alac : check readsamplesize to ensure get_bits() doesn't
[libav.git] / libavcodec / aac.c
CommitLineData
71e9a1b8
RS
1/*
2 * AAC decoder
3 * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
4 * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23/**
24 * @file aac.c
25 * AAC decoder
26 * @author Oded Shimon ( ods15 ods15 dyndns org )
27 * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
28 */
29
30/*
31 * supported tools
32 *
33 * Support? Name
34 * N (code in SoC repo) gain control
35 * Y block switching
36 * Y window shapes - standard
37 * N window shapes - Low Delay
38 * Y filterbank - standard
39 * N (code in SoC repo) filterbank - Scalable Sample Rate
40 * Y Temporal Noise Shaping
41 * N (code in SoC repo) Long Term Prediction
42 * Y intensity stereo
43 * Y channel coupling
44 * N frequency domain prediction
45 * Y Perceptual Noise Substitution
46 * Y Mid/Side stereo
47 * N Scalable Inverse AAC Quantization
48 * N Frequency Selective Switch
49 * N upsampling filter
50 * Y quantization & coding - AAC
51 * N quantization & coding - TwinVQ
52 * N quantization & coding - BSAC
53 * N AAC Error Resilience tools
54 * N Error Resilience payload syntax
55 * N Error Protection tool
56 * N CELP
57 * N Silence Compression
58 * N HVXC
59 * N HVXC 4kbits/s VR
60 * N Structured Audio tools
61 * N Structured Audio Sample Bank Format
62 * N MIDI
63 * N Harmonic and Individual Lines plus Noise
64 * N Text-To-Speech Interface
65 * N (in progress) Spectral Band Replication
66 * Y (not in this code) Layer-1
67 * Y (not in this code) Layer-2
68 * Y (not in this code) Layer-3
69 * N SinuSoidal Coding (Transient, Sinusoid, Noise)
70 * N (planned) Parametric Stereo
71 * N Direct Stream Transfer
72 *
73 * Note: - HE AAC v1 comprises LC AAC with Spectral Band Replication.
74 * - HE AAC v2 comprises LC AAC with Spectral Band Replication and
75 Parametric Stereo.
76 */
77
78
79#include "avcodec.h"
80#include "bitstream.h"
81#include "dsputil.h"
82
83#include "aac.h"
84#include "aactab.h"
cc0591da 85#include "aacdectab.h"
71e9a1b8
RS
86#include "mpeg4audio.h"
87
88#include <assert.h>
89#include <errno.h>
90#include <math.h>
91#include <string.h>
92
71e9a1b8
RS
93static VLC vlc_scalefactors;
94static VLC vlc_spectral[11];
95
96
9cc04edf 97/**
62a57fae
RS
98 * Configure output channel order based on the current program configuration element.
99 *
100 * @param che_pos current channel position configuration
101 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
102 *
103 * @return Returns error status. 0 - OK, !0 - error
104 */
105static int output_configure(AACContext *ac, enum ChannelPosition che_pos[4][MAX_ELEM_ID],
106 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID]) {
107 AVCodecContext *avctx = ac->avccontext;
108 int i, type, channels = 0;
109
110 if(!memcmp(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0])))
111 return 0; /* no change */
112
113 memcpy(che_pos, new_che_pos, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
114
115 /* Allocate or free elements depending on if they are in the
116 * current program configuration.
117 *
118 * Set up default 1:1 output mapping.
119 *
120 * For a 5.1 stream the output order will be:
f3399088 121 * [ Center ] [ Front Left ] [ Front Right ] [ LFE ] [ Surround Left ] [ Surround Right ]
62a57fae
RS
122 */
123
124 for(i = 0; i < MAX_ELEM_ID; i++) {
125 for(type = 0; type < 4; type++) {
126 if(che_pos[type][i]) {
127 if(!ac->che[type][i] && !(ac->che[type][i] = av_mallocz(sizeof(ChannelElement))))
128 return AVERROR(ENOMEM);
129 if(type != TYPE_CCE) {
130 ac->output_data[channels++] = ac->che[type][i]->ch[0].ret;
131 if(type == TYPE_CPE) {
132 ac->output_data[channels++] = ac->che[type][i]->ch[1].ret;
133 }
134 }
135 } else
136 av_freep(&ac->che[type][i]);
137 }
138 }
139
140 avctx->channels = channels;
141 return 0;
142}
143
144/**
9cc04edf
RS
145 * Decode an array of 4 bit element IDs, optionally interleaved with a stereo/mono switching bit.
146 *
147 * @param cpe_map Stereo (Channel Pair Element) map, NULL if stereo bit is not present.
148 * @param sce_map mono (Single Channel Element) map
149 * @param type speaker type/position for these channels
150 */
151static void decode_channel_map(enum ChannelPosition *cpe_map,
152 enum ChannelPosition *sce_map, enum ChannelPosition type, GetBitContext * gb, int n) {
153 while(n--) {
154 enum ChannelPosition *map = cpe_map && get_bits1(gb) ? cpe_map : sce_map; // stereo or mono map
155 map[get_bits(gb, 4)] = type;
156 }
157}
158
159/**
160 * Decode program configuration element; reference: table 4.2.
161 *
162 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
163 *
164 * @return Returns error status. 0 - OK, !0 - error
165 */
166static int decode_pce(AACContext * ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
167 GetBitContext * gb) {
168 int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc;
169
170 skip_bits(gb, 2); // object_type
171
172 ac->m4ac.sampling_index = get_bits(gb, 4);
173 if(ac->m4ac.sampling_index > 11) {
174 av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
175 return -1;
176 }
177 ac->m4ac.sample_rate = ff_mpeg4audio_sample_rates[ac->m4ac.sampling_index];
71e9a1b8
RS
178 num_front = get_bits(gb, 4);
179 num_side = get_bits(gb, 4);
180 num_back = get_bits(gb, 4);
181 num_lfe = get_bits(gb, 2);
182 num_assoc_data = get_bits(gb, 3);
183 num_cc = get_bits(gb, 4);
184
cc0591da
RS
185 if (get_bits1(gb))
186 skip_bits(gb, 4); // mono_mixdown_tag
187 if (get_bits1(gb))
188 skip_bits(gb, 4); // stereo_mixdown_tag
71e9a1b8 189
cc0591da
RS
190 if (get_bits1(gb))
191 skip_bits(gb, 3); // mixdown_coeff_index and pseudo_surround
71e9a1b8 192
cc0591da
RS
193 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_FRONT, gb, num_front);
194 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_SIDE, gb, num_side );
195 decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_BACK, gb, num_back );
196 decode_channel_map(NULL, new_che_pos[TYPE_LFE], AAC_CHANNEL_LFE, gb, num_lfe );
71e9a1b8
RS
197
198 skip_bits_long(gb, 4 * num_assoc_data);
199
cc0591da 200 decode_channel_map(new_che_pos[TYPE_CCE], new_che_pos[TYPE_CCE], AAC_CHANNEL_CC, gb, num_cc );
71e9a1b8
RS
201
202 align_get_bits(gb);
203
204 /* comment field, first byte is length */
205 skip_bits_long(gb, 8 * get_bits(gb, 8));
cc0591da
RS
206 return 0;
207}
71e9a1b8 208
9cc04edf
RS
209/**
210 * Set up channel positions based on a default channel configuration
211 * as specified in table 1.17.
212 *
213 * @param new_che_pos New channel position configuration - we only do something if it differs from the current one.
