【FFmpeg】avformat_find_stream_info
- 1.avformat_find_stream_info
- 1.1 初始化解析器（av_parser_init）
- 1.2 查找探测解码器（find_probe_decoder）
- 1.3 尝试打开解码器（avcodec_open2）
- 1.4 读取帧（read_frame_internal）
- 1.4.1 读取packet（ff_read_packet）
- 1.4.2 解析packet（parse_packet）
- 1.5 尝试解码一帧（try_decode_frame）
- 1.6 检查编码器参数（has_codec_parameters）
- 1.7 估算时长（estimate_timings）
- 2.小结
参考：
FFmpeg源代码简单分析：avformat_find_stream_info()
示例工程：
【FFmpeg】调用ffmpeg库实现264软编
【FFmpeg】调用ffmpeg库实现264软解
【FFmpeg】调用ffmpeg库进行RTMP推流和拉流
【FFmpeg】调用ffmpeg库进行SDL2解码后渲染
流程分析：
【FFmpeg】编码链路上主要函数的简单分析
【FFmpeg】解码链路上主要函数的简单分析
函数分析：
【FFmpeg】avformat_open_input函数
FFmpeg版本为7.0，avformat_find_stream_info内部函数的调用关系如下
1.avformat_find_stream_info
函数的主要功能是根据前面获取的AVFormatContext，结合输入的options，分析每个流（AVStream）的信息，主要内容包括：
（1）确定分析的最大时长
（2）对每个流当中的codec信息的获取
（2.1）初始化解析器（av_parser_init）
（2.2）查找探测解码器（find_probe_decoder）
（2.3）尝试打开解码器（avcodec_open2）
（3）frame信息的获取
（3.1）编解码器的检查和一些额外信息的检查
（3.2）读取帧（read_frame_internal）
（3.3）尝试解码一帧（try_decode_frame）
（4）到达文件末尾检查
（5）计算流当中时间相关信息
（6）结尾处理
这个函数对AVStream当中很多的变量进行了处理，比较复杂，只能做主体的记录，很多变量的含义不是很理解。我对这里的理解是，一个流当中应该主要考虑的问题包括：（1）流格式：这个在avformat_open_input当中已经获取了；（2）时间：流持续的时间（duration），起始时间（start_time），FPS，DTS等等；（3）数据内容：pix_fmt，width等。在这些复杂的流程之中，最重要的一个环节是解码流程，这里实际上进行了一帧的解码，才能够去获取帧级别的信息
int avformat_find_stream_info(AVFormatContext *ic, AVDictionary **options)
{
FFFormatContext *const si = ffformatcontext(ic);
int count = 0, ret = 0;
int64_t read_size;
AVPacket *pkt1 = si->pkt;
int64_t old_offset = avio_tell(ic->pb);
// new streams might appear, no options for those
int orig_nb_streams = ic->nb_streams;
int flush_codecs;
int64_t max_analyze_duration = ic->max_analyze_duration;
int64_t max_stream_analyze_duration;
int64_t max_subtitle_analyze_duration;
int64_t probesize = ic->probesize;
int eof_reached = 0;
int *missing_streams = av_opt_ptr(ic->iformat->priv_class, ic->priv_data, "missing_streams");
flush_codecs = probesize > 0;
av_opt_set_int(ic, "skip_clear", 1, AV_OPT_SEARCH_CHILDREN);
// ----- 1.确定分析最大时长 ----- //
max_stream_analyze_duration = max_analyze_duration;
max_subtitle_analyze_duration = max_analyze_duration;
// 计算最大的分析时长
// 字幕：30 * AV_TIME_BASE
// FLV：90 * AV_TIME_BASE
// MPEG：7 * AV_TIME_BASE
if (!max_analyze_duration) {
max_stream_analyze_duration =
max_analyze_duration = 5*AV_TIME_BASE;
max_subtitle_analyze_duration = 30*AV_TIME_BASE;
if (!strcmp(ic->iformat->name, "flv"))
max_stream_analyze_duration = 90*AV_TIME_BASE;
if (!strcmp(ic->iformat->name, "mpeg") || !strcmp(ic->iformat->name, "mpegts"))
max_stream_analyze_duration = 7*AV_TIME_BASE;
}
// 如果进行了probe，打印偏移量
if (ic->pb) {
FFIOContext *const ctx = ffiocontext(ic->pb);
av_log(ic, AV_LOG_DEBUG, "Before avformat_find_stream_info() pos: %"PRId64" bytes read:%"PRId64" seeks:%d nb_streams:%d\n",
avio_tell(ic->pb), ctx->bytes_read, ctx->seek_count, ic->nb_streams);
}
// 通常情况下，在调用avformat_find_stream_info之前，nb_stream为0
// 在avformat_find_stream_info中会获取nb_stream信息
// ----- 2.对每个流当中的codec信息的获取 ----- //
for (unsigned i = 0; i < ic->nb_streams; i++) {
const AVCodec *codec;
AVDictionary *thread_opt = NULL;
AVStream *const st = ic->streams[i];
FFStream *const sti = ffstream(st);
AVCodecContext *const avctx = sti->avctx;
/* check if the caller has overridden the codec id */
// only for the split stuff
/* 检查调用者是否已覆盖编解码器id */
if (!sti->parser && !(ic->flags & AVFMT_FLAG_NOPARSE) && sti->request_probe <= 0) {
// 初始化解析器
sti->parser = av_parser_init(st->codecpar->codec_id);
if (sti->parser) {
if (sti->need_parsing == AVSTREAM_PARSE_HEADERS) {
sti->parser->flags |= PARSER_FLAG_COMPLETE_FRAMES;
} else if (sti->need_parsing == AVSTREAM_PARSE_FULL_RAW) {
sti->parser->flags |= PARSER_FLAG_USE_CODEC_TS;
}
} else if (sti->need_parsing) {
av_log(ic, AV_LOG_VERBOSE, "parser not found for codec "
"%s, packets or times may be invalid.\n",
avcodec_get_name(st->codecpar->codec_id));
}
}
// 将codecpar赋值给avctx
ret = avcodec_parameters_to_context(avctx, st->codecpar);
if (ret < 0)
goto find_stream_info_err;
if (sti->request_probe <= 0)
sti->avctx_inited = 1;
// 查找并探测解码器
codec = find_probe_decoder(ic, st, st->codecpar->codec_id);
/* Force thread count to 1 since the H.264 decoder will not extract
* SPS and PPS to extradata during multi-threaded decoding. */
// 强制线程数为1，因为H.264解码器在多线程解码期间不会将SPS和PPS提取为额外数据
av_dict_set(options ? &options[i] : &thread_opt, "threads", "1", 0);
/* Force lowres to 0. The decoder might reduce the video size by the
* lowres factor, and we don't want that propagated to the stream's
* codecpar */
// 将lowres强制为0。解码器可能会减少视频大小的低分辨率因素，不希望传播到流的编解码器
av_dict_set(options ? &options[i] : &thread_opt, "lowres", "0", 0);
// 白名单的设置
if (ic->codec_whitelist)
av_dict_set(options ? &options[i] : &thread_opt, "codec_whitelist", ic->codec_whitelist, 0);
// Try to just open decoders, in case this is enough to get parameters.
// Also ensure that subtitle_header is properly set.
