summaryrefslogtreecommitdiff
path: root/libavcodec/takdec.c (plain)
blob: 5dfcca82ab77681ea5cdca2054c46e7ad18184b8
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
28#include "libavutil/internal.h"
29#include "libavutil/samplefmt.h"
30
31#define BITSTREAM_READER_LE
32#include "audiodsp.h"
33#include "thread.h"
34#include "avcodec.h"
35#include "internal.h"
36#include "unary.h"
37#include "tak.h"
38#include "takdsp.h"
39
40#define MAX_SUBFRAMES 8 ///< max number of subframes per channel
41#define MAX_PREDICTORS 256
42
43typedef struct MCDParam {
44 int8_t present; ///< decorrelation parameter availability for this channel
45 int8_t index; ///< index into array of decorrelation types
46 int8_t chan1;
47 int8_t chan2;
48} MCDParam;
49
50typedef struct TAKDecContext {
51 AVCodecContext *avctx; ///< parent AVCodecContext
52 AudioDSPContext adsp;
53 TAKDSPContext tdsp;
54 TAKStreamInfo ti;
55 GetBitContext gb; ///< bitstream reader initialized to start at the current frame
56
57 int uval;
58 int nb_samples; ///< number of samples in the current frame
59 uint8_t *decode_buffer;
60 unsigned int decode_buffer_size;
61 int32_t *decoded[TAK_MAX_CHANNELS]; ///< decoded samples for each channel
62
63 int8_t lpc_mode[TAK_MAX_CHANNELS];
64 int8_t sample_shift[TAK_MAX_CHANNELS]; ///< shift applied to every sample in the channel
65 int16_t predictors[MAX_PREDICTORS];
66 int nb_subframes; ///< number of subframes in the current frame
67 int16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples
68 int subframe_scale;
69
70 int8_t dmode; ///< channel decorrelation type in the current frame
71
72 MCDParam mcdparams[TAK_MAX_CHANNELS]; ///< multichannel decorrelation parameters
73
74 int8_t coding_mode[128];
75 DECLARE_ALIGNED(16, int16_t, filter)[MAX_PREDICTORS];
76 DECLARE_ALIGNED(16, int16_t, residues)[544];
77} TAKDecContext;
78
79static const int8_t mc_dmodes[] = { 1, 3, 4, 6, };
80
81static const uint16_t predictor_sizes[] = {
82 4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
83};
84
85static const struct CParam {
86 int init;
87 int escape;
88 int scale;
89 int aescape;
90 int bias;
91} xcodes[50] = {
92 { 0x01, 0x0000001, 0x0000001, 0x0000003, 0x0000008 },
93 { 0x02, 0x0000003, 0x0000001, 0x0000007, 0x0000006 },
94 { 0x03, 0x0000005, 0x0000002, 0x000000E, 0x000000D },
95 { 0x03, 0x0000003, 0x0000003, 0x000000D, 0x0000018 },
96 { 0x04, 0x000000B, 0x0000004, 0x000001C, 0x0000019 },
97 { 0x04, 0x0000006, 0x0000006, 0x000001A, 0x0000030 },
98 { 0x05, 0x0000016, 0x0000008, 0x0000038, 0x0000032 },
99 { 0x05, 0x000000C, 0x000000C, 0x0000034, 0x0000060 },
100 { 0x06, 0x000002C, 0x0000010, 0x0000070, 0x0000064 },
101 { 0x06, 0x0000018, 0x0000018, 0x0000068, 0x00000C0 },
102 { 0x07, 0x0000058, 0x0000020, 0x00000E0, 0x00000C8 },
103 { 0x07, 0x0000030, 0x0000030, 0x00000D0, 0x0000180 },
104 { 0x08, 0x00000B0, 0x0000040, 0x00001C0, 0x0000190 },
105 { 0x08, 0x0000060, 0x0000060, 0x00001A0, 0x0000300 },
106 { 0x09, 0x0000160, 0x0000080, 0x0000380, 0x0000320 },
107 { 0x09, 0x00000C0, 0x00000C0, 0x0000340, 0x0000600 },
108 { 0x0A, 0x00002C0, 0x0000100, 0x0000700, 0x0000640 },
109 { 0x0A, 0x0000180, 0x0000180, 0x0000680, 0x0000C00 },
110 { 0x0B, 0x0000580, 0x0000200, 0x0000E00, 0x0000C80 },
111 { 0x0B, 0x0000300, 0x0000300, 0x0000D00, 0x0001800 },
