X-Git-Url: https://git.carlh.net/gitweb/?a=blobdiff_plain;f=src%2Flib%2Fffmpeg_decoder.cc;h=148764162c20fe70bfdd9d4be2e7ed96faaf8d57;hb=ab624eaed4423b1813148e1c395092697b541139;hp=136843190d9cbd1cb46290eb8e5dcd7c209ee671;hpb=aa230169f8b59b7cb2da9a3bbb8ce5f7600285c0;p=dcpomatic.git diff --git a/src/lib/ffmpeg_decoder.cc b/src/lib/ffmpeg_decoder.cc index 136843190..148764162 100644 --- a/src/lib/ffmpeg_decoder.cc +++ b/src/lib/ffmpeg_decoder.cc @@ -58,11 +58,12 @@ using std::list; using boost::shared_ptr; using boost::optional; using boost::dynamic_pointer_cast; +using libdcp::Size; -FFmpegDecoder::FFmpegDecoder (shared_ptr f, shared_ptr o, Job* j) - : Decoder (f, o, j) - , VideoDecoder (f, o, j) - , AudioDecoder (f, o, j) +FFmpegDecoder::FFmpegDecoder (shared_ptr f, DecodeOptions o) + : Decoder (f, o) + , VideoDecoder (f, o) + , AudioDecoder (f, o) , _format_context (0) , _video_stream (-1) , _frame (0) @@ -78,7 +79,9 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr f, shared_ptrChanged.connect (bind (&FFmpegDecoder::film_changed, this, _1)); + if (!o.video_sync) { + _first_video = 0; + } } FFmpegDecoder::~FFmpegDecoder () @@ -124,7 +127,7 @@ FFmpegDecoder::setup_general () /* This is a hack; sometimes it seems that _audio_codec_context->channel_layout isn't set up, so bodge it here. No idea why we should have to do this. */ - + if (s->codec->channel_layout == 0) { s->codec->channel_layout = av_get_default_channel_layout (s->codec->channels); } @@ -194,7 +197,7 @@ FFmpegDecoder::setup_audio () void FFmpegDecoder::setup_subtitle () { - if (!_subtitle_stream) { + if (!_subtitle_stream || _subtitle_stream->id() >= _format_context->nb_streams) { return; } @@ -233,18 +236,20 @@ FFmpegDecoder::pass () int frame_finished; - while (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) { - filter_and_emit_video (_frame); + if (_opt.decode_video) { + while (avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) { + filter_and_emit_video (_frame); + } } - if (_audio_stream && _opt->decode_audio) { + if (_audio_stream && _opt.decode_audio) { while (avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) { int const data_size = av_samples_get_buffer_size ( 0, _audio_codec_context->channels, _frame->nb_samples, audio_sample_format (), 1 ); assert (_audio_codec_context->channels == _film->audio_channels()); - Audio (deinterleave_audio (_frame->data[0], data_size)); + Audio (deinterleave_audio (_frame->data, data_size)); } } @@ -254,8 +259,8 @@ FFmpegDecoder::pass () avcodec_get_frame_defaults (_frame); shared_ptr ffa = dynamic_pointer_cast (_audio_stream); - - if (_packet.stream_index == _video_stream) { + + if (_packet.stream_index == _video_stream && _opt.decode_video) { int frame_finished; int const r = avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet); @@ -265,14 +270,14 @@ FFmpegDecoder::pass () _film->log()->log (String::compose ("Used only %1 bytes of %2 in packet", r, _packet.size)); } - if (_opt->video_sync) { + if (_opt.video_sync) { out_with_sync (); } else { filter_and_emit_video (_frame); } } - } else if (ffa && _packet.stream_index == ffa->id() && _opt->decode_audio) { + } else if (ffa && _packet.stream_index == ffa->id() && _opt.decode_audio) { int frame_finished; if (avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, &_packet) >= 0 && frame_finished) { @@ -285,9 +290,9 @@ FFmpegDecoder::pass () was before this packet. Until then audio is thrown away. */ - if (_first_video && _first_video.get() <= source_pts_seconds) { + if ((_first_video && _first_video.get() <= source_pts_seconds) || !