c2f6dcf8a56ac3ada99fb22f79731e0ee97fb34c
[libav.git] / libavcodec / aacenc.c
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 libavcodec/aacenc.c
24 * AAC encoder
25 */
26
27 /***********************************
28 * TODOs:
29 * add sane pulse detection
30 * add temporal noise shaping
31 ***********************************/
32
33 #include "avcodec.h"
34 #include "put_bits.h"
35 #include "dsputil.h"
36 #include "mpeg4audio.h"
37
38 #include "aac.h"
39 #include "aactab.h"
40 #include "aacenc.h"
41
42 #include "psymodel.h"
43
44 static const uint8_t swb_size_1024_96[] = {
45 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8,
46 12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44,
47 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
48 };
49
50 static const uint8_t swb_size_1024_64[] = {
51 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8,
52 12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36,
53 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40
54 };
55
56 static const uint8_t swb_size_1024_48[] = {
57 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
58 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
59 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32,
60 96
61 };
62
63 static const uint8_t swb_size_1024_32[] = {
64 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
65 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
66 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
67 };
68
69 static const uint8_t swb_size_1024_24[] = {
70 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
71 12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28,
72 32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64
73 };
74
75 static const uint8_t swb_size_1024_16[] = {
76 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
77 12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28,
78 32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64
79 };
80
81 static const uint8_t swb_size_1024_8[] = {
82 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
83 16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28,
84 32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80
85 };
86
87 static const uint8_t *swb_size_1024[] = {
88 swb_size_1024_96, swb_size_1024_96, swb_size_1024_64,
89 swb_size_1024_48, swb_size_1024_48, swb_size_1024_32,
90 swb_size_1024_24, swb_size_1024_24, swb_size_1024_16,
91 swb_size_1024_16, swb_size_1024_16, swb_size_1024_8
92 };
93
94 static const uint8_t swb_size_128_96[] = {
95 4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36
96 };
97
98 static const uint8_t swb_size_128_48[] = {
99 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16
100 };
101
102 static const uint8_t swb_size_128_24[] = {
103 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20
104 };
105
106 static const uint8_t swb_size_128_16[] = {
107 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20
108 };
109
110 static const uint8_t swb_size_128_8[] = {
111 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20
112 };
113
114 static const uint8_t *swb_size_128[] = {
115 /* the last entry on the following row is swb_size_128_64 but is a
116 duplicate of swb_size_128_96 */
117 swb_size_128_96, swb_size_128_96, swb_size_128_96,
118 swb_size_128_48, swb_size_128_48, swb_size_128_48,
119 swb_size_128_24, swb_size_128_24, swb_size_128_16,
120 swb_size_128_16, swb_size_128_16, swb_size_128_8
121 };
122
123 /** default channel configurations */
124 static const uint8_t aac_chan_configs[6][5] = {
125 {1, TYPE_SCE}, // 1 channel - single channel element
126 {1, TYPE_CPE}, // 2 channels - channel pair
127 {2, TYPE_SCE, TYPE_CPE}, // 3 channels - center + stereo
128 {3, TYPE_SCE, TYPE_CPE, TYPE_SCE}, // 4 channels - front center + stereo + back center
129 {3, TYPE_SCE, TYPE_CPE, TYPE_CPE}, // 5 channels - front center + stereo + back stereo
130 {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE
131 };
132
133 /**
134 * Make AAC audio config object.
