一、get_audio_frame_duration函数
(一)get_audio_frame_duration函数的定义
get_audio_frame_duration函数定义在FFmpeg源码(本文演示用的FFmpeg源码版本为7.0.1)的源文件libavcodec/utils.c中:
static int get_audio_frame_duration(enum AVCodecID id, int sr, int ch, int ba,
uint32_t tag, int bits_per_coded_sample, int64_t bitrate,
uint8_t * extradata, int frame_size, int frame_bytes)
{
int bps = av_get_exact_bits_per_sample(id);
int framecount = (ba > 0 && frame_bytes / ba > 0) ? frame_bytes / ba : 1;
/* codecs with an exact constant bits per sample */
if (bps > 0 && ch > 0 && frame_bytes > 0 && ch < 32768 && bps < 32768)
return (frame_bytes * 8LL) / (bps * ch);
bps = bits_per_coded_sample;
/* codecs with a fixed packet duration */
switch (id) {
case AV_CODEC_ID_ADPCM_ADX: return 32;
case AV_CODEC_ID_ADPCM_IMA_QT: return 64;
case AV_CODEC_ID_ADPCM_EA_XAS: return 128;
case AV_CODEC_ID_AMR_NB:
case AV_CODEC_ID_EVRC:
case AV_CODEC_ID_GSM:
case AV_CODEC_ID_QCELP:
case AV_CODEC_ID_RA_288: return 160;
case AV_CODEC_ID_AMR_WB:
case AV_CODEC_ID_GSM_MS: return 320;
case AV_CODEC_ID_MP1: return 384;
case AV_CODEC_ID_ATRAC1: return 512;
case AV_CODEC_ID_ATRAC9:
case AV_CODEC_ID_ATRAC3:
if (framecount > INT_MAX/1024)
return 0;
return 1024 * framecount;
case AV_CODEC_ID_ATRAC3P: return 2048;
case AV_CODEC_ID_MP2:
case AV_CODEC_ID_MUSEPACK7: return 1152;
case AV_CODEC_ID_AC3: return 1536;
case AV_CODEC_ID_FTR: return 1024;
}
if (sr > 0) {
/* calc from sample rate */
if (id == AV_CODEC_ID_TTA)
return 256ll * sr / 245;
else if (id == AV_CODEC_ID_DST)
return 588ll * sr / 44100;
else if (id == AV_CODEC_ID_BINKAUDIO_DCT) {
if (sr / 22050 > 22)
return 0;
return (480 << (sr / 22050));
}
if (id == AV_CODEC_ID_MP3)
return sr <= 24000 ? 576 : 1152;
}
if (ba > 0) {
/* calc from block_align */
if (id == AV_CODEC_ID_SIPR) {
switch (ba) {
case 20: return 160;
case 19: return 144;
case 29: return 288;
case 37: return 480;
}
} else if (id == AV_CODEC_ID_ILBC) {
switch (ba) {
case 38: return 160;
case 50: return 240;
}
}
}
if (frame_bytes > 0) {
/* calc from frame_bytes only */
if (id == AV_CODEC_ID_TRUESPEECH)
return 240 * (frame_bytes / 32);
if (id == AV_CODEC_ID_NELLYMOSER)
return 256 * (frame_bytes / 64);
if (id == AV_CODEC_ID_RA_144)
return 160 * (frame_bytes / 20);
if (id == AV_CODEC_ID_APTX)
return 4 * (frame_bytes / 4);
if (id == AV_CODEC_ID_APTX_HD)
return 4 * (frame_bytes / 6);
if (bps > 0) {
/* calc from frame_bytes and bits_per_coded_sample */
if (id == AV_CODEC_ID_ADPCM_G726 || id == AV_CODEC_ID_ADPCM_G726LE)
return frame_bytes * 8 / bps;
}
if (ch > 0 && ch < INT_MAX/16) {
/* calc from frame_bytes and channels */
switch (id) {
case AV_CODEC_ID_FASTAUDIO:
return frame_bytes / (40 * ch) * 256;
case AV_CODEC_ID_ADPCM_IMA_MOFLEX:
return (frame_bytes - 4 * ch) / (128 * ch) * 256;
