第一章 视频渲染
第二章 音频(push)播放
第三章 音频(pull)播放(本章)
第四章 实现时钟同步
第五章 实现通用时钟同步
第六章 实现播放器
我们上一章讲了,ffmpeg解码sdl push的方式播放音频,调用流程简单,但是实现起来还是有点难度的。接下来讲的就是使用pull的方式播放音频,pull的方式即是使用回调的方式播放音频,在打开SDL音频设备的时候传入一个回调方法,SDL内部会按照一定频率回调这个方法,我们在回调方法中往音频缓冲写数据就能够播放声音了。
ffmpeg解码部分与《使用ffmpeg解码音频sdl(push)播放》一致,这里就不再赘述。
使用sdl前需要在最开始初始化sdl,全局只需要初始化一次即可。
if (SDL_Init(SDL_INIT_VIDEO | SDL_INIT_AUDIO | SDL_INIT_TIMER)) {
printf("Could not initialize SDL - %s\n", SDL_GetError());
return -1;
}
建议使用SDL_OpenAudioDevice打开设备,使用SDL_OpenAudio的话samples设置可能不生效,比较不灵活一点。pull的方式播放其实就是采用回调的方式播放,我们给下面代码的wanted_spec.callback赋值即可。
SDL_AudioSpec wanted_spec, spec;
int audioId = 0;
//打开设备
wanted_spec.channels = av_get_channel_layout_nb_channels(pCodecCtx->channel_layout);
wanted_spec.freq = pCodecCtx->sample_rate;
wanted_spec.format = AUDIO_F32SYS;
wanted_spec.silence = 0;
wanted_spec.samples = FFMAX(512, 2 << av_log2(wanted_spec.freq / 30));
//注册回调方法
wanted_spec.callback = audioCallback;
wanted_spec.userdata = NULL;
audioId = SDL_OpenAudioDevice(NULL, 0, &wanted_spec, &spec, 1);
if (audioId < 2)
{
printf("Open audio device error!\n");
goto end;
}
//开启播放
SDL_PauseAudioDevice(audioId, 0);
我们采用pull的方式播放,即注册回调方法,sdl会按照一定的频率触发回调,我们只需往回调参数的缓存指针写入音频数据即可。
//音频设备读取数据回调方法
void audioCallback(void* userdata, Uint8* stream, int len) {
//TODO:从音频队列中读取数据到设备缓冲
};
使用完成后需要销毁资源,如下所示,SDL_Quit并不是必要的,通常是程序退出才需要调用,这个时候调不调已经无所谓了。
if (audioId >= 2)
{
SDL_PauseAudioDevice(audioId, 1);
SDL_CloseAudioDevice(audioId);
}
SDL_Quit();
由于是采用回调的方式播放,必然需要一个队列往里面写入解码的数据,音频设备回调时再将数据读出。我们直接使用ffmpeg提供的音频队列即可。
AVAudioFifo是基于采样数的,所以初始化的时候需要设置音频格式以及队列长度采样数。直接使用音频设备的参数,队列长度要稍微长一点。
//使用音频队列
AVAudioFifo *fifo = av_audio_fifo_alloc(forceFormat, spec.channels, spec.samples * 10);
我们需要解码处往队列里写入数据
//解码后的音频数据
uint8_t* data;
//音频数据的采样数
size_t samples;
if (av_audio_fifo_space(fifo) >= samples)
{
av_audio_fifo_write(fifo, (void**)&data, samples);
}
在音频设备回调中读取队列中的音频数据
void audioCallback(void* userdata, Uint8* stream, int len) {
//从音频队列中读取数据到设备缓冲
av_audio_fifo_read(fifo, (void**)&stream, spec.samples);
};
if (fifo)
{
av_audio_fifo_free(fifo);
}
将上述代码合并起来形成一个完整的音频解码播放流程:
示例的sdk版本:ffmpeg 4.3、sdl2
windows、linux都可以正常运行
#include
#include
#include "libavformat/avformat.h"
#include "libavcodec/avcodec.h"
#include "libswscale/swscale.h"
#include "libswresample/swresample.h"
#include "libavutil/audio_fifo.h"
#undef main
AVAudioFifo* fifo = NULL;
SDL_AudioSpec wanted_spec, spec;
SDL_mutex* mtx = NULL;
static void audioCallback(void* userdata, Uint8* stream, int len) {
//由于AVAudioFifo非线程安全,且是子线程触发此回调,所以需要加锁
SDL_LockMutex(mtx);
//读取队列中的音频数据
av_audio_fifo_read(fifo, (void**)&stream, spec.samples);
SDL_UnlockMutex(mtx);
};
int main(int argc, char** argv) {
const char* input = "test_music.wav";
enum AVSampleFormat forceFormat;
AVFormatContext* pFormatCtx = NULL;
AVCodecContext* pCodecCtx = NULL;
const AVCodec* pCodec = NULL;
AVDictionary* opts = NULL;
AVPacket packet;
AVFrame* pFrame = NULL;
struct SwrContext* swr_ctx = NULL;