214 *
215 * @return Returns error status. 0 - OK, !0 - error
216 */
217static int set_default_channel_config(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
218 int channel_config)
219{
220 if(channel_config < 1 || channel_config > 7) {
221 av_log(ac->avccontext, AV_LOG_ERROR, "invalid default channel configuration (%d)\n",
222 channel_config);
223 return -1;
224 }
225
226 /* default channel configurations:
227 *
228 * 1ch : front center (mono)
229 * 2ch : L + R (stereo)
230 * 3ch : front center + L + R
231 * 4ch : front center + L + R + back center
232 * 5ch : front center + L + R + back stereo
233 * 6ch : front center + L + R + back stereo + LFE
234 * 7ch : front center + L + R + outer front left + outer front right + back stereo + LFE
235 */
236
237 if(channel_config != 2)
238 new_che_pos[TYPE_SCE][0] = AAC_CHANNEL_FRONT; // front center (or mono)
239 if(channel_config > 1)
240 new_che_pos[TYPE_CPE][0] = AAC_CHANNEL_FRONT; // L + R (or stereo)
241 if(channel_config == 4)
242 new_che_pos[TYPE_SCE][1] = AAC_CHANNEL_BACK; // back center
243 if(channel_config > 4)
244 new_che_pos[TYPE_CPE][(channel_config == 7) + 1]
245 = AAC_CHANNEL_BACK; // back stereo
246 if(channel_config > 5)
247 new_che_pos[TYPE_LFE][0] = AAC_CHANNEL_LFE; // LFE
248 if(channel_config == 7)
249 new_che_pos[TYPE_CPE][1] = AAC_CHANNEL_FRONT; // outer front left + outer front right
250
251 return 0;
252}
253
62a57fae
RS
254/**
255 * Decode GA "General Audio" specific configuration; reference: table 4.1.
256 *
257 * @return Returns error status. 0 - OK, !0 - error
258 */
259static int decode_ga_specific_config(AACContext * ac, GetBitContext * gb, int channel_config) {
260 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
261 int extension_flag, ret;
262
263 if(get_bits1(gb)) { // frameLengthFlag
264 av_log_missing_feature(ac->avccontext, "960/120 MDCT window is", 1);
9cc04edf
RS
265 return -1;
266 }
267
268 if (get_bits1(gb)) // dependsOnCoreCoder
269 skip_bits(gb, 14); // coreCoderDelay
270 extension_flag = get_bits1(gb);
271
272 if(ac->m4ac.object_type == AOT_AAC_SCALABLE ||
273 ac->m4ac.object_type == AOT_ER_AAC_SCALABLE)
274 skip_bits(gb, 3); // layerNr
275
276 memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
277 if (channel_config == 0) {
278 skip_bits(gb, 4); // element_instance_tag
279 if((ret = decode_pce(ac, new_che_pos, gb)))
280 return ret;
281 } else {
282 if((ret = set_default_channel_config(ac, new_che_pos, channel_config)))
283 return ret;
284 }
285 if((ret = output_configure(ac, ac->che_pos, new_che_pos)))
286 return ret;
287
288 if (extension_flag) {
289 switch (ac->m4ac.object_type) {
290 case AOT_ER_BSAC:
291 skip_bits(gb, 5); // numOfSubFrame
292 skip_bits(gb, 11); // layer_length
293 break;
294 case AOT_ER_AAC_LC:
295 case AOT_ER_AAC_LTP:
296 case AOT_ER_AAC_SCALABLE:
297 case AOT_ER_AAC_LD:
298 skip_bits(gb, 3); /* aacSectionDataResilienceFlag
299 * aacScalefactorDataResilienceFlag
300 * aacSpectralDataResilienceFlag
301 */
302 break;
303 }
304 skip_bits1(gb); // extensionFlag3 (TBD in version 3)
305 }
306 return 0;
307}
308
309/**
310 * Decode audio specific configuration; reference: table 1.13.
311 *
312 * @param data pointer to AVCodecContext extradata
313 * @param data_size size of AVCCodecContext extradata
314 *
315 * @return Returns error status. 0 - OK, !0 - error
316 */
317static int decode_audio_specific_config(AACContext * ac, void *data, int data_size) {
318 GetBitContext gb;
319 int i;
320
321 init_get_bits(&gb, data, data_size * 8);
322
323 if((i = ff_mpeg4audio_get_config(&ac->m4ac, data, data_size)) < 0)
324 return -1;
325 if(ac->m4ac.sampling_index > 11) {
326 av_log(ac->avccontext, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
327 return -1;
328 }
329
330 skip_bits_long(&gb, i);
331
332 switch (ac->m4ac.object_type) {
333 case AOT_AAC_LC:
334 if (decode_ga_specific_config(ac, &gb, ac->m4ac.chan_config))
335 return -1;
336 break;
337 default:
338 av_log(ac->avccontext, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n",
339 ac->m4ac.sbr == 1? "SBR+" : "", ac->m4ac.object_type);
340 return -1;
341 }
342 return 0;
343}
344
62a57fae
RS
345/**
346 * linear congruential pseudorandom number generator
347 *
348 * @param previous_val pointer to the current state of the generator
349 *
350 * @return Returns a 32-bit pseudorandom integer
351 */
352static av_always_inline int lcg_random(int previous_val) {
353 return previous_val * 1664525 + 1013904223;
354}
355
71e9a1b8
RS
356static av_cold int aac_decode_init(AVCodecContext * avccontext) {
357 AACContext * ac = avccontext->priv_data;
358 int i;
359
360 ac->avccontext = avccontext;
361
cc0591da
RS
362 if (avccontext->extradata_size <= 0 ||
363 decode_audio_specific_config(ac, avccontext->extradata, avccontext->extradata_size))
364 return -1;
365
9cc04edf 366 avccontext->sample_fmt = SAMPLE_FMT_S16;
71e9a1b8
RS
367 avccontext->sample_rate = ac->m4ac.sample_rate;
368 avccontext->frame_size = 1024;
369
370 AAC_INIT_VLC_STATIC( 0, 144);
371 AAC_INIT_VLC_STATIC( 1, 114);
372 AAC_INIT_VLC_STATIC( 2, 188);
373 AAC_INIT_VLC_STATIC( 3, 180);
374 AAC_INIT_VLC_STATIC( 4, 172);
375 AAC_INIT_VLC_STATIC( 5, 140);
376 AAC_INIT_VLC_STATIC( 6, 168);
377 AAC_INIT_VLC_STATIC( 7, 114);
378 AAC_INIT_VLC_STATIC( 8, 262);
379 AAC_INIT_VLC_STATIC( 9, 248);
380 AAC_INIT_VLC_STATIC(10, 384);
381
382 dsputil_init(&ac->dsp, avccontext);
383
9cc04edf
RS
384 ac->random_state = 0x1f2e3d4c;
385
71e9a1b8
RS
386 // -1024 - Compensate wrong IMDCT method.
387 // 32768 - Required to scale values to the correct range for the bias method
388 // for float to int16 conversion.
389
390 if(ac->dsp.float_to_int16 == ff_float_to_int16_c) {
391 ac->add_bias = 385.0f;
392 ac->sf_scale = 1. / (-1024. * 32768.);
393 ac->sf_offset = 0;
394 } else {
395 ac->add_bias = 0.0f;
396 ac->sf_scale = 1. / -1024.;
397 ac->sf_offset = 60;
398 }
399
400#ifndef CONFIG_HARDCODED_TABLES
cc0591da
RS
401 for (i = 0; i < 316; i++)
402 ff_aac_pow2sf_tab[i] = pow(2, (i - 200)/4.);
71e9a1b8
RS
403#endif /* CONFIG_HARDCODED_TABLES */
404
405 INIT_VLC_STATIC(&vlc_scalefactors, 7, sizeof(ff_aac_scalefactor_code)/sizeof(ff_aac_scalefactor_code[0]),
406 ff_aac_scalefactor_bits, sizeof(ff_aac_scalefactor_bits[0]), sizeof(ff_aac_scalefactor_bits[0]),
407 ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]),
408 352);
409
410 ff_mdct_init(&ac->mdct, 11, 1);
411 ff_mdct_init(&ac->mdct_small, 8, 1);
9ffd5c1c
RS
412 // window initialization
413 ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
414 ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
415 ff_sine_window_init(ff_sine_1024, 1024);
416 ff_sine_window_init(ff_sine_128, 128);
417
71e9a1b8
RS
418 return 0;
419}
420
9cc04edf
RS
421/**
422 * Skip data_stream_element; reference: table 4.10.
