,-1007,-1023,-1039,-1055,-1070,1724,1647,-1103,-1119,1631,1767,1662,1738,1708,1723,-1135,1780,1615,1779,1599,1677,1646,1778,1583,-1151,1777,1567,1737,1692,1765,1722,1707,1630,1751,1661,1764,1614,1736,1676,1763,1750,1645,1598,1721,1691,1762,1706,1582,1761,1566,-1167,1749,1629,767,766,751,765,494,494,735,764,719,749,734,763,447,447,748,718,477,506,431,491,446,476,461,505,415,430,475,445,504,399,460,489,414,503,383,474,429,459,502,502,746,752,488,398,501,473,413,472,486,271,480,270,-1439,-1455,1357,-1471,-1487,-1503,1341,1325,-1519,1489,1463,1403,1309,-1535,1372,1448,1418,1476,1356,1462,1387,-1551,1475,1340,1447,1402,1386,-1567,1068,1068,1474,1461,455,380,468,440,395,425,410,454,364,467,466,464,453,269,409,448,268,432,1371,1473,1432,1417,1308,1460,1355,1446,1459,1431,1083,1083,1401,1416,1458,1445,1067,1067,1370,1457,1051,1051,1291,1430,1385,1444,1354,1415,1400,1443,1082,1082,1173,1113,1186,1066,1185,1050,-1967,1158,1128,1172,1097,1171,1081,-1983,1157,1112,416,266,375,400,1170,1142,1127,1065,793,793,1169,1033,1156,1096,1141,1111,1155,1080,1126,1140,898,898,808,808,897,897,792,792,1095,1152,1032,1125,1110,1139,1079,1124,882,807,838,881,853,791,-2319,867,368,263,822,852,837,866,806,865,-2399,851,352,262,534,534,821,836,594,594,549,549,593,593,533,533,848,773,579,579,564,578,548,563,276,276,577,576,306,291,516,560,305,305,275,259, + -251,-892,-2058,-2620,-2828,-2957,-3023,-3039,1041,1041,1040,1040,769,769,769,769,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,-511,-527,-543,-559,1530,-575,-591,1528,1527,1407,1526,1391,1023,1023,1023,1023,1525,1375,1268,1268,1103,1103,1087,1087,1039,1039,1523,-604,815,815,815,815,510,495,509,479,508,463,507,447,431,505,415,399,-734,-782,1262,-815,1259,1244,-831,1258,1228,-847,-863,1196,-879,1253,987,987,748,-767,493,493,462,477,414,414,686,669,478,446,461,445,474,429,487,458,412,471,1266,1264,1009,1009,799,799,-1019,-1276,-1452,-1581,-1677,-1757,-1821,-1886,-1933,-1997,1257,1257,1483,1468,1512,1422,1497,1406,1467,1496,1421,1510,1134,1134,1225,1225,1466,1451,1374,1405,1252,1252,1358,1480,1164,1164,1251,1251,1238,1238,1389,1465,-1407,1054,1101,-1423,1207,-1439,830,830,1248,1038,1237,1117,1223,1148,1236,1208,411,426,395,410,379,269,1193,1222,1132,1235,1221,1116,976,976,1192,1162,1177,1220,1131,1191,963,963,-1647,961,780,-1663,558,558,994,993,437,408,393,407,829,978,813,797,947,-1743,721,721,377,392,844,950,828,890,706,706,812,859,796,960,948,843,934,874,571,571,-1919,690,555,689,421,346,539,539,944,779,918,873,932,842,903,888,570,570,931,917,674,674,-2575,1562,-2591,1609,-2607,1654,1322,1322,1441,1441,1696,1546,1683,1593,1669,1624,1426,1426,1321,1321,1639,1680,1425,1425,1305,1305,1545,1668,1608,1623,1667,1592,1638,1666,1320,1320,1652,1607,1409,1409,1304,1304,1288,1288,1664,1637,1395,1395,1335,1335,1622,1636,1394,1394,1319,1319,1606,1621,1392,1392,1137,1137,1137,1137,345,390,360,375,404,373,1047,-2751,-2767,-2783,1062,1121,1046,-2799,1077,-2815,1106,1061,789,789,1105,1104,263,355,310,340,325,354,352,262,339,324,1091,1076,1029,1090,1060,1075,833,833,788,788,1088,1028,818,818,803,803,561,561,531,531,816,771,546,546,289,274,288,258, + -253,-317,-381,-446,-478,-509,1279,1279,-811,-1179,-1451,-1756,-1900,-2028,-2189,-2253,-2333,-2414,-2445,-2511,-2526,1313,1298,-2559,1041,1041,1040,1040,1025,1025,1024,1024,1022,1007,1021,991,1020,975,1019,959,687,687,1018,1017,671,671,655,655,1016,1015,639,639,758,758,623,623,757,607,756,591,755,575,754,559,543,543,1009,783,-575,-621,-685,-749,496,-590,750,749,734,748,974,989,1003,958,988,973,1002,942,987,957,972,1001,926,986,941,971,956,1000,910,985,925,999,894,970,-1071,-1087,-1102,1390,-1135,1436,1509,1451,1374,-1151,1405,1358,1480,1420,-1167,1507,1494,1389,1342,1465,1435,1450,1326,1505,1310,1493,1373,1479,1404,1492,1464,1419,428,443,472,397,736,526,464,464,486,457,442,471,484,482,1357,1449,1434,1478,1388,1491,1341,1490,1325,1489,1463,1403,1309,1477,1372,1448,1418,1433,1476,1356,1462,1387,-1439,1475,1340,1447,1402,1474,1324,1461,1371,1473,269,448,1432,1417,1308,1460,-1711,1459,-1727,1441,1099,1099,1446,1386,1431,1401,-1743,1289,1083,1083,1160,1160,1458,1445,1067,1067,1370,1457,1307,1430,1129,1129,1098,1098,268,432,267,416,266,400,-1887,1144,1187,1082,1173,1113,1186,1066,1050,1158,1128,1143,1172,1097,1171,1081,420,391,1157,1112,1170,1142,1127,1065,1169,1049,1156,1096,1141,1111,1155,1080,1126,1154,1064,1153,1140,1095,1048,-2159,1125,1110,1137,-2175,823,823,1139,1138,807,807,384,264,368,263,868,838,853,791,867,822,852,837,866,806,865,790,-2319,851,821,836,352,262,850,805,849,-2399,533,533,835,820,336,261,578,548,563,577,532,532,832,772,562,562,547,547,305,275,560,515,290,290,288,258 }; + static const ma_uint8 tab32[] = { 130,162,193,209,44,28,76,140,9,9,9,9,9,9,9,9,190,254,222,238,126,94,157,157,109,61,173,205}; + static const ma_uint8 tab33[] = { 252,236,220,204,188,172,156,140,124,108,92,76,60,44,28,12 }; + static const ma_int16 tabindex[2*16] = { 0,32,64,98,0,132,180,218,292,364,426,538,648,746,0,1126,1460,1460,1460,1460,1460,1460,1460,1460,1842,1842,1842,1842,1842,1842,1842,1842 }; + static const ma_uint8 g_linbits[] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,2,3,4,6,8,10,13,4,5,6,7,8,9,11,13 }; +#define MA_DR_MP3_PEEK_BITS(n) (bs_cache >> (32 - (n))) +#define MA_DR_MP3_FLUSH_BITS(n) { bs_cache <<= (n); bs_sh += (n); } +#define MA_DR_MP3_CHECK_BITS while (bs_sh >= 0) { bs_cache |= (ma_uint32)*bs_next_ptr++ << bs_sh; bs_sh -= 8; } +#define MA_DR_MP3_BSPOS ((bs_next_ptr - bs->buf)*8 - 24 + bs_sh) + float one = 0.0f; + int ireg = 0, big_val_cnt = gr_info->big_values; + const ma_uint8 *sfb = gr_info->sfbtab; + const ma_uint8 *bs_next_ptr = bs->buf + bs->pos/8; + ma_uint32 bs_cache = (((bs_next_ptr[0]*256u + bs_next_ptr[1])*256u + bs_next_ptr[2])*256u + bs_next_ptr[3]) << (bs->pos & 7); + int pairs_to_decode, np, bs_sh = (bs->pos & 7) - 8; + bs_next_ptr += 4; + while (big_val_cnt > 0) + { + int tab_num = gr_info->table_select[ireg]; + int sfb_cnt = gr_info->region_count[ireg++]; + const ma_int16 *codebook = tabs + tabindex[tab_num]; + int linbits = g_linbits[tab_num]; + if (linbits) + { + do + { + np = *sfb++ / 2; + pairs_to_decode = MA_DR_MP3_MIN(big_val_cnt, np); + one = *scf++; + do + { + int j, w = 5; + int leaf = codebook[MA_DR_MP3_PEEK_BITS(w)]; + while (leaf < 0) + { + MA_DR_MP3_FLUSH_BITS(w); + w = leaf & 7; + leaf = codebook[MA_DR_MP3_PEEK_BITS(w) - (leaf >> 3)]; + } + MA_DR_MP3_FLUSH_BITS(leaf >> 8); + for (j = 0; j < 2; j++, dst++, leaf >>= 4) + { + int lsb = leaf & 0x0F; + if (lsb == 15) + { + lsb += MA_DR_MP3_PEEK_BITS(linbits); + MA_DR_MP3_FLUSH_BITS(linbits); + MA_DR_MP3_CHECK_BITS; + *dst = one*ma_dr_mp3_L3_pow_43(lsb)*((ma_int32)bs_cache < 0 ? -1: 1); + } else + { + *dst = g_ma_dr_mp3_pow43[16 + lsb - 16*(bs_cache >> 31)]*one; + } + MA_DR_MP3_FLUSH_BITS(lsb ? 1 : 0); + } + MA_DR_MP3_CHECK_BITS; + } while (--pairs_to_decode); + } while ((big_val_cnt -= np) > 0 && --sfb_cnt >= 0); + } else + { + do + { + np = *sfb++ / 2; + pairs_to_decode = MA_DR_MP3_MIN(big_val_cnt, np); + one = *scf++; + do + { + int j, w = 5; + int leaf = codebook[MA_DR_MP3_PEEK_BITS(w)]; + while (leaf < 0) + { + MA_DR_MP3_FLUSH_BITS(w); + w = leaf & 7; + leaf = codebook[MA_DR_MP3_PEEK_BITS(w) - (leaf >> 3)]; + } + MA_DR_MP3_FLUSH_BITS(leaf >> 8); + for (j = 0; j < 2; j++, dst++, leaf >>= 4) + { + int lsb = leaf & 0x0F; + *dst = g_ma_dr_mp3_pow43[16 + lsb - 16*(bs_cache >> 31)]*one; + MA_DR_MP3_FLUSH_BITS(lsb ? 1 : 0); + } + MA_DR_MP3_CHECK_BITS; + } while (--pairs_to_decode); + } while ((big_val_cnt -= np) > 0 && --sfb_cnt >= 0); + } + } + for (np = 1 - big_val_cnt;; dst += 4) + { + const ma_uint8 *codebook_count1 = (gr_info->count1_table) ? tab33 : tab32; + int leaf = codebook_count1[MA_DR_MP3_PEEK_BITS(4)]; + if (!(leaf & 8)) + { + leaf = codebook_count1[(leaf >> 3) + (bs_cache << 4 >> (32 - (leaf & 3)))]; + } + MA_DR_MP3_FLUSH_BITS(leaf & 7); + if (MA_DR_MP3_BSPOS > layer3gr_limit) + { + break; + } +#define MA_DR_MP3_RELOAD_SCALEFACTOR if (!--np) { np = *sfb++/2; if (!np) break; one = *scf++; } +#define MA_DR_MP3_DEQ_COUNT1(s) if (leaf & (128 >> s)) { dst[s] = ((ma_int32)bs_cache < 0) ? -one : one; MA_DR_MP3_FLUSH_BITS(1) } + MA_DR_MP3_RELOAD_SCALEFACTOR; + MA_DR_MP3_DEQ_COUNT1(0); + MA_DR_MP3_DEQ_COUNT1(1); + MA_DR_MP3_RELOAD_SCALEFACTOR; + MA_DR_MP3_DEQ_COUNT1(2); + MA_DR_MP3_DEQ_COUNT1(3); + MA_DR_MP3_CHECK_BITS; + } + bs->pos = layer3gr_limit; +} +static void ma_dr_mp3_L3_midside_stereo(float *left, int n) +{ + int i = 0; + float *right = left + 576; +#if MA_DR_MP3_HAVE_SIMD + if (ma_dr_mp3_have_simd()) + { + for (; i < n - 3; i += 4) + { + ma_dr_mp3_f4 vl = MA_DR_MP3_VLD(left + i); + ma_dr_mp3_f4 vr = MA_DR_MP3_VLD(right + i); + MA_DR_MP3_VSTORE(left + i, MA_DR_MP3_VADD(vl, vr)); + MA_DR_MP3_VSTORE(right + i, MA_DR_MP3_VSUB(vl, vr)); + } +#ifdef __GNUC__ + if (__builtin_constant_p(n % 4 == 0) && n % 4 == 0) + return; +#endif + } +#endif + for (; i < n; i++) + { + float a = left[i]; + float b = right[i]; + left[i] = a + b; + right[i] = a - b; + } +} +static void ma_dr_mp3_L3_intensity_stereo_band(float *left, int n, float kl, float kr) +{ + int i; + for (i = 0; i < n; i++) + { + left[i + 576] = left[i]*kr; + left[i] = left[i]*kl; + } +} +static void ma_dr_mp3_L3_stereo_top_band(const float *right, const ma_uint8 *sfb, int nbands, int max_band[3]) +{ + int i, k; + max_band[0] = max_band[1] = max_band[2] = -1; + for (i = 0; i < nbands; i++) + { + for (k = 0; k < sfb[i]; k += 2) + { + if (right[k] != 0 || right[k + 1] != 0) + { + max_band[i % 3] = i; + break; + } + } + right += sfb[i]; + } +} +static void ma_dr_mp3_L3_stereo_process(float *left, const ma_uint8 *ist_pos, const ma_uint8 *sfb, const ma_uint8 *hdr, int max_band[3], int mpeg2_sh) +{ + static const float g_pan[7*2] = { 0,1,0.21132487f,0.78867513f,0.36602540f,0.63397460f,0.5f,0.5f,0.63397460f,0.36602540f,0.78867513f,0.21132487f,1,0 }; + unsigned i, max_pos = MA_DR_MP3_HDR_TEST_MPEG1(hdr) ? 7 : 64; + for (i = 0; sfb[i]; i++) + { + unsigned ipos = ist_pos[i]; + if ((int)i > max_band[i % 3] && ipos < max_pos) + { + float kl, kr, s = MA_DR_MP3_HDR_TEST_MS_STEREO(hdr) ? 1.41421356f : 1; + if (MA_DR_MP3_HDR_TEST_MPEG1(hdr)) + { + kl = g_pan[2*ipos]; + kr = g_pan[2*ipos + 1]; + } else + { + kl = 1; + kr = ma_dr_mp3_L3_ldexp_q2(1, (ipos + 1) >> 1 << mpeg2_sh); + if (ipos & 1) + { + kl = kr; + kr = 1; + } + } + ma_dr_mp3_L3_intensity_stereo_band(left, sfb[i], kl*s, kr*s); + } else if (MA_DR_MP3_HDR_TEST_MS_STEREO(hdr)) + { + ma_dr_mp3_L3_midside_stereo(left, sfb[i]); + } + left += sfb[i]; + } +} +static void ma_dr_mp3_L3_intensity_stereo(float *left, ma_uint8 *ist_pos, const ma_dr_mp3_L3_gr_info *gr, const ma_uint8 *hdr) +{ + int max_band[3], n_sfb = gr->n_long_sfb + gr->n_short_sfb; + int i, max_blocks = gr->n_short_sfb ? 3 : 1; + ma_dr_mp3_L3_stereo_top_band(left + 576, gr->sfbtab, n_sfb, max_band); + if (gr->n_long_sfb) + { + max_band[0] = max_band[1] = max_band[2] = MA_DR_MP3_MAX(MA_DR_MP3_MAX(max_band[0], max_band[1]), max_band[2]); + } + for (i = 0; i < max_blocks; i++) + { + int default_pos = MA_DR_MP3_HDR_TEST_MPEG1(hdr) ? 3 : 0; + int itop = n_sfb - max_blocks + i; + int prev = itop - max_blocks; + ist_pos[itop] = (ma_uint8)(max_band[i] >= prev ? default_pos : ist_pos[prev]); + } + ma_dr_mp3_L3_stereo_process(left, ist_pos, gr->sfbtab, hdr, max_band, gr[1].scalefac_compress & 1); +} +static void ma_dr_mp3_L3_reorder(float *grbuf, float *scratch, const ma_uint8 *sfb) +{ + int i, len; + float *src = grbuf, *dst = scratch; + for (;0 != (len = *sfb); sfb += 3, src += 2*len) + { + for (i = 0; i < len; i++, src++) + { + *dst++ = src[0*len]; + *dst++ = src[1*len]; + *dst++ = src[2*len]; + } + } + MA_DR_MP3_COPY_MEMORY(grbuf, scratch, (dst - scratch)*sizeof(float)); +} +static void ma_dr_mp3_L3_antialias(float *grbuf, int nbands) +{ + static const float g_aa[2][8] = { + {0.85749293f,0.88174200f,0.94962865f,0.98331459f,0.99551782f,0.99916056f,0.99989920f,0.99999316f}, + {0.51449576f,0.47173197f,0.31337745f,0.18191320f,0.09457419f,0.04096558f,0.01419856f,0.00369997f} + }; + for (; nbands > 0; nbands--, grbuf += 18) + { + int i = 0; +#if MA_DR_MP3_HAVE_SIMD + if (ma_dr_mp3_have_simd()) for (; i < 8; i += 4) + { + ma_dr_mp3_f4 vu = MA_DR_MP3_VLD(grbuf + 18 + i); + ma_dr_mp3_f4 vd = MA_DR_MP3_VLD(grbuf + 14 - i); + ma_dr_mp3_f4 vc0 = MA_DR_MP3_VLD(g_aa[0] + i); + ma_dr_mp3_f4 vc1 = MA_DR_MP3_VLD(g_aa[1] + i); + vd = MA_DR_MP3_VREV(vd); + MA_DR_MP3_VSTORE(grbuf + 18 + i, MA_DR_MP3_VSUB(MA_DR_MP3_VMUL(vu, vc0), MA_DR_MP3_VMUL(vd, vc1))); + vd = MA_DR_MP3_VADD(MA_DR_MP3_VMUL(vu, vc1), MA_DR_MP3_VMUL(vd, vc0)); + MA_DR_MP3_VSTORE(grbuf + 14 - i, MA_DR_MP3_VREV(vd)); + } +#endif +#ifndef MA_DR_MP3_ONLY_SIMD + for(; i < 8; i++) + { + float u = grbuf[18 + i]; + float d = grbuf[17 - i]; + grbuf[18 + i] = u*g_aa[0][i] - d*g_aa[1][i]; + grbuf[17 - i] = u*g_aa[1][i] + d*g_aa[0][i]; + } +#endif + } +} +static void ma_dr_mp3_L3_dct3_9(float *y) +{ + float s0, s1, s2, s3, s4, s5, s6, s7, s8, t0, t2, t4; + s0 = y[0]; s2 = y[2]; s4 = y[4]; s6 = y[6]; s8 = y[8]; + t0 = s0 + s6*0.5f; + s0 -= s6; + t4 = (s4 + s2)*0.93969262f; + t2 = (s8 + s2)*0.76604444f; + s6 = (s4 - s8)*0.17364818f; + s4 += s8 - s2; + s2 = s0 - s4*0.5f; + y[4] = s4 + s0; + s8 = t0 - t2 + s6; + s0 = t0 - t4 + t2; + s4 = t0 + t4 - s6; + s1 = y[1]; s3 = y[3]; s5 = y[5]; s7 = y[7]; + s3 *= 0.86602540f; + t0 = (s5 + s1)*0.98480775f; + t4 = (s5 - s7)*0.34202014f; + t2 = (s1 + s7)*0.64278761f; + s1 = (s1 - s5 - s7)*0.86602540f; + s5 = t0 - s3 - t2; + s7 = t4 - s3 - t0; + s3 = t4 + s3 - t2; + y[0] = s4 - s7; + y[1] = s2 + s1; + y[2] = s0 - s3; + y[3] = s8 + s5; + y[5] = s8 - s5; + y[6] = s0 + s3; + y[7] = s2 - s1; + y[8] = s4 + s7; +} +static void ma_dr_mp3_L3_imdct36(float *grbuf, float *overlap, const float *window, int nbands) +{ + int i, j; + static const float g_twid9[18] = { + 0.73727734f,0.79335334f,0.84339145f,0.88701083f,0.92387953f,0.95371695f,0.97629601f,0.99144486f,0.99904822f,0.67559021f,0.60876143f,0.53729961f,0.46174861f,0.38268343f,0.30070580f,0.21643961f,0.13052619f,0.04361938f + }; + for (j = 0; j < nbands; j++, grbuf += 18, overlap += 9) + { + float co[9], si[9]; + co[0] = -grbuf[0]; + si[0] = grbuf[17]; + for (i = 0; i < 4; i++) + { + si[8 - 2*i] = grbuf[4*i + 1] - grbuf[4*i + 2]; + co[1 + 2*i] = grbuf[4*i + 1] + grbuf[4*i + 2]; + si[7 - 2*i] = grbuf[4*i + 4] - grbuf[4*i + 3]; + co[2 + 2*i] = -(grbuf[4*i + 3] + grbuf[4*i + 4]); + } + ma_dr_mp3_L3_dct3_9(co); + ma_dr_mp3_L3_dct3_9(si); + si[1] = -si[1]; + si[3] = -si[3]; + si[5] = -si[5]; + si[7] = -si[7]; + i = 0; +#if MA_DR_MP3_HAVE_SIMD + if (ma_dr_mp3_have_simd()) for (; i < 8; i += 4) + { + ma_dr_mp3_f4 vovl = MA_DR_MP3_VLD(overlap + i); + ma_dr_mp3_f4 vc = MA_DR_MP3_VLD(co + i); + ma_dr_mp3_f4 vs = MA_DR_MP3_VLD(si + i); + ma_dr_mp3_f4 vr0 = MA_DR_MP3_VLD(g_twid9 + i); + ma_dr_mp3_f4 vr1 = MA_DR_MP3_VLD(g_twid9 + 9 + i); + ma_dr_mp3_f4 vw0 = MA_DR_MP3_VLD(window + i); + ma_dr_mp3_f4 vw1 = MA_DR_MP3_VLD(window + 9 + i); + ma_dr_mp3_f4 vsum = MA_DR_MP3_VADD(MA_DR_MP3_VMUL(vc, vr1), MA_DR_MP3_VMUL(vs, vr0)); + MA_DR_MP3_VSTORE(overlap + i, MA_DR_MP3_VSUB(MA_DR_MP3_VMUL(vc, vr0), MA_DR_MP3_VMUL(vs, vr1))); + MA_DR_MP3_VSTORE(grbuf + i, MA_DR_MP3_VSUB(MA_DR_MP3_VMUL(vovl, vw0), MA_DR_MP3_VMUL(vsum, vw1))); + vsum = MA_DR_MP3_VADD(MA_DR_MP3_VMUL(vovl, vw1), MA_DR_MP3_VMUL(vsum, vw0)); + MA_DR_MP3_VSTORE(grbuf + 14 - i, MA_DR_MP3_VREV(vsum)); + } +#endif + for (; i < 9; i++) + { + float ovl = overlap[i]; + float sum = co[i]*g_twid9[9 + i] + si[i]*g_twid9[0 + i]; + overlap[i] = co[i]*g_twid9[0 + i] - si[i]*g_twid9[9 + i]; + grbuf[i] = ovl*window[0 + i] - sum*window[9 + i]; + grbuf[17 - i] = ovl*window[9 + i] + sum*window[0 + i]; + } + } +} +static void ma_dr_mp3_L3_idct3(float x0, float x1, float x2, float *dst) +{ + float m1 = x1*0.86602540f; + float a1 = x0 - x2*0.5f; + dst[1] = x0 + x2; + dst[0] = a1 + m1; + dst[2] = a1 - m1; +} +static void ma_dr_mp3_L3_imdct12(float *x, float *dst, float *overlap) +{ + static const float g_twid3[6] = { 0.79335334f,0.92387953f,0.99144486f, 0.60876143f,0.38268343f,0.13052619f }; + float co[3], si[3]; + int i; + ma_dr_mp3_L3_idct3(-x[0], x[6] + x[3], x[12] + x[9], co); + ma_dr_mp3_L3_idct3(x[15], x[12] - x[9], x[6] - x[3], si); + si[1] = -si[1]; + for (i = 0; i < 3; i++) + { + float ovl = overlap[i]; + float sum = co[i]*g_twid3[3 + i] + si[i]*g_twid3[0 + i]; + overlap[i] = co[i]*g_twid3[0 + i] - si[i]*g_twid3[3 + i]; + dst[i] = ovl*g_twid3[2 - i] - sum*g_twid3[5 - i]; + dst[5 - i] = ovl*g_twid3[5 - i] + sum*g_twid3[2 - i]; + } +} +static void ma_dr_mp3_L3_imdct_short(float *grbuf, float *overlap, int nbands) +{ + for (;nbands > 0; nbands--, overlap += 9, grbuf += 18) + { + float tmp[18]; + MA_DR_MP3_COPY_MEMORY(tmp, grbuf, sizeof(tmp)); + MA_DR_MP3_COPY_MEMORY(grbuf, overlap, 6*sizeof(float)); + ma_dr_mp3_L3_imdct12(tmp, grbuf + 6, overlap + 6); + ma_dr_mp3_L3_imdct12(tmp + 1, grbuf + 12, overlap + 6); + ma_dr_mp3_L3_imdct12(tmp + 2, overlap, overlap + 6); + } +} +static void ma_dr_mp3_L3_change_sign(float *grbuf) +{ + int b, i; + for (b = 0, grbuf += 18; b < 32; b += 2, grbuf += 36) + for (i = 1; i < 18; i += 2) + grbuf[i] = -grbuf[i]; +} +static void ma_dr_mp3_L3_imdct_gr(float *grbuf, float *overlap, unsigned block_type, unsigned n_long_bands) +{ + static const float g_mdct_window[2][18] = { + { 0.99904822f,0.99144486f,0.97629601f,0.95371695f,0.92387953f,0.88701083f,0.84339145f,0.79335334f,0.73727734f,0.04361938f,0.13052619f,0.21643961f,0.30070580f,0.38268343f,0.46174861f,0.53729961f,0.60876143f,0.67559021f }, + { 1,1,1,1,1,1,0.99144486f,0.92387953f,0.79335334f,0,0,0,0,0,0,0.13052619f,0.38268343f,0.60876143f } + }; + if (n_long_bands) + { + ma_dr_mp3_L3_imdct36(grbuf, overlap, g_mdct_window[0], n_long_bands); + grbuf += 18*n_long_bands; + overlap += 9*n_long_bands; + } + if (block_type == MA_DR_MP3_SHORT_BLOCK_TYPE) + ma_dr_mp3_L3_imdct_short(grbuf, overlap, 32 - n_long_bands); + else + ma_dr_mp3_L3_imdct36(grbuf, overlap, g_mdct_window[block_type == MA_DR_MP3_STOP_BLOCK_TYPE], 32 - n_long_bands); +} +static void ma_dr_mp3_L3_save_reservoir(ma_dr_mp3dec *h, ma_dr_mp3dec_scratch *s) +{ + int pos = (s->bs.pos + 7)/8u; + int remains = s->bs.limit/8u - pos; + if (remains > MA_DR_MP3_MAX_BITRESERVOIR_BYTES) + { + pos += remains - MA_DR_MP3_MAX_BITRESERVOIR_BYTES; + remains = MA_DR_MP3_MAX_BITRESERVOIR_BYTES; + } + if (remains > 0) + { + MA_DR_MP3_MOVE_MEMORY(h->reserv_buf, s->maindata + pos, remains); + } + h->reserv = remains; +} +static int ma_dr_mp3_L3_restore_reservoir(ma_dr_mp3dec *h, ma_dr_mp3_bs *bs, ma_dr_mp3dec_scratch *s, int main_data_begin) +{ + int frame_bytes = (bs->limit - bs->pos)/8; + int bytes_have = MA_DR_MP3_MIN(h->reserv, main_data_begin); + MA_DR_MP3_COPY_MEMORY(s->maindata, h->reserv_buf + MA_DR_MP3_MAX(0, h->reserv - main_data_begin), MA_DR_MP3_MIN(h->reserv, main_data_begin)); + MA_DR_MP3_COPY_MEMORY(s->maindata + bytes_have, bs->buf + bs->pos/8, frame_bytes); + ma_dr_mp3_bs_init(&s->bs, s->maindata, bytes_have + frame_bytes); + return h->reserv >= main_data_begin; +} +static void ma_dr_mp3_L3_decode(ma_dr_mp3dec *h, ma_dr_mp3dec_scratch *s, ma_dr_mp3_L3_gr_info *gr_info, int nch) +{ + int ch; + for (ch = 0; ch < nch; ch++) + { + int layer3gr_limit = s->bs.pos + gr_info[ch].part_23_length; + ma_dr_mp3_L3_decode_scalefactors(h->header, s->ist_pos[ch], &s->bs, gr_info + ch, s->scf, ch); + ma_dr_mp3_L3_huffman(s->grbuf[ch], &s->bs, gr_info + ch, s->scf, layer3gr_limit); + } + if (MA_DR_MP3_HDR_TEST_I_STEREO(h->header)) + { + ma_dr_mp3_L3_intensity_stereo(s->grbuf[0], s->ist_pos[1], gr_info, h->header); + } else if (MA_DR_MP3_HDR_IS_MS_STEREO(h->header)) + { + ma_dr_mp3_L3_midside_stereo(s->grbuf[0], 576); + } + for (ch = 0; ch < nch; ch++, gr_info++) + { + int aa_bands = 31; + int n_long_bands = (gr_info->mixed_block_flag ? 2 : 0) << (int)(MA_DR_MP3_HDR_GET_MY_SAMPLE_RATE(h->header) == 2); + if (gr_info->n_short_sfb) + { + aa_bands = n_long_bands - 1; + ma_dr_mp3_L3_reorder(s->grbuf[ch] + n_long_bands*18, s->syn[0], gr_info->sfbtab + gr_info->n_long_sfb); + } + ma_dr_mp3_L3_antialias(s->grbuf[ch], aa_bands); + ma_dr_mp3_L3_imdct_gr(s->grbuf[ch], h->mdct_overlap[ch], gr_info->block_type, n_long_bands); + ma_dr_mp3_L3_change_sign(s->grbuf[ch]); + } +} +static void ma_dr_mp3d_DCT_II(float *grbuf, int n) +{ + static const float g_sec[24] = { + 10.19000816f,0.50060302f,0.50241929f,3.40760851f,0.50547093f,0.52249861f,2.05778098f,0.51544732f,0.56694406f,1.48416460f,0.53104258f,0.64682180f,1.16943991f,0.55310392f,0.78815460f,0.97256821f,0.58293498f,1.06067765f,0.83934963f,0.62250412f,1.72244716f,0.74453628f,0.67480832f,5.10114861f + }; + int i, k = 0; +#if MA_DR_MP3_HAVE_SIMD + if (ma_dr_mp3_have_simd()) for (; k < n; k += 4) + { + ma_dr_mp3_f4 t[4][8], *x; + float *y = grbuf + k; + for (x = t[0], i = 0; i < 8; i++, x++) + { + ma_dr_mp3_f4 x0 = MA_DR_MP3_VLD(&y[i*18]); + ma_dr_mp3_f4 x1 = MA_DR_MP3_VLD(&y[(15 - i)*18]); + ma_dr_mp3_f4 x2 = MA_DR_MP3_VLD(&y[(16 + i)*18]); + ma_dr_mp3_f4 x3 = MA_DR_MP3_VLD(&y[(31 - i)*18]); + ma_dr_mp3_f4 t0 = MA_DR_MP3_VADD(x0, x3); + ma_dr_mp3_f4 t1 = MA_DR_MP3_VADD(x1, x2); + ma_dr_mp3_f4 t2 = MA_DR_MP3_VMUL_S(MA_DR_MP3_VSUB(x1, x2), g_sec[3*i + 0]); + ma_dr_mp3_f4 t3 = MA_DR_MP3_VMUL_S(MA_DR_MP3_VSUB(x0, x3), g_sec[3*i + 1]); + x[0] = MA_DR_MP3_VADD(t0, t1); + x[8] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VSUB(t0, t1), g_sec[3*i + 2]); + x[16] = MA_DR_MP3_VADD(t3, t2); + x[24] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VSUB(t3, t2), g_sec[3*i + 2]); + } + for (x = t[0], i = 0; i < 4; i++, x += 8) + { + ma_dr_mp3_f4 x0 = x[0], x1 = x[1], x2 = x[2], x3 = x[3], x4 = x[4], x5 = x[5], x6 = x[6], x7 = x[7], xt; + xt = MA_DR_MP3_VSUB(x0, x7); x0 = MA_DR_MP3_VADD(x0, x7); + x7 = MA_DR_MP3_VSUB(x1, x6); x1 = MA_DR_MP3_VADD(x1, x6); + x6 = MA_DR_MP3_VSUB(x2, x5); x2 = MA_DR_MP3_VADD(x2, x5); + x5 = MA_DR_MP3_VSUB(x3, x4); x3 = MA_DR_MP3_VADD(x3, x4); + x4 = MA_DR_MP3_VSUB(x0, x3); x0 = MA_DR_MP3_VADD(x0, x3); + x3 = MA_DR_MP3_VSUB(x1, x2); x1 = MA_DR_MP3_VADD(x1, x2); + x[0] = MA_DR_MP3_VADD(x0, x1); + x[4] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VSUB(x0, x1), 0.70710677f); + x5 = MA_DR_MP3_VADD(x5, x6); + x6 = MA_DR_MP3_VMUL_S(MA_DR_MP3_VADD(x6, x7), 0.70710677f); + x7 = MA_DR_MP3_VADD(x7, xt); + x3 = MA_DR_MP3_VMUL_S(MA_DR_MP3_VADD(x3, x4), 0.70710677f); + x5 = MA_DR_MP3_VSUB(x5, MA_DR_MP3_VMUL_S(x7, 0.198912367f)); + x7 = MA_DR_MP3_VADD(x7, MA_DR_MP3_VMUL_S(x5, 0.382683432f)); + x5 = MA_DR_MP3_VSUB(x5, MA_DR_MP3_VMUL_S(x7, 0.198912367f)); + x0 = MA_DR_MP3_VSUB(xt, x6); xt = MA_DR_MP3_VADD(xt, x6); + x[1] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VADD(xt, x7), 0.50979561f); + x[2] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VADD(x4, x3), 0.54119611f); + x[3] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VSUB(x0, x5), 0.60134488f); + x[5] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VADD(x0, x5), 0.89997619f); + x[6] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VSUB(x4, x3), 1.30656302f); + x[7] = MA_DR_MP3_VMUL_S(MA_DR_MP3_VSUB(xt, x7), 2.56291556f); + } + if (k > n - 3) + { +#if MA_DR_MP3_HAVE_SSE +#define MA_DR_MP3_VSAVE2(i, v) _mm_storel_pi((__m64 *)(void*)&y[i*18], v) +#else +#define MA_DR_MP3_VSAVE2(i, v) vst1_f32((float32_t *)&y[(i)*18], vget_low_f32(v)) +#endif + for (i = 0; i < 7; i++, y += 4*18) + { + ma_dr_mp3_f4 s = MA_DR_MP3_VADD(t[3][i], t[3][i + 1]); + MA_DR_MP3_VSAVE2(0, t[0][i]); + MA_DR_MP3_VSAVE2(1, MA_DR_MP3_VADD(t[2][i], s)); + MA_DR_MP3_VSAVE2(2, MA_DR_MP3_VADD(t[1][i], t[1][i + 1])); + MA_DR_MP3_VSAVE2(3, MA_DR_MP3_VADD(t[2][1 + i], s)); + } + MA_DR_MP3_VSAVE2(0, t[0][7]); + MA_DR_MP3_VSAVE2(1, MA_DR_MP3_VADD(t[2][7], t[3][7])); + MA_DR_MP3_VSAVE2(2, t[1][7]); + MA_DR_MP3_VSAVE2(3, t[3][7]); + } else + { +#define MA_DR_MP3_VSAVE4(i, v) MA_DR_MP3_VSTORE(&y[(i)*18], v) + for (i = 0; i < 7; i++, y += 4*18) + { + ma_dr_mp3_f4 s = MA_DR_MP3_VADD(t[3][i], t[3][i + 1]); + MA_DR_MP3_VSAVE4(0, t[0][i]); + MA_DR_MP3_VSAVE4(1, MA_DR_MP3_VADD(t[2][i], s)); + MA_DR_MP3_VSAVE4(2, MA_DR_MP3_VADD(t[1][i], t[1][i + 1])); + MA_DR_MP3_VSAVE4(3, MA_DR_MP3_VADD(t[2][1 + i], s)); + } + MA_DR_MP3_VSAVE4(0, t[0][7]); + MA_DR_MP3_VSAVE4(1, MA_DR_MP3_VADD(t[2][7], t[3][7])); + MA_DR_MP3_VSAVE4(2, t[1][7]); + MA_DR_MP3_VSAVE4(3, t[3][7]); + } + } else +#endif +#ifdef MA_DR_MP3_ONLY_SIMD + {} +#else + for (; k < n; k++) + { + float t[4][8], *x, *y = grbuf + k; + for (x = t[0], i = 0; i < 8; i++, x++) + { + float x0 = y[i*18]; + float x1 = y[(15 - i)*18]; + float x2 = y[(16 + i)*18]; + float x3 = y[(31 - i)*18]; + float t0 = x0 + x3; + float t1 = x1 + x2; + float t2 = (x1 - x2)*g_sec[3*i + 0]; + float t3 = (x0 - x3)*g_sec[3*i + 1]; + x[0] = t0 + t1; + x[8] = (t0 - t1)*g_sec[3*i + 2]; + x[16] = t3 + t2; + x[24] = (t3 - t2)*g_sec[3*i + 2]; + } + for (x = t[0], i = 0; i < 4; i++, x += 8) + { + float x0 = x[0], x1 = x[1], x2 = x[2], x3 = x[3], x4 = x[4], x5 = x[5], x6 = x[6], x7 = x[7], xt; + xt = x0 - x7; x0 += x7; + x7 = x1 - x6; x1 += x6; + x6 = x2 - x5; x2 += x5; + x5 = x3 - x4; x3 += x4; + x4 = x0 - x3; x0 += x3; + x3 = x1 - x2; x1 += x2; + x[0] = x0 + x1; + x[4] = (x0 - x1)*0.70710677f; + x5 = x5 + x6; + x6 = (x6 + x7)*0.70710677f; + x7 = x7 + xt; + x3 = (x3 + x4)*0.70710677f; + x5 -= x7*0.198912367f; + x7 += x5*0.382683432f; + x5 -= x7*0.198912367f; + x0 = xt - x6; xt += x6; + x[1] = (xt + x7)*0.50979561f; + x[2] = (x4 + x3)*0.54119611f; + x[3] = (x0 - x5)*0.60134488f; + x[5] = (x0 + x5)*0.89997619f; + x[6] = (x4 - x3)*1.30656302f; + x[7] = (xt - x7)*2.56291556f; + } + for (i = 0; i < 7; i++, y += 4*18) + { + y[0*18] = t[0][i]; + y[1*18] = t[2][i] + t[3][i] + t[3][i + 1]; + y[2*18] = t[1][i] + t[1][i + 1]; + y[3*18] = t[2][i + 1] + t[3][i] + t[3][i + 1]; + } + y[0*18] = t[0][7]; + y[1*18] = t[2][7] + t[3][7]; + y[2*18] = t[1][7]; + y[3*18] = t[3][7]; + } +#endif +} +#ifndef MA_DR_MP3_FLOAT_OUTPUT +typedef ma_int16 ma_dr_mp3d_sample_t; +static ma_int16 ma_dr_mp3d_scale_pcm(float sample) +{ + ma_int16 s; +#if MA_DR_MP3_HAVE_ARMV6 + ma_int32 s32 = (ma_int32)(sample + .5f); + s32 -= (s32 < 0); + s = (ma_int16)ma_dr_mp3_clip_int16_arm(s32); +#else + if (sample >= 32766.5f) return (ma_int16) 32767; + if (sample <= -32767.5f) return (ma_int16)-32768; + s = (ma_int16)(sample + .5f); + s -= (s < 0); +#endif + return s; +} +#else +typedef float ma_dr_mp3d_sample_t; +static float ma_dr_mp3d_scale_pcm(float sample) +{ + return sample*(1.f/32768.f); +} +#endif +static void ma_dr_mp3d_synth_pair(ma_dr_mp3d_sample_t *pcm, int nch, const float *z) +{ + float a; + a = (z[14*64] - z[ 0]) * 29; + a += (z[ 1*64] + z[13*64]) * 213; + a += (z[12*64] - z[ 2*64]) * 459; + a += (z[ 3*64] + z[11*64]) * 2037; + a += (z[10*64] - z[ 4*64]) * 5153; + a += (z[ 5*64] + z[ 9*64]) * 6574; + a += (z[ 8*64] - z[ 6*64]) * 37489; + a += z[ 7*64] * 75038; + pcm[0] = ma_dr_mp3d_scale_pcm(a); + z += 2; + a = z[14*64] * 104; + a += z[12*64] * 1567; + a += z[10*64] * 9727; + a += z[ 8*64] * 64019; + a += z[ 6*64] * -9975; + a += z[ 4*64] * -45; + a += z[ 2*64] * 146; + a += z[ 0*64] * -5; + pcm[16*nch] = ma_dr_mp3d_scale_pcm(a); +} +static void ma_dr_mp3d_synth(float *xl, ma_dr_mp3d_sample_t *dstl, int nch, float *lins) +{ + int i; + float *xr = xl + 576*(nch - 1); + ma_dr_mp3d_sample_t *dstr = dstl + (nch - 1); + static const float g_win[] = { + -1,26,-31,208,218,401,-519,2063,2000,4788,-5517,7134,5959,35640,-39336,74992, + -1,24,-35,202,222,347,-581,2080,1952,4425,-5879,7640,5288,33791,-41176,74856, + -1,21,-38,196,225,294,-645,2087,1893,4063,-6237,8092,4561,31947,-43006,74630, + -1,19,-41,190,227,244,-711,2085,1822,3705,-6589,8492,3776,30112,-44821,74313, + -1,17,-45,183,228,197,-779,2075,1739,3351,-6935,8840,2935,28289,-46617,73908, + -1,16,-49,176,228,153,-848,2057,1644,3004,-7271,9139,2037,26482,-48390,73415, + -2,14,-53,169,227,111,-919,2032,1535,2663,-7597,9389,1082,24694,-50137,72835, + -2,13,-58,161,224,72,-991,2001,1414,2330,-7910,9592,70,22929,-51853,72169, + -2,11,-63,154,221,36,-1064,1962,1280,2006,-8209,9750,-998,21189,-53534,71420, + -2,10,-68,147,215,2,-1137,1919,1131,1692,-8491,9863,-2122,19478,-55178,70590, + -3,9,-73,139,208,-29,-1210,1870,970,1388,-8755,9935,-3300,17799,-56778,69679, + -3,8,-79,132,200,-57,-1283,1817,794,1095,-8998,9966,-4533,16155,-58333,68692, + -4,7,-85,125,189,-83,-1356,1759,605,814,-9219,9959,-5818,14548,-59838,67629, + -4,7,-91,117,177,-106,-1428,1698,402,545,-9416,9916,-7154,12980,-61289,66494, + -5,6,-97,111,163,-127,-1498,1634,185,288,-9585,9838,-8540,11455,-62684,65290 + }; + float *zlin = lins + 15*64; + const float *w = g_win; + zlin[4*15] = xl[18*16]; + zlin[4*15 + 1] = xr[18*16]; + zlin[4*15 + 2] = xl[0]; + zlin[4*15 + 3] = xr[0]; + zlin[4*31] = xl[1 + 18*16]; + zlin[4*31 + 1] = xr[1 + 18*16]; + zlin[4*31 + 2] = xl[1]; + zlin[4*31 + 3] = xr[1]; + ma_dr_mp3d_synth_pair(dstr, nch, lins + 4*15 + 1); + ma_dr_mp3d_synth_pair(dstr + 32*nch, nch, lins + 4*15 + 64 + 1); + ma_dr_mp3d_synth_pair(dstl, nch, lins + 4*15); + ma_dr_mp3d_synth_pair(dstl + 32*nch, nch, lins + 4*15 + 64); +#if MA_DR_MP3_HAVE_SIMD + if (ma_dr_mp3_have_simd()) for (i = 14; i >= 0; i--) + { +#define MA_DR_MP3_VLOAD(k) ma_dr_mp3_f4 w0 = MA_DR_MP3_VSET(*w++); ma_dr_mp3_f4 w1 = MA_DR_MP3_VSET(*w++); ma_dr_mp3_f4 vz = MA_DR_MP3_VLD(&zlin[4*i - 64*k]); ma_dr_mp3_f4 vy = MA_DR_MP3_VLD(&zlin[4*i - 64*(15 - k)]); +#define MA_DR_MP3_V0(k) { MA_DR_MP3_VLOAD(k) b = MA_DR_MP3_VADD(MA_DR_MP3_VMUL(vz, w1), MA_DR_MP3_VMUL(vy, w0)) ; a = MA_DR_MP3_VSUB(MA_DR_MP3_VMUL(vz, w0), MA_DR_MP3_VMUL(vy, w1)); } +#define MA_DR_MP3_V1(k) { MA_DR_MP3_VLOAD(k) b = MA_DR_MP3_VADD(b, MA_DR_MP3_VADD(MA_DR_MP3_VMUL(vz, w1), MA_DR_MP3_VMUL(vy, w0))); a = MA_DR_MP3_VADD(a, MA_DR_MP3_VSUB(MA_DR_MP3_VMUL(vz, w0), MA_DR_MP3_VMUL(vy, w1))); } +#define MA_DR_MP3_V2(k) { MA_DR_MP3_VLOAD(k) b = MA_DR_MP3_VADD(b, MA_DR_MP3_VADD(MA_DR_MP3_VMUL(vz, w1), MA_DR_MP3_VMUL(vy, w0))); a = MA_DR_MP3_VADD(a, MA_DR_MP3_VSUB(MA_DR_MP3_VMUL(vy, w1), MA_DR_MP3_VMUL(vz, w0))); } + ma_dr_mp3_f4 a, b; + zlin[4*i] = xl[18*(31 - i)]; + zlin[4*i + 1] = xr[18*(31 - i)]; + zlin[4*i + 2] = xl[1 + 18*(31 - i)]; + zlin[4*i + 3] = xr[1 + 18*(31 - i)]; + zlin[4*i + 64] = xl[1 + 18*(1 + i)]; + zlin[4*i + 64 + 1] = xr[1 + 18*(1 + i)]; + zlin[4*i - 64 + 2] = xl[18*(1 + i)]; + zlin[4*i - 64 + 3] = xr[18*(1 + i)]; + MA_DR_MP3_V0(0) MA_DR_MP3_V2(1) MA_DR_MP3_V1(2) MA_DR_MP3_V2(3) MA_DR_MP3_V1(4) MA_DR_MP3_V2(5) MA_DR_MP3_V1(6) MA_DR_MP3_V2(7) + { +#ifndef MA_DR_MP3_FLOAT_OUTPUT +#if MA_DR_MP3_HAVE_SSE + static const ma_dr_mp3_f4 g_max = { 32767.0f, 32767.0f, 32767.0f, 32767.0f }; + static const ma_dr_mp3_f4 g_min = { -32768.0f, -32768.0f, -32768.0f, -32768.0f }; + __m128i pcm8 = _mm_packs_epi32(_mm_cvtps_epi32(_mm_max_ps(_mm_min_ps(a, g_max), g_min)), + _mm_cvtps_epi32(_mm_max_ps(_mm_min_ps(b, g_max), g_min))); + dstr[(15 - i)*nch] = (ma_int16)_mm_extract_epi16(pcm8, 1); + dstr[(17 + i)*nch] = (ma_int16)_mm_extract_epi16(pcm8, 5); + dstl[(15 - i)*nch] = (ma_int16)_mm_extract_epi16(pcm8, 0); + dstl[(17 + i)*nch] = (ma_int16)_mm_extract_epi16(pcm8, 4); + dstr[(47 - i)*nch] = (ma_int16)_mm_extract_epi16(pcm8, 3); + dstr[(49 + i)*nch] = (ma_int16)_mm_extract_epi16(pcm8, 7); + dstl[(47 - i)*nch] = (ma_int16)_mm_extract_epi16(pcm8, 2); + dstl[(49 + i)*nch] = (ma_int16)_mm_extract_epi16(pcm8, 6); +#else + int16x4_t pcma, pcmb; + a = MA_DR_MP3_VADD(a, MA_DR_MP3_VSET(0.5f)); + b = MA_DR_MP3_VADD(b, MA_DR_MP3_VSET(0.5f)); + pcma = vqmovn_s32(vqaddq_s32(vcvtq_s32_f32(a), vreinterpretq_s32_u32(vcltq_f32(a, MA_DR_MP3_VSET(0))))); + pcmb = vqmovn_s32(vqaddq_s32(vcvtq_s32_f32(b), vreinterpretq_s32_u32(vcltq_f32(b, MA_DR_MP3_VSET(0))))); + vst1_lane_s16(dstr + (15 - i)*nch, pcma, 1); + vst1_lane_s16(dstr + (17 + i)*nch, pcmb, 1); + vst1_lane_s16(dstl + (15 - i)*nch, pcma, 0); + vst1_lane_s16(dstl + (17 + i)*nch, pcmb, 0); + vst1_lane_s16(dstr + (47 - i)*nch, pcma, 3); + vst1_lane_s16(dstr + (49 + i)*nch, pcmb, 3); + vst1_lane_s16(dstl + (47 - i)*nch, pcma, 2); + vst1_lane_s16(dstl + (49 + i)*nch, pcmb, 2); +#endif +#else + #if MA_DR_MP3_HAVE_SSE + static const ma_dr_mp3_f4 g_scale = { 1.0f/32768.0f, 1.0f/32768.0f, 1.0f/32768.0f, 1.0f/32768.0f }; + #else + const ma_dr_mp3_f4 g_scale = vdupq_n_f32(1.0f/32768.0f); + #endif + a = MA_DR_MP3_VMUL(a, g_scale); + b = MA_DR_MP3_VMUL(b, g_scale); +#if MA_DR_MP3_HAVE_SSE + _mm_store_ss(dstr + (15 - i)*nch, _mm_shuffle_ps(a, a, _MM_SHUFFLE(1, 1, 1, 1))); + _mm_store_ss(dstr + (17 + i)*nch, _mm_shuffle_ps(b, b, _MM_SHUFFLE(1, 1, 1, 1))); + _mm_store_ss(dstl + (15 - i)*nch, _mm_shuffle_ps(a, a, _MM_SHUFFLE(0, 0, 0, 0))); + _mm_store_ss(dstl + (17 + i)*nch, _mm_shuffle_ps(b, b, _MM_SHUFFLE(0, 0, 0, 0))); + _mm_store_ss(dstr + (47 - i)*nch, _mm_shuffle_ps(a, a, _MM_SHUFFLE(3, 3, 3, 3))); + _mm_store_ss(dstr + (49 + i)*nch, _mm_shuffle_ps(b, b, _MM_SHUFFLE(3, 3, 3, 3))); + _mm_store_ss(dstl + (47 - i)*nch, _mm_shuffle_ps(a, a, _MM_SHUFFLE(2, 2, 2, 2))); + _mm_store_ss(dstl + (49 + i)*nch, _mm_shuffle_ps(b, b, _MM_SHUFFLE(2, 2, 2, 2))); +#else + vst1q_lane_f32(dstr + (15 - i)*nch, a, 1); + vst1q_lane_f32(dstr + (17 + i)*nch, b, 1); + vst1q_lane_f32(dstl + (15 - i)*nch, a, 0); + vst1q_lane_f32(dstl + (17 + i)*nch, b, 0); + vst1q_lane_f32(dstr + (47 - i)*nch, a, 3); + vst1q_lane_f32(dstr + (49 + i)*nch, b, 3); + vst1q_lane_f32(dstl + (47 - i)*nch, a, 2); + vst1q_lane_f32(dstl + (49 + i)*nch, b, 2); +#endif +#endif + } + } else +#endif +#ifdef MA_DR_MP3_ONLY_SIMD + {} +#else + for (i = 14; i >= 0; i--) + { +#define MA_DR_MP3_LOAD(k) float w0 = *w++; float w1 = *w++; float *vz = &zlin[4*i - k*64]; float *vy = &zlin[4*i - (15 - k)*64]; +#define MA_DR_MP3_S0(k) { int j; MA_DR_MP3_LOAD(k); for (j = 0; j < 4; j++) b[j] = vz[j]*w1 + vy[j]*w0, a[j] = vz[j]*w0 - vy[j]*w1; } +#define MA_DR_MP3_S1(k) { int j; MA_DR_MP3_LOAD(k); for (j = 0; j < 4; j++) b[j] += vz[j]*w1 + vy[j]*w0, a[j] += vz[j]*w0 - vy[j]*w1; } +#define MA_DR_MP3_S2(k) { int j; MA_DR_MP3_LOAD(k); for (j = 0; j < 4; j++) b[j] += vz[j]*w1 + vy[j]*w0, a[j] += vy[j]*w1 - vz[j]*w0; } + float a[4], b[4]; + zlin[4*i] = xl[18*(31 - i)]; + zlin[4*i + 1] = xr[18*(31 - i)]; + zlin[4*i + 2] = xl[1 + 18*(31 - i)]; + zlin[4*i + 3] = xr[1 + 18*(31 - i)]; + zlin[4*(i + 16)] = xl[1 + 18*(1 + i)]; + zlin[4*(i + 16) + 1] = xr[1 + 18*(1 + i)]; + zlin[4*(i - 16) + 2] = xl[18*(1 + i)]; + zlin[4*(i - 16) + 3] = xr[18*(1 + i)]; + MA_DR_MP3_S0(0) MA_DR_MP3_S2(1) MA_DR_MP3_S1(2) MA_DR_MP3_S2(3) MA_DR_MP3_S1(4) MA_DR_MP3_S2(5) MA_DR_MP3_S1(6) MA_DR_MP3_S2(7) + dstr[(15 - i)*nch] = ma_dr_mp3d_scale_pcm(a[1]); + dstr[(17 + i)*nch] = ma_dr_mp3d_scale_pcm(b[1]); + dstl[(15 - i)*nch] = ma_dr_mp3d_scale_pcm(a[0]); + dstl[(17 + i)*nch] = ma_dr_mp3d_scale_pcm(b[0]); + dstr[(47 - i)*nch] = ma_dr_mp3d_scale_pcm(a[3]); + dstr[(49 + i)*nch] = ma_dr_mp3d_scale_pcm(b[3]); + dstl[(47 - i)*nch] = ma_dr_mp3d_scale_pcm(a[2]); + dstl[(49 + i)*nch] = ma_dr_mp3d_scale_pcm(b[2]); + } +#endif +} +static void ma_dr_mp3d_synth_granule(float *qmf_state, float *grbuf, int nbands, int nch, ma_dr_mp3d_sample_t *pcm, float *lins) +{ + int i; + for (i = 0; i < nch; i++) + { + ma_dr_mp3d_DCT_II(grbuf + 576*i, nbands); + } + MA_DR_MP3_COPY_MEMORY(lins, qmf_state, sizeof(float)*15*64); + for (i = 0; i < nbands; i += 2) + { + ma_dr_mp3d_synth(grbuf + i, pcm + 32*nch*i, nch, lins + i*64); + } +#ifndef MA_DR_MP3_NONSTANDARD_BUT_LOGICAL + if (nch == 1) + { + for (i = 0; i < 15*64; i += 2) + { + qmf_state[i] = lins[nbands*64 + i]; + } + } else +#endif + { + MA_DR_MP3_COPY_MEMORY(qmf_state, lins + nbands*64, sizeof(float)*15*64); + } +} +static int ma_dr_mp3d_match_frame(const ma_uint8 *hdr, int mp3_bytes, int frame_bytes) +{ + int i, nmatch; + for (i = 0, nmatch = 0; nmatch < MA_DR_MP3_MAX_FRAME_SYNC_MATCHES; nmatch++) + { + i += ma_dr_mp3_hdr_frame_bytes(hdr + i, frame_bytes) + ma_dr_mp3_hdr_padding(hdr + i); + if (i + MA_DR_MP3_HDR_SIZE > mp3_bytes) + return nmatch > 0; + if (!ma_dr_mp3_hdr_compare(hdr, hdr + i)) + return 0; + } + return 1; +} +static int ma_dr_mp3d_find_frame(const ma_uint8 *mp3, int mp3_bytes, int *free_format_bytes, int *ptr_frame_bytes) +{ + int i, k; + for (i = 0; i < mp3_bytes - MA_DR_MP3_HDR_SIZE; i++, mp3++) + { + if (ma_dr_mp3_hdr_valid(mp3)) + { + int frame_bytes = ma_dr_mp3_hdr_frame_bytes(mp3, *free_format_bytes); + int frame_and_padding = frame_bytes + ma_dr_mp3_hdr_padding(mp3); + for (k = MA_DR_MP3_HDR_SIZE; !frame_bytes && k < MA_DR_MP3_MAX_FREE_FORMAT_FRAME_SIZE && i + 2*k < mp3_bytes - MA_DR_MP3_HDR_SIZE; k++) + { + if (ma_dr_mp3_hdr_compare(mp3, mp3 + k)) + { + int fb = k - ma_dr_mp3_hdr_padding(mp3); + int nextfb = fb + ma_dr_mp3_hdr_padding(mp3 + k); + if (i + k + nextfb + MA_DR_MP3_HDR_SIZE > mp3_bytes || !ma_dr_mp3_hdr_compare(mp3, mp3 + k + nextfb)) + continue; + frame_and_padding = k; + frame_bytes = fb; + *free_format_bytes = fb; + } + } + if ((frame_bytes && i + frame_and_padding <= mp3_bytes && + ma_dr_mp3d_match_frame(mp3, mp3_bytes - i, frame_bytes)) || + (!i && frame_and_padding == mp3_bytes)) + { + *ptr_frame_bytes = frame_and_padding; + return i; + } + *free_format_bytes = 0; + } + } + *ptr_frame_bytes = 0; + return mp3_bytes; +} +MA_API void ma_dr_mp3dec_init(ma_dr_mp3dec *dec) +{ + dec->header[0] = 0; +} +MA_API int ma_dr_mp3dec_decode_frame(ma_dr_mp3dec *dec, const ma_uint8 *mp3, int mp3_bytes, void *pcm, ma_dr_mp3dec_frame_info *info) +{ + int i = 0, igr, frame_size = 0, success = 1; + const ma_uint8 *hdr; + ma_dr_mp3_bs bs_frame[1]; + ma_dr_mp3dec_scratch scratch; + if (mp3_bytes > 4 && dec->header[0] == 0xff && ma_dr_mp3_hdr_compare(dec->header, mp3)) + { + frame_size = ma_dr_mp3_hdr_frame_bytes(mp3, dec->free_format_bytes) + ma_dr_mp3_hdr_padding(mp3); + if (frame_size != mp3_bytes && (frame_size + MA_DR_MP3_HDR_SIZE > mp3_bytes || !ma_dr_mp3_hdr_compare(mp3, mp3 + frame_size))) + { + frame_size = 0; + } + } + if (!frame_size) + { + MA_DR_MP3_ZERO_MEMORY(dec, sizeof(ma_dr_mp3dec)); + i = ma_dr_mp3d_find_frame(mp3, mp3_bytes, &dec->free_format_bytes, &frame_size); + if (!frame_size || i + frame_size > mp3_bytes) + { + info->frame_bytes = i; + return 0; + } + } + hdr = mp3 + i; + MA_DR_MP3_COPY_MEMORY(dec->header, hdr, MA_DR_MP3_HDR_SIZE); + info->frame_bytes = i + frame_size; + info->channels = MA_DR_MP3_HDR_IS_MONO(hdr) ? 1 : 2; + info->hz = ma_dr_mp3_hdr_sample_rate_hz(hdr); + info->layer = 4 - MA_DR_MP3_HDR_GET_LAYER(hdr); + info->bitrate_kbps = ma_dr_mp3_hdr_bitrate_kbps(hdr); + ma_dr_mp3_bs_init(bs_frame, hdr + MA_DR_MP3_HDR_SIZE, frame_size - MA_DR_MP3_HDR_SIZE); + if (MA_DR_MP3_HDR_IS_CRC(hdr)) + { + ma_dr_mp3_bs_get_bits(bs_frame, 16); + } + if (info->layer == 3) + { + int main_data_begin = ma_dr_mp3_L3_read_side_info(bs_frame, scratch.gr_info, hdr); + if (main_data_begin < 0 || bs_frame->pos > bs_frame->limit) + { + ma_dr_mp3dec_init(dec); + return 0; + } + success = ma_dr_mp3_L3_restore_reservoir(dec, bs_frame, &scratch, main_data_begin); + if (success && pcm != NULL) + { + for (igr = 0; igr < (MA_DR_MP3_HDR_TEST_MPEG1(hdr) ? 2 : 1); igr++, pcm = MA_DR_MP3_OFFSET_PTR(pcm, sizeof(ma_dr_mp3d_sample_t)*576*info->channels)) + { + MA_DR_MP3_ZERO_MEMORY(scratch.grbuf[0], 576*2*sizeof(float)); + ma_dr_mp3_L3_decode(dec, &scratch, scratch.gr_info + igr*info->channels, info->channels); + ma_dr_mp3d_synth_granule(dec->qmf_state, scratch.grbuf[0], 18, info->channels, (ma_dr_mp3d_sample_t*)pcm, scratch.syn[0]); + } + } + ma_dr_mp3_L3_save_reservoir(dec, &scratch); + } else + { +#ifdef MA_DR_MP3_ONLY_MP3 + return 0; +#else + ma_dr_mp3_L12_scale_info sci[1]; + if (pcm == NULL) { + return ma_dr_mp3_hdr_frame_samples(hdr); + } + ma_dr_mp3_L12_read_scale_info(hdr, bs_frame, sci); + MA_DR_MP3_ZERO_MEMORY(scratch.grbuf[0], 576*2*sizeof(float)); + for (i = 0, igr = 0; igr < 3; igr++) + { + if (12 == (i += ma_dr_mp3_L12_dequantize_granule(scratch.grbuf[0] + i, bs_frame, sci, info->layer | 1))) + { + i = 0; + ma_dr_mp3_L12_apply_scf_384(sci, sci->scf + igr, scratch.grbuf[0]); + ma_dr_mp3d_synth_granule(dec->qmf_state, scratch.grbuf[0], 12, info->channels, (ma_dr_mp3d_sample_t*)pcm, scratch.syn[0]); + MA_DR_MP3_ZERO_MEMORY(scratch.grbuf[0], 576*2*sizeof(float)); + pcm = MA_DR_MP3_OFFSET_PTR(pcm, sizeof(ma_dr_mp3d_sample_t)*384*info->channels); + } + if (bs_frame->pos > bs_frame->limit) + { + ma_dr_mp3dec_init(dec); + return 0; + } + } +#endif + } + return success*ma_dr_mp3_hdr_frame_samples(dec->header); +} +MA_API void ma_dr_mp3dec_f32_to_s16(const float *in, ma_int16 *out, size_t num_samples) +{ + size_t i = 0; +#if MA_DR_MP3_HAVE_SIMD + size_t aligned_count = num_samples & ~7; + for(; i < aligned_count; i+=8) + { + ma_dr_mp3_f4 scale = MA_DR_MP3_VSET(32768.0f); + ma_dr_mp3_f4 a = MA_DR_MP3_VMUL(MA_DR_MP3_VLD(&in[i ]), scale); + ma_dr_mp3_f4 b = MA_DR_MP3_VMUL(MA_DR_MP3_VLD(&in[i+4]), scale); +#if MA_DR_MP3_HAVE_SSE + ma_dr_mp3_f4 s16max = MA_DR_MP3_VSET( 32767.0f); + ma_dr_mp3_f4 s16min = MA_DR_MP3_VSET(-32768.0f); + __m128i pcm8 = _mm_packs_epi32(_mm_cvtps_epi32(_mm_max_ps(_mm_min_ps(a, s16max), s16min)), + _mm_cvtps_epi32(_mm_max_ps(_mm_min_ps(b, s16max), s16min))); + out[i ] = (ma_int16)_mm_extract_epi16(pcm8, 0); + out[i+1] = (ma_int16)_mm_extract_epi16(pcm8, 1); + out[i+2] = (ma_int16)_mm_extract_epi16(pcm8, 2); + out[i+3] = (ma_int16)_mm_extract_epi16(pcm8, 3); + out[i+4] = (ma_int16)_mm_extract_epi16(pcm8, 4); + out[i+5] = (ma_int16)_mm_extract_epi16(pcm8, 5); + out[i+6] = (ma_int16)_mm_extract_epi16(pcm8, 6); + out[i+7] = (ma_int16)_mm_extract_epi16(pcm8, 7); +#else + int16x4_t pcma, pcmb; + a = MA_DR_MP3_VADD(a, MA_DR_MP3_VSET(0.5f)); + b = MA_DR_MP3_VADD(b, MA_DR_MP3_VSET(0.5f)); + pcma = vqmovn_s32(vqaddq_s32(vcvtq_s32_f32(a), vreinterpretq_s32_u32(vcltq_f32(a, MA_DR_MP3_VSET(0))))); + pcmb = vqmovn_s32(vqaddq_s32(vcvtq_s32_f32(b), vreinterpretq_s32_u32(vcltq_f32(b, MA_DR_MP3_VSET(0))))); + vst1_lane_s16(out+i , pcma, 0); + vst1_lane_s16(out+i+1, pcma, 1); + vst1_lane_s16(out+i+2, pcma, 2); + vst1_lane_s16(out+i+3, pcma, 3); + vst1_lane_s16(out+i+4, pcmb, 0); + vst1_lane_s16(out+i+5, pcmb, 1); + vst1_lane_s16(out+i+6, pcmb, 2); + vst1_lane_s16(out+i+7, pcmb, 3); +#endif + } +#endif + for(; i < num_samples; i++) + { + float sample = in[i] * 32768.0f; + if (sample >= 32766.5f) + out[i] = (ma_int16) 32767; + else if (sample <= -32767.5f) + out[i] = (ma_int16)-32768; + else + { + short s = (ma_int16)(sample + .5f); + s -= (s < 0); + out[i] = s; + } + } +} +#ifndef MA_DR_MP3_SEEK_LEADING_MP3_FRAMES +#define MA_DR_MP3_SEEK_LEADING_MP3_FRAMES 2 +#endif +#define MA_DR_MP3_MIN_DATA_CHUNK_SIZE 16384 +#ifndef MA_DR_MP3_DATA_CHUNK_SIZE +#define MA_DR_MP3_DATA_CHUNK_SIZE (MA_DR_MP3_MIN_DATA_CHUNK_SIZE*4) +#endif +#define MA_DR_MP3_COUNTOF(x) (sizeof(x) / sizeof(x[0])) +#define MA_DR_MP3_CLAMP(x, lo, hi) (MA_DR_MP3_MAX(lo, MA_DR_MP3_MIN(x, hi))) +#ifndef MA_DR_MP3_PI_D +#define MA_DR_MP3_PI_D 3.14159265358979323846264 +#endif +#define MA_DR_MP3_DEFAULT_RESAMPLER_LPF_ORDER 2 +static MA_INLINE float ma_dr_mp3_mix_f32(float x, float y, float a) +{ + return x*(1-a) + y*a; +} +static MA_INLINE float ma_dr_mp3_mix_f32_fast(float x, float y, float a) +{ + float r0 = (y - x); + float r1 = r0*a; + return x + r1; +} +static MA_INLINE ma_uint32 ma_dr_mp3_gcf_u32(ma_uint32 a, ma_uint32 b) +{ + for (;;) { + if (b == 0) { + break; + } else { + ma_uint32 t = a; + a = b; + b = t % a; + } + } + return a; +} +static void* ma_dr_mp3__malloc_default(size_t sz, void* pUserData) +{ + (void)pUserData; + return MA_DR_MP3_MALLOC(sz); +} +static void* ma_dr_mp3__realloc_default(void* p, size_t sz, void* pUserData) +{ + (void)pUserData; + return MA_DR_MP3_REALLOC(p, sz); +} +static void ma_dr_mp3__free_default(void* p, void* pUserData) +{ + (void)pUserData; + MA_DR_MP3_FREE(p); +} +static void* ma_dr_mp3__malloc_from_callbacks(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks == NULL) { + return NULL; + } + if (pAllocationCallbacks->onMalloc != NULL) { + return pAllocationCallbacks->onMalloc(sz, pAllocationCallbacks->pUserData); + } + if (pAllocationCallbacks->onRealloc != NULL) { + return pAllocationCallbacks->onRealloc(NULL, sz, pAllocationCallbacks->pUserData); + } + return NULL; +} +static void* ma_dr_mp3__realloc_from_callbacks(void* p, size_t szNew, size_t szOld, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks == NULL) { + return NULL; + } + if (pAllocationCallbacks->onRealloc != NULL) { + return pAllocationCallbacks->onRealloc(p, szNew, pAllocationCallbacks->pUserData); + } + if (pAllocationCallbacks->onMalloc != NULL && pAllocationCallbacks->onFree != NULL) { + void* p2; + p2 = pAllocationCallbacks->onMalloc(szNew, pAllocationCallbacks->pUserData); + if (p2 == NULL) { + return NULL; + } + if (p != NULL) { + MA_DR_MP3_COPY_MEMORY(p2, p, szOld); + pAllocationCallbacks->onFree(p, pAllocationCallbacks->pUserData); + } + return p2; + } + return NULL; +} +static void ma_dr_mp3__free_from_callbacks(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (p == NULL || pAllocationCallbacks == NULL) { + return; + } + if (pAllocationCallbacks->onFree != NULL) { + pAllocationCallbacks->onFree(p, pAllocationCallbacks->pUserData); + } +} +static ma_allocation_callbacks ma_dr_mp3_copy_allocation_callbacks_or_defaults(const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks != NULL) { + return *pAllocationCallbacks; + } else { + ma_allocation_callbacks allocationCallbacks; + allocationCallbacks.pUserData = NULL; + allocationCallbacks.onMalloc = ma_dr_mp3__malloc_default; + allocationCallbacks.onRealloc = ma_dr_mp3__realloc_default; + allocationCallbacks.onFree = ma_dr_mp3__free_default; + return allocationCallbacks; + } +} +static size_t ma_dr_mp3__on_read(ma_dr_mp3* pMP3, void* pBufferOut, size_t bytesToRead) +{ + size_t bytesRead = pMP3->onRead(pMP3->pUserData, pBufferOut, bytesToRead); + pMP3->streamCursor += bytesRead; + return bytesRead; +} +static ma_bool32 ma_dr_mp3__on_seek(ma_dr_mp3* pMP3, int offset, ma_dr_mp3_seek_origin origin) +{ + MA_DR_MP3_ASSERT(offset >= 0); + if (!pMP3->onSeek(pMP3->pUserData, offset, origin)) { + return MA_FALSE; + } + if (origin == ma_dr_mp3_seek_origin_start) { + pMP3->streamCursor = (ma_uint64)offset; + } else { + pMP3->streamCursor += offset; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_mp3__on_seek_64(ma_dr_mp3* pMP3, ma_uint64 offset, ma_dr_mp3_seek_origin origin) +{ + if (offset <= 0x7FFFFFFF) { + return ma_dr_mp3__on_seek(pMP3, (int)offset, origin); + } + if (!ma_dr_mp3__on_seek(pMP3, 0x7FFFFFFF, ma_dr_mp3_seek_origin_start)) { + return MA_FALSE; + } + offset -= 0x7FFFFFFF; + while (offset > 0) { + if (offset <= 0x7FFFFFFF) { + if (!ma_dr_mp3__on_seek(pMP3, (int)offset, ma_dr_mp3_seek_origin_current)) { + return MA_FALSE; + } + offset = 0; + } else { + if (!ma_dr_mp3__on_seek(pMP3, 0x7FFFFFFF, ma_dr_mp3_seek_origin_current)) { + return MA_FALSE; + } + offset -= 0x7FFFFFFF; + } + } + return MA_TRUE; +} +static ma_uint32 ma_dr_mp3_decode_next_frame_ex__callbacks(ma_dr_mp3* pMP3, ma_dr_mp3d_sample_t* pPCMFrames) +{ + ma_uint32 pcmFramesRead = 0; + MA_DR_MP3_ASSERT(pMP3 != NULL); + MA_DR_MP3_ASSERT(pMP3->onRead != NULL); + if (pMP3->atEnd) { + return 0; + } + for (;;) { + ma_dr_mp3dec_frame_info info; + if (pMP3->dataSize < MA_DR_MP3_MIN_DATA_CHUNK_SIZE) { + size_t bytesRead; + if (pMP3->pData != NULL) { + MA_DR_MP3_MOVE_MEMORY(pMP3->pData, pMP3->pData + pMP3->dataConsumed, pMP3->dataSize); + } + pMP3->dataConsumed = 0; + if (pMP3->dataCapacity < MA_DR_MP3_DATA_CHUNK_SIZE) { + ma_uint8* pNewData; + size_t newDataCap; + newDataCap = MA_DR_MP3_DATA_CHUNK_SIZE; + pNewData = (ma_uint8*)ma_dr_mp3__realloc_from_callbacks(pMP3->pData, newDataCap, pMP3->dataCapacity, &pMP3->allocationCallbacks); + if (pNewData == NULL) { + return 0; + } + pMP3->pData = pNewData; + pMP3->dataCapacity = newDataCap; + } + bytesRead = ma_dr_mp3__on_read(pMP3, pMP3->pData + pMP3->dataSize, (pMP3->dataCapacity - pMP3->dataSize)); + if (bytesRead == 0) { + if (pMP3->dataSize == 0) { + pMP3->atEnd = MA_TRUE; + return 0; + } + } + pMP3->dataSize += bytesRead; + } + if (pMP3->dataSize > INT_MAX) { + pMP3->atEnd = MA_TRUE; + return 0; + } + MA_DR_MP3_ASSERT(pMP3->pData != NULL); + MA_DR_MP3_ASSERT(pMP3->dataCapacity > 0); + if (pMP3->pData == NULL) { + return 0; + } + pcmFramesRead = ma_dr_mp3dec_decode_frame(&pMP3->decoder, pMP3->pData + pMP3->dataConsumed, (int)pMP3->dataSize, pPCMFrames, &info); + if (info.frame_bytes > 0) { + pMP3->dataConsumed += (size_t)info.frame_bytes; + pMP3->dataSize -= (size_t)info.frame_bytes; + } + if (pcmFramesRead > 0) { + pcmFramesRead = ma_dr_mp3_hdr_frame_samples(pMP3->decoder.header); + pMP3->pcmFramesConsumedInMP3Frame = 0; + pMP3->pcmFramesRemainingInMP3Frame = pcmFramesRead; + pMP3->mp3FrameChannels = info.channels; + pMP3->mp3FrameSampleRate = info.hz; + break; + } else if (info.frame_bytes == 0) { + size_t bytesRead; + MA_DR_MP3_MOVE_MEMORY(pMP3->pData, pMP3->pData + pMP3->dataConsumed, pMP3->dataSize); + pMP3->dataConsumed = 0; + if (pMP3->dataCapacity == pMP3->dataSize) { + ma_uint8* pNewData; + size_t newDataCap; + newDataCap = pMP3->dataCapacity + MA_DR_MP3_DATA_CHUNK_SIZE; + pNewData = (ma_uint8*)ma_dr_mp3__realloc_from_callbacks(pMP3->pData, newDataCap, pMP3->dataCapacity, &pMP3->allocationCallbacks); + if (pNewData == NULL) { + return 0; + } + pMP3->pData = pNewData; + pMP3->dataCapacity = newDataCap; + } + bytesRead = ma_dr_mp3__on_read(pMP3, pMP3->pData + pMP3->dataSize, (pMP3->dataCapacity - pMP3->dataSize)); + if (bytesRead == 0) { + pMP3->atEnd = MA_TRUE; + return 0; + } + pMP3->dataSize += bytesRead; + } + }; + return pcmFramesRead; +} +static ma_uint32 ma_dr_mp3_decode_next_frame_ex__memory(ma_dr_mp3* pMP3, ma_dr_mp3d_sample_t* pPCMFrames) +{ + ma_uint32 pcmFramesRead = 0; + ma_dr_mp3dec_frame_info info; + MA_DR_MP3_ASSERT(pMP3 != NULL); + MA_DR_MP3_ASSERT(pMP3->memory.pData != NULL); + if (pMP3->atEnd) { + return 0; + } + for (;;) { + pcmFramesRead = ma_dr_mp3dec_decode_frame(&pMP3->decoder, pMP3->memory.pData + pMP3->memory.currentReadPos, (int)(pMP3->memory.dataSize - pMP3->memory.currentReadPos), pPCMFrames, &info); + if (pcmFramesRead > 0) { + pcmFramesRead = ma_dr_mp3_hdr_frame_samples(pMP3->decoder.header); + pMP3->pcmFramesConsumedInMP3Frame = 0; + pMP3->pcmFramesRemainingInMP3Frame = pcmFramesRead; + pMP3->mp3FrameChannels = info.channels; + pMP3->mp3FrameSampleRate = info.hz; + break; + } else if (info.frame_bytes > 0) { + pMP3->memory.currentReadPos += (size_t)info.frame_bytes; + } else { + break; + } + } + pMP3->memory.currentReadPos += (size_t)info.frame_bytes; + return pcmFramesRead; +} +static ma_uint32 ma_dr_mp3_decode_next_frame_ex(ma_dr_mp3* pMP3, ma_dr_mp3d_sample_t* pPCMFrames) +{ + if (pMP3->memory.pData != NULL && pMP3->memory.dataSize > 0) { + return ma_dr_mp3_decode_next_frame_ex__memory(pMP3, pPCMFrames); + } else { + return ma_dr_mp3_decode_next_frame_ex__callbacks(pMP3, pPCMFrames); + } +} +static ma_uint32 ma_dr_mp3_decode_next_frame(ma_dr_mp3* pMP3) +{ + MA_DR_MP3_ASSERT(pMP3 != NULL); + return ma_dr_mp3_decode_next_frame_ex(pMP3, (ma_dr_mp3d_sample_t*)pMP3->pcmFrames); +} +#if 0 +static ma_uint32 ma_dr_mp3_seek_next_frame(ma_dr_mp3* pMP3) +{ + ma_uint32 pcmFrameCount; + MA_DR_MP3_ASSERT(pMP3 != NULL); + pcmFrameCount = ma_dr_mp3_decode_next_frame_ex(pMP3, NULL); + if (pcmFrameCount == 0) { + return 0; + } + pMP3->currentPCMFrame += pcmFrameCount; + pMP3->pcmFramesConsumedInMP3Frame = pcmFrameCount; + pMP3->pcmFramesRemainingInMP3Frame = 0; + return pcmFrameCount; +} +#endif +static ma_bool32 ma_dr_mp3_init_internal(ma_dr_mp3* pMP3, ma_dr_mp3_read_proc onRead, ma_dr_mp3_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + MA_DR_MP3_ASSERT(pMP3 != NULL); + MA_DR_MP3_ASSERT(onRead != NULL); + ma_dr_mp3dec_init(&pMP3->decoder); + pMP3->onRead = onRead; + pMP3->onSeek = onSeek; + pMP3->pUserData = pUserData; + pMP3->allocationCallbacks = ma_dr_mp3_copy_allocation_callbacks_or_defaults(pAllocationCallbacks); + if (pMP3->allocationCallbacks.onFree == NULL || (pMP3->allocationCallbacks.onMalloc == NULL && pMP3->allocationCallbacks.onRealloc == NULL)) { + return MA_FALSE; + } + if (ma_dr_mp3_decode_next_frame(pMP3) == 0) { + ma_dr_mp3__free_from_callbacks(pMP3->pData, &pMP3->allocationCallbacks); + return MA_FALSE; + } + pMP3->channels = pMP3->mp3FrameChannels; + pMP3->sampleRate = pMP3->mp3FrameSampleRate; + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_mp3_init(ma_dr_mp3* pMP3, ma_dr_mp3_read_proc onRead, ma_dr_mp3_seek_proc onSeek, void* pUserData, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pMP3 == NULL || onRead == NULL) { + return MA_FALSE; + } + MA_DR_MP3_ZERO_OBJECT(pMP3); + return ma_dr_mp3_init_internal(pMP3, onRead, onSeek, pUserData, pAllocationCallbacks); +} +static size_t ma_dr_mp3__on_read_memory(void* pUserData, void* pBufferOut, size_t bytesToRead) +{ + ma_dr_mp3* pMP3 = (ma_dr_mp3*)pUserData; + size_t bytesRemaining; + MA_DR_MP3_ASSERT(pMP3 != NULL); + MA_DR_MP3_ASSERT(pMP3->memory.dataSize >= pMP3->memory.currentReadPos); + bytesRemaining = pMP3->memory.dataSize - pMP3->memory.currentReadPos; + if (bytesToRead > bytesRemaining) { + bytesToRead = bytesRemaining; + } + if (bytesToRead > 0) { + MA_DR_MP3_COPY_MEMORY(pBufferOut, pMP3->memory.pData + pMP3->memory.currentReadPos, bytesToRead); + pMP3->memory.currentReadPos += bytesToRead; + } + return bytesToRead; +} +static ma_bool32 ma_dr_mp3__on_seek_memory(void* pUserData, int byteOffset, ma_dr_mp3_seek_origin origin) +{ + ma_dr_mp3* pMP3 = (ma_dr_mp3*)pUserData; + MA_DR_MP3_ASSERT(pMP3 != NULL); + if (origin == ma_dr_mp3_seek_origin_current) { + if (byteOffset > 0) { + if (pMP3->memory.currentReadPos + byteOffset > pMP3->memory.dataSize) { + byteOffset = (int)(pMP3->memory.dataSize - pMP3->memory.currentReadPos); + } + } else { + if (pMP3->memory.currentReadPos < (size_t)-byteOffset) { + byteOffset = -(int)pMP3->memory.currentReadPos; + } + } + pMP3->memory.currentReadPos += byteOffset; + } else { + if ((ma_uint32)byteOffset <= pMP3->memory.dataSize) { + pMP3->memory.currentReadPos = byteOffset; + } else { + pMP3->memory.currentReadPos = pMP3->memory.dataSize; + } + } + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_mp3_init_memory(ma_dr_mp3* pMP3, const void* pData, size_t dataSize, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pMP3 == NULL) { + return MA_FALSE; + } + MA_DR_MP3_ZERO_OBJECT(pMP3); + if (pData == NULL || dataSize == 0) { + return MA_FALSE; + } + pMP3->memory.pData = (const ma_uint8*)pData; + pMP3->memory.dataSize = dataSize; + pMP3->memory.currentReadPos = 0; + return ma_dr_mp3_init_internal(pMP3, ma_dr_mp3__on_read_memory, ma_dr_mp3__on_seek_memory, pMP3, pAllocationCallbacks); +} +#ifndef MA_DR_MP3_NO_STDIO +#include +#include +static size_t ma_dr_mp3__on_read_stdio(void* pUserData, void* pBufferOut, size_t bytesToRead) +{ + return fread(pBufferOut, 1, bytesToRead, (FILE*)pUserData); +} +static ma_bool32 ma_dr_mp3__on_seek_stdio(void* pUserData, int offset, ma_dr_mp3_seek_origin origin) +{ + return fseek((FILE*)pUserData, offset, (origin == ma_dr_mp3_seek_origin_current) ? SEEK_CUR : SEEK_SET) == 0; +} +MA_API ma_bool32 ma_dr_mp3_init_file(ma_dr_mp3* pMP3, const char* pFilePath, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_bool32 result; + FILE* pFile; + if (ma_fopen(&pFile, pFilePath, "rb") != MA_SUCCESS) { + return MA_FALSE; + } + result = ma_dr_mp3_init(pMP3, ma_dr_mp3__on_read_stdio, ma_dr_mp3__on_seek_stdio, (void*)pFile, pAllocationCallbacks); + if (result != MA_TRUE) { + fclose(pFile); + return result; + } + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_mp3_init_file_w(ma_dr_mp3* pMP3, const wchar_t* pFilePath, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_bool32 result; + FILE* pFile; + if (ma_wfopen(&pFile, pFilePath, L"rb", pAllocationCallbacks) != MA_SUCCESS) { + return MA_FALSE; + } + result = ma_dr_mp3_init(pMP3, ma_dr_mp3__on_read_stdio, ma_dr_mp3__on_seek_stdio, (void*)pFile, pAllocationCallbacks); + if (result != MA_TRUE) { + fclose(pFile); + return result; + } + return MA_TRUE; +} +#endif +MA_API void ma_dr_mp3_uninit(ma_dr_mp3* pMP3) +{ + if (pMP3 == NULL) { + return; + } +#ifndef MA_DR_MP3_NO_STDIO + if (pMP3->onRead == ma_dr_mp3__on_read_stdio) { + FILE* pFile = (FILE*)pMP3->pUserData; + if (pFile != NULL) { + fclose(pFile); + pMP3->pUserData = NULL; + } + } +#endif + ma_dr_mp3__free_from_callbacks(pMP3->pData, &pMP3->allocationCallbacks); +} +#if defined(MA_DR_MP3_FLOAT_OUTPUT) +static void ma_dr_mp3_f32_to_s16(ma_int16* dst, const float* src, ma_uint64 sampleCount) +{ + ma_uint64 i; + ma_uint64 i4; + ma_uint64 sampleCount4; + i = 0; + sampleCount4 = sampleCount >> 2; + for (i4 = 0; i4 < sampleCount4; i4 += 1) { + float x0 = src[i+0]; + float x1 = src[i+1]; + float x2 = src[i+2]; + float x3 = src[i+3]; + x0 = ((x0 < -1) ? -1 : ((x0 > 1) ? 1 : x0)); + x1 = ((x1 < -1) ? -1 : ((x1 > 1) ? 1 : x1)); + x2 = ((x2 < -1) ? -1 : ((x2 > 1) ? 1 : x2)); + x3 = ((x3 < -1) ? -1 : ((x3 > 1) ? 1 : x3)); + x0 = x0 * 32767.0f; + x1 = x1 * 32767.0f; + x2 = x2 * 32767.0f; + x3 = x3 * 32767.0f; + dst[i+0] = (ma_int16)x0; + dst[i+1] = (ma_int16)x1; + dst[i+2] = (ma_int16)x2; + dst[i+3] = (ma_int16)x3; + i += 4; + } + for (; i < sampleCount; i += 1) { + float x = src[i]; + x = ((x < -1) ? -1 : ((x > 1) ? 1 : x)); + x = x * 32767.0f; + dst[i] = (ma_int16)x; + } +} +#endif +#if !defined(MA_DR_MP3_FLOAT_OUTPUT) +static void ma_dr_mp3_s16_to_f32(float* dst, const ma_int16* src, ma_uint64 sampleCount) +{ + ma_uint64 i; + for (i = 0; i < sampleCount; i += 1) { + float x = (float)src[i]; + x = x * 0.000030517578125f; + dst[i] = x; + } +} +#endif +static ma_uint64 ma_dr_mp3_read_pcm_frames_raw(ma_dr_mp3* pMP3, ma_uint64 framesToRead, void* pBufferOut) +{ + ma_uint64 totalFramesRead = 0; + MA_DR_MP3_ASSERT(pMP3 != NULL); + MA_DR_MP3_ASSERT(pMP3->onRead != NULL); + while (framesToRead > 0) { + ma_uint32 framesToConsume = (ma_uint32)MA_DR_MP3_MIN(pMP3->pcmFramesRemainingInMP3Frame, framesToRead); + if (pBufferOut != NULL) { + #if defined(MA_DR_MP3_FLOAT_OUTPUT) + float* pFramesOutF32 = (float*)MA_DR_MP3_OFFSET_PTR(pBufferOut, sizeof(float) * totalFramesRead * pMP3->channels); + float* pFramesInF32 = (float*)MA_DR_MP3_OFFSET_PTR(&pMP3->pcmFrames[0], sizeof(float) * pMP3->pcmFramesConsumedInMP3Frame * pMP3->mp3FrameChannels); + MA_DR_MP3_COPY_MEMORY(pFramesOutF32, pFramesInF32, sizeof(float) * framesToConsume * pMP3->channels); + #else + ma_int16* pFramesOutS16 = (ma_int16*)MA_DR_MP3_OFFSET_PTR(pBufferOut, sizeof(ma_int16) * totalFramesRead * pMP3->channels); + ma_int16* pFramesInS16 = (ma_int16*)MA_DR_MP3_OFFSET_PTR(&pMP3->pcmFrames[0], sizeof(ma_int16) * pMP3->pcmFramesConsumedInMP3Frame * pMP3->mp3FrameChannels); + MA_DR_MP3_COPY_MEMORY(pFramesOutS16, pFramesInS16, sizeof(ma_int16) * framesToConsume * pMP3->channels); + #endif + } + pMP3->currentPCMFrame += framesToConsume; + pMP3->pcmFramesConsumedInMP3Frame += framesToConsume; + pMP3->pcmFramesRemainingInMP3Frame -= framesToConsume; + totalFramesRead += framesToConsume; + framesToRead -= framesToConsume; + if (framesToRead == 0) { + break; + } + MA_DR_MP3_ASSERT(pMP3->pcmFramesRemainingInMP3Frame == 0); + if (ma_dr_mp3_decode_next_frame(pMP3) == 0) { + break; + } + } + return totalFramesRead; +} +MA_API ma_uint64 ma_dr_mp3_read_pcm_frames_f32(ma_dr_mp3* pMP3, ma_uint64 framesToRead, float* pBufferOut) +{ + if (pMP3 == NULL || pMP3->onRead == NULL) { + return 0; + } +#if defined(MA_DR_MP3_FLOAT_OUTPUT) + return ma_dr_mp3_read_pcm_frames_raw(pMP3, framesToRead, pBufferOut); +#else + { + ma_int16 pTempS16[8192]; + ma_uint64 totalPCMFramesRead = 0; + while (totalPCMFramesRead < framesToRead) { + ma_uint64 framesJustRead; + ma_uint64 framesRemaining = framesToRead - totalPCMFramesRead; + ma_uint64 framesToReadNow = MA_DR_MP3_COUNTOF(pTempS16) / pMP3->channels; + if (framesToReadNow > framesRemaining) { + framesToReadNow = framesRemaining; + } + framesJustRead = ma_dr_mp3_read_pcm_frames_raw(pMP3, framesToReadNow, pTempS16); + if (framesJustRead == 0) { + break; + } + ma_dr_mp3_s16_to_f32((float*)MA_DR_MP3_OFFSET_PTR(pBufferOut, sizeof(float) * totalPCMFramesRead * pMP3->channels), pTempS16, framesJustRead * pMP3->channels); + totalPCMFramesRead += framesJustRead; + } + return totalPCMFramesRead; + } +#endif +} +MA_API ma_uint64 ma_dr_mp3_read_pcm_frames_s16(ma_dr_mp3* pMP3, ma_uint64 framesToRead, ma_int16* pBufferOut) +{ + if (pMP3 == NULL || pMP3->onRead == NULL) { + return 0; + } +#if !defined(MA_DR_MP3_FLOAT_OUTPUT) + return ma_dr_mp3_read_pcm_frames_raw(pMP3, framesToRead, pBufferOut); +#else + { + float pTempF32[4096]; + ma_uint64 totalPCMFramesRead = 0; + while (totalPCMFramesRead < framesToRead) { + ma_uint64 framesJustRead; + ma_uint64 framesRemaining = framesToRead - totalPCMFramesRead; + ma_uint64 framesToReadNow = MA_DR_MP3_COUNTOF(pTempF32) / pMP3->channels; + if (framesToReadNow > framesRemaining) { + framesToReadNow = framesRemaining; + } + framesJustRead = ma_dr_mp3_read_pcm_frames_raw(pMP3, framesToReadNow, pTempF32); + if (framesJustRead == 0) { + break; + } + ma_dr_mp3_f32_to_s16((ma_int16*)MA_DR_MP3_OFFSET_PTR(pBufferOut, sizeof(ma_int16) * totalPCMFramesRead * pMP3->channels), pTempF32, framesJustRead * pMP3->channels); + totalPCMFramesRead += framesJustRead; + } + return totalPCMFramesRead; + } +#endif +} +static void ma_dr_mp3_reset(ma_dr_mp3* pMP3) +{ + MA_DR_MP3_ASSERT(pMP3 != NULL); + pMP3->pcmFramesConsumedInMP3Frame = 0; + pMP3->pcmFramesRemainingInMP3Frame = 0; + pMP3->currentPCMFrame = 0; + pMP3->dataSize = 0; + pMP3->atEnd = MA_FALSE; + ma_dr_mp3dec_init(&pMP3->decoder); +} +static ma_bool32 ma_dr_mp3_seek_to_start_of_stream(ma_dr_mp3* pMP3) +{ + MA_DR_MP3_ASSERT(pMP3 != NULL); + MA_DR_MP3_ASSERT(pMP3->onSeek != NULL); + if (!ma_dr_mp3__on_seek(pMP3, 0, ma_dr_mp3_seek_origin_start)) { + return MA_FALSE; + } + ma_dr_mp3_reset(pMP3); + return MA_TRUE; +} +static ma_bool32 ma_dr_mp3_seek_forward_by_pcm_frames__brute_force(ma_dr_mp3* pMP3, ma_uint64 frameOffset) +{ + ma_uint64 framesRead; +#if defined(MA_DR_MP3_FLOAT_OUTPUT) + framesRead = ma_dr_mp3_read_pcm_frames_f32(pMP3, frameOffset, NULL); +#else + framesRead = ma_dr_mp3_read_pcm_frames_s16(pMP3, frameOffset, NULL); +#endif + if (framesRead != frameOffset) { + return MA_FALSE; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_mp3_seek_to_pcm_frame__brute_force(ma_dr_mp3* pMP3, ma_uint64 frameIndex) +{ + MA_DR_MP3_ASSERT(pMP3 != NULL); + if (frameIndex == pMP3->currentPCMFrame) { + return MA_TRUE; + } + if (frameIndex < pMP3->currentPCMFrame) { + if (!ma_dr_mp3_seek_to_start_of_stream(pMP3)) { + return MA_FALSE; + } + } + MA_DR_MP3_ASSERT(frameIndex >= pMP3->currentPCMFrame); + return ma_dr_mp3_seek_forward_by_pcm_frames__brute_force(pMP3, (frameIndex - pMP3->currentPCMFrame)); +} +static ma_bool32 ma_dr_mp3_find_closest_seek_point(ma_dr_mp3* pMP3, ma_uint64 frameIndex, ma_uint32* pSeekPointIndex) +{ + ma_uint32 iSeekPoint; + MA_DR_MP3_ASSERT(pSeekPointIndex != NULL); + *pSeekPointIndex = 0; + if (frameIndex < pMP3->pSeekPoints[0].pcmFrameIndex) { + return MA_FALSE; + } + for (iSeekPoint = 0; iSeekPoint < pMP3->seekPointCount; ++iSeekPoint) { + if (pMP3->pSeekPoints[iSeekPoint].pcmFrameIndex > frameIndex) { + break; + } + *pSeekPointIndex = iSeekPoint; + } + return MA_TRUE; +} +static ma_bool32 ma_dr_mp3_seek_to_pcm_frame__seek_table(ma_dr_mp3* pMP3, ma_uint64 frameIndex) +{ + ma_dr_mp3_seek_point seekPoint; + ma_uint32 priorSeekPointIndex; + ma_uint16 iMP3Frame; + ma_uint64 leftoverFrames; + MA_DR_MP3_ASSERT(pMP3 != NULL); + MA_DR_MP3_ASSERT(pMP3->pSeekPoints != NULL); + MA_DR_MP3_ASSERT(pMP3->seekPointCount > 0); + if (ma_dr_mp3_find_closest_seek_point(pMP3, frameIndex, &priorSeekPointIndex)) { + seekPoint = pMP3->pSeekPoints[priorSeekPointIndex]; + } else { + seekPoint.seekPosInBytes = 0; + seekPoint.pcmFrameIndex = 0; + seekPoint.mp3FramesToDiscard = 0; + seekPoint.pcmFramesToDiscard = 0; + } + if (!ma_dr_mp3__on_seek_64(pMP3, seekPoint.seekPosInBytes, ma_dr_mp3_seek_origin_start)) { + return MA_FALSE; + } + ma_dr_mp3_reset(pMP3); + for (iMP3Frame = 0; iMP3Frame < seekPoint.mp3FramesToDiscard; ++iMP3Frame) { + ma_uint32 pcmFramesRead; + ma_dr_mp3d_sample_t* pPCMFrames; + pPCMFrames = NULL; + if (iMP3Frame == seekPoint.mp3FramesToDiscard-1) { + pPCMFrames = (ma_dr_mp3d_sample_t*)pMP3->pcmFrames; + } + pcmFramesRead = ma_dr_mp3_decode_next_frame_ex(pMP3, pPCMFrames); + if (pcmFramesRead == 0) { + return MA_FALSE; + } + } + pMP3->currentPCMFrame = seekPoint.pcmFrameIndex - seekPoint.pcmFramesToDiscard; + leftoverFrames = frameIndex - pMP3->currentPCMFrame; + return ma_dr_mp3_seek_forward_by_pcm_frames__brute_force(pMP3, leftoverFrames); +} +MA_API ma_bool32 ma_dr_mp3_seek_to_pcm_frame(ma_dr_mp3* pMP3, ma_uint64 frameIndex) +{ + if (pMP3 == NULL || pMP3->onSeek == NULL) { + return MA_FALSE; + } + if (frameIndex == 0) { + return ma_dr_mp3_seek_to_start_of_stream(pMP3); + } + if (pMP3->pSeekPoints != NULL && pMP3->seekPointCount > 0) { + return ma_dr_mp3_seek_to_pcm_frame__seek_table(pMP3, frameIndex); + } else { + return ma_dr_mp3_seek_to_pcm_frame__brute_force(pMP3, frameIndex); + } +} +MA_API ma_bool32 ma_dr_mp3_get_mp3_and_pcm_frame_count(ma_dr_mp3* pMP3, ma_uint64* pMP3FrameCount, ma_uint64* pPCMFrameCount) +{ + ma_uint64 currentPCMFrame; + ma_uint64 totalPCMFrameCount; + ma_uint64 totalMP3FrameCount; + if (pMP3 == NULL) { + return MA_FALSE; + } + if (pMP3->onSeek == NULL) { + return MA_FALSE; + } + currentPCMFrame = pMP3->currentPCMFrame; + if (!ma_dr_mp3_seek_to_start_of_stream(pMP3)) { + return MA_FALSE; + } + totalPCMFrameCount = 0; + totalMP3FrameCount = 0; + for (;;) { + ma_uint32 pcmFramesInCurrentMP3Frame; + pcmFramesInCurrentMP3Frame = ma_dr_mp3_decode_next_frame_ex(pMP3, NULL); + if (pcmFramesInCurrentMP3Frame == 0) { + break; + } + totalPCMFrameCount += pcmFramesInCurrentMP3Frame; + totalMP3FrameCount += 1; + } + if (!ma_dr_mp3_seek_to_start_of_stream(pMP3)) { + return MA_FALSE; + } + if (!ma_dr_mp3_seek_to_pcm_frame(pMP3, currentPCMFrame)) { + return MA_FALSE; + } + if (pMP3FrameCount != NULL) { + *pMP3FrameCount = totalMP3FrameCount; + } + if (pPCMFrameCount != NULL) { + *pPCMFrameCount = totalPCMFrameCount; + } + return MA_TRUE; +} +MA_API ma_uint64 ma_dr_mp3_get_pcm_frame_count(ma_dr_mp3* pMP3) +{ + ma_uint64 totalPCMFrameCount; + if (!ma_dr_mp3_get_mp3_and_pcm_frame_count(pMP3, NULL, &totalPCMFrameCount)) { + return 0; + } + return totalPCMFrameCount; +} +MA_API ma_uint64 ma_dr_mp3_get_mp3_frame_count(ma_dr_mp3* pMP3) +{ + ma_uint64 totalMP3FrameCount; + if (!ma_dr_mp3_get_mp3_and_pcm_frame_count(pMP3, &totalMP3FrameCount, NULL)) { + return 0; + } + return totalMP3FrameCount; +} +static void ma_dr_mp3__accumulate_running_pcm_frame_count(ma_dr_mp3* pMP3, ma_uint32 pcmFrameCountIn, ma_uint64* pRunningPCMFrameCount, float* pRunningPCMFrameCountFractionalPart) +{ + float srcRatio; + float pcmFrameCountOutF; + ma_uint32 pcmFrameCountOut; + srcRatio = (float)pMP3->mp3FrameSampleRate / (float)pMP3->sampleRate; + MA_DR_MP3_ASSERT(srcRatio > 0); + pcmFrameCountOutF = *pRunningPCMFrameCountFractionalPart + (pcmFrameCountIn / srcRatio); + pcmFrameCountOut = (ma_uint32)pcmFrameCountOutF; + *pRunningPCMFrameCountFractionalPart = pcmFrameCountOutF - pcmFrameCountOut; + *pRunningPCMFrameCount += pcmFrameCountOut; +} +typedef struct +{ + ma_uint64 bytePos; + ma_uint64 pcmFrameIndex; +} ma_dr_mp3__seeking_mp3_frame_info; +MA_API ma_bool32 ma_dr_mp3_calculate_seek_points(ma_dr_mp3* pMP3, ma_uint32* pSeekPointCount, ma_dr_mp3_seek_point* pSeekPoints) +{ + ma_uint32 seekPointCount; + ma_uint64 currentPCMFrame; + ma_uint64 totalMP3FrameCount; + ma_uint64 totalPCMFrameCount; + if (pMP3 == NULL || pSeekPointCount == NULL || pSeekPoints == NULL) { + return MA_FALSE; + } + seekPointCount = *pSeekPointCount; + if (seekPointCount == 0) { + return MA_FALSE; + } + currentPCMFrame = pMP3->currentPCMFrame; + if (!ma_dr_mp3_get_mp3_and_pcm_frame_count(pMP3, &totalMP3FrameCount, &totalPCMFrameCount)) { + return MA_FALSE; + } + if (totalMP3FrameCount < MA_DR_MP3_SEEK_LEADING_MP3_FRAMES+1) { + seekPointCount = 1; + pSeekPoints[0].seekPosInBytes = 0; + pSeekPoints[0].pcmFrameIndex = 0; + pSeekPoints[0].mp3FramesToDiscard = 0; + pSeekPoints[0].pcmFramesToDiscard = 0; + } else { + ma_uint64 pcmFramesBetweenSeekPoints; + ma_dr_mp3__seeking_mp3_frame_info mp3FrameInfo[MA_DR_MP3_SEEK_LEADING_MP3_FRAMES+1]; + ma_uint64 runningPCMFrameCount = 0; + float runningPCMFrameCountFractionalPart = 0; + ma_uint64 nextTargetPCMFrame; + ma_uint32 iMP3Frame; + ma_uint32 iSeekPoint; + if (seekPointCount > totalMP3FrameCount-1) { + seekPointCount = (ma_uint32)totalMP3FrameCount-1; + } + pcmFramesBetweenSeekPoints = totalPCMFrameCount / (seekPointCount+1); + if (!ma_dr_mp3_seek_to_start_of_stream(pMP3)) { + return MA_FALSE; + } + for (iMP3Frame = 0; iMP3Frame < MA_DR_MP3_SEEK_LEADING_MP3_FRAMES+1; ++iMP3Frame) { + ma_uint32 pcmFramesInCurrentMP3FrameIn; + MA_DR_MP3_ASSERT(pMP3->streamCursor >= pMP3->dataSize); + mp3FrameInfo[iMP3Frame].bytePos = pMP3->streamCursor - pMP3->dataSize; + mp3FrameInfo[iMP3Frame].pcmFrameIndex = runningPCMFrameCount; + pcmFramesInCurrentMP3FrameIn = ma_dr_mp3_decode_next_frame_ex(pMP3, NULL); + if (pcmFramesInCurrentMP3FrameIn == 0) { + return MA_FALSE; + } + ma_dr_mp3__accumulate_running_pcm_frame_count(pMP3, pcmFramesInCurrentMP3FrameIn, &runningPCMFrameCount, &runningPCMFrameCountFractionalPart); + } + nextTargetPCMFrame = 0; + for (iSeekPoint = 0; iSeekPoint < seekPointCount; ++iSeekPoint) { + nextTargetPCMFrame += pcmFramesBetweenSeekPoints; + for (;;) { + if (nextTargetPCMFrame < runningPCMFrameCount) { + pSeekPoints[iSeekPoint].seekPosInBytes = mp3FrameInfo[0].bytePos; + pSeekPoints[iSeekPoint].pcmFrameIndex = nextTargetPCMFrame; + pSeekPoints[iSeekPoint].mp3FramesToDiscard = MA_DR_MP3_SEEK_LEADING_MP3_FRAMES; + pSeekPoints[iSeekPoint].pcmFramesToDiscard = (ma_uint16)(nextTargetPCMFrame - mp3FrameInfo[MA_DR_MP3_SEEK_LEADING_MP3_FRAMES-1].pcmFrameIndex); + break; + } else { + size_t i; + ma_uint32 pcmFramesInCurrentMP3FrameIn; + for (i = 0; i < MA_DR_MP3_COUNTOF(mp3FrameInfo)-1; ++i) { + mp3FrameInfo[i] = mp3FrameInfo[i+1]; + } + mp3FrameInfo[MA_DR_MP3_COUNTOF(mp3FrameInfo)-1].bytePos = pMP3->streamCursor - pMP3->dataSize; + mp3FrameInfo[MA_DR_MP3_COUNTOF(mp3FrameInfo)-1].pcmFrameIndex = runningPCMFrameCount; + pcmFramesInCurrentMP3FrameIn = ma_dr_mp3_decode_next_frame_ex(pMP3, NULL); + if (pcmFramesInCurrentMP3FrameIn == 0) { + pSeekPoints[iSeekPoint].seekPosInBytes = mp3FrameInfo[0].bytePos; + pSeekPoints[iSeekPoint].pcmFrameIndex = nextTargetPCMFrame; + pSeekPoints[iSeekPoint].mp3FramesToDiscard = MA_DR_MP3_SEEK_LEADING_MP3_FRAMES; + pSeekPoints[iSeekPoint].pcmFramesToDiscard = (ma_uint16)(nextTargetPCMFrame - mp3FrameInfo[MA_DR_MP3_SEEK_LEADING_MP3_FRAMES-1].pcmFrameIndex); + break; + } + ma_dr_mp3__accumulate_running_pcm_frame_count(pMP3, pcmFramesInCurrentMP3FrameIn, &runningPCMFrameCount, &runningPCMFrameCountFractionalPart); + } + } + } + if (!ma_dr_mp3_seek_to_start_of_stream(pMP3)) { + return MA_FALSE; + } + if (!ma_dr_mp3_seek_to_pcm_frame(pMP3, currentPCMFrame)) { + return MA_FALSE; + } + } + *pSeekPointCount = seekPointCount; + return MA_TRUE; +} +MA_API ma_bool32 ma_dr_mp3_bind_seek_table(ma_dr_mp3* pMP3, ma_uint32 seekPointCount, ma_dr_mp3_seek_point* pSeekPoints) +{ + if (pMP3 == NULL) { + return MA_FALSE; + } + if (seekPointCount == 0 || pSeekPoints == NULL) { + pMP3->seekPointCount = 0; + pMP3->pSeekPoints = NULL; + } else { + pMP3->seekPointCount = seekPointCount; + pMP3->pSeekPoints = pSeekPoints; + } + return MA_TRUE; +} +static float* ma_dr_mp3__full_read_and_close_f32(ma_dr_mp3* pMP3, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount) +{ + ma_uint64 totalFramesRead = 0; + ma_uint64 framesCapacity = 0; + float* pFrames = NULL; + float temp[4096]; + MA_DR_MP3_ASSERT(pMP3 != NULL); + for (;;) { + ma_uint64 framesToReadRightNow = MA_DR_MP3_COUNTOF(temp) / pMP3->channels; + ma_uint64 framesJustRead = ma_dr_mp3_read_pcm_frames_f32(pMP3, framesToReadRightNow, temp); + if (framesJustRead == 0) { + break; + } + if (framesCapacity < totalFramesRead + framesJustRead) { + ma_uint64 oldFramesBufferSize; + ma_uint64 newFramesBufferSize; + ma_uint64 newFramesCap; + float* pNewFrames; + newFramesCap = framesCapacity * 2; + if (newFramesCap < totalFramesRead + framesJustRead) { + newFramesCap = totalFramesRead + framesJustRead; + } + oldFramesBufferSize = framesCapacity * pMP3->channels * sizeof(float); + newFramesBufferSize = newFramesCap * pMP3->channels * sizeof(float); + if (newFramesBufferSize > (ma_uint64)MA_SIZE_MAX) { + break; + } + pNewFrames = (float*)ma_dr_mp3__realloc_from_callbacks(pFrames, (size_t)newFramesBufferSize, (size_t)oldFramesBufferSize, &pMP3->allocationCallbacks); + if (pNewFrames == NULL) { + ma_dr_mp3__free_from_callbacks(pFrames, &pMP3->allocationCallbacks); + break; + } + pFrames = pNewFrames; + framesCapacity = newFramesCap; + } + MA_DR_MP3_COPY_MEMORY(pFrames + totalFramesRead*pMP3->channels, temp, (size_t)(framesJustRead*pMP3->channels*sizeof(float))); + totalFramesRead += framesJustRead; + if (framesJustRead != framesToReadRightNow) { + break; + } + } + if (pConfig != NULL) { + pConfig->channels = pMP3->channels; + pConfig->sampleRate = pMP3->sampleRate; + } + ma_dr_mp3_uninit(pMP3); + if (pTotalFrameCount) { + *pTotalFrameCount = totalFramesRead; + } + return pFrames; +} +static ma_int16* ma_dr_mp3__full_read_and_close_s16(ma_dr_mp3* pMP3, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount) +{ + ma_uint64 totalFramesRead = 0; + ma_uint64 framesCapacity = 0; + ma_int16* pFrames = NULL; + ma_int16 temp[4096]; + MA_DR_MP3_ASSERT(pMP3 != NULL); + for (;;) { + ma_uint64 framesToReadRightNow = MA_DR_MP3_COUNTOF(temp) / pMP3->channels; + ma_uint64 framesJustRead = ma_dr_mp3_read_pcm_frames_s16(pMP3, framesToReadRightNow, temp); + if (framesJustRead == 0) { + break; + } + if (framesCapacity < totalFramesRead + framesJustRead) { + ma_uint64 newFramesBufferSize; + ma_uint64 oldFramesBufferSize; + ma_uint64 newFramesCap; + ma_int16* pNewFrames; + newFramesCap = framesCapacity * 2; + if (newFramesCap < totalFramesRead + framesJustRead) { + newFramesCap = totalFramesRead + framesJustRead; + } + oldFramesBufferSize = framesCapacity * pMP3->channels * sizeof(ma_int16); + newFramesBufferSize = newFramesCap * pMP3->channels * sizeof(ma_int16); + if (newFramesBufferSize > (ma_uint64)MA_SIZE_MAX) { + break; + } + pNewFrames = (ma_int16*)ma_dr_mp3__realloc_from_callbacks(pFrames, (size_t)newFramesBufferSize, (size_t)oldFramesBufferSize, &pMP3->allocationCallbacks); + if (pNewFrames == NULL) { + ma_dr_mp3__free_from_callbacks(pFrames, &pMP3->allocationCallbacks); + break; + } + pFrames = pNewFrames; + framesCapacity = newFramesCap; + } + MA_DR_MP3_COPY_MEMORY(pFrames + totalFramesRead*pMP3->channels, temp, (size_t)(framesJustRead*pMP3->channels*sizeof(ma_int16))); + totalFramesRead += framesJustRead; + if (framesJustRead != framesToReadRightNow) { + break; + } + } + if (pConfig != NULL) { + pConfig->channels = pMP3->channels; + pConfig->sampleRate = pMP3->sampleRate; + } + ma_dr_mp3_uninit(pMP3); + if (pTotalFrameCount) { + *pTotalFrameCount = totalFramesRead; + } + return pFrames; +} +MA_API float* ma_dr_mp3_open_and_read_pcm_frames_f32(ma_dr_mp3_read_proc onRead, ma_dr_mp3_seek_proc onSeek, void* pUserData, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_mp3 mp3; + if (!ma_dr_mp3_init(&mp3, onRead, onSeek, pUserData, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_mp3__full_read_and_close_f32(&mp3, pConfig, pTotalFrameCount); +} +MA_API ma_int16* ma_dr_mp3_open_and_read_pcm_frames_s16(ma_dr_mp3_read_proc onRead, ma_dr_mp3_seek_proc onSeek, void* pUserData, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_mp3 mp3; + if (!ma_dr_mp3_init(&mp3, onRead, onSeek, pUserData, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_mp3__full_read_and_close_s16(&mp3, pConfig, pTotalFrameCount); +} +MA_API float* ma_dr_mp3_open_memory_and_read_pcm_frames_f32(const void* pData, size_t dataSize, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_mp3 mp3; + if (!ma_dr_mp3_init_memory(&mp3, pData, dataSize, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_mp3__full_read_and_close_f32(&mp3, pConfig, pTotalFrameCount); +} +MA_API ma_int16* ma_dr_mp3_open_memory_and_read_pcm_frames_s16(const void* pData, size_t dataSize, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_mp3 mp3; + if (!ma_dr_mp3_init_memory(&mp3, pData, dataSize, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_mp3__full_read_and_close_s16(&mp3, pConfig, pTotalFrameCount); +} +#ifndef MA_DR_MP3_NO_STDIO +MA_API float* ma_dr_mp3_open_file_and_read_pcm_frames_f32(const char* filePath, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_mp3 mp3; + if (!ma_dr_mp3_init_file(&mp3, filePath, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_mp3__full_read_and_close_f32(&mp3, pConfig, pTotalFrameCount); +} +MA_API ma_int16* ma_dr_mp3_open_file_and_read_pcm_frames_s16(const char* filePath, ma_dr_mp3_config* pConfig, ma_uint64* pTotalFrameCount, const ma_allocation_callbacks* pAllocationCallbacks) +{ + ma_dr_mp3 mp3; + if (!ma_dr_mp3_init_file(&mp3, filePath, pAllocationCallbacks)) { + return NULL; + } + return ma_dr_mp3__full_read_and_close_s16(&mp3, pConfig, pTotalFrameCount); +} +#endif +MA_API void* ma_dr_mp3_malloc(size_t sz, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks != NULL) { + return ma_dr_mp3__malloc_from_callbacks(sz, pAllocationCallbacks); + } else { + return ma_dr_mp3__malloc_default(sz, NULL); + } +} +MA_API void ma_dr_mp3_free(void* p, const ma_allocation_callbacks* pAllocationCallbacks) +{ + if (pAllocationCallbacks != NULL) { + ma_dr_mp3__free_from_callbacks(p, pAllocationCallbacks); + } else { + ma_dr_mp3__free_default(p, NULL); + } +} +#endif +/* dr_mp3_c end */ +#endif /* MA_DR_MP3_IMPLEMENTATION */ +#endif /* MA_NO_MP3 */ + + +/* End globally disabled warnings. */ +#if defined(_MSC_VER) + #pragma warning(pop) +#endif + +#endif /* miniaudio_c */ +#endif /* MINIAUDIO_IMPLEMENTATION */ + + +/* +This software is available as a choice of the following licenses. Choose +whichever you prefer. + +=============================================================================== +ALTERNATIVE 1 - Public Domain (www.unlicense.org) +=============================================================================== +This is free and unencumbered software released into the public domain. + +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. + +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +For more information, please refer to + +=============================================================================== +ALTERNATIVE 2 - MIT No Attribution +=============================================================================== +Copyright 2025 David Reid + +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +*/ diff --git a/platform/posix_impl.c b/platform/posix_impl.c new file mode 100644 index 0000000..1ad989f --- /dev/null +++ b/platform/posix_impl.c @@ -0,0 +1,44 @@ +#include "platform.h" +#include "compiler.h" +#include +#include +#include + +void platform_os_once_run(union platform_os_once *once, void (*run_fn)(void)) +{ + pthread_once(&once->_i, run_fn); +} + +size_t platform_os_getpagesize(void) +{ + // Try and get the page size, use macro as a last resort + long result = sysconf(_SC_PAGESIZE); + if (result < 0) + result = PAGE_SIZE; + return result; +} + +struct platform_os_timespec platform_os_time(int type) +{ + struct platform_os_timespec ts = {}; + ts.err = -1; + int posix_type = -1; + switch (type) { + case PLATFORM_OS_TIME_WALL: + posix_type = CLOCK_REALTIME; + break; + case PLATFORM_OS_TIME_MONO: + posix_type = CLOCK_MONOTONIC; + break; + case PLATFORM_OS_TIME_TICK__UNUSED: + default: + return ts; + } + struct timespec posix_ts; + if (clock_gettime(posix_type, &posix_ts) >= 0) { + ts.sec = posix_ts.tv_sec; + ts.nsec = posix_ts.tv_nsec; + ts.err = 0; + } + return ts; +} diff --git a/platform/sdl3.c b/platform/sdl3.c new file mode 100644 index 0000000..098f8fc --- /dev/null +++ b/platform/sdl3.c @@ -0,0 +1,342 @@ +#include +#include +#include +#include +#include +#include +#include "platform.h" +#include "os.h" +#include "core_callbacks.h" + +#define CONTROLLERS_MAX 32 +static struct { + size_t count; + unsigned int ids[CONTROLLERS_MAX]; + SDL_Gamepad *objs[CONTROLLERS_MAX]; +} platform_gamepads = {}; +static union platform_window platform_window = (union platform_window){(struct platform_window_i *)0}; +static bool platform_running = true; + +void platform_os_log_v(int type, const char *tag, const char *fmt, va_list ap) +{ + printf("[%s:%d] ", tag, type); + vprintf(fmt, ap); + if (!strchr(fmt, '\n')) + putc('\n', stdout); +} + +union platform_os_file platform_os_asset_open(const char *assetpath) +{ + union platform_os_file fp = (union platform_os_file){}; + + // +1 for path separator, +1 for null terminator + const char *base = SDL_GetBasePath(); + size_t base_size = strlen(base); + size_t fullpath_size = base_size + strlen(assetpath) + 2; + if (fullpath_size <= base_size) + return fp; + + char *fullpath = malloc(fullpath_size); + if (!fullpath) + return fp; + + int n = snprintf(fullpath, fullpath_size, "%s/%s", base, assetpath); + if (n < 0 || (size_t)n >= fullpath_size) + return fp; +#ifdef _WIN32 + // On Windows, change path sep to backslash + for (size_t i = 0, l = strlen(fullpath); i < l; i++) + if (fullpath[i] == '/') + fullpath[i] = '\\'; +#endif + fp._i = (struct platform_os_file_i *)fopen(fullpath, "rb"); + return fp; +} + +int platform_gfx_valid(union platform_window window) +{ + return !!(SDL_Window *)window._i; +} + +int platform_window_width(union platform_window window) +{ + if (!platform_gfx_valid(window)) + return 0; + + int w = 0; + SDL_GetWindowSizeInPixels((SDL_Window *)window._i, &w, (int *)0); + return w; +} + +int platform_window_height(union platform_window window) +{ + if (!platform_gfx_valid(window)) + return 0; + + int h = 0; + SDL_GetWindowSizeInPixels((SDL_Window *)window._i, (int *)0, &h); + return h; +} + +void platform_gfx_present(union platform_window window) +{ + if (platform_gfx_valid(window)) + SDL_GL_SwapWindow((SDL_Window *)window._i); +} + +// Convert SDL event constants to platform constants +// These functions are basically giant switch statements +// and they clutter up the code, so they're declared here +// and defined at the bottom of this file +static int platform_from_sdl_ev_type(int type); +static int platform_from_sdl_ev_keycode(int keycode); +static int platform_from_sdl_ev_gaxis(int axis); +static int platform_from_sdl_ev_gbutton(int button); + +static void platform_mainloop(void) +{ + SDL_Event s_ev; + if (platform_window._i) { + union platform_event p_ev = {}; + while (SDL_PollEvent(&s_ev) > 0) { + p_ev.type = platform_from_sdl_ev_type(s_ev.type); + switch(s_ev.type) { + case SDL_EVENT_QUIT: + platform_running = false; + return; + case SDL_EVENT_KEY_UP: + case SDL_EVENT_KEY_DOWN: + p_ev.key.which = s_ev.key.which; + p_ev.key.sym = platform_from_sdl_ev_keycode(s_ev.key.key); + p_ev.key.repeat = s_ev.key.repeat; + break; + case SDL_EVENT_MOUSE_BUTTON_UP: + case SDL_EVENT_MOUSE_BUTTON_DOWN: + p_ev.button.which = s_ev.button.which; + p_ev.button.pixels = (struct vector2f32){(float)s_ev.button.x, (float)s_ev.button.y}; + p_ev.button.index = s_ev.button.button; + p_ev.button.clicks = s_ev.button.clicks; + break; + case SDL_EVENT_MOUSE_MOTION: + p_ev.motion.pixels = (struct vector2f32){s_ev.motion.x, s_ev.motion.y}; + p_ev.motion.which = s_ev.motion.which; + break; + case SDL_EVENT_FINGER_UP: + case SDL_EVENT_FINGER_DOWN: + case SDL_EVENT_FINGER_MOTION: + // TODO: Test on a touchscreen (or even in an Android SDL build) + p_ev.finger.device = s_ev.tfinger.touchID; + p_ev.finger.id = s_ev.tfinger.fingerID; + p_ev.finger.pixels = (struct vector2f32){ + s_ev.tfinger.x * platform_window_width(platform_window), + s_ev.tfinger.y * platform_window_height(platform_window), + }; + p_ev.finger.pressure = s_ev.tfinger.pressure; + break; + case SDL_EVENT_GAMEPAD_ADDED: + case SDL_EVENT_GAMEPAD_REMOVED: + p_ev.gdevice.which = s_ev.gdevice.which; + break; + case SDL_EVENT_GAMEPAD_AXIS_MOTION: + p_ev.gaxis.which = s_ev.gaxis.which; + p_ev.gaxis.value = s_ev.gaxis.value; + p_ev.gaxis.axis = platform_from_sdl_ev_gaxis(s_ev.gaxis.axis); + break; + case SDL_EVENT_GAMEPAD_BUTTON_DOWN: + case SDL_EVENT_GAMEPAD_BUTTON_UP: + p_ev.gbutton.which = s_ev.gbutton.which; + p_ev.gbutton.button = platform_from_sdl_ev_gbutton(s_ev.gbutton.button); + break; + default: + break; + } + + // If the gamepad fails to be opened or closed, + // don't deliver the event. Otherwise, do. + if (s_ev.type == SDL_EVENT_GAMEPAD_ADDED) { + if (platform_gamepads.count >= CONTROLLERS_MAX) + continue; + SDL_Gamepad *gamepad = SDL_OpenGamepad(s_ev.gdevice.which); + if (!gamepad) + continue; + size_t idx = platform_gamepads.count++; + platform_gamepads.ids[idx] = s_ev.gdevice.which; + platform_gamepads.objs[idx] = gamepad; + } else if (s_ev.type == SDL_EVENT_GAMEPAD_REMOVED) { + SDL_Gamepad *gamepad = (SDL_Gamepad *)0; + for (size_t i = 0; i < platform_gamepads.count; i++) { + if (platform_gamepads.ids[i] == s_ev.gdevice.which) { + gamepad = platform_gamepads.objs[i]; + size_t last = --platform_gamepads.count; + platform_gamepads.ids[i] = platform_gamepads.ids[last]; + platform_gamepads.objs[i] = platform_gamepads.objs[last]; + break; + } + } + if (!gamepad) + continue; + SDL_CloseGamepad(gamepad); + } + core_callback_event(&p_ev, platform_window); + } + } + + // Do the updates and rendering + core_callback_render(platform_window); +} + +int main(int argc, char **argv) +{ + // SDL Init everything + // It's not safe to continue if things aren't initialized properly + // because we rely on SDL for general OS things as well. + int status = SDL_Init(SDL_INIT_VIDEO | SDL_INIT_EVENTS | SDL_INIT_GAMEPAD); + if (status < 0) { + os_loge("SDL Init failed! %s", SDL_GetError()); + return -status; + } + + // SDL caches this result, so to get more consistent performance + // on the first call from the game, do this in platform code. + (void)SDL_GetBasePath(); + + // Call the entry function before we initialize afx/gfx, similar to mobile + core_callback_entry(argc, argv); + + // Normally, on other platforms, not having a window created is fine + // because the user can terminate the application easily even without one. + // On PC and web, that's not the case, so just exit. + platform_window._i = (struct platform_window_i *)SDL_CreateWindow( + "Miss(le) Hell", + 450, 800, + SDL_WINDOW_OPENGL); + if (!platform_window._i) + os_logw("SDL_CreateWindow failed! %s", SDL_GetError()); + else if (!SDL_GL_CreateContext((SDL_Window *)platform_window._i)) { + os_logw("SDL_GL_CreateContext failed! %s", SDL_GetError()); + SDL_DestroyWindow((SDL_Window *)platform_window._i); + platform_window._i = (struct platform_window_i *)0; + } + // TODO: + // A headless option in a platform cmdline arg might be useful + if (!platform_window._i) { + os_loge("Window initialization failed! Exiting"); + return 1; + } + + os_logi("Hello World..."); + platform_gfx_startup(platform_window); + core_callback_gfx_startup(platform_window); + // TODO: I have no idea if this works on emscripten! Test it! +#ifdef EMSCRIPTEN + (void)platform_running; + emscripten_set_main_loop(platform_mainloop, 60, 1); +#else + while (platform_running) + platform_mainloop(); +#endif + core_callback_gfx_shutdown(platform_window); + platform_gfx_shutdown(platform_window); + os_logi("Goodbye world!"); + return 0; +} + +static int platform_from_sdl_ev_type(int type) +{ + switch (type) { + case SDL_EVENT_FINGER_DOWN: return PLATFORM_EVENT_FINGER_DOWN; + case SDL_EVENT_FINGER_UP: return PLATFORM_EVENT_FINGER_UP; + case SDL_EVENT_FINGER_MOTION: return PLATFORM_EVENT_FINGER_MOTION; + case SDL_EVENT_KEY_UP: return PLATFORM_EVENT_KEY_UP; + case SDL_EVENT_KEY_DOWN: return PLATFORM_EVENT_KEY_DOWN; + case SDL_EVENT_MOUSE_BUTTON_UP: return PLATFORM_EVENT_MOUSE_BUTTON_UP; + case SDL_EVENT_MOUSE_BUTTON_DOWN: return PLATFORM_EVENT_MOUSE_BUTTON_DOWN; + case SDL_EVENT_MOUSE_MOTION: return PLATFORM_EVENT_MOUSE_MOTION; + case SDL_EVENT_GAMEPAD_ADDED: return PLATFORM_EVENT_GAMEPAD_CONNECTED; + case SDL_EVENT_GAMEPAD_REMOVED: return PLATFORM_EVENT_GAMEPAD_DISCONNECTED; + case SDL_EVENT_GAMEPAD_AXIS_MOTION: return PLATFORM_EVENT_GAMEPAD_AXIS_MOTION; + case SDL_EVENT_GAMEPAD_BUTTON_DOWN: return PLATFORM_EVENT_GAMEPAD_BUTTON_DOWN; + case SDL_EVENT_GAMEPAD_BUTTON_UP: return PLATFORM_EVENT_GAMEPAD_BUTTON_UP; + } + return PLATFORM_EVENT_NONE; +} + +static int platform_from_sdl_ev_keycode(int keycode) +{ + switch (keycode) { + case SDLK_A: return PLATFORM_KEY_A; + case SDLK_B: return PLATFORM_KEY_B; + case SDLK_C: return PLATFORM_KEY_C; + case SDLK_D: return PLATFORM_KEY_D; + case SDLK_E: return PLATFORM_KEY_E; + case SDLK_F: return PLATFORM_KEY_F; + case SDLK_G: return PLATFORM_KEY_G; + case SDLK_H: return PLATFORM_KEY_H; + case SDLK_I: return PLATFORM_KEY_I; + case SDLK_J: return PLATFORM_KEY_J; + case SDLK_K: return PLATFORM_KEY_K; + case SDLK_L: return PLATFORM_KEY_L; + case SDLK_M: return PLATFORM_KEY_M; + case SDLK_N: return PLATFORM_KEY_N; + case SDLK_O: return PLATFORM_KEY_O; + case SDLK_P: return PLATFORM_KEY_P; + case SDLK_Q: return PLATFORM_KEY_Q; + case SDLK_R: return PLATFORM_KEY_R; + case SDLK_S: return PLATFORM_KEY_S; + case SDLK_T: return PLATFORM_KEY_T; + case SDLK_U: return PLATFORM_KEY_U; + case SDLK_V: return PLATFORM_KEY_V; + case SDLK_W: return PLATFORM_KEY_W; + case SDLK_X: return PLATFORM_KEY_X; + case SDLK_Y: return PLATFORM_KEY_Y; + case SDLK_Z: return PLATFORM_KEY_Z; + } + return PLATFORM_KEY_NONE; +} + +static int platform_from_sdl_ev_gaxis(int axis) +{ + switch (axis) + { + case SDL_GAMEPAD_AXIS_LEFTX: return PLATFORM_GAXIS_LEFTX; + case SDL_GAMEPAD_AXIS_LEFTY: return PLATFORM_GAXIS_LEFTY; + case SDL_GAMEPAD_AXIS_RIGHTX: return PLATFORM_GAXIS_RIGHTX; + case SDL_GAMEPAD_AXIS_RIGHTY: return PLATFORM_GAXIS_RIGHTY; + case SDL_GAMEPAD_AXIS_LEFT_TRIGGER: return PLATFORM_GAXIS_LEFT_TRIGGER; + case SDL_GAMEPAD_AXIS_RIGHT_TRIGGER: return PLATFORM_GAXIS_RIGHT_TRIGGER; + } + return PLATFORM_GBUTTON_NONE; +} + +static int platform_from_sdl_ev_gbutton(int button) +{ + switch (button) + { + case SDL_GAMEPAD_BUTTON_SOUTH: return PLATFORM_GBUTTON_SOUTH; + case SDL_GAMEPAD_BUTTON_EAST: return PLATFORM_GBUTTON_EAST; + case SDL_GAMEPAD_BUTTON_WEST: return PLATFORM_GBUTTON_WEST; + case SDL_GAMEPAD_BUTTON_NORTH: return PLATFORM_GBUTTON_NORTH; + case SDL_GAMEPAD_BUTTON_BACK: return PLATFORM_GBUTTON_BACK; + case SDL_GAMEPAD_BUTTON_GUIDE: return PLATFORM_GBUTTON_GUIDE; + case SDL_GAMEPAD_BUTTON_START: return PLATFORM_GBUTTON_START; + case SDL_GAMEPAD_BUTTON_LEFT_STICK: return PLATFORM_GBUTTON_LEFT_STICK; + case SDL_GAMEPAD_BUTTON_RIGHT_STICK: return PLATFORM_GBUTTON_RIGHT_STICK; + case SDL_GAMEPAD_BUTTON_LEFT_SHOULDER: return PLATFORM_GBUTTON_LEFT_SHOULDER; + case SDL_GAMEPAD_BUTTON_RIGHT_SHOULDER: return PLATFORM_GBUTTON_RIGHT_SHOULDER; + case SDL_GAMEPAD_BUTTON_DPAD_UP: return PLATFORM_GBUTTON_DPAD_UP; + case SDL_GAMEPAD_BUTTON_DPAD_DOWN: return PLATFORM_GBUTTON_DPAD_DOWN; + case SDL_GAMEPAD_BUTTON_DPAD_LEFT: return PLATFORM_GBUTTON_DPAD_LEFT; + case SDL_GAMEPAD_BUTTON_MISC1: return PLATFORM_GBUTTON_MISC1; + case SDL_GAMEPAD_BUTTON_RIGHT_PADDLE1: return PLATFORM_GBUTTON_RIGHT_PADDLE1; + case SDL_GAMEPAD_BUTTON_LEFT_PADDLE1: return PLATFORM_GBUTTON_LEFT_PADDLE1; + case SDL_GAMEPAD_BUTTON_RIGHT_PADDLE2: return PLATFORM_GBUTTON_RIGHT_PADDLE2; + case SDL_GAMEPAD_BUTTON_LEFT_PADDLE2: return PLATFORM_GBUTTON_LEFT_PADDLE2; + case SDL_GAMEPAD_BUTTON_TOUCHPAD: return PLATFORM_GBUTTON_TOUCHPAD; + case SDL_GAMEPAD_BUTTON_MISC2: return PLATFORM_GBUTTON_MISC2; + case SDL_GAMEPAD_BUTTON_MISC3: return PLATFORM_GBUTTON_MISC3; + case SDL_GAMEPAD_BUTTON_MISC4: return PLATFORM_GBUTTON_MISC4; + case SDL_GAMEPAD_BUTTON_MISC5: return PLATFORM_GBUTTON_MISC5; + case SDL_GAMEPAD_BUTTON_MISC6: return PLATFORM_GBUTTON_MISC6; + } + return PLATFORM_GBUTTON_NONE; +} diff --git a/platform/sokol_gfx.h b/platform/sokol_gfx.h new file mode 100644 index 0000000..df30cc7 --- /dev/null +++ b/platform/sokol_gfx.h @@ -0,0 +1,21518 @@ +#if defined(SOKOL_IMPL) && !defined(SOKOL_GFX_IMPL) +#define SOKOL_GFX_IMPL +#endif +#ifndef SOKOL_GFX_INCLUDED +/* + sokol_gfx.h -- simple 3D API wrapper + + Project URL: https://github.com/floooh/sokol + + Example code: https://github.com/floooh/sokol-samples + + Do this: + #define SOKOL_IMPL or + #define SOKOL_GFX_IMPL + before you include this file in *one* C or C++ file to create the + implementation. + + In the same place define one of the following to select the rendering + backend: + #define SOKOL_GLCORE + #define SOKOL_GLES3 + #define SOKOL_D3D11 + #define SOKOL_METAL + #define SOKOL_WGPU + #define SOKOL_DUMMY_BACKEND + + I.e. for the desktop GL it should look like this: + + #include ... + #include ... + #define SOKOL_IMPL + #define SOKOL_GLCORE + #include "sokol_gfx.h" + + The dummy backend replaces the platform-specific backend code with empty + stub functions. This is useful for writing tests that need to run on the + command line. + + Optionally provide the following defines with your own implementations: + + SOKOL_ASSERT(c) - your own assert macro (default: assert(c)) + SOKOL_UNREACHABLE() - a guard macro for unreachable code (default: assert(false)) + SOKOL_GFX_API_DECL - public function declaration prefix (default: extern) + SOKOL_API_DECL - same as SOKOL_GFX_API_DECL + SOKOL_API_IMPL - public function implementation prefix (default: -) + SOKOL_TRACE_HOOKS - enable trace hook callbacks (search below for TRACE HOOKS) + SOKOL_EXTERNAL_GL_LOADER - indicates that you're using your own GL loader, in this case + sokol_gfx.h will not include any platform GL headers and disable + the integrated Win32 GL loader + + If sokol_gfx.h is compiled as a DLL, define the following before + including the declaration or implementation: + + SOKOL_DLL + + On Windows, SOKOL_DLL will define SOKOL_GFX_API_DECL as __declspec(dllexport) + or __declspec(dllimport) as needed. + + If you want to compile without deprecated structs and functions, + define: + + SOKOL_NO_DEPRECATED + + Optionally define the following to force debug checks and validations + even in release mode: + + SOKOL_DEBUG - by default this is defined if _DEBUG is defined + + sokol_gfx DOES NOT: + =================== + - create a window, swapchain or the 3D-API context/device, you must do this + before sokol_gfx is initialized, and pass any required information + (like 3D device pointers) to the sokol_gfx initialization call + + - present the rendered frame, how this is done exactly usually depends + on how the window and 3D-API context/device was created + + - provide a unified shader language, instead 3D-API-specific shader + source-code or shader-bytecode must be provided (for the "official" + offline shader cross-compiler / code-generator, see here: + https://github.com/floooh/sokol-tools/blob/master/docs/sokol-shdc.md) + + + STEP BY STEP + ============ + --- to initialize sokol_gfx, after creating a window and a 3D-API + context/device, call: + + sg_setup(const sg_desc*) + + Depending on the selected 3D backend, sokol-gfx requires some + information about its runtime environment, like a GPU device pointer, + default swapchain pixel formats and so on. If you are using sokol_app.h + for the window system glue, you can use a helper function provided in + the sokol_glue.h header: + + #include "sokol_gfx.h" + #include "sokol_app.h" + #include "sokol_glue.h" + //... + sg_setup(&(sg_desc){ + .environment = sglue_environment(), + }); + + To get any logging output for errors and from the validation layer, you + need to provide a logging callback. Easiest way is through sokol_log.h: + + #include "sokol_log.h" + //... + sg_setup(&(sg_desc){ + //... + .logger.func = slog_func, + }); + + --- create resource objects (at least buffers, shaders and pipelines, + and optionally images, samplers and render-pass-attachments): + + sg_buffer sg_make_buffer(const sg_buffer_desc*) + sg_image sg_make_image(const sg_image_desc*) + sg_sampler sg_make_sampler(const sg_sampler_desc*) + sg_shader sg_make_shader(const sg_shader_desc*) + sg_pipeline sg_make_pipeline(const sg_pipeline_desc*) + sg_attachments sg_make_attachments(const sg_attachments_desc*) + + --- start a render- or compute-pass: + + sg_begin_pass(const sg_pass* pass); + + Typically, render passes render into an externally provided swapchain which + presents the rendering result on the display. Such a 'swapchain pass' + is started like this: + + sg_begin_pass(&(sg_pass){ .action = { ... }, .swapchain = sglue_swapchain() }) + + ...where .action is an sg_pass_action struct containing actions to be performed + at the start and end of a render pass (such as clearing the render surfaces to + a specific color), and .swapchain is an sg_swapchain struct with all the required + information to render into the swapchain's surfaces. + + To start an 'offscreen render pass' into sokol-gfx image objects, an sg_attachment + object handle is required instead of an sg_swapchain struct. An offscreen + pass is started like this (assuming attachments is an sg_attachments handle): + + sg_begin_pass(&(sg_pass){ .action = { ... }, .attachments = attachments }); + + To start a compute-pass, just set the .compute item to true: + + sg_begin_pass(&(sg_pass){ .compute = true }); + + --- set the pipeline state for the next draw call with: + + sg_apply_pipeline(sg_pipeline pip) + + --- fill an sg_bindings struct with the resource bindings for the next + draw- or dispatch-call (0..N vertex buffers, 0 or 1 index buffer, 0..N images, + samplers and storage-buffers), and call: + + sg_apply_bindings(const sg_bindings* bindings) + + to update the resource bindings. Note that in a compute pass, no vertex- + or index-buffer bindings are allowed and will be rejected by the validation + layer. + + --- optionally update shader uniform data with: + + sg_apply_uniforms(int ub_slot, const sg_range* data) + + Read the section 'UNIFORM DATA LAYOUT' to learn about the expected memory layout + of the uniform data passed into sg_apply_uniforms(). + + --- kick off a draw call with: + + sg_draw(int base_element, int num_elements, int num_instances) + + The sg_draw() function unifies all the different ways to render primitives + in a single call (indexed vs non-indexed rendering, and instanced vs non-instanced + rendering). In case of indexed rendering, base_element and num_element specify + indices in the currently bound index buffer. In case of non-indexed rendering + base_element and num_elements specify vertices in the currently bound + vertex-buffer(s). To perform instanced rendering, the rendering pipeline + must be setup for instancing (see sg_pipeline_desc below), a separate vertex buffer + containing per-instance data must be bound, and the num_instances parameter + must be > 1. + + --- ...or kick of a dispatch call to invoke a compute shader workload: + + sg_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) + + The dispatch args define the number of 'compute workgroups' processed + by the currently applied compute shader. + + --- finish the current pass with: + + sg_end_pass() + + --- when done with the current frame, call + + sg_commit() + + --- at the end of your program, shutdown sokol_gfx with: + + sg_shutdown() + + --- if you need to destroy resources before sg_shutdown(), call: + + sg_destroy_buffer(sg_buffer buf) + sg_destroy_image(sg_image img) + sg_destroy_sampler(sg_sampler smp) + sg_destroy_shader(sg_shader shd) + sg_destroy_pipeline(sg_pipeline pip) + sg_destroy_attachments(sg_attachments atts) + + --- to set a new viewport rectangle, call: + + sg_apply_viewport(int x, int y, int width, int height, bool origin_top_left) + + ...or if you want to specify the viewport rectangle with float values: + + sg_apply_viewportf(float x, float y, float width, float height, bool origin_top_left) + + --- to set a new scissor rect, call: + + sg_apply_scissor_rect(int x, int y, int width, int height, bool origin_top_left) + + ...or with float values: + + sg_apply_scissor_rectf(float x, float y, float width, float height, bool origin_top_left) + + Both sg_apply_viewport() and sg_apply_scissor_rect() must be called + inside a rendering pass (e.g. not in a compute pass, or outside a pass) + + Note that sg_begin_default_pass() and sg_begin_pass() will reset both the + viewport and scissor rectangles to cover the entire framebuffer. + + --- to update (overwrite) the content of buffer and image resources, call: + + sg_update_buffer(sg_buffer buf, const sg_range* data) + sg_update_image(sg_image img, const sg_image_data* data) + + Buffers and images to be updated must have been created with + SG_USAGE_DYNAMIC or SG_USAGE_STREAM. + + Only one update per frame is allowed for buffer and image resources when + using the sg_update_*() functions. The rationale is to have a simple + protection from the CPU scribbling over data the GPU is currently + using, or the CPU having to wait for the GPU + + Buffer and image updates can be partial, as long as a rendering + operation only references the valid (updated) data in the + buffer or image. + + --- to append a chunk of data to a buffer resource, call: + + int sg_append_buffer(sg_buffer buf, const sg_range* data) + + The difference to sg_update_buffer() is that sg_append_buffer() + can be called multiple times per frame to append new data to the + buffer piece by piece, optionally interleaved with draw calls referencing + the previously written data. + + sg_append_buffer() returns a byte offset to the start of the + written data, this offset can be assigned to + sg_bindings.vertex_buffer_offsets[n] or + sg_bindings.index_buffer_offset + + Code example: + + for (...) { + const void* data = ...; + const int num_bytes = ...; + int offset = sg_append_buffer(buf, &(sg_range) { .ptr=data, .size=num_bytes }); + bindings.vertex_buffer_offsets[0] = offset; + sg_apply_pipeline(pip); + sg_apply_bindings(&bindings); + sg_apply_uniforms(...); + sg_draw(...); + } + + A buffer to be used with sg_append_buffer() must have been created + with SG_USAGE_DYNAMIC or SG_USAGE_STREAM. + + If the application appends more data to the buffer then fits into + the buffer, the buffer will go into the "overflow" state for the + rest of the frame. + + Any draw calls attempting to render an overflown buffer will be + silently dropped (in debug mode this will also result in a + validation error). + + You can also check manually if a buffer is in overflow-state by calling + + bool sg_query_buffer_overflow(sg_buffer buf) + + You can manually check to see if an overflow would occur before adding + any data to a buffer by calling + + bool sg_query_buffer_will_overflow(sg_buffer buf, size_t size) + + NOTE: Due to restrictions in underlying 3D-APIs, appended chunks of + data will be 4-byte aligned in the destination buffer. This means + that there will be gaps in index buffers containing 16-bit indices + when the number of indices in a call to sg_append_buffer() is + odd. This isn't a problem when each call to sg_append_buffer() + is associated with one draw call, but will be problematic when + a single indexed draw call spans several appended chunks of indices. + + --- to check at runtime for optional features, limits and pixelformat support, + call: + + sg_features sg_query_features() + sg_limits sg_query_limits() + sg_pixelformat_info sg_query_pixelformat(sg_pixel_format fmt) + + --- if you need to call into the underlying 3D-API directly, you must call: + + sg_reset_state_cache() + + ...before calling sokol_gfx functions again + + --- you can inspect the original sg_desc structure handed to sg_setup() + by calling sg_query_desc(). This will return an sg_desc struct with + the default values patched in instead of any zero-initialized values + + --- you can get a desc struct matching the creation attributes of a + specific resource object via: + + sg_buffer_desc sg_query_buffer_desc(sg_buffer buf) + sg_image_desc sg_query_image_desc(sg_image img) + sg_sampler_desc sg_query_sampler_desc(sg_sampler smp) + sg_shader_desc sq_query_shader_desc(sg_shader shd) + sg_pipeline_desc sg_query_pipeline_desc(sg_pipeline pip) + sg_attachments_desc sg_query_attachments_desc(sg_attachments atts) + + ...but NOTE that the returned desc structs may be incomplete, only + creation attributes that are kept around internally after resource + creation will be filled in, and in some cases (like shaders) that's + very little. Any missing attributes will be set to zero. The returned + desc structs might still be useful as partial blueprint for creating + similar resources if filled up with the missing attributes. + + Calling the query-desc functions on an invalid resource will return + completely zeroed structs (it makes sense to check the resource state + with sg_query_*_state() first) + + --- you can query the default resource creation parameters through the functions + + sg_buffer_desc sg_query_buffer_defaults(const sg_buffer_desc* desc) + sg_image_desc sg_query_image_defaults(const sg_image_desc* desc) + sg_sampler_desc sg_query_sampler_defaults(const sg_sampler_desc* desc) + sg_shader_desc sg_query_shader_defaults(const sg_shader_desc* desc) + sg_pipeline_desc sg_query_pipeline_defaults(const sg_pipeline_desc* desc) + sg_attachments_desc sg_query_attachments_defaults(const sg_attachments_desc* desc) + + These functions take a pointer to a desc structure which may contain + zero-initialized items for default values. These zero-init values + will be replaced with their concrete values in the returned desc + struct. + + --- you can inspect various internal resource runtime values via: + + sg_buffer_info sg_query_buffer_info(sg_buffer buf) + sg_image_info sg_query_image_info(sg_image img) + sg_sampler_info sg_query_sampler_info(sg_sampler smp) + sg_shader_info sg_query_shader_info(sg_shader shd) + sg_pipeline_info sg_query_pipeline_info(sg_pipeline pip) + sg_attachments_info sg_query_attachments_info(sg_attachments atts) + + ...please note that the returned info-structs are tied quite closely + to sokol_gfx.h internals, and may change more often than other + public API functions and structs. + + --- you can query frame stats and control stats collection via: + + sg_query_frame_stats() + sg_enable_frame_stats() + sg_disable_frame_stats() + sg_frame_stats_enabled() + + --- you can ask at runtime what backend sokol_gfx.h has been compiled for: + + sg_backend sg_query_backend(void) + + --- call the following helper functions to compute the number of + bytes in a texture row or surface for a specific pixel format. + These functions might be helpful when preparing image data for consumption + by sg_make_image() or sg_update_image(): + + int sg_query_row_pitch(sg_pixel_format fmt, int width, int int row_align_bytes); + int sg_query_surface_pitch(sg_pixel_format fmt, int width, int height, int row_align_bytes); + + Width and height are generally in number pixels, but note that 'row' has different meaning + for uncompressed vs compressed pixel formats: for uncompressed formats, a row is identical + with a single line if pixels, while in compressed formats, one row is a line of *compression blocks*. + + This is why calling sg_query_surface_pitch() for a compressed pixel format and height + N, N+1, N+2, ... may return the same result. + + The row_align_bytes parammeter is for added flexibility. For image data that goes into + the sg_make_image() or sg_update_image() this should generally be 1, because these + functions take tightly packed image data as input no matter what alignment restrictions + exist in the backend 3D APIs. + + ON INITIALIZATION: + ================== + When calling sg_setup(), a pointer to an sg_desc struct must be provided + which contains initialization options. These options provide two types + of information to sokol-gfx: + + (1) upper bounds and limits needed to allocate various internal + data structures: + - the max number of resources of each type that can + be alive at the same time, this is used for allocating + internal pools + - the max overall size of uniform data that can be + updated per frame, including a worst-case alignment + per uniform update (this worst-case alignment is 256 bytes) + - the max size of all dynamic resource updates (sg_update_buffer, + sg_append_buffer and sg_update_image) per frame + - the max number of compute-dispatch calls in a compute pass + Not all of those limit values are used by all backends, but it is + good practice to provide them none-the-less. + + (2) 3D backend "environment information" in a nested sg_environment struct: + - pointers to backend-specific context- or device-objects (for instance + the D3D11, WebGPU or Metal device objects) + - defaults for external swapchain pixel formats and sample counts, + these will be used as default values in image and pipeline objects, + and the sg_swapchain struct passed into sg_begin_pass() + Usually you provide a complete sg_environment struct through + a helper function, as an example look at the sglue_environment() + function in the sokol_glue.h header. + + See the documentation block of the sg_desc struct below for more information. + + + ON RENDER PASSES + ================ + Relevant samples: + - https://floooh.github.io/sokol-html5/offscreen-sapp.html + - https://floooh.github.io/sokol-html5/offscreen-msaa-sapp.html + - https://floooh.github.io/sokol-html5/mrt-sapp.html + - https://floooh.github.io/sokol-html5/mrt-pixelformats-sapp.html + + A render pass groups rendering commands into a set of render target images + (called 'pass attachments'). Render target images can be used in subsequent + passes as textures (it is invalid to use the same image both as render target + and as texture in the same pass). + + The following sokol-gfx functions must only be called inside a render-pass: + + sg_apply_viewport[f] + sg_apply_scissor_rect[f] + sg_draw + + The following function may be called inside a render- or compute-pass, but + not outside a pass: + + sg_apply_pipeline + sg_apply_bindings + sg_apply_uniforms + + A frame must have at least one 'swapchain render pass' which renders into an + externally provided swapchain provided as an sg_swapchain struct to the + sg_begin_pass() function. If you use sokol_gfx.h together with sokol_app.h, + just call the sglue_swapchain() helper function in sokol_glue.h to + provide the swapchain information. Otherwise the following information + must be provided: + + - the color pixel-format of the swapchain's render surface + - an optional depth/stencil pixel format if the swapchain + has a depth/stencil buffer + - an optional sample-count for MSAA rendering + - NOTE: the above three values can be zero-initialized, in that + case the defaults from the sg_environment struct will be used that + had been passed to the sg_setup() function. + - a number of backend specific objects: + - GL/GLES3: just a GL framebuffer handle + - D3D11: + - an ID3D11RenderTargetView for the rendering surface + - if MSAA is used, an ID3D11RenderTargetView as + MSAA resolve-target + - an optional ID3D11DepthStencilView for the + depth/stencil buffer + - WebGPU + - a WGPUTextureView object for the rendering surface + - if MSAA is used, a WGPUTextureView object as MSAA resolve target + - an optional WGPUTextureView for the + - Metal (NOTE that the roles of provided surfaces is slightly + different in Metal than in D3D11 or WebGPU, notably, the + CAMetalDrawable is either rendered to directly, or serves + as MSAA resolve target): + - a CAMetalDrawable object which is either rendered + into directly, or in case of MSAA rendering, serves + as MSAA-resolve-target + - if MSAA is used, an multisampled MTLTexture where + rendering goes into + - an optional MTLTexture for the depth/stencil buffer + + It's recommended that you create a helper function which returns an + initialized sg_swapchain struct by value. This can then be directly plugged + into the sg_begin_pass function like this: + + sg_begin_pass(&(sg_pass){ .swapchain = sglue_swapchain() }); + + As an example for such a helper function check out the function sglue_swapchain() + in the sokol_glue.h header. + + For offscreen render passes, the render target images used in a render pass + are baked into an immutable sg_attachments object. + + For a simple offscreen scenario with one color-, one depth-stencil-render + target and without multisampling, creating an attachment object looks like this: + + First create two render target images, one with a color pixel format, + and one with the depth- or depth-stencil pixel format. Both images + must have the same dimensions: + + const sg_image color_img = sg_make_image(&(sg_image_desc){ + .render_target = true, + .width = 256, + .height = 256, + .pixel_format = SG_PIXELFORMAT_RGBA8, + .sample_count = 1, + }); + const sg_image depth_img = sg_make_image(&(sg_image_desc){ + .render_target = true, + .width = 256, + .height = 256, + .pixel_format = SG_PIXELFORMAT_DEPTH, + .sample_count = 1, + }); + + NOTE: when creating render target images, have in mind that some default values + are aligned with the default environment attributes in the sg_environment struct + that was passed into the sg_setup() call: + + - the default value for sg_image_desc.pixel_format is taken from + sg_environment.defaults.color_format + - the default value for sg_image_desc.sample_count is taken from + sg_environment.defaults.sample_count + - the default value for sg_image_desc.num_mipmaps is always 1 + + Next create an attachments object: + + const sg_attachments atts = sg_make_attachments(&(sg_attachments_desc){ + .colors[0].image = color_img, + .depth_stencil.image = depth_img, + }); + + This attachments object is then passed into the sg_begin_pass() function + in place of the swapchain struct: + + sg_begin_pass(&(sg_pass){ .attachments = atts }); + + Swapchain and offscreen passes form dependency trees each with a swapchain + pass at the root, offscreen passes as nodes, and render target images as + dependencies between passes. + + sg_pass_action structs are used to define actions that should happen at the + start and end of rendering passes (such as clearing pass attachments to a + specific color or depth-value, or performing an MSAA resolve operation at + the end of a pass). + + A typical sg_pass_action object which clears the color attachment to black + might look like this: + + const sg_pass_action = { + .colors[0] = { + .load_action = SG_LOADACTION_CLEAR, + .clear_value = { 0.0f, 0.0f, 0.0f, 1.0f } + } + }; + + This omits the defaults for the color attachment store action, and + the depth-stencil-attachments actions. The same pass action with the + defaults explicitly filled in would look like this: + + const sg_pass_action pass_action = { + .colors[0] = { + .load_action = SG_LOADACTION_CLEAR, + .store_action = SG_STOREACTION_STORE, + .clear_value = { 0.0f, 0.0f, 0.0f, 1.0f } + }, + .depth = = { + .load_action = SG_LOADACTION_CLEAR, + .store_action = SG_STOREACTION_DONTCARE, + .clear_value = 1.0f, + }, + .stencil = { + .load_action = SG_LOADACTION_CLEAR, + .store_action = SG_STOREACTION_DONTCARE, + .clear_value = 0 + } + }; + + With the sg_pass object and sg_pass_action struct in place everything + is ready now for the actual render pass: + + Using such this prepared sg_pass_action in a swapchain pass looks like + this: + + sg_begin_pass(&(sg_pass){ + .action = pass_action, + .swapchain = sglue_swapchain() + }); + ... + sg_end_pass(); + + ...of alternatively in one offscreen pass: + + sg_begin_pass(&(sg_pass){ + .action = pass_action, + .attachments = attachments, + }); + ... + sg_end_pass(); + + Offscreen rendering can also go into a mipmap, or a slice/face of + a cube-, array- or 3d-image (which some restrictions, for instance + it's not possible to create a 3D image with a depth/stencil pixel format, + these exceptions are generally caught by the sokol-gfx validation layer). + + The mipmap/slice selection happens at attachments creation time, for instance + to render into mipmap 2 of slice 3 of an array texture: + + const sg_attachments atts = sg_make_attachments(&(sg_attachments_desc){ + .colors[0] = { + .image = color_img, + .mip_level = 2, + .slice = 3, + }, + .depth_stencil.image = depth_img, + }); + + If MSAA offscreen rendering is desired, the multi-sample rendering result + must be 'resolved' into a separate 'resolve image', before that image can + be used as texture. + + Creating a simple attachments object for multisampled rendering requires + 3 attachment images: the color attachment image which has a sample + count > 1, a resolve attachment image of the same size and pixel format + but a sample count == 1, and a depth/stencil attachment image with + the same size and sample count as the color attachment image: + + const sg_image color_img = sg_make_image(&(sg_image_desc){ + .render_target = true, + .width = 256, + .height = 256, + .pixel_format = SG_PIXELFORMAT_RGBA8, + .sample_count = 4, + }); + const sg_image resolve_img = sg_make_image(&(sg_image_desc){ + .render_target = true, + .width = 256, + .height = 256, + .pixel_format = SG_PIXELFORMAT_RGBA8, + .sample_count = 1, + }); + const sg_image depth_img = sg_make_image(&(sg_image_desc){ + .render_target = true, + .width = 256, + .height = 256, + .pixel_format = SG_PIXELFORMAT_DEPTH, + .sample_count = 4, + }); + + ...create the attachments object: + + const sg_attachments atts = sg_make_attachments(&(sg_attachments_desc){ + .colors[0].image = color_img, + .resolves[0].image = resolve_img, + .depth_stencil.image = depth_img, + }); + + If an attachments object defines a resolve image in a specific resolve attachment slot, + an 'msaa resolve operation' will happen in sg_end_pass(). + + In this scenario, the content of the MSAA color attachment doesn't need to be + preserved (since it's only needed inside sg_end_pass for the msaa-resolve), so + the .store_action should be set to "don't care": + + const sg_pass_action = { + .colors[0] = { + .load_action = SG_LOADACTION_CLEAR, + .store_action = SG_STOREACTION_DONTCARE, + .clear_value = { 0.0f, 0.0f, 0.0f, 1.0f } + } + }; + + The actual render pass looks as usual: + + sg_begin_pass(&(sg_pass){ .action = pass_action, .attachments = atts }); + ... + sg_end_pass(); + + ...after sg_end_pass() the only difference to the non-msaa scenario is that the + rendering result which is going to be used as texture in a followup pass is + in 'resolve_img', not in 'color_img' (in fact, trying to bind color_img as a + texture would result in a validation error). + + + ON COMPUTE PASSES + ================= + Compute passes are used to update the content of storage resources + (currently only storage buffers) by running compute shader code on + the GPU. This will almost always be more efficient than computing + that same data on the CPU and uploading the data via `sg_update_buffer()`. + + NOTE: compute passes are only supported on the following platforms and + backends: + + - macOS and iOS with Metal + - Windows with D3D11 and OpenGL + - Linux with OpenGL or GLES3.1+ + - Web with WebGPU + - Android with GLES3.1+ + + ...this means compute shaders can't be used on the following platform/backend + combos (the same restrictions apply to using storage buffers without compute + shaders): + + - macOS with GL + - iOS with GLES3 + - Web with WebGL2 + + A compute pass is started with: + + sg_begin_pass(&(sg_pass){ .compute = true }); + + ...and finished with: + + sg_end_pass(); + + Typically the following functions will be called inside a compute pass: + + sg_apply_pipeline + sg_apply_bindings + sg_apply_uniforms + sg_dispatch + + The following functions are disallowed inside a compute pass + and will cause validation layer errors: + + sg_apply_viewport[f] + sg_apply_scissor_rect[f] + sg_draw + + Only special 'compute shaders' and 'compute pipelines' can be used in + compute passes. A compute shader only has a compute-function instead + of a vertex- and fragment-function pair, and it doesn't accept vertex- + and index-buffers as input, only storage-buffers, textures and non-filtering + samplers (more details on compute shaders in the following section). + + A compute pipeline is created by providing a compute shader object, + setting the .compute creation parameter to true and not defining any + 'render state': + + sg_pipeline pip = sg_make_pipeline(&(sg_pipeline_desc){ + .compute = true, + .shader = compute_shader, + }); + + The sg_apply_bindings and sg_apply_uniforms calls are the same as in + render passes, with the exception that no vertex- and index-buffers + can be bound in the sg_apply_bindings call. + + Finally to kick off a compute workload, call sg_dispatch with the + number of workgroups in the x, y and z-dimension: + + sg_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) + + Also see the following compute-shader samples: + + - https://floooh.github.io/sokol-webgpu/instancing-compute-sapp.html + - https://floooh.github.io/sokol-webgpu/computeboids-sapp.html + + + ON SHADER CREATION + ================== + sokol-gfx doesn't come with an integrated shader cross-compiler, instead + backend-specific shader sources or binary blobs need to be provided when + creating a shader object, along with reflection information about the + shader resource binding interface needed to bind sokol-gfx resources to the + proper shader inputs. + + The easiest way to provide all this shader creation data is to use the + sokol-shdc shader compiler tool to compile shaders from a common + GLSL syntax into backend-specific sources or binary blobs, along with + shader interface information and uniform blocks mapped to C structs. + + To create a shader using a C header which has been code-generated by sokol-shdc: + + // include the C header code-generated by sokol-shdc: + #include "myshader.glsl.h" + ... + + // create shader using a code-generated helper function from the C header: + sg_shader shd = sg_make_shader(myshader_shader_desc(sg_query_backend())); + + The samples in the 'sapp' subdirectory of the sokol-samples project + also use the sokol-shdc approach: + + https://github.com/floooh/sokol-samples/tree/master/sapp + + If you're planning to use sokol-shdc, you can stop reading here, instead + continue with the sokol-shdc documentation: + + https://github.com/floooh/sokol-tools/blob/master/docs/sokol-shdc.md + + To create shaders with backend-specific shader code or binary blobs, + the sg_make_shader() function requires the following information: + + - Shader code or shader binary blobs for the vertex- and fragment-, or the + compute-shader-stage: + - for the desktop GL backend, source code can be provided in '#version 410' or + '#version 430', version 430 is required when using storage buffers and + compute shaders support, but note that this is not available on macOS + - for the GLES3 backend, source code must be provided in '#version 300 es' syntax + - for the D3D11 backend, shaders can be provided as source or binary + blobs, the source code should be in HLSL4.0 (for compatibility with old + low-end GPUs) or preferably in HLSL5.0 syntax, note that when + shader source code is provided for the D3D11 backend, sokol-gfx will + dynamically load 'd3dcompiler_47.dll' + - for the Metal backends, shaders can be provided as source or binary blobs, the + MSL version should be in 'metal-1.1' (other versions may work but are not tested) + - for the WebGPU backend, shaders must be provided as WGSL source code + - optionally the following shader-code related attributes can be provided: + - an entry function name (only on D3D11 or Metal, but not OpenGL) + - on D3D11 only, a compilation target (default is "vs_4_0" and "ps_4_0") + + - Information about the input vertex attributes used by the vertex shader, + most of that backend-specific: + - An optional 'base type' (float, signed-/unsigned-int) for each vertex + attribute. When provided, this used by the validation layer to check + that the CPU-side input vertex format is compatible with the input + vertex declaration of the vertex shader. + - Metal: no location information needed since vertex attributes are always bound + by their attribute location defined in the shader via '[[attribute(N)]]' + - WebGPU: no location information needed since vertex attributes are always + bound by their attribute location defined in the shader via `@location(N)` + - GLSL: vertex attribute names can be optionally provided, in that case their + location will be looked up by name, otherwise, the vertex attribute location + can be defined with 'layout(location = N)' + - D3D11: a 'semantic name' and 'semantic index' must be provided for each vertex + attribute, e.g. if the vertex attribute is defined as 'TEXCOORD1' in the shader, + the semantic name would be 'TEXCOORD', and the semantic index would be '1' + + NOTE that vertex attributes currently must not have gaps. This requirement + may be relaxed in the future. + + - Specifically for Metal compute shaders, the 'number of threads per threadgroup' + must be provided. Normally this is extracted by sokol-shdc from the GLSL + shader source code. For instance the following statement in the input + GLSL: + + layout(local_size_x=64, local_size_y=1, local_size_z=1) in; + + ...will be communicated to the sokol-gfx Metal backend in the + code-generated sg_shader_desc struct: + + (sg_shader_desc){ + .mtl_threads_per_threadgroup = { .x = 64, .y = 1, .z = 1 }, + } + + - Information about each uniform block used in the shader: + - the shader stage of the uniform block (vertex, fragment or compute) + - the size of the uniform block in number of bytes + - a memory layout hint (currently 'native' or 'std140') where 'native' defines a + backend-specific memory layout which shouldn't be used for cross-platform code. + Only std140 guarantees a backend-agnostic memory layout. + - a backend-specific bind slot: + - D3D11/HLSL: the buffer register N (`register(bN)`) where N is 0..7 + - Metal/MSL: the buffer bind slot N (`[[buffer(N)]]`) where N is 0..7 + - WebGPU: the binding N in `@group(0) @binding(N)` where N is 0..15 + - For GLSL only: a description of the internal uniform block layout, which maps + member types and their offsets on the CPU side to uniform variable names + in the GLSL shader + - please also NOTE the documentation sections about UNIFORM DATA LAYOUT + and CROSS-BACKEND COMMON UNIFORM DATA LAYOUT below! + + - A description of each storage buffer used in the shader: + - the shader stage of the storage buffer + - a boolean 'readonly' flag, this is used for validation and hazard + tracking in some 3D backends. Note that in render passes, only + readonly storage buffer bindings are allowed. In compute passes, any + read/write storage buffer binding is assumed to be written to by the + compute shader. + - a backend-specific bind slot: + - D3D11/HLSL: + - for readonly storage buffer bindings: the texture register N + (`register(tN)`) where N is 0..23 (in HLSL, readonly storage + buffers and textures share the same bind space for + 'shader resource views') + - for read/write storage buffer buffer bindings: the UAV register N + (`register(uN)`) where N is 0..7 (in HLSL, readwrite storage + buffers use their own bind space for 'unordered access views') + - Metal/MSL: the buffer bind slot N (`[[buffer(N)]]`) where N is 8..15 + - WebGPU/WGSL: the binding N in `@group(0) @binding(N)` where N is 0..127 + - GL/GLSL: the buffer binding N in `layout(binding=N)` where N is 0..7 + - note that storage buffers are not supported on all backends + and platforms + + - A description of each texture/image used in the shader: + - the shader stage of the texture (vertex, fragment or compute) + - the expected image type: + - SG_IMAGETYPE_2D + - SG_IMAGETYPE_CUBE + - SG_IMAGETYPE_3D + - SG_IMAGETYPE_ARRAY + - the expected 'image sample type': + - SG_IMAGESAMPLETYPE_FLOAT + - SG_IMAGESAMPLETYPE_DEPTH + - SG_IMAGESAMPLETYPE_SINT + - SG_IMAGESAMPLETYPE_UINT + - SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT + - a flag whether the texture is expected to be multisampled + - a backend-specific bind slot: + - D3D11/HLSL: the texture register N (`register(tN)`) where N is 0..23 + (in HLSL, readonly storage buffers and texture share the same bind space) + - Metal/MSL: the texture bind slot N (`[[texture(N)]]`) where N is 0..15 + - WebGPU/WGSL: the binding N in `@group(0) @binding(N)` where N is 0..127 + + - A description of each sampler used in the shader: + - the shader stage of the sampler (vertex, fragment or compute) + - the expected sampler type: + - SG_SAMPLERTYPE_FILTERING, + - SG_SAMPLERTYPE_NONFILTERING, + - SG_SAMPLERTYPE_COMPARISON, + - a backend-specific bind slot: + - D3D11/HLSL: the sampler register N (`register(sN)`) where N is 0..15 + - Metal/MSL: the sampler bind slot N (`[[sampler(N)]]`) where N is 0..15 + - WebGPU/WGSL: the binding N in `@group(0) @binding(N)` where N is 0..127 + + - An array of 'image-sampler-pairs' used by the shader to sample textures, + for D3D11, Metal and WebGPU this is used for validation purposes to check + whether the texture and sampler are compatible with each other (especially + WebGPU is very picky about combining the correct + texture-sample-type with the correct sampler-type). For GLSL an + additional 'combined-image-sampler name' must be provided because 'OpenGL + style GLSL' cannot handle separate texture and sampler objects, but still + groups them into a traditional GLSL 'sampler object'. + + Compatibility rules for image-sample-type vs sampler-type are as follows: + + - SG_IMAGESAMPLETYPE_FLOAT => (SG_SAMPLERTYPE_FILTERING or SG_SAMPLERTYPE_NONFILTERING) + - SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT => SG_SAMPLERTYPE_NONFILTERING + - SG_IMAGESAMPLETYPE_SINT => SG_SAMPLERTYPE_NONFILTERING + - SG_IMAGESAMPLETYPE_UINT => SG_SAMPLERTYPE_NONFILTERING + - SG_IMAGESAMPLETYPE_DEPTH => SG_SAMPLERTYPE_COMPARISON + + Backend-specific bindslot ranges (not relevant when using sokol-shdc): + + - D3D11/HLSL: + - separate bindslot space per shader stage + - uniform blocks (as cbuffer): `register(b0..b7)` + - textures and readonly storage buffers: `register(t0..t23)` + - read/write storage buffers: `register(u0..u7)` + - samplers: `register(s0..s15)` + - Metal/MSL: + - separate bindslot space per shader stage + - uniform blocks: `[[buffer(0..7)]]` + - storage buffers: `[[buffer(8..15)]]` + - textures: `[[texture(0..15)]]` + - samplers: `[[sampler(0..15)]]` + - WebGPU/WGSL: + - common bindslot space across shader stages + - uniform blocks: `@group(0) @binding(0..15)` + - textures, samplers and storage buffers: `@group(1) @binding(0..127)` + - GL/GLSL: + - uniforms and image-samplers are bound by name + - storage buffers: `layout(std430, binding=0..7)` (common + bindslot space across shader stages) + + For example code of how to create backend-specific shader objects, + please refer to the following samples: + + - for D3D11: https://github.com/floooh/sokol-samples/tree/master/d3d11 + - for Metal: https://github.com/floooh/sokol-samples/tree/master/metal + - for OpenGL: https://github.com/floooh/sokol-samples/tree/master/glfw + - for GLES3: https://github.com/floooh/sokol-samples/tree/master/html5 + - for WebGPU: https://github.com/floooh/sokol-samples/tree/master/wgpu + + + ON SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT AND SG_SAMPLERTYPE_NONFILTERING + ======================================================================== + The WebGPU backend introduces the concept of 'unfilterable-float' textures, + which can only be combined with 'nonfiltering' samplers (this is a restriction + specific to WebGPU, but since the same sokol-gfx code should work across + all backend, the sokol-gfx validation layer also enforces this restriction + - the alternative would be undefined behaviour in some backend APIs on + some devices). + + The background is that some mobile devices (most notably iOS devices) can + not perform linear filtering when sampling textures with certain pixel + formats, most notable the 32F formats: + + - SG_PIXELFORMAT_R32F + - SG_PIXELFORMAT_RG32F + - SG_PIXELFORMAT_RGBA32F + + The information of whether a shader is going to be used with such an + unfilterable-float texture must already be provided in the sg_shader_desc + struct when creating the shader (see the above section "ON SHADER CREATION"). + + If you are using the sokol-shdc shader compiler, the information whether a + texture/sampler binding expects an 'unfilterable-float/nonfiltering' + texture/sampler combination cannot be inferred from the shader source + alone, you'll need to provide this hint via annotation-tags. For instance + here is an example from the ozz-skin-sapp.c sample shader which samples an + RGBA32F texture with skinning matrices in the vertex shader: + + ```glsl + @image_sample_type joint_tex unfilterable_float + uniform texture2D joint_tex; + @sampler_type smp nonfiltering + uniform sampler smp; + ``` + + This will result in SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT and + SG_SAMPLERTYPE_NONFILTERING being written to the code-generated + sg_shader_desc struct. + + + ON VERTEX FORMATS + ================= + Sokol-gfx implements the same strict mapping rules from CPU-side + vertex component formats to GPU-side vertex input data types: + + - float and packed normalized CPU-side formats must be used as + floating point base type in the vertex shader + - packed signed-integer CPU-side formats must be used as signed + integer base type in the vertex shader + - packed unsigned-integer CPU-side formats must be used as unsigned + integer base type in the vertex shader + + These mapping rules are enforced by the sokol-gfx validation layer, + but only when sufficient reflection information is provided in + `sg_shader_desc.attrs[].base_type`. This is the case when sokol-shdc + is used, otherwise the default base_type will be SG_SHADERATTRBASETYPE_UNDEFINED + which causes the sokol-gfx validation check to be skipped (of course you + can also provide the per-attribute base type information manually when + not using sokol-shdc). + + The detailed mapping rules from SG_VERTEXFORMAT_* to GLSL data types + are as follows: + + - FLOAT[*] => float, vec* + - BYTE4N => vec* (scaled to -1.0 .. +1.0) + - UBYTE4N => vec* (scaled to 0.0 .. +1.0) + - SHORT[*]N => vec* (scaled to -1.0 .. +1.0) + - USHORT[*]N => vec* (scaled to 0.0 .. +1.0) + - INT[*] => int, ivec* + - UINT[*] => uint, uvec* + - BYTE4 => int* + - UBYTE4 => uint* + - SHORT[*] => int* + - USHORT[*] => uint* + + NOTE that sokol-gfx only provides vertex formats with sizes of a multiple + of 4 (e.g. BYTE4N but not BYTE2N). This is because vertex components must + be 4-byte aligned anyway. + + + UNIFORM DATA LAYOUT: + ==================== + NOTE: if you use the sokol-shdc shader compiler tool, you don't need to worry + about the following details. + + The data that's passed into the sg_apply_uniforms() function must adhere to + specific layout rules so that the GPU shader finds the uniform block + items at the right offset. + + For the D3D11 and Metal backends, sokol-gfx only cares about the size of uniform + blocks, but not about the internal layout. The data will just be copied into + a uniform/constant buffer in a single operation and it's up you to arrange the + CPU-side layout so that it matches the GPU side layout. This also means that with + the D3D11 and Metal backends you are not limited to a 'cross-platform' subset + of uniform variable types. + + If you ever only use one of the D3D11, Metal *or* WebGPU backend, you can stop reading here. + + For the GL backends, the internal layout of uniform blocks matters though, + and you are limited to a small number of uniform variable types. This is + because sokol-gfx must be able to locate the uniform block members in order + to upload them to the GPU with glUniformXXX() calls. + + To describe the uniform block layout to sokol-gfx, the following information + must be passed to the sg_make_shader() call in the sg_shader_desc struct: + + - a hint about the used packing rule (either SG_UNIFORMLAYOUT_NATIVE or + SG_UNIFORMLAYOUT_STD140) + - a list of the uniform block members types in the correct order they + appear on the CPU side + + For example if the GLSL shader has the following uniform declarations: + + uniform mat4 mvp; + uniform vec2 offset0; + uniform vec2 offset1; + uniform vec2 offset2; + + ...and on the CPU side, there's a similar C struct: + + typedef struct { + float mvp[16]; + float offset0[2]; + float offset1[2]; + float offset2[2]; + } params_t; + + ...the uniform block description in the sg_shader_desc must look like this: + + sg_shader_desc desc = { + .vs.uniform_blocks[0] = { + .size = sizeof(params_t), + .layout = SG_UNIFORMLAYOUT_NATIVE, // this is the default and can be omitted + .uniforms = { + // order must be the same as in 'params_t': + [0] = { .name = "mvp", .type = SG_UNIFORMTYPE_MAT4 }, + [1] = { .name = "offset0", .type = SG_UNIFORMTYPE_VEC2 }, + [2] = { .name = "offset1", .type = SG_UNIFORMTYPE_VEC2 }, + [3] = { .name = "offset2", .type = SG_UNIFORMTYPE_VEC2 }, + } + } + }; + + With this information sokol-gfx can now compute the correct offsets of the data items + within the uniform block struct. + + The SG_UNIFORMLAYOUT_NATIVE packing rule works fine if only the GL backends are used, + but for proper D3D11/Metal/GL a subset of the std140 layout must be used which is + described in the next section: + + + CROSS-BACKEND COMMON UNIFORM DATA LAYOUT + ======================================== + For cross-platform / cross-3D-backend code it is important that the same uniform block + layout on the CPU side can be used for all sokol-gfx backends. To achieve this, + a common subset of the std140 layout must be used: + + - The uniform block layout hint in sg_shader_desc must be explicitly set to + SG_UNIFORMLAYOUT_STD140. + - Only the following GLSL uniform types can be used (with their associated sokol-gfx enums): + - float => SG_UNIFORMTYPE_FLOAT + - vec2 => SG_UNIFORMTYPE_FLOAT2 + - vec3 => SG_UNIFORMTYPE_FLOAT3 + - vec4 => SG_UNIFORMTYPE_FLOAT4 + - int => SG_UNIFORMTYPE_INT + - ivec2 => SG_UNIFORMTYPE_INT2 + - ivec3 => SG_UNIFORMTYPE_INT3 + - ivec4 => SG_UNIFORMTYPE_INT4 + - mat4 => SG_UNIFORMTYPE_MAT4 + - Alignment for those types must be as follows (in bytes): + - float => 4 + - vec2 => 8 + - vec3 => 16 + - vec4 => 16 + - int => 4 + - ivec2 => 8 + - ivec3 => 16 + - ivec4 => 16 + - mat4 => 16 + - Arrays are only allowed for the following types: vec4, int4, mat4. + + Note that the HLSL cbuffer layout rules are slightly different from the + std140 layout rules, this means that the cbuffer declarations in HLSL code + must be tweaked so that the layout is compatible with std140. + + The by far easiest way to tackle the common uniform block layout problem is + to use the sokol-shdc shader cross-compiler tool! + + + ON STORAGE BUFFERS + ================== + The two main purpose of storage buffers are: + + - to be populated by compute shaders with dynamically generated data + - for providing random-access data to all shader stages + + Storage buffers can be used to pass large amounts of random access structured + data from the CPU side to the shaders. They are similar to data textures, but are + more convenient to use both on the CPU and shader side since they can be accessed + in shaders as as a 1-dimensional array of struct items. + + Storage buffers are *NOT* supported on the following platform/backend combos: + + - macOS+GL (because storage buffers require GL 4.3, while macOS only goes up to GL 4.1) + - platforms which only support a GLES3.0 context (WebGL2 and iOS) + + To use storage buffers, the following steps are required: + + - write a shader which uses storage buffers (vertex- and fragment-shaders + can only read from storage buffers, while compute-shaders can both read + and write storage buffers) + - create one or more storage buffers via sg_make_buffer() with the + buffer type SG_BUFFERTYPE_STORAGEBUFFER + - when creating a shader via sg_make_shader(), populate the sg_shader_desc + struct with binding info (when using sokol-shdc, this step will be taken care + of automatically) + - which storage buffer bind slots on the vertex-, fragment- or compute-stage + are occupied + - whether the storage buffer on that bind slot is readonly (readonly + bindings are required for vertex- and fragment-shaders, and in compute + shaders the readonly flag is used to control hazard tracking in some + 3D backends) + + - when calling sg_apply_bindings(), apply the matching bind slots with the previously + created storage buffers + - ...and that's it. + + For more details, see the following backend-agnostic sokol samples: + + - simple vertex pulling from a storage buffer: + - C code: https://github.com/floooh/sokol-samples/blob/master/sapp/vertexpull-sapp.c + - shader: https://github.com/floooh/sokol-samples/blob/master/sapp/vertexpull-sapp.glsl + - instanced rendering via storage buffers (vertex- and instance-pulling): + - C code: https://github.com/floooh/sokol-samples/blob/master/sapp/instancing-pull-sapp.c + - shader: https://github.com/floooh/sokol-samples/blob/master/sapp/instancing-pull-sapp.glsl + - storage buffers both on the vertex- and fragment-stage: + - C code: https://github.com/floooh/sokol-samples/blob/master/sapp/sbuftex-sapp.c + - shader: https://github.com/floooh/sokol-samples/blob/master/sapp/sbuftex-sapp.glsl + - the Ozz animation sample rewritten to pull all rendering data from storage buffers: + - C code: https://github.com/floooh/sokol-samples/blob/master/sapp/ozz-storagebuffer-sapp.cc + - shader: https://github.com/floooh/sokol-samples/blob/master/sapp/ozz-storagebuffer-sapp.glsl + - the instancing sample modified to use compute shaders: + - C code: https://github.com/floooh/sokol-samples/blob/master/sapp/instancing-compute-sapp.c + - shader: https://github.com/floooh/sokol-samples/blob/master/sapp/instancing-compute-sapp.glsl + - the Compute Boids sample ported to sokol-gfx: + - C code: https://github.com/floooh/sokol-samples/blob/master/sapp/computeboids-sapp.c + - shader: https://github.com/floooh/sokol-samples/blob/master/sapp/computeboids-sapp.glsl + + ...also see the following backend-specific vertex pulling samples (those also don't use sokol-shdc): + + - D3D11: https://github.com/floooh/sokol-samples/blob/master/d3d11/vertexpulling-d3d11.c + - desktop GL: https://github.com/floooh/sokol-samples/blob/master/glfw/vertexpulling-glfw.c + - Metal: https://github.com/floooh/sokol-samples/blob/master/metal/vertexpulling-metal.c + - WebGPU: https://github.com/floooh/sokol-samples/blob/master/wgpu/vertexpulling-wgpu.c + + ...and the backend specific compute shader samples: + + - D3D11: https://github.com/floooh/sokol-samples/blob/master/d3d11/instancing-compute-d3d11.c + - desktop GL: https://github.com/floooh/sokol-samples/blob/master/glfw/instancing-compute-glfw.c + - Metal: https://github.com/floooh/sokol-samples/blob/master/metal/instancing-compute-metal.c + - WebGPU: https://github.com/floooh/sokol-samples/blob/master/wgpu/instancing-compute-wgpu.c + + Storage buffer shader authoring caveats when using sokol-shdc: + + - declare a read-only storage buffer interface block with `layout(binding=N) readonly buffer [name] { ... }` + (where 'N' is the index in `sg_bindings.storage_buffers[N]`) + - ...or a read/write storage buffer interface block with `layout(binding=N) buffer [name] { ... }` + - declare a struct which describes a single array item in the storage buffer interface block + - only put a single flexible array member into the storage buffer interface block + + E.g. a complete example in 'sokol-shdc GLSL': + + ```glsl + @vs + // declare a struct: + struct sb_vertex { + vec3 pos; + vec4 color; + } + // declare a buffer interface block with a single flexible struct array: + layout(binding=0) readonly buffer vertices { + sb_vertex vtx[]; + } + // in the shader function, access the storage buffer like this: + void main() { + vec3 pos = vtx[gl_VertexIndex].pos; + ... + } + @end + ``` + + In a compute shader you can read and write the same item in the same + storage buffer (but you'll have to be careful for random access since + many threads of the same compute function run in parallel): + + @cs + struct sb_item { + vec3 pos; + vec3 vel; + } + layout(binding=0) buffer items_ssbo { + sb_item items[]; + } + layout(local_size_x=64, local_size_y=1, local_size_z=1) in; + void main() { + uint idx = gl_GlobalInvocationID.x; + vec3 pos = items[idx].pos; + ... + items[idx].pos = pos; + } + @end + + Backend-specific storage-buffer caveats (not relevant when using sokol-shdc): + + D3D11: + - storage buffers are created as 'raw' Byte Address Buffers + (https://learn.microsoft.com/en-us/windows/win32/direct3d11/overviews-direct3d-11-resources-intro#raw-views-of-buffers) + - in HLSL, use a ByteAddressBuffer for readonly access of the buffer content: + (https://learn.microsoft.com/en-us/windows/win32/direct3dhlsl/sm5-object-byteaddressbuffer) + - ...or RWByteAddressBuffer for read/write access: + (https://learn.microsoft.com/en-us/windows/win32/direct3dhlsl/sm5-object-rwbyteaddressbuffer) + - readonly-storage buffers and textures are both bound as 'shader-resource-view' and + share the same bind slots (declared as `register(tN)` in HLSL), where N must be in the range 0..23) + - read/write storage buffers are bound as 'unordered-access-view' (declared as `register(uN)` in HLSL + where N is in the range 0..7) + + Metal: + - in Metal there is no internal difference between vertex-, uniform- and + storage-buffers, all are bound to the same 'buffer bind slots' with the + following reserved ranges: + - vertex shader stage: + - uniform buffers: slots 0..7 + - storage buffers: slots 8..15 + - vertex buffers: slots 15..23 + - fragment shader stage: + - uniform buffers: slots 0..7 + - storage buffers: slots 8..15 + - this means in MSL, storage buffer bindings start at [[buffer(8)]] both in + the vertex and fragment stage + + GL: + - the GL backend doesn't use name-lookup to find storage buffer bindings, this + means you must annotate buffers with `layout(std430, binding=N)` in GLSL + - ...where N is 0..7 in the vertex shader, and 8..15 in the fragment shader + + WebGPU: + - in WGSL, textures, samplers and storage buffers all use a shared + bindspace across all shader stages on bindgroup 1: + + `@group(1) @binding(0..127) + + + TRACE HOOKS: + ============ + sokol_gfx.h optionally allows to install "trace hook" callbacks for + each public API functions. When a public API function is called, and + a trace hook callback has been installed for this function, the + callback will be invoked with the parameters and result of the function. + This is useful for things like debugging- and profiling-tools, or + keeping track of resource creation and destruction. + + To use the trace hook feature: + + --- Define SOKOL_TRACE_HOOKS before including the implementation. + + --- Setup an sg_trace_hooks structure with your callback function + pointers (keep all function pointers you're not interested + in zero-initialized), optionally set the user_data member + in the sg_trace_hooks struct. + + --- Install the trace hooks by calling sg_install_trace_hooks(), + the return value of this function is another sg_trace_hooks + struct which contains the previously set of trace hooks. + You should keep this struct around, and call those previous + functions pointers from your own trace callbacks for proper + chaining. + + As an example of how trace hooks are used, have a look at the + imgui/sokol_gfx_imgui.h header which implements a realtime + debugging UI for sokol_gfx.h on top of Dear ImGui. + + + MEMORY ALLOCATION OVERRIDE + ========================== + You can override the memory allocation functions at initialization time + like this: + + void* my_alloc(size_t size, void* user_data) { + return malloc(size); + } + + void my_free(void* ptr, void* user_data) { + free(ptr); + } + + ... + sg_setup(&(sg_desc){ + // ... + .allocator = { + .alloc_fn = my_alloc, + .free_fn = my_free, + .user_data = ..., + } + }); + ... + + If no overrides are provided, malloc and free will be used. + + This only affects memory allocation calls done by sokol_gfx.h + itself though, not any allocations in OS libraries. + + + ERROR REPORTING AND LOGGING + =========================== + To get any logging information at all you need to provide a logging callback in the setup call + the easiest way is to use sokol_log.h: + + #include "sokol_log.h" + + sg_setup(&(sg_desc){ .logger.func = slog_func }); + + To override logging with your own callback, first write a logging function like this: + + void my_log(const char* tag, // e.g. 'sg' + uint32_t log_level, // 0=panic, 1=error, 2=warn, 3=info + uint32_t log_item_id, // SG_LOGITEM_* + const char* message_or_null, // a message string, may be nullptr in release mode + uint32_t line_nr, // line number in sokol_gfx.h + const char* filename_or_null, // source filename, may be nullptr in release mode + void* user_data) + { + ... + } + + ...and then setup sokol-gfx like this: + + sg_setup(&(sg_desc){ + .logger = { + .func = my_log, + .user_data = my_user_data, + } + }); + + The provided logging function must be reentrant (e.g. be callable from + different threads). + + If you don't want to provide your own custom logger it is highly recommended to use + the standard logger in sokol_log.h instead, otherwise you won't see any warnings or + errors. + + + COMMIT LISTENERS + ================ + It's possible to hook callback functions into sokol-gfx which are called from + inside sg_commit() in unspecified order. This is mainly useful for libraries + that build on top of sokol_gfx.h to be notified about the end/start of a frame. + + To add a commit listener, call: + + static void my_commit_listener(void* user_data) { + ... + } + + bool success = sg_add_commit_listener((sg_commit_listener){ + .func = my_commit_listener, + .user_data = ..., + }); + + The function returns false if the internal array of commit listeners is full, + or the same commit listener had already been added. + + If the function returns true, my_commit_listener() will be called each frame + from inside sg_commit(). + + By default, 1024 distinct commit listeners can be added, but this number + can be tweaked in the sg_setup() call: + + sg_setup(&(sg_desc){ + .max_commit_listeners = 2048, + }); + + An sg_commit_listener item is equal to another if both the function + pointer and user_data field are equal. + + To remove a commit listener: + + bool success = sg_remove_commit_listener((sg_commit_listener){ + .func = my_commit_listener, + .user_data = ..., + }); + + ...where the .func and .user_data field are equal to a previous + sg_add_commit_listener() call. The function returns true if the commit + listener item was found and removed, and false otherwise. + + + RESOURCE CREATION AND DESTRUCTION IN DETAIL + =========================================== + The 'vanilla' way to create resource objects is with the 'make functions': + + sg_buffer sg_make_buffer(const sg_buffer_desc* desc) + sg_image sg_make_image(const sg_image_desc* desc) + sg_sampler sg_make_sampler(const sg_sampler_desc* desc) + sg_shader sg_make_shader(const sg_shader_desc* desc) + sg_pipeline sg_make_pipeline(const sg_pipeline_desc* desc) + sg_attachments sg_make_attachments(const sg_attachments_desc* desc) + + This will result in one of three cases: + + 1. The returned handle is invalid. This happens when there are no more + free slots in the resource pool for this resource type. An invalid + handle is associated with the INVALID resource state, for instance: + + sg_buffer buf = sg_make_buffer(...) + if (sg_query_buffer_state(buf) == SG_RESOURCESTATE_INVALID) { + // buffer pool is exhausted + } + + 2. The returned handle is valid, but creating the underlying resource + has failed for some reason. This results in a resource object in the + FAILED state. The reason *why* resource creation has failed differ + by resource type. Look for log messages with more details. A failed + resource state can be checked with: + + sg_buffer buf = sg_make_buffer(...) + if (sg_query_buffer_state(buf) == SG_RESOURCESTATE_FAILED) { + // creating the resource has failed + } + + 3. And finally, if everything goes right, the returned resource is + in resource state VALID and ready to use. This can be checked + with: + + sg_buffer buf = sg_make_buffer(...) + if (sg_query_buffer_state(buf) == SG_RESOURCESTATE_VALID) { + // creating the resource has failed + } + + When calling the 'make functions', the created resource goes through a number + of states: + + - INITIAL: the resource slot associated with the new resource is currently + free (technically, there is no resource yet, just an empty pool slot) + - ALLOC: a handle for the new resource has been allocated, this just means + a pool slot has been reserved. + - VALID or FAILED: in VALID state any 3D API backend resource objects have + been successfully created, otherwise if anything went wrong, the resource + will be in FAILED state. + + Sometimes it makes sense to first grab a handle, but initialize the + underlying resource at a later time. For instance when loading data + asynchronously from a slow data source, you may know what buffers and + textures are needed at an early stage of the loading process, but actually + loading the buffer or texture content can only be completed at a later time. + + For such situations, sokol-gfx resource objects can be created in two steps. + You can allocate a handle upfront with one of the 'alloc functions': + + sg_buffer sg_alloc_buffer(void) + sg_image sg_alloc_image(void) + sg_sampler sg_alloc_sampler(void) + sg_shader sg_alloc_shader(void) + sg_pipeline sg_alloc_pipeline(void) + sg_attachments sg_alloc_attachments(void) + + This will return a handle with the underlying resource object in the + ALLOC state: + + sg_image img = sg_alloc_image(); + if (sg_query_image_state(img) == SG_RESOURCESTATE_ALLOC) { + // allocating an image handle has succeeded, otherwise + // the image pool is full + } + + Such an 'incomplete' handle can be used in most sokol-gfx rendering functions + without doing any harm, sokol-gfx will simply skip any rendering operation + that involve resources which are not in VALID state. + + At a later time (for instance once the texture has completed loading + asynchronously), the resource creation can be completed by calling one of + the 'init functions', those functions take an existing resource handle and + 'desc struct': + + void sg_init_buffer(sg_buffer buf, const sg_buffer_desc* desc) + void sg_init_image(sg_image img, const sg_image_desc* desc) + void sg_init_sampler(sg_sampler smp, const sg_sampler_desc* desc) + void sg_init_shader(sg_shader shd, const sg_shader_desc* desc) + void sg_init_pipeline(sg_pipeline pip, const sg_pipeline_desc* desc) + void sg_init_attachments(sg_attachments atts, const sg_attachments_desc* desc) + + The init functions expect a resource in ALLOC state, and after the function + returns, the resource will be either in VALID or FAILED state. Calling + an 'alloc function' followed by the matching 'init function' is fully + equivalent with calling the 'make function' alone. + + Destruction can also happen as a two-step process. The 'uninit functions' + will put a resource object from the VALID or FAILED state back into the + ALLOC state: + + void sg_uninit_buffer(sg_buffer buf) + void sg_uninit_image(sg_image img) + void sg_uninit_sampler(sg_sampler smp) + void sg_uninit_shader(sg_shader shd) + void sg_uninit_pipeline(sg_pipeline pip) + void sg_uninit_attachments(sg_attachments pass) + + Calling the 'uninit functions' with a resource that is not in the VALID or + FAILED state is a no-op. + + To finally free the pool slot for recycling call the 'dealloc functions': + + void sg_dealloc_buffer(sg_buffer buf) + void sg_dealloc_image(sg_image img) + void sg_dealloc_sampler(sg_sampler smp) + void sg_dealloc_shader(sg_shader shd) + void sg_dealloc_pipeline(sg_pipeline pip) + void sg_dealloc_attachments(sg_attachments atts) + + Calling the 'dealloc functions' on a resource that's not in ALLOC state is + a no-op, but will generate a warning log message. + + Calling an 'uninit function' and 'dealloc function' in sequence is equivalent + with calling the associated 'destroy function': + + void sg_destroy_buffer(sg_buffer buf) + void sg_destroy_image(sg_image img) + void sg_destroy_sampler(sg_sampler smp) + void sg_destroy_shader(sg_shader shd) + void sg_destroy_pipeline(sg_pipeline pip) + void sg_destroy_attachments(sg_attachments atts) + + The 'destroy functions' can be called on resources in any state and generally + do the right thing (for instance if the resource is in ALLOC state, the destroy + function will be equivalent to the 'dealloc function' and skip the 'uninit part'). + + And finally to close the circle, the 'fail functions' can be called to manually + put a resource in ALLOC state into the FAILED state: + + sg_fail_buffer(sg_buffer buf) + sg_fail_image(sg_image img) + sg_fail_sampler(sg_sampler smp) + sg_fail_shader(sg_shader shd) + sg_fail_pipeline(sg_pipeline pip) + sg_fail_attachments(sg_attachments atts) + + This is recommended if anything went wrong outside of sokol-gfx during asynchronous + resource setup (for instance a file loading operation failed). In this case, + the 'fail function' should be called instead of the 'init function'. + + Calling a 'fail function' on a resource that's not in ALLOC state is a no-op, + but will generate a warning log message. + + NOTE: that two-step resource creation usually only makes sense for buffers + and images, but not for samplers, shaders, pipelines or attachments. Most notably, trying + to create a pipeline object with a shader that's not in VALID state will + trigger a validation layer error, or if the validation layer is disabled, + result in a pipeline object in FAILED state. Same when trying to create + an attachments object with invalid image objects. + + + WEBGPU CAVEATS + ============== + For a general overview and design notes of the WebGPU backend see: + + https://floooh.github.io/2023/10/16/sokol-webgpu.html + + In general, don't expect an automatic speedup when switching from the WebGL2 + backend to the WebGPU backend. Some WebGPU functions currently actually + have a higher CPU overhead than similar WebGL2 functions, leading to the + paradoxical situation that some WebGPU code may be slower than similar WebGL2 + code. + + - when writing WGSL shader code by hand, a specific bind-slot convention + must be used: + + All uniform block structs must use `@group(0)` and bindings in the + range 0..15 + + @group(0) @binding(0..15) + + All textures, samplers and storage buffers must use `@group(1)` and + bindings must be in the range 0..127: + + @group(1) @binding(0..127) + + Note that the number of texture, sampler and storage buffer bindings + is still limited despite the large bind range: + + - up to 16 textures and sampler across all shader stages + - up to 8 storage buffers across all shader stages + + If you use sokol-shdc to generate WGSL shader code, you don't need to worry + about the above binding conventions since sokol-shdc. + + - The sokol-gfx WebGPU backend uses the sg_desc.uniform_buffer_size item + to allocate a single per-frame uniform buffer which must be big enough + to hold all data written by sg_apply_uniforms() during a single frame, + including a worst-case 256-byte alignment (e.g. each sg_apply_uniform + call will cost at least 256 bytes of uniform buffer size). The default size + is 4 MB, which is enough for 16384 sg_apply_uniform() calls per + frame (assuming the uniform data 'payload' is less than 256 bytes + per call). These rules are the same as for the Metal backend, so if + you are already using the Metal backend you'll be fine. + + - sg_apply_bindings(): the sokol-gfx WebGPU backend implements a bindgroup + cache to prevent excessive creation and destruction of BindGroup objects + when calling sg_apply_bindings(). The number of slots in the bindgroups + cache is defined in sg_desc.wgpu_bindgroups_cache_size when calling + sg_setup. The cache size must be a power-of-2 number, with the default being + 1024. The bindgroups cache behaviour can be observed by calling the new + function sg_query_frame_stats(), where the following struct items are + of interest: + + .wgpu.num_bindgroup_cache_hits + .wgpu.num_bindgroup_cache_misses + .wgpu.num_bindgroup_cache_collisions + .wgpu_num_bindgroup_cache_invalidates + .wgpu.num_bindgroup_cache_vs_hash_key_mismatch + + The value to pay attention to is `.wgpu.num_bindgroup_cache_collisions`, + if this number is consistently higher than a few percent of the + .wgpu.num_set_bindgroup value, it might be a good idea to bump the + bindgroups cache size to the next power-of-2. + + - sg_apply_viewport(): WebGPU currently has a unique restriction that viewport + rectangles must be contained entirely within the framebuffer. As a shitty + workaround sokol_gfx.h will clip incoming viewport rectangles against + the framebuffer, but this will distort the clipspace-to-screenspace mapping. + There's no proper way to handle this inside sokol_gfx.h, this must be fixed + in a future WebGPU update (see: https://github.com/gpuweb/gpuweb/issues/373 + and https://github.com/gpuweb/gpuweb/pull/5025) + + - The sokol shader compiler generally adds `diagnostic(off, derivative_uniformity);` + into the WGSL output. Currently only the Chrome WebGPU implementation seems + to recognize this. + + - Likewise, the following sokol-gfx pixel formats are not supported in WebGPU: + R16, R16SN, RG16, RG16SN, RGBA16, RGBA16SN. + Unlike unsupported vertex formats, unsupported pixel formats can be queried + in cross-backend code via sg_query_pixel_format() though. + + - The Emscripten WebGPU shim currently doesn't support the Closure minification + post-link-step (e.g. currently the emcc argument '--closure 1' or '--closure 2' + will generate broken Javascript code. + + - sokol-gfx requires the WebGPU device feature `depth32float-stencil8` to be enabled + (this should be widely supported) + + - sokol-gfx expects that the WebGPU device feature `float32-filterable` to *not* be + enabled (since this would exclude all iOS devices) + + + LICENSE + ======= + zlib/libpng license + + Copyright (c) 2018 Andre Weissflog + + This software is provided 'as-is', without any express or implied warranty. + In no event will the authors be held liable for any damages arising from the + use of this software. + + Permission is granted to anyone to use this software for any purpose, + including commercial applications, and to alter it and redistribute it + freely, subject to the following restrictions: + + 1. The origin of this software must not be misrepresented; you must not + claim that you wrote the original software. If you use this software in a + product, an acknowledgment in the product documentation would be + appreciated but is not required. + + 2. Altered source versions must be plainly marked as such, and must not + be misrepresented as being the original software. + + 3. This notice may not be removed or altered from any source + distribution. +*/ +#define SOKOL_GFX_INCLUDED (1) +#include // size_t +#include +#include + +#if defined(SOKOL_API_DECL) && !defined(SOKOL_GFX_API_DECL) +#define SOKOL_GFX_API_DECL SOKOL_API_DECL +#endif +#ifndef SOKOL_GFX_API_DECL +#if defined(_WIN32) && defined(SOKOL_DLL) && defined(SOKOL_GFX_IMPL) +#define SOKOL_GFX_API_DECL __declspec(dllexport) +#elif defined(_WIN32) && defined(SOKOL_DLL) +#define SOKOL_GFX_API_DECL __declspec(dllimport) +#else +#define SOKOL_GFX_API_DECL extern +#endif +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +/* + Resource id typedefs: + + sg_buffer: vertex- and index-buffers + sg_image: images used as textures and render-pass attachments + sg_sampler sampler objects describing how a texture is sampled in a shader + sg_shader: vertex- and fragment-shaders and shader interface information + sg_pipeline: associated shader and vertex-layouts, and render states + sg_attachments: a baked collection of render pass attachment images + + Instead of pointers, resource creation functions return a 32-bit + handle which uniquely identifies the resource object. + + The 32-bit resource id is split into a 16-bit pool index in the lower bits, + and a 16-bit 'generation counter' in the upper bits. The index allows fast + pool lookups, and combined with the generation-counter it allows to detect + 'dangling accesses' (trying to use an object which no longer exists, and + its pool slot has been reused for a new object) + + The resource ids are wrapped into a strongly-typed struct so that + trying to pass an incompatible resource id is a compile error. +*/ +typedef struct sg_buffer { uint32_t id; } sg_buffer; +typedef struct sg_image { uint32_t id; } sg_image; +typedef struct sg_sampler { uint32_t id; } sg_sampler; +typedef struct sg_shader { uint32_t id; } sg_shader; +typedef struct sg_pipeline { uint32_t id; } sg_pipeline; +typedef struct sg_attachments { uint32_t id; } sg_attachments; + +/* + sg_range is a pointer-size-pair struct used to pass memory blobs into + sokol-gfx. When initialized from a value type (array or struct), you can + use the SG_RANGE() macro to build an sg_range struct. For functions which + take either a sg_range pointer, or a (C++) sg_range reference, use the + SG_RANGE_REF macro as a solution which compiles both in C and C++. +*/ +typedef struct sg_range { + const void* ptr; + size_t size; +} sg_range; + +// disabling this for every includer isn't great, but the warnings are also quite pointless +#if defined(_MSC_VER) +#pragma warning(disable:4221) // /W4 only: nonstandard extension used: 'x': cannot be initialized using address of automatic variable 'y' +#pragma warning(disable:4204) // VS2015: nonstandard extension used: non-constant aggregate initializer +#endif +#if defined(__cplusplus) +#define SG_RANGE(x) sg_range{ &x, sizeof(x) } +#define SG_RANGE_REF(x) sg_range{ &x, sizeof(x) } +#else +#define SG_RANGE(x) (sg_range){ &x, sizeof(x) } +#define SG_RANGE_REF(x) &(sg_range){ &x, sizeof(x) } +#endif + +// various compile-time constants in the public API +enum { + SG_INVALID_ID = 0, + SG_NUM_INFLIGHT_FRAMES = 2, + SG_MAX_COLOR_ATTACHMENTS = 4, + SG_MAX_UNIFORMBLOCK_MEMBERS = 16, + SG_MAX_VERTEX_ATTRIBUTES = 16, + SG_MAX_MIPMAPS = 16, + SG_MAX_TEXTUREARRAY_LAYERS = 128, + SG_MAX_UNIFORMBLOCK_BINDSLOTS = 8, + SG_MAX_VERTEXBUFFER_BINDSLOTS = 8, + SG_MAX_IMAGE_BINDSLOTS = 16, + SG_MAX_SAMPLER_BINDSLOTS = 16, + SG_MAX_STORAGEBUFFER_BINDSLOTS = 8, + SG_MAX_IMAGE_SAMPLER_PAIRS = 16, +}; + +/* + sg_color + + An RGBA color value. +*/ +typedef struct sg_color { float r, g, b, a; } sg_color; + +/* + sg_backend + + The active 3D-API backend, use the function sg_query_backend() + to get the currently active backend. +*/ +typedef enum sg_backend { + SG_BACKEND_GLCORE, + SG_BACKEND_GLES3, + SG_BACKEND_D3D11, + SG_BACKEND_METAL_IOS, + SG_BACKEND_METAL_MACOS, + SG_BACKEND_METAL_SIMULATOR, + SG_BACKEND_WGPU, + SG_BACKEND_DUMMY, +} sg_backend; + +/* + sg_pixel_format + + sokol_gfx.h basically uses the same pixel formats as WebGPU, since these + are supported on most newer GPUs. + + A pixelformat name consist of three parts: + + - components (R, RG, RGB or RGBA) + - bit width per component (8, 16 or 32) + - component data type: + - unsigned normalized (no postfix) + - signed normalized (SN postfix) + - unsigned integer (UI postfix) + - signed integer (SI postfix) + - float (F postfix) + + Not all pixel formats can be used for everything, call sg_query_pixelformat() + to inspect the capabilities of a given pixelformat. The function returns + an sg_pixelformat_info struct with the following members: + + - sample: the pixelformat can be sampled as texture at least with + nearest filtering + - filter: the pixelformat can be sampled as texture with linear + filtering + - render: the pixelformat can be used as render-pass attachment + - blend: blending is supported when used as render-pass attachment + - msaa: multisample-antialiasing is supported when used + as render-pass attachment + - depth: the pixelformat can be used for depth-stencil attachments + - compressed: this is a block-compressed format + - bytes_per_pixel: the numbers of bytes in a pixel (0 for compressed formats) + + The default pixel format for texture images is SG_PIXELFORMAT_RGBA8. + + The default pixel format for render target images is platform-dependent + and taken from the sg_environment struct passed into sg_setup(). Typically + the default formats are: + + - for the Metal, D3D11 and WebGPU backends: SG_PIXELFORMAT_BGRA8 + - for GL backends: SG_PIXELFORMAT_RGBA8 +*/ +typedef enum sg_pixel_format { + _SG_PIXELFORMAT_DEFAULT, // value 0 reserved for default-init + SG_PIXELFORMAT_NONE, + + SG_PIXELFORMAT_R8, + SG_PIXELFORMAT_R8SN, + SG_PIXELFORMAT_R8UI, + SG_PIXELFORMAT_R8SI, + + SG_PIXELFORMAT_R16, + SG_PIXELFORMAT_R16SN, + SG_PIXELFORMAT_R16UI, + SG_PIXELFORMAT_R16SI, + SG_PIXELFORMAT_R16F, + SG_PIXELFORMAT_RG8, + SG_PIXELFORMAT_RG8SN, + SG_PIXELFORMAT_RG8UI, + SG_PIXELFORMAT_RG8SI, + + SG_PIXELFORMAT_R32UI, + SG_PIXELFORMAT_R32SI, + SG_PIXELFORMAT_R32F, + SG_PIXELFORMAT_RG16, + SG_PIXELFORMAT_RG16SN, + SG_PIXELFORMAT_RG16UI, + SG_PIXELFORMAT_RG16SI, + SG_PIXELFORMAT_RG16F, + SG_PIXELFORMAT_RGBA8, + SG_PIXELFORMAT_SRGB8A8, + SG_PIXELFORMAT_RGBA8SN, + SG_PIXELFORMAT_RGBA8UI, + SG_PIXELFORMAT_RGBA8SI, + SG_PIXELFORMAT_BGRA8, + SG_PIXELFORMAT_RGB10A2, + SG_PIXELFORMAT_RG11B10F, + SG_PIXELFORMAT_RGB9E5, + + SG_PIXELFORMAT_RG32UI, + SG_PIXELFORMAT_RG32SI, + SG_PIXELFORMAT_RG32F, + SG_PIXELFORMAT_RGBA16, + SG_PIXELFORMAT_RGBA16SN, + SG_PIXELFORMAT_RGBA16UI, + SG_PIXELFORMAT_RGBA16SI, + SG_PIXELFORMAT_RGBA16F, + + SG_PIXELFORMAT_RGBA32UI, + SG_PIXELFORMAT_RGBA32SI, + SG_PIXELFORMAT_RGBA32F, + + // NOTE: when adding/removing pixel formats before DEPTH, also update sokol_app.h/_SAPP_PIXELFORMAT_* + SG_PIXELFORMAT_DEPTH, + SG_PIXELFORMAT_DEPTH_STENCIL, + + // NOTE: don't put any new compressed format in front of here + SG_PIXELFORMAT_BC1_RGBA, + SG_PIXELFORMAT_BC2_RGBA, + SG_PIXELFORMAT_BC3_RGBA, + SG_PIXELFORMAT_BC3_SRGBA, + SG_PIXELFORMAT_BC4_R, + SG_PIXELFORMAT_BC4_RSN, + SG_PIXELFORMAT_BC5_RG, + SG_PIXELFORMAT_BC5_RGSN, + SG_PIXELFORMAT_BC6H_RGBF, + SG_PIXELFORMAT_BC6H_RGBUF, + SG_PIXELFORMAT_BC7_RGBA, + SG_PIXELFORMAT_BC7_SRGBA, + SG_PIXELFORMAT_ETC2_RGB8, + SG_PIXELFORMAT_ETC2_SRGB8, + SG_PIXELFORMAT_ETC2_RGB8A1, + SG_PIXELFORMAT_ETC2_RGBA8, + SG_PIXELFORMAT_ETC2_SRGB8A8, + SG_PIXELFORMAT_EAC_R11, + SG_PIXELFORMAT_EAC_R11SN, + SG_PIXELFORMAT_EAC_RG11, + SG_PIXELFORMAT_EAC_RG11SN, + + SG_PIXELFORMAT_ASTC_4x4_RGBA, + SG_PIXELFORMAT_ASTC_4x4_SRGBA, + + _SG_PIXELFORMAT_NUM, + _SG_PIXELFORMAT_FORCE_U32 = 0x7FFFFFFF +} sg_pixel_format; + +/* + Runtime information about a pixel format, returned by sg_query_pixelformat(). +*/ +typedef struct sg_pixelformat_info { + bool sample; // pixel format can be sampled in shaders at least with nearest filtering + bool filter; // pixel format can be sampled with linear filtering + bool render; // pixel format can be used as render-pass attachment + bool blend; // pixel format supports alpha-blending when used as render-pass attachment + bool msaa; // pixel format supports MSAA when used as render-pass attachment + bool depth; // pixel format is a depth format + bool compressed; // true if this is a hardware-compressed format + int bytes_per_pixel; // NOTE: this is 0 for compressed formats, use sg_query_row_pitch() / sg_query_surface_pitch() as alternative +} sg_pixelformat_info; + +/* + Runtime information about available optional features, returned by sg_query_features() +*/ +typedef struct sg_features { + bool origin_top_left; // framebuffer- and texture-origin is in top left corner + bool image_clamp_to_border; // border color and clamp-to-border uv-wrap mode is supported + bool mrt_independent_blend_state; // multiple-render-target rendering can use per-render-target blend state + bool mrt_independent_write_mask; // multiple-render-target rendering can use per-render-target color write masks + bool compute; // storage buffers and compute shaders are supported + bool msaa_image_bindings; // if true, multisampled images can be bound as texture resources +} sg_features; + +/* + Runtime information about resource limits, returned by sg_query_limit() +*/ +typedef struct sg_limits { + int max_image_size_2d; // max width/height of SG_IMAGETYPE_2D images + int max_image_size_cube; // max width/height of SG_IMAGETYPE_CUBE images + int max_image_size_3d; // max width/height/depth of SG_IMAGETYPE_3D images + int max_image_size_array; // max width/height of SG_IMAGETYPE_ARRAY images + int max_image_array_layers; // max number of layers in SG_IMAGETYPE_ARRAY images + int max_vertex_attrs; // max number of vertex attributes, clamped to SG_MAX_VERTEX_ATTRIBUTES + int gl_max_vertex_uniform_components; // <= GL_MAX_VERTEX_UNIFORM_COMPONENTS (only on GL backends) + int gl_max_combined_texture_image_units; // <= GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS (only on GL backends) +} sg_limits; + +/* + sg_resource_state + + The current state of a resource in its resource pool. + Resources start in the INITIAL state, which means the + pool slot is unoccupied and can be allocated. When a resource is + created, first an id is allocated, and the resource pool slot + is set to state ALLOC. After allocation, the resource is + initialized, which may result in the VALID or FAILED state. The + reason why allocation and initialization are separate is because + some resource types (e.g. buffers and images) might be asynchronously + initialized by the user application. If a resource which is not + in the VALID state is attempted to be used for rendering, rendering + operations will silently be dropped. + + The special INVALID state is returned in sg_query_xxx_state() if no + resource object exists for the provided resource id. +*/ +typedef enum sg_resource_state { + SG_RESOURCESTATE_INITIAL, + SG_RESOURCESTATE_ALLOC, + SG_RESOURCESTATE_VALID, + SG_RESOURCESTATE_FAILED, + SG_RESOURCESTATE_INVALID, + _SG_RESOURCESTATE_FORCE_U32 = 0x7FFFFFFF +} sg_resource_state; + +/* + sg_usage + + A resource usage hint describing the update strategy of + buffers and images. This is used in the sg_buffer_desc.usage + and sg_image_desc.usage members when creating buffers + and images: + + SG_USAGE_IMMUTABLE: the resource will never be updated with + new (CPU-side) data, instead the content of the + resource must be provided on creation + SG_USAGE_DYNAMIC: the resource will be updated infrequently + with new data (this could range from "once + after creation", to "quite often but not + every frame") + SG_USAGE_STREAM: the resource will be updated each frame + with new content + + The rendering backends use this hint to prevent that the + CPU needs to wait for the GPU when attempting to update + a resource that might be currently accessed by the GPU. + + Resource content is updated with the functions sg_update_buffer() or + sg_append_buffer() for buffer objects, and sg_update_image() for image + objects. For the sg_update_*() functions, only one update is allowed per + frame and resource object, while sg_append_buffer() can be called + multiple times per frame on the same buffer. The application must update + all data required for rendering (this means that the update data can be + smaller than the resource size, if only a part of the overall resource + size is used for rendering, you only need to make sure that the data that + *is* used is valid). + + The default usage is SG_USAGE_IMMUTABLE. +*/ +typedef enum sg_usage { + _SG_USAGE_DEFAULT, // value 0 reserved for default-init + SG_USAGE_IMMUTABLE, + SG_USAGE_DYNAMIC, + SG_USAGE_STREAM, + _SG_USAGE_NUM, + _SG_USAGE_FORCE_U32 = 0x7FFFFFFF +} sg_usage; + +/* + sg_buffer_type + + Indicates whether a buffer will be bound as vertex-, + index- or storage-buffer. + + Used in the sg_buffer_desc.type member when creating a buffer. + + The default value is SG_BUFFERTYPE_VERTEXBUFFER. +*/ +typedef enum sg_buffer_type { + _SG_BUFFERTYPE_DEFAULT, // value 0 reserved for default-init + SG_BUFFERTYPE_VERTEXBUFFER, + SG_BUFFERTYPE_INDEXBUFFER, + SG_BUFFERTYPE_STORAGEBUFFER, + _SG_BUFFERTYPE_NUM, + _SG_BUFFERTYPE_FORCE_U32 = 0x7FFFFFFF +} sg_buffer_type; + +/* + sg_index_type + + Indicates whether indexed rendering (fetching vertex-indices from an + index buffer) is used, and if yes, the index data type (16- or 32-bits). + + This is used in the sg_pipeline_desc.index_type member when creating a + pipeline object. + + The default index type is SG_INDEXTYPE_NONE. +*/ +typedef enum sg_index_type { + _SG_INDEXTYPE_DEFAULT, // value 0 reserved for default-init + SG_INDEXTYPE_NONE, + SG_INDEXTYPE_UINT16, + SG_INDEXTYPE_UINT32, + _SG_INDEXTYPE_NUM, + _SG_INDEXTYPE_FORCE_U32 = 0x7FFFFFFF +} sg_index_type; + +/* + sg_image_type + + Indicates the basic type of an image object (2D-texture, cubemap, + 3D-texture or 2D-array-texture). Used in the sg_image_desc.type member when + creating an image, and in sg_shader_image_desc to describe a sampled texture + in the shader (both must match and will be checked in the validation layer + when calling sg_apply_bindings). + + The default image type when creating an image is SG_IMAGETYPE_2D. +*/ +typedef enum sg_image_type { + _SG_IMAGETYPE_DEFAULT, // value 0 reserved for default-init + SG_IMAGETYPE_2D, + SG_IMAGETYPE_CUBE, + SG_IMAGETYPE_3D, + SG_IMAGETYPE_ARRAY, + _SG_IMAGETYPE_NUM, + _SG_IMAGETYPE_FORCE_U32 = 0x7FFFFFFF +} sg_image_type; + +/* + sg_image_sample_type + + The basic data type of a texture sample as expected by a shader. + Must be provided in sg_shader_image and used by the validation + layer in sg_apply_bindings() to check if the provided image object + is compatible with what the shader expects. Apart from the sokol-gfx + validation layer, WebGPU is the only backend API which actually requires + matching texture and sampler type to be provided upfront for validation + (other 3D APIs treat texture/sampler type mismatches as undefined behaviour). + + NOTE that the following texture pixel formats require the use + of SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT, combined with a sampler + of type SG_SAMPLERTYPE_NONFILTERING: + + - SG_PIXELFORMAT_R32F + - SG_PIXELFORMAT_RG32F + - SG_PIXELFORMAT_RGBA32F + + (when using sokol-shdc, also check out the meta tags `@image_sample_type` + and `@sampler_type`) +*/ +typedef enum sg_image_sample_type { + _SG_IMAGESAMPLETYPE_DEFAULT, // value 0 reserved for default-init + SG_IMAGESAMPLETYPE_FLOAT, + SG_IMAGESAMPLETYPE_DEPTH, + SG_IMAGESAMPLETYPE_SINT, + SG_IMAGESAMPLETYPE_UINT, + SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT, + _SG_IMAGESAMPLETYPE_NUM, + _SG_IMAGESAMPLETYPE_FORCE_U32 = 0x7FFFFFFF +} sg_image_sample_type; + +/* + sg_sampler_type + + The basic type of a texture sampler (sampling vs comparison) as + defined in a shader. Must be provided in sg_shader_sampler_desc. + + sg_image_sample_type and sg_sampler_type for a texture/sampler + pair must be compatible with each other, specifically only + the following pairs are allowed: + + - SG_IMAGESAMPLETYPE_FLOAT => (SG_SAMPLERTYPE_FILTERING or SG_SAMPLERTYPE_NONFILTERING) + - SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT => SG_SAMPLERTYPE_NONFILTERING + - SG_IMAGESAMPLETYPE_SINT => SG_SAMPLERTYPE_NONFILTERING + - SG_IMAGESAMPLETYPE_UINT => SG_SAMPLERTYPE_NONFILTERING + - SG_IMAGESAMPLETYPE_DEPTH => SG_SAMPLERTYPE_COMPARISON +*/ +typedef enum sg_sampler_type { + _SG_SAMPLERTYPE_DEFAULT, + SG_SAMPLERTYPE_FILTERING, + SG_SAMPLERTYPE_NONFILTERING, + SG_SAMPLERTYPE_COMPARISON, + _SG_SAMPLERTYPE_NUM, + _SG_SAMPLERTYPE_FORCE_U32, +} sg_sampler_type; + +/* + sg_cube_face + + The cubemap faces. Use these as indices in the sg_image_desc.content + array. +*/ +typedef enum sg_cube_face { + SG_CUBEFACE_POS_X, + SG_CUBEFACE_NEG_X, + SG_CUBEFACE_POS_Y, + SG_CUBEFACE_NEG_Y, + SG_CUBEFACE_POS_Z, + SG_CUBEFACE_NEG_Z, + SG_CUBEFACE_NUM, + _SG_CUBEFACE_FORCE_U32 = 0x7FFFFFFF +} sg_cube_face; + +/* + sg_primitive_type + + This is the common subset of 3D primitive types supported across all 3D + APIs. This is used in the sg_pipeline_desc.primitive_type member when + creating a pipeline object. + + The default primitive type is SG_PRIMITIVETYPE_TRIANGLES. +*/ +typedef enum sg_primitive_type { + _SG_PRIMITIVETYPE_DEFAULT, // value 0 reserved for default-init + SG_PRIMITIVETYPE_POINTS, + SG_PRIMITIVETYPE_LINES, + SG_PRIMITIVETYPE_LINE_STRIP, + SG_PRIMITIVETYPE_TRIANGLES, + SG_PRIMITIVETYPE_TRIANGLE_STRIP, + _SG_PRIMITIVETYPE_NUM, + _SG_PRIMITIVETYPE_FORCE_U32 = 0x7FFFFFFF +} sg_primitive_type; + +/* + sg_filter + + The filtering mode when sampling a texture image. This is + used in the sg_sampler_desc.min_filter, sg_sampler_desc.mag_filter + and sg_sampler_desc.mipmap_filter members when creating a sampler object. + + For the default is SG_FILTER_NEAREST. +*/ +typedef enum sg_filter { + _SG_FILTER_DEFAULT, // value 0 reserved for default-init + SG_FILTER_NEAREST, + SG_FILTER_LINEAR, + _SG_FILTER_NUM, + _SG_FILTER_FORCE_U32 = 0x7FFFFFFF +} sg_filter; + +/* + sg_wrap + + The texture coordinates wrapping mode when sampling a texture + image. This is used in the sg_image_desc.wrap_u, .wrap_v + and .wrap_w members when creating an image. + + The default wrap mode is SG_WRAP_REPEAT. + + NOTE: SG_WRAP_CLAMP_TO_BORDER is not supported on all backends + and platforms. To check for support, call sg_query_features() + and check the "clamp_to_border" boolean in the returned + sg_features struct. + + Platforms which don't support SG_WRAP_CLAMP_TO_BORDER will silently fall back + to SG_WRAP_CLAMP_TO_EDGE without a validation error. +*/ +typedef enum sg_wrap { + _SG_WRAP_DEFAULT, // value 0 reserved for default-init + SG_WRAP_REPEAT, + SG_WRAP_CLAMP_TO_EDGE, + SG_WRAP_CLAMP_TO_BORDER, + SG_WRAP_MIRRORED_REPEAT, + _SG_WRAP_NUM, + _SG_WRAP_FORCE_U32 = 0x7FFFFFFF +} sg_wrap; + +/* + sg_border_color + + The border color to use when sampling a texture, and the UV wrap + mode is SG_WRAP_CLAMP_TO_BORDER. + + The default border color is SG_BORDERCOLOR_OPAQUE_BLACK +*/ +typedef enum sg_border_color { + _SG_BORDERCOLOR_DEFAULT, // value 0 reserved for default-init + SG_BORDERCOLOR_TRANSPARENT_BLACK, + SG_BORDERCOLOR_OPAQUE_BLACK, + SG_BORDERCOLOR_OPAQUE_WHITE, + _SG_BORDERCOLOR_NUM, + _SG_BORDERCOLOR_FORCE_U32 = 0x7FFFFFFF +} sg_border_color; + +/* + sg_vertex_format + + The data type of a vertex component. This is used to describe + the layout of input vertex data when creating a pipeline object. + + NOTE that specific mapping rules exist from the CPU-side vertex + formats to the vertex attribute base type in the vertex shader code + (see doc header section 'ON VERTEX FORMATS'). +*/ +typedef enum sg_vertex_format { + SG_VERTEXFORMAT_INVALID, + SG_VERTEXFORMAT_FLOAT, + SG_VERTEXFORMAT_FLOAT2, + SG_VERTEXFORMAT_FLOAT3, + SG_VERTEXFORMAT_FLOAT4, + SG_VERTEXFORMAT_INT, + SG_VERTEXFORMAT_INT2, + SG_VERTEXFORMAT_INT3, + SG_VERTEXFORMAT_INT4, + SG_VERTEXFORMAT_UINT, + SG_VERTEXFORMAT_UINT2, + SG_VERTEXFORMAT_UINT3, + SG_VERTEXFORMAT_UINT4, + SG_VERTEXFORMAT_BYTE4, + SG_VERTEXFORMAT_BYTE4N, + SG_VERTEXFORMAT_UBYTE4, + SG_VERTEXFORMAT_UBYTE4N, + SG_VERTEXFORMAT_SHORT2, + SG_VERTEXFORMAT_SHORT2N, + SG_VERTEXFORMAT_USHORT2, + SG_VERTEXFORMAT_USHORT2N, + SG_VERTEXFORMAT_SHORT4, + SG_VERTEXFORMAT_SHORT4N, + SG_VERTEXFORMAT_USHORT4, + SG_VERTEXFORMAT_USHORT4N, + SG_VERTEXFORMAT_UINT10_N2, + SG_VERTEXFORMAT_HALF2, + SG_VERTEXFORMAT_HALF4, + _SG_VERTEXFORMAT_NUM, + _SG_VERTEXFORMAT_FORCE_U32 = 0x7FFFFFFF +} sg_vertex_format; + +/* + sg_vertex_step + + Defines whether the input pointer of a vertex input stream is advanced + 'per vertex' or 'per instance'. The default step-func is + SG_VERTEXSTEP_PER_VERTEX. SG_VERTEXSTEP_PER_INSTANCE is used with + instanced-rendering. + + The vertex-step is part of the vertex-layout definition + when creating pipeline objects. +*/ +typedef enum sg_vertex_step { + _SG_VERTEXSTEP_DEFAULT, // value 0 reserved for default-init + SG_VERTEXSTEP_PER_VERTEX, + SG_VERTEXSTEP_PER_INSTANCE, + _SG_VERTEXSTEP_NUM, + _SG_VERTEXSTEP_FORCE_U32 = 0x7FFFFFFF +} sg_vertex_step; + +/* + sg_uniform_type + + The data type of a uniform block member. This is used to + describe the internal layout of uniform blocks when creating + a shader object. This is only required for the GL backend, all + other backends will ignore the interior layout of uniform blocks. +*/ +typedef enum sg_uniform_type { + SG_UNIFORMTYPE_INVALID, + SG_UNIFORMTYPE_FLOAT, + SG_UNIFORMTYPE_FLOAT2, + SG_UNIFORMTYPE_FLOAT3, + SG_UNIFORMTYPE_FLOAT4, + SG_UNIFORMTYPE_INT, + SG_UNIFORMTYPE_INT2, + SG_UNIFORMTYPE_INT3, + SG_UNIFORMTYPE_INT4, + SG_UNIFORMTYPE_MAT4, + _SG_UNIFORMTYPE_NUM, + _SG_UNIFORMTYPE_FORCE_U32 = 0x7FFFFFFF +} sg_uniform_type; + +/* + sg_uniform_layout + + A hint for the interior memory layout of uniform blocks. This is + only relevant for the GL backend where the internal layout + of uniform blocks must be known to sokol-gfx. For all other backends the + internal memory layout of uniform blocks doesn't matter, sokol-gfx + will just pass uniform data as an opaque memory blob to the + 3D backend. + + SG_UNIFORMLAYOUT_NATIVE (default) + Native layout means that a 'backend-native' memory layout + is used. For the GL backend this means that uniforms + are packed tightly in memory (e.g. there are no padding + bytes). + + SG_UNIFORMLAYOUT_STD140 + The memory layout is a subset of std140. Arrays are only + allowed for the FLOAT4, INT4 and MAT4. Alignment is as + is as follows: + + FLOAT, INT: 4 byte alignment + FLOAT2, INT2: 8 byte alignment + FLOAT3, INT3: 16 byte alignment(!) + FLOAT4, INT4: 16 byte alignment + MAT4: 16 byte alignment + FLOAT4[], INT4[]: 16 byte alignment + + The overall size of the uniform block must be a multiple + of 16. + + For more information search for 'UNIFORM DATA LAYOUT' in the documentation block + at the start of the header. +*/ +typedef enum sg_uniform_layout { + _SG_UNIFORMLAYOUT_DEFAULT, // value 0 reserved for default-init + SG_UNIFORMLAYOUT_NATIVE, // default: layout depends on currently active backend + SG_UNIFORMLAYOUT_STD140, // std140: memory layout according to std140 + _SG_UNIFORMLAYOUT_NUM, + _SG_UNIFORMLAYOUT_FORCE_U32 = 0x7FFFFFFF +} sg_uniform_layout; + +/* + sg_cull_mode + + The face-culling mode, this is used in the + sg_pipeline_desc.cull_mode member when creating a + pipeline object. + + The default cull mode is SG_CULLMODE_NONE +*/ +typedef enum sg_cull_mode { + _SG_CULLMODE_DEFAULT, // value 0 reserved for default-init + SG_CULLMODE_NONE, + SG_CULLMODE_FRONT, + SG_CULLMODE_BACK, + _SG_CULLMODE_NUM, + _SG_CULLMODE_FORCE_U32 = 0x7FFFFFFF +} sg_cull_mode; + +/* + sg_face_winding + + The vertex-winding rule that determines a front-facing primitive. This + is used in the member sg_pipeline_desc.face_winding + when creating a pipeline object. + + The default winding is SG_FACEWINDING_CW (clockwise) +*/ +typedef enum sg_face_winding { + _SG_FACEWINDING_DEFAULT, // value 0 reserved for default-init + SG_FACEWINDING_CCW, + SG_FACEWINDING_CW, + _SG_FACEWINDING_NUM, + _SG_FACEWINDING_FORCE_U32 = 0x7FFFFFFF +} sg_face_winding; + +/* + sg_compare_func + + The compare-function for configuring depth- and stencil-ref tests + in pipeline objects, and for texture samplers which perform a comparison + instead of regular sampling operation. + + Used in the following structs: + + sg_pipeline_desc + .depth + .compare + .stencil + .front.compare + .back.compare + + sg_sampler_desc + .compare + + The default compare func for depth- and stencil-tests is + SG_COMPAREFUNC_ALWAYS. + + The default compare func for samplers is SG_COMPAREFUNC_NEVER. +*/ +typedef enum sg_compare_func { + _SG_COMPAREFUNC_DEFAULT, // value 0 reserved for default-init + SG_COMPAREFUNC_NEVER, + SG_COMPAREFUNC_LESS, + SG_COMPAREFUNC_EQUAL, + SG_COMPAREFUNC_LESS_EQUAL, + SG_COMPAREFUNC_GREATER, + SG_COMPAREFUNC_NOT_EQUAL, + SG_COMPAREFUNC_GREATER_EQUAL, + SG_COMPAREFUNC_ALWAYS, + _SG_COMPAREFUNC_NUM, + _SG_COMPAREFUNC_FORCE_U32 = 0x7FFFFFFF +} sg_compare_func; + +/* + sg_stencil_op + + The operation performed on a currently stored stencil-value when a + comparison test passes or fails. This is used when creating a pipeline + object in the following sg_pipeline_desc struct items: + + sg_pipeline_desc + .stencil + .front + .fail_op + .depth_fail_op + .pass_op + .back + .fail_op + .depth_fail_op + .pass_op + + The default value is SG_STENCILOP_KEEP. +*/ +typedef enum sg_stencil_op { + _SG_STENCILOP_DEFAULT, // value 0 reserved for default-init + SG_STENCILOP_KEEP, + SG_STENCILOP_ZERO, + SG_STENCILOP_REPLACE, + SG_STENCILOP_INCR_CLAMP, + SG_STENCILOP_DECR_CLAMP, + SG_STENCILOP_INVERT, + SG_STENCILOP_INCR_WRAP, + SG_STENCILOP_DECR_WRAP, + _SG_STENCILOP_NUM, + _SG_STENCILOP_FORCE_U32 = 0x7FFFFFFF +} sg_stencil_op; + +/* + sg_blend_factor + + The source and destination factors in blending operations. + This is used in the following members when creating a pipeline object: + + sg_pipeline_desc + .colors[i] + .blend + .src_factor_rgb + .dst_factor_rgb + .src_factor_alpha + .dst_factor_alpha + + The default value is SG_BLENDFACTOR_ONE for source + factors, and for the destination SG_BLENDFACTOR_ZERO if the associated + blend-op is ADD, SUBTRACT or REVERSE_SUBTRACT or SG_BLENDFACTOR_ONE + if the associated blend-op is MIN or MAX. +*/ +typedef enum sg_blend_factor { + _SG_BLENDFACTOR_DEFAULT, // value 0 reserved for default-init + SG_BLENDFACTOR_ZERO, + SG_BLENDFACTOR_ONE, + SG_BLENDFACTOR_SRC_COLOR, + SG_BLENDFACTOR_ONE_MINUS_SRC_COLOR, + SG_BLENDFACTOR_SRC_ALPHA, + SG_BLENDFACTOR_ONE_MINUS_SRC_ALPHA, + SG_BLENDFACTOR_DST_COLOR, + SG_BLENDFACTOR_ONE_MINUS_DST_COLOR, + SG_BLENDFACTOR_DST_ALPHA, + SG_BLENDFACTOR_ONE_MINUS_DST_ALPHA, + SG_BLENDFACTOR_SRC_ALPHA_SATURATED, + SG_BLENDFACTOR_BLEND_COLOR, + SG_BLENDFACTOR_ONE_MINUS_BLEND_COLOR, + SG_BLENDFACTOR_BLEND_ALPHA, + SG_BLENDFACTOR_ONE_MINUS_BLEND_ALPHA, + _SG_BLENDFACTOR_NUM, + _SG_BLENDFACTOR_FORCE_U32 = 0x7FFFFFFF +} sg_blend_factor; + +/* + sg_blend_op + + Describes how the source and destination values are combined in the + fragment blending operation. It is used in the following struct items + when creating a pipeline object: + + sg_pipeline_desc + .colors[i] + .blend + .op_rgb + .op_alpha + + The default value is SG_BLENDOP_ADD. +*/ +typedef enum sg_blend_op { + _SG_BLENDOP_DEFAULT, // value 0 reserved for default-init + SG_BLENDOP_ADD, + SG_BLENDOP_SUBTRACT, + SG_BLENDOP_REVERSE_SUBTRACT, + SG_BLENDOP_MIN, + SG_BLENDOP_MAX, + _SG_BLENDOP_NUM, + _SG_BLENDOP_FORCE_U32 = 0x7FFFFFFF +} sg_blend_op; + +/* + sg_color_mask + + Selects the active color channels when writing a fragment color to the + framebuffer. This is used in the members + sg_pipeline_desc.colors[i].write_mask when creating a pipeline object. + + The default colormask is SG_COLORMASK_RGBA (write all colors channels) + + NOTE: since the color mask value 0 is reserved for the default value + (SG_COLORMASK_RGBA), use SG_COLORMASK_NONE if all color channels + should be disabled. +*/ +typedef enum sg_color_mask { + _SG_COLORMASK_DEFAULT = 0, // value 0 reserved for default-init + SG_COLORMASK_NONE = 0x10, // special value for 'all channels disabled + SG_COLORMASK_R = 0x1, + SG_COLORMASK_G = 0x2, + SG_COLORMASK_RG = 0x3, + SG_COLORMASK_B = 0x4, + SG_COLORMASK_RB = 0x5, + SG_COLORMASK_GB = 0x6, + SG_COLORMASK_RGB = 0x7, + SG_COLORMASK_A = 0x8, + SG_COLORMASK_RA = 0x9, + SG_COLORMASK_GA = 0xA, + SG_COLORMASK_RGA = 0xB, + SG_COLORMASK_BA = 0xC, + SG_COLORMASK_RBA = 0xD, + SG_COLORMASK_GBA = 0xE, + SG_COLORMASK_RGBA = 0xF, + _SG_COLORMASK_FORCE_U32 = 0x7FFFFFFF +} sg_color_mask; + +/* + sg_load_action + + Defines the load action that should be performed at the start of a render pass: + + SG_LOADACTION_CLEAR: clear the render target + SG_LOADACTION_LOAD: load the previous content of the render target + SG_LOADACTION_DONTCARE: leave the render target in an undefined state + + This is used in the sg_pass_action structure. + + The default load action for all pass attachments is SG_LOADACTION_CLEAR, + with the values rgba = { 0.5f, 0.5f, 0.5f, 1.0f }, depth=1.0f and stencil=0. + + If you want to override the default behaviour, it is important to not + only set the clear color, but the 'action' field as well (as long as this + is _SG_LOADACTION_DEFAULT, the value fields will be ignored). +*/ +typedef enum sg_load_action { + _SG_LOADACTION_DEFAULT, + SG_LOADACTION_CLEAR, + SG_LOADACTION_LOAD, + SG_LOADACTION_DONTCARE, + _SG_LOADACTION_FORCE_U32 = 0x7FFFFFFF +} sg_load_action; + +/* + sg_store_action + + Defines the store action that should be performed at the end of a render pass: + + SG_STOREACTION_STORE: store the rendered content to the color attachment image + SG_STOREACTION_DONTCARE: allows the GPU to discard the rendered content +*/ +typedef enum sg_store_action { + _SG_STOREACTION_DEFAULT, + SG_STOREACTION_STORE, + SG_STOREACTION_DONTCARE, + _SG_STOREACTION_FORCE_U32 = 0x7FFFFFFF +} sg_store_action; + + +/* + sg_pass_action + + The sg_pass_action struct defines the actions to be performed + at the start and end of a render pass. + + - at the start of the pass: whether the render attachments should be cleared, + loaded with their previous content, or start in an undefined state + - for clear operations: the clear value (color, depth, or stencil values) + - at the end of the pass: whether the rendering result should be + stored back into the render attachment or discarded +*/ +typedef struct sg_color_attachment_action { + sg_load_action load_action; // default: SG_LOADACTION_CLEAR + sg_store_action store_action; // default: SG_STOREACTION_STORE + sg_color clear_value; // default: { 0.5f, 0.5f, 0.5f, 1.0f } +} sg_color_attachment_action; + +typedef struct sg_depth_attachment_action { + sg_load_action load_action; // default: SG_LOADACTION_CLEAR + sg_store_action store_action; // default: SG_STOREACTION_DONTCARE + float clear_value; // default: 1.0 +} sg_depth_attachment_action; + +typedef struct sg_stencil_attachment_action { + sg_load_action load_action; // default: SG_LOADACTION_CLEAR + sg_store_action store_action; // default: SG_STOREACTION_DONTCARE + uint8_t clear_value; // default: 0 +} sg_stencil_attachment_action; + +typedef struct sg_pass_action { + sg_color_attachment_action colors[SG_MAX_COLOR_ATTACHMENTS]; + sg_depth_attachment_action depth; + sg_stencil_attachment_action stencil; +} sg_pass_action; + +/* + sg_swapchain + + Used in sg_begin_pass() to provide details about an external swapchain + (pixel formats, sample count and backend-API specific render surface objects). + + The following information must be provided: + + - the width and height of the swapchain surfaces in number of pixels, + - the pixel format of the render- and optional msaa-resolve-surface + - the pixel format of the optional depth- or depth-stencil-surface + - the MSAA sample count for the render and depth-stencil surface + + If the pixel formats and MSAA sample counts are left zero-initialized, + their defaults are taken from the sg_environment struct provided in the + sg_setup() call. + + The width and height *must* be > 0. + + Additionally the following backend API specific objects must be passed in + as 'type erased' void pointers: + + GL: + - on all GL backends, a GL framebuffer object must be provided. This + can be zero for the default framebuffer. + + D3D11: + - an ID3D11RenderTargetView for the rendering surface, without + MSAA rendering this surface will also be displayed + - an optional ID3D11DepthStencilView for the depth- or depth/stencil + buffer surface + - when MSAA rendering is used, another ID3D11RenderTargetView + which serves as MSAA resolve target and will be displayed + + WebGPU (same as D3D11, except different types) + - a WGPUTextureView for the rendering surface, without + MSAA rendering this surface will also be displayed + - an optional WGPUTextureView for the depth- or depth/stencil + buffer surface + - when MSAA rendering is used, another WGPUTextureView + which serves as MSAA resolve target and will be displayed + + Metal (NOTE that the roles of provided surfaces is slightly different + than on D3D11 or WebGPU in case of MSAA vs non-MSAA rendering): + + - A current CAMetalDrawable (NOT an MTLDrawable!) which will be presented. + This will either be rendered to directly (if no MSAA is used), or serve + as MSAA-resolve target. + - an optional MTLTexture for the depth- or depth-stencil buffer + - an optional multisampled MTLTexture which serves as intermediate + rendering surface which will then be resolved into the + CAMetalDrawable. + + NOTE that for Metal you must use an ObjC __bridge cast to + properly tunnel the ObjC object id through a C void*, e.g.: + + swapchain.metal.current_drawable = (__bridge const void*) [mtkView currentDrawable]; + + On all other backends you shouldn't need to mess with the reference count. + + It's a good practice to write a helper function which returns an initialized + sg_swapchain structs, which can then be plugged directly into + sg_pass.swapchain. Look at the function sglue_swapchain() in the sokol_glue.h + as an example. +*/ +typedef struct sg_metal_swapchain { + const void* current_drawable; // CAMetalDrawable (NOT MTLDrawable!!!) + const void* depth_stencil_texture; // MTLTexture + const void* msaa_color_texture; // MTLTexture +} sg_metal_swapchain; + +typedef struct sg_d3d11_swapchain { + const void* render_view; // ID3D11RenderTargetView + const void* resolve_view; // ID3D11RenderTargetView + const void* depth_stencil_view; // ID3D11DepthStencilView +} sg_d3d11_swapchain; + +typedef struct sg_wgpu_swapchain { + const void* render_view; // WGPUTextureView + const void* resolve_view; // WGPUTextureView + const void* depth_stencil_view; // WGPUTextureView +} sg_wgpu_swapchain; + +typedef struct sg_gl_swapchain { + uint32_t framebuffer; // GL framebuffer object +} sg_gl_swapchain; + +typedef struct sg_swapchain { + int width; + int height; + int sample_count; + sg_pixel_format color_format; + sg_pixel_format depth_format; + sg_metal_swapchain metal; + sg_d3d11_swapchain d3d11; + sg_wgpu_swapchain wgpu; + sg_gl_swapchain gl; +} sg_swapchain; + +/* + sg_pass + + The sg_pass structure is passed as argument into the sg_begin_pass() + function. + + For a swapchain render pass, provide an sg_pass_action and sg_swapchain + struct (for instance via the sglue_swapchain() helper function from + sokol_glue.h): + + sg_begin_pass(&(sg_pass){ + .action = { ... }, + .swapchain = sglue_swapchain(), + }); + + For an offscreen render pass, provide an sg_pass_action struct and + an sg_attachments handle: + + sg_begin_pass(&(sg_pass){ + .action = { ... }, + .attachments = attachments, + }); + + You can also omit the .action object to get default pass action behaviour + (clear to color=grey, depth=1 and stencil=0). + + For a compute pass, just set the sg_pass.compute boolean to true: + + sg_begin_pass(&(sg_pass){ .compute = true }); +*/ +typedef struct sg_pass { + uint32_t _start_canary; + bool compute; + sg_pass_action action; + sg_attachments attachments; + sg_swapchain swapchain; + const char* label; + uint32_t _end_canary; +} sg_pass; + +/* + sg_bindings + + The sg_bindings structure defines the buffers, images and + samplers resource bindings for the next draw call. + + To update the resource bindings, call sg_apply_bindings() with + a pointer to a populated sg_bindings struct. Note that + sg_apply_bindings() must be called after sg_apply_pipeline() + and that bindings are not preserved across sg_apply_pipeline() + calls, even when the new pipeline uses the same 'bindings layout'. + + A resource binding struct contains: + + - 1..N vertex buffers + - 0..N vertex buffer offsets + - 0..1 index buffers + - 0..1 index buffer offsets + - 0..N images + - 0..N samplers + - 0..N storage buffers + + Where 'N' is defined in the following constants: + + - SG_MAX_VERTEXBUFFER_BINDSLOTS + - SG_MAX_IMAGE_BINDLOTS + - SG_MAX_SAMPLER_BINDSLOTS + - SG_MAX_STORAGEBUFFER_BINDGLOTS + + Note that inside compute passes vertex- and index-buffer-bindings are + disallowed. + + When using sokol-shdc for shader authoring, the `layout(binding=N)` + annotation in the shader code directly maps to the slot index for that + resource type in the bindings struct, for instance the following vertex- + and fragment-shader interface for sokol-shdc: + + @vs vs + layout(binding=0) uniform vs_params { ... }; + layout(binding=0) readonly buffer ssbo { ... }; + layout(binding=0) uniform texture2D vs_tex; + layout(binding=0) uniform sampler vs_smp; + ... + @end + + @fs fs + layout(binding=1) uniform fs_params { ... }; + layout(binding=1) uniform texture2D fs_tex; + layout(binding=1) uniform sampler fs_smp; + ... + @end + + ...would map to the following sg_bindings struct: + + const sg_bindings bnd = { + .vertex_buffers[0] = ..., + .images[0] = vs_tex, + .images[1] = fs_tex, + .samplers[0] = vs_smp, + .samplers[1] = fs_smp, + .storage_buffers[0] = ssbo, + }; + + ...alternatively you can use code-generated slot indices: + + const sg_bindings bnd = { + .vertex_buffers[0] = ..., + .images[IMG_vs_tex] = vs_tex, + .images[IMG_fs_tex] = fs_tex, + .samplers[SMP_vs_smp] = vs_smp, + .samplers[SMP_fs_smp] = fs_smp, + .storage_buffers[SBUF_ssbo] = ssbo, + }; + + Resource bindslots for a specific shader/pipeline may have gaps, and an + sg_bindings struct may have populated bind slots which are not used by a + specific shader. This allows to use the same sg_bindings struct across + different shader variants. + + When not using sokol-shdc, the bindslot indices in the sg_bindings + struct need to match the per-resource reflection info slot indices + in the sg_shader_desc struct (for details about that see the + sg_shader_desc struct documentation). + + The optional buffer offsets can be used to put different unrelated + chunks of vertex- and/or index-data into the same buffer objects. +*/ +typedef struct sg_bindings { + uint32_t _start_canary; + sg_buffer vertex_buffers[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + int vertex_buffer_offsets[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + sg_buffer index_buffer; + int index_buffer_offset; + sg_image images[SG_MAX_IMAGE_BINDSLOTS]; + sg_sampler samplers[SG_MAX_SAMPLER_BINDSLOTS]; + sg_buffer storage_buffers[SG_MAX_STORAGEBUFFER_BINDSLOTS]; + uint32_t _end_canary; +} sg_bindings; + +/* + sg_buffer_desc + + Creation parameters for sg_buffer objects, used in the + sg_make_buffer() call. + + The default configuration is: + + .size: 0 (*must* be >0 for buffers without data) + .type: SG_BUFFERTYPE_VERTEXBUFFER + .usage: SG_USAGE_IMMUTABLE + .data.ptr 0 (*must* be valid for immutable buffers) + .data.size 0 (*must* be > 0 for immutable buffers) + .label 0 (optional string label) + + For immutable buffers which are initialized with initial data, + keep the .size item zero-initialized, and set the size together with the + pointer to the initial data in the .data item. + + For immutable or mutable buffers without initial data, keep the .data item + zero-initialized, and set the buffer size in the .size item instead. + + NOTE: Immutable buffers without initial data are guaranteed to be + zero-initialized. For mutable (dynamic or streaming) buffers, the + initial content is undefined. + + You can also set both size values, but currently both size values must + be identical (this may change in the future when the dynamic resource + management may become more flexible). + + ADVANCED TOPIC: Injecting native 3D-API buffers: + + The following struct members allow to inject your own GL, Metal + or D3D11 buffers into sokol_gfx: + + .gl_buffers[SG_NUM_INFLIGHT_FRAMES] + .mtl_buffers[SG_NUM_INFLIGHT_FRAMES] + .d3d11_buffer + + You must still provide all other struct items except the .data item, and + these must match the creation parameters of the native buffers you + provide. For SG_USAGE_IMMUTABLE, only provide a single native 3D-API + buffer, otherwise you need to provide SG_NUM_INFLIGHT_FRAMES buffers + (only for GL and Metal, not D3D11). Providing multiple buffers for GL and + Metal is necessary because sokol_gfx will rotate through them when + calling sg_update_buffer() to prevent lock-stalls. + + Note that it is expected that immutable injected buffer have already been + initialized with content, and the .content member must be 0! + + Also you need to call sg_reset_state_cache() after calling native 3D-API + functions, and before calling any sokol_gfx function. +*/ +typedef struct sg_buffer_desc { + uint32_t _start_canary; + size_t size; + sg_buffer_type type; + sg_usage usage; + sg_range data; + const char* label; + // optionally inject backend-specific resources + uint32_t gl_buffers[SG_NUM_INFLIGHT_FRAMES]; + const void* mtl_buffers[SG_NUM_INFLIGHT_FRAMES]; + const void* d3d11_buffer; + const void* wgpu_buffer; + uint32_t _end_canary; +} sg_buffer_desc; + +/* + sg_image_data + + Defines the content of an image through a 2D array of sg_range structs. + The first array dimension is the cubemap face, and the second array + dimension the mipmap level. +*/ +typedef struct sg_image_data { + sg_range subimage[SG_CUBEFACE_NUM][SG_MAX_MIPMAPS]; +} sg_image_data; + +/* + sg_image_desc + + Creation parameters for sg_image objects, used in the sg_make_image() call. + + The default configuration is: + + .type: SG_IMAGETYPE_2D + .render_target: false + .width 0 (must be set to >0) + .height 0 (must be set to >0) + .num_slices 1 (3D textures: depth; array textures: number of layers) + .num_mipmaps: 1 + .usage: SG_USAGE_IMMUTABLE + .pixel_format: SG_PIXELFORMAT_RGBA8 for textures, or sg_desc.environment.defaults.color_format for render targets + .sample_count: 1 for textures, or sg_desc.environment.defaults.sample_count for render targets + .data an sg_image_data struct to define the initial content + .label 0 (optional string label for trace hooks) + + Q: Why is the default sample_count for render targets identical with the + "default sample count" from sg_desc.environment.defaults.sample_count? + + A: So that it matches the default sample count in pipeline objects. Even + though it is a bit strange/confusing that offscreen render targets by default + get the same sample count as 'default swapchains', but it's better that + an offscreen render target created with default parameters matches + a pipeline object created with default parameters. + + NOTE: + + Images with usage SG_USAGE_IMMUTABLE must be fully initialized by + providing a valid .data member which points to initialization data. + + ADVANCED TOPIC: Injecting native 3D-API textures: + + The following struct members allow to inject your own GL, Metal or D3D11 + textures into sokol_gfx: + + .gl_textures[SG_NUM_INFLIGHT_FRAMES] + .mtl_textures[SG_NUM_INFLIGHT_FRAMES] + .d3d11_texture + .d3d11_shader_resource_view + .wgpu_texture + .wgpu_texture_view + + For GL, you can also specify the texture target or leave it empty to use + the default texture target for the image type (GL_TEXTURE_2D for + SG_IMAGETYPE_2D etc) + + For D3D11 and WebGPU, either only provide a texture, or both a texture and + shader-resource-view / texture-view object. If you want to use access the + injected texture in a shader you *must* provide a shader-resource-view. + + The same rules apply as for injecting native buffers (see sg_buffer_desc + documentation for more details). +*/ +typedef struct sg_image_desc { + uint32_t _start_canary; + sg_image_type type; + bool render_target; + int width; + int height; + int num_slices; + int num_mipmaps; + sg_usage usage; + sg_pixel_format pixel_format; + int sample_count; + sg_image_data data; + const char* label; + // optionally inject backend-specific resources + uint32_t gl_textures[SG_NUM_INFLIGHT_FRAMES]; + uint32_t gl_texture_target; + const void* mtl_textures[SG_NUM_INFLIGHT_FRAMES]; + const void* d3d11_texture; + const void* d3d11_shader_resource_view; + const void* wgpu_texture; + const void* wgpu_texture_view; + uint32_t _end_canary; +} sg_image_desc; + +/* + sg_sampler_desc + + Creation parameters for sg_sampler objects, used in the sg_make_sampler() call + + .min_filter: SG_FILTER_NEAREST + .mag_filter: SG_FILTER_NEAREST + .mipmap_filter SG_FILTER_NEAREST + .wrap_u: SG_WRAP_REPEAT + .wrap_v: SG_WRAP_REPEAT + .wrap_w: SG_WRAP_REPEAT (only SG_IMAGETYPE_3D) + .min_lod 0.0f + .max_lod FLT_MAX + .border_color SG_BORDERCOLOR_OPAQUE_BLACK + .compare SG_COMPAREFUNC_NEVER + .max_anisotropy 1 (must be 1..16) + +*/ +typedef struct sg_sampler_desc { + uint32_t _start_canary; + sg_filter min_filter; + sg_filter mag_filter; + sg_filter mipmap_filter; + sg_wrap wrap_u; + sg_wrap wrap_v; + sg_wrap wrap_w; + float min_lod; + float max_lod; + sg_border_color border_color; + sg_compare_func compare; + uint32_t max_anisotropy; + const char* label; + // optionally inject backend-specific resources + uint32_t gl_sampler; + const void* mtl_sampler; + const void* d3d11_sampler; + const void* wgpu_sampler; + uint32_t _end_canary; +} sg_sampler_desc; + +/* + sg_shader_desc + + Used as parameter of sg_make_shader() to create a shader object which + communicates shader source or bytecode and shader interface + reflection information to sokol-gfx. + + If you use sokol-shdc you can ignore the following information since + the sg_shader_desc struct will be code generated. + + Otherwise you need to provide the following information to the + sg_make_shader() call: + + - a vertex- and fragment-shader function: + - the shader source or bytecode + - an optional entry point name + - for D3D11: an optional compile target when source code is provided + (the defaults are "vs_4_0" and "ps_4_0") + + - ...or alternatively, a compute function: + - the shader source or bytecode + - an optional entry point name + - for D3D11: an optional compile target when source code is provided + (the default is "cs_5_0") + + - vertex attributes required by some backends (not for compute shaders): + - the vertex attribute base type (undefined, float, signed int, unsigned int), + this information is only used in the validation layer to check that the + pipeline object vertex formats are compatible with the input vertex attribute + type used in the vertex shader. NOTE that the default base type + 'undefined' skips the validation layer check. + - for the GL backend: optional vertex attribute names used for name lookup + - for the D3D11 backend: semantic names and indices + + - only for compute shaders on the Metal backend: + - the workgroup size aka 'threads per thread-group' + + In other 3D APIs this is declared in the shader code: + - GLSL: `layout(local_size_x=x, local_size_y=y, local_size_y=z) in;` + - HLSL: `[numthreads(x, y, z)]` + - WGSL: `@workgroup_size(x, y, z)` + ...but in Metal the workgroup size is declared on the CPU side + + - reflection information for each uniform block used by the shader: + - the shader stage the uniform block appears in (SG_SHADERSTAGE_*) + - the size in bytes of the uniform block + - backend-specific bindslots: + - HLSL: the constant buffer register `register(b0..7)` + - MSL: the buffer attribute `[[buffer(0..7)]]` + - WGSL: the binding in `@group(0) @binding(0..15)` + - GLSL only: a description of the uniform block interior + - the memory layout standard (SG_UNIFORMLAYOUT_*) + - for each member in the uniform block: + - the member type (SG_UNIFORM_*) + - if the member is an array, the array count + - the member name + + - reflection information for each texture used by the shader: + - the shader stage the texture appears in (SG_SHADERSTAGE_*) + - the image type (SG_IMAGETYPE_*) + - the image-sample type (SG_IMAGESAMPLETYPE_*) + - whether the texture is multisampled + - backend specific bindslots: + - HLSL: the texture register `register(t0..23)` + - MSL: the texture attribute `[[texture(0..15)]]` + - WGSL: the binding in `@group(1) @binding(0..127)` + + - reflection information for each sampler used by the shader: + - the shader stage the sampler appears in (SG_SHADERSTAGE_*) + - the sampler type (SG_SAMPLERTYPE_*) + - backend specific bindslots: + - HLSL: the sampler register `register(s0..15)` + - MSL: the sampler attribute `[[sampler(0..15)]]` + - WGSL: the binding in `@group(0) @binding(0..127)` + + - reflection information for each storage buffer used by the shader: + - the shader stage the storage buffer appears in (SG_SHADERSTAGE_*) + - whether the storage buffer is readonly (currently this must + always be true) + - backend specific bindslots: + - HLSL: + - for readonly storage buffer bindings: `register(t0..23)` + - for read/write storage buffer bindings: `register(u0..7)` + - MSL: the buffer attribute `[[buffer(8..15)]]` + - WGSL: the binding in `@group(1) @binding(0..127)` + - GL: the binding in `layout(binding=0..7)` + + - reflection information for each combined image-sampler object + used by the shader: + - the shader stage (SG_SHADERSTAGE_*) + - the texture's array index in the sg_shader_desc.images[] array + - the sampler's array index in the sg_shader_desc.samplers[] array + - GLSL only: the name of the combined image-sampler object + + The number and order of items in the sg_shader_desc.attrs[] + array corresponds to the items in sg_pipeline_desc.layout.attrs. + + - sg_shader_desc.attrs[N] => sg_pipeline_desc.layout.attrs[N] + + NOTE that vertex attribute indices currently cannot have gaps. + + The items index in the sg_shader_desc.uniform_blocks[] array corresponds + to the ub_slot arg in sg_apply_uniforms(): + + - sg_shader_desc.uniform_blocks[N] => sg_apply_uniforms(N, ...) + + The items in the shader_desc images, samplers and storage_buffers + arrays correspond to the same array items in the sg_bindings struct: + + - sg_shader_desc.images[N] => sg_bindings.images[N] + - sg_shader_desc.samplers[N] => sg_bindings.samplers[N] + - sg_shader_desc.storage_buffers[N] => sg_bindings.storage_buffers[N] + + For all GL backends, shader source-code must be provided. For D3D11 and Metal, + either shader source-code or byte-code can be provided. + + NOTE that the uniform block, image, sampler and storage_buffer arrays + can have gaps. This allows to use the same sg_bindings struct for + different related shader variants. + + For D3D11, if source code is provided, the d3dcompiler_47.dll will be loaded + on demand. If this fails, shader creation will fail. When compiling HLSL + source code, you can provide an optional target string via + sg_shader_stage_desc.d3d11_target, the default target is "vs_4_0" for the + vertex shader stage and "ps_4_0" for the pixel shader stage. +*/ +typedef enum sg_shader_stage { + SG_SHADERSTAGE_NONE, + SG_SHADERSTAGE_VERTEX, + SG_SHADERSTAGE_FRAGMENT, + SG_SHADERSTAGE_COMPUTE, + _SG_SHADERSTAGE_FORCE_U32 = 0x7FFFFFFF, +} sg_shader_stage; + +typedef struct sg_shader_function { + const char* source; + sg_range bytecode; + const char* entry; + const char* d3d11_target; // default: "vs_4_0" or "ps_4_0" +} sg_shader_function; + +typedef enum sg_shader_attr_base_type { + SG_SHADERATTRBASETYPE_UNDEFINED, + SG_SHADERATTRBASETYPE_FLOAT, + SG_SHADERATTRBASETYPE_SINT, + SG_SHADERATTRBASETYPE_UINT, + _SG_SHADERATTRBASETYPE_FORCE_U32 = 0x7FFFFFFF, +} sg_shader_attr_base_type; + +typedef struct sg_shader_vertex_attr { + sg_shader_attr_base_type base_type; // default: UNDEFINED (disables validation) + const char* glsl_name; // [optional] GLSL attribute name + const char* hlsl_sem_name; // HLSL semantic name + uint8_t hlsl_sem_index; // HLSL semantic index +} sg_shader_vertex_attr; + +typedef struct sg_glsl_shader_uniform { + sg_uniform_type type; + uint16_t array_count; // 0 or 1 for scalars, >1 for arrays + const char* glsl_name; // glsl name binding is required on GL 4.1 and WebGL2 +} sg_glsl_shader_uniform; + +typedef struct sg_shader_uniform_block { + sg_shader_stage stage; + uint32_t size; + uint8_t hlsl_register_b_n; // HLSL register(bn) + uint8_t msl_buffer_n; // MSL [[buffer(n)]] + uint8_t wgsl_group0_binding_n; // WGSL @group(0) @binding(n) + sg_uniform_layout layout; + sg_glsl_shader_uniform glsl_uniforms[SG_MAX_UNIFORMBLOCK_MEMBERS]; +} sg_shader_uniform_block; + +typedef struct sg_shader_image { + sg_shader_stage stage; + sg_image_type image_type; + sg_image_sample_type sample_type; + bool multisampled; + uint8_t hlsl_register_t_n; // HLSL register(tn) bind slot + uint8_t msl_texture_n; // MSL [[texture(n)]] bind slot + uint8_t wgsl_group1_binding_n; // WGSL @group(1) @binding(n) bind slot +} sg_shader_image; + +typedef struct sg_shader_sampler { + sg_shader_stage stage; + sg_sampler_type sampler_type; + uint8_t hlsl_register_s_n; // HLSL register(sn) bind slot + uint8_t msl_sampler_n; // MSL [[sampler(n)]] bind slot + uint8_t wgsl_group1_binding_n; // WGSL @group(1) @binding(n) bind slot +} sg_shader_sampler; + +typedef struct sg_shader_storage_buffer { + sg_shader_stage stage; + bool readonly; + uint8_t hlsl_register_t_n; // HLSL register(tn) bind slot (for readonly access) + uint8_t hlsl_register_u_n; // HLSL register(un) bind slot (for read/write access) + uint8_t msl_buffer_n; // MSL [[buffer(n)]] bind slot + uint8_t wgsl_group1_binding_n; // WGSL @group(1) @binding(n) bind slot + uint8_t glsl_binding_n; // GLSL layout(binding=n) +} sg_shader_storage_buffer; + +typedef struct sg_shader_image_sampler_pair { + sg_shader_stage stage; + uint8_t image_slot; + uint8_t sampler_slot; + const char* glsl_name; // glsl name binding required because of GL 4.1 and WebGL2 +} sg_shader_image_sampler_pair; + +typedef struct sg_mtl_shader_threads_per_threadgroup { + int x, y, z; +} sg_mtl_shader_threads_per_threadgroup; + +typedef struct sg_shader_desc { + uint32_t _start_canary; + sg_shader_function vertex_func; + sg_shader_function fragment_func; + sg_shader_function compute_func; + sg_shader_vertex_attr attrs[SG_MAX_VERTEX_ATTRIBUTES]; + sg_shader_uniform_block uniform_blocks[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + sg_shader_storage_buffer storage_buffers[SG_MAX_STORAGEBUFFER_BINDSLOTS]; + sg_shader_image images[SG_MAX_IMAGE_BINDSLOTS]; + sg_shader_sampler samplers[SG_MAX_SAMPLER_BINDSLOTS]; + sg_shader_image_sampler_pair image_sampler_pairs[SG_MAX_IMAGE_SAMPLER_PAIRS]; + sg_mtl_shader_threads_per_threadgroup mtl_threads_per_threadgroup; + const char* label; + uint32_t _end_canary; +} sg_shader_desc; + +/* + sg_pipeline_desc + + The sg_pipeline_desc struct defines all creation parameters for an + sg_pipeline object, used as argument to the sg_make_pipeline() function: + + Pipeline objects come in two flavours: + + - render pipelines for use in render passes + - compute pipelines for use in compute passes + + A compute pipeline only requires a compute shader object but no + 'render state', while a render pipeline requires a vertex/fragment shader + object and additional render state declarations: + + - the vertex layout for all input vertex buffers + - a shader object + - the 3D primitive type (points, lines, triangles, ...) + - the index type (none, 16- or 32-bit) + - all the fixed-function-pipeline state (depth-, stencil-, blend-state, etc...) + + If the vertex data has no gaps between vertex components, you can omit + the .layout.buffers[].stride and layout.attrs[].offset items (leave them + default-initialized to 0), sokol-gfx will then compute the offsets and + strides from the vertex component formats (.layout.attrs[].format). + Please note that ALL vertex attribute offsets must be 0 in order for the + automatic offset computation to kick in. + + The default configuration is as follows: + + .compute: false (must be set to true for a compute pipeline) + .shader: 0 (must be initialized with a valid sg_shader id!) + .layout: + .buffers[]: vertex buffer layouts + .stride: 0 (if no stride is given it will be computed) + .step_func SG_VERTEXSTEP_PER_VERTEX + .step_rate 1 + .attrs[]: vertex attribute declarations + .buffer_index 0 the vertex buffer bind slot + .offset 0 (offsets can be omitted if the vertex layout has no gaps) + .format SG_VERTEXFORMAT_INVALID (must be initialized!) + .depth: + .pixel_format: sg_desc.context.depth_format + .compare: SG_COMPAREFUNC_ALWAYS + .write_enabled: false + .bias: 0.0f + .bias_slope_scale: 0.0f + .bias_clamp: 0.0f + .stencil: + .enabled: false + .front/back: + .compare: SG_COMPAREFUNC_ALWAYS + .fail_op: SG_STENCILOP_KEEP + .depth_fail_op: SG_STENCILOP_KEEP + .pass_op: SG_STENCILOP_KEEP + .read_mask: 0 + .write_mask: 0 + .ref: 0 + .color_count 1 + .colors[0..color_count] + .pixel_format sg_desc.context.color_format + .write_mask: SG_COLORMASK_RGBA + .blend: + .enabled: false + .src_factor_rgb: SG_BLENDFACTOR_ONE + .dst_factor_rgb: SG_BLENDFACTOR_ZERO + .op_rgb: SG_BLENDOP_ADD + .src_factor_alpha: SG_BLENDFACTOR_ONE + .dst_factor_alpha: SG_BLENDFACTOR_ZERO + .op_alpha: SG_BLENDOP_ADD + .primitive_type: SG_PRIMITIVETYPE_TRIANGLES + .index_type: SG_INDEXTYPE_NONE + .cull_mode: SG_CULLMODE_NONE + .face_winding: SG_FACEWINDING_CW + .sample_count: sg_desc.context.sample_count + .blend_color: (sg_color) { 0.0f, 0.0f, 0.0f, 0.0f } + .alpha_to_coverage_enabled: false + .label 0 (optional string label for trace hooks) +*/ +typedef struct sg_vertex_buffer_layout_state { + int stride; + sg_vertex_step step_func; + int step_rate; +} sg_vertex_buffer_layout_state; + +typedef struct sg_vertex_attr_state { + int buffer_index; + int offset; + sg_vertex_format format; +} sg_vertex_attr_state; + +typedef struct sg_vertex_layout_state { + sg_vertex_buffer_layout_state buffers[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + sg_vertex_attr_state attrs[SG_MAX_VERTEX_ATTRIBUTES]; +} sg_vertex_layout_state; + +typedef struct sg_stencil_face_state { + sg_compare_func compare; + sg_stencil_op fail_op; + sg_stencil_op depth_fail_op; + sg_stencil_op pass_op; +} sg_stencil_face_state; + +typedef struct sg_stencil_state { + bool enabled; + sg_stencil_face_state front; + sg_stencil_face_state back; + uint8_t read_mask; + uint8_t write_mask; + uint8_t ref; +} sg_stencil_state; + +typedef struct sg_depth_state { + sg_pixel_format pixel_format; + sg_compare_func compare; + bool write_enabled; + float bias; + float bias_slope_scale; + float bias_clamp; +} sg_depth_state; + +typedef struct sg_blend_state { + bool enabled; + sg_blend_factor src_factor_rgb; + sg_blend_factor dst_factor_rgb; + sg_blend_op op_rgb; + sg_blend_factor src_factor_alpha; + sg_blend_factor dst_factor_alpha; + sg_blend_op op_alpha; +} sg_blend_state; + +typedef struct sg_color_target_state { + sg_pixel_format pixel_format; + sg_color_mask write_mask; + sg_blend_state blend; +} sg_color_target_state; + +typedef struct sg_pipeline_desc { + uint32_t _start_canary; + bool compute; + sg_shader shader; + sg_vertex_layout_state layout; + sg_depth_state depth; + sg_stencil_state stencil; + int color_count; + sg_color_target_state colors[SG_MAX_COLOR_ATTACHMENTS]; + sg_primitive_type primitive_type; + sg_index_type index_type; + sg_cull_mode cull_mode; + sg_face_winding face_winding; + int sample_count; + sg_color blend_color; + bool alpha_to_coverage_enabled; + const char* label; + uint32_t _end_canary; +} sg_pipeline_desc; + +/* + sg_attachments_desc + + Creation parameters for an sg_attachments object, used as argument to the + sg_make_attachments() function. + + An attachments object bundles 0..4 color attachments, 0..4 msaa-resolve + attachments, and none or one depth-stencil attachmente for use + in a render pass. At least one color attachment or one depth-stencil + attachment must be provided (no color attachment and a depth-stencil + attachment is useful for a depth-only render pass). + + Each attachment definition consists of an image object, and two additional indices + describing which subimage the pass will render into: one mipmap index, and if the image + is a cubemap, array-texture or 3D-texture, the face-index, array-layer or + depth-slice. + + All attachments must have the same width and height. + + All color attachments and the depth-stencil attachment must have the + same sample count. + + If a resolve attachment is set, an MSAA-resolve operation from the + associated color attachment image into the resolve attachment image will take + place in the sg_end_pass() function. In this case, the color attachment + must have a (sample_count>1), and the resolve attachment a + (sample_count==1). The resolve attachment also must have the same pixel + format as the color attachment. + + NOTE that MSAA depth-stencil attachments cannot be msaa-resolved! +*/ +typedef struct sg_attachment_desc { + sg_image image; + int mip_level; + int slice; // cube texture: face; array texture: layer; 3D texture: slice +} sg_attachment_desc; + +typedef struct sg_attachments_desc { + uint32_t _start_canary; + sg_attachment_desc colors[SG_MAX_COLOR_ATTACHMENTS]; + sg_attachment_desc resolves[SG_MAX_COLOR_ATTACHMENTS]; + sg_attachment_desc depth_stencil; + const char* label; + uint32_t _end_canary; +} sg_attachments_desc; + +/* + sg_trace_hooks + + Installable callback functions to keep track of the sokol-gfx calls, + this is useful for debugging, or keeping track of resource creation + and destruction. + + Trace hooks are installed with sg_install_trace_hooks(), this returns + another sg_trace_hooks struct with the previous set of + trace hook function pointers. These should be invoked by the + new trace hooks to form a proper call chain. +*/ +typedef struct sg_trace_hooks { + void* user_data; + void (*reset_state_cache)(void* user_data); + void (*make_buffer)(const sg_buffer_desc* desc, sg_buffer result, void* user_data); + void (*make_image)(const sg_image_desc* desc, sg_image result, void* user_data); + void (*make_sampler)(const sg_sampler_desc* desc, sg_sampler result, void* user_data); + void (*make_shader)(const sg_shader_desc* desc, sg_shader result, void* user_data); + void (*make_pipeline)(const sg_pipeline_desc* desc, sg_pipeline result, void* user_data); + void (*make_attachments)(const sg_attachments_desc* desc, sg_attachments result, void* user_data); + void (*destroy_buffer)(sg_buffer buf, void* user_data); + void (*destroy_image)(sg_image img, void* user_data); + void (*destroy_sampler)(sg_sampler smp, void* user_data); + void (*destroy_shader)(sg_shader shd, void* user_data); + void (*destroy_pipeline)(sg_pipeline pip, void* user_data); + void (*destroy_attachments)(sg_attachments atts, void* user_data); + void (*update_buffer)(sg_buffer buf, const sg_range* data, void* user_data); + void (*update_image)(sg_image img, const sg_image_data* data, void* user_data); + void (*append_buffer)(sg_buffer buf, const sg_range* data, int result, void* user_data); + void (*begin_pass)(const sg_pass* pass, void* user_data); + void (*apply_viewport)(int x, int y, int width, int height, bool origin_top_left, void* user_data); + void (*apply_scissor_rect)(int x, int y, int width, int height, bool origin_top_left, void* user_data); + void (*apply_pipeline)(sg_pipeline pip, void* user_data); + void (*apply_bindings)(const sg_bindings* bindings, void* user_data); + void (*apply_uniforms)(int ub_index, const sg_range* data, void* user_data); + void (*draw)(int base_element, int num_elements, int num_instances, void* user_data); + void (*dispatch)(int num_groups_x, int num_groups_y, int num_groups_z, void* user_data); + void (*end_pass)(void* user_data); + void (*commit)(void* user_data); + void (*alloc_buffer)(sg_buffer result, void* user_data); + void (*alloc_image)(sg_image result, void* user_data); + void (*alloc_sampler)(sg_sampler result, void* user_data); + void (*alloc_shader)(sg_shader result, void* user_data); + void (*alloc_pipeline)(sg_pipeline result, void* user_data); + void (*alloc_attachments)(sg_attachments result, void* user_data); + void (*dealloc_buffer)(sg_buffer buf_id, void* user_data); + void (*dealloc_image)(sg_image img_id, void* user_data); + void (*dealloc_sampler)(sg_sampler smp_id, void* user_data); + void (*dealloc_shader)(sg_shader shd_id, void* user_data); + void (*dealloc_pipeline)(sg_pipeline pip_id, void* user_data); + void (*dealloc_attachments)(sg_attachments atts_id, void* user_data); + void (*init_buffer)(sg_buffer buf_id, const sg_buffer_desc* desc, void* user_data); + void (*init_image)(sg_image img_id, const sg_image_desc* desc, void* user_data); + void (*init_sampler)(sg_sampler smp_id, const sg_sampler_desc* desc, void* user_data); + void (*init_shader)(sg_shader shd_id, const sg_shader_desc* desc, void* user_data); + void (*init_pipeline)(sg_pipeline pip_id, const sg_pipeline_desc* desc, void* user_data); + void (*init_attachments)(sg_attachments atts_id, const sg_attachments_desc* desc, void* user_data); + void (*uninit_buffer)(sg_buffer buf_id, void* user_data); + void (*uninit_image)(sg_image img_id, void* user_data); + void (*uninit_sampler)(sg_sampler smp_id, void* user_data); + void (*uninit_shader)(sg_shader shd_id, void* user_data); + void (*uninit_pipeline)(sg_pipeline pip_id, void* user_data); + void (*uninit_attachments)(sg_attachments atts_id, void* user_data); + void (*fail_buffer)(sg_buffer buf_id, void* user_data); + void (*fail_image)(sg_image img_id, void* user_data); + void (*fail_sampler)(sg_sampler smp_id, void* user_data); + void (*fail_shader)(sg_shader shd_id, void* user_data); + void (*fail_pipeline)(sg_pipeline pip_id, void* user_data); + void (*fail_attachments)(sg_attachments atts_id, void* user_data); + void (*push_debug_group)(const char* name, void* user_data); + void (*pop_debug_group)(void* user_data); +} sg_trace_hooks; + +/* + sg_buffer_info + sg_image_info + sg_sampler_info + sg_shader_info + sg_pipeline_info + sg_attachments_info + + These structs contain various internal resource attributes which + might be useful for debug-inspection. Please don't rely on the + actual content of those structs too much, as they are quite closely + tied to sokol_gfx.h internals and may change more frequently than + the other public API elements. + + The *_info structs are used as the return values of the following functions: + + sg_query_buffer_info() + sg_query_image_info() + sg_query_sampler_info() + sg_query_shader_info() + sg_query_pipeline_info() + sg_query_attachments_info() +*/ +typedef struct sg_slot_info { + sg_resource_state state; // the current state of this resource slot + uint32_t res_id; // type-neutral resource if (e.g. sg_buffer.id) +} sg_slot_info; + +typedef struct sg_buffer_info { + sg_slot_info slot; // resource pool slot info + uint32_t update_frame_index; // frame index of last sg_update_buffer() + uint32_t append_frame_index; // frame index of last sg_append_buffer() + int append_pos; // current position in buffer for sg_append_buffer() + bool append_overflow; // is buffer in overflow state (due to sg_append_buffer) + int num_slots; // number of renaming-slots for dynamically updated buffers + int active_slot; // currently active write-slot for dynamically updated buffers +} sg_buffer_info; + +typedef struct sg_image_info { + sg_slot_info slot; // resource pool slot info + uint32_t upd_frame_index; // frame index of last sg_update_image() + int num_slots; // number of renaming-slots for dynamically updated images + int active_slot; // currently active write-slot for dynamically updated images +} sg_image_info; + +typedef struct sg_sampler_info { + sg_slot_info slot; // resource pool slot info +} sg_sampler_info; + +typedef struct sg_shader_info { + sg_slot_info slot; // resource pool slot info +} sg_shader_info; + +typedef struct sg_pipeline_info { + sg_slot_info slot; // resource pool slot info +} sg_pipeline_info; + +typedef struct sg_attachments_info { + sg_slot_info slot; // resource pool slot info +} sg_attachments_info; + +/* + sg_frame_stats + + Allows to track generic and backend-specific stats about a + render frame. Obtained by calling sg_query_frame_stats(). The returned + struct contains information about the *previous* frame. +*/ +typedef struct sg_frame_stats_gl { + uint32_t num_bind_buffer; + uint32_t num_active_texture; + uint32_t num_bind_texture; + uint32_t num_bind_sampler; + uint32_t num_use_program; + uint32_t num_render_state; + uint32_t num_vertex_attrib_pointer; + uint32_t num_vertex_attrib_divisor; + uint32_t num_enable_vertex_attrib_array; + uint32_t num_disable_vertex_attrib_array; + uint32_t num_uniform; + uint32_t num_memory_barriers; +} sg_frame_stats_gl; + +typedef struct sg_frame_stats_d3d11_pass { + uint32_t num_om_set_render_targets; + uint32_t num_clear_render_target_view; + uint32_t num_clear_depth_stencil_view; + uint32_t num_resolve_subresource; +} sg_frame_stats_d3d11_pass; + +typedef struct sg_frame_stats_d3d11_pipeline { + uint32_t num_rs_set_state; + uint32_t num_om_set_depth_stencil_state; + uint32_t num_om_set_blend_state; + uint32_t num_ia_set_primitive_topology; + uint32_t num_ia_set_input_layout; + uint32_t num_vs_set_shader; + uint32_t num_vs_set_constant_buffers; + uint32_t num_ps_set_shader; + uint32_t num_ps_set_constant_buffers; + uint32_t num_cs_set_shader; + uint32_t num_cs_set_constant_buffers; +} sg_frame_stats_d3d11_pipeline; + +typedef struct sg_frame_stats_d3d11_bindings { + uint32_t num_ia_set_vertex_buffers; + uint32_t num_ia_set_index_buffer; + uint32_t num_vs_set_shader_resources; + uint32_t num_vs_set_samplers; + uint32_t num_ps_set_shader_resources; + uint32_t num_ps_set_samplers; + uint32_t num_cs_set_shader_resources; + uint32_t num_cs_set_samplers; + uint32_t num_cs_set_unordered_access_views; +} sg_frame_stats_d3d11_bindings; + +typedef struct sg_frame_stats_d3d11_uniforms { + uint32_t num_update_subresource; +} sg_frame_stats_d3d11_uniforms; + +typedef struct sg_frame_stats_d3d11_draw { + uint32_t num_draw_indexed_instanced; + uint32_t num_draw_indexed; + uint32_t num_draw_instanced; + uint32_t num_draw; +} sg_frame_stats_d3d11_draw; + +typedef struct sg_frame_stats_d3d11 { + sg_frame_stats_d3d11_pass pass; + sg_frame_stats_d3d11_pipeline pipeline; + sg_frame_stats_d3d11_bindings bindings; + sg_frame_stats_d3d11_uniforms uniforms; + sg_frame_stats_d3d11_draw draw; + uint32_t num_map; + uint32_t num_unmap; +} sg_frame_stats_d3d11; + +typedef struct sg_frame_stats_metal_idpool { + uint32_t num_added; + uint32_t num_released; + uint32_t num_garbage_collected; +} sg_frame_stats_metal_idpool; + +typedef struct sg_frame_stats_metal_pipeline { + uint32_t num_set_blend_color; + uint32_t num_set_cull_mode; + uint32_t num_set_front_facing_winding; + uint32_t num_set_stencil_reference_value; + uint32_t num_set_depth_bias; + uint32_t num_set_render_pipeline_state; + uint32_t num_set_depth_stencil_state; +} sg_frame_stats_metal_pipeline; + +typedef struct sg_frame_stats_metal_bindings { + uint32_t num_set_vertex_buffer; + uint32_t num_set_vertex_texture; + uint32_t num_set_vertex_sampler_state; + uint32_t num_set_fragment_buffer; + uint32_t num_set_fragment_texture; + uint32_t num_set_fragment_sampler_state; + uint32_t num_set_compute_buffer; + uint32_t num_set_compute_texture; + uint32_t num_set_compute_sampler_state; +} sg_frame_stats_metal_bindings; + +typedef struct sg_frame_stats_metal_uniforms { + uint32_t num_set_vertex_buffer_offset; + uint32_t num_set_fragment_buffer_offset; + uint32_t num_set_compute_buffer_offset; +} sg_frame_stats_metal_uniforms; + +typedef struct sg_frame_stats_metal { + sg_frame_stats_metal_idpool idpool; + sg_frame_stats_metal_pipeline pipeline; + sg_frame_stats_metal_bindings bindings; + sg_frame_stats_metal_uniforms uniforms; +} sg_frame_stats_metal; + +typedef struct sg_frame_stats_wgpu_uniforms { + uint32_t num_set_bindgroup; + uint32_t size_write_buffer; +} sg_frame_stats_wgpu_uniforms; + +typedef struct sg_frame_stats_wgpu_bindings { + uint32_t num_set_vertex_buffer; + uint32_t num_skip_redundant_vertex_buffer; + uint32_t num_set_index_buffer; + uint32_t num_skip_redundant_index_buffer; + uint32_t num_create_bindgroup; + uint32_t num_discard_bindgroup; + uint32_t num_set_bindgroup; + uint32_t num_skip_redundant_bindgroup; + uint32_t num_bindgroup_cache_hits; + uint32_t num_bindgroup_cache_misses; + uint32_t num_bindgroup_cache_collisions; + uint32_t num_bindgroup_cache_invalidates; + uint32_t num_bindgroup_cache_hash_vs_key_mismatch; +} sg_frame_stats_wgpu_bindings; + +typedef struct sg_frame_stats_wgpu { + sg_frame_stats_wgpu_uniforms uniforms; + sg_frame_stats_wgpu_bindings bindings; +} sg_frame_stats_wgpu; + +typedef struct sg_frame_stats { + uint32_t frame_index; // current frame counter, starts at 0 + + uint32_t num_passes; + uint32_t num_apply_viewport; + uint32_t num_apply_scissor_rect; + uint32_t num_apply_pipeline; + uint32_t num_apply_bindings; + uint32_t num_apply_uniforms; + uint32_t num_draw; + uint32_t num_dispatch; + uint32_t num_update_buffer; + uint32_t num_append_buffer; + uint32_t num_update_image; + + uint32_t size_apply_uniforms; + uint32_t size_update_buffer; + uint32_t size_append_buffer; + uint32_t size_update_image; + + sg_frame_stats_gl gl; + sg_frame_stats_d3d11 d3d11; + sg_frame_stats_metal metal; + sg_frame_stats_wgpu wgpu; +} sg_frame_stats; + +/* + sg_log_item + + An enum with a unique item for each log message, warning, error + and validation layer message. Note that these messages are only + visible when a logger function is installed in the sg_setup() call. +*/ +#define _SG_LOG_ITEMS \ + _SG_LOGITEM_XMACRO(OK, "Ok") \ + _SG_LOGITEM_XMACRO(MALLOC_FAILED, "memory allocation failed") \ + _SG_LOGITEM_XMACRO(GL_TEXTURE_FORMAT_NOT_SUPPORTED, "pixel format not supported for texture (gl)") \ + _SG_LOGITEM_XMACRO(GL_3D_TEXTURES_NOT_SUPPORTED, "3d textures not supported (gl)") \ + _SG_LOGITEM_XMACRO(GL_ARRAY_TEXTURES_NOT_SUPPORTED, "array textures not supported (gl)") \ + _SG_LOGITEM_XMACRO(GL_STORAGEBUFFER_GLSL_BINDING_OUT_OF_RANGE, "GLSL storage buffer bindslot is out of range (must be 0..7) (gl)") \ + _SG_LOGITEM_XMACRO(GL_SHADER_COMPILATION_FAILED, "shader compilation failed (gl)") \ + _SG_LOGITEM_XMACRO(GL_SHADER_LINKING_FAILED, "shader linking failed (gl)") \ + _SG_LOGITEM_XMACRO(GL_VERTEX_ATTRIBUTE_NOT_FOUND_IN_SHADER, "vertex attribute not found in shader; NOTE: may be caused by GL driver's GLSL compiler removing unused globals") \ + _SG_LOGITEM_XMACRO(GL_UNIFORMBLOCK_NAME_NOT_FOUND_IN_SHADER, "uniform block name not found in shader; NOTE: may be caused by GL driver's GLSL compiler removing unused globals") \ + _SG_LOGITEM_XMACRO(GL_IMAGE_SAMPLER_NAME_NOT_FOUND_IN_SHADER, "image-sampler name not found in shader; NOTE: may be caused by GL driver's GLSL compiler removing unused globals") \ + _SG_LOGITEM_XMACRO(GL_FRAMEBUFFER_STATUS_UNDEFINED, "framebuffer completeness check failed with GL_FRAMEBUFFER_UNDEFINED (gl)") \ + _SG_LOGITEM_XMACRO(GL_FRAMEBUFFER_STATUS_INCOMPLETE_ATTACHMENT, "framebuffer completeness check failed with GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT (gl)") \ + _SG_LOGITEM_XMACRO(GL_FRAMEBUFFER_STATUS_INCOMPLETE_MISSING_ATTACHMENT, "framebuffer completeness check failed with GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT (gl)") \ + _SG_LOGITEM_XMACRO(GL_FRAMEBUFFER_STATUS_UNSUPPORTED, "framebuffer completeness check failed with GL_FRAMEBUFFER_UNSUPPORTED (gl)") \ + _SG_LOGITEM_XMACRO(GL_FRAMEBUFFER_STATUS_INCOMPLETE_MULTISAMPLE, "framebuffer completeness check failed with GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE (gl)") \ + _SG_LOGITEM_XMACRO(GL_FRAMEBUFFER_STATUS_UNKNOWN, "framebuffer completeness check failed (unknown reason) (gl)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_BUFFER_FAILED, "CreateBuffer() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_BUFFER_SRV_FAILED, "CreateShaderResourceView() failed for storage buffer (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_BUFFER_UAV_FAILED, "CreateUnorderedAccessView() failed for storage buffer (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_DEPTH_TEXTURE_UNSUPPORTED_PIXEL_FORMAT, "pixel format not supported for depth-stencil texture (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_DEPTH_TEXTURE_FAILED, "CreateTexture2D() failed for depth-stencil texture (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_2D_TEXTURE_UNSUPPORTED_PIXEL_FORMAT, "pixel format not supported for 2d-, cube- or array-texture (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_2D_TEXTURE_FAILED, "CreateTexture2D() failed for 2d-, cube- or array-texture (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_2D_SRV_FAILED, "CreateShaderResourceView() failed for 2d-, cube- or array-texture (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_3D_TEXTURE_UNSUPPORTED_PIXEL_FORMAT, "pixel format not supported for 3D texture (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_3D_TEXTURE_FAILED, "CreateTexture3D() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_3D_SRV_FAILED, "CreateShaderResourceView() failed for 3d texture (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_MSAA_TEXTURE_FAILED, "CreateTexture2D() failed for MSAA render target texture (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_SAMPLER_STATE_FAILED, "CreateSamplerState() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_UNIFORMBLOCK_HLSL_REGISTER_B_OUT_OF_RANGE, "uniform block 'hlsl_register_b_n' is out of range (must be 0..7)") \ + _SG_LOGITEM_XMACRO(D3D11_STORAGEBUFFER_HLSL_REGISTER_T_OUT_OF_RANGE, "storage buffer 'hlsl_register_t_n' is out of range (must be 0..23)") \ + _SG_LOGITEM_XMACRO(D3D11_STORAGEBUFFER_HLSL_REGISTER_U_OUT_OF_RANGE, "storage buffer 'hlsl_register_u_n' is out of range (must be 0..7)") \ + _SG_LOGITEM_XMACRO(D3D11_IMAGE_HLSL_REGISTER_T_OUT_OF_RANGE, "image 'hlsl_register_t_n' is out of range (must be 0..23)") \ + _SG_LOGITEM_XMACRO(D3D11_SAMPLER_HLSL_REGISTER_S_OUT_OF_RANGE, "sampler 'hlsl_register_s_n' is out of rang (must be 0..15)") \ + _SG_LOGITEM_XMACRO(D3D11_LOAD_D3DCOMPILER_47_DLL_FAILED, "loading d3dcompiler_47.dll failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_SHADER_COMPILATION_FAILED, "shader compilation failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_SHADER_COMPILATION_OUTPUT, "") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_CONSTANT_BUFFER_FAILED, "CreateBuffer() failed for uniform constant buffer (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_INPUT_LAYOUT_FAILED, "CreateInputLayout() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_RASTERIZER_STATE_FAILED, "CreateRasterizerState() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_DEPTH_STENCIL_STATE_FAILED, "CreateDepthStencilState() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_BLEND_STATE_FAILED, "CreateBlendState() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_RTV_FAILED, "CreateRenderTargetView() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_CREATE_DSV_FAILED, "CreateDepthStencilView() failed (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_MAP_FOR_UPDATE_BUFFER_FAILED, "Map() failed when updating buffer (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_MAP_FOR_APPEND_BUFFER_FAILED, "Map() failed when appending to buffer (d3d11)") \ + _SG_LOGITEM_XMACRO(D3D11_MAP_FOR_UPDATE_IMAGE_FAILED, "Map() failed when updating image (d3d11)") \ + _SG_LOGITEM_XMACRO(METAL_CREATE_BUFFER_FAILED, "failed to create buffer object (metal)") \ + _SG_LOGITEM_XMACRO(METAL_TEXTURE_FORMAT_NOT_SUPPORTED, "pixel format not supported for texture (metal)") \ + _SG_LOGITEM_XMACRO(METAL_CREATE_TEXTURE_FAILED, "failed to create texture object (metal)") \ + _SG_LOGITEM_XMACRO(METAL_CREATE_SAMPLER_FAILED, "failed to create sampler object (metal)") \ + _SG_LOGITEM_XMACRO(METAL_SHADER_COMPILATION_FAILED, "shader compilation failed (metal)") \ + _SG_LOGITEM_XMACRO(METAL_SHADER_CREATION_FAILED, "shader creation failed (metal)") \ + _SG_LOGITEM_XMACRO(METAL_SHADER_COMPILATION_OUTPUT, "") \ + _SG_LOGITEM_XMACRO(METAL_SHADER_ENTRY_NOT_FOUND, "shader entry function not found (metal)") \ + _SG_LOGITEM_XMACRO(METAL_UNIFORMBLOCK_MSL_BUFFER_SLOT_OUT_OF_RANGE, "uniform block 'msl_buffer_n' is out of range (must be 0..7)") \ + _SG_LOGITEM_XMACRO(METAL_STORAGEBUFFER_MSL_BUFFER_SLOT_OUT_OF_RANGE, "storage buffer 'msl_buffer_n' is out of range (must be 8..15)") \ + _SG_LOGITEM_XMACRO(METAL_IMAGE_MSL_TEXTURE_SLOT_OUT_OF_RANGE, "image 'msl_texture_n' is out of range (must be 0..15)") \ + _SG_LOGITEM_XMACRO(METAL_SAMPLER_MSL_SAMPLER_SLOT_OUT_OF_RANGE, "sampler 'msl_sampler_n' is out of range (must be 0..15)") \ + _SG_LOGITEM_XMACRO(METAL_CREATE_CPS_FAILED, "failed to create compute pipeline state (metal)") \ + _SG_LOGITEM_XMACRO(METAL_CREATE_CPS_OUTPUT, "") \ + _SG_LOGITEM_XMACRO(METAL_CREATE_RPS_FAILED, "failed to create render pipeline state (metal)") \ + _SG_LOGITEM_XMACRO(METAL_CREATE_RPS_OUTPUT, "") \ + _SG_LOGITEM_XMACRO(METAL_CREATE_DSS_FAILED, "failed to create depth stencil state (metal)") \ + _SG_LOGITEM_XMACRO(WGPU_BINDGROUPS_POOL_EXHAUSTED, "bindgroups pool exhausted (increase sg_desc.bindgroups_cache_size) (wgpu)") \ + _SG_LOGITEM_XMACRO(WGPU_BINDGROUPSCACHE_SIZE_GREATER_ONE, "sg_desc.wgpu_bindgroups_cache_size must be > 1 (wgpu)") \ + _SG_LOGITEM_XMACRO(WGPU_BINDGROUPSCACHE_SIZE_POW2, "sg_desc.wgpu_bindgroups_cache_size must be a power of 2 (wgpu)") \ + _SG_LOGITEM_XMACRO(WGPU_CREATEBINDGROUP_FAILED, "wgpuDeviceCreateBindGroup failed") \ + _SG_LOGITEM_XMACRO(WGPU_CREATE_BUFFER_FAILED, "wgpuDeviceCreateBuffer() failed") \ + _SG_LOGITEM_XMACRO(WGPU_CREATE_TEXTURE_FAILED, "wgpuDeviceCreateTexture() failed") \ + _SG_LOGITEM_XMACRO(WGPU_CREATE_TEXTURE_VIEW_FAILED, "wgpuTextureCreateView() failed") \ + _SG_LOGITEM_XMACRO(WGPU_CREATE_SAMPLER_FAILED, "wgpuDeviceCreateSampler() failed") \ + _SG_LOGITEM_XMACRO(WGPU_CREATE_SHADER_MODULE_FAILED, "wgpuDeviceCreateShaderModule() failed") \ + _SG_LOGITEM_XMACRO(WGPU_SHADER_CREATE_BINDGROUP_LAYOUT_FAILED, "wgpuDeviceCreateBindGroupLayout() for shader stage failed") \ + _SG_LOGITEM_XMACRO(WGPU_UNIFORMBLOCK_WGSL_GROUP0_BINDING_OUT_OF_RANGE, "uniform block 'wgsl_group0_binding_n' is out of range (must be 0..15)") \ + _SG_LOGITEM_XMACRO(WGPU_STORAGEBUFFER_WGSL_GROUP1_BINDING_OUT_OF_RANGE, "storage buffer 'wgsl_group1_binding_n' is out of range (must be 0..127)") \ + _SG_LOGITEM_XMACRO(WGPU_IMAGE_WGSL_GROUP1_BINDING_OUT_OF_RANGE, "image 'wgsl_group1_binding_n' is out of range (must be 0..127)") \ + _SG_LOGITEM_XMACRO(WGPU_SAMPLER_WGSL_GROUP1_BINDING_OUT_OF_RANGE, "sampler 'wgsl_group1_binding_n' is out of range (must be 0..127)") \ + _SG_LOGITEM_XMACRO(WGPU_CREATE_PIPELINE_LAYOUT_FAILED, "wgpuDeviceCreatePipelineLayout() failed") \ + _SG_LOGITEM_XMACRO(WGPU_CREATE_RENDER_PIPELINE_FAILED, "wgpuDeviceCreateRenderPipeline() failed") \ + _SG_LOGITEM_XMACRO(WGPU_CREATE_COMPUTE_PIPELINE_FAILED, "wgpuDeviceCreateComputePipeline() failed") \ + _SG_LOGITEM_XMACRO(WGPU_ATTACHMENTS_CREATE_TEXTURE_VIEW_FAILED, "wgpuTextureCreateView() failed in create attachments") \ + _SG_LOGITEM_XMACRO(IDENTICAL_COMMIT_LISTENER, "attempting to add identical commit listener") \ + _SG_LOGITEM_XMACRO(COMMIT_LISTENER_ARRAY_FULL, "commit listener array full") \ + _SG_LOGITEM_XMACRO(TRACE_HOOKS_NOT_ENABLED, "sg_install_trace_hooks() called, but SOKOL_TRACE_HOOKS is not defined") \ + _SG_LOGITEM_XMACRO(DEALLOC_BUFFER_INVALID_STATE, "sg_dealloc_buffer(): buffer must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(DEALLOC_IMAGE_INVALID_STATE, "sg_dealloc_image(): image must be in alloc state") \ + _SG_LOGITEM_XMACRO(DEALLOC_SAMPLER_INVALID_STATE, "sg_dealloc_sampler(): sampler must be in alloc state") \ + _SG_LOGITEM_XMACRO(DEALLOC_SHADER_INVALID_STATE, "sg_dealloc_shader(): shader must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(DEALLOC_PIPELINE_INVALID_STATE, "sg_dealloc_pipeline(): pipeline must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(DEALLOC_ATTACHMENTS_INVALID_STATE, "sg_dealloc_attachments(): attachments must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(INIT_BUFFER_INVALID_STATE, "sg_init_buffer(): buffer must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(INIT_IMAGE_INVALID_STATE, "sg_init_image(): image must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(INIT_SAMPLER_INVALID_STATE, "sg_init_sampler(): sampler must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(INIT_SHADER_INVALID_STATE, "sg_init_shader(): shader must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(INIT_PIPELINE_INVALID_STATE, "sg_init_pipeline(): pipeline must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(INIT_ATTACHMENTS_INVALID_STATE, "sg_init_attachments(): pass must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(UNINIT_BUFFER_INVALID_STATE, "sg_uninit_buffer(): buffer must be in VALID or FAILED state") \ + _SG_LOGITEM_XMACRO(UNINIT_IMAGE_INVALID_STATE, "sg_uninit_image(): image must be in VALID or FAILED state") \ + _SG_LOGITEM_XMACRO(UNINIT_SAMPLER_INVALID_STATE, "sg_uninit_sampler(): sampler must be in VALID or FAILED state") \ + _SG_LOGITEM_XMACRO(UNINIT_SHADER_INVALID_STATE, "sg_uninit_shader(): shader must be in VALID or FAILED state") \ + _SG_LOGITEM_XMACRO(UNINIT_PIPELINE_INVALID_STATE, "sg_uninit_pipeline(): pipeline must be in VALID or FAILED state") \ + _SG_LOGITEM_XMACRO(UNINIT_ATTACHMENTS_INVALID_STATE, "sg_uninit_attachments(): attachments must be in VALID or FAILED state") \ + _SG_LOGITEM_XMACRO(FAIL_BUFFER_INVALID_STATE, "sg_fail_buffer(): buffer must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(FAIL_IMAGE_INVALID_STATE, "sg_fail_image(): image must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(FAIL_SAMPLER_INVALID_STATE, "sg_fail_sampler(): sampler must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(FAIL_SHADER_INVALID_STATE, "sg_fail_shader(): shader must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(FAIL_PIPELINE_INVALID_STATE, "sg_fail_pipeline(): pipeline must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(FAIL_ATTACHMENTS_INVALID_STATE, "sg_fail_attachments(): attachments must be in ALLOC state") \ + _SG_LOGITEM_XMACRO(BUFFER_POOL_EXHAUSTED, "buffer pool exhausted") \ + _SG_LOGITEM_XMACRO(IMAGE_POOL_EXHAUSTED, "image pool exhausted") \ + _SG_LOGITEM_XMACRO(SAMPLER_POOL_EXHAUSTED, "sampler pool exhausted") \ + _SG_LOGITEM_XMACRO(SHADER_POOL_EXHAUSTED, "shader pool exhausted") \ + _SG_LOGITEM_XMACRO(PIPELINE_POOL_EXHAUSTED, "pipeline pool exhausted") \ + _SG_LOGITEM_XMACRO(PASS_POOL_EXHAUSTED, "pass pool exhausted") \ + _SG_LOGITEM_XMACRO(BEGINPASS_ATTACHMENT_INVALID, "sg_begin_pass: an attachment was provided that no longer exists") \ + _SG_LOGITEM_XMACRO(APPLY_BINDINGS_STORAGE_BUFFER_TRACKER_EXHAUSTED, "sg_apply_bindings: too many read/write storage buffers in pass (bump sg_desc.max_dispatch_calls_per_pass") \ + _SG_LOGITEM_XMACRO(DRAW_WITHOUT_BINDINGS, "attempting to draw without resource bindings") \ + _SG_LOGITEM_XMACRO(VALIDATE_BUFFERDESC_CANARY, "sg_buffer_desc not initialized") \ + _SG_LOGITEM_XMACRO(VALIDATE_BUFFERDESC_EXPECT_NONZERO_SIZE, "sg_buffer_desc.size must be greater zero") \ + _SG_LOGITEM_XMACRO(VALIDATE_BUFFERDESC_EXPECT_MATCHING_DATA_SIZE, "sg_buffer_desc.size and .data.size must be equal") \ + _SG_LOGITEM_XMACRO(VALIDATE_BUFFERDESC_EXPECT_ZERO_DATA_SIZE, "sg_buffer_desc.data.size expected to be zero") \ + _SG_LOGITEM_XMACRO(VALIDATE_BUFFERDESC_EXPECT_NO_DATA, "sg_buffer_desc.data.ptr must be null for dynamic/stream buffers") \ + _SG_LOGITEM_XMACRO(VALIDATE_BUFFERDESC_STORAGEBUFFER_SUPPORTED, "storage buffers not supported by the backend 3D API (requires OpenGL >= 4.3)") \ + _SG_LOGITEM_XMACRO(VALIDATE_BUFFERDESC_STORAGEBUFFER_SIZE_MULTIPLE_4, "size of storage buffers must be a multiple of 4") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDATA_NODATA, "sg_image_data: no data (.ptr and/or .size is zero)") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDATA_DATA_SIZE, "sg_image_data: data size doesn't match expected surface size") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_CANARY, "sg_image_desc not initialized") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_WIDTH, "sg_image_desc.width must be > 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_HEIGHT, "sg_image_desc.height must be > 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_RT_PIXELFORMAT, "invalid pixel format for render-target image") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_NONRT_PIXELFORMAT, "invalid pixel format for non-render-target image") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_MSAA_BUT_NO_RT, "non-render-target images cannot be multisampled") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_NO_MSAA_RT_SUPPORT, "MSAA not supported for this pixel format") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_MSAA_NUM_MIPMAPS, "MSAA images must have num_mipmaps == 1") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_MSAA_3D_IMAGE, "3D images cannot have a sample_count > 1") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_MSAA_CUBE_IMAGE, "cube images cannot have sample_count > 1") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_DEPTH_3D_IMAGE, "3D images cannot have a depth/stencil image format") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_RT_IMMUTABLE, "render target images must be SG_USAGE_IMMUTABLE") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_RT_NO_DATA, "render target images cannot be initialized with data") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_INJECTED_NO_DATA, "images with injected textures cannot be initialized with data") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_DYNAMIC_NO_DATA, "dynamic/stream images cannot be initialized with data") \ + _SG_LOGITEM_XMACRO(VALIDATE_IMAGEDESC_COMPRESSED_IMMUTABLE, "compressed images must be immutable") \ + _SG_LOGITEM_XMACRO(VALIDATE_SAMPLERDESC_CANARY, "sg_sampler_desc not initialized") \ + _SG_LOGITEM_XMACRO(VALIDATE_SAMPLERDESC_ANISTROPIC_REQUIRES_LINEAR_FILTERING, "sg_sampler_desc.max_anisotropy > 1 requires min/mag/mipmap_filter to be SG_FILTER_LINEAR") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_CANARY, "sg_shader_desc not initialized") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_VERTEX_SOURCE, "vertex shader source code expected") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_FRAGMENT_SOURCE, "fragment shader source code expected") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_COMPUTE_SOURCE, "compute shader source code expected") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_VERTEX_SOURCE_OR_BYTECODE, "vertex shader source or byte code expected") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_FRAGMENT_SOURCE_OR_BYTECODE, "fragment shader source or byte code expected") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_COMPUTE_SOURCE_OR_BYTECODE, "compute shader source or byte code expected") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_INVALID_SHADER_COMBO, "cannot combine compute shaders with vertex or fragment shaders") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_NO_BYTECODE_SIZE, "shader byte code length (in bytes) required") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_METAL_THREADS_PER_THREADGROUP, "sg_shader_desc.mtl_threads_per_threadgroup must be initialized for compute shaders (metal)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_NO_CONT_MEMBERS, "uniform block members must occupy continuous slots") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_SIZE_IS_ZERO, "bound uniform block size cannot be zero") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_METAL_BUFFER_SLOT_OUT_OF_RANGE, "uniform block 'msl_buffer_n' is out of range (must be 0..7)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_METAL_BUFFER_SLOT_COLLISION, "uniform block 'msl_buffer_n' must be unique across uniform blocks and storage buffers in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_HLSL_REGISTER_B_OUT_OF_RANGE, "uniform block 'hlsl_register_b_n' is out of range (must be 0..7)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_HLSL_REGISTER_B_COLLISION, "uniform block 'hlsl_register_b_n' must be unique across uniform blocks in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_WGSL_GROUP0_BINDING_OUT_OF_RANGE, "uniform block 'wgsl_group0_binding_n' is out of range (must be 0..15)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_WGSL_GROUP0_BINDING_COLLISION, "uniform block 'wgsl_group0_binding_n' must be unique across all uniform blocks") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_NO_MEMBERS, "GL backend requires uniform block member declarations") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_UNIFORM_GLSL_NAME, "uniform block member 'glsl_name' missing") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_SIZE_MISMATCH, "size of uniform block members doesn't match uniform block size") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_ARRAY_COUNT, "uniform array count must be >= 1") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_UNIFORMBLOCK_STD140_ARRAY_TYPE, "uniform arrays only allowed for FLOAT4, INT4, MAT4 in std140 layout") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_METAL_BUFFER_SLOT_OUT_OF_RANGE, "storage buffer 'msl_buffer_n' is out of range (must be 8..15)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_METAL_BUFFER_SLOT_COLLISION, "storage buffer 'msl_buffer_n' must be unique across uniform blocks and storage buffer in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_HLSL_REGISTER_T_OUT_OF_RANGE, "storage buffer 'hlsl_register_t_n' is out of range (must be 0..23)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_HLSL_REGISTER_T_COLLISION, "storage_buffer 'hlsl_register_t_n' must be unique across read-only storage buffers and images in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_HLSL_REGISTER_U_OUT_OF_RANGE, "storage buffer 'hlsl_register_u_n' is out of range (must be 0..7)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_HLSL_REGISTER_U_COLLISION, "storage_buffer 'hlsl_register_u_n' must be unique across read/write storage buffers in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_GLSL_BINDING_OUT_OF_RANGE, "storage buffer 'glsl_binding_n' is out of range (must be 0..7)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_GLSL_BINDING_COLLISION, "storage buffer 'glsl_binding_n' must be unique across shader stages") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_WGSL_GROUP1_BINDING_OUT_OF_RANGE, "storage buffer 'wgsl_group1_binding_n' is out of range (must be 0..127)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_STORAGEBUFFER_WGSL_GROUP1_BINDING_COLLISION, "storage buffer 'wgsl_group1_binding_n' must be unique across all images, samplers and storage buffers") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_METAL_TEXTURE_SLOT_OUT_OF_RANGE, "image 'msl_texture_n' is out of range (must be 0..15)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_METAL_TEXTURE_SLOT_COLLISION, "image 'msl_texture_n' must be unique in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_HLSL_REGISTER_T_OUT_OF_RANGE, "image 'hlsl_register_t_n' is out of range (must be 0..23)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_HLSL_REGISTER_T_COLLISION, "image 'hlsl_register_t_n' must be unique across images and storage buffers in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_WGSL_GROUP1_BINDING_OUT_OF_RANGE, "image 'wgsl_group1_binding_n' is out of range (must be 0..127)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_WGSL_GROUP1_BINDING_COLLISION, "image 'wgsl_group1_binding_n' must be unique across all images, samplers and storage buffers") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_SAMPLER_METAL_SAMPLER_SLOT_OUT_OF_RANGE, "sampler 'msl_sampler_n' is out of range (must be 0..15)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_SAMPLER_METAL_SAMPLER_SLOT_COLLISION, "sampler 'msl_sampler_n' must be unique in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_SAMPLER_HLSL_REGISTER_S_OUT_OF_RANGE, "sampler 'hlsl_register_s_n' is out of rang (must be 0..15)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_SAMPLER_HLSL_REGISTER_S_COLLISION, "sampler 'hlsl_register_s_n' must be unique in same shader stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_SAMPLER_WGSL_GROUP1_BINDING_OUT_OF_RANGE, "sampler 'wgsl_group1_binding_n' is out of range (must be 0..127)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_SAMPLER_WGSL_GROUP1_BINDING_COLLISION, "sampler 'wgsl_group1_binding_n' must be unique across all images, samplers and storage buffers") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_IMAGE_SLOT_OUT_OF_RANGE, "image-sampler-pair image slot index is out of range (sg_shader_desc.image_sampler_pairs[].image_slot)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_SAMPLER_SLOT_OUT_OF_RANGE, "image-sampler-pair sampler slot index is out of range (sg_shader_desc.image_sampler_pairs[].sampler_slot)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_IMAGE_STAGE_MISMATCH, "image-sampler-pair stage doesn't match referenced image stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_SAMPLER_STAGE_MISMATCH, "image-sampler-pair stage doesn't match referenced sampler stage") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_GLSL_NAME, "image-sampler-pair 'glsl_name' missing") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_NONFILTERING_SAMPLER_REQUIRED, "image sample type UNFILTERABLE_FLOAT, UINT, SINT can only be used with NONFILTERING sampler") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_COMPARISON_SAMPLER_REQUIRED, "image sample type DEPTH can only be used with COMPARISON sampler") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_IMAGE_NOT_REFERENCED_BY_IMAGE_SAMPLER_PAIRS, "one or more images are not referenced by by image-sampler-pairs (sg_shader_desc.image_sampler_pairs[].image_slot)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_SAMPLER_NOT_REFERENCED_BY_IMAGE_SAMPLER_PAIRS, "one or more samplers are not referenced by image-sampler-pairs (sg_shader_desc.image_sampler_pairs[].sampler_slot)") \ + _SG_LOGITEM_XMACRO(VALIDATE_SHADERDESC_ATTR_STRING_TOO_LONG, "vertex attribute name/semantic string too long (max len 16)") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_CANARY, "sg_pipeline_desc not initialized") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_SHADER, "sg_pipeline_desc.shader missing or invalid") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_COMPUTE_SHADER_EXPECTED, "sg_pipeline_desc.shader must be a compute shader") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_NO_COMPUTE_SHADER_EXPECTED, "sg_pipeline_desc.compute is false, but shader is a compute shader") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_NO_CONT_ATTRS, "sg_pipeline_desc.layout.attrs is not continuous") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_ATTR_BASETYPE_MISMATCH, "sg_pipeline_desc.layout.attrs[].format is incompatble with sg_shader_desc.attrs[].base_type") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_LAYOUT_STRIDE4, "sg_pipeline_desc.layout.buffers[].stride must be multiple of 4") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_ATTR_SEMANTICS, "D3D11 missing vertex attribute semantics in shader") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_SHADER_READONLY_STORAGEBUFFERS, "sg_pipeline_desc.shader: only readonly storage buffer bindings allowed in render pipelines") \ + _SG_LOGITEM_XMACRO(VALIDATE_PIPELINEDESC_BLENDOP_MINMAX_REQUIRES_BLENDFACTOR_ONE, "SG_BLENDOP_MIN/MAX requires all blend factors to be SG_BLENDFACTOR_ONE") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_CANARY, "sg_attachments_desc not initialized") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_NO_ATTACHMENTS, "sg_attachments_desc no color or depth-stencil attachments") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_NO_CONT_COLOR_ATTS, "color attachments must occupy continuous slots") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_IMAGE, "pass attachment image is not valid") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_MIPLEVEL, "pass attachment mip level is bigger than image has mipmaps") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_FACE, "pass attachment image is cubemap, but face index is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_LAYER, "pass attachment image is array texture, but layer index is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_SLICE, "pass attachment image is 3d texture, but slice value is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_IMAGE_NO_RT, "pass attachment image must be have render_target=true") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_COLOR_INV_PIXELFORMAT, "pass color-attachment images must be renderable color pixel format") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_INV_PIXELFORMAT, "pass depth-attachment image must be depth or depth-stencil pixel format") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_IMAGE_SIZES, "all pass attachments must have the same size") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_IMAGE_SAMPLE_COUNTS, "all pass attachments must have the same sample count") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_COLOR_IMAGE_MSAA, "pass resolve attachments must have a color attachment image with sample count > 1") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE, "pass resolve attachment image not valid") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_SAMPLE_COUNT, "pass resolve attachment image sample count must be 1") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_MIPLEVEL, "pass resolve attachment mip level is bigger than image has mipmaps") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_FACE, "pass resolve attachment is cubemap, but face index is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_LAYER, "pass resolve attachment is array texture, but layer index is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_SLICE, "pass resolve attachment is 3d texture, but slice value is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE_NO_RT, "pass resolve attachment image must have render_target=true") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE_SIZES, "pass resolve attachment size must match color attachment image size") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE_FORMAT, "pass resolve attachment pixel format must match color attachment pixel format") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE, "pass depth attachment image is not valid") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_MIPLEVEL, "pass depth attachment mip level is bigger than image has mipmaps") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_FACE, "pass depth attachment image is cubemap, but face index is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_LAYER, "pass depth attachment image is array texture, but layer index is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_SLICE, "pass depth attachment image is 3d texture, but slice value is too big") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE_NO_RT, "pass depth attachment image must be have render_target=true") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE_SIZES, "pass depth attachment image size must match color attachment image size") \ + _SG_LOGITEM_XMACRO(VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE_SAMPLE_COUNT, "pass depth attachment sample count must match color attachment sample count") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_CANARY, "sg_begin_pass: pass struct not initialized") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_EXPECT_NO_ATTACHMENTS, "sg_begin_pass: compute passes cannot have attachments") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_ATTACHMENTS_EXISTS, "sg_begin_pass: attachments object no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_ATTACHMENTS_VALID, "sg_begin_pass: attachments object not in resource state VALID") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_COLOR_ATTACHMENT_IMAGE, "sg_begin_pass: one or more color attachment images are not valid") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_RESOLVE_ATTACHMENT_IMAGE, "sg_begin_pass: one or more resolve attachment images are not valid") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_DEPTHSTENCIL_ATTACHMENT_IMAGE, "sg_begin_pass: one or more depth-stencil attachment images are not valid") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_WIDTH, "sg_begin_pass: expected pass.swapchain.width > 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_WIDTH_NOTSET, "sg_begin_pass: expected pass.swapchain.width == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_HEIGHT, "sg_begin_pass: expected pass.swapchain.height > 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_HEIGHT_NOTSET, "sg_begin_pass: expected pass.swapchain.height == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_SAMPLECOUNT, "sg_begin_pass: expected pass.swapchain.sample_count > 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_SAMPLECOUNT_NOTSET, "sg_begin_pass: expected pass.swapchain.sample_count == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_COLORFORMAT, "sg_begin_pass: expected pass.swapchain.color_format to be valid") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_COLORFORMAT_NOTSET, "sg_begin_pass: expected pass.swapchain.color_format to be unset") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_DEPTHFORMAT_NOTSET, "sg_begin_pass: expected pass.swapchain.depth_format to be unset") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_CURRENTDRAWABLE, "sg_begin_pass: expected pass.swapchain.metal.current_drawable != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_CURRENTDRAWABLE_NOTSET, "sg_begin_pass: expected pass.swapchain.metal.current_drawable == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_DEPTHSTENCILTEXTURE, "sg_begin_pass: expected pass.swapchain.metal.depth_stencil_texture != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_DEPTHSTENCILTEXTURE_NOTSET, "sg_begin_pass: expected pass.swapchain.metal.depth_stencil_texture == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_MSAACOLORTEXTURE, "sg_begin_pass: expected pass.swapchain.metal.msaa_color_texture != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_MSAACOLORTEXTURE_NOTSET, "sg_begin_pass: expected pass.swapchain.metal.msaa_color_texture == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RENDERVIEW, "sg_begin_pass: expected pass.swapchain.d3d11.render_view != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RENDERVIEW_NOTSET, "sg_begin_pass: expected pass.swapchain.d3d11.render_view == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RESOLVEVIEW, "sg_begin_pass: expected pass.swapchain.d3d11.resolve_view != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RESOLVEVIEW_NOTSET, "sg_begin_pass: expected pass.swapchain.d3d11.resolve_view == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_DEPTHSTENCILVIEW, "sg_begin_pass: expected pass.swapchain.d3d11.depth_stencil_view != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_DEPTHSTENCILVIEW_NOTSET, "sg_begin_pass: expected pass.swapchain.d3d11.depth_stencil_view == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RENDERVIEW, "sg_begin_pass: expected pass.swapchain.wgpu.render_view != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RENDERVIEW_NOTSET, "sg_begin_pass: expected pass.swapchain.wgpu.render_view == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RESOLVEVIEW, "sg_begin_pass: expected pass.swapchain.wgpu.resolve_view != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RESOLVEVIEW_NOTSET, "sg_begin_pass: expected pass.swapchain.wgpu.resolve_view == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_DEPTHSTENCILVIEW, "sg_begin_pass: expected pass.swapchain.wgpu.depth_stencil_view != 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_DEPTHSTENCILVIEW_NOTSET, "sg_begin_pass: expected pass.swapchain.wgpu.depth_stencil_view == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_BEGINPASS_SWAPCHAIN_GL_EXPECT_FRAMEBUFFER_NOTSET, "sg_begin_pass: expected pass.swapchain.gl.framebuffer == 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_AVP_RENDERPASS_EXPECTED, "sg_apply_viewport: must be called in a render pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_ASR_RENDERPASS_EXPECTED, "sg_apply_scissor_rect: must be called in a render pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_PIPELINE_VALID_ID, "sg_apply_pipeline: invalid pipeline id provided") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_PIPELINE_EXISTS, "sg_apply_pipeline: pipeline object no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_PIPELINE_VALID, "sg_apply_pipeline: pipeline object not in valid state") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_PASS_EXPECTED, "sg_apply_pipeline: must be called in a pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_SHADER_EXISTS, "sg_apply_pipeline: shader object no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_SHADER_VALID, "sg_apply_pipeline: shader object not in valid state") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_COMPUTEPASS_EXPECTED, "sg_apply_pipeline: trying to apply compute pipeline in render pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_RENDERPASS_EXPECTED, "sg_apply_pipeline: trying to apply render pipeline in compute pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_CURPASS_ATTACHMENTS_EXISTS, "sg_apply_pipeline: current pass attachments no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_CURPASS_ATTACHMENTS_VALID, "sg_apply_pipeline: current pass attachments not in valid state") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_ATT_COUNT, "sg_apply_pipeline: number of pipeline color attachments doesn't match number of pass color attachments") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_COLOR_FORMAT, "sg_apply_pipeline: pipeline color attachment pixel format doesn't match pass color attachment pixel format") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_DEPTH_FORMAT, "sg_apply_pipeline: pipeline depth pixel_format doesn't match pass depth attachment pixel format") \ + _SG_LOGITEM_XMACRO(VALIDATE_APIP_SAMPLE_COUNT, "sg_apply_pipeline: pipeline MSAA sample count doesn't match render pass attachment sample count") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_PASS_EXPECTED, "sg_apply_bindings: must be called in a pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EMPTY_BINDINGS, "sg_apply_bindings: the provided sg_bindings struct is empty") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_PIPELINE, "sg_apply_bindings: must be called after sg_apply_pipeline") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_PIPELINE_EXISTS, "sg_apply_bindings: currently applied pipeline object no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_PIPELINE_VALID, "sg_apply_bindings: currently applied pipeline object not in valid state") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_COMPUTE_EXPECTED_NO_VBS, "sg_apply_bindings: vertex buffer bindings not allowed in a compute pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_COMPUTE_EXPECTED_NO_IB, "sg_apply_bindings: index buffer binding not allowed in compute pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_VB, "sg_apply_bindings: vertex buffer binding is missing or buffer handle is invalid") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_VB_EXISTS, "sg_apply_bindings: vertex buffer no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_VB_TYPE, "sg_apply_bindings: buffer in vertex buffer slot is not a SG_BUFFERTYPE_VERTEXBUFFER") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_VB_OVERFLOW, "sg_apply_bindings: buffer in vertex buffer slot is overflown") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_NO_IB, "sg_apply_bindings: pipeline object defines indexed rendering, but no index buffer provided") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_IB, "sg_apply_bindings: pipeline object defines non-indexed rendering, but index buffer provided") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_IB_EXISTS, "sg_apply_bindings: index buffer no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_IB_TYPE, "sg_apply_bindings: buffer in index buffer slot is not a SG_BUFFERTYPE_INDEXBUFFER") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_IB_OVERFLOW, "sg_apply_bindings: buffer in index buffer slot is overflown") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_IMAGE_BINDING, "sg_apply_bindings: image binding is missing or the image handle is invalid") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_IMG_EXISTS, "sg_apply_bindings: bound image no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_IMAGE_TYPE_MISMATCH, "sg_apply_bindings: type of bound image doesn't match shader desc") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_MULTISAMPLED_IMAGE, "sg_apply_bindings: expected image with sample_count > 1") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_IMAGE_MSAA, "sg_apply_bindings: cannot bind image with sample_count>1") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_FILTERABLE_IMAGE, "sg_apply_bindings: filterable image expected") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_DEPTH_IMAGE, "sg_apply_bindings: depth image expected") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_SAMPLER_BINDING, "sg_apply_bindings: sampler binding is missing or the sampler handle is invalid") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_UNEXPECTED_SAMPLER_COMPARE_NEVER, "sg_apply_bindings: shader expects SG_SAMPLERTYPE_COMPARISON but sampler has SG_COMPAREFUNC_NEVER") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_SAMPLER_COMPARE_NEVER, "sg_apply_bindings: shader expects SG_SAMPLERTYPE_FILTERING or SG_SAMPLERTYPE_NONFILTERING but sampler doesn't have SG_COMPAREFUNC_NEVER") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_NONFILTERING_SAMPLER, "sg_apply_bindings: shader expected SG_SAMPLERTYPE_NONFILTERING, but sampler has SG_FILTER_LINEAR filters") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_SMP_EXISTS, "sg_apply_bindings: bound sampler no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_EXPECTED_STORAGEBUFFER_BINDING, "sg_apply_bindings: storage buffer binding is missing or the buffer handle is invalid") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_STORAGEBUFFER_EXISTS, "sg_apply_bindings: bound storage buffer no longer alive") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_STORAGEBUFFER_BINDING_BUFFERTYPE, "sg_apply_bindings: buffer bound to storage buffer slot is not of type storage buffer") \ + _SG_LOGITEM_XMACRO(VALIDATE_ABND_STORAGEBUFFER_READWRITE_IMMUTABLE, "sg_apply_bindings: storage buffers bound as read/write must have usage immutable") \ + _SG_LOGITEM_XMACRO(VALIDATE_AU_PASS_EXPECTED, "sg_apply_uniforms: must be called in a pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_AU_NO_PIPELINE, "sg_apply_uniforms: must be called after sg_apply_pipeline()") \ + _SG_LOGITEM_XMACRO(VALIDATE_AU_NO_UNIFORMBLOCK_AT_SLOT, "sg_apply_uniforms: no uniform block declaration at this shader stage UB slot") \ + _SG_LOGITEM_XMACRO(VALIDATE_AU_SIZE, "sg_apply_uniforms: data size doesn't match declared uniform block size") \ + _SG_LOGITEM_XMACRO(VALIDATE_DRAW_RENDERPASS_EXPECTED, "sg_draw: must be called in a render pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_DRAW_BASEELEMENT, "sg_draw: base_element cannot be < 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_DRAW_NUMELEMENTS, "sg_draw: num_elements cannot be < 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_DRAW_NUMINSTANCES, "sg_draw: num_instances cannot be < 0") \ + _SG_LOGITEM_XMACRO(VALIDATE_DRAW_REQUIRED_BINDINGS_OR_UNIFORMS_MISSING, "sg_draw: call to sg_apply_bindings() and/or sg_apply_uniforms() missing after sg_apply_pipeline()") \ + _SG_LOGITEM_XMACRO(VALIDATE_DISPATCH_COMPUTEPASS_EXPECTED, "sg_dispatch: must be called in a compute pass") \ + _SG_LOGITEM_XMACRO(VALIDATE_DISPATCH_NUMGROUPSX, "sg_dispatch: num_groups_x must be >=0 and <65536") \ + _SG_LOGITEM_XMACRO(VALIDATE_DISPATCH_NUMGROUPSY, "sg_dispatch: num_groups_y must be >=0 and <65536") \ + _SG_LOGITEM_XMACRO(VALIDATE_DISPATCH_NUMGROUPSZ, "sg_dispatch: num_groups_z must be >=0 and <65536") \ + _SG_LOGITEM_XMACRO(VALIDATE_DISPATCH_REQUIRED_BINDINGS_OR_UNIFORMS_MISSING, "sg_dispatch: call to sg_apply_bindings() and/or sg_apply_uniforms() missing after sg_apply_pipeline()") \ + _SG_LOGITEM_XMACRO(VALIDATE_UPDATEBUF_USAGE, "sg_update_buffer: cannot update immutable buffer") \ + _SG_LOGITEM_XMACRO(VALIDATE_UPDATEBUF_SIZE, "sg_update_buffer: update size is bigger than buffer size") \ + _SG_LOGITEM_XMACRO(VALIDATE_UPDATEBUF_ONCE, "sg_update_buffer: only one update allowed per buffer and frame") \ + _SG_LOGITEM_XMACRO(VALIDATE_UPDATEBUF_APPEND, "sg_update_buffer: cannot call sg_update_buffer and sg_append_buffer in same frame") \ + _SG_LOGITEM_XMACRO(VALIDATE_APPENDBUF_USAGE, "sg_append_buffer: cannot append to immutable buffer") \ + _SG_LOGITEM_XMACRO(VALIDATE_APPENDBUF_SIZE, "sg_append_buffer: overall appended size is bigger than buffer size") \ + _SG_LOGITEM_XMACRO(VALIDATE_APPENDBUF_UPDATE, "sg_append_buffer: cannot call sg_append_buffer and sg_update_buffer in same frame") \ + _SG_LOGITEM_XMACRO(VALIDATE_UPDIMG_USAGE, "sg_update_image: cannot update immutable image") \ + _SG_LOGITEM_XMACRO(VALIDATE_UPDIMG_ONCE, "sg_update_image: only one update allowed per image and frame") \ + _SG_LOGITEM_XMACRO(VALIDATION_FAILED, "validation layer checks failed") \ + +#define _SG_LOGITEM_XMACRO(item,msg) SG_LOGITEM_##item, +typedef enum sg_log_item { + _SG_LOG_ITEMS +} sg_log_item; +#undef _SG_LOGITEM_XMACRO + +/* + sg_desc + + The sg_desc struct contains configuration values for sokol_gfx, + it is used as parameter to the sg_setup() call. + + The default configuration is: + + .buffer_pool_size 128 + .image_pool_size 128 + .sampler_pool_size 64 + .shader_pool_size 32 + .pipeline_pool_size 64 + .attachments_pool_size 16 + .uniform_buffer_size 4 MB (4*1024*1024) + .max_dispatch_calls_per_pass 1024 + .max_commit_listeners 1024 + .disable_validation false + .mtl_force_managed_storage_mode false + .wgpu_disable_bindgroups_cache false + .wgpu_bindgroups_cache_size 1024 + + .allocator.alloc_fn 0 (in this case, malloc() will be called) + .allocator.free_fn 0 (in this case, free() will be called) + .allocator.user_data 0 + + .environment.defaults.color_format: default value depends on selected backend: + all GL backends: SG_PIXELFORMAT_RGBA8 + Metal and D3D11: SG_PIXELFORMAT_BGRA8 + WebGPU: *no default* (must be queried from WebGPU swapchain object) + .environment.defaults.depth_format: SG_PIXELFORMAT_DEPTH_STENCIL + .environment.defaults.sample_count: 1 + + Metal specific: + (NOTE: All Objective-C object references are transferred through + a bridged cast (__bridge const void*) to sokol_gfx, which will use an + unretained bridged cast (__bridge id) to retrieve the Objective-C + references back. Since the bridge cast is unretained, the caller + must hold a strong reference to the Objective-C object until sg_setup() + returns. + + .mtl_force_managed_storage_mode + when enabled, Metal buffers and texture resources are created in managed storage + mode, otherwise sokol-gfx will decide whether to create buffers and + textures in managed or shared storage mode (this is mainly a debugging option) + .mtl_use_command_buffer_with_retained_references + when true, the sokol-gfx Metal backend will use Metal command buffers which + bump the reference count of resource objects as long as they are inflight, + this is slower than the default command-buffer-with-unretained-references + method, this may be a workaround when confronted with lifetime validation + errors from the Metal validation layer until a proper fix has been implemented + .environment.metal.device + a pointer to the MTLDevice object + + D3D11 specific: + .environment.d3d11.device + a pointer to the ID3D11Device object, this must have been created + before sg_setup() is called + .environment.d3d11.device_context + a pointer to the ID3D11DeviceContext object + .d3d11_shader_debugging + set this to true to compile shaders which are provided as HLSL source + code with debug information and without optimization, this allows + shader debugging in tools like RenderDoc, to output source code + instead of byte code from sokol-shdc, omit the `--binary` cmdline + option + + WebGPU specific: + .wgpu_disable_bindgroups_cache + When this is true, the WebGPU backend will create and immediately + release a BindGroup object in the sg_apply_bindings() call, only + use this for debugging purposes. + .wgpu_bindgroups_cache_size + The size of the bindgroups cache for re-using BindGroup objects + between sg_apply_bindings() calls. The smaller the cache size, + the more likely are cache slot collisions which will cause + a BindGroups object to be destroyed and a new one created. + Use the information returned by sg_query_stats() to check + if this is a frequent occurrence, and increase the cache size as + needed (the default is 1024). + NOTE: wgpu_bindgroups_cache_size must be a power-of-2 number! + .environment.wgpu.device + a WGPUDevice handle + + When using sokol_gfx.h and sokol_app.h together, consider using the + helper function sglue_environment() in the sokol_glue.h header to + initialize the sg_desc.environment nested struct. sglue_environment() returns + a completely initialized sg_environment struct with information + provided by sokol_app.h. +*/ +typedef struct sg_environment_defaults { + sg_pixel_format color_format; + sg_pixel_format depth_format; + int sample_count; +} sg_environment_defaults; + +typedef struct sg_metal_environment { + const void* device; +} sg_metal_environment; + +typedef struct sg_d3d11_environment { + const void* device; + const void* device_context; +} sg_d3d11_environment; + +typedef struct sg_wgpu_environment { + const void* device; +} sg_wgpu_environment; + +typedef struct sg_environment { + sg_environment_defaults defaults; + sg_metal_environment metal; + sg_d3d11_environment d3d11; + sg_wgpu_environment wgpu; +} sg_environment; + +/* + sg_commit_listener + + Used with function sg_add_commit_listener() to add a callback + which will be called in sg_commit(). This is useful for libraries + building on top of sokol-gfx to be notified about when a frame + ends (instead of having to guess, or add a manual 'new-frame' + function. +*/ +typedef struct sg_commit_listener { + void (*func)(void* user_data); + void* user_data; +} sg_commit_listener; + +/* + sg_allocator + + Used in sg_desc to provide custom memory-alloc and -free functions + to sokol_gfx.h. If memory management should be overridden, both the + alloc_fn and free_fn function must be provided (e.g. it's not valid to + override one function but not the other). +*/ +typedef struct sg_allocator { + void* (*alloc_fn)(size_t size, void* user_data); + void (*free_fn)(void* ptr, void* user_data); + void* user_data; +} sg_allocator; + +/* + sg_logger + + Used in sg_desc to provide a logging function. Please be aware + that without logging function, sokol-gfx will be completely + silent, e.g. it will not report errors, warnings and + validation layer messages. For maximum error verbosity, + compile in debug mode (e.g. NDEBUG *not* defined) and provide a + compatible logger function in the sg_setup() call + (for instance the standard logging function from sokol_log.h). +*/ +typedef struct sg_logger { + void (*func)( + const char* tag, // always "sg" + uint32_t log_level, // 0=panic, 1=error, 2=warning, 3=info + uint32_t log_item_id, // SG_LOGITEM_* + const char* message_or_null, // a message string, may be nullptr in release mode + uint32_t line_nr, // line number in sokol_gfx.h + const char* filename_or_null, // source filename, may be nullptr in release mode + void* user_data); + void* user_data; +} sg_logger; + +typedef struct sg_desc { + uint32_t _start_canary; + int buffer_pool_size; + int image_pool_size; + int sampler_pool_size; + int shader_pool_size; + int pipeline_pool_size; + int attachments_pool_size; + int uniform_buffer_size; + int max_dispatch_calls_per_pass; // max expected number of dispatch calls per pass (default: 1024) + int max_commit_listeners; + bool disable_validation; // disable validation layer even in debug mode, useful for tests + bool d3d11_shader_debugging; // if true, HLSL shaders are compiled with D3DCOMPILE_DEBUG | D3DCOMPILE_SKIP_OPTIMIZATION + bool mtl_force_managed_storage_mode; // for debugging: use Metal managed storage mode for resources even with UMA + bool mtl_use_command_buffer_with_retained_references; // Metal: use a managed MTLCommandBuffer which ref-counts used resources + bool wgpu_disable_bindgroups_cache; // set to true to disable the WebGPU backend BindGroup cache + int wgpu_bindgroups_cache_size; // number of slots in the WebGPU bindgroup cache (must be 2^N) + sg_allocator allocator; + sg_logger logger; // optional log function override + sg_environment environment; + uint32_t _end_canary; +} sg_desc; + +// setup and misc functions +SOKOL_GFX_API_DECL void sg_setup(const sg_desc* desc); +SOKOL_GFX_API_DECL void sg_shutdown(void); +SOKOL_GFX_API_DECL bool sg_isvalid(void); +SOKOL_GFX_API_DECL void sg_reset_state_cache(void); +SOKOL_GFX_API_DECL sg_trace_hooks sg_install_trace_hooks(const sg_trace_hooks* trace_hooks); +SOKOL_GFX_API_DECL void sg_push_debug_group(const char* name); +SOKOL_GFX_API_DECL void sg_pop_debug_group(void); +SOKOL_GFX_API_DECL bool sg_add_commit_listener(sg_commit_listener listener); +SOKOL_GFX_API_DECL bool sg_remove_commit_listener(sg_commit_listener listener); + +// resource creation, destruction and updating +SOKOL_GFX_API_DECL sg_buffer sg_make_buffer(const sg_buffer_desc* desc); +SOKOL_GFX_API_DECL sg_image sg_make_image(const sg_image_desc* desc); +SOKOL_GFX_API_DECL sg_sampler sg_make_sampler(const sg_sampler_desc* desc); +SOKOL_GFX_API_DECL sg_shader sg_make_shader(const sg_shader_desc* desc); +SOKOL_GFX_API_DECL sg_pipeline sg_make_pipeline(const sg_pipeline_desc* desc); +SOKOL_GFX_API_DECL sg_attachments sg_make_attachments(const sg_attachments_desc* desc); +SOKOL_GFX_API_DECL void sg_destroy_buffer(sg_buffer buf); +SOKOL_GFX_API_DECL void sg_destroy_image(sg_image img); +SOKOL_GFX_API_DECL void sg_destroy_sampler(sg_sampler smp); +SOKOL_GFX_API_DECL void sg_destroy_shader(sg_shader shd); +SOKOL_GFX_API_DECL void sg_destroy_pipeline(sg_pipeline pip); +SOKOL_GFX_API_DECL void sg_destroy_attachments(sg_attachments atts); +SOKOL_GFX_API_DECL void sg_update_buffer(sg_buffer buf, const sg_range* data); +SOKOL_GFX_API_DECL void sg_update_image(sg_image img, const sg_image_data* data); +SOKOL_GFX_API_DECL int sg_append_buffer(sg_buffer buf, const sg_range* data); +SOKOL_GFX_API_DECL bool sg_query_buffer_overflow(sg_buffer buf); +SOKOL_GFX_API_DECL bool sg_query_buffer_will_overflow(sg_buffer buf, size_t size); + +// render and compute functions +SOKOL_GFX_API_DECL void sg_begin_pass(const sg_pass* pass); +SOKOL_GFX_API_DECL void sg_apply_viewport(int x, int y, int width, int height, bool origin_top_left); +SOKOL_GFX_API_DECL void sg_apply_viewportf(float x, float y, float width, float height, bool origin_top_left); +SOKOL_GFX_API_DECL void sg_apply_scissor_rect(int x, int y, int width, int height, bool origin_top_left); +SOKOL_GFX_API_DECL void sg_apply_scissor_rectf(float x, float y, float width, float height, bool origin_top_left); +SOKOL_GFX_API_DECL void sg_apply_pipeline(sg_pipeline pip); +SOKOL_GFX_API_DECL void sg_apply_bindings(const sg_bindings* bindings); +SOKOL_GFX_API_DECL void sg_apply_uniforms(int ub_slot, const sg_range* data); +SOKOL_GFX_API_DECL void sg_draw(int base_element, int num_elements, int num_instances); +SOKOL_GFX_API_DECL void sg_dispatch(int num_groups_x, int num_groups_y, int num_groups_z); +SOKOL_GFX_API_DECL void sg_end_pass(void); +SOKOL_GFX_API_DECL void sg_commit(void); + +// getting information +SOKOL_GFX_API_DECL sg_desc sg_query_desc(void); +SOKOL_GFX_API_DECL sg_backend sg_query_backend(void); +SOKOL_GFX_API_DECL sg_features sg_query_features(void); +SOKOL_GFX_API_DECL sg_limits sg_query_limits(void); +SOKOL_GFX_API_DECL sg_pixelformat_info sg_query_pixelformat(sg_pixel_format fmt); +SOKOL_GFX_API_DECL int sg_query_row_pitch(sg_pixel_format fmt, int width, int row_align_bytes); +SOKOL_GFX_API_DECL int sg_query_surface_pitch(sg_pixel_format fmt, int width, int height, int row_align_bytes); +// get current state of a resource (INITIAL, ALLOC, VALID, FAILED, INVALID) +SOKOL_GFX_API_DECL sg_resource_state sg_query_buffer_state(sg_buffer buf); +SOKOL_GFX_API_DECL sg_resource_state sg_query_image_state(sg_image img); +SOKOL_GFX_API_DECL sg_resource_state sg_query_sampler_state(sg_sampler smp); +SOKOL_GFX_API_DECL sg_resource_state sg_query_shader_state(sg_shader shd); +SOKOL_GFX_API_DECL sg_resource_state sg_query_pipeline_state(sg_pipeline pip); +SOKOL_GFX_API_DECL sg_resource_state sg_query_attachments_state(sg_attachments atts); +// get runtime information about a resource +SOKOL_GFX_API_DECL sg_buffer_info sg_query_buffer_info(sg_buffer buf); +SOKOL_GFX_API_DECL sg_image_info sg_query_image_info(sg_image img); +SOKOL_GFX_API_DECL sg_sampler_info sg_query_sampler_info(sg_sampler smp); +SOKOL_GFX_API_DECL sg_shader_info sg_query_shader_info(sg_shader shd); +SOKOL_GFX_API_DECL sg_pipeline_info sg_query_pipeline_info(sg_pipeline pip); +SOKOL_GFX_API_DECL sg_attachments_info sg_query_attachments_info(sg_attachments atts); +// get desc structs matching a specific resource (NOTE that not all creation attributes may be provided) +SOKOL_GFX_API_DECL sg_buffer_desc sg_query_buffer_desc(sg_buffer buf); +SOKOL_GFX_API_DECL sg_image_desc sg_query_image_desc(sg_image img); +SOKOL_GFX_API_DECL sg_sampler_desc sg_query_sampler_desc(sg_sampler smp); +SOKOL_GFX_API_DECL sg_shader_desc sg_query_shader_desc(sg_shader shd); +SOKOL_GFX_API_DECL sg_pipeline_desc sg_query_pipeline_desc(sg_pipeline pip); +SOKOL_GFX_API_DECL sg_attachments_desc sg_query_attachments_desc(sg_attachments atts); +// get resource creation desc struct with their default values replaced +SOKOL_GFX_API_DECL sg_buffer_desc sg_query_buffer_defaults(const sg_buffer_desc* desc); +SOKOL_GFX_API_DECL sg_image_desc sg_query_image_defaults(const sg_image_desc* desc); +SOKOL_GFX_API_DECL sg_sampler_desc sg_query_sampler_defaults(const sg_sampler_desc* desc); +SOKOL_GFX_API_DECL sg_shader_desc sg_query_shader_defaults(const sg_shader_desc* desc); +SOKOL_GFX_API_DECL sg_pipeline_desc sg_query_pipeline_defaults(const sg_pipeline_desc* desc); +SOKOL_GFX_API_DECL sg_attachments_desc sg_query_attachments_defaults(const sg_attachments_desc* desc); +// assorted query functions +SOKOL_GFX_API_DECL size_t sg_query_buffer_size(sg_buffer buf); +SOKOL_GFX_API_DECL sg_buffer_type sg_query_buffer_type(sg_buffer buf); +SOKOL_GFX_API_DECL sg_usage sg_query_buffer_usage(sg_buffer buf); +SOKOL_GFX_API_DECL sg_image_type sg_query_image_type(sg_image img); +SOKOL_GFX_API_DECL int sg_query_image_width(sg_image img); +SOKOL_GFX_API_DECL int sg_query_image_height(sg_image img); +SOKOL_GFX_API_DECL int sg_query_image_num_slices(sg_image img); +SOKOL_GFX_API_DECL int sg_query_image_num_mipmaps(sg_image img); +SOKOL_GFX_API_DECL sg_pixel_format sg_query_image_pixelformat(sg_image img); +SOKOL_GFX_API_DECL sg_usage sg_query_image_usage(sg_image img); +SOKOL_GFX_API_DECL int sg_query_image_sample_count(sg_image img); + +// separate resource allocation and initialization (for async setup) +SOKOL_GFX_API_DECL sg_buffer sg_alloc_buffer(void); +SOKOL_GFX_API_DECL sg_image sg_alloc_image(void); +SOKOL_GFX_API_DECL sg_sampler sg_alloc_sampler(void); +SOKOL_GFX_API_DECL sg_shader sg_alloc_shader(void); +SOKOL_GFX_API_DECL sg_pipeline sg_alloc_pipeline(void); +SOKOL_GFX_API_DECL sg_attachments sg_alloc_attachments(void); +SOKOL_GFX_API_DECL void sg_dealloc_buffer(sg_buffer buf); +SOKOL_GFX_API_DECL void sg_dealloc_image(sg_image img); +SOKOL_GFX_API_DECL void sg_dealloc_sampler(sg_sampler smp); +SOKOL_GFX_API_DECL void sg_dealloc_shader(sg_shader shd); +SOKOL_GFX_API_DECL void sg_dealloc_pipeline(sg_pipeline pip); +SOKOL_GFX_API_DECL void sg_dealloc_attachments(sg_attachments attachments); +SOKOL_GFX_API_DECL void sg_init_buffer(sg_buffer buf, const sg_buffer_desc* desc); +SOKOL_GFX_API_DECL void sg_init_image(sg_image img, const sg_image_desc* desc); +SOKOL_GFX_API_DECL void sg_init_sampler(sg_sampler smg, const sg_sampler_desc* desc); +SOKOL_GFX_API_DECL void sg_init_shader(sg_shader shd, const sg_shader_desc* desc); +SOKOL_GFX_API_DECL void sg_init_pipeline(sg_pipeline pip, const sg_pipeline_desc* desc); +SOKOL_GFX_API_DECL void sg_init_attachments(sg_attachments attachments, const sg_attachments_desc* desc); +SOKOL_GFX_API_DECL void sg_uninit_buffer(sg_buffer buf); +SOKOL_GFX_API_DECL void sg_uninit_image(sg_image img); +SOKOL_GFX_API_DECL void sg_uninit_sampler(sg_sampler smp); +SOKOL_GFX_API_DECL void sg_uninit_shader(sg_shader shd); +SOKOL_GFX_API_DECL void sg_uninit_pipeline(sg_pipeline pip); +SOKOL_GFX_API_DECL void sg_uninit_attachments(sg_attachments atts); +SOKOL_GFX_API_DECL void sg_fail_buffer(sg_buffer buf); +SOKOL_GFX_API_DECL void sg_fail_image(sg_image img); +SOKOL_GFX_API_DECL void sg_fail_sampler(sg_sampler smp); +SOKOL_GFX_API_DECL void sg_fail_shader(sg_shader shd); +SOKOL_GFX_API_DECL void sg_fail_pipeline(sg_pipeline pip); +SOKOL_GFX_API_DECL void sg_fail_attachments(sg_attachments atts); + +// frame stats +SOKOL_GFX_API_DECL void sg_enable_frame_stats(void); +SOKOL_GFX_API_DECL void sg_disable_frame_stats(void); +SOKOL_GFX_API_DECL bool sg_frame_stats_enabled(void); +SOKOL_GFX_API_DECL sg_frame_stats sg_query_frame_stats(void); + +/* Backend-specific structs and functions, these may come in handy for mixing + sokol-gfx rendering with 'native backend' rendering functions. + + This group of functions will be expanded as needed. +*/ + +typedef struct sg_d3d11_buffer_info { + const void* buf; // ID3D11Buffer* +} sg_d3d11_buffer_info; + +typedef struct sg_d3d11_image_info { + const void* tex2d; // ID3D11Texture2D* + const void* tex3d; // ID3D11Texture3D* + const void* res; // ID3D11Resource* (either tex2d or tex3d) + const void* srv; // ID3D11ShaderResourceView* +} sg_d3d11_image_info; + +typedef struct sg_d3d11_sampler_info { + const void* smp; // ID3D11SamplerState* +} sg_d3d11_sampler_info; + +typedef struct sg_d3d11_shader_info { + const void* cbufs[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; // ID3D11Buffer* (constant buffers by bind slot) + const void* vs; // ID3D11VertexShader* + const void* fs; // ID3D11PixelShader* +} sg_d3d11_shader_info; + +typedef struct sg_d3d11_pipeline_info { + const void* il; // ID3D11InputLayout* + const void* rs; // ID3D11RasterizerState* + const void* dss; // ID3D11DepthStencilState* + const void* bs; // ID3D11BlendState* +} sg_d3d11_pipeline_info; + +typedef struct sg_d3d11_attachments_info { + const void* color_rtv[SG_MAX_COLOR_ATTACHMENTS]; // ID3D11RenderTargetView + const void* resolve_rtv[SG_MAX_COLOR_ATTACHMENTS]; // ID3D11RenderTargetView + const void* dsv; // ID3D11DepthStencilView +} sg_d3d11_attachments_info; + +typedef struct sg_mtl_buffer_info { + const void* buf[SG_NUM_INFLIGHT_FRAMES]; // id + int active_slot; +} sg_mtl_buffer_info; + +typedef struct sg_mtl_image_info { + const void* tex[SG_NUM_INFLIGHT_FRAMES]; // id + int active_slot; +} sg_mtl_image_info; + +typedef struct sg_mtl_sampler_info { + const void* smp; // id +} sg_mtl_sampler_info; + +typedef struct sg_mtl_shader_info { + const void* vertex_lib; // id + const void* fragment_lib; // id + const void* vertex_func; // id + const void* fragment_func; // id +} sg_mtl_shader_info; + +typedef struct sg_mtl_pipeline_info { + const void* rps; // id + const void* dss; // id +} sg_mtl_pipeline_info; + +typedef struct sg_wgpu_buffer_info { + const void* buf; // WGPUBuffer +} sg_wgpu_buffer_info; + +typedef struct sg_wgpu_image_info { + const void* tex; // WGPUTexture + const void* view; // WGPUTextureView +} sg_wgpu_image_info; + +typedef struct sg_wgpu_sampler_info { + const void* smp; // WGPUSampler +} sg_wgpu_sampler_info; + +typedef struct sg_wgpu_shader_info { + const void* vs_mod; // WGPUShaderModule + const void* fs_mod; // WGPUShaderModule + const void* bgl; // WGPUBindGroupLayout; +} sg_wgpu_shader_info; + +typedef struct sg_wgpu_pipeline_info { + const void* render_pipeline; // WGPURenderPipeline + const void* compute_pipeline; // WGPUComputePipeline +} sg_wgpu_pipeline_info; + +typedef struct sg_wgpu_attachments_info { + const void* color_view[SG_MAX_COLOR_ATTACHMENTS]; // WGPUTextureView + const void* resolve_view[SG_MAX_COLOR_ATTACHMENTS]; // WGPUTextureView + const void* ds_view; // WGPUTextureView +} sg_wgpu_attachments_info; + +typedef struct sg_gl_buffer_info { + uint32_t buf[SG_NUM_INFLIGHT_FRAMES]; + int active_slot; +} sg_gl_buffer_info; + +typedef struct sg_gl_image_info { + uint32_t tex[SG_NUM_INFLIGHT_FRAMES]; + uint32_t tex_target; + uint32_t msaa_render_buffer; + int active_slot; +} sg_gl_image_info; + +typedef struct sg_gl_sampler_info { + uint32_t smp; +} sg_gl_sampler_info; + +typedef struct sg_gl_shader_info { + uint32_t prog; +} sg_gl_shader_info; + +typedef struct sg_gl_attachments_info { + uint32_t framebuffer; + uint32_t msaa_resolve_framebuffer[SG_MAX_COLOR_ATTACHMENTS]; +} sg_gl_attachments_info; + +// D3D11: return ID3D11Device +SOKOL_GFX_API_DECL const void* sg_d3d11_device(void); +// D3D11: return ID3D11DeviceContext +SOKOL_GFX_API_DECL const void* sg_d3d11_device_context(void); +// D3D11: get internal buffer resource objects +SOKOL_GFX_API_DECL sg_d3d11_buffer_info sg_d3d11_query_buffer_info(sg_buffer buf); +// D3D11: get internal image resource objects +SOKOL_GFX_API_DECL sg_d3d11_image_info sg_d3d11_query_image_info(sg_image img); +// D3D11: get internal sampler resource objects +SOKOL_GFX_API_DECL sg_d3d11_sampler_info sg_d3d11_query_sampler_info(sg_sampler smp); +// D3D11: get internal shader resource objects +SOKOL_GFX_API_DECL sg_d3d11_shader_info sg_d3d11_query_shader_info(sg_shader shd); +// D3D11: get internal pipeline resource objects +SOKOL_GFX_API_DECL sg_d3d11_pipeline_info sg_d3d11_query_pipeline_info(sg_pipeline pip); +// D3D11: get internal pass resource objects +SOKOL_GFX_API_DECL sg_d3d11_attachments_info sg_d3d11_query_attachments_info(sg_attachments atts); + +// Metal: return __bridge-casted MTLDevice +SOKOL_GFX_API_DECL const void* sg_mtl_device(void); +// Metal: return __bridge-casted MTLRenderCommandEncoder when inside render pass (otherwise zero) +SOKOL_GFX_API_DECL const void* sg_mtl_render_command_encoder(void); +// Metal: return __bridge-casted MTLComputeCommandEncoder when inside compute pass (otherwise zero) +SOKOL_GFX_API_DECL const void* sg_mtl_compute_command_encoder(void); +// Metal: get internal __bridge-casted buffer resource objects +SOKOL_GFX_API_DECL sg_mtl_buffer_info sg_mtl_query_buffer_info(sg_buffer buf); +// Metal: get internal __bridge-casted image resource objects +SOKOL_GFX_API_DECL sg_mtl_image_info sg_mtl_query_image_info(sg_image img); +// Metal: get internal __bridge-casted sampler resource objects +SOKOL_GFX_API_DECL sg_mtl_sampler_info sg_mtl_query_sampler_info(sg_sampler smp); +// Metal: get internal __bridge-casted shader resource objects +SOKOL_GFX_API_DECL sg_mtl_shader_info sg_mtl_query_shader_info(sg_shader shd); +// Metal: get internal __bridge-casted pipeline resource objects +SOKOL_GFX_API_DECL sg_mtl_pipeline_info sg_mtl_query_pipeline_info(sg_pipeline pip); + +// WebGPU: return WGPUDevice object +SOKOL_GFX_API_DECL const void* sg_wgpu_device(void); +// WebGPU: return WGPUQueue object +SOKOL_GFX_API_DECL const void* sg_wgpu_queue(void); +// WebGPU: return this frame's WGPUCommandEncoder +SOKOL_GFX_API_DECL const void* sg_wgpu_command_encoder(void); +// WebGPU: return WGPURenderPassEncoder of current pass (returns 0 when outside pass or in a compute pass) +SOKOL_GFX_API_DECL const void* sg_wgpu_render_pass_encoder(void); +// WebGPU: return WGPUComputePassEncoder of current pass (returns 0 when outside pass or in a render pass) +SOKOL_GFX_API_DECL const void* sg_wgpu_compute_pass_encoder(void); +// WebGPU: get internal buffer resource objects +SOKOL_GFX_API_DECL sg_wgpu_buffer_info sg_wgpu_query_buffer_info(sg_buffer buf); +// WebGPU: get internal image resource objects +SOKOL_GFX_API_DECL sg_wgpu_image_info sg_wgpu_query_image_info(sg_image img); +// WebGPU: get internal sampler resource objects +SOKOL_GFX_API_DECL sg_wgpu_sampler_info sg_wgpu_query_sampler_info(sg_sampler smp); +// WebGPU: get internal shader resource objects +SOKOL_GFX_API_DECL sg_wgpu_shader_info sg_wgpu_query_shader_info(sg_shader shd); +// WebGPU: get internal pipeline resource objects +SOKOL_GFX_API_DECL sg_wgpu_pipeline_info sg_wgpu_query_pipeline_info(sg_pipeline pip); +// WebGPU: get internal pass resource objects +SOKOL_GFX_API_DECL sg_wgpu_attachments_info sg_wgpu_query_attachments_info(sg_attachments atts); + +// GL: get internal buffer resource objects +SOKOL_GFX_API_DECL sg_gl_buffer_info sg_gl_query_buffer_info(sg_buffer buf); +// GL: get internal image resource objects +SOKOL_GFX_API_DECL sg_gl_image_info sg_gl_query_image_info(sg_image img); +// GL: get internal sampler resource objects +SOKOL_GFX_API_DECL sg_gl_sampler_info sg_gl_query_sampler_info(sg_sampler smp); +// GL: get internal shader resource objects +SOKOL_GFX_API_DECL sg_gl_shader_info sg_gl_query_shader_info(sg_shader shd); +// GL: get internal pass resource objects +SOKOL_GFX_API_DECL sg_gl_attachments_info sg_gl_query_attachments_info(sg_attachments atts); + +#ifdef __cplusplus +} // extern "C" + +// reference-based equivalents for c++ +inline void sg_setup(const sg_desc& desc) { return sg_setup(&desc); } + +inline sg_buffer sg_make_buffer(const sg_buffer_desc& desc) { return sg_make_buffer(&desc); } +inline sg_image sg_make_image(const sg_image_desc& desc) { return sg_make_image(&desc); } +inline sg_sampler sg_make_sampler(const sg_sampler_desc& desc) { return sg_make_sampler(&desc); } +inline sg_shader sg_make_shader(const sg_shader_desc& desc) { return sg_make_shader(&desc); } +inline sg_pipeline sg_make_pipeline(const sg_pipeline_desc& desc) { return sg_make_pipeline(&desc); } +inline sg_attachments sg_make_attachments(const sg_attachments_desc& desc) { return sg_make_attachments(&desc); } +inline void sg_update_image(sg_image img, const sg_image_data& data) { return sg_update_image(img, &data); } + +inline void sg_begin_pass(const sg_pass& pass) { return sg_begin_pass(&pass); } +inline void sg_apply_bindings(const sg_bindings& bindings) { return sg_apply_bindings(&bindings); } +inline void sg_apply_uniforms(int ub_slot, const sg_range& data) { return sg_apply_uniforms(ub_slot, &data); } + +inline sg_buffer_desc sg_query_buffer_defaults(const sg_buffer_desc& desc) { return sg_query_buffer_defaults(&desc); } +inline sg_image_desc sg_query_image_defaults(const sg_image_desc& desc) { return sg_query_image_defaults(&desc); } +inline sg_sampler_desc sg_query_sampler_defaults(const sg_sampler_desc& desc) { return sg_query_sampler_defaults(&desc); } +inline sg_shader_desc sg_query_shader_defaults(const sg_shader_desc& desc) { return sg_query_shader_defaults(&desc); } +inline sg_pipeline_desc sg_query_pipeline_defaults(const sg_pipeline_desc& desc) { return sg_query_pipeline_defaults(&desc); } +inline sg_attachments_desc sg_query_attachments_defaults(const sg_attachments_desc& desc) { return sg_query_attachments_defaults(&desc); } + +inline void sg_init_buffer(sg_buffer buf, const sg_buffer_desc& desc) { return sg_init_buffer(buf, &desc); } +inline void sg_init_image(sg_image img, const sg_image_desc& desc) { return sg_init_image(img, &desc); } +inline void sg_init_sampler(sg_sampler smp, const sg_sampler_desc& desc) { return sg_init_sampler(smp, &desc); } +inline void sg_init_shader(sg_shader shd, const sg_shader_desc& desc) { return sg_init_shader(shd, &desc); } +inline void sg_init_pipeline(sg_pipeline pip, const sg_pipeline_desc& desc) { return sg_init_pipeline(pip, &desc); } +inline void sg_init_attachments(sg_attachments atts, const sg_attachments_desc& desc) { return sg_init_attachments(atts, &desc); } + +inline void sg_update_buffer(sg_buffer buf_id, const sg_range& data) { return sg_update_buffer(buf_id, &data); } +inline int sg_append_buffer(sg_buffer buf_id, const sg_range& data) { return sg_append_buffer(buf_id, &data); } +#endif +#endif // SOKOL_GFX_INCLUDED + +// ██ ███ ███ ██████ ██ ███████ ███ ███ ███████ ███ ██ ████████ █████ ████████ ██ ██████ ███ ██ +// ██ ████ ████ ██ ██ ██ ██ ████ ████ ██ ████ ██ ██ ██ ██ ██ ██ ██ ██ ████ ██ +// ██ ██ ████ ██ ██████ ██ █████ ██ ████ ██ █████ ██ ██ ██ ██ ███████ ██ ██ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ███████ ███████ ██ ██ ███████ ██ ████ ██ ██ ██ ██ ██ ██████ ██ ████ +// +// >>implementation +#ifdef SOKOL_GFX_IMPL +#define SOKOL_GFX_IMPL_INCLUDED (1) + +#if !(defined(SOKOL_GLCORE)||defined(SOKOL_GLES3)||defined(SOKOL_D3D11)||defined(SOKOL_METAL)||defined(SOKOL_WGPU)||defined(SOKOL_DUMMY_BACKEND)) +#error "Please select a backend with SOKOL_GLCORE, SOKOL_GLES3, SOKOL_D3D11, SOKOL_METAL, SOKOL_WGPU or SOKOL_DUMMY_BACKEND" +#endif +#if defined(SOKOL_MALLOC) || defined(SOKOL_CALLOC) || defined(SOKOL_FREE) +#error "SOKOL_MALLOC/CALLOC/FREE macros are no longer supported, please use sg_desc.allocator to override memory allocation functions" +#endif + +#include // malloc, free, qsort +#include // memset +#include // FLT_MAX + +#ifndef SOKOL_API_IMPL + #define SOKOL_API_IMPL +#endif +#ifndef SOKOL_DEBUG + #ifndef NDEBUG + #define SOKOL_DEBUG + #endif +#endif +#ifndef SOKOL_ASSERT + #include + #define SOKOL_ASSERT(c) assert(c) +#endif +#ifndef SOKOL_UNREACHABLE + #define SOKOL_UNREACHABLE SOKOL_ASSERT(false) +#endif + +#ifndef _SOKOL_PRIVATE + #if defined(__GNUC__) || defined(__clang__) + #define _SOKOL_PRIVATE __attribute__((unused)) static + #else + #define _SOKOL_PRIVATE static + #endif +#endif + +#ifndef _SOKOL_UNUSED + #define _SOKOL_UNUSED(x) (void)(x) +#endif + +#if defined(SOKOL_TRACE_HOOKS) +#define _SG_TRACE_ARGS(fn, ...) if (_sg.hooks.fn) { _sg.hooks.fn(__VA_ARGS__, _sg.hooks.user_data); } +#define _SG_TRACE_NOARGS(fn) if (_sg.hooks.fn) { _sg.hooks.fn(_sg.hooks.user_data); } +#else +#define _SG_TRACE_ARGS(fn, ...) +#define _SG_TRACE_NOARGS(fn) +#endif + +// default clear values +#ifndef SG_DEFAULT_CLEAR_RED +#define SG_DEFAULT_CLEAR_RED (0.5f) +#endif +#ifndef SG_DEFAULT_CLEAR_GREEN +#define SG_DEFAULT_CLEAR_GREEN (0.5f) +#endif +#ifndef SG_DEFAULT_CLEAR_BLUE +#define SG_DEFAULT_CLEAR_BLUE (0.5f) +#endif +#ifndef SG_DEFAULT_CLEAR_ALPHA +#define SG_DEFAULT_CLEAR_ALPHA (1.0f) +#endif +#ifndef SG_DEFAULT_CLEAR_DEPTH +#define SG_DEFAULT_CLEAR_DEPTH (1.0f) +#endif +#ifndef SG_DEFAULT_CLEAR_STENCIL +#define SG_DEFAULT_CLEAR_STENCIL (0) +#endif + +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable:4115) // named type definition in parentheses +#pragma warning(disable:4505) // unreferenced local function has been removed +#pragma warning(disable:4201) // nonstandard extension used: nameless struct/union (needed by d3d11.h) +#pragma warning(disable:4054) // 'type cast': from function pointer +#pragma warning(disable:4055) // 'type cast': from data pointer +#endif + +#if defined(SOKOL_D3D11) + #ifndef D3D11_NO_HELPERS + #define D3D11_NO_HELPERS + #endif + #ifndef WIN32_LEAN_AND_MEAN + #define WIN32_LEAN_AND_MEAN + #endif + #ifndef NOMINMAX + #define NOMINMAX + #endif + #include + #include + #ifdef _MSC_VER + #pragma comment (lib, "kernel32") + #pragma comment (lib, "user32") + #pragma comment (lib, "dxgi") + #pragma comment (lib, "d3d11") + #endif +#elif defined(SOKOL_METAL) + // see https://clang.llvm.org/docs/LanguageExtensions.html#automatic-reference-counting + #if !defined(__cplusplus) + #if __has_feature(objc_arc) && !__has_feature(objc_arc_fields) + #error "sokol_gfx.h requires __has_feature(objc_arc_field) if ARC is enabled (use a more recent compiler version)" + #endif + #endif + #include + #include + #if defined(TARGET_OS_IPHONE) && !TARGET_OS_IPHONE + #define _SG_TARGET_MACOS (1) + #else + #define _SG_TARGET_IOS (1) + #if defined(TARGET_IPHONE_SIMULATOR) && TARGET_IPHONE_SIMULATOR + #define _SG_TARGET_IOS_SIMULATOR (1) + #endif + #endif + #import + #import // needed for CAMetalDrawable +#elif defined(SOKOL_WGPU) + #include + #if defined(__EMSCRIPTEN__) + #include + #endif +#elif defined(SOKOL_GLCORE) || defined(SOKOL_GLES3) + #define _SOKOL_ANY_GL (1) + + // include platform specific GL headers (or on Win32: use an embedded GL loader) + #if !defined(SOKOL_EXTERNAL_GL_LOADER) + #if defined(_WIN32) + #if defined(SOKOL_GLCORE) && !defined(SOKOL_EXTERNAL_GL_LOADER) + #ifndef WIN32_LEAN_AND_MEAN + #define WIN32_LEAN_AND_MEAN + #endif + #ifndef NOMINMAX + #define NOMINMAX + #endif + #include + #define _SOKOL_USE_WIN32_GL_LOADER (1) + #pragma comment (lib, "kernel32") // GetProcAddress() + #define _SOKOL_GL_HAS_COMPUTE (1) + #endif + #elif defined(__APPLE__) + #include + #ifndef GL_SILENCE_DEPRECATION + #define GL_SILENCE_DEPRECATION + #endif + #if defined(TARGET_OS_IPHONE) && !TARGET_OS_IPHONE + #include + #else + #include + #include + #endif + #elif defined(__EMSCRIPTEN__) + #if defined(SOKOL_GLES3) + #include + #endif + #elif defined(__ANDROID__) + #define _SOKOL_GL_HAS_COMPUTE (1) + #include + #elif defined(__linux__) || defined(__unix__) + #define _SOKOL_GL_HAS_COMPUTE (1) + #if defined(SOKOL_GLCORE) + #define GL_GLEXT_PROTOTYPES + #include + #else + #include + #include + #endif + #endif + #endif + + // optional GL loader definitions (only on Win32) + #if defined(_SOKOL_USE_WIN32_GL_LOADER) + #define __gl_h_ 1 + #define __gl32_h_ 1 + #define __gl31_h_ 1 + #define __GL_H__ 1 + #define __glext_h_ 1 + #define __GLEXT_H_ 1 + #define __gltypes_h_ 1 + #define __glcorearb_h_ 1 + #define __gl_glcorearb_h_ 1 + #define GL_APIENTRY APIENTRY + + typedef unsigned int GLenum; + typedef unsigned int GLuint; + typedef int GLsizei; + typedef char GLchar; + typedef ptrdiff_t GLintptr; + typedef ptrdiff_t GLsizeiptr; + typedef double GLclampd; + typedef unsigned short GLushort; + typedef unsigned char GLubyte; + typedef unsigned char GLboolean; + typedef uint64_t GLuint64; + typedef double GLdouble; + typedef unsigned short GLhalf; + typedef float GLclampf; + typedef unsigned int GLbitfield; + typedef signed char GLbyte; + typedef short GLshort; + typedef void GLvoid; + typedef int64_t GLint64; + typedef float GLfloat; + typedef int GLint; + #define GL_INT_2_10_10_10_REV 0x8D9F + #define GL_R32F 0x822E + #define GL_PROGRAM_POINT_SIZE 0x8642 + #define GL_DEPTH_ATTACHMENT 0x8D00 + #define GL_DEPTH_STENCIL_ATTACHMENT 0x821A + #define GL_COLOR_ATTACHMENT2 0x8CE2 + #define GL_COLOR_ATTACHMENT0 0x8CE0 + #define GL_R16F 0x822D + #define GL_COLOR_ATTACHMENT22 0x8CF6 + #define GL_DRAW_FRAMEBUFFER 0x8CA9 + #define GL_FRAMEBUFFER_COMPLETE 0x8CD5 + #define GL_NUM_EXTENSIONS 0x821D + #define GL_INFO_LOG_LENGTH 0x8B84 + #define GL_VERTEX_SHADER 0x8B31 + #define GL_INCR 0x1E02 + #define GL_DYNAMIC_DRAW 0x88E8 + #define GL_STATIC_DRAW 0x88E4 + #define GL_TEXTURE_CUBE_MAP_POSITIVE_Z 0x8519 + #define GL_TEXTURE_CUBE_MAP 0x8513 + #define GL_FUNC_SUBTRACT 0x800A + #define GL_FUNC_REVERSE_SUBTRACT 0x800B + #define GL_CONSTANT_COLOR 0x8001 + #define GL_DECR_WRAP 0x8508 + #define GL_R8 0x8229 + #define GL_LINEAR_MIPMAP_LINEAR 0x2703 + #define GL_ELEMENT_ARRAY_BUFFER 0x8893 + #define GL_SHORT 0x1402 + #define GL_DEPTH_TEST 0x0B71 + #define GL_TEXTURE_CUBE_MAP_NEGATIVE_Y 0x8518 + #define GL_LINK_STATUS 0x8B82 + #define GL_TEXTURE_CUBE_MAP_POSITIVE_Y 0x8517 + #define GL_SAMPLE_ALPHA_TO_COVERAGE 0x809E + #define GL_RGBA16F 0x881A + #define GL_CONSTANT_ALPHA 0x8003 + #define GL_READ_FRAMEBUFFER 0x8CA8 + #define GL_TEXTURE0 0x84C0 + #define GL_TEXTURE_MIN_LOD 0x813A + #define GL_CLAMP_TO_EDGE 0x812F + #define GL_UNSIGNED_SHORT_5_6_5 0x8363 + #define GL_TEXTURE_WRAP_R 0x8072 + #define GL_UNSIGNED_SHORT_5_5_5_1 0x8034 + #define GL_NEAREST_MIPMAP_NEAREST 0x2700 + #define GL_UNSIGNED_SHORT_4_4_4_4 0x8033 + #define GL_SRC_ALPHA_SATURATE 0x0308 + #define GL_STREAM_DRAW 0x88E0 + #define GL_ONE 1 + #define GL_NEAREST_MIPMAP_LINEAR 0x2702 + #define GL_RGB10_A2 0x8059 + #define GL_RGBA8 0x8058 + #define GL_SRGB8_ALPHA8 0x8C43 + #define GL_COLOR_ATTACHMENT1 0x8CE1 + #define GL_RGBA4 0x8056 + #define GL_RGB8 0x8051 + #define GL_ARRAY_BUFFER 0x8892 + #define GL_STENCIL 0x1802 + #define GL_TEXTURE_2D 0x0DE1 + #define GL_DEPTH 0x1801 + #define GL_FRONT 0x0404 + #define GL_STENCIL_BUFFER_BIT 0x00000400 + #define GL_REPEAT 0x2901 + #define GL_RGBA 0x1908 + #define GL_TEXTURE_CUBE_MAP_POSITIVE_X 0x8515 + #define GL_DECR 0x1E03 + #define GL_FRAGMENT_SHADER 0x8B30 + #define GL_COMPUTE_SHADER 0x91B9 + #define GL_FLOAT 0x1406 + #define GL_TEXTURE_MAX_LOD 0x813B + #define GL_DEPTH_COMPONENT 0x1902 + #define GL_ONE_MINUS_DST_ALPHA 0x0305 + #define GL_COLOR 0x1800 + #define GL_TEXTURE_2D_ARRAY 0x8C1A + #define GL_TRIANGLES 0x0004 + #define GL_UNSIGNED_BYTE 0x1401 + #define GL_TEXTURE_MAG_FILTER 0x2800 + #define GL_ONE_MINUS_CONSTANT_ALPHA 0x8004 + #define GL_NONE 0 + #define GL_SRC_COLOR 0x0300 + #define GL_BYTE 0x1400 + #define GL_TEXTURE_CUBE_MAP_NEGATIVE_Z 0x851A + #define GL_LINE_STRIP 0x0003 + #define GL_TEXTURE_3D 0x806F + #define GL_CW 0x0900 + #define GL_LINEAR 0x2601 + #define GL_RENDERBUFFER 0x8D41 + #define GL_GEQUAL 0x0206 + #define GL_COLOR_BUFFER_BIT 0x00004000 + #define GL_RGBA32F 0x8814 + #define GL_BLEND 0x0BE2 + #define GL_ONE_MINUS_SRC_ALPHA 0x0303 + #define GL_ONE_MINUS_CONSTANT_COLOR 0x8002 + #define GL_TEXTURE_WRAP_T 0x2803 + #define GL_TEXTURE_WRAP_S 0x2802 + #define GL_TEXTURE_MIN_FILTER 0x2801 + #define GL_LINEAR_MIPMAP_NEAREST 0x2701 + #define GL_EXTENSIONS 0x1F03 + #define GL_NO_ERROR 0 + #define GL_REPLACE 0x1E01 + #define GL_KEEP 0x1E00 + #define GL_CCW 0x0901 + #define GL_TEXTURE_CUBE_MAP_NEGATIVE_X 0x8516 + #define GL_RGB 0x1907 + #define GL_TRIANGLE_STRIP 0x0005 + #define GL_FALSE 0 + #define GL_ZERO 0 + #define GL_CULL_FACE 0x0B44 + #define GL_INVERT 0x150A + #define GL_INT 0x1404 + #define GL_UNSIGNED_INT 0x1405 + #define GL_UNSIGNED_SHORT 0x1403 + #define GL_NEAREST 0x2600 + #define GL_SCISSOR_TEST 0x0C11 + #define GL_LEQUAL 0x0203 + #define GL_STENCIL_TEST 0x0B90 + #define GL_DITHER 0x0BD0 + #define GL_DEPTH_COMPONENT32F 0x8CAC + #define GL_EQUAL 0x0202 + #define GL_FRAMEBUFFER 0x8D40 + #define GL_RGB5 0x8050 + #define GL_LINES 0x0001 + #define GL_DEPTH_BUFFER_BIT 0x00000100 + #define GL_SRC_ALPHA 0x0302 + #define GL_INCR_WRAP 0x8507 + #define GL_LESS 0x0201 + #define GL_MULTISAMPLE 0x809D + #define GL_FRAMEBUFFER_BINDING 0x8CA6 + #define GL_BACK 0x0405 + #define GL_ALWAYS 0x0207 + #define GL_FUNC_ADD 0x8006 + #define GL_ONE_MINUS_DST_COLOR 0x0307 + #define GL_NOTEQUAL 0x0205 + #define GL_DST_COLOR 0x0306 + #define GL_COMPILE_STATUS 0x8B81 + #define GL_RED 0x1903 + #define GL_COLOR_ATTACHMENT3 0x8CE3 + #define GL_DST_ALPHA 0x0304 + #define GL_RGB5_A1 0x8057 + #define GL_GREATER 0x0204 + #define GL_POLYGON_OFFSET_FILL 0x8037 + #define GL_TRUE 1 + #define GL_NEVER 0x0200 + #define GL_POINTS 0x0000 + #define GL_ONE_MINUS_SRC_COLOR 0x0301 + #define GL_MIRRORED_REPEAT 0x8370 + #define GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS 0x8B4D + #define GL_R11F_G11F_B10F 0x8C3A + #define GL_UNSIGNED_INT_10F_11F_11F_REV 0x8C3B + #define GL_RGB9_E5 0x8C3D + #define GL_UNSIGNED_INT_5_9_9_9_REV 0x8C3E + #define GL_RGBA32UI 0x8D70 + #define GL_RGB32UI 0x8D71 + #define GL_RGBA16UI 0x8D76 + #define GL_RGB16UI 0x8D77 + #define GL_RGBA8UI 0x8D7C + #define GL_RGB8UI 0x8D7D + #define GL_RGBA32I 0x8D82 + #define GL_RGB32I 0x8D83 + #define GL_RGBA16I 0x8D88 + #define GL_RGB16I 0x8D89 + #define GL_RGBA8I 0x8D8E + #define GL_RGB8I 0x8D8F + #define GL_RED_INTEGER 0x8D94 + #define GL_RG 0x8227 + #define GL_RG_INTEGER 0x8228 + #define GL_R8 0x8229 + #define GL_R16 0x822A + #define GL_RG8 0x822B + #define GL_RG16 0x822C + #define GL_R16F 0x822D + #define GL_R32F 0x822E + #define GL_RG16F 0x822F + #define GL_RG32F 0x8230 + #define GL_R8I 0x8231 + #define GL_R8UI 0x8232 + #define GL_R16I 0x8233 + #define GL_R16UI 0x8234 + #define GL_R32I 0x8235 + #define GL_R32UI 0x8236 + #define GL_RG8I 0x8237 + #define GL_RG8UI 0x8238 + #define GL_RG16I 0x8239 + #define GL_RG16UI 0x823A + #define GL_RG32I 0x823B + #define GL_RG32UI 0x823C + #define GL_RGBA_INTEGER 0x8D99 + #define GL_R8_SNORM 0x8F94 + #define GL_RG8_SNORM 0x8F95 + #define GL_RGB8_SNORM 0x8F96 + #define GL_RGBA8_SNORM 0x8F97 + #define GL_R16_SNORM 0x8F98 + #define GL_RG16_SNORM 0x8F99 + #define GL_RGB16_SNORM 0x8F9A + #define GL_RGBA16_SNORM 0x8F9B + #define GL_RGBA16 0x805B + #define GL_MAX_TEXTURE_SIZE 0x0D33 + #define GL_MAX_CUBE_MAP_TEXTURE_SIZE 0x851C + #define GL_MAX_3D_TEXTURE_SIZE 0x8073 + #define GL_MAX_ARRAY_TEXTURE_LAYERS 0x88FF + #define GL_MAX_VERTEX_ATTRIBS 0x8869 + #define GL_CLAMP_TO_BORDER 0x812D + #define GL_TEXTURE_BORDER_COLOR 0x1004 + #define GL_CURRENT_PROGRAM 0x8B8D + #define GL_MAX_VERTEX_UNIFORM_COMPONENTS 0x8B4A + #define GL_UNPACK_ALIGNMENT 0x0CF5 + #define GL_FRAMEBUFFER_SRGB 0x8DB9 + #define GL_TEXTURE_COMPARE_MODE 0x884C + #define GL_TEXTURE_COMPARE_FUNC 0x884D + #define GL_COMPARE_REF_TO_TEXTURE 0x884E + #define GL_TEXTURE_CUBE_MAP_SEAMLESS 0x884F + #define GL_TEXTURE_MAX_LEVEL 0x813D + #define GL_FRAMEBUFFER_UNDEFINED 0x8219 + #define GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT 0x8CD6 + #define GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT 0x8CD7 + #define GL_FRAMEBUFFER_UNSUPPORTED 0x8CDD + #define GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE 0x8D56 + #define GL_MAJOR_VERSION 0x821B + #define GL_MINOR_VERSION 0x821C + #define GL_TEXTURE_2D_MULTISAMPLE 0x9100 + #define GL_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9102 + #define GL_SHADER_STORAGE_BARRIER_BIT 0x2000 + #define GL_MIN 0x8007 + #define GL_MAX 0x8008 + #endif + + #ifndef GL_UNSIGNED_INT_2_10_10_10_REV + #define GL_UNSIGNED_INT_2_10_10_10_REV 0x8368 + #endif + #ifndef GL_UNSIGNED_INT_24_8 + #define GL_UNSIGNED_INT_24_8 0x84FA + #endif + #ifndef GL_TEXTURE_MAX_ANISOTROPY_EXT + #define GL_TEXTURE_MAX_ANISOTROPY_EXT 0x84FE + #endif + #ifndef GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT + #define GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT 0x84FF + #endif + #ifndef GL_COMPRESSED_RGBA_S3TC_DXT1_EXT + #define GL_COMPRESSED_RGBA_S3TC_DXT1_EXT 0x83F1 + #endif + #ifndef GL_COMPRESSED_RGBA_S3TC_DXT3_EXT + #define GL_COMPRESSED_RGBA_S3TC_DXT3_EXT 0x83F2 + #endif + #ifndef GL_COMPRESSED_RGBA_S3TC_DXT5_EXT + #define GL_COMPRESSED_RGBA_S3TC_DXT5_EXT 0x83F3 + #endif + #ifndef GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT + #define GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT 0x8C4F + #endif + #ifndef GL_COMPRESSED_RED_RGTC1 + #define GL_COMPRESSED_RED_RGTC1 0x8DBB + #endif + #ifndef GL_COMPRESSED_SIGNED_RED_RGTC1 + #define GL_COMPRESSED_SIGNED_RED_RGTC1 0x8DBC + #endif + #ifndef GL_COMPRESSED_RED_GREEN_RGTC2 + #define GL_COMPRESSED_RED_GREEN_RGTC2 0x8DBD + #endif + #ifndef GL_COMPRESSED_SIGNED_RED_GREEN_RGTC2 + #define GL_COMPRESSED_SIGNED_RED_GREEN_RGTC2 0x8DBE + #endif + #ifndef GL_COMPRESSED_RGBA_BPTC_UNORM_ARB + #define GL_COMPRESSED_RGBA_BPTC_UNORM_ARB 0x8E8C + #endif + #ifndef GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM_ARB + #define GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM_ARB 0x8E8D + #endif + #ifndef GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT_ARB + #define GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT_ARB 0x8E8E + #endif + #ifndef GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT_ARB + #define GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT_ARB 0x8E8F + #endif + #ifndef GL_COMPRESSED_RGB8_ETC2 + #define GL_COMPRESSED_RGB8_ETC2 0x9274 + #endif + #ifndef GL_COMPRESSED_SRGB8_ETC2 + #define GL_COMPRESSED_SRGB8_ETC2 0x9275 + #endif + #ifndef GL_COMPRESSED_RGBA8_ETC2_EAC + #define GL_COMPRESSED_RGBA8_ETC2_EAC 0x9278 + #endif + #ifndef GL_COMPRESSED_SRGB8_ALPHA8_ETC2_EAC + #define GL_COMPRESSED_SRGB8_ALPHA8_ETC2_EAC 0x9279 + #endif + #ifndef GL_COMPRESSED_RGB8_PUNCHTHROUGH_ALPHA1_ETC2 + #define GL_COMPRESSED_RGB8_PUNCHTHROUGH_ALPHA1_ETC2 0x9276 + #endif + #ifndef GL_COMPRESSED_R11_EAC + #define GL_COMPRESSED_R11_EAC 0x9270 + #endif + #ifndef GL_COMPRESSED_SIGNED_R11_EAC + #define GL_COMPRESSED_SIGNED_R11_EAC 0x9271 + #endif + #ifndef GL_COMPRESSED_RG11_EAC + #define GL_COMPRESSED_RG11_EAC 0x9272 + #endif + #ifndef GL_COMPRESSED_SIGNED_RG11_EAC + #define GL_COMPRESSED_SIGNED_RG11_EAC 0x9273 + #endif + #ifndef GL_COMPRESSED_RGBA_ASTC_4x4_KHR + #define GL_COMPRESSED_RGBA_ASTC_4x4_KHR 0x93B0 + #endif + #ifndef GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR + #define GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR 0x93D0 + #endif + #ifndef GL_DEPTH24_STENCIL8 + #define GL_DEPTH24_STENCIL8 0x88F0 + #endif + #ifndef GL_HALF_FLOAT + #define GL_HALF_FLOAT 0x140B + #endif + #ifndef GL_DEPTH_STENCIL + #define GL_DEPTH_STENCIL 0x84F9 + #endif + #ifndef GL_LUMINANCE + #define GL_LUMINANCE 0x1909 + #endif + #ifndef GL_COMPUTE_SHADER + #define GL_COMPUTE_SHADER 0x91B9 + #endif + #ifndef _SG_GL_CHECK_ERROR + #define _SG_GL_CHECK_ERROR() { SOKOL_ASSERT(glGetError() == GL_NO_ERROR); } + #endif +#endif + +#if defined(SOKOL_GLES3) + // on WebGL2, GL_FRAMEBUFFER_UNDEFINED technically doesn't exist (it is defined + // in the Emscripten headers, but may not exist in other WebGL2 shims) + // see: https://github.com/floooh/sokol/pull/933 + #ifndef GL_FRAMEBUFFER_UNDEFINED + #define GL_FRAMEBUFFER_UNDEFINED 0x8219 + #endif +#endif + +// make some GL constants generally available to simplify compilation, +// use of those constants will be filtered by runtime flags +#ifndef GL_SHADER_STORAGE_BUFFER +#define GL_SHADER_STORAGE_BUFFER 0x90D2 +#endif + +// ███████ ████████ ██████ ██ ██ ██████ ████████ ███████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ███████ ██ ██████ ██ ██ ██ ██ ███████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ███████ ██ ██ ██ ██████ ██████ ██ ███████ +// +// >>structs +// resource pool slots +typedef struct { + uint32_t id; + sg_resource_state state; +} _sg_slot_t; + +// resource pool housekeeping struct +typedef struct { + int size; + int queue_top; + uint32_t* gen_ctrs; + int* free_queue; +} _sg_pool_t; + +_SOKOL_PRIVATE void _sg_pool_init(_sg_pool_t* pool, int num); +_SOKOL_PRIVATE void _sg_pool_discard(_sg_pool_t* pool); +_SOKOL_PRIVATE int _sg_pool_alloc_index(_sg_pool_t* pool); +_SOKOL_PRIVATE void _sg_pool_free_index(_sg_pool_t* pool, int slot_index); +_SOKOL_PRIVATE void _sg_slot_reset(_sg_slot_t* slot); +_SOKOL_PRIVATE uint32_t _sg_slot_alloc(_sg_pool_t* pool, _sg_slot_t* slot, int slot_index); +_SOKOL_PRIVATE int _sg_slot_index(uint32_t id); + +// resource func forward decls +struct _sg_pools_s; +struct _sg_buffer_s; +_SOKOL_PRIVATE struct _sg_buffer_s* _sg_lookup_buffer(const struct _sg_pools_s* p, uint32_t buf_id); + +// resource tracking (for keeping track of gpu-written storage resources +typedef struct { + uint32_t size; + uint32_t cur; + uint32_t* items; +} _sg_tracker_t; + +_SOKOL_PRIVATE void _sg_tracker_init(_sg_tracker_t* tracker, uint32_t num); +_SOKOL_PRIVATE void _sg_tracker_discard(_sg_tracker_t* tracker); +_SOKOL_PRIVATE void _sg_tracker_reset(_sg_tracker_t* tracker); +_SOKOL_PRIVATE bool _sg_tracker_add(_sg_tracker_t* tracker, uint32_t res_id); + +// constants +enum { + _SG_STRING_SIZE = 32, + _SG_SLOT_SHIFT = 16, + _SG_SLOT_MASK = (1<<_SG_SLOT_SHIFT)-1, + _SG_MAX_POOL_SIZE = (1<<_SG_SLOT_SHIFT), + _SG_DEFAULT_BUFFER_POOL_SIZE = 128, + _SG_DEFAULT_IMAGE_POOL_SIZE = 128, + _SG_DEFAULT_SAMPLER_POOL_SIZE = 64, + _SG_DEFAULT_SHADER_POOL_SIZE = 32, + _SG_DEFAULT_PIPELINE_POOL_SIZE = 64, + _SG_DEFAULT_ATTACHMENTS_POOL_SIZE = 16, + _SG_DEFAULT_UB_SIZE = 4 * 1024 * 1024, + _SG_DEFAULT_MAX_DISPATCH_CALLS_PER_PASS = 1024, + _SG_DEFAULT_MAX_COMMIT_LISTENERS = 1024, + _SG_DEFAULT_WGPU_BINDGROUP_CACHE_SIZE = 1024, +}; + +// fixed-size string +typedef struct { + char buf[_SG_STRING_SIZE]; +} _sg_str_t; + +// helper macros +#define _sg_def(val, def) (((val) == 0) ? (def) : (val)) +#define _sg_def_flt(val, def) (((val) == 0.0f) ? (def) : (val)) +#define _sg_min(a,b) (((a)<(b))?(a):(b)) +#define _sg_max(a,b) (((a)>(b))?(a):(b)) +#define _sg_clamp(v,v0,v1) (((v)<(v0))?(v0):(((v)>(v1))?(v1):(v))) +#define _sg_fequal(val,cmp,delta) ((((val)-(cmp))> -(delta))&&(((val)-(cmp))<(delta))) +#define _sg_ispow2(val) ((val&(val-1))==0) +#define _sg_stats_add(key,val) {if(_sg.stats_enabled){ _sg.stats.key+=val;}} + +_SOKOL_PRIVATE void* _sg_malloc_clear(size_t size); +_SOKOL_PRIVATE void _sg_free(void* ptr); +_SOKOL_PRIVATE void _sg_clear(void* ptr, size_t size); + +typedef struct { + int size; + int append_pos; + bool append_overflow; + uint32_t update_frame_index; + uint32_t append_frame_index; + int num_slots; + int active_slot; + sg_buffer_type type; + sg_usage usage; +} _sg_buffer_common_t; + +_SOKOL_PRIVATE void _sg_buffer_common_init(_sg_buffer_common_t* cmn, const sg_buffer_desc* desc) { + cmn->size = (int)desc->size; + cmn->append_pos = 0; + cmn->append_overflow = false; + cmn->update_frame_index = 0; + cmn->append_frame_index = 0; + cmn->num_slots = (desc->usage == SG_USAGE_IMMUTABLE) ? 1 : SG_NUM_INFLIGHT_FRAMES; + cmn->active_slot = 0; + cmn->type = desc->type; + cmn->usage = desc->usage; +} + +typedef struct { + uint32_t upd_frame_index; + int num_slots; + int active_slot; + sg_image_type type; + bool render_target; + int width; + int height; + int num_slices; + int num_mipmaps; + sg_usage usage; + sg_pixel_format pixel_format; + int sample_count; +} _sg_image_common_t; + +_SOKOL_PRIVATE void _sg_image_common_init(_sg_image_common_t* cmn, const sg_image_desc* desc) { + cmn->upd_frame_index = 0; + cmn->num_slots = (desc->usage == SG_USAGE_IMMUTABLE) ? 1 : SG_NUM_INFLIGHT_FRAMES; + cmn->active_slot = 0; + cmn->type = desc->type; + cmn->render_target = desc->render_target; + cmn->width = desc->width; + cmn->height = desc->height; + cmn->num_slices = desc->num_slices; + cmn->num_mipmaps = desc->num_mipmaps; + cmn->usage = desc->usage; + cmn->pixel_format = desc->pixel_format; + cmn->sample_count = desc->sample_count; +} + +typedef struct { + sg_filter min_filter; + sg_filter mag_filter; + sg_filter mipmap_filter; + sg_wrap wrap_u; + sg_wrap wrap_v; + sg_wrap wrap_w; + float min_lod; + float max_lod; + sg_border_color border_color; + sg_compare_func compare; + uint32_t max_anisotropy; +} _sg_sampler_common_t; + +_SOKOL_PRIVATE void _sg_sampler_common_init(_sg_sampler_common_t* cmn, const sg_sampler_desc* desc) { + cmn->min_filter = desc->min_filter; + cmn->mag_filter = desc->mag_filter; + cmn->mipmap_filter = desc->mipmap_filter; + cmn->wrap_u = desc->wrap_u; + cmn->wrap_v = desc->wrap_v; + cmn->wrap_w = desc->wrap_w; + cmn->min_lod = desc->min_lod; + cmn->max_lod = desc->max_lod; + cmn->border_color = desc->border_color; + cmn->compare = desc->compare; + cmn->max_anisotropy = desc->max_anisotropy; +} + +typedef struct { + sg_shader_attr_base_type base_type; +} _sg_shader_attr_t; + +typedef struct { + sg_shader_stage stage; + uint32_t size; +} _sg_shader_uniform_block_t; + +typedef struct { + sg_shader_stage stage; + bool readonly; +} _sg_shader_storage_buffer_t; + +typedef struct { + sg_shader_stage stage; + sg_image_type image_type; + sg_image_sample_type sample_type; + bool multisampled; +} _sg_shader_image_t; + +typedef struct { + sg_shader_stage stage; + sg_sampler_type sampler_type; +} _sg_shader_sampler_t; + +// combined image sampler mappings, only needed on GL +typedef struct { + sg_shader_stage stage; + uint8_t image_slot; + uint8_t sampler_slot; +} _sg_shader_image_sampler_t; + +typedef struct { + uint32_t required_bindings_and_uniforms; + bool is_compute; + _sg_shader_attr_t attrs[SG_MAX_VERTEX_ATTRIBUTES]; + _sg_shader_uniform_block_t uniform_blocks[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + _sg_shader_storage_buffer_t storage_buffers[SG_MAX_STORAGEBUFFER_BINDSLOTS]; + _sg_shader_image_t images[SG_MAX_IMAGE_BINDSLOTS]; + _sg_shader_sampler_t samplers[SG_MAX_SAMPLER_BINDSLOTS]; + _sg_shader_image_sampler_t image_samplers[SG_MAX_IMAGE_SAMPLER_PAIRS]; +} _sg_shader_common_t; + +_SOKOL_PRIVATE void _sg_shader_common_init(_sg_shader_common_t* cmn, const sg_shader_desc* desc) { + cmn->is_compute = desc->compute_func.source || desc->compute_func.bytecode.ptr; + for (size_t i = 0; i < SG_MAX_VERTEX_ATTRIBUTES; i++) { + cmn->attrs[i].base_type = desc->attrs[i].base_type; + } + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + const sg_shader_uniform_block* src = &desc->uniform_blocks[i]; + _sg_shader_uniform_block_t* dst = &cmn->uniform_blocks[i]; + if (src->stage != SG_SHADERSTAGE_NONE) { + cmn->required_bindings_and_uniforms |= (1 << i); + dst->stage = src->stage; + dst->size = src->size; + } + } + const uint32_t required_bindings_flag = (1 << SG_MAX_UNIFORMBLOCK_BINDSLOTS); + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + const sg_shader_storage_buffer* src = &desc->storage_buffers[i]; + _sg_shader_storage_buffer_t* dst = &cmn->storage_buffers[i]; + if (src->stage != SG_SHADERSTAGE_NONE) { + cmn->required_bindings_and_uniforms |= required_bindings_flag; + dst->stage = src->stage; + dst->readonly = src->readonly; + } + } + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + const sg_shader_image* src = &desc->images[i]; + _sg_shader_image_t* dst = &cmn->images[i]; + if (src->stage != SG_SHADERSTAGE_NONE) { + cmn->required_bindings_and_uniforms |= required_bindings_flag; + dst->stage = src->stage; + dst->image_type = src->image_type; + dst->sample_type = src->sample_type; + dst->multisampled = src->multisampled; + } + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + const sg_shader_sampler* src = &desc->samplers[i]; + _sg_shader_sampler_t* dst = &cmn->samplers[i]; + if (src->stage != SG_SHADERSTAGE_NONE) { + cmn->required_bindings_and_uniforms |= required_bindings_flag; + dst->stage = src->stage; + dst->sampler_type = src->sampler_type; + } + } + for (size_t i = 0; i < SG_MAX_IMAGE_SAMPLER_PAIRS; i++) { + const sg_shader_image_sampler_pair* src = &desc->image_sampler_pairs[i]; + _sg_shader_image_sampler_t* dst = &cmn->image_samplers[i]; + if (src->stage != SG_SHADERSTAGE_NONE) { + dst->stage = src->stage; + SOKOL_ASSERT((src->image_slot >= 0) && (src->image_slot < SG_MAX_IMAGE_BINDSLOTS)); + SOKOL_ASSERT(desc->images[src->image_slot].stage == src->stage); + dst->image_slot = src->image_slot; + SOKOL_ASSERT((src->sampler_slot >= 0) && (src->sampler_slot < SG_MAX_SAMPLER_BINDSLOTS)); + SOKOL_ASSERT(desc->samplers[src->sampler_slot].stage == src->stage); + dst->sampler_slot = src->sampler_slot; + } + } +} + +typedef struct { + bool vertex_buffer_layout_active[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + bool use_instanced_draw; + bool is_compute; + uint32_t required_bindings_and_uniforms; + sg_shader shader_id; + sg_vertex_layout_state layout; + sg_depth_state depth; + sg_stencil_state stencil; + int color_count; + sg_color_target_state colors[SG_MAX_COLOR_ATTACHMENTS]; + sg_primitive_type primitive_type; + sg_index_type index_type; + sg_cull_mode cull_mode; + sg_face_winding face_winding; + int sample_count; + sg_color blend_color; + bool alpha_to_coverage_enabled; +} _sg_pipeline_common_t; + +_SOKOL_PRIVATE void _sg_pipeline_common_init(_sg_pipeline_common_t* cmn, const sg_pipeline_desc* desc) { + SOKOL_ASSERT((desc->color_count >= 0) && (desc->color_count <= SG_MAX_COLOR_ATTACHMENTS)); + + // FIXME: most of this isn't needed for compute pipelines + + const uint32_t required_bindings_flag = (1 << SG_MAX_UNIFORMBLOCK_BINDSLOTS); + for (int i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + const sg_vertex_attr_state* a_state = &desc->layout.attrs[i]; + if (a_state->format != SG_VERTEXFORMAT_INVALID) { + SOKOL_ASSERT(a_state->buffer_index < SG_MAX_VERTEXBUFFER_BINDSLOTS); + cmn->vertex_buffer_layout_active[a_state->buffer_index] = true; + cmn->required_bindings_and_uniforms |= required_bindings_flag; + } + } + cmn->is_compute = desc->compute; + cmn->use_instanced_draw = false; + cmn->shader_id = desc->shader; + cmn->layout = desc->layout; + cmn->depth = desc->depth; + cmn->stencil = desc->stencil; + cmn->color_count = desc->color_count; + for (int i = 0; i < desc->color_count; i++) { + cmn->colors[i] = desc->colors[i]; + } + cmn->primitive_type = desc->primitive_type; + cmn->index_type = desc->index_type; + if (cmn->index_type != SG_INDEXTYPE_NONE) { + cmn->required_bindings_and_uniforms |= required_bindings_flag; + } + cmn->cull_mode = desc->cull_mode; + cmn->face_winding = desc->face_winding; + cmn->sample_count = desc->sample_count; + cmn->blend_color = desc->blend_color; + cmn->alpha_to_coverage_enabled = desc->alpha_to_coverage_enabled; +} + +typedef struct { + sg_image image_id; + int mip_level; + int slice; +} _sg_attachment_common_t; + +typedef struct { + int width; + int height; + int num_colors; + _sg_attachment_common_t colors[SG_MAX_COLOR_ATTACHMENTS]; + _sg_attachment_common_t resolves[SG_MAX_COLOR_ATTACHMENTS]; + _sg_attachment_common_t depth_stencil; +} _sg_attachments_common_t; + +_SOKOL_PRIVATE void _sg_attachment_common_init(_sg_attachment_common_t* cmn, const sg_attachment_desc* desc) { + cmn->image_id = desc->image; + cmn->mip_level = desc->mip_level; + cmn->slice = desc->slice; +} + +_SOKOL_PRIVATE void _sg_attachments_common_init(_sg_attachments_common_t* cmn, const sg_attachments_desc* desc, int width, int height) { + SOKOL_ASSERT((width > 0) && (height > 0)); + cmn->width = width; + cmn->height = height; + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + if (desc->colors[i].image.id != SG_INVALID_ID) { + cmn->num_colors++; + _sg_attachment_common_init(&cmn->colors[i], &desc->colors[i]); + _sg_attachment_common_init(&cmn->resolves[i], &desc->resolves[i]); + } + } + if (desc->depth_stencil.image.id != SG_INVALID_ID) { + _sg_attachment_common_init(&cmn->depth_stencil, &desc->depth_stencil); + } +} + +#if defined(SOKOL_DUMMY_BACKEND) +typedef struct _sg_buffer_s { + _sg_slot_t slot; + _sg_buffer_common_t cmn; +} _sg_dummy_buffer_t; +typedef _sg_dummy_buffer_t _sg_buffer_t; + +typedef struct _sg_image_s { + _sg_slot_t slot; + _sg_image_common_t cmn; +} _sg_dummy_image_t; +typedef _sg_dummy_image_t _sg_image_t; + +typedef struct _sg_sampler_s { + _sg_slot_t slot; + _sg_sampler_common_t cmn; +} _sg_dummy_sampler_t; +typedef _sg_dummy_sampler_t _sg_sampler_t; + +typedef struct _sg_shader_s { + _sg_slot_t slot; + _sg_shader_common_t cmn; +} _sg_dummy_shader_t; +typedef _sg_dummy_shader_t _sg_shader_t; + +typedef struct _sg_pipeline_s { + _sg_slot_t slot; + _sg_shader_t* shader; + _sg_pipeline_common_t cmn; +} _sg_dummy_pipeline_t; +typedef _sg_dummy_pipeline_t _sg_pipeline_t; + +typedef struct { + _sg_image_t* image; +} _sg_dummy_attachment_t; + +typedef struct _sg_attachments_s { + _sg_slot_t slot; + _sg_attachments_common_t cmn; + struct { + _sg_dummy_attachment_t colors[SG_MAX_COLOR_ATTACHMENTS]; + _sg_dummy_attachment_t resolves[SG_MAX_COLOR_ATTACHMENTS]; + _sg_dummy_attachment_t depth_stencil; + } dmy; +} _sg_dummy_attachments_t; +typedef _sg_dummy_attachments_t _sg_attachments_t; + +#elif defined(_SOKOL_ANY_GL) + +typedef struct _sg_buffer_s { + _sg_slot_t slot; + _sg_buffer_common_t cmn; + struct { + GLuint buf[SG_NUM_INFLIGHT_FRAMES]; + bool injected; // if true, external buffers were injected with sg_buffer_desc.gl_buffers + bool gpu_dirty; // true if modified by GPU shader but memory barrier hasn't been issued yet + } gl; +} _sg_gl_buffer_t; +typedef _sg_gl_buffer_t _sg_buffer_t; + +typedef struct _sg_image_s { + _sg_slot_t slot; + _sg_image_common_t cmn; + struct { + GLenum target; + GLuint msaa_render_buffer; + GLuint tex[SG_NUM_INFLIGHT_FRAMES]; + bool injected; // if true, external textures were injected with sg_image_desc.gl_textures + } gl; +} _sg_gl_image_t; +typedef _sg_gl_image_t _sg_image_t; + +typedef struct _sg_sampler_s { + _sg_slot_t slot; + _sg_sampler_common_t cmn; + struct { + GLuint smp; + bool injected; // true if external sampler was injects in sg_sampler_desc.gl_sampler + } gl; +} _sg_gl_sampler_t; +typedef _sg_gl_sampler_t _sg_sampler_t; + +typedef struct { + GLint gl_loc; + sg_uniform_type type; + uint16_t count; + uint16_t offset; +} _sg_gl_uniform_t; + +typedef struct { + int num_uniforms; + _sg_gl_uniform_t uniforms[SG_MAX_UNIFORMBLOCK_MEMBERS]; +} _sg_gl_uniform_block_t; + +typedef struct { + _sg_str_t name; +} _sg_gl_shader_attr_t; + +typedef struct _sg_shader_s { + _sg_slot_t slot; + _sg_shader_common_t cmn; + struct { + GLuint prog; + _sg_gl_shader_attr_t attrs[SG_MAX_VERTEX_ATTRIBUTES]; + _sg_gl_uniform_block_t uniform_blocks[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + uint8_t sbuf_binding[SG_MAX_STORAGEBUFFER_BINDSLOTS]; + int8_t tex_slot[SG_MAX_IMAGE_SAMPLER_PAIRS]; // GL texture unit index + } gl; +} _sg_gl_shader_t; +typedef _sg_gl_shader_t _sg_shader_t; + +typedef struct { + int8_t vb_index; // -1 if attr is not enabled + int8_t divisor; // -1 if not initialized + uint8_t stride; + uint8_t size; + uint8_t normalized; + int offset; + GLenum type; + sg_shader_attr_base_type base_type; +} _sg_gl_attr_t; + +typedef struct _sg_pipeline_s { + _sg_slot_t slot; + _sg_pipeline_common_t cmn; + _sg_shader_t* shader; + struct { + _sg_gl_attr_t attrs[SG_MAX_VERTEX_ATTRIBUTES]; + sg_depth_state depth; + sg_stencil_state stencil; + sg_primitive_type primitive_type; + sg_blend_state blend; + sg_color_mask color_write_mask[SG_MAX_COLOR_ATTACHMENTS]; + sg_cull_mode cull_mode; + sg_face_winding face_winding; + int sample_count; + bool alpha_to_coverage_enabled; + } gl; +} _sg_gl_pipeline_t; +typedef _sg_gl_pipeline_t _sg_pipeline_t; + +typedef struct { + _sg_image_t* image; +} _sg_gl_attachment_t; + +typedef struct _sg_attachments_s { + _sg_slot_t slot; + _sg_attachments_common_t cmn; + struct { + GLuint fb; + _sg_gl_attachment_t colors[SG_MAX_COLOR_ATTACHMENTS]; + _sg_gl_attachment_t resolves[SG_MAX_COLOR_ATTACHMENTS]; + _sg_gl_attachment_t depth_stencil; + GLuint msaa_resolve_framebuffer[SG_MAX_COLOR_ATTACHMENTS]; + } gl; +} _sg_gl_attachments_t; +typedef _sg_gl_attachments_t _sg_attachments_t; + +typedef struct { + _sg_gl_attr_t gl_attr; + GLuint gl_vbuf; +} _sg_gl_cache_attr_t; + +typedef struct { + GLenum target; + GLuint texture; + GLuint sampler; +} _sg_gl_cache_texture_sampler_bind_slot; + +#define _SG_GL_MAX_SBUF_BINDINGS (SG_MAX_STORAGEBUFFER_BINDSLOTS) +#define _SG_GL_MAX_IMG_SMP_BINDINGS (SG_MAX_IMAGE_SAMPLER_PAIRS) +typedef struct { + sg_depth_state depth; + sg_stencil_state stencil; + sg_blend_state blend; + sg_color_mask color_write_mask[SG_MAX_COLOR_ATTACHMENTS]; + sg_cull_mode cull_mode; + sg_face_winding face_winding; + bool polygon_offset_enabled; + int sample_count; + sg_color blend_color; + bool alpha_to_coverage_enabled; + _sg_gl_cache_attr_t attrs[SG_MAX_VERTEX_ATTRIBUTES]; + GLuint vertex_buffer; + GLuint index_buffer; + GLuint storage_buffer; // general bind point + GLuint storage_buffers[_SG_GL_MAX_SBUF_BINDINGS]; + GLuint stored_vertex_buffer; + GLuint stored_index_buffer; + GLuint stored_storage_buffer; + GLuint prog; + _sg_gl_cache_texture_sampler_bind_slot texture_samplers[_SG_GL_MAX_IMG_SMP_BINDINGS]; + _sg_gl_cache_texture_sampler_bind_slot stored_texture_sampler; + int cur_ib_offset; + GLenum cur_primitive_type; + GLenum cur_index_type; + GLenum cur_active_texture; + _sg_pipeline_t* cur_pipeline; + sg_pipeline cur_pipeline_id; +} _sg_gl_state_cache_t; + +typedef struct { + bool valid; + GLuint vao; + _sg_gl_state_cache_t cache; + bool ext_anisotropic; + GLint max_anisotropy; + sg_store_action color_store_actions[SG_MAX_COLOR_ATTACHMENTS]; + sg_store_action depth_store_action; + sg_store_action stencil_store_action; + #if _SOKOL_USE_WIN32_GL_LOADER + HINSTANCE opengl32_dll; + #endif +} _sg_gl_backend_t; + +#elif defined(SOKOL_D3D11) + +typedef struct _sg_buffer_s { + _sg_slot_t slot; + _sg_buffer_common_t cmn; + struct { + ID3D11Buffer* buf; + ID3D11ShaderResourceView* srv; + ID3D11UnorderedAccessView* uav; + } d3d11; +} _sg_d3d11_buffer_t; +typedef _sg_d3d11_buffer_t _sg_buffer_t; + +typedef struct _sg_image_s { + _sg_slot_t slot; + _sg_image_common_t cmn; + struct { + DXGI_FORMAT format; + ID3D11Texture2D* tex2d; + ID3D11Texture3D* tex3d; + ID3D11Resource* res; // either tex2d or tex3d + ID3D11ShaderResourceView* srv; + } d3d11; +} _sg_d3d11_image_t; +typedef _sg_d3d11_image_t _sg_image_t; + +typedef struct _sg_sampler_s { + _sg_slot_t slot; + _sg_sampler_common_t cmn; + struct { + ID3D11SamplerState* smp; + } d3d11; +} _sg_d3d11_sampler_t; +typedef _sg_d3d11_sampler_t _sg_sampler_t; + +typedef struct { + _sg_str_t sem_name; + int sem_index; +} _sg_d3d11_shader_attr_t; + +#define _SG_D3D11_MAX_STAGE_UB_BINDINGS (SG_MAX_UNIFORMBLOCK_BINDSLOTS) +#define _SG_D3D11_MAX_STAGE_SRV_BINDINGS (SG_MAX_IMAGE_BINDSLOTS + SG_MAX_STORAGEBUFFER_BINDSLOTS) +#define _SG_D3D11_MAX_STAGE_UAV_BINDINGS (SG_MAX_STORAGEBUFFER_BINDSLOTS) +#define _SG_D3D11_MAX_STAGE_SMP_BINDINGS (SG_MAX_SAMPLER_BINDSLOTS) + +typedef struct _sg_shader_s { + _sg_slot_t slot; + _sg_shader_common_t cmn; + struct { + _sg_d3d11_shader_attr_t attrs[SG_MAX_VERTEX_ATTRIBUTES]; + ID3D11VertexShader* vs; + ID3D11PixelShader* fs; + ID3D11ComputeShader* cs; + void* vs_blob; + size_t vs_blob_length; + uint8_t ub_register_b_n[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + uint8_t img_register_t_n[SG_MAX_IMAGE_BINDSLOTS]; + uint8_t smp_register_s_n[SG_MAX_SAMPLER_BINDSLOTS]; + uint8_t sbuf_register_t_n[SG_MAX_STORAGEBUFFER_BINDSLOTS]; + uint8_t sbuf_register_u_n[SG_MAX_STORAGEBUFFER_BINDSLOTS]; + ID3D11Buffer* all_cbufs[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + ID3D11Buffer* vs_cbufs[_SG_D3D11_MAX_STAGE_UB_BINDINGS]; + ID3D11Buffer* fs_cbufs[_SG_D3D11_MAX_STAGE_UB_BINDINGS]; + ID3D11Buffer* cs_cbufs[_SG_D3D11_MAX_STAGE_UB_BINDINGS]; + } d3d11; +} _sg_d3d11_shader_t; +typedef _sg_d3d11_shader_t _sg_shader_t; + +typedef struct _sg_pipeline_s { + _sg_slot_t slot; + _sg_pipeline_common_t cmn; + _sg_shader_t* shader; + struct { + UINT stencil_ref; + UINT vb_strides[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + D3D_PRIMITIVE_TOPOLOGY topology; + DXGI_FORMAT index_format; + ID3D11InputLayout* il; + ID3D11RasterizerState* rs; + ID3D11DepthStencilState* dss; + ID3D11BlendState* bs; + } d3d11; +} _sg_d3d11_pipeline_t; +typedef _sg_d3d11_pipeline_t _sg_pipeline_t; + +typedef struct { + _sg_image_t* image; + union { + ID3D11RenderTargetView* rtv; + ID3D11DepthStencilView* dsv; + } view; +} _sg_d3d11_attachment_t; + +typedef struct _sg_attachments_s { + _sg_slot_t slot; + _sg_attachments_common_t cmn; + struct { + _sg_d3d11_attachment_t colors[SG_MAX_COLOR_ATTACHMENTS]; + _sg_d3d11_attachment_t resolves[SG_MAX_COLOR_ATTACHMENTS]; + _sg_d3d11_attachment_t depth_stencil; + } d3d11; +} _sg_d3d11_attachments_t; +typedef _sg_d3d11_attachments_t _sg_attachments_t; + +typedef struct { + bool valid; + ID3D11Device* dev; + ID3D11DeviceContext* ctx; + bool use_indexed_draw; + bool use_instanced_draw; + _sg_pipeline_t* cur_pipeline; + sg_pipeline cur_pipeline_id; + struct { + ID3D11RenderTargetView* render_view; + ID3D11RenderTargetView* resolve_view; + } cur_pass; + // on-demand loaded d3dcompiler_47.dll handles + HINSTANCE d3dcompiler_dll; + bool d3dcompiler_dll_load_failed; + pD3DCompile D3DCompile_func; + // global subresourcedata array for texture updates + D3D11_SUBRESOURCE_DATA subres_data[SG_MAX_MIPMAPS * SG_MAX_TEXTUREARRAY_LAYERS]; +} _sg_d3d11_backend_t; + +#elif defined(SOKOL_METAL) + +#if defined(_SG_TARGET_MACOS) || defined(_SG_TARGET_IOS_SIMULATOR) +#define _SG_MTL_UB_ALIGN (256) +#else +#define _SG_MTL_UB_ALIGN (16) +#endif +#define _SG_MTL_INVALID_SLOT_INDEX (0) + +typedef struct { + uint32_t frame_index; // frame index at which it is safe to release this resource + int slot_index; +} _sg_mtl_release_item_t; + +typedef struct { + NSMutableArray* pool; + int num_slots; + int free_queue_top; + int* free_queue; + int release_queue_front; + int release_queue_back; + _sg_mtl_release_item_t* release_queue; +} _sg_mtl_idpool_t; + +typedef struct _sg_buffer_s { + _sg_slot_t slot; + _sg_buffer_common_t cmn; + struct { + int buf[SG_NUM_INFLIGHT_FRAMES]; // index into _sg_mtl_pool + } mtl; +} _sg_mtl_buffer_t; +typedef _sg_mtl_buffer_t _sg_buffer_t; + +typedef struct _sg_image_s { + _sg_slot_t slot; + _sg_image_common_t cmn; + struct { + int tex[SG_NUM_INFLIGHT_FRAMES]; + } mtl; +} _sg_mtl_image_t; +typedef _sg_mtl_image_t _sg_image_t; + +typedef struct _sg_sampler_s { + _sg_slot_t slot; + _sg_sampler_common_t cmn; + struct { + int sampler_state; + } mtl; +} _sg_mtl_sampler_t; +typedef _sg_mtl_sampler_t _sg_sampler_t; + +typedef struct { + int mtl_lib; + int mtl_func; +} _sg_mtl_shader_func_t; + +typedef struct _sg_shader_s { + _sg_slot_t slot; + _sg_shader_common_t cmn; + struct { + _sg_mtl_shader_func_t vertex_func; + _sg_mtl_shader_func_t fragment_func; + _sg_mtl_shader_func_t compute_func; + MTLSize threads_per_threadgroup; + uint8_t ub_buffer_n[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + uint8_t img_texture_n[SG_MAX_IMAGE_BINDSLOTS]; + uint8_t smp_sampler_n[SG_MAX_SAMPLER_BINDSLOTS]; + uint8_t sbuf_buffer_n[SG_MAX_STORAGEBUFFER_BINDSLOTS]; + } mtl; +} _sg_mtl_shader_t; +typedef _sg_mtl_shader_t _sg_shader_t; + +typedef struct _sg_pipeline_s { + _sg_slot_t slot; + _sg_pipeline_common_t cmn; + _sg_shader_t* shader; + struct { + MTLPrimitiveType prim_type; + int index_size; + MTLIndexType index_type; + MTLCullMode cull_mode; + MTLWinding winding; + uint32_t stencil_ref; + MTLSize threads_per_threadgroup; + int cps; // MTLComputePipelineState + int rps; // MTLRenderPipelineState + int dss; // MTLDepthStencilState + } mtl; +} _sg_mtl_pipeline_t; +typedef _sg_mtl_pipeline_t _sg_pipeline_t; + +typedef struct { + _sg_image_t* image; +} _sg_mtl_attachment_t; + +typedef struct _sg_attachments_s { + _sg_slot_t slot; + _sg_attachments_common_t cmn; + struct { + _sg_mtl_attachment_t colors[SG_MAX_COLOR_ATTACHMENTS]; + _sg_mtl_attachment_t resolves[SG_MAX_COLOR_ATTACHMENTS]; + _sg_mtl_attachment_t depth_stencil; + } mtl; +} _sg_mtl_attachments_t; +typedef _sg_mtl_attachments_t _sg_attachments_t; + +// resource binding state cache +#define _SG_MTL_MAX_STAGE_UB_BINDINGS (SG_MAX_UNIFORMBLOCK_BINDSLOTS) +#define _SG_MTL_MAX_STAGE_UB_SBUF_BINDINGS (_SG_MTL_MAX_STAGE_UB_BINDINGS + SG_MAX_STORAGEBUFFER_BINDSLOTS) +#define _SG_MTL_MAX_STAGE_BUFFER_BINDINGS (_SG_MTL_MAX_STAGE_UB_SBUF_BINDINGS + SG_MAX_VERTEXBUFFER_BINDSLOTS) +#define _SG_MTL_MAX_STAGE_IMAGE_BINDINGS (SG_MAX_IMAGE_BINDSLOTS) +#define _SG_MTL_MAX_STAGE_SAMPLER_BINDINGS (SG_MAX_SAMPLER_BINDSLOTS) +typedef struct { + const _sg_pipeline_t* cur_pipeline; + sg_pipeline cur_pipeline_id; + const _sg_buffer_t* cur_indexbuffer; + sg_buffer cur_indexbuffer_id; + int cur_indexbuffer_offset; + int cur_vs_buffer_offsets[_SG_MTL_MAX_STAGE_BUFFER_BINDINGS]; + sg_buffer cur_vs_buffer_ids[_SG_MTL_MAX_STAGE_BUFFER_BINDINGS]; + sg_buffer cur_fs_buffer_ids[_SG_MTL_MAX_STAGE_BUFFER_BINDINGS]; + sg_buffer cur_cs_buffer_ids[_SG_MTL_MAX_STAGE_BUFFER_BINDINGS]; + sg_image cur_vs_image_ids[_SG_MTL_MAX_STAGE_IMAGE_BINDINGS]; + sg_image cur_fs_image_ids[_SG_MTL_MAX_STAGE_IMAGE_BINDINGS]; + sg_image cur_cs_image_ids[_SG_MTL_MAX_STAGE_IMAGE_BINDINGS]; + sg_sampler cur_vs_sampler_ids[_SG_MTL_MAX_STAGE_SAMPLER_BINDINGS]; + sg_sampler cur_fs_sampler_ids[_SG_MTL_MAX_STAGE_SAMPLER_BINDINGS]; + sg_sampler cur_cs_sampler_ids[_SG_MTL_MAX_STAGE_SAMPLER_BINDINGS]; +} _sg_mtl_state_cache_t; + +typedef struct { + bool valid; + bool use_shared_storage_mode; + uint32_t cur_frame_rotate_index; + int ub_size; + int cur_ub_offset; + uint8_t* cur_ub_base_ptr; + _sg_mtl_state_cache_t state_cache; + _sg_mtl_idpool_t idpool; + dispatch_semaphore_t sem; + id device; + id cmd_queue; + id cmd_buffer; + id render_cmd_encoder; + id compute_cmd_encoder; + id cur_drawable; + id uniform_buffers[SG_NUM_INFLIGHT_FRAMES]; +} _sg_mtl_backend_t; + +#elif defined(SOKOL_WGPU) + +#define _SG_WGPU_ROWPITCH_ALIGN (256) +#define _SG_WGPU_MAX_UNIFORM_UPDATE_SIZE (1<<16) // also see WGPULimits.maxUniformBufferBindingSize +#define _SG_WGPU_NUM_BINDGROUPS (2) // 0: uniforms, 1: images, samplers, storage buffers +#define _SG_WGPU_UB_BINDGROUP_INDEX (0) +#define _SG_WGPU_IMG_SMP_SBUF_BINDGROUP_INDEX (1) +#define _SG_WGPU_MAX_UB_BINDGROUP_ENTRIES (SG_MAX_UNIFORMBLOCK_BINDSLOTS) +#define _SG_WGPU_MAX_UB_BINDGROUP_BIND_SLOTS (2 * SG_MAX_UNIFORMBLOCK_BINDSLOTS) +#define _SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES (SG_MAX_IMAGE_BINDSLOTS + SG_MAX_SAMPLER_BINDSLOTS + SG_MAX_STORAGEBUFFER_BINDSLOTS) +#define _SG_WGPU_MAX_IMG_SMP_SBUF_BIND_SLOTS (128) + +typedef struct _sg_buffer_s { + _sg_slot_t slot; + _sg_buffer_common_t cmn; + struct { + WGPUBuffer buf; + } wgpu; +} _sg_wgpu_buffer_t; +typedef _sg_wgpu_buffer_t _sg_buffer_t; + +typedef struct _sg_image_s { + _sg_slot_t slot; + _sg_image_common_t cmn; + struct { + WGPUTexture tex; + WGPUTextureView view; + } wgpu; +} _sg_wgpu_image_t; +typedef _sg_wgpu_image_t _sg_image_t; + +typedef struct _sg_sampler_s { + _sg_slot_t slot; + _sg_sampler_common_t cmn; + struct { + WGPUSampler smp; + } wgpu; +} _sg_wgpu_sampler_t; +typedef _sg_wgpu_sampler_t _sg_sampler_t; + +typedef struct { + WGPUShaderModule module; + _sg_str_t entry; +} _sg_wgpu_shader_func_t; + +typedef struct _sg_shader_s { + _sg_slot_t slot; + _sg_shader_common_t cmn; + struct { + _sg_wgpu_shader_func_t vertex_func; + _sg_wgpu_shader_func_t fragment_func; + _sg_wgpu_shader_func_t compute_func; + WGPUBindGroupLayout bgl_ub; + WGPUBindGroup bg_ub; + WGPUBindGroupLayout bgl_img_smp_sbuf; + // a mapping of sokol-gfx bind slots to setBindGroup dynamic-offset-array indices + uint8_t ub_num_dynoffsets; + uint8_t ub_dynoffsets[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + // indexed by sokol-gfx bind slot: + uint8_t ub_grp0_bnd_n[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + uint8_t img_grp1_bnd_n[SG_MAX_IMAGE_BINDSLOTS]; + uint8_t smp_grp1_bnd_n[SG_MAX_SAMPLER_BINDSLOTS]; + uint8_t sbuf_grp1_bnd_n[SG_MAX_STORAGEBUFFER_BINDSLOTS]; + } wgpu; +} _sg_wgpu_shader_t; +typedef _sg_wgpu_shader_t _sg_shader_t; + +typedef struct _sg_pipeline_s { + _sg_slot_t slot; + _sg_pipeline_common_t cmn; + _sg_shader_t* shader; + struct { + WGPURenderPipeline rpip; + WGPUComputePipeline cpip; + WGPUColor blend_color; + } wgpu; +} _sg_wgpu_pipeline_t; +typedef _sg_wgpu_pipeline_t _sg_pipeline_t; + +typedef struct { + _sg_image_t* image; + WGPUTextureView view; +} _sg_wgpu_attachment_t; + +typedef struct _sg_attachments_s { + _sg_slot_t slot; + _sg_attachments_common_t cmn; + struct { + _sg_wgpu_attachment_t colors[SG_MAX_COLOR_ATTACHMENTS]; + _sg_wgpu_attachment_t resolves[SG_MAX_COLOR_ATTACHMENTS]; + _sg_wgpu_attachment_t depth_stencil; + } wgpu; +} _sg_wgpu_attachments_t; +typedef _sg_wgpu_attachments_t _sg_attachments_t; + +// a pool of per-frame uniform buffers +typedef struct { + uint32_t num_bytes; + uint32_t offset; // current offset into buf + uint8_t* staging; // intermediate buffer for uniform data updates + WGPUBuffer buf; // the GPU-side uniform buffer + uint32_t bind_offsets[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; // NOTE: index is sokol-gfx ub slot index! +} _sg_wgpu_uniform_buffer_t; + +typedef struct { + uint32_t id; +} _sg_wgpu_bindgroup_handle_t; + +typedef enum { + _SG_WGPU_BINDGROUPSCACHEITEMTYPE_NONE = 0, + _SG_WGPU_BINDGROUPSCACHEITEMTYPE_IMAGE = 0x00001111, + _SG_WGPU_BINDGROUPSCACHEITEMTYPE_SAMPLER = 0x00002222, + _SG_WGPU_BINDGROUPSCACHEITEMTYPE_STORAGEBUFFER = 0x00003333, + _SG_WGPU_BINDGROUPSCACHEITEMTYPE_PIPELINE = 0x00004444, +} _sg_wgpu_bindgroups_cache_item_type_t; + +#define _SG_WGPU_BINDGROUPSCACHEKEY_NUM_ITEMS (1 + _SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES) +typedef struct { + uint64_t hash; + // the format of cache key items is BBBBTTTTIIIIIIII + // where + // - BBBB is 2x the WGPU binding + // - TTTT is the _sg_wgpu_bindgroups_cache_item_type_t + // - IIIIIIII is the resource id + // + // where the item type is a per-resource-type bit pattern + uint64_t items[_SG_WGPU_BINDGROUPSCACHEKEY_NUM_ITEMS]; +} _sg_wgpu_bindgroups_cache_key_t; + +typedef struct { + uint32_t num; // must be 2^n + uint32_t index_mask; // mask to turn hash into valid index + _sg_wgpu_bindgroup_handle_t* items; +} _sg_wgpu_bindgroups_cache_t; + +typedef struct { + _sg_slot_t slot; + WGPUBindGroup bindgroup; + _sg_wgpu_bindgroups_cache_key_t key; +} _sg_wgpu_bindgroup_t; + +typedef struct { + _sg_pool_t pool; + _sg_wgpu_bindgroup_t* bindgroups; +} _sg_wgpu_bindgroups_pool_t; + +typedef struct { + struct { + sg_buffer buffer; + uint64_t offset; + } vbs[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + struct { + sg_buffer buffer; + uint64_t offset; + } ib; + _sg_wgpu_bindgroup_handle_t bg; +} _sg_wgpu_bindings_cache_t; + +// the WGPU backend state +typedef struct { + bool valid; + bool use_indexed_draw; + WGPUDevice dev; + WGPUSupportedLimits limits; + WGPUQueue queue; + WGPUCommandEncoder cmd_enc; + WGPURenderPassEncoder rpass_enc; + WGPUComputePassEncoder cpass_enc; + WGPUBindGroup empty_bind_group; + const _sg_pipeline_t* cur_pipeline; + sg_pipeline cur_pipeline_id; + _sg_wgpu_uniform_buffer_t uniform; + _sg_wgpu_bindings_cache_t bindings_cache; + _sg_wgpu_bindgroups_cache_t bindgroups_cache; + _sg_wgpu_bindgroups_pool_t bindgroups_pool; +} _sg_wgpu_backend_t; +#endif + +// POOL STRUCTS + +// this *MUST* remain 0 +#define _SG_INVALID_SLOT_INDEX (0) + +typedef struct _sg_pools_s { + _sg_pool_t buffer_pool; + _sg_pool_t image_pool; + _sg_pool_t sampler_pool; + _sg_pool_t shader_pool; + _sg_pool_t pipeline_pool; + _sg_pool_t attachments_pool; + _sg_buffer_t* buffers; + _sg_image_t* images; + _sg_sampler_t* samplers; + _sg_shader_t* shaders; + _sg_pipeline_t* pipelines; + _sg_attachments_t* attachments; +} _sg_pools_t; + +typedef struct { + int num; // number of allocated commit listener items + int upper; // the current upper index (no valid items past this point) + sg_commit_listener* items; +} _sg_commit_listeners_t; + +// resolved resource bindings struct +typedef struct { + _sg_pipeline_t* pip; + int vb_offsets[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + int ib_offset; + _sg_buffer_t* vbs[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + _sg_buffer_t* ib; + _sg_image_t* imgs[SG_MAX_IMAGE_BINDSLOTS]; + _sg_sampler_t* smps[SG_MAX_SAMPLER_BINDSLOTS]; + _sg_buffer_t* sbufs[SG_MAX_STORAGEBUFFER_BINDSLOTS]; +} _sg_bindings_t; + +typedef struct { + bool sample; + bool filter; + bool render; + bool blend; + bool msaa; + bool depth; +} _sg_pixelformat_info_t; + +typedef struct { + bool valid; + sg_desc desc; // original desc with default values patched in + uint32_t frame_index; + struct { + bool valid; + bool in_pass; + bool is_compute; + sg_attachments atts_id; // SG_INVALID_ID in a swapchain pass + _sg_attachments_t* atts; // 0 in a swapchain pass + int width; + int height; + struct { + sg_pixel_format color_fmt; + sg_pixel_format depth_fmt; + int sample_count; + } swapchain; + } cur_pass; + sg_pipeline cur_pipeline; + bool next_draw_valid; + uint32_t required_bindings_and_uniforms; // used to check that bindings and uniforms are applied after applying pipeline + uint32_t applied_bindings_and_uniforms; // bits 0..7: uniform blocks, bit 8: bindings + #if defined(SOKOL_DEBUG) + sg_log_item validate_error; + #endif + struct { + _sg_tracker_t readwrite_sbufs; // tracks read/write storage buffers used in compute pass + } compute; + _sg_pools_t pools; + sg_backend backend; + sg_features features; + sg_limits limits; + _sg_pixelformat_info_t formats[_SG_PIXELFORMAT_NUM]; + bool stats_enabled; + sg_frame_stats stats; + sg_frame_stats prev_stats; + #if defined(_SOKOL_ANY_GL) + _sg_gl_backend_t gl; + #elif defined(SOKOL_METAL) + _sg_mtl_backend_t mtl; + #elif defined(SOKOL_D3D11) + _sg_d3d11_backend_t d3d11; + #elif defined(SOKOL_WGPU) + _sg_wgpu_backend_t wgpu; + #endif + #if defined(SOKOL_TRACE_HOOKS) + sg_trace_hooks hooks; + #endif + _sg_commit_listeners_t commit_listeners; +} _sg_state_t; +static _sg_state_t _sg; + +// ██ ██████ ██████ ██████ ██ ███ ██ ██████ +// ██ ██ ██ ██ ██ ██ ████ ██ ██ +// ██ ██ ██ ██ ███ ██ ███ ██ ██ ██ ██ ██ ███ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ███████ ██████ ██████ ██████ ██ ██ ████ ██████ +// +// >>logging +#if defined(SOKOL_DEBUG) +#define _SG_LOGITEM_XMACRO(item,msg) #item ": " msg, +static const char* _sg_log_messages[] = { + _SG_LOG_ITEMS +}; +#undef _SG_LOGITEM_XMACRO +#endif // SOKOL_DEBUG + +#define _SG_PANIC(code) _sg_log(SG_LOGITEM_ ##code, 0, 0, __LINE__) +#define _SG_ERROR(code) _sg_log(SG_LOGITEM_ ##code, 1, 0, __LINE__) +#define _SG_WARN(code) _sg_log(SG_LOGITEM_ ##code, 2, 0, __LINE__) +#define _SG_INFO(code) _sg_log(SG_LOGITEM_ ##code, 3, 0, __LINE__) +#define _SG_LOGMSG(code,msg) _sg_log(SG_LOGITEM_ ##code, 3, msg, __LINE__) +#define _SG_VALIDATE(cond,code) if (!(cond)){ _sg.validate_error = SG_LOGITEM_ ##code; _sg_log(SG_LOGITEM_ ##code, 1, 0, __LINE__); } + +static void _sg_log(sg_log_item log_item, uint32_t log_level, const char* msg, uint32_t line_nr) { + if (_sg.desc.logger.func) { + const char* filename = 0; + #if defined(SOKOL_DEBUG) + filename = __FILE__; + if (0 == msg) { + msg = _sg_log_messages[log_item]; + } + #endif + _sg.desc.logger.func("sg", log_level, (uint32_t)log_item, msg, line_nr, filename, _sg.desc.logger.user_data); + } else { + // for log level PANIC it would be 'undefined behaviour' to continue + if (log_level == 0) { + abort(); + } + } +} + +// ███ ███ ███████ ███ ███ ██████ ██████ ██ ██ +// ████ ████ ██ ████ ████ ██ ██ ██ ██ ██ ██ +// ██ ████ ██ █████ ██ ████ ██ ██ ██ ██████ ████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██ ██ ███████ ██ ██ ██████ ██ ██ ██ +// +// >>memory + +// a helper macro to clear a struct with potentially ARC'ed ObjC references +#if defined(SOKOL_METAL) + #if defined(__cplusplus) + #define _SG_CLEAR_ARC_STRUCT(type, item) { item = type(); } + #else + #define _SG_CLEAR_ARC_STRUCT(type, item) { item = (type) { 0 }; } + #endif +#else + #define _SG_CLEAR_ARC_STRUCT(type, item) { _sg_clear(&item, sizeof(item)); } +#endif + +_SOKOL_PRIVATE void _sg_clear(void* ptr, size_t size) { + SOKOL_ASSERT(ptr && (size > 0)); + memset(ptr, 0, size); +} + +_SOKOL_PRIVATE void* _sg_malloc(size_t size) { + SOKOL_ASSERT(size > 0); + void* ptr; + if (_sg.desc.allocator.alloc_fn) { + ptr = _sg.desc.allocator.alloc_fn(size, _sg.desc.allocator.user_data); + } else { + ptr = malloc(size); + } + if (0 == ptr) { + _SG_PANIC(MALLOC_FAILED); + } + return ptr; +} + +_SOKOL_PRIVATE void* _sg_malloc_clear(size_t size) { + void* ptr = _sg_malloc(size); + _sg_clear(ptr, size); + return ptr; +} + +_SOKOL_PRIVATE void _sg_free(void* ptr) { + if (_sg.desc.allocator.free_fn) { + _sg.desc.allocator.free_fn(ptr, _sg.desc.allocator.user_data); + } else { + free(ptr); + } +} + +_SOKOL_PRIVATE bool _sg_strempty(const _sg_str_t* str) { + return 0 == str->buf[0]; +} + +_SOKOL_PRIVATE const char* _sg_strptr(const _sg_str_t* str) { + return &str->buf[0]; +} + +_SOKOL_PRIVATE void _sg_strcpy(_sg_str_t* dst, const char* src) { + SOKOL_ASSERT(dst); + if (src) { + #if defined(_MSC_VER) + strncpy_s(dst->buf, _SG_STRING_SIZE, src, (_SG_STRING_SIZE-1)); + #else + strncpy(dst->buf, src, _SG_STRING_SIZE); + #endif + dst->buf[_SG_STRING_SIZE-1] = 0; + } else { + _sg_clear(dst->buf, _SG_STRING_SIZE); + } +} + +// ██ ██ ███████ ██ ██████ ███████ ██████ ███████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ███████ █████ ██ ██████ █████ ██████ ███████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██ ██ ███████ ███████ ██ ███████ ██ ██ ███████ +// +// >>helpers +_SOKOL_PRIVATE uint32_t _sg_align_u32(uint32_t val, uint32_t align) { + SOKOL_ASSERT((align > 0) && ((align & (align - 1)) == 0)); + return (val + (align - 1)) & ~(align - 1); +} + +typedef struct { int x, y, w, h; } _sg_recti_t; + +_SOKOL_PRIVATE _sg_recti_t _sg_clipi(int x, int y, int w, int h, int clip_width, int clip_height) { + x = _sg_min(_sg_max(0, x), clip_width-1); + y = _sg_min(_sg_max(0, y), clip_height-1); + if ((x + w) > clip_width) { + w = clip_width - x; + } + if ((y + h) > clip_height) { + h = clip_height - y; + } + w = _sg_max(w, 1); + h = _sg_max(h, 1); + const _sg_recti_t res = { x, y, w, h }; + return res; +} + +_SOKOL_PRIVATE int _sg_vertexformat_bytesize(sg_vertex_format fmt) { + switch (fmt) { + case SG_VERTEXFORMAT_FLOAT: return 4; + case SG_VERTEXFORMAT_FLOAT2: return 8; + case SG_VERTEXFORMAT_FLOAT3: return 12; + case SG_VERTEXFORMAT_FLOAT4: return 16; + case SG_VERTEXFORMAT_INT: return 4; + case SG_VERTEXFORMAT_INT2: return 8; + case SG_VERTEXFORMAT_INT3: return 12; + case SG_VERTEXFORMAT_INT4: return 16; + case SG_VERTEXFORMAT_UINT: return 4; + case SG_VERTEXFORMAT_UINT2: return 8; + case SG_VERTEXFORMAT_UINT3: return 12; + case SG_VERTEXFORMAT_UINT4: return 16; + case SG_VERTEXFORMAT_BYTE4: return 4; + case SG_VERTEXFORMAT_BYTE4N: return 4; + case SG_VERTEXFORMAT_UBYTE4: return 4; + case SG_VERTEXFORMAT_UBYTE4N: return 4; + case SG_VERTEXFORMAT_SHORT2: return 4; + case SG_VERTEXFORMAT_SHORT2N: return 4; + case SG_VERTEXFORMAT_USHORT2: return 4; + case SG_VERTEXFORMAT_USHORT2N: return 4; + case SG_VERTEXFORMAT_SHORT4: return 8; + case SG_VERTEXFORMAT_SHORT4N: return 8; + case SG_VERTEXFORMAT_USHORT4: return 8; + case SG_VERTEXFORMAT_USHORT4N: return 8; + case SG_VERTEXFORMAT_UINT10_N2: return 4; + case SG_VERTEXFORMAT_HALF2: return 4; + case SG_VERTEXFORMAT_HALF4: return 8; + case SG_VERTEXFORMAT_INVALID: return 0; + default: + SOKOL_UNREACHABLE; + return -1; + } +} + +_SOKOL_PRIVATE const char* _sg_vertexformat_to_string(sg_vertex_format fmt) { + switch (fmt) { + case SG_VERTEXFORMAT_FLOAT: return "FLOAT"; + case SG_VERTEXFORMAT_FLOAT2: return "FLOAT2"; + case SG_VERTEXFORMAT_FLOAT3: return "FLOAT3"; + case SG_VERTEXFORMAT_FLOAT4: return "FLOAT4"; + case SG_VERTEXFORMAT_INT: return "INT"; + case SG_VERTEXFORMAT_INT2: return "INT2"; + case SG_VERTEXFORMAT_INT3: return "INT3"; + case SG_VERTEXFORMAT_INT4: return "INT4"; + case SG_VERTEXFORMAT_UINT: return "UINT"; + case SG_VERTEXFORMAT_UINT2: return "UINT2"; + case SG_VERTEXFORMAT_UINT3: return "UINT3"; + case SG_VERTEXFORMAT_UINT4: return "UINT4"; + case SG_VERTEXFORMAT_BYTE4: return "BYTE4"; + case SG_VERTEXFORMAT_BYTE4N: return "BYTE4N"; + case SG_VERTEXFORMAT_UBYTE4: return "UBYTE4"; + case SG_VERTEXFORMAT_UBYTE4N: return "UBYTE4N"; + case SG_VERTEXFORMAT_SHORT2: return "SHORT2"; + case SG_VERTEXFORMAT_SHORT2N: return "SHORT2N"; + case SG_VERTEXFORMAT_USHORT2: return "USHORT2"; + case SG_VERTEXFORMAT_USHORT2N: return "USHORT2N"; + case SG_VERTEXFORMAT_SHORT4: return "SHORT4"; + case SG_VERTEXFORMAT_SHORT4N: return "SHORT4N"; + case SG_VERTEXFORMAT_USHORT4: return "USHORT4"; + case SG_VERTEXFORMAT_USHORT4N: return "USHORT4N"; + case SG_VERTEXFORMAT_UINT10_N2: return "UINT10_N2"; + case SG_VERTEXFORMAT_HALF2: return "HALF2"; + case SG_VERTEXFORMAT_HALF4: return "HALF4"; + default: + SOKOL_UNREACHABLE; + return "INVALID"; + } +} + +_SOKOL_PRIVATE const char* _sg_shaderattrbasetype_to_string(sg_shader_attr_base_type b) { + switch (b) { + case SG_SHADERATTRBASETYPE_UNDEFINED: return "UNDEFINED"; + case SG_SHADERATTRBASETYPE_FLOAT: return "FLOAT"; + case SG_SHADERATTRBASETYPE_SINT: return "SINT"; + case SG_SHADERATTRBASETYPE_UINT: return "UINT"; + default: + SOKOL_UNREACHABLE; + return "INVALID"; + } +} + +_SOKOL_PRIVATE sg_shader_attr_base_type _sg_vertexformat_basetype(sg_vertex_format fmt) { + switch (fmt) { + case SG_VERTEXFORMAT_FLOAT: + case SG_VERTEXFORMAT_FLOAT2: + case SG_VERTEXFORMAT_FLOAT3: + case SG_VERTEXFORMAT_FLOAT4: + case SG_VERTEXFORMAT_HALF2: + case SG_VERTEXFORMAT_HALF4: + case SG_VERTEXFORMAT_BYTE4N: + case SG_VERTEXFORMAT_UBYTE4N: + case SG_VERTEXFORMAT_SHORT2N: + case SG_VERTEXFORMAT_USHORT2N: + case SG_VERTEXFORMAT_SHORT4N: + case SG_VERTEXFORMAT_USHORT4N: + case SG_VERTEXFORMAT_UINT10_N2: + return SG_SHADERATTRBASETYPE_FLOAT; + case SG_VERTEXFORMAT_INT: + case SG_VERTEXFORMAT_INT2: + case SG_VERTEXFORMAT_INT3: + case SG_VERTEXFORMAT_INT4: + case SG_VERTEXFORMAT_BYTE4: + case SG_VERTEXFORMAT_SHORT2: + case SG_VERTEXFORMAT_SHORT4: + return SG_SHADERATTRBASETYPE_SINT; + case SG_VERTEXFORMAT_UINT: + case SG_VERTEXFORMAT_UINT2: + case SG_VERTEXFORMAT_UINT3: + case SG_VERTEXFORMAT_UINT4: + case SG_VERTEXFORMAT_UBYTE4: + case SG_VERTEXFORMAT_USHORT2: + case SG_VERTEXFORMAT_USHORT4: + return SG_SHADERATTRBASETYPE_UINT; + default: + SOKOL_UNREACHABLE; + return SG_SHADERATTRBASETYPE_UNDEFINED; + } +} + +_SOKOL_PRIVATE uint32_t _sg_uniform_alignment(sg_uniform_type type, int array_count, sg_uniform_layout ub_layout) { + if (ub_layout == SG_UNIFORMLAYOUT_NATIVE) { + return 1; + } else { + SOKOL_ASSERT(array_count > 0); + if (array_count == 1) { + switch (type) { + case SG_UNIFORMTYPE_FLOAT: + case SG_UNIFORMTYPE_INT: + return 4; + case SG_UNIFORMTYPE_FLOAT2: + case SG_UNIFORMTYPE_INT2: + return 8; + case SG_UNIFORMTYPE_FLOAT3: + case SG_UNIFORMTYPE_FLOAT4: + case SG_UNIFORMTYPE_INT3: + case SG_UNIFORMTYPE_INT4: + return 16; + case SG_UNIFORMTYPE_MAT4: + return 16; + default: + SOKOL_UNREACHABLE; + return 1; + } + } else { + return 16; + } + } +} + +_SOKOL_PRIVATE uint32_t _sg_uniform_size(sg_uniform_type type, int array_count, sg_uniform_layout ub_layout) { + SOKOL_ASSERT(array_count > 0); + if (array_count == 1) { + switch (type) { + case SG_UNIFORMTYPE_FLOAT: + case SG_UNIFORMTYPE_INT: + return 4; + case SG_UNIFORMTYPE_FLOAT2: + case SG_UNIFORMTYPE_INT2: + return 8; + case SG_UNIFORMTYPE_FLOAT3: + case SG_UNIFORMTYPE_INT3: + return 12; + case SG_UNIFORMTYPE_FLOAT4: + case SG_UNIFORMTYPE_INT4: + return 16; + case SG_UNIFORMTYPE_MAT4: + return 64; + default: + SOKOL_UNREACHABLE; + return 0; + } + } else { + if (ub_layout == SG_UNIFORMLAYOUT_NATIVE) { + switch (type) { + case SG_UNIFORMTYPE_FLOAT: + case SG_UNIFORMTYPE_INT: + return 4 * (uint32_t)array_count; + case SG_UNIFORMTYPE_FLOAT2: + case SG_UNIFORMTYPE_INT2: + return 8 * (uint32_t)array_count; + case SG_UNIFORMTYPE_FLOAT3: + case SG_UNIFORMTYPE_INT3: + return 12 * (uint32_t)array_count; + case SG_UNIFORMTYPE_FLOAT4: + case SG_UNIFORMTYPE_INT4: + return 16 * (uint32_t)array_count; + case SG_UNIFORMTYPE_MAT4: + return 64 * (uint32_t)array_count; + default: + SOKOL_UNREACHABLE; + return 0; + } + } else { + switch (type) { + case SG_UNIFORMTYPE_FLOAT: + case SG_UNIFORMTYPE_FLOAT2: + case SG_UNIFORMTYPE_FLOAT3: + case SG_UNIFORMTYPE_FLOAT4: + case SG_UNIFORMTYPE_INT: + case SG_UNIFORMTYPE_INT2: + case SG_UNIFORMTYPE_INT3: + case SG_UNIFORMTYPE_INT4: + return 16 * (uint32_t)array_count; + case SG_UNIFORMTYPE_MAT4: + return 64 * (uint32_t)array_count; + default: + SOKOL_UNREACHABLE; + return 0; + } + } + } +} + +_SOKOL_PRIVATE bool _sg_is_compressed_pixel_format(sg_pixel_format fmt) { + switch (fmt) { + case SG_PIXELFORMAT_BC1_RGBA: + case SG_PIXELFORMAT_BC2_RGBA: + case SG_PIXELFORMAT_BC3_RGBA: + case SG_PIXELFORMAT_BC3_SRGBA: + case SG_PIXELFORMAT_BC4_R: + case SG_PIXELFORMAT_BC4_RSN: + case SG_PIXELFORMAT_BC5_RG: + case SG_PIXELFORMAT_BC5_RGSN: + case SG_PIXELFORMAT_BC6H_RGBF: + case SG_PIXELFORMAT_BC6H_RGBUF: + case SG_PIXELFORMAT_BC7_RGBA: + case SG_PIXELFORMAT_BC7_SRGBA: + case SG_PIXELFORMAT_ETC2_RGB8: + case SG_PIXELFORMAT_ETC2_SRGB8: + case SG_PIXELFORMAT_ETC2_RGB8A1: + case SG_PIXELFORMAT_ETC2_RGBA8: + case SG_PIXELFORMAT_ETC2_SRGB8A8: + case SG_PIXELFORMAT_EAC_R11: + case SG_PIXELFORMAT_EAC_R11SN: + case SG_PIXELFORMAT_EAC_RG11: + case SG_PIXELFORMAT_EAC_RG11SN: + case SG_PIXELFORMAT_ASTC_4x4_RGBA: + case SG_PIXELFORMAT_ASTC_4x4_SRGBA: + return true; + default: + return false; + } +} + +_SOKOL_PRIVATE bool _sg_is_valid_rendertarget_color_format(sg_pixel_format fmt) { + const int fmt_index = (int) fmt; + SOKOL_ASSERT((fmt_index >= 0) && (fmt_index < _SG_PIXELFORMAT_NUM)); + return _sg.formats[fmt_index].render && !_sg.formats[fmt_index].depth; +} + +_SOKOL_PRIVATE bool _sg_is_valid_rendertarget_depth_format(sg_pixel_format fmt) { + const int fmt_index = (int) fmt; + SOKOL_ASSERT((fmt_index >= 0) && (fmt_index < _SG_PIXELFORMAT_NUM)); + return _sg.formats[fmt_index].render && _sg.formats[fmt_index].depth; +} + +_SOKOL_PRIVATE bool _sg_is_depth_or_depth_stencil_format(sg_pixel_format fmt) { + return (SG_PIXELFORMAT_DEPTH == fmt) || (SG_PIXELFORMAT_DEPTH_STENCIL == fmt); +} + +_SOKOL_PRIVATE bool _sg_is_depth_stencil_format(sg_pixel_format fmt) { + return (SG_PIXELFORMAT_DEPTH_STENCIL == fmt); +} + +_SOKOL_PRIVATE int _sg_pixelformat_bytesize(sg_pixel_format fmt) { + switch (fmt) { + case SG_PIXELFORMAT_R8: + case SG_PIXELFORMAT_R8SN: + case SG_PIXELFORMAT_R8UI: + case SG_PIXELFORMAT_R8SI: + return 1; + case SG_PIXELFORMAT_R16: + case SG_PIXELFORMAT_R16SN: + case SG_PIXELFORMAT_R16UI: + case SG_PIXELFORMAT_R16SI: + case SG_PIXELFORMAT_R16F: + case SG_PIXELFORMAT_RG8: + case SG_PIXELFORMAT_RG8SN: + case SG_PIXELFORMAT_RG8UI: + case SG_PIXELFORMAT_RG8SI: + return 2; + case SG_PIXELFORMAT_R32UI: + case SG_PIXELFORMAT_R32SI: + case SG_PIXELFORMAT_R32F: + case SG_PIXELFORMAT_RG16: + case SG_PIXELFORMAT_RG16SN: + case SG_PIXELFORMAT_RG16UI: + case SG_PIXELFORMAT_RG16SI: + case SG_PIXELFORMAT_RG16F: + case SG_PIXELFORMAT_RGBA8: + case SG_PIXELFORMAT_SRGB8A8: + case SG_PIXELFORMAT_RGBA8SN: + case SG_PIXELFORMAT_RGBA8UI: + case SG_PIXELFORMAT_RGBA8SI: + case SG_PIXELFORMAT_BGRA8: + case SG_PIXELFORMAT_RGB10A2: + case SG_PIXELFORMAT_RG11B10F: + case SG_PIXELFORMAT_RGB9E5: + return 4; + case SG_PIXELFORMAT_RG32UI: + case SG_PIXELFORMAT_RG32SI: + case SG_PIXELFORMAT_RG32F: + case SG_PIXELFORMAT_RGBA16: + case SG_PIXELFORMAT_RGBA16SN: + case SG_PIXELFORMAT_RGBA16UI: + case SG_PIXELFORMAT_RGBA16SI: + case SG_PIXELFORMAT_RGBA16F: + return 8; + case SG_PIXELFORMAT_RGBA32UI: + case SG_PIXELFORMAT_RGBA32SI: + case SG_PIXELFORMAT_RGBA32F: + return 16; + case SG_PIXELFORMAT_DEPTH: + case SG_PIXELFORMAT_DEPTH_STENCIL: + return 4; + default: + SOKOL_UNREACHABLE; + return 0; + } +} + +_SOKOL_PRIVATE int _sg_roundup(int val, int round_to) { + return (val+(round_to-1)) & ~(round_to-1); +} + +_SOKOL_PRIVATE uint32_t _sg_roundup_u32(uint32_t val, uint32_t round_to) { + return (val+(round_to-1)) & ~(round_to-1); +} + +_SOKOL_PRIVATE uint64_t _sg_roundup_u64(uint64_t val, uint64_t round_to) { + return (val+(round_to-1)) & ~(round_to-1); +} + +_SOKOL_PRIVATE bool _sg_multiple_u64(uint64_t val, uint64_t of) { + return (val & (of-1)) == 0; +} + +/* return row pitch for an image + + see ComputePitch in https://github.com/microsoft/DirectXTex/blob/master/DirectXTex/DirectXTexUtil.cpp +*/ +_SOKOL_PRIVATE int _sg_row_pitch(sg_pixel_format fmt, int width, int row_align) { + int pitch; + switch (fmt) { + case SG_PIXELFORMAT_BC1_RGBA: + case SG_PIXELFORMAT_BC4_R: + case SG_PIXELFORMAT_BC4_RSN: + case SG_PIXELFORMAT_ETC2_RGB8: + case SG_PIXELFORMAT_ETC2_SRGB8: + case SG_PIXELFORMAT_ETC2_RGB8A1: + case SG_PIXELFORMAT_EAC_R11: + case SG_PIXELFORMAT_EAC_R11SN: + pitch = ((width + 3) / 4) * 8; + pitch = pitch < 8 ? 8 : pitch; + break; + case SG_PIXELFORMAT_BC2_RGBA: + case SG_PIXELFORMAT_BC3_RGBA: + case SG_PIXELFORMAT_BC3_SRGBA: + case SG_PIXELFORMAT_BC5_RG: + case SG_PIXELFORMAT_BC5_RGSN: + case SG_PIXELFORMAT_BC6H_RGBF: + case SG_PIXELFORMAT_BC6H_RGBUF: + case SG_PIXELFORMAT_BC7_RGBA: + case SG_PIXELFORMAT_BC7_SRGBA: + case SG_PIXELFORMAT_ETC2_RGBA8: + case SG_PIXELFORMAT_ETC2_SRGB8A8: + case SG_PIXELFORMAT_EAC_RG11: + case SG_PIXELFORMAT_EAC_RG11SN: + case SG_PIXELFORMAT_ASTC_4x4_RGBA: + case SG_PIXELFORMAT_ASTC_4x4_SRGBA: + pitch = ((width + 3) / 4) * 16; + pitch = pitch < 16 ? 