FFmpeg  2.8.17
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Modules Pages
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 FFmpeg.
6  *
7  * FFmpeg 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  * FFmpeg 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 FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 /**
23  * @file
24  * AAC encoder
25  */
26 
27 /***********************************
28  * TODOs:
29  * add sane pulse detection
30  ***********************************/
31 
32 #include "libavutil/float_dsp.h"
33 #include "libavutil/opt.h"
34 #include "avcodec.h"
35 #include "put_bits.h"
36 #include "internal.h"
37 #include "mpeg4audio.h"
38 #include "kbdwin.h"
39 #include "sinewin.h"
40 
41 #include "aac.h"
42 #include "aactab.h"
43 #include "aacenc.h"
44 #include "aacenctab.h"
45 #include "aacenc_utils.h"
46 
47 #include "psymodel.h"
48 
49 /**
50  * Make AAC audio config object.
51  * @see 1.6.2.1 "Syntax - AudioSpecificConfig"
52  */
54 {
55  PutBitContext pb;
56  AACEncContext *s = avctx->priv_data;
57 
58  init_put_bits(&pb, avctx->extradata, avctx->extradata_size);
59  put_bits(&pb, 5, s->profile+1); //profile
60  put_bits(&pb, 4, s->samplerate_index); //sample rate index
61  put_bits(&pb, 4, s->channels);
62  //GASpecificConfig
63  put_bits(&pb, 1, 0); //frame length - 1024 samples
64  put_bits(&pb, 1, 0); //does not depend on core coder
65  put_bits(&pb, 1, 0); //is not extension
66 
67  //Explicitly Mark SBR absent
68  put_bits(&pb, 11, 0x2b7); //sync extension
69  put_bits(&pb, 5, AOT_SBR);
70  put_bits(&pb, 1, 0);
71  flush_put_bits(&pb);
72 }
73 
74 #define WINDOW_FUNC(type) \
75 static void apply_ ##type ##_window(AVFloatDSPContext *fdsp, \
76  SingleChannelElement *sce, \
77  const float *audio)
78 
79 WINDOW_FUNC(only_long)
80 {
81  const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
82  const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
83  float *out = sce->ret_buf;
84 
85  fdsp->vector_fmul (out, audio, lwindow, 1024);
86  fdsp->vector_fmul_reverse(out + 1024, audio + 1024, pwindow, 1024);
87 }
88 
89 WINDOW_FUNC(long_start)
90 {
91  const float *lwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
92  const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
93  float *out = sce->ret_buf;
94 
95  fdsp->vector_fmul(out, audio, lwindow, 1024);
96  memcpy(out + 1024, audio + 1024, sizeof(out[0]) * 448);
97  fdsp->vector_fmul_reverse(out + 1024 + 448, audio + 1024 + 448, swindow, 128);
98  memset(out + 1024 + 576, 0, sizeof(out[0]) * 448);
99 }
100 
101 WINDOW_FUNC(long_stop)
102 {
103  const float *lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
104  const float *swindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
105  float *out = sce->ret_buf;
106 
107  memset(out, 0, sizeof(out[0]) * 448);
108  fdsp->vector_fmul(out + 448, audio + 448, swindow, 128);
109  memcpy(out + 576, audio + 576, sizeof(out[0]) * 448);
110  fdsp->vector_fmul_reverse(out + 1024, audio + 1024, lwindow, 1024);
111 }
112 
113 WINDOW_FUNC(eight_short)
114 {
115  const float *swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
116  const float *pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
117  const float *in = audio + 448;
118  float *out = sce->ret_buf;
119  int w;
120 
121  for (w = 0; w < 8; w++) {
122  fdsp->vector_fmul (out, in, w ? pwindow : swindow, 128);
123  out += 128;
124  in += 128;
125  fdsp->vector_fmul_reverse(out, in, swindow, 128);
126  out += 128;
127  }
128 }
129 
130 static void (*const apply_window[4])(AVFloatDSPContext *fdsp,
132  const float *audio) = {
133  [ONLY_LONG_SEQUENCE] = apply_only_long_window,
134  [LONG_START_SEQUENCE] = apply_long_start_window,
135  [EIGHT_SHORT_SEQUENCE] = apply_eight_short_window,
136  [LONG_STOP_SEQUENCE] = apply_long_stop_window
137 };
138 
140  float *audio)
141 {
142  int i;
143  float *output = sce->ret_buf;
144 
145  apply_window[sce->ics.window_sequence[0]](s->fdsp, sce, audio);
146 
148  s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
149  else
150  for (i = 0; i < 1024; i += 128)
151  s->mdct128.mdct_calc(&s->mdct128, &sce->coeffs[i], output + i*2);
152  memcpy(audio, audio + 1024, sizeof(audio[0]) * 1024);
153  memcpy(sce->pcoeffs, sce->coeffs, sizeof(sce->pcoeffs));
154 }
155 
156 /**
157  * Encode ics_info element.
158  * @see Table 4.6 (syntax of ics_info)
159  */
161 {
162  int w;
163 
164  put_bits(&s->pb, 1, 0); // ics_reserved bit
165  put_bits(&s->pb, 2, info->window_sequence[0]);
166  put_bits(&s->pb, 1, info->use_kb_window[0]);
167  if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
168  put_bits(&s->pb, 6, info->max_sfb);
169  put_bits(&s->pb, 1, !!info->predictor_present);
170  } else {
171  put_bits(&s->pb, 4, info->max_sfb);
172  for (w = 1; w < 8; w++)
173  put_bits(&s->pb, 1, !info->group_len[w]);
174  }
175 }
176 
177 /**
178  * Encode MS data.
179  * @see 4.6.8.1 "Joint Coding - M/S Stereo"
180  */
182 {
183  int i, w;
184 
185  put_bits(pb, 2, cpe->ms_mode);
186  if (cpe->ms_mode == 1)
187  for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w])
188  for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
189  put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
190 }
191 
192 /**
193  * Produce integer coefficients from scalefactors provided by the model.
