2 Copyright (C) 2012 Carl Hetherington <cth@carlh.net>
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 /** @file src/decoder.cc
21 * @brief Parent class for decoders of content.
26 #include <boost/lexical_cast.hpp>
28 #include <libavfilter/avfiltergraph.h>
29 #include <libavfilter/buffersrc.h>
30 #if (LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 77) || LIBAVFILTER_VERSION_MAJOR == 3
31 #include <libavfilter/avcodec.h>
32 #include <libavfilter/buffersink.h>
33 #elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
34 #include <libavfilter/vsrc_buffer.h>
36 #include <libavformat/avio.h>
41 #include "film_state.h"
43 #include "exceptions.h"
49 #include "delay_line.h"
50 #include "ffmpeg_compatibility.h"
54 using namespace boost;
56 /** @param s FilmState of the Film.
58 * @param j Job that we are running within, or 0
59 * @param l Log to use.
60 * @param minimal true to do the bare minimum of work; just run through the content. Useful for acquiring
61 * accurate frame counts as quickly as possible. This generates no video or audio output.
62 * @param ignore_length Ignore the content's claimed length when computing progress.
64 Decoder::Decoder (boost::shared_ptr<const FilmState> s, boost::shared_ptr<const Options> o, Job* j, Log* l, bool minimal, bool ignore_length)
70 , _ignore_length (ignore_length)
72 , _buffer_src_context (0)
73 , _buffer_sink_context (0)
74 , _have_setup_video_filters (false)
77 , _audio_frames_processed (0)
79 if (_opt->decode_video_frequency != 0 && _fs->length() == 0) {
80 throw DecodeError ("cannot do a partial decode if length == 0");
89 /** Start off a decode processing run */
91 Decoder::process_begin ()
93 _delay_in_bytes = _fs->audio_delay() * _fs->audio_sample_rate() * _fs->audio_channels() * _fs->bytes_per_sample() / 1000;
95 _delay_line = new DelayLine (_delay_in_bytes);
97 _audio_frames_processed = 0;
100 /** Finish off a decode processing run */
102 Decoder::process_end ()
104 if (_delay_in_bytes < 0) {
105 uint8_t remainder[-_delay_in_bytes];
106 _delay_line->get_remaining (remainder);
107 _audio_frames_processed += _delay_in_bytes / (_fs->audio_channels() * _fs->bytes_per_sample());
108 emit_audio (remainder, _delay_in_bytes);
111 /* If we cut the decode off, the audio may be short; push some silence
112 in to get it to the right length.
115 int64_t const audio_short_by_frames =
116 ((int64_t) _fs->dcp_length() * _fs->target_sample_rate() / _fs->frames_per_second())
117 - _audio_frames_processed;
119 if (audio_short_by_frames >= 0) {
122 s << "Adding " << audio_short_by_frames << " frames of silence to the end.";
123 _log->log (s.str ());
125 int64_t bytes = audio_short_by_frames * _fs->audio_channels() * _fs->bytes_per_sample();
127 int64_t const silence_size = 64 * 1024;
128 uint8_t silence[silence_size];
129 memset (silence, 0, silence_size);
132 int64_t const t = min (bytes, silence_size);
133 emit_audio (silence, t);
139 /** Start decoding */
145 if (_job && _ignore_length) {
146 _job->set_progress_unknown ();
149 while (pass () == false) {
150 if (_job && !_ignore_length) {
151 _job->set_progress (float (_video_frame) / _fs->dcp_length());
158 /** Run one pass. This may or may not generate any actual video / audio data;
159 * some decoders may require several passes to generate a single frame.
160 * @return true if we have finished processing all data; otherwise false.
165 if (!_have_setup_video_filters) {
166 setup_video_filters ();
167 _have_setup_video_filters = true;
170 if (_video_frame >= _fs->dcp_length()) {
177 /** Called by subclasses to tell the world that some audio data is ready
178 * @param data Audio data, in FilmState::audio_sample_format.
179 * @param size Number of bytes of data.
182 Decoder::process_audio (uint8_t* data, int size)
184 /* Push into the delay line */
185 size = _delay_line->feed (data, size);
187 emit_audio (data, size);
191 Decoder::emit_audio (uint8_t* data, int size)
193 /* Deinterleave and convert to float */
195 float* samples[_fs->audio_channels()];
196 int const total_samples = size / _fs->bytes_per_sample();
197 int const frames = total_samples / _fs->audio_channels();
198 for (int i = 0; i < _fs->audio_channels(); ++i) {
199 samples[i] = new float[frames];
202 switch (_fs->audio_sample_format()) {
203 case AV_SAMPLE_FMT_S16:
208 for (int i = 0; i < total_samples; ++i) {
209 /* unsigned sample */
210 int const ou = p[0] | (p[1] << 8);
212 int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
214 samples[channel][sample] = float(os) / 0x8000;
216 cout << samples[channel][sample] << " from s16\n";
219 if (channel == _fs->audio_channels()) {
229 case AV_SAMPLE_FMT_FLTP:
231 float* p = reinterpret_cast<float*> (data);
232 for (int i = 0; i < _fs->audio_channels(); ++i) {
233 for (int j = 0; j < frames; ++j) {
234 samples[i][j] = *p++;
235 cout << samples[i][j] << " from float.\n";
246 /* Maybe apply gain */
247 if (_fs->audio_gain() != 0) {
248 float const linear_gain = pow (10, _fs->audio_gain() / 20);
249 for (int i = 0; i < _fs->audio_channels(); ++i) {
250 for (int j = 0; j < frames; ++j) {
251 samples[i][j] *= linear_gain;
256 /* Update the number of audio frames we've pushed to the encoder */
257 _audio_frames_processed += frames;
259 Audio (samples, frames);
261 for (int i = 0; i < _fs->audio_channels(); ++i) {
266 /** Called by subclasses to tell the world that some video data is ready.