423 */
424static void skip_data_stream_element(GetBitContext * gb) {
71e9a1b8
RS
425 int byte_align = get_bits1(gb);
426 int count = get_bits(gb, 8);
427 if (count == 255)
428 count += get_bits(gb, 8);
429 if (byte_align)
430 align_get_bits(gb);
431 skip_bits_long(gb, 8 * count);
432}
433
434/**
9cc04edf
RS
435 * Decode Individual Channel Stream info; reference: table 4.6.
436 *
437 * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
438 */
439static int decode_ics_info(AACContext * ac, IndividualChannelStream * ics, GetBitContext * gb, int common_window) {
440 if (get_bits1(gb)) {
441 av_log(ac->avccontext, AV_LOG_ERROR, "Reserved bit set.\n");
442 memset(ics, 0, sizeof(IndividualChannelStream));
443 return -1;
444 }
445 ics->window_sequence[1] = ics->window_sequence[0];
446 ics->window_sequence[0] = get_bits(gb, 2);
447 ics->use_kb_window[1] = ics->use_kb_window[0];
448 ics->use_kb_window[0] = get_bits1(gb);
449 ics->num_window_groups = 1;
450 ics->group_len[0] = 1;
9ffd5c1c
RS
451 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
452 int i;
453 ics->max_sfb = get_bits(gb, 4);
454 for (i = 0; i < 7; i++) {
455 if (get_bits1(gb)) {
456 ics->group_len[ics->num_window_groups-1]++;
457 } else {
458 ics->num_window_groups++;
459 ics->group_len[ics->num_window_groups-1] = 1;
460 }
461 }
462 ics->num_windows = 8;
463 ics->swb_offset = swb_offset_128[ac->m4ac.sampling_index];
464 ics->num_swb = ff_aac_num_swb_128[ac->m4ac.sampling_index];
465 ics->tns_max_bands = tns_max_bands_128[ac->m4ac.sampling_index];
466 } else {
467 ics->max_sfb = get_bits(gb, 6);
468 ics->num_windows = 1;
469 ics->swb_offset = swb_offset_1024[ac->m4ac.sampling_index];
470 ics->num_swb = ff_aac_num_swb_1024[ac->m4ac.sampling_index];
471 ics->tns_max_bands = tns_max_bands_1024[ac->m4ac.sampling_index];
62a57fae
RS
472 if (get_bits1(gb)) {
473 av_log_missing_feature(ac->avccontext, "Predictor bit set but LTP is", 1);
474 memset(ics, 0, sizeof(IndividualChannelStream));
475 return -1;
476 }
477 }
478
479 if(ics->max_sfb > ics->num_swb) {
480 av_log(ac->avccontext, AV_LOG_ERROR,
481 "Number of scalefactor bands in group (%d) exceeds limit (%d).\n",
482 ics->max_sfb, ics->num_swb);
483 memset(ics, 0, sizeof(IndividualChannelStream));
484 return -1;
485 }
486
9cc04edf
RS
487 return 0;
488}
489
490/**
9cc04edf
RS
491 * Decode band types (section_data payload); reference: table 4.46.
492 *
493 * @param band_type array of the used band type
494 * @param band_type_run_end array of the last scalefactor band of a band type run
495 *
496 * @return Returns error status. 0 - OK, !0 - error
497 */
498static int decode_band_types(AACContext * ac, enum BandType band_type[120],
cc0591da
RS
499 int band_type_run_end[120], GetBitContext * gb, IndividualChannelStream * ics) {
500 int g, idx = 0;
501 const int bits = (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) ? 3 : 5;
502 for (g = 0; g < ics->num_window_groups; g++) {
503 int k = 0;
504 while (k < ics->max_sfb) {
505 uint8_t sect_len = k;
506 int sect_len_incr;
507 int sect_band_type = get_bits(gb, 4);
508 if (sect_band_type == 12) {
509 av_log(ac->avccontext, AV_LOG_ERROR, "invalid band type\n");
510 return -1;
511 }
512 while ((sect_len_incr = get_bits(gb, bits)) == (1 << bits)-1)
513 sect_len += sect_len_incr;
514 sect_len += sect_len_incr;
515 if (sect_len > ics->max_sfb) {
516 av_log(ac->avccontext, AV_LOG_ERROR,
517 "Number of bands (%d) exceeds limit (%d).\n",
518 sect_len, ics->max_sfb);
519 return -1;
520 }
9ffd5c1c
RS
521 for (; k < sect_len; k++) {
522 band_type [idx] = sect_band_type;
523 band_type_run_end[idx++] = sect_len;
524 }
9cc04edf
RS
525 }
526 }
527 return 0;
528}
cc0591da 529
9cc04edf
RS
530/**
531 * Decode scalefactors; reference: table 4.47.
cc0591da 532 *
cc0591da
RS
533 * @param global_gain first scalefactor value as scalefactors are differentially coded
534 * @param band_type array of the used band type
535 * @param band_type_run_end array of the last scalefactor band of a band type run
536 * @param sf array of scalefactors or intensity stereo positions
537 *
538 * @return Returns error status. 0 - OK, !0 - error
539 */
540static int decode_scalefactors(AACContext * ac, float sf[120], GetBitContext * gb,
9edae4ad 541 unsigned int global_gain, IndividualChannelStream * ics,
cc0591da
RS
542 enum BandType band_type[120], int band_type_run_end[120]) {
543 const int sf_offset = ac->sf_offset + (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE ? 12 : 0);
544 int g, i, idx = 0;
545 int offset[3] = { global_gain, global_gain - 90, 100 };
546 int noise_flag = 1;
547 static const char *sf_str[3] = { "Global gain", "Noise gain", "Intensity stereo position" };
cc0591da
RS
548 for (g = 0; g < ics->num_window_groups; g++) {
549 for (i = 0; i < ics->max_sfb;) {
550 int run_end = band_type_run_end[idx];
551 if (band_type[idx] == ZERO_BT) {
552 for(; i < run_end; i++, idx++)
553 sf[idx] = 0.;
554 }else if((band_type[idx] == INTENSITY_BT) || (band_type[idx] == INTENSITY_BT2)) {
cc0591da
RS
555 for(; i < run_end; i++, idx++) {
556 offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
557 if(offset[2] > 255U) {
558 av_log(ac->avccontext, AV_LOG_ERROR,
559 "%s (%d) out of range.\n", sf_str[2], offset[2]);
560 return -1;
561 }
562 sf[idx] = ff_aac_pow2sf_tab[-offset[2] + 300];
cc0591da
RS
563 }
564 }else if(band_type[idx] == NOISE_BT) {
565 for(; i < run_end; i++, idx++) {
566 if(noise_flag-- > 0)
567 offset[1] += get_bits(gb, 9) - 256;
568 else
569 offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
570 if(offset[1] > 255U) {
571 av_log(ac->avccontext, AV_LOG_ERROR,
572 "%s (%d) out of range.\n", sf_str[1], offset[1]);
573 return -1;
574 }
575 sf[idx] = -ff_aac_pow2sf_tab[ offset[1] + sf_offset];
cc0591da
RS
576 }
577 }else {
578 for(; i < run_end; i++, idx++) {
579 offset[0] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
580 if(offset[0] > 255U) {
581 av_log(ac->avccontext, AV_LOG_ERROR,
582 "%s (%d) out of range.\n", sf_str[0], offset[0]);
583 return -1;
584 }
585 sf[idx] = -ff_aac_pow2sf_tab[ offset[0] + sf_offset];
cc0591da
RS
586 }
587 }
588 }
589 }
590 return 0;
591}
592
593/**
594 * Decode pulse data; reference: table 4.7.
