FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
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 
28 #include "libavutil/internal.h"
29 #include "libavutil/samplefmt.h"
30 #include "tak.h"
31 #include "takdsp.h"
32 #include "audiodsp.h"
33 #include "thread.h"
34 #include "avcodec.h"
35 #include "internal.h"
36 #include "unary.h"
37 
38 #define MAX_SUBFRAMES 8 ///< max number of subframes per channel
39 #define MAX_PREDICTORS 256
40 
41 typedef struct MCDParam {
42  int8_t present; ///< decorrelation parameter availability for this channel
43  int8_t index; ///< index into array of decorrelation types
44  int8_t chan1;
45  int8_t chan2;
46 } MCDParam;
47 
48 typedef struct TAKDecContext {
49  AVCodecContext *avctx; ///< parent AVCodecContext
53  GetBitContext gb; ///< bitstream reader initialized to start at the current frame
54 
55  int uval;
56  int nb_samples; ///< number of samples in the current frame
58  unsigned int decode_buffer_size;
59  int32_t *decoded[TAK_MAX_CHANNELS]; ///< decoded samples for each channel
60 
62  int8_t sample_shift[TAK_MAX_CHANNELS]; ///< shift applied to every sample in the channel
64  int nb_subframes; ///< number of subframes in the current frame
65  int16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples
67 
68  int8_t dmode; ///< channel decorrelation type in the current frame
69 
70  MCDParam mcdparams[TAK_MAX_CHANNELS]; ///< multichannel decorrelation parameters
71 
72  int8_t coding_mode[128];
74  DECLARE_ALIGNED(16, int16_t, residues)[544];
76 
77 static const int8_t mc_dmodes[] = { 1, 3, 4, 6, };
78 
79 static const uint16_t predictor_sizes[] = {
80  4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
81 };
82 
83 static const struct CParam {
84  int init;
85  int escape;
86  int scale;
87  int aescape;
88  int bias;
89 } xcodes[50] = {
90  { 0x01, 0x0000001, 0x0000001, 0x0000003, 0x0000008 },
91  { 0x02, 0x0000003, 0x0000001, 0x0000007, 0x0000006 },
92  { 0x03, 0x0000005, 0x0000002, 0x000000E, 0x000000D },
93  { 0x03, 0x0000003, 0x0000003, 0x000000D, 0x0000018 },
94  { 0x04, 0x000000B, 0x0000004, 0x000001C, 0x0000019 },
95  { 0x04, 0x0000006, 0x0000006, 0x000001A, 0x0000030 },
96  { 0x05, 0x0000016, 0x0000008, 0x0000038, 0x0000032 },
97  { 0x05, 0x000000C, 0x000000C, 0x0000034, 0x0000060 },
98  { 0x06, 0x000002C, 0x0000010, 0x0000070, 0x0000064 },
99  { 0x06, 0x0000018, 0x0000018, 0x0000068, 0x00000C0 },
100  { 0x07, 0x0000058, 0x0000020, 0x00000E0, 0x00000C8 },
101  { 0x07, 0x0000030, 0x0000030, 0x00000D0, 0x0000180 },
102  { 0x08, 0x00000B0, 0x0000040, 0x00001C0, 0x0000190 },
103  { 0x08, 0x0000060, 0x0000060, 0x00001A0, 0x0000300 },
104  { 0x09, 0x0000160, 0x0000080, 0x0000380, 0x0000320 },
105  { 0x09, 0x00000C0, 0x00000C0, 0x0000340, 0x0000600 },
106  { 0x0A, 0x00002C0, 0x0000100, 0x0000700, 0x0000640 },
107  { 0x0A, 0x0000180, 0x0000180, 0x0000680, 0x0000C00 },
108  { 0x0B, 0x0000580, 0x0000200, 0x0000E00, 0x0000C80 },
109  { 0x0B, 0x0000300, 0x0000300, 0x0000D00, 0x0001800 },
110  { 0x0C, 0x0000B00, 0x0000400, 0x0001C00, 0x0001900 },
111  { 0x0C, 0x0000600, 0x0000600, 0x0001A00, 0x0003000 },
112  { 0x0D, 0x0001600, 0x0000800, 0x0003800, 0x0003200 },
113  { 0x0D, 0x0000C00, 0x0000C00, 0x0003400, 0x0006000 },
114  { 0x0E, 0x0002C00, 0x0001000, 0x0007000, 0x0006400 },
115  { 0x0E, 0x0001800, 0x0001800, 0x0006800, 0x000C000 },
116  { 0x0F, 0x0005800, 0x0002000, 0x000E000, 0x000C800 },
117  { 0x0F, 0x0003000, 0x0003000, 0x000D000, 0x0018000 },
118  { 0x10, 0x000B000, 0x0004000, 0x001C000, 0x0019000 },
119  { 0x10, 0x0006000, 0x0006000, 0x001A000, 0x0030000 },
120  { 0x11, 0x0016000, 0x0008000, 0x0038000, 0x0032000 },
121  { 0x11, 0x000C000, 0x000C000, 0x0034000, 0x0060000 },
122  { 0x12, 0x002C000, 0x0010000, 0x0070000, 0x0064000 },
123  { 0x12, 0x0018000, 0x0018000, 0x0068000, 0x00C0000 },
124  { 0x13, 0x0058000, 0x0020000, 0x00E0000, 0x00C8000 },
125  { 0x13, 0x0030000, 0x0030000, 0x00D0000, 0x0180000 },
126  { 0x14, 0x00B0000, 0x0040000, 0x01C0000, 0x0190000 },
127  { 0x14, 0x0060000, 0x0060000, 0x01A0000, 0x0300000 },
128  { 0x15, 0x0160000, 0x0080000, 0x0380000, 0x0320000 },
129  { 0x15, 0x00C0000, 0x00C0000, 0x0340000, 0x0600000 },
130  { 0x16, 0x02C0000, 0x0100000, 0x0700000, 0x0640000 },
131  { 0x16, 0x0180000, 0x0180000, 0x0680000, 0x0C00000 },
132  { 0x17, 0x0580000, 0x0200000, 0x0E00000, 0x0C80000 },
