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 * 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 44static 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 50static 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 56static 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 63static 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 69static 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 75static 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 81static 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 87static 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 94static const uint8_t swb_size_128_96[] = { 95 4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36 96}; 97 98static const uint8_t swb_size_128_48[] = { 99 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16 100}; 101 102static 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 106static 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 110static 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 114static 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 */ 124static 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 */ 137static 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 153static 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 if (avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW) { 174 av_log(avctx, AV_LOG_ERROR, "Unsupported profile %d\n", avctx->profile); 175 return -1; 176 } 177 if (1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * avctx->channels) { 178 av_log(avctx, AV_LOG_ERROR, "Too many bits per frame requested\n"); 179 return -1; 180 } 181 s->samplerate_index = i; 182 183 dsputil_init(&s->dsp, avctx); 184 ff_mdct_init(&s->mdct1024, 11, 0, 1.0); 185 ff_mdct_init(&s->mdct128, 8, 0, 1.0); 186 // window init 187 ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024); 188 ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128); 189 ff_init_ff_sine_windows(10); 190 ff_init_ff_sine_windows(7); 191 192 s->samples = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0])); 193 s->cpe = av_mallocz(sizeof(ChannelElement) * aac_chan_configs[avctx->channels-1][0]); 194 avctx->extradata = av_malloc(2); 195 avctx->extradata_size = 2; 196 put_audio_specific_config(avctx); 197 198 sizes[0] = swb_size_1024[i]; 199 sizes[1] = swb_size_128[i]; 200 lengths[0] = ff_aac_num_swb_1024[i]; 201 lengths[1] = ff_aac_num_swb_128[i]; 202 ff_psy_init(&s->psy, avctx, 2, sizes, lengths); 203 s->psypp = ff_psy_preprocess_init(avctx); 204 s->coder = &ff_aac_coders[0]; 205 206 s->lambda = avctx->global_quality ? avctx->global_quality : 120; 207#if !CONFIG_HARDCODED_TABLES 208 for (i = 0; i < 428; i++) 209 ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.); 210#endif /* CONFIG_HARDCODED_TABLES */ 211 212 if (avctx->channels > 5) 213 av_log(avctx, AV_LOG_ERROR, "This encoder does not yet enforce the restrictions on LFEs. " 214 "The output will most likely be an illegal bitstream.\n"); 215 216 return 0; 217} 218 219static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s, 220 SingleChannelElement *sce, short *audio, int channel) 221{ 222 int i, j, k; 223 const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024; 224 const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128; 225 const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128; 226 227 if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) { 228 memcpy(s->output, sce->saved, sizeof(float)*1024); 229 if (sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE) { 230 memset(s->output, 0, sizeof(s->output[0]) * 448); 231 for (i = 448; i < 576; i++) 232 s->output[i] = sce->saved[i] * pwindow[i - 448]; 233 for (i = 576; i < 704; i++) 234 s->output[i] = sce->saved[i]; 235 } 236 if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) { 237 j = channel; 238 for (i = 0; i < 1024; i++, j += avctx->channels) { 239 s->output[i+1024] = audio[j] * lwindow[1024 - i - 1]; 240 sce->saved[i] = audio[j] * lwindow[i]; 241 } 242 } else { 243 j = channel; 244 for (i = 0; i < 