FFMPEG + SDL音频播放分析

目录 [hide]

• 1 抽象流程：
• 2 关键实现：
  • 2.1 main()函数
  • 2.2 decode_thread()读取文件信息和音频包
  • 2.3 stream_component_open()：设置音频参数和打开设备
  • 2.4 audio_callback(): 回调函数，向SDL缓冲区填充数据
  • 2.5 audio_decode_frame()：解码音频
• 3 FFMPEG结构体
  • 3.1 channel_layout_map
• 4 FFMPEG宏定义
  • 4.1 Audio channel convenience macros
• 5 SDL宏定义
  • 5.1 SDL_AudioSpec format

抽象流程：

设置SDL的音频参数 —-> 打开声音设备，播放静音 —-> ffmpeg读取音频流中数据放入队列 —-> SDL调用用户设置的函数来获取音频数据 —-> 播放音频

SDL内部维护了一个buffer来存放解码后的数据，这个buffer中的数据来源是我们注册的回调函数（audio_callback），audio_callback调用audio_decode_frame来做具体的音频解码工作，需要引起注意的是：从流中读取出的一个音频包(avpacket)可能含有多个音频桢(avframe)，所以需要多次调用avcodec_decode_audio4来完成整个包的解码，解码出来的数据存放在我们自己的缓冲中（audio_buf2）。SDL每一次回调都会引起数据从audio_buf2拷贝到SDL内部缓冲区，当audio_buf2中的数据大于SDL的缓冲区大小时，需要分多次拷贝。

关键实现：

main()函数

1	int main(int argc, char **argv){

2	SDL_Event event; //SDL事件变量

3	VideoState    *is; // 纪录视频及解码器等信息的大结构体

4	is = (VideoState*) av_mallocz(sizeof(VideoState));

5	if(argc < 2){

6	fprintf(stderr, "Usage: play <file>\n");

7

exit(1);

8

}

9	av_register_all(); //注册所有ffmpeg的解码器

10	/* 初始化SDL，这里只实用了AUDIO，如果有视频，好需要SDL_INIT_VIDEO等等 */

11	if(SDL_Init(SDL_INIT_AUDIO)){

12	fprintf(stderr, "Count not initialize SDL - %s\n", SDL_GetError());

13

exit(1);

14

}

15	is_strlcpy(is->filename, argv[1], sizeof(is->filename));

16	/* 创建一个SDL线程来做视频解码工作，主线程进入SDL事件循环 */

17	is->parse_tid = SDL_CreateThread(decode_thread, is);

18	if(!is->parse_tid){

19	SDL_WaitEvent(&event);

20	switch(event.type){

21	case FF_QUIT_EVENT:

22	case SDL_QUIT:

23	is->quit = 1;

24	SDL_Quit();

25

exit(0);

26

break;

27

default:

28

break;

29

}

30

}

31

return 0;

32

}

decode_thread()读取文件信息和音频包

1	static int decode_thread(void *arg){

2	VideoState *is = (VideoState*)arg;

3	AVFormatContext *ic = NULL;

4	AVPacket pkt1, *packet = &pkt1;

5	int ret, i, audio_index = -1;

6

7	is->audioStream = -1;

8	global_video_state = is;

9	/*  使用ffmpeg打开视频，解码器等 常规工作 */

10	if(avFormat_open_input(&ic, is->filename, NULL,  NULL) != 0)  {

11	fprintf(stderr, "open file error: %s\n", is->filename);

12	return -1;

13

}

14	is->ic = ic;

15	if(avformat_find_stream_info(ic, NULL) < 0){

16	fprintf(stderr, "find stream info error\n");

17	return -1;

18

}

19	av_dump_format(ic, 0, is->filename, 0);

20	for(i  = 0; i < ic->nb_streams; i++){

21	if(ic->streams[i])->codec->codec_type == AVMEDIA_TYPE_AUDIO && audio_index == -1){

22	audio_index = i;

23

break;