133  { 0x17, 0x0300000, 0x0300000, 0x0D00000, 0x1800000 },
134  { 0x18, 0x0B00000, 0x0400000, 0x1C00000, 0x1900000 },
135  { 0x18, 0x0600000, 0x0600000, 0x1A00000, 0x3000000 },
136  { 0x19, 0x1600000, 0x0800000, 0x3800000, 0x3200000 },
137  { 0x19, 0x0C00000, 0x0C00000, 0x3400000, 0x6000000 },
138  { 0x1A, 0x2C00000, 0x1000000, 0x7000000, 0x6400000 },
139  { 0x1A, 0x1800000, 0x1800000, 0x6800000, 0xC000000 },
140 };
141 
142 static int set_bps_params(AVCodecContext *avctx)
143 {
144  switch (avctx->bits_per_raw_sample) {
145  case 8:
146  avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
147  break;
148  case 16:
150  break;
151  case 24:
153  break;
154  default:
155  av_log(avctx, AV_LOG_ERROR, "invalid/unsupported bits per sample: %d\n",
156  avctx->bits_per_raw_sample);
157  return AVERROR_INVALIDDATA;
158  }
159 
160  return 0;
161 }
162 
164 {
165  TAKDecContext *s = avctx->priv_data;
166  int shift;
167 
168  if (avctx->sample_rate < 11025) {
169  shift = 3;
170  } else if (avctx->sample_rate < 22050) {
171  shift = 2;
172  } else if (avctx->sample_rate < 44100) {
173  shift = 1;
174  } else {
175  shift = 0;
176  }
177  s->uval = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << shift;
178  s->subframe_scale = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << 1;
179 }
180 
182 {
183  TAKDecContext *s = avctx->priv_data;
184 
185  ff_audiodsp_init(&s->adsp);
186  ff_takdsp_init(&s->tdsp);
187 
188  s->avctx = avctx;
190 
191  set_sample_rate_params(avctx);
192 
193  return set_bps_params(avctx);
194 }
195 
196 static void decode_lpc(int32_t *coeffs, int mode, int length)
197 {
198  int i;
199 
200  if (length < 2)
201  return;
202 
203  if (mode == 1) {
204  int a1 = *coeffs++;
205  for (i = 0; i < length - 1 >> 1; i++) {
206  *coeffs += a1;
207  coeffs[1] += *coeffs;
208  a1 = coeffs[1];
209  coeffs += 2;
210  }
211  if (length - 1 & 1)
212  *coeffs += a1;
213  } else if (mode == 2) {
214  int a1 = coeffs[1];
215  int a2 = a1 + *coeffs;
216  coeffs[1] = a2;
217  if (length > 2) {
218  coeffs += 2;
219  for (i = 0; i < length - 2 >> 1; i++) {
220  int a3 = *coeffs + a1;
221  int a4 = a3 + a2;
222  *coeffs = a4;
223  a1 = coeffs[1] + a3;
224  a2 = a1 + a4;
225  coeffs[1] = a2;
226  coeffs += 2;
227  }
228  if (length & 1)
229  *coeffs += a1 + a2;
230  }
231  } else if (mode == 3) {
232  int a1 = coeffs[1];
233  int a2 = a1 + *coeffs;
234  coeffs[1] = a2;
235  if (length > 2) {
236  int a3 = coeffs[2];
237  int a4 = a3 + a1;
238  int a5 = a4 + a2;
239  coeffs[2] = a5;
240  coeffs += 3;
241  for (i = 0; i < length - 3; i++) {
242  a3 += *coeffs;
243  a4 += a3;
244  a5 += a4;
245  *coeffs = a5;
246  coeffs++;
247  }
248  }
249  }
250 }
251 
252 static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
253 {
254  struct CParam code;
255  GetBitContext *gb = &s->gb;
256  int i;
257 
258  if (!mode) {
259  memset(decoded, 0, len * sizeof(*decoded));
260  return 0;
261  }
262 
263  if (mode > FF_ARRAY_ELEMS(xcodes))
264  return AVERROR_INVALIDDATA;
265  code = xcodes[mode - 1];
266 
267  for (i = 0; i < len; i++) {
268  int x = get_bits_long(gb, code.init);
269  if (x >= code.escape && get_bits1(gb)) {
270  x |= 1 << code.init;
271  if (x >= code.aescape) {
272  int scale = get_unary(gb, 1, 9);
273  if (scale == 9) {
274  int scale_bits = get_bits(gb, 3);
275  if (scale_bits > 0) {
276  if (scale_bits == 7) {
277  scale_bits += get_bits(gb, 5);
278  if (scale_bits > 29)
279  return AVERROR_INVALIDDATA;
280  }
281  scale = get_bits_long(gb, scale_bits) + 1;
282  x += code.scale * scale;
283  }
284  x += code.bias;
285  } else
286  x += code.scale * scale - code.escape;
287  } else
288  x -= code.escape;
289  }
290  decoded[i] = (x >> 1) ^ -(x & 1);
291  }
292 
293  return 0;
294 }
295 
296 static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
297 {
298  GetBitContext *gb = &s->gb;
299  int i, mode, ret;
300 
301  if (length > s->nb_samples)
302  return AVERROR_INVALIDDATA;
303 
304  if (get_bits1(gb)) {
305  int wlength, rval;
306 
307  wlength = length / s->uval;
308 
309  rval = length - (wlength * s->uval);
310 
311  if (rval < s->uval / 2)
312  rval += s->uval;
313  else
314  wlength++;
315 
316  if (wlength <= 1 || wlength > 128)
317  return AVERROR_INVALIDDATA;
318 
319  s->coding_mode[0] = mode = get_bits(gb, 6);
320 
321  for (i = 1; i < wlength; i++) {