448; i++, j += avctx->channels) 245 s->output[i+1024] = audio[j]; 246 for (i = 448; i < 576; i++, j += avctx->channels) 247 s->output[i+1024] = audio[j] * swindow[576 - i - 1]; 248 memset(s->output+1024+576, 0, sizeof(s->output[0]) * 448); 249 j = channel; 250 for (i = 0; i < 1024; i++, j += avctx->channels) 251 sce->saved[i] = audio[j]; 252 } 253 ff_mdct_calc(&s->mdct1024, sce->coeffs, s->output); 254 } else { 255 j = channel; 256 for (k = 0; k < 1024; k += 128) { 257 for (i = 448 + k; i < 448 + k + 256; i++) 258 s->output[i - 448 - k] = (i < 1024) 259 ? sce->saved[i] 260 : audio[channel + (i-1024)*avctx->channels]; 261 s->dsp.vector_fmul (s->output, k ? swindow : pwindow, 128); 262 s->dsp.vector_fmul_reverse(s->output+128, s->output+128, swindow, 128); 263 ff_mdct_calc(&s->mdct128, sce->coeffs + k, s->output); 264 } 265 j = channel; 266 for (i = 0; i < 1024; i++, j += avctx->channels) 267 sce->saved[i] = audio[j]; 268 } 269} 270 271/** 272 * Encode ics_info element. 273 * @see Table 4.6 (syntax of ics_info) 274 */ 275static void put_ics_info(AACEncContext *s, IndividualChannelStream *info) 276{ 277 int w; 278 279 put_bits(&s->pb, 1, 0); // ics_reserved bit 280 put_bits(&s->pb, 2, info->window_sequence[0]); 281 put_bits(&s->pb, 1, info->use_kb_window[0]); 282 if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) { 283 put_bits(&s->pb, 6, info->max_sfb); 284 put_bits(&s->pb, 1, 0); // no prediction 285 } else { 286 put_bits(&s->pb, 4, info->max_sfb); 287 for (w = 1; w < 8; w++) 288 put_bits(&s->pb, 1, !info->group_len[w]); 289 } 290} 291 292/** 293 * Encode MS data. 294 * @see 4.6.8.1 "Joint Coding - M/S Stereo" 295 */ 296static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe) 297{ 298 int i, w; 299 300 put_bits(pb, 2, cpe->ms_mode); 301 if (cpe->ms_mode == 1) 302 for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w]) 303 for (i = 0; i < cpe->ch[0].ics.max_sfb; i++) 304 put_bits(pb, 1, cpe->ms_mask[w*16 + i]); 305} 306 307/** 308 * Produce integer coefficients from scalefactors provided by the model. 309 */ 310static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans) 311{ 312 int i, w, w2, g, ch; 313 int start, sum, maxsfb, cmaxsfb; 314 315 for (ch = 0; ch < chans; ch++) { 316 IndividualChannelStream *ics = &cpe->ch[ch].ics; 317 start = 0; 318 maxsfb = 0; 319 cpe->ch[ch].pulse.num_pulse = 0; 320 for (w = 0; w < ics->num_windows*16; w += 16) { 321 for (g = 0; g < ics->num_swb; g++) { 322 sum = 0; 323 //apply M/S 324 if (!ch && cpe->ms_mask[w + g]) { 325 for (i = 0; i < ics->swb_sizes[g]; i++) { 326 cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0; 327 cpe->ch[1].coeffs[start+i] = cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i]; 328 } 329 } 330 start += ics->swb_sizes[g]; 331 } 332 for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--) 333 ; 334 maxsfb = FFMAX(maxsfb, cmaxsfb); 335 } 336 ics->max_sfb = maxsfb; 337 338 //adjust zero bands for window groups 339 for (w = 0; w < ics->num_windows; w += ics->group_len[w]) { 340 for (g = 0; g < ics->max_sfb; g++) { 341 i = 1; 342 for (w2 = w; w2 < w + ics->group_len[w]; w2++) { 343 if (!cpe->ch[ch].zeroes[w2*16 + g]) { 344 i = 0; 345 break; 346 } 347 } 348 cpe->ch[ch].zeroes[w*16 + g] = i; 349 } 350 } 351 } 352 353 if (chans > 1 && cpe->common_window) { 354 IndividualChannelStream *ics0 = &cpe->ch[0].ics; 355 IndividualChannelStream *ics1 = &cpe->ch[1].ics; 356 int msc = 0; 357 ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb); 358 ics1->max_sfb = ics0->max_sfb; 359 for (w = 0; w < ics0->num_windows*16; w += 16) 360 for (i = 0; i < ics0->max_sfb; i++) 361 if (cpe->ms_mask[w+i]) 362 msc++; 363 if (msc == 0 || ics0->max_sfb == 0) 364 cpe->ms_mode = 0; 365 else 366 cpe->ms_mode = msc < ics0->max_sfb ? 1 : 2; 367 } 368} 369 370/** 371 * Encode scalefactor band coding type. 372 */ 373static void encode_band_info(AACEncContext *s, SingleChannelElement *sce) 374{ 375 int w; 376 377 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) 378 s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda); 379} 380 381/** 382 * Encode scalefactors. 