// 尝试只打开解码器部分，如果这足以获取所需的参数，那么就进行这个操作
// 同时也要保证字幕头信息被合理设置
if (!has_codec_parameters(st, NULL) && sti->request_probe <= 0 ||
st->codecpar->codec_type == AVMEDIA_TYPE_SUBTITLE) {
if (codec && !avctx->codec)
// 尝试打开codec
if (avcodec_open2(avctx, codec, options ? &options[i] : &thread_opt) < 0)
av_log(ic, AV_LOG_WARNING,
"Failed to open codec in %s\n", __func__);
}
if (!options)
av_dict_free(&thread_opt);
}
read_size = 0;
// ----- 3.frame信息的获取 ----- //
for (;;) {
const AVPacket *pkt;
AVStream *st;
FFStream *sti;
AVCodecContext *avctx;
int analyzed_all_streams;
unsigned i;
// 检查用户是否请求中断与cb相关的阻塞函数
if (ff_check_interrupt(&ic->interrupt_callback)) {
ret = AVERROR_EXIT;
av_log(ic, AV_LOG_DEBUG, "interrupted\n");
break;
}
/* check if one codec still needs to be handled */
// 检查编解码器是否仍然需要处理
for (i = 0; i < ic->nb_streams; i++) {
AVStream *const st = ic->streams[i];
FFStream *const sti = ffstream(st);
int fps_analyze_framecount = 20;
int count;
// 如果没有codec的参数，则退出循环
if (!has_codec_parameters(st, NULL))
break;
/* If the timebase is coarse (like the usual millisecond precision
* of mkv), we need to analyze more frames to reliably arrive at
* the correct fps. */
// 如果时间基准很粗略（比如mkv的一般毫秒级精度），为了确保准确地确定帧率，我们需要分析更多的帧
// 换句话说，为了得到稳定的帧速率测量，当视频的时间精度较低时，必须对更多的帧进行分析和处理
if (av_q2d(st->time_base) > 0.0005) // av_q2d: 将AVRational转换成double类型
fps_analyze_framecount *= 2;
// 1. 时间基础不可靠吗?
// 2. 这是一种启发式方法，可以在快速接受报头中的值与进行一些额外检查之间取得平衡。
// 3. 老的DivX和Xvid通常有一些毫无意义的时间基础，比如1fps或2fps。MPEG-2通常滥用字段重复标志来存储不同的帧率
// 还有一些“可变”fps文件也需要检测
if (!tb_unreliable(ic, st)) //
fps_analyze_framecount = 0;
if (ic->fps_probe_size >= 0)
fps_analyze_framecount = ic->fps_probe_size;
// #define AV_DISPOSITION_ATTACHED_PIC (1 << 10)
// 流作为attached picture/“cover art”(例如ID3v2中的APIC帧)存储在文件中。除非发生查找，否则将在从文件读取的前几个
// 包中返回与它关联的第一个(通常是唯一的)包。它也可以在avstream.attachd_pic中随时访问
if (st->disposition & AV_DISPOSITION_ATTACHED_PIC)
fps_analyze_framecount = 0;
/* variable fps and no guess at the real fps */
// 可变的FPS，不猜测真实的FPS
count = (ic->iformat->flags & AVFMT_NOTIMESTAMPS) ?
sti->info->codec_info_duration_fields/2 :
sti->info->duration_count;
if (!(st->r_frame_rate.num && st->avg_frame_rate.num) &&
st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
if (count < fps_analyze_framecount)
break;
}
// Look at the first 3 frames if there is evidence of frame delay
// but the decoder delay is not set.
// 看看前3帧，如果有帧延迟的证据，但解码器延迟没有设置
if (sti->info->frame_delay_evidence && count < 2 && sti->avctx->has_b_frames == 0)
break;
if (!sti->avctx->extradata &&
(!sti->extract_extradata.inited || sti->extract_extradata.bsf) &&
extract_extradata_check(st)) // 额外数据的检查
break;
if (sti->first_dts == AV_NOPTS_VALUE &&
(!(ic->iformat->flags & AVFMT_NOTIMESTAMPS) || sti->need_parsing == AVSTREAM_PARSE_FULL_RAW) &&
sti->codec_info_nb_frames < ((st->disposition & AV_DISPOSITION_ATTACHED_PIC) ? 1 : ic->max_ts_probe) &&
(st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO ||
st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO))
break;
}
analyzed_all_streams = 0;
// missing_streams的赋值在别的地方，例如对于FLV格式，missing_streams的处理方式为
// flv->missing_streams = flags & (FLV_HEADER_FLAG_HASVIDEO | FLV_HEADER_FLAG_HASAUDIO);
// 如果没有丢失的流信息，
if (!missing_streams || !*missing_streams)
if (i == ic->nb_streams) {
analyzed_all_streams = 1;
/* NOTE: If the format has no header, then we need to read some
* packets to get most of the streams, so we cannot stop here. */
// 如果格式没有头，那么我们需要读取一些数据包来获取大部分流，所以我们不能在这里停止
if (!(ic->ctx_flags & AVFMTCTX_NOHEADER)) {
/* If we found the info for all the codecs, we can stop. */
// 如果找到所有的codecs，可以停止
ret = count;
av_log(ic, AV_LOG_DEBUG, "All info found\n");
flush_codecs = 0;
break;
}
}
/* We did not get all the codec info, but we read too much data. */
// 没有得到所有的编解码器信息，但读取了太多的数据
if (read_size >= probesize) {
ret = count;
av_log(ic, AV_LOG_DEBUG,
"Probe buffer size limit of %"PRId64" bytes reached\n", probesize);
for (unsigned i = 0; i < ic->nb_streams; i++) {
AVStream *const st = ic->streams[i];
FFStream *const sti = ffstream(st);
if (!st->r_frame_rate.num &&
sti->info->duration_count <= 1 &&
st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO &&
strcmp(ic->iformat->name, "image2"))
av_log(ic, AV_LOG_WARNING,
"Stream #%d: not enough frames to estimate rate; "
"consider increasing probesize\n", i);
}
break;
}
/* NOTE: A new stream can be added there if no header in file
* (AVFMTCTX_NOHEADER). */
// 如果文件中没有头文件，可以添加一个新的流(AVFMTCTX_NOHEADER)
// 读取帧的信息
ret = read_frame_internal(ic, pkt1);
if (ret == AVERROR(EAGAIN))
continue;
if (ret < 0) {
/* EOF or error*/
eof_reached = 1;
break;
}
if (!(ic->flags & AVFMT_FLAG_NOBUFFER)) {
// 将AVPacket加入到list当中
ret = avpriv_packet_list_put(&si->packet_buffer,
pkt1, NULL, 0);
if (ret < 0)
goto unref_then_goto_end;
pkt = &si->packet_buffer.tail->pkt;
} else {
pkt = pkt1;
}
st = ic->streams[pkt->stream_index];
sti = ffstream(st);
if (!(st->disposition & AV_DISPOSITION_ATTACHED_PIC))
read_size += pkt->size;
avctx = sti->avctx;
if (!sti->avctx_inited) {
ret = avcodec_parameters_to_context(avctx, st->codecpar);
if (ret < 0)
goto unref_then_goto_end;
sti->avctx_inited = 1;
}
if (pkt->dts != AV_NOPTS_VALUE && sti->codec_info_nb_frames > 1) {
