Skip to content

Commit

Permalink
ffmpeg: use new decode API
Browse files Browse the repository at this point in the history 
This is a bit messy, mainly due to timestamp handling.

decode_video() relied on the fact that it could set dts on a flush/drain
packet. This is not possible with the old API, and won't be. (I think
doing this was very questionable with the old API. Flush packets should
not contain any information; they just cause a FIFO to be emptied.) This
is replaced with checking the best_effort_timestamp for AV_NOPTS_VALUE,
and using the suggested DTS in the drain case.

The fate-cavs test still fails due to dropping the last frame. This
happens because the timestamp of the last frame goes backwards
(ffprobe -show_frames shows the same thing). I suspect that this
"worked" due to the best effort timestamp logic picking the DTS
over the decreasing PTS. Since this logic is in libavcodec (where
it probably shouldn't be), this can't be easily fixed. The timestamps
of the cavs samples are weird anyway, so I chose not to fix it.

Another strange thing is the timestamp handling in the video path of
process_input_packet (after the decode_video() call). It looks like
the code to increase next_dts and next_pts should be run every time
a frame is decoded - but it's needed even if output is skipped.
  • Loading branch information
@BtbN
wm4 authored and BtbN committed Sep 18, 2016
1 parent 8fab5c7 commit 5bc339b
Show file tree
Hide file tree
Showing 2 changed files with 75 additions and 52 deletions.
126 changes: 75 additions & 51 deletions ffmpeg.c
View file
Expand Up @@ -1943,6 +1943,33 @@ static void check_decode_result(InputStream *ist, int *got_output, int ret)
}
}

// This does not quite work like avcodec_decode_audio4/avcodec_decode_video2.
// There is the following difference: if you got a frame, you must call
// it again with pkt=NULL. pkt==NULL is treated differently from pkt.size==0
// (pkt==NULL means get more output, pkt.size==0 is a flush/drain packet)
static int decode(AVCodecContext *avctx, AVFrame *frame, int *got_frame, AVPacket *pkt)
{
int ret;

*got_frame = 0;

if (pkt) {
ret = avcodec_send_packet(avctx, pkt);
// In particular, we don't expect AVERROR(EAGAIN), because we read all
// decoded frames with avcodec_receive_frame() until done.
if (ret < 0)
return ret == AVERROR_EOF ? 0 : ret;
}

ret = avcodec_receive_frame(avctx, frame);
if (ret < 0 && ret != AVERROR(EAGAIN) && ret != AVERROR_EOF)
return ret;
if (ret >= 0)
*got_frame = 1;

return 0;
}

static int decode_audio(InputStream *ist, AVPacket *pkt, int *got_output)
{
AVFrame *decoded_frame, *f;
Expand All @@ -1957,7 +1984,7 @@ static int decode_audio(InputStream *ist, AVPacket *pkt, int *got_output)
decoded_frame = ist->decoded_frame;

update_benchmark(NULL);
ret = avcodec_decode_audio4(avctx, decoded_frame, got_output, pkt);
ret = decode(avctx, decoded_frame, got_output, pkt);
update_benchmark("decode_audio %d.%d", ist->file_index, ist->st->index);

if (ret >= 0 && avctx->sample_rate <= 0) {
Expand Down Expand Up @@ -2033,14 +2060,13 @@ static int decode_audio(InputStream *ist, AVPacket *pkt, int *got_output)
} else if (decoded_frame->pkt_pts != AV_NOPTS_VALUE) {
decoded_frame->pts = decoded_frame->pkt_pts;
decoded_frame_tb = ist->st->time_base;
} else if (pkt->pts != AV_NOPTS_VALUE) {
} else if (pkt && pkt->pts != AV_NOPTS_VALUE) {
decoded_frame->pts = pkt->pts;
decoded_frame_tb = ist->st->time_base;
}else {
decoded_frame->pts = ist->dts;
decoded_frame_tb = AV_TIME_BASE_Q;
}
pkt->pts = AV_NOPTS_VALUE;
if (decoded_frame->pts != AV_NOPTS_VALUE)
decoded_frame->pts = av_rescale_delta(decoded_frame_tb, decoded_frame->pts,
(AVRational){1, avctx->sample_rate}, decoded_frame->nb_samples, &ist->filter_in_rescale_delta_last,
Expand Down Expand Up @@ -2068,23 +2094,28 @@ static int decode_audio(InputStream *ist, AVPacket *pkt, int *got_output)
return err < 0 ? err : ret;
}