24

}

25

}

26	if(audio_index >= 0) {

27	/* 所有设置SDL音频流信息的步骤都在这个函数里完成 */

28	stream_component_open(is, audio_index);

29

}

30	if(is->audioStream < 0){

31	fprintf(stderr, "could not open codecs for file: %s\n", is->filename);

32	goto fail;

33

}

34	/* 读包的主循环， av_read_frame不停的从文件中读取数据包（这里只取音频包）*/

35

for(;;){

36	if(is->quit) break;

37	/* 这里audioq.size是指队列中的所有数据包带的音频数据的总量，并不是包的数量 */

38	if(is->audioq.size > MAX_AUDIO_SIZE){

39	SDL_Delay(10); // 毫秒

40

continue;

41

}

42	ret = av_read_frame(is->ic, packet);

43	if(ret < 0){

44	if(ret == AVERROR_EOF || url_feof(is->ic->pb))    break;

45	if(is->ic->pb && is->ic->pb->error)    break;

46

contiue;

47

}

48	if(packet->stream_index == is->audioStream){

49	packet_queue_put(&is->audioq, packet);

50

} else{

51	av_free_packet(packet);

52

}

53

}

54	while(!is->quit)    SDL_Delay(100);

55

fail: {

56	SDL_Event event;

57	event.type = FF_QUIT_EVENT;

58	event.user.data1 = is;

59	SDL_PushEvent(&event);

60

}

61

return 0;

62

}

stream_component_open()：设置音频参数和打开设备

1	int stream_component_open(videoState *is, int stream_index){

2	AVFormatContext *ic = is->ic;

3	AVCodecContext *codecCtx;

4	AVCodec *codec;

5	/* 在用SDL_OpenAudio()打开音频设备的时候需要这两个参数*/

6	/* wanted_spec是我们期望设置的属性，spec是系统最终接受的参数 */

7	/* 我们需要检查系统接受的参数是否正确 */

8	SDL_AudioSpec wanted_spec, spec;

9	int64_t wanted_channel_layout = 0; // 声道布局(SDL中的具体定义见“FFMPEG结构体”部分)

10	int wanted_nb_channels; // 声道数

11	/*  SDL支持的声道数为 1, 2, 4, 6 */

12	/*  后面我们会使用这个数组来纠正不支持的声道数目 */

13	const int next_nb_channels[] = { 0, 0, 1, 6,  2, 6, 4, 6 };

14

15	if(stream_index < 0 || stream_index >= ic->nb_streams)    return -1;

16	codecCtx = ic->streams[stream_index]->codec;

17	wanted_nb_channels = codecCtx->channels;

18	if(!wanted_channel_layout || wanted_nb_channels != av_get_channel_layout_nb_channels(wanted_channel_layout)) {

19	wanted_channel_layout = av_get_default_channel_lauout(wanted_channel_nb_channels);

20	wanted_channel_layout &= ~AV_CH_LAYOUT_STEREO_DOWNMIX;

21

}

22	wanted_spec.channels = av_get_channels_layout_nb_channels(wanted_channel_layout);

23	wanted_spec.freq = codecCtx->sample_rate;

24	if(wanted_spec.freq <= 0 || wanted_spec.channels <=0){

25	fprintf(stderr, "Invaild sample rate or channel count!\n");

26	return -1;

27

}

28	wanted_spec.format = AUDIO_S16SYS; // 具体含义请查看“SDL宏定义”部分

29	wanted_spec.silence = 0; // 0指示静音

30	wanted_spec.samples = SDL_AUDIO_BUFFER_SIZE; // 自定义SDL缓冲区大小

31	wanted_spec.callback = audio_callback; // 音频解码的关键回调函数

32	wanted_spec.userdata = is; // 传给上面回调函数的外带数据

33

34	/*  打开音频设备，这里使用一个while来循环尝试打开不同的声道数(由上面 */

35	/*  next_nb_channels数组指定）直到成功打开，或者全部失败 */

36	while(SDL_OpenAudio(&wanted_spec, &spec) < 0){

37	fprintf(stderr, "SDL_OpenAudio(%d channels): %s\n", wanted_spec.channels, SDL_GetError());

