Simplify timestamp reordering by using the new API.
[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:
121 * [ Front Left ] [ Front Right ] [ Center ] [ LFE ] [ Surround Left ] [ Surround Right ]
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++) {
618 tns->n_filt[w] = get_bits(gb, 2 - is8);
619
620 if (tns->n_filt[w])
621 coef_res = get_bits1(gb);
622
623 for (filt = 0; filt < tns->n_filt[w]; filt++) {
624 int tmp2_idx;
625 tns->length[w][filt] = get_bits(gb, 6 - 2*is8);
626
627 if ((tns->order[w][filt] = get_bits(gb, 5 - 2*is8)) > tns_max_order) {
628 av_log(ac->avccontext, AV_LOG_ERROR, "TNS filter order %d is greater than maximum %d.",
629 tns->order[w][filt], tns_max_order);
630 tns->order[w][filt] = 0;
631 return -1;
632 }
633 tns->direction[w][filt] = get_bits1(gb);
634 coef_compress = get_bits1(gb);
635 coef_len = coef_res + 3 - coef_compress;
636 tmp2_idx = 2*coef_compress + coef_res;
637
638 for (i = 0; i < tns->order[w][filt]; i++)
639 tns->coef[w][filt][i] = tns_tmp2_map[tmp2_idx][get_bits(gb, coef_len)];
640 }
641 }
642 return 0;
643}
644
645/**
9cc04edf
RS
646 * Decode Mid/Side data; reference: table 4.54.
647 *
648 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
649 * [1] mask is decoded from bitstream; [2] mask is all 1s;
650 * [3] reserved for scalable AAC
651 */
652static void decode_mid_side_stereo(ChannelElement * cpe, GetBitContext * gb,
653 int ms_present) {
62a57fae
RS
654 int idx;
655 if (ms_present == 1) {
656 for (idx = 0; idx < cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb; idx++)
657 cpe->ms_mask[idx] = get_bits1(gb);
658 } else if (ms_present == 2) {
659 memset(cpe->ms_mask, 1, cpe->ch[0].ics.num_window_groups * cpe->ch[0].ics.max_sfb * sizeof(cpe->ms_mask[0]));
660 }
661}
9cc04edf
RS
662
663/**
9ffd5c1c
RS
664 * Decode spectral data; reference: table 4.50.
665 * Dequantize and scale spectral data; reference: 4.6.3.3.
666 *
667 * @param coef array of dequantized, scaled spectral data
668 * @param sf array of scalefactors or intensity stereo positions
669 * @param pulse_present set if pulses are present
670 * @param pulse pointer to pulse data struct
671 * @param band_type array of the used band type
672 *
673 * @return Returns error status. 0 - OK, !0 - error
674 */
675static int decode_spectrum_and_dequant(AACContext * ac, float coef[1024], GetBitContext * gb, float sf[120],
676 int pulse_present, const Pulse * pulse, const IndividualChannelStream * ics, enum BandType band_type[120]) {
677 int i, k, g, idx = 0;
678 const int c = 1024/ics->num_windows;
679 const uint16_t * offsets = ics->swb_offset;
680 float *coef_base = coef;
681
682 for (g = 0; g < ics->num_windows; g++)
683 memset(coef + g * 128 + offsets[ics->max_sfb], 0, sizeof(float)*(c - offsets[ics->max_sfb]));
684
685 for (g = 0; g < ics->num_window_groups; g++) {
686 for (i = 0; i < ics->max_sfb; i++, idx++) {
687 const int cur_band_type = band_type[idx];
688 const int dim = cur_band_type >= FIRST_PAIR_BT ? 2 : 4;
689 const int is_cb_unsigned = IS_CODEBOOK_UNSIGNED(cur_band_type);
690 int group;
691 if (cur_band_type == ZERO_BT) {
692 for (group = 0; group < ics->group_len[g]; group++) {
693 memset(coef + group * 128 + offsets[i], 0, (offsets[i+1] - offsets[i])*sizeof(float));
694 }
695 }else if (cur_band_type == NOISE_BT) {
696 const float scale = sf[idx] / ((offsets[i+1] - offsets[i]) * PNS_MEAN_ENERGY);
697 for (group = 0; group < ics->group_len[g]; group++) {
698 for (k = offsets[i]; k < offsets[i+1]; k++) {
699 ac->random_state = lcg_random(ac->random_state);
700 coef[group*128+k] = ac->random_state * scale;
701 }
702 }
703 }else if (cur_band_type != INTENSITY_BT2 && cur_band_type != INTENSITY_BT) {
704 for (group = 0; group < ics->group_len[g]; group++) {
