mirror/miniaudio
1
0
Fork 0
mirror of https://github.com/mackron/miniaudio.git synced 2026-04-21 15:56:58 +02:00
Code Issues Packages Projects Releases Wiki Activity

Add resampling test.

Browse Source
This commit is contained in:
David Reid
2026-02-14 20:26:20 +10:00
parent d286a97ab1
commit dbf391611d
2 changed files with 557 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CMakeLists.txt
+66
View File
@@ -498,6 +498,51 @@ if(NOT MINIAUDIO_NO_LIBOPUS)
endif()
endif()
# libsamplerate
if(NOT TARGET samplerate)
# Try pkg-config first
find_package(PkgConfig QUIET)
if(PkgConfig_FOUND)
pkg_check_modules(samplerate IMPORTED_TARGET samplerate)
endif()
if(TARGET PkgConfig::samplerate)
add_library(samplerate ALIAS PkgConfig::samplerate)
else()
message(STATUS "libsamplerate not found.")
endif()
endif()
# speexdsp
if(NOT TARGET speexdsp)
# Try pkg-config first
find_package(PkgConfig QUIET)
if(PkgConfig_FOUND)
pkg_check_modules(speexdsp IMPORTED_TARGET speexdsp)
endif()
if(TARGET PkgConfig::speexdsp)
add_library(speexdsp ALIAS PkgConfig::speexdsp)
else()
message(STATUS "speexdsp not found.")
endif()
endif()
# soxr
if(NOT TARGET soxr)
# Try pkg-config first
find_package(PkgConfig QUIET)
if(PkgConfig_FOUND)
pkg_check_modules(soxr IMPORTED_TARGET soxr)
endif()
if(TARGET PkgConfig::soxr)
add_library(soxr ALIAS PkgConfig::soxr)
else()
message(STATUS "soxr not found.")
endif()
endif()
# SteamAudio has an annoying SDK setup. In the lib folder there is a folder for each platform. We need to specify the
# platform we're compiling for.
@@ -837,6 +882,27 @@ if(MINIAUDIO_BUILD_TESTS)
add_miniaudio_test(miniaudio_generation generation/generation.c)
add_test(NAME miniaudio_generation COMMAND miniaudio_generation)
add_miniaudio_test(miniaudio_resampling resampling/resampling.c)
#add_test(NAME miniaudio_resampling COMMAND miniaudio_resampling)
if(TARGET samplerate)
target_link_libraries(miniaudio_resampling PRIVATE samplerate)
target_compile_definitions(miniaudio_resampling PRIVATE MA_HAS_LIBSAMPLERATE)
else()
message(STATUS "libsamplerate not found. Sample rate conversion will be disabled in miniaudio_resampling.")
endif()
if(TARGET speexdsp)
target_link_libraries(miniaudio_resampling PRIVATE speexdsp)
target_compile_definitions(miniaudio_resampling PRIVATE MA_HAS_SPEEXDSP)
else()
message(STATUS "speexdsp not found. Speex resampler will be disabled in miniaudio_resampling.")
endif()
if(TARGET soxr)
target_link_libraries(miniaudio_resampling PRIVATE soxr)
target_compile_definitions(miniaudio_resampling PRIVATE MA_HAS_SOXR)
else()
message(STATUS "soxr not found. SoX resampler will be disabled in miniaudio_resampling.")
endif()
endif()
# Examples
tests/resampling/resampling.c
+491
View File
@@ -0,0 +1,491 @@
#define MA_NO_PIPEWIRE
#if 1
#include "../../miniaudio.c"
#else
#define MINIAUDIO_IMPLEMENTATION
#include "../../miniaudio-11.h"
#endif
#ifdef MA_HAS_LIBSAMPLERATE
#include <samplerate.h>
#endif
#ifdef MA_HAS_SPEEXDSP
#include <speex/speex_resampler.h>
#endif
#ifdef MA_HAS_SOXR
#include <soxr.h>
#endif
#define RESAMPLING_LPF_ORDER 0
#define RESAMPLING_FRAMES_PER_ITERATION 1024 * 1024
#define RESAMPLING_ITERATION_COUNT 100
#define resampling_api_miniaudio 0
#define resampling_api_libsamplerate 1
#define resampling_api_speex 2
#define resampling_api_soxr 4
typedef int resampling_api;
const char* resampling_api_to_string(resampling_api api)
{
if (api == resampling_api_miniaudio) {
return "miniaudio";
}
if (api == resampling_api_libsamplerate) {
return "libsamplerate";
}
if (api == resampling_api_speex) {
return "speex";
}
if (api == resampling_api_soxr) {
return "soxr";
}
MA_ASSERT(!"Resampling API needs to be added to resampling_api_to_string().");
return "unknown";
}
/* Modify this to control the resampling API. Comment out to force miniaudio. */
/*#define RESAMPLING_LISTENING_TEST_API resampling_api_libsamplerate*/
#if defined(RESAMPLING_LISTENING_TEST_API)
#if RESAMPLING_LISTENING_TEST_API == resampling_api_libsamplerate && !defined(MA_HAS_LIBSAMPLERATE)
#error "libsamplerate is unavailable and cannot be used."
