aacenc.c
Go to the documentation of this file.
1 /*
2  * AAC encoder
3  * Copyright (C) 2008 Konstantin Shishkov
4  *
5  * This file is part of Libav.
6  *
7  * Libav is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * Libav is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with Libav; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
27 /***********************************
28  * TODOs:
29  * add sane pulse detection
30  * add temporal noise shaping
31  ***********************************/
32 
33 #include "libavutil/float_dsp.h"
34 #include "libavutil/opt.h"
35 #include "avcodec.h"
36 #include "put_bits.h"
37 #include "dsputil.h"
38 #include "internal.h"
39 #include "mpeg4audio.h"
40 #include "kbdwin.h"
41 #include "sinewin.h"
42 
43 #include "aac.h"
44 #include "aactab.h"
45 #include "aacenc.h"
46 
47 #include "psymodel.h"
48 
49 #define AAC_MAX_CHANNELS 6
50 
51 #define ERROR_IF(cond, ...) \
52  if (cond) { \
53  av_log(avctx, AV_LOG_ERROR, __VA_ARGS__); \
54  return AVERROR(EINVAL); \
55  }
56 
57 float ff_aac_pow34sf_tab[428];
58 
59 static const uint8_t swb_size_1024_96[] = {
60  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8,
61  12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44,
62  64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
63 };
64 
65 static const uint8_t swb_size_1024_64[] = {
66  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8,
67  12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36,
68  40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40
69 };
70 
71 static const uint8_t swb_size_1024_48[] = {
72  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
73  12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
74  32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32,
75  96
76 };
77 
78 static const uint8_t swb_size_1024_32[] = {
79  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
80  12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
81  32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
82 };
83 
84 static const uint8_t swb_size_1024_24[] = {
85  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
86  12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28,
87  32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64
88 };
89 
90 static const uint8_t swb_size_1024_16[] = {
91  8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
92  12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28,
93  32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64
94 };
95 
96 static const uint8_t swb_size_1024_8[] = {
97  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
98  16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28,
99  32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80
100 };
101 
102 static const uint8_t *swb_size_1024[] = {
107 };
108 
109 static const uint8_t swb_size_128_96[] = {
110  4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36
111 };
112 
113 static const uint8_t swb_size_128_48[] = {
114  4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16
115 };
116 
117 static const uint8_t swb_size_128_24[] = {
118  4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20
119 };
120 
121 static const uint8_t swb_size_128_16[] = {
122  4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20
123 };
124 
125 static const uint8_t swb_size_128_8[] = {
126  4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20
127 };
128 
129 static const uint8_t *swb_size_128[] = {
130  /* the last entry on the following row is swb_size_128_64 but is a
131  duplicate of swb_size_128_96 */
136 };
137 
139 static const uint8_t aac_chan_configs[6][5] = {
140  {1, TYPE_SCE}, // 1 channel - single channel element
141  {1, TYPE_CPE}, // 2 channels - channel pair
142  {2, TYPE_SCE, TYPE_CPE}, // 3 channels - center + stereo
143  {3, TYPE_SCE, TYPE_CPE, TYPE_SCE}, // 4 channels - front center + stereo + back center
144  {3, TYPE_SCE, TYPE_CPE, TYPE_CPE}, // 5 channels - front center + stereo + back stereo
145  {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE
146 };
147 
152  { 0 },
153  { 0, 1 },
154  { 2, 0, 1 },
155  { 2, 0, 1, 3 },
156  { 2, 0, 1, 3, 4 },
157  { 2, 0, 1, 4, 5, 3 },
158 };
159 
165 {
166  PutBitContext pb;
167  AACEncContext *s = avctx->priv_data;
168 
169  init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8);
170  put_bits(&pb, 5, 2); //object type - AAC-LC
171  put_bits(&pb, 4, s->samplerate_index); //sample rate index
172  put_bits(&pb, 4, s->channels);
173  //GASpecificConfig
174  put_bits(&pb, 1, 0); //frame length - 1024 samples
175  put_bits(&pb, 1, 0); //does not depend on core coder
176  put_bits(&pb, 1, 0); //is not extension
177 
178  //Explicitly Mark SBR absent
179  put_bits(&pb, 11, 0x2b7); //sync extension
180  put_bits(&pb, 5, AOT_SBR);
181  put_bits(&pb, 1, 0);
182  flush_put_bits(&pb);
183 }
184 
185 #define WINDOW_FUNC(type) \
186 static void apply_ ##type ##_window(DSPContext *dsp, AVFloatDSPContext *fdsp, \