705 for (k = offsets[i]; k < offsets[i+1]; k += dim) {
706 const int index = get_vlc2(gb, vlc_spectral[cur_band_type - 1].table, 6, 3);
707 const int coef_tmp_idx = (group << 7) + k;
708 const float *vq_ptr;
709 int j;
710 if(index >= ff_aac_spectral_sizes[cur_band_type - 1]) {
711 av_log(ac->avccontext, AV_LOG_ERROR,
712 "Read beyond end of ff_aac_codebook_vectors[%d][]. index %d >= %d\n",
713 cur_band_type - 1, index, ff_aac_spectral_sizes[cur_band_type - 1]);
714 return -1;
715 }
716 vq_ptr = &ff_aac_codebook_vectors[cur_band_type - 1][index * dim];
717 if (is_cb_unsigned) {
718 for (j = 0; j < dim; j++)
719 if (vq_ptr[j])
720 coef[coef_tmp_idx + j] = 1 - 2*(int)get_bits1(gb);
721 }else {
722 for (j = 0; j < dim; j++)
723 coef[coef_tmp_idx + j] = 1.0f;
724 }
725 if (cur_band_type == ESC_BT) {
726 for (j = 0; j < 2; j++) {
727 if (vq_ptr[j] == 64.0f) {
728 int n = 4;
729 /* The total length of escape_sequence must be < 22 bits according
730 to the specification (i.e. max is 11111111110xxxxxxxxxx). */
731 while (get_bits1(gb) && n < 15) n++;
732 if(n == 15) {
733 av_log(ac->avccontext, AV_LOG_ERROR, "error in spectral data, ESC overflow\n");
734 return -1;
735 }
736 n = (1<<n) + get_bits(gb, n);
737 coef[coef_tmp_idx + j] *= cbrtf(fabsf(n)) * n;
738 }else
739 coef[coef_tmp_idx + j] *= vq_ptr[j];
740 }
741 }else
742 for (j = 0; j < dim; j++)
743 coef[coef_tmp_idx + j] *= vq_ptr[j];
744 for (j = 0; j < dim; j++)
745 coef[coef_tmp_idx + j] *= sf[idx];
746 }
747 }
748 }
749 }
750 coef += ics->group_len[g]<<7;
751 }
752
753 if (pulse_present) {
754 for(i = 0; i < pulse->num_pulse; i++){
755 float co = coef_base[ pulse->pos[i] ];
756 float ico = co / sqrtf(sqrtf(fabsf(co))) + pulse->amp[i];
757 coef_base[ pulse->pos[i] ] = cbrtf(fabsf(ico)) * ico;
758 }
759 }
760 return 0;
761}
762
763/**
9cc04edf
RS
764 * Decode an individual_channel_stream payload; reference: table 4.44.
765 *
766 * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
767 * @param scale_flag scalable [1] or non-scalable [0] AAC (Unused until scalable AAC is implemented.)
768 *
769 * @return Returns error status. 0 - OK, !0 - error
770 */
771static int decode_ics(AACContext * ac, SingleChannelElement * sce, GetBitContext * gb, int common_window, int scale_flag) {
9cc04edf
RS
772 Pulse pulse;
773 TemporalNoiseShaping * tns = &sce->tns;
774 IndividualChannelStream * ics = &sce->ics;
775 float * out = sce->coeffs;
776 int global_gain, pulse_present = 0;
777
848a5815
RS
778 /* This assignment is to silence a GCC warning about the variable being used
779 * uninitialized when in fact it always is.
9cc04edf
RS
780 */
781 pulse.num_pulse = 0;
9cc04edf
RS
782
783 global_gain = get_bits(gb, 8);
784
785 if (!common_window && !scale_flag) {
786 if (decode_ics_info(ac, ics, gb, 0) < 0)
787 return -1;
788 }
789
790 if (decode_band_types(ac, sce->band_type, sce->band_type_run_end, gb, ics) < 0)
791 return -1;
792 if (decode_scalefactors(ac, sce->sf, gb, global_gain, ics, sce->band_type, sce->band_type_run_end) < 0)
793 return -1;
794
795 pulse_present = 0;
796 if (!scale_flag) {
797 if ((pulse_present = get_bits1(gb))) {
798 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
799 av_log(ac->avccontext, AV_LOG_ERROR, "Pulse tool not allowed in eight short sequence.\n");
800 return -1;
801 }
848a5815 802 decode_pulses(&pulse, gb, ics->swb_offset);
9cc04edf
RS
803 }
804 if ((tns->present = get_bits1(gb)) && decode_tns(ac, tns, gb, ics))
805 return -1;
806 if (get_bits1(gb)) {
807 av_log_missing_feature(ac->avccontext, "SSR", 1);
808 return -1;
809 }
810 }
811
848a5815 812 if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0)
9cc04edf 813 return -1;
9cc04edf
RS
814 return 0;
815}
816
817/**
9ffd5c1c
RS
818 * Mid/Side stereo decoding; reference: 4.6.8.1.3.