/* check for non-increasing dts */
if (sti->info->fps_last_dts != AV_NOPTS_VALUE &&
sti->info->fps_last_dts >= pkt->dts) {
av_log(ic, AV_LOG_DEBUG,
"Non-increasing DTS in stream %d: packet %d with DTS "
"%"PRId64", packet %d with DTS %"PRId64"\n",
st->index, sti->info->fps_last_dts_idx,
sti->info->fps_last_dts, sti->codec_info_nb_frames,
pkt->dts);
sti->info->fps_first_dts =
sti->info->fps_last_dts = AV_NOPTS_VALUE;
}
/* Check for a discontinuity in dts. If the difference in dts
* is more than 1000 times the average packet duration in the
* sequence, we treat it as a discontinuity. */
// 检查dts中的不连续。如果dts的差异超过序列中平均数据包持续时间的1000倍，我们将其视为不连续
if (sti->info->fps_last_dts != AV_NOPTS_VALUE &&
sti->info->fps_last_dts_idx > sti->info->fps_first_dts_idx &&
(pkt->dts - (uint64_t)sti->info->fps_last_dts) / 1000 >
(sti->info->fps_last_dts - (uint64_t)sti->info->fps_first_dts) /
(sti->info->fps_last_dts_idx - sti->info->fps_first_dts_idx)) {
av_log(ic, AV_LOG_WARNING,
"DTS discontinuity in stream %d: packet %d with DTS "
"%"PRId64", packet %d with DTS %"PRId64"\n",
st->index, sti->info->fps_last_dts_idx,
sti->info->fps_last_dts, sti->codec_info_nb_frames,
pkt->dts);
sti->info->fps_first_dts =
sti->info->fps_last_dts = AV_NOPTS_VALUE;
}
/* update stored dts values */
// 更新存储的DTS值
if (sti->info->fps_first_dts == AV_NOPTS_VALUE) {
sti->info->fps_first_dts = pkt->dts;
sti->info->fps_first_dts_idx = sti->codec_info_nb_frames;
}
sti->info->fps_last_dts = pkt->dts;
sti->info->fps_last_dts_idx = sti->codec_info_nb_frames;
}
if (sti->codec_info_nb_frames > 1) {
int64_t t = 0;
int64_t limit;
if (st->time_base.den > 0)
t = av_rescale_q(sti->info->codec_info_duration, st->time_base, AV_TIME_BASE_Q);
if (st->avg_frame_rate.num > 0)
t = FFMAX(t, av_rescale_q(sti->codec_info_nb_frames, av_inv_q(st->avg_frame_rate), AV_TIME_BASE_Q));
if ( t == 0
&& sti->codec_info_nb_frames > 30
&& sti->info->fps_first_dts != AV_NOPTS_VALUE
&& sti->info->fps_last_dts != AV_NOPTS_VALUE) {
int64_t dur = av_sat_sub64(sti->info->fps_last_dts, sti->info->fps_first_dts);
t = FFMAX(t, av_rescale_q(dur, st->time_base, AV_TIME_BASE_Q));
}
if (analyzed_all_streams) limit = max_analyze_duration;
else if (avctx->codec_type == AVMEDIA_TYPE_SUBTITLE) limit = max_subtitle_analyze_duration;
else limit = max_stream_analyze_duration;
if (t >= limit) {
av_log(ic, AV_LOG_VERBOSE, "max_analyze_duration %"PRId64" reached at %"PRId64" microseconds st:%d\n",
limit,
t, pkt->stream_index);
if (ic->flags & AVFMT_FLAG_NOBUFFER)
av_packet_unref(pkt1);
break;
}
if (pkt->duration > 0) {
const int fields = sti->codec_desc && (sti->codec_desc->props & AV_CODEC_PROP_FIELDS);
if (avctx->codec_type == AVMEDIA_TYPE_SUBTITLE && pkt->pts != AV_NOPTS_VALUE && st->start_time != AV_NOPTS_VALUE && pkt->pts >= st->start_time
&& (uint64_t)pkt->pts - st->start_time < INT64_MAX
) {
sti->info->codec_info_duration = FFMIN(pkt->pts - st->start_time, sti->info->codec_info_duration + pkt->duration);
} else
sti->info->codec_info_duration += pkt->duration;
sti->info->codec_info_duration_fields += sti->parser && sti->need_parsing && fields
? sti->parser->repeat_pict + 1 : 2;
}
}
if (st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
#if FF_API_R_FRAME_RATE
ff_rfps_add_frame(ic, st, pkt->dts);
#endif
if (pkt->dts != pkt->pts && pkt->dts != AV_NOPTS_VALUE && pkt->pts != AV_NOPTS_VALUE)
sti->info->frame_delay_evidence = 1;
}
if (!sti->avctx->extradata) {
ret = extract_extradata(si, st, pkt);
if (ret < 0)
goto unref_then_goto_end;
}
/* If still no information, we try to open the codec and to
* decompress the frame. We try to avoid that in most cases as
* it takes longer and uses more memory. For MPEG-4, we need to
* decompress for QuickTime.
*
* If AV_CODEC_CAP_CHANNEL_CONF is set this will force decoding of at
* least one frame of codec data, this makes sure the codec initializes
* the channel configuration and does not only trust the values from
* the container. */
// 1.如果仍然没有信息，尝试打开编解码器并解压缩帧。尽量避免在大多数情况下使用，因为它需要更长的时间和使用更多的内存
// 对于MPEG-4，需要为QuickTime解压
// 2.如果设置了AV_CODEC_CAP_CHANNEL_CONF，这将强制解码至少一帧编解码器数据，这确保编解码器初始化通道配置，
// 并且不仅信任来自容器的值
try_decode_frame(ic, st, pkt,
(options && i < orig_nb_streams) ? &options[i] : NULL);
if (ic->flags & AVFMT_FLAG_NOBUFFER)
av_packet_unref(pkt1);
sti->codec_info_nb_frames++;
count++;
}
// ----- 4.到达文件末尾检查 ----- //
if (eof_reached) {
for (unsigned stream_index = 0; stream_index < ic->nb_streams; stream_index++) {
AVStream *const st = ic->streams[stream_index];
AVCodecContext *const avctx = ffstream(st)->avctx;
if (!has_codec_parameters(st, NULL)) {
const AVCodec *codec = find_probe_decoder(ic, st, st->codecpar->codec_id);
if (codec && !avctx->codec) {
AVDictionary *opts = NULL;
if (ic->codec_whitelist)
av_dict_set(&opts, "codec_whitelist", ic->codec_whitelist, 0);
if (avcodec_open2(avctx, codec, (options && stream_index < orig_nb_streams) ? &options[stream_index] : &opts) < 0)
av_log(ic, AV_LOG_WARNING,
"Failed to open codec in %s\n", __func__);
av_dict_free(&opts);
}
}
// EOF already reached while reading the stream above.
// So continue with reoordering DTS with whatever delay we have.