187  SingleChannelElement *sce, \
188  const float *audio)
189 
190 WINDOW_FUNC(only_long)
191 {
192  const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
193  const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
194  float *out = sce->ret_buf;
195 
196  fdsp->vector_fmul (out, audio, lwindow, 1024);
197  dsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
198 }
199 
200 WINDOW_FUNC(long_start)
201 {
202  const float *lwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
203  const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
204  float *out = sce->ret_buf;
205 
206  fdsp->vector_fmul(out, audio, lwindow, 1024);
207  memcpy(out + 1024, audio + 1024, sizeof(out[0]) * 448);
208  dsp->vector_fmul_reverse(out + 1024 + 448, audio + 1024 + 448, swindow, 128);
209  memset(out + 1024 + 576, 0, sizeof(out[0]) * 448);
210 }
211 
212 WINDOW_FUNC(long_stop)
213 {
214  const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
215  const float *swindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
216  float *out = sce->ret_buf;
217 
218  memset(out, 0, sizeof(out[0]) * 448);
219  fdsp->vector_fmul(out + 448, audio + 448, swindow, 128);
220  memcpy(out + 576, audio + 576, sizeof(out[0]) * 448);
221  dsp->vector_fmul_reverse(out + 1024, audio + 1024, lwindow, 1024);
222 }
223 
224 WINDOW_FUNC(eight_short)
225 {
226  const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
227  const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
228  const float *in = audio + 448;
229  float *out = sce->ret_buf;
230  int w;
231 
232  for (w = 0; w < 8; w++) {
233  fdsp->vector_fmul (out, in, w ? pwindow : swindow, 128);
234  out += 128;
235  in += 128;
236  dsp->vector_fmul_reverse(out, in, swindow, 128);
237  out += 128;
238  }
239 }
240 
241 static void (*const apply_window[4])(DSPContext *dsp, AVFloatDSPContext *fdsp,
243  const float *audio) = {
244  [ONLY_LONG_SEQUENCE] = apply_only_long_window,
245  [LONG_START_SEQUENCE] = apply_long_start_window,
246  [EIGHT_SHORT_SEQUENCE] = apply_eight_short_window,
247  [LONG_STOP_SEQUENCE] = apply_long_stop_window
248 };
249 
251  float *audio)
252 {
253  int i;
254  float *output = sce->ret_buf;
255 
256  apply_window[sce->ics.window_sequence[0]](&s->dsp, &s->fdsp, sce, audio);
257 
259  s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
260  else
261  for (i = 0; i < 1024; i += 128)
262  s->mdct128.mdct_calc(&s->mdct128, sce->coeffs + i, output + i*2);
263  memcpy(audio, audio + 1024, sizeof(audio[0]) * 1024);
264 }
265 
271 {
272  int w;
273 
274  put_bits(&s->pb, 1, 0); // ics_reserved bit
275  put_bits(&s->pb, 2, info->window_sequence[0]);
276  put_bits(&s->pb, 1, info->use_kb_window[0]);
277  if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
278  put_bits(&s->pb, 6, info->max_sfb);
279  put_bits(&s->pb, 1, 0); // no prediction
280  } else {
281  put_bits(&s->pb, 4, info->max_sfb);
282  for (w = 1; w < 8; w++)
283  put_bits(&s->pb, 1, !info->group_len[w]);
284  }
285 }
286 
292 {
293  int i, w;
294 
295  put_bits(pb, 2, cpe->ms_mode);
296  if (cpe->ms_mode == 1)
297  for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w])
298  for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
299  put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
300 }
301 
305 static void adjust_frame_information(ChannelElement *cpe, int chans)
306 {
307  int i, w, w2, g, ch;
308  int start, maxsfb, cmaxsfb;
309 
310  for (ch = 0; ch < chans; ch++) {
311  IndividualChannelStream *ics = &cpe->ch[ch].ics;
312  start = 0;
313  maxsfb = 0;
314  cpe->ch[ch].pulse.num_pulse = 0;
315  for (w = 0; w < ics->num_windows*16; w += 16) {
316  for (g = 0; g < ics->num_swb; g++) {
317  //apply M/S
318  if (cpe->common_window && !ch && cpe->ms_mask[w + g]) {
319  for (i = 0; i < ics->swb_sizes[g]; i++) {
320  cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
321  cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
322  }
323  }
324  start += ics->swb_sizes[g];
325  }
326  for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--)
327  ;
328  maxsfb = FFMAX(maxsfb, cmaxsfb);
329  }
330  ics->max_sfb = maxsfb;
331 
332  //adjust zero bands for window groups
333  for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
334  for (g = 0; g < ics->max_sfb; g++) {
335  i = 1;
336  for (w2 = w; w2 < w + ics->group_len[w]; w2++) {
337  if (!cpe->ch[ch].zeroes[w2*16 + g]) {
338  i = 0;
339  break;
340  }
341  }
342  cpe->ch[ch].zeroes[w*16 + g] = i;
343  }
344  }
345  }
346 
347  if (chans > 1 && cpe->common_window) {
348  IndividualChannelStream *ics0 = &cpe->ch[0].ics;
349  IndividualChannelStream *ics1 = &cpe->ch[1].ics;
350  int msc = 0;
351  ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
352  ics1->max_sfb = ics0->max_sfb;
353  for (w = 0; w < ics0->num_windows*16; w += 16)
354  for (i = 0; i < ics0->max_sfb; i++)
355  if (cpe->ms_mask[w+i])
356  msc++;
357  if (msc == 0 || ics0->max_sfb == 0)
358  cpe->ms_mode = 0;
359  else
