cosmetics: K&R coding style
[libav.git] / libavcodec / aac.c
CommitLineData
71e9a1b8
RS
1/*
2 * AAC decoder
3 * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
4 * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
5 *
6 * This file is part of FFmpeg.
7 *
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.
12 *
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.
17 *
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
21 */
22
23/**
bad5537e 24 * @file libavcodec/aac.c
71e9a1b8
RS
25 * AAC decoder
26 * @author Oded Shimon ( ods15 ods15 dyndns org )
27 * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
28 */
29
30/*
31 * supported tools
32 *
33 * Support? Name
34 * N (code in SoC repo) gain control
35 * Y block switching
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
42 * Y intensity stereo
43 * Y channel coupling
7633a041 44 * Y frequency domain prediction
71e9a1b8
RS
45 * Y Perceptual Noise Substitution
46 * Y Mid/Side stereo
47 * N Scalable Inverse AAC Quantization
48 * N Frequency Selective Switch
49 * N upsampling filter
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
56 * N CELP
57 * N Silence Compression
58 * N HVXC
59 * N HVXC 4kbits/s VR
60 * N Structured Audio tools
61 * N Structured Audio Sample Bank Format
62 * N MIDI
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
72 *
73 * Note: - HE AAC v1 comprises LC AAC with Spectral Band Replication.
74 * - HE AAC v2 comprises LC AAC with Spectral Band Replication and
75 Parametric Stereo.
76 */
77
78
79#include "avcodec.h"
dbbec0c2 80#include "internal.h"
9106a698 81#include "get_bits.h"
71e9a1b8 82#include "dsputil.h"
1be0fc29 83#include "lpc.h"
71e9a1b8
RS
84
85#include "aac.h"
86#include "aactab.h"
cc0591da 87#include "aacdectab.h"
71e9a1b8 88#include "mpeg4audio.h"
158b3912 89#include "aac_parser.h"
71e9a1b8
RS
90
91#include <assert.h>
92#include <errno.h>
93#include <math.h>
94#include <string.h>
95
4a39ccb4
AC
96union float754 { float f; uint32_t i; };
97
71e9a1b8
RS
98static VLC vlc_scalefactors;
99static VLC vlc_spectral[11];
100
101
bb5c0988
AC
102static ChannelElement* get_che(AACContext *ac, int type, int elem_id) {
103 static const int8_t tags_per_config[16] = { 0, 1, 1, 2, 3, 3, 4, 5, 0, 0, 0, 0, 0, 0, 0, 0 };
104 if (ac->tag_che_map[type][elem_id]) {
105 return ac->tag_che_map[type][elem_id];
106 }
107 if (ac->tags_mapped >= tags_per_config[ac->m4ac.chan_config]) {
108 return NULL;
109 }
110 switch (ac->m4ac.chan_config) {
111 case 7:
112 if (ac->tags_mapped == 3 && type == TYPE_CPE) {
113 ac->tags_mapped++;
114 return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][2];
115 }
116 case 6:
117 /* Some streams incorrectly code 5.1 audio as SCE[0] CPE[0] CPE[1] SCE[1]
118 instead of SCE[0] CPE[0] CPE[0] LFE[0]. If we seem to have
119 encountered such a stream, transfer the LFE[0] element to SCE[1] */
120 if (ac->tags_mapped == tags_per_config[ac->m4ac.chan_config] - 1 && (type == TYPE_LFE || type == TYPE_SCE)) {
121 ac->tags_mapped++;
122 return ac->tag_che_map[type][elem_id] = ac->che[TYPE_LFE][0];
123 }
124 case 5:
125 if (ac->tags_mapped == 2 && type == TYPE_CPE) {
126 ac->tags_mapped++;
127 return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][1];
128 }
129 case 4:
130 if (ac->tags_mapped == 2 && ac->m4ac.chan_config == 4 && type == TYPE_SCE) {
131 ac->tags_mapped++;
132 return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][1];
133 }
134 case 3:
135 case 2:
136 if (ac->tags_mapped == (ac->m4ac.chan_config != 2) && type == TYPE_CPE) {
137 ac->tags_mapped++;
138 return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][0];
139 } else if (ac->m4ac.chan_config == 2) {
140 return NULL;
141 }
142 case 1:
143 if (!ac->tags_mapped && type == TYPE_SCE) {
144 ac->tags_mapped++;
145 return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][0];
146 }
147 default:
148 return NULL;
149 }
150}
151
9cc04edf 152/**
62a57fae
RS
153 * Configure output channel order based on the current program configuration element.
154 *
155 * @param che_pos current channel position configuration
156 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
157 *
158 * @return Returns error status. 0 - OK, !0 - error
159 */
160static int output_configure(AACContext *ac, enum ChannelPosition che_pos[4][MAX_ELEM_ID],
bb5c0988 161 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID], int channel_config) {
62a57fae
RS
162 AVCodecContext *avctx = ac->avccontext;
163 int i, type, channels = 0;
164
62a57fae
RS
165 memcpy(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
166
167 /* Allocate or free elements depending on if they are in the
168 * current program configuration.
169 *
170 * Set up default 1:1 output mapping.
171 *
172 * For a 5.1 stream the output order will be:
f3399088 173 * [ Center ] [ Front Left ] [ Front Right ] [ LFE ] [ Surround Left ] [ Surround Right ]
62a57fae
RS
174 */
175
176 for(i = 0; i < MAX_ELEM_ID; i++) {
177 for(type = 0; type < 4; type++) {
178 if(che_pos[type][i]) {
179 if(!ac->che[type][i] && !(ac->che[type][i] = av_mallocz(sizeof(ChannelElement))))
180 return AVERROR(ENOMEM);
181 if(type != TYPE_CCE) {
182 ac->output_data[channels++] = ac->che[type][i]->ch[0].ret;
183 if(type == TYPE_CPE) {
184 ac->output_data[channels++] = ac->che[type][i]->ch[1].ret;
185 }
186 }
187 } else
188 av_freep(&ac->che[type][i]);
189 }
190 }
191
bb5c0988
AC
192 if (channel_config) {
193 memset(ac->tag_che_map, 0, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));
194 ac->tags_mapped = 0;
195 } else {
196 memcpy(ac->tag_che_map, ac->che, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));
197 ac->tags_mapped = 4*MAX_ELEM_ID;
198 }
199
62a57fae 200 avctx->channels = channels;
bb5c0988 201
6308765c
AC
202 ac->output_configured = 1;
203
62a57fae
RS
204 return 0;
205}
206
207/**
9cc04edf
RS
208 * Decode an array of 4 bit element IDs, optionally interleaved with a stereo/mono switching bit.
209 *
210 * @param cpe_map Stereo (Channel Pair Element) map, NULL if stereo bit is not present.
211 * @param sce_map mono (Single Channel Element) map
212 * @param type speaker type/position for these channels
213 */
214static void decode_channel_map(enum ChannelPosition *cpe_map,
215 enum ChannelPosition *sce_map, enum ChannelPosition type, GetBitContext * gb, int n) {
216 while(n--) {
217 enum ChannelPosition *map = cpe_map && get_bits1(gb) ? cpe_map : sce_map; // stereo or mono map
218 map[get_bits(gb, 4)] = type;
219 }
220}
221
222/**
223 * Decode program configuration element; reference: table 4.2.
224 *
225 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
226 *
227 * @return Returns error status. 0 - OK, !0 - error
228 */
229static int decode_pce(AACContext * ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
230 GetBitContext * gb) {
99665a21 231 int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc, sampling_index;
9cc04edf
RS
232
233 skip_bits(gb, 2); // object_type
234
99665a21 235 sampling_index = get_bits(gb, 4);
401a9950
AC
236 if (ac->m4ac.sampling_index != sampling_index)
237 av_log(ac->avccontext, AV_LOG_WARNING, "Sample rate index in program config element does not match the sample rate index configured by the container.\n");
238
71e9a1b8
RS
239 num_front = get_bits(gb, 4);
240 num_side = get_bits(gb, 4);
241 num_back = get_bits(gb, 4);
242 num_lfe = get_bits(gb, 2);
243 num_assoc_data = get_bits(gb, 3);
244 num_cc = get_bits(gb, 4);
245
cc0591da
RS
246 if (get_bits1(gb))
247 skip_bits(gb, 4); // mono_mixdown_tag
248 if (get_bits1(gb))
249 skip_bits(gb, 4); // stereo_mixdown_tag
71e9a1b8 250
cc0591da
RS
251 if (get_bits1(gb))
252 skip_bits(gb, 3); // mixdown_coeff_index and pseudo_surround
71e9a1b8 253
cc0591da
RS
254 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_FRONT, gb, num_front);
255 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_SIDE, gb, num_side );
256 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_BACK, gb, num_back );
257 decode_channel_map(NULL, new_che_pos[TYPE_LFE], AAC_CHANNEL_LFE, gb, num_lfe );
71e9a1b8
RS
258
259 skip_bits_long(gb, 4 * num_assoc_data);
260
cc0591da 261 decode_channel_map(new_che_pos[TYPE_CCE], new_che_pos[TYPE_CCE], AAC_CHANNEL_CC, gb, num_cc );
71e9a1b8
RS
262
263 align_get_bits(gb);
264
265 /* comment field, first byte is length */
266 skip_bits_long(gb, 8 * get_bits(gb, 8));
cc0591da
RS
267 return 0;
268}
71e9a1b8 269
9cc04edf
RS
270/**
271 * Set up channel positions based on a default channel configuration
272 * as specified in table 1.17.