383 */ 384static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s, 385 SingleChannelElement *sce) 386{ 387 int off = sce->sf_idx[0], diff; 388 int i, w; 389 390 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { 391 for (i = 0; i < sce->ics.max_sfb; i++) { 392 if (!sce->zeroes[w*16 + i]) { 393 diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO; 394 if (diff < 0 || diff > 120) 395 av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n"); 396 off = sce->sf_idx[w*16 + i]; 397 put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]); 398 } 399 } 400 } 401} 402 403/** 404 * Encode pulse data. 405 */ 406static void encode_pulses(AACEncContext *s, Pulse *pulse) 407{ 408 int i; 409 410 put_bits(&s->pb, 1, !!pulse->num_pulse); 411 if (!pulse->num_pulse) 412 return; 413 414 put_bits(&s->pb, 2, pulse->num_pulse - 1); 415 put_bits(&s->pb, 6, pulse->start); 416 for (i = 0; i < pulse->num_pulse; i++) { 417 put_bits(&s->pb, 5, pulse->pos[i]); 418 put_bits(&s->pb, 4, pulse->amp[i]); 419 } 420} 421 422/** 423 * Encode spectral coefficients processed by psychoacoustic model. 424 */ 425static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce) 426{ 427 int start, i, w, w2; 428 429 for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) { 430 start = 0; 431 for (i = 0; i < sce->ics.max_sfb; i++) { 432 if (sce->zeroes[w*16 + i]) { 433 start += sce->ics.swb_sizes[i]; 434 continue; 435 } 436 for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++) 437 s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128, 438 sce->ics.swb_sizes[i], 439 sce->sf_idx[w*16 + i], 440 sce->band_type[w*16 + i], 441 s->lambda); 442 start += sce->ics.swb_sizes[i]; 443 } 444 } 445} 446 447/** 448 * Encode one channel of audio data. 449 */ 450static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s, 451 SingleChannelElement *sce, 452 int common_window) 453{ 454 put_bits(&s->pb, 8, sce->sf_idx[0]); 455 if (!common_window) 456 put_ics_info(s, &sce->ics); 457 encode_band_info(s, sce); 458 encode_scale_factors(avctx, s, sce); 459 encode_pulses(s, &sce->pulse); 460 put_bits(&s->pb, 1, 0); //tns 461 put_bits(&s->pb, 1, 0); //ssr 462 encode_spectral_coeffs(s, sce); 463 return 0; 464} 465 466/** 467 * Write some auxiliary information about the created AAC file. 468 */ 469static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s, 470 const char *name) 471{ 472 int i, namelen, padbits; 473 474 namelen = strlen(name) + 2; 475 put_bits(&s->pb, 3, TYPE_FIL); 476 put_bits(&s->pb, 4, FFMIN(namelen, 15)); 477 if (namelen >= 15) 478 put_bits(&s->pb, 8, namelen - 16); 479 put_bits(&s->pb, 4, 0); //extension type - filler 480 padbits = 8 - (put_bits_count(&s->pb) & 7); 481 align_put_bits(&s->pb); 482 for (i = 0; i < namelen - 2; i++) 483 put_bits(&s->pb, 8, name[i]); 484 put_bits(&s->pb, 12 - padbits, 0); 485} 486 487static int aac_encode_frame(AVCodecContext *avctx, 488 uint8_t *frame, int buf_size, void *data) 489{ 490 AACEncContext *s = avctx->priv_data; 491 int16_t *samples = s->samples, *samples2, *la; 492 ChannelElement *cpe; 493 int i, j, chans, tag, start_ch; 494 const uint8_t *chan_map = aac_chan_configs[avctx->channels-1]; 495 int chan_el_counter[4]; 496 FFPsyWindowInfo windows[avctx->channels]; 497 498 if (s->last_frame) 499 return 0; 500 if (data) { 501 if (!s->psypp) { 502 memcpy(s->samples + 1024 * avctx->channels, data, 503 1024 * avctx->channels * sizeof(s->samples[0])); 504 } else { 505 start_ch = 0; 506 samples2 = s->samples + 1024 * avctx->channels; 507 for (i = 0; i < chan_map[0]; i++) { 508 tag = chan_map[i+1]; 509 chans = tag == TYPE_CPE ? 2 : 1; 510 ff_psy_preprocess(s->psypp, (uint16_t*)data + start_ch, 511 samples2 + start_ch, start_ch, chans); 512 start_ch += chans; 513 } 514 } 515 } 516 if (!avctx->frame_number) { 517 memcpy(s->samples, s->samples + 1024 * avctx->channels, 518 1024 * avctx->channels * sizeof(s->samples[0])); 519 return 0; 520 } 521 522 start_ch = 0; 523 for (i = 0; i < chan_map[0]; i++) { 524 FFPsyWindowInfo* wi = windows + start_ch; 525 tag = chan_map[i+1]; 526 chans = tag == TYPE_CPE ? 