X-Git-Url: https://main.carlh.net/gitweb/?p=dcpomatic.git;a=blobdiff_plain;f=src%2Flib%2Faudio_decoder.cc;h=f7f147bd92e9b80f58cc46f5bd91c34fb551d6fd;hp=16a03a8e93593cdc5879f1e33f72c93742660ffb;hb=0241df1707c7ea5658f471828ff6dc944e21af42;hpb=9ba99535750e19c341f6ff535c6c8991658a8fbb diff --git a/src/lib/audio_decoder.cc b/src/lib/audio_decoder.cc index 16a03a8e9..f7f147bd9 100644 --- a/src/lib/audio_decoder.cc +++ b/src/lib/audio_decoder.cc @@ -1,5 +1,5 @@ /* - Copyright (C) 2012-2017 Carl Hetherington + Copyright (C) 2012-2021 Carl Hetherington This file is part of DCP-o-matic. @@ -21,126 +21,157 @@ #include "audio_decoder.h" #include "audio_buffers.h" #include "audio_content.h" +#include "dcpomatic_log.h" #include "log.h" #include "resampler.h" #include "compose.hpp" -#include #include #include "i18n.h" -#define LOG_GENERAL(...) _log->log (String::compose (__VA_ARGS__), LogEntry::TYPE_GENERAL); - using std::cout; using std::map; using std::pair; -using boost::shared_ptr; +using std::shared_ptr; +using std::make_shared; using boost::optional; +using namespace dcpomatic; -AudioDecoder::AudioDecoder (Decoder* parent, shared_ptr content, shared_ptr log) - : DecoderPart (parent, log) +AudioDecoder::AudioDecoder (Decoder* parent, shared_ptr content, bool fast) + : DecoderPart (parent) , _content (content) + , _fast (fast) { /* Set up _positions so that we have one for each stream */ - BOOST_FOREACH (AudioStreamPtr i, content->streams ()) { + for (auto i: content->streams ()) { _positions[i] = 0; } } +/** @param time_already_delayed true if the delay should not be added to time */ void -AudioDecoder::emit (AudioStreamPtr stream, shared_ptr data, ContentTime time) +AudioDecoder::emit (shared_ptr film, AudioStreamPtr stream, shared_ptr data, ContentTime time, bool time_already_delayed) { if (ignore ()) { return; } + /* Amount of error we will tolerate on audio timestamps; see comment below. + * We'll use 1 24fps video frame at 48kHz as this seems to be roughly how + * ffplay does it. + */ + static Frame const slack_frames = 48000 / 24; + + int const resampled_rate = _content->resampled_frame_rate(film); + if (!time_already_delayed) { + time += ContentTime::from_seconds (_content->delay() / 1000.0); + } + + auto reset = false; if (_positions[stream] == 0) { /* This is the first data we have received since initialisation or seek. Set the position based on the ContentTime that was given. After this first time - we just count samples, as it seems that ContentTimes are unreliable from - FFmpegDecoder (not quite continuous; perhaps due to some rounding error). + we just count samples unless the timestamp is more than slack_frames away + from where we think it should be. This is because ContentTimes seem to be + slightly unreliable from FFmpegDecoder (i.e. not sample accurate), but we still + need to obey them sometimes otherwise we get sync problems such as #1833. */ if (_content->delay() > 0) { /* Insert silence to give the delay */ silence (_content->delay ()); } - time += ContentTime::from_seconds (_content->delay() / 1000.0); - _positions[stream] = time.frames_round (stream->frame_rate ()); + reset = true; + } else if (std::abs(_positions[stream] - time.frames_round(resampled_rate)) > slack_frames) { + reset = true; + LOG_GENERAL ( + "Reset audio position: was %1, new data at %2, slack: %3 frames", + _positions[stream], + time.frames_round(resampled_rate), + std::abs(_positions[stream] - time.frames_round(resampled_rate)) + ); + } + + if (reset) { + _positions[stream] = time.frames_round (resampled_rate); } shared_ptr resampler; - map >::iterator i = _resamplers.find(stream); + auto i = _resamplers.find(stream); if (i != _resamplers.end ()) { resampler = i->second; } else { - if (stream->frame_rate() != _content->resampled_frame_rate()) { + if (stream->frame_rate() != resampled_rate) { LOG_GENERAL ( "Creating new resampler from %1 to %2 with %3 channels", stream->frame_rate(), - _content->resampled_frame_rate(), + resampled_rate, stream->channels() ); - resampler.reset (new Resampler (stream->frame_rate(), _content->resampled_frame_rate(), stream->channels())); + resampler = make_shared(stream->frame_rate(), resampled_rate, stream->channels()); + if (_fast) { + resampler->set_fast (); + } _resamplers[stream] = resampler; } } if (resampler) { - shared_ptr ro = resampler->run (data); + auto ro = resampler->run (data); if (ro->frames() == 0) { return; } data = ro; } - _positions[stream] += Data(stream, ContentAudio (data, _positions[stream])).get_value_or(0); + Data(stream, ContentAudio (data, _positions[stream])); + _positions[stream] += data->frames(); } /** @return Time just after the last thing that was emitted from a given stream */ ContentTime -AudioDecoder::stream_position (AudioStreamPtr stream) const +AudioDecoder::stream_position (shared_ptr film, AudioStreamPtr stream) const { - map::const_iterator i = _positions.find (stream); + auto i = _positions.find (stream); DCPOMATIC_ASSERT (i != _positions.end ()); - return ContentTime::from_frames (i->second, _content->resampled_frame_rate()); + return ContentTime::from_frames (i->second, _content->resampled_frame_rate(film)); } -ContentTime -AudioDecoder::position () const +boost::optional +AudioDecoder::position (shared_ptr film) const { optional p; - for (map::const_iterator i = _positions.begin(); i != _positions.end(); ++i) { - ContentTime const ct = stream_position (i->first); + for (auto i: _positions) { + auto const ct = stream_position (film, i.first); if (!p || ct < *p) { p = ct; } } - return p.get_value_or(ContentTime()); + return p; } void AudioDecoder::seek () { - for (map >::iterator i = _resamplers.begin(); i != _resamplers.end(); ++i) { - i->second->flush (); - i->second->reset (); + for (auto i: _resamplers) { + i.second->flush (); + i.second->reset (); } - for (map::iterator i = _positions.begin(); i != _positions.end(); ++i) { - i->second = 0; + for (auto& i: _positions) { + i.second = 0; } } void AudioDecoder::flush () { - for (map >::iterator i = _resamplers.begin(); i != _resamplers.end(); ++i) { - shared_ptr ro = i->second->flush (); + for (auto const& i: _resamplers) { + auto ro = i.second->flush (); if (ro->frames() > 0) { - Data (i->first, ContentAudio (ro, _positions[i->first])); - _positions[i->first] += ro->frames(); + Data (i.first, ContentAudio (ro, _positions[i.first])); + _positions[i.first] += ro->frames(); } } @@ -153,9 +184,9 @@ AudioDecoder::flush () void AudioDecoder::silence (int milliseconds) { - BOOST_FOREACH (AudioStreamPtr i, _content->streams ()) { + for (auto i: _content->streams()) { int const samples = ContentTime::from_seconds(milliseconds / 1000.0).frames_round(i->frame_rate()); - shared_ptr silence (new AudioBuffers (i->channels(), samples)); + auto silence = make_shared(i->channels(), samples); silence->make_silent (); Data (i, ContentAudio (silence, _positions[i])); }