source: src/io/source_avcodec.c @ 549928e

feature/autosinkfeature/constantqfeature/pitchshiftfeature/pydocstringsfeature/timestretchpitchshiftsamplertimestretchyinfft+
Last change on this file since 549928e was 549928e, checked in by Paul Brossier <piem@piem.org>, 7 years ago

wscript: make sure all libav libraries are installed to build source_avcodec

Signed-off-by: Paul Brossier <piem@piem.org>

  • Property mode set to 100644
File size: 11.5 KB
Line 
1/*
2  Copyright (C) 2013 Paul Brossier <piem@aubio.org>
3
4  This file is part of aubio.
5
6  aubio is free software: you can redistribute it and/or modify
7  it under the terms of the GNU General Public License as published by
8  the Free Software Foundation, either version 3 of the License, or
9  (at your option) any later version.
10
11  aubio is distributed in the hope that it will be useful,
12  but WITHOUT ANY WARRANTY; without even the implied warranty of
13  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  GNU General Public License for more details.
15
16  You should have received a copy of the GNU General Public License
17  along with aubio.  If not, see <http://www.gnu.org/licenses/>.
18
19*/
20
21
22#include "config.h"
23
24#ifdef HAVE_LIBAV
25
26#include <sndfile.h>
27#include <libavcodec/avcodec.h>
28#include <libavformat/avformat.h>
29#include <libavresample/avresample.h>
30#include <libavutil/opt.h>
31#include <stdlib.h>
32
33#include "aubio_priv.h"
34#include "fvec.h"
35#include "fmat.h"
36#include "source_avcodec.h"
37
38#define AUBIO_AVCODEC_MAX_BUFFER_SIZE FF_MIN_BUFFER_SIZE
39
40struct _aubio_source_avcodec_t {
41  uint_t hop_size;
42  uint_t samplerate;
43  uint_t channels;
44
45  // some data about the file
46  char_t *path;
47  uint_t input_samplerate;
48  uint_t input_channels;
49
50  // avcodec stuff
51  AVFormatContext *avFormatCtx;
52  AVCodecContext *avCodecCtx;
53  AVFrame *avFrame;
54  AVAudioResampleContext *avr;
55  float *output;
56  uint_t read_samples;
57  uint_t read_index;
58  sint_t selected_stream;
59  uint_t eof;
60  uint_t multi;
61};
62
63// hack to create or re-create the context the first time _do or _do_multi is called
64void aubio_source_avcodec_reset_resampler(aubio_source_avcodec_t * s, uint_t multi);
65
66aubio_source_avcodec_t * new_aubio_source_avcodec(char_t * path, uint_t samplerate, uint_t hop_size) {
67  aubio_source_avcodec_t * s = AUBIO_NEW(aubio_source_avcodec_t);
68  int err;
69  if (path == NULL) {
70    AUBIO_ERR("Aborted opening null path\n");
71    goto beach;
72  }
73  if ((sint_t)samplerate < 0) {
74    AUBIO_ERR("Can not open %s with samplerate %d\n", path, samplerate);
75    goto beach;
76  }
77  if ((sint_t)hop_size <= 0) {
78    AUBIO_ERR("Can not open %s with hop_size %d\n", path, hop_size);
79    goto beach;
80  }
81
82  s->hop_size = hop_size;
83  s->channels = 1;
84  s->path = path;
85
86  // register all formats and codecs
87  av_register_all();
88
89  // if path[0] != '/'
90  //avformat_network_init();
91
92  // try opening the file and get some info about it
93  AVFormatContext *avFormatCtx = s->avFormatCtx;
94  avFormatCtx = NULL;
95  if ( (err = avformat_open_input(&avFormatCtx, s->path, NULL, NULL) ) < 0 ) {
96    char errorstr[256];
97    av_strerror (err, errorstr, sizeof(errorstr));
98    AUBIO_ERR("Failed opening %s (%s)\n", s->path, errorstr);
99    goto beach;