112 { 0x0C, 0x0000B00, 0x0000400, 0x0001C00, 0x0001900 },
113 { 0x0C, 0x0000600, 0x0000600, 0x0001A00, 0x0003000 },
114 { 0x0D, 0x0001600, 0x0000800, 0x0003800, 0x0003200 },
115 { 0x0D, 0x0000C00, 0x0000C00, 0x0003400, 0x0006000 },
116 { 0x0E, 0x0002C00, 0x0001000, 0x0007000, 0x0006400 },
117 { 0x0E, 0x0001800, 0x0001800, 0x0006800, 0x000C000 },
118 { 0x0F, 0x0005800, 0x0002000, 0x000E000, 0x000C800 },
119 { 0x0F, 0x0003000, 0x0003000, 0x000D000, 0x0018000 },
120 { 0x10, 0x000B000, 0x0004000, 0x001C000, 0x0019000 },
121 { 0x10, 0x0006000, 0x0006000, 0x001A000, 0x0030000 },
122 { 0x11, 0x0016000, 0x0008000, 0x0038000, 0x0032000 },
123 { 0x11, 0x000C000, 0x000C000, 0x0034000, 0x0060000 },
124 { 0x12, 0x002C000, 0x0010000, 0x0070000, 0x0064000 },
125 { 0x12, 0x0018000, 0x0018000, 0x0068000, 0x00C0000 },
126 { 0x13, 0x0058000, 0x0020000, 0x00E0000, 0x00C8000 },
127 { 0x13, 0x0030000, 0x0030000, 0x00D0000, 0x0180000 },
128 { 0x14, 0x00B0000, 0x0040000, 0x01C0000, 0x0190000 },
129 { 0x14, 0x0060000, 0x0060000, 0x01A0000, 0x0300000 },
130 { 0x15, 0x0160000, 0x0080000, 0x0380000, 0x0320000 },
131 { 0x15, 0x00C0000, 0x00C0000, 0x0340000, 0x0600000 },
132 { 0x16, 0x02C0000, 0x0100000, 0x0700000, 0x0640000 },
133 { 0x16, 0x0180000, 0x0180000, 0x0680000, 0x0C00000 },
134 { 0x17, 0x0580000, 0x0200000, 0x0E00000, 0x0C80000 },
135 { 0x17, 0x0300000, 0x0300000, 0x0D00000, 0x1800000 },
136 { 0x18, 0x0B00000, 0x0400000, 0x1C00000, 0x1900000 },
137 { 0x18, 0x0600000, 0x0600000, 0x1A00000, 0x3000000 },
138 { 0x19, 0x1600000, 0x0800000, 0x3800000, 0x3200000 },
139 { 0x19, 0x0C00000, 0x0C00000, 0x3400000, 0x6000000 },
140 { 0x1A, 0x2C00000, 0x1000000, 0x7000000, 0x6400000 },
141 { 0x1A, 0x1800000, 0x1800000, 0x6800000, 0xC000000 },
142};
143
144static int set_bps_params(AVCodecContext *avctx)
145{
146 switch (avctx->bits_per_raw_sample) {
147 case 8:
148 avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
149 break;
150 case 16:
151 avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
152 break;
153 case 24:
154 avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
155 break;
156 default:
157 av_log(avctx, AV_LOG_ERROR, "invalid/unsupported bits per sample: %d\n",
158 avctx->bits_per_raw_sample);
159 return AVERROR_INVALIDDATA;
160 }
161
162 return 0;
163}
164
165static void set_sample_rate_params(AVCodecContext *avctx)
166{
167 TAKDecContext *s = avctx->priv_data;
168 int shift;
169
170 if (avctx->sample_rate < 11025) {
171 shift = 3;
172 } else if (avctx->sample_rate < 22050) {
173 shift = 2;
174 } else if (avctx->sample_rate < 44100) {
175 shift = 1;
176 } else {
177 shift = 0;
178 }
179 s->uval = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << shift;
180 s->subframe_scale = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << 1;
181}
182
183static av_cold int tak_decode_init(AVCodecContext *avctx)
184{
185 TAKDecContext *s = avctx->priv_data;
186
187 ff_audiodsp_init(&s->adsp);
188 ff_takdsp_init(&s->tdsp);
189
190 s->avctx = avctx;
191 avctx->bits_per_raw_sample = avctx->bits_per_coded_sample;
192
193 set_sample_rate_params(avctx);
194
195 return set_bps_params(avctx);
196}
197
198static void decode_lpc(int32_t *coeffs, int mode, int length)
199{
200 int i;
201
202 if (length < 2)
203 return;
204
205 if (mode == 1) {
