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>
32 #elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
33 #include <libavfilter/vsrc_buffer.h>
35 #include <libavformat/avio.h>
40 #include "film_state.h"
42 #include "exceptions.h"
48 #include "delay_line.h"
49 #include "ffmpeg_compatibility.h"
52 using namespace boost;
54 /** @param s FilmState of the Film.
56 * @param j Job that we are running within, or 0
57 * @param l Log to use.
58 * @param minimal true to do the bare minimum of work; just run through the content. Useful for acquiring
59 * accurate frame counts as quickly as possible. This generates no video or audio output.
60 * @param ignore_length Ignore the content's claimed length when computing progress.
62 Decoder::Decoder (boost::shared_ptr<const FilmState> s, boost::shared_ptr<const Options> o, Job* j, Log* l, bool minimal, bool ignore_length)
68 , _ignore_length (ignore_length)
70 , _buffer_src_context (0)
71 , _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");
87 /** Start off a decode processing run */
89 Decoder::process_begin ()
91 _delay_in_bytes = _fs->audio_delay * _fs->audio_sample_rate * _fs->audio_channels * _fs->bytes_per_sample() / 1000;
93 _delay_line = new DelayLine (_delay_in_bytes);
95 _audio_frames_processed = 0;
98 /** Finish off a decode processing run */
100 Decoder::process_end ()
102 if (_delay_in_bytes < 0) {
103 uint8_t remainder[-_delay_in_bytes];
104 _delay_line->get_remaining (remainder);
105 _audio_frames_processed += _delay_in_bytes / (_fs->audio_channels * _fs->bytes_per_sample());
106 Audio (remainder, _delay_in_bytes);
109 /* If we cut the decode off, the audio may be short; push some silence
110 in to get it to the right length.
113 int64_t const audio_short_by_frames =
114 ((int64_t) decoding_frames() * _fs->audio_sample_rate / _fs->frames_per_second)
115 - _audio_frames_processed;
117 if (audio_short_by_frames >= 0) {
118 int bytes = audio_short_by_frames * _fs->audio_channels * _fs->bytes_per_sample();
120 int const silence_size = 64 * 1024;
121 uint8_t silence[silence_size];
122 memset (silence, 0, silence_size);
125 int const t = min (bytes, silence_size);
132 /** Start decoding */
138 if (_job && _ignore_length) {
139 _job->set_progress_unknown ();
142 while (pass () == false) {
143 if (_job && !_ignore_length) {
144 _job->set_progress (float (_video_frame) / decoding_frames ());
151 /** @return Number of frames that we will be decoding */
153 Decoder::decoding_frames () const
155 if (_opt->num_frames > 0) {
156 return _opt->num_frames;
162 /** Run one pass. This may or may not generate any actual video / audio data;
163 * some decoders may require several passes to generate a single frame.
164 * @return true if we have finished processing all data; otherwise false.
169 if (!_have_setup_video_filters) {
170 setup_video_filters ();
171 _have_setup_video_filters = true;
174 if (_opt->num_frames != 0 && _video_frame >= _opt->num_frames) {
181 /** Called by subclasses to tell the world that some audio data is ready
182 * @param data Interleaved audio data, in FilmState::audio_sample_format.
183 * @param size Number of bytes of data.
186 Decoder::process_audio (uint8_t* data, int size)
188 /* Samples per channel */
189 int const samples = size / _fs->bytes_per_sample();
191 /* Maybe apply gain */
192 if (_fs->audio_gain != 0) {
193 float const linear_gain = pow (10, _fs->audio_gain / 20);
195 switch (_fs->audio_sample_format) {
196 case AV_SAMPLE_FMT_S16:
197 for (int i = 0; i < samples; ++i) {
198 /* XXX: assumes little-endian; also we should probably be dithering here */
200 /* unsigned sample */
201 int const ou = p[0] | (p[1] << 8);
204 int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
206 /* signed sample with altered gain */
207 int const gs = int (os * linear_gain);
209 /* unsigned sample with altered gain */
210 int const gu = gs > 0 ? gs : (0x10000 + gs);
214 p[1] = (gu & 0xff00) >> 8;
223 /* Update the number of audio frames we've pushed to the encoder */
224 _audio_frames_processed += size / (_fs->audio_channels * _fs->bytes_per_sample ());
226 /* Push into the delay line and then tell the world what we've got */
227 int available = _delay_line->feed (data, size);
228 Audio (data, available);
231 /** Called by subclasses to tell the world that some video data is ready.