case AV_CODEC_ID_ADPCM_AFC:
return frame_bytes / (9 * ch) * 16;
case AV_CODEC_ID_ADPCM_PSX:
case AV_CODEC_ID_ADPCM_DTK:
frame_bytes /= 16 * ch;
if (frame_bytes > INT_MAX / 28)
return 0;
return frame_bytes * 28;
case AV_CODEC_ID_ADPCM_4XM:
case AV_CODEC_ID_ADPCM_IMA_ACORN:
case AV_CODEC_ID_ADPCM_IMA_DAT4:
case AV_CODEC_ID_ADPCM_IMA_ISS:
return (frame_bytes - 4 * ch) * 2 / ch;
case AV_CODEC_ID_ADPCM_IMA_SMJPEG:
return (frame_bytes - 4) * 2 / ch;
case AV_CODEC_ID_ADPCM_IMA_AMV:
return (frame_bytes - 8) * 2;
case AV_CODEC_ID_ADPCM_THP:
case AV_CODEC_ID_ADPCM_THP_LE:
if (extradata)
return frame_bytes * 14LL / (8 * ch);
break;
case AV_CODEC_ID_ADPCM_XA:
return (frame_bytes / 128) * 224 / ch;
case AV_CODEC_ID_INTERPLAY_DPCM:
return (frame_bytes - 6 - ch) / ch;
case AV_CODEC_ID_ROQ_DPCM:
return (frame_bytes - 8) / ch;
case AV_CODEC_ID_XAN_DPCM:
return (frame_bytes - 2 * ch) / ch;
case AV_CODEC_ID_MACE3:
return 3 * frame_bytes / ch;
case AV_CODEC_ID_MACE6:
return 6 * frame_bytes / ch;
case AV_CODEC_ID_PCM_LXF:
return 2 * (frame_bytes / (5 * ch));
case AV_CODEC_ID_IAC:
case AV_CODEC_ID_IMC:
return 4 * frame_bytes / ch;
}
if (tag) {
/* calc from frame_bytes, channels, and codec_tag */
if (id == AV_CODEC_ID_SOL_DPCM) {
if (tag == 3)
return frame_bytes / ch;
else
return frame_bytes * 2 / ch;
}
}
if (ba > 0) {
/* calc from frame_bytes, channels, and block_align */
int blocks = frame_bytes / ba;
int64_t tmp = 0;
switch (id) {
case AV_CODEC_ID_ADPCM_IMA_WAV:
if (bps < 2 || bps > 5)
return 0;
tmp = blocks * (1LL + (ba - 4 * ch) / (bps * ch) * 8);
break;
case AV_CODEC_ID_ADPCM_IMA_DK3:
tmp = blocks * (((ba - 16LL) * 2 / 3 * 4) / ch);
break;
case AV_CODEC_ID_ADPCM_IMA_DK4:
tmp = blocks * (1 + (ba - 4LL * ch) * 2 / ch);
break;
case AV_CODEC_ID_ADPCM_IMA_RAD:
tmp = blocks * ((ba - 4LL * ch) * 2 / ch);
break;
case AV_CODEC_ID_ADPCM_MS:
tmp = blocks * (2 + (ba - 7LL * ch) * 2LL / ch);
break;
case AV_CODEC_ID_ADPCM_MTAF:
tmp = blocks * (ba - 16LL) * 2 / ch;
break;
case AV_CODEC_ID_ADPCM_XMD:
tmp = blocks * 32;
break;
}
if (tmp) {
if (tmp != (int)tmp)
return 0;
return tmp;
}
}
if (bps > 0) {
/* calc from frame_bytes, channels, and bits_per_coded_sample */
switch (id) {
case AV_CODEC_ID_PCM_DVD:
if(bps<4 || frame_bytes<3)
return 0;
return 2 * ((frame_bytes - 3) / ((bps * 2 / 8) * ch));
case AV_CODEC_ID_PCM_BLURAY:
if(bps<4 || frame_bytes<4)
return 0;
return (frame_bytes - 4) / ((FFALIGN(ch, 2) * bps) / 8);
case AV_CODEC_ID_S302M:
return 2 * (frame_bytes / ((bps + 4) / 4)) / ch;
}
}
}
}
/* Fall back on using frame_size */
if (frame_size > 1 && frame_bytes)
return frame_size;
//For WMA we currently have no other means to calculate duration thus we
//do it here by assuming CBR, which is true for all known cases.