// 在读取上面的流时已经达到EOF。所以继续重新排序DTS不管有什么延迟
if (si->packet_buffer.head && !has_decode_delay_been_guessed(st)) {
update_dts_from_pts(ic, stream_index, si->packet_buffer.head);
}
}
}
// 清理decoders
if (flush_codecs) {
AVPacket *empty_pkt = si->pkt;
int err = 0;
av_packet_unref(empty_pkt);
for (unsigned i = 0; i < ic->nb_streams; i++) {
AVStream *const st = ic->streams[i];
FFStream *const sti = ffstream(st);
/* flush the decoders */
if (sti->info->found_decoder == 1) {
err = try_decode_frame(ic, st, empty_pkt,
(options && i < orig_nb_streams)
? &options[i] : NULL);
if (err < 0) {
av_log(ic, AV_LOG_INFO,
"decoding for stream %d failed\n", st->index);
}
}
}
}
// ----- 5.计算流当中时间相关信息 ----- //
// 计算帧率
ff_rfps_calculate(ic);
for (unsigned i = 0; i < ic->nb_streams; i++) {
AVStream *const st = ic->streams[i];
FFStream *const sti = ffstream(st);
AVCodecContext *const avctx = sti->avctx;
if (avctx->codec_type == AVMEDIA_TYPE_VIDEO) {
if (avctx->codec_id == AV_CODEC_ID_RAWVIDEO && !avctx->codec_tag && !avctx->bits_per_coded_sample) {
// 返回一个表示与像素格式pix_fmt相关联的fourCC代码的值，如果找不到相关的fourCC代码，则返回0
uint32_t tag= avcodec_pix_fmt_to_codec_tag(avctx->pix_fmt);
// 查找pixel的fmt
if (avpriv_pix_fmt_find(PIX_FMT_LIST_RAW, tag) == avctx->pix_fmt)
avctx->codec_tag= tag;
}
/* estimate average framerate if not set by demuxer */
// 估计平均帧率，如果没有被demuxer所设置
if (sti->info->codec_info_duration_fields &&
!st->avg_frame_rate.num &&
sti->info->codec_info_duration) {
int best_fps = 0;
double best_error = 0.01;
AVRational codec_frame_rate = avctx->framerate;
if (sti->info->codec_info_duration >= INT64_MAX / st->time_base.num / 2||
sti->info->codec_info_duration_fields >= INT64_MAX / st->time_base.den ||
sti->info->codec_info_duration < 0)
continue;
av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
sti->info->codec_info_duration_fields * (int64_t) st->time_base.den,
sti->info->codec_info_duration * 2 * (int64_t) st->time_base.num, 60000);
/* Round guessed framerate to a "standard" framerate if it's
* within 1% of the original estimate. */
// 如果猜测的帧率在原始估计的1%以内，则将其舍入为“标准”帧率
for (int j = 0; j < MAX_STD_TIMEBASES; j++) {
AVRational std_fps = { get_std_framerate(j), 12 * 1001 };
double error = fabs(av_q2d(st->avg_frame_rate) /
av_q2d(std_fps) - 1);
if (error < best_error) {
best_error = error;
best_fps = std_fps.num;
}
if (si->prefer_codec_framerate && codec_frame_rate.num > 0 && codec_frame_rate.den > 0) {
error = fabs(av_q2d(codec_frame_rate) /
av_q2d(std_fps) - 1);
if (error < best_error) {
best_error = error;
best_fps = std_fps.num;
}
}
}
if (best_fps)
av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
best_fps, 12 * 1001, INT_MAX);
}
if (!st->r_frame_rate.num) {
const AVCodecDescriptor *desc = sti->codec_desc;
AVRational mul = (AVRational){ desc && (desc->props & AV_CODEC_PROP_FIELDS) ? 2 : 1, 1 };
AVRational fr = av_mul_q(avctx->framerate, mul);
if (fr.num && fr.den && av_cmp_q(st->time_base, av_inv_q(fr)) <= 0) {
st->r_frame_rate = fr;
} else {
st->r_frame_rate.num = st->time_base.den;
st->r_frame_rate.den = st->time_base.num;
}
}
st->codecpar->framerate = avctx->framerate;
if (sti->display_aspect_ratio.num && sti->display_aspect_ratio.den) {
AVRational hw_ratio = { avctx->height, avctx->width };
st->sample_aspect_ratio = av_mul_q(sti->display_aspect_ratio,
hw_ratio);
}
} else if (avctx->codec_type == AVMEDIA_TYPE_AUDIO) { // 音频的检查
if (!avctx->bits_per_coded_sample)
avctx->bits_per_coded_sample =
av_get_bits_per_sample(avctx->codec_id);
// set stream disposition based on audio service type
switch (avctx->audio_service_type) {
case AV_AUDIO_SERVICE_TYPE_EFFECTS:
st->disposition = AV_DISPOSITION_CLEAN_EFFECTS;
break;
case AV_AUDIO_SERVICE_TYPE_VISUALLY_IMPAIRED:
st->disposition = AV_DISPOSITION_VISUAL_IMPAIRED;
break;
case AV_AUDIO_SERVICE_TYPE_HEARING_IMPAIRED:
st->disposition = AV_DISPOSITION_HEARING_IMPAIRED;
break;
case AV_AUDIO_SERVICE_TYPE_COMMENTARY:
st->disposition = AV_DISPOSITION_COMMENT;
break;
case AV_AUDIO_SERVICE_TYPE_KARAOKE:
st->disposition = AV_DISPOSITION_KARAOKE;
break;
}
}
}
// 估算视频流的起始时间，持续时间
if (probesize)
estimate_timings(ic, old_offset);
av_opt_set_int(ic, "skip_clear", 0, AV_OPT_SEARCH_CHILDREN);
// ----- 6.结尾处理 ----- //
if (ret >= 0 && ic->nb_streams)
/* We could not have all the codec parameters before EOF. */
// 我们不可能在EOF之前得到所有的编解码器参数
ret = -1;
for (unsigned i = 0; i < ic->nb_streams; i++) {
AVStream *const st = ic->streams[i];
FFStream *const sti = ffstream(st);
const char *errmsg;
/* if no packet was ever seen, update context now for has_codec_parameters */
// 如果没有看到packet，现在更新上下文为has_codec_parameters
if (!sti->avctx_inited) {
if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO &&
st->codecpar->format == AV_SAMPLE_FMT_NONE)
st->codecpar->format = sti->avctx->sample_fmt;
ret = avcodec_parameters_to_context(sti->avctx, st->codecpar);
if (ret < 0)
goto find_stream_info_err;
}
if (!has_codec_parameters(st, &errmsg)) {
char buf[256];
avcodec_string(buf, sizeof(buf), sti->avctx, 0);
av_log(ic, AV_LOG_WARNING,
"Could not find codec parameters for stream %d (%s): %s\n"
"Consider increasing the value for the 'analyzeduration' (%"PRId64") and 'probesize' (%"PRId64") options\n",
i, buf, errmsg, ic->max_analyze_duration, ic->probesize);
} else {
ret = 0;
}
}
// 计算chapters
ret = compute_chapters_end(ic);
if (ret < 0)
goto find_stream_info_err;
/* update the stream parameters from the internal codec contexts */
// 从内部编解码器上下文更新流参数
for (unsigned i = 0; i < ic->nb_streams; i++) {
AVStream *const st = ic->streams[i];
FFStream *const sti = ffstream(st);
if (sti->avctx_inited) {
ret = avcodec_parameters_from_context(st->codecpar, sti->avctx);
if (ret < 0)
goto find_stream_info_err;
if (sti->avctx->rc_buffer_size > 0 || sti->avctx->rc_max_rate > 0 ||
sti->avctx->rc_min_rate) {
size_t cpb_size;