819 */
820static void apply_mid_side_stereo(ChannelElement * cpe) {
821 const IndividualChannelStream * ics = &cpe->ch[0].ics;
822 float *ch0 = cpe->ch[0].coeffs;
823 float *ch1 = cpe->ch[1].coeffs;
824 int g, i, k, group, idx = 0;
825 const uint16_t * offsets = ics->swb_offset;
826 for (g = 0; g < ics->num_window_groups; g++) {
827 for (i = 0; i < ics->max_sfb; i++, idx++) {
828 if (cpe->ms_mask[idx] &&
829 cpe->ch[0].band_type[idx] < NOISE_BT && cpe->ch[1].band_type[idx] < NOISE_BT) {
830 for (group = 0; group < ics->group_len[g]; group++) {
831 for (k = offsets[i]; k < offsets[i+1]; k++) {
832 float tmp = ch0[group*128 + k] - ch1[group*128 + k];
833 ch0[group*128 + k] += ch1[group*128 + k];
834 ch1[group*128 + k] = tmp;
835 }
836 }
837 }
838 }
839 ch0 += ics->group_len[g]*128;
840 ch1 += ics->group_len[g]*128;
841 }
842}
843
844/**
845 * intensity stereo decoding; reference: 4.6.8.2.3
846 *
847 * @param ms_present Indicates mid/side stereo presence. [0] mask is all 0s;
848 * [1] mask is decoded from bitstream; [2] mask is all 1s;
849 * [3] reserved for scalable AAC
850 */
851static void apply_intensity_stereo(ChannelElement * cpe, int ms_present) {
852 const IndividualChannelStream * ics = &cpe->ch[1].ics;
853 SingleChannelElement * sce1 = &cpe->ch[1];
854 float *coef0 = cpe->ch[0].coeffs, *coef1 = cpe->ch[1].coeffs;
855 const uint16_t * offsets = ics->swb_offset;
856 int g, group, i, k, idx = 0;
857 int c;
858 float scale;
859 for (g = 0; g < ics->num_window_groups; g++) {
860 for (i = 0; i < ics->max_sfb;) {
861 if (sce1->band_type[idx] == INTENSITY_BT || sce1->band_type[idx] == INTENSITY_BT2) {
862 const int bt_run_end = sce1->band_type_run_end[idx];
863 for (; i < bt_run_end; i++, idx++) {
864 c = -1 + 2 * (sce1->band_type[idx] - 14);
865 if (ms_present)
866 c *= 1 - 2 * cpe->ms_mask[idx];
867 scale = c * sce1->sf[idx];
868 for (group = 0; group < ics->group_len[g]; group++)
869 for (k = offsets[i]; k < offsets[i+1]; k++)
870 coef1[group*128 + k] = scale * coef0[group*128 + k];
871 }
872 } else {
873 int bt_run_end = sce1->band_type_run_end[idx];
874 idx += bt_run_end - i;
875 i = bt_run_end;
876 }
877 }
878 coef0 += ics->group_len[g]*128;
879 coef1 += ics->group_len[g]*128;
880 }
881}
882
883/**
9cc04edf
RS
884 * Decode a channel_pair_element; reference: table 4.4.
885 *
886 * @param elem_id Identifies the instance of a syntax element.
887 *
888 * @return Returns error status. 0 - OK, !0 - error
889 */
890static int decode_cpe(AACContext * ac, GetBitContext * gb, int elem_id) {
891 int i, ret, common_window, ms_present = 0;
892 ChannelElement * cpe;
893
894 cpe = ac->che[TYPE_CPE][elem_id];
895 common_window = get_bits1(gb);
896 if (common_window) {
897 if (decode_ics_info(ac, &cpe->ch[0].ics, gb, 1))
898 return -1;
899 i = cpe->ch[1].ics.use_kb_window[0];
900 cpe->ch[1].ics = cpe->ch[0].ics;
901 cpe->ch[1].ics.use_kb_window[1] = i;
902 ms_present = get_bits(gb, 2);
903 if(ms_present == 3) {
904 av_log(ac->avccontext, AV_LOG_ERROR, "ms_present = 3 is reserved.\n");
905 return -1;
906 } else if(ms_present)
907 decode_mid_side_stereo(cpe, gb, ms_present);
908 }
909 if ((ret = decode_ics(ac, &cpe->ch[0], gb, common_window, 0)))
910 return ret;
911 if ((ret = decode_ics(ac, &cpe->ch[1], gb, common_window, 0)))
912 return ret;
913
914 if (common_window && ms_present)
915 apply_mid_side_stereo(cpe);
916
848a5815 917 apply_intensity_stereo(cpe, ms_present);
9cc04edf
RS
918 return 0;
919}
920
9ffd5c1c
RS
921/**
922 * Decode coupling_channel_element; reference: table 4.8.
