3 * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
4 * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26 * @author Oded Shimon ( ods15 ods15 dyndns org )
27 * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
34 * N (code in SoC repo) gain control
36 * Y window shapes - standard
37 * N window shapes - Low Delay
38 * Y filterbank - standard
39 * N (code in SoC repo) filterbank - Scalable Sample Rate
40 * Y Temporal Noise Shaping
41 * N (code in SoC repo) Long Term Prediction
44 * N frequency domain prediction
45 * Y Perceptual Noise Substitution
47 * N Scalable Inverse AAC Quantization
48 * N Frequency Selective Switch
50 * Y quantization & coding - AAC
51 * N quantization & coding - TwinVQ
52 * N quantization & coding - BSAC
53 * N AAC Error Resilience tools
54 * N Error Resilience payload syntax
55 * N Error Protection tool
57 * N Silence Compression
60 * N Structured Audio tools
61 * N Structured Audio Sample Bank Format
63 * N Harmonic and Individual Lines plus Noise
64 * N Text-To-Speech Interface
65 * N (in progress) Spectral Band Replication
66 * Y (not in this code) Layer-1
67 * Y (not in this code) Layer-2
68 * Y (not in this code) Layer-3
69 * N SinuSoidal Coding (Transient, Sinusoid, Noise)
70 * N (planned) Parametric Stereo
71 * N Direct Stream Transfer
73 * Note: - HE AAC v1 comprises LC AAC with Spectral Band Replication.
74 * - HE AAC v2 comprises LC AAC with Spectral Band Replication and
80 #include "bitstream.h"
85 #include "aacdectab.h"
86 #include "mpeg4audio.h"
93 static VLC vlc_scalefactors;
94 static VLC vlc_spectral[11];
98 * Configure output channel order based on the current program configuration element.
100 * @param che_pos current channel position configuration
101 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
103 * @return Returns error status. 0 - OK, !0 - error
105 static int output_configure(AACContext *ac, enum ChannelPosition che_pos[4][MAX_ELEM_ID],
106 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID]) {
107 AVCodecContext *avctx = ac->avccontext;
108 int i, type, channels = 0;
110 if(!memcmp(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0])))
111 return 0; /* no change */
113 memcpy(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
115 /* Allocate or free elements depending on if they are in the
116 * current program configuration.
118 * Set up default 1:1 output mapping.
120 * For a 5.1 stream the output order will be:
121 * [ Front Left ] [ Front Right ] [ Center ] [ LFE ] [ Surround Left ] [ Surround Right ]
124 for(i = 0; i < MAX_ELEM_ID; i++) {
125 for(type = 0; type < 4; type++) {
126 if(che_pos[type][i]) {
127 if(!ac->che[type][i] && !(ac->che[type][i] = av_mallocz(sizeof(ChannelElement))))
128 return AVERROR(ENOMEM);
129 if(type != TYPE_CCE) {
130 ac->output_data[channels++] = ac->che[type][i]->ch[0].ret;
131 if(type == TYPE_CPE) {
132 ac->output_data[channels++] = ac->che[type][i]->ch[1].ret;
136 av_freep(&ac->che[type][i]);
140 avctx->channels = channels;
145 * Decode an array of 4 bit element IDs, optionally interleaved with a stereo/mono switching bit.
147 * @param cpe_map Stereo (Channel Pair Element) map, NULL if stereo bit is not present.
148 * @param sce_map mono (Single Channel Element) map
149 * @param type speaker type/position for these channels
151 static void decode_channel_map(enum ChannelPosition *cpe_map,
152 enum ChannelPosition *sce_map, enum ChannelPosition type, GetBitContext * gb, int n) {
154 enum ChannelPosition *map = cpe_map && get_bits1(gb) ? cpe_map : sce_map; // stereo or mono map
155 map[get_bits(gb, 4)] = type;
160 * Decode program configuration element; reference: table 4.2.
162 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
164 * @return Returns error status. 0 - OK, !0 - error
166 static int decode_pce(AACContext * ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
167 GetBitContext * gb) {
168 int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc;
170 skip_bits(gb, 2); // object_type
172 ac->m4ac.sampling_index = get_bits(gb, 4);
173 if(ac->m4ac.sampling_index > 11) {
174 av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
177 ac->m4ac.sample_rate = ff_mpeg4audio_sample_rates[ac->m4ac.sampling_index];
178 num_front = get_bits(gb, 4);
179 num_side = get_bits(gb, 4);
180 num_back = get_bits(gb, 4);
181 num_lfe = get_bits(gb, 2);
182 num_assoc_data = get_bits(gb, 3);
183 num_cc = get_bits(gb, 4);
186 skip_bits(gb, 4); // mono_mixdown_tag
188 skip_bits(gb, 4); // stereo_mixdown_tag
191 skip_bits(gb, 3); // mixdown_coeff_index and pseudo_surround
193 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_FRONT, gb, num_front);
194 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_SIDE, gb, num_side );
195 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_BACK, gb, num_back );
196 decode_channel_map(NULL, new_che_pos[TYPE_LFE], AAC_CHANNEL_LFE, gb, num_lfe );
198 skip_bits_long(gb, 4 * num_assoc_data);
200 decode_channel_map(new_che_pos[TYPE_CCE], new_che_pos[TYPE_CCE], AAC_CHANNEL_CC, gb, num_cc );
204 /* comment field, first byte is length */
205 skip_bits_long(gb, 8 * get_bits(gb, 8));
210 * Set up channel positions based on a default channel configuration
211 * as specified in table 1.17.