38	wanted_spec.channels = next_nb_channels[FFMIN(7, wanted_spec.channels)]; // FFMIN()由ffmpeg定义的宏，返回较小的数

39	if(!wanted_spec.channels){

40	fprintf(stderr, "No more channel to try\n");

41	return -1;

42

}

43	wanted_channel_layout = av_get_default_channel_layout(wanted_spec.channels);

44

}

45	/* 检查实际使用的配置（保存在spec,由SDL_OpenAudio()填充） */

46	if(spec.format != AUDIO_S16SYS){

47	fprintf(stderr, "SDL advised audio format %d is not supported\n", spec.format);

48	return -1;

49

}

50	if(spec.channels != wanted_spec.channels) {

51	wanted_channel_layout = av_get_default_channel_layout(spec.channels);

52	if(!wanted_channel_layout){

53	fprintf(stderr, "SDL advised channel count %d is not support\n", spec.channels);

54	return -1;

55

}

56

}

57	/* 把设置好的参数保存到大结构中 */

58	is->audio_src_fmt = is->audio_tgt_fmt = AV_SAMPLE_FMT_S16;

59	is->audio_src_freq = is->audio_tgt_freq = spec.freq;

60	is->audio_src_channel_layout = is->audio_tgt_layout = wanted_channel_layout;

61	is->audio_src_channels = is->audio_tat_channels = spec.channels;

62

63	codec = avcodec_find_decoder(codecCtx>codec_id);

64	if(!codec || (avcodec_open2(codecCtx, codec, NULL) < 0)){

65	fprintf(stderr, "Unsupported codec!\n");

66	return -1;

67

}

68	ic->streams[stream_index]->discard = AVDISCARD_DEFAULT; //具体含义请查看“FFMPEG宏定义”部分

69	is->audioStream = stream_index;

70	is->audio_st = ic->streams[stream_index];

71	is->audio_buf_size = 0;

72	is->audio_buf_index = 0;

73	memset(&is->audio_pkt, 0, sizeof(is->audio_pkt));

74	packet_queue_init(&is->audioq);

75	SDL_PauseAudio(0); // 开始播放静音

76

}

audio_callback(): 回调函数，向SDL缓冲区填充数据

1	void audio_callback(void *userdata, Uint8 *stream, int len){

2	VideoState *is = (VideoState*)userdata;

3	int len1, audio_data_size;

4

5	/*   len是由SDL传入的SDL缓冲区的大小，如果这个缓冲未满，我们就一直往里填充数据 */

6	while(len > 0){

7	/*  audio_buf_index 和 audio_buf_size 标示我们自己用来放置解码出来的数据的缓冲区，*/

8	/*   这些数据待copy到SDL缓冲区， 当audio_buf_index >= audio_buf_size的时候意味着我*/

9	/*   们的缓冲为空，没有数据可供copy，这时候需要调用audio_decode_frame来解码出更

10	/*   多的桢数据 */

11	if(is->audio_buf_index >= is->audio_buf_size){

12	audio_data_size = audio_decode_frame(is);

13	/* audio_data_size < 0 标示没能解码出数据，我们默认播放静音 */

14	is(audio_data_size < 0){

15	is->audio_buf_size = 1024;

16	/* 清零，静音 */

17	memset(is->audio_buf, 0, is->audio_buf_size);

18

} else{

19	is->audio_buf_size = audio_data_size;

20

}

21	is->audio_buf_index = 0;

22

}

23	/*  查看stream可用空间，决定一次copy多少数据，剩下的下次继续copy */

24	len1 = is->audio_buf_size - is->audio_buf_index;

25	if(len1 > len)    len1 = len;

26

27	memcpy(stream, (uint8_t*)is->audio_buf + is->audio_buf_index, len1);

28	len -= len1;

29	stream += len1;

30	is->audio_buf_index += len1;

31

}

32

}

audio_decode_frame()：解码音频

1	int audio_decode_frame(VideoState *is){

2	int len1, len2, decoded_data_size;

