FFmpeg
Loading...
Searching...
No Matches
takdec.c
Go to the documentation of this file.
1/*
2 * TAK decoder
3 * Copyright (c) 2012 Paul B Mahol
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
23 * @file
24 * TAK (Tom's lossless Audio Kompressor) decoder
25 * @author Paul B Mahol
26 */
27
29#include "libavutil/internal.h"
30#include "libavutil/mem.h"
32#include "libavutil/samplefmt.h"
33
34#define CACHED_BITSTREAM_READER !ARCH_X86_32
35#define BITSTREAM_READER_LE
36#include "audiodsp.h"
37#include "thread.h"
38#include "avcodec.h"
39#include "codec_internal.h"
40#include "unary.h"
41#include "tak.h"
42#include "takdsp.h"
43
44#define MAX_SUBFRAMES 8 ///< max number of subframes per channel
45#define MAX_PREDICTORS 256
46
47typedef struct MCDParam {
48 int8_t present; ///< decorrelation parameter availability for this channel
49 int8_t index; ///< index into array of decorrelation types
50 int8_t chan1;
51 int8_t chan2;
52} MCDParam;
53
54typedef struct TAKDecContext {
55 AVCodecContext *avctx; ///< parent AVCodecContext
59 GetBitContext gb; ///< bitstream reader initialized to start at the current frame
60
61 int uval;
62 int nb_samples; ///< number of samples in the current frame
63 uint8_t *decode_buffer;
64 unsigned int decode_buffer_size;
65 int32_t *decoded[TAK_MAX_CHANNELS]; ///< decoded samples for each channel
66
68 int8_t sample_shift[TAK_MAX_CHANNELS]; ///< shift applied to every sample in the channel
70 int nb_subframes; ///< number of subframes in the current frame
71 int16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples
73
74 int8_t dmode; ///< channel decorrelation type in the current frame
75
76 MCDParam mcdparams[TAK_MAX_CHANNELS]; ///< multichannel decorrelation parameters
77
78 int8_t coding_mode[128];
80 DECLARE_ALIGNED(16, int16_t, residues)[544];
82
83static const int8_t mc_dmodes[] = { 1, 3, 4, 6, };
84
85static const uint16_t predictor_sizes[] = {
86 4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
87};
88
89static const struct CParam {
90 int init;
91 int escape;
92 int scale;
94 int bias;
95} xcodes[50] = {
96 { 0x01, 0x0000001, 0x0000001, 0x0000003, 0x0000008 },
97 { 0x02, 0x0000003, 0x0000001, 0x0000007, 0x0000006 },
98 { 0x03, 0x0000005, 0x0000002, 0x000000E, 0x000000D },
99 { 0x03, 0x0000003, 0x0000003, 0x000000D, 0x0000018 },
100 { 0x04, 0x000000B, 0x0000004, 0x000001C, 0x0000019 },
101 { 0x04, 0x0000006, 0x0000006, 0x000001A, 0x0000030 },
102 { 0x05, 0x0000016, 0x0000008, 0x0000038, 0x0000032 },
103 { 0x05, 0x000000C, 0x000000C, 0x0000034, 0x0000060 },
104 { 0x06, 0x000002C, 0x0000010, 0x0000070, 0x0000064 },
105 { 0x06, 0x0000018, 0x0000018, 0x0000068, 0x00000C0 },
106 { 0x07, 0x0000058, 0x0000020, 0x00000E0, 0x00000C8 },
107 { 0x07, 0x0000030, 0x0000030, 0x00000D0, 0x0000180 },
108 { 0x08, 0x00000B0, 0x0000040, 0x00001C0, 0x0000190 },
109 { 0x08, 0x0000060, 0x0000060, 0x00001A0, 0x0000300 },
110 { 0x09, 0x0000160, 0x0000080, 0x0000380, 0x0000320 },
111 { 0x09, 0x00000C0, 0x00000C0, 0x0000340, 0x0000600 },
112 { 0x0A, 0x00002C0, 0x0000100, 0x0000700, 0x0000640 },
113 { 0x0A, 0x0000180, 0x0000180, 0x0000680, 0x0000C00 },
114 { 0x0B, 0x0000580, 0x0000200, 0x0000E00, 0x0000C80 },
115 { 0x0B, 0x0000300, 0x0000300, 0x0000D00, 0x0001800 },
116 { 0x0C, 0x0000B00, 0x0000400, 0x0001C00, 0x0001900 },
117 { 0x0C, 0x0000600, 0x0000600, 0x0001A00, 0x0003000 },
118 { 0x0D, 0x0001600, 0x0000800, 0x0003800, 0x0003200 },
119 { 0x0D, 0x0000C00, 0x0000C00, 0x0003400, 0x0006000 },
120 { 0x0E, 0x0002C00, 0x0001000, 0x0007000, 0x0006400 },
121 { 0x0E, 0x0001800, 0x0001800, 0x0006800, 0x000C000 },
122 { 0x0F, 0x0005800, 0x0002000, 0x000E000, 0x000C800 },
123 { 0x0F, 0x0003000, 0x0003000, 0x000D000, 0x0018000 },
124 { 0x10, 0x000B000, 0x0004000, 0x001C000, 0x0019000 },
125 { 0x10, 0x0006000, 0x0006000, 0x001A000, 0x0030000 },
126 { 0x11, 0x0016000, 0x0008000, 0x0038000, 0x0032000 },
127 { 0x11, 0x000C000, 0x000C000, 0x0034000, 0x0060000 },
128 { 0x12, 0x002C000, 0x0010000, 0x0070000, 0x0064000 },
129 { 0x12, 0x0018000, 0x0018000, 0x0068000, 0x00C0000 },
130 { 0x13, 0x0058000, 0x0020000, 0x00E0000, 0x00C8000 },
131 { 0x13, 0x0030000, 0x0030000, 0x00D0000, 0x0180000 },