360  cpe->ms_mode = msc < ics0->max_sfb * ics0->num_windows ? 1 : 2;
361  }
362 }
363 
368 {
369  int w;
370 
371  for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
372  s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
373 }
374 
380 {
381  int off = sce->sf_idx[0], diff;
382  int i, w;
383 
384  for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
385  for (i = 0; i < sce->ics.max_sfb; i++) {
386  if (!sce->zeroes[w*16 + i]) {
387  diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO;
388  if (diff < 0 || diff > 120)
389  av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n");
390  off = sce->sf_idx[w*16 + i];
392  }
393  }
394  }
395 }
396 
400 static void encode_pulses(AACEncContext *s, Pulse *pulse)
401 {
402  int i;
403 
404  put_bits(&s->pb, 1, !!pulse->num_pulse);
405  if (!pulse->num_pulse)
406  return;
407 
408  put_bits(&s->pb, 2, pulse->num_pulse - 1);
409  put_bits(&s->pb, 6, pulse->start);
410  for (i = 0; i < pulse->num_pulse; i++) {
411  put_bits(&s->pb, 5, pulse->pos[i]);
412  put_bits(&s->pb, 4, pulse->amp[i]);
413  }
414 }
415 
420 {
421  int start, i, w, w2;
422 
423  for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
424  start = 0;
425  for (i = 0; i < sce->ics.max_sfb; i++) {
426  if (sce->zeroes[w*16 + i]) {
427  start += sce->ics.swb_sizes[i];
428  continue;
429  }
430  for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++)
431  s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128,
432  sce->ics.swb_sizes[i],
433  sce->sf_idx[w*16 + i],
434  sce->band_type[w*16 + i],
435  s->lambda);
436  start += sce->ics.swb_sizes[i];
437  }
438  }
439 }
440 
446  int common_window)
447 {
448  put_bits(&s->pb, 8, sce->sf_idx[0]);
449  if (!common_window)
450  put_ics_info(s, &sce->ics);
451  encode_band_info(s, sce);
452  encode_scale_factors(avctx, s, sce);
453  encode_pulses(s, &sce->pulse);
454  put_bits(&s->pb, 1, 0); //tns
455  put_bits(&s->pb, 1, 0); //ssr
456  encode_spectral_coeffs(s, sce);
457  return 0;
458 }
459 
463 static void put_bitstream_info(AACEncContext *s, const char *name)
464 {
465  int i, namelen, padbits;
466 
467  namelen = strlen(name) + 2;
468  put_bits(&s->pb, 3, TYPE_FIL);
469  put_bits(&s->pb, 4, FFMIN(namelen, 15));
470  if (namelen >= 15)
471  put_bits(&s->pb, 8, namelen - 14);
472  put_bits(&s->pb, 4, 0); //extension type - filler
473  padbits = -put_bits_count(&s->pb) & 7;
475  for (i = 0; i < namelen - 2; i++)
476  put_bits(&s->pb, 8, name[i]);
477  put_bits(&s->pb, 12 - padbits, 0);
478 }
479 
480 /*
481  * Copy input samples.
482  * Channels are reordered from Libav's default order to AAC order.
483  */
484 static void copy_input_samples(AACEncContext *s, const AVFrame *frame)
485 {
486  int ch;
487  int end = 2048 + (frame ? frame->nb_samples : 0);
488  const uint8_t *channel_map = aac_chan_maps[s->channels - 1];
489 
490  /* copy and remap input samples */
491  for (ch = 0; ch < s->channels; ch++) {
492  /* copy last 1024 samples of previous frame to the start of the current frame */
493  memcpy(&s->planar_samples[ch][1024], &s->planar_samples[ch][2048], 1024 * sizeof(s->planar_samples[0][0]));
494 
495  /* copy new samples and zero any remaining samples */
496  if (frame) {
497  memcpy(&s->planar_samples[ch][2048],
498  frame->extended_data[channel_map[ch]],
499  frame->nb_samples * sizeof(s->planar_samples[0][0]));
500  }
501  memset(&s->planar_samples[ch][end], 0,
502  (3072 - end) * sizeof(s->planar_samples[0][0]));
503  }
504 }
505 
506 static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
507  const AVFrame *frame, int *got_packet_ptr)
508 {
509  AACEncContext *s = avctx->priv_data;
510  float **samples = s->planar_samples, *samples2, *la, *overlap;
511  ChannelElement *cpe;
512  int i, ch, w, g, chans, tag, start_ch, ret;
513  int chan_el_counter[4];
515 
516  if (s->last_frame == 2)
517  return 0;
518 
519  /* add current frame to queue */
520  if (frame) {
521  if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
522  return ret;
523  }
524 
525  copy_input_samples(s, frame);
526  if (s->psypp)
528 
529  if (!avctx->frame_number)
530  return 0;
531 
532  start_ch = 0;
533  for (i = 0; i < s->chan_map[0]; i++) {
534  FFPsyWindowInfo* wi = windows + start_ch;
535  tag = s->chan_map[i+1];
536  chans = tag == TYPE_CPE ? 2 : 1;
537  cpe = &s->cpe[i];
538  for (ch = 0; ch < chans; ch++) {
539  IndividualChannelStream *ics = &cpe->ch[ch].ics;
540  int cur_channel = start_ch + ch;
541  overlap = &samples[cur_channel][0];
542  samples2 = overlap + 1024;
543  la = samples2 + (448+64);
544  if (!frame)
545  la = NULL;
546  if (tag == TYPE_LFE) {
547  wi[ch].window_type[0] = ONLY_LONG_SEQUENCE;
548  wi[ch].window_shape = 0;
549  wi[ch].num_windows = 1;
550  wi[ch].grouping[0] = 1;
551 
552  /* Only the lowest 12 coefficients are used in a LFE channel.
553  * The expression below results in only the bottom 8 coefficients
554  * being used for 11.025kHz to 16kHz sample rates.