923 *
924 * @param elem_id Identifies the instance of a syntax element.
925 *
926 * @return Returns error status. 0 - OK, !0 - error
927 */
928static int decode_cce(AACContext * ac, GetBitContext * gb, ChannelElement * che) {
929 int num_gain = 0;
930 int c, g, sfb, ret, idx = 0;
931 int sign;
932 float scale;
933 SingleChannelElement * sce = &che->ch[0];
934 ChannelCoupling * coup = &che->coup;
935
62a57fae
RS
936 coup->coupling_point = 2*get_bits1(gb);
937 coup->num_coupled = get_bits(gb, 3);
938 for (c = 0; c <= coup->num_coupled; c++) {
939 num_gain++;
940 coup->type[c] = get_bits1(gb) ? TYPE_CPE : TYPE_SCE;
941 coup->id_select[c] = get_bits(gb, 4);
942 if (coup->type[c] == TYPE_CPE) {
943 coup->ch_select[c] = get_bits(gb, 2);
944 if (coup->ch_select[c] == 3)
945 num_gain++;
946 } else
947 coup->ch_select[c] = 1;
948 }
949 coup->coupling_point += get_bits1(gb);
950
951 if (coup->coupling_point == 2) {
952 av_log(ac->avccontext, AV_LOG_ERROR,
953 "Independently switched CCE with 'invalid' domain signalled.\n");
954 memset(coup, 0, sizeof(ChannelCoupling));
955 return -1;
956 }
957
958 sign = get_bits(gb, 1);
959 scale = pow(2., pow(2., get_bits(gb, 2) - 3));
960
961 if ((ret = decode_ics(ac, sce, gb, 0, 0)))
962 return ret;
963
964 for (c = 0; c < num_gain; c++) {
965 int cge = 1;
966 int gain = 0;
967 float gain_cache = 1.;
968 if (c) {
969 cge = coup->coupling_point == AFTER_IMDCT ? 1 : get_bits1(gb);
970 gain = cge ? get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60: 0;
971 gain_cache = pow(scale, gain);
972 }
973 for (g = 0; g < sce->ics.num_window_groups; g++)
974 for (sfb = 0; sfb < sce->ics.max_sfb; sfb++, idx++)
975 if (sce->band_type[idx] != ZERO_BT) {
976 if (!cge) {
977 int t = get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
978 if (t) {
979 int s = 1;
980 if (sign) {
981 s -= 2 * (t & 0x1);
982 t >>= 1;
983 }
984 gain += t;
985 gain_cache = pow(scale, gain) * s;
986 }
987 }
988 coup->gain[c][idx] = gain_cache;
989 }
990 }
991 return 0;
992}
993
9cc04edf
RS
994/**
995 * Decode Spectral Band Replication extension data; reference: table 4.55.
cc0591da
RS
996 *
997 * @param crc flag indicating the presence of CRC checksum
998 * @param cnt length of TYPE_FIL syntactic element in bytes
9cc04edf 999 *
cc0591da
RS
1000 * @return Returns number of bytes consumed from the TYPE_FIL element.
1001 */
1002static int decode_sbr_extension(AACContext * ac, GetBitContext * gb, int crc, int cnt) {
1003 // TODO : sbr_extension implementation
9edae4ad 1004 av_log_missing_feature(ac->avccontext, "SBR", 0);
cc0591da
RS
1005 skip_bits_long(gb, 8*cnt - 4); // -4 due to reading extension type
1006 return cnt;
1007}
1008
9cc04edf 1009/**
62a57fae
RS
1010 * Parse whether channels are to be excluded from Dynamic Range Compression; reference: table 4.53.
1011 *
1012 * @return Returns number of bytes consumed.