3	AVPacket *pkt = &is->audio_pkt;

4	int got_frame = 0;

5	int64_t dec_channel_layout;

6	int wanted_nb_samples, resampled_data_size;

7

8

for(;;){

9	while(is->audio_pkt_size > 0){

10	if(!is->audio_frame){

11	if(!(is->audio_frame = avacodec_alloc_frame())){

12	return AVERROR(ENOMEM);

13

}

14

} else

15	avcodec_get_frame_defaults(is->audio_frame);

16

17	len1 = avcodec_decode_audio4(is->audio_st_codec, is->audio_frame, got_frame, pkt);

18	/* 解码错误，跳过整个包 */

19	if(len1 < 0){

20	is->audio_pkt_size = 0;

21

break;

22

}

23	is->audio_pkt_data += len1;

24	is->audio_pkt_size -= len1;

25	if(!got_frame)   continue;

26	/* 计算解码出来的桢需要的缓冲大小 */

27	decoded_data_size = av_samples_get_buffer_size(NULL,

28	is->audio_frame_channels,

29	is->audio_frame_nb_samples,

30	is->audio_frame_format, 1);

31	dec_channel_layout = (is->audio_frame->channel_layout && is->audio_frame->channels

32	== av_get_channel_layout_nb_channels(is->audio_frame->channel_layout))

33	? is->audio_frame->channel_layout : av_get_default_channel_layout(is->audio_frame->channels);

34	wanted_nb_samples =  is->audio_frame->nb_samples;

35	if (is->audio_frame->format != is->audio_src_fmt ||

36	dec_channel_layout != is->audio_src_channel_layout ||

37	is->audio_frame->sample_rate != is->audio_src_freq ||

38	(wanted_nb_samples != is->audio_frame->nb_samples && !is->swr_ctx)) {

39	if (is->swr_ctx) swr_free(&is->swr_ctx);

40	is->swr_ctx = swr_alloc_set_opts(NULL,

41	is->audio_tgt_channel_layout,

42	is->audio_tgt_fmt,

43	is->audio_tgt_freq,

44	dec_channel_layout,

45	is->audio_frame->format,

46	is->audio_frame->sample_rate,

47

0, NULL);

48	if (!is->swr_ctx || swr_init(is->swr_ctx) < 0) {

49	fprintf(stderr, "swr_init() failed\n");

50

break;

51

}

52	is->audio_src_channel_layout = dec_channel_layout;

53	is->audio_src_channels = is->audio_st->codec->channels;

54	is->audio_src_freq = is->audio_st->codec->sample_rate;

55	is->audio_src_fmt = is->audio_st->codec->sample_fmt;

56

}

57	/* 这里我们可以对采样数进行调整，增加或者减少，一般可以用来做声画同步 */

58	if (is->swr_ctx) {

59	const uint8_t **in = (const uint8_t **)is->audio_frame->extended_data;

60	uint8_t *out[] = { is->audio_buf2 };

61	if (wanted_nb_samples != is->audio_frame->nb_samples) {

62	if(swr_set_compensation(is->swr_ctx,

63	(wanted_nb_samples - is->audio_frame->nb_samples)*is->audio_tgt_freq/is->audio_frame->sample_rate,

64	wanted_nb_samples * is->audio_tgt_freq/is->audio_frame->sample_rate) < 0) {

65	fprintf(stderr, "swr_set_compensation() failed\n");

66

break;

67

}

68

}

69	len2 = swr_convert(is->swr_ctx, out,

70	sizeof(is->audio_buf2)/is->audio_tgt_channels/av_get_bytes_per_sample(is->audio_tgt_fmt),

71	in, is->audio_frame->nb_samples);

72	if (len2 < 0) {

73	fprintf(stderr, "swr_convert() failed\n");

74

break;

75

}

76	if(len2 == sizeof(is->audio_buf2)/is->audio_tgt_channels/av_get_bytes_per_sample(is->audio_tgt_fmt)) {

77	fprintf(stderr, "warning: audio buffer is probably too small\n");