132 { 0x14, 0x00B0000, 0x0040000, 0x01C0000, 0x0190000 },
133 { 0x14, 0x0060000, 0x0060000, 0x01A0000, 0x0300000 },
134 { 0x15, 0x0160000, 0x0080000, 0x0380000, 0x0320000 },
135 { 0x15, 0x00C0000, 0x00C0000, 0x0340000, 0x0600000 },
136 { 0x16, 0x02C0000, 0x0100000, 0x0700000, 0x0640000 },
137 { 0x16, 0x0180000, 0x0180000, 0x0680000, 0x0C00000 },
138 { 0x17, 0x0580000, 0x0200000, 0x0E00000, 0x0C80000 },
139 { 0x17, 0x0300000, 0x0300000, 0x0D00000, 0x1800000 },
140 { 0x18, 0x0B00000, 0x0400000, 0x1C00000, 0x1900000 },
141 { 0x18, 0x0600000, 0x0600000, 0x1A00000, 0x3000000 },
142 { 0x19, 0x1600000, 0x0800000, 0x3800000, 0x3200000 },
143 { 0x19, 0x0C00000, 0x0C00000, 0x3400000, 0x6000000 },
144 { 0x1A, 0x2C00000, 0x1000000, 0x7000000, 0x6400000 },
145 { 0x1A, 0x1800000, 0x1800000, 0x6800000, 0xC000000 },
147
149{
150 switch (avctx->bits_per_raw_sample) {
151 case 8:
153 break;
154 case 16:
156 break;
157 case 24:
159 break;
160 default:
161 av_log(avctx, AV_LOG_ERROR, "invalid/unsupported bits per sample: %d\n",
162 avctx->bits_per_raw_sample);
163 return AVERROR_INVALIDDATA;
164 }
165
166 return 0;
167}
168
170{
171 TAKDecContext *s = avctx->priv_data;
172 int shift;
173
174 if (avctx->sample_rate < 11025) {
175 shift = 3;
176 } else if (avctx->sample_rate < 22050) {
177 shift = 2;
178 } else if (avctx->sample_rate < 44100) {
179 shift = 1;
180 } else {
181 shift = 0;
182 }
183 s->uval = FFALIGN(avctx->sample_rate + 511LL >> 9, 4) << shift;
184 s->subframe_scale = FFALIGN(avctx->sample_rate + 511LL >> 9, 4) << 1;
185}
186
188{
189 TAKDecContext *s = avctx->priv_data;
190
191 ff_audiodsp_init(&s->adsp);
192 ff_takdsp_init(&s->tdsp);
193
194 s->avctx = avctx;
196
198
199 return set_bps_params(avctx);
200}
201
202static void decode_lpc(int32_t *coeffs, int mode, int length)
203{
204 int i;
205
206 if (length < 2)
207 return;
208
209 if (mode == 1) {
210 unsigned a1 = *coeffs++;
211 for (i = 0; i < length - 1 >> 1; i++) {
212 *coeffs += a1;
213 coeffs[1] += (unsigned)*coeffs;
214 a1 = coeffs[1];
215 coeffs += 2;
216 }
217 if (length - 1 & 1)
218 *coeffs += a1;
219 } else if (mode == 2) {
220 unsigned a1 = coeffs[1];
221 unsigned a2 = a1 + *coeffs;
222 coeffs[1] = a2;
223 if (length > 2) {
224 coeffs += 2;
225 for (i = 0; i < length - 2 >> 1; i++) {
226 unsigned a3 = *coeffs + a1;
227 unsigned a4 = a3 + a2;
228 *coeffs = a4;
229 a1 = coeffs[1] + a3;
230 a2 = a1 + a4;
231 coeffs[1] = a2;
232 coeffs += 2;
233 }
234 if (length & 1)
235 *coeffs += a1 + a2;
236 }
237 } else if (mode == 3) {
238 unsigned a1 = coeffs[1];
239 unsigned a2 = a1 + *coeffs;
240 coeffs[1] = a2;
241 if (length > 2) {
242 unsigned a3 = coeffs[2];
243 unsigned a4 = a3 + a1;
244 unsigned a5 = a4 + a2;
245 coeffs[2] = a5;
246 coeffs += 3;
247 for (i = 0; i < length - 3; i++) {
248 a3 += *coeffs;
249 a4 += a3;
250 a5 += a4;
251 *coeffs = a5;
252 coeffs++;
253 }
254 }
255 }
256}
257
258static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
259{
260 struct CParam code;
261 GetBitContext *gb = &s->gb;
262 int i;
263
264 if (!mode) {
265 memset(decoded, 0, len * sizeof(*decoded));
266 return 0;
267 }
268
270 return AVERROR_INVALIDDATA;
271 code = xcodes[mode - 1];
272
273 for (i = 0; i < len; i++) {
274 unsigned x = get_bits_long(gb, code.init);
275 if (x >= code.escape && get_bits1(gb)) {
276 x |= 1 << code.init;
277 if (x >= code.aescape) {
278 unsigned scale = get_unary(gb, 1, 9);
279 if (scale == 9) {
280 int scale_bits = get_bits(gb, 3);
281 if (scale_bits > 0) {
282 if (scale_bits == 7) {
283 scale_bits += get_bits(gb, 5);
284 if (scale_bits > 29)
285 return AVERROR_INVALIDDATA;
286 }
287 scale = get_bits_long(gb, scale_bits) + 1;
288 x += code.scale * scale;
289 }
290 x += code.bias;
291 } else
292 x += code.scale * scale - code.escape;
293 } else
294 x -= code.escape;
295 }
296 decoded[i] = (x >> 1) ^ -(x & 1);
297 }
298
299 return 0;
300}
301
302static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
303{
304 GetBitContext *gb = &s->gb;
305 int i, mode, ret;
306
307 if (length > s->nb_samples)
308 return AVERROR_INVALIDDATA;
309
310 if (get_bits1(gb)) {
311 int wlength, rval;
312
313 wlength = length / s->uval;
314
315 rval = length - (wlength * s->uval);
316
317 if (rval < s->uval / 2)