1013 */
1014static int decode_drc_channel_exclusions(DynamicRangeControl *che_drc, GetBitContext * gb) {
1015 int i;
1016 int num_excl_chan = 0;
1017
1018 do {
1019 for (i = 0; i < 7; i++)
1020 che_drc->exclude_mask[num_excl_chan++] = get_bits1(gb);
1021 } while (num_excl_chan < MAX_CHANNELS - 7 && get_bits1(gb));
1022
1023 return num_excl_chan / 7;
1024}
1025
1026/**
9cc04edf
RS
1027 * Decode dynamic range information; reference: table 4.52.
1028 *
1029 * @param cnt length of TYPE_FIL syntactic element in bytes
1030 *
1031 * @return Returns number of bytes consumed.
1032 */
1033static int decode_dynamic_range(DynamicRangeControl *che_drc, GetBitContext * gb, int cnt) {
1034 int n = 1;
1035 int drc_num_bands = 1;
1036 int i;
1037
1038 /* pce_tag_present? */
1039 if(get_bits1(gb)) {
1040 che_drc->pce_instance_tag = get_bits(gb, 4);
1041 skip_bits(gb, 4); // tag_reserved_bits
1042 n++;
1043 }
1044
1045 /* excluded_chns_present? */
1046 if(get_bits1(gb)) {
1047 n += decode_drc_channel_exclusions(che_drc, gb);
1048 }
1049
1050 /* drc_bands_present? */
1051 if (get_bits1(gb)) {
1052 che_drc->band_incr = get_bits(gb, 4);
1053 che_drc->interpolation_scheme = get_bits(gb, 4);
1054 n++;
1055 drc_num_bands += che_drc->band_incr;
1056 for (i = 0; i < drc_num_bands; i++) {
1057 che_drc->band_top[i] = get_bits(gb, 8);
1058 n++;
1059 }
1060 }
1061
1062 /* prog_ref_level_present? */
1063 if (get_bits1(gb)) {
1064 che_drc->prog_ref_level = get_bits(gb, 7);
1065 skip_bits1(gb); // prog_ref_level_reserved_bits
1066 n++;
1067 }
1068
1069 for (i = 0; i < drc_num_bands; i++) {
1070 che_drc->dyn_rng_sgn[i] = get_bits1(gb);
1071 che_drc->dyn_rng_ctl[i] = get_bits(gb, 7);
1072 n++;
1073 }
1074
1075 return n;
1076}
1077
1078/**
1079 * Decode extension data (incomplete); reference: table 4.51.
1080 *
1081 * @param cnt length of TYPE_FIL syntactic element in bytes
1082 *
1083 * @return Returns number of bytes consumed
1084 */
1085static int decode_extension_payload(AACContext * ac, GetBitContext * gb, int cnt) {
cc0591da
RS
1086 int crc_flag = 0;
1087 int res = cnt;
1088 switch (get_bits(gb, 4)) { // extension type
1089 case EXT_SBR_DATA_CRC:
1090 crc_flag++;
1091 case EXT_SBR_DATA:
1092 res = decode_sbr_extension(ac, gb, crc_flag, cnt);
1093 break;
1094 case EXT_DYNAMIC_RANGE:
1095 res = decode_dynamic_range(&ac->che_drc, gb, cnt);
1096 break;
1097 case EXT_FILL:
1098 case EXT_FILL_DATA:
1099 case EXT_DATA_ELEMENT:
1100 default:
1101 skip_bits_long(gb, 8*cnt - 4);
1102 break;
1103 };
1104 return res;
1105}
1106
1dece0d2
RS
1107 start = ics->swb_offset[FFMIN(bottom, mmm)];
1108 end = ics->swb_offset[FFMIN( top, mmm)];
1109 if ((size = end - start) <= 0)
1110 continue;
1111 if (tns->direction[w][filt]) {
1112 inc = -1; start = end - 1;
1113 } else {
1114 inc = 1;
1115 }
1116 start += w * 128;
1117
1118 // ar filter
1119 for (m = 0; m < size; m++, start += inc)
1120 for (i = 1; i <= FFMIN(m, order); i++)
1121 coef[start] -= coef[start - i*inc] * lpc[i];
1122 }
1123 }
1124}
1125
cc0591da 1126/**
9cc04edf
RS
1127 * Conduct IMDCT and windowing.