555  */
556  ics->num_swb = s->samplerate_index >= 8 ? 1 : 3;
557  } else {
558  wi[ch] = s->psy.model->window(&s->psy, samples2, la, cur_channel,
559  ics->window_sequence[0]);
560  }
561  ics->window_sequence[1] = ics->window_sequence[0];
562  ics->window_sequence[0] = wi[ch].window_type[0];
563  ics->use_kb_window[1] = ics->use_kb_window[0];
564  ics->use_kb_window[0] = wi[ch].window_shape;
565  ics->num_windows = wi[ch].num_windows;
566  ics->swb_sizes = s->psy.bands [ics->num_windows == 8];
567  ics->num_swb = tag == TYPE_LFE ? ics->num_swb : s->psy.num_bands[ics->num_windows == 8];
568  for (w = 0; w < ics->num_windows; w++)
569  ics->group_len[w] = wi[ch].grouping[w];
570 
571  apply_window_and_mdct(s, &cpe->ch[ch], overlap);
572  }
573  start_ch += chans;
574  }
575  if ((ret = ff_alloc_packet(avpkt, 768 * s->channels))) {
576  av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
577  return ret;
578  }
579 
580  do {
581  int frame_bits;
582 
583  init_put_bits(&s->pb, avpkt->data, avpkt->size);
584 
585  if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT))
587  start_ch = 0;
588  memset(chan_el_counter, 0, sizeof(chan_el_counter));
589  for (i = 0; i < s->chan_map[0]; i++) {
590  FFPsyWindowInfo* wi = windows + start_ch;
591  const float *coeffs[2];
592  tag = s->chan_map[i+1];
593  chans = tag == TYPE_CPE ? 2 : 1;
594  cpe = &s->cpe[i];
595  put_bits(&s->pb, 3, tag);
596  put_bits(&s->pb, 4, chan_el_counter[tag]++);
597  for (ch = 0; ch < chans; ch++)
598  coeffs[ch] = cpe->ch[ch].coeffs;
599  s->psy.model->analyze(&s->psy, start_ch, coeffs, wi);
600  for (ch = 0; ch < chans; ch++) {
601  s->cur_channel = start_ch * 2 + ch;
602  s->coder->search_for_quantizers(avctx, s, &cpe->ch[ch], s->lambda);
603  }
604  cpe->common_window = 0;
605  if (chans > 1
606  && wi[0].window_type[0] == wi[1].window_type[0]
607  && wi[0].window_shape == wi[1].window_shape) {
608 
609  cpe->common_window = 1;
610  for (w = 0; w < wi[0].num_windows; w++) {
611  if (wi[0].grouping[w] != wi[1].grouping[w]) {
612  cpe->common_window = 0;
613  break;
614  }
615  }
616  }
617  s->cur_channel = start_ch * 2;
618  if (s->options.stereo_mode && cpe->common_window) {
619  if (s->options.stereo_mode > 0) {
620  IndividualChannelStream *ics = &cpe->ch[0].ics;
621  for (w = 0; w < ics->num_windows; w += ics->group_len[w])
622  for (g = 0; g < ics->num_swb; g++)
623  cpe->ms_mask[w*16+g] = 1;
624  } else if (s->coder->search_for_ms) {
625  s->coder->search_for_ms(s, cpe, s->lambda);
626  }
627  }
628  adjust_frame_information(cpe, chans);
629  if (chans == 2) {
630  put_bits(&s->pb, 1, cpe->common_window);
631  if (cpe->common_window) {
632  put_ics_info(s, &cpe->ch[0].ics);
633  encode_ms_info(&s->pb, cpe);
634  }
635  }
636  for (ch = 0; ch < chans; ch++) {
637  s->cur_channel = start_ch + ch;
638  encode_individual_channel(avctx, s, &cpe->ch[ch], cpe->common_window);
639  }
640  start_ch += chans;
641  }
642 
643  frame_bits = put_bits_count(&s->pb);
644  if (frame_bits <= 6144 * s->channels - 3) {
645  s->psy.bitres.bits = frame_bits / s->channels;
646  break;
647  }
648 
649  s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
650 
651  } while (1);
652 
653  put_bits(&s->pb, 3, TYPE_END);
654  flush_put_bits(&s->pb);
655  avctx->frame_bits = put_bits_count(&s->pb);
656 
657  // rate control stuff
658  if (!(avctx->flags & CODEC_FLAG_QSCALE)) {
659  float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
660  s->lambda *= ratio;
661  s->lambda = FFMIN(s->lambda, 65536.f);
662  }
663 
664  if (!frame)
665  s->last_frame++;
666 
667  ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
668  &avpkt->duration);
669 
670  avpkt->size = put_bits_count(&s->pb) >> 3;
671  *got_packet_ptr = 1;
672  return 0;
673 }
674 
676 {
677  AACEncContext *s = avctx->priv_data;
678 
679  ff_mdct_end(&s->mdct1024);
680  ff_mdct_end(&s->mdct128);
681  ff_psy_end(&s->psy);
682  if (s->psypp)
684  av_freep(&s->buffer.samples);
685  av_freep(&s->cpe);
686  ff_af_queue_close(&s->afq);
687 #if FF_API_OLD_ENCODE_AUDIO
688  av_freep(&avctx->coded_frame);
689 #endif
690  return 0;
691 }
692 
694 {
695  int ret = 0;
696 
697  ff_dsputil_init(&s->dsp, avctx);
699 