78	swr_init(is->swr_ctx);

79

}

80	is->audio_buf = is->audio_buf2;

81	resampled_data_size = len2*is->audio_tgt_channels*av_get_bytes_per_sample(is->audio_tgt_fmt);

82

} else {

83	resampled_data_size = decoded_data_size;

84	is->audio_buf = is->audio_frame->data[0];

85

}

86	/*  返回得到的数据 */

87	return resampled_data_size;

88

}

89	if (pkt->data) av_free_packet(pkt);

90	memset(pkt, 0, sizeof(*pkt));

91	if (is->quit) return -1;

92	if (packet_queue_get(&is->audioq, pkt, 1) < 0) return -1;

93	is->audio_pkt_data = pkt->data;

94	is->audio_pkt_size = pkt->size;

95

96

}

97

}

FFMPEG结构体

channel_layout_map

1	static const struct {

2	const char *name;

3	int nb_channels;

4	uint64_t layout;

5	} channel_layout_map[] = {

6	{ "mono", 1, AV_CH_LAYOUT_MONO },

7	{ "stereo", 2, AV_CH_LAYOUT_STEREO },

8	{ "2.1", 3, AV_CH_LAYOUT_2POINT1 },

9	{ "3.0", 3, AV_CH_LAYOUT_SURROUND },

10	{ "3.0(back)", 3, AV_CH_LAYOUT_2_1 },

11	{ "4.0", 4, AV_CH_LAYOUT_4POINT0 },

12	{ "quad", 4, AV_CH_LAYOUT_QUAD },

13	{ "quad(side)", 4, AV_CH_LAYOUT_2_2 },

14	{ "3.1", 4, AV_CH_LAYOUT_3POINT1 },

15	{ "5.0", 5, AV_CH_LAYOUT_5POINT0_BACK },

16	{ "5.0(side)", 5, AV_CH_LAYOUT_5POINT0 },

17	{ "4.1", 5, AV_CH_LAYOUT_4POINT1 },

18	{ "5.1", 6, AV_CH_LAYOUT_5POINT1_BACK },

19	{ "5.1(side)", 6, AV_CH_LAYOUT_5POINT1 },

20	{ "6.0", 6, AV_CH_LAYOUT_6POINT0 },

21	{ "6.0(front)", 6, AV_CH_LAYOUT_6POINT0_FRONT },

22	{ "hexagonal", 6, AV_CH_LAYOUT_HEXAGONAL },

23	{ "6.1", 7, AV_CH_LAYOUT_6POINT1 },

24	{ "6.1", 7, AV_CH_LAYOUT_6POINT1_BACK },

25	{ "6.1(front)", 7, AV_CH_LAYOUT_6POINT1_FRONT },

26	{ "7.0", 7, AV_CH_LAYOUT_7POINT0 },

27	{ "7.0(front)", 7, AV_CH_LAYOUT_7POINT0_FRONT },

28	{ "7.1", 8, AV_CH_LAYOUT_7POINT1 },

29	{ "7.1(wide)", 8, AV_CH_LAYOUT_7POINT1_WIDE },

30	{ "octagonal", 8, AV_CH_LAYOUT_OCTAGONAL },

31	{ "downmix", 2, AV_CH_LAYOUT_STEREO_DOWNMIX, },

32

};

FFMPEG宏定义

Audio channel convenience macros

1	#define AV_CH_LAYOUT_MONO              (AV_CH_FRONT_CENTER)

2	#define AV_CH_LAYOUT_STEREO            (AV_CH_FRONT_LEFT|AV_CH_FRONT_RIGHT)

3	#define AV_CH_LAYOUT_2POINT1           (AV_CH_LAYOUT_STEREO|AV_CH_LOW_FREQUENCY)

4	#define AV_CH_LAYOUT_2_1               (AV_CH_LAYOUT_STEREO|AV_CH_BACK_CENTER)

5	#define AV_CH_LAYOUT_SURROUND          (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER)

