3 * Copyright (C) 2008 Konstantin Shishkov
5 * This file is part of FFmpeg.
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.
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.
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
27 /***********************************
29 * add sane pulse detection
30 * add temporal noise shaping
31 ***********************************/
33 #include "libavutil/opt.h"
37 #include "mpeg4audio.h"
47 #define AAC_MAX_CHANNELS 6
49 static const uint8_t swb_size_1024_96[] = {
50 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8,
51 12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44,
52 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
55 static const uint8_t swb_size_1024_64[] = {
56 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8,
57 12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36,
58 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40
61 static const uint8_t swb_size_1024_48[] = {
62 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
63 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
64 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32,
68 static const uint8_t swb_size_1024_32[] = {
69 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
70 12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
71 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
74 static const uint8_t swb_size_1024_24[] = {
75 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
76 12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28,
77 32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64
80 static const uint8_t swb_size_1024_16[] = {
81 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
82 12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28,
83 32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64
86 static const uint8_t swb_size_1024_8[] = {
87 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
88 16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28,
89 32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80
92 static const uint8_t *swb_size_1024[] = {
93 swb_size_1024_96, swb_size_1024_96, swb_size_1024_64,
94 swb_size_1024_48, swb_size_1024_48, swb_size_1024_32,
95 swb_size_1024_24, swb_size_1024_24, swb_size_1024_16,
96 swb_size_1024_16, swb_size_1024_16, swb_size_1024_8
99 static const uint8_t swb_size_128_96[] = {
100 4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36
103 static const uint8_t swb_size_128_48[] = {
104 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16
107 static const uint8_t swb_size_128_24[] = {
108 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20
111 static const uint8_t swb_size_128_16[] = {
112 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20
115 static const uint8_t swb_size_128_8[] = {
116 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20
119 static const uint8_t *swb_size_128[] = {
120 /* the last entry on the following row is swb_size_128_64 but is a
121 duplicate of swb_size_128_96 */
122 swb_size_128_96, swb_size_128_96, swb_size_128_96,
123 swb_size_128_48, swb_size_128_48, swb_size_128_48,
124 swb_size_128_24, swb_size_128_24, swb_size_128_16,
125 swb_size_128_16, swb_size_128_16, swb_size_128_8
128 /** default channel configurations */
129 static const uint8_t aac_chan_configs[6][5] = {
130 {1, TYPE_SCE}, // 1 channel - single channel element
131 {1, TYPE_CPE}, // 2 channels - channel pair
132 {2, TYPE_SCE, TYPE_CPE}, // 3 channels - center + stereo
133 {3, TYPE_SCE, TYPE_CPE, TYPE_SCE}, // 4 channels - front center + stereo + back center
134 {3, TYPE_SCE, TYPE_CPE, TYPE_CPE}, // 5 channels - front center + stereo + back stereo
135 {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE
138 static const uint8_t channel_maps[][AAC_MAX_CHANNELS] = {
144 { 2, 0, 1, 4, 5, 3 },
148 * Make AAC audio config object.
149 * @see 1.6.2.1 "Syntax - AudioSpecificConfig"
151 static void put_audio_specific_config(AVCodecContext *avctx)
154 AACEncContext *s = avctx->priv_data;
156 init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8);
157 put_bits(&pb, 5, 2); //object type - AAC-LC
158 put_bits(&pb, 4, s->samplerate_index); //sample rate index
159 put_bits(&pb, 4, avctx->channels);
161 put_bits(&pb, 1, 0); //frame length - 1024 samples
162 put_bits(&pb, 1, 0); //does not depend on core coder
163 put_bits(&pb, 1, 0); //is not extension
165 //Explicitly Mark SBR absent
166 put_bits(&pb, 11, 0x2b7); //sync extension
167 put_bits(&pb, 5, AOT_SBR);
172 static av_cold int aac_encode_init(AVCodecContext *avctx)