700  // window init
705 
706  if (ret = ff_mdct_init(&s->mdct1024, 11, 0, 32768.0))
707  return ret;
708  if (ret = ff_mdct_init(&s->mdct128, 8, 0, 32768.0))
709  return ret;
710 
711  return 0;
712 }
713 
715 {
716  int ch;
717  FF_ALLOCZ_OR_GOTO(avctx, s->buffer.samples, 3 * 1024 * s->channels * sizeof(s->buffer.samples[0]), alloc_fail);
718  FF_ALLOCZ_OR_GOTO(avctx, s->cpe, sizeof(ChannelElement) * s->chan_map[0], alloc_fail);
719  FF_ALLOCZ_OR_GOTO(avctx, avctx->extradata, 5 + FF_INPUT_BUFFER_PADDING_SIZE, alloc_fail);
720 
721  for(ch = 0; ch < s->channels; ch++)
722  s->planar_samples[ch] = s->buffer.samples + 3 * 1024 * ch;
723 
725  if (!(avctx->coded_frame = avcodec_alloc_frame()))
726  goto alloc_fail;
727 #endif
728 
729  return 0;
730 alloc_fail:
731  return AVERROR(ENOMEM);
732 }
733 
735 {
736  AACEncContext *s = avctx->priv_data;
737  int i, ret = 0;
738  const uint8_t *sizes[2];
739  uint8_t grouping[AAC_MAX_CHANNELS];
740  int lengths[2];
741 
742  avctx->frame_size = 1024;
743 
744  for (i = 0; i < 16; i++)
746  break;
747 
748  s->channels = avctx->channels;
749 
750  ERROR_IF(i == 16,
751  "Unsupported sample rate %d\n", avctx->sample_rate);
753  "Unsupported number of channels: %d\n", s->channels);
755  "Unsupported profile %d\n", avctx->profile);
756  ERROR_IF(1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * s->channels,
757  "Too many bits per frame requested\n");
758 
759  s->samplerate_index = i;
760 
762 
763  if (ret = dsp_init(avctx, s))
764  goto fail;
765 
766  if (ret = alloc_buffers(avctx, s))
767  goto fail;
768 
769  avctx->extradata_size = 5;
771 
772  sizes[0] = swb_size_1024[i];
773  sizes[1] = swb_size_128[i];
774  lengths[0] = ff_aac_num_swb_1024[i];
775  lengths[1] = ff_aac_num_swb_128[i];
776  for (i = 0; i < s->chan_map[0]; i++)
777  grouping[i] = s->chan_map[i + 1] == TYPE_CPE;
778  if (ret = ff_psy_init(&s->psy, avctx, 2, sizes, lengths, s->chan_map[0], grouping))
779  goto fail;
780  s->psypp = ff_psy_preprocess_init(avctx);
781  s->coder = &ff_aac_coders[2];
782 
783  s->lambda = avctx->global_quality ? avctx->global_quality : 120;
784 
786 
787  for (i = 0; i < 428; i++)
788  ff_aac_pow34sf_tab[i] = sqrt(ff_aac_pow2sf_tab[i] * sqrt(ff_aac_pow2sf_tab[i]));
789 
790  avctx->delay = 1024;
791  ff_af_queue_init(avctx, &s->afq);
792 
793  return 0;
794 fail:
795  aac_encode_end(avctx);
796  return ret;
797 }
798 
799 #define AACENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
800 static const AVOption aacenc_options[] = {
801  {"stereo_mode", "Stereo coding method", offsetof(AACEncContext, options.stereo_mode), AV_OPT_TYPE_INT, {.i64 = 0}, -1, 1, AACENC_FLAGS, "stereo_mode"},
802  {"auto", "Selected by the Encoder", 0, AV_OPT_TYPE_CONST, {.i64 = -1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
803  {"ms_off", "Disable Mid/Side coding", 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
804  {"ms_force", "Force Mid/Side for the whole frame if possible", 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
805  {NULL}
806 };
807 
808 static const AVClass aacenc_class = {
809  "AAC encoder",
813 };
814 
816  .name = "aac",
817  .type = AVMEDIA_TYPE_AUDIO,
818  .id = AV_CODEC_ID_AAC,
819  .priv_data_size = sizeof(AACEncContext),
821  .encode2 = aac_encode_frame,
823  .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY |
825  .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
827  .long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
828  .priv_class = &aacenc_class,
829 };
static const int16_t coeffs[28]
static const uint8_t aac_chan_configs[6][5]
default channel configurations
Definition: aacenc.c:139
av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx)
Definition: dsputil.c:2656
void(* search_for_ms)(struct AACEncContext *s, ChannelElement *cpe, const float lambda)
Definition: aacenc.h:47
uint8_t use_kb_window[2]
If set, use Kaiser-Bessel window, otherwise use a sinus window.
Definition: aac.h:161
static int16_t * samples
This structure describes decoded (raw) audio or video data.
Definition: avcodec.h:989
static const uint8_t swb_size_1024_64[]
Definition: aacenc.c:65
int grouping[8]
window grouping (for e.g. AAC)
Definition: psymodel.h:67
AVOption.