16 : pitch; + break; + default: + pitch = width * _sg_pixelformat_bytesize(fmt); + break; + } + pitch = _sg_roundup(pitch, row_align); + return pitch; +} + +// compute the number of rows in a surface depending on pixel format +_SOKOL_PRIVATE int _sg_num_rows(sg_pixel_format fmt, int height) { + int num_rows; + switch (fmt) { + case SG_PIXELFORMAT_BC1_RGBA: + case SG_PIXELFORMAT_BC4_R: + case SG_PIXELFORMAT_BC4_RSN: + case SG_PIXELFORMAT_ETC2_RGB8: + case SG_PIXELFORMAT_ETC2_SRGB8: + case SG_PIXELFORMAT_ETC2_RGB8A1: + case SG_PIXELFORMAT_ETC2_RGBA8: + case SG_PIXELFORMAT_ETC2_SRGB8A8: + case SG_PIXELFORMAT_EAC_R11: + case SG_PIXELFORMAT_EAC_R11SN: + case SG_PIXELFORMAT_EAC_RG11: + case SG_PIXELFORMAT_EAC_RG11SN: + case SG_PIXELFORMAT_BC2_RGBA: + case SG_PIXELFORMAT_BC3_RGBA: + case SG_PIXELFORMAT_BC3_SRGBA: + case SG_PIXELFORMAT_BC5_RG: + case SG_PIXELFORMAT_BC5_RGSN: + case SG_PIXELFORMAT_BC6H_RGBF: + case SG_PIXELFORMAT_BC6H_RGBUF: + case SG_PIXELFORMAT_BC7_RGBA: + case SG_PIXELFORMAT_BC7_SRGBA: + case SG_PIXELFORMAT_ASTC_4x4_RGBA: + case SG_PIXELFORMAT_ASTC_4x4_SRGBA: + num_rows = ((height + 3) / 4); + break; + default: + num_rows = height; + break; + } + if (num_rows < 1) { + num_rows = 1; + } + return num_rows; +} + +// return size of a mipmap level +_SOKOL_PRIVATE int _sg_miplevel_dim(int base_dim, int mip_level) { + return _sg_max(base_dim >> mip_level, 1); +} + +/* return pitch of a 2D subimage / texture slice + see ComputePitch in https://github.com/microsoft/DirectXTex/blob/master/DirectXTex/DirectXTexUtil.cpp +*/ +_SOKOL_PRIVATE int _sg_surface_pitch(sg_pixel_format fmt, int width, int height, int row_align) { + int num_rows = _sg_num_rows(fmt, height); + return num_rows * _sg_row_pitch(fmt, width, row_align); +} + +// capability table pixel format helper functions +_SOKOL_PRIVATE void _sg_pixelformat_all(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->filter = true; + pfi->blend = true; + pfi->render = true; + pfi->msaa = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_s(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_sf(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->filter = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_sr(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->render = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_sfr(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->filter = true; + pfi->render = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_srmd(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->render = true; + pfi->msaa = true; + pfi->depth = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_srm(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->render = true; + pfi->msaa = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_sfrm(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->filter = true; + pfi->render = true; + pfi->msaa = true; +} +_SOKOL_PRIVATE void _sg_pixelformat_sbrm(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->blend = true; + pfi->render = true; + pfi->msaa = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_sbr(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->blend = true; + pfi->render = true; +} + +_SOKOL_PRIVATE void _sg_pixelformat_sfbr(_sg_pixelformat_info_t* pfi) { + pfi->sample = true; + pfi->filter = true; + pfi->blend = true; + pfi->render = true; +} + +_SOKOL_PRIVATE sg_pass_action _sg_pass_action_defaults(const sg_pass_action* action) { + SOKOL_ASSERT(action); + sg_pass_action res = *action; + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + if (res.colors[i].load_action == _SG_LOADACTION_DEFAULT) { + res.colors[i].load_action = SG_LOADACTION_CLEAR; + res.colors[i].clear_value.r = SG_DEFAULT_CLEAR_RED; + res.colors[i].clear_value.g = SG_DEFAULT_CLEAR_GREEN; + res.colors[i].clear_value.b = SG_DEFAULT_CLEAR_BLUE; + res.colors[i].clear_value.a = SG_DEFAULT_CLEAR_ALPHA; + } + if (res.colors[i].store_action == _SG_STOREACTION_DEFAULT) { + res.colors[i].store_action = SG_STOREACTION_STORE; + } + } + if (res.depth.load_action == _SG_LOADACTION_DEFAULT) { + res.depth.load_action = SG_LOADACTION_CLEAR; + res.depth.clear_value = SG_DEFAULT_CLEAR_DEPTH; + } + if (res.depth.store_action == _SG_STOREACTION_DEFAULT) { + res.depth.store_action = SG_STOREACTION_DONTCARE; + } + if (res.stencil.load_action == _SG_LOADACTION_DEFAULT) { + res.stencil.load_action = SG_LOADACTION_CLEAR; + res.stencil.clear_value = SG_DEFAULT_CLEAR_STENCIL; + } + if (res.stencil.store_action == _SG_STOREACTION_DEFAULT) { + res.stencil.store_action = SG_STOREACTION_DONTCARE; + } + return res; +} + +// ██████ ██ ██ ███ ███ ███ ███ ██ ██ ██████ █████ ██████ ██ ██ ███████ ███ ██ ██████ +// ██ ██ ██ ██ ████ ████ ████ ████ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ████ ██ ██ ██ +// ██ ██ ██ ██ ██ ████ ██ ██ ████ ██ ████ ██████ ███████ ██ █████ █████ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██████ ██████ ██ ██ ██ ██ ██ ██████ ██ ██ ██████ ██ ██ ███████ ██ ████ ██████ +// +// >>dummy backend +#if defined(SOKOL_DUMMY_BACKEND) + +_SOKOL_PRIVATE void _sg_dummy_setup_backend(const sg_desc* desc) { + SOKOL_ASSERT(desc); + _SOKOL_UNUSED(desc); + _sg.backend = SG_BACKEND_DUMMY; + for (int i = SG_PIXELFORMAT_R8; i < SG_PIXELFORMAT_BC1_RGBA; i++) { + _sg.formats[i].sample = true; + _sg.formats[i].filter = true; + _sg.formats[i].render = true; + _sg.formats[i].blend = true; + _sg.formats[i].msaa = true; + } + _sg.formats[SG_PIXELFORMAT_DEPTH].depth = true; + _sg.formats[SG_PIXELFORMAT_DEPTH_STENCIL].depth = true; +} + +_SOKOL_PRIVATE void _sg_dummy_discard_backend(void) { + // empty +} + +_SOKOL_PRIVATE void _sg_dummy_reset_state_cache(void) { + // empty +} + +_SOKOL_PRIVATE sg_resource_state _sg_dummy_create_buffer(_sg_buffer_t* buf, const sg_buffer_desc* desc) { + SOKOL_ASSERT(buf && desc); + _SOKOL_UNUSED(buf); + _SOKOL_UNUSED(desc); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_dummy_discard_buffer(_sg_buffer_t* buf) { + SOKOL_ASSERT(buf); + _SOKOL_UNUSED(buf); +} + +_SOKOL_PRIVATE sg_resource_state _sg_dummy_create_image(_sg_image_t* img, const sg_image_desc* desc) { + SOKOL_ASSERT(img && desc); + _SOKOL_UNUSED(img); + _SOKOL_UNUSED(desc); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_dummy_discard_image(_sg_image_t* img) { + SOKOL_ASSERT(img); + _SOKOL_UNUSED(img); +} + +_SOKOL_PRIVATE sg_resource_state _sg_dummy_create_sampler(_sg_sampler_t* smp, const sg_sampler_desc* desc) { + SOKOL_ASSERT(smp && desc); + _SOKOL_UNUSED(smp); + _SOKOL_UNUSED(desc); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_dummy_discard_sampler(_sg_sampler_t* smp) { + SOKOL_ASSERT(smp); + _SOKOL_UNUSED(smp); +} + +_SOKOL_PRIVATE sg_resource_state _sg_dummy_create_shader(_sg_shader_t* shd, const sg_shader_desc* desc) { + SOKOL_ASSERT(shd && desc); + _SOKOL_UNUSED(shd); + _SOKOL_UNUSED(desc); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_dummy_discard_shader(_sg_shader_t* shd) { + SOKOL_ASSERT(shd); + _SOKOL_UNUSED(shd); +} + +_SOKOL_PRIVATE sg_resource_state _sg_dummy_create_pipeline(_sg_pipeline_t* pip, _sg_shader_t* shd, const sg_pipeline_desc* desc) { + SOKOL_ASSERT(pip && desc); + _SOKOL_UNUSED(desc); + pip->shader = shd; + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_dummy_discard_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + _SOKOL_UNUSED(pip); +} + +_SOKOL_PRIVATE sg_resource_state _sg_dummy_create_attachments(_sg_attachments_t* atts, _sg_image_t** color_images, _sg_image_t** resolve_images, _sg_image_t* ds_img, const sg_attachments_desc* desc) { + SOKOL_ASSERT(atts && desc); + SOKOL_ASSERT(color_images && resolve_images); + + for (int i = 0; i < atts->cmn.num_colors; i++) { + const sg_attachment_desc* color_desc = &desc->colors[i]; + _SOKOL_UNUSED(color_desc); + SOKOL_ASSERT(color_desc->image.id != SG_INVALID_ID); + SOKOL_ASSERT(0 == atts->dmy.colors[i].image); + SOKOL_ASSERT(color_images[i] && (color_images[i]->slot.id == color_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_color_format(color_images[i]->cmn.pixel_format)); + atts->dmy.colors[i].image = color_images[i]; + + const sg_attachment_desc* resolve_desc = &desc->resolves[i]; + if (resolve_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(0 == atts->dmy.resolves[i].image); + SOKOL_ASSERT(resolve_images[i] && (resolve_images[i]->slot.id == resolve_desc->image.id)); + SOKOL_ASSERT(color_images[i] && (color_images[i]->cmn.pixel_format == resolve_images[i]->cmn.pixel_format)); + atts->dmy.resolves[i].image = resolve_images[i]; + } + } + + SOKOL_ASSERT(0 == atts->dmy.depth_stencil.image); + const sg_attachment_desc* ds_desc = &desc->depth_stencil; + if (ds_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(ds_img && (ds_img->slot.id == ds_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_depth_format(ds_img->cmn.pixel_format)); + atts->dmy.depth_stencil.image = ds_img; + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_dummy_discard_attachments(_sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + _SOKOL_UNUSED(atts); +} + +_SOKOL_PRIVATE _sg_image_t* _sg_dummy_attachments_color_image(const _sg_attachments_t* atts, int index) { + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + return atts->dmy.colors[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_dummy_attachments_resolve_image(const _sg_attachments_t* atts, int index) { + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + return atts->dmy.resolves[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_dummy_attachments_ds_image(const _sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + return atts->dmy.depth_stencil.image; +} + +_SOKOL_PRIVATE void _sg_dummy_begin_pass(const sg_pass* pass) { + SOKOL_ASSERT(pass); + _SOKOL_UNUSED(pass); +} + +_SOKOL_PRIVATE void _sg_dummy_end_pass(void) { + // empty +} + +_SOKOL_PRIVATE void _sg_dummy_commit(void) { + // empty +} + +_SOKOL_PRIVATE void _sg_dummy_apply_viewport(int x, int y, int w, int h, bool origin_top_left) { + _SOKOL_UNUSED(x); + _SOKOL_UNUSED(y); + _SOKOL_UNUSED(w); + _SOKOL_UNUSED(h); + _SOKOL_UNUSED(origin_top_left); +} + +_SOKOL_PRIVATE void _sg_dummy_apply_scissor_rect(int x, int y, int w, int h, bool origin_top_left) { + _SOKOL_UNUSED(x); + _SOKOL_UNUSED(y); + _SOKOL_UNUSED(w); + _SOKOL_UNUSED(h); + _SOKOL_UNUSED(origin_top_left); +} + +_SOKOL_PRIVATE void _sg_dummy_apply_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + _SOKOL_UNUSED(pip); +} + +_SOKOL_PRIVATE bool _sg_dummy_apply_bindings(_sg_bindings_t* bnd) { + SOKOL_ASSERT(bnd); + SOKOL_ASSERT(bnd->pip); + _SOKOL_UNUSED(bnd); + return true; +} + +_SOKOL_PRIVATE void _sg_dummy_apply_uniforms(int ub_slot, const sg_range* data) { + _SOKOL_UNUSED(ub_slot); + _SOKOL_UNUSED(data); +} + +_SOKOL_PRIVATE void _sg_dummy_draw(int base_element, int num_elements, int num_instances) { + _SOKOL_UNUSED(base_element); + _SOKOL_UNUSED(num_elements); + _SOKOL_UNUSED(num_instances); +} + +_SOKOL_PRIVATE void _sg_dummy_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) { + _SOKOL_UNUSED(num_groups_x); + _SOKOL_UNUSED(num_groups_y); + _SOKOL_UNUSED(num_groups_z); +} + +_SOKOL_PRIVATE void _sg_dummy_update_buffer(_sg_buffer_t* buf, const sg_range* data) { + SOKOL_ASSERT(buf && data && data->ptr && (data->size > 0)); + _SOKOL_UNUSED(data); + if (++buf->cmn.active_slot >= buf->cmn.num_slots) { + buf->cmn.active_slot = 0; + } +} + +_SOKOL_PRIVATE bool _sg_dummy_append_buffer(_sg_buffer_t* buf, const sg_range* data, bool new_frame) { + SOKOL_ASSERT(buf && data && data->ptr && (data->size > 0)); + _SOKOL_UNUSED(data); + if (new_frame) { + if (++buf->cmn.active_slot >= buf->cmn.num_slots) { + buf->cmn.active_slot = 0; + } + } + return true; +} + +_SOKOL_PRIVATE void _sg_dummy_update_image(_sg_image_t* img, const sg_image_data* data) { + SOKOL_ASSERT(img && data); + _SOKOL_UNUSED(data); + if (++img->cmn.active_slot >= img->cmn.num_slots) { + img->cmn.active_slot = 0; + } +} + +// ██████ ██████ ███████ ███ ██ ██████ ██ ██████ █████ ██████ ██ ██ ███████ ███ ██ ██████ +// ██ ██ ██ ██ ██ ████ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ████ ██ ██ ██ +// ██ ██ ██████ █████ ██ ██ ██ ██ ███ ██ ██████ ███████ ██ █████ █████ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██████ ██ ███████ ██ ████ ██████ ███████ ██████ ██ ██ ██████ ██ ██ ███████ ██ ████ ██████ +// +// >>opengl backend +#elif defined(_SOKOL_ANY_GL) + +// optional GL loader for win32 +#if defined(_SOKOL_USE_WIN32_GL_LOADER) + +#ifndef SG_GL_FUNCS_EXT +#define SG_GL_FUNCS_EXT +#endif + +// X Macro list of GL function names and signatures +#define _SG_GL_FUNCS \ + SG_GL_FUNCS_EXT \ + _SG_XMACRO(glBindVertexArray, void, (GLuint array)) \ + _SG_XMACRO(glFramebufferTextureLayer, void, (GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer)) \ + _SG_XMACRO(glGenFramebuffers, void, (GLsizei n, GLuint * framebuffers)) \ + _SG_XMACRO(glBindFramebuffer, void, (GLenum target, GLuint framebuffer)) \ + _SG_XMACRO(glBindRenderbuffer, void, (GLenum target, GLuint renderbuffer)) \ + _SG_XMACRO(glGetStringi, const GLubyte *, (GLenum name, GLuint index)) \ + _SG_XMACRO(glClearBufferfi, void, (GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil)) \ + _SG_XMACRO(glClearBufferfv, void, (GLenum buffer, GLint drawbuffer, const GLfloat * value)) \ + _SG_XMACRO(glClearBufferuiv, void, (GLenum buffer, GLint drawbuffer, const GLuint * value)) \ + _SG_XMACRO(glClearBufferiv, void, (GLenum buffer, GLint drawbuffer, const GLint * value)) \ + _SG_XMACRO(glDeleteRenderbuffers, void, (GLsizei n, const GLuint * renderbuffers)) \ + _SG_XMACRO(glUniform1fv, void, (GLint location, GLsizei count, const GLfloat * value)) \ + _SG_XMACRO(glUniform2fv, void, (GLint location, GLsizei count, const GLfloat * value)) \ + _SG_XMACRO(glUniform3fv, void, (GLint location, GLsizei count, const GLfloat * value)) \ + _SG_XMACRO(glUniform4fv, void, (GLint location, GLsizei count, const GLfloat * value)) \ + _SG_XMACRO(glUniform1iv, void, (GLint location, GLsizei count, const GLint * value)) \ + _SG_XMACRO(glUniform2iv, void, (GLint location, GLsizei count, const GLint * value)) \ + _SG_XMACRO(glUniform3iv, void, (GLint location, GLsizei count, const GLint * value)) \ + _SG_XMACRO(glUniform4iv, void, (GLint location, GLsizei count, const GLint * value)) \ + _SG_XMACRO(glUniformMatrix4fv, void, (GLint location, GLsizei count, GLboolean transpose, const GLfloat * value)) \ + _SG_XMACRO(glUseProgram, void, (GLuint program)) \ + _SG_XMACRO(glShaderSource, void, (GLuint shader, GLsizei count, const GLchar *const* string, const GLint * length)) \ + _SG_XMACRO(glLinkProgram, void, (GLuint program)) \ + _SG_XMACRO(glGetUniformLocation, GLint, (GLuint program, const GLchar * name)) \ + _SG_XMACRO(glGetShaderiv, void, (GLuint shader, GLenum pname, GLint * params)) \ + _SG_XMACRO(glGetProgramInfoLog, void, (GLuint program, GLsizei bufSize, GLsizei * length, GLchar * infoLog)) \ + _SG_XMACRO(glGetAttribLocation, GLint, (GLuint program, const GLchar * name)) \ + _SG_XMACRO(glDisableVertexAttribArray, void, (GLuint index)) \ + _SG_XMACRO(glDeleteShader, void, (GLuint shader)) \ + _SG_XMACRO(glDeleteProgram, void, (GLuint program)) \ + _SG_XMACRO(glCompileShader, void, (GLuint shader)) \ + _SG_XMACRO(glStencilFuncSeparate, void, (GLenum face, GLenum func, GLint ref, GLuint mask)) \ + _SG_XMACRO(glStencilOpSeparate, void, (GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass)) \ + _SG_XMACRO(glRenderbufferStorageMultisample, void, (GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height)) \ + _SG_XMACRO(glDrawBuffers, void, (GLsizei n, const GLenum * bufs)) \ + _SG_XMACRO(glVertexAttribDivisor, void, (GLuint index, GLuint divisor)) \ + _SG_XMACRO(glBufferSubData, void, (GLenum target, GLintptr offset, GLsizeiptr size, const void * data)) \ + _SG_XMACRO(glGenBuffers, void, (GLsizei n, GLuint * buffers)) \ + _SG_XMACRO(glCheckFramebufferStatus, GLenum, (GLenum target)) \ + _SG_XMACRO(glFramebufferRenderbuffer, void, (GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer)) \ + _SG_XMACRO(glCompressedTexImage2D, void, (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const void * data)) \ + _SG_XMACRO(glCompressedTexImage3D, void, (GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLsizei imageSize, const void * data)) \ + _SG_XMACRO(glActiveTexture, void, (GLenum texture)) \ + _SG_XMACRO(glTexSubImage3D, void, (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, const void * pixels)) \ + _SG_XMACRO(glRenderbufferStorage, void, (GLenum target, GLenum internalformat, GLsizei width, GLsizei height)) \ + _SG_XMACRO(glGenTextures, void, (GLsizei n, GLuint * textures)) \ + _SG_XMACRO(glPolygonOffset, void, (GLfloat factor, GLfloat units)) \ + _SG_XMACRO(glDrawElements, void, (GLenum mode, GLsizei count, GLenum type, const void * indices)) \ + _SG_XMACRO(glDeleteFramebuffers, void, (GLsizei n, const GLuint * framebuffers)) \ + _SG_XMACRO(glBlendEquationSeparate, void, (GLenum modeRGB, GLenum modeAlpha)) \ + _SG_XMACRO(glDeleteTextures, void, (GLsizei n, const GLuint * textures)) \ + _SG_XMACRO(glGetProgramiv, void, (GLuint program, GLenum pname, GLint * params)) \ + _SG_XMACRO(glBindTexture, void, (GLenum target, GLuint texture)) \ + _SG_XMACRO(glTexImage3D, void, (GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLenum format, GLenum type, const void * pixels)) \ + _SG_XMACRO(glCreateShader, GLuint, (GLenum type)) \ + _SG_XMACRO(glTexSubImage2D, void, (GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLenum type, const void * pixels)) \ + _SG_XMACRO(glFramebufferTexture2D, void, (GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level)) \ + _SG_XMACRO(glCreateProgram, GLuint, (void)) \ + _SG_XMACRO(glViewport, void, (GLint x, GLint y, GLsizei width, GLsizei height)) \ + _SG_XMACRO(glDeleteBuffers, void, (GLsizei n, const GLuint * buffers)) \ + _SG_XMACRO(glDrawArrays, void, (GLenum mode, GLint first, GLsizei count)) \ + _SG_XMACRO(glDrawElementsInstanced, void, (GLenum mode, GLsizei count, GLenum type, const void * indices, GLsizei instancecount)) \ + _SG_XMACRO(glVertexAttribPointer, void, (GLuint index, GLint size, GLenum type, GLboolean normalized, GLsizei stride, const void * pointer)) \ + _SG_XMACRO(glVertexAttribIPointer, void, (GLuint index, GLint size, GLenum type, GLsizei stride, const void * pointer)) \ + _SG_XMACRO(glUniform1i, void, (GLint location, GLint v0)) \ + _SG_XMACRO(glDisable, void, (GLenum cap)) \ + _SG_XMACRO(glColorMask, void, (GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha)) \ + _SG_XMACRO(glColorMaski, void, (GLuint buf, GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha)) \ + _SG_XMACRO(glBindBuffer, void, (GLenum target, GLuint buffer)) \ + _SG_XMACRO(glDeleteVertexArrays, void, (GLsizei n, const GLuint * arrays)) \ + _SG_XMACRO(glDepthMask, void, (GLboolean flag)) \ + _SG_XMACRO(glDrawArraysInstanced, void, (GLenum mode, GLint first, GLsizei count, GLsizei instancecount)) \ + _SG_XMACRO(glScissor, void, (GLint x, GLint y, GLsizei width, GLsizei height)) \ + _SG_XMACRO(glGenRenderbuffers, void, (GLsizei n, GLuint * renderbuffers)) \ + _SG_XMACRO(glBufferData, void, (GLenum target, GLsizeiptr size, const void * data, GLenum usage)) \ + _SG_XMACRO(glBlendFuncSeparate, void, (GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha)) \ + _SG_XMACRO(glTexParameteri, void, (GLenum target, GLenum pname, GLint param)) \ + _SG_XMACRO(glGetIntegerv, void, (GLenum pname, GLint * data)) \ + _SG_XMACRO(glEnable, void, (GLenum cap)) \ + _SG_XMACRO(glBlitFramebuffer, void, (GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter)) \ + _SG_XMACRO(glStencilMask, void, (GLuint mask)) \ + _SG_XMACRO(glAttachShader, void, (GLuint program, GLuint shader)) \ + _SG_XMACRO(glGetError, GLenum, (void)) \ + _SG_XMACRO(glBlendColor, void, (GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha)) \ + _SG_XMACRO(glTexParameterf, void, (GLenum target, GLenum pname, GLfloat param)) \ + _SG_XMACRO(glTexParameterfv, void, (GLenum target, GLenum pname, const GLfloat* params)) \ + _SG_XMACRO(glGetShaderInfoLog, void, (GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * infoLog)) \ + _SG_XMACRO(glDepthFunc, void, (GLenum func)) \ + _SG_XMACRO(glStencilOp , void, (GLenum fail, GLenum zfail, GLenum zpass)) \ + _SG_XMACRO(glStencilFunc, void, (GLenum func, GLint ref, GLuint mask)) \ + _SG_XMACRO(glEnableVertexAttribArray, void, (GLuint index)) \ + _SG_XMACRO(glBlendFunc, void, (GLenum sfactor, GLenum dfactor)) \ + _SG_XMACRO(glReadBuffer, void, (GLenum src)) \ + _SG_XMACRO(glTexImage2D, void, (GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLint border, GLenum format, GLenum type, const void * pixels)) \ + _SG_XMACRO(glGenVertexArrays, void, (GLsizei n, GLuint * arrays)) \ + _SG_XMACRO(glFrontFace, void, (GLenum mode)) \ + _SG_XMACRO(glCullFace, void, (GLenum mode)) \ + _SG_XMACRO(glPixelStorei, void, (GLenum pname, GLint param)) \ + _SG_XMACRO(glBindSampler, void, (GLuint unit, GLuint sampler)) \ + _SG_XMACRO(glGenSamplers, void, (GLsizei n, GLuint* samplers)) \ + _SG_XMACRO(glSamplerParameteri, void, (GLuint sampler, GLenum pname, GLint param)) \ + _SG_XMACRO(glSamplerParameterf, void, (GLuint sampler, GLenum pname, GLfloat param)) \ + _SG_XMACRO(glSamplerParameterfv, void, (GLuint sampler, GLenum pname, const GLfloat* params)) \ + _SG_XMACRO(glDeleteSamplers, void, (GLsizei n, const GLuint* samplers)) \ + _SG_XMACRO(glBindBufferBase, void, (GLenum target, GLuint index, GLuint buffer)) \ + _SG_XMACRO(glTexImage2DMultisample, void, (GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height, GLboolean fixedsamplelocations)) \ + _SG_XMACRO(glTexImage3DMultisample, void, (GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLboolean fixedsamplelocations)) \ + _SG_XMACRO(glDispatchCompute, void, (GLuint num_groups_x, GLuint num_groups_y, GLuint num_groups_z)) \ + _SG_XMACRO(glMemoryBarrier, void, (GLbitfield barriers)) + +// generate GL function pointer typedefs +#define _SG_XMACRO(name, ret, args) typedef ret (GL_APIENTRY* PFN_ ## name) args; +_SG_GL_FUNCS +#undef _SG_XMACRO + +// generate GL function pointers +#define _SG_XMACRO(name, ret, args) static PFN_ ## name name; +_SG_GL_FUNCS +#undef _SG_XMACRO + +// helper function to lookup GL functions in GL DLL +typedef PROC (WINAPI * _sg_wglGetProcAddress)(LPCSTR); +_SOKOL_PRIVATE void* _sg_gl_getprocaddr(const char* name, _sg_wglGetProcAddress wgl_getprocaddress) { + void* proc_addr = (void*) wgl_getprocaddress(name); + if (0 == proc_addr) { + proc_addr = (void*) GetProcAddress(_sg.gl.opengl32_dll, name); + } + SOKOL_ASSERT(proc_addr); + return proc_addr; +} + +// populate GL function pointers +_SOKOL_PRIVATE void _sg_gl_load_opengl(void) { + SOKOL_ASSERT(0 == _sg.gl.opengl32_dll); + _sg.gl.opengl32_dll = LoadLibraryA("opengl32.dll"); + SOKOL_ASSERT(_sg.gl.opengl32_dll); + _sg_wglGetProcAddress wgl_getprocaddress = (_sg_wglGetProcAddress) GetProcAddress(_sg.gl.opengl32_dll, "wglGetProcAddress"); + SOKOL_ASSERT(wgl_getprocaddress); + #define _SG_XMACRO(name, ret, args) name = (PFN_ ## name) _sg_gl_getprocaddr(#name, wgl_getprocaddress); + _SG_GL_FUNCS + #undef _SG_XMACRO +} + +_SOKOL_PRIVATE void _sg_gl_unload_opengl(void) { + SOKOL_ASSERT(_sg.gl.opengl32_dll); + FreeLibrary(_sg.gl.opengl32_dll); + _sg.gl.opengl32_dll = 0; +} +#endif // _SOKOL_USE_WIN32_GL_LOADER + +//-- type translation ---------------------------------------------------------- +_SOKOL_PRIVATE GLenum _sg_gl_buffer_target(sg_buffer_type t) { + switch (t) { + case SG_BUFFERTYPE_VERTEXBUFFER: return GL_ARRAY_BUFFER; + case SG_BUFFERTYPE_INDEXBUFFER: return GL_ELEMENT_ARRAY_BUFFER; + case SG_BUFFERTYPE_STORAGEBUFFER: return GL_SHADER_STORAGE_BUFFER; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_texture_target(sg_image_type t, int sample_count) { + #if defined(SOKOL_GLCORE) + const bool msaa = sample_count > 1; + if (msaa) { + switch (t) { + case SG_IMAGETYPE_2D: return GL_TEXTURE_2D_MULTISAMPLE; + case SG_IMAGETYPE_ARRAY: return GL_TEXTURE_2D_MULTISAMPLE_ARRAY; + default: SOKOL_UNREACHABLE; return 0; + } + } else { + switch (t) { + case SG_IMAGETYPE_2D: return GL_TEXTURE_2D; + case SG_IMAGETYPE_CUBE: return GL_TEXTURE_CUBE_MAP; + case SG_IMAGETYPE_3D: return GL_TEXTURE_3D; + case SG_IMAGETYPE_ARRAY: return GL_TEXTURE_2D_ARRAY; + default: SOKOL_UNREACHABLE; return 0; + } + } + #else + SOKOL_ASSERT(sample_count == 1); _SOKOL_UNUSED(sample_count); + switch (t) { + case SG_IMAGETYPE_2D: return GL_TEXTURE_2D; + case SG_IMAGETYPE_CUBE: return GL_TEXTURE_CUBE_MAP; + case SG_IMAGETYPE_3D: return GL_TEXTURE_3D; + case SG_IMAGETYPE_ARRAY: return GL_TEXTURE_2D_ARRAY; + default: SOKOL_UNREACHABLE; return 0; + } + #endif +} + +_SOKOL_PRIVATE GLenum _sg_gl_usage(sg_usage u) { + switch (u) { + case SG_USAGE_IMMUTABLE: return GL_STATIC_DRAW; + case SG_USAGE_DYNAMIC: return GL_DYNAMIC_DRAW; + case SG_USAGE_STREAM: return GL_STREAM_DRAW; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_shader_stage(sg_shader_stage stage) { + switch (stage) { + case SG_SHADERSTAGE_VERTEX: return GL_VERTEX_SHADER; + case SG_SHADERSTAGE_FRAGMENT: return GL_FRAGMENT_SHADER; + case SG_SHADERSTAGE_COMPUTE: return GL_COMPUTE_SHADER; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLint _sg_gl_vertexformat_size(sg_vertex_format fmt) { + switch (fmt) { + case SG_VERTEXFORMAT_FLOAT: return 1; + case SG_VERTEXFORMAT_FLOAT2: return 2; + case SG_VERTEXFORMAT_FLOAT3: return 3; + case SG_VERTEXFORMAT_FLOAT4: return 4; + case SG_VERTEXFORMAT_INT: return 1; + case SG_VERTEXFORMAT_INT2: return 2; + case SG_VERTEXFORMAT_INT3: return 3; + case SG_VERTEXFORMAT_INT4: return 4; + case SG_VERTEXFORMAT_UINT: return 1; + case SG_VERTEXFORMAT_UINT2: return 2; + case SG_VERTEXFORMAT_UINT3: return 3; + case SG_VERTEXFORMAT_UINT4: return 4; + case SG_VERTEXFORMAT_BYTE4: return 4; + case SG_VERTEXFORMAT_BYTE4N: return 4; + case SG_VERTEXFORMAT_UBYTE4: return 4; + case SG_VERTEXFORMAT_UBYTE4N: return 4; + case SG_VERTEXFORMAT_SHORT2: return 2; + case SG_VERTEXFORMAT_SHORT2N: return 2; + case SG_VERTEXFORMAT_USHORT2: return 2; + case SG_VERTEXFORMAT_USHORT2N: return 2; + case SG_VERTEXFORMAT_SHORT4: return 4; + case SG_VERTEXFORMAT_SHORT4N: return 4; + case SG_VERTEXFORMAT_USHORT4: return 4; + case SG_VERTEXFORMAT_USHORT4N: return 4; + case SG_VERTEXFORMAT_UINT10_N2: return 4; + case SG_VERTEXFORMAT_HALF2: return 2; + case SG_VERTEXFORMAT_HALF4: return 4; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_vertexformat_type(sg_vertex_format fmt) { + switch (fmt) { + case SG_VERTEXFORMAT_FLOAT: + case SG_VERTEXFORMAT_FLOAT2: + case SG_VERTEXFORMAT_FLOAT3: + case SG_VERTEXFORMAT_FLOAT4: + return GL_FLOAT; + case SG_VERTEXFORMAT_INT: + case SG_VERTEXFORMAT_INT2: + case SG_VERTEXFORMAT_INT3: + case SG_VERTEXFORMAT_INT4: + return GL_INT; + case SG_VERTEXFORMAT_UINT: + case SG_VERTEXFORMAT_UINT2: + case SG_VERTEXFORMAT_UINT3: + case SG_VERTEXFORMAT_UINT4: + return GL_UNSIGNED_INT; + case SG_VERTEXFORMAT_BYTE4: + case SG_VERTEXFORMAT_BYTE4N: + return GL_BYTE; + case SG_VERTEXFORMAT_UBYTE4: + case SG_VERTEXFORMAT_UBYTE4N: + return GL_UNSIGNED_BYTE; + case SG_VERTEXFORMAT_SHORT2: + case SG_VERTEXFORMAT_SHORT2N: + case SG_VERTEXFORMAT_SHORT4: + case SG_VERTEXFORMAT_SHORT4N: + return GL_SHORT; + case SG_VERTEXFORMAT_USHORT2: + case SG_VERTEXFORMAT_USHORT2N: + case SG_VERTEXFORMAT_USHORT4: + case SG_VERTEXFORMAT_USHORT4N: + return GL_UNSIGNED_SHORT; + case SG_VERTEXFORMAT_UINT10_N2: + return GL_UNSIGNED_INT_2_10_10_10_REV; + case SG_VERTEXFORMAT_HALF2: + case SG_VERTEXFORMAT_HALF4: + return GL_HALF_FLOAT; + default: + SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLboolean _sg_gl_vertexformat_normalized(sg_vertex_format fmt) { + switch (fmt) { + case SG_VERTEXFORMAT_BYTE4N: + case SG_VERTEXFORMAT_UBYTE4N: + case SG_VERTEXFORMAT_SHORT2N: + case SG_VERTEXFORMAT_USHORT2N: + case SG_VERTEXFORMAT_SHORT4N: + case SG_VERTEXFORMAT_USHORT4N: + case SG_VERTEXFORMAT_UINT10_N2: + return GL_TRUE; + default: + return GL_FALSE; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_primitive_type(sg_primitive_type t) { + switch (t) { + case SG_PRIMITIVETYPE_POINTS: return GL_POINTS; + case SG_PRIMITIVETYPE_LINES: return GL_LINES; + case SG_PRIMITIVETYPE_LINE_STRIP: return GL_LINE_STRIP; + case SG_PRIMITIVETYPE_TRIANGLES: return GL_TRIANGLES; + case SG_PRIMITIVETYPE_TRIANGLE_STRIP: return GL_TRIANGLE_STRIP; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_index_type(sg_index_type t) { + switch (t) { + case SG_INDEXTYPE_NONE: return 0; + case SG_INDEXTYPE_UINT16: return GL_UNSIGNED_SHORT; + case SG_INDEXTYPE_UINT32: return GL_UNSIGNED_INT; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_compare_func(sg_compare_func cmp) { + switch (cmp) { + case SG_COMPAREFUNC_NEVER: return GL_NEVER; + case SG_COMPAREFUNC_LESS: return GL_LESS; + case SG_COMPAREFUNC_EQUAL: return GL_EQUAL; + case SG_COMPAREFUNC_LESS_EQUAL: return GL_LEQUAL; + case SG_COMPAREFUNC_GREATER: return GL_GREATER; + case SG_COMPAREFUNC_NOT_EQUAL: return GL_NOTEQUAL; + case SG_COMPAREFUNC_GREATER_EQUAL: return GL_GEQUAL; + case SG_COMPAREFUNC_ALWAYS: return GL_ALWAYS; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_stencil_op(sg_stencil_op op) { + switch (op) { + case SG_STENCILOP_KEEP: return GL_KEEP; + case SG_STENCILOP_ZERO: return GL_ZERO; + case SG_STENCILOP_REPLACE: return GL_REPLACE; + case SG_STENCILOP_INCR_CLAMP: return GL_INCR; + case SG_STENCILOP_DECR_CLAMP: return GL_DECR; + case SG_STENCILOP_INVERT: return GL_INVERT; + case SG_STENCILOP_INCR_WRAP: return GL_INCR_WRAP; + case SG_STENCILOP_DECR_WRAP: return GL_DECR_WRAP; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_blend_factor(sg_blend_factor f) { + switch (f) { + case SG_BLENDFACTOR_ZERO: return GL_ZERO; + case SG_BLENDFACTOR_ONE: return GL_ONE; + case SG_BLENDFACTOR_SRC_COLOR: return GL_SRC_COLOR; + case SG_BLENDFACTOR_ONE_MINUS_SRC_COLOR: return GL_ONE_MINUS_SRC_COLOR; + case SG_BLENDFACTOR_SRC_ALPHA: return GL_SRC_ALPHA; + case SG_BLENDFACTOR_ONE_MINUS_SRC_ALPHA: return GL_ONE_MINUS_SRC_ALPHA; + case SG_BLENDFACTOR_DST_COLOR: return GL_DST_COLOR; + case SG_BLENDFACTOR_ONE_MINUS_DST_COLOR: return GL_ONE_MINUS_DST_COLOR; + case SG_BLENDFACTOR_DST_ALPHA: return GL_DST_ALPHA; + case SG_BLENDFACTOR_ONE_MINUS_DST_ALPHA: return GL_ONE_MINUS_DST_ALPHA; + case SG_BLENDFACTOR_SRC_ALPHA_SATURATED: return GL_SRC_ALPHA_SATURATE; + case SG_BLENDFACTOR_BLEND_COLOR: return GL_CONSTANT_COLOR; + case SG_BLENDFACTOR_ONE_MINUS_BLEND_COLOR: return GL_ONE_MINUS_CONSTANT_COLOR; + case SG_BLENDFACTOR_BLEND_ALPHA: return GL_CONSTANT_ALPHA; + case SG_BLENDFACTOR_ONE_MINUS_BLEND_ALPHA: return GL_ONE_MINUS_CONSTANT_ALPHA; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_blend_op(sg_blend_op op) { + switch (op) { + case SG_BLENDOP_ADD: return GL_FUNC_ADD; + case SG_BLENDOP_SUBTRACT: return GL_FUNC_SUBTRACT; + case SG_BLENDOP_REVERSE_SUBTRACT: return GL_FUNC_REVERSE_SUBTRACT; + case SG_BLENDOP_MIN: return GL_MIN; + case SG_BLENDOP_MAX: return GL_MAX; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_min_filter(sg_filter min_f, sg_filter mipmap_f) { + if (min_f == SG_FILTER_NEAREST) { + switch (mipmap_f) { + case SG_FILTER_NEAREST: return GL_NEAREST_MIPMAP_NEAREST; + case SG_FILTER_LINEAR: return GL_NEAREST_MIPMAP_LINEAR; + default: SOKOL_UNREACHABLE; return (GLenum)0; + } + } else if (min_f == SG_FILTER_LINEAR) { + switch (mipmap_f) { + case SG_FILTER_NEAREST: return GL_LINEAR_MIPMAP_NEAREST; + case SG_FILTER_LINEAR: return GL_LINEAR_MIPMAP_LINEAR; + default: SOKOL_UNREACHABLE; return (GLenum)0; + } + } else { + SOKOL_UNREACHABLE; return (GLenum)0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_mag_filter(sg_filter mag_f) { + if (mag_f == SG_FILTER_NEAREST) { + return GL_NEAREST; + } else { + return GL_LINEAR; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_wrap(sg_wrap w) { + switch (w) { + case SG_WRAP_CLAMP_TO_EDGE: return GL_CLAMP_TO_EDGE; + #if defined(SOKOL_GLCORE) + case SG_WRAP_CLAMP_TO_BORDER: return GL_CLAMP_TO_BORDER; + #else + case SG_WRAP_CLAMP_TO_BORDER: return GL_CLAMP_TO_EDGE; + #endif + case SG_WRAP_REPEAT: return GL_REPEAT; + case SG_WRAP_MIRRORED_REPEAT: return GL_MIRRORED_REPEAT; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_teximage_type(sg_pixel_format fmt) { + switch (fmt) { + case SG_PIXELFORMAT_R8: + case SG_PIXELFORMAT_R8UI: + case SG_PIXELFORMAT_RG8: + case SG_PIXELFORMAT_RG8UI: + case SG_PIXELFORMAT_RGBA8: + case SG_PIXELFORMAT_SRGB8A8: + case SG_PIXELFORMAT_RGBA8UI: + case SG_PIXELFORMAT_BGRA8: + return GL_UNSIGNED_BYTE; + case SG_PIXELFORMAT_R8SN: + case SG_PIXELFORMAT_R8SI: + case SG_PIXELFORMAT_RG8SN: + case SG_PIXELFORMAT_RG8SI: + case SG_PIXELFORMAT_RGBA8SN: + case SG_PIXELFORMAT_RGBA8SI: + return GL_BYTE; + case SG_PIXELFORMAT_R16: + case SG_PIXELFORMAT_R16UI: + case SG_PIXELFORMAT_RG16: + case SG_PIXELFORMAT_RG16UI: + case SG_PIXELFORMAT_RGBA16: + case SG_PIXELFORMAT_RGBA16UI: + return GL_UNSIGNED_SHORT; + case SG_PIXELFORMAT_R16SN: + case SG_PIXELFORMAT_R16SI: + case SG_PIXELFORMAT_RG16SN: + case SG_PIXELFORMAT_RG16SI: + case SG_PIXELFORMAT_RGBA16SN: + case SG_PIXELFORMAT_RGBA16SI: + return GL_SHORT; + case SG_PIXELFORMAT_R16F: + case SG_PIXELFORMAT_RG16F: + case SG_PIXELFORMAT_RGBA16F: + return GL_HALF_FLOAT; + case SG_PIXELFORMAT_R32UI: + case SG_PIXELFORMAT_RG32UI: + case SG_PIXELFORMAT_RGBA32UI: + return GL_UNSIGNED_INT; + case SG_PIXELFORMAT_R32SI: + case SG_PIXELFORMAT_RG32SI: + case SG_PIXELFORMAT_RGBA32SI: + return GL_INT; + case SG_PIXELFORMAT_R32F: + case SG_PIXELFORMAT_RG32F: + case SG_PIXELFORMAT_RGBA32F: + return GL_FLOAT; + case SG_PIXELFORMAT_RGB10A2: + return GL_UNSIGNED_INT_2_10_10_10_REV; + case SG_PIXELFORMAT_RG11B10F: + return GL_UNSIGNED_INT_10F_11F_11F_REV; + case SG_PIXELFORMAT_RGB9E5: + return GL_UNSIGNED_INT_5_9_9_9_REV; + case SG_PIXELFORMAT_DEPTH: + return GL_FLOAT; + case SG_PIXELFORMAT_DEPTH_STENCIL: + return GL_UNSIGNED_INT_24_8; + default: + SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_teximage_format(sg_pixel_format fmt) { + switch (fmt) { + case SG_PIXELFORMAT_R8: + case SG_PIXELFORMAT_R8SN: + case SG_PIXELFORMAT_R16: + case SG_PIXELFORMAT_R16SN: + case SG_PIXELFORMAT_R16F: + case SG_PIXELFORMAT_R32F: + return GL_RED; + case SG_PIXELFORMAT_R8UI: + case SG_PIXELFORMAT_R8SI: + case SG_PIXELFORMAT_R16UI: + case SG_PIXELFORMAT_R16SI: + case SG_PIXELFORMAT_R32UI: + case SG_PIXELFORMAT_R32SI: + return GL_RED_INTEGER; + case SG_PIXELFORMAT_RG8: + case SG_PIXELFORMAT_RG8SN: + case SG_PIXELFORMAT_RG16: + case SG_PIXELFORMAT_RG16SN: + case SG_PIXELFORMAT_RG16F: + case SG_PIXELFORMAT_RG32F: + return GL_RG; + case SG_PIXELFORMAT_RG8UI: + case SG_PIXELFORMAT_RG8SI: + case SG_PIXELFORMAT_RG16UI: + case SG_PIXELFORMAT_RG16SI: + case SG_PIXELFORMAT_RG32UI: + case SG_PIXELFORMAT_RG32SI: + return GL_RG_INTEGER; + case SG_PIXELFORMAT_RGBA8: + case SG_PIXELFORMAT_SRGB8A8: + case SG_PIXELFORMAT_RGBA8SN: + case SG_PIXELFORMAT_RGBA16: + case SG_PIXELFORMAT_RGBA16SN: + case SG_PIXELFORMAT_RGBA16F: + case SG_PIXELFORMAT_RGBA32F: + case SG_PIXELFORMAT_RGB10A2: + return GL_RGBA; + case SG_PIXELFORMAT_RGBA8UI: + case SG_PIXELFORMAT_RGBA8SI: + case SG_PIXELFORMAT_RGBA16UI: + case SG_PIXELFORMAT_RGBA16SI: + case SG_PIXELFORMAT_RGBA32UI: + case SG_PIXELFORMAT_RGBA32SI: + return GL_RGBA_INTEGER; + case SG_PIXELFORMAT_RG11B10F: + case SG_PIXELFORMAT_RGB9E5: + return GL_RGB; + case SG_PIXELFORMAT_DEPTH: + return GL_DEPTH_COMPONENT; + case SG_PIXELFORMAT_DEPTH_STENCIL: + return GL_DEPTH_STENCIL; + case SG_PIXELFORMAT_BC1_RGBA: + return GL_COMPRESSED_RGBA_S3TC_DXT1_EXT; + case SG_PIXELFORMAT_BC2_RGBA: + return GL_COMPRESSED_RGBA_S3TC_DXT3_EXT; + case SG_PIXELFORMAT_BC3_RGBA: + return GL_COMPRESSED_RGBA_S3TC_DXT5_EXT; + case SG_PIXELFORMAT_BC3_SRGBA: + return GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT; + case SG_PIXELFORMAT_BC4_R: + return GL_COMPRESSED_RED_RGTC1; + case SG_PIXELFORMAT_BC4_RSN: + return GL_COMPRESSED_SIGNED_RED_RGTC1; + case SG_PIXELFORMAT_BC5_RG: + return GL_COMPRESSED_RED_GREEN_RGTC2; + case SG_PIXELFORMAT_BC5_RGSN: + return GL_COMPRESSED_SIGNED_RED_GREEN_RGTC2; + case SG_PIXELFORMAT_BC6H_RGBF: + return GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT_ARB; + case SG_PIXELFORMAT_BC6H_RGBUF: + return GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT_ARB; + case SG_PIXELFORMAT_BC7_RGBA: + return GL_COMPRESSED_RGBA_BPTC_UNORM_ARB; + case SG_PIXELFORMAT_BC7_SRGBA: + return GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM_ARB; + case SG_PIXELFORMAT_ETC2_RGB8: + return GL_COMPRESSED_RGB8_ETC2; + case SG_PIXELFORMAT_ETC2_SRGB8: + return GL_COMPRESSED_SRGB8_ETC2; + case SG_PIXELFORMAT_ETC2_RGB8A1: + return GL_COMPRESSED_RGB8_PUNCHTHROUGH_ALPHA1_ETC2; + case SG_PIXELFORMAT_ETC2_RGBA8: + return GL_COMPRESSED_RGBA8_ETC2_EAC; + case SG_PIXELFORMAT_ETC2_SRGB8A8: + return GL_COMPRESSED_SRGB8_ALPHA8_ETC2_EAC; + case SG_PIXELFORMAT_EAC_R11: + return GL_COMPRESSED_R11_EAC; + case SG_PIXELFORMAT_EAC_R11SN: + return GL_COMPRESSED_SIGNED_R11_EAC; + case SG_PIXELFORMAT_EAC_RG11: + return GL_COMPRESSED_RG11_EAC; + case SG_PIXELFORMAT_EAC_RG11SN: + return GL_COMPRESSED_SIGNED_RG11_EAC; + case SG_PIXELFORMAT_ASTC_4x4_RGBA: + return GL_COMPRESSED_RGBA_ASTC_4x4_KHR; + case SG_PIXELFORMAT_ASTC_4x4_SRGBA: + return GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR; + default: + SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_teximage_internal_format(sg_pixel_format fmt) { + switch (fmt) { + case SG_PIXELFORMAT_R8: return GL_R8; + case SG_PIXELFORMAT_R8SN: return GL_R8_SNORM; + case SG_PIXELFORMAT_R8UI: return GL_R8UI; + case SG_PIXELFORMAT_R8SI: return GL_R8I; + #if !defined(SOKOL_GLES3) + case SG_PIXELFORMAT_R16: return GL_R16; + case SG_PIXELFORMAT_R16SN: return GL_R16_SNORM; + #endif + case SG_PIXELFORMAT_R16UI: return GL_R16UI; + case SG_PIXELFORMAT_R16SI: return GL_R16I; + case SG_PIXELFORMAT_R16F: return GL_R16F; + case SG_PIXELFORMAT_RG8: return GL_RG8; + case SG_PIXELFORMAT_RG8SN: return GL_RG8_SNORM; + case SG_PIXELFORMAT_RG8UI: return GL_RG8UI; + case SG_PIXELFORMAT_RG8SI: return GL_RG8I; + case SG_PIXELFORMAT_R32UI: return GL_R32UI; + case SG_PIXELFORMAT_R32SI: return GL_R32I; + case SG_PIXELFORMAT_R32F: return GL_R32F; + #if !defined(SOKOL_GLES3) + case SG_PIXELFORMAT_RG16: return GL_RG16; + case SG_PIXELFORMAT_RG16SN: return GL_RG16_SNORM; + #endif + case SG_PIXELFORMAT_RG16UI: return GL_RG16UI; + case SG_PIXELFORMAT_RG16SI: return GL_RG16I; + case SG_PIXELFORMAT_RG16F: return GL_RG16F; + case SG_PIXELFORMAT_RGBA8: return GL_RGBA8; + case SG_PIXELFORMAT_SRGB8A8: return GL_SRGB8_ALPHA8; + case SG_PIXELFORMAT_RGBA8SN: return GL_RGBA8_SNORM; + case SG_PIXELFORMAT_RGBA8UI: return GL_RGBA8UI; + case SG_PIXELFORMAT_RGBA8SI: return GL_RGBA8I; + case SG_PIXELFORMAT_RGB10A2: return GL_RGB10_A2; + case SG_PIXELFORMAT_RG11B10F: return GL_R11F_G11F_B10F; + case SG_PIXELFORMAT_RGB9E5: return GL_RGB9_E5; + case SG_PIXELFORMAT_RG32UI: return GL_RG32UI; + case SG_PIXELFORMAT_RG32SI: return GL_RG32I; + case SG_PIXELFORMAT_RG32F: return GL_RG32F; + #if !defined(SOKOL_GLES3) + case SG_PIXELFORMAT_RGBA16: return GL_RGBA16; + case SG_PIXELFORMAT_RGBA16SN: return GL_RGBA16_SNORM; + #endif + case SG_PIXELFORMAT_RGBA16UI: return GL_RGBA16UI; + case SG_PIXELFORMAT_RGBA16SI: return GL_RGBA16I; + case SG_PIXELFORMAT_RGBA16F: return GL_RGBA16F; + case SG_PIXELFORMAT_RGBA32UI: return GL_RGBA32UI; + case SG_PIXELFORMAT_RGBA32SI: return GL_RGBA32I; + case SG_PIXELFORMAT_RGBA32F: return GL_RGBA32F; + case SG_PIXELFORMAT_DEPTH: return GL_DEPTH_COMPONENT32F; + case SG_PIXELFORMAT_DEPTH_STENCIL: return GL_DEPTH24_STENCIL8; + case SG_PIXELFORMAT_BC1_RGBA: return GL_COMPRESSED_RGBA_S3TC_DXT1_EXT; + case SG_PIXELFORMAT_BC2_RGBA: return GL_COMPRESSED_RGBA_S3TC_DXT3_EXT; + case SG_PIXELFORMAT_BC3_RGBA: return GL_COMPRESSED_RGBA_S3TC_DXT5_EXT; + case SG_PIXELFORMAT_BC3_SRGBA: return GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT; + case SG_PIXELFORMAT_BC4_R: return GL_COMPRESSED_RED_RGTC1; + case SG_PIXELFORMAT_BC4_RSN: return GL_COMPRESSED_SIGNED_RED_RGTC1; + case SG_PIXELFORMAT_BC5_RG: return GL_COMPRESSED_RED_GREEN_RGTC2; + case SG_PIXELFORMAT_BC5_RGSN: return GL_COMPRESSED_SIGNED_RED_GREEN_RGTC2; + case SG_PIXELFORMAT_BC6H_RGBF: return GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT_ARB; + case SG_PIXELFORMAT_BC6H_RGBUF: return GL_COMPRESSED_RGB_BPTC_UNSIGNED_FLOAT_ARB; + case SG_PIXELFORMAT_BC7_RGBA: return GL_COMPRESSED_RGBA_BPTC_UNORM_ARB; + case SG_PIXELFORMAT_BC7_SRGBA: return GL_COMPRESSED_SRGB_ALPHA_BPTC_UNORM_ARB; + case SG_PIXELFORMAT_ETC2_RGB8: return GL_COMPRESSED_RGB8_ETC2; + case SG_PIXELFORMAT_ETC2_SRGB8: return GL_COMPRESSED_SRGB8_ETC2; + case SG_PIXELFORMAT_ETC2_RGB8A1: return GL_COMPRESSED_RGB8_PUNCHTHROUGH_ALPHA1_ETC2; + case SG_PIXELFORMAT_ETC2_RGBA8: return GL_COMPRESSED_RGBA8_ETC2_EAC; + case SG_PIXELFORMAT_ETC2_SRGB8A8: return GL_COMPRESSED_SRGB8_ALPHA8_ETC2_EAC; + case SG_PIXELFORMAT_EAC_R11: return GL_COMPRESSED_R11_EAC; + case SG_PIXELFORMAT_EAC_R11SN: return GL_COMPRESSED_SIGNED_R11_EAC; + case SG_PIXELFORMAT_EAC_RG11: return GL_COMPRESSED_RG11_EAC; + case SG_PIXELFORMAT_EAC_RG11SN: return GL_COMPRESSED_SIGNED_RG11_EAC; + case SG_PIXELFORMAT_ASTC_4x4_RGBA: return GL_COMPRESSED_RGBA_ASTC_4x4_KHR; + case SG_PIXELFORMAT_ASTC_4x4_SRGBA: return GL_COMPRESSED_SRGB8_ALPHA8_ASTC_4x4_KHR; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_cubeface_target(int face_index) { + switch (face_index) { + case 0: return GL_TEXTURE_CUBE_MAP_POSITIVE_X; + case 1: return GL_TEXTURE_CUBE_MAP_NEGATIVE_X; + case 2: return GL_TEXTURE_CUBE_MAP_POSITIVE_Y; + case 3: return GL_TEXTURE_CUBE_MAP_NEGATIVE_Y; + case 4: return GL_TEXTURE_CUBE_MAP_POSITIVE_Z; + case 5: return GL_TEXTURE_CUBE_MAP_NEGATIVE_Z; + default: SOKOL_UNREACHABLE; return 0; + } +} + +// see: https://www.khronos.org/registry/OpenGL-Refpages/es3.0/html/glTexImage2D.xhtml +_SOKOL_PRIVATE void _sg_gl_init_pixelformats(bool has_bgra) { + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_R8SN]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_R8UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_R8SI]); + #if !defined(SOKOL_GLES3) + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R16]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R16SN]); + #endif + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_R16UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_R16SI]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RG8SN]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG8UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG8SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R32UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R32SI]); + #if !defined(SOKOL_GLES3) + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG16]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG16SN]); + #endif + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG16UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG16SI]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_SRGB8A8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RGBA8SN]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA8UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA8SI]); + if (has_bgra) { + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_BGRA8]); + } + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGB10A2]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RGB9E5]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG32UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG32SI]); + #if !defined(SOKOL_GLES3) + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA16]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA16SN]); + #endif + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA16UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA16SI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA32UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA32SI]); + _sg_pixelformat_srmd(&_sg.formats[SG_PIXELFORMAT_DEPTH]); + _sg_pixelformat_srmd(&_sg.formats[SG_PIXELFORMAT_DEPTH_STENCIL]); +} + +// FIXME: OES_half_float_blend +_SOKOL_PRIVATE void _sg_gl_init_pixelformats_half_float(bool has_colorbuffer_half_float) { + if (has_colorbuffer_half_float) { + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R16F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG16F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA16F]); + } else { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_R16F]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RG16F]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RGBA16F]); + } +} + +_SOKOL_PRIVATE void _sg_gl_init_pixelformats_float(bool has_colorbuffer_float, bool has_texture_float_linear, bool has_float_blend) { + if (has_texture_float_linear) { + if (has_colorbuffer_float) { + if (has_float_blend) { + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R32F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG32F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + } else { + _sg_pixelformat_sfrm(&_sg.formats[SG_PIXELFORMAT_R32F]); + _sg_pixelformat_sfrm(&_sg.formats[SG_PIXELFORMAT_RG32F]); + _sg_pixelformat_sfrm(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + } + _sg_pixelformat_sfrm(&_sg.formats[SG_PIXELFORMAT_RG11B10F]); + } else { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_R32F]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RG32F]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RG11B10F]); + } + } else { + if (has_colorbuffer_float) { + _sg_pixelformat_sbrm(&_sg.formats[SG_PIXELFORMAT_R32F]); + _sg_pixelformat_sbrm(&_sg.formats[SG_PIXELFORMAT_RG32F]); + _sg_pixelformat_sbrm(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG11B10F]); + } else { + _sg_pixelformat_s(&_sg.formats[SG_PIXELFORMAT_R32F]); + _sg_pixelformat_s(&_sg.formats[SG_PIXELFORMAT_RG32F]); + _sg_pixelformat_s(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + _sg_pixelformat_s(&_sg.formats[SG_PIXELFORMAT_RG11B10F]); + } + } +} + +_SOKOL_PRIVATE void _sg_gl_init_pixelformats_s3tc(void) { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC1_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC2_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC3_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC3_SRGBA]); +} + +_SOKOL_PRIVATE void _sg_gl_init_pixelformats_rgtc(void) { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC4_R]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC4_RSN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC5_RG]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC5_RGSN]); +} + +_SOKOL_PRIVATE void _sg_gl_init_pixelformats_bptc(void) { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC6H_RGBF]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC6H_RGBUF]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC7_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC7_SRGBA]); +} + +_SOKOL_PRIVATE void _sg_gl_init_pixelformats_etc2(void) { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGB8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_SRGB8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGB8A1]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGBA8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_SRGB8A8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_R11]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_R11SN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_RG11]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_RG11SN]); +} + +_SOKOL_PRIVATE void _sg_gl_init_pixelformats_astc(void) { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ASTC_4x4_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ASTC_4x4_SRGBA]); + } + +_SOKOL_PRIVATE void _sg_gl_init_limits(void) { + _SG_GL_CHECK_ERROR(); + GLint gl_int; + glGetIntegerv(GL_MAX_TEXTURE_SIZE, &gl_int); + _SG_GL_CHECK_ERROR(); + _sg.limits.max_image_size_2d = gl_int; + _sg.limits.max_image_size_array = gl_int; + glGetIntegerv(GL_MAX_CUBE_MAP_TEXTURE_SIZE, &gl_int); + _SG_GL_CHECK_ERROR(); + _sg.limits.max_image_size_cube = gl_int; + glGetIntegerv(GL_MAX_VERTEX_ATTRIBS, &gl_int); + _SG_GL_CHECK_ERROR(); + if (gl_int > SG_MAX_VERTEX_ATTRIBUTES) { + gl_int = SG_MAX_VERTEX_ATTRIBUTES; + } + _sg.limits.max_vertex_attrs = gl_int; + glGetIntegerv(GL_MAX_VERTEX_UNIFORM_COMPONENTS, &gl_int); + _SG_GL_CHECK_ERROR(); + _sg.limits.gl_max_vertex_uniform_components = gl_int; + glGetIntegerv(GL_MAX_3D_TEXTURE_SIZE, &gl_int); + _SG_GL_CHECK_ERROR(); + _sg.limits.max_image_size_3d = gl_int; + glGetIntegerv(GL_MAX_ARRAY_TEXTURE_LAYERS, &gl_int); + _SG_GL_CHECK_ERROR(); + _sg.limits.max_image_array_layers = gl_int; + if (_sg.gl.ext_anisotropic) { + glGetIntegerv(GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT, &gl_int); + _SG_GL_CHECK_ERROR(); + _sg.gl.max_anisotropy = gl_int; + } else { + _sg.gl.max_anisotropy = 1; + } + glGetIntegerv(GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS, &gl_int); + _SG_GL_CHECK_ERROR(); + _sg.limits.gl_max_combined_texture_image_units = gl_int; +} + +#if defined(SOKOL_GLCORE) +_SOKOL_PRIVATE void _sg_gl_init_caps_glcore(void) { + _sg.backend = SG_BACKEND_GLCORE; + + GLint major_version = 0; + GLint minor_version = 0; + glGetIntegerv(GL_MAJOR_VERSION, &major_version); + glGetIntegerv(GL_MINOR_VERSION, &minor_version); + const int version = major_version * 100 + minor_version * 10; + _sg.features.origin_top_left = false; + _sg.features.image_clamp_to_border = true; + _sg.features.mrt_independent_blend_state = false; + _sg.features.mrt_independent_write_mask = true; + _sg.features.compute = version >= 430; + #if defined(__APPLE__) + _sg.features.msaa_image_bindings = false; + #else + _sg.features.msaa_image_bindings = true; + #endif + + // scan extensions + bool has_s3tc = false; // BC1..BC3 + bool has_rgtc = false; // BC4 and BC5 + bool has_bptc = false; // BC6H and BC7 + bool has_etc2 = false; + bool has_astc = false; + GLint num_ext = 0; + glGetIntegerv(GL_NUM_EXTENSIONS, &num_ext); + for (int i = 0; i < num_ext; i++) { + const char* ext = (const char*) glGetStringi(GL_EXTENSIONS, (GLuint)i); + if (ext) { + if (strstr(ext, "_texture_compression_s3tc")) { + has_s3tc = true; + } else if (strstr(ext, "_texture_compression_rgtc")) { + has_rgtc = true; + } else if (strstr(ext, "_texture_compression_bptc")) { + has_bptc = true; + } else if (strstr(ext, "_ES3_compatibility")) { + has_etc2 = true; + } else if (strstr(ext, "_texture_filter_anisotropic")) { + _sg.gl.ext_anisotropic = true; + } else if (strstr(ext, "_texture_compression_astc_ldr")) { + has_astc = true; + } + } + } + + // limits + _sg_gl_init_limits(); + + // pixel formats + const bool has_bgra = false; // not a bug + const bool has_colorbuffer_float = true; + const bool has_colorbuffer_half_float = true; + const bool has_texture_float_linear = true; // FIXME??? + const bool has_float_blend = true; + _sg_gl_init_pixelformats(has_bgra); + _sg_gl_init_pixelformats_float(has_colorbuffer_float, has_texture_float_linear, has_float_blend); + _sg_gl_init_pixelformats_half_float(has_colorbuffer_half_float); + if (has_s3tc) { + _sg_gl_init_pixelformats_s3tc(); + } + if (has_rgtc) { + _sg_gl_init_pixelformats_rgtc(); + } + if (has_bptc) { + _sg_gl_init_pixelformats_bptc(); + } + if (has_etc2) { + _sg_gl_init_pixelformats_etc2(); + } + if (has_astc) { + _sg_gl_init_pixelformats_astc(); + } +} +#endif + +#if defined(SOKOL_GLES3) +_SOKOL_PRIVATE void _sg_gl_init_caps_gles3(void) { + _sg.backend = SG_BACKEND_GLES3; + + GLint major_version = 0; + GLint minor_version = 0; + glGetIntegerv(GL_MAJOR_VERSION, &major_version); + glGetIntegerv(GL_MINOR_VERSION, &minor_version); + const int version = major_version * 100 + minor_version * 10; + _sg.features.origin_top_left = false; + _sg.features.image_clamp_to_border = false; + _sg.features.mrt_independent_blend_state = false; + _sg.features.mrt_independent_write_mask = false; + _sg.features.compute = version >= 310; + _sg.features.msaa_image_bindings = false; + + bool has_s3tc = false; // BC1..BC3 + bool has_rgtc = false; // BC4 and BC5 + bool has_bptc = false; // BC6H and BC7 + #if defined(__EMSCRIPTEN__) + bool has_etc2 = false; + #else + bool has_etc2 = true; + #endif + bool has_astc = false; + bool has_colorbuffer_float = false; + bool has_colorbuffer_half_float = false; + bool has_texture_float_linear = false; + bool has_float_blend = false; + GLint num_ext = 0; + glGetIntegerv(GL_NUM_EXTENSIONS, &num_ext); + for (int i = 0; i < num_ext; i++) { + const char* ext = (const char*) glGetStringi(GL_EXTENSIONS, (GLuint)i); + if (ext) { + if (strstr(ext, "_texture_compression_s3tc")) { + has_s3tc = true; + } else if (strstr(ext, "_compressed_texture_s3tc")) { + has_s3tc = true; + } else if (strstr(ext, "_texture_compression_rgtc")) { + has_rgtc = true; + } else if (strstr(ext, "_texture_compression_bptc")) { + has_bptc = true; + } else if (strstr(ext, "_compressed_texture_etc")) { + has_etc2 = true; + } else if (strstr(ext, "_compressed_texture_astc")) { + has_astc = true; + } else if (strstr(ext, "_color_buffer_float")) { + has_colorbuffer_float = true; + } else if (strstr(ext, "_color_buffer_half_float")) { + has_colorbuffer_half_float = true; + } else if (strstr(ext, "_texture_float_linear")) { + has_texture_float_linear = true; + } else if (strstr(ext, "_float_blend")) { + has_float_blend = true; + } else if (strstr(ext, "_texture_filter_anisotropic")) { + _sg.gl.ext_anisotropic = true; + } + } + } + + /* on WebGL2, color_buffer_float also includes 16-bit formats + see: https://developer.mozilla.org/en-US/docs/Web/API/EXT_color_buffer_float + */ + #if defined(__EMSCRIPTEN__) + if (!has_colorbuffer_half_float && has_colorbuffer_float) { + has_colorbuffer_half_float = has_colorbuffer_float; + } + #endif + + // limits + _sg_gl_init_limits(); + + // pixel formats + const bool has_bgra = false; // not a bug + _sg_gl_init_pixelformats(has_bgra); + _sg_gl_init_pixelformats_float(has_colorbuffer_float, has_texture_float_linear, has_float_blend); + _sg_gl_init_pixelformats_half_float(has_colorbuffer_half_float); + if (has_s3tc) { + _sg_gl_init_pixelformats_s3tc(); + } + if (has_rgtc) { + _sg_gl_init_pixelformats_rgtc(); + } + if (has_bptc) { + _sg_gl_init_pixelformats_bptc(); + } + if (has_etc2) { + _sg_gl_init_pixelformats_etc2(); + } + if (has_astc) { + _sg_gl_init_pixelformats_astc(); + } +} +#endif + +//-- state cache implementation ------------------------------------------------ +_SOKOL_PRIVATE void _sg_gl_cache_clear_buffer_bindings(bool force) { + if (force || (_sg.gl.cache.vertex_buffer != 0)) { + glBindBuffer(GL_ARRAY_BUFFER, 0); + _sg.gl.cache.vertex_buffer = 0; + _sg_stats_add(gl.num_bind_buffer, 1); + } + if (force || (_sg.gl.cache.index_buffer != 0)) { + glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0); + _sg.gl.cache.index_buffer = 0; + _sg_stats_add(gl.num_bind_buffer, 1); + } + if (force || (_sg.gl.cache.storage_buffer != 0)) { + if (_sg.features.compute) { + glBindBuffer(GL_SHADER_STORAGE_BUFFER, 0); + } + _sg.gl.cache.storage_buffer = 0; + _sg_stats_add(gl.num_bind_buffer, 1); + } + for (size_t i = 0; i < _SG_GL_MAX_SBUF_BINDINGS; i++) { + if (force || (_sg.gl.cache.storage_buffers[i] != 0)) { + if (_sg.features.compute) { + glBindBufferBase(GL_SHADER_STORAGE_BUFFER, (GLuint)i, 0); + } + _sg.gl.cache.storage_buffers[i] = 0; + _sg_stats_add(gl.num_bind_buffer, 1); + } + } +} + +_SOKOL_PRIVATE void _sg_gl_cache_bind_buffer(GLenum target, GLuint buffer) { + SOKOL_ASSERT((GL_ARRAY_BUFFER == target) || (GL_ELEMENT_ARRAY_BUFFER == target) || (GL_SHADER_STORAGE_BUFFER == target)); + if (target == GL_ARRAY_BUFFER) { + if (_sg.gl.cache.vertex_buffer != buffer) { + _sg.gl.cache.vertex_buffer = buffer; + glBindBuffer(target, buffer); + _sg_stats_add(gl.num_bind_buffer, 1); + } + } else if (target == GL_ELEMENT_ARRAY_BUFFER) { + if (_sg.gl.cache.index_buffer != buffer) { + _sg.gl.cache.index_buffer = buffer; + glBindBuffer(target, buffer); + _sg_stats_add(gl.num_bind_buffer, 1); + } + } else if (target == GL_SHADER_STORAGE_BUFFER) { + if (_sg.gl.cache.storage_buffer != buffer) { + _sg.gl.cache.storage_buffer = buffer; + if (_sg.features.compute) { + glBindBuffer(target, buffer); + } + _sg_stats_add(gl.num_bind_buffer, 1); + } + } else { + SOKOL_UNREACHABLE; + } +} + +_SOKOL_PRIVATE void _sg_gl_cache_bind_storage_buffer(uint8_t glsl_binding_n, GLuint buffer) { + SOKOL_ASSERT(glsl_binding_n < _SG_GL_MAX_SBUF_BINDINGS); + if (_sg.gl.cache.storage_buffers[glsl_binding_n] != buffer) { + _sg.gl.cache.storage_buffers[glsl_binding_n] = buffer; + _sg.gl.cache.storage_buffer = buffer; // not a bug + if (_sg.features.compute) { + glBindBufferBase(GL_SHADER_STORAGE_BUFFER, glsl_binding_n, buffer); + } + _sg_stats_add(gl.num_bind_buffer, 1); + } +} + +_SOKOL_PRIVATE void _sg_gl_cache_store_buffer_binding(GLenum target) { + if (target == GL_ARRAY_BUFFER) { + _sg.gl.cache.stored_vertex_buffer = _sg.gl.cache.vertex_buffer; + } else if (target == GL_ELEMENT_ARRAY_BUFFER) { + _sg.gl.cache.stored_index_buffer = _sg.gl.cache.index_buffer; + } else if (target == GL_SHADER_STORAGE_BUFFER) { + _sg.gl.cache.stored_storage_buffer = _sg.gl.cache.storage_buffer; + } else { + SOKOL_UNREACHABLE; + } +} + +_SOKOL_PRIVATE void _sg_gl_cache_restore_buffer_binding(GLenum target) { + if (target == GL_ARRAY_BUFFER) { + if (_sg.gl.cache.stored_vertex_buffer != 0) { + // we only care about restoring valid ids + _sg_gl_cache_bind_buffer(target, _sg.gl.cache.stored_vertex_buffer); + _sg.gl.cache.stored_vertex_buffer = 0; + } + } else if (target == GL_ELEMENT_ARRAY_BUFFER) { + if (_sg.gl.cache.stored_index_buffer != 0) { + // we only care about restoring valid ids + _sg_gl_cache_bind_buffer(target, _sg.gl.cache.stored_index_buffer); + _sg.gl.cache.stored_index_buffer = 0; + } + } else if (target == GL_SHADER_STORAGE_BUFFER) { + if (_sg.gl.cache.stored_storage_buffer != 0) { + // we only care about restoring valid ids + _sg_gl_cache_bind_buffer(target, _sg.gl.cache.stored_storage_buffer); + _sg.gl.cache.stored_storage_buffer = 0; + } + } else { + SOKOL_UNREACHABLE; + } +} + +// called from _sg_gl_discard_buffer() +_SOKOL_PRIVATE void _sg_gl_cache_invalidate_buffer(GLuint buf) { + if (buf == _sg.gl.cache.vertex_buffer) { + _sg.gl.cache.vertex_buffer = 0; + glBindBuffer(GL_ARRAY_BUFFER, 0); + _sg_stats_add(gl.num_bind_buffer, 1); + } + if (buf == _sg.gl.cache.index_buffer) { + _sg.gl.cache.index_buffer = 0; + glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0); + _sg_stats_add(gl.num_bind_buffer, 1); + } + if (buf == _sg.gl.cache.storage_buffer) { + _sg.gl.cache.storage_buffer = 0; + glBindBuffer(GL_SHADER_STORAGE_BUFFER, 0); + _sg_stats_add(gl.num_bind_buffer, 1); + } + for (size_t i = 0; i < _SG_GL_MAX_SBUF_BINDINGS; i++) { + if (buf == _sg.gl.cache.storage_buffers[i]) { + _sg.gl.cache.storage_buffers[i] = 0; + _sg.gl.cache.storage_buffer = 0; // not a bug! + glBindBufferBase(GL_SHADER_STORAGE_BUFFER, (GLuint)i, 0); + _sg_stats_add(gl.num_bind_buffer, 1); + } + } + if (buf == _sg.gl.cache.stored_vertex_buffer) { + _sg.gl.cache.stored_vertex_buffer = 0; + } + if (buf == _sg.gl.cache.stored_index_buffer) { + _sg.gl.cache.stored_index_buffer = 0; + } + if (buf == _sg.gl.cache.stored_storage_buffer) { + _sg.gl.cache.stored_storage_buffer = 0; + } + for (int i = 0; i < SG_MAX_VERTEX_ATTRIBUTES; i++) { + if (buf == _sg.gl.cache.attrs[i].gl_vbuf) { + _sg.gl.cache.attrs[i].gl_vbuf = 0; + } + } +} + +_SOKOL_PRIVATE void _sg_gl_cache_active_texture(GLenum texture) { + _SG_GL_CHECK_ERROR(); + if (_sg.gl.cache.cur_active_texture != texture) { + _sg.gl.cache.cur_active_texture = texture; + glActiveTexture(texture); + _sg_stats_add(gl.num_active_texture, 1); + } + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE void _sg_gl_cache_clear_texture_sampler_bindings(bool force) { + _SG_GL_CHECK_ERROR(); + for (int i = 0; (i < _SG_GL_MAX_IMG_SMP_BINDINGS) && (i < _sg.limits.gl_max_combined_texture_image_units); i++) { + if (force || (_sg.gl.cache.texture_samplers[i].texture != 0)) { + GLenum gl_texture_unit = (GLenum) (GL_TEXTURE0 + i); + glActiveTexture(gl_texture_unit); + _sg_stats_add(gl.num_active_texture, 1); + glBindTexture(GL_TEXTURE_2D, 0); + glBindTexture(GL_TEXTURE_CUBE_MAP, 0); + glBindTexture(GL_TEXTURE_3D, 0); + glBindTexture(GL_TEXTURE_2D_ARRAY, 0); + _sg_stats_add(gl.num_bind_texture, 4); + glBindSampler((GLuint)i, 0); + _sg_stats_add(gl.num_bind_sampler, 1); + _sg.gl.cache.texture_samplers[i].target = 0; + _sg.gl.cache.texture_samplers[i].texture = 0; + _sg.gl.cache.texture_samplers[i].sampler = 0; + _sg.gl.cache.cur_active_texture = gl_texture_unit; + } + } + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE void _sg_gl_cache_bind_texture_sampler(int8_t gl_tex_slot, GLenum target, GLuint texture, GLuint sampler) { + /* it's valid to call this function with target=0 and/or texture=0 + target=0 will unbind the previous binding, texture=0 will clear + the new binding + */ + SOKOL_ASSERT((gl_tex_slot >= 0) && (gl_tex_slot < _SG_GL_MAX_IMG_SMP_BINDINGS)); + if (gl_tex_slot >= _sg.limits.gl_max_combined_texture_image_units) { + return; + } + _SG_GL_CHECK_ERROR(); + _sg_gl_cache_texture_sampler_bind_slot* slot = &_sg.gl.cache.texture_samplers[gl_tex_slot]; + if ((slot->target != target) || (slot->texture != texture) || (slot->sampler != sampler)) { + _sg_gl_cache_active_texture((GLenum)(GL_TEXTURE0 + gl_tex_slot)); + // if the target has changed, clear the previous binding on that target + if ((target != slot->target) && (slot->target != 0)) { + glBindTexture(slot->target, 0); + _SG_GL_CHECK_ERROR(); + _sg_stats_add(gl.num_bind_texture, 1); + } + // apply new binding (can be 0 to unbind) + if (target != 0) { + glBindTexture(target, texture); + _SG_GL_CHECK_ERROR(); + _sg_stats_add(gl.num_bind_texture, 1); + } + // apply new sampler (can be 0 to unbind) + glBindSampler((GLuint)gl_tex_slot, sampler); + _SG_GL_CHECK_ERROR(); + _sg_stats_add(gl.num_bind_sampler, 1); + + slot->target = target; + slot->texture = texture; + slot->sampler = sampler; + } +} + +_SOKOL_PRIVATE void _sg_gl_cache_store_texture_sampler_binding(int8_t gl_tex_slot) { + SOKOL_ASSERT((gl_tex_slot >= 0) && (gl_tex_slot < _SG_GL_MAX_IMG_SMP_BINDINGS)); + _sg.gl.cache.stored_texture_sampler = _sg.gl.cache.texture_samplers[gl_tex_slot]; +} + +_SOKOL_PRIVATE void _sg_gl_cache_restore_texture_sampler_binding(int8_t gl_tex_slot) { + SOKOL_ASSERT((gl_tex_slot >= 0) && (gl_tex_slot < _SG_GL_MAX_IMG_SMP_BINDINGS)); + _sg_gl_cache_texture_sampler_bind_slot* slot = &_sg.gl.cache.stored_texture_sampler; + if (slot->texture != 0) { + // we only care about restoring valid ids + SOKOL_ASSERT(slot->target != 0); + _sg_gl_cache_bind_texture_sampler(gl_tex_slot, slot->target, slot->texture, slot->sampler); + slot->target = 0; + slot->texture = 0; + slot->sampler = 0; + } +} + +// called from _sg_gl_discard_texture() and _sg_gl_discard_sampler() +_SOKOL_PRIVATE void _sg_gl_cache_invalidate_texture_sampler(GLuint tex, GLuint smp) { + _SG_GL_CHECK_ERROR(); + for (size_t i = 0; i < _SG_GL_MAX_IMG_SMP_BINDINGS; i++) { + _sg_gl_cache_texture_sampler_bind_slot* slot = &_sg.gl.cache.texture_samplers[i]; + if ((0 != slot->target) && ((tex == slot->texture) || (smp == slot->sampler))) { + _sg_gl_cache_active_texture((GLenum)(GL_TEXTURE0 + i)); + glBindTexture(slot->target, 0); + _SG_GL_CHECK_ERROR(); + _sg_stats_add(gl.num_bind_texture, 1); + glBindSampler((GLuint)i, 0); + _SG_GL_CHECK_ERROR(); + _sg_stats_add(gl.num_bind_sampler, 1); + slot->target = 0; + slot->texture = 0; + slot->sampler = 0; + } + } + if ((tex == _sg.gl.cache.stored_texture_sampler.texture) || (smp == _sg.gl.cache.stored_texture_sampler.sampler)) { + _sg.gl.cache.stored_texture_sampler.target = 0; + _sg.gl.cache.stored_texture_sampler.texture = 0; + _sg.gl.cache.stored_texture_sampler.sampler = 0; + } +} + +// called from _sg_gl_discard_shader() +_SOKOL_PRIVATE void _sg_gl_cache_invalidate_program(GLuint prog) { + if (prog == _sg.gl.cache.prog) { + _sg.gl.cache.prog = 0; + glUseProgram(0); + _sg_stats_add(gl.num_use_program, 1); + } +} + +// called from _sg_gl_discard_pipeline() +_SOKOL_PRIVATE void _sg_gl_cache_invalidate_pipeline(_sg_pipeline_t* pip) { + if (pip == _sg.gl.cache.cur_pipeline) { + _sg.gl.cache.cur_pipeline = 0; + _sg.gl.cache.cur_pipeline_id.id = SG_INVALID_ID; + } +} + +_SOKOL_PRIVATE void _sg_gl_reset_state_cache(void) { + _SG_GL_CHECK_ERROR(); + glBindVertexArray(_sg.gl.vao); + _SG_GL_CHECK_ERROR(); + _sg_clear(&_sg.gl.cache, sizeof(_sg.gl.cache)); + _sg_gl_cache_clear_buffer_bindings(true); + _SG_GL_CHECK_ERROR(); + _sg_gl_cache_clear_texture_sampler_bindings(true); + _SG_GL_CHECK_ERROR(); + for (int i = 0; i < _sg.limits.max_vertex_attrs; i++) { + _sg_gl_attr_t* attr = &_sg.gl.cache.attrs[i].gl_attr; + attr->vb_index = -1; + attr->divisor = -1; + glDisableVertexAttribArray((GLuint)i); + _SG_GL_CHECK_ERROR(); + _sg_stats_add(gl.num_disable_vertex_attrib_array, 1); + } + _sg.gl.cache.cur_primitive_type = GL_TRIANGLES; + + // shader program + glGetIntegerv(GL_CURRENT_PROGRAM, (GLint*)&_sg.gl.cache.prog); + _SG_GL_CHECK_ERROR(); + + // depth and stencil state + _sg.gl.cache.depth.compare = SG_COMPAREFUNC_ALWAYS; + _sg.gl.cache.stencil.front.compare = SG_COMPAREFUNC_ALWAYS; + _sg.gl.cache.stencil.front.fail_op = SG_STENCILOP_KEEP; + _sg.gl.cache.stencil.front.depth_fail_op = SG_STENCILOP_KEEP; + _sg.gl.cache.stencil.front.pass_op = SG_STENCILOP_KEEP; + _sg.gl.cache.stencil.back.compare = SG_COMPAREFUNC_ALWAYS; + _sg.gl.cache.stencil.back.fail_op = SG_STENCILOP_KEEP; + _sg.gl.cache.stencil.back.depth_fail_op = SG_STENCILOP_KEEP; + _sg.gl.cache.stencil.back.pass_op = SG_STENCILOP_KEEP; + glEnable(GL_DEPTH_TEST); + glDepthFunc(GL_ALWAYS); + glDepthMask(GL_FALSE); + glDisable(GL_STENCIL_TEST); + glStencilFunc(GL_ALWAYS, 0, 0); + glStencilOp(GL_KEEP, GL_KEEP, GL_KEEP); + glStencilMask(0); + _sg_stats_add(gl.num_render_state, 7); + + // blend state + _sg.gl.cache.blend.src_factor_rgb = SG_BLENDFACTOR_ONE; + _sg.gl.cache.blend.dst_factor_rgb = SG_BLENDFACTOR_ZERO; + _sg.gl.cache.blend.op_rgb = SG_BLENDOP_ADD; + _sg.gl.cache.blend.src_factor_alpha = SG_BLENDFACTOR_ONE; + _sg.gl.cache.blend.dst_factor_alpha = SG_BLENDFACTOR_ZERO; + _sg.gl.cache.blend.op_alpha = SG_BLENDOP_ADD; + glDisable(GL_BLEND); + glBlendFuncSeparate(GL_ONE, GL_ZERO, GL_ONE, GL_ZERO); + glBlendEquationSeparate(GL_FUNC_ADD, GL_FUNC_ADD); + glBlendColor(0.0f, 0.0f, 0.0f, 0.0f); + _sg_stats_add(gl.num_render_state, 4); + + // standalone state + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + _sg.gl.cache.color_write_mask[i] = SG_COLORMASK_RGBA; + } + _sg.gl.cache.cull_mode = SG_CULLMODE_NONE; + _sg.gl.cache.face_winding = SG_FACEWINDING_CW; + _sg.gl.cache.sample_count = 1; + glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); + glPolygonOffset(0.0f, 0.0f); + glDisable(GL_POLYGON_OFFSET_FILL); + glDisable(GL_CULL_FACE); + glFrontFace(GL_CW); + glCullFace(GL_BACK); + glEnable(GL_SCISSOR_TEST); + glDisable(GL_SAMPLE_ALPHA_TO_COVERAGE); + glEnable(GL_DITHER); + glDisable(GL_POLYGON_OFFSET_FILL); + _sg_stats_add(gl.num_render_state, 10); + #if defined(SOKOL_GLCORE) + glEnable(GL_MULTISAMPLE); + glEnable(GL_PROGRAM_POINT_SIZE); + _sg_stats_add(gl.num_render_state, 2); + #endif +} + +_SOKOL_PRIVATE void _sg_gl_setup_backend(const sg_desc* desc) { + _SOKOL_UNUSED(desc); + + // assumes that _sg.gl is already zero-initialized + _sg.gl.valid = true; + + #if defined(_SOKOL_USE_WIN32_GL_LOADER) + _sg_gl_load_opengl(); + #endif + + // clear initial GL error state + #if defined(SOKOL_DEBUG) + while (glGetError() != GL_NO_ERROR); + #endif + #if defined(SOKOL_GLCORE) + _sg_gl_init_caps_glcore(); + #elif defined(SOKOL_GLES3) + _sg_gl_init_caps_gles3(); + #endif + + glGenVertexArrays(1, &_sg.gl.vao); + glBindVertexArray(_sg.gl.vao); + _SG_GL_CHECK_ERROR(); + // incoming texture data is generally expected to be packed tightly + glPixelStorei(GL_UNPACK_ALIGNMENT, 1); + #if defined(SOKOL_GLCORE) + // enable seamless cubemap sampling (only desktop GL) + glEnable(GL_TEXTURE_CUBE_MAP_SEAMLESS); + #endif + _sg_gl_reset_state_cache(); +} + +_SOKOL_PRIVATE void _sg_gl_discard_backend(void) { + SOKOL_ASSERT(_sg.gl.valid); + if (_sg.gl.vao) { + glDeleteVertexArrays(1, &_sg.gl.vao); + } + #if defined(_SOKOL_USE_WIN32_GL_LOADER) + _sg_gl_unload_opengl(); + #endif + _sg.gl.valid = false; +} + +//-- GL backend resource creation and destruction ------------------------------ +_SOKOL_PRIVATE sg_resource_state _sg_gl_create_buffer(_sg_buffer_t* buf, const sg_buffer_desc* desc) { + SOKOL_ASSERT(buf && desc); + _SG_GL_CHECK_ERROR(); + buf->gl.injected = (0 != desc->gl_buffers[0]); + const GLenum gl_target = _sg_gl_buffer_target(buf->cmn.type); + const GLenum gl_usage = _sg_gl_usage(buf->cmn.usage); + for (int slot = 0; slot < buf->cmn.num_slots; slot++) { + GLuint gl_buf = 0; + if (buf->gl.injected) { + SOKOL_ASSERT(desc->gl_buffers[slot]); + gl_buf = desc->gl_buffers[slot]; + } else { + glGenBuffers(1, &gl_buf); + SOKOL_ASSERT(gl_buf); + _sg_gl_cache_store_buffer_binding(gl_target); + _sg_gl_cache_bind_buffer(gl_target, gl_buf); + glBufferData(gl_target, buf->cmn.size, 0, gl_usage); + if (buf->cmn.usage == SG_USAGE_IMMUTABLE) { + if (desc->data.ptr) { + glBufferSubData(gl_target, 0, buf->cmn.size, desc->data.ptr); + } else { + // setup a zero-initialized buffer (don't explicitly need to do this on WebGL) + #if !defined(__EMSCRIPTEN__) + void* ptr = _sg_malloc_clear((size_t)buf->cmn.size); + glBufferSubData(gl_target, 0, buf->cmn.size, ptr); + _sg_free(ptr); + #endif + } + } + _sg_gl_cache_restore_buffer_binding(gl_target); + } + buf->gl.buf[slot] = gl_buf; + } + _SG_GL_CHECK_ERROR(); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_gl_discard_buffer(_sg_buffer_t* buf) { + SOKOL_ASSERT(buf); + _SG_GL_CHECK_ERROR(); + for (int slot = 0; slot < buf->cmn.num_slots; slot++) { + if (buf->gl.buf[slot]) { + _sg_gl_cache_invalidate_buffer(buf->gl.buf[slot]); + if (!buf->gl.injected) { + glDeleteBuffers(1, &buf->gl.buf[slot]); + } + } + } + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE bool _sg_gl_supported_texture_format(sg_pixel_format fmt) { + const int fmt_index = (int) fmt; + SOKOL_ASSERT((fmt_index > SG_PIXELFORMAT_NONE) && (fmt_index < _SG_PIXELFORMAT_NUM)); + return _sg.formats[fmt_index].sample; +} + +_SOKOL_PRIVATE sg_resource_state _sg_gl_create_image(_sg_image_t* img, const sg_image_desc* desc) { + SOKOL_ASSERT(img && desc); + _SG_GL_CHECK_ERROR(); + const bool msaa = img->cmn.sample_count > 1; + img->gl.injected = (0 != desc->gl_textures[0]); + + // check if texture format is support + if (!_sg_gl_supported_texture_format(img->cmn.pixel_format)) { + _SG_ERROR(GL_TEXTURE_FORMAT_NOT_SUPPORTED); + return SG_RESOURCESTATE_FAILED; + } + const GLenum gl_internal_format = _sg_gl_teximage_internal_format(img->cmn.pixel_format); + + // GLES3/WebGL2/macOS doesn't have support for multisampled textures, so create a render buffer object instead + if (!_sg.features.msaa_image_bindings && img->cmn.render_target && msaa) { + glGenRenderbuffers(1, &img->gl.msaa_render_buffer); + glBindRenderbuffer(GL_RENDERBUFFER, img->gl.msaa_render_buffer); + glRenderbufferStorageMultisample(GL_RENDERBUFFER, img->cmn.sample_count, gl_internal_format, img->cmn.width, img->cmn.height); + } else if (img->gl.injected) { + img->gl.target = _sg_gl_texture_target(img->cmn.type, img->cmn.sample_count); + // inject externally GL textures + for (int slot = 0; slot < img->cmn.num_slots; slot++) { + SOKOL_ASSERT(desc->gl_textures[slot]); + img->gl.tex[slot] = desc->gl_textures[slot]; + } + if (desc->gl_texture_target) { + img->gl.target = (GLenum)desc->gl_texture_target; + } + } else { + // create our own GL texture(s) + img->gl.target = _sg_gl_texture_target(img->cmn.type, img->cmn.sample_count); + const GLenum gl_format = _sg_gl_teximage_format(img->cmn.pixel_format); + const bool is_compressed = _sg_is_compressed_pixel_format(img->cmn.pixel_format); + for (int slot = 0; slot < img->cmn.num_slots; slot++) { + glGenTextures(1, &img->gl.tex[slot]); + SOKOL_ASSERT(img->gl.tex[slot]); + _sg_gl_cache_store_texture_sampler_binding(0); + _sg_gl_cache_bind_texture_sampler(0, img->gl.target, img->gl.tex[slot], 0); + glTexParameteri(img->gl.target, GL_TEXTURE_MAX_LEVEL, img->cmn.num_mipmaps - 1); + + // NOTE: workaround for https://issues.chromium.org/issues/355605685 + // FIXME: on GLES3 and GL 4.3 (e.g. not macOS) the texture initialization + // should be rewritten to use glTexStorage + glTexSubImage + bool tex_storage_allocated = false; + #if defined(__EMSCRIPTEN__) + if (desc->data.subimage[0][0].ptr == 0) { + SOKOL_ASSERT(!msaa); + tex_storage_allocated = true; + if ((SG_IMAGETYPE_2D == img->cmn.type) || (SG_IMAGETYPE_CUBE == img->cmn.type)) { + glTexStorage2D(img->gl.target, img->cmn.num_mipmaps, gl_internal_format, img->cmn.width, img->cmn.height); + } else if ((SG_IMAGETYPE_3D == img->cmn.type) || (SG_IMAGETYPE_ARRAY == img->cmn.type)) { + glTexStorage3D(img->gl.target, img->cmn.num_mipmaps, gl_internal_format, img->cmn.width, img->cmn.height, img->cmn.num_slices); + } + } + #endif + if (!tex_storage_allocated) { + const int num_faces = img->cmn.type == SG_IMAGETYPE_CUBE ? 6 : 1; + int data_index = 0; + for (int face_index = 0; face_index < num_faces; face_index++) { + for (int mip_index = 0; mip_index < img->cmn.num_mipmaps; mip_index++, data_index++) { + GLenum gl_img_target = img->gl.target; + if (SG_IMAGETYPE_CUBE == img->cmn.type) { + gl_img_target = _sg_gl_cubeface_target(face_index); + } + const GLvoid* data_ptr = desc->data.subimage[face_index][mip_index].ptr; + const int mip_width = _sg_miplevel_dim(img->cmn.width, mip_index); + const int mip_height = _sg_miplevel_dim(img->cmn.height, mip_index); + if ((SG_IMAGETYPE_2D == img->cmn.type) || (SG_IMAGETYPE_CUBE == img->cmn.type)) { + if (is_compressed) { + SOKOL_ASSERT(!msaa); + const GLsizei data_size = (GLsizei) desc->data.subimage[face_index][mip_index].size; + glCompressedTexImage2D(gl_img_target, mip_index, gl_internal_format, + mip_width, mip_height, 0, data_size, data_ptr); + } else { + const GLenum gl_type = _sg_gl_teximage_type(img->cmn.pixel_format); + #if defined(SOKOL_GLCORE) && !defined(__APPLE__) + if (msaa) { + glTexImage2DMultisample(gl_img_target, img->cmn.sample_count, gl_internal_format, + mip_width, mip_height, GL_TRUE); + } else { + glTexImage2D(gl_img_target, mip_index, (GLint)gl_internal_format, + mip_width, mip_height, 0, gl_format, gl_type, data_ptr); + } + #else + SOKOL_ASSERT(!msaa); + glTexImage2D(gl_img_target, mip_index, (GLint)gl_internal_format, + mip_width, mip_height, 0, gl_format, gl_type, data_ptr); + #endif + } + } else if ((SG_IMAGETYPE_3D == img->cmn.type) || (SG_IMAGETYPE_ARRAY == img->cmn.type)) { + int mip_depth = img->cmn.num_slices; + if (SG_IMAGETYPE_3D == img->cmn.type) { + mip_depth = _sg_miplevel_dim(mip_depth, mip_index); + } + if (is_compressed) { + SOKOL_ASSERT(!msaa); + const GLsizei data_size = (GLsizei) desc->data.subimage[face_index][mip_index].size; + glCompressedTexImage3D(gl_img_target, mip_index, gl_internal_format, + mip_width, mip_height, mip_depth, 0, data_size, data_ptr); + } else { + const GLenum gl_type = _sg_gl_teximage_type(img->cmn.pixel_format); + #if defined(SOKOL_GLCORE) && !defined(__APPLE__) + if (msaa) { + // NOTE: only for array textures, not actual 3D textures! + glTexImage3DMultisample(gl_img_target, img->cmn.sample_count, gl_internal_format, + mip_width, mip_height, mip_depth, GL_TRUE); + } else { + glTexImage3D(gl_img_target, mip_index, (GLint)gl_internal_format, + mip_width, mip_height, mip_depth, 0, gl_format, gl_type, data_ptr); + } + #else + SOKOL_ASSERT(!msaa); + glTexImage3D(gl_img_target, mip_index, (GLint)gl_internal_format, + mip_width, mip_height, mip_depth, 0, gl_format, gl_type, data_ptr); + #endif + } + } + } + } + } + _sg_gl_cache_restore_texture_sampler_binding(0); + } + } + _SG_GL_CHECK_ERROR(); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_gl_discard_image(_sg_image_t* img) { + SOKOL_ASSERT(img); + _SG_GL_CHECK_ERROR(); + for (int slot = 0; slot < img->cmn.num_slots; slot++) { + if (img->gl.tex[slot]) { + _sg_gl_cache_invalidate_texture_sampler(img->gl.tex[slot], 0); + if (!img->gl.injected) { + glDeleteTextures(1, &img->gl.tex[slot]); + } + } + } + if (img->gl.msaa_render_buffer) { + glDeleteRenderbuffers(1, &img->gl.msaa_render_buffer); + } + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE sg_resource_state _sg_gl_create_sampler(_sg_sampler_t* smp, const sg_sampler_desc* desc) { + SOKOL_ASSERT(smp && desc); + _SG_GL_CHECK_ERROR(); + smp->gl.injected = (0 != desc->gl_sampler); + if (smp->gl.injected) { + smp->gl.smp = (GLuint) desc->gl_sampler; + } else { + glGenSamplers(1, &smp->gl.smp); + SOKOL_ASSERT(smp->gl.smp); + + const GLenum gl_min_filter = _sg_gl_min_filter(smp->cmn.min_filter, smp->cmn.mipmap_filter); + const GLenum gl_mag_filter = _sg_gl_mag_filter(smp->cmn.mag_filter); + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_MIN_FILTER, (GLint)gl_min_filter); + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_MAG_FILTER, (GLint)gl_mag_filter); + // GL spec has strange defaults for mipmap min/max lod: -1000 to +1000 + const float min_lod = _sg_clamp(desc->min_lod, 0.0f, 1000.0f); + const float max_lod = _sg_clamp(desc->max_lod, 0.0f, 1000.0f); + glSamplerParameterf(smp->gl.smp, GL_TEXTURE_MIN_LOD, min_lod); + glSamplerParameterf(smp->gl.smp, GL_TEXTURE_MAX_LOD, max_lod); + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_WRAP_S, (GLint)_sg_gl_wrap(smp->cmn.wrap_u)); + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_WRAP_T, (GLint)_sg_gl_wrap(smp->cmn.wrap_v)); + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_WRAP_R, (GLint)_sg_gl_wrap(smp->cmn.wrap_w)); + #if defined(SOKOL_GLCORE) + float border[4]; + switch (smp->cmn.border_color) { + case SG_BORDERCOLOR_TRANSPARENT_BLACK: + border[0] = 0.0f; border[1] = 0.0f; border[2] = 0.0f; border[3] = 0.0f; + break; + case SG_BORDERCOLOR_OPAQUE_WHITE: + border[0] = 1.0f; border[1] = 1.0f; border[2] = 1.0f; border[3] = 1.0f; + break; + default: + border[0] = 0.0f; border[1] = 0.0f; border[2] = 0.0f; border[3] = 1.0f; + break; + } + glSamplerParameterfv(smp->gl.smp, GL_TEXTURE_BORDER_COLOR, border); + #endif + if (smp->cmn.compare != SG_COMPAREFUNC_NEVER) { + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_COMPARE_MODE, GL_COMPARE_REF_TO_TEXTURE); + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_COMPARE_FUNC, (GLint)_sg_gl_compare_func(smp->cmn.compare)); + } else { + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_COMPARE_MODE, GL_NONE); + } + if (_sg.gl.ext_anisotropic && (smp->cmn.max_anisotropy > 1)) { + GLint max_aniso = (GLint) smp->cmn.max_anisotropy; + if (max_aniso > _sg.gl.max_anisotropy) { + max_aniso = _sg.gl.max_anisotropy; + } + glSamplerParameteri(smp->gl.smp, GL_TEXTURE_MAX_ANISOTROPY_EXT, max_aniso); + } + } + _SG_GL_CHECK_ERROR(); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_gl_discard_sampler(_sg_sampler_t* smp) { + SOKOL_ASSERT(smp); + _SG_GL_CHECK_ERROR(); + _sg_gl_cache_invalidate_texture_sampler(0, smp->gl.smp); + if (!smp->gl.injected) { + glDeleteSamplers(1, &smp->gl.smp); + } + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE GLuint _sg_gl_compile_shader(sg_shader_stage stage, const char* src) { + SOKOL_ASSERT(src); + _SG_GL_CHECK_ERROR(); + GLuint gl_shd = glCreateShader(_sg_gl_shader_stage(stage)); + glShaderSource(gl_shd, 1, &src, 0); + glCompileShader(gl_shd); + GLint compile_status = 0; + glGetShaderiv(gl_shd, GL_COMPILE_STATUS, &compile_status); + if (!compile_status) { + // compilation failed, log error and delete shader + GLint log_len = 0; + glGetShaderiv(gl_shd, GL_INFO_LOG_LENGTH, &log_len); + if (log_len > 0) { + GLchar* log_buf = (GLchar*) _sg_malloc((size_t)log_len); + glGetShaderInfoLog(gl_shd, log_len, &log_len, log_buf); + _SG_ERROR(GL_SHADER_COMPILATION_FAILED); + _SG_LOGMSG(GL_SHADER_COMPILATION_FAILED, log_buf); + _sg_free(log_buf); + } + glDeleteShader(gl_shd); + gl_shd = 0; + } + _SG_GL_CHECK_ERROR(); + return gl_shd; +} + +// NOTE: this is an out-of-range check for GLSL bindslots that's also active in release mode +_SOKOL_PRIVATE bool _sg_gl_ensure_glsl_bindslot_ranges(const sg_shader_desc* desc) { + SOKOL_ASSERT(desc); + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (desc->storage_buffers[i].glsl_binding_n >= _SG_GL_MAX_SBUF_BINDINGS) { + _SG_ERROR(GL_STORAGEBUFFER_GLSL_BINDING_OUT_OF_RANGE); + return false; + } + } + return true; +} + +_SOKOL_PRIVATE sg_resource_state _sg_gl_create_shader(_sg_shader_t* shd, const sg_shader_desc* desc) { + SOKOL_ASSERT(shd && desc); + SOKOL_ASSERT(!shd->gl.prog); + _SG_GL_CHECK_ERROR(); + + // perform a fatal range-check on GLSL bindslots that's also active + // in release mode to avoid potential out-of-bounds array accesses + if (!_sg_gl_ensure_glsl_bindslot_ranges(desc)) { + return SG_RESOURCESTATE_FAILED; + } + + // copy the optional vertex attribute names over + for (int i = 0; i < SG_MAX_VERTEX_ATTRIBUTES; i++) { + _sg_strcpy(&shd->gl.attrs[i].name, desc->attrs[i].glsl_name); + } + + const bool has_vs = desc->vertex_func.source; + const bool has_fs = desc->fragment_func.source; + const bool has_cs = desc->compute_func.source; + SOKOL_ASSERT((has_vs && has_fs) || has_cs); + GLuint gl_prog = glCreateProgram(); + if (has_vs && has_fs) { + GLuint gl_vs = _sg_gl_compile_shader(SG_SHADERSTAGE_VERTEX, desc->vertex_func.source); + GLuint gl_fs = _sg_gl_compile_shader(SG_SHADERSTAGE_FRAGMENT, desc->fragment_func.source); + if (!(gl_vs && gl_fs)) { + glDeleteProgram(gl_prog); + if (gl_vs) { glDeleteShader(gl_vs); } + if (gl_fs) { glDeleteShader(gl_fs); } + return SG_RESOURCESTATE_FAILED; + } + glAttachShader(gl_prog, gl_vs); + glAttachShader(gl_prog, gl_fs); + glLinkProgram(gl_prog); + glDeleteShader(gl_vs); + glDeleteShader(gl_fs); + _SG_GL_CHECK_ERROR(); + } else if (has_cs) { + GLuint gl_cs = _sg_gl_compile_shader(SG_SHADERSTAGE_COMPUTE, desc->compute_func.source); + if (!gl_cs) { + glDeleteProgram(gl_prog); + return SG_RESOURCESTATE_FAILED; + } + glAttachShader(gl_prog, gl_cs); + glLinkProgram(gl_prog); + glDeleteShader(gl_cs); + _SG_GL_CHECK_ERROR(); + } else { + SOKOL_UNREACHABLE; + } + GLint link_status; + glGetProgramiv(gl_prog, GL_LINK_STATUS, &link_status); + if (!link_status) { + GLint log_len = 0; + glGetProgramiv(gl_prog, GL_INFO_LOG_LENGTH, &log_len); + if (log_len > 0) { + GLchar* log_buf = (GLchar*) _sg_malloc((size_t)log_len); + glGetProgramInfoLog(gl_prog, log_len, &log_len, log_buf); + _SG_ERROR(GL_SHADER_LINKING_FAILED); + _SG_LOGMSG(GL_SHADER_LINKING_FAILED, log_buf); + _sg_free(log_buf); + } + glDeleteProgram(gl_prog); + return SG_RESOURCESTATE_FAILED; + } + shd->gl.prog = gl_prog; + + // resolve uniforms + _SG_GL_CHECK_ERROR(); + for (size_t ub_index = 0; ub_index < SG_MAX_UNIFORMBLOCK_BINDSLOTS; ub_index++) { + const sg_shader_uniform_block* ub_desc = &desc->uniform_blocks[ub_index]; + if (ub_desc->stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(ub_desc->size > 0); + _sg_gl_uniform_block_t* ub = &shd->gl.uniform_blocks[ub_index]; + SOKOL_ASSERT(ub->num_uniforms == 0); + uint32_t cur_uniform_offset = 0; + for (int u_index = 0; u_index < SG_MAX_UNIFORMBLOCK_MEMBERS; u_index++) { + const sg_glsl_shader_uniform* u_desc = &ub_desc->glsl_uniforms[u_index]; + if (u_desc->type == SG_UNIFORMTYPE_INVALID) { + break; + } + const uint32_t u_align = _sg_uniform_alignment(u_desc->type, u_desc->array_count, ub_desc->layout); + const uint32_t u_size = _sg_uniform_size(u_desc->type, u_desc->array_count, ub_desc->layout); + cur_uniform_offset = _sg_align_u32(cur_uniform_offset, u_align); + _sg_gl_uniform_t* u = &ub->uniforms[u_index]; + u->type = u_desc->type; + u->count = (uint16_t) u_desc->array_count; + u->offset = (uint16_t) cur_uniform_offset; + SOKOL_ASSERT(u_desc->glsl_name); + u->gl_loc = glGetUniformLocation(gl_prog, u_desc->glsl_name); + if (u->gl_loc == -1) { + _SG_WARN(GL_UNIFORMBLOCK_NAME_NOT_FOUND_IN_SHADER); + _SG_LOGMSG(GL_UNIFORMBLOCK_NAME_NOT_FOUND_IN_SHADER, u_desc->glsl_name); + } + cur_uniform_offset += u_size; + ub->num_uniforms++; + } + if (ub_desc->layout == SG_UNIFORMLAYOUT_STD140) { + cur_uniform_offset = _sg_align_u32(cur_uniform_offset, 16); + } + SOKOL_ASSERT(ub_desc->size == (size_t)cur_uniform_offset); + _SOKOL_UNUSED(cur_uniform_offset); + } + + // copy storage buffer bind slots + for (size_t sbuf_index = 0; sbuf_index < SG_MAX_STORAGEBUFFER_BINDSLOTS; sbuf_index++) { + const sg_shader_storage_buffer* sbuf_desc = &desc->storage_buffers[sbuf_index]; + if (sbuf_desc->stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(sbuf_desc->glsl_binding_n < _SG_GL_MAX_SBUF_BINDINGS); + shd->gl.sbuf_binding[sbuf_index] = sbuf_desc->glsl_binding_n; + } + + // record image sampler location in shader program + _SG_GL_CHECK_ERROR(); + GLuint cur_prog = 0; + glGetIntegerv(GL_CURRENT_PROGRAM, (GLint*)&cur_prog); + glUseProgram(gl_prog); + GLint gl_tex_slot = 0; + for (size_t img_smp_index = 0; img_smp_index < SG_MAX_IMAGE_SAMPLER_PAIRS; img_smp_index++) { + const sg_shader_image_sampler_pair* img_smp_desc = &desc->image_sampler_pairs[img_smp_index]; + if (img_smp_desc->stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(img_smp_desc->glsl_name); + GLint gl_loc = glGetUniformLocation(gl_prog, img_smp_desc->glsl_name); + if (gl_loc != -1) { + glUniform1i(gl_loc, gl_tex_slot); + shd->gl.tex_slot[img_smp_index] = (int8_t)gl_tex_slot++; + } else { + shd->gl.tex_slot[img_smp_index] = -1; + _SG_WARN(GL_IMAGE_SAMPLER_NAME_NOT_FOUND_IN_SHADER); + _SG_LOGMSG(GL_IMAGE_SAMPLER_NAME_NOT_FOUND_IN_SHADER, img_smp_desc->glsl_name); + } + } + + // it's legal to call glUseProgram with 0 + glUseProgram(cur_prog); + _SG_GL_CHECK_ERROR(); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_gl_discard_shader(_sg_shader_t* shd) { + SOKOL_ASSERT(shd); + _SG_GL_CHECK_ERROR(); + if (shd->gl.prog) { + _sg_gl_cache_invalidate_program(shd->gl.prog); + glDeleteProgram(shd->gl.prog); + } + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE sg_resource_state _sg_gl_create_pipeline(_sg_pipeline_t* pip, _sg_shader_t* shd, const sg_pipeline_desc* desc) { + SOKOL_ASSERT(pip && shd && desc); + SOKOL_ASSERT((pip->shader == 0) && (pip->cmn.shader_id.id != SG_INVALID_ID)); + SOKOL_ASSERT(desc->shader.id == shd->slot.id); + SOKOL_ASSERT(shd->gl.prog); + SOKOL_ASSERT(_sg.limits.max_vertex_attrs <= SG_MAX_VERTEX_ATTRIBUTES); + pip->shader = shd; + if (pip->cmn.is_compute) { + // shortcut for compute pipelines + return SG_RESOURCESTATE_VALID; + } + pip->gl.primitive_type = desc->primitive_type; + pip->gl.depth = desc->depth; + pip->gl.stencil = desc->stencil; + // FIXME: blend color and write mask per draw-buffer-attachment (requires GL4) + pip->gl.blend = desc->colors[0].blend; + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + pip->gl.color_write_mask[i] = desc->colors[i].write_mask; + } + pip->gl.cull_mode = desc->cull_mode; + pip->gl.face_winding = desc->face_winding; + pip->gl.sample_count = desc->sample_count; + pip->gl.alpha_to_coverage_enabled = desc->alpha_to_coverage_enabled; + + // NOTE: GLSL compilers may remove unused vertex attributes so we can't rely + // on the 'prepopulated' vertex_buffer_layout_active[] state and need to + // fill this array from scratch with the actual info after GLSL compilation + for (int i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + pip->cmn.vertex_buffer_layout_active[i] = false; + } + + // resolve vertex attributes + for (int attr_index = 0; attr_index < SG_MAX_VERTEX_ATTRIBUTES; attr_index++) { + pip->gl.attrs[attr_index].vb_index = -1; + } + for (int attr_index = 0; attr_index < _sg.limits.max_vertex_attrs; attr_index++) { + const sg_vertex_attr_state* a_state = &desc->layout.attrs[attr_index]; + if (a_state->format == SG_VERTEXFORMAT_INVALID) { + break; + } + SOKOL_ASSERT(a_state->buffer_index < SG_MAX_VERTEXBUFFER_BINDSLOTS); + const sg_vertex_buffer_layout_state* l_state = &desc->layout.buffers[a_state->buffer_index]; + const sg_vertex_step step_func = l_state->step_func; + const int step_rate = l_state->step_rate; + GLint attr_loc = attr_index; + if (!_sg_strempty(&shd->gl.attrs[attr_index].name)) { + attr_loc = glGetAttribLocation(pip->shader->gl.prog, _sg_strptr(&shd->gl.attrs[attr_index].name)); + } + if (attr_loc != -1) { + SOKOL_ASSERT(attr_loc < (GLint)_sg.limits.max_vertex_attrs); + _sg_gl_attr_t* gl_attr = &pip->gl.attrs[attr_loc]; + SOKOL_ASSERT(gl_attr->vb_index == -1); + gl_attr->vb_index = (int8_t) a_state->buffer_index; + if (step_func == SG_VERTEXSTEP_PER_VERTEX) { + gl_attr->divisor = 0; + } else { + gl_attr->divisor = (int8_t) step_rate; + pip->cmn.use_instanced_draw = true; + } + SOKOL_ASSERT(l_state->stride > 0); + gl_attr->stride = (uint8_t) l_state->stride; + gl_attr->offset = a_state->offset; + gl_attr->size = (uint8_t) _sg_gl_vertexformat_size(a_state->format); + gl_attr->type = _sg_gl_vertexformat_type(a_state->format); + gl_attr->normalized = _sg_gl_vertexformat_normalized(a_state->format); + gl_attr->base_type = _sg_vertexformat_basetype(a_state->format); + pip->cmn.vertex_buffer_layout_active[a_state->buffer_index] = true; + } else { + _SG_WARN(GL_VERTEX_ATTRIBUTE_NOT_FOUND_IN_SHADER); + _SG_LOGMSG(GL_VERTEX_ATTRIBUTE_NOT_FOUND_IN_SHADER, _sg_strptr(&shd->gl.attrs[attr_index].name)); + } + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_gl_discard_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + _sg_gl_cache_invalidate_pipeline(pip); +} + +_SOKOL_PRIVATE void _sg_gl_fb_attach_texture(const _sg_gl_attachment_t* gl_att, const _sg_attachment_common_t* cmn_att, GLenum gl_att_type) { + const _sg_image_t* img = gl_att->image; + SOKOL_ASSERT(img); + const GLuint gl_tex = img->gl.tex[0]; + SOKOL_ASSERT(gl_tex); + const GLuint gl_target = img->gl.target; + SOKOL_ASSERT(gl_target); + const int mip_level = cmn_att->mip_level; + const int slice = cmn_att->slice; + switch (img->cmn.type) { + case SG_IMAGETYPE_2D: + glFramebufferTexture2D(GL_FRAMEBUFFER, gl_att_type, gl_target, gl_tex, mip_level); + break; + case SG_IMAGETYPE_CUBE: + glFramebufferTexture2D(GL_FRAMEBUFFER, gl_att_type, _sg_gl_cubeface_target(slice), gl_tex, mip_level); + break; + default: + glFramebufferTextureLayer(GL_FRAMEBUFFER, gl_att_type, gl_tex, mip_level, slice); + break; + } +} + +_SOKOL_PRIVATE GLenum _sg_gl_depth_stencil_attachment_type(const _sg_gl_attachment_t* ds_att) { + const _sg_image_t* img = ds_att->image; + SOKOL_ASSERT(img); + if (_sg_is_depth_stencil_format(img->cmn.pixel_format)) { + return GL_DEPTH_STENCIL_ATTACHMENT; + } else { + return GL_DEPTH_ATTACHMENT; + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_gl_create_attachments(_sg_attachments_t* atts, _sg_image_t** color_images, _sg_image_t** resolve_images, _sg_image_t* ds_image, const sg_attachments_desc* desc) { + SOKOL_ASSERT(atts && desc); + SOKOL_ASSERT(color_images && resolve_images); + _SG_GL_CHECK_ERROR(); + + // copy image pointers + for (int i = 0; i < atts->cmn.num_colors; i++) { + const sg_attachment_desc* color_desc = &desc->colors[i]; + _SOKOL_UNUSED(color_desc); + SOKOL_ASSERT(color_desc->image.id != SG_INVALID_ID); + SOKOL_ASSERT(0 == atts->gl.colors[i].image); + SOKOL_ASSERT(color_images[i] && (color_images[i]->slot.id == color_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_color_format(color_images[i]->cmn.pixel_format)); + atts->gl.colors[i].image = color_images[i]; + + const sg_attachment_desc* resolve_desc = &desc->resolves[i]; + if (resolve_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(0 == atts->gl.resolves[i].image); + SOKOL_ASSERT(resolve_images[i] && (resolve_images[i]->slot.id == resolve_desc->image.id)); + SOKOL_ASSERT(color_images[i] && (color_images[i]->cmn.pixel_format == resolve_images[i]->cmn.pixel_format)); + atts->gl.resolves[i].image = resolve_images[i]; + } + } + SOKOL_ASSERT(0 == atts->gl.depth_stencil.image); + const sg_attachment_desc* ds_desc = &desc->depth_stencil; + if (ds_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(ds_image && (ds_image->slot.id == ds_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_depth_format(ds_image->cmn.pixel_format)); + atts->gl.depth_stencil.image = ds_image; + } + + // store current framebuffer binding (restored at end of function) + GLuint gl_orig_fb; + glGetIntegerv(GL_FRAMEBUFFER_BINDING, (GLint*)&gl_orig_fb); + + // create a framebuffer object + glGenFramebuffers(1, &atts->gl.fb); + glBindFramebuffer(GL_FRAMEBUFFER, atts->gl.fb); + + // attach color attachments to framebuffer + for (int i = 0; i < atts->cmn.num_colors; i++) { + const _sg_image_t* color_img = atts->gl.colors[i].image; + SOKOL_ASSERT(color_img); + const GLuint gl_msaa_render_buffer = color_img->gl.msaa_render_buffer; + if (gl_msaa_render_buffer) { + glFramebufferRenderbuffer(GL_FRAMEBUFFER, (GLenum)(GL_COLOR_ATTACHMENT0+i), GL_RENDERBUFFER, gl_msaa_render_buffer); + } else { + const GLenum gl_att_type = (GLenum)(GL_COLOR_ATTACHMENT0 + i); + _sg_gl_fb_attach_texture(&atts->gl.colors[i], &atts->cmn.colors[i], gl_att_type); + } + } + // attach depth-stencil attachment + if (atts->gl.depth_stencil.image) { + const GLenum gl_att = _sg_gl_depth_stencil_attachment_type(&atts->gl.depth_stencil); + const _sg_image_t* ds_img = atts->gl.depth_stencil.image; + const GLuint gl_msaa_render_buffer = ds_img->gl.msaa_render_buffer; + if (gl_msaa_render_buffer) { + glFramebufferRenderbuffer(GL_FRAMEBUFFER, gl_att, GL_RENDERBUFFER, gl_msaa_render_buffer); + } else { + const GLenum gl_att_type = _sg_gl_depth_stencil_attachment_type(&atts->gl.depth_stencil); + _sg_gl_fb_attach_texture(&atts->gl.depth_stencil, &atts->cmn.depth_stencil, gl_att_type); + } + } + + // check if framebuffer is complete + { + const GLenum fb_status = glCheckFramebufferStatus(GL_FRAMEBUFFER); + if (fb_status != GL_FRAMEBUFFER_COMPLETE) { + switch (fb_status) { + case GL_FRAMEBUFFER_UNDEFINED: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_UNDEFINED); + break; + case GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_INCOMPLETE_ATTACHMENT); + break; + case GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_INCOMPLETE_MISSING_ATTACHMENT); + break; + case GL_FRAMEBUFFER_UNSUPPORTED: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_UNSUPPORTED); + break; + case GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_INCOMPLETE_MULTISAMPLE); + break; + default: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_UNKNOWN); + break; + } + return SG_RESOURCESTATE_FAILED; + } + } + + // setup color attachments for the framebuffer + static const GLenum gl_draw_bufs[SG_MAX_COLOR_ATTACHMENTS] = { + GL_COLOR_ATTACHMENT0, + GL_COLOR_ATTACHMENT1, + GL_COLOR_ATTACHMENT2, + GL_COLOR_ATTACHMENT3 + }; + glDrawBuffers(atts->cmn.num_colors, gl_draw_bufs); + + // create MSAA resolve framebuffers if necessary + for (int i = 0; i < atts->cmn.num_colors; i++) { + _sg_gl_attachment_t* gl_resolve_att = &atts->gl.resolves[i]; + if (gl_resolve_att->image) { + _sg_attachment_common_t* cmn_resolve_att = &atts->cmn.resolves[i]; + SOKOL_ASSERT(0 == atts->gl.msaa_resolve_framebuffer[i]); + glGenFramebuffers(1, &atts->gl.msaa_resolve_framebuffer[i]); + glBindFramebuffer(GL_FRAMEBUFFER, atts->gl.msaa_resolve_framebuffer[i]); + _sg_gl_fb_attach_texture(gl_resolve_att, cmn_resolve_att, GL_COLOR_ATTACHMENT0); + // check if framebuffer is complete + const GLenum fb_status = glCheckFramebufferStatus(GL_FRAMEBUFFER); + if (fb_status != GL_FRAMEBUFFER_COMPLETE) { + switch (fb_status) { + case GL_FRAMEBUFFER_UNDEFINED: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_UNDEFINED); + break; + case GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_INCOMPLETE_ATTACHMENT); + break; + case GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_INCOMPLETE_MISSING_ATTACHMENT); + break; + case GL_FRAMEBUFFER_UNSUPPORTED: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_UNSUPPORTED); + break; + case GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_INCOMPLETE_MULTISAMPLE); + break; + default: + _SG_ERROR(GL_FRAMEBUFFER_STATUS_UNKNOWN); + break; + } + return SG_RESOURCESTATE_FAILED; + } + // setup color attachments for the framebuffer + glDrawBuffers(1, &gl_draw_bufs[0]); + } + } + + // restore original framebuffer binding + glBindFramebuffer(GL_FRAMEBUFFER, gl_orig_fb); + _SG_GL_CHECK_ERROR(); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_gl_discard_attachments(_sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + _SG_GL_CHECK_ERROR(); + if (0 != atts->gl.fb) { + glDeleteFramebuffers(1, &atts->gl.fb); + } + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + if (atts->gl.msaa_resolve_framebuffer[i]) { + glDeleteFramebuffers(1, &atts->gl.msaa_resolve_framebuffer[i]); + } + } + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE _sg_image_t* _sg_gl_attachments_color_image(const _sg_attachments_t* atts, int index) { + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + return atts->gl.colors[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_gl_attachments_resolve_image(const _sg_attachments_t* atts, int index) { + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + return atts->gl.resolves[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_gl_attachments_ds_image(const _sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + return atts->gl.depth_stencil.image; +} + +_SOKOL_PRIVATE void _sg_gl_begin_pass(const sg_pass* pass) { + // FIXME: what if a texture used as render target is still bound, should we + // unbind all currently bound textures in begin pass? + SOKOL_ASSERT(pass); + _SG_GL_CHECK_ERROR(); + + // early out if this a compute pass + if (pass->compute) { + return; + } + + const _sg_attachments_t* atts = _sg.cur_pass.atts; + const sg_swapchain* swapchain = &pass->swapchain; + const sg_pass_action* action = &pass->action; + + // bind the render pass framebuffer + // + // FIXME: Disabling SRGB conversion for the default framebuffer is + // a crude hack to make behaviour for sRGB render target textures + // identical with the Metal and D3D11 swapchains created by sokol-app. + // + // This will need a cleaner solution (e.g. allowing to configure + // sokol_app.h with an sRGB or RGB framebuffer. + if (atts) { + // offscreen pass + SOKOL_ASSERT(atts->gl.fb); + #if defined(SOKOL_GLCORE) + glEnable(GL_FRAMEBUFFER_SRGB); + #endif + glBindFramebuffer(GL_FRAMEBUFFER, atts->gl.fb); + } else { + // default pass + #if defined(SOKOL_GLCORE) + glDisable(GL_FRAMEBUFFER_SRGB); + #endif + // NOTE: on some platforms, the default framebuffer of a context + // is null, so we can't actually assert here that the + // framebuffer has been provided + glBindFramebuffer(GL_FRAMEBUFFER, swapchain->gl.framebuffer); + } + glViewport(0, 0, _sg.cur_pass.width, _sg.cur_pass.height); + glScissor(0, 0, _sg.cur_pass.width, _sg.cur_pass.height); + + // number of color attachments + const int num_color_atts = atts ? atts->cmn.num_colors : 1; + + // clear color and depth-stencil attachments if needed + bool clear_any_color = false; + for (int i = 0; i < num_color_atts; i++) { + if (SG_LOADACTION_CLEAR == action->colors[i].load_action) { + clear_any_color = true; + break; + } + } + const bool clear_depth = (action->depth.load_action == SG_LOADACTION_CLEAR); + const bool clear_stencil = (action->stencil.load_action == SG_LOADACTION_CLEAR); + + bool need_pip_cache_flush = false; + if (clear_any_color) { + bool need_color_mask_flush = false; + // NOTE: not a bug to iterate over all possible color attachments + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + if (SG_COLORMASK_RGBA != _sg.gl.cache.color_write_mask[i]) { + need_pip_cache_flush = true; + need_color_mask_flush = true; + _sg.gl.cache.color_write_mask[i] = SG_COLORMASK_RGBA; + } + } + if (need_color_mask_flush) { + glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); + } + } + if (clear_depth) { + if (!_sg.gl.cache.depth.write_enabled) { + need_pip_cache_flush = true; + _sg.gl.cache.depth.write_enabled = true; + glDepthMask(GL_TRUE); + } + if (_sg.gl.cache.depth.compare != SG_COMPAREFUNC_ALWAYS) { + need_pip_cache_flush = true; + _sg.gl.cache.depth.compare = SG_COMPAREFUNC_ALWAYS; + glDepthFunc(GL_ALWAYS); + } + } + if (clear_stencil) { + if (_sg.gl.cache.stencil.write_mask != 0xFF) { + need_pip_cache_flush = true; + _sg.gl.cache.stencil.write_mask = 0xFF; + glStencilMask(0xFF); + } + } + if (need_pip_cache_flush) { + // we messed with the state cache directly, need to clear cached + // pipeline to force re-evaluation in next sg_apply_pipeline() + _sg.gl.cache.cur_pipeline = 0; + _sg.gl.cache.cur_pipeline_id.id = SG_INVALID_ID; + } + for (int i = 0; i < num_color_atts; i++) { + if (action->colors[i].load_action == SG_LOADACTION_CLEAR) { + glClearBufferfv(GL_COLOR, i, &action->colors[i].clear_value.r); + } + } + if ((atts == 0) || (atts->gl.depth_stencil.image)) { + if (clear_depth && clear_stencil) { + glClearBufferfi(GL_DEPTH_STENCIL, 0, action->depth.clear_value, action->stencil.clear_value); + } else if (clear_depth) { + glClearBufferfv(GL_DEPTH, 0, &action->depth.clear_value); + } else if (clear_stencil) { + GLint val = (GLint) action->stencil.clear_value; + glClearBufferiv(GL_STENCIL, 0, &val); + } + } + // keep store actions for end-pass + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + _sg.gl.color_store_actions[i] = action->colors[i].store_action; + } + _sg.gl.depth_store_action = action->depth.store_action; + _sg.gl.stencil_store_action = action->stencil.store_action; + + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE void _sg_gl_end_pass(void) { + _SG_GL_CHECK_ERROR(); + + if (_sg.cur_pass.atts) { + const _sg_attachments_t* atts = _sg.cur_pass.atts; + SOKOL_ASSERT(atts->slot.id == _sg.cur_pass.atts_id.id); + bool fb_read_bound = false; + bool fb_draw_bound = false; + const int num_color_atts = atts->cmn.num_colors; + for (int i = 0; i < num_color_atts; i++) { + // perform MSAA resolve if needed + if (atts->gl.msaa_resolve_framebuffer[i] != 0) { + if (!fb_read_bound) { + SOKOL_ASSERT(atts->gl.fb); + glBindFramebuffer(GL_READ_FRAMEBUFFER, atts->gl.fb); + fb_read_bound = true; + } + const int w = atts->gl.colors[i].image->cmn.width; + const int h = atts->gl.colors[i].image->cmn.height; + glBindFramebuffer(GL_DRAW_FRAMEBUFFER, atts->gl.msaa_resolve_framebuffer[i]); + glReadBuffer((GLenum)(GL_COLOR_ATTACHMENT0 + i)); + glBlitFramebuffer(0, 0, w, h, 0, 0, w, h, GL_COLOR_BUFFER_BIT, GL_NEAREST); + fb_draw_bound = true; + } + } + + // invalidate framebuffers + _SOKOL_UNUSED(fb_draw_bound); + #if defined(SOKOL_GLES3) + // need to restore framebuffer binding before invalidate if the MSAA resolve had changed the binding + if (fb_draw_bound) { + glBindFramebuffer(GL_FRAMEBUFFER, atts->gl.fb); + } + GLenum invalidate_atts[SG_MAX_COLOR_ATTACHMENTS + 2] = { 0 }; + int att_index = 0; + for (int i = 0; i < num_color_atts; i++) { + if (_sg.gl.color_store_actions[i] == SG_STOREACTION_DONTCARE) { + invalidate_atts[att_index++] = (GLenum)(GL_COLOR_ATTACHMENT0 + i); + } + } + if ((_sg.gl.depth_store_action == SG_STOREACTION_DONTCARE) && (_sg.cur_pass.atts->cmn.depth_stencil.image_id.id != SG_INVALID_ID)) { + invalidate_atts[att_index++] = GL_DEPTH_ATTACHMENT; + } + if ((_sg.gl.stencil_store_action == SG_STOREACTION_DONTCARE) && (_sg.cur_pass.atts->cmn.depth_stencil.image_id.id != SG_INVALID_ID)) { + invalidate_atts[att_index++] = GL_STENCIL_ATTACHMENT; + } + if (att_index > 0) { + glInvalidateFramebuffer(GL_DRAW_FRAMEBUFFER, att_index, invalidate_atts); + } + #endif + } + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE void _sg_gl_apply_viewport(int x, int y, int w, int h, bool origin_top_left) { + y = origin_top_left ? (_sg.cur_pass.height - (y+h)) : y; + glViewport(x, y, w, h); +} + +_SOKOL_PRIVATE void _sg_gl_apply_scissor_rect(int x, int y, int w, int h, bool origin_top_left) { + y = origin_top_left ? (_sg.cur_pass.height - (y+h)) : y; + glScissor(x, y, w, h); +} + +_SOKOL_PRIVATE void _sg_gl_apply_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + SOKOL_ASSERT(pip->shader && (pip->cmn.shader_id.id == pip->shader->slot.id)); + _SG_GL_CHECK_ERROR(); + if ((_sg.gl.cache.cur_pipeline != pip) || (_sg.gl.cache.cur_pipeline_id.id != pip->slot.id)) { + _sg.gl.cache.cur_pipeline = pip; + _sg.gl.cache.cur_pipeline_id.id = pip->slot.id; + + // bind shader program + if (pip->shader->gl.prog != _sg.gl.cache.prog) { + _sg.gl.cache.prog = pip->shader->gl.prog; + glUseProgram(pip->shader->gl.prog); + _sg_stats_add(gl.num_use_program, 1); + } + + // if this is a compute pass, can early-out here + if (pip->cmn.is_compute) { + _SG_GL_CHECK_ERROR(); + return; + } + + // update render pipeline state + _sg.gl.cache.cur_primitive_type = _sg_gl_primitive_type(pip->gl.primitive_type); + _sg.gl.cache.cur_index_type = _sg_gl_index_type(pip->cmn.index_type); + + // update depth state + { + const sg_depth_state* state_ds = &pip->gl.depth; + sg_depth_state* cache_ds = &_sg.gl.cache.depth; + if (state_ds->compare != cache_ds->compare) { + cache_ds->compare = state_ds->compare; + glDepthFunc(_sg_gl_compare_func(state_ds->compare)); + _sg_stats_add(gl.num_render_state, 1); + } + if (state_ds->write_enabled != cache_ds->write_enabled) { + cache_ds->write_enabled = state_ds->write_enabled; + glDepthMask(state_ds->write_enabled); + _sg_stats_add(gl.num_render_state, 1); + } + if (!_sg_fequal(state_ds->bias, cache_ds->bias, 0.000001f) || + !_sg_fequal(state_ds->bias_slope_scale, cache_ds->bias_slope_scale, 0.000001f)) + { + /* according to ANGLE's D3D11 backend: + D3D11 SlopeScaledDepthBias ==> GL polygonOffsetFactor + D3D11 DepthBias ==> GL polygonOffsetUnits + DepthBiasClamp has no meaning on GL + */ + cache_ds->bias = state_ds->bias; + cache_ds->bias_slope_scale = state_ds->bias_slope_scale; + glPolygonOffset(state_ds->bias_slope_scale, state_ds->bias); + _sg_stats_add(gl.num_render_state, 1); + bool po_enabled = true; + if (_sg_fequal(state_ds->bias, 0.0f, 0.000001f) && + _sg_fequal(state_ds->bias_slope_scale, 0.0f, 0.000001f)) + { + po_enabled = false; + } + if (po_enabled != _sg.gl.cache.polygon_offset_enabled) { + _sg.gl.cache.polygon_offset_enabled = po_enabled; + if (po_enabled) { + glEnable(GL_POLYGON_OFFSET_FILL); + } else { + glDisable(GL_POLYGON_OFFSET_FILL); + } + _sg_stats_add(gl.num_render_state, 1); + } + } + } + + // update stencil state + { + const sg_stencil_state* state_ss = &pip->gl.stencil; + sg_stencil_state* cache_ss = &_sg.gl.cache.stencil; + if (state_ss->enabled != cache_ss->enabled) { + cache_ss->enabled = state_ss->enabled; + if (state_ss->enabled) { + glEnable(GL_STENCIL_TEST); + } else { + glDisable(GL_STENCIL_TEST); + } + _sg_stats_add(gl.num_render_state, 1); + } + if (state_ss->write_mask != cache_ss->write_mask) { + cache_ss->write_mask = state_ss->write_mask; + glStencilMask(state_ss->write_mask); + _sg_stats_add(gl.num_render_state, 1); + } + for (int i = 0; i < 2; i++) { + const sg_stencil_face_state* state_sfs = (i==0)? &state_ss->front : &state_ss->back; + sg_stencil_face_state* cache_sfs = (i==0)? &cache_ss->front : &cache_ss->back; + GLenum gl_face = (i==0)? GL_FRONT : GL_BACK; + if ((state_sfs->compare != cache_sfs->compare) || + (state_ss->read_mask != cache_ss->read_mask) || + (state_ss->ref != cache_ss->ref)) + { + cache_sfs->compare = state_sfs->compare; + glStencilFuncSeparate(gl_face, + _sg_gl_compare_func(state_sfs->compare), + state_ss->ref, + state_ss->read_mask); + _sg_stats_add(gl.num_render_state, 1); + } + if ((state_sfs->fail_op != cache_sfs->fail_op) || + (state_sfs->depth_fail_op != cache_sfs->depth_fail_op) || + (state_sfs->pass_op != cache_sfs->pass_op)) + { + cache_sfs->fail_op = state_sfs->fail_op; + cache_sfs->depth_fail_op = state_sfs->depth_fail_op; + cache_sfs->pass_op = state_sfs->pass_op; + glStencilOpSeparate(gl_face, + _sg_gl_stencil_op(state_sfs->fail_op), + _sg_gl_stencil_op(state_sfs->depth_fail_op), + _sg_gl_stencil_op(state_sfs->pass_op)); + _sg_stats_add(gl.num_render_state, 1); + } + } + cache_ss->read_mask = state_ss->read_mask; + cache_ss->ref = state_ss->ref; + } + + if (pip->cmn.color_count > 0) { + // update blend state + // FIXME: separate blend state per color attachment + const sg_blend_state* state_bs = &pip->gl.blend; + sg_blend_state* cache_bs = &_sg.gl.cache.blend; + if (state_bs->enabled != cache_bs->enabled) { + cache_bs->enabled = state_bs->enabled; + if (state_bs->enabled) { + glEnable(GL_BLEND); + } else { + glDisable(GL_BLEND); + } + _sg_stats_add(gl.num_render_state, 1); + } + if ((state_bs->src_factor_rgb != cache_bs->src_factor_rgb) || + (state_bs->dst_factor_rgb != cache_bs->dst_factor_rgb) || + (state_bs->src_factor_alpha != cache_bs->src_factor_alpha) || + (state_bs->dst_factor_alpha != cache_bs->dst_factor_alpha)) + { + cache_bs->src_factor_rgb = state_bs->src_factor_rgb; + cache_bs->dst_factor_rgb = state_bs->dst_factor_rgb; + cache_bs->src_factor_alpha = state_bs->src_factor_alpha; + cache_bs->dst_factor_alpha = state_bs->dst_factor_alpha; + glBlendFuncSeparate(_sg_gl_blend_factor(state_bs->src_factor_rgb), + _sg_gl_blend_factor(state_bs->dst_factor_rgb), + _sg_gl_blend_factor(state_bs->src_factor_alpha), + _sg_gl_blend_factor(state_bs->dst_factor_alpha)); + _sg_stats_add(gl.num_render_state, 1); + } + if ((state_bs->op_rgb != cache_bs->op_rgb) || (state_bs->op_alpha != cache_bs->op_alpha)) { + cache_bs->op_rgb = state_bs->op_rgb; + cache_bs->op_alpha = state_bs->op_alpha; + glBlendEquationSeparate(_sg_gl_blend_op(state_bs->op_rgb), _sg_gl_blend_op(state_bs->op_alpha)); + _sg_stats_add(gl.num_render_state, 1); + } + + // standalone color target state + for (GLuint i = 0; i < (GLuint)pip->cmn.color_count; i++) { + if (pip->gl.color_write_mask[i] != _sg.gl.cache.color_write_mask[i]) { + const sg_color_mask cm = pip->gl.color_write_mask[i]; + _sg.gl.cache.color_write_mask[i] = cm; + #ifdef SOKOL_GLCORE + glColorMaski(i, + (cm & SG_COLORMASK_R) != 0, + (cm & SG_COLORMASK_G) != 0, + (cm & SG_COLORMASK_B) != 0, + (cm & SG_COLORMASK_A) != 0); + #else + if (0 == i) { + glColorMask((cm & SG_COLORMASK_R) != 0, + (cm & SG_COLORMASK_G) != 0, + (cm & SG_COLORMASK_B) != 0, + (cm & SG_COLORMASK_A) != 0); + } + #endif + _sg_stats_add(gl.num_render_state, 1); + } + } + + if (!_sg_fequal(pip->cmn.blend_color.r, _sg.gl.cache.blend_color.r, 0.0001f) || + !_sg_fequal(pip->cmn.blend_color.g, _sg.gl.cache.blend_color.g, 0.0001f) || + !_sg_fequal(pip->cmn.blend_color.b, _sg.gl.cache.blend_color.b, 0.0001f) || + !_sg_fequal(pip->cmn.blend_color.a, _sg.gl.cache.blend_color.a, 0.0001f)) + { + sg_color c = pip->cmn.blend_color; + _sg.gl.cache.blend_color = c; + glBlendColor(c.r, c.g, c.b, c.a); + _sg_stats_add(gl.num_render_state, 1); + } + } // pip->cmn.color_count > 0 + + if (pip->gl.cull_mode != _sg.gl.cache.cull_mode) { + _sg.gl.cache.cull_mode = pip->gl.cull_mode; + if (SG_CULLMODE_NONE == pip->gl.cull_mode) { + glDisable(GL_CULL_FACE); + _sg_stats_add(gl.num_render_state, 1); + } else { + glEnable(GL_CULL_FACE); + GLenum gl_mode = (SG_CULLMODE_FRONT == pip->gl.cull_mode) ? GL_FRONT : GL_BACK; + glCullFace(gl_mode); + _sg_stats_add(gl.num_render_state, 2); + } + } + if (pip->gl.face_winding != _sg.gl.cache.face_winding) { + _sg.gl.cache.face_winding = pip->gl.face_winding; + GLenum gl_winding = (SG_FACEWINDING_CW == pip->gl.face_winding) ? GL_CW : GL_CCW; + glFrontFace(gl_winding); + _sg_stats_add(gl.num_render_state, 1); + } + if (pip->gl.alpha_to_coverage_enabled != _sg.gl.cache.alpha_to_coverage_enabled) { + _sg.gl.cache.alpha_to_coverage_enabled = pip->gl.alpha_to_coverage_enabled; + if (pip->gl.alpha_to_coverage_enabled) { + glEnable(GL_SAMPLE_ALPHA_TO_COVERAGE); + } else { + glDisable(GL_SAMPLE_ALPHA_TO_COVERAGE); + } + _sg_stats_add(gl.num_render_state, 1); + } + #ifdef SOKOL_GLCORE + if (pip->gl.sample_count != _sg.gl.cache.sample_count) { + _sg.gl.cache.sample_count = pip->gl.sample_count; + if (pip->gl.sample_count > 1) { + glEnable(GL_MULTISAMPLE); + } else { + glDisable(GL_MULTISAMPLE); + } + _sg_stats_add(gl.num_render_state, 1); + } + #endif + + } + _SG_GL_CHECK_ERROR(); +} + +#if defined _SOKOL_GL_HAS_COMPUTE +_SOKOL_PRIVATE void _sg_gl_handle_memory_barriers(const _sg_shader_t* shd, const _sg_bindings_t* bnd) { + if (!_sg.features.compute) { + return; + } + // NOTE: currently only storage buffers can be GPU-written, and storage + // buffers cannot be bound as vertex- or index-buffers. + bool needs_barrier = false; + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (shd->cmn.storage_buffers[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + _sg_buffer_t* buf = bnd->sbufs[i]; + // if this buffer has pending GPU changes, issue a memory barrier + if (buf->gl.gpu_dirty) { + buf->gl.gpu_dirty = false; + needs_barrier = true; + } + // if this binding is going to be written by the GPU set the buffer to 'gpu_dirty' + if (!shd->cmn.storage_buffers[i].readonly) { + buf->gl.gpu_dirty = true; + } + } + if (needs_barrier) { + glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT); + _sg_stats_add(gl.num_memory_barriers, 1); + } +} +#endif + +_SOKOL_PRIVATE bool _sg_gl_apply_bindings(_sg_bindings_t* bnd) { + SOKOL_ASSERT(bnd); + SOKOL_ASSERT(bnd->pip && bnd->pip->shader); + SOKOL_ASSERT(bnd->pip->shader->slot.id == bnd->pip->cmn.shader_id.id); + _SG_GL_CHECK_ERROR(); + const _sg_shader_t* shd = bnd->pip->shader; + + // take care of storage buffer memory barriers + #if defined(_SOKOL_GL_HAS_COMPUTE) + _sg_gl_handle_memory_barriers(shd, bnd); + #endif + + // bind combined image-samplers + _SG_GL_CHECK_ERROR(); + for (size_t img_smp_index = 0; img_smp_index < SG_MAX_IMAGE_SAMPLER_PAIRS; img_smp_index++) { + const _sg_shader_image_sampler_t* img_smp = &shd->cmn.image_samplers[img_smp_index]; + if (img_smp->stage == SG_SHADERSTAGE_NONE) { + continue; + } + const int8_t gl_tex_slot = (GLint)shd->gl.tex_slot[img_smp_index]; + if (gl_tex_slot != -1) { + SOKOL_ASSERT(img_smp->image_slot < SG_MAX_IMAGE_BINDSLOTS); + SOKOL_ASSERT(img_smp->sampler_slot < SG_MAX_SAMPLER_BINDSLOTS); + const _sg_image_t* img = bnd->imgs[img_smp->image_slot]; + const _sg_sampler_t* smp = bnd->smps[img_smp->sampler_slot]; + SOKOL_ASSERT(img); + SOKOL_ASSERT(smp); + const GLenum gl_tgt = img->gl.target; + const GLuint gl_tex = img->gl.tex[img->cmn.active_slot]; + const GLuint gl_smp = smp->gl.smp; + _sg_gl_cache_bind_texture_sampler(gl_tex_slot, gl_tgt, gl_tex, gl_smp); + } + } + _SG_GL_CHECK_ERROR(); + + // bind storage buffers + for (size_t sbuf_index = 0; sbuf_index < SG_MAX_STORAGEBUFFER_BINDSLOTS; sbuf_index++) { + if (shd->cmn.storage_buffers[sbuf_index].stage == SG_SHADERSTAGE_NONE) { + continue; + } + const _sg_buffer_t* sbuf = bnd->sbufs[sbuf_index]; + const uint8_t binding = shd->gl.sbuf_binding[sbuf_index]; + GLuint gl_sbuf = sbuf->gl.buf[sbuf->cmn.active_slot]; + _sg_gl_cache_bind_storage_buffer(binding, gl_sbuf); + } + _SG_GL_CHECK_ERROR(); + + // if compute-pipeline, early out here + if (bnd->pip->cmn.is_compute) { + return true; + } + + // index buffer (can be 0) + const GLuint gl_ib = bnd->ib ? bnd->ib->gl.buf[bnd->ib->cmn.active_slot] : 0; + _sg_gl_cache_bind_buffer(GL_ELEMENT_ARRAY_BUFFER, gl_ib); + _sg.gl.cache.cur_ib_offset = bnd->ib_offset; + + // vertex attributes + for (GLuint attr_index = 0; attr_index < (GLuint)_sg.limits.max_vertex_attrs; attr_index++) { + _sg_gl_attr_t* attr = &bnd->pip->gl.attrs[attr_index]; + _sg_gl_cache_attr_t* cache_attr = &_sg.gl.cache.attrs[attr_index]; + bool cache_attr_dirty = false; + int vb_offset = 0; + GLuint gl_vb = 0; + if (attr->vb_index >= 0) { + // attribute is enabled + SOKOL_ASSERT(attr->vb_index < SG_MAX_VERTEXBUFFER_BINDSLOTS); + _sg_buffer_t* vb = bnd->vbs[attr->vb_index]; + SOKOL_ASSERT(vb); + gl_vb = vb->gl.buf[vb->cmn.active_slot]; + vb_offset = bnd->vb_offsets[attr->vb_index] + attr->offset; + if ((gl_vb != cache_attr->gl_vbuf) || + (attr->size != cache_attr->gl_attr.size) || + (attr->type != cache_attr->gl_attr.type) || + (attr->normalized != cache_attr->gl_attr.normalized) || + (attr->base_type != cache_attr->gl_attr.base_type) || + (attr->stride != cache_attr->gl_attr.stride) || + (vb_offset != cache_attr->gl_attr.offset) || + (cache_attr->gl_attr.divisor != attr->divisor)) + { + _sg_gl_cache_bind_buffer(GL_ARRAY_BUFFER, gl_vb); + if (attr->base_type == SG_SHADERATTRBASETYPE_FLOAT) { + glVertexAttribPointer(attr_index, attr->size, attr->type, attr->normalized, attr->stride, (const GLvoid*)(GLintptr)vb_offset); + } else { + glVertexAttribIPointer(attr_index, attr->size, attr->type, attr->stride, (const GLvoid*)(GLintptr)vb_offset); + } + _sg_stats_add(gl.num_vertex_attrib_pointer, 1); + glVertexAttribDivisor(attr_index, (GLuint)attr->divisor); + _sg_stats_add(gl.num_vertex_attrib_divisor, 1); + cache_attr_dirty = true; + } + if (cache_attr->gl_attr.vb_index == -1) { + glEnableVertexAttribArray(attr_index); + _sg_stats_add(gl.num_enable_vertex_attrib_array, 1); + cache_attr_dirty = true; + } + } else { + // attribute is disabled + if (cache_attr->gl_attr.vb_index != -1) { + glDisableVertexAttribArray(attr_index); + _sg_stats_add(gl.num_disable_vertex_attrib_array, 1); + cache_attr_dirty = true; + } + } + if (cache_attr_dirty) { + cache_attr->gl_attr = *attr; + cache_attr->gl_attr.offset = vb_offset; + cache_attr->gl_vbuf = gl_vb; + } + } + _SG_GL_CHECK_ERROR(); + return true; +} + +_SOKOL_PRIVATE void _sg_gl_apply_uniforms(int ub_slot, const sg_range* data) { + SOKOL_ASSERT(_sg.gl.cache.cur_pipeline); + SOKOL_ASSERT((ub_slot >= 0) && (ub_slot < SG_MAX_UNIFORMBLOCK_BINDSLOTS)); + const _sg_pipeline_t* pip = _sg.gl.cache.cur_pipeline; + SOKOL_ASSERT(pip && pip->shader); + SOKOL_ASSERT(pip->slot.id == _sg.gl.cache.cur_pipeline_id.id); + const _sg_shader_t* shd = pip->shader; + SOKOL_ASSERT(shd->slot.id == pip->cmn.shader_id.id); + SOKOL_ASSERT(SG_SHADERSTAGE_NONE != shd->cmn.uniform_blocks[ub_slot].stage); + SOKOL_ASSERT(data->size == shd->cmn.uniform_blocks[ub_slot].size); + const _sg_gl_uniform_block_t* gl_ub = &shd->gl.uniform_blocks[ub_slot]; + for (int u_index = 0; u_index < gl_ub->num_uniforms; u_index++) { + const _sg_gl_uniform_t* u = &gl_ub->uniforms[u_index]; + SOKOL_ASSERT(u->type != SG_UNIFORMTYPE_INVALID); + if (u->gl_loc == -1) { + continue; + } + _sg_stats_add(gl.num_uniform, 1); + GLfloat* fptr = (GLfloat*) (((uint8_t*)data->ptr) + u->offset); + GLint* iptr = (GLint*) (((uint8_t*)data->ptr) + u->offset); + switch (u->type) { + case SG_UNIFORMTYPE_INVALID: + break; + case SG_UNIFORMTYPE_FLOAT: + glUniform1fv(u->gl_loc, u->count, fptr); + break; + case SG_UNIFORMTYPE_FLOAT2: + glUniform2fv(u->gl_loc, u->count, fptr); + break; + case SG_UNIFORMTYPE_FLOAT3: + glUniform3fv(u->gl_loc, u->count, fptr); + break; + case SG_UNIFORMTYPE_FLOAT4: + glUniform4fv(u->gl_loc, u->count, fptr); + break; + case SG_UNIFORMTYPE_INT: + glUniform1iv(u->gl_loc, u->count, iptr); + break; + case SG_UNIFORMTYPE_INT2: + glUniform2iv(u->gl_loc, u->count, iptr); + break; + case SG_UNIFORMTYPE_INT3: + glUniform3iv(u->gl_loc, u->count, iptr); + break; + case SG_UNIFORMTYPE_INT4: + glUniform4iv(u->gl_loc, u->count, iptr); + break; + case SG_UNIFORMTYPE_MAT4: + glUniformMatrix4fv(u->gl_loc, u->count, GL_FALSE, fptr); + break; + default: + SOKOL_UNREACHABLE; + break; + } + } +} + +_SOKOL_PRIVATE void _sg_gl_draw(int base_element, int num_elements, int num_instances) { + SOKOL_ASSERT(_sg.gl.cache.cur_pipeline); + const GLenum i_type = _sg.gl.cache.cur_index_type; + const GLenum p_type = _sg.gl.cache.cur_primitive_type; + const bool use_instanced_draw = (num_instances > 1) || (_sg.gl.cache.cur_pipeline->cmn.use_instanced_draw); + if (0 != i_type) { + // indexed rendering + const int i_size = (i_type == GL_UNSIGNED_SHORT) ? 2 : 4; + const int ib_offset = _sg.gl.cache.cur_ib_offset; + const GLvoid* indices = (const GLvoid*)(GLintptr)(base_element*i_size+ib_offset); + if (use_instanced_draw) { + glDrawElementsInstanced(p_type, num_elements, i_type, indices, num_instances); + } else { + glDrawElements(p_type, num_elements, i_type, indices); + } + } else { + // non-indexed rendering + if (use_instanced_draw) { + glDrawArraysInstanced(p_type, base_element, num_elements, num_instances); + } else { + glDrawArrays(p_type, base_element, num_elements); + } + } +} + +_SOKOL_PRIVATE void _sg_gl_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) { + #if defined(_SOKOL_GL_HAS_COMPUTE) + if (!_sg.features.compute) { + return; + } + glDispatchCompute((GLuint)num_groups_x, (GLuint)num_groups_y, (GLuint)num_groups_z); + #else + (void)num_groups_x; (void)num_groups_y; (void)num_groups_z; + #endif +} + +_SOKOL_PRIVATE void _sg_gl_commit(void) { + // "soft" clear bindings (only those that are actually bound) + _sg_gl_cache_clear_buffer_bindings(false); + _sg_gl_cache_clear_texture_sampler_bindings(false); +} + +_SOKOL_PRIVATE void _sg_gl_update_buffer(_sg_buffer_t* buf, const sg_range* data) { + SOKOL_ASSERT(buf && data && data->ptr && (data->size > 0)); + // only one update per buffer per frame allowed + if (++buf->cmn.active_slot >= buf->cmn.num_slots) { + buf->cmn.active_slot = 0; + } + GLenum gl_tgt = _sg_gl_buffer_target(buf->cmn.type); + SOKOL_ASSERT(buf->cmn.active_slot < SG_NUM_INFLIGHT_FRAMES); + GLuint gl_buf = buf->gl.buf[buf->cmn.active_slot]; + SOKOL_ASSERT(gl_buf); + _SG_GL_CHECK_ERROR(); + _sg_gl_cache_store_buffer_binding(gl_tgt); + _sg_gl_cache_bind_buffer(gl_tgt, gl_buf); + glBufferSubData(gl_tgt, 0, (GLsizeiptr)data->size, data->ptr); + _sg_gl_cache_restore_buffer_binding(gl_tgt); + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE void _sg_gl_append_buffer(_sg_buffer_t* buf, const sg_range* data, bool new_frame) { + SOKOL_ASSERT(buf && data && data->ptr && (data->size > 0)); + if (new_frame) { + if (++buf->cmn.active_slot >= buf->cmn.num_slots) { + buf->cmn.active_slot = 0; + } + } + GLenum gl_tgt = _sg_gl_buffer_target(buf->cmn.type); + SOKOL_ASSERT(buf->cmn.active_slot < SG_NUM_INFLIGHT_FRAMES); + GLuint gl_buf = buf->gl.buf[buf->cmn.active_slot]; + SOKOL_ASSERT(gl_buf); + _SG_GL_CHECK_ERROR(); + _sg_gl_cache_store_buffer_binding(gl_tgt); + _sg_gl_cache_bind_buffer(gl_tgt, gl_buf); + glBufferSubData(gl_tgt, buf->cmn.append_pos, (GLsizeiptr)data->size, data->ptr); + _sg_gl_cache_restore_buffer_binding(gl_tgt); + _SG_GL_CHECK_ERROR(); +} + +_SOKOL_PRIVATE void _sg_gl_update_image(_sg_image_t* img, const sg_image_data* data) { + SOKOL_ASSERT(img && data); + // only one update per image per frame allowed + if (++img->cmn.active_slot >= img->cmn.num_slots) { + img->cmn.active_slot = 0; + } + SOKOL_ASSERT(img->cmn.active_slot < SG_NUM_INFLIGHT_FRAMES); + SOKOL_ASSERT(0 != img->gl.tex[img->cmn.active_slot]); + _sg_gl_cache_store_texture_sampler_binding(0); + _sg_gl_cache_bind_texture_sampler(0, img->gl.target, img->gl.tex[img->cmn.active_slot], 0); + const GLenum gl_img_format = _sg_gl_teximage_format(img->cmn.pixel_format); + const GLenum gl_img_type = _sg_gl_teximage_type(img->cmn.pixel_format); + const int num_faces = img->cmn.type == SG_IMAGETYPE_CUBE ? 6 : 1; + const int num_mips = img->cmn.num_mipmaps; + for (int face_index = 0; face_index < num_faces; face_index++) { + for (int mip_index = 0; mip_index < num_mips; mip_index++) { + GLenum gl_img_target = img->gl.target; + if (SG_IMAGETYPE_CUBE == img->cmn.type) { + gl_img_target = _sg_gl_cubeface_target(face_index); + } + const GLvoid* data_ptr = data->subimage[face_index][mip_index].ptr; + int mip_width = _sg_miplevel_dim(img->cmn.width, mip_index); + int mip_height = _sg_miplevel_dim(img->cmn.height, mip_index); + if ((SG_IMAGETYPE_2D == img->cmn.type) || (SG_IMAGETYPE_CUBE == img->cmn.type)) { + glTexSubImage2D(gl_img_target, mip_index, + 0, 0, + mip_width, mip_height, + gl_img_format, gl_img_type, + data_ptr); + } else if ((SG_IMAGETYPE_3D == img->cmn.type) || (SG_IMAGETYPE_ARRAY == img->cmn.type)) { + int mip_depth = img->cmn.num_slices; + if (SG_IMAGETYPE_3D == img->cmn.type) { + mip_depth = _sg_miplevel_dim(img->cmn.num_slices, mip_index); + } + glTexSubImage3D(gl_img_target, mip_index, + 0, 0, 0, + mip_width, mip_height, mip_depth, + gl_img_format, gl_img_type, + data_ptr); + + } + } + } + _sg_gl_cache_restore_texture_sampler_binding(0); +} + +// ██████ ██████ ██████ ██ ██ ██████ █████ ██████ ██ ██ ███████ ███ ██ ██████ +// ██ ██ ██ ██ ██ ███ ███ ██ ██ ██ ██ ██ ██ ██ ██ ████ ██ ██ ██ +// ██ ██ █████ ██ ██ ██ ██ ██████ ███████ ██ █████ █████ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██████ ██████ ██████ ██ ██ ██████ ██ ██ ██████ ██ ██ ███████ ██ ████ ██████ +// +// >>d3d11 backend +#elif defined(SOKOL_D3D11) + +#if defined(__cplusplus) +#define _sg_d3d11_AddRef(self) (self)->AddRef() +#else +#define _sg_d3d11_AddRef(self) (self)->lpVtbl->AddRef(self) +#endif + +#if defined(__cplusplus) +#define _sg_d3d11_Release(self) (self)->Release() +#else +#define _sg_d3d11_Release(self) (self)->lpVtbl->Release(self) +#endif + +// NOTE: This needs to be a macro since we can't use the polymorphism in C. It's called on many kinds of resources. +// NOTE: Based on microsoft docs, it's fine to call this with pData=NULL if DataSize is also zero. +#if defined(__cplusplus) +#define _sg_d3d11_SetPrivateData(self, guid, DataSize, pData) (self)->SetPrivateData(guid, DataSize, pData) +#else +#define _sg_d3d11_SetPrivateData(self, guid, DataSize, pData) (self)->lpVtbl->SetPrivateData(self, guid, DataSize, pData) +#endif + +#if defined(__cplusplus) +#define _sg_win32_refguid(guid) guid +#else +#define _sg_win32_refguid(guid) &guid +#endif + +static const GUID _sg_d3d11_WKPDID_D3DDebugObjectName = { 0x429b8c22,0x9188,0x4b0c, {0x87,0x42,0xac,0xb0,0xbf,0x85,0xc2,0x00} }; + +#if defined(SOKOL_DEBUG) +#define _sg_d3d11_setlabel(self, label) _sg_d3d11_SetPrivateData(self, _sg_win32_refguid(_sg_d3d11_WKPDID_D3DDebugObjectName), label ? (UINT)strlen(label) : 0, label) +#else +#define _sg_d3d11_setlabel(self, label) +#endif + + +//-- D3D11 C/C++ wrappers ------------------------------------------------------ +static inline HRESULT _sg_d3d11_CheckFormatSupport(ID3D11Device* self, DXGI_FORMAT Format, UINT* pFormatSupport) { + #if defined(__cplusplus) + return self->CheckFormatSupport(Format, pFormatSupport); + #else + return self->lpVtbl->CheckFormatSupport(self, Format, pFormatSupport); + #endif +} + +static inline void _sg_d3d11_OMSetRenderTargets(ID3D11DeviceContext* self, UINT NumViews, ID3D11RenderTargetView* const* ppRenderTargetViews, ID3D11DepthStencilView *pDepthStencilView) { + #if defined(__cplusplus) + self->OMSetRenderTargets(NumViews, ppRenderTargetViews, pDepthStencilView); + #else + self->lpVtbl->OMSetRenderTargets(self, NumViews, ppRenderTargetViews, pDepthStencilView); + #endif +} + +static inline void _sg_d3d11_RSSetState(ID3D11DeviceContext* self, ID3D11RasterizerState* pRasterizerState) { + #if defined(__cplusplus) + self->RSSetState(pRasterizerState); + #else + self->lpVtbl->RSSetState(self, pRasterizerState); + #endif +} + +static inline void _sg_d3d11_OMSetDepthStencilState(ID3D11DeviceContext* self, ID3D11DepthStencilState* pDepthStencilState, UINT StencilRef) { + #if defined(__cplusplus) + self->OMSetDepthStencilState(pDepthStencilState, StencilRef); + #else + self->lpVtbl->OMSetDepthStencilState(self, pDepthStencilState, StencilRef); + #endif +} + +static inline void _sg_d3d11_OMSetBlendState(ID3D11DeviceContext* self, ID3D11BlendState* pBlendState, const FLOAT BlendFactor[4], UINT SampleMask) { + #if defined(__cplusplus) + self->OMSetBlendState(pBlendState, BlendFactor, SampleMask); + #else + self->lpVtbl->OMSetBlendState(self, pBlendState, BlendFactor, SampleMask); + #endif +} + +static inline void _sg_d3d11_IASetVertexBuffers(ID3D11DeviceContext* self, UINT StartSlot, UINT NumBuffers, ID3D11Buffer* const* ppVertexBuffers, const UINT* pStrides, const UINT* pOffsets) { + #if defined(__cplusplus) + self->IASetVertexBuffers(StartSlot, NumBuffers, ppVertexBuffers, pStrides, pOffsets); + #else + self->lpVtbl->IASetVertexBuffers(self, StartSlot, NumBuffers, ppVertexBuffers, pStrides, pOffsets); + #endif +} + +static inline void _sg_d3d11_IASetIndexBuffer(ID3D11DeviceContext* self, ID3D11Buffer* pIndexBuffer, DXGI_FORMAT Format, UINT Offset) { + #if defined(__cplusplus) + self->IASetIndexBuffer(pIndexBuffer, Format, Offset); + #else + self->lpVtbl->IASetIndexBuffer(self, pIndexBuffer, Format, Offset); + #endif +} + +static inline void _sg_d3d11_IASetInputLayout(ID3D11DeviceContext* self, ID3D11InputLayout* pInputLayout) { + #if defined(__cplusplus) + self->IASetInputLayout(pInputLayout); + #else + self->lpVtbl->IASetInputLayout(self, pInputLayout); + #endif +} + +static inline void _sg_d3d11_VSSetShader(ID3D11DeviceContext* self, ID3D11VertexShader* pVertexShader, ID3D11ClassInstance* const* ppClassInstances, UINT NumClassInstances) { + #if defined(__cplusplus) + self->VSSetShader(pVertexShader, ppClassInstances, NumClassInstances); + #else + self->lpVtbl->VSSetShader(self, pVertexShader, ppClassInstances, NumClassInstances); + #endif +} + +static inline void _sg_d3d11_PSSetShader(ID3D11DeviceContext* self, ID3D11PixelShader* pPixelShader, ID3D11ClassInstance* const* ppClassInstances, UINT NumClassInstances) { + #if defined(__cplusplus) + self->PSSetShader(pPixelShader, ppClassInstances, NumClassInstances); + #else + self->lpVtbl->PSSetShader(self, pPixelShader, ppClassInstances, NumClassInstances); + #endif +} + +static inline void _sg_d3d11_CSSetShader(ID3D11DeviceContext* self, ID3D11ComputeShader* pComputeShader, ID3D11ClassInstance* const* ppClassInstances, UINT NumClassInstances) { + #if defined(__cplusplus) + self->CSSetShader(pComputeShader, ppClassInstances, NumClassInstances); + #else + self->lpVtbl->CSSetShader(self, pComputeShader, ppClassInstances, NumClassInstances); + #endif +} + +static inline void _sg_d3d11_VSSetConstantBuffers(ID3D11DeviceContext* self, UINT StartSlot, UINT NumBuffers, ID3D11Buffer* const* ppConstantBuffers) { + #if defined(__cplusplus) + self->VSSetConstantBuffers(StartSlot, NumBuffers, ppConstantBuffers); + #else + self->lpVtbl->VSSetConstantBuffers(self, StartSlot, NumBuffers, ppConstantBuffers); + #endif +} + +static inline void _sg_d3d11_PSSetConstantBuffers(ID3D11DeviceContext* self, UINT StartSlot, UINT NumBuffers, ID3D11Buffer* const* ppConstantBuffers) { + #if defined(__cplusplus) + self->PSSetConstantBuffers(StartSlot, NumBuffers, ppConstantBuffers); + #else + self->lpVtbl->PSSetConstantBuffers(self, StartSlot, NumBuffers, ppConstantBuffers); + #endif +} + +static inline void _sg_d3d11_CSSetConstantBuffers(ID3D11DeviceContext* self, UINT StartSlot, UINT NumBuffers, ID3D11Buffer* const* ppConstantBuffers) { + #if defined(__cplusplus) + self->CSSetConstantBuffers(StartSlot, NumBuffers, ppConstantBuffers); + #else + self->lpVtbl->CSSetConstantBuffers(self, StartSlot, NumBuffers, ppConstantBuffers); + #endif +} + +static inline void _sg_d3d11_VSSetShaderResources(ID3D11DeviceContext* self, UINT StartSlot, UINT NumViews, ID3D11ShaderResourceView* const* ppShaderResourceViews) { + #if defined(__cplusplus) + self->VSSetShaderResources(StartSlot, NumViews, ppShaderResourceViews); + #else + self->lpVtbl->VSSetShaderResources(self, StartSlot, NumViews, ppShaderResourceViews); + #endif +} + +static inline void _sg_d3d11_PSSetShaderResources(ID3D11DeviceContext* self, UINT StartSlot, UINT NumViews, ID3D11ShaderResourceView* const* ppShaderResourceViews) { + #if defined(__cplusplus) + self->PSSetShaderResources(StartSlot, NumViews, ppShaderResourceViews); + #else + self->lpVtbl->PSSetShaderResources(self, StartSlot, NumViews, ppShaderResourceViews); + #endif +} + +static inline void _sg_d3d11_CSSetShaderResources(ID3D11DeviceContext* self, UINT StartSlot, UINT NumViews, ID3D11ShaderResourceView* const* ppShaderResourceViews) { + #if defined(__cplusplus) + self->CSSetShaderResources(StartSlot, NumViews, ppShaderResourceViews); + #else + self->lpVtbl->CSSetShaderResources(self, StartSlot, NumViews, ppShaderResourceViews); + #endif +} + +static inline void _sg_d3d11_VSSetSamplers(ID3D11DeviceContext* self, UINT StartSlot, UINT NumSamplers, ID3D11SamplerState* const* ppSamplers) { + #if defined(__cplusplus) + self->VSSetSamplers(StartSlot, NumSamplers, ppSamplers); + #else + self->lpVtbl->VSSetSamplers(self, StartSlot, NumSamplers, ppSamplers); + #endif +} + +static inline void _sg_d3d11_PSSetSamplers(ID3D11DeviceContext* self, UINT StartSlot, UINT NumSamplers, ID3D11SamplerState* const* ppSamplers) { + #if defined(__cplusplus) + self->PSSetSamplers(StartSlot, NumSamplers, ppSamplers); + #else + self->lpVtbl->PSSetSamplers(self, StartSlot, NumSamplers, ppSamplers); + #endif +} + +static inline void _sg_d3d11_CSSetSamplers(ID3D11DeviceContext* self, UINT StartSlot, UINT NumSamplers, ID3D11SamplerState* const* ppSamplers) { + #if defined(__cplusplus) + self->CSSetSamplers(StartSlot, NumSamplers, ppSamplers); + #else + self->lpVtbl->CSSetSamplers(self, StartSlot, NumSamplers, ppSamplers); + #endif +} + +static inline void _sg_d3d11_CSSetUnorderedAccessViews(ID3D11DeviceContext* self, UINT StartSlot, UINT NumUAVs, ID3D11UnorderedAccessView* const* ppUnorderedAccessViews, const UINT* pUAVInitialCounts) { + #if defined(__cplusplus) + self->CSSetUnorderedAccessViews(StartSlot, NumUAVs, ppUnorderedAccessViews, pUAVInitialCounts); + #else + self->lpVtbl->CSSetUnorderedAccessViews(self, StartSlot, NumUAVs, ppUnorderedAccessViews, pUAVInitialCounts); + #endif +} + +static inline HRESULT _sg_d3d11_CreateBuffer(ID3D11Device* self, const D3D11_BUFFER_DESC* pDesc, const D3D11_SUBRESOURCE_DATA* pInitialData, ID3D11Buffer** ppBuffer) { + #if defined(__cplusplus) + return self->CreateBuffer(pDesc, pInitialData, ppBuffer); + #else + return self->lpVtbl->CreateBuffer(self, pDesc, pInitialData, ppBuffer); + #endif +} + +static inline HRESULT _sg_d3d11_CreateTexture2D(ID3D11Device* self, const D3D11_TEXTURE2D_DESC* pDesc, const D3D11_SUBRESOURCE_DATA* pInitialData, ID3D11Texture2D** ppTexture2D) { + #if defined(__cplusplus) + return self->CreateTexture2D(pDesc, pInitialData, ppTexture2D); + #else + return self->lpVtbl->CreateTexture2D(self, pDesc, pInitialData, ppTexture2D); + #endif +} + +static inline HRESULT _sg_d3d11_CreateShaderResourceView(ID3D11Device* self, ID3D11Resource* pResource, const D3D11_SHADER_RESOURCE_VIEW_DESC* pDesc, ID3D11ShaderResourceView** ppSRView) { + #if defined(__cplusplus) + return self->CreateShaderResourceView(pResource, pDesc, ppSRView); + #else + return self->lpVtbl->CreateShaderResourceView(self, pResource, pDesc, ppSRView); + #endif +} + +static inline HRESULT _sg_d3d11_CreateUnorderedAccessView(ID3D11Device* self, ID3D11Resource* pResource, const D3D11_UNORDERED_ACCESS_VIEW_DESC* pDesc, ID3D11UnorderedAccessView** ppUAVView) { + #if defined(__cplusplus) + return self->CreateUnorderedAccessView(pResource, pDesc, ppUAVView); + #else + return self->lpVtbl->CreateUnorderedAccessView(self, pResource, pDesc, ppUAVView); + #endif +} + +static inline void _sg_d3d11_GetResource(ID3D11View* self, ID3D11Resource** ppResource) { + #if defined(__cplusplus) + self->GetResource(ppResource); + #else + self->lpVtbl->GetResource(self, ppResource); + #endif +} + +static inline HRESULT _sg_d3d11_CreateTexture3D(ID3D11Device* self, const D3D11_TEXTURE3D_DESC* pDesc, const D3D11_SUBRESOURCE_DATA* pInitialData, ID3D11Texture3D** ppTexture3D) { + #if defined(__cplusplus) + return self->CreateTexture3D(pDesc, pInitialData, ppTexture3D); + #else + return self->lpVtbl->CreateTexture3D(self, pDesc, pInitialData, ppTexture3D); + #endif +} + +static inline HRESULT _sg_d3d11_CreateSamplerState(ID3D11Device* self, const D3D11_SAMPLER_DESC* pSamplerDesc, ID3D11SamplerState** ppSamplerState) { + #if defined(__cplusplus) + return self->CreateSamplerState(pSamplerDesc, ppSamplerState); + #else + return self->lpVtbl->CreateSamplerState(self, pSamplerDesc, ppSamplerState); + #endif +} + +static inline LPVOID _sg_d3d11_GetBufferPointer(ID3D10Blob* self) { + #if defined(__cplusplus) + return self->GetBufferPointer(); + #else + return self->lpVtbl->GetBufferPointer(self); + #endif +} + +static inline SIZE_T _sg_d3d11_GetBufferSize(ID3D10Blob* self) { + #if defined(__cplusplus) + return self->GetBufferSize(); + #else + return self->lpVtbl->GetBufferSize(self); + #endif +} + +static inline HRESULT _sg_d3d11_CreateVertexShader(ID3D11Device* self, const void* pShaderBytecode, SIZE_T BytecodeLength, ID3D11ClassLinkage* pClassLinkage, ID3D11VertexShader** ppVertexShader) { + #if defined(__cplusplus) + return self->CreateVertexShader(pShaderBytecode, BytecodeLength, pClassLinkage, ppVertexShader); + #else + return self->lpVtbl->CreateVertexShader(self, pShaderBytecode, BytecodeLength, pClassLinkage, ppVertexShader); + #endif +} + +static inline HRESULT _sg_d3d11_CreatePixelShader(ID3D11Device* self, const void* pShaderBytecode, SIZE_T BytecodeLength, ID3D11ClassLinkage* pClassLinkage, ID3D11PixelShader** ppPixelShader) { + #if defined(__cplusplus) + return self->CreatePixelShader(pShaderBytecode, BytecodeLength, pClassLinkage, ppPixelShader); + #else + return self->lpVtbl->CreatePixelShader(self, pShaderBytecode, BytecodeLength, pClassLinkage, ppPixelShader); + #endif +} + +static inline HRESULT _sg_d3d11_CreateComputeShader(ID3D11Device* self, const void* pShaderBytecode, SIZE_T BytecodeLength, ID3D11ClassLinkage* pClassLinkage, ID3D11ComputeShader** ppComputeShader) { + #if defined(__cplusplus) + return self->CreateComputeShader(pShaderBytecode, BytecodeLength, pClassLinkage, ppComputeShader); + #else + return self->lpVtbl->CreateComputeShader(self, pShaderBytecode, BytecodeLength, pClassLinkage, ppComputeShader); + #endif +} + +static inline HRESULT _sg_d3d11_CreateInputLayout(ID3D11Device* self, const D3D11_INPUT_ELEMENT_DESC* pInputElementDescs, UINT NumElements, const void* pShaderBytecodeWithInputSignature, SIZE_T BytecodeLength, ID3D11InputLayout **ppInputLayout) { + #if defined(__cplusplus) + return self->CreateInputLayout(pInputElementDescs, NumElements, pShaderBytecodeWithInputSignature, BytecodeLength, ppInputLayout); + #else + return self->lpVtbl->CreateInputLayout(self, pInputElementDescs, NumElements, pShaderBytecodeWithInputSignature, BytecodeLength, ppInputLayout); + #endif +} + +static inline HRESULT _sg_d3d11_CreateRasterizerState(ID3D11Device* self, const D3D11_RASTERIZER_DESC* pRasterizerDesc, ID3D11RasterizerState** ppRasterizerState) { + #if defined(__cplusplus) + return self->CreateRasterizerState(pRasterizerDesc, ppRasterizerState); + #else + return self->lpVtbl->CreateRasterizerState(self, pRasterizerDesc, ppRasterizerState); + #endif +} + +static inline HRESULT _sg_d3d11_CreateDepthStencilState(ID3D11Device* self, const D3D11_DEPTH_STENCIL_DESC* pDepthStencilDesc, ID3D11DepthStencilState** ppDepthStencilState) { + #if defined(__cplusplus) + return self->CreateDepthStencilState(pDepthStencilDesc, ppDepthStencilState); + #else + return self->lpVtbl->CreateDepthStencilState(self, pDepthStencilDesc, ppDepthStencilState); + #endif +} + +static inline HRESULT _sg_d3d11_CreateBlendState(ID3D11Device* self, const D3D11_BLEND_DESC* pBlendStateDesc, ID3D11BlendState** ppBlendState) { + #if defined(__cplusplus) + return self->CreateBlendState(pBlendStateDesc, ppBlendState); + #else + return self->lpVtbl->CreateBlendState(self, pBlendStateDesc, ppBlendState); + #endif +} + +static inline HRESULT _sg_d3d11_CreateRenderTargetView(ID3D11Device* self, ID3D11Resource *pResource, const D3D11_RENDER_TARGET_VIEW_DESC* pDesc, ID3D11RenderTargetView** ppRTView) { + #if defined(__cplusplus) + return self->CreateRenderTargetView(pResource, pDesc, ppRTView); + #else + return self->lpVtbl->CreateRenderTargetView(self, pResource, pDesc, ppRTView); + #endif +} + +static inline HRESULT _sg_d3d11_CreateDepthStencilView(ID3D11Device* self, ID3D11Resource* pResource, const D3D11_DEPTH_STENCIL_VIEW_DESC* pDesc, ID3D11DepthStencilView** ppDepthStencilView) { + #if defined(__cplusplus) + return self->CreateDepthStencilView(pResource, pDesc, ppDepthStencilView); + #else + return self->lpVtbl->CreateDepthStencilView(self, pResource, pDesc, ppDepthStencilView); + #endif +} + +static inline void _sg_d3d11_RSSetViewports(ID3D11DeviceContext* self, UINT NumViewports, const D3D11_VIEWPORT* pViewports) { + #if defined(__cplusplus) + self->RSSetViewports(NumViewports, pViewports); + #else + self->lpVtbl->RSSetViewports(self, NumViewports, pViewports); + #endif +} + +static inline void _sg_d3d11_RSSetScissorRects(ID3D11DeviceContext* self, UINT NumRects, const D3D11_RECT* pRects) { + #if defined(__cplusplus) + self->RSSetScissorRects(NumRects, pRects); + #else + self->lpVtbl->RSSetScissorRects(self, NumRects, pRects); + #endif +} + +static inline void _sg_d3d11_ClearRenderTargetView(ID3D11DeviceContext* self, ID3D11RenderTargetView* pRenderTargetView, const FLOAT ColorRGBA[4]) { + #if defined(__cplusplus) + self->ClearRenderTargetView(pRenderTargetView, ColorRGBA); + #else + self->lpVtbl->ClearRenderTargetView(self, pRenderTargetView, ColorRGBA); + #endif +} + +static inline void _sg_d3d11_ClearDepthStencilView(ID3D11DeviceContext* self, ID3D11DepthStencilView* pDepthStencilView, UINT ClearFlags, FLOAT Depth, UINT8 Stencil) { + #if defined(__cplusplus) + self->ClearDepthStencilView(pDepthStencilView, ClearFlags, Depth, Stencil); + #else + self->lpVtbl->ClearDepthStencilView(self, pDepthStencilView, ClearFlags, Depth, Stencil); + #endif +} + +static inline void _sg_d3d11_ResolveSubresource(ID3D11DeviceContext* self, ID3D11Resource* pDstResource, UINT DstSubresource, ID3D11Resource* pSrcResource, UINT SrcSubresource, DXGI_FORMAT Format) { + #if defined(__cplusplus) + self->ResolveSubresource(pDstResource, DstSubresource, pSrcResource, SrcSubresource, Format); + #else + self->lpVtbl->ResolveSubresource(self, pDstResource, DstSubresource, pSrcResource, SrcSubresource, Format); + #endif +} + +static inline void _sg_d3d11_IASetPrimitiveTopology(ID3D11DeviceContext* self, D3D11_PRIMITIVE_TOPOLOGY Topology) { + #if defined(__cplusplus) + self->IASetPrimitiveTopology(Topology); + #else + self->lpVtbl->IASetPrimitiveTopology(self, Topology); + #endif +} + +static inline void _sg_d3d11_UpdateSubresource(ID3D11DeviceContext* self, ID3D11Resource* pDstResource, UINT DstSubresource, const D3D11_BOX* pDstBox, const void* pSrcData, UINT SrcRowPitch, UINT SrcDepthPitch) { + #if defined(__cplusplus) + self->UpdateSubresource(pDstResource, DstSubresource, pDstBox, pSrcData, SrcRowPitch, SrcDepthPitch); + #else + self->lpVtbl->UpdateSubresource(self, pDstResource, DstSubresource, pDstBox, pSrcData, SrcRowPitch, SrcDepthPitch); + #endif +} + +static inline void _sg_d3d11_DrawIndexed(ID3D11DeviceContext* self, UINT IndexCount, UINT StartIndexLocation, INT BaseVertexLocation) { + #if defined(__cplusplus) + self->DrawIndexed(IndexCount, StartIndexLocation, BaseVertexLocation); + #else + self->lpVtbl->DrawIndexed(self, IndexCount, StartIndexLocation, BaseVertexLocation); + #endif +} + +static inline void _sg_d3d11_DrawIndexedInstanced(ID3D11DeviceContext* self, UINT IndexCountPerInstance, UINT InstanceCount, UINT StartIndexLocation, INT BaseVertexLocation, UINT StartInstanceLocation) { + #if defined(__cplusplus) + self->DrawIndexedInstanced(IndexCountPerInstance, InstanceCount, StartIndexLocation, BaseVertexLocation, StartInstanceLocation); + #else + self->lpVtbl->DrawIndexedInstanced(self, IndexCountPerInstance, InstanceCount, StartIndexLocation, BaseVertexLocation, StartInstanceLocation); + #endif +} + +static inline void _sg_d3d11_Draw(ID3D11DeviceContext* self, UINT VertexCount, UINT StartVertexLocation) { + #if defined(__cplusplus) + self->Draw(VertexCount, StartVertexLocation); + #else + self->lpVtbl->Draw(self, VertexCount, StartVertexLocation); + #endif +} + +static inline void _sg_d3d11_DrawInstanced(ID3D11DeviceContext* self, UINT VertexCountPerInstance, UINT InstanceCount, UINT StartVertexLocation, UINT StartInstanceLocation) { + #if defined(__cplusplus) + self->DrawInstanced(VertexCountPerInstance, InstanceCount, StartVertexLocation, StartInstanceLocation); + #else + self->lpVtbl->DrawInstanced(self, VertexCountPerInstance, InstanceCount, StartVertexLocation, StartInstanceLocation); + #endif +} + +static inline void _sg_d3d11_Dispatch(ID3D11DeviceContext* self, UINT ThreadGroupCountX, UINT ThreadGroupCountY, UINT ThreadGroupCountZ) { + #if defined(__cplusplus) + self->Dispatch(ThreadGroupCountX, ThreadGroupCountY, ThreadGroupCountZ); + #else + self->lpVtbl->Dispatch(self, ThreadGroupCountX, ThreadGroupCountY, ThreadGroupCountZ); + #endif +} + +static inline HRESULT _sg_d3d11_Map(ID3D11DeviceContext* self, ID3D11Resource* pResource, UINT Subresource, D3D11_MAP MapType, UINT MapFlags, D3D11_MAPPED_SUBRESOURCE* pMappedResource) { + #if defined(__cplusplus) + return self->Map(pResource, Subresource, MapType, MapFlags, pMappedResource); + #else + return self->lpVtbl->Map(self, pResource, Subresource, MapType, MapFlags, pMappedResource); + #endif +} + +static inline void _sg_d3d11_Unmap(ID3D11DeviceContext* self, ID3D11Resource* pResource, UINT Subresource) { + #if defined(__cplusplus) + self->Unmap(pResource, Subresource); + #else + self->lpVtbl->Unmap(self, pResource, Subresource); + #endif +} + +static inline void _sg_d3d11_ClearState(ID3D11DeviceContext* self) { + #if defined(__cplusplus) + self->ClearState(); + #else + self->lpVtbl->ClearState(self); + #endif +} + +//-- enum translation functions ------------------------------------------------ +_SOKOL_PRIVATE D3D11_USAGE _sg_d3d11_usage(sg_usage usg) { + switch (usg) { + case SG_USAGE_IMMUTABLE: + return D3D11_USAGE_IMMUTABLE; + case SG_USAGE_DYNAMIC: + case SG_USAGE_STREAM: + return D3D11_USAGE_DYNAMIC; + default: + SOKOL_UNREACHABLE; + return (D3D11_USAGE) 0; + } +} + +_SOKOL_PRIVATE D3D11_USAGE _sg_d3d11_buffer_usage(sg_usage usg, sg_buffer_type type) { + switch (usg) { + case SG_USAGE_IMMUTABLE: + if (type == SG_BUFFERTYPE_STORAGEBUFFER) { + return D3D11_USAGE_DEFAULT; + } else { + return D3D11_USAGE_IMMUTABLE; + } + case SG_USAGE_DYNAMIC: + case SG_USAGE_STREAM: + return D3D11_USAGE_DYNAMIC; + default: + SOKOL_UNREACHABLE; + return (D3D11_USAGE) 0; + } +} + +_SOKOL_PRIVATE UINT _sg_d3d11_buffer_bind_flags(sg_usage usg, sg_buffer_type t) { + switch (t) { + case SG_BUFFERTYPE_VERTEXBUFFER: + return D3D11_BIND_VERTEX_BUFFER; + case SG_BUFFERTYPE_INDEXBUFFER: + return D3D11_BIND_INDEX_BUFFER; + case SG_BUFFERTYPE_STORAGEBUFFER: + if (usg == SG_USAGE_IMMUTABLE) { + return D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_UNORDERED_ACCESS; + } else { + return D3D11_BIND_SHADER_RESOURCE; + } + default: + SOKOL_UNREACHABLE; + return 0; + } +} + +_SOKOL_PRIVATE UINT _sg_d3d11_buffer_misc_flags(sg_buffer_type t) { + switch (t) { + case SG_BUFFERTYPE_VERTEXBUFFER: + case SG_BUFFERTYPE_INDEXBUFFER: + return 0; + case SG_BUFFERTYPE_STORAGEBUFFER: + return D3D11_RESOURCE_MISC_BUFFER_ALLOW_RAW_VIEWS; + default: + SOKOL_UNREACHABLE; + return 0; + } +} + +_SOKOL_PRIVATE UINT _sg_d3d11_cpu_access_flags(sg_usage usg) { + switch (usg) { + case SG_USAGE_IMMUTABLE: + return 0; + case SG_USAGE_DYNAMIC: + case SG_USAGE_STREAM: + return D3D11_CPU_ACCESS_WRITE; + default: + SOKOL_UNREACHABLE; + return 0; + } +} + +_SOKOL_PRIVATE DXGI_FORMAT _sg_d3d11_texture_pixel_format(sg_pixel_format fmt) { + switch (fmt) { + case SG_PIXELFORMAT_R8: return DXGI_FORMAT_R8_UNORM; + case SG_PIXELFORMAT_R8SN: return DXGI_FORMAT_R8_SNORM; + case SG_PIXELFORMAT_R8UI: return DXGI_FORMAT_R8_UINT; + case SG_PIXELFORMAT_R8SI: return DXGI_FORMAT_R8_SINT; + case SG_PIXELFORMAT_R16: return DXGI_FORMAT_R16_UNORM; + case SG_PIXELFORMAT_R16SN: return DXGI_FORMAT_R16_SNORM; + case SG_PIXELFORMAT_R16UI: return DXGI_FORMAT_R16_UINT; + case SG_PIXELFORMAT_R16SI: return DXGI_FORMAT_R16_SINT; + case SG_PIXELFORMAT_R16F: return DXGI_FORMAT_R16_FLOAT; + case SG_PIXELFORMAT_RG8: return DXGI_FORMAT_R8G8_UNORM; + case SG_PIXELFORMAT_RG8SN: return DXGI_FORMAT_R8G8_SNORM; + case SG_PIXELFORMAT_RG8UI: return DXGI_FORMAT_R8G8_UINT; + case SG_PIXELFORMAT_RG8SI: return DXGI_FORMAT_R8G8_SINT; + case SG_PIXELFORMAT_R32UI: return DXGI_FORMAT_R32_UINT; + case SG_PIXELFORMAT_R32SI: return DXGI_FORMAT_R32_SINT; + case SG_PIXELFORMAT_R32F: return DXGI_FORMAT_R32_FLOAT; + case SG_PIXELFORMAT_RG16: return DXGI_FORMAT_R16G16_UNORM; + case SG_PIXELFORMAT_RG16SN: return DXGI_FORMAT_R16G16_SNORM; + case SG_PIXELFORMAT_RG16UI: return DXGI_FORMAT_R16G16_UINT; + case SG_PIXELFORMAT_RG16SI: return DXGI_FORMAT_R16G16_SINT; + case SG_PIXELFORMAT_RG16F: return DXGI_FORMAT_R16G16_FLOAT; + case SG_PIXELFORMAT_RGBA8: return DXGI_FORMAT_R8G8B8A8_UNORM; + case SG_PIXELFORMAT_SRGB8A8: return DXGI_FORMAT_R8G8B8A8_UNORM_SRGB; + case SG_PIXELFORMAT_RGBA8SN: return DXGI_FORMAT_R8G8B8A8_SNORM; + case SG_PIXELFORMAT_RGBA8UI: return DXGI_FORMAT_R8G8B8A8_UINT; + case SG_PIXELFORMAT_RGBA8SI: return DXGI_FORMAT_R8G8B8A8_SINT; + case SG_PIXELFORMAT_BGRA8: return DXGI_FORMAT_B8G8R8A8_UNORM; + case SG_PIXELFORMAT_RGB10A2: return DXGI_FORMAT_R10G10B10A2_UNORM; + case SG_PIXELFORMAT_RG11B10F: return DXGI_FORMAT_R11G11B10_FLOAT; + case SG_PIXELFORMAT_RGB9E5: return DXGI_FORMAT_R9G9B9E5_SHAREDEXP; + case SG_PIXELFORMAT_RG32UI: return DXGI_FORMAT_R32G32_UINT; + case SG_PIXELFORMAT_RG32SI: return DXGI_FORMAT_R32G32_SINT; + case SG_PIXELFORMAT_RG32F: return DXGI_FORMAT_R32G32_FLOAT; + case SG_PIXELFORMAT_RGBA16: return DXGI_FORMAT_R16G16B16A16_UNORM; + case SG_PIXELFORMAT_RGBA16SN: return DXGI_FORMAT_R16G16B16A16_SNORM; + case SG_PIXELFORMAT_RGBA16UI: return DXGI_FORMAT_R16G16B16A16_UINT; + case SG_PIXELFORMAT_RGBA16SI: return DXGI_FORMAT_R16G16B16A16_SINT; + case SG_PIXELFORMAT_RGBA16F: return DXGI_FORMAT_R16G16B16A16_FLOAT; + case SG_PIXELFORMAT_RGBA32UI: return DXGI_FORMAT_R32G32B32A32_UINT; + case SG_PIXELFORMAT_RGBA32SI: return DXGI_FORMAT_R32G32B32A32_SINT; + case SG_PIXELFORMAT_RGBA32F: return DXGI_FORMAT_R32G32B32A32_FLOAT; + case SG_PIXELFORMAT_DEPTH: return DXGI_FORMAT_R32_TYPELESS; + case SG_PIXELFORMAT_DEPTH_STENCIL: return DXGI_FORMAT_R24G8_TYPELESS; + case SG_PIXELFORMAT_BC1_RGBA: return DXGI_FORMAT_BC1_UNORM; + case SG_PIXELFORMAT_BC2_RGBA: return DXGI_FORMAT_BC2_UNORM; + case SG_PIXELFORMAT_BC3_RGBA: return DXGI_FORMAT_BC3_UNORM; + case SG_PIXELFORMAT_BC3_SRGBA: return DXGI_FORMAT_BC3_UNORM_SRGB; + case SG_PIXELFORMAT_BC4_R: return DXGI_FORMAT_BC4_UNORM; + case SG_PIXELFORMAT_BC4_RSN: return DXGI_FORMAT_BC4_SNORM; + case SG_PIXELFORMAT_BC5_RG: return DXGI_FORMAT_BC5_UNORM; + case SG_PIXELFORMAT_BC5_RGSN: return DXGI_FORMAT_BC5_SNORM; + case SG_PIXELFORMAT_BC6H_RGBF: return DXGI_FORMAT_BC6H_SF16; + case SG_PIXELFORMAT_BC6H_RGBUF: return DXGI_FORMAT_BC6H_UF16; + case SG_PIXELFORMAT_BC7_RGBA: return DXGI_FORMAT_BC7_UNORM; + case SG_PIXELFORMAT_BC7_SRGBA: return DXGI_FORMAT_BC7_UNORM_SRGB; + default: return DXGI_FORMAT_UNKNOWN; + }; +} + +_SOKOL_PRIVATE DXGI_FORMAT _sg_d3d11_srv_pixel_format(sg_pixel_format fmt) { + if (fmt == SG_PIXELFORMAT_DEPTH) { + return DXGI_FORMAT_R32_FLOAT; + } else if (fmt == SG_PIXELFORMAT_DEPTH_STENCIL) { + return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + } else { + return _sg_d3d11_texture_pixel_format(fmt); + } +} + +_SOKOL_PRIVATE DXGI_FORMAT _sg_d3d11_dsv_pixel_format(sg_pixel_format fmt) { + if (fmt == SG_PIXELFORMAT_DEPTH) { + return DXGI_FORMAT_D32_FLOAT; + } else if (fmt == SG_PIXELFORMAT_DEPTH_STENCIL) { + return DXGI_FORMAT_D24_UNORM_S8_UINT; + } else { + return _sg_d3d11_texture_pixel_format(fmt); + } +} + +_SOKOL_PRIVATE DXGI_FORMAT _sg_d3d11_rtv_pixel_format(sg_pixel_format fmt) { + if (fmt == SG_PIXELFORMAT_DEPTH) { + return DXGI_FORMAT_R32_FLOAT; + } else if (fmt == SG_PIXELFORMAT_DEPTH_STENCIL) { + return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + } else { + return _sg_d3d11_texture_pixel_format(fmt); + } +} + +_SOKOL_PRIVATE D3D11_PRIMITIVE_TOPOLOGY _sg_d3d11_primitive_topology(sg_primitive_type prim_type) { + switch (prim_type) { + case SG_PRIMITIVETYPE_POINTS: return D3D11_PRIMITIVE_TOPOLOGY_POINTLIST; + case SG_PRIMITIVETYPE_LINES: return D3D11_PRIMITIVE_TOPOLOGY_LINELIST; + case SG_PRIMITIVETYPE_LINE_STRIP: return D3D11_PRIMITIVE_TOPOLOGY_LINESTRIP; + case SG_PRIMITIVETYPE_TRIANGLES: return D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST; + case SG_PRIMITIVETYPE_TRIANGLE_STRIP: return D3D11_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP; + default: SOKOL_UNREACHABLE; return (D3D11_PRIMITIVE_TOPOLOGY) 0; + } +} + +_SOKOL_PRIVATE DXGI_FORMAT _sg_d3d11_index_format(sg_index_type index_type) { + switch (index_type) { + case SG_INDEXTYPE_NONE: return DXGI_FORMAT_UNKNOWN; + case SG_INDEXTYPE_UINT16: return DXGI_FORMAT_R16_UINT; + case SG_INDEXTYPE_UINT32: return DXGI_FORMAT_R32_UINT; + default: SOKOL_UNREACHABLE; return (DXGI_FORMAT) 0; + } +} + +_SOKOL_PRIVATE D3D11_FILTER _sg_d3d11_filter(sg_filter min_f, sg_filter mag_f, sg_filter mipmap_f, bool comparison, uint32_t max_anisotropy) { + uint32_t d3d11_filter = 0; + if (max_anisotropy > 1) { + // D3D11_FILTER_ANISOTROPIC = 0x55, + d3d11_filter |= 0x55; + } else { + // D3D11_FILTER_MIN_MAG_MIP_POINT = 0, + // D3D11_FILTER_MIN_MAG_POINT_MIP_LINEAR = 0x1, + // D3D11_FILTER_MIN_POINT_MAG_LINEAR_MIP_POINT = 0x4, + // D3D11_FILTER_MIN_POINT_MAG_MIP_LINEAR = 0x5, + // D3D11_FILTER_MIN_LINEAR_MAG_MIP_POINT = 0x10, + // D3D11_FILTER_MIN_LINEAR_MAG_POINT_MIP_LINEAR = 0x11, + // D3D11_FILTER_MIN_MAG_LINEAR_MIP_POINT = 0x14, + // D3D11_FILTER_MIN_MAG_MIP_LINEAR = 0x15, + if (mipmap_f == SG_FILTER_LINEAR) { + d3d11_filter |= 0x01; + } + if (mag_f == SG_FILTER_LINEAR) { + d3d11_filter |= 0x04; + } + if (min_f == SG_FILTER_LINEAR) { + d3d11_filter |= 0x10; + } + } + // D3D11_FILTER_COMPARISON_MIN_MAG_MIP_POINT = 0x80, + // D3D11_FILTER_COMPARISON_MIN_MAG_POINT_MIP_LINEAR = 0x81, + // D3D11_FILTER_COMPARISON_MIN_POINT_MAG_LINEAR_MIP_POINT = 0x84, + // D3D11_FILTER_COMPARISON_MIN_POINT_MAG_MIP_LINEAR = 0x85, + // D3D11_FILTER_COMPARISON_MIN_LINEAR_MAG_MIP_POINT = 0x90, + // D3D11_FILTER_COMPARISON_MIN_LINEAR_MAG_POINT_MIP_LINEAR = 0x91, + // D3D11_FILTER_COMPARISON_MIN_MAG_LINEAR_MIP_POINT = 0x94, + // D3D11_FILTER_COMPARISON_MIN_MAG_MIP_LINEAR = 0x95, + // D3D11_FILTER_COMPARISON_ANISOTROPIC = 0xd5, + if (comparison) { + d3d11_filter |= 0x80; + } + return (D3D11_FILTER)d3d11_filter; +} + +_SOKOL_PRIVATE D3D11_TEXTURE_ADDRESS_MODE _sg_d3d11_address_mode(sg_wrap m) { + switch (m) { + case SG_WRAP_REPEAT: return D3D11_TEXTURE_ADDRESS_WRAP; + case SG_WRAP_CLAMP_TO_EDGE: return D3D11_TEXTURE_ADDRESS_CLAMP; + case SG_WRAP_CLAMP_TO_BORDER: return D3D11_TEXTURE_ADDRESS_BORDER; + case SG_WRAP_MIRRORED_REPEAT: return D3D11_TEXTURE_ADDRESS_MIRROR; + default: SOKOL_UNREACHABLE; return (D3D11_TEXTURE_ADDRESS_MODE) 0; + } +} + +_SOKOL_PRIVATE DXGI_FORMAT _sg_d3d11_vertex_format(sg_vertex_format fmt) { + switch (fmt) { + case SG_VERTEXFORMAT_FLOAT: return DXGI_FORMAT_R32_FLOAT; + case SG_VERTEXFORMAT_FLOAT2: return DXGI_FORMAT_R32G32_FLOAT; + case SG_VERTEXFORMAT_FLOAT3: return DXGI_FORMAT_R32G32B32_FLOAT; + case SG_VERTEXFORMAT_FLOAT4: return DXGI_FORMAT_R32G32B32A32_FLOAT; + case SG_VERTEXFORMAT_INT: return DXGI_FORMAT_R32_SINT; + case SG_VERTEXFORMAT_INT2: return DXGI_FORMAT_R32G32_SINT; + case SG_VERTEXFORMAT_INT3: return DXGI_FORMAT_R32G32B32_SINT; + case SG_VERTEXFORMAT_INT4: return DXGI_FORMAT_R32G32B32A32_SINT; + case SG_VERTEXFORMAT_UINT: return DXGI_FORMAT_R32_UINT; + case SG_VERTEXFORMAT_UINT2: return DXGI_FORMAT_R32G32_UINT; + case SG_VERTEXFORMAT_UINT3: return DXGI_FORMAT_R32G32B32_UINT; + case SG_VERTEXFORMAT_UINT4: return DXGI_FORMAT_R32G32B32A32_UINT; + case SG_VERTEXFORMAT_BYTE4: return DXGI_FORMAT_R8G8B8A8_SINT; + case SG_VERTEXFORMAT_BYTE4N: return DXGI_FORMAT_R8G8B8A8_SNORM; + case SG_VERTEXFORMAT_UBYTE4: return DXGI_FORMAT_R8G8B8A8_UINT; + case SG_VERTEXFORMAT_UBYTE4N: return DXGI_FORMAT_R8G8B8A8_UNORM; + case SG_VERTEXFORMAT_SHORT2: return DXGI_FORMAT_R16G16_SINT; + case SG_VERTEXFORMAT_SHORT2N: return DXGI_FORMAT_R16G16_SNORM; + case SG_VERTEXFORMAT_USHORT2: return DXGI_FORMAT_R16G16_UINT; + case SG_VERTEXFORMAT_USHORT2N: return DXGI_FORMAT_R16G16_UNORM; + case SG_VERTEXFORMAT_SHORT4: return DXGI_FORMAT_R16G16B16A16_SINT; + case SG_VERTEXFORMAT_SHORT4N: return DXGI_FORMAT_R16G16B16A16_SNORM; + case SG_VERTEXFORMAT_USHORT4: return DXGI_FORMAT_R16G16B16A16_UINT; + case SG_VERTEXFORMAT_USHORT4N: return DXGI_FORMAT_R16G16B16A16_UNORM; + case SG_VERTEXFORMAT_UINT10_N2: return DXGI_FORMAT_R10G10B10A2_UNORM; + case SG_VERTEXFORMAT_HALF2: return DXGI_FORMAT_R16G16_FLOAT; + case SG_VERTEXFORMAT_HALF4: return DXGI_FORMAT_R16G16B16A16_FLOAT; + default: SOKOL_UNREACHABLE; return (DXGI_FORMAT) 0; + } +} + +_SOKOL_PRIVATE D3D11_INPUT_CLASSIFICATION _sg_d3d11_input_classification(sg_vertex_step step) { + switch (step) { + case SG_VERTEXSTEP_PER_VERTEX: return D3D11_INPUT_PER_VERTEX_DATA; + case SG_VERTEXSTEP_PER_INSTANCE: return D3D11_INPUT_PER_INSTANCE_DATA; + default: SOKOL_UNREACHABLE; return (D3D11_INPUT_CLASSIFICATION) 0; + } +} + +_SOKOL_PRIVATE D3D11_CULL_MODE _sg_d3d11_cull_mode(sg_cull_mode m) { + switch (m) { + case SG_CULLMODE_NONE: return D3D11_CULL_NONE; + case SG_CULLMODE_FRONT: return D3D11_CULL_FRONT; + case SG_CULLMODE_BACK: return D3D11_CULL_BACK; + default: SOKOL_UNREACHABLE; return (D3D11_CULL_MODE) 0; + } +} + +_SOKOL_PRIVATE D3D11_COMPARISON_FUNC _sg_d3d11_compare_func(sg_compare_func f) { + switch (f) { + case SG_COMPAREFUNC_NEVER: return D3D11_COMPARISON_NEVER; + case SG_COMPAREFUNC_LESS: return D3D11_COMPARISON_LESS; + case SG_COMPAREFUNC_EQUAL: return D3D11_COMPARISON_EQUAL; + case SG_COMPAREFUNC_LESS_EQUAL: return D3D11_COMPARISON_LESS_EQUAL; + case SG_COMPAREFUNC_GREATER: return D3D11_COMPARISON_GREATER; + case SG_COMPAREFUNC_NOT_EQUAL: return D3D11_COMPARISON_NOT_EQUAL; + case SG_COMPAREFUNC_GREATER_EQUAL: return D3D11_COMPARISON_GREATER_EQUAL; + case SG_COMPAREFUNC_ALWAYS: return D3D11_COMPARISON_ALWAYS; + default: SOKOL_UNREACHABLE; return (D3D11_COMPARISON_FUNC) 0; + } +} + +_SOKOL_PRIVATE D3D11_STENCIL_OP _sg_d3d11_stencil_op(sg_stencil_op op) { + switch (op) { + case SG_STENCILOP_KEEP: return D3D11_STENCIL_OP_KEEP; + case SG_STENCILOP_ZERO: return D3D11_STENCIL_OP_ZERO; + case SG_STENCILOP_REPLACE: return D3D11_STENCIL_OP_REPLACE; + case SG_STENCILOP_INCR_CLAMP: return D3D11_STENCIL_OP_INCR_SAT; + case SG_STENCILOP_DECR_CLAMP: return D3D11_STENCIL_OP_DECR_SAT; + case SG_STENCILOP_INVERT: return D3D11_STENCIL_OP_INVERT; + case SG_STENCILOP_INCR_WRAP: return D3D11_STENCIL_OP_INCR; + case SG_STENCILOP_DECR_WRAP: return D3D11_STENCIL_OP_DECR; + default: SOKOL_UNREACHABLE; return (D3D11_STENCIL_OP) 0; + } +} + +_SOKOL_PRIVATE D3D11_BLEND _sg_d3d11_blend_factor(sg_blend_factor f) { + switch (f) { + case SG_BLENDFACTOR_ZERO: return D3D11_BLEND_ZERO; + case SG_BLENDFACTOR_ONE: return D3D11_BLEND_ONE; + case SG_BLENDFACTOR_SRC_COLOR: return D3D11_BLEND_SRC_COLOR; + case SG_BLENDFACTOR_ONE_MINUS_SRC_COLOR: return D3D11_BLEND_INV_SRC_COLOR; + case SG_BLENDFACTOR_SRC_ALPHA: return D3D11_BLEND_SRC_ALPHA; + case SG_BLENDFACTOR_ONE_MINUS_SRC_ALPHA: return D3D11_BLEND_INV_SRC_ALPHA; + case SG_BLENDFACTOR_DST_COLOR: return D3D11_BLEND_DEST_COLOR; + case SG_BLENDFACTOR_ONE_MINUS_DST_COLOR: return D3D11_BLEND_INV_DEST_COLOR; + case SG_BLENDFACTOR_DST_ALPHA: return D3D11_BLEND_DEST_ALPHA; + case SG_BLENDFACTOR_ONE_MINUS_DST_ALPHA: return D3D11_BLEND_INV_DEST_ALPHA; + case SG_BLENDFACTOR_SRC_ALPHA_SATURATED: return D3D11_BLEND_SRC_ALPHA_SAT; + case SG_BLENDFACTOR_BLEND_COLOR: return D3D11_BLEND_BLEND_FACTOR; + case SG_BLENDFACTOR_ONE_MINUS_BLEND_COLOR: return D3D11_BLEND_INV_BLEND_FACTOR; + case SG_BLENDFACTOR_BLEND_ALPHA: return D3D11_BLEND_BLEND_FACTOR; + case SG_BLENDFACTOR_ONE_MINUS_BLEND_ALPHA: return D3D11_BLEND_INV_BLEND_FACTOR; + default: SOKOL_UNREACHABLE; return (D3D11_BLEND) 0; + } +} + +_SOKOL_PRIVATE D3D11_BLEND_OP _sg_d3d11_blend_op(sg_blend_op op) { + switch (op) { + case SG_BLENDOP_ADD: return D3D11_BLEND_OP_ADD; + case SG_BLENDOP_SUBTRACT: return D3D11_BLEND_OP_SUBTRACT; + case SG_BLENDOP_REVERSE_SUBTRACT: return D3D11_BLEND_OP_REV_SUBTRACT; + case SG_BLENDOP_MIN: return D3D11_BLEND_OP_MIN; + case SG_BLENDOP_MAX: return D3D11_BLEND_OP_MAX; + default: SOKOL_UNREACHABLE; return (D3D11_BLEND_OP) 0; + } +} + +_SOKOL_PRIVATE UINT8 _sg_d3d11_color_write_mask(sg_color_mask m) { + UINT8 res = 0; + if (m & SG_COLORMASK_R) { + res |= D3D11_COLOR_WRITE_ENABLE_RED; + } + if (m & SG_COLORMASK_G) { + res |= D3D11_COLOR_WRITE_ENABLE_GREEN; + } + if (m & SG_COLORMASK_B) { + res |= D3D11_COLOR_WRITE_ENABLE_BLUE; + } + if (m & SG_COLORMASK_A) { + res |= D3D11_COLOR_WRITE_ENABLE_ALPHA; + } + return res; +} + +_SOKOL_PRIVATE UINT _sg_d3d11_dxgi_fmt_caps(DXGI_FORMAT dxgi_fmt) { + UINT dxgi_fmt_caps = 0; + if (dxgi_fmt != DXGI_FORMAT_UNKNOWN) { + HRESULT hr = _sg_d3d11_CheckFormatSupport(_sg.d3d11.dev, dxgi_fmt, &dxgi_fmt_caps); + SOKOL_ASSERT(SUCCEEDED(hr) || (E_FAIL == hr)); + if (!SUCCEEDED(hr)) { + dxgi_fmt_caps = 0; + } + } + return dxgi_fmt_caps; +} + +// see: https://docs.microsoft.com/en-us/windows/win32/direct3d11/overviews-direct3d-11-resources-limits#resource-limits-for-feature-level-11-hardware +_SOKOL_PRIVATE void _sg_d3d11_init_caps(void) { + _sg.backend = SG_BACKEND_D3D11; + + _sg.features.origin_top_left = true; + _sg.features.image_clamp_to_border = true; + _sg.features.mrt_independent_blend_state = true; + _sg.features.mrt_independent_write_mask = true; + _sg.features.compute = true; + _sg.features.msaa_image_bindings = true; + + _sg.limits.max_image_size_2d = 16 * 1024; + _sg.limits.max_image_size_cube = 16 * 1024; + _sg.limits.max_image_size_3d = 2 * 1024; + _sg.limits.max_image_size_array = 16 * 1024; + _sg.limits.max_image_array_layers = 2 * 1024; + _sg.limits.max_vertex_attrs = SG_MAX_VERTEX_ATTRIBUTES; + + // see: https://docs.microsoft.com/en-us/windows/win32/api/d3d11/ne-d3d11-d3d11_format_support + for (int fmt = (SG_PIXELFORMAT_NONE+1); fmt < _SG_PIXELFORMAT_NUM; fmt++) { + const UINT srv_dxgi_fmt_caps = _sg_d3d11_dxgi_fmt_caps(_sg_d3d11_srv_pixel_format((sg_pixel_format)fmt)); + const UINT rtv_dxgi_fmt_caps = _sg_d3d11_dxgi_fmt_caps(_sg_d3d11_rtv_pixel_format((sg_pixel_format)fmt)); + const UINT dsv_dxgi_fmt_caps = _sg_d3d11_dxgi_fmt_caps(_sg_d3d11_dsv_pixel_format((sg_pixel_format)fmt)); + _sg_pixelformat_info_t* info = &_sg.formats[fmt]; + const bool render = 0 != (rtv_dxgi_fmt_caps & D3D11_FORMAT_SUPPORT_RENDER_TARGET); + const bool depth = 0 != (dsv_dxgi_fmt_caps & D3D11_FORMAT_SUPPORT_DEPTH_STENCIL); + info->sample = 0 != (srv_dxgi_fmt_caps & D3D11_FORMAT_SUPPORT_TEXTURE2D); + info->filter = 0 != (srv_dxgi_fmt_caps & D3D11_FORMAT_SUPPORT_SHADER_SAMPLE); + info->render = render || depth; + if (depth) { + info->blend = 0 != (dsv_dxgi_fmt_caps & D3D11_FORMAT_SUPPORT_BLENDABLE); + info->msaa = 0 != (dsv_dxgi_fmt_caps & D3D11_FORMAT_SUPPORT_MULTISAMPLE_RENDERTARGET); + } else { + info->blend = 0 != (rtv_dxgi_fmt_caps & D3D11_FORMAT_SUPPORT_BLENDABLE); + info->msaa = 0 != (rtv_dxgi_fmt_caps & D3D11_FORMAT_SUPPORT_MULTISAMPLE_RENDERTARGET); + } + info->depth = depth; + } +} + +_SOKOL_PRIVATE void _sg_d3d11_setup_backend(const sg_desc* desc) { + // assume _sg.d3d11 already is zero-initialized + SOKOL_ASSERT(desc); + SOKOL_ASSERT(desc->environment.d3d11.device); + SOKOL_ASSERT(desc->environment.d3d11.device_context); + _sg.d3d11.valid = true; + _sg.d3d11.dev = (ID3D11Device*) desc->environment.d3d11.device; + _sg.d3d11.ctx = (ID3D11DeviceContext*) desc->environment.d3d11.device_context; + _sg_d3d11_init_caps(); +} + +_SOKOL_PRIVATE void _sg_d3d11_discard_backend(void) { + SOKOL_ASSERT(_sg.d3d11.valid); + _sg.d3d11.valid = false; +} + +_SOKOL_PRIVATE void _sg_d3d11_clear_state(void) { + // clear all the device context state, so that resource refs don't keep stuck in the d3d device context + _sg_d3d11_ClearState(_sg.d3d11.ctx); +} + +_SOKOL_PRIVATE void _sg_d3d11_reset_state_cache(void) { + // there's currently no state cache in the D3D11 backend, so this is a no-op +} + +_SOKOL_PRIVATE sg_resource_state _sg_d3d11_create_buffer(_sg_buffer_t* buf, const sg_buffer_desc* desc) { + SOKOL_ASSERT(buf && desc); + SOKOL_ASSERT(!buf->d3d11.buf); + const bool injected = (0 != desc->d3d11_buffer); + if (injected) { + buf->d3d11.buf = (ID3D11Buffer*) desc->d3d11_buffer; + _sg_d3d11_AddRef(buf->d3d11.buf); + // FIXME: for storage buffers also need to inject resource view + } else { + D3D11_BUFFER_DESC d3d11_buf_desc; + _sg_clear(&d3d11_buf_desc, sizeof(d3d11_buf_desc)); + d3d11_buf_desc.ByteWidth = (UINT)buf->cmn.size; + d3d11_buf_desc.Usage = _sg_d3d11_buffer_usage(buf->cmn.usage, buf->cmn.type); + d3d11_buf_desc.BindFlags = _sg_d3d11_buffer_bind_flags(buf->cmn.usage, buf->cmn.type); + d3d11_buf_desc.CPUAccessFlags = _sg_d3d11_cpu_access_flags(buf->cmn.usage); + d3d11_buf_desc.MiscFlags = _sg_d3d11_buffer_misc_flags(buf->cmn.type); + D3D11_SUBRESOURCE_DATA* init_data_ptr = 0; + D3D11_SUBRESOURCE_DATA init_data; + _sg_clear(&init_data, sizeof(init_data)); + if (buf->cmn.usage == SG_USAGE_IMMUTABLE) { + // D3D11 doesn't allow creating immutable buffers without data, so need + // to explicitly provide a zero-initialized memory buffer + if (desc->data.ptr) { + init_data.pSysMem = desc->data.ptr; + } else { + init_data.pSysMem = (const void*)_sg_malloc_clear((size_t)buf->cmn.size); + } + init_data_ptr = &init_data; + } + HRESULT hr = _sg_d3d11_CreateBuffer(_sg.d3d11.dev, &d3d11_buf_desc, init_data_ptr, &buf->d3d11.buf); + if (init_data.pSysMem && (desc->data.ptr == 0)) { + _sg_free((void*)init_data.pSysMem); + } + if (!(SUCCEEDED(hr) && buf->d3d11.buf)) { + _SG_ERROR(D3D11_CREATE_BUFFER_FAILED); + return SG_RESOURCESTATE_FAILED; + } + + // for storage buffers need to create a shader-resource-view + // for read-only access, and an unordered-access-view for + // read-write access + if (buf->cmn.type == SG_BUFFERTYPE_STORAGEBUFFER) { + SOKOL_ASSERT(_sg_multiple_u64((uint64_t)buf->cmn.size, 4)); + D3D11_SHADER_RESOURCE_VIEW_DESC d3d11_srv_desc; + _sg_clear(&d3d11_srv_desc, sizeof(d3d11_srv_desc)); + d3d11_srv_desc.Format = DXGI_FORMAT_R32_TYPELESS; + d3d11_srv_desc.ViewDimension = D3D11_SRV_DIMENSION_BUFFEREX; + d3d11_srv_desc.BufferEx.FirstElement = 0; + d3d11_srv_desc.BufferEx.NumElements = ((UINT)buf->cmn.size) / 4; + d3d11_srv_desc.BufferEx.Flags = D3D11_BUFFEREX_SRV_FLAG_RAW; + hr = _sg_d3d11_CreateShaderResourceView(_sg.d3d11.dev, (ID3D11Resource*)buf->d3d11.buf, &d3d11_srv_desc, &buf->d3d11.srv); + if (!(SUCCEEDED(hr) && buf->d3d11.srv)) { + _SG_ERROR(D3D11_CREATE_BUFFER_SRV_FAILED); + return SG_RESOURCESTATE_FAILED; + } + if (buf->cmn.usage == SG_USAGE_IMMUTABLE) { + D3D11_UNORDERED_ACCESS_VIEW_DESC d3d11_uav_desc; + _sg_clear(&d3d11_uav_desc, sizeof(d3d11_uav_desc)); + d3d11_uav_desc.Format = DXGI_FORMAT_R32_TYPELESS; + d3d11_uav_desc.ViewDimension = D3D11_UAV_DIMENSION_BUFFER; + d3d11_uav_desc.Buffer.FirstElement = 0; + d3d11_uav_desc.Buffer.NumElements = ((UINT)buf->cmn.size) / 4; + d3d11_uav_desc.Buffer.Flags = D3D11_BUFFER_UAV_FLAG_RAW; + hr = _sg_d3d11_CreateUnorderedAccessView(_sg.d3d11.dev, (ID3D11Resource*)buf->d3d11.buf, &d3d11_uav_desc, &buf->d3d11.uav); + if (!(SUCCEEDED(hr) && buf->d3d11.uav)) { + _SG_ERROR(D3D11_CREATE_BUFFER_UAV_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } + } + _sg_d3d11_setlabel(buf->d3d11.buf, desc->label); + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_d3d11_discard_buffer(_sg_buffer_t* buf) { + SOKOL_ASSERT(buf); + if (buf->d3d11.buf) { + _sg_d3d11_Release(buf->d3d11.buf); + } + if (buf->d3d11.srv) { + _sg_d3d11_Release(buf->d3d11.srv); + } + if (buf->d3d11.uav) { + _sg_d3d11_Release(buf->d3d11.uav); + } +} + +_SOKOL_PRIVATE void _sg_d3d11_fill_subres_data(const _sg_image_t* img, const sg_image_data* data) { + const int num_faces = (img->cmn.type == SG_IMAGETYPE_CUBE) ? 6:1; + const int num_slices = (img->cmn.type == SG_IMAGETYPE_ARRAY) ? img->cmn.num_slices:1; + int subres_index = 0; + for (int face_index = 0; face_index < num_faces; face_index++) { + for (int slice_index = 0; slice_index < num_slices; slice_index++) { + for (int mip_index = 0; mip_index < img->cmn.num_mipmaps; mip_index++, subres_index++) { + SOKOL_ASSERT(subres_index < (SG_MAX_MIPMAPS * SG_MAX_TEXTUREARRAY_LAYERS)); + D3D11_SUBRESOURCE_DATA* subres_data = &_sg.d3d11.subres_data[subres_index]; + const int mip_width = _sg_miplevel_dim(img->cmn.width, mip_index); + const int mip_height = _sg_miplevel_dim(img->cmn.height, mip_index); + const sg_range* subimg_data = &(data->subimage[face_index][mip_index]); + const size_t slice_size = subimg_data->size / (size_t)num_slices; + const size_t slice_offset = slice_size * (size_t)slice_index; + const uint8_t* ptr = (const uint8_t*) subimg_data->ptr; + subres_data->pSysMem = ptr + slice_offset; + subres_data->SysMemPitch = (UINT)_sg_row_pitch(img->cmn.pixel_format, mip_width, 1); + if (img->cmn.type == SG_IMAGETYPE_3D) { + // FIXME? const int mip_depth = _sg_miplevel_dim(img->depth, mip_index); + subres_data->SysMemSlicePitch = (UINT)_sg_surface_pitch(img->cmn.pixel_format, mip_width, mip_height, 1); + } else { + subres_data->SysMemSlicePitch = 0; + } + } + } + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_d3d11_create_image(_sg_image_t* img, const sg_image_desc* desc) { + SOKOL_ASSERT(img && desc); + SOKOL_ASSERT((0 == img->d3d11.tex2d) && (0 == img->d3d11.tex3d) && (0 == img->d3d11.res) && (0 == img->d3d11.srv)); + HRESULT hr; + + const bool injected = (0 != desc->d3d11_texture); + const bool msaa = (img->cmn.sample_count > 1); + SOKOL_ASSERT(!(msaa && (img->cmn.type == SG_IMAGETYPE_CUBE))); + img->d3d11.format = _sg_d3d11_texture_pixel_format(img->cmn.pixel_format); + if (img->d3d11.format == DXGI_FORMAT_UNKNOWN) { + _SG_ERROR(D3D11_CREATE_2D_TEXTURE_UNSUPPORTED_PIXEL_FORMAT); + return SG_RESOURCESTATE_FAILED; + } + + // prepare initial content pointers + D3D11_SUBRESOURCE_DATA* init_data = 0; + if (!injected && (img->cmn.usage == SG_USAGE_IMMUTABLE) && !img->cmn.render_target) { + _sg_d3d11_fill_subres_data(img, &desc->data); + init_data = _sg.d3d11.subres_data; + } + if (img->cmn.type != SG_IMAGETYPE_3D) { + // 2D-, cube- or array-texture + // first check for injected texture and/or resource view + if (injected) { + img->d3d11.tex2d = (ID3D11Texture2D*) desc->d3d11_texture; + _sg_d3d11_AddRef(img->d3d11.tex2d); + img->d3d11.srv = (ID3D11ShaderResourceView*) desc->d3d11_shader_resource_view; + if (img->d3d11.srv) { + _sg_d3d11_AddRef(img->d3d11.srv); + } + } else { + // if not injected, create 2D texture + D3D11_TEXTURE2D_DESC d3d11_tex_desc; + _sg_clear(&d3d11_tex_desc, sizeof(d3d11_tex_desc)); + d3d11_tex_desc.Width = (UINT)img->cmn.width; + d3d11_tex_desc.Height = (UINT)img->cmn.height; + d3d11_tex_desc.MipLevels = (UINT)img->cmn.num_mipmaps; + switch (img->cmn.type) { + case SG_IMAGETYPE_ARRAY: d3d11_tex_desc.ArraySize = (UINT)img->cmn.num_slices; break; + case SG_IMAGETYPE_CUBE: d3d11_tex_desc.ArraySize = 6; break; + default: d3d11_tex_desc.ArraySize = 1; break; + } + d3d11_tex_desc.Format = img->d3d11.format; + d3d11_tex_desc.BindFlags = D3D11_BIND_SHADER_RESOURCE; + if (img->cmn.render_target) { + d3d11_tex_desc.Usage = D3D11_USAGE_DEFAULT; + if (_sg_is_depth_or_depth_stencil_format(img->cmn.pixel_format)) { + d3d11_tex_desc.BindFlags |= D3D11_BIND_DEPTH_STENCIL; + } else { + d3d11_tex_desc.BindFlags |= D3D11_BIND_RENDER_TARGET; + } + d3d11_tex_desc.CPUAccessFlags = 0; + } else { + d3d11_tex_desc.Usage = _sg_d3d11_usage(img->cmn.usage); + d3d11_tex_desc.CPUAccessFlags = _sg_d3d11_cpu_access_flags(img->cmn.usage); + } + d3d11_tex_desc.SampleDesc.Count = (UINT)img->cmn.sample_count; + d3d11_tex_desc.SampleDesc.Quality = (UINT) (msaa ? D3D11_STANDARD_MULTISAMPLE_PATTERN : 0); + d3d11_tex_desc.MiscFlags = (img->cmn.type == SG_IMAGETYPE_CUBE) ? D3D11_RESOURCE_MISC_TEXTURECUBE : 0; + + hr = _sg_d3d11_CreateTexture2D(_sg.d3d11.dev, &d3d11_tex_desc, init_data, &img->d3d11.tex2d); + if (!(SUCCEEDED(hr) && img->d3d11.tex2d)) { + _SG_ERROR(D3D11_CREATE_2D_TEXTURE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(img->d3d11.tex2d, desc->label); + + // create shader-resource-view for 2D texture + D3D11_SHADER_RESOURCE_VIEW_DESC d3d11_srv_desc; + _sg_clear(&d3d11_srv_desc, sizeof(d3d11_srv_desc)); + d3d11_srv_desc.Format = _sg_d3d11_srv_pixel_format(img->cmn.pixel_format); + switch (img->cmn.type) { + case SG_IMAGETYPE_2D: + d3d11_srv_desc.ViewDimension = msaa ? D3D11_SRV_DIMENSION_TEXTURE2DMS : D3D11_SRV_DIMENSION_TEXTURE2D; + d3d11_srv_desc.Texture2D.MipLevels = (UINT)img->cmn.num_mipmaps; + break; + case SG_IMAGETYPE_CUBE: + d3d11_srv_desc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURECUBE; + d3d11_srv_desc.TextureCube.MipLevels = (UINT)img->cmn.num_mipmaps; + break; + case SG_IMAGETYPE_ARRAY: + d3d11_srv_desc.ViewDimension = msaa ? D3D11_SRV_DIMENSION_TEXTURE2DMSARRAY : D3D11_SRV_DIMENSION_TEXTURE2DARRAY; + d3d11_srv_desc.Texture2DArray.MipLevels = (UINT)img->cmn.num_mipmaps; + d3d11_srv_desc.Texture2DArray.ArraySize = (UINT)img->cmn.num_slices; + break; + default: + SOKOL_UNREACHABLE; break; + } + hr = _sg_d3d11_CreateShaderResourceView(_sg.d3d11.dev, (ID3D11Resource*)img->d3d11.tex2d, &d3d11_srv_desc, &img->d3d11.srv); + if (!(SUCCEEDED(hr) && img->d3d11.srv)) { + _SG_ERROR(D3D11_CREATE_2D_SRV_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(img->d3d11.srv, desc->label); + } + SOKOL_ASSERT(img->d3d11.tex2d); + img->d3d11.res = (ID3D11Resource*)img->d3d11.tex2d; + _sg_d3d11_AddRef(img->d3d11.res); + } else { + // 3D texture - same procedure, first check if injected, than create non-injected + if (injected) { + img->d3d11.tex3d = (ID3D11Texture3D*) desc->d3d11_texture; + _sg_d3d11_AddRef(img->d3d11.tex3d); + img->d3d11.srv = (ID3D11ShaderResourceView*) desc->d3d11_shader_resource_view; + if (img->d3d11.srv) { + _sg_d3d11_AddRef(img->d3d11.srv); + } + } else { + // not injected, create 3d texture + D3D11_TEXTURE3D_DESC d3d11_tex_desc; + _sg_clear(&d3d11_tex_desc, sizeof(d3d11_tex_desc)); + d3d11_tex_desc.Width = (UINT)img->cmn.width; + d3d11_tex_desc.Height = (UINT)img->cmn.height; + d3d11_tex_desc.Depth = (UINT)img->cmn.num_slices; + d3d11_tex_desc.MipLevels = (UINT)img->cmn.num_mipmaps; + d3d11_tex_desc.Format = img->d3d11.format; + if (img->cmn.render_target) { + d3d11_tex_desc.Usage = D3D11_USAGE_DEFAULT; + d3d11_tex_desc.BindFlags = D3D11_BIND_RENDER_TARGET; + d3d11_tex_desc.CPUAccessFlags = 0; + } else { + d3d11_tex_desc.Usage = _sg_d3d11_usage(img->cmn.usage); + d3d11_tex_desc.BindFlags = D3D11_BIND_SHADER_RESOURCE; + d3d11_tex_desc.CPUAccessFlags = _sg_d3d11_cpu_access_flags(img->cmn.usage); + } + if (img->d3d11.format == DXGI_FORMAT_UNKNOWN) { + _SG_ERROR(D3D11_CREATE_3D_TEXTURE_UNSUPPORTED_PIXEL_FORMAT); + return SG_RESOURCESTATE_FAILED; + } + hr = _sg_d3d11_CreateTexture3D(_sg.d3d11.dev, &d3d11_tex_desc, init_data, &img->d3d11.tex3d); + if (!(SUCCEEDED(hr) && img->d3d11.tex3d)) { + _SG_ERROR(D3D11_CREATE_3D_TEXTURE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(img->d3d11.tex3d, desc->label); + + // create shader-resource-view for 3D texture + if (!msaa) { + D3D11_SHADER_RESOURCE_VIEW_DESC d3d11_srv_desc; + _sg_clear(&d3d11_srv_desc, sizeof(d3d11_srv_desc)); + d3d11_srv_desc.Format = _sg_d3d11_srv_pixel_format(img->cmn.pixel_format); + d3d11_srv_desc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE3D; + d3d11_srv_desc.Texture3D.MipLevels = (UINT)img->cmn.num_mipmaps; + hr = _sg_d3d11_CreateShaderResourceView(_sg.d3d11.dev, (ID3D11Resource*)img->d3d11.tex3d, &d3d11_srv_desc, &img->d3d11.srv); + if (!(SUCCEEDED(hr) && img->d3d11.srv)) { + _SG_ERROR(D3D11_CREATE_3D_SRV_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(img->d3d11.srv, desc->label); + } + } + SOKOL_ASSERT(img->d3d11.tex3d); + img->d3d11.res = (ID3D11Resource*)img->d3d11.tex3d; + _sg_d3d11_AddRef(img->d3d11.res); + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_d3d11_discard_image(_sg_image_t* img) { + SOKOL_ASSERT(img); + if (img->d3d11.tex2d) { + _sg_d3d11_Release(img->d3d11.tex2d); + } + if (img->d3d11.tex3d) { + _sg_d3d11_Release(img->d3d11.tex3d); + } + if (img->d3d11.res) { + _sg_d3d11_Release(img->d3d11.res); + } + if (img->d3d11.srv) { + _sg_d3d11_Release(img->d3d11.srv); + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_d3d11_create_sampler(_sg_sampler_t* smp, const sg_sampler_desc* desc) { + SOKOL_ASSERT(smp && desc); + SOKOL_ASSERT(0 == smp->d3d11.smp); + const bool injected = (0 != desc->d3d11_sampler); + if (injected) { + smp->d3d11.smp = (ID3D11SamplerState*)desc->d3d11_sampler; + _sg_d3d11_AddRef(smp->d3d11.smp); + } else { + D3D11_SAMPLER_DESC d3d11_smp_desc; + _sg_clear(&d3d11_smp_desc, sizeof(d3d11_smp_desc)); + d3d11_smp_desc.Filter = _sg_d3d11_filter(desc->min_filter, desc->mag_filter, desc->mipmap_filter, desc->compare != SG_COMPAREFUNC_NEVER, desc->max_anisotropy); + d3d11_smp_desc.AddressU = _sg_d3d11_address_mode(desc->wrap_u); + d3d11_smp_desc.AddressV = _sg_d3d11_address_mode(desc->wrap_v); + d3d11_smp_desc.AddressW = _sg_d3d11_address_mode(desc->wrap_w); + d3d11_smp_desc.MipLODBias = 0.0f; // FIXME? + switch (desc->border_color) { + case SG_BORDERCOLOR_TRANSPARENT_BLACK: + // all 0.0f + break; + case SG_BORDERCOLOR_OPAQUE_WHITE: + for (int i = 0; i < 4; i++) { + d3d11_smp_desc.BorderColor[i] = 1.0f; + } + break; + default: + // opaque black + d3d11_smp_desc.BorderColor[3] = 1.0f; + break; + } + d3d11_smp_desc.MaxAnisotropy = desc->max_anisotropy; + d3d11_smp_desc.ComparisonFunc = _sg_d3d11_compare_func(desc->compare); + d3d11_smp_desc.MinLOD = desc->min_lod; + d3d11_smp_desc.MaxLOD = desc->max_lod; + HRESULT hr = _sg_d3d11_CreateSamplerState(_sg.d3d11.dev, &d3d11_smp_desc, &smp->d3d11.smp); + if (!(SUCCEEDED(hr) && smp->d3d11.smp)) { + _SG_ERROR(D3D11_CREATE_SAMPLER_STATE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(smp->d3d11.smp, desc->label); + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_d3d11_discard_sampler(_sg_sampler_t* smp) { + SOKOL_ASSERT(smp); + if (smp->d3d11.smp) { + _sg_d3d11_Release(smp->d3d11.smp); + } +} + +_SOKOL_PRIVATE bool _sg_d3d11_load_d3dcompiler_dll(void) { + if ((0 == _sg.d3d11.d3dcompiler_dll) && !_sg.d3d11.d3dcompiler_dll_load_failed) { + _sg.d3d11.d3dcompiler_dll = LoadLibraryA("d3dcompiler_47.dll"); + if (0 == _sg.d3d11.d3dcompiler_dll) { + // don't attempt to load missing DLL in the future + _SG_ERROR(D3D11_LOAD_D3DCOMPILER_47_DLL_FAILED); + _sg.d3d11.d3dcompiler_dll_load_failed = true; + return false; + } + // look up function pointers + _sg.d3d11.D3DCompile_func = (pD3DCompile)(void*) GetProcAddress(_sg.d3d11.d3dcompiler_dll, "D3DCompile"); + SOKOL_ASSERT(_sg.d3d11.D3DCompile_func); + } + return 0 != _sg.d3d11.d3dcompiler_dll; +} + +_SOKOL_PRIVATE ID3DBlob* _sg_d3d11_compile_shader(const sg_shader_function* shd_func) { + if (!_sg_d3d11_load_d3dcompiler_dll()) { + return NULL; + } + SOKOL_ASSERT(shd_func->d3d11_target); + UINT flags1 = D3DCOMPILE_PACK_MATRIX_COLUMN_MAJOR; + if (_sg.desc.d3d11_shader_debugging) { + flags1 |= D3DCOMPILE_DEBUG | D3DCOMPILE_SKIP_OPTIMIZATION; + } else { + flags1 |= D3DCOMPILE_OPTIMIZATION_LEVEL3; + } + ID3DBlob* output = NULL; + ID3DBlob* errors_or_warnings = NULL; + HRESULT hr = _sg.d3d11.D3DCompile_func( + shd_func->source, // pSrcData + strlen(shd_func->source), // SrcDataSize + NULL, // pSourceName + NULL, // pDefines + NULL, // pInclude + shd_func->entry ? shd_func->entry : "main", // pEntryPoint + shd_func->d3d11_target, // pTarget + flags1, // Flags1 + 0, // Flags2 + &output, // ppCode + &errors_or_warnings); // ppErrorMsgs + if (FAILED(hr)) { + _SG_ERROR(D3D11_SHADER_COMPILATION_FAILED); + } + if (errors_or_warnings) { + _SG_WARN(D3D11_SHADER_COMPILATION_OUTPUT); + _SG_LOGMSG(D3D11_SHADER_COMPILATION_OUTPUT, (LPCSTR)_sg_d3d11_GetBufferPointer(errors_or_warnings)); + _sg_d3d11_Release(errors_or_warnings); errors_or_warnings = NULL; + } + if (FAILED(hr)) { + // just in case, usually output is NULL here + if (output) { + _sg_d3d11_Release(output); + output = NULL; + } + } + return output; +} + +// NOTE: this is an out-of-range check for HLSL bindslots that's also active in release mode +_SOKOL_PRIVATE bool _sg_d3d11_ensure_hlsl_bindslot_ranges(const sg_shader_desc* desc) { + SOKOL_ASSERT(desc); + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + if (desc->uniform_blocks[i].hlsl_register_b_n >= _SG_D3D11_MAX_STAGE_UB_BINDINGS) { + _SG_ERROR(D3D11_UNIFORMBLOCK_HLSL_REGISTER_B_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (desc->storage_buffers[i].hlsl_register_t_n >= _SG_D3D11_MAX_STAGE_SRV_BINDINGS) { + _SG_ERROR(D3D11_STORAGEBUFFER_HLSL_REGISTER_T_OUT_OF_RANGE); + return false; + } + if (desc->storage_buffers[i].hlsl_register_u_n >= _SG_D3D11_MAX_STAGE_UAV_BINDINGS) { + _SG_ERROR(D3D11_STORAGEBUFFER_HLSL_REGISTER_U_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + if (desc->images[i].hlsl_register_t_n >= _SG_D3D11_MAX_STAGE_SRV_BINDINGS) { + _SG_ERROR(D3D11_IMAGE_HLSL_REGISTER_T_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + if (desc->samplers[i].hlsl_register_s_n >= _SG_D3D11_MAX_STAGE_SMP_BINDINGS) { + _SG_ERROR(D3D11_SAMPLER_HLSL_REGISTER_S_OUT_OF_RANGE); + return false; + } + } + return true; +} + +_SOKOL_PRIVATE sg_resource_state _sg_d3d11_create_shader(_sg_shader_t* shd, const sg_shader_desc* desc) { + SOKOL_ASSERT(shd && desc); + SOKOL_ASSERT(!shd->d3d11.vs && !shd->d3d11.fs && !shd->d3d11.cs && !shd->d3d11.vs_blob); + HRESULT hr; + + // perform a range-check on HLSL bindslots that's also active in release + // mode to avoid potential out-of-bounds array accesses + if (!_sg_d3d11_ensure_hlsl_bindslot_ranges(desc)) { + return SG_RESOURCESTATE_FAILED; + } + + // copy vertex attribute semantic names and indices + for (size_t i = 0; i < SG_MAX_VERTEX_ATTRIBUTES; i++) { + _sg_strcpy(&shd->d3d11.attrs[i].sem_name, desc->attrs[i].hlsl_sem_name); + shd->d3d11.attrs[i].sem_index = desc->attrs[i].hlsl_sem_index; + } + + // copy HLSL bind slots + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + shd->d3d11.ub_register_b_n[i] = desc->uniform_blocks[i].hlsl_register_b_n; + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + shd->d3d11.sbuf_register_t_n[i] = desc->storage_buffers[i].hlsl_register_t_n; + shd->d3d11.sbuf_register_u_n[i] = desc->storage_buffers[i].hlsl_register_u_n; + } + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + shd->d3d11.img_register_t_n[i] = desc->images[i].hlsl_register_t_n; + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + shd->d3d11.smp_register_s_n[i] = desc->samplers[i].hlsl_register_s_n; + } + + // create a D3D constant buffer for each uniform block + for (size_t ub_index = 0; ub_index < SG_MAX_UNIFORMBLOCK_BINDSLOTS; ub_index++) { + const sg_shader_stage stage = desc->uniform_blocks[ub_index].stage; + if (stage == SG_SHADERSTAGE_NONE) { + continue; + } + const _sg_shader_uniform_block_t* ub = &shd->cmn.uniform_blocks[ub_index]; + ID3D11Buffer* cbuf = 0; + D3D11_BUFFER_DESC cb_desc; + _sg_clear(&cb_desc, sizeof(cb_desc)); + cb_desc.ByteWidth = (UINT)_sg_roundup((int)ub->size, 16); + cb_desc.Usage = D3D11_USAGE_DEFAULT; + cb_desc.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + hr = _sg_d3d11_CreateBuffer(_sg.d3d11.dev, &cb_desc, NULL, &cbuf); + if (!(SUCCEEDED(hr) && cbuf)) { + _SG_ERROR(D3D11_CREATE_CONSTANT_BUFFER_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(cbuf, desc->label); + shd->d3d11.all_cbufs[ub_index] = cbuf; + + const uint8_t d3d11_slot = shd->d3d11.ub_register_b_n[ub_index]; + SOKOL_ASSERT(d3d11_slot < _SG_D3D11_MAX_STAGE_UB_BINDINGS); + if (stage == SG_SHADERSTAGE_VERTEX) { + SOKOL_ASSERT(0 == shd->d3d11.vs_cbufs[d3d11_slot]); + shd->d3d11.vs_cbufs[d3d11_slot] = cbuf; + } else if (stage == SG_SHADERSTAGE_FRAGMENT) { + SOKOL_ASSERT(0 == shd->d3d11.fs_cbufs[d3d11_slot]); + shd->d3d11.fs_cbufs[d3d11_slot] = cbuf; + } else if (stage == SG_SHADERSTAGE_COMPUTE) { + SOKOL_ASSERT(0 == shd->d3d11.cs_cbufs[d3d11_slot]); + shd->d3d11.cs_cbufs[d3d11_slot] = cbuf; + } else { + SOKOL_UNREACHABLE; + } + } + + // create shader functions + const bool has_vs = desc->vertex_func.bytecode.ptr || desc->vertex_func.source; + const bool has_fs = desc->fragment_func.bytecode.ptr || desc->fragment_func.source; + const bool has_cs = desc->compute_func.bytecode.ptr || desc->compute_func.source; + bool vs_valid = false; bool fs_valid = false; bool cs_valid = false; + if (has_vs) { + const void* vs_ptr = 0; SIZE_T vs_length = 0; + ID3DBlob* vs_blob = 0; + if (desc->vertex_func.bytecode.ptr) { + SOKOL_ASSERT(desc->vertex_func.bytecode.size > 0); + vs_ptr = desc->vertex_func.bytecode.ptr; + vs_length = desc->vertex_func.bytecode.size; + } else { + SOKOL_ASSERT(desc->vertex_func.source); + vs_blob = _sg_d3d11_compile_shader(&desc->vertex_func); + if (vs_blob) { + vs_ptr = _sg_d3d11_GetBufferPointer(vs_blob); + vs_length = _sg_d3d11_GetBufferSize(vs_blob); + } + } + if (vs_ptr && (vs_length > 0)) { + hr = _sg_d3d11_CreateVertexShader(_sg.d3d11.dev, vs_ptr, vs_length, NULL, &shd->d3d11.vs); + vs_valid = SUCCEEDED(hr) && shd->d3d11.vs; + } + // set label, and need to store a copy of the vertex shader blob for the pipeline creation + if (vs_valid) { + _sg_d3d11_setlabel(shd->d3d11.vs, desc->label); + shd->d3d11.vs_blob_length = vs_length; + shd->d3d11.vs_blob = _sg_malloc((size_t)vs_length); + SOKOL_ASSERT(shd->d3d11.vs_blob); + memcpy(shd->d3d11.vs_blob, vs_ptr, vs_length); + } + if (vs_blob) { + _sg_d3d11_Release(vs_blob); + } + } + if (has_fs) { + const void* fs_ptr = 0; SIZE_T fs_length = 0; + ID3DBlob* fs_blob = 0; + if (desc->fragment_func.bytecode.ptr) { + SOKOL_ASSERT(desc->fragment_func.bytecode.size > 0); + fs_ptr = desc->fragment_func.bytecode.ptr; + fs_length = desc->fragment_func.bytecode.size; + } else { + SOKOL_ASSERT(desc->fragment_func.source); + fs_blob = _sg_d3d11_compile_shader(&desc->fragment_func); + if (fs_blob) { + fs_ptr = _sg_d3d11_GetBufferPointer(fs_blob); + fs_length = _sg_d3d11_GetBufferSize(fs_blob); + } + } + if (fs_ptr && (fs_length > 0)) { + hr = _sg_d3d11_CreatePixelShader(_sg.d3d11.dev, fs_ptr, fs_length, NULL, &shd->d3d11.fs); + fs_valid = SUCCEEDED(hr) && shd->d3d11.fs; + } + if (fs_valid) { + _sg_d3d11_setlabel(shd->d3d11.fs, desc->label); + } + if (fs_blob) { + _sg_d3d11_Release(fs_blob); + } + } + if (has_cs) { + const void* cs_ptr = 0; SIZE_T cs_length = 0; + ID3DBlob* cs_blob = 0; + if (desc->compute_func.bytecode.ptr) { + SOKOL_ASSERT(desc->compute_func.bytecode.size > 0); + cs_ptr = desc->compute_func.bytecode.ptr; + cs_length = desc->compute_func.bytecode.size; + } else { + SOKOL_ASSERT(desc->compute_func.source); + cs_blob = _sg_d3d11_compile_shader(&desc->compute_func); + if (cs_blob) { + cs_ptr = _sg_d3d11_GetBufferPointer(cs_blob); + cs_length = _sg_d3d11_GetBufferSize(cs_blob); + } + } + if (cs_ptr && (cs_length > 0)) { + hr = _sg_d3d11_CreateComputeShader(_sg.d3d11.dev, cs_ptr, cs_length, NULL, &shd->d3d11.cs); + cs_valid = SUCCEEDED(hr) && shd->d3d11.cs; + } + if (cs_blob) { + _sg_d3d11_Release(cs_blob); + } + } + if ((vs_valid && fs_valid) || cs_valid) { + return SG_RESOURCESTATE_VALID; + } else { + return SG_RESOURCESTATE_FAILED; + } +} + +_SOKOL_PRIVATE void _sg_d3d11_discard_shader(_sg_shader_t* shd) { + SOKOL_ASSERT(shd); + if (shd->d3d11.vs) { + _sg_d3d11_Release(shd->d3d11.vs); + } + if (shd->d3d11.fs) { + _sg_d3d11_Release(shd->d3d11.fs); + } + if (shd->d3d11.cs) { + _sg_d3d11_Release(shd->d3d11.cs); + } + if (shd->d3d11.vs_blob) { + _sg_free(shd->d3d11.vs_blob); + } + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + if (shd->d3d11.all_cbufs[i]) { + _sg_d3d11_Release(shd->d3d11.all_cbufs[i]); + } + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_d3d11_create_pipeline(_sg_pipeline_t* pip, _sg_shader_t* shd, const sg_pipeline_desc* desc) { + SOKOL_ASSERT(pip && shd && desc); + SOKOL_ASSERT(desc->shader.id == shd->slot.id); + SOKOL_ASSERT(shd->slot.state == SG_RESOURCESTATE_VALID); + + pip->shader = shd; + + // if this is a compute pipeline, we're done here + if (pip->cmn.is_compute) { + return SG_RESOURCESTATE_VALID; + } + + // a render pipeline... + SOKOL_ASSERT(shd->d3d11.vs_blob && shd->d3d11.vs_blob_length > 0); + SOKOL_ASSERT(!pip->d3d11.il && !pip->d3d11.rs && !pip->d3d11.dss && !pip->d3d11.bs); + + pip->d3d11.index_format = _sg_d3d11_index_format(pip->cmn.index_type); + pip->d3d11.topology = _sg_d3d11_primitive_topology(desc->primitive_type); + pip->d3d11.stencil_ref = desc->stencil.ref; + + // create input layout object + HRESULT hr; + D3D11_INPUT_ELEMENT_DESC d3d11_comps[SG_MAX_VERTEX_ATTRIBUTES]; + _sg_clear(d3d11_comps, sizeof(d3d11_comps)); + size_t attr_index = 0; + for (; attr_index < SG_MAX_VERTEX_ATTRIBUTES; attr_index++) { + const sg_vertex_attr_state* a_state = &desc->layout.attrs[attr_index]; + if (a_state->format == SG_VERTEXFORMAT_INVALID) { + break; + } + SOKOL_ASSERT(a_state->buffer_index < SG_MAX_VERTEXBUFFER_BINDSLOTS); + SOKOL_ASSERT(pip->cmn.vertex_buffer_layout_active[a_state->buffer_index]); + const sg_vertex_buffer_layout_state* l_state = &desc->layout.buffers[a_state->buffer_index]; + const sg_vertex_step step_func = l_state->step_func; + const int step_rate = l_state->step_rate; + D3D11_INPUT_ELEMENT_DESC* d3d11_comp = &d3d11_comps[attr_index]; + d3d11_comp->SemanticName = _sg_strptr(&shd->d3d11.attrs[attr_index].sem_name); + d3d11_comp->SemanticIndex = (UINT)shd->d3d11.attrs[attr_index].sem_index; + d3d11_comp->Format = _sg_d3d11_vertex_format(a_state->format); + d3d11_comp->InputSlot = (UINT)a_state->buffer_index; + d3d11_comp->AlignedByteOffset = (UINT)a_state->offset; + d3d11_comp->InputSlotClass = _sg_d3d11_input_classification(step_func); + if (SG_VERTEXSTEP_PER_INSTANCE == step_func) { + d3d11_comp->InstanceDataStepRate = (UINT)step_rate; + pip->cmn.use_instanced_draw = true; + } + } + for (size_t layout_index = 0; layout_index < SG_MAX_VERTEXBUFFER_BINDSLOTS; layout_index++) { + if (pip->cmn.vertex_buffer_layout_active[layout_index]) { + const sg_vertex_buffer_layout_state* l_state = &desc->layout.buffers[layout_index]; + SOKOL_ASSERT(l_state->stride > 0); + pip->d3d11.vb_strides[layout_index] = (UINT)l_state->stride; + } else { + pip->d3d11.vb_strides[layout_index] = 0; + } + } + if (attr_index > 0) { + hr = _sg_d3d11_CreateInputLayout(_sg.d3d11.dev, + d3d11_comps, // pInputElementDesc + (UINT)attr_index, // NumElements + shd->d3d11.vs_blob, // pShaderByteCodeWithInputSignature + shd->d3d11.vs_blob_length, // BytecodeLength + &pip->d3d11.il); + if (!(SUCCEEDED(hr) && pip->d3d11.il)) { + _SG_ERROR(D3D11_CREATE_INPUT_LAYOUT_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(pip->d3d11.il, desc->label); + } + + // create rasterizer state + D3D11_RASTERIZER_DESC rs_desc; + _sg_clear(&rs_desc, sizeof(rs_desc)); + rs_desc.FillMode = D3D11_FILL_SOLID; + rs_desc.CullMode = _sg_d3d11_cull_mode(desc->cull_mode); + rs_desc.FrontCounterClockwise = desc->face_winding == SG_FACEWINDING_CCW; + rs_desc.DepthBias = (INT) pip->cmn.depth.bias; + rs_desc.DepthBiasClamp = pip->cmn.depth.bias_clamp; + rs_desc.SlopeScaledDepthBias = pip->cmn.depth.bias_slope_scale; + rs_desc.DepthClipEnable = TRUE; + rs_desc.ScissorEnable = TRUE; + rs_desc.MultisampleEnable = desc->sample_count > 1; + rs_desc.AntialiasedLineEnable = FALSE; + hr = _sg_d3d11_CreateRasterizerState(_sg.d3d11.dev, &rs_desc, &pip->d3d11.rs); + if (!