#endif
#endif
#ifndef RESAMPLING_LISTENING_TEST_API
#define RESAMPLING_LISTENING_TEST_API resampling_api_miniaudio
#endif
typedef struct
{
ma_resampler resampler;
ma_waveform waveform;
ma_uint8 cache[4096];
ma_uint32 cachedFrameCount;
ma_uint32 cachedFrameCap;
float sampleRateStep;
float sampleRateRatio;
float minSampleRateRatio;
float maxSampleRateRatio;
#ifdef MA_HAS_LIBSAMPLERATE
SRC_STATE* pLSRState;
#endif
} resampler_data_callback_data;
static void resampler_data_callback(ma_device* pDevice, void* pFramesOut, const void* pFramesIn, ma_uint32 frameCount)
{
resampler_data_callback_data* pResamplerData = (resampler_data_callback_data*)pDevice->pUserData;
ma_uint32 framesWritten;
ma_uint32 bpf;
bpf = ma_get_bytes_per_frame(pResamplerData->resampler.format, pResamplerData->resampler.channels);
framesWritten = 0;
while (framesWritten < frameCount) {
/* We resample from the cache. */
while (pResamplerData->cachedFrameCount > 0 && framesWritten < frameCount) {
#if RESAMPLING_LISTENING_TEST_API == resampling_api_miniaudio
{
ma_uint64 frameCountIn;
ma_uint64 frameCountOut;
frameCountIn = pResamplerData->cachedFrameCount;
frameCountOut = frameCount - framesWritten;
ma_resampler_process_pcm_frames(&pResamplerData->resampler, pResamplerData->cache, &frameCountIn, ma_offset_ptr(pFramesOut, framesWritten * bpf), &frameCountOut);
MA_MOVE_MEMORY(pResamplerData->cache, ma_offset_ptr(pResamplerData->cache, frameCountIn * bpf), (pResamplerData->cachedFrameCount - frameCountIn) * bpf);
pResamplerData->cachedFrameCount -= frameCountIn;
framesWritten += frameCountOut;
}
#endif
#if RESAMPLING_LISTENING_TEST_API == resampling_api_libsamplerate
{
SRC_DATA data;
MA_ZERO_OBJECT(&data);
data.data_in = (const float*)pResamplerData->cache;
data.data_out = (float*)ma_offset_ptr(pFramesOut, framesWritten * bpf);
data.input_frames = pResamplerData->cachedFrameCount;
data.output_frames = frameCount - framesWritten;
data.src_ratio = 1 / pResamplerData->sampleRateRatio;
src_process(pResamplerData->pLSRState, &data);
MA_MOVE_MEMORY(pResamplerData->cache, ma_offset_ptr(pResamplerData->cache, data.input_frames_used * bpf), (pResamplerData->cachedFrameCount - data.input_frames_used) * bpf);
pResamplerData->cachedFrameCount -= data.input_frames_used;
framesWritten += data.output_frames_gen;
}
#endif
}
/* Reload the cache if necessary. */
if (pResamplerData->cachedFrameCount == 0) {
ma_waveform_read_pcm_frames(&pResamplerData->waveform, pResamplerData->cache, pResamplerData->cachedFrameCap, NULL);
pResamplerData->cachedFrameCount = pResamplerData->cachedFrameCap;
}
}
/* Sweep the sample rate. We'll just adjust the output rate. */
pResamplerData->sampleRateRatio += pResamplerData->sampleRateStep;
/* */ if (pResamplerData->sampleRateRatio > pResamplerData->maxSampleRateRatio) {
pResamplerData->sampleRateRatio = pResamplerData->maxSampleRateRatio;
pResamplerData->sampleRateStep = -pResamplerData->sampleRateStep;
} else if (pResamplerData->sampleRateRatio < pResamplerData->minSampleRateRatio) {
pResamplerData->sampleRateRatio = pResamplerData->minSampleRateRatio;
pResamplerData->sampleRateStep = -pResamplerData->sampleRateStep;
}
#if RESAMPLING_LISTENING_TEST_API == resampling_api_miniaudio
{
ma_resampler_set_rate_ratio(&pResamplerData->resampler, pResamplerData->sampleRateRatio);
}
#endif
#if RESAMPLING_LISTENING_TEST_API == resampling_api_libsamplerate
{
src_set_ratio(pResamplerData->pLSRState, 1 / pResamplerData->sampleRateRatio);