174 AACEncContext *s = avctx->priv_data;
176 const uint8_t *sizes[2];
177 uint8_t grouping[AAC_MAX_CHANNELS];
180 avctx->frame_size = 1024;
182 for (i = 0; i < 16; i++)
183 if (avctx->sample_rate == avpriv_mpeg4audio_sample_rates[i])
186 av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate);
189 if (avctx->channels > AAC_MAX_CHANNELS) {
190 av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels);
193 if (avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW) {
194 av_log(avctx, AV_LOG_ERROR, "Unsupported profile %d\n", avctx->profile);
197 if (1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * avctx->channels) {
198 av_log(avctx, AV_LOG_ERROR, "Too many bits per frame requested\n");
201 s->samplerate_index = i;
203 dsputil_init(&s->dsp, avctx);
204 ff_mdct_init(&s->mdct1024, 11, 0, 1.0);
205 ff_mdct_init(&s->mdct128, 8, 0, 1.0);
207 ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
208 ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
209 ff_init_ff_sine_windows(10);
210 ff_init_ff_sine_windows(7);
212 s->chan_map = aac_chan_configs[avctx->channels-1];
213 s->samples = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0]));
214 s->cpe = av_mallocz(sizeof(ChannelElement) * s->chan_map[0]);
215 avctx->extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
216 avctx->extradata_size = 5;
217 put_audio_specific_config(avctx);
219 sizes[0] = swb_size_1024[i];
220 sizes[1] = swb_size_128[i];
221 lengths[0] = ff_aac_num_swb_1024[i];
222 lengths[1] = ff_aac_num_swb_128[i];
223 for (i = 0; i < s->chan_map[0]; i++)
224 grouping[i] = s->chan_map[i + 1] == TYPE_CPE;
225 ff_psy_init(&s->psy, avctx, 2, sizes, lengths, s->chan_map[0], grouping);
226 s->psypp = ff_psy_preprocess_init(avctx);
227 s->coder = &ff_aac_coders[s->options.aac_coder];
229 s->lambda = avctx->global_quality ? avctx->global_quality : 120;
236 static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s,
237 SingleChannelElement *sce, short *audio)
240 const int chans = avctx->channels;
241 const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
242 const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
243 const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
244 float *output = sce->ret;
246 if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
247 memcpy(output, sce->saved, sizeof(float)*1024);
248 if (sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE) {
249 memset(output, 0, sizeof(output[0]) * 448);
250 for (i = 448; i < 576; i++)
251 output[i] = sce->saved[i] * pwindow[i - 448];
252 for (i = 576; i < 704; i++)
253 output[i] = sce->saved[i];
255 if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) {
256 for (i = 0; i < 1024; i++) {
257 output[i+1024] = audio[i * chans] * lwindow[1024 - i - 1];
258 sce->saved[i] = audio[i * chans] * lwindow[i];
261 for (i = 0; i < 448; i++)
262 output[i+1024] = audio[i * chans];
264 output[i+1024] = audio[i * chans] * swindow[576 - i - 1];
265 memset(output+1024+576, 0, sizeof(output[0]) * 448);
266 for (i = 0; i < 1024; i++)
267 sce->saved[i] = audio[i * chans];
269 s->mdct1024.mdct_calc(&s->mdct1024, sce->coeffs, output);
271 for (k = 0; k < 1024; k += 128) {
272 for (i = 448 + k; i < 448 + k + 256; i++)
273 output[i - 448 - k] = (i < 1024)
275 : audio[(i-1024)*chans];
276 s->dsp.vector_fmul (output, output, k ? swindow : pwindow, 128);
277 s->dsp.vector_fmul_reverse(output+128, output+128, swindow, 128);
278 s->mdct128.mdct_calc(&s->mdct128, sce->coeffs + k, output);
280 for (i = 0; i < 1024; i++)
281 sce->saved[i] = audio[i * chans];
286 * Encode ics_info element.
287 * @see Table 4.6 (syntax of ics_info)
289 static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
293 put_bits(&s->pb, 1, 0); // ics_reserved bit
294 put_bits(&s->pb, 2, info->window_sequence[0]);
295 put_bits(&s->pb, 1, info->use_kb_window[0]);
296 if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
297 put_bits(&s->pb, 6, info->max_sfb);
298 put_bits(&s->pb, 1, 0); // no prediction
300 put_bits(&s->pb, 4, info->max_sfb);
301 for (w = 1; w < 8; w++)
302 put_bits(&s->pb, 1, !info->group_len[w]);
308 * @see 4.6.8.1 "Joint Coding - M/S Stereo"
310 static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
314 put_bits(pb, 2, cpe->ms_mode);
315 if (cpe->ms_mode == 1)
316 for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w])
317 for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
318 put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
322 * Produce integer coefficients from scalefactors provided by the model.