213 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
215 * @return Returns error status. 0 - OK, !0 - error
217 static int set_default_channel_config(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
220 if(channel_config < 1 || channel_config > 7) {
221 av_log(ac->avccontext, AV_LOG_ERROR, "invalid default channel configuration (%d)\n",
226 /* default channel configurations:
228 * 1ch : front center (mono)
229 * 2ch : L + R (stereo)
230 * 3ch : front center + L + R
231 * 4ch : front center + L + R + back center
232 * 5ch : front center + L + R + back stereo
233 * 6ch : front center + L + R + back stereo + LFE
234 * 7ch : front center + L + R + outer front left + outer front right + back stereo + LFE
237 if(channel_config != 2)
238 new_che_pos[TYPE_SCE][0] = AAC_CHANNEL_FRONT; // front center (or mono)
239 if(channel_config > 1)
240 new_che_pos[TYPE_CPE][0] = AAC_CHANNEL_FRONT; // L + R (or stereo)
241 if(channel_config == 4)
242 new_che_pos[TYPE_SCE][1] = AAC_CHANNEL_BACK; // back center
243 if(channel_config > 4)
244 new_che_pos[TYPE_CPE][(channel_config == 7) + 1]
245 = AAC_CHANNEL_BACK; // back stereo
246 if(channel_config > 5)
247 new_che_pos[TYPE_LFE][0] = AAC_CHANNEL_LFE; // LFE
248 if(channel_config == 7)
249 new_che_pos[TYPE_CPE][1] = AAC_CHANNEL_FRONT; // outer front left + outer front right
255 * Decode GA "General Audio" specific configuration; reference: table 4.1.
257 * @return Returns error status. 0 - OK, !0 - error
259 static int decode_ga_specific_config(AACContext * ac, GetBitContext * gb, int channel_config) {
260 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
261 int extension_flag, ret;
263 if(get_bits1(gb)) { // frameLengthFlag
264 av_log_missing_feature(ac->avccontext, "960/120 MDCT window is", 1);
268 if (get_bits1(gb)) // dependsOnCoreCoder
269 skip_bits(gb, 14); // coreCoderDelay
270 extension_flag = get_bits1(gb);
272 if(ac->m4ac.object_type == AOT_AAC_SCALABLE ||
273 ac->m4ac.object_type == AOT_ER_AAC_SCALABLE)
274 skip_bits(gb, 3); // layerNr
276 memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
277 if (channel_config == 0) {
278 skip_bits(gb, 4); // element_instance_tag
279 if((ret = decode_pce(ac, new_che_pos, gb)))
282 if((ret = set_default_channel_config(ac, new_che_pos, channel_config)))
285 if((ret = output_configure(ac, ac->che_pos, new_che_pos)))
288 if (extension_flag) {
289 switch (ac->m4ac.object_type) {
291 skip_bits(gb, 5); // numOfSubFrame
292 skip_bits(gb, 11); // layer_length
296 case AOT_ER_AAC_SCALABLE:
298 skip_bits(gb, 3); /* aacSectionDataResilienceFlag
299 * aacScalefactorDataResilienceFlag
300 * aacSpectralDataResilienceFlag
304 skip_bits1(gb); // extensionFlag3 (TBD in version 3)
310 * Decode audio specific configuration; reference: table 1.13.
312 * @param data pointer to AVCodecContext extradata
313 * @param data_size size of AVCCodecContext extradata
315 * @return Returns error status. 0 - OK, !0 - error
317 static int decode_audio_specific_config(AACContext * ac, void *data, int data_size) {
321 init_get_bits(&gb, data, data_size * 8);
323 if((i = ff_mpeg4audio_get_config(&ac->m4ac, data, data_size)) < 0)
325 if(ac->m4ac.sampling_index > 11) {
326 av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
330 skip_bits_long(&gb, i);
332 switch (ac->m4ac.object_type) {
334 if (decode_ga_specific_config(ac, &gb, ac->m4ac.chan_config))
338 av_log(ac->avccontext, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n",
339 ac->m4ac.sbr == 1? "SBR+" : "", ac->m4ac.object_type);
346 * linear congruential pseudorandom number generator
348 * @param previous_val pointer to the current state of the generator
350 * @return Returns a 32-bit pseudorandom integer
352 static av_always_inline int lcg_random(int previous_val) {
353 return previous_val * 1664525 + 1013904223;
356 static av_cold int aac_decode_init(AVCodecContext * avccontext) {
357 AACContext * ac = avccontext->priv_data;
360 ac->avccontext = avccontext;
362 if (avccontext->extradata_size <= 0 ||
363 decode_audio_specific_config(ac, avccontext->extradata, avccontext->extradata_size))
366 avccontext->sample_fmt = SAMPLE_FMT_S16;
367 avccontext->sample_rate = ac->m4ac.sample_rate;
368 avccontext->frame_size = 1024;
370 AAC_INIT_VLC_STATIC( 0, 144);
371 AAC_INIT_VLC_STATIC( 1, 114);
372 AAC_INIT_VLC_STATIC( 2, 188);
373 AAC_INIT_VLC_STATIC( 3, 180);
374 AAC_INIT_VLC_STATIC( 4, 172);
375 AAC_INIT_VLC_STATIC( 5, 140);
376 AAC_INIT_VLC_STATIC( 6, 168);
377 AAC_INIT_VLC_STATIC( 7, 114);
378 AAC_INIT_VLC_STATIC( 8, 262);
379 AAC_INIT_VLC_STATIC( 9, 248);
380 AAC_INIT_VLC_STATIC(10, 384);
382 dsputil_init(&ac->dsp, avccontext);
384 ac->random_state = 0x1f2e3d4c;
386 // -1024 - Compensate wrong IMDCT method.
387 // 32768 - Required to scale values to the correct range for the bias method
388 // for float to int16 conversion.
390 if(ac->dsp.float_to_int16 == ff_float_to_int16_c) {
391 ac->add_bias = 385.0f;
392 ac->sf_scale = 1. / (-1024. * 32768.);
396 ac->sf_scale = 1. / -1024.;
400 #ifndef CONFIG_HARDCODED_TABLES
401 for (i = 0; i < 316; i++)
402 ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.);
403 #endif /* CONFIG_HARDCODED_TABLES */
405 INIT_VLC_STATIC(&vlc_scalefactors, 7, sizeof(ff_aac_scalefactor_code)/sizeof(ff_aac_scalefactor_code[0]),
406 ff_aac_scalefactor_bits, sizeof(ff_aac_scalefactor_bits[0]), sizeof(ff_aac_scalefactor_bits[0]),
407 ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]),
410 ff_mdct_init(&ac->mdct, 11, 1);
411 ff_mdct_init(&ac->mdct_small, 8, 1);
412 // window initialization
413 ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
414 ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
415 ff_sine_window_init(ff_sine_1024, 1024);
416 ff_sine_window_init(ff_sine_128, 128);
422 * Skip data_stream_element; reference: table 4.10.
424 static void skip_data_stream_element(GetBitContext * gb) {
425 int byte_align = get_bits1(gb);
426 int count = get_bits(gb, 8);
428 count += get_bits(gb, 8);
431 skip_bits_long(gb, 8 * count);
435 * Decode Individual Channel Stream info; reference: table 4.6.