206 int a1 = *coeffs++;
207 for (i = 0; i < length - 1 >> 1; i++) {
208 *coeffs += a1;
209 coeffs[1] += *coeffs;
210 a1 = coeffs[1];
211 coeffs += 2;
212 }
213 if (length - 1 & 1)
214 *coeffs += a1;
215 } else if (mode == 2) {
216 int a1 = coeffs[1];
217 int a2 = a1 + *coeffs;
218 coeffs[1] = a2;
219 if (length > 2) {
220 coeffs += 2;
221 for (i = 0; i < length - 2 >> 1; i++) {
222 int a3 = *coeffs + a1;
223 int a4 = a3 + a2;
224 *coeffs = a4;
225 a1 = coeffs[1] + a3;
226 a2 = a1 + a4;
227 coeffs[1] = a2;
228 coeffs += 2;
229 }
230 if (length & 1)
231 *coeffs += a1 + a2;
232 }
233 } else if (mode == 3) {
234 int a1 = coeffs[1];
235 int a2 = a1 + *coeffs;
236 coeffs[1] = a2;
237 if (length > 2) {
238 int a3 = coeffs[2];
239 int a4 = a3 + a1;
240 int a5 = a4 + a2;
241 coeffs[2] = a5;
242 coeffs += 3;
243 for (i = 0; i < length - 3; i++) {
244 a3 += *coeffs;
245 a4 += a3;
246 a5 += a4;
247 *coeffs = a5;
248 coeffs++;
249 }
250 }
251 }
252}
253
254static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
255{
256 struct CParam code;
257 GetBitContext *gb = &s->gb;
258 int i;
259
260 if (!mode) {
261 memset(decoded, 0, len * sizeof(*decoded));
262 return 0;
263 }
264
265 if (mode > FF_ARRAY_ELEMS(xcodes))
266 return AVERROR_INVALIDDATA;
267 code = xcodes[mode - 1];
268
269 for (i = 0; i < len; i++) {
270 int x = get_bits_long(gb, code.init);
271 if (x >= code.escape && get_bits1(gb)) {
272 x |= 1 << code.init;
273 if (x >= code.aescape) {
274 int scale = get_unary(gb, 1, 9);
275 if (scale == 9) {
276 int scale_bits = get_bits(gb, 3);
277 if (scale_bits > 0) {
278 if (scale_bits == 7) {
279 scale_bits += get_bits(gb, 5);
280 if (scale_bits > 29)
281 return AVERROR_INVALIDDATA;
282 }
283 scale = get_bits_long(gb, scale_bits) + 1;
284 x += code.scale * scale;
285 }
286 x += code.bias;
287 } else
288 x += code.scale * scale - code.escape;
289 } else
290 x -= code.escape;
291 }
292 decoded[i] = (x >> 1) ^ -(x & 1);
293 }
294
295 return 0;
296}
297
298static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
299{
300 GetBitContext *gb = &s->gb;
301 int i, mode, ret;
302
303 if (length > s->nb_samples)
304 return AVERROR_INVALIDDATA;
305
306 if (get_bits1(gb)) {
307 int wlength, rval;
308
309 wlength = length / s->uval;
310
311 rval = length - (wlength * s->uval);
312
313 if (rval < s->uval / 2)
314 rval += s->uval;
315 else
316 wlength++;
317
318 if (wlength <= 1 || wlength > 128)
319 return AVERROR_INVALIDDATA;
320
321 s->coding_mode[0] = mode = get_bits(gb, 6);
322
323 for (i = 1; i < wlength; i++) {
324 int c = get_unary(gb, 1, 6);
325
326 switch (c) {
327 case 6:
328 mode = get_bits(gb, 6);
329 break;
330 case 5:
331 case 4:
332 case 3: {
333 /* mode += sign ? (1 - c) : (c - 1) */
334 int sign = get_bits1(gb);
335 mode += (-sign ^ (c - 1)) + sign;
336 break;
337 }
338 case 2:
339 mode++;
340 break;
341 case 1:
342 mode--;
343 break;
344 }
345 s->coding_mode[i] = mode;
346 }
347
348 i = 0;
349 while (i < wlength) {
350 int len = 0;
351
352 mode = s->coding_mode[i];
353 do {
354 if (i >= wlength - 1)
355 len += rval;
356 else
357 len += s->uval;
358 i++;
359
360 if (i == wlength)
361 break;
362 } while (s->coding_mode[i] == mode);
363
364 if ((ret = decode_segment(s, mode, decoded, len)) < 0)
365 return ret;
366 decoded += len;