273 *
274 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
275 *
276 * @return Returns error status. 0 - OK, !0 - error
277 */
278static int set_default_channel_config(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
279 int channel_config)
280{
281 if(channel_config < 1 || channel_config > 7) {
282 av_log(ac->avccontext, AV_LOG_ERROR, "invalid default channel configuration (%d)\n",
283 channel_config);
284 return -1;
285 }
286
287 /* default channel configurations:
288 *
289 * 1ch : front center (mono)
290 * 2ch : L + R (stereo)
291 * 3ch : front center + L + R
292 * 4ch : front center + L + R + back center
293 * 5ch : front center + L + R + back stereo
294 * 6ch : front center + L + R + back stereo + LFE
295 * 7ch : front center + L + R + outer front left + outer front right + back stereo + LFE
296 */
297
298 if(channel_config != 2)
299 new_che_pos[TYPE_SCE][0] = AAC_CHANNEL_FRONT; // front center (or mono)
300 if(channel_config > 1)
301 new_che_pos[TYPE_CPE][0] = AAC_CHANNEL_FRONT; // L + R (or stereo)
302 if(channel_config == 4)
303 new_che_pos[TYPE_SCE][1] = AAC_CHANNEL_BACK; // back center
304 if(channel_config > 4)
305 new_che_pos[TYPE_CPE][(channel_config == 7) + 1]
306 = AAC_CHANNEL_BACK; // back stereo
307 if(channel_config > 5)
308 new_che_pos[TYPE_LFE][0] = AAC_CHANNEL_LFE; // LFE
309 if(channel_config == 7)
310 new_che_pos[TYPE_CPE][1] = AAC_CHANNEL_FRONT; // outer front left + outer front right
311
312 return 0;
313}
314
62a57fae
RS
315/**
316 * Decode GA "General Audio" specific configuration; reference: table 4.1.
317 *
318 * @return Returns error status. 0 - OK, !0 - error
319 */
320static int decode_ga_specific_config(AACContext * ac, GetBitContext * gb, int channel_config) {
321 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
322 int extension_flag, ret;
323
324 if(get_bits1(gb)) { // frameLengthFlag
ce863d7f 325 av_log_missing_feature(ac->avccontext, "960/120 MDCT window is", 1);
9cc04edf
RS
326 return -1;
327 }
328
329 if (get_bits1(gb)) // dependsOnCoreCoder
330 skip_bits(gb, 14); // coreCoderDelay
331 extension_flag = get_bits1(gb);
332
333 if(ac->m4ac.object_type == AOT_AAC_SCALABLE ||
334 ac->m4ac.object_type == AOT_ER_AAC_SCALABLE)
335 skip_bits(gb, 3); // layerNr
336
337 memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
338 if (channel_config == 0) {
339 skip_bits(gb, 4); // element_instance_tag
340 if((ret = decode_pce(ac, new_che_pos, gb)))
341 return ret;
342 } else {
343 if((ret = set_default_channel_config(ac, new_che_pos, channel_config)))
344 return ret;
345 }
bb5c0988 346 if((ret = output_configure(ac, ac->che_pos, new_che_pos, channel_config)))
9cc04edf
RS
347 return ret;
348
349 if (extension_flag) {
350 switch (ac->m4ac.object_type) {
351 case AOT_ER_BSAC:
352 skip_bits(gb, 5); // numOfSubFrame
353 skip_bits(gb, 11); // layer_length
354 break;
355 case AOT_ER_AAC_LC:
356 case AOT_ER_AAC_LTP:
357 case AOT_ER_AAC_SCALABLE:
358 case AOT_ER_AAC_LD:
359 skip_bits(gb, 3); /* aacSectionDataResilienceFlag
360 * aacScalefactorDataResilienceFlag
361 * aacSpectralDataResilienceFlag
362 */
363 break;
364 }
365 skip_bits1(gb); // extensionFlag3 (TBD in version 3)
366 }
367 return 0;
368}
369
370/**
371 * Decode audio specific configuration; reference: table 1.13.
372 *
373 * @param data pointer to AVCodecContext extradata
374 * @param data_size size of AVCCodecContext extradata
375 *
376 * @return Returns error status. 0 - OK, !0 - error
377 */
378static int decode_audio_specific_config(AACContext * ac, void *data, int data_size) {
379 GetBitContext gb;
380 int i;
381
382 init_get_bits(&gb, data, data_size * 8);
383
384 if((i = ff_mpeg4audio_get_config(&ac->m4ac, data, data_size)) < 0)
385 return -1;
30272450 386 if(ac->m4ac.sampling_index > 12) {
9cc04edf
RS
387 av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
388 return -1;
389 }
390
391 skip_bits_long(&gb, i);
392
393 switch (ac->m4ac.object_type) {
7633a041 394 case AOT_AAC_MAIN:
9cc04edf
RS
395 case AOT_AAC_LC:
396 if (decode_ga_specific_config(ac, &gb, ac->m4ac.chan_config))
397 return -1;
398 break;
399 default:
400 av_log(ac->avccontext, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n",
401 ac->m4ac.sbr == 1? "SBR+" : "", ac->m4ac.object_type);
402 return -1;
403 }
404 return 0;
405}
406
62a57fae
RS
407/**
408 * linear congruential pseudorandom number generator
409 *
410 * @param previous_val pointer to the current state of the generator
411 *
412 * @return Returns a 32-bit pseudorandom integer
413 */
414static av_always_inline int lcg_random(int previous_val) {
415 return previous_val * 1664525 + 1013904223;
416}
417
7633a041
AC
418static void reset_predict_state(PredictorState * ps) {
419 ps->r0 = 0.0f;
420 ps->r1 = 0.0f;
421 ps->cor0 = 0.0f;
422 ps->cor1 = 0.0f;
423 ps->var0 = 1.0f;
424 ps->var1 = 1.0f;
425}
426
427static void reset_all_predictors(PredictorState * ps) {
428 int i;
429 for (i = 0; i < MAX_PREDICTORS; i++)
430 reset_predict_state(&ps[i]);
431}
432
433static void reset_predictor_group(PredictorState * ps, int group_num) {
434 int i;
435 for (i = group_num-1; i < MAX_PREDICTORS; i+=30)
436 reset_predict_state(&ps[i]);
437}
438
71e9a1b8
RS
439static av_cold int aac_decode_init(AVCodecContext * avccontext) {
440 AACContext * ac = avccontext->priv_data;
441 int i;
442
443 ac->avccontext = avccontext;
444
158b3912
RS
445 if (avccontext->extradata_size > 0) {
446 if(decode_audio_specific_config(ac, avccontext->extradata, avccontext->extradata_size))
447 return -1;
448 avccontext->sample_rate = ac->m4ac.sample_rate;
449 } else if (avccontext->channels > 0) {
158b3912 450 ac->m4ac.sample_rate = avccontext->sample_rate;
158b3912 451 }
cc0591da 452
9cc04edf 453 avccontext->sample_fmt = SAMPLE_FMT_S16;
71e9a1b8
RS
454 avccontext->frame_size = 1024;
455
456 AAC_INIT_VLC_STATIC( 0, 144);
457 AAC_INIT_VLC_STATIC( 1, 114);
458 AAC_INIT_VLC_STATIC( 2, 188);
459 AAC_INIT_VLC_STATIC( 3, 180);
460 AAC_INIT_VLC_STATIC( 4, 172);
461 AAC_INIT_VLC_STATIC( 5, 140);
462 AAC_INIT_VLC_STATIC( 6, 168);
463 AAC_INIT_VLC_STATIC( 7, 114);
464 AAC_INIT_VLC_STATIC( 8, 262);
465 AAC_INIT_VLC_STATIC( 9, 248);
466 AAC_INIT_VLC_STATIC(10, 384);
467
468 dsputil_init(&ac->dsp, avccontext);
469
9cc04edf
RS
470 ac->random_state = 0x1f2e3d4c;
471
71e9a1b8
RS
472 // -1024 - Compensate wrong IMDCT method.
473 // 32768 - Required to scale values to the correct range for the bias method
474 // for float to int16 conversion.
475
476 if(ac->dsp.float_to_int16 == ff_float_to_int16_c) {
477 ac->add_bias = 385.0f;
478 ac->sf_scale = 1. / (-1024. * 32768.);
479 ac->sf_offset = 0;
480 } else {
481 ac->add_bias = 0.0f;
482 ac->sf_scale = 1. / -1024.;
483 ac->sf_offset = 60;
484 }
485
b250f9c6 486#if !CONFIG_HARDCODED_TABLES
d0ee5021 487 for (i = 0; i < 428; i++)
cc0591da 488 ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.);
71e9a1b8
RS
489#endif /* CONFIG_HARDCODED_TABLES */
490
37d3e066 491 INIT_VLC_STATIC(&vlc_scalefactors,7,FF_ARRAY_ELEMS(ff_aac_scalefactor_code),
71e9a1b8
RS
492 ff_aac_scalefactor_bits, sizeof(ff_aac_scalefactor_bits[0]), sizeof(ff_aac_scalefactor_bits[0]),
493 ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]),
494 352);
495
7d485f16
SS
496 ff_mdct_init(&ac->mdct, 11, 1, 1.0);
497 ff_mdct_init(&ac->mdct_small, 8, 1, 1.0);
9ffd5c1c
RS
498 // window initialization
499 ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
500 ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
501 ff_sine_window_init(ff_sine_1024, 1024);
502 ff_sine_window_init(ff_sine_128, 128);
503
71e9a1b8
RS
504 return 0;
505}
506
9cc04edf
RS
507/**
508 * Skip data_stream_element; reference: table 4.10.