uint8_t* outBuffer = NULL;
int audioindex = -1;
int exitFlag = 0;
int isLoop = 1;
double framerate;
int audioId = 0;
memset(&packet, 0, sizeof(AVPacket));
//初始化SDL
if (SDL_Init(SDL_INIT_VIDEO | SDL_INIT_AUDIO | SDL_INIT_TIMER)) {
printf("Could not initialize SDL - %s\n", SDL_GetError());
return -1;
}
//打开输入流
if (avformat_open_input(&pFormatCtx, input, NULL, NULL) != 0) {
printf("Couldn't open input stream.\n");
goto end;
}
//查找输入流信息
if (avformat_find_stream_info(pFormatCtx, NULL) < 0) {
printf("Couldn't find stream information.\n");
goto end;
}
//获取音频流
for (unsigned i = 0; i < pFormatCtx->nb_streams; i++)
if (pFormatCtx->streams[i]->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
audioindex = i;
break;
}
if (audioindex == -1) {
printf("Didn't find a audio stream.\n");
goto end;
}
//创建解码上下文
pCodecCtx = avcodec_alloc_context3(NULL);
if (pCodecCtx == NULL)
{
printf("Could not allocate AVCodecContext\n");
goto end;
}
//获取解码器
if (avcodec_parameters_to_context(pCodecCtx, pFormatCtx->streams[audioindex]->codecpar) < 0)
{
printf("Could not init AVCodecContext\n");
goto end;
}
pCodec = avcodec_find_decoder(pCodecCtx->codec_id);
if (pCodec == NULL) {
printf("Codec not found.\n");
goto end;
}
//使用多线程解码
if (!av_dict_get(opts, "threads", NULL, 0))
av_dict_set(&opts, "threads", "auto", 0);
//打开解码器
if (avcodec_open2(pCodecCtx, pCodec, &opts) < 0) {
printf("Could not open codec.\n");
goto end;
}
if (pCodecCtx->sample_fmt == AV_SAMPLE_FMT_NONE)
{
printf("Unknown sample foramt.\n");
goto end;
}
if (pCodecCtx->sample_rate <= 0 || pFormatCtx->streams[audioindex]->codecpar->channels <= 0)
{
printf("Invalid sample rate or channel count!\n");
goto end;
}
//打开设备
wanted_spec.channels = pFormatCtx->streams[audioindex]->codecpar->channels;
wanted_spec.freq = pCodecCtx->sample_rate;
wanted_spec.format = AUDIO_F32SYS;
wanted_spec.silence = 0;
wanted_spec.samples = FFMAX(512, 2 << av_log2(wanted_spec.freq / 30));
wanted_spec.callback = audioCallback;
wanted_spec.userdata = NULL;
audioId = SDL_OpenAudioDevice(NULL, 0, &wanted_spec, &spec, 1);
if (audioId < 2)
{
printf("Open audio device error!\n");
goto end;
}
switch (spec.format)
{
case AUDIO_S16SYS:
forceFormat = AV_SAMPLE_FMT_S16;
break;
case AUDIO_S32SYS:
forceFormat = AV_SAMPLE_FMT_S32;
break;
case AUDIO_F32SYS:
forceFormat = AV_SAMPLE_FMT_FLT;
break;
default:
printf("audio device format was not surported!\n");
goto end;
break;
}
pFrame = av_frame_alloc();
if (!pFrame)
{
printf("alloc frame failed!\n");
goto end;
}
//使用音频队列
fifo = av_audio_fifo_alloc(forceFormat, spec.channels, spec.samples * 10);
if (!fifo)
{
printf("alloc fifo failed!\n");
goto end;
}
mtx = SDL_CreateMutex();
if (!mtx)
{
printf("alloc mutex failed!\n");
goto end;
}
SDL_PauseAudioDevice(audioId, 0);
start:
while (!exitFlag)
{
//读取包
int gotPacket = av_read_frame(pFormatCtx, &packet) == 0;
if (!gotPacket || packet.stream_index == audioindex)
//!gotPacket:未获取到packet需要将解码器的缓存flush,所以还需要进一次解码流程。
{
//发送包