367 }
368 } else {
369 mode = get_bits(gb, 6);
370 if ((ret = decode_segment(s, mode, decoded, length)) < 0)
371 return ret;
372 }
373
374 return 0;
375}
376
377static int get_bits_esc4(GetBitContext *gb)
378{
379 if (get_bits1(gb))
380 return get_bits(gb, 4) + 1;
381 else
382 return 0;
383}
384
385static int decode_subframe(TAKDecContext *s, int32_t *decoded,
386 int subframe_size, int prev_subframe_size)
387{
388 GetBitContext *gb = &s->gb;
389 int x, y, i, j, ret = 0;
390 int dshift, size, filter_quant, filter_order;
391 int tfilter[MAX_PREDICTORS];
392
393 if (!get_bits1(gb))
394 return decode_residues(s, decoded, subframe_size);
395
396 filter_order = predictor_sizes[get_bits(gb, 4)];
397
398 if (prev_subframe_size > 0 && get_bits1(gb)) {
399 if (filter_order > prev_subframe_size)
400 return AVERROR_INVALIDDATA;
401
402 decoded -= filter_order;
403 subframe_size += filter_order;
404
405 if (filter_order > subframe_size)
406 return AVERROR_INVALIDDATA;
407 } else {
408 int lpc_mode;
409
410 if (filter_order > subframe_size)
411 return AVERROR_INVALIDDATA;
412
413 lpc_mode = get_bits(gb, 2);
414 if (lpc_mode > 2)
415 return AVERROR_INVALIDDATA;
416
417 if ((ret = decode_residues(s, decoded, filter_order)) < 0)
418 return ret;
419
420 if (lpc_mode)
421 decode_lpc(decoded, lpc_mode, filter_order);
422 }
423
424 dshift = get_bits_esc4(gb);
425 size = get_bits1(gb) + 6;
426
427 filter_quant = 10;
428 if (get_bits1(gb)) {
429 filter_quant -= get_bits(gb, 3) + 1;
430 if (filter_quant < 3)
431 return AVERROR_INVALIDDATA;
432 }
433
434 s->predictors[0] = get_sbits(gb, 10);
435 s->predictors[1] = get_sbits(gb, 10);
436 s->predictors[2] = get_sbits(gb, size) << (10 - size);
437 s->predictors[3] = get_sbits(gb, size) << (10 - size);
438 if (filter_order > 4) {
439 int tmp = size - get_bits1(gb);
440
441 for (i = 4; i < filter_order; i++) {
442 if (!(i & 3))
443 x = tmp - get_bits(gb, 2);
444 s->predictors[i] = get_sbits(gb, x) << (10 - size);
445 }
446 }
447
448 tfilter[0] = s->predictors[0] << 6;
449 for (i = 1; i < filter_order; i++) {
450 int32_t *p1 = &tfilter[0];
451 int32_t *p2 = &tfilter[i - 1];
452
453 for (j = 0; j < (i + 1) / 2; j++) {
454 x = *p1 + (s->predictors[i] * *p2 + 256 >> 9);
455 *p2 += s->predictors[i] * *p1 + 256 >> 9;
456 *p1++ = x;
457 p2--;
458 }
459
460 tfilter[i] = s->predictors[i] << 6;
461 }
462
463 x = 1 << (32 - (15 - filter_quant));
464 y = 1 << ((15 - filter_quant) - 1);
465 for (i = 0, j = filter_order - 1; i < filter_order / 2; i++, j--) {
466 s->filter[j] = x - ((tfilter[i] + y) >> (15 - filter_quant));
467 s->filter[i] = x - ((tfilter[j] + y) >> (15 - filter_quant));
468 }
469
470 if ((ret = decode_residues(s, &decoded[filter_order],
471 subframe_size - filter_order)) < 0)
472 return ret;
473
474 for (i = 0; i < filter_order; i++)
475 s->residues[i] = *decoded++ >> dshift;
476
477 y = FF_ARRAY_ELEMS(s->residues) - filter_order;
478 x = subframe_size - filter_order;
479 while (x > 0) {
480 int tmp = FFMIN(y, x);
481
482 for (i = 0; i < tmp; i++) {
483 int v = 1 << (filter_quant - 1);
484
485 if (filter_order & -16)
486 v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
487 filter_order & -16);
488 for (j = filter_order & -16; j < filter_order; j += 4) {
489 v += s->residues[i + j + 3] * s->filter[j + 3] +