Definition: opt.h:233
uint8_t ** bands
scalefactor band sizes for possible frame sizes
Definition: psymodel.h:82
Definition: aac.h:204
void(* mdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input)
Definition: fft.h:83
static const AVClass aacenc_class
Definition: aacenc.c:808
av_cold void ff_kbd_window_init(float *window, float alpha, int n)
Generate a Kaiser-Bessel Derived Window.
Definition: kbdwin.c:26
#define SCALE_DIFF_ZERO
codebook index corresponding to zero scalefactor indices difference
Definition: aac.h:143
Definition: aac.h:57
AVFrame * coded_frame
the picture in the bitstream
Definition: avcodec.h:2725
Definition: aac.h:50
Definition: aac.h:51
av_cold void ff_psy_preprocess_end(struct FFPsyPreprocessContext *ctx)
Cleanup audio preprocessing module.
Definition: psymodel.c:130
int size
Definition: avcodec.h:916
struct AACEncContext::@3 buffer
AACCoefficientsEncoder * coder
Definition: aacenc.h:72
void avpriv_align_put_bits(PutBitContext *s)
Pad the bitstream with zeros up to the next byte boundary.
Definition: bitstream.c:45
static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
Encode ics_info element.
Definition: aacenc.c:270
#define AAC_MAX_CHANNELS
Definition: aacenc.c:49
int common_window
Set if channels share a common 'IndividualChannelStream' in bitstream.
Definition: aac.h:250
av_cold int ff_psy_init(FFPsyContext *ctx, AVCodecContext *avctx, int num_lens, const uint8_t **bands, const int *num_bands, int num_groups, const uint8_t *group_map)
Initialize psychoacoustic model.
Definition: psymodel.c:31
uint8_t ms_mask[128]
Set if mid/side stereo is used for each scalefactor window band.
Definition: aac.h:252
static const uint8_t swb_size_128_8[]
Definition: aacenc.c:125
float lambda
Definition: aacenc.h:75
int profile
profile
Definition: avcodec.h:2815
AVCodec.
Definition: avcodec.h:2960
static const uint8_t swb_size_1024_8[]
Definition: aacenc.c:96
static const uint8_t swb_size_128_96[]
Definition: aacenc.c:109
static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
Encode spectral coefficients processed by psychoacoustic model.
Definition: aacenc.c:419
int * num_bands
number of scalefactor bands for possible frame sizes
Definition: psymodel.h:83
void av_freep(void *arg)
Free a memory block which has been allocated with av_malloc(z)() or av_realloc() and set the pointer ...
Definition: mem.c:151
const uint8_t ff_aac_num_swb_128[]
Definition: aactab.c:43
AACEncOptions options
encoding options
Definition: aacenc.h:57
AAC encoder context.
Definition: aacenc.h:55
uint8_t
AVOptions.
#define WINDOW_FUNC(type)
Definition: aacenc.c:185
void(* quantize_and_encode_band)(struct AACEncContext *s, PutBitContext *pb, const float *in, int size, int scale_idx, int cb, const float lambda)
Definition: aacenc.h:45
SingleChannelElement ch[2]
Definition: aac.h:254
int samplerate_index
MPEG-4 samplerate index.
Definition: aacenc.h:65
Definition: aac.h:53
const uint8_t * chan_map
channel configuration map
Definition: aacenc.h:67
const uint8_t ff_aac_scalefactor_bits[121]
Definition: aactab.c:70
uint8_t * extradata
some codecs need / can use extradata like Huffman tables.
Definition: avcodec.h:1454
const char * name
AudioFrameQueue afq
Definition: aacenc.h:76
static const uint8_t swb_size_1024_48[]
Definition: aacenc.c:71
uint8_t * data
Definition: avcodec.h:915
uint32_t tag
Definition: movenc.c:802
AVFloatDSPContext fdsp
Definition: aacenc.h:62
int duration
Duration of this packet in AVStream->time_base units, 0 if unknown.
Definition: avcodec.h:937
static void adjust_frame_information(ChannelElement *cpe, int chans)
Produce integer coefficients from scalefactors provided by the model.
Definition: aacenc.c:305
static int init(AVCodecParserContext *s)
Definition: h264_parser.c:335
float, planar
Definition: samplefmt.h:60
static const AVOption aacenc_options[]
Definition: aacenc.c:800
static const uint8_t swb_size_1024_24[]
Definition: aacenc.c:84
float coeffs[1024]
coefficients for IMDCT
Definition: aac.h:237
static void(*const apply_window[4])(DSPContext *dsp, AVFloatDSPContext *fdsp, SingleChannelElement *sce, const float *audio)
Definition: aacenc.c:241
static const int sizes[][2]
Definition: img2dec.c:46
const uint8_t ff_aac_num_swb_1024[]
Definition: aactab.c:39
int last_frame
Definition: aacenc.h:74
sample_fmts
Definition: avconv_filter.c:63
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:88
int stereo_mode
Definition: aacenc.h:35
g
Definition: yuv2rgb.c:540
float ff_aac_kbd_long_1024[1024]
Definition: aactab.c:36
int flags
CODEC_FLAG_*.
Definition: avcodec.h:1434
int amp[4]
Definition: aac.h:208
void av_log(void *avcl, int level, const char *fmt,...)
Definition: log.c:146
const char * name
Name of the codec implementation.
