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)
73 , _have_setup_video_filters (false)
76 , _audio_frames_processed (0)
78 if (_opt->decode_video_frequency != 0 && _fs->length == 0) {
79 throw DecodeError ("cannot do a partial decode if length == 0");
88 /** Start off a decode processing run */
90 Decoder::process_begin ()
92 _delay_in_bytes = _fs->audio_delay * _fs->audio_sample_rate * _fs->audio_channels * _fs->bytes_per_sample() / 1000;
94 _delay_line = new DelayLine (_delay_in_bytes);
96 _audio_frames_processed = 0;
99 /** Finish off a decode processing run */
101 Decoder::process_end ()
103 if (_delay_in_bytes < 0) {
104 uint8_t remainder[-_delay_in_bytes];
105 _delay_line->get_remaining (remainder);
106 _audio_frames_processed += _delay_in_bytes / (_fs->audio_channels * _fs->bytes_per_sample());
107 Audio (remainder, _delay_in_bytes);
110 /* If we cut the decode off, the audio may be short; push some silence
111 in to get it to the right length.
114 int64_t const audio_short_by_frames =
115 ((int64_t) _fs->dcp_length() * _fs->target_sample_rate() / _fs->frames_per_second)
116 - _audio_frames_processed;
118 if (audio_short_by_frames >= 0) {
121 s << "Adding " << audio_short_by_frames << " frames of silence to the end.";
122 _log->log (s.str ());
124 int64_t bytes = audio_short_by_frames * _fs->audio_channels * _fs->bytes_per_sample();
126 int64_t const silence_size = 64 * 1024;
127 uint8_t silence[silence_size];
128 memset (silence, 0, silence_size);
131 int64_t const t = min (bytes, silence_size);
138 /** Start decoding */
144 if (_job && _ignore_length) {
145 _job->set_progress_unknown ();
148 while (pass () == false) {
149 if (_job && !_ignore_length) {
150 _job->set_progress (float (_video_frame) / _fs->dcp_length());
157 /** Run one pass. This may or may not generate any actual video / audio data;
158 * some decoders may require several passes to generate a single frame.
159 * @return true if we have finished processing all data; otherwise false.
164 if (!_have_setup_video_filters) {
165 setup_video_filters ();
166 _have_setup_video_filters = true;
169 if (_video_frame >= _fs->dcp_length()) {
176 /** Called by subclasses to tell the world that some audio data is ready
177 * @param data Interleaved audio data, in FilmState::audio_sample_format.
178 * @param size Number of bytes of data.
181 Decoder::process_audio (uint8_t* data, int size)
183 /* Samples per channel */
184 int const samples = size / _fs->bytes_per_sample();
186 /* Maybe apply gain */
187 if (_fs->audio_gain != 0) {
188 float const linear_gain = pow (10, _fs->audio_gain / 20);
190 switch (_fs->audio_sample_format) {
191 case AV_SAMPLE_FMT_S16:
192 for (int i = 0; i < samples; ++i) {
193 /* XXX: assumes little-endian; also we should probably be dithering here */
195 /* unsigned sample */
196 int const ou = p[0] | (p[1] << 8);
199 int const os = ou >= 0x8000 ? (- 0x10000 + ou) : ou;
201 /* signed sample with altered gain */
202 int const gs = int (os * linear_gain);
204 /* unsigned sample with altered gain */
205 int const gu = gs > 0 ? gs : (0x10000 + gs);
209 p[1] = (gu & 0xff00) >> 8;
218 /* Update the number of audio frames we've pushed to the encoder */
219 _audio_frames_processed += size / (_fs->audio_channels * _fs->bytes_per_sample ());
221 /* Push into the delay line and then tell the world what we've got */
222 int available = _delay_line->feed (data, size);
223 Audio (data, available);
226 /** Called by subclasses to tell the world that some video data is ready.
227 * We do some post-processing / filtering then emit it for listeners.
228 * @param frame to decode; caller manages memory.