100  }
101
102  // try to make sure max_analyze_duration is big enough for most songs
103  avFormatCtx->max_analyze_duration *= 100;
104
105  // retrieve stream information
106  if ( (err = avformat_find_stream_info(avFormatCtx, NULL)) < 0 ) {
107    char errorstr[256];
108    av_strerror (err, errorstr, sizeof(errorstr));
109    AUBIO_ERR("Could not find stream information " "for %s (%s)\n", s->path,
110        errorstr);
111    goto beach;
112  }
113
114  // dump information about file onto standard error
115  //av_dump_format(avFormatCtx, 0, s->path, 0);
116
117  // look for the first audio stream
118  uint_t i;
119  sint_t selected_stream = -1;
120  for (i = 0; i < avFormatCtx->nb_streams; i++) {
121    if (avFormatCtx->streams[i]->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
122      if (selected_stream == -1) {
123        selected_stream = i;
124      } else {
125        AUBIO_WRN("More than one audio stream in %s, "
126            "taking the first one\n", s->path);
127      }
128    }
129  }
130  if (selected_stream == -1) {
131    AUBIO_ERR("No audio stream in %s\n", s->path);
132    goto beach;
133  }
134  //AUBIO_DBG("Taking stream %d in file %s\n", selected_stream, s->path);
135  s->selected_stream = selected_stream;
136
137  AVCodecContext *avCodecCtx = s->avCodecCtx;
138  avCodecCtx = avFormatCtx->streams[selected_stream]->codec;
139  AVCodec *codec = avcodec_find_decoder(avCodecCtx->codec_id);
140  if (codec == NULL) {
141    AUBIO_ERR("Could not find decoder for %s", s->path);
142    goto beach;
143  }
144
145  if ( ( err = avcodec_open2(avCodecCtx, codec, NULL) ) < 0) {
146    char errorstr[256];
147    av_strerror (err, errorstr, sizeof(errorstr));
148    AUBIO_ERR("Could not load codec for %s (%s)\n", s->path, errorstr);
149    goto beach;
150  }
151
152  /* get input specs */
153  s->input_samplerate = avCodecCtx->sample_rate;
154  s->input_channels   = avCodecCtx->channels;
155  //AUBIO_DBG("input_samplerate: %d\n", s->input_samplerate);
156  //AUBIO_DBG("input_channels: %d\n", s->input_channels);
157
158  if (samplerate == 0) {
159    samplerate = s->input_samplerate;
160    //AUBIO_DBG("sampling rate set to 0, automagically adjusting to %d\n", samplerate);
161  }
162  s->samplerate = samplerate;
163
164  if (s->samplerate >  s->input_samplerate) {
165    AUBIO_WRN("upsampling %s from %d to %d\n", s->path,
166        s->input_samplerate, s->samplerate);
167  }
168
169  AVFrame *avFrame = s->avFrame;
170  avFrame = avcodec_alloc_frame();
171  if (!avFrame) {
172    AUBIO_ERR("Could not allocate frame for (%s)\n", s->path);
173  }
174
175  /* allocate output for avr */
176  s->output = (float *)av_malloc(AUBIO_AVCODEC_MAX_BUFFER_SIZE * sizeof(float));
177
178  s->read_samples = 0;
179  s->read_index = 0;
180
181  s->avFormatCtx = avFormatCtx;
182  s->avCodecCtx = avCodecCtx;
183  s->avFrame = avFrame;
184
185  // default to mono output
186  aubio_source_avcodec_reset_resampler(s, 0);
187
188  s->eof = 0;
189  s->multi = 0;
190
191  //av_log_set_level(AV_LOG_QUIET);
192
193  return s;
194
195beach:
196  //AUBIO_ERR("can not read %s at samplerate %dHz with a hop_size of %d\n",
197  //    s->path, s->samplerate, s->hop_size);
198  del_aubio_source_avcodec(s);
199  return NULL;
200}
201
202void aubio_source_avcodec_reset_resampler(aubio_source_avcodec_t * s, uint_t multi) {