595 */
848a5815 596static void decode_pulses(Pulse * pulse, GetBitContext * gb, const uint16_t * swb_offset) {
cc0591da
RS
597 int i;
598 pulse->num_pulse = get_bits(gb, 2) + 1;
848a5815
RS
599 pulse->pos[0] = get_bits(gb, 5) + swb_offset[get_bits(gb, 6)];
600 pulse->amp[0] = get_bits(gb, 4);
601 for (i = 1; i < pulse->num_pulse; i++) {
602 pulse->pos[i] = get_bits(gb, 5) + pulse->pos[i-1];
603 pulse->amp[i] = get_bits(gb, 4);
cc0591da
RS
604 }
605}
606
607/**
1dece0d2
RS
608 * Decode Temporal Noise Shaping data; reference: table 4.48.
609 *
610 * @return Returns error status. 0 - OK, !0 - error
611 */
612static int decode_tns(AACContext * ac, TemporalNoiseShaping * tns,
613 GetBitContext * gb, const IndividualChannelStream * ics) {
614 int w, filt, i, coef_len, coef_res, coef_compress;
615 const int is8 = ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE;
616 const int tns_max_order = is8 ? 7 : ac->m4ac.object_type == AOT_AAC_MAIN ? 20 : 12;
617 for (w = 0; w < ics->num_windows; w++) {
fbd91d7c 618 if ((tns->n_filt[w] = get_bits(gb, 2 - is8))) {
1dece0d2
RS
619 coef_res = get_bits1(gb);
620
65b20b24
RS
621 for (filt = 0; filt < tns->n_filt[w]; filt++) {
622 int tmp2_idx;
623 tns->length[w][filt] = get_bits(gb, 6 - 2*is8);
624
625 if ((tns->order[w][filt] = get_bits(gb, 5 - 2*is8)) > tns_max_order) {
626 av_log(ac->avccontext, AV_LOG_ERROR, "TNS filter order %d is greater than maximum %d.",
627 tns->order[w][filt], tns_max_order);
628 tns->order[w][filt] = 0;
629 return -1;
630 }
631 tns->direction[w][filt] = get_bits1(gb);
632 coef_compress = get_bits1(gb);
633 coef_len = coef_res + 3 - coef_compress;
634 tmp2_idx = 2*coef_compress + coef_res;
1dece0d2 635
65b20b24
RS
636 for (i = 0; i < tns->order[w][filt]; i++)
637 tns->coef[w][filt][i] = tns_tmp2_map[tmp2_idx][get_bits(gb, coef_len)];
638 }
fbd91d7c 639 }
1dece0d2
RS
640 }
641 return 0;
642}
643
644/**
9cc04edf
RS
645 * Decode Mid/Side data; reference: table 4.54.
646 *
647 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
648 * [1] mask is decoded from bitstream; [2] mask is all 1s;
649 * [3] reserved for scalable AAC
650 */
651static void decode_mid_side_stereo(ChannelElement * cpe, GetBitContext * gb,
652 int ms_present) {
62a57fae
RS
653 int idx;
654 if (ms_present == 1) {
655 for (idx = 0; idx < cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb; idx++)
656 cpe->ms_mask[idx] = get_bits1(gb);
657 } else if (ms_present == 2) {
658 memset(cpe->ms_mask, 1, cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb * sizeof(cpe->ms_mask[0]));
659 }
660}
9cc04edf
RS
661
662/**
9ffd5c1c
RS
663 * Decode spectral data; reference: table 4.50.
664 * Dequantize and scale spectral data; reference: 4.6.3.3.
665 *
666 * @param coef array of dequantized, scaled spectral data
667 * @param sf array of scalefactors or intensity stereo positions
668 * @param pulse_present set if pulses are present
669 * @param pulse pointer to pulse data struct
670 * @param band_type array of the used band type
671 *
672 * @return Returns error status. 0 - OK, !0 - error
673 */
674static int decode_spectrum_and_dequant(AACContext * ac, float coef[1024], GetBitContext * gb, float sf[120],
675 int pulse_present, const Pulse * pulse, const IndividualChannelStream * ics, enum BandType band_type[120]) {
676 int i, k, g, idx = 0;
677 const int c = 1024/ics->num_windows;
678 const uint16_t * offsets = ics->swb_offset;
679 float *coef_base = coef;
680
681 for (g = 0; g < ics->num_windows; g++)
682 memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float)*(c - offsets[ics->max_sfb]));
683
684 for (g = 0; g < ics->num_window_groups; g++) {
685 for (i = 0; i < ics->max_sfb; i++, idx++) {
686 const int cur_band_type = band_type[idx];
687 const int dim = cur_band_type >= FIRST_PAIR_BT ? 2 : 4;
688 const int is_cb_unsigned = IS_CODEBOOK_UNSIGNED(cur_band_type);
689 int group;
690 if (cur_band_type == ZERO_BT) {
691 for (group = 0; group < ics->group_len[g]; group++) {
692 memset(coef + group * 128 + offsets[i], 0, (offsets[i+1] - offsets[i])*sizeof(float));
693 }
694 }else if (cur_band_type == NOISE_BT) {
695 const float scale = sf[idx] / ((offsets[i+1] - offsets[i]) * PNS_MEAN_ENERGY);
696 for (group = 0; group < ics->group_len[g]; group++) {
697 for (k = offsets[i]; k < offsets[i+1]; k++) {
698 ac->random_state = lcg_random(ac->random_state);
699 coef[group*128+k] = ac->random_state * scale;
700 }
701 }
702 }else if (cur_band_type != INTENSITY_BT2 && cur_band_type != INTENSITY_BT) {
703 for (group = 0; group < ics->group_len[g]; group++) {
704 for (k = offsets[i]; k < offsets[i+1]; k += dim) {
705 const int index = get_vlc2(gb, vlc_spectral[cur_band_type - 1].table, 6, 3);
706 const int coef_tmp_idx = (group << 7) + k;
707 const float *vq_ptr;
708 int j;
709 if(index >= ff_aac_spectral_sizes[cur_band_type - 1]) {
710 av_log(ac->avccontext, AV_LOG_ERROR,
711 "Read beyond end of ff_aac_codebook_vectors[%d][]. index %d >= %d\n",
712 cur_band_type - 1, index, ff_aac_spectral_sizes[cur_band_type - 1]);
713 return -1;
714 }
715 vq_ptr = &ff_aac_codebook_vectors[cur_band_type - 1][index * dim];
716 if (is_cb_unsigned) {
717 for (j = 0; j < dim; j++)
718 if (vq_ptr[j])
719 coef[coef_tmp_idx + j] = 1 - 2*(int)get_bits1(gb);
720 }else {
721 for (j = 0; j < dim; j++)
722 coef[coef_tmp_idx + j] = 1.0f;
723 }
724 if (cur_band_type == ESC_BT) {
725 for (j = 0; j < 2; j++) {
726 if (vq_ptr[j] == 64.0f) {
727 int n = 4;
728 /* The total length of escape_sequence must be < 22 bits according
729 to the specification (i.e. max is 11111111110xxxxxxxxxx). */
730 while (get_bits1(gb) && n < 15) n++;
731 if(n == 15) {
732 av_log(ac->avccontext, AV_LOG_ERROR, "error in spectral data, ESC overflow\n");
733 return -1;
734 }
735 n = (1<<n) + get_bits(gb, n);
736 coef[coef_tmp_idx + j] *= cbrtf(fabsf(n)) * n;
737 }else
738 coef[coef_tmp_idx + j] *= vq_ptr[j];
739 }
740 }else
741 for (j = 0; j < dim; j++)
742 coef[coef_tmp_idx + j] *= vq_ptr[j];
743 for (j = 0; j < dim; j++)
744 coef[coef_tmp_idx + j] *= sf[idx];
745 }
746 }
747 }
748 }
749 coef += ics->group_len[g]<<7;
750 }
751
752 if (pulse_present) {
753 for(i = 0; i < pulse->num_pulse; i++){
754 float co = coef_base[ pulse->pos[i] ];
755 float ico = co / sqrtf(sqrtf(fabsf(co))) + pulse->amp[i];
756 coef_base[ pulse->pos[i] ] = cbrtf(fabsf(ico)) * ico;
757 }
758 }
759 return 0;
760}
761
762/**
9cc04edf
RS
763 * Decode an individual_channel_stream payload; reference: table 4.44.