490 s->residues[i + j + 2] * s->filter[j + 2] +
491 s->residues[i + j + 1] * s->filter[j + 1] +
492 s->residues[i + j ] * s->filter[j ];
493 }
494 v = (av_clip_intp2(v >> filter_quant, 13) << dshift) - *decoded;
495 *decoded++ = v;
496 s->residues[filter_order + i] = v >> dshift;
497 }
498
499 x -= tmp;
500 if (x > 0)
501 memcpy(s->residues, &s->residues[y], 2 * filter_order);
502 }
503
504 emms_c();
505
506 return 0;
507}
508
509static int decode_channel(TAKDecContext *s, int chan)
510{
511 AVCodecContext *avctx = s->avctx;
512 GetBitContext *gb = &s->gb;
513 int32_t *decoded = s->decoded[chan];
514 int left = s->nb_samples - 1;
515 int i = 0, ret, prev = 0;
516
517 s->sample_shift[chan] = get_bits_esc4(gb);
518 if (s->sample_shift[chan] >= avctx->bits_per_raw_sample)
519 return AVERROR_INVALIDDATA;
520
521 *decoded++ = get_sbits(gb, avctx->bits_per_raw_sample - s->sample_shift[chan]);
522 s->lpc_mode[chan] = get_bits(gb, 2);
523 s->nb_subframes = get_bits(gb, 3) + 1;
524
525 if (s->nb_subframes > 1) {
526 if (get_bits_left(gb) < (s->nb_subframes - 1) * 6)
527 return AVERROR_INVALIDDATA;
528
529 for (; i < s->nb_subframes - 1; i++) {
530 int v = get_bits(gb, 6);
531
532 s->subframe_len[i] = (v - prev) * s->subframe_scale;
533 if (s->subframe_len[i] <= 0)
534 return AVERROR_INVALIDDATA;
535
536 left -= s->subframe_len[i];
537 prev = v;
538 }
539
540 if (left <= 0)
541 return AVERROR_INVALIDDATA;
542 }
543 s->subframe_len[i] = left;
544
545 prev = 0;
546 for (i = 0; i < s->nb_subframes; i++) {
547 if ((ret = decode_subframe(s, decoded, s->subframe_len[i], prev)) < 0)
548 return ret;
549 decoded += s->subframe_len[i];
550 prev = s->subframe_len[i];
551 }
552
553 return 0;
554}
555
556static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
557{
558 GetBitContext *gb = &s->gb;
559 int32_t *p1 = s->decoded[c1] + (s->dmode > 5);
560 int32_t *p2 = s->decoded[c2] + (s->dmode > 5);
561 int32_t bp1 = p1[0];
562 int32_t bp2 = p2[0];
563 int i;
564 int dshift, dfactor;
565
566 length += s->dmode < 6;
567
568 switch (s->dmode) {
569 case 1: /* left/side */
570 s->tdsp.decorrelate_ls(p1, p2, length);
571 break;
572 case 2: /* side/right */
573 s->tdsp.decorrelate_sr(p1, p2, length);
574 break;
575 case 3: /* side/mid */
576 s->tdsp.decorrelate_sm(p1, p2, length);
577 break;
578 case 4: /* side/left with scale factor */
579 FFSWAP(int32_t*, p1, p2);
580 FFSWAP(int32_t, bp1, bp2);
581 case 5: /* side/right with scale factor */
582 dshift = get_bits_esc4(gb);
583 dfactor = get_sbits(gb, 10);
584 s->tdsp.decorrelate_sf(p1, p2, length, dshift, dfactor);
585 break;
586 case 6:
587 FFSWAP(int32_t*, p1, p2);
588 case 7: {
589 int length2, order_half, filter_order, dval1, dval2;
590 int tmp, x, code_size;
591
592 if (length < 256)
593 return AVERROR_INVALIDDATA;
594
595 dshift = get_bits_esc4(gb);
596 filter_order = 8 << get_bits1(gb);
597 dval1 = get_bits1(gb);
598 dval2 = get_bits1(gb);
599
600 for (i = 0; i < filter_order; i++) {
601 if (!(i & 3))
602 code_size = 14 - get_bits(gb, 3);
603 s->filter[i] = get_sbits(gb, code_size);
604 }
605
606 order_half = filter_order / 2;
607 length2 = length - (filter_order - 1);
608
609 /* decorrelate beginning samples */
610 if (dval1) {
611 for (i = 0; i < order_half; i++) {
612 int32_t a = p1[i];