322  int c = get_unary(gb, 1, 6);
323 
324  switch (c) {
325  case 6:
326  mode = get_bits(gb, 6);
327  break;
328  case 5:
329  case 4:
330  case 3: {
331  /* mode += sign ? (1 - c) : (c - 1) */
332  int sign = get_bits1(gb);
333  mode += (-sign ^ (c - 1)) + sign;
334  break;
335  }
336  case 2:
337  mode++;
338  break;
339  case 1:
340  mode--;
341  break;
342  }
343  s->coding_mode[i] = mode;
344  }
345 
346  i = 0;
347  while (i < wlength) {
348  int len = 0;
349 
350  mode = s->coding_mode[i];
351  do {
352  if (i >= wlength - 1)
353  len += rval;
354  else
355  len += s->uval;
356  i++;
357 
358  if (i == wlength)
359  break;
360  } while (s->coding_mode[i] == mode);
361 
362  if ((ret = decode_segment(s, mode, decoded, len)) < 0)
363  return ret;
364  decoded += len;
365  }
366  } else {
367  mode = get_bits(gb, 6);
368  if ((ret = decode_segment(s, mode, decoded, length)) < 0)
369  return ret;
370  }
371 
372  return 0;
373 }
374 
376 {
377  if (get_bits1(gb))
378  return get_bits(gb, 4) + 1;
379  else
380  return 0;
381 }
382 
383 static int decode_subframe(TAKDecContext *s, int32_t *decoded,
384  int subframe_size, int prev_subframe_size)
385 {
386  GetBitContext *gb = &s->gb;
387  int x, y, i, j, ret = 0;
388  int dshift, size, filter_quant, filter_order;
389  int tfilter[MAX_PREDICTORS];
390 
391  if (!get_bits1(gb))
392  return decode_residues(s, decoded, subframe_size);
393 
394  filter_order = predictor_sizes[get_bits(gb, 4)];
395 
396  if (prev_subframe_size > 0 && get_bits1(gb)) {
397  if (filter_order > prev_subframe_size)
398  return AVERROR_INVALIDDATA;
399 
400  decoded -= filter_order;
401  subframe_size += filter_order;
402 
403  if (filter_order > subframe_size)
404  return AVERROR_INVALIDDATA;
405  } else {
406  int lpc_mode;
407 
408  if (filter_order > subframe_size)
409  return AVERROR_INVALIDDATA;
410 
411  lpc_mode = get_bits(gb, 2);
412  if (lpc_mode > 2)
413  return AVERROR_INVALIDDATA;
414 
415  if ((ret = decode_residues(s, decoded, filter_order)) < 0)
416  return ret;
417 
418  if (lpc_mode)
419  decode_lpc(decoded, lpc_mode, filter_order);
420  }
421 
422  dshift = get_bits_esc4(gb);
423  size = get_bits1(gb) + 6;
424 
425  filter_quant = 10;
426  if (get_bits1(gb)) {
427  filter_quant -= get_bits(gb, 3) + 1;
428  if (filter_quant < 3)
429  return AVERROR_INVALIDDATA;
430  }
431 
432  s->predictors[0] = get_sbits(gb, 10);
433  s->predictors[1] = get_sbits(gb, 10);
434  s->predictors[2] = get_sbits(gb, size) << (10 - size);
435  s->predictors[3] = get_sbits(gb, size) << (10 - size);
436  if (filter_order > 4) {
437  int tmp = size - get_bits1(gb);
438 
439  for (i = 4; i < filter_order; i++) {
440  if (!(i & 3))
441  x = tmp - get_bits(gb, 2);
442  s->predictors[i] = get_sbits(gb, x) << (10 - size);
443  }
444  }
445 
446  tfilter[0] = s->predictors[0] << 6;
447  for (i = 1; i < filter_order; i++) {
448  int32_t *p1 = &tfilter[0];
449  int32_t *p2 = &tfilter[i - 1];
450 
451  for (j = 0; j < (i + 1) / 2; j++) {
452  x = *p1 + (s->predictors[i] * *p2 + 256 >> 9);
453  *p2 += s->predictors[i] * *p1 + 256 >> 9;
454  *p1++ = x;
455  p2--;
456  }
457 
458  tfilter[i] = s->predictors[i] << 6;
459  }
460 
461  x = 1 << (32 - (15 - filter_quant));
462  y = 1 << ((15 - filter_quant) - 1);
463  for (i = 0, j = filter_order - 1; i < filter_order / 2; i++, j--) {
464  s->filter[j] = x - ((tfilter[i] + y) >> (15 - filter_quant));
465  s->filter[i] = x - ((tfilter[j] + y) >> (15 - filter_quant));
466  }
467 
468  if ((ret = decode_residues(s, &decoded[filter_order],
469  subframe_size - filter_order)) < 0)
470  return ret;
471 
472  for (i = 0; i < filter_order; i++)
473  s->residues[i] = *decoded++ >> dshift;
474 
475  y = FF_ARRAY_ELEMS(s->residues) - filter_order;
476  x = subframe_size - filter_order;
477  while (x > 0) {
478  int tmp = FFMIN(y, x);
479 
480  for (i = 0; i < tmp; i++) {
481  int v = 1 << (filter_quant - 1);
482 
483  if (filter_order & -16)
484  v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
485  filter_order & -16);
486  for (j = filter_order & -16; j < filter_order; j += 4) {
487  v += s->residues[i + j + 3] * s->filter[j + 3] +
488  s->residues[i + j + 2] * s->filter[j + 2] +
489  s->residues[i + j + 1] * s->filter[j + 1] +
490  s->residues[i + j ] * s->filter[j ];
491  }