509 */
510static void skip_data_stream_element(GetBitContext * gb) {
71e9a1b8
RS
511 int byte_align = get_bits1(gb);
512 int count = get_bits(gb, 8);
513 if (count == 255)
514 count += get_bits(gb, 8);
515 if (byte_align)
516 align_get_bits(gb);
517 skip_bits_long(gb, 8 * count);
518}
519
7633a041
AC
520static int decode_prediction(AACContext * ac, IndividualChannelStream * ics, GetBitContext * gb) {
521 int sfb;
522 if (get_bits1(gb)) {
523 ics->predictor_reset_group = get_bits(gb, 5);
524 if (ics->predictor_reset_group == 0 || ics->predictor_reset_group > 30) {
525 av_log(ac->avccontext, AV_LOG_ERROR, "Invalid Predictor Reset Group.\n");
526 return -1;
527 }
528 }
529 for (sfb = 0; sfb < FFMIN(ics->max_sfb, ff_aac_pred_sfb_max[ac->m4ac.sampling_index]); sfb++) {
530 ics->prediction_used[sfb] = get_bits1(gb);
531 }
532 return 0;
533}
534
71e9a1b8 535/**
9cc04edf
RS
536 * Decode Individual Channel Stream info; reference: table 4.6.
537 *
538 * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
539 */
540static int decode_ics_info(AACContext * ac, IndividualChannelStream * ics, GetBitContext * gb, int common_window) {
541 if (get_bits1(gb)) {
542 av_log(ac->avccontext, AV_LOG_ERROR, "Reserved bit set.\n");
543 memset(ics, 0, sizeof(IndividualChannelStream));
544 return -1;
545 }
546 ics->window_sequence[1] = ics->window_sequence[0];
547 ics->window_sequence[0] = get_bits(gb, 2);
548 ics->use_kb_window[1] = ics->use_kb_window[0];
549 ics->use_kb_window[0] = get_bits1(gb);
550 ics->num_window_groups = 1;
551 ics->group_len[0] = 1;
9ffd5c1c
RS
552 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
553 int i;
554 ics->max_sfb = get_bits(gb, 4);
555 for (i = 0; i < 7; i++) {
556 if (get_bits1(gb)) {
557 ics->group_len[ics->num_window_groups-1]++;
558 } else {
559 ics->num_window_groups++;
560 ics->group_len[ics->num_window_groups-1] = 1;
561 }
562 }
563 ics->num_windows = 8;
d4eeef0c
AC
564 ics->swb_offset = ff_swb_offset_128[ac->m4ac.sampling_index];
565 ics->num_swb = ff_aac_num_swb_128[ac->m4ac.sampling_index];
85e9296f 566 ics->tns_max_bands = ff_tns_max_bands_128[ac->m4ac.sampling_index];
7633a041 567 ics->predictor_present = 0;
9ffd5c1c
RS
568 } else {
569 ics->max_sfb = get_bits(gb, 6);
570 ics->num_windows = 1;
d4eeef0c
AC
571 ics->swb_offset = ff_swb_offset_1024[ac->m4ac.sampling_index];
572 ics->num_swb = ff_aac_num_swb_1024[ac->m4ac.sampling_index];
85e9296f 573 ics->tns_max_bands = ff_tns_max_bands_1024[ac->m4ac.sampling_index];
7633a041
AC
574 ics->predictor_present = get_bits1(gb);
575 ics->predictor_reset_group = 0;
576 if (ics->predictor_present) {
577 if (ac->m4ac.object_type == AOT_AAC_MAIN) {
578 if (decode_prediction(ac, ics, gb)) {
579 memset(ics, 0, sizeof(IndividualChannelStream));
580 return -1;
581 }
582 } else if (ac->m4ac.object_type == AOT_AAC_LC) {
583 av_log(ac->avccontext, AV_LOG_ERROR, "Prediction is not allowed in AAC-LC.\n");
584 memset(ics, 0, sizeof(IndividualChannelStream));
585 return -1;
586 } else {
ce863d7f 587 av_log_missing_feature(ac->avccontext, "Predictor bit set but LTP is", 1);
8f5aaa6d
RS
588 memset(ics, 0, sizeof(IndividualChannelStream));
589 return -1;
7633a041 590 }
62a57fae
RS
591 }
592 }
593
594 if(ics->max_sfb > ics->num_swb) {
595 av_log(ac->avccontext, AV_LOG_ERROR,
596 "Number of scalefactor bands in group (%d) exceeds limit (%d).\n",
597 ics->max_sfb, ics->num_swb);
598 memset(ics, 0, sizeof(IndividualChannelStream));
599 return -1;
600 }
601
9cc04edf
RS
602 return 0;
603}
604
605/**
9cc04edf
RS
606 * Decode band types (section_data payload); reference: table 4.46.
607 *
608 * @param band_type array of the used band type
609 * @param band_type_run_end array of the last scalefactor band of a band type run
610 *
611 * @return Returns error status. 0 - OK, !0 - error
612 */
613static int decode_band_types(AACContext * ac, enum BandType band_type[120],
cc0591da
RS
614 int band_type_run_end[120], GetBitContext * gb, IndividualChannelStream * ics) {
615 int g, idx = 0;
616 const int bits = (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) ? 3 : 5;
617 for (g = 0; g < ics->num_window_groups; g++) {
618 int k = 0;
619 while (k < ics->max_sfb) {
620 uint8_t sect_len = k;
621 int sect_len_incr;
622 int sect_band_type = get_bits(gb, 4);
623 if (sect_band_type == 12) {
624 av_log(ac->avccontext, AV_LOG_ERROR, "invalid band type\n");
625 return -1;
626 }
627 while ((sect_len_incr = get_bits(gb, bits)) == (1 << bits)-1)
628 sect_len += sect_len_incr;
629 sect_len += sect_len_incr;
630 if (sect_len > ics->max_sfb) {
631 av_log(ac->avccontext, AV_LOG_ERROR,
632 "Number of bands (%d) exceeds limit (%d).\n",
633 sect_len, ics->max_sfb);
634 return -1;
635 }
9ffd5c1c
RS
636 for (; k < sect_len; k++) {
637 band_type [idx] = sect_band_type;
638 band_type_run_end[idx++] = sect_len;
639 }
9cc04edf
RS
640 }
641 }
642 return 0;
643}
cc0591da 644
9cc04edf
RS
645/**
646 * Decode scalefactors; reference: table 4.47.
cc0591da 647 *
cc0591da
RS
648 * @param global_gain first scalefactor value as scalefactors are differentially coded
649 * @param band_type array of the used band type
650 * @param band_type_run_end array of the last scalefactor band of a band type run
651 * @param sf array of scalefactors or intensity stereo positions
652 *
653 * @return Returns error status. 0 - OK, !0 - error
654 */
655static int decode_scalefactors(AACContext * ac, float sf[120], GetBitContext * gb,
9edae4ad 656 unsigned int global_gain, IndividualChannelStream * ics,
cc0591da
RS
657 enum BandType band_type[120], int band_type_run_end[120]) {
658 const int sf_offset = ac->sf_offset + (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE ? 12 : 0);
659 int g, i, idx = 0;
660 int offset[3] = { global_gain, global_gain - 90, 100 };
661 int noise_flag = 1;
662 static const char *sf_str[3] = { "Global gain", "Noise gain", "Intensity stereo position" };
cc0591da
RS
663 for (g = 0; g < ics->num_window_groups; g++) {
664 for (i = 0; i < ics->max_sfb;) {
665 int run_end = band_type_run_end[idx];
666 if (band_type[idx] == ZERO_BT) {
667 for(; i < run_end; i++, idx++)
668 sf[idx] = 0.;
669 }else if((band_type[idx] == INTENSITY_BT) || (band_type[idx] == INTENSITY_BT2)) {
cc0591da
RS
670 for(; i < run_end; i++, idx++) {
671 offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
672 if(offset[2] > 255U) {
673 av_log(ac->avccontext, AV_LOG_ERROR,
674 "%s (%d) out of range.\n", sf_str[2], offset[2]);
675 return -1;
676 }
677 sf[idx] = ff_aac_pow2sf_tab[-offset[2] + 300];
cc0591da
RS
678 }
679 }else if(band_type[idx] == NOISE_BT) {
680 for(; i < run_end; i++, idx++) {
681 if(noise_flag-- > 0)
682 offset[1] += get_bits(gb, 9) - 256;
683 else
684 offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
685 if(offset[1] > 255U) {
686 av_log(ac->avccontext, AV_LOG_ERROR,
687 "%s (%d) out of range.\n", sf_str[1], offset[1]);
688 return -1;
689 }
d0ee5021 690 sf[idx] = -ff_aac_pow2sf_tab[ offset[1] + sf_offset + 100];
cc0591da
RS
691 }
692 }else {
693 for(; i < run_end; i++, idx++) {
694 offset[0] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
695 if(offset[0] > 255U) {
696 av_log(ac->avccontext, AV_LOG_ERROR,
697 "%s (%d) out of range.\n", sf_str[0], offset[0]);
698 return -1;
699 }
700 sf[idx] = -ff_aac_pow2sf_tab[ offset[0] + sf_offset];
cc0591da
RS
701 }
702 }
703 }
704 }
705 return 0;
706}
707
708/**
709 * Decode pulse data; reference: table 4.7.
710 */
aac0eda4
AC
711static int decode_pulses(Pulse * pulse, GetBitContext * gb, const uint16_t * swb_offset, int num_swb) {
712 int i, pulse_swb;
cc0591da 713 pulse->num_pulse = get_bits(gb, 2) + 1;
aac0eda4
AC
714 pulse_swb = get_bits(gb, 6);
715 if (pulse_swb >= num_swb)
716 return -1;
717 pulse->pos[0] = swb_offset[pulse_swb];
408992ba 718 pulse->pos[0] += get_bits(gb, 5);
aac0eda4
AC
719 if (pulse->pos[0] > 1023)
720 return -1;
848a5815
RS
721 pulse->amp[0] = get_bits(gb, 4);
722 for (i = 1; i < pulse->num_pulse; i++) {
723 pulse->pos[i] = get_bits(gb, 5) + pulse->pos[i-1];
aac0eda4
AC
724 if (pulse->pos[i] > 1023)
725 return -1;
848a5815 726 pulse->amp[i] = get_bits(gb, 4);
cc0591da 727 }
aac0eda4 728 return 0;
cc0591da
RS
729}
730
731/**
1dece0d2
RS
732 * Decode Temporal Noise Shaping data; reference: table 4.48.
733 *
734 * @return Returns error status. 0 - OK, !0 - error
735 */
736static int decode_tns(AACContext * ac, TemporalNoiseShaping * tns,
737 GetBitContext * gb, const IndividualChannelStream * ics) {
738 int w, filt, i, coef_len, coef_res, coef_compress;
739 const int is8 = ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE;
740 const int tns_max_order = is8 ? 7 : ac->m4ac.object_type == AOT_AAC_MAIN ? 20 : 12;
741 for (w = 0; w < ics->num_windows; w++) {
fbd91d7c 742 if ((tns->n_filt[w] = get_bits(gb, 2 - is8))) {
1dece0d2
RS
743 coef_res = get_bits1(gb);
744
65b20b24
RS
745 for (filt = 0; filt < tns->n_filt[w]; filt++) {
746 int tmp2_idx;
747 tns->length[w][filt] = get_bits(gb, 6 - 2*is8);
748
749 if ((tns->order[w][filt] = get_bits(gb, 5 - 2*is8)) > tns_max_order) {
750 av_log(ac->avccontext, AV_LOG_ERROR, "TNS filter order %d is greater than maximum %d.",
751 tns->order[w][filt], tns_max_order);
752 tns->order[w][filt] = 0;
753 return -1;
754 }
51673647 755 if (tns->order[w][filt]) {
35445d29
RS
756 tns->direction[w][filt] = get_bits1(gb);
757 coef_compress = get_bits1(gb);
758 coef_len = coef_res + 3 - coef_compress;
759 tmp2_idx = 2*coef_compress + coef_res;
1dece0d2 760
35445d29
RS
761 for (i = 0; i < tns->order[w][filt]; i++)
762 tns->coef[w][filt][i] = tns_tmp2_map[tmp2_idx][get_bits(gb, coef_len)];
51673647 763 }
65b20b24 764 }
fbd91d7c 765 }
1dece0d2
RS
766 }
767 return 0;
768}
769
770/**
9cc04edf
RS
771 * Decode Mid/Side data; reference: table 4.54.
772 *
773 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
774 * [1] mask is decoded from bitstream; [2] mask is all 1s;
775 * [3] reserved for scalable AAC
776 */
777static void decode_mid_side_stereo(ChannelElement * cpe, GetBitContext * gb,
778 int ms_present) {
62a57fae
RS
779 int idx;
780 if (ms_present == 1) {
781 for (idx = 0; idx < cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb; idx++)
782 cpe->ms_mask[idx] = get_bits1(gb);
783 } else if (ms_present == 2) {
784 memset(cpe->ms_mask, 1, cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb * sizeof(cpe->ms_mask[0]));
785 }
786}
9cc04edf
RS
787
788/**
9ffd5c1c
RS
789 * Decode spectral data; reference: table 4.50.
790 * Dequantize and scale spectral data; reference: 4.6.3.3.
791 *
792 * @param coef array of dequantized, scaled spectral data
793 * @param sf array of scalefactors or intensity stereo positions
794 * @param pulse_present set if pulses are present
795 * @param pulse pointer to pulse data struct
796 * @param band_type array of the used band type
797 *
798 * @return Returns error status. 0 - OK, !0 - error
799 */
800static int decode_spectrum_and_dequant(AACContext * ac, float coef[1024], GetBitContext * gb, float sf[120],
801 int pulse_present, const Pulse * pulse, const IndividualChannelStream * ics, enum BandType band_type[120]) {
802 int i, k, g, idx = 0;
803 const int c = 1024/ics->num_windows;
804 const uint16_t * offsets = ics->swb_offset;
805 float *coef_base = coef;
4a904903 806 static const float sign_lookup[] = { 1.0f, -1.0f };
9ffd5c1c
RS
807
808 for (g = 0; g < ics->num_windows; g++)
809 memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float)*(c - offsets[ics->max_sfb]));
810
811 for (g = 0; g < ics->num_window_groups; g++) {
812 for (i = 0; i < ics->max_sfb; i++, idx++) {
813 const int cur_band_type = band_type[idx];
814 const int dim = cur_band_type >= FIRST_PAIR_BT ? 2 : 4;
815 const int is_cb_unsigned = IS_CODEBOOK_UNSIGNED(cur_band_type);
816 int group;
aab54133 817 if (cur_band_type == ZERO_BT || cur_band_type == INTENSITY_BT2 || cur_band_type == INTENSITY_BT) {
9ffd5c1c
RS
818 for (group = 0; group < ics->group_len[g]; group++) {
819 memset(coef + group * 128 + offsets[i], 0, (offsets[i+1] - offsets[i])*sizeof(float));
820 }
821 }else if (cur_band_type == NOISE_BT) {
9ffd5c1c 822 for (group = 0; group < ics->group_len[g]; group++) {
d0ee5021
AC
823 float scale;
824 float band_energy = 0;
9ffd5c1c
RS
825 for (k = offsets[i]; k < offsets[i+1]; k++) {
826 ac->random_state = lcg_random(ac->random_state);
d0ee5021
AC
827 coef[group*128+k] = ac->random_state;
828 band_energy += coef[group*128+k]*coef[group*128+k];
829 }
830 scale = sf[idx] / sqrtf(band_energy);
831 for (k = offsets[i]; k < offsets[i+1]; k++) {
832 coef[group*128+k] *= scale;
9ffd5c1c
RS
833 }
834 }
aab54133 835 }else {
9ffd5c1c
RS
836 for (group = 0; group < ics->group_len[g]; group++) {
837 for (k = offsets[i]; k < offsets[i+1]; k += dim) {
838 const int index = get_vlc2(gb, vlc_spectral[cur_band_type - 1].table, 6, 3);
839 const int coef_tmp_idx = (group << 7) + k;
840 const float *vq_ptr;
841 int j;
842 if(index >= ff_aac_spectral_sizes[cur_band_type - 1]) {
843 av_log(ac->avccontext, AV_LOG_ERROR,
844 "Read beyond end of ff_aac_codebook_vectors[%d][]. index %d >= %d\n",
845 cur_band_type - 1, index, ff_aac_spectral_sizes[cur_band_type - 1]);
846 return -1;
847 }
848 vq_ptr = &ff_aac_codebook_vectors[cur_band_type - 1][index * dim];
849 if (is_cb_unsigned) {
4a904903
AC
850 if (vq_ptr[0]) coef[coef_tmp_idx ] = sign_lookup[get_bits1(gb)];
851 if (vq_ptr[1]) coef[coef_tmp_idx + 1] = sign_lookup[get_bits1(gb)];
c0893c3a 852 if (dim == 4) {
4a904903
AC
853 if (vq_ptr[2]) coef[coef_tmp_idx + 2] = sign_lookup[get_bits1(gb)];
854 if (vq_ptr[3]) coef[coef_tmp_idx + 3] = sign_lookup[get_bits1(gb)];
c0893c3a 855 }
e8d5c07b
AC
856 if (cur_band_type == ESC_BT) {
857 for (j = 0; j < 2; j++) {
858 if (vq_ptr[j] == 64.0f) {
859 int n = 4;
860 /* The total length of escape_sequence must be < 22 bits according
861 to the specification (i.e. max is 11111111110xxxxxxxxxx). */
862 while (get_bits1(gb) && n < 15) n++;
863 if(n == 15) {
864 av_log(ac->avccontext, AV_LOG_ERROR, "error in spectral data, ESC overflow\n");
865 return -1;
866 }
867 n = (1<<n) + get_bits(gb, n);
868 coef[coef_tmp_idx + j] *= cbrtf(n) * n;
869 }else
870 coef[coef_tmp_idx + j] *= vq_ptr[j];
871 }
872 }else
873 {
874 coef[coef_tmp_idx ] *= vq_ptr[0];
875 coef[coef_tmp_idx + 1] *= vq_ptr[1];
876 if (dim == 4) {
877 coef[coef_tmp_idx + 2] *= vq_ptr[2];
878 coef[coef_tmp_idx + 3] *= vq_ptr[3];
879 }
9ffd5c1c 880 }
21e8c46c
AC
881 }else {
882 coef[coef_tmp_idx ] = vq_ptr[0];
883 coef[coef_tmp_idx + 1] = vq_ptr[1];
884 if (dim == 4) {
885 coef[coef_tmp_idx + 2] = vq_ptr[2];
886 coef[coef_tmp_idx + 3] = vq_ptr[3];
887 }
888 }
c0893c3a
AC
889 coef[coef_tmp_idx ] *= sf[idx];
890 coef[coef_tmp_idx + 1] *= sf[idx];
891 if (dim == 4) {
892 coef[coef_tmp_idx + 2] *= sf[idx];
893 coef[coef_tmp_idx + 3] *= sf[idx];
894 }
9ffd5c1c
RS
895 }
896 }
897 }
898 }
899 coef += ics->group_len[g]<<7;
900 }
901
902 if (pulse_present) {
51436848 903 idx = 0;
9ffd5c1c
RS
904 for(i = 0; i < pulse->num_pulse; i++){
905 float co = coef_base[ pulse->pos[i] ];
51436848
AC
906 while(offsets[idx + 1] <= pulse->pos[i])
907 idx++;
908 if (band_type[idx] != NOISE_BT && sf[idx]) {
70735a3f
RS
909 float ico = -pulse->amp[i];
910 if (co) {
911 co /= sf[idx];
912 ico = co / sqrtf(sqrtf(fabsf(co))) + (co > 0 ? -ico : ico);
913 }
914 coef_base[ pulse->pos[i] ] = cbrtf(fabsf(ico)) * ico * sf[idx];
51436848 915 }
9ffd5c1c
RS
916 }
917 }
918 return 0;
919}
920
7633a041 921static av_always_inline float flt16_round(float pf) {
4a39ccb4
AC
922 union float754 tmp;
923 tmp.f = pf;
924 tmp.i = (tmp.i + 0x00008000U) & 0xFFFF0000U;
925 return tmp.f;
7633a041
AC
926}
927
928static av_always_inline float flt16_even(float pf) {
4a39ccb4
AC
929 union float754 tmp;
930 tmp.f = pf;
931 tmp.i = (tmp.i + 0x00007FFFU + (tmp.i & 0x00010000U>>16)) & 0xFFFF0000U;
932 return tmp.f;
7633a041
AC
933}
934
935static av_always_inline float flt16_trunc(float pf) {
4a39ccb4
AC
936 union float754 pun;
937 pun.f = pf;
938 pun.i &= 0xFFFF0000U;
939 return pun.f;
7633a041
AC
940}
941
942static void predict(AACContext * ac, PredictorState * ps, float* coef, int output_enable) {
943 const float a = 0.953125; // 61.0/64
944 const float alpha = 0.90625; // 29.0/32
945 float e0, e1;
946 float pv;
947 float k1, k2;
948
949 k1 = ps->var0 > 1 ? ps->cor0 * flt16_even(a / ps->var0) : 0;
950 k2 = ps->var1 > 1 ? ps->cor1 * flt16_even(a / ps->var1) : 0;
951
952 pv = flt16_round(k1 * ps->r0 + k2 * ps->r1);
953 if (output_enable)
954 *coef += pv * ac->sf_scale;
955
956 e0 = *coef / ac->sf_scale;
957 e1 = e0 - k1 * ps->r0;
958
959 ps->cor1 = flt16_trunc(alpha * ps->cor1 + ps->r1 * e1);
960 ps->var1 = flt16_trunc(alpha * ps->var1 + 0.5 * (ps->r1 * ps->r1 + e1 * e1));
961 ps->cor0 = flt16_trunc(alpha * ps->cor0 + ps->r0 * e0);
962 ps->var0 = flt16_trunc(alpha * ps->var0 + 0.5 * (ps->r0 * ps->r0 + e0 * e0));
963
964 ps->r1 = flt16_trunc(a * (ps->r0 - k1 * e0));
965 ps->r0 = flt16_trunc(a * e0);
966}
967
968/**
969 * Apply AAC-Main style frequency domain prediction.
970 */
971static void apply_prediction(AACContext * ac, SingleChannelElement * sce) {
972 int sfb, k;
973
974 if (!sce->ics.predictor_initialized) {
aab54133 975 reset_all_predictors(sce->predictor_state);
7633a041
AC
976 sce->ics.predictor_initialized = 1;
977 }
978
979 if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
980 for (sfb = 0; sfb < ff_aac_pred_sfb_max[ac->m4ac.sampling_index]; sfb++) {
981 for (k = sce->ics.swb_offset[sfb]; k < sce->ics.swb_offset[sfb + 1]; k++) {
aab54133 982 predict(ac, &sce->predictor_state[k], &sce->coeffs[k],
7633a041
AC
983 sce->ics.predictor_present && sce->ics.prediction_used[sfb]);
984 }
985 }
986 if (sce->ics.predictor_reset_group)
aab54133 987 reset_predictor_group(sce->predictor_state, sce->ics.predictor_reset_group);
7633a041 988 } else
aab54133 989 reset_all_predictors(sce->predictor_state);
7633a041
AC
990}
991
9ffd5c1c 992/**
9cc04edf
RS
993 * Decode an individual_channel_stream payload; reference: table 4.44.
994 *
995 * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
996 * @param scale_flag scalable [1] or non-scalable [0] AAC (Unused until scalable AAC is implemented.)
997 *
998 * @return Returns error status. 0 - OK, !0 - error
999 */
1000static int decode_ics(AACContext * ac, SingleChannelElement * sce, GetBitContext * gb, int common_window, int scale_flag) {
9cc04edf
RS
1001 Pulse pulse;
1002 TemporalNoiseShaping * tns = &sce->tns;
1003 IndividualChannelStream * ics = &sce->ics;
1004 float * out = sce->coeffs;
1005 int global_gain, pulse_present = 0;
1006
848a5815
RS
1007 /* This assignment is to silence a GCC warning about the variable being used
1008 * uninitialized when in fact it always is.
9cc04edf
RS
1009 */
1010 pulse.num_pulse = 0;
9cc04edf
RS
1011
1012 global_gain = get_bits(gb, 8);
1013
1014 if (!common_window && !scale_flag) {
1015 if (decode_ics_info(ac, ics, gb, 0) < 0)
1016 return -1;
1017 }
1018
1019 if (decode_band_types(ac, sce->band_type, sce->band_type_run_end, gb, ics) < 0)
1020 return -1;
1021 if (decode_scalefactors(ac, sce->sf, gb, global_gain, ics, sce->band_type, sce->band_type_run_end) < 0)
1022 return -1;
1023
1024 pulse_present = 0;
1025 if (!scale_flag) {
1026 if ((pulse_present = get_bits1(gb))) {
1027 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
1028 av_log(ac->avccontext, AV_LOG_ERROR, "Pulse tool not allowed in eight short sequence.\n");
1029 return -1;
1030 }
aac0eda4
AC
1031 if (decode_pulses(&pulse, gb, ics->swb_offset, ics->num_swb)) {
1032 av_log(ac->avccontext, AV_LOG_ERROR, "Pulse data corrupt or invalid.\n");
1033 return -1;
1034 }
9cc04edf
RS
1035 }
1036 if ((tns->present = get_bits1(gb)) && decode_tns(ac, tns, gb, ics))
1037 return -1;
1038 if (get_bits1(gb)) {
ce863d7f 1039 av_log_missing_feature(ac->avccontext, "SSR", 1);
9cc04edf
RS
1040 return -1;
1041 }
1042 }
1043
848a5815 1044 if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0)
9cc04edf 1045 return -1;
7633a041 1046
aab54133 1047 if(ac->m4ac.object_type == AOT_AAC_MAIN && !common_window)
7633a041
AC
1048 apply_prediction(ac, sce);
1049
9cc04edf
RS
1050 return 0;
1051}
1052
1053/**
9ffd5c1c
RS
1054 * Mid/Side stereo decoding; reference: 4.6.8.1.3.
1055 */
1056static void apply_mid_side_stereo(ChannelElement * cpe) {
1057 const IndividualChannelStream * ics = &cpe->ch[0].ics;
1058 float *ch0 = cpe->ch[0].coeffs;
1059 float *ch1 = cpe->ch[1].coeffs;
1060 int g, i, k, group, idx = 0;
1061 const uint16_t * offsets = ics->swb_offset;
1062 for (g = 0; g < ics->num_window_groups; g++) {
1063 for (i = 0; i < ics->max_sfb; i++, idx++) {
1064 if (cpe->ms_mask[idx] &&
1065 cpe->ch[0].band_type[idx] < NOISE_BT && cpe->ch[1].band_type[idx] < NOISE_BT) {
1066 for (group = 0; group < ics->group_len[g]; group++) {
1067 for (k = offsets[i]; k < offsets[i+1]; k++) {
1068 float tmp = ch0[group*128 + k] - ch1[group*128 + k];
1069 ch0[group*128 + k] += ch1[group*128 + k];
1070 ch1[group*128 + k] = tmp;
1071 }
1072 }
1073 }
1074 }
1075 ch0 += ics->group_len[g]*128;
1076 ch1 += ics->group_len[g]*128;
1077 }
1078}
1079
1080/**
1081 * intensity stereo decoding; reference: 4.6.8.2.3
1082 *
1083 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
1084 * [1] mask is decoded from bitstream; [2] mask is all 1s;
1085 * [3] reserved for scalable AAC
1086 */
1087static void apply_intensity_stereo(ChannelElement * cpe, int ms_present) {
1088 const IndividualChannelStream * ics = &cpe->ch[1].ics;
1089 SingleChannelElement * sce1 = &cpe->ch[1];
1090 float *coef0 = cpe->ch[0].coeffs, *coef1 = cpe->ch[1].coeffs;
1091 const uint16_t * offsets = ics->swb_offset;
1092 int g, group, i, k, idx = 0;
1093 int c;
1094 float scale;
1095 for (g = 0; g < ics->num_window_groups; g++) {
1096 for (i = 0; i < ics->max_sfb;) {
1097 if (sce1->band_type[idx] == INTENSITY_BT || sce1->band_type[idx] == INTENSITY_BT2) {
1098 const int bt_run_end = sce1->band_type_run_end[idx];
1099 for (; i < bt_run_end; i++, idx++) {
1100 c = -1 + 2 * (sce1->band_type[idx] - 14);
1101 if (ms_present)
1102 c *= 1 - 2 * cpe->ms_mask[idx];
1103 scale = c * sce1->sf[idx];
1104 for (group = 0; group < ics->group_len[g]; group++)
1105 for (k = offsets[i]; k < offsets[i+1]; k++)
1106 coef1[group*128 + k] = scale * coef0[group*128 + k];
1107 }
1108 } else {
1109 int bt_run_end = sce1->band_type_run_end[idx];
1110 idx += bt_run_end - i;
1111 i = bt_run_end;
1112 }
1113 }
1114 coef0 += ics->group_len[g]*128;
1115 coef1 += ics->group_len[g]*128;
1116 }
1117}
1118
1119/**
9cc04edf
RS
1120 * Decode a channel_pair_element; reference: table 4.4.
1121 *
1122 * @param elem_id Identifies the instance of a syntax element.
1123 *
1124 * @return Returns error status. 0 - OK, !0 - error
1125 */
9ef88afa 1126static int decode_cpe(AACContext * ac, GetBitContext * gb, ChannelElement * cpe) {
9cc04edf 1127 int i, ret, common_window, ms_present = 0;
9cc04edf 1128
9cc04edf
RS
1129 common_window = get_bits1(gb);
1130 if (common_window) {
1131 if (decode_ics_info(ac, &cpe->ch[0].ics, gb, 1))
1132 return -1;
1133 i = cpe->ch[1].ics.use_kb_window[0];
1134 cpe->ch[1].ics = cpe->ch[0].ics;
1135 cpe->ch[1].ics.use_kb_window[1] = i;
1136 ms_present = get_bits(gb, 2);
1137 if(ms_present == 3) {
1138 av_log(ac->avccontext, AV_LOG_ERROR, "ms_present = 3 is reserved.\n");
1139 return -1;
1140 } else if(ms_present)
1141 decode_mid_side_stereo(cpe, gb, ms_present);
1142 }
1143 if ((ret = decode_ics(ac, &cpe->ch[0], gb, common_window, 0)))
1144 return ret;
1145 if ((ret = decode_ics(ac, &cpe->ch[1], gb, common_window, 0)))
1146 return ret;
1147
aab54133
AC
1148 if (common_window) {
1149 if (ms_present)
897c1703 1150 apply_mid_side_stereo(cpe);
aab54133
AC
1151 if (ac->m4ac.object_type == AOT_AAC_MAIN) {
1152 apply_prediction(ac, &cpe->ch[0]);
1153 apply_prediction(ac, &cpe->ch[1]);
1154 }
1155 }
9cc04edf 1156
848a5815 1157 apply_intensity_stereo(cpe, ms_present);
9cc04edf
RS
1158 return 0;
1159}
1160
9ffd5c1c
RS
1161/**
1162 * Decode coupling_channel_element; reference: table 4.8.
1163 *
1164 * @param elem_id Identifies the instance of a syntax element.
1165 *
1166 * @return Returns error status. 0 - OK, !0 - error
1167 */
1168static int decode_cce(AACContext * ac, GetBitContext * gb, ChannelElement * che) {
1169 int num_gain = 0;
341b28c0 1170 int c, g, sfb, ret;
9ffd5c1c
RS
1171 int sign;
1172 float scale;
1173 SingleChannelElement * sce = &che->ch[0];
1174 ChannelCoupling * coup = &che->coup;
1175
62a57fae
RS
1176 coup->coupling_point = 2*get_bits1(gb);
1177 coup->num_coupled = get_bits(gb, 3);
1178 for (c = 0; c <= coup->num_coupled; c++) {
1179 num_gain++;
1180 coup->type[c] = get_bits1(gb) ? TYPE_CPE : TYPE_SCE;
1181 coup->id_select[c] = get_bits(gb, 4);
1182 if (coup->type[c] == TYPE_CPE) {
1183 coup->ch_select[c] = get_bits(gb, 2);
1184 if (coup->ch_select[c] == 3)
1185 num_gain++;
1186 } else
88de95c2 1187 coup->ch_select[c] = 2;
62a57fae 1188 }
04964ee8 1189 coup->coupling_point += get_bits1(gb) || (coup->coupling_point>>1);
62a57fae
RS
1190
1191 sign = get_bits(gb, 1);
c8947a56 1192 scale = pow(2., pow(2., (int)get_bits(gb, 2) - 3));
62a57fae
RS
1193
1194 if ((ret = decode_ics(ac, sce, gb, 0, 0)))
1195 return ret;
1196
1197 for (c = 0; c < num_gain; c++) {
341b28c0 1198 int idx = 0;
62a57fae
RS
1199 int cge = 1;
1200 int gain = 0;
1201 float gain_cache = 1.;
1202 if (c) {
1203 cge = coup->coupling_point == AFTER_IMDCT ? 1 : get_bits1(gb);
1204 gain = cge ? get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60: 0;
88de95c2 1205 gain_cache = pow(scale, -gain);
62a57fae 1206 }
f1ade11e
AC
1207 if (coup->coupling_point == AFTER_IMDCT) {
1208 coup->gain[c][0] = gain_cache;
1209 } else {
03b12747
AC
1210 for (g = 0; g < sce->ics.num_window_groups; g++) {
1211 for (sfb = 0; sfb < sce->ics.max_sfb; sfb++, idx++) {
1212 if (sce->band_type[idx] != ZERO_BT) {
1213 if (!cge) {
1214 int t = get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
1215 if (t) {
1216 int s = 1;
1217 t = gain += t;
1218 if (sign) {
1219 s -= 2 * (t & 0x1);
1220 t >>= 1;
1221 }
1222 gain_cache = pow(scale, -t) * s;
62a57fae 1223 }
62a57fae 1224 }
03b12747 1225 coup->gain[c][idx] = gain_cache;
62a57fae 1226 }
62a57fae 1227 }
f80a8ca5
RS
1228 }
1229 }
62a57fae
RS
1230 }
1231 return 0;
1232}
1233
9cc04edf
RS
1234/**
1235 * Decode Spectral Band Replication extension data; reference: table 4.55.
cc0591da
RS
1236 *
1237 * @param crc flag indicating the presence of CRC checksum
1238 * @param cnt length of TYPE_FIL syntactic element in bytes
9cc04edf 1239 *
cc0591da
RS
1240 * @return Returns number of bytes consumed from the TYPE_FIL element.
1241 */
1242static int decode_sbr_extension(AACContext * ac, GetBitContext * gb, int crc, int cnt) {
1243 // TODO : sbr_extension implementation
ce863d7f 1244 av_log_missing_feature(ac->avccontext, "SBR", 0);
cc0591da
RS
1245 skip_bits_long(gb, 8*cnt - 4); // -4 due to reading extension type
1246 return cnt;
1247}
1248
9cc04edf 1249/**
62a57fae
RS
1250 * Parse whether channels are to be excluded from Dynamic Range Compression; reference: table 4.53.
1251 *
1252 * @return Returns number of bytes consumed.
1253 */
1254static int decode_drc_channel_exclusions(DynamicRangeControl *che_drc, GetBitContext * gb) {
1255 int i;
1256 int num_excl_chan = 0;
1257
1258 do {
1259 for (i = 0; i < 7; i++)
1260 che_drc->exclude_mask[num_excl_chan++] = get_bits1(gb);
1261 } while (num_excl_chan < MAX_CHANNELS - 7 && get_bits1(gb));
1262
1263 return num_excl_chan / 7;
1264}
1265
1266/**
9cc04edf
RS
1267 * Decode dynamic range information; reference: table 4.52.
1268 *
1269 * @param cnt length of TYPE_FIL syntactic element in bytes
1270 *
1271 * @return Returns number of bytes consumed.
1272 */
1273static int decode_dynamic_range(DynamicRangeControl *che_drc, GetBitContext * gb, int cnt) {
1274 int n = 1;
1275 int drc_num_bands = 1;
1276 int i;
1277
1278 /* pce_tag_present? */
1279 if(get_bits1(gb)) {
1280 che_drc->pce_instance_tag = get_bits(gb, 4);
1281 skip_bits(gb, 4); // tag_reserved_bits
1282 n++;
1283 }
1284
1285 /* excluded_chns_present? */
1286 if(get_bits1(gb)) {
1287 n += decode_drc_channel_exclusions(che_drc, gb);
1288 }
1289
1290 /* drc_bands_present? */
1291 if (get_bits1(gb)) {
1292 che_drc->band_incr = get_bits(gb, 4);
1293 che_drc->interpolation_scheme = get_bits(gb, 4);
1294 n++;
1295 drc_num_bands += che_drc->band_incr;
1296 for (i = 0; i < drc_num_bands; i++) {
1297 che_drc->band_top[i] = get_bits(gb, 8);
1298 n++;
1299 }
1300 }
1301
1302 /* prog_ref_level_present? */
1303 if (get_bits1(gb)) {
1304 che_drc->prog_ref_level = get_bits(gb, 7);
1305 skip_bits1(gb); // prog_ref_level_reserved_bits
1306 n++;
1307 }
1308
1309 for (i = 0; i < drc_num_bands; i++) {
1310 che_drc->dyn_rng_sgn[i] = get_bits1(gb);
1311 che_drc->dyn_rng_ctl[i] = get_bits(gb, 7);
1312 n++;
1313 }
1314
1315 return n;
1316}
1317
1318/**
1319 * Decode extension data (incomplete); reference: table 4.51.
1320 *
1321 * @param cnt length of TYPE_FIL syntactic element in bytes
1322 *
1323 * @return Returns number of bytes consumed
1324 */
1325static int decode_extension_payload(AACContext * ac, GetBitContext * gb, int cnt) {
cc0591da
RS
1326 int crc_flag = 0;
1327 int res = cnt;
1328 switch (get_bits(gb, 4)) { // extension type
1329 case EXT_SBR_DATA_CRC:
1330 crc_flag++;
1331 case EXT_SBR_DATA:
1332 res = decode_sbr_extension(ac, gb, crc_flag, cnt);
1333 break;
1334 case EXT_DYNAMIC_RANGE:
1335 res = decode_dynamic_range(&ac->che_drc, gb, cnt);
1336 break;
1337 case EXT_FILL:
1338 case EXT_FILL_DATA:
1339 case EXT_DATA_ELEMENT:
1340 default:
1341 skip_bits_long(gb, 8*cnt - 4);
1342 break;
1343 };
1344 return res;
1345}
1346
7d8f3de4
RS
1347/**
1348 * Decode Temporal Noise Shaping filter coefficients and apply all-pole filters; reference: 4.6.9.3.
1349 *
1350 * @param decode 1 if tool is used normally, 0 if tool is used in LTP.
1351 * @param coef spectral coefficients
1352 */
1353static void apply_tns(float coef[1024], TemporalNoiseShaping * tns, IndividualChannelStream * ics, int decode) {
1354 const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
1098e8d2 1355 int w, filt, m, i;
7d8f3de4
RS
1356 int bottom, top, order, start, end, size, inc;
1357 float lpc[TNS_MAX_ORDER];
1358
1359 for (w = 0; w < ics->num_windows; w++) {
1360 bottom = ics->num_swb;
1361 for (filt = 0; filt < tns->n_filt[w]; filt++) {
1362 top = bottom;
1363 bottom = FFMAX(0, top - tns->length[w][filt]);
1364 order = tns->order[w][filt];
1365 if (order == 0)
1366 continue;
1367
1be0fc29
VS
1368 // tns_decode_coef
1369 compute_lpc_coefs(tns->coef[w][filt], order, lpc, 0, 0, 0);
7d8f3de4 1370
1dece0d2
RS
1371 start = ics->swb_offset[FFMIN(bottom, mmm)];
1372 end = ics->swb_offset[FFMIN( top, mmm)];
1373 if ((size = end - start) <= 0)
1374 continue;
1375 if (tns->direction[w][filt]) {
1376 inc = -1; start = end - 1;
1377 } else {
1378 inc = 1;
1379 }
1380 start += w * 128;
1381
1382 // ar filter
1383 for (m = 0; m < size; m++, start += inc)
1384 for (i = 1; i <= FFMIN(m, order); i++)
7d8f3de4 1385 coef[start] -= coef[start - i*inc] * lpc[i-1];
1dece0d2
RS
1386 }
1387 }
1388}
1389
cc0591da 1390/**
9cc04edf
RS
1391 * Conduct IMDCT and windowing.
1392 */
1393static void imdct_and_windowing(AACContext * ac, SingleChannelElement * sce) {
1394 IndividualChannelStream * ics = &sce->ics;
1395 float * in = sce->coeffs;
1396 float * out = sce->ret;
1397 float * saved = sce->saved;
848a5815
RS
1398 const float * swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
1399 const float * lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
1400 const float * swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
9cc04edf 1401 float * buf = ac->buf_mdct;
8153f144 1402 float * temp = ac->temp;
9cc04edf
RS
1403 int i;
1404
f4990558 1405 // imdct
62a57fae
RS
1406 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
1407 if (ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE)
1408 av_log(ac->avccontext, AV_LOG_WARNING,
1409 "Transition from an ONLY_LONG or LONG_STOP to an EIGHT_SHORT sequence detected. "
1410 "If you heard an audible artifact, please submit the sample to the FFmpeg developers.\n");
b0f5852a
RS
1411 for (i = 0; i < 1024; i += 128)
1412 ff_imdct_half(&ac->mdct_small, buf + i, in + i);
f4990558 1413 } else
b0f5852a 1414 ff_imdct_half(&ac->mdct, buf, in);
f4990558
RS
1415
1416 /* window overlapping
1417 * NOTE: To simplify the overlapping code, all 'meaningless' short to long
1418 * and long to short transitions are considered to be short to short
1419 * transitions. This leaves just two cases (long to long and short to short)
1420 * with a little special sauce for EIGHT_SHORT_SEQUENCE.
1421 */
1422 if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
1423 (ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
b0f5852a 1424 ac->dsp.vector_fmul_window( out, saved, buf, lwindow_prev, ac->add_bias, 512);
f4990558 1425 } else {
db38c386
RS
1426 for (i = 0; i < 448; i++)
1427 out[i] = saved[i] + ac->add_bias;
62a57fae 1428
f4990558 1429 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
b0f5852a
RS
1430 ac->dsp.vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, ac->add_bias, 64);
1431 ac->dsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, ac->add_bias, 64);
1432 ac->dsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, ac->add_bias, 64);
1433 ac->dsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, ac->add_bias, 64);
1434 ac->dsp.vector_fmul_window(temp, buf + 3*128 + 64, buf + 4*128, swindow, ac->add_bias, 64);
1435 memcpy( out + 448 + 4*128, temp, 64 * sizeof(float));
f4990558 1436 } else {
b0f5852a 1437 ac->dsp.vector_fmul_window(out + 448, saved + 448, buf, swindow_prev, ac->add_bias, 64);
db38c386 1438 for (i = 576; i < 1024; i++)
b0f5852a 1439 out[i] = buf[i-512] + ac->add_bias;
f4990558
RS
1440 }
1441 }
62a57fae 1442
f4990558
RS
1443 // buffer update
1444 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
b0f5852a
RS
1445 for (i = 0; i < 64; i++)
1446 saved[i] = temp[64 + i] - ac->add_bias;
1447 ac->dsp.vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 0, 64);
1448 ac->dsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 0, 64);
1449 ac->dsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 0, 64);
1450 memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(float));
f4990558 1451 } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
b0f5852a
RS
1452 memcpy( saved, buf + 512, 448 * sizeof(float));
1453 memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(float));
f4990558 1454 } else { // LONG_STOP or ONLY_LONG
b0f5852a 1455 memcpy( saved, buf + 512, 512 * sizeof(float));
62a57fae
RS
1456 }
1457}
1458
9cc04edf 1459/**
cc0591da
RS
1460 * Apply dependent channel coupling (applied before IMDCT).
1461 *
1462 * @param index index into coupling gain array
1463 */
fbdae895
AC
1464static void apply_dependent_coupling(AACContext * ac, SingleChannelElement * target, ChannelElement * cce, int index) {
1465 IndividualChannelStream * ics = &cce->ch[0].ics;
cc0591da 1466 const uint16_t * offsets = ics->swb_offset;
fbdae895
AC
1467 float * dest = target->coeffs;
1468 const float * src = cce->ch[0].coeffs;
cc0591da
RS
1469 int g, i, group, k, idx = 0;
1470 if(ac->m4ac.object_type == AOT_AAC_LTP) {
1471 av_log(ac->avccontext, AV_LOG_ERROR,
1472 "Dependent coupling is not supported together with LTP\n");
1473 return;
1474 }
1475 for (g = 0; g < ics->num_window_groups; g++) {
1476 for (i = 0; i < ics->max_sfb; i++, idx++) {
fbdae895 1477 if (cce->ch[0].band_type[idx] != ZERO_BT) {
cfd937b0 1478 const float gain = cce->coup.gain[index][idx];
cc0591da
RS
1479 for (group = 0; group < ics->group_len[g]; group++) {
1480 for (k = offsets[i]; k < offsets[i+1]; k++) {
1481 // XXX dsputil-ize
cfd937b0 1482 dest[group*128+k] += gain * src[group*128+k];
cc0591da
RS
1483 }
1484 }
1485 }
1486 }
1487 dest += ics->group_len[g]*128;
1488 src += ics->group_len[g]*128;
1489 }
1490}
1491
1492/**
1493 * Apply independent channel coupling (applied after IMDCT).
1494 *
1495 * @param index index into coupling gain array
1496 */
fbdae895 1497static void apply_independent_coupling(AACContext * ac, SingleChannelElement * target, ChannelElement * cce, int index) {
cc0591da 1498 int i;
039821a8
AC
1499 const float gain = cce->coup.gain[index][0];
1500 const float bias = ac->add_bias;
1501 const float* src = cce->ch[0].ret;
1502 float* dest = target->ret;
1503
cc0591da 1504 for (i = 0; i < 1024; i++)
039821a8 1505 dest[i] += gain * (src[i] - bias);
cc0591da
RS
1506}
1507
9ffd5c1c
RS
1508/**
1509 * channel coupling transformation interface
1510 *
1511 * @param index index into coupling gain array
1512 * @param apply_coupling_method pointer to (in)dependent coupling function
1513 */
1514static void apply_channel_coupling(AACContext * ac, ChannelElement * cc,
88de95c2 1515 enum RawDataBlockType type, int elem_id, enum CouplingPoint coupling_point,
fbdae895 1516 void (*apply_coupling_method)(AACContext * ac, SingleChannelElement * target, ChannelElement * cce, int index))
9ffd5c1c 1517{
88de95c2
AC
1518 int i, c;
1519
1520 for (i = 0; i < MAX_ELEM_ID; i++) {
1521 ChannelElement *cce = ac->che[TYPE_CCE][i];
1522 int index = 0;
1523
1524 if (cce && cce->coup.coupling_point == coupling_point) {
1525 ChannelCoupling * coup = &cce->coup;
1526
1527 for (c = 0; c <= coup->num_coupled; c++) {
1528 if (coup->type[c] == type && coup->id_select[c] == elem_id) {
1529 if (coup->ch_select[c] != 1) {
1530 apply_coupling_method(ac, &cc->ch[0], cce, index);
1531 if (coup->ch_select[c] != 0)
1532 index++;
1533 }
1534 if (coup->ch_select[c] != 2)
1535 apply_coupling_method(ac, &cc->ch[1], cce, index++);
1536 } else
1537 index += 1 + (coup->ch_select[c] == 3);
9ffd5c1c 1538 }
9ffd5c1c
RS
1539 }
1540 }
1541}
1542
1543/**
1544 * Convert spectral data to float samples, applying all supported tools as appropriate.
1545 */
1546static void spectral_to_sample(AACContext * ac) {
6735e2c8 1547 int i, type;
88de95c2
AC
1548 for(type = 3; type >= 0; type--) {
1549 for (i = 0; i < MAX_ELEM_ID; i++) {
9ffd5c1c
RS
1550 ChannelElement *che = ac->che[type][i];
1551 if(che) {
88de95c2
AC
1552 if(type <= TYPE_CPE)
1553 apply_channel_coupling(ac, che, type, i, BEFORE_TNS, apply_dependent_coupling);
9ffd5c1c
RS
1554 if(che->ch[0].tns.present)
1555 apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1);
1556 if(che->ch[1].tns.present)
1557 apply_tns(che->ch[1].coeffs, &che->ch[1].tns, &che->ch[1].ics, 1);
88de95c2
AC
1558 if(type <= TYPE_CPE)
1559 apply_channel_coupling(ac, che, type, i, BETWEEN_TNS_AND_IMDCT, apply_dependent_coupling);
1560 if(type != TYPE_CCE || che->coup.coupling_point == AFTER_IMDCT)
1561 imdct_and_windowing(ac, &che->ch[0]);
9ffd5c1c
RS
1562 if(type == TYPE_CPE)
1563 imdct_and_windowing(ac, &che->ch[1]);
88de95c2
AC
1564 if(type <= TYPE_CCE)
1565 apply_channel_coupling(ac, che, type, i, AFTER_IMDCT, apply_independent_coupling);
62a57fae
RS
1566 }
1567 }
1568 }
1569}
1570
158b3912
RS
1571static int parse_adts_frame_header(AACContext * ac, GetBitContext * gb) {
1572
1573 int size;
1574 AACADTSHeaderInfo hdr_info;
1575
1576 size = ff_aac_parse_header(gb, &hdr_info);
1577 if (size > 0) {
6308765c
AC
1578 if (!ac->output_configured && hdr_info.chan_config) {
1579 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
1580 memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
158b3912 1581 ac->m4ac.chan_config = hdr_info.chan_config;
6308765c
AC
1582 if (set_default_channel_config(ac, new_che_pos, hdr_info.chan_config))
1583 return -7;
1584 if (output_configure(ac, ac->che_pos, new_che_pos, 1))
1585 return -7;
1586 }
158b3912
RS
1587 ac->m4ac.sample_rate = hdr_info.sample_rate;
1588 ac->m4ac.sampling_index = hdr_info.sampling_index;
1589 ac->m4ac.object_type = hdr_info.object_type;
7d87e2ce
AC
1590 if (hdr_info.num_aac_frames == 1) {
1591 if (!hdr_info.crc_absent)
1592 skip_bits(gb, 16);
1593 } else {
ce863d7f 1594 av_log_missing_feature(ac->avccontext, "More than one AAC RDB per ADTS frame is", 0);
7d87e2ce
AC
1595 return -1;
1596 }
51741a82 1597 }
158b3912
RS
1598 return size;
1599}
1600
7a00bbad
TB
1601static int aac_decode_frame(AVCodecContext * avccontext, void * data, int * data_size, AVPacket *avpkt) {
1602 const uint8_t *buf = avpkt->data;
1603 int buf_size = avpkt->size;
62a57fae 1604 AACContext * ac = avccontext->priv_data;
bb5c0988 1605 ChannelElement * che = NULL;
62a57fae
RS
1606 GetBitContext gb;
1607 enum RawDataBlockType elem_type;
1608 int err, elem_id, data_size_tmp;
1609
1610 init_get_bits(&gb, buf, buf_size*8);
1611
158b3912 1612 if (show_bits(&gb, 12) == 0xfff) {
5967e141 1613 if (parse_adts_frame_header(ac, &gb) < 0) {
158b3912
RS
1614 av_log(avccontext, AV_LOG_ERROR, "Error decoding AAC frame header.\n");
1615 return -1;
1616 }
30272450 1617 if (ac->m4ac.sampling_index > 12) {
f418b861
JM
1618 av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
1619 return -1;
1620 }
158b3912
RS
1621 }
1622
62a57fae
RS
1623 // parse
1624 while ((elem_type = get_bits(&gb, 3)) != TYPE_END) {
1625 elem_id = get_bits(&gb, 4);
62a57fae 1626
bb5c0988 1627 if(elem_type < TYPE_DSE && !(che=get_che(ac, elem_type, elem_id))) {
5f401ee0
RS
1628 av_log(ac->avccontext, AV_LOG_ERROR, "channel element %d.%d is not allocated\n", elem_type, elem_id);
1629 return -1;
62a57fae
RS
1630 }
1631
1632 switch (elem_type) {
1633
1634 case TYPE_SCE:
bb5c0988 1635 err = decode_ics(ac, &che->ch[0], &gb, 0, 0);
62a57fae
RS
1636 break;
1637
1638 case TYPE_CPE:
bb5c0988 1639 err = decode_cpe(ac, &gb, che);
62a57fae
RS
1640 break;
1641
1642 case TYPE_CCE:
bb5c0988 1643 err = decode_cce(ac, &gb, che);
62a57fae
RS
1644 break;
1645
1646 case TYPE_LFE:
bb5c0988 1647 err = decode_ics(ac, &che->ch[0], &gb, 0, 0);
62a57fae
RS
1648 break;
1649
1650 case TYPE_DSE:
1651 skip_data_stream_element(&gb);
1652 err = 0;
1653 break;
1654
1655 case TYPE_PCE:
1656 {
1657 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
1658 memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
1659 if((err = decode_pce(ac, new_che_pos, &gb)))
1660 break;
6308765c
AC
1661 if (ac->output_configured)
1662 av_log(avccontext, AV_LOG_ERROR,
1663 "Not evaluating a further program_config_element as this construct is dubious at best.\n");
1664 else
36f51773 1665 err = output_configure(ac, ac->che_pos, new_che_pos, 0);
62a57fae
RS
1666 break;
1667 }
1668
1669 case TYPE_FIL:
1670 if (elem_id == 15)
1671 elem_id += get_bits(&gb, 8) - 1;
1672 while (elem_id > 0)
1673 elem_id -= decode_extension_payload(ac, &gb, elem_id);
1674 err = 0; /* FIXME */
1675 break;
1676
1677 default:
1678 err = -1; /* should not happen, but keeps compiler happy */
1679 break;
1680 }
1681
1682 if(err)
1683 return err;
1684 }
1685
1686 spectral_to_sample(ac);
1687
9cc04edf
RS
1688 if (!ac->is_saved) {
1689 ac->is_saved = 1;
1690 *data_size = 0;
848a5815 1691 return buf_size;
9cc04edf
RS
1692 }
1693
1694 data_size_tmp = 1024 * avccontext->channels * sizeof(int16_t);
1695 if(*data_size < data_size_tmp) {
1696 av_log(avccontext, AV_LOG_ERROR,
1697 "Output buffer too small (%d) or trying to output too many samples (%d) for this frame.\n",
1698 *data_size, data_size_tmp);
1699 return -1;
1700 }
1701 *data_size = data_size_tmp;
1702
1703 ac->dsp.float_to_int16_interleave(data, (const float **)ac->output_data, 1024, avccontext->channels);
1704
1705 return buf_size;
1706}
1707
71e9a1b8
RS
1708static av_cold int aac_decode_close(AVCodecContext * avccontext) {
1709 AACContext * ac = avccontext->priv_data;
9edae4ad 1710 int i, type;
71e9a1b8 1711
cc0591da 1712 for (i = 0; i < MAX_ELEM_ID; i++) {
9edae4ad
RS
1713 for(type = 0; type < 4; type++)
1714 av_freep(&ac->che[type][i]);
71e9a1b8
RS
1715 }
1716
1717 ff_mdct_end(&ac->mdct);
1718 ff_mdct_end(&ac->mdct_small);
71e9a1b8
RS
1719 return 0 ;
1720}
1721
1722AVCodec aac_decoder = {
1723 "aac",
1724 CODEC_TYPE_AUDIO,
1725 CODEC_ID_AAC,
1726 sizeof(AACContext),
1727 aac_decode_init,
1728 NULL,
1729 aac_decode_close,
1730 aac_decode_frame,
1731 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
cc0591da 1732 .sample_fmts = (enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE},
71e9a1b8 1733};