613 int32_t b = p2[i];
614 p1[i] = a + b;
615 }
616 }
617
618 /* decorrelate ending samples */
619 if (dval2) {
620 for (i = length2 + order_half; i < length; i++) {
621 int32_t a = p1[i];
622 int32_t b = p2[i];
623 p1[i] = a + b;
624 }
625 }
626
627
628 for (i = 0; i < filter_order; i++)
629 s->residues[i] = *p2++ >> dshift;
630
631 p1 += order_half;
632 x = FF_ARRAY_ELEMS(s->residues) - filter_order;
633 for (; length2 > 0; length2 -= tmp) {
634 tmp = FFMIN(length2, x);
635
636 for (i = 0; i < tmp - (tmp == length2); i++)
637 s->residues[filter_order + i] = *p2++ >> dshift;
638
639 for (i = 0; i < tmp; i++) {
640 int v = 1 << 9;
641
642 if (filter_order == 16) {
643 v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
644 filter_order);
645 } else {
646 v += s->residues[i + 7] * s->filter[7] +
647 s->residues[i + 6] * s->filter[6] +
648 s->residues[i + 5] * s->filter[5] +
649 s->residues[i + 4] * s->filter[4] +
650 s->residues[i + 3] * s->filter[3] +
651 s->residues[i + 2] * s->filter[2] +
652 s->residues[i + 1] * s->filter[1] +
653 s->residues[i ] * s->filter[0];
654 }
655
656 v = (av_clip_intp2(v >> 10, 13) << dshift) - *p1;
657 *p1++ = v;
658 }
659
660 memmove(s->residues, &s->residues[tmp], 2 * filter_order);
661 }
662
663 emms_c();
664 break;
665 }
666 }
667
668 if (s->dmode > 0 && s->dmode < 6) {
669 p1[0] = bp1;
670 p2[0] = bp2;
671 }
672
673 return 0;
674}
675
676static int tak_decode_frame(AVCodecContext *avctx, void *data,
677 int *got_frame_ptr, AVPacket *pkt)
678{
679 TAKDecContext *s = avctx->priv_data;
680 AVFrame *frame = data;
681 ThreadFrame tframe = { .f = data };
682 GetBitContext *gb = &s->gb;
683 int chan, i, ret, hsize;
684
685 if (pkt->size < TAK_MIN_FRAME_HEADER_BYTES)
686 return AVERROR_INVALIDDATA;
687
688 if ((ret = init_get_bits8(gb, pkt->data, pkt->size)) < 0)
689 return ret;
690
691 if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
692 return ret;
693
694 hsize = get_bits_count(gb) / 8;
695 if (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_COMPLIANT)) {
696 if (ff_tak_check_crc(pkt->data, hsize)) {
697 av_log(avctx, AV_LOG_ERROR, "CRC error\n");
698 if (avctx->err_recognition & AV_EF_EXPLODE)
699 return AVERROR_INVALIDDATA;
700 }
701 }
702
703 if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
704 s->ti.codec != TAK_CODEC_MULTICHANNEL) {
705 avpriv_report_missing_feature(avctx, "TAK codec type %d", s->ti.codec);
706 return AVERROR_PATCHWELCOME;
707 }
708 if (s->ti.data_type) {
709 av_log(avctx, AV_LOG_ERROR,
710 "unsupported data type: %d\n", s->ti.data_type);
711 return AVERROR_INVALIDDATA;
712 }
713 if (s->ti.codec == TAK_CODEC_MONO_STEREO && s->ti.channels > 2) {
714 av_log(avctx, AV_LOG_ERROR,
715 "invalid number of channels: %d\n", s->ti.channels);
716 return AVERROR_INVALIDDATA;
717 }
718 if (s->ti.channels > 6) {
719 av_log(avctx, AV_LOG_ERROR,
720 "unsupported number of channels: %d\n", s->ti.channels);
721 return AVERROR_INVALIDDATA;
722 }
723
724 if (s->ti.frame_samples <= 0) {
725 av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
726 return AVERROR_INVALIDDATA;
727 }
728
729 avctx->bits_per_raw_sample = s->ti.bps;
730 if ((ret = set_bps_params(avctx)) < 0)
731 return ret;
732 if (s->ti.sample_rate != avctx->sample_rate) {
733 avctx->sample_rate = s->ti.sample_rate;