764 *
765 * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
766 * @param scale_flag scalable [1] or non-scalable [0] AAC (Unused until scalable AAC is implemented.)
767 *
768 * @return Returns error status. 0 - OK, !0 - error
769 */
770static int decode_ics(AACContext * ac, SingleChannelElement * sce, GetBitContext * gb, int common_window, int scale_flag) {
9cc04edf
RS
771 Pulse pulse;
772 TemporalNoiseShaping * tns = &sce->tns;
773 IndividualChannelStream * ics = &sce->ics;
774 float * out = sce->coeffs;
775 int global_gain, pulse_present = 0;
776
848a5815
RS
777 /* This assignment is to silence a GCC warning about the variable being used
778 * uninitialized when in fact it always is.
9cc04edf
RS
779 */
780 pulse.num_pulse = 0;
9cc04edf
RS
781
782 global_gain = get_bits(gb, 8);
783
784 if (!common_window && !scale_flag) {
785 if (decode_ics_info(ac, ics, gb, 0) < 0)
786 return -1;
787 }
788
789 if (decode_band_types(ac, sce->band_type, sce->band_type_run_end, gb, ics) < 0)
790 return -1;
791 if (decode_scalefactors(ac, sce->sf, gb, global_gain, ics, sce->band_type, sce->band_type_run_end) < 0)
792 return -1;
793
794 pulse_present = 0;
795 if (!scale_flag) {
796 if ((pulse_present = get_bits1(gb))) {
797 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
798 av_log(ac->avccontext, AV_LOG_ERROR, "Pulse tool not allowed in eight short sequence.\n");
799 return -1;
800 }
848a5815 801 decode_pulses(&pulse, gb, ics->swb_offset);
9cc04edf
RS
802 }
803 if ((tns->present = get_bits1(gb)) && decode_tns(ac, tns, gb, ics))
804 return -1;
805 if (get_bits1(gb)) {
806 av_log_missing_feature(ac->avccontext, "SSR", 1);
807 return -1;
808 }
809 }
810
848a5815 811 if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0)
9cc04edf 812 return -1;
9cc04edf
RS
813 return 0;
814}
815
816/**
9ffd5c1c
RS
817 * Mid/Side stereo decoding; reference: 4.6.8.1.3.
818 */
819static void apply_mid_side_stereo(ChannelElement * cpe) {
820 const IndividualChannelStream * ics = &cpe->ch[0].ics;
821 float *ch0 = cpe->ch[0].coeffs;
822 float *ch1 = cpe->ch[1].coeffs;
823 int g, i, k, group, idx = 0;
824 const uint16_t * offsets = ics->swb_offset;
825 for (g = 0; g < ics->num_window_groups; g++) {
826 for (i = 0; i < ics->max_sfb; i++, idx++) {
827 if (cpe->ms_mask[idx] &&
828 cpe->ch[0].band_type[idx] < NOISE_BT && cpe->ch[1].band_type[idx] < NOISE_BT) {
829 for (group = 0; group < ics->group_len[g]; group++) {
830 for (k = offsets[i]; k < offsets[i+1]; k++) {
831 float tmp = ch0[group*128 + k] - ch1[group*128 + k];
832 ch0[group*128 + k] += ch1[group*128 + k];
833 ch1[group*128 + k] = tmp;
834 }
835 }
836 }
837 }
838 ch0 += ics->group_len[g]*128;
839 ch1 += ics->group_len[g]*128;
840 }
841}
842
843/**
844 * intensity stereo decoding; reference: 4.6.8.2.3
845 *
846 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
847 * [1] mask is decoded from bitstream; [2] mask is all 1s;
848 * [3] reserved for scalable AAC
849 */
850static void apply_intensity_stereo(ChannelElement * cpe, int ms_present) {
851 const IndividualChannelStream * ics = &cpe->ch[1].ics;
852 SingleChannelElement * sce1 = &cpe->ch[1];
853 float *coef0 = cpe->ch[0].coeffs, *coef1 = cpe->ch[1].coeffs;
854 const uint16_t * offsets = ics->swb_offset;
855 int g, group, i, k, idx = 0;
856 int c;
857 float scale;
858 for (g = 0; g < ics->num_window_groups; g++) {
859 for (i = 0; i < ics->max_sfb;) {
860 if (sce1->band_type[idx] == INTENSITY_BT || sce1->band_type[idx] == INTENSITY_BT2) {
861 const int bt_run_end = sce1->band_type_run_end[idx];
862 for (; i < bt_run_end; i++, idx++) {
863 c = -1 + 2 * (sce1->band_type[idx] - 14);
864 if (ms_present)
865 c *= 1 - 2 * cpe->ms_mask[idx];
866 scale = c * sce1->sf[idx];
867 for (group = 0; group < ics->group_len[g]; group++)
868 for (k = offsets[i]; k < offsets[i+1]; k++)
869 coef1[group*128 + k] = scale * coef0[group*128 + k];
870 }
871 } else {
872 int bt_run_end = sce1->band_type_run_end[idx];
873 idx += bt_run_end - i;
874 i = bt_run_end;
875 }
876 }
877 coef0 += ics->group_len[g]*128;
878 coef1 += ics->group_len[g]*128;
879 }
880}
881
882/**
9cc04edf
RS
883 * Decode a channel_pair_element; reference: table 4.4.
884 *
885 * @param elem_id Identifies the instance of a syntax element.
886 *
887 * @return Returns error status. 0 - OK, !0 - error
888 */
889static int decode_cpe(AACContext * ac, GetBitContext * gb, int elem_id) {
890 int i, ret, common_window, ms_present = 0;
891 ChannelElement * cpe;
892
893 cpe = ac->che[TYPE_CPE][elem_id];
894 common_window = get_bits1(gb);
895 if (common_window) {
896 if (decode_ics_info(ac, &cpe->ch[0].ics, gb, 1))
897 return -1;
898 i = cpe->ch[1].ics.use_kb_window[0];
899 cpe->ch[1].ics = cpe->ch[0].ics;
900 cpe->ch[1].ics.use_kb_window[1] = i;
901 ms_present = get_bits(gb, 2);
902 if(ms_present == 3) {
903 av_log(ac->avccontext, AV_LOG_ERROR, "ms_present = 3 is reserved.\n");
904 return -1;
905 } else if(ms_present)
906 decode_mid_side_stereo(cpe, gb, ms_present);
907 }
908 if ((ret = decode_ics(ac, &cpe->ch[0], gb, common_window, 0)))
909 return ret;
910 if ((ret = decode_ics(ac, &cpe->ch[1], gb, common_window, 0)))
911 return ret;
912
913 if (common_window && ms_present)
914 apply_mid_side_stereo(cpe);
915
848a5815 916 apply_intensity_stereo(cpe, ms_present);
9cc04edf
RS
917 return 0;
918}
919
9ffd5c1c
RS
920/**
921 * Decode coupling_channel_element; reference: table 4.8.
922 *
923 * @param elem_id Identifies the instance of a syntax element.
