X-Git-Url: https://main.carlh.net/gitweb/?a=blobdiff_plain;f=src%2Flib%2Fffmpeg_decoder.cc;h=25ca9bb8882f658385ac677f86a36c685ba5b404;hb=db468a15e50c8491d4b8462ad0676be905f49065;hp=462db283ab8875dca384d96c045d44531513f4f7;hpb=ab1a666e724911c41dfe08fc96748b38ace839c1;p=dcpomatic.git diff --git a/src/lib/ffmpeg_decoder.cc b/src/lib/ffmpeg_decoder.cc index 462db283a..25ca9bb88 100644 --- a/src/lib/ffmpeg_decoder.cc +++ b/src/lib/ffmpeg_decoder.cc @@ -41,7 +41,6 @@ extern "C" { #include "transcoder.h" #include "job.h" #include "filter.h" -#include "options.h" #include "exceptions.h" #include "image.h" #include "util.h" @@ -62,10 +61,11 @@ using boost::optional; using boost::dynamic_pointer_cast; using libdcp::Size; -FFmpegDecoder::FFmpegDecoder (shared_ptr f, DecodeOptions o) - : Decoder (f, o) - , VideoDecoder (f, o) - , AudioDecoder (f, o) +FFmpegDecoder::FFmpegDecoder (shared_ptr f, shared_ptr c, bool video, bool audio, bool subtitles, bool video_sync) + : Decoder (f) + , VideoDecoder (f) + , AudioDecoder (f) + , _ffmpeg_content (c) , _format_context (0) , _video_stream (-1) , _frame (0) @@ -75,13 +75,17 @@ FFmpegDecoder::FFmpegDecoder (shared_ptr f, DecodeOptions o) , _audio_codec (0) , _subtitle_codec_context (0) , _subtitle_codec (0) + , _decode_video (video) + , _decode_audio (audio) + , _decode_subtitles (subtitles) + , _video_sync (video_sync) { setup_general (); setup_video (); setup_audio (); setup_subtitle (); - if (!o.video_sync) { + if (!video_sync) { _first_video = 0; } } @@ -110,8 +114,8 @@ FFmpegDecoder::setup_general () { av_register_all (); - if (avformat_open_input (&_format_context, _film->content_path().c_str(), 0, 0) < 0) { - throw OpenFileError (_film->content_path ()); + if (avformat_open_input (&_format_context, _ffmpeg_content->file().string().c_str(), 0, 0) < 0) { + throw OpenFileError (_ffmpeg_content->file().string ()); } if (avformat_find_stream_info (_format_context, 0) < 0) { @@ -135,17 +139,11 @@ FFmpegDecoder::setup_general () } _audio_streams.push_back ( - shared_ptr ( - new FFmpegAudioStream (stream_name (s), i, s->codec->sample_rate, s->codec->channel_layout) - ) + FFmpegAudioStream (stream_name (s), i, s->codec->sample_rate, s->codec->channel_layout) ); } else if (s->codec->codec_type == AVMEDIA_TYPE_SUBTITLE) { - _subtitle_streams.push_back ( - shared_ptr ( - new SubtitleStream (stream_name (s), i) - ) - ); + _subtitle_streams.push_back (FFmpegSubtitleStream (stream_name (s), i)); } } @@ -177,14 +175,11 @@ FFmpegDecoder::setup_video () void FFmpegDecoder::setup_audio () { - if (!_audio_stream) { + if (!_ffmpeg_content->audio_stream ()) { return; } - shared_ptr ffa = dynamic_pointer_cast (_audio_stream); - assert (ffa); - - _audio_codec_context = _format_context->streams[ffa->id()]->codec; + _audio_codec_context = _format_context->streams[_ffmpeg_content->audio_stream()->id]->codec; _audio_codec = avcodec_find_decoder (_audio_codec_context->codec_id); if (_audio_codec == 0) { @@ -199,11 +194,11 @@ FFmpegDecoder::setup_audio () void FFmpegDecoder::setup_subtitle () { - if (!_subtitle_stream || _subtitle_stream->id() >= int (_format_context->nb_streams)) { + if (!_ffmpeg_content->subtitle_stream() || _ffmpeg_content->subtitle_stream()->id >= int (_format_context->nb_streams)) { return; } - _subtitle_codec_context = _format_context->streams[_subtitle_stream->id()]->codec; + _subtitle_codec_context = _format_context->streams[_ffmpeg_content->subtitle_stream()->id]->codec; _subtitle_codec = avcodec_find_decoder (_subtitle_codec_context->codec_id); if (_subtitle_codec == 0) { @@ -238,21 +233,14 @@ FFmpegDecoder::pass () int frame_finished; - if (_opt.decode_video) { + if (_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) { - 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, data_size)); - } + if (_ffmpeg_content->audio_stream() && _decode_audio) { + decode_audio_packet (); } return true; @@ -260,9 +248,7 @@ FFmpegDecoder::pass () avcodec_get_frame_defaults (_frame); - shared_ptr ffa = dynamic_pointer_cast (_audio_stream); - - if (_packet.stream_index == _video_stream && _opt.decode_video) { + if (_packet.stream_index == _video_stream && _decode_video) { int frame_finished; int const r = avcodec_decode_video2 (_video_codec_context, _frame, &frame_finished, &_packet); @@ -272,63 +258,16 @@ FFmpegDecoder::pass () _film->log()->log (String::compose (N_("Used only %1 bytes of %2 in packet"), r, _packet.size)); } - if (_opt.video_sync) { + if (_video_sync) { out_with_sync (); } else { filter_and_emit_video (_frame); } } - } 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) { - - /* Where we are in the source, in seconds */ - double const source_pts_seconds = av_q2d (_format_context->streams[_packet.stream_index]->time_base) - * av_frame_get_best_effort_timestamp(_frame); - - /* We only decode audio if we've had our first video packet through, and if it - was before this packet. Until then audio is thrown away. - */ - - if ((_first_video && _first_video.get() <= source_pts_seconds) || !_opt.decode_video) { - - 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 - first video frame. Push some silence to make up any gap between our first - video frame and our first audio. - */ - - /* frames of silence that we must push */ - int const s = rint ((_first_audio.get() - _first_video.get()) * ffa->sample_rate ()); - - _film->log()->log ( - String::compose ( - N_("First video at %1, first audio at %2, pushing %3 audio frames of silence for %4 channels (%5 bytes per sample)"), - _first_video.get(), _first_audio.get(), s, ffa->channels(), bytes_per_audio_sample() - ) - ); - - if (s) { - shared_ptr audio (new AudioBuffers (ffa->channels(), s)); - audio->make_silent (); - Audio (audio); - } - } - - 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, data_size)); - } - } - - } else if (_subtitle_stream && _packet.stream_index == _subtitle_stream->id() && _opt.decode_subtitles && _first_video) { + } else if (_ffmpeg_content->audio_stream() && _packet.stream_index == _ffmpeg_content->audio_stream()->id && _decode_audio) { + decode_audio_packet (); + } else if (_ffmpeg_content->subtitle_stream() && _packet.stream_index == _ffmpeg_content->subtitle_stream()->id && _decode_subtitles && _first_video) { int got_subtitle; AVSubtitle sub; @@ -360,19 +299,16 @@ FFmpegDecoder::pass () shared_ptr FFmpegDecoder::deinterleave_audio (uint8_t** data, int size) { - assert (_film->audio_channels()); + assert (_ffmpeg_content->audio_channels()); assert (bytes_per_audio_sample()); - shared_ptr ffa = dynamic_pointer_cast (_audio_stream); - assert (ffa); - /* Deinterleave and convert to float */ - assert ((size % (bytes_per_audio_sample() * ffa->channels())) == 0); + assert ((size % (bytes_per_audio_sample() * _ffmpeg_content->audio_channels())) == 0); int const total_samples = size / bytes_per_audio_sample(); - int const frames = total_samples / _film->audio_channels(); - shared_ptr audio (new AudioBuffers (ffa->channels(), frames)); + int const frames = total_samples / _ffmpeg_content->audio_channels(); + shared_ptr audio (new AudioBuffers (_ffmpeg_content->audio_channels(), frames)); switch (audio_sample_format()) { case AV_SAMPLE_FMT_S16: @@ -384,7 +320,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t** data, int size) audio->data(channel)[sample] = float(*p++) / (1 << 15); ++channel; - if (channel == _film->audio_channels()) { + if (channel == _ffmpeg_content->audio_channels()) { channel = 0; ++sample; } @@ -395,7 +331,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t** data, int size) case AV_SAMPLE_FMT_S16P: { int16_t** p = reinterpret_cast (data); - for (int i = 0; i < _film->audio_channels(); ++i) { + for (int i = 0; i < _ffmpeg_content->audio_channels(); ++i) { for (int j = 0; j < frames; ++j) { audio->data(i)[j] = static_cast(p[i][j]) / (1 << 15); } @@ -412,7 +348,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t** data, int size) audio->data(channel)[sample] = static_cast(*p++) / (1 << 31); ++channel; - if (channel == _film->audio_channels()) { + if (channel == _ffmpeg_content->audio_channels()) { channel = 0; ++sample; } @@ -429,7 +365,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t** data, int size) audio->data(channel)[sample] = *p++; ++channel; - if (channel == _film->audio_channels()) { + if (channel == _ffmpeg_content->audio_channels()) { channel = 0; ++sample; } @@ -440,7 +376,7 @@ FFmpegDecoder::deinterleave_audio (uint8_t** data, int size) case AV_SAMPLE_FMT_FLTP: { float** p = reinterpret_cast (data); - for (int i = 0; i < _film->audio_channels(); ++i) { + for (int