318 rval += s->uval;
319 else
320 wlength++;
321
322 if (wlength <= 1 || wlength > 128)
323 return AVERROR_INVALIDDATA;
324
325 s->coding_mode[0] = mode = get_bits(gb, 6);
326
327 for (i = 1; i < wlength; i++) {
328 int c = get_unary(gb, 1, 6);
329
330 switch (c) {
331 case 6:
332 mode = get_bits(gb, 6);
333 break;
334 case 5:
335 case 4:
336 case 3: {
337 /* mode += sign ? (1 - c) : (c - 1) */
338 int sign = get_bits1(gb);
339 mode += (-sign ^ (c - 1)) + sign;
340 break;
341 }
342 case 2:
343 mode++;
344 break;
345 case 1:
346 mode--;
347 break;
348 }
349 s->coding_mode[i] = mode;
350 }
351
352 i = 0;
353 while (i < wlength) {
354 int len = 0;
355
356 mode = s->coding_mode[i];
357 do {
358 if (i >= wlength - 1)
359 len += rval;
360 else
361 len += s->uval;
362 i++;
363
364 if (i == wlength)
365 break;
366 } while (s->coding_mode[i] == mode);
367
368 if ((ret = decode_segment(s, mode, decoded, len)) < 0)
369 return ret;
370 decoded += len;
371 }
372 } else {
373 mode = get_bits(gb, 6);
374 if ((ret = decode_segment(s, mode, decoded, length)) < 0)
375 return ret;
376 }
377
378 return 0;
379}
380
382{
383 if (get_bits1(gb))
384 return get_bits(gb, 4) + 1;
385 else
386 return 0;
387}
388
390 int subframe_size, int prev_subframe_size)
391{
392 GetBitContext *gb = &s->gb;
393 int x, y, i, j, ret = 0;
394 int dshift, size, filter_quant, filter_order;
395 int tfilter[MAX_PREDICTORS];
396
397 if (!get_bits1(gb))
398 return decode_residues(s, decoded, subframe_size);
399
400 filter_order = predictor_sizes[get_bits(gb, 4)];
401
402 if (prev_subframe_size > 0 && get_bits1(gb)) {
403 if (filter_order > prev_subframe_size)
404 return AVERROR_INVALIDDATA;
405
406 decoded -= filter_order;
407 subframe_size += filter_order;
408
409 if (filter_order > subframe_size)
410 return AVERROR_INVALIDDATA;
411 } else {
412 int lpc_mode;
413
414 if (filter_order > subframe_size)
415 return AVERROR_INVALIDDATA;
416
417 lpc_mode = get_bits(gb, 2);
418 if (lpc_mode > 2)
419 return AVERROR_INVALIDDATA;
420
421 if ((ret = decode_residues(s, decoded, filter_order)) < 0)
422 return ret;
423
424 if (lpc_mode)
425 decode_lpc(decoded, lpc_mode, filter_order);
426 }
427
428 dshift = get_bits_esc4(gb);
429 size = get_bits1(gb) + 6;
430
431 filter_quant = 10;
432 if (get_bits1(gb)) {
433 filter_quant -= get_bits(gb, 3) + 1;
434 if (filter_quant < 3)
435 return AVERROR_INVALIDDATA;
436 }
437
438 if (get_bits_left(gb) < 2*10 + 2*size)
439 return AVERROR_INVALIDDATA;
440
441 s->predictors[0] = get_sbits(gb, 10);
442 s->predictors[1] = get_sbits(gb, 10);
443 s->predictors[2] = get_sbits(gb, size) * (1 << (10 - size));
444 s->predictors[3] = get_sbits(gb, size) * (1 << (10 - size));
445 if (filter_order > 4) {
446 int tmp = size - get_bits1(gb);
447
448 for (i = 4; i < filter_order; i++) {
449 if (!(i & 3))
450 x = tmp - get_bits(gb, 2);
451 s->predictors[i] = get_sbits(gb, x) * (1 << (10 - size));
452 }
453 }
454
455 tfilter[0] = s->predictors[0] * 64;
456 for (i = 1; i < filter_order; i++) {
457 uint32_t *p1 = &tfilter[0];
458 uint32_t *p2 = &tfilter[i - 1];
459
460 for (j = 0; j < (i + 1) / 2; j++) {
461 x = *p1 + ((int32_t)(s->predictors[i] * *p2 + 256) >> 9);
462 *p2 += (int32_t)(s->predictors[i] * *p1 + 256) >> 9;
463 *p1++ = x;
464 p2--;
465 }
466
467 tfilter[i] = s->predictors[i] * 64;
468 }
469
470 x = 1 << (32 - (15 - filter_quant));
471 y = 1 << ((15 - filter_quant) - 1);
472 for (i = 0, j = filter_order - 1; i < filter_order / 2; i++, j--) {
473 s->filter[j] = x - ((tfilter[i] + y) >> (15 - filter_quant));
474 s->filter[i] = x - ((tfilter[j] + y) >> (15 - filter_quant));
475 }
476
477 if ((ret = decode_residues(s, &decoded[filter_order],
478 subframe_size - filter_order)) < 0)
479 return ret;
480
481 for (i = 0; i < filter_order; i++)
482 s->residues[i] = *decoded++ >> dshift;
483
484 y = FF_ARRAY_ELEMS(s->residues) - filter_order;
485 x = subframe_size - filter_order;
486 while (x > 0) {
487 int tmp = FFMIN(y, x);
488
489 for (i = 0; i < tmp; i++) {
490 int v = 1 << (filter_quant - 1);
491
492 if (filter_order & -16)
493 v += (unsigned)s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
494 filter_order & -16);
495 for (j = filter_order & -16; j < filter_order; j += 4) {