924 *
925 * @return Returns error status. 0 - OK, !0 - error
926 */
927static int decode_cce(AACContext * ac, GetBitContext * gb, ChannelElement * che) {
928 int num_gain = 0;
929 int c, g, sfb, ret, idx = 0;
930 int sign;
931 float scale;
932 SingleChannelElement * sce = &che->ch[0];
933 ChannelCoupling * coup = &che->coup;
934
62a57fae
RS
935 coup->coupling_point = 2*get_bits1(gb);
936 coup->num_coupled = get_bits(gb, 3);
937 for (c = 0; c <= coup->num_coupled; c++) {
938 num_gain++;
939 coup->type[c] = get_bits1(gb) ? TYPE_CPE : TYPE_SCE;
940 coup->id_select[c] = get_bits(gb, 4);
941 if (coup->type[c] == TYPE_CPE) {
942 coup->ch_select[c] = get_bits(gb, 2);
943 if (coup->ch_select[c] == 3)
944 num_gain++;
945 } else
946 coup->ch_select[c] = 1;
947 }
948 coup->coupling_point += get_bits1(gb);
949
950 if (coup->coupling_point == 2) {
951 av_log(ac->avccontext, AV_LOG_ERROR,
952 "Independently switched CCE with 'invalid' domain signalled.\n");
953 memset(coup, 0, sizeof(ChannelCoupling));
954 return -1;
955 }
956
957 sign = get_bits(gb, 1);
958 scale = pow(2., pow(2., get_bits(gb, 2) - 3));
959
960 if ((ret = decode_ics(ac, sce, gb, 0, 0)))
961 return ret;
962
963 for (c = 0; c < num_gain; c++) {
964 int cge = 1;
965 int gain = 0;
966 float gain_cache = 1.;
967 if (c) {
968 cge = coup->coupling_point == AFTER_IMDCT ? 1 : get_bits1(gb);
969 gain = cge ? get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60: 0;
970 gain_cache = pow(scale, gain);
971 }
972 for (g = 0; g < sce->ics.num_window_groups; g++)
973 for (sfb = 0; sfb < sce->ics.max_sfb; sfb++, idx++)
974 if (sce->band_type[idx] != ZERO_BT) {
975 if (!cge) {
976 int t = get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
977 if (t) {
978 int s = 1;
979 if (sign) {
980 s -= 2 * (t & 0x1);
981 t >>= 1;
982 }
983 gain += t;
984 gain_cache = pow(scale, gain) * s;
985 }
986 }
987 coup->gain[c][idx] = gain_cache;
988 }
989 }
990 return 0;
991}
992
9cc04edf
RS
993/**
994 * Decode Spectral Band Replication extension data; reference: table 4.55.
cc0591da
RS
995 *
996 * @param crc flag indicating the presence of CRC checksum
997 * @param cnt length of TYPE_FIL syntactic element in bytes
9cc04edf 998 *
cc0591da
RS
999 * @return Returns number of bytes consumed from the TYPE_FIL element.
1000 */
1001static int decode_sbr_extension(AACContext * ac, GetBitContext * gb, int crc, int cnt) {
1002 // TODO : sbr_extension implementation
9edae4ad 1003 av_log_missing_feature(ac->avccontext, "SBR", 0);
cc0591da
RS
1004 skip_bits_long(gb, 8*cnt - 4); // -4 due to reading extension type
1005 return cnt;
1006}
1007
9cc04edf 1008/**
62a57fae
RS
1009 * Parse whether channels are to be excluded from Dynamic Range Compression; reference: table 4.53.
1010 *
1011 * @return Returns number of bytes consumed.
1012 */
1013static int decode_drc_channel_exclusions(DynamicRangeControl *che_drc, GetBitContext * gb) {
1014 int i;
1015 int num_excl_chan = 0;
1016
1017 do {
1018 for (i = 0; i < 7; i++)
1019 che_drc->exclude_mask[num_excl_chan++] = get_bits1(gb);
1020 } while (num_excl_chan < MAX_CHANNELS - 7 && get_bits1(gb));
1021
1022 return num_excl_chan / 7;
1023}
1024
1025/**
9cc04edf
RS
1026 * Decode dynamic range information; reference: table 4.52.
1027 *
1028 * @param cnt length of TYPE_FIL syntactic element in bytes
1029 *
1030 * @return Returns number of bytes consumed.
1031 */
1032static int decode_dynamic_range(DynamicRangeControl *che_drc, GetBitContext * gb, int cnt) {
1033 int n = 1;
1034 int drc_num_bands = 1;
1035 int i;
1036
1037 /* pce_tag_present? */
1038 if(get_bits1(gb)) {
1039 che_drc->pce_instance_tag = get_bits(gb, 4);
1040 skip_bits(gb, 4); // tag_reserved_bits
1041 n++;
1042 }
1043
1044 /* excluded_chns_present? */
1045 if(get_bits1(gb)) {
1046 n += decode_drc_channel_exclusions(che_drc, gb);
1047 }
1048
1049 /* drc_bands_present? */
1050 if (get_bits1(gb)) {
1051 che_drc->band_incr = get_bits(gb, 4);
1052 che_drc->interpolation_scheme = get_bits(gb, 4);
1053 n++;
1054 drc_num_bands += che_drc->band_incr;
1055 for (i = 0; i < drc_num_bands; i++) {
1056 che_drc->band_top[i] = get_bits(gb, 8);
1057 n++;
1058 }
1059 }
1060
1061 /* prog_ref_level_present? */
1062 if (get_bits1(gb)) {
1063 che_drc->prog_ref_level = get_bits(gb, 7);
1064 skip_bits1(gb); // prog_ref_level_reserved_bits
1065 n++;
1066 }
1067
1068 for (i = 0; i < drc_num_bands; i++) {
1069 che_drc->dyn_rng_sgn[i] = get_bits1(gb);
1070 che_drc->dyn_rng_ctl[i] = get_bits(gb, 7);
1071 n++;
1072 }
1073
1074 return n;
1075}
1076
1077/**
1078 * Decode extension data (incomplete); reference: table 4.51.
1079 *
1080 * @param cnt length of TYPE_FIL syntactic element in bytes
1081 *
1082 * @return Returns number of bytes consumed
1083 */
1084static int decode_extension_payload(AACContext * ac, GetBitContext * gb, int cnt) {
cc0591da
RS
1085 int crc_flag = 0;
1086 int res = cnt;
1087 switch (get_bits(gb, 4)) { // extension type
1088 case EXT_SBR_DATA_CRC:
1089 crc_flag++;
1090 case EXT_SBR_DATA:
1091 res = decode_sbr_extension(ac, gb, crc_flag, cnt);
1092 break;
1093 case EXT_DYNAMIC_RANGE:
1094 res = decode_dynamic_range(&ac->che_drc, gb, cnt);
1095 break;
1096 case EXT_FILL:
1097 case EXT_FILL_DATA:
1098 case EXT_DATA_ELEMENT:
1099 default:
1100 skip_bits_long(gb, 8*cnt - 4);
1101 break;
1102 };
1103 return res;
1104}
1105
7d8f3de4
RS
1106/**
1107 * Decode Temporal Noise Shaping filter coefficients and apply all-pole filters; reference: 4.6.9.3.
1108 *
1109 * @param decode 1 if tool is used normally, 0 if tool is used in LTP.
1110 * @param coef spectral coefficients
1111 */
1112static void apply_tns(float coef[1024], TemporalNoiseShaping * tns, IndividualChannelStream * ics, int decode) {
1113 const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
1098e8d2 1114 int w, filt, m, i;
7d8f3de4
RS
1115 int bottom, top, order, start, end, size, inc;
1116 float lpc[TNS_MAX_ORDER];
1117
1118 for (w = 0; w < ics->num_windows; w++) {
1119 bottom = ics->num_swb;
1120 for (filt = 0; filt < tns->n_filt[w]; filt++) {
1121 top = bottom;
1122 bottom = FFMAX(0, top - tns->length[w][filt]);
1123 order = tns->order[w][filt];
1124 if (order == 0)
1125 continue;
1126
1127 /* tns_decode_coef
1128 * FIXME: This duplicates the functionality of some double code in lpc.c.
