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 _delay_in_bytes = _fs->audio_delay * _fs->audio_sample_rate * _fs->audio_channels * _fs->bytes_per_sample() / 1000;
109 _delay_line = new DelayLine (_delay_in_bytes);
111 _audio_frames_processed = 0;
115 Decoder::process_end ()
121 uint8_t buffer[256 * _fs->bytes_per_sample() * _fs->audio_channels];
126 int const frames = swr_convert (_swr_context, out, 256, 0, 0);
129 throw DecodeError ("could not run sample-rate converter");
137 int available = _delay_line->feed (buffer, frames * _fs->audio_channels * _fs->bytes_per_sample());
138 Audio (buffer, available);
141 cout << "mopped up " << mop << "\n";
143 swr_free (&_swr_context);
146 if (_delay_in_bytes < 0) {
147 uint8_t remainder[-_delay_in_bytes];
148 _delay_line->get_remaining (remainder);
149 _audio_frames_processed += _delay_in_bytes / (_fs->audio_channels * _fs->bytes_per_sample());
150 Audio (remainder, _delay_in_bytes);
153 /* If we cut the decode off, the audio may be short; push some silence
154 in to get it to the right length.
157 int const audio_short_by_frames =
158 (decoding_frames() * dcp_audio_sample_rate (_fs->audio_sample_rate) / _fs->frames_per_second)
159 - _audio_frames_processed;
161 int bytes = audio_short_by_frames * _fs->audio_channels * _fs->bytes_per_sample();
163 int const silence_size = 64 * 1024;
164 uint8_t silence[silence_size];
165 memset (silence, 0, silence_size);
168 int const t = min (bytes, silence_size);
174 /** Start decoding */
180 if (_job && _ignore_length) {
181 _job->set_progress_unknown ();
184 while (pass () == false) {
185 if (_job && !_ignore_length) {
186 _job->set_progress (float (_video_frame) / decoding_frames ());
193 /** @return Number of frames that we will be decoding */
195 Decoder::decoding_frames () const
197 if (_opt->num_frames > 0) {
198 return _opt->num_frames;
204 /** Run one pass. This may or may not generate any actual video / audio data;
205 * some decoders may require several passes to generate a single frame.
206 * @return true if we have finished processing all data; otherwise false.
211 if (!_have_setup_video_filters) {
212 setup_video_filters ();
213 _have_setup_video_filters = true;
216 if (_opt->num_frames != 0 && _video_frame >= _opt->num_frames) {
223 /** Called by subclasses to tell the world that some audio data is ready
224 * @param data Interleaved audio data, in FilmState::audio_sample_format.
225 * @param size Number of bytes of data.
228 Decoder::process_audio (uint8_t* data, int size)
230 /* Here's samples per channel */
231 int const samples = size / _fs->bytes_per_sample();
233 /* And here's frames (where 1 frame is a collection of samples, 1 for each channel,
234 so for 5.1 a frame would be 6 samples)
236 int const frames = samples / _fs->audio_channels;
238 /* Maybe apply gain */
239 if (_fs->audio_gain != 0) {
240 float const linear_gain = pow (10, _fs->audio_gain / 20);
242 switch (_fs->audio_sample_format) {
243 case AV_SAMPLE_FMT_S16:
244 for (int i = 0; i < samples; ++i) {
245 /* XXX: assumes little-endian; also we should probably be dithering here */
247 /* unsigned sample */
248 int const ou = p[0] | (p[1] << 8);
251 int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
253 /* signed sample with altered gain */
254 int const gs = int (os * linear_gain);
256 /* unsigned sample with altered gain */
257 int const gu = gs > 0 ? gs : (0x10000 + gs);
261 p[1] = (gu & 0xff00) >> 8;
270 /* This is a buffer we might use if we are sample-rate converting;
271 it will need freeing if so.