if (bitrate > 0 && frame_bytes > 0 && sr > 0 && ba > 1) {
if (id == AV_CODEC_ID_WMAV1 || id == AV_CODEC_ID_WMAV2)
return (frame_bytes * 8LL * sr) / bitrate;
}
return 0;
}
该函数的作用是:计算某个音频packet占用的时间值,也就是AVPacket结构体中的成员变量duration的值。注意:该时间值以AVStream的time_base为单位,还需要额外跟time_base进行运算,才能得到以秒为单位的时间值。
形参id:输入型参数。编解码器唯一标识符,指明码流对应的标准。
形参sr:输入型参数。音频的采样频率。
形参ch:输入型参数。声道数量。
形参ba:输入型参数。“区块对齐”,即每个采样点所需的字节数。
形参tag:输入型参数。关于编解码器的附加信息,一般可忽略。
形参bits_per_coded_sample:输入型参数。音频的采样位数。
形参bitrate:输入型参数。音频的码率,单位为bits/s。
形参extradata:输入型参数。初始化解码器所需的额外二进制数据,依赖于编解码器,一般可忽略。
形参frame_size:输入型参数。一般可忽略。
形参frame_bytes:输入型参数。该音频packet的大小,单位为字节。
返回值:该音频packet占用的时间值。
(二)get_audio_frame_duration函数的内部实现分析
大部分情况下get_audio_frame_duration函数可以化简为:
static int get_audio_frame_duration(enum AVCodecID id, int sr, int ch, int ba,
uint32_t tag, int bits_per_coded_sample, int64_t bitrate,
uint8_t * extradata, int frame_size, int frame_bytes)
{
int bps = av_get_exact_bits_per_sample(id);
/* codecs with an exact constant bits per sample */
if (bps > 0 && ch > 0 && frame_bytes > 0 && ch < 32768 && bps < 32768)
return (frame_bytes * 8LL) / (bps * ch);
}
get_audio_frame_duration函数中,在一开始不会直接用形参bits_per_coded_sample存贮的采样位数,而是会先通过av_get_exact_bits_per_sample函数得到音频的采样位数:
int bps = av_get_exact_bits_per_sample(id);
然后判断音频采样位数是否在0到32768之内、声道数量是否在0到32768之内、packet的大小是否大于0。如果都满足条件,返回 packet的大小(单位为字节)×8÷(音频的采样位数×声道数量)的结果,该结果就是该音频packet占用的时间值。比如,如果packet的大小为16384字节、音频的采样位数为16位、声道数为2,则该音频packet占用的时间值为:16384×8÷(16×2)= 4096:
/* codecs with an exact constant bits per sample */
if (bps > 0 && ch > 0 && frame_bytes > 0 && ch < 32768 && bps < 32768)
return (frame_bytes * 8LL) / (bps * ch);
二、av_get_audio_frame_duration2函数
av_get_audio_frame_duration2函数定义在源文件libavcodec/utils.c中:
int av_get_audio_frame_duration2(AVCodecParameters *par, int frame_bytes)
{
int channels = par->ch_layout.nb_channels;
int duration;
duration = get_audio_frame_duration(par->codec_id, par->sample_rate,
channels, par->block_align,
par->codec_tag, par->bits_per_coded_sample,
par->bit_rate, par->extradata, par->frame_size,
frame_bytes);
return FFMAX(0, duration);
}
该函数的作用是:计算某个音频packet占用的时间值,也就是AVPacket结构体中的成员变量duration的值。注意:该时间值以AVStream的time_base为单位,还需要额外跟time_base进行运算,才能得到以秒为单位的时间值。
形参par:输入型参数。指向一个AVCodecParameters(保存音视频流的基本参数信息)的结构体。
形参frame_bytes:输入型参数。该音频packet的大小,单位为字节。
返回值:该音频packet占用的时间值。
可以看到av_get_audio_frame_duration2函数内部就是通过调用get_audio_frame_duration函数来计算某个音频packet占用的时间值。