1128 */
1129static void imdct_and_windowing(AACContext * ac, SingleChannelElement * sce) {
1130 IndividualChannelStream * ics = &sce->ics;
1131 float * in = sce->coeffs;
1132 float * out = sce->ret;
1133 float * saved = sce->saved;
848a5815
RS
1134 const float * lwindow = ics->use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
1135 const float * swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
1136 const float * lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
1137 const float * swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
9cc04edf
RS
1138 float * buf = ac->buf_mdct;
1139 int i;
1140
62a57fae
RS
1141 if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
1142 if (ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE)
1143 av_log(ac->avccontext, AV_LOG_WARNING,
1144 "Transition from an ONLY_LONG or LONG_STOP to an EIGHT_SHORT sequence detected. "
1145 "If you heard an audible artifact, please submit the sample to the FFmpeg developers.\n");
1146 for (i = 0; i < 2048; i += 256) {
1147 ff_imdct_calc(&ac->mdct_small, buf + i, in + i/2);
1148 ac->dsp.vector_fmul_reverse(ac->revers + i/2, buf + i + 128, swindow, 128);
1149 }
1150 for (i = 0; i < 448; i++) out[i] = saved[i] + ac->add_bias;
1151
1152 ac->dsp.vector_fmul_add_add(out + 448 + 0*128, buf + 0*128, swindow_prev, saved + 448 , ac->add_bias, 128, 1);
1153 ac->dsp.vector_fmul_add_add(out + 448 + 1*128, buf + 2*128, swindow, ac->revers + 0*128, ac->add_bias, 128, 1);
1154 ac->dsp.vector_fmul_add_add(out + 448 + 2*128, buf + 4*128, swindow, ac->revers + 1*128, ac->add_bias, 128, 1);
1155 ac->dsp.vector_fmul_add_add(out + 448 + 3*128, buf + 6*128, swindow, ac->revers + 2*128, ac->add_bias, 128, 1);
1156 ac->dsp.vector_fmul_add_add(out + 448 + 4*128, buf + 8*128, swindow, ac->revers + 3*128, ac->add_bias, 64, 1);
1157
1158#if 0
1159 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);
1160 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);
1161 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);
1162 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);
1163#endif
1164
1165 ac->dsp.vector_fmul_add_add(saved, buf + 1024 + 64, swindow + 64, ac->revers + 3*128+64, 0, 64, 1);
1166 ac->dsp.vector_fmul_add_add(saved + 64, buf + 1024 + 2*128, swindow, ac->revers + 4*128, 0, 128, 1);
1167 ac->dsp.vector_fmul_add_add(saved + 192, buf + 1024 + 4*128, swindow, ac->revers + 5*128, 0, 128, 1);
1168 ac->dsp.vector_fmul_add_add(saved + 320, buf + 1024 + 6*128, swindow, ac->revers + 6*128, 0, 128, 1);
1169 memcpy( saved + 448, ac->revers + 7*128, 128 * sizeof(float));
1170 memset( saved + 576, 0, 448 * sizeof(float));
1171 } else {
1172 ff_imdct_calc(&ac->mdct, buf, in);
1173 if (ics->window_sequence[0] == LONG_STOP_SEQUENCE) {
1174 for (i = 0; i < 448; i++) out[i] = saved[i] + ac->add_bias;
1175 ac->dsp.vector_fmul_add_add(out + 448, buf + 448, swindow_prev, saved + 448, ac->add_bias, 128, 1);
1176 for (i = 576; i < 1024; i++) out[i] = buf[i] + saved[i] + ac->add_bias;
1177 } else {
1178 ac->dsp.vector_fmul_add_add(out, buf, lwindow_prev, saved, ac->add_bias, 1024, 1);
1179 }
1180 if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
1181 memcpy(saved, buf + 1024, 448 * sizeof(float));
1182 ac->dsp.vector_fmul_reverse(saved + 448, buf + 1024 + 448, swindow, 128);
1183 memset(saved + 576, 0, 448 * sizeof(float));
1184 } else {
1185 ac->dsp.vector_fmul_reverse(saved, buf + 1024, lwindow, 1024);
1186 }
1187 }
1188}
1189
9cc04edf 1190/**
cc0591da
RS
1191 * Apply dependent channel coupling (applied before IMDCT).