AVCPBProperties *props = av_cpb_properties_alloc(&cpb_size);
if (props) {
if (sti->avctx->rc_buffer_size > 0)
props->buffer_size = sti->avctx->rc_buffer_size;
if (sti->avctx->rc_min_rate > 0)
props->min_bitrate = sti->avctx->rc_min_rate;
if (sti->avctx->rc_max_rate > 0)
props->max_bitrate = sti->avctx->rc_max_rate;
if (!av_packet_side_data_add(&st->codecpar->coded_side_data,
&st->codecpar->nb_coded_side_data,
AV_PKT_DATA_CPB_PROPERTIES,
(uint8_t *)props, cpb_size, 0))
av_free(props);
}
}
}
sti->avctx_inited = 0;
#if FF_API_AVSTREAM_SIDE_DATA
FF_DISABLE_DEPRECATION_WARNINGS
if (st->codecpar->nb_coded_side_data > 0) {
av_assert0(!st->side_data && !st->nb_side_data);
st->side_data = av_calloc(st->codecpar->nb_coded_side_data, sizeof(*st->side_data));
if (!st->side_data) {
ret = AVERROR(ENOMEM);
goto find_stream_info_err;
}
for (int j = 0; j < st->codecpar->nb_coded_side_data; j++) {
uint8_t *data = av_memdup(st->codecpar->coded_side_data[j].data,
st->codecpar->coded_side_data[j].size);
if (!data) {
ret = AVERROR(ENOMEM);
goto find_stream_info_err;
}
st->side_data[j].type = st->codecpar->coded_side_data[j].type;
st->side_data[j].size = st->codecpar->coded_side_data[j].size;
st->side_data[j].data = data;
st->nb_side_data++;
}
}
FF_ENABLE_DEPRECATION_WARNINGS
#endif
}
find_stream_info_err:
for (unsigned i = 0; i < ic->nb_streams; i++) {
AVStream *const st = ic->streams[i];
FFStream *const sti = ffstream(st);
int err;
if (sti->info) {
av_freep(&sti->info->duration_error);
av_freep(&sti->info);
}
err = codec_close(sti);
if (err < 0 && ret >= 0)
ret = err;
av_bsf_free(&sti->extract_extradata.bsf);
}
if (ic->pb) {
FFIOContext *const ctx = ffiocontext(ic->pb);
av_log(ic, AV_LOG_DEBUG, "After avformat_find_stream_info() pos: %"PRId64" bytes read:%"PRId64" seeks:%d frames:%d\n",
avio_tell(ic->pb), ctx->bytes_read, ctx->seek_count, count);
}
return ret;
unref_then_goto_end:
av_packet_unref(pkt1);
goto find_stream_info_err;
}
1.1 初始化解析器（av_parser_init）
参考【FFmpeg】解码链路上主要函数的简单分析
1.2 查找探测解码器（find_probe_decoder）
find_probe_decoder首先检查是否是H264格式，随后调用ff_find_decoder进行解码器的查找
static const AVCodec *find_probe_decoder(AVFormatContext *s, const AVStream *st, enum AVCodecID codec_id)
{
const AVCodec *codec;
#if CONFIG_H264_DECODER
/* Other parts of the code assume this decoder to be used for h264,
* so force it if possible. */
// 直接返回H264格式
if (codec_id == AV_CODEC_ID_H264)
return avcodec_find_decoder_by_name("h264");
#endif
// 查找解码器
codec = ff_find_decoder(s, st, codec_id);
if (!codec)
return NULL;
// 解码器的能力中包括了避免进行probe，则重新寻找一个支持probe的解码器
if (codec->capabilities & AV_CODEC_CAP_AVOID_PROBING) {
const AVCodec *probe_codec = NULL;
void *iter = NULL;
while ((probe_codec = av_codec_iterate(&iter))) {
if (probe_codec->id == codec->id &&
av_codec_is_decoder(probe_codec) &&
!(probe_codec->capabilities & (AV_CODEC_CAP_AVOID_PROBING | AV_CODEC_CAP_EXPERIMENTAL))) {
return probe_codec;
}
}
}
// 如果实在找不到支持probe的解码器，则返回指定的解码器
return codec;
}
ff_find_decoder的调用方式为
const AVCodec *ff_find_decoder(AVFormatContext *s, const AVStream *st,
enum AVCodecID codec_id)
{
switch (st->codecpar->codec_type) {
case AVMEDIA_TYPE_VIDEO:
if (s->video_codec) return s->video_codec;
break;
case AVMEDIA_TYPE_AUDIO:
if (s->audio_codec) return s->audio_codec;
break;
case AVMEDIA_TYPE_SUBTITLE:
if (s->subtitle_codec) return s->subtitle_codec;
break;
}
return avcodec_find_decoder(codec_id);
}
可以看到，ff_find_decoder根据codec_type来确定decoder的类型，视频、音频或字幕，如果不是这三种类型，再调用avcodec_find_decoder确定codec，avcodec_find_decoder函数在其他文中已经记录过，大致的意思是遍历支持的decoder，查看是否有与codec_id匹配的decoder
1.3 尝试打开解码器（avcodec_open2）
参考【FFmpeg】编码链路上主要函数的简单分析
1.4 读取帧（read_frame_internal）
该函数用于读取一帧，通过调用ff_read_packet来读取packet
static int read_frame_internal(AVFormatContext *s, AVPacket *pkt)
{
FFFormatContext *const si = ffformatcontext(s);
int ret, got_packet = 0;
AVDictionary *metadata = NULL;
while (!got_packet && !si->parse_queue.head) {
AVStream *st;
FFStream *sti;
/* read next packet */
// 读取一个packet，具体会调用.read_packet实现，根据不同格式有所不同
// 例如对于FLV格式，ff_read_packet内部会调用flv_read_packet()读取一个packet
ret = ff_read_packet(s, pkt);
if (ret < 0) {
if (ret == AVERROR(EAGAIN))
return ret;
/* flush the parsers */
for (unsigned i = 0; i < s->nb_streams; i++) {
AVStream *const st = s->streams[i];
FFStream *const sti = ffstream(st);
if (sti->parser && sti->need_parsing)
parse_packet(s, pkt, st->index, 1); // 解析packet
}
/* all remaining packets are now in parse_queue =>
* really terminate parsing */
break;
}
ret = 0;
st = s->streams[pkt->stream_index];
sti = ffstream(st);
st->event_flags |= AVSTREAM_EVENT_FLAG_NEW_PACKETS;
/* update context if required */
// 更新一个context，需要先关闭decoder
if (sti->need_context_update) {
if (avcodec_is_open(sti->avctx)) {
av_log(s, AV_LOG_DEBUG, "Demuxer context update while decoder is open, closing and trying to re-open\n");
ret = codec_close(sti);
sti->info->found_decoder = 0;
if (ret < 0)
return ret;
}
/* close parser, because it depends on the codec */
if (sti->parser && sti->avctx->codec_id != st->codecpar->codec_id) {
av_parser_close(sti->parser);
sti->parser = NULL;
}
// 将codecpar的信息拷贝给avctx
ret = avcodec_parameters_to_context(sti->avctx, st->codecpar);
if (ret < 0) {
av_packet_unref(pkt);
return ret;
}
sti->codec_desc = avcodec_descriptor_get(sti->avctx->codec_id);
sti->need_context_update = 0;
}
// 检查时间戳
if (pkt->pts != AV_NOPTS_VALUE &&
pkt->dts != AV_NOPTS_VALUE &&
pkt->pts < pkt->dts) {
av_log(s, AV_LOG_WARNING,
"Invalid timestamps stream=%d, pts=%s, dts=%s, size=%d\n",
pkt->stream_index,
av_ts2str(pkt->pts),
av_ts2str(pkt->dts),
pkt->size);
}
if (s->debug & FF_FDEBUG_TS)
av_log(s, AV_LOG_DEBUG,
"ff_read_packet stream=%d, pts=%s, dts=%s, size=%d, duration=%"PRId64", flags=%d\n",
pkt->stream_index,
av_ts2str(pkt->pts),
av_ts2str(pkt->dts),
pkt->size, pkt->duration, pkt->flags);
// 重启parser
