More accurate calculation of export video pts; may fix #1663.
[dcpomatic.git] / src / lib / audio_decoder.cc
index e1c93ac77bcb23de9a5888cdc2babd36ffdef442..a5e86f22b8e352d529fa39e69149ad10cca9fec4 100644 (file)
 /*
-    Copyright (C) 2012 Carl Hetherington <cth@carlh.net>
+    Copyright (C) 2012-2018 Carl Hetherington <cth@carlh.net>
 
-    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 <http://www.gnu.org/licenses/>.
 
 */
 
 #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 "compose.hpp"
+#include <boost/foreach.hpp>
+#include <iostream>
 
 #include "i18n.h"
 
-using std::stringstream;
-using boost::optional;
+using std::cout;
+using std::map;
+using std::pair;
 using boost::shared_ptr;
+using boost::optional;
+using namespace dcpomatic;
+
+AudioDecoder::AudioDecoder (Decoder* parent, shared_ptr<const AudioContent> 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 ()) {
+               _positions[i] = 0;
+       }
+}
 
-AudioDecoder::AudioDecoder (shared_ptr<const Film> f, shared_ptr<const AudioContent> c)
-       : Decoder (f)
-       , _audio_content (c)
-       , _output_audio_frame_rate (_audio_content->output_audio_frame_rate (f))
+void
+AudioDecoder::emit (shared_ptr<const Film> film, AudioStreamPtr stream, shared_ptr<const AudioBuffers> data, ContentTime time)
 {
-       if (_audio_content->content_audio_frame_rate() != _output_audio_frame_rate) {
-
-               stringstream s;
-               s << String::compose ("Will resample audio from %1 to %2", _audio_content->content_audio_frame_rate(), _output_audio_frame_rate);
-               _film->log()->log (s.str ());
-
-               /* We will be using planar float data when we call the
-                  resampler.  As far as I can see, the audio channel
-                  layout is not necessary for our purposes; it seems
-                  only to be used get the number of channels and
-                  decide if rematrixing is needed.  It won't be, since
-                  input and output layouts are the same.
+       if (ignore ()) {
+               return;
+       }
+
+       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).
                */
+               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 (_content->resampled_frame_rate(film));
+       }
 
-               _swr_context = swr_alloc_set_opts (
-                       0,
-                       av_get_default_channel_layout (MAX_AUDIO_CHANNELS),
-                       AV_SAMPLE_FMT_FLTP,
-                       _output_audio_frame_rate,
-                       av_get_default_channel_layout (MAX_AUDIO_CHANNELS),
-                       AV_SAMPLE_FMT_FLTP,
-                       _audio_content->content_audio_frame_rate(),
-                       0, 0
-                       );
-               
-               swr_init (_swr_context);
+       shared_ptr<Resampler> resampler;
+       ResamplerMap::iterator i = _resamplers.find(stream);
+       if (i != _resamplers.end ()) {
+               resampler = i->second;
        } else {
-               _swr_context = 0;
+               if (stream->frame_rate() != _content->resampled_frame_rate(film)) {
+                       LOG_GENERAL (
+                               "Creating new resampler from %1 to %2 with %3 channels",
+                               stream->frame_rate(),
+                               _content->resampled_frame_rate(film),
+                               stream->channels()
+                               );
+
+                       resampler.reset (new Resampler (stream->frame_rate(), _content->resampled_frame_rate(film), stream->channels()));
+                       if (_fast) {
+                               resampler->set_fast ();
+                       }
+                       _resamplers[stream] = resampler;
+               }
        }
-}
 
-AudioDecoder::~AudioDecoder ()
-{
-       if (_swr_context) {
-               swr_free (&_swr_context);
+       if (resampler) {
+               shared_ptr<const AudioBuffers> ro = resampler->run (data);
+               if (ro->frames() == 0) {
+                       return;
+               }
+               data = ro;
        }
+
+       Data(stream, ContentAudio (data, _positions[stream]));
+       _positions[stream] += data->frames();
 }
-       
 
