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() * bytes_per_audio_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() * bytes_per_audio_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 && _opt->decode_audio) {
121 _log->log (String::compose ("DCP length is %1; %2 frames of audio processed.", _fs->dcp_length(), _audio_frames_processed));
122 _log->log (String::compose ("Adding %1 frames of silence to the end.", audio_short_by_frames));
124 int64_t bytes = audio_short_by_frames * _fs->audio_channels() * bytes_per_audio_sample();
126 int64_t const silence_size = 16 * 1024 * _fs->audio_channels() * bytes_per_audio_sample();
127 uint8_t silence[silence_size];
128 memset (silence, 0, silence_size);
131 int64_t const t = min (bytes, silence_size);
132 emit_audio (silence, t);
138 /** Start decoding */
144 if (_job && _ignore_length) {
145 _job->set_progress_unknown ();
148 while (pass () == false) {
149 if (_job && !_ignore_length) {
150 _job->set_progress (float (_video_frame) / _fs->dcp_length());
157 /** Run one pass. This may or may not generate any actual video / audio data;
158 * some decoders may require several passes to generate a single frame.
159 * @return true if we have finished processing all data; otherwise false.
164 if (!_have_setup_video_filters) {
165 setup_video_filters ();
166 _have_setup_video_filters = true;
169 if (_video_frame >= _fs->dcp_length()) {
176 /** Called by subclasses to tell the world that some audio data is ready
177 * @param data Audio data, in FilmState::audio_sample_format.
178 * @param size Number of bytes of data.
181 Decoder::process_audio (uint8_t* data, int size)
183 /* Push into the delay line */
184 size = _delay_line->feed (data, size);
186 emit_audio (data, size);
190 Decoder::emit_audio (uint8_t* data, int size)
192 /* Deinterleave and convert to float */
194 assert ((size % (bytes_per_audio_sample() * _fs->audio_channels())) == 0);
196 int const total_samples = size / bytes_per_audio_sample();
197 int const frames = total_samples / _fs->audio_channels();
198 shared_ptr<AudioBuffers> audio (new AudioBuffers (_fs->audio_channels(), frames));
200 switch (audio_sample_format()) {
201 case AV_SAMPLE_FMT_S16:
206 for (int i = 0; i < total_samples; ++i) {
207 /* unsigned sample */
208 int const ou = p[0] | (p[1] << 8);
210 int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
212 audio->data(channel)[sample] = float(os) / 0x8000;
215 if (channel == _fs->audio_channels()) {
225 case AV_SAMPLE_FMT_FLTP:
227 float* p = reinterpret_cast<float*> (data);
228 for (int i = 0; i < _fs->audio_channels(); ++i) {
229 memcpy (audio->data(i), p, frames * sizeof(float));
239 /* Maybe apply gain */
240 if (_fs->audio_gain() != 0) {
241 float const linear_gain = pow (10, _fs->audio_gain() / 20);
242 for (int i = 0; i < _fs->audio_channels(); ++i) {
243 for (int j = 0; j < frames; ++j) {
244 audio->data(i)[j] *= linear_gain;
249 /* Update the number of audio frames we've pushed to the encoder */
250 _audio_frames_processed += frames;
255 /** Called by subclasses to tell the world that some video data is ready.
256 * We do some post-processing / filtering then emit it for listeners.
257 * @param frame to decode; caller manages memory.