267 * We do some post-processing / filtering then emit it for listeners.
268 * @param frame to decode; caller manages memory.
271 Decoder::process_video (AVFrame* frame)
278 /* Use FilmState::length here as our one may be wrong */
281 if (_opt->decode_video_frequency != 0) {
282 gap = _fs->length() / _opt->decode_video_frequency;
285 if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) {
290 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 61
292 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0) < 0) {
293 throw DecodeError ("could not push buffer into filter chain.");
296 #elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
299 par.num = sample_aspect_ratio_numerator ();
300 par.den = sample_aspect_ratio_denominator ();
302 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) {
303 throw DecodeError ("could not push buffer into filter chain.");
308 if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) {
309 throw DecodeError ("could not push buffer into filter chain.");
314 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15 && LIBAVFILTER_VERSION_MINOR <= 61
315 while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
317 while (av_buffersink_read (_buffer_sink_context, 0)) {
320 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15
322 int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
324 throw DecodeError ("could not request filtered frame");
327 AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf;
331 AVFilterBufferRef* filter_buffer;
332 if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) {
339 /* This takes ownership of filter_buffer */
340 shared_ptr<Image> image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer));
342 if (_opt->black_after > 0 && _video_frame > _opt->black_after) {
343 image->make_black ();
346 shared_ptr<Subtitle> sub;
347 if (_timed_subtitle && _timed_subtitle->displayed_at (double (last_video_frame()) / rint (_fs->frames_per_second()))) {
348 sub = _timed_subtitle->subtitle ();
351 TIMING ("Decoder emits %1", _video_frame);
352 Video (image, _video_frame, sub);
359 /** Set up a video filtering chain to include cropping and any filters that are specified
363 Decoder::setup_video_filters ()
366 Size size_after_crop;
368 if (_opt->apply_crop) {
369 size_after_crop = _fs->cropped_size (native_size ());
370 fs << crop_string (Position (_fs->crop().left, _fs->crop().top), size_after_crop);
372 size_after_crop = native_size ();
373 fs << crop_string (Position (0, 0), size_after_crop);
376 string filters = Filter::ffmpeg_strings (_fs->filters()).first;
377 if (!filters.empty ()) {
381 filters += fs.str ();
383 avfilter_register_all ();
385 AVFilterGraph* graph = avfilter_graph_alloc();
387 throw DecodeError ("Could not create filter graph.");
390 AVFilter* buffer_src = avfilter_get_by_name("buffer");
391 if (buffer_src == 0) {
392 throw DecodeError ("Could not find buffer src filter");
395 AVFilter* buffer_sink = get_sink ();
398 a << native_size().width << ":"
399 << native_size().height << ":"
400 << pixel_format() << ":"
401 << time_base_numerator() << ":"
402 << time_base_denominator() << ":"
403 << sample_aspect_ratio_numerator() << ":"
404 << sample_aspect_ratio_denominator();
408 if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
409 throw DecodeError ("could not create buffer source");
412 AVBufferSinkParams* sink_params = av_buffersink_params_alloc ();
413 PixelFormat* pixel_fmts = new PixelFormat[2];
414 pixel_fmts[0] = pixel_format ();
415 pixel_fmts[1] = PIX_FMT_NONE;
416 sink_params->pixel_fmts = pixel_fmts;
418 if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, sink_params, graph) < 0) {
419 throw DecodeError ("could not create buffer sink.");
422 AVFilterInOut* outputs = avfilter_inout_alloc ();
423 outputs->name = av_strdup("in");
424 outputs->filter_ctx = _buffer_src_context;
425 outputs->pad_idx = 0;
428 AVFilterInOut* inputs = avfilter_inout_alloc ();
429 inputs->name = av_strdup("out");
430 inputs->filter_ctx = _buffer_sink_context;
434 _log->log ("Using filter chain `" + filters + "'");
436 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
437 if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) {
438 throw DecodeError ("could not set up filter graph.");
441 if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
442 throw DecodeError ("could not set up filter graph.");
446 if (avfilter_graph_config (graph, 0) < 0) {
447 throw DecodeError ("could not configure filter graph.");
450 /* XXX: leaking `inputs' / `outputs' ? */
454 Decoder::process_subtitle (shared_ptr<TimedSubtitle> s)
458 if (_opt->apply_crop) {
459 Position const p = _timed_subtitle->subtitle()->position ();
460 _timed_subtitle->subtitle()->set_position (Position (p.x - _fs->crop().left, p.y - _fs->crop().top));