437 * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
439 static int decode_ics_info(AACContext * ac, IndividualChannelStream * ics, GetBitContext * gb, int common_window) {
441 av_log(ac->avccontext, AV_LOG_ERROR, "Reserved bit set.\n");
442 memset(ics, 0, sizeof(IndividualChannelStream));
445 ics->window_sequence[1] = ics->window_sequence[0];
446 ics->window_sequence[0] = get_bits(gb, 2);
447 ics->use_kb_window[1] = ics->use_kb_window[0];
448 ics->use_kb_window[0] = get_bits1(gb);
449 ics->num_window_groups = 1;
450 ics->group_len[0] = 1;
451 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
453 ics->max_sfb = get_bits(gb, 4);
454 for (i = 0; i < 7; i++) {
456 ics->group_len[ics->num_window_groups-1]++;
458 ics->num_window_groups++;
459 ics->group_len[ics->num_window_groups-1] = 1;
462 ics->num_windows = 8;
463 ics->swb_offset = swb_offset_128[ac->m4ac.sampling_index];
464 ics->num_swb = ff_aac_num_swb_128[ac->m4ac.sampling_index];
465 ics->tns_max_bands = tns_max_bands_128[ac->m4ac.sampling_index];
467 ics->max_sfb = get_bits(gb, 6);
468 ics->num_windows = 1;
469 ics->swb_offset = swb_offset_1024[ac->m4ac.sampling_index];
470 ics->num_swb = ff_aac_num_swb_1024[ac->m4ac.sampling_index];
471 ics->tns_max_bands = tns_max_bands_1024[ac->m4ac.sampling_index];
473 av_log_missing_feature(ac->avccontext, "Predictor bit set but LTP is", 1);
474 memset(ics, 0, sizeof(IndividualChannelStream));
479 if(ics->max_sfb > ics->num_swb) {
480 av_log(ac->avccontext, AV_LOG_ERROR,
481 "Number of scalefactor bands in group (%d) exceeds limit (%d).\n",
482 ics->max_sfb, ics->num_swb);
483 memset(ics, 0, sizeof(IndividualChannelStream));
491 * Decode band types (section_data payload); reference: table 4.46.
493 * @param band_type array of the used band type
494 * @param band_type_run_end array of the last scalefactor band of a band type run
496 * @return Returns error status. 0 - OK, !0 - error
498 static int decode_band_types(AACContext * ac, enum BandType band_type[120],
499 int band_type_run_end[120], GetBitContext * gb, IndividualChannelStream * ics) {
501 const int bits = (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) ? 3 : 5;
502 for (g = 0; g < ics->num_window_groups; g++) {
504 while (k < ics->max_sfb) {
505 uint8_t sect_len = k;
507 int sect_band_type = get_bits(gb, 4);
508 if (sect_band_type == 12) {
509 av_log(ac->avccontext, AV_LOG_ERROR, "invalid band type\n");
512 while ((sect_len_incr = get_bits(gb, bits)) == (1 << bits)-1)
513 sect_len += sect_len_incr;
514 sect_len += sect_len_incr;
515 if (sect_len > ics->max_sfb) {
516 av_log(ac->avccontext, AV_LOG_ERROR,
517 "Number of bands (%d) exceeds limit (%d).\n",
518 sect_len, ics->max_sfb);
521 for (; k < sect_len; k++) {
522 band_type [idx] = sect_band_type;
523 band_type_run_end[idx++] = sect_len;
531 * Decode scalefactors; reference: table 4.47.
533 * @param global_gain first scalefactor value as scalefactors are differentially coded
534 * @param band_type array of the used band type
535 * @param band_type_run_end array of the last scalefactor band of a band type run
536 * @param sf array of scalefactors or intensity stereo positions
538 * @return Returns error status. 0 - OK, !0 - error
540 static int decode_scalefactors(AACContext * ac, float sf[120], GetBitContext * gb,
541 unsigned int global_gain, IndividualChannelStream * ics,
542 enum BandType band_type[120], int band_type_run_end[120]) {
543 const int sf_offset = ac->sf_offset + (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE ? 12 : 0);
545 int offset[3] = { global_gain, global_gain - 90, 100 };
547 static const char *sf_str[3] = { "Global gain", "Noise gain", "Intensity stereo position" };
548 for (g = 0; g < ics->num_window_groups; g++) {
549 for (i = 0; i < ics->max_sfb;) {
550 int run_end = band_type_run_end[idx];
551 if (band_type[idx] == ZERO_BT) {
552 for(; i < run_end; i++, idx++)
554 }else if((band_type[idx] == INTENSITY_BT) || (band_type[idx] == INTENSITY_BT2)) {
555 for(; i < run_end; i++, idx++) {
556 offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
557 if(offset[2] > 255U) {
558 av_log(ac->avccontext, AV_LOG_ERROR,
559 "%s (%d) out of range.\n", sf_str[2], offset[2]);
562 sf[idx] = ff_aac_pow2sf_tab[-offset[2] + 300];
564 }else if(band_type[idx] == NOISE_BT) {
565 for(; i < run_end; i++, idx++) {
567 offset[1] += get_bits(gb, 9) - 256;
569 offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
570 if(offset[1] > 255U) {
571 av_log(ac->avccontext, AV_LOG_ERROR,
572 "%s (%d) out of range.\n", sf_str[1], offset[1]);
575 sf[idx] = -ff_aac_pow2sf_tab[ offset[1] + sf_offset];
578 for(; i < run_end; i++, idx++) {
579 offset[0] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
580 if(offset[0] > 255U) {
581 av_log(ac->avccontext, AV_LOG_ERROR,
582 "%s (%d) out of range.\n", sf_str[0], offset[0]);
585 sf[idx] = -ff_aac_pow2sf_tab[ offset[0] + sf_offset];
594 * Decode pulse data; reference: table 4.7.
596 static void decode_pulses(Pulse * pulse, GetBitContext * gb, const uint16_t * swb_offset) {
598 pulse->num_pulse = get_bits(gb, 2) + 1;
599 pulse->pos[0] = get_bits(gb, 5) + swb_offset[get_bits(gb, 6)];
600 pulse->amp[0] = get_bits(gb, 4);
601 for (i = 1; i < pulse->num_pulse; i++) {
602 pulse->pos[i] = get_bits(gb, 5) + pulse->pos[i-1];
603 pulse->amp[i] = get_bits(gb, 4);
608 * Decode Temporal Noise Shaping data; reference: table 4.48.