324 static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans)
327 int start, maxsfb, cmaxsfb;
329 for (ch = 0; ch < chans; ch++) {
330 IndividualChannelStream *ics = &cpe->ch[ch].ics;
333 cpe->ch[ch].pulse.num_pulse = 0;
334 for (w = 0; w < ics->num_windows*16; w += 16) {
335 for (g = 0; g < ics->num_swb; g++) {
337 if (cpe->common_window && !ch && cpe->ms_mask[w + g]) {
338 for (i = 0; i < ics->swb_sizes[g]; i++) {
339 cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
340 cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
343 start += ics->swb_sizes[g];
345 for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--)
347 maxsfb = FFMAX(maxsfb, cmaxsfb);
349 ics->max_sfb = maxsfb;
351 //adjust zero bands for window groups
352 for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
353 for (g = 0; g < ics->max_sfb; g++) {
355 for (w2 = w; w2 < w + ics->group_len[w]; w2++) {
356 if (!cpe->ch[ch].zeroes[w2*16 + g]) {
361 cpe->ch[ch].zeroes[w*16 + g] = i;
366 if (chans > 1 && cpe->common_window) {
367 IndividualChannelStream *ics0 = &cpe->ch[0].ics;
368 IndividualChannelStream *ics1 = &cpe->ch[1].ics;
370 ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
371 ics1->max_sfb = ics0->max_sfb;
372 for (w = 0; w < ics0->num_windows*16; w += 16)
373 for (i = 0; i < ics0->max_sfb; i++)
374 if (cpe->ms_mask[w+i])
376 if (msc == 0 || ics0->max_sfb == 0)
379 cpe->ms_mode = msc < ics0->max_sfb * ics0->num_windows ? 1 : 2;
384 * Encode scalefactor band coding type.
386 static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
390 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
391 s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
395 * Encode scalefactors.
397 static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s,
398 SingleChannelElement *sce)
400 int off = sce->sf_idx[0], diff;
403 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
404 for (i = 0; i < sce->ics.max_sfb; i++) {
405 if (!sce->zeroes[w*16 + i]) {
406 diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO;
407 if (diff < 0 || diff > 120)
408 av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n");
409 off = sce->sf_idx[w*16 + i];
410 put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
419 static void encode_pulses(AACEncContext *s, Pulse *pulse)
423 put_bits(&s->pb, 1, !!pulse->num_pulse);
424 if (!pulse->num_pulse)
427 put_bits(&s->pb, 2, pulse->num_pulse - 1);
428 put_bits(&s->pb, 6, pulse->start);
429 for (i = 0; i < pulse->num_pulse; i++) {
430 put_bits(&s->pb, 5, pulse->pos[i]);
431 put_bits(&s->pb, 4, pulse->amp[i]);
436 * Encode spectral coefficients processed by psychoacoustic model.
438 static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
442 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
444 for (i = 0; i < sce->ics.max_sfb; i++) {
445 if (sce->zeroes[w*16 + i]) {
446 start += sce->ics.swb_sizes[i];
449 for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++)
450 s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128,
451 sce->ics.swb_sizes[i],
452 sce->sf_idx[w*16 + i],
453 sce->band_type[w*16 + i],
455 start += sce->ics.swb_sizes[i];
461 * Encode one channel of audio data.
463 static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
464 SingleChannelElement *sce,
467 put_bits(&s->pb, 8, sce->sf_idx[0]);
469 put_ics_info(s, &sce->ics);
470 encode_band_info(s, sce);
471 encode_scale_factors(avctx, s, sce);
472 encode_pulses(s, &sce->pulse);
473 put_bits(&s->pb, 1, 0); //tns
474 put_bits(&s->pb, 1, 0); //ssr
475 encode_spectral_coeffs(s, sce);
480 * Write some auxiliary information about the created AAC file.
