Merge master.
[dcpomatic.git] / src / lib / decoder.cc
index 34c04b260502ab75c424dc9635c2aee3e1515956..2f66aff68b4fb63fc5c5804472fe76eb9c41169d 100644 (file)
 using std::string;
 using std::stringstream;
 using std::min;
+using std::pair;
 using std::list;
 using boost::shared_ptr;
+using boost::optional;
 
 /** @param f Film.
  *  @param o Options.
  *  @param j Job that we are running within, or 0
+ *  @param minimal true to do the bare minimum of work; just run through the content.  Useful for acquiring
+ *  accurate frame counts as quickly as possible.  This generates no video or audio output.
  */
 Decoder::Decoder (boost::shared_ptr<Film> f, boost::shared_ptr<const Options> o, Job* j)
        : _film (f)
        , _opt (o)
        , _job (j)
-       , _video_frame_index (0)
+       , _video_frame (0)
+       , _audio_frame (0)
        , _delay_line (0)
-       , _delay_in_bytes (0)
-       , _audio_frames_processed (0)
+       , _delay_in_frames (0)
 {
        
 }
@@ -64,61 +68,67 @@ Decoder::~Decoder ()
        delete _delay_line;
 }
 
-/** Start off a decode processing run */
+/** Start off a decode processing run.  This should only be called once on
+ *  a given Decoder object.
+ */
 void
 Decoder::process_begin ()
 {
-       _delay_in_bytes = _film->audio_delay() * audio_sample_rate() * audio_channels() * bytes_per_audio_sample() / 1000;
-       delete _delay_line;
-       _delay_line = new DelayLine (_delay_in_bytes);
-
-       _audio_frames_processed = 0;
+       if (_audio_stream) {
+               _delay_in_frames = _film->audio_delay() * _audio_stream.get().sample_rate() / 1000;
+               _delay_line = new DelayLine (_audio_stream.get().channels(), _delay_in_frames);
+       }
 }
 
 /** Finish off a decode processing run */
 void
 Decoder::process_end ()
 {
-       if (_delay_in_bytes < 0) {
-               uint8_t remainder[-_delay_in_bytes];
-               _delay_line->get_remaining (remainder);
-               _audio_frames_processed += _delay_in_bytes / (audio_channels() * bytes_per_audio_sample());
-               emit_audio (remainder, -_delay_in_bytes);
+       if (_delay_in_frames < 0 && _opt->decode_audio && _audio_stream) {
+               shared_ptr<AudioBuffers> b (new AudioBuffers (_audio_stream.get().channels(), -_delay_in_frames));
+               b->make_silent ();
+               emit_audio (b);
        }
 
-       /* If we cut the decode off, the audio may be short; push some silence
-          in to get it to the right length.
-       */
+       if (_opt->decode_audio && _audio_stream) {
+
+               /* Ensure that our video and audio emissions are the same length */
+
+               int64_t audio_short_by_frames = video_frames_to_audio_frames (_video_frame, _audio_stream.get().sample_rate(), frames_per_second()) - _audio_frame;
+
+               _film->log()->log (
+                       String::compose (
+                               "Decoder has emitted %1 video frames (which equals %2 audio frames) and %3 audio frames",
+                               _video_frame,
+                               video_frames_to_audio_frames (_video_frame, _audio_stream.get().sample_rate(), frames_per_second()),
+                               _audio_frame
+                               )
+                       );
 
-       int64_t const video_length_in_audio_frames = ((int64_t) video_frame_index() * audio_sample_rate() / frames_per_second());
-       int64_t const audio_short_by_frames = video_length_in_audio_frames - _audio_frames_processed;
+               if (audio_short_by_frames < 0) {
 