203  if ( (multi != s->multi) || (s->avr == NULL) ) {
204    int64_t input_layout = av_get_default_channel_layout(s->input_channels);
205    uint_t output_channels = multi ? s->input_channels : 1;
206    int64_t output_layout = av_get_default_channel_layout(output_channels);
207    if (s->avr != NULL) {
208      avresample_close( s->avr );
209      av_free ( s->avr );
210      s->avr = NULL;
211    }
212    AVAudioResampleContext *avr = s->avr;
213    avr = avresample_alloc_context();
214
215    av_opt_set_int(avr, "in_channel_layout",  input_layout,           0);
216    av_opt_set_int(avr, "out_channel_layout", output_layout,          0);
217    av_opt_set_int(avr, "in_sample_rate",     s->input_samplerate,    0);
218    av_opt_set_int(avr, "out_sample_rate",    s->samplerate,          0);
219    av_opt_set_int(avr, "in_sample_fmt",      s->avCodecCtx->sample_fmt, 0);
220    av_opt_set_int(avr, "out_sample_fmt",     AV_SAMPLE_FMT_FLT,      0);
221    int err;
222    if ( ( err = avresample_open(avr) ) < 0) {
223      char errorstr[256];
224      av_strerror (err, errorstr, sizeof(errorstr));
225      AUBIO_ERR("Could not open AVAudioResampleContext for %s (%s)\n",
226          s->path, errorstr);
227      //goto beach;
228      return;
229    }
230    s->avr = avr;
231    s->multi = multi;
232  }
233}
234
235void aubio_source_avcodec_readframe(aubio_source_avcodec_t *s, uint_t * read_samples) {
236  AVFormatContext *avFormatCtx = s->avFormatCtx;
237  AVCodecContext *avCodecCtx = s->avCodecCtx;
238  AVFrame *avFrame = s->avFrame;
239  AVPacket avPacket;
240  av_init_packet (&avPacket);
241  AVAudioResampleContext *avr = s->avr;
242  float *output = s->output;
243  *read_samples = 0;
244
245  do
246  {
247    int err = av_read_frame (avFormatCtx, &avPacket);
248    if (err == AVERROR_EOF) {
249      s->eof = 1;
250      goto beach;
251    }
252    if (err != 0) {
253      char errorstr[256];
254      av_strerror (err, errorstr, sizeof(errorstr));
255      AUBIO_ERR("Could not read frame in %s (%s)\n", s->path, errorstr);
256      goto beach;
257    }
258  } while (avPacket.stream_index != s->selected_stream);
259
260  int got_frame = 0;
261  int len = avcodec_decode_audio4(avCodecCtx, avFrame, &got_frame, &avPacket);
262
263  if (len < 0) {
264    AUBIO_ERR("Error while decoding %s\n", s->path);
265    goto beach;
266  }
267  if (got_frame == 0) {
268    //AUBIO_ERR("Could not get frame for (%s)\n", s->path);
269    goto beach;
270  }
271
272  int in_linesize = 0;
273  av_samples_get_buffer_size(&in_linesize, avCodecCtx->channels,
274      avFrame->nb_samples, avCodecCtx->sample_fmt, 1);
275  int in_samples = avFrame->nb_samples;
276  int out_linesize = 0;
277  int max_out_samples = AUBIO_AVCODEC_MAX_BUFFER_SIZE;
278  int out_samples = avresample_convert ( avr,
279        (uint8_t **)&output, out_linesize, max_out_samples,
280        (uint8_t **)avFrame->data, in_linesize, in_samples);
281  if (out_samples <= 0) {
282    AUBIO_ERR("No sample found while converting frame (%s)\n", s->path);
283    goto beach;
284  }
285
286  *read_samples = out_samples;
287
288beach:
289  s->avFormatCtx = avFormatCtx;
290  s->avCodecCtx = avCodecCtx;
291  s->avFrame = avFrame;
292  s->avr = avr;
293  s->output = output;
294
295  av_free_packet(&avPacket);
296}
297
298void aubio_source_avcodec_do(aubio_source_avcodec_t * s, fvec_t * read_data, uint_t * read){