if (sti->need_parsing && !sti->parser && !(s->flags & AVFMT_FLAG_NOPARSE)) {
sti->parser = av_parser_init(st->codecpar->codec_id);
if (!sti->parser) {
av_log(s, AV_LOG_VERBOSE, "parser not found for codec "
"%s, packets or times may be invalid.\n",
avcodec_get_name(st->codecpar->codec_id));
/* no parser available: just output the raw packets */
sti->need_parsing = AVSTREAM_PARSE_NONE;
} else if (sti->need_parsing == AVSTREAM_PARSE_HEADERS)
sti->parser->flags |= PARSER_FLAG_COMPLETE_FRAMES;
else if (sti->need_parsing == AVSTREAM_PARSE_FULL_ONCE)
sti->parser->flags |= PARSER_FLAG_ONCE;
else if (sti->need_parsing == AVSTREAM_PARSE_FULL_RAW)
sti->parser->flags |= PARSER_FLAG_USE_CODEC_TS;
}
if (!sti->need_parsing || !sti->parser) {
/* no parsing needed: we just output the packet as is */
// 无需解析
compute_pkt_fields(s, st, NULL, pkt, AV_NOPTS_VALUE, AV_NOPTS_VALUE);
if ((s->iformat->flags & AVFMT_GENERIC_INDEX) &&
(pkt->flags & AV_PKT_FLAG_KEY) && pkt->dts != AV_NOPTS_VALUE) {
ff_reduce_index(s, st->index);
av_add_index_entry(st, pkt->pos, pkt->dts,
0, 0, AVINDEX_KEYFRAME);
}
got_packet = 1;
} else if (st->discard < AVDISCARD_ALL) {
if ((ret = parse_packet(s, pkt, pkt->stream_index, 0)) < 0)
return ret;
st->codecpar->sample_rate = sti->avctx->sample_rate;
st->codecpar->bit_rate = sti->avctx->bit_rate;
ret = av_channel_layout_copy(&st->codecpar->ch_layout, &sti->avctx->ch_layout);
if (ret < 0)
return ret;
st->codecpar->codec_id = sti->avctx->codec_id;
} else {
/* free packet */
av_packet_unref(pkt);
}
if (pkt->flags & AV_PKT_FLAG_KEY)
sti->skip_to_keyframe = 0;
if (sti->skip_to_keyframe) {
av_packet_unref(pkt);
got_packet = 0;
}
}
// 删除列表中最老的AVPacket并返回它
if (!got_packet && si->parse_queue.head)
ret = avpriv_packet_list_get(&si->parse_queue, pkt);
if (ret >= 0) {
AVStream *const st = s->streams[pkt->stream_index];
FFStream *const sti = ffstream(st);
int discard_padding = 0;
if (sti->first_discard_sample && pkt->pts != AV_NOPTS_VALUE) {
int64_t pts = pkt->pts - (is_relative(pkt->pts) ? RELATIVE_TS_BASE : 0);
int64_t sample = ts_to_samples(st, pts);
int64_t duration = ts_to_samples(st, pkt->duration);
int64_t end_sample = sample + duration;
if (duration > 0 && end_sample >= sti->first_discard_sample &&
sample < sti->last_discard_sample)
discard_padding = FFMIN(end_sample - sti->first_discard_sample, duration);
}
if (sti->start_skip_samples && (pkt->pts == 0 || pkt->pts == RELATIVE_TS_BASE))
sti->skip_samples = sti->start_skip_samples;
sti->skip_samples = FFMAX(0, sti->skip_samples);
if (sti->skip_samples || discard_padding) {
uint8_t *p = av_packet_new_side_data(pkt, AV_PKT_DATA_SKIP_SAMPLES, 10);
if (p) {
AV_WL32(p, sti->skip_samples);
AV_WL32(p + 4, discard_padding);
av_log(s, AV_LOG_DEBUG, "demuxer injecting skip %u / discard %u\n",
(unsigned)sti->skip_samples, (unsigned)discard_padding);
}
sti->skip_samples = 0;
}
/*
side data和extra data：
（1）side data和extra data都与编码器和解码器有关，但用途和功能有所不同
（2）side data指的是编码过程中生成的辅助信息，对编解码过程可能很重要，对于某些格式，可能包含重建图像所必须的码表或其他数据
（3）extra data指的是编解码过程中的额外信息，不直接参与图像编解码，但对解码过程有帮助，可以是解码器的初始化信息、编码参数等
*/
#if FF_API_AVSTREAM_SIDE_DATA
if (sti->inject_global_side_data) {
for (int i = 0; i < st->codecpar->nb_coded_side_data; i++) {
const AVPacketSideData *const src_sd = &st->codecpar->coded_side_data[i];
uint8_t *dst_data;
if (av_packet_get_side_data(pkt, src_sd->type, NULL))
continue;
dst_data = av_packet_new_side_data(pkt, src_sd->type, src_sd->size);
if (!dst_data) {
av_log(s, AV_LOG_WARNING, "Could not inject global side data\n");
continue;
}
memcpy(dst_data, src_sd->data, src_sd->size);
}
sti->inject_global_side_data = 0;
}
#endif
}
// meta data指的是与媒体文件相关联的描述性信息，如标题、作者、版权等等。这些信息通常不直接影响编解码过程，但有助于描述和组织媒体内容
if (!si->metafree) {
int metaret = av_opt_get_dict_val(s, "metadata", AV_OPT_SEARCH_CHILDREN, &metadata);
if (metadata) {
s->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
av_dict_copy(&s->metadata, metadata, 0);
av_dict_free(&metadata);
av_opt_set_dict_val(s, "metadata", NULL, AV_OPT_SEARCH_CHILDREN);
}
si->metafree = metaret == AVERROR_OPTION_NOT_FOUND;
}
if (s->debug & FF_FDEBUG_TS)
av_log(s, AV_LOG_DEBUG,
"read_frame_internal stream=%d, pts=%s, dts=%s, "
"size=%d, duration=%"PRId64", flags=%d\n",
pkt->stream_index,
av_ts2str(pkt->pts),
av_ts2str(pkt->dts),
pkt->size, pkt->duration, pkt->flags);
/* A demuxer might have returned EOF because of an IO error, let's
* propagate this back to the user. */
if (ret == AVERROR_EOF && s->pb && s->pb->error < 0 && s->pb->error != AVERROR(EAGAIN))
ret = s->pb->error;
return ret;
}
在上面的代码中，进行了很多的检查和赋值，其中最核心的功能为读取packet（ff_read_packet）和解析packet（parse_packet），ff_read_packet实现的方式如下
1.4.1 读取packet（ff_read_packet）
ff_read_packet的主要功能是读取packet，存入AVPacket中。在代码中，最重要的地方为read_packet，具体执行了读取packet的任务。如果是FLV格式，则会调用flv_read_packet进行读取。假设现在编解码过程中刚执行了avformat_input_open，此时nb_stream为0，很多由nb_stream决定的循环不会进入。nb_stream的创建会再flv_read_packet中调用create_stream中实现，这略去部分代码， 因为read_packet函数非常重要，后续在别的文章中详细记录
int ff_read_packet(AVFormatContext *s, AVPacket *pkt)
{
FFFormatContext *const si = ffformatcontext(s);
int err;
#if FF_API_INIT_PACKET
FF_DISABLE_DEPRECATION_WARNINGS
pkt->data = NULL;
pkt->size = 0;
av_init_packet(pkt);
FF_ENABLE_DEPRECATION_WARNINGS
#else
av_packet_unref(pkt);
#endif
for (;;) {
PacketListEntry *pktl = si->raw_packet_buffer.head;
if (pktl) {
AVStream *const st = s->streams[pktl->pkt.stream_index];
if (si->raw_packet_buffer_size >= s->probesize)
if ((err = probe_codec(s, st, NULL)) < 0)
return err;
if (ffstream(st)->request_probe <= 0) {
avpriv_packet_list_get(&si->raw_packet_buffer, pkt);
si->raw_packet_buffer_size -= pkt->size;
return 0;
}
}
// 读取packet
err = ffifmt(s->iformat)->read_packet(s, pkt);
if (err < 0) {
av_packet_unref(pkt);
/* Some demuxers return FFERROR_REDO when they consume
data and discard it (ignored streams, junk, extradata).