135 * @see 1.6.2.1 "Syntax - AudioSpecificConfig"
136 */
137 static void put_audio_specific_config(AVCodecContext *avctx)
138 {
139 PutBitContext pb;
140 AACEncContext *s = avctx->priv_data;
141
142 init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8);
143 put_bits(&pb, 5, 2); //object type - AAC-LC
144 put_bits(&pb, 4, s->samplerate_index); //sample rate index
145 put_bits(&pb, 4, avctx->channels);
146 //GASpecificConfig
147 put_bits(&pb, 1, 0); //frame length - 1024 samples
148 put_bits(&pb, 1, 0); //does not depend on core coder
149 put_bits(&pb, 1, 0); //is not extension
150 flush_put_bits(&pb);
151 }
152
153 static av_cold int aac_encode_init(AVCodecContext *avctx)
154 {
155 AACEncContext *s = avctx->priv_data;
156 int i;
157 const uint8_t *sizes[2];
158 int lengths[2];
159
160 avctx->frame_size = 1024;
161
162 for (i = 0; i < 16; i++)
163 if (avctx->sample_rate == ff_mpeg4audio_sample_rates[i])
164 break;
165 if (i == 16) {
166 av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate);
167 return -1;
168 }
169 if (avctx->channels > 6) {
170 av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels);
171 return -1;
172 }
173 s->samplerate_index = i;
174
175 dsputil_init(&s->dsp, avctx);
176 ff_mdct_init(&s->mdct1024, 11, 0, 1.0);
177 ff_mdct_init(&s->mdct128, 8, 0, 1.0);
178 // window init
179 ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
180 ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
181 ff_sine_window_init(ff_sine_1024, 1024);
182 ff_sine_window_init(ff_sine_128, 128);
183
184 s->samples = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0]));
185 s->cpe = av_mallocz(sizeof(ChannelElement) * aac_chan_configs[avctx->channels-1][0]);
186 avctx->extradata = av_malloc(2);
187 avctx->extradata_size = 2;
188 put_audio_specific_config(avctx);
189
190 sizes[0] = swb_size_1024[i];
191 sizes[1] = swb_size_128[i];
192 lengths[0] = ff_aac_num_swb_1024[i];
193 lengths[1] = ff_aac_num_swb_128[i];
194 ff_psy_init(&s->psy, avctx, 2, sizes, lengths);
195 s->psypp = ff_psy_preprocess_init(avctx);
196 s->coder = &ff_aac_coders[0];
197
198 s->lambda = avctx->global_quality ? avctx->global_quality : 120;
199 #if !CONFIG_HARDCODED_TABLES
200 for (i = 0; i < 428; i++)
201 ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.);
202 #endif /* CONFIG_HARDCODED_TABLES */
203
204 if (avctx->channels > 5)
205 av_log(avctx, AV_LOG_ERROR, "This encoder does not yet enforce the restrictions on LFEs. "
206 "The output will most likely be an illegal bitstream.\n");
207
208 return 0;
209 }
210
211 static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s,
212 SingleChannelElement *sce, short *audio, int channel)
213 {
214 int i, j, k;
215 const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
216 const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
217 const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
218
219 if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
220 memcpy(s->output, sce->saved, sizeof(float)*1024);
221 if (sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE) {
222 memset(s->output, 0, sizeof(s->output[0]) * 448);
223 for (i = 448; i < 576; i++)
224 s->output[i] = sce->saved[i] * pwindow[i - 448];
225 for (i = 576; i < 704; i++)
226 s->output[i] = sce->saved[i];
227 }
228 if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) {
229 j = channel;
230 for (i = 0; i < 1024; i++, j += avctx->channels) {
231 s->output[i+1024] = audio[j] * lwindow[1024 - i - 1];
232 sce->saved[i] = audio[j] * lwindow[i];
233 }
234 } else {
235 j = channel;
236 for (i = 0; i < 448; i++, j += avctx->channels)
237 s->output[i+1024] = audio[j];
238 for (i = 448; i < 576; i++, j += avctx->channels)
239 s->output[i+1024] = audio[j] * swindow[576 - i - 1];
240 memset(s->output+1024+576, 0, sizeof(s->output[0]) * 448);
241 j = channel;
242 for (i = 0; i < 1024; i++, j += avctx->channels)
243 sce->saved[i] = audio[j];
244 }
245 ff_mdct_calc(&s->mdct1024, sce->coeffs, s->output);
246 } else {
247 j = channel;
248 for (k = 0; k < 1024; k += 128) {
249 for (i = 448 + k; i < 448 + k + 256; i++)
250 s->output[i - 448 - k] = (i < 1024)
251 ? sce->saved[i]
252 : audio[channel + (i-1024)*avctx->channels];
253 s->dsp.vector_fmul (s->output, k ? swindow : pwindow, 128);
254 s->dsp.vector_fmul_reverse(s->output+128, s->output+128, swindow, 128);
255 ff_mdct_calc(&s->mdct128, sce->coeffs + k, s->output);
256 }
257 j = channel;
258 for (i = 0; i < 1024; i++, j += avctx->channels)
259 sce->saved[i] = audio[j];
260 }
261 }
262
263 /**
264 * Encode ics_info element.
265 * @see Table 4.6 (syntax of ics_info)
266 */
267 static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
268 {
269 int w;
270
271 put_bits(&s->pb, 1, 0); // ics_reserved bit
272 put_bits(&s->pb, 2, info->window_sequence[0]);
273 put_bits(&s->pb, 1, info->use_kb_window[0]);
274 if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
275 put_bits(&s->pb, 6, info->max_sfb);
276 put_bits(&s->pb, 1, 0); // no prediction
277 } else {
278 put_bits(&s->pb, 4, info->max_sfb);
279 for (w = 1; w < 8; w++) {
280 put_bits(&s->pb, 1, !info->group_len[w]);
281 }
282 }
283 }
284
285 /**
286 * Encode MS data.