static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output, int eof)
{
AVFrame *decoded_frame, *f;
int i, ret = 0, err = 0, resample_changed;
int64_t best_effort_timestamp;
int64_t dts;
AVRational *frame_sample_aspect;
AVPacket avpkt;

if (!ist->decoded_frame && !(ist->decoded_frame = av_frame_alloc()))
return AVERROR(ENOMEM);
if (!ist->filter_frame && !(ist->filter_frame = av_frame_alloc()))
return AVERROR(ENOMEM);
decoded_frame = ist->decoded_frame;
pkt->dts = av_rescale_q(ist->dts, AV_TIME_BASE_Q, ist->st->time_base);
dts = av_rescale_q(ist->dts, AV_TIME_BASE_Q, ist->st->time_base);
if (pkt) {
avpkt = *pkt;
avpkt.dts = dts;
}

update_benchmark(NULL);
ret = avcodec_decode_video2(ist->dec_ctx,
decoded_frame, got_output, pkt);
ret = decode(ist->dec_ctx, decoded_frame, got_output, pkt ? &avpkt : NULL);
update_benchmark("decode_video %d.%d", ist->file_index, ist->st->index);

// The following line may be required in some cases where there is no parser
Expand Down Expand Up @@ -2134,6 +2165,8 @@ static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
ist->hwaccel_retrieved_pix_fmt = decoded_frame->format;

best_effort_timestamp= av_frame_get_best_effort_timestamp(decoded_frame);
if(eof && best_effort_timestamp == AV_NOPTS_VALUE)
best_effort_timestamp = dts;
if(best_effort_timestamp != AV_NOPTS_VALUE) {
int64_t ts = av_rescale_q(decoded_frame->pts = best_effort_timestamp, ist->st->time_base, AV_TIME_BASE_Q);

Expand All @@ -2152,8 +2185,6 @@ static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
ist->st->time_base.num, ist->st->time_base.den);
}

pkt->size = 0;

if (ist->st->sample_aspect_ratio.num)
decoded_frame->sample_aspect_ratio = ist->st->sample_aspect_ratio;

Expand Down Expand Up @@ -2192,12 +2223,12 @@ static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
break;
} else
f = decoded_frame;
ret = av_buffersrc_add_frame_flags(ist->filters[i]->filter, f, AV_BUFFERSRC_FLAG_PUSH);
if (ret == AVERROR_EOF) {
ret = 0; /* ignore */
} else if (ret < 0) {
err = av_buffersrc_add_frame_flags(ist->filters[i]->filter, f, AV_BUFFERSRC_FLAG_PUSH);
if (err == AVERROR_EOF) {
err = 0; /* ignore */
} else if (err < 0) {
av_log(NULL, AV_LOG_FATAL,
"Failed to inject frame into filter network: %s\n", av_err2str(ret));
"Failed to inject frame into filter network: %s\n", av_err2str(err));
exit_program(1);
}
}
Expand Down Expand Up @@ -2282,7 +2313,8 @@ static int send_filter_eof(InputStream *ist)
static int process_input_packet(InputStream *ist, const AVPacket *pkt, int no_eof)
{
int ret = 0, i;
int got_output = 0;
int repeating = 0;
int eof_reached = !pkt;

AVPacket avpkt;
if (!ist->saw_first_ts) {
Expand All @@ -2305,50 +2337,52 @@ static int process_input_packet(InputStream *ist, const AVPacket *pkt, int no_eo
av_init_packet(&avpkt);
avpkt.data = NULL;
avpkt.size = 0;
goto handle_eof;
} else {
avpkt = *pkt;
}

if (pkt->dts != AV_NOPTS_VALUE) {
if (pkt && pkt->dts != AV_NOPTS_VALUE) {
ist->next_dts = ist->dts = av_rescale_q(pkt->dts, ist->st->time_base, AV_TIME_BASE_Q);
if (ist->dec_ctx->codec_type != AVMEDIA_TYPE_VIDEO || !ist->decoding_needed)
ist->next_pts = ist->pts = ist->dts;
}