_opt.decode_video) { - if (!_first_audio) { + if (!_first_audio && _opt.decode_video) { _first_audio = source_pts_seconds; /* This is our first audio frame, and if we've arrived here we must have had our @@ -317,11 +322,11 @@ FFmpegDecoder::pass () ); assert (_audio_codec_context->channels == _film->audio_channels()); - Audio (deinterleave_audio (_frame->data[0], data_size)); + Audio (deinterleave_audio (_frame->data, data_size)); } } - } else if (_subtitle_stream && _packet.stream_index == _subtitle_stream->id() && _opt->decode_subtitles && _first_video) { + } else if (_subtitle_stream && _packet.stream_index == _subtitle_stream->id() && _opt.decode_subtitles && _first_video) { int got_subtitle; AVSubtitle sub; @@ -332,7 +337,7 @@ FFmpegDecoder::pass () if (sub.num_rects > 0) { shared_ptr ts; try { - emit_subtitle (shared_ptr (new TimedSubtitle (sub, _first_video.get()))); + emit_subtitle (shared_ptr (new TimedSubtitle (sub))); } catch (...) { /* some problem with the subtitle; we probably didn't understand it */ } @@ -347,8 +352,11 @@ FFmpegDecoder::pass () return false; } +/** @param data pointer to array of pointers to buffers. + * Only the first buffer will be used for non-planar data, otherwise there will be one per channel. + */ shared_ptr -FFmpegDecoder::deinterleave_audio (uint8_t* data, int size) +FFmpegDecoder::deinterleave_audio (uint8_t** data, int size) { assert (_film->audio_channels()); assert (bytes_per_audio_sample()); @@ -367,7 +375,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t* data, int size) switch (audio_sample_format()) { case AV_SAMPLE_FMT_S16: { - int16_t* p = (int16_t *) data; + int16_t* p = reinterpret_cast (data[0]); int sample = 0; int channel = 0; for (int i = 0; i < total_samples; ++i) { @@ -382,13 +390,24 @@ FFmpegDecoder::deinterleave_audio (uint8_t* data, int size) } break; + case AV_SAMPLE_FMT_S16P: + { + int16_t** p = reinterpret_cast (data); + for (int i = 0; i < _film->audio_channels(); ++i) { + for (int j = 0; j < frames; ++j) { + audio->data(i)[j] = static_cast(p[i][j]) / (1 << 15); + } + } + } + break; + case AV_SAMPLE_FMT_S32: { - int32_t* p = (int32_t *) data; + int32_t* p = reinterpret_cast (data[0]); int sample = 0; int channel = 0; for (int i = 0; i < total_samples; ++i) { - audio->data(channel)[sample] = float(*p++) / (1 << 31); + audio->data(channel)[sample] = static_cast(*p++) / (1 << 31); ++channel; if (channel == _film->audio_channels()) { @@ -397,19 +416,36 @@ FFmpegDecoder::deinterleave_audio (uint8_t* data, int size) } } } + break; + + case AV_SAMPLE_FMT_FLT: + { + float* p = reinterpret_cast (data[0]); + int sample = 0; + int channel = 0; + for (int i = 0; i < total_samples; ++i) { + audio->data(channel)[sample] = *p++; + ++channel; + if (channel == _film->audio_channels()) { + channel = 0; + ++sample; + } + } + } + break; + case AV_SAMPLE_FMT_FLTP: { - float* p = reinterpret_cast (data); + float** p = reinterpret_cast (data); for (int i = 0; i < _film->audio_channels(); ++i) { - memcpy (audio->data(i), p, frames * sizeof(float)); - p += frames; + memcpy (audio->data(i), p[i], frames * sizeof(float)); } } break; default: - assert (false); + throw DecodeError (String::compose ("Unrecognised audio sample format (%1)", static_cast (audio_sample_format()))); } return audio; @@ -437,10 +473,10 @@ FFmpegDecoder::audio_sample_format () const return _audio_codec_context->sample_fmt; } -Size +libdcp::Size FFmpegDecoder::native_size () const { - return Size (_video_codec_context->width, _video_codec_context->height); + return libdcp::Size (_video_codec_context->width, _video_codec_context->height); } PixelFormat @@ -516,6 +552,7 @@ FFmpegDecoder::set_subtitle_stream (shared_ptr s) { VideoDecoder::set_subtitle_stream (s); setup_subtitle (); + OutputChanged (); } void @@ -526,12 +563,12 @@ FFmpegDecoder::filter_and_emit_video (AVFrame* frame) shared_ptr graph; list >::iterator i = _filter_graphs.begin(); - while (i != _filter_graphs.end() && !(*i)->can_process (Size (frame->width, frame->height), (AVPixelFormat) frame->format)) { + while (i != _filter_graphs.end() && !(*i)->can_process (libdcp::Size (frame->width, frame->height), (AVPixelFormat) frame->format)) { ++i; } if (i == _filter_graphs.end ()) { - graph.reset (new FilterGraph (_film, this, Size (frame->width, frame->height), (AVPixelFormat) frame->format)); + graph.reset (new FilterGraph (_film, this, libdcp::Size (frame->width, frame->height), (AVPixelFormat) frame->format)); _filter_graphs.push_back (graph); _film->log()->log (String::compose ("New graph for %1x%2, pixel format %3", frame->width, frame->height, frame->format)); } else { @@ -540,20 +577,39 @@ FFmpegDecoder::filter_and_emit_video (AVFrame* frame) list > images = graph->process (frame); - SourceFrame const sf = av_q2d (_format_context->streams[_video_stream]->time_base) - * av_frame_get_best_effort_timestamp(_frame) * frames_per_second(); - for (list >::iterator i = images.begin(); i != images.end(); ++i) { - emit_video (*i, sf); + emit_video (*i, frame_time ()); } } bool -FFmpegDecoder::seek (SourceFrame f) +FFmpegDecoder::seek (double p) { - int64_t const t = static_cast(f) / (av_q2d (_format_context->streams[_video_stream]->time_base) * frames_per_second()); - int const r = av_seek_frame (_format_context, _video_stream, t, 0); + return do_seek (p, false); +} + +bool +FFmpegDecoder::seek_to_last () +{ + /* This AVSEEK_FLAG_BACKWARD in do_seek is a bit of a hack; without it, if we ask for a seek to the same place as last time + (used when we change decoder parameters and want to re-fetch the frame) we end up going forwards rather than + staying in the same place. + */ + return do_seek (last_source_time(), true); +} + +bool +FFmpegDecoder::do_seek (double p, bool backwards) +{ + int64_t const vt = p / av_q2d (_format_context->streams[_video_stream]->time_base); + + int const r = av_seek_frame (_format_context, _video_stream, vt, backwards ? AVSEEK_FLAG_BACKWARD : 0); + avcodec_flush_buffers (_video_codec_context); + if (_subtitle_codec_context) { + avcodec_flush_buffers (_subtitle_codec_context); + } + return r < 0; } @@ -641,7 +697,7 @@ FFmpegDecoder::out_with_sync () String::compose ( "Extra video frame inserted at %1s; source frame %2, source PTS %3 (at %4 fps)", out_pts_seconds, video_frame(), source_pts_seconds, frames_per_second() - ) + ) ); } } @@ -660,6 +716,7 @@ FFmpegDecoder::film_changed (Film::Property p) { switch (p) { case Film::CROP: + case Film::FILTERS: { boost::mutex::scoped_lock lm (_filter_graphs_mutex); _filter_graphs.clear (); @@ -679,3 +736,9 @@ FFmpegDecoder::length () const return (double(_format_context->duration) / AV_TIME_BASE) * frames_per_second(); } +double +FFmpegDecoder::frame_time () const +{ + return av_frame_get_best_effort_timestamp(_frame) * av_q2d (_format_context->streams[_video_stream]->time_base); +} +