1192 *
1193 * @param index index into coupling gain array
1194 */
1195static void apply_dependent_coupling(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index) {
1196 IndividualChannelStream * ics = &cc->ch[0].ics;
1197 const uint16_t * offsets = ics->swb_offset;
1198 float * dest = sce->coeffs;
1199 const float * src = cc->ch[0].coeffs;
1200 int g, i, group, k, idx = 0;
1201 if(ac->m4ac.object_type == AOT_AAC_LTP) {
1202 av_log(ac->avccontext, AV_LOG_ERROR,
1203 "Dependent coupling is not supported together with LTP\n");
1204 return;
1205 }
1206 for (g = 0; g < ics->num_window_groups; g++) {
1207 for (i = 0; i < ics->max_sfb; i++, idx++) {
1208 if (cc->ch[0].band_type[idx] != ZERO_BT) {
cc0591da
RS
1209 for (group = 0; group < ics->group_len[g]; group++) {
1210 for (k = offsets[i]; k < offsets[i+1]; k++) {
1211 // XXX dsputil-ize
9edae4ad 1212 dest[group*128+k] += cc->coup.gain[index][idx] * src[group*128+k];
cc0591da
RS
1213 }
1214 }
1215 }
1216 }
1217 dest += ics->group_len[g]*128;
1218 src += ics->group_len[g]*128;
1219 }
1220}
1221
1222/**
1223 * Apply independent channel coupling (applied after IMDCT).
1224 *
1225 * @param index index into coupling gain array
1226 */
1227static void apply_independent_coupling(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index) {
1228 int i;
cc0591da 1229 for (i = 0; i < 1024; i++)
9edae4ad 1230 sce->ret[i] += cc->coup.gain[index][0] * (cc->ch[0].ret[i] - ac->add_bias);
cc0591da
RS
1231}
1232
9ffd5c1c
RS
1233/**
1234 * channel coupling transformation interface
1235 *
1236 * @param index index into coupling gain array
1237 * @param apply_coupling_method pointer to (in)dependent coupling function
1238 */
1239static void apply_channel_coupling(AACContext * ac, ChannelElement * cc,
1240 void (*apply_coupling_method)(AACContext * ac, SingleChannelElement * sce, ChannelElement * cc, int index))
1241{
1242 int c;
1243 int index = 0;
1244 ChannelCoupling * coup = &cc->coup;
1245 for (c = 0; c <= coup->num_coupled; c++) {
1246 if (ac->che[coup->type[c]][coup->id_select[c]]) {
1247 if (coup->ch_select[c] != 2) {
1248 apply_coupling_method(ac, &ac->che[coup->type[c]][coup->id_select[c]]->ch[0], cc, index);
1249 if (coup->ch_select[c] != 0)
1250 index++;
1251 }
1252 if (coup->ch_select[c] != 1)
1253 apply_coupling_method(ac, &ac->che[coup->type[c]][coup->id_select[c]]->ch[1], cc, index++);
1254 } else {
1255 av_log(ac->avccontext, AV_LOG_ERROR,
1256 "coupling target %sE[%d] not available\n",
1257 coup->type[c] == TYPE_CPE ? "CP" : "SC", coup->id_select[c]);
1258 break;
1259 }
1260 }
1261}
1262
1263/**
1264 * Convert spectral data to float samples, applying all supported tools as appropriate.
1265 */
1266static void spectral_to_sample(AACContext * ac) {
1267 int i, type;
1268 for (i = 0; i < MAX_ELEM_ID; i++) {
1269 for(type = 0; type < 4; type++) {
1270 ChannelElement *che = ac->che[type][i];
1271 if(che) {
1272 if(che->coup.coupling_point == BEFORE_TNS)
1273 apply_channel_coupling(ac, che, apply_dependent_coupling);
1274 if(che->ch[0].tns.present)
1275 apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1);
1276 if(che->ch[1].tns.present)
1277 apply_tns(che->ch[1].coeffs, &che->ch[1].tns, &che->ch[1].ics, 1);
1278 if(che->coup.coupling_point == BETWEEN_TNS_AND_IMDCT)
1279 apply_channel_coupling(ac, che, apply_dependent_coupling);
1280 imdct_and_windowing(ac, &che->ch[0]);
1281 if(type == TYPE_CPE)
1282 imdct_and_windowing(ac, &che->ch[1]);
1283 if(che->coup.coupling_point == AFTER_IMDCT)
1284 apply_channel_coupling(ac, che, apply_independent_coupling);
62a57fae
RS
1285 }
1286 }
1287 }
1288}
1289