We must re-call the demuxer to get the real packet. */
if (err == FFERROR_REDO)
continue;
if (!pktl || err == AVERROR(EAGAIN))
return err;
for (unsigned i = 0; i < s->nb_streams; i++) {
AVStream *const st = s->streams[i];
FFStream *const sti = ffstream(st);
if (sti->probe_packets || sti->request_probe > 0)
if ((err = probe_codec(s, st, NULL)) < 0)
return err;
av_assert0(sti->request_probe <= 0);
}
continue;
}
err = av_packet_make_refcounted(pkt);
if (err < 0) {
av_packet_unref(pkt);
return err;
}
err = handle_new_packet(s, pkt, 1);
if (err <= 0) /* Error or passthrough */
return err;
}
}
以FLV格式为例，会使用flv_read_packet进行packet的读取
static int flv_read_packet(AVFormatContext *s, AVPacket *pkt)
{
// ...
/* now find stream */
for (i = 0; i < s->nb_streams; i++) {
st = s->streams[i];
if (stream_type == FLV_STREAM_TYPE_AUDIO) {
if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO &&
(s->audio_codec_id || flv_same_audio_codec(st->codecpar, flags)))
break;
} else if (stream_type == FLV_STREAM_TYPE_VIDEO) {
if (st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO &&
(s->video_codec_id || flv_same_video_codec(st->codecpar, video_codec_id)))
break;
} else if (stream_type == FLV_STREAM_TYPE_SUBTITLE) {
if (st->codecpar->codec_type == AVMEDIA_TYPE_SUBTITLE)
break;
} else if (stream_type == FLV_STREAM_TYPE_DATA) {
if (st->codecpar->codec_type == AVMEDIA_TYPE_DATA)
break;
}
}
/* 创建新的stream，同时会将nb_stream + 1 */
// 这里也表明了在调用avformat_open_input之后nb_stream为0，但是
// 在调用了avformat_find_stream_info之后nb_stream不为0的原因
if (i == s->nb_streams) {
static const enum AVMediaType stream_types[] = {AVMEDIA_TYPE_VIDEO, AVMEDIA_TYPE_AUDIO, AVMEDIA_TYPE_SUBTITLE, AVMEDIA_TYPE_DATA};
st = create_stream(s, stream_types[stream_type]);
if (!st)
return AVERROR(ENOMEM);
}
// ...
}
1.4.2 解析packet（parse_packet）
解析packet中，最核心是调用了av_parser_parse2函数进行前面packet的解析工作，略去一些代码，在其他文章中记录
/**
* Parse a packet, add all split parts to parse_queue.
*
* @param pkt Packet to parse; must not be NULL.
* @param flush Indicates whether to flush. If set, pkt must be blank.
*/
// 解析一个packet，并且将所有解析拆分的部分添加到parse_queue
static int parse_packet(AVFormatContext *s, AVPacket *pkt,
int stream_index, int flush)
{
// ...
while (size > 0 || (flush && got_output)) {
int64_t next_pts = pkt->pts;
int64_t next_dts = pkt->dts;
int len;
//
len = av_parser_parse2(sti->parser, sti->avctx,
&out_pkt->data, &out_pkt->size, data, size,
pkt->pts, pkt->dts, pkt->pos);
// ...
1.5 尝试解码一帧（try_decode_frame）
如果在前面仍然没有获取到信息，需要尝试打开解码器进行解码，不过需要尽量避免这个操作，因为会带来额外的时间开销。具体来说，会调用avcodec_send_packet将packet送入解码器进行解码，再调用avcodec_receive_frame获取已经解码的frame的信息
/* returns 1 or 0 if or if not decoded data was returned, or a negative error */
static int try_decode_frame(AVFormatContext *s, AVStream *st,
const AVPacket *pkt, AVDictionary **options)
{
FFStream *const sti = ffstream(st);
AVCodecContext *const avctx = sti->avctx;
const AVCodec *codec;
int got_picture = 1, ret = 0;
AVFrame *frame = av_frame_alloc();
AVSubtitle subtitle;
int do_skip_frame = 0;
enum AVDiscard skip_frame;
int pkt_to_send = pkt->size > 0;
if (!frame)
return AVERROR(ENOMEM);
if (!avcodec_is_open(avctx) &&
sti->info->found_decoder <= 0 &&
(st->codecpar->codec_id != -sti->info->found_decoder || !st->codecpar->codec_id)) {
AVDictionary *thread_opt = NULL;
codec = find_probe_decoder(s, st, st->codecpar->codec_id);
if (!codec) {
sti->info->found_decoder = -st->codecpar->codec_id;
ret = -1;
goto fail;
}
/* Force thread count to 1 since the H.264 decoder will not extract
* SPS and PPS to extradata during multi-threaded decoding. */
// 强制线程数为1，因为H.264解码器在多线程解码期间不会将SPS和PPS提取为额外数据
av_dict_set(options ? options : &thread_opt, "threads", "1", 0);
/* Force lowres to 0. The decoder might reduce the video size by the
* lowres factor, and we don't want that propagated to the stream's
* codecpar */
av_dict_set(options ? options : &thread_opt, "lowres", "0", 0);
if (s->codec_whitelist)
av_dict_set(options ? options : &thread_opt, "codec_whitelist", s->codec_whitelist, 0);
ret = avcodec_open2(avctx, codec, options ? options : &thread_opt);
if (!options)
av_dict_free(&thread_opt);
if (ret < 0) {
sti->info->found_decoder = -avctx->codec_id;
goto fail;
}
sti->info->found_decoder = 1;
} else if (!sti->info->found_decoder)
sti->info->found_decoder = 1;
if (sti->info->found_decoder < 0) {
ret = -1;
goto fail;
}
if (avpriv_codec_get_cap_skip_frame_fill_param(avctx->codec)) {
do_skip_frame = 1;
skip_frame = avctx->skip_frame;
avctx->skip_frame = AVDISCARD_ALL;
}
while ((pkt_to_send || (!pkt->data && got_picture)) &&
ret >= 0 &&
(!has_codec_parameters(st, NULL) || !has_decode_delay_been_guessed(st) ||
(!sti->codec_info_nb_frames &&
(avctx->codec->capabilities & AV_CODEC_CAP_CHANNEL_CONF)))) {
got_picture = 0;
// 如果是视频或者是音频信息
if (avctx->codec_type == AVMEDIA_TYPE_VIDEO ||
avctx->codec_type == AVMEDIA_TYPE_AUDIO) {
// 将packet送入解码器
ret = avcodec_send_packet(avctx, pkt);
if (ret < 0 && ret != AVERROR(EAGAIN) && ret != AVERROR_EOF)
break;
if (ret >= 0)
pkt_to_send = 0;
// 获取已经解码的frame
ret = avcodec_receive_frame(avctx, frame);
if (ret >= 0)
got_picture = 1;
if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF)
ret = 0;