194  */
195 static void adjust_frame_information(ChannelElement *cpe, int chans)
196 {
197  int i, w, w2, g, ch;
198  int maxsfb, cmaxsfb;
199 
200  for (ch = 0; ch < chans; ch++) {
201  IndividualChannelStream *ics = &cpe->ch[ch].ics;
202  maxsfb = 0;
203  cpe->ch[ch].pulse.num_pulse = 0;
204  for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
205  for (w2 = 0; w2 < ics->group_len[w]; w2++) {
206  for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w*16+cmaxsfb-1]; cmaxsfb--)
207  ;
208  maxsfb = FFMAX(maxsfb, cmaxsfb);
209  }
210  }
211  ics->max_sfb = maxsfb;
212 
213  //adjust zero bands for window groups
214  for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
215  for (g = 0; g < ics->max_sfb; g++) {
216  i = 1;
217  for (w2 = w; w2 < w + ics->group_len[w]; w2++) {
218  if (!cpe->ch[ch].zeroes[w2*16 + g]) {
219  i = 0;
220  break;
221  }
222  }
223  cpe->ch[ch].zeroes[w*16 + g] = i;
224  }
225  }
226  }
227 
228  if (chans > 1 && cpe->common_window) {
229  IndividualChannelStream *ics0 = &cpe->ch[0].ics;
230  IndividualChannelStream *ics1 = &cpe->ch[1].ics;
231  int msc = 0;
232  ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
233  ics1->max_sfb = ics0->max_sfb;
234  for (w = 0; w < ics0->num_windows*16; w += 16)
235  for (i = 0; i < ics0->max_sfb; i++)
236  if (cpe->ms_mask[w+i])
237  msc++;
238  if (msc == 0 || ics0->max_sfb == 0)
239  cpe->ms_mode = 0;
240  else
241  cpe->ms_mode = msc < ics0->max_sfb * ics0->num_windows ? 1 : 2;
242  }
243 }
244 
246 {
247  int w, w2, g, i;
248  IndividualChannelStream *ics = &cpe->ch[0].ics;
249  if (!cpe->common_window)
250  return;
251  for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
252  for (w2 = 0; w2 < ics->group_len[w]; w2++) {
253  int start = (w+w2) * 128;
254  for (g = 0; g < ics->num_swb; g++) {
255  int p = -1 + 2 * (cpe->ch[1].band_type[w*16+g] - 14);
256  float scale = cpe->ch[0].is_ener[w*16+g];
257  if (!cpe->is_mask[w*16 + g]) {
258  start += ics->swb_sizes[g];
259  continue;
260  }
261  for (i = 0; i < ics->swb_sizes[g]; i++) {
262  float sum = (cpe->ch[0].coeffs[start+i] + p*cpe->ch[1].coeffs[start+i])*scale;
263  cpe->ch[0].coeffs[start+i] = sum;
264  cpe->ch[1].coeffs[start+i] = 0.0f;
265  }
266  start += ics->swb_sizes[g];
267  }
268  }
269  }
270 }
271 
273 {
274  int w, w2, g, i;
275  IndividualChannelStream *ics = &cpe->ch[0].ics;
276  if (!cpe->common_window)
277  return;
278  for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
279  for (w2 = 0; w2 < ics->group_len[w]; w2++) {
280  int start = (w+w2) * 128;
281  for (g = 0; g < ics->num_swb; g++) {
282  if (!cpe->ms_mask[w*16 + g]) {
283  start += ics->swb_sizes[g];
284  continue;
285  }
286  for (i = 0; i < ics->swb_sizes[g]; i++) {
287  float L = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) * 0.5f;
288  float R = L - cpe->ch[1].coeffs[start+i];
289  cpe->ch[0].coeffs[start+i] = L;
290  cpe->ch[1].coeffs[start+i] = R;
291  }
292  start += ics->swb_sizes[g];
293  }
294  }
295  }
296 }
297 
298 /**
299  * Encode scalefactor band coding type.
300  */
302 {
303  int w;
304 
307 
308  for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
309  s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
310 }
311 
312 /**
313  * Encode scalefactors.
314  */
317 {
318  int diff, off_sf = sce->sf_idx[0], off_pns = sce->sf_idx[0] - NOISE_OFFSET;
319  int off_is = 0, noise_flag = 1;
320  int i, w;
321 
322  for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
323  for (i = 0; i < sce->ics.max_sfb; i++) {
324  if (!sce->zeroes[w*16 + i]) {
325  if (sce->band_type[w*16 + i] == NOISE_BT) {
326  diff = sce->sf_idx[w*16 + i] - off_pns;
327  off_pns = sce->sf_idx[w*16 + i];
328  if (noise_flag-- > 0) {
329  put_bits(&s->pb, NOISE_PRE_BITS, diff + NOISE_PRE);
330  continue;
331  }
332  } else if (sce->band_type[w*16 + i] == INTENSITY_BT ||
333  sce->band_type[w*16 + i] == INTENSITY_BT2) {
334  diff = sce->sf_idx[w*16 + i] - off_is;
335  off_is = sce->sf_idx[w*16 + i];
336  } else {
337  diff = sce->sf_idx[w*16 + i] - off_sf;
338  off_sf = sce->sf_idx[w*16 + i];
339  }
340  diff += SCALE_DIFF_ZERO;
341  av_assert0(diff >= 0 && diff <= 120);
343  }
344  }
345  }
346 }
347 
348 /**
349  * Encode pulse data.
350  */
351 static void encode_pulses(AACEncContext *s, Pulse *pulse)
352 {
353  int i;
354 
355  put_bits(&s->pb, 1, !!pulse->num_pulse);
356  if (!pulse->num_pulse)
357  return;
358 
359  put_bits(&s->pb, 2, pulse->num_pulse - 1);
360  put_bits(&s->pb, 6, pulse->start);
361  for (i = 0; i < pulse->num_pulse; i++) {
362  put_bits(&s->pb, 5, pulse->pos[i]);
363  put_bits(&s->pb, 4, pulse->amp[i]);
364  }
365 }
366 
367 /**
368  * Encode spectral coefficients processed by psychoacoustic model.
369  */
371 {
372  int start, i, w, w2;
373 
374  for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
375  start = 0;
376  for (i = 0; i < sce->ics.max_sfb; i++) {
377  if (sce->zeroes[w*16 + i]) {
378  start += sce->ics.swb_sizes[i];
379  continue;
380  }
381  for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++) {
382  s->coder->quantize_and_encode_band(s, &s->pb,
383  &sce->coeffs[start + w2*128],
384  NULL, sce->ics.swb_sizes[i],
385  sce->sf_idx[w*16 + i],
386  sce->band_type[w*16 + i],
387  s->lambda,
388  sce->ics.window_clipping[w]);
389  }
390  start += sce->ics.swb_sizes[i];
391  }
392  }
393 }
394 
395 /**
396  * Downscale spectral coefficients for near-clipping windows to avoid artifacts
397  */
399 {
400  int start, i, j, w;
401 
402  if (sce->ics.clip_avoidance_factor < 1.0f) {
403  for (w = 0; w < sce->ics.num_windows; w++) {
404  start = 0;
405  for (i = 0; i < sce->ics.max_sfb; i++) {
406  float *swb_coeffs = &sce->coeffs[start + w*128];
407  for (j = 0; j < sce->ics.swb_sizes[i]; j++)
408  swb_coeffs[j] *= sce->ics.clip_avoidance_factor;
409  start += sce->ics.swb_sizes[i];
410  }
411  }
412  }
413 }
414 
415 /**
416  * Encode one channel of audio data.
