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"
53 using namespace boost;
55 /** @param s FilmState of the Film.
57 * @param j Job that we are running within, or 0
58 * @param l Log to use.
59 * @param minimal true to do the bare minimum of work; just run through the content. Useful for acquiring
60 * accurate frame counts as quickly as possible. This generates no video or audio output.
61 * @param ignore_length Ignore the content's claimed length when computing progress.
63 Decoder::Decoder (boost::shared_ptr<const FilmState> s, boost::shared_ptr<const Options> o, Job* j, Log* l, bool minimal, bool ignore_length)
69 , _ignore_length (ignore_length)
71 , _buffer_src_context (0)
72 , _buffer_sink_context (0)
76 , _have_setup_video_filters (false)
79 , _audio_frames_processed (0)
81 if (_opt->decode_video_frequency != 0 && _fs->length == 0) {
82 throw DecodeError ("cannot do a partial decode if length == 0");
91 /** Start off a decode processing run */
93 Decoder::process_begin ()
95 if (_fs->audio_sample_rate != dcp_audio_sample_rate (_fs->audio_sample_rate)) {
97 _swr_context = swr_alloc_set_opts (
99 audio_channel_layout(),
100 audio_sample_format(),
101 dcp_audio_sample_rate (_fs->audio_sample_rate),
102 audio_channel_layout(),
103 audio_sample_format(),
104 _fs->audio_sample_rate,
108 swr_init (_swr_context);
110 throw DecodeError ("Cannot resample audio as libswresample is not present");
118 _delay_in_bytes = _fs->audio_delay * _fs->audio_sample_rate * _fs->audio_channels * _fs->bytes_per_sample() / 1000;
120 _delay_line = new DelayLine (_delay_in_bytes);
122 _audio_frames_processed = 0;
125 /** Finish off a decode processing run */
127 Decoder::process_end ()
134 uint8_t buffer[256 * _fs->bytes_per_sample() * _fs->audio_channels];
139 int const frames = swr_convert (_swr_context, out, 256, 0, 0);
142 throw DecodeError ("could not run sample-rate converter");
150 int available = _delay_line->feed (buffer, frames * _fs->audio_channels * _fs->bytes_per_sample());
151 Audio (buffer, available);
154 swr_free (&_swr_context);
158 if (_delay_in_bytes < 0) {
159 uint8_t remainder[-_delay_in_bytes];
160 _delay_line->get_remaining (remainder);
161 _audio_frames_processed += _delay_in_bytes / (_fs->audio_channels * _fs->bytes_per_sample());
162 Audio (remainder, _delay_in_bytes);
165 /* If we cut the decode off, the audio may be short; push some silence
166 in to get it to the right length.
169 int64_t const audio_short_by_frames =
170 ((int64_t) decoding_frames() * dcp_audio_sample_rate (_fs->audio_sample_rate) / _fs->frames_per_second)
171 - _audio_frames_processed;
173 if (audio_short_by_frames >= 0) {
174 int bytes = audio_short_by_frames * _fs->audio_channels * _fs->bytes_per_sample();
176 int const silence_size = 64 * 1024;
177 uint8_t silence[silence_size];
178 memset (silence, 0, silence_size);
181 int const t = min (bytes, silence_size);
188 /** Start decoding */
194 if (_job && _ignore_length) {
195 _job->set_progress_unknown ();
198 while (pass () == false) {
199 if (_job && !_ignore_length) {
200 _job->set_progress (float (_video_frame) / decoding_frames ());
207 /** @return Number of frames that we will be decoding */
209 Decoder::decoding_frames () const
211 if (_opt->num_frames > 0) {
212 return _opt->num_frames;
218 /** Run one pass. This may or may not generate any actual video / audio data;
219 * some decoders may require several passes to generate a single frame.
220 * @return true if we have finished processing all data; otherwise false.
225 if (!_have_setup_video_filters) {
226 setup_video_filters ();
227 _have_setup_video_filters = true;
230 if (_opt->num_frames != 0 && _video_frame >= _opt->num_frames) {
237 /** Called by subclasses to tell the world that some audio data is ready
238 * @param data Interleaved audio data, in FilmState::audio_sample_format.
239 * @param size Number of bytes of data.
242 Decoder::process_audio (uint8_t* data, int size)
244 /* Here's samples per channel */
245 int const samples = size / _fs->bytes_per_sample();
248 /* And here's frames (where 1 frame is a collection of samples, 1 for each channel,
249 so for 5.1 a frame would be 6 samples)
251 int const frames = samples / _fs->audio_channels;
254 /* Maybe apply gain */
255 if (_fs->audio_gain != 0) {
256 float const linear_gain = pow (10, _fs->audio_gain / 20);
258 switch (_fs->audio_sample_format) {
259 case AV_SAMPLE_FMT_S16:
260 for (int i = 0; i < samples; ++i) {
261 /* XXX: assumes little-endian; also we should probably be dithering here */
263 /* unsigned sample */
264 int const ou = p[0] | (p[1] << 8);
267 int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
269 /* signed sample with altered gain */
270 int const gs = int (os * linear_gain);
272 /* unsigned sample with altered gain */
273 int const gu = gs > 0 ? gs : (0x10000 + gs);
277 p[1] = (gu & 0xff00) >> 8;
286 /* This is a buffer we might use if we are sample-rate converting;
287 it will need freeing if so.