6	#define AV_CH_LAYOUT_3POINT1           (AV_CH_LAYOUT_SURROUND|AV_CH_LOW_FREQUENCY)

7	#define AV_CH_LAYOUT_4POINT0           (AV_CH_LAYOUT_SURROUND|AV_CH_BACK_CENTER)

8	#define AV_CH_LAYOUT_4POINT1           (AV_CH_LAYOUT_4POINT0|AV_CH_LOW_FREQUENCY)

9	#define AV_CH_LAYOUT_2_2               (AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT)

10	#define AV_CH_LAYOUT_QUAD              (AV_CH_LAYOUT_STEREO|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT)

11	#define AV_CH_LAYOUT_5POINT0           (AV_CH_LAYOUT_SURROUND|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT)

12	#define AV_CH_LAYOUT_5POINT1           (AV_CH_LAYOUT_5POINT0|AV_CH_LOW_FREQUENCY)

13	#define AV_CH_LAYOUT_5POINT0_BACK      (AV_CH_LAYOUT_SURROUND|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT)

14	#define AV_CH_LAYOUT_5POINT1_BACK      (AV_CH_LAYOUT_5POINT0_BACK|AV_CH_LOW_FREQUENCY)

15	#define AV_CH_LAYOUT_6POINT0           (AV_CH_LAYOUT_5POINT0|AV_CH_BACK_CENTER)

16	#define AV_CH_LAYOUT_6POINT0_FRONT     (AV_CH_LAYOUT_2_2|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER)

17	#define AV_CH_LAYOUT_HEXAGONAL         (AV_CH_LAYOUT_5POINT0_BACK|AV_CH_BACK_CENTER)

18	#define AV_CH_LAYOUT_6POINT1           (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER)

19	#define AV_CH_LAYOUT_6POINT1_BACK      (AV_CH_LAYOUT_5POINT1_BACK|AV_CH_BACK_CENTER)

20	#define AV_CH_LAYOUT_6POINT1_FRONT     (AV_CH_LAYOUT_6POINT0_FRONT|AV_CH_LOW_FREQUENCY)

21	#define AV_CH_LAYOUT_7POINT0           (AV_CH_LAYOUT_5POINT0|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT)

22	#define AV_CH_LAYOUT_7POINT0_FRONT     (AV_CH_LAYOUT_5POINT0|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER)

23	#define AV_CH_LAYOUT_7POINT1           (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT)

24	#define AV_CH_LAYOUT_7POINT1_WIDE      (AV_CH_LAYOUT_5POINT1|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER)

25	#define AV_CH_LAYOUT_7POINT1_WIDE_BACK (AV_CH_LAYOUT_5POINT1_BACK|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER)

26	#define AV_CH_LAYOUT_OCTAGONAL         (AV_CH_LAYOUT_5POINT0|AV_CH_BACK_LEFT|AV_CH_BACK_CENTER|AV_CH_BACK_RIGHT)

27	#define AV_CH_LAYOUT_STEREO_DOWNMIX    (AV_CH_STEREO_LEFT|AV_CH_STEREO_RIGHT)

SDL宏定义

SDL_AudioSpec format

查看源代码

 

打印 帮助

1	AUDIO_U8           Unsigned 8-bit samples

2	AUDIO_S8            Signed 8-bit samples

3	AUDIO_U16LSB    Unsigned 16-bit samples, in little-endian byte order

4	AUDIO_S16LSB    Signed 16-bit samples, in little-endian byte order

5	AUDIO_U16MSB    Unsigned 16-bit samples, in big-endian byte order

6	AUDIO_S16MSB    Signed 16-bit samples, in big-endian byte order

7	AUDIO_U16           same as AUDIO_U16LSB (for backwards compatability probably)

8	AUDIO_S16           same as AUDIO_S16LSB (for backwards compatability probably)

9	AUDIO_U16SYS    Unsigned 16-bit samples, in system byte order

10	AUDIO_S16SYS     Signed 16-bit samples, in system byte order

git clone https://github.com/lnmcc/musicPlayer.git