482 static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s,
485 int i, namelen, padbits;
487 namelen = strlen(name) + 2;
488 put_bits(&s->pb, 3, TYPE_FIL);
489 put_bits(&s->pb, 4, FFMIN(namelen, 15));
491 put_bits(&s->pb, 8, namelen - 14);
492 put_bits(&s->pb, 4, 0); //extension type - filler
493 padbits = 8 - (put_bits_count(&s->pb) & 7);
494 avpriv_align_put_bits(&s->pb);
495 for (i = 0; i < namelen - 2; i++)
496 put_bits(&s->pb, 8, name[i]);
497 put_bits(&s->pb, 12 - padbits, 0);
500 static int aac_encode_frame(AVCodecContext *avctx,
501 uint8_t *frame, int buf_size, void *data)
503 AACEncContext *s = avctx->priv_data;
504 int16_t *samples = s->samples, *samples2, *la;
506 int i, ch, w, g, chans, tag, start_ch;
507 int chan_el_counter[4];
508 FFPsyWindowInfo windows[AAC_MAX_CHANNELS];
514 if (avctx->channels <= 2) {
515 memcpy(s->samples + 1024 * avctx->channels, data,
516 1024 * avctx->channels * sizeof(s->samples[0]));
518 for (i = 0; i < 1024; i++)
519 for (ch = 0; ch < avctx->channels; ch++)
520 s->samples[(i + 1024) * avctx->channels + ch] =
521 ((int16_t*)data)[i * avctx->channels +
522 channel_maps[avctx->channels-1][ch]];
526 samples2 = s->samples + 1024 * avctx->channels;
527 for (i = 0; i < s->chan_map[0]; i++) {
528 tag = s->chan_map[i+1];
529 chans = tag == TYPE_CPE ? 2 : 1;
530 ff_psy_preprocess(s->psypp,
531 (uint16_t*)data + channel_maps[avctx->channels-1][start_ch],
532 samples2 + start_ch, start_ch, chans);
537 if (!avctx->frame_number) {
538 memcpy(s->samples, s->samples + 1024 * avctx->channels,
539 1024 * avctx->channels * sizeof(s->samples[0]));
544 for (i = 0; i < s->chan_map[0]; i++) {
545 FFPsyWindowInfo* wi = windows + start_ch;
546 tag = s->chan_map[i+1];
547 chans = tag == TYPE_CPE ? 2 : 1;
549 for (ch = 0; ch < chans; ch++) {
550 IndividualChannelStream *ics = &cpe->ch[ch].ics;
551 int cur_channel = start_ch + ch;
552 samples2 = samples + cur_channel;
553 la = samples2 + (448+64) * avctx->channels;
556 if (tag == TYPE_LFE) {
557 wi[ch].window_type[0] = ONLY_LONG_SEQUENCE;
558 wi[ch].window_shape = 0;
559 wi[ch].num_windows = 1;
560 wi[ch].grouping[0] = 1;
562 /* Only the lowest 12 coefficients are used in a LFE channel.
563 * The expression below results in only the bottom 8 coefficients
564 * being used for 11.025kHz to 16kHz sample rates.
566 ics->num_swb = s->samplerate_index >= 8 ? 1 : 3;
568 wi[ch] = s->psy.model->window(&s->psy, samples2, la, cur_channel,
569 ics->window_sequence[0]);
571 ics->window_sequence[1] = ics->window_sequence[0];
572 ics->window_sequence[0] = wi[ch].window_type[0];
573 ics->use_kb_window[1] = ics->use_kb_window[0];
574 ics->use_kb_window[0] = wi[ch].window_shape;
575 ics->num_windows = wi[ch].num_windows;
576 ics->swb_sizes = s->psy.bands [ics->num_windows == 8];
577 ics->num_swb = tag == TYPE_LFE ? ics->num_swb : s->psy.num_bands[ics->num_windows == 8];
578 for (w = 0; w < ics->num_windows; w++)
579 ics->group_len[w] = wi[ch].grouping[w];
581 apply_window_and_mdct(avctx, s, &cpe->ch[ch], samples2);
587 init_put_bits(&s->pb, frame, buf_size*8);
588 if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT))
589 put_bitstream_info(avctx, s, LIBAVCODEC_IDENT);
591 memset(chan_el_counter, 0, sizeof(chan_el_counter));
592 for (i = 0; i < s->chan_map[0]; i++) {
593 FFPsyWindowInfo* wi = windows + start_ch;
594 const float *coeffs[2];
595 tag = s->chan_map[i+1];
596 chans = tag == TYPE_CPE ? 2 : 1;
598 put_bits(&s->pb, 3, tag);
599 put_bits(&s->pb, 4, chan_el_counter[tag]++);