417  */
420  int common_window)
421 {
422  put_bits(&s->pb, 8, sce->sf_idx[0]);
423  if (!common_window) {
424  put_ics_info(s, &sce->ics);
425  if (s->coder->encode_main_pred)
426  s->coder->encode_main_pred(s, sce);
427  }
428  encode_band_info(s, sce);
429  encode_scale_factors(avctx, s, sce);
430  encode_pulses(s, &sce->pulse);
431  put_bits(&s->pb, 1, !!sce->tns.present);
432  if (s->coder->encode_tns_info)
433  s->coder->encode_tns_info(s, sce);
434  put_bits(&s->pb, 1, 0); //ssr
435  encode_spectral_coeffs(s, sce);
436  return 0;
437 }
438 
439 /**
440  * Write some auxiliary information about the created AAC file.
441  */
442 static void put_bitstream_info(AACEncContext *s, const char *name)
443 {
444  int i, namelen, padbits;
445 
446  namelen = strlen(name) + 2;
447  put_bits(&s->pb, 3, TYPE_FIL);
448  put_bits(&s->pb, 4, FFMIN(namelen, 15));
449  if (namelen >= 15)
450  put_bits(&s->pb, 8, namelen - 14);
451  put_bits(&s->pb, 4, 0); //extension type - filler
452  padbits = -put_bits_count(&s->pb) & 7;
454  for (i = 0; i < namelen - 2; i++)
455  put_bits(&s->pb, 8, name[i]);
456  put_bits(&s->pb, 12 - padbits, 0);
457 }
458 
459 /*
460  * Copy input samples.
461  * Channels are reordered from libavcodec's default order to AAC order.
462  */
464 {
465  int ch;
466  int end = 2048 + (frame ? frame->nb_samples : 0);
467  const uint8_t *channel_map = aac_chan_maps[s->channels - 1];
468 
469  /* copy and remap input samples */
470  for (ch = 0; ch < s->channels; ch++) {
471  /* copy last 1024 samples of previous frame to the start of the current frame */
472  memcpy(&s->planar_samples[ch][1024], &s->planar_samples[ch][2048], 1024 * sizeof(s->planar_samples[0][0]));
473 
474  /* copy new samples and zero any remaining samples */
475  if (frame) {
476  memcpy(&s->planar_samples[ch][2048],
477  frame->extended_data[channel_map[ch]],
478  frame->nb_samples * sizeof(s->planar_samples[0][0]));
479  }
480  memset(&s->planar_samples[ch][end], 0,
481  (3072 - end) * sizeof(s->planar_samples[0][0]));
482  }
483 }
484 
485 static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
486  const AVFrame *frame, int *got_packet_ptr)
487 {
488  AACEncContext *s = avctx->priv_data;
489  float **samples = s->planar_samples, *samples2, *la, *overlap;
490  ChannelElement *cpe;
492  int i, ch, w, chans, tag, start_ch, ret;
493  int ms_mode = 0, is_mode = 0, tns_mode = 0, pred_mode = 0;
494  int chan_el_counter[4];
496  int k;
497 
498  if (s->last_frame == 2)
499  return 0;
500 
501  /* add current frame to queue */
502  if (frame) {
503  if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
504  return ret;
505  }
506 
507  copy_input_samples(s, frame);
508  if (s->psypp)
510 
511  if (!avctx->frame_number)
512  return 0;
513 
514  start_ch = 0;
515  for (i = 0; i < s->chan_map[0]; i++) {
516  FFPsyWindowInfo* wi = windows + start_ch;
517  tag = s->chan_map[i+1];
518  chans = tag == TYPE_CPE ? 2 : 1;
519  cpe = &s->cpe[i];
520  for (ch = 0; ch < chans; ch++) {
521  IndividualChannelStream *ics = &cpe->ch[ch].ics;
522  int cur_channel = start_ch + ch;
523  float clip_avoidance_factor;
524  overlap = &samples[cur_channel][0];
525  samples2 = overlap + 1024;
526  la = samples2 + (448+64);
527  if (!frame)
528  la = NULL;
529  if (tag == TYPE_LFE) {
530  wi[ch].window_type[0] = ONLY_LONG_SEQUENCE;
531  wi[ch].window_shape = 0;
532  wi[ch].num_windows = 1;
533  wi[ch].grouping[0] = 1;
534 
535  /* Only the lowest 12 coefficients are used in a LFE channel.
536  * The expression below results in only the bottom 8 coefficients
537  * being used for 11.025kHz to 16kHz sample rates.
538  */
539  ics->num_swb = s->samplerate_index >= 8 ? 1 : 3;
540  } else {
541  wi[ch] = s->psy.model->window(&s->psy, samples2, la, cur_channel,
542  ics->window_sequence[0]);
543  }
544  ics->window_sequence[1] = ics->window_sequence[0];
545  ics->window_sequence[0] = wi[ch].window_type[0];
546  ics->use_kb_window[1] = ics->use_kb_window[0];
547  ics->use_kb_window[0] = wi[ch].window_shape;
548  ics->num_windows = wi[ch].num_windows;
549  ics->swb_sizes = s->psy.bands [ics->num_windows == 8];
550  ics->num_swb = tag == TYPE_LFE ? ics->num_swb : s->psy.num_bands[ics->num_windows == 8];
551  ics->max_sfb = FFMIN(ics->max_sfb, ics->num_swb);
552  ics->swb_offset = wi[ch].window_type[0] == EIGHT_SHORT_SEQUENCE ?
555  ics->tns_max_bands = wi[ch].window_type[0] == EIGHT_SHORT_SEQUENCE ?