1129 */
1130 for (m = 0; m < order; m++) {
4b0044b7
RS
1131 float tmp;
1132 lpc[m] = tns->coef[w][filt][m];
1133 for (i = 0; i < m/2; i++) {
1134 tmp = lpc[i];
1135 lpc[i] += lpc[m] * lpc[m-1-i];
1136 lpc[m-1-i] += lpc[m] * tmp;
1137 }
1138 if(m & 1)
1139 lpc[i] += lpc[m] * lpc[i];
7d8f3de4
RS
1140 }
1141
1dece0d2
RS
1142 start = ics->swb_offset[FFMIN(bottom, mmm)];
1143 end = ics->swb_offset[FFMIN( top, mmm)];
1144 if ((size = end - start) <= 0)
1145 continue;
1146 if (tns->direction[w][filt]) {
1147 inc = -1; start = end - 1;
1148 } else {
1149 inc = 1;
1150 }
1151 start += w * 128;
1152
1153 // ar filter
1154 for (m = 0; m < size; m++, start += inc)
1155 for (i = 1; i <= FFMIN(m, order); i++)
7d8f3de4 1156 coef[start] -= coef[start - i*inc] * lpc[i-1];
1dece0d2
RS
1157 }
1158 }
1159}
1160
cc0591da 1161/**
9cc04edf
RS
1162 * Conduct IMDCT and windowing.
1163 */
1164static void imdct_and_windowing(AACContext * ac, SingleChannelElement * sce) {
1165 IndividualChannelStream * ics = &sce->ics;
1166 float * in = sce->coeffs;
1167 float * out = sce->ret;
1168 float * saved = sce->saved;
848a5815
RS
1169 const float * lwindow = ics->use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
1170 const float * swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
1171 const float * lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
1172 const float * swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
9cc04edf
RS
1173 float * buf = ac->buf_mdct;
1174 int i;
1175
62a57fae
RS
1176 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
1177 if (ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE)
1178 av_log(ac->avccontext, AV_LOG_WARNING,
1179 "Transition from an ONLY_LONG or LONG_STOP to an EIGHT_SHORT sequence detected. "
1180 "If you heard an audible artifact, please submit the sample to the FFmpeg developers.\n");
1181 for (i = 0; i < 2048; i += 256) {
1182 ff_imdct_calc(&ac->mdct_small, buf + i, in + i/2);
1183 ac->dsp.vector_fmul_reverse(ac->revers + i/2, buf + i + 128, swindow, 128);
1184 }
1185 for (i = 0; i < 448; i++) out[i] = saved[i] + ac->add_bias;
1186
1187 ac->dsp.vector_fmul_add_add(out + 448 + 0*128, buf + 0*128, swindow_prev, saved + 448 , ac->add_bias, 128, 1);
1188 ac->dsp.vector_fmul_add_add(out + 448 + 1*128, buf + 2*128, swindow, ac->revers + 0*128, ac->add_bias, 128, 1);
1189 ac->dsp.vector_fmul_add_add(out + 448 + 2*128, buf + 4*128, swindow, ac->revers + 1*128, ac->add_bias, 128, 1);
1190 ac->dsp.vector_fmul_add_add(out + 448 + 3*128, buf + 6*128, swindow, ac->revers + 2*128, ac->add_bias, 128, 1);
1191 ac->dsp.vector_fmul_add_add(out + 448 + 4*128, buf + 8*128, swindow, ac->revers + 3*128, ac->add_bias, 64, 1);
1192
1193#if 0
1194 vector_fmul_add_add_add(&ac->dsp, out + 448 + 1*128, buf + 2*128, swindow, saved + 448 + 1*128, ac->revers + 0*128, ac->add_bias, 128);
1195 vector_fmul_add_add_add(&ac->dsp, out + 448 + 2*128, buf + 4*128, swindow, saved + 448 + 2*128, ac->revers + 1*128, ac->add_bias, 128);
1196 vector_fmul_add_add_add(&ac->dsp, out + 448 + 3*128, buf + 6*128, swindow, saved + 448 + 3*128, ac->revers + 2*128, ac->add_bias, 128);
1197 vector_fmul_add_add_add(&ac->dsp, out + 448 + 4*128, buf + 8*128, swindow, saved + 448 + 4*128, ac->revers + 3*128, ac->add_bias, 64);
1198#endif
1199
1200 ac->dsp.vector_fmul_add_add(saved, buf + 1024 + 64, swindow + 64, ac->revers + 3*128+64, 0, 64, 1);
1201 ac->dsp.vector_fmul_add_add(saved + 64, buf + 1024 + 2*128, swindow, ac->revers + 4*128, 0, 128, 1);
1202 ac->dsp.vector_fmul_add_add(saved + 192, buf + 1024 + 4*128, swindow, ac->revers + 5*128, 0, 128, 1);
1203 ac->dsp.vector_fmul_add_add(saved + 320, buf + 1024 + 6*128, swindow, ac->revers + 6*128, 0, 128, 1);
1204 memcpy( saved + 448, ac->revers + 7*128, 128 * sizeof(float));
1205 memset( saved + 576, 0, 448 * sizeof(float));
1206 } else {
1207 ff_imdct_calc(&ac->mdct, buf, in);
1208 if (ics->window_sequence[0] == LONG_STOP_SEQUENCE) {
1209 for (i = 0; i < 448; i++) out[i] = saved[i] + ac->add_bias;
1210 ac->dsp.vector_fmul_add_add(out + 448, buf + 448, swindow_prev, saved + 448, ac->add_bias, 128, 1);
1211 for (i = 576; i < 1024; i++) out[i] = buf[i] + saved[i] + ac->add_bias;
1212 } else {
1213 ac->dsp.vector_fmul_add_add(out, buf, lwindow_prev, saved, ac->add_bias, 1024, 1);
1214 }
1215 if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
1216 memcpy(saved, buf + 1024, 448 * sizeof(float));
1217 ac->dsp.vector_fmul_reverse(saved + 448, buf + 1024 + 448, swindow, 128);
1218 memset(saved + 576, 0, 448 * sizeof(float));
1219 } else {
1220 ac->dsp.vector_fmul_reverse(saved, buf + 1024, lwindow, 1024);
1221 }
1222 }
1223}
1224
9cc04edf 1225/**
cc0591da
RS
1226 * Apply dependent channel coupling (applied before IMDCT).
1227 *
1228 * @param index index into coupling gain array
1229 */
1230static void apply_dependent_coupling(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index) {
1231 IndividualChannelStream * ics = &cc->ch[0].ics;
1232 const uint16_t * offsets = ics->swb_offset;
1233 float * dest = sce->coeffs;
1234 const float * src = cc->ch[0].coeffs;
1235 int g, i, group, k, idx = 0;
1236 if(ac->m4ac.object_type == AOT_AAC_LTP) {
1237 av_log(ac->avccontext, AV_LOG_ERROR,
1238 "Dependent coupling is not supported together with LTP\n");
1239 return;
1240 }
1241 for (g = 0; g < ics->num_window_groups; g++) {
1242 for (i = 0; i < ics->max_sfb; i++, idx++) {
1243 if (cc->ch[0].band_type[idx] != ZERO_BT) {
cc0591da
RS
1244 for (group = 0; group < ics->group_len[g]; group++) {
1245 for (k = offsets[i]; k < offsets[i+1]; k++) {
1246 // XXX dsputil-ize
9edae4ad 1247 dest[group*128+k] += cc->coup.gain[index][idx] * src[group*128+k];
cc0591da
RS
1248 }
1249 }
1250 }
1251 }
1252 dest += ics->group_len[g]*128;
1253 src += ics->group_len[g]*128;
1254 }
1255}
1256
1257/**
1258 * Apply independent channel coupling (applied after IMDCT).