289 uint8_t* out_buffer = 0;
291 /* Maybe sample-rate convert */
295 uint8_t const * in[2] = {
300 /* Compute the resampled frame count and add 32 for luck */
301 int const out_buffer_size_frames = ceil (frames * float (dcp_audio_sample_rate (_fs->audio_sample_rate)) / _fs->audio_sample_rate) + 32;
302 int const out_buffer_size_bytes = out_buffer_size_frames * _fs->audio_channels * _fs->bytes_per_sample();
303 out_buffer = new uint8_t[out_buffer_size_bytes];
311 int out_frames = swr_convert (_swr_context, out, out_buffer_size_frames, in, frames);
312 if (out_frames < 0) {
313 throw DecodeError ("could not run sample-rate converter");
316 /* And point our variables at the resampled audio */
318 size = out_frames * _fs->audio_channels * _fs->bytes_per_sample();
322 /* Update the number of audio frames we've pushed to the encoder */
323 _audio_frames_processed += size / (_fs->audio_channels * _fs->bytes_per_sample ());
325 /* Push into the delay line and then tell the world what we've got */
326 int available = _delay_line->feed (data, size);
327 Audio (data, available);
329 /* Delete the sample-rate conversion buffer, if it exists */
333 /** Called by subclasses to tell the world that some video data is ready.
334 * We do some post-processing / filtering then emit it for listeners.
335 * @param frame to decode; caller manages memory.
338 Decoder::process_video (AVFrame* frame)
345 /* Use FilmState::length here as our one may be wrong */
348 if (_opt->decode_video_frequency != 0) {
349 gap = _fs->length / _opt->decode_video_frequency;
352 if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) {
357 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 61
359 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0) < 0) {
360 throw DecodeError ("could not push buffer into filter chain.");
363 #elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
366 par.num = sample_aspect_ratio_numerator ();
367 par.den = sample_aspect_ratio_denominator ();
369 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) {
370 throw DecodeError ("could not push buffer into filter chain.");
375 if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) {
376 throw DecodeError ("could not push buffer into filter chain.");
381 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15 && LIBAVFILTER_VERSION_MINOR <= 61
382 while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
384 while (av_buffersink_read (_buffer_sink_context, 0)) {
387 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15
389 int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
391 throw DecodeError ("could not request filtered frame");
394 AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf;
398 AVFilterBufferRef* filter_buffer;
399 if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) {
406 /* This takes ownership of filter_buffer */
407 shared_ptr<Image> image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer));
409 if (_opt->black_after > 0 && _video_frame > _opt->black_after) {
410 image->make_black ();
413 TIMING ("Decoder emits %1", _video_frame);
414 Video (image, _video_frame);
421 /** Set up a video filtering chain to include cropping and any filters that are specified
425 Decoder::setup_video_filters ()
428 Size size_after_crop;
430 if (_opt->apply_crop) {
431 size_after_crop = _fs->cropped_size (native_size ());
432 fs << crop_string (Position (_fs->crop.left, _fs->crop.top), size_after_crop);
434 size_after_crop = native_size ();
435 fs << crop_string (Position (0, 0), size_after_crop);
438 string filters = Filter::ffmpeg_strings (_fs->filters).first;
439 if (!filters.empty ()) {
443 filters += fs.str ();
445 avfilter_register_all ();
447 AVFilterGraph* graph = avfilter_graph_alloc();
449 throw DecodeError ("Could not create filter graph.");
452 AVFilter* buffer_src = avfilter_get_by_name("buffer");
453 if (buffer_src == 0) {
454 throw DecodeError ("Could not find buffer src filter");
457 AVFilter* buffer_sink = get_sink ();
460 a << native_size().width << ":"
461 << native_size().height << ":"
462 << pixel_format() << ":"
463 << time_base_numerator() << ":"
464 << time_base_denominator() << ":"
465 << sample_aspect_ratio_numerator() << ":"
466 << sample_aspect_ratio_denominator();
470 if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
471 throw DecodeError ("could not create buffer source");
474 AVBufferSinkParams* sink_params = av_buffersink_params_alloc ();
475 PixelFormat* pixel_fmts = new PixelFormat[2];
476 pixel_fmts[0] = pixel_format ();
477 pixel_fmts[1] = PIX_FMT_NONE;
478 sink_params->pixel_fmts = pixel_fmts;
480 if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, sink_params, graph) < 0) {
481 throw DecodeError ("could not create buffer sink.");
484 AVFilterInOut* outputs = avfilter_inout_alloc ();
485 outputs->name = av_strdup("in");
486 outputs->filter_ctx = _buffer_src_context;
487 outputs->pad_idx = 0;
490 AVFilterInOut* inputs = avfilter_inout_alloc ();
491 inputs->name = av_strdup("out");
492 inputs->filter_ctx = _buffer_sink_context;
496 _log->log ("Using filter chain `" + filters + "'");
498 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
499 if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) {
500 throw DecodeError ("could not set up filter graph.");
503 if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
504 throw DecodeError ("could not set up filter graph.");
508 if (avfilter_graph_config (graph, 0) < 0) {
509 throw DecodeError ("could not configure filter graph.");
512 /* XXX: leaking `inputs' / `outputs' ? */