}
#endif
(void)pFramesIn;
}
void resampler_listening_test(void)
{
ma_result result;
resampler_data_callback_data callbackData;
ma_resampler_config resamplerConfig;
ma_waveform_config waveformConfig;
ma_device_config deviceConfig;
ma_device device;
ma_format format;
ma_uint32 channels = 2;
printf("Resampling Listening Test API: %s\n", resampling_api_to_string(RESAMPLING_LISTENING_TEST_API));
#if RESAMPLING_LISTENING_TEST_API == resampling_api_miniaudio
{
format = ma_format_f32; /* Switch this between f32 and s16 as required when testing miniaudio. */
}
#endif
#if RESAMPLING_LISTENING_TEST_API == resampling_api_libsamplerate
{
format = ma_format_f32; /* Can only use f32 with libsamplerate. */
callbackData.pLSRState = src_new(SRC_SINC_FASTEST /*SRC_SINC_BEST_QUALITY*/, channels, NULL);
}
#endif
resamplerConfig = ma_resampler_config_init(format, channels, 48000, 48000, ma_resample_algorithm_linear);
resamplerConfig.linear.lpfOrder = RESAMPLING_LPF_ORDER;
result = ma_resampler_init(&resamplerConfig, NULL, &callbackData.resampler);
if (result != MA_SUCCESS) {
printf("Failed to initialize resampler.");
return;
}
waveformConfig = ma_waveform_config_init(resamplerConfig.format, resamplerConfig.channels, resamplerConfig.sampleRateIn, ma_waveform_type_sine, 0.2f, 440);
result = ma_waveform_init(&waveformConfig, &callbackData.waveform);
if (result != MA_SUCCESS) {
printf("Failed to initialize waveform.");
return;
}
callbackData.cachedFrameCount = 0;
callbackData.cachedFrameCap = sizeof(callbackData.cache) / ma_get_bytes_per_frame(resamplerConfig.format, resamplerConfig.channels);
/* Default. */
#if 1
callbackData.sampleRateStep = 0.01f;
callbackData.minSampleRateRatio = 0.2f;
callbackData.maxSampleRateRatio = 5.0f;
#endif
/* Low frequency upsample. */
#if 0
callbackData.sampleRateStep = 0.001f;
callbackData.minSampleRateRatio = 0.01f;
callbackData.maxSampleRateRatio = 0.09f;
#endif
/* High frequency downsample */
#if 0
callbackData.sampleRateStep = 0.01f;
callbackData.minSampleRateRatio = 4.0f;
callbackData.maxSampleRateRatio = 6.0f;
#endif
#if 1
callbackData.sampleRateStep = 0.002f;
callbackData.minSampleRateRatio = 0.002f;
callbackData.maxSampleRateRatio = 0.2f;
#endif
callbackData.sampleRateRatio = callbackData.minSampleRateRatio;
deviceConfig = ma_device_config_init(ma_device_type_playback);
deviceConfig.playback.format = resamplerConfig.format;
deviceConfig.playback.channels = resamplerConfig.channels;
deviceConfig.sampleRate = resamplerConfig.sampleRateIn;
deviceConfig.periodSizeInFrames = /*2048; //*/256;
deviceConfig.dataCallback = resampler_data_callback;
deviceConfig.pUserData = &callbackData;
result = ma_device_init(NULL, &deviceConfig, &device);
if (result != MA_SUCCESS) {
printf("Failed to initialize device.");
return;
}
ma_device_start(&device);
ma_sleep(100000);
ma_device_uninit(&device);
ma_waveform_uninit(&callbackData.waveform);
ma_resampler_uninit(&callbackData.resampler, NULL);
}
static void profile_resampler_miniaudio(resampling_api api, ma_format format, ma_uint32 channels, ma_uint32 sampleRateIn, ma_uint32 sampleRateOut, ma_uint32 lpfOrder)
{
ma_result result;
ma_timer timer;
double startTime;
void* pSamplesIn = NULL;
void* pSamplesOut = NULL;
ma_uint32 bpf = ma_get_bytes_per_frame(format, channels);
int i;
ma_uint32 framesOutCap = 1024;
printf("%s: %s, %s, LPF %u, %u > %u: ", resampling_api_to_string(api), ((format == ma_format_s16) ? "s16" : "f32"), ((channels == 1) ? "mono" : "stereo"), lpfOrder, sampleRateIn, sampleRateOut);