287 * @see 4.6.8.1 "Joint Coding - M/S Stereo"
288 */
289 static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
290 {
291 int i, w;
292
293 put_bits(pb, 2, cpe->ms_mode);
294 if (cpe->ms_mode == 1) {
295 for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w]) {
296 for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
297 put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
298 }
299 }
300 }
301
302 /**
303 * Produce integer coefficients from scalefactors provided by the model.
304 */
305 static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans)
306 {
307 int i, w, w2, g, ch;
308 int start, sum, 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 sum = 0;
318 //apply M/S
319 if (!ch && cpe->ms_mask[w + g]) {
320 for (i = 0; i < ics->swb_sizes[g]; i++) {
321 cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
322 cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
323 }
324 }
325 start += ics->swb_sizes[g];
326 }
327 for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--);
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]) msc++;
356 if (msc == 0 || ics0->max_sfb == 0)
357 cpe->ms_mode = 0;
358 else
359 cpe->ms_mode = msc < ics0->max_sfb ? 1 : 2;
360 }
361 }
362
363 /**
364 * Encode scalefactor band coding type.
365 */
366 static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
367 {
368 int w;
369
370 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
371 s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
372 }
373 }
374
375 /**
376 * Encode scalefactors.
377 */
378 static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s,
379 SingleChannelElement *sce)
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];
391 put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
392 }
393 }
394 }
395 }
396
397 /**
398 * Encode pulse data.
399 */
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
416 /**
417 * Encode spectral coefficients processed by psychoacoustic model.
418 */
419 static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
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 }
437 start += sce->ics.swb_sizes[i];
438 }
439 }
440 }
441
442 /**
443 * Encode one channel of audio data.
444 */
445 static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
446 SingleChannelElement *sce,
447 int common_window)
448 {
449 put_bits(&s->pb, 8, sce->sf_idx[0]);
450 if (!common_window)
451 put_ics_info(s, &sce->ics);
452 encode_band_info(s, sce);
453 encode_scale_factors(avctx, s, sce);
454 encode_pulses(s, &sce->pulse);
455 put_bits(&s->pb, 1, 0); //tns
456 put_bits(&s->pb, 1, 0); //ssr
457 encode_spectral_coeffs(s, sce);
458 return 0;
459 }
460
461 /**
462 * Write some auxiliary information about the created AAC file.
463 */
464 static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s,
465 const char *name)
466 {
467 int i, namelen, padbits;
468
469 namelen = strlen(name) + 2;
470 put_bits(&s->pb, 3, TYPE_FIL);
471 put_bits(&s->pb, 4, FFMIN(namelen, 15));
472 if (namelen >= 15)
473 put_bits(&s->pb, 8, namelen - 16);
474 put_bits(&s->pb, 4, 0); //extension type - filler
475 padbits = 8 - (put_bits_count(&s->pb) & 7);
476 align_put_bits(&s->pb);
477 for (i = 0; i < namelen - 2; i++)
478 put_bits(&s->pb, 8, name[i]);
479 put_bits(&s->pb, 12 - padbits, 0);
480 }
481
482 static int aac_encode_frame(AVCodecContext *avctx,
483 uint8_t *frame, int buf_size, void *data)
484 {
485 AACEncContext *s = avctx->priv_data;
486 int16_t *samples = s->samples, *samples2, *la;
487 ChannelElement *cpe;
488 int i, j, chans, tag, start_ch;
489 const uint8_t *chan_map = aac_chan_configs[avctx->channels-1];
490 int chan_el_counter[4];
491
492 if (s->last_frame)
493 return 0;
494 if (data) {
495 if (!s->psypp) {
496 memcpy(s->samples + 1024 * avctx->channels, data,
497 1024 * avctx->channels * sizeof(s->samples[0]));
498 } else {
499 start_ch = 0;
500 samples2 = s->samples + 1024 * avctx->channels;
501 for (i = 0; i < chan_map[0]; i++) {
502 tag = chan_map[i+1];
503 chans = tag == TYPE_CPE ? 2 : 1;