492  v = (av_clip_intp2(v >> filter_quant, 13) << dshift) - *decoded;
493  *decoded++ = v;
494  s->residues[filter_order + i] = v >> dshift;
495  }
496 
497  x -= tmp;
498  if (x > 0)
499  memcpy(s->residues, &s->residues[y], 2 * filter_order);
500  }
501 
502  emms_c();
503 
504  return 0;
505 }
506 
507 static int decode_channel(TAKDecContext *s, int chan)
508 {
509  AVCodecContext *avctx = s->avctx;
510  GetBitContext *gb = &s->gb;
511  int32_t *decoded = s->decoded[chan];
512  int left = s->nb_samples - 1;
513  int i = 0, ret, prev = 0;
514 
515  s->sample_shift[chan] = get_bits_esc4(gb);
516  if (s->sample_shift[chan] >= avctx->bits_per_raw_sample)
517  return AVERROR_INVALIDDATA;
518 
519  *decoded++ = get_sbits(gb, avctx->bits_per_raw_sample - s->sample_shift[chan]);
520  s->lpc_mode[chan] = get_bits(gb, 2);
521  s->nb_subframes = get_bits(gb, 3) + 1;
522 
523  if (s->nb_subframes > 1) {
524  if (get_bits_left(gb) < (s->nb_subframes - 1) * 6)
525  return AVERROR_INVALIDDATA;
526 
527  for (; i < s->nb_subframes - 1; i++) {
528  int v = get_bits(gb, 6);
529 
530  s->subframe_len[i] = (v - prev) * s->subframe_scale;
531  if (s->subframe_len[i] <= 0)
532  return AVERROR_INVALIDDATA;
533 
534  left -= s->subframe_len[i];
535  prev = v;
536  }
537 
538  if (left <= 0)
539  return AVERROR_INVALIDDATA;
540  }
541  s->subframe_len[i] = left;
542 
543  prev = 0;
544  for (i = 0; i < s->nb_subframes; i++) {
545  if ((ret = decode_subframe(s, decoded, s->subframe_len[i], prev)) < 0)
546  return ret;
547  decoded += s->subframe_len[i];
548  prev = s->subframe_len[i];
549  }
550 
551  return 0;
552 }
553 
554 static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
555 {
556  GetBitContext *gb = &s->gb;
557  int32_t *p1 = s->decoded[c1] + (s->dmode > 5);
558  int32_t *p2 = s->decoded[c2] + (s->dmode > 5);
559  int32_t bp1 = p1[0];
560  int32_t bp2 = p2[0];
561  int i;
562  int dshift, dfactor;
563 
564  length += s->dmode < 6;
565 
566  switch (s->dmode) {
567  case 1: /* left/side */
568  s->tdsp.decorrelate_ls(p1, p2, length);
569  break;
570  case 2: /* side/right */
571  s->tdsp.decorrelate_sr(p1, p2, length);
572  break;
573  case 3: /* side/mid */
574  s->tdsp.decorrelate_sm(p1, p2, length);
575  break;
576  case 4: /* side/left with scale factor */
577  FFSWAP(int32_t*, p1, p2);
578  FFSWAP(int32_t, bp1, bp2);
579  case 5: /* side/right with scale factor */
580  dshift = get_bits_esc4(gb);
581  dfactor = get_sbits(gb, 10);
582  s->tdsp.decorrelate_sf(p1, p2, length, dshift, dfactor);
583  break;
584  case 6:
585  FFSWAP(int32_t*, p1, p2);
586  case 7: {
587  int length2, order_half, filter_order, dval1, dval2;
588  int tmp, x, code_size;
589 
590  if (length < 256)
591  return AVERROR_INVALIDDATA;
592 
593  dshift = get_bits_esc4(gb);
594  filter_order = 8 << get_bits1(gb);
595  dval1 = get_bits1(gb);
596  dval2 = get_bits1(gb);
597 
598  for (i = 0; i < filter_order; i++) {
599  if (!(i & 3))
600  code_size = 14 - get_bits(gb, 3);
601  s->filter[i] = get_sbits(gb, code_size);
602  }
603 
604  order_half = filter_order / 2;
605  length2 = length - (filter_order - 1);
606 
607  /* decorrelate beginning samples */
608  if (dval1) {
609  for (i = 0; i < order_half; i++) {
610  int32_t a = p1[i];
611  int32_t b = p2[i];
612  p1[i] = a + b;
613  }
614  }
615 
616  /* decorrelate ending samples */
617  if (dval2) {
618  for (i = length2 + order_half; i < length; i++) {
619  int32_t a = p1[i];
620  int32_t b = p2[i];
621  p1[i] = a + b;
622  }
623  }
624 
625 
626  for (i = 0; i < filter_order; i++)
627  s->residues[i] = *p2++ >> dshift;
628 
629  p1 += order_half;
630  x = FF_ARRAY_ELEMS(s->residues) - filter_order;
631  for (; length2 > 0; length2 -= tmp) {
632  tmp = FFMIN(length2, x);
633 
634  for (i = 0; i < tmp - (tmp == length2); i++)
635  s->residues[filter_order + i] = *p2++ >> dshift;
636 
637  for (i = 0; i < tmp; i++) {
638  int v = 1 << 9;
639 
640  if (filter_order == 16) {
641  v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
642  filter_order);
643  } else {
644  v += s->residues[i + 7] * s->filter[7] +
645  s->residues[i + 6] * s->filter[6] +
646  s->residues[i + 5] * s->filter[5] +
647  s->residues[i + 4] * s->filter[4] +
648  s->residues[i + 3] * s->filter[3] +
649  s->residues[i + 2] * s->filter[2] +
650  s->residues[i + 1] * s->filter[1] +
651  s->residues[i ] * s->filter[0];