ma_timer_init(&timer);
if (api == resampling_api_libsamplerate && format != ma_format_f32) {
printf("Format not supported.\n");
return;
}
pSamplesIn = ma_malloc(RESAMPLING_FRAMES_PER_ITERATION * bpf, NULL);
pSamplesOut = ma_malloc(framesOutCap * bpf, NULL);
ma_debug_fill_pcm_frames_with_sine_wave((float*)pSamplesIn, RESAMPLING_FRAMES_PER_ITERATION, format, channels, sampleRateIn); /* The float* cast is to work around an API bug in miniaudio v0.11. It's harmless. */
if (api == resampling_api_miniaudio) {
ma_linear_resampler_config resamplerConfig;
ma_linear_resampler resampler;
resamplerConfig = ma_linear_resampler_config_init(format, channels, sampleRateIn, sampleRateOut);
resamplerConfig.lpfOrder = lpfOrder;
result = ma_linear_resampler_init(&resamplerConfig, NULL, &resampler);
if (result != MA_SUCCESS) {
printf("Failed to initialize resampler.\n");
return;
}
startTime = ma_timer_get_time_in_seconds(&timer);
{
for (i = 0; i < RESAMPLING_ITERATION_COUNT; i += 1) {
ma_uint64 totalFramesRead = 0;
while (totalFramesRead < RESAMPLING_FRAMES_PER_ITERATION) {
ma_uint64 framesIn = RESAMPLING_FRAMES_PER_ITERATION - totalFramesRead;
ma_uint64 framesOut = framesOutCap;
ma_linear_resampler_process_pcm_frames(&resampler, ma_offset_pcm_frames_ptr(pSamplesIn, totalFramesRead, format, channels), &framesIn, pSamplesOut, &framesOut);
totalFramesRead += framesIn;
}
}
}
printf("%f\n", ma_timer_get_time_in_seconds(&timer) - startTime);
}
if (api == resampling_api_libsamplerate) {
#ifdef MA_HAS_LIBSAMPLERATE
{
#if 0
SRC_SINC_BEST_QUALITY
SRC_SINC_MEDIUM_QUALITY
SRC_SINC_FASTEST
SRC_ZERO_ORDER_HOLD
SRC_LINEAR
#endif
SRC_STATE* pState = src_new(SRC_LINEAR, channels, NULL);
startTime = ma_timer_get_time_in_seconds(&timer);
{
for (i = 0; i < RESAMPLING_ITERATION_COUNT; i += 1) {
ma_uint64 totalFramesRead = 0;
while (totalFramesRead < RESAMPLING_FRAMES_PER_ITERATION) {
SRC_DATA data;
data.data_in = ma_offset_pcm_frames_ptr(pSamplesIn, totalFramesRead, format, channels);
data.data_out = pSamplesOut;
data.input_frames = RESAMPLING_FRAMES_PER_ITERATION - totalFramesRead;
data.output_frames = framesOutCap;
data.src_ratio = (double)sampleRateOut / sampleRateIn; /* I think libsamplerate's ratio is out/in, whereas miniaudio is in/out. Advice welcome if I am wrong about this. */
src_process(pState, &data);
totalFramesRead += data.output_frames_gen;
}
}
}
printf("%f\n", ma_timer_get_time_in_seconds(&timer) - startTime);
}
#else
{
printf("libsamplerate not available.\n");
}
#endif
}
if (api == resampling_api_speex) {
#ifdef MA_HAS_SPEEXDSP
{
int err;
SpeexResamplerState* pState = speex_resampler_init(channels, sampleRateIn, sampleRateOut, SPEEX_RESAMPLER_QUALITY_MIN, &err);
if (err != RESAMPLER_ERR_SUCCESS) {
printf("Failed to initialize speex resampler.\n");
return;
}
startTime = ma_timer_get_time_in_seconds(&timer);
{
for (i = 0; i < RESAMPLING_ITERATION_COUNT; i += 1) {
ma_uint64 totalFramesRead = 0;
while (totalFramesRead < RESAMPLING_FRAMES_PER_ITERATION) {
spx_uint32_t in_len = (spx_uint32_t)(RESAMPLING_FRAMES_PER_ITERATION - totalFramesRead);
spx_uint32_t out_len = (spx_uint32_t)framesOutCap;
if (format == ma_format_f32) {
speex_resampler_process_interleaved_float(pState, ma_offset_pcm_frames_ptr(pSamplesIn, totalFramesRead, format, channels), &in_len, pSamplesOut, &out_len);
} else {