496 v += s->residues[i + j + 3] * (unsigned)s->filter[j + 3] +
497 s->residues[i + j + 2] * (unsigned)s->filter[j + 2] +
498 s->residues[i + j + 1] * (unsigned)s->filter[j + 1] +
499 s->residues[i + j ] * (unsigned)s->filter[j ];
500 }
501 v = (av_clip_intp2(v >> filter_quant, 13) * (1 << dshift)) - (unsigned)*decoded;
502 *decoded++ = v;
503 s->residues[filter_order + i] = v >> dshift;
504 }
505
506 x -= tmp;
507 if (x > 0)
508 memcpy(s->residues, &s->residues[y], 2 * filter_order);
509 }
510
511 return 0;
512}
513
514static int decode_channel(TAKDecContext *s, int chan)
515{
516 AVCodecContext *avctx = s->avctx;
517 GetBitContext *gb = &s->gb;
518 int32_t *decoded = s->decoded[chan];
519 int left = s->nb_samples - 1;
520 int i = 0, ret, prev = 0;
521
522 s->sample_shift[chan] = get_bits_esc4(gb);
523 if (s->sample_shift[chan] >= avctx->bits_per_raw_sample)
524 return AVERROR_INVALIDDATA;
525
526 *decoded++ = get_sbits(gb, avctx->bits_per_raw_sample - s->sample_shift[chan]);
527 s->lpc_mode[chan] = get_bits(gb, 2);
528 s->nb_subframes = get_bits(gb, 3) + 1;
529
530 if (s->nb_subframes > 1) {
531 if (get_bits_left(gb) < (s->nb_subframes - 1) * 6)
532 return AVERROR_INVALIDDATA;
533
534 for (; i < s->nb_subframes - 1; i++) {
535 int v = get_bits(gb, 6);
536
537 s->subframe_len[i] = (v - prev) * s->subframe_scale;
538 if (s->subframe_len[i] <= 0)
539 return AVERROR_INVALIDDATA;
540
541 left -= s->subframe_len[i];
542 prev = v;
543 }
544
545 if (left <= 0)
546 return AVERROR_INVALIDDATA;
547 }
548 s->subframe_len[i] = left;
549
550 prev = 0;
551 for (i = 0; i < s->nb_subframes; i++) {
552 if ((ret = decode_subframe(s, decoded, s->subframe_len[i], prev)) < 0)
553 return ret;
554 decoded += s->subframe_len[i];
555 prev = s->subframe_len[i];
556 }
557
558 return 0;
559}
560
561static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
562{
563 GetBitContext *gb = &s->gb;
564 int32_t *p1 = s->decoded[c1] + (s->dmode > 5);
565 int32_t *p2 = s->decoded[c2] + (s->dmode > 5);
566 int32_t bp1 = p1[0];
567 int32_t bp2 = p2[0];
568 int i;
569 int dshift, dfactor;
570
571 length += s->dmode < 6;
572
573 switch (s->dmode) {
574 case 1: /* left/side */
575 s->tdsp.decorrelate_ls(p1, p2, length);
576 break;
577 case 2: /* side/right */
578 s->tdsp.decorrelate_sr(p1, p2, length);
579 break;
580 case 3: /* side/mid */
581 s->tdsp.decorrelate_sm(p1, p2, length);
582 break;
583 case 4: /* side/left with scale factor */
584 FFSWAP(int32_t*, p1, p2);
585 FFSWAP(int32_t, bp1, bp2);
587 case 5: /* side/right with scale factor */
588 dshift = get_bits_esc4(gb);
589 dfactor = get_sbits(gb, 10);
590 s->tdsp.decorrelate_sf(p1, p2, length, dshift, dfactor);
591 break;
592 case 6:
593 FFSWAP(int32_t*, p1, p2);
595 case 7: {
596 int length2, order_half, filter_order, dval1, dval2;
597 int tmp, x, code_size;
598
599 if (length < 256)
600 return AVERROR_INVALIDDATA;
601
602 dshift = get_bits_esc4(gb);
603 filter_order = 8 << get_bits1(gb);
604 dval1 = get_bits1(gb);
605 dval2 = get_bits1(gb);
606
607 for (i = 0; i < filter_order; i++) {
608 if (!(i & 3))
609 code_size = 14 - get_bits(gb, 3);
610 s->filter[i] = get_sbits(gb, code_size);
611 }
612
613 order_half = filter_order / 2;
614 length2 = length - (filter_order - 1);
615
616 /* decorrelate beginning samples */
617 if (dval1) {
618 for (i = 0; i < order_half; i++) {
619 int32_t a = p1[i];
620 int32_t b = p2[i];
621 p1[i] = a + b;
622 }
623 }
624
625 /* decorrelate ending samples */
626 if (dval2) {
627 for (i = length2 + order_half; i < length; i++) {
628 int32_t a = p1[i];
629 int32_t b = p2[i];
630 p1[i] = a + b;
631 }
632 }
633
634
635 for (i = 0; i < filter_order; i++)
636 s->residues[i] = *p2++ >> dshift;
637
638 p1 += order_half;
639 x = FF_ARRAY_ELEMS(s->residues) - filter_order;
640 for (; length2 > 0; length2 -= tmp) {
641 tmp = FFMIN(length2, x);
642
643 for (i = 0; i < tmp - (tmp == length2); i++)
644 s->residues[filter_order + i] = *p2++ >> dshift;
645
646 for (i = 0; i < tmp; i++) {
647 int v = 1 << 9;
648
649 if (filter_order == 16) {
650 v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
651 filter_order);
652 } else {
653 v += s->residues[i + 7] * s->filter[7] +
