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.
27 #include <libavfilter/avfiltergraph.h>
28 #include <libavfilter/buffersrc.h>
29 #ifndef DVDOMATIC_FFMPEG_0_8_3
30 #include <libavfilter/avcodec.h>
31 #include <libavfilter/buffersink.h>
33 #include <libavformat/avio.h>
35 #include <samplerate.h>
39 #include "film_state.h"
41 #include "exceptions.h"
47 #include "delay_line.h"
48 #include "ffmpeg_compatibility.h"
51 using namespace boost;
53 /** @param s FilmState of the Film.
55 * @param j Job that we are running within, or 0
56 * @param l Log to use.
57 * @param minimal true to do the bare minimum of work; just run through the content. Useful for acquiring
58 * accurate frame counts as quickly as possible. This generates no video or audio output.
59 * @param ignore_length Ignore the content's claimed length when computing progress.
61 Decoder::Decoder (boost::shared_ptr<const FilmState> s, boost::shared_ptr<const Options> o, Job* j, Log* l, bool minimal, bool ignore_length)
67 , _ignore_length (ignore_length)
69 , _buffer_src_context (0)
70 , _buffer_sink_context (0)
72 , _have_setup_video_filters (false)
75 , _audio_frames_processed (0)
77 if (_opt->decode_video_frequency != 0 && _fs->length == 0) {
78 throw DecodeError ("cannot do a partial decode if length == 0");
88 Decoder::process_begin ()
90 if (_fs->audio_sample_rate != dcp_audio_sample_rate (_fs->audio_sample_rate)) {
91 _swr_context = swr_alloc_set_opts (
93 audio_channel_layout(),
94 audio_sample_format(),
95 dcp_audio_sample_rate (_fs->audio_sample_rate),
96 audio_channel_layout(),
97 audio_sample_format(),
98 _fs->audio_sample_rate,
102 swr_init (_swr_context);
107 /* This assumes 2 bytes per sample */
108 _delay_in_bytes = _fs->audio_delay * _fs->audio_sample_rate * _fs->audio_channels * 2 / 1000;
110 _delay_line = new DelayLine (_delay_in_bytes);
112 _audio_frames_processed = 0;
116 Decoder::process_end ()
122 uint8_t buffer[256 * 2 * _fs->audio_channels];
127 int const frames = swr_convert (_swr_context, out, 256, 0, 0);
130 throw DecodeError ("could not run sample-rate converter");
138 int available = _delay_line->feed (buffer, frames * _fs->audio_channels * 2);
139 Audio (buffer, available);
142 cout << "mopped up " << mop << "\n";
144 swr_free (&_swr_context);
147 if (_delay_in_bytes < 0) {
148 uint8_t remainder[-_delay_in_bytes];
149 _delay_line->get_remaining (remainder);
150 _audio_frames_processed += _delay_in_bytes / (audio_channels() * 2);
151 Audio (remainder, _delay_in_bytes);
154 /* If we cut the decode off, the audio may be short; push some silence
155 in to get it to the right length.
158 int const audio_short_by_frames = (decoding_frames() * dcp_audio_sample_rate (_fs->audio_sample_rate) / _fs->frames_per_second) - _audio_frames_processed;
160 int bytes = audio_short_by_frames * audio_channels() * 2;
162 int const silence_size = 64 * 1024;
163 uint8_t silence[silence_size];
164 memset (silence, 0, silence_size);
167 int const t = min (bytes, silence_size);
173 /** Start decoding */
179 if (_job && _ignore_length) {
180 _job->set_progress_unknown ();
183 while (pass () == false) {
184 if (_job && !_ignore_length) {
185 _job->set_progress (float (_video_frame) / decoding_frames ());
192 /** @return Number of frames that we will be decoding */
194 Decoder::decoding_frames () const
196 if (_opt->num_frames > 0) {
197 return _opt->num_frames;
203 /** Run one pass. This may or may not generate any actual video / audio data;
204 * some decoders may require several passes to generate a single frame.
205 * @return true if we have finished processing all data; otherwise false.