734 set_sample_rate_params(avctx);
735 }
736 if (s->ti.ch_layout)
737 avctx->channel_layout = s->ti.ch_layout;
738 avctx->channels = s->ti.channels;
739
740 s->nb_samples = s->ti.last_frame_samples ? s->ti.last_frame_samples
741 : s->ti.frame_samples;
742
743 frame->nb_samples = s->nb_samples;
744 if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0)
745 return ret;
746 ff_thread_finish_setup(avctx);
747
748 if (avctx->bits_per_raw_sample <= 16) {
749 int buf_size = av_samples_get_buffer_size(NULL, avctx->channels,
750 s->nb_samples,
751 AV_SAMPLE_FMT_S32P, 0);
752 if (buf_size < 0)
753 return buf_size;
754 av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
755 if (!s->decode_buffer)
756 return AVERROR(ENOMEM);
757 ret = av_samples_fill_arrays((uint8_t **)s->decoded, NULL,
758 s->decode_buffer, avctx->channels,
759 s->nb_samples, AV_SAMPLE_FMT_S32P, 0);
760 if (ret < 0)
761 return ret;
762 } else {
763 for (chan = 0; chan < avctx->channels; chan++)
764 s->decoded[chan] = (int32_t *)frame->extended_data[chan];
765 }
766
767 if (s->nb_samples < 16) {
768 for (chan = 0; chan < avctx->channels; chan++) {
769 int32_t *decoded = s->decoded[chan];
770 for (i = 0; i < s->nb_samples; i++)
771 decoded[i] = get_sbits(gb, avctx->bits_per_raw_sample);
772 }
773 } else {
774 if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
775 for (chan = 0; chan < avctx->channels; chan++)
776 if (ret = decode_channel(s, chan))
777 return ret;
778
779 if (avctx->channels == 2) {
780 s->nb_subframes = get_bits(gb, 1) + 1;
781 if (s->nb_subframes > 1) {
782 s->subframe_len[1] = get_bits(gb, 6);
783 }
784
785 s->dmode = get_bits(gb, 3);
786 if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
787 return ret;
788 }
789 } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
790 if (get_bits1(gb)) {
791 int ch_mask = 0;
792
793 chan = get_bits(gb, 4) + 1;
794 if (chan > avctx->channels)
795 return AVERROR_INVALIDDATA;
796
797 for (i = 0; i < chan; i++) {
798 int nbit = get_bits(gb, 4);
799
800 if (nbit >= avctx->channels)
801 return AVERROR_INVALIDDATA;
802
803 if (ch_mask & 1 << nbit)
804 return AVERROR_INVALIDDATA;
805
806 s->mcdparams[i].present = get_bits1(gb);
807 if (s->mcdparams[i].present) {
808 s->mcdparams[i].index = get_bits(gb, 2);
809 s->mcdparams[i].chan2 = get_bits(gb, 4);
810 if (s->mcdparams[i].chan2 >= avctx->channels) {
811 av_log(avctx, AV_LOG_ERROR,
812 "invalid channel 2 (%d) for %d channel(s)\n",
813 s->mcdparams[i].chan2, avctx->channels);
814 return AVERROR_INVALIDDATA;
815 }
816 if (s->mcdparams[i].index == 1) {
817 if ((nbit == s->mcdparams[i].chan2) ||
818 (ch_mask & 1 << s->mcdparams[i].chan2))
819 return AVERROR_INVALIDDATA;
820
821 ch_mask |= 1 << s->mcdparams[i].chan2;
822 } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
823 return AVERROR_INVALIDDATA;
824 }
825 }
826 s->mcdparams[i].chan1 = nbit;
827
828 ch_mask |= 1 << nbit;
829 }
830 } else {
831 chan = avctx->channels;
832 for (i = 0; i < chan; i++) {
833 s->mcdparams[i].present = 0;
834 s->mcdparams[i].chan1 = i;
835 }
836 }
837
838 for (i = 0; i < chan; i++) {
839 if (s->mcdparams[i].present && s->mcdparams[i].index == 1)
840 if (ret = decode_channel(s, s->mcdparams[i].chan2))
841 return ret;
842
843 if (ret = decode_channel(s, s->mcdparams[i].chan1))