610 * @return Returns error status. 0 - OK, !0 - error
612 static int decode_tns(AACContext * ac, TemporalNoiseShaping * tns,
613 GetBitContext * gb, const IndividualChannelStream * ics) {
614 int w, filt, i, coef_len, coef_res, coef_compress;
615 const int is8 = ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE;
616 const int tns_max_order = is8 ? 7 : ac->m4ac.object_type == AOT_AAC_MAIN ? 20 : 12;
617 for (w = 0; w < ics->num_windows; w++) {
618 tns->n_filt[w] = get_bits(gb, 2 - is8);
621 coef_res = get_bits1(gb);
623 for (filt = 0; filt < tns->n_filt[w]; filt++) {
625 tns->length[w][filt] = get_bits(gb, 6 - 2*is8);
627 if ((tns->order[w][filt] = get_bits(gb, 5 - 2*is8)) > tns_max_order) {
628 av_log(ac->avccontext, AV_LOG_ERROR, "TNS filter order %d is greater than maximum %d.",
629 tns->order[w][filt], tns_max_order);
630 tns->order[w][filt] = 0;
633 tns->direction[w][filt] = get_bits1(gb);
634 coef_compress = get_bits1(gb);
635 coef_len = coef_res + 3 - coef_compress;
636 tmp2_idx = 2*coef_compress + coef_res;
638 for (i = 0; i < tns->order[w][filt]; i++)
639 tns->coef[w][filt][i] = tns_tmp2_map[tmp2_idx][get_bits(gb, coef_len)];
646 * Decode Mid/Side data; reference: table 4.54.
648 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
649 * [1] mask is decoded from bitstream; [2] mask is all 1s;
650 * [3] reserved for scalable AAC
652 static void decode_mid_side_stereo(ChannelElement * cpe, GetBitContext * gb,
655 if (ms_present == 1) {
656 for (idx = 0; idx < cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb; idx++)
657 cpe->ms_mask[idx] = get_bits1(gb);
658 } else if (ms_present == 2) {
659 memset(cpe->ms_mask, 1, cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb * sizeof(cpe->ms_mask[0]));
664 * Decode spectral data; reference: table 4.50.
665 * Dequantize and scale spectral data; reference: 4.6.3.3.
667 * @param coef array of dequantized, scaled spectral data
668 * @param sf array of scalefactors or intensity stereo positions
669 * @param pulse_present set if pulses are present
670 * @param pulse pointer to pulse data struct
671 * @param band_type array of the used band type
673 * @return Returns error status. 0 - OK, !0 - error
675 static int decode_spectrum_and_dequant(AACContext * ac, float coef[1024], GetBitContext * gb, float sf[120],
676 int pulse_present, const Pulse * pulse, const IndividualChannelStream * ics, enum BandType band_type[120]) {
677 int i, k, g, idx = 0;
678 const int c = 1024/ics->num_windows;
679 const uint16_t * offsets = ics->swb_offset;
680 float *coef_base = coef;
682 for (g = 0; g < ics->num_windows; g++)
683 memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float)*(c - offsets[ics->max_sfb]));
685 for (g = 0; g < ics->num_window_groups; g++) {
686 for (i = 0; i < ics->max_sfb; i++, idx++) {
687 const int cur_band_type = band_type[idx];
688 const int dim = cur_band_type >= FIRST_PAIR_BT ? 2 : 4;
689 const int is_cb_unsigned = IS_CODEBOOK_UNSIGNED(cur_band_type);
691 if (cur_band_type == ZERO_BT) {
692 for (group = 0; group < ics->group_len[g]; group++) {
693 memset(coef + group * 128 + offsets[i], 0, (offsets[i+1] - offsets[i])*sizeof(float));
695 }else if (cur_band_type == NOISE_BT) {
696 const float scale = sf[idx] / ((offsets[i+1] - offsets[i]) * PNS_MEAN_ENERGY);
697 for (group = 0; group < ics->group_len[g]; group++) {
698 for (k = offsets[i]; k < offsets[i+1]; k++) {
699 ac->random_state = lcg_random(ac->random_state);
700 coef[group*128+k] = ac->random_state * scale;
703 }else if (cur_band_type != INTENSITY_BT2 && cur_band_type != INTENSITY_BT) {
704 for (group = 0; group < ics->group_len[g]; group++) {
705 for (k = offsets[i]; k < offsets[i+1]; k += dim) {
706 const int index = get_vlc2(gb, vlc_spectral[cur_band_type - 1].table, 6, 3);
707 const int coef_tmp_idx = (group << 7) + k;
710 if(index >= ff_aac_spectral_sizes[cur_band_type - 1]) {
711 av_log(ac->avccontext, AV_LOG_ERROR,
712 "Read beyond end of ff_aac_codebook_vectors[%d][]. index %d >= %d\n",
713 cur_band_type - 1, index, ff_aac_spectral_sizes[cur_band_type - 1]);
716 vq_ptr = &ff_aac_codebook_vectors[cur_band_type - 1][index * dim];
717 if (is_cb_unsigned) {
718 for (j = 0; j < dim; j++)
720 coef[coef_tmp_idx + j] = 1 - 2*(int)get_bits1(gb);
722 for (j = 0; j < dim; j++)
723 coef[coef_tmp_idx + j] = 1.0f;
725 if (cur_band_type == ESC_BT) {
726 for (j = 0; j < 2; j++) {
727 if (vq_ptr[j] == 64.0f) {
729 /* The total length of escape_sequence must be < 22 bits according
730 to the specification (i.e. max is 11111111110xxxxxxxxxx). */
731 while (get_bits1(gb) && n < 15) n++;
733 av_log(ac->avccontext, AV_LOG_ERROR, "error in spectral data, ESC overflow\n");
736 n = (1<<n) + get_bits(gb, n);
737 coef[coef_tmp_idx + j] *= cbrtf(fabsf(n)) * n;
739 coef[coef_tmp_idx + j] *= vq_ptr[j];
742 for (j = 0; j < dim; j++)
743 coef[coef_tmp_idx + j] *= vq_ptr[j];
744 for (j = 0; j < dim; j++)
745 coef[coef_tmp_idx + j] *= sf[idx];
750 coef += ics->group_len[g]<<7;
754 for(i = 0; i < pulse->num_pulse; i++){
755 float co = coef_base[ pulse->pos[i] ];
756 float ico = co / sqrtf(sqrtf(fabsf(co))) + pulse->amp[i];
757 coef_base[ pulse->pos[i] ] = cbrtf(fabsf(ico)) * ico;
764 * Decode an individual_channel_stream payload; reference: table 4.44.