// while we have more to decode or while the decoder did output something on EOF
while (ist->decoding_needed && (avpkt.size > 0 || (!pkt && got_output))) {
int duration;
handle_eof:
while (ist->decoding_needed && (!pkt || avpkt.size > 0)) {
int duration = 0;
int got_output = 0;

ist->pts = ist->next_pts;
ist->dts = ist->next_dts;

switch (ist->dec_ctx->codec_type) {
case AVMEDIA_TYPE_AUDIO:
ret = decode_audio (ist, &avpkt, &got_output);
ret = decode_audio (ist, repeating ? NULL : &avpkt, &got_output);
break;
case AVMEDIA_TYPE_VIDEO:
ret = decode_video (ist, &avpkt, &got_output);
if (avpkt.duration) {
duration = av_rescale_q(avpkt.duration, ist->st->time_base, AV_TIME_BASE_Q);
} else if(ist->dec_ctx->framerate.num != 0 && ist->dec_ctx->framerate.den != 0) {
int ticks= av_stream_get_parser(ist->st) ? av_stream_get_parser(ist->st)->repeat_pict+1 : ist->dec_ctx->ticks_per_frame;
duration = ((int64_t)AV_TIME_BASE *
ist->dec_ctx->framerate.den * ticks) /
ist->dec_ctx->framerate.num / ist->dec_ctx->ticks_per_frame;
} else
duration = 0;
ret = decode_video (ist, repeating ? NULL : &avpkt, &got_output, !pkt);
if (!repeating || !pkt || got_output) {
if (pkt && pkt->duration) {
duration = av_rescale_q(pkt->duration, ist->st->time_base, AV_TIME_BASE_Q);
} else if(ist->dec_ctx->framerate.num != 0 && ist->dec_ctx->framerate.den != 0) {
int ticks= av_stream_get_parser(ist->st) ? av_stream_get_parser(ist->st)->repeat_pict+1 : ist->dec_ctx->ticks_per_frame;
duration = ((int64_t)AV_TIME_BASE *
ist->dec_ctx->framerate.den * ticks) /
ist->dec_ctx->framerate.num / ist->dec_ctx->ticks_per_frame;
}

if(ist->dts != AV_NOPTS_VALUE && duration) {
ist->next_dts += duration;
}else
ist->next_dts = AV_NOPTS_VALUE;
if(ist->dts != AV_NOPTS_VALUE && duration) {
ist->next_dts += duration;
}else
ist->next_dts = AV_NOPTS_VALUE;
}

if (got_output)
ist->next_pts += duration; //FIXME the duration is not correct in some cases
break;
case AVMEDIA_TYPE_SUBTITLE:
if (repeating)
break;
ret = transcode_subtitles(ist, &avpkt, &got_output);
break;
default:
Expand All @@ -2363,26 +2397,16 @@ static int process_input_packet(InputStream *ist, const AVPacket *pkt, int no_eo
break;
}

avpkt.dts=
avpkt.pts= AV_NOPTS_VALUE;

// touch data and size only if not EOF
if (pkt) {
if(ist->dec_ctx->codec_type != AVMEDIA_TYPE_AUDIO)
ret = avpkt.size;
avpkt.data += ret;
avpkt.size -= ret;
}
if (!got_output) {
continue;
}
if (got_output && !pkt)
if (!got_output)
break;

eof_reached = 0;
repeating = 1;
}

/* after flushing, send an EOF on all the filter inputs attached to the stream */
/* except when looping we need to flush but not to send an EOF */
if (!pkt && ist->decoding_needed && !got_output && !no_eof) {
if (!pkt && ist->decoding_needed && eof_reached && !no_eof) {
int ret = send_filter_eof(ist);
if (ret < 0) {
av_log(NULL, AV_LOG_FATAL, "Error marking filters as finished\n");
Expand Down Expand Up @@ -2426,7 +2450,7 @@ static int process_input_packet(InputStream *ist, const AVPacket *pkt, int no_eo
do_streamcopy(ist, ost, pkt);
}

return got_output;
return !eof_reached;
}

static void print_sdp(void)
Expand Down
1 change: 0 additions & 1 deletion tests/ref/fate/cavs
View file
Expand Up @@ -173,4 +173,3 @@
0, 167, 167, 1, 622080, 0xdcb4cee8
0, 168, 168, 1, 622080, 0xb41172e5
0, 169, 169, 1, 622080, 0x56c72478
0, 170, 170, 1, 622080, 0x84ff3af9

0 comments on commit 5bc339b

Please sign in to comment.