558  clip_avoidance_factor = 0.0f;
559  for (w = 0; w < ics->num_windows; w++)
560  ics->group_len[w] = wi[ch].grouping[w];
561  for (w = 0; w < ics->num_windows; w++) {
562  if (wi[ch].clipping[w] > CLIP_AVOIDANCE_FACTOR) {
563  ics->window_clipping[w] = 1;
564  clip_avoidance_factor = FFMAX(clip_avoidance_factor, wi[ch].clipping[w]);
565  } else {
566  ics->window_clipping[w] = 0;
567  }
568  }
569  if (clip_avoidance_factor > CLIP_AVOIDANCE_FACTOR) {
570  ics->clip_avoidance_factor = CLIP_AVOIDANCE_FACTOR / clip_avoidance_factor;
571  } else {
572  ics->clip_avoidance_factor = 1.0f;
573  }
574 
575  apply_window_and_mdct(s, &cpe->ch[ch], overlap);
576 
577  for (k = 0; k < 1024; k++) {
578  if (!(fabs(cpe->ch[ch].coeffs[k]) < 1E16)) { // Ensure headroom for energy calculation
579  av_log(avctx, AV_LOG_ERROR, "Input contains (near) NaN/+-Inf\n");
580  return AVERROR(EINVAL);
581  }
582  }
583  avoid_clipping(s, &cpe->ch[ch]);
584  }
585  start_ch += chans;
586  }
587  if ((ret = ff_alloc_packet2(avctx, avpkt, 8192 * s->channels, 0)) < 0)
588  return ret;
589  do {
590  int frame_bits;
591 
592  init_put_bits(&s->pb, avpkt->data, avpkt->size);
593 
594  if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & AV_CODEC_FLAG_BITEXACT))
596  start_ch = 0;
597  memset(chan_el_counter, 0, sizeof(chan_el_counter));
598  for (i = 0; i < s->chan_map[0]; i++) {
599  FFPsyWindowInfo* wi = windows + start_ch;
600  const float *coeffs[2];
601  tag = s->chan_map[i+1];
602  chans = tag == TYPE_CPE ? 2 : 1;
603  cpe = &s->cpe[i];
604  cpe->common_window = 0;
605  memset(cpe->is_mask, 0, sizeof(cpe->is_mask));
606  memset(cpe->ms_mask, 0, sizeof(cpe->ms_mask));
607  put_bits(&s->pb, 3, tag);
608  put_bits(&s->pb, 4, chan_el_counter[tag]++);
609  for (ch = 0; ch < chans; ch++) {
610  sce = &cpe->ch[ch];
611  coeffs[ch] = sce->coeffs;
612  sce->ics.predictor_present = 0;
613  memset(&sce->ics.prediction_used, 0, sizeof(sce->ics.prediction_used));
614  memset(&sce->tns, 0, sizeof(TemporalNoiseShaping));
615  for (w = 0; w < 128; w++)
616  if (sce->band_type[w] > RESERVED_BT)
617  sce->band_type[w] = 0;
618  }
619  s->psy.model->analyze(&s->psy, start_ch, coeffs, wi);
620  for (ch = 0; ch < chans; ch++) {
621  s->cur_channel = start_ch + ch;
622  s->coder->search_for_quantizers(avctx, s, &cpe->ch[ch], s->lambda);
623  }
624  if (chans > 1
625  && wi[0].window_type[0] == wi[1].window_type[0]
626  && wi[0].window_shape == wi[1].window_shape) {
627 
628  cpe->common_window = 1;
629  for (w = 0; w < wi[0].num_windows; w++) {
630  if (wi[0].grouping[w] != wi[1].grouping[w]) {
631  cpe->common_window = 0;
632  break;
633  }
634  }
635  }
636  for (ch = 0; ch < chans; ch++) { /* TNS and PNS */
637  sce = &cpe->ch[ch];
638  s->cur_channel = start_ch + ch;
639  if (s->options.pns && s->coder->search_for_pns)
640  s->coder->search_for_pns(s, avctx, sce);
641  if (s->options.tns && s->coder->search_for_tns)
642  s->coder->search_for_tns(s, sce);
643  if (s->options.tns && s->coder->apply_tns_filt)
644  s->coder->apply_tns_filt(s, sce);
645  if (sce->tns.present)
646  tns_mode = 1;
647  }
648  s->cur_channel = start_ch;
649  if (s->options.intensity_stereo) { /* Intensity Stereo */
650  if (s->coder->search_for_is)
651  s->coder->search_for_is(s, avctx, cpe);
652  if (cpe->is_mode) is_mode = 1;
654  }
655  if (s->options.pred) { /* Prediction */
656  for (ch = 0; ch < chans; ch++) {
657  sce = &cpe->ch[ch];
658  s->cur_channel = start_ch + ch;
659  if (s->options.pred && s->coder->search_for_pred)
660  s->coder->search_for_pred(s, sce);
661  if (cpe->ch[ch].ics.predictor_present) pred_mode = 1;
662  }
664  s->coder->adjust_common_prediction(s, cpe);
665  for (ch = 0; ch < chans; ch++) {
666  sce = &cpe->ch[ch];
667  s->cur_channel = start_ch + ch;
668  if (s->options.pred && s->coder->apply_main_pred)
669  s->coder->apply_main_pred(s, sce);
670  }
671  s->cur_channel = start_ch;
672  }
673  if (s->options.stereo_mode) { /* Mid/Side stereo */
674  if (s->options.stereo_mode == -1 && s->coder->search_for_ms)
675  s->coder->search_for_ms(s, cpe);
676  else if (cpe->common_window)
677  memset(cpe->ms_mask, 1, sizeof(cpe->ms_mask));
678  for (w = 0; w < 128; w++)
679  cpe->ms_mask[w] = cpe->is_mask[w] ? 0 : cpe->ms_mask[w];
681  }
682  adjust_frame_information(cpe, chans);
683  if (chans == 2) {
684  put_bits(&s->pb, 1, cpe->common_window);
685  if (cpe->common_window) {
686  put_ics_info(s, &cpe->ch[0].ics);
687  if (s->coder->encode_main_pred)
688  s->coder->encode_main_pred(s, &cpe->ch[0]);
689  encode_ms_info(&s->pb, cpe);
690  if (cpe->ms_mode) ms_mode = 1;
691  }
692  }
693  for (ch = 0; ch < chans; ch++) {
694  s->cur_channel = start_ch + ch;
695  encode_individual_channel(avctx, s, &cpe->ch[ch], cpe->common_window);
696  }
697  start_ch += chans;
698  }
699 
700  frame_bits = put_bits_count(&s->pb);
701  if (frame_bits <= 6144 * s->channels - 3) {
702  s->psy.bitres.bits = frame_bits / s->channels;
703  break;
704  }
705  if (is_mode || ms_mode || tns_mode || pred_mode) {
706  for (i = 0; i < s->chan_map[0]; i++) {
707  // Must restore coeffs
708  chans = tag == TYPE_CPE ? 2 : 1;
709  cpe = &s->cpe[i];
710  for (ch = 0; ch < chans; ch++)
711  memcpy(cpe->ch[ch].coeffs, cpe->ch[ch].pcoeffs, sizeof(cpe->ch[ch].coeffs));
712  }
713  }
714 
715  s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
716 
717  } while (1);
718 
719  put_bits(&s->pb, 3, TYPE_END);
720  flush_put_bits(&s->pb);
721  avctx->frame_bits = put_bits_count(&s->pb);
722 
723  // rate control stuff
724  if (!(avctx->flags & AV_CODEC_FLAG_QSCALE)) {
725  float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
726  s->lambda *= ratio;
727  s->lambda = FFMIN(s->lambda, 65536.f);
728  }
729 
730  if (!frame)
731  s->last_frame++;
732 
733  ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
734  &avpkt->duration);
735 
736  avpkt->size = put_bits_count(&s->pb) >> 3;
737  *got_packet_ptr = 1;
738  return 0;
739 }
740 
742 {
743  AACEncContext *s = avctx->priv_data;
744 
745  ff_mdct_end(&s->mdct1024);
746  ff_mdct_end(&s->mdct128);
747  ff_psy_end(&s->psy);
748  ff_lpc_end(&s->lpc);
749  if (s->psypp)
751  av_freep(&s->buffer.samples);
752  av_freep(&s->cpe);
753  av_freep(&s->fdsp);
754  ff_af_queue_close(&s->afq);
755  return 0;
756 }
757 
759 {
760  int ret = 0;
761 
763  if (!s->fdsp)
764  return AVERROR(ENOMEM);
765 
766  // window init
771 
772  if ((ret = ff_mdct_init(&s->mdct1024, 11, 0, 32768.0)) < 0)
773  return ret;
774  if ((ret = ff_mdct_init(&s->mdct128, 8, 0, 32768.0)) < 0)
775  return ret;