766 * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
767 * @param scale_flag scalable [1] or non-scalable [0] AAC (Unused until scalable AAC is implemented.)
769 * @return Returns error status. 0 - OK, !0 - error
771 static int decode_ics(AACContext * ac, SingleChannelElement * sce, GetBitContext * gb, int common_window, int scale_flag) {
773 TemporalNoiseShaping * tns = &sce->tns;
774 IndividualChannelStream * ics = &sce->ics;
775 float * out = sce->coeffs;
776 int global_gain, pulse_present = 0;
778 /* This assignment is to silence a GCC warning about the variable being used
779 * uninitialized when in fact it always is.
783 global_gain = get_bits(gb, 8);
785 if (!common_window && !scale_flag) {
786 if (decode_ics_info(ac, ics, gb, 0) < 0)
790 if (decode_band_types(ac, sce->band_type, sce->band_type_run_end, gb, ics) < 0)
792 if (decode_scalefactors(ac, sce->sf, gb, global_gain, ics, sce->band_type, sce->band_type_run_end) < 0)
797 if ((pulse_present = get_bits1(gb))) {
798 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
799 av_log(ac->avccontext, AV_LOG_ERROR, "Pulse tool not allowed in eight short sequence.\n");
802 decode_pulses(&pulse, gb, ics->swb_offset);
804 if ((tns->present = get_bits1(gb)) && decode_tns(ac, tns, gb, ics))
807 av_log_missing_feature(ac->avccontext, "SSR", 1);
812 if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0)
818 * Mid/Side stereo decoding; reference: 4.6.8.1.3.
820 static void apply_mid_side_stereo(ChannelElement * cpe) {
821 const IndividualChannelStream * ics = &cpe->ch[0].ics;
822 float *ch0 = cpe->ch[0].coeffs;
823 float *ch1 = cpe->ch[1].coeffs;
824 int g, i, k, group, idx = 0;
825 const uint16_t * offsets = ics->swb_offset;
826 for (g = 0; g < ics->num_window_groups; g++) {
827 for (i = 0; i < ics->max_sfb; i++, idx++) {
828 if (cpe->ms_mask[idx] &&
829 cpe->ch[0].band_type[idx] < NOISE_BT && cpe->ch[1].band_type[idx] < NOISE_BT) {
830 for (group = 0; group < ics->group_len[g]; group++) {
831 for (k = offsets[i]; k < offsets[i+1]; k++) {
832 float tmp = ch0[group*128 + k] - ch1[group*128 + k];
833 ch0[group*128 + k] += ch1[group*128 + k];
834 ch1[group*128 + k] = tmp;
839 ch0 += ics->group_len[g]*128;
840 ch1 += ics->group_len[g]*128;
845 * intensity stereo decoding; reference: 4.6.8.2.3
847 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
848 * [1] mask is decoded from bitstream; [2] mask is all 1s;
849 * [3] reserved for scalable AAC
851 static void apply_intensity_stereo(ChannelElement * cpe, int ms_present) {
852 const IndividualChannelStream * ics = &cpe->ch[1].ics;
853 SingleChannelElement * sce1 = &cpe->ch[1];
854 float *coef0 = cpe->ch[0].coeffs, *coef1 = cpe->ch[1].coeffs;
855 const uint16_t * offsets = ics->swb_offset;
856 int g, group, i, k, idx = 0;
859 for (g = 0; g < ics->num_window_groups; g++) {
860 for (i = 0; i < ics->max_sfb;) {
861 if (sce1->band_type[idx] == INTENSITY_BT || sce1->band_type[idx] == INTENSITY_BT2) {
862 const int bt_run_end = sce1->band_type_run_end[idx];
863 for (; i < bt_run_end; i++, idx++) {
864 c = -1 + 2 * (sce1->band_type[idx] - 14);
866 c *= 1 - 2 * cpe->ms_mask[idx];
867 scale = c * sce1->sf[idx];
868 for (group = 0; group < ics->group_len[g]; group++)
869 for (k = offsets[i]; k < offsets[i+1]; k++)
870 coef1[group*128 + k] = scale * coef0[group*128 + k];
873 int bt_run_end = sce1->band_type_run_end[idx];
874 idx += bt_run_end - i;
878 coef0 += ics->group_len[g]*128;
879 coef1 += ics->group_len[g]*128;
884 * Decode a channel_pair_element; reference: table 4.4.
886 * @param elem_id Identifies the instance of a syntax element.
888 * @return Returns error status. 0 - OK, !0 - error
890 static int decode_cpe(AACContext * ac, GetBitContext * gb, int elem_id) {
891 int i, ret, common_window, ms_present = 0;
892 ChannelElement * cpe;
894 cpe = ac->che[TYPE_CPE][elem_id];
895 common_window = get_bits1(gb);
897 if (decode_ics_info(ac, &cpe->ch[0].ics, gb, 1))
899 i = cpe->ch[1].ics.use_kb_window[0];
900 cpe->ch[1].ics = cpe->ch[0].ics;
901 cpe->ch[1].ics.use_kb_window[1] = i;
902 ms_present = get_bits(gb, 2);
903 if(ms_present == 3) {
904 av_log(ac->avccontext, AV_LOG_ERROR, "ms_present = 3 is reserved.\n");
906 } else if(ms_present)
907 decode_mid_side_stereo(cpe, gb, ms_present);
909 if ((ret = decode_ics(ac, &cpe->ch[0], gb, common_window, 0)))
911 if ((ret = decode_ics(ac, &cpe->ch[1], gb, common_window, 0)))
914 if (common_window && ms_present)
915 apply_mid_side_stereo(cpe);
917 apply_intensity_stereo(cpe, ms_present);
922 * Decode coupling_channel_element; reference: table 4.8.
