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 #if (LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 77) || LIBAVFILTER_VERSION_MAJOR == 3
30 #include <libavfilter/avcodec.h>
31 #include <libavfilter/buffersink.h>
33 #include <libavformat/avio.h>
38 #include "film_state.h"
40 #include "exceptions.h"
46 #include "delay_line.h"
47 #include "ffmpeg_compatibility.h"
50 using namespace boost;
52 /** @param s FilmState of the Film.
54 * @param j Job that we are running within, or 0
55 * @param l Log to use.
56 * @param minimal true to do the bare minimum of work; just run through the content. Useful for acquiring
57 * accurate frame counts as quickly as possible. This generates no video or audio output.
58 * @param ignore_length Ignore the content's claimed length when computing progress.
60 Decoder::Decoder (boost::shared_ptr<const FilmState> s, boost::shared_ptr<const Options> o, Job* j, Log* l, bool minimal, bool ignore_length)
66 , _ignore_length (ignore_length)
68 , _buffer_src_context (0)
69 , _buffer_sink_context (0)
71 , _have_setup_video_filters (false)
74 , _audio_frames_processed (0)
76 if (_opt->decode_video_frequency != 0 && _fs->length == 0) {
77 throw DecodeError ("cannot do a partial decode if length == 0");
87 Decoder::process_begin ()
89 if (_fs->audio_sample_rate != dcp_audio_sample_rate (_fs->audio_sample_rate)) {
90 _swr_context = swr_alloc_set_opts (
92 audio_channel_layout(),
93 audio_sample_format(),
94 dcp_audio_sample_rate (_fs->audio_sample_rate),
95 audio_channel_layout(),
96 audio_sample_format(),
97 _fs->audio_sample_rate,
101 swr_init (_swr_context);
106 _delay_in_bytes = _fs->audio_delay * _fs->audio_sample_rate * _fs->audio_channels * _fs->bytes_per_sample() / 1000;
108 _delay_line = new DelayLine (_delay_in_bytes);
110 _audio_frames_processed = 0;
114 Decoder::process_end ()
120 uint8_t buffer[256 * _fs->bytes_per_sample() * _fs->audio_channels];
125 int const frames = swr_convert (_swr_context, out, 256, 0, 0);
128 throw DecodeError ("could not run sample-rate converter");
136 int available = _delay_line->feed (buffer, frames * _fs->audio_channels * _fs->bytes_per_sample());
137 Audio (buffer, available);
140 swr_free (&_swr_context);
143 if (_delay_in_bytes < 0) {
144 uint8_t remainder[-_delay_in_bytes];
145 _delay_line->get_remaining (remainder);
146 _audio_frames_processed += _delay_in_bytes / (_fs->audio_channels * _fs->bytes_per_sample());
147 Audio (remainder, _delay_in_bytes);
150 /* If we cut the decode off, the audio may be short; push some silence
151 in to get it to the right length.
154 int const audio_short_by_frames =
155 (decoding_frames() * dcp_audio_sample_rate (_fs->audio_sample_rate) / _fs->frames_per_second)
156 - _audio_frames_processed;
158 int bytes = audio_short_by_frames * _fs->audio_channels * _fs->bytes_per_sample();
160 int const silence_size = 64 * 1024;
161 uint8_t silence[silence_size];
162 memset (silence, 0, silence_size);
165 int const t = min (bytes, silence_size);
171 /** Start decoding */
177 if (_job && _ignore_length) {
178 _job->set_progress_unknown ();
181 while (pass () == false) {
182 if (_job && !_ignore_length) {
183 _job->set_progress (float (_video_frame) / decoding_frames ());
190 /** @return Number of frames that we will be decoding */
192 Decoder::decoding_frames () const
194 if (_opt->num_frames > 0) {
195 return _opt->num_frames;
201 /** Run one pass. This may or may not generate any actual video / audio data;
202 * some decoders may require several passes to generate a single frame.
203 * @return true if we have finished processing all data; otherwise false.
208 if (!_have_setup_video_filters) {
209 setup_video_filters ();
210 _have_setup_video_filters = true;
213 if (_opt->num_frames != 0 && _video_frame >= _opt->num_frames) {
220 /** Called by subclasses to tell the world that some audio data is ready
221 * @param data Interleaved audio data, in FilmState::audio_sample_format.
222 * @param size Number of bytes of data.
225 Decoder::process_audio (uint8_t* data, int size)
227 /* Here's samples per channel */
228 int const samples = size / _fs->bytes_per_sample();
230 /* And here's frames (where 1 frame is a collection of samples, 1 for each channel,
231 so for 5.1 a frame would be 6 samples)
233 int const frames = samples / _fs->audio_channels;
235 /* Maybe apply gain */
236 if (_fs->audio_gain != 0) {
237 float const linear_gain = pow (10, _fs->audio_gain / 20);
239 switch (_fs->audio_sample_format) {
240 case AV_SAMPLE_FMT_S16:
241 for (int i = 0; i < samples; ++i) {
242 /* XXX: assumes little-endian; also we should probably be dithering here */
244 /* unsigned sample */
245 int const ou = p[0] | (p[1] << 8);
248 int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
250 /* signed sample with altered gain */
251 int const gs = int (os * linear_gain);
253 /* unsigned sample with altered gain */
254 int const gu = gs > 0 ? gs : (0x10000 + gs);
258 p[1] = (gu & 0xff00) >> 8;
267 /* This is a buffer we might use if we are sample-rate converting;
268 it will need freeing if so.