speex_resampler_process_interleaved_int(pState, ma_offset_pcm_frames_ptr(pSamplesIn, totalFramesRead, format, channels), &in_len, pSamplesOut, &out_len);
}
totalFramesRead += in_len;
}
}
}
printf("%f\n", ma_timer_get_time_in_seconds(&timer) - startTime);
speex_resampler_destroy(pState);
}
#else
{
printf("speex not available.\n");
}
#endif
}
if (api == resampling_api_soxr) {
#ifdef MA_HAS_SOXR
{
soxr_error_t error;
soxr_io_spec_t ioSpec;
soxr_quality_spec_t qualitySpec;
soxr_t pState;
ioSpec = soxr_io_spec((format == ma_format_f32) ? SOXR_FLOAT32_I : SOXR_INT16_I, (format == ma_format_f32) ? SOXR_FLOAT32_I : SOXR_INT16_I);
qualitySpec = soxr_quality_spec(SOXR_QQ, 0);
pState = soxr_create((double)sampleRateIn, (double)sampleRateOut, channels, &error, &ioSpec, &qualitySpec, NULL);
if (error != NULL) {
printf("Failed to initialize soxr resampler.\n");
return;
}
startTime = ma_timer_get_time_in_seconds(&timer);
{
for (i = 0; i < RESAMPLING_ITERATION_COUNT; i += 1) {
ma_uint64 totalFramesRead = 0;
while (totalFramesRead < RESAMPLING_FRAMES_PER_ITERATION) {
size_t in_len = RESAMPLING_FRAMES_PER_ITERATION - totalFramesRead;
size_t out_len = 0;
soxr_process(pState, ma_offset_pcm_frames_ptr(pSamplesIn, totalFramesRead, format, channels), in_len, &in_len, pSamplesOut, framesOutCap, &out_len);
totalFramesRead += in_len;
}
}
}
printf("%f\n", ma_timer_get_time_in_seconds(&timer) - startTime);
soxr_delete(pState);
}
#else
{
printf("soxr not available.\n");
}
#endif
}
}
void profile_resampler(void)
{
unsigned int denormalsState;
denormalsState = ma_disable_denormals();
{
#if 1
/* miniaudio */
profile_resampler_miniaudio(resampling_api_miniaudio, ma_format_s16, 1, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_miniaudio, ma_format_s16, 2, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_miniaudio, ma_format_f32, 1, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_miniaudio, ma_format_f32, 2, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_miniaudio, ma_format_s16, 1, 48000, 44100, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_miniaudio, ma_format_s16, 2, 48000, 44100, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_miniaudio, ma_format_f32, 1, 48000, 44100, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_miniaudio, ma_format_f32, 2, 48000, 44100, RESAMPLING_LPF_ORDER);
/* libsamplerate */
profile_resampler_miniaudio(resampling_api_libsamplerate, ma_format_f32, 1, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_libsamplerate, ma_format_f32, 2, 44100, 48000, RESAMPLING_LPF_ORDER);
/* speex */
profile_resampler_miniaudio(resampling_api_speex, ma_format_s16, 1, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_speex, ma_format_s16, 2, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_speex, ma_format_f32, 1, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_speex, ma_format_f32, 2, 44100, 48000, RESAMPLING_LPF_ORDER);
/* soxr */
profile_resampler_miniaudio(resampling_api_soxr, ma_format_s16, 1, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_soxr, ma_format_s16, 2, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_soxr, ma_format_f32, 1, 44100, 48000, RESAMPLING_LPF_ORDER);
profile_resampler_miniaudio(resampling_api_soxr, ma_format_f32, 2, 44100, 48000, RESAMPLING_LPF_ORDER);
#endif
}
ma_restore_denormals(denormalsState);
/* Finish off with a listening test. */
resampler_listening_test();
}
int main(int argc, char** argv)
{
profile_resampler();
(void)argc;
(void)argv;
return 0;
}