(SUCCEEDED(hr) && pip->d3d11.rs)) { + _SG_ERROR(D3D11_CREATE_RASTERIZER_STATE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(pip->d3d11.rs, desc->label); + + // create depth-stencil state + D3D11_DEPTH_STENCIL_DESC dss_desc; + _sg_clear(&dss_desc, sizeof(dss_desc)); + dss_desc.DepthEnable = TRUE; + dss_desc.DepthWriteMask = desc->depth.write_enabled ? D3D11_DEPTH_WRITE_MASK_ALL : D3D11_DEPTH_WRITE_MASK_ZERO; + dss_desc.DepthFunc = _sg_d3d11_compare_func(desc->depth.compare); + dss_desc.StencilEnable = desc->stencil.enabled; + dss_desc.StencilReadMask = desc->stencil.read_mask; + dss_desc.StencilWriteMask = desc->stencil.write_mask; + const sg_stencil_face_state* sf = &desc->stencil.front; + dss_desc.FrontFace.StencilFailOp = _sg_d3d11_stencil_op(sf->fail_op); + dss_desc.FrontFace.StencilDepthFailOp = _sg_d3d11_stencil_op(sf->depth_fail_op); + dss_desc.FrontFace.StencilPassOp = _sg_d3d11_stencil_op(sf->pass_op); + dss_desc.FrontFace.StencilFunc = _sg_d3d11_compare_func(sf->compare); + const sg_stencil_face_state* sb = &desc->stencil.back; + dss_desc.BackFace.StencilFailOp = _sg_d3d11_stencil_op(sb->fail_op); + dss_desc.BackFace.StencilDepthFailOp = _sg_d3d11_stencil_op(sb->depth_fail_op); + dss_desc.BackFace.StencilPassOp = _sg_d3d11_stencil_op(sb->pass_op); + dss_desc.BackFace.StencilFunc = _sg_d3d11_compare_func(sb->compare); + hr = _sg_d3d11_CreateDepthStencilState(_sg.d3d11.dev, &dss_desc, &pip->d3d11.dss); + if (!(SUCCEEDED(hr) && pip->d3d11.dss)) { + _SG_ERROR(D3D11_CREATE_DEPTH_STENCIL_STATE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(pip->d3d11.dss, desc->label); + + // create blend state + D3D11_BLEND_DESC bs_desc; + _sg_clear(&bs_desc, sizeof(bs_desc)); + bs_desc.AlphaToCoverageEnable = desc->alpha_to_coverage_enabled; + bs_desc.IndependentBlendEnable = TRUE; + { + size_t i = 0; + for (i = 0; i < (size_t)desc->color_count; i++) { + const sg_blend_state* src = &desc->colors[i].blend; + D3D11_RENDER_TARGET_BLEND_DESC* dst = &bs_desc.RenderTarget[i]; + dst->BlendEnable = src->enabled; + dst->SrcBlend = _sg_d3d11_blend_factor(src->src_factor_rgb); + dst->DestBlend = _sg_d3d11_blend_factor(src->dst_factor_rgb); + dst->BlendOp = _sg_d3d11_blend_op(src->op_rgb); + dst->SrcBlendAlpha = _sg_d3d11_blend_factor(src->src_factor_alpha); + dst->DestBlendAlpha = _sg_d3d11_blend_factor(src->dst_factor_alpha); + dst->BlendOpAlpha = _sg_d3d11_blend_op(src->op_alpha); + dst->RenderTargetWriteMask = _sg_d3d11_color_write_mask(desc->colors[i].write_mask); + } + for (; i < 8; i++) { + D3D11_RENDER_TARGET_BLEND_DESC* dst = &bs_desc.RenderTarget[i]; + dst->BlendEnable = FALSE; + dst->SrcBlend = dst->SrcBlendAlpha = D3D11_BLEND_ONE; + dst->DestBlend = dst->DestBlendAlpha = D3D11_BLEND_ZERO; + dst->BlendOp = dst->BlendOpAlpha = D3D11_BLEND_OP_ADD; + dst->RenderTargetWriteMask = D3D11_COLOR_WRITE_ENABLE_ALL; + } + } + hr = _sg_d3d11_CreateBlendState(_sg.d3d11.dev, &bs_desc, &pip->d3d11.bs); + if (!(SUCCEEDED(hr) && pip->d3d11.bs)) { + _SG_ERROR(D3D11_CREATE_BLEND_STATE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(pip->d3d11.bs, desc->label); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_d3d11_discard_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + if (pip == _sg.d3d11.cur_pipeline) { + _sg.d3d11.cur_pipeline = 0; + _sg.d3d11.cur_pipeline_id.id = SG_INVALID_ID; + } + if (pip->d3d11.il) { + _sg_d3d11_Release(pip->d3d11.il); + } + if (pip->d3d11.rs) { + _sg_d3d11_Release(pip->d3d11.rs); + } + if (pip->d3d11.dss) { + _sg_d3d11_Release(pip->d3d11.dss); + } + if (pip->d3d11.bs) { + _sg_d3d11_Release(pip->d3d11.bs); + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_d3d11_create_attachments(_sg_attachments_t* atts, _sg_image_t** color_images, _sg_image_t** resolve_images, _sg_image_t* ds_img, const sg_attachments_desc* desc) { + SOKOL_ASSERT(atts && desc); + SOKOL_ASSERT(color_images && resolve_images); + SOKOL_ASSERT(_sg.d3d11.dev); + + // copy image pointers + for (size_t i = 0; i < (size_t)atts->cmn.num_colors; i++) { + const sg_attachment_desc* color_desc = &desc->colors[i]; + _SOKOL_UNUSED(color_desc); + SOKOL_ASSERT(color_desc->image.id != SG_INVALID_ID); + SOKOL_ASSERT(0 == atts->d3d11.colors[i].image); + SOKOL_ASSERT(color_images[i] && (color_images[i]->slot.id == color_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_color_format(color_images[i]->cmn.pixel_format)); + atts->d3d11.colors[i].image = color_images[i]; + + const sg_attachment_desc* resolve_desc = &desc->resolves[i]; + if (resolve_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(0 == atts->d3d11.resolves[i].image); + SOKOL_ASSERT(resolve_images[i] && (resolve_images[i]->slot.id == resolve_desc->image.id)); + SOKOL_ASSERT(color_images[i] && (color_images[i]->cmn.pixel_format == resolve_images[i]->cmn.pixel_format)); + atts->d3d11.resolves[i].image = resolve_images[i]; + } + } + SOKOL_ASSERT(0 == atts->d3d11.depth_stencil.image); + const sg_attachment_desc* ds_desc = &desc->depth_stencil; + if (ds_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(ds_img && (ds_img->slot.id == ds_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_depth_format(ds_img->cmn.pixel_format)); + atts->d3d11.depth_stencil.image = ds_img; + } + + // create render-target views + for (size_t i = 0; i < (size_t)atts->cmn.num_colors; i++) { + const _sg_attachment_common_t* cmn_color_att = &atts->cmn.colors[i]; + const _sg_image_t* color_img = color_images[i]; + SOKOL_ASSERT(0 == atts->d3d11.colors[i].view.rtv); + const bool msaa = color_img->cmn.sample_count > 1; + D3D11_RENDER_TARGET_VIEW_DESC d3d11_rtv_desc; + _sg_clear(&d3d11_rtv_desc, sizeof(d3d11_rtv_desc)); + d3d11_rtv_desc.Format = _sg_d3d11_rtv_pixel_format(color_img->cmn.pixel_format); + if (color_img->cmn.type == SG_IMAGETYPE_2D) { + if (msaa) { + d3d11_rtv_desc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DMS; + } else { + d3d11_rtv_desc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D; + d3d11_rtv_desc.Texture2D.MipSlice = (UINT)cmn_color_att->mip_level; + } + } else if ((color_img->cmn.type == SG_IMAGETYPE_CUBE) || (color_img->cmn.type == SG_IMAGETYPE_ARRAY)) { + if (msaa) { + d3d11_rtv_desc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DMSARRAY; + d3d11_rtv_desc.Texture2DMSArray.FirstArraySlice = (UINT)cmn_color_att->slice; + d3d11_rtv_desc.Texture2DMSArray.ArraySize = 1; + } else { + d3d11_rtv_desc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + d3d11_rtv_desc.Texture2DArray.MipSlice = (UINT)cmn_color_att->mip_level; + d3d11_rtv_desc.Texture2DArray.FirstArraySlice = (UINT)cmn_color_att->slice; + d3d11_rtv_desc.Texture2DArray.ArraySize = 1; + } + } else { + SOKOL_ASSERT(color_img->cmn.type == SG_IMAGETYPE_3D); + SOKOL_ASSERT(!msaa); + d3d11_rtv_desc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE3D; + d3d11_rtv_desc.Texture3D.MipSlice = (UINT)cmn_color_att->mip_level; + d3d11_rtv_desc.Texture3D.FirstWSlice = (UINT)cmn_color_att->slice; + d3d11_rtv_desc.Texture3D.WSize = 1; + } + SOKOL_ASSERT(color_img->d3d11.res); + HRESULT hr = _sg_d3d11_CreateRenderTargetView(_sg.d3d11.dev, color_img->d3d11.res, &d3d11_rtv_desc, &atts->d3d11.colors[i].view.rtv); + if (!(SUCCEEDED(hr) && atts->d3d11.colors[i].view.rtv)) { + _SG_ERROR(D3D11_CREATE_RTV_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(atts->d3d11.colors[i].view.rtv, desc->label); + } + SOKOL_ASSERT(0 == atts->d3d11.depth_stencil.view.dsv); + if (ds_desc->image.id != SG_INVALID_ID) { + const _sg_attachment_common_t* cmn_ds_att = &atts->cmn.depth_stencil; + const bool msaa = ds_img->cmn.sample_count > 1; + D3D11_DEPTH_STENCIL_VIEW_DESC d3d11_dsv_desc; + _sg_clear(&d3d11_dsv_desc, sizeof(d3d11_dsv_desc)); + d3d11_dsv_desc.Format = _sg_d3d11_dsv_pixel_format(ds_img->cmn.pixel_format); + SOKOL_ASSERT(ds_img && ds_img->cmn.type != SG_IMAGETYPE_3D); + if (ds_img->cmn.type == SG_IMAGETYPE_2D) { + if (msaa) { + d3d11_dsv_desc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DMS; + } else { + d3d11_dsv_desc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2D; + d3d11_dsv_desc.Texture2D.MipSlice = (UINT)cmn_ds_att->mip_level; + } + } else if ((ds_img->cmn.type == SG_IMAGETYPE_CUBE) || (ds_img->cmn.type == SG_IMAGETYPE_ARRAY)) { + if (msaa) { + d3d11_dsv_desc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DMSARRAY; + d3d11_dsv_desc.Texture2DMSArray.FirstArraySlice = (UINT)cmn_ds_att->slice; + d3d11_dsv_desc.Texture2DMSArray.ArraySize = 1; + } else { + d3d11_dsv_desc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DARRAY; + d3d11_dsv_desc.Texture2DArray.MipSlice = (UINT)cmn_ds_att->mip_level; + d3d11_dsv_desc.Texture2DArray.FirstArraySlice = (UINT)cmn_ds_att->slice; + d3d11_dsv_desc.Texture2DArray.ArraySize = 1; + } + } + SOKOL_ASSERT(ds_img->d3d11.res); + HRESULT hr = _sg_d3d11_CreateDepthStencilView(_sg.d3d11.dev, ds_img->d3d11.res, &d3d11_dsv_desc, &atts->d3d11.depth_stencil.view.dsv); + if (!(SUCCEEDED(hr) && atts->d3d11.depth_stencil.view.dsv)) { + _SG_ERROR(D3D11_CREATE_DSV_FAILED); + return SG_RESOURCESTATE_FAILED; + } + _sg_d3d11_setlabel(atts->d3d11.depth_stencil.view.dsv, desc->label); + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_d3d11_discard_attachments(_sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + for (size_t i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + if (atts->d3d11.colors[i].view.rtv) { + _sg_d3d11_Release(atts->d3d11.colors[i].view.rtv); + } + if (atts->d3d11.resolves[i].view.rtv) { + _sg_d3d11_Release(atts->d3d11.resolves[i].view.rtv); + } + } + if (atts->d3d11.depth_stencil.view.dsv) { + _sg_d3d11_Release(atts->d3d11.depth_stencil.view.dsv); + } +} + +_SOKOL_PRIVATE _sg_image_t* _sg_d3d11_attachments_color_image(const _sg_attachments_t* atts, int index) { + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + return atts->d3d11.colors[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_d3d11_attachments_resolve_image(const _sg_attachments_t* atts, int index) { + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + return atts->d3d11.resolves[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_d3d11_attachments_ds_image(const _sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + return atts->d3d11.depth_stencil.image; +} + +_SOKOL_PRIVATE void _sg_d3d11_begin_pass(const sg_pass* pass) { + SOKOL_ASSERT(pass); + if (_sg.cur_pass.is_compute) { + // nothing to do in compute passes + return; + } + const _sg_attachments_t* atts = _sg.cur_pass.atts; + const sg_swapchain* swapchain = &pass->swapchain; + const sg_pass_action* action = &pass->action; + + int num_rtvs = 0; + ID3D11RenderTargetView* rtvs[SG_MAX_COLOR_ATTACHMENTS] = { 0 }; + ID3D11DepthStencilView* dsv = 0; + _sg.d3d11.cur_pass.render_view = 0; + _sg.d3d11.cur_pass.resolve_view = 0; + if (atts) { + num_rtvs = atts->cmn.num_colors; + for (size_t i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + rtvs[i] = atts->d3d11.colors[i].view.rtv; + } + dsv = atts->d3d11.depth_stencil.view.dsv; + } else { + // NOTE: depth-stencil-view is optional + SOKOL_ASSERT(swapchain->d3d11.render_view); + num_rtvs = 1; + rtvs[0] = (ID3D11RenderTargetView*) swapchain->d3d11.render_view; + dsv = (ID3D11DepthStencilView*) swapchain->d3d11.depth_stencil_view; + _sg.d3d11.cur_pass.render_view = (ID3D11RenderTargetView*) swapchain->d3d11.render_view; + _sg.d3d11.cur_pass.resolve_view = (ID3D11RenderTargetView*) swapchain->d3d11.resolve_view; + } + // apply the render-target- and depth-stencil-views + _sg_d3d11_OMSetRenderTargets(_sg.d3d11.ctx, SG_MAX_COLOR_ATTACHMENTS, rtvs, dsv); + _sg_stats_add(d3d11.pass.num_om_set_render_targets, 1); + + // set viewport and scissor rect to cover whole screen + D3D11_VIEWPORT vp; + _sg_clear(&vp, sizeof(vp)); + vp.Width = (FLOAT) _sg.cur_pass.width; + vp.Height = (FLOAT) _sg.cur_pass.height; + vp.MaxDepth = 1.0f; + _sg_d3d11_RSSetViewports(_sg.d3d11.ctx, 1, &vp); + D3D11_RECT rect; + rect.left = 0; + rect.top = 0; + rect.right = _sg.cur_pass.width; + rect.bottom = _sg.cur_pass.height; + _sg_d3d11_RSSetScissorRects(_sg.d3d11.ctx, 1, &rect); + + // perform clear action + for (size_t i = 0; i < (size_t)num_rtvs; i++) { + if (action->colors[i].load_action == SG_LOADACTION_CLEAR) { + _sg_d3d11_ClearRenderTargetView(_sg.d3d11.ctx, rtvs[i], (float*)&action->colors[i].clear_value); + _sg_stats_add(d3d11.pass.num_clear_render_target_view, 1); + } + } + UINT ds_flags = 0; + if (action->depth.load_action == SG_LOADACTION_CLEAR) { + ds_flags |= D3D11_CLEAR_DEPTH; + } + if (action->stencil.load_action == SG_LOADACTION_CLEAR) { + ds_flags |= D3D11_CLEAR_STENCIL; + } + if ((0 != ds_flags) && dsv) { + _sg_d3d11_ClearDepthStencilView(_sg.d3d11.ctx, dsv, ds_flags, action->depth.clear_value, action->stencil.clear_value); + _sg_stats_add(d3d11.pass.num_clear_depth_stencil_view, 1); + } +} + +// D3D11CalcSubresource only exists for C++ +_SOKOL_PRIVATE UINT _sg_d3d11_calcsubresource(UINT mip_slice, UINT array_slice, UINT mip_levels) { + return mip_slice + array_slice * mip_levels; +} + +_SOKOL_PRIVATE void _sg_d3d11_end_pass(void) { + SOKOL_ASSERT(_sg.d3d11.ctx); + + if (!_sg.cur_pass.is_compute) { + // need to resolve MSAA render attachments into texture? + if (_sg.cur_pass.atts_id.id != SG_INVALID_ID) { + // ...for offscreen pass... + SOKOL_ASSERT(_sg.cur_pass.atts && _sg.cur_pass.atts->slot.id == _sg.cur_pass.atts_id.id); + for (size_t i = 0; i < (size_t)_sg.cur_pass.atts->cmn.num_colors; i++) { + const _sg_image_t* resolve_img = _sg.cur_pass.atts->d3d11.resolves[i].image; + if (resolve_img) { + const _sg_image_t* color_img = _sg.cur_pass.atts->d3d11.colors[i].image; + const _sg_attachment_common_t* cmn_color_att = &_sg.cur_pass.atts->cmn.colors[i]; + const _sg_attachment_common_t* cmn_resolve_att = &_sg.cur_pass.atts->cmn.resolves[i]; + SOKOL_ASSERT(resolve_img->slot.id == cmn_resolve_att->image_id.id); + SOKOL_ASSERT(color_img && (color_img->slot.id == cmn_color_att->image_id.id)); + SOKOL_ASSERT(color_img->cmn.sample_count > 1); + SOKOL_ASSERT(resolve_img->cmn.sample_count == 1); + const UINT src_subres = _sg_d3d11_calcsubresource( + (UINT)cmn_color_att->mip_level, + (UINT)cmn_color_att->slice, + (UINT)color_img->cmn.num_mipmaps); + const UINT dst_subres = _sg_d3d11_calcsubresource( + (UINT)cmn_resolve_att->mip_level, + (UINT)cmn_resolve_att->slice, + (UINT)resolve_img->cmn.num_mipmaps); + _sg_d3d11_ResolveSubresource(_sg.d3d11.ctx, + resolve_img->d3d11.res, + dst_subres, + color_img->d3d11.res, + src_subres, + color_img->d3d11.format); + _sg_stats_add(d3d11.pass.num_resolve_subresource, 1); + } + } + } else { + // ...for swapchain pass... + if (_sg.d3d11.cur_pass.resolve_view) { + SOKOL_ASSERT(_sg.d3d11.cur_pass.render_view); + SOKOL_ASSERT(_sg.cur_pass.swapchain.sample_count > 1); + SOKOL_ASSERT(_sg.cur_pass.swapchain.color_fmt > SG_PIXELFORMAT_NONE); + ID3D11Resource* d3d11_render_res = 0; + ID3D11Resource* d3d11_resolve_res = 0; + _sg_d3d11_GetResource((ID3D11View*)_sg.d3d11.cur_pass.render_view, &d3d11_render_res); + _sg_d3d11_GetResource((ID3D11View*)_sg.d3d11.cur_pass.resolve_view, &d3d11_resolve_res); + SOKOL_ASSERT(d3d11_render_res); + SOKOL_ASSERT(d3d11_resolve_res); + const sg_pixel_format color_fmt = _sg.cur_pass.swapchain.color_fmt; + _sg_d3d11_ResolveSubresource(_sg.d3d11.ctx, d3d11_resolve_res, 0, d3d11_render_res, 0, _sg_d3d11_rtv_pixel_format(color_fmt)); + _sg_d3d11_Release(d3d11_render_res); + _sg_d3d11_Release(d3d11_resolve_res); + _sg_stats_add(d3d11.pass.num_resolve_subresource, 1); + } + } + } + _sg.d3d11.cur_pass.render_view = 0; + _sg.d3d11.cur_pass.resolve_view = 0; + _sg.d3d11.cur_pipeline = 0; + _sg.d3d11.cur_pipeline_id.id = SG_INVALID_ID; + _sg_d3d11_clear_state(); +} + +_SOKOL_PRIVATE void _sg_d3d11_apply_viewport(int x, int y, int w, int h, bool origin_top_left) { + SOKOL_ASSERT(_sg.d3d11.ctx); + D3D11_VIEWPORT vp; + vp.TopLeftX = (FLOAT) x; + vp.TopLeftY = (FLOAT) (origin_top_left ? y : (_sg.cur_pass.height - (y + h))); + vp.Width = (FLOAT) w; + vp.Height = (FLOAT) h; + vp.MinDepth = 0.0f; + vp.MaxDepth = 1.0f; + _sg_d3d11_RSSetViewports(_sg.d3d11.ctx, 1, &vp); +} + +_SOKOL_PRIVATE void _sg_d3d11_apply_scissor_rect(int x, int y, int w, int h, bool origin_top_left) { + SOKOL_ASSERT(_sg.d3d11.ctx); + D3D11_RECT rect; + rect.left = x; + rect.top = (origin_top_left ? y : (_sg.cur_pass.height - (y + h))); + rect.right = x + w; + rect.bottom = origin_top_left ? (y + h) : (_sg.cur_pass.height - y); + _sg_d3d11_RSSetScissorRects(_sg.d3d11.ctx, 1, &rect); +} + +_SOKOL_PRIVATE void _sg_d3d11_apply_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + SOKOL_ASSERT(pip->shader && (pip->cmn.shader_id.id == pip->shader->slot.id)); + SOKOL_ASSERT(_sg.d3d11.ctx); + + _sg.d3d11.cur_pipeline = pip; + _sg.d3d11.cur_pipeline_id.id = pip->slot.id; + + if (pip->cmn.is_compute) { + // a compute pipeline + SOKOL_ASSERT(pip->shader->d3d11.cs); + _sg_d3d11_CSSetShader(_sg.d3d11.ctx, pip->shader->d3d11.cs, NULL, 0); + _sg_d3d11_CSSetConstantBuffers(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_UB_BINDINGS, pip->shader->d3d11.cs_cbufs); + _sg_stats_add(d3d11.pipeline.num_cs_set_shader, 1); + _sg_stats_add(d3d11.pipeline.num_cs_set_constant_buffers, 1); + } else { + // a render pipeline + SOKOL_ASSERT(pip->d3d11.rs && pip->d3d11.bs && pip->d3d11.dss); + SOKOL_ASSERT(pip->shader->d3d11.vs); + SOKOL_ASSERT(pip->shader->d3d11.fs); + + _sg.d3d11.use_indexed_draw = (pip->d3d11.index_format != DXGI_FORMAT_UNKNOWN); + _sg.d3d11.use_instanced_draw = pip->cmn.use_instanced_draw; + + _sg_d3d11_RSSetState(_sg.d3d11.ctx, pip->d3d11.rs); + _sg_d3d11_OMSetDepthStencilState(_sg.d3d11.ctx, pip->d3d11.dss, pip->d3d11.stencil_ref); + _sg_d3d11_OMSetBlendState(_sg.d3d11.ctx, pip->d3d11.bs, (float*)&pip->cmn.blend_color, 0xFFFFFFFF); + _sg_d3d11_IASetPrimitiveTopology(_sg.d3d11.ctx, pip->d3d11.topology); + _sg_d3d11_IASetInputLayout(_sg.d3d11.ctx, pip->d3d11.il); + _sg_d3d11_VSSetShader(_sg.d3d11.ctx, pip->shader->d3d11.vs, NULL, 0); + _sg_d3d11_VSSetConstantBuffers(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_UB_BINDINGS, pip->shader->d3d11.vs_cbufs); + _sg_d3d11_PSSetShader(_sg.d3d11.ctx, pip->shader->d3d11.fs, NULL, 0); + _sg_d3d11_PSSetConstantBuffers(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_UB_BINDINGS, pip->shader->d3d11.fs_cbufs); + _sg_stats_add(d3d11.pipeline.num_rs_set_state, 1); + _sg_stats_add(d3d11.pipeline.num_om_set_depth_stencil_state, 1); + _sg_stats_add(d3d11.pipeline.num_om_set_blend_state, 1); + _sg_stats_add(d3d11.pipeline.num_ia_set_primitive_topology, 1); + _sg_stats_add(d3d11.pipeline.num_ia_set_input_layout, 1); + _sg_stats_add(d3d11.pipeline.num_vs_set_shader, 1); + _sg_stats_add(d3d11.pipeline.num_vs_set_constant_buffers, 1); + _sg_stats_add(d3d11.pipeline.num_ps_set_shader, 1); + _sg_stats_add(d3d11.pipeline.num_ps_set_constant_buffers, 1); + } +} + +_SOKOL_PRIVATE bool _sg_d3d11_apply_bindings(_sg_bindings_t* bnd) { + SOKOL_ASSERT(bnd); + SOKOL_ASSERT(bnd->pip && bnd->pip->shader); + SOKOL_ASSERT(bnd->pip->shader->slot.id == bnd->pip->cmn.shader_id.id); + SOKOL_ASSERT(_sg.d3d11.ctx); + const _sg_shader_t* shd = bnd->pip->shader; + const bool is_compute = bnd->pip->cmn.is_compute; + + // gather all the D3D11 resources into arrays + ID3D11Buffer* d3d11_ib = bnd->ib ? bnd->ib->d3d11.buf : 0; + ID3D11Buffer* d3d11_vbs[SG_MAX_VERTEXBUFFER_BINDSLOTS] = {0}; + UINT d3d11_vb_offsets[SG_MAX_VERTEXBUFFER_BINDSLOTS] = {0}; + ID3D11ShaderResourceView* d3d11_vs_srvs[_SG_D3D11_MAX_STAGE_SRV_BINDINGS] = {0}; + ID3D11ShaderResourceView* d3d11_fs_srvs[_SG_D3D11_MAX_STAGE_SRV_BINDINGS] = {0}; + ID3D11ShaderResourceView* d3d11_cs_srvs[_SG_D3D11_MAX_STAGE_SRV_BINDINGS] = {0}; + ID3D11SamplerState* d3d11_vs_smps[_SG_D3D11_MAX_STAGE_SMP_BINDINGS] = {0}; + ID3D11SamplerState* d3d11_fs_smps[_SG_D3D11_MAX_STAGE_SMP_BINDINGS] = {0}; + ID3D11SamplerState* d3d11_cs_smps[_SG_D3D11_MAX_STAGE_SMP_BINDINGS] = {0}; + ID3D11UnorderedAccessView* d3d11_cs_uavs[_SG_D3D11_MAX_STAGE_UAV_BINDINGS] = {0}; + + if (!is_compute) { + for (size_t i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + const _sg_buffer_t* vb = bnd->vbs[i]; + if (vb == 0) { + continue; + } + SOKOL_ASSERT(vb->d3d11.buf); + d3d11_vbs[i] = vb->d3d11.buf; + d3d11_vb_offsets[i] = (UINT)bnd->vb_offsets[i]; + } + } + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + const _sg_image_t* img = bnd->imgs[i]; + if (img == 0) { + continue; + } + const sg_shader_stage stage = shd->cmn.images[i].stage; + SOKOL_ASSERT(stage != SG_SHADERSTAGE_NONE); + const uint8_t d3d11_slot = shd->d3d11.img_register_t_n[i]; + SOKOL_ASSERT(d3d11_slot < _SG_D3D11_MAX_STAGE_SRV_BINDINGS); + SOKOL_ASSERT(img->d3d11.srv); + ID3D11ShaderResourceView* d3d11_srv = img->d3d11.srv; + switch (stage) { + case SG_SHADERSTAGE_VERTEX: d3d11_vs_srvs[d3d11_slot] = d3d11_srv; break; + case SG_SHADERSTAGE_FRAGMENT: d3d11_fs_srvs[d3d11_slot] = d3d11_srv; break; + case SG_SHADERSTAGE_COMPUTE: d3d11_cs_srvs[d3d11_slot] = d3d11_srv; break; + default: SOKOL_UNREACHABLE; + } + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + const _sg_buffer_t* sbuf = bnd->sbufs[i]; + if (sbuf == 0) { + continue; + } + const sg_shader_stage stage = shd->cmn.storage_buffers[i].stage; + SOKOL_ASSERT(stage != SG_SHADERSTAGE_NONE); + if (shd->cmn.storage_buffers[i].readonly) { + SOKOL_ASSERT(sbuf->d3d11.srv); + const uint8_t d3d11_slot = shd->d3d11.sbuf_register_t_n[i]; + SOKOL_ASSERT(d3d11_slot < _SG_D3D11_MAX_STAGE_SRV_BINDINGS); + ID3D11ShaderResourceView* d3d11_srv = sbuf->d3d11.srv; + switch (stage) { + case SG_SHADERSTAGE_VERTEX: d3d11_vs_srvs[d3d11_slot] = d3d11_srv; break; + case SG_SHADERSTAGE_FRAGMENT: d3d11_fs_srvs[d3d11_slot] = d3d11_srv; break; + case SG_SHADERSTAGE_COMPUTE: d3d11_cs_srvs[d3d11_slot] = d3d11_srv; break; + default: SOKOL_UNREACHABLE; + } + } else { + SOKOL_ASSERT(sbuf->d3d11.uav); + SOKOL_ASSERT(stage == SG_SHADERSTAGE_COMPUTE); + const uint8_t d3d11_slot = shd->d3d11.sbuf_register_u_n[i]; + SOKOL_ASSERT(d3d11_slot < _SG_D3D11_MAX_STAGE_UAV_BINDINGS); + d3d11_cs_uavs[d3d11_slot] = sbuf->d3d11.uav; + } + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + const _sg_sampler_t* smp = bnd->smps[i]; + if (smp == 0) { + continue; + } + const sg_shader_stage stage = shd->cmn.samplers[i].stage; + SOKOL_ASSERT(stage != SG_SHADERSTAGE_NONE); + const uint8_t d3d11_slot = shd->d3d11.smp_register_s_n[i]; + SOKOL_ASSERT(d3d11_slot < _SG_D3D11_MAX_STAGE_SMP_BINDINGS); + SOKOL_ASSERT(smp->d3d11.smp); + ID3D11SamplerState* d3d11_smp = smp->d3d11.smp; + switch (stage) { + case SG_SHADERSTAGE_VERTEX: d3d11_vs_smps[d3d11_slot] = d3d11_smp; break; + case SG_SHADERSTAGE_FRAGMENT: d3d11_fs_smps[d3d11_slot] = d3d11_smp; break; + case SG_SHADERSTAGE_COMPUTE: d3d11_cs_smps[d3d11_slot] = d3d11_smp; break; + default: SOKOL_UNREACHABLE; + } + } + if (is_compute) { + _sg_d3d11_CSSetShaderResources(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_SRV_BINDINGS, d3d11_cs_srvs); + _sg_d3d11_CSSetSamplers(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_SMP_BINDINGS, d3d11_cs_smps); + _sg_d3d11_CSSetUnorderedAccessViews(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_UAV_BINDINGS, d3d11_cs_uavs, NULL); + _sg_stats_add(d3d11.bindings.num_cs_set_shader_resources, 1); + _sg_stats_add(d3d11.bindings.num_cs_set_samplers, 1); + _sg_stats_add(d3d11.bindings.num_cs_set_unordered_access_views, 1); + } else { + _sg_d3d11_IASetVertexBuffers(_sg.d3d11.ctx, 0, SG_MAX_VERTEXBUFFER_BINDSLOTS, d3d11_vbs, bnd->pip->d3d11.vb_strides, d3d11_vb_offsets); + _sg_d3d11_IASetIndexBuffer(_sg.d3d11.ctx, d3d11_ib, bnd->pip->d3d11.index_format, (UINT)bnd->ib_offset); + _sg_d3d11_VSSetShaderResources(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_SRV_BINDINGS, d3d11_vs_srvs); + _sg_d3d11_PSSetShaderResources(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_SRV_BINDINGS, d3d11_fs_srvs); + _sg_d3d11_VSSetSamplers(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_SMP_BINDINGS, d3d11_vs_smps); + _sg_d3d11_PSSetSamplers(_sg.d3d11.ctx, 0, _SG_D3D11_MAX_STAGE_SMP_BINDINGS, d3d11_fs_smps); + _sg_stats_add(d3d11.bindings.num_ia_set_vertex_buffers, 1); + _sg_stats_add(d3d11.bindings.num_ia_set_index_buffer, 1); + _sg_stats_add(d3d11.bindings.num_vs_set_shader_resources, 1); + _sg_stats_add(d3d11.bindings.num_ps_set_shader_resources, 1); + _sg_stats_add(d3d11.bindings.num_vs_set_samplers, 1); + _sg_stats_add(d3d11.bindings.num_ps_set_samplers, 1); + } + return true; +} + +_SOKOL_PRIVATE void _sg_d3d11_apply_uniforms(int ub_slot, const sg_range* data) { + SOKOL_ASSERT(_sg.d3d11.ctx); + SOKOL_ASSERT((ub_slot >= 0) && (ub_slot < SG_MAX_UNIFORMBLOCK_BINDSLOTS)); + SOKOL_ASSERT(_sg.d3d11.cur_pipeline && _sg.d3d11.cur_pipeline->slot.id == _sg.d3d11.cur_pipeline_id.id); + const _sg_shader_t* shd = _sg.d3d11.cur_pipeline->shader; + SOKOL_ASSERT(shd && (shd->slot.id == _sg.d3d11.cur_pipeline->cmn.shader_id.id)); + SOKOL_ASSERT(data->size == shd->cmn.uniform_blocks[ub_slot].size); + + ID3D11Buffer* cbuf = shd->d3d11.all_cbufs[ub_slot]; + SOKOL_ASSERT(cbuf); + _sg_d3d11_UpdateSubresource(_sg.d3d11.ctx, (ID3D11Resource*)cbuf, 0, NULL, data->ptr, 0, 0); + _sg_stats_add(d3d11.uniforms.num_update_subresource, 1); +} + +_SOKOL_PRIVATE void _sg_d3d11_draw(int base_element, int num_elements, int num_instances) { + const bool use_instanced_draw = (num_instances > 1) || (_sg.d3d11.use_instanced_draw); + if (_sg.d3d11.use_indexed_draw) { + if (use_instanced_draw) { + _sg_d3d11_DrawIndexedInstanced(_sg.d3d11.ctx, (UINT)num_elements, (UINT)num_instances, (UINT)base_element, 0, 0); + _sg_stats_add(d3d11.draw.num_draw_indexed_instanced, 1); + } else { + _sg_d3d11_DrawIndexed(_sg.d3d11.ctx, (UINT)num_elements, (UINT)base_element, 0); + _sg_stats_add(d3d11.draw.num_draw_indexed, 1); + } + } else { + if (use_instanced_draw) { + _sg_d3d11_DrawInstanced(_sg.d3d11.ctx, (UINT)num_elements, (UINT)num_instances, (UINT)base_element, 0); + _sg_stats_add(d3d11.draw.num_draw_instanced, 1); + } else { + _sg_d3d11_Draw(_sg.d3d11.ctx, (UINT)num_elements, (UINT)base_element); + _sg_stats_add(d3d11.draw.num_draw, 1); + } + } +} + +_SOKOL_PRIVATE void _sg_d3d11_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) { + _sg_d3d11_Dispatch(_sg.d3d11.ctx, (UINT)num_groups_x, (UINT)num_groups_y, (UINT)num_groups_z); +} + +_SOKOL_PRIVATE void _sg_d3d11_commit(void) { + // empty +} + +_SOKOL_PRIVATE void _sg_d3d11_update_buffer(_sg_buffer_t* buf, const sg_range* data) { + SOKOL_ASSERT(buf && data && data->ptr && (data->size > 0)); + SOKOL_ASSERT(_sg.d3d11.ctx); + SOKOL_ASSERT(buf->d3d11.buf); + D3D11_MAPPED_SUBRESOURCE d3d11_msr; + HRESULT hr = _sg_d3d11_Map(_sg.d3d11.ctx, (ID3D11Resource*)buf->d3d11.buf, 0, D3D11_MAP_WRITE_DISCARD, 0, &d3d11_msr); + _sg_stats_add(d3d11.num_map, 1); + if (SUCCEEDED(hr)) { + memcpy(d3d11_msr.pData, data->ptr, data->size); + _sg_d3d11_Unmap(_sg.d3d11.ctx, (ID3D11Resource*)buf->d3d11.buf, 0); + _sg_stats_add(d3d11.num_unmap, 1); + } else { + _SG_ERROR(D3D11_MAP_FOR_UPDATE_BUFFER_FAILED); + } +} + +_SOKOL_PRIVATE void _sg_d3d11_append_buffer(_sg_buffer_t* buf, const sg_range* data, bool new_frame) { + SOKOL_ASSERT(buf && data && data->ptr && (data->size > 0)); + SOKOL_ASSERT(_sg.d3d11.ctx); + SOKOL_ASSERT(buf->d3d11.buf); + D3D11_MAP map_type = new_frame ? D3D11_MAP_WRITE_DISCARD : D3D11_MAP_WRITE_NO_OVERWRITE; + D3D11_MAPPED_SUBRESOURCE d3d11_msr; + HRESULT hr = _sg_d3d11_Map(_sg.d3d11.ctx, (ID3D11Resource*)buf->d3d11.buf, 0, map_type, 0, &d3d11_msr); + _sg_stats_add(d3d11.num_map, 1); + if (SUCCEEDED(hr)) { + uint8_t* dst_ptr = (uint8_t*)d3d11_msr.pData + buf->cmn.append_pos; + memcpy(dst_ptr, data->ptr, data->size); + _sg_d3d11_Unmap(_sg.d3d11.ctx, (ID3D11Resource*)buf->d3d11.buf, 0); + _sg_stats_add(d3d11.num_unmap, 1); + } else { + _SG_ERROR(D3D11_MAP_FOR_APPEND_BUFFER_FAILED); + } +} + +// see: https://learn.microsoft.com/en-us/windows/win32/direct3d11/overviews-direct3d-11-resources-subresources +// also see: https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nf-d3d11-d3d11calcsubresource +_SOKOL_PRIVATE void _sg_d3d11_update_image(_sg_image_t* img, const sg_image_data* data) { + SOKOL_ASSERT(img && data); + SOKOL_ASSERT(_sg.d3d11.ctx); + SOKOL_ASSERT(img->d3d11.res); + const int num_faces = (img->cmn.type == SG_IMAGETYPE_CUBE) ? 6:1; + const int num_slices = (img->cmn.type == SG_IMAGETYPE_ARRAY) ? img->cmn.num_slices:1; + const int num_depth_slices = (img->cmn.type == SG_IMAGETYPE_3D) ? img->cmn.num_slices:1; + UINT subres_index = 0; + HRESULT hr; + D3D11_MAPPED_SUBRESOURCE d3d11_msr; + for (int face_index = 0; face_index < num_faces; face_index++) { + for (int slice_index = 0; slice_index < num_slices; slice_index++) { + for (int mip_index = 0; mip_index < img->cmn.num_mipmaps; mip_index++, subres_index++) { + SOKOL_ASSERT(subres_index < (SG_MAX_MIPMAPS * SG_MAX_TEXTUREARRAY_LAYERS)); + const int mip_width = _sg_miplevel_dim(img->cmn.width, mip_index); + const int mip_height = _sg_miplevel_dim(img->cmn.height, mip_index); + const int src_row_pitch = _sg_row_pitch(img->cmn.pixel_format, mip_width, 1); + const int src_depth_pitch = _sg_surface_pitch(img->cmn.pixel_format, mip_width, mip_height, 1); + const sg_range* subimg_data = &(data->subimage[face_index][mip_index]); + const size_t slice_size = subimg_data->size / (size_t)num_slices; + SOKOL_ASSERT(slice_size == (size_t)(src_depth_pitch * num_depth_slices)); + const size_t slice_offset = slice_size * (size_t)slice_index; + const uint8_t* slice_ptr = ((const uint8_t*)subimg_data->ptr) + slice_offset; + hr = _sg_d3d11_Map(_sg.d3d11.ctx, img->d3d11.res, subres_index, D3D11_MAP_WRITE_DISCARD, 0, &d3d11_msr); + _sg_stats_add(d3d11.num_map, 1); + if (SUCCEEDED(hr)) { + const uint8_t* src_ptr = slice_ptr; + uint8_t* dst_ptr = (uint8_t*)d3d11_msr.pData; + for (int depth_index = 0; depth_index < num_depth_slices; depth_index++) { + if (src_row_pitch == (int)d3d11_msr.RowPitch) { + const size_t copy_size = slice_size / (size_t)num_depth_slices; + SOKOL_ASSERT((copy_size * (size_t)num_depth_slices) == slice_size); + memcpy(dst_ptr, src_ptr, copy_size); + } else { + SOKOL_ASSERT(src_row_pitch < (int)d3d11_msr.RowPitch); + const uint8_t* src_row_ptr = src_ptr; + uint8_t* dst_row_ptr = dst_ptr; + for (int row_index = 0; row_index < mip_height; row_index++) { + memcpy(dst_row_ptr, src_row_ptr, (size_t)src_row_pitch); + src_row_ptr += src_row_pitch; + dst_row_ptr += d3d11_msr.RowPitch; + } + } + src_ptr += src_depth_pitch; + dst_ptr += d3d11_msr.DepthPitch; + } + _sg_d3d11_Unmap(_sg.d3d11.ctx, img->d3d11.res, subres_index); + _sg_stats_add(d3d11.num_unmap, 1); + } else { + _SG_ERROR(D3D11_MAP_FOR_UPDATE_IMAGE_FAILED); + } + } + } + } +} + +// ███ ███ ███████ ████████ █████ ██ ██████ █████ ██████ ██ ██ ███████ ███ ██ ██████ +// ████ ████ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ████ ██ ██ ██ +// ██ ████ ██ █████ ██ ███████ ██ ██████ ███████ ██ █████ █████ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██ ██ ███████ ██ ██ ██ ███████ ██████ ██ ██ ██████ ██ ██ ███████ ██ ████ ██████ +// +// >>metal backend +#elif defined(SOKOL_METAL) + +#if __has_feature(objc_arc) +#define _SG_OBJC_RETAIN(obj) { } +#define _SG_OBJC_RELEASE(obj) { obj = nil; } +#else +#define _SG_OBJC_RETAIN(obj) { [obj retain]; } +#define _SG_OBJC_RELEASE(obj) { [obj release]; obj = nil; } +#endif + +//-- enum translation functions ------------------------------------------------ +_SOKOL_PRIVATE MTLLoadAction _sg_mtl_load_action(sg_load_action a) { + switch (a) { + case SG_LOADACTION_CLEAR: return MTLLoadActionClear; + case SG_LOADACTION_LOAD: return MTLLoadActionLoad; + case SG_LOADACTION_DONTCARE: return MTLLoadActionDontCare; + default: SOKOL_UNREACHABLE; return (MTLLoadAction)0; + } +} + +_SOKOL_PRIVATE MTLStoreAction _sg_mtl_store_action(sg_store_action a, bool resolve) { + switch (a) { + case SG_STOREACTION_STORE: + if (resolve) { + return MTLStoreActionStoreAndMultisampleResolve; + } else { + return MTLStoreActionStore; + } + break; + case SG_STOREACTION_DONTCARE: + if (resolve) { + return MTLStoreActionMultisampleResolve; + } else { + return MTLStoreActionDontCare; + } + break; + default: SOKOL_UNREACHABLE; return (MTLStoreAction)0; + } +} + +_SOKOL_PRIVATE MTLResourceOptions _sg_mtl_resource_options_storage_mode_managed_or_shared(void) { + #if defined(_SG_TARGET_MACOS) + if (_sg.mtl.use_shared_storage_mode) { + return MTLResourceStorageModeShared; + } else { + return MTLResourceStorageModeManaged; + } + #else + // MTLResourceStorageModeManaged is not even defined on iOS SDK + return MTLResourceStorageModeShared; + #endif +} + +_SOKOL_PRIVATE MTLResourceOptions _sg_mtl_buffer_resource_options(sg_usage usg) { + switch (usg) { + case SG_USAGE_IMMUTABLE: + return _sg_mtl_resource_options_storage_mode_managed_or_shared(); + case SG_USAGE_DYNAMIC: + case SG_USAGE_STREAM: + return MTLResourceCPUCacheModeWriteCombined | _sg_mtl_resource_options_storage_mode_managed_or_shared(); + default: + SOKOL_UNREACHABLE; + return 0; + } +} + +_SOKOL_PRIVATE MTLVertexStepFunction _sg_mtl_step_function(sg_vertex_step step) { + switch (step) { + case SG_VERTEXSTEP_PER_VERTEX: return MTLVertexStepFunctionPerVertex; + case SG_VERTEXSTEP_PER_INSTANCE: return MTLVertexStepFunctionPerInstance; + default: SOKOL_UNREACHABLE; return (MTLVertexStepFunction)0; + } +} + +_SOKOL_PRIVATE MTLVertexFormat _sg_mtl_vertex_format(sg_vertex_format fmt) { + switch (fmt) { + case SG_VERTEXFORMAT_FLOAT: return MTLVertexFormatFloat; + case SG_VERTEXFORMAT_FLOAT2: return MTLVertexFormatFloat2; + case SG_VERTEXFORMAT_FLOAT3: return MTLVertexFormatFloat3; + case SG_VERTEXFORMAT_FLOAT4: return MTLVertexFormatFloat4; + case SG_VERTEXFORMAT_INT: return MTLVertexFormatInt; + case SG_VERTEXFORMAT_INT2: return MTLVertexFormatInt2; + case SG_VERTEXFORMAT_INT3: return MTLVertexFormatInt3; + case SG_VERTEXFORMAT_INT4: return MTLVertexFormatInt4; + case SG_VERTEXFORMAT_UINT: return MTLVertexFormatUInt; + case SG_VERTEXFORMAT_UINT2: return MTLVertexFormatUInt2; + case SG_VERTEXFORMAT_UINT3: return MTLVertexFormatUInt3; + case SG_VERTEXFORMAT_UINT4: return MTLVertexFormatUInt4; + case SG_VERTEXFORMAT_BYTE4: return MTLVertexFormatChar4; + case SG_VERTEXFORMAT_BYTE4N: return MTLVertexFormatChar4Normalized; + case SG_VERTEXFORMAT_UBYTE4: return MTLVertexFormatUChar4; + case SG_VERTEXFORMAT_UBYTE4N: return MTLVertexFormatUChar4Normalized; + case SG_VERTEXFORMAT_SHORT2: return MTLVertexFormatShort2; + case SG_VERTEXFORMAT_SHORT2N: return MTLVertexFormatShort2Normalized; + case SG_VERTEXFORMAT_USHORT2: return MTLVertexFormatUShort2; + case SG_VERTEXFORMAT_USHORT2N: return MTLVertexFormatUShort2Normalized; + case SG_VERTEXFORMAT_SHORT4: return MTLVertexFormatShort4; + case SG_VERTEXFORMAT_SHORT4N: return MTLVertexFormatShort4Normalized; + case SG_VERTEXFORMAT_USHORT4: return MTLVertexFormatUShort4; + case SG_VERTEXFORMAT_USHORT4N: return MTLVertexFormatUShort4Normalized; + case SG_VERTEXFORMAT_UINT10_N2: return MTLVertexFormatUInt1010102Normalized; + case SG_VERTEXFORMAT_HALF2: return MTLVertexFormatHalf2; + case SG_VERTEXFORMAT_HALF4: return MTLVertexFormatHalf4; + default: SOKOL_UNREACHABLE; return (MTLVertexFormat)0; + } +} + +_SOKOL_PRIVATE MTLPrimitiveType _sg_mtl_primitive_type(sg_primitive_type t) { + switch (t) { + case SG_PRIMITIVETYPE_POINTS: return MTLPrimitiveTypePoint; + case SG_PRIMITIVETYPE_LINES: return MTLPrimitiveTypeLine; + case SG_PRIMITIVETYPE_LINE_STRIP: return MTLPrimitiveTypeLineStrip; + case SG_PRIMITIVETYPE_TRIANGLES: return MTLPrimitiveTypeTriangle; + case SG_PRIMITIVETYPE_TRIANGLE_STRIP: return MTLPrimitiveTypeTriangleStrip; + default: SOKOL_UNREACHABLE; return (MTLPrimitiveType)0; + } +} + +_SOKOL_PRIVATE MTLPixelFormat _sg_mtl_pixel_format(sg_pixel_format fmt) { + switch (fmt) { + case SG_PIXELFORMAT_R8: return MTLPixelFormatR8Unorm; + case SG_PIXELFORMAT_R8SN: return MTLPixelFormatR8Snorm; + case SG_PIXELFORMAT_R8UI: return MTLPixelFormatR8Uint; + case SG_PIXELFORMAT_R8SI: return MTLPixelFormatR8Sint; + case SG_PIXELFORMAT_R16: return MTLPixelFormatR16Unorm; + case SG_PIXELFORMAT_R16SN: return MTLPixelFormatR16Snorm; + case SG_PIXELFORMAT_R16UI: return MTLPixelFormatR16Uint; + case SG_PIXELFORMAT_R16SI: return MTLPixelFormatR16Sint; + case SG_PIXELFORMAT_R16F: return MTLPixelFormatR16Float; + case SG_PIXELFORMAT_RG8: return MTLPixelFormatRG8Unorm; + case SG_PIXELFORMAT_RG8SN: return MTLPixelFormatRG8Snorm; + case SG_PIXELFORMAT_RG8UI: return MTLPixelFormatRG8Uint; + case SG_PIXELFORMAT_RG8SI: return MTLPixelFormatRG8Sint; + case SG_PIXELFORMAT_R32UI: return MTLPixelFormatR32Uint; + case SG_PIXELFORMAT_R32SI: return MTLPixelFormatR32Sint; + case SG_PIXELFORMAT_R32F: return MTLPixelFormatR32Float; + case SG_PIXELFORMAT_RG16: return MTLPixelFormatRG16Unorm; + case SG_PIXELFORMAT_RG16SN: return MTLPixelFormatRG16Snorm; + case SG_PIXELFORMAT_RG16UI: return MTLPixelFormatRG16Uint; + case SG_PIXELFORMAT_RG16SI: return MTLPixelFormatRG16Sint; + case SG_PIXELFORMAT_RG16F: return MTLPixelFormatRG16Float; + case SG_PIXELFORMAT_RGBA8: return MTLPixelFormatRGBA8Unorm; + case SG_PIXELFORMAT_SRGB8A8: return MTLPixelFormatRGBA8Unorm_sRGB; + case SG_PIXELFORMAT_RGBA8SN: return MTLPixelFormatRGBA8Snorm; + case SG_PIXELFORMAT_RGBA8UI: return MTLPixelFormatRGBA8Uint; + case SG_PIXELFORMAT_RGBA8SI: return MTLPixelFormatRGBA8Sint; + case SG_PIXELFORMAT_BGRA8: return MTLPixelFormatBGRA8Unorm; + case SG_PIXELFORMAT_RGB10A2: return MTLPixelFormatRGB10A2Unorm; + case SG_PIXELFORMAT_RG11B10F: return MTLPixelFormatRG11B10Float; + case SG_PIXELFORMAT_RGB9E5: return MTLPixelFormatRGB9E5Float; + case SG_PIXELFORMAT_RG32UI: return MTLPixelFormatRG32Uint; + case SG_PIXELFORMAT_RG32SI: return MTLPixelFormatRG32Sint; + case SG_PIXELFORMAT_RG32F: return MTLPixelFormatRG32Float; + case SG_PIXELFORMAT_RGBA16: return MTLPixelFormatRGBA16Unorm; + case SG_PIXELFORMAT_RGBA16SN: return MTLPixelFormatRGBA16Snorm; + case SG_PIXELFORMAT_RGBA16UI: return MTLPixelFormatRGBA16Uint; + case SG_PIXELFORMAT_RGBA16SI: return MTLPixelFormatRGBA16Sint; + case SG_PIXELFORMAT_RGBA16F: return MTLPixelFormatRGBA16Float; + case SG_PIXELFORMAT_RGBA32UI: return MTLPixelFormatRGBA32Uint; + case SG_PIXELFORMAT_RGBA32SI: return MTLPixelFormatRGBA32Sint; + case SG_PIXELFORMAT_RGBA32F: return MTLPixelFormatRGBA32Float; + case SG_PIXELFORMAT_DEPTH: return MTLPixelFormatDepth32Float; + case SG_PIXELFORMAT_DEPTH_STENCIL: return MTLPixelFormatDepth32Float_Stencil8; + #if defined(_SG_TARGET_MACOS) + case SG_PIXELFORMAT_BC1_RGBA: return MTLPixelFormatBC1_RGBA; + case SG_PIXELFORMAT_BC2_RGBA: return MTLPixelFormatBC2_RGBA; + case SG_PIXELFORMAT_BC3_RGBA: return MTLPixelFormatBC3_RGBA; + case SG_PIXELFORMAT_BC3_SRGBA: return MTLPixelFormatBC3_RGBA_sRGB; + case SG_PIXELFORMAT_BC4_R: return MTLPixelFormatBC4_RUnorm; + case SG_PIXELFORMAT_BC4_RSN: return MTLPixelFormatBC4_RSnorm; + case SG_PIXELFORMAT_BC5_RG: return MTLPixelFormatBC5_RGUnorm; + case SG_PIXELFORMAT_BC5_RGSN: return MTLPixelFormatBC5_RGSnorm; + case SG_PIXELFORMAT_BC6H_RGBF: return MTLPixelFormatBC6H_RGBFloat; + case SG_PIXELFORMAT_BC6H_RGBUF: return MTLPixelFormatBC6H_RGBUfloat; + case SG_PIXELFORMAT_BC7_RGBA: return MTLPixelFormatBC7_RGBAUnorm; + case SG_PIXELFORMAT_BC7_SRGBA: return MTLPixelFormatBC7_RGBAUnorm_sRGB; + #else + case SG_PIXELFORMAT_ETC2_RGB8: return MTLPixelFormatETC2_RGB8; + case SG_PIXELFORMAT_ETC2_SRGB8: return MTLPixelFormatETC2_RGB8_sRGB; + case SG_PIXELFORMAT_ETC2_RGB8A1: return MTLPixelFormatETC2_RGB8A1; + case SG_PIXELFORMAT_ETC2_RGBA8: return MTLPixelFormatEAC_RGBA8; + case SG_PIXELFORMAT_ETC2_SRGB8A8: return MTLPixelFormatEAC_RGBA8_sRGB; + case SG_PIXELFORMAT_EAC_R11: return MTLPixelFormatEAC_R11Unorm; + case SG_PIXELFORMAT_EAC_R11SN: return MTLPixelFormatEAC_R11Snorm; + case SG_PIXELFORMAT_EAC_RG11: return MTLPixelFormatEAC_RG11Unorm; + case SG_PIXELFORMAT_EAC_RG11SN: return MTLPixelFormatEAC_RG11Snorm; + case SG_PIXELFORMAT_ASTC_4x4_RGBA: return MTLPixelFormatASTC_4x4_LDR; + case SG_PIXELFORMAT_ASTC_4x4_SRGBA: return MTLPixelFormatASTC_4x4_sRGB; + #endif + default: return MTLPixelFormatInvalid; + } +} + +_SOKOL_PRIVATE MTLColorWriteMask _sg_mtl_color_write_mask(sg_color_mask m) { + MTLColorWriteMask mtl_mask = MTLColorWriteMaskNone; + if (m & SG_COLORMASK_R) { + mtl_mask |= MTLColorWriteMaskRed; + } + if (m & SG_COLORMASK_G) { + mtl_mask |= MTLColorWriteMaskGreen; + } + if (m & SG_COLORMASK_B) { + mtl_mask |= MTLColorWriteMaskBlue; + } + if (m & SG_COLORMASK_A) { + mtl_mask |= MTLColorWriteMaskAlpha; + } + return mtl_mask; +} + +_SOKOL_PRIVATE MTLBlendOperation _sg_mtl_blend_op(sg_blend_op op) { + switch (op) { + case SG_BLENDOP_ADD: return MTLBlendOperationAdd; + case SG_BLENDOP_SUBTRACT: return MTLBlendOperationSubtract; + case SG_BLENDOP_REVERSE_SUBTRACT: return MTLBlendOperationReverseSubtract; + case SG_BLENDOP_MIN: return MTLBlendOperationMin; + case SG_BLENDOP_MAX: return MTLBlendOperationMax; + default: SOKOL_UNREACHABLE; return (MTLBlendOperation)0; + } +} + +_SOKOL_PRIVATE MTLBlendFactor _sg_mtl_blend_factor(sg_blend_factor f) { + switch (f) { + case SG_BLENDFACTOR_ZERO: return MTLBlendFactorZero; + case SG_BLENDFACTOR_ONE: return MTLBlendFactorOne; + case SG_BLENDFACTOR_SRC_COLOR: return MTLBlendFactorSourceColor; + case SG_BLENDFACTOR_ONE_MINUS_SRC_COLOR: return MTLBlendFactorOneMinusSourceColor; + case SG_BLENDFACTOR_SRC_ALPHA: return MTLBlendFactorSourceAlpha; + case SG_BLENDFACTOR_ONE_MINUS_SRC_ALPHA: return MTLBlendFactorOneMinusSourceAlpha; + case SG_BLENDFACTOR_DST_COLOR: return MTLBlendFactorDestinationColor; + case SG_BLENDFACTOR_ONE_MINUS_DST_COLOR: return MTLBlendFactorOneMinusDestinationColor; + case SG_BLENDFACTOR_DST_ALPHA: return MTLBlendFactorDestinationAlpha; + case SG_BLENDFACTOR_ONE_MINUS_DST_ALPHA: return MTLBlendFactorOneMinusDestinationAlpha; + case SG_BLENDFACTOR_SRC_ALPHA_SATURATED: return MTLBlendFactorSourceAlphaSaturated; + case SG_BLENDFACTOR_BLEND_COLOR: return MTLBlendFactorBlendColor; + case SG_BLENDFACTOR_ONE_MINUS_BLEND_COLOR: return MTLBlendFactorOneMinusBlendColor; + case SG_BLENDFACTOR_BLEND_ALPHA: return MTLBlendFactorBlendAlpha; + case SG_BLENDFACTOR_ONE_MINUS_BLEND_ALPHA: return MTLBlendFactorOneMinusBlendAlpha; + default: SOKOL_UNREACHABLE; return (MTLBlendFactor)0; + } +} + +_SOKOL_PRIVATE MTLCompareFunction _sg_mtl_compare_func(sg_compare_func f) { + switch (f) { + case SG_COMPAREFUNC_NEVER: return MTLCompareFunctionNever; + case SG_COMPAREFUNC_LESS: return MTLCompareFunctionLess; + case SG_COMPAREFUNC_EQUAL: return MTLCompareFunctionEqual; + case SG_COMPAREFUNC_LESS_EQUAL: return MTLCompareFunctionLessEqual; + case SG_COMPAREFUNC_GREATER: return MTLCompareFunctionGreater; + case SG_COMPAREFUNC_NOT_EQUAL: return MTLCompareFunctionNotEqual; + case SG_COMPAREFUNC_GREATER_EQUAL: return MTLCompareFunctionGreaterEqual; + case SG_COMPAREFUNC_ALWAYS: return MTLCompareFunctionAlways; + default: SOKOL_UNREACHABLE; return (MTLCompareFunction)0; + } +} + +_SOKOL_PRIVATE MTLStencilOperation _sg_mtl_stencil_op(sg_stencil_op op) { + switch (op) { + case SG_STENCILOP_KEEP: return MTLStencilOperationKeep; + case SG_STENCILOP_ZERO: return MTLStencilOperationZero; + case SG_STENCILOP_REPLACE: return MTLStencilOperationReplace; + case SG_STENCILOP_INCR_CLAMP: return MTLStencilOperationIncrementClamp; + case SG_STENCILOP_DECR_CLAMP: return MTLStencilOperationDecrementClamp; + case SG_STENCILOP_INVERT: return MTLStencilOperationInvert; + case SG_STENCILOP_INCR_WRAP: return MTLStencilOperationIncrementWrap; + case SG_STENCILOP_DECR_WRAP: return MTLStencilOperationDecrementWrap; + default: SOKOL_UNREACHABLE; return (MTLStencilOperation)0; + } +} + +_SOKOL_PRIVATE MTLCullMode _sg_mtl_cull_mode(sg_cull_mode m) { + switch (m) { + case SG_CULLMODE_NONE: return MTLCullModeNone; + case SG_CULLMODE_FRONT: return MTLCullModeFront; + case SG_CULLMODE_BACK: return MTLCullModeBack; + default: SOKOL_UNREACHABLE; return (MTLCullMode)0; + } +} + +_SOKOL_PRIVATE MTLWinding _sg_mtl_winding(sg_face_winding w) { + switch (w) { + case SG_FACEWINDING_CW: return MTLWindingClockwise; + case SG_FACEWINDING_CCW: return MTLWindingCounterClockwise; + default: SOKOL_UNREACHABLE; return (MTLWinding)0; + } +} + +_SOKOL_PRIVATE MTLIndexType _sg_mtl_index_type(sg_index_type t) { + switch (t) { + case SG_INDEXTYPE_UINT16: return MTLIndexTypeUInt16; + case SG_INDEXTYPE_UINT32: return MTLIndexTypeUInt32; + default: SOKOL_UNREACHABLE; return (MTLIndexType)0; + } +} + +_SOKOL_PRIVATE int _sg_mtl_index_size(sg_index_type t) { + switch (t) { + case SG_INDEXTYPE_NONE: return 0; + case SG_INDEXTYPE_UINT16: return 2; + case SG_INDEXTYPE_UINT32: return 4; + default: SOKOL_UNREACHABLE; return 0; + } +} + +_SOKOL_PRIVATE MTLTextureType _sg_mtl_texture_type(sg_image_type t) { + switch (t) { + case SG_IMAGETYPE_2D: return MTLTextureType2D; + case SG_IMAGETYPE_CUBE: return MTLTextureTypeCube; + case SG_IMAGETYPE_3D: return MTLTextureType3D; + case SG_IMAGETYPE_ARRAY: return MTLTextureType2DArray; + default: SOKOL_UNREACHABLE; return (MTLTextureType)0; + } +} + +_SOKOL_PRIVATE MTLSamplerAddressMode _sg_mtl_address_mode(sg_wrap w) { + if (_sg.features.image_clamp_to_border) { + if (@available(macOS 12.0, iOS 14.0, *)) { + // border color feature available + switch (w) { + case SG_WRAP_REPEAT: return MTLSamplerAddressModeRepeat; + case SG_WRAP_CLAMP_TO_EDGE: return MTLSamplerAddressModeClampToEdge; + case SG_WRAP_CLAMP_TO_BORDER: return MTLSamplerAddressModeClampToBorderColor; + case SG_WRAP_MIRRORED_REPEAT: return MTLSamplerAddressModeMirrorRepeat; + default: SOKOL_UNREACHABLE; return (MTLSamplerAddressMode)0; + } + } + } + // fallthrough: clamp to border no supported + switch (w) { + case SG_WRAP_REPEAT: return MTLSamplerAddressModeRepeat; + case SG_WRAP_CLAMP_TO_EDGE: return MTLSamplerAddressModeClampToEdge; + case SG_WRAP_CLAMP_TO_BORDER: return MTLSamplerAddressModeClampToEdge; + case SG_WRAP_MIRRORED_REPEAT: return MTLSamplerAddressModeMirrorRepeat; + default: SOKOL_UNREACHABLE; return (MTLSamplerAddressMode)0; + } +} + +_SOKOL_PRIVATE API_AVAILABLE(ios(14.0), macos(12.0)) MTLSamplerBorderColor _sg_mtl_border_color(sg_border_color c) { + switch (c) { + case SG_BORDERCOLOR_TRANSPARENT_BLACK: return MTLSamplerBorderColorTransparentBlack; + case SG_BORDERCOLOR_OPAQUE_BLACK: return MTLSamplerBorderColorOpaqueBlack; + case SG_BORDERCOLOR_OPAQUE_WHITE: return MTLSamplerBorderColorOpaqueWhite; + default: SOKOL_UNREACHABLE; return (MTLSamplerBorderColor)0; + } +} + +_SOKOL_PRIVATE MTLSamplerMinMagFilter _sg_mtl_minmag_filter(sg_filter f) { + switch (f) { + case SG_FILTER_NEAREST: + return MTLSamplerMinMagFilterNearest; + case SG_FILTER_LINEAR: + return MTLSamplerMinMagFilterLinear; + default: + SOKOL_UNREACHABLE; return (MTLSamplerMinMagFilter)0; + } +} + +_SOKOL_PRIVATE MTLSamplerMipFilter _sg_mtl_mipmap_filter(sg_filter f) { + switch (f) { + case SG_FILTER_NEAREST: + return MTLSamplerMipFilterNearest; + case SG_FILTER_LINEAR: + return MTLSamplerMipFilterLinear; + default: + SOKOL_UNREACHABLE; return (MTLSamplerMipFilter)0; + } +} + +_SOKOL_PRIVATE size_t _sg_mtl_vertexbuffer_bindslot(size_t sokol_bindslot) { + return sokol_bindslot + _SG_MTL_MAX_STAGE_UB_SBUF_BINDINGS; +} + +//-- a pool for all Metal resource objects, with deferred release queue --------- +_SOKOL_PRIVATE void _sg_mtl_init_pool(const sg_desc* desc) { + _sg.mtl.idpool.num_slots = 2 * + ( + 2 * desc->buffer_pool_size + + 4 * desc->image_pool_size + + 1 * desc->sampler_pool_size + + 4 * desc->shader_pool_size + + 2 * desc->pipeline_pool_size + + desc->attachments_pool_size + + 128 + ); + _sg.mtl.idpool.pool = [NSMutableArray arrayWithCapacity:(NSUInteger)_sg.mtl.idpool.num_slots]; + _SG_OBJC_RETAIN(_sg.mtl.idpool.pool); + NSNull* null = [NSNull null]; + for (int i = 0; i < _sg.mtl.idpool.num_slots; i++) { + [_sg.mtl.idpool.pool addObject:null]; + } + SOKOL_ASSERT([_sg.mtl.idpool.pool count] == (NSUInteger)_sg.mtl.idpool.num_slots); + // a queue of currently free slot indices + _sg.mtl.idpool.free_queue_top = 0; + _sg.mtl.idpool.free_queue = (int*)_sg_malloc_clear((size_t)_sg.mtl.idpool.num_slots * sizeof(int)); + // pool slot 0 is reserved! + for (int i = _sg.mtl.idpool.num_slots-1; i >= 1; i--) { + _sg.mtl.idpool.free_queue[_sg.mtl.idpool.free_queue_top++] = i; + } + // a circular queue which holds release items (frame index when a resource is to be released, and the resource's pool index + _sg.mtl.idpool.release_queue_front = 0; + _sg.mtl.idpool.release_queue_back = 0; + _sg.mtl.idpool.release_queue = (_sg_mtl_release_item_t*)_sg_malloc_clear((size_t)_sg.mtl.idpool.num_slots * sizeof(_sg_mtl_release_item_t)); + for (int i = 0; i < _sg.mtl.idpool.num_slots; i++) { + _sg.mtl.idpool.release_queue[i].frame_index = 0; + _sg.mtl.idpool.release_queue[i].slot_index = _SG_MTL_INVALID_SLOT_INDEX; + } +} + +_SOKOL_PRIVATE void _sg_mtl_destroy_pool(void) { + _sg_free(_sg.mtl.idpool.release_queue); _sg.mtl.idpool.release_queue = 0; + _sg_free(_sg.mtl.idpool.free_queue); _sg.mtl.idpool.free_queue = 0; + _SG_OBJC_RELEASE(_sg.mtl.idpool.pool); +} + +// get a new free resource pool slot +_SOKOL_PRIVATE int _sg_mtl_alloc_pool_slot(void) { + SOKOL_ASSERT(_sg.mtl.idpool.free_queue_top > 0); + const int slot_index = _sg.mtl.idpool.free_queue[--_sg.mtl.idpool.free_queue_top]; + SOKOL_ASSERT((slot_index > 0) && (slot_index < _sg.mtl.idpool.num_slots)); + return slot_index; +} + +// put a free resource pool slot back into the free-queue +_SOKOL_PRIVATE void _sg_mtl_free_pool_slot(int slot_index) { + SOKOL_ASSERT(_sg.mtl.idpool.free_queue_top < _sg.mtl.idpool.num_slots); + SOKOL_ASSERT((slot_index > 0) && (slot_index < _sg.mtl.idpool.num_slots)); + _sg.mtl.idpool.free_queue[_sg.mtl.idpool.free_queue_top++] = slot_index; +} + +// add an MTLResource to the pool, return pool index or 0 if input was 'nil' +_SOKOL_PRIVATE int _sg_mtl_add_resource(id res) { + if (nil == res) { + return _SG_MTL_INVALID_SLOT_INDEX; + } + _sg_stats_add(metal.idpool.num_added, 1); + const int slot_index = _sg_mtl_alloc_pool_slot(); + // NOTE: the NSMutableArray will take ownership of its items + SOKOL_ASSERT([NSNull null] == _sg.mtl.idpool.pool[(NSUInteger)slot_index]); + _sg.mtl.idpool.pool[(NSUInteger)slot_index] = res; + return slot_index; +} + +/* mark an MTLResource for release, this will put the resource into the + deferred-release queue, and the resource will then be released N frames later, + the special pool index 0 will be ignored (this means that a nil + value was provided to _sg_mtl_add_resource() +*/ +_SOKOL_PRIVATE void _sg_mtl_release_resource(uint32_t frame_index, int slot_index) { + if (slot_index == _SG_MTL_INVALID_SLOT_INDEX) { + return; + } + _sg_stats_add(metal.idpool.num_released, 1); + SOKOL_ASSERT((slot_index > 0) && (slot_index < _sg.mtl.idpool.num_slots)); + SOKOL_ASSERT([NSNull null] != _sg.mtl.idpool.pool[(NSUInteger)slot_index]); + int release_index = _sg.mtl.idpool.release_queue_front++; + if (_sg.mtl.idpool.release_queue_front >= _sg.mtl.idpool.num_slots) { + // wrap-around + _sg.mtl.idpool.release_queue_front = 0; + } + // release queue full? + SOKOL_ASSERT(_sg.mtl.idpool.release_queue_front != _sg.mtl.idpool.release_queue_back); + SOKOL_ASSERT(0 == _sg.mtl.idpool.release_queue[release_index].frame_index); + const uint32_t safe_to_release_frame_index = frame_index + SG_NUM_INFLIGHT_FRAMES + 1; + _sg.mtl.idpool.release_queue[release_index].frame_index = safe_to_release_frame_index; + _sg.mtl.idpool.release_queue[release_index].slot_index = slot_index; +} + +// run garbage-collection pass on all resources in the release-queue +_SOKOL_PRIVATE void _sg_mtl_garbage_collect(uint32_t frame_index) { + while (_sg.mtl.idpool.release_queue_back != _sg.mtl.idpool.release_queue_front) { + if (frame_index < _sg.mtl.idpool.release_queue[_sg.mtl.idpool.release_queue_back].frame_index) { + // don't need to check further, release-items past this are too young + break; + } + _sg_stats_add(metal.idpool.num_garbage_collected, 1); + // safe to release this resource + const int slot_index = _sg.mtl.idpool.release_queue[_sg.mtl.idpool.release_queue_back].slot_index; + SOKOL_ASSERT((slot_index > 0) && (slot_index < _sg.mtl.idpool.num_slots)); + // note: the NSMutableArray takes ownership of its items, assigning an NSNull object will + // release the object, no matter if using ARC or not + SOKOL_ASSERT(_sg.mtl.idpool.pool[(NSUInteger)slot_index] != [NSNull null]); + _sg.mtl.idpool.pool[(NSUInteger)slot_index] = [NSNull null]; + // put the now free pool index back on the free queue + _sg_mtl_free_pool_slot(slot_index); + // reset the release queue slot and advance the back index + _sg.mtl.idpool.release_queue[_sg.mtl.idpool.release_queue_back].frame_index = 0; + _sg.mtl.idpool.release_queue[_sg.mtl.idpool.release_queue_back].slot_index = _SG_MTL_INVALID_SLOT_INDEX; + _sg.mtl.idpool.release_queue_back++; + if (_sg.mtl.idpool.release_queue_back >= _sg.mtl.idpool.num_slots) { + // wrap-around + _sg.mtl.idpool.release_queue_back = 0; + } + } +} + +_SOKOL_PRIVATE id _sg_mtl_id(int slot_index) { + return _sg.mtl.idpool.pool[(NSUInteger)slot_index]; +} + +_SOKOL_PRIVATE void _sg_mtl_clear_state_cache(void) { + _sg_clear(&_sg.mtl.state_cache, sizeof(_sg.mtl.state_cache)); +} + +// https://developer.apple.com/metal/Metal-Feature-Set-Tables.pdf +_SOKOL_PRIVATE void _sg_mtl_init_caps(void) { + #if defined(_SG_TARGET_MACOS) + _sg.backend = SG_BACKEND_METAL_MACOS; + #elif defined(_SG_TARGET_IOS) + #if defined(_SG_TARGET_IOS_SIMULATOR) + _sg.backend = SG_BACKEND_METAL_SIMULATOR; + #else + _sg.backend = SG_BACKEND_METAL_IOS; + #endif + #endif + _sg.features.origin_top_left = true; + _sg.features.mrt_independent_blend_state = true; + _sg.features.mrt_independent_write_mask = true; + _sg.features.compute = true; + _sg.features.msaa_image_bindings = true; + + _sg.features.image_clamp_to_border = false; + #if (MAC_OS_X_VERSION_MAX_ALLOWED >= 120000) || (__IPHONE_OS_VERSION_MAX_ALLOWED >= 140000) + if (@available(macOS 12.0, iOS 14.0, *)) { + _sg.features.image_clamp_to_border = [_sg.mtl.device supportsFamily:MTLGPUFamilyApple7] + || [_sg.mtl.device supportsFamily:MTLGPUFamilyMac2]; + #if (MAC_OS_X_VERSION_MAX_ALLOWED >= 130000) || (__IPHONE_OS_VERSION_MAX_ALLOWED >= 160000) + if (!_sg.features.image_clamp_to_border) { + if (@available(macOS 13.0, iOS 16.0, *)) { + _sg.features.image_clamp_to_border = [_sg.mtl.device supportsFamily:MTLGPUFamilyMetal3]; + } + } + #endif + } + #endif + + #if defined(_SG_TARGET_MACOS) + _sg.limits.max_image_size_2d = 16 * 1024; + _sg.limits.max_image_size_cube = 16 * 1024; + _sg.limits.max_image_size_3d = 2 * 1024; + _sg.limits.max_image_size_array = 16 * 1024; + _sg.limits.max_image_array_layers = 2 * 1024; + #else + // FIXME: newer iOS devices support 16k textures + _sg.limits.max_image_size_2d = 8 * 1024; + _sg.limits.max_image_size_cube = 8 * 1024; + _sg.limits.max_image_size_3d = 2 * 1024; + _sg.limits.max_image_size_array = 8 * 1024; + _sg.limits.max_image_array_layers = 2 * 1024; + #endif + _sg.limits.max_vertex_attrs = SG_MAX_VERTEX_ATTRIBUTES; + + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R8SN]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_R8UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_R8SI]); + #if defined(_SG_TARGET_MACOS) + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R16]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R16SN]); + #else + _sg_pixelformat_sfbr(&_sg.formats[SG_PIXELFORMAT_R16]); + _sg_pixelformat_sfbr(&_sg.formats[SG_PIXELFORMAT_R16SN]); + #endif + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_R16UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_R16SI]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R16F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG8SN]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG8UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG8SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R32UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R32SI]); + #if defined(_SG_TARGET_MACOS) + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R32F]); + #else + _sg_pixelformat_sbr(&_sg.formats[SG_PIXELFORMAT_R32F]); + #endif + #if defined(_SG_TARGET_MACOS) + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG16]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG16SN]); + #else + _sg_pixelformat_sfbr(&_sg.formats[SG_PIXELFORMAT_RG16]); + _sg_pixelformat_sfbr(&_sg.formats[SG_PIXELFORMAT_RG16SN]); + #endif + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG16UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG16SI]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG16F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_SRGB8A8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA8SN]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA8UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA8SI]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_BGRA8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGB10A2]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG11B10F]); + #if defined(_SG_TARGET_MACOS) + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RGB9E5]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG32UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RG32SI]); + #else + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGB9E5]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG32UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG32SI]); + #endif + #if defined(_SG_TARGET_MACOS) + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG32F]); + #else + _sg_pixelformat_sbr(&_sg.formats[SG_PIXELFORMAT_RG32F]); + #endif + #if defined(_SG_TARGET_MACOS) + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA16]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA16SN]); + #else + _sg_pixelformat_sfbr(&_sg.formats[SG_PIXELFORMAT_RGBA16]); + _sg_pixelformat_sfbr(&_sg.formats[SG_PIXELFORMAT_RGBA16SN]); + #endif + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA16UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA16SI]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA16F]); + #if defined(_SG_TARGET_MACOS) + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA32UI]); + _sg_pixelformat_srm(&_sg.formats[SG_PIXELFORMAT_RGBA32SI]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + #else + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA32UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA32SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + #endif + _sg_pixelformat_srmd(&_sg.formats[SG_PIXELFORMAT_DEPTH]); + _sg_pixelformat_srmd(&_sg.formats[SG_PIXELFORMAT_DEPTH_STENCIL]); + #if defined(_SG_TARGET_MACOS) + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC1_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC2_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC3_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC3_SRGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC4_R]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC4_RSN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC5_RG]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC5_RGSN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC6H_RGBF]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC6H_RGBUF]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC7_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC7_SRGBA]); + #else + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGB8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_SRGB8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGB8A1]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGBA8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_SRGB8A8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_R11]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_R11SN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_RG11]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_RG11SN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ASTC_4x4_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ASTC_4x4_SRGBA]); + + #endif +} + +//-- main Metal backend state and functions ------------------------------------ +_SOKOL_PRIVATE void _sg_mtl_setup_backend(const sg_desc* desc) { + // assume already zero-initialized + SOKOL_ASSERT(desc); + SOKOL_ASSERT(desc->environment.metal.device); + SOKOL_ASSERT(desc->uniform_buffer_size > 0); + _sg_mtl_init_pool(desc); + _sg_mtl_clear_state_cache(); + _sg.mtl.valid = true; + _sg.mtl.ub_size = desc->uniform_buffer_size; + _sg.mtl.sem = dispatch_semaphore_create(SG_NUM_INFLIGHT_FRAMES); + _sg.mtl.device = (__bridge id) desc->environment.metal.device; + _sg.mtl.cmd_queue = [_sg.mtl.device newCommandQueue]; + + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + _sg.mtl.uniform_buffers[i] = [_sg.mtl.device + newBufferWithLength:(NSUInteger)_sg.mtl.ub_size + options:MTLResourceCPUCacheModeWriteCombined|MTLResourceStorageModeShared + ]; + #if defined(SOKOL_DEBUG) + _sg.mtl.uniform_buffers[i].label = [NSString stringWithFormat:@"sg-uniform-buffer.%d", i]; + #endif + } + + if (desc->mtl_force_managed_storage_mode) { + _sg.mtl.use_shared_storage_mode = false; + } else if (@available(macOS 10.15, iOS 13.0, *)) { + // on Intel Macs, always use managed resources even though the + // device says it supports unified memory (because of texture restrictions) + const bool is_apple_gpu = [_sg.mtl.device supportsFamily:MTLGPUFamilyApple1]; + if (!is_apple_gpu) { + _sg.mtl.use_shared_storage_mode = false; + } else { + _sg.mtl.use_shared_storage_mode = true; + } + } else { + #if defined(_SG_TARGET_MACOS) + _sg.mtl.use_shared_storage_mode = false; + #else + _sg.mtl.use_shared_storage_mode = true; + #endif + } + _sg_mtl_init_caps(); +} + +_SOKOL_PRIVATE void _sg_mtl_discard_backend(void) { + SOKOL_ASSERT(_sg.mtl.valid); + // wait for the last frame to finish + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + dispatch_semaphore_wait(_sg.mtl.sem, DISPATCH_TIME_FOREVER); + } + // semaphore must be "relinquished" before destruction + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + dispatch_semaphore_signal(_sg.mtl.sem); + } + _sg_mtl_garbage_collect(_sg.frame_index + SG_NUM_INFLIGHT_FRAMES + 2); + _sg_mtl_destroy_pool(); + _sg.mtl.valid = false; + + _SG_OBJC_RELEASE(_sg.mtl.sem); + _SG_OBJC_RELEASE(_sg.mtl.device); + _SG_OBJC_RELEASE(_sg.mtl.cmd_queue); + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + _SG_OBJC_RELEASE(_sg.mtl.uniform_buffers[i]); + } + // NOTE: MTLCommandBuffer, MTLRenderCommandEncoder and MTLComputeCommandEncoder are auto-released + _sg.mtl.cmd_buffer = nil; + _sg.mtl.render_cmd_encoder = nil; + _sg.mtl.compute_cmd_encoder = nil; +} + +_SOKOL_PRIVATE void _sg_mtl_reset_state_cache(void) { + _sg_mtl_clear_state_cache(); +} + +_SOKOL_PRIVATE sg_resource_state _sg_mtl_create_buffer(_sg_buffer_t* buf, const sg_buffer_desc* desc) { + SOKOL_ASSERT(buf && desc); + SOKOL_ASSERT(buf->cmn.size > 0); + const bool injected = (0 != desc->mtl_buffers[0]); + MTLResourceOptions mtl_options = _sg_mtl_buffer_resource_options(buf->cmn.usage); + for (int slot = 0; slot < buf->cmn.num_slots; slot++) { + id mtl_buf; + if (injected) { + SOKOL_ASSERT(desc->mtl_buffers[slot]); + mtl_buf = (__bridge id) desc->mtl_buffers[slot]; + } else { + if (desc->data.ptr) { + SOKOL_ASSERT(desc->data.size > 0); + mtl_buf = [_sg.mtl.device newBufferWithBytes:desc->data.ptr length:(NSUInteger)buf->cmn.size options:mtl_options]; + } else { + // this is guaranteed to zero-initialize the buffer + mtl_buf = [_sg.mtl.device newBufferWithLength:(NSUInteger)buf->cmn.size options:mtl_options]; + } + if (nil == mtl_buf) { + _SG_ERROR(METAL_CREATE_BUFFER_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } + #if defined(SOKOL_DEBUG) + if (desc->label) { + mtl_buf.label = [NSString stringWithFormat:@"%s.%d", desc->label, slot]; + } + #endif + buf->mtl.buf[slot] = _sg_mtl_add_resource(mtl_buf); + _SG_OBJC_RELEASE(mtl_buf); + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_mtl_discard_buffer(_sg_buffer_t* buf) { + SOKOL_ASSERT(buf); + for (int slot = 0; slot < buf->cmn.num_slots; slot++) { + // it's valid to call release resource with '0' + _sg_mtl_release_resource(_sg.frame_index, buf->mtl.buf[slot]); + } +} + +_SOKOL_PRIVATE void _sg_mtl_copy_image_data(const _sg_image_t* img, __unsafe_unretained id mtl_tex, const sg_image_data* data) { + const int num_faces = (img->cmn.type == SG_IMAGETYPE_CUBE) ? 6:1; + const int num_slices = (img->cmn.type == SG_IMAGETYPE_ARRAY) ? img->cmn.num_slices : 1; + for (int face_index = 0; face_index < num_faces; face_index++) { + for (int mip_index = 0; mip_index < img->cmn.num_mipmaps; mip_index++) { + SOKOL_ASSERT(data->subimage[face_index][mip_index].ptr); + SOKOL_ASSERT(data->subimage[face_index][mip_index].size > 0); + const uint8_t* data_ptr = (const uint8_t*)data->subimage[face_index][mip_index].ptr; + const int mip_width = _sg_miplevel_dim(img->cmn.width, mip_index); + const int mip_height = _sg_miplevel_dim(img->cmn.height, mip_index); + int bytes_per_row = _sg_row_pitch(img->cmn.pixel_format, mip_width, 1); + int bytes_per_slice = _sg_surface_pitch(img->cmn.pixel_format, mip_width, mip_height, 1); + /* bytesPerImage special case: https://developer.apple.com/documentation/metal/mtltexture/1515679-replaceregion + + "Supply a nonzero value only when you copy data to a MTLTextureType3D type texture" + */ + MTLRegion region; + int bytes_per_image; + if (img->cmn.type == SG_IMAGETYPE_3D) { + const int mip_depth = _sg_miplevel_dim(img->cmn.num_slices, mip_index); + region = MTLRegionMake3D(0, 0, 0, (NSUInteger)mip_width, (NSUInteger)mip_height, (NSUInteger)mip_depth); + bytes_per_image = bytes_per_slice; + // FIXME: apparently the minimal bytes_per_image size for 3D texture is 4 KByte... somehow need to handle this + } else { + region = MTLRegionMake2D(0, 0, (NSUInteger)mip_width, (NSUInteger)mip_height); + bytes_per_image = 0; + } + + for (int slice_index = 0; slice_index < num_slices; slice_index++) { + const int mtl_slice_index = (img->cmn.type == SG_IMAGETYPE_CUBE) ? face_index : slice_index; + const int slice_offset = slice_index * bytes_per_slice; + SOKOL_ASSERT((slice_offset + bytes_per_slice) <= (int)data->subimage[face_index][mip_index].size); + [mtl_tex replaceRegion:region + mipmapLevel:(NSUInteger)mip_index + slice:(NSUInteger)mtl_slice_index + withBytes:data_ptr + slice_offset + bytesPerRow:(NSUInteger)bytes_per_row + bytesPerImage:(NSUInteger)bytes_per_image]; + } + } + } +} + +// initialize MTLTextureDescriptor with common attributes +_SOKOL_PRIVATE bool _sg_mtl_init_texdesc_common(MTLTextureDescriptor* mtl_desc, _sg_image_t* img) { + mtl_desc.textureType = _sg_mtl_texture_type(img->cmn.type); + mtl_desc.pixelFormat = _sg_mtl_pixel_format(img->cmn.pixel_format); + if (MTLPixelFormatInvalid == mtl_desc.pixelFormat) { + _SG_ERROR(METAL_TEXTURE_FORMAT_NOT_SUPPORTED); + return false; + } + mtl_desc.width = (NSUInteger)img->cmn.width; + mtl_desc.height = (NSUInteger)img->cmn.height; + if (SG_IMAGETYPE_3D == img->cmn.type) { + mtl_desc.depth = (NSUInteger)img->cmn.num_slices; + } else { + mtl_desc.depth = 1; + } + mtl_desc.mipmapLevelCount = (NSUInteger)img->cmn.num_mipmaps; + if (SG_IMAGETYPE_ARRAY == img->cmn.type) { + mtl_desc.arrayLength = (NSUInteger)img->cmn.num_slices; + } else { + mtl_desc.arrayLength = 1; + } + mtl_desc.usage = MTLTextureUsageShaderRead; + MTLResourceOptions res_options = 0; + if (img->cmn.usage != SG_USAGE_IMMUTABLE) { + res_options |= MTLResourceCPUCacheModeWriteCombined; + } + res_options |= _sg_mtl_resource_options_storage_mode_managed_or_shared(); + mtl_desc.resourceOptions = res_options; + return true; +} + +// initialize MTLTextureDescriptor with rendertarget attributes +_SOKOL_PRIVATE void _sg_mtl_init_texdesc_rt(MTLTextureDescriptor* mtl_desc, _sg_image_t* img) { + SOKOL_ASSERT(img->cmn.render_target); + _SOKOL_UNUSED(img); + mtl_desc.usage = MTLTextureUsageShaderRead | MTLTextureUsageRenderTarget; + mtl_desc.resourceOptions = MTLResourceStorageModePrivate; +} + +// initialize MTLTextureDescriptor with MSAA attributes +_SOKOL_PRIVATE void _sg_mtl_init_texdesc_rt_msaa(MTLTextureDescriptor* mtl_desc, _sg_image_t* img) { + SOKOL_ASSERT(img->cmn.sample_count > 1); + mtl_desc.usage = MTLTextureUsageShaderRead | MTLTextureUsageRenderTarget; + mtl_desc.resourceOptions = MTLResourceStorageModePrivate; + mtl_desc.textureType = MTLTextureType2DMultisample; + mtl_desc.sampleCount = (NSUInteger)img->cmn.sample_count; +} + +_SOKOL_PRIVATE sg_resource_state _sg_mtl_create_image(_sg_image_t* img, const sg_image_desc* desc) { + SOKOL_ASSERT(img && desc); + const bool injected = (0 != desc->mtl_textures[0]); + + // first initialize all Metal resource pool slots to 'empty' + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + img->mtl.tex[i] = _sg_mtl_add_resource(nil); + } + + // initialize a Metal texture descriptor + MTLTextureDescriptor* mtl_desc = [[MTLTextureDescriptor alloc] init]; + if (!_sg_mtl_init_texdesc_common(mtl_desc, img)) { + _SG_OBJC_RELEASE(mtl_desc); + return SG_RESOURCESTATE_FAILED; + } + if (img->cmn.render_target) { + if (img->cmn.sample_count > 1) { + _sg_mtl_init_texdesc_rt_msaa(mtl_desc, img); + } else { + _sg_mtl_init_texdesc_rt(mtl_desc, img); + } + } + for (int slot = 0; slot < img->cmn.num_slots; slot++) { + id mtl_tex; + if (injected) { + SOKOL_ASSERT(desc->mtl_textures[slot]); + mtl_tex = (__bridge id) desc->mtl_textures[slot]; + } else { + mtl_tex = [_sg.mtl.device newTextureWithDescriptor:mtl_desc]; + if (nil == mtl_tex) { + _SG_OBJC_RELEASE(mtl_desc); + _SG_ERROR(METAL_CREATE_TEXTURE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + if ((img->cmn.usage == SG_USAGE_IMMUTABLE) && !img->cmn.render_target) { + _sg_mtl_copy_image_data(img, mtl_tex, &desc->data); + } + } + #if defined(SOKOL_DEBUG) + if (desc->label) { + mtl_tex.label = [NSString stringWithFormat:@"%s.%d", desc->label, slot]; + } + #endif + img->mtl.tex[slot] = _sg_mtl_add_resource(mtl_tex); + _SG_OBJC_RELEASE(mtl_tex); + } + _SG_OBJC_RELEASE(mtl_desc); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_mtl_discard_image(_sg_image_t* img) { + SOKOL_ASSERT(img); + // it's valid to call release resource with a 'null resource' + for (int slot = 0; slot < img->cmn.num_slots; slot++) { + _sg_mtl_release_resource(_sg.frame_index, img->mtl.tex[slot]); + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_mtl_create_sampler(_sg_sampler_t* smp, const sg_sampler_desc* desc) { + SOKOL_ASSERT(smp && desc); + id mtl_smp; + const bool injected = (0 != desc->mtl_sampler); + if (injected) { + SOKOL_ASSERT(desc->mtl_sampler); + mtl_smp = (__bridge id) desc->mtl_sampler; + } else { + MTLSamplerDescriptor* mtl_desc = [[MTLSamplerDescriptor alloc] init]; + mtl_desc.sAddressMode = _sg_mtl_address_mode(desc->wrap_u); + mtl_desc.tAddressMode = _sg_mtl_address_mode(desc->wrap_v); + mtl_desc.rAddressMode = _sg_mtl_address_mode(desc->wrap_w); + if (_sg.features.image_clamp_to_border) { + if (@available(macOS 12.0, iOS 14.0, *)) { + mtl_desc.borderColor = _sg_mtl_border_color(desc->border_color); + } + } + mtl_desc.minFilter = _sg_mtl_minmag_filter(desc->min_filter); + mtl_desc.magFilter = _sg_mtl_minmag_filter(desc->mag_filter); + mtl_desc.mipFilter = _sg_mtl_mipmap_filter(desc->mipmap_filter); + mtl_desc.lodMinClamp = desc->min_lod; + mtl_desc.lodMaxClamp = desc->max_lod; + // FIXME: lodAverage? + mtl_desc.maxAnisotropy = desc->max_anisotropy; + mtl_desc.normalizedCoordinates = YES; + mtl_desc.compareFunction = _sg_mtl_compare_func(desc->compare); + #if defined(SOKOL_DEBUG) + if (desc->label) { + mtl_desc.label = [NSString stringWithUTF8String:desc->label]; + } + #endif + mtl_smp = [_sg.mtl.device newSamplerStateWithDescriptor:mtl_desc]; + _SG_OBJC_RELEASE(mtl_desc); + if (nil == mtl_smp) { + _SG_ERROR(METAL_CREATE_SAMPLER_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } + smp->mtl.sampler_state = _sg_mtl_add_resource(mtl_smp); + _SG_OBJC_RELEASE(mtl_smp); + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_mtl_discard_sampler(_sg_sampler_t* smp) { + SOKOL_ASSERT(smp); + // it's valid to call release resource with a 'null resource' + _sg_mtl_release_resource(_sg.frame_index, smp->mtl.sampler_state); +} + +_SOKOL_PRIVATE id _sg_mtl_compile_library(const char* src) { + NSError* err = NULL; + id lib = [_sg.mtl.device + newLibraryWithSource:[NSString stringWithUTF8String:src] + options:nil + error:&err + ]; + if (err) { + _SG_ERROR(METAL_SHADER_COMPILATION_FAILED); + _SG_LOGMSG(METAL_SHADER_COMPILATION_OUTPUT, [err.localizedDescription UTF8String]); + } + return lib; +} + +_SOKOL_PRIVATE id _sg_mtl_library_from_bytecode(const void* ptr, size_t num_bytes) { + NSError* err = NULL; + dispatch_data_t lib_data = dispatch_data_create(ptr, num_bytes, NULL, DISPATCH_DATA_DESTRUCTOR_DEFAULT); + id lib = [_sg.mtl.device newLibraryWithData:lib_data error:&err]; + if (err) { + _SG_ERROR(METAL_SHADER_CREATION_FAILED); + _SG_LOGMSG(METAL_SHADER_COMPILATION_OUTPUT, [err.localizedDescription UTF8String]); + } + _SG_OBJC_RELEASE(lib_data); + return lib; +} + +_SOKOL_PRIVATE bool _sg_mtl_create_shader_func(const sg_shader_function* func, const char* label, const char* label_ext, _sg_mtl_shader_func_t* res) { + SOKOL_ASSERT(res->mtl_lib == _SG_MTL_INVALID_SLOT_INDEX); + SOKOL_ASSERT(res->mtl_func == _SG_MTL_INVALID_SLOT_INDEX); + id mtl_lib = nil; + if (func->bytecode.ptr) { + SOKOL_ASSERT(func->bytecode.size > 0); + mtl_lib = _sg_mtl_library_from_bytecode(func->bytecode.ptr, func->bytecode.size); + } else if (func->source) { + mtl_lib = _sg_mtl_compile_library(func->source); + } + if (mtl_lib == nil) { + return false; + } + #if defined(SOKOL_DEBUG) + if (label) { + SOKOL_ASSERT(label_ext); + mtl_lib.label = [NSString stringWithFormat:@"%s.%s", label, label_ext]; + } + #else + _SOKOL_UNUSED(label); + _SOKOL_UNUSED(label_ext); + #endif + SOKOL_ASSERT(func->entry); + id mtl_func = [mtl_lib newFunctionWithName:[NSString stringWithUTF8String:func->entry]]; + if (mtl_func == nil) { + _SG_ERROR(METAL_SHADER_ENTRY_NOT_FOUND); + _SG_OBJC_RELEASE(mtl_lib); + return false; + } + res->mtl_lib = _sg_mtl_add_resource(mtl_lib); + res->mtl_func = _sg_mtl_add_resource(mtl_func); + _SG_OBJC_RELEASE(mtl_lib); + _SG_OBJC_RELEASE(mtl_func); + return true; +} + +_SOKOL_PRIVATE void _sg_mtl_discard_shader_func(const _sg_mtl_shader_func_t* func) { + // it is valid to call _sg_mtl_release_resource with a 'null resource' + _sg_mtl_release_resource(_sg.frame_index, func->mtl_func); + _sg_mtl_release_resource(_sg.frame_index, func->mtl_lib); +} + +// NOTE: this is an out-of-range check for MSL bindslots that's also active in release mode +_SOKOL_PRIVATE bool _sg_mtl_ensure_msl_bindslot_ranges(const sg_shader_desc* desc) { + SOKOL_ASSERT(desc); + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + if (desc->uniform_blocks[i].msl_buffer_n >= _SG_MTL_MAX_STAGE_UB_BINDINGS) { + _SG_ERROR(METAL_UNIFORMBLOCK_MSL_BUFFER_SLOT_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (desc->storage_buffers[i].msl_buffer_n >= _SG_MTL_MAX_STAGE_UB_SBUF_BINDINGS) { + _SG_ERROR(METAL_STORAGEBUFFER_MSL_BUFFER_SLOT_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + if (desc->images[i].msl_texture_n >= _SG_MTL_MAX_STAGE_IMAGE_BINDINGS) { + _SG_ERROR(METAL_IMAGE_MSL_TEXTURE_SLOT_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + if (desc->samplers[i].msl_sampler_n >= _SG_MTL_MAX_STAGE_SAMPLER_BINDINGS) { + _SG_ERROR(METAL_SAMPLER_MSL_SAMPLER_SLOT_OUT_OF_RANGE); + return false; + } + } + return true; +} + +_SOKOL_PRIVATE sg_resource_state _sg_mtl_create_shader(_sg_shader_t* shd, const sg_shader_desc* desc) { + SOKOL_ASSERT(shd && desc); + + // do a MSL bindslot range check also in release mode, and if that fails, + // also fail shader creation + if (!_sg_mtl_ensure_msl_bindslot_ranges(desc)) { + return SG_RESOURCESTATE_FAILED; + } + + shd->mtl.threads_per_threadgroup = MTLSizeMake( + (NSUInteger)desc->mtl_threads_per_threadgroup.x, + (NSUInteger)desc->mtl_threads_per_threadgroup.y, + (NSUInteger)desc->mtl_threads_per_threadgroup.z); + + // copy resource bindslot mappings + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + shd->mtl.ub_buffer_n[i] = desc->uniform_blocks[i].msl_buffer_n; + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + shd->mtl.sbuf_buffer_n[i] = desc->storage_buffers[i].msl_buffer_n; + } + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + shd->mtl.img_texture_n[i] = desc->images[i].msl_texture_n; + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + shd->mtl.smp_sampler_n[i] = desc->samplers[i].msl_sampler_n; + } + + // create metal library and function objects + bool shd_valid = true; + if (desc->vertex_func.source || desc->vertex_func.bytecode.ptr) { + shd_valid &= _sg_mtl_create_shader_func(&desc->vertex_func, desc->label, "vs", &shd->mtl.vertex_func); + } + if (desc->fragment_func.source || desc->fragment_func.bytecode.ptr) { + shd_valid &= _sg_mtl_create_shader_func(&desc->fragment_func, desc->label, "fs", &shd->mtl.fragment_func); + } + if (desc->compute_func.source || desc->compute_func.bytecode.ptr) { + shd_valid &= _sg_mtl_create_shader_func(&desc->compute_func, desc->label, "cs", &shd->mtl.compute_func); + } + if (!shd_valid) { + _sg_mtl_discard_shader_func(&shd->mtl.vertex_func); + _sg_mtl_discard_shader_func(&shd->mtl.fragment_func); + _sg_mtl_discard_shader_func(&shd->mtl.compute_func); + } + return shd_valid ? SG_RESOURCESTATE_VALID : SG_RESOURCESTATE_FAILED; +} + +_SOKOL_PRIVATE void _sg_mtl_discard_shader(_sg_shader_t* shd) { + SOKOL_ASSERT(shd); + _sg_mtl_discard_shader_func(&shd->mtl.vertex_func); + _sg_mtl_discard_shader_func(&shd->mtl.fragment_func); + _sg_mtl_discard_shader_func(&shd->mtl.compute_func); +} + +_SOKOL_PRIVATE sg_resource_state _sg_mtl_create_pipeline(_sg_pipeline_t* pip, _sg_shader_t* shd, const sg_pipeline_desc* desc) { + SOKOL_ASSERT(pip && shd && desc); + SOKOL_ASSERT(desc->shader.id == shd->slot.id); + + pip->shader = shd; + + if (pip->cmn.is_compute) { + NSError* err = NULL; + MTLComputePipelineDescriptor* cp_desc = [[MTLComputePipelineDescriptor alloc] init]; + cp_desc.computeFunction = _sg_mtl_id(shd->mtl.compute_func.mtl_func); + cp_desc.threadGroupSizeIsMultipleOfThreadExecutionWidth = true; + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + const sg_shader_stage stage = shd->cmn.storage_buffers[i].stage; + SOKOL_ASSERT((stage != SG_SHADERSTAGE_VERTEX) && (stage != SG_SHADERSTAGE_FRAGMENT)); + if ((stage == SG_SHADERSTAGE_COMPUTE) && shd->cmn.storage_buffers[i].readonly) { + const NSUInteger mtl_slot = shd->mtl.sbuf_buffer_n[i]; + cp_desc.buffers[mtl_slot].mutability = MTLMutabilityImmutable; + } + } + #if defined(SOKOL_DEBUG) + if (desc->label) { + cp_desc.label = [NSString stringWithFormat:@"%s", desc->label]; + } + #endif + id mtl_cps = [_sg.mtl.device + newComputePipelineStateWithDescriptor:cp_desc + options:MTLPipelineOptionNone + reflection:nil + error:&err]; + _SG_OBJC_RELEASE(cp_desc); + if (nil == mtl_cps) { + SOKOL_ASSERT(err); + _SG_ERROR(METAL_CREATE_CPS_FAILED); + _SG_LOGMSG(METAL_CREATE_CPS_OUTPUT, [err.localizedDescription UTF8String]); + return SG_RESOURCESTATE_FAILED; + } + pip->mtl.cps = _sg_mtl_add_resource(mtl_cps); + _SG_OBJC_RELEASE(mtl_cps); + pip->mtl.threads_per_threadgroup = shd->mtl.threads_per_threadgroup; + } else { + sg_primitive_type prim_type = desc->primitive_type; + pip->mtl.prim_type = _sg_mtl_primitive_type(prim_type); + pip->mtl.index_size = _sg_mtl_index_size(pip->cmn.index_type); + if (SG_INDEXTYPE_NONE != pip->cmn.index_type) { + pip->mtl.index_type = _sg_mtl_index_type(pip->cmn.index_type); + } + pip->mtl.cull_mode = _sg_mtl_cull_mode(desc->cull_mode); + pip->mtl.winding = _sg_mtl_winding(desc->face_winding); + pip->mtl.stencil_ref = desc->stencil.ref; + + // create vertex-descriptor + MTLVertexDescriptor* vtx_desc = [MTLVertexDescriptor vertexDescriptor]; + for (NSUInteger attr_index = 0; attr_index < SG_MAX_VERTEX_ATTRIBUTES; attr_index++) { + const sg_vertex_attr_state* a_state = &desc->layout.attrs[attr_index]; + if (a_state->format == SG_VERTEXFORMAT_INVALID) { + break; + } + SOKOL_ASSERT(a_state->buffer_index < SG_MAX_VERTEXBUFFER_BINDSLOTS); + SOKOL_ASSERT(pip->cmn.vertex_buffer_layout_active[a_state->buffer_index]); + vtx_desc.attributes[attr_index].format = _sg_mtl_vertex_format(a_state->format); + vtx_desc.attributes[attr_index].offset = (NSUInteger)a_state->offset; + vtx_desc.attributes[attr_index].bufferIndex = _sg_mtl_vertexbuffer_bindslot((size_t)a_state->buffer_index); + } + for (NSUInteger layout_index = 0; layout_index < SG_MAX_VERTEXBUFFER_BINDSLOTS; layout_index++) { + if (pip->cmn.vertex_buffer_layout_active[layout_index]) { + const sg_vertex_buffer_layout_state* l_state = &desc->layout.buffers[layout_index]; + const NSUInteger mtl_vb_slot = _sg_mtl_vertexbuffer_bindslot(layout_index); + SOKOL_ASSERT(l_state->stride > 0); + vtx_desc.layouts[mtl_vb_slot].stride = (NSUInteger)l_state->stride; + vtx_desc.layouts[mtl_vb_slot].stepFunction = _sg_mtl_step_function(l_state->step_func); + vtx_desc.layouts[mtl_vb_slot].stepRate = (NSUInteger)l_state->step_rate; + if (SG_VERTEXSTEP_PER_INSTANCE == l_state->step_func) { + // NOTE: not actually used in _sg_mtl_draw() + pip->cmn.use_instanced_draw = true; + } + } + } + + // render-pipeline descriptor + MTLRenderPipelineDescriptor* rp_desc = [[MTLRenderPipelineDescriptor alloc] init]; + rp_desc.vertexDescriptor = vtx_desc; + SOKOL_ASSERT(shd->mtl.vertex_func.mtl_func != _SG_MTL_INVALID_SLOT_INDEX); + rp_desc.vertexFunction = _sg_mtl_id(shd->mtl.vertex_func.mtl_func); + SOKOL_ASSERT(shd->mtl.fragment_func.mtl_func != _SG_MTL_INVALID_SLOT_INDEX); + rp_desc.fragmentFunction = _sg_mtl_id(shd->mtl.fragment_func.mtl_func); + rp_desc.rasterSampleCount = (NSUInteger)desc->sample_count; + rp_desc.alphaToCoverageEnabled = desc->alpha_to_coverage_enabled; + rp_desc.alphaToOneEnabled = NO; + rp_desc.rasterizationEnabled = YES; + rp_desc.depthAttachmentPixelFormat = _sg_mtl_pixel_format(desc->depth.pixel_format); + if (desc->depth.pixel_format == SG_PIXELFORMAT_DEPTH_STENCIL) { + rp_desc.stencilAttachmentPixelFormat = _sg_mtl_pixel_format(desc->depth.pixel_format); + } + for (NSUInteger i = 0; i < (NSUInteger)desc->color_count; i++) { + SOKOL_ASSERT(i < SG_MAX_COLOR_ATTACHMENTS); + const sg_color_target_state* cs = &desc->colors[i]; + rp_desc.colorAttachments[i].pixelFormat = _sg_mtl_pixel_format(cs->pixel_format); + rp_desc.colorAttachments[i].writeMask = _sg_mtl_color_write_mask(cs->write_mask); + rp_desc.colorAttachments[i].blendingEnabled = cs->blend.enabled; + rp_desc.colorAttachments[i].alphaBlendOperation = _sg_mtl_blend_op(cs->blend.op_alpha); + rp_desc.colorAttachments[i].rgbBlendOperation = _sg_mtl_blend_op(cs->blend.op_rgb); + rp_desc.colorAttachments[i].destinationAlphaBlendFactor = _sg_mtl_blend_factor(cs->blend.dst_factor_alpha); + rp_desc.colorAttachments[i].destinationRGBBlendFactor = _sg_mtl_blend_factor(cs->blend.dst_factor_rgb); + rp_desc.colorAttachments[i].sourceAlphaBlendFactor = _sg_mtl_blend_factor(cs->blend.src_factor_alpha); + rp_desc.colorAttachments[i].sourceRGBBlendFactor = _sg_mtl_blend_factor(cs->blend.src_factor_rgb); + } + // set buffer mutability for all read-only buffers (vertex buffers and read-only storage buffers) + for (size_t i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + if (pip->cmn.vertex_buffer_layout_active[i]) { + const NSUInteger mtl_slot = _sg_mtl_vertexbuffer_bindslot(i); + rp_desc.vertexBuffers[mtl_slot].mutability = MTLMutabilityImmutable; + } + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + const NSUInteger mtl_slot = shd->mtl.sbuf_buffer_n[i]; + const sg_shader_stage stage = shd->cmn.storage_buffers[i].stage; + SOKOL_ASSERT(stage != SG_SHADERSTAGE_COMPUTE); + if (stage == SG_SHADERSTAGE_VERTEX) { + SOKOL_ASSERT(shd->cmn.storage_buffers[i].readonly); + rp_desc.vertexBuffers[mtl_slot].mutability = MTLMutabilityImmutable; + } else if (stage == SG_SHADERSTAGE_FRAGMENT) { + SOKOL_ASSERT(shd->cmn.storage_buffers[i].readonly); + rp_desc.fragmentBuffers[mtl_slot].mutability = MTLMutabilityImmutable; + } + } + #if defined(SOKOL_DEBUG) + if (desc->label) { + rp_desc.label = [NSString stringWithFormat:@"%s", desc->label]; + } + #endif + NSError* err = NULL; + id mtl_rps = [_sg.mtl.device newRenderPipelineStateWithDescriptor:rp_desc error:&err]; + _SG_OBJC_RELEASE(rp_desc); + if (nil == mtl_rps) { + SOKOL_ASSERT(err); + _SG_ERROR(METAL_CREATE_RPS_FAILED); + _SG_LOGMSG(METAL_CREATE_RPS_OUTPUT, [err.localizedDescription UTF8String]); + return SG_RESOURCESTATE_FAILED; + } + pip->mtl.rps = _sg_mtl_add_resource(mtl_rps); + _SG_OBJC_RELEASE(mtl_rps); + + // depth-stencil-state + MTLDepthStencilDescriptor* ds_desc = [[MTLDepthStencilDescriptor alloc] init]; + ds_desc.depthCompareFunction = _sg_mtl_compare_func(desc->depth.compare); + ds_desc.depthWriteEnabled = desc->depth.write_enabled; + if (desc->stencil.enabled) { + const sg_stencil_face_state* sb = &desc->stencil.back; + ds_desc.backFaceStencil = [[MTLStencilDescriptor alloc] init]; + ds_desc.backFaceStencil.stencilFailureOperation = _sg_mtl_stencil_op(sb->fail_op); + ds_desc.backFaceStencil.depthFailureOperation = _sg_mtl_stencil_op(sb->depth_fail_op); + ds_desc.backFaceStencil.depthStencilPassOperation = _sg_mtl_stencil_op(sb->pass_op); + ds_desc.backFaceStencil.stencilCompareFunction = _sg_mtl_compare_func(sb->compare); + ds_desc.backFaceStencil.readMask = desc->stencil.read_mask; + ds_desc.backFaceStencil.writeMask = desc->stencil.write_mask; + const sg_stencil_face_state* sf = &desc->stencil.front; + ds_desc.frontFaceStencil = [[MTLStencilDescriptor alloc] init]; + ds_desc.frontFaceStencil.stencilFailureOperation = _sg_mtl_stencil_op(sf->fail_op); + ds_desc.frontFaceStencil.depthFailureOperation = _sg_mtl_stencil_op(sf->depth_fail_op); + ds_desc.frontFaceStencil.depthStencilPassOperation = _sg_mtl_stencil_op(sf->pass_op); + ds_desc.frontFaceStencil.stencilCompareFunction = _sg_mtl_compare_func(sf->compare); + ds_desc.frontFaceStencil.readMask = desc->stencil.read_mask; + ds_desc.frontFaceStencil.writeMask = desc->stencil.write_mask; + } + #if defined(SOKOL_DEBUG) + if (desc->label) { + ds_desc.label = [NSString stringWithFormat:@"%s.dss", desc->label]; + } + #endif + id mtl_dss = [_sg.mtl.device newDepthStencilStateWithDescriptor:ds_desc]; + _SG_OBJC_RELEASE(ds_desc); + if (nil == mtl_dss) { + _SG_ERROR(METAL_CREATE_DSS_FAILED); + return SG_RESOURCESTATE_FAILED; + } + pip->mtl.dss = _sg_mtl_add_resource(mtl_dss); + _SG_OBJC_RELEASE(mtl_dss); + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_mtl_discard_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + // it's valid to call release resource with a 'null resource' + _sg_mtl_release_resource(_sg.frame_index, pip->mtl.cps); + _sg_mtl_release_resource(_sg.frame_index, pip->mtl.rps); + _sg_mtl_release_resource(_sg.frame_index, pip->mtl.dss); +} + +_SOKOL_PRIVATE sg_resource_state _sg_mtl_create_attachments(_sg_attachments_t* atts, _sg_image_t** color_images, _sg_image_t** resolve_images, _sg_image_t* ds_img, const sg_attachments_desc* desc) { + SOKOL_ASSERT(atts && desc); + SOKOL_ASSERT(color_images && resolve_images); + + // copy image pointers + for (int i = 0; i < atts->cmn.num_colors; i++) { + const sg_attachment_desc* color_desc = &desc->colors[i]; + _SOKOL_UNUSED(color_desc); + SOKOL_ASSERT(color_desc->image.id != SG_INVALID_ID); + SOKOL_ASSERT(0 == atts->mtl.colors[i].image); + SOKOL_ASSERT(color_images[i] && (color_images[i]->slot.id == color_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_color_format(color_images[i]->cmn.pixel_format)); + atts->mtl.colors[i].image = color_images[i]; + + const sg_attachment_desc* resolve_desc = &desc->resolves[i]; + if (resolve_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(0 == atts->mtl.resolves[i].image); + SOKOL_ASSERT(resolve_images[i] && (resolve_images[i]->slot.id == resolve_desc->image.id)); + SOKOL_ASSERT(color_images[i] && (color_images[i]->cmn.pixel_format == resolve_images[i]->cmn.pixel_format)); + atts->mtl.resolves[i].image = resolve_images[i]; + } + } + SOKOL_ASSERT(0 == atts->mtl.depth_stencil.image); + const sg_attachment_desc* ds_desc = &desc->depth_stencil; + if (ds_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(ds_img && (ds_img->slot.id == ds_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_depth_format(ds_img->cmn.pixel_format)); + atts->mtl.depth_stencil.image = ds_img; + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_mtl_discard_attachments(_sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + _SOKOL_UNUSED(atts); +} + +_SOKOL_PRIVATE _sg_image_t* _sg_mtl_attachments_color_image(const _sg_attachments_t* atts, int index) { + // NOTE: may return null + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + return atts->mtl.colors[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_mtl_attachments_resolve_image(const _sg_attachments_t* atts, int index) { + // NOTE: may return null + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + return atts->mtl.resolves[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_mtl_attachments_ds_image(const _sg_attachments_t* atts) { + // NOTE: may return null + SOKOL_ASSERT(atts); + return atts->mtl.depth_stencil.image; +} + +_SOKOL_PRIVATE void _sg_mtl_bind_uniform_buffers(void) { + // In the Metal backend, uniform buffer bindings happen once in sg_begin_pass() and + // remain valid for the entire pass. Only binding offsets will be updated + // in sg_apply_uniforms() + if (_sg.cur_pass.is_compute) { + SOKOL_ASSERT(nil != _sg.mtl.compute_cmd_encoder); + for (size_t slot = 0; slot < SG_MAX_UNIFORMBLOCK_BINDSLOTS; slot++) { + [_sg.mtl.compute_cmd_encoder + setBuffer:_sg.mtl.uniform_buffers[_sg.mtl.cur_frame_rotate_index] + offset:0 + atIndex:slot]; + } + } else { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + for (size_t slot = 0; slot < SG_MAX_UNIFORMBLOCK_BINDSLOTS; slot++) { + [_sg.mtl.render_cmd_encoder + setVertexBuffer:_sg.mtl.uniform_buffers[_sg.mtl.cur_frame_rotate_index] + offset:0 + atIndex:slot]; + [_sg.mtl.render_cmd_encoder + setFragmentBuffer:_sg.mtl.uniform_buffers[_sg.mtl.cur_frame_rotate_index] + offset:0 + atIndex:slot]; + } + } +} + +_SOKOL_PRIVATE void _sg_mtl_begin_compute_pass(const sg_pass* pass) { + SOKOL_ASSERT(pass); (void)pass; + SOKOL_ASSERT(nil != _sg.mtl.cmd_buffer); + SOKOL_ASSERT(nil == _sg.mtl.compute_cmd_encoder); + SOKOL_ASSERT(nil == _sg.mtl.render_cmd_encoder); + + // NOTE: we actually want computeCommandEncoderWithDispatchType:MTLDispatchTypeConcurrent, but + // that requires bumping the macOS base version to 10.14 + _sg.mtl.compute_cmd_encoder = [_sg.mtl.cmd_buffer computeCommandEncoder]; + if (nil == _sg.mtl.compute_cmd_encoder) { + _sg.cur_pass.valid = false; + return; + } + + #if defined(SOKOL_DEBUG) + if (pass->label) { + _sg.mtl.compute_cmd_encoder.label = [NSString stringWithUTF8String:pass->label]; + } + #endif +} + +_SOKOL_PRIVATE void _sg_mtl_begin_render_pass(const sg_pass* pass) { + SOKOL_ASSERT(pass); + SOKOL_ASSERT(nil != _sg.mtl.cmd_buffer); + SOKOL_ASSERT(nil == _sg.mtl.render_cmd_encoder); + SOKOL_ASSERT(nil == _sg.mtl.compute_cmd_encoder); + + const _sg_attachments_t* atts = _sg.cur_pass.atts; + const sg_swapchain* swapchain = &pass->swapchain; + const sg_pass_action* action = &pass->action; + + MTLRenderPassDescriptor* pass_desc = [MTLRenderPassDescriptor renderPassDescriptor]; + SOKOL_ASSERT(pass_desc); + if (atts) { + // setup pass descriptor for offscreen rendering + SOKOL_ASSERT(atts->slot.state == SG_RESOURCESTATE_VALID); + for (NSUInteger i = 0; i < (NSUInteger)atts->cmn.num_colors; i++) { + const _sg_attachment_common_t* cmn_color_att = &atts->cmn.colors[i]; + const _sg_mtl_attachment_t* mtl_color_att = &atts->mtl.colors[i]; + const _sg_image_t* color_att_img = mtl_color_att->image; + const _sg_attachment_common_t* cmn_resolve_att = &atts->cmn.resolves[i]; + const _sg_mtl_attachment_t* mtl_resolve_att = &atts->mtl.resolves[i]; + const _sg_image_t* resolve_att_img = mtl_resolve_att->image; + SOKOL_ASSERT(color_att_img->slot.state == SG_RESOURCESTATE_VALID); + SOKOL_ASSERT(color_att_img->slot.id == cmn_color_att->image_id.id); + SOKOL_ASSERT(color_att_img->mtl.tex[color_att_img->cmn.active_slot] != _SG_MTL_INVALID_SLOT_INDEX); + pass_desc.colorAttachments[i].loadAction = _sg_mtl_load_action(action->colors[i].load_action); + pass_desc.colorAttachments[i].storeAction = _sg_mtl_store_action(action->colors[i].store_action, resolve_att_img != 0); + sg_color c = action->colors[i].clear_value; + pass_desc.colorAttachments[i].clearColor = MTLClearColorMake(c.r, c.g, c.b, c.a); + pass_desc.colorAttachments[i].texture = _sg_mtl_id(color_att_img->mtl.tex[color_att_img->cmn.active_slot]); + pass_desc.colorAttachments[i].level = (NSUInteger)cmn_color_att->mip_level; + switch (color_att_img->cmn.type) { + case SG_IMAGETYPE_CUBE: + case SG_IMAGETYPE_ARRAY: + pass_desc.colorAttachments[i].slice = (NSUInteger)cmn_color_att->slice; + break; + case SG_IMAGETYPE_3D: + pass_desc.colorAttachments[i].depthPlane = (NSUInteger)cmn_color_att->slice; + break; + default: break; + } + if (resolve_att_img) { + SOKOL_ASSERT(resolve_att_img->slot.state == SG_RESOURCESTATE_VALID); + SOKOL_ASSERT(resolve_att_img->slot.id == cmn_resolve_att->image_id.id); + SOKOL_ASSERT(resolve_att_img->mtl.tex[resolve_att_img->cmn.active_slot] != _SG_MTL_INVALID_SLOT_INDEX); + pass_desc.colorAttachments[i].resolveTexture = _sg_mtl_id(resolve_att_img->mtl.tex[resolve_att_img->cmn.active_slot]); + pass_desc.colorAttachments[i].resolveLevel = (NSUInteger)cmn_resolve_att->mip_level; + switch (resolve_att_img->cmn.type) { + case SG_IMAGETYPE_CUBE: + case SG_IMAGETYPE_ARRAY: + pass_desc.colorAttachments[i].resolveSlice = (NSUInteger)cmn_resolve_att->slice; + break; + case SG_IMAGETYPE_3D: + pass_desc.colorAttachments[i].resolveDepthPlane = (NSUInteger)cmn_resolve_att->slice; + break; + default: break; + } + } + } + const _sg_image_t* ds_att_img = atts->mtl.depth_stencil.image; + if (0 != ds_att_img) { + SOKOL_ASSERT(ds_att_img->slot.state == SG_RESOURCESTATE_VALID); + SOKOL_ASSERT(ds_att_img->slot.id == atts->cmn.depth_stencil.image_id.id); + SOKOL_ASSERT(ds_att_img->mtl.tex[ds_att_img->cmn.active_slot] != _SG_MTL_INVALID_SLOT_INDEX); + pass_desc.depthAttachment.texture = _sg_mtl_id(ds_att_img->mtl.tex[ds_att_img->cmn.active_slot]); + pass_desc.depthAttachment.loadAction = _sg_mtl_load_action(action->depth.load_action); + pass_desc.depthAttachment.storeAction = _sg_mtl_store_action(action->depth.store_action, false); + pass_desc.depthAttachment.clearDepth = action->depth.clear_value; + const _sg_attachment_common_t* cmn_ds_att = &atts->cmn.depth_stencil; + switch (ds_att_img->cmn.type) { + case SG_IMAGETYPE_CUBE: + case SG_IMAGETYPE_ARRAY: + pass_desc.depthAttachment.slice = (NSUInteger)cmn_ds_att->slice; + break; + case SG_IMAGETYPE_3D: + pass_desc.depthAttachment.resolveDepthPlane = (NSUInteger)cmn_ds_att->slice; + break; + default: break; + } + if (_sg_is_depth_stencil_format(ds_att_img->cmn.pixel_format)) { + pass_desc.stencilAttachment.texture = _sg_mtl_id(ds_att_img->mtl.tex[ds_att_img->cmn.active_slot]); + pass_desc.stencilAttachment.loadAction = _sg_mtl_load_action(action->stencil.load_action); + pass_desc.stencilAttachment.storeAction = _sg_mtl_store_action(action->depth.store_action, false); + pass_desc.stencilAttachment.clearStencil = action->stencil.clear_value; + switch (ds_att_img->cmn.type) { + case SG_IMAGETYPE_CUBE: + case SG_IMAGETYPE_ARRAY: + pass_desc.stencilAttachment.slice = (NSUInteger)cmn_ds_att->slice; + break; + case SG_IMAGETYPE_3D: + pass_desc.stencilAttachment.resolveDepthPlane = (NSUInteger)cmn_ds_att->slice; + break; + default: break; + } + } + } + } else { + // setup pass descriptor for swapchain rendering + // + // NOTE: at least in macOS Sonoma this no longer seems to be the case, the + // current drawable is also valid in a minimized window + // === + // an MTKView current_drawable will not be valid if window is minimized, don't do any rendering in this case + if (0 == swapchain->metal.current_drawable) { + _sg.cur_pass.valid = false; + return; + } + // pin the swapchain resources into memory so that they outlive their command buffer + // (this is necessary because the command buffer doesn't retain references) + int pass_desc_ref = _sg_mtl_add_resource(pass_desc); + _sg_mtl_release_resource(_sg.frame_index, pass_desc_ref); + + _sg.mtl.cur_drawable = (__bridge id) swapchain->metal.current_drawable; + if (swapchain->sample_count > 1) { + // multi-sampling: render into msaa texture, resolve into drawable texture + id msaa_tex = (__bridge id) swapchain->metal.msaa_color_texture; + SOKOL_ASSERT(msaa_tex != nil); + pass_desc.colorAttachments[0].texture = msaa_tex; + pass_desc.colorAttachments[0].resolveTexture = _sg.mtl.cur_drawable.texture; + pass_desc.colorAttachments[0].storeAction = MTLStoreActionMultisampleResolve; + } else { + // non-msaa: render into current_drawable + pass_desc.colorAttachments[0].texture = _sg.mtl.cur_drawable.texture; + pass_desc.colorAttachments[0].storeAction = MTLStoreActionStore; + } + pass_desc.colorAttachments[0].loadAction = _sg_mtl_load_action(action->colors[0].load_action); + const sg_color c = action->colors[0].clear_value; + pass_desc.colorAttachments[0].clearColor = MTLClearColorMake(c.r, c.g, c.b, c.a); + + // optional depth-stencil texture + if (swapchain->metal.depth_stencil_texture) { + id ds_tex = (__bridge id) swapchain->metal.depth_stencil_texture; + SOKOL_ASSERT(ds_tex != nil); + pass_desc.depthAttachment.texture = ds_tex; + pass_desc.depthAttachment.storeAction = MTLStoreActionDontCare; + pass_desc.depthAttachment.loadAction = _sg_mtl_load_action(action->depth.load_action); + pass_desc.depthAttachment.clearDepth = action->depth.clear_value; + if (_sg_is_depth_stencil_format(swapchain->depth_format)) { + pass_desc.stencilAttachment.texture = ds_tex; + pass_desc.stencilAttachment.storeAction = MTLStoreActionDontCare; + pass_desc.stencilAttachment.loadAction = _sg_mtl_load_action(action->stencil.load_action); + pass_desc.stencilAttachment.clearStencil = action->stencil.clear_value; + } + } + } + + // NOTE: at least in macOS Sonoma, the following is no longer the case, a valid + // render command encoder is also returned in a minimized window + // === + // create a render command encoder, this might return nil if window is minimized + _sg.mtl.render_cmd_encoder = [_sg.mtl.cmd_buffer renderCommandEncoderWithDescriptor:pass_desc]; + if (nil == _sg.mtl.render_cmd_encoder) { + _sg.cur_pass.valid = false; + return; + } + + #if defined(SOKOL_DEBUG) + if (pass->label) { + _sg.mtl.render_cmd_encoder.label = [NSString stringWithUTF8String:pass->label]; + } + #endif +} + +_SOKOL_PRIVATE void _sg_mtl_begin_pass(const sg_pass* pass) { + SOKOL_ASSERT(pass); + SOKOL_ASSERT(_sg.mtl.cmd_queue); + SOKOL_ASSERT(nil == _sg.mtl.compute_cmd_encoder); + SOKOL_ASSERT(nil == _sg.mtl.render_cmd_encoder); + SOKOL_ASSERT(nil == _sg.mtl.cur_drawable); + _sg_mtl_clear_state_cache(); + + // if this is the first pass in the frame, create one command buffer and blit-cmd-encoder for the entire frame + if (nil == _sg.mtl.cmd_buffer) { + // block until the oldest frame in flight has finished + dispatch_semaphore_wait(_sg.mtl.sem, DISPATCH_TIME_FOREVER); + if (_sg.desc.mtl_use_command_buffer_with_retained_references) { + _sg.mtl.cmd_buffer = [_sg.mtl.cmd_queue commandBuffer]; + } else { + _sg.mtl.cmd_buffer = [_sg.mtl.cmd_queue commandBufferWithUnretainedReferences]; + } + [_sg.mtl.cmd_buffer enqueue]; + [_sg.mtl.cmd_buffer addCompletedHandler:^(id cmd_buf) { + // NOTE: this code is called on a different thread! + _SOKOL_UNUSED(cmd_buf); + dispatch_semaphore_signal(_sg.mtl.sem); + }]; + } + + // if this is first pass in frame, get uniform buffer base pointer + if (0 == _sg.mtl.cur_ub_base_ptr) { + _sg.mtl.cur_ub_base_ptr = (uint8_t*)[_sg.mtl.uniform_buffers[_sg.mtl.cur_frame_rotate_index] contents]; + } + + if (pass->compute) { + _sg_mtl_begin_compute_pass(pass); + } else { + _sg_mtl_begin_render_pass(pass); + } + + // bind uniform buffers, those bindings remain valid for the entire pass + if (_sg.cur_pass.valid) { + _sg_mtl_bind_uniform_buffers(); + } +} + +_SOKOL_PRIVATE void _sg_mtl_end_pass(void) { + if (nil != _sg.mtl.render_cmd_encoder) { + [_sg.mtl.render_cmd_encoder endEncoding]; + // NOTE: MTLRenderCommandEncoder is autoreleased + _sg.mtl.render_cmd_encoder = nil; + } + if (nil != _sg.mtl.compute_cmd_encoder) { + [_sg.mtl.compute_cmd_encoder endEncoding]; + // NOTE: MTLComputeCommandEncoder is autoreleased + _sg.mtl.compute_cmd_encoder = nil; + + // synchronize any managed buffers written by the GPU + #if defined(_SG_TARGET_MACOS) + if (_sg_mtl_resource_options_storage_mode_managed_or_shared() == MTLResourceStorageModeManaged) { + if (_sg.compute.readwrite_sbufs.cur > 0) { + id blit_cmd_encoder = [_sg.mtl.cmd_buffer blitCommandEncoder]; + for (uint32_t i = 0; i < _sg.compute.readwrite_sbufs.cur; i++) { + _sg_buffer_t* sbuf = _sg_lookup_buffer(&_sg.pools, _sg.compute.readwrite_sbufs.items[i]); + if (sbuf) { + [blit_cmd_encoder synchronizeResource:_sg_mtl_id(sbuf->mtl.buf[sbuf->cmn.active_slot])]; + } + } + [blit_cmd_encoder endEncoding]; + } + } + #endif + } + // if this is a swapchain pass, present the drawable + if (nil != _sg.mtl.cur_drawable) { + [_sg.mtl.cmd_buffer presentDrawable:_sg.mtl.cur_drawable]; + _sg.mtl.cur_drawable = nil; + } +} + +_SOKOL_PRIVATE void _sg_mtl_commit(void) { + SOKOL_ASSERT(nil == _sg.mtl.render_cmd_encoder); + SOKOL_ASSERT(nil == _sg.mtl.compute_cmd_encoder); + SOKOL_ASSERT(nil != _sg.mtl.cmd_buffer); + + // commit the frame's command buffer + [_sg.mtl.cmd_buffer commit]; + + // garbage-collect resources pending for release + _sg_mtl_garbage_collect(_sg.frame_index); + + // rotate uniform buffer slot + if (++_sg.mtl.cur_frame_rotate_index >= SG_NUM_INFLIGHT_FRAMES) { + _sg.mtl.cur_frame_rotate_index = 0; + } + _sg.mtl.cur_ub_offset = 0; + _sg.mtl.cur_ub_base_ptr = 0; + // NOTE: MTLCommandBuffer is autoreleased + _sg.mtl.cmd_buffer = nil; +} + +_SOKOL_PRIVATE void _sg_mtl_apply_viewport(int x, int y, int w, int h, bool origin_top_left) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + SOKOL_ASSERT(_sg.cur_pass.height > 0); + MTLViewport vp; + vp.originX = (double) x; + vp.originY = (double) (origin_top_left ? y : (_sg.cur_pass.height - (y + h))); + vp.width = (double) w; + vp.height = (double) h; + vp.znear = 0.0; + vp.zfar = 1.0; + [_sg.mtl.render_cmd_encoder setViewport:vp]; +} + +_SOKOL_PRIVATE void _sg_mtl_apply_scissor_rect(int x, int y, int w, int h, bool origin_top_left) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + SOKOL_ASSERT(_sg.cur_pass.width > 0); + SOKOL_ASSERT(_sg.cur_pass.height > 0); + // clip against framebuffer rect + const _sg_recti_t clip = _sg_clipi(x, y, w, h, _sg.cur_pass.width, _sg.cur_pass.height); + MTLScissorRect r; + r.x = (NSUInteger)clip.x; + r.y = (NSUInteger) (origin_top_left ? clip.y : (_sg.cur_pass.height - (clip.y + clip.h))); + r.width = (NSUInteger)clip.w; + r.height = (NSUInteger)clip.h; + [_sg.mtl.render_cmd_encoder setScissorRect:r]; +} + +_SOKOL_PRIVATE void _sg_mtl_apply_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + SOKOL_ASSERT(pip->shader && (pip->cmn.shader_id.id == pip->shader->slot.id)); + if (_sg.mtl.state_cache.cur_pipeline_id.id != pip->slot.id) { + _sg.mtl.state_cache.cur_pipeline = pip; + _sg.mtl.state_cache.cur_pipeline_id.id = pip->slot.id; + if (pip->cmn.is_compute) { + SOKOL_ASSERT(_sg.cur_pass.is_compute); + SOKOL_ASSERT(nil != _sg.mtl.compute_cmd_encoder); + SOKOL_ASSERT(pip->mtl.cps != _SG_MTL_INVALID_SLOT_INDEX); + [_sg.mtl.compute_cmd_encoder setComputePipelineState:_sg_mtl_id(pip->mtl.cps)]; + } else { + SOKOL_ASSERT(!_sg.cur_pass.is_compute); + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + sg_color c = pip->cmn.blend_color; + [_sg.mtl.render_cmd_encoder setBlendColorRed:c.r green:c.g blue:c.b alpha:c.a]; + _sg_stats_add(metal.pipeline.num_set_blend_color, 1); + [_sg.mtl.render_cmd_encoder setCullMode:pip->mtl.cull_mode]; + _sg_stats_add(metal.pipeline.num_set_cull_mode, 1); + [_sg.mtl.render_cmd_encoder setFrontFacingWinding:pip->mtl.winding]; + _sg_stats_add(metal.pipeline.num_set_front_facing_winding, 1); + [_sg.mtl.render_cmd_encoder setStencilReferenceValue:pip->mtl.stencil_ref]; + _sg_stats_add(metal.pipeline.num_set_stencil_reference_value, 1); + [_sg.mtl.render_cmd_encoder setDepthBias:pip->cmn.depth.bias slopeScale:pip->cmn.depth.bias_slope_scale clamp:pip->cmn.depth.bias_clamp]; + _sg_stats_add(metal.pipeline.num_set_depth_bias, 1); + SOKOL_ASSERT(pip->mtl.rps != _SG_MTL_INVALID_SLOT_INDEX); + [_sg.mtl.render_cmd_encoder setRenderPipelineState:_sg_mtl_id(pip->mtl.rps)]; + _sg_stats_add(metal.pipeline.num_set_render_pipeline_state, 1); + SOKOL_ASSERT(pip->mtl.dss != _SG_MTL_INVALID_SLOT_INDEX); + [_sg.mtl.render_cmd_encoder setDepthStencilState:_sg_mtl_id(pip->mtl.dss)]; + _sg_stats_add(metal.pipeline.num_set_depth_stencil_state, 1); + } + } +} + +_SOKOL_PRIVATE bool _sg_mtl_apply_bindings(_sg_bindings_t* bnd) { + SOKOL_ASSERT(bnd); + SOKOL_ASSERT(bnd->pip); + SOKOL_ASSERT(bnd->pip && bnd->pip->shader); + SOKOL_ASSERT(bnd->pip->shader->slot.id == bnd->pip->cmn.shader_id.id); + const _sg_shader_t* shd = bnd->pip->shader; + + // don't set vertex- and index-buffers in compute passes + if (!_sg.cur_pass.is_compute) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + // store index buffer binding, this will be needed later in sg_draw() + _sg.mtl.state_cache.cur_indexbuffer = bnd->ib; + _sg.mtl.state_cache.cur_indexbuffer_offset = bnd->ib_offset; + if (bnd->ib) { + SOKOL_ASSERT(bnd->pip->cmn.index_type != SG_INDEXTYPE_NONE); + _sg.mtl.state_cache.cur_indexbuffer_id.id = bnd->ib->slot.id; + } else { + SOKOL_ASSERT(bnd->pip->cmn.index_type == SG_INDEXTYPE_NONE); + _sg.mtl.state_cache.cur_indexbuffer_id.id = SG_INVALID_ID; + } + // apply vertex buffers + for (size_t i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + const _sg_buffer_t* vb = bnd->vbs[i]; + if (vb == 0) { + continue; + } + const NSUInteger mtl_slot = _sg_mtl_vertexbuffer_bindslot(i); + SOKOL_ASSERT(mtl_slot < _SG_MTL_MAX_STAGE_BUFFER_BINDINGS); + const int vb_offset = bnd->vb_offsets[i]; + if ((_sg.mtl.state_cache.cur_vs_buffer_ids[mtl_slot].id != vb->slot.id) || + (_sg.mtl.state_cache.cur_vs_buffer_offsets[mtl_slot] != vb_offset)) + { + _sg.mtl.state_cache.cur_vs_buffer_offsets[mtl_slot] = vb_offset; + if (_sg.mtl.state_cache.cur_vs_buffer_ids[mtl_slot].id != vb->slot.id) { + // vertex buffer has changed + _sg.mtl.state_cache.cur_vs_buffer_ids[mtl_slot].id = vb->slot.id; + SOKOL_ASSERT(vb->mtl.buf[vb->cmn.active_slot] != _SG_MTL_INVALID_SLOT_INDEX); + [_sg.mtl.render_cmd_encoder setVertexBuffer:_sg_mtl_id(vb->mtl.buf[vb->cmn.active_slot]) + offset:(NSUInteger)vb_offset + atIndex:mtl_slot]; + } else { + // only vertex buffer offset has changed + [_sg.mtl.render_cmd_encoder setVertexBufferOffset:(NSUInteger)vb_offset atIndex:mtl_slot]; + } + _sg_stats_add(metal.bindings.num_set_vertex_buffer, 1); + } + } + } + + // apply image bindings + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + const _sg_image_t* img = bnd->imgs[i]; + if (img == 0) { + continue; + } + SOKOL_ASSERT(img->mtl.tex[img->cmn.active_slot] != _SG_MTL_INVALID_SLOT_INDEX); + const sg_shader_stage stage = shd->cmn.images[i].stage; + SOKOL_ASSERT((stage == SG_SHADERSTAGE_VERTEX) || (stage == SG_SHADERSTAGE_FRAGMENT) || (stage == SG_SHADERSTAGE_COMPUTE)); + const NSUInteger mtl_slot = shd->mtl.img_texture_n[i]; + SOKOL_ASSERT(mtl_slot < _SG_MTL_MAX_STAGE_IMAGE_BINDINGS); + if (stage == SG_SHADERSTAGE_VERTEX) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + if (_sg.mtl.state_cache.cur_vs_image_ids[mtl_slot].id != img->slot.id) { + _sg.mtl.state_cache.cur_vs_image_ids[mtl_slot].id = img->slot.id; + [_sg.mtl.render_cmd_encoder setVertexTexture:_sg_mtl_id(img->mtl.tex[img->cmn.active_slot]) atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_vertex_texture, 1); + } + } else if (stage == SG_SHADERSTAGE_FRAGMENT) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + if (_sg.mtl.state_cache.cur_fs_image_ids[mtl_slot].id != img->slot.id) { + _sg.mtl.state_cache.cur_fs_image_ids[mtl_slot].id = img->slot.id; + [_sg.mtl.render_cmd_encoder setFragmentTexture:_sg_mtl_id(img->mtl.tex[img->cmn.active_slot]) atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_fragment_texture, 1); + } + } else if (stage == SG_SHADERSTAGE_COMPUTE) { + SOKOL_ASSERT(nil != _sg.mtl.compute_cmd_encoder); + if (_sg.mtl.state_cache.cur_cs_image_ids[mtl_slot].id != img->slot.id) { + _sg.mtl.state_cache.cur_cs_image_ids[mtl_slot].id = img->slot.id; + [_sg.mtl.compute_cmd_encoder setTexture:_sg_mtl_id(img->mtl.tex[img->cmn.active_slot]) atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_compute_texture, 1); + } + } + } + + // apply sampler bindings + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + const _sg_sampler_t* smp = bnd->smps[i]; + if (smp == 0) { + continue; + } + SOKOL_ASSERT(smp->mtl.sampler_state != _SG_MTL_INVALID_SLOT_INDEX); + const sg_shader_stage stage = shd->cmn.samplers[i].stage; + SOKOL_ASSERT((stage == SG_SHADERSTAGE_VERTEX) || (stage == SG_SHADERSTAGE_FRAGMENT) || (stage == SG_SHADERSTAGE_COMPUTE)); + const NSUInteger mtl_slot = shd->mtl.smp_sampler_n[i]; + SOKOL_ASSERT(mtl_slot < _SG_MTL_MAX_STAGE_SAMPLER_BINDINGS); + if (stage == SG_SHADERSTAGE_VERTEX) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + if (_sg.mtl.state_cache.cur_vs_sampler_ids[mtl_slot].id != smp->slot.id) { + _sg.mtl.state_cache.cur_vs_sampler_ids[mtl_slot].id = smp->slot.id; + [_sg.mtl.render_cmd_encoder setVertexSamplerState:_sg_mtl_id(smp->mtl.sampler_state) atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_vertex_sampler_state, 1); + } + } else if (stage == SG_SHADERSTAGE_FRAGMENT) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + if (_sg.mtl.state_cache.cur_fs_sampler_ids[mtl_slot].id != smp->slot.id) { + _sg.mtl.state_cache.cur_fs_sampler_ids[mtl_slot].id = smp->slot.id; + [_sg.mtl.render_cmd_encoder setFragmentSamplerState:_sg_mtl_id(smp->mtl.sampler_state) atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_fragment_sampler_state, 1); + } + } else if (stage == SG_SHADERSTAGE_COMPUTE) { + SOKOL_ASSERT(nil != _sg.mtl.compute_cmd_encoder); + if (_sg.mtl.state_cache.cur_cs_sampler_ids[mtl_slot].id != smp->slot.id) { + _sg.mtl.state_cache.cur_cs_sampler_ids[mtl_slot].id = smp->slot.id; + [_sg.mtl.compute_cmd_encoder setSamplerState:_sg_mtl_id(smp->mtl.sampler_state) atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_compute_sampler_state, 1); + } + } + } + + // apply storage buffer bindings + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + const _sg_buffer_t* sbuf = bnd->sbufs[i]; + if (sbuf == 0) { + continue; + } + SOKOL_ASSERT(sbuf->mtl.buf[sbuf->cmn.active_slot] != _SG_MTL_INVALID_SLOT_INDEX); + const sg_shader_stage stage = shd->cmn.storage_buffers[i].stage; + SOKOL_ASSERT((stage == SG_SHADERSTAGE_VERTEX) || (stage == SG_SHADERSTAGE_FRAGMENT) || (stage == SG_SHADERSTAGE_COMPUTE)); + const NSUInteger mtl_slot = shd->mtl.sbuf_buffer_n[i]; + SOKOL_ASSERT(mtl_slot < _SG_MTL_MAX_STAGE_UB_SBUF_BINDINGS); + if (stage == SG_SHADERSTAGE_VERTEX) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + if (_sg.mtl.state_cache.cur_vs_buffer_ids[mtl_slot].id != sbuf->slot.id) { + _sg.mtl.state_cache.cur_vs_buffer_ids[mtl_slot].id = sbuf->slot.id; + [_sg.mtl.render_cmd_encoder setVertexBuffer:_sg_mtl_id(sbuf->mtl.buf[sbuf->cmn.active_slot]) offset:0 atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_vertex_buffer, 1); + } + } else if (stage == SG_SHADERSTAGE_FRAGMENT) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + if (_sg.mtl.state_cache.cur_fs_buffer_ids[mtl_slot].id != sbuf->slot.id) { + _sg.mtl.state_cache.cur_fs_buffer_ids[mtl_slot].id = sbuf->slot.id; + [_sg.mtl.render_cmd_encoder setFragmentBuffer:_sg_mtl_id(sbuf->mtl.buf[sbuf->cmn.active_slot]) offset:0 atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_fragment_buffer, 1); + } + } else if (stage == SG_SHADERSTAGE_COMPUTE) { + SOKOL_ASSERT(nil != _sg.mtl.compute_cmd_encoder); + if (_sg.mtl.state_cache.cur_cs_buffer_ids[mtl_slot].id != sbuf->slot.id) { + _sg.mtl.state_cache.cur_cs_buffer_ids[mtl_slot].id = sbuf->slot.id; + [_sg.mtl.compute_cmd_encoder setBuffer:_sg_mtl_id(sbuf->mtl.buf[sbuf->cmn.active_slot]) offset:0 atIndex:mtl_slot]; + _sg_stats_add(metal.bindings.num_set_compute_buffer, 1); + } + } + } + return true; +} + +_SOKOL_PRIVATE void _sg_mtl_apply_uniforms(int ub_slot, const sg_range* data) { + SOKOL_ASSERT((ub_slot >= 0) && (ub_slot < SG_MAX_UNIFORMBLOCK_BINDSLOTS)); + SOKOL_ASSERT(((size_t)_sg.mtl.cur_ub_offset + data->size) <= (size_t)_sg.mtl.ub_size); + SOKOL_ASSERT((_sg.mtl.cur_ub_offset & (_SG_MTL_UB_ALIGN-1)) == 0); + const _sg_pipeline_t* pip = _sg.mtl.state_cache.cur_pipeline; + SOKOL_ASSERT(pip && pip->shader); + SOKOL_ASSERT(pip->slot.id == _sg.mtl.state_cache.cur_pipeline_id.id); + const _sg_shader_t* shd = pip->shader; + SOKOL_ASSERT(shd->slot.id == pip->cmn.shader_id.id); + SOKOL_ASSERT(data->size == shd->cmn.uniform_blocks[ub_slot].size); + + const sg_shader_stage stage = shd->cmn.uniform_blocks[ub_slot].stage; + const NSUInteger mtl_slot = shd->mtl.ub_buffer_n[ub_slot]; + + // copy to global uniform buffer, record offset into cmd encoder, and advance offset + uint8_t* dst = &_sg.mtl.cur_ub_base_ptr[_sg.mtl.cur_ub_offset]; + memcpy(dst, data->ptr, data->size); + if (stage == SG_SHADERSTAGE_VERTEX) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + [_sg.mtl.render_cmd_encoder setVertexBufferOffset:(NSUInteger)_sg.mtl.cur_ub_offset atIndex:mtl_slot]; + _sg_stats_add(metal.uniforms.num_set_vertex_buffer_offset, 1); + } else if (stage == SG_SHADERSTAGE_FRAGMENT) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + [_sg.mtl.render_cmd_encoder setFragmentBufferOffset:(NSUInteger)_sg.mtl.cur_ub_offset atIndex:mtl_slot]; + _sg_stats_add(metal.uniforms.num_set_fragment_buffer_offset, 1); + } else if (stage == SG_SHADERSTAGE_COMPUTE) { + SOKOL_ASSERT(nil != _sg.mtl.compute_cmd_encoder); + [_sg.mtl.compute_cmd_encoder setBufferOffset:(NSUInteger)_sg.mtl.cur_ub_offset atIndex:mtl_slot]; + _sg_stats_add(metal.uniforms.num_set_compute_buffer_offset, 1); + } else { + SOKOL_UNREACHABLE; + } + _sg.mtl.cur_ub_offset = _sg_roundup(_sg.mtl.cur_ub_offset + (int)data->size, _SG_MTL_UB_ALIGN); +} + +_SOKOL_PRIVATE void _sg_mtl_draw(int base_element, int num_elements, int num_instances) { + SOKOL_ASSERT(nil != _sg.mtl.render_cmd_encoder); + SOKOL_ASSERT(_sg.mtl.state_cache.cur_pipeline && (_sg.mtl.state_cache.cur_pipeline->slot.id == _sg.mtl.state_cache.cur_pipeline_id.id)); + if (SG_INDEXTYPE_NONE != _sg.mtl.state_cache.cur_pipeline->cmn.index_type) { + // indexed rendering + SOKOL_ASSERT(_sg.mtl.state_cache.cur_indexbuffer && (_sg.mtl.state_cache.cur_indexbuffer->slot.id == _sg.mtl.state_cache.cur_indexbuffer_id.id)); + const _sg_buffer_t* ib = _sg.mtl.state_cache.cur_indexbuffer; + SOKOL_ASSERT(ib->mtl.buf[ib->cmn.active_slot] != _SG_MTL_INVALID_SLOT_INDEX); + const NSUInteger index_buffer_offset = (NSUInteger) (_sg.mtl.state_cache.cur_indexbuffer_offset + base_element * _sg.mtl.state_cache.cur_pipeline->mtl.index_size); + [_sg.mtl.render_cmd_encoder drawIndexedPrimitives:_sg.mtl.state_cache.cur_pipeline->mtl.prim_type + indexCount:(NSUInteger)num_elements + indexType:_sg.mtl.state_cache.cur_pipeline->mtl.index_type + indexBuffer:_sg_mtl_id(ib->mtl.buf[ib->cmn.active_slot]) + indexBufferOffset:index_buffer_offset + instanceCount:(NSUInteger)num_instances]; + } else { + // non-indexed rendering + [_sg.mtl.render_cmd_encoder drawPrimitives:_sg.mtl.state_cache.cur_pipeline->mtl.prim_type + vertexStart:(NSUInteger)base_element + vertexCount:(NSUInteger)num_elements + instanceCount:(NSUInteger)num_instances]; + } +} + +_SOKOL_PRIVATE void _sg_mtl_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) { + SOKOL_ASSERT(nil != _sg.mtl.compute_cmd_encoder); + SOKOL_ASSERT(_sg.mtl.state_cache.cur_pipeline && (_sg.mtl.state_cache.cur_pipeline->slot.id == _sg.mtl.state_cache.cur_pipeline_id.id)); + const _sg_pipeline_t* cur_pip = _sg.mtl.state_cache.cur_pipeline; + const MTLSize thread_groups = MTLSizeMake( + (NSUInteger)num_groups_x, + (NSUInteger)num_groups_y, + (NSUInteger)num_groups_z); + const MTLSize threads_per_threadgroup = cur_pip->mtl.threads_per_threadgroup; + [_sg.mtl.compute_cmd_encoder dispatchThreadgroups:thread_groups threadsPerThreadgroup:threads_per_threadgroup]; +} + +_SOKOL_PRIVATE void _sg_mtl_update_buffer(_sg_buffer_t* buf, const sg_range* data) { + SOKOL_ASSERT(buf && data && data->ptr && (data->size > 0)); + if (++buf->cmn.active_slot >= buf->cmn.num_slots) { + buf->cmn.active_slot = 0; + } + __unsafe_unretained id mtl_buf = _sg_mtl_id(buf->mtl.buf[buf->cmn.active_slot]); + void* dst_ptr = [mtl_buf contents]; + memcpy(dst_ptr, data->ptr, data->size); + #if defined(_SG_TARGET_MACOS) + if (_sg_mtl_resource_options_storage_mode_managed_or_shared() == MTLResourceStorageModeManaged) { + [mtl_buf didModifyRange:NSMakeRange(0, data->size)]; + } + #endif +} + +_SOKOL_PRIVATE void _sg_mtl_append_buffer(_sg_buffer_t* buf, const sg_range* data, bool new_frame) { + SOKOL_ASSERT(buf && data && data->ptr && (data->size > 0)); + if (new_frame) { + if (++buf->cmn.active_slot >= buf->cmn.num_slots) { + buf->cmn.active_slot = 0; + } + } + __unsafe_unretained id mtl_buf = _sg_mtl_id(buf->mtl.buf[buf->cmn.active_slot]); + uint8_t* dst_ptr = (uint8_t*) [mtl_buf contents]; + dst_ptr += buf->cmn.append_pos; + memcpy(dst_ptr, data->ptr, data->size); + #if defined(_SG_TARGET_MACOS) + if (_sg_mtl_resource_options_storage_mode_managed_or_shared() == MTLResourceStorageModeManaged) { + [mtl_buf didModifyRange:NSMakeRange((NSUInteger)buf->cmn.append_pos, (NSUInteger)data->size)]; + } + #endif +} + +_SOKOL_PRIVATE void _sg_mtl_update_image(_sg_image_t* img, const sg_image_data* data) { + SOKOL_ASSERT(img && data); + if (++img->cmn.active_slot >= img->cmn.num_slots) { + img->cmn.active_slot = 0; + } + __unsafe_unretained id mtl_tex = _sg_mtl_id(img->mtl.tex[img->cmn.active_slot]); + _sg_mtl_copy_image_data(img, mtl_tex, data); +} + +_SOKOL_PRIVATE void _sg_mtl_push_debug_group(const char* name) { + SOKOL_ASSERT(name); + if (_sg.mtl.render_cmd_encoder) { + [_sg.mtl.render_cmd_encoder pushDebugGroup:[NSString stringWithUTF8String:name]]; + } else if (_sg.mtl.compute_cmd_encoder) { + [_sg.mtl.compute_cmd_encoder pushDebugGroup:[NSString stringWithUTF8String:name]]; + } +} + +_SOKOL_PRIVATE void _sg_mtl_pop_debug_group(void) { + if (_sg.mtl.render_cmd_encoder) { + [_sg.mtl.render_cmd_encoder popDebugGroup]; + } else if (_sg.mtl.compute_cmd_encoder) { + [_sg.mtl.compute_cmd_encoder popDebugGroup]; + } +} + +// ██ ██ ███████ ██████ ██████ ██████ ██ ██ ██████ █████ ██████ ██ ██ ███████ ███ ██ ██████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ████ ██ ██ ██ +// ██ █ ██ █████ ██████ ██ ███ ██████ ██ ██ ██████ ███████ ██ █████ █████ ██ ██ ██ ██ ██ +// ██ ███ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ███ ███ ███████ ██████ ██████ ██ ██████ ██████ ██ ██ ██████ ██ ██ ███████ ██ ████ ██████ +// +// >>webgpu +// >>wgpu +#elif defined(SOKOL_WGPU) + +#if !defined(__EMSCRIPTEN__) +// FIXME: webgpu.h differences between Dawn and Emscripten webgpu.h +#define wgpuBufferReference wgpuBufferAddRef +#define wgpuTextureReference wgpuTextureAddRef +#define wgpuTextureViewReference wgpuTextureViewAddRef +#define wgpuSamplerReference wgpuSamplerAddRef +#define WGPUSType_ShaderModuleWGSLDescriptor WGPUSType_ShaderSourceWGSL +_SOKOL_PRIVATE WGPUStringView _sg_wgpu_stringview(const char* str) { + WGPUStringView res; + if (str) { + res.data = str; + res.length = strlen(str); + } else { + res.data = 0; + res.length = 0; + } + return res; +} +_SOKOL_PRIVATE WGPUOptionalBool _sg_wgpu_optional_bool(bool b) { + return b ? WGPUOptionalBool_True : WGPUOptionalBool_False; +} +#else +#define _sg_wgpu_stringview(str) str +#define _sg_wgpu_optional_bool(b) (b) +#endif + +_SOKOL_PRIVATE WGPUBufferUsage _sg_wgpu_buffer_usage(sg_buffer_type t, sg_usage u) { + // FIXME: change to WGPUBufferUsage once Emscripten and Dawn webgpu.h agree + int res = 0; + if (SG_BUFFERTYPE_VERTEXBUFFER == t) { + res = WGPUBufferUsage_Vertex; + } else if (SG_BUFFERTYPE_STORAGEBUFFER == t) { + res = WGPUBufferUsage_Storage; + } else { + res = WGPUBufferUsage_Index; + } + if (SG_USAGE_IMMUTABLE != u) { + res |= WGPUBufferUsage_CopyDst; + } + return (WGPUBufferUsage)res; +} + +_SOKOL_PRIVATE WGPULoadOp _sg_wgpu_load_op(WGPUTextureView view, sg_load_action a) { + if (0 == view) { + return WGPULoadOp_Undefined; + } else switch (a) { + case SG_LOADACTION_CLEAR: + case SG_LOADACTION_DONTCARE: + return WGPULoadOp_Clear; + case SG_LOADACTION_LOAD: + return WGPULoadOp_Load; + default: + SOKOL_UNREACHABLE; + return WGPULoadOp_Force32; + } +} + +_SOKOL_PRIVATE WGPUStoreOp _sg_wgpu_store_op(WGPUTextureView view, sg_store_action a) { + if (0 == view) { + return WGPUStoreOp_Undefined; + } else switch (a) { + case SG_STOREACTION_STORE: + return WGPUStoreOp_Store; + case SG_STOREACTION_DONTCARE: + return WGPUStoreOp_Discard; + default: + SOKOL_UNREACHABLE; + return WGPUStoreOp_Force32; + } +} + +_SOKOL_PRIVATE WGPUTextureViewDimension _sg_wgpu_texture_view_dimension(sg_image_type t) { + switch (t) { + case SG_IMAGETYPE_2D: return WGPUTextureViewDimension_2D; + case SG_IMAGETYPE_CUBE: return WGPUTextureViewDimension_Cube; + case SG_IMAGETYPE_3D: return WGPUTextureViewDimension_3D; + case SG_IMAGETYPE_ARRAY: return WGPUTextureViewDimension_2DArray; + default: SOKOL_UNREACHABLE; return WGPUTextureViewDimension_Force32; + } +} + +_SOKOL_PRIVATE WGPUTextureDimension _sg_wgpu_texture_dimension(sg_image_type t) { + if (SG_IMAGETYPE_3D == t) { + return WGPUTextureDimension_3D; + } else { + return WGPUTextureDimension_2D; + } +} + +_SOKOL_PRIVATE WGPUTextureSampleType _sg_wgpu_texture_sample_type(sg_image_sample_type t, bool msaa) { + switch (t) { + case SG_IMAGESAMPLETYPE_FLOAT: return msaa ? WGPUTextureSampleType_UnfilterableFloat : WGPUTextureSampleType_Float; + case SG_IMAGESAMPLETYPE_DEPTH: return WGPUTextureSampleType_Depth; + case SG_IMAGESAMPLETYPE_SINT: return WGPUTextureSampleType_Sint; + case SG_IMAGESAMPLETYPE_UINT: return WGPUTextureSampleType_Uint; + case SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT: return WGPUTextureSampleType_UnfilterableFloat; + default: SOKOL_UNREACHABLE; return WGPUTextureSampleType_Force32; + } +} + +_SOKOL_PRIVATE WGPUSamplerBindingType _sg_wgpu_sampler_binding_type(sg_sampler_type t) { + switch (t) { + case SG_SAMPLERTYPE_FILTERING: return WGPUSamplerBindingType_Filtering; + case SG_SAMPLERTYPE_COMPARISON: return WGPUSamplerBindingType_Comparison; + case SG_SAMPLERTYPE_NONFILTERING: return WGPUSamplerBindingType_NonFiltering; + default: SOKOL_UNREACHABLE; return WGPUSamplerBindingType_Force32; + } +} + +_SOKOL_PRIVATE WGPUAddressMode _sg_wgpu_sampler_address_mode(sg_wrap m) { + switch (m) { + case SG_WRAP_REPEAT: + return WGPUAddressMode_Repeat; + case SG_WRAP_CLAMP_TO_EDGE: + case SG_WRAP_CLAMP_TO_BORDER: + return WGPUAddressMode_ClampToEdge; + case SG_WRAP_MIRRORED_REPEAT: + return WGPUAddressMode_MirrorRepeat; + default: + SOKOL_UNREACHABLE; + return WGPUAddressMode_Force32; + } +} + +_SOKOL_PRIVATE WGPUFilterMode _sg_wgpu_sampler_minmag_filter(sg_filter f) { + switch (f) { + case SG_FILTER_NEAREST: + return WGPUFilterMode_Nearest; + case SG_FILTER_LINEAR: + return WGPUFilterMode_Linear; + default: + SOKOL_UNREACHABLE; + return WGPUFilterMode_Force32; + } +} + +_SOKOL_PRIVATE WGPUMipmapFilterMode _sg_wgpu_sampler_mipmap_filter(sg_filter f) { + switch (f) { + case SG_FILTER_NEAREST: + return WGPUMipmapFilterMode_Nearest; + case SG_FILTER_LINEAR: + return WGPUMipmapFilterMode_Linear; + default: + SOKOL_UNREACHABLE; + return WGPUMipmapFilterMode_Force32; + } +} + +_SOKOL_PRIVATE WGPUIndexFormat _sg_wgpu_indexformat(sg_index_type t) { + // NOTE: there's no WGPUIndexFormat_None + return (t == SG_INDEXTYPE_UINT16) ? WGPUIndexFormat_Uint16 : WGPUIndexFormat_Uint32; +} + +_SOKOL_PRIVATE WGPUIndexFormat _sg_wgpu_stripindexformat(sg_primitive_type prim_type, sg_index_type idx_type) { + if (idx_type == SG_INDEXTYPE_NONE) { + return WGPUIndexFormat_Undefined; + } else if ((prim_type == SG_PRIMITIVETYPE_LINE_STRIP) || (prim_type == SG_PRIMITIVETYPE_TRIANGLE_STRIP)) { + return _sg_wgpu_indexformat(idx_type); + } else { + return WGPUIndexFormat_Undefined; + } +} + +_SOKOL_PRIVATE WGPUVertexStepMode _sg_wgpu_stepmode(sg_vertex_step s) { + return (s == SG_VERTEXSTEP_PER_VERTEX) ? WGPUVertexStepMode_Vertex : WGPUVertexStepMode_Instance; +} + +_SOKOL_PRIVATE WGPUVertexFormat _sg_wgpu_vertexformat(sg_vertex_format f) { + switch (f) { + case SG_VERTEXFORMAT_FLOAT: return WGPUVertexFormat_Float32; + case SG_VERTEXFORMAT_FLOAT2: return WGPUVertexFormat_Float32x2; + case SG_VERTEXFORMAT_FLOAT3: return WGPUVertexFormat_Float32x3; + case SG_VERTEXFORMAT_FLOAT4: return WGPUVertexFormat_Float32x4; + case SG_VERTEXFORMAT_INT: return WGPUVertexFormat_Sint32; + case SG_VERTEXFORMAT_INT2: return WGPUVertexFormat_Sint32x2; + case SG_VERTEXFORMAT_INT3: return WGPUVertexFormat_Sint32x3; + case SG_VERTEXFORMAT_INT4: return WGPUVertexFormat_Sint32x4; + case SG_VERTEXFORMAT_UINT: return WGPUVertexFormat_Uint32; + case SG_VERTEXFORMAT_UINT2: return WGPUVertexFormat_Uint32x2; + case SG_VERTEXFORMAT_UINT3: return WGPUVertexFormat_Uint32x3; + case SG_VERTEXFORMAT_UINT4: return WGPUVertexFormat_Uint32x4; + case SG_VERTEXFORMAT_BYTE4: return WGPUVertexFormat_Sint8x4; + case SG_VERTEXFORMAT_BYTE4N: return WGPUVertexFormat_Snorm8x4; + case SG_VERTEXFORMAT_UBYTE4: return WGPUVertexFormat_Uint8x4; + case SG_VERTEXFORMAT_UBYTE4N: return WGPUVertexFormat_Unorm8x4; + case SG_VERTEXFORMAT_SHORT2: return WGPUVertexFormat_Sint16x2; + case SG_VERTEXFORMAT_SHORT2N: return WGPUVertexFormat_Snorm16x2; + case SG_VERTEXFORMAT_USHORT2: return WGPUVertexFormat_Uint16x2; + case SG_VERTEXFORMAT_USHORT2N: return WGPUVertexFormat_Unorm16x2; + case SG_VERTEXFORMAT_SHORT4: return WGPUVertexFormat_Sint16x4; + case SG_VERTEXFORMAT_SHORT4N: return WGPUVertexFormat_Snorm16x4; + case SG_VERTEXFORMAT_USHORT4: return WGPUVertexFormat_Uint16x4; + case SG_VERTEXFORMAT_USHORT4N: return WGPUVertexFormat_Unorm16x4; + case SG_VERTEXFORMAT_UINT10_N2: return WGPUVertexFormat_Unorm10_10_10_2; + case SG_VERTEXFORMAT_HALF2: return WGPUVertexFormat_Float16x2; + case SG_VERTEXFORMAT_HALF4: return WGPUVertexFormat_Float16x4; + default: + SOKOL_UNREACHABLE; + return WGPUVertexFormat_Force32; + } +} + +_SOKOL_PRIVATE WGPUPrimitiveTopology _sg_wgpu_topology(sg_primitive_type t) { + switch (t) { + case SG_PRIMITIVETYPE_POINTS: return WGPUPrimitiveTopology_PointList; + case SG_PRIMITIVETYPE_LINES: return WGPUPrimitiveTopology_LineList; + case SG_PRIMITIVETYPE_LINE_STRIP: return WGPUPrimitiveTopology_LineStrip; + case SG_PRIMITIVETYPE_TRIANGLES: return WGPUPrimitiveTopology_TriangleList; + case SG_PRIMITIVETYPE_TRIANGLE_STRIP: return WGPUPrimitiveTopology_TriangleStrip; + default: + SOKOL_UNREACHABLE; + return WGPUPrimitiveTopology_Force32; + } +} + +_SOKOL_PRIVATE WGPUFrontFace _sg_wgpu_frontface(sg_face_winding fw) { + return (fw == SG_FACEWINDING_CCW) ? WGPUFrontFace_CCW : WGPUFrontFace_CW; +} + +_SOKOL_PRIVATE WGPUCullMode _sg_wgpu_cullmode(sg_cull_mode cm) { + switch (cm) { + case SG_CULLMODE_NONE: return WGPUCullMode_None; + case SG_CULLMODE_FRONT: return WGPUCullMode_Front; + case SG_CULLMODE_BACK: return WGPUCullMode_Back; + default: + SOKOL_UNREACHABLE; + return WGPUCullMode_Force32; + } +} + +_SOKOL_PRIVATE WGPUTextureFormat _sg_wgpu_textureformat(sg_pixel_format p) { + switch (p) { + case SG_PIXELFORMAT_NONE: return WGPUTextureFormat_Undefined; + case SG_PIXELFORMAT_R8: return WGPUTextureFormat_R8Unorm; + case SG_PIXELFORMAT_R8SN: return WGPUTextureFormat_R8Snorm; + case SG_PIXELFORMAT_R8UI: return WGPUTextureFormat_R8Uint; + case SG_PIXELFORMAT_R8SI: return WGPUTextureFormat_R8Sint; + case SG_PIXELFORMAT_R16UI: return WGPUTextureFormat_R16Uint; + case SG_PIXELFORMAT_R16SI: return WGPUTextureFormat_R16Sint; + case SG_PIXELFORMAT_R16F: return WGPUTextureFormat_R16Float; + case SG_PIXELFORMAT_RG8: return WGPUTextureFormat_RG8Unorm; + case SG_PIXELFORMAT_RG8SN: return WGPUTextureFormat_RG8Snorm; + case SG_PIXELFORMAT_RG8UI: return WGPUTextureFormat_RG8Uint; + case SG_PIXELFORMAT_RG8SI: return WGPUTextureFormat_RG8Sint; + case SG_PIXELFORMAT_R32UI: return WGPUTextureFormat_R32Uint; + case SG_PIXELFORMAT_R32SI: return WGPUTextureFormat_R32Sint; + case SG_PIXELFORMAT_R32F: return WGPUTextureFormat_R32Float; + case SG_PIXELFORMAT_RG16UI: return WGPUTextureFormat_RG16Uint; + case SG_PIXELFORMAT_RG16SI: return WGPUTextureFormat_RG16Sint; + case SG_PIXELFORMAT_RG16F: return WGPUTextureFormat_RG16Float; + case SG_PIXELFORMAT_RGBA8: return WGPUTextureFormat_RGBA8Unorm; + case SG_PIXELFORMAT_SRGB8A8: return WGPUTextureFormat_RGBA8UnormSrgb; + case SG_PIXELFORMAT_RGBA8SN: return WGPUTextureFormat_RGBA8Snorm; + case SG_PIXELFORMAT_RGBA8UI: return WGPUTextureFormat_RGBA8Uint; + case SG_PIXELFORMAT_RGBA8SI: return WGPUTextureFormat_RGBA8Sint; + case SG_PIXELFORMAT_BGRA8: return WGPUTextureFormat_BGRA8Unorm; + case SG_PIXELFORMAT_RGB10A2: return WGPUTextureFormat_RGB10A2Unorm; + case SG_PIXELFORMAT_RG11B10F: return WGPUTextureFormat_RG11B10Ufloat; + case SG_PIXELFORMAT_RG32UI: return WGPUTextureFormat_RG32Uint; + case SG_PIXELFORMAT_RG32SI: return WGPUTextureFormat_RG32Sint; + case SG_PIXELFORMAT_RG32F: return WGPUTextureFormat_RG32Float; + case SG_PIXELFORMAT_RGBA16UI: return WGPUTextureFormat_RGBA16Uint; + case SG_PIXELFORMAT_RGBA16SI: return WGPUTextureFormat_RGBA16Sint; + case SG_PIXELFORMAT_RGBA16F: return WGPUTextureFormat_RGBA16Float; + case SG_PIXELFORMAT_RGBA32UI: return WGPUTextureFormat_RGBA32Uint; + case SG_PIXELFORMAT_RGBA32SI: return WGPUTextureFormat_RGBA32Sint; + case SG_PIXELFORMAT_RGBA32F: return WGPUTextureFormat_RGBA32Float; + case SG_PIXELFORMAT_DEPTH: return WGPUTextureFormat_Depth32Float; + case SG_PIXELFORMAT_DEPTH_STENCIL: return WGPUTextureFormat_Depth32FloatStencil8; + case SG_PIXELFORMAT_BC1_RGBA: return WGPUTextureFormat_BC1RGBAUnorm; + case SG_PIXELFORMAT_BC2_RGBA: return WGPUTextureFormat_BC2RGBAUnorm; + case SG_PIXELFORMAT_BC3_RGBA: return WGPUTextureFormat_BC3RGBAUnorm; + case SG_PIXELFORMAT_BC3_SRGBA: return WGPUTextureFormat_BC3RGBAUnormSrgb; + case SG_PIXELFORMAT_BC4_R: return WGPUTextureFormat_BC4RUnorm; + case SG_PIXELFORMAT_BC4_RSN: return WGPUTextureFormat_BC4RSnorm; + case SG_PIXELFORMAT_BC5_RG: return WGPUTextureFormat_BC5RGUnorm; + case SG_PIXELFORMAT_BC5_RGSN: return WGPUTextureFormat_BC5RGSnorm; + case SG_PIXELFORMAT_BC6H_RGBF: return WGPUTextureFormat_BC6HRGBFloat; + case SG_PIXELFORMAT_BC6H_RGBUF: return WGPUTextureFormat_BC6HRGBUfloat; + case SG_PIXELFORMAT_BC7_RGBA: return WGPUTextureFormat_BC7RGBAUnorm; + case SG_PIXELFORMAT_BC7_SRGBA: return WGPUTextureFormat_BC7RGBAUnormSrgb; + case SG_PIXELFORMAT_ETC2_RGB8: return WGPUTextureFormat_ETC2RGB8Unorm; + case SG_PIXELFORMAT_ETC2_RGB8A1: return WGPUTextureFormat_ETC2RGB8A1Unorm; + case SG_PIXELFORMAT_ETC2_RGBA8: return WGPUTextureFormat_ETC2RGBA8Unorm; + case SG_PIXELFORMAT_ETC2_SRGB8: return WGPUTextureFormat_ETC2RGB8UnormSrgb; + case SG_PIXELFORMAT_ETC2_SRGB8A8: return WGPUTextureFormat_ETC2RGBA8UnormSrgb; + case SG_PIXELFORMAT_EAC_R11: return WGPUTextureFormat_EACR11Unorm; + case SG_PIXELFORMAT_EAC_R11SN: return WGPUTextureFormat_EACR11Snorm; + case SG_PIXELFORMAT_EAC_RG11: return WGPUTextureFormat_EACRG11Unorm; + case SG_PIXELFORMAT_EAC_RG11SN: return WGPUTextureFormat_EACRG11Snorm; + case SG_PIXELFORMAT_RGB9E5: return WGPUTextureFormat_RGB9E5Ufloat; + case SG_PIXELFORMAT_ASTC_4x4_RGBA: return WGPUTextureFormat_ASTC4x4Unorm; + case SG_PIXELFORMAT_ASTC_4x4_SRGBA: return WGPUTextureFormat_ASTC4x4UnormSrgb; + // NOT SUPPORTED + case SG_PIXELFORMAT_R16: + case SG_PIXELFORMAT_R16SN: + case SG_PIXELFORMAT_RG16: + case SG_PIXELFORMAT_RG16SN: + case SG_PIXELFORMAT_RGBA16: + case SG_PIXELFORMAT_RGBA16SN: + return WGPUTextureFormat_Undefined; + + default: + SOKOL_UNREACHABLE; + return WGPUTextureFormat_Force32; + } +} + +_SOKOL_PRIVATE WGPUCompareFunction _sg_wgpu_comparefunc(sg_compare_func f) { + switch (f) { + case SG_COMPAREFUNC_NEVER: return WGPUCompareFunction_Never; + case SG_COMPAREFUNC_LESS: return WGPUCompareFunction_Less; + case SG_COMPAREFUNC_EQUAL: return WGPUCompareFunction_Equal; + case SG_COMPAREFUNC_LESS_EQUAL: return WGPUCompareFunction_LessEqual; + case SG_COMPAREFUNC_GREATER: return WGPUCompareFunction_Greater; + case SG_COMPAREFUNC_NOT_EQUAL: return WGPUCompareFunction_NotEqual; + case SG_COMPAREFUNC_GREATER_EQUAL: return WGPUCompareFunction_GreaterEqual; + case SG_COMPAREFUNC_ALWAYS: return WGPUCompareFunction_Always; + default: + SOKOL_UNREACHABLE; + return WGPUCompareFunction_Force32; + } +} + +_SOKOL_PRIVATE WGPUStencilOperation _sg_wgpu_stencilop(sg_stencil_op op) { + switch (op) { + case SG_STENCILOP_KEEP: return WGPUStencilOperation_Keep; + case SG_STENCILOP_ZERO: return WGPUStencilOperation_Zero; + case SG_STENCILOP_REPLACE: return WGPUStencilOperation_Replace; + case SG_STENCILOP_INCR_CLAMP: return WGPUStencilOperation_IncrementClamp; + case SG_STENCILOP_DECR_CLAMP: return WGPUStencilOperation_DecrementClamp; + case SG_STENCILOP_INVERT: return WGPUStencilOperation_Invert; + case SG_STENCILOP_INCR_WRAP: return WGPUStencilOperation_IncrementWrap; + case SG_STENCILOP_DECR_WRAP: return WGPUStencilOperation_DecrementWrap; + default: + SOKOL_UNREACHABLE; + return WGPUStencilOperation_Force32; + } +} + +_SOKOL_PRIVATE WGPUBlendOperation _sg_wgpu_blendop(sg_blend_op op) { + switch (op) { + case SG_BLENDOP_ADD: return WGPUBlendOperation_Add; + case SG_BLENDOP_SUBTRACT: return WGPUBlendOperation_Subtract; + case SG_BLENDOP_REVERSE_SUBTRACT: return WGPUBlendOperation_ReverseSubtract; + case SG_BLENDOP_MIN: return WGPUBlendOperation_Min; + case SG_BLENDOP_MAX: return WGPUBlendOperation_Max; + default: + SOKOL_UNREACHABLE; + return WGPUBlendOperation_Force32; + } +} + +_SOKOL_PRIVATE WGPUBlendFactor _sg_wgpu_blendfactor(sg_blend_factor f) { + switch (f) { + case SG_BLENDFACTOR_ZERO: return WGPUBlendFactor_Zero; + case SG_BLENDFACTOR_ONE: return WGPUBlendFactor_One; + case SG_BLENDFACTOR_SRC_COLOR: return WGPUBlendFactor_Src; + case SG_BLENDFACTOR_ONE_MINUS_SRC_COLOR: return WGPUBlendFactor_OneMinusSrc; + case SG_BLENDFACTOR_SRC_ALPHA: return WGPUBlendFactor_SrcAlpha; + case SG_BLENDFACTOR_ONE_MINUS_SRC_ALPHA: return WGPUBlendFactor_OneMinusSrcAlpha; + case SG_BLENDFACTOR_DST_COLOR: return WGPUBlendFactor_Dst; + case SG_BLENDFACTOR_ONE_MINUS_DST_COLOR: return WGPUBlendFactor_OneMinusDst; + case SG_BLENDFACTOR_DST_ALPHA: return WGPUBlendFactor_DstAlpha; + case SG_BLENDFACTOR_ONE_MINUS_DST_ALPHA: return WGPUBlendFactor_OneMinusDstAlpha; + case SG_BLENDFACTOR_SRC_ALPHA_SATURATED: return WGPUBlendFactor_SrcAlphaSaturated; + case SG_BLENDFACTOR_BLEND_COLOR: return WGPUBlendFactor_Constant; + case SG_BLENDFACTOR_ONE_MINUS_BLEND_COLOR: return WGPUBlendFactor_OneMinusConstant; + // FIXME: separate blend alpha value not supported? + case SG_BLENDFACTOR_BLEND_ALPHA: return WGPUBlendFactor_Constant; + case SG_BLENDFACTOR_ONE_MINUS_BLEND_ALPHA: return WGPUBlendFactor_OneMinusConstant; + default: + SOKOL_UNREACHABLE; + return WGPUBlendFactor_Force32; + } +} + +_SOKOL_PRIVATE WGPUColorWriteMask _sg_wgpu_colorwritemask(uint8_t m) { + // FIXME: change to WGPUColorWriteMask once Emscripten and Dawn webgpu.h agree + int res = 0; + if (0 != (m & SG_COLORMASK_R)) { + res |= WGPUColorWriteMask_Red; + } + if (0 != (m & SG_COLORMASK_G)) { + res |= WGPUColorWriteMask_Green; + } + if (0 != (m & SG_COLORMASK_B)) { + res |= WGPUColorWriteMask_Blue; + } + if (0 != (m & SG_COLORMASK_A)) { + res |= WGPUColorWriteMask_Alpha; + } + return (WGPUColorWriteMask)res; +} + +_SOKOL_PRIVATE WGPUShaderStage _sg_wgpu_shader_stage(sg_shader_stage stage) { + switch (stage) { + case SG_SHADERSTAGE_VERTEX: return WGPUShaderStage_Vertex; + case SG_SHADERSTAGE_FRAGMENT: return WGPUShaderStage_Fragment; + case SG_SHADERSTAGE_COMPUTE: return WGPUShaderStage_Compute; + default: SOKOL_UNREACHABLE; return WGPUShaderStage_None; + } +} + +_SOKOL_PRIVATE void _sg_wgpu_init_caps(void) { + _sg.backend = SG_BACKEND_WGPU; + _sg.features.origin_top_left = true; + _sg.features.image_clamp_to_border = false; + _sg.features.mrt_independent_blend_state = true; + _sg.features.mrt_independent_write_mask = true; + _sg.features.compute = true; + _sg.features.msaa_image_bindings = true; + + wgpuDeviceGetLimits(_sg.wgpu.dev, &_sg.wgpu.limits); + + const WGPULimits* l = &_sg.wgpu.limits.limits; + _sg.limits.max_image_size_2d = (int) l->maxTextureDimension2D; + _sg.limits.max_image_size_cube = (int) l->maxTextureDimension2D; // not a bug, see: https://github.com/gpuweb/gpuweb/issues/1327 + _sg.limits.max_image_size_3d = (int) l->maxTextureDimension3D; + _sg.limits.max_image_size_array = (int) l->maxTextureDimension2D; + _sg.limits.max_image_array_layers = (int) l->maxTextureArrayLayers; + _sg.limits.max_vertex_attrs = SG_MAX_VERTEX_ATTRIBUTES; + + // NOTE: no WGPUTextureFormat_R16Unorm + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_SRGB8A8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_BGRA8]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_R16F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RG16F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGBA16F]); + _sg_pixelformat_all(&_sg.formats[SG_PIXELFORMAT_RGB10A2]); + + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_R8SN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RG8SN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RGBA8SN]); + + // FIXME: can be made renderable via extension + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RG11B10F]); + + // NOTE: msaa rendering is possible in WebGPU, but no resolve + // which is a combination that's not currently supported in sokol-gfx + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R8UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R8SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG8UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG8SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA8UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA8SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R16UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R16SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG16UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG16SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA16UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA16SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R32UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R32SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG32UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG32SI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA32UI]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA32SI]); + + if (wgpuDeviceHasFeature(_sg.wgpu.dev, WGPUFeatureName_Float32Filterable)) { + _sg_pixelformat_sfr(&_sg.formats[SG_PIXELFORMAT_R32F]); + _sg_pixelformat_sfr(&_sg.formats[SG_PIXELFORMAT_RG32F]); + _sg_pixelformat_sfr(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + } else { + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_R32F]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RG32F]); + _sg_pixelformat_sr(&_sg.formats[SG_PIXELFORMAT_RGBA32F]); + } + + _sg_pixelformat_srmd(&_sg.formats[SG_PIXELFORMAT_DEPTH]); + _sg_pixelformat_srmd(&_sg.formats[SG_PIXELFORMAT_DEPTH_STENCIL]); + + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_RGB9E5]); + + if (wgpuDeviceHasFeature(_sg.wgpu.dev, WGPUFeatureName_TextureCompressionBC)) { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC1_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC2_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC3_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC3_SRGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC4_R]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC4_RSN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC5_RG]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC5_RGSN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC6H_RGBF]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC6H_RGBUF]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC7_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_BC7_SRGBA]); + } + if (wgpuDeviceHasFeature(_sg.wgpu.dev, WGPUFeatureName_TextureCompressionETC2)) { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGB8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_SRGB8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGB8A1]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_RGBA8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ETC2_SRGB8A8]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_R11]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_R11SN]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_RG11]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_EAC_RG11SN]); + } + + if (wgpuDeviceHasFeature(_sg.wgpu.dev, WGPUFeatureName_TextureCompressionASTC)) { + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ASTC_4x4_RGBA]); + _sg_pixelformat_sf(&_sg.formats[SG_PIXELFORMAT_ASTC_4x4_SRGBA]); + } +} + +_SOKOL_PRIVATE void _sg_wgpu_uniform_buffer_init(const sg_desc* desc) { + SOKOL_ASSERT(0 == _sg.wgpu.uniform.staging); + SOKOL_ASSERT(0 == _sg.wgpu.uniform.buf); + + // Add the max-uniform-update size (64 KB) to the requested buffer size, + // this is to prevent validation errors in the WebGPU implementation + // if the entire buffer size is used per frame. 64 KB is the allowed + // max uniform update size on NVIDIA + // + // FIXME: is this still needed? + _sg.wgpu.uniform.num_bytes = (uint32_t)(desc->uniform_buffer_size + _SG_WGPU_MAX_UNIFORM_UPDATE_SIZE); + _sg.wgpu.uniform.staging = (uint8_t*)_sg_malloc(_sg.wgpu.uniform.num_bytes); + + WGPUBufferDescriptor ub_desc; + _sg_clear(&ub_desc, sizeof(ub_desc)); + ub_desc.size = _sg.wgpu.uniform.num_bytes; + ub_desc.usage = WGPUBufferUsage_Uniform|WGPUBufferUsage_CopyDst; + _sg.wgpu.uniform.buf = wgpuDeviceCreateBuffer(_sg.wgpu.dev, &ub_desc); + SOKOL_ASSERT(_sg.wgpu.uniform.buf); +} + +_SOKOL_PRIVATE void _sg_wgpu_uniform_buffer_discard(void) { + if (_sg.wgpu.uniform.buf) { + wgpuBufferRelease(_sg.wgpu.uniform.buf); + _sg.wgpu.uniform.buf = 0; + } + if (_sg.wgpu.uniform.staging) { + _sg_free(_sg.wgpu.uniform.staging); + _sg.wgpu.uniform.staging = 0; + } +} + +_SOKOL_PRIVATE void _sg_wgpu_uniform_buffer_on_commit(void) { + wgpuQueueWriteBuffer(_sg.wgpu.queue, _sg.wgpu.uniform.buf, 0, _sg.wgpu.uniform.staging, _sg.wgpu.uniform.offset); + _sg_stats_add(wgpu.uniforms.size_write_buffer, _sg.wgpu.uniform.offset); + _sg.wgpu.uniform.offset = 0; + _sg_clear(_sg.wgpu.uniform.bind_offsets, sizeof(_sg.wgpu.uniform.bind_offsets)); +} + +_SOKOL_PRIVATE void _sg_wgpu_bindgroups_pool_init(const sg_desc* desc) { + SOKOL_ASSERT((desc->wgpu_bindgroups_cache_size > 0) && (desc->wgpu_bindgroups_cache_size < _SG_MAX_POOL_SIZE)); + _sg_wgpu_bindgroups_pool_t* p = &_sg.wgpu.bindgroups_pool; + SOKOL_ASSERT(0 == p->bindgroups); + const int pool_size = desc->wgpu_bindgroups_cache_size; + _sg_pool_init(&p->pool, pool_size); + size_t pool_byte_size = sizeof(_sg_wgpu_bindgroup_t) * (size_t)p->pool.size; + p->bindgroups = (_sg_wgpu_bindgroup_t*) _sg_malloc_clear(pool_byte_size); +} + +_SOKOL_PRIVATE void _sg_wgpu_bindgroups_pool_discard(void) { + _sg_wgpu_bindgroups_pool_t* p = &_sg.wgpu.bindgroups_pool; + SOKOL_ASSERT(p->bindgroups); + _sg_free(p->bindgroups); p->bindgroups = 0; + _sg_pool_discard(&p->pool); +} + +_SOKOL_PRIVATE _sg_wgpu_bindgroup_t* _sg_wgpu_bindgroup_at(uint32_t bg_id) { + SOKOL_ASSERT(SG_INVALID_ID != bg_id); + _sg_wgpu_bindgroups_pool_t* p = &_sg.wgpu.bindgroups_pool; + int slot_index = _sg_slot_index(bg_id); + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < p->pool.size)); + return &p->bindgroups[slot_index]; +} + +_SOKOL_PRIVATE _sg_wgpu_bindgroup_t* _sg_wgpu_lookup_bindgroup(uint32_t bg_id) { + if (SG_INVALID_ID != bg_id) { + _sg_wgpu_bindgroup_t* bg = _sg_wgpu_bindgroup_at(bg_id); + if (bg->slot.id == bg_id) { + return bg; + } + } + return 0; +} + +_SOKOL_PRIVATE _sg_wgpu_bindgroup_handle_t _sg_wgpu_alloc_bindgroup(void) { + _sg_wgpu_bindgroups_pool_t* p = &_sg.wgpu.bindgroups_pool; + _sg_wgpu_bindgroup_handle_t res; + int slot_index = _sg_pool_alloc_index(&p->pool); + if (_SG_INVALID_SLOT_INDEX != slot_index) { + res.id = _sg_slot_alloc(&p->pool, &p->bindgroups[slot_index].slot, slot_index); + } else { + res.id = SG_INVALID_ID; + _SG_ERROR(WGPU_BINDGROUPS_POOL_EXHAUSTED); + } + return res; +} + +_SOKOL_PRIVATE void _sg_wgpu_dealloc_bindgroup(_sg_wgpu_bindgroup_t* bg) { + SOKOL_ASSERT(bg && (bg->slot.state == SG_RESOURCESTATE_ALLOC) && (bg->slot.id != SG_INVALID_ID)); + _sg_wgpu_bindgroups_pool_t* p = &_sg.wgpu.bindgroups_pool; + _sg_pool_free_index(&p->pool, _sg_slot_index(bg->slot.id)); + _sg_slot_reset(&bg->slot); +} + +_SOKOL_PRIVATE void _sg_wgpu_reset_bindgroup_to_alloc_state(_sg_wgpu_bindgroup_t* bg) { + SOKOL_ASSERT(bg); + _sg_slot_t slot = bg->slot; + _sg_clear(bg, sizeof(_sg_wgpu_bindgroup_t)); + bg->slot = slot; + bg->slot.state = SG_RESOURCESTATE_ALLOC; +} + +// MurmurHash64B (see: https://github.com/aappleby/smhasher/blob/61a0530f28277f2e850bfc39600ce61d02b518de/src/MurmurHash2.cpp#L142) +_SOKOL_PRIVATE uint64_t _sg_wgpu_hash(const void* key, int len, uint64_t seed) { + const uint32_t m = 0x5bd1e995; + const int r = 24; + uint32_t h1 = (uint32_t)seed ^ (uint32_t)len; + uint32_t h2 = (uint32_t)(seed >> 32); + const uint32_t * data = (const uint32_t *)key; + while (len >= 8) { + uint32_t k1 = *data++; + k1 *= m; k1 ^= k1 >> r; k1 *= m; + h1 *= m; h1 ^= k1; + len -= 4; + uint32_t k2 = *data++; + k2 *= m; k2 ^= k2 >> r; k2 *= m; + h2 *= m; h2 ^= k2; + len -= 4; + } + if (len >= 4) { + uint32_t k1 = *data++; + k1 *= m; k1 ^= k1 >> r; k1 *= m; + h1 *= m; h1 ^= k1; + len -= 4; + } + switch(len) { + case 3: h2 ^= (uint32_t)(((unsigned char*)data)[2] << 16); + case 2: h2 ^= (uint32_t)(((unsigned char*)data)[1] << 8); + case 1: h2 ^= ((unsigned char*)data)[0]; + h2 *= m; + }; + h1 ^= h2 >> 18; h1 *= m; + h2 ^= h1 >> 22; h2 *= m; + h1 ^= h2 >> 17; h1 *= m; + h2 ^= h1 >> 19; h2 *= m; + uint64_t h = h1; + h = (h << 32) | h2; + return h; +} + +_SOKOL_PRIVATE uint64_t _sg_wgpu_bindgroups_cache_item(_sg_wgpu_bindgroups_cache_item_type_t type, uint8_t wgpu_binding, uint32_t id) { + // key pattern is bbbbttttiiiiiiii + const uint64_t bb = (uint64_t)wgpu_binding; + const uint64_t tttt = (uint64_t)type; + const uint64_t iiiiiiii = (uint64_t)id; + return (bb << 56) | (bb << 48) | (tttt << 32) | iiiiiiii; +} + +_SOKOL_PRIVATE uint64_t _sg_wgpu_bindgroups_cache_pip_item(uint32_t id) { + return _sg_wgpu_bindgroups_cache_item(_SG_WGPU_BINDGROUPSCACHEITEMTYPE_PIPELINE, 0xFF, id); +} + +_SOKOL_PRIVATE uint64_t _sg_wgpu_bindgroups_cache_image_item(uint8_t wgpu_binding, uint32_t id) { + return _sg_wgpu_bindgroups_cache_item(_SG_WGPU_BINDGROUPSCACHEITEMTYPE_IMAGE, wgpu_binding, id); +} + +_SOKOL_PRIVATE uint64_t _sg_wgpu_bindgroups_cache_sampler_item(uint8_t wgpu_binding, uint32_t id) { + return _sg_wgpu_bindgroups_cache_item(_SG_WGPU_BINDGROUPSCACHEITEMTYPE_SAMPLER, wgpu_binding, id); +} + +_SOKOL_PRIVATE uint64_t _sg_wgpu_bindgroups_cache_sbuf_item(uint8_t wgpu_binding, uint32_t id) { + return _sg_wgpu_bindgroups_cache_item(_SG_WGPU_BINDGROUPSCACHEITEMTYPE_STORAGEBUFFER, wgpu_binding, id); +} + +_SOKOL_PRIVATE void _sg_wgpu_init_bindgroups_cache_key(_sg_wgpu_bindgroups_cache_key_t* key, const _sg_bindings_t* bnd) { + SOKOL_ASSERT(bnd); + SOKOL_ASSERT(bnd->pip); + const _sg_shader_t* shd = bnd->pip->shader; + SOKOL_ASSERT(shd && shd->slot.id == bnd->pip->cmn.shader_id.id); + + _sg_clear(key->items, sizeof(key->items)); + key->items[0] = _sg_wgpu_bindgroups_cache_pip_item(bnd->pip->slot.id); + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + if (shd->cmn.images[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(bnd->imgs[i]); + const size_t item_idx = i + 1; + SOKOL_ASSERT(item_idx < _SG_WGPU_BINDGROUPSCACHEKEY_NUM_ITEMS); + SOKOL_ASSERT(0 == key->items[item_idx]); + const uint8_t wgpu_binding = shd->wgpu.img_grp1_bnd_n[i]; + const uint32_t id = bnd->imgs[i]->slot.id; + key->items[item_idx] = _sg_wgpu_bindgroups_cache_image_item(wgpu_binding, id); + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + if (shd->cmn.samplers[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(bnd->smps[i]); + const size_t item_idx = i + 1 + SG_MAX_IMAGE_BINDSLOTS; + SOKOL_ASSERT(item_idx < _SG_WGPU_BINDGROUPSCACHEKEY_NUM_ITEMS); + SOKOL_ASSERT(0 == key->items[item_idx]); + const uint8_t wgpu_binding = shd->wgpu.smp_grp1_bnd_n[i]; + const uint32_t id = bnd->smps[i]->slot.id; + key->items[item_idx] = _sg_wgpu_bindgroups_cache_sampler_item(wgpu_binding, id); + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (shd->cmn.storage_buffers[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(bnd->sbufs[i]); + const size_t item_idx = i + 1 + SG_MAX_IMAGE_BINDSLOTS + SG_MAX_SAMPLER_BINDSLOTS; + SOKOL_ASSERT(item_idx < _SG_WGPU_BINDGROUPSCACHEKEY_NUM_ITEMS); + SOKOL_ASSERT(0 == key->items[item_idx]); + const uint8_t wgpu_binding = shd->wgpu.sbuf_grp1_bnd_n[i]; + const uint32_t id = bnd->sbufs[i]->slot.id; + key->items[item_idx] = _sg_wgpu_bindgroups_cache_sbuf_item(wgpu_binding, id); + } + key->hash = _sg_wgpu_hash(&key->items, (int)sizeof(key->items), 0x1234567887654321); +} + +_SOKOL_PRIVATE bool _sg_wgpu_compare_bindgroups_cache_key(_sg_wgpu_bindgroups_cache_key_t* k0, _sg_wgpu_bindgroups_cache_key_t* k1) { + SOKOL_ASSERT(k0 && k1); + if (k0->hash != k1->hash) { + return false; + } + if (memcmp(&k0->items, &k1->items, sizeof(k0->items)) != 0) { + _sg_stats_add(wgpu.bindings.num_bindgroup_cache_hash_vs_key_mismatch, 1); + return false; + } + return true; +} + +_SOKOL_PRIVATE _sg_wgpu_bindgroup_t* _sg_wgpu_create_bindgroup(_sg_bindings_t* bnd) { + SOKOL_ASSERT(_sg.wgpu.dev); + SOKOL_ASSERT(bnd->pip); + const _sg_shader_t* shd = bnd->pip->shader; + SOKOL_ASSERT(shd && (shd->slot.id == bnd->pip->cmn.shader_id.id)); + _sg_stats_add(wgpu.bindings.num_create_bindgroup, 1); + _sg_wgpu_bindgroup_handle_t bg_id = _sg_wgpu_alloc_bindgroup(); + if (bg_id.id == SG_INVALID_ID) { + return 0; + } + _sg_wgpu_bindgroup_t* bg = _sg_wgpu_bindgroup_at(bg_id.id); + SOKOL_ASSERT(bg && (bg->slot.state == SG_RESOURCESTATE_ALLOC)); + + // create wgpu bindgroup object (also see _sg_wgpu_create_shader()) + WGPUBindGroupLayout bgl = bnd->pip->shader->wgpu.bgl_img_smp_sbuf; + SOKOL_ASSERT(bgl); + WGPUBindGroupEntry bg_entries[_SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES]; + _sg_clear(&bg_entries, sizeof(bg_entries)); + size_t bgl_index = 0; + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + if (shd->cmn.images[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(bnd->imgs[i]); + SOKOL_ASSERT(bgl_index < _SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES); + WGPUBindGroupEntry* bg_entry = &bg_entries[bgl_index]; + bg_entry->binding = shd->wgpu.img_grp1_bnd_n[i]; + bg_entry->textureView = bnd->imgs[i]->wgpu.view; + bgl_index += 1; + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + if (shd->cmn.samplers[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(bnd->smps[i]); + SOKOL_ASSERT(bgl_index < _SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES); + WGPUBindGroupEntry* bg_entry = &bg_entries[bgl_index]; + bg_entry->binding = shd->wgpu.smp_grp1_bnd_n[i]; + bg_entry->sampler = bnd->smps[i]->wgpu.smp; + bgl_index += 1; + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (shd->cmn.storage_buffers[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + SOKOL_ASSERT(bnd->sbufs[i]); + SOKOL_ASSERT(bgl_index < _SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES); + WGPUBindGroupEntry* bg_entry = &bg_entries[bgl_index]; + bg_entry->binding = shd->wgpu.sbuf_grp1_bnd_n[i]; + bg_entry->buffer = bnd->sbufs[i]->wgpu.buf; + bg_entry->size = (uint64_t) bnd->sbufs[i]->cmn.size; + bgl_index += 1; + } + WGPUBindGroupDescriptor bg_desc; + _sg_clear(&bg_desc, sizeof(bg_desc)); + bg_desc.layout = bgl; + bg_desc.entryCount = bgl_index; + bg_desc.entries = bg_entries; + bg->bindgroup = wgpuDeviceCreateBindGroup(_sg.wgpu.dev, &bg_desc); + if (bg->bindgroup == 0) { + _SG_ERROR(WGPU_CREATEBINDGROUP_FAILED); + bg->slot.state = SG_RESOURCESTATE_FAILED; + return bg; + } + _sg_wgpu_init_bindgroups_cache_key(&bg->key, bnd); + bg->slot.state = SG_RESOURCESTATE_VALID; + return bg; +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_bindgroup(_sg_wgpu_bindgroup_t* bg) { + SOKOL_ASSERT(bg); + _sg_stats_add(wgpu.bindings.num_discard_bindgroup, 1); + if (bg->slot.state == SG_RESOURCESTATE_VALID) { + if (bg->bindgroup) { + wgpuBindGroupRelease(bg->bindgroup); + bg->bindgroup = 0; + } + _sg_wgpu_reset_bindgroup_to_alloc_state(bg); + SOKOL_ASSERT(bg->slot.state == SG_RESOURCESTATE_ALLOC); + } + if (bg->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_wgpu_dealloc_bindgroup(bg); + SOKOL_ASSERT(bg->slot.state == SG_RESOURCESTATE_INITIAL); + } +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_all_bindgroups(void) { + _sg_wgpu_bindgroups_pool_t* p = &_sg.wgpu.bindgroups_pool; + for (int i = 0; i < p->pool.size; i++) { + sg_resource_state state = p->bindgroups[i].slot.state; + if ((state == SG_RESOURCESTATE_VALID) || (state == SG_RESOURCESTATE_FAILED)) { + _sg_wgpu_discard_bindgroup(&p->bindgroups[i]); + } + } +} + +_SOKOL_PRIVATE void _sg_wgpu_bindgroups_cache_init(const sg_desc* desc) { + SOKOL_ASSERT(desc); + SOKOL_ASSERT(_sg.wgpu.bindgroups_cache.num == 0); + SOKOL_ASSERT(_sg.wgpu.bindgroups_cache.index_mask == 0); + SOKOL_ASSERT(_sg.wgpu.bindgroups_cache.items == 0); + const int num = desc->wgpu_bindgroups_cache_size; + if (num <= 1) { + _SG_PANIC(WGPU_BINDGROUPSCACHE_SIZE_GREATER_ONE); + } + if (!_sg_ispow2(num)) { + _SG_PANIC(WGPU_BINDGROUPSCACHE_SIZE_POW2); + } + _sg.wgpu.bindgroups_cache.num = (uint32_t)desc->wgpu_bindgroups_cache_size; + _sg.wgpu.bindgroups_cache.index_mask = _sg.wgpu.bindgroups_cache.num - 1; + size_t size_in_bytes = sizeof(_sg_wgpu_bindgroup_handle_t) * (size_t)num; + _sg.wgpu.bindgroups_cache.items = (_sg_wgpu_bindgroup_handle_t*)_sg_malloc_clear(size_in_bytes); +} + +_SOKOL_PRIVATE void _sg_wgpu_bindgroups_cache_discard(void) { + if (_sg.wgpu.bindgroups_cache.items) { + _sg_free(_sg.wgpu.bindgroups_cache.items); + _sg.wgpu.bindgroups_cache.items = 0; + } + _sg.wgpu.bindgroups_cache.num = 0; + _sg.wgpu.bindgroups_cache.index_mask = 0; +} + +_SOKOL_PRIVATE void _sg_wgpu_bindgroups_cache_set(uint64_t hash, uint32_t bg_id) { + uint32_t index = hash & _sg.wgpu.bindgroups_cache.index_mask; + SOKOL_ASSERT(index < _sg.wgpu.bindgroups_cache.num); + SOKOL_ASSERT(_sg.wgpu.bindgroups_cache.items); + _sg.wgpu.bindgroups_cache.items[index].id = bg_id; +} + +_SOKOL_PRIVATE uint32_t _sg_wgpu_bindgroups_cache_get(uint64_t hash) { + uint32_t index = hash & _sg.wgpu.bindgroups_cache.index_mask; + SOKOL_ASSERT(index < _sg.wgpu.bindgroups_cache.num); + SOKOL_ASSERT(_sg.wgpu.bindgroups_cache.items); + return _sg.wgpu.bindgroups_cache.items[index].id; +} + +// called from wgpu resource destroy functions to also invalidate any +// bindgroups cache slot and bindgroup referencing that resource +_SOKOL_PRIVATE void _sg_wgpu_bindgroups_cache_invalidate(_sg_wgpu_bindgroups_cache_item_type_t type, uint32_t id) { + const uint64_t key_mask = 0x0000FFFFFFFFFFFF; + const uint64_t key_item = _sg_wgpu_bindgroups_cache_item(type, 0, id) & key_mask; + SOKOL_ASSERT(_sg.wgpu.bindgroups_cache.items); + for (uint32_t cache_item_idx = 0; cache_item_idx < _sg.wgpu.bindgroups_cache.num; cache_item_idx++) { + const uint32_t bg_id = _sg.wgpu.bindgroups_cache.items[cache_item_idx].id; + if (bg_id != SG_INVALID_ID) { + _sg_wgpu_bindgroup_t* bg = _sg_wgpu_lookup_bindgroup(bg_id); + SOKOL_ASSERT(bg && (bg->slot.state == SG_RESOURCESTATE_VALID)); + // check if resource is in bindgroup, if yes discard bindgroup and invalidate cache slot + bool invalidate_cache_item = false; + for (int key_item_idx = 0; key_item_idx < _SG_WGPU_BINDGROUPSCACHEKEY_NUM_ITEMS; key_item_idx++) { + if ((bg->key.items[key_item_idx] & key_mask) == key_item) { + invalidate_cache_item = true; + break; + } + } + if (invalidate_cache_item) { + _sg_wgpu_discard_bindgroup(bg); bg = 0; + _sg_wgpu_bindgroups_cache_set(cache_item_idx, SG_INVALID_ID); + _sg_stats_add(wgpu.bindings.num_bindgroup_cache_invalidates, 1); + } + } + } +} + +_SOKOL_PRIVATE void _sg_wgpu_bindings_cache_clear(void) { + memset(&_sg.wgpu.bindings_cache, 0, sizeof(_sg.wgpu.bindings_cache)); +} + +_SOKOL_PRIVATE bool _sg_wgpu_bindings_cache_vb_dirty(size_t index, const _sg_buffer_t* vb, uint64_t offset) { + SOKOL_ASSERT((index >= 0) && (index < SG_MAX_VERTEXBUFFER_BINDSLOTS)); + if (vb) { + return (_sg.wgpu.bindings_cache.vbs[index].buffer.id != vb->slot.id) + || (_sg.wgpu.bindings_cache.vbs[index].offset != offset); + } else { + return _sg.wgpu.bindings_cache.vbs[index].buffer.id != SG_INVALID_ID; + } +} + +_SOKOL_PRIVATE void _sg_wgpu_bindings_cache_vb_update(size_t index, const _sg_buffer_t* vb, uint64_t offset) { + SOKOL_ASSERT((index >= 0) && (index < SG_MAX_VERTEXBUFFER_BINDSLOTS)); + if (vb) { + _sg.wgpu.bindings_cache.vbs[index].buffer.id = vb->slot.id; + _sg.wgpu.bindings_cache.vbs[index].offset = offset; + } else { + _sg.wgpu.bindings_cache.vbs[index].buffer.id = SG_INVALID_ID; + _sg.wgpu.bindings_cache.vbs[index].offset = 0; + } +} + +_SOKOL_PRIVATE bool _sg_wgpu_bindings_cache_ib_dirty(const _sg_buffer_t* ib, uint64_t offset) { + if (ib) { + return (_sg.wgpu.bindings_cache.ib.buffer.id != ib->slot.id) + || (_sg.wgpu.bindings_cache.ib.offset != offset); + } else { + return _sg.wgpu.bindings_cache.ib.buffer.id != SG_INVALID_ID; + } +} + +_SOKOL_PRIVATE void _sg_wgpu_bindings_cache_ib_update(const _sg_buffer_t* ib, uint64_t offset) { + if (ib) { + _sg.wgpu.bindings_cache.ib.buffer.id = ib->slot.id; + _sg.wgpu.bindings_cache.ib.offset = offset; + } else { + _sg.wgpu.bindings_cache.ib.buffer.id = SG_INVALID_ID; + _sg.wgpu.bindings_cache.ib.offset = 0; + } +} + +_SOKOL_PRIVATE bool _sg_wgpu_bindings_cache_bg_dirty(const _sg_wgpu_bindgroup_t* bg) { + if (bg) { + return _sg.wgpu.bindings_cache.bg.id != bg->slot.id; + } else { + return _sg.wgpu.bindings_cache.bg.id != SG_INVALID_ID; + } +} + +_SOKOL_PRIVATE void _sg_wgpu_bindings_cache_bg_update(const _sg_wgpu_bindgroup_t* bg) { + if (bg) { + _sg.wgpu.bindings_cache.bg.id = bg->slot.id; + } else { + _sg.wgpu.bindings_cache.bg.id = SG_INVALID_ID; + } +} + +_SOKOL_PRIVATE void _sg_wgpu_set_img_smp_sbuf_bindgroup(_sg_wgpu_bindgroup_t* bg) { + if (_sg_wgpu_bindings_cache_bg_dirty(bg)) { + _sg_wgpu_bindings_cache_bg_update(bg); + _sg_stats_add(wgpu.bindings.num_set_bindgroup, 1); + if (_sg.cur_pass.is_compute) { + SOKOL_ASSERT(_sg.wgpu.cpass_enc); + if (bg) { + SOKOL_ASSERT(bg->slot.state == SG_RESOURCESTATE_VALID); + SOKOL_ASSERT(bg->bindgroup); + wgpuComputePassEncoderSetBindGroup(_sg.wgpu.cpass_enc, _SG_WGPU_IMG_SMP_SBUF_BINDGROUP_INDEX, bg->bindgroup, 0, 0); + } else { + wgpuComputePassEncoderSetBindGroup(_sg.wgpu.cpass_enc, _SG_WGPU_IMG_SMP_SBUF_BINDGROUP_INDEX, _sg.wgpu.empty_bind_group, 0, 0); + } + } else { + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + if (bg) { + SOKOL_ASSERT(bg->slot.state == SG_RESOURCESTATE_VALID); + SOKOL_ASSERT(bg->bindgroup); + wgpuRenderPassEncoderSetBindGroup(_sg.wgpu.rpass_enc, _SG_WGPU_IMG_SMP_SBUF_BINDGROUP_INDEX, bg->bindgroup, 0, 0); + } else { + wgpuRenderPassEncoderSetBindGroup(_sg.wgpu.rpass_enc, _SG_WGPU_IMG_SMP_SBUF_BINDGROUP_INDEX, _sg.wgpu.empty_bind_group, 0, 0); + } + } + } else { + _sg_stats_add(wgpu.bindings.num_skip_redundant_bindgroup, 1); + } +} + +_SOKOL_PRIVATE bool _sg_wgpu_apply_bindgroup(_sg_bindings_t* bnd) { + if (!_sg.desc.wgpu_disable_bindgroups_cache) { + _sg_wgpu_bindgroup_t* bg = 0; + _sg_wgpu_bindgroups_cache_key_t key; + _sg_wgpu_init_bindgroups_cache_key(&key, bnd); + uint32_t bg_id = _sg_wgpu_bindgroups_cache_get(key.hash); + if (bg_id != SG_INVALID_ID) { + // potential cache hit + bg = _sg_wgpu_lookup_bindgroup(bg_id); + SOKOL_ASSERT(bg && (bg->slot.state == SG_RESOURCESTATE_VALID)); + if (!_sg_wgpu_compare_bindgroups_cache_key(&key, &bg->key)) { + // cache collision, need to delete cached bindgroup + _sg_stats_add(wgpu.bindings.num_bindgroup_cache_collisions, 1); + _sg_wgpu_discard_bindgroup(bg); + _sg_wgpu_bindgroups_cache_set(key.hash, SG_INVALID_ID); + bg = 0; + } else { + _sg_stats_add(wgpu.bindings.num_bindgroup_cache_hits, 1); + } + } else { + _sg_stats_add(wgpu.bindings.num_bindgroup_cache_misses, 1); + } + if (bg == 0) { + // either no cache entry yet, or cache collision, create new bindgroup and store in cache + bg = _sg_wgpu_create_bindgroup(bnd); + _sg_wgpu_bindgroups_cache_set(key.hash, bg->slot.id); + } + if (bg && bg->slot.state == SG_RESOURCESTATE_VALID) { + _sg_wgpu_set_img_smp_sbuf_bindgroup(bg); + } else { + return false; + } + } else { + // bindgroups cache disabled, create and destroy bindgroup on the fly (expensive!) + _sg_wgpu_bindgroup_t* bg = _sg_wgpu_create_bindgroup(bnd); + if (bg) { + if (bg->slot.state == SG_RESOURCESTATE_VALID) { + _sg_wgpu_set_img_smp_sbuf_bindgroup(bg); + } + _sg_wgpu_discard_bindgroup(bg); + } else { + return false; + } + } + return true; +} + +_SOKOL_PRIVATE bool _sg_wgpu_apply_index_buffer(_sg_bindings_t* bnd) { + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + const _sg_buffer_t* ib = bnd->ib; + uint64_t offset = (uint64_t)bnd->ib_offset; + if (_sg_wgpu_bindings_cache_ib_dirty(ib, offset)) { + _sg_wgpu_bindings_cache_ib_update(ib, offset); + if (ib) { + const WGPUIndexFormat format = _sg_wgpu_indexformat(bnd->pip->cmn.index_type); + const uint64_t buf_size = (uint64_t)ib->cmn.size; + SOKOL_ASSERT(buf_size > offset); + const uint64_t max_bytes = buf_size - offset; + wgpuRenderPassEncoderSetIndexBuffer(_sg.wgpu.rpass_enc, ib->wgpu.buf, format, offset, max_bytes); + /* FIXME: the else-pass should actually set a null index buffer, but that doesn't seem to work yet + } else { + wgpuRenderPassEncoderSetIndexBuffer(_sg.wgpu.rpass_enc, 0, WGPUIndexFormat_Undefined, 0, 0); + */ + } + _sg_stats_add(wgpu.bindings.num_set_index_buffer, 1); + } else { + _sg_stats_add(wgpu.bindings.num_skip_redundant_index_buffer, 1); + } + return true; +} + +_SOKOL_PRIVATE bool _sg_wgpu_apply_vertex_buffers(_sg_bindings_t* bnd) { + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + for (uint32_t slot = 0; slot < SG_MAX_VERTEXBUFFER_BINDSLOTS; slot++) { + const _sg_buffer_t* vb = bnd->vbs[slot]; + const uint64_t offset = (uint64_t)bnd->vb_offsets[slot]; + if (_sg_wgpu_bindings_cache_vb_dirty(slot, vb, offset)) { + _sg_wgpu_bindings_cache_vb_update(slot, vb, offset); + if (vb) { + const uint64_t buf_size = (uint64_t)vb->cmn.size; + SOKOL_ASSERT(buf_size > offset); + const uint64_t max_bytes = buf_size - offset; + wgpuRenderPassEncoderSetVertexBuffer(_sg.wgpu.rpass_enc, slot, vb->wgpu.buf, offset, max_bytes); + /* FIXME: the else-pass should actually set a null vertex buffer, but that doesn't seem to work yet + } else { + wgpuRenderPassEncoderSetVertexBuffer(_sg.wgpu.rpass_enc, slot, 0, 0, 0); + */ + } + _sg_stats_add(wgpu.bindings.num_set_vertex_buffer, 1); + } else { + _sg_stats_add(wgpu.bindings.num_skip_redundant_vertex_buffer, 1); + } + } + return true; +} + +_SOKOL_PRIVATE void _sg_wgpu_setup_backend(const sg_desc* desc) { + SOKOL_ASSERT(desc); + SOKOL_ASSERT(desc->environment.wgpu.device); + SOKOL_ASSERT(desc->uniform_buffer_size > 0); + _sg.backend = SG_BACKEND_WGPU; + _sg.wgpu.valid = true; + _sg.wgpu.dev = (WGPUDevice) desc->environment.wgpu.device; + _sg.wgpu.queue = wgpuDeviceGetQueue(_sg.wgpu.dev); + SOKOL_ASSERT(_sg.wgpu.queue); + + _sg_wgpu_init_caps(); + _sg_wgpu_uniform_buffer_init(desc); + _sg_wgpu_bindgroups_pool_init(desc); + _sg_wgpu_bindgroups_cache_init(desc); + _sg_wgpu_bindings_cache_clear(); + + // create an empty bind group + WGPUBindGroupLayoutDescriptor bgl_desc; + _sg_clear(&bgl_desc, sizeof(bgl_desc)); + WGPUBindGroupLayout empty_bgl = wgpuDeviceCreateBindGroupLayout(_sg.wgpu.dev, &bgl_desc); + SOKOL_ASSERT(empty_bgl); + WGPUBindGroupDescriptor bg_desc; + _sg_clear(&bg_desc, sizeof(bg_desc)); + bg_desc.layout = empty_bgl; + _sg.wgpu.empty_bind_group = wgpuDeviceCreateBindGroup(_sg.wgpu.dev, &bg_desc); + SOKOL_ASSERT(_sg.wgpu.empty_bind_group); + wgpuBindGroupLayoutRelease(empty_bgl); + + // create initial per-frame command encoder + WGPUCommandEncoderDescriptor cmd_enc_desc; + _sg_clear(&cmd_enc_desc, sizeof(cmd_enc_desc)); + _sg.wgpu.cmd_enc = wgpuDeviceCreateCommandEncoder(_sg.wgpu.dev, &cmd_enc_desc); + SOKOL_ASSERT(_sg.wgpu.cmd_enc); +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_backend(void) { + SOKOL_ASSERT(_sg.wgpu.valid); + SOKOL_ASSERT(_sg.wgpu.cmd_enc); + _sg.wgpu.valid = false; + _sg_wgpu_discard_all_bindgroups(); + _sg_wgpu_bindgroups_cache_discard(); + _sg_wgpu_bindgroups_pool_discard(); + _sg_wgpu_uniform_buffer_discard(); + wgpuBindGroupRelease(_sg.wgpu.empty_bind_group); _sg.wgpu.empty_bind_group = 0; + wgpuCommandEncoderRelease(_sg.wgpu.cmd_enc); _sg.wgpu.cmd_enc = 0; + wgpuQueueRelease(_sg.wgpu.queue); _sg.wgpu.queue = 0; +} + +_SOKOL_PRIVATE void _sg_wgpu_reset_state_cache(void) { + _sg_wgpu_bindings_cache_clear(); +} + +_SOKOL_PRIVATE sg_resource_state _sg_wgpu_create_buffer(_sg_buffer_t* buf, const sg_buffer_desc* desc) { + SOKOL_ASSERT(buf && desc); + SOKOL_ASSERT(buf->cmn.size > 0); + const bool injected = (0 != desc->wgpu_buffer); + if (injected) { + buf->wgpu.buf = (WGPUBuffer) desc->wgpu_buffer; + wgpuBufferReference(buf->wgpu.buf); + } else { + // buffer mapping size must be multiple of 4, so round up buffer size (only a problem + // with index buffers containing odd number of indices) + const uint64_t wgpu_buf_size = _sg_roundup_u64((uint64_t)buf->cmn.size, 4); + const bool map_at_creation = (SG_USAGE_IMMUTABLE == buf->cmn.usage) && (desc->data.ptr); + + WGPUBufferDescriptor wgpu_buf_desc; + _sg_clear(&wgpu_buf_desc, sizeof(wgpu_buf_desc)); + wgpu_buf_desc.usage = _sg_wgpu_buffer_usage(buf->cmn.type, buf->cmn.usage); + wgpu_buf_desc.size = wgpu_buf_size; + wgpu_buf_desc.mappedAtCreation = map_at_creation; + wgpu_buf_desc.label = _sg_wgpu_stringview(desc->label); + buf->wgpu.buf = wgpuDeviceCreateBuffer(_sg.wgpu.dev, &wgpu_buf_desc); + if (0 == buf->wgpu.buf) { + _SG_ERROR(WGPU_CREATE_BUFFER_FAILED); + return SG_RESOURCESTATE_FAILED; + } + // NOTE: assume that WebGPU creates zero-initialized buffers + if (map_at_creation) { + SOKOL_ASSERT(desc->data.ptr && (desc->data.size > 0)); + SOKOL_ASSERT(desc->data.size <= (size_t)buf->cmn.size); + // FIXME: inefficient on WASM + void* ptr = wgpuBufferGetMappedRange(buf->wgpu.buf, 0, wgpu_buf_size); + SOKOL_ASSERT(ptr); + memcpy(ptr, desc->data.ptr, desc->data.size); + wgpuBufferUnmap(buf->wgpu.buf); + } + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_buffer(_sg_buffer_t* buf) { + SOKOL_ASSERT(buf); + if (buf->cmn.type == SG_BUFFERTYPE_STORAGEBUFFER) { + _sg_wgpu_bindgroups_cache_invalidate(_SG_WGPU_BINDGROUPSCACHEITEMTYPE_STORAGEBUFFER, buf->slot.id); + } + if (buf->wgpu.buf) { + wgpuBufferRelease(buf->wgpu.buf); + } +} + +_SOKOL_PRIVATE void _sg_wgpu_copy_buffer_data(const _sg_buffer_t* buf, uint64_t offset, const sg_range* data) { + SOKOL_ASSERT((offset + data->size) <= (size_t)buf->cmn.size); + // WebGPU's write-buffer requires the size to be a multiple of four, so we may need to split the copy + // operation into two writeBuffer calls + uint64_t clamped_size = data->size & ~3UL; + uint64_t extra_size = data->size & 3UL; + SOKOL_ASSERT(extra_size < 4); + wgpuQueueWriteBuffer(_sg.wgpu.queue, buf->wgpu.buf, offset, data->ptr, clamped_size); + if (extra_size > 0) { + const uint64_t extra_src_offset = clamped_size; + const uint64_t extra_dst_offset = offset + clamped_size; + uint8_t extra_data[4] = { 0 }; + uint8_t* extra_src_ptr = ((uint8_t*)data->ptr) + extra_src_offset; + for (size_t i = 0; i < extra_size; i++) { + extra_data[i] = extra_src_ptr[i]; + } + wgpuQueueWriteBuffer(_sg.wgpu.queue, buf->wgpu.buf, extra_dst_offset, extra_src_ptr, 4); + } +} + +_SOKOL_PRIVATE void _sg_wgpu_copy_image_data(const _sg_image_t* img, WGPUTexture wgpu_tex, const sg_image_data* data) { + WGPUTextureDataLayout wgpu_layout; + _sg_clear(&wgpu_layout, sizeof(wgpu_layout)); + WGPUImageCopyTexture wgpu_copy_tex; + _sg_clear(&wgpu_copy_tex, sizeof(wgpu_copy_tex)); + wgpu_copy_tex.texture = wgpu_tex; + wgpu_copy_tex.aspect = WGPUTextureAspect_All; + WGPUExtent3D wgpu_extent; + _sg_clear(&wgpu_extent, sizeof(wgpu_extent)); + const int num_faces = (img->cmn.type == SG_IMAGETYPE_CUBE) ? 6 : 1; + for (int face_index = 0; face_index < num_faces; face_index++) { + for (int mip_index = 0; mip_index < img->cmn.num_mipmaps; mip_index++) { + wgpu_copy_tex.mipLevel = (uint32_t)mip_index; + wgpu_copy_tex.origin.z = (uint32_t)face_index; + int mip_width = _sg_miplevel_dim(img->cmn.width, mip_index); + int mip_height = _sg_miplevel_dim(img->cmn.height, mip_index); + int mip_slices; + switch (img->cmn.type) { + case SG_IMAGETYPE_CUBE: + mip_slices = 1; + break; + case SG_IMAGETYPE_3D: + mip_slices = _sg_miplevel_dim(img->cmn.num_slices, mip_index); + break; + default: + mip_slices = img->cmn.num_slices; + break; + } + const int row_pitch = _sg_row_pitch(img->cmn.pixel_format, mip_width, 1); + const int num_rows = _sg_num_rows(img->cmn.pixel_format, mip_height); + if (_sg_is_compressed_pixel_format(img->cmn.pixel_format)) { + mip_width = _sg_roundup(mip_width, 4); + mip_height = _sg_roundup(mip_height, 4); + } + wgpu_layout.offset = 0; + wgpu_layout.bytesPerRow = (uint32_t)row_pitch; + wgpu_layout.rowsPerImage = (uint32_t)num_rows; + wgpu_extent.width = (uint32_t)mip_width; + wgpu_extent.height = (uint32_t)mip_height; + wgpu_extent.depthOrArrayLayers = (uint32_t)mip_slices; + const sg_range* mip_data = &data->subimage[face_index][mip_index]; + wgpuQueueWriteTexture(_sg.wgpu.queue, &wgpu_copy_tex, mip_data->ptr, mip_data->size, &wgpu_layout, &wgpu_extent); + } + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_wgpu_create_image(_sg_image_t* img, const sg_image_desc* desc) { + SOKOL_ASSERT(img && desc); + const bool injected = (0 != desc->wgpu_texture); + if (injected) { + img->wgpu.tex = (WGPUTexture)desc->wgpu_texture; + wgpuTextureReference(img->wgpu.tex); + img->wgpu.view = (WGPUTextureView)desc->wgpu_texture_view; + if (img->wgpu.view) { + wgpuTextureViewReference(img->wgpu.view); + } + } else { + WGPUTextureDescriptor wgpu_tex_desc; + _sg_clear(&wgpu_tex_desc, sizeof(wgpu_tex_desc)); + wgpu_tex_desc.label = _sg_wgpu_stringview(desc->label); + wgpu_tex_desc.usage = WGPUTextureUsage_TextureBinding|WGPUTextureUsage_CopyDst; + if (desc->render_target) { + wgpu_tex_desc.usage |= WGPUTextureUsage_RenderAttachment; + } + wgpu_tex_desc.dimension = _sg_wgpu_texture_dimension(img->cmn.type); + wgpu_tex_desc.size.width = (uint32_t) img->cmn.width; + wgpu_tex_desc.size.height = (uint32_t) img->cmn.height; + if (desc->type == SG_IMAGETYPE_CUBE) { + wgpu_tex_desc.size.depthOrArrayLayers = 6; + } else { + wgpu_tex_desc.size.depthOrArrayLayers = (uint32_t) img->cmn.num_slices; + } + wgpu_tex_desc.format = _sg_wgpu_textureformat(img->cmn.pixel_format); + wgpu_tex_desc.mipLevelCount = (uint32_t) img->cmn.num_mipmaps; + wgpu_tex_desc.sampleCount = (uint32_t) img->cmn.sample_count; + img->wgpu.tex = wgpuDeviceCreateTexture(_sg.wgpu.dev, &wgpu_tex_desc); + if (0 == img->wgpu.tex) { + _SG_ERROR(WGPU_CREATE_TEXTURE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + if ((img->cmn.usage == SG_USAGE_IMMUTABLE) && !img->cmn.render_target) { + _sg_wgpu_copy_image_data(img, img->wgpu.tex, &desc->data); + } + WGPUTextureViewDescriptor wgpu_texview_desc; + _sg_clear(&wgpu_texview_desc, sizeof(wgpu_texview_desc)); + wgpu_texview_desc.label = _sg_wgpu_stringview(desc->label); + wgpu_texview_desc.dimension = _sg_wgpu_texture_view_dimension(img->cmn.type); + wgpu_texview_desc.mipLevelCount = (uint32_t)img->cmn.num_mipmaps; + if (img->cmn.type == SG_IMAGETYPE_CUBE) { + wgpu_texview_desc.arrayLayerCount = 6; + } else if (img->cmn.type == SG_IMAGETYPE_ARRAY) { + wgpu_texview_desc.arrayLayerCount = (uint32_t)img->cmn.num_slices; + } else { + wgpu_texview_desc.arrayLayerCount = 1; + } + if (_sg_is_depth_or_depth_stencil_format(img->cmn.pixel_format)) { + wgpu_texview_desc.aspect = WGPUTextureAspect_DepthOnly; + } else { + wgpu_texview_desc.aspect = WGPUTextureAspect_All; + } + img->wgpu.view = wgpuTextureCreateView(img->wgpu.tex, &wgpu_texview_desc); + if (0 == img->wgpu.view) { + _SG_ERROR(WGPU_CREATE_TEXTURE_VIEW_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_image(_sg_image_t* img) { + SOKOL_ASSERT(img); + _sg_wgpu_bindgroups_cache_invalidate(_SG_WGPU_BINDGROUPSCACHEITEMTYPE_IMAGE, img->slot.id); + if (img->wgpu.view) { + wgpuTextureViewRelease(img->wgpu.view); + img->wgpu.view = 0; + } + if (img->wgpu.tex) { + wgpuTextureRelease(img->wgpu.tex); + img->wgpu.tex = 0; + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_wgpu_create_sampler(_sg_sampler_t* smp, const sg_sampler_desc* desc) { + SOKOL_ASSERT(smp && desc); + SOKOL_ASSERT(_sg.wgpu.dev); + const bool injected = (0 != desc->wgpu_sampler); + if (injected) { + smp->wgpu.smp = (WGPUSampler) desc->wgpu_sampler; + wgpuSamplerReference(smp->wgpu.smp); + } else { + WGPUSamplerDescriptor wgpu_desc; + _sg_clear(&wgpu_desc, sizeof(wgpu_desc)); + wgpu_desc.label = _sg_wgpu_stringview(desc->label); + wgpu_desc.addressModeU = _sg_wgpu_sampler_address_mode(desc->wrap_u); + wgpu_desc.addressModeV = _sg_wgpu_sampler_address_mode(desc->wrap_v); + wgpu_desc.addressModeW = _sg_wgpu_sampler_address_mode(desc->wrap_w); + wgpu_desc.magFilter = _sg_wgpu_sampler_minmag_filter(desc->mag_filter); + wgpu_desc.minFilter = _sg_wgpu_sampler_minmag_filter(desc->min_filter); + wgpu_desc.mipmapFilter = _sg_wgpu_sampler_mipmap_filter(desc->mipmap_filter); + wgpu_desc.lodMinClamp = desc->min_lod; + wgpu_desc.lodMaxClamp = desc->max_lod; + wgpu_desc.compare = _sg_wgpu_comparefunc(desc->compare); + if (wgpu_desc.compare == WGPUCompareFunction_Never) { + wgpu_desc.compare = WGPUCompareFunction_Undefined; + } + wgpu_desc.maxAnisotropy = (uint16_t)desc->max_anisotropy; + smp->wgpu.smp = wgpuDeviceCreateSampler(_sg.wgpu.dev, &wgpu_desc); + if (0 == smp->wgpu.smp) { + _SG_ERROR(WGPU_CREATE_SAMPLER_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_sampler(_sg_sampler_t* smp) { + SOKOL_ASSERT(smp); + _sg_wgpu_bindgroups_cache_invalidate(_SG_WGPU_BINDGROUPSCACHEITEMTYPE_SAMPLER, smp->slot.id); + if (smp->wgpu.smp) { + wgpuSamplerRelease(smp->wgpu.smp); + smp->wgpu.smp = 0; + } +} + +_SOKOL_PRIVATE _sg_wgpu_shader_func_t _sg_wgpu_create_shader_func(const sg_shader_function* func, const char* label) { + SOKOL_ASSERT(func); + SOKOL_ASSERT(func->source); + SOKOL_ASSERT(func->entry); + + _sg_wgpu_shader_func_t res; + _sg_clear(&res, sizeof(res)); + _sg_strcpy(&res.entry, func->entry); + + WGPUShaderModuleWGSLDescriptor wgpu_shdmod_wgsl_desc; + _sg_clear(&wgpu_shdmod_wgsl_desc, sizeof(wgpu_shdmod_wgsl_desc)); + wgpu_shdmod_wgsl_desc.chain.sType = WGPUSType_ShaderModuleWGSLDescriptor; + wgpu_shdmod_wgsl_desc.code = _sg_wgpu_stringview(func->source); + + WGPUShaderModuleDescriptor wgpu_shdmod_desc; + _sg_clear(&wgpu_shdmod_desc, sizeof(wgpu_shdmod_desc)); + wgpu_shdmod_desc.nextInChain = &wgpu_shdmod_wgsl_desc.chain; + wgpu_shdmod_desc.label = _sg_wgpu_stringview(label); + + // NOTE: if compilation fails we won't actually find out in this call since + // it always returns a valid module handle, and the GetCompilationInfo() call + // is asynchronous + res.module = wgpuDeviceCreateShaderModule(_sg.wgpu.dev, &wgpu_shdmod_desc); + if (0 == res.module) { + _SG_ERROR(WGPU_CREATE_SHADER_MODULE_FAILED); + } + return res; +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_shader_func(_sg_wgpu_shader_func_t* func) { + if (func->module) { + wgpuShaderModuleRelease(func->module); + func->module = 0; + } +} + +typedef struct { uint8_t sokol_slot, wgpu_slot; } _sg_wgpu_dynoffset_mapping_t; + +_SOKOL_PRIVATE int _sg_wgpu_dynoffset_cmp(const void* a, const void* b) { + const _sg_wgpu_dynoffset_mapping_t* aa = (const _sg_wgpu_dynoffset_mapping_t*)a; + const _sg_wgpu_dynoffset_mapping_t* bb = (const _sg_wgpu_dynoffset_mapping_t*)b; + if (aa->wgpu_slot < bb->wgpu_slot) return -1; + else if (aa->wgpu_slot > bb->wgpu_slot) return 1; + return 0; +} + +// NOTE: this is an out-of-range check for WGSL bindslots that's also active in release mode +_SOKOL_PRIVATE bool _sg_wgpu_ensure_wgsl_bindslot_ranges(const sg_shader_desc* desc) { + SOKOL_ASSERT(desc); + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + if (desc->uniform_blocks[i].wgsl_group0_binding_n >= _SG_WGPU_MAX_UB_BINDGROUP_BIND_SLOTS) { + _SG_ERROR(WGPU_UNIFORMBLOCK_WGSL_GROUP0_BINDING_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (desc->storage_buffers[i].wgsl_group1_binding_n >= _SG_WGPU_MAX_IMG_SMP_SBUF_BIND_SLOTS) { + _SG_ERROR(WGPU_STORAGEBUFFER_WGSL_GROUP1_BINDING_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + if (desc->images[i].wgsl_group1_binding_n >= _SG_WGPU_MAX_IMG_SMP_SBUF_BIND_SLOTS) { + _SG_ERROR(WGPU_IMAGE_WGSL_GROUP1_BINDING_OUT_OF_RANGE); + return false; + } + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + if (desc->samplers[i].wgsl_group1_binding_n >= _SG_WGPU_MAX_IMG_SMP_SBUF_BIND_SLOTS) { + _SG_ERROR(WGPU_SAMPLER_WGSL_GROUP1_BINDING_OUT_OF_RANGE); + return false; + } + } + return true; +} + +_SOKOL_PRIVATE sg_resource_state _sg_wgpu_create_shader(_sg_shader_t* shd, const sg_shader_desc* desc) { + SOKOL_ASSERT(shd && desc); + SOKOL_ASSERT(shd->wgpu.vertex_func.module == 0); + SOKOL_ASSERT(shd->wgpu.fragment_func.module == 0); + SOKOL_ASSERT(shd->wgpu.compute_func.module == 0); + SOKOL_ASSERT(shd->wgpu.bgl_ub == 0); + SOKOL_ASSERT(shd->wgpu.bg_ub == 0); + SOKOL_ASSERT(shd->wgpu.bgl_img_smp_sbuf == 0); + + // do a release-mode bounds-check on wgsl bindslots, even though out-of-range + // bindslots can't cause out-of-bounds accesses in the wgpu backend, this + // is done to be consistent with the other backends + if (!