654 s->residues[i + 6] * s->filter[6] +
655 s->residues[i + 5] * s->filter[5] +
656 s->residues[i + 4] * s->filter[4] +
657 s->residues[i + 3] * s->filter[3] +
658 s->residues[i + 2] * s->filter[2] +
659 s->residues[i + 1] * s->filter[1] +
660 s->residues[i ] * s->filter[0];
661 }
662
663 v = av_clip_intp2(v >> 10, 13) * (1U << dshift) - *p1;
664 *p1++ = v;
665 }
666
667 memmove(s->residues, &s->residues[tmp], 2 * filter_order);
668 }
669 break;
670 }
671 }
672
673 if (s->dmode > 0 && s->dmode < 6) {
674 p1[0] = bp1;
675 p2[0] = bp2;
676 }
677
678 return 0;
679}
680
682 int *got_frame_ptr, AVPacket *pkt)
683{
684 TAKDecContext *s = avctx->priv_data;
685 GetBitContext *gb = &s->gb;
686 int chan, i, ret, hsize;
687
689 return AVERROR_INVALIDDATA;
690
691 if ((ret = init_get_bits8(gb, pkt->data, pkt->size)) < 0)
692 return ret;
693
694 if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
695 return ret;
696
697 hsize = get_bits_count(gb) / 8;
699 if (ff_tak_check_crc(pkt->data, hsize)) {
700 av_log(avctx, AV_LOG_ERROR, "CRC error\n");
701 if (avctx->err_recognition & AV_EF_EXPLODE)
702 return AVERROR_INVALIDDATA;
703 }
704 }
705
706 if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
707 s->ti.codec != TAK_CODEC_MULTICHANNEL) {
708 avpriv_report_missing_feature(avctx, "TAK codec type %d", s->ti.codec);
710 }
711 if (s->ti.data_type) {
712 av_log(avctx, AV_LOG_ERROR,
713 "unsupported data type: %d\n", s->ti.data_type);
714 return AVERROR_INVALIDDATA;
715 }
716 if (s->ti.codec == TAK_CODEC_MONO_STEREO && s->ti.channels > 2) {
717 av_log(avctx, AV_LOG_ERROR,
718 "invalid number of channels: %d\n", s->ti.channels);
719 return AVERROR_INVALIDDATA;
720 }
721 if (s->ti.channels > 6) {
722 av_log(avctx, AV_LOG_ERROR,
723 "unsupported number of channels: %d\n", s->ti.channels);
724 return AVERROR_INVALIDDATA;
725 }
726
727 if (s->ti.frame_samples <= 0) {
728 av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
729 return AVERROR_INVALIDDATA;
730 }
731
732 avctx->bits_per_raw_sample = s->ti.bps;
733 if ((ret = set_bps_params(avctx)) < 0)
734 return ret;
735 if (s->ti.sample_rate != avctx->sample_rate) {
736 avctx->sample_rate = s->ti.sample_rate;
738 }
739
741 if (s->ti.ch_layout) {
742 av_channel_layout_from_mask(&avctx->ch_layout, s->ti.ch_layout);
743 } else {
745 avctx->ch_layout.nb_channels = s->ti.channels;
746 }
747
748 s->nb_samples = s->ti.last_frame_samples ? s->ti.last_frame_samples
749 : s->ti.frame_samples;
750
751 frame->nb_samples = s->nb_samples;
752 if ((ret = ff_thread_get_buffer(avctx, frame, 0)) < 0)
753 return ret;
755
756 if (avctx->bits_per_raw_sample <= 16) {
758 s->nb_samples,
760 if (buf_size < 0)
761 return buf_size;
762 av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
763 if (!s->decode_buffer)
764 return AVERROR(ENOMEM);
765 ret = av_samples_fill_arrays((uint8_t **)s->decoded, NULL,
766 s->decode_buffer, avctx->ch_layout.nb_channels,
767 s->nb_samples, AV_SAMPLE_FMT_S32P, 0);
768 if (ret < 0)
769 return ret;
770 } else {
771 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++)
772 s->decoded[chan] = (int32_t *)frame->extended_data[chan];
773 }
774
775 if (s->nb_samples < 16) {
776 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
777 int32_t *decoded = s->decoded[chan];
778 for (i = 0; i < s->nb_samples; i++)
779 decoded[i] = get_sbits(gb, avctx->bits_per_raw_sample);
780 }
781 } else {
782 if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
783 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++)
784 if (ret = decode_channel(s, chan))
785 return ret;
786
787 if (avctx->ch_layout.nb_channels == 2) {
788 s->nb_subframes = get_bits(gb, 1) + 1;
789 if (s->nb_subframes > 1) {
790 s->subframe_len[1] = get_bits(gb, 6);
791 }
792
793 s->dmode = get_bits(gb, 3);
794 if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
795 return ret;
796 }
797 } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
798 if (get_bits1(gb)) {
799 int ch_mask = 0;
800
801 chan = get_bits(gb, 4) + 1;
802 if (chan > avctx->ch_layout.nb_channels)
803 return AVERROR_INVALIDDATA;
804
805 for (i = 0; i < chan; i++) {
806 int nbit = get_bits(gb, 4);
807
808 if (nbit >= avctx->ch_layout.nb_channels)