652  }
653 
654  v = (av_clip_intp2(v >> 10, 13) << dshift) - *p1;
655  *p1++ = v;
656  }
657 
658  memmove(s->residues, &s->residues[tmp], 2 * filter_order);
659  }
660 
661  emms_c();
662  break;
663  }
664  }
665 
666  if (s->dmode > 0 && s->dmode < 6) {
667  p1[0] = bp1;
668  p2[0] = bp2;
669  }
670 
671  return 0;
672 }
673 
674 static int tak_decode_frame(AVCodecContext *avctx, void *data,
675  int *got_frame_ptr, AVPacket *pkt)
676 {
677  TAKDecContext *s = avctx->priv_data;
678  AVFrame *frame = data;
679  ThreadFrame tframe = { .f = data };
680  GetBitContext *gb = &s->gb;
681  int chan, i, ret, hsize;
682 
683  if (pkt->size < TAK_MIN_FRAME_HEADER_BYTES)
684  return AVERROR_INVALIDDATA;
685 
686  if ((ret = init_get_bits8(gb, pkt->data, pkt->size)) < 0)
687  return ret;
688 
689  if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
690  return ret;
691 
692  hsize = get_bits_count(gb) / 8;
694  if (ff_tak_check_crc(pkt->data, hsize)) {
695  av_log(avctx, AV_LOG_ERROR, "CRC error\n");
696  if (avctx->err_recognition & AV_EF_EXPLODE)
697  return AVERROR_INVALIDDATA;
698  }
699  }
700 
701  if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
703  av_log(avctx, AV_LOG_ERROR, "unsupported codec: %d\n", s->ti.codec);
704  return AVERROR_PATCHWELCOME;
705  }
706  if (s->ti.data_type) {
707  av_log(avctx, AV_LOG_ERROR,
708  "unsupported data type: %d\n", s->ti.data_type);
709  return AVERROR_INVALIDDATA;
710  }
711  if (s->ti.codec == TAK_CODEC_MONO_STEREO && s->ti.channels > 2) {
712  av_log(avctx, AV_LOG_ERROR,
713  "invalid number of channels: %d\n", s->ti.channels);
714  return AVERROR_INVALIDDATA;
715  }
716  if (s->ti.channels > 6) {
717  av_log(avctx, AV_LOG_ERROR,
718  "unsupported number of channels: %d\n", s->ti.channels);
719  return AVERROR_INVALIDDATA;
720  }
721 
722  if (s->ti.frame_samples <= 0) {
723  av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
724  return AVERROR_INVALIDDATA;
725  }
726 
727  avctx->bits_per_raw_sample = s->ti.bps;
728  if ((ret = set_bps_params(avctx)) < 0)
729  return ret;
730  if (s->ti.sample_rate != avctx->sample_rate) {
731  avctx->sample_rate = s->ti.sample_rate;
732  set_sample_rate_params(avctx);
733  }
734  if (s->ti.ch_layout)
735  avctx->channel_layout = s->ti.ch_layout;
736  avctx->channels = s->ti.channels;
737 
739  : s->ti.frame_samples;
740 
741  frame->nb_samples = s->nb_samples;
742  if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0)
743  return ret;
744  ff_thread_finish_setup(avctx);
745 
746  if (avctx->bits_per_raw_sample <= 16) {
747  int buf_size = av_samples_get_buffer_size(NULL, avctx->channels,
748  s->nb_samples,
749  AV_SAMPLE_FMT_S32P, 0);
750  if (buf_size < 0)
751  return buf_size;
752  av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
753  if (!s->decode_buffer)
754  return AVERROR(ENOMEM);
756  s->decode_buffer, avctx->channels,
758  if (ret < 0)
759  return ret;
760  } else {
761  for (chan = 0; chan < avctx->channels; chan++)
762  s->decoded[chan] = (int32_t *)frame->extended_data[chan];
763  }
764 
765  if (s->nb_samples < 16) {
766  for (chan = 0; chan < avctx->channels; chan++) {
767  int32_t *decoded = s->decoded[chan];
768  for (i = 0; i < s->nb_samples; i++)
769  decoded[i] = get_sbits(gb, avctx->bits_per_raw_sample);
770  }
771  } else {
772  if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
773  for (chan = 0; chan < avctx->channels; chan++)
774  if (ret = decode_channel(s, chan))
775  return ret;
776 
777  if (avctx->channels == 2) {
778  s->nb_subframes = get_bits(gb, 1) + 1;
779  if (s->nb_subframes > 1) {
780  s->subframe_len[1] = get_bits(gb, 6);
781  }
782 
783  s->dmode = get_bits(gb, 3);
784  if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
785  return ret;
786  }
787  } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
788  if (get_bits1(gb)) {
789  int ch_mask = 0;
790 
791  chan = get_bits(gb, 4) + 1;
792  if (chan > avctx->channels)
793  return AVERROR_INVALIDDATA;
794 
795  for (i = 0; i < chan; i++) {
796  int nbit = get_bits(gb, 4);
797 
798  if (nbit >= avctx->channels)
799  return AVERROR_INVALIDDATA;
800 
801  if (ch_mask & 1 << nbit)
802  return AVERROR_INVALIDDATA;
803 
804  s->mcdparams[i].present = get_bits1(gb);
805  if (s->mcdparams[i].present) {
806  s->mcdparams[i].index = get_bits(gb, 2);
807  s->mcdparams[i].chan2 = get_bits(gb, 4);
