X-Git-Url: https://main.carlh.net/gitweb/?a=blobdiff_plain;f=src%2Flib%2Faudio_decoder.cc;h=2d02043b5948d1071d0de18982e14abfcf03c12e;hb=ac34066d5e448d1984d11a180be74e31b6e13b5c;hp=9606c378c62d3babd09cf28c94b131c312aeaf39;hpb=b666a794a130386bc01ede2143ef40bd6973eb32;p=dcpomatic.git diff --git a/src/lib/audio_decoder.cc b/src/lib/audio_decoder.cc index 9606c378c..2d02043b5 100644 --- a/src/lib/audio_decoder.cc +++ b/src/lib/audio_decoder.cc @@ -1,135 +1,200 @@ /* - Copyright (C) 2012-2014 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 content) - : _audio_content (content) - , _decoded_audio (shared_ptr (new AudioBuffers (content->audio_channels(), 0)), 0) +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; } } -shared_ptr -AudioDecoder::get_audio (AudioFrame frame, AudioFrame length, bool accurate) -{ - shared_ptr dec; - AudioFrame const end = frame + length - 1; - - if (frame < _decoded_audio.frame || end > (_decoded_audio.frame + length * 4)) { - /* Either we have no decoded data, or what we do have is a long way from what we want: seek */ - seek (ContentTime::from_frames (frame, _audio_content->content_audio_frame_rate()), accurate); +/** @param time_already_delayed true if the delay should not be added to time */ +void +AudioDecoder::emit (shared_ptr film, AudioStreamPtr stream, shared_ptr data, ContentTime time, bool time_already_delayed) +{ + if (ignore ()) { + return; } - /* Now enough pass() calls will either: - * (a) give us what we want, or - * (b) hit the end of the decoder. - * - * If we are being accurate, we want the right frames, - * otherwise any frames will do. + /* 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. */ - if (accurate) { - while (!pass() && _decoded_audio.audio->frames() < length) {} + 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 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 ()); + } + 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; + auto i = _resamplers.find(stream); + if (i != _resamplers.end()) { + resampler = i->second; } else { - while (!pass() && (_decoded_audio.frame > frame || (_decoded_audio.frame + _decoded_audio.audio->frames()) < end)) {} + 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; + } } - - /* Clean up decoded */ - AudioFrame const decoded_offset = frame - _decoded_audio.frame; - AudioFrame const amount_left = _decoded_audio.audio->frames() - decoded_offset; - _decoded_audio.audio->move (decoded_offset, 0, amount_left); - _decoded_audio.audio->set_frames (amount_left); + if (resampler) { + auto ro = resampler->run (data); + if (ro->frames() == 0) { + return; + } + data = ro; + } + + Data(stream, ContentAudio (data, _positions[stream])); + _positions[stream] += data->frames(); +} - shared_ptr out (new AudioBuffers (_decoded_audio.audio->channels(), length)); - out->copy_from (_decoded_audio.audio.get(), length, frame - _decoded_audio.frame, 0); - return shared_ptr (new ContentAudio (out, frame)); +/** @return Time just after the last thing that was emitted from a given stream */ +ContentTime +AudioDecoder::stream_position (shared_ptr film, AudioStreamPtr stream) const +{ + auto i = _positions.find (stream); + DCPOMATIC_ASSERT (i != _positions.end ()); + return ContentTime::from_frames (i->second, _content->resampled_frame_rate(film)); } -/** Called by subclasses when audio data is ready. - * - * Audio timestamping is made hard by many factors, but perhaps the most entertaining 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. - */ -void -AudioDecoder::audio (shared_ptr data, ContentTime time) + +boost::optional +AudioDecoder::position (shared_ptr film) const { - if (_resampler) { - data = _resampler->run (data); + optional p; + for (auto i: _positions) { + auto const ct = stream_position (film, i.first); + if (!p || ct < *p) { + p = ct; + } } - if (!_audio_position) { - _audio_position = time.frames (_audio_content->output_audio_frame_rate ()); - } + return p; +} - assert (_audio_position >= (_decoded_audio.frame + _decoded_audio.audio->frames())); - /* Resize _decoded_audio to fit the new data */ - _decoded_audio.audio->ensure_size (_audio_position.get() + data->frames() - _decoded_audio.frame); +void +AudioDecoder::seek () +{ + for (auto i: _resamplers) { + i.second->flush (); + i.second->reset (); + } - /* Copy new data in */ - _decoded_audio.audio->copy_from (data.get(), data->frames(), 0, _audio_position.get() - _decoded_audio.frame); - _audio_position = _audio_position.get() + data->frames (); + for (auto& i: _positions) { + i.second = 0; + } } -/* XXX: called? */ + 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 */ + silence (-_content->delay ()); } - */ } + void -AudioDecoder::seek (ContentTime, bool) +AudioDecoder::silence (int milliseconds) { - _audio_position.reset (); + for (auto i: _content->streams()) { + int const samples = ContentTime::from_seconds(milliseconds / 1000.0).frames_round(i->frame_rate()); + auto silence = make_shared(i->channels(), samples); + silence->make_silent (); + Data (i, ContentAudio (silence, _positions[i])); + } }