924 * @param elem_id Identifies the instance of a syntax element.
926 * @return Returns error status. 0 - OK, !0 - error
928 static int decode_cce(AACContext * ac, GetBitContext * gb, ChannelElement * che) {
930 int c, g, sfb, ret, idx = 0;
933 SingleChannelElement * sce = &che->ch[0];
934 ChannelCoupling * coup = &che->coup;
936 coup->coupling_point = 2*get_bits1(gb);
937 coup->num_coupled = get_bits(gb, 3);
938 for (c = 0; c <= coup->num_coupled; c++) {
940 coup->type[c] = get_bits1(gb) ? TYPE_CPE : TYPE_SCE;
941 coup->id_select[c] = get_bits(gb, 4);
942 if (coup->type[c] == TYPE_CPE) {
943 coup->ch_select[c] = get_bits(gb, 2);
944 if (coup->ch_select[c] == 3)
947 coup->ch_select[c] = 1;
949 coup->coupling_point += get_bits1(gb);
951 if (coup->coupling_point == 2) {
952 av_log(ac->avccontext, AV_LOG_ERROR,
953 "Independently switched CCE with 'invalid' domain signalled.\n");
954 memset(coup, 0, sizeof(ChannelCoupling));
958 sign = get_bits(gb, 1);
959 scale = pow(2., pow(2., get_bits(gb, 2) - 3));
961 if ((ret = decode_ics(ac, sce, gb, 0, 0)))
964 for (c = 0; c < num_gain; c++) {
967 float gain_cache = 1.;
969 cge = coup->coupling_point == AFTER_IMDCT ? 1 : get_bits1(gb);
970 gain = cge ? get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60: 0;
971 gain_cache = pow(scale, gain);
973 for (g = 0; g < sce->ics.num_window_groups; g++)
974 for (sfb = 0; sfb < sce->ics.max_sfb; sfb++, idx++)
975 if (sce->band_type[idx] != ZERO_BT) {
977 int t = get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
985 gain_cache = pow(scale, gain) * s;
988 coup->gain[c][idx] = gain_cache;
995 * Decode Spectral Band Replication extension data; reference: table 4.55.
997 * @param crc flag indicating the presence of CRC checksum
998 * @param cnt length of TYPE_FIL syntactic element in bytes
1000 * @return Returns number of bytes consumed from the TYPE_FIL element.
1002 static int decode_sbr_extension(AACContext * ac, GetBitContext * gb, int crc, int cnt) {
1003 // TODO : sbr_extension implementation
1004 av_log_missing_feature(ac->avccontext, "SBR", 0);
1005 skip_bits_long(gb, 8*cnt - 4); // -4 due to reading extension type
1010 * Parse whether channels are to be excluded from Dynamic Range Compression; reference: table 4.53.
1012 * @return Returns number of bytes consumed.
1014 static int decode_drc_channel_exclusions(DynamicRangeControl *che_drc, GetBitContext * gb) {
1016 int num_excl_chan = 0;
1019 for (i = 0; i < 7; i++)
1020 che_drc->exclude_mask[num_excl_chan++] = get_bits1(gb);
1021 } while (num_excl_chan < MAX_CHANNELS - 7 && get_bits1(gb));
1023 return num_excl_chan / 7;
1027 * Decode dynamic range information; reference: table 4.52.
1029 * @param cnt length of TYPE_FIL syntactic element in bytes
1031 * @return Returns number of bytes consumed.
1033 static int decode_dynamic_range(DynamicRangeControl *che_drc, GetBitContext * gb, int cnt) {
1035 int drc_num_bands = 1;
1038 /* pce_tag_present? */
1040 che_drc->pce_instance_tag = get_bits(gb, 4);
1041 skip_bits(gb, 4); // tag_reserved_bits
1045 /* excluded_chns_present? */
1047 n += decode_drc_channel_exclusions(che_drc, gb);
1050 /* drc_bands_present? */
1051 if (get_bits1(gb)) {
1052 che_drc->band_incr = get_bits(gb, 4);
1053 che_drc->interpolation_scheme = get_bits(gb, 4);
1055 drc_num_bands += che_drc->band_incr;
1056 for (i = 0; i < drc_num_bands; i++) {
1057 che_drc->band_top[i] = get_bits(gb, 8);
1062 /* prog_ref_level_present? */
1063 if (get_bits1(gb)) {
1064 che_drc->prog_ref_level = get_bits(gb, 7);
1065 skip_bits1(gb); // prog_ref_level_reserved_bits
1069 for (i = 0; i < drc_num_bands; i++) {
1070 che_drc->dyn_rng_sgn[i] = get_bits1(gb);
1071 che_drc->dyn_rng_ctl[i] = get_bits(gb, 7);
1079 * Decode extension data (incomplete); reference: table 4.51.
1081 * @param cnt length of TYPE_FIL syntactic element in bytes
1083 * @return Returns number of bytes consumed
1085 static int decode_extension_payload(AACContext * ac, GetBitContext * gb, int cnt) {
1088 switch (get_bits(gb, 4)) { // extension type
1089 case EXT_SBR_DATA_CRC:
1092 res = decode_sbr_extension(ac, gb, crc_flag, cnt);
1094 case EXT_DYNAMIC_RANGE:
1095 res = decode_dynamic_range(&ac->che_drc, gb, cnt);
1099 case EXT_DATA_ELEMENT:
1101 skip_bits_long(gb, 8*cnt - 4);
1108 * Decode Temporal Noise Shaping filter coefficients and apply all-pole filters; reference: 4.6.9.3.
1110 * @param decode 1 if tool is used normally, 0 if tool is used in LTP.
1111 * @param coef spectral coefficients
1113 static void apply_tns(float coef[1024], TemporalNoiseShaping * tns, IndividualChannelStream * ics, int decode) {
1114 const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
1115 int w, filt, m, i, ib;
1116 int bottom, top, order, start, end, size, inc;
1117 float lpc[TNS_MAX_ORDER];
1119 for (w = 0; w < ics->num_windows; w++) {
1120 bottom = ics->num_swb;
1121 for (filt = 0; filt < tns->n_filt[w]; filt++) {
1123 bottom = FFMAX(0, top - tns->length[w][filt]);
1124 order = tns->order[w][filt];
1129 * FIXME: This duplicates the functionality of some double code in lpc.c.