776 
777  return 0;
778 }
779 
781 {
782  int ch;
783  FF_ALLOCZ_ARRAY_OR_GOTO(avctx, s->buffer.samples, s->channels, 3 * 1024 * sizeof(s->buffer.samples[0]), alloc_fail);
784  FF_ALLOCZ_ARRAY_OR_GOTO(avctx, s->cpe, s->chan_map[0], sizeof(ChannelElement), alloc_fail);
785  FF_ALLOCZ_OR_GOTO(avctx, avctx->extradata, 5 + AV_INPUT_BUFFER_PADDING_SIZE, alloc_fail);
786 
787  for(ch = 0; ch < s->channels; ch++)
788  s->planar_samples[ch] = s->buffer.samples + 3 * 1024 * ch;
789 
790  return 0;
791 alloc_fail:
792  return AVERROR(ENOMEM);
793 }
794 
796 {
797  AACEncContext *s = avctx->priv_data;
798  int i, ret = 0;
799  const uint8_t *sizes[2];
800  uint8_t grouping[AAC_MAX_CHANNELS];
801  int lengths[2];
802 
803  avctx->frame_size = 1024;
804 
805  for (i = 0; i < 16; i++)
807  break;
808 
809  s->channels = avctx->channels;
810 
812  "Unsupported sample rate %d\n", avctx->sample_rate);
814  "Unsupported number of channels: %d\n", s->channels);
815  WARN_IF(1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * s->channels,
816  "Too many bits per frame requested, clamping to max\n");
817  if (avctx->profile == FF_PROFILE_AAC_MAIN) {
818  s->options.pred = 1;
819  } else if ((avctx->profile == FF_PROFILE_AAC_LOW ||
820  avctx->profile == FF_PROFILE_UNKNOWN) && s->options.pred) {
821  s->profile = 0; /* Main */
822  WARN_IF(1, "Prediction requested, changing profile to AAC-Main\n");
823  } else if (avctx->profile == FF_PROFILE_AAC_LOW ||
824  avctx->profile == FF_PROFILE_UNKNOWN) {
825  s->profile = 1; /* Low */
826  } else {
827  ERROR_IF(1, "Unsupported profile %d\n", avctx->profile);
828  }
829 
830  if (s->options.aac_coder != AAC_CODER_TWOLOOP) {
831  s->options.intensity_stereo = 0;
832  s->options.pns = 0;
833  }
834 
835  avctx->bit_rate = (int)FFMIN(
836  6144 * s->channels / 1024.0 * avctx->sample_rate,
837  avctx->bit_rate);
838 
839  s->samplerate_index = i;
840 
842 
843  if ((ret = dsp_init(avctx, s)) < 0)
844  goto fail;
845 
846  if ((ret = alloc_buffers(avctx, s)) < 0)
847  goto fail;
848 
849  avctx->extradata_size = 5;
851 
852  sizes[0] = ff_aac_swb_size_1024[i];
853  sizes[1] = ff_aac_swb_size_128[i];
854  lengths[0] = ff_aac_num_swb_1024[i];
855  lengths[1] = ff_aac_num_swb_128[i];
856  for (i = 0; i < s->chan_map[0]; i++)
857  grouping[i] = s->chan_map[i + 1] == TYPE_CPE;
858  if ((ret = ff_psy_init(&s->psy, avctx, 2, sizes, lengths,
859  s->chan_map[0], grouping)) < 0)
860  goto fail;
861  s->psypp = ff_psy_preprocess_init(avctx);
864 
865  if (HAVE_MIPSDSPR1)
867 
868  s->lambda = avctx->global_quality > 0 ? avctx->global_quality : 120;
869 
871 
872  avctx->initial_padding = 1024;
873  ff_af_queue_init(avctx, &s->afq);
874 
875  return 0;
876 fail:
877  aac_encode_end(avctx);
878  return ret;
879 }
880 
881 #define AACENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
882 static const AVOption aacenc_options[] = {
883  {"stereo_mode", "Stereo coding method", offsetof(AACEncContext, options.stereo_mode), AV_OPT_TYPE_INT, {.i64 = 0}, -1, 1, AACENC_FLAGS, "stereo_mode"},
884  {"auto", "Selected by the Encoder", 0, AV_OPT_TYPE_CONST, {.i64 = -1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
885  {"ms_off", "Disable Mid/Side coding", 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
886  {"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"},
887  {"aac_coder", "Coding algorithm", offsetof(AACEncContext, options.aac_coder), AV_OPT_TYPE_INT, {.i64 = AAC_CODER_TWOLOOP}, 0, AAC_CODER_NB-1, AACENC_FLAGS, "aac_coder"},
888  {"faac", "FAAC-inspired method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_FAAC}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
889  {"anmr", "ANMR method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_ANMR}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
890  {"twoloop", "Two loop searching method", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_TWOLOOP}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
891  {"fast", "Constant quantizer", 0, AV_OPT_TYPE_CONST, {.i64 = AAC_CODER_FAST}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_coder"},
892  {"aac_pns", "Perceptual Noise Substitution", offsetof(AACEncContext, options.pns), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, AACENC_FLAGS, "aac_pns"},
893  {"disable", "Disable perceptual noise substitution", 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_pns"},
894  {"enable", "Enable perceptual noise substitution", 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_pns"},
895  {"aac_is", "Intensity stereo coding", offsetof(AACEncContext, options.intensity_stereo), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, AACENC_FLAGS, "intensity_stereo"},
896  {"disable", "Disable intensity stereo coding", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, INT_MIN, INT_MAX, AACENC_FLAGS, "intensity_stereo"},
897  {"enable", "Enable intensity stereo coding", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, INT_MIN, INT_MAX, AACENC_FLAGS, "intensity_stereo"},
898  {"aac_tns", "Temporal noise shaping", offsetof(AACEncContext, options.tns), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AACENC_FLAGS, "aac_tns"},
899  {"disable", "Disable temporal noise shaping", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_tns"},
900  {"enable", "Enable temporal noise shaping", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_tns"},
901  {"aac_pred", "AAC-Main prediction", offsetof(AACEncContext, options.pred), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AACENC_FLAGS, "aac_pred"},
902  {"disable", "Disable AAC-Main prediction", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_pred"},
903  {"enable", "Enable AAC-Main prediction", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, INT_MIN, INT_MAX, AACENC_FLAGS, "aac_pred"},
904  {NULL}
905 };
906 
907 static const AVClass aacenc_class = {
908  "AAC encoder",
912 };
913 
915  .name = "aac",
916  .long_name = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
917  .type = AVMEDIA_TYPE_AUDIO,
918  .id = AV_CODEC_ID_AAC,
919  .priv_data_size = sizeof(AACEncContext),
921  .encode2 = aac_encode_frame,
922  .close = aac_encode_end,
926  .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
928  .priv_class = &aacenc_class,
929 };
float, planar
Definition: samplefmt.h:70
#define NULL
Definition: coverity.c:32
const char * s
Definition: avisynth_c.h:631
void(* search_for_pns)(struct AACEncContext *s, AVCodecContext *avctx, SingleChannelElement *sce)
Definition: aacenc.h:68
Band types following are encoded differently from others.