809 return AVERROR_INVALIDDATA;
810
811 if (ch_mask & 1 << nbit)
812 return AVERROR_INVALIDDATA;
813
814 s->mcdparams[i].present = get_bits1(gb);
815 if (s->mcdparams[i].present) {
816 s->mcdparams[i].index = get_bits(gb, 2);
817 s->mcdparams[i].chan2 = get_bits(gb, 4);
818 if (s->mcdparams[i].chan2 >= avctx->ch_layout.nb_channels) {
819 av_log(avctx, AV_LOG_ERROR,
820 "invalid channel 2 (%d) for %d channel(s)\n",
821 s->mcdparams[i].chan2, avctx->ch_layout.nb_channels);
822 return AVERROR_INVALIDDATA;
823 }
824 if (s->mcdparams[i].index == 1) {
825 if ((nbit == s->mcdparams[i].chan2) ||
826 (ch_mask & 1 << s->mcdparams[i].chan2))
827 return AVERROR_INVALIDDATA;
828
829 ch_mask |= 1 << s->mcdparams[i].chan2;
830 } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
831 return AVERROR_INVALIDDATA;
832 }
833 }
834 s->mcdparams[i].chan1 = nbit;
835
836 ch_mask |= 1 << nbit;
837 }
838 } else {
839 chan = avctx->ch_layout.nb_channels;
840 for (i = 0; i < chan; i++) {
841 s->mcdparams[i].present = 0;
842 s->mcdparams[i].chan1 = i;
843 }
844 }
845
846 for (i = 0; i < chan; i++) {
847 if (s->mcdparams[i].present && s->mcdparams[i].index == 1)
848 if (ret = decode_channel(s, s->mcdparams[i].chan2))
849 return ret;
850
851 if (ret = decode_channel(s, s->mcdparams[i].chan1))
852 return ret;
853
854 if (s->mcdparams[i].present) {
855 s->dmode = mc_dmodes[s->mcdparams[i].index];
856 if (ret = decorrelate(s,
857 s->mcdparams[i].chan2,
858 s->mcdparams[i].chan1,
859 s->nb_samples - 1))
860 return ret;
861 }
862 }
863 }
864
865 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
866 int32_t *decoded = s->decoded[chan];
867
868 if (s->lpc_mode[chan])
869 decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
870
871 if (s->sample_shift[chan] > 0)
872 for (i = 0; i < s->nb_samples; i++)
873 decoded[i] *= 1U << s->sample_shift[chan];
874 }
875 }
876
877 align_get_bits(gb);
878 skip_bits(gb, 24);
879 if (get_bits_left(gb) < 0)
880 av_log(avctx, AV_LOG_DEBUG, "overread\n");
881 else if (get_bits_left(gb) > 0)
882 av_log(avctx, AV_LOG_DEBUG, "underread\n");
883
885 if (ff_tak_check_crc(pkt->data + hsize,
886 get_bits_count(gb) / 8 - hsize)) {
887 av_log(avctx, AV_LOG_ERROR, "CRC error\n");
888 if (avctx->err_recognition & AV_EF_EXPLODE)
889 return AVERROR_INVALIDDATA;
890 }
891 }
892
893 /* convert to output buffer */
894 switch (avctx->sample_fmt) {
896 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
897 uint8_t *samples = (uint8_t *)frame->extended_data[chan];
898 int32_t *decoded = s->decoded[chan];
899 for (i = 0; i < s->nb_samples; i++)
900 samples[i] = decoded[i] + 0x80U;
901 }
902 break;
904 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
905 int16_t *samples = (int16_t *)frame->extended_data[chan];
906 int32_t *decoded = s->decoded[chan];
907 for (i = 0; i < s->nb_samples; i++)
908 samples[i] = decoded[i];
909 }
910 break;
912 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
913 int32_t *samples = (int32_t *)frame->extended_data[chan];
914 for (i = 0; i < s->nb_samples; i++)
915 samples[i] *= 1U << 8;
916 }
917 break;
918 }
919
920 *got_frame_ptr = 1;
921
922 return pkt->size;
923}
924
925#if HAVE_THREADS
926static int update_thread_context(AVCodecContext *dst,
927 const AVCodecContext *src)
928{
929 TAKDecContext *tsrc = src->priv_data;
930 TAKDecContext *tdst = dst->priv_data;
931
932 if (dst == src)
933 return 0;
934 memcpy(&tdst->ti, &tsrc->ti, sizeof(TAKStreamInfo));
935 return 0;
936}
937#endif
938
940{
941 TAKDecContext *s = avctx->priv_data;
942
943 av_freep(&s->decode_buffer);
944
945 return 0;
946}
947
949 .p.name = "tak",
950 CODEC_LONG_NAME("TAK (Tom's lossless Audio Kompressor)"),
951 .p.type = AVMEDIA_TYPE_AUDIO,
952 .p.id = AV_CODEC_ID_TAK,
953 .priv_data_size = sizeof(TAKDecContext),
957 UPDATE_THREAD_CONTEXT(update_thread_context),
959};
#define MAX_PREDICTORS
Definition aac.h:89
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition dsp.h:87
const FFCodec ff_tak_decoder
Definition takdec.c:948
static av_cold void close(AVCodecParserContext *s)
Definition apv_parser.c:197
int32_t
Libavcodec external API header.
static int BS_FUNC left(const BSCTX *bc)
Return the number of the bits left in a buffer.