Definition: avcodec.h:2967
int num_windows
number of windows in a frame
Definition: psymodel.h:66
static void copy_input_samples(AACEncContext *s, const AVFrame *frame)
Definition: aacenc.c:484
uint8_t max_sfb
number of scalefactor bands per group
Definition: aac.h:159
static void put_bits(PutBitContext *s, int n, unsigned int value)
Write up to 31 bits into a bitstream.
Definition: put_bits.h:136
#define ff_mdct_init
Definition: fft.h:146
Definition: aac.h:56
int num_swb
number of scalefactor window bands
Definition: aac.h:167
int ff_af_queue_add(AudioFrameQueue *afq, const AVFrame *f)
Add a frame to the queue.
void(* search_for_quantizers)(AVCodecContext *avctx, struct AACEncContext *s, SingleChannelElement *sce, const float lambda)
Definition: aacenc.h:41
int off
Definition: dsputil_bfin.c:28
static int put_bits_count(PutBitContext *s)
Definition: put_bits.h:70
#define AACENC_FLAGS
Definition: aacenc.c:799
AVFrame * avcodec_alloc_frame(void)
Allocate an AVFrame and set its fields to default values.
Definition: utils.c:616
struct AACEncContext AACEncContext
AAC encoder context.
int bit_rate
the average bitrate
Definition: avcodec.h:1404
enum WindowSequence window_sequence[2]
Definition: aac.h:160
int cur_channel
Definition: aacenc.h:73
static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, const AVFrame *frame, int *got_packet_ptr)
Definition: aacenc.c:506
void(* analyze)(FFPsyContext *ctx, int channel, const float **coeffs, const FFPsyWindowInfo *wi)
Perform psychoacoustic analysis and set band info (threshold, energy) for a group of channels...
Definition: psymodel.h:122
static const uint8_t aac_chan_maps[AAC_MAX_CHANNELS][AAC_MAX_CHANNELS]
Table to remap channels from Libav's default order to AAC order.
Definition: aacenc.c:151
int pos[4]
Definition: aac.h:207
int channels
channel count
Definition: aacenc.h:66
int ff_alloc_packet(AVPacket *avpkt, int size)
Check AVPacket size and/or allocate data.
Definition: utils.c:878
AAC definitions and structures.
FFTContext mdct128
short (128 samples) frame transform context
Definition: aacenc.h:60
PutBitContext pb
Definition: aacenc.h:58
float ff_aac_pow34sf_tab[428]
Definition: aacenc.c:57
static const uint8_t swb_size_128_48[]
Definition: aacenc.c:113
static const uint8_t swb_size_128_24[]
Definition: aacenc.c:117
LIBAVUTIL_VERSION_INT
Definition: eval.c:52
static const uint8_t swb_size_1024_16[]
Definition: aacenc.c:90
static av_cold int aac_encode_end(AVCodecContext *avctx)
Definition: aacenc.c:675
int frame_size
Number of samples per channel in an audio frame.
Definition: avcodec.h:2124
static const uint8_t swb_size_1024_32[]
Definition: aacenc.c:78
NULL
Definition: eval.c:52
external API header
static void put_audio_specific_config(AVCodecContext *avctx)
Make AAC audio config object.
Definition: aacenc.c:164
AV_SAMPLE_FMT_NONE
Definition: avconv_filter.c:63
int sample_rate
samples per second
Definition: avcodec.h:2104
float ff_aac_kbd_short_128[128]
Definition: aactab.c:37
static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
Encode MS data.
Definition: aacenc.c:291
FFPsyWindowInfo(* window)(FFPsyContext *ctx, const float *audio, const float *la, int channel, int prev_type)
Suggest window sequence for channel.
Definition: psymodel.h:112
av_default_item_name
Definition: dnxhdenc.c:43
int frame_bits
number of bits used for the previously encoded frame
Definition: avcodec.h:2495
main external API structure.
Definition: avcodec.h:1339
static void(WINAPI *cond_broadcast)(pthread_cond_t *cond)
static void close(AVCodecParserContext *s)
Definition: h264_parser.c:326
int bits
number of bits used in the bitresevoir
Definition: psymodel.h:88
DSPContext dsp
Definition: aacenc.h:61
IndividualChannelStream ics
Definition: aac.h:229
void(* encode_window_bands_info)(struct AACEncContext *s, SingleChannelElement *sce, int win, int group_len, const float lambda)
Definition: aacenc.h:43
int extradata_size
Definition: avcodec.h:1455
uint8_t group_len[8]
Definition: aac.h:163
Describe the class of an AVClass context structure.
Definition: log.h:33
static void put_bitstream_info(AACEncContext *s, const char *name)
Write some auxiliary information about the created AAC file.
Definition: aacenc.c:463
static const uint8_t swb_size_1024_96[]
Definition: aacenc.c:59
int window_shape
window shape (sine/KBD/whatever)
Definition: psymodel.h:65
static void encode_pulses(AACEncContext *s, Pulse *pulse)
Encode pulse data.
Definition: aacenc.c:400
static const uint8_t swb_size_128_16[]
Definition: aacenc.c:121
const uint8_t * swb_sizes
table of scalefactor band sizes for a particular window
Definition: aac.h:166
struct FFPsyContext::@51 bitres
FFPsyContext psy
Definition: aacenc.h:70
const uint32_t ff_aac_scalefactor_code[121]
Definition: aactab.c:51
static av_cold int alloc_buffers(AVCodecContext *avctx, AACEncContext *s)
Definition: aacenc.c:714
const struct FFPsyModel * model
encoder-specific model functions
Definition: psymodel.h:76
int ms_mode
Signals mid/side stereo flags coding mode (used by encoder)
Definition: aac.h:251
static const uint8_t * swb_size_1024[]
Definition: aacenc.c:102
struct FFPsyPreprocessContext * psypp
Definition: aacenc.h:71
int global_quality
Global quality for codecs which cannot change it per frame.
Definition: avcodec.h:1420
uint8_t zeroes[128]
band is not coded (used by encoder)
Definition: aac.h:236
int sf_idx[128]
scalefactor indices (used by encoder)
Definition: aac.h:235
AVCodec ff_aac_encoder
Definition: aacenc.c:815
const int avpriv_mpeg4audio_sample_rates[16]
Definition: mpeg4audio.c:55
const OptionDef options[]
Definition: avserver.c:4665
Y Spectral Band Replication.
Definition: mpeg4audio.h:64
float * samples
Definition: aacenc.h:81
static av_cold int aac_encode_init(AVCodecContext *avctx)
Definition: aacenc.c:734
common internal api header.
static void flush_put_bits(PutBitContext *s)
Pad the end of the output stream with zeros.
Definition: put_bits.h:86
Single Channel Element - used for both SCE and LFE elements.
Definition: aac.h:228
windowing related information
Definition: psymodel.h:63
#define ff_mdct_end
Definition: fft.h:147
av_cold struct FFPsyPreprocessContext * ff_psy_preprocess_init(AVCodecContext *avctx)
psychoacoustic model audio preprocessing initialization
Definition: psymodel.c:95
AACCoefficientsEncoder ff_aac_coders[]
Definition: aaccoder.c:1115
void ff_psy_preprocess(struct FFPsyPreprocessContext *ctx, float **audio, int channels)
Preprocess several channel in audio frame in order to compress it better.
Definition: psymodel.c:118
static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce)
Encode scalefactors.
Definition: aacenc.c:378
AVSampleFormat
Audio Sample Formats.
Definition: samplefmt.h:49
ChannelElement * cpe
channel elements
Definition: aacenc.h:69
Individual Channel Stream.
Definition: aac.h:158
float ff_aac_pow2sf_tab[428]
Definition: aac_tablegen.h:32
static void init_put_bits(PutBitContext *s, uint8_t *buffer, int buffer_size)
Initialize the PutBitContext s.
Definition: put_bits.h:52
static const uint8_t * swb_size_128[]
Definition: aacenc.c:129
DSP utils.
channel element - generic struct for SCE/CPE/CCE/LFE
Definition: aac.h:248
void * priv_data
Definition: avcodec.h:1382
int start
Definition: aac.h:206
FFTContext mdct1024
long (1024 samples) frame transform context
Definition: aacenc.h:59
#define ERROR_IF(cond,...)
Definition: aacenc.c:51
void ff_af_queue_init(AVCodecContext *avctx, AudioFrameQueue *afq)
Initialize AudioFrameQueue.
void ff_af_queue_remove(AudioFrameQueue *afq, int nb_samples, int64_t *pts, int *duration)
Remove frame(s) from the queue.
int channels
number of audio channels
Definition: avcodec.h:2105
int num_pulse
Definition: aac.h:205
static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
Encode scalefactor band coding type.
Definition: aacenc.c:367
void ff_af_queue_close(AudioFrameQueue *afq)
Close AudioFrameQueue.
enum BandType band_type[128]
band types
Definition: aac.h:232
int frame_number
Frame counter, set by libavcodec.
Definition: avcodec.h:2135
float ret_buf[2048]
PCM output buffer.
Definition: aac.h:239
void ff_aac_tableinit(void)
Definition: aac_tablegen.h:34
static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce, int common_window)
Encode one channel of audio data.
Definition: aacenc.c:444
void avpriv_float_dsp_init(AVFloatDSPContext *fdsp, int bit_exact)
Initialize a float DSP context.
Definition: float_dsp.c:55
static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce, float *audio)
Definition: aacenc.c:250
uint8_t ** extended_data
pointers to the data planes/channels.
Definition: avcodec.h:1028
AAC data declarations.
av_cold void ff_psy_end(FFPsyContext *ctx)
Cleanup model context at the end.
Definition: psymodel.c:76
This structure stores compressed data.
Definition: avcodec.h:898
int delay
Codec delay.
Definition: avcodec.h:1497
int window_type[3]
window type (short/long/transitional, etc.) - current, previous and next
Definition: psymodel.h:64
int nb_samples
number of audio samples (per channel) described by this frame
Definition: avcodec.h:1042
static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
Definition: aacenc.c:693
int64_t pts
Presentation timestamp in AVStream->time_base units; the time at which the decompressed packet will b...
Definition: avcodec.h:908
void ff_init_ff_sine_windows(int index)
initialize the specified entry of ff_sine_windows
#define FF_ALLOCZ_OR_GOTO(ctx, p, size, label)
Definition: internal.h:69
DSPContext.
Definition: dsputil.h:194
float * planar_samples[6]
saved preprocessed input
Definition: aacenc.h:63
if(!(ptr_align%ac->ptr_align)&&samples_align >=aligned_len)
bitstream writer API