_sg_wgpu_ensure_wgsl_bindslot_ranges(desc)) { + return SG_RESOURCESTATE_FAILED; + } + + // build shader modules + bool shd_valid = true; + if (desc->vertex_func.source) { + shd->wgpu.vertex_func = _sg_wgpu_create_shader_func(&desc->vertex_func, desc->label); + shd_valid &= shd->wgpu.vertex_func.module != 0; + } + if (desc->fragment_func.source) { + shd->wgpu.fragment_func = _sg_wgpu_create_shader_func(&desc->fragment_func, desc->label); + shd_valid &= shd->wgpu.fragment_func.module != 0; + } + if (desc->compute_func.source) { + shd->wgpu.compute_func = _sg_wgpu_create_shader_func(&desc->compute_func, desc->label); + shd_valid &= shd->wgpu.compute_func.module != 0; + } + if (!shd_valid) { + _sg_wgpu_discard_shader_func(&shd->wgpu.vertex_func); + _sg_wgpu_discard_shader_func(&shd->wgpu.fragment_func); + _sg_wgpu_discard_shader_func(&shd->wgpu.compute_func); + return SG_RESOURCESTATE_FAILED; + } + + // create bind group layout and bind group for uniform blocks + // NOTE also need to create a mapping of sokol ub bind slots to array indices + // for the dynamic offsets array in the setBindGroup call + SOKOL_ASSERT(_SG_WGPU_MAX_UB_BINDGROUP_ENTRIES <= _SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES); + WGPUBindGroupLayoutEntry bgl_entries[_SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES]; + _sg_clear(bgl_entries, sizeof(bgl_entries)); + WGPUBindGroupLayoutDescriptor bgl_desc; + _sg_clear(&bgl_desc, sizeof(bgl_desc)); + WGPUBindGroupEntry bg_entries[_SG_WGPU_MAX_IMG_SMP_SBUF_BINDGROUP_ENTRIES]; + _sg_clear(&bg_entries, sizeof(bg_entries)); + WGPUBindGroupDescriptor bg_desc; + _sg_clear(&bg_desc, sizeof(bg_desc)); + _sg_wgpu_dynoffset_mapping_t dynoffset_map[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + _sg_clear(dynoffset_map, sizeof(dynoffset_map)); + size_t bgl_index = 0; + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + if (shd->cmn.uniform_blocks[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + shd->wgpu.ub_grp0_bnd_n[i] = desc->uniform_blocks[i].wgsl_group0_binding_n; + WGPUBindGroupEntry* bg_entry = &bg_entries[bgl_index]; + WGPUBindGroupLayoutEntry* bgl_entry = &bgl_entries[bgl_index]; + bgl_entry->binding = shd->wgpu.ub_grp0_bnd_n[i]; + bgl_entry->visibility = _sg_wgpu_shader_stage(shd->cmn.uniform_blocks[i].stage); + bgl_entry->buffer.type = WGPUBufferBindingType_Uniform; + bgl_entry->buffer.hasDynamicOffset = true; + bg_entry->binding = bgl_entry->binding; + bg_entry->buffer = _sg.wgpu.uniform.buf; + bg_entry->size = _SG_WGPU_MAX_UNIFORM_UPDATE_SIZE; + dynoffset_map[i].sokol_slot = i; + dynoffset_map[i].wgpu_slot = bgl_entry->binding; + bgl_index += 1; + } + bgl_desc.entryCount = bgl_index; + bgl_desc.entries = bgl_entries; + shd->wgpu.bgl_ub = wgpuDeviceCreateBindGroupLayout(_sg.wgpu.dev, &bgl_desc); + SOKOL_ASSERT(shd->wgpu.bgl_ub); + bg_desc.layout = shd->wgpu.bgl_ub; + bg_desc.entryCount = bgl_index; + bg_desc.entries = bg_entries; + shd->wgpu.bg_ub = wgpuDeviceCreateBindGroup(_sg.wgpu.dev, &bg_desc); + SOKOL_ASSERT(shd->wgpu.bg_ub); + + // sort the dynoffset_map by wgpu bindings, this is because the + // dynamic offsets of the WebGPU setBindGroup call must be in + // 'binding order', not 'bindgroup entry order' + qsort(dynoffset_map, bgl_index, sizeof(_sg_wgpu_dynoffset_mapping_t), _sg_wgpu_dynoffset_cmp); + shd->wgpu.ub_num_dynoffsets = bgl_index; + for (uint8_t i = 0; i < bgl_index; i++) { + const uint8_t sokol_slot = dynoffset_map[i].sokol_slot; + shd->wgpu.ub_dynoffsets[sokol_slot] = i; + } + + // create bind group layout for images, samplers and storage buffers + _sg_clear(bgl_entries, sizeof(bgl_entries)); + _sg_clear(&bgl_desc, sizeof(bgl_desc)); + bgl_index = 0; + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + if (shd->cmn.images[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + const bool msaa = shd->cmn.images[i].multisampled; + shd->wgpu.img_grp1_bnd_n[i] = desc->images[i].wgsl_group1_binding_n; + WGPUBindGroupLayoutEntry* bgl_entry = &bgl_entries[bgl_index]; + bgl_entry->binding = shd->wgpu.img_grp1_bnd_n[i]; + bgl_entry->visibility = _sg_wgpu_shader_stage(shd->cmn.images[i].stage); + bgl_entry->texture.viewDimension = _sg_wgpu_texture_view_dimension(shd->cmn.images[i].image_type); + bgl_entry->texture.sampleType = _sg_wgpu_texture_sample_type(shd->cmn.images[i].sample_type, msaa); + bgl_entry->texture.multisampled = msaa; + bgl_index += 1; + } + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + if (shd->cmn.samplers[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + shd->wgpu.smp_grp1_bnd_n[i] = desc->samplers[i].wgsl_group1_binding_n; + WGPUBindGroupLayoutEntry* bgl_entry = &bgl_entries[bgl_index]; + bgl_entry->binding = shd->wgpu.smp_grp1_bnd_n[i]; + bgl_entry->visibility = _sg_wgpu_shader_stage(shd->cmn.samplers[i].stage); + bgl_entry->sampler.type = _sg_wgpu_sampler_binding_type(shd->cmn.samplers[i].sampler_type); + bgl_index += 1; + } + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (shd->cmn.storage_buffers[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + shd->wgpu.sbuf_grp1_bnd_n[i] = desc->storage_buffers[i].wgsl_group1_binding_n; + WGPUBindGroupLayoutEntry* bgl_entry = &bgl_entries[bgl_index]; + bgl_entry->binding = shd->wgpu.sbuf_grp1_bnd_n[i]; + bgl_entry->visibility = _sg_wgpu_shader_stage(shd->cmn.storage_buffers[i].stage); + if (shd->cmn.storage_buffers[i].readonly) { + bgl_entry->buffer.type = WGPUBufferBindingType_ReadOnlyStorage; + } else { + bgl_entry->buffer.type = WGPUBufferBindingType_Storage; + } + bgl_index += 1; + } + bgl_desc.entryCount = bgl_index; + bgl_desc.entries = bgl_entries; + shd->wgpu.bgl_img_smp_sbuf = wgpuDeviceCreateBindGroupLayout(_sg.wgpu.dev, &bgl_desc); + if (shd->wgpu.bgl_img_smp_sbuf == 0) { + _SG_ERROR(WGPU_SHADER_CREATE_BINDGROUP_LAYOUT_FAILED); + return SG_RESOURCESTATE_FAILED; + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_shader(_sg_shader_t* shd) { + SOKOL_ASSERT(shd); + _sg_wgpu_discard_shader_func(&shd->wgpu.vertex_func); + _sg_wgpu_discard_shader_func(&shd->wgpu.fragment_func); + _sg_wgpu_discard_shader_func(&shd->wgpu.compute_func); + if (shd->wgpu.bgl_ub) { + wgpuBindGroupLayoutRelease(shd->wgpu.bgl_ub); + shd->wgpu.bgl_ub = 0; + } + if (shd->wgpu.bg_ub) { + wgpuBindGroupRelease(shd->wgpu.bg_ub); + shd->wgpu.bg_ub = 0; + } + if (shd->wgpu.bgl_img_smp_sbuf) { + wgpuBindGroupLayoutRelease(shd->wgpu.bgl_img_smp_sbuf); + shd->wgpu.bgl_img_smp_sbuf = 0; + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_wgpu_create_pipeline(_sg_pipeline_t* pip, _sg_shader_t* shd, const sg_pipeline_desc* desc) { + SOKOL_ASSERT(pip && shd && desc); + SOKOL_ASSERT(desc->shader.id == shd->slot.id); + SOKOL_ASSERT(shd->wgpu.bgl_ub); + SOKOL_ASSERT(shd->wgpu.bgl_img_smp_sbuf); + pip->shader = shd; + + pip->wgpu.blend_color.r = (double) desc->blend_color.r; + pip->wgpu.blend_color.g = (double) desc->blend_color.g; + pip->wgpu.blend_color.b = (double) desc->blend_color.b; + pip->wgpu.blend_color.a = (double) desc->blend_color.a; + + // - @group(0) for uniform blocks + // - @group(1) for all image, sampler and storagebuffer resources + WGPUBindGroupLayout wgpu_bgl[_SG_WGPU_NUM_BINDGROUPS]; + _sg_clear(&wgpu_bgl, sizeof(wgpu_bgl)); + wgpu_bgl[_SG_WGPU_UB_BINDGROUP_INDEX ] = shd->wgpu.bgl_ub; + wgpu_bgl[_SG_WGPU_IMG_SMP_SBUF_BINDGROUP_INDEX] = shd->wgpu.bgl_img_smp_sbuf; + WGPUPipelineLayoutDescriptor wgpu_pl_desc; + _sg_clear(&wgpu_pl_desc, sizeof(wgpu_pl_desc)); + wgpu_pl_desc.bindGroupLayoutCount = _SG_WGPU_NUM_BINDGROUPS; + wgpu_pl_desc.bindGroupLayouts = &wgpu_bgl[0]; + const WGPUPipelineLayout wgpu_pip_layout = wgpuDeviceCreatePipelineLayout(_sg.wgpu.dev, &wgpu_pl_desc); + if (0 == wgpu_pip_layout) { + _SG_ERROR(WGPU_CREATE_PIPELINE_LAYOUT_FAILED); + return SG_RESOURCESTATE_FAILED; + } + SOKOL_ASSERT(wgpu_pip_layout); + + if (pip->cmn.is_compute) { + WGPUComputePipelineDescriptor wgpu_pip_desc; + _sg_clear(&wgpu_pip_desc, sizeof(wgpu_pip_desc)); + wgpu_pip_desc.label = _sg_wgpu_stringview(desc->label); + wgpu_pip_desc.layout = wgpu_pip_layout; + wgpu_pip_desc.compute.module = shd->wgpu.compute_func.module; + wgpu_pip_desc.compute.entryPoint = shd->wgpu.compute_func.entry.buf; + pip->wgpu.cpip = wgpuDeviceCreateComputePipeline(_sg.wgpu.dev, &wgpu_pip_desc); + wgpuPipelineLayoutRelease(wgpu_pip_layout); + if (0 == pip->wgpu.cpip) { + _SG_ERROR(WGPU_CREATE_COMPUTE_PIPELINE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } else { + WGPUVertexBufferLayout wgpu_vb_layouts[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + _sg_clear(wgpu_vb_layouts, sizeof(wgpu_vb_layouts)); + WGPUVertexAttribute wgpu_vtx_attrs[SG_MAX_VERTEXBUFFER_BINDSLOTS][SG_MAX_VERTEX_ATTRIBUTES]; + _sg_clear(wgpu_vtx_attrs, sizeof(wgpu_vtx_attrs)); + int wgpu_vb_num = 0; + for (int vb_idx = 0; vb_idx < SG_MAX_VERTEXBUFFER_BINDSLOTS; vb_idx++, wgpu_vb_num++) { + const sg_vertex_buffer_layout_state* vbl_state = &desc->layout.buffers[vb_idx]; + if (0 == vbl_state->stride) { + break; + } + wgpu_vb_layouts[vb_idx].arrayStride = (uint64_t)vbl_state->stride; + wgpu_vb_layouts[vb_idx].stepMode = _sg_wgpu_stepmode(vbl_state->step_func); + wgpu_vb_layouts[vb_idx].attributes = &wgpu_vtx_attrs[vb_idx][0]; + } + for (int va_idx = 0; va_idx < SG_MAX_VERTEX_ATTRIBUTES; va_idx++) { + const sg_vertex_attr_state* va_state = &desc->layout.attrs[va_idx]; + if (SG_VERTEXFORMAT_INVALID == va_state->format) { + break; + } + const int vb_idx = va_state->buffer_index; + SOKOL_ASSERT(vb_idx < SG_MAX_VERTEXBUFFER_BINDSLOTS); + SOKOL_ASSERT(pip->cmn.vertex_buffer_layout_active[vb_idx]); + const size_t wgpu_attr_idx = wgpu_vb_layouts[vb_idx].attributeCount; + wgpu_vb_layouts[vb_idx].attributeCount += 1; + wgpu_vtx_attrs[vb_idx][wgpu_attr_idx].format = _sg_wgpu_vertexformat(va_state->format); + wgpu_vtx_attrs[vb_idx][wgpu_attr_idx].offset = (uint64_t)va_state->offset; + wgpu_vtx_attrs[vb_idx][wgpu_attr_idx].shaderLocation = (uint32_t)va_idx; + } + + WGPURenderPipelineDescriptor wgpu_pip_desc; + _sg_clear(&wgpu_pip_desc, sizeof(wgpu_pip_desc)); + WGPUDepthStencilState wgpu_ds_state; + _sg_clear(&wgpu_ds_state, sizeof(wgpu_ds_state)); + WGPUFragmentState wgpu_frag_state; + _sg_clear(&wgpu_frag_state, sizeof(wgpu_frag_state)); + WGPUColorTargetState wgpu_ctgt_state[SG_MAX_COLOR_ATTACHMENTS]; + _sg_clear(&wgpu_ctgt_state, sizeof(wgpu_ctgt_state)); + WGPUBlendState wgpu_blend_state[SG_MAX_COLOR_ATTACHMENTS]; + _sg_clear(&wgpu_blend_state, sizeof(wgpu_blend_state)); + wgpu_pip_desc.label = _sg_wgpu_stringview(desc->label); + wgpu_pip_desc.layout = wgpu_pip_layout; + wgpu_pip_desc.vertex.module = shd->wgpu.vertex_func.module; + wgpu_pip_desc.vertex.entryPoint = shd->wgpu.vertex_func.entry.buf; + wgpu_pip_desc.vertex.bufferCount = (size_t)wgpu_vb_num; + wgpu_pip_desc.vertex.buffers = &wgpu_vb_layouts[0]; + wgpu_pip_desc.primitive.topology = _sg_wgpu_topology(desc->primitive_type); + wgpu_pip_desc.primitive.stripIndexFormat = _sg_wgpu_stripindexformat(desc->primitive_type, desc->index_type); + wgpu_pip_desc.primitive.frontFace = _sg_wgpu_frontface(desc->face_winding); + wgpu_pip_desc.primitive.cullMode = _sg_wgpu_cullmode(desc->cull_mode); + if (SG_PIXELFORMAT_NONE != desc->depth.pixel_format) { + wgpu_ds_state.format = _sg_wgpu_textureformat(desc->depth.pixel_format); + wgpu_ds_state.depthWriteEnabled = _sg_wgpu_optional_bool(desc->depth.write_enabled); + wgpu_ds_state.depthCompare = _sg_wgpu_comparefunc(desc->depth.compare); + wgpu_ds_state.stencilFront.compare = _sg_wgpu_comparefunc(desc->stencil.front.compare); + wgpu_ds_state.stencilFront.failOp = _sg_wgpu_stencilop(desc->stencil.front.fail_op); + wgpu_ds_state.stencilFront.depthFailOp = _sg_wgpu_stencilop(desc->stencil.front.depth_fail_op); + wgpu_ds_state.stencilFront.passOp = _sg_wgpu_stencilop(desc->stencil.front.pass_op); + wgpu_ds_state.stencilBack.compare = _sg_wgpu_comparefunc(desc->stencil.back.compare); + wgpu_ds_state.stencilBack.failOp = _sg_wgpu_stencilop(desc->stencil.back.fail_op); + wgpu_ds_state.stencilBack.depthFailOp = _sg_wgpu_stencilop(desc->stencil.back.depth_fail_op); + wgpu_ds_state.stencilBack.passOp = _sg_wgpu_stencilop(desc->stencil.back.pass_op); + wgpu_ds_state.stencilReadMask = desc->stencil.read_mask; + wgpu_ds_state.stencilWriteMask = desc->stencil.write_mask; + wgpu_ds_state.depthBias = (int32_t)desc->depth.bias; + wgpu_ds_state.depthBiasSlopeScale = desc->depth.bias_slope_scale; + wgpu_ds_state.depthBiasClamp = desc->depth.bias_clamp; + wgpu_pip_desc.depthStencil = &wgpu_ds_state; + } + wgpu_pip_desc.multisample.count = (uint32_t)desc->sample_count; + wgpu_pip_desc.multisample.mask = 0xFFFFFFFF; + wgpu_pip_desc.multisample.alphaToCoverageEnabled = desc->alpha_to_coverage_enabled; + if (desc->color_count > 0) { + wgpu_frag_state.module = shd->wgpu.fragment_func.module; + wgpu_frag_state.entryPoint = shd->wgpu.fragment_func.entry.buf; + wgpu_frag_state.targetCount = (size_t)desc->color_count; + wgpu_frag_state.targets = &wgpu_ctgt_state[0]; + for (int i = 0; i < desc->color_count; i++) { + SOKOL_ASSERT(i < SG_MAX_COLOR_ATTACHMENTS); + wgpu_ctgt_state[i].format = _sg_wgpu_textureformat(desc->colors[i].pixel_format); + wgpu_ctgt_state[i].writeMask = _sg_wgpu_colorwritemask(desc->colors[i].write_mask); + if (desc->colors[i].blend.enabled) { + wgpu_ctgt_state[i].blend = &wgpu_blend_state[i]; + wgpu_blend_state[i].color.operation = _sg_wgpu_blendop(desc->colors[i].blend.op_rgb); + wgpu_blend_state[i].color.srcFactor = _sg_wgpu_blendfactor(desc->colors[i].blend.src_factor_rgb); + wgpu_blend_state[i].color.dstFactor = _sg_wgpu_blendfactor(desc->colors[i].blend.dst_factor_rgb); + wgpu_blend_state[i].alpha.operation = _sg_wgpu_blendop(desc->colors[i].blend.op_alpha); + wgpu_blend_state[i].alpha.srcFactor = _sg_wgpu_blendfactor(desc->colors[i].blend.src_factor_alpha); + wgpu_blend_state[i].alpha.dstFactor = _sg_wgpu_blendfactor(desc->colors[i].blend.dst_factor_alpha); + } + } + wgpu_pip_desc.fragment = &wgpu_frag_state; + } + pip->wgpu.rpip = wgpuDeviceCreateRenderPipeline(_sg.wgpu.dev, &wgpu_pip_desc); + wgpuPipelineLayoutRelease(wgpu_pip_layout); + if (0 == pip->wgpu.rpip) { + _SG_ERROR(WGPU_CREATE_RENDER_PIPELINE_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + _sg_wgpu_bindgroups_cache_invalidate(_SG_WGPU_BINDGROUPSCACHEITEMTYPE_PIPELINE, pip->slot.id); + if (pip == _sg.wgpu.cur_pipeline) { + _sg.wgpu.cur_pipeline = 0; + _sg.wgpu.cur_pipeline_id.id = SG_INVALID_ID; + } + if (pip->wgpu.rpip) { + wgpuRenderPipelineRelease(pip->wgpu.rpip); + pip->wgpu.rpip = 0; + } + if (pip->wgpu.cpip) { + wgpuComputePipelineRelease(pip->wgpu.cpip); + pip->wgpu.cpip = 0; + } +} + +_SOKOL_PRIVATE sg_resource_state _sg_wgpu_create_attachments(_sg_attachments_t* atts, _sg_image_t** color_images, _sg_image_t** resolve_images, _sg_image_t* ds_img, const sg_attachments_desc* desc) { + SOKOL_ASSERT(atts && desc); + SOKOL_ASSERT(color_images && resolve_images); + + // copy image pointers and create renderable wgpu texture views + for (int i = 0; i < atts->cmn.num_colors; i++) { + const sg_attachment_desc* color_desc = &desc->colors[i]; + _SOKOL_UNUSED(color_desc); + SOKOL_ASSERT(color_desc->image.id != SG_INVALID_ID); + SOKOL_ASSERT(0 == atts->wgpu.colors[i].image); + SOKOL_ASSERT(color_images[i] && (color_images[i]->slot.id == color_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_color_format(color_images[i]->cmn.pixel_format)); + SOKOL_ASSERT(color_images[i]->wgpu.tex); + atts->wgpu.colors[i].image = color_images[i]; + + WGPUTextureViewDescriptor wgpu_color_view_desc; + _sg_clear(&wgpu_color_view_desc, sizeof(wgpu_color_view_desc)); + wgpu_color_view_desc.baseMipLevel = (uint32_t) color_desc->mip_level; + wgpu_color_view_desc.mipLevelCount = 1; + wgpu_color_view_desc.baseArrayLayer = (uint32_t) color_desc->slice; + wgpu_color_view_desc.arrayLayerCount = 1; + atts->wgpu.colors[i].view = wgpuTextureCreateView(color_images[i]->wgpu.tex, &wgpu_color_view_desc); + if (0 == atts->wgpu.colors[i].view) { + _SG_ERROR(WGPU_ATTACHMENTS_CREATE_TEXTURE_VIEW_FAILED); + return SG_RESOURCESTATE_FAILED; + } + + const sg_attachment_desc* resolve_desc = &desc->resolves[i]; + if (resolve_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(0 == atts->wgpu.resolves[i].image); + SOKOL_ASSERT(resolve_images[i] && (resolve_images[i]->slot.id == resolve_desc->image.id)); + SOKOL_ASSERT(color_images[i] && (color_images[i]->cmn.pixel_format == resolve_images[i]->cmn.pixel_format)); + SOKOL_ASSERT(resolve_images[i]->wgpu.tex); + atts->wgpu.resolves[i].image = resolve_images[i]; + + WGPUTextureViewDescriptor wgpu_resolve_view_desc; + _sg_clear(&wgpu_resolve_view_desc, sizeof(wgpu_resolve_view_desc)); + wgpu_resolve_view_desc.baseMipLevel = (uint32_t) resolve_desc->mip_level; + wgpu_resolve_view_desc.mipLevelCount = 1; + wgpu_resolve_view_desc.baseArrayLayer = (uint32_t) resolve_desc->slice; + wgpu_resolve_view_desc.arrayLayerCount = 1; + atts->wgpu.resolves[i].view = wgpuTextureCreateView(resolve_images[i]->wgpu.tex, &wgpu_resolve_view_desc); + if (0 == atts->wgpu.resolves[i].view) { + _SG_ERROR(WGPU_ATTACHMENTS_CREATE_TEXTURE_VIEW_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } + } + SOKOL_ASSERT(0 == atts->wgpu.depth_stencil.image); + const sg_attachment_desc* ds_desc = &desc->depth_stencil; + if (ds_desc->image.id != SG_INVALID_ID) { + SOKOL_ASSERT(ds_img && (ds_img->slot.id == ds_desc->image.id)); + SOKOL_ASSERT(_sg_is_valid_rendertarget_depth_format(ds_img->cmn.pixel_format)); + SOKOL_ASSERT(ds_img->wgpu.tex); + atts->wgpu.depth_stencil.image = ds_img; + + WGPUTextureViewDescriptor wgpu_ds_view_desc; + _sg_clear(&wgpu_ds_view_desc, sizeof(wgpu_ds_view_desc)); + wgpu_ds_view_desc.baseMipLevel = (uint32_t) ds_desc->mip_level; + wgpu_ds_view_desc.mipLevelCount = 1; + wgpu_ds_view_desc.baseArrayLayer = (uint32_t) ds_desc->slice; + wgpu_ds_view_desc.arrayLayerCount = 1; + atts->wgpu.depth_stencil.view = wgpuTextureCreateView(ds_img->wgpu.tex, &wgpu_ds_view_desc); + if (0 == atts->wgpu.depth_stencil.view) { + _SG_ERROR(WGPU_ATTACHMENTS_CREATE_TEXTURE_VIEW_FAILED); + return SG_RESOURCESTATE_FAILED; + } + } + return SG_RESOURCESTATE_VALID; +} + +_SOKOL_PRIVATE void _sg_wgpu_discard_attachments(_sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + for (int i = 0; i < atts->cmn.num_colors; i++) { + if (atts->wgpu.colors[i].view) { + wgpuTextureViewRelease(atts->wgpu.colors[i].view); + atts->wgpu.colors[i].view = 0; + } + if (atts->wgpu.resolves[i].view) { + wgpuTextureViewRelease(atts->wgpu.resolves[i].view); + atts->wgpu.resolves[i].view = 0; + } + } + if (atts->wgpu.depth_stencil.view) { + wgpuTextureViewRelease(atts->wgpu.depth_stencil.view); + atts->wgpu.depth_stencil.view = 0; + } +} + +_SOKOL_PRIVATE _sg_image_t* _sg_wgpu_attachments_color_image(const _sg_attachments_t* atts, int index) { + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + // NOTE: may return null + return atts->wgpu.colors[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_wgpu_attachments_resolve_image(const _sg_attachments_t* atts, int index) { + SOKOL_ASSERT(atts && (index >= 0) && (index < SG_MAX_COLOR_ATTACHMENTS)); + // NOTE: may return null + return atts->wgpu.resolves[index].image; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_wgpu_attachments_ds_image(const _sg_attachments_t* atts) { + // NOTE: may return null + SOKOL_ASSERT(atts); + return atts->wgpu.depth_stencil.image; +} + +_SOKOL_PRIVATE void _sg_wgpu_init_color_att(WGPURenderPassColorAttachment* wgpu_att, const sg_color_attachment_action* action, WGPUTextureView color_view, WGPUTextureView resolve_view) { + wgpu_att->depthSlice = WGPU_DEPTH_SLICE_UNDEFINED; + wgpu_att->view = color_view; + wgpu_att->resolveTarget = resolve_view; + wgpu_att->loadOp = _sg_wgpu_load_op(color_view, action->load_action); + wgpu_att->storeOp = _sg_wgpu_store_op(color_view, action->store_action); + wgpu_att->clearValue.r = action->clear_value.r; + wgpu_att->clearValue.g = action->clear_value.g; + wgpu_att->clearValue.b = action->clear_value.b; + wgpu_att->clearValue.a = action->clear_value.a; +} + +_SOKOL_PRIVATE void _sg_wgpu_init_ds_att(WGPURenderPassDepthStencilAttachment* wgpu_att, const sg_pass_action* action, sg_pixel_format fmt, WGPUTextureView view) { + wgpu_att->view = view; + wgpu_att->depthLoadOp = _sg_wgpu_load_op(view, action->depth.load_action); + wgpu_att->depthStoreOp = _sg_wgpu_store_op(view, action->depth.store_action); + wgpu_att->depthClearValue = action->depth.clear_value; + wgpu_att->depthReadOnly = false; + if (_sg_is_depth_stencil_format(fmt)) { + wgpu_att->stencilLoadOp = _sg_wgpu_load_op(view, action->stencil.load_action); + wgpu_att->stencilStoreOp = _sg_wgpu_store_op(view, action->stencil.store_action); + } else { + wgpu_att->stencilLoadOp = WGPULoadOp_Undefined; + wgpu_att->stencilStoreOp = WGPUStoreOp_Undefined; + } + wgpu_att->stencilClearValue = action->stencil.clear_value; + wgpu_att->stencilReadOnly = false; +} + +_SOKOL_PRIVATE void _sg_wgpu_begin_compute_pass(const sg_pass* pass) { + WGPUComputePassDescriptor wgpu_pass_desc; + _sg_clear(&wgpu_pass_desc, sizeof(wgpu_pass_desc)); + wgpu_pass_desc.label = _sg_wgpu_stringview(pass->label); + _sg.wgpu.cpass_enc = wgpuCommandEncoderBeginComputePass(_sg.wgpu.cmd_enc, &wgpu_pass_desc); + SOKOL_ASSERT(_sg.wgpu.cpass_enc); + // clear initial bindings + wgpuComputePassEncoderSetBindGroup(_sg.wgpu.cpass_enc, _SG_WGPU_UB_BINDGROUP_INDEX, _sg.wgpu.empty_bind_group, 0, 0); + wgpuComputePassEncoderSetBindGroup(_sg.wgpu.cpass_enc, _SG_WGPU_IMG_SMP_SBUF_BINDGROUP_INDEX, _sg.wgpu.empty_bind_group, 0, 0); + _sg_stats_add(wgpu.bindings.num_set_bindgroup, 1); +} + +_SOKOL_PRIVATE void _sg_wgpu_begin_render_pass(const sg_pass* pass) { + const _sg_attachments_t* atts = _sg.cur_pass.atts; + const sg_swapchain* swapchain = &pass->swapchain; + const sg_pass_action* action = &pass->action; + + WGPURenderPassDescriptor wgpu_pass_desc; + WGPURenderPassColorAttachment wgpu_color_att[SG_MAX_COLOR_ATTACHMENTS]; + WGPURenderPassDepthStencilAttachment wgpu_ds_att; + _sg_clear(&wgpu_pass_desc, sizeof(wgpu_pass_desc)); + _sg_clear(&wgpu_color_att, sizeof(wgpu_color_att)); + _sg_clear(&wgpu_ds_att, sizeof(wgpu_ds_att)); + wgpu_pass_desc.label = _sg_wgpu_stringview(pass->label); + if (atts) { + SOKOL_ASSERT(atts->slot.state == SG_RESOURCESTATE_VALID); + for (int i = 0; i < atts->cmn.num_colors; i++) { + _sg_wgpu_init_color_att(&wgpu_color_att[i], &action->colors[i], atts->wgpu.colors[i].view, atts->wgpu.resolves[i].view); + } + wgpu_pass_desc.colorAttachmentCount = (size_t)atts->cmn.num_colors; + wgpu_pass_desc.colorAttachments = &wgpu_color_att[0]; + if (atts->wgpu.depth_stencil.image) { + _sg_wgpu_init_ds_att(&wgpu_ds_att, action, atts->wgpu.depth_stencil.image->cmn.pixel_format, atts->wgpu.depth_stencil.view); + wgpu_pass_desc.depthStencilAttachment = &wgpu_ds_att; + } + } else { + WGPUTextureView wgpu_color_view = (WGPUTextureView) swapchain->wgpu.render_view; + WGPUTextureView wgpu_resolve_view = (WGPUTextureView) swapchain->wgpu.resolve_view; + WGPUTextureView wgpu_depth_stencil_view = (WGPUTextureView) swapchain->wgpu.depth_stencil_view; + _sg_wgpu_init_color_att(&wgpu_color_att[0], &action->colors[0], wgpu_color_view, wgpu_resolve_view); + wgpu_pass_desc.colorAttachmentCount = 1; + wgpu_pass_desc.colorAttachments = &wgpu_color_att[0]; + if (wgpu_depth_stencil_view) { + SOKOL_ASSERT(swapchain->depth_format > SG_PIXELFORMAT_NONE); + _sg_wgpu_init_ds_att(&wgpu_ds_att, action, swapchain->depth_format, wgpu_depth_stencil_view); + wgpu_pass_desc.depthStencilAttachment = &wgpu_ds_att; + } + } + _sg.wgpu.rpass_enc = wgpuCommandEncoderBeginRenderPass(_sg.wgpu.cmd_enc, &wgpu_pass_desc); + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + + wgpuRenderPassEncoderSetBindGroup(_sg.wgpu.rpass_enc, _SG_WGPU_UB_BINDGROUP_INDEX, _sg.wgpu.empty_bind_group, 0, 0); + wgpuRenderPassEncoderSetBindGroup(_sg.wgpu.rpass_enc, _SG_WGPU_IMG_SMP_SBUF_BINDGROUP_INDEX, _sg.wgpu.empty_bind_group, 0, 0); + _sg_stats_add(wgpu.bindings.num_set_bindgroup, 1); +} + +_SOKOL_PRIVATE void _sg_wgpu_begin_pass(const sg_pass* pass) { + SOKOL_ASSERT(pass); + SOKOL_ASSERT(_sg.wgpu.dev); + SOKOL_ASSERT(_sg.wgpu.cmd_enc); + SOKOL_ASSERT(0 == _sg.wgpu.rpass_enc); + SOKOL_ASSERT(0 == _sg.wgpu.cpass_enc); + + _sg.wgpu.cur_pipeline = 0; + _sg.wgpu.cur_pipeline_id.id = SG_INVALID_ID; + _sg_wgpu_bindings_cache_clear(); + + if (pass->compute) { + _sg_wgpu_begin_compute_pass(pass); + } else { + _sg_wgpu_begin_render_pass(pass); + } +} + +_SOKOL_PRIVATE void _sg_wgpu_end_pass(void) { + if (_sg.wgpu.rpass_enc) { + wgpuRenderPassEncoderEnd(_sg.wgpu.rpass_enc); + wgpuRenderPassEncoderRelease(_sg.wgpu.rpass_enc); + _sg.wgpu.rpass_enc = 0; + } + if (_sg.wgpu.cpass_enc) { + wgpuComputePassEncoderEnd(_sg.wgpu.cpass_enc); + wgpuComputePassEncoderRelease(_sg.wgpu.cpass_enc); + _sg.wgpu.cpass_enc = 0; + } +} + +_SOKOL_PRIVATE void _sg_wgpu_commit(void) { + SOKOL_ASSERT(_sg.wgpu.cmd_enc); + + _sg_wgpu_uniform_buffer_on_commit(); + + WGPUCommandBufferDescriptor cmd_buf_desc; + _sg_clear(&cmd_buf_desc, sizeof(cmd_buf_desc)); + WGPUCommandBuffer wgpu_cmd_buf = wgpuCommandEncoderFinish(_sg.wgpu.cmd_enc, &cmd_buf_desc); + SOKOL_ASSERT(wgpu_cmd_buf); + wgpuCommandEncoderRelease(_sg.wgpu.cmd_enc); + _sg.wgpu.cmd_enc = 0; + + wgpuQueueSubmit(_sg.wgpu.queue, 1, &wgpu_cmd_buf); + wgpuCommandBufferRelease(wgpu_cmd_buf); + + // create a new render-command-encoder for next frame + WGPUCommandEncoderDescriptor cmd_enc_desc; + _sg_clear(&cmd_enc_desc, sizeof(cmd_enc_desc)); + _sg.wgpu.cmd_enc = wgpuDeviceCreateCommandEncoder(_sg.wgpu.dev, &cmd_enc_desc); +} + +_SOKOL_PRIVATE void _sg_wgpu_apply_viewport(int x, int y, int w, int h, bool origin_top_left) { + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + // FIXME FIXME FIXME: CLIPPING THE VIEWPORT HERE IS WRONG!!! + // (but currently required because WebGPU insists that the viewport rectangle must be + // fully contained inside the framebuffer, but this doesn't make any sense, and also + // isn't required by the backend APIs) + const _sg_recti_t clip = _sg_clipi(x, y, w, h, _sg.cur_pass.width, _sg.cur_pass.height); + float xf = (float) clip.x; + float yf = (float) (origin_top_left ? clip.y : (_sg.cur_pass.height - (clip.y + clip.h))); + float wf = (float) clip.w; + float hf = (float) clip.h; + wgpuRenderPassEncoderSetViewport(_sg.wgpu.rpass_enc, xf, yf, wf, hf, 0.0f, 1.0f); +} + +_SOKOL_PRIVATE void _sg_wgpu_apply_scissor_rect(int x, int y, int w, int h, bool origin_top_left) { + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + const _sg_recti_t clip = _sg_clipi(x, y, w, h, _sg.cur_pass.width, _sg.cur_pass.height); + uint32_t sx = (uint32_t) clip.x; + uint32_t sy = (uint32_t) (origin_top_left ? clip.y : (_sg.cur_pass.height - (clip.y + clip.h))); + uint32_t sw = (uint32_t) clip.w; + uint32_t sh = (uint32_t) clip.h; + wgpuRenderPassEncoderSetScissorRect(_sg.wgpu.rpass_enc, sx, sy, sw, sh); +} + +_SOKOL_PRIVATE void _sg_wgpu_set_ub_bindgroup(const _sg_shader_t* shd) { + // NOTE: dynamic offsets must be in binding order, not in BindGroupEntry order + SOKOL_ASSERT(shd->wgpu.ub_num_dynoffsets < SG_MAX_UNIFORMBLOCK_BINDSLOTS); + uint32_t dyn_offsets[SG_MAX_UNIFORMBLOCK_BINDSLOTS]; + _sg_clear(dyn_offsets, sizeof(dyn_offsets)); + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + if (shd->cmn.uniform_blocks[i].stage == SG_SHADERSTAGE_NONE) { + continue; + } + uint8_t dynoffset_index = shd->wgpu.ub_dynoffsets[i]; + SOKOL_ASSERT(dynoffset_index < shd->wgpu.ub_num_dynoffsets); + dyn_offsets[dynoffset_index] = _sg.wgpu.uniform.bind_offsets[i]; + } + if (_sg.cur_pass.is_compute) { + SOKOL_ASSERT(_sg.wgpu.cpass_enc); + wgpuComputePassEncoderSetBindGroup(_sg.wgpu.cpass_enc, + _SG_WGPU_UB_BINDGROUP_INDEX, + shd->wgpu.bg_ub, + shd->wgpu.ub_num_dynoffsets, + dyn_offsets); + } else { + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + wgpuRenderPassEncoderSetBindGroup(_sg.wgpu.rpass_enc, + _SG_WGPU_UB_BINDGROUP_INDEX, + shd->wgpu.bg_ub, + shd->wgpu.ub_num_dynoffsets, + dyn_offsets); + } +} + +_SOKOL_PRIVATE void _sg_wgpu_apply_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + SOKOL_ASSERT(pip->shader && (pip->shader->slot.id == pip->cmn.shader_id.id)); + _sg.wgpu.cur_pipeline = pip; + _sg.wgpu.cur_pipeline_id.id = pip->slot.id; + if (pip->cmn.is_compute) { + SOKOL_ASSERT(_sg.cur_pass.is_compute); + SOKOL_ASSERT(pip->wgpu.cpip); + SOKOL_ASSERT(_sg.wgpu.cpass_enc); + wgpuComputePassEncoderSetPipeline(_sg.wgpu.cpass_enc, pip->wgpu.cpip); + } else { + SOKOL_ASSERT(!_sg.cur_pass.is_compute); + SOKOL_ASSERT(pip->wgpu.rpip); + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + _sg.wgpu.use_indexed_draw = (pip->cmn.index_type != SG_INDEXTYPE_NONE); + wgpuRenderPassEncoderSetPipeline(_sg.wgpu.rpass_enc, pip->wgpu.rpip); + wgpuRenderPassEncoderSetBlendConstant(_sg.wgpu.rpass_enc, &pip->wgpu.blend_color); + wgpuRenderPassEncoderSetStencilReference(_sg.wgpu.rpass_enc, pip->cmn.stencil.ref); + } + // bind groups must be set because pipelines without uniform blocks or resource bindings + // will still create 'empty' BindGroupLayouts + _sg_wgpu_set_ub_bindgroup(pip->shader); + _sg_wgpu_set_img_smp_sbuf_bindgroup(0); // this will set the 'empty bind group' +} + +_SOKOL_PRIVATE bool _sg_wgpu_apply_bindings(_sg_bindings_t* bnd) { + SOKOL_ASSERT(bnd); + SOKOL_ASSERT(bnd->pip->shader && (bnd->pip->cmn.shader_id.id == bnd->pip->shader->slot.id)); + bool retval = true; + if (!_sg.cur_pass.is_compute) { + retval &= _sg_wgpu_apply_index_buffer(bnd); + retval &= _sg_wgpu_apply_vertex_buffers(bnd); + } + retval &= _sg_wgpu_apply_bindgroup(bnd); + return retval; +} + +_SOKOL_PRIVATE void _sg_wgpu_apply_uniforms(int ub_slot, const sg_range* data) { + const uint32_t alignment = _sg.wgpu.limits.limits.minUniformBufferOffsetAlignment; + SOKOL_ASSERT(_sg.wgpu.uniform.staging); + SOKOL_ASSERT((ub_slot >= 0) && (ub_slot < SG_MAX_UNIFORMBLOCK_BINDSLOTS)); + SOKOL_ASSERT((_sg.wgpu.uniform.offset + data->size) <= _sg.wgpu.uniform.num_bytes); + SOKOL_ASSERT((_sg.wgpu.uniform.offset & (alignment - 1)) == 0); + const _sg_pipeline_t* pip = _sg.wgpu.cur_pipeline; + SOKOL_ASSERT(pip && pip->shader); + SOKOL_ASSERT(pip->slot.id == _sg.wgpu.cur_pipeline_id.id); + const _sg_shader_t* shd = pip->shader; + SOKOL_ASSERT(shd->slot.id == pip->cmn.shader_id.id); + SOKOL_ASSERT(data->size == shd->cmn.uniform_blocks[ub_slot].size); + SOKOL_ASSERT(data->size <= _SG_WGPU_MAX_UNIFORM_UPDATE_SIZE); + + _sg_stats_add(wgpu.uniforms.num_set_bindgroup, 1); + memcpy(_sg.wgpu.uniform.staging + _sg.wgpu.uniform.offset, data->ptr, data->size); + _sg.wgpu.uniform.bind_offsets[ub_slot] = _sg.wgpu.uniform.offset; + _sg.wgpu.uniform.offset = _sg_roundup_u32(_sg.wgpu.uniform.offset + (uint32_t)data->size, alignment); + + _sg_wgpu_set_ub_bindgroup(shd); +} + +_SOKOL_PRIVATE void _sg_wgpu_draw(int base_element, int num_elements, int num_instances) { + SOKOL_ASSERT(_sg.wgpu.rpass_enc); + SOKOL_ASSERT(_sg.wgpu.cur_pipeline && (_sg.wgpu.cur_pipeline->slot.id == _sg.wgpu.cur_pipeline_id.id)); + if (SG_INDEXTYPE_NONE != _sg.wgpu.cur_pipeline->cmn.index_type) { + wgpuRenderPassEncoderDrawIndexed(_sg.wgpu.rpass_enc, (uint32_t)num_elements, (uint32_t)num_instances, (uint32_t)base_element, 0, 0); + } else { + wgpuRenderPassEncoderDraw(_sg.wgpu.rpass_enc, (uint32_t)num_elements, (uint32_t)num_instances, (uint32_t)base_element, 0); + } +} + +_SOKOL_PRIVATE void _sg_wgpu_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) { + SOKOL_ASSERT(_sg.wgpu.cpass_enc); + wgpuComputePassEncoderDispatchWorkgroups(_sg.wgpu.cpass_enc, + (uint32_t)num_groups_x, + (uint32_t)num_groups_y, + (uint32_t)num_groups_z); +} + +_SOKOL_PRIVATE void _sg_wgpu_update_buffer(_sg_buffer_t* buf, const sg_range* data) { + SOKOL_ASSERT(data && data->ptr && (data->size > 0)); + SOKOL_ASSERT(buf); + _sg_wgpu_copy_buffer_data(buf, 0, data); +} + +_SOKOL_PRIVATE void _sg_wgpu_append_buffer(_sg_buffer_t* buf, const sg_range* data, bool new_frame) { + SOKOL_ASSERT(data && data->ptr && (data->size > 0)); + _SOKOL_UNUSED(new_frame); + _sg_wgpu_copy_buffer_data(buf, (uint64_t)buf->cmn.append_pos, data); +} + +_SOKOL_PRIVATE void _sg_wgpu_update_image(_sg_image_t* img, const sg_image_data* data) { + SOKOL_ASSERT(img && data); + _sg_wgpu_copy_image_data(img, img->wgpu.tex, data); +} +#endif + +// ██████ ███████ ███ ██ ███████ ██████ ██ ██████ ██████ █████ ██████ ██ ██ ███████ ███ ██ ██████ +// ██ ██ ████ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ████ ██ ██ ██ +// ██ ███ █████ ██ ██ ██ █████ ██████ ██ ██ ██████ ███████ ██ █████ █████ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██████ ███████ ██ ████ ███████ ██ ██ ██ ██████ ██████ ██ ██ ██████ ██ ██ ███████ ██ ████ ██████ +// +// >>generic backend +static inline void _sg_setup_backend(const sg_desc* desc) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_setup_backend(desc); + #elif defined(SOKOL_METAL) + _sg_mtl_setup_backend(desc); + #elif defined(SOKOL_D3D11) + _sg_d3d11_setup_backend(desc); + #elif defined(SOKOL_WGPU) + _sg_wgpu_setup_backend(desc); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_setup_backend(desc); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_discard_backend(void) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_discard_backend(); + #elif defined(SOKOL_METAL) + _sg_mtl_discard_backend(); + #elif defined(SOKOL_D3D11) + _sg_d3d11_discard_backend(); + #elif defined(SOKOL_WGPU) + _sg_wgpu_discard_backend(); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_discard_backend(); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_reset_state_cache(void) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_reset_state_cache(); + #elif defined(SOKOL_METAL) + _sg_mtl_reset_state_cache(); + #elif defined(SOKOL_D3D11) + _sg_d3d11_reset_state_cache(); + #elif defined(SOKOL_WGPU) + _sg_wgpu_reset_state_cache(); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_reset_state_cache(); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline sg_resource_state _sg_create_buffer(_sg_buffer_t* buf, const sg_buffer_desc* desc) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_create_buffer(buf, desc); + #elif defined(SOKOL_METAL) + return _sg_mtl_create_buffer(buf, desc); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_create_buffer(buf, desc); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_create_buffer(buf, desc); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_create_buffer(buf, desc); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_discard_buffer(_sg_buffer_t* buf) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_discard_buffer(buf); + #elif defined(SOKOL_METAL) + _sg_mtl_discard_buffer(buf); + #elif defined(SOKOL_D3D11) + _sg_d3d11_discard_buffer(buf); + #elif defined(SOKOL_WGPU) + _sg_wgpu_discard_buffer(buf); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_discard_buffer(buf); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline sg_resource_state _sg_create_image(_sg_image_t* img, const sg_image_desc* desc) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_create_image(img, desc); + #elif defined(SOKOL_METAL) + return _sg_mtl_create_image(img, desc); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_create_image(img, desc); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_create_image(img, desc); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_create_image(img, desc); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_discard_image(_sg_image_t* img) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_discard_image(img); + #elif defined(SOKOL_METAL) + _sg_mtl_discard_image(img); + #elif defined(SOKOL_D3D11) + _sg_d3d11_discard_image(img); + #elif defined(SOKOL_WGPU) + _sg_wgpu_discard_image(img); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_discard_image(img); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline sg_resource_state _sg_create_sampler(_sg_sampler_t* smp, const sg_sampler_desc* desc) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_create_sampler(smp, desc); + #elif defined(SOKOL_METAL) + return _sg_mtl_create_sampler(smp, desc); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_create_sampler(smp, desc); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_create_sampler(smp, desc); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_create_sampler(smp, desc); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_discard_sampler(_sg_sampler_t* smp) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_discard_sampler(smp); + #elif defined(SOKOL_METAL) + _sg_mtl_discard_sampler(smp); + #elif defined(SOKOL_D3D11) + _sg_d3d11_discard_sampler(smp); + #elif defined(SOKOL_WGPU) + _sg_wgpu_discard_sampler(smp); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_discard_sampler(smp); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline sg_resource_state _sg_create_shader(_sg_shader_t* shd, const sg_shader_desc* desc) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_create_shader(shd, desc); + #elif defined(SOKOL_METAL) + return _sg_mtl_create_shader(shd, desc); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_create_shader(shd, desc); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_create_shader(shd, desc); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_create_shader(shd, desc); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_discard_shader(_sg_shader_t* shd) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_discard_shader(shd); + #elif defined(SOKOL_METAL) + _sg_mtl_discard_shader(shd); + #elif defined(SOKOL_D3D11) + _sg_d3d11_discard_shader(shd); + #elif defined(SOKOL_WGPU) + _sg_wgpu_discard_shader(shd); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_discard_shader(shd); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline sg_resource_state _sg_create_pipeline(_sg_pipeline_t* pip, _sg_shader_t* shd, const sg_pipeline_desc* desc) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_create_pipeline(pip, shd, desc); + #elif defined(SOKOL_METAL) + return _sg_mtl_create_pipeline(pip, shd, desc); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_create_pipeline(pip, shd, desc); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_create_pipeline(pip, shd, desc); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_create_pipeline(pip, shd, desc); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_discard_pipeline(_sg_pipeline_t* pip) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_discard_pipeline(pip); + #elif defined(SOKOL_METAL) + _sg_mtl_discard_pipeline(pip); + #elif defined(SOKOL_D3D11) + _sg_d3d11_discard_pipeline(pip); + #elif defined(SOKOL_WGPU) + _sg_wgpu_discard_pipeline(pip); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_discard_pipeline(pip); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline sg_resource_state _sg_create_attachments(_sg_attachments_t* atts, _sg_image_t** color_images, _sg_image_t** resolve_images, _sg_image_t* ds_image, const sg_attachments_desc* desc) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_create_attachments(atts, color_images, resolve_images, ds_image, desc); + #elif defined(SOKOL_METAL) + return _sg_mtl_create_attachments(atts, color_images, resolve_images, ds_image, desc); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_create_attachments(atts, color_images, resolve_images, ds_image, desc); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_create_attachments(atts, color_images, resolve_images, ds_image, desc); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_create_attachments(atts, color_images, resolve_images, ds_image, desc); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_discard_attachments(_sg_attachments_t* atts) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_discard_attachments(atts); + #elif defined(SOKOL_METAL) + _sg_mtl_discard_attachments(atts); + #elif defined(SOKOL_D3D11) + _sg_d3d11_discard_attachments(atts); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_discard_attachments(atts); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_discard_attachments(atts); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline _sg_image_t* _sg_attachments_color_image(const _sg_attachments_t* atts, int index) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_attachments_color_image(atts, index); + #elif defined(SOKOL_METAL) + return _sg_mtl_attachments_color_image(atts, index); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_attachments_color_image(atts, index); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_attachments_color_image(atts, index); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_attachments_color_image(atts, index); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline _sg_image_t* _sg_attachments_resolve_image(const _sg_attachments_t* atts, int index) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_attachments_resolve_image(atts, index); + #elif defined(SOKOL_METAL) + return _sg_mtl_attachments_resolve_image(atts, index); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_attachments_resolve_image(atts, index); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_attachments_resolve_image(atts, index); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_attachments_resolve_image(atts, index); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline _sg_image_t* _sg_attachments_ds_image(const _sg_attachments_t* atts) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_attachments_ds_image(atts); + #elif defined(SOKOL_METAL) + return _sg_mtl_attachments_ds_image(atts); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_attachments_ds_image(atts); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_attachments_ds_image(atts); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_attachments_ds_image(atts); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_begin_pass(const sg_pass* pass) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_begin_pass(pass); + #elif defined(SOKOL_METAL) + _sg_mtl_begin_pass(pass); + #elif defined(SOKOL_D3D11) + _sg_d3d11_begin_pass(pass); + #elif defined(SOKOL_WGPU) + _sg_wgpu_begin_pass(pass); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_begin_pass(pass); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_end_pass(void) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_end_pass(); + #elif defined(SOKOL_METAL) + _sg_mtl_end_pass(); + #elif defined(SOKOL_D3D11) + _sg_d3d11_end_pass(); + #elif defined(SOKOL_WGPU) + _sg_wgpu_end_pass(); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_end_pass(); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_apply_viewport(int x, int y, int w, int h, bool origin_top_left) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_apply_viewport(x, y, w, h, origin_top_left); + #elif defined(SOKOL_METAL) + _sg_mtl_apply_viewport(x, y, w, h, origin_top_left); + #elif defined(SOKOL_D3D11) + _sg_d3d11_apply_viewport(x, y, w, h, origin_top_left); + #elif defined(SOKOL_WGPU) + _sg_wgpu_apply_viewport(x, y, w, h, origin_top_left); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_apply_viewport(x, y, w, h, origin_top_left); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_apply_scissor_rect(int x, int y, int w, int h, bool origin_top_left) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_apply_scissor_rect(x, y, w, h, origin_top_left); + #elif defined(SOKOL_METAL) + _sg_mtl_apply_scissor_rect(x, y, w, h, origin_top_left); + #elif defined(SOKOL_D3D11) + _sg_d3d11_apply_scissor_rect(x, y, w, h, origin_top_left); + #elif defined(SOKOL_WGPU) + _sg_wgpu_apply_scissor_rect(x, y, w, h, origin_top_left); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_apply_scissor_rect(x, y, w, h, origin_top_left); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_apply_pipeline(_sg_pipeline_t* pip) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_apply_pipeline(pip); + #elif defined(SOKOL_METAL) + _sg_mtl_apply_pipeline(pip); + #elif defined(SOKOL_D3D11) + _sg_d3d11_apply_pipeline(pip); + #elif defined(SOKOL_WGPU) + _sg_wgpu_apply_pipeline(pip); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_apply_pipeline(pip); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline bool _sg_apply_bindings(_sg_bindings_t* bnd) { + #if defined(_SOKOL_ANY_GL) + return _sg_gl_apply_bindings(bnd); + #elif defined(SOKOL_METAL) + return _sg_mtl_apply_bindings(bnd); + #elif defined(SOKOL_D3D11) + return _sg_d3d11_apply_bindings(bnd); + #elif defined(SOKOL_WGPU) + return _sg_wgpu_apply_bindings(bnd); + #elif defined(SOKOL_DUMMY_BACKEND) + return _sg_dummy_apply_bindings(bnd); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_apply_uniforms(int ub_slot, const sg_range* data) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_apply_uniforms(ub_slot, data); + #elif defined(SOKOL_METAL) + _sg_mtl_apply_uniforms(ub_slot, data); + #elif defined(SOKOL_D3D11) + _sg_d3d11_apply_uniforms(ub_slot, data); + #elif defined(SOKOL_WGPU) + _sg_wgpu_apply_uniforms(ub_slot, data); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_apply_uniforms(ub_slot, data); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_draw(int base_element, int num_elements, int num_instances) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_draw(base_element, num_elements, num_instances); + #elif defined(SOKOL_METAL) + _sg_mtl_draw(base_element, num_elements, num_instances); + #elif defined(SOKOL_D3D11) + _sg_d3d11_draw(base_element, num_elements, num_instances); + #elif defined(SOKOL_WGPU) + _sg_wgpu_draw(base_element, num_elements, num_instances); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_draw(base_element, num_elements, num_instances); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_dispatch(num_groups_x, num_groups_y, num_groups_z); + #elif defined(SOKOL_METAL) + _sg_mtl_dispatch(num_groups_x, num_groups_y, num_groups_z); + #elif defined(SOKOL_D3D11) + _sg_d3d11_dispatch(num_groups_x, num_groups_y, num_groups_z); + #elif defined(SOKOL_WGPU) + _sg_wgpu_dispatch(num_groups_x, num_groups_y, num_groups_z); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_dispatch(num_groups_x, num_groups_y, num_groups_z); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_commit(void) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_commit(); + #elif defined(SOKOL_METAL) + _sg_mtl_commit(); + #elif defined(SOKOL_D3D11) + _sg_d3d11_commit(); + #elif defined(SOKOL_WGPU) + _sg_wgpu_commit(); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_commit(); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_update_buffer(_sg_buffer_t* buf, const sg_range* data) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_update_buffer(buf, data); + #elif defined(SOKOL_METAL) + _sg_mtl_update_buffer(buf, data); + #elif defined(SOKOL_D3D11) + _sg_d3d11_update_buffer(buf, data); + #elif defined(SOKOL_WGPU) + _sg_wgpu_update_buffer(buf, data); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_update_buffer(buf, data); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_append_buffer(_sg_buffer_t* buf, const sg_range* data, bool new_frame) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_append_buffer(buf, data, new_frame); + #elif defined(SOKOL_METAL) + _sg_mtl_append_buffer(buf, data, new_frame); + #elif defined(SOKOL_D3D11) + _sg_d3d11_append_buffer(buf, data, new_frame); + #elif defined(SOKOL_WGPU) + _sg_wgpu_append_buffer(buf, data, new_frame); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_append_buffer(buf, data, new_frame); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_update_image(_sg_image_t* img, const sg_image_data* data) { + #if defined(_SOKOL_ANY_GL) + _sg_gl_update_image(img, data); + #elif defined(SOKOL_METAL) + _sg_mtl_update_image(img, data); + #elif defined(SOKOL_D3D11) + _sg_d3d11_update_image(img, data); + #elif defined(SOKOL_WGPU) + _sg_wgpu_update_image(img, data); + #elif defined(SOKOL_DUMMY_BACKEND) + _sg_dummy_update_image(img, data); + #else + #error("INVALID BACKEND"); + #endif +} + +static inline void _sg_push_debug_group(const char* name) { + #if defined(SOKOL_METAL) + _sg_mtl_push_debug_group(name); + #else + _SOKOL_UNUSED(name); + #endif +} + +static inline void _sg_pop_debug_group(void) { + #if defined(SOKOL_METAL) + _sg_mtl_pop_debug_group(); + #endif +} + +// ██████ ██████ ██████ ██ +// ██ ██ ██ ██ ██ ██ ██ +// ██████ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ +// ██ ██████ ██████ ███████ +// +// >>pool +_SOKOL_PRIVATE void _sg_pool_init(_sg_pool_t* pool, int num) { + SOKOL_ASSERT(pool && (num >= 1)); + // slot 0 is reserved for the 'invalid id', so bump the pool size by 1 + pool->size = num + 1; + pool->queue_top = 0; + // generation counters indexable by pool slot index, slot 0 is reserved + size_t gen_ctrs_size = sizeof(uint32_t) * (size_t)pool->size; + pool->gen_ctrs = (uint32_t*)_sg_malloc_clear(gen_ctrs_size); + // it's not a bug to only reserve 'num' here + pool->free_queue = (int*) _sg_malloc_clear(sizeof(int) * (size_t)num); + // never allocate the zero-th pool item since the invalid id is 0 + for (int i = pool->size-1; i >= 1; i--) { + pool->free_queue[pool->queue_top++] = i; + } +} + +_SOKOL_PRIVATE void _sg_pool_discard(_sg_pool_t* pool) { + SOKOL_ASSERT(pool); + SOKOL_ASSERT(pool->free_queue); + _sg_free(pool->free_queue); + pool->free_queue = 0; + SOKOL_ASSERT(pool->gen_ctrs); + _sg_free(pool->gen_ctrs); + pool->gen_ctrs = 0; + pool->size = 0; + pool->queue_top = 0; +} + +_SOKOL_PRIVATE int _sg_pool_alloc_index(_sg_pool_t* pool) { + SOKOL_ASSERT(pool); + SOKOL_ASSERT(pool->free_queue); + if (pool->queue_top > 0) { + int slot_index = pool->free_queue[--pool->queue_top]; + SOKOL_ASSERT((slot_index > 0) && (slot_index < pool->size)); + return slot_index; + } else { + // pool exhausted + return _SG_INVALID_SLOT_INDEX; + } +} + +_SOKOL_PRIVATE void _sg_pool_free_index(_sg_pool_t* pool, int slot_index) { + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < pool->size)); + SOKOL_ASSERT(pool); + SOKOL_ASSERT(pool->free_queue); + SOKOL_ASSERT(pool->queue_top < pool->size); + #ifdef SOKOL_DEBUG + // debug check against double-free + for (int i = 0; i < pool->queue_top; i++) { + SOKOL_ASSERT(pool->free_queue[i] != slot_index); + } + #endif + pool->free_queue[pool->queue_top++] = slot_index; + SOKOL_ASSERT(pool->queue_top <= (pool->size-1)); +} + +_SOKOL_PRIVATE void _sg_slot_reset(_sg_slot_t* slot) { + SOKOL_ASSERT(slot); + _sg_clear(slot, sizeof(_sg_slot_t)); +} + +_SOKOL_PRIVATE void _sg_reset_buffer_to_alloc_state(_sg_buffer_t* buf) { + SOKOL_ASSERT(buf); + _sg_slot_t slot = buf->slot; + _sg_clear(buf, sizeof(*buf)); + buf->slot = slot; + buf->slot.state = SG_RESOURCESTATE_ALLOC; +} + +_SOKOL_PRIVATE void _sg_reset_image_to_alloc_state(_sg_image_t* img) { + SOKOL_ASSERT(img); + _sg_slot_t slot = img->slot; + _sg_clear(img, sizeof(*img)); + img->slot = slot; + img->slot.state = SG_RESOURCESTATE_ALLOC; +} + +_SOKOL_PRIVATE void _sg_reset_sampler_to_alloc_state(_sg_sampler_t* smp) { + SOKOL_ASSERT(smp); + _sg_slot_t slot = smp->slot; + _sg_clear(smp, sizeof(*smp)); + smp->slot = slot; + smp->slot.state = SG_RESOURCESTATE_ALLOC; +} + +_SOKOL_PRIVATE void _sg_reset_shader_to_alloc_state(_sg_shader_t* shd) { + SOKOL_ASSERT(shd); + _sg_slot_t slot = shd->slot; + _sg_clear(shd, sizeof(*shd)); + shd->slot = slot; + shd->slot.state = SG_RESOURCESTATE_ALLOC; +} + +_SOKOL_PRIVATE void _sg_reset_pipeline_to_alloc_state(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip); + _sg_slot_t slot = pip->slot; + _sg_clear(pip, sizeof(*pip)); + pip->slot = slot; + pip->slot.state = SG_RESOURCESTATE_ALLOC; +} + +_SOKOL_PRIVATE void _sg_reset_attachments_to_alloc_state(_sg_attachments_t* atts) { + SOKOL_ASSERT(atts); + _sg_slot_t slot = atts->slot; + _sg_clear(atts, sizeof(*atts)); + atts->slot = slot; + atts->slot.state = SG_RESOURCESTATE_ALLOC; +} + +_SOKOL_PRIVATE void _sg_setup_pools(_sg_pools_t* p, const sg_desc* desc) { + SOKOL_ASSERT(p); + SOKOL_ASSERT(desc); + // note: the pools here will have an additional item, since slot 0 is reserved + SOKOL_ASSERT((desc->buffer_pool_size > 0) && (desc->buffer_pool_size < _SG_MAX_POOL_SIZE)); + _sg_pool_init(&p->buffer_pool, desc->buffer_pool_size); + size_t buffer_pool_byte_size = sizeof(_sg_buffer_t) * (size_t)p->buffer_pool.size; + p->buffers = (_sg_buffer_t*) _sg_malloc_clear(buffer_pool_byte_size); + + SOKOL_ASSERT((desc->image_pool_size > 0) && (desc->image_pool_size < _SG_MAX_POOL_SIZE)); + _sg_pool_init(&p->image_pool, desc->image_pool_size); + size_t image_pool_byte_size = sizeof(_sg_image_t) * (size_t)p->image_pool.size; + p->images = (_sg_image_t*) _sg_malloc_clear(image_pool_byte_size); + + SOKOL_ASSERT((desc->sampler_pool_size > 0) && (desc->sampler_pool_size < _SG_MAX_POOL_SIZE)); + _sg_pool_init(&p->sampler_pool, desc->sampler_pool_size); + size_t sampler_pool_byte_size = sizeof(_sg_sampler_t) * (size_t)p->sampler_pool.size; + p->samplers = (_sg_sampler_t*) _sg_malloc_clear(sampler_pool_byte_size); + + SOKOL_ASSERT((desc->shader_pool_size > 0) && (desc->shader_pool_size < _SG_MAX_POOL_SIZE)); + _sg_pool_init(&p->shader_pool, desc->shader_pool_size); + size_t shader_pool_byte_size = sizeof(_sg_shader_t) * (size_t)p->shader_pool.size; + p->shaders = (_sg_shader_t*) _sg_malloc_clear(shader_pool_byte_size); + + SOKOL_ASSERT((desc->pipeline_pool_size > 0) && (desc->pipeline_pool_size < _SG_MAX_POOL_SIZE)); + _sg_pool_init(&p->pipeline_pool, desc->pipeline_pool_size); + size_t pipeline_pool_byte_size = sizeof(_sg_pipeline_t) * (size_t)p->pipeline_pool.size; + p->pipelines = (_sg_pipeline_t*) _sg_malloc_clear(pipeline_pool_byte_size); + + SOKOL_ASSERT((desc->attachments_pool_size > 0) && (desc->attachments_pool_size < _SG_MAX_POOL_SIZE)); + _sg_pool_init(&p->attachments_pool, desc->attachments_pool_size); + size_t attachments_pool_byte_size = sizeof(_sg_attachments_t) * (size_t)p->attachments_pool.size; + p->attachments = (_sg_attachments_t*) _sg_malloc_clear(attachments_pool_byte_size); +} + +_SOKOL_PRIVATE void _sg_discard_pools(_sg_pools_t* p) { + SOKOL_ASSERT(p); + _sg_free(p->attachments); p->attachments = 0; + _sg_free(p->pipelines); p->pipelines = 0; + _sg_free(p->shaders); p->shaders = 0; + _sg_free(p->samplers); p->samplers = 0; + _sg_free(p->images); p->images = 0; + _sg_free(p->buffers); p->buffers = 0; + _sg_pool_discard(&p->attachments_pool); + _sg_pool_discard(&p->pipeline_pool); + _sg_pool_discard(&p->shader_pool); + _sg_pool_discard(&p->sampler_pool); + _sg_pool_discard(&p->image_pool); + _sg_pool_discard(&p->buffer_pool); +} + +/* allocate the slot at slot_index: + - bump the slot's generation counter + - create a resource id from the generation counter and slot index + - set the slot's id to this id + - set the slot's state to ALLOC + - return the resource id +*/ +_SOKOL_PRIVATE uint32_t _sg_slot_alloc(_sg_pool_t* pool, _sg_slot_t* slot, int slot_index) { + /* FIXME: add handling for an overflowing generation counter, + for now, just overflow (another option is to disable + the slot) + */ + SOKOL_ASSERT(pool && pool->gen_ctrs); + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < pool->size)); + SOKOL_ASSERT(slot->id == SG_INVALID_ID); + SOKOL_ASSERT(slot->state == SG_RESOURCESTATE_INITIAL); + uint32_t ctr = ++pool->gen_ctrs[slot_index]; + slot->id = (ctr<<_SG_SLOT_SHIFT)|(slot_index & _SG_SLOT_MASK); + slot->state = SG_RESOURCESTATE_ALLOC; + return slot->id; +} + +// extract slot index from id +_SOKOL_PRIVATE int _sg_slot_index(uint32_t id) { + int slot_index = (int) (id & _SG_SLOT_MASK); + SOKOL_ASSERT(_SG_INVALID_SLOT_INDEX != slot_index); + return slot_index; +} + +// returns pointer to resource by id without matching id check +_SOKOL_PRIVATE _sg_buffer_t* _sg_buffer_at(const _sg_pools_t* p, uint32_t buf_id) { + SOKOL_ASSERT(p && (SG_INVALID_ID != buf_id)); + int slot_index = _sg_slot_index(buf_id); + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < p->buffer_pool.size)); + return &p->buffers[slot_index]; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_image_at(const _sg_pools_t* p, uint32_t img_id) { + SOKOL_ASSERT(p && (SG_INVALID_ID != img_id)); + int slot_index = _sg_slot_index(img_id); + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < p->image_pool.size)); + return &p->images[slot_index]; +} + +_SOKOL_PRIVATE _sg_sampler_t* _sg_sampler_at(const _sg_pools_t* p, uint32_t smp_id) { + SOKOL_ASSERT(p && (SG_INVALID_ID != smp_id)); + int slot_index = _sg_slot_index(smp_id); + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < p->sampler_pool.size)); + return &p->samplers[slot_index]; +} + +_SOKOL_PRIVATE _sg_shader_t* _sg_shader_at(const _sg_pools_t* p, uint32_t shd_id) { + SOKOL_ASSERT(p && (SG_INVALID_ID != shd_id)); + int slot_index = _sg_slot_index(shd_id); + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < p->shader_pool.size)); + return &p->shaders[slot_index]; +} + +_SOKOL_PRIVATE _sg_pipeline_t* _sg_pipeline_at(const _sg_pools_t* p, uint32_t pip_id) { + SOKOL_ASSERT(p && (SG_INVALID_ID != pip_id)); + int slot_index = _sg_slot_index(pip_id); + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < p->pipeline_pool.size)); + return &p->pipelines[slot_index]; +} + +_SOKOL_PRIVATE _sg_attachments_t* _sg_attachments_at(const _sg_pools_t* p, uint32_t atts_id) { + SOKOL_ASSERT(p && (SG_INVALID_ID != atts_id)); + int slot_index = _sg_slot_index(atts_id); + SOKOL_ASSERT((slot_index > _SG_INVALID_SLOT_INDEX) && (slot_index < p->attachments_pool.size)); + return &p->attachments[slot_index]; +} + +// returns pointer to resource with matching id check, may return 0 +_SOKOL_PRIVATE _sg_buffer_t* _sg_lookup_buffer(const _sg_pools_t* p, uint32_t buf_id) { + if (SG_INVALID_ID != buf_id) { + _sg_buffer_t* buf = _sg_buffer_at(p, buf_id); + if (buf->slot.id == buf_id) { + return buf; + } + } + return 0; +} + +_SOKOL_PRIVATE _sg_image_t* _sg_lookup_image(const _sg_pools_t* p, uint32_t img_id) { + if (SG_INVALID_ID != img_id) { + _sg_image_t* img = _sg_image_at(p, img_id); + if (img->slot.id == img_id) { + return img; + } + } + return 0; +} + +_SOKOL_PRIVATE _sg_sampler_t* _sg_lookup_sampler(const _sg_pools_t* p, uint32_t smp_id) { + if (SG_INVALID_ID != smp_id) { + _sg_sampler_t* smp = _sg_sampler_at(p, smp_id); + if (smp->slot.id == smp_id) { + return smp; + } + } + return 0; +} + +_SOKOL_PRIVATE _sg_shader_t* _sg_lookup_shader(const _sg_pools_t* p, uint32_t shd_id) { + SOKOL_ASSERT(p); + if (SG_INVALID_ID != shd_id) { + _sg_shader_t* shd = _sg_shader_at(p, shd_id); + if (shd->slot.id == shd_id) { + return shd; + } + } + return 0; +} + +_SOKOL_PRIVATE _sg_pipeline_t* _sg_lookup_pipeline(const _sg_pools_t* p, uint32_t pip_id) { + SOKOL_ASSERT(p); + if (SG_INVALID_ID != pip_id) { + _sg_pipeline_t* pip = _sg_pipeline_at(p, pip_id); + if (pip->slot.id == pip_id) { + return pip; + } + } + return 0; +} + +_SOKOL_PRIVATE _sg_attachments_t* _sg_lookup_attachments(const _sg_pools_t* p, uint32_t atts_id) { + SOKOL_ASSERT(p); + if (SG_INVALID_ID != atts_id) { + _sg_attachments_t* atts = _sg_attachments_at(p, atts_id); + if (atts->slot.id == atts_id) { + return atts; + } + } + return 0; +} + +_SOKOL_PRIVATE void _sg_discard_all_resources(_sg_pools_t* p) { + /* this is a bit dumb since it loops over all pool slots to + find the occupied slots, on the other hand it is only ever + executed at shutdown + NOTE: ONLY EXECUTE THIS AT SHUTDOWN + ...because the free queues will not be reset + and the resource slots not be cleared! + */ + for (int i = 1; i < p->buffer_pool.size; i++) { + sg_resource_state state = p->buffers[i].slot.state; + if ((state == SG_RESOURCESTATE_VALID) || (state == SG_RESOURCESTATE_FAILED)) { + _sg_discard_buffer(&p->buffers[i]); + } + } + for (int i = 1; i < p->image_pool.size; i++) { + sg_resource_state state = p->images[i].slot.state; + if ((state == SG_RESOURCESTATE_VALID) || (state == SG_RESOURCESTATE_FAILED)) { + _sg_discard_image(&p->images[i]); + } + } + for (int i = 1; i < p->sampler_pool.size; i++) { + sg_resource_state state = p->samplers[i].slot.state; + if ((state == SG_RESOURCESTATE_VALID) || (state == SG_RESOURCESTATE_FAILED)) { + _sg_discard_sampler(&p->samplers[i]); + } + } + for (int i = 1; i < p->shader_pool.size; i++) { + sg_resource_state state = p->shaders[i].slot.state; + if ((state == SG_RESOURCESTATE_VALID) || (state == SG_RESOURCESTATE_FAILED)) { + _sg_discard_shader(&p->shaders[i]); + } + } + for (int i = 1; i < p->pipeline_pool.size; i++) { + sg_resource_state state = p->pipelines[i].slot.state; + if ((state == SG_RESOURCESTATE_VALID) || (state == SG_RESOURCESTATE_FAILED)) { + _sg_discard_pipeline(&p->pipelines[i]); + } + } + for (int i = 1; i < p->attachments_pool.size; i++) { + sg_resource_state state = p->attachments[i].slot.state; + if ((state == SG_RESOURCESTATE_VALID) || (state == SG_RESOURCESTATE_FAILED)) { + _sg_discard_attachments(&p->attachments[i]); + } + } +} + +// ████████ ██████ █████ ██████ ██ ██ ███████ ██████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██ ██████ ███████ ██ █████ █████ ██████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██████ ██ ██ ███████ ██ ██ +// +// >>tracker +_SOKOL_PRIVATE void _sg_tracker_init(_sg_tracker_t* tracker, uint32_t num) { + SOKOL_ASSERT(tracker); + SOKOL_ASSERT(num > 0); + SOKOL_ASSERT(0 == tracker->size); + SOKOL_ASSERT(0 == tracker->cur); + SOKOL_ASSERT(0 == tracker->items); + tracker->size = (uint32_t)num; + tracker->items = (uint32_t*)_sg_malloc_clear(num * sizeof(uint32_t)); +} + +_SOKOL_PRIVATE void _sg_tracker_discard(_sg_tracker_t* tracker) { + SOKOL_ASSERT(tracker); + if (tracker->items) { + _sg_free(tracker->items); + } + tracker->size = 0; + tracker->cur = 0; + tracker->items = 0; +} + +_SOKOL_PRIVATE void _sg_tracker_reset(_sg_tracker_t* tracker) { + SOKOL_ASSERT(tracker && tracker->items); + tracker->cur = 0; +} + +_SOKOL_PRIVATE bool _sg_tracker_add(_sg_tracker_t* tracker, uint32_t res_id) { + SOKOL_ASSERT(tracker && tracker->items); + if (tracker->cur < tracker->size) { + tracker->items[tracker->cur++] = res_id; + return true; + } else { + return false; + } +} + +// ██ ██ █████ ██ ██ ██████ █████ ████████ ██ ██████ ███ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ████ ██ +// ██ ██ ███████ ██ ██ ██ ██ ███████ ██ ██ ██ ██ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ████ ██ ██ ███████ ██ ██████ ██ ██ ██ ██ ██████ ██ ████ +// +// >>validation +#if defined(SOKOL_DEBUG) +_SOKOL_PRIVATE void _sg_validate_begin(void) { + _sg.validate_error = SG_LOGITEM_OK; +} + +_SOKOL_PRIVATE bool _sg_validate_end(void) { + if (_sg.validate_error != SG_LOGITEM_OK) { + #if !defined(SOKOL_VALIDATE_NON_FATAL) + _SG_PANIC(VALIDATION_FAILED); + return false; + #else + return false; + #endif + } else { + return true; + } +} +#endif + +_SOKOL_PRIVATE bool _sg_validate_buffer_desc(const sg_buffer_desc* desc) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(desc); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(desc); + _sg_validate_begin(); + _SG_VALIDATE(desc->_start_canary == 0, VALIDATE_BUFFERDESC_CANARY); + _SG_VALIDATE(desc->_end_canary == 0, VALIDATE_BUFFERDESC_CANARY); + _SG_VALIDATE(desc->size > 0, VALIDATE_BUFFERDESC_EXPECT_NONZERO_SIZE); + bool injected = (0 != desc->gl_buffers[0]) || + (0 != desc->mtl_buffers[0]) || + (0 != desc->d3d11_buffer) || + (0 != desc->wgpu_buffer); + if (!injected && (desc->usage == SG_USAGE_IMMUTABLE)) { + if (desc->data.ptr) { + _SG_VALIDATE(desc->size == desc->data.size, VALIDATE_BUFFERDESC_EXPECT_MATCHING_DATA_SIZE); + } else { + _SG_VALIDATE(desc->data.size == 0, VALIDATE_BUFFERDESC_EXPECT_ZERO_DATA_SIZE); + } + } else { + _SG_VALIDATE(0 == desc->data.ptr, VALIDATE_BUFFERDESC_EXPECT_NO_DATA); + _SG_VALIDATE(desc->data.size == 0, VALIDATE_BUFFERDESC_EXPECT_ZERO_DATA_SIZE); + } + if (desc->type == SG_BUFFERTYPE_STORAGEBUFFER) { + _SG_VALIDATE(_sg.features.compute, VALIDATE_BUFFERDESC_STORAGEBUFFER_SUPPORTED); + _SG_VALIDATE(_sg_multiple_u64(desc->size, 4), VALIDATE_BUFFERDESC_STORAGEBUFFER_SIZE_MULTIPLE_4); + } + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE void _sg_validate_image_data(const sg_image_data* data, sg_pixel_format fmt, int width, int height, int num_faces, int num_mips, int num_slices) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(data); + _SOKOL_UNUSED(fmt); + _SOKOL_UNUSED(width); + _SOKOL_UNUSED(height); + _SOKOL_UNUSED(num_faces); + _SOKOL_UNUSED(num_mips); + _SOKOL_UNUSED(num_slices); + #else + for (int face_index = 0; face_index < num_faces; face_index++) { + for (int mip_index = 0; mip_index < num_mips; mip_index++) { + const bool has_data = data->subimage[face_index][mip_index].ptr != 0; + const bool has_size = data->subimage[face_index][mip_index].size > 0; + _SG_VALIDATE(has_data && has_size, VALIDATE_IMAGEDATA_NODATA); + const int mip_width = _sg_miplevel_dim(width, mip_index); + const int mip_height = _sg_miplevel_dim(height, mip_index); + const int bytes_per_slice = _sg_surface_pitch(fmt, mip_width, mip_height, 1); + const int expected_size = bytes_per_slice * num_slices; + _SG_VALIDATE(expected_size == (int)data->subimage[face_index][mip_index].size, VALIDATE_IMAGEDATA_DATA_SIZE); + } + } + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_image_desc(const sg_image_desc* desc) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(desc); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(desc); + _sg_validate_begin(); + _SG_VALIDATE(desc->_start_canary == 0, VALIDATE_IMAGEDESC_CANARY); + _SG_VALIDATE(desc->_end_canary == 0, VALIDATE_IMAGEDESC_CANARY); + _SG_VALIDATE(desc->width > 0, VALIDATE_IMAGEDESC_WIDTH); + _SG_VALIDATE(desc->height > 0, VALIDATE_IMAGEDESC_HEIGHT); + const sg_pixel_format fmt = desc->pixel_format; + const sg_usage usage = desc->usage; + const bool injected = (0 != desc->gl_textures[0]) || + (0 != desc->mtl_textures[0]) || + (0 != desc->d3d11_texture) || + (0 != desc->wgpu_texture); + if (_sg_is_depth_or_depth_stencil_format(fmt)) { + _SG_VALIDATE(desc->type != SG_IMAGETYPE_3D, VALIDATE_IMAGEDESC_DEPTH_3D_IMAGE); + } + if (desc->render_target) { + SOKOL_ASSERT(((int)fmt >= 0) && ((int)fmt < _SG_PIXELFORMAT_NUM)); + _SG_VALIDATE(_sg.formats[fmt].render, VALIDATE_IMAGEDESC_RT_PIXELFORMAT); + _SG_VALIDATE(usage == SG_USAGE_IMMUTABLE, VALIDATE_IMAGEDESC_RT_IMMUTABLE); + _SG_VALIDATE(desc->data.subimage[0][0].ptr==0, VALIDATE_IMAGEDESC_RT_NO_DATA); + if (desc->sample_count > 1) { + _SG_VALIDATE(_sg.formats[fmt].msaa, VALIDATE_IMAGEDESC_NO_MSAA_RT_SUPPORT); + _SG_VALIDATE(desc->num_mipmaps == 1, VALIDATE_IMAGEDESC_MSAA_NUM_MIPMAPS); + _SG_VALIDATE(desc->type != SG_IMAGETYPE_3D, VALIDATE_IMAGEDESC_MSAA_3D_IMAGE); + _SG_VALIDATE(desc->type != SG_IMAGETYPE_CUBE, VALIDATE_IMAGEDESC_MSAA_CUBE_IMAGE); + } + } else { + _SG_VALIDATE(desc->sample_count == 1, VALIDATE_IMAGEDESC_MSAA_BUT_NO_RT); + const bool valid_nonrt_fmt = !_sg_is_valid_rendertarget_depth_format(fmt); + _SG_VALIDATE(valid_nonrt_fmt, VALIDATE_IMAGEDESC_NONRT_PIXELFORMAT); + const bool is_compressed = _sg_is_compressed_pixel_format(desc->pixel_format); + const bool is_immutable = (usage == SG_USAGE_IMMUTABLE); + if (is_compressed) { + _SG_VALIDATE(is_immutable, VALIDATE_IMAGEDESC_COMPRESSED_IMMUTABLE); + } + if (!injected && is_immutable) { + // image desc must have valid data + _sg_validate_image_data(&desc->data, + desc->pixel_format, + desc->width, + desc->height, + (desc->type == SG_IMAGETYPE_CUBE) ? 6 : 1, + desc->num_mipmaps, + desc->num_slices); + } else { + // image desc must not have data + for (int face_index = 0; face_index < SG_CUBEFACE_NUM; face_index++) { + for (int mip_index = 0; mip_index < SG_MAX_MIPMAPS; mip_index++) { + const bool no_data = 0 == desc->data.subimage[face_index][mip_index].ptr; + const bool no_size = 0 == desc->data.subimage[face_index][mip_index].size; + if (injected) { + _SG_VALIDATE(no_data && no_size, VALIDATE_IMAGEDESC_INJECTED_NO_DATA); + } + if (!is_immutable) { + _SG_VALIDATE(no_data && no_size, VALIDATE_IMAGEDESC_DYNAMIC_NO_DATA); + } + } + } + } + } + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_sampler_desc(const sg_sampler_desc* desc) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(desc); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(desc); + _sg_validate_begin(); + _SG_VALIDATE(desc->_start_canary == 0, VALIDATE_SAMPLERDESC_CANARY); + _SG_VALIDATE(desc->_end_canary == 0, VALIDATE_SAMPLERDESC_CANARY); + // restriction from WebGPU: when anisotropy > 1, all filters must be linear + if (desc->max_anisotropy > 1) { + _SG_VALIDATE((desc->min_filter == SG_FILTER_LINEAR) + && (desc->mag_filter == SG_FILTER_LINEAR) + && (desc->mipmap_filter == SG_FILTER_LINEAR), + VALIDATE_SAMPLERDESC_ANISTROPIC_REQUIRES_LINEAR_FILTERING); + } + return _sg_validate_end(); + #endif +} + +typedef struct { + uint64_t lo, hi; +} _sg_u128_t; + +_SOKOL_PRIVATE _sg_u128_t _sg_u128(void) { + _sg_u128_t res; + _sg_clear(&res, sizeof(res)); + return res; +} + +_SOKOL_PRIVATE _sg_u128_t _sg_validate_set_slot_bit(_sg_u128_t bits, sg_shader_stage stage, uint8_t slot) { + switch (stage) { + case SG_SHADERSTAGE_NONE: + SOKOL_ASSERT(slot < 128); + if (slot < 64) { + bits.lo |= 1ULL << slot; + } else { + bits.hi |= 1ULL << (slot - 64); + } + break; + case SG_SHADERSTAGE_VERTEX: + SOKOL_ASSERT(slot < 64); + bits.lo |= 1ULL << slot; + break; + case SG_SHADERSTAGE_FRAGMENT: + SOKOL_ASSERT(slot < 64); + bits.hi |= 1ULL << slot; + break; + case SG_SHADERSTAGE_COMPUTE: + SOKOL_ASSERT(slot < 64); + bits.lo |= 1ULL << slot; + break; + default: + SOKOL_UNREACHABLE; + break; + } + return bits; +} + +_SOKOL_PRIVATE bool _sg_validate_slot_bits(_sg_u128_t bits, sg_shader_stage stage, uint8_t slot) { + _sg_u128_t mask = _sg_u128(); + switch (stage) { + case SG_SHADERSTAGE_NONE: + SOKOL_ASSERT(slot < 128); + if (slot < 64) { + mask.lo = 1ULL << slot; + } else { + mask.hi = 1ULL << (slot - 64); + } + break; + case SG_SHADERSTAGE_VERTEX: + SOKOL_ASSERT(slot < 64); + mask.lo = 1ULL << slot; + break; + case SG_SHADERSTAGE_FRAGMENT: + SOKOL_ASSERT(slot < 64); + mask.hi = 1ULL << slot; + break; + case SG_SHADERSTAGE_COMPUTE: + SOKOL_ASSERT(slot < 64); + mask.lo = 1ULL << slot; + break; + default: + SOKOL_UNREACHABLE; + break; + } + return ((bits.lo & mask.lo) == 0) && ((bits.hi & mask.hi) == 0); +} + +_SOKOL_PRIVATE bool _sg_validate_shader_desc(const sg_shader_desc* desc) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(desc); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(desc); + bool is_compute_shader = (desc->compute_func.source != 0) || (desc->compute_func.bytecode.ptr != 0); + _sg_validate_begin(); + _SG_VALIDATE(desc->_start_canary == 0, VALIDATE_SHADERDESC_CANARY); + _SG_VALIDATE(desc->_end_canary == 0, VALIDATE_SHADERDESC_CANARY); + #if defined(SOKOL_GLCORE) || defined(SOKOL_GLES3) || defined(SOKOL_WGPU) + // on GL or WebGPU, must provide shader source code + if (is_compute_shader) { + _SG_VALIDATE(0 != desc->compute_func.source, VALIDATE_SHADERDESC_COMPUTE_SOURCE); + } else { + _SG_VALIDATE(0 != desc->vertex_func.source, VALIDATE_SHADERDESC_VERTEX_SOURCE); + _SG_VALIDATE(0 != desc->fragment_func.source, VALIDATE_SHADERDESC_FRAGMENT_SOURCE); + } + #elif defined(SOKOL_METAL) || defined(SOKOL_D3D11) + // on Metal or D3D11, must provide shader source code or byte code + if (is_compute_shader) { + _SG_VALIDATE((0 != desc->compute_func.source) || (0 != desc->compute_func.bytecode.ptr), VALIDATE_SHADERDESC_COMPUTE_SOURCE_OR_BYTECODE); + } else { + _SG_VALIDATE((0 != desc->vertex_func.source)|| (0 != desc->vertex_func.bytecode.ptr), VALIDATE_SHADERDESC_VERTEX_SOURCE_OR_BYTECODE); + _SG_VALIDATE((0 != desc->fragment_func.source) || (0 != desc->fragment_func.bytecode.ptr), VALIDATE_SHADERDESC_FRAGMENT_SOURCE_OR_BYTECODE); + } + #else + // Dummy Backend, don't require source or bytecode + #endif + if (is_compute_shader) { + _SG_VALIDATE((0 == desc->vertex_func.source) && (0 == desc->vertex_func.bytecode.ptr), VALIDATE_SHADERDESC_INVALID_SHADER_COMBO); + _SG_VALIDATE((0 == desc->fragment_func.source) && (0 == desc->fragment_func.bytecode.ptr), VALIDATE_SHADERDESC_INVALID_SHADER_COMBO); + } else { + _SG_VALIDATE((0 == desc->compute_func.source) && (0 == desc->compute_func.bytecode.ptr), VALIDATE_SHADERDESC_INVALID_SHADER_COMBO); + } + #if defined(SOKOL_METAL) + if (is_compute_shader) { + _SG_VALIDATE(desc->mtl_threads_per_threadgroup.x > 0, VALIDATE_SHADERDESC_METAL_THREADS_PER_THREADGROUP); + _SG_VALIDATE(desc->mtl_threads_per_threadgroup.y > 0, VALIDATE_SHADERDESC_METAL_THREADS_PER_THREADGROUP); + _SG_VALIDATE(desc->mtl_threads_per_threadgroup.z > 0, VALIDATE_SHADERDESC_METAL_THREADS_PER_THREADGROUP); + } + #endif + for (size_t i = 0; i < SG_MAX_VERTEX_ATTRIBUTES; i++) { + if (desc->attrs[i].glsl_name) { + _SG_VALIDATE(strlen(desc->attrs[i].glsl_name) < _SG_STRING_SIZE, VALIDATE_SHADERDESC_ATTR_STRING_TOO_LONG); + } + if (desc->attrs[i].hlsl_sem_name) { + _SG_VALIDATE(strlen(desc->attrs[i].hlsl_sem_name) < _SG_STRING_SIZE, VALIDATE_SHADERDESC_ATTR_STRING_TOO_LONG); + } + } + // if shader byte code, the size must also be provided + if (0 != desc->vertex_func.bytecode.ptr) { + _SG_VALIDATE(desc->vertex_func.bytecode.size > 0, VALIDATE_SHADERDESC_NO_BYTECODE_SIZE); + } + if (0 != desc->fragment_func.bytecode.ptr) { + _SG_VALIDATE(desc->fragment_func.bytecode.size > 0, VALIDATE_SHADERDESC_NO_BYTECODE_SIZE); + } + if (0 != desc->compute_func.bytecode.ptr) { + _SG_VALIDATE(desc->compute_func.bytecode.size > 0, VALIDATE_SHADERDESC_NO_BYTECODE_SIZE); + } + + #if defined(SOKOL_METAL) + _sg_u128_t msl_buf_bits = _sg_u128(); + _sg_u128_t msl_tex_bits = _sg_u128(); + _sg_u128_t msl_smp_bits = _sg_u128(); + #elif defined(SOKOL_D3D11) + _sg_u128_t hlsl_buf_bits = _sg_u128(); + _sg_u128_t hlsl_srv_bits = _sg_u128(); + _sg_u128_t hlsl_uav_bits = _sg_u128(); + _sg_u128_t hlsl_smp_bits = _sg_u128(); + #elif defined(_SOKOL_ANY_GL) + _sg_u128_t glsl_bnd_bits = _sg_u128(); + #elif defined(SOKOL_WGPU) + _sg_u128_t wgsl_group0_bits = _sg_u128(); + _sg_u128_t wgsl_group1_bits = _sg_u128(); + #endif + for (size_t ub_idx = 0; ub_idx < SG_MAX_UNIFORMBLOCK_BINDSLOTS; ub_idx++) { + const sg_shader_uniform_block* ub_desc = &desc->uniform_blocks[ub_idx]; + if (ub_desc->stage == SG_SHADERSTAGE_NONE) { + continue; + } + _SG_VALIDATE(ub_desc->size > 0, VALIDATE_SHADERDESC_UNIFORMBLOCK_SIZE_IS_ZERO); + #if defined(SOKOL_METAL) + _SG_VALIDATE(ub_desc->msl_buffer_n < _SG_MTL_MAX_STAGE_UB_BINDINGS, VALIDATE_SHADERDESC_UNIFORMBLOCK_METAL_BUFFER_SLOT_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(msl_buf_bits, ub_desc->stage, ub_desc->msl_buffer_n), VALIDATE_SHADERDESC_UNIFORMBLOCK_METAL_BUFFER_SLOT_COLLISION); + msl_buf_bits = _sg_validate_set_slot_bit(msl_buf_bits, ub_desc->stage, ub_desc->msl_buffer_n); + #elif defined(SOKOL_D3D11) + _SG_VALIDATE(ub_desc->hlsl_register_b_n < _SG_D3D11_MAX_STAGE_UB_BINDINGS, VALIDATE_SHADERDESC_UNIFORMBLOCK_HLSL_REGISTER_B_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(hlsl_buf_bits, ub_desc->stage, ub_desc->hlsl_register_b_n), VALIDATE_SHADERDESC_UNIFORMBLOCK_HLSL_REGISTER_B_COLLISION); + hlsl_buf_bits = _sg_validate_set_slot_bit(hlsl_buf_bits, ub_desc->stage, ub_desc->hlsl_register_b_n); + #elif defined(SOKOL_WGPU) + _SG_VALIDATE(ub_desc->wgsl_group0_binding_n < _SG_WGPU_MAX_UB_BINDGROUP_BIND_SLOTS, VALIDATE_SHADERDESC_UNIFORMBLOCK_WGSL_GROUP0_BINDING_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(wgsl_group0_bits, SG_SHADERSTAGE_NONE, ub_desc->wgsl_group0_binding_n), VALIDATE_SHADERDESC_UNIFORMBLOCK_WGSL_GROUP0_BINDING_COLLISION); + wgsl_group0_bits = _sg_validate_set_slot_bit(wgsl_group0_bits, SG_SHADERSTAGE_NONE, ub_desc->wgsl_group0_binding_n); + #endif + #if defined(_SOKOL_ANY_GL) + bool uniforms_continuous = true; + uint32_t uniform_offset = 0; + int num_uniforms = 0; + for (size_t u_index = 0; u_index < SG_MAX_UNIFORMBLOCK_MEMBERS; u_index++) { + const sg_glsl_shader_uniform* u_desc = &ub_desc->glsl_uniforms[u_index]; + if (u_desc->type != SG_UNIFORMTYPE_INVALID) { + _SG_VALIDATE(uniforms_continuous, VALIDATE_SHADERDESC_UNIFORMBLOCK_NO_CONT_MEMBERS); + _SG_VALIDATE(u_desc->glsl_name, VALIDATE_SHADERDESC_UNIFORMBLOCK_UNIFORM_GLSL_NAME); + const int array_count = u_desc->array_count; + _SG_VALIDATE(array_count > 0, VALIDATE_SHADERDESC_UNIFORMBLOCK_ARRAY_COUNT); + const uint32_t u_align = _sg_uniform_alignment(u_desc->type, array_count, ub_desc->layout); + const uint32_t u_size = _sg_uniform_size(u_desc->type, array_count, ub_desc->layout); + uniform_offset = _sg_align_u32(uniform_offset, u_align); + uniform_offset += u_size; + num_uniforms++; + // with std140, arrays are only allowed for FLOAT4, INT4, MAT4 + if (ub_desc->layout == SG_UNIFORMLAYOUT_STD140) { + if (array_count > 1) { + _SG_VALIDATE((u_desc->type == SG_UNIFORMTYPE_FLOAT4) || (u_desc->type == SG_UNIFORMTYPE_INT4) || (u_desc->type == SG_UNIFORMTYPE_MAT4), VALIDATE_SHADERDESC_UNIFORMBLOCK_STD140_ARRAY_TYPE); + } + } + } else { + uniforms_continuous = false; + } + } + if (ub_desc->layout == SG_UNIFORMLAYOUT_STD140) { + uniform_offset = _sg_align_u32(uniform_offset, 16); + } + _SG_VALIDATE((size_t)uniform_offset == ub_desc->size, VALIDATE_SHADERDESC_UNIFORMBLOCK_SIZE_MISMATCH); + _SG_VALIDATE(num_uniforms > 0, VALIDATE_SHADERDESC_UNIFORMBLOCK_NO_MEMBERS); + #endif + } + + for (size_t sbuf_idx = 0; sbuf_idx < SG_MAX_STORAGEBUFFER_BINDSLOTS; sbuf_idx++) { + const sg_shader_storage_buffer* sbuf_desc = &desc->storage_buffers[sbuf_idx]; + if (sbuf_desc->stage == SG_SHADERSTAGE_NONE) { + continue; + } + #if defined(SOKOL_METAL) + _SG_VALIDATE((sbuf_desc->msl_buffer_n >= _SG_MTL_MAX_STAGE_UB_BINDINGS) && (sbuf_desc->msl_buffer_n < _SG_MTL_MAX_STAGE_UB_SBUF_BINDINGS), VALIDATE_SHADERDESC_STORAGEBUFFER_METAL_BUFFER_SLOT_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(msl_buf_bits, sbuf_desc->stage, sbuf_desc->msl_buffer_n), VALIDATE_SHADERDESC_STORAGEBUFFER_METAL_BUFFER_SLOT_COLLISION); + msl_buf_bits = _sg_validate_set_slot_bit(msl_buf_bits, sbuf_desc->stage, sbuf_desc->msl_buffer_n); + #elif defined(SOKOL_D3D11) + if (sbuf_desc->readonly) { + _SG_VALIDATE(sbuf_desc->hlsl_register_t_n < _SG_D3D11_MAX_STAGE_SRV_BINDINGS, VALIDATE_SHADERDESC_STORAGEBUFFER_HLSL_REGISTER_T_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(hlsl_srv_bits, sbuf_desc->stage, sbuf_desc->hlsl_register_t_n), VALIDATE_SHADERDESC_STORAGEBUFFER_HLSL_REGISTER_T_COLLISION); + hlsl_srv_bits = _sg_validate_set_slot_bit(hlsl_srv_bits, sbuf_desc->stage, sbuf_desc->hlsl_register_t_n); + } else { + _SG_VALIDATE(sbuf_desc->hlsl_register_u_n < _SG_D3D11_MAX_STAGE_UAV_BINDINGS, VALIDATE_SHADERDESC_STORAGEBUFFER_HLSL_REGISTER_U_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(hlsl_uav_bits, sbuf_desc->stage, sbuf_desc->hlsl_register_u_n), VALIDATE_SHADERDESC_STORAGEBUFFER_HLSL_REGISTER_U_COLLISION); + hlsl_uav_bits = _sg_validate_set_slot_bit(hlsl_uav_bits, sbuf_desc->stage, sbuf_desc->hlsl_register_u_n); + } + #elif defined(_SOKOL_ANY_GL) + _SG_VALIDATE(sbuf_desc->glsl_binding_n < _SG_GL_MAX_SBUF_BINDINGS, VALIDATE_SHADERDESC_STORAGEBUFFER_GLSL_BINDING_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(glsl_bnd_bits, SG_SHADERSTAGE_NONE, sbuf_desc->glsl_binding_n), VALIDATE_SHADERDESC_STORAGEBUFFER_GLSL_BINDING_COLLISION); + glsl_bnd_bits = _sg_validate_set_slot_bit(glsl_bnd_bits, SG_SHADERSTAGE_NONE, sbuf_desc->glsl_binding_n); + #elif defined(SOKOL_WGPU) + _SG_VALIDATE(sbuf_desc->wgsl_group1_binding_n < _SG_WGPU_MAX_IMG_SMP_SBUF_BIND_SLOTS, VALIDATE_SHADERDESC_STORAGEBUFFER_WGSL_GROUP1_BINDING_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(wgsl_group1_bits, SG_SHADERSTAGE_NONE, sbuf_desc->wgsl_group1_binding_n), VALIDATE_SHADERDESC_STORAGEBUFFER_WGSL_GROUP1_BINDING_COLLISION); + wgsl_group1_bits = _sg_validate_set_slot_bit(wgsl_group1_bits, SG_SHADERSTAGE_NONE, sbuf_desc->wgsl_group1_binding_n); + #endif + } + + uint32_t img_slot_mask = 0; + for (size_t img_idx = 0; img_idx < SG_MAX_IMAGE_BINDSLOTS; img_idx++) { + const sg_shader_image* img_desc = &desc->images[img_idx]; + if (img_desc->stage == SG_SHADERSTAGE_NONE) { + continue; + } + img_slot_mask |= (1 << img_idx); + #if defined(SOKOL_METAL) + _SG_VALIDATE(img_desc->msl_texture_n < _SG_MTL_MAX_STAGE_IMAGE_BINDINGS, VALIDATE_SHADERDESC_IMAGE_METAL_TEXTURE_SLOT_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(msl_tex_bits, img_desc->stage, img_desc->msl_texture_n), VALIDATE_SHADERDESC_IMAGE_METAL_TEXTURE_SLOT_COLLISION); + msl_tex_bits = _sg_validate_set_slot_bit(msl_tex_bits, img_desc->stage, img_desc->msl_texture_n); + #elif defined(SOKOL_D3D11) + _SG_VALIDATE(img_desc->hlsl_register_t_n < _SG_D3D11_MAX_STAGE_SRV_BINDINGS, VALIDATE_SHADERDESC_IMAGE_HLSL_REGISTER_T_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(hlsl_srv_bits, img_desc->stage, img_desc->hlsl_register_t_n), VALIDATE_SHADERDESC_IMAGE_HLSL_REGISTER_T_COLLISION); + hlsl_srv_bits = _sg_validate_set_slot_bit(hlsl_srv_bits, img_desc->stage, img_desc->hlsl_register_t_n); + #elif defined(SOKOL_WGPU) + _SG_VALIDATE(img_desc->wgsl_group1_binding_n < _SG_WGPU_MAX_IMG_SMP_SBUF_BIND_SLOTS, VALIDATE_SHADERDESC_IMAGE_WGSL_GROUP1_BINDING_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(wgsl_group1_bits, SG_SHADERSTAGE_NONE, img_desc->wgsl_group1_binding_n), VALIDATE_SHADERDESC_IMAGE_WGSL_GROUP1_BINDING_COLLISION); + wgsl_group1_bits = _sg_validate_set_slot_bit(wgsl_group1_bits, SG_SHADERSTAGE_NONE, img_desc->wgsl_group1_binding_n); + #endif + } + + uint32_t smp_slot_mask = 0; + for (size_t smp_idx = 0; smp_idx < SG_MAX_SAMPLER_BINDSLOTS; smp_idx++) { + const sg_shader_sampler* smp_desc = &desc->samplers[smp_idx]; + if (smp_desc->stage == SG_SHADERSTAGE_NONE) { + continue; + } + smp_slot_mask |= (1 << smp_idx); + #if defined(SOKOL_METAL) + _SG_VALIDATE(smp_desc->msl_sampler_n < _SG_MTL_MAX_STAGE_SAMPLER_BINDINGS, VALIDATE_SHADERDESC_SAMPLER_METAL_SAMPLER_SLOT_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(msl_smp_bits, smp_desc->stage, smp_desc->msl_sampler_n), VALIDATE_SHADERDESC_SAMPLER_METAL_SAMPLER_SLOT_COLLISION); + msl_smp_bits = _sg_validate_set_slot_bit(msl_smp_bits, smp_desc->stage, smp_desc->msl_sampler_n); + #elif defined(SOKOL_D3D11) + _SG_VALIDATE(smp_desc->hlsl_register_s_n < _SG_D3D11_MAX_STAGE_SMP_BINDINGS, VALIDATE_SHADERDESC_SAMPLER_HLSL_REGISTER_S_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(hlsl_smp_bits, smp_desc->stage, smp_desc->hlsl_register_s_n), VALIDATE_SHADERDESC_SAMPLER_HLSL_REGISTER_S_COLLISION); + hlsl_smp_bits = _sg_validate_set_slot_bit(hlsl_smp_bits, smp_desc->stage, smp_desc->hlsl_register_s_n); + #elif defined(SOKOL_WGPU) + _SG_VALIDATE(smp_desc->wgsl_group1_binding_n < _SG_WGPU_MAX_IMG_SMP_SBUF_BIND_SLOTS, VALIDATE_SHADERDESC_SAMPLER_WGSL_GROUP1_BINDING_OUT_OF_RANGE); + _SG_VALIDATE(_sg_validate_slot_bits(wgsl_group1_bits, SG_SHADERSTAGE_NONE, smp_desc->wgsl_group1_binding_n), VALIDATE_SHADERDESC_SAMPLER_WGSL_GROUP1_BINDING_COLLISION); + wgsl_group1_bits = _sg_validate_set_slot_bit(wgsl_group1_bits, SG_SHADERSTAGE_NONE, smp_desc->wgsl_group1_binding_n); + #endif + } + + uint32_t ref_img_slot_mask = 0; + uint32_t ref_smp_slot_mask = 0; + for (size_t img_smp_idx = 0; img_smp_idx < SG_MAX_IMAGE_SAMPLER_PAIRS; img_smp_idx++) { + const sg_shader_image_sampler_pair* img_smp_desc = &desc->image_sampler_pairs[img_smp_idx]; + if (img_smp_desc->stage == SG_SHADERSTAGE_NONE) { + continue; + } + #if defined(_SOKOL_ANY_GL) + _SG_VALIDATE(img_smp_desc->glsl_name != 0, VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_GLSL_NAME); + #endif + const bool img_slot_in_range = img_smp_desc->image_slot < SG_MAX_IMAGE_BINDSLOTS; + const bool smp_slot_in_range = img_smp_desc->sampler_slot < SG_MAX_SAMPLER_BINDSLOTS; + _SG_VALIDATE(img_slot_in_range, VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_IMAGE_SLOT_OUT_OF_RANGE); + _SG_VALIDATE(smp_slot_in_range, VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_SAMPLER_SLOT_OUT_OF_RANGE); + if (img_slot_in_range && smp_slot_in_range) { + ref_img_slot_mask |= 1 << img_smp_desc->image_slot; + ref_smp_slot_mask |= 1 << img_smp_desc->sampler_slot; + const sg_shader_image* img_desc = &desc->images[img_smp_desc->image_slot]; + const sg_shader_sampler* smp_desc = &desc->samplers[img_smp_desc->sampler_slot]; + _SG_VALIDATE(img_desc->stage == img_smp_desc->stage, VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_IMAGE_STAGE_MISMATCH); + _SG_VALIDATE(smp_desc->stage == img_smp_desc->stage, VALIDATE_SHADERDESC_IMAGE_SAMPLER_PAIR_SAMPLER_STAGE_MISMATCH); + const bool needs_nonfiltering = (img_desc->sample_type == SG_IMAGESAMPLETYPE_UINT) + || (img_desc->sample_type == SG_IMAGESAMPLETYPE_SINT) + || (img_desc->sample_type == SG_IMAGESAMPLETYPE_UNFILTERABLE_FLOAT); + const bool needs_comparison = img_desc->sample_type == SG_IMAGESAMPLETYPE_DEPTH; + if (needs_nonfiltering) { + _SG_VALIDATE(needs_nonfiltering && (smp_desc->sampler_type == SG_SAMPLERTYPE_NONFILTERING), VALIDATE_SHADERDESC_NONFILTERING_SAMPLER_REQUIRED); + } + if (needs_comparison) { + _SG_VALIDATE(needs_comparison && (smp_desc->sampler_type == SG_SAMPLERTYPE_COMPARISON), VALIDATE_SHADERDESC_COMPARISON_SAMPLER_REQUIRED); + } + } + } + // each image and sampler must be referenced by an image sampler + _SG_VALIDATE(img_slot_mask == ref_img_slot_mask, VALIDATE_SHADERDESC_IMAGE_NOT_REFERENCED_BY_IMAGE_SAMPLER_PAIRS); + _SG_VALIDATE(smp_slot_mask == ref_smp_slot_mask, VALIDATE_SHADERDESC_SAMPLER_NOT_REFERENCED_BY_IMAGE_SAMPLER_PAIRS); + + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_pipeline_desc(const sg_pipeline_desc* desc) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(desc); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(desc); + _sg_validate_begin(); + _SG_VALIDATE(desc->_start_canary == 0, VALIDATE_PIPELINEDESC_CANARY); + _SG_VALIDATE(desc->_end_canary == 0, VALIDATE_PIPELINEDESC_CANARY); + _SG_VALIDATE(desc->shader.id != SG_INVALID_ID, VALIDATE_PIPELINEDESC_SHADER); + const _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, desc->shader.id); + _SG_VALIDATE(0 != shd, VALIDATE_PIPELINEDESC_SHADER); + if (shd) { + _SG_VALIDATE(shd->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_PIPELINEDESC_SHADER); + if (desc->compute) { + _SG_VALIDATE(shd->cmn.is_compute, VALIDATE_PIPELINEDESC_COMPUTE_SHADER_EXPECTED); + } else { + _SG_VALIDATE(!shd->cmn.is_compute, VALIDATE_PIPELINEDESC_NO_COMPUTE_SHADER_EXPECTED); + bool attrs_cont = true; + for (size_t attr_index = 0; attr_index < SG_MAX_VERTEX_ATTRIBUTES; attr_index++) { + const sg_vertex_attr_state* a_state = &desc->layout.attrs[attr_index]; + if (a_state->format == SG_VERTEXFORMAT_INVALID) { + attrs_cont = false; + continue; + } + _SG_VALIDATE(attrs_cont, VALIDATE_PIPELINEDESC_NO_CONT_ATTRS); + SOKOL_ASSERT(a_state->buffer_index < SG_MAX_VERTEXBUFFER_BINDSLOTS); + // vertex format must match expected shader attribute base type (if provided) + if (shd->cmn.attrs[attr_index].base_type != SG_SHADERATTRBASETYPE_UNDEFINED) { + if (_sg_vertexformat_basetype(a_state->format) != shd->cmn.attrs[attr_index].base_type) { + _SG_VALIDATE(false, VALIDATE_PIPELINEDESC_ATTR_BASETYPE_MISMATCH); + _SG_LOGMSG(VALIDATE_PIPELINEDESC_ATTR_BASETYPE_MISMATCH, "attr format:"); + _SG_LOGMSG(VALIDATE_PIPELINEDESC_ATTR_BASETYPE_MISMATCH, _sg_vertexformat_to_string(a_state->format)); + _SG_LOGMSG(VALIDATE_PIPELINEDESC_ATTR_BASETYPE_MISMATCH, "shader attr base type:"); + _SG_LOGMSG(VALIDATE_PIPELINEDESC_ATTR_BASETYPE_MISMATCH, _sg_shaderattrbasetype_to_string(shd->cmn.attrs[attr_index].base_type)); + } + } + #if defined(SOKOL_D3D11) + // on D3D11, semantic names (and semantic indices) must be provided + _SG_VALIDATE(!_sg_strempty(&shd->d3d11.attrs[attr_index].sem_name), VALIDATE_PIPELINEDESC_ATTR_SEMANTICS); + #endif + } + // must only use readonly storage buffer bindings in render pipelines + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (shd->cmn.storage_buffers[i].stage != SG_SHADERSTAGE_NONE) { + _SG_VALIDATE(shd->cmn.storage_buffers[i].readonly, VALIDATE_PIPELINEDESC_SHADER_READONLY_STORAGEBUFFERS); + } + } + for (int buf_index = 0; buf_index < SG_MAX_VERTEXBUFFER_BINDSLOTS; buf_index++) { + const sg_vertex_buffer_layout_state* l_state = &desc->layout.buffers[buf_index]; + if (l_state->stride == 0) { + continue; + } + _SG_VALIDATE(_sg_multiple_u64((uint64_t)l_state->stride, 4), VALIDATE_PIPELINEDESC_LAYOUT_STRIDE4); + } + } + } + for (size_t color_index = 0; color_index < (size_t)desc->color_count; color_index++) { + const sg_blend_state* bs = &desc->colors[color_index].blend; + if ((bs->op_rgb == SG_BLENDOP_MIN) || (bs->op_rgb == SG_BLENDOP_MAX)) { + _SG_VALIDATE((bs->src_factor_rgb == SG_BLENDFACTOR_ONE) && (bs->dst_factor_rgb == SG_BLENDFACTOR_ONE), VALIDATE_PIPELINEDESC_BLENDOP_MINMAX_REQUIRES_BLENDFACTOR_ONE); + } + if ((bs->op_alpha == SG_BLENDOP_MIN) || (bs->op_alpha == SG_BLENDOP_MAX)) { + _SG_VALIDATE((bs->src_factor_alpha == SG_BLENDFACTOR_ONE) && (bs->dst_factor_alpha == SG_BLENDFACTOR_ONE), VALIDATE_PIPELINEDESC_BLENDOP_MINMAX_REQUIRES_BLENDFACTOR_ONE); + } + } + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_attachments_desc(const sg_attachments_desc* desc) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(desc); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(desc); + _sg_validate_begin(); + _SG_VALIDATE(desc->_start_canary == 0, VALIDATE_ATTACHMENTSDESC_CANARY); + _SG_VALIDATE(desc->_end_canary == 0, VALIDATE_ATTACHMENTSDESC_CANARY); + bool atts_cont = true; + int color_width = -1, color_height = -1, color_sample_count = -1; + bool has_color_atts = false; + for (int att_index = 0; att_index < SG_MAX_COLOR_ATTACHMENTS; att_index++) { + const sg_attachment_desc* att = &desc->colors[att_index]; + if (att->image.id == SG_INVALID_ID) { + atts_cont = false; + continue; + } + _SG_VALIDATE(atts_cont, VALIDATE_ATTACHMENTSDESC_NO_CONT_COLOR_ATTS); + has_color_atts = true; + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, att->image.id); + _SG_VALIDATE(img, VALIDATE_ATTACHMENTSDESC_IMAGE); + if (0 != img) { + _SG_VALIDATE(img->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_ATTACHMENTSDESC_IMAGE); + _SG_VALIDATE(img->cmn.render_target, VALIDATE_ATTACHMENTSDESC_IMAGE_NO_RT); + _SG_VALIDATE(att->mip_level < img->cmn.num_mipmaps, VALIDATE_ATTACHMENTSDESC_MIPLEVEL); + if (img->cmn.type == SG_IMAGETYPE_CUBE) { + _SG_VALIDATE(att->slice < 6, VALIDATE_ATTACHMENTSDESC_FACE); + } else if (img->cmn.type == SG_IMAGETYPE_ARRAY) { + _SG_VALIDATE(att->slice < img->cmn.num_slices, VALIDATE_ATTACHMENTSDESC_LAYER); + } else if (img->cmn.type == SG_IMAGETYPE_3D) { + _SG_VALIDATE(att->slice < img->cmn.num_slices, VALIDATE_ATTACHMENTSDESC_SLICE); + } + if (att_index == 0) { + color_width = _sg_miplevel_dim(img->cmn.width, att->mip_level); + color_height = _sg_miplevel_dim(img->cmn.height, att->mip_level); + color_sample_count = img->cmn.sample_count; + } else { + _SG_VALIDATE(color_width == _sg_miplevel_dim(img->cmn.width, att->mip_level), VALIDATE_ATTACHMENTSDESC_IMAGE_SIZES); + _SG_VALIDATE(color_height == _sg_miplevel_dim(img->cmn.height, att->mip_level), VALIDATE_ATTACHMENTSDESC_IMAGE_SIZES); + _SG_VALIDATE(color_sample_count == img->cmn.sample_count, VALIDATE_ATTACHMENTSDESC_IMAGE_SAMPLE_COUNTS); + } + _SG_VALIDATE(_sg_is_valid_rendertarget_color_format(img->cmn.pixel_format), VALIDATE_ATTACHMENTSDESC_COLOR_INV_PIXELFORMAT); + + // check resolve attachment + const sg_attachment_desc* res_att = &desc->resolves[att_index]; + if (res_att->image.id != SG_INVALID_ID) { + // associated color attachment must be MSAA + _SG_VALIDATE(img->cmn.sample_count > 1, VALIDATE_ATTACHMENTSDESC_RESOLVE_COLOR_IMAGE_MSAA); + const _sg_image_t* res_img = _sg_lookup_image(&_sg.pools, res_att->image.id); + _SG_VALIDATE(res_img, VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE); + if (res_img != 0) { + _SG_VALIDATE(res_img->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE); + _SG_VALIDATE(res_img->cmn.render_target, VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE_NO_RT); + _SG_VALIDATE(res_img->cmn.sample_count == 1, VALIDATE_ATTACHMENTSDESC_RESOLVE_SAMPLE_COUNT); + _SG_VALIDATE(res_att->mip_level < res_img->cmn.num_mipmaps, VALIDATE_ATTACHMENTSDESC_RESOLVE_MIPLEVEL); + if (res_img->cmn.type == SG_IMAGETYPE_CUBE) { + _SG_VALIDATE(res_att->slice < 6, VALIDATE_ATTACHMENTSDESC_RESOLVE_FACE); + } else if (res_img->cmn.type == SG_IMAGETYPE_ARRAY) { + _SG_VALIDATE(res_att->slice < res_img->cmn.num_slices, VALIDATE_ATTACHMENTSDESC_RESOLVE_LAYER); + } else if (res_img->cmn.type == SG_IMAGETYPE_3D) { + _SG_VALIDATE(res_att->slice < res_img->cmn.num_slices, VALIDATE_ATTACHMENTSDESC_RESOLVE_SLICE); + } + _SG_VALIDATE(img->cmn.pixel_format == res_img->cmn.pixel_format, VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE_FORMAT); + _SG_VALIDATE(color_width == _sg_miplevel_dim(res_img->cmn.width, res_att->mip_level), VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE_SIZES); + _SG_VALIDATE(color_height == _sg_miplevel_dim(res_img->cmn.height, res_att->mip_level), VALIDATE_ATTACHMENTSDESC_RESOLVE_IMAGE_SIZES); + } + } + } + } + bool has_depth_stencil_att = false; + if (desc->depth_stencil.image.id != SG_INVALID_ID) { + const sg_attachment_desc* att = &desc->depth_stencil; + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, att->image.id); + _SG_VALIDATE(img, VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE); + has_depth_stencil_att = true; + if (img) { + _SG_VALIDATE(img->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE); + _SG_VALIDATE(att->mip_level < img->cmn.num_mipmaps, VALIDATE_ATTACHMENTSDESC_DEPTH_MIPLEVEL); + if (img->cmn.type == SG_IMAGETYPE_CUBE) { + _SG_VALIDATE(att->slice < 6, VALIDATE_ATTACHMENTSDESC_DEPTH_FACE); + } else if (img->cmn.type == SG_IMAGETYPE_ARRAY) { + _SG_VALIDATE(att->slice < img->cmn.num_slices, VALIDATE_ATTACHMENTSDESC_DEPTH_LAYER); + } else if (img->cmn.type == SG_IMAGETYPE_3D) { + // NOTE: this can't actually happen because of VALIDATE_IMAGEDESC_DEPTH_3D_IMAGE + _SG_VALIDATE(att->slice < img->cmn.num_slices, VALIDATE_ATTACHMENTSDESC_DEPTH_SLICE); + } + _SG_VALIDATE(img->cmn.render_target, VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE_NO_RT); + _SG_VALIDATE((color_width == -1) || (color_width == _sg_miplevel_dim(img->cmn.width, att->mip_level)), VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE_SIZES); + _SG_VALIDATE((color_height == -1) || (color_height == _sg_miplevel_dim(img->cmn.height, att->mip_level)), VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE_SIZES); + _SG_VALIDATE((color_sample_count == -1) || (color_sample_count == img->cmn.sample_count), VALIDATE_ATTACHMENTSDESC_DEPTH_IMAGE_SAMPLE_COUNT); + _SG_VALIDATE(_sg_is_valid_rendertarget_depth_format(img->cmn.pixel_format), VALIDATE_ATTACHMENTSDESC_DEPTH_INV_PIXELFORMAT); + } + } + _SG_VALIDATE(has_color_atts || has_depth_stencil_att, VALIDATE_ATTACHMENTSDESC_NO_ATTACHMENTS); + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_begin_pass(const sg_pass* pass) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(pass); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + const bool is_compute_pass = pass->compute; + const bool is_swapchain_pass = !is_compute_pass && (pass->attachments.id == SG_INVALID_ID); + const bool is_offscreen_pass = !(is_compute_pass || is_swapchain_pass); + _sg_validate_begin(); + _SG_VALIDATE(pass->_start_canary == 0, VALIDATE_BEGINPASS_CANARY); + _SG_VALIDATE(pass->_end_canary == 0, VALIDATE_BEGINPASS_CANARY); + if (is_compute_pass) { + // this is a compute pass + _SG_VALIDATE(pass->attachments.id == SG_INVALID_ID, VALIDATE_BEGINPASS_EXPECT_NO_ATTACHMENTS); + } else if (is_swapchain_pass) { + // this is a swapchain pass + _SG_VALIDATE(pass->swapchain.width > 0, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_WIDTH); + _SG_VALIDATE(pass->swapchain.height > 0, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_HEIGHT); + _SG_VALIDATE(pass->swapchain.sample_count > 0, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_SAMPLECOUNT); + _SG_VALIDATE(pass->swapchain.color_format > SG_PIXELFORMAT_NONE, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_COLORFORMAT); + // NOTE: depth buffer is optional, so depth_format is allowed to be invalid + // NOTE: the GL framebuffer handle may actually be 0 + #if defined(SOKOL_METAL) + _SG_VALIDATE(pass->swapchain.metal.current_drawable != 0, VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_CURRENTDRAWABLE); + if (pass->swapchain.depth_format == SG_PIXELFORMAT_NONE) { + _SG_VALIDATE(pass->swapchain.metal.depth_stencil_texture == 0, VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_DEPTHSTENCILTEXTURE_NOTSET); + } else { + _SG_VALIDATE(pass->swapchain.metal.depth_stencil_texture != 0, VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_DEPTHSTENCILTEXTURE); + } + if (pass->swapchain.sample_count > 1) { + _SG_VALIDATE(pass->swapchain.metal.msaa_color_texture != 0, VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_MSAACOLORTEXTURE); + } else { + _SG_VALIDATE(pass->swapchain.metal.msaa_color_texture == 0, VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_MSAACOLORTEXTURE_NOTSET); + } + #elif defined(SOKOL_D3D11) + _SG_VALIDATE(pass->swapchain.d3d11.render_view != 0, VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RENDERVIEW); + if (pass->swapchain.depth_format == SG_PIXELFORMAT_NONE) { + _SG_VALIDATE(pass->swapchain.d3d11.depth_stencil_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_DEPTHSTENCILVIEW_NOTSET); + } else { + _SG_VALIDATE(pass->swapchain.d3d11.depth_stencil_view != 0, VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_DEPTHSTENCILVIEW); + } + if (pass->swapchain.sample_count > 1) { + _SG_VALIDATE(pass->swapchain.d3d11.resolve_view != 0, VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RESOLVEVIEW); + } else { + _SG_VALIDATE(pass->swapchain.d3d11.resolve_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RESOLVEVIEW_NOTSET); + } + #elif defined(SOKOL_WGPU) + _SG_VALIDATE(pass->swapchain.wgpu.render_view != 0, VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RENDERVIEW); + if (pass->swapchain.depth_format == SG_PIXELFORMAT_NONE) { + _SG_VALIDATE(pass->swapchain.wgpu.depth_stencil_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_DEPTHSTENCILVIEW_NOTSET); + } else { + _SG_VALIDATE(pass->swapchain.wgpu.depth_stencil_view != 0, VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_DEPTHSTENCILVIEW); + } + if (pass->swapchain.sample_count > 1) { + _SG_VALIDATE(pass->swapchain.wgpu.resolve_view != 0, VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RESOLVEVIEW); + } else { + _SG_VALIDATE(pass->swapchain.wgpu.resolve_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RESOLVEVIEW_NOTSET); + } + #endif + } else { + // this is an 'offscreen pass' + const _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, pass->attachments.id); + if (atts) { + _SG_VALIDATE(atts->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_BEGINPASS_ATTACHMENTS_VALID); + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + const _sg_attachment_common_t* color_att = &atts->cmn.colors[i]; + const _sg_image_t* color_img = _sg_attachments_color_image(atts, i); + if (color_img) { + _SG_VALIDATE(color_img->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_BEGINPASS_COLOR_ATTACHMENT_IMAGE); + _SG_VALIDATE(color_img->slot.id == color_att->image_id.id, VALIDATE_BEGINPASS_COLOR_ATTACHMENT_IMAGE); + } + const _sg_attachment_common_t* resolve_att = &atts->cmn.resolves[i]; + const _sg_image_t* resolve_img = _sg_attachments_resolve_image(atts, i); + if (resolve_img) { + _SG_VALIDATE(resolve_img->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_BEGINPASS_RESOLVE_ATTACHMENT_IMAGE); + _SG_VALIDATE(resolve_img->slot.id == resolve_att->image_id.id, VALIDATE_BEGINPASS_RESOLVE_ATTACHMENT_IMAGE); + } + } + const _sg_image_t* ds_img = _sg_attachments_ds_image(atts); + if (ds_img) { + const _sg_attachment_common_t* att = &atts->cmn.depth_stencil; + _SG_VALIDATE(ds_img->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_BEGINPASS_DEPTHSTENCIL_ATTACHMENT_IMAGE); + _SG_VALIDATE(ds_img->slot.id == att->image_id.id, VALIDATE_BEGINPASS_DEPTHSTENCIL_ATTACHMENT_IMAGE); + } + } else { + _SG_VALIDATE(atts != 0, VALIDATE_BEGINPASS_ATTACHMENTS_EXISTS); + } + } + if (is_compute_pass || is_offscreen_pass) { + _SG_VALIDATE(pass->swapchain.width == 0, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_WIDTH_NOTSET); + _SG_VALIDATE(pass->swapchain.height == 0, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_HEIGHT_NOTSET); + _SG_VALIDATE(pass->swapchain.sample_count == 0, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_SAMPLECOUNT_NOTSET); + _SG_VALIDATE(pass->swapchain.color_format == _SG_PIXELFORMAT_DEFAULT, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_COLORFORMAT_NOTSET); + _SG_VALIDATE(pass->swapchain.depth_format == _SG_PIXELFORMAT_DEFAULT, VALIDATE_BEGINPASS_SWAPCHAIN_EXPECT_DEPTHFORMAT_NOTSET); + #if defined(SOKOL_METAL) + _SG_VALIDATE(pass->swapchain.metal.current_drawable == 0, VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_CURRENTDRAWABLE_NOTSET); + _SG_VALIDATE(pass->swapchain.metal.depth_stencil_texture == 0, VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_DEPTHSTENCILTEXTURE_NOTSET); + _SG_VALIDATE(pass->swapchain.metal.msaa_color_texture == 0, VALIDATE_BEGINPASS_SWAPCHAIN_METAL_EXPECT_MSAACOLORTEXTURE_NOTSET); + #elif defined(SOKOL_D3D11) + _SG_VALIDATE(pass->swapchain.d3d11.render_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RENDERVIEW_NOTSET); + _SG_VALIDATE(pass->swapchain.d3d11.depth_stencil_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_DEPTHSTENCILVIEW_NOTSET); + _SG_VALIDATE(pass->swapchain.d3d11.resolve_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_D3D11_EXPECT_RESOLVEVIEW_NOTSET); + #elif defined(SOKOL_WGPU) + _SG_VALIDATE(pass->swapchain.wgpu.render_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RENDERVIEW_NOTSET); + _SG_VALIDATE(pass->swapchain.wgpu.depth_stencil_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_DEPTHSTENCILVIEW_NOTSET); + _SG_VALIDATE(pass->swapchain.wgpu.resolve_view == 0, VALIDATE_BEGINPASS_SWAPCHAIN_WGPU_EXPECT_RESOLVEVIEW_NOTSET); + #elif defined(_SOKOL_ANY_GL) + _SG_VALIDATE(pass->swapchain.gl.framebuffer == 0, VALIDATE_BEGINPASS_SWAPCHAIN_GL_EXPECT_FRAMEBUFFER_NOTSET); + #endif + } + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_apply_viewport(int x, int y, int width, int height, bool origin_top_left) { + _SOKOL_UNUSED(x); + _SOKOL_UNUSED(y); + _SOKOL_UNUSED(width); + _SOKOL_UNUSED(height); + _SOKOL_UNUSED(origin_top_left); + #if !defined(SOKOL_DEBUG) + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + _sg_validate_begin(); + _SG_VALIDATE(_sg.cur_pass.in_pass && !_sg.cur_pass.is_compute, VALIDATE_AVP_RENDERPASS_EXPECTED); + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_apply_scissor_rect(int x, int y, int width, int height, bool origin_top_left) { + _SOKOL_UNUSED(x); + _SOKOL_UNUSED(y); + _SOKOL_UNUSED(width); + _SOKOL_UNUSED(height); + _SOKOL_UNUSED(origin_top_left); + #if !defined(SOKOL_DEBUG) + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + _sg_validate_begin(); + _SG_VALIDATE(_sg.cur_pass.in_pass && !_sg.cur_pass.is_compute, VALIDATE_ASR_RENDERPASS_EXPECTED); + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_apply_pipeline(sg_pipeline pip_id) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(pip_id); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + _sg_validate_begin(); + // the pipeline object must be alive and valid + _SG_VALIDATE(pip_id.id != SG_INVALID_ID, VALIDATE_APIP_PIPELINE_VALID_ID); + const _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + _SG_VALIDATE(pip != 0, VALIDATE_APIP_PIPELINE_EXISTS); + if (!pip) { + return _sg_validate_end(); + } + _SG_VALIDATE(pip->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_APIP_PIPELINE_VALID); + // the pipeline's shader must be alive and valid + SOKOL_ASSERT(pip->shader); + _SG_VALIDATE(_sg.cur_pass.in_pass, VALIDATE_APIP_PASS_EXPECTED); + _SG_VALIDATE(pip->shader->slot.id == pip->cmn.shader_id.id, VALIDATE_APIP_SHADER_EXISTS); + _SG_VALIDATE(pip->shader->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_APIP_SHADER_VALID); + if (pip->cmn.is_compute) { + _SG_VALIDATE(_sg.cur_pass.is_compute, VALIDATE_APIP_COMPUTEPASS_EXPECTED); + } else { + _SG_VALIDATE(!_sg.cur_pass.is_compute, VALIDATE_APIP_RENDERPASS_EXPECTED); + // check that pipeline attributes match current pass attributes + if (_sg.cur_pass.atts_id.id != SG_INVALID_ID) { + // an offscreen pass + const _sg_attachments_t* atts = _sg.cur_pass.atts; + SOKOL_ASSERT(atts); + _SG_VALIDATE(atts->slot.id == _sg.cur_pass.atts_id.id, VALIDATE_APIP_CURPASS_ATTACHMENTS_EXISTS); + _SG_VALIDATE(atts->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_APIP_CURPASS_ATTACHMENTS_VALID); + + _SG_VALIDATE(pip->cmn.color_count == atts->cmn.num_colors, VALIDATE_APIP_ATT_COUNT); + for (int i = 0; i < pip->cmn.color_count; i++) { + const _sg_image_t* att_img = _sg_attachments_color_image(atts, i); + _SG_VALIDATE(pip->cmn.colors[i].pixel_format == att_img->cmn.pixel_format, VALIDATE_APIP_COLOR_FORMAT); + _SG_VALIDATE(pip->cmn.sample_count == att_img->cmn.sample_count, VALIDATE_APIP_SAMPLE_COUNT); + } + const _sg_image_t* att_dsimg = _sg_attachments_ds_image(atts); + if (att_dsimg) { + _SG_VALIDATE(pip->cmn.depth.pixel_format == att_dsimg->cmn.pixel_format, VALIDATE_APIP_DEPTH_FORMAT); + } else { + _SG_VALIDATE(pip->cmn.depth.pixel_format == SG_PIXELFORMAT_NONE, VALIDATE_APIP_DEPTH_FORMAT); + } + } else { + // default pass + _SG_VALIDATE(pip->cmn.color_count == 1, VALIDATE_APIP_ATT_COUNT); + _SG_VALIDATE(pip->cmn.colors[0].pixel_format == _sg.cur_pass.swapchain.color_fmt, VALIDATE_APIP_COLOR_FORMAT); + _SG_VALIDATE(pip->cmn.depth.pixel_format == _sg.cur_pass.swapchain.depth_fmt, VALIDATE_APIP_DEPTH_FORMAT); + _SG_VALIDATE(pip->cmn.sample_count == _sg.cur_pass.swapchain.sample_count, VALIDATE_APIP_SAMPLE_COUNT); + } + } + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_apply_bindings(const sg_bindings* bindings) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(bindings); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + _sg_validate_begin(); + + // must be called in a pass + _SG_VALIDATE(_sg.cur_pass.in_pass, VALIDATE_ABND_PASS_EXPECTED); + + // bindings must not be empty + bool has_any_bindings = bindings->index_buffer.id != SG_INVALID_ID; + if (!has_any_bindings) for (size_t i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + has_any_bindings |= bindings->vertex_buffers[i].id != SG_INVALID_ID; + } + if (!has_any_bindings) for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + has_any_bindings |= bindings->images[i].id != SG_INVALID_ID; + } + if (!has_any_bindings) for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + has_any_bindings |= bindings->samplers[i].id != SG_INVALID_ID; + } + if (!has_any_bindings) for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + has_any_bindings |= bindings->storage_buffers[i].id != SG_INVALID_ID; + } + _SG_VALIDATE(has_any_bindings, VALIDATE_ABND_EMPTY_BINDINGS); + + // a pipeline object must have been applied + _SG_VALIDATE(_sg.cur_pipeline.id != SG_INVALID_ID, VALIDATE_ABND_PIPELINE); + const _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, _sg.cur_pipeline.id); + _SG_VALIDATE(pip != 0, VALIDATE_ABND_PIPELINE_EXISTS); + if (!pip) { + return _sg_validate_end(); + } + _SG_VALIDATE(pip->slot.state == SG_RESOURCESTATE_VALID, VALIDATE_ABND_PIPELINE_VALID); + SOKOL_ASSERT(pip->shader && (pip->cmn.shader_id.id == pip->shader->slot.id)); + const _sg_shader_t* shd = pip->shader; + + if (_sg.cur_pass.is_compute) { + for (size_t i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + _SG_VALIDATE(bindings->vertex_buffers[i].id == SG_INVALID_ID, VALIDATE_ABND_COMPUTE_EXPECTED_NO_VBS); + } + } else { + // has expected vertex buffers, and vertex buffers still exist + for (size_t i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + if (pip->cmn.vertex_buffer_layout_active[i]) { + _SG_VALIDATE(bindings->vertex_buffers[i].id != SG_INVALID_ID, VALIDATE_ABND_EXPECTED_VB); + // buffers in vertex-buffer-slots must be of type SG_BUFFERTYPE_VERTEXBUFFER + if (bindings->vertex_buffers[i].id != SG_INVALID_ID) { + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, bindings->vertex_buffers[i].id); + _SG_VALIDATE(buf != 0, VALIDATE_ABND_VB_EXISTS); + if (buf && buf->slot.state == SG_RESOURCESTATE_VALID) { + _SG_VALIDATE(SG_BUFFERTYPE_VERTEXBUFFER == buf->cmn.type, VALIDATE_ABND_VB_TYPE); + _SG_VALIDATE(!buf->cmn.append_overflow, VALIDATE_ABND_VB_OVERFLOW); + } + } + } + } + } + + if (_sg.cur_pass.is_compute) { + _SG_VALIDATE(bindings->index_buffer.id == SG_INVALID_ID, VALIDATE_ABND_COMPUTE_EXPECTED_NO_IB); + } else { + // index buffer expected or not, and index buffer still exists + if (pip->cmn.index_type == SG_INDEXTYPE_NONE) { + // pipeline defines non-indexed rendering, but index buffer provided + _SG_VALIDATE(bindings->index_buffer.id == SG_INVALID_ID, VALIDATE_ABND_IB); + } else { + // pipeline defines indexed rendering, but no index buffer provided + _SG_VALIDATE(bindings->index_buffer.id != SG_INVALID_ID, VALIDATE_ABND_NO_IB); + } + if (bindings->index_buffer.id != SG_INVALID_ID) { + // buffer in index-buffer-slot must be of type SG_BUFFERTYPE_INDEXBUFFER + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, bindings->index_buffer.id); + _SG_VALIDATE(buf != 0, VALIDATE_ABND_IB_EXISTS); + if (buf && buf->slot.state == SG_RESOURCESTATE_VALID) { + _SG_VALIDATE(SG_BUFFERTYPE_INDEXBUFFER == buf->cmn.type, VALIDATE_ABND_IB_TYPE); + _SG_VALIDATE(!buf->cmn.append_overflow, VALIDATE_ABND_IB_OVERFLOW); + } + } + } + + // has expected images + for (size_t i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + if (shd->cmn.images[i].stage != SG_SHADERSTAGE_NONE) { + _SG_VALIDATE(bindings->images[i].id != SG_INVALID_ID, VALIDATE_ABND_EXPECTED_IMAGE_BINDING); + if (bindings->images[i].id != SG_INVALID_ID) { + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, bindings->images[i].id); + _SG_VALIDATE(img != 0, VALIDATE_ABND_IMG_EXISTS); + if (img && img->slot.state == SG_RESOURCESTATE_VALID) { + _SG_VALIDATE(img->cmn.type == shd->cmn.images[i].image_type, VALIDATE_ABND_IMAGE_TYPE_MISMATCH); + if (!_sg.features.msaa_image_bindings) { + _SG_VALIDATE(img->cmn.sample_count == 1, VALIDATE_ABND_IMAGE_MSAA); + } + if (shd->cmn.images[i].multisampled) { + _SG_VALIDATE(img->cmn.sample_count > 1, VALIDATE_ABND_EXPECTED_MULTISAMPLED_IMAGE); + } + const _sg_pixelformat_info_t* info = &_sg.formats[img->cmn.pixel_format]; + switch (shd->cmn.images[i].sample_type) { + case SG_IMAGESAMPLETYPE_FLOAT: + _SG_VALIDATE(info->filter, VALIDATE_ABND_EXPECTED_FILTERABLE_IMAGE); + break; + case SG_IMAGESAMPLETYPE_DEPTH: + _SG_VALIDATE(info->depth, VALIDATE_ABND_EXPECTED_DEPTH_IMAGE); + break; + default: + break; + } + } + } + } + } + + // has expected samplers + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + if (shd->cmn.samplers[i].stage != SG_SHADERSTAGE_NONE) { + _SG_VALIDATE(bindings->samplers[i].id != SG_INVALID_ID, VALIDATE_ABND_EXPECTED_SAMPLER_BINDING); + if (bindings->samplers[i].id != SG_INVALID_ID) { + const _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, bindings->samplers[i].id); + _SG_VALIDATE(smp != 0, VALIDATE_ABND_SMP_EXISTS); + if (smp) { + if (shd->cmn.samplers[i].sampler_type == SG_SAMPLERTYPE_COMPARISON) { + _SG_VALIDATE(smp->cmn.compare != SG_COMPAREFUNC_NEVER, VALIDATE_ABND_UNEXPECTED_SAMPLER_COMPARE_NEVER); + } else { + _SG_VALIDATE(smp->cmn.compare == SG_COMPAREFUNC_NEVER, VALIDATE_ABND_EXPECTED_SAMPLER_COMPARE_NEVER); + } + if (shd->cmn.samplers[i].sampler_type == SG_SAMPLERTYPE_NONFILTERING) { + const bool nonfiltering = (smp->cmn.min_filter != SG_FILTER_LINEAR) + && (smp->cmn.mag_filter != SG_FILTER_LINEAR) + && (smp->cmn.mipmap_filter != SG_FILTER_LINEAR); + _SG_VALIDATE(nonfiltering, VALIDATE_ABND_EXPECTED_NONFILTERING_SAMPLER); + } + } + } + } + } + + // has expected storage buffers + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (shd->cmn.storage_buffers[i].stage != SG_SHADERSTAGE_NONE) { + _SG_VALIDATE(bindings->storage_buffers[i].id != SG_INVALID_ID, VALIDATE_ABND_EXPECTED_STORAGEBUFFER_BINDING); + if (bindings->storage_buffers[i].id != SG_INVALID_ID) { + const _sg_buffer_t* sbuf = _sg_lookup_buffer(&_sg.pools, bindings->storage_buffers[i].id); + _SG_VALIDATE(sbuf != 0, VALIDATE_ABND_STORAGEBUFFER_EXISTS); + if (sbuf) { + _SG_VALIDATE(sbuf->cmn.type == SG_BUFFERTYPE_STORAGEBUFFER, VALIDATE_ABND_STORAGEBUFFER_BINDING_BUFFERTYPE); + // read/write bindings are only allowed for immutable buffers + if (!shd->cmn.storage_buffers[i].readonly) { + _SG_VALIDATE(sbuf->cmn.usage == SG_USAGE_IMMUTABLE, VALIDATE_ABND_STORAGEBUFFER_READWRITE_IMMUTABLE); + } + } + } + } + } + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_apply_uniforms(int ub_slot, const sg_range* data) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(ub_slot); + _SOKOL_UNUSED(data); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT((ub_slot >= 0) && (ub_slot < SG_MAX_UNIFORMBLOCK_BINDSLOTS)); + _sg_validate_begin(); + _SG_VALIDATE(_sg.cur_pass.in_pass, VALIDATE_AU_PASS_EXPECTED); + _SG_VALIDATE(_sg.cur_pipeline.id != SG_INVALID_ID, VALIDATE_AU_NO_PIPELINE); + const _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, _sg.cur_pipeline.id); + SOKOL_ASSERT(pip && (pip->slot.id == _sg.cur_pipeline.id)); + SOKOL_ASSERT(pip->shader && (pip->shader->slot.id == pip->cmn.shader_id.id)); + + const _sg_shader_t* shd = pip->shader; + _SG_VALIDATE(shd->cmn.uniform_blocks[ub_slot].stage != SG_SHADERSTAGE_NONE, VALIDATE_AU_NO_UNIFORMBLOCK_AT_SLOT); + _SG_VALIDATE(data->size == shd->cmn.uniform_blocks[ub_slot].size, VALIDATE_AU_SIZE); + + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_draw(int base_element, int num_elements, int num_instances) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(base_element); + _SOKOL_UNUSED(num_elements); + _SOKOL_UNUSED(num_instances); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + _sg_validate_begin(); + _SG_VALIDATE(_sg.cur_pass.in_pass && !_sg.cur_pass.is_compute, VALIDATE_DRAW_RENDERPASS_EXPECTED); + _SG_VALIDATE(base_element >= 0, VALIDATE_DRAW_BASEELEMENT); + _SG_VALIDATE(num_elements >= 0, VALIDATE_DRAW_NUMELEMENTS); + _SG_VALIDATE(num_instances >= 0, VALIDATE_DRAW_NUMINSTANCES); + _SG_VALIDATE(_sg.required_bindings_and_uniforms == _sg.applied_bindings_and_uniforms, VALIDATE_DRAW_REQUIRED_BINDINGS_OR_UNIFORMS_MISSING); + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(num_groups_x); + _SOKOL_UNUSED(num_groups_y); + _SOKOL_UNUSED(num_groups_z); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + _sg_validate_begin(); + _SG_VALIDATE(_sg.cur_pass.in_pass && _sg.cur_pass.is_compute, VALIDATE_DISPATCH_COMPUTEPASS_EXPECTED); + _SG_VALIDATE((num_groups_x >= 0) && (num_groups_x < (1<<16)), VALIDATE_DISPATCH_NUMGROUPSX); + _SG_VALIDATE((num_groups_y >= 0) && (num_groups_y < (1<<16)), VALIDATE_DISPATCH_NUMGROUPSY); + _SG_VALIDATE((num_groups_z >= 0) && (num_groups_z < (1<<16)), VALIDATE_DISPATCH_NUMGROUPSZ); + _SG_VALIDATE(_sg.required_bindings_and_uniforms == _sg.applied_bindings_and_uniforms, VALIDATE_DRAW_REQUIRED_BINDINGS_OR_UNIFORMS_MISSING); + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_update_buffer(const _sg_buffer_t* buf, const sg_range* data) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(buf); + _SOKOL_UNUSED(data); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(buf && data && data->ptr); + _sg_validate_begin(); + _SG_VALIDATE(buf->cmn.usage != SG_USAGE_IMMUTABLE, VALIDATE_UPDATEBUF_USAGE); + _SG_VALIDATE(buf->cmn.size >= (int)data->size, VALIDATE_UPDATEBUF_SIZE); + _SG_VALIDATE(buf->cmn.update_frame_index != _sg.frame_index, VALIDATE_UPDATEBUF_ONCE); + _SG_VALIDATE(buf->cmn.append_frame_index != _sg.frame_index, VALIDATE_UPDATEBUF_APPEND); + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_append_buffer(const _sg_buffer_t* buf, const sg_range* data) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(buf); + _SOKOL_UNUSED(data); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(buf && data && data->ptr); + _sg_validate_begin(); + _SG_VALIDATE(buf->cmn.usage != SG_USAGE_IMMUTABLE, VALIDATE_APPENDBUF_USAGE); + _SG_VALIDATE(buf->cmn.size >= (buf->cmn.append_pos + (int)data->size), VALIDATE_APPENDBUF_SIZE); + _SG_VALIDATE(buf->cmn.update_frame_index != _sg.frame_index, VALIDATE_APPENDBUF_UPDATE); + return _sg_validate_end(); + #endif +} + +_SOKOL_PRIVATE bool _sg_validate_update_image(const _sg_image_t* img, const sg_image_data* data) { + #if !defined(SOKOL_DEBUG) + _SOKOL_UNUSED(img); + _SOKOL_UNUSED(data); + return true; + #else + if (_sg.desc.disable_validation) { + return true; + } + SOKOL_ASSERT(img && data); + _sg_validate_begin(); + _SG_VALIDATE(img->cmn.usage != SG_USAGE_IMMUTABLE, VALIDATE_UPDIMG_USAGE); + _SG_VALIDATE(img->cmn.upd_frame_index != _sg.frame_index, VALIDATE_UPDIMG_ONCE); + _sg_validate_image_data(data, + img->cmn.pixel_format, + img->cmn.width, + img->cmn.height, + (img->cmn.type == SG_IMAGETYPE_CUBE) ? 6 : 1, + img->cmn.num_mipmaps, + img->cmn.num_slices); + return _sg_validate_end(); + #endif +} + +// ██████ ███████ ███████ ██████ ██ ██ ██████ ██████ ███████ ███████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██████ █████ ███████ ██ ██ ██ ██ ██████ ██ █████ ███████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██ ██ ███████ ███████ ██████ ██████ ██ ██ ██████ ███████ ███████ +// +// >>resources +_SOKOL_PRIVATE sg_buffer_desc _sg_buffer_desc_defaults(const sg_buffer_desc* desc) { + sg_buffer_desc def = *desc; + def.type = _sg_def(def.type, SG_BUFFERTYPE_VERTEXBUFFER); + def.usage = _sg_def(def.usage, SG_USAGE_IMMUTABLE); + if (def.size == 0) { + def.size = def.data.size; + } + return def; +} + +_SOKOL_PRIVATE sg_image_desc _sg_image_desc_defaults(const sg_image_desc* desc) { + sg_image_desc def = *desc; + def.type = _sg_def(def.type, SG_IMAGETYPE_2D); + def.num_slices = _sg_def(def.num_slices, 1); + def.num_mipmaps = _sg_def(def.num_mipmaps, 1); + def.usage = _sg_def(def.usage, SG_USAGE_IMMUTABLE); + if (desc->render_target) { + def.pixel_format = _sg_def(def.pixel_format, _sg.desc.environment.defaults.color_format); + def.sample_count = _sg_def(def.sample_count, _sg.desc.environment.defaults.sample_count); + } else { + def.pixel_format = _sg_def(def.pixel_format, SG_PIXELFORMAT_RGBA8); + def.sample_count = _sg_def(def.sample_count, 1); + } + return def; +} + +_SOKOL_PRIVATE sg_sampler_desc _sg_sampler_desc_defaults(const sg_sampler_desc* desc) { + sg_sampler_desc def = *desc; + def.min_filter = _sg_def(def.min_filter, SG_FILTER_NEAREST); + def.mag_filter = _sg_def(def.mag_filter, SG_FILTER_NEAREST); + def.mipmap_filter = _sg_def(def.mipmap_filter, SG_FILTER_NEAREST); + def.wrap_u = _sg_def(def.wrap_u, SG_WRAP_REPEAT); + def.wrap_v = _sg_def(def.wrap_v, SG_WRAP_REPEAT); + def.wrap_w = _sg_def(def.wrap_w, SG_WRAP_REPEAT); + def.max_lod = _sg_def_flt(def.max_lod, FLT_MAX); + def.border_color = _sg_def(def.border_color, SG_BORDERCOLOR_OPAQUE_BLACK); + def.compare = _sg_def(def.compare, SG_COMPAREFUNC_NEVER); + def.max_anisotropy = _sg_def(def.max_anisotropy, 1); + return def; +} + +_SOKOL_PRIVATE sg_shader_desc _sg_shader_desc_defaults(const sg_shader_desc* desc) { + sg_shader_desc def = *desc; + #if defined(SOKOL_METAL) + def.vertex_func.entry = _sg_def(def.vertex_func.entry, "_main"); + def.fragment_func.entry = _sg_def(def.fragment_func.entry, "_main"); + def.compute_func.entry = _sg_def(def.compute_func.entry, "_main"); + #else + def.vertex_func.entry = _sg_def(def.vertex_func.entry, "main"); + def.fragment_func.entry = _sg_def(def.fragment_func.entry, "main"); + def.compute_func.entry = _sg_def(def.compute_func.entry, "main"); + #endif + #if defined(SOKOL_D3D11) + if (def.vertex_func.source) { + def.vertex_func.d3d11_target = _sg_def(def.vertex_func.d3d11_target, "vs_4_0"); + } + if (def.fragment_func.source) { + def.fragment_func.d3d11_target = _sg_def(def.fragment_func.d3d11_target, "ps_4_0"); + } + if (def.compute_func.source) { + def.compute_func.d3d11_target = _sg_def(def.fragment_func.d3d11_target,"cs_5_0"); + } + #endif + def.mtl_threads_per_threadgroup.y = _sg_def(desc->mtl_threads_per_threadgroup.y, 1); + def.mtl_threads_per_threadgroup.z = _sg_def(desc->mtl_threads_per_threadgroup.z, 1); + for (size_t ub_index = 0; ub_index < SG_MAX_UNIFORMBLOCK_BINDSLOTS; ub_index++) { + sg_shader_uniform_block* ub_desc = &def.uniform_blocks[ub_index]; + if (ub_desc->stage != SG_SHADERSTAGE_NONE) { + ub_desc->layout = _sg_def(ub_desc->layout, SG_UNIFORMLAYOUT_NATIVE); + for (size_t u_index = 0; u_index < SG_MAX_UNIFORMBLOCK_MEMBERS; u_index++) { + sg_glsl_shader_uniform* u_desc = &ub_desc->glsl_uniforms[u_index]; + if (u_desc->type == SG_UNIFORMTYPE_INVALID) { + break; + } + u_desc->array_count = _sg_def(u_desc->array_count, 1); + } + } + } + for (size_t img_index = 0; img_index < SG_MAX_IMAGE_BINDSLOTS; img_index++) { + sg_shader_image* img_desc = &def.images[img_index]; + if (img_desc->stage != SG_SHADERSTAGE_NONE) { + img_desc->image_type = _sg_def(img_desc->image_type, SG_IMAGETYPE_2D); + img_desc->sample_type = _sg_def(img_desc->sample_type, SG_IMAGESAMPLETYPE_FLOAT); + } + } + for (size_t smp_index = 0; smp_index < SG_MAX_SAMPLER_BINDSLOTS; smp_index++) { + sg_shader_sampler* smp_desc = &def.samplers[smp_index]; + if (smp_desc->stage != SG_SHADERSTAGE_NONE) { + smp_desc->sampler_type = _sg_def(smp_desc->sampler_type, SG_SAMPLERTYPE_FILTERING); + } + } + return def; +} + +_SOKOL_PRIVATE sg_pipeline_desc _sg_pipeline_desc_defaults(const sg_pipeline_desc* desc) { + sg_pipeline_desc def = *desc; + + // FIXME: should we actually do all this stuff for a compute pipeline? + + def.primitive_type = _sg_def(def.primitive_type, SG_PRIMITIVETYPE_TRIANGLES); + def.index_type = _sg_def(def.index_type, SG_INDEXTYPE_NONE); + def.cull_mode = _sg_def(def.cull_mode, SG_CULLMODE_NONE); + def.face_winding = _sg_def(def.face_winding, SG_FACEWINDING_CW); + def.sample_count = _sg_def(def.sample_count, _sg.desc.environment.defaults.sample_count); + + def.stencil.front.compare = _sg_def(def.stencil.front.compare, SG_COMPAREFUNC_ALWAYS); + def.stencil.front.fail_op = _sg_def(def.stencil.front.fail_op, SG_STENCILOP_KEEP); + def.stencil.front.depth_fail_op = _sg_def(def.stencil.front.depth_fail_op, SG_STENCILOP_KEEP); + def.stencil.front.pass_op = _sg_def(def.stencil.front.pass_op, SG_STENCILOP_KEEP); + def.stencil.back.compare = _sg_def(def.stencil.back.compare, SG_COMPAREFUNC_ALWAYS); + def.stencil.back.fail_op = _sg_def(def.stencil.back.fail_op, SG_STENCILOP_KEEP); + def.stencil.back.depth_fail_op = _sg_def(def.stencil.back.depth_fail_op, SG_STENCILOP_KEEP); + def.stencil.back.pass_op = _sg_def(def.stencil.back.pass_op, SG_STENCILOP_KEEP); + + def.depth.compare = _sg_def(def.depth.compare, SG_COMPAREFUNC_ALWAYS); + def.depth.pixel_format = _sg_def(def.depth.pixel_format, _sg.desc.environment.defaults.depth_format); + if (def.colors[0].pixel_format == SG_PIXELFORMAT_NONE) { + // special case depth-only rendering, enforce a color count of 0 + def.color_count = 0; + } else { + def.color_count = _sg_def(def.color_count, 1); + } + if (def.color_count > SG_MAX_COLOR_ATTACHMENTS) { + def.color_count = SG_MAX_COLOR_ATTACHMENTS; + } + for (int i = 0; i < def.color_count; i++) { + sg_color_target_state* cs = &def.colors[i]; + cs->pixel_format = _sg_def(cs->pixel_format, _sg.desc.environment.defaults.color_format); + cs->write_mask = _sg_def(cs->write_mask, SG_COLORMASK_RGBA); + sg_blend_state* bs = &def.colors[i].blend; + bs->op_rgb = _sg_def(bs->op_rgb, SG_BLENDOP_ADD); + bs->src_factor_rgb = _sg_def(bs->src_factor_rgb, SG_BLENDFACTOR_ONE); + if ((bs->op_rgb == SG_BLENDOP_MIN) || (bs->op_rgb == SG_BLENDOP_MAX)) { + bs->dst_factor_rgb = _sg_def(bs->dst_factor_rgb, SG_BLENDFACTOR_ONE); + } else { + bs->dst_factor_rgb = _sg_def(bs->dst_factor_rgb, SG_BLENDFACTOR_ZERO); + } + bs->op_alpha = _sg_def(bs->op_alpha, SG_BLENDOP_ADD); + bs->src_factor_alpha = _sg_def(bs->src_factor_alpha, SG_BLENDFACTOR_ONE); + if ((bs->op_alpha == SG_BLENDOP_MIN) || (bs->op_alpha == SG_BLENDOP_MAX)) { + bs->dst_factor_alpha = _sg_def(bs->dst_factor_alpha, SG_BLENDFACTOR_ONE); + } else { + bs->dst_factor_alpha = _sg_def(bs->dst_factor_alpha, SG_BLENDFACTOR_ZERO); + } + } + + for (int attr_index = 0; attr_index < SG_MAX_VERTEX_ATTRIBUTES; attr_index++) { + sg_vertex_attr_state* a_state = &def.layout.attrs[attr_index]; + if (a_state->format == SG_VERTEXFORMAT_INVALID) { + break; + } + SOKOL_ASSERT(a_state->buffer_index < SG_MAX_VERTEXBUFFER_BINDSLOTS); + sg_vertex_buffer_layout_state* l_state = &def.layout.buffers[a_state->buffer_index]; + l_state->step_func = _sg_def(l_state->step_func, SG_VERTEXSTEP_PER_VERTEX); + l_state->step_rate = _sg_def(l_state->step_rate, 1); + } + + // resolve vertex layout strides and offsets + int auto_offset[SG_MAX_VERTEXBUFFER_BINDSLOTS]; + _sg_clear(auto_offset, sizeof(auto_offset)); + bool use_auto_offset = true; + for (int attr_index = 0; attr_index < SG_MAX_VERTEX_ATTRIBUTES; attr_index++) { + // to use computed offsets, *all* attr offsets must be 0 + if (def.layout.attrs[attr_index].offset != 0) { + use_auto_offset = false; + } + } + for (int attr_index = 0; attr_index < SG_MAX_VERTEX_ATTRIBUTES; attr_index++) { + sg_vertex_attr_state* a_state = &def.layout.attrs[attr_index]; + if (a_state->format == SG_VERTEXFORMAT_INVALID) { + break; + } + SOKOL_ASSERT(a_state->buffer_index < SG_MAX_VERTEXBUFFER_BINDSLOTS); + if (use_auto_offset) { + a_state->offset = auto_offset[a_state->buffer_index]; + } + auto_offset[a_state->buffer_index] += _sg_vertexformat_bytesize(a_state->format); + } + // compute vertex strides if needed + for (int buf_index = 0; buf_index < SG_MAX_VERTEXBUFFER_BINDSLOTS; buf_index++) { + sg_vertex_buffer_layout_state* l_state = &def.layout.buffers[buf_index]; + if (l_state->stride == 0) { + l_state->stride = auto_offset[buf_index]; + } + } + + return def; +} + +_SOKOL_PRIVATE sg_attachments_desc _sg_attachments_desc_defaults(const sg_attachments_desc* desc) { + sg_attachments_desc def = *desc; + return def; +} + +_SOKOL_PRIVATE sg_buffer _sg_alloc_buffer(void) { + sg_buffer res; + int slot_index = _sg_pool_alloc_index(&_sg.pools.buffer_pool); + if (_SG_INVALID_SLOT_INDEX != slot_index) { + res.id = _sg_slot_alloc(&_sg.pools.buffer_pool, &_sg.pools.buffers[slot_index].slot, slot_index); + } else { + res.id = SG_INVALID_ID; + _SG_ERROR(BUFFER_POOL_EXHAUSTED); + } + return res; +} + +_SOKOL_PRIVATE sg_image _sg_alloc_image(void) { + sg_image res; + int slot_index = _sg_pool_alloc_index(&_sg.pools.image_pool); + if (_SG_INVALID_SLOT_INDEX != slot_index) { + res.id = _sg_slot_alloc(&_sg.pools.image_pool, &_sg.pools.images[slot_index].slot, slot_index); + } else { + res.id = SG_INVALID_ID; + _SG_ERROR(IMAGE_POOL_EXHAUSTED); + } + return res; +} + +_SOKOL_PRIVATE sg_sampler _sg_alloc_sampler(void) { + sg_sampler res; + int slot_index = _sg_pool_alloc_index(&_sg.pools.sampler_pool); + if (_SG_INVALID_SLOT_INDEX != slot_index) { + res.id = _sg_slot_alloc(&_sg.pools.sampler_pool, &_sg.pools.samplers[slot_index].slot, slot_index); + } else { + res.id = SG_INVALID_ID; + _SG_ERROR(SAMPLER_POOL_EXHAUSTED); + } + return res; +} + +_SOKOL_PRIVATE sg_shader _sg_alloc_shader(void) { + sg_shader res; + int slot_index = _sg_pool_alloc_index(&_sg.pools.shader_pool); + if (_SG_INVALID_SLOT_INDEX != slot_index) { + res.id = _sg_slot_alloc(&_sg.pools.shader_pool, &_sg.pools.shaders[slot_index].slot, slot_index); + } else { + res.id = SG_INVALID_ID; + _SG_ERROR(SHADER_POOL_EXHAUSTED); + } + return res; +} + +_SOKOL_PRIVATE sg_pipeline _sg_alloc_pipeline(void) { + sg_pipeline res; + int slot_index = _sg_pool_alloc_index(&_sg.pools.pipeline_pool); + if (_SG_INVALID_SLOT_INDEX != slot_index) { + res.id =_sg_slot_alloc(&_sg.pools.pipeline_pool, &_sg.pools.pipelines[slot_index].slot, slot_index); + } else { + res.id = SG_INVALID_ID; + _SG_ERROR(PIPELINE_POOL_EXHAUSTED); + } + return res; +} + +_SOKOL_PRIVATE sg_attachments _sg_alloc_attachments(void) { + sg_attachments res; + int slot_index = _sg_pool_alloc_index(&_sg.pools.attachments_pool); + if (_SG_INVALID_SLOT_INDEX != slot_index) { + res.id = _sg_slot_alloc(&_sg.pools.attachments_pool, &_sg.pools.attachments[slot_index].slot, slot_index); + } else { + res.id = SG_INVALID_ID; + _SG_ERROR(PASS_POOL_EXHAUSTED); + } + return res; +} + +_SOKOL_PRIVATE void _sg_dealloc_buffer(_sg_buffer_t* buf) { + SOKOL_ASSERT(buf && (buf->slot.state == SG_RESOURCESTATE_ALLOC) && (buf->slot.id != SG_INVALID_ID)); + _sg_pool_free_index(&_sg.pools.buffer_pool, _sg_slot_index(buf->slot.id)); + _sg_slot_reset(&buf->slot); +} + +_SOKOL_PRIVATE void _sg_dealloc_image(_sg_image_t* img) { + SOKOL_ASSERT(img && (img->slot.state == SG_RESOURCESTATE_ALLOC) && (img->slot.id != SG_INVALID_ID)); + _sg_pool_free_index(&_sg.pools.image_pool, _sg_slot_index(img->slot.id)); + _sg_slot_reset(&img->slot); +} + +_SOKOL_PRIVATE void _sg_dealloc_sampler(_sg_sampler_t* smp) { + SOKOL_ASSERT(smp && (smp->slot.state == SG_RESOURCESTATE_ALLOC) && (smp->slot.id != SG_INVALID_ID)); + _sg_pool_free_index(&_sg.pools.sampler_pool, _sg_slot_index(smp->slot.id)); + _sg_slot_reset(&smp->slot); +} + +_SOKOL_PRIVATE void _sg_dealloc_shader(_sg_shader_t* shd) { + SOKOL_ASSERT(shd && (shd->slot.state == SG_RESOURCESTATE_ALLOC) && (shd->slot.id != SG_INVALID_ID)); + _sg_pool_free_index(&_sg.pools.shader_pool, _sg_slot_index(shd->slot.id)); + _sg_slot_reset(&shd->slot); +} + +_SOKOL_PRIVATE void _sg_dealloc_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip && (pip->slot.state == SG_RESOURCESTATE_ALLOC) && (pip->slot.id != SG_INVALID_ID)); + _sg_pool_free_index(&_sg.pools.pipeline_pool, _sg_slot_index(pip->slot.id)); + _sg_slot_reset(&pip->slot); +} + +_SOKOL_PRIVATE void _sg_dealloc_attachments(_sg_attachments_t* atts) { + SOKOL_ASSERT(atts && (atts->slot.state == SG_RESOURCESTATE_ALLOC) && (atts->slot.id != SG_INVALID_ID)); + _sg_pool_free_index(&_sg.pools.attachments_pool, _sg_slot_index(atts->slot.id)); + _sg_slot_reset(&atts->slot); +} + +_SOKOL_PRIVATE void _sg_init_buffer(_sg_buffer_t* buf, const sg_buffer_desc* desc) { + SOKOL_ASSERT(buf && (buf->slot.state == SG_RESOURCESTATE_ALLOC)); + SOKOL_ASSERT(desc); + if (_sg_validate_buffer_desc(desc)) { + _sg_buffer_common_init(&buf->cmn, desc); + buf->slot.state = _sg_create_buffer(buf, desc); + } else { + buf->slot.state = SG_RESOURCESTATE_FAILED; + } + SOKOL_ASSERT((buf->slot.state == SG_RESOURCESTATE_VALID)||(buf->slot.state == SG_RESOURCESTATE_FAILED)); +} + +_SOKOL_PRIVATE void _sg_init_image(_sg_image_t* img, const sg_image_desc* desc) { + SOKOL_ASSERT(img && (img->slot.state == SG_RESOURCESTATE_ALLOC)); + SOKOL_ASSERT(desc); + if (_sg_validate_image_desc(desc)) { + _sg_image_common_init(&img->cmn, desc); + img->slot.state = _sg_create_image(img, desc); + } else { + img->slot.state = SG_RESOURCESTATE_FAILED; + } + SOKOL_ASSERT((img->slot.state == SG_RESOURCESTATE_VALID)||(img->slot.state == SG_RESOURCESTATE_FAILED)); +} + +_SOKOL_PRIVATE void _sg_init_sampler(_sg_sampler_t* smp, const sg_sampler_desc* desc) { + SOKOL_ASSERT(smp && (smp->slot.state == SG_RESOURCESTATE_ALLOC)); + SOKOL_ASSERT(desc); + if (_sg_validate_sampler_desc(desc)) { + _sg_sampler_common_init(&smp->cmn, desc); + smp->slot.state = _sg_create_sampler(smp, desc); + } else { + smp->slot.state = SG_RESOURCESTATE_FAILED; + } + SOKOL_ASSERT((smp->slot.state == SG_RESOURCESTATE_VALID)||(smp->slot.state == SG_RESOURCESTATE_FAILED)); +} + +_SOKOL_PRIVATE void _sg_init_shader(_sg_shader_t* shd, const sg_shader_desc* desc) { + SOKOL_ASSERT(shd && (shd->slot.state == SG_RESOURCESTATE_ALLOC)); + SOKOL_ASSERT(desc); + if (_sg_validate_shader_desc(desc)) { + _sg_shader_common_init(&shd->cmn, desc); + shd->slot.state = _sg_create_shader(shd, desc); + } else { + shd->slot.state = SG_RESOURCESTATE_FAILED; + } + SOKOL_ASSERT((shd->slot.state == SG_RESOURCESTATE_VALID)||(shd->slot.state == SG_RESOURCESTATE_FAILED)); +} + +_SOKOL_PRIVATE void _sg_init_pipeline(_sg_pipeline_t* pip, const sg_pipeline_desc* desc) { + SOKOL_ASSERT(pip && (pip->slot.state == SG_RESOURCESTATE_ALLOC)); + SOKOL_ASSERT(desc); + if (_sg_validate_pipeline_desc(desc)) { + _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, desc->shader.id); + if (shd && (shd->slot.state == SG_RESOURCESTATE_VALID)) { + _sg_pipeline_common_init(&pip->cmn, desc); + pip->slot.state = _sg_create_pipeline(pip, shd, desc); + } else { + pip->slot.state = SG_RESOURCESTATE_FAILED; + } + } else { + pip->slot.state = SG_RESOURCESTATE_FAILED; + } + SOKOL_ASSERT((pip->slot.state == SG_RESOURCESTATE_VALID)||(pip->slot.state == SG_RESOURCESTATE_FAILED)); +} + +_SOKOL_PRIVATE void _sg_init_attachments(_sg_attachments_t* atts, const sg_attachments_desc* desc) { + SOKOL_ASSERT(atts && atts->slot.state == SG_RESOURCESTATE_ALLOC); + SOKOL_ASSERT(desc); + if (_sg_validate_attachments_desc(desc)) { + // lookup pass attachment image pointers + _sg_image_t* color_images[SG_MAX_COLOR_ATTACHMENTS] = { 0 }; + _sg_image_t* resolve_images[SG_MAX_COLOR_ATTACHMENTS] = { 0 }; + _sg_image_t* ds_image = 0; + // NOTE: validation already checked that all surfaces are same width/height + int width = 0; + int height = 0; + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + if (desc->colors[i].image.id) { + color_images[i] = _sg_lookup_image(&_sg.pools, desc->colors[i].image.id); + if (!(color_images[i] && color_images[i]->slot.state == SG_RESOURCESTATE_VALID)) { + atts->slot.state = SG_RESOURCESTATE_FAILED; + return; + } + const int mip_level = desc->colors[i].mip_level; + width = _sg_miplevel_dim(color_images[i]->cmn.width, mip_level); + height = _sg_miplevel_dim(color_images[i]->cmn.height, mip_level); + } + if (desc->resolves[i].image.id) { + resolve_images[i] = _sg_lookup_image(&_sg.pools, desc->resolves[i].image.id); + if (!(resolve_images[i] && resolve_images[i]->slot.state == SG_RESOURCESTATE_VALID)) { + atts->slot.state = SG_RESOURCESTATE_FAILED; + return; + } + } + } + if (desc->depth_stencil.image.id) { + ds_image = _sg_lookup_image(&_sg.pools, desc->depth_stencil.image.id); + if (!(ds_image && ds_image->slot.state == SG_RESOURCESTATE_VALID)) { + atts->slot.state = SG_RESOURCESTATE_FAILED; + return; + } + const int mip_level = desc->depth_stencil.mip_level; + width = _sg_miplevel_dim(ds_image->cmn.width, mip_level); + height = _sg_miplevel_dim(ds_image->cmn.height, mip_level); + } + _sg_attachments_common_init(&atts->cmn, desc, width, height); + atts->slot.state = _sg_create_attachments(atts, color_images, resolve_images, ds_image, desc); + } else { + atts->slot.state = SG_RESOURCESTATE_FAILED; + } + SOKOL_ASSERT((atts->slot.state == SG_RESOURCESTATE_VALID)||(atts->slot.state == SG_RESOURCESTATE_FAILED)); +} + +_SOKOL_PRIVATE void _sg_uninit_buffer(_sg_buffer_t* buf) { + SOKOL_ASSERT(buf && ((buf->slot.state == SG_RESOURCESTATE_VALID) || (buf->slot.state == SG_RESOURCESTATE_FAILED))); + _sg_discard_buffer(buf); + _sg_reset_buffer_to_alloc_state(buf); +} + +_SOKOL_PRIVATE void _sg_uninit_image(_sg_image_t* img) { + SOKOL_ASSERT(img && ((img->slot.state == SG_RESOURCESTATE_VALID) || (img->slot.state == SG_RESOURCESTATE_FAILED))); + _sg_discard_image(img); + _sg_reset_image_to_alloc_state(img); +} + +_SOKOL_PRIVATE void _sg_uninit_sampler(_sg_sampler_t* smp) { + SOKOL_ASSERT(smp && ((smp->slot.state == SG_RESOURCESTATE_VALID) || (smp->slot.state == SG_RESOURCESTATE_FAILED))); + _sg_discard_sampler(smp); + _sg_reset_sampler_to_alloc_state(smp); +} + +_SOKOL_PRIVATE void _sg_uninit_shader(_sg_shader_t* shd) { + SOKOL_ASSERT(shd && ((shd->slot.state == SG_RESOURCESTATE_VALID) || (shd->slot.state == SG_RESOURCESTATE_FAILED))); + _sg_discard_shader(shd); + _sg_reset_shader_to_alloc_state(shd); +} + +_SOKOL_PRIVATE void _sg_uninit_pipeline(_sg_pipeline_t* pip) { + SOKOL_ASSERT(pip && ((pip->slot.state == SG_RESOURCESTATE_VALID) || (pip->slot.state == SG_RESOURCESTATE_FAILED))); + _sg_discard_pipeline(pip); + _sg_reset_pipeline_to_alloc_state(pip); +} + +_SOKOL_PRIVATE void _sg_uninit_attachments(_sg_attachments_t* atts) { + SOKOL_ASSERT(atts && ((atts->slot.state == SG_RESOURCESTATE_VALID) || (atts->slot.state == SG_RESOURCESTATE_FAILED))); + _sg_discard_attachments(atts); + _sg_reset_attachments_to_alloc_state(atts); +} + +_SOKOL_PRIVATE void _sg_setup_commit_listeners(const sg_desc* desc) { + SOKOL_ASSERT(desc->max_commit_listeners > 0); + SOKOL_ASSERT(0 == _sg.commit_listeners.items); + SOKOL_ASSERT(0 == _sg.commit_listeners.num); + SOKOL_ASSERT(0 == _sg.commit_listeners.upper); + _sg.commit_listeners.num = desc->max_commit_listeners; + const size_t size = (size_t)_sg.commit_listeners.num * sizeof(sg_commit_listener); + _sg.commit_listeners.items = (sg_commit_listener*)_sg_malloc_clear(size); +} + +_SOKOL_PRIVATE void _sg_discard_commit_listeners(void) { + SOKOL_ASSERT(0 != _sg.commit_listeners.items); + _sg_free(_sg.commit_listeners.items); + _sg.commit_listeners.items = 0; +} + +_SOKOL_PRIVATE void _sg_notify_commit_listeners(void) { + SOKOL_ASSERT(_sg.commit_listeners.items); + for (int i = 0; i < _sg.commit_listeners.upper; i++) { + const sg_commit_listener* listener = &_sg.commit_listeners.items[i]; + if (listener->func) { + listener->func(listener->user_data); + } + } +} + +_SOKOL_PRIVATE bool _sg_add_commit_listener(const sg_commit_listener* new_listener) { + SOKOL_ASSERT(new_listener && new_listener->func); + SOKOL_ASSERT(_sg.commit_listeners.items); + // first check if the listener hadn't been added already + for (int i = 0; i < _sg.commit_listeners.upper; i++) { + const sg_commit_listener* slot = &_sg.commit_listeners.items[i]; + if ((slot->func == new_listener->func) && (slot->user_data == new_listener->user_data)) { + _SG_ERROR(IDENTICAL_COMMIT_LISTENER); + return false; + } + } + // first try to plug a hole + sg_commit_listener* slot = 0; + for (int i = 0; i < _sg.commit_listeners.upper; i++) { + if (_sg.commit_listeners.items[i].func == 0) { + slot = &_sg.commit_listeners.items[i]; + break; + } + } + if (!slot) { + // append to end + if (_sg.commit_listeners.upper < _sg.commit_listeners.num) { + slot = &_sg.commit_listeners.items[_sg.commit_listeners.upper++]; + } + } + if (!slot) { + _SG_ERROR(COMMIT_LISTENER_ARRAY_FULL); + return false; + } + *slot = *new_listener; + return true; +} + +_SOKOL_PRIVATE bool _sg_remove_commit_listener(const sg_commit_listener* listener) { + SOKOL_ASSERT(listener && listener->func); + SOKOL_ASSERT(_sg.commit_listeners.items); + for (int i = 0; i < _sg.commit_listeners.upper; i++) { + sg_commit_listener* slot = &_sg.commit_listeners.items[i]; + // both the function pointer and user data must match! + if ((slot->func == listener->func) && (slot->user_data == listener->user_data)) { + slot->func = 0; + slot->user_data = 0; + // NOTE: since _sg_add_commit_listener() already catches duplicates, + // we don't need to worry about them here + return true; + } + } + return false; +} + +_SOKOL_PRIVATE void _sg_setup_compute(const sg_desc* desc) { + SOKOL_ASSERT(desc && (desc->max_dispatch_calls_per_pass > 0)); + const uint32_t max_tracked_sbufs = (uint32_t)desc->max_dispatch_calls_per_pass * SG_MAX_STORAGEBUFFER_BINDSLOTS; + _sg_tracker_init(&_sg.compute.readwrite_sbufs, max_tracked_sbufs); +} + +_SOKOL_PRIVATE void _sg_discard_compute(void) { + _sg_tracker_discard(&_sg.compute.readwrite_sbufs); +} + +_SOKOL_PRIVATE void _sg_compute_pass_track_storage_buffer(_sg_buffer_t* sbuf, bool readonly) { + SOKOL_ASSERT(sbuf); + if (!readonly) { + _sg_tracker_add(&_sg.compute.readwrite_sbufs, sbuf->slot.id); + } +} + +_SOKOL_PRIVATE void _sg_compute_on_endpass(void) { + SOKOL_ASSERT(_sg.cur_pass.in_pass); + SOKOL_ASSERT(_sg.cur_pass.is_compute); + _sg_tracker_reset(&_sg.compute.readwrite_sbufs); +} + +_SOKOL_PRIVATE sg_desc _sg_desc_defaults(const sg_desc* desc) { + /* + NOTE: on WebGPU, the default color pixel format MUST be provided, + cannot be a default compile-time constant. + */ + sg_desc res = *desc; + #if defined(SOKOL_WGPU) + SOKOL_ASSERT(SG_PIXELFORMAT_NONE < res.environment.defaults.color_format); + #elif defined(SOKOL_METAL) || defined(SOKOL_D3D11) + res.environment.defaults.color_format = _sg_def(res.environment.defaults.color_format, SG_PIXELFORMAT_BGRA8); + #else + res.environment.defaults.color_format = _sg_def(res.environment.defaults.color_format, SG_PIXELFORMAT_RGBA8); + #endif + res.environment.defaults.depth_format = _sg_def(res.environment.defaults.depth_format, SG_PIXELFORMAT_DEPTH_STENCIL); + res.environment.defaults.sample_count = _sg_def(res.environment.defaults.sample_count, 1); + res.buffer_pool_size = _sg_def(res.buffer_pool_size, _SG_DEFAULT_BUFFER_POOL_SIZE); + res.image_pool_size = _sg_def(res.image_pool_size, _SG_DEFAULT_IMAGE_POOL_SIZE); + res.sampler_pool_size = _sg_def(res.sampler_pool_size, _SG_DEFAULT_SAMPLER_POOL_SIZE); + res.shader_pool_size = _sg_def(res.shader_pool_size, _SG_DEFAULT_SHADER_POOL_SIZE); + res.pipeline_pool_size = _sg_def(res.pipeline_pool_size, _SG_DEFAULT_PIPELINE_POOL_SIZE); + res.attachments_pool_size = _sg_def(res.attachments_pool_size, _SG_DEFAULT_ATTACHMENTS_POOL_SIZE); + res.uniform_buffer_size = _sg_def(res.uniform_buffer_size, _SG_DEFAULT_UB_SIZE); + res.max_dispatch_calls_per_pass = _sg_def(res.max_dispatch_calls_per_pass, _SG_DEFAULT_MAX_DISPATCH_CALLS_PER_PASS); + res.max_commit_listeners = _sg_def(res.max_commit_listeners, _SG_DEFAULT_MAX_COMMIT_LISTENERS); + res.wgpu_bindgroups_cache_size = _sg_def(res.wgpu_bindgroups_cache_size, _SG_DEFAULT_WGPU_BINDGROUP_CACHE_SIZE); + return res; +} + +_SOKOL_PRIVATE sg_pass _sg_pass_defaults(const sg_pass* pass) { + sg_pass res = *pass; + if (!res.compute) { + if (res.compute && res.attachments.id == SG_INVALID_ID) { + // this is a swapchain-pass + res.swapchain.sample_count = _sg_def(res.swapchain.sample_count, _sg.desc.environment.defaults.sample_count); + res.swapchain.color_format = _sg_def(res.swapchain.color_format, _sg.desc.environment.defaults.color_format); + res.swapchain.depth_format = _sg_def(res.swapchain.depth_format, _sg.desc.environment.defaults.depth_format); + } + res.action = _sg_pass_action_defaults(&res.action); + } + return res; +} + +// ██████ ██ ██ ██████ ██ ██ ██████ +// ██ ██ ██ ██ ██ ██ ██ ██ ██ +// ██████ ██ ██ ██████ ██ ██ ██ +// ██ ██ ██ ██ ██ ██ ██ ██ +// ██ ██████ ██████ ███████ ██ ██████ +// +// >>public +SOKOL_API_IMPL void sg_setup(const sg_desc* desc) { + SOKOL_ASSERT(desc); + SOKOL_ASSERT((desc->_start_canary == 0) && (desc->_end_canary == 0)); + SOKOL_ASSERT((desc->allocator.alloc_fn && desc->allocator.free_fn) || (!desc->allocator.alloc_fn && !desc->allocator.free_fn)); + _SG_CLEAR_ARC_STRUCT(_sg_state_t, _sg); + _sg.desc = _sg_desc_defaults(desc); + _sg_setup_pools(&_sg.pools, &_sg.desc); + _sg_setup_compute(&_sg.desc); + _sg_setup_commit_listeners(&_sg.desc); + _sg.frame_index = 1; + _sg.stats_enabled = true; + _sg_setup_backend(&_sg.desc); + _sg.valid = true; +} + +SOKOL_API_IMPL void sg_shutdown(void) { + _sg_discard_all_resources(&_sg.pools); + _sg_discard_backend(); + _sg_discard_commit_listeners(); + _sg_discard_compute(); + _sg_discard_pools(&_sg.pools); + _SG_CLEAR_ARC_STRUCT(_sg_state_t, _sg); +} + +SOKOL_API_IMPL bool sg_isvalid(void) { + return _sg.valid; +} + +SOKOL_API_IMPL sg_desc sg_query_desc(void) { + SOKOL_ASSERT(_sg.valid); + return _sg.desc; +} + +SOKOL_API_IMPL sg_backend sg_query_backend(void) { + SOKOL_ASSERT(_sg.valid); + return _sg.backend; +} + +SOKOL_API_IMPL sg_features sg_query_features(void) { + SOKOL_ASSERT(_sg.valid); + return _sg.features; +} + +SOKOL_API_IMPL sg_limits sg_query_limits(void) { + SOKOL_ASSERT(_sg.valid); + return _sg.limits; +} + +SOKOL_API_IMPL sg_pixelformat_info sg_query_pixelformat(sg_pixel_format fmt) { + SOKOL_ASSERT(_sg.valid); + int fmt_index = (int) fmt; + SOKOL_ASSERT((fmt_index > SG_PIXELFORMAT_NONE) && (fmt_index < _SG_PIXELFORMAT_NUM)); + const _sg_pixelformat_info_t* src = &_sg.formats[fmt_index]; + sg_pixelformat_info res; + _sg_clear(&res, sizeof(res)); + res.sample = src->sample; + res.filter = src->filter; + res.render = src->render; + res.blend = src->blend; + res.msaa = src->msaa; + res.depth = src->depth; + res.compressed = _sg_is_compressed_pixel_format(fmt); + if (!res.compressed) { + res.bytes_per_pixel = _sg_pixelformat_bytesize(fmt); + } + return res; +} + +SOKOL_API_IMPL int sg_query_row_pitch(sg_pixel_format fmt, int width, int row_align_bytes) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(width > 0); + SOKOL_ASSERT((row_align_bytes > 0) && _sg_ispow2(row_align_bytes)); + SOKOL_ASSERT(((int)fmt > SG_PIXELFORMAT_NONE) && ((int)fmt < _SG_PIXELFORMAT_NUM)); + return _sg_row_pitch(fmt, width, row_align_bytes); +} + +SOKOL_API_IMPL int sg_query_surface_pitch(sg_pixel_format fmt, int width, int height, int row_align_bytes) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT((width > 0) && (height > 0)); + SOKOL_ASSERT((row_align_bytes > 0) && _sg_ispow2(row_align_bytes)); + SOKOL_ASSERT(((int)fmt > SG_PIXELFORMAT_NONE) && ((int)fmt < _SG_PIXELFORMAT_NUM)); + return _sg_surface_pitch(fmt, width, height, row_align_bytes); +} + +SOKOL_API_IMPL sg_frame_stats sg_query_frame_stats(void) { + SOKOL_ASSERT(_sg.valid); + return _sg.prev_stats; +} + +SOKOL_API_IMPL sg_trace_hooks sg_install_trace_hooks(const sg_trace_hooks* trace_hooks) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(trace_hooks); + _SOKOL_UNUSED(trace_hooks); + #if defined(SOKOL_TRACE_HOOKS) + sg_trace_hooks old_hooks = _sg.hooks; + _sg.hooks = *trace_hooks; + #else + static sg_trace_hooks old_hooks; + _SG_WARN(TRACE_HOOKS_NOT_ENABLED); + #endif + return old_hooks; +} + +SOKOL_API_IMPL sg_buffer sg_alloc_buffer(void) { + SOKOL_ASSERT(_sg.valid); + sg_buffer res = _sg_alloc_buffer(); + _SG_TRACE_ARGS(alloc_buffer, res); + return res; +} + +SOKOL_API_IMPL sg_image sg_alloc_image(void) { + SOKOL_ASSERT(_sg.valid); + sg_image res = _sg_alloc_image(); + _SG_TRACE_ARGS(alloc_image, res); + return res; +} + +SOKOL_API_IMPL sg_sampler sg_alloc_sampler(void) { + SOKOL_ASSERT(_sg.valid); + sg_sampler res = _sg_alloc_sampler(); + _SG_TRACE_ARGS(alloc_sampler, res); + return res; +} + +SOKOL_API_IMPL sg_shader sg_alloc_shader(void) { + SOKOL_ASSERT(_sg.valid); + sg_shader res = _sg_alloc_shader(); + _SG_TRACE_ARGS(alloc_shader, res); + return res; +} + +SOKOL_API_IMPL sg_pipeline sg_alloc_pipeline(void) { + SOKOL_ASSERT(_sg.valid); + sg_pipeline res = _sg_alloc_pipeline(); + _SG_TRACE_ARGS(alloc_pipeline, res); + return res; +} + +SOKOL_API_IMPL sg_attachments sg_alloc_attachments(void) { + SOKOL_ASSERT(_sg.valid); + sg_attachments res = _sg_alloc_attachments(); + _SG_TRACE_ARGS(alloc_attachments, res); + return res; +} + +SOKOL_API_IMPL void sg_dealloc_buffer(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + if (buf->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_buffer(buf); + } else { + _SG_ERROR(DEALLOC_BUFFER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(dealloc_buffer, buf_id); +} + +SOKOL_API_IMPL void sg_dealloc_image(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + if (img->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_image(img); + } else { + _SG_ERROR(DEALLOC_IMAGE_INVALID_STATE); + } + } + _SG_TRACE_ARGS(dealloc_image, img_id); +} + +SOKOL_API_IMPL void sg_dealloc_sampler(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + if (smp->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_sampler(smp); + } else { + _SG_ERROR(DEALLOC_SAMPLER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(dealloc_sampler, smp_id); +} + +SOKOL_API_IMPL void sg_dealloc_shader(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + if (shd->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_shader(shd); + } else { + _SG_ERROR(DEALLOC_SHADER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(dealloc_shader, shd_id); +} + +SOKOL_API_IMPL void sg_dealloc_pipeline(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + if (pip->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_pipeline(pip); + } else { + _SG_ERROR(DEALLOC_PIPELINE_INVALID_STATE); + } + } + _SG_TRACE_ARGS(dealloc_pipeline, pip_id); +} + +SOKOL_API_IMPL void sg_dealloc_attachments(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + if (atts->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_attachments(atts); + } else { + _SG_ERROR(DEALLOC_ATTACHMENTS_INVALID_STATE); + } + } + _SG_TRACE_ARGS(dealloc_attachments, atts_id); +} + +SOKOL_API_IMPL void sg_init_buffer(sg_buffer buf_id, const sg_buffer_desc* desc) { + SOKOL_ASSERT(_sg.valid); + sg_buffer_desc desc_def = _sg_buffer_desc_defaults(desc); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + if (buf->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_init_buffer(buf, &desc_def); + SOKOL_ASSERT((buf->slot.state == SG_RESOURCESTATE_VALID) || (buf->slot.state == SG_RESOURCESTATE_FAILED)); + } else { + _SG_ERROR(INIT_BUFFER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(init_buffer, buf_id, &desc_def); +} + +SOKOL_API_IMPL void sg_init_image(sg_image img_id, const sg_image_desc* desc) { + SOKOL_ASSERT(_sg.valid); + sg_image_desc desc_def = _sg_image_desc_defaults(desc); + _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + if (img->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_init_image(img, &desc_def); + SOKOL_ASSERT((img->slot.state == SG_RESOURCESTATE_VALID) || (img->slot.state == SG_RESOURCESTATE_FAILED)); + } else { + _SG_ERROR(INIT_IMAGE_INVALID_STATE); + } + } + _SG_TRACE_ARGS(init_image, img_id, &desc_def); +} + +SOKOL_API_IMPL void sg_init_sampler(sg_sampler smp_id, const sg_sampler_desc* desc) { + SOKOL_ASSERT(_sg.valid); + sg_sampler_desc desc_def = _sg_sampler_desc_defaults(desc); + _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + if (smp->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_init_sampler(smp, &desc_def); + SOKOL_ASSERT((smp->slot.state == SG_RESOURCESTATE_VALID) || (smp->slot.state == SG_RESOURCESTATE_FAILED)); + } else { + _SG_ERROR(INIT_SAMPLER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(init_sampler, smp_id, &desc_def); +} + +SOKOL_API_IMPL void sg_init_shader(sg_shader shd_id, const sg_shader_desc* desc) { + SOKOL_ASSERT(_sg.valid); + sg_shader_desc desc_def = _sg_shader_desc_defaults(desc); + _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + if (shd->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_init_shader(shd, &desc_def); + SOKOL_ASSERT((shd->slot.state == SG_RESOURCESTATE_VALID) || (shd->slot.state == SG_RESOURCESTATE_FAILED)); + } else { + _SG_ERROR(INIT_SHADER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(init_shader, shd_id, &desc_def); +} + +SOKOL_API_IMPL void sg_init_pipeline(sg_pipeline pip_id, const sg_pipeline_desc* desc) { + SOKOL_ASSERT(_sg.valid); + sg_pipeline_desc desc_def = _sg_pipeline_desc_defaults(desc); + _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + if (pip->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_init_pipeline(pip, &desc_def); + SOKOL_ASSERT((pip->slot.state == SG_RESOURCESTATE_VALID) || (pip->slot.state == SG_RESOURCESTATE_FAILED)); + } else { + _SG_ERROR(INIT_PIPELINE_INVALID_STATE); + } + } + _SG_TRACE_ARGS(init_pipeline, pip_id, &desc_def); +} + +SOKOL_API_IMPL void sg_init_attachments(sg_attachments atts_id, const sg_attachments_desc* desc) { + SOKOL_ASSERT(_sg.valid); + sg_attachments_desc desc_def = _sg_attachments_desc_defaults(desc); + _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + if (atts->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_init_attachments(atts, &desc_def); + SOKOL_ASSERT((atts->slot.state == SG_RESOURCESTATE_VALID) || (atts->slot.state == SG_RESOURCESTATE_FAILED)); + } else { + _SG_ERROR(INIT_ATTACHMENTS_INVALID_STATE); + } + } + _SG_TRACE_ARGS(init_attachments, atts_id, &desc_def); +} + +SOKOL_API_IMPL void sg_uninit_buffer(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + if ((buf->slot.state == SG_RESOURCESTATE_VALID) || (buf->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_buffer(buf); + SOKOL_ASSERT(buf->slot.state == SG_RESOURCESTATE_ALLOC); + } else { + _SG_ERROR(UNINIT_BUFFER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(uninit_buffer, buf_id); +} + +SOKOL_API_IMPL void sg_uninit_image(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + if ((img->slot.state == SG_RESOURCESTATE_VALID) || (img->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_image(img); + SOKOL_ASSERT(img->slot.state == SG_RESOURCESTATE_ALLOC); + } else { + _SG_ERROR(UNINIT_IMAGE_INVALID_STATE); + } + } + _SG_TRACE_ARGS(uninit_image, img_id); +} + +SOKOL_API_IMPL void sg_uninit_sampler(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + if ((smp->slot.state == SG_RESOURCESTATE_VALID) || (smp->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_sampler(smp); + SOKOL_ASSERT(smp->slot.state == SG_RESOURCESTATE_ALLOC); + } else { + _SG_ERROR(UNINIT_SAMPLER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(uninit_sampler, smp_id); +} + +SOKOL_API_IMPL void sg_uninit_shader(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + if ((shd->slot.state == SG_RESOURCESTATE_VALID) || (shd->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_shader(shd); + SOKOL_ASSERT(shd->slot.state == SG_RESOURCESTATE_ALLOC); + } else { + _SG_ERROR(UNINIT_SHADER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(uninit_shader, shd_id); +} + +SOKOL_API_IMPL void sg_uninit_pipeline(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + if ((pip->slot.state == SG_RESOURCESTATE_VALID) || (pip->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_pipeline(pip); + SOKOL_ASSERT(pip->slot.state == SG_RESOURCESTATE_ALLOC); + } else { + _SG_ERROR(UNINIT_PIPELINE_INVALID_STATE); + } + } + _SG_TRACE_ARGS(uninit_pipeline, pip_id); +} + +SOKOL_API_IMPL void sg_uninit_attachments(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + if ((atts->slot.state == SG_RESOURCESTATE_VALID) || (atts->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_attachments(atts); + SOKOL_ASSERT(atts->slot.state == SG_RESOURCESTATE_ALLOC); + } else { + _SG_ERROR(UNINIT_ATTACHMENTS_INVALID_STATE); + } + } + _SG_TRACE_ARGS(uninit_attachments, atts_id); +} + +SOKOL_API_IMPL void sg_fail_buffer(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + if (buf->slot.state == SG_RESOURCESTATE_ALLOC) { + buf->slot.state = SG_RESOURCESTATE_FAILED; + } else { + _SG_ERROR(FAIL_BUFFER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(fail_buffer, buf_id); +} + +SOKOL_API_IMPL void sg_fail_image(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + if (img->slot.state == SG_RESOURCESTATE_ALLOC) { + img->slot.state = SG_RESOURCESTATE_FAILED; + } else { + _SG_ERROR(FAIL_IMAGE_INVALID_STATE); + } + } + _SG_TRACE_ARGS(fail_image, img_id); +} + +SOKOL_API_IMPL void sg_fail_sampler(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + if (smp->slot.state == SG_RESOURCESTATE_ALLOC) { + smp->slot.state = SG_RESOURCESTATE_FAILED; + } else { + _SG_ERROR(FAIL_SAMPLER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(fail_sampler, smp_id); +} + +SOKOL_API_IMPL void sg_fail_shader(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + if (shd->slot.state == SG_RESOURCESTATE_ALLOC) { + shd->slot.state = SG_RESOURCESTATE_FAILED; + } else { + _SG_ERROR(FAIL_SHADER_INVALID_STATE); + } + } + _SG_TRACE_ARGS(fail_shader, shd_id); +} + +SOKOL_API_IMPL void sg_fail_pipeline(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + if (pip->slot.state == SG_RESOURCESTATE_ALLOC) { + pip->slot.state = SG_RESOURCESTATE_FAILED; + } else { + _SG_ERROR(FAIL_PIPELINE_INVALID_STATE); + } + } + _SG_TRACE_ARGS(fail_pipeline, pip_id); +} + +SOKOL_API_IMPL void sg_fail_attachments(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + if (atts->slot.state == SG_RESOURCESTATE_ALLOC) { + atts->slot.state = SG_RESOURCESTATE_FAILED; + } else { + _SG_ERROR(FAIL_ATTACHMENTS_INVALID_STATE); + } + } + _SG_TRACE_ARGS(fail_attachments, atts_id); +} + +SOKOL_API_IMPL sg_resource_state sg_query_buffer_state(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + sg_resource_state res = buf ? buf->slot.state : SG_RESOURCESTATE_INVALID; + return res; +} + +SOKOL_API_IMPL sg_resource_state sg_query_image_state(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + sg_resource_state res = img ? img->slot.state : SG_RESOURCESTATE_INVALID; + return res; +} + +SOKOL_API_IMPL sg_resource_state sg_query_sampler_state(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + sg_resource_state res = smp ? smp->slot.state : SG_RESOURCESTATE_INVALID; + return res; +} + +SOKOL_API_IMPL sg_resource_state sg_query_shader_state(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + sg_resource_state res = shd ? shd->slot.state : SG_RESOURCESTATE_INVALID; + return res; +} + +SOKOL_API_IMPL sg_resource_state sg_query_pipeline_state(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + sg_resource_state res = pip ? pip->slot.state : SG_RESOURCESTATE_INVALID; + return res; +} + +SOKOL_API_IMPL sg_resource_state sg_query_attachments_state(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + sg_resource_state res = atts ? atts->slot.state : SG_RESOURCESTATE_INVALID; + return res; +} + +SOKOL_API_IMPL sg_buffer sg_make_buffer(const sg_buffer_desc* desc) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(desc); + sg_buffer_desc desc_def = _sg_buffer_desc_defaults(desc); + sg_buffer buf_id = _sg_alloc_buffer(); + if (buf_id.id != SG_INVALID_ID) { + _sg_buffer_t* buf = _sg_buffer_at(&_sg.pools, buf_id.id); + SOKOL_ASSERT(buf && (buf->slot.state == SG_RESOURCESTATE_ALLOC)); + _sg_init_buffer(buf, &desc_def); + SOKOL_ASSERT((buf->slot.state == SG_RESOURCESTATE_VALID) || (buf->slot.state == SG_RESOURCESTATE_FAILED)); + } + _SG_TRACE_ARGS(make_buffer, &desc_def, buf_id); + return buf_id; +} + +SOKOL_API_IMPL sg_image sg_make_image(const sg_image_desc* desc) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(desc); + sg_image_desc desc_def = _sg_image_desc_defaults(desc); + sg_image img_id = _sg_alloc_image(); + if (img_id.id != SG_INVALID_ID) { + _sg_image_t* img = _sg_image_at(&_sg.pools, img_id.id); + SOKOL_ASSERT(img && (img->slot.state == SG_RESOURCESTATE_ALLOC)); + _sg_init_image(img, &desc_def); + SOKOL_ASSERT((img->slot.state == SG_RESOURCESTATE_VALID) || (img->slot.state == SG_RESOURCESTATE_FAILED)); + } + _SG_TRACE_ARGS(make_image, &desc_def, img_id); + return img_id; +} + +SOKOL_API_IMPL sg_sampler sg_make_sampler(const sg_sampler_desc* desc) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(desc); + sg_sampler_desc desc_def = _sg_sampler_desc_defaults(desc); + sg_sampler smp_id = _sg_alloc_sampler(); + if (smp_id.id != SG_INVALID_ID) { + _sg_sampler_t* smp = _sg_sampler_at(&_sg.pools, smp_id.id); + SOKOL_ASSERT(smp && (smp->slot.state == SG_RESOURCESTATE_ALLOC)); + _sg_init_sampler(smp, &desc_def); + SOKOL_ASSERT((smp->slot.state == SG_RESOURCESTATE_VALID) || (smp->slot.state == SG_RESOURCESTATE_FAILED)); + } + _SG_TRACE_ARGS(make_sampler, &desc_def, smp_id); + return smp_id; +} + +SOKOL_API_IMPL sg_shader sg_make_shader(const sg_shader_desc* desc) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(desc); + sg_shader_desc desc_def = _sg_shader_desc_defaults(desc); + sg_shader shd_id = _sg_alloc_shader(); + if (shd_id.id != SG_INVALID_ID) { + _sg_shader_t* shd = _sg_shader_at(&_sg.pools, shd_id.id); + SOKOL_ASSERT(shd && (shd->slot.state == SG_RESOURCESTATE_ALLOC)); + _sg_init_shader(shd, &desc_def); + SOKOL_ASSERT((shd->slot.state == SG_RESOURCESTATE_VALID) || (shd->slot.state == SG_RESOURCESTATE_FAILED)); + } + _SG_TRACE_ARGS(make_shader, &desc_def, shd_id); + return shd_id; +} + +SOKOL_API_IMPL sg_pipeline sg_make_pipeline(const sg_pipeline_desc* desc) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(desc); + sg_pipeline_desc desc_def = _sg_pipeline_desc_defaults(desc); + sg_pipeline pip_id = _sg_alloc_pipeline(); + if (pip_id.id != SG_INVALID_ID) { + _sg_pipeline_t* pip = _sg_pipeline_at(&_sg.pools, pip_id.id); + SOKOL_ASSERT(pip && (pip->slot.state == SG_RESOURCESTATE_ALLOC)); + _sg_init_pipeline(pip, &desc_def); + SOKOL_ASSERT((pip->slot.state == SG_RESOURCESTATE_VALID) || (pip->slot.state == SG_RESOURCESTATE_FAILED)); + } + _SG_TRACE_ARGS(make_pipeline, &desc_def, pip_id); + return pip_id; +} + +SOKOL_API_IMPL sg_attachments sg_make_attachments(const sg_attachments_desc* desc) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(desc); + sg_attachments_desc desc_def = _sg_attachments_desc_defaults(desc); + sg_attachments atts_id = _sg_alloc_attachments(); + if (atts_id.id != SG_INVALID_ID) { + _sg_attachments_t* atts = _sg_attachments_at(&_sg.pools, atts_id.id); + SOKOL_ASSERT(atts && (atts->slot.state == SG_RESOURCESTATE_ALLOC)); + _sg_init_attachments(atts, &desc_def); + SOKOL_ASSERT((atts->slot.state == SG_RESOURCESTATE_VALID) || (atts->slot.state == SG_RESOURCESTATE_FAILED)); + } + _SG_TRACE_ARGS(make_attachments, &desc_def, atts_id); + return atts_id; +} + +SOKOL_API_IMPL void sg_destroy_buffer(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + _SG_TRACE_ARGS(destroy_buffer, buf_id); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + if ((buf->slot.state == SG_RESOURCESTATE_VALID) || (buf->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_buffer(buf); + SOKOL_ASSERT(buf->slot.state == SG_RESOURCESTATE_ALLOC); + } + if (buf->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_buffer(buf); + SOKOL_ASSERT(buf->slot.state == SG_RESOURCESTATE_INITIAL); + } + } +} + +SOKOL_API_IMPL void sg_destroy_image(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + _SG_TRACE_ARGS(destroy_image, img_id); + _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + if ((img->slot.state == SG_RESOURCESTATE_VALID) || (img->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_image(img); + SOKOL_ASSERT(img->slot.state == SG_RESOURCESTATE_ALLOC); + } + if (img->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_image(img); + SOKOL_ASSERT(img->slot.state == SG_RESOURCESTATE_INITIAL); + } + } +} + +SOKOL_API_IMPL void sg_destroy_sampler(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + _SG_TRACE_ARGS(destroy_sampler, smp_id); + _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + if ((smp->slot.state == SG_RESOURCESTATE_VALID) || (smp->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_sampler(smp); + SOKOL_ASSERT(smp->slot.state == SG_RESOURCESTATE_ALLOC); + } + if (smp->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_sampler(smp); + SOKOL_ASSERT(smp->slot.state == SG_RESOURCESTATE_INITIAL); + } + } +} + +SOKOL_API_IMPL void sg_destroy_shader(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + _SG_TRACE_ARGS(destroy_shader, shd_id); + _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + if ((shd->slot.state == SG_RESOURCESTATE_VALID) || (shd->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_shader(shd); + SOKOL_ASSERT(shd->slot.state == SG_RESOURCESTATE_ALLOC); + } + if (shd->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_shader(shd); + SOKOL_ASSERT(shd->slot.state == SG_RESOURCESTATE_INITIAL); + } + } +} + +SOKOL_API_IMPL void sg_destroy_pipeline(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + _SG_TRACE_ARGS(destroy_pipeline, pip_id); + _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + if ((pip->slot.state == SG_RESOURCESTATE_VALID) || (pip->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_pipeline(pip); + SOKOL_ASSERT(pip->slot.state == SG_RESOURCESTATE_ALLOC); + } + if (pip->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_pipeline(pip); + SOKOL_ASSERT(pip->slot.state == SG_RESOURCESTATE_INITIAL); + } + } +} + +SOKOL_API_IMPL void sg_destroy_attachments(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + _SG_TRACE_ARGS(destroy_attachments, atts_id); + _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + if ((atts->slot.state == SG_RESOURCESTATE_VALID) || (atts->slot.state == SG_RESOURCESTATE_FAILED)) { + _sg_uninit_attachments(atts); + SOKOL_ASSERT(atts->slot.state == SG_RESOURCESTATE_ALLOC); + } + if (atts->slot.state == SG_RESOURCESTATE_ALLOC) { + _sg_dealloc_attachments(atts); + SOKOL_ASSERT(atts->slot.state == SG_RESOURCESTATE_INITIAL); + } + } +} + +SOKOL_API_IMPL void sg_begin_pass(const sg_pass* pass) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(!_sg.cur_pass.valid); + SOKOL_ASSERT(!_sg.cur_pass.in_pass); + SOKOL_ASSERT(pass); + SOKOL_ASSERT((pass->_start_canary == 0) && (pass->_end_canary == 0)); + const sg_pass pass_def = _sg_pass_defaults(pass); + if (!_sg_validate_begin_pass(&pass_def)) { + return; + } + if (!pass_def.compute) { + if (pass_def.attachments.id != SG_INVALID_ID) { + // an offscreen pass + SOKOL_ASSERT(_sg.cur_pass.atts == 0); + _sg.cur_pass.atts = _sg_lookup_attachments(&_sg.pools, pass_def.attachments.id); + if (0 == _sg.cur_pass.atts) { + _SG_ERROR(BEGINPASS_ATTACHMENT_INVALID); + return; + } + _sg.cur_pass.atts_id = pass_def.attachments; + _sg.cur_pass.width = _sg.cur_pass.atts->cmn.width; + _sg.cur_pass.height = _sg.cur_pass.atts->cmn.height; + } else { + // a swapchain pass + SOKOL_ASSERT(pass_def.swapchain.width > 0); + SOKOL_ASSERT(pass_def.swapchain.height > 0); + SOKOL_ASSERT(pass_def.swapchain.color_format > SG_PIXELFORMAT_NONE); + SOKOL_ASSERT(pass_def.swapchain.sample_count > 0); + _sg.cur_pass.width = pass_def.swapchain.width; + _sg.cur_pass.height = pass_def.swapchain.height; + _sg.cur_pass.swapchain.color_fmt = pass_def.swapchain.color_format; + _sg.cur_pass.swapchain.depth_fmt = pass_def.swapchain.depth_format; + _sg.cur_pass.swapchain.sample_count = pass_def.swapchain.sample_count; + } + } + _sg.cur_pass.valid = true; // may be overruled by backend begin-pass functions + _sg.cur_pass.in_pass = true; + _sg.cur_pass.is_compute = pass_def.compute; + _sg_begin_pass(&pass_def); + _SG_TRACE_ARGS(begin_pass, &pass_def); +} + +SOKOL_API_IMPL void sg_apply_viewport(int x, int y, int width, int height, bool origin_top_left) { + SOKOL_ASSERT(_sg.valid); + #if defined(SOKOL_DEBUG) + if (!_sg_validate_apply_viewport(x, y, width, height, origin_top_left)) { + return; + } + #endif + _sg_stats_add(num_apply_viewport, 1); + if (!_sg.cur_pass.valid) { + return; + } + _sg_apply_viewport(x, y, width, height, origin_top_left); + _SG_TRACE_ARGS(apply_viewport, x, y, width, height, origin_top_left); +} + +SOKOL_API_IMPL void sg_apply_viewportf(float x, float y, float width, float height, bool origin_top_left) { + sg_apply_viewport((int)x, (int)y, (int)width, (int)height, origin_top_left); +} + +SOKOL_API_IMPL void sg_apply_scissor_rect(int x, int y, int width, int height, bool origin_top_left) { + SOKOL_ASSERT(_sg.valid); + #if defined(SOKOL_DEBUG) + if (!_sg_validate_apply_scissor_rect(x, y, width, height, origin_top_left)) { + return; + } + #endif + _sg_stats_add(num_apply_scissor_rect, 1); + if (!_sg.cur_pass.valid) { + return; + } + _sg_apply_scissor_rect(x, y, width, height, origin_top_left); + _SG_TRACE_ARGS(apply_scissor_rect, x, y, width, height, origin_top_left); +} + +SOKOL_API_IMPL void sg_apply_scissor_rectf(float x, float y, float width, float height, bool origin_top_left) { + sg_apply_scissor_rect((int)x, (int)y, (int)width, (int)height, origin_top_left); +} + +SOKOL_API_IMPL void sg_apply_pipeline(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + _sg_stats_add(num_apply_pipeline, 1); + if (!_sg_validate_apply_pipeline(pip_id)) { + _sg.next_draw_valid = false; + return; + } + if (!_sg.cur_pass.valid) { + return; + } + _sg.cur_pipeline = pip_id; + _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + SOKOL_ASSERT(pip); + + _sg.next_draw_valid = (SG_RESOURCESTATE_VALID == pip->slot.state); + if (!_sg.next_draw_valid) { + return; + } + + SOKOL_ASSERT(pip->shader && (pip->shader->slot.id == pip->cmn.shader_id.id)); + _sg_apply_pipeline(pip); + + // set the expected bindings and uniform block flags + _sg.required_bindings_and_uniforms = pip->cmn.required_bindings_and_uniforms | pip->shader->cmn.required_bindings_and_uniforms; + _sg.applied_bindings_and_uniforms = 0; + + _SG_TRACE_ARGS(apply_pipeline, pip_id); +} + +SOKOL_API_IMPL void sg_apply_bindings(const sg_bindings* bindings) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(bindings); + SOKOL_ASSERT((bindings->_start_canary == 0) && (bindings->_end_canary==0)); + _sg_stats_add(num_apply_bindings, 1); + _sg.applied_bindings_and_uniforms |= (1 << SG_MAX_UNIFORMBLOCK_BINDSLOTS); + if (!_sg_validate_apply_bindings(bindings)) { + _sg.next_draw_valid = false; + return; + } + if (!_sg.cur_pass.valid) { + return; + } + if (!_sg.next_draw_valid) { + return; + } + + _sg_bindings_t bnd; + _sg_clear(&bnd, sizeof(bnd)); + bnd.pip = _sg_lookup_pipeline(&_sg.pools, _sg.cur_pipeline.id); + if (0 == bnd.pip) { + _sg.next_draw_valid = false; + } + SOKOL_ASSERT(bnd.pip->shader && (bnd.pip->cmn.shader_id.id == bnd.pip->shader->slot.id)); + const _sg_shader_t* shd = bnd.pip->shader; + + if (!_sg.cur_pass.is_compute) { + for (size_t i = 0; i < SG_MAX_VERTEXBUFFER_BINDSLOTS; i++) { + if (bnd.pip->cmn.vertex_buffer_layout_active[i]) { + SOKOL_ASSERT(bindings->vertex_buffers[i].id != SG_INVALID_ID); + bnd.vbs[i] = _sg_lookup_buffer(&_sg.pools, bindings->vertex_buffers[i].id); + bnd.vb_offsets[i] = bindings->vertex_buffer_offsets[i]; + if (bnd.vbs[i]) { + _sg.next_draw_valid &= (SG_RESOURCESTATE_VALID == bnd.vbs[i]->slot.state); + _sg.next_draw_valid &= !bnd.vbs[i]->cmn.append_overflow; + } else { + _sg.next_draw_valid = false; + } + } + } + if (bindings->index_buffer.id) { + bnd.ib = _sg_lookup_buffer(&_sg.pools, bindings->index_buffer.id); + bnd.ib_offset = bindings->index_buffer_offset; + if (bnd.ib) { + _sg.next_draw_valid &= (SG_RESOURCESTATE_VALID == bnd.ib->slot.state); + _sg.next_draw_valid &= !bnd.ib->cmn.append_overflow; + } else { + _sg.next_draw_valid = false; + } + } + } + + for (int i = 0; i < SG_MAX_IMAGE_BINDSLOTS; i++) { + if (shd->cmn.images[i].stage != SG_SHADERSTAGE_NONE) { + SOKOL_ASSERT(bindings->images[i].id != SG_INVALID_ID); + bnd.imgs[i] = _sg_lookup_image(&_sg.pools, bindings->images[i].id); + if (bnd.imgs[i]) { + _sg.next_draw_valid &= (SG_RESOURCESTATE_VALID == bnd.imgs[i]->slot.state); + } else { + _sg.next_draw_valid = false; + } + } + } + + for (size_t i = 0; i < SG_MAX_SAMPLER_BINDSLOTS; i++) { + if (shd->cmn.samplers[i].stage != SG_SHADERSTAGE_NONE) { + SOKOL_ASSERT(bindings->samplers[i].id != SG_INVALID_ID); + bnd.smps[i] = _sg_lookup_sampler(&_sg.pools, bindings->samplers[i].id); + if (bnd.smps[i]) { + _sg.next_draw_valid &= (SG_RESOURCESTATE_VALID == bnd.smps[i]->slot.state); + } else { + _sg.next_draw_valid = false; + } + } + } + + for (size_t i = 0; i < SG_MAX_STORAGEBUFFER_BINDSLOTS; i++) { + if (shd->cmn.storage_buffers[i].stage != SG_SHADERSTAGE_NONE) { + SOKOL_ASSERT(bindings->storage_buffers[i].id != SG_INVALID_ID); + bnd.sbufs[i] = _sg_lookup_buffer(&_sg.pools, bindings->storage_buffers[i].id); + if (bnd.sbufs[i]) { + _sg.next_draw_valid &= (SG_RESOURCESTATE_VALID == bnd.sbufs[i]->slot.state); + if (_sg.cur_pass.is_compute) { + _sg_compute_pass_track_storage_buffer(bnd.sbufs[i], shd->cmn.storage_buffers[i].readonly); + } + } else { + _sg.next_draw_valid = false; + } + } + } + + if (_sg.next_draw_valid) { + _sg.next_draw_valid &= _sg_apply_bindings(&bnd); + _SG_TRACE_ARGS(apply_bindings, bindings); + } +} + +SOKOL_API_IMPL void sg_apply_uniforms(int ub_slot, const sg_range* data) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT((ub_slot >= 0) && (ub_slot < SG_MAX_UNIFORMBLOCK_BINDSLOTS)); + SOKOL_ASSERT(data && data->ptr && (data->size > 0)); + _sg_stats_add(num_apply_uniforms, 1); + _sg_stats_add(size_apply_uniforms, (uint32_t)data->size); + _sg.applied_bindings_and_uniforms |= 1 << ub_slot; + if (!_sg_validate_apply_uniforms(ub_slot, data)) { + _sg.next_draw_valid = false; + return; + } + if (!_sg.cur_pass.valid) { + return; + } + if (!_sg.next_draw_valid) { + return; + } + _sg_apply_uniforms(ub_slot, data); + _SG_TRACE_ARGS(apply_uniforms, ub_slot, data); +} + +SOKOL_API_IMPL void sg_draw(int base_element, int num_elements, int num_instances) { + SOKOL_ASSERT(_sg.valid); + #if defined(SOKOL_DEBUG) + if (!_sg_validate_draw(base_element, num_elements, num_instances)) { + return; + } + #endif + _sg_stats_add(num_draw, 1); + if (!_sg.cur_pass.valid) { + return; + } + if (!_sg.next_draw_valid) { + return; + } + // skip no-op draws + if ((0 == num_elements) || (0 == num_instances)) { + return; + } + _sg_draw(base_element, num_elements, num_instances); + _SG_TRACE_ARGS(draw, base_element, num_elements, num_instances); +} + +SOKOL_API_IMPL void sg_dispatch(int num_groups_x, int num_groups_y, int num_groups_z) { + SOKOL_ASSERT(_sg.valid); + #if defined(SOKOL_DEBUG) + if (!_sg_validate_dispatch(num_groups_x, num_groups_y, num_groups_z)) { + return; + } + #endif + _sg_stats_add(num_dispatch, 1); + if (!_sg.cur_pass.valid) { + return; + } + if (!_sg.next_draw_valid) { + return; + } + // skip no-op dispatches + if ((0 == num_groups_x) || (0 == num_groups_y) || (0 == num_groups_z)) { + return; + } + _sg_dispatch(num_groups_x, num_groups_y, num_groups_z); + _SG_TRACE_ARGS(dispatch, num_groups_x, num_groups_y, num_groups_z); +} + +SOKOL_API_IMPL void sg_end_pass(void) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(_sg.cur_pass.in_pass); + _sg_stats_add(num_passes, 1); + // NOTE: don't exit early if !_sg.cur_pass.valid + _sg_end_pass(); + _sg.cur_pipeline.id = SG_INVALID_ID; + if (_sg.cur_pass.is_compute) { + _sg_compute_on_endpass(); + } + _sg_clear(&_sg.cur_pass, sizeof(_sg.cur_pass)); + _SG_TRACE_NOARGS(end_pass); +} + +SOKOL_API_IMPL void sg_commit(void) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(!_sg.cur_pass.valid); + SOKOL_ASSERT(!_sg.cur_pass.in_pass); + _sg_commit(); + _sg.stats.frame_index = _sg.frame_index; + _sg.prev_stats = _sg.stats; + _sg_clear(&_sg.stats, sizeof(_sg.stats)); + _sg_notify_commit_listeners(); + _SG_TRACE_NOARGS(commit); + _sg.frame_index++; +} + +SOKOL_API_IMPL void sg_reset_state_cache(void) { + SOKOL_ASSERT(_sg.valid); + _sg_reset_state_cache(); + _SG_TRACE_NOARGS(reset_state_cache); +} + +SOKOL_API_IMPL void sg_update_buffer(sg_buffer buf_id, const sg_range* data) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(data && data->ptr && (data->size > 0)); + _sg_stats_add(num_update_buffer, 1); + _sg_stats_add(size_update_buffer, (uint32_t)data->size); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if ((data->size > 0) && buf && (buf->slot.state == SG_RESOURCESTATE_VALID)) { + if (_sg_validate_update_buffer(buf, data)) { + SOKOL_ASSERT(data->size <= (size_t)buf->cmn.size); + // only one update allowed per buffer and frame + SOKOL_ASSERT(buf->cmn.update_frame_index != _sg.frame_index); + // update and append on same buffer in same frame not allowed + SOKOL_ASSERT(buf->cmn.append_frame_index != _sg.frame_index); + _sg_update_buffer(buf, data); + buf->cmn.update_frame_index = _sg.frame_index; + } + } + _SG_TRACE_ARGS(update_buffer, buf_id, data); +} + +SOKOL_API_IMPL int sg_append_buffer(sg_buffer buf_id, const sg_range* data) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(data && data->ptr); + _sg_stats_add(num_append_buffer, 1); + _sg_stats_add(size_append_buffer, (uint32_t)data->size); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + int result; + if (buf) { + // rewind append cursor in a new frame + if (buf->cmn.append_frame_index != _sg.frame_index) { + buf->cmn.append_pos = 0; + buf->cmn.append_overflow = false; + } + if (((size_t)buf->cmn.append_pos + data->size) > (size_t)buf->cmn.size) { + buf->cmn.append_overflow = true; + } + const int start_pos = buf->cmn.append_pos; + // NOTE: the multiple-of-4 requirement for the buffer offset is coming + // from WebGPU, but we want identical behaviour between backends + SOKOL_ASSERT(_sg_multiple_u64((uint64_t)start_pos, 4)); + if (buf->slot.state == SG_RESOURCESTATE_VALID) { + if (_sg_validate_append_buffer(buf, data)) { + if (!buf->cmn.append_overflow && (data->size > 0)) { + // update and append on same buffer in same frame not allowed + SOKOL_ASSERT(buf->cmn.update_frame_index != _sg.frame_index); + _sg_append_buffer(buf, data, buf->cmn.append_frame_index != _sg.frame_index); + buf->cmn.append_pos += (int) _sg_roundup_u64(data->size, 4); + buf->cmn.append_frame_index = _sg.frame_index; + } + } + } + result = start_pos; + } else { + // FIXME: should we return -1 here? + result = 0; + } + _SG_TRACE_ARGS(append_buffer, buf_id, data, result); + return result; +} + +SOKOL_API_IMPL bool sg_query_buffer_overflow(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + bool result = buf ? buf->cmn.append_overflow : false; + return result; +} + +SOKOL_API_IMPL bool sg_query_buffer_will_overflow(sg_buffer buf_id, size_t size) { + SOKOL_ASSERT(_sg.valid); + _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + bool result = false; + if (buf) { + int append_pos = buf->cmn.append_pos; + // rewind append cursor in a new frame + if (buf->cmn.append_frame_index != _sg.frame_index) { + append_pos = 0; + } + if ((append_pos + _sg_roundup((int)size, 4)) > buf->cmn.size) { + result = true; + } + } + return result; +} + +SOKOL_API_IMPL void sg_update_image(sg_image img_id, const sg_image_data* data) { + SOKOL_ASSERT(_sg.valid); + _sg_stats_add(num_update_image, 1); + for (int face_index = 0; face_index < SG_CUBEFACE_NUM; face_index++) { + for (int mip_index = 0; mip_index < SG_MAX_MIPMAPS; mip_index++) { + if (data->subimage[face_index][mip_index].size == 0) { + break; + } + _sg_stats_add(size_update_image, (uint32_t)data->subimage[face_index][mip_index].size); + } + } + _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img && img->slot.state == SG_RESOURCESTATE_VALID) { + if (_sg_validate_update_image(img, data)) { + SOKOL_ASSERT(img->cmn.upd_frame_index != _sg.frame_index); + _sg_update_image(img, data); + img->cmn.upd_frame_index = _sg.frame_index; + } + } + _SG_TRACE_ARGS(update_image, img_id, data); +} + +SOKOL_API_IMPL void sg_push_debug_group(const char* name) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(name); + _sg_push_debug_group(name); + _SG_TRACE_ARGS(push_debug_group, name); +} + +SOKOL_API_IMPL void sg_pop_debug_group(void) { + SOKOL_ASSERT(_sg.valid); + _sg_pop_debug_group(); + _SG_TRACE_NOARGS(pop_debug_group); +} + +SOKOL_API_IMPL bool sg_add_commit_listener(sg_commit_listener listener) { + SOKOL_ASSERT(_sg.valid); + return _sg_add_commit_listener(&listener); +} + +SOKOL_API_IMPL bool sg_remove_commit_listener(sg_commit_listener listener) { + SOKOL_ASSERT(_sg.valid); + return _sg_remove_commit_listener(&listener); +} + +SOKOL_API_IMPL void sg_enable_frame_stats(void) { + SOKOL_ASSERT(_sg.valid); + _sg.stats_enabled = true; +} + +SOKOL_API_IMPL void sg_disable_frame_stats(void) { + SOKOL_ASSERT(_sg.valid); + _sg.stats_enabled = false; +} + +SOKOL_API_IMPL bool sg_frame_stats_enabled(void) { + return _sg.stats_enabled; +} + +SOKOL_API_IMPL sg_buffer_info sg_query_buffer_info(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + sg_buffer_info info; + _sg_clear(&info, sizeof(info)); + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + info.slot.state = buf->slot.state; + info.slot.res_id = buf->slot.id; + info.update_frame_index = buf->cmn.update_frame_index; + info.append_frame_index = buf->cmn.append_frame_index; + info.append_pos = buf->cmn.append_pos; + info.append_overflow = buf->cmn.append_overflow; + #if defined(SOKOL_D3D11) + info.num_slots = 1; + info.active_slot = 0; + #else + info.num_slots = buf->cmn.num_slots; + info.active_slot = buf->cmn.active_slot; + #endif + } + return info; +} + +SOKOL_API_IMPL sg_image_info sg_query_image_info(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + sg_image_info info; + _sg_clear(&info, sizeof(info)); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + info.slot.state = img->slot.state; + info.slot.res_id = img->slot.id; + info.upd_frame_index = img->cmn.upd_frame_index; + #if defined(SOKOL_D3D11) + info.num_slots = 1; + info.active_slot = 0; + #else + info.num_slots = img->cmn.num_slots; + info.active_slot = img->cmn.active_slot; + #endif + } + return info; +} + +SOKOL_API_IMPL sg_sampler_info sg_query_sampler_info(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + sg_sampler_info info; + _sg_clear(&info, sizeof(info)); + const _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + info.slot.state = smp->slot.state; + info.slot.res_id = smp->slot.id; + } + return info; +} + +SOKOL_API_IMPL sg_shader_info sg_query_shader_info(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + sg_shader_info info; + _sg_clear(&info, sizeof(info)); + const _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + info.slot.state = shd->slot.state; + info.slot.res_id = shd->slot.id; + } + return info; +} + +SOKOL_API_IMPL sg_pipeline_info sg_query_pipeline_info(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + sg_pipeline_info info; + _sg_clear(&info, sizeof(info)); + const _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + info.slot.state = pip->slot.state; + info.slot.res_id = pip->slot.id; + } + return info; +} + +SOKOL_API_IMPL sg_attachments_info sg_query_attachments_info(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + sg_attachments_info info; + _sg_clear(&info, sizeof(info)); + const _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + info.slot.state = atts->slot.state; + info.slot.res_id = atts->slot.id; + } + return info; +} + +SOKOL_API_IMPL sg_buffer_desc sg_query_buffer_desc(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + sg_buffer_desc desc; + _sg_clear(&desc, sizeof(desc)); + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + desc.size = (size_t)buf->cmn.size; + desc.type = buf->cmn.type; + desc.usage = buf->cmn.usage; + } + return desc; +} + +SOKOL_API_IMPL size_t sg_query_buffer_size(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + return (size_t)buf->cmn.size; + } + return 0; +} + +SOKOL_API_IMPL sg_buffer_type sg_query_buffer_type(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + return buf->cmn.type; + } + return _SG_BUFFERTYPE_DEFAULT; +} + +SOKOL_API_IMPL sg_usage sg_query_buffer_usage(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + return buf->cmn.usage; + } + return _SG_USAGE_DEFAULT; +} + +SOKOL_API_IMPL sg_image_desc sg_query_image_desc(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + sg_image_desc desc; + _sg_clear(&desc, sizeof(desc)); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + desc.type = img->cmn.type; + desc.render_target = img->cmn.render_target; + desc.width = img->cmn.width; + desc.height = img->cmn.height; + desc.num_slices = img->cmn.num_slices; + desc.num_mipmaps = img->cmn.num_mipmaps; + desc.usage = img->cmn.usage; + desc.pixel_format = img->cmn.pixel_format; + desc.sample_count = img->cmn.sample_count; + } + return desc; +} + +SOKOL_API_IMPL sg_image_type sg_query_image_type(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + return img->cmn.type; + } + return _SG_IMAGETYPE_DEFAULT; +} + +SOKOL_API_IMPL int sg_query_image_width(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + return img->cmn.width; + } + return 0; +} + +SOKOL_API_IMPL int sg_query_image_height(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + return img->cmn.height; + } + return 0; +} + +SOKOL_API_IMPL int sg_query_image_num_slices(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + return img->cmn.num_slices; + } + return 0; +} + +SOKOL_API_IMPL int sg_query_image_num_mipmaps(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + return img->cmn.num_mipmaps; + } + return 0; +} + +SOKOL_API_IMPL sg_pixel_format sg_query_image_pixelformat(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + return img->cmn.pixel_format; + } + return _SG_PIXELFORMAT_DEFAULT; +} + +SOKOL_API_IMPL sg_usage sg_query_image_usage(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + return img->cmn.usage; + } + return _SG_USAGE_DEFAULT; +} + +SOKOL_API_IMPL int sg_query_image_sample_count(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + SOKOL_ASSERT(_sg.valid); + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + return img->cmn.sample_count; + } + return 0; +} + +SOKOL_API_IMPL sg_sampler_desc sg_query_sampler_desc(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + sg_sampler_desc desc; + _sg_clear(&desc, sizeof(desc)); + const _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + desc.min_filter = smp->cmn.min_filter; + desc.mag_filter = smp->cmn.mag_filter; + desc.mipmap_filter = smp->cmn.mipmap_filter; + desc.wrap_u = smp->cmn.wrap_u; + desc.wrap_v = smp->cmn.wrap_v; + desc.wrap_w = smp->cmn.wrap_w; + desc.min_lod = smp->cmn.min_lod; + desc.max_lod = smp->cmn.max_lod; + desc.border_color = smp->cmn.border_color; + desc.compare = smp->cmn.compare; + desc.max_anisotropy = smp->cmn.max_anisotropy; + } + return desc; +} + +SOKOL_API_IMPL sg_shader_desc sg_query_shader_desc(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + sg_shader_desc desc; + _sg_clear(&desc, sizeof(desc)); + const _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + for (size_t ub_idx = 0; ub_idx < SG_MAX_UNIFORMBLOCK_BINDSLOTS; ub_idx++) { + sg_shader_uniform_block* ub_desc = &desc.uniform_blocks[ub_idx]; + const _sg_shader_uniform_block_t* ub = &shd->cmn.uniform_blocks[ub_idx]; + ub_desc->stage = ub->stage; + ub_desc->size = ub->size; + } + for (size_t sbuf_idx = 0; sbuf_idx < SG_MAX_STORAGEBUFFER_BINDSLOTS; sbuf_idx++) { + sg_shader_storage_buffer* sbuf_desc = &desc.storage_buffers[sbuf_idx]; + const _sg_shader_storage_buffer_t* sbuf = &shd->cmn.storage_buffers[sbuf_idx]; + sbuf_desc->stage = sbuf->stage; + sbuf_desc->readonly = sbuf->readonly; + } + for (size_t img_idx = 0; img_idx < SG_MAX_IMAGE_BINDSLOTS; img_idx++) { + sg_shader_image* img_desc = &desc.images[img_idx]; + const _sg_shader_image_t* img = &shd->cmn.images[img_idx]; + img_desc->stage = img->stage; + img_desc->image_type = img->image_type; + img_desc->sample_type = img->sample_type; + img_desc->multisampled = img->multisampled; + } + for (size_t smp_idx = 0; smp_idx < SG_MAX_SAMPLER_BINDSLOTS; smp_idx++) { + sg_shader_sampler* smp_desc = &desc.samplers[smp_idx]; + const _sg_shader_sampler_t* smp = &shd->cmn.samplers[smp_idx]; + smp_desc->stage = smp->stage; + smp_desc->sampler_type = smp->sampler_type; + } + for (size_t img_smp_idx = 0; img_smp_idx < SG_MAX_IMAGE_SAMPLER_PAIRS; img_smp_idx++) { + sg_shader_image_sampler_pair* img_smp_desc = &desc.image_sampler_pairs[img_smp_idx]; + const _sg_shader_image_sampler_t* img_smp = &shd->cmn.image_samplers[img_smp_idx]; + img_smp_desc->stage = img_smp->stage; + img_smp_desc->image_slot = img_smp->image_slot; + img_smp_desc->sampler_slot = img_smp->sampler_slot; + } + } + return desc; +} + +SOKOL_API_IMPL sg_pipeline_desc sg_query_pipeline_desc(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + sg_pipeline_desc desc; + _sg_clear(&desc, sizeof(desc)); + const _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + desc.compute = pip->cmn.is_compute; + desc.shader = pip->cmn.shader_id; + desc.layout = pip->cmn.layout; + desc.depth = pip->cmn.depth; + desc.stencil = pip->cmn.stencil; + desc.color_count = pip->cmn.color_count; + for (int i = 0; i < pip->cmn.color_count; i++) { + desc.colors[i] = pip->cmn.colors[i]; + } + desc.primitive_type = pip->cmn.primitive_type; + desc.index_type = pip->cmn.index_type; + desc.cull_mode = pip->cmn.cull_mode; + desc.face_winding = pip->cmn.face_winding; + desc.sample_count = pip->cmn.sample_count; + desc.blend_color = pip->cmn.blend_color; + desc.alpha_to_coverage_enabled = pip->cmn.alpha_to_coverage_enabled; + } + return desc; +} + +SOKOL_API_IMPL sg_attachments_desc sg_query_attachments_desc(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + sg_attachments_desc desc; + _sg_clear(&desc, sizeof(desc)); + const _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + for (int i = 0; i < atts->cmn.num_colors; i++) { + desc.colors[i].image = atts->cmn.colors[i].image_id; + desc.colors[i].mip_level = atts->cmn.colors[i].mip_level; + desc.colors[i].slice = atts->cmn.colors[i].slice; + } + desc.depth_stencil.image = atts->cmn.depth_stencil.image_id; + desc.depth_stencil.mip_level = atts->cmn.depth_stencil.mip_level; + desc.depth_stencil.slice = atts->cmn.depth_stencil.slice; + } + return desc; +} + +SOKOL_API_IMPL sg_buffer_desc sg_query_buffer_defaults(const sg_buffer_desc* desc) { + SOKOL_ASSERT(_sg.valid && desc); + return _sg_buffer_desc_defaults(desc); +} + +SOKOL_API_IMPL sg_image_desc sg_query_image_defaults(const sg_image_desc* desc) { + SOKOL_ASSERT(_sg.valid && desc); + return _sg_image_desc_defaults(desc); +} + +SOKOL_API_IMPL sg_sampler_desc sg_query_sampler_defaults(const sg_sampler_desc* desc) { + SOKOL_ASSERT(_sg.valid && desc); + return _sg_sampler_desc_defaults(desc); +} + +SOKOL_API_IMPL sg_shader_desc sg_query_shader_defaults(const sg_shader_desc* desc) { + SOKOL_ASSERT(_sg.valid && desc); + return _sg_shader_desc_defaults(desc); +} + +SOKOL_API_IMPL sg_pipeline_desc sg_query_pipeline_defaults(const sg_pipeline_desc* desc) { + SOKOL_ASSERT(_sg.valid && desc); + return _sg_pipeline_desc_defaults(desc); +} + +SOKOL_API_IMPL sg_attachments_desc sg_query_attachments_defaults(const sg_attachments_desc* desc) { + SOKOL_ASSERT(_sg.valid && desc); + return _sg_attachments_desc_defaults(desc); +} + +SOKOL_API_IMPL const void* sg_d3d11_device(void) { + #if defined(SOKOL_D3D11) + return (const void*) _sg.d3d11.dev; + #else + return 0; + #endif +} + +SOKOL_API_IMPL const void* sg_d3d11_device_context(void) { + #if defined(SOKOL_D3D11) + return (const void*) _sg.d3d11.ctx; + #else + return 0; + #endif +} + +SOKOL_API_IMPL sg_d3d11_buffer_info sg_d3d11_query_buffer_info(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + sg_d3d11_buffer_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_D3D11) + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + res.buf = (const void*) buf->d3d11.buf; + } + #else + _SOKOL_UNUSED(buf_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_d3d11_image_info sg_d3d11_query_image_info(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + sg_d3d11_image_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_D3D11) + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + res.tex2d = (const void*) img->d3d11.tex2d; + res.tex3d = (const void*) img->d3d11.tex3d; + res.res = (const void*) img->d3d11.res; + res.srv = (const void*) img->d3d11.srv; + } + #else + _SOKOL_UNUSED(img_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_d3d11_sampler_info sg_d3d11_query_sampler_info(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + sg_d3d11_sampler_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_D3D11) + const _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + res.smp = (const void*) smp->d3d11.smp; + } + #else + _SOKOL_UNUSED(smp_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_d3d11_shader_info sg_d3d11_query_shader_info(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + sg_d3d11_shader_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_D3D11) + const _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + for (size_t i = 0; i < SG_MAX_UNIFORMBLOCK_BINDSLOTS; i++) { + res.cbufs[i] = (const void*) shd->d3d11.all_cbufs[i]; + } + res.vs = (const void*) shd->d3d11.vs; + res.fs = (const void*) shd->d3d11.fs; + } + #else + _SOKOL_UNUSED(shd_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_d3d11_pipeline_info sg_d3d11_query_pipeline_info(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + sg_d3d11_pipeline_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_D3D11) + const _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + res.il = (const void*) pip->d3d11.il; + res.rs = (const void*) pip->d3d11.rs; + res.dss = (const void*) pip->d3d11.dss; + res.bs = (const void*) pip->d3d11.bs; + } + #else + _SOKOL_UNUSED(pip_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_d3d11_attachments_info sg_d3d11_query_attachments_info(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + sg_d3d11_attachments_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_D3D11) + const _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + res.color_rtv[i] = (const void*) atts->d3d11.colors[i].view.rtv; + res.resolve_rtv[i] = (const void*) atts->d3d11.resolves[i].view.rtv; + } + res.dsv = (const void*) atts->d3d11.depth_stencil.view.dsv; + } + #else + _SOKOL_UNUSED(atts_id); + #endif + return res; +} + +SOKOL_API_IMPL const void* sg_mtl_device(void) { + #if defined(SOKOL_METAL) + if (nil != _sg.mtl.device) { + return (__bridge const void*) _sg.mtl.device; + } else { + return 0; + } + #else + return 0; + #endif +} + +SOKOL_API_IMPL const void* sg_mtl_render_command_encoder(void) { + #if defined(SOKOL_METAL) + if (nil != _sg.mtl.render_cmd_encoder) { + return (__bridge const void*) _sg.mtl.render_cmd_encoder; + } else { + return 0; + } + #else + return 0; + #endif +} + +SOKOL_API_IMPL const void* sg_mtl_compute_command_encoder(void) { + #if defined(SOKOL_METAL) + if (nil != _sg.mtl.compute_cmd_encoder) { + return (__bridge const void*) _sg.mtl.compute_cmd_encoder; + } else { + return 0; + } + #else + return 0; + #endif +} + +SOKOL_API_IMPL sg_mtl_buffer_info sg_mtl_query_buffer_info(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + sg_mtl_buffer_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_METAL) + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + if (buf->mtl.buf[i] != 0) { + res.buf[i] = (__bridge void*) _sg_mtl_id(buf->mtl.buf[i]); + } + } + res.active_slot = buf->cmn.active_slot; + } + #else + _SOKOL_UNUSED(buf_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_mtl_image_info sg_mtl_query_image_info(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + sg_mtl_image_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_METAL) + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + if (img->mtl.tex[i] != 0) { + res.tex[i] = (__bridge void*) _sg_mtl_id(img->mtl.tex[i]); + } + } + res.active_slot = img->cmn.active_slot; + } + #else + _SOKOL_UNUSED(img_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_mtl_sampler_info sg_mtl_query_sampler_info(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + sg_mtl_sampler_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_METAL) + const _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + if (smp->mtl.sampler_state != 0) { + res.smp = (__bridge void*) _sg_mtl_id(smp->mtl.sampler_state); + } + } + #else + _SOKOL_UNUSED(smp_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_mtl_shader_info sg_mtl_query_shader_info(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + sg_mtl_shader_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_METAL) + const _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + const int vertex_lib = shd->mtl.vertex_func.mtl_lib; + const int vertex_func = shd->mtl.vertex_func.mtl_func; + const int fragment_lib = shd->mtl.fragment_func.mtl_lib; + const int fragment_func = shd->mtl.fragment_func.mtl_func; + if (vertex_lib != 0) { + res.vertex_lib = (__bridge void*) _sg_mtl_id(vertex_lib); + } + if (fragment_lib != 0) { + res.fragment_lib = (__bridge void*) _sg_mtl_id(fragment_lib); + } + if (vertex_func != 0) { + res.vertex_func = (__bridge void*) _sg_mtl_id(vertex_func); + } + if (fragment_func != 0) { + res.fragment_func = (__bridge void*) _sg_mtl_id(fragment_func); + } + } + #else + _SOKOL_UNUSED(shd_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_mtl_pipeline_info sg_mtl_query_pipeline_info(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + sg_mtl_pipeline_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_METAL) + const _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + if (pip->mtl.rps != 0) { + res.rps = (__bridge void*) _sg_mtl_id(pip->mtl.rps); + } + if (pip->mtl.dss != 0) { + res.dss = (__bridge void*) _sg_mtl_id(pip->mtl.dss); + } + } + #else + _SOKOL_UNUSED(pip_id); + #endif + return res; +} + +SOKOL_API_IMPL const void* sg_wgpu_device(void) { + #if defined(SOKOL_WGPU) + return (const void*) _sg.wgpu.dev; + #else + return 0; + #endif +} + +SOKOL_API_IMPL const void* sg_wgpu_queue(void) { + #if defined(SOKOL_WGPU) + return (const void*) _sg.wgpu.queue; + #else + return 0; + #endif +} + +SOKOL_API_IMPL const void* sg_wgpu_command_encoder(void) { + #if defined(SOKOL_WGPU) + return (const void*) _sg.wgpu.cmd_enc; + #else + return 0; + #endif +} + +SOKOL_API_IMPL const void* sg_wgpu_render_pass_encoder(void) { + #if defined(SOKOL_WGPU) + return (const void*) _sg.wgpu.rpass_enc; + #else + return 0; + #endif +} + +SOKOL_API_IMPL const void* sg_wgpu_compute_pass_encoder(void) { + #if defined(SOKOL_WGPU) + return (const void*) _sg.wgpu.cpass_enc; + #else + return 0; + #endif +} + +SOKOL_API_IMPL sg_wgpu_buffer_info sg_wgpu_query_buffer_info(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + sg_wgpu_buffer_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_WGPU) + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + res.buf = (const void*) buf->wgpu.buf; + } + #else + _SOKOL_UNUSED(buf_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_wgpu_image_info sg_wgpu_query_image_info(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + sg_wgpu_image_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_WGPU) + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + res.tex = (const void*) img->wgpu.tex; + res.view = (const void*) img->wgpu.view; + } + #else + _SOKOL_UNUSED(img_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_wgpu_sampler_info sg_wgpu_query_sampler_info(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + sg_wgpu_sampler_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_WGPU) + const _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + res.smp = (const void*) smp->wgpu.smp; + } + #else + _SOKOL_UNUSED(smp_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_wgpu_shader_info sg_wgpu_query_shader_info(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + sg_wgpu_shader_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_WGPU) + const _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + res.vs_mod = (const void*) shd->wgpu.vertex_func.module; + res.fs_mod = (const void*) shd->wgpu.fragment_func.module; + res.bgl = (const void*) shd->wgpu.bgl_img_smp_sbuf; + } + #else + _SOKOL_UNUSED(shd_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_wgpu_pipeline_info sg_wgpu_query_pipeline_info(sg_pipeline pip_id) { + SOKOL_ASSERT(_sg.valid); + sg_wgpu_pipeline_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_WGPU) + const _sg_pipeline_t* pip = _sg_lookup_pipeline(&_sg.pools, pip_id.id); + if (pip) { + res.render_pipeline = (const void*) pip->wgpu.rpip; + res.compute_pipeline = (const void*) pip->wgpu.cpip; + } + #else + _SOKOL_UNUSED(pip_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_wgpu_attachments_info sg_wgpu_query_attachments_info(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + sg_wgpu_attachments_info res; + _sg_clear(&res, sizeof(res)); + #if defined(SOKOL_WGPU) + const _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + res.color_view[i] = (const void*) atts->wgpu.colors[i].view; + res.resolve_view[i] = (const void*) atts->wgpu.resolves[i].view; + } + res.ds_view = (const void*) atts->wgpu.depth_stencil.view; + } + #else + _SOKOL_UNUSED(atts_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_gl_buffer_info sg_gl_query_buffer_info(sg_buffer buf_id) { + SOKOL_ASSERT(_sg.valid); + sg_gl_buffer_info res; + _sg_clear(&res, sizeof(res)); + #if defined(_SOKOL_ANY_GL) + const _sg_buffer_t* buf = _sg_lookup_buffer(&_sg.pools, buf_id.id); + if (buf) { + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + res.buf[i] = buf->gl.buf[i]; + } + res.active_slot = buf->cmn.active_slot; + } + #else + _SOKOL_UNUSED(buf_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_gl_image_info sg_gl_query_image_info(sg_image img_id) { + SOKOL_ASSERT(_sg.valid); + sg_gl_image_info res; + _sg_clear(&res, sizeof(res)); + #if defined(_SOKOL_ANY_GL) + const _sg_image_t* img = _sg_lookup_image(&_sg.pools, img_id.id); + if (img) { + for (int i = 0; i < SG_NUM_INFLIGHT_FRAMES; i++) { + res.tex[i] = img->gl.tex[i]; + } + res.tex_target = img->gl.target; + res.msaa_render_buffer = img->gl.msaa_render_buffer; + res.active_slot = img->cmn.active_slot; + } + #else + _SOKOL_UNUSED(img_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_gl_sampler_info sg_gl_query_sampler_info(sg_sampler smp_id) { + SOKOL_ASSERT(_sg.valid); + sg_gl_sampler_info res; + _sg_clear(&res, sizeof(res)); + #if defined(_SOKOL_ANY_GL) + const _sg_sampler_t* smp = _sg_lookup_sampler(&_sg.pools, smp_id.id); + if (smp) { + res.smp = smp->gl.smp; + } + #else + _SOKOL_UNUSED(smp_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_gl_shader_info sg_gl_query_shader_info(sg_shader shd_id) { + SOKOL_ASSERT(_sg.valid); + sg_gl_shader_info res; + _sg_clear(&res, sizeof(res)); + #if defined(_SOKOL_ANY_GL) + const _sg_shader_t* shd = _sg_lookup_shader(&_sg.pools, shd_id.id); + if (shd) { + res.prog = shd->gl.prog; + } + #else + _SOKOL_UNUSED(shd_id); + #endif + return res; +} + +SOKOL_API_IMPL sg_gl_attachments_info sg_gl_query_attachments_info(sg_attachments atts_id) { + SOKOL_ASSERT(_sg.valid); + sg_gl_attachments_info res; + _sg_clear(&res, sizeof(res)); + #if defined(_SOKOL_ANY_GL) + const _sg_attachments_t* atts = _sg_lookup_attachments(&_sg.pools, atts_id.id); + if (atts) { + res.framebuffer = atts->gl.fb; + for (int i = 0; i < SG_MAX_COLOR_ATTACHMENTS; i++) { + res.msaa_resolve_framebuffer[i] = atts->gl.msaa_resolve_framebuffer[i]; + } + } + #else + _SOKOL_UNUSED(atts_id); + #endif + return res; +} + +#ifdef _MSC_VER +#pragma warning(pop) +#endif + +#endif // SOKOL_GFX_IMPL diff --git a/platform/sokol_impl.c b/platform/sokol_impl.c new file mode 100644 index 0000000..acd59cf --- /dev/null +++ b/platform/sokol_impl.c @@ -0,0 +1,191 @@ +// Implementation of platform code when using Sokol +// The platform must define backends (i.e. SOKOL_GLES3) +// and other platform-specific goodies. Then this +// file takes care of the rest! +#include "sokol_gfx.h" +#include "os.h" + +static void sokol_logger(const char *tag, uint32_t verbosity, uint32_t id, const char *msg_or_null, uint32_t line, const char *fname_or_null, void *unused) +{ + (void)unused; + (void)tag; + (void)id; + OSLOG(verbosity + LOG_FATAL, "From sokol graphics! [%s:%u] %s", fname_or_null, line, msg_or_null); +} + +// TODO: +// Sokol does support custom allocators +// Platform GFX allocators aren't exposed, but could be useful for debugging +void platform_gfx_startup(union platform_window window) +{ + if (!platform_gfx_valid(window)) return; + + sg_environment env = { + .defaults = { + .color_format = SG_PIXELFORMAT_RGBA8, + .depth_format = SG_PIXELFORMAT_DEPTH_STENCIL, + .sample_count = 1, + }, + }; + sg_setup(&(sg_desc){ + .environment = env, + .logger.func = sokol_logger, + }); +} +void platform_gfx_render_begin(union platform_window window) +{ + if (!platform_gfx_valid(window)) return; + + sg_swapchain sc = { + .width = platform_window_width(window), + .height = platform_window_height(window), + .color_format = SG_PIXELFORMAT_RGBA8, + .depth_format = SG_PIXELFORMAT_DEPTH_STENCIL, + .sample_count = 1, + }; + sg_pass_action pa = { + .colors[0] = { + .load_action = SG_LOADACTION_CLEAR, + .clear_value = { 0.0f, 0.0f, 0.0f, 1.0f } + } + }; + sg_begin_pass(&(sg_pass){ .action = pa, .swapchain = sc }); +} +void platform_gfx_render_end(union platform_window window) +{ + if (!platform_gfx_valid(window)) return; + + sg_end_pass(); + sg_commit(); +} +void platform_gfx_shutdown(union platform_window window) +{ + if (!platform_gfx_valid(window)) return; + + sg_shutdown(); +} + +union platform_gfx_shader platform_gfx_shader_construct(union platform_window window, const struct platform_gfx_shader_desc *desc) +{ + union platform_gfx_shader s = {}; + if (!platform_gfx_valid(window)) return s; + + s.sgid = sg_make_shader(&(sg_shader_desc){ + .vertex_func.source = desc->vertex, + .fragment_func.source = desc->fragment, + .images[0].stage = SG_SHADERSTAGE_FRAGMENT, + .samplers[0].stage = SG_SHADERSTAGE_FRAGMENT, + .image_sampler_pairs[0] = { + .stage = SG_SHADERSTAGE_FRAGMENT, + .image_slot = 0, + .sampler_slot = 0, + .glsl_name = "tex", + }, + }).id; + return s; +} +void platform_gfx_shader_destruct(union platform_window window, union platform_gfx_shader s) +{ + if (!platform_gfx_valid(window)) return; + + sg_destroy_shader((sg_shader){s.sgid}); +} + +union platform_gfx_pipeline platform_gfx_pipeline_construct(union platform_window window, const struct platform_gfx_pipeline_desc *desc) +{ + union platform_gfx_pipeline p = {}; + if (!platform_gfx_valid(window)) return p; + + p.sgid = sg_make_pipeline(&(sg_pipeline_desc){ + .shader = (sg_shader){desc->shader.sgid}, + .layout = { + .attrs = { + [0].format=SG_VERTEXFORMAT_FLOAT2, + [1].format=SG_VERTEXFORMAT_FLOAT2, + }, + }, + .colors[0].blend = { + .enabled = true, + .src_factor_rgb = SG_BLENDFACTOR_SRC_ALPHA, + .dst_factor_rgb = SG_BLENDFACTOR_ONE_MINUS_SRC_ALPHA, + }, + }).id; + return p; +} +void platform_gfx_pipeline_destruct(union platform_window window, union platform_gfx_pipeline p) +{ + if (!platform_gfx_valid(window)) return; + + sg_destroy_pipeline((sg_pipeline){p.sgid}); +} +void platform_gfx_pipeline_bind(union platform_window window, union platform_gfx_pipeline p) +{ + if (!platform_gfx_valid(window)) return; + + sg_apply_pipeline((sg_pipeline){p.sgid}); +} + +union platform_gfx_texture platform_gfx_texture_construct(union platform_window window, const struct platform_gfx_texture_desc *desc) +{ + union platform_gfx_texture t = {}; + if (!platform_gfx_valid(window)) return t; + + int format = 0; + switch (desc->format) { + case IMAGE_FORMAT_RGBX8: format = SG_PIXELFORMAT_RGBA8; break; + case IMAGE_FORMAT_RGBA8: format = SG_PIXELFORMAT_RGBA8; break; + case IMAGE_FORMAT_R8: format = SG_PIXELFORMAT_R8; break; + } + t.sgid = sg_make_image(&(sg_image_desc){ + .width = desc->width, + .height = desc->height, + .data.subimage[0][0] = { .ptr = desc->pixels, .size = desc->bytesize }, + .pixel_format = format, + .sample_count = 1, + }).id; + return t; +} +void platform_gfx_texture_destruct(union platform_window window, union platform_gfx_texture texture) +{ + if (!platform_gfx_valid(window)) return; + + sg_destroy_image((sg_image){texture.sgid}); +} + +union platform_gfx_sampler platform_gfx_sampler_construct(union platform_window window, const struct platform_gfx_sampler_desc *desc) +{ + union platform_gfx_sampler s = {}; + if (!platform_gfx_valid(window)) return s; + + // TODO: Allow specification of filters and the like + (void)desc; + s.sgid = sg_make_sampler(&(sg_sampler_desc){ + .min_filter = SG_FILTER_NEAREST, + .mag_filter = SG_FILTER_NEAREST, + }).id; + return s; +} +void platform_gfx_sampler_destruct(union platform_window window, union platform_gfx_sampler s) +{ + if (!platform_gfx_valid(window)) return; + + sg_destroy_sampler((sg_sampler){s.sgid}); +} + +// TODO: Should bind textures and samplers rather than just have one +// Thankfully is good enough to deal with the simple 2D case now... +void platform_gfx_vertices_draw(union platform_window window, union platform_gfx_texture texture, union platform_gfx_sampler sampler, const void *data, size_t bytelen, size_t vertices_count) +{ + (void)texture; + (void)sampler; + if (!platform_gfx_valid(window)) return; + + sg_buffer buf = sg_make_buffer(&(sg_buffer_desc){ .data = { .ptr = data, .size = bytelen } }); + sg_apply_bindings(&(sg_bindings){ + .vertex_buffers = { buf }, + .images[0] = (sg_image){texture.sgid}, + .samplers[0] = (sg_sampler){sampler.sgid}, + }); + sg_draw(0, vertices_count, 1); + sg_destroy_buffer(buf); +} diff --git a/platform/stb_image.h b/platform/stb_image.h new file mode 100644 index 0000000..9eedabe --- /dev/null +++ b/platform/stb_image.h @@ -0,0 +1,7988 @@ +/* stb_image - v2.30 - public domain image loader - http://nothings.org/stb + no warranty implied; use at your own risk + + Do this: + #define STB_IMAGE_IMPLEMENTATION + before you include this file in *one* C or C++ file to create the implementation. + + // i.e. it should look like this: + #include ... + #include ... + #include ... + #define STB_IMAGE_IMPLEMENTATION + #include "stb_image.h" + + You can #define STBI_ASSERT(x) before the #include to avoid using assert.h. + And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free + + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib) + PNG 1/2/4/8/16-bit-per-channel + + TGA (not sure what subset, if a subset) + BMP non-1bpp, non-RLE + PSD (composited view only, no extra channels, 8/16 bit-per-channel) + + GIF (*comp always reports as 4-channel) + HDR (radiance rgbE format) + PIC (Softimage PIC) + PNM (PPM and PGM binary only) + + Animated GIF still needs a proper API, but here's one way to do it: + http://gist.github.com/urraka/685d9a6340b26b830d49 + + - decode from memory or through FILE (define STBI_NO_STDIO to remove code) + - decode from arbitrary I/O callbacks + - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON) + + Full documentation under "DOCUMENTATION" below. + + +LICENSE + + See end of file for license information. + +RECENT REVISION HISTORY: + + 2.30 (2024-05-31) avoid erroneous gcc warning + 2.29 (2023-05-xx) optimizations + 2.28 (2023-01-29) many error fixes, security errors, just tons of stuff + 2.27 (2021-07-11) document stbi_info better, 16-bit PNM support, bug fixes + 2.26 (2020-07-13) many minor fixes + 2.25 (2020-02-02) fix warnings + 2.24 (2020-02-02) fix warnings; thread-local failure_reason and flip_vertically + 2.23 (2019-08-11) fix clang static analysis warning + 2.22 (2019-03-04) gif fixes, fix warnings + 2.21 (2019-02-25) fix typo in comment + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings + 2.16 (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes + 2.15 (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-12-04) experimental 16-bit API, only for PNG so far; fixes + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64 + RGB-format JPEG; remove white matting in PSD; + allocate large structures on the stack; + correct channel count for PNG & BMP + 2.10 (2016-01-22) avoid warning introduced in 2.09 + 2.09 (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED + + See end of file for full revision history. + + + ============================ Contributors ========================= + + Image formats Extensions, features + Sean Barrett (jpeg, png, bmp) Jetro Lauha (stbi_info) + Nicolas Schulz (hdr, psd) Martin "SpartanJ" Golini (stbi_info) + Jonathan Dummer (tga) James "moose2000" Brown (iPhone PNG) + Jean-Marc Lienher (gif) Ben "Disch" Wenger (io callbacks) + Tom Seddon (pic) Omar Cornut (1/2/4-bit PNG) + Thatcher Ulrich (psd) Nicolas Guillemot (vertical flip) + Ken Miller (pgm, ppm) Richard Mitton (16-bit PSD) + github:urraka (animated gif) Junggon Kim (PNM comments) + Christopher Forseth (animated gif) Daniel Gibson (16-bit TGA) + socks-the-fox (16-bit PNG) + Jeremy Sawicki (handle all ImageNet JPGs) + Optimizations & bugfixes Mikhail Morozov (1-bit BMP) + Fabian "ryg" Giesen Anael Seghezzi (is-16-bit query) + Arseny Kapoulkine Simon Breuss (16-bit PNM) + John-Mark Allen + Carmelo J Fdez-Aguera + + Bug & warning fixes + Marc LeBlanc David Woo Guillaume George Martins Mozeiko + Christpher Lloyd Jerry Jansson Joseph Thomson Blazej Dariusz Roszkowski + Phil Jordan Dave Moore Roy Eltham + Hayaki Saito Nathan Reed Won Chun + Luke Graham Johan Duparc Nick Verigakis the Horde3D community + Thomas Ruf Ronny Chevalier github:rlyeh + Janez Zemva John Bartholomew Michal Cichon github:romigrou + Jonathan Blow Ken Hamada Tero Hanninen github:svdijk + Eugene Golushkov Laurent Gomila Cort Stratton github:snagar + Aruelien Pocheville Sergio Gonzalez Thibault Reuille github:Zelex + Cass Everitt Ryamond Barbiero github:grim210 + Paul Du Bois Engin Manap Aldo Culquicondor github:sammyhw + Philipp Wiesemann Dale Weiler Oriol Ferrer Mesia github:phprus + Josh Tobin Neil Bickford Matthew Gregan github:poppolopoppo + Julian Raschke Gregory Mullen Christian Floisand github:darealshinji + Baldur Karlsson Kevin Schmidt JR Smith github:Michaelangel007 + Brad Weinberger Matvey Cherevko github:mosra + Luca Sas Alexander Veselov Zack Middleton [reserved] + Ryan C. Gordon [reserved] [reserved] + DO NOT ADD YOUR NAME HERE + + Jacko Dirks + + To add your name to the credits, pick a random blank space in the middle and fill it. + 80% of merge conflicts on stb PRs are due to people adding their name at the end + of the credits. +*/ + +#ifndef STBI_INCLUDE_STB_IMAGE_H +#define STBI_INCLUDE_STB_IMAGE_H + +// DOCUMENTATION +// +// Limitations: +// - no 12-bit-per-channel JPEG +// - no JPEGs with arithmetic coding +// - GIF always returns *comp=4 +// +// Basic usage (see HDR discussion below for HDR usage): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// // ... but 'n' will always be the number that it would have been if you said 0 +// stbi_image_free(data); +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *channels_in_file -- outputs # of image components in image file +// int desired_channels -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data, or NULL on an allocation failure or if the image is +// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'desired_channels' if desired_channels is non-zero, or +// *channels_in_file otherwise. If desired_channels is non-zero, +// *channels_in_file has the number of components that _would_ have been +// output otherwise. E.g. if you set desired_channels to 4, you will always +// get RGBA output, but you can check *channels_in_file to see if it's trivially +// opaque because e.g. there were only 3 channels in the source image. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *channels_in_file will be unchanged. The function +// stbi_failure_reason() can be queried for an extremely brief, end-user +// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS +// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized. +// +// To query the width, height and component count of an image without having to +// decode the full file, you can use the stbi_info family of functions: +// +// int x,y,n,ok; +// ok = stbi_info(filename, &x, &y, &n); +// // returns ok=1 and sets x, y, n if image is a supported format, +// // 0 otherwise. +// +// Note that stb_image pervasively uses ints in its public API for sizes, +// including sizes of memory buffers. This is now part of the API and thus +// hard to change without causing breakage. As a result, the various image +// loaders all have certain limits on image size; these differ somewhat +// by format but generally boil down to either just under 2GB or just under +// 1GB. When the decoded image would be larger than this, stb_image decoding +// will fail. +// +// Additionally, stb_image will reject image files that have any of their +// dimensions set to a larger value than the configurable STBI_MAX_DIMENSIONS, +// which defaults to 2**24 = 16777216 pixels. Due to the above memory limit, +// the only way to have an image with such dimensions load correctly +// is for it to have a rather extreme aspect ratio. Either way, the +// assumption here is that such larger images are likely to be malformed +// or malicious. If you do need to load an image with individual dimensions +// larger than that, and it still fits in the overall size limit, you can +// #define STBI_MAX_DIMENSIONS on your own to be something larger. +// +// =========================================================================== +// +// UNICODE: +// +// If compiling for Windows and you wish to use Unicode filenames, compile +// with +// #define STBI_WINDOWS_UTF8 +// and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert +// Windows wchar_t filenames to utf8. +// +// =========================================================================== +// +// Philosophy +// +// stb libraries are designed with the following priorities: +// +// 1. easy to use +// 2. easy to maintain +// 3. good performance +// +// Sometimes I let "good performance" creep up in priority over "easy to maintain", +// and for best performance I may provide less-easy-to-use APIs that give higher +// performance, in addition to the easy-to-use ones. Nevertheless, it's important +// to keep in mind that from the standpoint of you, a client of this library, +// all you care about is #1 and #3, and stb libraries DO NOT emphasize #3 above all. +// +// Some secondary priorities arise directly from the first two, some of which +// provide more explicit reasons why performance can't be emphasized. +// +// - Portable ("ease of use") +// - Small source code footprint ("easy to maintain") +// - No dependencies ("ease of use") +// +// =========================================================================== +// +// I/O callbacks +// +// I/O callbacks allow you to read from arbitrary sources, like packaged +// files or some other source. Data read from callbacks are processed +// through a small internal buffer (currently 128 bytes) to try to reduce +// overhead. +// +// The three functions you must define are "read" (reads some bytes of data), +// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end). +// +// =========================================================================== +// +// SIMD support +// +// The JPEG decoder will try to automatically use SIMD kernels on x86 when +// supported by the compiler. For ARM Neon support, you must explicitly +// request it. +// +// (The old do-it-yourself SIMD API is no longer supported in the current +// code.) +// +// On x86, SSE2 will automatically be used when available based on a run-time +// test; if not, the generic C versions are used as a fall-back. On ARM targets, +// the typical path is to have separate builds for NEON and non-NEON devices +// (at least this is true for iOS and Android). Therefore, the NEON support is +// toggled by a build flag: define STBI_NEON to get NEON loops. +// +// If for some reason you do not want to use any of SIMD code, or if +// you have issues compiling it, you can disable it entirely by +// defining STBI_NO_SIMD. +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image supports loading HDR images in general, and currently the Radiance +// .HDR file format specifically. You can still load any file through the existing +// interface; if you attempt to load an HDR file, it will be automatically remapped +// to LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); +// +// =========================================================================== +// +// iPhone PNG support: +// +// We optionally support converting iPhone-formatted PNGs (which store +// premultiplied BGRA) back to RGB, even though they're internally encoded +// differently. To enable this conversion, call +// stbi_convert_iphone_png_to_rgb(1). +// +// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per +// pixel to remove any premultiplied alpha *only* if the image file explicitly +// says there's premultiplied data (currently only happens in iPhone images, +// and only if iPhone convert-to-rgb processing is on). +// +// =========================================================================== +// +// ADDITIONAL CONFIGURATION +// +// - You can suppress implementation of any of the decoders to reduce +// your code footprint by #defining one or more of the following +// symbols before creating the implementation. +// +// STBI_NO_JPEG +// STBI_NO_PNG +// STBI_NO_BMP +// STBI_NO_PSD +// STBI_NO_TGA +// STBI_NO_GIF +// STBI_NO_HDR +// STBI_NO_PIC +// STBI_NO_PNM (.ppm and .pgm) +// +// - You can request *only* certain decoders and suppress all other ones +// (this will be more forward-compatible, as addition of new decoders +// doesn't require you to disable them explicitly): +// +// STBI_ONLY_JPEG +// STBI_ONLY_PNG +// STBI_ONLY_BMP +// STBI_ONLY_PSD +// STBI_ONLY_TGA +// STBI_ONLY_GIF +// STBI_ONLY_HDR +// STBI_ONLY_PIC +// STBI_ONLY_PNM (.ppm and .pgm) +// +// - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still +// want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB +// +// - If you define STBI_MAX_DIMENSIONS, stb_image will reject images greater +// than that size (in either width or height) without further processing. +// This is to let programs in the wild set an upper bound to prevent +// denial-of-service attacks on untrusted data, as one could generate a +// valid image of gigantic dimensions and force stb_image to allocate a +// huge block of memory and spend disproportionate time decoding it. By +// default this is set to (1 << 24), which is 16777216, but that's still +// very big. + +#ifndef STBI_NO_STDIO +#include +#endif // STBI_NO_STDIO + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for desired_channels + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4 +}; + +#include +typedef unsigned char stbi_uc; +typedef unsigned short stbi_us; + +#ifdef __cplusplus +extern "C" { +#endif + +#ifndef STBIDEF +#ifdef STB_IMAGE_STATIC +#define STBIDEF static +#else +#define STBIDEF extern +#endif +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// PRIMARY API - works on images of any type +// + +// +// load image by filename, open file, or memory buffer +// + +typedef struct +{ + int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read + void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative + int (*eof) (void *user); // returns nonzero if we are at end of file/data +} stbi_io_callbacks; + +//////////////////////////////////// +// +// 8-bits-per-channel interface +// + +STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +// for stbi_load_from_file, file pointer is left pointing immediately after image +#endif + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +#endif + +#ifdef STBI_WINDOWS_UTF8 +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input); +#endif + +//////////////////////////////////// +// +// 16-bits-per-channel interface +// + +STBIDEF stbi_us *stbi_load_16_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_us *stbi_load_16 (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +#endif + +//////////////////////////////////// +// +// float-per-channel interface +// +#ifndef STBI_NO_LINEAR + STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + + #ifndef STBI_NO_STDIO + STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); + #endif +#endif + +#ifndef STBI_NO_HDR + STBIDEF void stbi_hdr_to_ldr_gamma(float gamma); + STBIDEF void stbi_hdr_to_ldr_scale(float scale); +#endif // STBI_NO_HDR + +#ifndef STBI_NO_LINEAR + STBIDEF void stbi_ldr_to_hdr_gamma(float gamma); + STBIDEF void stbi_ldr_to_hdr_scale(float scale); +#endif // STBI_NO_LINEAR + +// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user); +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len); +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename); +STBIDEF int stbi_is_hdr_from_file(FILE *f); +#endif // STBI_NO_STDIO + + +// get a VERY brief reason for failure +// on most compilers (and ALL modern mainstream compilers) this is threadsafe +STBIDEF const char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +STBIDEF void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp); +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len); +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user); + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info (char const *filename, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +STBIDEF int stbi_is_16_bit (char const *filename); +STBIDEF int stbi_is_16_bit_from_file(FILE *f); +#endif + + + +// for image formats that explicitly notate that they have premultiplied alpha, +// we just return the colors as stored in the file. set this flag to force +// unpremultiplication. results are undefined if the unpremultiply overflow. +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply); + +// indicate whether we should process iphone images back to canonical format, +// or just pass them through "as-is" +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert); + +// flip the image vertically, so the first pixel in the output array is the bottom left +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip); + +// as above, but only applies to images loaded on the thread that calls the function +// this function is only available if your compiler supports thread-local variables; +// calling it will fail to link if your compiler doesn't +STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply); +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert); +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip); + +// ZLIB client - used by PNG, available for other purposes + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen); +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header); +STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // STBI_INCLUDE_STB_IMAGE_H + +#ifdef STB_IMAGE_IMPLEMENTATION + +#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \ + || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \ + || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \ + || defined(STBI_ONLY_ZLIB) + #ifndef STBI_ONLY_JPEG + #define STBI_NO_JPEG + #endif + #ifndef STBI_ONLY_PNG + #define STBI_NO_PNG + #endif + #ifndef STBI_ONLY_BMP + #define STBI_NO_BMP + #endif + #ifndef STBI_ONLY_PSD + #define STBI_NO_PSD + #endif + #ifndef STBI_ONLY_TGA + #define STBI_NO_TGA + #endif + #ifndef STBI_ONLY_GIF + #define STBI_NO_GIF + #endif + #ifndef STBI_ONLY_HDR + #define STBI_NO_HDR + #endif + #ifndef STBI_ONLY_PIC + #define STBI_NO_PIC + #endif + #ifndef STBI_ONLY_PNM + #define STBI_NO_PNM + #endif +#endif + +#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB) +#define STBI_NO_ZLIB +#endif + + +#include +#include // ptrdiff_t on osx +#include +#include +#include + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) +#include // ldexp, pow +#endif + +#ifndef STBI_NO_STDIO +#include +#endif + +#ifndef STBI_ASSERT +#include +#define STBI_ASSERT(x) assert(x) +#endif + +#ifdef __cplusplus +#define STBI_EXTERN extern "C" +#else +#define STBI_EXTERN extern +#endif + + +#ifndef _MSC_VER + #ifdef __cplusplus + #define stbi_inline inline + #else + #define stbi_inline + #endif +#else + #define stbi_inline __forceinline +#endif + +#ifndef STBI_NO_THREAD_LOCALS + #if defined(__cplusplus) && __cplusplus >= 201103L + #define STBI_THREAD_LOCAL thread_local + #elif defined(__GNUC__) && __GNUC__ < 5 + #define STBI_THREAD_LOCAL __thread + #elif defined(_MSC_VER) + #define STBI_THREAD_LOCAL __declspec(thread) + #elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L && !defined(__STDC_NO_THREADS__) + #define STBI_THREAD_LOCAL _Thread_local + #endif + + #ifndef STBI_THREAD_LOCAL + #if defined(__GNUC__) + #define STBI_THREAD_LOCAL __thread + #endif + #endif +#endif + +#if defined(_MSC_VER) || defined(__SYMBIAN32__) +typedef unsigned short stbi__uint16; +typedef signed short stbi__int16; +typedef unsigned int stbi__uint32; +typedef signed int stbi__int32; +#else +#include +typedef uint16_t stbi__uint16; +typedef int16_t stbi__int16; +typedef uint32_t stbi__uint32; +typedef int32_t stbi__int32; +#endif + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1]; + +#ifdef _MSC_VER +#define STBI_NOTUSED(v) (void)(v) +#else +#define STBI_NOTUSED(v) (void)sizeof(v) +#endif + +#ifdef _MSC_VER +#define STBI_HAS_LROTL +#endif + +#ifdef STBI_HAS_LROTL + #define stbi_lrot(x,y) _lrotl(x,y) +#else + #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (-(y) & 31))) +#endif + +#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED)) +// ok +#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED) +// ok +#else +#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)." +#endif + +#ifndef STBI_MALLOC +#define STBI_MALLOC(sz) malloc(sz) +#define STBI_REALLOC(p,newsz) realloc(p,newsz) +#define STBI_FREE(p) free(p) +#endif + +#ifndef STBI_REALLOC_SIZED +#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz) +#endif + +// x86/x64 detection +#if defined(__x86_64__) || defined(_M_X64) +#define STBI__X64_TARGET +#elif defined(__i386) || defined(_M_IX86) +#define STBI__X86_TARGET +#endif + +#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__SSE2__) && !defined(STBI_NO_SIMD) +// gcc doesn't support sse2 intrinsics unless you compile with -msse2, +// which in turn means it gets to use SSE2 everywhere. This is unfortunate, +// but previous attempts to provide the SSE2 functions with runtime +// detection caused numerous issues. The way architecture extensions are +// exposed in GCC/Clang is, sadly, not really suited for one-file libs. +// New behavior: if compiled with -msse2, we use SSE2 without any +// detection; if not, we don't use it at all. +#define STBI_NO_SIMD +#endif + +#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD) +// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET +// +// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the +// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant. +// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not +// simultaneously enabling "-mstackrealign". +// +// See https://github.com/nothings/stb/issues/81 for more information. +// +// So default to no SSE2 on 32-bit MinGW. If you've read this far and added +// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2. +#define STBI_NO_SIMD +#endif + +#if !defined(STBI_NO_SIMD) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET)) +#define STBI_SSE2 +#include + +#ifdef _MSC_VER + +#if _MSC_VER >= 1400 // not VC6 +#include // __cpuid +static int stbi__cpuid3(void) +{ + int info[4]; + __cpuid(info,1); + return info[3]; +} +#else +static int stbi__cpuid3(void) +{ + int res; + __asm { + mov eax,1 + cpuid + mov res,edx + } + return res; +} +#endif + +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + int info3 = stbi__cpuid3(); + return ((info3 >> 26) & 1) != 0; +} +#endif + +#else // assume GCC-style if not VC++ +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + // If we're even attempting to compile this on GCC/Clang, that means + // -msse2 is on, which means the compiler is allowed to use SSE2 + // instructions at will, and so are we. + return 1; +} +#endif + +#endif +#endif + +// ARM NEON +#if defined(STBI_NO_SIMD) && defined(STBI_NEON) +#undef STBI_NEON +#endif + +#ifdef STBI_NEON +#include +#ifdef _MSC_VER +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name +#else +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) +#endif +#endif + +#ifndef STBI_SIMD_ALIGN +#define STBI_SIMD_ALIGN(type, name) type name +#endif + +#ifndef STBI_MAX_DIMENSIONS +#define STBI_MAX_DIMENSIONS (1 << 24) +#endif + +/////////////////////////////////////////////// +// +// stbi__context struct and start_xxx functions + +// stbi__context structure is our basic context used by all images, so it +// contains all the IO context, plus some basic image information +typedef struct +{ + stbi__uint32 img_x, img_y; + int img_n, img_out_n; + + stbi_io_callbacks io; + void *io_user_data; + + int read_from_callbacks; + int buflen; + stbi_uc buffer_start[128]; + int callback_already_read; + + stbi_uc *img_buffer, *img_buffer_end; + stbi_uc *img_buffer_original, *img_buffer_original_end; +} stbi__context; + + +static void stbi__refill_buffer(stbi__context *s); + +// initialize a memory-decode context +static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len) +{ + s->io.read = NULL; + s->read_from_callbacks = 0; + s->callback_already_read = 0; + s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer; + s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len; +} + +// initialize a callback-based context +static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user) +{ + s->io = *c; + s->io_user_data = user; + s->buflen = sizeof(s->buffer_start); + s->read_from_callbacks = 1; + s->callback_already_read = 0; + s->img_buffer = s->img_buffer_original = s->buffer_start; + stbi__refill_buffer(s); + s->img_buffer_original_end = s->img_buffer_end; +} + +#ifndef STBI_NO_STDIO + +static int stbi__stdio_read(void *user, char *data, int size) +{ + return (int) fread(data,1,size,(FILE*) user); +} + +static void stbi__stdio_skip(void *user, int n) +{ + int ch; + fseek((FILE*) user, n, SEEK_CUR); + ch = fgetc((FILE*) user); /* have to read a byte to reset feof()'s flag */ + if (ch != EOF) { + ungetc(ch, (FILE *) user); /* push byte back onto stream if valid. */ + } +} + +static int stbi__stdio_eof(void *user) +{ + return feof((FILE*) user) || ferror((FILE *) user); +} + +static stbi_io_callbacks stbi__stdio_callbacks = +{ + stbi__stdio_read, + stbi__stdio_skip, + stbi__stdio_eof, +}; + +static void stbi__start_file(stbi__context *s, FILE *f) +{ + stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f); +} + +//static void stop_file(stbi__context *s) { } + +#endif // !STBI_NO_STDIO + +static void stbi__rewind(stbi__context *s) +{ + // conceptually rewind SHOULD rewind to the beginning of the stream, + // but we just rewind to the beginning of the initial buffer, because + // we only use it after doing 'test', which only ever looks at at most 92 bytes + s->img_buffer = s->img_buffer_original; + s->img_buffer_end = s->img_buffer_original_end; +} + +enum +{ + STBI_ORDER_RGB, + STBI_ORDER_BGR +}; + +typedef struct +{ + int bits_per_channel; + int num_channels; + int channel_order; +} stbi__result_info; + +#ifndef STBI_NO_JPEG +static int stbi__jpeg_test(stbi__context *s); +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNG +static int stbi__png_test(stbi__context *s); +static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__png_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_BMP +static int stbi__bmp_test(stbi__context *s); +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_TGA +static int stbi__tga_test(stbi__context *s); +static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s); +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc); +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__psd_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_HDR +static int stbi__hdr_test(stbi__context *s); +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_test(stbi__context *s); +static void *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_GIF +static int stbi__gif_test(stbi__context *s); +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNM +static int stbi__pnm_test(stbi__context *s); +static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__pnm_is16(stbi__context *s); +#endif + +static +#ifdef STBI_THREAD_LOCAL +STBI_THREAD_LOCAL +#endif +const char *stbi__g_failure_reason; + +STBIDEF const char *stbi_failure_reason(void) +{ + return stbi__g_failure_reason; +} + +#ifndef STBI_NO_FAILURE_STRINGS +static int stbi__err(const char *str) +{ + stbi__g_failure_reason = str; + return 0; +} +#endif + +static void *stbi__malloc(size_t size) +{ + return STBI_MALLOC(size); +} + +// stb_image uses ints pervasively, including for offset calculations. +// therefore the largest decoded image size we can support with the +// current code, even on 64-bit targets, is INT_MAX. this is not a +// significant limitation for the intended use case. +// +// we do, however, need to make sure our size calculations don't +// overflow. hence a few helper functions for size calculations that +// multiply integers together, making sure that they're non-negative +// and no overflow occurs. + +// return 1 if the sum is valid, 0 on overflow. +// negative terms are considered invalid. +static int stbi__addsizes_valid(int a, int b) +{ + if (b < 0) return 0; + // now 0 <= b <= INT_MAX, hence also + // 0 <= INT_MAX - b <= INTMAX. + // And "a + b <= INT_MAX" (which might overflow) is the + // same as a <= INT_MAX - b (no overflow) + return a <= INT_MAX - b; +} + +// returns 1 if the product is valid, 0 on overflow. +// negative factors are considered invalid. +static int stbi__mul2sizes_valid(int a, int b) +{ + if (a < 0 || b < 0) return 0; + if (b == 0) return 1; // mul-by-0 is always safe + // portable way to check for no overflows in a*b + return a <= INT_MAX/b; +} + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow +static int stbi__mad2sizes_valid(int a, int b, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add); +} +#endif + +// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow +static int stbi__mad3sizes_valid(int a, int b, int c, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__addsizes_valid(a*b*c, add); +} + +// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add); +} +#endif + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// mallocs with size overflow checking +static void *stbi__malloc_mad2(int a, int b, int add) +{ + if (!stbi__mad2sizes_valid(a, b, add)) return NULL; + return stbi__malloc(a*b + add); +} +#endif + +static void *stbi__malloc_mad3(int a, int b, int c, int add) +{ + if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL; + return stbi__malloc(a*b*c + add); +} + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static void *stbi__malloc_mad4(int a, int b, int c, int d, int add) +{ + if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL; + return stbi__malloc(a*b*c*d + add); +} +#endif + +// returns 1 if the sum of two signed ints is valid (between -2^31 and 2^31-1 inclusive), 0 on overflow. +static int stbi__addints_valid(int a, int b) +{ + if ((a >= 0) != (b >= 0)) return 1; // a and b have different signs, so no overflow + if (a < 0 && b < 0) return a >= INT_MIN - b; // same as a + b >= INT_MIN; INT_MIN - b cannot overflow since b < 0. + return a <= INT_MAX - b; +} + +// returns 1 if the product of two ints fits in a signed short, 0 on overflow. +static int stbi__mul2shorts_valid(int a, int b) +{ + if (b == 0 || b == -1) return 1; // multiplication by 0 is always 0; check for -1 so SHRT_MIN/b doesn't overflow + if ((a >= 0) == (b >= 0)) return a <= SHRT_MAX/b; // product is positive, so similar to mul2sizes_valid + if (b < 0) return a <= SHRT_MIN / b; // same as a * b >= SHRT_MIN + return a >= SHRT_MIN / b; +} + +// stbi__err - error +// stbi__errpf - error returning pointer to float +// stbi__errpuc - error returning pointer to unsigned char + +#ifdef STBI_NO_FAILURE_STRINGS + #define stbi__err(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define stbi__err(x,y) stbi__err(y) +#else + #define stbi__err(x,y) stbi__err(x) +#endif + +#define stbi__errpf(x,y) ((float *)(size_t) (stbi__err(x,y)?NULL:NULL)) +#define stbi__errpuc(x,y) ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL)) + +STBIDEF void stbi_image_free(void *retval_from_stbi_load) +{ + STBI_FREE(retval_from_stbi_load); +} + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +#endif + +#ifndef STBI_NO_HDR +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +static int stbi__vertically_flip_on_load_global = 0; + +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_global = flag_true_if_should_flip; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__vertically_flip_on_load stbi__vertically_flip_on_load_global +#else +static STBI_THREAD_LOCAL int stbi__vertically_flip_on_load_local, stbi__vertically_flip_on_load_set; + +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_local = flag_true_if_should_flip; + stbi__vertically_flip_on_load_set = 1; +} + +#define stbi__vertically_flip_on_load (stbi__vertically_flip_on_load_set \ + ? stbi__vertically_flip_on_load_local \ + : stbi__vertically_flip_on_load_global) +#endif // STBI_THREAD_LOCAL + +static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields + ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed + ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order + ri->num_channels = 0; + + // test the formats with a very explicit header first (at least a FOURCC + // or distinctive magic number first) + #ifndef STBI_NO_PNG + if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_BMP + if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_GIF + if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PSD + if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc); + #else + STBI_NOTUSED(bpc); + #endif + #ifndef STBI_NO_PIC + if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp, ri); + #endif + + // then the formats that can end up attempting to load with just 1 or 2 + // bytes matching expectations; these are prone to false positives, so + // try them later + #ifndef STBI_NO_JPEG + if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PNM + if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp, ri); + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri); + return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + + #ifndef STBI_NO_TGA + // test tga last because it's a crappy test! + if (stbi__tga_test(s)) + return stbi__tga_load(s,x,y,comp,req_comp, ri); + #endif + + return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt"); +} + +static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi_uc *reduced; + + reduced = (stbi_uc *) stbi__malloc(img_len); + if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling + + STBI_FREE(orig); + return reduced; +} + +static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi__uint16 *enlarged; + + enlarged = (stbi__uint16 *) stbi__malloc(img_len*2); + if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff + + STBI_FREE(orig); + return enlarged; +} + +static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel) +{ + int row; + size_t bytes_per_row = (size_t)w * bytes_per_pixel; + stbi_uc temp[2048]; + stbi_uc *bytes = (stbi_uc *)image; + + for (row = 0; row < (h>>1); row++) { + stbi_uc *row0 = bytes + row*bytes_per_row; + stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row; + // swap row0 with row1 + size_t bytes_left = bytes_per_row; + while (bytes_left) { + size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp); + memcpy(temp, row0, bytes_copy); + memcpy(row0, row1, bytes_copy); + memcpy(row1, temp, bytes_copy); + row0 += bytes_copy; + row1 += bytes_copy; + bytes_left -= bytes_copy; + } + } +} + +#ifndef STBI_NO_GIF +static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel) +{ + int slice; + int slice_size = w * h * bytes_per_pixel; + + stbi_uc *bytes = (stbi_uc *)image; + for (slice = 0; slice < z; ++slice) { + stbi__vertical_flip(bytes, w, h, bytes_per_pixel); + bytes += slice_size; + } +} +#endif + +static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 8) { + result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 8; + } + + // @TODO: move stbi__convert_format to here + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc)); + } + + return (unsigned char *) result; +} + +static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 16) { + result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 16; + } + + // @TODO: move stbi__convert_format16 to here + // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16)); + } + + return (stbi__uint16 *) result; +} + +#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR) +static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp) +{ + if (stbi__vertically_flip_on_load && result != NULL) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(float)); + } +} +#endif + +#ifndef STBI_NO_STDIO + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide); +STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default); +#endif + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input) +{ + return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL); +} +#endif + +static FILE *stbi__fopen(char const *filename, char const *mode) +{ + FILE *f; +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) + wchar_t wMode[64]; + wchar_t wFilename[1024]; + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename))) + return 0; + + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode))) + return 0; + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != _wfopen_s(&f, wFilename, wMode)) + f = 0; +#else + f = _wfopen(wFilename, wMode); +#endif + +#elif defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != fopen_s(&f, filename, mode)) + f=0; +#else + f = fopen(filename, mode); +#endif + return f; +} + + +STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + unsigned char *result; + if (!f) return stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} + +STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__uint16 *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_16bit(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} + +STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + stbi__uint16 *result; + if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file_16(f,x,y,comp,req_comp); + fclose(f); + return result; +} + + +#endif //!STBI_NO_STDIO + +STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_mem(&s,buffer,len); + + result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp); + if (stbi__vertically_flip_on_load) { + stbi__vertical_flip_slices( result, *x, *y, *z, *comp ); + } + + return result; +} +#endif + +#ifndef STBI_NO_LINEAR +static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + stbi__result_info ri; + float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri); + if (hdr_data) + stbi__float_postprocess(hdr_data,x,y,comp,req_comp); + return hdr_data; + } + #endif + data = stbi__load_and_postprocess_8bit(s, x, y, comp, req_comp); + if (data) + return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return stbi__errpf("unknown image type", "Image not of any known type, or corrupt"); +} + +STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + float *result; + FILE *f = stbi__fopen(filename, "rb"); + if (!f) return stbi__errpf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_file(&s,f); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} +#endif // !STBI_NO_STDIO + +#endif // !STBI_NO_LINEAR + +// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is +// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always +// reports false! + +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(buffer); + STBI_NOTUSED(len); + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +STBIDEF int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + long pos = ftell(f); + int res; + stbi__context s; + stbi__start_file(&s,f); + res = stbi__hdr_test(&s); + fseek(f, pos, SEEK_SET); + return res; + #else + STBI_NOTUSED(f); + return 0; + #endif +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(clbk); + STBI_NOTUSED(user); + return 0; + #endif +} + +#ifndef STBI_NO_LINEAR +static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f; + +STBIDEF void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; } +STBIDEF void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; } +#endif + +static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f; + +STBIDEF void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; } +STBIDEF void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; } + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +enum +{ + STBI__SCAN_load=0, + STBI__SCAN_type, + STBI__SCAN_header +}; + +static void stbi__refill_buffer(stbi__context *s) +{ + int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen); + s->callback_already_read += (int) (s->img_buffer - s->img_buffer_original); + if (n == 0) { + // at end of file, treat same as if from memory, but need to handle case + // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file + s->read_from_callbacks = 0; + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start+1; + *s->img_buffer = 0; + } else { + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start + n; + } +} + +stbi_inline static stbi_uc stbi__get8(stbi__context *s) +{ + if (s->img_buffer < s->img_buffer_end) + return *s->img_buffer++; + if (s->read_from_callbacks) { + stbi__refill_buffer(s); + return *s->img_buffer++; + } + return 0; +} + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_HDR) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +stbi_inline static int stbi__at_eof(stbi__context *s) +{ + if (s->io.read) { + if (!(s->io.eof)(s->io_user_data)) return 0; + // if feof() is true, check if buffer = end + // special case: we've only got the special 0 character at the end + if (s->read_from_callbacks == 0) return 1; + } + + return s->img_buffer >= s->img_buffer_end; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) +// nothing +#else +static void stbi__skip(stbi__context *s, int n) +{ + if (n == 0) return; // already there! + if (n < 0) { + s->img_buffer = s->img_buffer_end; + return; + } + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + s->img_buffer = s->img_buffer_end; + (s->io.skip)(s->io_user_data, n - blen); + return; + } + } + s->img_buffer += n; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_TGA) && defined(STBI_NO_HDR) && defined(STBI_NO_PNM) +// nothing +#else +static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n) +{ + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + int res, count; + + memcpy(buffer, s->img_buffer, blen); + + count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen); + res = (count == (n-blen)); + s->img_buffer = s->img_buffer_end; + return res; + } + } + + if (s->img_buffer+n <= s->img_buffer_end) { + memcpy(buffer, s->img_buffer, n); + s->img_buffer += n; + return 1; + } else + return 0; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static int stbi__get16be(stbi__context *s) +{ + int z = stbi__get8(s); + return (z << 8) + stbi__get8(s); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static stbi__uint32 stbi__get32be(stbi__context *s) +{ + stbi__uint32 z = stbi__get16be(s); + return (z << 16) + stbi__get16be(s); +} +#endif + +#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) +// nothing +#else +static int stbi__get16le(stbi__context *s) +{ + int z = stbi__get8(s); + return z + (stbi__get8(s) << 8); +} +#endif + +#ifndef STBI_NO_BMP +static stbi__uint32 stbi__get32le(stbi__context *s) +{ + stbi__uint32 z = stbi__get16le(s); + z += (stbi__uint32)stbi__get16le(s) << 16; + return z; +} +#endif + +#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static stbi_uc stbi__compute_y(int r, int g, int b) +{ + return (stbi_uc) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) stbi__malloc_mad3(req_comp, x, y, 0); + if (good == NULL) { + STBI_FREE(data); + return stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + unsigned char *src = data + j * x * img_n ; + unsigned char *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 stbi__compute_y_16(int r, int g, int b) +{ + return (stbi__uint16) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + stbi__uint16 *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2); + if (good == NULL) { + STBI_FREE(data); + return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + stbi__uint16 *src = data + j * x * img_n ; + stbi__uint16 *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return (stbi__uint16*) stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output; + if (!data) return NULL; + output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0); + if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale); + } + } + if (n < comp) { + for (i=0; i < x*y; ++i) { + output[i*comp + n] = data[i*comp + n]/255.0f; + } + } + STBI_FREE(data); + return output; +} +#endif + +#ifndef STBI_NO_HDR +#define stbi__float2int(x) ((int) (x)) +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output; + if (!data) return NULL; + output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0); + if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + } + STBI_FREE(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder +// +// simple implementation +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - some SIMD kernels for common paths on targets with SSE2/NEON +// - uses a lot of intermediate memory, could cache poorly + +#ifndef STBI_NO_JPEG + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + stbi_uc fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + stbi__uint16 code[256]; + stbi_uc values[256]; + stbi_uc size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} stbi__huffman; + +typedef struct +{ + stbi__context *s; + stbi__huffman huff_dc[4]; + stbi__huffman huff_ac[4]; + stbi__uint16 dequant[4][64]; + stbi__int16 fast_ac[4][1 << FAST_BITS]; + +// sizes for components, interleaved MCUs + int img_h_max, img_v_max; + int img_mcu_x, img_mcu_y; + int img_mcu_w, img_mcu_h; + +// definition of jpeg image component + struct + { + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + stbi_uc *data; + void *raw_data, *raw_coeff; + stbi_uc *linebuf; + short *coeff; // progressive only + int coeff_w, coeff_h; // number of 8x8 coefficient blocks + } img_comp[4]; + + stbi__uint32 code_buffer; // jpeg entropy-coded buffer + int code_bits; // number of valid bits + unsigned char marker; // marker seen while filling entropy buffer + int nomore; // flag if we saw a marker so must stop + + int progressive; + int spec_start; + int spec_end; + int succ_high; + int succ_low; + int eob_run; + int jfif; + int app14_color_transform; // Adobe APP14 tag + int rgb; + + int scan_n, order[4]; + int restart_interval, todo; + +// kernels + void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]); + void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step); + stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs); +} stbi__jpeg; + +static int stbi__build_huffman(stbi__huffman *h, int *count) +{ + int i,j,k=0; + unsigned int code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) { + for (j=0; j < count[i]; ++j) { + h->size[k++] = (stbi_uc) (i+1); + if(k >= 257) return stbi__err("bad size list","Corrupt JPEG"); + } + } + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (stbi__uint16) (code++); + if (code-1 >= (1u << j)) return stbi__err("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (stbi_uc) i; + } + } + } + return 1; +} + +// build a table that decodes both magnitude and value of small ACs in +// one go. +static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h) +{ + int i; + for (i=0; i < (1 << FAST_BITS); ++i) { + stbi_uc fast = h->fast[i]; + fast_ac[i] = 0; + if (fast < 255) { + int rs = h->values[fast]; + int run = (rs >> 4) & 15; + int magbits = rs & 15; + int len = h->size[fast]; + + if (magbits && len + magbits <= FAST_BITS) { + // magnitude code followed by receive_extend code + int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits); + int m = 1 << (magbits - 1); + if (k < m) k += (~0U << magbits) + 1; + // if the result is small enough, we can fit it in fast_ac table + if (k >= -128 && k <= 127) + fast_ac[i] = (stbi__int16) ((k * 256) + (run * 16) + (len + magbits)); + } + } + } +} + +static void stbi__grow_buffer_unsafe(stbi__jpeg *j) +{ + do { + unsigned int b = j->nomore ? 0 : stbi__get8(j->s); + if (b == 0xff) { + int c = stbi__get8(j->s); + while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes + if (c != 0) { + j->marker = (unsigned char) c; + j->nomore = 1; + return; + } + } + j->code_buffer |= b << (24 - j->code_bits); + j->code_bits += 8; + } while (j->code_bits <= 24); +} + +// (1 << n) - 1 +static const stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h) +{ + unsigned int temp; + int c,k; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + int s = h->size[k]; + if (s > j->code_bits) + return -1; + j->code_buffer <<= s; + j->code_bits -= s; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + temp = j->code_buffer >> 16; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + j->code_bits -= 16; + return -1; + } + + if (k > j->code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k]; + if(c < 0 || c >= 256) // symbol id out of bounds! + return -1; + STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + j->code_bits -= k; + j->code_buffer <<= k; + return h->values[c]; +} + +// bias[n] = (-1<code_bits < n) stbi__grow_buffer_unsafe(j); + if (j->code_bits < n) return 0; // ran out of bits from stream, return 0s intead of continuing + + sgn = j->code_buffer >> 31; // sign bit always in MSB; 0 if MSB clear (positive), 1 if MSB set (negative) + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k + (stbi__jbias[n] & (sgn - 1)); +} + +// get some unsigned bits +stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n) +{ + unsigned int k; + if (j->code_bits < n) stbi__grow_buffer_unsafe(j); + if (j->code_bits < n) return 0; // ran out of bits from stream, return 0s intead of continuing + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k; +} + +stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j) +{ + unsigned int k; + if (j->code_bits < 1) stbi__grow_buffer_unsafe(j); + if (j->code_bits < 1) return 0; // ran out of bits from stream, return 0s intead of continuing + k = j->code_buffer; + j->code_buffer <<= 1; + --j->code_bits; + return k & 0x80000000; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static const stbi_uc stbi__jpeg_dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi__uint16 *dequant) +{ + int diff,dc,k; + int t; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0 || t > 15) return stbi__err("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? stbi__extend_receive(j, t) : 0; + if (!stbi__addints_valid(j->img_comp[b].dc_pred, diff)) return stbi__err("bad delta","Corrupt JPEG"); + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + if (!stbi__mul2shorts_valid(dc, dequant[0])) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + data[0] = (short) (dc * dequant[0]); + + // decode AC components, see JPEG spec + k = 1; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + if (s > j->code_bits) return stbi__err("bad huffman code", "Combined length longer than code bits available"); + j->code_buffer <<= s; + j->code_bits -= s; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * dequant[zig]); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]); + } + } + } while (k < 64); + return 1; +} + +static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b) +{ + int diff,dc; + int t; + if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + if (j->succ_high == 0) { + // first scan for DC coefficient, must be first + memset(data,0,64*sizeof(data[0])); // 0 all the ac values now + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0 || t > 15) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + diff = t ? stbi__extend_receive(j, t) : 0; + + if (!stbi__addints_valid(j->img_comp[b].dc_pred, diff)) return stbi__err("bad delta", "Corrupt JPEG"); + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + if (!stbi__mul2shorts_valid(dc, 1 << j->succ_low)) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + data[0] = (short) (dc * (1 << j->succ_low)); + } else { + // refinement scan for DC coefficient + if (stbi__jpeg_get_bit(j)) + data[0] += (short) (1 << j->succ_low); + } + return 1; +} + +// @OPTIMIZE: store non-zigzagged during the decode passes, +// and only de-zigzag when dequantizing +static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac) +{ + int k; + if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->succ_high == 0) { + int shift = j->succ_low; + + if (j->eob_run) { + --j->eob_run; + return 1; + } + + k = j->spec_start; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + if (s > j->code_bits) return stbi__err("bad huffman code", "Combined length longer than code bits available"); + j->code_buffer <<= s; + j->code_bits -= s; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * (1 << shift)); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r); + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + --j->eob_run; + break; + } + k += 16; + } else { + k += r; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * (1 << shift)); + } + } + } while (k <= j->spec_end); + } else { + // refinement scan for these AC coefficients + + short bit = (short) (1 << j->succ_low); + + if (j->eob_run) { + --j->eob_run; + for (k = j->spec_start; k <= j->spec_end; ++k) { + short *p = &data[stbi__jpeg_dezigzag[k]]; + if (*p != 0) + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } + } else { + k = j->spec_start; + do { + int r,s; + int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r) - 1; + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + r = 64; // force end of block + } else { + // r=15 s=0 should write 16 0s, so we just do + // a run of 15 0s and then write s (which is 0), + // so we don't have to do anything special here + } + } else { + if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG"); + // sign bit + if (stbi__jpeg_get_bit(j)) + s = bit; + else + s = -bit; + } + + // advance by r + while (k <= j->spec_end) { + short *p = &data[stbi__jpeg_dezigzag[k++]]; + if (*p != 0) { + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } else { + if (r == 0) { + *p = (short) s; + break; + } + --r; + } + } + } while (k <= j->spec_end); + } + } + return 1; +} + +// take a -128..127 value and stbi__clamp it and convert to 0..255 +stbi_inline static stbi_uc stbi__clamp(int x) +{ + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (stbi_uc) x; +} + +#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5))) +#define stbi__fsh(x) ((x) * 4096) + +// derived from jidctint -- DCT_ISLOW +#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * stbi__f2f(0.5411961f); \ + t2 = p1 + p3*stbi__f2f(-1.847759065f); \ + t3 = p1 + p2*stbi__f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = stbi__fsh(p2+p3); \ + t1 = stbi__fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*stbi__f2f( 1.175875602f); \ + t0 = t0*stbi__f2f( 0.298631336f); \ + t1 = t1*stbi__f2f( 2.053119869f); \ + t2 = t2*stbi__f2f( 3.072711026f); \ + t3 = t3*stbi__f2f( 1.501321110f); \ + p1 = p5 + p1*stbi__f2f(-0.899976223f); \ + p2 = p5 + p2*stbi__f2f(-2.562915447f); \ + p3 = p3*stbi__f2f(-1.961570560f); \ + p4 = p4*stbi__f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64]) +{ + int i,val[64],*v=val; + stbi_uc *o; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0]*4; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + // so we want to round that, which means adding 0.5 * 1<<17, + // aka 65536. Also, we'll end up with -128 to 127 that we want + // to encode as 0..255 by adding 128, so we'll add that before the shift + x0 += 65536 + (128<<17); + x1 += 65536 + (128<<17); + x2 += 65536 + (128<<17); + x3 += 65536 + (128<<17); + // tried computing the shifts into temps, or'ing the temps to see + // if any were out of range, but that was slower + o[0] = stbi__clamp((x0+t3) >> 17); + o[7] = stbi__clamp((x0-t3) >> 17); + o[1] = stbi__clamp((x1+t2) >> 17); + o[6] = stbi__clamp((x1-t2) >> 17); + o[2] = stbi__clamp((x2+t1) >> 17); + o[5] = stbi__clamp((x2-t1) >> 17); + o[3] = stbi__clamp((x3+t0) >> 17); + o[4] = stbi__clamp((x3-t0) >> 17); + } +} + +#ifdef STBI_SSE2 +// sse2 integer IDCT. not the fastest possible implementation but it +// produces bit-identical results to the generic C version so it's +// fully "transparent". +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + // This is constructed to match our regular (generic) integer IDCT exactly. + __m128i row0, row1, row2, row3, row4, row5, row6, row7; + __m128i tmp; + + // dot product constant: even elems=x, odd elems=y + #define dct_const(x,y) _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y)) + + // out(0) = c0[even]*x + c0[odd]*y (c0, x, y 16-bit, out 32-bit) + // out(1) = c1[even]*x + c1[odd]*y + #define dct_rot(out0,out1, x,y,c0,c1) \ + __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \ + __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \ + __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \ + __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \ + __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \ + __m128i out1##_h = _mm_madd_epi16(c0##hi, c1) + + // out = in << 12 (in 16-bit, out 32-bit) + #define dct_widen(out, in) \ + __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \ + __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4) + + // wide add + #define dct_wadd(out, a, b) \ + __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_add_epi32(a##_h, b##_h) + + // wide sub + #define dct_wsub(out, a, b) \ + __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_sub_epi32(a##_h, b##_h) + + // butterfly a/b, add bias, then shift by "s" and pack + #define dct_bfly32o(out0, out1, a,b,bias,s) \ + { \ + __m128i abiased_l = _mm_add_epi32(a##_l, bias); \ + __m128i abiased_h = _mm_add_epi32(a##_h, bias); \ + dct_wadd(sum, abiased, b); \ + dct_wsub(dif, abiased, b); \ + out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \ + out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \ + } + + // 8-bit interleave step (for transposes) + #define dct_interleave8(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi8(a, b); \ + b = _mm_unpackhi_epi8(tmp, b) + + // 16-bit interleave step (for transposes) + #define dct_interleave16(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi16(a, b); \ + b = _mm_unpackhi_epi16(tmp, b) + + #define dct_pass(bias,shift) \ + { \ + /* even part */ \ + dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \ + __m128i sum04 = _mm_add_epi16(row0, row4); \ + __m128i dif04 = _mm_sub_epi16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \ + dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \ + __m128i sum17 = _mm_add_epi16(row1, row7); \ + __m128i sum35 = _mm_add_epi16(row3, row5); \ + dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \ + dct_wadd(x4, y0o, y4o); \ + dct_wadd(x5, y1o, y5o); \ + dct_wadd(x6, y2o, y5o); \ + dct_wadd(x7, y3o, y4o); \ + dct_bfly32o(row0,row7, x0,x7,bias,shift); \ + dct_bfly32o(row1,row6, x1,x6,bias,shift); \ + dct_bfly32o(row2,row5, x2,x5,bias,shift); \ + dct_bfly32o(row3,row4, x3,x4,bias,shift); \ + } + + __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f)); + __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f)); + __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f)); + __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f)); + __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f)); + __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f)); + __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f)); + __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f)); + + // rounding biases in column/row passes, see stbi__idct_block for explanation. + __m128i bias_0 = _mm_set1_epi32(512); + __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17)); + + // load + row0 = _mm_load_si128((const __m128i *) (data + 0*8)); + row1 = _mm_load_si128((const __m128i *) (data + 1*8)); + row2 = _mm_load_si128((const __m128i *) (data + 2*8)); + row3 = _mm_load_si128((const __m128i *) (data + 3*8)); + row4 = _mm_load_si128((const __m128i *) (data + 4*8)); + row5 = _mm_load_si128((const __m128i *) (data + 5*8)); + row6 = _mm_load_si128((const __m128i *) (data + 6*8)); + row7 = _mm_load_si128((const __m128i *) (data + 7*8)); + + // column pass + dct_pass(bias_0, 10); + + { + // 16bit 8x8 transpose pass 1 + dct_interleave16(row0, row4); + dct_interleave16(row1, row5); + dct_interleave16(row2, row6); + dct_interleave16(row3, row7); + + // transpose pass 2 + dct_interleave16(row0, row2); + dct_interleave16(row1, row3); + dct_interleave16(row4, row6); + dct_interleave16(row5, row7); + + // transpose pass 3 + dct_interleave16(row0, row1); + dct_interleave16(row2, row3); + dct_interleave16(row4, row5); + dct_interleave16(row6, row7); + } + + // row pass + dct_pass(bias_1, 17); + + { + // pack + __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7 + __m128i p1 = _mm_packus_epi16(row2, row3); + __m128i p2 = _mm_packus_epi16(row4, row5); + __m128i p3 = _mm_packus_epi16(row6, row7); + + // 8bit 8x8 transpose pass 1 + dct_interleave8(p0, p2); // a0e0a1e1... + dct_interleave8(p1, p3); // c0g0c1g1... + + // transpose pass 2 + dct_interleave8(p0, p1); // a0c0e0g0... + dct_interleave8(p2, p3); // b0d0f0h0... + + // transpose pass 3 + dct_interleave8(p0, p2); // a0b0c0d0... + dct_interleave8(p1, p3); // a4b4c4d4... + + // store + _mm_storel_epi64((__m128i *) out, p0); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p2); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p1); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p3); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e)); + } + +#undef dct_const +#undef dct_rot +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_interleave8 +#undef dct_interleave16 +#undef dct_pass +} + +#endif // STBI_SSE2 + +#ifdef STBI_NEON + +// NEON integer IDCT. should produce bit-identical +// results to the generic C version. +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + int16x8_t row0, row1, row2, row3, row4, row5, row6, row7; + + int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f)); + int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f)); + int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f)); + int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f)); + int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f)); + int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f)); + int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f)); + int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f)); + int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f)); + int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f)); + int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f)); + int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f)); + +#define dct_long_mul(out, inq, coeff) \ + int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff) + +#define dct_long_mac(out, acc, inq, coeff) \ + int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff) + +#define dct_widen(out, inq) \ + int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \ + int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12) + +// wide add +#define dct_wadd(out, a, b) \ + int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vaddq_s32(a##_h, b##_h) + +// wide sub +#define dct_wsub(out, a, b) \ + int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vsubq_s32(a##_h, b##_h) + +// butterfly a/b, then shift using "shiftop" by "s" and pack +#define dct_bfly32o(out0,out1, a,b,shiftop,s) \ + { \ + dct_wadd(sum, a, b); \ + dct_wsub(dif, a, b); \ + out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \ + out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \ + } + +#define dct_pass(shiftop, shift) \ + { \ + /* even part */ \ + int16x8_t sum26 = vaddq_s16(row2, row6); \ + dct_long_mul(p1e, sum26, rot0_0); \ + dct_long_mac(t2e, p1e, row6, rot0_1); \ + dct_long_mac(t3e, p1e, row2, rot0_2); \ + int16x8_t sum04 = vaddq_s16(row0, row4); \ + int16x8_t dif04 = vsubq_s16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + int16x8_t sum15 = vaddq_s16(row1, row5); \ + int16x8_t sum17 = vaddq_s16(row1, row7); \ + int16x8_t sum35 = vaddq_s16(row3, row5); \ + int16x8_t sum37 = vaddq_s16(row3, row7); \ + int16x8_t sumodd = vaddq_s16(sum17, sum35); \ + dct_long_mul(p5o, sumodd, rot1_0); \ + dct_long_mac(p1o, p5o, sum17, rot1_1); \ + dct_long_mac(p2o, p5o, sum35, rot1_2); \ + dct_long_mul(p3o, sum37, rot2_0); \ + dct_long_mul(p4o, sum15, rot2_1); \ + dct_wadd(sump13o, p1o, p3o); \ + dct_wadd(sump24o, p2o, p4o); \ + dct_wadd(sump23o, p2o, p3o); \ + dct_wadd(sump14o, p1o, p4o); \ + dct_long_mac(x4, sump13o, row7, rot3_0); \ + dct_long_mac(x5, sump24o, row5, rot3_1); \ + dct_long_mac(x6, sump23o, row3, rot3_2); \ + dct_long_mac(x7, sump14o, row1, rot3_3); \ + dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \ + dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \ + dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \ + dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \ + } + + // load + row0 = vld1q_s16(data + 0*8); + row1 = vld1q_s16(data + 1*8); + row2 = vld1q_s16(data + 2*8); + row3 = vld1q_s16(data + 3*8); + row4 = vld1q_s16(data + 4*8); + row5 = vld1q_s16(data + 5*8); + row6 = vld1q_s16(data + 6*8); + row7 = vld1q_s16(data + 7*8); + + // add DC bias + row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0)); + + // column pass + dct_pass(vrshrn_n_s32, 10); + + // 16bit 8x8 transpose + { +// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively. +// whether compilers actually get this is another story, sadly. +#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); } +#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); } + + // pass 1 + dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6 + dct_trn16(row2, row3); + dct_trn16(row4, row5); + dct_trn16(row6, row7); + + // pass 2 + dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4 + dct_trn32(row1, row3); + dct_trn32(row4, row6); + dct_trn32(row5, row7); + + // pass 3 + dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0 + dct_trn64(row1, row5); + dct_trn64(row2, row6); + dct_trn64(row3, row7); + +#undef dct_trn16 +#undef dct_trn32 +#undef dct_trn64 + } + + // row pass + // vrshrn_n_s32 only supports shifts up to 16, we need + // 17. so do a non-rounding shift of 16 first then follow + // up with a rounding shift by 1. + dct_pass(vshrn_n_s32, 16); + + { + // pack and round + uint8x8_t p0 = vqrshrun_n_s16(row0, 1); + uint8x8_t p1 = vqrshrun_n_s16(row1, 1); + uint8x8_t p2 = vqrshrun_n_s16(row2, 1); + uint8x8_t p3 = vqrshrun_n_s16(row3, 1); + uint8x8_t p4 = vqrshrun_n_s16(row4, 1); + uint8x8_t p5 = vqrshrun_n_s16(row5, 1); + uint8x8_t p6 = vqrshrun_n_s16(row6, 1); + uint8x8_t p7 = vqrshrun_n_s16(row7, 1); + + // again, these can translate into one instruction, but often don't. +#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); } +#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); } + + // sadly can't use interleaved stores here since we only write + // 8 bytes to each scan line! + + // 8x8 8-bit transpose pass 1 + dct_trn8_8(p0, p1); + dct_trn8_8(p2, p3); + dct_trn8_8(p4, p5); + dct_trn8_8(p6, p7); + + // pass 2 + dct_trn8_16(p0, p2); + dct_trn8_16(p1, p3); + dct_trn8_16(p4, p6); + dct_trn8_16(p5, p7); + + // pass 3 + dct_trn8_32(p0, p4); + dct_trn8_32(p1, p5); + dct_trn8_32(p2, p6); + dct_trn8_32(p3, p7); + + // store + vst1_u8(out, p0); out += out_stride; + vst1_u8(out, p1); out += out_stride; + vst1_u8(out, p2); out += out_stride; + vst1_u8(out, p3); out += out_stride; + vst1_u8(out, p4); out += out_stride; + vst1_u8(out, p5); out += out_stride; + vst1_u8(out, p6); out += out_stride; + vst1_u8(out, p7); + +#undef dct_trn8_8 +#undef dct_trn8_16 +#undef dct_trn8_32 + } + +#undef dct_long_mul +#undef dct_long_mac +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_pass +} + +#endif // STBI_NEON + +#define STBI__MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static stbi_uc stbi__get_marker(stbi__jpeg *j) +{ + stbi_uc x; + if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; } + x = stbi__get8(j->s); + if (x != 0xff) return STBI__MARKER_none; + while (x == 0xff) + x = stbi__get8(j->s); // consume repeated 0xff fill bytes + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +#define STBI__RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, stbi__jpeg_reset the entropy decoder and +// the dc prediction +static void stbi__jpeg_reset(stbi__jpeg *j) +{ + j->code_bits = 0; + j->code_buffer = 0; + j->nomore = 0; + j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = j->img_comp[3].dc_pred = 0; + j->marker = STBI__MARKER_none; + j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff; + j->eob_run = 0; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int stbi__parse_entropy_coded_data(stbi__jpeg *z) +{ + stbi__jpeg_reset(z); + if (!z->progressive) { + if (z->scan_n == 1) { + int i,j; + STBI_SIMD_ALIGN(short, data[64]); + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + STBI_SIMD_ALIGN(short, data[64]); + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x)*8; + int y2 = (j*z->img_comp[n].v + y)*8; + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data); + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } else { + if (z->scan_n == 1) { + int i,j; + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + if (z->spec_start == 0) { + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } else { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha])) + return 0; + } + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x); + int y2 = (j*z->img_comp[n].v + y); + short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w); + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } +} + +static void stbi__jpeg_dequantize(short *data, stbi__uint16 *dequant) +{ + int i; + for (i=0; i < 64; ++i) + data[i] *= dequant[i]; +} + +static void stbi__jpeg_finish(stbi__jpeg *z) +{ + if (z->progressive) { + // dequantize and idct the data + int i,j,n; + for (n=0; n < z->s->img_n; ++n) { + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]); + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + } + } + } + } +} + +static int stbi__process_marker(stbi__jpeg *z, int m) +{ + int L; + switch (m) { + case STBI__MARKER_none: // no marker found + return stbi__err("expected marker","Corrupt JPEG"); + + case 0xDD: // DRI - specify restart interval + if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG"); + z->restart_interval = stbi__get16be(z->s); + return 1; + + case 0xDB: // DQT - define quantization table + L = stbi__get16be(z->s)-2; + while (L > 0) { + int q = stbi__get8(z->s); + int p = q >> 4, sixteen = (p != 0); + int t = q & 15,i; + if (p != 0 && p != 1) return stbi__err("bad DQT type","Corrupt JPEG"); + if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG"); + + for (i=0; i < 64; ++i) + z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s)); + L -= (sixteen ? 129 : 65); + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = stbi__get16be(z->s)-2; + while (L > 0) { + stbi_uc *v; + int sizes[16],i,n=0; + int q = stbi__get8(z->s); + int tc = q >> 4; + int th = q & 15; + if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = stbi__get8(z->s); + n += sizes[i]; + } + if(n > 256) return stbi__err("bad DHT header","Corrupt JPEG"); // Loop over i < n would write past end of values! + L -= 17; + if (tc == 0) { + if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0; + v = z->huff_dc[th].values; + } else { + if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0; + v = z->huff_ac[th].values; + } + for (i=0; i < n; ++i) + v[i] = stbi__get8(z->s); + if (tc != 0) + stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th); + L -= n; + } + return L==0; + } + + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + L = stbi__get16be(z->s); + if (L < 2) { + if (m == 0xFE) + return stbi__err("bad COM len","Corrupt JPEG"); + else + return stbi__err("bad APP len","Corrupt JPEG"); + } + L -= 2; + + if (m == 0xE0 && L >= 5) { // JFIF APP0 segment + static const unsigned char tag[5] = {'J','F','I','F','\0'}; + int ok = 1; + int i; + for (i=0; i < 5; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 5; + if (ok) + z->jfif = 1; + } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment + static const unsigned char tag[6] = {'A','d','o','b','e','\0'}; + int ok = 1; + int i; + for (i=0; i < 6; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 6; + if (ok) { + stbi__get8(z->s); // version + stbi__get16be(z->s); // flags0 + stbi__get16be(z->s); // flags1 + z->app14_color_transform = stbi__get8(z->s); // color transform + L -= 6; + } + } + + stbi__skip(z->s, L); + return 1; + } + + return stbi__err("unknown marker","Corrupt JPEG"); +} + +// after we see SOS +static int stbi__process_scan_header(stbi__jpeg *z) +{ + int i; + int Ls = stbi__get16be(z->s); + z->scan_n = stbi__get8(z->s); + if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG"); + for (i=0; i < z->scan_n; ++i) { + int id = stbi__get8(z->s), which; + int q = stbi__get8(z->s); + for (which = 0; which < z->s->img_n; ++which) + if (z->img_comp[which].id == id) + break; + if (which == z->s->img_n) return 0; // no match + z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG"); + z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG"); + z->order[i] = which; + } + + { + int aa; + z->spec_start = stbi__get8(z->s); + z->spec_end = stbi__get8(z->s); // should be 63, but might be 0 + aa = stbi__get8(z->s); + z->succ_high = (aa >> 4); + z->succ_low = (aa & 15); + if (z->progressive) { + if (z->spec_start > 63 || z->spec_end > 63 || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13) + return stbi__err("bad SOS", "Corrupt JPEG"); + } else { + if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG"); + if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG"); + z->spec_end = 63; + } + } + + return 1; +} + +static int stbi__free_jpeg_components(stbi__jpeg *z, int ncomp, int why) +{ + int i; + for (i=0; i < ncomp; ++i) { + if (z->img_comp[i].raw_data) { + STBI_FREE(z->img_comp[i].raw_data); + z->img_comp[i].raw_data = NULL; + z->img_comp[i].data = NULL; + } + if (z->img_comp[i].raw_coeff) { + STBI_FREE(z->img_comp[i].raw_coeff); + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].coeff = 0; + } + if (z->img_comp[i].linebuf) { + STBI_FREE(z->img_comp[i].linebuf); + z->img_comp[i].linebuf = NULL; + } + } + return why; +} + +static int stbi__process_frame_header(stbi__jpeg *z, int scan) +{ + stbi__context *s = z->s; + int Lf,p,i,q, h_max=1,v_max=1,c; + Lf = stbi__get16be(s); if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG + p = stbi__get8(s); if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + s->img_y = stbi__get16be(s); if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + s->img_x = stbi__get16be(s); if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + c = stbi__get8(s); + if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG"); + s->img_n = c; + for (i=0; i < c; ++i) { + z->img_comp[i].data = NULL; + z->img_comp[i].linebuf = NULL; + } + + if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG"); + + z->rgb = 0; + for (i=0; i < s->img_n; ++i) { + static const unsigned char rgb[3] = { 'R', 'G', 'B' }; + z->img_comp[i].id = stbi__get8(s); + if (s->img_n == 3 && z->img_comp[i].id == rgb[i]) + ++z->rgb; + q = stbi__get8(s); + z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG"); + z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG"); + z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG"); + } + + if (scan != STBI__SCAN_load) return 1; + + if (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) return stbi__err("too large", "Image too large to decode"); + + for (i=0; i < s->img_n; ++i) { + if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h; + if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v; + } + + // check that plane subsampling factors are integer ratios; our resamplers can't deal with fractional ratios + // and I've never seen a non-corrupted JPEG file actually use them + for (i=0; i < s->img_n; ++i) { + if (h_max % z->img_comp[i].h != 0) return stbi__err("bad H","Corrupt JPEG"); + if (v_max % z->img_comp[i].v != 0) return stbi__err("bad V","Corrupt JPEG"); + } + + // compute interleaved mcu info + z->img_h_max = h_max; + z->img_v_max = v_max; + z->img_mcu_w = h_max * 8; + z->img_mcu_h = v_max * 8; + // these sizes can't be more than 17 bits + z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w; + z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h; + + for (i=0; i < s->img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max; + z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + // + // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier) + // so these muls can't overflow with 32-bit ints (which we require) + z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8; + z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8; + z->img_comp[i].coeff = 0; + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].linebuf = NULL; + z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15); + if (z->img_comp[i].raw_data == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + // align blocks for idct using mmx/sse + z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); + if (z->progressive) { + // w2, h2 are multiples of 8 (see above) + z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8; + z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8; + z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15); + if (z->img_comp[i].raw_coeff == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15); + } + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define stbi__DNL(x) ((x) == 0xdc) +#define stbi__SOI(x) ((x) == 0xd8) +#define stbi__EOI(x) ((x) == 0xd9) +#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2) +#define stbi__SOS(x) ((x) == 0xda) + +#define stbi__SOF_progressive(x) ((x) == 0xc2) + +static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan) +{ + int m; + z->jfif = 0; + z->app14_color_transform = -1; // valid values are 0,1,2 + z->marker = STBI__MARKER_none; // initialize cached marker to empty + m = stbi__get_marker(z); + if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG"); + if (scan == STBI__SCAN_type) return 1; + m = stbi__get_marker(z); + while (!stbi__SOF(m)) { + if (!stbi__process_marker(z,m)) return 0; + m = stbi__get_marker(z); + while (m == STBI__MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG"); + m = stbi__get_marker(z); + } + } + z->progressive = stbi__SOF_progressive(m); + if (!stbi__process_frame_header(z, scan)) return 0; + return 1; +} + +static stbi_uc stbi__skip_jpeg_junk_at_end(stbi__jpeg *j) +{ + // some JPEGs have junk at end, skip over it but if we find what looks + // like a valid marker, resume there + while (!stbi__at_eof(j->s)) { + stbi_uc x = stbi__get8(j->s); + while (x == 0xff) { // might be a marker + if (stbi__at_eof(j->s)) return STBI__MARKER_none; + x = stbi__get8(j->s); + if (x != 0x00 && x != 0xff) { + // not a stuffed zero or lead-in to another marker, looks + // like an actual marker, return it + return x; + } + // stuffed zero has x=0 now which ends the loop, meaning we go + // back to regular scan loop. + // repeated 0xff keeps trying to read the next byte of the marker. + } + } + return STBI__MARKER_none; +} + +// decode image to YCbCr format +static int stbi__decode_jpeg_image(stbi__jpeg *j) +{ + int m; + for (m = 0; m < 4; m++) { + j->img_comp[m].raw_data = NULL; + j->img_comp[m].raw_coeff = NULL; + } + j->restart_interval = 0; + if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0; + m = stbi__get_marker(j); + while (!stbi__EOI(m)) { + if (stbi__SOS(m)) { + if (!stbi__process_scan_header(j)) return 0; + if (!stbi__parse_entropy_coded_data(j)) return 0; + if (j->marker == STBI__MARKER_none ) { + j->marker = stbi__skip_jpeg_junk_at_end(j); + // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0 + } + m = stbi__get_marker(j); + if (STBI__RESTART(m)) + m = stbi__get_marker(j); + } else if (stbi__DNL(m)) { + int Ld = stbi__get16be(j->s); + stbi__uint32 NL = stbi__get16be(j->s); + if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG"); + if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG"); + m = stbi__get_marker(j); + } else { + if (!stbi__process_marker(j, m)) return 1; + m = stbi__get_marker(j); + } + } + if (j->progressive) + stbi__jpeg_finish(j); + return 1; +} + +// static jfif-centered resampling (across block boundaries) + +typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1, + int w, int hs); + +#define stbi__div4(x) ((stbi_uc) ((x) >> 2)) + +static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + STBI_NOTUSED(out); + STBI_NOTUSED(in_far); + STBI_NOTUSED(w); + STBI_NOTUSED(hs); + return in_near; +} + +static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples vertically for every one in input + int i; + STBI_NOTUSED(hs); + for (i=0; i < w; ++i) + out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2); + return out; +} + +static stbi_uc* stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples horizontally for every one in input + int i; + stbi_uc *input = in_near; + + if (w == 1) { + // if only one sample, can't do any interpolation + out[0] = out[1] = input[0]; + return out; + } + + out[0] = input[0]; + out[1] = stbi__div4(input[0]*3 + input[1] + 2); + for (i=1; i < w-1; ++i) { + int n = 3*input[i]+2; + out[i*2+0] = stbi__div4(n+input[i-1]); + out[i*2+1] = stbi__div4(n+input[i+1]); + } + out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2); + out[i*2+1] = input[w-1]; + + STBI_NOTUSED(in_far); + STBI_NOTUSED(hs); + + return out; +} + +#define stbi__div16(x) ((stbi_uc) ((x) >> 4)) + +static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i,t0,t1; + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + out[0] = stbi__div4(t1+2); + for (i=1; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i=0,t0,t1; + + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + // process groups of 8 pixels for as long as we can. + // note we can't handle the last pixel in a row in this loop + // because we need to handle the filter boundary conditions. + for (; i < ((w-1) & ~7); i += 8) { +#if defined(STBI_SSE2) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + __m128i zero = _mm_setzero_si128(); + __m128i farb = _mm_loadl_epi64((__m128i *) (in_far + i)); + __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i)); + __m128i farw = _mm_unpacklo_epi8(farb, zero); + __m128i nearw = _mm_unpacklo_epi8(nearb, zero); + __m128i diff = _mm_sub_epi16(farw, nearw); + __m128i nears = _mm_slli_epi16(nearw, 2); + __m128i curr = _mm_add_epi16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + __m128i prv0 = _mm_slli_si128(curr, 2); + __m128i nxt0 = _mm_srli_si128(curr, 2); + __m128i prev = _mm_insert_epi16(prv0, t1, 0); + __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + __m128i bias = _mm_set1_epi16(8); + __m128i curs = _mm_slli_epi16(curr, 2); + __m128i prvd = _mm_sub_epi16(prev, curr); + __m128i nxtd = _mm_sub_epi16(next, curr); + __m128i curb = _mm_add_epi16(curs, bias); + __m128i even = _mm_add_epi16(prvd, curb); + __m128i odd = _mm_add_epi16(nxtd, curb); + + // interleave even and odd pixels, then undo scaling. + __m128i int0 = _mm_unpacklo_epi16(even, odd); + __m128i int1 = _mm_unpackhi_epi16(even, odd); + __m128i de0 = _mm_srli_epi16(int0, 4); + __m128i de1 = _mm_srli_epi16(int1, 4); + + // pack and write output + __m128i outv = _mm_packus_epi16(de0, de1); + _mm_storeu_si128((__m128i *) (out + i*2), outv); +#elif defined(STBI_NEON) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + uint8x8_t farb = vld1_u8(in_far + i); + uint8x8_t nearb = vld1_u8(in_near + i); + int16x8_t diff = vreinterpretq_s16_u16(vsubl_u8(farb, nearb)); + int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2)); + int16x8_t curr = vaddq_s16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + int16x8_t prv0 = vextq_s16(curr, curr, 7); + int16x8_t nxt0 = vextq_s16(curr, curr, 1); + int16x8_t prev = vsetq_lane_s16(t1, prv0, 0); + int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + int16x8_t curs = vshlq_n_s16(curr, 2); + int16x8_t prvd = vsubq_s16(prev, curr); + int16x8_t nxtd = vsubq_s16(next, curr); + int16x8_t even = vaddq_s16(curs, prvd); + int16x8_t odd = vaddq_s16(curs, nxtd); + + // undo scaling and round, then store with even/odd phases interleaved + uint8x8x2_t o; + o.val[0] = vqrshrun_n_s16(even, 4); + o.val[1] = vqrshrun_n_s16(odd, 4); + vst2_u8(out + i*2, o); +#endif + + // "previous" value for next iter + t1 = 3*in_near[i+7] + in_far[i+7]; + } + + t0 = t1; + t1 = 3*in_near[i] + in_far[i]; + out[i*2] = stbi__div16(3*t1 + t0 + 8); + + for (++i; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} +#endif + +static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // resample with nearest-neighbor + int i,j; + STBI_NOTUSED(in_far); + for (i=0; i < w; ++i) + for (j=0; j < hs; ++j) + out[i*hs+j] = in_near[i]; + return out; +} + +// this is a reduced-precision calculation of YCbCr-to-RGB introduced +// to make sure the code produces the same results in both SIMD and scalar +#define stbi__float2fixed(x) (((int) ((x) * 4096.0f + 0.5f)) << 8) +static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* stbi__float2fixed(1.40200f); + g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step) +{ + int i = 0; + +#ifdef STBI_SSE2 + // step == 3 is pretty ugly on the final interleave, and i'm not convinced + // it's useful in practice (you wouldn't use it for textures, for example). + // so just accelerate step == 4 case. + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + __m128i signflip = _mm_set1_epi8(-0x80); + __m128i cr_const0 = _mm_set1_epi16( (short) ( 1.40200f*4096.0f+0.5f)); + __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f)); + __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f)); + __m128i cb_const1 = _mm_set1_epi16( (short) ( 1.77200f*4096.0f+0.5f)); + __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128); + __m128i xw = _mm_set1_epi16(255); // alpha channel + + for (; i+7 < count; i += 8) { + // load + __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i)); + __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i)); + __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i)); + __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128 + __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128 + + // unpack to short (and left-shift cr, cb by 8) + __m128i yw = _mm_unpacklo_epi8(y_bias, y_bytes); + __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased); + __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased); + + // color transform + __m128i yws = _mm_srli_epi16(yw, 4); + __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw); + __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw); + __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1); + __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1); + __m128i rws = _mm_add_epi16(cr0, yws); + __m128i gwt = _mm_add_epi16(cb0, yws); + __m128i bws = _mm_add_epi16(yws, cb1); + __m128i gws = _mm_add_epi16(gwt, cr1); + + // descale + __m128i rw = _mm_srai_epi16(rws, 4); + __m128i bw = _mm_srai_epi16(bws, 4); + __m128i gw = _mm_srai_epi16(gws, 4); + + // back to byte, set up for transpose + __m128i brb = _mm_packus_epi16(rw, bw); + __m128i gxb = _mm_packus_epi16(gw, xw); + + // transpose to interleave channels + __m128i t0 = _mm_unpacklo_epi8(brb, gxb); + __m128i t1 = _mm_unpackhi_epi8(brb, gxb); + __m128i o0 = _mm_unpacklo_epi16(t0, t1); + __m128i o1 = _mm_unpackhi_epi16(t0, t1); + + // store + _mm_storeu_si128((__m128i *) (out + 0), o0); + _mm_storeu_si128((__m128i *) (out + 16), o1); + out += 32; + } + } +#endif + +#ifdef STBI_NEON + // in this version, step=3 support would be easy to add. but is there demand? + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + uint8x8_t signflip = vdup_n_u8(0x80); + int16x8_t cr_const0 = vdupq_n_s16( (short) ( 1.40200f*4096.0f+0.5f)); + int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f)); + int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f)); + int16x8_t cb_const1 = vdupq_n_s16( (short) ( 1.77200f*4096.0f+0.5f)); + + for (; i+7 < count; i += 8) { + // load + uint8x8_t y_bytes = vld1_u8(y + i); + uint8x8_t cr_bytes = vld1_u8(pcr + i); + uint8x8_t cb_bytes = vld1_u8(pcb + i); + int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip)); + int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip)); + + // expand to s16 + int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4)); + int16x8_t crw = vshll_n_s8(cr_biased, 7); + int16x8_t cbw = vshll_n_s8(cb_biased, 7); + + // color transform + int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0); + int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0); + int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1); + int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1); + int16x8_t rws = vaddq_s16(yws, cr0); + int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1); + int16x8_t bws = vaddq_s16(yws, cb1); + + // undo scaling, round, convert to byte + uint8x8x4_t o; + o.val[0] = vqrshrun_n_s16(rws, 4); + o.val[1] = vqrshrun_n_s16(gws, 4); + o.val[2] = vqrshrun_n_s16(bws, 4); + o.val[3] = vdup_n_u8(255); + + // store, interleaving r/g/b/a + vst4_u8(out, o); + out += 8*4; + } + } +#endif + + for (; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* stbi__float2fixed(1.40200f); + g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} +#endif + +// set up the kernels +static void stbi__setup_jpeg(stbi__jpeg *j) +{ + j->idct_block_kernel = stbi__idct_block; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2; + +#ifdef STBI_SSE2 + if (stbi__sse2_available()) { + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; + } +#endif + +#ifdef STBI_NEON + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; +#endif +} + +// clean up the temporary component buffers +static void stbi__cleanup_jpeg(stbi__jpeg *j) +{ + stbi__free_jpeg_components(j, j->s->img_n, 0); +} + +typedef struct +{ + resample_row_func resample; + stbi_uc *line0,*line1; + int hs,vs; // expansion factor in each axis + int w_lores; // horizontal pixels pre-expansion + int ystep; // how far through vertical expansion we are + int ypos; // which pre-expansion row we're on +} stbi__resample; + +// fast 0..255 * 0..255 => 0..255 rounded multiplication +static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y) +{ + unsigned int t = x*y + 128; + return (stbi_uc) ((t + (t >>8)) >> 8); +} + +static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n, is_rgb; + z->s->img_n = 0; // make stbi__cleanup_jpeg safe + + // validate req_comp + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + + // load a jpeg image from whichever source, but leave in YCbCr format + if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : z->s->img_n >= 3 ? 3 : 1; + + is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif)); + + if (z->s->img_n == 3 && n < 3 && !is_rgb) + decode_n = 1; + else + decode_n = z->s->img_n; + + // nothing to do if no components requested; check this now to avoid + // accessing uninitialized coutput[0] later + if (decode_n <= 0) { stbi__cleanup_jpeg(z); return NULL; } + + // resample and color-convert + { + int k; + unsigned int i,j; + stbi_uc *output; + stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL }; + + stbi__resample res_comp[4]; + + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + + // allocate line buffer big enough for upsampling off the edges + // with upsample factor of 4 + z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3); + if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + r->hs = z->img_h_max / z->img_comp[k].h; + r->vs = z->img_v_max / z->img_comp[k].v; + r->ystep = r->vs >> 1; + r->w_lores = (z->s->img_x + r->hs-1) / r->hs; + r->ypos = 0; + r->line0 = r->line1 = z->img_comp[k].data; + + if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1; + else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2; + else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2; + else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel; + else r->resample = stbi__resample_row_generic; + } + + // can't error after this so, this is safe + output = (stbi_uc *) stbi__malloc_mad3(n, z->s->img_x, z->s->img_y, 1); + if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + // now go ahead and resample + for (j=0; j < z->s->img_y; ++j) { + stbi_uc *out = output + n * z->s->img_x * j; + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + int y_bot = r->ystep >= (r->vs >> 1); + coutput[k] = r->resample(z->img_comp[k].linebuf, + y_bot ? r->line1 : r->line0, + y_bot ? r->line0 : r->line1, + r->w_lores, r->hs); + if (++r->ystep >= r->vs) { + r->ystep = 0; + r->line0 = r->line1; + if (++r->ypos < z->img_comp[k].y) + r->line1 += z->img_comp[k].w2; + } + } + if (n >= 3) { + stbi_uc *y = coutput[0]; + if (z->s->img_n == 3) { + if (is_rgb) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = y[i]; + out[1] = coutput[1][i]; + out[2] = coutput[2][i]; + out[3] = 255; + out += n; + } + } else { + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else if (z->s->img_n == 4) { + if (z->app14_color_transform == 0) { // CMYK + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(coutput[0][i], m); + out[1] = stbi__blinn_8x8(coutput[1][i], m); + out[2] = stbi__blinn_8x8(coutput[2][i], m); + out[3] = 255; + out += n; + } + } else if (z->app14_color_transform == 2) { // YCCK + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(255 - out[0], m); + out[1] = stbi__blinn_8x8(255 - out[1], m); + out[2] = stbi__blinn_8x8(255 - out[2], m); + out += n; + } + } else { // YCbCr + alpha? Ignore the fourth channel for now + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else + for (i=0; i < z->s->img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n==3 + out += n; + } + } else { + if (is_rgb) { + if (n == 1) + for (i=0; i < z->s->img_x; ++i) + *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + else { + for (i=0; i < z->s->img_x; ++i, out += 2) { + out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + out[1] = 255; + } + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 0) { + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + stbi_uc r = stbi__blinn_8x8(coutput[0][i], m); + stbi_uc g = stbi__blinn_8x8(coutput[1][i], m); + stbi_uc b = stbi__blinn_8x8(coutput[2][i], m); + out[0] = stbi__compute_y(r, g, b); + out[1] = 255; + out += n; + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 2) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]); + out[1] = 255; + out += n; + } + } else { + stbi_uc *y = coutput[0]; + if (n == 1) + for (i=0; i < z->s->img_x; ++i) out[i] = y[i]; + else + for (i=0; i < z->s->img_x; ++i) { *out++ = y[i]; *out++ = 255; } + } + } + } + stbi__cleanup_jpeg(z); + *out_x = z->s->img_x; + *out_y = z->s->img_y; + if (comp) *comp = z->s->img_n >= 3 ? 3 : 1; // report original components, not output + return output; + } +} + +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + unsigned char* result; + stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__errpuc("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + STBI_NOTUSED(ri); + j->s = s; + stbi__setup_jpeg(j); + result = load_jpeg_image(j, x,y,comp,req_comp); + STBI_FREE(j); + return result; +} + +static int stbi__jpeg_test(stbi__context *s) +{ + int r; + stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__err("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + j->s = s; + stbi__setup_jpeg(j); + r = stbi__decode_jpeg_header(j, STBI__SCAN_type); + stbi__rewind(s); + STBI_FREE(j); + return r; +} + +static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp) +{ + if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) { + stbi__rewind( j->s ); + return 0; + } + if (x) *x = j->s->img_x; + if (y) *y = j->s->img_y; + if (comp) *comp = j->s->img_n >= 3 ? 3 : 1; + return 1; +} + +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp) +{ + int result; + stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg))); + if (!j) return stbi__err("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + j->s = s; + result = stbi__jpeg_info_raw(j, x, y, comp); + STBI_FREE(j); + return result; +} +#endif + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +#ifndef STBI_NO_ZLIB + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables +#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1) +#define STBI__ZNSYMS 288 // number of symbols in literal/length alphabet + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + stbi__uint16 fast[1 << STBI__ZFAST_BITS]; + stbi__uint16 firstcode[16]; + int maxcode[17]; + stbi__uint16 firstsymbol[16]; + stbi_uc size[STBI__ZNSYMS]; + stbi__uint16 value[STBI__ZNSYMS]; +} stbi__zhuffman; + +stbi_inline static int stbi__bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +stbi_inline static int stbi__bit_reverse(int v, int bits) +{ + STBI_ASSERT(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return stbi__bitreverse16(v) >> (16-bits); +} + +static int stbi__zbuild_huffman(stbi__zhuffman *z, const stbi_uc *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 0, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + if (sizes[i] > (1 << i)) + return stbi__err("bad sizes", "Corrupt PNG"); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (stbi__uint16) code; + z->firstsymbol[i] = (stbi__uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i); + z->size [c] = (stbi_uc ) s; + z->value[c] = (stbi__uint16) i; + if (s <= STBI__ZFAST_BITS) { + int j = stbi__bit_reverse(next_code[s],s); + while (j < (1 << STBI__ZFAST_BITS)) { + z->fast[j] = fastv; + j += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +typedef struct +{ + stbi_uc *zbuffer, *zbuffer_end; + int num_bits; + int hit_zeof_once; + stbi__uint32 code_buffer; + + char *zout; + char *zout_start; + char *zout_end; + int z_expandable; + + stbi__zhuffman z_length, z_distance; +} stbi__zbuf; + +stbi_inline static int stbi__zeof(stbi__zbuf *z) +{ + return (z->zbuffer >= z->zbuffer_end); +} + +stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z) +{ + return stbi__zeof(z) ? 0 : *z->zbuffer++; +} + +static void stbi__fill_bits(stbi__zbuf *z) +{ + do { + if (z->code_buffer >= (1U << z->num_bits)) { + z->zbuffer = z->zbuffer_end; /* treat this as EOF so we fail. */ + return; + } + z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits; + z->num_bits += 8; + } while (z->num_bits <= 24); +} + +stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n) +{ + unsigned int k; + if (z->num_bits < n) stbi__fill_bits(z); + k = z->code_buffer & ((1 << n) - 1); + z->code_buffer >>= n; + z->num_bits -= n; + return k; +} + +static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s,k; + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = stbi__bit_reverse(a->code_buffer, 16); + for (s=STBI__ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s >= 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + if (b >= STBI__ZNSYMS) return -1; // some data was corrupt somewhere! + if (z->size[b] != s) return -1; // was originally an assert, but report failure instead. + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; +} + +stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s; + if (a->num_bits < 16) { + if (stbi__zeof(a)) { + if (!a->hit_zeof_once) { + // This is the first time we hit eof, insert 16 extra padding btis + // to allow us to keep going; if we actually consume any of them + // though, that is invalid data. This is caught later. + a->hit_zeof_once = 1; + a->num_bits += 16; // add 16 implicit zero bits + } else { + // We already inserted our extra 16 padding bits and are again + // out, this stream is actually prematurely terminated. + return -1; + } + } else { + stbi__fill_bits(a); + } + } + b = z->fast[a->code_buffer & STBI__ZFAST_MASK]; + if (b) { + s = b >> 9; + a->code_buffer >>= s; + a->num_bits -= s; + return b & 511; + } + return stbi__zhuffman_decode_slowpath(a, z); +} + +static int stbi__zexpand(stbi__zbuf *z, char *zout, int n) // need to make room for n bytes +{ + char *q; + unsigned int cur, limit, old_limit; + z->zout = zout; + if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG"); + cur = (unsigned int) (z->zout - z->zout_start); + limit = old_limit = (unsigned) (z->zout_end - z->zout_start); + if (UINT_MAX - cur < (unsigned) n) return stbi__err("outofmem", "Out of memory"); + while (cur + n > limit) { + if(limit > UINT_MAX / 2) return stbi__err("outofmem", "Out of memory"); + limit *= 2; + } + q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit); + STBI_NOTUSED(old_limit); + if (q == NULL) return stbi__err("outofmem", "Out of memory"); + z->zout_start = q; + z->zout = q + cur; + z->zout_end = q + limit; + return 1; +} + +static const int stbi__zlength_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static const int stbi__zlength_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static const int stbi__zdist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int stbi__parse_huffman_block(stbi__zbuf *a) +{ + char *zout = a->zout; + for(;;) { + int z = stbi__zhuffman_decode(a, &a->z_length); + if (z < 256) { + if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes + if (zout >= a->zout_end) { + if (!stbi__zexpand(a, zout, 1)) return 0; + zout = a->zout; + } + *zout++ = (char) z; + } else { + stbi_uc *p; + int len,dist; + if (z == 256) { + a->zout = zout; + if (a->hit_zeof_once && a->num_bits < 16) { + // The first time we hit zeof, we inserted 16 extra zero bits into our bit + // buffer so the decoder can just do its speculative decoding. But if we + // actually consumed any of those bits (which is the case when num_bits < 16), + // the stream actually read past the end so it is malformed. + return stbi__err("unexpected end","Corrupt PNG"); + } + return 1; + } + if (z >= 286) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, length codes 286 and 287 must not appear in compressed data + z -= 257; + len = stbi__zlength_base[z]; + if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]); + z = stbi__zhuffman_decode(a, &a->z_distance); + if (z < 0 || z >= 30) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, distance codes 30 and 31 must not appear in compressed data + dist = stbi__zdist_base[z]; + if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]); + if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG"); + if (len > a->zout_end - zout) { + if (!stbi__zexpand(a, zout, len)) return 0; + zout = a->zout; + } + p = (stbi_uc *) (zout - dist); + if (dist == 1) { // run of one byte; common in images. + stbi_uc v = *p; + if (len) { do *zout++ = v; while (--len); } + } else { + if (len) { do *zout++ = *p++; while (--len); } + } + } + } +} + +static int stbi__compute_huffman_codes(stbi__zbuf *a) +{ + static const stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + stbi__zhuffman z_codelength; + stbi_uc lencodes[286+32+137];//padding for maximum single op + stbi_uc codelength_sizes[19]; + int i,n; + + int hlit = stbi__zreceive(a,5) + 257; + int hdist = stbi__zreceive(a,5) + 1; + int hclen = stbi__zreceive(a,4) + 4; + int ntot = hlit + hdist; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = stbi__zreceive(a,3); + codelength_sizes[length_dezigzag[i]] = (stbi_uc) s; + } + if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < ntot) { + int c = stbi__zhuffman_decode(a, &z_codelength); + if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG"); + if (c < 16) + lencodes[n++] = (stbi_uc) c; + else { + stbi_uc fill = 0; + if (c == 16) { + c = stbi__zreceive(a,2)+3; + if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG"); + fill = lencodes[n-1]; + } else if (c == 17) { + c = stbi__zreceive(a,3)+3; + } else if (c == 18) { + c = stbi__zreceive(a,7)+11; + } else { + return stbi__err("bad codelengths", "Corrupt PNG"); + } + if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG"); + memset(lencodes+n, fill, c); + n += c; + } + } + if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG"); + if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int stbi__parse_uncompressed_block(stbi__zbuf *a) +{ + stbi_uc header[4]; + int len,nlen,k; + if (a->num_bits & 7) + stbi__zreceive(a, a->num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (a->num_bits > 0) { + header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check + a->code_buffer >>= 8; + a->num_bits -= 8; + } + if (a->num_bits < 0) return stbi__err("zlib corrupt","Corrupt PNG"); + // now fill header the normal way + while (k < 4) + header[k++] = stbi__zget8(a); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG"); + if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG"); + if (a->zout + len > a->zout_end) + if (!stbi__zexpand(a, a->zout, len)) return 0; + memcpy(a->zout, a->zbuffer, len); + a->zbuffer += len; + a->zout += len; + return 1; +} + +static int stbi__parse_zlib_header(stbi__zbuf *a) +{ + int cmf = stbi__zget8(a); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = stbi__zget8(a); + if (stbi__zeof(a)) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +static const stbi_uc stbi__zdefault_length[STBI__ZNSYMS] = +{ + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8 +}; +static const stbi_uc stbi__zdefault_distance[32] = +{ + 5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5 +}; +/* +Init algorithm: +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8; + for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9; + for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7; + for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8; + + for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5; +} +*/ + +static int stbi__parse_zlib(stbi__zbuf *a, int parse_header) +{ + int final, type; + if (parse_header) + if (!stbi__parse_zlib_header(a)) return 0; + a->num_bits = 0; + a->code_buffer = 0; + a->hit_zeof_once = 0; + do { + final = stbi__zreceive(a,1); + type = stbi__zreceive(a,2); + if (type == 0) { + if (!stbi__parse_uncompressed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , STBI__ZNSYMS)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0; + } else { + if (!stbi__compute_huffman_codes(a)) return 0; + } + if (!stbi__parse_huffman_block(a)) return 0; + } + } while (!final); + return 1; +} + +static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return stbi__parse_zlib(a, parse_header); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 1)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer+len; + if (stbi__do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 0)) + return (int) (a.zout - a.zout_start); + else + return -1; +} +#endif + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + +#ifndef STBI_NO_PNG +typedef struct +{ + stbi__uint32 length; + stbi__uint32 type; +} stbi__pngchunk; + +static stbi__pngchunk stbi__get_chunk_header(stbi__context *s) +{ + stbi__pngchunk c; + c.length = stbi__get32be(s); + c.type = stbi__get32be(s); + return c; +} + +static int stbi__check_png_header(stbi__context *s) +{ + static const stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG"); + return 1; +} + +typedef struct +{ + stbi__context *s; + stbi_uc *idata, *expanded, *out; + int depth; +} stbi__png; + + +enum { + STBI__F_none=0, + STBI__F_sub=1, + STBI__F_up=2, + STBI__F_avg=3, + STBI__F_paeth=4, + // synthetic filter used for first scanline to avoid needing a dummy row of 0s + STBI__F_avg_first +}; + +static stbi_uc first_row_filter[5] = +{ + STBI__F_none, + STBI__F_sub, + STBI__F_none, + STBI__F_avg_first, + STBI__F_sub // Paeth with b=c=0 turns out to be equivalent to sub +}; + +static int stbi__paeth(int a, int b, int c) +{ + // This formulation looks very different from the reference in the PNG spec, but is + // actually equivalent and has favorable data dependencies and admits straightforward + // generation of branch-free code, which helps performance significantly. + int thresh = c*3 - (a + b); + int lo = a < b ? a : b; + int hi = a < b ? b : a; + int t0 = (hi <= thresh) ? lo : c; + int t1 = (thresh <= lo) ? hi : t0; + return t1; +} + +static const stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 }; + +// adds an extra all-255 alpha channel +// dest == src is legal +// img_n must be 1 or 3 +static void stbi__create_png_alpha_expand8(stbi_uc *dest, stbi_uc *src, stbi__uint32 x, int img_n) +{ + int i; + // must process data backwards since we allow dest==src + if (img_n == 1) { + for (i=x-1; i >= 0; --i) { + dest[i*2+1] = 255; + dest[i*2+0] = src[i]; + } + } else { + STBI_ASSERT(img_n == 3); + for (i=x-1; i >= 0; --i) { + dest[i*4+3] = 255; + dest[i*4+2] = src[i*3+2]; + dest[i*4+1] = src[i*3+1]; + dest[i*4+0] = src[i*3+0]; + } + } +} + +// create the png data from post-deflated data +static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color) +{ + int bytes = (depth == 16 ? 2 : 1); + stbi__context *s = a->s; + stbi__uint32 i,j,stride = x*out_n*bytes; + stbi__uint32 img_len, img_width_bytes; + stbi_uc *filter_buf; + int all_ok = 1; + int k; + int img_n = s->img_n; // copy it into a local for later + + int output_bytes = out_n*bytes; + int filter_bytes = img_n*bytes; + int width = x; + + STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1); + a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into + if (!a->out) return stbi__err("outofmem", "Out of memory"); + + // note: error exits here don't need to clean up a->out individually, + // stbi__do_png always does on error. + if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG"); + img_width_bytes = (((img_n * x * depth) + 7) >> 3); + if (!stbi__mad2sizes_valid(img_width_bytes, y, img_width_bytes)) return stbi__err("too large", "Corrupt PNG"); + img_len = (img_width_bytes + 1) * y; + + // we used to check for exact match between raw_len and img_len on non-interlaced PNGs, + // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros), + // so just check for raw_len < img_len always. + if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG"); + + // Allocate two scan lines worth of filter workspace buffer. + filter_buf = (stbi_uc *) stbi__malloc_mad2(img_width_bytes, 2, 0); + if (!filter_buf) return stbi__err("outofmem", "Out of memory"); + + // Filtering for low-bit-depth images + if (depth < 8) { + filter_bytes = 1; + width = img_width_bytes; + } + + for (j=0; j < y; ++j) { + // cur/prior filter buffers alternate + stbi_uc *cur = filter_buf + (j & 1)*img_width_bytes; + stbi_uc *prior = filter_buf + (~j & 1)*img_width_bytes; + stbi_uc *dest = a->out + stride*j; + int nk = width * filter_bytes; + int filter = *raw++; + + // check filter type + if (filter > 4) { + all_ok = stbi__err("invalid filter","Corrupt PNG"); + break; + } + + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + + // perform actual filtering + switch (filter) { + case STBI__F_none: + memcpy(cur, raw, nk); + break; + case STBI__F_sub: + memcpy(cur, raw, filter_bytes); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); + break; + case STBI__F_up: + for (k = 0; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + prior[k]); + break; + case STBI__F_avg: + for (k = 0; k < filter_bytes; ++k) + cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); + break; + case STBI__F_paeth: + for (k = 0; k < filter_bytes; ++k) + cur[k] = STBI__BYTECAST(raw[k] + prior[k]); // prior[k] == stbi__paeth(0,prior[k],0) + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes], prior[k], prior[k-filter_bytes])); + break; + case STBI__F_avg_first: + memcpy(cur, raw, filter_bytes); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); + break; + } + + raw += nk; + + // expand decoded bits in cur to dest, also adding an extra alpha channel if desired + if (depth < 8) { + stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range + stbi_uc *in = cur; + stbi_uc *out = dest; + stbi_uc inb = 0; + stbi__uint32 nsmp = x*img_n; + + // expand bits to bytes first + if (depth == 4) { + for (i=0; i < nsmp; ++i) { + if ((i & 1) == 0) inb = *in++; + *out++ = scale * (inb >> 4); + inb <<= 4; + } + } else if (depth == 2) { + for (i=0; i < nsmp; ++i) { + if ((i & 3) == 0) inb = *in++; + *out++ = scale * (inb >> 6); + inb <<= 2; + } + } else { + STBI_ASSERT(depth == 1); + for (i=0; i < nsmp; ++i) { + if ((i & 7) == 0) inb = *in++; + *out++ = scale * (inb >> 7); + inb <<= 1; + } + } + + // insert alpha=255 values if desired + if (img_n != out_n) + stbi__create_png_alpha_expand8(dest, dest, x, img_n); + } else if (depth == 8) { + if (img_n == out_n) + memcpy(dest, cur, x*img_n); + else + stbi__create_png_alpha_expand8(dest, cur, x, img_n); + } else if (depth == 16) { + // convert the image data from big-endian to platform-native + stbi__uint16 *dest16 = (stbi__uint16*)dest; + stbi__uint32 nsmp = x*img_n; + + if (img_n == out_n) { + for (i = 0; i < nsmp; ++i, ++dest16, cur += 2) + *dest16 = (cur[0] << 8) | cur[1]; + } else { + STBI_ASSERT(img_n+1 == out_n); + if (img_n == 1) { + for (i = 0; i < x; ++i, dest16 += 2, cur += 2) { + dest16[0] = (cur[0] << 8) | cur[1]; + dest16[1] = 0xffff; + } + } else { + STBI_ASSERT(img_n == 3); + for (i = 0; i < x; ++i, dest16 += 4, cur += 6) { + dest16[0] = (cur[0] << 8) | cur[1]; + dest16[1] = (cur[2] << 8) | cur[3]; + dest16[2] = (cur[4] << 8) | cur[5]; + dest16[3] = 0xffff; + } + } + } + } + } + + STBI_FREE(filter_buf); + if (!all_ok) return 0; + + return 1; +} + +static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced) +{ + int bytes = (depth == 16 ? 2 : 1); + int out_bytes = out_n * bytes; + stbi_uc *final; + int p; + if (!interlaced) + return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color); + + // de-interlacing + final = (stbi_uc *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0); + if (!final) return stbi__err("outofmem", "Out of memory"); + for (p=0; p < 7; ++p) { + int xorig[] = { 0,4,0,2,0,1,0 }; + int yorig[] = { 0,0,4,0,2,0,1 }; + int xspc[] = { 8,8,4,4,2,2,1 }; + int yspc[] = { 8,8,8,4,4,2,2 }; + int i,j,x,y; + // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1 + x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p]; + y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p]; + if (x && y) { + stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y; + if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) { + STBI_FREE(final); + return 0; + } + for (j=0; j < y; ++j) { + for (i=0; i < x; ++i) { + int out_y = j*yspc[p]+yorig[p]; + int out_x = i*xspc[p]+xorig[p]; + memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes, + a->out + (j*x+i)*out_bytes, out_bytes); + } + } + STBI_FREE(a->out); + image_data += img_len; + image_data_len -= img_len; + } + } + a->out = final; + + return 1; +} + +static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi__uint16 *p = (stbi__uint16*) z->out; + + // compute color-based transparency, assuming we've + // already got 65535 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i = 0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 65535); + p += 2; + } + } else { + for (i = 0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n) +{ + stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y; + stbi_uc *p, *temp_out, *orig = a->out; + + p = (stbi_uc *) stbi__malloc_mad2(pixel_count, pal_img_n, 0); + if (p == NULL) return stbi__err("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + STBI_FREE(a->out); + a->out = temp_out; + + STBI_NOTUSED(len); + + return 1; +} + +static int stbi__unpremultiply_on_load_global = 0; +static int stbi__de_iphone_flag_global = 0; + +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_global = flag_true_if_should_unpremultiply; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_global = flag_true_if_should_convert; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__unpremultiply_on_load stbi__unpremultiply_on_load_global +#define stbi__de_iphone_flag stbi__de_iphone_flag_global +#else +static STBI_THREAD_LOCAL int stbi__unpremultiply_on_load_local, stbi__unpremultiply_on_load_set; +static STBI_THREAD_LOCAL int stbi__de_iphone_flag_local, stbi__de_iphone_flag_set; + +STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_local = flag_true_if_should_unpremultiply; + stbi__unpremultiply_on_load_set = 1; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_local = flag_true_if_should_convert; + stbi__de_iphone_flag_set = 1; +} + +#define stbi__unpremultiply_on_load (stbi__unpremultiply_on_load_set \ + ? stbi__unpremultiply_on_load_local \ + : stbi__unpremultiply_on_load_global) +#define stbi__de_iphone_flag (stbi__de_iphone_flag_set \ + ? stbi__de_iphone_flag_local \ + : stbi__de_iphone_flag_global) +#endif // STBI_THREAD_LOCAL + +static void stbi__de_iphone(stbi__png *z) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + if (s->img_out_n == 3) { // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 3; + } + } else { + STBI_ASSERT(s->img_out_n == 4); + if (stbi__unpremultiply_on_load) { + // convert bgr to rgb and unpremultiply + for (i=0; i < pixel_count; ++i) { + stbi_uc a = p[3]; + stbi_uc t = p[0]; + if (a) { + stbi_uc half = a / 2; + p[0] = (p[2] * 255 + half) / a; + p[1] = (p[1] * 255 + half) / a; + p[2] = ( t * 255 + half) / a; + } else { + p[0] = p[2]; + p[2] = t; + } + p += 4; + } + } else { + // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 4; + } + } + } +} + +#define STBI__PNG_TYPE(a,b,c,d) (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d)) + +static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) +{ + stbi_uc palette[1024], pal_img_n=0; + stbi_uc has_trans=0, tc[3]={0}; + stbi__uint16 tc16[3]; + stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k,interlace=0, color=0, is_iphone=0; + stbi__context *s = z->s; + + z->expanded = NULL; + z->idata = NULL; + z->out = NULL; + + if (!stbi__check_png_header(s)) return 0; + + if (scan == STBI__SCAN_type) return 1; + + for (;;) { + stbi__pngchunk c = stbi__get_chunk_header(s); + switch (c.type) { + case STBI__PNG_TYPE('C','g','B','I'): + is_iphone = 1; + stbi__skip(s, c.length); + break; + case STBI__PNG_TYPE('I','H','D','R'): { + int comp,filter; + if (!first) return stbi__err("multiple IHDR","Corrupt PNG"); + first = 0; + if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG"); + s->img_x = stbi__get32be(s); + s->img_y = stbi__get32be(s); + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + z->depth = stbi__get8(s); if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16) return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only"); + color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3 && z->depth == 16) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG"); + comp = stbi__get8(s); if (comp) return stbi__err("bad comp method","Corrupt PNG"); + filter= stbi__get8(s); if (filter) return stbi__err("bad filter method","Corrupt PNG"); + interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG"); + if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode"); + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + s->img_n = 1; + if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG"); + } + // even with SCAN_header, have to scan to see if we have a tRNS + break; + } + + case STBI__PNG_TYPE('P','L','T','E'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = stbi__get8(s); + palette[i*4+1] = stbi__get8(s); + palette[i*4+2] = stbi__get8(s); + palette[i*4+3] = 255; + } + break; + } + + case STBI__PNG_TYPE('t','R','N','S'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; } + if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = stbi__get8(s); + } else { + if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG"); + if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG"); + has_trans = 1; + // non-paletted with tRNS = constant alpha. if header-scanning, we can stop now. + if (scan == STBI__SCAN_header) { ++s->img_n; return 1; } + if (z->depth == 16) { + for (k = 0; k < s->img_n && k < 3; ++k) // extra loop test to suppress false GCC warning + tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is + } else { + for (k = 0; k < s->img_n && k < 3; ++k) + tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger + } + } + break; + } + + case STBI__PNG_TYPE('I','D','A','T'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG"); + if (scan == STBI__SCAN_header) { + // header scan definitely stops at first IDAT + if (pal_img_n) + s->img_n = pal_img_n; + return 1; + } + if (c.length > (1u << 30)) return stbi__err("IDAT size limit", "IDAT section larger than 2^30 bytes"); + if ((int)(ioff + c.length) < (int)ioff) return 0; + if (ioff + c.length > idata_limit) { + stbi__uint32 idata_limit_old = idata_limit; + stbi_uc *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + STBI_NOTUSED(idata_limit_old); + p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory"); + z->idata = p; + } + if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG"); + ioff += c.length; + break; + } + + case STBI__PNG_TYPE('I','E','N','D'): { + stbi__uint32 raw_len, bpl; + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (scan != STBI__SCAN_load) return 1; + if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG"); + // initial guess for decoded data size to avoid unnecessary reallocs + bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component + raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */; + z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone); + if (z->expanded == NULL) return 0; // zlib should set error + STBI_FREE(z->idata); z->idata = NULL; + if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + s->img_out_n = s->img_n+1; + else + s->img_out_n = s->img_n; + if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0; + if (has_trans) { + if (z->depth == 16) { + if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0; + } else { + if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0; + } + } + if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2) + stbi__de_iphone(z); + if (pal_img_n) { + // pal_img_n == 3 or 4 + s->img_n = pal_img_n; // record the actual colors we had + s->img_out_n = pal_img_n; + if (req_comp >= 3) s->img_out_n = req_comp; + if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n)) + return 0; + } else if (has_trans) { + // non-paletted image with tRNS -> source image has (constant) alpha + ++s->img_n; + } + STBI_FREE(z->expanded); z->expanded = NULL; + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + return 1; + } + + default: + // if critical, fail + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + // not threadsafe + static char invalid_chunk[] = "XXXX PNG chunk not known"; + invalid_chunk[0] = STBI__BYTECAST(c.type >> 24); + invalid_chunk[1] = STBI__BYTECAST(c.type >> 16); + invalid_chunk[2] = STBI__BYTECAST(c.type >> 8); + invalid_chunk[3] = STBI__BYTECAST(c.type >> 0); + #endif + return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type"); + } + stbi__skip(s, c.length); + break; + } + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + } +} + +static void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri) +{ + void *result=NULL; + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) { + if (p->depth <= 8) + ri->bits_per_channel = 8; + else if (p->depth == 16) + ri->bits_per_channel = 16; + else + return stbi__errpuc("bad bits_per_channel", "PNG not supported: unsupported color depth"); + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s->img_out_n) { + if (ri->bits_per_channel == 8) + result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + else + result = stbi__convert_format16((stbi__uint16 *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + p->s->img_out_n = req_comp; + if (result == NULL) return result; + } + *x = p->s->img_x; + *y = p->s->img_y; + if (n) *n = p->s->img_n; + } + STBI_FREE(p->out); p->out = NULL; + STBI_FREE(p->expanded); p->expanded = NULL; + STBI_FREE(p->idata); p->idata = NULL; + + return result; +} + +static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi__png p; + p.s = s; + return stbi__do_png(&p, x,y,comp,req_comp, ri); +} + +static int stbi__png_test(stbi__context *s) +{ + int r; + r = stbi__check_png_header(s); + stbi__rewind(s); + return r; +} + +static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp) +{ + if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) { + stbi__rewind( p->s ); + return 0; + } + if (x) *x = p->s->img_x; + if (y) *y = p->s->img_y; + if (comp) *comp = p->s->img_n; + return 1; +} + +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__png p; + p.s = s; + return stbi__png_info_raw(&p, x, y, comp); +} + +static int stbi__png_is16(stbi__context *s) +{ + stbi__png p; + p.s = s; + if (!stbi__png_info_raw(&p, NULL, NULL, NULL)) + return 0; + if (p.depth != 16) { + stbi__rewind(p.s); + return 0; + } + return 1; +} +#endif + +// Microsoft/Windows BMP image + +#ifndef STBI_NO_BMP +static int stbi__bmp_test_raw(stbi__context *s) +{ + int r; + int sz; + if (stbi__get8(s) != 'B') return 0; + if (stbi__get8(s) != 'M') return 0; + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + stbi__get32le(s); // discard data offset + sz = stbi__get32le(s); + r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124); + return r; +} + +static int stbi__bmp_test(stbi__context *s) +{ + int r = stbi__bmp_test_raw(s); + stbi__rewind(s); + return r; +} + + +// returns 0..31 for the highest set bit +static int stbi__high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) { n += 16; z >>= 16; } + if (z >= 0x00100) { n += 8; z >>= 8; } + if (z >= 0x00010) { n += 4; z >>= 4; } + if (z >= 0x00004) { n += 2; z >>= 2; } + if (z >= 0x00002) { n += 1;/* >>= 1;*/ } + return n; +} + +static int stbi__bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +// extract an arbitrarily-aligned N-bit value (N=bits) +// from v, and then make it 8-bits long and fractionally +// extend it to full full range. +static int stbi__shiftsigned(unsigned int v, int shift, int bits) +{ + static unsigned int mul_table[9] = { + 0, + 0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/, + 0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/, + }; + static unsigned int shift_table[9] = { + 0, 0,0,1,0,2,4,6,0, + }; + if (shift < 0) + v <<= -shift; + else + v >>= shift; + STBI_ASSERT(v < 256); + v >>= (8-bits); + STBI_ASSERT(bits >= 0 && bits <= 8); + return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits]; +} + +typedef struct +{ + int bpp, offset, hsz; + unsigned int mr,mg,mb,ma, all_a; + int extra_read; +} stbi__bmp_data; + +static int stbi__bmp_set_mask_defaults(stbi__bmp_data *info, int compress) +{ + // BI_BITFIELDS specifies masks explicitly, don't override + if (compress == 3) + return 1; + + if (compress == 0) { + if (info->bpp == 16) { + info->mr = 31u << 10; + info->mg = 31u << 5; + info->mb = 31u << 0; + } else if (info->bpp == 32) { + info->mr = 0xffu << 16; + info->mg = 0xffu << 8; + info->mb = 0xffu << 0; + info->ma = 0xffu << 24; + info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0 + } else { + // otherwise, use defaults, which is all-0 + info->mr = info->mg = info->mb = info->ma = 0; + } + return 1; + } + return 0; // error +} + +static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info) +{ + int hsz; + if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP"); + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + info->offset = stbi__get32le(s); + info->hsz = hsz = stbi__get32le(s); + info->mr = info->mg = info->mb = info->ma = 0; + info->extra_read = 14; + + if (info->offset < 0) return stbi__errpuc("bad BMP", "bad BMP"); + + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown"); + if (hsz == 12) { + s->img_x = stbi__get16le(s); + s->img_y = stbi__get16le(s); + } else { + s->img_x = stbi__get32le(s); + s->img_y = stbi__get32le(s); + } + if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP"); + info->bpp = stbi__get16le(s); + if (hsz != 12) { + int compress = stbi__get32le(s); + if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE"); + if (compress >= 4) return stbi__errpuc("BMP JPEG/PNG", "BMP type not supported: unsupported compression"); // this includes PNG/JPEG modes + if (compress == 3 && info->bpp != 16 && info->bpp != 32) return stbi__errpuc("bad BMP", "bad BMP"); // bitfields requires 16 or 32 bits/pixel + stbi__get32le(s); // discard sizeof + stbi__get32le(s); // discard hres + stbi__get32le(s); // discard vres + stbi__get32le(s); // discard colorsused + stbi__get32le(s); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + } + if (info->bpp == 16 || info->bpp == 32) { + if (compress == 0) { + stbi__bmp_set_mask_defaults(info, compress); + } else if (compress == 3) { + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->extra_read += 12; + // not documented, but generated by photoshop and handled by mspaint + if (info->mr == info->mg && info->mg == info->mb) { + // ?!?!? + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else { + // V4/V5 header + int i; + if (hsz != 108 && hsz != 124) + return stbi__errpuc("bad BMP", "bad BMP"); + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->ma = stbi__get32le(s); + if (compress != 3) // override mr/mg/mb unless in BI_BITFIELDS mode, as per docs + stbi__bmp_set_mask_defaults(info, compress); + stbi__get32le(s); // discard color space + for (i=0; i < 12; ++i) + stbi__get32le(s); // discard color space parameters + if (hsz == 124) { + stbi__get32le(s); // discard rendering intent + stbi__get32le(s); // discard offset of profile data + stbi__get32le(s); // discard size of profile data + stbi__get32le(s); // discard reserved + } + } + } + return (void *) 1; +} + + +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + unsigned int mr=0,mg=0,mb=0,ma=0, all_a; + stbi_uc pal[256][4]; + int psize=0,i,j,width; + int flip_vertically, pad, target; + stbi__bmp_data info; + STBI_NOTUSED(ri); + + info.all_a = 255; + if (stbi__bmp_parse_header(s, &info) == NULL) + return NULL; // error code already set + + flip_vertically = ((int) s->img_y) > 0; + s->img_y = abs((int) s->img_y); + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + mr = info.mr; + mg = info.mg; + mb = info.mb; + ma = info.ma; + all_a = info.all_a; + + if (info.hsz == 12) { + if (info.bpp < 24) + psize = (info.offset - info.extra_read - 24) / 3; + } else { + if (info.bpp < 16) + psize = (info.offset - info.extra_read - info.hsz) >> 2; + } + if (psize == 0) { + // accept some number of extra bytes after the header, but if the offset points either to before + // the header ends or implies a large amount of extra data, reject the file as malformed + int bytes_read_so_far = s->callback_already_read + (int)(s->img_buffer - s->img_buffer_original); + int header_limit = 1024; // max we actually read is below 256 bytes currently. + int extra_data_limit = 256*4; // what ordinarily goes here is a palette; 256 entries*4 bytes is its max size. + if (bytes_read_so_far <= 0 || bytes_read_so_far > header_limit) { + return stbi__errpuc("bad header", "Corrupt BMP"); + } + // we established that bytes_read_so_far is positive and sensible. + // the first half of this test rejects offsets that are either too small positives, or + // negative, and guarantees that info.offset >= bytes_read_so_far > 0. this in turn + // ensures the number computed in the second half of the test can't overflow. + if (info.offset < bytes_read_so_far || info.offset - bytes_read_so_far > extra_data_limit) { + return stbi__errpuc("bad offset", "Corrupt BMP"); + } else { + stbi__skip(s, info.offset - bytes_read_so_far); + } + } + + if (info.bpp == 24 && ma == 0xff000000) + s->img_n = 3; + else + s->img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = s->img_n; // if they want monochrome, we'll post-convert + + // sanity-check size + if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0)) + return stbi__errpuc("too large", "Corrupt BMP"); + + out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + if (info.bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); } + for (i=0; i < psize; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + if (info.hsz != 12) stbi__get8(s); + pal[i][3] = 255; + } + stbi__skip(s, info.offset - info.extra_read - info.hsz - psize * (info.hsz == 12 ? 3 : 4)); + if (info.bpp == 1) width = (s->img_x + 7) >> 3; + else if (info.bpp == 4) width = (s->img_x + 1) >> 1; + else if (info.bpp == 8) width = s->img_x; + else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); } + pad = (-width)&3; + if (info.bpp == 1) { + for (j=0; j < (int) s->img_y; ++j) { + int bit_offset = 7, v = stbi__get8(s); + for (i=0; i < (int) s->img_x; ++i) { + int color = (v>>bit_offset)&0x1; + out[z++] = pal[color][0]; + out[z++] = pal[color][1]; + out[z++] = pal[color][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + if((--bit_offset) < 0) { + bit_offset = 7; + v = stbi__get8(s); + } + } + stbi__skip(s, pad); + } + } else { + for (j=0; j < (int) s->img_y; ++j) { + for (i=0; i < (int) s->img_x; i += 2) { + int v=stbi__get8(s),v2=0; + if (info.bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + v = (info.bpp == 8) ? stbi__get8(s) : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + stbi__skip(s, pad); + } + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + stbi__skip(s, info.offset - info.extra_read - info.hsz); + if (info.bpp == 24) width = 3 * s->img_x; + else if (info.bpp == 16) width = 2*s->img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (info.bpp == 24) { + easy = 1; + } else if (info.bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + // right shift amt to put high bit in position #7 + rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr); + gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg); + bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb); + ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma); + if (rcount > 8 || gcount > 8 || bcount > 8 || acount > 8) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + } + for (j=0; j < (int) s->img_y; ++j) { + if (easy) { + for (i=0; i < (int) s->img_x; ++i) { + unsigned char a; + out[z+2] = stbi__get8(s); + out[z+1] = stbi__get8(s); + out[z+0] = stbi__get8(s); + z += 3; + a = (easy == 2 ? stbi__get8(s) : 255); + all_a |= a; + if (target == 4) out[z++] = a; + } + } else { + int bpp = info.bpp; + for (i=0; i < (int) s->img_x; ++i) { + stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s)); + unsigned int a; + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount)); + a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255); + all_a |= a; + if (target == 4) out[z++] = STBI__BYTECAST(a); + } + } + stbi__skip(s, pad); + } + } + + // if alpha channel is all 0s, replace with all 255s + if (target == 4 && all_a == 0) + for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4) + out[i] = 255; + + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) s->img_y>>1; ++j) { + stbi_uc *p1 = out + j *s->img_x*target; + stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target; + for (i=0; i < (int) s->img_x*target; ++i) { + t = p1[i]; p1[i] = p2[i]; p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + return out; +} +#endif + +// Targa Truevision - TGA +// by Jonathan Dummer +#ifndef STBI_NO_TGA +// returns STBI_rgb or whatever, 0 on error +static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16) +{ + // only RGB or RGBA (incl. 16bit) or grey allowed + if (is_rgb16) *is_rgb16 = 0; + switch(bits_per_pixel) { + case 8: return STBI_grey; + case 16: if(is_grey) return STBI_grey_alpha; + // fallthrough + case 15: if(is_rgb16) *is_rgb16 = 1; + return STBI_rgb; + case 24: // fallthrough + case 32: return bits_per_pixel/8; + default: return 0; + } +} + +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp) +{ + int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp; + int sz, tga_colormap_type; + stbi__get8(s); // discard Offset + tga_colormap_type = stbi__get8(s); // colormap type + if( tga_colormap_type > 1 ) { + stbi__rewind(s); + return 0; // only RGB or indexed allowed + } + tga_image_type = stbi__get8(s); // image type + if ( tga_colormap_type == 1 ) { // colormapped (paletted) image + if (tga_image_type != 1 && tga_image_type != 9) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip image x and y origin + tga_colormap_bpp = sz; + } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE + if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) { + stbi__rewind(s); + return 0; // only RGB or grey allowed, +/- RLE + } + stbi__skip(s,9); // skip colormap specification and image x/y origin + tga_colormap_bpp = 0; + } + tga_w = stbi__get16le(s); + if( tga_w < 1 ) { + stbi__rewind(s); + return 0; // test width + } + tga_h = stbi__get16le(s); + if( tga_h < 1 ) { + stbi__rewind(s); + return 0; // test height + } + tga_bits_per_pixel = stbi__get8(s); // bits per pixel + stbi__get8(s); // ignore alpha bits + if (tga_colormap_bpp != 0) { + if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) { + // when using a colormap, tga_bits_per_pixel is the size of the indexes + // I don't think anything but 8 or 16bit indexes makes sense + stbi__rewind(s); + return 0; + } + tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL); + } else { + tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL); + } + if(!tga_comp) { + stbi__rewind(s); + return 0; + } + if (x) *x = tga_w; + if (y) *y = tga_h; + if (comp) *comp = tga_comp; + return 1; // seems to have passed everything +} + +static int stbi__tga_test(stbi__context *s) +{ + int res = 0; + int sz, tga_color_type; + stbi__get8(s); // discard Offset + tga_color_type = stbi__get8(s); // color type + if ( tga_color_type > 1 ) goto errorEnd; // only RGB or indexed allowed + sz = stbi__get8(s); // image type + if ( tga_color_type == 1 ) { // colormapped (paletted) image + if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9 + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + stbi__skip(s,4); // skip image x and y origin + } else { // "normal" image w/o colormap + if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE + stbi__skip(s,9); // skip colormap specification and image x/y origin + } + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test width + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test height + sz = stbi__get8(s); // bits per pixel + if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + + res = 1; // if we got this far, everything's good and we can return 1 instead of 0 + +errorEnd: + stbi__rewind(s); + return res; +} + +// read 16bit value and convert to 24bit RGB +static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out) +{ + stbi__uint16 px = (stbi__uint16)stbi__get16le(s); + stbi__uint16 fiveBitMask = 31; + // we have 3 channels with 5bits each + int r = (px >> 10) & fiveBitMask; + int g = (px >> 5) & fiveBitMask; + int b = px & fiveBitMask; + // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later + out[0] = (stbi_uc)((r * 255)/31); + out[1] = (stbi_uc)((g * 255)/31); + out[2] = (stbi_uc)((b * 255)/31); + + // some people claim that the most significant bit might be used for alpha + // (possibly if an alpha-bit is set in the "image descriptor byte") + // but that only made 16bit test images completely translucent.. + // so let's treat all 15 and 16bit TGAs as RGB with no alpha. +} + +static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + // read in the TGA header stuff + int tga_offset = stbi__get8(s); + int tga_indexed = stbi__get8(s); + int tga_image_type = stbi__get8(s); + int tga_is_RLE = 0; + int tga_palette_start = stbi__get16le(s); + int tga_palette_len = stbi__get16le(s); + int tga_palette_bits = stbi__get8(s); + int tga_x_origin = stbi__get16le(s); + int tga_y_origin = stbi__get16le(s); + int tga_width = stbi__get16le(s); + int tga_height = stbi__get16le(s); + int tga_bits_per_pixel = stbi__get8(s); + int tga_comp, tga_rgb16=0; + int tga_inverted = stbi__get8(s); + // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?) + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4] = {0}; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + STBI_NOTUSED(ri); + STBI_NOTUSED(tga_x_origin); // @TODO + STBI_NOTUSED(tga_y_origin); // @TODO + + if (tga_height > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (tga_width > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // do a tiny bit of precessing + if ( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // If I'm paletted, then I'll use the number of bits from the palette + if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16); + else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16); + + if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency + return stbi__errpuc("bad format", "Can't find out TGA pixelformat"); + + // tga info + *x = tga_width; + *y = tga_height; + if (comp) *comp = tga_comp; + + if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0)) + return stbi__errpuc("too large", "Corrupt TGA"); + + tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0); + if (!tga_data) return stbi__errpuc("outofmem", "Out of memory"); + + // skip to the data's starting position (offset usually = 0) + stbi__skip(s, tga_offset ); + + if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) { + for (i=0; i < tga_height; ++i) { + int row = tga_inverted ? tga_height -i - 1 : i; + stbi_uc *tga_row = tga_data + row*tga_width*tga_comp; + stbi__getn(s, tga_row, tga_width * tga_comp); + } + } else { + // do I need to load a palette? + if ( tga_indexed) + { + if (tga_palette_len == 0) { /* you have to have at least one entry! */ + STBI_FREE(tga_data); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + + // any data to skip? (offset usually = 0) + stbi__skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0); + if (!tga_palette) { + STBI_FREE(tga_data); + return stbi__errpuc("outofmem", "Out of memory"); + } + if (tga_rgb16) { + stbi_uc *pal_entry = tga_palette; + STBI_ASSERT(tga_comp == STBI_rgb); + for (i=0; i < tga_palette_len; ++i) { + stbi__tga_read_rgb16(s, pal_entry); + pal_entry += tga_comp; + } + } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) { + STBI_FREE(tga_data); + STBI_FREE(tga_palette); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + } + // load the data + for (i=0; i < tga_width * tga_height; ++i) + { + // if I'm in RLE mode, do I need to get a RLE stbi__pngchunk? + if ( tga_is_RLE ) + { + if ( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = stbi__get8(s); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if ( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if ( read_next_pixel ) + { + // load however much data we did have + if ( tga_indexed ) + { + // read in index, then perform the lookup + int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s); + if ( pal_idx >= tga_palette_len ) { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_comp; + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else if(tga_rgb16) { + STBI_ASSERT(tga_comp == STBI_rgb); + stbi__tga_read_rgb16(s, raw_data); + } else { + // read in the data raw + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = stbi__get8(s); + } + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + + // copy data + for (j = 0; j < tga_comp; ++j) + tga_data[i*tga_comp+j] = raw_data[j]; + + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if ( tga_inverted ) + { + for (j = 0; j*2 < tga_height; ++j) + { + int index1 = j * tga_width * tga_comp; + int index2 = (tga_height - 1 - j) * tga_width * tga_comp; + for (i = tga_width * tga_comp; i > 0; --i) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if ( tga_palette != NULL ) + { + STBI_FREE( tga_palette ); + } + } + + // swap RGB - if the source data was RGB16, it already is in the right order + if (tga_comp >= 3 && !tga_rgb16) + { + unsigned char* tga_pixel = tga_data; + for (i=0; i < tga_width * tga_height; ++i) + { + unsigned char temp = tga_pixel[0]; + tga_pixel[0] = tga_pixel[2]; + tga_pixel[2] = temp; + tga_pixel += tga_comp; + } + } + + // convert to target component count + if (req_comp && req_comp != tga_comp) + tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height); + + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + STBI_NOTUSED(tga_palette_start); + // OK, done + return tga_data; +} +#endif + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s) +{ + int r = (stbi__get32be(s) == 0x38425053); + stbi__rewind(s); + return r; +} + +static int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount) +{ + int count, nleft, len; + + count = 0; + while ((nleft = pixelCount - count) > 0) { + len = stbi__get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + if (len > nleft) return 0; // corrupt data + count += len; + while (len) { + *p = stbi__get8(s); + p += 4; + len--; + } + } else if (len > 128) { + stbi_uc val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len = 257 - len; + if (len > nleft) return 0; // corrupt data + val = stbi__get8(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + + return 1; +} + +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + int pixelCount; + int channelCount, compression; + int channel, i; + int bitdepth; + int w,h; + stbi_uc *out; + STBI_NOTUSED(ri); + + // Check identifier + if (stbi__get32be(s) != 0x38425053) // "8BPS" + return stbi__errpuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (stbi__get16be(s) != 1) + return stbi__errpuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + stbi__skip(s, 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) + return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = stbi__get32be(s); + w = stbi__get32be(s); + + if (h > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (w > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // Make sure the depth is 8 bits. + bitdepth = stbi__get16be(s); + if (bitdepth != 8 && bitdepth != 16) + return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (stbi__get16be(s) != 3) + return stbi__errpuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + stbi__skip(s,stbi__get32be(s) ); + + // Skip the image resources. (resolution, pen tool paths, etc) + stbi__skip(s, stbi__get32be(s) ); + + // Skip the reserved data. + stbi__skip(s, stbi__get32be(s) ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = stbi__get16be(s); + if (compression > 1) + return stbi__errpuc("bad compression", "PSD has an unknown compression format"); + + // Check size + if (!stbi__mad3sizes_valid(4, w, h, 0)) + return stbi__errpuc("too large", "Corrupt PSD"); + + // Create the destination image. + + if (!compression && bitdepth == 16 && bpc == 16) { + out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0); + ri->bits_per_channel = 16; + } else + out = (stbi_uc *) stbi__malloc(4 * w*h); + + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceded by a 2-byte data count for each row in the data, + // which we're going to just skip. + stbi__skip(s, h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + stbi_uc *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++, p += 4) + *p = (channel == 3 ? 255 : 0); + } else { + // Read the RLE data. + if (!stbi__psd_decode_rle(s, p, pixelCount)) { + STBI_FREE(out); + return stbi__errpuc("corrupt", "bad RLE data"); + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit (or 16-bit) value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + if (channel >= channelCount) { + // Fill this channel with default data. + if (bitdepth == 16 && bpc == 16) { + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + stbi__uint16 val = channel == 3 ? 65535 : 0; + for (i = 0; i < pixelCount; i++, q += 4) + *q = val; + } else { + stbi_uc *p = out+channel; + stbi_uc val = channel == 3 ? 255 : 0; + for (i = 0; i < pixelCount; i++, p += 4) + *p = val; + } + } else { + if (ri->bits_per_channel == 16) { // output bpc + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + for (i = 0; i < pixelCount; i++, q += 4) + *q = (stbi__uint16) stbi__get16be(s); + } else { + stbi_uc *p = out+channel; + if (bitdepth == 16) { // input bpc + for (i = 0; i < pixelCount; i++, p += 4) + *p = (stbi_uc) (stbi__get16be(s) >> 8); + } else { + for (i = 0; i < pixelCount; i++, p += 4) + *p = stbi__get8(s); + } + } + } + } + } + + // remove weird white matte from PSD + if (channelCount >= 4) { + if (ri->bits_per_channel == 16) { + for (i=0; i < w*h; ++i) { + stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i; + if (pixel[3] != 0 && pixel[3] != 65535) { + float a = pixel[3] / 65535.0f; + float ra = 1.0f / a; + float inv_a = 65535.0f * (1 - ra); + pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a); + pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a); + pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a); + } + } + } else { + for (i=0; i < w*h; ++i) { + unsigned char *pixel = out + 4*i; + if (pixel[3] != 0 && pixel[3] != 255) { + float a = pixel[3] / 255.0f; + float ra = 1.0f / a; + float inv_a = 255.0f * (1 - ra); + pixel[0] = (unsigned char) (pixel[0]*ra + inv_a); + pixel[1] = (unsigned char) (pixel[1]*ra + inv_a); + pixel[2] = (unsigned char) (pixel[2]*ra + inv_a); + } + } + } + } + + // convert to desired output format + if (req_comp && req_comp != 4) { + if (ri->bits_per_channel == 16) + out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h); + else + out = stbi__convert_format(out, 4, req_comp, w, h); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + if (comp) *comp = 4; + *y = h; + *x = w; + + return out; +} +#endif + +// ************************************************************************************************* +// Softimage PIC loader +// by Tom Seddon +// +// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format +// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/ + +#ifndef STBI_NO_PIC +static int stbi__pic_is4(stbi__context *s,const char *str) +{ + int i; + for (i=0; i<4; ++i) + if (stbi__get8(s) != (stbi_uc)str[i]) + return 0; + + return 1; +} + +static int stbi__pic_test_core(stbi__context *s) +{ + int i; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) + return 0; + + for(i=0;i<84;++i) + stbi__get8(s); + + if (!stbi__pic_is4(s,"PICT")) + return 0; + + return 1; +} + +typedef struct +{ + stbi_uc size,type,channel; +} stbi__pic_packet; + +static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest) +{ + int mask=0x80, i; + + for (i=0; i<4; ++i, mask>>=1) { + if (channel & mask) { + if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short"); + dest[i]=stbi__get8(s); + } + } + + return dest; +} + +static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src) +{ + int mask=0x80,i; + + for (i=0;i<4; ++i, mask>>=1) + if (channel&mask) + dest[i]=src[i]; +} + +static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result) +{ + int act_comp=0,num_packets=0,y,chained; + stbi__pic_packet packets[10]; + + // this will (should...) cater for even some bizarre stuff like having data + // for the same channel in multiple packets. + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return stbi__errpuc("bad format","too many packets"); + + packet = &packets[num_packets++]; + + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + + act_comp |= packet->channel; + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (reading packets)"); + if (packet->size != 8) return stbi__errpuc("bad format","packet isn't 8bpp"); + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel? + + for(y=0; ytype) { + default: + return stbi__errpuc("bad format","packet has bad compression type"); + + case 0: {//uncompressed + int x; + + for(x=0;xchannel,dest)) + return 0; + break; + } + + case 1://Pure RLE + { + int left=width, i; + + while (left>0) { + stbi_uc count,value[4]; + + count=stbi__get8(s); + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pure read count)"); + + if (count > left) + count = (stbi_uc) left; + + if (!stbi__readval(s,packet->channel,value)) return 0; + + for(i=0; ichannel,dest,value); + left -= count; + } + } + break; + + case 2: {//Mixed RLE + int left=width; + while (left>0) { + int count = stbi__get8(s), i; + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (mixed read count)"); + + if (count >= 128) { // Repeated + stbi_uc value[4]; + + if (count==128) + count = stbi__get16be(s); + else + count -= 127; + if (count > left) + return stbi__errpuc("bad file","scanline overrun"); + + if (!stbi__readval(s,packet->channel,value)) + return 0; + + for(i=0;ichannel,dest,value); + } else { // Raw + ++count; + if (count>left) return stbi__errpuc("bad file","scanline overrun"); + + for(i=0;ichannel,dest)) + return 0; + } + left-=count; + } + break; + } + } + } + } + + return result; +} + +static void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri) +{ + stbi_uc *result; + int i, x,y, internal_comp; + STBI_NOTUSED(ri); + + if (!comp) comp = &internal_comp; + + for (i=0; i<92; ++i) + stbi__get8(s); + + x = stbi__get16be(s); + y = stbi__get16be(s); + + if (y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)"); + if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC image too large to decode"); + + stbi__get32be(s); //skip `ratio' + stbi__get16be(s); //skip `fields' + stbi__get16be(s); //skip `pad' + + // intermediate buffer is RGBA + result = (stbi_uc *) stbi__malloc_mad3(x, y, 4, 0); + if (!result) return stbi__errpuc("outofmem", "Out of memory"); + memset(result, 0xff, x*y*4); + + if (!stbi__pic_load_core(s,x,y,comp, result)) { + STBI_FREE(result); + result=0; + } + *px = x; + *py = y; + if (req_comp == 0) req_comp = *comp; + result=stbi__convert_format(result,4,req_comp,x,y); + + return result; +} + +static int stbi__pic_test(stbi__context *s) +{ + int r = stbi__pic_test_core(s); + stbi__rewind(s); + return r; +} +#endif + +// ************************************************************************************************* +// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb + +#ifndef STBI_NO_GIF +typedef struct +{ + stbi__int16 prefix; + stbi_uc first; + stbi_uc suffix; +} stbi__gif_lzw; + +typedef struct +{ + int w,h; + stbi_uc *out; // output buffer (always 4 components) + stbi_uc *background; // The current "background" as far as a gif is concerned + stbi_uc *history; + int flags, bgindex, ratio, transparent, eflags; + stbi_uc pal[256][4]; + stbi_uc lpal[256][4]; + stbi__gif_lzw codes[8192]; + stbi_uc *color_table; + int parse, step; + int lflags; + int start_x, start_y; + int max_x, max_y; + int cur_x, cur_y; + int line_size; + int delay; +} stbi__gif; + +static int stbi__gif_test_raw(stbi__context *s) +{ + int sz; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0; + sz = stbi__get8(s); + if (sz != '9' && sz != '7') return 0; + if (stbi__get8(s) != 'a') return 0; + return 1; +} + +static int stbi__gif_test(stbi__context *s) +{ + int r = stbi__gif_test_raw(s); + stbi__rewind(s); + return r; +} + +static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp) +{ + int i; + for (i=0; i < num_entries; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + pal[i][3] = transp == i ? 0 : 255; + } +} + +static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info) +{ + stbi_uc version; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') + return stbi__err("not GIF", "Corrupt GIF"); + + version = stbi__get8(s); + if (version != '7' && version != '9') return stbi__err("not GIF", "Corrupt GIF"); + if (stbi__get8(s) != 'a') return stbi__err("not GIF", "Corrupt GIF"); + + stbi__g_failure_reason = ""; + g->w = stbi__get16le(s); + g->h = stbi__get16le(s); + g->flags = stbi__get8(s); + g->bgindex = stbi__get8(s); + g->ratio = stbi__get8(s); + g->transparent = -1; + + if (g->w > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (g->h > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + + if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments + + if (is_info) return 1; + + if (g->flags & 0x80) + stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1); + + return 1; +} + +static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__gif* g = (stbi__gif*) stbi__malloc(sizeof(stbi__gif)); + if (!g) return stbi__err("outofmem", "Out of memory"); + if (!stbi__gif_header(s, g, comp, 1)) { + STBI_FREE(g); + stbi__rewind( s ); + return 0; + } + if (x) *x = g->w; + if (y) *y = g->h; + STBI_FREE(g); + return 1; +} + +static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code) +{ + stbi_uc *p, *c; + int idx; + + // recurse to decode the prefixes, since the linked-list is backwards, + // and working backwards through an interleaved image would be nasty + if (g->codes[code].prefix >= 0) + stbi__out_gif_code(g, g->codes[code].prefix); + + if (g->cur_y >= g->max_y) return; + + idx = g->cur_x + g->cur_y; + p = &g->out[idx]; + g->history[idx / 4] = 1; + + c = &g->color_table[g->codes[code].suffix * 4]; + if (c[3] > 128) { // don't render transparent pixels; + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = c[3]; + } + g->cur_x += 4; + + if (g->cur_x >= g->max_x) { + g->cur_x = g->start_x; + g->cur_y += g->step; + + while (g->cur_y >= g->max_y && g->parse > 0) { + g->step = (1 << g->parse) * g->line_size; + g->cur_y = g->start_y + (g->step >> 1); + --g->parse; + } + } +} + +static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g) +{ + stbi_uc lzw_cs; + stbi__int32 len, init_code; + stbi__uint32 first; + stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear; + stbi__gif_lzw *p; + + lzw_cs = stbi__get8(s); + if (lzw_cs > 12) return NULL; + clear = 1 << lzw_cs; + first = 1; + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + bits = 0; + valid_bits = 0; + for (init_code = 0; init_code < clear; init_code++) { + g->codes[init_code].prefix = -1; + g->codes[init_code].first = (stbi_uc) init_code; + g->codes[init_code].suffix = (stbi_uc) init_code; + } + + // support no starting clear code + avail = clear+2; + oldcode = -1; + + len = 0; + for(;;) { + if (valid_bits < codesize) { + if (len == 0) { + len = stbi__get8(s); // start new block + if (len == 0) + return g->out; + } + --len; + bits |= (stbi__int32) stbi__get8(s) << valid_bits; + valid_bits += 8; + } else { + stbi__int32 code = bits & codemask; + bits >>= codesize; + valid_bits -= codesize; + // @OPTIMIZE: is there some way we can accelerate the non-clear path? + if (code == clear) { // clear code + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + avail = clear + 2; + oldcode = -1; + first = 0; + } else if (code == clear + 1) { // end of stream code + stbi__skip(s, len); + while ((len = stbi__get8(s)) > 0) + stbi__skip(s,len); + return g->out; + } else if (code <= avail) { + if (first) { + return stbi__errpuc("no clear code", "Corrupt GIF"); + } + + if (oldcode >= 0) { + p = &g->codes[avail++]; + if (avail > 8192) { + return stbi__errpuc("too many codes", "Corrupt GIF"); + } + + p->prefix = (stbi__int16) oldcode; + p->first = g->codes[oldcode].first; + p->suffix = (code == avail) ? p->first : g->codes[code].first; + } else if (code == avail) + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + + stbi__out_gif_code(g, (stbi__uint16) code); + + if ((avail & codemask) == 0 && avail <= 0x0FFF) { + codesize++; + codemask = (1 << codesize) - 1; + } + + oldcode = code; + } else { + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + } + } + } +} + +// this function is designed to support animated gifs, although stb_image doesn't support it +// two back is the image from two frames ago, used for a very specific disposal format +static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back) +{ + int dispose; + int first_frame; + int pi; + int pcount; + STBI_NOTUSED(req_comp); + + // on first frame, any non-written pixels get the background colour (non-transparent) + first_frame = 0; + if (g->out == 0) { + if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header + if (!stbi__mad3sizes_valid(4, g->w, g->h, 0)) + return stbi__errpuc("too large", "GIF image is too large"); + pcount = g->w * g->h; + g->out = (stbi_uc *) stbi__malloc(4 * pcount); + g->background = (stbi_uc *) stbi__malloc(4 * pcount); + g->history = (stbi_uc *) stbi__malloc(pcount); + if (!g->out || !g->background || !g->history) + return stbi__errpuc("outofmem", "Out of memory"); + + // image is treated as "transparent" at the start - ie, nothing overwrites the current background; + // background colour is only used for pixels that are not rendered first frame, after that "background" + // color refers to the color that was there the previous frame. + memset(g->out, 0x00, 4 * pcount); + memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent) + memset(g->history, 0x00, pcount); // pixels that were affected previous frame + first_frame = 1; + } else { + // second frame - how do we dispose of the previous one? + dispose = (g->eflags & 0x1C) >> 2; + pcount = g->w * g->h; + + if ((dispose == 3) && (two_back == 0)) { + dispose = 2; // if I don't have an image to revert back to, default to the old background + } + + if (dispose == 3) { // use previous graphic + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 ); + } + } + } else if (dispose == 2) { + // restore what was changed last frame to background before that frame; + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 ); + } + } + } else { + // This is a non-disposal case eithe way, so just + // leave the pixels as is, and they will become the new background + // 1: do not dispose + // 0: not specified. + } + + // background is what out is after the undoing of the previou frame; + memcpy( g->background, g->out, 4 * g->w * g->h ); + } + + // clear my history; + memset( g->history, 0x00, g->w * g->h ); // pixels that were affected previous frame + + for (;;) { + int tag = stbi__get8(s); + switch (tag) { + case 0x2C: /* Image Descriptor */ + { + stbi__int32 x, y, w, h; + stbi_uc *o; + + x = stbi__get16le(s); + y = stbi__get16le(s); + w = stbi__get16le(s); + h = stbi__get16le(s); + if (((x + w) > (g->w)) || ((y + h) > (g->h))) + return stbi__errpuc("bad Image Descriptor", "Corrupt GIF"); + + g->line_size = g->w * 4; + g->start_x = x * 4; + g->start_y = y * g->line_size; + g->max_x = g->start_x + w * 4; + g->max_y = g->start_y + h * g->line_size; + g->cur_x = g->start_x; + g->cur_y = g->start_y; + + // if the width of the specified rectangle is 0, that means + // we may not see *any* pixels or the image is malformed; + // to make sure this is caught, move the current y down to + // max_y (which is what out_gif_code checks). + if (w == 0) + g->cur_y = g->max_y; + + g->lflags = stbi__get8(s); + + if (g->lflags & 0x40) { + g->step = 8 * g->line_size; // first interlaced spacing + g->parse = 3; + } else { + g->step = g->line_size; + g->parse = 0; + } + + if (g->lflags & 0x80) { + stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1); + g->color_table = (stbi_uc *) g->lpal; + } else if (g->flags & 0x80) { + g->color_table = (stbi_uc *) g->pal; + } else + return stbi__errpuc("missing color table", "Corrupt GIF"); + + o = stbi__process_gif_raster(s, g); + if (!o) return NULL; + + // if this was the first frame, + pcount = g->w * g->h; + if (first_frame && (g->bgindex > 0)) { + // if first frame, any pixel not drawn to gets the background color + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi] == 0) { + g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be; + memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 ); + } + } + } + + return o; + } + + case 0x21: // Comment Extension. + { + int len; + int ext = stbi__get8(s); + if (ext == 0xF9) { // Graphic Control Extension. + len = stbi__get8(s); + if (len == 4) { + g->eflags = stbi__get8(s); + g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths. + + // unset old transparent + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 255; + } + if (g->eflags & 0x01) { + g->transparent = stbi__get8(s); + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 0; + } + } else { + // don't need transparent + stbi__skip(s, 1); + g->transparent = -1; + } + } else { + stbi__skip(s, len); + break; + } + } + while ((len = stbi__get8(s)) != 0) { + stbi__skip(s, len); + } + break; + } + + case 0x3B: // gif stream termination code + return (stbi_uc *) s; // using '1' causes warning on some compilers + + default: + return stbi__errpuc("unknown code", "Corrupt GIF"); + } + } +} + +static void *stbi__load_gif_main_outofmem(stbi__gif *g, stbi_uc *out, int **delays) +{ + STBI_FREE(g->out); + STBI_FREE(g->history); + STBI_FREE(g->background); + + if (out) STBI_FREE(out); + if (delays && *delays) STBI_FREE(*delays); + return stbi__errpuc("outofmem", "Out of memory"); +} + +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + if (stbi__gif_test(s)) { + int layers = 0; + stbi_uc *u = 0; + stbi_uc *out = 0; + stbi_uc *two_back = 0; + stbi__gif g; + int stride; + int out_size = 0; + int delays_size = 0; + + STBI_NOTUSED(out_size); + STBI_NOTUSED(delays_size); + + memset(&g, 0, sizeof(g)); + if (delays) { + *delays = 0; + } + + do { + u = stbi__gif_load_next(s, &g, comp, req_comp, two_back); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + + if (u) { + *x = g.w; + *y = g.h; + ++layers; + stride = g.w * g.h * 4; + + if (out) { + void *tmp = (stbi_uc*) STBI_REALLOC_SIZED( out, out_size, layers * stride ); + if (!tmp) + return stbi__load_gif_main_outofmem(&g, out, delays); + else { + out = (stbi_uc*) tmp; + out_size = layers * stride; + } + + if (delays) { + int *new_delays = (int*) STBI_REALLOC_SIZED( *delays, delays_size, sizeof(int) * layers ); + if (!new_delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + *delays = new_delays; + delays_size = layers * sizeof(int); + } + } else { + out = (stbi_uc*)stbi__malloc( layers * stride ); + if (!out) + return stbi__load_gif_main_outofmem(&g, out, delays); + out_size = layers * stride; + if (delays) { + *delays = (int*) stbi__malloc( layers * sizeof(int) ); + if (!*delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + delays_size = layers * sizeof(int); + } + } + memcpy( out + ((layers - 1) * stride), u, stride ); + if (layers >= 2) { + two_back = out - 2 * stride; + } + + if (delays) { + (*delays)[layers - 1U] = g.delay; + } + } + } while (u != 0); + + // free temp buffer; + STBI_FREE(g.out); + STBI_FREE(g.history); + STBI_FREE(g.background); + + // do the final conversion after loading everything; + if (req_comp && req_comp != 4) + out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h); + + *z = layers; + return out; + } else { + return stbi__errpuc("not GIF", "Image was not as a gif type."); + } +} + +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *u = 0; + stbi__gif g; + memset(&g, 0, sizeof(g)); + STBI_NOTUSED(ri); + + u = stbi__gif_load_next(s, &g, comp, req_comp, 0); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + if (u) { + *x = g.w; + *y = g.h; + + // moved conversion to after successful load so that the same + // can be done for multiple frames. + if (req_comp && req_comp != 4) + u = stbi__convert_format(u, 4, req_comp, g.w, g.h); + } else if (g.out) { + // if there was an error and we allocated an image buffer, free it! + STBI_FREE(g.out); + } + + // free buffers needed for multiple frame loading; + STBI_FREE(g.history); + STBI_FREE(g.background); + + return u; +} + +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp) +{ + return stbi__gif_info_raw(s,x,y,comp); +} +#endif + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int stbi__hdr_test_core(stbi__context *s, const char *signature) +{ + int i; + for (i=0; signature[i]; ++i) + if (stbi__get8(s) != signature[i]) + return 0; + stbi__rewind(s); + return 1; +} + +static int stbi__hdr_test(stbi__context* s) +{ + int r = stbi__hdr_test_core(s, "#?RADIANCE\n"); + stbi__rewind(s); + if(!r) { + r = stbi__hdr_test_core(s, "#?RGBE\n"); + stbi__rewind(s); + } + return r; +} + +#define STBI__HDR_BUFLEN 1024 +static char *stbi__hdr_gettoken(stbi__context *z, char *buffer) +{ + int len=0; + char c = '\0'; + + c = (char) stbi__get8(z); + + while (!stbi__at_eof(z) && c != '\n') { + buffer[len++] = c; + if (len == STBI__HDR_BUFLEN-1) { + // flush to end of line + while (!stbi__at_eof(z) && stbi__get8(z) != '\n') + ; + break; + } + c = (char) stbi__get8(z); + } + + buffer[len] = 0; + return buffer; +} + +static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if ( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + const char *headerToken; + STBI_NOTUSED(ri); + + // Check identifier + headerToken = stbi__hdr_gettoken(s,buffer); + if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 0) + return stbi__errpf("not HDR", "Corrupt HDR image"); + + // Parse header + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return stbi__errpf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = (int) strtol(token, NULL, 10); + + if (height > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + if (width > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + + *x = width; + *y = height; + + if (comp) *comp = 3; + if (req_comp == 0) req_comp = 3; + + if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0)) + return stbi__errpf("too large", "HDR image is too large"); + + // Read data + hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0); + if (!hdr_data) + return stbi__errpf("outofmem", "Out of memory"); + + // Load image data + // image data is stored as some number of sca + if ( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + stbi__getn(s, rgbe, 4); + stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = stbi__get8(s); + c2 = stbi__get8(s); + len = stbi__get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + stbi_uc rgbe[4]; + rgbe[0] = (stbi_uc) c1; + rgbe[1] = (stbi_uc) c2; + rgbe[2] = (stbi_uc) len; + rgbe[3] = (stbi_uc) stbi__get8(s); + stbi__hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + STBI_FREE(scanline); + goto main_decode_loop; // yes, this makes no sense + } + len <<= 8; + len |= stbi__get8(s); + if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) { + scanline = (stbi_uc *) stbi__malloc_mad2(width, 4, 0); + if (!scanline) { + STBI_FREE(hdr_data); + return stbi__errpf("outofmem", "Out of memory"); + } + } + + for (k = 0; k < 4; ++k) { + int nleft; + i = 0; + while ((nleft = width - i) > 0) { + count = stbi__get8(s); + if (count > 128) { + // Run + value = stbi__get8(s); + count -= 128; + if ((count == 0) || (count > nleft)) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + if ((count == 0) || (count > nleft)) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = stbi__get8(s); + } + } + } + for (i=0; i < width; ++i) + stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + if (scanline) + STBI_FREE(scanline); + } + + return hdr_data; +} + +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int dummy; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + if (stbi__hdr_test(s) == 0) { + stbi__rewind( s ); + return 0; + } + + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) { + stbi__rewind( s ); + return 0; + } + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *y = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *x = (int) strtol(token, NULL, 10); + *comp = 3; + return 1; +} +#endif // STBI_NO_HDR + +#ifndef STBI_NO_BMP +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp) +{ + void *p; + stbi__bmp_data info; + + info.all_a = 255; + p = stbi__bmp_parse_header(s, &info); + if (p == NULL) { + stbi__rewind( s ); + return 0; + } + if (x) *x = s->img_x; + if (y) *y = s->img_y; + if (comp) { + if (info.bpp == 24 && info.ma == 0xff000000) + *comp = 3; + else + *comp = info.ma ? 4 : 3; + } + return 1; +} +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp) +{ + int channelCount, dummy, depth; + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + *y = stbi__get32be(s); + *x = stbi__get32be(s); + depth = stbi__get16be(s); + if (depth != 8 && depth != 16) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 3) { + stbi__rewind( s ); + return 0; + } + *comp = 4; + return 1; +} + +static int stbi__psd_is16(stbi__context *s) +{ + int channelCount, depth; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + STBI_NOTUSED(stbi__get32be(s)); + STBI_NOTUSED(stbi__get32be(s)); + depth = stbi__get16be(s); + if (depth != 16) { + stbi__rewind( s ); + return 0; + } + return 1; +} +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp) +{ + int act_comp=0,num_packets=0,chained,dummy; + stbi__pic_packet packets[10]; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) { + stbi__rewind(s); + return 0; + } + + stbi__skip(s, 88); + + *x = stbi__get16be(s); + *y = stbi__get16be(s); + if (stbi__at_eof(s)) { + stbi__rewind( s); + return 0; + } + if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) { + stbi__rewind( s ); + return 0; + } + + stbi__skip(s, 8); + + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return 0; + + packet = &packets[num_packets++]; + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + act_comp |= packet->channel; + + if (stbi__at_eof(s)) { + stbi__rewind( s ); + return 0; + } + if (packet->size != 8) { + stbi__rewind( s ); + return 0; + } + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); + + return 1; +} +#endif + +// ************************************************************************************************* +// Portable Gray Map and Portable Pixel Map loader +// by Ken Miller +// +// PGM: http://netpbm.sourceforge.net/doc/pgm.html +// PPM: http://netpbm.sourceforge.net/doc/ppm.html +// +// Known limitations: +// Does not support comments in the header section +// Does not support ASCII image data (formats P2 and P3) + +#ifndef STBI_NO_PNM + +static int stbi__pnm_test(stbi__context *s) +{ + char p, t; + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind( s ); + return 0; + } + return 1; +} + +static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + STBI_NOTUSED(ri); + + ri->bits_per_channel = stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n); + if (ri->bits_per_channel == 0) + return 0; + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + + if (!stbi__mad4sizes_valid(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0)) + return stbi__errpuc("too large", "PNM too large"); + + out = (stbi_uc *) stbi__malloc_mad4(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + if (!stbi__getn(s, out, s->img_n * s->img_x * s->img_y * (ri->bits_per_channel / 8))) { + STBI_FREE(out); + return stbi__errpuc("bad PNM", "PNM file truncated"); + } + + if (req_comp && req_comp != s->img_n) { + if (ri->bits_per_channel == 16) { + out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, s->img_n, req_comp, s->img_x, s->img_y); + } else { + out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y); + } + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + return out; +} + +static int stbi__pnm_isspace(char c) +{ + return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r'; +} + +static void stbi__pnm_skip_whitespace(stbi__context *s, char *c) +{ + for (;;) { + while (!stbi__at_eof(s) && stbi__pnm_isspace(*c)) + *c = (char) stbi__get8(s); + + if (stbi__at_eof(s) || *c != '#') + break; + + while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' ) + *c = (char) stbi__get8(s); + } +} + +static int stbi__pnm_isdigit(char c) +{ + return c >= '0' && c <= '9'; +} + +static int stbi__pnm_getinteger(stbi__context *s, char *c) +{ + int value = 0; + + while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) { + value = value*10 + (*c - '0'); + *c = (char) stbi__get8(s); + if((value > 214748364) || (value == 214748364 && *c > '7')) + return stbi__err("integer parse overflow", "Parsing an integer in the PPM header overflowed a 32-bit int"); + } + + return value; +} + +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp) +{ + int maxv, dummy; + char c, p, t; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + stbi__rewind(s); + + // Get identifier + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind(s); + return 0; + } + + *comp = (t == '6') ? 