#define i(width, name, range_min, range_max)
Definition cbs_h264.c:63
#define s(width, name)
Definition cbs_vp9.c:198
#define UPDATE_THREAD_CONTEXT(func)
#define FF_CODEC_DECODE_CB(func)
#define CODEC_LONG_NAME(str)
#define av_clip_intp2
Definition common.h:121
#define NULL
Definition coverity.c:32
#define AV_EF_CRCCHECK
Verify checksums embedded in the bitstream (could be of either encoded or decoded data,...
Definition defs.h:48
#define AV_EF_COMPLIANT
consider all spec non compliances as errors
Definition defs.h:55
#define AV_EF_EXPLODE
abort decoding on minor error detection
Definition defs.h:51
static AVPacket * pkt
static AVFrame * frame
int(* init)(AVBSFContext *ctx)
Definition dts2pts.c:608
static unsigned int get_bits_long(GetBitContext *s, int n)
Read 0-32 bits.
Definition get_bits.h:424
static int get_sbits(GetBitContext *s, int n)
Definition get_bits.h:322
static int get_bits_left(GetBitContext *gb)
Definition get_bits.h:688
static unsigned int get_bits1(GetBitContext *s)
Definition get_bits.h:391
static void skip_bits(GetBitContext *s, int n)
Definition get_bits.h:383
static int init_get_bits8(GetBitContext *s, const uint8_t *buffer, int byte_size)
Initialize GetBitContext.
Definition get_bits.h:544
static const uint8_t * align_get_bits(GetBitContext *s)
Definition get_bits.h:560
static int get_bits_count(const GetBitContext *s)
Definition get_bits.h:254
static unsigned int get_bits(GetBitContext *s, int n)
Read 1-25 bits.
Definition get_bits.h:337
#define AV_CODEC_CAP_DR1
Codec uses get_buffer() or get_encode_buffer() for allocating buffers and supports custom allocators.
Definition codec.h:49
#define AV_CODEC_CAP_CHANNEL_CONF
Codec should fill in channel configuration and samplerate instead of container.
Definition codec.h:94
#define AV_CODEC_CAP_FRAME_THREADS
Codec supports frame-level multithreading.
Definition codec.h:98
@ AV_CODEC_ID_TAK
Definition codec_id.h:515
void av_channel_layout_uninit(AVChannelLayout *channel_layout)
Free any allocated data in the channel layout and reset the channel count to 0.
int av_channel_layout_from_mask(AVChannelLayout *channel_layout, uint64_t mask)
Initialize a native channel layout from a bitmask indicating which channels are present.
@ AV_CHANNEL_ORDER_UNSPEC
Only the channel count is specified, without any further information about the channel order.
#define AVERROR_PATCHWELCOME
Not yet implemented in FFmpeg, patches welcome.
Definition error.h:64
#define AVERROR_INVALIDDATA
Invalid data found when processing input.
Definition error.h:61
#define AVERROR(e)
Definition error.h:45
#define AV_LOG_DEBUG
Stuff which is only useful for libav* developers.
Definition log.h:231
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition log.h:210
void av_fast_malloc(void *ptr, unsigned int *size, size_t min_size)
Allocate a buffer, reusing the given one if large enough.
Definition mem.c:555
@ AVMEDIA_TYPE_AUDIO
Definition avutil.h:201
int av_samples_get_buffer_size(int *linesize, int nb_channels, int nb_samples, enum AVSampleFormat sample_fmt, int align)
Get the required buffer size for the given audio parameters.
Definition samplefmt.c:121
@ AV_SAMPLE_FMT_S16P
signed 16 bits, planar
Definition samplefmt.h:64
@ AV_SAMPLE_FMT_U8P
unsigned 8 bits, planar
Definition samplefmt.h:63
@ AV_SAMPLE_FMT_S32P
signed 32 bits, planar
Definition samplefmt.h:65
int av_samples_fill_arrays(uint8_t **audio_data, int *linesize, const uint8_t *buf, int nb_channels, int nb_samples, enum AVSampleFormat sample_fmt, int align)
Fill plane data pointers and linesize for samples with sample format sample_fmt.
Definition samplefmt.c:153
int a
#define b
Definition input.c:43
static void scale(int *out, const int *in, const int w, const int h, const int shift)
Definition intra.c:278
av_cold void ff_audiodsp_init(AudioDSPContext *c)
Definition audiodsp.c:65
static int shift(int a, int b)
Definition bonk.c:261
static int get_bits_esc4(GetBitContext *gb)
Definition takdec.c:381
static const uint16_t predictor_sizes[]
Definition takdec.c:85
static int decode_channel(TAKDecContext *s, int chan)
Definition takdec.c:514
static const int8_t mc_dmodes[]
Definition takdec.c:83
static int decode_subframe(TAKDecContext *s, int32_t *decoded, int subframe_size, int prev_subframe_size)
Definition takdec.c:389
static av_cold int tak_decode_close(AVCodecContext *avctx)
Definition takdec.c:939
static const struct CParam xcodes[50]
static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
Definition takdec.c:561
static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
Definition takdec.c:302
static void decode_lpc(int32_t *coeffs, int mode, int length)
Definition takdec.c:202
static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
Definition takdec.c:258
#define MAX_SUBFRAMES
max number of subframes per channel
Definition takdec.c:44
static av_cold int tak_decode_init(AVCodecContext *avctx)
Definition takdec.c:187
static int set_bps_params(AVCodecContext *avctx)
Definition takdec.c:148
static int tak_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_frame_ptr, AVPacket *pkt)
Definition takdec.c:681
static void set_sample_rate_params(AVCodecContext *avctx)
Definition takdec.c:169
#define MAX_PREDICTORS
Definition takdec.c:45
av_cold void ff_takdsp_init(TAKDSPContext *c)
Definition takdsp.c:73
Multithreading API for decoders.