Definition: aac.h:86
uint8_t use_kb_window[2]
If set, use Kaiser-Bessel window, otherwise use a sine window.
Definition: aac.h:174
AACCoefficientsEncoder ff_aac_coders[AAC_CODER_NB]
Definition: aaccoder.c:958
This structure describes decoded (raw) audio or video data.
Definition: frame.h:171
#define FF_ALLOCZ_ARRAY_OR_GOTO(ctx, p, nelem, elsize, label)
Definition: internal.h:159
int grouping[8]
window grouping (for e.g. AAC)
Definition: psymodel.h:68
AVOption.
Definition: opt.h:255
uint8_t ** bands
scalefactor band sizes for possible frame sizes
Definition: psymodel.h:84
Definition: aac.h:221
void(* mdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input)
Definition: fft.h:109
static const AVClass aacenc_class
Definition: aacenc.c:907
static void put_bits(Jpeg2000EncoderContext *s, int val, int n)
put n times val bit
Definition: j2kenc.c:205
av_cold void ff_kbd_window_init(float *window, float alpha, int n)
Generate a Kaiser-Bessel Derived Window.
Definition: kbdwin.c:26
float pcoeffs[1024]
coefficients for IMDCT, pristine
Definition: aac.h:257
#define LIBAVUTIL_VERSION_INT
Definition: version.h:62
#define SCALE_DIFF_ZERO
codebook index corresponding to zero scalefactor indices difference
Definition: aac.h:152
uint8_t window_clipping[8]
set if a certain window is near clipping
Definition: aac.h:188
Definition: aac.h:63
const char * g
Definition: vf_curves.c:108
static av_cold int init(AVCodecContext *avctx)
Definition: avrndec.c:35
Definition: aac.h:57
av_cold void ff_psy_preprocess_end(struct FFPsyPreprocessContext *ctx)
Cleanup audio preprocessing module.
Definition: psymodel.c:150
#define WARN_IF(cond,...)
Definition: aacenc_utils.h:137
int size
Definition: avcodec.h:1434
const int ff_aac_swb_size_1024_len
Definition: aacenctab.c:108
AACCoefficientsEncoder * coder
Definition: aacenc.h:98
void avpriv_align_put_bits(PutBitContext *s)
Pad the bitstream with zeros up to the next byte boundary.
Definition: bitstream.c:48
static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
Encode ics_info element.
Definition: aacenc.c:160
int common_window
Set if channels share a common 'IndividualChannelStream' in bitstream.
Definition: aac.h:273
const uint8_t * ff_aac_swb_size_1024[]
Definition: aacenctab.c:99
#define AV_CODEC_CAP_EXPERIMENTAL
Codec is experimental and is thus avoided in favor of non experimental encoders.
Definition: avcodec.h:912
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:276
float lambda
Definition: aacenc.h:101
#define NOISE_PRE
preamble for NOISE_BT, put in bitstream with the first noise band
Definition: aac.h:154
int profile
profile
Definition: avcodec.h:3125
AVCodec.
Definition: avcodec.h:3482
void(* search_for_tns)(struct AACEncContext *s, SingleChannelElement *sce)
Definition: aacenc.h:69
static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
Encode spectral coefficients processed by psychoacoustic model.
Definition: aacenc.c:370
Spectral data are scaled white noise not coded in the bitstream.
Definition: aac.h:87
struct FFPsyContext::@78 bitres
int * num_bands
number of scalefactor bands for possible frame sizes
Definition: psymodel.h:85
const uint8_t ff_aac_num_swb_128[]
Definition: aactab.c:53
const uint16_t * swb_offset
table of offsets to the lowest spectral coefficient of a scalefactor band, sfb, for a particular wind...
Definition: aac.h:178
#define AV_CODEC_CAP_DELAY
Encoder or decoder requires flushing with NULL input at the end in order to give the complete and cor...
Definition: avcodec.h:882
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Definition: avassert.h:37
supported_samplerates
AACEncOptions options
encoding options
Definition: aacenc.h:82
AAC encoder context.
Definition: aacenc.h:80
uint8_t
#define av_cold
Definition: attributes.h:74
AVOptions.
int intensity_stereo
Definition: aacenc.h:50
#define WINDOW_FUNC(type)
Definition: aacenc.c:74
LPCContext lpc
used by TNS
Definition: aacenc.h:90
void ff_aac_coder_init_mips(AACEncContext *c)
SingleChannelElement ch[2]
Definition: aac.h:279
int samplerate_index
MPEG-4 samplerate index.
Definition: aacenc.h:91
Definition: aac.h:59
av_cold void ff_af_queue_init(AVCodecContext *avctx, AudioFrameQueue *afq)
Initialize AudioFrameQueue.
static av_cold int end(AVCodecContext *avctx)
Definition: avrndec.c:90
const uint8_t * chan_map
channel configuration map
Definition: aacenc.h:93
TemporalNoiseShaping tns
Definition: aac.h:247
#define FF_PROFILE_UNKNOWN
Definition: avcodec.h:3126
const uint8_t ff_aac_scalefactor_bits[121]
Definition: aactab.c:80
uint8_t * extradata
some codecs need / can use extradata like Huffman tables.
Definition: avcodec.h:1627
AudioFrameQueue afq
Definition: aacenc.h:102
static AVFrame * frame
static const uint8_t aac_chan_maps[AAC_MAX_CHANNELS][AAC_MAX_CHANNELS]
Table to remap channels from libavcodec's default order to AAC order.
Definition: aacenctab.h:59
uint8_t * data
Definition: avcodec.h:1433
const uint8_t * ff_aac_swb_size_128[]
Definition: aacenctab.c:91
uint32_t tag
Definition: movenc.c:1339
Scalefactor data are intensity stereo positions (in phase).
Definition: aac.h:89
int profile
copied from avctx
Definition: aacenc.h:89
int duration
Duration of this packet in AVStream->time_base units, 0 if unknown.
Definition: avcodec.h:1451
const OptionDef options[]
Definition: ffserver.c:3810
static void adjust_frame_information(ChannelElement *cpe, int chans)
Produce integer coefficients from scalefactors provided by the model.