260 Decoder::process_video (AVFrame* frame)
267 /* Use FilmState::length here as our one may be wrong */
270 if (_opt->decode_video_frequency != 0) {
271 gap = _fs->length() / _opt->decode_video_frequency;
274 if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) {
279 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 61
281 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0) < 0) {
282 throw DecodeError ("could not push buffer into filter chain.");
285 #elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
288 par.num = sample_aspect_ratio_numerator ();
289 par.den = sample_aspect_ratio_denominator ();
291 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) {
292 throw DecodeError ("could not push buffer into filter chain.");
297 if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) {
298 throw DecodeError ("could not push buffer into filter chain.");
303 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15 && LIBAVFILTER_VERSION_MINOR <= 61
304 while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
306 while (av_buffersink_read (_buffer_sink_context, 0)) {
309 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15
311 int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
313 throw DecodeError ("could not request filtered frame");
316 AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf;
320 AVFilterBufferRef* filter_buffer;
321 if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) {
328 /* This takes ownership of filter_buffer */
329 shared_ptr<Image> image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer));
331 if (_opt->black_after > 0 && _video_frame > _opt->black_after) {
332 image->make_black ();
335 shared_ptr<Subtitle> sub;
336 if (_timed_subtitle && _timed_subtitle->displayed_at (double (last_video_frame()) / rint (_fs->frames_per_second()))) {
337 sub = _timed_subtitle->subtitle ();
340 TIMING ("Decoder emits %1", _video_frame);
341 Video (image, _video_frame, sub);
348 /** Set up a video filtering chain to include cropping and any filters that are specified
352 Decoder::setup_video_filters ()
355 Size size_after_crop;
357 if (_opt->apply_crop) {
358 size_after_crop = _fs->cropped_size (native_size ());
359 fs << crop_string (Position (_fs->crop().left, _fs->crop().top), size_after_crop);
361 size_after_crop = native_size ();
362 fs << crop_string (Position (0, 0), size_after_crop);
365 string filters = Filter::ffmpeg_strings (_fs->filters()).first;
366 if (!filters.empty ()) {
370 filters += fs.str ();
372 avfilter_register_all ();
374 AVFilterGraph* graph = avfilter_graph_alloc();
376 throw DecodeError ("Could not create filter graph.");
379 AVFilter* buffer_src = avfilter_get_by_name("buffer");
380 if (buffer_src == 0) {
381 throw DecodeError ("Could not find buffer src filter");
384 AVFilter* buffer_sink = get_sink ();
387 a << native_size().width << ":"
388 << native_size().height << ":"
389 << pixel_format() << ":"
390 << time_base_numerator() << ":"
391 << time_base_denominator() << ":"
392 << sample_aspect_ratio_numerator() << ":"
393 << sample_aspect_ratio_denominator();
397 if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
398 throw DecodeError ("could not create buffer source");
401 AVBufferSinkParams* sink_params = av_buffersink_params_alloc ();
402 PixelFormat* pixel_fmts = new PixelFormat[2];
403 pixel_fmts[0] = pixel_format ();
404 pixel_fmts[1] = PIX_FMT_NONE;
405 sink_params->pixel_fmts = pixel_fmts;
407 if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, sink_params, graph) < 0) {
408 throw DecodeError ("could not create buffer sink.");
411 AVFilterInOut* outputs = avfilter_inout_alloc ();
412 outputs->name = av_strdup("in");
413 outputs->filter_ctx = _buffer_src_context;
414 outputs->pad_idx = 0;
417 AVFilterInOut* inputs = avfilter_inout_alloc ();
418 inputs->name = av_strdup("out");
419 inputs->filter_ctx = _buffer_sink_context;
423 _log->log ("Using filter chain `" + filters + "'");
425 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
426 if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) {
427 throw DecodeError ("could not set up filter graph.");
430 if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
431 throw DecodeError ("could not set up filter graph.");
435 if (avfilter_graph_config (graph, 0) < 0) {
436 throw DecodeError ("could not configure filter graph.");
439 /* XXX: leaking `inputs' / `outputs' ? */
443 Decoder::process_subtitle (shared_ptr<TimedSubtitle> s)
447 if (_opt->apply_crop) {
448 Position const p = _timed_subtitle->subtitle()->position ();
449 _timed_subtitle->subtitle()->set_position (Position (p.x - _fs->crop().left, p.y - _fs->crop().top));
455 Decoder::bytes_per_audio_sample () const
457 return av_get_bytes_per_sample (audio_sample_format ());