273 uint8_t* out_buffer = 0;
275 /* Maybe sample-rate convert */
278 uint8_t const * in[2] = {
283 /* Compute the resampled frame count and add 32 for luck */
284 int const out_buffer_size_frames = ceil (frames * float (dcp_audio_sample_rate (_fs->audio_sample_rate)) / _fs->audio_sample_rate) + 32;
285 int const out_buffer_size_bytes = out_buffer_size_frames * _fs->audio_channels * _fs->bytes_per_sample();
286 out_buffer = new uint8_t[out_buffer_size_bytes];
294 int out_frames = swr_convert (_swr_context, out, out_buffer_size_frames, in, frames);
295 if (out_frames < 0) {
296 throw DecodeError ("could not run sample-rate converter");
299 /* And point our variables at the resampled audio */
301 size = out_frames * _fs->audio_channels * _fs->bytes_per_sample();
304 /* Update the number of audio frames we've pushed to the encoder */
305 _audio_frames_processed += size / (_fs->audio_channels * _fs->bytes_per_sample ());
307 /* Push into the delay line and then tell the world what we've got */
308 int available = _delay_line->feed (data, size);
309 Audio (data, available);
311 /* Delete the sample-rate conversion buffer, if it exists */
315 /** Called by subclasses to tell the world that some video data is ready.
316 * We do some post-processing / filtering then emit it for listeners.
317 * @param frame to decode; caller manages memory.
320 Decoder::process_video (AVFrame* frame)
327 /* Use FilmState::length here as our one may be wrong */
330 if (_opt->decode_video_frequency != 0) {
331 gap = _fs->length / _opt->decode_video_frequency;
334 if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) {
339 #ifdef DVDOMATIC_FFMPEG_0_8_3
342 par.num = sample_aspect_ratio_numerator ();
343 par.den = sample_aspect_ratio_denominator ();
345 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) {
346 throw DecodeError ("could not push buffer into filter chain.");
351 if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) {
352 throw DecodeError ("could not push buffer into filter chain.");
357 #ifdef DVDOMATIC_FFMPEG_0_8_3
358 while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
360 while (av_buffersink_read (_buffer_sink_context, 0)) {
363 #ifdef DVDOMATIC_FFMPEG_0_8_3
365 int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
367 throw DecodeError ("could not request filtered frame");
370 AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf;
374 AVFilterBufferRef* filter_buffer;
375 if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) {
382 /* This takes ownership of filter_buffer */
383 shared_ptr<Image> image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer));
385 if (_opt->black_after > 0 && _video_frame > _opt->black_after) {
386 image->make_black ();
389 Video (image, _video_frame);
396 /** Set up a video filtering chain to include cropping and any filters that are specified
400 Decoder::setup_video_filters ()
403 Size size_after_crop;
405 if (_opt->apply_crop) {
406 size_after_crop = _fs->cropped_size (native_size ());
407 fs << crop_string (Position (_fs->left_crop, _fs->top_crop), size_after_crop);
409 size_after_crop = native_size ();
410 fs << crop_string (Position (0, 0), size_after_crop);
413 string filters = Filter::ffmpeg_strings (_fs->filters).first;
414 if (!filters.empty ()) {
418 filters += fs.str ();
420 avfilter_register_all ();
422 AVFilterGraph* graph = avfilter_graph_alloc();
424 throw DecodeError ("Could not create filter graph.");
427 AVFilter* buffer_src = avfilter_get_by_name("buffer");
428 if (buffer_src == 0) {
429 throw DecodeError ("Could not find buffer src filter");
432 AVFilter* buffer_sink = get_sink ();
435 a << native_size().width << ":"
436 << native_size().height << ":"
437 << pixel_format() << ":"
438 << time_base_numerator() << ":"
439 << time_base_denominator() << ":"
440 << sample_aspect_ratio_numerator() << ":"
441 << sample_aspect_ratio_denominator();
444 if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
445 throw DecodeError ("could not create buffer source");
448 enum PixelFormat pixel_formats[] = { pixel_format(), PIX_FMT_NONE };
449 if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, pixel_formats, graph) < 0) {
450 throw DecodeError ("could not create buffer sink.");
453 AVFilterInOut* outputs = avfilter_inout_alloc ();
454 outputs->name = av_strdup("in");
455 outputs->filter_ctx = _buffer_src_context;
456 outputs->pad_idx = 0;
459 AVFilterInOut* inputs = avfilter_inout_alloc ();
460 inputs->name = av_strdup("out");
461 inputs->filter_ctx = _buffer_sink_context;
465 _log->log ("Using filter chain `" + filters + "'");
466 #ifdef DVDOMATIC_FFMPEG_0_8_3
467 if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) {
469 if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
472 throw DecodeError ("could not set up filter graph.");
475 if (avfilter_graph_config (graph, 0) < 0) {
476 throw DecodeError ("could not configure filter graph.");
479 /* XXX: leaking `inputs' / `outputs' ? */