Subs successfully exported with thumbs.
[dcpomatic.git] / src / lib / j2k_wav_encoder.cc
index 86c3ae13f5d39862f496914c07182cc8f5c8613e..7bcac483b7778aaa66ffb85bc9a0cde2432b6656 100644 (file)
@@ -105,13 +105,15 @@ J2KWAVEncoder::close_sound_files ()
 }      
 
 void
-J2KWAVEncoder::process_video (shared_ptr<Image> yuv, int frame)
+J2KWAVEncoder::process_video (shared_ptr<Image> yuv, int frame, shared_ptr<Subtitle> sub)
 {
        boost::mutex::scoped_lock lock (_worker_mutex);
 
        /* Wait until the queue has gone down a bit */
        while (_queue.size() >= _worker_threads.size() * 2 && !_process_end) {
+               TIMING ("decoder sleeps with queue of %1", _queue.size());
                _worker_condition.wait (lock);
+               TIMING ("decoder wakes with queue of %1", _queue.size());
        }
 
        if (_process_end) {
@@ -121,9 +123,10 @@ J2KWAVEncoder::process_video (shared_ptr<Image> yuv, int frame)
        /* Only do the processing if we don't already have a file for this frame */
        if (!boost::filesystem::exists (_opt->frame_out_path (frame, false))) {
                pair<string, string> const s = Filter::ffmpeg_strings (_fs->filters);
+               TIMING ("adding to queue of %1", _queue.size ());
                _queue.push_back (boost::shared_ptr<DCPVideoFrame> (
                                          new DCPVideoFrame (
-                                                 yuv, _opt->out_size, _opt->padding, _fs->scaler, frame, _fs->frames_per_second, s.second,
+                                                 yuv, sub, _opt->out_size, _opt->padding, _fs->scaler, frame, _fs->frames_per_second, s.second,
                                                  Config::instance()->colour_lut_index (), Config::instance()->j2k_bandwidth (),
                                                  _log
                                                  )
@@ -145,6 +148,8 @@ J2KWAVEncoder::encoder_thread (ServerDescription* server)
        int remote_backoff = 0;
        
        while (1) {
+
+               TIMING ("encoder thread %1 sleeps", boost::this_thread::get_id());
                boost::mutex::scoped_lock lock (_worker_mutex);
                while (_queue.empty () && !_process_end) {
                        _worker_condition.wait (lock);
@@ -154,7 +159,9 @@ J2KWAVEncoder::encoder_thread (ServerDescription* server)
                        return;
                }
 
+               TIMING ("encoder thread %1 wakes with queue of %2", boost::this_thread::get_id(), _queue.size());
                boost::shared_ptr<DCPVideoFrame> vf = _queue.front ();
+               _log->log (String::compose ("Encoder thread %1 pops frame %2 from queue", boost::this_thread::get_id(), vf->frame()));
                _queue.pop_front ();
                
                lock.unlock ();
@@ -166,9 +173,7 @@ J2KWAVEncoder::encoder_thread (ServerDescription* server)
                                encoded = vf->encode_remotely (server);
 
                                if (remote_backoff > 0) {
-                                       stringstream s;
-                                       s << server->host_name() << " was lost, but now she is found; removing backoff";
-                                       _log->log (s.str ());
+                                       _log->log (String::compose ("%1 was lost, but now she is found; removing backoff", server->host_name ()));
                                }
                                
                                /* This job succeeded, so remove any backoff */
@@ -179,18 +184,20 @@ J2KWAVEncoder::encoder_thread (ServerDescription* server)
                                        /* back off more */
                                        remote_backoff += 10;
                                }
-                               stringstream s;
-                               s << "Remote encode of " << vf->frame() << " on " << server->host_name() << " failed (" << e.what() << "); thread sleeping for " << remote_backoff << "s.";
-                               _log->log (s.str ());
+                               _log->log (
+                                       String::compose (
+                                               "Remote encode of %1 on %2 failed (%3); thread sleeping for %4s",
+                                               vf->frame(), server->host_name(), e.what(), remote_backoff)
+                                       );
                        }
                                