231 Decoder::process_video (AVFrame* frame)
238 /* Use FilmState::length here as our one may be wrong */
241 if (_opt->decode_video_frequency != 0) {
242 gap = _fs->length / _opt->decode_video_frequency;
245 if (_opt->decode_video_frequency != 0 && gap != 0 && (_video_frame % gap) != 0) {
250 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 53 && LIBAVFILTER_VERSION_MINOR <= 61
252 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0) < 0) {
253 throw DecodeError ("could not push buffer into filter chain.");
256 #elif LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
259 par.num = sample_aspect_ratio_numerator ();
260 par.den = sample_aspect_ratio_denominator ();
262 if (av_vsrc_buffer_add_frame (_buffer_src_context, frame, 0, par) < 0) {
263 throw DecodeError ("could not push buffer into filter chain.");
268 if (av_buffersrc_write_frame (_buffer_src_context, frame) < 0) {
269 throw DecodeError ("could not push buffer into filter chain.");
274 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15 && LIBAVFILTER_VERSION_MINOR <= 61
275 while (avfilter_poll_frame (_buffer_sink_context->inputs[0])) {
277 while (av_buffersink_read (_buffer_sink_context, 0)) {
280 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR >= 15
282 int r = avfilter_request_frame (_buffer_sink_context->inputs[0]);
284 throw DecodeError ("could not request filtered frame");
287 AVFilterBufferRef* filter_buffer = _buffer_sink_context->inputs[0]->cur_buf;
291 AVFilterBufferRef* filter_buffer;
292 if (av_buffersink_get_buffer_ref (_buffer_sink_context, &filter_buffer, 0) < 0) {
299 /* This takes ownership of filter_buffer */
300 shared_ptr<Image> image (new FilterBufferImage ((PixelFormat) frame->format, filter_buffer));
302 if (_opt->black_after > 0 && _video_frame > _opt->black_after) {
303 image->make_black ();
308 TIMING ("Decoder emits %1", _video_frame);
309 Video (image, _video_frame);
316 /** Set up a video filtering chain to include cropping and any filters that are specified
320 Decoder::setup_video_filters ()
323 Size size_after_crop;
325 if (_opt->apply_crop) {
326 size_after_crop = _fs->cropped_size (native_size ());
327 fs << crop_string (Position (_fs->crop.left, _fs->crop.top), size_after_crop);
329 size_after_crop = native_size ();
330 fs << crop_string (Position (0, 0), size_after_crop);
333 string filters = Filter::ffmpeg_strings (_fs->filters).first;
334 if (!filters.empty ()) {
338 filters += fs.str ();
340 avfilter_register_all ();
342 AVFilterGraph* graph = avfilter_graph_alloc();
344 throw DecodeError ("Could not create filter graph.");
347 AVFilter* buffer_src = avfilter_get_by_name("buffer");
348 if (buffer_src == 0) {
349 throw DecodeError ("Could not find buffer src filter");
352 AVFilter* buffer_sink = get_sink ();
355 a << native_size().width << ":"
356 << native_size().height << ":"
357 << pixel_format() << ":"
358 << time_base_numerator() << ":"
359 << time_base_denominator() << ":"
360 << sample_aspect_ratio_numerator() << ":"
361 << sample_aspect_ratio_denominator();
365 if ((r = avfilter_graph_create_filter (&_buffer_src_context, buffer_src, "in", a.str().c_str(), 0, graph)) < 0) {
366 throw DecodeError ("could not create buffer source");
369 AVBufferSinkParams* sink_params = av_buffersink_params_alloc ();
370 PixelFormat* pixel_fmts = new PixelFormat[2];
371 pixel_fmts[0] = pixel_format ();
372 pixel_fmts[1] = PIX_FMT_NONE;
373 sink_params->pixel_fmts = pixel_fmts;
375 if (avfilter_graph_create_filter (&_buffer_sink_context, buffer_sink, "out", 0, sink_params, graph) < 0) {
376 throw DecodeError ("could not create buffer sink.");
379 AVFilterInOut* outputs = avfilter_inout_alloc ();
380 outputs->name = av_strdup("in");
381 outputs->filter_ctx = _buffer_src_context;
382 outputs->pad_idx = 0;
385 AVFilterInOut* inputs = avfilter_inout_alloc ();
386 inputs->name = av_strdup("out");
387 inputs->filter_ctx = _buffer_sink_context;
391 _log->log ("Using filter chain `" + filters + "'");
393 #if LIBAVFILTER_VERSION_MAJOR == 2 && LIBAVFILTER_VERSION_MINOR == 15
394 if (avfilter_graph_parse (graph, filters.c_str(), inputs, outputs, 0) < 0) {
395 throw DecodeError ("could not set up filter graph.");
398 if (avfilter_graph_parse (graph, filters.c_str(), &inputs, &outputs, 0) < 0) {
399 throw DecodeError ("could not set up filter graph.");
403 if (avfilter_graph_config (graph, 0) < 0) {
404 throw DecodeError ("could not configure filter graph.");
407 /* XXX: leaking `inputs' / `outputs' ? */