504 ff_psy_preprocess(s->psypp, (uint16_t*)data + start_ch,
505 samples2 + start_ch, start_ch, chans);
506 start_ch += chans;
507 }
508 }
509 }
510 if (!avctx->frame_number) {
511 memcpy(s->samples, s->samples + 1024 * avctx->channels,
512 1024 * avctx->channels * sizeof(s->samples[0]));
513 return 0;
514 }
515
516 init_put_bits(&s->pb, frame, buf_size*8);
517 if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT)) {
518 put_bitstream_info(avctx, s, LIBAVCODEC_IDENT);
519 }
520 start_ch = 0;
521 memset(chan_el_counter, 0, sizeof(chan_el_counter));
522 for (i = 0; i < chan_map[0]; i++) {
523 FFPsyWindowInfo wi[2];
524 tag = chan_map[i+1];
525 chans = tag == TYPE_CPE ? 2 : 1;
526 cpe = &s->cpe[i];
527 samples2 = samples + start_ch;
528 la = samples2 + 1024 * avctx->channels + start_ch;
529 if (!data) la = NULL;
530 for (j = 0; j < chans; j++) {
531 IndividualChannelStream *ics = &cpe->ch[j].ics;
532 int k;
533 wi[j] = ff_psy_suggest_window(&s->psy, samples2, la, start_ch + j, ics->window_sequence[0]);
534 ics->window_sequence[1] = ics->window_sequence[0];
535 ics->window_sequence[0] = wi[j].window_type[0];
536 ics->use_kb_window[1] = ics->use_kb_window[0];
537 ics->use_kb_window[0] = wi[j].window_shape;
538 ics->num_windows = wi[j].num_windows;
539 ics->swb_sizes = s->psy.bands [ics->num_windows == 8];
540 ics->num_swb = s->psy.num_bands[ics->num_windows == 8];
541 for (k = 0; k < ics->num_windows; k++)
542 ics->group_len[k] = wi[j].grouping[k];
543
544 s->cur_channel = start_ch + j;
545 apply_window_and_mdct(avctx, s, &cpe->ch[j], samples2, j);
546 s->coder->search_for_quantizers(avctx, s, &cpe->ch[j], s->lambda);
547 }
548 cpe->common_window = 0;
549 if (chans > 1
550 && wi[0].window_type[0] == wi[1].window_type[0]
551 && wi[0].window_shape == wi[1].window_shape) {
552
553 cpe->common_window = 1;
554 for (j = 0; j < wi[0].num_windows; j++) {
555 if (wi[0].grouping[j] != wi[1].grouping[j]) {
556 cpe->common_window = 0;
557 break;
558 }
559 }
560 }
561 if (cpe->common_window && s->coder->search_for_ms)
562 s->coder->search_for_ms(s, cpe, s->lambda);
563 adjust_frame_information(s, cpe, chans);
564 put_bits(&s->pb, 3, tag);
565 put_bits(&s->pb, 4, chan_el_counter[tag]++);
566 if (chans == 2) {
567 put_bits(&s->pb, 1, cpe->common_window);
568 if (cpe->common_window) {
569 put_ics_info(s, &cpe->ch[0].ics);
570 encode_ms_info(&s->pb, cpe);
571 }
572 }
573 for (j = 0; j < chans; j++) {
574 s->cur_channel = start_ch + j;
575 ff_psy_set_band_info(&s->psy, s->cur_channel, cpe->ch[j].coeffs, &wi[j]);
576 encode_individual_channel(avctx, s, &cpe->ch[j], cpe->common_window);
577 }
578 start_ch += chans;
579 }
580
581 put_bits(&s->pb, 3, TYPE_END);
582 flush_put_bits(&s->pb);
583 avctx->frame_bits = put_bits_count(&s->pb);
584
585 // rate control stuff
586 if (!(avctx->flags & CODEC_FLAG_QSCALE)) {
587 float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
588 s->lambda *= ratio;
589 }
590
591 if (avctx->frame_bits > 6144*avctx->channels) {
592 av_log(avctx, AV_LOG_ERROR, "input buffer violation %d > %d.\n",
593 avctx->frame_bits, 6144*avctx->channels);
594 }
595
596 if (!data)
597 s->last_frame = 1;
598 memcpy(s->samples, s->samples + 1024 * avctx->channels,
599 1024 * avctx->channels * sizeof(s->samples[0]));
600 return put_bits_count(&s->pb)>>3;
601 }
602
603 static av_cold int aac_encode_end(AVCodecContext *avctx)
604 {
605 AACEncContext *s = avctx->priv_data;
606
607 ff_mdct_end(&s->mdct1024);
608 ff_mdct_end(&s->mdct128);
609 ff_psy_end(&s->psy);
610 ff_psy_preprocess_end(s->psypp);
611 av_freep(&s->samples);
612 av_freep(&s->cpe);
613 return 0;
614 }
615
616 AVCodec aac_encoder = {
617 "aac",
618 CODEC_TYPE_AUDIO,
619 CODEC_ID_AAC,
620 sizeof(AACEncContext),
621 aac_encode_init,
622 aac_encode_frame,
623 aac_encode_end,
624 .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY,
625 .sample_fmts = (enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE},
626 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
627 };