} else if (avctx->codec_type == AVMEDIA_TYPE_SUBTITLE) { // 如果是字幕信息
ret = avcodec_decode_subtitle2(avctx, &subtitle,
&got_picture, pkt);
if (got_picture)
avsubtitle_free(&subtitle);
if (ret >= 0)
pkt_to_send = 0;
}
if (ret >= 0) {
if (got_picture)
sti->nb_decoded_frames++; // 已经解码的帧数
ret = got_picture;
}
}
fail:
if (do_skip_frame) {
avctx->skip_frame = skip_frame;
}
av_frame_free(&frame);
return ret;
}
1.6 检查编码器参数（has_codec_parameters）
该函数会检查视频或者音频的参数是否都进行了配置，如果没有进行配置则报出错误
static int has_codec_parameters(const AVStream *st, const char **errmsg_ptr)
{
const FFStream *const sti = cffstream(st);
const AVCodecContext *const avctx = sti->avctx;
#define FAIL(errmsg) do { \
if (errmsg_ptr) \
*errmsg_ptr = errmsg; \
return 0; \
} while (0)
if ( avctx->codec_id == AV_CODEC_ID_NONE
&& avctx->codec_type != AVMEDIA_TYPE_DATA)
FAIL("unknown codec");
switch (avctx->codec_type) {
case AVMEDIA_TYPE_AUDIO:
if (!avctx->frame_size && determinable_frame_size(avctx))
FAIL("unspecified frame size");
if (sti->info->found_decoder >= 0 &&
avctx->sample_fmt == AV_SAMPLE_FMT_NONE)
FAIL("unspecified sample format");
if (!avctx->sample_rate)
FAIL("unspecified sample rate");
if (!avctx->ch_layout.nb_channels)
FAIL("unspecified number of channels");
if (sti->info->found_decoder >= 0 && !sti->nb_decoded_frames && avctx->codec_id == AV_CODEC_ID_DTS)
FAIL("no decodable DTS frames");
break;
case AVMEDIA_TYPE_VIDEO:
if (!avctx->width)
FAIL("unspecified size");
if (sti->info->found_decoder >= 0 && avctx->pix_fmt == AV_PIX_FMT_NONE)
FAIL("unspecified pixel format");
if (st->codecpar->codec_id == AV_CODEC_ID_RV30 || st->codecpar->codec_id == AV_CODEC_ID_RV40)
if (!st->sample_aspect_ratio.num && !st->codecpar->sample_aspect_ratio.num && !sti->codec_info_nb_frames)
FAIL("no frame in rv30/40 and no sar");
break;
case AVMEDIA_TYPE_SUBTITLE:
if (avctx->codec_id == AV_CODEC_ID_HDMV_PGS_SUBTITLE && !avctx->width)
FAIL("unspecified size");
break;
case AVMEDIA_TYPE_DATA:
if (avctx->codec_id == AV_CODEC_ID_NONE) return 1;
}
return 1;
}
1.7 估算时长（estimate_timings）
该函数用于估算流的起始时间，持续时间
static void estimate_timings(AVFormatContext *ic, int64_t old_offset)
{
int64_t file_size;
/* get the file size, if possible */
if (ic->iformat->flags & AVFMT_NOFILE) {
file_size = 0;
} else {
file_size = avio_size(ic->pb);
file_size = FFMAX(0, file_size);
}
// 确定估算的方法
if ((!strcmp(ic->iformat->name, "mpeg") ||
!strcmp(ic->iformat->name, "mpegts")) &&
file_size && (ic->pb->seekable & AVIO_SEEKABLE_NORMAL)) {
/* get accurate estimate from the PTSes */
estimate_timings_from_pts(ic, old_offset);
ic->duration_estimation_method = AVFMT_DURATION_FROM_PTS;
} else if (has_duration(ic)) {
/* at least one component has timings - we use them for all
* the components */
fill_all_stream_timings(ic);
/* nut demuxer estimate the duration from PTS */
if (!strcmp(ic->iformat->name, "nut"))
ic->duration_estimation_method = AVFMT_DURATION_FROM_PTS;
else
ic->duration_estimation_method = AVFMT_DURATION_FROM_STREAM;
} else {
/* less precise: use bitrate info */
estimate_timings_from_bit_rate(ic);
ic->duration_estimation_method = AVFMT_DURATION_FROM_BITRATE;
}
// 更新时间流的计时信息
update_stream_timings(ic);
for (unsigned i = 0; i < ic->nb_streams; i++) {
AVStream *const st = ic->streams[i];
if (st->time_base.den)
av_log(ic, AV_LOG_TRACE, "stream %u: start_time: %s duration: %s\n", i,
av_ts2timestr(st->start_time, &st->time_base),
av_ts2timestr(st->duration, &st->time_base));
}
av_log(ic, AV_LOG_TRACE,
"format: start_time: %s duration: %s (estimate from %s) bitrate=%"PRId64" kb/s\n",
av_ts2timestr(ic->start_time, &AV_TIME_BASE_Q),
av_ts2timestr(ic->duration, &AV_TIME_BASE_Q),
duration_estimate_name(ic->duration_estimation_method),
(int64_t)ic->bit_rate / 1000);
}
其中，估算的方法定义为
/**
* The duration of a video can be estimated through various ways, and this enum can be used
* to know how the duration was estimated.
*/
enum AVDurationEstimationMethod {
// 由PTS精确估计
AVFMT_DURATION_FROM_PTS, ///< Duration accurately estimated from PTSes
// 从具有已知持续时间的流估计的持续时间
AVFMT_DURATION_FROM_STREAM, ///< Duration estimated from a stream with a known duration
// 从比特率估计的持续时间(不太准确)
AVFMT_DURATION_FROM_BITRATE ///< Duration estimated from bitrate (less accurate)
};
在这里，一共使用了3种方法来进行duration的估计，分别是：
（1）AVFMT_DURATION_FROM_PTS
精确的估计方法，从注释上看，且仅用于MPEG-PS格式的文件，调用estimate_timings_from_pts()实现，通过利用最后一个AVPacket的pts和第一个AVPacket的pts相减得到
（2）AVFMT_DURATION_FROM_STREAM
粗略估计的方法，从具有已知持续时间的流估计的持续时间，调用fill_all_stream_timings()实现。具体来说，首先更新流的信息，然后遍历每个流，如果某个流的起始时间戳为0，会从别的流当中拷贝一份起始时间戳start_time和持续时间duration给到当前的流
（3）AVFMT_DURATION_FROM_BITRATE
不太准确估计的方法，从比特率估计的持续时间，调用estimate_timings_from_bit_rate()实现。具体来说，如果AVFormatContext当中没有bit_rate，则把AVFormatContext下面包含的所有AVStream的bit_rate加起来，作为AVFormatContext的bit_rate；如果有bit_rate，则直接认为这个bit_rate是正确的。持续时间duration由file_size除以bit_rate获得。
2.小结
avformat_find_stream_info主要的功能是解析数据源当中的流信息，它是基于avformat_open_input中获取的格式进行的，其中涉及到的关键流信息包括音频信息，视频信息，字幕信息，其中视频信息中又包括了起始时间戳，持续时间，AVCodec信息（width，height等），stream数量等等，这是极其重要的函数，需要重点关注
CSDN : https://blog.csdn.net/weixin_42877471
Github : https://github.com/DoFulangChen
![[PyTorch]：加速Pytorch 模型训练的几种方法（几行代码），最快提升八倍（附实验记录）](https://img-blog.csdnimg.cn/img_convert/145c4c5115995c7a98da0293ede96bc9.png)