1131 for (m = 0; m < order; m++) {
1133 lpc[m] = tns->coef[w][filt][m];
1134 for (i = 0; i < m/2; i++) {
1136 lpc[i] += lpc[m] * lpc[m-1-i];
1137 lpc[m-1-i] += lpc[m] * tmp;
1140 lpc[i] += lpc[m] * lpc[i];
1143 start = ics->swb_offset[FFMIN(bottom, mmm)];
1144 end = ics->swb_offset[FFMIN( top, mmm)];
1145 if ((size = end - start) <= 0)
1147 if (tns->direction[w][filt]) {
1148 inc = -1; start = end - 1;
1155 for (m = 0; m < size; m++, start += inc)
1156 for (i = 1; i <= FFMIN(m, order); i++)
1157 coef[start] -= coef[start - i*inc] * lpc[i-1];
1163 * Conduct IMDCT and windowing.
1165 static void imdct_and_windowing(AACContext * ac, SingleChannelElement * sce) {
1166 IndividualChannelStream * ics = &sce->ics;
1167 float * in = sce->coeffs;
1168 float * out = sce->ret;
1169 float * saved = sce->saved;
1170 const float * lwindow = ics->use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
1171 const float * swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
1172 const float * lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
1173 const float * swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
1174 float * buf = ac->buf_mdct;
1177 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
1178 if (ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE)
1179 av_log(ac->avccontext, AV_LOG_WARNING,
1180 "Transition from an ONLY_LONG or LONG_STOP to an EIGHT_SHORT sequence detected. "
1181 "If you heard an audible artifact, please submit the sample to the FFmpeg developers.\n");
1182 for (i = 0; i < 2048; i += 256) {
1183 ff_imdct_calc(&ac->mdct_small, buf + i, in + i/2);
1184 ac->dsp.vector_fmul_reverse(ac->revers + i/2, buf + i + 128, swindow, 128);
1186 for (i = 0; i < 448; i++) out[i] = saved[i] + ac->add_bias;
1188 ac->dsp.vector_fmul_add_add(out + 448 + 0*128, buf + 0*128, swindow_prev, saved + 448 , ac->add_bias, 128, 1);
1189 ac->dsp.vector_fmul_add_add(out + 448 + 1*128, buf + 2*128, swindow, ac->revers + 0*128, ac->add_bias, 128, 1);
1190 ac->dsp.vector_fmul_add_add(out + 448 + 2*128, buf + 4*128, swindow, ac->revers + 1*128, ac->add_bias, 128, 1);
1191 ac->dsp.vector_fmul_add_add(out + 448 + 3*128, buf + 6*128, swindow, ac->revers + 2*128, ac->add_bias, 128, 1);
1192 ac->dsp.vector_fmul_add_add(out + 448 + 4*128, buf + 8*128, swindow, ac->revers + 3*128, ac->add_bias, 64, 1);
1195 vector_fmul_add_add_add(&ac->dsp, out + 448 + 1*128, buf + 2*128, swindow, saved + 448 + 1*128, ac->revers + 0*128, ac->add_bias, 128);
1196 vector_fmul_add_add_add(&ac->dsp, out + 448 + 2*128, buf + 4*128, swindow, saved + 448 + 2*128, ac->revers + 1*128, ac->add_bias, 128);
1197 vector_fmul_add_add_add(&ac->dsp, out + 448 + 3*128, buf + 6*128, swindow, saved + 448 + 3*128, ac->revers + 2*128, ac->add_bias, 128);
1198 vector_fmul_add_add_add(&ac->dsp, out + 448 + 4*128, buf + 8*128, swindow, saved + 448 + 4*128, ac->revers + 3*128, ac->add_bias, 64);
1201 ac->dsp.vector_fmul_add_add(saved, buf + 1024 + 64, swindow + 64, ac->revers + 3*128+64, 0, 64, 1);
1202 ac->dsp.vector_fmul_add_add(saved + 64, buf + 1024 + 2*128, swindow, ac->revers + 4*128, 0, 128, 1);
1203 ac->dsp.vector_fmul_add_add(saved + 192, buf + 1024 + 4*128, swindow, ac->revers + 5*128, 0, 128, 1);
1204 ac->dsp.vector_fmul_add_add(saved + 320, buf + 1024 + 6*128, swindow, ac->revers + 6*128, 0, 128, 1);
1205 memcpy( saved + 448, ac->revers + 7*128, 128 * sizeof(float));
1206 memset( saved + 576, 0, 448 * sizeof(float));
1208 ff_imdct_calc(&ac->mdct, buf, in);
1209 if (ics->window_sequence[0] == LONG_STOP_SEQUENCE) {
1210 for (i = 0; i < 448; i++) out[i] = saved[i] + ac->add_bias;
1211 ac->dsp.vector_fmul_add_add(out + 448, buf + 448, swindow_prev, saved + 448, ac->add_bias, 128, 1);
1212 for (i = 576; i < 1024; i++) out[i] = buf[i] + saved[i] + ac->add_bias;
1214 ac->dsp.vector_fmul_add_add(out, buf, lwindow_prev, saved, ac->add_bias, 1024, 1);
1216 if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
1217 memcpy(saved, buf + 1024, 448 * sizeof(float));
1218 ac->dsp.vector_fmul_reverse(saved + 448, buf + 1024 + 448, swindow, 128);
1219 memset(saved + 576, 0, 448 * sizeof(float));
1221 ac->dsp.vector_fmul_reverse(saved, buf + 1024, lwindow, 1024);
1227 * Apply dependent channel coupling (applied before IMDCT).
1229 * @param index index into coupling gain array
1231 static void apply_dependent_coupling(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index) {
1232 IndividualChannelStream * ics = &cc->ch[0].ics;
1233 const uint16_t * offsets = ics->swb_offset;
1234 float * dest = sce->coeffs;
1235 const float * src = cc->ch[0].coeffs;
1236 int g, i, group, k, idx = 0;
1237 if(ac->m4ac.object_type == AOT_AAC_LTP) {
1238 av_log(ac->avccontext, AV_LOG_ERROR,
1239 "Dependent coupling is not supported together with LTP\n");
1242 for (g = 0; g < ics->num_window_groups; g++) {
1243 for (i = 0; i < ics->max_sfb; i++, idx++) {
1244 if (cc->ch[0].band_type[idx] != ZERO_BT) {
1245 for (group = 0; group < ics->group_len[g]; group++) {
1246 for (k = offsets[i]; k < offsets[i+1]; k++) {
1248 dest[group*128+k] += cc->coup.gain[index][idx] * src[group*128+k];
1253 dest += ics->group_len[g]*128;
1254 src += ics->group_len[g]*128;
1259 * Apply independent channel coupling (applied after IMDCT).