844 return ret;
845
846 if (s->mcdparams[i].present) {
847 s->dmode = mc_dmodes[s->mcdparams[i].index];
848 if (ret = decorrelate(s,
849 s->mcdparams[i].chan2,
850 s->mcdparams[i].chan1,
851 s->nb_samples - 1))
852 return ret;
853 }
854 }
855 }
856
857 for (chan = 0; chan < avctx->channels; chan++) {
858 int32_t *decoded = s->decoded[chan];
859
860 if (s->lpc_mode[chan])
861 decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
862
863 if (s->sample_shift[chan] > 0)
864 for (i = 0; i < s->nb_samples; i++)
865 decoded[i] <<= s->sample_shift[chan];
866 }
867 }
868
869 align_get_bits(gb);
870 skip_bits(gb, 24);
871 if (get_bits_left(gb) < 0)
872 av_log(avctx, AV_LOG_DEBUG, "overread\n");
873 else if (get_bits_left(gb) > 0)
874 av_log(avctx, AV_LOG_DEBUG, "underread\n");
875
876 if (avctx->err_recognition & (AV_EF_CRCCHECK | AV_EF_COMPLIANT)) {
877 if (ff_tak_check_crc(pkt->data + hsize,
878 get_bits_count(gb) / 8 - hsize)) {
879 av_log(avctx, AV_LOG_ERROR, "CRC error\n");
880 if (avctx->err_recognition & AV_EF_EXPLODE)
881 return AVERROR_INVALIDDATA;
882 }
883 }
884
885 /* convert to output buffer */
886 switch (avctx->sample_fmt) {
887 case AV_SAMPLE_FMT_U8P:
888 for (chan = 0; chan < avctx->channels; chan++) {
889 uint8_t *samples = (uint8_t *)frame->extended_data[chan];
890 int32_t *decoded = s->decoded[chan];
891 for (i = 0; i < s->nb_samples; i++)
892 samples[i] = decoded[i] + 0x80;
893 }
894 break;
895 case AV_SAMPLE_FMT_S16P:
896 for (chan = 0; chan < avctx->channels; chan++) {
897 int16_t *samples = (int16_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];
901 }
902 break;
903 case AV_SAMPLE_FMT_S32P:
904 for (chan = 0; chan < avctx->channels; chan++) {
905 int32_t *samples = (int32_t *)frame->extended_data[chan];
906 for (i = 0; i < s->nb_samples; i++)
907 samples[i] <<= 8;
908 }
909 break;
910 }
911
912 *got_frame_ptr = 1;
913
914 return pkt->size;
915}
916
917#if HAVE_THREADS
918static int init_thread_copy(AVCodecContext *avctx)
919{
920 TAKDecContext *s = avctx->priv_data;
921 s->avctx = avctx;
922 return 0;
923}
924
925static int update_thread_context(AVCodecContext *dst,
926 const AVCodecContext *src)
927{
928 TAKDecContext *tsrc = src->priv_data;
929 TAKDecContext *tdst = dst->priv_data;
930
931 if (dst == src)
932 return 0;
933 memcpy(&tdst->ti, &tsrc->ti, sizeof(TAKStreamInfo));
934 return 0;
935}
936#endif
937
938static av_cold int tak_decode_close(AVCodecContext *avctx)
939{
940 TAKDecContext *s = avctx->priv_data;
941
942 av_freep(&s->decode_buffer);
943
944 return 0;
945}
946
947AVCodec ff_tak_decoder = {
948 .name = "tak",
949 .long_name = NULL_IF_CONFIG_SMALL("TAK (Tom's lossless Audio Kompressor)"),
950 .type = AVMEDIA_TYPE_AUDIO,
951 .id = AV_CODEC_ID_TAK,
952 .priv_data_size = sizeof(TAKDecContext),
953 .init = tak_decode_init,
954 .close = tak_decode_close,
955 .decode = tak_decode_frame,
956 .init_thread_copy = ONLY_IF_THREADS_ENABLED(init_thread_copy),
957 .update_thread_context = ONLY_IF_THREADS_ENABLED(update_thread_context),
958 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
959 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
960 AV_SAMPLE_FMT_S16P,
961 AV_SAMPLE_FMT_S32P,
962 AV_SAMPLE_FMT_NONE },
963};
964
generated by cgit at 2024-05-17 03:24:04 (GMT)