i = 0; i < _ffmpeg_content->audio_channels(); ++i) { memcpy (audio->data(i), p[i], frames * sizeof(float)); } } @@ -542,21 +478,6 @@ FFmpegDecoder::bytes_per_audio_sample () const return av_get_bytes_per_sample (audio_sample_format ()); } -void -FFmpegDecoder::set_audio_stream (shared_ptr s) -{ - AudioDecoder::set_audio_stream (s); - setup_audio (); -} - -void -FFmpegDecoder::set_subtitle_stream (shared_ptr s) -{ - VideoDecoder::set_subtitle_stream (s); - setup_subtitle (); - OutputChanged (); -} - void FFmpegDecoder::filter_and_emit_video (AVFrame* frame) { @@ -615,58 +536,6 @@ FFmpegDecoder::do_seek (double p, bool backwards) return r < 0; } -shared_ptr -FFmpegAudioStream::create (string t, optional v) -{ - if (!v) { - /* version < 1; no type in the string, and there's only FFmpeg streams anyway */ - return shared_ptr (new FFmpegAudioStream (t, v)); - } - - stringstream s (t); - string type; - s >> type; - if (type != N_("ffmpeg")) { - return shared_ptr (); - } - - return shared_ptr (new FFmpegAudioStream (t, v)); -} - -FFmpegAudioStream::FFmpegAudioStream (string t, optional version) -{ - stringstream n (t); - - int name_index = 4; - if (!version) { - name_index = 2; - int channels; - n >> _id >> channels; - _channel_layout = av_get_default_channel_layout (channels); - _sample_rate = 0; - } else { - string type; - /* Current (marked version 1) */ - n >> type >> _id >> _sample_rate >> _channel_layout; - assert (type == N_("ffmpeg")); - } - - for (int i = 0; i < name_index; ++i) { - size_t const s = t.find (' '); - if (s != string::npos) { - t = t.substr (s + 1); - } - } - - _name = t; -} - -string -FFmpegAudioStream::to_string () const -{ - return String::compose (N_("ffmpeg %1 %2 %3 %4"), _id, _sample_rate, _channel_layout, _name); -} - void FFmpegDecoder::out_with_sync () { @@ -732,8 +601,8 @@ FFmpegDecoder::film_changed (Film::Property p) } /** @return Length (in video frames) according to our content's header */ -SourceFrame -FFmpegDecoder::length () const +ContentVideoFrame +FFmpegDecoder::video_length () const { return (double(_format_context->duration) / AV_TIME_BASE) * frames_per_second(); } @@ -744,3 +613,68 @@ FFmpegDecoder::frame_time () const return av_frame_get_best_effort_timestamp(_frame) * av_q2d (_format_context->streams[_video_stream]->time_base); } +void +FFmpegDecoder::decode_audio_packet () +{ + /* Audio packets can contain multiple frames, so we may have to call avcodec_decode_audio4 + several times. + */ + + AVPacket copy_packet = _packet; + + while (copy_packet.size > 0) { + + int frame_finished; + int const decode_result = avcodec_decode_audio4 (_audio_codec_context, _frame, &frame_finished, ©_packet); + if (decode_result >= 0 && frame_finished) { + + /* Where we are in the source, in seconds */ + double const source_pts_seconds = av_q2d (_format_context->streams[copy_packet.stream_index]->time_base) + * av_frame_get_best_effort_timestamp(_frame); + + /* We only decode audio if we've had our first video packet through, and if it + was before this packet. Until then audio is thrown away. + */ + + if ((_first_video && _first_video.get() <= source_pts_seconds) || !_decode_video) { + + if (!_first_audio && _decode_video) { + _first_audio = source_pts_seconds; + + /* This is our first audio frame, and if we've arrived here we must have had our + first video frame. Push some silence to make up any gap between our first + video frame and our first audio. + */ + + /* frames of silence that we must push */ + int const s = rint ((_first_audio.get() - _first_video.get()) * _ffmpeg_content->audio_frame_rate ()); + + _film->log()->log ( + String::compose ( + N_("First video at %1, first audio at %2, pushing %3 audio frames of silence for %4 channels (%5 bytes per sample)"), + _first_video.get(), _first_audio.get(), s, _ffmpeg_content->audio_channels(), bytes_per_audio_sample() + ) + ); + + if (s) { + shared_ptr audio (new AudioBuffers (_ffmpeg_content->audio_channels(), s)); + audio->make_silent (); + Audio (audio); + } + } + + 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 == _ffmpeg_content->audio_channels()); + Audio (deinterleave_audio (_frame->data, data_size)); + } + } + + if (decode_result >= 0) { + copy_packet.data += decode_result; + copy_packet.size -= decode_result; + } + } +}