270 uint8_t* out_buffer = 0;
272 /* Maybe sample-rate convert */
275 uint8_t const * in[2] = {
280 /* Compute the resampled frame count and add 32 for luck */
281 int const out_buffer_size_frames = ceil (frames * float (dcp_audio_sample_rate (_fs->audio_sample_rate)) / _fs->audio_sample_rate) + 32;
282 int const out_buffer_size_bytes = out_buffer_size_frames * _fs->audio_channels * _fs->bytes_per_sample();
283 out_buffer = new uint8_t[out_buffer_size_bytes];
291 int out_frames = swr_convert (_swr_context, out, out_buffer_size_frames, in, frames);
292 if (out_frames < 0) {
293 throw DecodeError ("could not run sample-rate converter");
296 /* And point our variables at the resampled audio */
298 size = out_frames * _fs->audio_channels * _fs->bytes_per_sample();
301 /* Update the number of audio frames we've pushed to the encoder */
302 _audio_frames_processed += size / (_fs->audio_channels * _fs->bytes_per_sample ());
304 /* Push into the delay line and then tell the world what we've got */
305 int available = _delay_line->feed (data, size);
306 Audio (data, available);
308 /* Delete the sample-rate conversion buffer, if it exists */
312 /** Called by subclasses to tell the world that some video data is ready.
313 * We do some post-processing / filtering then emit it for listeners.
314 * @param frame to decode; caller manages memory.
317 Decoder::process_video (AVFrame* frame)
324 /* Use FilmState::length here as our one may be wrong */
327 if (_opt->decode_video_frequency != 0) {
328 gap = _fs->length / _opt->decode_video_frequency;
331 if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) {
336 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 61
338 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0) < 0) {
339 throw DecodeError ("could not push buffer into filter chain.");
347 par.num = sample_aspect_ratio_numerator ();
348 par.den = sample_aspect_ratio_denominator ();
350 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) {
351 throw DecodeError ("could not push buffer into filter chain.");
356 if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) {
357 throw DecodeError ("could not push buffer into filter chain.");
362 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 23 && LIBAVFILTER_VERSION_MINOR <= 61
363 while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
365 while (av_buffersink_read (_buffer_sink_context, 0)) {
368 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 53
370 int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
372 throw DecodeError ("could not request filtered frame");
375 AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf;
379 AVFilterBufferRef* filter_buffer;
380 if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) {
387 /* This takes ownership of filter_buffer */
388 shared_ptr<Image> image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer));
390 if (_opt->black_after > 0 && _video_frame > _opt->black_after) {
391 image->make_black ();
394 Video (image, _video_frame);
401 /** Set up a video filtering chain to include cropping and any filters that are specified
405 Decoder::setup_video_filters ()
408 Size size_after_crop;
410 if (_opt->apply_crop) {
411 size_after_crop = _fs->cropped_size (native_size ());
412 fs << crop_string (Position (_fs->crop.left, _fs->crop.top), size_after_crop);
414 size_after_crop = native_size ();
415 fs << crop_string (Position (0, 0), size_after_crop);
418 string filters = Filter::ffmpeg_strings (_fs->filters).first;
419 if (!filters.empty ()) {
423 filters += fs.str ();
425 avfilter_register_all ();
427 AVFilterGraph* graph = avfilter_graph_alloc();
429 throw DecodeError ("Could not create filter graph.");
432 AVFilter* buffer_src = avfilter_get_by_name("buffer");
433 if (buffer_src == 0) {
434 throw DecodeError ("Could not find buffer src filter");
437 AVFilter* buffer_sink = avfilter_get_by_name("buffersink");
438 if (buffer_sink == 0) {
439 throw DecodeError ("Could not create buffer sink filter");
443 a << native_size().width << ":"
444 << native_size().height << ":"
445 << pixel_format() << ":"
446 << time_base_numerator() << ":"
447 << time_base_denominator() << ":"
448 << sample_aspect_ratio_numerator() << ":"
449 << sample_aspect_ratio_denominator();
452 if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
453 throw DecodeError ("could not create buffer source");
456 enum PixelFormat pixel_formats[] = { pixel_format(), PIX_FMT_NONE };
457 if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, pixel_formats, graph) < 0) {
458 throw DecodeError ("could not create buffer sink.");
461 AVFilterInOut* outputs = avfilter_inout_alloc ();
462 outputs->name = av_strdup("in");
463 outputs->filter_ctx = _buffer_src_context;
464 outputs->pad_idx = 0;
467 AVFilterInOut* inputs = avfilter_inout_alloc ();
468 inputs->name = av_strdup("out");
469 inputs->filter_ctx = _buffer_sink_context;
473 _log->log ("Using filter chain `" + filters + "'");
474 if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
475 throw DecodeError ("could not set up filter graph.");
478 if (avfilter_graph_config (graph, 0) < 0) {
479 throw DecodeError ("could not configure filter graph.");
482 /* XXX: leaking `inputs' / `outputs' ? */