1290static int aac_decode_frame(AVCodecContext * avccontext, void * data, int * data_size, const uint8_t * buf, int buf_size) {
1291 AACContext * ac = avccontext->priv_data;
1292 GetBitContext gb;
1293 enum RawDataBlockType elem_type;
1294 int err, elem_id, data_size_tmp;
1295
1296 init_get_bits(&gb, buf, buf_size*8);
1297
1298 // parse
1299 while ((elem_type = get_bits(&gb, 3)) != TYPE_END) {
1300 elem_id = get_bits(&gb, 4);
1301 err = -1;
1302
1303 if(elem_type == TYPE_SCE && elem_id == 1 &&
1304 !ac->che[TYPE_SCE][elem_id] && ac->che[TYPE_LFE][0]) {
1305 /* Some streams incorrectly code 5.1 audio as SCE[0] CPE[0] CPE[1] SCE[1]
1306 instead of SCE[0] CPE[0] CPE[0] LFE[0]. If we seem to have
1307 encountered such a stream, transfer the LFE[0] element to SCE[1] */
1308 ac->che[TYPE_SCE][elem_id] = ac->che[TYPE_LFE][0];
1309 ac->che[TYPE_LFE][0] = NULL;
1310 }
1311 if(elem_type && elem_type < TYPE_DSE) {
1312 if(!ac->che[elem_type][elem_id])
1313 return -1;
1314 if(elem_type != TYPE_CCE)
1315 ac->che[elem_type][elem_id]->coup.coupling_point = 4;
1316 }
1317
1318 switch (elem_type) {
1319
1320 case TYPE_SCE:
1321 err = decode_ics(ac, &ac->che[TYPE_SCE][elem_id]->ch[0], &gb, 0, 0);
1322 break;
1323
1324 case TYPE_CPE:
1325 err = decode_cpe(ac, &gb, elem_id);
1326 break;
1327
1328 case TYPE_CCE:
1329 err = decode_cce(ac, &gb, ac->che[TYPE_SCE][elem_id]);
1330 break;
1331
1332 case TYPE_LFE:
1333 err = decode_ics(ac, &ac->che[TYPE_LFE][elem_id]->ch[0], &gb, 0, 0);
1334 break;
1335
1336 case TYPE_DSE:
1337 skip_data_stream_element(&gb);
1338 err = 0;
1339 break;
1340
1341 case TYPE_PCE:
1342 {
1343 enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
1344 memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
1345 if((err = decode_pce(ac, new_che_pos, &gb)))
1346 break;
1347 err = output_configure(ac, ac->che_pos, new_che_pos);
1348 break;
1349 }
1350
1351 case TYPE_FIL:
1352 if (elem_id == 15)
1353 elem_id += get_bits(&gb, 8) - 1;
1354 while (elem_id > 0)
1355 elem_id -= decode_extension_payload(ac, &gb, elem_id);
1356 err = 0; /* FIXME */
1357 break;
1358
1359 default:
1360 err = -1; /* should not happen, but keeps compiler happy */
1361 break;
1362 }
1363
1364 if(err)
1365 return err;
1366 }
1367
1368 spectral_to_sample(ac);
1369
9cc04edf
RS
1370 if (!ac->is_saved) {
1371 ac->is_saved = 1;
1372 *data_size = 0;
848a5815 1373 return buf_size;
9cc04edf
RS
1374 }
1375
1376 data_size_tmp = 1024 * avccontext->channels * sizeof(int16_t);
1377 if(*data_size < data_size_tmp) {
1378 av_log(avccontext, AV_LOG_ERROR,
1379 "Output buffer too small (%d) or trying to output too many samples (%d) for this frame.\n",
1380 *data_size, data_size_tmp);
1381 return -1;
1382 }
1383 *data_size = data_size_tmp;
1384
1385 ac->dsp.float_to_int16_interleave(data, (const float **)ac->output_data, 1024, avccontext->channels);
1386
1387 return buf_size;
1388}
1389
71e9a1b8
RS
1390static av_cold int aac_decode_close(AVCodecContext * avccontext) {
1391 AACContext * ac = avccontext->priv_data;
9edae4ad 1392 int i, type;
71e9a1b8 1393
cc0591da 1394 for (i = 0; i < MAX_ELEM_ID; i++) {
9edae4ad
RS
1395 for(type = 0; type < 4; type++)
1396 av_freep(&ac->che[type][i]);
71e9a1b8
RS
1397 }
1398
1399 ff_mdct_end(&ac->mdct);
1400 ff_mdct_end(&ac->mdct_small);
71e9a1b8
RS
1401 return 0 ;
1402}
1403
1404AVCodec aac_decoder = {
1405 "aac",
1406 CODEC_TYPE_AUDIO,
1407 CODEC_ID_AAC,
1408 sizeof(AACContext),
1409 aac_decode_init,
1410 NULL,
1411 aac_decode_close,
1412 aac_decode_frame,
1413 .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
cc0591da 1414 .sample_fmts = (enum SampleFormat[]){SAMPLE_FMT_S16,SAMPLE_FMT_NONE},
71e9a1b8 1415};