                } else {
                        try {
+                               TIMING ("encoder thread %1 begins local encode of %2", boost::this_thread::get_id(), vf->frame());
                                encoded = vf->encode_locally ();
+                               TIMING ("encoder thread %1 finishes local encode of %2", boost::this_thread::get_id(), vf->frame());
                        } catch (std::exception& e) {
-                               stringstream s;
-                               s << "Local encode failed " << e.what() << ".";
-                               _log->log (s.str ());
+                               _log->log (String::compose ("Local encode failed (%1)", e.what ()));
                        }
                }
 
@@ -199,6 +206,7 @@ J2KWAVEncoder::encoder_thread (ServerDescription* server)
                        frame_done (vf->frame ());
                } else {
                        lock.lock ();
+                       _log->log (String::compose ("Encoder thread %1 pushes frame %2 back onto queue after failure", boost::this_thread::get_id(), vf->frame()));
                        _queue.push_front (vf);
                        lock.unlock ();
                }
@@ -215,13 +223,18 @@ J2KWAVEncoder::encoder_thread (ServerDescription* server)
 void
 J2KWAVEncoder::process_begin (int64_t audio_channel_layout, AVSampleFormat audio_sample_format)
 {
-       if ((_fs->audio_sample_rate != dcp_audio_sample_rate (_fs->audio_sample_rate)) || (rint (_fs->frames_per_second) != _fs->frames_per_second)) {
-#ifdef HAVE_SWRESAMPLE         
+       if (_fs->audio_sample_rate != _fs->target_sample_rate ()) {
+#ifdef HAVE_SWRESAMPLE
+
+               stringstream s;
+               s << "Will resample audio from " << _fs->audio_sample_rate << " to " << _fs->target_sample_rate();
+               _log->log (s.str ());
+               
                _swr_context = swr_alloc_set_opts (
                        0,
                        audio_channel_layout,
                        audio_sample_format,
-                       target_sample_rate(),
+                       _fs->target_sample_rate(),
                        audio_channel_layout,
                        audio_sample_format,
                        _fs->audio_sample_rate,
@@ -281,28 +294,24 @@ J2KWAVEncoder::process_end ()
        */
 
        for (list<shared_ptr<DCPVideoFrame> >::iterator i = _queue.begin(); i != _queue.end(); ++i) {
-               stringstream s;
-               s << "Encode left-over frame " << (*i)->frame();
-               _log->log (s.str ());
+               _log->log (String::compose ("Encode left-over frame %1", (*i)->frame ()));
                try {
                        shared_ptr<EncodedData> e = (*i)->encode_locally ();
                        e->write (_opt, (*i)->frame ());
                        frame_done ((*i)->frame ());
                } catch (std::exception& e) {
-                       stringstream s;
-                       s << "Local encode failed " << e.what() << ".";
-                       _log->log (s.str ());
+                       _log->log (String::compose ("Local encode failed (%1)", e.what ()));
                }
        }
 
 #if HAVE_SWRESAMPLE    
        if (_swr_context) {
 
-               int mop = 0;
                while (1) {
                        uint8_t buffer[256 * _fs->bytes_per_sample() * _fs->audio_channels];
-                       uint8_t* out[1] = {
-                               buffer
+                       uint8_t* out[2] = {
+                               buffer,
+                               0
                        };
 
                        int const frames = swr_convert (_swr_context, out, 256, 0, 0);
@@ -315,8 +324,7 @@ J2KWAVEncoder::process_end ()
                                break;
                        }
 
-                       mop += frames;
-                       write_audio (buffer, frames);
+                       write_audio (buffer, frames * _fs->bytes_per_sample() * _fs->audio_channels);
                }
 
                swr_free (&_swr_context);
@@ -360,7 +368,7 @@ J2KWAVEncoder::process_audio (uint8_t* data, int size)
                int const frames = samples / _fs->audio_channels;
 