808  if (s->mcdparams[i].chan2 >= avctx->channels) {
809  av_log(avctx, AV_LOG_ERROR,
810  "invalid channel 2 (%d) for %d channel(s)\n",
811  s->mcdparams[i].chan2, avctx->channels);
812  return AVERROR_INVALIDDATA;
813  }
814  if (s->mcdparams[i].index == 1) {
815  if ((nbit == s->mcdparams[i].chan2) ||
816  (ch_mask & 1 << s->mcdparams[i].chan2))
817  return AVERROR_INVALIDDATA;
818 
819  ch_mask |= 1 << s->mcdparams[i].chan2;
820  } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
821  return AVERROR_INVALIDDATA;
822  }
823  }
824  s->mcdparams[i].chan1 = nbit;
825 
826  ch_mask |= 1 << nbit;
827  }
828  } else {
829  chan = avctx->channels;
830  for (i = 0; i < chan; i++) {
831  s->mcdparams[i].present = 0;
832  s->mcdparams[i].chan1 = i;
833  }
834  }
835 
836  for (i = 0; i < chan; i++) {
837  if (s->mcdparams[i].present && s->mcdparams[i].index == 1)
838  if (ret = decode_channel(s, s->mcdparams[i].chan2))
839  return ret;
840 
841  if (ret = decode_channel(s, s->mcdparams[i].chan1))
842  return ret;
843 
844  if (s->mcdparams[i].present) {
845  s->dmode = mc_dmodes[s->mcdparams[i].index];
846  if (ret = decorrelate(s,
847  s->mcdparams[i].chan2,
848  s->mcdparams[i].chan1,
849  s->nb_samples - 1))
850  return ret;
851  }
852  }
853  }
854 
855  for (chan = 0; chan < avctx->channels; chan++) {
856  int32_t *decoded = s->decoded[chan];
857 
858  if (s->lpc_mode[chan])
859  decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
860 
861  if (s->sample_shift[chan] > 0)
862  for (i = 0; i < s->nb_samples; i++)
863  decoded[i] <<= s->sample_shift[chan];
864  }
865  }
866 
867  align_get_bits(gb);
868  skip_bits(gb, 24);
869  if (get_bits_left(gb) < 0)
870  av_log(avctx, AV_LOG_DEBUG, "overread\n");
871  else if (get_bits_left(gb) > 0)
872  av_log(avctx, AV_LOG_DEBUG, "underread\n");
873 
875  if (ff_tak_check_crc(pkt->data + hsize,
876  get_bits_count(gb) / 8 - hsize)) {
877  av_log(avctx, AV_LOG_ERROR, "CRC error\n");
878  if (avctx->err_recognition & AV_EF_EXPLODE)
879  return AVERROR_INVALIDDATA;
880  }
881  }
882 
883  /* convert to output buffer */
884  switch (avctx->sample_fmt) {
885  case AV_SAMPLE_FMT_U8P:
886  for (chan = 0; chan < avctx->channels; chan++) {
887  uint8_t *samples = (uint8_t *)frame->extended_data[chan];
888  int32_t *decoded = s->decoded[chan];
889  for (i = 0; i < s->nb_samples; i++)
890  samples[i] = decoded[i] + 0x80;
891  }
892  break;
893  case AV_SAMPLE_FMT_S16P:
894  for (chan = 0; chan < avctx->channels; chan++) {
895  int16_t *samples = (int16_t *)frame->extended_data[chan];
896  int32_t *decoded = s->decoded[chan];
897  for (i = 0; i < s->nb_samples; i++)
898  samples[i] = decoded[i];
899  }
900  break;
901  case AV_SAMPLE_FMT_S32P:
902  for (chan = 0; chan < avctx->channels; chan++) {
903  int32_t *samples = (int32_t *)frame->extended_data[chan];
904  for (i = 0; i < s->nb_samples; i++)
905  samples[i] <<= 8;
906  }
907  break;
908  }
909 
910  *got_frame_ptr = 1;
911 
912  return pkt->size;
913 }
914 
915 #if HAVE_THREADS
916 static int init_thread_copy(AVCodecContext *avctx)
917 {
918  TAKDecContext *s = avctx->priv_data;
919  s->avctx = avctx;
920  return 0;
921 }
922 
923 static int update_thread_context(AVCodecContext *dst,
924  const AVCodecContext *src)
925 {
926  TAKDecContext *tsrc = src->priv_data;
927  TAKDecContext *tdst = dst->priv_data;
928 
929  if (dst == src)
930  return 0;
931  memcpy(&tdst->ti, &tsrc->ti, sizeof(TAKStreamInfo));
932  return 0;
933 }
934 #endif
935 
937 {
938  TAKDecContext *s = avctx->priv_data;
939 
940  av_freep(&s->decode_buffer);
941 
942  return 0;
943 }
944 
946  .name = "tak",
947  .long_name = NULL_IF_CONFIG_SMALL("TAK (Tom's lossless Audio Kompressor)"),
948  .type = AVMEDIA_TYPE_AUDIO,
949  .id = AV_CODEC_ID_TAK,
950  .priv_data_size = sizeof(TAKDecContext),
952  .close = tak_decode_close,
955  .update_thread_context = ONLY_IF_THREADS_ENABLED(update_thread_context),
957  .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
961 };
#define TAK_MAX_CHANNELS
Definition: tak.h:65
#define NULL
Definition: coverity.c:32
static int set_bps_params(AVCodecContext *avctx)
Definition: takdec.c:142
const char * s
Definition: avisynth_c.h:631
static const uint16_t predictor_sizes[]
Definition: takdec.c:79
#define AVERROR_INVALIDDATA
Invalid data found when processing input.