Macro definitions for various function/variable attributes.
#define av_fallthrough
Definition attributes.h:67
#define av_cold
Definition attributes.h:117
common internal API header
void avpriv_report_missing_feature(void *avc, const char *msg,...) av_printf_format(2
Log a generic warning message about a missing feature.
#define FFSWAP(type, a, b)
Definition macros.h:52
#define FFMIN(a, b)
Definition macros.h:49
#define FFALIGN(x, a)
Definition macros.h:78
Memory handling functions.
#define DECLARE_ALIGNED(n, t, v)
Declare a variable that is aligned in memory.
static const uint64_t c2
Definition murmur3.c:53
static const uint64_t c1
Definition murmur3.c:52
int ff_thread_get_buffer(AVCodecContext *avctx, AVFrame *f, int flags)
Wrapper around get_buffer() for frame-multithreaded codecs.
void ff_thread_finish_setup(AVCodecContext *avctx)
If the codec defines update_thread_context(), call this when they are ready for the next thread to st...
#define FF_ARRAY_ELEMS(a)
const uint8_t * code
Definition spdifenc.c:433
enum AVChannelOrder order
Channel order used in this layout.
int nb_channels
Number of channels in this layout.
main external API structure.
Definition avcodec.h:443
AVChannelLayout ch_layout
Audio channel layout.
Definition avcodec.h:1055
enum AVSampleFormat sample_fmt
audio sample format
Definition avcodec.h:1047
int bits_per_coded_sample
bits per sample/pixel from the demuxer (needed for huffyuv).
Definition avcodec.h:1564
int bits_per_raw_sample
Bits per sample/pixel of internal libavcodec pixel/sample format.
Definition avcodec.h:1571
int sample_rate
samples per second
Definition avcodec.h:1040
void * priv_data
Definition avcodec.h:470
int err_recognition
Error recognition; may misdetect some more or less valid parts as errors.
Definition avcodec.h:1416
This structure describes decoded (raw) audio or video data.
Definition frame.h:472
This structure stores compressed data.
Definition packet.h:580
int scale
Definition takdec.c:92
int bias
Definition takdec.c:94
int init
Definition takdec.c:90
int aescape
Definition takdec.c:93
int escape
Definition takdec.c:91
int8_t index
index into array of decorrelation types
Definition takdec.c:49
int8_t chan2
Definition takdec.c:51
int8_t chan1
Definition takdec.c:50
int8_t present
decorrelation parameter availability for this channel
Definition takdec.c:48
int8_t dmode
channel decorrelation type in the current frame
Definition takdec.c:74
int16_t filter[MAX_PREDICTORS]
Definition takdec.c:79
int16_t subframe_len[MAX_SUBFRAMES]
subframe length in samples
Definition takdec.c:71
int8_t coding_mode[128]
Definition takdec.c:78
uint8_t * decode_buffer
Definition takdec.c:63
int subframe_scale
Definition takdec.c:72
int16_t predictors[MAX_PREDICTORS]
Definition takdec.c:69
int8_t lpc_mode[TAK_MAX_CHANNELS]
Definition takdec.c:67
GetBitContext gb
bitstream reader initialized to start at the current frame
Definition takdec.c:59
TAKDSPContext tdsp
Definition takdec.c:57
int nb_subframes
number of subframes in the current frame
Definition takdec.c:70
MCDParam mcdparams[TAK_MAX_CHANNELS]
multichannel decorrelation parameters
Definition takdec.c:76
int32_t * decoded[TAK_MAX_CHANNELS]
decoded samples for each channel
Definition takdec.c:65
unsigned int decode_buffer_size
Definition takdec.c:64
int nb_samples
number of samples in the current frame
Definition takdec.c:62
AVCodecContext * avctx
parent AVCodecContext
Definition takdec.c:55
TAKStreamInfo ti
Definition takdec.c:58
int8_t sample_shift[TAK_MAX_CHANNELS]
shift applied to every sample in the channel
Definition takdec.c:68
AudioDSPContext adsp
Definition takdec.c:56
int16_t residues[544]
Definition takdec.c:80
Definition swscale.c:71
#define av_freep(p)
#define av_log(a,...)
int ff_tak_decode_frame_header(void *logctx, GetBitContext *gb, TAKStreamInfo *ti, int log_level_offset)
Validate and decode a frame header.
Definition tak.c:147
int ff_tak_check_crc(const uint8_t *buf, unsigned int buf_size)
Definition tak.c:79
TAK (Tom's lossless Audio Kompressor) decoder/demuxer common functions.
#define TAK_MAX_CHANNELS
Definition tak.h:62
#define TAK_MIN_FRAME_HEADER_BYTES
Definition tak.h:95
@ TAK_CODEC_MONO_STEREO
Definition tak.h:98
@ TAK_CODEC_MULTICHANNEL
Definition tak.h:99
static uint8_t tmp[40]
Definition aes_ctr.c:52
#define src
Definition vp8dsp.c:248
int size
static int get_unary(GetBitContext *gb, int stop, int len)
Get unary code of limited length.
Definition unary.h:46
static double a3(void *priv, double x, double y)
Definition vf_xfade.c:2031
static double a2(void *priv, double x, double y)
Definition vf_xfade.c:2030
static double a1(void *priv, double x, double y)
Definition vf_xfade.c:2029
int len
static double c[64]