299  if (s->multi == 1) aubio_source_avcodec_reset_resampler(s, 0);
300  uint_t i;
301  uint_t end = 0;
302  uint_t total_wrote = 0;
303  while (total_wrote < s->hop_size) {
304    end = MIN(s->read_samples - s->read_index, s->hop_size - total_wrote);
305    for (i = 0; i < end; i++) {
306      read_data->data[i + total_wrote] = s->output[i + s->read_index];
307    }
308    total_wrote += end;
309    if (total_wrote < s->hop_size) {
310      uint_t avcodec_read = 0;
311      aubio_source_avcodec_readframe(s, &avcodec_read);
312      s->read_samples = avcodec_read;
313      s->read_index = 0;
314      if (s->eof) {
315        break;
316      }
317    } else {
318      s->read_index += end;
319    }
320  }
321  if (total_wrote < s->hop_size) {
322    for (i = end; i < s->hop_size; i++) {
323      read_data->data[i] = 0.;
324    }
325  }
326  *read = total_wrote;
327}
328
329void aubio_source_avcodec_do_multi(aubio_source_avcodec_t * s, fmat_t * read_data, uint_t * read){
330  if (s->multi == 0) aubio_source_avcodec_reset_resampler(s, 1);
331  uint_t i,j;
332  uint_t end = 0;
333  uint_t total_wrote = 0;
334  while (total_wrote < s->hop_size) {
335    end = MIN(s->read_samples - s->read_index, s->hop_size - total_wrote);
336    for (j = 0; j < read_data->height; j++) {
337      for (i = 0; i < end; i++) {
338        read_data->data[j][i + total_wrote] =
339          s->output[(i + s->read_index) * s->input_channels + j];
340      }
341    }
342    total_wrote += end;
343    if (total_wrote < s->hop_size) {
344      uint_t avcodec_read = 0;
345      aubio_source_avcodec_readframe(s, &avcodec_read);
346      s->read_samples = avcodec_read;
347      s->read_index = 0;
348      if (s->eof) {
349        break;
350      }
351    } else {
352      s->read_index += end;
353    }
354  }
355  if (total_wrote < s->hop_size) {
356    for (j = 0; j < read_data->height; j++) {
357      for (i = end; i < s->hop_size; i++) {
358        read_data->data[j][i] = 0.;
359      }
360    }
361  }
362  *read = total_wrote;
363}
364
365uint_t aubio_source_avcodec_get_samplerate(aubio_source_avcodec_t * s) {
366  return s->samplerate;
367}
368
369uint_t aubio_source_avcodec_get_channels(aubio_source_avcodec_t * s) {
370  return s->input_channels;
371}
372
373uint_t aubio_source_avcodec_seek (aubio_source_avcodec_t * s, uint_t pos) {
374  //uint_t resampled_pos = (uint_t)ROUND(pos * s->input_samplerate * 1. / s->samplerate);
375  AUBIO_ERR("Not implemented, _seek(%d) file %s", pos, s->path);
376  return -1; //sf_seek (s->handle, resampled_pos, SEEK_SET);
377}
378
379void del_aubio_source_avcodec(aubio_source_avcodec_t * s){
380  if (!s) return;
381  if (s->output != NULL) {
382    av_free(s->output);
383  }
384  if (s->avr != NULL) {
385    avresample_close( s->avr );
386    av_free ( s->avr );
387  }
388  s->avr = NULL;
389  if (s->avFrame != NULL) {
390    avcodec_free_frame( &(s->avFrame) );
391  }
392  s->avFrame = NULL;
393  if (s->avCodecCtx != NULL) {
394    avcodec_close ( s->avCodecCtx );
395  }
396  s->avCodecCtx = NULL;
397  if (s->avFormatCtx != NULL) {
398    avformat_close_input ( &(s->avFormatCtx) );
399  }
400  s->avFrame = NULL;
401  s->avFormatCtx = NULL;
402  AUBIO_FREE(s);
403}
404
405#endif /* HAVE_LIBAV */
Note: See TracBrowser for help on using the repository browser.