600 for (ch = 0; ch < chans; ch++)
601 coeffs[ch] = cpe->ch[ch].coeffs;
602 s->psy.model->analyze(&s->psy, start_ch, coeffs, wi);
603 for (ch = 0; ch < chans; ch++) {
604 s->cur_channel = start_ch * 2 + ch;
605 s->coder->search_for_quantizers(avctx, s, &cpe->ch[ch], s->lambda);
607 cpe->common_window = 0;
609 && wi[0].window_type[0] == wi[1].window_type[0]
610 && wi[0].window_shape == wi[1].window_shape) {
612 cpe->common_window = 1;
613 for (w = 0; w < wi[0].num_windows; w++) {
614 if (wi[0].grouping[w] != wi[1].grouping[w]) {
615 cpe->common_window = 0;
620 s->cur_channel = start_ch * 2;
621 if (s->options.stereo_mode && cpe->common_window) {
622 if (s->options.stereo_mode > 0) {
623 IndividualChannelStream *ics = &cpe->ch[0].ics;
624 for (w = 0; w < ics->num_windows; w += ics->group_len[w])
625 for (g = 0; g < ics->num_swb; g++)
626 cpe->ms_mask[w*16+g] = 1;
627 } else if (s->coder->search_for_ms) {
628 s->coder->search_for_ms(s, cpe, s->lambda);
631 adjust_frame_information(s, cpe, chans);
633 put_bits(&s->pb, 1, cpe->common_window);
634 if (cpe->common_window) {
635 put_ics_info(s, &cpe->ch[0].ics);
636 encode_ms_info(&s->pb, cpe);
639 for (ch = 0; ch < chans; ch++) {
640 s->cur_channel = start_ch + ch;
641 encode_individual_channel(avctx, s, &cpe->ch[ch], cpe->common_window);
646 frame_bits = put_bits_count(&s->pb);
647 if (frame_bits <= 6144 * avctx->channels - 3) {
648 s->psy.bitres.bits = frame_bits / avctx->channels;
652 s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
656 put_bits(&s->pb, 3, TYPE_END);
657 flush_put_bits(&s->pb);
658 avctx->frame_bits = put_bits_count(&s->pb);
660 // rate control stuff
661 if (!(avctx->flags & CODEC_FLAG_QSCALE)) {
662 float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
664 s->lambda = FFMIN(s->lambda, 65536.f);
669 memcpy(s->samples, s->samples + 1024 * avctx->channels,
670 1024 * avctx->channels * sizeof(s->samples[0]));
671 return put_bits_count(&s->pb)>>3;
674 static av_cold int aac_encode_end(AVCodecContext *avctx)
676 AACEncContext *s = avctx->priv_data;
678 ff_mdct_end(&s->mdct1024);
679 ff_mdct_end(&s->mdct128);
681 ff_psy_preprocess_end(s->psypp);
682 av_freep(&s->samples);
687 #define AACENC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
688 static const AVOption aacenc_options[] = {
689 {"stereo_mode", "Stereo coding method", offsetof(AACEncContext, options.stereo_mode), AV_OPT_TYPE_INT, {.dbl = 0}, -1, 1, AACENC_FLAGS, "stereo_mode"},
690 {"auto", "Selected by the Encoder", 0, AV_OPT_TYPE_CONST, {.dbl = -1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
691 {"ms_off", "Disable Mid/Side coding", 0, AV_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
692 {"ms_force", "Force Mid/Side for the whole frame if possible", 0, AV_OPT_TYPE_CONST, {.dbl = 1 }, INT_MIN, INT_MAX, AACENC_FLAGS, "stereo_mode"},
693 {"aac_coder", "", offsetof(AACEncContext, options.aac_coder), AV_OPT_TYPE_INT, {.dbl = 2}, 0, AAC_CODER_NB-1, AACENC_FLAGS},
697 static const AVClass aacenc_class = {
699 av_default_item_name,
701 LIBAVUTIL_VERSION_INT,
704 AVCodec ff_aac_encoder = {
706 .type = AVMEDIA_TYPE_AUDIO,
708 .priv_data_size = sizeof(AACEncContext),
709 .init = aac_encode_init,
710 .encode = aac_encode_frame,
711 .close = aac_encode_end,
712 .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL,
713 .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE},
714 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
715 .priv_class = &aacenc_class,