1259 *
1260 * @param index index into coupling gain array
1261 */
1262static void apply_independent_coupling(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index) {
1263 int i;
cc0591da 1264 for (i = 0; i < 1024; i++)
9edae4ad 1265 sce->ret[i] += cc->coup.gain[index][0] * (cc->ch[0].ret[i] - ac->add_bias);
cc0591da
RS
1266}
1267
9ffd5c1c
RS
1268/**
1269 * channel coupling transformation interface
1270 *
1271 * @param index index into coupling gain array
1272 * @param apply_coupling_method pointer to (in)dependent coupling function
1273 */
1274static void apply_channel_coupling(AACContext * ac, ChannelElement * cc,
1275 void (*apply_coupling_method)(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index))
1276{
1277 int c;
1278 int index = 0;
1279 ChannelCoupling * coup = &cc->coup;
1280 for (c = 0; c <= coup->num_coupled; c++) {
1281 if (ac->che[coup->type[c]][coup->id_select[c]]) {
1282 if (coup->ch_select[c] != 2) {
1283 apply_coupling_method(ac, &ac->che[coup->type[c]][coup->id_select[c]]->ch[0], cc, index);
1284 if (coup->ch_select[c] != 0)
1285 index++;
1286 }
1287 if (coup->ch_select[c] != 1)
1288 apply_coupling_method(ac, &ac->che[coup->type[c]][coup->id_select[c]]->ch[1], cc, index++);
1289 } else {
1290 av_log(ac->avccontext, AV_LOG_ERROR,
1291 "coupling target %sE[%d] not available\n",
1292 coup->type[c] == TYPE_CPE ? "CP" : "SC", coup->id_select[c]);
1293 break;
1294 }
1295 }
1296}
1297
1298/**
1299 * Convert spectral data to float samples, applying all supported tools as appropriate.
1300 */
1301static void spectral_to_sample(AACContext * ac) {
1302 int i, type;
1303 for (i = 0; i < MAX_ELEM_ID; i++) {
1304 for(type = 0; type < 4; type++) {
1305 ChannelElement *che = ac->che[type][i];
1306 if(che) {
1307 if(che->coup.coupling_point == BEFORE_TNS)
1308 apply_channel_coupling(ac, che, apply_dependent_coupling);
1309 if(che->ch[0].tns.present)
1310 apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1);
1311 if(che->ch[1].tns.present)
1312 apply_tns(che->ch[1].coeffs, &che->ch[1].tns, &che->ch[1].ics, 1);
1313 if(che->coup.coupling_point == BETWEEN_TNS_AND_IMDCT)
1314 apply_channel_coupling(ac, che, apply_dependent_coupling);
1315 imdct_and_windowing(ac, &che->ch[0]);
1316 if(type == TYPE_CPE)
1317 imdct_and_windowing(ac, &che->ch[1]);
1318 if(che->coup.coupling_point == AFTER_IMDCT)
1319 apply_channel_coupling(ac, che, apply_independent_coupling);
62a57fae
RS
1320 }
1321 }
1322 }
1323}
1324
1325static int aac_decode_frame(AVCodecContext * avccontext, void * data, int * data_size, const uint8_t * buf, int buf_size) {
1326 AACContext * ac = avccontext->priv_data;
1327 GetBitContext gb;
1328 enum RawDataBlockType elem_type;
1329 int err, elem_id, data_size_tmp;
1330
1331 init_get_bits(&gb, buf, buf_size*8);
1332
1333 // parse
1334 while ((elem_type = get_bits(&gb, 3)) != TYPE_END) {
1335 elem_id = get_bits(&gb, 4);
1336 err = -1;
1337
1338 if(elem_type == TYPE_SCE && elem_id == 1 &&
1339 !ac->che[TYPE_SCE][elem_id] && ac->che[TYPE_LFE][0]) {
1340 /* Some streams incorrectly code 5.1 audio as SCE[0] CPE[0] CPE[1] SCE[1]
1341 instead of SCE[0] CPE[0] CPE[0] LFE[0]. If we seem to have
1342 encountered such a stream, transfer the LFE[0] element to SCE[1] */
1343 ac->che[TYPE_SCE][elem_id] = ac->che[TYPE_LFE][0];
1344 ac->che[TYPE_LFE][0] = NULL;
1345 }
1346 if(elem_type && elem_type < TYPE_DSE) {
1347 if(!ac->che[elem_type][elem_id])
1348 return -1;
1349 if(elem_type != TYPE_CCE)
1350 ac->che[elem_type][elem_id]->coup.coupling_point = 4;
1351 }
1352
1353 switch (elem_type) {
1354
1355 case TYPE_SCE:
1356 err = decode_ics(ac, &ac->che[TYPE_SCE][elem_id]->ch[0], &gb, 0, 0);
1357 break;
1358
1359 case TYPE_CPE:
1360 err = decode_cpe(ac, &gb, elem_id);
1361 break;
1362
1363 case TYPE_CCE:
1364 err = decode_cce(ac, &gb, ac->che[TYPE_SCE][elem_id]);
1365 break;
1366
1367 case TYPE_LFE:
1368 err = decode_ics(ac, &ac->che[TYPE_LFE][elem_id]->ch[0], &gb, 0, 0);
1369 break;
1370
1371 case TYPE_DSE:
1372 skip_data_stream_element(&gb);
1373 err = 0;
1374 break;
1375
1376 case TYPE_PCE:
1377 {
1378 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
1379 memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
1380 if((err = decode_pce(ac, new_che_pos, &gb)))
1381 break;
1382 err = output_configure(ac, ac->che_pos, new_che_pos);
1383 break;
1384 }
1385
1386 case TYPE_FIL:
1387 if (elem_id == 15)
1388 elem_id += get_bits(&gb, 8) - 1;
1389 while (elem_id > 0)
1390 elem_id -= decode_extension_payload(ac, &gb, elem_id);
1391 err = 0; /* FIXME */
1392 break;
1393
1394 default:
1395 err = -1; /* should not happen, but keeps compiler happy */
1396 break;
1397 }
1398
1399 if(err)
1400 return err;
1401 }
1402
1403 spectral_to_sample(ac);
1404
9cc04edf
RS
1405 if (!ac->is_saved) {
1406 ac->is_saved = 1;
1407 *data_size = 0;
848a5815 1408 return buf_size;
9cc04edf
RS
1409 }
1410
1411 data_size_tmp = 1024 * avccontext->channels * sizeof(int16_t);
1412 if(*data_size < data_size_tmp) {
1413 av_log(avccontext, AV_LOG_ERROR,
1414 "Output buffer too small (%d) or trying to output too many samples (%d) for this frame.\n",
1415 *data_size, data_size_tmp);
1416 return -1;
1417 }
1418 *data_size = data_size_tmp;
1419
1420 ac->dsp.float_to_int16_interleave(data, (const float **)ac->output_data, 1024, avccontext->channels);
1421
1422 return buf_size;
1423}
1424
71e9a1b8
RS
1425static av_cold int aac_decode_close(AVCodecContext * avccontext) {
1426 AACContext * ac = avccontext->priv_data;
9edae4ad 1427 int i, type;
71e9a1b8 1428
cc0591da 1429 for (i = 0; i < MAX_ELEM_ID; i++) {
9edae4ad
RS
1430 for(type = 0; type < 4; type++)
1431 av_freep(&ac->che[type][i]);
71e9a1b8
RS
1432 }
1433
1434 ff_mdct_end(&ac->mdct);
1435 ff_mdct_end(&ac->mdct_small);
71e9a1b8
RS
1436 return 0 ;
1437}
1438
1439AVCodec aac_decoder = {
1440 "aac",
1441 CODEC_TYPE_AUDIO,
1442 CODEC_ID_AAC,
1443 sizeof(AACContext),
1444 aac_decode_init,
1445 NULL,
1446 aac_decode_close,
1447 aac_decode_frame,
1448 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
cc0591da 1449 .sample_fmts = (enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE},
71e9a1b8 1450};