3 : 1; // '5' is 1-component .pgm; '6' is 3-component .ppm + + c = (char) stbi__get8(s); + stbi__pnm_skip_whitespace(s, &c); + + *x = stbi__pnm_getinteger(s, &c); // read width + if(*x == 0) + return stbi__err("invalid width", "PPM image header had zero or overflowing width"); + stbi__pnm_skip_whitespace(s, &c); + + *y = stbi__pnm_getinteger(s, &c); // read height + if (*y == 0) + return stbi__err("invalid width", "PPM image header had zero or overflowing width"); + stbi__pnm_skip_whitespace(s, &c); + + maxv = stbi__pnm_getinteger(s, &c); // read max value + if (maxv > 65535) + return stbi__err("max value > 65535", "PPM image supports only 8-bit and 16-bit images"); + else if (maxv > 255) + return 16; + else + return 8; +} + +static int stbi__pnm_is16(stbi__context *s) +{ + if (stbi__pnm_info(s, NULL, NULL, NULL) == 16) + return 1; + return 0; +} +#endif + +static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp) +{ + #ifndef STBI_NO_JPEG + if (stbi__jpeg_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNG + if (stbi__png_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_GIF + if (stbi__gif_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_BMP + if (stbi__bmp_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PIC + if (stbi__pic_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_info(s, x, y, comp)) return 1; + #endif + + // test tga last because it's a crappy test! + #ifndef STBI_NO_TGA + if (stbi__tga_info(s, x, y, comp)) + return 1; + #endif + return stbi__err("unknown image type", "Image not of any known type, or corrupt"); +} + +static int stbi__is_16_main(stbi__context *s) +{ + #ifndef STBI_NO_PNG + if (stbi__png_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_is16(s)) return 1; + #endif + return 0; +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_info_from_file(f, x, y, comp); + fclose(f); + return result; +} + +STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__info_main(&s,x,y,comp); + fseek(f,pos,SEEK_SET); + return r; +} + +STBIDEF int stbi_is_16_bit(char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_is_16_bit_from_file(f); + fclose(f); + return result; +} + +STBIDEF int stbi_is_16_bit_from_file(FILE *f) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__is_16_main(&s); + fseek(f,pos,SEEK_SET); + return r; +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__is_16_main(&s); +} + +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__is_16_main(&s); +} + +#endif // STB_IMAGE_IMPLEMENTATION + +/* + revision history: + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug + 1-bit BMP + *_is_16_bit api + avoid warnings + 2.16 (2017-07-23) all functions have 16-bit variants; + STBI_NO_STDIO works again; + compilation fixes; + fix rounding in unpremultiply; + optimize vertical flip; + disable raw_len validation; + documentation fixes + 2.15 (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode; + warning fixes; disable run-time SSE detection on gcc; + uniform handling of optional "return" values; + thread-safe initialization of zlib tables + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-11-29) add 16-bit API, only supported for PNG right now + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) allocate large structures on the stack + remove white matting for transparent PSD + fix reported channel count for PNG & BMP + re-enable SSE2 in non-gcc 64-bit + support RGB-formatted JPEG + read 16-bit PNGs (only as 8-bit) + 2.10 (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED + 2.09 (2016-01-16) allow comments in PNM files + 16-bit-per-pixel TGA (not bit-per-component) + info() for TGA could break due to .hdr handling + info() for BMP to shares code instead of sloppy parse + can use STBI_REALLOC_SIZED if allocator doesn't support realloc + code cleanup + 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA + 2.07 (2015-09-13) fix compiler warnings + partial animated GIF support + limited 16-bpc PSD support + #ifdef unused functions + bug with < 92 byte PIC,PNM,HDR,TGA + 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value + 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning + 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit + 2.03 (2015-04-12) extra corruption checking (mmozeiko) + stbi_set_flip_vertically_on_load (nguillemot) + fix NEON support; fix mingw support + 2.02 (2015-01-19) fix incorrect assert, fix warning + 2.01 (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2 + 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG + 2.00 (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg) + progressive JPEG (stb) + PGM/PPM support (Ken Miller) + STBI_MALLOC,STBI_REALLOC,STBI_FREE + GIF bugfix -- seemingly never worked + STBI_NO_*, STBI_ONLY_* + 1.48 (2014-12-14) fix incorrectly-named assert() + 1.47 (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb) + optimize PNG (ryg) + fix bug in interlaced PNG with user-specified channel count (stb) + 1.46 (2014-08-26) + fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG + 1.45 (2014-08-16) + fix MSVC-ARM internal compiler error by wrapping malloc + 1.44 (2014-08-07) + various warning fixes from Ronny Chevalier + 1.43 (2014-07-15) + fix MSVC-only compiler problem in code changed in 1.42 + 1.42 (2014-07-09) + don't define _CRT_SECURE_NO_WARNINGS (affects user code) + fixes to stbi__cleanup_jpeg path + added STBI_ASSERT to avoid requiring assert.h + 1.41 (2014-06-25) + fix search&replace from 1.36 that messed up comments/error messages + 1.40 (2014-06-22) + fix gcc struct-initialization warning + 1.39 (2014-06-15) + fix to TGA optimization when req_comp != number of components in TGA; + fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite) + add support for BMP version 5 (more ignored fields) + 1.38 (2014-06-06) + suppress MSVC warnings on integer casts truncating values + fix accidental rename of 'skip' field of I/O + 1.37 (2014-06-04) + remove duplicate typedef + 1.36 (2014-06-03) + convert to header file single-file library + if de-iphone isn't set, load iphone images color-swapped instead of returning NULL + 1.35 (2014-05-27) + various warnings + fix broken STBI_SIMD path + fix bug where stbi_load_from_file no longer left file pointer in correct place + fix broken non-easy path for 32-bit BMP (possibly never used) + TGA optimization by Arseny Kapoulkine + 1.34 (unknown) + use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case + 1.33 (2011-07-14) + make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements + 1.32 (2011-07-13) + support for "info" function for all supported filetypes (SpartanJ) + 1.31 (2011-06-20) + a few more leak fixes, bug in PNG handling (SpartanJ) + 1.30 (2011-06-11) + added ability to load files via callbacks to accomidate custom input streams (Ben Wenger) + removed deprecated format-specific test/load functions + removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway + error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha) + fix inefficiency in decoding 32-bit BMP (David Woo) + 1.29 (2010-08-16) + various warning fixes from Aurelien Pocheville + 1.28 (2010-08-01) + fix bug in GIF palette transparency (SpartanJ) + 1.27 (2010-08-01) + cast-to-stbi_uc to fix warnings + 1.26 (2010-07-24) + fix bug in file buffering for PNG reported by SpartanJ + 1.25 (2010-07-17) + refix trans_data warning (Won Chun) + 1.24 (2010-07-12) + perf improvements reading from files on platforms with lock-heavy fgetc() + minor perf improvements for jpeg + deprecated type-specific functions so we'll get feedback if they're needed + attempt to fix trans_data warning (Won Chun) + 1.23 fixed bug in iPhone support + 1.22 (2010-07-10) + removed image *writing* support + stbi_info support from Jetro Lauha + GIF support from Jean-Marc Lienher + iPhone PNG-extensions from James Brown + warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva) + 1.21 fix use of 'stbi_uc' in header (reported by jon blow) + 1.20 added support for Softimage PIC, by Tom Seddon + 1.19 bug in interlaced PNG corruption check (found by ryg) + 1.18 (2008-08-02) + fix a threading bug (local mutable static) + 1.17 support interlaced PNG + 1.16 major bugfix - stbi__convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug + header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant + 0.50 (2006-11-19) + first released version +*/ + + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/platform/stb_impl.c b/platform/stb_impl.c new file mode 100644 index 0000000..18d641c --- /dev/null +++ b/platform/stb_impl.c @@ -0,0 +1,112 @@ +// Implementation of platform code when using stb. +// stb is pretty straightforward so no need for any +// fancy platform-specific goodies like sokol_impl +#include "stb_rect_pack.h" +#include "stb_truetype.h" +#include "stb_image.h" +#include "os.h" +#include "compiler.h" + +THREADLOCAL static struct allocator *stb_allocator = &os_allocator_default; + +// For STB image +void *os_malloc_emulate_stb(size_t size) { return os_malloc_emulate_a(size, stb_allocator); } +void *os_realloc_emulate_stb(void *p, size_t newsize) { return os_realloc_emulate_a(p, newsize, stb_allocator); } +void os_free_emulate_stb(void *p) { return os_free_emulate_a(p, stb_allocator); } + +// For STB ttf +void *os_malloc_emulate_stb_a(size_t size, void *a) { return os_malloc_emulate_a(size, (struct allocator *)a); } +void os_free_emulate_stb_a(void *p, void *a) { return os_free_emulate_a(p, (struct allocator *)a); } + +struct image platform_os_image_decode(const void *mem, size_t size, struct allocator *a) +{ + stb_allocator = a; + struct image img = {}; + if (size > INT_MAX || size < 0) + return img; + int x,y,channels; + void *p = stbi_load_from_memory(mem, (int)size, &x, &y, &channels, 0); + if (!p) + return img; + img._i = os_blob_from_emulate(p); + img.pixels = p; + img.width = x; + img.height = y; + img.bytesize = img.width * img.height * channels; + img.format = IMAGE_FORMAT_RGBA8; // TODO: Probably... IDK + return img; +} + +struct platform_gfx_font_i +{ + struct platform_gfx_font_desc desc; + stbtt_fontinfo info; +}; +union platform_gfx_font platform_gfx_font_construct(union platform_window window, const struct platform_gfx_font_desc *desc) +{ + (void)window; + + struct platform_gfx_font_i font; + memset(&font, 0, sizeof(font)); + memcpy(&font.desc, desc, sizeof(font.desc)); + + if (!stbtt_InitFont(&font.info, desc->data.memory, 0)) + return (union platform_gfx_font){}; + + struct platform_gfx_font_i *p_font = os_malloc_emulate_a(sizeof(*p_font), desc->a); + if (p_font) + memcpy(p_font, &font, sizeof(*p_font)); + return (union platform_gfx_font){p_font}; +} +void platform_gfx_font_destruct(union platform_window window, union platform_gfx_font font) +{ + (void)window; + + // Free the data provided to the font and then the font itself + // For STBTT, it looks like the font itself doesn't make dynamic + // allocations when the font is init'd, only when it rasterises + os_free(font._i->desc.data_free, font._i->desc.data); + os_free_emulate_a(font._i, font._i->desc.a); +} +struct image platform_gfx_font_texture(union platform_window window, union platform_gfx_font font, float fontsize, struct platform_gfx_font_bmglyph *glyphs) +{ + (void)window; + + // TODO: We want to take various unicode codepoint ranges. + // For now, we only support ASCII + int tw = (int)(fontsize * 12); + int th = (int)(fontsize * 12); + + struct blob b = os_alloc(font._i->desc.a, tw*th, 0); + if (!b.memory) + return (struct image){}; + stbtt_pack_context spc; + if (stbtt_PackBegin(&spc, b.memory, tw, th, 0, 1, font._i->desc.a)) { + // Printable ASCII ranges from 0x20 ( ) to 0x7e (~) + // NOTE: stbtt_packedchar cast works here for now and is easy + // If you change the bmglyph struct, you will pay dearly + int success = stbtt_PackFontRange(&spc, font._i->desc.data.memory, 0, fontsize, 0x20, 0x7e - 0x20, (stbtt_packedchar *)glyphs); + stbtt_PackEnd(&spc); + + if (success) + return (struct image){b, b.memory, b.size, tw, th, IMAGE_FORMAT_R8}; + os_free(font._i->desc.a, b); + } + return (struct image){}; +} +struct platform_gfx_font_info platform_gfx_font_info_get(union platform_window window, union platform_gfx_font f) +{ + (void)window; + struct platform_gfx_font_i *font = f._i; + if (!font) + return (struct platform_gfx_font_info){}; + + struct platform_gfx_font_info info; + stbtt_GetFontVMetrics(&font->info, &info.ascent, &info.descent, &info.linegap); + return info; +} +struct allocator *platform_gfx_font_allocator(union platform_window window, union platform_gfx_font font) +{ + (void)window; + return font._i->desc.a; +} diff --git a/platform/stb_impl.h b/platform/stb_impl.h new file mode 100644 index 0000000..3728b1e --- /dev/null +++ b/platform/stb_impl.h @@ -0,0 +1,19 @@ +#ifndef STB_IMPL_H +#define STB_IMPL_H + +#if defined(STB_IMAGE_IMPLEMENTATION) && (!defined(STBI_MALLOC) || !defined(STBI_REALLOC) || !defined(STBI_FREE)) +#error If you're compiling this file with STB_IMAGE_IMPLEMENTATION, include the STBI malloc functions +#endif +#if defined(STB_TRUETYPE_IMPLEMENTATION) && (!defined(STBTT_malloc) || !defined(STBTT_free)) +#error If you're compiling this file with STB_TRUETYPE_IMPLEMENTATION, include the STBTT malloc functions +#endif + +#include "os.h" +extern struct allocator *stb_allocator; +void *os_malloc_emulate_stb(size_t size); +void *os_realloc_emulate_stb(void *p, size_t newsize); +void os_free_emulate_stb(void *p); +void *os_malloc_emulate_stb_a(size_t size, void *a); +void os_free_emulate_stb_a(void *p, void *a); + +#endif diff --git a/platform/stb_rect_pack.h b/platform/stb_rect_pack.h new file mode 100644 index 0000000..6a633ce --- /dev/null +++ b/platform/stb_rect_pack.h @@ -0,0 +1,623 @@ +// stb_rect_pack.h - v1.01 - public domain - rectangle packing +// Sean Barrett 2014 +// +// Useful for e.g. packing rectangular textures into an atlas. +// Does not do rotation. +// +// Before #including, +// +// #define STB_RECT_PACK_IMPLEMENTATION +// +// in the file that you want to have the implementation. +// +// Not necessarily the awesomest packing method, but better than +// the totally naive one in stb_truetype (which is primarily what +// this is meant to replace). +// +// Has only had a few tests run, may have issues. +// +// More docs to come. +// +// No memory allocations; uses qsort() and assert() from stdlib. +// Can override those by defining STBRP_SORT and STBRP_ASSERT. +// +// This library currently uses the Skyline Bottom-Left algorithm. +// +// Please note: better rectangle packers are welcome! Please +// implement them to the same API, but with a different init +// function. +// +// Credits +// +// Library +// Sean Barrett +// Minor features +// Martins Mozeiko +// github:IntellectualKitty +// +// Bugfixes / warning fixes +// Jeremy Jaussaud +// Fabian Giesen +// +// Version history: +// +// 1.01 (2021-07-11) always use large rect mode, expose STBRP__MAXVAL in public section +// 1.00 (2019-02-25) avoid small space waste; gracefully fail too-wide rectangles +// 0.99 (2019-02-07) warning fixes +// 0.11 (2017-03-03) return packing success/fail result +// 0.10 (2016-10-25) remove cast-away-const to avoid warnings +// 0.09 (2016-08-27) fix compiler warnings +// 0.08 (2015-09-13) really fix bug with empty rects (w=0 or h=0) +// 0.07 (2015-09-13) fix bug with empty rects (w=0 or h=0) +// 0.06 (2015-04-15) added STBRP_SORT to allow replacing qsort +// 0.05: added STBRP_ASSERT to allow replacing assert +// 0.04: fixed minor bug in STBRP_LARGE_RECTS support +// 0.01: initial release +// +// LICENSE +// +// See end of file for license information. + +////////////////////////////////////////////////////////////////////////////// +// +// INCLUDE SECTION +// + +#ifndef STB_INCLUDE_STB_RECT_PACK_H +#define STB_INCLUDE_STB_RECT_PACK_H + +#define STB_RECT_PACK_VERSION 1 + +#ifdef STBRP_STATIC +#define STBRP_DEF static +#else +#define STBRP_DEF extern +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +typedef struct stbrp_context stbrp_context; +typedef struct stbrp_node stbrp_node; +typedef struct stbrp_rect stbrp_rect; + +typedef int stbrp_coord; + +#define STBRP__MAXVAL 0x7fffffff +// Mostly for internal use, but this is the maximum supported coordinate value. + +STBRP_DEF int stbrp_pack_rects (stbrp_context *context, stbrp_rect *rects, int num_rects); +// Assign packed locations to rectangles. The rectangles are of type +// 'stbrp_rect' defined below, stored in the array 'rects', and there +// are 'num_rects' many of them. +// +// Rectangles which are successfully packed have the 'was_packed' flag +// set to a non-zero value and 'x' and 'y' store the minimum location +// on each axis (i.e. bottom-left in cartesian coordinates, top-left +// if you imagine y increasing downwards). Rectangles which do not fit +// have the 'was_packed' flag set to 0. +// +// You should not try to access the 'rects' array from another thread +// while this function is running, as the function temporarily reorders +// the array while it executes. +// +// To pack into another rectangle, you need to call stbrp_init_target +// again. To continue packing into the same rectangle, you can call +// this function again. Calling this multiple times with multiple rect +// arrays will probably produce worse packing results than calling it +// a single time with the full rectangle array, but the option is +// available. +// +// The function returns 1 if all of the rectangles were successfully +// packed and 0 otherwise. + +struct stbrp_rect +{ + // reserved for your use: + int id; + + // input: + stbrp_coord w, h; + + // output: + stbrp_coord x, y; + int was_packed; // non-zero if valid packing + +}; // 16 bytes, nominally + + +STBRP_DEF void stbrp_init_target (stbrp_context *context, int width, int height, stbrp_node *nodes, int num_nodes); +// Initialize a rectangle packer to: +// pack a rectangle that is 'width' by 'height' in dimensions +// using temporary storage provided by the array 'nodes', which is 'num_nodes' long +// +// You must call this function every time you start packing into a new target. +// +// There is no "shutdown" function. The 'nodes' memory must stay valid for +// the following stbrp_pack_rects() call (or calls), but can be freed after +// the call (or calls) finish. +// +// Note: to guarantee best results, either: +// 1. make sure 'num_nodes' >= 'width' +// or 2. call stbrp_allow_out_of_mem() defined below with 'allow_out_of_mem = 1' +// +// If you don't do either of the above things, widths will be quantized to multiples +// of small integers to guarantee the algorithm doesn't run out of temporary storage. +// +// If you do #2, then the non-quantized algorithm will be used, but the algorithm +// may run out of temporary storage and be unable to pack some rectangles. + +STBRP_DEF void stbrp_setup_allow_out_of_mem (stbrp_context *context, int allow_out_of_mem); +// Optionally call this function after init but before doing any packing to +// change the handling of the out-of-temp-memory scenario, described above. +// If you call init again, this will be reset to the default (false). + + +STBRP_DEF void stbrp_setup_heuristic (stbrp_context *context, int heuristic); +// Optionally select which packing heuristic the library should use. Different +// heuristics will produce better/worse results for different data sets. +// If you call init again, this will be reset to the default. + +enum +{ + STBRP_HEURISTIC_Skyline_default=0, + STBRP_HEURISTIC_Skyline_BL_sortHeight = STBRP_HEURISTIC_Skyline_default, + STBRP_HEURISTIC_Skyline_BF_sortHeight +}; + + +////////////////////////////////////////////////////////////////////////////// +// +// the details of the following structures don't matter to you, but they must +// be visible so you can handle the memory allocations for them + +struct stbrp_node +{ + stbrp_coord x,y; + stbrp_node *next; +}; + +struct stbrp_context +{ + int width; + int height; + int align; + int init_mode; + int heuristic; + int num_nodes; + stbrp_node *active_head; + stbrp_node *free_head; + stbrp_node extra[2]; // we allocate two extra nodes so optimal user-node-count is 'width' not 'width+2' +}; + +#ifdef __cplusplus +} +#endif + +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// IMPLEMENTATION SECTION +// + +#ifdef STB_RECT_PACK_IMPLEMENTATION +#ifndef STBRP_SORT +#include +#define STBRP_SORT qsort +#endif + +#ifndef STBRP_ASSERT +#include +#define STBRP_ASSERT assert +#endif + +#ifdef _MSC_VER +#define STBRP__NOTUSED(v) (void)(v) +#define STBRP__CDECL __cdecl +#else +#define STBRP__NOTUSED(v) (void)sizeof(v) +#define STBRP__CDECL +#endif + +enum +{ + STBRP__INIT_skyline = 1 +}; + +STBRP_DEF void stbrp_setup_heuristic(stbrp_context *context, int heuristic) +{ + switch (context->init_mode) { + case STBRP__INIT_skyline: + STBRP_ASSERT(heuristic == STBRP_HEURISTIC_Skyline_BL_sortHeight || heuristic == STBRP_HEURISTIC_Skyline_BF_sortHeight); + context->heuristic = heuristic; + break; + default: + STBRP_ASSERT(0); + } +} + +STBRP_DEF void stbrp_setup_allow_out_of_mem(stbrp_context *context, int allow_out_of_mem) +{ + if (allow_out_of_mem) + // if it's ok to run out of memory, then don't bother aligning them; + // this gives better packing, but may fail due to OOM (even though + // the rectangles easily fit). @TODO a smarter approach would be to only + // quantize once we've hit OOM, then we could get rid of this parameter. + context->align = 1; + else { + // if it's not ok to run out of memory, then quantize the widths + // so that num_nodes is always enough nodes. + // + // I.e. num_nodes * align >= width + // align >= width / num_nodes + // align = ceil(width/num_nodes) + + context->align = (context->width + context->num_nodes-1) / context->num_nodes; + } +} + +STBRP_DEF void stbrp_init_target(stbrp_context *context, int width, int height, stbrp_node *nodes, int num_nodes) +{ + int i; + + for (i=0; i < num_nodes-1; ++i) + nodes[i].next = &nodes[i+1]; + nodes[i].next = NULL; + context->init_mode = STBRP__INIT_skyline; + context->heuristic = STBRP_HEURISTIC_Skyline_default; + context->free_head = &nodes[0]; + context->active_head = &context->extra[0]; + context->width = width; + context->height = height; + context->num_nodes = num_nodes; + stbrp_setup_allow_out_of_mem(context, 0); + + // node 0 is the full width, node 1 is the sentinel (lets us not store width explicitly) + context->extra[0].x = 0; + context->extra[0].y = 0; + context->extra[0].next = &context->extra[1]; + context->extra[1].x = (stbrp_coord) width; + context->extra[1].y = (1<<30); + context->extra[1].next = NULL; +} + +// find minimum y position if it starts at x1 +static int stbrp__skyline_find_min_y(stbrp_context *c, stbrp_node *first, int x0, int width, int *pwaste) +{ + stbrp_node *node = first; + int x1 = x0 + width; + int min_y, visited_width, waste_area; + + STBRP__NOTUSED(c); + + STBRP_ASSERT(first->x <= x0); + + #if 0 + // skip in case we're past the node + while (node->next->x <= x0) + ++node; + #else + STBRP_ASSERT(node->next->x > x0); // we ended up handling this in the caller for efficiency + #endif + + STBRP_ASSERT(node->x <= x0); + + min_y = 0; + waste_area = 0; + visited_width = 0; + while (node->x < x1) { + if (node->y > min_y) { + // raise min_y higher. + // we've accounted for all waste up to min_y, + // but we'll now add more waste for everything we've visted + waste_area += visited_width * (node->y - min_y); + min_y = node->y; + // the first time through, visited_width might be reduced + if (node->x < x0) + visited_width += node->next->x - x0; + else + visited_width += node->next->x - node->x; + } else { + // add waste area + int under_width = node->next->x - node->x; + if (under_width + visited_width > width) + under_width = width - visited_width; + waste_area += under_width * (min_y - node->y); + visited_width += under_width; + } + node = node->next; + } + + *pwaste = waste_area; + return min_y; +} + +typedef struct +{ + int x,y; + stbrp_node **prev_link; +} stbrp__findresult; + +static stbrp__findresult stbrp__skyline_find_best_pos(stbrp_context *c, int width, int height) +{ + int best_waste = (1<<30), best_x, best_y = (1 << 30); + stbrp__findresult fr; + stbrp_node **prev, *node, *tail, **best = NULL; + + // align to multiple of c->align + width = (width + c->align - 1); + width -= width % c->align; + STBRP_ASSERT(width % c->align == 0); + + // if it can't possibly fit, bail immediately + if (width > c->width || height > c->height) { + fr.prev_link = NULL; + fr.x = fr.y = 0; + return fr; + } + + node = c->active_head; + prev = &c->active_head; + while (node->x + width <= c->width) { + int y,waste; + y = stbrp__skyline_find_min_y(c, node, node->x, width, &waste); + if (c->heuristic == STBRP_HEURISTIC_Skyline_BL_sortHeight) { // actually just want to test BL + // bottom left + if (y < best_y) { + best_y = y; + best = prev; + } + } else { + // best-fit + if (y + height <= c->height) { + // can only use it if it first vertically + if (y < best_y || (y == best_y && waste < best_waste)) { + best_y = y; + best_waste = waste; + best = prev; + } + } + } + prev = &node->next; + node = node->next; + } + + best_x = (best == NULL) ? 0 : (*best)->x; + + // if doing best-fit (BF), we also have to try aligning right edge to each node position + // + // e.g, if fitting + // + // ____________________ + // |____________________| + // + // into + // + // | | + // | ____________| + // |____________| + // + // then right-aligned reduces waste, but bottom-left BL is always chooses left-aligned + // + // This makes BF take about 2x the time + + if (c->heuristic == STBRP_HEURISTIC_Skyline_BF_sortHeight) { + tail = c->active_head; + node = c->active_head; + prev = &c->active_head; + // find first node that's admissible + while (tail->x < width) + tail = tail->next; + while (tail) { + int xpos = tail->x - width; + int y,waste; + STBRP_ASSERT(xpos >= 0); + // find the left position that matches this + while (node->next->x <= xpos) { + prev = &node->next; + node = node->next; + } + STBRP_ASSERT(node->next->x > xpos && node->x <= xpos); + y = stbrp__skyline_find_min_y(c, node, xpos, width, &waste); + if (y + height <= c->height) { + if (y <= best_y) { + if (y < best_y || waste < best_waste || (waste==best_waste && xpos < best_x)) { + best_x = xpos; + STBRP_ASSERT(y <= best_y); + best_y = y; + best_waste = waste; + best = prev; + } + } + } + tail = tail->next; + } + } + + fr.prev_link = best; + fr.x = best_x; + fr.y = best_y; + return fr; +} + +static stbrp__findresult stbrp__skyline_pack_rectangle(stbrp_context *context, int width, int height) +{ + // find best position according to heuristic + stbrp__findresult res = stbrp__skyline_find_best_pos(context, width, height); + stbrp_node *node, *cur; + + // bail if: + // 1. it failed + // 2. the best node doesn't fit (we don't always check this) + // 3. we're out of memory + if (res.prev_link == NULL || res.y + height > context->height || context->free_head == NULL) { + res.prev_link = NULL; + return res; + } + + // on success, create new node + node = context->free_head; + node->x = (stbrp_coord) res.x; + node->y = (stbrp_coord) (res.y + height); + + context->free_head = node->next; + + // insert the new node into the right starting point, and + // let 'cur' point to the remaining nodes needing to be + // stiched back in + + cur = *res.prev_link; + if (cur->x < res.x) { + // preserve the existing one, so start testing with the next one + stbrp_node *next = cur->next; + cur->next = node; + cur = next; + } else { + *res.prev_link = node; + } + + // from here, traverse cur and free the nodes, until we get to one + // that shouldn't be freed + while (cur->next && cur->next->x <= res.x + width) { + stbrp_node *next = cur->next; + // move the current node to the free list + cur->next = context->free_head; + context->free_head = cur; + cur = next; + } + + // stitch the list back in + node->next = cur; + + if (cur->x < res.x + width) + cur->x = (stbrp_coord) (res.x + width); + +#ifdef _DEBUG + cur = context->active_head; + while (cur->x < context->width) { + STBRP_ASSERT(cur->x < cur->next->x); + cur = cur->next; + } + STBRP_ASSERT(cur->next == NULL); + + { + int count=0; + cur = context->active_head; + while (cur) { + cur = cur->next; + ++count; + } + cur = context->free_head; + while (cur) { + cur = cur->next; + ++count; + } + STBRP_ASSERT(count == context->num_nodes+2); + } +#endif + + return res; +} + +static int STBRP__CDECL rect_height_compare(const void *a, const void *b) +{ + const stbrp_rect *p = (const stbrp_rect *) a; + const stbrp_rect *q = (const stbrp_rect *) b; + if (p->h > q->h) + return -1; + if (p->h < q->h) + return 1; + return (p->w > q->w) ? -1 : (p->w < q->w); +} + +static int STBRP__CDECL rect_original_order(const void *a, const void *b) +{ + const stbrp_rect *p = (const stbrp_rect *) a; + const stbrp_rect *q = (const stbrp_rect *) b; + return (p->was_packed < q->was_packed) ? -1 : (p->was_packed > q->was_packed); +} + +STBRP_DEF int stbrp_pack_rects(stbrp_context *context, stbrp_rect *rects, int num_rects) +{ + int i, all_rects_packed = 1; + + // we use the 'was_packed' field internally to allow sorting/unsorting + for (i=0; i < num_rects; ++i) { + rects[i].was_packed = i; + } + + // sort according to heuristic + STBRP_SORT(rects, num_rects, sizeof(rects[0]), rect_height_compare); + + for (i=0; i < num_rects; ++i) { + if (rects[i].w == 0 || rects[i].h == 0) { + rects[i].x = rects[i].y = 0; // empty rect needs no space + } else { + stbrp__findresult fr = stbrp__skyline_pack_rectangle(context, rects[i].w, rects[i].h); + if (fr.prev_link) { + rects[i].x = (stbrp_coord) fr.x; + rects[i].y = (stbrp_coord) fr.y; + } else { + rects[i].x = rects[i].y = STBRP__MAXVAL; + } + } + } + + // unsort + STBRP_SORT(rects, num_rects, sizeof(rects[0]), rect_original_order); + + // set was_packed flags and all_rects_packed status + for (i=0; i < num_rects; ++i) { + rects[i].was_packed = !(rects[i].x == STBRP__MAXVAL && rects[i].y == STBRP__MAXVAL); + if (!rects[i].was_packed) + all_rects_packed = 0; + } + + // return the all_rects_packed status + return all_rects_packed; +} +#endif + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/platform/stb_truetype.h b/platform/stb_truetype.h new file mode 100644 index 0000000..90a5c2e --- /dev/null +++ b/platform/stb_truetype.h @@ -0,0 +1,5079 @@ +// stb_truetype.h - v1.26 - public domain +// authored from 2009-2021 by Sean Barrett / RAD Game Tools +// +// ======================================================================= +// +// NO SECURITY GUARANTEE -- DO NOT USE THIS ON UNTRUSTED FONT FILES +// +// This library does no range checking of the offsets found in the file, +// meaning an attacker can use it to read arbitrary memory. +// +// ======================================================================= +// +// This library processes TrueType files: +// parse files +// extract glyph metrics +// extract glyph shapes +// render glyphs to one-channel bitmaps with antialiasing (box filter) +// render glyphs to one-channel SDF bitmaps (signed-distance field/function) +// +// Todo: +// non-MS cmaps +// crashproof on bad data +// hinting? (no longer patented) +// cleartype-style AA? +// optimize: use simple memory allocator for intermediates +// optimize: build edge-list directly from curves +// optimize: rasterize directly from curves? +// +// ADDITIONAL CONTRIBUTORS +// +// Mikko Mononen: compound shape support, more cmap formats +// Tor Andersson: kerning, subpixel rendering +// Dougall Johnson: OpenType / Type 2 font handling +// Daniel Ribeiro Maciel: basic GPOS-based kerning +// +// Misc other: +// Ryan Gordon +// Simon Glass +// github:IntellectualKitty +// Imanol Celaya +// Daniel Ribeiro Maciel +// +// Bug/warning reports/fixes: +// "Zer" on mollyrocket Fabian "ryg" Giesen github:NiLuJe +// Cass Everitt Martins Mozeiko github:aloucks +// stoiko (Haemimont Games) Cap Petschulat github:oyvindjam +// Brian Hook Omar Cornut github:vassvik +// Walter van Niftrik Ryan Griege +// David Gow Peter LaValle +// David Given Sergey Popov +// Ivan-Assen Ivanov Giumo X. Clanjor +// Anthony Pesch Higor Euripedes +// Johan Duparc Thomas Fields +// Hou Qiming Derek Vinyard +// Rob Loach Cort Stratton +// Kenney Phillis Jr. Brian Costabile +// Ken Voskuil (kaesve) Yakov Galka +// +// VERSION HISTORY +// +// 1.26 (2021-08-28) fix broken rasterizer +// 1.25 (2021-07-11) many fixes +// 1.24 (2020-02-05) fix warning +// 1.23 (2020-02-02) query SVG data for glyphs; query whole kerning table (but only kern not GPOS) +// 1.22 (2019-08-11) minimize missing-glyph duplication; fix kerning if both 'GPOS' and 'kern' are defined +// 1.21 (2019-02-25) fix warning +// 1.20 (2019-02-07) PackFontRange skips missing codepoints; GetScaleFontVMetrics() +// 1.19 (2018-02-11) GPOS kerning, STBTT_fmod +// 1.18 (2018-01-29) add missing function +// 1.17 (2017-07-23) make more arguments const; doc fix +// 1.16 (2017-07-12) SDF support +// 1.15 (2017-03-03) make more arguments const +// 1.14 (2017-01-16) num-fonts-in-TTC function +// 1.13 (2017-01-02) support OpenType fonts, certain Apple fonts +// 1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual +// 1.11 (2016-04-02) fix unused-variable warning +// 1.10 (2016-04-02) user-defined fabs(); rare memory leak; remove duplicate typedef +// 1.09 (2016-01-16) warning fix; avoid crash on outofmem; use allocation userdata properly +// 1.08 (2015-09-13) document stbtt_Rasterize(); fixes for vertical & horizontal edges +// 1.07 (2015-08-01) allow PackFontRanges to accept arrays of sparse codepoints; +// variant PackFontRanges to pack and render in separate phases; +// fix stbtt_GetFontOFfsetForIndex (never worked for non-0 input?); +// fixed an assert() bug in the new rasterizer +// replace assert() with STBTT_assert() in new rasterizer +// +// Full history can be found at the end of this file. +// +// LICENSE +// +// See end of file for license information. +// +// USAGE +// +// Include this file in whatever places need to refer to it. In ONE C/C++ +// file, write: +// #define STB_TRUETYPE_IMPLEMENTATION +// before the #include of this file. This expands out the actual +// implementation into that C/C++ file. +// +// To make the implementation private to the file that generates the implementation, +// #define STBTT_STATIC +// +// Simple 3D API (don't ship this, but it's fine for tools and quick start) +// stbtt_BakeFontBitmap() -- bake a font to a bitmap for use as texture +// stbtt_GetBakedQuad() -- compute quad to draw for a given char +// +// Improved 3D API (more shippable): +// #include "stb_rect_pack.h" -- optional, but you really want it +// stbtt_PackBegin() +// stbtt_PackSetOversampling() -- for improved quality on small fonts +// stbtt_PackFontRanges() -- pack and renders +// stbtt_PackEnd() +// stbtt_GetPackedQuad() +// +// "Load" a font file from a memory buffer (you have to keep the buffer loaded) +// stbtt_InitFont() +// stbtt_GetFontOffsetForIndex() -- indexing for TTC font collections +// stbtt_GetNumberOfFonts() -- number of fonts for TTC font collections +// +// Render a unicode codepoint to a bitmap +// stbtt_GetCodepointBitmap() -- allocates and returns a bitmap +// stbtt_MakeCodepointBitmap() -- renders into bitmap you provide +// stbtt_GetCodepointBitmapBox() -- how big the bitmap must be +// +// Character advance/positioning +// stbtt_GetCodepointHMetrics() +// stbtt_GetFontVMetrics() +// stbtt_GetFontVMetricsOS2() +// stbtt_GetCodepointKernAdvance() +// +// Starting with version 1.06, the rasterizer was replaced with a new, +// faster and generally-more-precise rasterizer. The new rasterizer more +// accurately measures pixel coverage for anti-aliasing, except in the case +// where multiple shapes overlap, in which case it overestimates the AA pixel +// coverage. Thus, anti-aliasing of intersecting shapes may look wrong. If +// this turns out to be a problem, you can re-enable the old rasterizer with +// #define STBTT_RASTERIZER_VERSION 1 +// which will incur about a 15% speed hit. +// +// ADDITIONAL DOCUMENTATION +// +// Immediately after this block comment are a series of sample programs. +// +// After the sample programs is the "header file" section. This section +// includes documentation for each API function. +// +// Some important concepts to understand to use this library: +// +// Codepoint +// Characters are defined by unicode codepoints, e.g. 65 is +// uppercase A, 231 is lowercase c with a cedilla, 0x7e30 is +// the hiragana for "ma". +// +// Glyph +// A visual character shape (every codepoint is rendered as +// some glyph) +// +// Glyph index +// A font-specific integer ID representing a glyph +// +// Baseline +// Glyph shapes are defined relative to a baseline, which is the +// bottom of uppercase characters. Characters extend both above +// and below the baseline. +// +// Current Point +// As you draw text to the screen, you keep track of a "current point" +// which is the origin of each character. The current point's vertical +// position is the baseline. Even "baked fonts" use this model. +// +// Vertical Font Metrics +// The vertical qualities of the font, used to vertically position +// and space the characters. See docs for stbtt_GetFontVMetrics. +// +// Font Size in Pixels or Points +// The preferred interface for specifying font sizes in stb_truetype +// is to specify how tall the font's vertical extent should be in pixels. +// If that sounds good enough, skip the next paragraph. +// +// Most font APIs instead use "points", which are a common typographic +// measurement for describing font size, defined as 72 points per inch. +// stb_truetype provides a point API for compatibility. However, true +// "per inch" conventions don't make much sense on computer displays +// since different monitors have different number of pixels per +// inch. For example, Windows traditionally uses a convention that +// there are 96 pixels per inch, thus making 'inch' measurements have +// nothing to do with inches, and thus effectively defining a point to +// be 1.333 pixels. Additionally, the TrueType font data provides +// an explicit scale factor to scale a given font's glyphs to points, +// but the author has observed that this scale factor is often wrong +// for non-commercial fonts, thus making fonts scaled in points +// according to the TrueType spec incoherently sized in practice. +// +// DETAILED USAGE: +// +// Scale: +// Select how high you want the font to be, in points or pixels. +// Call ScaleForPixelHeight or ScaleForMappingEmToPixels to compute +// a scale factor SF that will be used by all other functions. +// +// Baseline: +// You need to select a y-coordinate that is the baseline of where +// your text will appear. Call GetFontBoundingBox to get the baseline-relative +// bounding box for all characters. SF*-y0 will be the distance in pixels +// that the worst-case character could extend above the baseline, so if +// you want the top edge of characters to appear at the top of the +// screen where y=0, then you would set the baseline to SF*-y0. +// +// Current point: +// Set the current point where the first character will appear. The +// first character could extend left of the current point; this is font +// dependent. You can either choose a current point that is the leftmost +// point and hope, or add some padding, or check the bounding box or +// left-side-bearing of the first character to be displayed and set +// the current point based on that. +// +// Displaying a character: +// Compute the bounding box of the character. It will contain signed values +// relative to . I.e. if it returns x0,y0,x1,y1, +// then the character should be displayed in the rectangle from +// to = 32 && *text < 128) { + stbtt_aligned_quad q; + stbtt_GetBakedQuad(cdata, 512,512, *text-32, &x,&y,&q,1);//1=opengl & d3d10+,0=d3d9 + glTexCoord2f(q.s0,q.t0); glVertex2f(q.x0,q.y0); + glTexCoord2f(q.s1,q.t0); glVertex2f(q.x1,q.y0); + glTexCoord2f(q.s1,q.t1); glVertex2f(q.x1,q.y1); + glTexCoord2f(q.s0,q.t1); glVertex2f(q.x0,q.y1); + } + ++text; + } + glEnd(); +} +#endif +// +// +////////////////////////////////////////////////////////////////////////////// +// +// Complete program (this compiles): get a single bitmap, print as ASCII art +// +#if 0 +#include +#define STB_TRUETYPE_IMPLEMENTATION // force following include to generate implementation +#include "stb_truetype.h" + +char ttf_buffer[1<<25]; + +int main(int argc, char **argv) +{ + stbtt_fontinfo font; + unsigned char *bitmap; + int w,h,i,j,c = (argc > 1 ? atoi(argv[1]) : 'a'), s = (argc > 2 ? atoi(argv[2]) : 20); + + fread(ttf_buffer, 1, 1<<25, fopen(argc > 3 ? argv[3] : "c:/windows/fonts/arialbd.ttf", "rb")); + + stbtt_InitFont(&font, ttf_buffer, stbtt_GetFontOffsetForIndex(ttf_buffer,0)); + bitmap = stbtt_GetCodepointBitmap(&font, 0,stbtt_ScaleForPixelHeight(&font, s), c, &w, &h, 0,0); + + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) + putchar(" .:ioVM@"[bitmap[j*w+i]>>5]); + putchar('\n'); + } + return 0; +} +#endif +// +// Output: +// +// .ii. +// @@@@@@. +// V@Mio@@o +// :i. V@V +// :oM@@M +// :@@@MM@M +// @@o o@M +// :@@. M@M +// @@@o@@@@ +// :M@@V:@@. +// +////////////////////////////////////////////////////////////////////////////// +// +// Complete program: print "Hello World!" banner, with bugs +// +#if 0 +char buffer[24<<20]; +unsigned char screen[20][79]; + +int main(int arg, char **argv) +{ + stbtt_fontinfo font; + int i,j,ascent,baseline,ch=0; + float scale, xpos=2; // leave a little padding in case the character extends left + char *text = "Heljo World!"; // intentionally misspelled to show 'lj' brokenness + + fread(buffer, 1, 1000000, fopen("c:/windows/fonts/arialbd.ttf", "rb")); + stbtt_InitFont(&font, buffer, 0); + + scale = stbtt_ScaleForPixelHeight(&font, 15); + stbtt_GetFontVMetrics(&font, &ascent,0,0); + baseline = (int) (ascent*scale); + + while (text[ch]) { + int advance,lsb,x0,y0,x1,y1; + float x_shift = xpos - (float) floor(xpos); + stbtt_GetCodepointHMetrics(&font, text[ch], &advance, &lsb); + stbtt_GetCodepointBitmapBoxSubpixel(&font, text[ch], scale,scale,x_shift,0, &x0,&y0,&x1,&y1); + stbtt_MakeCodepointBitmapSubpixel(&font, &screen[baseline + y0][(int) xpos + x0], x1-x0,y1-y0, 79, scale,scale,x_shift,0, text[ch]); + // note that this stomps the old data, so where character boxes overlap (e.g. 'lj') it's wrong + // because this API is really for baking character bitmaps into textures. if you want to render + // a sequence of characters, you really need to render each bitmap to a temp buffer, then + // "alpha blend" that into the working buffer + xpos += (advance * scale); + if (text[ch+1]) + xpos += scale*stbtt_GetCodepointKernAdvance(&font, text[ch],text[ch+1]); + ++ch; + } + + for (j=0; j < 20; ++j) { + for (i=0; i < 78; ++i) + putchar(" .:ioVM@"[screen[j][i]>>5]); + putchar('\n'); + } + + return 0; +} +#endif + + +////////////////////////////////////////////////////////////////////////////// +////////////////////////////////////////////////////////////////////////////// +//// +//// INTEGRATION WITH YOUR CODEBASE +//// +//// The following sections allow you to supply alternate definitions +//// of C library functions used by stb_truetype, e.g. if you don't +//// link with the C runtime library. + +#ifdef STB_TRUETYPE_IMPLEMENTATION + // #define your own (u)stbtt_int8/16/32 before including to override this + #ifndef stbtt_uint8 + typedef unsigned char stbtt_uint8; + typedef signed char stbtt_int8; + typedef unsigned short stbtt_uint16; + typedef signed short stbtt_int16; + typedef unsigned int stbtt_uint32; + typedef signed int stbtt_int32; + #endif + + typedef char stbtt__check_size32[sizeof(stbtt_int32)==4 ? 1 : -1]; + typedef char stbtt__check_size16[sizeof(stbtt_int16)==2 ? 1 : -1]; + + // e.g. #define your own STBTT_ifloor/STBTT_iceil() to avoid math.h + #ifndef STBTT_ifloor + #include + #define STBTT_ifloor(x) ((int) floor(x)) + #define STBTT_iceil(x) ((int) ceil(x)) + #endif + + #ifndef STBTT_sqrt + #include + #define STBTT_sqrt(x) sqrt(x) + #define STBTT_pow(x,y) pow(x,y) + #endif + + #ifndef STBTT_fmod + #include + #define STBTT_fmod(x,y) fmod(x,y) + #endif + + #ifndef STBTT_cos + #include + #define STBTT_cos(x) cos(x) + #define STBTT_acos(x) acos(x) + #endif + + #ifndef STBTT_fabs + #include + #define STBTT_fabs(x) fabs(x) + #endif + + // #define your own functions "STBTT_malloc" / "STBTT_free" to avoid malloc.h + #ifndef STBTT_malloc + #include + #define STBTT_malloc(x,u) ((void)(u),malloc(x)) + #define STBTT_free(x,u) ((void)(u),free(x)) + #endif + + #ifndef STBTT_assert + #include + #define STBTT_assert(x) assert(x) + #endif + + #ifndef STBTT_strlen + #include + #define STBTT_strlen(x) strlen(x) + #endif + + #ifndef STBTT_memcpy + #include + #define STBTT_memcpy memcpy + #define STBTT_memset memset + #endif +#endif + +/////////////////////////////////////////////////////////////////////////////// +/////////////////////////////////////////////////////////////////////////////// +//// +//// INTERFACE +//// +//// + +#ifndef __STB_INCLUDE_STB_TRUETYPE_H__ +#define __STB_INCLUDE_STB_TRUETYPE_H__ + +#ifdef STBTT_STATIC +#define STBTT_DEF static +#else +#define STBTT_DEF extern +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +// private structure +typedef struct +{ + unsigned char *data; + int cursor; + int size; +} stbtt__buf; + +////////////////////////////////////////////////////////////////////////////// +// +// TEXTURE BAKING API +// +// If you use this API, you only have to call two functions ever. +// + +typedef struct +{ + unsigned short x0,y0,x1,y1; // coordinates of bbox in bitmap + float xoff,yoff,xadvance; +} stbtt_bakedchar; + +STBTT_DEF int stbtt_BakeFontBitmap(const unsigned char *data, int offset, // font location (use offset=0 for plain .ttf) + float pixel_height, // height of font in pixels + unsigned char *pixels, int pw, int ph, // bitmap to be filled in + int first_char, int num_chars, // characters to bake + stbtt_bakedchar *chardata); // you allocate this, it's num_chars long +// if return is positive, the first unused row of the bitmap +// if return is negative, returns the negative of the number of characters that fit +// if return is 0, no characters fit and no rows were used +// This uses a very crappy packing. + +typedef struct +{ + float x0,y0,s0,t0; // top-left + float x1,y1,s1,t1; // bottom-right +} stbtt_aligned_quad; + +STBTT_DEF void stbtt_GetBakedQuad(const stbtt_bakedchar *chardata, int pw, int ph, // same data as above + int char_index, // character to display + float *xpos, float *ypos, // pointers to current position in screen pixel space + stbtt_aligned_quad *q, // output: quad to draw + int opengl_fillrule); // true if opengl fill rule; false if DX9 or earlier +// Call GetBakedQuad with char_index = 'character - first_char', and it +// creates the quad you need to draw and advances the current position. +// +// The coordinate system used assumes y increases downwards. +// +// Characters will extend both above and below the current position; +// see discussion of "BASELINE" above. +// +// It's inefficient; you might want to c&p it and optimize it. + +STBTT_DEF void stbtt_GetScaledFontVMetrics(const unsigned char *fontdata, int index, float size, float *ascent, float *descent, float *lineGap); +// Query the font vertical metrics without having to create a font first. + + +////////////////////////////////////////////////////////////////////////////// +// +// NEW TEXTURE BAKING API +// +// This provides options for packing multiple fonts into one atlas, not +// perfectly but better than nothing. + +typedef struct +{ + unsigned short x0,y0,x1,y1; // coordinates of bbox in bitmap + float xoff,yoff,xadvance; + float xoff2,yoff2; +} stbtt_packedchar; + +typedef struct stbtt_pack_context stbtt_pack_context; +typedef struct stbtt_fontinfo stbtt_fontinfo; +#ifndef STB_RECT_PACK_VERSION +typedef struct stbrp_rect stbrp_rect; +#endif + +STBTT_DEF int stbtt_PackBegin(stbtt_pack_context *spc, unsigned char *pixels, int width, int height, int stride_in_bytes, int padding, void *alloc_context); +// Initializes a packing context stored in the passed-in stbtt_pack_context. +// Future calls using this context will pack characters into the bitmap passed +// in here: a 1-channel bitmap that is width * height. stride_in_bytes is +// the distance from one row to the next (or 0 to mean they are packed tightly +// together). "padding" is the amount of padding to leave between each +// character (normally you want '1' for bitmaps you'll use as textures with +// bilinear filtering). +// +// Returns 0 on failure, 1 on success. + +STBTT_DEF void stbtt_PackEnd (stbtt_pack_context *spc); +// Cleans up the packing context and frees all memory. + +#define STBTT_POINT_SIZE(x) (-(x)) + +STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size, + int first_unicode_char_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range); +// Creates character bitmaps from the font_index'th font found in fontdata (use +// font_index=0 if you don't know what that is). It creates num_chars_in_range +// bitmaps for characters with unicode values starting at first_unicode_char_in_range +// and increasing. Data for how to render them is stored in chardata_for_range; +// pass these to stbtt_GetPackedQuad to get back renderable quads. +// +// font_size is the full height of the character from ascender to descender, +// as computed by stbtt_ScaleForPixelHeight. To use a point size as computed +// by stbtt_ScaleForMappingEmToPixels, wrap the point size in STBTT_POINT_SIZE() +// and pass that result as 'font_size': +// ..., 20 , ... // font max minus min y is 20 pixels tall +// ..., STBTT_POINT_SIZE(20), ... // 'M' is 20 pixels tall + +typedef struct +{ + float font_size; + int first_unicode_codepoint_in_range; // if non-zero, then the chars are continuous, and this is the first codepoint + int *array_of_unicode_codepoints; // if non-zero, then this is an array of unicode codepoints + int num_chars; + stbtt_packedchar *chardata_for_range; // output + unsigned char h_oversample, v_oversample; // don't set these, they're used internally +} stbtt_pack_range; + +STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges); +// Creates character bitmaps from multiple ranges of characters stored in +// ranges. This will usually create a better-packed bitmap than multiple +// calls to stbtt_PackFontRange. Note that you can call this multiple +// times within a single PackBegin/PackEnd. + +STBTT_DEF void stbtt_PackSetOversampling(stbtt_pack_context *spc, unsigned int h_oversample, unsigned int v_oversample); +// Oversampling a font increases the quality by allowing higher-quality subpixel +// positioning, and is especially valuable at smaller text sizes. +// +// This function sets the amount of oversampling for all following calls to +// stbtt_PackFontRange(s) or stbtt_PackFontRangesGatherRects for a given +// pack context. The default (no oversampling) is achieved by h_oversample=1 +// and v_oversample=1. The total number of pixels required is +// h_oversample*v_oversample larger than the default; for example, 2x2 +// oversampling requires 4x the storage of 1x1. For best results, render +// oversampled textures with bilinear filtering. Look at the readme in +// stb/tests/oversample for information about oversampled fonts +// +// To use with PackFontRangesGather etc., you must set it before calls +// call to PackFontRangesGatherRects. + +STBTT_DEF void stbtt_PackSetSkipMissingCodepoints(stbtt_pack_context *spc, int skip); +// If skip != 0, this tells stb_truetype to skip any codepoints for which +// there is no corresponding glyph. If skip=0, which is the default, then +// codepoints without a glyph recived the font's "missing character" glyph, +// typically an empty box by convention. + +STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int ph, // same data as above + int char_index, // character to display + float *xpos, float *ypos, // pointers to current position in screen pixel space + stbtt_aligned_quad *q, // output: quad to draw + int align_to_integer); + +STBTT_DEF int stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects); +STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect *rects, int num_rects); +STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects); +// Calling these functions in sequence is roughly equivalent to calling +// stbtt_PackFontRanges(). If you more control over the packing of multiple +// fonts, or if you want to pack custom data into a font texture, take a look +// at the source to of stbtt_PackFontRanges() and create a custom version +// using these functions, e.g. call GatherRects multiple times, +// building up a single array of rects, then call PackRects once, +// then call RenderIntoRects repeatedly. This may result in a +// better packing than calling PackFontRanges multiple times +// (or it may not). + +// this is an opaque structure that you shouldn't mess with which holds +// all the context needed from PackBegin to PackEnd. +struct stbtt_pack_context { + void *user_allocator_context; + void *pack_info; + int width; + int height; + int stride_in_bytes; + int padding; + int skip_missing; + unsigned int h_oversample, v_oversample; + unsigned char *pixels; + void *nodes; +}; + +////////////////////////////////////////////////////////////////////////////// +// +// FONT LOADING +// +// + +STBTT_DEF int stbtt_GetNumberOfFonts(const unsigned char *data); +// This function will determine the number of fonts in a font file. TrueType +// collection (.ttc) files may contain multiple fonts, while TrueType font +// (.ttf) files only contain one font. The number of fonts can be used for +// indexing with the previous function where the index is between zero and one +// less than the total fonts. If an error occurs, -1 is returned. + +STBTT_DEF int stbtt_GetFontOffsetForIndex(const unsigned char *data, int index); +// Each .ttf/.ttc file may have more than one font. Each font has a sequential +// index number starting from 0. Call this function to get the font offset for +// a given index; it returns -1 if the index is out of range. A regular .ttf +// file will only define one font and it always be at offset 0, so it will +// return '0' for index 0, and -1 for all other indices. + +// The following structure is defined publicly so you can declare one on +// the stack or as a global or etc, but you should treat it as opaque. +struct stbtt_fontinfo +{ + void * userdata; + unsigned char * data; // pointer to .ttf file + int fontstart; // offset of start of font + + int numGlyphs; // number of glyphs, needed for range checking + + int loca,head,glyf,hhea,hmtx,kern,gpos,svg; // table locations as offset from start of .ttf + int index_map; // a cmap mapping for our chosen character encoding + int indexToLocFormat; // format needed to map from glyph index to glyph + + stbtt__buf cff; // cff font data + stbtt__buf charstrings; // the charstring index + stbtt__buf gsubrs; // global charstring subroutines index + stbtt__buf subrs; // private charstring subroutines index + stbtt__buf fontdicts; // array of font dicts + stbtt__buf fdselect; // map from glyph to fontdict +}; + +STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data, int offset); +// Given an offset into the file that defines a font, this function builds +// the necessary cached info for the rest of the system. You must allocate +// the stbtt_fontinfo yourself, and stbtt_InitFont will fill it out. You don't +// need to do anything special to free it, because the contents are pure +// value data with no additional data structures. Returns 0 on failure. + + +////////////////////////////////////////////////////////////////////////////// +// +// CHARACTER TO GLYPH-INDEX CONVERSIOn + +STBTT_DEF int stbtt_FindGlyphIndex(const stbtt_fontinfo *info, int unicode_codepoint); +// If you're going to perform multiple operations on the same character +// and you want a speed-up, call this function with the character you're +// going to process, then use glyph-based functions instead of the +// codepoint-based functions. +// Returns 0 if the character codepoint is not defined in the font. + + +////////////////////////////////////////////////////////////////////////////// +// +// CHARACTER PROPERTIES +// + +STBTT_DEF float stbtt_ScaleForPixelHeight(const stbtt_fontinfo *info, float pixels); +// computes a scale factor to produce a font whose "height" is 'pixels' tall. +// Height is measured as the distance from the highest ascender to the lowest +// descender; in other words, it's equivalent to calling stbtt_GetFontVMetrics +// and computing: +// scale = pixels / (ascent - descent) +// so if you prefer to measure height by the ascent only, use a similar calculation. + +STBTT_DEF float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels); +// computes a scale factor to produce a font whose EM size is mapped to +// 'pixels' tall. This is probably what traditional APIs compute, but +// I'm not positive. + +STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, int *descent, int *lineGap); +// ascent is the coordinate above the baseline the font extends; descent +// is the coordinate below the baseline the font extends (i.e. it is typically negative) +// lineGap is the spacing between one row's descent and the next row's ascent... +// so you should advance the vertical position by "*ascent - *descent + *lineGap" +// these are expressed in unscaled coordinates, so you must multiply by +// the scale factor for a given size + +STBTT_DEF int stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap); +// analogous to GetFontVMetrics, but returns the "typographic" values from the OS/2 +// table (specific to MS/Windows TTF files). +// +// Returns 1 on success (table present), 0 on failure. + +STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1); +// the bounding box around all possible characters + +STBTT_DEF void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing); +// leftSideBearing is the offset from the current horizontal position to the left edge of the character +// advanceWidth is the offset from the current horizontal position to the next horizontal position +// these are expressed in unscaled coordinates + +STBTT_DEF int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2); +// an additional amount to add to the 'advance' value between ch1 and ch2 + +STBTT_DEF int stbtt_GetCodepointBox(const stbtt_fontinfo *info, int codepoint, int *x0, int *y0, int *x1, int *y1); +// Gets the bounding box of the visible part of the glyph, in unscaled coordinates + +STBTT_DEF void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing); +STBTT_DEF int stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2); +STBTT_DEF int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1); +// as above, but takes one or more glyph indices for greater efficiency + +typedef struct stbtt_kerningentry +{ + int glyph1; // use stbtt_FindGlyphIndex + int glyph2; + int advance; +} stbtt_kerningentry; + +STBTT_DEF int stbtt_GetKerningTableLength(const stbtt_fontinfo *info); +STBTT_DEF int stbtt_GetKerningTable(const stbtt_fontinfo *info, stbtt_kerningentry* table, int table_length); +// Retrieves a complete list of all of the kerning pairs provided by the font +// stbtt_GetKerningTable never writes more than table_length entries and returns how many entries it did write. +// The table will be sorted by (a.glyph1 == b.glyph1)?(a.glyph2 < b.glyph2):(a.glyph1 < b.glyph1) + +////////////////////////////////////////////////////////////////////////////// +// +// GLYPH SHAPES (you probably don't need these, but they have to go before +// the bitmaps for C declaration-order reasons) +// + +#ifndef STBTT_vmove // you can predefine these to use different values (but why?) + enum { + STBTT_vmove=1, + STBTT_vline, + STBTT_vcurve, + STBTT_vcubic + }; +#endif + +#ifndef stbtt_vertex // you can predefine this to use different values + // (we share this with other code at RAD) + #define stbtt_vertex_type short // can't use stbtt_int16 because that's not visible in the header file + typedef struct + { + stbtt_vertex_type x,y,cx,cy,cx1,cy1; + unsigned char type,padding; + } stbtt_vertex; +#endif + +STBTT_DEF int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index); +// returns non-zero if nothing is drawn for this glyph + +STBTT_DEF int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices); +STBTT_DEF int stbtt_GetGlyphShape(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **vertices); +// returns # of vertices and fills *vertices with the pointer to them +// these are expressed in "unscaled" coordinates +// +// The shape is a series of contours. Each one starts with +// a STBTT_moveto, then consists of a series of mixed +// STBTT_lineto and STBTT_curveto segments. A lineto +// draws a line from previous endpoint to its x,y; a curveto +// draws a quadratic bezier from previous endpoint to +// its x,y, using cx,cy as the bezier control point. + +STBTT_DEF void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *vertices); +// frees the data allocated above + +STBTT_DEF unsigned char *stbtt_FindSVGDoc(const stbtt_fontinfo *info, int gl); +STBTT_DEF int stbtt_GetCodepointSVG(const stbtt_fontinfo *info, int unicode_codepoint, const char **svg); +STBTT_DEF int stbtt_GetGlyphSVG(const stbtt_fontinfo *info, int gl, const char **svg); +// fills svg with the character's SVG data. +// returns data size or 0 if SVG not found. + +////////////////////////////////////////////////////////////////////////////// +// +// BITMAP RENDERING +// + +STBTT_DEF void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata); +// frees the bitmap allocated below + +STBTT_DEF unsigned char *stbtt_GetCodepointBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int codepoint, int *width, int *height, int *xoff, int *yoff); +// allocates a large-enough single-channel 8bpp bitmap and renders the +// specified character/glyph at the specified scale into it, with +// antialiasing. 0 is no coverage (transparent), 255 is fully covered (opaque). +// *width & *height are filled out with the width & height of the bitmap, +// which is stored left-to-right, top-to-bottom. +// +// xoff/yoff are the offset it pixel space from the glyph origin to the top-left of the bitmap + +STBTT_DEF unsigned char *stbtt_GetCodepointBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint, int *width, int *height, int *xoff, int *yoff); +// the same as stbtt_GetCodepoitnBitmap, but you can specify a subpixel +// shift for the character + +STBTT_DEF void stbtt_MakeCodepointBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int codepoint); +// the same as stbtt_GetCodepointBitmap, but you pass in storage for the bitmap +// in the form of 'output', with row spacing of 'out_stride' bytes. the bitmap +// is clipped to out_w/out_h bytes. Call stbtt_GetCodepointBitmapBox to get the +// width and height and positioning info for it first. + +STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint); +// same as stbtt_MakeCodepointBitmap, but you can specify a subpixel +// shift for the character + +STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint); +// same as stbtt_MakeCodepointBitmapSubpixel, but prefiltering +// is performed (see stbtt_PackSetOversampling) + +STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1); +// get the bbox of the bitmap centered around the glyph origin; so the +// bitmap width is ix1-ix0, height is iy1-iy0, and location to place +// the bitmap top left is (leftSideBearing*scale,iy0). +// (Note that the bitmap uses y-increases-down, but the shape uses +// y-increases-up, so CodepointBitmapBox and CodepointBox are inverted.) + +STBTT_DEF void stbtt_GetCodepointBitmapBoxSubpixel(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1); +// same as stbtt_GetCodepointBitmapBox, but you can specify a subpixel +// shift for the character + +// the following functions are equivalent to the above functions, but operate +// on glyph indices instead of Unicode codepoints (for efficiency) +STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff); +STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff); +STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph); +STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph); +STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int glyph); +STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1); +STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1); + + +// @TODO: don't expose this structure +typedef struct +{ + int w,h,stride; + unsigned char *pixels; +} stbtt__bitmap; + +// rasterize a shape with quadratic beziers into a bitmap +STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result, // 1-channel bitmap to draw into + float flatness_in_pixels, // allowable error of curve in pixels + stbtt_vertex *vertices, // array of vertices defining shape + int num_verts, // number of vertices in above array + float scale_x, float scale_y, // scale applied to input vertices + float shift_x, float shift_y, // translation applied to input vertices + int x_off, int y_off, // another translation applied to input + int invert, // if non-zero, vertically flip shape + void *userdata); // context for to STBTT_MALLOC + +////////////////////////////////////////////////////////////////////////////// +// +// Signed Distance Function (or Field) rendering + +STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata); +// frees the SDF bitmap allocated below + +STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff); +STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff); +// These functions compute a discretized SDF field for a single character, suitable for storing +// in a single-channel texture, sampling with bilinear filtering, and testing against +// larger than some threshold to produce scalable fonts. +// info -- the font +// scale -- controls the size of the resulting SDF bitmap, same as it would be creating a regular bitmap +// glyph/codepoint -- the character to generate the SDF for +// padding -- extra "pixels" around the character which are filled with the distance to the character (not 0), +// which allows effects like bit outlines +// onedge_value -- value 0-255 to test the SDF against to reconstruct the character (i.e. the isocontour of the character) +// pixel_dist_scale -- what value the SDF should increase by when moving one SDF "pixel" away from the edge (on the 0..255 scale) +// if positive, > onedge_value is inside; if negative, < onedge_value is inside +// width,height -- output height & width of the SDF bitmap (including padding) +// xoff,yoff -- output origin of the character +// return value -- a 2D array of bytes 0..255, width*height in size +// +// pixel_dist_scale & onedge_value are a scale & bias that allows you to make +// optimal use of the limited 0..255 for your application, trading off precision +// and special effects. SDF values outside the range 0..255 are clamped to 0..255. +// +// Example: +// scale = stbtt_ScaleForPixelHeight(22) +// padding = 5 +// onedge_value = 180 +// pixel_dist_scale = 180/5.0 = 36.0 +// +// This will create an SDF bitmap in which the character is about 22 pixels +// high but the whole bitmap is about 22+5+5=32 pixels high. To produce a filled +// shape, sample the SDF at each pixel and fill the pixel if the SDF value +// is greater than or equal to 180/255. (You'll actually want to antialias, +// which is beyond the scope of this example.) Additionally, you can compute +// offset outlines (e.g. to stroke the character border inside & outside, +// or only outside). For example, to fill outside the character up to 3 SDF +// pixels, you would compare against (180-36.0*3)/255 = 72/255. The above +// choice of variables maps a range from 5 pixels outside the shape to +// 2 pixels inside the shape to 0..255; this is intended primarily for apply +// outside effects only (the interior range is needed to allow proper +// antialiasing of the font at *smaller* sizes) +// +// The function computes the SDF analytically at each SDF pixel, not by e.g. +// building a higher-res bitmap and approximating it. In theory the quality +// should be as high as possible for an SDF of this size & representation, but +// unclear if this is true in practice (perhaps building a higher-res bitmap +// and computing from that can allow drop-out prevention). +// +// The algorithm has not been optimized at all, so expect it to be slow +// if computing lots of characters or very large sizes. + + + +////////////////////////////////////////////////////////////////////////////// +// +// Finding the right font... +// +// You should really just solve this offline, keep your own tables +// of what font is what, and don't try to get it out of the .ttf file. +// That's because getting it out of the .ttf file is really hard, because +// the names in the file can appear in many possible encodings, in many +// possible languages, and e.g. if you need a case-insensitive comparison, +// the details of that depend on the encoding & language in a complex way +// (actually underspecified in truetype, but also gigantic). +// +// But you can use the provided functions in two possible ways: +// stbtt_FindMatchingFont() will use *case-sensitive* comparisons on +// unicode-encoded names to try to find the font you want; +// you can run this before calling stbtt_InitFont() +// +// stbtt_GetFontNameString() lets you get any of the various strings +// from the file yourself and do your own comparisons on them. +// You have to have called stbtt_InitFont() first. + + +STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *fontdata, const char *name, int flags); +// returns the offset (not index) of the font that matches, or -1 if none +// if you use STBTT_MACSTYLE_DONTCARE, use a font name like "Arial Bold". +// if you use any other flag, use a font name like "Arial"; this checks +// the 'macStyle' header field; i don't know if fonts set this consistently +#define STBTT_MACSTYLE_DONTCARE 0 +#define STBTT_MACSTYLE_BOLD 1 +#define STBTT_MACSTYLE_ITALIC 2 +#define STBTT_MACSTYLE_UNDERSCORE 4 +#define STBTT_MACSTYLE_NONE 8 // <= not same as 0, this makes us check the bitfield is 0 + +STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2); +// returns 1/0 whether the first string interpreted as utf8 is identical to +// the second string interpreted as big-endian utf16... useful for strings from next func + +STBTT_DEF const char *stbtt_GetFontNameString(const stbtt_fontinfo *font, int *length, int platformID, int encodingID, int languageID, int nameID); +// returns the string (which may be big-endian double byte, e.g. for unicode) +// and puts the length in bytes in *length. +// +// some of the values for the IDs are below; for more see the truetype spec: +// http://developer.apple.com/textfonts/TTRefMan/RM06/Chap6name.html +// http://www.microsoft.com/typography/otspec/name.htm + +enum { // platformID + STBTT_PLATFORM_ID_UNICODE =0, + STBTT_PLATFORM_ID_MAC =1, + STBTT_PLATFORM_ID_ISO =2, + STBTT_PLATFORM_ID_MICROSOFT =3 +}; + +enum { // encodingID for STBTT_PLATFORM_ID_UNICODE + STBTT_UNICODE_EID_UNICODE_1_0 =0, + STBTT_UNICODE_EID_UNICODE_1_1 =1, + STBTT_UNICODE_EID_ISO_10646 =2, + STBTT_UNICODE_EID_UNICODE_2_0_BMP=3, + STBTT_UNICODE_EID_UNICODE_2_0_FULL=4 +}; + +enum { // encodingID for STBTT_PLATFORM_ID_MICROSOFT + STBTT_MS_EID_SYMBOL =0, + STBTT_MS_EID_UNICODE_BMP =1, + STBTT_MS_EID_SHIFTJIS =2, + STBTT_MS_EID_UNICODE_FULL =10 +}; + +enum { // encodingID for STBTT_PLATFORM_ID_MAC; same as Script Manager codes + STBTT_MAC_EID_ROMAN =0, STBTT_MAC_EID_ARABIC =4, + STBTT_MAC_EID_JAPANESE =1, STBTT_MAC_EID_HEBREW =5, + STBTT_MAC_EID_CHINESE_TRAD =2, STBTT_MAC_EID_GREEK =6, + STBTT_MAC_EID_KOREAN =3, STBTT_MAC_EID_RUSSIAN =7 +}; + +enum { // languageID for STBTT_PLATFORM_ID_MICROSOFT; same as LCID... + // problematic because there are e.g. 16 english LCIDs and 16 arabic LCIDs + STBTT_MS_LANG_ENGLISH =0x0409, STBTT_MS_LANG_ITALIAN =0x0410, + STBTT_MS_LANG_CHINESE =0x0804, STBTT_MS_LANG_JAPANESE =0x0411, + STBTT_MS_LANG_DUTCH =0x0413, STBTT_MS_LANG_KOREAN =0x0412, + STBTT_MS_LANG_FRENCH =0x040c, STBTT_MS_LANG_RUSSIAN =0x0419, + STBTT_MS_LANG_GERMAN =0x0407, STBTT_MS_LANG_SPANISH =0x0409, + STBTT_MS_LANG_HEBREW =0x040d, STBTT_MS_LANG_SWEDISH =0x041D +}; + +enum { // languageID for STBTT_PLATFORM_ID_MAC + STBTT_MAC_LANG_ENGLISH =0 , STBTT_MAC_LANG_JAPANESE =11, + STBTT_MAC_LANG_ARABIC =12, STBTT_MAC_LANG_KOREAN =23, + STBTT_MAC_LANG_DUTCH =4 , STBTT_MAC_LANG_RUSSIAN =32, + STBTT_MAC_LANG_FRENCH =1 , STBTT_MAC_LANG_SPANISH =6 , + STBTT_MAC_LANG_GERMAN =2 , STBTT_MAC_LANG_SWEDISH =5 , + STBTT_MAC_LANG_HEBREW =10, STBTT_MAC_LANG_CHINESE_SIMPLIFIED =33, + STBTT_MAC_LANG_ITALIAN =3 , STBTT_MAC_LANG_CHINESE_TRAD =19 +}; + +#ifdef __cplusplus +} +#endif + +#endif // __STB_INCLUDE_STB_TRUETYPE_H__ + +/////////////////////////////////////////////////////////////////////////////// +/////////////////////////////////////////////////////////////////////////////// +//// +//// IMPLEMENTATION +//// +//// + +#ifdef STB_TRUETYPE_IMPLEMENTATION + +#ifndef STBTT_MAX_OVERSAMPLE +#define STBTT_MAX_OVERSAMPLE 8 +#endif + +#if STBTT_MAX_OVERSAMPLE > 255 +#error "STBTT_MAX_OVERSAMPLE cannot be > 255" +#endif + +typedef int stbtt__test_oversample_pow2[(STBTT_MAX_OVERSAMPLE & (STBTT_MAX_OVERSAMPLE-1)) == 0 ? 1 : -1]; + +#ifndef STBTT_RASTERIZER_VERSION +#define STBTT_RASTERIZER_VERSION 2 +#endif + +#ifdef _MSC_VER +#define STBTT__NOTUSED(v) (void)(v) +#else +#define STBTT__NOTUSED(v) (void)sizeof(v) +#endif + +////////////////////////////////////////////////////////////////////////// +// +// stbtt__buf helpers to parse data from file +// + +static stbtt_uint8 stbtt__buf_get8(stbtt__buf *b) +{ + if (b->cursor >= b->size) + return 0; + return b->data[b->cursor++]; +} + +static stbtt_uint8 stbtt__buf_peek8(stbtt__buf *b) +{ + if (b->cursor >= b->size) + return 0; + return b->data[b->cursor]; +} + +static void stbtt__buf_seek(stbtt__buf *b, int o) +{ + STBTT_assert(!(o > b->size || o < 0)); + b->cursor = (o > b->size || o < 0) ? b->size : o; +} + +static void stbtt__buf_skip(stbtt__buf *b, int o) +{ + stbtt__buf_seek(b, b->cursor + o); +} + +static stbtt_uint32 stbtt__buf_get(stbtt__buf *b, int n) +{ + stbtt_uint32 v = 0; + int i; + STBTT_assert(n >= 1 && n <= 4); + for (i = 0; i < n; i++) + v = (v << 8) | stbtt__buf_get8(b); + return v; +} + +static stbtt__buf stbtt__new_buf(const void *p, size_t size) +{ + stbtt__buf r; + STBTT_assert(size < 0x40000000); + r.data = (stbtt_uint8*) p; + r.size = (int) size; + r.cursor = 0; + return r; +} + +#define stbtt__buf_get16(b) stbtt__buf_get((b), 2) +#define stbtt__buf_get32(b) stbtt__buf_get((b), 4) + +static stbtt__buf stbtt__buf_range(const stbtt__buf *b, int o, int s) +{ + stbtt__buf r = stbtt__new_buf(NULL, 0); + if (o < 0 || s < 0 || o > b->size || s > b->size - o) return r; + r.data = b->data + o; + r.size = s; + return r; +} + +static stbtt__buf stbtt__cff_get_index(stbtt__buf *b) +{ + int count, start, offsize; + start = b->cursor; + count = stbtt__buf_get16(b); + if (count) { + offsize = stbtt__buf_get8(b); + STBTT_assert(offsize >= 1 && offsize <= 4); + stbtt__buf_skip(b, offsize * count); + stbtt__buf_skip(b, stbtt__buf_get(b, offsize) - 1); + } + return stbtt__buf_range(b, start, b->cursor - start); +} + +static stbtt_uint32 stbtt__cff_int(stbtt__buf *b) +{ + int b0 = stbtt__buf_get8(b); + if (b0 >= 32 && b0 <= 246) return b0 - 139; + else if (b0 >= 247 && b0 <= 250) return (b0 - 247)*256 + stbtt__buf_get8(b) + 108; + else if (b0 >= 251 && b0 <= 254) return -(b0 - 251)*256 - stbtt__buf_get8(b) - 108; + else if (b0 == 28) return stbtt__buf_get16(b); + else if (b0 == 29) return stbtt__buf_get32(b); + STBTT_assert(0); + return 0; +} + +static void stbtt__cff_skip_operand(stbtt__buf *b) { + int v, b0 = stbtt__buf_peek8(b); + STBTT_assert(b0 >= 28); + if (b0 == 30) { + stbtt__buf_skip(b, 1); + while (b->cursor < b->size) { + v = stbtt__buf_get8(b); + if ((v & 0xF) == 0xF || (v >> 4) == 0xF) + break; + } + } else { + stbtt__cff_int(b); + } +} + +static stbtt__buf stbtt__dict_get(stbtt__buf *b, int key) +{ + stbtt__buf_seek(b, 0); + while (b->cursor < b->size) { + int start = b->cursor, end, op; + while (stbtt__buf_peek8(b) >= 28) + stbtt__cff_skip_operand(b); + end = b->cursor; + op = stbtt__buf_get8(b); + if (op == 12) op = stbtt__buf_get8(b) | 0x100; + if (op == key) return stbtt__buf_range(b, start, end-start); + } + return stbtt__buf_range(b, 0, 0); +} + +static void stbtt__dict_get_ints(stbtt__buf *b, int key, int outcount, stbtt_uint32 *out) +{ + int i; + stbtt__buf operands = stbtt__dict_get(b, key); + for (i = 0; i < outcount && operands.cursor < operands.size; i++) + out[i] = stbtt__cff_int(&operands); +} + +static int stbtt__cff_index_count(stbtt__buf *b) +{ + stbtt__buf_seek(b, 0); + return stbtt__buf_get16(b); +} + +static stbtt__buf stbtt__cff_index_get(stbtt__buf b, int i) +{ + int count, offsize, start, end; + stbtt__buf_seek(&b, 0); + count = stbtt__buf_get16(&b); + offsize = stbtt__buf_get8(&b); + STBTT_assert(i >= 0 && i < count); + STBTT_assert(offsize >= 1 && offsize <= 4); + stbtt__buf_skip(&b, i*offsize); + start = stbtt__buf_get(&b, offsize); + end = stbtt__buf_get(&b, offsize); + return stbtt__buf_range(&b, 2+(count+1)*offsize+start, end - start); +} + +////////////////////////////////////////////////////////////////////////// +// +// accessors to parse data from file +// + +// on platforms that don't allow misaligned reads, if we want to allow +// truetype fonts that aren't padded to alignment, define ALLOW_UNALIGNED_TRUETYPE + +#define ttBYTE(p) (* (stbtt_uint8 *) (p)) +#define ttCHAR(p) (* (stbtt_int8 *) (p)) +#define ttFixed(p) ttLONG(p) + +static stbtt_uint16 ttUSHORT(stbtt_uint8 *p) { return p[0]*256 + p[1]; } +static stbtt_int16 ttSHORT(stbtt_uint8 *p) { return p[0]*256 + p[1]; } +static stbtt_uint32 ttULONG(stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } +static stbtt_int32 ttLONG(stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } + +#define stbtt_tag4(p,c0,c1,c2,c3) ((p)[0] == (c0) && (p)[1] == (c1) && (p)[2] == (c2) && (p)[3] == (c3)) +#define stbtt_tag(p,str) stbtt_tag4(p,str[0],str[1],str[2],str[3]) + +static int stbtt__isfont(stbtt_uint8 *font) +{ + // check the version number + if (stbtt_tag4(font, '1',0,0,0)) return 1; // TrueType 1 + if (stbtt_tag(font, "typ1")) return 1; // TrueType with type 1 font -- we don't support this! + if (stbtt_tag(font, "OTTO")) return 1; // OpenType with CFF + if (stbtt_tag4(font, 0,1,0,0)) return 1; // OpenType 1.0 + if (stbtt_tag(font, "true")) return 1; // Apple specification for TrueType fonts + return 0; +} + +// @OPTIMIZE: binary search +static stbtt_uint32 stbtt__find_table(stbtt_uint8 *data, stbtt_uint32 fontstart, const char *tag) +{ + stbtt_int32 num_tables = ttUSHORT(data+fontstart+4); + stbtt_uint32 tabledir = fontstart + 12; + stbtt_int32 i; + for (i=0; i < num_tables; ++i) { + stbtt_uint32 loc = tabledir + 16*i; + if (stbtt_tag(data+loc+0, tag)) + return ttULONG(data+loc+8); + } + return 0; +} + +static int stbtt_GetFontOffsetForIndex_internal(unsigned char *font_collection, int index) +{ + // if it's just a font, there's only one valid index + if (stbtt__isfont(font_collection)) + return index == 0 ? 0 : -1; + + // check if it's a TTC + if (stbtt_tag(font_collection, "ttcf")) { + // version 1? + if (ttULONG(font_collection+4) == 0x00010000 || ttULONG(font_collection+4) == 0x00020000) { + stbtt_int32 n = ttLONG(font_collection+8); + if (index >= n) + return -1; + return ttULONG(font_collection+12+index*4); + } + } + return -1; +} + +static int stbtt_GetNumberOfFonts_internal(unsigned char *font_collection) +{ + // if it's just a font, there's only one valid font + if (stbtt__isfont(font_collection)) + return 1; + + // check if it's a TTC + if (stbtt_tag(font_collection, "ttcf")) { + // version 1? + if (ttULONG(font_collection+4) == 0x00010000 || ttULONG(font_collection+4) == 0x00020000) { + return ttLONG(font_collection+8); + } + } + return 0; +} + +static stbtt__buf stbtt__get_subrs(stbtt__buf cff, stbtt__buf fontdict) +{ + stbtt_uint32 subrsoff = 0, private_loc[2] = { 0, 0 }; + stbtt__buf pdict; + stbtt__dict_get_ints(&fontdict, 18, 2, private_loc); + if (!private_loc[1] || !private_loc[0]) return stbtt__new_buf(NULL, 0); + pdict = stbtt__buf_range(&cff, private_loc[1], private_loc[0]); + stbtt__dict_get_ints(&pdict, 19, 1, &subrsoff); + if (!subrsoff) return stbtt__new_buf(NULL, 0); + stbtt__buf_seek(&cff, private_loc[1]+subrsoff); + return stbtt__cff_get_index(&cff); +} + +// since most people won't use this, find this table the first time it's needed +static int stbtt__get_svg(stbtt_fontinfo *info) +{ + stbtt_uint32 t; + if (info->svg < 0) { + t = stbtt__find_table(info->data, info->fontstart, "SVG "); + if (t) { + stbtt_uint32 offset = ttULONG(info->data + t + 2); + info->svg = t + offset; + } else { + info->svg = 0; + } + } + return info->svg; +} + +static int stbtt_InitFont_internal(stbtt_fontinfo *info, unsigned char *data, int fontstart) +{ + stbtt_uint32 cmap, t; + stbtt_int32 i,numTables; + + info->data = data; + info->fontstart = fontstart; + info->cff = stbtt__new_buf(NULL, 0); + + cmap = stbtt__find_table(data, fontstart, "cmap"); // required + info->loca = stbtt__find_table(data, fontstart, "loca"); // required + info->head = stbtt__find_table(data, fontstart, "head"); // required + info->glyf = stbtt__find_table(data, fontstart, "glyf"); // required + info->hhea = stbtt__find_table(data, fontstart, "hhea"); // required + info->hmtx = stbtt__find_table(data, fontstart, "hmtx"); // required + info->kern = stbtt__find_table(data, fontstart, "kern"); // not required + info->gpos = stbtt__find_table(data, fontstart, "GPOS"); // not required + + if (!cmap || !info->head || !info->hhea || !info->hmtx) + return 0; + if (info->glyf) { + // required for truetype + if (!info->loca) return 0; + } else { + // initialization for CFF / Type2 fonts (OTF) + stbtt__buf b, topdict, topdictidx; + stbtt_uint32 cstype = 2, charstrings = 0, fdarrayoff = 0, fdselectoff = 0; + stbtt_uint32 cff; + + cff = stbtt__find_table(data, fontstart, "CFF "); + if (!cff) return 0; + + info->fontdicts = stbtt__new_buf(NULL, 0); + info->fdselect = stbtt__new_buf(NULL, 0); + + // @TODO this should use size from table (not 512MB) + info->cff = stbtt__new_buf(data+cff, 512*1024*1024); + b = info->cff; + + // read the header + stbtt__buf_skip(&b, 2); + stbtt__buf_seek(&b, stbtt__buf_get8(&b)); // hdrsize + + // @TODO the name INDEX could list multiple fonts, + // but we just use the first one. + stbtt__cff_get_index(&b); // name INDEX + topdictidx = stbtt__cff_get_index(&b); + topdict = stbtt__cff_index_get(topdictidx, 0); + stbtt__cff_get_index(&b); // string INDEX + info->gsubrs = stbtt__cff_get_index(&b); + + stbtt__dict_get_ints(&topdict, 17, 1, &charstrings); + stbtt__dict_get_ints(&topdict, 0x100 | 6, 1, &cstype); + stbtt__dict_get_ints(&topdict, 0x100 | 36, 1, &fdarrayoff); + stbtt__dict_get_ints(&topdict, 0x100 | 37, 1, &fdselectoff); + info->subrs = stbtt__get_subrs(b, topdict); + + // we only support Type 2 charstrings + if (cstype != 2) return 0; + if (charstrings == 0) return 0; + + if (fdarrayoff) { + // looks like a CID font + if (!fdselectoff) return 0; + stbtt__buf_seek(&b, fdarrayoff); + info->fontdicts = stbtt__cff_get_index(&b); + info->fdselect = stbtt__buf_range(&b, fdselectoff, b.size-fdselectoff); + } + + stbtt__buf_seek(&b, charstrings); + info->charstrings = stbtt__cff_get_index(&b); + } + + t = stbtt__find_table(data, fontstart, "maxp"); + if (t) + info->numGlyphs = ttUSHORT(data+t+4); + else + info->numGlyphs = 0xffff; + + info->svg = -1; + + // find a cmap encoding table we understand *now* to avoid searching + // later. (todo: could make this installable) + // the same regardless of glyph. + numTables = ttUSHORT(data + cmap + 2); + info->index_map = 0; + for (i=0; i < numTables; ++i) { + stbtt_uint32 encoding_record = cmap + 4 + 8 * i; + // find an encoding we understand: + switch(ttUSHORT(data+encoding_record)) { + case STBTT_PLATFORM_ID_MICROSOFT: + switch (ttUSHORT(data+encoding_record+2)) { + case STBTT_MS_EID_UNICODE_BMP: + case STBTT_MS_EID_UNICODE_FULL: + // MS/Unicode + info->index_map = cmap + ttULONG(data+encoding_record+4); + break; + } + break; + case STBTT_PLATFORM_ID_UNICODE: + // Mac/iOS has these + // all the encodingIDs are unicode, so we don't bother to check it + info->index_map = cmap + ttULONG(data+encoding_record+4); + break; + } + } + if (info->index_map == 0) + return 0; + + info->indexToLocFormat = ttUSHORT(data+info->head + 50); + return 1; +} + +STBTT_DEF int stbtt_FindGlyphIndex(const stbtt_fontinfo *info, int unicode_codepoint) +{ + stbtt_uint8 *data = info->data; + stbtt_uint32 index_map = info->index_map; + + stbtt_uint16 format = ttUSHORT(data + index_map + 0); + if (format == 0) { // apple byte encoding + stbtt_int32 bytes = ttUSHORT(data + index_map + 2); + if (unicode_codepoint < bytes-6) + return ttBYTE(data + index_map + 6 + unicode_codepoint); + return 0; + } else if (format == 6) { + stbtt_uint32 first = ttUSHORT(data + index_map + 6); + stbtt_uint32 count = ttUSHORT(data + index_map + 8); + if ((stbtt_uint32) unicode_codepoint >= first && (stbtt_uint32) unicode_codepoint < first+count) + return ttUSHORT(data + index_map + 10 + (unicode_codepoint - first)*2); + return 0; + } else if (format == 2) { + STBTT_assert(0); // @TODO: high-byte mapping for japanese/chinese/korean + return 0; + } else if (format == 4) { // standard mapping for windows fonts: binary search collection of ranges + stbtt_uint16 segcount = ttUSHORT(data+index_map+6) >> 1; + stbtt_uint16 searchRange = ttUSHORT(data+index_map+8) >> 1; + stbtt_uint16 entrySelector = ttUSHORT(data+index_map+10); + stbtt_uint16 rangeShift = ttUSHORT(data+index_map+12) >> 1; + + // do a binary search of the segments + stbtt_uint32 endCount = index_map + 14; + stbtt_uint32 search = endCount; + + if (unicode_codepoint > 0xffff) + return 0; + + // they lie from endCount .. endCount + segCount + // but searchRange is the nearest power of two, so... + if (unicode_codepoint >= ttUSHORT(data + search + rangeShift*2)) + search += rangeShift*2; + + // now decrement to bias correctly to find smallest + search -= 2; + while (entrySelector) { + stbtt_uint16 end; + searchRange >>= 1; + end = ttUSHORT(data + search + searchRange*2); + if (unicode_codepoint > end) + search += searchRange*2; + --entrySelector; + } + search += 2; + + { + stbtt_uint16 offset, start, last; + stbtt_uint16 item = (stbtt_uint16) ((search - endCount) >> 1); + + start = ttUSHORT(data + index_map + 14 + segcount*2 + 2 + 2*item); + last = ttUSHORT(data + endCount + 2*item); + if (unicode_codepoint < start || unicode_codepoint > last) + return 0; + + offset = ttUSHORT(data + index_map + 14 + segcount*6 + 2 + 2*item); + if (offset == 0) + return (stbtt_uint16) (unicode_codepoint + ttSHORT(data + index_map + 14 + segcount*4 + 2 + 2*item)); + + return ttUSHORT(data + offset + (unicode_codepoint-start)*2 + index_map + 14 + segcount*6 + 2 + 2*item); + } + } else if (format == 12 || format == 13) { + stbtt_uint32 ngroups = ttULONG(data+index_map+12); + stbtt_int32 low,high; + low = 0; high = (stbtt_int32)ngroups; + // Binary search the right group. + while (low < high) { + stbtt_int32 mid = low + ((high-low) >> 1); // rounds down, so low <= mid < high + stbtt_uint32 start_char = ttULONG(data+index_map+16+mid*12); + stbtt_uint32 end_char = ttULONG(data+index_map+16+mid*12+4); + if ((stbtt_uint32) unicode_codepoint < start_char) + high = mid; + else if ((stbtt_uint32) unicode_codepoint > end_char) + low = mid+1; + else { + stbtt_uint32 start_glyph = ttULONG(data+index_map+16+mid*12+8); + if (format == 12) + return start_glyph + unicode_codepoint-start_char; + else // format == 13 + return start_glyph; + } + } + return 0; // not found + } + // @TODO + STBTT_assert(0); + return 0; +} + +STBTT_DEF int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices) +{ + return stbtt_GetGlyphShape(info, stbtt_FindGlyphIndex(info, unicode_codepoint), vertices); +} + +static void stbtt_setvertex(stbtt_vertex *v, stbtt_uint8 type, stbtt_int32 x, stbtt_int32 y, stbtt_int32 cx, stbtt_int32 cy) +{ + v->type = type; + v->x = (stbtt_int16) x; + v->y = (stbtt_int16) y; + v->cx = (stbtt_int16) cx; + v->cy = (stbtt_int16) cy; +} + +static int stbtt__GetGlyfOffset(const stbtt_fontinfo *info, int glyph_index) +{ + int g1,g2; + + STBTT_assert(!info->cff.size); + + if (glyph_index >= info->numGlyphs) return -1; // glyph index out of range + if (info->indexToLocFormat >= 2) return -1; // unknown index->glyph map format + + if (info->indexToLocFormat == 0) { + g1 = info->glyf + ttUSHORT(info->data + info->loca + glyph_index * 2) * 2; + g2 = info->glyf + ttUSHORT(info->data + info->loca + glyph_index * 2 + 2) * 2; + } else { + g1 = info->glyf + ttULONG (info->data + info->loca + glyph_index * 4); + g2 = info->glyf + ttULONG (info->data + info->loca + glyph_index * 4 + 4); + } + + return g1==g2 ? -1 : g1; // if length is 0, return -1 +} + +static int stbtt__GetGlyphInfoT2(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1); + +STBTT_DEF int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1) +{ + if (info->cff.size) { + stbtt__GetGlyphInfoT2(info, glyph_index, x0, y0, x1, y1); + } else { + int g = stbtt__GetGlyfOffset(info, glyph_index); + if (g < 0) return 0; + + if (x0) *x0 = ttSHORT(info->data + g + 2); + if (y0) *y0 = ttSHORT(info->data + g + 4); + if (x1) *x1 = ttSHORT(info->data + g + 6); + if (y1) *y1 = ttSHORT(info->data + g + 8); + } + return 1; +} + +STBTT_DEF int stbtt_GetCodepointBox(const stbtt_fontinfo *info, int codepoint, int *x0, int *y0, int *x1, int *y1) +{ + return stbtt_GetGlyphBox(info, stbtt_FindGlyphIndex(info,codepoint), x0,y0,x1,y1); +} + +STBTT_DEF int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index) +{ + stbtt_int16 numberOfContours; + int g; + if (info->cff.size) + return stbtt__GetGlyphInfoT2(info, glyph_index, NULL, NULL, NULL, NULL) == 0; + g = stbtt__GetGlyfOffset(info, glyph_index); + if (g < 0) return 1; + numberOfContours = ttSHORT(info->data + g); + return numberOfContours == 0; +} + +static int stbtt__close_shape(stbtt_vertex *vertices, int num_vertices, int was_off, int start_off, + stbtt_int32 sx, stbtt_int32 sy, stbtt_int32 scx, stbtt_int32 scy, stbtt_int32 cx, stbtt_int32 cy) +{ + if (start_off) { + if (was_off) + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, (cx+scx)>>1, (cy+scy)>>1, cx,cy); + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, sx,sy,scx,scy); + } else { + if (was_off) + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve,sx,sy,cx,cy); + else + stbtt_setvertex(&vertices[num_vertices++], STBTT_vline,sx,sy,0,0); + } + return num_vertices; +} + +static int stbtt__GetGlyphShapeTT(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) +{ + stbtt_int16 numberOfContours; + stbtt_uint8 *endPtsOfContours; + stbtt_uint8 *data = info->data; + stbtt_vertex *vertices=0; + int num_vertices=0; + int g = stbtt__GetGlyfOffset(info, glyph_index); + + *pvertices = NULL; + + if (g < 0) return 0; + + numberOfContours = ttSHORT(data + g); + + if (numberOfContours > 0) { + stbtt_uint8 flags=0,flagcount; + stbtt_int32 ins, i,j=0,m,n, next_move, was_off=0, off, start_off=0; + stbtt_int32 x,y,cx,cy,sx,sy, scx,scy; + stbtt_uint8 *points; + endPtsOfContours = (data + g + 10); + ins = ttUSHORT(data + g + 10 + numberOfContours * 2); + points = data + g + 10 + numberOfContours * 2 + 2 + ins; + + n = 1+ttUSHORT(endPtsOfContours + numberOfContours*2-2); + + m = n + 2*numberOfContours; // a loose bound on how many vertices we might need + vertices = (stbtt_vertex *) STBTT_malloc(m * sizeof(vertices[0]), info->userdata); + if (vertices == 0) + return 0; + + next_move = 0; + flagcount=0; + + // in first pass, we load uninterpreted data into the allocated array + // above, shifted to the end of the array so we won't overwrite it when + // we create our final data starting from the front + + off = m - n; // starting offset for uninterpreted data, regardless of how m ends up being calculated + + // first load flags + + for (i=0; i < n; ++i) { + if (flagcount == 0) { + flags = *points++; + if (flags & 8) + flagcount = *points++; + } else + --flagcount; + vertices[off+i].type = flags; + } + + // now load x coordinates + x=0; + for (i=0; i < n; ++i) { + flags = vertices[off+i].type; + if (flags & 2) { + stbtt_int16 dx = *points++; + x += (flags & 16) ? dx : -dx; // ??? + } else { + if (!(flags & 16)) { + x = x + (stbtt_int16) (points[0]*256 + points[1]); + points += 2; + } + } + vertices[off+i].x = (stbtt_int16) x; + } + + // now load y coordinates + y=0; + for (i=0; i < n; ++i) { + flags = vertices[off+i].type; + if (flags & 4) { + stbtt_int16 dy = *points++; + y += (flags & 32) ? dy : -dy; // ??? + } else { + if (!(flags & 32)) { + y = y + (stbtt_int16) (points[0]*256 + points[1]); + points += 2; + } + } + vertices[off+i].y = (stbtt_int16) y; + } + + // now convert them to our format + num_vertices=0; + sx = sy = cx = cy = scx = scy = 0; + for (i=0; i < n; ++i) { + flags = vertices[off+i].type; + x = (stbtt_int16) vertices[off+i].x; + y = (stbtt_int16) vertices[off+i].y; + + if (next_move == i) { + if (i != 0) + num_vertices = stbtt__close_shape(vertices, num_vertices, was_off, start_off, sx,sy,scx,scy,cx,cy); + + // now start the new one + start_off = !(flags & 1); + if (start_off) { + // if we start off with an off-curve point, then when we need to find a point on the curve + // where we can start, and we need to save some state for when we wraparound. + scx = x; + scy = y; + if (!(vertices[off+i+1].type & 1)) { + // next point is also a curve point, so interpolate an on-point curve + sx = (x + (stbtt_int32) vertices[off+i+1].x) >> 1; + sy = (y + (stbtt_int32) vertices[off+i+1].y) >> 1; + } else { + // otherwise just use the next point as our start point + sx = (stbtt_int32) vertices[off+i+1].x; + sy = (stbtt_int32) vertices[off+i+1].y; + ++i; // we're using point i+1 as the starting point, so skip it + } + } else { + sx = x; + sy = y; + } + stbtt_setvertex(&vertices[num_vertices++], STBTT_vmove,sx,sy,0,0); + was_off = 0; + next_move = 1 + ttUSHORT(endPtsOfContours+j*2); + ++j; + } else { + if (!(flags & 1)) { // if it's a curve + if (was_off) // two off-curve control points in a row means interpolate an on-curve midpoint + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, (cx+x)>>1, (cy+y)>>1, cx, cy); + cx = x; + cy = y; + was_off = 1; + } else { + if (was_off) + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, x,y, cx, cy); + else + stbtt_setvertex(&vertices[num_vertices++], STBTT_vline, x,y,0,0); + was_off = 0; + } + } + } + num_vertices = stbtt__close_shape(vertices, num_vertices, was_off, start_off, sx,sy,scx,scy,cx,cy); + } else if (numberOfContours < 0) { + // Compound shapes. + int more = 1; + stbtt_uint8 *comp = data + g + 10; + num_vertices = 0; + vertices = 0; + while (more) { + stbtt_uint16 flags, gidx; + int comp_num_verts = 0, i; + stbtt_vertex *comp_verts = 0, *tmp = 0; + float mtx[6] = {1,0,0,1,0,0}, m, n; + + flags = ttSHORT(comp); comp+=2; + gidx = ttSHORT(comp); comp+=2; + + if (flags & 2) { // XY values + if (flags & 1) { // shorts + mtx[4] = ttSHORT(comp); comp+=2; + mtx[5] = ttSHORT(comp); comp+=2; + } else { + mtx[4] = ttCHAR(comp); comp+=1; + mtx[5] = ttCHAR(comp); comp+=1; + } + } + else { + // @TODO handle matching point + STBTT_assert(0); + } + if (flags & (1<<3)) { // WE_HAVE_A_SCALE + mtx[0] = mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[1] = mtx[2] = 0; + } else if (flags & (1<<6)) { // WE_HAVE_AN_X_AND_YSCALE + mtx[0] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[1] = mtx[2] = 0; + mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; + } else if (flags & (1<<7)) { // WE_HAVE_A_TWO_BY_TWO + mtx[0] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[1] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[2] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; + } + + // Find transformation scales. + m = (float) STBTT_sqrt(mtx[0]*mtx[0] + mtx[1]*mtx[1]); + n = (float) STBTT_sqrt(mtx[2]*mtx[2] + mtx[3]*mtx[3]); + + // Get indexed glyph. + comp_num_verts = stbtt_GetGlyphShape(info, gidx, &comp_verts); + if (comp_num_verts > 0) { + // Transform vertices. + for (i = 0; i < comp_num_verts; ++i) { + stbtt_vertex* v = &comp_verts[i]; + stbtt_vertex_type x,y; + x=v->x; y=v->y; + v->x = (stbtt_vertex_type)(m * (mtx[0]*x + mtx[2]*y + mtx[4])); + v->y = (stbtt_vertex_type)(n * (mtx[1]*x + mtx[3]*y + mtx[5])); + x=v->cx; y=v->cy; + v->cx = (stbtt_vertex_type)(m * (mtx[0]*x + mtx[2]*y + mtx[4])); + v->cy = (stbtt_vertex_type)(n * (mtx[1]*x + mtx[3]*y + mtx[5])); + } + // Append vertices. + tmp = (stbtt_vertex*)STBTT_malloc((num_vertices+comp_num_verts)*sizeof(stbtt_vertex), info->userdata); + if (!tmp) { + if (vertices) STBTT_free(vertices, info->userdata); + if (comp_verts) STBTT_free(comp_verts, info->userdata); + return 0; + } + if (num_vertices > 0 && vertices) STBTT_memcpy(tmp, vertices, num_vertices*sizeof(stbtt_vertex)); + STBTT_memcpy(tmp+num_vertices, comp_verts, comp_num_verts*sizeof(stbtt_vertex)); + if (vertices) STBTT_free(vertices, info->userdata); + vertices = tmp; + STBTT_free(comp_verts, info->userdata); + num_vertices += comp_num_verts; + } + // More components ? + more = flags & (1<<5); + } + } else { + // numberOfCounters == 0, do nothing + } + + *pvertices = vertices; + return num_vertices; +} + +typedef struct +{ + int bounds; + int started; + float first_x, first_y; + float x, y; + stbtt_int32 min_x, max_x, min_y, max_y; + + stbtt_vertex *pvertices; + int num_vertices; +} stbtt__csctx; + +#define STBTT__CSCTX_INIT(bounds) {bounds,0, 0,0, 0,0, 0,0,0,0, NULL, 0} + +static void stbtt__track_vertex(stbtt__csctx *c, stbtt_int32 x, stbtt_int32 y) +{ + if (x > c->max_x || !c->started) c->max_x = x; + if (y > c->max_y || !c->started) c->max_y = y; + if (x < c->min_x || !c->started) c->min_x = x; + if (y < c->min_y || !c->started) c->min_y = y; + c->started = 1; +} + +static void stbtt__csctx_v(stbtt__csctx *c, stbtt_uint8 type, stbtt_int32 x, stbtt_int32 y, stbtt_int32 cx, stbtt_int32 cy, stbtt_int32 cx1, stbtt_int32 cy1) +{ + if (c->bounds) { + stbtt__track_vertex(c, x, y); + if (type == STBTT_vcubic) { + stbtt__track_vertex(c, cx, cy); + stbtt__track_vertex(c, cx1, cy1); + } + } else { + stbtt_setvertex(&c->pvertices[c->num_vertices], type, x, y, cx, cy); + c->pvertices[c->num_vertices].cx1 = (stbtt_int16) cx1; + c->pvertices[c->num_vertices].cy1 = (stbtt_int16) cy1; + } + c->num_vertices++; +} + +static void stbtt__csctx_close_shape(stbtt__csctx *ctx) +{ + if (ctx->first_x != ctx->x || ctx->first_y != ctx->y) + stbtt__csctx_v(ctx, STBTT_vline, (int)ctx->first_x, (int)ctx->first_y, 0, 0, 0, 0); +} + +static void stbtt__csctx_rmove_to(stbtt__csctx *ctx, float dx, float dy) +{ + stbtt__csctx_close_shape(ctx); + ctx->first_x = ctx->x = ctx->x + dx; + ctx->first_y = ctx->y = ctx->y + dy; + stbtt__csctx_v(ctx, STBTT_vmove, (int)ctx->x, (int)ctx->y, 0, 0, 0, 0); +} + +static void stbtt__csctx_rline_to(stbtt__csctx *ctx, float dx, float dy) +{ + ctx->x += dx; + ctx->y += dy; + stbtt__csctx_v(ctx, STBTT_vline, (int)ctx->x, (int)ctx->y, 0, 0, 0, 0); +} + +static void stbtt__csctx_rccurve_to(stbtt__csctx *ctx, float dx1, float dy1, float dx2, float dy2, float dx3, float dy3) +{ + float cx1 = ctx->x + dx1; + float cy1 = ctx->y + dy1; + float cx2 = cx1 + dx2; + float cy2 = cy1 + dy2; + ctx->x = cx2 + dx3; + ctx->y = cy2 + dy3; + stbtt__csctx_v(ctx, STBTT_vcubic, (int)ctx->x, (int)ctx->y, (int)cx1, (int)cy1, (int)cx2, (int)cy2); +} + +static stbtt__buf stbtt__get_subr(stbtt__buf idx, int n) +{ + int count = stbtt__cff_index_count(&idx); + int bias = 107; + if (count >= 33900) + bias = 32768; + else if (count >= 1240) + bias = 1131; + n += bias; + if (n < 0 || n >= count) + return stbtt__new_buf(NULL, 0); + return stbtt__cff_index_get(idx, n); +} + +static stbtt__buf stbtt__cid_get_glyph_subrs(const stbtt_fontinfo *info, int glyph_index) +{ + stbtt__buf fdselect = info->fdselect; + int nranges, start, end, v, fmt, fdselector = -1, i; + + stbtt__buf_seek(&fdselect, 0); + fmt = stbtt__buf_get8(&fdselect); + if (fmt == 0) { + // untested + stbtt__buf_skip(&fdselect, glyph_index); + fdselector = stbtt__buf_get8(&fdselect); + } else if (fmt == 3) { + nranges = stbtt__buf_get16(&fdselect); + start = stbtt__buf_get16(&fdselect); + for (i = 0; i < nranges; i++) { + v = stbtt__buf_get8(&fdselect); + end = stbtt__buf_get16(&fdselect); + if (glyph_index >= start && glyph_index < end) { + fdselector = v; + break; + } + start = end; + } + } + if (fdselector == -1) stbtt__new_buf(NULL, 0); + return stbtt__get_subrs(info->cff, stbtt__cff_index_get(info->fontdicts, fdselector)); +} + +static int stbtt__run_charstring(const stbtt_fontinfo *info, int glyph_index, stbtt__csctx *c) +{ + int in_header = 1, maskbits = 0, subr_stack_height = 0, sp = 0, v, i, b0; + int has_subrs = 0, clear_stack; + float s[48]; + stbtt__buf subr_stack[10], subrs = info->subrs, b; + float f; + +#define STBTT__CSERR(s) (0) + + // this currently ignores the initial width value, which isn't needed if we have hmtx + b = stbtt__cff_index_get(info->charstrings, glyph_index); + while (b.cursor < b.size) { + i = 0; + clear_stack = 1; + b0 = stbtt__buf_get8(&b); + switch (b0) { + // @TODO implement hinting + case 0x13: // hintmask + case 0x14: // cntrmask + if (in_header) + maskbits += (sp / 2); // implicit "vstem" + in_header = 0; + stbtt__buf_skip(&b, (maskbits + 7) / 8); + break; + + case 0x01: // hstem + case 0x03: // vstem + case 0x12: // hstemhm + case 0x17: // vstemhm + maskbits += (sp / 2); + break; + + case 0x15: // rmoveto + in_header = 0; + if (sp < 2) return STBTT__CSERR("rmoveto stack"); + stbtt__csctx_rmove_to(c, s[sp-2], s[sp-1]); + break; + case 0x04: // vmoveto + in_header = 0; + if (sp < 1) return STBTT__CSERR("vmoveto stack"); + stbtt__csctx_rmove_to(c, 0, s[sp-1]); + break; + case 0x16: // hmoveto + in_header = 0; + if (sp < 1) return STBTT__CSERR("hmoveto stack"); + stbtt__csctx_rmove_to(c, s[sp-1], 0); + break; + + case 0x05: // rlineto + if (sp < 2) return STBTT__CSERR("rlineto stack"); + for (; i + 1 < sp; i += 2) + stbtt__csctx_rline_to(c, s[i], s[i+1]); + break; + + // hlineto/vlineto and vhcurveto/hvcurveto alternate horizontal and vertical + // starting from a different place. + + case 0x07: // vlineto + if (sp < 1) return STBTT__CSERR("vlineto stack"); + goto vlineto; + case 0x06: // hlineto + if (sp < 1) return STBTT__CSERR("hlineto stack"); + for (;;) { + if (i >= sp) break; + stbtt__csctx_rline_to(c, s[i], 0); + i++; + vlineto: + if (i >= sp) break; + stbtt__csctx_rline_to(c, 0, s[i]); + i++; + } + break; + + case 0x1F: // hvcurveto + if (sp < 4) return STBTT__CSERR("hvcurveto stack"); + goto hvcurveto; + case 0x1E: // vhcurveto + if (sp < 4) return STBTT__CSERR("vhcurveto stack"); + for (;;) { + if (i + 3 >= sp) break; + stbtt__csctx_rccurve_to(c, 0, s[i], s[i+1], s[i+2], s[i+3], (sp - i == 5) ? s[i + 4] : 0.0f); + i += 4; + hvcurveto: + if (i + 3 >= sp) break; + stbtt__csctx_rccurve_to(c, s[i], 0, s[i+1], s[i+2], (sp - i == 5) ? s[i+4] : 0.0f, s[i+3]); + i += 4; + } + break; + + case 0x08: // rrcurveto + if (sp < 6) return STBTT__CSERR("rcurveline stack"); + for (; i + 5 < sp; i += 6) + stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); + break; + + case 0x18: // rcurveline + if (sp < 8) return STBTT__CSERR("rcurveline stack"); + for (; i + 5 < sp - 2; i += 6) + stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); + if (i + 1 >= sp) return STBTT__CSERR("rcurveline stack"); + stbtt__csctx_rline_to(c, s[i], s[i+1]); + break; + + case 0x19: // rlinecurve + if (sp < 8) return STBTT__CSERR("rlinecurve stack"); + for (; i + 1 < sp - 6; i += 2) + stbtt__csctx_rline_to(c, s[i], s[i+1]); + if (i + 5 >= sp) return STBTT__CSERR("rlinecurve stack"); + stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); + break; + + case 0x1A: // vvcurveto + case 0x1B: // hhcurveto + if (sp < 4) return STBTT__CSERR("(vv|hh)curveto stack"); + f = 0.0; + if (sp & 1) { f = s[i]; i++; } + for (; i + 3 < sp; i += 4) { + if (b0 == 0x1B) + stbtt__csctx_rccurve_to(c, s[i], f, s[i+1], s[i+2], s[i+3], 0.0); + else + stbtt__csctx_rccurve_to(c, f, s[i], s[i+1], s[i+2], 0.0, s[i+3]); + f = 0.0; + } + break; + + case 0x0A: // callsubr + if (!has_subrs) { + if (info->fdselect.size) + subrs = stbtt__cid_get_glyph_subrs(info, glyph_index); + has_subrs = 1; + } + // FALLTHROUGH + case 0x1D: // callgsubr + if (sp < 1) return STBTT__CSERR("call(g|)subr stack"); + v = (int) s[--sp]; + if (subr_stack_height >= 10) return STBTT__CSERR("recursion limit"); + subr_stack[subr_stack_height++] = b; + b = stbtt__get_subr(b0 == 0x0A ? subrs : info->gsubrs, v); + if (b.size == 0) return STBTT__CSERR("subr not found"); + b.cursor = 0; + clear_stack = 0; + break; + + case 0x0B: // return + if (subr_stack_height <= 0) return STBTT__CSERR("return outside subr"); + b = subr_stack[--subr_stack_height]; + clear_stack = 0; + break; + + case 0x0E: // endchar + stbtt__csctx_close_shape(c); + return 1; + + case 0x0C: { // two-byte escape + float dx1, dx2, dx3, dx4, dx5, dx6, dy1, dy2, dy3, dy4, dy5, dy6; + float dx, dy; + int b1 = stbtt__buf_get8(&b); + switch (b1) { + // @TODO These "flex" implementations ignore the flex-depth and resolution, + // and always draw beziers. + case 0x22: // hflex + if (sp < 7) return STBTT__CSERR("hflex stack"); + dx1 = s[0]; + dx2 = s[1]; + dy2 = s[2]; + dx3 = s[3]; + dx4 = s[4]; + dx5 = s[5]; + dx6 = s[6]; + stbtt__csctx_rccurve_to(c, dx1, 0, dx2, dy2, dx3, 0); + stbtt__csctx_rccurve_to(c, dx4, 0, dx5, -dy2, dx6, 0); + break; + + case 0x23: // flex + if (sp < 13) return STBTT__CSERR("flex stack"); + dx1 = s[0]; + dy1 = s[1]; + dx2 = s[2]; + dy2 = s[3]; + dx3 = s[4]; + dy3 = s[5]; + dx4 = s[6]; + dy4 = s[7]; + dx5 = s[8]; + dy5 = s[9]; + dx6 = s[10]; + dy6 = s[11]; + //fd is s[12] + stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, dy3); + stbtt__csctx_rccurve_to(c, dx4, dy4, dx5, dy5, dx6, dy6); + break; + + case 0x24: // hflex1 + if (sp < 9) return STBTT__CSERR("hflex1 stack"); + dx1 = s[0]; + dy1 = s[1]; + dx2 = s[2]; + dy2 = s[3]; + dx3 = s[4]; + dx4 = s[5]; + dx5 = s[6]; + dy5 = s[7]; + dx6 = s[8]; + stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, 0); + stbtt__csctx_rccurve_to(c, dx4, 0, dx5, dy5, dx6, -(dy1+dy2+dy5)); + break; + + case 0x25: // flex1 + if (sp < 11) return STBTT__CSERR("flex1 stack"); + dx1 = s[0]; + dy1 = s[1]; + dx2 = s[2]; + dy2 = s[3]; + dx3 = s[4]; + dy3 = s[5]; + dx4 = s[6]; + dy4 = s[7]; + dx5 = s[8]; + dy5 = s[9]; + dx6 = dy6 = s[10]; + dx = dx1+dx2+dx3+dx4+dx5; + dy = dy1+dy2+dy3+dy4+dy5; + if (STBTT_fabs(dx) > STBTT_fabs(dy)) + dy6 = -dy; + else + dx6 = -dx; + stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, dy3); + stbtt__csctx_rccurve_to(c, dx4, dy4, dx5, dy5, dx6, dy6); + break; + + default: + return STBTT__CSERR("unimplemented"); + } + } break; + + default: + if (b0 != 255 && b0 != 28 && b0 < 32) + return STBTT__CSERR("reserved operator"); + + // push immediate + if (b0 == 255) { + f = (float)(stbtt_int32)stbtt__buf_get32(&b) / 0x10000; + } else { + stbtt__buf_skip(&b, -1); + f = (float)(stbtt_int16)stbtt__cff_int(&b); + } + if (sp >= 48) return STBTT__CSERR("push stack overflow"); + s[sp++] = f; + clear_stack = 0; + break; + } + if (clear_stack) sp = 0; + } + return STBTT__CSERR("no endchar"); + +#undef STBTT__CSERR +} + +static int stbtt__GetGlyphShapeT2(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) +{ + // runs the charstring twice, once to count and once to output (to avoid realloc) + stbtt__csctx count_ctx = STBTT__CSCTX_INIT(1); + stbtt__csctx output_ctx = STBTT__CSCTX_INIT(0); + if (stbtt__run_charstring(info, glyph_index, &count_ctx)) { + *pvertices = (stbtt_vertex*)STBTT_malloc(count_ctx.num_vertices*sizeof(stbtt_vertex), info->userdata); + output_ctx.pvertices = *pvertices; + if (stbtt__run_charstring(info, glyph_index, &output_ctx)) { + STBTT_assert(output_ctx.num_vertices == count_ctx.num_vertices); + return output_ctx.num_vertices; + } + } + *pvertices = NULL; + return 0; +} + +static int stbtt__GetGlyphInfoT2(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1) +{ + stbtt__csctx c = STBTT__CSCTX_INIT(1); + int r = stbtt__run_charstring(info, glyph_index, &c); + if (x0) *x0 = r ? c.min_x : 0; + if (y0) *y0 = r ? c.min_y : 0; + if (x1) *x1 = r ? c.max_x : 0; + if (y1) *y1 = r ? c.max_y : 0; + return r ? c.num_vertices : 0; +} + +STBTT_DEF int stbtt_GetGlyphShape(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) +{ + if (!info->cff.size) + return stbtt__GetGlyphShapeTT(info, glyph_index, pvertices); + else + return stbtt__GetGlyphShapeT2(info, glyph_index, pvertices); +} + +STBTT_DEF void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing) +{ + stbtt_uint16 numOfLongHorMetrics = ttUSHORT(info->data+info->hhea + 34); + if (glyph_index < numOfLongHorMetrics) { + if (advanceWidth) *advanceWidth = ttSHORT(info->data + info->hmtx + 4*glyph_index); + if (leftSideBearing) *leftSideBearing = ttSHORT(info->data + info->hmtx + 4*glyph_index + 2); + } else { + if (advanceWidth) *advanceWidth = ttSHORT(info->data + info->hmtx + 4*(numOfLongHorMetrics-1)); + if (leftSideBearing) *leftSideBearing = ttSHORT(info->data + info->hmtx + 4*numOfLongHorMetrics + 2*(glyph_index - numOfLongHorMetrics)); + } +} + +STBTT_DEF int stbtt_GetKerningTableLength(const stbtt_fontinfo *info) +{ + stbtt_uint8 *data = info->data + info->kern; + + // we only look at the first table. it must be 'horizontal' and format 0. + if (!info->kern) + return 0; + if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 + return 0; + if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format + return 0; + + return ttUSHORT(data+10); +} + +STBTT_DEF int stbtt_GetKerningTable(const stbtt_fontinfo *info, stbtt_kerningentry* table, int table_length) +{ + stbtt_uint8 *data = info->data + info->kern; + int k, length; + + // we only look at the first table. it must be 'horizontal' and format 0. + if (!info->kern) + return 0; + if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 + return 0; + if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format + return 0; + + length = ttUSHORT(data+10); + if (table_length < length) + length = table_length; + + for (k = 0; k < length; k++) + { + table[k].glyph1 = ttUSHORT(data+18+(k*6)); + table[k].glyph2 = ttUSHORT(data+20+(k*6)); + table[k].advance = ttSHORT(data+22+(k*6)); + } + + return length; +} + +static int stbtt__GetGlyphKernInfoAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2) +{ + stbtt_uint8 *data = info->data + info->kern; + stbtt_uint32 needle, straw; + int l, r, m; + + // we only look at the first table. it must be 'horizontal' and format 0. + if (!info->kern) + return 0; + if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 + return 0; + if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format + return 0; + + l = 0; + r = ttUSHORT(data+10) - 1; + needle = glyph1 << 16 | glyph2; + while (l <= r) { + m = (l + r) >> 1; + straw = ttULONG(data+18+(m*6)); // note: unaligned read + if (needle < straw) + r = m - 1; + else if (needle > straw) + l = m + 1; + else + return ttSHORT(data+22+(m*6)); + } + return 0; +} + +static stbtt_int32 stbtt__GetCoverageIndex(stbtt_uint8 *coverageTable, int glyph) +{ + stbtt_uint16 coverageFormat = ttUSHORT(coverageTable); + switch (coverageFormat) { + case 1: { + stbtt_uint16 glyphCount = ttUSHORT(coverageTable + 2); + + // Binary search. + stbtt_int32 l=0, r=glyphCount-1, m; + int straw, needle=glyph; + while (l <= r) { + stbtt_uint8 *glyphArray = coverageTable + 4; + stbtt_uint16 glyphID; + m = (l + r) >> 1; + glyphID = ttUSHORT(glyphArray + 2 * m); + straw = glyphID; + if (needle < straw) + r = m - 1; + else if (needle > straw) + l = m + 1; + else { + return m; + } + } + break; + } + + case 2: { + stbtt_uint16 rangeCount = ttUSHORT(coverageTable + 2); + stbtt_uint8 *rangeArray = coverageTable + 4; + + // Binary search. + stbtt_int32 l=0, r=rangeCount-1, m; + int strawStart, strawEnd, needle=glyph; + while (l <= r) { + stbtt_uint8 *rangeRecord; + m = (l + r) >> 1; + rangeRecord = rangeArray + 6 * m; + strawStart = ttUSHORT(rangeRecord); + strawEnd = ttUSHORT(rangeRecord + 2); + if (needle < strawStart) + r = m - 1; + else if (needle > strawEnd) + l = m + 1; + else { + stbtt_uint16 startCoverageIndex = ttUSHORT(rangeRecord + 4); + return startCoverageIndex + glyph - strawStart; + } + } + break; + } + + default: return -1; // unsupported + } + + return -1; +} + +static stbtt_int32 stbtt__GetGlyphClass(stbtt_uint8 *classDefTable, int glyph) +{ + stbtt_uint16 classDefFormat = ttUSHORT(classDefTable); + switch (classDefFormat) + { + case 1: { + stbtt_uint16 startGlyphID = ttUSHORT(classDefTable + 2); + stbtt_uint16 glyphCount = ttUSHORT(classDefTable + 4); + stbtt_uint8 *classDef1ValueArray = classDefTable + 6; + + if (glyph >= startGlyphID && glyph < startGlyphID + glyphCount) + return (stbtt_int32)ttUSHORT(classDef1ValueArray + 2 * (glyph - startGlyphID)); + break; + } + + case 2: { + stbtt_uint16 classRangeCount = ttUSHORT(classDefTable + 2); + stbtt_uint8 *classRangeRecords = classDefTable + 4; + + // Binary search. + stbtt_int32 l=0, r=classRangeCount-1, m; + int strawStart, strawEnd, needle=glyph; + while (l <= r) { + stbtt_uint8 *classRangeRecord; + m = (l + r) >> 1; + classRangeRecord = classRangeRecords + 6 * m; + strawStart = ttUSHORT(classRangeRecord); + strawEnd = ttUSHORT(classRangeRecord + 2); + if (needle < strawStart) + r = m - 1; + else if (needle > strawEnd) + l = m + 1; + else + return (stbtt_int32)ttUSHORT(classRangeRecord + 4); + } + break; + } + + default: + return -1; // Unsupported definition type, return an error. + } + + // "All glyphs not assigned to a class fall into class 0". (OpenType spec) + return 0; +} + +// Define to STBTT_assert(x) if you want to break on unimplemented formats. +#define STBTT_GPOS_TODO_assert(x) + +static stbtt_int32 stbtt__GetGlyphGPOSInfoAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2) +{ + stbtt_uint16 lookupListOffset; + stbtt_uint8 *lookupList; + stbtt_uint16 lookupCount; + stbtt_uint8 *data; + stbtt_int32 i, sti; + + if (!info->gpos) return 0; + + data = info->data + info->gpos; + + if (ttUSHORT(data+0) != 1) return 0; // Major version 1 + if (ttUSHORT(data+2) != 0) return 0; // Minor version 0 + + lookupListOffset = ttUSHORT(data+8); + lookupList = data + lookupListOffset; + lookupCount = ttUSHORT(lookupList); + + for (i=0; i= pairSetCount) return 0; + + needle=glyph2; + r=pairValueCount-1; + l=0; + + // Binary search. + while (l <= r) { + stbtt_uint16 secondGlyph; + stbtt_uint8 *pairValue; + m = (l + r) >> 1; + pairValue = pairValueArray + (2 + valueRecordPairSizeInBytes) * m; + secondGlyph = ttUSHORT(pairValue); + straw = secondGlyph; + if (needle < straw) + r = m - 1; + else if (needle > straw) + l = m + 1; + else { + stbtt_int16 xAdvance = ttSHORT(pairValue + 2); + return xAdvance; + } + } + } else + return 0; + break; + } + + case 2: { + stbtt_uint16 valueFormat1 = ttUSHORT(table + 4); + stbtt_uint16 valueFormat2 = ttUSHORT(table + 6); + if (valueFormat1 == 4 && valueFormat2 == 0) { // Support more formats? + stbtt_uint16 classDef1Offset = ttUSHORT(table + 8); + stbtt_uint16 classDef2Offset = ttUSHORT(table + 10); + int glyph1class = stbtt__GetGlyphClass(table + classDef1Offset, glyph1); + int glyph2class = stbtt__GetGlyphClass(table + classDef2Offset, glyph2); + + stbtt_uint16 class1Count = ttUSHORT(table + 12); + stbtt_uint16 class2Count = ttUSHORT(table + 14); + stbtt_uint8 *class1Records, *class2Records; + stbtt_int16 xAdvance; + + if (glyph1class < 0 || glyph1class >= class1Count) return 0; // malformed + if (glyph2class < 0 || glyph2class >= class2Count) return 0; // malformed + + class1Records = table + 16; + class2Records = class1Records + 2 * (glyph1class * class2Count); + xAdvance = ttSHORT(class2Records + 2 * glyph2class); + return xAdvance; + } else + return 0; + break; + } + + default: + return 0; // Unsupported position format + } + } + } + + return 0; +} + +STBTT_DEF int stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int g1, int g2) +{ + int xAdvance = 0; + + if (info->gpos) + xAdvance += stbtt__GetGlyphGPOSInfoAdvance(info, g1, g2); + else if (info->kern) + xAdvance += stbtt__GetGlyphKernInfoAdvance(info, g1, g2); + + return xAdvance; +} + +STBTT_DEF int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2) +{ + if (!info->kern && !info->gpos) // if no kerning table, don't waste time looking up both codepoint->glyphs + return 0; + return stbtt_GetGlyphKernAdvance(info, stbtt_FindGlyphIndex(info,ch1), stbtt_FindGlyphIndex(info,ch2)); +} + +STBTT_DEF void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing) +{ + stbtt_GetGlyphHMetrics(info, stbtt_FindGlyphIndex(info,codepoint), advanceWidth, leftSideBearing); +} + +STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, int *descent, int *lineGap) +{ + if (ascent ) *ascent = ttSHORT(info->data+info->hhea + 4); + if (descent) *descent = ttSHORT(info->data+info->hhea + 6); + if (lineGap) *lineGap = ttSHORT(info->data+info->hhea + 8); +} + +STBTT_DEF int stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap) +{ + int tab = stbtt__find_table(info->data, info->fontstart, "OS/2"); + if (!tab) + return 0; + if (typoAscent ) *typoAscent = ttSHORT(info->data+tab + 68); + if (typoDescent) *typoDescent = ttSHORT(info->data+tab + 70); + if (typoLineGap) *typoLineGap = ttSHORT(info->data+tab + 72); + return 1; +} + +STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1) +{ + *x0 = ttSHORT(info->data + info->head + 36); + *y0 = ttSHORT(info->data + info->head + 38); + *x1 = ttSHORT(info->data + info->head + 40); + *y1 = ttSHORT(info->data + info->head + 42); +} + +STBTT_DEF float stbtt_ScaleForPixelHeight(const stbtt_fontinfo *info, float height) +{ + int fheight = ttSHORT(info->data + info->hhea + 4) - ttSHORT(info->data + info->hhea + 6); + return (float) height / fheight; +} + +STBTT_DEF float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels) +{ + int unitsPerEm = ttUSHORT(info->data + info->head + 18); + return pixels / unitsPerEm; +} + +STBTT_DEF void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *v) +{ + STBTT_free(v, info->userdata); +} + +STBTT_DEF stbtt_uint8 *stbtt_FindSVGDoc(const stbtt_fontinfo *info, int gl) +{ + int i; + stbtt_uint8 *data = info->data; + stbtt_uint8 *svg_doc_list = data + stbtt__get_svg((stbtt_fontinfo *) info); + + int numEntries = ttUSHORT(svg_doc_list); + stbtt_uint8 *svg_docs = svg_doc_list + 2; + + for(i=0; i= ttUSHORT(svg_doc)) && (gl <= ttUSHORT(svg_doc + 2))) + return svg_doc; + } + return 0; +} + +STBTT_DEF int stbtt_GetGlyphSVG(const stbtt_fontinfo *info, int gl, const char **svg) +{ + stbtt_uint8 *data = info->data; + stbtt_uint8 *svg_doc; + + if (info->svg == 0) + return 0; + + svg_doc = stbtt_FindSVGDoc(info, gl); + if (svg_doc != NULL) { + *svg = (char *) data + info->svg + ttULONG(svg_doc + 4); + return ttULONG(svg_doc + 8); + } else { + return 0; + } +} + +STBTT_DEF int stbtt_GetCodepointSVG(const stbtt_fontinfo *info, int unicode_codepoint, const char **svg) +{ + return stbtt_GetGlyphSVG(info, stbtt_FindGlyphIndex(info, unicode_codepoint), svg); +} + +////////////////////////////////////////////////////////////////////////////// +// +// antialiasing software rasterizer +// + +STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1) +{ + int x0=0,y0=0,x1,y1; // =0 suppresses compiler warning + if (!stbtt_GetGlyphBox(font, glyph, &x0,&y0,&x1,&y1)) { + // e.g. space character + if (ix0) *ix0 = 0; + if (iy0) *iy0 = 0; + if (ix1) *ix1 = 0; + if (iy1) *iy1 = 0; + } else { + // move to integral bboxes (treating pixels as little squares, what pixels get touched)? + if (ix0) *ix0 = STBTT_ifloor( x0 * scale_x + shift_x); + if (iy0) *iy0 = STBTT_ifloor(-y1 * scale_y + shift_y); + if (ix1) *ix1 = STBTT_iceil ( x1 * scale_x + shift_x); + if (iy1) *iy1 = STBTT_iceil (-y0 * scale_y + shift_y); + } +} + +STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1) +{ + stbtt_GetGlyphBitmapBoxSubpixel(font, glyph, scale_x, scale_y,0.0f,0.0f, ix0, iy0, ix1, iy1); +} + +STBTT_DEF void stbtt_GetCodepointBitmapBoxSubpixel(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1) +{ + stbtt_GetGlyphBitmapBoxSubpixel(font, stbtt_FindGlyphIndex(font,codepoint), scale_x, scale_y,shift_x,shift_y, ix0,iy0,ix1,iy1); +} + +STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1) +{ + stbtt_GetCodepointBitmapBoxSubpixel(font, codepoint, scale_x, scale_y,0.0f,0.0f, ix0,iy0,ix1,iy1); +} + +////////////////////////////////////////////////////////////////////////////// +// +// Rasterizer + +typedef struct stbtt__hheap_chunk +{ + struct stbtt__hheap_chunk *next; +} stbtt__hheap_chunk; + +typedef struct stbtt__hheap +{ + struct stbtt__hheap_chunk *head; + void *first_free; + int num_remaining_in_head_chunk; +} stbtt__hheap; + +static void *stbtt__hheap_alloc(stbtt__hheap *hh, size_t size, void *userdata) +{ + if (hh->first_free) { + void *p = hh->first_free; + hh->first_free = * (void **) p; + return p; + } else { + if (hh->num_remaining_in_head_chunk == 0) { + int count = (size < 32 ? 2000 : size < 128 ? 800 : 100); + stbtt__hheap_chunk *c = (stbtt__hheap_chunk *) STBTT_malloc(sizeof(stbtt__hheap_chunk) + size * count, userdata); + if (c == NULL) + return NULL; + c->next = hh->head; + hh->head = c; + hh->num_remaining_in_head_chunk = count; + } + --hh->num_remaining_in_head_chunk; + return (char *) (hh->head) + sizeof(stbtt__hheap_chunk) + size * hh->num_remaining_in_head_chunk; + } +} + +static void stbtt__hheap_free(stbtt__hheap *hh, void *p) +{ + *(void **) p = hh->first_free; + hh->first_free = p; +} + +static void stbtt__hheap_cleanup(stbtt__hheap *hh, void *userdata) +{ + stbtt__hheap_chunk *c = hh->head; + while (c) { + stbtt__hheap_chunk *n = c->next; + STBTT_free(c, userdata); + c = n; + } +} + +typedef struct stbtt__edge { + float x0,y0, x1,y1; + int invert; +} stbtt__edge; + + +typedef struct stbtt__active_edge +{ + struct stbtt__active_edge *next; + #if STBTT_RASTERIZER_VERSION==1 + int x,dx; + float ey; + int direction; + #elif STBTT_RASTERIZER_VERSION==2 + float fx,fdx,fdy; + float direction; + float sy; + float ey; + #else + #error "Unrecognized value of STBTT_RASTERIZER_VERSION" + #endif +} stbtt__active_edge; + +#if STBTT_RASTERIZER_VERSION == 1 +#define STBTT_FIXSHIFT 10 +#define STBTT_FIX (1 << STBTT_FIXSHIFT) +#define STBTT_FIXMASK (STBTT_FIX-1) + +static stbtt__active_edge *stbtt__new_active(stbtt__hheap *hh, stbtt__edge *e, int off_x, float start_point, void *userdata) +{ + stbtt__active_edge *z = (stbtt__active_edge *) stbtt__hheap_alloc(hh, sizeof(*z), userdata); + float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0); + STBTT_assert(z != NULL); + if (!z) return z; + + // round dx down to avoid overshooting + if (dxdy < 0) + z->dx = -STBTT_ifloor(STBTT_FIX * -dxdy); + else + z->dx = STBTT_ifloor(STBTT_FIX * dxdy); + + z->x = STBTT_ifloor(STBTT_FIX * e->x0 + z->dx * (start_point - e->y0)); // use z->dx so when we offset later it's by the same amount + z->x -= off_x * STBTT_FIX; + + z->ey = e->y1; + z->next = 0; + z->direction = e->invert ? 1 : -1; + return z; +} +#elif STBTT_RASTERIZER_VERSION == 2 +static stbtt__active_edge *stbtt__new_active(stbtt__hheap *hh, stbtt__edge *e, int off_x, float start_point, void *userdata) +{ + stbtt__active_edge *z = (stbtt__active_edge *) stbtt__hheap_alloc(hh, sizeof(*z), userdata); + float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0); + STBTT_assert(z != NULL); + //STBTT_assert(e->y0 <= start_point); + if (!z) return z; + z->fdx = dxdy; + z->fdy = dxdy != 0.0f ? (1.0f/dxdy) : 0.0f; + z->fx = e->x0 + dxdy * (start_point - e->y0); + z->fx -= off_x; + z->direction = e->invert ? 1.0f : -1.0f; + z->sy = e->y0; + z->ey = e->y1; + z->next = 0; + return z; +} +#else +#error "Unrecognized value of STBTT_RASTERIZER_VERSION" +#endif + +#if STBTT_RASTERIZER_VERSION == 1 +// note: this routine clips fills that extend off the edges... ideally this +// wouldn't happen, but it could happen if the truetype glyph bounding boxes +// are wrong, or if the user supplies a too-small bitmap +static void stbtt__fill_active_edges(unsigned char *scanline, int len, stbtt__active_edge *e, int max_weight) +{ + // non-zero winding fill + int x0=0, w=0; + + while (e) { + if (w == 0) { + // if we're currently at zero, we need to record the edge start point + x0 = e->x; w += e->direction; + } else { + int x1 = e->x; w += e->direction; + // if we went to zero, we need to draw + if (w == 0) { + int i = x0 >> STBTT_FIXSHIFT; + int j = x1 >> STBTT_FIXSHIFT; + + if (i < len && j >= 0) { + if (i == j) { + // x0,x1 are the same pixel, so compute combined coverage + scanline[i] = scanline[i] + (stbtt_uint8) ((x1 - x0) * max_weight >> STBTT_FIXSHIFT); + } else { + if (i >= 0) // add antialiasing for x0 + scanline[i] = scanline[i] + (stbtt_uint8) (((STBTT_FIX - (x0 & STBTT_FIXMASK)) * max_weight) >> STBTT_FIXSHIFT); + else + i = -1; // clip + + if (j < len) // add antialiasing for x1 + scanline[j] = scanline[j] + (stbtt_uint8) (((x1 & STBTT_FIXMASK) * max_weight) >> STBTT_FIXSHIFT); + else + j = len; // clip + + for (++i; i < j; ++i) // fill pixels between x0 and x1 + scanline[i] = scanline[i] + (stbtt_uint8) max_weight; + } + } + } + } + + e = e->next; + } +} + +static void stbtt__rasterize_sorted_edges(stbtt__bitmap *result, stbtt__edge *e, int n, int vsubsample, int off_x, int off_y, void *userdata) +{ + stbtt__hheap hh = { 0, 0, 0 }; + stbtt__active_edge *active = NULL; + int y,j=0; + int max_weight = (255 / vsubsample); // weight per vertical scanline + int s; // vertical subsample index + unsigned char scanline_data[512], *scanline; + + if (result->w > 512) + scanline = (unsigned char *) STBTT_malloc(result->w, userdata); + else + scanline = scanline_data; + + y = off_y * vsubsample; + e[n].y0 = (off_y + result->h) * (float) vsubsample + 1; + + while (j < result->h) { + STBTT_memset(scanline, 0, result->w); + for (s=0; s < vsubsample; ++s) { + // find center of pixel for this scanline + float scan_y = y + 0.5f; + stbtt__active_edge **step = &active; + + // update all active edges; + // remove all active edges that terminate before the center of this scanline + while (*step) { + stbtt__active_edge * z = *step; + if (z->ey <= scan_y) { + *step = z->next; // delete from list + STBTT_assert(z->direction); + z->direction = 0; + stbtt__hheap_free(&hh, z); + } else { + z->x += z->dx; // advance to position for current scanline + step = &((*step)->next); // advance through list + } + } + + // resort the list if needed + for(;;) { + int changed=0; + step = &active; + while (*step && (*step)->next) { + if ((*step)->x > (*step)->next->x) { + stbtt__active_edge *t = *step; + stbtt__active_edge *q = t->next; + + t->next = q->next; + q->next = t; + *step = q; + changed = 1; + } + step = &(*step)->next; + } + if (!changed) break; + } + + // insert all edges that start before the center of this scanline -- omit ones that also end on this scanline + while (e->y0 <= scan_y) { + if (e->y1 > scan_y) { + stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y, userdata); + if (z != NULL) { + // find insertion point + if (active == NULL) + active = z; + else if (z->x < active->x) { + // insert at front + z->next = active; + active = z; + } else { + // find thing to insert AFTER + stbtt__active_edge *p = active; + while (p->next && p->next->x < z->x) + p = p->next; + // at this point, p->next->x is NOT < z->x + z->next = p->next; + p->next = z; + } + } + } + ++e; + } + + // now process all active edges in XOR fashion + if (active) + stbtt__fill_active_edges(scanline, result->w, active, max_weight); + + ++y; + } + STBTT_memcpy(result->pixels + j * result->stride, scanline, result->w); + ++j; + } + + stbtt__hheap_cleanup(&hh, userdata); + + if (scanline != scanline_data) + STBTT_free(scanline, userdata); +} + +#elif STBTT_RASTERIZER_VERSION == 2 + +// the edge passed in here does not cross the vertical line at x or the vertical line at x+1 +// (i.e. it has already been clipped to those) +static void stbtt__handle_clipped_edge(float *scanline, int x, stbtt__active_edge *e, float x0, float y0, float x1, float y1) +{ + if (y0 == y1) return; + STBTT_assert(y0 < y1); + STBTT_assert(e->sy <= e->ey); + if (y0 > e->ey) return; + if (y1 < e->sy) return; + if (y0 < e->sy) { + x0 += (x1-x0) * (e->sy - y0) / (y1-y0); + y0 = e->sy; + } + if (y1 > e->ey) { + x1 += (x1-x0) * (e->ey - y1) / (y1-y0); + y1 = e->ey; + } + + if (x0 == x) + STBTT_assert(x1 <= x+1); + else if (x0 == x+1) + STBTT_assert(x1 >= x); + else if (x0 <= x) + STBTT_assert(x1 <= x); + else if (x0 >= x+1) + STBTT_assert(x1 >= x+1); + else + STBTT_assert(x1 >= x && x1 <= x+1); + + if (x0 <= x && x1 <= x) + scanline[x] += e->direction * (y1-y0); + else if (x0 >= x+1 && x1 >= x+1) + ; + else { + STBTT_assert(x0 >= x && x0 <= x+1 && x1 >= x && x1 <= x+1); + scanline[x] += e->direction * (y1-y0) * (1-((x0-x)+(x1-x))/2); // coverage = 1 - average x position + } +} + +static float stbtt__sized_trapezoid_area(float height, float top_width, float bottom_width) +{ + STBTT_assert(top_width >= 0); + STBTT_assert(bottom_width >= 0); + return (top_width + bottom_width) / 2.0f * height; +} + +static float stbtt__position_trapezoid_area(float height, float tx0, float tx1, float bx0, float bx1) +{ + return stbtt__sized_trapezoid_area(height, tx1 - tx0, bx1 - bx0); +} + +static float stbtt__sized_triangle_area(float height, float width) +{ + return height * width / 2; +} + +static void stbtt__fill_active_edges_new(float *scanline, float *scanline_fill, int len, stbtt__active_edge *e, float y_top) +{ + float y_bottom = y_top+1; + + while (e) { + // brute force every pixel + + // compute intersection points with top & bottom + STBTT_assert(e->ey >= y_top); + + if (e->fdx == 0) { + float x0 = e->fx; + if (x0 < len) { + if (x0 >= 0) { + stbtt__handle_clipped_edge(scanline,(int) x0,e, x0,y_top, x0,y_bottom); + stbtt__handle_clipped_edge(scanline_fill-1,(int) x0+1,e, x0,y_top, x0,y_bottom); + } else { + stbtt__handle_clipped_edge(scanline_fill-1,0,e, x0,y_top, x0,y_bottom); + } + } + } else { + float x0 = e->fx; + float dx = e->fdx; + float xb = x0 + dx; + float x_top, x_bottom; + float sy0,sy1; + float dy = e->fdy; + STBTT_assert(e->sy <= y_bottom && e->ey >= y_top); + + // compute endpoints of line segment clipped to this scanline (if the + // line segment starts on this scanline. x0 is the intersection of the + // line with y_top, but that may be off the line segment. + if (e->sy > y_top) { + x_top = x0 + dx * (e->sy - y_top); + sy0 = e->sy; + } else { + x_top = x0; + sy0 = y_top; + } + if (e->ey < y_bottom) { + x_bottom = x0 + dx * (e->ey - y_top); + sy1 = e->ey; + } else { + x_bottom = xb; + sy1 = y_bottom; + } + + if (x_top >= 0 && x_bottom >= 0 && x_top < len && x_bottom < len) { + // from here on, we don't have to range check x values + + if ((int) x_top == (int) x_bottom) { + float height; + // simple case, only spans one pixel + int x = (int) x_top; + height = (sy1 - sy0) * e->direction; + STBTT_assert(x >= 0 && x < len); + scanline[x] += stbtt__position_trapezoid_area(height, x_top, x+1.0f, x_bottom, x+1.0f); + scanline_fill[x] += height; // everything right of this pixel is filled + } else { + int x,x1,x2; + float y_crossing, y_final, step, sign, area; + // covers 2+ pixels + if (x_top > x_bottom) { + // flip scanline vertically; signed area is the same + float t; + sy0 = y_bottom - (sy0 - y_top); + sy1 = y_bottom - (sy1 - y_top); + t = sy0, sy0 = sy1, sy1 = t; + t = x_bottom, x_bottom = x_top, x_top = t; + dx = -dx; + dy = -dy; + t = x0, x0 = xb, xb = t; + } + STBTT_assert(dy >= 0); + STBTT_assert(dx >= 0); + + x1 = (int) x_top; + x2 = (int) x_bottom; + // compute intersection with y axis at x1+1 + y_crossing = y_top + dy * (x1+1 - x0); + + // compute intersection with y axis at x2 + y_final = y_top + dy * (x2 - x0); + + // x1 x_top x2 x_bottom + // y_top +------|-----+------------+------------+--------|---+------------+ + // | | | | | | + // | | | | | | + // sy0 | Txxxxx|............|............|............|............| + // y_crossing | *xxxxx.......|............|............|............| + // | | xxxxx..|............|............|............| + // | | /- xx*xxxx........|............|............| + // | | dy < | xxxxxx..|............|............| + // y_final | | \- | xx*xxx.........|............| + // sy1 | | | | xxxxxB...|............| + // | | | | | | + // | | | | | | + // y_bottom +------------+------------+------------+------------+------------+ + // + // goal is to measure the area covered by '.' in each pixel + + // if x2 is right at the right edge of x1, y_crossing can blow up, github #1057 + // @TODO: maybe test against sy1 rather than y_bottom? + if (y_crossing > y_bottom) + y_crossing = y_bottom; + + sign = e->direction; + + // area of the rectangle covered from sy0..y_crossing + area = sign * (y_crossing-sy0); + + // area of the triangle (x_top,sy0), (x1+1,sy0), (x1+1,y_crossing) + scanline[x1] += stbtt__sized_triangle_area(area, x1+1 - x_top); + + // check if final y_crossing is blown up; no test case for this + if (y_final > y_bottom) { + y_final = y_bottom; + dy = (y_final - y_crossing ) / (x2 - (x1+1)); // if denom=0, y_final = y_crossing, so y_final <= y_bottom + } + + // in second pixel, area covered by line segment found in first pixel + // is always a rectangle 1 wide * the height of that line segment; this + // is exactly what the variable 'area' stores. it also gets a contribution + // from the line segment within it. the THIRD pixel will get the first + // pixel's rectangle contribution, the second pixel's rectangle contribution, + // and its own contribution. the 'own contribution' is the same in every pixel except + // the leftmost and rightmost, a trapezoid that slides down in each pixel. + // the second pixel's contribution to the third pixel will be the + // rectangle 1 wide times the height change in the second pixel, which is dy. + + step = sign * dy * 1; // dy is dy/dx, change in y for every 1 change in x, + // which multiplied by 1-pixel-width is how much pixel area changes for each step in x + // so the area advances by 'step' every time + + for (x = x1+1; x < x2; ++x) { + scanline[x] += area + step/2; // area of trapezoid is 1*step/2 + area += step; + } + STBTT_assert(STBTT_fabs(area) <= 1.01f); // accumulated error from area += step unless we round step down + STBTT_assert(sy1 > y_final-0.01f); + + // area covered in the last pixel is the rectangle from all the pixels to the left, + // plus the trapezoid filled by the line segment in this pixel all the way to the right edge + scanline[x2] += area + sign * stbtt__position_trapezoid_area(sy1-y_final, (float) x2, x2+1.0f, x_bottom, x2+1.0f); + + // the rest of the line is filled based on the total height of the line segment in this pixel + scanline_fill[x2] += sign * (sy1-sy0); + } + } else { + // if edge goes outside of box we're drawing, we require + // clipping logic. since this does not match the intended use + // of this library, we use a different, very slow brute + // force implementation + // note though that this does happen some of the time because + // x_top and x_bottom can be extrapolated at the top & bottom of + // the shape and actually lie outside the bounding box + int x; + for (x=0; x < len; ++x) { + // cases: + // + // there can be up to two intersections with the pixel. any intersection + // with left or right edges can be handled by splitting into two (or three) + // regions. intersections with top & bottom do not necessitate case-wise logic. + // + // the old way of doing this found the intersections with the left & right edges, + // then used some simple logic to produce up to three segments in sorted order + // from top-to-bottom. however, this had a problem: if an x edge was epsilon + // across the x border, then the corresponding y position might not be distinct + // from the other y segment, and it might ignored as an empty segment. to avoid + // that, we need to explicitly produce segments based on x positions. + + // rename variables to clearly-defined pairs + float y0 = y_top; + float x1 = (float) (x); + float x2 = (float) (x+1); + float x3 = xb; + float y3 = y_bottom; + + // x = e->x + e->dx * (y-y_top) + // (y-y_top) = (x - e->x) / e->dx + // y = (x - e->x) / e->dx + y_top + float y1 = (x - x0) / dx + y_top; + float y2 = (x+1 - x0) / dx + y_top; + + if (x0 < x1 && x3 > x2) { // three segments descending down-right + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); + stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x2,y2); + stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); + } else if (x3 < x1 && x0 > x2) { // three segments descending down-left + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); + stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x1,y1); + stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); + } else if (x0 < x1 && x3 > x1) { // two segments across x, down-right + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); + stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); + } else if (x3 < x1 && x0 > x1) { // two segments across x, down-left + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); + stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); + } else if (x0 < x2 && x3 > x2) { // two segments across x+1, down-right + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); + stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); + } else if (x3 < x2 && x0 > x2) { // two segments across x+1, down-left + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); + stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); + } else { // one segment + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x3,y3); + } + } + } + } + e = e->next; + } +} + +// directly AA rasterize edges w/o supersampling +static void stbtt__rasterize_sorted_edges(stbtt__bitmap *result, stbtt__edge *e, int n, int vsubsample, int off_x, int off_y, void *userdata) +{ + stbtt__hheap hh = { 0, 0, 0 }; + stbtt__active_edge *active = NULL; + int y,j=0, i; + float scanline_data[129], *scanline, *scanline2; + + STBTT__NOTUSED(vsubsample); + + if (result->w > 64) + scanline = (float *) STBTT_malloc((result->w*2+1) * sizeof(float), userdata); + else + scanline = scanline_data; + + scanline2 = scanline + result->w; + + y = off_y; + e[n].y0 = (float) (off_y + result->h) + 1; + + while (j < result->h) { + // find center of pixel for this scanline + float scan_y_top = y + 0.0f; + float scan_y_bottom = y + 1.0f; + stbtt__active_edge **step = &active; + + STBTT_memset(scanline , 0, result->w*sizeof(scanline[0])); + STBTT_memset(scanline2, 0, (result->w+1)*sizeof(scanline[0])); + + // update all active edges; + // remove all active edges that terminate before the top of this scanline + while (*step) { + stbtt__active_edge * z = *step; + if (z->ey <= scan_y_top) { + *step = z->next; // delete from list + STBTT_assert(z->direction); + z->direction = 0; + stbtt__hheap_free(&hh, z); + } else { + step = &((*step)->next); // advance through list + } + } + + // insert all edges that start before the bottom of this scanline + while (e->y0 <= scan_y_bottom) { + if (e->y0 != e->y1) { + stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y_top, userdata); + if (z != NULL) { + if (j == 0 && off_y != 0) { + if (z->ey < scan_y_top) { + // this can happen due to subpixel positioning and some kind of fp rounding error i think + z->ey = scan_y_top; + } + } + STBTT_assert(z->ey >= scan_y_top); // if we get really unlucky a tiny bit of an edge can be out of bounds + // insert at front + z->next = active; + active = z; + } + } + ++e; + } + + // now process all active edges + if (active) + stbtt__fill_active_edges_new(scanline, scanline2+1, result->w, active, scan_y_top); + + { + float sum = 0; + for (i=0; i < result->w; ++i) { + float k; + int m; + sum += scanline2[i]; + k = scanline[i] + sum; + k = (float) STBTT_fabs(k)*255 + 0.5f; + m = (int) k; + if (m > 255) m = 255; + result->pixels[j*result->stride + i] = (unsigned char) m; + } + } + // advance all the edges + step = &active; + while (*step) { + stbtt__active_edge *z = *step; + z->fx += z->fdx; // advance to position for current scanline + step = &((*step)->next); // advance through list + } + + ++y; + ++j; + } + + stbtt__hheap_cleanup(&hh, userdata); + + if (scanline != scanline_data) + STBTT_free(scanline, userdata); +} +#else +#error "Unrecognized value of STBTT_RASTERIZER_VERSION" +#endif + +#define STBTT__COMPARE(a,b) ((a)->y0 < (b)->y0) + +static void stbtt__sort_edges_ins_sort(stbtt__edge *p, int n) +{ + int i,j; + for (i=1; i < n; ++i) { + stbtt__edge t = p[i], *a = &t; + j = i; + while (j > 0) { + stbtt__edge *b = &p[j-1]; + int c = STBTT__COMPARE(a,b); + if (!c) break; + p[j] = p[j-1]; + --j; + } + if (i != j) + p[j] = t; + } +} + +static void stbtt__sort_edges_quicksort(stbtt__edge *p, int n) +{ + /* threshold for transitioning to insertion sort */ + while (n > 12) { + stbtt__edge t; + int c01,c12,c,m,i,j; + + /* compute median of three */ + m = n >> 1; + c01 = STBTT__COMPARE(&p[0],&p[m]); + c12 = STBTT__COMPARE(&p[m],&p[n-1]); + /* if 0 >= mid >= end, or 0 < mid < end, then use mid */ + if (c01 != c12) { + /* otherwise, we'll need to swap something else to middle */ + int z; + c = STBTT__COMPARE(&p[0],&p[n-1]); + /* 0>mid && midn => n; 0 0 */ + /* 0n: 0>n => 0; 0 n */ + z = (c == c12) ? 0 : n-1; + t = p[z]; + p[z] = p[m]; + p[m] = t; + } + /* now p[m] is the median-of-three */ + /* swap it to the beginning so it won't move around */ + t = p[0]; + p[0] = p[m]; + p[m] = t; + + /* partition loop */ + i=1; + j=n-1; + for(;;) { + /* handling of equality is crucial here */ + /* for sentinels & efficiency with duplicates */ + for (;;++i) { + if (!STBTT__COMPARE(&p[i], &p[0])) break; + } + for (;;--j) { + if (!STBTT__COMPARE(&p[0], &p[j])) break; + } + /* make sure we haven't crossed */ + if (i >= j) break; + t = p[i]; + p[i] = p[j]; + p[j] = t; + + ++i; + --j; + } + /* recurse on smaller side, iterate on larger */ + if (j < (n-i)) { + stbtt__sort_edges_quicksort(p,j); + p = p+i; + n = n-i; + } else { + stbtt__sort_edges_quicksort(p+i, n-i); + n = j; + } + } +} + +static void stbtt__sort_edges(stbtt__edge *p, int n) +{ + stbtt__sort_edges_quicksort(p, n); + stbtt__sort_edges_ins_sort(p, n); +} + +typedef struct +{ + float x,y; +} stbtt__point; + +static void stbtt__rasterize(stbtt__bitmap *result, stbtt__point *pts, int *wcount, int windings, float scale_x, float scale_y, float shift_x, float shift_y, int off_x, int off_y, int invert, void *userdata) +{ + float y_scale_inv = invert ? -scale_y : scale_y; + stbtt__edge *e; + int n,i,j,k,m; +#if STBTT_RASTERIZER_VERSION == 1 + int vsubsample = result->h < 8 ? 15 : 5; +#elif STBTT_RASTERIZER_VERSION == 2 + int vsubsample = 1; +#else + #error "Unrecognized value of STBTT_RASTERIZER_VERSION" +#endif + // vsubsample should divide 255 evenly; otherwise we won't reach full opacity + + // now we have to blow out the windings into explicit edge lists + n = 0; + for (i=0; i < windings; ++i) + n += wcount[i]; + + e = (stbtt__edge *) STBTT_malloc(sizeof(*e) * (n+1), userdata); // add an extra one as a sentinel + if (e == 0) return; + n = 0; + + m=0; + for (i=0; i < windings; ++i) { + stbtt__point *p = pts + m; + m += wcount[i]; + j = wcount[i]-1; + for (k=0; k < wcount[i]; j=k++) { + int a=k,b=j; + // skip the edge if horizontal + if (p[j].y == p[k].y) + continue; + // add edge from j to k to the list + e[n].invert = 0; + if (invert ? p[j].y > p[k].y : p[j].y < p[k].y) { + e[n].invert = 1; + a=j,b=k; + } + e[n].x0 = p[a].x * scale_x + shift_x; + e[n].y0 = (p[a].y * y_scale_inv + shift_y) * vsubsample; + e[n].x1 = p[b].x * scale_x + shift_x; + e[n].y1 = (p[b].y * y_scale_inv + shift_y) * vsubsample; + ++n; + } + } + + // now sort the edges by their highest point (should snap to integer, and then by x) + //STBTT_sort(e, n, sizeof(e[0]), stbtt__edge_compare); + stbtt__sort_edges(e, n); + + // now, traverse the scanlines and find the intersections on each scanline, use xor winding rule + stbtt__rasterize_sorted_edges(result, e, n, vsubsample, off_x, off_y, userdata); + + STBTT_free(e, userdata); +} + +static void stbtt__add_point(stbtt__point *points, int n, float x, float y) +{ + if (!points) return; // during first pass, it's unallocated + points[n].x = x; + points[n].y = y; +} + +// tessellate until threshold p is happy... @TODO warped to compensate for non-linear stretching +static int stbtt__tesselate_curve(stbtt__point *points, int *num_points, float x0, float y0, float x1, float y1, float x2, float y2, float objspace_flatness_squared, int n) +{ + // midpoint + float mx = (x0 + 2*x1 + x2)/4; + float my = (y0 + 2*y1 + y2)/4; + // versus directly drawn line + float dx = (x0+x2)/2 - mx; + float dy = (y0+y2)/2 - my; + if (n > 16) // 65536 segments on one curve better be enough! + return 1; + if (dx*dx+dy*dy > objspace_flatness_squared) { // half-pixel error allowed... need to be smaller if AA + stbtt__tesselate_curve(points, num_points, x0,y0, (x0+x1)/2.0f,(y0+y1)/2.0f, mx,my, objspace_flatness_squared,n+1); + stbtt__tesselate_curve(points, num_points, mx,my, (x1+x2)/2.0f,(y1+y2)/2.0f, x2,y2, objspace_flatness_squared,n+1); + } else { + stbtt__add_point(points, *num_points,x2,y2); + *num_points = *num_points+1; + } + return 1; +} + +static void stbtt__tesselate_cubic(stbtt__point *points, int *num_points, float x0, float y0, float x1, float y1, float x2, float y2, float x3, float y3, float objspace_flatness_squared, int n) +{ + // @TODO this "flatness" calculation is just made-up nonsense that seems to work well enough + float dx0 = x1-x0; + float dy0 = y1-y0; + float dx1 = x2-x1; + float dy1 = y2-y1; + float dx2 = x3-x2; + float dy2 = y3-y2; + float dx = x3-x0; + float dy = y3-y0; + float longlen = (float) (STBTT_sqrt(dx0*dx0+dy0*dy0)+STBTT_sqrt(dx1*dx1+dy1*dy1)+STBTT_sqrt(dx2*dx2+dy2*dy2)); + float shortlen = (float) STBTT_sqrt(dx*dx+dy*dy); + float flatness_squared = longlen*longlen-shortlen*shortlen; + + if (n > 16) // 65536 segments on one curve better be enough! + return; + + if (flatness_squared > objspace_flatness_squared) { + float x01 = (x0+x1)/2; + float y01 = (y0+y1)/2; + float x12 = (x1+x2)/2; + float y12 = (y1+y2)/2; + float x23 = (x2+x3)/2; + float y23 = (y2+y3)/2; + + float xa = (x01+x12)/2; + float ya = (y01+y12)/2; + float xb = (x12+x23)/2; + float yb = (y12+y23)/2; + + float mx = (xa+xb)/2; + float my = (ya+yb)/2; + + stbtt__tesselate_cubic(points, num_points, x0,y0, x01,y01, xa,ya, mx,my, objspace_flatness_squared,n+1); + stbtt__tesselate_cubic(points, num_points, mx,my, xb,yb, x23,y23, x3,y3, objspace_flatness_squared,n+1); + } else { + stbtt__add_point(points, *num_points,x3,y3); + *num_points = *num_points+1; + } +} + +// returns number of contours +static stbtt__point *stbtt_FlattenCurves(stbtt_vertex *vertices, int num_verts, float objspace_flatness, int **contour_lengths, int *num_contours, void *userdata) +{ + stbtt__point *points=0; + int num_points=0; + + float objspace_flatness_squared = objspace_flatness * objspace_flatness; + int i,n=0,start=0, pass; + + // count how many "moves" there are to get the contour count + for (i=0; i < num_verts; ++i) + if (vertices[i].type == STBTT_vmove) + ++n; + + *num_contours = n; + if (n == 0) return 0; + + *contour_lengths = (int *) STBTT_malloc(sizeof(**contour_lengths) * n, userdata); + + if (*contour_lengths == 0) { + *num_contours = 0; + return 0; + } + + // make two passes through the points so we don't need to realloc + for (pass=0; pass < 2; ++pass) { + float x=0,y=0; + if (pass == 1) { + points = (stbtt__point *) STBTT_malloc(num_points * sizeof(points[0]), userdata); + if (points == NULL) goto error; + } + num_points = 0; + n= -1; + for (i=0; i < num_verts; ++i) { + switch (vertices[i].type) { + case STBTT_vmove: + // start the next contour + if (n >= 0) + (*contour_lengths)[n] = num_points - start; + ++n; + start = num_points; + + x = vertices[i].x, y = vertices[i].y; + stbtt__add_point(points, num_points++, x,y); + break; + case STBTT_vline: + x = vertices[i].x, y = vertices[i].y; + stbtt__add_point(points, num_points++, x, y); + break; + case STBTT_vcurve: + stbtt__tesselate_curve(points, &num_points, x,y, + vertices[i].cx, vertices[i].cy, + vertices[i].x, vertices[i].y, + objspace_flatness_squared, 0); + x = vertices[i].x, y = vertices[i].y; + break; + case STBTT_vcubic: + stbtt__tesselate_cubic(points, &num_points, x,y, + vertices[i].cx, vertices[i].cy, + vertices[i].cx1, vertices[i].cy1, + vertices[i].x, vertices[i].y, + objspace_flatness_squared, 0); + x = vertices[i].x, y = vertices[i].y; + break; + } + } + (*contour_lengths)[n] = num_points - start; + } + + return points; +error: + STBTT_free(points, userdata); + STBTT_free(*contour_lengths, userdata); + *contour_lengths = 0; + *num_contours = 0; + return NULL; +} + +STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result, float flatness_in_pixels, stbtt_vertex *vertices, int num_verts, float scale_x, float scale_y, float shift_x, float shift_y, int x_off, int y_off, int invert, void *userdata) +{ + float scale = scale_x > scale_y ? scale_y : scale_x; + int winding_count = 0; + int *winding_lengths = NULL; + stbtt__point *windings = stbtt_FlattenCurves(vertices, num_verts, flatness_in_pixels / scale, &winding_lengths, &winding_count, userdata); + if (windings) { + stbtt__rasterize(result, windings, winding_lengths, winding_count, scale_x, scale_y, shift_x, shift_y, x_off, y_off, invert, userdata); + STBTT_free(winding_lengths, userdata); + STBTT_free(windings, userdata); + } +} + +STBTT_DEF void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata) +{ + STBTT_free(bitmap, userdata); +} + +STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff) +{ + int ix0,iy0,ix1,iy1; + stbtt__bitmap gbm; + stbtt_vertex *vertices; + int num_verts = stbtt_GetGlyphShape(info, glyph, &vertices); + + if (scale_x == 0) scale_x = scale_y; + if (scale_y == 0) { + if (scale_x == 0) { + STBTT_free(vertices, info->userdata); + return NULL; + } + scale_y = scale_x; + } + + stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale_x, scale_y, shift_x, shift_y, &ix0,&iy0,&ix1,&iy1); + + // now we get the size + gbm.w = (ix1 - ix0); + gbm.h = (iy1 - iy0); + gbm.pixels = NULL; // in case we error + + if (width ) *width = gbm.w; + if (height) *height = gbm.h; + if (xoff ) *xoff = ix0; + if (yoff ) *yoff = iy0; + + if (gbm.w && gbm.h) { + gbm.pixels = (unsigned char *) STBTT_malloc(gbm.w * gbm.h, info->userdata); + if (gbm.pixels) { + gbm.stride = gbm.w; + + stbtt_Rasterize(&gbm, 0.35f, vertices, num_verts, scale_x, scale_y, shift_x, shift_y, ix0, iy0, 1, info->userdata); + } + } + STBTT_free(vertices, info->userdata); + return gbm.pixels; +} + +STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff) +{ + return stbtt_GetGlyphBitmapSubpixel(info, scale_x, scale_y, 0.0f, 0.0f, glyph, width, height, xoff, yoff); +} + +STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph) +{ + int ix0,iy0; + stbtt_vertex *vertices; + int num_verts = stbtt_GetGlyphShape(info, glyph, &vertices); + stbtt__bitmap gbm; + + stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale_x, scale_y, shift_x, shift_y, &ix0,&iy0,0,0); + gbm.pixels = output; + gbm.w = out_w; + gbm.h = out_h; + gbm.stride = out_stride; + + if (gbm.w && gbm.h) + stbtt_Rasterize(&gbm, 0.35f, vertices, num_verts, scale_x, scale_y, shift_x, shift_y, ix0,iy0, 1, info->userdata); + + STBTT_free(vertices, info->userdata); +} + +STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph) +{ + stbtt_MakeGlyphBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, 0.0f,0.0f, glyph); +} + +STBTT_DEF unsigned char *stbtt_GetCodepointBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint, int *width, int *height, int *xoff, int *yoff) +{ + return stbtt_GetGlyphBitmapSubpixel(info, scale_x, scale_y,shift_x,shift_y, stbtt_FindGlyphIndex(info,codepoint), width,height,xoff,yoff); +} + +STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint) +{ + stbtt_MakeGlyphBitmapSubpixelPrefilter(info, output, out_w, out_h, out_stride, scale_x, scale_y, shift_x, shift_y, oversample_x, oversample_y, sub_x, sub_y, stbtt_FindGlyphIndex(info,codepoint)); +} + +STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint) +{ + stbtt_MakeGlyphBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, shift_x, shift_y, stbtt_FindGlyphIndex(info,codepoint)); +} + +STBTT_DEF unsigned char *stbtt_GetCodepointBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int codepoint, int *width, int *height, int *xoff, int *yoff) +{ + return stbtt_GetCodepointBitmapSubpixel(info, scale_x, scale_y, 0.0f,0.0f, codepoint, width,height,xoff,yoff); +} + +STBTT_DEF void stbtt_MakeCodepointBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int codepoint) +{ + stbtt_MakeCodepointBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, 0.0f,0.0f, codepoint); +} + +////////////////////////////////////////////////////////////////////////////// +// +// bitmap baking +// +// This is SUPER-CRAPPY packing to keep source code small + +static int stbtt_BakeFontBitmap_internal(unsigned char *data, int offset, // font location (use offset=0 for plain .ttf) + float pixel_height, // height of font in pixels + unsigned char *pixels, int pw, int ph, // bitmap to be filled in + int first_char, int num_chars, // characters to bake + stbtt_bakedchar *chardata) +{ + float scale; + int x,y,bottom_y, i; + stbtt_fontinfo f; + f.userdata = NULL; + if (!stbtt_InitFont(&f, data, offset)) + return -1; + STBTT_memset(pixels, 0, pw*ph); // background of 0 around pixels + x=y=1; + bottom_y = 1; + + scale = stbtt_ScaleForPixelHeight(&f, pixel_height); + + for (i=0; i < num_chars; ++i) { + int advance, lsb, x0,y0,x1,y1,gw,gh; + int g = stbtt_FindGlyphIndex(&f, first_char + i); + stbtt_GetGlyphHMetrics(&f, g, &advance, &lsb); + stbtt_GetGlyphBitmapBox(&f, g, scale,scale, &x0,&y0,&x1,&y1); + gw = x1-x0; + gh = y1-y0; + if (x + gw + 1 >= pw) + y = bottom_y, x = 1; // advance to next row + if (y + gh + 1 >= ph) // check if it fits vertically AFTER potentially moving to next row + return -i; + STBTT_assert(x+gw < pw); + STBTT_assert(y+gh < ph); + stbtt_MakeGlyphBitmap(&f, pixels+x+y*pw, gw,gh,pw, scale,scale, g); + chardata[i].x0 = (stbtt_int16) x; + chardata[i].y0 = (stbtt_int16) y; + chardata[i].x1 = (stbtt_int16) (x + gw); + chardata[i].y1 = (stbtt_int16) (y + gh); + chardata[i].xadvance = scale * advance; + chardata[i].xoff = (float) x0; + chardata[i].yoff = (float) y0; + x = x + gw + 1; + if (y+gh+1 > bottom_y) + bottom_y = y+gh+1; + } + return bottom_y; +} + +STBTT_DEF void stbtt_GetBakedQuad(const stbtt_bakedchar *chardata, int pw, int ph, int char_index, float *xpos, float *ypos, stbtt_aligned_quad *q, int opengl_fillrule) +{ + float d3d_bias = opengl_fillrule ? 0 : -0.5f; + float ipw = 1.0f / pw, iph = 1.0f / ph; + const stbtt_bakedchar *b = chardata + char_index; + int round_x = STBTT_ifloor((*xpos + b->xoff) + 0.5f); + int round_y = STBTT_ifloor((*ypos + b->yoff) + 0.5f); + + q->x0 = round_x + d3d_bias; + q->y0 = round_y + d3d_bias; + q->x1 = round_x + b->x1 - b->x0 + d3d_bias; + q->y1 = round_y + b->y1 - b->y0 + d3d_bias; + + q->s0 = b->x0 * ipw; + q->t0 = b->y0 * iph; + q->s1 = b->x1 * ipw; + q->t1 = b->y1 * iph; + + *xpos += b->xadvance; +} + +////////////////////////////////////////////////////////////////////////////// +// +// rectangle packing replacement routines if you don't have stb_rect_pack.h +// + +#ifndef STB_RECT_PACK_VERSION + +typedef int stbrp_coord; + +//////////////////////////////////////////////////////////////////////////////////// +// // +// // +// COMPILER WARNING ?!?!? // +// // +// // +// if you get a compile warning due to these symbols being defined more than // +// once, move #include "stb_rect_pack.h" before #include "stb_truetype.h" // +// // +//////////////////////////////////////////////////////////////////////////////////// + +typedef struct +{ + int width,height; + int x,y,bottom_y; +} stbrp_context; + +typedef struct +{ + unsigned char x; +} stbrp_node; + +struct stbrp_rect +{ + stbrp_coord x,y; + int id,w,h,was_packed; +}; + +static void stbrp_init_target(stbrp_context *con, int pw, int ph, stbrp_node *nodes, int num_nodes) +{ + con->width = pw; + con->height = ph; + con->x = 0; + con->y = 0; + con->bottom_y = 0; + STBTT__NOTUSED(nodes); + STBTT__NOTUSED(num_nodes); +} + +static void stbrp_pack_rects(stbrp_context *con, stbrp_rect *rects, int num_rects) +{ + int i; + for (i=0; i < num_rects; ++i) { + if (con->x + rects[i].w > con->width) { + con->x = 0; + con->y = con->bottom_y; + } + if (con->y + rects[i].h > con->height) + break; + rects[i].x = con->x; + rects[i].y = con->y; + rects[i].was_packed = 1; + con->x += rects[i].w; + if (con->y + rects[i].h > con->bottom_y) + con->bottom_y = con->y + rects[i].h; + } + for ( ; i < num_rects; ++i) + rects[i].was_packed = 0; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// bitmap baking +// +// This is SUPER-AWESOME (tm Ryan Gordon) packing using stb_rect_pack.h. If +// stb_rect_pack.h isn't available, it uses the BakeFontBitmap strategy. + +STBTT_DEF int stbtt_PackBegin(stbtt_pack_context *spc, unsigned char *pixels, int pw, int ph, int stride_in_bytes, int padding, void *alloc_context) +{ + stbrp_context *context = (stbrp_context *) STBTT_malloc(sizeof(*context) ,alloc_context); + int num_nodes = pw - padding; + stbrp_node *nodes = (stbrp_node *) STBTT_malloc(sizeof(*nodes ) * num_nodes,alloc_context); + + if (context == NULL || nodes == NULL) { + if (context != NULL) STBTT_free(context, alloc_context); + if (nodes != NULL) STBTT_free(nodes , alloc_context); + return 0; + } + + spc->user_allocator_context = alloc_context; + spc->width = pw; + spc->height = ph; + spc->pixels = pixels; + spc->pack_info = context; + spc->nodes = nodes; + spc->padding = padding; + spc->stride_in_bytes = stride_in_bytes != 0 ? stride_in_bytes : pw; + spc->h_oversample = 1; + spc->v_oversample = 1; + spc->skip_missing = 0; + + stbrp_init_target(context, pw-padding, ph-padding, nodes, num_nodes); + + if (pixels) + STBTT_memset(pixels, 0, pw*ph); // background of 0 around pixels + + return 1; +} + +STBTT_DEF void stbtt_PackEnd (stbtt_pack_context *spc) +{ + STBTT_free(spc->nodes , spc->user_allocator_context); + STBTT_free(spc->pack_info, spc->user_allocator_context); +} + +STBTT_DEF void stbtt_PackSetOversampling(stbtt_pack_context *spc, unsigned int h_oversample, unsigned int v_oversample) +{ + STBTT_assert(h_oversample <= STBTT_MAX_OVERSAMPLE); + STBTT_assert(v_oversample <= STBTT_MAX_OVERSAMPLE); + if (h_oversample <= STBTT_MAX_OVERSAMPLE) + spc->h_oversample = h_oversample; + if (v_oversample <= STBTT_MAX_OVERSAMPLE) + spc->v_oversample = v_oversample; +} + +STBTT_DEF void stbtt_PackSetSkipMissingCodepoints(stbtt_pack_context *spc, int skip) +{ + spc->skip_missing = skip; +} + +#define STBTT__OVER_MASK (STBTT_MAX_OVERSAMPLE-1) + +static void stbtt__h_prefilter(unsigned char *pixels, int w, int h, int stride_in_bytes, unsigned int kernel_width) +{ + unsigned char buffer[STBTT_MAX_OVERSAMPLE]; + int safe_w = w - kernel_width; + int j; + STBTT_memset(buffer, 0, STBTT_MAX_OVERSAMPLE); // suppress bogus warning from VS2013 -analyze + for (j=0; j < h; ++j) { + int i; + unsigned int total; + STBTT_memset(buffer, 0, kernel_width); + + total = 0; + + // make kernel_width a constant in common cases so compiler can optimize out the divide + switch (kernel_width) { + case 2: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / 2); + } + break; + case 3: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / 3); + } + break; + case 4: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / 4); + } + break; + case 5: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / 5); + } + break; + default: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / kernel_width); + } + break; + } + + for (; i < w; ++i) { + STBTT_assert(pixels[i] == 0); + total -= buffer[i & STBTT__OVER_MASK]; + pixels[i] = (unsigned char) (total / kernel_width); + } + + pixels += stride_in_bytes; + } +} + +static void stbtt__v_prefilter(unsigned char *pixels, int w, int h, int stride_in_bytes, unsigned int kernel_width) +{ + unsigned char buffer[STBTT_MAX_OVERSAMPLE]; + int safe_h = h - kernel_width; + int j; + STBTT_memset(buffer, 0, STBTT_MAX_OVERSAMPLE); // suppress bogus warning from VS2013 -analyze + for (j=0; j < w; ++j) { + int i; + unsigned int total; + STBTT_memset(buffer, 0, kernel_width); + + total = 0; + + // make kernel_width a constant in common cases so compiler can optimize out the divide + switch (kernel_width) { + case 2: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / 2); + } + break; + case 3: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / 3); + } + break; + case 4: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / 4); + } + break; + case 5: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / 5); + } + break; + default: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / kernel_width); + } + break; + } + + for (; i < h; ++i) { + STBTT_assert(pixels[i*stride_in_bytes] == 0); + total -= buffer[i & STBTT__OVER_MASK]; + pixels[i*stride_in_bytes] = (unsigned char) (total / kernel_width); + } + + pixels += 1; + } +} + +static float stbtt__oversample_shift(int oversample) +{ + if (!oversample) + return 0.0f; + + // The prefilter is a box filter of width "oversample", + // which shifts phase by (oversample - 1)/2 pixels in + // oversampled space. We want to shift in the opposite + // direction to counter this. + return (float)-(oversample - 1) / (2.0f * (float)oversample); +} + +// rects array must be big enough to accommodate all characters in the given ranges +STBTT_DEF int stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects) +{ + int i,j,k; + int missing_glyph_added = 0; + + k=0; + for (i=0; i < num_ranges; ++i) { + float fh = ranges[i].font_size; + float scale = fh > 0 ? stbtt_ScaleForPixelHeight(info, fh) : stbtt_ScaleForMappingEmToPixels(info, -fh); + ranges[i].h_oversample = (unsigned char) spc->h_oversample; + ranges[i].v_oversample = (unsigned char) spc->v_oversample; + for (j=0; j < ranges[i].num_chars; ++j) { + int x0,y0,x1,y1; + int codepoint = ranges[i].array_of_unicode_codepoints == NULL ? ranges[i].first_unicode_codepoint_in_range + j : ranges[i].array_of_unicode_codepoints[j]; + int glyph = stbtt_FindGlyphIndex(info, codepoint); + if (glyph == 0 && (spc->skip_missing || missing_glyph_added)) { + rects[k].w = rects[k].h = 0; + } else { + stbtt_GetGlyphBitmapBoxSubpixel(info,glyph, + scale * spc->h_oversample, + scale * spc->v_oversample, + 0,0, + &x0,&y0,&x1,&y1); + rects[k].w = (stbrp_coord) (x1-x0 + spc->padding + spc->h_oversample-1); + rects[k].h = (stbrp_coord) (y1-y0 + spc->padding + spc->v_oversample-1); + if (glyph == 0) + missing_glyph_added = 1; + } + ++k; + } + } + + return k; +} + +STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int prefilter_x, int prefilter_y, float *sub_x, float *sub_y, int glyph) +{ + stbtt_MakeGlyphBitmapSubpixel(info, + output, + out_w - (prefilter_x - 1), + out_h - (prefilter_y - 1), + out_stride, + scale_x, + scale_y, + shift_x, + shift_y, + glyph); + + if (prefilter_x > 1) + stbtt__h_prefilter(output, out_w, out_h, out_stride, prefilter_x); + + if (prefilter_y > 1) + stbtt__v_prefilter(output, out_w, out_h, out_stride, prefilter_y); + + *sub_x = stbtt__oversample_shift(prefilter_x); + *sub_y = stbtt__oversample_shift(prefilter_y); +} + +// rects array must be big enough to accommodate all characters in the given ranges +STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects) +{ + int i,j,k, missing_glyph = -1, return_value = 1; + + // save current values + int old_h_over = spc->h_oversample; + int old_v_over = spc->v_oversample; + + k = 0; + for (i=0; i < num_ranges; ++i) { + float fh = ranges[i].font_size; + float scale = fh > 0 ? stbtt_ScaleForPixelHeight(info, fh) : stbtt_ScaleForMappingEmToPixels(info, -fh); + float recip_h,recip_v,sub_x,sub_y; + spc->h_oversample = ranges[i].h_oversample; + spc->v_oversample = ranges[i].v_oversample; + recip_h = 1.0f / spc->h_oversample; + recip_v = 1.0f / spc->v_oversample; + sub_x = stbtt__oversample_shift(spc->h_oversample); + sub_y = stbtt__oversample_shift(spc->v_oversample); + for (j=0; j < ranges[i].num_chars; ++j) { + stbrp_rect *r = &rects[k]; + if (r->was_packed && r->w != 0 && r->h != 0) { + stbtt_packedchar *bc = &ranges[i].chardata_for_range[j]; + int advance, lsb, x0,y0,x1,y1; + int codepoint = ranges[i].array_of_unicode_codepoints == NULL ? ranges[i].first_unicode_codepoint_in_range + j : ranges[i].array_of_unicode_codepoints[j]; + int glyph = stbtt_FindGlyphIndex(info, codepoint); + stbrp_coord pad = (stbrp_coord) spc->padding; + + // pad on left and top + r->x += pad; + r->y += pad; + r->w -= pad; + r->h -= pad; + stbtt_GetGlyphHMetrics(info, glyph, &advance, &lsb); + stbtt_GetGlyphBitmapBox(info, glyph, + scale * spc->h_oversample, + scale * spc->v_oversample, + &x0,&y0,&x1,&y1); + stbtt_MakeGlyphBitmapSubpixel(info, + spc->pixels + r->x + r->y*spc->stride_in_bytes, + r->w - spc->h_oversample+1, + r->h - spc->v_oversample+1, + spc->stride_in_bytes, + scale * spc->h_oversample, + scale * spc->v_oversample, + 0,0, + glyph); + + if (spc->h_oversample > 1) + stbtt__h_prefilter(spc->pixels + r->x + r->y*spc->stride_in_bytes, + r->w, r->h, spc->stride_in_bytes, + spc->h_oversample); + + if (spc->v_oversample > 1) + stbtt__v_prefilter(spc->pixels + r->x + r->y*spc->stride_in_bytes, + r->w, r->h, spc->stride_in_bytes, + spc->v_oversample); + + bc->x0 = (stbtt_int16) r->x; + bc->y0 = (stbtt_int16) r->y; + bc->x1 = (stbtt_int16) (r->x + r->w); + bc->y1 = (stbtt_int16) (r->y + r->h); + bc->xadvance = scale * advance; + bc->xoff = (float) x0 * recip_h + sub_x; + bc->yoff = (float) y0 * recip_v + sub_y; + bc->xoff2 = (x0 + r->w) * recip_h + sub_x; + bc->yoff2 = (y0 + r->h) * recip_v + sub_y; + + if (glyph == 0) + missing_glyph = j; + } else if (spc->skip_missing) { + return_value = 0; + } else if (r->was_packed && r->w == 0 && r->h == 0 && missing_glyph >= 0) { + ranges[i].chardata_for_range[j] = ranges[i].chardata_for_range[missing_glyph]; + } else { + return_value = 0; // if any fail, report failure + } + + ++k; + } + } + + // restore original values + spc->h_oversample = old_h_over; + spc->v_oversample = old_v_over; + + return return_value; +} + +STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect *rects, int num_rects) +{ + stbrp_pack_rects((stbrp_context *) spc->pack_info, rects, num_rects); +} + +STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges) +{ + stbtt_fontinfo info; + int i,j,n, return_value = 1; + //stbrp_context *context = (stbrp_context *) spc->pack_info; + stbrp_rect *rects; + + // flag all characters as NOT packed + for (i=0; i < num_ranges; ++i) + for (j=0; j < ranges[i].num_chars; ++j) + ranges[i].chardata_for_range[j].x0 = + ranges[i].chardata_for_range[j].y0 = + ranges[i].chardata_for_range[j].x1 = + ranges[i].chardata_for_range[j].y1 = 0; + + n = 0; + for (i=0; i < num_ranges; ++i) + n += ranges[i].num_chars; + + rects = (stbrp_rect *) STBTT_malloc(sizeof(*rects) * n, spc->user_allocator_context); + if (rects == NULL) + return 0; + + info.userdata = spc->user_allocator_context; + stbtt_InitFont(&info, fontdata, stbtt_GetFontOffsetForIndex(fontdata,font_index)); + + n = stbtt_PackFontRangesGatherRects(spc, &info, ranges, num_ranges, rects); + + stbtt_PackFontRangesPackRects(spc, rects, n); + + return_value = stbtt_PackFontRangesRenderIntoRects(spc, &info, ranges, num_ranges, rects); + + STBTT_free(rects, spc->user_allocator_context); + return return_value; +} + +STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size, + int first_unicode_codepoint_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range) +{ + stbtt_pack_range range; + range.first_unicode_codepoint_in_range = first_unicode_codepoint_in_range; + range.array_of_unicode_codepoints = NULL; + range.num_chars = num_chars_in_range; + range.chardata_for_range = chardata_for_range; + range.font_size = font_size; + return stbtt_PackFontRanges(spc, fontdata, font_index, &range, 1); +} + +STBTT_DEF void stbtt_GetScaledFontVMetrics(const unsigned char *fontdata, int index, float size, float *ascent, float *descent, float *lineGap) +{ + int i_ascent, i_descent, i_lineGap; + float scale; + stbtt_fontinfo info; + stbtt_InitFont(&info, fontdata, stbtt_GetFontOffsetForIndex(fontdata, index)); + scale = size > 0 ? stbtt_ScaleForPixelHeight(&info, size) : stbtt_ScaleForMappingEmToPixels(&info, -size); + stbtt_GetFontVMetrics(&info, &i_ascent, &i_descent, &i_lineGap); + *ascent = (float) i_ascent * scale; + *descent = (float) i_descent * scale; + *lineGap = (float) i_lineGap * scale; +} + +STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int ph, int char_index, float *xpos, float *ypos, stbtt_aligned_quad *q, int align_to_integer) +{ + float ipw = 1.0f / pw, iph = 1.0f / ph; + const stbtt_packedchar *b = chardata + char_index; + + if (align_to_integer) { + float x = (float) STBTT_ifloor((*xpos + b->xoff) + 0.5f); + float y = (float) STBTT_ifloor((*ypos + b->yoff) + 0.5f); + q->x0 = x; + q->y0 = y; + q->x1 = x + b->xoff2 - b->xoff; + q->y1 = y + b->yoff2 - b->yoff; + } else { + q->x0 = *xpos + b->xoff; + q->y0 = *ypos + b->yoff; + q->x1 = *xpos + b->xoff2; + q->y1 = *ypos + b->yoff2; + } + + q->s0 = b->x0 * ipw; + q->t0 = b->y0 * iph; + q->s1 = b->x1 * ipw; + q->t1 = b->y1 * iph; + + *xpos += b->xadvance; +} + +////////////////////////////////////////////////////////////////////////////// +// +// sdf computation +// + +#define STBTT_min(a,b) ((a) < (b) ? (a) : (b)) +#define STBTT_max(a,b) ((a) < (b) ? (b) : (a)) + +static int stbtt__ray_intersect_bezier(float orig[2], float ray[2], float q0[2], float q1[2], float q2[2], float hits[2][2]) +{ + float q0perp = q0[1]*ray[0] - q0[0]*ray[1]; + float q1perp = q1[1]*ray[0] - q1[0]*ray[1]; + float q2perp = q2[1]*ray[0] - q2[0]*ray[1]; + float roperp = orig[1]*ray[0] - orig[0]*ray[1]; + + float a = q0perp - 2*q1perp + q2perp; + float b = q1perp - q0perp; + float c = q0perp - roperp; + + float s0 = 0., s1 = 0.; + int num_s = 0; + + if (a != 0.0) { + float discr = b*b - a*c; + if (discr > 0.0) { + float rcpna = -1 / a; + float d = (float) STBTT_sqrt(discr); + s0 = (b+d) * rcpna; + s1 = (b-d) * rcpna; + if (s0 >= 0.0 && s0 <= 1.0) + num_s = 1; + if (d > 0.0 && s1 >= 0.0 && s1 <= 1.0) { + if (num_s == 0) s0 = s1; + ++num_s; + } + } + } else { + // 2*b*s + c = 0 + // s = -c / (2*b) + s0 = c / (-2 * b); + if (s0 >= 0.0 && s0 <= 1.0) + num_s = 1; + } + + if (num_s == 0) + return 0; + else { + float rcp_len2 = 1 / (ray[0]*ray[0] + ray[1]*ray[1]); + float rayn_x = ray[0] * rcp_len2, rayn_y = ray[1] * rcp_len2; + + float q0d = q0[0]*rayn_x + q0[1]*rayn_y; + float q1d = q1[0]*rayn_x + q1[1]*rayn_y; + float q2d = q2[0]*rayn_x + q2[1]*rayn_y; + float rod = orig[0]*rayn_x + orig[1]*rayn_y; + + float q10d = q1d - q0d; + float q20d = q2d - q0d; + float q0rd = q0d - rod; + + hits[0][0] = q0rd + s0*(2.0f - 2.0f*s0)*q10d + s0*s0*q20d; + hits[0][1] = a*s0+b; + + if (num_s > 1) { + hits[1][0] = q0rd + s1*(2.0f - 2.0f*s1)*q10d + s1*s1*q20d; + hits[1][1] = a*s1+b; + return 2; + } else { + return 1; + } + } +} + +static int equal(float *a, float *b) +{ + return (a[0] == b[0] && a[1] == b[1]); +} + +static int stbtt__compute_crossings_x(float x, float y, int nverts, stbtt_vertex *verts) +{ + int i; + float orig[2], ray[2] = { 1, 0 }; + float y_frac; + int winding = 0; + + // make sure y never passes through a vertex of the shape + y_frac = (float) STBTT_fmod(y, 1.0f); + if (y_frac < 0.01f) + y += 0.01f; + else if (y_frac > 0.99f) + y -= 0.01f; + + orig[0] = x; + orig[1] = y; + + // test a ray from (-infinity,y) to (x,y) + for (i=0; i < nverts; ++i) { + if (verts[i].type == STBTT_vline) { + int x0 = (int) verts[i-1].x, y0 = (int) verts[i-1].y; + int x1 = (int) verts[i ].x, y1 = (int) verts[i ].y; + if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) { + float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0; + if (x_inter < x) + winding += (y0 < y1) ? 1 : -1; + } + } + if (verts[i].type == STBTT_vcurve) { + int x0 = (int) verts[i-1].x , y0 = (int) verts[i-1].y ; + int x1 = (int) verts[i ].cx, y1 = (int) verts[i ].cy; + int x2 = (int) verts[i ].x , y2 = (int) verts[i ].y ; + int ax = STBTT_min(x0,STBTT_min(x1,x2)), ay = STBTT_min(y0,STBTT_min(y1,y2)); + int by = STBTT_max(y0,STBTT_max(y1,y2)); + if (y > ay && y < by && x > ax) { + float q0[2],q1[2],q2[2]; + float hits[2][2]; + q0[0] = (float)x0; + q0[1] = (float)y0; + q1[0] = (float)x1; + q1[1] = (float)y1; + q2[0] = (float)x2; + q2[1] = (float)y2; + if (equal(q0,q1) || equal(q1,q2)) { + x0 = (int)verts[i-1].x; + y0 = (int)verts[i-1].y; + x1 = (int)verts[i ].x; + y1 = (int)verts[i ].y; + if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) { + float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0; + if (x_inter < x) + winding += (y0 < y1) ? 1 : -1; + } + } else { + int num_hits = stbtt__ray_intersect_bezier(orig, ray, q0, q1, q2, hits); + if (num_hits >= 1) + if (hits[0][0] < 0) + winding += (hits[0][1] < 0 ? -1 : 1); + if (num_hits >= 2) + if (hits[1][0] < 0) + winding += (hits[1][1] < 0 ? -1 : 1); + } + } + } + } + return winding; +} + +static float stbtt__cuberoot( float x ) +{ + if (x<0) + return -(float) STBTT_pow(-x,1.0f/3.0f); + else + return (float) STBTT_pow( x,1.0f/3.0f); +} + +// x^3 + a*x^2 + b*x + c = 0 +static int stbtt__solve_cubic(float a, float b, float c, float* r) +{ + float s = -a / 3; + float p = b - a*a / 3; + float q = a * (2*a*a - 9*b) / 27 + c; + float p3 = p*p*p; + float d = q*q + 4*p3 / 27; + if (d >= 0) { + float z = (float) STBTT_sqrt(d); + float u = (-q + z) / 2; + float v = (-q - z) / 2; + u = stbtt__cuberoot(u); + v = stbtt__cuberoot(v); + r[0] = s + u + v; + return 1; + } else { + float u = (float) STBTT_sqrt(-p/3); + float v = (float) STBTT_acos(-STBTT_sqrt(-27/p3) * q / 2) / 3; // p3 must be negative, since d is negative + float m = (float) STBTT_cos(v); + float n = (float) STBTT_cos(v-3.141592/2)*1.732050808f; + r[0] = s + u * 2 * m; + r[1] = s - u * (m + n); + r[2] = s - u * (m - n); + + //STBTT_assert( STBTT_fabs(((r[0]+a)*r[0]+b)*r[0]+c) < 0.05f); // these asserts may not be safe at all scales, though they're in bezier t parameter units so maybe? + //STBTT_assert( STBTT_fabs(((r[1]+a)*r[1]+b)*r[1]+c) < 0.05f); + //STBTT_assert( STBTT_fabs(((r[2]+a)*r[2]+b)*r[2]+c) < 0.05f); + return 3; + } +} + +STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff) +{ + float scale_x = scale, scale_y = scale; + int ix0,iy0,ix1,iy1; + int w,h; + unsigned char *data; + + if (scale == 0) return NULL; + + stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale, scale, 0.0f,0.0f, &ix0,&iy0,&ix1,&iy1); + + // if empty, return NULL + if (ix0 == ix1 || iy0 == iy1) + return NULL; + + ix0 -= padding; + iy0 -= padding; + ix1 += padding; + iy1 += padding; + + w = (ix1 - ix0); + h = (iy1 - iy0); + + if (width ) *width = w; + if (height) *height = h; + if (xoff ) *xoff = ix0; + if (yoff ) *yoff = iy0; + + // invert for y-downwards bitmaps + scale_y = -scale_y; + + { + // distance from singular values (in the same units as the pixel grid) + const float eps = 1./1024, eps2 = eps*eps; + int x,y,i,j; + float *precompute; + stbtt_vertex *verts; + int num_verts = stbtt_GetGlyphShape(info, glyph, &verts); + data = (unsigned char *) STBTT_malloc(w * h, info->userdata); + precompute = (float *) STBTT_malloc(num_verts * sizeof(float), info->userdata); + + for (i=0,j=num_verts-1; i < num_verts; j=i++) { + if (verts[i].type == STBTT_vline) { + float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y; + float x1 = verts[j].x*scale_x, y1 = verts[j].y*scale_y; + float dist = (float) STBTT_sqrt((x1-x0)*(x1-x0) + (y1-y0)*(y1-y0)); + precompute[i] = (dist < eps) ? 