232 * We do some post-processing / filtering then emit it for listeners.
233 * @param frame to decode; caller manages memory.
236 Decoder::process_video (AVFrame* frame)
243 /* Use FilmState::length here as our one may be wrong */
246 if (_opt->decode_video_frequency != 0) {
247 gap = _fs->length / _opt->decode_video_frequency;
250 if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) {
255 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 61
257 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0) < 0) {
258 throw DecodeError ("could not push buffer into filter chain.");
261 #elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
264 par.num = sample_aspect_ratio_numerator ();
265 par.den = sample_aspect_ratio_denominator ();
267 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) {
268 throw DecodeError ("could not push buffer into filter chain.");
273 if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) {
274 throw DecodeError ("could not push buffer into filter chain.");
279 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15 && LIBAVFILTER_VERSION_MINOR <= 61
280 while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
282 while (av_buffersink_read (_buffer_sink_context, 0)) {
285 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15
287 int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
289 throw DecodeError ("could not request filtered frame");
292 AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf;
296 AVFilterBufferRef* filter_buffer;
297 if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) {
304 /* This takes ownership of filter_buffer */
305 shared_ptr<Image> image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer));
307 if (_opt->black_after > 0 && _video_frame > _opt->black_after) {
308 image->make_black ();
311 Video (image, _video_frame);
318 /** Set up a video filtering chain to include cropping and any filters that are specified
322 Decoder::setup_video_filters ()
325 Size size_after_crop;
327 if (_opt->apply_crop) {
328 size_after_crop = _fs->cropped_size (native_size ());
329 fs << crop_string (Position (_fs->crop.left, _fs->crop.top), size_after_crop);
331 size_after_crop = native_size ();
332 fs << crop_string (Position (0, 0), size_after_crop);
335 string filters = Filter::ffmpeg_strings (_fs->filters).first;
336 if (!filters.empty ()) {
340 filters += fs.str ();
342 avfilter_register_all ();
344 AVFilterGraph* graph = avfilter_graph_alloc();
346 throw DecodeError ("Could not create filter graph.");
349 AVFilter* buffer_src = avfilter_get_by_name("buffer");
350 if (buffer_src == 0) {
351 throw DecodeError ("Could not find buffer src filter");
354 AVFilter* buffer_sink = get_sink ();
357 a << native_size().width << ":"
358 << native_size().height << ":"
359 << pixel_format() << ":"
360 << time_base_numerator() << ":"
361 << time_base_denominator() << ":"
362 << sample_aspect_ratio_numerator() << ":"
363 << sample_aspect_ratio_denominator();
367 if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
368 throw DecodeError ("could not create buffer source");
371 AVBufferSinkParams* sink_params = av_buffersink_params_alloc ();
372 PixelFormat* pixel_fmts = new PixelFormat[2];
373 pixel_fmts[0] = pixel_format ();
374 pixel_fmts[1] = PIX_FMT_NONE;
375 sink_params->pixel_fmts = pixel_fmts;
377 if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, sink_params, graph) < 0) {
378 throw DecodeError ("could not create buffer sink.");
381 AVFilterInOut* outputs = avfilter_inout_alloc ();
382 outputs->name = av_strdup("in");
383 outputs->filter_ctx = _buffer_src_context;
384 outputs->pad_idx = 0;
387 AVFilterInOut* inputs = avfilter_inout_alloc ();
388 inputs->name = av_strdup("out");
389 inputs->filter_ctx = _buffer_sink_context;
393 _log->log ("Using filter chain `" + filters + "'");
395 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
396 if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) {
397 throw DecodeError ("could not set up filter graph.");
400 if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
401 throw DecodeError ("could not set up filter graph.");
405 if (avfilter_graph_config (graph, 0) < 0) {
406 throw DecodeError ("could not configure filter graph.");
409 /* XXX: leaking `inputs' / `outputs' ? */