                /* Compute the resampled frame count and add 32 for luck */
-               int const out_buffer_size_frames = ceil (frames * target_sample_rate() / _fs->audio_sample_rate) + 32;
+               int const out_buffer_size_frames = ceil (frames * _fs->target_sample_rate() / _fs->audio_sample_rate) + 32;
                int const out_buffer_size_bytes = out_buffer_size_frames * _fs->audio_channels * _fs->bytes_per_sample();
                out_buffer = new uint8_t[out_buffer_size_bytes];
 
@@ -370,7 +378,7 @@ J2KWAVEncoder::process_audio (uint8_t* data, int size)
                };
 
                /* Resample audio */
-               int out_frames = swr_convert (_swr_context, out, out_buffer_size_frames, in, size);
+               int out_frames = swr_convert (_swr_context, out, out_buffer_size_frames, in, frames);
                if (out_frames < 0) {
                        throw EncodeError ("could not run sample-rate converter");
                }
@@ -390,12 +398,12 @@ J2KWAVEncoder::process_audio (uint8_t* data, int size)
 void
 J2KWAVEncoder::write_audio (uint8_t* data, int size)
 {
-       /* Size of a sample in bytes */
-       int const sample_size = 2;
-
-       /* XXX: we are assuming that sample_size is right, the _deinterleave_buffer_size is a multiple
-          of the sample size and that data_size is a multiple of _fs->audio_channels * sample_size.
+       /* XXX: we are assuming that the _deinterleave_buffer_size is a multiple
+          of the sample size and that size is a multiple of _fs->audio_channels * sample_size.
        */
+
+       assert ((size % (_fs->audio_channels * _fs->bytes_per_sample())) == 0);
+       assert ((_deinterleave_buffer_size % _fs->bytes_per_sample()) == 0);
        
        /* XXX: this code is very tricksy and it must be possible to make it simpler ... */
        
@@ -407,17 +415,17 @@ J2KWAVEncoder::write_audio (uint8_t* data, int size)
                /* How many bytes of the deinterleaved data to do this time */
                int this_time = min (remaining / _fs->audio_channels, _deinterleave_buffer_size);
                for (int i = 0; i < _fs->audio_channels; ++i) {
-                       for (int j = 0; j < this_time; j += sample_size) {
-                               for (int k = 0; k < sample_size; ++k) {
+                       for (int j = 0; j < this_time; j += _fs->bytes_per_sample()) {
+                               for (int k = 0; k < _fs->bytes_per_sample(); ++k) {
                                        int const to = j + k;
-                                       int const from = position + (i * sample_size) + (j * _fs->audio_channels) + k;
+                                       int const from = position + (i * _fs->bytes_per_sample()) + (j * _fs->audio_channels) + k;
                                        _deinterleave_buffer[to] = data[from];
                                }
                        }
                        
                        switch (_fs->audio_sample_format) {
                        case AV_SAMPLE_FMT_S16:
-                               sf_write_short (_sound_files[i], (const short *) _deinterleave_buffer, this_time / sample_size);
+                               sf_write_short (_sound_files[i], (const short *) _deinterleave_buffer, this_time / _fs->bytes_per_sample());
                                break;
                        default:
                                throw EncodeError ("unknown audio sample format");
@@ -429,22 +437,3 @@ J2KWAVEncoder::write_audio (uint8_t* data, int size)
        }
 }
 
-int
-J2KWAVEncoder::target_sample_rate () const
-{
-       double t = dcp_audio_sample_rate (_fs->audio_sample_rate);
-       if (rint (_fs->frames_per_second) != _fs->frames_per_second) {
-               if (_fs->frames_per_second == 23.976) {
-                       /* 24fps drop-frame ie 24 * 1000 / 1001 frames per second;
-                          hence we need to resample the audio to dcp_audio_sample_rate * 1000 / 1001
-                          so that when we play it back at dcp_audio_sample_rate it is sped up
-                          by the same amount that the video is
-                       */
-                       t *= double(1000) / 1001;
-               } else {
-                       throw EncodeError ("unknown fractional frame rate");
-               }
-       }
-
-       return rint (t);
-}