0.0f : 1.0f / dist; + } else if (verts[i].type == STBTT_vcurve) { + float x2 = verts[j].x *scale_x, y2 = verts[j].y *scale_y; + float x1 = verts[i].cx*scale_x, y1 = verts[i].cy*scale_y; + float x0 = verts[i].x *scale_x, y0 = verts[i].y *scale_y; + float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2; + float len2 = bx*bx + by*by; + if (len2 >= eps2) + precompute[i] = 1.0f / len2; + else + precompute[i] = 0.0f; + } else + precompute[i] = 0.0f; + } + + for (y=iy0; y < iy1; ++y) { + for (x=ix0; x < ix1; ++x) { + float val; + float min_dist = 999999.0f; + float sx = (float) x + 0.5f; + float sy = (float) y + 0.5f; + float x_gspace = (sx / scale_x); + float y_gspace = (sy / scale_y); + + int winding = stbtt__compute_crossings_x(x_gspace, y_gspace, num_verts, verts); // @OPTIMIZE: this could just be a rasterization, but needs to be line vs. non-tesselated curves so a new path + + for (i=0; i < num_verts; ++i) { + float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y; + + if (verts[i].type == STBTT_vline && precompute[i] != 0.0f) { + float x1 = verts[i-1].x*scale_x, y1 = verts[i-1].y*scale_y; + + float dist,dist2 = (x0-sx)*(x0-sx) + (y0-sy)*(y0-sy); + if (dist2 < min_dist*min_dist) + min_dist = (float) STBTT_sqrt(dist2); + + // coarse culling against bbox + //if (sx > STBTT_min(x0,x1)-min_dist && sx < STBTT_max(x0,x1)+min_dist && + // sy > STBTT_min(y0,y1)-min_dist && sy < STBTT_max(y0,y1)+min_dist) + dist = (float) STBTT_fabs((x1-x0)*(y0-sy) - (y1-y0)*(x0-sx)) * precompute[i]; + STBTT_assert(i != 0); + if (dist < min_dist) { + // check position along line + // x' = x0 + t*(x1-x0), y' = y0 + t*(y1-y0) + // minimize (x'-sx)*(x'-sx)+(y'-sy)*(y'-sy) + float dx = x1-x0, dy = y1-y0; + float px = x0-sx, py = y0-sy; + // minimize (px+t*dx)^2 + (py+t*dy)^2 = px*px + 2*px*dx*t + t^2*dx*dx + py*py + 2*py*dy*t + t^2*dy*dy + // derivative: 2*px*dx + 2*py*dy + (2*dx*dx+2*dy*dy)*t, set to 0 and solve + float t = -(px*dx + py*dy) / (dx*dx + dy*dy); + if (t >= 0.0f && t <= 1.0f) + min_dist = dist; + } + } else if (verts[i].type == STBTT_vcurve) { + float x2 = verts[i-1].x *scale_x, y2 = verts[i-1].y *scale_y; + float x1 = verts[i ].cx*scale_x, y1 = verts[i ].cy*scale_y; + float box_x0 = STBTT_min(STBTT_min(x0,x1),x2); + float box_y0 = STBTT_min(STBTT_min(y0,y1),y2); + float box_x1 = STBTT_max(STBTT_max(x0,x1),x2); + float box_y1 = STBTT_max(STBTT_max(y0,y1),y2); + // coarse culling against bbox to avoid computing cubic unnecessarily + if (sx > box_x0-min_dist && sx < box_x1+min_dist && sy > box_y0-min_dist && sy < box_y1+min_dist) { + int num=0; + float ax = x1-x0, ay = y1-y0; + float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2; + float mx = x0 - sx, my = y0 - sy; + float res[3] = {0.f,0.f,0.f}; + float px,py,t,it,dist2; + float a_inv = precompute[i]; + if (a_inv == 0.0) { // if a_inv is 0, it's 2nd degree so use quadratic formula + float a = 3*(ax*bx + ay*by); + float b = 2*(ax*ax + ay*ay) + (mx*bx+my*by); + float c = mx*ax+my*ay; + if (STBTT_fabs(a) < eps2) { // if a is 0, it's linear + if (STBTT_fabs(b) >= eps2) { + res[num++] = -c/b; + } + } else { + float discriminant = b*b - 4*a*c; + if (discriminant < 0) + num = 0; + else { + float root = (float) STBTT_sqrt(discriminant); + res[0] = (-b - root)/(2*a); + res[1] = (-b + root)/(2*a); + num = 2; // don't bother distinguishing 1-solution case, as code below will still work + } + } + } else { + float b = 3*(ax*bx + ay*by) * a_inv; // could precompute this as it doesn't depend on sample point + float c = (2*(ax*ax + ay*ay) + (mx*bx+my*by)) * a_inv; + float d = (mx*ax+my*ay) * a_inv; + num = stbtt__solve_cubic(b, c, d, res); + } + dist2 = (x0-sx)*(x0-sx) + (y0-sy)*(y0-sy); + if (dist2 < min_dist*min_dist) + min_dist = (float) STBTT_sqrt(dist2); + + if (num >= 1 && res[0] >= 0.0f && res[0] <= 1.0f) { + t = res[0], it = 1.0f - t; + px = it*it*x0 + 2*t*it*x1 + t*t*x2; + py = it*it*y0 + 2*t*it*y1 + t*t*y2; + dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); + if (dist2 < min_dist * min_dist) + min_dist = (float) STBTT_sqrt(dist2); + } + if (num >= 2 && res[1] >= 0.0f && res[1] <= 1.0f) { + t = res[1], it = 1.0f - t; + px = it*it*x0 + 2*t*it*x1 + t*t*x2; + py = it*it*y0 + 2*t*it*y1 + t*t*y2; + dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); + if (dist2 < min_dist * min_dist) + min_dist = (float) STBTT_sqrt(dist2); + } + if (num >= 3 && res[2] >= 0.0f && res[2] <= 1.0f) { + t = res[2], it = 1.0f - t; + px = it*it*x0 + 2*t*it*x1 + t*t*x2; + py = it*it*y0 + 2*t*it*y1 + t*t*y2; + dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); + if (dist2 < min_dist * min_dist) + min_dist = (float) STBTT_sqrt(dist2); + } + } + } + } + if (winding == 0) + min_dist = -min_dist; // if outside the shape, value is negative + val = onedge_value + pixel_dist_scale * min_dist; + if (val < 0) + val = 0; + else if (val > 255) + val = 255; + data[(y-iy0)*w+(x-ix0)] = (unsigned char) val; + } + } + STBTT_free(precompute, info->userdata); + STBTT_free(verts, info->userdata); + } + return data; +} + +STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff) +{ + return stbtt_GetGlyphSDF(info, scale, stbtt_FindGlyphIndex(info, codepoint), padding, onedge_value, pixel_dist_scale, width, height, xoff, yoff); +} + +STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata) +{ + STBTT_free(bitmap, userdata); +} + +////////////////////////////////////////////////////////////////////////////// +// +// font name matching -- recommended not to use this +// + +// check if a utf8 string contains a prefix which is the utf16 string; if so return length of matching utf8 string +static stbtt_int32 stbtt__CompareUTF8toUTF16_bigendian_prefix(stbtt_uint8 *s1, stbtt_int32 len1, stbtt_uint8 *s2, stbtt_int32 len2) +{ + stbtt_int32 i=0; + + // convert utf16 to utf8 and compare the results while converting + while (len2) { + stbtt_uint16 ch = s2[0]*256 + s2[1]; + if (ch < 0x80) { + if (i >= len1) return -1; + if (s1[i++] != ch) return -1; + } else if (ch < 0x800) { + if (i+1 >= len1) return -1; + if (s1[i++] != 0xc0 + (ch >> 6)) return -1; + if (s1[i++] != 0x80 + (ch & 0x3f)) return -1; + } else if (ch >= 0xd800 && ch < 0xdc00) { + stbtt_uint32 c; + stbtt_uint16 ch2 = s2[2]*256 + s2[3]; + if (i+3 >= len1) return -1; + c = ((ch - 0xd800) << 10) + (ch2 - 0xdc00) + 0x10000; + if (s1[i++] != 0xf0 + (c >> 18)) return -1; + if (s1[i++] != 0x80 + ((c >> 12) & 0x3f)) return -1; + if (s1[i++] != 0x80 + ((c >> 6) & 0x3f)) return -1; + if (s1[i++] != 0x80 + ((c ) & 0x3f)) return -1; + s2 += 2; // plus another 2 below + len2 -= 2; + } else if (ch >= 0xdc00 && ch < 0xe000) { + return -1; + } else { + if (i+2 >= len1) return -1; + if (s1[i++] != 0xe0 + (ch >> 12)) return -1; + if (s1[i++] != 0x80 + ((ch >> 6) & 0x3f)) return -1; + if (s1[i++] != 0x80 + ((ch ) & 0x3f)) return -1; + } + s2 += 2; + len2 -= 2; + } + return i; +} + +static int stbtt_CompareUTF8toUTF16_bigendian_internal(char *s1, int len1, char *s2, int len2) +{ + return len1 == stbtt__CompareUTF8toUTF16_bigendian_prefix((stbtt_uint8*) s1, len1, (stbtt_uint8*) s2, len2); +} + +// returns results in whatever encoding you request... but note that 2-byte encodings +// will be BIG-ENDIAN... use stbtt_CompareUTF8toUTF16_bigendian() to compare +STBTT_DEF const char *stbtt_GetFontNameString(const stbtt_fontinfo *font, int *length, int platformID, int encodingID, int languageID, int nameID) +{ + stbtt_int32 i,count,stringOffset; + stbtt_uint8 *fc = font->data; + stbtt_uint32 offset = font->fontstart; + stbtt_uint32 nm = stbtt__find_table(fc, offset, "name"); + if (!nm) return NULL; + + count = ttUSHORT(fc+nm+2); + stringOffset = nm + ttUSHORT(fc+nm+4); + for (i=0; i < count; ++i) { + stbtt_uint32 loc = nm + 6 + 12 * i; + if (platformID == ttUSHORT(fc+loc+0) && encodingID == ttUSHORT(fc+loc+2) + && languageID == ttUSHORT(fc+loc+4) && nameID == ttUSHORT(fc+loc+6)) { + *length = ttUSHORT(fc+loc+8); + return (const char *) (fc+stringOffset+ttUSHORT(fc+loc+10)); + } + } + return NULL; +} + +static int stbtt__matchpair(stbtt_uint8 *fc, stbtt_uint32 nm, stbtt_uint8 *name, stbtt_int32 nlen, stbtt_int32 target_id, stbtt_int32 next_id) +{ + stbtt_int32 i; + stbtt_int32 count = ttUSHORT(fc+nm+2); + stbtt_int32 stringOffset = nm + ttUSHORT(fc+nm+4); + + for (i=0; i < count; ++i) { + stbtt_uint32 loc = nm + 6 + 12 * i; + stbtt_int32 id = ttUSHORT(fc+loc+6); + if (id == target_id) { + // find the encoding + stbtt_int32 platform = ttUSHORT(fc+loc+0), encoding = ttUSHORT(fc+loc+2), language = ttUSHORT(fc+loc+4); + + // is this a Unicode encoding? + if (platform == 0 || (platform == 3 && encoding == 1) || (platform == 3 && encoding == 10)) { + stbtt_int32 slen = ttUSHORT(fc+loc+8); + stbtt_int32 off = ttUSHORT(fc+loc+10); + + // check if there's a prefix match + stbtt_int32 matchlen = stbtt__CompareUTF8toUTF16_bigendian_prefix(name, nlen, fc+stringOffset+off,slen); + if (matchlen >= 0) { + // check for target_id+1 immediately following, with same encoding & language + if (i+1 < count && ttUSHORT(fc+loc+12+6) == next_id && ttUSHORT(fc+loc+12) == platform && ttUSHORT(fc+loc+12+2) == encoding && ttUSHORT(fc+loc+12+4) == language) { + slen = ttUSHORT(fc+loc+12+8); + off = ttUSHORT(fc+loc+12+10); + if (slen == 0) { + if (matchlen == nlen) + return 1; + } else if (matchlen < nlen && name[matchlen] == ' ') { + ++matchlen; + if (stbtt_CompareUTF8toUTF16_bigendian_internal((char*) (name+matchlen), nlen-matchlen, (char*)(fc+stringOffset+off),slen)) + return 1; + } + } else { + // if nothing immediately following + if (matchlen == nlen) + return 1; + } + } + } + + // @TODO handle other encodings + } + } + return 0; +} + +static int stbtt__matches(stbtt_uint8 *fc, stbtt_uint32 offset, stbtt_uint8 *name, stbtt_int32 flags) +{ + stbtt_int32 nlen = (stbtt_int32) STBTT_strlen((char *) name); + stbtt_uint32 nm,hd; + if (!stbtt__isfont(fc+offset)) return 0; + + // check italics/bold/underline flags in macStyle... + if (flags) { + hd = stbtt__find_table(fc, offset, "head"); + if ((ttUSHORT(fc+hd+44) & 7) != (flags & 7)) return 0; + } + + nm = stbtt__find_table(fc, offset, "name"); + if (!nm) return 0; + + if (flags) { + // if we checked the macStyle flags, then just check the family and ignore the subfamily + if (stbtt__matchpair(fc, nm, name, nlen, 16, -1)) return 1; + if (stbtt__matchpair(fc, nm, name, nlen, 1, -1)) return 1; + if (stbtt__matchpair(fc, nm, name, nlen, 3, -1)) return 1; + } else { + if (stbtt__matchpair(fc, nm, name, nlen, 16, 17)) return 1; + if (stbtt__matchpair(fc, nm, name, nlen, 1, 2)) return 1; + if (stbtt__matchpair(fc, nm, name, nlen, 3, -1)) return 1; + } + + return 0; +} + +static int stbtt_FindMatchingFont_internal(unsigned char *font_collection, char *name_utf8, stbtt_int32 flags) +{ + stbtt_int32 i; + for (i=0;;++i) { + stbtt_int32 off = stbtt_GetFontOffsetForIndex(font_collection, i); + if (off < 0) return off; + if (stbtt__matches((stbtt_uint8 *) font_collection, off, (stbtt_uint8*) name_utf8, flags)) + return off; + } +} + +#if defined(__GNUC__) || defined(__clang__) +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wcast-qual" +#endif + +STBTT_DEF int stbtt_BakeFontBitmap(const unsigned char *data, int offset, + float pixel_height, unsigned char *pixels, int pw, int ph, + int first_char, int num_chars, stbtt_bakedchar *chardata) +{ + return stbtt_BakeFontBitmap_internal((unsigned char *) data, offset, pixel_height, pixels, pw, ph, first_char, num_chars, chardata); +} + +STBTT_DEF int stbtt_GetFontOffsetForIndex(const unsigned char *data, int index) +{ + return stbtt_GetFontOffsetForIndex_internal((unsigned char *) data, index); +} + +STBTT_DEF int stbtt_GetNumberOfFonts(const unsigned char *data) +{ + return stbtt_GetNumberOfFonts_internal((unsigned char *) data); +} + +STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data, int offset) +{ + return stbtt_InitFont_internal(info, (unsigned char *) data, offset); +} + +STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *fontdata, const char *name, int flags) +{ + return stbtt_FindMatchingFont_internal((unsigned char *) fontdata, (char *) name, flags); +} + +STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2) +{ + return stbtt_CompareUTF8toUTF16_bigendian_internal((char *) s1, len1, (char *) s2, len2); +} + +#if defined(__GNUC__) || defined(__clang__) +#pragma GCC diagnostic pop +#endif + +#endif // STB_TRUETYPE_IMPLEMENTATION + + +// FULL VERSION HISTORY +// +// 1.25 (2021-07-11) many fixes +// 1.24 (2020-02-05) fix warning +// 1.23 (2020-02-02) query SVG data for glyphs; query whole kerning table (but only kern not GPOS) +// 1.22 (2019-08-11) minimize missing-glyph duplication; fix kerning if both 'GPOS' and 'kern' are defined +// 1.21 (2019-02-25) fix warning +// 1.20 (2019-02-07) PackFontRange skips missing codepoints; GetScaleFontVMetrics() +// 1.19 (2018-02-11) OpenType GPOS kerning (horizontal only), STBTT_fmod +// 1.18 (2018-01-29) add missing function +// 1.17 (2017-07-23) make more arguments const; doc fix +// 1.16 (2017-07-12) SDF support +// 1.15 (2017-03-03) make more arguments const +// 1.14 (2017-01-16) num-fonts-in-TTC function +// 1.13 (2017-01-02) support OpenType fonts, certain Apple fonts +// 1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual +// 1.11 (2016-04-02) fix unused-variable warning +// 1.10 (2016-04-02) allow user-defined fabs() replacement +// fix memory leak if fontsize=0.0 +// fix warning from duplicate typedef +// 1.09 (2016-01-16) warning fix; avoid crash on outofmem; use alloc userdata for PackFontRanges +// 1.08 (2015-09-13) document stbtt_Rasterize(); fixes for vertical & horizontal edges +// 1.07 (2015-08-01) allow PackFontRanges to accept arrays of sparse codepoints; +// allow PackFontRanges to pack and render in separate phases; +// fix stbtt_GetFontOFfsetForIndex (never worked for non-0 input?); +// fixed an assert() bug in the new rasterizer +// replace assert() with STBTT_assert() in new rasterizer +// 1.06 (2015-07-14) performance improvements (~35% faster on x86 and x64 on test machine) +// also more precise AA rasterizer, except if shapes overlap +// remove need for STBTT_sort +// 1.05 (2015-04-15) fix misplaced definitions for STBTT_STATIC +// 1.04 (2015-04-15) typo in example +// 1.03 (2015-04-12) STBTT_STATIC, fix memory leak in new packing, various fixes +// 1.02 (2014-12-10) fix various warnings & compile issues w/ stb_rect_pack, C++ +// 1.01 (2014-12-08) fix subpixel position when oversampling to exactly match +// non-oversampled; STBTT_POINT_SIZE for packed case only +// 1.00 (2014-12-06) add new PackBegin etc. API, w/ support for oversampling +// 0.99 (2014-09-18) fix multiple bugs with subpixel rendering (ryg) +// 0.9 (2014-08-07) support certain mac/iOS fonts without an MS platformID +// 0.8b (2014-07-07) fix a warning +// 0.8 (2014-05-25) fix a few more warnings +// 0.7 (2013-09-25) bugfix: subpixel glyph bug fixed in 0.5 had come back +// 0.6c (2012-07-24) improve documentation +// 0.6b (2012-07-20) fix a few more warnings +// 0.6 (2012-07-17) fix warnings; added stbtt_ScaleForMappingEmToPixels, +// stbtt_GetFontBoundingBox, stbtt_IsGlyphEmpty +// 0.5 (2011-12-09) bugfixes: +// subpixel glyph renderer computed wrong bounding box +// first vertex of shape can be off-curve (FreeSans) +// 0.4b (2011-12-03) fixed an error in the font baking example +// 0.4 (2011-12-01) kerning, subpixel rendering (tor) +// bugfixes for: +// codepoint-to-glyph conversion using table fmt=12 +// codepoint-to-glyph conversion using table fmt=4 +// stbtt_GetBakedQuad with non-square texture (Zer) +// updated Hello World! sample to use kerning and subpixel +// fixed some warnings +// 0.3 (2009-06-24) cmap fmt=12, compound shapes (MM) +// userdata, malloc-from-userdata, non-zero fill (stb) +// 0.2 (2009-03-11) Fix unsigned/signed char warnings +// 0.1 (2009-03-09) First public release +// + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/platform/stb_vorbis.c b/platform/stb_vorbis.c new file mode 100644 index 0000000..3e5c250 --- /dev/null +++ b/platform/stb_vorbis.c @@ -0,0 +1,5584 @@ +// Ogg Vorbis audio decoder - v1.22 - public domain +// http://nothings.org/stb_vorbis/ +// +// Original version written by Sean Barrett in 2007. +// +// Originally sponsored by RAD Game Tools. Seeking implementation +// sponsored by Phillip Bennefall, Marc Andersen, Aaron Baker, +// Elias Software, Aras Pranckevicius, and Sean Barrett. +// +// LICENSE +// +// See end of file for license information. +// +// Limitations: +// +// - floor 0 not supported (used in old ogg vorbis files pre-2004) +// - lossless sample-truncation at beginning ignored +// - cannot concatenate multiple vorbis streams +// - sample positions are 32-bit, limiting seekable 192Khz +// files to around 6 hours (Ogg supports 64-bit) +// +// Feature contributors: +// Dougall Johnson (sample-exact seeking) +// +// Bugfix/warning contributors: +// Terje Mathisen Niklas Frykholm Andy Hill +// Casey Muratori John Bolton Gargaj +// Laurent Gomila Marc LeBlanc Ronny Chevalier +// Bernhard Wodo Evan Balster github:alxprd +// Tom Beaumont Ingo Leitgeb Nicolas Guillemot +// Phillip Bennefall Rohit Thiago Goulart +// github:manxorist Saga Musix github:infatum +// Timur Gagiev Maxwell Koo Peter Waller +// github:audinowho Dougall Johnson David Reid +// github:Clownacy Pedro J. Estebanez Remi Verschelde +// AnthoFoxo github:morlat Gabriel Ravier +// +// Partial history: +// 1.22 - 2021-07-11 - various small fixes +// 1.21 - 2021-07-02 - fix bug for files with no comments +// 1.20 - 2020-07-11 - several small fixes +// 1.19 - 2020-02-05 - warnings +// 1.18 - 2020-02-02 - fix seek bugs; parse header comments; misc warnings etc. +// 1.17 - 2019-07-08 - fix CVE-2019-13217..CVE-2019-13223 (by ForAllSecure) +// 1.16 - 2019-03-04 - fix warnings +// 1.15 - 2019-02-07 - explicit failure if Ogg Skeleton data is found +// 1.14 - 2018-02-11 - delete bogus dealloca usage +// 1.13 - 2018-01-29 - fix truncation of last frame (hopefully) +// 1.12 - 2017-11-21 - limit residue begin/end to blocksize/2 to avoid large temp allocs in bad/corrupt files +// 1.11 - 2017-07-23 - fix MinGW compilation +// 1.10 - 2017-03-03 - more robust seeking; fix negative ilog(); clear error in open_memory +// 1.09 - 2016-04-04 - back out 'truncation of last frame' fix from previous version +// 1.08 - 2016-04-02 - warnings; setup memory leaks; truncation of last frame +// 1.07 - 2015-01-16 - fixes for crashes on invalid files; warning fixes; const +// 1.06 - 2015-08-31 - full, correct support for seeking API (Dougall Johnson) +// some crash fixes when out of memory or with corrupt files +// fix some inappropriately signed shifts +// 1.05 - 2015-04-19 - don't define __forceinline if it's redundant +// 1.04 - 2014-08-27 - fix missing const-correct case in API +// 1.03 - 2014-08-07 - warning fixes +// 1.02 - 2014-07-09 - declare qsort comparison as explicitly _cdecl in Windows +// 1.01 - 2014-06-18 - fix stb_vorbis_get_samples_float (interleaved was correct) +// 1.0 - 2014-05-26 - fix memory leaks; fix warnings; fix bugs in >2-channel; +// (API change) report sample rate for decode-full-file funcs +// +// See end of file for full version history. + + +////////////////////////////////////////////////////////////////////////////// +// +// HEADER BEGINS HERE +// + +#ifndef STB_VORBIS_INCLUDE_STB_VORBIS_H +#define STB_VORBIS_INCLUDE_STB_VORBIS_H + +#if defined(STB_VORBIS_NO_CRT) && !defined(STB_VORBIS_NO_STDIO) +#define STB_VORBIS_NO_STDIO 1 +#endif + +#ifndef STB_VORBIS_NO_STDIO +#include +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +/////////// THREAD SAFETY + +// Individual stb_vorbis* handles are not thread-safe; you cannot decode from +// them from multiple threads at the same time. However, you can have multiple +// stb_vorbis* handles and decode from them independently in multiple thrads. + + +/////////// MEMORY ALLOCATION + +// normally stb_vorbis uses malloc() to allocate memory at startup, +// and alloca() to allocate temporary memory during a frame on the +// stack. (Memory consumption will depend on the amount of setup +// data in the file and how you set the compile flags for speed +// vs. size. In my test files the maximal-size usage is ~150KB.) +// +// You can modify the wrapper functions in the source (setup_malloc, +// setup_temp_malloc, temp_malloc) to change this behavior, or you +// can use a simpler allocation model: you pass in a buffer from +// which stb_vorbis will allocate _all_ its memory (including the +// temp memory). "open" may fail with a VORBIS_outofmem if you +// do not pass in enough data; there is no way to determine how +// much you do need except to succeed (at which point you can +// query get_info to find the exact amount required. yes I know +// this is lame). +// +// If you pass in a non-NULL buffer of the type below, allocation +// will occur from it as described above. Otherwise just pass NULL +// to use malloc()/alloca() + +typedef struct +{ + char *alloc_buffer; + int alloc_buffer_length_in_bytes; +} stb_vorbis_alloc; + + +/////////// FUNCTIONS USEABLE WITH ALL INPUT MODES + +typedef struct stb_vorbis stb_vorbis; + +typedef struct +{ + unsigned int sample_rate; + int channels; + + unsigned int setup_memory_required; + unsigned int setup_temp_memory_required; + unsigned int temp_memory_required; + + int max_frame_size; +} stb_vorbis_info; + +typedef struct +{ + char *vendor; + + int comment_list_length; + char **comment_list; +} stb_vorbis_comment; + +// get general information about the file +extern stb_vorbis_info stb_vorbis_get_info(stb_vorbis *f); + +// get ogg comments +extern stb_vorbis_comment stb_vorbis_get_comment(stb_vorbis *f); + +// get the last error detected (clears it, too) +extern int stb_vorbis_get_error(stb_vorbis *f); + +// close an ogg vorbis file and free all memory in use +extern void stb_vorbis_close(stb_vorbis *f); + +// this function returns the offset (in samples) from the beginning of the +// file that will be returned by the next decode, if it is known, or -1 +// otherwise. after a flush_pushdata() call, this may take a while before +// it becomes valid again. +// NOT WORKING YET after a seek with PULLDATA API +extern int stb_vorbis_get_sample_offset(stb_vorbis *f); + +// returns the current seek point within the file, or offset from the beginning +// of the memory buffer. In pushdata mode it returns 0. +extern unsigned int stb_vorbis_get_file_offset(stb_vorbis *f); + +/////////// PUSHDATA API + +#ifndef STB_VORBIS_NO_PUSHDATA_API + +// this API allows you to get blocks of data from any source and hand +// them to stb_vorbis. you have to buffer them; stb_vorbis will tell +// you how much it used, and you have to give it the rest next time; +// and stb_vorbis may not have enough data to work with and you will +// need to give it the same data again PLUS more. Note that the Vorbis +// specification does not bound the size of an individual frame. + +extern stb_vorbis *stb_vorbis_open_pushdata( + const unsigned char * datablock, int datablock_length_in_bytes, + int *datablock_memory_consumed_in_bytes, + int *error, + const stb_vorbis_alloc *alloc_buffer); +// create a vorbis decoder by passing in the initial data block containing +// the ogg&vorbis headers (you don't need to do parse them, just provide +// the first N bytes of the file--you're told if it's not enough, see below) +// on success, returns an stb_vorbis *, does not set error, returns the amount of +// data parsed/consumed on this call in *datablock_memory_consumed_in_bytes; +// on failure, returns NULL on error and sets *error, does not change *datablock_memory_consumed +// if returns NULL and *error is VORBIS_need_more_data, then the input block was +// incomplete and you need to pass in a larger block from the start of the file + +extern int stb_vorbis_decode_frame_pushdata( + stb_vorbis *f, + const unsigned char *datablock, int datablock_length_in_bytes, + int *channels, // place to write number of float * buffers + float ***output, // place to write float ** array of float * buffers + int *samples // place to write number of output samples + ); +// decode a frame of audio sample data if possible from the passed-in data block +// +// return value: number of bytes we used from datablock +// +// possible cases: +// 0 bytes used, 0 samples output (need more data) +// N bytes used, 0 samples output (resynching the stream, keep going) +// N bytes used, M samples output (one frame of data) +// note that after opening a file, you will ALWAYS get one N-bytes,0-sample +// frame, because Vorbis always "discards" the first frame. +// +// Note that on resynch, stb_vorbis will rarely consume all of the buffer, +// instead only datablock_length_in_bytes-3 or less. This is because it wants +// to avoid missing parts of a page header if they cross a datablock boundary, +// without writing state-machiney code to record a partial detection. +// +// The number of channels returned are stored in *channels (which can be +// NULL--it is always the same as the number of channels reported by +// get_info). *output will contain an array of float* buffers, one per +// channel. In other words, (*output)[0][0] contains the first sample from +// the first channel, and (*output)[1][0] contains the first sample from +// the second channel. +// +// *output points into stb_vorbis's internal output buffer storage; these +// buffers are owned by stb_vorbis and application code should not free +// them or modify their contents. They are transient and will be overwritten +// once you ask for more data to get decoded, so be sure to grab any data +// you need before then. + +extern void stb_vorbis_flush_pushdata(stb_vorbis *f); +// inform stb_vorbis that your next datablock will not be contiguous with +// previous ones (e.g. you've seeked in the data); future attempts to decode +// frames will cause stb_vorbis to resynchronize (as noted above), and +// once it sees a valid Ogg page (typically 4-8KB, as large as 64KB), it +// will begin decoding the _next_ frame. +// +// if you want to seek using pushdata, you need to seek in your file, then +// call stb_vorbis_flush_pushdata(), then start calling decoding, then once +// decoding is returning you data, call stb_vorbis_get_sample_offset, and +// if you don't like the result, seek your file again and repeat. +#endif + + +////////// PULLING INPUT API + +#ifndef STB_VORBIS_NO_PULLDATA_API +// This API assumes stb_vorbis is allowed to pull data from a source-- +// either a block of memory containing the _entire_ vorbis stream, or a +// FILE * that you or it create, or possibly some other reading mechanism +// if you go modify the source to replace the FILE * case with some kind +// of callback to your code. (But if you don't support seeking, you may +// just want to go ahead and use pushdata.) + +#if !defined(STB_VORBIS_NO_STDIO) && !defined(STB_VORBIS_NO_INTEGER_CONVERSION) +extern int stb_vorbis_decode_filename(const char *filename, int *channels, int *sample_rate, short **output); +#endif +#if !defined(STB_VORBIS_NO_INTEGER_CONVERSION) +extern int stb_vorbis_decode_memory(const unsigned char *mem, int len, int *channels, int *sample_rate, short **output); +#endif +// decode an entire file and output the data interleaved into a malloc()ed +// buffer stored in *output. The return value is the number of samples +// decoded, or -1 if the file could not be opened or was not an ogg vorbis file. +// When you're done with it, just free() the pointer returned in *output. + +extern stb_vorbis * stb_vorbis_open_memory(const unsigned char *data, int len, + int *error, const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from an ogg vorbis stream in memory (note +// this must be the entire stream!). on failure, returns NULL and sets *error + +#ifndef STB_VORBIS_NO_STDIO +extern stb_vorbis * stb_vorbis_open_filename(const char *filename, + int *error, const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from a filename via fopen(). on failure, +// returns NULL and sets *error (possibly to VORBIS_file_open_failure). + +extern stb_vorbis * stb_vorbis_open_file(FILE *f, int close_handle_on_close, + int *error, const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from an open FILE *, looking for a stream at +// the _current_ seek point (ftell). on failure, returns NULL and sets *error. +// note that stb_vorbis must "own" this stream; if you seek it in between +// calls to stb_vorbis, it will become confused. Moreover, if you attempt to +// perform stb_vorbis_seek_*() operations on this file, it will assume it +// owns the _entire_ rest of the file after the start point. Use the next +// function, stb_vorbis_open_file_section(), to limit it. + +extern stb_vorbis * stb_vorbis_open_file_section(FILE *f, int close_handle_on_close, + int *error, const stb_vorbis_alloc *alloc_buffer, unsigned int len); +// create an ogg vorbis decoder from an open FILE *, looking for a stream at +// the _current_ seek point (ftell); the stream will be of length 'len' bytes. +// on failure, returns NULL and sets *error. note that stb_vorbis must "own" +// this stream; if you seek it in between calls to stb_vorbis, it will become +// confused. +#endif + +extern int stb_vorbis_seek_frame(stb_vorbis *f, unsigned int sample_number); +extern int stb_vorbis_seek(stb_vorbis *f, unsigned int sample_number); +// these functions seek in the Vorbis file to (approximately) 'sample_number'. +// after calling seek_frame(), the next call to get_frame_*() will include +// the specified sample. after calling stb_vorbis_seek(), the next call to +// stb_vorbis_get_samples_* will start with the specified sample. If you +// do not need to seek to EXACTLY the target sample when using get_samples_*, +// you can also use seek_frame(). + +extern int stb_vorbis_seek_start(stb_vorbis *f); +// this function is equivalent to stb_vorbis_seek(f,0) + +extern unsigned int stb_vorbis_stream_length_in_samples(stb_vorbis *f); +extern float stb_vorbis_stream_length_in_seconds(stb_vorbis *f); +// these functions return the total length of the vorbis stream + +extern int stb_vorbis_get_frame_float(stb_vorbis *f, int *channels, float ***output); +// decode the next frame and return the number of samples. the number of +// channels returned are stored in *channels (which can be NULL--it is always +// the same as the number of channels reported by get_info). *output will +// contain an array of float* buffers, one per channel. These outputs will +// be overwritten on the next call to stb_vorbis_get_frame_*. +// +// You generally should not intermix calls to stb_vorbis_get_frame_*() +// and stb_vorbis_get_samples_*(), since the latter calls the former. + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +extern int stb_vorbis_get_frame_short_interleaved(stb_vorbis *f, int num_c, short *buffer, int num_shorts); +extern int stb_vorbis_get_frame_short (stb_vorbis *f, int num_c, short **buffer, int num_samples); +#endif +// decode the next frame and return the number of *samples* per channel. +// Note that for interleaved data, you pass in the number of shorts (the +// size of your array), but the return value is the number of samples per +// channel, not the total number of samples. +// +// The data is coerced to the number of channels you request according to the +// channel coercion rules (see below). You must pass in the size of your +// buffer(s) so that stb_vorbis will not overwrite the end of the buffer. +// The maximum buffer size needed can be gotten from get_info(); however, +// the Vorbis I specification implies an absolute maximum of 4096 samples +// per channel. + +// Channel coercion rules: +// Let M be the number of channels requested, and N the number of channels present, +// and Cn be the nth channel; let stereo L be the sum of all L and center channels, +// and stereo R be the sum of all R and center channels (channel assignment from the +// vorbis spec). +// M N output +// 1 k sum(Ck) for all k +// 2 * stereo L, stereo R +// k l k > l, the first l channels, then 0s +// k l k <= l, the first k channels +// Note that this is not _good_ surround etc. mixing at all! It's just so +// you get something useful. + +extern int stb_vorbis_get_samples_float_interleaved(stb_vorbis *f, int channels, float *buffer, int num_floats); +extern int stb_vorbis_get_samples_float(stb_vorbis *f, int channels, float **buffer, int num_samples); +// gets num_samples samples, not necessarily on a frame boundary--this requires +// buffering so you have to supply the buffers. DOES NOT APPLY THE COERCION RULES. +// Returns the number of samples stored per channel; it may be less than requested +// at the end of the file. If there are no more samples in the file, returns 0. + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +extern int stb_vorbis_get_samples_short_interleaved(stb_vorbis *f, int channels, short *buffer, int num_shorts); +extern int stb_vorbis_get_samples_short(stb_vorbis *f, int channels, short **buffer, int num_samples); +#endif +// gets num_samples samples, not necessarily on a frame boundary--this requires +// buffering so you have to supply the buffers. Applies the coercion rules above +// to produce 'channels' channels. Returns the number of samples stored per channel; +// it may be less than requested at the end of the file. If there are no more +// samples in the file, returns 0. + +#endif + +//////// ERROR CODES + +enum STBVorbisError +{ + VORBIS__no_error, + + VORBIS_need_more_data=1, // not a real error + + VORBIS_invalid_api_mixing, // can't mix API modes + VORBIS_outofmem, // not enough memory + VORBIS_feature_not_supported, // uses floor 0 + VORBIS_too_many_channels, // STB_VORBIS_MAX_CHANNELS is too small + VORBIS_file_open_failure, // fopen() failed + VORBIS_seek_without_length, // can't seek in unknown-length file + + VORBIS_unexpected_eof=10, // file is truncated? + VORBIS_seek_invalid, // seek past EOF + + // decoding errors (corrupt/invalid stream) -- you probably + // don't care about the exact details of these + + // vorbis errors: + VORBIS_invalid_setup=20, + VORBIS_invalid_stream, + + // ogg errors: + VORBIS_missing_capture_pattern=30, + VORBIS_invalid_stream_structure_version, + VORBIS_continued_packet_flag_invalid, + VORBIS_incorrect_stream_serial_number, + VORBIS_invalid_first_page, + VORBIS_bad_packet_type, + VORBIS_cant_find_last_page, + VORBIS_seek_failed, + VORBIS_ogg_skeleton_not_supported +}; + + +#ifdef __cplusplus +} +#endif + +#endif // STB_VORBIS_INCLUDE_STB_VORBIS_H +// +// HEADER ENDS HERE +// +////////////////////////////////////////////////////////////////////////////// + +#ifndef STB_VORBIS_HEADER_ONLY + +// global configuration settings (e.g. set these in the project/makefile), +// or just set them in this file at the top (although ideally the first few +// should be visible when the header file is compiled too, although it's not +// crucial) + +// STB_VORBIS_NO_PUSHDATA_API +// does not compile the code for the various stb_vorbis_*_pushdata() +// functions +// #define STB_VORBIS_NO_PUSHDATA_API + +// STB_VORBIS_NO_PULLDATA_API +// does not compile the code for the non-pushdata APIs +// #define STB_VORBIS_NO_PULLDATA_API + +// STB_VORBIS_NO_STDIO +// does not compile the code for the APIs that use FILE *s internally +// or externally (implied by STB_VORBIS_NO_PULLDATA_API) +// #define STB_VORBIS_NO_STDIO + +// STB_VORBIS_NO_INTEGER_CONVERSION +// does not compile the code for converting audio sample data from +// float to integer (implied by STB_VORBIS_NO_PULLDATA_API) +// #define STB_VORBIS_NO_INTEGER_CONVERSION + +// STB_VORBIS_NO_FAST_SCALED_FLOAT +// does not use a fast float-to-int trick to accelerate float-to-int on +// most platforms which requires endianness be defined correctly. +//#define STB_VORBIS_NO_FAST_SCALED_FLOAT + + +// STB_VORBIS_MAX_CHANNELS [number] +// globally define this to the maximum number of channels you need. +// The spec does not put a restriction on channels except that +// the count is stored in a byte, so 255 is the hard limit. +// Reducing this saves about 16 bytes per value, so using 16 saves +// (255-16)*16 or around 4KB. Plus anything other memory usage +// I forgot to account for. Can probably go as low as 8 (7.1 audio), +// 6 (5.1 audio), or 2 (stereo only). +#ifndef STB_VORBIS_MAX_CHANNELS +#define STB_VORBIS_MAX_CHANNELS 16 // enough for anyone? +#endif + +// STB_VORBIS_PUSHDATA_CRC_COUNT [number] +// after a flush_pushdata(), stb_vorbis begins scanning for the +// next valid page, without backtracking. when it finds something +// that looks like a page, it streams through it and verifies its +// CRC32. Should that validation fail, it keeps scanning. But it's +// possible that _while_ streaming through to check the CRC32 of +// one candidate page, it sees another candidate page. This #define +// determines how many "overlapping" candidate pages it can search +// at once. Note that "real" pages are typically ~4KB to ~8KB, whereas +// garbage pages could be as big as 64KB, but probably average ~16KB. +// So don't hose ourselves by scanning an apparent 64KB page and +// missing a ton of real ones in the interim; so minimum of 2 +#ifndef STB_VORBIS_PUSHDATA_CRC_COUNT +#define STB_VORBIS_PUSHDATA_CRC_COUNT 4 +#endif + +// STB_VORBIS_FAST_HUFFMAN_LENGTH [number] +// sets the log size of the huffman-acceleration table. Maximum +// supported value is 24. with larger numbers, more decodings are O(1), +// but the table size is larger so worse cache missing, so you'll have +// to probe (and try multiple ogg vorbis files) to find the sweet spot. +#ifndef STB_VORBIS_FAST_HUFFMAN_LENGTH +#define STB_VORBIS_FAST_HUFFMAN_LENGTH 10 +#endif + +// STB_VORBIS_FAST_BINARY_LENGTH [number] +// sets the log size of the binary-search acceleration table. this +// is used in similar fashion to the fast-huffman size to set initial +// parameters for the binary search + +// STB_VORBIS_FAST_HUFFMAN_INT +// The fast huffman tables are much more efficient if they can be +// stored as 16-bit results instead of 32-bit results. This restricts +// the codebooks to having only 65535 possible outcomes, though. +// (At least, accelerated by the huffman table.) +#ifndef STB_VORBIS_FAST_HUFFMAN_INT +#define STB_VORBIS_FAST_HUFFMAN_SHORT +#endif + +// STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH +// If the 'fast huffman' search doesn't succeed, then stb_vorbis falls +// back on binary searching for the correct one. This requires storing +// extra tables with the huffman codes in sorted order. Defining this +// symbol trades off space for speed by forcing a linear search in the +// non-fast case, except for "sparse" codebooks. +// #define STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH + +// STB_VORBIS_DIVIDES_IN_RESIDUE +// stb_vorbis precomputes the result of the scalar residue decoding +// that would otherwise require a divide per chunk. you can trade off +// space for time by defining this symbol. +// #define STB_VORBIS_DIVIDES_IN_RESIDUE + +// STB_VORBIS_DIVIDES_IN_CODEBOOK +// vorbis VQ codebooks can be encoded two ways: with every case explicitly +// stored, or with all elements being chosen from a small range of values, +// and all values possible in all elements. By default, stb_vorbis expands +// this latter kind out to look like the former kind for ease of decoding, +// because otherwise an integer divide-per-vector-element is required to +// unpack the index. If you define STB_VORBIS_DIVIDES_IN_CODEBOOK, you can +// trade off storage for speed. +//#define STB_VORBIS_DIVIDES_IN_CODEBOOK + +#ifdef STB_VORBIS_CODEBOOK_SHORTS +#error "STB_VORBIS_CODEBOOK_SHORTS is no longer supported as it produced incorrect results for some input formats" +#endif + +// STB_VORBIS_DIVIDE_TABLE +// this replaces small integer divides in the floor decode loop with +// table lookups. made less than 1% difference, so disabled by default. + +// STB_VORBIS_NO_INLINE_DECODE +// disables the inlining of the scalar codebook fast-huffman decode. +// might save a little codespace; useful for debugging +// #define STB_VORBIS_NO_INLINE_DECODE + +// STB_VORBIS_NO_DEFER_FLOOR +// Normally we only decode the floor without synthesizing the actual +// full curve. We can instead synthesize the curve immediately. This +// requires more memory and is very likely slower, so I don't think +// you'd ever want to do it except for debugging. +// #define STB_VORBIS_NO_DEFER_FLOOR + + + + +////////////////////////////////////////////////////////////////////////////// + +#ifdef STB_VORBIS_NO_PULLDATA_API + #define STB_VORBIS_NO_INTEGER_CONVERSION + #define STB_VORBIS_NO_STDIO +#endif + +#if defined(STB_VORBIS_NO_CRT) && !defined(STB_VORBIS_NO_STDIO) + #define STB_VORBIS_NO_STDIO 1 +#endif + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT + + // only need endianness for fast-float-to-int, which we don't + // use for pushdata + + #ifndef STB_VORBIS_BIG_ENDIAN + #define STB_VORBIS_ENDIAN 0 + #else + #define STB_VORBIS_ENDIAN 1 + #endif + +#endif +#endif + + +#ifndef STB_VORBIS_NO_STDIO +#include +#endif + +#ifndef STB_VORBIS_NO_CRT + #include + #include + #include + #include + + // find definition of alloca if it's not in stdlib.h: + #if defined(_MSC_VER) || defined(__MINGW32__) + #include + #endif + #if defined(__linux__) || defined(__linux) || defined(__sun__) || defined(__EMSCRIPTEN__) || defined(__NEWLIB__) + #include + #endif +#else // STB_VORBIS_NO_CRT + #define NULL 0 + #define malloc(s) 0 + #define free(s) ((void) 0) + #define realloc(s) 0 +#endif // STB_VORBIS_NO_CRT + +#include + +#ifdef __MINGW32__ + // eff you mingw: + // "fixed": + // http://sourceforge.net/p/mingw-w64/mailman/message/32882927/ + // "no that broke the build, reverted, who cares about C": + // http://sourceforge.net/p/mingw-w64/mailman/message/32890381/ + #ifdef __forceinline + #undef __forceinline + #endif + #define __forceinline + #ifndef alloca + #define alloca __builtin_alloca + #endif +#elif !defined(_MSC_VER) + #if __GNUC__ + #define __forceinline inline + #else + #define __forceinline + #endif +#endif + +#if STB_VORBIS_MAX_CHANNELS > 256 +#error "Value of STB_VORBIS_MAX_CHANNELS outside of allowed range" +#endif + +#if STB_VORBIS_FAST_HUFFMAN_LENGTH > 24 +#error "Value of STB_VORBIS_FAST_HUFFMAN_LENGTH outside of allowed range" +#endif + + +#if 0 +#include +#define CHECK(f) _CrtIsValidHeapPointer(f->channel_buffers[1]) +#else +#define CHECK(f) ((void) 0) +#endif + +#define MAX_BLOCKSIZE_LOG 13 // from specification +#define MAX_BLOCKSIZE (1 << MAX_BLOCKSIZE_LOG) + + +typedef unsigned char uint8; +typedef signed char int8; +typedef unsigned short uint16; +typedef signed short int16; +typedef unsigned int uint32; +typedef signed int int32; + +#ifndef TRUE +#define TRUE 1 +#define FALSE 0 +#endif + +typedef float codetype; + +#ifdef _MSC_VER +#define STBV_NOTUSED(v) (void)(v) +#else +#define STBV_NOTUSED(v) (void)sizeof(v) +#endif + +// @NOTE +// +// Some arrays below are tagged "//varies", which means it's actually +// a variable-sized piece of data, but rather than malloc I assume it's +// small enough it's better to just allocate it all together with the +// main thing +// +// Most of the variables are specified with the smallest size I could pack +// them into. It might give better performance to make them all full-sized +// integers. It should be safe to freely rearrange the structures or change +// the sizes larger--nothing relies on silently truncating etc., nor the +// order of variables. + +#define FAST_HUFFMAN_TABLE_SIZE (1 << STB_VORBIS_FAST_HUFFMAN_LENGTH) +#define FAST_HUFFMAN_TABLE_MASK (FAST_HUFFMAN_TABLE_SIZE - 1) + +typedef struct +{ + int dimensions, entries; + uint8 *codeword_lengths; + float minimum_value; + float delta_value; + uint8 value_bits; + uint8 lookup_type; + uint8 sequence_p; + uint8 sparse; + uint32 lookup_values; + codetype *multiplicands; + uint32 *codewords; + #ifdef STB_VORBIS_FAST_HUFFMAN_SHORT + int16 fast_huffman[FAST_HUFFMAN_TABLE_SIZE]; + #else + int32 fast_huffman[FAST_HUFFMAN_TABLE_SIZE]; + #endif + uint32 *sorted_codewords; + int *sorted_values; + int sorted_entries; +} Codebook; + +typedef struct +{ + uint8 order; + uint16 rate; + uint16 bark_map_size; + uint8 amplitude_bits; + uint8 amplitude_offset; + uint8 number_of_books; + uint8 book_list[16]; // varies +} Floor0; + +typedef struct +{ + uint8 partitions; + uint8 partition_class_list[32]; // varies + uint8 class_dimensions[16]; // varies + uint8 class_subclasses[16]; // varies + uint8 class_masterbooks[16]; // varies + int16 subclass_books[16][8]; // varies + uint16 Xlist[31*8+2]; // varies + uint8 sorted_order[31*8+2]; + uint8 neighbors[31*8+2][2]; + uint8 floor1_multiplier; + uint8 rangebits; + int values; +} Floor1; + +typedef union +{ + Floor0 floor0; + Floor1 floor1; +} Floor; + +typedef struct +{ + uint32 begin, end; + uint32 part_size; + uint8 classifications; + uint8 classbook; + uint8 **classdata; + int16 (*residue_books)[8]; +} Residue; + +typedef struct +{ + uint8 magnitude; + uint8 angle; + uint8 mux; +} MappingChannel; + +typedef struct +{ + uint16 coupling_steps; + MappingChannel *chan; + uint8 submaps; + uint8 submap_floor[15]; // varies + uint8 submap_residue[15]; // varies +} Mapping; + +typedef struct +{ + uint8 blockflag; + uint8 mapping; + uint16 windowtype; + uint16 transformtype; +} Mode; + +typedef struct +{ + uint32 goal_crc; // expected crc if match + int bytes_left; // bytes left in packet + uint32 crc_so_far; // running crc + int bytes_done; // bytes processed in _current_ chunk + uint32 sample_loc; // granule pos encoded in page +} CRCscan; + +typedef struct +{ + uint32 page_start, page_end; + uint32 last_decoded_sample; +} ProbedPage; + +struct stb_vorbis +{ + // user-accessible info + unsigned int sample_rate; + int channels; + + unsigned int setup_memory_required; + unsigned int temp_memory_required; + unsigned int setup_temp_memory_required; + + char *vendor; + int comment_list_length; + char **comment_list; + + // input config +#ifndef STB_VORBIS_NO_STDIO + FILE *f; + uint32 f_start; + int close_on_free; +#endif + + uint8 *stream; + uint8 *stream_start; + uint8 *stream_end; + + uint32 stream_len; + + uint8 push_mode; + + // the page to seek to when seeking to start, may be zero + uint32 first_audio_page_offset; + + // p_first is the page on which the first audio packet ends + // (but not necessarily the page on which it starts) + ProbedPage p_first, p_last; + + // memory management + stb_vorbis_alloc alloc; + int setup_offset; + int temp_offset; + + // run-time results + int eof; + enum STBVorbisError error; + + // user-useful data + + // header info + int blocksize[2]; + int blocksize_0, blocksize_1; + int codebook_count; + Codebook *codebooks; + int floor_count; + uint16 floor_types[64]; // varies + Floor *floor_config; + int residue_count; + uint16 residue_types[64]; // varies + Residue *residue_config; + int mapping_count; + Mapping *mapping; + int mode_count; + Mode mode_config[64]; // varies + + uint32 total_samples; + + // decode buffer + float *channel_buffers[STB_VORBIS_MAX_CHANNELS]; + float *outputs [STB_VORBIS_MAX_CHANNELS]; + + float *previous_window[STB_VORBIS_MAX_CHANNELS]; + int previous_length; + + #ifndef STB_VORBIS_NO_DEFER_FLOOR + int16 *finalY[STB_VORBIS_MAX_CHANNELS]; + #else + float *floor_buffers[STB_VORBIS_MAX_CHANNELS]; + #endif + + uint32 current_loc; // sample location of next frame to decode + int current_loc_valid; + + // per-blocksize precomputed data + + // twiddle factors + float *A[2],*B[2],*C[2]; + float *window[2]; + uint16 *bit_reverse[2]; + + // current page/packet/segment streaming info + uint32 serial; // stream serial number for verification + int last_page; + int segment_count; + uint8 segments[255]; + uint8 page_flag; + uint8 bytes_in_seg; + uint8 first_decode; + int next_seg; + int last_seg; // flag that we're on the last segment + int last_seg_which; // what was the segment number of the last seg? + uint32 acc; + int valid_bits; + int packet_bytes; + int end_seg_with_known_loc; + uint32 known_loc_for_packet; + int discard_samples_deferred; + uint32 samples_output; + + // push mode scanning + int page_crc_tests; // only in push_mode: number of tests active; -1 if not searching +#ifndef STB_VORBIS_NO_PUSHDATA_API + CRCscan scan[STB_VORBIS_PUSHDATA_CRC_COUNT]; +#endif + + // sample-access + int channel_buffer_start; + int channel_buffer_end; +}; + +#if defined(STB_VORBIS_NO_PUSHDATA_API) + #define IS_PUSH_MODE(f) FALSE +#elif defined(STB_VORBIS_NO_PULLDATA_API) + #define IS_PUSH_MODE(f) TRUE +#else + #define IS_PUSH_MODE(f) ((f)->push_mode) +#endif + +typedef struct stb_vorbis vorb; + +static int error(vorb *f, enum STBVorbisError e) +{ + f->error = e; + if (!f->eof && e != VORBIS_need_more_data) { + f->error=e; // breakpoint for debugging + } + return 0; +} + + +// these functions are used for allocating temporary memory +// while decoding. if you can afford the stack space, use +// alloca(); otherwise, provide a temp buffer and it will +// allocate out of those. + +#define array_size_required(count,size) (count*(sizeof(void *)+(size))) + +#define temp_alloc(f,size) (f->alloc.alloc_buffer ? setup_temp_malloc(f,size) : alloca(size)) +#define temp_free(f,p) (void)0 +#define temp_alloc_save(f) ((f)->temp_offset) +#define temp_alloc_restore(f,p) ((f)->temp_offset = (p)) + +#define temp_block_array(f,count,size) make_block_array(temp_alloc(f,array_size_required(count,size)), count, size) + +// given a sufficiently large block of memory, make an array of pointers to subblocks of it +static void *make_block_array(void *mem, int count, int size) +{ + int i; + void ** p = (void **) mem; + char *q = (char *) (p + count); + for (i=0; i < count; ++i) { + p[i] = q; + q += size; + } + return p; +} + +static void *setup_malloc(vorb *f, int sz) +{ + sz = (sz+7) & ~7; // round up to nearest 8 for alignment of future allocs. + f->setup_memory_required += sz; + if (f->alloc.alloc_buffer) { + void *p = (char *) f->alloc.alloc_buffer + f->setup_offset; + if (f->setup_offset + sz > f->temp_offset) return NULL; + f->setup_offset += sz; + return p; + } + return sz ? malloc(sz) : NULL; +} + +static void setup_free(vorb *f, void *p) +{ + if (f->alloc.alloc_buffer) return; // do nothing; setup mem is a stack + free(p); +} + +static void *setup_temp_malloc(vorb *f, int sz) +{ + sz = (sz+7) & ~7; // round up to nearest 8 for alignment of future allocs. + if (f->alloc.alloc_buffer) { + if (f->temp_offset - sz < f->setup_offset) return NULL; + f->temp_offset -= sz; + return (char *) f->alloc.alloc_buffer + f->temp_offset; + } + return malloc(sz); +} + +static void setup_temp_free(vorb *f, void *p, int sz) +{ + if (f->alloc.alloc_buffer) { + f->temp_offset += (sz+7)&~7; + return; + } + free(p); +} + +#define CRC32_POLY 0x04c11db7 // from spec + +static uint32 crc_table[256]; +static void crc32_init(void) +{ + int i,j; + uint32 s; + for(i=0; i < 256; i++) { + for (s=(uint32) i << 24, j=0; j < 8; ++j) + s = (s << 1) ^ (s >= (1U<<31) ? CRC32_POLY : 0); + crc_table[i] = s; + } +} + +static __forceinline uint32 crc32_update(uint32 crc, uint8 byte) +{ + return (crc << 8) ^ crc_table[byte ^ (crc >> 24)]; +} + + +// used in setup, and for huffman that doesn't go fast path +static unsigned int bit_reverse(unsigned int n) +{ + n = ((n & 0xAAAAAAAA) >> 1) | ((n & 0x55555555) << 1); + n = ((n & 0xCCCCCCCC) >> 2) | ((n & 0x33333333) << 2); + n = ((n & 0xF0F0F0F0) >> 4) | ((n & 0x0F0F0F0F) << 4); + n = ((n & 0xFF00FF00) >> 8) | ((n & 0x00FF00FF) << 8); + return (n >> 16) | (n << 16); +} + +static float square(float x) +{ + return x*x; +} + +// this is a weird definition of log2() for which log2(1) = 1, log2(2) = 2, log2(4) = 3 +// as required by the specification. fast(?) implementation from stb.h +// @OPTIMIZE: called multiple times per-packet with "constants"; move to setup +static int ilog(int32 n) +{ + static signed char log2_4[16] = { 0,1,2,2,3,3,3,3,4,4,4,4,4,4,4,4 }; + + if (n < 0) return 0; // signed n returns 0 + + // 2 compares if n < 16, 3 compares otherwise (4 if signed or n > 1<<29) + if (n < (1 << 14)) + if (n < (1 << 4)) return 0 + log2_4[n ]; + else if (n < (1 << 9)) return 5 + log2_4[n >> 5]; + else return 10 + log2_4[n >> 10]; + else if (n < (1 << 24)) + if (n < (1 << 19)) return 15 + log2_4[n >> 15]; + else return 20 + log2_4[n >> 20]; + else if (n < (1 << 29)) return 25 + log2_4[n >> 25]; + else return 30 + log2_4[n >> 30]; +} + +#ifndef M_PI + #define M_PI 3.14159265358979323846264f // from CRC +#endif + +// code length assigned to a value with no huffman encoding +#define NO_CODE 255 + +/////////////////////// LEAF SETUP FUNCTIONS ////////////////////////// +// +// these functions are only called at setup, and only a few times +// per file + +static float float32_unpack(uint32 x) +{ + // from the specification + uint32 mantissa = x & 0x1fffff; + uint32 sign = x & 0x80000000; + uint32 exp = (x & 0x7fe00000) >> 21; + double res = sign ? -(double)mantissa : (double)mantissa; + return (float) ldexp((float)res, (int)exp-788); +} + + +// zlib & jpeg huffman tables assume that the output symbols +// can either be arbitrarily arranged, or have monotonically +// increasing frequencies--they rely on the lengths being sorted; +// this makes for a very simple generation algorithm. +// vorbis allows a huffman table with non-sorted lengths. This +// requires a more sophisticated construction, since symbols in +// order do not map to huffman codes "in order". +static void add_entry(Codebook *c, uint32 huff_code, int symbol, int count, int len, uint32 *values) +{ + if (!c->sparse) { + c->codewords [symbol] = huff_code; + } else { + c->codewords [count] = huff_code; + c->codeword_lengths[count] = len; + values [count] = symbol; + } +} + +static int compute_codewords(Codebook *c, uint8 *len, int n, uint32 *values) +{ + int i,k,m=0; + uint32 available[32]; + + memset(available, 0, sizeof(available)); + // find the first entry + for (k=0; k < n; ++k) if (len[k] < NO_CODE) break; + if (k == n) { assert(c->sorted_entries == 0); return TRUE; } + assert(len[k] < 32); // no error return required, code reading lens checks this + // add to the list + add_entry(c, 0, k, m++, len[k], values); + // add all available leaves + for (i=1; i <= len[k]; ++i) + available[i] = 1U << (32-i); + // note that the above code treats the first case specially, + // but it's really the same as the following code, so they + // could probably be combined (except the initial code is 0, + // and I use 0 in available[] to mean 'empty') + for (i=k+1; i < n; ++i) { + uint32 res; + int z = len[i], y; + if (z == NO_CODE) continue; + assert(z < 32); // no error return required, code reading lens checks this + // find lowest available leaf (should always be earliest, + // which is what the specification calls for) + // note that this property, and the fact we can never have + // more than one free leaf at a given level, isn't totally + // trivial to prove, but it seems true and the assert never + // fires, so! + while (z > 0 && !available[z]) --z; + if (z == 0) { return FALSE; } + res = available[z]; + available[z] = 0; + add_entry(c, bit_reverse(res), i, m++, len[i], values); + // propagate availability up the tree + if (z != len[i]) { + for (y=len[i]; y > z; --y) { + assert(available[y] == 0); + available[y] = res + (1 << (32-y)); + } + } + } + return TRUE; +} + +// accelerated huffman table allows fast O(1) match of all symbols +// of length <= STB_VORBIS_FAST_HUFFMAN_LENGTH +static void compute_accelerated_huffman(Codebook *c) +{ + int i, len; + for (i=0; i < FAST_HUFFMAN_TABLE_SIZE; ++i) + c->fast_huffman[i] = -1; + + len = c->sparse ? c->sorted_entries : c->entries; + #ifdef STB_VORBIS_FAST_HUFFMAN_SHORT + if (len > 32767) len = 32767; // largest possible value we can encode! + #endif + for (i=0; i < len; ++i) { + if (c->codeword_lengths[i] <= STB_VORBIS_FAST_HUFFMAN_LENGTH) { + uint32 z = c->sparse ? bit_reverse(c->sorted_codewords[i]) : c->codewords[i]; + // set table entries for all bit combinations in the higher bits + while (z < FAST_HUFFMAN_TABLE_SIZE) { + c->fast_huffman[z] = i; + z += 1 << c->codeword_lengths[i]; + } + } + } +} + +#ifdef _MSC_VER +#define STBV_CDECL __cdecl +#else +#define STBV_CDECL +#endif + +static int STBV_CDECL uint32_compare(const void *p, const void *q) +{ + uint32 x = * (uint32 *) p; + uint32 y = * (uint32 *) q; + return x < y ? -1 : x > y; +} + +static int include_in_sort(Codebook *c, uint8 len) +{ + if (c->sparse) { assert(len != NO_CODE); return TRUE; } + if (len == NO_CODE) return FALSE; + if (len > STB_VORBIS_FAST_HUFFMAN_LENGTH) return TRUE; + return FALSE; +} + +// if the fast table above doesn't work, we want to binary +// search them... need to reverse the bits +static void compute_sorted_huffman(Codebook *c, uint8 *lengths, uint32 *values) +{ + int i, len; + // build a list of all the entries + // OPTIMIZATION: don't include the short ones, since they'll be caught by FAST_HUFFMAN. + // this is kind of a frivolous optimization--I don't see any performance improvement, + // but it's like 4 extra lines of code, so. + if (!c->sparse) { + int k = 0; + for (i=0; i < c->entries; ++i) + if (include_in_sort(c, lengths[i])) + c->sorted_codewords[k++] = bit_reverse(c->codewords[i]); + assert(k == c->sorted_entries); + } else { + for (i=0; i < c->sorted_entries; ++i) + c->sorted_codewords[i] = bit_reverse(c->codewords[i]); + } + + qsort(c->sorted_codewords, c->sorted_entries, sizeof(c->sorted_codewords[0]), uint32_compare); + c->sorted_codewords[c->sorted_entries] = 0xffffffff; + + len = c->sparse ? c->sorted_entries : c->entries; + // now we need to indicate how they correspond; we could either + // #1: sort a different data structure that says who they correspond to + // #2: for each sorted entry, search the original list to find who corresponds + // #3: for each original entry, find the sorted entry + // #1 requires extra storage, #2 is slow, #3 can use binary search! + for (i=0; i < len; ++i) { + int huff_len = c->sparse ? lengths[values[i]] : lengths[i]; + if (include_in_sort(c,huff_len)) { + uint32 code = bit_reverse(c->codewords[i]); + int x=0, n=c->sorted_entries; + while (n > 1) { + // invariant: sc[x] <= code < sc[x+n] + int m = x + (n >> 1); + if (c->sorted_codewords[m] <= code) { + x = m; + n -= (n>>1); + } else { + n >>= 1; + } + } + assert(c->sorted_codewords[x] == code); + if (c->sparse) { + c->sorted_values[x] = values[i]; + c->codeword_lengths[x] = huff_len; + } else { + c->sorted_values[x] = i; + } + } + } +} + +// only run while parsing the header (3 times) +static int vorbis_validate(uint8 *data) +{ + static uint8 vorbis[6] = { 'v', 'o', 'r', 'b', 'i', 's' }; + return memcmp(data, vorbis, 6) == 0; +} + +// called from setup only, once per code book +// (formula implied by specification) +static int lookup1_values(int entries, int dim) +{ + int r = (int) floor(exp((float) log((float) entries) / dim)); + if ((int) floor(pow((float) r+1, dim)) <= entries) // (int) cast for MinGW warning; + ++r; // floor() to avoid _ftol() when non-CRT + if (pow((float) r+1, dim) <= entries) + return -1; + if ((int) floor(pow((float) r, dim)) > entries) + return -1; + return r; +} + +// called twice per file +static void compute_twiddle_factors(int n, float *A, float *B, float *C) +{ + int n4 = n >> 2, n8 = n >> 3; + int k,k2; + + for (k=k2=0; k < n4; ++k,k2+=2) { + A[k2 ] = (float) cos(4*k*M_PI/n); + A[k2+1] = (float) -sin(4*k*M_PI/n); + B[k2 ] = (float) cos((k2+1)*M_PI/n/2) * 0.5f; + B[k2+1] = (float) sin((k2+1)*M_PI/n/2) * 0.5f; + } + for (k=k2=0; k < n8; ++k,k2+=2) { + C[k2 ] = (float) cos(2*(k2+1)*M_PI/n); + C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n); + } +} + +static void compute_window(int n, float *window) +{ + int n2 = n >> 1, i; + for (i=0; i < n2; ++i) + window[i] = (float) sin(0.5 * M_PI * square((float) sin((i - 0 + 0.5) / n2 * 0.5 * M_PI))); +} + +static void compute_bitreverse(int n, uint16 *rev) +{ + int ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + int i, n8 = n >> 3; + for (i=0; i < n8; ++i) + rev[i] = (bit_reverse(i) >> (32-ld+3)) << 2; +} + +static int init_blocksize(vorb *f, int b, int n) +{ + int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3; + f->A[b] = (float *) setup_malloc(f, sizeof(float) * n2); + f->B[b] = (float *) setup_malloc(f, sizeof(float) * n2); + f->C[b] = (float *) setup_malloc(f, sizeof(float) * n4); + if (!f->A[b] || !f->B[b] || !f->C[b]) return error(f, VORBIS_outofmem); + compute_twiddle_factors(n, f->A[b], f->B[b], f->C[b]); + f->window[b] = (float *) setup_malloc(f, sizeof(float) * n2); + if (!f->window[b]) return error(f, VORBIS_outofmem); + compute_window(n, f->window[b]); + f->bit_reverse[b] = (uint16 *) setup_malloc(f, sizeof(uint16) * n8); + if (!f->bit_reverse[b]) return error(f, VORBIS_outofmem); + compute_bitreverse(n, f->bit_reverse[b]); + return TRUE; +} + +static void neighbors(uint16 *x, int n, int *plow, int *phigh) +{ + int low = -1; + int high = 65536; + int i; + for (i=0; i < n; ++i) { + if (x[i] > low && x[i] < x[n]) { *plow = i; low = x[i]; } + if (x[i] < high && x[i] > x[n]) { *phigh = i; high = x[i]; } + } +} + +// this has been repurposed so y is now the original index instead of y +typedef struct +{ + uint16 x,id; +} stbv__floor_ordering; + +static int STBV_CDECL point_compare(const void *p, const void *q) +{ + stbv__floor_ordering *a = (stbv__floor_ordering *) p; + stbv__floor_ordering *b = (stbv__floor_ordering *) q; + return a->x < b->x ? -1 : a->x > b->x; +} + +// +/////////////////////// END LEAF SETUP FUNCTIONS ////////////////////////// + + +#if defined(STB_VORBIS_NO_STDIO) + #define USE_MEMORY(z) TRUE +#else + #define USE_MEMORY(z) ((z)->stream) +#endif + +static uint8 get8(vorb *z) +{ + if (USE_MEMORY(z)) { + if (z->stream >= z->stream_end) { z->eof = TRUE; return 0; } + return *z->stream++; + } + + #ifndef STB_VORBIS_NO_STDIO + { + int c = fgetc(z->f); + if (c == EOF) { z->eof = TRUE; return 0; } + return c; + } + #endif +} + +static uint32 get32(vorb *f) +{ + uint32 x; + x = get8(f); + x += get8(f) << 8; + x += get8(f) << 16; + x += (uint32) get8(f) << 24; + return x; +} + +static int getn(vorb *z, uint8 *data, int n) +{ + if (USE_MEMORY(z)) { + if (z->stream+n > z->stream_end) { z->eof = 1; return 0; } + memcpy(data, z->stream, n); + z->stream += n; + return 1; + } + + #ifndef STB_VORBIS_NO_STDIO + if (fread(data, n, 1, z->f) == 1) + return 1; + else { + z->eof = 1; + return 0; + } + #endif +} + +static void skip(vorb *z, int n) +{ + if (USE_MEMORY(z)) { + z->stream += n; + if (z->stream >= z->stream_end) z->eof = 1; + return; + } + #ifndef STB_VORBIS_NO_STDIO + { + long x = ftell(z->f); + fseek(z->f, x+n, SEEK_SET); + } + #endif +} + +static int set_file_offset(stb_vorbis *f, unsigned int loc) +{ + #ifndef STB_VORBIS_NO_PUSHDATA_API + if (f->push_mode) return 0; + #endif + f->eof = 0; + if (USE_MEMORY(f)) { + if (f->stream_start + loc >= f->stream_end || f->stream_start + loc < f->stream_start) { + f->stream = f->stream_end; + f->eof = 1; + return 0; + } else { + f->stream = f->stream_start + loc; + return 1; + } + } + #ifndef STB_VORBIS_NO_STDIO + if (loc + f->f_start < loc || loc >= 0x80000000) { + loc = 0x7fffffff; + f->eof = 1; + } else { + loc += f->f_start; + } + if (!fseek(f->f, loc, SEEK_SET)) + return 1; + f->eof = 1; + fseek(f->f, f->f_start, SEEK_END); + return 0; + #endif +} + + +static uint8 ogg_page_header[4] = { 0x4f, 0x67, 0x67, 0x53 }; + +static int capture_pattern(vorb *f) +{ + if (0x4f != get8(f)) return FALSE; + if (0x67 != get8(f)) return FALSE; + if (0x67 != get8(f)) return FALSE; + if (0x53 != get8(f)) return FALSE; + return TRUE; +} + +#define PAGEFLAG_continued_packet 1 +#define PAGEFLAG_first_page 2 +#define PAGEFLAG_last_page 4 + +static int start_page_no_capturepattern(vorb *f) +{ + uint32 loc0,loc1,n; + if (f->first_decode && !IS_PUSH_MODE(f)) { + f->p_first.page_start = stb_vorbis_get_file_offset(f) - 4; + } + // stream structure version + if (0 != get8(f)) return error(f, VORBIS_invalid_stream_structure_version); + // header flag + f->page_flag = get8(f); + // absolute granule position + loc0 = get32(f); + loc1 = get32(f); + // @TODO: validate loc0,loc1 as valid positions? + // stream serial number -- vorbis doesn't interleave, so discard + get32(f); + //if (f->serial != get32(f)) return error(f, VORBIS_incorrect_stream_serial_number); + // page sequence number + n = get32(f); + f->last_page = n; + // CRC32 + get32(f); + // page_segments + f->segment_count = get8(f); + if (!getn(f, f->segments, f->segment_count)) + return error(f, VORBIS_unexpected_eof); + // assume we _don't_ know any the sample position of any segments + f->end_seg_with_known_loc = -2; + if (loc0 != ~0U || loc1 != ~0U) { + int i; + // determine which packet is the last one that will complete + for (i=f->segment_count-1; i >= 0; --i) + if (f->segments[i] < 255) + break; + // 'i' is now the index of the _last_ segment of a packet that ends + if (i >= 0) { + f->end_seg_with_known_loc = i; + f->known_loc_for_packet = loc0; + } + } + if (f->first_decode) { + int i,len; + len = 0; + for (i=0; i < f->segment_count; ++i) + len += f->segments[i]; + len += 27 + f->segment_count; + f->p_first.page_end = f->p_first.page_start + len; + f->p_first.last_decoded_sample = loc0; + } + f->next_seg = 0; + return TRUE; +} + +static int start_page(vorb *f) +{ + if (!capture_pattern(f)) return error(f, VORBIS_missing_capture_pattern); + return start_page_no_capturepattern(f); +} + +static int start_packet(vorb *f) +{ + while (f->next_seg == -1) { + if (!start_page(f)) return FALSE; + if (f->page_flag & PAGEFLAG_continued_packet) + return error(f, VORBIS_continued_packet_flag_invalid); + } + f->last_seg = FALSE; + f->valid_bits = 0; + f->packet_bytes = 0; + f->bytes_in_seg = 0; + // f->next_seg is now valid + return TRUE; +} + +static int maybe_start_packet(vorb *f) +{ + if (f->next_seg == -1) { + int x = get8(f); + if (f->eof) return FALSE; // EOF at page boundary is not an error! + if (0x4f != x ) return error(f, VORBIS_missing_capture_pattern); + if (0x67 != get8(f)) return error(f, VORBIS_missing_capture_pattern); + if (0x67 != get8(f)) return error(f, VORBIS_missing_capture_pattern); + if (0x53 != get8(f)) return error(f, VORBIS_missing_capture_pattern); + if (!start_page_no_capturepattern(f)) return FALSE; + if (f->page_flag & PAGEFLAG_continued_packet) { + // set up enough state that we can read this packet if we want, + // e.g. during recovery + f->last_seg = FALSE; + f->bytes_in_seg = 0; + return error(f, VORBIS_continued_packet_flag_invalid); + } + } + return start_packet(f); +} + +static int next_segment(vorb *f) +{ + int len; + if (f->last_seg) return 0; + if (f->next_seg == -1) { + f->last_seg_which = f->segment_count-1; // in case start_page fails + if (!start_page(f)) { f->last_seg = 1; return 0; } + if (!(f->page_flag & PAGEFLAG_continued_packet)) return error(f, VORBIS_continued_packet_flag_invalid); + } + len = f->segments[f->next_seg++]; + if (len < 255) { + f->last_seg = TRUE; + f->last_seg_which = f->next_seg-1; + } + if (f->next_seg >= f->segment_count) + f->next_seg = -1; + assert(f->bytes_in_seg == 0); + f->bytes_in_seg = len; + return len; +} + +#define EOP (-1) +#define INVALID_BITS (-1) + +static int get8_packet_raw(vorb *f) +{ + if (!f->bytes_in_seg) { // CLANG! + if (f->last_seg) return EOP; + else if (!next_segment(f)) return EOP; + } + assert(f->bytes_in_seg > 0); + --f->bytes_in_seg; + ++f->packet_bytes; + return get8(f); +} + +static int get8_packet(vorb *f) +{ + int x = get8_packet_raw(f); + f->valid_bits = 0; + return x; +} + +static int get32_packet(vorb *f) +{ + uint32 x; + x = get8_packet(f); + x += get8_packet(f) << 8; + x += get8_packet(f) << 16; + x += (uint32) get8_packet(f) << 24; + return x; +} + +static void flush_packet(vorb *f) +{ + while (get8_packet_raw(f) != EOP); +} + +// @OPTIMIZE: this is the secondary bit decoder, so it's probably not as important +// as the huffman decoder? +static uint32 get_bits(vorb *f, int n) +{ + uint32 z; + + if (f->valid_bits < 0) return 0; + if (f->valid_bits < n) { + if (n > 24) { + // the accumulator technique below would not work correctly in this case + z = get_bits(f, 24); + z += get_bits(f, n-24) << 24; + return z; + } + if (f->valid_bits == 0) f->acc = 0; + while (f->valid_bits < n) { + int z = get8_packet_raw(f); + if (z == EOP) { + f->valid_bits = INVALID_BITS; + return 0; + } + f->acc += z << f->valid_bits; + f->valid_bits += 8; + } + } + + assert(f->valid_bits >= n); + z = f->acc & ((1 << n)-1); + f->acc >>= n; + f->valid_bits -= n; + return z; +} + +// @OPTIMIZE: primary accumulator for huffman +// expand the buffer to as many bits as possible without reading off end of packet +// it might be nice to allow f->valid_bits and f->acc to be stored in registers, +// e.g. cache them locally and decode locally +static __forceinline void prep_huffman(vorb *f) +{ + if (f->valid_bits <= 24) { + if (f->valid_bits == 0) f->acc = 0; + do { + int z; + if (f->last_seg && !f->bytes_in_seg) return; + z = get8_packet_raw(f); + if (z == EOP) return; + f->acc += (unsigned) z << f->valid_bits; + f->valid_bits += 8; + } while (f->valid_bits <= 24); + } +} + +enum +{ + VORBIS_packet_id = 1, + VORBIS_packet_comment = 3, + VORBIS_packet_setup = 5 +}; + +static int codebook_decode_scalar_raw(vorb *f, Codebook *c) +{ + int i; + prep_huffman(f); + + if (c->codewords == NULL && c->sorted_codewords == NULL) + return -1; + + // cases to use binary search: sorted_codewords && !c->codewords + // sorted_codewords && c->entries > 8 + if (c->entries > 8 ? c->sorted_codewords!=NULL : !c->codewords) { + // binary search + uint32 code = bit_reverse(f->acc); + int x=0, n=c->sorted_entries, len; + + while (n > 1) { + // invariant: sc[x] <= code < sc[x+n] + int m = x + (n >> 1); + if (c->sorted_codewords[m] <= code) { + x = m; + n -= (n>>1); + } else { + n >>= 1; + } + } + // x is now the sorted index + if (!c->sparse) x = c->sorted_values[x]; + // x is now sorted index if sparse, or symbol otherwise + len = c->codeword_lengths[x]; + if (f->valid_bits >= len) { + f->acc >>= len; + f->valid_bits -= len; + return x; + } + + f->valid_bits = 0; + return -1; + } + + // if small, linear search + assert(!c->sparse); + for (i=0; i < c->entries; ++i) { + if (c->codeword_lengths[i] == NO_CODE) continue; + if (c->codewords[i] == (f->acc & ((1 << c->codeword_lengths[i])-1))) { + if (f->valid_bits >= c->codeword_lengths[i]) { + f->acc >>= c->codeword_lengths[i]; + f->valid_bits -= c->codeword_lengths[i]; + return i; + } + f->valid_bits = 0; + return -1; + } + } + + error(f, VORBIS_invalid_stream); + f->valid_bits = 0; + return -1; +} + +#ifndef STB_VORBIS_NO_INLINE_DECODE + +#define DECODE_RAW(var, f,c) \ + if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) \ + prep_huffman(f); \ + var = f->acc & FAST_HUFFMAN_TABLE_MASK; \ + var = c->fast_huffman[var]; \ + if (var >= 0) { \ + int n = c->codeword_lengths[var]; \ + f->acc >>= n; \ + f->valid_bits -= n; \ + if (f->valid_bits < 0) { f->valid_bits = 0; var = -1; } \ + } else { \ + var = codebook_decode_scalar_raw(f,c); \ + } + +#else + +static int codebook_decode_scalar(vorb *f, Codebook *c) +{ + int i; + if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) + prep_huffman(f); + // fast huffman table lookup + i = f->acc & FAST_HUFFMAN_TABLE_MASK; + i = c->fast_huffman[i]; + if (i >= 0) { + f->acc >>= c->codeword_lengths[i]; + f->valid_bits -= c->codeword_lengths[i]; + if (f->valid_bits < 0) { f->valid_bits = 0; return -1; } + return i; + } + return codebook_decode_scalar_raw(f,c); +} + +#define DECODE_RAW(var,f,c) var = codebook_decode_scalar(f,c); + +#endif + +#define DECODE(var,f,c) \ + DECODE_RAW(var,f,c) \ + if (c->sparse) var = c->sorted_values[var]; + +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + #define DECODE_VQ(var,f,c) DECODE_RAW(var,f,c) +#else + #define DECODE_VQ(var,f,c) DECODE(var,f,c) +#endif + + + + + + +// CODEBOOK_ELEMENT_FAST is an optimization for the CODEBOOK_FLOATS case +// where we avoid one addition +#define CODEBOOK_ELEMENT(c,off) (c->multiplicands[off]) +#define CODEBOOK_ELEMENT_FAST(c,off) (c->multiplicands[off]) +#define CODEBOOK_ELEMENT_BASE(c) (0) + +static int codebook_decode_start(vorb *f, Codebook *c) +{ + int z = -1; + + // type 0 is only legal in a scalar context + if (c->lookup_type == 0) + error(f, VORBIS_invalid_stream); + else { + DECODE_VQ(z,f,c); + if (c->sparse) assert(z < c->sorted_entries); + if (z < 0) { // check for EOP + if (!f->bytes_in_seg) + if (f->last_seg) + return z; + error(f, VORBIS_invalid_stream); + } + } + return z; +} + +static int codebook_decode(vorb *f, Codebook *c, float *output, int len) +{ + int i,z = codebook_decode_start(f,c); + if (z < 0) return FALSE; + if (len > c->dimensions) len = c->dimensions; + +#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + float last = CODEBOOK_ELEMENT_BASE(c); + int div = 1; + for (i=0; i < len; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c,off) + last; + output[i] += val; + if (c->sequence_p) last = val + c->minimum_value; + div *= c->lookup_values; + } + return TRUE; + } +#endif + + z *= c->dimensions; + if (c->sequence_p) { + float last = CODEBOOK_ELEMENT_BASE(c); + for (i=0; i < len; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + output[i] += val; + last = val + c->minimum_value; + } + } else { + float last = CODEBOOK_ELEMENT_BASE(c); + for (i=0; i < len; ++i) { + output[i] += CODEBOOK_ELEMENT_FAST(c,z+i) + last; + } + } + + return TRUE; +} + +static int codebook_decode_step(vorb *f, Codebook *c, float *output, int len, int step) +{ + int i,z = codebook_decode_start(f,c); + float last = CODEBOOK_ELEMENT_BASE(c); + if (z < 0) return FALSE; + if (len > c->dimensions) len = c->dimensions; + +#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int div = 1; + for (i=0; i < len; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c,off) + last; + output[i*step] += val; + if (c->sequence_p) last = val; + div *= c->lookup_values; + } + return TRUE; + } +#endif + + z *= c->dimensions; + for (i=0; i < len; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + output[i*step] += val; + if (c->sequence_p) last = val; + } + + return TRUE; +} + +static int codebook_decode_deinterleave_repeat(vorb *f, Codebook *c, float **outputs, int ch, int *c_inter_p, int *p_inter_p, int len, int total_decode) +{ + int c_inter = *c_inter_p; + int p_inter = *p_inter_p; + int i,z, effective = c->dimensions; + + // type 0 is only legal in a scalar context + if (c->lookup_type == 0) return error(f, VORBIS_invalid_stream); + + while (total_decode > 0) { + float last = CODEBOOK_ELEMENT_BASE(c); + DECODE_VQ(z,f,c); + #ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + assert(!c->sparse || z < c->sorted_entries); + #endif + if (z < 0) { + if (!f->bytes_in_seg) + if (f->last_seg) return FALSE; + return error(f, VORBIS_invalid_stream); + } + + // if this will take us off the end of the buffers, stop short! + // we check by computing the length of the virtual interleaved + // buffer (len*ch), our current offset within it (p_inter*ch)+(c_inter), + // and the length we'll be using (effective) + if (c_inter + p_inter*ch + effective > len * ch) { + effective = len*ch - (p_inter*ch - c_inter); + } + + #ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int div = 1; + for (i=0; i < effective; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c,off) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { c_inter = 0; ++p_inter; } + if (c->sequence_p) last = val; + div *= c->lookup_values; + } + } else + #endif + { + z *= c->dimensions; + if (c->sequence_p) { + for (i=0; i < effective; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { c_inter = 0; ++p_inter; } + last = val; + } + } else { + for (i=0; i < effective; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { c_inter = 0; ++p_inter; } + } + } + } + + total_decode -= effective; + } + *c_inter_p = c_inter; + *p_inter_p = p_inter; + return TRUE; +} + +static int predict_point(int x, int x0, int x1, int y0, int y1) +{ + int dy = y1 - y0; + int adx = x1 - x0; + // @OPTIMIZE: force int division to round in the right direction... is this necessary on x86? + int err = abs(dy) * (x - x0); + int off = err / adx; + return dy < 0 ? y0 - off : y0 + off; +} + +// the following table is block-copied from the specification +static float inverse_db_table[256] = +{ + 1.0649863e-07f, 1.1341951e-07f, 1.2079015e-07f, 1.2863978e-07f, + 1.3699951e-07f, 1.4590251e-07f, 1.5538408e-07f, 1.6548181e-07f, + 1.7623575e-07f, 1.8768855e-07f, 1.9988561e-07f, 2.1287530e-07f, + 2.2670913e-07f, 2.4144197e-07f, 2.5713223e-07f, 2.7384213e-07f, + 2.9163793e-07f, 3.1059021e-07f, 3.3077411e-07f, 3.5226968e-07f, + 3.7516214e-07f, 3.9954229e-07f, 4.2550680e-07f, 4.5315863e-07f, + 4.8260743e-07f, 5.1396998e-07f, 5.4737065e-07f, 5.8294187e-07f, + 6.2082472e-07f, 6.6116941e-07f, 7.0413592e-07f, 7.4989464e-07f, + 7.9862701e-07f, 8.5052630e-07f, 9.0579828e-07f, 9.6466216e-07f, + 1.0273513e-06f, 1.0941144e-06f, 1.1652161e-06f, 1.2409384e-06f, + 1.3215816e-06f, 1.4074654e-06f, 1.4989305e-06f, 1.5963394e-06f, + 1.7000785e-06f, 1.8105592e-06f, 1.9282195e-06f, 2.0535261e-06f, + 2.1869758e-06f, 2.3290978e-06f, 2.4804557e-06f, 2.6416497e-06f, + 2.8133190e-06f, 2.9961443e-06f, 3.1908506e-06f, 3.3982101e-06f, + 3.6190449e-06f, 3.8542308e-06f, 4.1047004e-06f, 4.3714470e-06f, + 4.6555282e-06f, 4.9580707e-06f, 5.2802740e-06f, 5.6234160e-06f, + 5.9888572e-06f, 6.3780469e-06f, 6.7925283e-06f, 7.2339451e-06f, + 7.7040476e-06f, 8.2047000e-06f, 8.7378876e-06f, 9.3057248e-06f, + 9.9104632e-06f, 1.0554501e-05f, 1.1240392e-05f, 1.1970856e-05f, + 1.2748789e-05f, 1.3577278e-05f, 1.4459606e-05f, 1.5399272e-05f, + 1.6400004e-05f, 1.7465768e-05f, 1.8600792e-05f, 1.9809576e-05f, + 2.1096914e-05f, 2.2467911e-05f, 2.3928002e-05f, 2.5482978e-05f, + 2.7139006e-05f, 2.8902651e-05f, 3.0780908e-05f, 3.2781225e-05f, + 3.4911534e-05f, 3.7180282e-05f, 3.9596466e-05f, 4.2169667e-05f, + 4.4910090e-05f, 4.7828601e-05f, 5.0936773e-05f, 5.4246931e-05f, + 5.7772202e-05f, 6.1526565e-05f, 6.5524908e-05f, 6.9783085e-05f, + 7.4317983e-05f, 7.9147585e-05f, 8.4291040e-05f, 8.9768747e-05f, + 9.5602426e-05f, 0.00010181521f, 0.00010843174f, 0.00011547824f, + 0.00012298267f, 0.00013097477f, 0.00013948625f, 0.00014855085f, + 0.00015820453f, 0.00016848555f, 0.00017943469f, 0.00019109536f, + 0.00020351382f, 0.00021673929f, 0.00023082423f, 0.00024582449f, + 0.00026179955f, 0.00027881276f, 0.00029693158f, 0.00031622787f, + 0.00033677814f, 0.00035866388f, 0.00038197188f, 0.00040679456f, + 0.00043323036f, 0.00046138411f, 0.00049136745f, 0.00052329927f, + 0.00055730621f, 0.00059352311f, 0.00063209358f, 0.00067317058f, + 0.00071691700f, 0.00076350630f, 0.00081312324f, 0.00086596457f, + 0.00092223983f, 0.00098217216f, 0.0010459992f, 0.0011139742f, + 0.0011863665f, 0.0012634633f, 0.0013455702f, 0.0014330129f, + 0.0015261382f, 0.0016253153f, 0.0017309374f, 0.0018434235f, + 0.0019632195f, 0.0020908006f, 0.0022266726f, 0.0023713743f, + 0.0025254795f, 0.0026895994f, 0.0028643847f, 0.0030505286f, + 0.0032487691f, 0.0034598925f, 0.0036847358f, 0.0039241906f, + 0.0041792066f, 0.0044507950f, 0.0047400328f, 0.0050480668f, + 0.0053761186f, 0.0057254891f, 0.0060975636f, 0.0064938176f, + 0.0069158225f, 0.0073652516f, 0.0078438871f, 0.0083536271f, + 0.0088964928f, 0.009474637f, 0.010090352f, 0.010746080f, + 0.011444421f, 0.012188144f, 0.012980198f, 0.013823725f, + 0.014722068f, 0.015678791f, 0.016697687f, 0.017782797f, + 0.018938423f, 0.020169149f, 0.021479854f, 0.022875735f, + 0.024362330f, 0.025945531f, 0.027631618f, 0.029427276f, + 0.031339626f, 0.033376252f, 0.035545228f, 0.037855157f, + 0.040315199f, 0.042935108f, 0.045725273f, 0.048696758f, + 0.051861348f, 0.055231591f, 0.058820850f, 0.062643361f, + 0.066714279f, 0.071049749f, 0.075666962f, 0.080584227f, + 0.085821044f, 0.091398179f, 0.097337747f, 0.10366330f, + 0.11039993f, 0.11757434f, 0.12521498f, 0.13335215f, + 0.14201813f, 0.15124727f, 0.16107617f, 0.17154380f, + 0.18269168f, 0.19456402f, 0.20720788f, 0.22067342f, + 0.23501402f, 0.25028656f, 0.26655159f, 0.28387361f, + 0.30232132f, 0.32196786f, 0.34289114f, 0.36517414f, + 0.38890521f, 0.41417847f, 0.44109412f, 0.46975890f, + 0.50028648f, 0.53279791f, 0.56742212f, 0.60429640f, + 0.64356699f, 0.68538959f, 0.72993007f, 0.77736504f, + 0.82788260f, 0.88168307f, 0.9389798f, 1.0f +}; + + +// @OPTIMIZE: if you want to replace this bresenham line-drawing routine, +// note that you must produce bit-identical output to decode correctly; +// this specific sequence of operations is specified in the spec (it's +// drawing integer-quantized frequency-space lines that the encoder +// expects to be exactly the same) +// ... also, isn't the whole point of Bresenham's algorithm to NOT +// have to divide in the setup? sigh. +#ifndef STB_VORBIS_NO_DEFER_FLOOR +#define LINE_OP(a,b) a *= b +#else +#define LINE_OP(a,b) a = b +#endif + +#ifdef STB_VORBIS_DIVIDE_TABLE +#define DIVTAB_NUMER 32 +#define DIVTAB_DENOM 64 +int8 integer_divide_table[DIVTAB_NUMER][DIVTAB_DENOM]; // 2KB +#endif + +static __forceinline void draw_line(float *output, int x0, int y0, int x1, int y1, int n) +{ + int dy = y1 - y0; + int adx = x1 - x0; + int ady = abs(dy); + int base; + int x=x0,y=y0; + int err = 0; + int sy; + +#ifdef STB_VORBIS_DIVIDE_TABLE + if (adx < DIVTAB_DENOM && ady < DIVTAB_NUMER) { + if (dy < 0) { + base = -integer_divide_table[ady][adx]; + sy = base-1; + } else { + base = integer_divide_table[ady][adx]; + sy = base+1; + } + } else { + base = dy / adx; + if (dy < 0) + sy = base - 1; + else + sy = base+1; + } +#else + base = dy / adx; + if (dy < 0) + sy = base - 1; + else + sy = base+1; +#endif + ady -= abs(base) * adx; + if (x1 > n) x1 = n; + if (x < x1) { + LINE_OP(output[x], inverse_db_table[y&255]); + for (++x; x < x1; ++x) { + err += ady; + if (err >= adx) { + err -= adx; + y += sy; + } else + y += base; + LINE_OP(output[x], inverse_db_table[y&255]); + } + } +} + +static int residue_decode(vorb *f, Codebook *book, float *target, int offset, int n, int rtype) +{ + int k; + if (rtype == 0) { + int step = n / book->dimensions; + for (k=0; k < step; ++k) + if (!codebook_decode_step(f, book, target+offset+k, n-offset-k, step)) + return FALSE; + } else { + for (k=0; k < n; ) { + if (!codebook_decode(f, book, target+offset, n-k)) + return FALSE; + k += book->dimensions; + offset += book->dimensions; + } + } + return TRUE; +} + +// n is 1/2 of the blocksize -- +// specification: "Correct per-vector decode length is [n]/2" +static void decode_residue(vorb *f, float *residue_buffers[], int ch, int n, int rn, uint8 *do_not_decode) +{ + int i,j,pass; + Residue *r = f->residue_config + rn; + int rtype = f->residue_types[rn]; + int c = r->classbook; + int classwords = f->codebooks[c].dimensions; + unsigned int actual_size = rtype == 2 ? n*2 : n; + unsigned int limit_r_begin = (r->begin < actual_size ? r->begin : actual_size); + unsigned int limit_r_end = (r->end < actual_size ? r->end : actual_size); + int n_read = limit_r_end - limit_r_begin; + int part_read = n_read / r->part_size; + int temp_alloc_point = temp_alloc_save(f); + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + uint8 ***part_classdata = (uint8 ***) temp_block_array(f,f->channels, part_read * sizeof(**part_classdata)); + #else + int **classifications = (int **) temp_block_array(f,f->channels, part_read * sizeof(**classifications)); + #endif + + CHECK(f); + + for (i=0; i < ch; ++i) + if (!do_not_decode[i]) + memset(residue_buffers[i], 0, sizeof(float) * n); + + if (rtype == 2 && ch != 1) { + for (j=0; j < ch; ++j) + if (!do_not_decode[j]) + break; + if (j == ch) + goto done; + + for (pass=0; pass < 8; ++pass) { + int pcount = 0, class_set = 0; + if (ch == 2) { + while (pcount < part_read) { + int z = r->begin + pcount*r->part_size; + int c_inter = (z & 1), p_inter = z>>1; + if (pass == 0) { + Codebook *c = f->codebooks+r->classbook; + int q; + DECODE(q,f,c); + if (q == EOP) goto done; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[0][class_set] = r->classdata[q]; + #else + for (i=classwords-1; i >= 0; --i) { + classifications[0][i+pcount] = q % r->classifications; + q /= r->classifications; + } + #endif + } + for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { + int z = r->begin + pcount*r->part_size; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[0][class_set][i]; + #else + int c = classifications[0][pcount]; + #endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + Codebook *book = f->codebooks + b; + #ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; + #else + // saves 1% + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; + #endif + } else { + z += r->part_size; + c_inter = z & 1; + p_inter = z >> 1; + } + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; + #endif + } + } else if (ch > 2) { + while (pcount < part_read) { + int z = r->begin + pcount*r->part_size; + int c_inter = z % ch, p_inter = z/ch; + if (pass == 0) { + Codebook *c = f->codebooks+r->classbook; + int q; + DECODE(q,f,c); + if (q == EOP) goto done; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[0][class_set] = r->classdata[q]; + #else + for (i=classwords-1; i >= 0; --i) { + classifications[0][i+pcount] = q % r->classifications; + q /= r->classifications; + } + #endif + } + for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { + int z = r->begin + pcount*r->part_size; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[0][class_set][i]; + #else + int c = classifications[0][pcount]; + #endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + Codebook *book = f->codebooks + b; + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; + } else { + z += r->part_size; + c_inter = z % ch; + p_inter = z / ch; + } + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; + #endif + } + } + } + goto done; + } + CHECK(f); + + for (pass=0; pass < 8; ++pass) { + int pcount = 0, class_set=0; + while (pcount < part_read) { + if (pass == 0) { + for (j=0; j < ch; ++j) { + if (!do_not_decode[j]) { + Codebook *c = f->codebooks+r->classbook; + int temp; + DECODE(temp,f,c); + if (temp == EOP) goto done; + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[j][class_set] = r->classdata[temp]; + #else + for (i=classwords-1; i >= 0; --i) { + classifications[j][i+pcount] = temp % r->classifications; + temp /= r->classifications; + } + #endif + } + } + } + for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { + for (j=0; j < ch; ++j) { + if (!do_not_decode[j]) { + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[j][class_set][i]; + #else + int c = classifications[j][pcount]; + #endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + float *target = residue_buffers[j]; + int offset = r->begin + pcount * r->part_size; + int n = r->part_size; + Codebook *book = f->codebooks + b; + if (!residue_decode(f, book, target, offset, n, rtype)) + goto done; + } + } + } + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; + #endif + } + } + done: + CHECK(f); + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + temp_free(f,part_classdata); + #else + temp_free(f,classifications); + #endif + temp_alloc_restore(f,temp_alloc_point); +} + + +#if 0 +// slow way for debugging +void inverse_mdct_slow(float *buffer, int n) +{ + int i,j; + int n2 = n >> 1; + float *x = (float *) malloc(sizeof(*x) * n2); + memcpy(x, buffer, sizeof(*x) * n2); + for (i=0; i < n; ++i) { + float acc = 0; + for (j=0; j < n2; ++j) + // formula from paper: + //acc += n/4.0f * x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1)); + // formula from wikipedia + //acc += 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5)); + // these are equivalent, except the formula from the paper inverts the multiplier! + // however, what actually works is NO MULTIPLIER!?! + //acc += 64 * 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5)); + acc += x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1)); + buffer[i] = acc; + } + free(x); +} +#elif 0 +// same as above, but just barely able to run in real time on modern machines +void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype) +{ + float mcos[16384]; + int i,j; + int n2 = n >> 1, nmask = (n << 2) -1; + float *x = (float *) malloc(sizeof(*x) * n2); + memcpy(x, buffer, sizeof(*x) * n2); + for (i=0; i < 4*n; ++i) + mcos[i] = (float) cos(M_PI / 2 * i / n); + + for (i=0; i < n; ++i) { + float acc = 0; + for (j=0; j < n2; ++j) + acc += x[j] * mcos[(2 * i + 1 + n2)*(2*j+1) & nmask]; + buffer[i] = acc; + } + free(x); +} +#elif 0 +// transform to use a slow dct-iv; this is STILL basically trivial, +// but only requires half as many ops +void dct_iv_slow(float *buffer, int n) +{ + float mcos[16384]; + float x[2048]; + int i,j; + int n2 = n >> 1, nmask = (n << 3) - 1; + memcpy(x, buffer, sizeof(*x) * n); + for (i=0; i < 8*n; ++i) + mcos[i] = (float) cos(M_PI / 4 * i / n); + for (i=0; i < n; ++i) { + float acc = 0; + for (j=0; j < n; ++j) + acc += x[j] * mcos[((2 * i + 1)*(2*j+1)) & nmask]; + buffer[i] = acc; + } +} + +void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype) +{ + int i, n4 = n >> 2, n2 = n >> 1, n3_4 = n - n4; + float temp[4096]; + + memcpy(temp, buffer, n2 * sizeof(float)); + dct_iv_slow(temp, n2); // returns -c'-d, a-b' + + for (i=0; i < n4 ; ++i) buffer[i] = temp[i+n4]; // a-b' + for ( ; i < n3_4; ++i) buffer[i] = -temp[n3_4 - i - 1]; // b-a', c+d' + for ( ; i < n ; ++i) buffer[i] = -temp[i - n3_4]; // c'+d +} +#endif + +#ifndef LIBVORBIS_MDCT +#define LIBVORBIS_MDCT 0 +#endif + +#if LIBVORBIS_MDCT +// directly call the vorbis MDCT using an interface documented +// by Jeff Roberts... useful for performance comparison +typedef struct +{ + int n; + int log2n; + + float *trig; + int *bitrev; + + float scale; +} mdct_lookup; + +extern void mdct_init(mdct_lookup *lookup, int n); +extern void mdct_clear(mdct_lookup *l); +extern void mdct_backward(mdct_lookup *init, float *in, float *out); + +mdct_lookup M1,M2; + +void inverse_mdct(float *buffer, int n, vorb *f, int blocktype) +{ + mdct_lookup *M; + if (M1.n == n) M = &M1; + else if (M2.n == n) M = &M2; + else if (M1.n == 0) { mdct_init(&M1, n); M = &M1; } + else { + if (M2.n) __asm int 3; + mdct_init(&M2, n); + M = &M2; + } + + mdct_backward(M, buffer, buffer); +} +#endif + + +// the following were split out into separate functions while optimizing; +// they could be pushed back up but eh. __forceinline showed no change; +// they're probably already being inlined. +static void imdct_step3_iter0_loop(int n, float *e, int i_off, int k_off, float *A) +{ + float *ee0 = e + i_off; + float *ee2 = ee0 + k_off; + int i; + + assert((n & 3) == 0); + for (i=(n>>2); i > 0; --i) { + float k00_20, k01_21; + k00_20 = ee0[ 0] - ee2[ 0]; + k01_21 = ee0[-1] - ee2[-1]; + ee0[ 0] += ee2[ 0];//ee0[ 0] = ee0[ 0] + ee2[ 0]; + ee0[-1] += ee2[-1];//ee0[-1] = ee0[-1] + ee2[-1]; + ee2[ 0] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-1] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-2] - ee2[-2]; + k01_21 = ee0[-3] - ee2[-3]; + ee0[-2] += ee2[-2];//ee0[-2] = ee0[-2] + ee2[-2]; + ee0[-3] += ee2[-3];//ee0[-3] = ee0[-3] + ee2[-3]; + ee2[-2] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-3] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-4] - ee2[-4]; + k01_21 = ee0[-5] - ee2[-5]; + ee0[-4] += ee2[-4];//ee0[-4] = ee0[-4] + ee2[-4]; + ee0[-5] += ee2[-5];//ee0[-5] = ee0[-5] + ee2[-5]; + ee2[-4] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-5] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-6] - ee2[-6]; + k01_21 = ee0[-7] - ee2[-7]; + ee0[-6] += ee2[-6];//ee0[-6] = ee0[-6] + ee2[-6]; + ee0[-7] += ee2[-7];//ee0[-7] = ee0[-7] + ee2[-7]; + ee2[-6] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-7] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + ee0 -= 8; + ee2 -= 8; + } +} + +static void imdct_step3_inner_r_loop(int lim, float *e, int d0, int k_off, float *A, int k1) +{ + int i; + float k00_20, k01_21; + + float *e0 = e + d0; + float *e2 = e0 + k_off; + + for (i=lim >> 2; i > 0; --i) { + k00_20 = e0[-0] - e2[-0]; + k01_21 = e0[-1] - e2[-1]; + e0[-0] += e2[-0];//e0[-0] = e0[-0] + e2[-0]; + e0[-1] += e2[-1];//e0[-1] = e0[-1] + e2[-1]; + e2[-0] = (k00_20)*A[0] - (k01_21) * A[1]; + e2[-1] = (k01_21)*A[0] + (k00_20) * A[1]; + + A += k1; + + k00_20 = e0[-2] - e2[-2]; + k01_21 = e0[-3] - e2[-3]; + e0[-2] += e2[-2];//e0[-2] = e0[-2] + e2[-2]; + e0[-3] += e2[-3];//e0[-3] = e0[-3] + e2[-3]; + e2[-2] = (k00_20)*A[0] - (k01_21) * A[1]; + e2[-3] = (k01_21)*A[0] + (k00_20) * A[1]; + + A += k1; + + k00_20 = e0[-4] - e2[-4]; + k01_21 = e0[-5] - e2[-5]; + e0[-4] += e2[-4];//e0[-4] = e0[-4] + e2[-4]; + e0[-5] += e2[-5];//e0[-5] = e0[-5] + e2[-5]; + e2[-4] = (k00_20)*A[0] - (k01_21) * A[1]; + e2[-5] = (k01_21)*A[0] + (k00_20) * A[1]; + + A += k1; + + k00_20 = e0[-6] - e2[-6]; + k01_21 = e0[-7] - e2[-7]; + e0[-6] += e2[-6];//e0[-6] = e0[-6] + e2[-6]; + e0[-7] += e2[-7];//e0[-7] = e0[-7] + e2[-7]; + e2[-6] = (k00_20)*A[0] - (k01_21) * A[1]; + e2[-7] = (k01_21)*A[0] + (k00_20) * A[1]; + + e0 -= 8; + e2 -= 8; + + A += k1; + } +} + +static void imdct_step3_inner_s_loop(int n, float *e, int i_off, int k_off, float *A, int a_off, int k0) +{ + int i; + float A0 = A[0]; + float A1 = A[0+1]; + float A2 = A[0+a_off]; + float A3 = A[0+a_off+1]; + float A4 = A[0+a_off*2+0]; + float A5 = A[0+a_off*2+1]; + float A6 = A[0+a_off*3+0]; + float A7 = A[0+a_off*3+1]; + + float k00,k11; + + float *ee0 = e +i_off; + float *ee2 = ee0+k_off; + + for (i=n; i > 0; --i) { + k00 = ee0[ 0] - ee2[ 0]; + k11 = ee0[-1] - ee2[-1]; + ee0[ 0] = ee0[ 0] + ee2[ 0]; + ee0[-1] = ee0[-1] + ee2[-1]; + ee2[ 0] = (k00) * A0 - (k11) * A1; + ee2[-1] = (k11) * A0 + (k00) * A1; + + k00 = ee0[-2] - ee2[-2]; + k11 = ee0[-3] - ee2[-3]; + ee0[-2] = ee0[-2] + ee2[-2]; + ee0[-3] = ee0[-3] + ee2[-3]; + ee2[-2] = (k00) * A2 - (k11) * A3; + ee2[-3] = (k11) * A2 + (k00) * A3; + + k00 = ee0[-4] - ee2[-4]; + k11 = ee0[-5] - ee2[-5]; + ee0[-4] = ee0[-4] + ee2[-4]; + ee0[-5] = ee0[-5] + ee2[-5]; + ee2[-4] = (k00) * A4 - (k11) * A5; + ee2[-5] = (k11) * A4 + (k00) * A5; + + k00 = ee0[-6] - ee2[-6]; + k11 = ee0[-7] - ee2[-7]; + ee0[-6] = ee0[-6] + ee2[-6]; + ee0[-7] = ee0[-7] + ee2[-7]; + ee2[-6] = (k00) * A6 - (k11) * A7; + ee2[-7] = (k11) * A6 + (k00) * A7; + + ee0 -= k0; + ee2 -= k0; + } +} + +static __forceinline void iter_54(float *z) +{ + float k00,k11,k22,k33; + float y0,y1,y2,y3; + + k00 = z[ 0] - z[-4]; + y0 = z[ 0] + z[-4]; + y2 = z[-2] + z[-6]; + k22 = z[-2] - z[-6]; + + z[-0] = y0 + y2; // z0 + z4 + z2 + z6 + z[-2] = y0 - y2; // z0 + z4 - z2 - z6 + + // done with y0,y2 + + k33 = z[-3] - z[-7]; + + z[-4] = k00 + k33; // z0 - z4 + z3 - z7 + z[-6] = k00 - k33; // z0 - z4 - z3 + z7 + + // done with k33 + + k11 = z[-1] - z[-5]; + y1 = z[-1] + z[-5]; + y3 = z[-3] + z[-7]; + + z[-1] = y1 + y3; // z1 + z5 + z3 + z7 + z[-3] = y1 - y3; // z1 + z5 - z3 - z7 + z[-5] = k11 - k22; // z1 - z5 + z2 - z6 + z[-7] = k11 + k22; // z1 - z5 - z2 + z6 +} + +static void imdct_step3_inner_s_loop_ld654(int n, float *e, int i_off, float *A, int base_n) +{ + int a_off = base_n >> 3; + float A2 = A[0+a_off]; + float *z = e + i_off; + float *base = z - 16 * n; + + while (z > base) { + float k00,k11; + float l00,l11; + + k00 = z[-0] - z[ -8]; + k11 = z[-1] - z[ -9]; + l00 = z[-2] - z[-10]; + l11 = z[-3] - z[-11]; + z[ -0] = z[-0] + z[ -8]; + z[ -1] = z[-1] + z[ -9]; + z[ -2] = z[-2] + z[-10]; + z[ -3] = z[-3] + z[-11]; + z[ -8] = k00; + z[ -9] = k11; + z[-10] = (l00+l11) * A2; + z[-11] = (l11-l00) * A2; + + k00 = z[ -4] - z[-12]; + k11 = z[ -5] - z[-13]; + l00 = z[ -6] - z[-14]; + l11 = z[ -7] - z[-15]; + z[ -4] = z[ -4] + z[-12]; + z[ -5] = z[ -5] + z[-13]; + z[ -6] = z[ -6] + z[-14]; + z[ -7] = z[ -7] + z[-15]; + z[-12] = k11; + z[-13] = -k00; + z[-14] = (l11-l00) * A2; + z[-15] = (l00+l11) * -A2; + + iter_54(z); + iter_54(z-8); + z -= 16; + } +} + +static void inverse_mdct(float *buffer, int n, vorb *f, int blocktype) +{ + int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l; + int ld; + // @OPTIMIZE: reduce register pressure by using fewer variables? + int save_point = temp_alloc_save(f); + float *buf2 = (float *) temp_alloc(f, n2 * sizeof(*buf2)); + float *u=NULL,*v=NULL; + // twiddle factors + float *A = f->A[blocktype]; + + // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio" + // See notes about bugs in that paper in less-optimal implementation 'inverse_mdct_old' after this function. + + // kernel from paper + + + // merged: + // copy and reflect spectral data + // step 0 + + // note that it turns out that the items added together during + // this step are, in fact, being added to themselves (as reflected + // by step 0). inexplicable inefficiency! this became obvious + // once I combined the passes. + + // so there's a missing 'times 2' here (for adding X to itself). + // this propagates through linearly to the end, where the numbers + // are 1/2 too small, and need to be compensated for. + + { + float *d,*e, *AA, *e_stop; + d = &buf2[n2-2]; + AA = A; + e = &buffer[0]; + e_stop = &buffer[n2]; + while (e != e_stop) { + d[1] = (e[0] * AA[0] - e[2]*AA[1]); + d[0] = (e[0] * AA[1] + e[2]*AA[0]); + d -= 2; + AA += 2; + e += 4; + } + + e = &buffer[n2-3]; + while (d >= buf2) { + d[1] = (-e[2] * AA[0] - -e[0]*AA[1]); + d[0] = (-e[2] * AA[1] + -e[0]*AA[0]); + d -= 2; + AA += 2; + e -= 4; + } + } + + // now we use symbolic names for these, so that we can + // possibly swap their meaning as we change which operations + // are in place + + u = buffer; + v = buf2; + + // step 2 (paper output is w, now u) + // this could be in place, but the data ends up in the wrong + // place... _somebody_'s got to swap it, so this is nominated + { + float *AA = &A[n2-8]; + float *d0,*d1, *e0, *e1; + + e0 = &v[n4]; + e1 = &v[0]; + + d0 = &u[n4]; + d1 = &u[0]; + + while (AA >= A) { + float v40_20, v41_21; + + v41_21 = e0[1] - e1[1]; + v40_20 = e0[0] - e1[0]; + d0[1] = e0[1] + e1[1]; + d0[0] = e0[0] + e1[0]; + d1[1] = v41_21*AA[4] - v40_20*AA[5]; + d1[0] = v40_20*AA[4] + v41_21*AA[5]; + + v41_21 = e0[3] - e1[3]; + v40_20 = e0[2] - e1[2]; + d0[3] = e0[3] + e1[3]; + d0[2] = e0[2] + e1[2]; + d1[3] = v41_21*AA[0] - v40_20*AA[1]; + d1[2] = v40_20*AA[0] + v41_21*AA[1]; + + AA -= 8; + + d0 += 4; + d1 += 4; + e0 += 4; + e1 += 4; + } + } + + // step 3 + ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + + // optimized step 3: + + // the original step3 loop can be nested r inside s or s inside r; + // it's written originally as s inside r, but this is dumb when r + // iterates many times, and s few. So I have two copies of it and + // switch between them halfway. + + // this is iteration 0 of step 3 + imdct_step3_iter0_loop(n >> 4, u, n2-1-n4*0, -(n >> 3), A); + imdct_step3_iter0_loop(n >> 4, u, n2-1-n4*1, -(n >> 3), A); + + // this is iteration 1 of step 3 + imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*0, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*1, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*2, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*3, -(n >> 4), A, 16); + + l=2; + for (; l < (ld-3)>>1; ++l) { + int k0 = n >> (l+2), k0_2 = k0>>1; + int lim = 1 << (l+1); + int i; + for (i=0; i < lim; ++i) + imdct_step3_inner_r_loop(n >> (l+4), u, n2-1 - k0*i, -k0_2, A, 1 << (l+3)); + } + + for (; l < ld-6; ++l) { + int k0 = n >> (l+2), k1 = 1 << (l+3), k0_2 = k0>>1; + int rlim = n >> (l+6), r; + int lim = 1 << (l+1); + int i_off; + float *A0 = A; + i_off = n2-1; + for (r=rlim; r > 0; --r) { + imdct_step3_inner_s_loop(lim, u, i_off, -k0_2, A0, k1, k0); + A0 += k1*4; + i_off -= 8; + } + } + + // iterations with count: + // ld-6,-5,-4 all interleaved together + // the big win comes from getting rid of needless flops + // due to the constants on pass 5 & 4 being all 1 and 0; + // combining them to be simultaneous to improve cache made little difference + imdct_step3_inner_s_loop_ld654(n >> 5, u, n2-1, A, n); + + // output is u + + // step 4, 5, and 6 + // cannot be in-place because of step 5 + { + uint16 *bitrev = f->bit_reverse[blocktype]; + // weirdly, I'd have thought reading sequentially and writing + // erratically would have been better than vice-versa, but in + // fact that's not what my testing showed. (That is, with + // j = bitreverse(i), do you read i and write j, or read j and write i.) + + float *d0 = &v[n4-4]; + float *d1 = &v[n2-4]; + while (d0 >= v) { + int k4; + + k4 = bitrev[0]; + d1[3] = u[k4+0]; + d1[2] = u[k4+1]; + d0[3] = u[k4+2]; + d0[2] = u[k4+3]; + + k4 = bitrev[1]; + d1[1] = u[k4+0]; + d1[0] = u[k4+1]; + d0[1] = u[k4+2]; + d0[0] = u[k4+3]; + + d0 -= 4; + d1 -= 4; + bitrev += 2; + } + } + // (paper output is u, now v) + + + // data must be in buf2 + assert(v == buf2); + + // step 7 (paper output is v, now v) + // this is now in place + { + float *C = f->C[blocktype]; + float *d, *e; + + d = v; + e = v + n2 - 4; + + while (d < e) { + float a02,a11,b0,b1,b2,b3; + + a02 = d[0] - e[2]; + a11 = d[1] + e[3]; + + b0 = C[1]*a02 + C[0]*a11; + b1 = C[1]*a11 - C[0]*a02; + + b2 = d[0] + e[ 2]; + b3 = d[1] - e[ 3]; + + d[0] = b2 + b0; + d[1] = b3 + b1; + e[2] = b2 - b0; + e[3] = b1 - b3; + + a02 = d[2] - e[0]; + a11 = d[3] + e[1]; + + b0 = C[3]*a02 + C[2]*a11; + b1 = C[3]*a11 - C[2]*a02; + + b2 = d[2] + e[ 0]; + b3 = d[3] - e[ 1]; + + d[2] = b2 + b0; + d[3] = b3 + b1; + e[0] = b2 - b0; + e[1] = b1 - b3; + + C += 4; + d += 4; + e -= 4; + } + } + + // data must be in buf2 + + + // step 8+decode (paper output is X, now buffer) + // this generates pairs of data a la 8 and pushes them directly through + // the decode kernel (pushing rather than pulling) to avoid having + // to make another pass later + + // this cannot POSSIBLY be in place, so we refer to the buffers directly + + { + float *d0,*d1,*d2,*d3; + + float *B = f->B[blocktype] + n2 - 8; + float *e = buf2 + n2 - 8; + d0 = &buffer[0]; + d1 = &buffer[n2-4]; + d2 = &buffer[n2]; + d3 = &buffer[n-4]; + while (e >= v) { + float p0,p1,p2,p3; + + p3 = e[6]*B[7] - e[7]*B[6]; + p2 = -e[6]*B[6] - e[7]*B[7]; + + d0[0] = p3; + d1[3] = - p3; + d2[0] = p2; + d3[3] = p2; + + p1 = e[4]*B[5] - e[5]*B[4]; + p0 = -e[4]*B[4] - e[5]*B[5]; + + d0[1] = p1; + d1[2] = - p1; + d2[1] = p0; + d3[2] = p0; + + p3 = e[2]*B[3] - e[3]*B[2]; + p2 = -e[2]*B[2] - e[3]*B[3]; + + d0[2] = p3; + d1[1] = - p3; + d2[2] = p2; + d3[1] = p2; + + p1 = e[0]*B[1] - e[1]*B[0]; + p0 = -e[0]*B[0] - e[1]*B[1]; + + d0[3] = p1; + d1[0] = - p1; + d2[3] = p0; + d3[0] = p0; + + B -= 8; + e -= 8; + d0 += 4; + d2 += 4; + d1 -= 4; + d3 -= 4; + } + } + + temp_free(f,buf2); + temp_alloc_restore(f,save_point); +} + +#if 0 +// this is the original version of the above code, if you want to optimize it from scratch +void inverse_mdct_naive(float *buffer, int n) +{ + float s; + float A[1 << 12], B[1 << 12], C[1 << 11]; + int i,k,k2,k4, n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l; + int n3_4 = n - n4, ld; + // how can they claim this only uses N words?! + // oh, because they're only used sparsely, whoops + float u[1 << 13], X[1 << 13], v[1 << 13], w[1 << 13]; + // set up twiddle factors + + for (k=k2=0; k < n4; ++k,k2+=2) { + A[k2 ] = (float) cos(4*k*M_PI/n); + A[k2+1] = (float) -sin(4*k*M_PI/n); + B[k2 ] = (float) cos((k2+1)*M_PI/n/2); + B[k2+1] = (float) sin((k2+1)*M_PI/n/2); + } + for (k=k2=0; k < n8; ++k,k2+=2) { + C[k2 ] = (float) cos(2*(k2+1)*M_PI/n); + C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n); + } + + // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio" + // Note there are bugs in that pseudocode, presumably due to them attempting + // to rename the arrays nicely rather than representing the way their actual + // implementation bounces buffers back and forth. As a result, even in the + // "some formulars corrected" version, a direct implementation fails. These + // are noted below as "paper bug". + + // copy and reflect spectral data + for (k=0; k < n2; ++k) u[k] = buffer[k]; + for ( ; k < n ; ++k) u[k] = -buffer[n - k - 1]; + // kernel from paper + // step 1 + for (k=k2=k4=0; k < n4; k+=1, k2+=2, k4+=4) { + v[n-k4-1] = (u[k4] - u[n-k4-1]) * A[k2] - (u[k4+2] - u[n-k4-3])*A[k2+1]; + v[n-k4-3] = (u[k4] - u[n-k4-1]) * A[k2+1] + (u[k4+2] - u[n-k4-3])*A[k2]; + } + // step 2 + for (k=k4=0; k < n8; k+=1, k4+=4) { + w[n2+3+k4] = v[n2+3+k4] + v[k4+3]; + w[n2+1+k4] = v[n2+1+k4] + v[k4+1]; + w[k4+3] = (v[n2+3+k4] - v[k4+3])*A[n2-4-k4] - (v[n2+1+k4]-v[k4+1])*A[n2-3-k4]; + w[k4+1] = (v[n2+1+k4] - v[k4+1])*A[n2-4-k4] + (v[n2+3+k4]-v[k4+3])*A[n2-3-k4]; + } + // step 3 + ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + for (l=0; l < ld-3; ++l) { + int k0 = n >> (l+2), k1 = 1 << (l+3); + int rlim = n >> (l+4), r4, r; + int s2lim = 1 << (l+2), s2; + for (r=r4=0; r < rlim; r4+=4,++r) { + for (s2=0; s2 < s2lim; s2+=2) { + u[n-1-k0*s2-r4] = w[n-1-k0*s2-r4] + w[n-1-k0*(s2+1)-r4]; + u[n-3-k0*s2-r4] = w[n-3-k0*s2-r4] + w[n-3-k0*(s2+1)-r4]; + u[n-1-k0*(s2+1)-r4] = (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1] + - (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1+1]; + u[n-3-k0*(s2+1)-r4] = (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1] + + (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1+1]; + } + } + if (l+1 < ld-3) { + // paper bug: ping-ponging of u&w here is omitted + memcpy(w, u, sizeof(u)); + } + } + + // step 4 + for (i=0; i < n8; ++i) { + int j = bit_reverse(i) >> (32-ld+3); + assert(j < n8); + if (i == j) { + // paper bug: original code probably swapped in place; if copying, + // need to directly copy in this case + int i8 = i << 3; + v[i8+1] = u[i8+1]; + v[i8+3] = u[i8+3]; + v[i8+5] = u[i8+5]; + v[i8+7] = u[i8+7]; + } else if (i < j) { + int i8 = i << 3, j8 = j << 3; + v[j8+1] = u[i8+1], v[i8+1] = u[j8 + 1]; + v[j8+3] = u[i8+3], v[i8+3] = u[j8 + 3]; + v[j8+5] = u[i8+5], v[i8+5] = u[j8 + 5]; + v[j8+7] = u[i8+7], v[i8+7] = u[j8 + 7]; + } + } + // step 5 + for (k=0; k < n2; ++k) { + w[k] = v[k*2+1]; + } + // step 6 + for (k=k2=k4=0; k < n8; ++k, k2 += 2, k4 += 4) { + u[n-1-k2] = w[k4]; + u[n-2-k2] = w[k4+1]; + u[n3_4 - 1 - k2] = w[k4+2]; + u[n3_4 - 2 - k2] = w[k4+3]; + } + // step 7 + for (k=k2=0; k < n8; ++k, k2 += 2) { + v[n2 + k2 ] = ( u[n2 + k2] + u[n-2-k2] + C[k2+1]*(u[n2+k2]-u[n-2-k2]) + C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2; + v[n-2 - k2] = ( u[n2 + k2] + u[n-2-k2] - C[k2+1]*(u[n2+k2]-u[n-2-k2]) - C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2; + v[n2+1+ k2] = ( u[n2+1+k2] - u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2; + v[n-1 - k2] = (-u[n2+1+k2] + u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2; + } + // step 8 + for (k=k2=0; k < n4; ++k,k2 += 2) { + X[k] = v[k2+n2]*B[k2 ] + v[k2+1+n2]*B[k2+1]; + X[n2-1-k] = v[k2+n2]*B[k2+1] - v[k2+1+n2]*B[k2 ]; + } + + // decode kernel to output + // determined the following value experimentally + // (by first figuring out what made inverse_mdct_slow work); then matching that here + // (probably vorbis encoder premultiplies by n or n/2, to save it on the decoder?) + s = 0.5; // theoretically would be n4 + + // [[[ note! the s value of 0.5 is compensated for by the B[] in the current code, + // so it needs to use the "old" B values to behave correctly, or else + // set s to 1.0 ]]] + for (i=0; i < n4 ; ++i) buffer[i] = s * X[i+n4]; + for ( ; i < n3_4; ++i) buffer[i] = -s * X[n3_4 - i - 1]; + for ( ; i < n ; ++i) buffer[i] = -s * X[i - n3_4]; +} +#endif + +static float *get_window(vorb *f, int len) +{ + len <<= 1; + if (len == f->blocksize_0) return f->window[0]; + if (len == f->blocksize_1) return f->window[1]; + return NULL; +} + +#ifndef STB_VORBIS_NO_DEFER_FLOOR +typedef int16 YTYPE; +#else +typedef int YTYPE; +#endif +static int do_floor(vorb *f, Mapping *map, int i, int n, float *target, YTYPE *finalY, uint8 *step2_flag) +{ + int n2 = n >> 1; + int s = map->chan[i].mux, floor; + floor = map->submap_floor[s]; + if (f->floor_types[floor] == 0) { + return error(f, VORBIS_invalid_stream); + } else { + Floor1 *g = &f->floor_config[floor].floor1; + int j,q; + int lx = 0, ly = finalY[0] * g->floor1_multiplier; + for (q=1; q < g->values; ++q) { + j = g->sorted_order[q]; + #ifndef STB_VORBIS_NO_DEFER_FLOOR + STBV_NOTUSED(step2_flag); + if (finalY[j] >= 0) + #else + if (step2_flag[j]) + #endif + { + int hy = finalY[j] * g->floor1_multiplier; + int hx = g->Xlist[j]; + if (lx != hx) + draw_line(target, lx,ly, hx,hy, n2); + CHECK(f); + lx = hx, ly = hy; + } + } + if (lx < n2) { + // optimization of: draw_line(target, lx,ly, n,ly, n2); + for (j=lx; j < n2; ++j) + LINE_OP(target[j], inverse_db_table[ly]); + CHECK(f); + } + } + return TRUE; +} + +// The meaning of "left" and "right" +// +// For a given frame: +// we compute samples from 0..n +// window_center is n/2 +// we'll window and mix the samples from left_start to left_end with data from the previous frame +// all of the samples from left_end to right_start can be output without mixing; however, +// this interval is 0-length except when transitioning between short and long frames +// all of the samples from right_start to right_end need to be mixed with the next frame, +// which we don't have, so those get saved in a buffer +// frame N's right_end-right_start, the number of samples to mix with the next frame, +// has to be the same as frame N+1's left_end-left_start (which they are by +// construction) + +static int vorbis_decode_initial(vorb *f, int *p_left_start, int *p_left_end, int *p_right_start, int *p_right_end, int *mode) +{ + Mode *m; + int i, n, prev, next, window_center; + f->channel_buffer_start = f->channel_buffer_end = 0; + + retry: + if (f->eof) return FALSE; + if (!maybe_start_packet(f)) + return FALSE; + // check packet type + if (get_bits(f,1) != 0) { + if (IS_PUSH_MODE(f)) + return error(f,VORBIS_bad_packet_type); + while (EOP != get8_packet(f)); + goto retry; + } + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + + i = get_bits(f, ilog(f->mode_count-1)); + if (i == EOP) return FALSE; + if (i >= f->mode_count) return FALSE; + *mode = i; + m = f->mode_config + i; + if (m->blockflag) { + n = f->blocksize_1; + prev = get_bits(f,1); + next = get_bits(f,1); + } else { + prev = next = 0; + n = f->blocksize_0; + } + +// WINDOWING + + window_center = n >> 1; + if (m->blockflag && !prev) { + *p_left_start = (n - f->blocksize_0) >> 2; + *p_left_end = (n + f->blocksize_0) >> 2; + } else { + *p_left_start = 0; + *p_left_end = window_center; + } + if (m->blockflag && !next) { + *p_right_start = (n*3 - f->blocksize_0) >> 2; + *p_right_end = (n*3 + f->blocksize_0) >> 2; + } else { + *p_right_start = window_center; + *p_right_end = n; + } + + return TRUE; +} + +static int vorbis_decode_packet_rest(vorb *f, int *len, Mode *m, int left_start, int left_end, int right_start, int right_end, int *p_left) +{ + Mapping *map; + int i,j,k,n,n2; + int zero_channel[256]; + int really_zero_channel[256]; + +// WINDOWING + + STBV_NOTUSED(left_end); + n = f->blocksize[m->blockflag]; + map = &f->mapping[m->mapping]; + +// FLOORS + n2 = n >> 1; + + CHECK(f); + + for (i=0; i < f->channels; ++i) { + int s = map->chan[i].mux, floor; + zero_channel[i] = FALSE; + floor = map->submap_floor[s]; + if (f->floor_types[floor] == 0) { + return error(f, VORBIS_invalid_stream); + } else { + Floor1 *g = &f->floor_config[floor].floor1; + if (get_bits(f, 1)) { + short *finalY; + uint8 step2_flag[256]; + static int range_list[4] = { 256, 128, 86, 64 }; + int range = range_list[g->floor1_multiplier-1]; + int offset = 2; + finalY = f->finalY[i]; + finalY[0] = get_bits(f, ilog(range)-1); + finalY[1] = get_bits(f, ilog(range)-1); + for (j=0; j < g->partitions; ++j) { + int pclass = g->partition_class_list[j]; + int cdim = g->class_dimensions[pclass]; + int cbits = g->class_subclasses[pclass]; + int csub = (1 << cbits)-1; + int cval = 0; + if (cbits) { + Codebook *c = f->codebooks + g->class_masterbooks[pclass]; + DECODE(cval,f,c); + } + for (k=0; k < cdim; ++k) { + int book = g->subclass_books[pclass][cval & csub]; + cval = cval >> cbits; + if (book >= 0) { + int temp; + Codebook *c = f->codebooks + book; + DECODE(temp,f,c); + finalY[offset++] = temp; + } else + finalY[offset++] = 0; + } + } + if (f->valid_bits == INVALID_BITS) goto error; // behavior according to spec + step2_flag[0] = step2_flag[1] = 1; + for (j=2; j < g->values; ++j) { + int low, high, pred, highroom, lowroom, room, val; + low = g->neighbors[j][0]; + high = g->neighbors[j][1]; + //neighbors(g->Xlist, j, &low, &high); + pred = predict_point(g->Xlist[j], g->Xlist[low], g->Xlist[high], finalY[low], finalY[high]); + val = finalY[j]; + highroom = range - pred; + lowroom = pred; + if (highroom < lowroom) + room = highroom * 2; + else + room = lowroom * 2; + if (val) { + step2_flag[low] = step2_flag[high] = 1; + step2_flag[j] = 1; + if (val >= room) + if (highroom > lowroom) + finalY[j] = val - lowroom + pred; + else + finalY[j] = pred - val + highroom - 1; + else + if (val & 1) + finalY[j] = pred - ((val+1)>>1); + else + finalY[j] = pred + (val>>1); + } else { + step2_flag[j] = 0; + finalY[j] = pred; + } + } + +#ifdef STB_VORBIS_NO_DEFER_FLOOR + do_floor(f, map, i, n, f->floor_buffers[i], finalY, step2_flag); +#else + // defer final floor computation until _after_ residue + for (j=0; j < g->values; ++j) { + if (!step2_flag[j]) + finalY[j] = -1; + } +#endif + } else { + error: + zero_channel[i] = TRUE; + } + // So we just defer everything else to later + + // at this point we've decoded the floor into buffer + } + } + CHECK(f); + // at this point we've decoded all floors + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + + // re-enable coupled channels if necessary + memcpy(really_zero_channel, zero_channel, sizeof(really_zero_channel[0]) * f->channels); + for (i=0; i < map->coupling_steps; ++i) + if (!zero_channel[map->chan[i].magnitude] || !zero_channel[map->chan[i].angle]) { + zero_channel[map->chan[i].magnitude] = zero_channel[map->chan[i].angle] = FALSE; + } + + CHECK(f); +// RESIDUE DECODE + for (i=0; i < map->submaps; ++i) { + float *residue_buffers[STB_VORBIS_MAX_CHANNELS]; + int r; + uint8 do_not_decode[256]; + int ch = 0; + for (j=0; j < f->channels; ++j) { + if (map->chan[j].mux == i) { + if (zero_channel[j]) { + do_not_decode[ch] = TRUE; + residue_buffers[ch] = NULL; + } else { + do_not_decode[ch] = FALSE; + residue_buffers[ch] = f->channel_buffers[j]; + } + ++ch; + } + } + r = map->submap_residue[i]; + decode_residue(f, residue_buffers, ch, n2, r, do_not_decode); + } + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + CHECK(f); + +// INVERSE COUPLING + for (i = map->coupling_steps-1; i >= 0; --i) { + int n2 = n >> 1; + float *m = f->channel_buffers[map->chan[i].magnitude]; + float *a = f->channel_buffers[map->chan[i].angle ]; + for (j=0; j < n2; ++j) { + float a2,m2; + if (m[j] > 0) + if (a[j] > 0) + m2 = m[j], a2 = m[j] - a[j]; + else + a2 = m[j], m2 = m[j] + a[j]; + else + if (a[j] > 0) + m2 = m[j], a2 = m[j] + a[j]; + else + a2 = m[j], m2 = m[j] - a[j]; + m[j] = m2; + a[j] = a2; + } + } + CHECK(f); + + // finish decoding the floors +#ifndef STB_VORBIS_NO_DEFER_FLOOR + for (i=0; i < f->channels; ++i) { + if (really_zero_channel[i]) { + memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2); + } else { + do_floor(f, map, i, n, f->channel_buffers[i], f->finalY[i], NULL); + } + } +#else + for (i=0; i < f->channels; ++i) { + if (really_zero_channel[i]) { + memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2); + } else { + for (j=0; j < n2; ++j) + f->channel_buffers[i][j] *= f->floor_buffers[i][j]; + } + } +#endif + +// INVERSE MDCT + CHECK(f); + for (i=0; i < f->channels; ++i) + inverse_mdct(f->channel_buffers[i], n, f, m->blockflag); + CHECK(f); + + // this shouldn't be necessary, unless we exited on an error + // and want to flush to get to the next packet + flush_packet(f); + + if (f->first_decode) { + // assume we start so first non-discarded sample is sample 0 + // this isn't to spec, but spec would require us to read ahead + // and decode the size of all current frames--could be done, + // but presumably it's not a commonly used feature + f->current_loc = 0u - n2; // start of first frame is positioned for discard (NB this is an intentional unsigned overflow/wrap-around) + // we might have to discard samples "from" the next frame too, + // if we're lapping a large block then a small at the start? + f->discard_samples_deferred = n - right_end; + f->current_loc_valid = TRUE; + f->first_decode = FALSE; + } else if (f->discard_samples_deferred) { + if (f->discard_samples_deferred >= right_start - left_start) { + f->discard_samples_deferred -= (right_start - left_start); + left_start = right_start; + *p_left = left_start; + } else { + left_start += f->discard_samples_deferred; + *p_left = left_start; + f->discard_samples_deferred = 0; + } + } else if (f->previous_length == 0 && f->current_loc_valid) { + // we're recovering from a seek... that means we're going to discard + // the samples from this packet even though we know our position from + // the last page header, so we need to update the position based on + // the discarded samples here + // but wait, the code below is going to add this in itself even + // on a discard, so we don't need to do it here... + } + + // check if we have ogg information about the sample # for this packet + if (f->last_seg_which == f->end_seg_with_known_loc) { + // if we have a valid current loc, and this is final: + if (f->current_loc_valid && (f->page_flag & PAGEFLAG_last_page)) { + uint32 current_end = f->known_loc_for_packet; + // then let's infer the size of the (probably) short final frame + if (current_end < f->current_loc + (right_end-left_start)) { + if (current_end < f->current_loc) { + // negative truncation, that's impossible! + *len = 0; + } else { + *len = current_end - f->current_loc; + } + *len += left_start; // this doesn't seem right, but has no ill effect on my test files + if (*len > right_end) *len = right_end; // this should never happen + f->current_loc += *len; + return TRUE; + } + } + // otherwise, just set our sample loc + // guess that the ogg granule pos refers to the _middle_ of the + // last frame? + // set f->current_loc to the position of left_start + f->current_loc = f->known_loc_for_packet - (n2-left_start); + f->current_loc_valid = TRUE; + } + if (f->current_loc_valid) + f->current_loc += (right_start - left_start); + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + *len = right_end; // ignore samples after the window goes to 0 + CHECK(f); + + return TRUE; +} + +static int vorbis_decode_packet(vorb *f, int *len, int *p_left, int *p_right) +{ + int mode, left_end, right_end; + if (!vorbis_decode_initial(f, p_left, &left_end, p_right, &right_end, &mode)) return 0; + return vorbis_decode_packet_rest(f, len, f->mode_config + mode, *p_left, left_end, *p_right, right_end, p_left); +} + +static int vorbis_finish_frame(stb_vorbis *f, int len, int left, int right) +{ + int prev,i,j; + // we use right&left (the start of the right- and left-window sin()-regions) + // to determine how much to return, rather than inferring from the rules + // (same result, clearer code); 'left' indicates where our sin() window + // starts, therefore where the previous window's right edge starts, and + // therefore where to start mixing from the previous buffer. 'right' + // indicates where our sin() ending-window starts, therefore that's where + // we start saving, and where our returned-data ends. + + // mixin from previous window + if (f->previous_length) { + int i,j, n = f->previous_length; + float *w = get_window(f, n); + if (w == NULL) return 0; + for (i=0; i < f->channels; ++i) { + for (j=0; j < n; ++j) + f->channel_buffers[i][left+j] = + f->channel_buffers[i][left+j]*w[ j] + + f->previous_window[i][ j]*w[n-1-j]; + } + } + + prev = f->previous_length; + + // last half of this data becomes previous window + f->previous_length = len - right; + + // @OPTIMIZE: could avoid this copy by double-buffering the + // output (flipping previous_window with channel_buffers), but + // then previous_window would have to be 2x as large, and + // channel_buffers couldn't be temp mem (although they're NOT + // currently temp mem, they could be (unless we want to level + // performance by spreading out the computation)) + for (i=0; i < f->channels; ++i) + for (j=0; right+j < len; ++j) + f->previous_window[i][j] = f->channel_buffers[i][right+j]; + + if (!prev) + // there was no previous packet, so this data isn't valid... + // this isn't entirely true, only the would-have-overlapped data + // isn't valid, but this seems to be what the spec requires + return 0; + + // truncate a short frame + if (len < right) right = len; + + f->samples_output += right-left; + + return right - left; +} + +static int vorbis_pump_first_frame(stb_vorbis *f) +{ + int len, right, left, res; + res = vorbis_decode_packet(f, &len, &left, &right); + if (res) + vorbis_finish_frame(f, len, left, right); + return res; +} + +#ifndef STB_VORBIS_NO_PUSHDATA_API +static int is_whole_packet_present(stb_vorbis *f) +{ + // make sure that we have the packet available before continuing... + // this requires a full ogg parse, but we know we can fetch from f->stream + + // instead of coding this out explicitly, we could save the current read state, + // read the next packet with get8() until end-of-packet, check f->eof, then + // reset the state? but that would be slower, esp. since we'd have over 256 bytes + // of state to restore (primarily the page segment table) + + int s = f->next_seg, first = TRUE; + uint8 *p = f->stream; + + if (s != -1) { // if we're not starting the packet with a 'continue on next page' flag + for (; s < f->segment_count; ++s) { + p += f->segments[s]; + if (f->segments[s] < 255) // stop at first short segment + break; + } + // either this continues, or it ends it... + if (s == f->segment_count) + s = -1; // set 'crosses page' flag + if (p > f->stream_end) return error(f, VORBIS_need_more_data); + first = FALSE; + } + for (; s == -1;) { + uint8 *q; + int n; + + // check that we have the page header ready + if (p + 26 >= f->stream_end) return error(f, VORBIS_need_more_data); + // validate the page + if (memcmp(p, ogg_page_header, 4)) return error(f, VORBIS_invalid_stream); + if (p[4] != 0) return error(f, VORBIS_invalid_stream); + if (first) { // the first segment must NOT have 'continued_packet', later ones MUST + if (f->previous_length) + if ((p[5] & PAGEFLAG_continued_packet)) return error(f, VORBIS_invalid_stream); + // if no previous length, we're resynching, so we can come in on a continued-packet, + // which we'll just drop + } else { + if (!(p[5] & PAGEFLAG_continued_packet)) return error(f, VORBIS_invalid_stream); + } + n = p[26]; // segment counts + q = p+27; // q points to segment table + p = q + n; // advance past header + // make sure we've read the segment table + if (p > f->stream_end) return error(f, VORBIS_need_more_data); + for (s=0; s < n; ++s) { + p += q[s]; + if (q[s] < 255) + break; + } + if (s == n) + s = -1; // set 'crosses page' flag + if (p > f->stream_end) return error(f, VORBIS_need_more_data); + first = FALSE; + } + return TRUE; +} +#endif // !STB_VORBIS_NO_PUSHDATA_API + +static int start_decoder(vorb *f) +{ + uint8 header[6], x,y; + int len,i,j,k, max_submaps = 0; + int longest_floorlist=0; + + // first page, first packet + f->first_decode = TRUE; + + if (!start_page(f)) return FALSE; + // validate page flag + if (!(f->page_flag & PAGEFLAG_first_page)) return error(f, VORBIS_invalid_first_page); + if (f->page_flag & PAGEFLAG_last_page) return error(f, VORBIS_invalid_first_page); + if (f->page_flag & PAGEFLAG_continued_packet) return error(f, VORBIS_invalid_first_page); + // check for expected packet length + if (f->segment_count != 1) return error(f, VORBIS_invalid_first_page); + if (f->segments[0] != 30) { + // check for the Ogg skeleton fishead identifying header to refine our error + if (f->segments[0] == 64 && + getn(f, header, 6) && + header[0] == 'f' && + header[1] == 'i' && + header[2] == 's' && + header[3] == 'h' && + header[4] == 'e' && + header[5] == 'a' && + get8(f) == 'd' && + get8(f) == '\0') return error(f, VORBIS_ogg_skeleton_not_supported); + else + return error(f, VORBIS_invalid_first_page); + } + + // read packet + // check packet header + if (get8(f) != VORBIS_packet_id) return error(f, VORBIS_invalid_first_page); + if (!getn(f, header, 6)) return error(f, VORBIS_unexpected_eof); + if (!vorbis_validate(header)) return error(f, VORBIS_invalid_first_page); + // vorbis_version + if (get32(f) != 0) return error(f, VORBIS_invalid_first_page); + f->channels = get8(f); if (!f->channels) return error(f, VORBIS_invalid_first_page); + if (f->channels > STB_VORBIS_MAX_CHANNELS) return error(f, VORBIS_too_many_channels); + f->sample_rate = get32(f); if (!f->sample_rate) return error(f, VORBIS_invalid_first_page); + get32(f); // bitrate_maximum + get32(f); // bitrate_nominal + get32(f); // bitrate_minimum + x = get8(f); + { + int log0,log1; + log0 = x & 15; + log1 = x >> 4; + f->blocksize_0 = 1 << log0; + f->blocksize_1 = 1 << log1; + if (log0 < 6 || log0 > 13) return error(f, VORBIS_invalid_setup); + if (log1 < 6 || log1 > 13) return error(f, VORBIS_invalid_setup); + if (log0 > log1) return error(f, VORBIS_invalid_setup); + } + + // framing_flag + x = get8(f); + if (!(x & 1)) return error(f, VORBIS_invalid_first_page); + + // second packet! + if (!start_page(f)) return FALSE; + + if (!start_packet(f)) return FALSE; + + if (!next_segment(f)) return FALSE; + + if (get8_packet(f) != VORBIS_packet_comment) return error(f, VORBIS_invalid_setup); + for (i=0; i < 6; ++i) header[i] = get8_packet(f); + if (!vorbis_validate(header)) return error(f, VORBIS_invalid_setup); + //file vendor + len = get32_packet(f); + f->vendor = (char*)setup_malloc(f, sizeof(char) * (len+1)); + if (f->vendor == NULL) return error(f, VORBIS_outofmem); + for(i=0; i < len; ++i) { + f->vendor[i] = get8_packet(f); + } + f->vendor[len] = (char)'\0'; + //user comments + f->comment_list_length = get32_packet(f); + f->comment_list = NULL; + if (f->comment_list_length > 0) + { + f->comment_list = (char**) setup_malloc(f, sizeof(char*) * (f->comment_list_length)); + if (f->comment_list == NULL) return error(f, VORBIS_outofmem); + } + + for(i=0; i < f->comment_list_length; ++i) { + len = get32_packet(f); + f->comment_list[i] = (char*)setup_malloc(f, sizeof(char) * (len+1)); + if (f->comment_list[i] == NULL) return error(f, VORBIS_outofmem); + + for(j=0; j < len; ++j) { + f->comment_list[i][j] = get8_packet(f); + } + f->comment_list[i][len] = (char)'\0'; + } + + // framing_flag + x = get8_packet(f); + if (!(x & 1)) return error(f, VORBIS_invalid_setup); + + + skip(f, f->bytes_in_seg); + f->bytes_in_seg = 0; + + do { + len = next_segment(f); + skip(f, len); + f->bytes_in_seg = 0; + } while (len); + + // third packet! + if (!start_packet(f)) return FALSE; + + #ifndef STB_VORBIS_NO_PUSHDATA_API + if (IS_PUSH_MODE(f)) { + if (!is_whole_packet_present(f)) { + // convert error in ogg header to write type + if (f->error == VORBIS_invalid_stream) + f->error = VORBIS_invalid_setup; + return FALSE; + } + } + #endif + + crc32_init(); // always init it, to avoid multithread race conditions + + if (get8_packet(f) != VORBIS_packet_setup) return error(f, VORBIS_invalid_setup); + for (i=0; i < 6; ++i) header[i] = get8_packet(f); + if (!vorbis_validate(header)) return error(f, VORBIS_invalid_setup); + + // codebooks + + f->codebook_count = get_bits(f,8) + 1; + f->codebooks = (Codebook *) setup_malloc(f, sizeof(*f->codebooks) * f->codebook_count); + if (f->codebooks == NULL) return error(f, VORBIS_outofmem); + memset(f->codebooks, 0, sizeof(*f->codebooks) * f->codebook_count); + for (i=0; i < f->codebook_count; ++i) { + uint32 *values; + int ordered, sorted_count; + int total=0; + uint8 *lengths; + Codebook *c = f->codebooks+i; + CHECK(f); + x = get_bits(f, 8); if (x != 0x42) return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); if (x != 0x43) return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); if (x != 0x56) return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); + c->dimensions = (get_bits(f, 8)<<8) + x; + x = get_bits(f, 8); + y = get_bits(f, 8); + c->entries = (get_bits(f, 8)<<16) + (y<<8) + x; + ordered = get_bits(f,1); + c->sparse = ordered ? 0 : get_bits(f,1); + + if (c->dimensions == 0 && c->entries != 0) return error(f, VORBIS_invalid_setup); + + if (c->sparse) + lengths = (uint8 *) setup_temp_malloc(f, c->entries); + else + lengths = c->codeword_lengths = (uint8 *) setup_malloc(f, c->entries); + + if (!lengths) return error(f, VORBIS_outofmem); + + if (ordered) { + int current_entry = 0; + int current_length = get_bits(f,5) + 1; + while (current_entry < c->entries) { + int limit = c->entries - current_entry; + int n = get_bits(f, ilog(limit)); + if (current_length >= 32) return error(f, VORBIS_invalid_setup); + if (current_entry + n > (int) c->entries) { return error(f, VORBIS_invalid_setup); } + memset(lengths + current_entry, current_length, n); + current_entry += n; + ++current_length; + } + } else { + for (j=0; j < c->entries; ++j) { + int present = c->sparse ? get_bits(f,1) : 1; + if (present) { + lengths[j] = get_bits(f, 5) + 1; + ++total; + if (lengths[j] == 32) + return error(f, VORBIS_invalid_setup); + } else { + lengths[j] = NO_CODE; + } + } + } + + if (c->sparse && total >= c->entries >> 2) { + // convert sparse items to non-sparse! + if (c->entries > (int) f->setup_temp_memory_required) + f->setup_temp_memory_required = c->entries; + + c->codeword_lengths = (uint8 *) setup_malloc(f, c->entries); + if (c->codeword_lengths == NULL) return error(f, VORBIS_outofmem); + memcpy(c->codeword_lengths, lengths, c->entries); + setup_temp_free(f, lengths, c->entries); // note this is only safe if there have been no intervening temp mallocs! + lengths = c->codeword_lengths; + c->sparse = 0; + } + + // compute the size of the sorted tables + if (c->sparse) { + sorted_count = total; + } else { + sorted_count = 0; + #ifndef STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH + for (j=0; j < c->entries; ++j) + if (lengths[j] > STB_VORBIS_FAST_HUFFMAN_LENGTH && lengths[j] != NO_CODE) + ++sorted_count; + #endif + } + + c->sorted_entries = sorted_count; + values = NULL; + + CHECK(f); + if (!c->sparse) { + c->codewords = (uint32 *) setup_malloc(f, sizeof(c->codewords[0]) * c->entries); + if (!c->codewords) return error(f, VORBIS_outofmem); + } else { + unsigned int size; + if (c->sorted_entries) { + c->codeword_lengths = (uint8 *) setup_malloc(f, c->sorted_entries); + if (!c->codeword_lengths) return error(f, VORBIS_outofmem); + c->codewords = (uint32 *) setup_temp_malloc(f, sizeof(*c->codewords) * c->sorted_entries); + if (!c->codewords) return error(f, VORBIS_outofmem); + values = (uint32 *) setup_temp_malloc(f, sizeof(*values) * c->sorted_entries); + if (!values) return error(f, VORBIS_outofmem); + } + size = c->entries + (sizeof(*c->codewords) + sizeof(*values)) * c->sorted_entries; + if (size > f->setup_temp_memory_required) + f->setup_temp_memory_required = size; + } + + if (!compute_codewords(c, lengths, c->entries, values)) { + if (c->sparse) setup_temp_free(f, values, 0); + return error(f, VORBIS_invalid_setup); + } + + if (c->sorted_entries) { + // allocate an extra slot for sentinels + c->sorted_codewords = (uint32 *) setup_malloc(f, sizeof(*c->sorted_codewords) * (c->sorted_entries+1)); + if (c->sorted_codewords == NULL) return error(f, VORBIS_outofmem); + // allocate an extra slot at the front so that c->sorted_values[-1] is defined + // so that we can catch that case without an extra if + c->sorted_values = ( int *) setup_malloc(f, sizeof(*c->sorted_values ) * (c->sorted_entries+1)); + if (c->sorted_values == NULL) return error(f, VORBIS_outofmem); + ++c->sorted_values; + c->sorted_values[-1] = -1; + compute_sorted_huffman(c, lengths, values); + } + + if (c->sparse) { + setup_temp_free(f, values, sizeof(*values)*c->sorted_entries); + setup_temp_free(f, c->codewords, sizeof(*c->codewords)*c->sorted_entries); + setup_temp_free(f, lengths, c->entries); + c->codewords = NULL; + } + + compute_accelerated_huffman(c); + + CHECK(f); + c->lookup_type = get_bits(f, 4); + if (c->lookup_type > 2) return error(f, VORBIS_invalid_setup); + if (c->lookup_type > 0) { + uint16 *mults; + c->minimum_value = float32_unpack(get_bits(f, 32)); + c->delta_value = float32_unpack(get_bits(f, 32)); + c->value_bits = get_bits(f, 4)+1; + c->sequence_p = get_bits(f,1); + if (c->lookup_type == 1) { + int values = lookup1_values(c->entries, c->dimensions); + if (values < 0) return error(f, VORBIS_invalid_setup); + c->lookup_values = (uint32) values; + } else { + c->lookup_values = c->entries * c->dimensions; + } + if (c->lookup_values == 0) return error(f, VORBIS_invalid_setup); + mults = (uint16 *) setup_temp_malloc(f, sizeof(mults[0]) * c->lookup_values); + if (mults == NULL) return error(f, VORBIS_outofmem); + for (j=0; j < (int) c->lookup_values; ++j) { + int q = get_bits(f, c->value_bits); + if (q == EOP) { setup_temp_free(f,mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_invalid_setup); } + mults[j] = q; + } + +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int len, sparse = c->sparse; + float last=0; + // pre-expand the lookup1-style multiplicands, to avoid a divide in the inner loop + if (sparse) { + if (c->sorted_entries == 0) goto skip; + c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->sorted_entries * c->dimensions); + } else + c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->entries * c->dimensions); + if (c->multiplicands == NULL) { setup_temp_free(f,mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_outofmem); } + len = sparse ? c->sorted_entries : c->entries; + for (j=0; j < len; ++j) { + unsigned int z = sparse ? c->sorted_values[j] : j; + unsigned int div=1; + for (k=0; k < c->dimensions; ++k) { + int off = (z / div) % c->lookup_values; + float val = mults[off]*c->delta_value + c->minimum_value + last; + c->multiplicands[j*c->dimensions + k] = val; + if (c->sequence_p) + last = val; + if (k+1 < c->dimensions) { + if (div > UINT_MAX / (unsigned int) c->lookup_values) { + setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values); + return error(f, VORBIS_invalid_setup); + } + div *= c->lookup_values; + } + } + } + c->lookup_type = 2; + } + else +#endif + { + float last=0; + CHECK(f); + c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->lookup_values); + if (c->multiplicands == NULL) { setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_outofmem); } + for (j=0; j < (int) c->lookup_values; ++j) { + float val = mults[j] * c->delta_value + c->minimum_value + last; + c->multiplicands[j] = val; + if (c->sequence_p) + last = val; + } + } +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + skip:; +#endif + setup_temp_free(f, mults, sizeof(mults[0])*c->lookup_values); + + CHECK(f); + } + CHECK(f); + } + + // time domain transfers (notused) + + x = get_bits(f, 6) + 1; + for (i=0; i < x; ++i) { + uint32 z = get_bits(f, 16); + if (z != 0) return error(f, VORBIS_invalid_setup); + } + + // Floors + f->floor_count = get_bits(f, 6)+1; + f->floor_config = (Floor *) setup_malloc(f, f->floor_count * sizeof(*f->floor_config)); + if (f->floor_config == NULL) return error(f, VORBIS_outofmem); + for (i=0; i < f->floor_count; ++i) { + f->floor_types[i] = get_bits(f, 16); + if (f->floor_types[i] > 1) return error(f, VORBIS_invalid_setup); + if (f->floor_types[i] == 0) { + Floor0 *g = &f->floor_config[i].floor0; + g->order = get_bits(f,8); + g->rate = get_bits(f,16); + g->bark_map_size = get_bits(f,16); + g->amplitude_bits = get_bits(f,6); + g->amplitude_offset = get_bits(f,8); + g->number_of_books = get_bits(f,4) + 1; + for (j=0; j < g->number_of_books; ++j) + g->book_list[j] = get_bits(f,8); + return error(f, VORBIS_feature_not_supported); + } else { + stbv__floor_ordering p[31*8+2]; + Floor1 *g = &f->floor_config[i].floor1; + int max_class = -1; + g->partitions = get_bits(f, 5); + for (j=0; j < g->partitions; ++j) { + g->partition_class_list[j] = get_bits(f, 4); + if (g->partition_class_list[j] > max_class) + max_class = g->partition_class_list[j]; + } + for (j=0; j <= max_class; ++j) { + g->class_dimensions[j] = get_bits(f, 3)+1; + g->class_subclasses[j] = get_bits(f, 2); + if (g->class_subclasses[j]) { + g->class_masterbooks[j] = get_bits(f, 8); + if (g->class_masterbooks[j] >= f->codebook_count) return error(f, VORBIS_invalid_setup); + } + for (k=0; k < 1 << g->class_subclasses[j]; ++k) { + g->subclass_books[j][k] = (int16)get_bits(f,8)-1; + if (g->subclass_books[j][k] >= f->codebook_count) return error(f, VORBIS_invalid_setup); + } + } + g->floor1_multiplier = get_bits(f,2)+1; + g->rangebits = get_bits(f,4); + g->Xlist[0] = 0; + g->Xlist[1] = 1 << g->rangebits; + g->values = 2; + for (j=0; j < g->partitions; ++j) { + int c = g->partition_class_list[j]; + for (k=0; k < g->class_dimensions[c]; ++k) { + g->Xlist[g->values] = get_bits(f, g->rangebits); + ++g->values; + } + } + // precompute the sorting + for (j=0; j < g->values; ++j) { + p[j].x = g->Xlist[j]; + p[j].id = j; + } + qsort(p, g->values, sizeof(p[0]), point_compare); + for (j=0; j < g->values-1; ++j) + if (p[j].x == p[j+1].x) + return error(f, VORBIS_invalid_setup); + for (j=0; j < g->values; ++j) + g->sorted_order[j] = (uint8) p[j].id; + // precompute the neighbors + for (j=2; j < g->values; ++j) { + int low = 0,hi = 0; + neighbors(g->Xlist, j, &low,&hi); + g->neighbors[j][0] = low; + g->neighbors[j][1] = hi; + } + + if (g->values > longest_floorlist) + longest_floorlist = g->values; + } + } + + // Residue + f->residue_count = get_bits(f, 6)+1; + f->residue_config = (Residue *) setup_malloc(f, f->residue_count * sizeof(f->residue_config[0])); + if (f->residue_config == NULL) return error(f, VORBIS_outofmem); + memset(f->residue_config, 0, f->residue_count * sizeof(f->residue_config[0])); + for (i=0; i < f->residue_count; ++i) { + uint8 residue_cascade[64]; + Residue *r = f->residue_config+i; + f->residue_types[i] = get_bits(f, 16); + if (f->residue_types[i] > 2) return error(f, VORBIS_invalid_setup); + r->begin = get_bits(f, 24); + r->end = get_bits(f, 24); + if (r->end < r->begin) return error(f, VORBIS_invalid_setup); + r->part_size = get_bits(f,24)+1; + r->classifications = get_bits(f,6)+1; + r->classbook = get_bits(f,8); + if (r->classbook >= f->codebook_count) return error(f, VORBIS_invalid_setup); + for (j=0; j < r->classifications; ++j) { + uint8 high_bits=0; + uint8 low_bits=get_bits(f,3); + if (get_bits(f,1)) + high_bits = get_bits(f,5); + residue_cascade[j] = high_bits*8 + low_bits; + } + r->residue_books = (short (*)[8]) setup_malloc(f, sizeof(r->residue_books[0]) * r->classifications); + if (r->residue_books == NULL) return error(f, VORBIS_outofmem); + for (j=0; j < r->classifications; ++j) { + for (k=0; k < 8; ++k) { + if (residue_cascade[j] & (1 << k)) { + r->residue_books[j][k] = get_bits(f, 8); + if (r->residue_books[j][k] >= f->codebook_count) return error(f, VORBIS_invalid_setup); + } else { + r->residue_books[j][k] = -1; + } + } + } + // precompute the classifications[] array to avoid inner-loop mod/divide + // call it 'classdata' since we already have r->classifications + r->classdata = (uint8 **) setup_malloc(f, sizeof(*r->classdata) * f->codebooks[r->classbook].entries); + if (!r->classdata) return error(f, VORBIS_outofmem); + memset(r->classdata, 0, sizeof(*r->classdata) * f->codebooks[r->classbook].entries); + for (j=0; j < f->codebooks[r->classbook].entries; ++j) { + int classwords = f->codebooks[r->classbook].dimensions; + int temp = j; + r->classdata[j] = (uint8 *) setup_malloc(f, sizeof(r->classdata[j][0]) * classwords); + if (r->classdata[j] == NULL) return error(f, VORBIS_outofmem); + for (k=classwords-1; k >= 0; --k) { + r->classdata[j][k] = temp % r->classifications; + temp /= r->classifications; + } + } + } + + f->mapping_count = get_bits(f,6)+1; + f->mapping = (Mapping *) setup_malloc(f, f->mapping_count * sizeof(*f->mapping)); + if (f->mapping == NULL) return error(f, VORBIS_outofmem); + memset(f->mapping, 0, f->mapping_count * sizeof(*f->mapping)); + for (i=0; i < f->mapping_count; ++i) { + Mapping *m = f->mapping + i; + int mapping_type = get_bits(f,16); + if (mapping_type != 0) return error(f, VORBIS_invalid_setup); + m->chan = (MappingChannel *) setup_malloc(f, f->channels * sizeof(*m->chan)); + if (m->chan == NULL) return error(f, VORBIS_outofmem); + if (get_bits(f,1)) + m->submaps = get_bits(f,4)+1; + else + m->submaps = 1; + if (m->submaps > max_submaps) + max_submaps = m->submaps; + if (get_bits(f,1)) { + m->coupling_steps = get_bits(f,8)+1; + if (m->coupling_steps > f->channels) return error(f, VORBIS_invalid_setup); + for (k=0; k < m->coupling_steps; ++k) { + m->chan[k].magnitude = get_bits(f, ilog(f->channels-1)); + m->chan[k].angle = get_bits(f, ilog(f->channels-1)); + if (m->chan[k].magnitude >= f->channels) return error(f, VORBIS_invalid_setup); + if (m->chan[k].angle >= f->channels) return error(f, VORBIS_invalid_setup); + if (m->chan[k].magnitude == m->chan[k].angle) return error(f, VORBIS_invalid_setup); + } + } else + m->coupling_steps = 0; + + // reserved field + if (get_bits(f,2)) return error(f, VORBIS_invalid_setup); + if (m->submaps > 1) { + for (j=0; j < f->channels; ++j) { + m->chan[j].mux = get_bits(f, 4); + if (m->chan[j].mux >= m->submaps) return error(f, VORBIS_invalid_setup); + } + } else + // @SPECIFICATION: this case is missing from the spec + for (j=0; j < f->channels; ++j) + m->chan[j].mux = 0; + + for (j=0; j < m->submaps; ++j) { + get_bits(f,8); // discard + m->submap_floor[j] = get_bits(f,8); + m->submap_residue[j] = get_bits(f,8); + if (m->submap_floor[j] >= f->floor_count) return error(f, VORBIS_invalid_setup); + if (m->submap_residue[j] >= f->residue_count) return error(f, VORBIS_invalid_setup); + } + } + + // Modes + f->mode_count = get_bits(f, 6)+1; + for (i=0; i < f->mode_count; ++i) { + Mode *m = f->mode_config+i; + m->blockflag = get_bits(f,1); + m->windowtype = get_bits(f,16); + m->transformtype = get_bits(f,16); + m->mapping = get_bits(f,8); + if (m->windowtype != 0) return error(f, VORBIS_invalid_setup); + if (m->transformtype != 0) return error(f, VORBIS_invalid_setup); + if (m->mapping >= f->mapping_count) return error(f, VORBIS_invalid_setup); + } + + flush_packet(f); + + f->previous_length = 0; + + for (i=0; i < f->channels; ++i) { + f->channel_buffers[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1); + f->previous_window[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1/2); + f->finalY[i] = (int16 *) setup_malloc(f, sizeof(int16) * longest_floorlist); + if (f->channel_buffers[i] == NULL || f->previous_window[i] == NULL || f->finalY[i] == NULL) return error(f, VORBIS_outofmem); + memset(f->channel_buffers[i], 0, sizeof(float) * f->blocksize_1); + #ifdef STB_VORBIS_NO_DEFER_FLOOR + f->floor_buffers[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1/2); + if (f->floor_buffers[i] == NULL) return error(f, VORBIS_outofmem); + #endif + } + + if (!init_blocksize(f, 0, f->blocksize_0)) return FALSE; + if (!init_blocksize(f, 1, f->blocksize_1)) return FALSE; + f->blocksize[0] = f->blocksize_0; + f->blocksize[1] = f->blocksize_1; + +#ifdef STB_VORBIS_DIVIDE_TABLE + if (integer_divide_table[1][1]==0) + for (i=0; i < DIVTAB_NUMER; ++i) + for (j=1; j < DIVTAB_DENOM; ++j) + integer_divide_table[i][j] = i / j; +#endif + + // compute how much temporary memory is needed + + // 1. + { + uint32 imdct_mem = (f->blocksize_1 * sizeof(float) >> 1); + uint32 classify_mem; + int i,max_part_read=0; + for (i=0; i < f->residue_count; ++i) { + Residue *r = f->residue_config + i; + unsigned int actual_size = f->blocksize_1 / 2; + unsigned int limit_r_begin = r->begin < actual_size ? r->begin : actual_size; + unsigned int limit_r_end = r->end < actual_size ? r->end : actual_size; + int n_read = limit_r_end - limit_r_begin; + int part_read = n_read / r->part_size; + if (part_read > max_part_read) + max_part_read = part_read; + } + #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + classify_mem = f->channels * (sizeof(void*) + max_part_read * sizeof(uint8 *)); + #else + classify_mem = f->channels * (sizeof(void*) + max_part_read * sizeof(int *)); + #endif + + // maximum reasonable partition size is f->blocksize_1 + + f->temp_memory_required = classify_mem; + if (imdct_mem > f->temp_memory_required) + f->temp_memory_required = imdct_mem; + } + + + if (f->alloc.alloc_buffer) { + assert(f->temp_offset == f->alloc.alloc_buffer_length_in_bytes); + // check if there's enough temp memory so we don't error later + if (f->setup_offset + sizeof(*f) + f->temp_memory_required > (unsigned) f->temp_offset) + return error(f, VORBIS_outofmem); + } + + // @TODO: stb_vorbis_seek_start expects first_audio_page_offset to point to a page + // without PAGEFLAG_continued_packet, so this either points to the first page, or + // the page after the end of the headers. It might be cleaner to point to a page + // in the middle of the headers, when that's the page where the first audio packet + // starts, but we'd have to also correctly skip the end of any continued packet in + // stb_vorbis_seek_start. + if (f->next_seg == -1) { + f->first_audio_page_offset = stb_vorbis_get_file_offset(f); + } else { + f->first_audio_page_offset = 0; + } + + return TRUE; +} + +static void vorbis_deinit(stb_vorbis *p) +{ + int i,j; + + setup_free(p, p->vendor); + for (i=0; i < p->comment_list_length; ++i) { + setup_free(p, p->comment_list[i]); + } + setup_free(p, p->comment_list); + + if (p->residue_config) { + for (i=0; i < p->residue_count; ++i) { + Residue *r = p->residue_config+i; + if (r->classdata) { + for (j=0; j < p->codebooks[r->classbook].entries; ++j) + setup_free(p, r->classdata[j]); + setup_free(p, r->classdata); + } + setup_free(p, r->residue_books); + } + } + + if (p->codebooks) { + CHECK(p); + for (i=0; i < p->codebook_count; ++i) { + Codebook *c = p->codebooks + i; + setup_free(p, c->codeword_lengths); + setup_free(p, c->multiplicands); + setup_free(p, c->codewords); + setup_free(p, c->sorted_codewords); + // c->sorted_values[-1] is the first entry in the array + setup_free(p, c->sorted_values ? c->sorted_values-1 : NULL); + } + setup_free(p, p->codebooks); + } + setup_free(p, p->floor_config); + setup_free(p, p->residue_config); + if (p->mapping) { + for (i=0; i < p->mapping_count; ++i) + setup_free(p, p->mapping[i].chan); + setup_free(p, p->mapping); + } + CHECK(p); + for (i=0; i < p->channels && i < STB_VORBIS_MAX_CHANNELS; ++i) { + setup_free(p, p->channel_buffers[i]); + setup_free(p, p->previous_window[i]); + #ifdef STB_VORBIS_NO_DEFER_FLOOR + setup_free(p, p->floor_buffers[i]); + #endif + setup_free(p, p->finalY[i]); + } + for (i=0; i < 2; ++i) { + setup_free(p, p->A[i]); + setup_free(p, p->B[i]); + setup_free(p, p->C[i]); + setup_free(p, p->window[i]); + setup_free(p, p->bit_reverse[i]); + } + #ifndef STB_VORBIS_NO_STDIO + if (p->close_on_free) fclose(p->f); + #endif +} + +void stb_vorbis_close(stb_vorbis *p) +{ + if (p == NULL) return; + vorbis_deinit(p); + setup_free(p,p); +} + +static void vorbis_init(stb_vorbis *p, const stb_vorbis_alloc *z) +{ + memset(p, 0, sizeof(*p)); // NULL out all malloc'd pointers to start + if (z) { + p->alloc = *z; + p->alloc.alloc_buffer_length_in_bytes &= ~7; + p->temp_offset = p->alloc.alloc_buffer_length_in_bytes; + } + p->eof = 0; + p->error = VORBIS__no_error; + p->stream = NULL; + p->codebooks = NULL; + p->page_crc_tests = -1; + #ifndef STB_VORBIS_NO_STDIO + p->close_on_free = FALSE; + p->f = NULL; + #endif +} + +int stb_vorbis_get_sample_offset(stb_vorbis *f) +{ + if (f->current_loc_valid) + return f->current_loc; + else + return -1; +} + +stb_vorbis_info stb_vorbis_get_info(stb_vorbis *f) +{ + stb_vorbis_info d; + d.channels = f->channels; + d.sample_rate = f->sample_rate; + d.setup_memory_required = f->setup_memory_required; + d.setup_temp_memory_required = f->setup_temp_memory_required; + d.temp_memory_required = f->temp_memory_required; + d.max_frame_size = f->blocksize_1 >> 1; + return d; +} + +stb_vorbis_comment stb_vorbis_get_comment(stb_vorbis *f) +{ + stb_vorbis_comment d; + d.vendor = f->vendor; + d.comment_list_length = f->comment_list_length; + d.comment_list = f->comment_list; + return d; +} + +int stb_vorbis_get_error(stb_vorbis *f) +{ + int e = f->error; + f->error = VORBIS__no_error; + return e; +} + +static stb_vorbis * vorbis_alloc(stb_vorbis *f) +{ + stb_vorbis *p = (stb_vorbis *) setup_malloc(f, sizeof(*p)); + return p; +} + +#ifndef STB_VORBIS_NO_PUSHDATA_API + +void stb_vorbis_flush_pushdata(stb_vorbis *f) +{ + f->previous_length = 0; + f->page_crc_tests = 0; + f->discard_samples_deferred = 0; + f->current_loc_valid = FALSE; + f->first_decode = FALSE; + f->samples_output = 0; + f->channel_buffer_start = 0; + f->channel_buffer_end = 0; +} + +static int vorbis_search_for_page_pushdata(vorb *f, uint8 *data, int data_len) +{ + int i,n; + for (i=0; i < f->page_crc_tests; ++i) + f->scan[i].bytes_done = 0; + + // if we have room for more scans, search for them first, because + // they may cause us to stop early if their header is incomplete + if (f->page_crc_tests < STB_VORBIS_PUSHDATA_CRC_COUNT) { + if (data_len < 4) return 0; + data_len -= 3; // need to look for 4-byte sequence, so don't miss + // one that straddles a boundary + for (i=0; i < data_len; ++i) { + if (data[i] == 0x4f) { + if (0==memcmp(data+i, ogg_page_header, 4)) { + int j,len; + uint32 crc; + // make sure we have the whole page header + if (i+26 >= data_len || i+27+data[i+26] >= data_len) { + // only read up to this page start, so hopefully we'll + // have the whole page header start next time + data_len = i; + break; + } + // ok, we have it all; compute the length of the page + len = 27 + data[i+26]; + for (j=0; j < data[i+26]; ++j) + len += data[i+27+j]; + // scan everything up to the embedded crc (which we must 0) + crc = 0; + for (j=0; j < 22; ++j) + crc = crc32_update(crc, data[i+j]); + // now process 4 0-bytes + for ( ; j < 26; ++j) + crc = crc32_update(crc, 0); + // len is the total number of bytes we need to scan + n = f->page_crc_tests++; + f->scan[n].bytes_left = len-j; + f->scan[n].crc_so_far = crc; + f->scan[n].goal_crc = data[i+22] + (data[i+23] << 8) + (data[i+24]<<16) + (data[i+25]<<24); + // if the last frame on a page is continued to the next, then + // we can't recover the sample_loc immediately + if (data[i+27+data[i+26]-1] == 255) + f->scan[n].sample_loc = ~0; + else + f->scan[n].sample_loc = data[i+6] + (data[i+7] << 8) + (data[i+ 8]<<16) + (data[i+ 9]<<24); + f->scan[n].bytes_done = i+j; + if (f->page_crc_tests == STB_VORBIS_PUSHDATA_CRC_COUNT) + break; + // keep going if we still have room for more + } + } + } + } + + for (i=0; i < f->page_crc_tests;) { + uint32 crc; + int j; + int n = f->scan[i].bytes_done; + int m = f->scan[i].bytes_left; + if (m > data_len - n) m = data_len - n; + // m is the bytes to scan in the current chunk + crc = f->scan[i].crc_so_far; + for (j=0; j < m; ++j) + crc = crc32_update(crc, data[n+j]); + f->scan[i].bytes_left -= m; + f->scan[i].crc_so_far = crc; + if (f->scan[i].bytes_left == 0) { + // does it match? + if (f->scan[i].crc_so_far == f->scan[i].goal_crc) { + // Houston, we have page + data_len = n+m; // consumption amount is wherever that scan ended + f->page_crc_tests = -1; // drop out of page scan mode + f->previous_length = 0; // decode-but-don't-output one frame + f->next_seg = -1; // start a new page + f->current_loc = f->scan[i].sample_loc; // set the current sample location + // to the amount we'd have decoded had we decoded this page + f->current_loc_valid = f->current_loc != ~0U; + return data_len; + } + // delete entry + f->scan[i] = f->scan[--f->page_crc_tests]; + } else { + ++i; + } + } + + return data_len; +} + +// return value: number of bytes we used +int stb_vorbis_decode_frame_pushdata( + stb_vorbis *f, // the file we're decoding + const uint8 *data, int data_len, // the memory available for decoding + int *channels, // place to write number of float * buffers + float ***output, // place to write float ** array of float * buffers + int *samples // place to write number of output samples + ) +{ + int i; + int len,right,left; + + if (!IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); + + if (f->page_crc_tests >= 0) { + *samples = 0; + return vorbis_search_for_page_pushdata(f, (uint8 *) data, data_len); + } + + f->stream = (uint8 *) data; + f->stream_end = (uint8 *) data + data_len; + f->error = VORBIS__no_error; + + // check that we have the entire packet in memory + if (!is_whole_packet_present(f)) { + *samples = 0; + return 0; + } + + if (!vorbis_decode_packet(f, &len, &left, &right)) { + // save the actual error we encountered + enum STBVorbisError error = f->error; + if (error == VORBIS_bad_packet_type) { + // flush and resynch + f->error = VORBIS__no_error; + while (get8_packet(f) != EOP) + if (f->eof) break; + *samples = 0; + return (int) (f->stream - data); + } + if (error == VORBIS_continued_packet_flag_invalid) { + if (f->previous_length == 0) { + // we may be resynching, in which case it's ok to hit one + // of these; just discard the packet + f->error = VORBIS__no_error; + while (get8_packet(f) != EOP) + if (f->eof) break; + *samples = 0; + return (int) (f->stream - data); + } + } + // if we get an error while parsing, what to do? + // well, it DEFINITELY won't work to continue from where we are! + stb_vorbis_flush_pushdata(f); + // restore the error that actually made us bail + f->error = error; + *samples = 0; + return 1; + } + + // success! + len = vorbis_finish_frame(f, len, left, right); + for (i=0; i < f->channels; ++i) + f->outputs[i] = f->channel_buffers[i] + left; + + if (channels) *channels = f->channels; + *samples = len; + *output = f->outputs; + return (int) (f->stream - data); +} + +stb_vorbis *stb_vorbis_open_pushdata( + const unsigned char *data, int data_len, // the memory available for decoding + int *data_used, // only defined if result is not NULL + int *error, const stb_vorbis_alloc *alloc) +{ + stb_vorbis *f, p; + vorbis_init(&p, alloc); + p.stream = (uint8 *) data; + p.stream_end = (uint8 *) data + data_len; + p.push_mode = TRUE; + if (!start_decoder(&p)) { + if (p.eof) + *error = VORBIS_need_more_data; + else + *error = p.error; + vorbis_deinit(&p); + return NULL; + } + f = vorbis_alloc(&p); + if (f) { + *f = p; + *data_used = (int) (f->stream - data); + *error = 0; + return f; + } else { + vorbis_deinit(&p); + return NULL; + } +} +#endif // STB_VORBIS_NO_PUSHDATA_API + +unsigned int stb_vorbis_get_file_offset(stb_vorbis *f) +{ + #ifndef STB_VORBIS_NO_PUSHDATA_API + if (f->push_mode) return 0; + #endif + if (USE_MEMORY(f)) return (unsigned int) (f->stream - f->stream_start); + #ifndef STB_VORBIS_NO_STDIO + return (unsigned int) (ftell(f->f) - f->f_start); + #endif +} + +#ifndef STB_VORBIS_NO_PULLDATA_API +// +// DATA-PULLING API +// + +static uint32 vorbis_find_page(stb_vorbis *f, uint32 *end, uint32 *last) +{ + for(;;) { + int n; + if (f->eof) return 0; + n = get8(f); + if (n == 0x4f) { // page header candidate + unsigned int retry_loc = stb_vorbis_get_file_offset(f); + int i; + // check if we're off the end of a file_section stream + if (retry_loc - 25 > f->stream_len) + return 0; + // check the rest of the header + for (i=1; i < 4; ++i) + if (get8(f) != ogg_page_header[i]) + break; + if (f->eof) return 0; + if (i == 4) { + uint8 header[27]; + uint32 i, crc, goal, len; + for (i=0; i < 4; ++i) + header[i] = ogg_page_header[i]; + for (; i < 27; ++i) + header[i] = get8(f); + if (f->eof) return 0; + if (header[4] != 0) goto invalid; + goal = header[22] + (header[23] << 8) + (header[24]<<16) + ((uint32)header[25]<<24); + for (i=22; i < 26; ++i) + header[i] = 0; + crc = 0; + for (i=0; i < 27; ++i) + crc = crc32_update(crc, header[i]); + len = 0; + for (i=0; i < header[26]; ++i) { + int s = get8(f); + crc = crc32_update(crc, s); + len += s; + } + if (len && f->eof) return 0; + for (i=0; i < len; ++i) + crc = crc32_update(crc, get8(f)); + // finished parsing probable page + if (crc == goal) { + // we could now check that it's either got the last + // page flag set, OR it's followed by the capture + // pattern, but I guess TECHNICALLY you could have + // a file with garbage between each ogg page and recover + // from it automatically? So even though that paranoia + // might decrease the chance of an invalid decode by + // another 2^32, not worth it since it would hose those + // invalid-but-useful files? + if (end) + *end = stb_vorbis_get_file_offset(f); + if (last) { + if (header[5] & 0x04) + *last = 1; + else + *last = 0; + } + set_file_offset(f, retry_loc-1); + return 1; + } + } + invalid: + // not a valid page, so rewind and look for next one + set_file_offset(f, retry_loc); + } + } +} + + +#define SAMPLE_unknown 0xffffffff + +// seeking is implemented with a binary search, which narrows down the range to +// 64K, before using a linear search (because finding the synchronization +// pattern can be expensive, and the chance we'd find the end page again is +// relatively high for small ranges) +// +// two initial interpolation-style probes are used at the start of the search +// to try to bound either side of the binary search sensibly, while still +// working in O(log n) time if they fail. + +static int get_seek_page_info(stb_vorbis *f, ProbedPage *z) +{ + uint8 header[27], lacing[255]; + int i,len; + + // record where the page starts + z->page_start = stb_vorbis_get_file_offset(f); + + // parse the header + getn(f, header, 27); + if (header[0] != 'O' || header[1] != 'g' || header[2] != 'g' || header[3] != 'S') + return 0; + getn(f, lacing, header[26]); + + // determine the length of the payload + len = 0; + for (i=0; i < header[26]; ++i) + len += lacing[i]; + + // this implies where the page ends + z->page_end = z->page_start + 27 + header[26] + len; + + // read the last-decoded sample out of the data + z->last_decoded_sample = header[6] + (header[7] << 8) + (header[8] << 16) + (header[9] << 24); + + // restore file state to where we were + set_file_offset(f, z->page_start); + return 1; +} + +// rarely used function to seek back to the preceding page while finding the +// start of a packet +static int go_to_page_before(stb_vorbis *f, unsigned int limit_offset) +{ + unsigned int previous_safe, end; + + // now we want to seek back 64K from the limit + if (limit_offset >= 65536 && limit_offset-65536 >= f->first_audio_page_offset) + previous_safe = limit_offset - 65536; + else + previous_safe = f->first_audio_page_offset; + + set_file_offset(f, previous_safe); + + while (vorbis_find_page(f, &end, NULL)) { + if (end >= limit_offset && stb_vorbis_get_file_offset(f) < limit_offset) + return 1; + set_file_offset(f, end); + } + + return 0; +} + +// implements the search logic for finding a page and starting decoding. if +// the function succeeds, current_loc_valid will be true and current_loc will +// be less than or equal to the provided sample number (the closer the +// better). +static int seek_to_sample_coarse(stb_vorbis *f, uint32 sample_number) +{ + ProbedPage left, right, mid; + int i, start_seg_with_known_loc, end_pos, page_start; + uint32 delta, stream_length, padding, last_sample_limit; + double offset = 0.0, bytes_per_sample = 0.0; + int probe = 0; + + // find the last page and validate the target sample + stream_length = stb_vorbis_stream_length_in_samples(f); + if (stream_length == 0) return error(f, VORBIS_seek_without_length); + if (sample_number > stream_length) return error(f, VORBIS_seek_invalid); + + // this is the maximum difference between the window-center (which is the + // actual granule position value), and the right-start (which the spec + // indicates should be the granule position (give or take one)). + padding = ((f->blocksize_1 - f->blocksize_0) >> 2); + if (sample_number < padding) + last_sample_limit = 0; + else + last_sample_limit = sample_number - padding; + + left = f->p_first; + while (left.last_decoded_sample == ~0U) { + // (untested) the first page does not have a 'last_decoded_sample' + set_file_offset(f, left.page_end); + if (!get_seek_page_info(f, &left)) goto error; + } + + right = f->p_last; + assert(right.last_decoded_sample != ~0U); + + // starting from the start is handled differently + if (last_sample_limit <= left.last_decoded_sample) { + if (stb_vorbis_seek_start(f)) { + if (f->current_loc > sample_number) + return error(f, VORBIS_seek_failed); + return 1; + } + return 0; + } + + while (left.page_end != right.page_start) { + assert(left.page_end < right.page_start); + // search range in bytes + delta = right.page_start - left.page_end; + if (delta <= 65536) { + // there's only 64K left to search - handle it linearly + set_file_offset(f, left.page_end); + } else { + if (probe < 2) { + if (probe == 0) { + // first probe (interpolate) + double data_bytes = right.page_end - left.page_start; + bytes_per_sample = data_bytes / right.last_decoded_sample; + offset = left.page_start + bytes_per_sample * (last_sample_limit - left.last_decoded_sample); + } else { + // second probe (try to bound the other side) + double error = ((double) last_sample_limit - mid.last_decoded_sample) * bytes_per_sample; + if (error >= 0 && error < 8000) error = 8000; + if (error < 0 && error > -8000) error = -8000; + offset += error * 2; + } + + // ensure the offset is valid + if (offset < left.page_end) + offset = left.page_end; + if (offset > right.page_start - 65536) + offset = right.page_start - 65536; + + set_file_offset(f, (unsigned int) offset); + } else { + // binary search for large ranges (offset by 32K to ensure + // we don't hit the right page) + set_file_offset(f, left.page_end + (delta / 2) - 32768); + } + + if (!vorbis_find_page(f, NULL, NULL)) goto error; + } + + for (;;) { + if (!get_seek_page_info(f, &mid)) goto error; + if (mid.last_decoded_sample != ~0U) break; + // (untested) no frames end on this page + set_file_offset(f, mid.page_end); + assert(mid.page_start < right.page_start); + } + + // if we've just found the last page again then we're in a tricky file, + // and we're close enough (if it wasn't an interpolation probe). + if (mid.page_start == right.page_start) { + if (probe >= 2 || delta <= 65536) + break; + } else { + if (last_sample_limit < mid.last_decoded_sample) + right = mid; + else + left = mid; + } + + ++probe; + } + + // seek back to start of the last packet + page_start = left.page_start; + set_file_offset(f, page_start); + if (!start_page(f)) return error(f, VORBIS_seek_failed); + end_pos = f->end_seg_with_known_loc; + assert(end_pos >= 0); + + for (;;) { + for (i = end_pos; i > 0; --i) + if (f->segments[i-1] != 255) + break; + + start_seg_with_known_loc = i; + + if (start_seg_with_known_loc > 0 || !(f->page_flag & PAGEFLAG_continued_packet)) + break; + + // (untested) the final packet begins on an earlier page + if (!go_to_page_before(f, page_start)) + goto error; + + page_start = stb_vorbis_get_file_offset(f); + if (!start_page(f)) goto error; + end_pos = f->segment_count - 1; + } + + // prepare to start decoding + f->current_loc_valid = FALSE; + f->last_seg = FALSE; + f->valid_bits = 0; + f->packet_bytes = 0; + f->bytes_in_seg = 0; + f->previous_length = 0; + f->next_seg = start_seg_with_known_loc; + + for (i = 0; i < start_seg_with_known_loc; i++) + skip(f, f->segments[i]); + + // start decoding (optimizable - this frame is generally discarded) + if (!vorbis_pump_first_frame(f)) + return 0; + if (f->current_loc > sample_number) + return error(f, VORBIS_seek_failed); + return 1; + +error: + // try to restore the file to a valid state + stb_vorbis_seek_start(f); + return error(f, VORBIS_seek_failed); +} + +// the same as vorbis_decode_initial, but without advancing +static int peek_decode_initial(vorb *f, int *p_left_start, int *p_left_end, int *p_right_start, int *p_right_end, int *mode) +{ + int bits_read, bytes_read; + + if (!vorbis_decode_initial(f, p_left_start, p_left_end, p_right_start, p_right_end, mode)) + return 0; + + // either 1 or 2 bytes were read, figure out which so we can rewind + bits_read = 1 + ilog(f->mode_count-1); + if (f->mode_config[*mode].blockflag) + bits_read += 2; + bytes_read = (bits_read + 7) / 8; + + f->bytes_in_seg += bytes_read; + f->packet_bytes -= bytes_read; + skip(f, -bytes_read); + if (f->next_seg == -1) + f->next_seg = f->segment_count - 1; + else + f->next_seg--; + f->valid_bits = 0; + + return 1; +} + +int stb_vorbis_seek_frame(stb_vorbis *f, unsigned int sample_number) +{ + uint32 max_frame_samples; + + if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); + + // fast page-level search + if (!seek_to_sample_coarse(f, sample_number)) + return 0; + + assert(f->current_loc_valid); + assert(f->current_loc <= sample_number); + + // linear search for the relevant packet + max_frame_samples = (f->blocksize_1*3 - f->blocksize_0) >> 2; + while (f->current_loc < sample_number) { + int left_start, left_end, right_start, right_end, mode, frame_samples; + if (!peek_decode_initial(f, &left_start, &left_end, &right_start, &right_end, &mode)) + return error(f, VORBIS_seek_failed); + // calculate the number of samples returned by the next frame + frame_samples = right_start - left_start; + if (f->current_loc + frame_samples > sample_number) { + return 1; // the next frame will contain the sample + } else if (f->current_loc + frame_samples + max_frame_samples > sample_number) { + // there's a chance the frame after this could contain the sample + vorbis_pump_first_frame(f); + } else { + // this frame is too early to be relevant + f->current_loc += frame_samples; + f->previous_length = 0; + maybe_start_packet(f); + flush_packet(f); + } + } + // the next frame should start with the sample + if (f->current_loc != sample_number) return error(f, VORBIS_seek_failed); + return 1; +} + +int stb_vorbis_seek(stb_vorbis *f, unsigned int sample_number) +{ + if (!stb_vorbis_seek_frame(f, sample_number)) + return 0; + + if (sample_number != f->current_loc) { + int n; + uint32 frame_start = f->current_loc; + stb_vorbis_get_frame_float(f, &n, NULL); + assert(sample_number > frame_start); + assert(f->channel_buffer_start + (int) (sample_number-frame_start) <= f->channel_buffer_end); + f->channel_buffer_start += (sample_number - frame_start); + } + + return 1; +} + +int stb_vorbis_seek_start(stb_vorbis *f) +{ + if (IS_PUSH_MODE(f)) { return error(f, VORBIS_invalid_api_mixing); } + set_file_offset(f, f->first_audio_page_offset); + f->previous_length = 0; + f->first_decode = TRUE; + f->next_seg = -1; + return vorbis_pump_first_frame(f); +} + +unsigned int stb_vorbis_stream_length_in_samples(stb_vorbis *f) +{ + unsigned int restore_offset, previous_safe; + unsigned int end, last_page_loc; + + if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); + if (!f->total_samples) { + unsigned int last; + uint32 lo,hi; + char header[6]; + + // first, store the current decode position so we can restore it + restore_offset = stb_vorbis_get_file_offset(f); + + // now we want to seek back 64K from the end (the last page must + // be at most a little less than 64K, but let's allow a little slop) + if (f->stream_len >= 65536 && f->stream_len-65536 >= f->first_audio_page_offset) + previous_safe = f->stream_len - 65536; + else + previous_safe = f->first_audio_page_offset; + + set_file_offset(f, previous_safe); + // previous_safe is now our candidate 'earliest known place that seeking + // to will lead to the final page' + + if (!vorbis_find_page(f, &end, &last)) { + // if we can't find a page, we're hosed! + f->error = VORBIS_cant_find_last_page; + f->total_samples = 0xffffffff; + goto done; + } + + // check if there are more pages + last_page_loc = stb_vorbis_get_file_offset(f); + + // stop when the last_page flag is set, not when we reach eof; + // this allows us to stop short of a 'file_section' end without + // explicitly checking the length of the section + while (!last) { + set_file_offset(f, end); + if (!vorbis_find_page(f, &end, &last)) { + // the last page we found didn't have the 'last page' flag + // set. whoops! + break; + } + //previous_safe = last_page_loc+1; // NOTE: not used after this point, but note for debugging + last_page_loc = stb_vorbis_get_file_offset(f); + } + + set_file_offset(f, last_page_loc); + + // parse the header + getn(f, (unsigned char *)header, 6); + // extract the absolute granule position + lo = get32(f); + hi = get32(f); + if (lo == 0xffffffff && hi == 0xffffffff) { + f->error = VORBIS_cant_find_last_page; + f->total_samples = SAMPLE_unknown; + goto done; + } + if (hi) + lo = 0xfffffffe; // saturate + f->total_samples = lo; + + f->p_last.page_start = last_page_loc; + f->p_last.page_end = end; + f->p_last.last_decoded_sample = lo; + + done: + set_file_offset(f, restore_offset); + } + return f->total_samples == SAMPLE_unknown ? 0 : f->total_samples; +} + +float stb_vorbis_stream_length_in_seconds(stb_vorbis *f) +{ + return stb_vorbis_stream_length_in_samples(f) / (float) f->sample_rate; +} + + + +int stb_vorbis_get_frame_float(stb_vorbis *f, int *channels, float ***output) +{ + int len, right,left,i; + if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); + + if (!vorbis_decode_packet(f, &len, &left, &right)) { + f->channel_buffer_start = f->channel_buffer_end = 0; + return 0; + } + + len = vorbis_finish_frame(f, len, left, right); + for (i=0; i < f->channels; ++i) + f->outputs[i] = f->channel_buffers[i] + left; + + f->channel_buffer_start = left; + f->channel_buffer_end = left+len; + + if (channels) *channels = f->channels; + if (output) *output = f->outputs; + return len; +} + +#ifndef STB_VORBIS_NO_STDIO + +stb_vorbis * stb_vorbis_open_file_section(FILE *file, int close_on_free, int *error, const stb_vorbis_alloc *alloc, unsigned int length) +{ + stb_vorbis *f, p; + vorbis_init(&p, alloc); + p.f = file; + p.f_start = (uint32) ftell(file); + p.stream_len = length; + p.close_on_free = close_on_free; + if (start_decoder(&p)) { + f = vorbis_alloc(&p); + if (f) { + *f = p; + vorbis_pump_first_frame(f); + return f; + } + } + if (error) *error = p.error; + vorbis_deinit(&p); + return NULL; +} + +stb_vorbis * stb_vorbis_open_file(FILE *file, int close_on_free, int *error, const stb_vorbis_alloc *alloc) +{ + unsigned int len, start; + start = (unsigned int) ftell(file); + fseek(file, 0, SEEK_END); + len = (unsigned int) (ftell(file) - start); + fseek(file, start, SEEK_SET); + return stb_vorbis_open_file_section(file, close_on_free, error, alloc, len); +} + +stb_vorbis * stb_vorbis_open_filename(const char *filename, int *error, const stb_vorbis_alloc *alloc) +{ + FILE *f; +#if defined(_WIN32) && defined(__STDC_WANT_SECURE_LIB__) + if (0 != fopen_s(&f, filename, "rb")) + f = NULL; +#else + f = fopen(filename, "rb"); +#endif + if (f) + return stb_vorbis_open_file(f, TRUE, error, alloc); + if (error) *error = VORBIS_file_open_failure; + return NULL; +} +#endif // STB_VORBIS_NO_STDIO + +stb_vorbis * stb_vorbis_open_memory(const unsigned char *data, int len, int *error, const stb_vorbis_alloc *alloc) +{ + stb_vorbis *f, p; + if (!data) { + if (error) *error = VORBIS_unexpected_eof; + return NULL; + } + vorbis_init(&p, alloc); + p.stream = (uint8 *) data; + p.stream_end = (uint8 *) data + len; + p.stream_start = (uint8 *) p.stream; + p.stream_len = len; + p.push_mode = FALSE; + if (start_decoder(&p)) { + f = vorbis_alloc(&p); + if (f) { + *f = p; + vorbis_pump_first_frame(f); + if (error) *error = VORBIS__no_error; + return f; + } + } + if (error) *error = p.error; + vorbis_deinit(&p); + return NULL; +} + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +#define PLAYBACK_MONO 1 +#define PLAYBACK_LEFT 2 +#define PLAYBACK_RIGHT 4 + +#define L (PLAYBACK_LEFT | PLAYBACK_MONO) +#define C (PLAYBACK_LEFT | PLAYBACK_RIGHT | PLAYBACK_MONO) +#define R (PLAYBACK_RIGHT | PLAYBACK_MONO) + +static int8 channel_position[7][6] = +{ + { 0 }, + { C }, + { L, R }, + { L, C, R }, + { L, R, L, R }, + { L, C, R, L, R }, + { L, C, R, L, R, C }, +}; + + +#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT + typedef union { + float f; + int i; + } float_conv; + typedef char stb_vorbis_float_size_test[sizeof(float)==4 && sizeof(int) == 4]; + #define FASTDEF(x) float_conv x + // add (1<<23) to convert to int, then divide by 2^SHIFT, then add 0.5/2^SHIFT to round + #define MAGIC(SHIFT) (1.5f * (1 << (23-SHIFT)) + 0.5f/(1 << SHIFT)) + #define ADDEND(SHIFT) (((150-SHIFT) << 23) + (1 << 22)) + #define FAST_SCALED_FLOAT_TO_INT(temp,x,s) (temp.f = (x) + MAGIC(s), temp.i - ADDEND(s)) + #define check_endianness() +#else + #define FAST_SCALED_FLOAT_TO_INT(temp,x,s) ((int) ((x) * (1 << (s)))) + #define check_endianness() + #define FASTDEF(x) +#endif + +static void copy_samples(short *dest, float *src, int len) +{ + int i; + check_endianness(); + for (i=0; i < len; ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp, src[i],15); + if ((unsigned int) (v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + dest[i] = v; + } +} + +static void compute_samples(int mask, short *output, int num_c, float **data, int d_offset, int len) +{ + #define STB_BUFFER_SIZE 32 + float buffer[STB_BUFFER_SIZE]; + int i,j,o,n = STB_BUFFER_SIZE; + check_endianness(); + for (o = 0; o < len; o += STB_BUFFER_SIZE) { + memset(buffer, 0, sizeof(buffer)); + if (o + n > len) n = len - o; + for (j=0; j < num_c; ++j) { + if (channel_position[num_c][j] & mask) { + for (i=0; i < n; ++i) + buffer[i] += data[j][d_offset+o+i]; + } + } + for (i=0; i < n; ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp,buffer[i],15); + if ((unsigned int) (v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + output[o+i] = v; + } + } + #undef STB_BUFFER_SIZE +} + +static void compute_stereo_samples(short *output, int num_c, float **data, int d_offset, int len) +{ + #define STB_BUFFER_SIZE 32 + float buffer[STB_BUFFER_SIZE]; + int i,j,o,n = STB_BUFFER_SIZE >> 1; + // o is the offset in the source data + check_endianness(); + for (o = 0; o < len; o += STB_BUFFER_SIZE >> 1) { + // o2 is the offset in the output data + int o2 = o << 1; + memset(buffer, 0, sizeof(buffer)); + if (o + n > len) n = len - o; + for (j=0; j < num_c; ++j) { + int m = channel_position[num_c][j] & (PLAYBACK_LEFT | PLAYBACK_RIGHT); + if (m == (PLAYBACK_LEFT | PLAYBACK_RIGHT)) { + for (i=0; i < n; ++i) { + buffer[i*2+0] += data[j][d_offset+o+i]; + buffer[i*2+1] += data[j][d_offset+o+i]; + } + } else if (m == PLAYBACK_LEFT) { + for (i=0; i < n; ++i) { + buffer[i*2+0] += data[j][d_offset+o+i]; + } + } else if (m == PLAYBACK_RIGHT) { + for (i=0; i < n; ++i) { + buffer[i*2+1] += data[j][d_offset+o+i]; + } + } + } + for (i=0; i < (n<<1); ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp,buffer[i],15); + if ((unsigned int) (v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + output[o2+i] = v; + } + } + #undef STB_BUFFER_SIZE +} + +static void convert_samples_short(int buf_c, short **buffer, int b_offset, int data_c, float **data, int d_offset, int samples) +{ + int i; + if (buf_c != data_c && buf_c <= 2 && data_c <= 6) { + static int channel_selector[3][2] = { {0}, {PLAYBACK_MONO}, {PLAYBACK_LEFT, PLAYBACK_RIGHT} }; + for (i=0; i < buf_c; ++i) + compute_samples(channel_selector[buf_c][i], buffer[i]+b_offset, data_c, data, d_offset, samples); + } else { + int limit = buf_c < data_c ? buf_c : data_c; + for (i=0; i < limit; ++i) + copy_samples(buffer[i]+b_offset, data[i]+d_offset, samples); + for ( ; i < buf_c; ++i) + memset(buffer[i]+b_offset, 0, sizeof(short) * samples); + } +} + +int stb_vorbis_get_frame_short(stb_vorbis *f, int num_c, short **buffer, int num_samples) +{ + float **output = NULL; + int len = stb_vorbis_get_frame_float(f, NULL, &output); + if (len > num_samples) len = num_samples; + if (len) + convert_samples_short(num_c, buffer, 0, f->channels, output, 0, len); + return len; +} + +static void convert_channels_short_interleaved(int buf_c, short *buffer, int data_c, float **data, int d_offset, int len) +{ + int i; + check_endianness(); + if (buf_c != data_c && buf_c <= 2 && data_c <= 6) { + assert(buf_c == 2); + for (i=0; i < buf_c; ++i) + compute_stereo_samples(buffer, data_c, data, d_offset, len); + } else { + int limit = buf_c < data_c ? buf_c : data_c; + int j; + for (j=0; j < len; ++j) { + for (i=0; i < limit; ++i) { + FASTDEF(temp); + float f = data[i][d_offset+j]; + int v = FAST_SCALED_FLOAT_TO_INT(temp, f,15);//data[i][d_offset+j],15); + if ((unsigned int) (v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + *buffer++ = v; + } + for ( ; i < buf_c; ++i) + *buffer++ = 0; + } + } +} + +int stb_vorbis_get_frame_short_interleaved(stb_vorbis *f, int num_c, short *buffer, int num_shorts) +{ + float **output; + int len; + if (num_c == 1) return stb_vorbis_get_frame_short(f,num_c,&buffer, num_shorts); + len = stb_vorbis_get_frame_float(f, NULL, &output); + if (len) { + if (len*num_c > num_shorts) len = num_shorts / num_c; + convert_channels_short_interleaved(num_c, buffer, f->channels, output, 0, len); + } + return len; +} + +int stb_vorbis_get_samples_short_interleaved(stb_vorbis *f, int channels, short *buffer, int num_shorts) +{ + float **outputs; + int len = num_shorts / channels; + int n=0; + while (n < len) { + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n+k >= len) k = len - n; + if (k) + convert_channels_short_interleaved(channels, buffer, f->channels, f->channel_buffers, f->channel_buffer_start, k); + buffer += k*channels; + n += k; + f->channel_buffer_start += k; + if (n == len) break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break; + } + return n; +} + +int stb_vorbis_get_samples_short(stb_vorbis *f, int channels, short **buffer, int len) +{ + float **outputs; + int n=0; + while (n < len) { + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n+k >= len) k = len - n; + if (k) + convert_samples_short(channels, buffer, n, f->channels, f->channel_buffers, f->channel_buffer_start, k); + n += k; + f->channel_buffer_start += k; + if (n == len) break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break; + } + return n; +} + +#ifndef STB_VORBIS_NO_STDIO +int stb_vorbis_decode_filename(const char *filename, int *channels, int *sample_rate, short **output) +{ + int data_len, offset, total, limit, error; + short *data; + stb_vorbis *v = stb_vorbis_open_filename(filename, &error, NULL); + if (v == NULL) return -1; + limit = v->channels * 4096; + *channels = v->channels; + if (sample_rate) + *sample_rate = v->sample_rate; + offset = data_len = 0; + total = limit; + data = (short *) malloc(total * sizeof(*data)); + if (data == NULL) { + stb_vorbis_close(v); + return -2; + } + for (;;) { + int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data+offset, total-offset); + if (n == 0) break; + data_len += n; + offset += n * v->channels; + if (offset + limit > total) { + short *data2; + total *= 2; + data2 = (short *) realloc(data, total * sizeof(*data)); + if (data2 == NULL) { + free(data); + stb_vorbis_close(v); + return -2; + } + data = data2; + } + } + *output = data; + stb_vorbis_close(v); + return data_len; +} +#endif // NO_STDIO + +int stb_vorbis_decode_memory(const uint8 *mem, int len, int *channels, int *sample_rate, short **output) +{ + int data_len, offset, total, limit, error; + short *data; + stb_vorbis *v = stb_vorbis_open_memory(mem, len, &error, NULL); + if (v == NULL) return -1; + limit = v->channels * 4096; + *channels = v->channels; + if (sample_rate) + *sample_rate = v->sample_rate; + offset = data_len = 0; + total = limit; + data = (short *) malloc(total * sizeof(*data)); + if (data == NULL) { + stb_vorbis_close(v); + return -2; + } + for (;;) { + int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data+offset, total-offset); + if (n == 0) break; + data_len += n; + offset += n * v->channels; + if (offset + limit > total) { + short *data2; + total *= 2; + data2 = (short *) realloc(data, total * sizeof(*data)); + if (data2 == NULL) { + free(data); + stb_vorbis_close(v); + return -2; + } + data = data2; + } + } + *output = data; + stb_vorbis_close(v); + return data_len; +} +#endif // STB_VORBIS_NO_INTEGER_CONVERSION + +int stb_vorbis_get_samples_float_interleaved(stb_vorbis *f, int channels, float *buffer, int num_floats) +{ + float **outputs; + int len = num_floats / channels; + int n=0; + int z = f->channels; + if (z > channels) z = channels; + while (n < len) { + int i,j; + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n+k >= len) k = len - n; + for (j=0; j < k; ++j) { + for (i=0; i < z; ++i) + *buffer++ = f->channel_buffers[i][f->channel_buffer_start+j]; + for ( ; i < channels; ++i) + *buffer++ = 0; + } + n += k; + f->channel_buffer_start += k; + if (n == len) + break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) + break; + } + return n; +} + +int stb_vorbis_get_samples_float(stb_vorbis *f, int channels, float **buffer, int num_samples) +{ + float **outputs; + int n=0; + int z = f->channels; + if (z > channels) z = channels; + while (n < num_samples) { + int i; + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n+k >= num_samples) k = num_samples - n; + if (k) { + for (i=0; i < z; ++i) + memcpy(buffer[i]+n, f->channel_buffers[i]+f->channel_buffer_start, sizeof(float)*k); + for ( ; i < channels; ++i) + memset(buffer[i]+n, 0, sizeof(float) * k); + } + n += k; + f->channel_buffer_start += k; + if (n == num_samples) + break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) + break; + } + return n; +} +#endif // STB_VORBIS_NO_PULLDATA_API + +/* Version history + 1.17 - 2019-07-08 - fix CVE-2019-13217, -13218, -13219, -13220, -13221, -13222, -13223 + found with Mayhem by ForAllSecure + 1.16 - 2019-03-04 - fix warnings + 1.15 - 2019-02-07 - explicit failure if Ogg Skeleton data is found + 1.14 - 2018-02-11 - delete bogus dealloca usage + 1.13 - 2018-01-29 - fix truncation of last frame (hopefully) + 1.12 - 2017-11-21 - limit residue begin/end to blocksize/2 to avoid large temp allocs in bad/corrupt files + 1.11 - 2017-07-23 - fix MinGW compilation + 1.10 - 2017-03-03 - more robust seeking; fix negative ilog(); clear error in open_memory + 1.09 - 2016-04-04 - back out 'avoid discarding last frame' fix from previous version + 1.08 - 2016-04-02 - fixed multiple warnings; fix setup memory leaks; + avoid discarding last frame of audio data + 1.07 - 2015-01-16 - fixed some warnings, fix mingw, const-correct API + some more crash fixes when out of memory or with corrupt files + 1.06 - 2015-08-31 - full, correct support for seeking API (Dougall Johnson) + some crash fixes when out of memory or with corrupt files + 1.05 - 2015-04-19 - don't define __forceinline if it's redundant + 1.04 - 2014-08-27 - fix missing const-correct case in API + 1.03 - 2014-08-07 - Warning fixes + 1.02 - 2014-07-09 - Declare qsort compare function _cdecl on windows + 1.01 - 2014-06-18 - fix stb_vorbis_get_samples_float + 1.0 - 2014-05-26 - fix memory leaks; fix warnings; fix bugs in multichannel + (API change) report sample rate for decode-full-file funcs + 0.99996 - bracket #include for macintosh compilation by Laurent Gomila + 0.99995 - use union instead of pointer-cast for fast-float-to-int to avoid alias-optimization problem + 0.99994 - change fast-float-to-int to work in single-precision FPU mode, remove endian-dependence + 0.99993 - remove assert that fired on legal files with empty tables + 0.99992 - rewind-to-start + 0.99991 - bugfix to stb_vorbis_get_samples_short by Bernhard Wodo + 0.9999 - (should have been 0.99990) fix no-CRT support, compiling as C++ + 0.9998 - add a full-decode function with a memory source + 0.9997 - fix a bug in the read-from-FILE case in 0.9996 addition + 0.9996 - query length of vorbis stream in samples/seconds + 0.9995 - bugfix to another optimization that only happened in certain files + 0.9994 - bugfix to one of the optimizations that caused significant (but inaudible?) errors + 0.9993 - performance improvements; runs in 99% to 104% of time of reference implementation + 0.9992 - performance improvement of IMDCT; now performs close to reference implementation + 0.9991 - performance improvement of IMDCT + 0.999 - (should have been 0.9990) performance improvement of IMDCT + 0.998 - no-CRT support from Casey Muratori + 0.997 - bugfixes for bugs found by Terje Mathisen + 0.996 - bugfix: fast-huffman decode initialized incorrectly for sparse codebooks; fixing gives 10% speedup - found by Terje Mathisen + 0.995 - bugfix: fix to 'effective' overrun detection - found by Terje Mathisen + 0.994 - bugfix: garbage decode on final VQ symbol of a non-multiple - found by Terje Mathisen + 0.993 - bugfix: pushdata API required 1 extra byte for empty page (failed to consume final page if empty) - found by Terje Mathisen + 0.992 - fixes for MinGW warning + 0.991 - turn fast-float-conversion on by default + 0.990 - fix push-mode seek recovery if you seek into the headers + 0.98b - fix to bad release of 0.98 + 0.98 - fix push-mode seek recovery; robustify float-to-int and support non-fast mode + 0.97 - builds under c++ (typecasting, don't use 'class' keyword) + 0.96 - somehow MY 0.95 was right, but the web one was wrong, so here's my 0.95 rereleased as 0.96, fixes a typo in the clamping code + 0.95 - clamping code for 16-bit functions + 0.94 - not publically released + 0.93 - fixed all-zero-floor case (was decoding garbage) + 0.92 - fixed a memory leak + 0.91 - conditional compiles to omit parts of the API and the infrastructure to support them: STB_VORBIS_NO_PULLDATA_API, STB_VORBIS_NO_PUSHDATA_API, STB_VORBIS_NO_STDIO, STB_VORBIS_NO_INTEGER_CONVERSION + 0.90 - first public release +*/ + +#endif // STB_VORBIS_HEADER_ONLY + + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/platform/stdio_impl.c b/platform/stdio_impl.c new file mode 100644 index 0000000..0260f0a --- /dev/null +++ b/platform/stdio_impl.c @@ -0,0 +1,48 @@ +#include +#include "platform.h" + +// TODO: Should handle _FILE_OFFSET_BITS for seeks + +ssize_t platform_os_file_read(union platform_os_file fp, void *buf, size_t size) +{ + if (!fp._i) return -1; + + size_t n = fread(buf, 1, size, (FILE *)fp._i); + if (n == 0 && ferror((FILE *)fp._i)) { + clearerr((FILE *)fp._i); + return -1; + } + return (ssize_t)n; +} +ssize_t platform_os_file_write(union platform_os_file fp, const void *buf, size_t size) +{ + if (!fp._i) return -1; + + size_t n = fwrite(buf, 1, size, (FILE *)fp._i); + if (n == 0 && ferror((FILE *)fp._i)) { + clearerr((FILE *)fp._i); + return -1; + } + return (ssize_t)n; +} +int64_t platform_os_file_seek(union platform_os_file fp, int64_t delta, int from) +{ + if (!fp._i) return -1; + + int whence = -1; + switch (from) { + case PLATFORM_OS_SEEK_SET: whence = SEEK_SET; break; + case PLATFORM_OS_SEEK_CUR: whence = SEEK_CUR; break; + case PLATFORM_OS_SEEK_END: whence = SEEK_END; break; + } + int64_t pos = fseek((FILE *)fp._i, delta, whence); + if (pos != 0) + pos = -1; + else + pos = ftell((FILE *)fp._i); + return pos; +} +void platform_os_file_close(union platform_os_file fp) +{ + fclose((FILE *)fp._i); +} diff --git a/platform/windows_impl.c b/platform/windows_impl.c new file mode 100644 index 0000000..61aece1 --- /dev/null +++ b/platform/windows_impl.c @@ -0,0 +1,50 @@ +#include "platform.h" +#include "compiler.h" +#include + +static BOOL platform_run_wrapper(PINIT_ONCE InitOnce, PVOID Parameter, PVOID *Context) +{ + (void)InitOnce, (void)Context; + ((void(*)(void))Parameter)(); + return TRUE; +} + +void platform_os_once_run(union platform_os_once *once, void (*run_fn)(void)) +{ + InitOnceExecuteOnce(&once->_i, platform_run_wrapper, run_fn, (LPVOID *)0); +} + +size_t platform_os_getpagesize(void) +{ + SYSTEM_INFO si; + GetSystemInfo(&si); + return si.dwPageSize; +} + +struct platform_os_timespec platform_os_time(int type) +{ + struct platform_os_timespec ts = {}; + ts.err = -1; + switch (type) { + // TODO: _time64 is simple but extremely inaccurate! + // Use a more accurate counter. + case PLATFORM_OS_TIME_WALL: + ts.sec = _time64((__time64_t *)0); + ts.nsec = 0; + ts.err = 0; + break; + // TODO: GetTicks is simple but not accurate + // Use a more accurate counter + case PLATFORM_OS_TIME_MONO: + ts.sec = (int64_t)GetTickCount64(); + ts.nsec = (ts.sec % 1000) * 1000000; + ts.sec /= 1000; + ts.err = 0; + break; + case PLATFORM_OS_TIME_TICK__UNUSED: + FALLTHROUGH; + default: + break; + } + return ts; +} diff --git a/scripts/android/build-arch.sh b/scripts/android/build-arch.sh new file mode 100755 index 0000000..2cf1669 --- /dev/null +++ b/scripts/android/build-arch.sh @@ -0,0 +1,72 @@ +#!/bin/sh +. scripts/android/toolchain.sh + +ABI="" +case "$1" in + aarch64) JNI_BN="arm64-v8a" ;; + armv7a) JNI_BN="armeabi-v7a" ; ABI="eabi" ;; + i686) JNI_BN="x86" ;; + x86_64) JNI_BN="x86_64" ;; + *) echo "Usage: $0 {aarch64,armv7a,i686,x86_64} [debug|release]" ; exit 1 ;; +esac +ARCH="$1" + +MODE="$2" +if [ "$MODE" = "" ] ; then + MODE=debug +fi +case "$MODE" in + debug) MODE_FLAGS="-g -Og" ;; + release) MODE_FLAGS="-O2 -flto" ;; + *) echo "$0: Specify 'release' or 'debug' as the mode, not '$MODE'" ; exit 2 ;; +esac + +# Compiler variables +CC="$ANDROID_TOOLCHAIN/bin/clang -target ${ARCH}-linux-android${ABI}${SDK_VERSION}" +CC_SYSROOT="$ANDROID_TOOLCHAIN"/sysroot +CFLAGS="$MODE_FLAGS --sysroot=$CC_SYSROOT -iquote include -fPIC -Wall -Wextra -Werror -Wimplicit-fallthrough -DOSLOG_TAG_DEFAULT=\"${APP_NAME}\"" +LDFLAGS="-lm -lc -llog -landroid -ldl -lEGL -lGLESv3 -laaudio" + +# JNI directory and path where our app goes +JNI_DIR=lib/"$JNI_BN" +OBJ_DIR="$BUILD_DIR"/obj/"$ARCH"/"$MODE" +APP_LIB="lib${APP_NAME}.so" + +# Execute commands and error out on the first bad command +set -e +set -x + +# Android assets +mkdir -p "$OBJ_DIR" "$BUILD_DIR/$JNI_DIR" + +# We want to only build a source if it's older than the obj +# Once you see miniaudio go brrrrrrrrrrrrr during compile, +# you'll see why I bothered with this... +compile_obj () { + if [ ! "$1" -ot "$OBJ_DIR/${1##*/}.o" ] ; then $CC $3 -c "$1" -o "$OBJ_DIR/${1##*/}.o" $CFLAGS $2 ; fi +} + +# Certain pieces of code that we trust, like 3rd party libs, can return warnings but shouldn't error out +# When you compile header files (like miniaudio.h), don't forget -Wno-unused-function and -xc or you'll have a bad time +# Also, I'm so sorry about stb... Trust me, all that is important. +compile_obj "$NATIVE_APP_GLUE_DIR"/android_native_app_glue.c "-Wno-error -Wno-implicit-fallthrough -I $NATIVE_APP_GLUE_DIR" +compile_obj platform/stb_vorbis.c "-Wno-error -Wno-implicit-fallthrough" +compile_obj platform/miniaudio.h "-DMINIAUDIO_IMPLEMENTATION -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" +compile_obj platform/sokol_gfx.h "-DSOKOL_IMPL -DSOKOL_GLES3 -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" +compile_obj platform/stb_image.h "-DSTB_IMAGE_IMPLEMENTATION -DSTBI_MALLOC=os_malloc_emulate_stb -DSTBI_REALLOC=os_realloc_emulate_stb -DSTBI_FREE=os_free_emulate_stb -include platform/stb_impl.h -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" +compile_obj platform/stb_truetype.h "-DSTB_TRUETYPE_IMPLEMENTATION -DSTBTT_malloc=os_malloc_emulate_stb_a -DSTBTT_free=os_free_emulate_stb_a -include platform/stb_impl.h -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" +compile_obj platform/stb_rect_pack.h "-DSTB_RECT_PACK_IMPLEMENTATION -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" + +# Compile our platform code that's not third party +# android.c is always rebuilt because WAIT_FOR_DEBUGGER +# emits extra code depending on if it's defined or not +touch platform/android.c +compile_obj platform/android.c "-I $NATIVE_APP_GLUE_DIR -D WAIT_FOR_DEBUGGER=0$WAIT" +compile_obj platform/sokol_impl.c +compile_obj platform/stb_impl.c +compile_obj platform/posix_impl.c +compile_obj platform/stdio_impl.c +for SRC in core/*.c game/*.c ; do + compile_obj "$SRC" +done +$CC -shared "$OBJ_DIR"/*.o -o "$BUILD_DIR/$JNI_DIR/$APP_LIB" $CFLAGS $LDFLAGS diff --git a/scripts/android/build.sh b/scripts/android/build.sh new file mode 100755 index 0000000..b399d96 --- /dev/null +++ b/scripts/android/build.sh @@ -0,0 +1,24 @@ +#!/bin/sh + +. scripts/android/toolchain.sh +set -e +set -x + +rm -f "$BUILD_DIR"/lib/*/lib*.so + +# NOTE: +# I primarily debug on a physical test phone which is an AARCH64 device +# If you test on an emulator, change the default arch to be x86_64 +MODE="$1" +if [ "$MODE" = "" ] ; then + MODE=debug +fi +case "$MODE" in + debug) ARCHS=aarch64 ;; + release) ARCHS="aarch64 armv7a i686 x86_64" ;; + *) echo "$0: Specify 'debug' or 'release' for mode, not '$MODE'" ; exit 2 ;; +esac +for ARCH in $ARCHS ; do + scripts/android/build-arch.sh $ARCH $MODE +done +scripts/android/package.sh diff --git a/scripts/android/debug.sh b/scripts/android/debug.sh new file mode 100755 index 0000000..1dc95a3 --- /dev/null +++ b/scripts/android/debug.sh @@ -0,0 +1,31 @@ +#!/bin/sh +. scripts/android/toolchain.sh + +# It's very important that paths be taken literally at this point +# Magic from MSYS was appreciated before but not now... +export MSYS_NO_PATHCONV=1 +ADB="$ANDROID_HOME/platform-tools/adb" +"$ADB" push "$ANDROID_TOOLCHAIN/lib/clang/20/lib/linux/aarch64/lldb-server" /data/local/tmp/lldb-server +"$ADB" shell "monkey -p $APP_PKG -c android.intent.category.LAUNCHER 1" +PROC_PID="$("$ADB" shell "ps -A | grep $APP_PKG" | awk '{print $2}' | head -1)" +if [ "$PROC_PID" = "" ] ; then + echo "Could not get PID of $APP_PKG even though it was launched? Strange..." + exit 1 +fi +DATA_DIR="$("$ADB" shell "run-as $APP_PKG pwd")" +"$ADB" forward tcp:4444 localfilesystem:"$DATA_DIR"/lldb-server.socket +"$ADB" shell "run-as $APP_PKG sh -c 'rm -f lldb-server.socket ; cp /data/local/tmp/lldb-server . ; chmod +x . ; chmod 700 lldb-server ; ./lldb-server g \"unix://\$PWD/lldb-server.socket\" --attach $PROC_PID'" & + +# Normally not a good idea to do this for SIGBUSes and SIGSEGVs, +# but a lot of those get thrown around! Debugging is basically unusable +# if we stop at every one of them. I hate Android.... +"$ANDROID_TOOLCHAIN/bin/lldb$EXTCMD" \ + -O 'proc hand -n true -p true -s false SIGSEGV SIGBUS' \ + -O 'gdb-remote localhost:4444' \ + +# These settings were found in ndk-gdb, but they make a lot of SIGSTOPS +# that make debugging also unusable, so I do away with these settings. +# You can use them in an actual crash, but probably not beforehand. + #-O "target create '$BUILD_DIR/lib/arm64-v8a/libhbatbp.so'" \ + #-O "target modules search-paths add / $SYSROOT_DIR/" \ + #-O "settings append target.exec-search-paths $SYSROOT_DIR/system/bin $SYSROOT_DIR/system/lib64" \ diff --git a/scripts/android/install.sh b/scripts/android/install.sh new file mode 100755 index 0000000..1909ffe --- /dev/null +++ b/scripts/android/install.sh @@ -0,0 +1,4 @@ +#!/bin/sh +. scripts/android/toolchain.sh + +"$ANDROID_HOME"/platform-tools/adb install build/android/app-signed.apk diff --git a/scripts/android/log.sh b/scripts/android/log.sh new file mode 100755 index 0000000..1007805 --- /dev/null +++ b/scripts/android/log.sh @@ -0,0 +1,4 @@ +#!/bin/sh +. scripts/android/toolchain.sh + +"$ADB" logcat "*:F AndroidRuntime:E ${APP_NAME}" diff --git a/scripts/android/package.sh b/scripts/android/package.sh new file mode 100755 index 0000000..d125b40 --- /dev/null +++ b/scripts/android/package.sh @@ -0,0 +1,54 @@ +#!/bin/sh +. scripts/android/toolchain.sh + +# Keystore variables +KEYSTORE="$BUILD_DIR/$APP_NAME.keystore" +KEYSTORE_PASS="GenericPassword" + +# Report progress on script +set -x + +# Create the keystore if it doesn't exist yet +if [ ! -f "$KEYSTORE" ] ; then + "$JAVA_HOME"/bin/keytool -genkeypair -validity 10000 -dname "CN=hbat,O=Android,C=ES" -keystore "$KEYSTORE" -storepass "$KEYSTORE_PASS" -keypass "$KEYSTORE_PASS" -alias hbatKey -keyalg RSA +fi + +mkdir -p "$BUILD_DIR"/res/drawable-ldpi && cp "$LOGO_LDPI_DIR/$LOGO_NAME$LOGO_EXT" "$BUILD_DIR"/res/drawable-ldpi/"$LOGO_NAME$LOGO_EXT" +mkdir -p "$BUILD_DIR"/res/drawable-mdpi && cp "$LOGO_MDPI_DIR/$LOGO_NAME$LOGO_EXT" "$BUILD_DIR"/res/drawable-mdpi/"$LOGO_NAME$LOGO_EXT" +mkdir -p "$BUILD_DIR"/res/drawable-hdpi && cp "$LOGO_HDPI_DIR/$LOGO_NAME$LOGO_EXT" "$BUILD_DIR"/res/drawable-hdpi/"$LOGO_NAME$LOGO_EXT" +mkdir -p "$BUILD_DIR"/res/values + + +# Create Android Manifest and text resources +cat >"$BUILD_DIR"/res/values/strings.xml < +${APP_TITLE} +STRINGS +cat >"$BUILD_DIR"/AndroidManifest.xml < + + + + + + + + + + + + + +MANIFEST + +# Package it using aapt and apksigner +"$BUILD_TOOLS"/aapt package -f -M "$BUILD_DIR"/AndroidManifest.xml -S "$BUILD_DIR"/res -A assets -I "$ANDROID_HOME"/platforms/android-"$SDK_VERSION"/android.jar -F "$APK_UNSIGNED" +(cd "$BUILD_DIR" && "$BUILD_TOOLS"/aapt add app.apk lib/*/lib${APP_NAME}.so) +"$BUILD_TOOLS"/zipalign -f 4 "$APK_UNSIGNED" "$APK_SIGNED" +"$BUILD_TOOLS"/apksigner"$EXTBAT" sign --ks "$KEYSTORE" --ks-pass pass:"$KEYSTORE_PASS" "$APK_SIGNED" diff --git a/scripts/android/sysroot.sh b/scripts/android/sysroot.sh new file mode 100755 index 0000000..400660d --- /dev/null +++ b/scripts/android/sysroot.sh @@ -0,0 +1,13 @@ +#!/bin/sh +. scripts/android/toolchain.sh + +# MSYS shouldn't convert paths here +export MSYS_NO_PATHCONV=1 +mkdir -p "$SYSROOT_DIR"/system/bin +for BIN in app_process64 linker64 ; do + "$ADB" pull /system/bin/"$BIN" "$SYSROOT_DIR"/system/bin/"$BIN" +done +mkdir -p "$SYSROOT_DIR"/system/lib64 +for LIB in libc.so libm.so libdl.so ; do + "$ADB" pull /system/lib64/"$LIB" "$SYSROOT_DIR"/system/lib64/"$LIB" +done diff --git a/scripts/android/toolchain.sh b/scripts/android/toolchain.sh new file mode 100755 index 0000000..9088ab2 --- /dev/null +++ b/scripts/android/toolchain.sh @@ -0,0 +1,77 @@ +#!/bin/sh + +# App name and organization +APP_NAME="misslehell" +APP_TITLE="Miss(le) Hell" +APP_ORG="org.misomosispi" +APP_PKG="${ORG}.${APP_NAME}" + +# Other Android Manifest variables +ORIENTATION="portrait" +MIN_SDK="29" +LOGO_NAME="explosion" +LOGO_EXT=".png" +LOGO_LDPI_DIR="logo/ldpi" +LOGO_MDPI_DIR="logo/mdpi" +LOGO_HDPI_DIR="logo/hdpi" + +# Android toolchain variables +# Use 'source' on this file and you'll have the variables + +# Check environment variables +if [ "$ANDROID_HOME" = "" ]; then + ANDROID_HOME="$ANDROID_SDK_HOME" +fi +if [ ! -d "$ANDROID_HOME" ]; then + echo "ANDROID_HOME environment variable is not set to a directory!" + echo "You can download Android's cmdline tools from https://developer.android.com/studio#command-line-tools-only" + echo "Then install them in \$ANDROID_HOME/cmdline-tools/latest" + exit 63 +fi +if [ ! -d "$JAVA_HOME" ] ; then + echo "JAVA_HOME environment variable is not set to a directory!" + echo "You can download OpenJDK 24 from https://jdk.java.net/24" + echo "Then install it in \$JAVA_HOME" + exit 62 +fi + +# Get latest build tools, NDK, SDK +# The engine should keep working for the latest and greatest NDK +BUILD_TOOLS_VERSION="36.0.0" +NDK_VERSION="29.0.13113456" +SDK_VERSION="35" + +# Tools directories which are checked for existence +# If they don't exist, instruct the user to install them and exit +BUILD_TOOLS="$ANDROID_HOME"/build-tools/"$BUILD_TOOLS_VERSION" +PLATFORM_TOOLS="$ANDROID_HOME"/platform-tools +PLATFORM_DIR="$ANDROID_HOME"/platforms/android-"$SDK_VERSION" +ANDROID_NDK="$ANDROID_HOME"/ndk/"$NDK_VERSION" +if [ ! -d "$BUILD_TOOLS" -o ! -d "$PLATFORM_TOOLS" -o ! -d "$PLATFORM_DIR" -o ! -d "$ANDROID_NDK" ] ; then + echo "You do not have the requested SDK/NDK/tool versions for building this project!" + echo "Either change the versions in this script (scripts/android/toolchain.sh) or run this sdkmanager command..." + echo "sdkmanager 'platform-tools' 'platforms;android-$SDK_VERSION' 'ndk;$NDK_VERSION' 'build-tools;$BUILD_TOOLS_VERSION'" + exit 61 +fi + +# Android NDK Toolchain directories +ANDROID_TOOLCHAIN_BASE="$ANDROID_NDK"/toolchains/llvm/prebuilt +ANDROID_TOOLCHAIN="$ANDROID_TOOLCHAIN_BASE"/linux-x86_64 +EXTBAT="" +EXTCMD="" +if [ ! -d "$ANDROID_TOOLCHAIN" ] ; then + ANDROID_TOOLCHAIN="$ANDROID_TOOLCHAIN_BASE"/windows-x86_64 + EXTBAT=".bat" + EXTCMD=".cmd" +fi +NDK_SOURCES="$ANDROID_NDK"/sources/android +NATIVE_APP_GLUE_DIR="$NDK_SOURCES"/native_app_glue + +# Build,sysroot directory and final signed APK +BUILD_DIR=build/android +SYSROOT_DIR=sysroot/android +APK_UNSIGNED="$BUILD_DIR/app.apk" +APK_SIGNED="$BUILD_DIR/app-signed.apk" + +# Convenience variables +ADB="$ANDROID_HOME/platform-tools/adb" diff --git a/scripts/linux/build.sh b/scripts/linux/build.sh new file mode 100755 index 0000000..cbca0b2 --- /dev/null +++ b/scripts/linux/build.sh @@ -0,0 +1,66 @@ +#!/bin/sh +. scripts/linux/toolchain.sh + +MODE="$1" +if [ "$MODE" = "" ] ; then + MODE=debug +fi +case "$MODE" in + debug) MODE_FLAGS="-g -Og" ;; + release) MODE_FLAGS="-O2" ;; + *) echo "$0: Specify 'release' or 'debug' as the mode, not '$MODE'" ; exit 2 ;; +esac + +# Default to using gcc and its build flags +CC="clang" +CFLAGS="$MODE_FLAGS $(pkg-config --cflags sdl3 gl) -iquote include -Wall -Wextra -Werror -Wimplicit-fallthrough" +LDFLAGS="$(pkg-config --libs sdl3 gl) -lm" + +OBJ_DIR="$BUILD_DIR/obj/$MODE" +compile_obj () { + if [ ! "$1" -ot "$OBJ_DIR/${1##*/}.o" ] ; then "$CC" $3 -c "$1" -o "$OBJ_DIR/${1##*/}.o" $CFLAGS $2 ; fi +} + +APP_UNSIGNED_DIR="$PWD/${APP_UNSIGNED%/*}" +SDL3_UNSIGNED="$APP_UNSIGNED_DIR/SDL3.dll" +package_asset () { + if [ ! "$1" -ot "$APP_UNSIGNED_DIR/$1" ] ; then cp "$1" "$APP_UNSIGNED_DIR/$1" ; fi +} + +set -e +set -x + +# Create the directory for object files +mkdir -p "$OBJ_DIR" + +# Compile third-party code +compile_obj platform/stb_vorbis.c "-Wno-error -Wno-implicit-fallthrough" +compile_obj platform/miniaudio.h "-DMINIAUDIO_IMPLEMENTATION -Wno-implicit-fallthrough -Wno-error -Wno-unused-function" "-xc" +compile_obj platform/sokol_gfx.h "-DSOKOL_IMPL -DSOKOL_GLCORE -Wno-implicit-fallthrough -Wno-error -Wno-unused-function" "-xc" +compile_obj platform/stb_image.h "-DSTB_IMAGE_IMPLEMENTATION -DSTBI_MALLOC=os_malloc_emulate_stb -DSTBI_REALLOC=os_realloc_emulate_stb -DSTBI_FREE=os_free_emulate_stb -include platform/stb_impl.h -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" +compile_obj platform/stb_truetype.h "-DSTB_TRUETYPE_IMPLEMENTATION -DSTBTT_malloc=os_malloc_emulate_stb_a -DSTBTT_free=os_free_emulate_stb_a -include platform/stb_impl.h -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" +compile_obj platform/stb_rect_pack.h "-DSTB_RECT_PACK_IMPLEMENTATION -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" + +#TODO: Implement platform functions and then impl files can be compiled +compile_obj platform/sokol_impl.c +compile_obj platform/stb_impl.c +compile_obj platform/posix_impl.c +compile_obj platform/stdio_impl.c + +# Compile core and game code +for SRC in core/*.c game/*.c ; do + compile_obj "$SRC" +done + +# We use SDL3 as a generic backend for various builds like Linux +compile_obj platform/sdl3.c + +"$CC" "$OBJ_DIR"/*.o -o "$APP_UNSIGNED" $CFLAGS $LDFLAGS + +mkdir -p "$APP_UNSIGNED_DIR/binary" +mkdir -p "$APP_UNSIGNED_DIR/text" +cd assets/ +for ASSET in text/* binary/* ; do + package_asset "$ASSET" +done +cd "$OLDPWD" diff --git a/scripts/linux/toolchain.sh b/scripts/linux/toolchain.sh new file mode 100755 index 0000000..a0f7f6c --- /dev/null +++ b/scripts/linux/toolchain.sh @@ -0,0 +1,13 @@ +#!/bin/sh + +if ! pkg-config --cflags --libs sdl3 gl >/dev/null ; then + echo "SDL3 or OpenGL not found by pkg-config" + echo "You can install on Debian by running this command" + echo "sudo apt install libsdl3-dev libgl-dev" + exit 63 +fi + +# Build directory and final SDL lib,app location +# We don't do signing on Linux (yet?) +BUILD_DIR=build/linux +APP_UNSIGNED="$BUILD_DIR/app" diff --git a/scripts/windows/build.sh b/scripts/windows/build.sh new file mode 100644 index 0000000..a7e0806 --- /dev/null +++ b/scripts/windows/build.sh @@ -0,0 +1,69 @@ +#!/bin/sh +. scripts/windows/toolchain.sh + +MODE="$1" +if [ "$MODE" = "" ] ; then + MODE=debug +fi +case "$MODE" in + debug) MODE_FLAGS="-g -Og" ;; + release) MODE_FLAGS="-O2" ;; + *) echo "$0: Specify 'release' or 'debug' as the mode, not '$MODE'" ; exit 2 ;; +esac + +# Default to using clang by MSVC and its build flags +# Also disable CRT warnings on Windows because we compile +# on POSIX and want to keep APIs and errors consistent +CC="$MSVC_CLANG" +CFLAGS="$MODE_FLAGS -iquote include -I $SDL3_HOME/include -Wall -Wextra -Werror -Wimplicit-fallthrough -D_CRT_SECURE_NO_WARNINGS" +LDFLAGS="-L $SDL3_HOME/lib/x64 -lsdl3" + +OBJ_DIR="$BUILD_DIR/obj/$MODE" +compile_obj () { + if [ ! "$1" -ot "$OBJ_DIR/${1##*/}.o" ] ; then "$CC" $3 -c "$1" -o "$OBJ_DIR/${1##*/}.o" $CFLAGS $2 ; fi +} + +APP_UNSIGNED_DIR="$PWD/${APP_UNSIGNED%/*}" +SDL3_UNSIGNED="$APP_UNSIGNED_DIR/SDL3.dll" +package_asset () { + if [ ! "$1" -ot "$APP_UNSIGNED_DIR/$1" ] ; then cp "$1" "$APP_UNSIGNED_DIR/$1" ; fi +} + +set -e +set -x + +# Create the directory for object files +mkdir -p "$OBJ_DIR" + +# Compile third-party code +compile_obj platform/stb_vorbis.c "-Wno-error -Wno-implicit-fallthrough" +compile_obj platform/miniaudio.h "-DMINIAUDIO_IMPLEMENTATION -Wno-implicit-fallthrough -Wno-error -Wno-unused-function" "-xc" +compile_obj platform/sokol_gfx.h "-DSOKOL_IMPL -DSOKOL_GLCORE -Wno-implicit-fallthrough -Wno-error -Wno-unused-function" "-xc" +compile_obj platform/stb_image.h "-DSTB_IMAGE_IMPLEMENTATION -DSTBI_MALLOC=os_malloc_emulate_stb -DSTBI_REALLOC=os_realloc_emulate_stb -DSTBI_FREE=os_free_emulate_stb -include platform/stb_impl.h -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" +compile_obj platform/stb_truetype.h "-DSTB_TRUETYPE_IMPLEMENTATION -DSTBTT_malloc=os_malloc_emulate_stb_a -DSTBTT_free=os_free_emulate_stb_a -include platform/stb_impl.h -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" +compile_obj platform/stb_rect_pack.h "-DSTB_RECT_PACK_IMPLEMENTATION -Wno-error -Wno-implicit-fallthrough -Wno-unused-function" "-xc" + +#TODO: Implement platform functions and then impl files can be compiled +compile_obj platform/sokol_impl.c +compile_obj platform/stb_impl.c +compile_obj platform/windows_impl.c +compile_obj platform/stdio_impl.c + +# Compile core and game code +for SRC in core/*.c game/*.c ; do + compile_obj "$SRC" +done + +# We use SDL3 as a generic backend for various builds like Windows +compile_obj platform/sdl3.c + +"$CC" "$OBJ_DIR"/*.o -o "$APP_UNSIGNED" $CFLAGS $LDFLAGS +cp "$SDL3_HOME/lib/x64/SDL3.dll" "$SDL3_UNSIGNED" + +mkdir -p "$APP_UNSIGNED_DIR/binary" +mkdir -p "$APP_UNSIGNED_DIR/text" +cd assets/ +for ASSET in text/* binary/* ; do + package_asset "$ASSET" +done +cd "$OLDPWD" diff --git a/scripts/windows/toolchain.sh b/scripts/windows/toolchain.sh new file mode 100644 index 0000000..d50b859 --- /dev/null +++ b/scripts/windows/toolchain.sh @@ -0,0 +1,30 @@ +#!/bin/sh + +# Paths for MSVC / CLANG +#MSVC_VER=14.41.34120 +MSVC_TOOLS="/c/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools" +#MSVC_CL="$MSVC_TOOLS/MSVC/$MSVC_VER/bin/HostX64/x64" +MSVC_CLANG="$MSVC_TOOLS/Llvm/x64/bin/clang" + +# Build directory and final SDL lib,app location +# We don't do signing on Windows (yet?) +BUILD_DIR=build/windows +APP_UNSIGNED="$BUILD_DIR/app.exe" + +if [ ! -d "$SDL3_HOME" ] ; then + echo "SDL3_HOME is not set to a directory!" + echo "You can download SDL3 from https://github.com/libsdl-org/SDL/releases by choosing a 3.x.x release" + echo "Then install it in \$SDL3_HOME" + exit 63 +fi +if [ ! -d "$MSVC_TOOLS" ] ; then + echo "Path '$MSVC_TOOLS' is not a directory" + echo "Either change this path in this script (scripts/windows/toolchain.sh) or install MSVC + LLVM tools for Visual Studio 2022" + echo "You can download Visual Studio Community 2022 from https://visualstudio.microsoft.com/thank-you-downloading-visual-studio/?sku=Community&channel=Release&version=VS2022" + exit 62 +fi +if [ ! -f "$MSVC_CLANG" ] ; then + echo "Path '$MSVC_CLANG' is not a file" + echo "Either change this path in this script (scripts/windows/toolchain.sh) or install LLVM for Visual Studio 2022" + exit 61 +fi