if (avcodec_send_packet(pCodecCtx, &packet) < 0)
{
printf("Decode error.\n");
av_packet_unref(&packet);
goto end;
}
//接收解码的帧
while (avcodec_receive_frame(pCodecCtx, pFrame) == 0) {
uint8_t* data;
size_t dataSize;
size_t samples;
if (forceFormat != pCodecCtx->sample_fmt || spec.freq != pFrame->sample_rate || spec.channels != pFrame->channels)
//重采样
{
//计算输入采样数
int out_count = (int64_t)pFrame->nb_samples * spec.freq / pFrame->sample_rate + 256;
//计算输出数据大小
int out_size = av_samples_get_buffer_size(NULL, spec.channels, out_count, forceFormat, 0);
//输入数据指针
const uint8_t** in = (const uint8_t**)pFrame->extended_data;
//输出缓冲区指针
uint8_t** out = &outBuffer;
int len2 = 0;
if (out_size < 0) {
av_log(NULL, AV_LOG_ERROR, "av_samples_get_buffer_size() failed\n");
goto end;
}
if (!swr_ctx)
//初始化重采样对象
{
swr_ctx = swr_alloc_set_opts(NULL, av_get_default_channel_layout(spec.channels), forceFormat, spec.freq, av_get_default_channel_layout(pFormatCtx->streams[audioindex]->codecpar->channels) , pCodecCtx->sample_fmt, pCodecCtx->sample_rate, 0, NULL);
if (!swr_ctx || swr_init(swr_ctx) < 0) {
av_log(NULL, AV_LOG_ERROR, "swr_alloc_set_opts() failed\n");
goto end;
}
}
if (!outBuffer)
//申请输出缓冲区
{
outBuffer = (uint8_t*)av_mallocz(out_size);
}
//执行重采样
len2 = swr_convert(swr_ctx, out, out_count, in, pFrame->nb_samples);
if (len2 < 0) {
av_log(NULL, AV_LOG_ERROR, "swr_convert() failed\n");
goto end;
}
//取得输出数据
data = outBuffer;
//输出数据长度
dataSize = av_samples_get_buffer_size(0, spec.channels, len2, forceFormat, 1);
samples = len2;
}
else
{
data = pFrame->data[0];
dataSize = av_samples_get_buffer_size(pFrame->linesize, pFrame->channels, pFrame->nb_samples, forceFormat, 0);
samples = pFrame->nb_samples;
}
//写入队列
while (1) {
SDL_LockMutex(mtx);
if (av_audio_fifo_space(fifo) >= samples)
{
av_audio_fifo_write(fifo, (void**)&data, samples);
SDL_UnlockMutex(mtx);
break;
}
SDL_UnlockMutex(mtx);
//队列可用空间不足则延时等待
SDL_Delay((dataSize) * 1000.0 / (spec.freq * av_get_bytes_per_sample(forceFormat) * spec.channels) - 1);
}
}
}
av_packet_unref(&packet);
if (!gotPacket)
{
//循环播放时flush出缓存帧后需要调用此方法才能重新解码。
avcodec_flush_buffers(pCodecCtx);
break;
}
}
if (!exitFlag)
{
if (isLoop)
{
//定位到起点
if (avformat_seek_file(pFormatCtx, -1, 0, 0, 0, AVSEEK_FLAG_FRAME) >= 0)
{
goto start;
}
}
}
end:
//销毁资源
if (fifo)
{
av_audio_fifo_free(fifo);
}
if (audioId >= 2)
{
SDL_PauseAudioDevice(audioId, 1);
SDL_CloseAudioDevice(audioId);
}
if (mtx)
{
SDL_DestroyMutex(mtx);
}
if (pFrame)
{
if (pFrame->format != -1)
{
av_frame_unref(pFrame);
}
av_frame_free(&pFrame);
}
if (packet.data)
{
av_packet_unref(&packet);
}
if (pCodecCtx)
{
avcodec_close(pCodecCtx);
avcodec_free_context(&pCodecCtx);
}
if (pFormatCtx)
avformat_close_input(&pFormatCtx);
if (pFormatCtx)
avformat_free_context(pFormatCtx);
swr_free(&swr_ctx);
av_dict_free(&opts);
if (outBuffer)
av_free(outBuffer);
SDL_Quit();
return 0;
}
以上就是今天要讲的内容,使用pull的方式播发音频比push简单很多,而且更加灵活可以继续实现更复杂的功能比如多路音频合并,使用push则难以实现。我们唯一要注意的就是保证线程安全,对队列的读写也非常简单,长度控制直接通过延时就可以做到。