X-Git-Url: https://git.carlh.net/gitweb/?a=blobdiff_plain;f=src%2Flib%2Faudio_decoder.cc;h=61ff5d265c526e6de31e44d5bfb7eb6972883fba;hb=2b0e9dd97a5773f52eba5704903b82e90f4c6f63;hp=8d3b0e1288ea5b53627a9cfb82619b0ccc125c7e;hpb=58588c485eac7c488a574fe923576f109ec0134e;p=dcpomatic.git diff --git a/src/lib/audio_decoder.cc b/src/lib/audio_decoder.cc index 8d3b0e128..61ff5d265 100644 --- a/src/lib/audio_decoder.cc +++ b/src/lib/audio_decoder.cc @@ -1,89 +1,213 @@ /* - Copyright (C) 2012 Carl Hetherington + Copyright (C) 2012-2021 Carl Hetherington - This program is free software; you can redistribute it and/or modify + This file is part of DCP-o-matic. + + DCP-o-matic is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. - This program is distributed in the hope that it will be useful, + DCP-o-matic is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License - along with this program; if not, write to the Free Software - Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + along with DCP-o-matic. If not, see . */ + #include "audio_decoder.h" #include "audio_buffers.h" -#include "exceptions.h" +#include "audio_content.h" +#include "dcpomatic_log.h" #include "log.h" #include "resampler.h" -#include "util.h" -#include "film.h" +#include "compose.hpp" +#include #include "i18n.h" -using std::stringstream; -using std::list; -using std::pair; + using std::cout; +using std::shared_ptr; +using std::make_shared; using boost::optional; -using boost::shared_ptr; +using namespace dcpomatic; -AudioDecoder::AudioDecoder (shared_ptr film, shared_ptr content) - : Decoder (film) - , _audio_content (content) + +AudioDecoder::AudioDecoder (Decoder* parent, shared_ptr content, bool fast) + : DecoderPart (parent) + , _content (content) + , _fast (fast) { - if (content->output_audio_frame_rate() != content->content_audio_frame_rate() && content->audio_channels ()) { - _resampler.reset (new Resampler (content->content_audio_frame_rate(), content->output_audio_frame_rate(), content->audio_channels ())); + /* Set up _positions so that we have one for each stream */ + for (auto i: content->streams ()) { + _positions[i] = 0; } } -/** Audio timestamping is made hard by many factors, but the final nail in the coffin is resampling. - * We have to assume that we are feeding continuous data into the resampler, and so we get continuous - * data out. Hence we do the timestamping here, post-resampler, just by counting samples. - * - * The time is passed in here so that after a seek we can set up our _audio_position. The - * time is ignored once this has been done. - */ + +/** @param time_already_delayed true if the delay should not be added to time */ void -AudioDecoder::audio (shared_ptr data, ContentTime time) +AudioDecoder::emit(shared_ptr film, AudioStreamPtr stream, shared_ptr data, ContentTime time, bool flushing) +{ + if (ignore ()) { + return; + } + + int const resampled_rate = _content->resampled_frame_rate(film); + if (!flushing) { + time += ContentTime::from_seconds (_content->delay() / 1000.0); + } + + /* Amount of error we will tolerate on audio timestamps; see comment below. + * We'll use 1 24fps video frame as this seems to be roughly how ffplay does it. + */ + Frame const slack_frames = resampled_rate / 24; + + /* first_since_seek is set to true if this is the first data we have + received since initialisation or seek. We'll set the position based + on the ContentTime that was given. After this first time 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. + */ + + auto const first_since_seek = _positions[stream] == 0; + auto const need_reset = !first_since_seek && (std::abs(_positions[stream] - time.frames_round(resampled_rate)) > slack_frames); + + if (need_reset) { + 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 (first_since_seek || need_reset) { + _positions[stream] = time.frames_round (resampled_rate); + } + + if (first_since_seek && _content->delay() > 0) { + silence (stream, _content->delay()); + } + + shared_ptr resampler; + auto i = _resamplers.find(stream); + if (i != _resamplers.end()) { + resampler = i->second; + } else { + if (stream->frame_rate() != resampled_rate) { + LOG_GENERAL ( + "Creating new resampler from %1 to %2 with %3 channels", + stream->frame_rate(), + resampled_rate, + stream->channels() + ); + + resampler = make_shared(stream->frame_rate(), resampled_rate, stream->channels()); + if (_fast) { + resampler->set_fast (); + } + _resamplers[stream] = resampler; + } + } + + if (resampler && !flushing) { + /* It can be the the data here has a different number of channels than the stream + * it comes from (e.g. the files decoded by FFmpegDecoder sometimes have a random + * frame, often at the end, with more channels). Insert silence or discard channels + * here. + */ + if (resampler->channels() != data->channels()) { + LOG_WARNING("Received audio data with an unexpected channel count of %1 instead of %2", data->channels(), resampler->channels()); + auto data_copy = data->clone(); + data_copy->set_channels(resampler->channels()); + data = resampler->run(data_copy); + } else { + data = resampler->run(data); + } + + if (data->frames() == 0) { + return; + } + } + + 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 (shared_ptr film, AudioStreamPtr stream) const { - if (_resampler) { - data = _resampler->run (data); + auto i = _positions.find (stream); + DCPOMATIC_ASSERT (i != _positions.end ()); + return ContentTime::from_frames (i->second, _content->resampled_frame_rate(film)); +} + + +boost::optional +AudioDecoder::position (shared_ptr film) const +{ + optional p; + for (auto i: _positions) { + auto const ct = stream_position (film, i.first); + if (!p || ct < *p) { + p = ct; + } } - if (!_audio_position) { - shared_ptr film = _film.lock (); - assert (film); - FrameRateChange frc = film->active_frame_rate_change (_audio_content->position ()); - _audio_position = (double (time) / frc.speed_up) * film->audio_frame_rate() / TIME_HZ; + return p; +} + + +void +AudioDecoder::seek () +{ + for (auto i: _resamplers) { + i.second->flush (); + i.second->reset (); } - _pending.push_back (shared_ptr (new DecodedAudio (data, _audio_position.get ()))); - _audio_position = _audio_position.get() + data->frames (); + for (auto& i: _positions) { + i.second = 0; + } } + void AudioDecoder::flush () { - if (!_resampler) { - return; + 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(); + } } - shared_ptr b = _resampler->flush (); - if (b) { - _pending.push_back (shared_ptr (new DecodedAudio (b, _audio_position.get ()))); - _audio_position = _audio_position.get() + b->frames (); + if (_content->delay() < 0) { + /* Finish off with the gap caused by the delay */ + for (auto stream: _content->streams()) { + silence (stream, -_content->delay()); + } } } + void -AudioDecoder::seek (ContentTime, bool) +AudioDecoder::silence (AudioStreamPtr stream, int milliseconds) { - _audio_position.reset (); + int const samples = ContentTime::from_seconds(milliseconds / 1000.0).frames_round(stream->frame_rate()); + auto silence = make_shared(stream->channels(), samples); + silence->make_silent (); + Data (stream, ContentAudio(silence, _positions[stream])); }