2 : 1; 527 cpe = &s->cpe[i]; 528 samples2 = samples + start_ch; 529 la = samples2 + 1024 * avctx->channels + start_ch; 530 if (!data) 531 la = NULL; 532 for (j = 0; j < chans; j++) { 533 IndividualChannelStream *ics = &cpe->ch[j].ics; 534 int k; 535 wi[j] = ff_psy_suggest_window(&s->psy, samples2, la, start_ch + j, ics->window_sequence[0]); 536 ics->window_sequence[1] = ics->window_sequence[0]; 537 ics->window_sequence[0] = wi[j].window_type[0]; 538 ics->use_kb_window[1] = ics->use_kb_window[0]; 539 ics->use_kb_window[0] = wi[j].window_shape; 540 ics->num_windows = wi[j].num_windows; 541 ics->swb_sizes = s->psy.bands [ics->num_windows == 8]; 542 ics->num_swb = s->psy.num_bands[ics->num_windows == 8]; 543 for (k = 0; k < ics->num_windows; k++) 544 ics->group_len[k] = wi[j].grouping[k]; 545 546 s->cur_channel = start_ch + j; 547 apply_window_and_mdct(avctx, s, &cpe->ch[j], samples2, j); 548 } 549 start_ch += chans; 550 } 551 do { 552 int frame_bits; 553 init_put_bits(&s->pb, frame, buf_size*8); 554 if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT)) 555 put_bitstream_info(avctx, s, LIBAVCODEC_IDENT); 556 start_ch = 0; 557 memset(chan_el_counter, 0, sizeof(chan_el_counter)); 558 for (i = 0; i < chan_map[0]; i++) { 559 FFPsyWindowInfo* wi = windows + start_ch; 560 tag = chan_map[i+1]; 561 chans = tag == TYPE_CPE ? 2 : 1; 562 cpe = &s->cpe[i]; 563 for (j = 0; j < chans; j++) { 564 s->cur_channel = start_ch + j; 565 s->coder->search_for_quantizers(avctx, s, &cpe->ch[j], s->lambda); 566 } 567 cpe->common_window = 0; 568 if (chans > 1 569 && wi[0].window_type[0] == wi[1].window_type[0] 570 && wi[0].window_shape == wi[1].window_shape) { 571 572 cpe->common_window = 1; 573 for (j = 0; j < wi[0].num_windows; j++) { 574 if (wi[0].grouping[j] != wi[1].grouping[j]) { 575 cpe->common_window = 0; 576 break; 577 } 578 } 579 } 580 s->cur_channel = start_ch; 581 if (cpe->common_window && s->coder->search_for_ms) 582 s->coder->search_for_ms(s, cpe, s->lambda); 583 adjust_frame_information(s, cpe, chans); 584 put_bits(&s->pb, 3, tag); 585 put_bits(&s->pb, 4, chan_el_counter[tag]++); 586 if (chans == 2) { 587 put_bits(&s->pb, 1, cpe->common_window); 588 if (cpe->common_window) { 589 put_ics_info(s, &cpe->ch[0].ics); 590 encode_ms_info(&s->pb, cpe); 591 } 592 } 593 for (j = 0; j < chans; j++) { 594 s->cur_channel = start_ch + j; 595 ff_psy_set_band_info(&s->psy, s->cur_channel, cpe->ch[j].coeffs, &wi[j]); 596 encode_individual_channel(avctx, s, &cpe->ch[j], cpe->common_window); 597 } 598 start_ch += chans; 599 } 600 601 frame_bits = put_bits_count(&s->pb); 602 if (frame_bits <= 6144 * avctx->channels - 3) 603 break; 604 605 s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits; 606 607 } while (1); 608 609 put_bits(&s->pb, 3, TYPE_END); 610 flush_put_bits(&s->pb); 611 avctx->frame_bits = put_bits_count(&s->pb); 612 613 // rate control stuff 614 if (!(avctx->flags & CODEC_FLAG_QSCALE)) { 615 float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits; 616 s->lambda *= ratio; 617 s->lambda = FFMIN(s->lambda, 65536.f); 618 } 619 620 if (!data) 621 s->last_frame = 1; 622 memcpy(s->samples, s->samples + 1024 * avctx->channels, 623 1024 * avctx->channels * sizeof(s->samples[0])); 624 return put_bits_count(&s->pb)>>3; 625} 626 627static av_cold int aac_encode_end(AVCodecContext *avctx) 628{ 629 AACEncContext *s = avctx->priv_data; 630 631 ff_mdct_end(&s->mdct1024); 632 ff_mdct_end(&s->mdct128); 633 ff_psy_end(&s->psy); 634 ff_psy_preprocess_end(s->psypp); 635 av_freep(&s->samples); 636 av_freep(&s->cpe); 637 return 0; 638} 639 640AVCodec aac_encoder = { 641 "aac", 642 AVMEDIA_TYPE_AUDIO, 643 CODEC_ID_AAC, 644 sizeof(AACEncContext), 645 aac_encode_init, 646 aac_encode_frame, 647 aac_encode_end, 648 .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL, 649 .sample_fmts = (const enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE}, 650 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"), 651}; 652