-#if 0
-void
-AudioDecoder::process_end ()
+/** @return Time just after the last thing that was emitted from a given stream */
+ContentTime
+AudioDecoder::stream_position (shared_ptr<const Film> film, AudioStreamPtr stream) const
 {
-       if (_swr_context) {
-
-               shared_ptr<AudioBuffers> out (new AudioBuffers (_film->audio_mapping().dcp_channels(), 256));
-                       
-               while (1) {
-                       int const frames = swr_convert (_swr_context, (uint8_t **) out->data(), 256, 0, 0);
+       PositionMap::const_iterator i = _positions.find (stream);
+       DCPOMATIC_ASSERT (i != _positions.end ());
+       return ContentTime::from_frames (i->second, _content->resampled_frame_rate(film));
+}
 
-                       if (frames < 0) {
-                               throw EncodeError (_("could not run sample-rate converter"));
-                       }
+boost::optional<ContentTime>
+AudioDecoder::position (shared_ptr<const Film> film) const
+{
+       optional<ContentTime> p;
+       for (PositionMap::const_iterator i = _positions.begin(); i != _positions.end(); ++i) {
+               ContentTime const ct = stream_position (film, i->first);
+               if (!p || ct < *p) {
+                       p = ct;
+               }
+       }
 
-                       if (frames == 0) {
-                               break;
-                       }
+       return p;
+}
 
-                       out->set_frames (frames);
-                       _writer->write (out);
-               }
+void
+AudioDecoder::seek ()
+{
+       for (ResamplerMap::iterator i = _resamplers.begin(); i != _resamplers.end(); ++i) {
+               i->second->flush ();
+               i->second->reset ();
+       }
 
+       for (PositionMap::iterator i = _positions.begin(); i != _positions.end(); ++i) {
+               i->second = 0;
        }
 }
-#endif
 
 void
-AudioDecoder::emit_audio (shared_ptr<const AudioBuffers> data, Time time)
+AudioDecoder::flush ()
 {
-       /* XXX: map audio to 5.1 */
-       
-       /* Maybe sample-rate convert */
-       if (_swr_context) {
-
-               /* Compute the resampled frames count and add 32 for luck */
-               int const max_resampled_frames = ceil ((int64_t) data->frames() * _output_audio_frame_rate / _audio_content->content_audio_frame_rate()) + 32;
-
-               shared_ptr<AudioBuffers> resampled (new AudioBuffers (MAX_AUDIO_CHANNELS, max_resampled_frames));
-
-               /* Resample audio */
-               int const resampled_frames = swr_convert (
-                       _swr_context, (uint8_t **) resampled->data(), max_resampled_frames, (uint8_t const **) data->data(), data->frames()
-                       );
-               
-               if (resampled_frames < 0) {
-                       throw EncodeError (_("could not run sample-rate converter"));
+       for (ResamplerMap::iterator i = _resamplers.begin(); i != _resamplers.end(); ++i) {
+               shared_ptr<const AudioBuffers> ro = i->second->flush ();
+               if (ro->frames() > 0) {
+                       Data (i->first, ContentAudio (ro, _positions[i->first]));
+                       _positions[i->first] += ro->frames();
                }
-
-               resampled->set_frames (resampled_frames);
-               
-               /* And point our variables at the resampled audio */
-               data = resampled;
        }
 
-       Audio (data, time);
+       if (_content->delay() < 0) {
+               /* Finish off with the gap caused by the delay */
+               silence (-_content->delay ());
+       }
 }
 
-               
+void
+AudioDecoder::silence (int milliseconds)
+{
+       BOOST_FOREACH (AudioStreamPtr i, _content->streams ()) {
+               int const samples = ContentTime::from_seconds(milliseconds / 1000.0).frames_round(i->frame_rate());
+               shared_ptr<AudioBuffers> silence (new AudioBuffers (i->channels(), samples));
+               silence->make_silent ();
+               Data (i, ContentAudio (silence, _positions[i]));
+       }
+}