Definition: aacenc.c:195
#define av_log(a,...)
static const AVOption aacenc_options[]
Definition: aacenc.c:882
float coeffs[1024]
coefficients for IMDCT, maybe processed
Definition: aac.h:258
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:176
#define FF_PROFILE_AAC_MAIN
Definition: avcodec.h:3129
av_default_item_name
static const int sizes[][2]
Definition: img2dec.c:48
#define AVERROR(e)
Definition: error.h:43
const uint8_t ff_aac_num_swb_1024[]
Definition: aactab.c:41
int last_frame
Definition: aacenc.h:100
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:178
float is_ener[128]
Intensity stereo pos (used by encoder)
Definition: aac.h:255
int stereo_mode
Definition: aacenc.h:45
int initial_padding
Audio only.
Definition: avcodec.h:3304
float ff_aac_kbd_long_1024[1024]
Definition: aactab.c:36
int flags
AV_CODEC_FLAG_*.
Definition: avcodec.h:1607
static const int mpeg4audio_sample_rates[16]
Definition: aacenctab.h:70
int amp[4]
Definition: aac.h:225
const char * name
Name of the codec implementation.
Definition: avcodec.h:3489
int num_windows
number of windows in a frame
Definition: psymodel.h:67
static void copy_input_samples(AACEncContext *s, const AVFrame *frame)
Definition: aacenc.c:463
uint8_t max_sfb
number of scalefactor bands per group
Definition: aac.h:172
#define ff_mdct_init
Definition: fft.h:167
Definition: aac.h:62
void(* set_special_band_scalefactors)(struct AACEncContext *s, SingleChannelElement *sce)
Definition: aacenc.h:67
int num_swb
number of scalefactor window bands
Definition: aac.h:180
int ff_af_queue_add(AudioFrameQueue *afq, const AVFrame *f)
Add a frame to the queue.
#define FFMAX(a, b)
Definition: common.h:90
Libavcodec external API header.
#define fail()
Definition: checkasm.h:57
void(* search_for_quantizers)(AVCodecContext *avctx, struct AACEncContext *s, SingleChannelElement *sce, const float lambda)
Definition: aacenc.h:56
static int put_bits_count(PutBitContext *s)
Definition: put_bits.h:85
#define AACENC_FLAGS
Definition: aacenc.c:881
void(* search_for_is)(struct AACEncContext *s, AVCodecContext *avctx, ChannelElement *cpe)
Definition: aacenc.h:71
int bit_rate
the average bitrate
Definition: avcodec.h:1577
enum WindowSequence window_sequence[2]
Definition: aac.h:173
#define AV_CODEC_FLAG_BITEXACT
Use only bitexact stuff (except (I)DCT).
Definition: avcodec.h:788
av_cold void ff_lpc_end(LPCContext *s)
Uninitialize LPCContext.
Definition: lpc.c:316
#define AV_CODEC_FLAG_QSCALE
Use fixed qscale.
Definition: avcodec.h:735
#define AV_CODEC_CAP_SMALL_LAST_FRAME
Codec can be fed a final frame with a smaller size.
Definition: avcodec.h:887
int cur_channel
Definition: aacenc.h:99
#define FFMIN(a, b)
Definition: common.h:92
static void apply_intensity_stereo(ChannelElement *cpe)
Definition: aacenc.c:245
static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, const AVFrame *frame, int *got_packet_ptr)
Definition: aacenc.c:485
void(* apply_tns_filt)(struct AACEncContext *s, SingleChannelElement *sce)
Definition: aacenc.h:66
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:124
int pos[4]
Definition: aac.h:224
void(* quantize_and_encode_band)(struct AACEncContext *s, PutBitContext *pb, const float *in, float *out, int size, int scale_idx, int cb, const float lambda, int rtz)
Definition: aacenc.h:60
int channels
channel count
Definition: aacenc.h:92
void(* encode_main_pred)(struct AACEncContext *s, SingleChannelElement *sce)
Definition: aacenc.h:63
AAC definitions and structures.
void(* adjust_common_prediction)(struct AACEncContext *s, ChannelElement *cpe)
Definition: aacenc.h:64
const uint8_t ff_tns_max_bands_1024[]
Definition: aactab.c:1272
FFTContext mdct128
short (128 samples) frame transform context
Definition: aacenc.h:85
PutBitContext pb
Definition: aacenc.h:83
static void(*const apply_window[4])(AVFloatDSPContext *fdsp, SingleChannelElement *sce, const float *audio)
Definition: aacenc.c:130
void(* search_for_pred)(struct AACEncContext *s, SingleChannelElement *sce)
Definition: aacenc.h:72
#define L(x)
Definition: vp56_arith.h:36
AVFloatDSPContext * fdsp
Definition: aacenc.h:86
static av_cold int aac_encode_end(AVCodecContext *avctx)
Definition: aacenc.c:741
int frame_size
Number of samples per channel in an audio frame.
Definition: avcodec.h:2292
#define CLIP_AVOIDANCE_FACTOR
Definition: aac.h:53
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:59
void(* search_for_ms)(struct AACEncContext *s, ChannelElement *cpe)
Definition: aacenc.h:70
static void put_audio_specific_config(AVCodecContext *avctx)
Make AAC audio config object.
Definition: aacenc.c:53
Temporal Noise Shaping.
Definition: aac.h:195
int sample_rate
samples per second
Definition: avcodec.h:2272
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:181
FFPsyWindowInfo(* window)(FFPsyContext *ctx, const float *audio, const float *la, int channel, int prev_type)
Suggest window sequence for channel.
Definition: psymodel.h:114
int frame_bits
number of bits used for the previously encoded frame
Definition: avcodec.h:2774
void(* apply_main_pred)(struct AACEncContext *s, SingleChannelElement *sce)
Definition: aacenc.h:65
main external API structure.
Definition: avcodec.h:1512
static void(WINAPI *cond_broadcast)(pthread_cond_t *cond)
int bits
number of bits used in the bitresevoir
Definition: psymodel.h:90
#define NOISE_PRE_BITS
length of preamble
Definition: aac.h:155
#define FF_PROFILE_AAC_LOW
Definition: avcodec.h:3130
Levinson-Durbin recursion.
Definition: lpc.h:47
IndividualChannelStream ics
Definition: aac.h:246
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> in
void(* encode_window_bands_info)(struct AACEncContext *s, SingleChannelElement *sce, int win, int group_len, const float lambda)
Definition: aacenc.h:58
int extradata_size
Definition: avcodec.h:1628
uint8_t group_len[8]
Definition: aac.h:176
Describe the class of an AVClass context structure.
Definition: log.h:67
static void put_bitstream_info(AACEncContext *s, const char *name)
Write some auxiliary information about the created AAC file.
Definition: aacenc.c:442
const int ff_aac_swb_size_128_len
Definition: aacenctab.c:107
int window_shape
window shape (sine/KBD/whatever)
Definition: psymodel.h:66
static void encode_pulses(AACEncContext *s, Pulse *pulse)
Encode pulse data.
Definition: aacenc.c:351
const uint8_t * swb_sizes
table of scalefactor band sizes for a particular window
Definition: aac.h:179
#define TNS_MAX_ORDER
Definition: aac.h:50
av_cold AVFloatDSPContext * avpriv_float_dsp_alloc(int bit_exact)
Allocate a float DSP context.