1261 * @param index index into coupling gain array
1263 static void apply_independent_coupling(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index) {
1265 for (i = 0; i < 1024; i++)
1266 sce->ret[i] += cc->coup.gain[index][0] * (cc->ch[0].ret[i] - ac->add_bias);
1270 * channel coupling transformation interface
1272 * @param index index into coupling gain array
1273 * @param apply_coupling_method pointer to (in)dependent coupling function
1275 static void apply_channel_coupling(AACContext * ac, ChannelElement * cc,
1276 void (*apply_coupling_method)(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index))
1280 ChannelCoupling * coup = &cc->coup;
1281 for (c = 0; c <= coup->num_coupled; c++) {
1282 if (ac->che[coup->type[c]][coup->id_select[c]]) {
1283 if (coup->ch_select[c] != 2) {
1284 apply_coupling_method(ac, &ac->che[coup->type[c]][coup->id_select[c]]->ch[0], cc, index);
1285 if (coup->ch_select[c] != 0)
1288 if (coup->ch_select[c] != 1)
1289 apply_coupling_method(ac, &ac->che[coup->type[c]][coup->id_select[c]]->ch[1], cc, index++);
1291 av_log(ac->avccontext, AV_LOG_ERROR,
1292 "coupling target %sE[%d] not available\n",
1293 coup->type[c] == TYPE_CPE ? "CP" : "SC", coup->id_select[c]);
1300 * Convert spectral data to float samples, applying all supported tools as appropriate.
1302 static void spectral_to_sample(AACContext * ac) {
1304 for (i = 0; i < MAX_ELEM_ID; i++) {
1305 for(type = 0; type < 4; type++) {
1306 ChannelElement *che = ac->che[type][i];
1308 if(che->coup.coupling_point == BEFORE_TNS)
1309 apply_channel_coupling(ac, che, apply_dependent_coupling);
1310 if(che->ch[0].tns.present)
1311 apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1);
1312 if(che->ch[1].tns.present)
1313 apply_tns(che->ch[1].coeffs, &che->ch[1].tns, &che->ch[1].ics, 1);
1314 if(che->coup.coupling_point == BETWEEN_TNS_AND_IMDCT)
1315 apply_channel_coupling(ac, che, apply_dependent_coupling);
1316 imdct_and_windowing(ac, &che->ch[0]);
1317 if(type == TYPE_CPE)
1318 imdct_and_windowing(ac, &che->ch[1]);
1319 if(che->coup.coupling_point == AFTER_IMDCT)
1320 apply_channel_coupling(ac, che, apply_independent_coupling);
1326 static int aac_decode_frame(AVCodecContext * avccontext, void * data, int * data_size, const uint8_t * buf, int buf_size) {
1327 AACContext * ac = avccontext->priv_data;
1329 enum RawDataBlockType elem_type;
1330 int err, elem_id, data_size_tmp;
1332 init_get_bits(&gb, buf, buf_size*8);
1335 while ((elem_type = get_bits(&gb, 3)) != TYPE_END) {
1336 elem_id = get_bits(&gb, 4);
1339 if(elem_type == TYPE_SCE && elem_id == 1 &&
1340 !ac->che[TYPE_SCE][elem_id] && ac->che[TYPE_LFE][0]) {
1341 /* Some streams incorrectly code 5.1 audio as SCE[0] CPE[0] CPE[1] SCE[1]
1342 instead of SCE[0] CPE[0] CPE[0] LFE[0]. If we seem to have
1343 encountered such a stream, transfer the LFE[0] element to SCE[1] */
1344 ac->che[TYPE_SCE][elem_id] = ac->che[TYPE_LFE][0];
1345 ac->che[TYPE_LFE][0] = NULL;
1347 if(elem_type && elem_type < TYPE_DSE) {
1348 if(!ac->che[elem_type][elem_id])
1350 if(elem_type != TYPE_CCE)
1351 ac->che[elem_type][elem_id]->coup.coupling_point = 4;
1354 switch (elem_type) {
1357 err = decode_ics(ac, &ac->che[TYPE_SCE][elem_id]->ch[0], &gb, 0, 0);
1361 err = decode_cpe(ac, &gb, elem_id);
1365 err = decode_cce(ac, &gb, ac->che[TYPE_SCE][elem_id]);
1369 err = decode_ics(ac, &ac->che[TYPE_LFE][elem_id]->ch[0], &gb, 0, 0);
1373 skip_data_stream_element(&gb);
1379 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
1380 memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
1381 if((err = decode_pce(ac, new_che_pos, &gb)))
1383 err = output_configure(ac, ac->che_pos, new_che_pos);
1389 elem_id += get_bits(&gb, 8) - 1;
1391 elem_id -= decode_extension_payload(ac, &gb, elem_id);
1392 err = 0; /* FIXME */
1396 err = -1; /* should not happen, but keeps compiler happy */
1404 spectral_to_sample(ac);
1406 if (!ac->is_saved) {
1412 data_size_tmp = 1024 * avccontext->channels * sizeof(int16_t);
1413 if(*data_size < data_size_tmp) {
1414 av_log(avccontext, AV_LOG_ERROR,
1415 "Output buffer too small (%d) or trying to output too many samples (%d) for this frame.\n",
1416 *data_size, data_size_tmp);
1419 *data_size = data_size_tmp;
1421 ac->dsp.float_to_int16_interleave(data, (const float **)ac->output_data, 1024, avccontext->channels);
1426 static av_cold int aac_decode_close(AVCodecContext * avccontext) {
1427 AACContext * ac = avccontext->priv_data;
1430 for (i = 0; i < MAX_ELEM_ID; i++) {
1431 for(type = 0; type < 4; type++)
1432 av_freep(&ac->che[type][i]);
1435 ff_mdct_end(&ac->mdct);
1436 ff_mdct_end(&ac->mdct_small);
1440 AVCodec aac_decoder = {
1449 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
1450 .sample_fmts = (enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE},