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 * Copyright (c) 2008-2013 Alex Converse <alex.converse@gmail.com>
6 *
7 * AAC LATM decoder
8 * Copyright (c) 2008-2010 Paul Kendall <paul@kcbbs.gen.nz>
9 * Copyright (c) 2010 Janne Grunau <janne-libav@jannau.net>
10 *
11 * AAC decoder fixed-point implementation
12 * Copyright (c) 2013
13 * MIPS Technologies, Inc., California.
14 *
15 * This file is part of FFmpeg.
16 *
17 * FFmpeg is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU Lesser General Public
19 * License as published by the Free Software Foundation; either
20 * version 2.1 of the License, or (at your option) any later version.
21 *
22 * FFmpeg is distributed in the hope that it will be useful,
23 * but WITHOUT ANY WARRANTY; without even the implied warranty of
24 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
25 * Lesser General Public License for more details.
26 *
27 * You should have received a copy of the GNU Lesser General Public
28 * License along with FFmpeg; if not, write to the Free Software
29 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
30 */
31
32 /**
33 * @file
34 * AAC decoder
35 * @author Oded Shimon ( ods15 ods15 dyndns org )
36 * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
37 *
38 * AAC decoder fixed-point implementation
39 * @author Stanislav Ocovaj ( stanislav.ocovaj imgtec com )
40 * @author Nedeljko Babic ( nedeljko.babic imgtec com )
41 */
42
43 /*
44 * supported tools
45 *
46 * Support? Name
47 * N (code in SoC repo) gain control
48 * Y block switching
49 * Y window shapes - standard
50 * N window shapes - Low Delay
51 * Y filterbank - standard
52 * N (code in SoC repo) filterbank - Scalable Sample Rate
53 * Y Temporal Noise Shaping
54 * Y Long Term Prediction
55 * Y intensity stereo
56 * Y channel coupling
57 * Y frequency domain prediction
58 * Y Perceptual Noise Substitution
59 * Y Mid/Side stereo
60 * N Scalable Inverse AAC Quantization
61 * N Frequency Selective Switch
62 * N upsampling filter
63 * Y quantization & coding - AAC
64 * N quantization & coding - TwinVQ
65 * N quantization & coding - BSAC
66 * N AAC Error Resilience tools
67 * N Error Resilience payload syntax
68 * N Error Protection tool
69 * N CELP
70 * N Silence Compression
71 * N HVXC
72 * N HVXC 4kbits/s VR
73 * N Structured Audio tools
74 * N Structured Audio Sample Bank Format
75 * N MIDI
76 * N Harmonic and Individual Lines plus Noise
77 * N Text-To-Speech Interface
78 * Y Spectral Band Replication
79 * Y (not in this code) Layer-1
80 * Y (not in this code) Layer-2
81 * Y (not in this code) Layer-3
82 * N SinuSoidal Coding (Transient, Sinusoid, Noise)
83 * Y Parametric Stereo
84 * N Direct Stream Transfer
85 * Y (not in fixed point code) Enhanced AAC Low Delay (ER AAC ELD)
86 *
87 * Note: - HE AAC v1 comprises LC AAC with Spectral Band Replication.
88 * - HE AAC v2 comprises LC AAC with Spectral Band Replication and
89 Parametric Stereo.
90 */
91
92 #include "libavutil/thread.h"
93
94 static VLC vlc_scalefactors;
95 static VLC vlc_spectral[11];
96
97 static int output_configure(AACContext *ac,
98 uint8_t layout_map[MAX_ELEM_ID*4][3], int tags,
99 enum OCStatus oc_type, int get_new_frame);
100
101 #define overread_err "Input buffer exhausted before END element found\n"
102
count_channels(uint8_t (* layout)[3],int tags)103 static int count_channels(uint8_t (*layout)[3], int tags)
104 {
105 int i, sum = 0;
106 for (i = 0; i < tags; i++) {
107 int syn_ele = layout[i][0];
108 int pos = layout[i][2];
109 sum += (1 + (syn_ele == TYPE_CPE)) *
110 (pos != AAC_CHANNEL_OFF && pos != AAC_CHANNEL_CC);
111 }
112 return sum;
113 }
114
115 /**
116 * Check for the channel element in the current channel position configuration.
117 * If it exists, make sure the appropriate element is allocated and map the
118 * channel order to match the internal FFmpeg channel layout.
119 *
120 * @param che_pos current channel position configuration
121 * @param type channel element type
122 * @param id channel element id
123 * @param channels count of the number of channels in the configuration
124 *
125 * @return Returns error status. 0 - OK, !0 - error
126 */
che_configure(AACContext * ac,enum ChannelPosition che_pos,int type,int id,int * channels)127 static av_cold int che_configure(AACContext *ac,
128 enum ChannelPosition che_pos,
129 int type, int id, int *channels)
130 {
131 if (*channels >= MAX_CHANNELS)
132 return AVERROR_INVALIDDATA;
133 if (che_pos) {
134 if (!ac->che[type][id]) {
135 if (!(ac->che[type][id] = av_mallocz(sizeof(ChannelElement))))
136 return AVERROR(ENOMEM);
137 AAC_RENAME(ff_aac_sbr_ctx_init)(ac, &ac->che[type][id]->sbr, type);
138 }
139 if (type != TYPE_CCE) {
140 if (*channels >= MAX_CHANNELS - (type == TYPE_CPE || (type == TYPE_SCE && ac->oc[1].m4ac.ps == 1))) {
141 av_log(ac->avctx, AV_LOG_ERROR, "Too many channels\n");
142 return AVERROR_INVALIDDATA;
143 }
144 ac->output_element[(*channels)++] = &ac->che[type][id]->ch[0];
145 if (type == TYPE_CPE ||
146 (type == TYPE_SCE && ac->oc[1].m4ac.ps == 1)) {
147 ac->output_element[(*channels)++] = &ac->che[type][id]->ch[1];
148 }
149 }
150 } else {
151 if (ac->che[type][id])
152 AAC_RENAME(ff_aac_sbr_ctx_close)(&ac->che[type][id]->sbr);
153 av_freep(&ac->che[type][id]);
154 }
155 return 0;
156 }
157
frame_configure_elements(AVCodecContext * avctx)158 static int frame_configure_elements(AVCodecContext *avctx)
159 {
160 AACContext *ac = avctx->priv_data;
161 int type, id, ch, ret;
162
163 /* set channel pointers to internal buffers by default */
164 for (type = 0; type < 4; type++) {
165 for (id = 0; id < MAX_ELEM_ID; id++) {
166 ChannelElement *che = ac->che[type][id];
167 if (che) {
168 che->ch[0].ret = che->ch[0].ret_buf;
169 che->ch[1].ret = che->ch[1].ret_buf;
170 }
171 }
172 }
173
174 /* get output buffer */
175 av_frame_unref(ac->frame);
176 if (!avctx->channels)
177 return 1;
178
179 ac->frame->nb_samples = 2048;
180 if ((ret = ff_get_buffer(avctx, ac->frame, 0)) < 0)
181 return ret;
182
183 /* map output channel pointers to AVFrame data */
184 for (ch = 0; ch < avctx->channels; ch++) {
185 if (ac->output_element[ch])
186 ac->output_element[ch]->ret = (INTFLOAT *)ac->frame->extended_data[ch];
187 }
188
189 return 0;
190 }
191
192 struct elem_to_channel {
193 uint64_t av_position;
194 uint8_t syn_ele;
195 uint8_t elem_id;
196 uint8_t aac_position;
197 };
198
assign_pair(struct elem_to_channel e2c_vec[MAX_ELEM_ID],uint8_t (* layout_map)[3],int offset,uint64_t left,uint64_t right,int pos)199 static int assign_pair(struct elem_to_channel e2c_vec[MAX_ELEM_ID],
200 uint8_t (*layout_map)[3], int offset, uint64_t left,
201 uint64_t right, int pos)
202 {
203 if (layout_map[offset][0] == TYPE_CPE) {
204 e2c_vec[offset] = (struct elem_to_channel) {
205 .av_position = left | right,
206 .syn_ele = TYPE_CPE,
207 .elem_id = layout_map[offset][1],
208 .aac_position = pos
209 };
210 return 1;
211 } else {
212 e2c_vec[offset] = (struct elem_to_channel) {
213 .av_position = left,
214 .syn_ele = TYPE_SCE,
215 .elem_id = layout_map[offset][1],
216 .aac_position = pos
217 };
218 e2c_vec[offset + 1] = (struct elem_to_channel) {
219 .av_position = right,
220 .syn_ele = TYPE_SCE,
221 .elem_id = layout_map[offset + 1][1],
222 .aac_position = pos
223 };
224 return 2;
225 }
226 }
227
count_paired_channels(uint8_t (* layout_map)[3],int tags,int pos,int * current)228 static int count_paired_channels(uint8_t (*layout_map)[3], int tags, int pos,
229 int *current)
230 {
231 int num_pos_channels = 0;
232 int first_cpe = 0;
233 int sce_parity = 0;
234 int i;
235 for (i = *current; i < tags; i++) {
236 if (layout_map[i][2] != pos)
237 break;
238 if (layout_map[i][0] == TYPE_CPE) {
239 if (sce_parity) {
240 if (pos == AAC_CHANNEL_FRONT && !first_cpe) {
241 sce_parity = 0;
242 } else {
243 return -1;
244 }
245 }
246 num_pos_channels += 2;
247 first_cpe = 1;
248 } else {
249 num_pos_channels++;
250 sce_parity ^= 1;
251 }
252 }
253 if (sce_parity &&
254 ((pos == AAC_CHANNEL_FRONT && first_cpe) || pos == AAC_CHANNEL_SIDE))
255 return -1;
256 *current = i;
257 return num_pos_channels;
258 }
259
sniff_channel_order(uint8_t (* layout_map)[3],int tags)260 static uint64_t sniff_channel_order(uint8_t (*layout_map)[3], int tags)
261 {
262 int i, n, total_non_cc_elements;
263 struct elem_to_channel e2c_vec[4 * MAX_ELEM_ID] = { { 0 } };
264 int num_front_channels, num_side_channels, num_back_channels;
265 uint64_t layout;
266
267 if (FF_ARRAY_ELEMS(e2c_vec) < tags)
268 return 0;
269
270 i = 0;
271 num_front_channels =
272 count_paired_channels(layout_map, tags, AAC_CHANNEL_FRONT, &i);
273 if (num_front_channels < 0)
274 return 0;
275 num_side_channels =
276 count_paired_channels(layout_map, tags, AAC_CHANNEL_SIDE, &i);
277 if (num_side_channels < 0)
278 return 0;
279 num_back_channels =
280 count_paired_channels(layout_map, tags, AAC_CHANNEL_BACK, &i);
281 if (num_back_channels < 0)
282 return 0;
283
284 if (num_side_channels == 0 && num_back_channels >= 4) {
285 num_side_channels = 2;
286 num_back_channels -= 2;
287 }
288
289 i = 0;
290 if (num_front_channels & 1) {
291 e2c_vec[i] = (struct elem_to_channel) {
292 .av_position = AV_CH_FRONT_CENTER,
293 .syn_ele = TYPE_SCE,
294 .elem_id = layout_map[i][1],
295 .aac_position = AAC_CHANNEL_FRONT
296 };
297 i++;
298 num_front_channels--;
299 }
300 if (num_front_channels >= 4) {
301 i += assign_pair(e2c_vec, layout_map, i,
302 AV_CH_FRONT_LEFT_OF_CENTER,
303 AV_CH_FRONT_RIGHT_OF_CENTER,
304 AAC_CHANNEL_FRONT);
305 num_front_channels -= 2;
306 }
307 if (num_front_channels >= 2) {
308 i += assign_pair(e2c_vec, layout_map, i,
309 AV_CH_FRONT_LEFT,
310 AV_CH_FRONT_RIGHT,
311 AAC_CHANNEL_FRONT);
312 num_front_channels -= 2;
313 }
314 while (num_front_channels >= 2) {
315 i += assign_pair(e2c_vec, layout_map, i,
316 UINT64_MAX,
317 UINT64_MAX,
318 AAC_CHANNEL_FRONT);
319 num_front_channels -= 2;
320 }
321
322 if (num_side_channels >= 2) {
323 i += assign_pair(e2c_vec, layout_map, i,
324 AV_CH_SIDE_LEFT,
325 AV_CH_SIDE_RIGHT,
326 AAC_CHANNEL_FRONT);
327 num_side_channels -= 2;
328 }
329 while (num_side_channels >= 2) {
330 i += assign_pair(e2c_vec, layout_map, i,
331 UINT64_MAX,
332 UINT64_MAX,
333 AAC_CHANNEL_SIDE);
334 num_side_channels -= 2;
335 }
336
337 while (num_back_channels >= 4) {
338 i += assign_pair(e2c_vec, layout_map, i,
339 UINT64_MAX,
340 UINT64_MAX,
341 AAC_CHANNEL_BACK);
342 num_back_channels -= 2;
343 }
344 if (num_back_channels >= 2) {
345 i += assign_pair(e2c_vec, layout_map, i,
346 AV_CH_BACK_LEFT,
347 AV_CH_BACK_RIGHT,
348 AAC_CHANNEL_BACK);
349 num_back_channels -= 2;
350 }
351 if (num_back_channels) {
352 e2c_vec[i] = (struct elem_to_channel) {
353 .av_position = AV_CH_BACK_CENTER,
354 .syn_ele = TYPE_SCE,
355 .elem_id = layout_map[i][1],
356 .aac_position = AAC_CHANNEL_BACK
357 };
358 i++;
359 num_back_channels--;
360 }
361
362 if (i < tags && layout_map[i][2] == AAC_CHANNEL_LFE) {
363 e2c_vec[i] = (struct elem_to_channel) {
364 .av_position = AV_CH_LOW_FREQUENCY,
365 .syn_ele = TYPE_LFE,
366 .elem_id = layout_map[i][1],
367 .aac_position = AAC_CHANNEL_LFE
368 };
369 i++;
370 }
371 while (i < tags && layout_map[i][2] == AAC_CHANNEL_LFE) {
372 e2c_vec[i] = (struct elem_to_channel) {
373 .av_position = UINT64_MAX,
374 .syn_ele = TYPE_LFE,
375 .elem_id = layout_map[i][1],
376 .aac_position = AAC_CHANNEL_LFE
377 };
378 i++;
379 }
380
381 // Must choose a stable sort
382 total_non_cc_elements = n = i;
383 do {
384 int next_n = 0;
385 for (i = 1; i < n; i++)
386 if (e2c_vec[i - 1].av_position > e2c_vec[i].av_position) {
387 FFSWAP(struct elem_to_channel, e2c_vec[i - 1], e2c_vec[i]);
388 next_n = i;
389 }
390 n = next_n;
391 } while (n > 0);
392
393 layout = 0;
394 for (i = 0; i < total_non_cc_elements; i++) {
395 layout_map[i][0] = e2c_vec[i].syn_ele;
396 layout_map[i][1] = e2c_vec[i].elem_id;
397 layout_map[i][2] = e2c_vec[i].aac_position;
398 if (e2c_vec[i].av_position != UINT64_MAX) {
399 layout |= e2c_vec[i].av_position;
400 }
401 }
402
403 return layout;
404 }
405
406 /**
407 * Save current output configuration if and only if it has been locked.
408 */
push_output_configuration(AACContext * ac)409 static int push_output_configuration(AACContext *ac) {
410 int pushed = 0;
411
412 if (ac->oc[1].status == OC_LOCKED || ac->oc[0].status == OC_NONE) {
413 ac->oc[0] = ac->oc[1];
414 pushed = 1;
415 }
416 ac->oc[1].status = OC_NONE;
417 return pushed;
418 }
419
420 /**
421 * Restore the previous output configuration if and only if the current
422 * configuration is unlocked.
423 */
pop_output_configuration(AACContext * ac)424 static void pop_output_configuration(AACContext *ac) {
425 if (ac->oc[1].status != OC_LOCKED && ac->oc[0].status != OC_NONE) {
426 ac->oc[1] = ac->oc[0];
427 ac->avctx->channels = ac->oc[1].channels;
428 ac->avctx->channel_layout = ac->oc[1].channel_layout;
429 output_configure(ac, ac->oc[1].layout_map, ac->oc[1].layout_map_tags,
430 ac->oc[1].status, 0);
431 }
432 }
433
434 /**
435 * Configure output channel order based on the current program
436 * configuration element.
437 *
438 * @return Returns error status. 0 - OK, !0 - error
439 */
output_configure(AACContext * ac,uint8_t layout_map[MAX_ELEM_ID * 4][3],int tags,enum OCStatus oc_type,int get_new_frame)440 static int output_configure(AACContext *ac,
441 uint8_t layout_map[MAX_ELEM_ID * 4][3], int tags,
442 enum OCStatus oc_type, int get_new_frame)
443 {
444 AVCodecContext *avctx = ac->avctx;
445 int i, channels = 0, ret;
446 uint64_t layout = 0;
447 uint8_t id_map[TYPE_END][MAX_ELEM_ID] = {{ 0 }};
448 uint8_t type_counts[TYPE_END] = { 0 };
449
450 if (ac->oc[1].layout_map != layout_map) {
451 memcpy(ac->oc[1].layout_map, layout_map, tags * sizeof(layout_map[0]));
452 ac->oc[1].layout_map_tags = tags;
453 }
454 for (i = 0; i < tags; i++) {
455 int type = layout_map[i][0];
456 int id = layout_map[i][1];
457 id_map[type][id] = type_counts[type]++;
458 if (id_map[type][id] >= MAX_ELEM_ID) {
459 avpriv_request_sample(ac->avctx, "Too large remapped id");
460 return AVERROR_PATCHWELCOME;
461 }
462 }
463 // Try to sniff a reasonable channel order, otherwise output the
464 // channels in the order the PCE declared them.
465 if (avctx->request_channel_layout != AV_CH_LAYOUT_NATIVE)
466 layout = sniff_channel_order(layout_map, tags);
467 for (i = 0; i < tags; i++) {
468 int type = layout_map[i][0];
469 int id = layout_map[i][1];
470 int iid = id_map[type][id];
471 int position = layout_map[i][2];
472 // Allocate or free elements depending on if they are in the
473 // current program configuration.
474 ret = che_configure(ac, position, type, iid, &channels);
475 if (ret < 0)
476 return ret;
477 ac->tag_che_map[type][id] = ac->che[type][iid];
478 }
479 if (ac->oc[1].m4ac.ps == 1 && channels == 2) {
480 if (layout == AV_CH_FRONT_CENTER) {
481 layout = AV_CH_FRONT_LEFT|AV_CH_FRONT_RIGHT;
482 } else {
483 layout = 0;
484 }
485 }
486
487 if (layout) avctx->channel_layout = layout;
488 ac->oc[1].channel_layout = layout;
489 avctx->channels = ac->oc[1].channels = channels;
490 ac->oc[1].status = oc_type;
491
492 if (get_new_frame) {
493 if ((ret = frame_configure_elements(ac->avctx)) < 0)
494 return ret;
495 }
496
497 return 0;
498 }
499
flush(AVCodecContext * avctx)500 static void flush(AVCodecContext *avctx)
501 {
502 AACContext *ac= avctx->priv_data;
503 int type, i, j;
504
505 for (type = 3; type >= 0; type--) {
506 for (i = 0; i < MAX_ELEM_ID; i++) {
507 ChannelElement *che = ac->che[type][i];
508 if (che) {
509 for (j = 0; j <= 1; j++) {
510 memset(che->ch[j].saved, 0, sizeof(che->ch[j].saved));
511 }
512 }
513 }
514 }
515 }
516
517 /**
518 * Set up channel positions based on a default channel configuration
519 * as specified in table 1.17.
520 *
521 * @return Returns error status. 0 - OK, !0 - error
522 */
set_default_channel_config(AACContext * ac,AVCodecContext * avctx,uint8_t (* layout_map)[3],int * tags,int channel_config)523 static int set_default_channel_config(AACContext *ac, AVCodecContext *avctx,
524 uint8_t (*layout_map)[3],
525 int *tags,
526 int channel_config)
527 {
528 if (channel_config < 1 || (channel_config > 7 && channel_config < 11) ||
529 channel_config > 12) {
530 av_log(avctx, AV_LOG_ERROR,
531 "invalid default channel configuration (%d)\n",
532 channel_config);
533 return AVERROR_INVALIDDATA;
534 }
535 *tags = tags_per_config[channel_config];
536 memcpy(layout_map, aac_channel_layout_map[channel_config - 1],
537 *tags * sizeof(*layout_map));
538
539 /*
540 * AAC specification has 7.1(wide) as a default layout for 8-channel streams.
541 * However, at least Nero AAC encoder encodes 7.1 streams using the default
542 * channel config 7, mapping the side channels of the original audio stream
543 * to the second AAC_CHANNEL_FRONT pair in the AAC stream. Similarly, e.g. FAAD
544 * decodes the second AAC_CHANNEL_FRONT pair as side channels, therefore decoding
545 * the incorrect streams as if they were correct (and as the encoder intended).
546 *
547 * As actual intended 7.1(wide) streams are very rare, default to assuming a
548 * 7.1 layout was intended.
549 */
550 if (channel_config == 7 && avctx->strict_std_compliance < FF_COMPLIANCE_STRICT && (!ac || !ac->warned_71_wide++)) {
551 av_log(avctx, AV_LOG_INFO, "Assuming an incorrectly encoded 7.1 channel layout"
552 " instead of a spec-compliant 7.1(wide) layout, use -strict %d to decode"
553 " according to the specification instead.\n", FF_COMPLIANCE_STRICT);
554 layout_map[2][2] = AAC_CHANNEL_SIDE;
555 }
556
557 return 0;
558 }
559
get_che(AACContext * ac,int type,int elem_id)560 static ChannelElement *get_che(AACContext *ac, int type, int elem_id)
561 {
562 /* For PCE based channel configurations map the channels solely based
563 * on tags. */
564 if (!ac->oc[1].m4ac.chan_config) {
565 return ac->tag_che_map[type][elem_id];
566 }
567 // Allow single CPE stereo files to be signalled with mono configuration.
568 if (!ac->tags_mapped && type == TYPE_CPE &&
569 ac->oc[1].m4ac.chan_config == 1) {
570 uint8_t layout_map[MAX_ELEM_ID*4][3];
571 int layout_map_tags;
572 push_output_configuration(ac);
573
574 av_log(ac->avctx, AV_LOG_DEBUG, "mono with CPE\n");
575
576 if (set_default_channel_config(ac, ac->avctx, layout_map,
577 &layout_map_tags, 2) < 0)
578 return NULL;
579 if (output_configure(ac, layout_map, layout_map_tags,
580 OC_TRIAL_FRAME, 1) < 0)
581 return NULL;
582
583 ac->oc[1].m4ac.chan_config = 2;
584 ac->oc[1].m4ac.ps = 0;
585 }
586 // And vice-versa
587 if (!ac->tags_mapped && type == TYPE_SCE &&
588 ac->oc[1].m4ac.chan_config == 2) {
589 uint8_t layout_map[MAX_ELEM_ID * 4][3];
590 int layout_map_tags;
591 push_output_configuration(ac);
592
593 av_log(ac->avctx, AV_LOG_DEBUG, "stereo with SCE\n");
594
595 if (set_default_channel_config(ac, ac->avctx, layout_map,
596 &layout_map_tags, 1) < 0)
597 return NULL;
598 if (output_configure(ac, layout_map, layout_map_tags,
599 OC_TRIAL_FRAME, 1) < 0)
600 return NULL;
601
602 ac->oc[1].m4ac.chan_config = 1;
603 if (ac->oc[1].m4ac.sbr)
604 ac->oc[1].m4ac.ps = -1;
605 }
606 /* For indexed channel configurations map the channels solely based
607 * on position. */
608 switch (ac->oc[1].m4ac.chan_config) {
609 case 12:
610 case 7:
611 if (ac->tags_mapped == 3 && type == TYPE_CPE) {
612 ac->tags_mapped++;
613 return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][2];
614 }
615 case 11:
616 if (ac->tags_mapped == 2 &&
617 ac->oc[1].m4ac.chan_config == 11 &&
618 type == TYPE_SCE) {
619 ac->tags_mapped++;
620 return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][1];
621 }
622 case 6:
623 /* Some streams incorrectly code 5.1 audio as
624 * SCE[0] CPE[0] CPE[1] SCE[1]
625 * instead of
626 * SCE[0] CPE[0] CPE[1] LFE[0].
627 * If we seem to have encountered such a stream, transfer
628 * the LFE[0] element to the SCE[1]'s mapping */
629 if (ac->tags_mapped == tags_per_config[ac->oc[1].m4ac.chan_config] - 1 && (type == TYPE_LFE || type == TYPE_SCE)) {
630 if (!ac->warned_remapping_once && (type != TYPE_LFE || elem_id != 0)) {
631 av_log(ac->avctx, AV_LOG_WARNING,
632 "This stream seems to incorrectly report its last channel as %s[%d], mapping to LFE[0]\n",
633 type == TYPE_SCE ? "SCE" : "LFE", elem_id);
634 ac->warned_remapping_once++;
635 }
636 ac->tags_mapped++;
637 return ac->tag_che_map[type][elem_id] = ac->che[TYPE_LFE][0];
638 }
639 case 5:
640 if (ac->tags_mapped == 2 && type == TYPE_CPE) {
641 ac->tags_mapped++;
642 return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][1];
643 }
644 case 4:
645 /* Some streams incorrectly code 4.0 audio as
646 * SCE[0] CPE[0] LFE[0]
647 * instead of
648 * SCE[0] CPE[0] SCE[1].
649 * If we seem to have encountered such a stream, transfer
650 * the SCE[1] element to the LFE[0]'s mapping */
651 if (ac->tags_mapped == tags_per_config[ac->oc[1].m4ac.chan_config] - 1 && (type == TYPE_LFE || type == TYPE_SCE)) {
652 if (!ac->warned_remapping_once && (type != TYPE_SCE || elem_id != 1)) {
653 av_log(ac->avctx, AV_LOG_WARNING,
654 "This stream seems to incorrectly report its last channel as %s[%d], mapping to SCE[1]\n",
655 type == TYPE_SCE ? "SCE" : "LFE", elem_id);
656 ac->warned_remapping_once++;
657 }
658 ac->tags_mapped++;
659 return ac->tag_che_map[type][elem_id] = ac->che[TYPE_SCE][1];
660 }
661 if (ac->tags_mapped == 2 &&
662 ac->oc[1].m4ac.chan_config == 4 &&
663 type == TYPE_SCE) {
664 ac->tags_mapped++;
665 return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][1];
666 }
667 case 3:
668 case 2:
669 if (ac->tags_mapped == (ac->oc[1].m4ac.chan_config != 2) &&
670 type == TYPE_CPE) {
671 ac->tags_mapped++;
672 return ac->tag_che_map[TYPE_CPE][elem_id] = ac->che[TYPE_CPE][0];
673 } else if (ac->oc[1].m4ac.chan_config == 2) {
674 return NULL;
675 }
676 case 1:
677 if (!ac->tags_mapped && type == TYPE_SCE) {
678 ac->tags_mapped++;
679 return ac->tag_che_map[TYPE_SCE][elem_id] = ac->che[TYPE_SCE][0];
680 }
681 default:
682 return NULL;
683 }
684 }
685
686 /**
687 * Decode an array of 4 bit element IDs, optionally interleaved with a
688 * stereo/mono switching bit.
689 *
690 * @param type speaker type/position for these channels
691 */
decode_channel_map(uint8_t layout_map[][3],enum ChannelPosition type,GetBitContext * gb,int n)692 static void decode_channel_map(uint8_t layout_map[][3],
693 enum ChannelPosition type,
694 GetBitContext *gb, int n)
695 {
696 while (n--) {
697 enum RawDataBlockType syn_ele;
698 switch (type) {
699 case AAC_CHANNEL_FRONT:
700 case AAC_CHANNEL_BACK:
701 case AAC_CHANNEL_SIDE:
702 syn_ele = get_bits1(gb);
703 break;
704 case AAC_CHANNEL_CC:
705 skip_bits1(gb);
706 syn_ele = TYPE_CCE;
707 break;
708 case AAC_CHANNEL_LFE:
709 syn_ele = TYPE_LFE;
710 break;
711 default:
712 // AAC_CHANNEL_OFF has no channel map
713 av_assert0(0);
714 }
715 layout_map[0][0] = syn_ele;
716 layout_map[0][1] = get_bits(gb, 4);
717 layout_map[0][2] = type;
718 layout_map++;
719 }
720 }
721
relative_align_get_bits(GetBitContext * gb,int reference_position)722 static inline void relative_align_get_bits(GetBitContext *gb,
723 int reference_position) {
724 int n = (reference_position - get_bits_count(gb) & 7);
725 if (n)
726 skip_bits(gb, n);
727 }
728
729 /**
730 * Decode program configuration element; reference: table 4.2.
731 *
732 * @return Returns error status. 0 - OK, !0 - error
733 */
decode_pce(AVCodecContext * avctx,MPEG4AudioConfig * m4ac,uint8_t (* layout_map)[3],GetBitContext * gb,int byte_align_ref)734 static int decode_pce(AVCodecContext *avctx, MPEG4AudioConfig *m4ac,
735 uint8_t (*layout_map)[3],
736 GetBitContext *gb, int byte_align_ref)
737 {
738 int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc;
739 int sampling_index;
740 int comment_len;
741 int tags;
742
743 skip_bits(gb, 2); // object_type
744
745 sampling_index = get_bits(gb, 4);
746 if (m4ac->sampling_index != sampling_index)
747 av_log(avctx, AV_LOG_WARNING,
748 "Sample rate index in program config element does not "
749 "match the sample rate index configured by the container.\n");
750
751 num_front = get_bits(gb, 4);
752 num_side = get_bits(gb, 4);
753 num_back = get_bits(gb, 4);
754 num_lfe = get_bits(gb, 2);
755 num_assoc_data = get_bits(gb, 3);
756 num_cc = get_bits(gb, 4);
757
758 if (get_bits1(gb))
759 skip_bits(gb, 4); // mono_mixdown_tag
760 if (get_bits1(gb))
761 skip_bits(gb, 4); // stereo_mixdown_tag
762
763 if (get_bits1(gb))
764 skip_bits(gb, 3); // mixdown_coeff_index and pseudo_surround
765
766 if (get_bits_left(gb) < 5 * (num_front + num_side + num_back + num_cc) + 4 *(num_lfe + num_assoc_data + num_cc)) {
767 av_log(avctx, AV_LOG_ERROR, "decode_pce: " overread_err);
768 return -1;
769 }
770 decode_channel_map(layout_map , AAC_CHANNEL_FRONT, gb, num_front);
771 tags = num_front;
772 decode_channel_map(layout_map + tags, AAC_CHANNEL_SIDE, gb, num_side);
773 tags += num_side;
774 decode_channel_map(layout_map + tags, AAC_CHANNEL_BACK, gb, num_back);
775 tags += num_back;
776 decode_channel_map(layout_map + tags, AAC_CHANNEL_LFE, gb, num_lfe);
777 tags += num_lfe;
778
779 skip_bits_long(gb, 4 * num_assoc_data);
780
781 decode_channel_map(layout_map + tags, AAC_CHANNEL_CC, gb, num_cc);
782 tags += num_cc;
783
784 relative_align_get_bits(gb, byte_align_ref);
785
786 /* comment field, first byte is length */
787 comment_len = get_bits(gb, 8) * 8;
788 if (get_bits_left(gb) < comment_len) {
789 av_log(avctx, AV_LOG_ERROR, "decode_pce: " overread_err);
790 return AVERROR_INVALIDDATA;
791 }
792 skip_bits_long(gb, comment_len);
793 return tags;
794 }
795
796 /**
797 * Decode GA "General Audio" specific configuration; reference: table 4.1.
798 *
799 * @param ac pointer to AACContext, may be null
800 * @param avctx pointer to AVCCodecContext, used for logging
801 *
802 * @return Returns error status. 0 - OK, !0 - error
803 */
decode_ga_specific_config(AACContext * ac,AVCodecContext * avctx,GetBitContext * gb,int get_bit_alignment,MPEG4AudioConfig * m4ac,int channel_config)804 static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx,
805 GetBitContext *gb,
806 int get_bit_alignment,
807 MPEG4AudioConfig *m4ac,
808 int channel_config)
809 {
810 int extension_flag, ret, ep_config, res_flags;
811 uint8_t layout_map[MAX_ELEM_ID*4][3];
812 int tags = 0;
813
814 #if USE_FIXED
815 if (get_bits1(gb)) { // frameLengthFlag
816 avpriv_report_missing_feature(avctx, "Fixed point 960/120 MDCT window");
817 return AVERROR_PATCHWELCOME;
818 }
819 m4ac->frame_length_short = 0;
820 #else
821 m4ac->frame_length_short = get_bits1(gb);
822 if (m4ac->frame_length_short && m4ac->sbr == 1) {
823 avpriv_report_missing_feature(avctx, "SBR with 960 frame length");
824 if (ac) ac->warned_960_sbr = 1;
825 m4ac->sbr = 0;
826 m4ac->ps = 0;
827 }
828 #endif
829
830 if (get_bits1(gb)) // dependsOnCoreCoder
831 skip_bits(gb, 14); // coreCoderDelay
832 extension_flag = get_bits1(gb);
833
834 if (m4ac->object_type == AOT_AAC_SCALABLE ||
835 m4ac->object_type == AOT_ER_AAC_SCALABLE)
836 skip_bits(gb, 3); // layerNr
837
838 if (channel_config == 0) {
839 skip_bits(gb, 4); // element_instance_tag
840 tags = decode_pce(avctx, m4ac, layout_map, gb, get_bit_alignment);
841 if (tags < 0)
842 return tags;
843 } else {
844 if ((ret = set_default_channel_config(ac, avctx, layout_map,
845 &tags, channel_config)))
846 return ret;
847 }
848
849 if (count_channels(layout_map, tags) > 1) {
850 m4ac->ps = 0;
851 } else if (m4ac->sbr == 1 && m4ac->ps == -1)
852 m4ac->ps = 1;
853
854 if (ac && (ret = output_configure(ac, layout_map, tags, OC_GLOBAL_HDR, 0)))
855 return ret;
856
857 if (extension_flag) {
858 switch (m4ac->object_type) {
859 case AOT_ER_BSAC:
860 skip_bits(gb, 5); // numOfSubFrame
861 skip_bits(gb, 11); // layer_length
862 break;
863 case AOT_ER_AAC_LC:
864 case AOT_ER_AAC_LTP:
865 case AOT_ER_AAC_SCALABLE:
866 case AOT_ER_AAC_LD:
867 res_flags = get_bits(gb, 3);
868 if (res_flags) {
869 avpriv_report_missing_feature(avctx,
870 "AAC data resilience (flags %x)",
871 res_flags);
872 return AVERROR_PATCHWELCOME;
873 }
874 break;
875 }
876 skip_bits1(gb); // extensionFlag3 (TBD in version 3)
877 }
878 switch (m4ac->object_type) {
879 case AOT_ER_AAC_LC:
880 case AOT_ER_AAC_LTP:
881 case AOT_ER_AAC_SCALABLE:
882 case AOT_ER_AAC_LD:
883 ep_config = get_bits(gb, 2);
884 if (ep_config) {
885 avpriv_report_missing_feature(avctx,
886 "epConfig %d", ep_config);
887 return AVERROR_PATCHWELCOME;
888 }
889 }
890 return 0;
891 }
892
decode_eld_specific_config(AACContext * ac,AVCodecContext * avctx,GetBitContext * gb,MPEG4AudioConfig * m4ac,int channel_config)893 static int decode_eld_specific_config(AACContext *ac, AVCodecContext *avctx,
894 GetBitContext *gb,
895 MPEG4AudioConfig *m4ac,
896 int channel_config)
897 {
898 int ret, ep_config, res_flags;
899 uint8_t layout_map[MAX_ELEM_ID*4][3];
900 int tags = 0;
901 const int ELDEXT_TERM = 0;
902
903 m4ac->ps = 0;
904 m4ac->sbr = 0;
905 #if USE_FIXED
906 if (get_bits1(gb)) { // frameLengthFlag
907 avpriv_request_sample(avctx, "960/120 MDCT window");
908 return AVERROR_PATCHWELCOME;
909 }
910 #else
911 m4ac->frame_length_short = get_bits1(gb);
912 #endif
913 res_flags = get_bits(gb, 3);
914 if (res_flags) {
915 avpriv_report_missing_feature(avctx,
916 "AAC data resilience (flags %x)",
917 res_flags);
918 return AVERROR_PATCHWELCOME;
919 }
920
921 if (get_bits1(gb)) { // ldSbrPresentFlag
922 avpriv_report_missing_feature(avctx,
923 "Low Delay SBR");
924 return AVERROR_PATCHWELCOME;
925 }
926
927 while (get_bits(gb, 4) != ELDEXT_TERM) {
928 int len = get_bits(gb, 4);
929 if (len == 15)
930 len += get_bits(gb, 8);
931 if (len == 15 + 255)
932 len += get_bits(gb, 16);
933 if (get_bits_left(gb) < len * 8 + 4) {
934 av_log(avctx, AV_LOG_ERROR, overread_err);
935 return AVERROR_INVALIDDATA;
936 }
937 skip_bits_long(gb, 8 * len);
938 }
939
940 if ((ret = set_default_channel_config(ac, avctx, layout_map,
941 &tags, channel_config)))
942 return ret;
943
944 if (ac && (ret = output_configure(ac, layout_map, tags, OC_GLOBAL_HDR, 0)))
945 return ret;
946
947 ep_config = get_bits(gb, 2);
948 if (ep_config) {
949 avpriv_report_missing_feature(avctx,
950 "epConfig %d", ep_config);
951 return AVERROR_PATCHWELCOME;
952 }
953 return 0;
954 }
955
956 /**
957 * Decode audio specific configuration; reference: table 1.13.
958 *
959 * @param ac pointer to AACContext, may be null
960 * @param avctx pointer to AVCCodecContext, used for logging
961 * @param m4ac pointer to MPEG4AudioConfig, used for parsing
962 * @param gb buffer holding an audio specific config
963 * @param get_bit_alignment relative alignment for byte align operations
964 * @param sync_extension look for an appended sync extension
965 *
966 * @return Returns error status or number of consumed bits. <0 - error
967 */
decode_audio_specific_config_gb(AACContext * ac,AVCodecContext * avctx,MPEG4AudioConfig * m4ac,GetBitContext * gb,int get_bit_alignment,int sync_extension)968 static int decode_audio_specific_config_gb(AACContext *ac,
969 AVCodecContext *avctx,
970 MPEG4AudioConfig *m4ac,
971 GetBitContext *gb,
972 int get_bit_alignment,
973 int sync_extension)
974 {
975 int i, ret;
976 GetBitContext gbc = *gb;
977
978 if ((i = ff_mpeg4audio_get_config_gb(m4ac, &gbc, sync_extension, avctx)) < 0)
979 return AVERROR_INVALIDDATA;
980
981 if (m4ac->sampling_index > 12) {
982 av_log(avctx, AV_LOG_ERROR,
983 "invalid sampling rate index %d\n",
984 m4ac->sampling_index);
985 return AVERROR_INVALIDDATA;
986 }
987 if (m4ac->object_type == AOT_ER_AAC_LD &&
988 (m4ac->sampling_index < 3 || m4ac->sampling_index > 7)) {
989 av_log(avctx, AV_LOG_ERROR,
990 "invalid low delay sampling rate index %d\n",
991 m4ac->sampling_index);
992 return AVERROR_INVALIDDATA;
993 }
994
995 skip_bits_long(gb, i);
996
997 switch (m4ac->object_type) {
998 case AOT_AAC_MAIN:
999 case AOT_AAC_LC:
1000 case AOT_AAC_SSR:
1001 case AOT_AAC_LTP:
1002 case AOT_ER_AAC_LC:
1003 case AOT_ER_AAC_LD:
1004 if ((ret = decode_ga_specific_config(ac, avctx, gb, get_bit_alignment,
1005 m4ac, m4ac->chan_config)) < 0)
1006 return ret;
1007 break;
1008 case AOT_ER_AAC_ELD:
1009 if ((ret = decode_eld_specific_config(ac, avctx, gb,
1010 m4ac, m4ac->chan_config)) < 0)
1011 return ret;
1012 break;
1013 default:
1014 avpriv_report_missing_feature(avctx,
1015 "Audio object type %s%d",
1016 m4ac->sbr == 1 ? "SBR+" : "",
1017 m4ac->object_type);
1018 return AVERROR(ENOSYS);
1019 }
1020
1021 ff_dlog(avctx,
1022 "AOT %d chan config %d sampling index %d (%d) SBR %d PS %d\n",
1023 m4ac->object_type, m4ac->chan_config, m4ac->sampling_index,
1024 m4ac->sample_rate, m4ac->sbr,
1025 m4ac->ps);
1026
1027 return get_bits_count(gb);
1028 }
1029
decode_audio_specific_config(AACContext * ac,AVCodecContext * avctx,MPEG4AudioConfig * m4ac,const uint8_t * data,int64_t bit_size,int sync_extension)1030 static int decode_audio_specific_config(AACContext *ac,
1031 AVCodecContext *avctx,
1032 MPEG4AudioConfig *m4ac,
1033 const uint8_t *data, int64_t bit_size,
1034 int sync_extension)
1035 {
1036 int i, ret;
1037 GetBitContext gb;
1038
1039 if (bit_size < 0 || bit_size > INT_MAX) {
1040 av_log(avctx, AV_LOG_ERROR, "Audio specific config size is invalid\n");
1041 return AVERROR_INVALIDDATA;
1042 }
1043
1044 ff_dlog(avctx, "audio specific config size %d\n", (int)bit_size >> 3);
1045 for (i = 0; i < bit_size >> 3; i++)
1046 ff_dlog(avctx, "%02x ", data[i]);
1047 ff_dlog(avctx, "\n");
1048
1049 if ((ret = init_get_bits(&gb, data, bit_size)) < 0)
1050 return ret;
1051
1052 return decode_audio_specific_config_gb(ac, avctx, m4ac, &gb, 0,
1053 sync_extension);
1054 }
1055
1056 /**
1057 * linear congruential pseudorandom number generator
1058 *
1059 * @param previous_val pointer to the current state of the generator
1060 *
1061 * @return Returns a 32-bit pseudorandom integer
1062 */
lcg_random(unsigned previous_val)1063 static av_always_inline int lcg_random(unsigned previous_val)
1064 {
1065 union { unsigned u; int s; } v = { previous_val * 1664525u + 1013904223 };
1066 return v.s;
1067 }
1068
reset_all_predictors(PredictorState * ps)1069 static void reset_all_predictors(PredictorState *ps)
1070 {
1071 int i;
1072 for (i = 0; i < MAX_PREDICTORS; i++)
1073 reset_predict_state(&ps[i]);
1074 }
1075
sample_rate_idx(int rate)1076 static int sample_rate_idx (int rate)
1077 {
1078 if (92017 <= rate) return 0;
1079 else if (75132 <= rate) return 1;
1080 else if (55426 <= rate) return 2;
1081 else if (46009 <= rate) return 3;
1082 else if (37566 <= rate) return 4;
1083 else if (27713 <= rate) return 5;
1084 else if (23004 <= rate) return 6;
1085 else if (18783 <= rate) return 7;
1086 else if (13856 <= rate) return 8;
1087 else if (11502 <= rate) return 9;
1088 else if (9391 <= rate) return 10;
1089 else return 11;
1090 }
1091
reset_predictor_group(PredictorState * ps,int group_num)1092 static void reset_predictor_group(PredictorState *ps, int group_num)
1093 {
1094 int i;
1095 for (i = group_num - 1; i < MAX_PREDICTORS; i += 30)
1096 reset_predict_state(&ps[i]);
1097 }
1098
1099 #define AAC_INIT_VLC_STATIC(num, size) \
1100 INIT_VLC_STATIC(&vlc_spectral[num], 8, ff_aac_spectral_sizes[num], \
1101 ff_aac_spectral_bits[num], sizeof(ff_aac_spectral_bits[num][0]), \
1102 sizeof(ff_aac_spectral_bits[num][0]), \
1103 ff_aac_spectral_codes[num], sizeof(ff_aac_spectral_codes[num][0]), \
1104 sizeof(ff_aac_spectral_codes[num][0]), \
1105 size);
1106
1107 static void aacdec_init(AACContext *ac);
1108
aac_static_table_init(void)1109 static av_cold void aac_static_table_init(void)
1110 {
1111 AAC_INIT_VLC_STATIC( 0, 304);
1112 AAC_INIT_VLC_STATIC( 1, 270);
1113 AAC_INIT_VLC_STATIC( 2, 550);
1114 AAC_INIT_VLC_STATIC( 3, 300);
1115 AAC_INIT_VLC_STATIC( 4, 328);
1116 AAC_INIT_VLC_STATIC( 5, 294);
1117 AAC_INIT_VLC_STATIC( 6, 306);
1118 AAC_INIT_VLC_STATIC( 7, 268);
1119 AAC_INIT_VLC_STATIC( 8, 510);
1120 AAC_INIT_VLC_STATIC( 9, 366);
1121 AAC_INIT_VLC_STATIC(10, 462);
1122
1123 AAC_RENAME(ff_aac_sbr_init)();
1124
1125 ff_aac_tableinit();
1126
1127 INIT_VLC_STATIC(&vlc_scalefactors, 7,
1128 FF_ARRAY_ELEMS(ff_aac_scalefactor_code),
1129 ff_aac_scalefactor_bits,
1130 sizeof(ff_aac_scalefactor_bits[0]),
1131 sizeof(ff_aac_scalefactor_bits[0]),
1132 ff_aac_scalefactor_code,
1133 sizeof(ff_aac_scalefactor_code[0]),
1134 sizeof(ff_aac_scalefactor_code[0]),
1135 352);
1136
1137 // window initialization
1138 AAC_RENAME(ff_kbd_window_init)(AAC_RENAME(ff_aac_kbd_long_1024), 4.0, 1024);
1139 AAC_RENAME(ff_kbd_window_init)(AAC_RENAME(ff_aac_kbd_short_128), 6.0, 128);
1140 #if !USE_FIXED
1141 AAC_RENAME(ff_kbd_window_init)(AAC_RENAME(ff_aac_kbd_long_960), 4.0, 960);
1142 AAC_RENAME(ff_kbd_window_init)(AAC_RENAME(ff_aac_kbd_short_120), 6.0, 120);
1143 AAC_RENAME(ff_sine_window_init)(AAC_RENAME(ff_sine_960), 960);
1144 AAC_RENAME(ff_sine_window_init)(AAC_RENAME(ff_sine_120), 120);
1145 #endif
1146 AAC_RENAME(ff_init_ff_sine_windows)(10);
1147 AAC_RENAME(ff_init_ff_sine_windows)( 9);
1148 AAC_RENAME(ff_init_ff_sine_windows)( 7);
1149
1150 AAC_RENAME(ff_cbrt_tableinit)();
1151 }
1152
1153 static AVOnce aac_table_init = AV_ONCE_INIT;
1154
aac_decode_init(AVCodecContext * avctx)1155 static av_cold int aac_decode_init(AVCodecContext *avctx)
1156 {
1157 AACContext *ac = avctx->priv_data;
1158 int ret;
1159
1160 if (avctx->sample_rate > 96000)
1161 return AVERROR_INVALIDDATA;
1162
1163 ret = ff_thread_once(&aac_table_init, &aac_static_table_init);
1164 if (ret != 0)
1165 return AVERROR_UNKNOWN;
1166
1167 ac->avctx = avctx;
1168 ac->oc[1].m4ac.sample_rate = avctx->sample_rate;
1169
1170 aacdec_init(ac);
1171 #if USE_FIXED
1172 avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
1173 #else
1174 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
1175 #endif /* USE_FIXED */
1176
1177 if (avctx->extradata_size > 0) {
1178 if ((ret = decode_audio_specific_config(ac, ac->avctx, &ac->oc[1].m4ac,
1179 avctx->extradata,
1180 avctx->extradata_size * 8LL,
1181 1)) < 0)
1182 return ret;
1183 } else {
1184 int sr, i;
1185 uint8_t layout_map[MAX_ELEM_ID*4][3];
1186 int layout_map_tags;
1187
1188 sr = sample_rate_idx(avctx->sample_rate);
1189 ac->oc[1].m4ac.sampling_index = sr;
1190 ac->oc[1].m4ac.channels = avctx->channels;
1191 ac->oc[1].m4ac.sbr = -1;
1192 ac->oc[1].m4ac.ps = -1;
1193
1194 for (i = 0; i < FF_ARRAY_ELEMS(ff_mpeg4audio_channels); i++)
1195 if (ff_mpeg4audio_channels[i] == avctx->channels)
1196 break;
1197 if (i == FF_ARRAY_ELEMS(ff_mpeg4audio_channels)) {
1198 i = 0;
1199 }
1200 ac->oc[1].m4ac.chan_config = i;
1201
1202 if (ac->oc[1].m4ac.chan_config) {
1203 int ret = set_default_channel_config(ac, avctx, layout_map,
1204 &layout_map_tags, ac->oc[1].m4ac.chan_config);
1205 if (!ret)
1206 output_configure(ac, layout_map, layout_map_tags,
1207 OC_GLOBAL_HDR, 0);
1208 else if (avctx->err_recognition & AV_EF_EXPLODE)
1209 return AVERROR_INVALIDDATA;
1210 }
1211 }
1212
1213 if (avctx->channels > MAX_CHANNELS) {
1214 av_log(avctx, AV_LOG_ERROR, "Too many channels\n");
1215 return AVERROR_INVALIDDATA;
1216 }
1217
1218 #if USE_FIXED
1219 ac->fdsp = avpriv_alloc_fixed_dsp(avctx->flags & AV_CODEC_FLAG_BITEXACT);
1220 #else
1221 ac->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
1222 #endif /* USE_FIXED */
1223 if (!ac->fdsp) {
1224 return AVERROR(ENOMEM);
1225 }
1226
1227 ac->random_state = 0x1f2e3d4c;
1228
1229 AAC_RENAME_32(ff_mdct_init)(&ac->mdct, 11, 1, 1.0 / RANGE15(1024.0));
1230 AAC_RENAME_32(ff_mdct_init)(&ac->mdct_ld, 10, 1, 1.0 / RANGE15(512.0));
1231 AAC_RENAME_32(ff_mdct_init)(&ac->mdct_small, 8, 1, 1.0 / RANGE15(128.0));
1232 AAC_RENAME_32(ff_mdct_init)(&ac->mdct_ltp, 11, 0, RANGE15(-2.0));
1233 #if !USE_FIXED
1234 ret = ff_mdct15_init(&ac->mdct120, 1, 3, 1.0f/(16*1024*120*2));
1235 if (ret < 0)
1236 return ret;
1237 ret = ff_mdct15_init(&ac->mdct480, 1, 5, 1.0f/(16*1024*960));
1238 if (ret < 0)
1239 return ret;
1240 ret = ff_mdct15_init(&ac->mdct960, 1, 6, 1.0f/(16*1024*960*2));
1241 if (ret < 0)
1242 return ret;
1243 #endif
1244
1245 return 0;
1246 }
1247
1248 /**
1249 * Skip data_stream_element; reference: table 4.10.
1250 */
skip_data_stream_element(AACContext * ac,GetBitContext * gb)1251 static int skip_data_stream_element(AACContext *ac, GetBitContext *gb)
1252 {
1253 int byte_align = get_bits1(gb);
1254 int count = get_bits(gb, 8);
1255 if (count == 255)
1256 count += get_bits(gb, 8);
1257 if (byte_align)
1258 align_get_bits(gb);
1259
1260 if (get_bits_left(gb) < 8 * count) {
1261 av_log(ac->avctx, AV_LOG_ERROR, "skip_data_stream_element: "overread_err);
1262 return AVERROR_INVALIDDATA;
1263 }
1264 skip_bits_long(gb, 8 * count);
1265 return 0;
1266 }
1267
decode_prediction(AACContext * ac,IndividualChannelStream * ics,GetBitContext * gb)1268 static int decode_prediction(AACContext *ac, IndividualChannelStream *ics,
1269 GetBitContext *gb)
1270 {
1271 int sfb;
1272 if (get_bits1(gb)) {
1273 ics->predictor_reset_group = get_bits(gb, 5);
1274 if (ics->predictor_reset_group == 0 ||
1275 ics->predictor_reset_group > 30) {
1276 av_log(ac->avctx, AV_LOG_ERROR,
1277 "Invalid Predictor Reset Group.\n");
1278 return AVERROR_INVALIDDATA;
1279 }
1280 }
1281 for (sfb = 0; sfb < FFMIN(ics->max_sfb, ff_aac_pred_sfb_max[ac->oc[1].m4ac.sampling_index]); sfb++) {
1282 ics->prediction_used[sfb] = get_bits1(gb);
1283 }
1284 return 0;
1285 }
1286
1287 /**
1288 * Decode Long Term Prediction data; reference: table 4.xx.
1289 */
decode_ltp(LongTermPrediction * ltp,GetBitContext * gb,uint8_t max_sfb)1290 static void decode_ltp(LongTermPrediction *ltp,
1291 GetBitContext *gb, uint8_t max_sfb)
1292 {
1293 int sfb;
1294
1295 ltp->lag = get_bits(gb, 11);
1296 ltp->coef = ltp_coef[get_bits(gb, 3)];
1297 for (sfb = 0; sfb < FFMIN(max_sfb, MAX_LTP_LONG_SFB); sfb++)
1298 ltp->used[sfb] = get_bits1(gb);
1299 }
1300
1301 /**
1302 * Decode Individual Channel Stream info; reference: table 4.6.
1303 */
decode_ics_info(AACContext * ac,IndividualChannelStream * ics,GetBitContext * gb)1304 static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics,
1305 GetBitContext *gb)
1306 {
1307 const MPEG4AudioConfig *const m4ac = &ac->oc[1].m4ac;
1308 const int aot = m4ac->object_type;
1309 const int sampling_index = m4ac->sampling_index;
1310 int ret_fail = AVERROR_INVALIDDATA;
1311
1312 if (aot != AOT_ER_AAC_ELD) {
1313 if (get_bits1(gb)) {
1314 av_log(ac->avctx, AV_LOG_ERROR, "Reserved bit set.\n");
1315 if (ac->avctx->err_recognition & AV_EF_BITSTREAM)
1316 return AVERROR_INVALIDDATA;
1317 }
1318 ics->window_sequence[1] = ics->window_sequence[0];
1319 ics->window_sequence[0] = get_bits(gb, 2);
1320 if (aot == AOT_ER_AAC_LD &&
1321 ics->window_sequence[0] != ONLY_LONG_SEQUENCE) {
1322 av_log(ac->avctx, AV_LOG_ERROR,
1323 "AAC LD is only defined for ONLY_LONG_SEQUENCE but "
1324 "window sequence %d found.\n", ics->window_sequence[0]);
1325 ics->window_sequence[0] = ONLY_LONG_SEQUENCE;
1326 return AVERROR_INVALIDDATA;
1327 }
1328 ics->use_kb_window[1] = ics->use_kb_window[0];
1329 ics->use_kb_window[0] = get_bits1(gb);
1330 }
1331 ics->num_window_groups = 1;
1332 ics->group_len[0] = 1;
1333 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
1334 int i;
1335 ics->max_sfb = get_bits(gb, 4);
1336 for (i = 0; i < 7; i++) {
1337 if (get_bits1(gb)) {
1338 ics->group_len[ics->num_window_groups - 1]++;
1339 } else {
1340 ics->num_window_groups++;
1341 ics->group_len[ics->num_window_groups - 1] = 1;
1342 }
1343 }
1344 ics->num_windows = 8;
1345 if (m4ac->frame_length_short) {
1346 ics->swb_offset = ff_swb_offset_120[sampling_index];
1347 ics->num_swb = ff_aac_num_swb_120[sampling_index];
1348 } else {
1349 ics->swb_offset = ff_swb_offset_128[sampling_index];
1350 ics->num_swb = ff_aac_num_swb_128[sampling_index];
1351 }
1352 ics->tns_max_bands = ff_tns_max_bands_128[sampling_index];
1353 ics->predictor_present = 0;
1354 } else {
1355 ics->max_sfb = get_bits(gb, 6);
1356 ics->num_windows = 1;
1357 if (aot == AOT_ER_AAC_LD || aot == AOT_ER_AAC_ELD) {
1358 if (m4ac->frame_length_short) {
1359 ics->swb_offset = ff_swb_offset_480[sampling_index];
1360 ics->num_swb = ff_aac_num_swb_480[sampling_index];
1361 ics->tns_max_bands = ff_tns_max_bands_480[sampling_index];
1362 } else {
1363 ics->swb_offset = ff_swb_offset_512[sampling_index];
1364 ics->num_swb = ff_aac_num_swb_512[sampling_index];
1365 ics->tns_max_bands = ff_tns_max_bands_512[sampling_index];
1366 }
1367 if (!ics->num_swb || !ics->swb_offset) {
1368 ret_fail = AVERROR_BUG;
1369 goto fail;
1370 }
1371 } else {
1372 if (m4ac->frame_length_short) {
1373 ics->num_swb = ff_aac_num_swb_960[sampling_index];
1374 ics->swb_offset = ff_swb_offset_960[sampling_index];
1375 } else {
1376 ics->num_swb = ff_aac_num_swb_1024[sampling_index];
1377 ics->swb_offset = ff_swb_offset_1024[sampling_index];
1378 }
1379 ics->tns_max_bands = ff_tns_max_bands_1024[sampling_index];
1380 }
1381 if (aot != AOT_ER_AAC_ELD) {
1382 ics->predictor_present = get_bits1(gb);
1383 ics->predictor_reset_group = 0;
1384 }
1385 if (ics->predictor_present) {
1386 if (aot == AOT_AAC_MAIN) {
1387 if (decode_prediction(ac, ics, gb)) {
1388 goto fail;
1389 }
1390 } else if (aot == AOT_AAC_LC ||
1391 aot == AOT_ER_AAC_LC) {
1392 av_log(ac->avctx, AV_LOG_ERROR,
1393 "Prediction is not allowed in AAC-LC.\n");
1394 goto fail;
1395 } else {
1396 if (aot == AOT_ER_AAC_LD) {
1397 av_log(ac->avctx, AV_LOG_ERROR,
1398 "LTP in ER AAC LD not yet implemented.\n");
1399 ret_fail = AVERROR_PATCHWELCOME;
1400 goto fail;
1401 }
1402 if ((ics->ltp.present = get_bits(gb, 1)))
1403 decode_ltp(&ics->ltp, gb, ics->max_sfb);
1404 }
1405 }
1406 }
1407
1408 if (ics->max_sfb > ics->num_swb) {
1409 av_log(ac->avctx, AV_LOG_ERROR,
1410 "Number of scalefactor bands in group (%d) "
1411 "exceeds limit (%d).\n",
1412 ics->max_sfb, ics->num_swb);
1413 goto fail;
1414 }
1415
1416 return 0;
1417 fail:
1418 ics->max_sfb = 0;
1419 return ret_fail;
1420 }
1421
1422 /**
1423 * Decode band types (section_data payload); reference: table 4.46.
1424 *
1425 * @param band_type array of the used band type
1426 * @param band_type_run_end array of the last scalefactor band of a band type run
1427 *
1428 * @return Returns error status. 0 - OK, !0 - error
1429 */
decode_band_types(AACContext * ac,enum BandType band_type[120],int band_type_run_end[120],GetBitContext * gb,IndividualChannelStream * ics)1430 static int decode_band_types(AACContext *ac, enum BandType band_type[120],
1431 int band_type_run_end[120], GetBitContext *gb,
1432 IndividualChannelStream *ics)
1433 {
1434 int g, idx = 0;
1435 const int bits = (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) ? 3 : 5;
1436 for (g = 0; g < ics->num_window_groups; g++) {
1437 int k = 0;
1438 while (k < ics->max_sfb) {
1439 uint8_t sect_end = k;
1440 int sect_len_incr;
1441 int sect_band_type = get_bits(gb, 4);
1442 if (sect_band_type == 12) {
1443 av_log(ac->avctx, AV_LOG_ERROR, "invalid band type\n");
1444 return AVERROR_INVALIDDATA;
1445 }
1446 do {
1447 sect_len_incr = get_bits(gb, bits);
1448 sect_end += sect_len_incr;
1449 if (get_bits_left(gb) < 0) {
1450 av_log(ac->avctx, AV_LOG_ERROR, "decode_band_types: "overread_err);
1451 return AVERROR_INVALIDDATA;
1452 }
1453 if (sect_end > ics->max_sfb) {
1454 av_log(ac->avctx, AV_LOG_ERROR,
1455 "Number of bands (%d) exceeds limit (%d).\n",
1456 sect_end, ics->max_sfb);
1457 return AVERROR_INVALIDDATA;
1458 }
1459 } while (sect_len_incr == (1 << bits) - 1);
1460 for (; k < sect_end; k++) {
1461 band_type [idx] = sect_band_type;
1462 band_type_run_end[idx++] = sect_end;
1463 }
1464 }
1465 }
1466 return 0;
1467 }
1468
1469 /**
1470 * Decode scalefactors; reference: table 4.47.
1471 *
1472 * @param global_gain first scalefactor value as scalefactors are differentially coded
1473 * @param band_type array of the used band type
1474 * @param band_type_run_end array of the last scalefactor band of a band type run
1475 * @param sf array of scalefactors or intensity stereo positions
1476 *
1477 * @return Returns error status. 0 - OK, !0 - error
1478 */
decode_scalefactors(AACContext * ac,INTFLOAT sf[120],GetBitContext * gb,unsigned int global_gain,IndividualChannelStream * ics,enum BandType band_type[120],int band_type_run_end[120])1479 static int decode_scalefactors(AACContext *ac, INTFLOAT sf[120], GetBitContext *gb,
1480 unsigned int global_gain,
1481 IndividualChannelStream *ics,
1482 enum BandType band_type[120],
1483 int band_type_run_end[120])
1484 {
1485 int g, i, idx = 0;
1486 int offset[3] = { global_gain, global_gain - NOISE_OFFSET, 0 };
1487 int clipped_offset;
1488 int noise_flag = 1;
1489 for (g = 0; g < ics->num_window_groups; g++) {
1490 for (i = 0; i < ics->max_sfb;) {
1491 int run_end = band_type_run_end[idx];
1492 if (band_type[idx] == ZERO_BT) {
1493 for (; i < run_end; i++, idx++)
1494 sf[idx] = FIXR(0.);
1495 } else if ((band_type[idx] == INTENSITY_BT) ||
1496 (band_type[idx] == INTENSITY_BT2)) {
1497 for (; i < run_end; i++, idx++) {
1498 offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - SCALE_DIFF_ZERO;
1499 clipped_offset = av_clip(offset[2], -155, 100);
1500 if (offset[2] != clipped_offset) {
1501 avpriv_request_sample(ac->avctx,
1502 "If you heard an audible artifact, there may be a bug in the decoder. "
1503 "Clipped intensity stereo position (%d -> %d)",
1504 offset[2], clipped_offset);
1505 }
1506 #if USE_FIXED
1507 sf[idx] = 100 - clipped_offset;
1508 #else
1509 sf[idx] = ff_aac_pow2sf_tab[-clipped_offset + POW_SF2_ZERO];
1510 #endif /* USE_FIXED */
1511 }
1512 } else if (band_type[idx] == NOISE_BT) {
1513 for (; i < run_end; i++, idx++) {
1514 if (noise_flag-- > 0)
1515 offset[1] += get_bits(gb, NOISE_PRE_BITS) - NOISE_PRE;
1516 else
1517 offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - SCALE_DIFF_ZERO;
1518 clipped_offset = av_clip(offset[1], -100, 155);
1519 if (offset[1] != clipped_offset) {
1520 avpriv_request_sample(ac->avctx,
1521 "If you heard an audible artifact, there may be a bug in the decoder. "
1522 "Clipped noise gain (%d -> %d)",
1523 offset[1], clipped_offset);
1524 }
1525 #if USE_FIXED
1526 sf[idx] = -(100 + clipped_offset);
1527 #else
1528 sf[idx] = -ff_aac_pow2sf_tab[clipped_offset + POW_SF2_ZERO];
1529 #endif /* USE_FIXED */
1530 }
1531 } else {
1532 for (; i < run_end; i++, idx++) {
1533 offset[0] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - SCALE_DIFF_ZERO;
1534 if (offset[0] > 255U) {
1535 av_log(ac->avctx, AV_LOG_ERROR,
1536 "Scalefactor (%d) out of range.\n", offset[0]);
1537 return AVERROR_INVALIDDATA;
1538 }
1539 #if USE_FIXED
1540 sf[idx] = -offset[0];
1541 #else
1542 sf[idx] = -ff_aac_pow2sf_tab[offset[0] - 100 + POW_SF2_ZERO];
1543 #endif /* USE_FIXED */
1544 }
1545 }
1546 }
1547 }
1548 return 0;
1549 }
1550
1551 /**
1552 * Decode pulse data; reference: table 4.7.
1553 */
decode_pulses(Pulse * pulse,GetBitContext * gb,const uint16_t * swb_offset,int num_swb)1554 static int decode_pulses(Pulse *pulse, GetBitContext *gb,
1555 const uint16_t *swb_offset, int num_swb)
1556 {
1557 int i, pulse_swb;
1558 pulse->num_pulse = get_bits(gb, 2) + 1;
1559 pulse_swb = get_bits(gb, 6);
1560 if (pulse_swb >= num_swb)
1561 return -1;
1562 pulse->pos[0] = swb_offset[pulse_swb];
1563 pulse->pos[0] += get_bits(gb, 5);
1564 if (pulse->pos[0] >= swb_offset[num_swb])
1565 return -1;
1566 pulse->amp[0] = get_bits(gb, 4);
1567 for (i = 1; i < pulse->num_pulse; i++) {
1568 pulse->pos[i] = get_bits(gb, 5) + pulse->pos[i - 1];
1569 if (pulse->pos[i] >= swb_offset[num_swb])
1570 return -1;
1571 pulse->amp[i] = get_bits(gb, 4);
1572 }
1573 return 0;
1574 }
1575
1576 /**
1577 * Decode Temporal Noise Shaping data; reference: table 4.48.
1578 *
1579 * @return Returns error status. 0 - OK, !0 - error
1580 */
decode_tns(AACContext * ac,TemporalNoiseShaping * tns,GetBitContext * gb,const IndividualChannelStream * ics)1581 static int decode_tns(AACContext *ac, TemporalNoiseShaping *tns,
1582 GetBitContext *gb, const IndividualChannelStream *ics)
1583 {
1584 int w, filt, i, coef_len, coef_res, coef_compress;
1585 const int is8 = ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE;
1586 const int tns_max_order = is8 ? 7 : ac->oc[1].m4ac.object_type == AOT_AAC_MAIN ? 20 : 12;
1587 for (w = 0; w < ics->num_windows; w++) {
1588 if ((tns->n_filt[w] = get_bits(gb, 2 - is8))) {
1589 coef_res = get_bits1(gb);
1590
1591 for (filt = 0; filt < tns->n_filt[w]; filt++) {
1592 int tmp2_idx;
1593 tns->length[w][filt] = get_bits(gb, 6 - 2 * is8);
1594
1595 if ((tns->order[w][filt] = get_bits(gb, 5 - 2 * is8)) > tns_max_order) {
1596 av_log(ac->avctx, AV_LOG_ERROR,
1597 "TNS filter order %d is greater than maximum %d.\n",
1598 tns->order[w][filt], tns_max_order);
1599 tns->order[w][filt] = 0;
1600 return AVERROR_INVALIDDATA;
1601 }
1602 if (tns->order[w][filt]) {
1603 tns->direction[w][filt] = get_bits1(gb);
1604 coef_compress = get_bits1(gb);
1605 coef_len = coef_res + 3 - coef_compress;
1606 tmp2_idx = 2 * coef_compress + coef_res;
1607
1608 for (i = 0; i < tns->order[w][filt]; i++)
1609 tns->coef[w][filt][i] = tns_tmp2_map[tmp2_idx][get_bits(gb, coef_len)];
1610 }
1611 }
1612 }
1613 }
1614 return 0;
1615 }
1616
1617 /**
1618 * Decode Mid/Side data; reference: table 4.54.
1619 *
1620 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
1621 * [1] mask is decoded from bitstream; [2] mask is all 1s;
1622 * [3] reserved for scalable AAC
1623 */
decode_mid_side_stereo(ChannelElement * cpe,GetBitContext * gb,int ms_present)1624 static void decode_mid_side_stereo(ChannelElement *cpe, GetBitContext *gb,
1625 int ms_present)
1626 {
1627 int idx;
1628 int max_idx = cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb;
1629 if (ms_present == 1) {
1630 for (idx = 0; idx < max_idx; idx++)
1631 cpe->ms_mask[idx] = get_bits1(gb);
1632 } else if (ms_present == 2) {
1633 memset(cpe->ms_mask, 1, max_idx * sizeof(cpe->ms_mask[0]));
1634 }
1635 }
1636
1637 /**
1638 * Decode spectral data; reference: table 4.50.
1639 * Dequantize and scale spectral data; reference: 4.6.3.3.
1640 *
1641 * @param coef array of dequantized, scaled spectral data
1642 * @param sf array of scalefactors or intensity stereo positions
1643 * @param pulse_present set if pulses are present
1644 * @param pulse pointer to pulse data struct
1645 * @param band_type array of the used band type
1646 *
1647 * @return Returns error status. 0 - OK, !0 - error
1648 */
decode_spectrum_and_dequant(AACContext * ac,INTFLOAT coef[1024],GetBitContext * gb,const INTFLOAT sf[120],int pulse_present,const Pulse * pulse,const IndividualChannelStream * ics,enum BandType band_type[120])1649 static int decode_spectrum_and_dequant(AACContext *ac, INTFLOAT coef[1024],
1650 GetBitContext *gb, const INTFLOAT sf[120],
1651 int pulse_present, const Pulse *pulse,
1652 const IndividualChannelStream *ics,
1653 enum BandType band_type[120])
1654 {
1655 int i, k, g, idx = 0;
1656 const int c = 1024 / ics->num_windows;
1657 const uint16_t *offsets = ics->swb_offset;
1658 INTFLOAT *coef_base = coef;
1659
1660 for (g = 0; g < ics->num_windows; g++)
1661 memset(coef + g * 128 + offsets[ics->max_sfb], 0,
1662 sizeof(INTFLOAT) * (c - offsets[ics->max_sfb]));
1663
1664 for (g = 0; g < ics->num_window_groups; g++) {
1665 unsigned g_len = ics->group_len[g];
1666
1667 for (i = 0; i < ics->max_sfb; i++, idx++) {
1668 const unsigned cbt_m1 = band_type[idx] - 1;
1669 INTFLOAT *cfo = coef + offsets[i];
1670 int off_len = offsets[i + 1] - offsets[i];
1671 int group;
1672
1673 if (cbt_m1 >= INTENSITY_BT2 - 1) {
1674 for (group = 0; group < (AAC_SIGNE)g_len; group++, cfo+=128) {
1675 memset(cfo, 0, off_len * sizeof(*cfo));
1676 }
1677 } else if (cbt_m1 == NOISE_BT - 1) {
1678 for (group = 0; group < (AAC_SIGNE)g_len; group++, cfo+=128) {
1679 INTFLOAT band_energy;
1680 #if USE_FIXED
1681 for (k = 0; k < off_len; k++) {
1682 ac->random_state = lcg_random(ac->random_state);
1683 cfo[k] = ac->random_state >> 3;
1684 }
1685
1686 band_energy = ac->fdsp->scalarproduct_fixed(cfo, cfo, off_len);
1687 band_energy = fixed_sqrt(band_energy, 31);
1688 noise_scale(cfo, sf[idx], band_energy, off_len);
1689 #else
1690 float scale;
1691
1692 for (k = 0; k < off_len; k++) {
1693 ac->random_state = lcg_random(ac->random_state);
1694 cfo[k] = ac->random_state;
1695 }
1696
1697 band_energy = ac->fdsp->scalarproduct_float(cfo, cfo, off_len);
1698 scale = sf[idx] / sqrtf(band_energy);
1699 ac->fdsp->vector_fmul_scalar(cfo, cfo, scale, off_len);
1700 #endif /* USE_FIXED */
1701 }
1702 } else {
1703 #if !USE_FIXED
1704 const float *vq = ff_aac_codebook_vector_vals[cbt_m1];
1705 #endif /* !USE_FIXED */
1706 const uint16_t *cb_vector_idx = ff_aac_codebook_vector_idx[cbt_m1];
1707 VLC_TYPE (*vlc_tab)[2] = vlc_spectral[cbt_m1].table;
1708 OPEN_READER(re, gb);
1709
1710 switch (cbt_m1 >> 1) {
1711 case 0:
1712 for (group = 0; group < (AAC_SIGNE)g_len; group++, cfo+=128) {
1713 INTFLOAT *cf = cfo;
1714 int len = off_len;
1715
1716 do {
1717 int code;
1718 unsigned cb_idx;
1719
1720 UPDATE_CACHE(re, gb);
1721 GET_VLC(code, re, gb, vlc_tab, 8, 2);
1722 cb_idx = cb_vector_idx[code];
1723 #if USE_FIXED
1724 cf = DEC_SQUAD(cf, cb_idx);
1725 #else
1726 cf = VMUL4(cf, vq, cb_idx, sf + idx);
1727 #endif /* USE_FIXED */
1728 } while (len -= 4);
1729 }
1730 break;
1731
1732 case 1:
1733 for (group = 0; group < (AAC_SIGNE)g_len; group++, cfo+=128) {
1734 INTFLOAT *cf = cfo;
1735 int len = off_len;
1736
1737 do {
1738 int code;
1739 unsigned nnz;
1740 unsigned cb_idx;
1741 uint32_t bits;
1742
1743 UPDATE_CACHE(re, gb);
1744 GET_VLC(code, re, gb, vlc_tab, 8, 2);
1745 cb_idx = cb_vector_idx[code];
1746 nnz = cb_idx >> 8 & 15;
1747 bits = nnz ? GET_CACHE(re, gb) : 0;
1748 LAST_SKIP_BITS(re, gb, nnz);
1749 #if USE_FIXED
1750 cf = DEC_UQUAD(cf, cb_idx, bits);
1751 #else
1752 cf = VMUL4S(cf, vq, cb_idx, bits, sf + idx);
1753 #endif /* USE_FIXED */
1754 } while (len -= 4);
1755 }
1756 break;
1757
1758 case 2:
1759 for (group = 0; group < (AAC_SIGNE)g_len; group++, cfo+=128) {
1760 INTFLOAT *cf = cfo;
1761 int len = off_len;
1762
1763 do {
1764 int code;
1765 unsigned cb_idx;
1766
1767 UPDATE_CACHE(re, gb);
1768 GET_VLC(code, re, gb, vlc_tab, 8, 2);
1769 cb_idx = cb_vector_idx[code];
1770 #if USE_FIXED
1771 cf = DEC_SPAIR(cf, cb_idx);
1772 #else
1773 cf = VMUL2(cf, vq, cb_idx, sf + idx);
1774 #endif /* USE_FIXED */
1775 } while (len -= 2);
1776 }
1777 break;
1778
1779 case 3:
1780 case 4:
1781 for (group = 0; group < (AAC_SIGNE)g_len; group++, cfo+=128) {
1782 INTFLOAT *cf = cfo;
1783 int len = off_len;
1784
1785 do {
1786 int code;
1787 unsigned nnz;
1788 unsigned cb_idx;
1789 unsigned sign;
1790
1791 UPDATE_CACHE(re, gb);
1792 GET_VLC(code, re, gb, vlc_tab, 8, 2);
1793 cb_idx = cb_vector_idx[code];
1794 nnz = cb_idx >> 8 & 15;
1795 sign = nnz ? SHOW_UBITS(re, gb, nnz) << (cb_idx >> 12) : 0;
1796 LAST_SKIP_BITS(re, gb, nnz);
1797 #if USE_FIXED
1798 cf = DEC_UPAIR(cf, cb_idx, sign);
1799 #else
1800 cf = VMUL2S(cf, vq, cb_idx, sign, sf + idx);
1801 #endif /* USE_FIXED */
1802 } while (len -= 2);
1803 }
1804 break;
1805
1806 default:
1807 for (group = 0; group < (AAC_SIGNE)g_len; group++, cfo+=128) {
1808 #if USE_FIXED
1809 int *icf = cfo;
1810 int v;
1811 #else
1812 float *cf = cfo;
1813 uint32_t *icf = (uint32_t *) cf;
1814 #endif /* USE_FIXED */
1815 int len = off_len;
1816
1817 do {
1818 int code;
1819 unsigned nzt, nnz;
1820 unsigned cb_idx;
1821 uint32_t bits;
1822 int j;
1823
1824 UPDATE_CACHE(re, gb);
1825 GET_VLC(code, re, gb, vlc_tab, 8, 2);
1826
1827 if (!code) {
1828 *icf++ = 0;
1829 *icf++ = 0;
1830 continue;
1831 }
1832
1833 cb_idx = cb_vector_idx[code];
1834 nnz = cb_idx >> 12;
1835 nzt = cb_idx >> 8;
1836 bits = SHOW_UBITS(re, gb, nnz) << (32-nnz);
1837 LAST_SKIP_BITS(re, gb, nnz);
1838
1839 for (j = 0; j < 2; j++) {
1840 if (nzt & 1<<j) {
1841 uint32_t b;
1842 int n;
1843 /* The total length of escape_sequence must be < 22 bits according
1844 to the specification (i.e. max is 111111110xxxxxxxxxxxx). */
1845 UPDATE_CACHE(re, gb);
1846 b = GET_CACHE(re, gb);
1847 b = 31 - av_log2(~b);
1848
1849 if (b > 8) {
1850 av_log(ac->avctx, AV_LOG_ERROR, "error in spectral data, ESC overflow\n");
1851 return AVERROR_INVALIDDATA;
1852 }
1853
1854 SKIP_BITS(re, gb, b + 1);
1855 b += 4;
1856 n = (1 << b) + SHOW_UBITS(re, gb, b);
1857 LAST_SKIP_BITS(re, gb, b);
1858 #if USE_FIXED
1859 v = n;
1860 if (bits & 1U<<31)
1861 v = -v;
1862 *icf++ = v;
1863 #else
1864 *icf++ = ff_cbrt_tab[n] | (bits & 1U<<31);
1865 #endif /* USE_FIXED */
1866 bits <<= 1;
1867 } else {
1868 #if USE_FIXED
1869 v = cb_idx & 15;
1870 if (bits & 1U<<31)
1871 v = -v;
1872 *icf++ = v;
1873 #else
1874 unsigned v = ((const uint32_t*)vq)[cb_idx & 15];
1875 *icf++ = (bits & 1U<<31) | v;
1876 #endif /* USE_FIXED */
1877 bits <<= !!v;
1878 }
1879 cb_idx >>= 4;
1880 }
1881 } while (len -= 2);
1882 #if !USE_FIXED
1883 ac->fdsp->vector_fmul_scalar(cfo, cfo, sf[idx], off_len);
1884 #endif /* !USE_FIXED */
1885 }
1886 }
1887
1888 CLOSE_READER(re, gb);
1889 }
1890 }
1891 coef += g_len << 7;
1892 }
1893
1894 if (pulse_present) {
1895 idx = 0;
1896 for (i = 0; i < pulse->num_pulse; i++) {
1897 INTFLOAT co = coef_base[ pulse->pos[i] ];
1898 while (offsets[idx + 1] <= pulse->pos[i])
1899 idx++;
1900 if (band_type[idx] != NOISE_BT && sf[idx]) {
1901 INTFLOAT ico = -pulse->amp[i];
1902 #if USE_FIXED
1903 if (co) {
1904 ico = co + (co > 0 ? -ico : ico);
1905 }
1906 coef_base[ pulse->pos[i] ] = ico;
1907 #else
1908 if (co) {
1909 co /= sf[idx];
1910 ico = co / sqrtf(sqrtf(fabsf(co))) + (co > 0 ? -ico : ico);
1911 }
1912 coef_base[ pulse->pos[i] ] = cbrtf(fabsf(ico)) * ico * sf[idx];
1913 #endif /* USE_FIXED */
1914 }
1915 }
1916 }
1917 #if USE_FIXED
1918 coef = coef_base;
1919 idx = 0;
1920 for (g = 0; g < ics->num_window_groups; g++) {
1921 unsigned g_len = ics->group_len[g];
1922
1923 for (i = 0; i < ics->max_sfb; i++, idx++) {
1924 const unsigned cbt_m1 = band_type[idx] - 1;
1925 int *cfo = coef + offsets[i];
1926 int off_len = offsets[i + 1] - offsets[i];
1927 int group;
1928
1929 if (cbt_m1 < NOISE_BT - 1) {
1930 for (group = 0; group < (int)g_len; group++, cfo+=128) {
1931 ac->vector_pow43(cfo, off_len);
1932 ac->subband_scale(cfo, cfo, sf[idx], 34, off_len, ac->avctx);
1933 }
1934 }
1935 }
1936 coef += g_len << 7;
1937 }
1938 #endif /* USE_FIXED */
1939 return 0;
1940 }
1941
1942 /**
1943 * Apply AAC-Main style frequency domain prediction.
1944 */
apply_prediction(AACContext * ac,SingleChannelElement * sce)1945 static void apply_prediction(AACContext *ac, SingleChannelElement *sce)
1946 {
1947 int sfb, k;
1948
1949 if (!sce->ics.predictor_initialized) {
1950 reset_all_predictors(sce->predictor_state);
1951 sce->ics.predictor_initialized = 1;
1952 }
1953
1954 if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
1955 for (sfb = 0;
1956 sfb < ff_aac_pred_sfb_max[ac->oc[1].m4ac.sampling_index];
1957 sfb++) {
1958 for (k = sce->ics.swb_offset[sfb];
1959 k < sce->ics.swb_offset[sfb + 1];
1960 k++) {
1961 predict(&sce->predictor_state[k], &sce->coeffs[k],
1962 sce->ics.predictor_present &&
1963 sce->ics.prediction_used[sfb]);
1964 }
1965 }
1966 if (sce->ics.predictor_reset_group)
1967 reset_predictor_group(sce->predictor_state,
1968 sce->ics.predictor_reset_group);
1969 } else
1970 reset_all_predictors(sce->predictor_state);
1971 }
1972
decode_gain_control(SingleChannelElement * sce,GetBitContext * gb)1973 static void decode_gain_control(SingleChannelElement * sce, GetBitContext * gb)
1974 {
1975 // wd_num, wd_test, aloc_size
1976 static const uint8_t gain_mode[4][3] = {
1977 {1, 0, 5}, // ONLY_LONG_SEQUENCE = 0,
1978 {2, 1, 2}, // LONG_START_SEQUENCE,
1979 {8, 0, 2}, // EIGHT_SHORT_SEQUENCE,
1980 {2, 1, 5}, // LONG_STOP_SEQUENCE
1981 };
1982
1983 const int mode = sce->ics.window_sequence[0];
1984 uint8_t bd, wd, ad;
1985
1986 // FIXME: Store the gain control data on |sce| and do something with it.
1987 uint8_t max_band = get_bits(gb, 2);
1988 for (bd = 0; bd < max_band; bd++) {
1989 for (wd = 0; wd < gain_mode[mode][0]; wd++) {
1990 uint8_t adjust_num = get_bits(gb, 3);
1991 for (ad = 0; ad < adjust_num; ad++) {
1992 skip_bits(gb, 4 + ((wd == 0 && gain_mode[mode][1])
1993 ? 4
1994 : gain_mode[mode][2]));
1995 }
1996 }
1997 }
1998 }
1999
2000 /**
2001 * Decode an individual_channel_stream payload; reference: table 4.44.
2002 *
2003 * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
2004 * @param scale_flag scalable [1] or non-scalable [0] AAC (Unused until scalable AAC is implemented.)
2005 *
2006 * @return Returns error status. 0 - OK, !0 - error
2007 */
decode_ics(AACContext * ac,SingleChannelElement * sce,GetBitContext * gb,int common_window,int scale_flag)2008 static int decode_ics(AACContext *ac, SingleChannelElement *sce,
2009 GetBitContext *gb, int common_window, int scale_flag)
2010 {
2011 Pulse pulse;
2012 TemporalNoiseShaping *tns = &sce->tns;
2013 IndividualChannelStream *ics = &sce->ics;
2014 INTFLOAT *out = sce->coeffs;
2015 int global_gain, eld_syntax, er_syntax, pulse_present = 0;
2016 int ret;
2017
2018 eld_syntax = ac->oc[1].m4ac.object_type == AOT_ER_AAC_ELD;
2019 er_syntax = ac->oc[1].m4ac.object_type == AOT_ER_AAC_LC ||
2020 ac->oc[1].m4ac.object_type == AOT_ER_AAC_LTP ||
2021 ac->oc[1].m4ac.object_type == AOT_ER_AAC_LD ||
2022 ac->oc[1].m4ac.object_type == AOT_ER_AAC_ELD;
2023
2024 /* This assignment is to silence a GCC warning about the variable being used
2025 * uninitialized when in fact it always is.
2026 */
2027 pulse.num_pulse = 0;
2028
2029 global_gain = get_bits(gb, 8);
2030
2031 if (!common_window && !scale_flag) {
2032 ret = decode_ics_info(ac, ics, gb);
2033 if (ret < 0)
2034 goto fail;
2035 }
2036
2037 if ((ret = decode_band_types(ac, sce->band_type,
2038 sce->band_type_run_end, gb, ics)) < 0)
2039 goto fail;
2040 if ((ret = decode_scalefactors(ac, sce->sf, gb, global_gain, ics,
2041 sce->band_type, sce->band_type_run_end)) < 0)
2042 goto fail;
2043
2044 pulse_present = 0;
2045 if (!scale_flag) {
2046 if (!eld_syntax && (pulse_present = get_bits1(gb))) {
2047 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
2048 av_log(ac->avctx, AV_LOG_ERROR,
2049 "Pulse tool not allowed in eight short sequence.\n");
2050 ret = AVERROR_INVALIDDATA;
2051 goto fail;
2052 }
2053 if (decode_pulses(&pulse, gb, ics->swb_offset, ics->num_swb)) {
2054 av_log(ac->avctx, AV_LOG_ERROR,
2055 "Pulse data corrupt or invalid.\n");
2056 ret = AVERROR_INVALIDDATA;
2057 goto fail;
2058 }
2059 }
2060 tns->present = get_bits1(gb);
2061 if (tns->present && !er_syntax) {
2062 ret = decode_tns(ac, tns, gb, ics);
2063 if (ret < 0)
2064 goto fail;
2065 }
2066 if (!eld_syntax && get_bits1(gb)) {
2067 decode_gain_control(sce, gb);
2068 if (!ac->warned_gain_control) {
2069 avpriv_report_missing_feature(ac->avctx, "Gain control");
2070 ac->warned_gain_control = 1;
2071 }
2072 }
2073 // I see no textual basis in the spec for this occurring after SSR gain
2074 // control, but this is what both reference and real implmentations do
2075 if (tns->present && er_syntax) {
2076 ret = decode_tns(ac, tns, gb, ics);
2077 if (ret < 0)
2078 goto fail;
2079 }
2080 }
2081
2082 ret = decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present,
2083 &pulse, ics, sce->band_type);
2084 if (ret < 0)
2085 goto fail;
2086
2087 if (ac->oc[1].m4ac.object_type == AOT_AAC_MAIN && !common_window)
2088 apply_prediction(ac, sce);
2089
2090 return 0;
2091 fail:
2092 tns->present = 0;
2093 return ret;
2094 }
2095
2096 /**
2097 * Mid/Side stereo decoding; reference: 4.6.8.1.3.
2098 */
apply_mid_side_stereo(AACContext * ac,ChannelElement * cpe)2099 static void apply_mid_side_stereo(AACContext *ac, ChannelElement *cpe)
2100 {
2101 const IndividualChannelStream *ics = &cpe->ch[0].ics;
2102 INTFLOAT *ch0 = cpe->ch[0].coeffs;
2103 INTFLOAT *ch1 = cpe->ch[1].coeffs;
2104 int g, i, group, idx = 0;
2105 const uint16_t *offsets = ics->swb_offset;
2106 for (g = 0; g < ics->num_window_groups; g++) {
2107 for (i = 0; i < ics->max_sfb; i++, idx++) {
2108 if (cpe->ms_mask[idx] &&
2109 cpe->ch[0].band_type[idx] < NOISE_BT &&
2110 cpe->ch[1].band_type[idx] < NOISE_BT) {
2111 #if USE_FIXED
2112 for (group = 0; group < ics->group_len[g]; group++) {
2113 ac->fdsp->butterflies_fixed(ch0 + group * 128 + offsets[i],
2114 ch1 + group * 128 + offsets[i],
2115 offsets[i+1] - offsets[i]);
2116 #else
2117 for (group = 0; group < ics->group_len[g]; group++) {
2118 ac->fdsp->butterflies_float(ch0 + group * 128 + offsets[i],
2119 ch1 + group * 128 + offsets[i],
2120 offsets[i+1] - offsets[i]);
2121 #endif /* USE_FIXED */
2122 }
2123 }
2124 }
2125 ch0 += ics->group_len[g] * 128;
2126 ch1 += ics->group_len[g] * 128;
2127 }
2128 }
2129
2130 /**
2131 * intensity stereo decoding; reference: 4.6.8.2.3
2132 *
2133 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
2134 * [1] mask is decoded from bitstream; [2] mask is all 1s;
2135 * [3] reserved for scalable AAC
2136 */
2137 static void apply_intensity_stereo(AACContext *ac,
2138 ChannelElement *cpe, int ms_present)
2139 {
2140 const IndividualChannelStream *ics = &cpe->ch[1].ics;
2141 SingleChannelElement *sce1 = &cpe->ch[1];
2142 INTFLOAT *coef0 = cpe->ch[0].coeffs, *coef1 = cpe->ch[1].coeffs;
2143 const uint16_t *offsets = ics->swb_offset;
2144 int g, group, i, idx = 0;
2145 int c;
2146 INTFLOAT scale;
2147 for (g = 0; g < ics->num_window_groups; g++) {
2148 for (i = 0; i < ics->max_sfb;) {
2149 if (sce1->band_type[idx] == INTENSITY_BT ||
2150 sce1->band_type[idx] == INTENSITY_BT2) {
2151 const int bt_run_end = sce1->band_type_run_end[idx];
2152 for (; i < bt_run_end; i++, idx++) {
2153 c = -1 + 2 * (sce1->band_type[idx] - 14);
2154 if (ms_present)
2155 c *= 1 - 2 * cpe->ms_mask[idx];
2156 scale = c * sce1->sf[idx];
2157 for (group = 0; group < ics->group_len[g]; group++)
2158 #if USE_FIXED
2159 ac->subband_scale(coef1 + group * 128 + offsets[i],
2160 coef0 + group * 128 + offsets[i],
2161 scale,
2162 23,
2163 offsets[i + 1] - offsets[i] ,ac->avctx);
2164 #else
2165 ac->fdsp->vector_fmul_scalar(coef1 + group * 128 + offsets[i],
2166 coef0 + group * 128 + offsets[i],
2167 scale,
2168 offsets[i + 1] - offsets[i]);
2169 #endif /* USE_FIXED */
2170 }
2171 } else {
2172 int bt_run_end = sce1->band_type_run_end[idx];
2173 idx += bt_run_end - i;
2174 i = bt_run_end;
2175 }
2176 }
2177 coef0 += ics->group_len[g] * 128;
2178 coef1 += ics->group_len[g] * 128;
2179 }
2180 }
2181
2182 /**
2183 * Decode a channel_pair_element; reference: table 4.4.
2184 *
2185 * @return Returns error status. 0 - OK, !0 - error
2186 */
2187 static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe)
2188 {
2189 int i, ret, common_window, ms_present = 0;
2190 int eld_syntax = ac->oc[1].m4ac.object_type == AOT_ER_AAC_ELD;
2191
2192 common_window = eld_syntax || get_bits1(gb);
2193 if (common_window) {
2194 if (decode_ics_info(ac, &cpe->ch[0].ics, gb))
2195 return AVERROR_INVALIDDATA;
2196 i = cpe->ch[1].ics.use_kb_window[0];
2197 cpe->ch[1].ics = cpe->ch[0].ics;
2198 cpe->ch[1].ics.use_kb_window[1] = i;
2199 if (cpe->ch[1].ics.predictor_present &&
2200 (ac->oc[1].m4ac.object_type != AOT_AAC_MAIN))
2201 if ((cpe->ch[1].ics.ltp.present = get_bits(gb, 1)))
2202 decode_ltp(&cpe->ch[1].ics.ltp, gb, cpe->ch[1].ics.max_sfb);
2203 ms_present = get_bits(gb, 2);
2204 if (ms_present == 3) {
2205 av_log(ac->avctx, AV_LOG_ERROR, "ms_present = 3 is reserved.\n");
2206 return AVERROR_INVALIDDATA;
2207 } else if (ms_present)
2208 decode_mid_side_stereo(cpe, gb, ms_present);
2209 }
2210 if ((ret = decode_ics(ac, &cpe->ch[0], gb, common_window, 0)))
2211 return ret;
2212 if ((ret = decode_ics(ac, &cpe->ch[1], gb, common_window, 0)))
2213 return ret;
2214
2215 if (common_window) {
2216 if (ms_present)
2217 apply_mid_side_stereo(ac, cpe);
2218 if (ac->oc[1].m4ac.object_type == AOT_AAC_MAIN) {
2219 apply_prediction(ac, &cpe->ch[0]);
2220 apply_prediction(ac, &cpe->ch[1]);
2221 }
2222 }
2223
2224 apply_intensity_stereo(ac, cpe, ms_present);
2225 return 0;
2226 }
2227
2228 static const float cce_scale[] = {
2229 1.09050773266525765921, //2^(1/8)
2230 1.18920711500272106672, //2^(1/4)
2231 M_SQRT2,
2232 2,
2233 };
2234
2235 /**
2236 * Decode coupling_channel_element; reference: table 4.8.
2237 *
2238 * @return Returns error status. 0 - OK, !0 - error
2239 */
2240 static int decode_cce(AACContext *ac, GetBitContext *gb, ChannelElement *che)
2241 {
2242 int num_gain = 0;
2243 int c, g, sfb, ret;
2244 int sign;
2245 INTFLOAT scale;
2246 SingleChannelElement *sce = &che->ch[0];
2247 ChannelCoupling *coup = &che->coup;
2248
2249 coup->coupling_point = 2 * get_bits1(gb);
2250 coup->num_coupled = get_bits(gb, 3);
2251 for (c = 0; c <= coup->num_coupled; c++) {
2252 num_gain++;
2253 coup->type[c] = get_bits1(gb) ? TYPE_CPE : TYPE_SCE;
2254 coup->id_select[c] = get_bits(gb, 4);
2255 if (coup->type[c] == TYPE_CPE) {
2256 coup->ch_select[c] = get_bits(gb, 2);
2257 if (coup->ch_select[c] == 3)
2258 num_gain++;
2259 } else
2260 coup->ch_select[c] = 2;
2261 }
2262 coup->coupling_point += get_bits1(gb) || (coup->coupling_point >> 1);
2263
2264 sign = get_bits(gb, 1);
2265 #if USE_FIXED
2266 scale = get_bits(gb, 2);
2267 #else
2268 scale = cce_scale[get_bits(gb, 2)];
2269 #endif
2270
2271 if ((ret = decode_ics(ac, sce, gb, 0, 0)))
2272 return ret;
2273
2274 for (c = 0; c < num_gain; c++) {
2275 int idx = 0;
2276 int cge = 1;
2277 int gain = 0;
2278 INTFLOAT gain_cache = FIXR10(1.);
2279 if (c) {
2280 cge = coup->coupling_point == AFTER_IMDCT ? 1 : get_bits1(gb);
2281 gain = cge ? get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60: 0;
2282 gain_cache = GET_GAIN(scale, gain);
2283 #if USE_FIXED
2284 if ((abs(gain_cache)-1024) >> 3 > 30)
2285 return AVERROR(ERANGE);
2286 #endif
2287 }
2288 if (coup->coupling_point == AFTER_IMDCT) {
2289 coup->gain[c][0] = gain_cache;
2290 } else {
2291 for (g = 0; g < sce->ics.num_window_groups; g++) {
2292 for (sfb = 0; sfb < sce->ics.max_sfb; sfb++, idx++) {
2293 if (sce->band_type[idx] != ZERO_BT) {
2294 if (!cge) {
2295 int t = get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
2296 if (t) {
2297 int s = 1;
2298 t = gain += t;
2299 if (sign) {
2300 s -= 2 * (t & 0x1);
2301 t >>= 1;
2302 }
2303 gain_cache = GET_GAIN(scale, t) * s;
2304 #if USE_FIXED
2305 if ((abs(gain_cache)-1024) >> 3 > 30)
2306 return AVERROR(ERANGE);
2307 #endif
2308 }
2309 }
2310 coup->gain[c][idx] = gain_cache;
2311 }
2312 }
2313 }
2314 }
2315 }
2316 return 0;
2317 }
2318
2319 /**
2320 * Parse whether channels are to be excluded from Dynamic Range Compression; reference: table 4.53.
2321 *
2322 * @return Returns number of bytes consumed.
2323 */
2324 static int decode_drc_channel_exclusions(DynamicRangeControl *che_drc,
2325 GetBitContext *gb)
2326 {
2327 int i;
2328 int num_excl_chan = 0;
2329
2330 do {
2331 for (i = 0; i < 7; i++)
2332 che_drc->exclude_mask[num_excl_chan++] = get_bits1(gb);
2333 } while (num_excl_chan < MAX_CHANNELS - 7 && get_bits1(gb));
2334
2335 return num_excl_chan / 7;
2336 }
2337
2338 /**
2339 * Decode dynamic range information; reference: table 4.52.
2340 *
2341 * @return Returns number of bytes consumed.
2342 */
2343 static int decode_dynamic_range(DynamicRangeControl *che_drc,
2344 GetBitContext *gb)
2345 {
2346 int n = 1;
2347 int drc_num_bands = 1;
2348 int i;
2349
2350 /* pce_tag_present? */
2351 if (get_bits1(gb)) {
2352 che_drc->pce_instance_tag = get_bits(gb, 4);
2353 skip_bits(gb, 4); // tag_reserved_bits
2354 n++;
2355 }
2356
2357 /* excluded_chns_present? */
2358 if (get_bits1(gb)) {
2359 n += decode_drc_channel_exclusions(che_drc, gb);
2360 }
2361
2362 /* drc_bands_present? */
2363 if (get_bits1(gb)) {
2364 che_drc->band_incr = get_bits(gb, 4);
2365 che_drc->interpolation_scheme = get_bits(gb, 4);
2366 n++;
2367 drc_num_bands += che_drc->band_incr;
2368 for (i = 0; i < drc_num_bands; i++) {
2369 che_drc->band_top[i] = get_bits(gb, 8);
2370 n++;
2371 }
2372 }
2373
2374 /* prog_ref_level_present? */
2375 if (get_bits1(gb)) {
2376 che_drc->prog_ref_level = get_bits(gb, 7);
2377 skip_bits1(gb); // prog_ref_level_reserved_bits
2378 n++;
2379 }
2380
2381 for (i = 0; i < drc_num_bands; i++) {
2382 che_drc->dyn_rng_sgn[i] = get_bits1(gb);
2383 che_drc->dyn_rng_ctl[i] = get_bits(gb, 7);
2384 n++;
2385 }
2386
2387 return n;
2388 }
2389
2390 static int decode_fill(AACContext *ac, GetBitContext *gb, int len) {
2391 uint8_t buf[256];
2392 int i, major, minor;
2393
2394 if (len < 13+7*8)
2395 goto unknown;
2396
2397 get_bits(gb, 13); len -= 13;
2398
2399 for(i=0; i+1<sizeof(buf) && len>=8; i++, len-=8)
2400 buf[i] = get_bits(gb, 8);
2401
2402 buf[i] = 0;
2403 if (ac->avctx->debug & FF_DEBUG_PICT_INFO)
2404 av_log(ac->avctx, AV_LOG_DEBUG, "FILL:%s\n", buf);
2405
2406 if (sscanf(buf, "libfaac %d.%d", &major, &minor) == 2){
2407 ac->avctx->internal->skip_samples = 1024;
2408 }
2409
2410 unknown:
2411 skip_bits_long(gb, len);
2412
2413 return 0;
2414 }
2415
2416 /**
2417 * Decode extension data (incomplete); reference: table 4.51.
2418 *
2419 * @param cnt length of TYPE_FIL syntactic element in bytes
2420 *
2421 * @return Returns number of bytes consumed
2422 */
2423 static int decode_extension_payload(AACContext *ac, GetBitContext *gb, int cnt,
2424 ChannelElement *che, enum RawDataBlockType elem_type)
2425 {
2426 int crc_flag = 0;
2427 int res = cnt;
2428 int type = get_bits(gb, 4);
2429
2430 if (ac->avctx->debug & FF_DEBUG_STARTCODE)
2431 av_log(ac->avctx, AV_LOG_DEBUG, "extension type: %d len:%d\n", type, cnt);
2432
2433 switch (type) { // extension type
2434 case EXT_SBR_DATA_CRC:
2435 crc_flag++;
2436 case EXT_SBR_DATA:
2437 if (!che) {
2438 av_log(ac->avctx, AV_LOG_ERROR, "SBR was found before the first channel element.\n");
2439 return res;
2440 } else if (ac->oc[1].m4ac.frame_length_short) {
2441 if (!ac->warned_960_sbr)
2442 avpriv_report_missing_feature(ac->avctx,
2443 "SBR with 960 frame length");
2444 ac->warned_960_sbr = 1;
2445 skip_bits_long(gb, 8 * cnt - 4);
2446 return res;
2447 } else if (!ac->oc[1].m4ac.sbr) {
2448 av_log(ac->avctx, AV_LOG_ERROR, "SBR signaled to be not-present but was found in the bitstream.\n");
2449 skip_bits_long(gb, 8 * cnt - 4);
2450 return res;
2451 } else if (ac->oc[1].m4ac.sbr == -1 && ac->oc[1].status == OC_LOCKED) {
2452 av_log(ac->avctx, AV_LOG_ERROR, "Implicit SBR was found with a first occurrence after the first frame.\n");
2453 skip_bits_long(gb, 8 * cnt - 4);
2454 return res;
2455 } else if (ac->oc[1].m4ac.ps == -1 && ac->oc[1].status < OC_LOCKED && ac->avctx->channels == 1) {
2456 ac->oc[1].m4ac.sbr = 1;
2457 ac->oc[1].m4ac.ps = 1;
2458 ac->avctx->profile = FF_PROFILE_AAC_HE_V2;
2459 output_configure(ac, ac->oc[1].layout_map, ac->oc[1].layout_map_tags,
2460 ac->oc[1].status, 1);
2461 } else {
2462 ac->oc[1].m4ac.sbr = 1;
2463 ac->avctx->profile = FF_PROFILE_AAC_HE;
2464 }
2465 res = AAC_RENAME(ff_decode_sbr_extension)(ac, &che->sbr, gb, crc_flag, cnt, elem_type);
2466 break;
2467 case EXT_DYNAMIC_RANGE:
2468 res = decode_dynamic_range(&ac->che_drc, gb);
2469 break;
2470 case EXT_FILL:
2471 decode_fill(ac, gb, 8 * cnt - 4);
2472 break;
2473 case EXT_FILL_DATA:
2474 case EXT_DATA_ELEMENT:
2475 default:
2476 skip_bits_long(gb, 8 * cnt - 4);
2477 break;
2478 };
2479 return res;
2480 }
2481
2482 /**
2483 * Decode Temporal Noise Shaping filter coefficients and apply all-pole filters; reference: 4.6.9.3.
2484 *
2485 * @param decode 1 if tool is used normally, 0 if tool is used in LTP.
2486 * @param coef spectral coefficients
2487 */
2488 static void apply_tns(INTFLOAT coef_param[1024], TemporalNoiseShaping *tns,
2489 IndividualChannelStream *ics, int decode)
2490 {
2491 const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
2492 int w, filt, m, i;
2493 int bottom, top, order, start, end, size, inc;
2494 INTFLOAT lpc[TNS_MAX_ORDER];
2495 INTFLOAT tmp[TNS_MAX_ORDER+1];
2496 UINTFLOAT *coef = coef_param;
2497
2498 if(!mmm)
2499 return;
2500
2501 for (w = 0; w < ics->num_windows; w++) {
2502 bottom = ics->num_swb;
2503 for (filt = 0; filt < tns->n_filt[w]; filt++) {
2504 top = bottom;
2505 bottom = FFMAX(0, top - tns->length[w][filt]);
2506 order = tns->order[w][filt];
2507 if (order == 0)
2508 continue;
2509
2510 // tns_decode_coef
2511 AAC_RENAME(compute_lpc_coefs)(tns->coef[w][filt], order, lpc, 0, 0, 0);
2512
2513 start = ics->swb_offset[FFMIN(bottom, mmm)];
2514 end = ics->swb_offset[FFMIN( top, mmm)];
2515 if ((size = end - start) <= 0)
2516 continue;
2517 if (tns->direction[w][filt]) {
2518 inc = -1;
2519 start = end - 1;
2520 } else {
2521 inc = 1;
2522 }
2523 start += w * 128;
2524
2525 if (decode) {
2526 // ar filter
2527 for (m = 0; m < size; m++, start += inc)
2528 for (i = 1; i <= FFMIN(m, order); i++)
2529 coef[start] -= AAC_MUL26((INTFLOAT)coef[start - i * inc], lpc[i - 1]);
2530 } else {
2531 // ma filter
2532 for (m = 0; m < size; m++, start += inc) {
2533 tmp[0] = coef[start];
2534 for (i = 1; i <= FFMIN(m, order); i++)
2535 coef[start] += AAC_MUL26(tmp[i], lpc[i - 1]);
2536 for (i = order; i > 0; i--)
2537 tmp[i] = tmp[i - 1];
2538 }
2539 }
2540 }
2541 }
2542 }
2543
2544 /**
2545 * Apply windowing and MDCT to obtain the spectral
2546 * coefficient from the predicted sample by LTP.
2547 */
2548 static void windowing_and_mdct_ltp(AACContext *ac, INTFLOAT *out,
2549 INTFLOAT *in, IndividualChannelStream *ics)
2550 {
2551 const INTFLOAT *lwindow = ics->use_kb_window[0] ? AAC_RENAME(ff_aac_kbd_long_1024) : AAC_RENAME(ff_sine_1024);
2552 const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME(ff_aac_kbd_short_128) : AAC_RENAME(ff_sine_128);
2553 const INTFLOAT *lwindow_prev = ics->use_kb_window[1] ? AAC_RENAME(ff_aac_kbd_long_1024) : AAC_RENAME(ff_sine_1024);
2554 const INTFLOAT *swindow_prev = ics->use_kb_window[1] ? AAC_RENAME(ff_aac_kbd_short_128) : AAC_RENAME(ff_sine_128);
2555
2556 if (ics->window_sequence[0] != LONG_STOP_SEQUENCE) {
2557 ac->fdsp->vector_fmul(in, in, lwindow_prev, 1024);
2558 } else {
2559 memset(in, 0, 448 * sizeof(*in));
2560 ac->fdsp->vector_fmul(in + 448, in + 448, swindow_prev, 128);
2561 }
2562 if (ics->window_sequence[0] != LONG_START_SEQUENCE) {
2563 ac->fdsp->vector_fmul_reverse(in + 1024, in + 1024, lwindow, 1024);
2564 } else {
2565 ac->fdsp->vector_fmul_reverse(in + 1024 + 448, in + 1024 + 448, swindow, 128);
2566 memset(in + 1024 + 576, 0, 448 * sizeof(*in));
2567 }
2568 ac->mdct_ltp.mdct_calc(&ac->mdct_ltp, out, in);
2569 }
2570
2571 /**
2572 * Apply the long term prediction
2573 */
2574 static void apply_ltp(AACContext *ac, SingleChannelElement *sce)
2575 {
2576 const LongTermPrediction *ltp = &sce->ics.ltp;
2577 const uint16_t *offsets = sce->ics.swb_offset;
2578 int i, sfb;
2579
2580 if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
2581 INTFLOAT *predTime = sce->ret;
2582 INTFLOAT *predFreq = ac->buf_mdct;
2583 int16_t num_samples = 2048;
2584
2585 if (ltp->lag < 1024)
2586 num_samples = ltp->lag + 1024;
2587 for (i = 0; i < num_samples; i++)
2588 predTime[i] = AAC_MUL30(sce->ltp_state[i + 2048 - ltp->lag], ltp->coef);
2589 memset(&predTime[i], 0, (2048 - i) * sizeof(*predTime));
2590
2591 ac->windowing_and_mdct_ltp(ac, predFreq, predTime, &sce->ics);
2592
2593 if (sce->tns.present)
2594 ac->apply_tns(predFreq, &sce->tns, &sce->ics, 0);
2595
2596 for (sfb = 0; sfb < FFMIN(sce->ics.max_sfb, MAX_LTP_LONG_SFB); sfb++)
2597 if (ltp->used[sfb])
2598 for (i = offsets[sfb]; i < offsets[sfb + 1]; i++)
2599 sce->coeffs[i] += (UINTFLOAT)predFreq[i];
2600 }
2601 }
2602
2603 /**
2604 * Update the LTP buffer for next frame
2605 */
2606 static void update_ltp(AACContext *ac, SingleChannelElement *sce)
2607 {
2608 IndividualChannelStream *ics = &sce->ics;
2609 INTFLOAT *saved = sce->saved;
2610 INTFLOAT *saved_ltp = sce->coeffs;
2611 const INTFLOAT *lwindow = ics->use_kb_window[0] ? AAC_RENAME(ff_aac_kbd_long_1024) : AAC_RENAME(ff_sine_1024);
2612 const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME(ff_aac_kbd_short_128) : AAC_RENAME(ff_sine_128);
2613 int i;
2614
2615 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
2616 memcpy(saved_ltp, saved, 512 * sizeof(*saved_ltp));
2617 memset(saved_ltp + 576, 0, 448 * sizeof(*saved_ltp));
2618 ac->fdsp->vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960, &swindow[64], 64);
2619
2620 for (i = 0; i < 64; i++)
2621 saved_ltp[i + 512] = AAC_MUL31(ac->buf_mdct[1023 - i], swindow[63 - i]);
2622 } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
2623 memcpy(saved_ltp, ac->buf_mdct + 512, 448 * sizeof(*saved_ltp));
2624 memset(saved_ltp + 576, 0, 448 * sizeof(*saved_ltp));
2625 ac->fdsp->vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960, &swindow[64], 64);
2626
2627 for (i = 0; i < 64; i++)
2628 saved_ltp[i + 512] = AAC_MUL31(ac->buf_mdct[1023 - i], swindow[63 - i]);
2629 } else { // LONG_STOP or ONLY_LONG
2630 ac->fdsp->vector_fmul_reverse(saved_ltp, ac->buf_mdct + 512, &lwindow[512], 512);
2631
2632 for (i = 0; i < 512; i++)
2633 saved_ltp[i + 512] = AAC_MUL31(ac->buf_mdct[1023 - i], lwindow[511 - i]);
2634 }
2635
2636 memcpy(sce->ltp_state, sce->ltp_state+1024, 1024 * sizeof(*sce->ltp_state));
2637 memcpy(sce->ltp_state+1024, sce->ret, 1024 * sizeof(*sce->ltp_state));
2638 memcpy(sce->ltp_state+2048, saved_ltp, 1024 * sizeof(*sce->ltp_state));
2639 }
2640
2641 /**
2642 * Conduct IMDCT and windowing.
2643 */
2644 static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce)
2645 {
2646 IndividualChannelStream *ics = &sce->ics;
2647 INTFLOAT *in = sce->coeffs;
2648 INTFLOAT *out = sce->ret;
2649 INTFLOAT *saved = sce->saved;
2650 const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME(ff_aac_kbd_short_128) : AAC_RENAME(ff_sine_128);
2651 const INTFLOAT *lwindow_prev = ics->use_kb_window[1] ? AAC_RENAME(ff_aac_kbd_long_1024) : AAC_RENAME(ff_sine_1024);
2652 const INTFLOAT *swindow_prev = ics->use_kb_window[1] ? AAC_RENAME(ff_aac_kbd_short_128) : AAC_RENAME(ff_sine_128);
2653 INTFLOAT *buf = ac->buf_mdct;
2654 INTFLOAT *temp = ac->temp;
2655 int i;
2656
2657 // imdct
2658 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
2659 for (i = 0; i < 1024; i += 128)
2660 ac->mdct_small.imdct_half(&ac->mdct_small, buf + i, in + i);
2661 } else {
2662 ac->mdct.imdct_half(&ac->mdct, buf, in);
2663 #if USE_FIXED
2664 for (i=0; i<1024; i++)
2665 buf[i] = (buf[i] + 4LL) >> 3;
2666 #endif /* USE_FIXED */
2667 }
2668
2669 /* window overlapping
2670 * NOTE: To simplify the overlapping code, all 'meaningless' short to long
2671 * and long to short transitions are considered to be short to short
2672 * transitions. This leaves just two cases (long to long and short to short)
2673 * with a little special sauce for EIGHT_SHORT_SEQUENCE.
2674 */
2675 if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
2676 (ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
2677 ac->fdsp->vector_fmul_window( out, saved, buf, lwindow_prev, 512);
2678 } else {
2679 memcpy( out, saved, 448 * sizeof(*out));
2680
2681 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
2682 ac->fdsp->vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, 64);
2683 ac->fdsp->vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, 64);
2684 ac->fdsp->vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, 64);
2685 ac->fdsp->vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, 64);
2686 ac->fdsp->vector_fmul_window(temp, buf + 3*128 + 64, buf + 4*128, swindow, 64);
2687 memcpy( out + 448 + 4*128, temp, 64 * sizeof(*out));
2688 } else {
2689 ac->fdsp->vector_fmul_window(out + 448, saved + 448, buf, swindow_prev, 64);
2690 memcpy( out + 576, buf + 64, 448 * sizeof(*out));
2691 }
2692 }
2693
2694 // buffer update
2695 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
2696 memcpy( saved, temp + 64, 64 * sizeof(*saved));
2697 ac->fdsp->vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 64);
2698 ac->fdsp->vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 64);
2699 ac->fdsp->vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 64);
2700 memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(*saved));
2701 } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
2702 memcpy( saved, buf + 512, 448 * sizeof(*saved));
2703 memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(*saved));
2704 } else { // LONG_STOP or ONLY_LONG
2705 memcpy( saved, buf + 512, 512 * sizeof(*saved));
2706 }
2707 }
2708
2709 /**
2710 * Conduct IMDCT and windowing.
2711 */
2712 static void imdct_and_windowing_960(AACContext *ac, SingleChannelElement *sce)
2713 {
2714 #if !USE_FIXED
2715 IndividualChannelStream *ics = &sce->ics;
2716 INTFLOAT *in = sce->coeffs;
2717 INTFLOAT *out = sce->ret;
2718 INTFLOAT *saved = sce->saved;
2719 const INTFLOAT *swindow = ics->use_kb_window[0] ? AAC_RENAME(ff_aac_kbd_short_120) : AAC_RENAME(ff_sine_120);
2720 const INTFLOAT *lwindow_prev = ics->use_kb_window[1] ? AAC_RENAME(ff_aac_kbd_long_960) : AAC_RENAME(ff_sine_960);
2721 const INTFLOAT *swindow_prev = ics->use_kb_window[1] ? AAC_RENAME(ff_aac_kbd_short_120) : AAC_RENAME(ff_sine_120);
2722 INTFLOAT *buf = ac->buf_mdct;
2723 INTFLOAT *temp = ac->temp;
2724 int i;
2725
2726 // imdct
2727 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
2728 for (i = 0; i < 8; i++)
2729 ac->mdct120->imdct_half(ac->mdct120, buf + i * 120, in + i * 128, 1);
2730 } else {
2731 ac->mdct960->imdct_half(ac->mdct960, buf, in, 1);
2732 }
2733
2734 /* window overlapping
2735 * NOTE: To simplify the overlapping code, all 'meaningless' short to long
2736 * and long to short transitions are considered to be short to short
2737 * transitions. This leaves just two cases (long to long and short to short)
2738 * with a little special sauce for EIGHT_SHORT_SEQUENCE.
2739 */
2740
2741 if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
2742 (ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
2743 ac->fdsp->vector_fmul_window( out, saved, buf, lwindow_prev, 480);
2744 } else {
2745 memcpy( out, saved, 420 * sizeof(*out));
2746
2747 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
2748 ac->fdsp->vector_fmul_window(out + 420 + 0*120, saved + 420, buf + 0*120, swindow_prev, 60);
2749 ac->fdsp->vector_fmul_window(out + 420 + 1*120, buf + 0*120 + 60, buf + 1*120, swindow, 60);
2750 ac->fdsp->vector_fmul_window(out + 420 + 2*120, buf + 1*120 + 60, buf + 2*120, swindow, 60);
2751 ac->fdsp->vector_fmul_window(out + 420 + 3*120, buf + 2*120 + 60, buf + 3*120, swindow, 60);
2752 ac->fdsp->vector_fmul_window(temp, buf + 3*120 + 60, buf + 4*120, swindow, 60);
2753 memcpy( out + 420 + 4*120, temp, 60 * sizeof(*out));
2754 } else {
2755 ac->fdsp->vector_fmul_window(out + 420, saved + 420, buf, swindow_prev, 60);
2756 memcpy( out + 540, buf + 60, 420 * sizeof(*out));
2757 }
2758 }
2759
2760 // buffer update
2761 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
2762 memcpy( saved, temp + 60, 60 * sizeof(*saved));
2763 ac->fdsp->vector_fmul_window(saved + 60, buf + 4*120 + 60, buf + 5*120, swindow, 60);
2764 ac->fdsp->vector_fmul_window(saved + 180, buf + 5*120 + 60, buf + 6*120, swindow, 60);
2765 ac->fdsp->vector_fmul_window(saved + 300, buf + 6*120 + 60, buf + 7*120, swindow, 60);
2766 memcpy( saved + 420, buf + 7*120 + 60, 60 * sizeof(*saved));
2767 } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
2768 memcpy( saved, buf + 480, 420 * sizeof(*saved));
2769 memcpy( saved + 420, buf + 7*120 + 60, 60 * sizeof(*saved));
2770 } else { // LONG_STOP or ONLY_LONG
2771 memcpy( saved, buf + 480, 480 * sizeof(*saved));
2772 }
2773 #endif
2774 }
2775 static void imdct_and_windowing_ld(AACContext *ac, SingleChannelElement *sce)
2776 {
2777 IndividualChannelStream *ics = &sce->ics;
2778 INTFLOAT *in = sce->coeffs;
2779 INTFLOAT *out = sce->ret;
2780 INTFLOAT *saved = sce->saved;
2781 INTFLOAT *buf = ac->buf_mdct;
2782 #if USE_FIXED
2783 int i;
2784 #endif /* USE_FIXED */
2785
2786 // imdct
2787 ac->mdct.imdct_half(&ac->mdct_ld, buf, in);
2788
2789 #if USE_FIXED
2790 for (i = 0; i < 1024; i++)
2791 buf[i] = (buf[i] + 2) >> 2;
2792 #endif /* USE_FIXED */
2793
2794 // window overlapping
2795 if (ics->use_kb_window[1]) {
2796 // AAC LD uses a low overlap sine window instead of a KBD window
2797 memcpy(out, saved, 192 * sizeof(*out));
2798 ac->fdsp->vector_fmul_window(out + 192, saved + 192, buf, AAC_RENAME(ff_sine_128), 64);
2799 memcpy( out + 320, buf + 64, 192 * sizeof(*out));
2800 } else {
2801 ac->fdsp->vector_fmul_window(out, saved, buf, AAC_RENAME(ff_sine_512), 256);
2802 }
2803
2804 // buffer update
2805 memcpy(saved, buf + 256, 256 * sizeof(*saved));
2806 }
2807
2808 static void imdct_and_windowing_eld(AACContext *ac, SingleChannelElement *sce)
2809 {
2810 INTFLOAT *in = sce->coeffs;
2811 INTFLOAT *out = sce->ret;
2812 INTFLOAT *saved = sce->saved;
2813 INTFLOAT *buf = ac->buf_mdct;
2814 int i;
2815 const int n = ac->oc[1].m4ac.frame_length_short ? 480 : 512;
2816 const int n2 = n >> 1;
2817 const int n4 = n >> 2;
2818 const INTFLOAT *const window = n == 480 ? AAC_RENAME(ff_aac_eld_window_480) :
2819 AAC_RENAME(ff_aac_eld_window_512);
2820
2821 // Inverse transform, mapped to the conventional IMDCT by
2822 // Chivukula, R.K.; Reznik, Y.A.; Devarajan, V.,
2823 // "Efficient algorithms for MPEG-4 AAC-ELD, AAC-LD and AAC-LC filterbanks,"
2824 // International Conference on Audio, Language and Image Processing, ICALIP 2008.
2825 // URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4590245&isnumber=4589950
2826 for (i = 0; i < n2; i+=2) {
2827 INTFLOAT temp;
2828 temp = in[i ]; in[i ] = -in[n - 1 - i]; in[n - 1 - i] = temp;
2829 temp = -in[i + 1]; in[i + 1] = in[n - 2 - i]; in[n - 2 - i] = temp;
2830 }
2831 #if !USE_FIXED
2832 if (n == 480)
2833 ac->mdct480->imdct_half(ac->mdct480, buf, in, 1);
2834 else
2835 #endif
2836 ac->mdct.imdct_half(&ac->mdct_ld, buf, in);
2837
2838 #if USE_FIXED
2839 for (i = 0; i < 1024; i++)
2840 buf[i] = (buf[i] + 1) >> 1;
2841 #endif /* USE_FIXED */
2842
2843 for (i = 0; i < n; i+=2) {
2844 buf[i] = -buf[i];
2845 }
2846 // Like with the regular IMDCT at this point we still have the middle half
2847 // of a transform but with even symmetry on the left and odd symmetry on
2848 // the right
2849
2850 // window overlapping
2851 // The spec says to use samples [0..511] but the reference decoder uses
2852 // samples [128..639].
2853 for (i = n4; i < n2; i ++) {
2854 out[i - n4] = AAC_MUL31( buf[ n2 - 1 - i] , window[i - n4]) +
2855 AAC_MUL31( saved[ i + n2] , window[i + n - n4]) +
2856 AAC_MUL31(-saved[n + n2 - 1 - i] , window[i + 2*n - n4]) +
2857 AAC_MUL31(-saved[ 2*n + n2 + i] , window[i + 3*n - n4]);
2858 }
2859 for (i = 0; i < n2; i ++) {
2860 out[n4 + i] = AAC_MUL31( buf[ i] , window[i + n2 - n4]) +
2861 AAC_MUL31(-saved[ n - 1 - i] , window[i + n2 + n - n4]) +
2862 AAC_MUL31(-saved[ n + i] , window[i + n2 + 2*n - n4]) +
2863 AAC_MUL31( saved[2*n + n - 1 - i] , window[i + n2 + 3*n - n4]);
2864 }
2865 for (i = 0; i < n4; i ++) {
2866 out[n2 + n4 + i] = AAC_MUL31( buf[ i + n2] , window[i + n - n4]) +
2867 AAC_MUL31(-saved[n2 - 1 - i] , window[i + 2*n - n4]) +
2868 AAC_MUL31(-saved[n + n2 + i] , window[i + 3*n - n4]);
2869 }
2870
2871 // buffer update
2872 memmove(saved + n, saved, 2 * n * sizeof(*saved));
2873 memcpy( saved, buf, n * sizeof(*saved));
2874 }
2875
2876 /**
2877 * channel coupling transformation interface
2878 *
2879 * @param apply_coupling_method pointer to (in)dependent coupling function
2880 */
2881 static void apply_channel_coupling(AACContext *ac, ChannelElement *cc,
2882 enum RawDataBlockType type, int elem_id,
2883 enum CouplingPoint coupling_point,
2884 void (*apply_coupling_method)(AACContext *ac, SingleChannelElement *target, ChannelElement *cce, int index))
2885 {
2886 int i, c;
2887
2888 for (i = 0; i < MAX_ELEM_ID; i++) {
2889 ChannelElement *cce = ac->che[TYPE_CCE][i];
2890 int index = 0;
2891
2892 if (cce && cce->coup.coupling_point == coupling_point) {
2893 ChannelCoupling *coup = &cce->coup;
2894
2895 for (c = 0; c <= coup->num_coupled; c++) {
2896 if (coup->type[c] == type && coup->id_select[c] == elem_id) {
2897 if (coup->ch_select[c] != 1) {
2898 apply_coupling_method(ac, &cc->ch[0], cce, index);
2899 if (coup->ch_select[c] != 0)
2900 index++;
2901 }
2902 if (coup->ch_select[c] != 2)
2903 apply_coupling_method(ac, &cc->ch[1], cce, index++);
2904 } else
2905 index += 1 + (coup->ch_select[c] == 3);
2906 }
2907 }
2908 }
2909 }
2910
2911 /**
2912 * Convert spectral data to samples, applying all supported tools as appropriate.
2913 */
2914 static void spectral_to_sample(AACContext *ac, int samples)
2915 {
2916 int i, type;
2917 void (*imdct_and_window)(AACContext *ac, SingleChannelElement *sce);
2918 switch (ac->oc[1].m4ac.object_type) {
2919 case AOT_ER_AAC_LD:
2920 imdct_and_window = imdct_and_windowing_ld;
2921 break;
2922 case AOT_ER_AAC_ELD:
2923 imdct_and_window = imdct_and_windowing_eld;
2924 break;
2925 default:
2926 if (ac->oc[1].m4ac.frame_length_short)
2927 imdct_and_window = imdct_and_windowing_960;
2928 else
2929 imdct_and_window = ac->imdct_and_windowing;
2930 }
2931 for (type = 3; type >= 0; type--) {
2932 for (i = 0; i < MAX_ELEM_ID; i++) {
2933 ChannelElement *che = ac->che[type][i];
2934 if (che && che->present) {
2935 if (type <= TYPE_CPE)
2936 apply_channel_coupling(ac, che, type, i, BEFORE_TNS, AAC_RENAME(apply_dependent_coupling));
2937 if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP) {
2938 if (che->ch[0].ics.predictor_present) {
2939 if (che->ch[0].ics.ltp.present)
2940 ac->apply_ltp(ac, &che->ch[0]);
2941 if (che->ch[1].ics.ltp.present && type == TYPE_CPE)
2942 ac->apply_ltp(ac, &che->ch[1]);
2943 }
2944 }
2945 if (che->ch[0].tns.present)
2946 ac->apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1);
2947 if (che->ch[1].tns.present)
2948 ac->apply_tns(che->ch[1].coeffs, &che->ch[1].tns, &che->ch[1].ics, 1);
2949 if (type <= TYPE_CPE)
2950 apply_channel_coupling(ac, che, type, i, BETWEEN_TNS_AND_IMDCT, AAC_RENAME(apply_dependent_coupling));
2951 if (type != TYPE_CCE || che->coup.coupling_point == AFTER_IMDCT) {
2952 imdct_and_window(ac, &che->ch[0]);
2953 if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP)
2954 ac->update_ltp(ac, &che->ch[0]);
2955 if (type == TYPE_CPE) {
2956 imdct_and_window(ac, &che->ch[1]);
2957 if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP)
2958 ac->update_ltp(ac, &che->ch[1]);
2959 }
2960 if (ac->oc[1].m4ac.sbr > 0) {
2961 AAC_RENAME(ff_sbr_apply)(ac, &che->sbr, type, che->ch[0].ret, che->ch[1].ret);
2962 }
2963 }
2964 if (type <= TYPE_CCE)
2965 apply_channel_coupling(ac, che, type, i, AFTER_IMDCT, AAC_RENAME(apply_independent_coupling));
2966
2967 #if USE_FIXED
2968 {
2969 int j;
2970 /* preparation for resampler */
2971 for(j = 0; j<samples; j++){
2972 che->ch[0].ret[j] = (int32_t)av_clip64((int64_t)che->ch[0].ret[j]*128, INT32_MIN, INT32_MAX-0x8000)+0x8000;
2973 if(type == TYPE_CPE)
2974 che->ch[1].ret[j] = (int32_t)av_clip64((int64_t)che->ch[1].ret[j]*128, INT32_MIN, INT32_MAX-0x8000)+0x8000;
2975 }
2976 }
2977 #endif /* USE_FIXED */
2978 che->present = 0;
2979 } else if (che) {
2980 av_log(ac->avctx, AV_LOG_VERBOSE, "ChannelElement %d.%d missing \n", type, i);
2981 }
2982 }
2983 }
2984 }
2985
2986 static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb)
2987 {
2988 int size;
2989 AACADTSHeaderInfo hdr_info;
2990 uint8_t layout_map[MAX_ELEM_ID*4][3];
2991 int layout_map_tags, ret;
2992
2993 size = ff_adts_header_parse(gb, &hdr_info);
2994 if (size > 0) {
2995 if (!ac->warned_num_aac_frames && hdr_info.num_aac_frames != 1) {
2996 // This is 2 for "VLB " audio in NSV files.
2997 // See samples/nsv/vlb_audio.
2998 avpriv_report_missing_feature(ac->avctx,
2999 "More than one AAC RDB per ADTS frame");
3000 ac->warned_num_aac_frames = 1;
3001 }
3002 push_output_configuration(ac);
3003 if (hdr_info.chan_config) {
3004 ac->oc[1].m4ac.chan_config = hdr_info.chan_config;
3005 if ((ret = set_default_channel_config(ac, ac->avctx,
3006 layout_map,
3007 &layout_map_tags,
3008 hdr_info.chan_config)) < 0)
3009 return ret;
3010 if ((ret = output_configure(ac, layout_map, layout_map_tags,
3011 FFMAX(ac->oc[1].status,
3012 OC_TRIAL_FRAME), 0)) < 0)
3013 return ret;
3014 } else {
3015 ac->oc[1].m4ac.chan_config = 0;
3016 /**
3017 * dual mono frames in Japanese DTV can have chan_config 0
3018 * WITHOUT specifying PCE.
3019 * thus, set dual mono as default.
3020 */
3021 if (ac->dmono_mode && ac->oc[0].status == OC_NONE) {
3022 layout_map_tags = 2;
3023 layout_map[0][0] = layout_map[1][0] = TYPE_SCE;
3024 layout_map[0][2] = layout_map[1][2] = AAC_CHANNEL_FRONT;
3025 layout_map[0][1] = 0;
3026 layout_map[1][1] = 1;
3027 if (output_configure(ac, layout_map, layout_map_tags,
3028 OC_TRIAL_FRAME, 0))
3029 return -7;
3030 }
3031 }
3032 ac->oc[1].m4ac.sample_rate = hdr_info.sample_rate;
3033 ac->oc[1].m4ac.sampling_index = hdr_info.sampling_index;
3034 ac->oc[1].m4ac.object_type = hdr_info.object_type;
3035 ac->oc[1].m4ac.frame_length_short = 0;
3036 if (ac->oc[0].status != OC_LOCKED ||
3037 ac->oc[0].m4ac.chan_config != hdr_info.chan_config ||
3038 ac->oc[0].m4ac.sample_rate != hdr_info.sample_rate) {
3039 ac->oc[1].m4ac.sbr = -1;
3040 ac->oc[1].m4ac.ps = -1;
3041 }
3042 if (!hdr_info.crc_absent)
3043 skip_bits(gb, 16);
3044 }
3045 return size;
3046 }
3047
3048 static int aac_decode_er_frame(AVCodecContext *avctx, void *data,
3049 int *got_frame_ptr, GetBitContext *gb)
3050 {
3051 AACContext *ac = avctx->priv_data;
3052 const MPEG4AudioConfig *const m4ac = &ac->oc[1].m4ac;
3053 ChannelElement *che;
3054 int err, i;
3055 int samples = m4ac->frame_length_short ? 960 : 1024;
3056 int chan_config = m4ac->chan_config;
3057 int aot = m4ac->object_type;
3058
3059 if (aot == AOT_ER_AAC_LD || aot == AOT_ER_AAC_ELD)
3060 samples >>= 1;
3061
3062 ac->frame = data;
3063
3064 if ((err = frame_configure_elements(avctx)) < 0)
3065 return err;
3066
3067 // The FF_PROFILE_AAC_* defines are all object_type - 1
3068 // This may lead to an undefined profile being signaled
3069 ac->avctx->profile = aot - 1;
3070
3071 ac->tags_mapped = 0;
3072
3073 if (chan_config < 0 || (chan_config >= 8 && chan_config < 11) || chan_config >= 13) {
3074 avpriv_request_sample(avctx, "Unknown ER channel configuration %d",
3075 chan_config);
3076 return AVERROR_INVALIDDATA;
3077 }
3078 for (i = 0; i < tags_per_config[chan_config]; i++) {
3079 const int elem_type = aac_channel_layout_map[chan_config-1][i][0];
3080 const int elem_id = aac_channel_layout_map[chan_config-1][i][1];
3081 if (!(che=get_che(ac, elem_type, elem_id))) {
3082 av_log(ac->avctx, AV_LOG_ERROR,
3083 "channel element %d.%d is not allocated\n",
3084 elem_type, elem_id);
3085 return AVERROR_INVALIDDATA;
3086 }
3087 che->present = 1;
3088 if (aot != AOT_ER_AAC_ELD)
3089 skip_bits(gb, 4);
3090 switch (elem_type) {
3091 case TYPE_SCE:
3092 err = decode_ics(ac, &che->ch[0], gb, 0, 0);
3093 break;
3094 case TYPE_CPE:
3095 err = decode_cpe(ac, gb, che);
3096 break;
3097 case TYPE_LFE:
3098 err = decode_ics(ac, &che->ch[0], gb, 0, 0);
3099 break;
3100 }
3101 if (err < 0)
3102 return err;
3103 }
3104
3105 spectral_to_sample(ac, samples);
3106
3107 if (!ac->frame->data[0] && samples) {
3108 av_log(avctx, AV_LOG_ERROR, "no frame data found\n");
3109 return AVERROR_INVALIDDATA;
3110 }
3111
3112 ac->frame->nb_samples = samples;
3113 ac->frame->sample_rate = avctx->sample_rate;
3114 *got_frame_ptr = 1;
3115
3116 skip_bits_long(gb, get_bits_left(gb));
3117 return 0;
3118 }
3119
3120 static int aac_decode_frame_int(AVCodecContext *avctx, void *data,
3121 int *got_frame_ptr, GetBitContext *gb, AVPacket *avpkt)
3122 {
3123 AACContext *ac = avctx->priv_data;
3124 ChannelElement *che = NULL, *che_prev = NULL;
3125 enum RawDataBlockType elem_type, che_prev_type = TYPE_END;
3126 int err, elem_id;
3127 int samples = 0, multiplier, audio_found = 0, pce_found = 0;
3128 int is_dmono, sce_count = 0;
3129 int payload_alignment;
3130 uint8_t che_presence[4][MAX_ELEM_ID] = {{0}};
3131
3132 ac->frame = data;
3133
3134 if (show_bits(gb, 12) == 0xfff) {
3135 if ((err = parse_adts_frame_header(ac, gb)) < 0) {
3136 av_log(avctx, AV_LOG_ERROR, "Error decoding AAC frame header.\n");
3137 goto fail;
3138 }
3139 if (ac->oc[1].m4ac.sampling_index > 12) {
3140 av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->oc[1].m4ac.sampling_index);
3141 err = AVERROR_INVALIDDATA;
3142 goto fail;
3143 }
3144 }
3145
3146 if ((err = frame_configure_elements(avctx)) < 0)
3147 goto fail;
3148
3149 // The FF_PROFILE_AAC_* defines are all object_type - 1
3150 // This may lead to an undefined profile being signaled
3151 ac->avctx->profile = ac->oc[1].m4ac.object_type - 1;
3152
3153 payload_alignment = get_bits_count(gb);
3154 ac->tags_mapped = 0;
3155 // parse
3156 while ((elem_type = get_bits(gb, 3)) != TYPE_END) {
3157 elem_id = get_bits(gb, 4);
3158
3159 if (avctx->debug & FF_DEBUG_STARTCODE)
3160 av_log(avctx, AV_LOG_DEBUG, "Elem type:%x id:%x\n", elem_type, elem_id);
3161
3162 if (!avctx->channels && elem_type != TYPE_PCE) {
3163 err = AVERROR_INVALIDDATA;
3164 goto fail;
3165 }
3166
3167 if (elem_type < TYPE_DSE) {
3168 if (che_presence[elem_type][elem_id]) {
3169 int error = che_presence[elem_type][elem_id] > 1;
3170 av_log(ac->avctx, error ? AV_LOG_ERROR : AV_LOG_DEBUG, "channel element %d.%d duplicate\n",
3171 elem_type, elem_id);
3172 if (error) {
3173 err = AVERROR_INVALIDDATA;
3174 goto fail;
3175 }
3176 }
3177 che_presence[elem_type][elem_id]++;
3178
3179 if (!(che=get_che(ac, elem_type, elem_id))) {
3180 av_log(ac->avctx, AV_LOG_ERROR, "channel element %d.%d is not allocated\n",
3181 elem_type, elem_id);
3182 err = AVERROR_INVALIDDATA;
3183 goto fail;
3184 }
3185 samples = ac->oc[1].m4ac.frame_length_short ? 960 : 1024;
3186 che->present = 1;
3187 }
3188
3189 switch (elem_type) {
3190
3191 case TYPE_SCE:
3192 err = decode_ics(ac, &che->ch[0], gb, 0, 0);
3193 audio_found = 1;
3194 sce_count++;
3195 break;
3196
3197 case TYPE_CPE:
3198 err = decode_cpe(ac, gb, che);
3199 audio_found = 1;
3200 break;
3201
3202 case TYPE_CCE:
3203 err = decode_cce(ac, gb, che);
3204 break;
3205
3206 case TYPE_LFE:
3207 err = decode_ics(ac, &che->ch[0], gb, 0, 0);
3208 audio_found = 1;
3209 break;
3210
3211 case TYPE_DSE:
3212 err = skip_data_stream_element(ac, gb);
3213 break;
3214
3215 case TYPE_PCE: {
3216 uint8_t layout_map[MAX_ELEM_ID*4][3];
3217 int tags;
3218
3219 int pushed = push_output_configuration(ac);
3220 if (pce_found && !pushed) {
3221 err = AVERROR_INVALIDDATA;
3222 goto fail;
3223 }
3224
3225 tags = decode_pce(avctx, &ac->oc[1].m4ac, layout_map, gb,
3226 payload_alignment);
3227 if (tags < 0) {
3228 err = tags;
3229 break;
3230 }
3231 if (pce_found) {
3232 av_log(avctx, AV_LOG_ERROR,
3233 "Not evaluating a further program_config_element as this construct is dubious at best.\n");
3234 pop_output_configuration(ac);
3235 } else {
3236 err = output_configure(ac, layout_map, tags, OC_TRIAL_PCE, 1);
3237 if (!err)
3238 ac->oc[1].m4ac.chan_config = 0;
3239 pce_found = 1;
3240 }
3241 break;
3242 }
3243
3244 case TYPE_FIL:
3245 if (elem_id == 15)
3246 elem_id += get_bits(gb, 8) - 1;
3247 if (get_bits_left(gb) < 8 * elem_id) {
3248 av_log(avctx, AV_LOG_ERROR, "TYPE_FIL: "overread_err);
3249 err = AVERROR_INVALIDDATA;
3250 goto fail;
3251 }
3252 err = 0;
3253 while (elem_id > 0) {
3254 int ret = decode_extension_payload(ac, gb, elem_id, che_prev, che_prev_type);
3255 if (ret < 0) {
3256 err = ret;
3257 break;
3258 }
3259 elem_id -= ret;
3260 }
3261 break;
3262
3263 default:
3264 err = AVERROR_BUG; /* should not happen, but keeps compiler happy */
3265 break;
3266 }
3267
3268 if (elem_type < TYPE_DSE) {
3269 che_prev = che;
3270 che_prev_type = elem_type;
3271 }
3272
3273 if (err)
3274 goto fail;
3275
3276 if (get_bits_left(gb) < 3) {
3277 av_log(avctx, AV_LOG_ERROR, overread_err);
3278 err = AVERROR_INVALIDDATA;
3279 goto fail;
3280 }
3281 }
3282
3283 if (!avctx->channels) {
3284 *got_frame_ptr = 0;
3285 return 0;
3286 }
3287
3288 multiplier = (ac->oc[1].m4ac.sbr == 1) ? ac->oc[1].m4ac.ext_sample_rate > ac->oc[1].m4ac.sample_rate : 0;
3289 samples <<= multiplier;
3290
3291 spectral_to_sample(ac, samples);
3292
3293 if (ac->oc[1].status && audio_found) {
3294 avctx->sample_rate = ac->oc[1].m4ac.sample_rate << multiplier;
3295 avctx->frame_size = samples;
3296 ac->oc[1].status = OC_LOCKED;
3297 }
3298
3299 if (multiplier)
3300 avctx->internal->skip_samples_multiplier = 2;
3301
3302 if (!ac->frame->data[0] && samples) {
3303 av_log(avctx, AV_LOG_ERROR, "no frame data found\n");
3304 err = AVERROR_INVALIDDATA;
3305 goto fail;
3306 }
3307
3308 if (samples) {
3309 ac->frame->nb_samples = samples;
3310 ac->frame->sample_rate = avctx->sample_rate;
3311 } else
3312 av_frame_unref(ac->frame);
3313 *got_frame_ptr = !!samples;
3314
3315 /* for dual-mono audio (SCE + SCE) */
3316 is_dmono = ac->dmono_mode && sce_count == 2 &&
3317 ac->oc[1].channel_layout == (AV_CH_FRONT_LEFT | AV_CH_FRONT_RIGHT);
3318 if (is_dmono) {
3319 if (ac->dmono_mode == 1)
3320 ((AVFrame *)data)->data[1] =((AVFrame *)data)->data[0];
3321 else if (ac->dmono_mode == 2)
3322 ((AVFrame *)data)->data[0] =((AVFrame *)data)->data[1];
3323 }
3324
3325 return 0;
3326 fail:
3327 pop_output_configuration(ac);
3328 return err;
3329 }
3330
3331 static int aac_decode_frame(AVCodecContext *avctx, void *data,
3332 int *got_frame_ptr, AVPacket *avpkt)
3333 {
3334 AACContext *ac = avctx->priv_data;
3335 const uint8_t *buf = avpkt->data;
3336 int buf_size = avpkt->size;
3337 GetBitContext gb;
3338 int buf_consumed;
3339 int buf_offset;
3340 int err;
3341 int new_extradata_size;
3342 const uint8_t *new_extradata = av_packet_get_side_data(avpkt,
3343 AV_PKT_DATA_NEW_EXTRADATA,
3344 &new_extradata_size);
3345 int jp_dualmono_size;
3346 const uint8_t *jp_dualmono = av_packet_get_side_data(avpkt,
3347 AV_PKT_DATA_JP_DUALMONO,
3348 &jp_dualmono_size);
3349
3350 if (new_extradata) {
3351 /* discard previous configuration */
3352 ac->oc[1].status = OC_NONE;
3353 err = decode_audio_specific_config(ac, ac->avctx, &ac->oc[1].m4ac,
3354 new_extradata,
3355 new_extradata_size * 8LL, 1);
3356 if (err < 0) {
3357 return err;
3358 }
3359 }
3360
3361 ac->dmono_mode = 0;
3362 if (jp_dualmono && jp_dualmono_size > 0)
3363 ac->dmono_mode = 1 + *jp_dualmono;
3364 if (ac->force_dmono_mode >= 0)
3365 ac->dmono_mode = ac->force_dmono_mode;
3366
3367 if (INT_MAX / 8 <= buf_size)
3368 return AVERROR_INVALIDDATA;
3369
3370 if ((err = init_get_bits8(&gb, buf, buf_size)) < 0)
3371 return err;
3372
3373 switch (ac->oc[1].m4ac.object_type) {
3374 case AOT_ER_AAC_LC:
3375 case AOT_ER_AAC_LTP:
3376 case AOT_ER_AAC_LD:
3377 case AOT_ER_AAC_ELD:
3378 err = aac_decode_er_frame(avctx, data, got_frame_ptr, &gb);
3379 break;
3380 default:
3381 err = aac_decode_frame_int(avctx, data, got_frame_ptr, &gb, avpkt);
3382 }
3383 if (err < 0)
3384 return err;
3385
3386 buf_consumed = (get_bits_count(&gb) + 7) >> 3;
3387 for (buf_offset = buf_consumed; buf_offset < buf_size; buf_offset++)
3388 if (buf[buf_offset])
3389 break;
3390
3391 return buf_size > buf_offset ? buf_consumed : buf_size;
3392 }
3393
3394 static av_cold int aac_decode_close(AVCodecContext *avctx)
3395 {
3396 AACContext *ac = avctx->priv_data;
3397 int i, type;
3398
3399 for (i = 0; i < MAX_ELEM_ID; i++) {
3400 for (type = 0; type < 4; type++) {
3401 if (ac->che[type][i])
3402 AAC_RENAME(ff_aac_sbr_ctx_close)(&ac->che[type][i]->sbr);
3403 av_freep(&ac->che[type][i]);
3404 }
3405 }
3406
3407 ff_mdct_end(&ac->mdct);
3408 ff_mdct_end(&ac->mdct_small);
3409 ff_mdct_end(&ac->mdct_ld);
3410 ff_mdct_end(&ac->mdct_ltp);
3411 #if !USE_FIXED
3412 ff_mdct15_uninit(&ac->mdct120);
3413 ff_mdct15_uninit(&ac->mdct480);
3414 ff_mdct15_uninit(&ac->mdct960);
3415 #endif
3416 av_freep(&ac->fdsp);
3417 return 0;
3418 }
3419
3420 static void aacdec_init(AACContext *c)
3421 {
3422 c->imdct_and_windowing = imdct_and_windowing;
3423 c->apply_ltp = apply_ltp;
3424 c->apply_tns = apply_tns;
3425 c->windowing_and_mdct_ltp = windowing_and_mdct_ltp;
3426 c->update_ltp = update_ltp;
3427 #if USE_FIXED
3428 c->vector_pow43 = vector_pow43;
3429 c->subband_scale = subband_scale;
3430 #endif
3431
3432 #if !USE_FIXED
3433 if(ARCH_MIPS)
3434 ff_aacdec_init_mips(c);
3435 #endif /* !USE_FIXED */
3436 }
3437 /**
3438 * AVOptions for Japanese DTV specific extensions (ADTS only)
3439 */
3440 #define AACDEC_FLAGS AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
3441 static const AVOption options[] = {
3442 {"dual_mono_mode", "Select the channel to decode for dual mono",
3443 offsetof(AACContext, force_dmono_mode), AV_OPT_TYPE_INT, {.i64=-1}, -1, 2,
3444 AACDEC_FLAGS, "dual_mono_mode"},
3445
3446 {"auto", "autoselection", 0, AV_OPT_TYPE_CONST, {.i64=-1}, INT_MIN, INT_MAX, AACDEC_FLAGS, "dual_mono_mode"},
3447 {"main", "Select Main/Left channel", 0, AV_OPT_TYPE_CONST, {.i64= 1}, INT_MIN, INT_MAX, AACDEC_FLAGS, "dual_mono_mode"},
3448 {"sub" , "Select Sub/Right channel", 0, AV_OPT_TYPE_CONST, {.i64= 2}, INT_MIN, INT_MAX, AACDEC_FLAGS, "dual_mono_mode"},
3449 {"both", "Select both channels", 0, AV_OPT_TYPE_CONST, {.i64= 0}, INT_MIN, INT_MAX, AACDEC_FLAGS, "dual_mono_mode"},
3450
3451 {NULL},
3452 };
3453
3454 static const AVClass aac_decoder_class = {
3455 .class_name = "AAC decoder",
3456 .item_name = av_default_item_name,
3457 .option = options,
3458 .version = LIBAVUTIL_VERSION_INT,
3459 };
3460