Definition: float_dsp.c:143
int ff_alloc_packet2(AVCodecContext *avctx, AVPacket *avpkt, int64_t size, int64_t min_size)
Check AVPacket size and/or allocate data.
Definition: utils.c:1804
FFPsyContext psy
Definition: aacenc.h:96
const uint32_t ff_aac_scalefactor_code[121]
Definition: aactab.c:61
static av_cold int alloc_buffers(AVCodecContext *avctx, AACEncContext *s)
Definition: aacenc.c:780
const struct FFPsyModel * model
encoder-specific model functions
Definition: psymodel.h:78
av_cold int ff_lpc_init(LPCContext *s, int blocksize, int max_order, enum FFLPCType lpc_type)
Initialize LPCContext.
Definition: lpc.c:294
#define AAC_MAX_CHANNELS
Definition: aacenctab.h:39
struct AACEncContext::@28 buffer
int ms_mode
Signals mid/side stereo flags coding mode (used by encoder)
Definition: aac.h:274
AAC encoder data.
const uint8_t ff_tns_max_bands_128[]
Definition: aactab.c:1284
struct FFPsyPreprocessContext * psypp
Definition: aacenc.h:97
#define NOISE_OFFSET
subtracted from global gain, used as offset for the preamble
Definition: aac.h:156
void(* encode_tns_info)(struct AACEncContext *s, SingleChannelElement *sce)
Definition: aacenc.h:62
int global_quality
Global quality for codecs which cannot change it per frame.
Definition: avcodec.h:1593
uint8_t zeroes[128]
band is not coded (used by encoder)
Definition: aac.h:254
int sf_idx[128]
scalefactor indices (used by encoder)
Definition: aac.h:253
AVCodec ff_aac_encoder
Definition: aacenc.c:914
uint8_t is_mode
Set if any bands have been encoded using intensity stereo (used by encoder)
Definition: aac.h:275
const int avpriv_mpeg4audio_sample_rates[16]
Definition: mpeg4audio.c:57
int aac_coder
Definition: aacenc.h:46
Scalefactor data are intensity stereo positions (out of phase).
Definition: aac.h:88
av_cold void ff_aac_tableinit(void)
Definition: aac_tablegen.h:35
Y Spectral Band Replication.
Definition: mpeg4audio.h:65
float * samples
Definition: aacenc.h:107
uint8_t prediction_used[41]
Definition: aac.h:187
static av_cold int aac_encode_init(AVCodecContext *avctx)
Definition: aacenc.c:795
common internal api header.
static void flush_put_bits(PutBitContext *s)
Pad the end of the output stream with zeros.
Definition: put_bits.h:101
if(ret< 0)
Definition: vf_mcdeint.c:280
AAC encoder utilities.
Single Channel Element - used for both SCE and LFE elements.
Definition: aac.h:245
windowing related information
Definition: psymodel.h:64
#define ff_mdct_end
Definition: fft.h:168
av_cold struct FFPsyPreprocessContext * ff_psy_preprocess_init(AVCodecContext *avctx)
psychoacoustic model audio preprocessing initialization
Definition: psymodel.c:102
const uint16_t *const ff_swb_offset_1024[]
Definition: aactab.c:1230
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:137
static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce)
Encode scalefactors.
Definition: aacenc.c:315
ChannelElement * cpe
channel elements
Definition: aacenc.h:95
Individual Channel Stream.
Definition: aac.h:171
float clip_avoidance_factor
set if any window is near clipping to the necessary atennuation factor to avoid it ...
Definition: aac.h:189
static void init_put_bits(PutBitContext *s, uint8_t *buffer, int buffer_size)
Initialize the PutBitContext s.
Definition: put_bits.h:48
#define ERROR_IF(cond,...)
Definition: aacenc_utils.h:131
#define AV_INPUT_BUFFER_PADDING_SIZE
Required number of additionally allocated bytes at the end of the input bitstream for decoding...
Definition: avcodec.h:636
channel element - generic struct for SCE/CPE/CCE/LFE
Definition: aac.h:270
void * priv_data
Definition: avcodec.h:1554
int start
Definition: aac.h:223
FFTContext mdct1024
long (1024 samples) frame transform context
Definition: aacenc.h:84
static av_always_inline int diff(const uint32_t a, const uint32_t b)
static const int16_t coeffs[]
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:2273
int num_pulse
Definition: aac.h:222
static const uint8_t aac_chan_configs[6][5]
default channel configurations
Definition: aacenctab.h:47
static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
Encode scalefactor band coding type.
Definition: aacenc.c:301
static void apply_mid_side_stereo(ChannelElement *cpe)
Definition: aacenc.c:272
#define HAVE_MIPSDSPR1
Definition: config.h:76
void ff_af_queue_close(AudioFrameQueue *afq)
Close AudioFrameQueue.
enum BandType band_type[128]
band types
Definition: aac.h:249
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> out
#define LIBAVCODEC_IDENT
Definition: version.h:43
static enum AVSampleFormat sample_fmts[]
Definition: adpcmenc.c:701
int frame_number
Frame counter, set by libavcodec.
Definition: avcodec.h:2303
float ret_buf[2048]
PCM output buffer.
Definition: aac.h:260
#define av_freep(p)
void INT64 start
Definition: avisynth_c.h:553
Definition: vf_geq.c:46
static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s, SingleChannelElement *sce, int common_window)
Encode one channel of audio data.
Definition: aacenc.c:418
static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce, float *audio)
Definition: aacenc.c:139
const uint16_t *const ff_swb_offset_128[]
Definition: aactab.c:1254
uint8_t is_mask[128]
Set if intensity stereo is used (used by encoder)
Definition: aac.h:277
uint8_t ** extended_data
pointers to the data planes/channels.
Definition: frame.h:215
AAC data declarations.
av_cold void ff_psy_end(FFPsyContext *ctx)
Cleanup model context at the end.
Definition: psymodel.c:82
This structure stores compressed data.
Definition: avcodec.h:1410
static void avoid_clipping(AACEncContext *s, SingleChannelElement *sce)
Downscale spectral coefficients for near-clipping windows to avoid artifacts.
Definition: aacenc.c:398
int window_type[3]
window type (short/long/transitional, etc.) - current, previous and next
Definition: psymodel.h:65
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:225
static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
Definition: aacenc.c:758
int64_t pts
Presentation timestamp in AVStream->time_base units; the time at which the decompressed packet will b...
Definition: avcodec.h:1426
for(j=16;j >0;--j)
int pred
Definition: aacenc.h:49
#define FF_ALLOCZ_OR_GOTO(ctx, p, size, label)
Definition: internal.h:141
void AAC_RENAME() ff_init_ff_sine_windows(int index)
initialize the specified entry of ff_sine_windows
float * planar_samples[6]
saved preprocessed input
Definition: aacenc.h:87
const char * name
Definition: opengl_enc.c:103
bitstream writer API