-       _film->log()->log (
-               String::compose ("Source length is %1 (%2 audio frames); %3 frames of audio processed.",
-                                video_frame_index(),
-                                video_length_in_audio_frames,
-                                _audio_frames_processed)
-               );
+                       _film->log()->log (String::compose ("Emitted %1 too many audio frames", -audio_short_by_frames));
+                       
+                       /* We have emitted more audio than video.  Emit enough black video frames so that we reverse this */
+                       int const black_video_frames = ceil (-audio_short_by_frames * frames_per_second() / _audio_stream.get().sample_rate());
+
+                       _film->log()->log (String::compose ("Emitting %1 frames of black video", black_video_frames));
+
+                       shared_ptr<Image> black (new CompactImage (pixel_format(), native_size()));
+                       black->make_black ();
+                       for (int i = 0; i < black_video_frames; ++i) {
+                               emit_video (black, shared_ptr<Subtitle> ());
+                       }
+
+                       /* Now recompute our check value */
+                       audio_short_by_frames = video_frames_to_audio_frames (_video_frame, _audio_stream.get().sample_rate(), frames_per_second()) - _audio_frame;
+               }
        
-       if (audio_short_by_frames >= 0 && _opt->decode_audio) {
-
-               _film->log()->log (String::compose ("Source length is %1; %2 frames of audio processed.", video_frame_index(), _audio_frames_processed));
-               _film->log()->log (String::compose ("Adding %1 frames of silence to the end.", audio_short_by_frames));
-
-               /* XXX: this is slightly questionable; does memset () give silence with all
-                  sample formats?
-               */
-
-               int64_t bytes = audio_short_by_frames * _film->audio_channels() * bytes_per_audio_sample();
-               
-               int64_t const silence_size = 16 * 1024 * _film->audio_channels() * bytes_per_audio_sample();
-               uint8_t silence[silence_size];
-               memset (silence, 0, silence_size);
-               
-               while (bytes) {
-                       int64_t const t = min (bytes, silence_size);
-                       emit_audio (silence, t);
-                       bytes -= t;
+               if (audio_short_by_frames > 0) {
+                       _film->log()->log (String::compose ("Emitted %1 too few audio frames", audio_short_by_frames));
+                       shared_ptr<AudioBuffers> b (new AudioBuffers (_audio_stream.get().channels(), audio_short_by_frames));
+                       b->make_silent ();
+                       emit_audio (b);
                }
        }
 }
@@ -135,8 +145,7 @@ Decoder::go ()
 
        while (pass () == false) {
                if (_job && _film->dcp_length()) {
-                       SourceFrame const p = _video_frame_index - _film->dcp_trim_start();
-                       _job->set_progress (float (p) / _film->dcp_length().get());
+                       _job->set_progress (float (_video_frame) / _film->length().get());
                }
        }
 
@@ -148,94 +157,20 @@ Decoder::go ()
  *  @param size Number of bytes of data.
  */
 void
-Decoder::process_audio (uint8_t* data, int size)
-{
-       /* Push into the delay line */
-       size = _delay_line->feed (data, size);
-
-       emit_audio (data, size);
-}
-
-void
-Decoder::emit_audio (uint8_t* data, int size)
+Decoder::process_audio (shared_ptr<AudioBuffers> audio)
 {
-       if (size == 0) {
-               return;
-       }
-       
-       assert (_film->audio_channels());
-       assert (bytes_per_audio_sample());
-       
-       /* Deinterleave and convert to float */
-
-       assert ((size % (bytes_per_audio_sample() * audio_channels())) == 0);
-
-       int const total_samples = size / bytes_per_audio_sample();
-       int const frames = total_samples / _film->audio_channels();
-       shared_ptr<AudioBuffers> audio (new AudioBuffers (audio_channels(), frames));
-
-       switch (audio_sample_format()) {
-       case AV_SAMPLE_FMT_S16:
-       {
-               int16_t* p = (int16_t *) data;
-               int sample = 0;
-               int channel = 0;
-               for (int i = 0; i < total_samples; ++i) {
-                       audio->data(channel)[sample] = float(*p++) / (1 << 15);
-
-                       ++channel;
-                       if (channel == _film->audio_channels()) {
-                               channel = 0;
-                               ++sample;
-                       }
-               }
-       }
-       break;
-
-       case AV_SAMPLE_FMT_S32:
-       {
-               int32_t* p = (int32_t *) data;
-               int sample = 0;
-               int channel = 0;
-               for (int i = 0; i < total_samples; ++i) {
-                       audio->data(channel)[sample] = float(*p++) / (1 << 31);
-
-                       ++channel;
-                       if (channel == _film->audio_channels()) {
-                               channel = 0;
-                               ++sample;
-                       }
-               }
-       }
-
-       case AV_SAMPLE_FMT_FLTP:
-       {
-               float* p = reinterpret_cast<float*> (data);
-               for (int i = 0; i < _film->audio_channels(); ++i) {
-                       memcpy (audio->data(i), p, frames * sizeof(float));
-                       p += frames;
-               }
-       }
-       break;
-
-       default:
-               assert (false);
-       }
-
        /* Maybe apply gain */
        if (_film->audio_gain() != 0) {
                float const linear_gain = pow (10, _film->audio_gain() / 20);
-               for (int i = 0; i < _film->audio_channels(); ++i) {
-                       for (int j = 0; j < frames; ++j) {
+               for (int i = 0; i < audio->channels(); ++i) {
+                       for (int j = 0; j < audio->frames(); ++j) {
                                audio->data(i)[j] *= linear_gain;
                        }
                }
        }
 