Definition: error.h:59
int ff_tak_decode_frame_header(AVCodecContext *avctx, GetBitContext *gb, TAKStreamInfo *ti, int log_level_offset)
Validate and decode a frame header.
Definition: tak.c:126
int32_t(* scalarproduct_int16)(const int16_t *v1, const int16_t *v2, int len)
Calculate scalar product of two vectors.
Definition: audiodsp.h:29
static int shift(int a, int b)
Definition: sonic.c:82
int16_t subframe_len[MAX_SUBFRAMES]
subframe length in samples
Definition: takdec.c:65
This structure describes decoded (raw) audio or video data.
Definition: frame.h:180
int channels
Definition: tak.h:134
ptrdiff_t const GLvoid * data
Definition: opengl_enc.c:101
av_cold void ff_takdsp_init(TAKDSPContext *c)
Definition: takdsp.c:73
static unsigned int get_bits(GetBitContext *s, int n)
Read 1-25 bits.
Definition: get_bits.h:260
static int init_thread_copy(AVCodecContext *avctx)
Definition: tta.c:390
AVFrame * f
Definition: thread.h:36
TAKDSPContext tdsp
Definition: takdec.c:51
int aescape
Definition: takdec.c:87
static av_cold int init(AVCodecContext *avctx)
Definition: avrndec.c:35
MCDParam mcdparams[TAK_MAX_CHANNELS]
multichannel decorrelation parameters
Definition: takdec.c:70
int16_t filter[MAX_PREDICTORS]
Definition: takdec.c:73
int size
Definition: avcodec.h:1578
const char * b
Definition: vf_curves.c:109
#define AV_EF_COMPLIANT
consider all spec non compliances as errors
Definition: avcodec.h:2944
static int decode_subframe(TAKDecContext *s, int32_t *decoded, int subframe_size, int prev_subframe_size)
Definition: takdec.c:383
#define DECLARE_ALIGNED(n, t, v)
Definition: mem.h:53
av_cold void ff_audiodsp_init(AudioDSPContext *c)
Definition: audiodsp.c:106
uint64_t ch_layout
Definition: tak.h:139
#define a1
Definition: regdef.h:47
static AVPacket pkt
int bits_per_raw_sample
Bits per sample/pixel of internal libavcodec pixel/sample format.
Definition: avcodec.h:3042
static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
Definition: takdec.c:252
void(* decorrelate_ls)(int32_t *p1, int32_t *p2, int length)
Definition: takdsp.h:25
AVCodec.
Definition: avcodec.h:3535
static int get_sbits(GetBitContext *s, int n)
Definition: get_bits.h:245
int8_t lpc_mode[TAK_MAX_CHANNELS]
Definition: takdec.c:61
unsigned int decode_buffer_size
Definition: takdec.c:58
#define a3
Definition: regdef.h:49
int init
Definition: takdec.c:84
enum AVSampleFormat sample_fmt
audio sample format
Definition: avcodec.h:2411
uint8_t
#define av_cold
Definition: attributes.h:82
mode
Definition: f_perms.c:27
static av_cold int tak_decode_close(AVCodecContext *avctx)
Definition: takdec.c:936
Multithreading support functions.
int ff_tak_check_crc(const uint8_t *buf, unsigned int buf_size)
Definition: tak.c:75
static AVFrame * frame
AVCodecContext * avctx
parent AVCodecContext
Definition: takdec.c:49
uint8_t * data
Definition: avcodec.h:1577
static int get_bits_count(const GetBitContext *s)
Definition: get_bits.h:212
static const uint64_t c1
Definition: murmur3.c:49
int escape
Definition: takdec.c:85
ptrdiff_t size
Definition: opengl_enc.c:101
int bits_per_coded_sample
bits per sample/pixel from the demuxer (needed for huffyuv).
Definition: avcodec.h:3035
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 FFALIGN(x, a)
Definition: macros.h:48
int data_type
Definition: tak.h:132
static int decode_channel(TAKDecContext *s, int chan)
Definition: takdec.c:507
#define av_log(a,...)
#define MAX_SUBFRAMES
max number of subframes per channel
Definition: takdec.c:38
int8_t index
index into array of decorrelation types
Definition: takdec.c:43
static int get_bits_left(GetBitContext *gb)
Definition: get_bits.h:607
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:176
int uval
Definition: takdec.c:55
#define AVERROR(e)
Definition: error.h:43
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:176
#define AV_LOG_DEBUG
Stuff which is only useful for libav* developers.
Definition: log.h:197
GLsizei GLsizei * length
Definition: opengl_enc.c:115
const char * name
Name of the codec implementation.
Definition: avcodec.h:3542
int8_t present
decorrelation parameter availability for this channel
Definition: takdec.c:42
int last_frame_samples
Definition: tak.h:138
static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
Definition: takdec.c:296
#define AV_CODEC_CAP_FRAME_THREADS
Codec supports frame-level multithreading.
Definition: avcodec.h:1016
uint64_t channel_layout
Audio channel layout.
Definition: avcodec.h:2454
#define ONLY_IF_THREADS_ENABLED(x)
Define a function with only the non-default version specified.
Definition: internal.h:215
int bps
Definition: tak.h:135
common internal API header
int8_t coding_mode[128]
Definition: takdec.c:72
static av_cold int tak_decode_init(AVCodecContext *avctx)
Definition: takdec.c:181
int8_t chan2
Definition: takdec.c:45
void(* decorrelate_sr)(int32_t *p1, int32_t *p2, int length)
Definition: takdsp.h:26
int err_recognition
Error recognition; may misdetect some more or less valid parts as errors.
Definition: avcodec.h:2929
#define FFMIN(a, b)
Definition: common.h:96
signed 32 bits, planar
Definition: samplefmt.h:68
static const struct CParam xcodes[50]
void(* decorrelate_sm)(int32_t *p1, int32_t *p2, int length)
Definition: takdsp.h:27
int scale
Definition: takdec.c:86
static uint8_t tmp[8]
Definition: des-test.c:38
int32_t
#define a2
Definition: regdef.h:48
static int get_bits_esc4(GetBitContext *gb)
Definition: takdec.c:375
void(* decorrelate_sf)(int32_t *p1, int32_t *p2, int length, int dshift, int dfactor)
Definition: takdsp.h:28
#define AV_EF_EXPLODE
abort decoding on minor error detection
Definition: avcodec.h:2940
unsigned 8 bits, planar
Definition: samplefmt.h:66
AudioDSPContext adsp
Definition: takdec.c:50
#define src
Definition: vp9dsp.c:530
uint8_t * decode_buffer
Definition: takdec.c:57
static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
Definition: takdec.c:554
static void decode_lpc(int32_t *coeffs, int mode, int length)
Definition: takdec.c:196
#define FF_ARRAY_ELEMS(a)
#define AVERROR_PATCHWELCOME
Not yet implemented in FFmpeg, patches welcome.
Definition: error.h:62
AVCodec ff_tak_decoder
Definition: takdec.c:945
Definition: takdec.c:83
#define a5
Definition: regdef.h:51
Libavcodec external API header.
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:58
#define TAK_MIN_FRAME_HEADER_BYTES
Definition: tak.h:98
int sample_rate
samples per second
Definition: avcodec.h:2403
int16_t predictors[MAX_PREDICTORS]
Definition: takdec.c:63
static int init_get_bits8(GetBitContext *s, const uint8_t *buffer, int byte_size)
Initialize GetBitContext.
Definition: get_bits.h:449
int ff_thread_get_buffer(AVCodecContext *avctx, ThreadFrame *f, int flags)
Wrapper around get_buffer() for frame-multithreaded codecs.
main external API structure.
Definition: avcodec.h:1642
enum TAKCodecType codec
Definition: tak.h:131
int32_t * decoded[TAK_MAX_CHANNELS]
decoded samples for each channel
Definition: takdec.c:59
static unsigned int get_bits1(GetBitContext *s)
Definition: get_bits.h:312
static void skip_bits(GetBitContext *s, int n)
Definition: get_bits.h:305
int frame_samples
Definition: tak.h:137
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:117
TAK (Tom's lossless Audio Kompressor) decoder/demuxer common functions.
GetBitContext gb
bitstream reader initialized to start at the current frame
Definition: takdec.c:53
int8_t dmode
channel decorrelation type in the current frame
Definition: takdec.c:68
#define AV_EF_CRCCHECK
Verify checksums embedded in the bitstream (could be of either encoded or decoded data...
Definition: avcodec.h:2937
static unsigned int get_bits_long(GetBitContext *s, int n)
Read 0-32 bits.
Definition: get_bits.h:345
int sample_rate
Definition: tak.h:133
static int decode(AVCodecContext *avctx, void *data, int *got_sub, AVPacket *avpkt)
Definition: ccaption_dec.c:722
int bias
Definition: takdec.c:88
int nb_samples
number of samples in the current frame
Definition: takdec.c:56
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:499
TAKStreamInfo ti
Definition: takdec.c:52
common internal api header.
if(ret< 0)
Definition: vf_mcdeint.c:282
int8_t chan1
Definition: takdec.c:44
static int get_unary(GetBitContext *gb, int stop, int len)
Get unary code of limited length.
Definition: unary.h:33
static double c[64]
int16_t residues[544]
Definition: takdec.c:74
static const uint64_t c2
Definition: murmur3.c:50
static const int8_t mc_dmodes[]
Definition: takdec.c:77
int subframe_scale
Definition: takdec.c:66
int nb_subframes
number of subframes in the current frame
Definition: takdec.c:64
#define a4
Definition: regdef.h:50
void * priv_data
Definition: avcodec.h:1684
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:149
static int tak_decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, AVPacket *pkt)
Definition: takdec.c:674
static const int16_t coeffs[]
int len
int channels
number of audio channels
Definition: avcodec.h:2404
static const uint8_t * align_get_bits(GetBitContext *s)
Definition: get_bits.h:457
#define MAX_PREDICTORS
Definition: takdec.c:39
static void set_sample_rate_params(AVCodecContext *avctx)
Definition: takdec.c:163
#define av_freep(p)
int8_t sample_shift[TAK_MAX_CHANNELS]
shift applied to every sample in the channel
Definition: takdec.c:62
signed 16 bits, planar
Definition: samplefmt.h:67
#define FFSWAP(type, a, b)
Definition: common.h:99
uint8_t ** extended_data
pointers to the data planes/channels.
Definition: frame.h:227
This structure stores compressed data.
Definition: avcodec.h:1554
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:237
#define AV_CODEC_CAP_DR1
Codec uses get_buffer() for allocating buffers and supports custom allocators.
Definition: avcodec.h:953
for(j=16;j >0;--j)