210 if (!_have_setup_video_filters) {
211 setup_video_filters ();
212 _have_setup_video_filters = true;
215 if (_opt->num_frames != 0 && _video_frame >= _opt->num_frames) {
222 /** Called by subclasses to tell the world that some audio data is ready */
224 Decoder::process_audio (uint8_t* data, int channels, int size)
226 int const samples = size / 2;
227 int const frames = samples / channels;
229 if (_fs->audio_gain != 0) {
230 float const linear_gain = pow (10, _fs->audio_gain / 20);
232 switch (_fs->audio_sample_format) {
233 case AV_SAMPLE_FMT_S16:
234 for (int i = 0; i < samples; ++i) {
235 /* XXX: assumes little-endian; also we should probably be dithering here */
236 int const ou = p[0] | (p[1] << 8);
237 int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
238 int const gs = int (os * linear_gain);
239 int const gu = gs > 0 ? gs : (0x10000 + gs);
241 p[1] = (gu & 0xff00) >> 8;
252 uint8_t const * in[1] = {
256 int const out_buffer_size_frames = ceil (frames * float (dcp_audio_sample_rate (_fs->audio_sample_rate)) / _fs->audio_sample_rate) + 32;
257 int const out_buffer_size_bytes = out_buffer_size_frames * channels * 2;
258 uint8_t out_buffer[out_buffer_size_bytes];
264 int out_frames = swr_convert (_swr_context, out, out_buffer_size_frames, in, frames);
265 if (out_frames < 0) {
266 throw DecodeError ("could not run sample-rate converter");
269 size = out_frames * channels * 2;
272 /* Update the number of audio frames we've pushed to the encoder;
273 2 is the hard-coded (!) number of bytes per sample.
275 _audio_frames_processed += size / (channels * 2);
277 int available = _delay_line->feed (data, size);
278 Audio (data, available);
281 /** Called by subclasses to tell the world that some video data is ready.
282 * We do some post-processing / filtering then emit it for listeners.
283 * @param frame to decode; caller manages memory.
286 Decoder::process_video (AVFrame* frame)
293 /* Use FilmState::length here as our one may be wrong */
296 if (_opt->decode_video_frequency != 0) {
297 gap = _fs->length / _opt->decode_video_frequency;
300 if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) {
305 #ifdef DVDOMATIC_FFMPEG_0_8_3
308 par.num = sample_aspect_ratio_numerator ();
309 par.den = sample_aspect_ratio_denominator ();
311 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) {
312 throw DecodeError ("could not push buffer into filter chain.");
317 if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) {
318 throw DecodeError ("could not push buffer into filter chain.");
323 #ifdef DVDOMATIC_FFMPEG_0_8_3
324 while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
326 while (av_buffersink_read (_buffer_sink_context, 0)) {
329 #ifdef DVDOMATIC_FFMPEG_0_8_3
331 int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
333 throw DecodeError ("could not request filtered frame");
336 AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf;
340 AVFilterBufferRef* filter_buffer;
341 if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) {
348 /* This takes ownership of filter_buffer */
349 shared_ptr<Image> image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer));
351 if (_opt->black_after > 0 && _video_frame > _opt->black_after) {
352 image->make_black ();
355 Video (image, _video_frame);
362 Decoder::setup_video_filters ()
365 Size size_after_crop;
367 if (_opt->apply_crop) {
368 size_after_crop = _fs->cropped_size (native_size ());
369 fs << crop_string (Position (_fs->left_crop, _fs->top_crop), size_after_crop);
371 size_after_crop = native_size ();
372 fs << crop_string (Position (0, 0), size_after_crop);
375 string filters = Filter::ffmpeg_strings (_fs->filters).first;
376 if (!filters.empty ()) {
380 filters += fs.str ();
382 avfilter_register_all ();
384 AVFilterGraph* graph = avfilter_graph_alloc();
386 throw DecodeError ("Could not create filter graph.");
389 AVFilter* buffer_src = avfilter_get_by_name("buffer");
390 if (buffer_src == 0) {
391 throw DecodeError ("Could not find buffer src filter");
394 AVFilter* buffer_sink = get_sink ();
397 a << native_size().width << ":"
398 << native_size().height << ":"
399 << pixel_format() << ":"
400 << time_base_numerator() << ":"
401 << time_base_denominator() << ":"
402 << sample_aspect_ratio_numerator() << ":"
403 << sample_aspect_ratio_denominator();
406 if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
407 throw DecodeError ("could not create buffer source");
410 enum PixelFormat pixel_formats[] = { pixel_format(), PIX_FMT_NONE };
411 if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, pixel_formats, graph) < 0) {
412 throw DecodeError ("could not create buffer sink.");
415 AVFilterInOut* outputs = avfilter_inout_alloc ();
416 outputs->name = av_strdup("in");
417 outputs->filter_ctx = _buffer_src_context;
418 outputs->pad_idx = 0;
421 AVFilterInOut* inputs = avfilter_inout_alloc ();
422 inputs->name = av_strdup("out");
423 inputs->filter_ctx = _buffer_sink_context;
427 _log->log ("Using filter chain `" + filters + "'");
428 #ifdef DVDOMATIC_FFMPEG_0_8_3
429 if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) {
431 if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
434 throw DecodeError ("could not set up filter graph.");
437 if (avfilter_graph_config (graph, 0) < 0) {
438 throw DecodeError ("could not configure filter graph.");
441 /* XXX: leaking `inputs' / `outputs' ? */