-       /* Update the number of audio frames we've pushed to the encoder */
-       _audio_frames_processed += audio->frames ();
-
-       Audio (audio);
+       _delay_line->feed (audio);
+       emit_audio (audio);
 }
 
 /** Called by subclasses to tell the world that some video data is ready.
@@ -245,20 +180,6 @@ Decoder::emit_audio (uint8_t* data, int size)
 void
 Decoder::process_video (AVFrame* frame)
 {
-       assert (_film->length());
-       
-       /* Use Film::length here as our one may be wrong */
-
-       if (_opt->decode_video_skip != 0 && (_video_frame_index % _opt->decode_video_skip) != 0) {
-               ++_video_frame_index;
-               return;
-       }
-
-       if (_film->dcp_trim_start() > _video_frame_index || (_film->length().get() - _film->dcp_trim_end()) < _video_frame_index) {
-               ++_video_frame_index;
-               return;
-       }
-
        shared_ptr<FilterGraph> graph;
 
        list<shared_ptr<FilterGraph> >::iterator i = _filter_graphs.begin();
@@ -278,15 +199,11 @@ Decoder::process_video (AVFrame* frame)
 
        for (list<shared_ptr<Image> >::iterator i = images.begin(); i != images.end(); ++i) {
                shared_ptr<Subtitle> sub;
-               if (_timed_subtitle && _timed_subtitle->displayed_at (double (video_frame_index()) / _film->frames_per_second())) {
+               if (_timed_subtitle && _timed_subtitle->displayed_at (double (video_frame()) / _film->frames_per_second())) {
                        sub = _timed_subtitle->subtitle ();
                }
-               
-               TIMING ("Decoder emits %1", _video_frame_index);
-               Video (*i, _video_frame_index, sub);
-               ++_video_frame_index;
-               _last_image = *i;
-               _last_subtitle = sub;
+
+               emit_video (*i, sub);
        }
 }
 
@@ -297,9 +214,26 @@ Decoder::repeat_last_video ()
                _last_image.reset (new CompactImage (pixel_format(), native_size()));
                _last_image->make_black ();
        }
-       
-       Video (_last_image, _video_frame_index, _last_subtitle);
-       ++_video_frame_index;
+
+       emit_video (_last_image, _last_subtitle);
+}
+
+void
+Decoder::emit_video (shared_ptr<Image> image, shared_ptr<Subtitle> sub)
+{
+       TIMING ("Decoder emits %1", _video_frame);
+       Video (image, _video_frame, sub);
+       ++_video_frame;
+
+       _last_image = image;
+       _last_subtitle = sub;
+}
+
+void
+Decoder::emit_audio (shared_ptr<AudioBuffers> audio)
+{
+       Audio (audio, _audio_frame);
+       _audio_frame += audio->frames ();
 }
 
 void
@@ -313,9 +247,14 @@ Decoder::process_subtitle (shared_ptr<TimedSubtitle> s)
        }
 }
 
+void
+Decoder::set_audio_stream (optional<AudioStream> s)
+{
+       _audio_stream = s;
+}
 
-int
-Decoder::bytes_per_audio_sample () const
+void
+Decoder::set_subtitle_stream (optional<SubtitleStream> s)
 {
-       return av_get_bytes_per_sample (audio_sample_format ());
+       _subtitle_stream = s;
 }