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 = 3 - (avctx->sample_rate / 11025);
167  shift = FFMAX(0, shift);
168  s->uval = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << shift;
169  s->subframe_scale = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << 1;
170 }
171 
173 {
174  TAKDecContext *s = avctx->priv_data;
175 
176  ff_audiodsp_init(&s->adsp);
177  ff_takdsp_init(&s->tdsp);
178 
179  s->avctx = avctx;
181 
182  set_sample_rate_params(avctx);
183 
184  return set_bps_params(avctx);
185 }
186 
187 static void decode_lpc(int32_t *coeffs, int mode, int length)
188 {
189  int i;
190 
191  if (length < 2)
192  return;
193 
194  if (mode == 1) {
195  int a1 = *coeffs++;
196  for (i = 0; i < length - 1 >> 1; i++) {
197  *coeffs += a1;
198  coeffs[1] += *coeffs;
199  a1 = coeffs[1];
200  coeffs += 2;
201  }
202  if (length - 1 & 1)
203  *coeffs += a1;
204  } else if (mode == 2) {
205  int a1 = coeffs[1];
206  int a2 = a1 + *coeffs;
207  coeffs[1] = a2;
208  if (length > 2) {
209  coeffs += 2;
210  for (i = 0; i < length - 2 >> 1; i++) {
211  int a3 = *coeffs + a1;
212  int a4 = a3 + a2;
213  *coeffs = a4;
214  a1 = coeffs[1] + a3;
215  a2 = a1 + a4;
216  coeffs[1] = a2;
217  coeffs += 2;
218  }
219  if (length & 1)
220  *coeffs += a1 + a2;
221  }
222  } else if (mode == 3) {
223  int a1 = coeffs[1];
224  int a2 = a1 + *coeffs;
225  coeffs[1] = a2;
226  if (length > 2) {
227  int a3 = coeffs[2];
228  int a4 = a3 + a1;
229  int a5 = a4 + a2;
230  coeffs += 3;
231  for (i = 0; i < length - 3; i++) {
232  a3 += *coeffs;
233  a4 += a3;
234  a5 += a4;
235  *coeffs = a5;
236  coeffs++;
237  }
238  }
239  }
240 }
241 
242 static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
243 {
244  struct CParam code;
245  GetBitContext *gb = &s->gb;
246  int i;
247 
248  if (!mode) {
249  memset(decoded, 0, len * sizeof(*decoded));
250  return 0;
251  }
252 
253  if (mode > FF_ARRAY_ELEMS(xcodes))
254  return AVERROR_INVALIDDATA;
255  code = xcodes[mode - 1];
256 
257  for (i = 0; i < len; i++) {
258  int x = get_bits_long(gb, code.init);
259  if (x >= code.escape && get_bits1(gb)) {
260  x |= 1 << code.init;
261  if (x >= code.aescape) {
262  int scale = get_unary(gb, 1, 9);
263  if (scale == 9) {
264  int scale_bits = get_bits(gb, 3);
265  if (scale_bits > 0) {
266  if (scale_bits == 7) {
267  scale_bits += get_bits(gb, 5);
268  if (scale_bits > 29)
269  return AVERROR_INVALIDDATA;
270  }
271  scale = get_bits_long(gb, scale_bits) + 1;
272  x += code.scale * scale;
273  }
274  x += code.bias;
275  } else
276  x += code.scale * scale - code.escape;
277  } else
278  x -= code.escape;
279  }
280  decoded[i] = (x >> 1) ^ -(x & 1);
281  }
282 
283  return 0;
284 }
285 
286 static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
287 {
288  GetBitContext *gb = &s->gb;
289  int i, mode, ret;
290 
291  if (length > s->nb_samples)
292  return AVERROR_INVALIDDATA;
293 
294  if (get_bits1(gb)) {
295  int wlength, rval;
296 
297  wlength = length / s->uval;
298 
299  rval = length - (wlength * s->uval);
300 
301  if (rval < s->uval / 2)
302  rval += s->uval;
303  else
304  wlength++;
305 
306  if (wlength <= 1 || wlength > 128)
307  return AVERROR_INVALIDDATA;
308 
309  s->coding_mode[0] = mode = get_bits(gb, 6);
310 
311  for (i = 1; i < wlength; i++) {
312  int c = get_unary(gb, 1, 6);
313 
314  switch (c) {
315  case 6:
316  mode = get_bits(gb, 6);
317  break;
318  case 5:
319  case 4:
320  case 3: {
321  /* mode += sign ? (1 - c) : (c - 1) */
322  int sign = get_bits1(gb);
323  mode += (-sign ^ (c - 1)) + sign;
324  break;
325  }
326  case 2:
327  mode++;
328  break;
329  case 1:
330  mode--;
331  break;
332  }
333  s->coding_mode[i] = mode;
334  }
335 
336  i = 0;
337  while (i < wlength) {
338  int len = 0;
339 
340  mode = s->coding_mode[i];
341  do {
342  if (i >= wlength - 1)
343  len += rval;
344  else
345  len += s->uval;
346  i++;
347 
348  if (i == wlength)
349  break;
350  } while (s->coding_mode[i] == mode);
351 
352  if ((ret = decode_segment(s, mode, decoded, len)) < 0)
353  return ret;
354  decoded += len;
355  }
356  } else {
357  mode = get_bits(gb, 6);
358  if ((ret = decode_segment(s, mode, decoded, length)) < 0)
359  return ret;
360  }
361 
362  return 0;
363 }
364 
366 {
367  if (get_bits1(gb))
368  return get_bits(gb, 4) + 1;
369  else
370  return 0;
371 }
372 
373 static int decode_subframe(TAKDecContext *s, int32_t *decoded,
374  int subframe_size, int prev_subframe_size)
375 {
376  GetBitContext *gb = &s->gb;
377  int x, y, i, j, ret = 0;
378  int dshift, size, filter_quant, filter_order;
379  int tfilter[MAX_PREDICTORS];
380 
381  if (!get_bits1(gb))
382  return decode_residues(s, decoded, subframe_size);
383 
384  filter_order = predictor_sizes[get_bits(gb, 4)];
385 
386  if (prev_subframe_size > 0 && get_bits1(gb)) {
387  if (filter_order > prev_subframe_size)
388  return AVERROR_INVALIDDATA;
389 
390  decoded -= filter_order;
391  subframe_size += filter_order;
392 
393  if (filter_order > subframe_size)
394  return AVERROR_INVALIDDATA;
395  } else {
396  int lpc_mode;
397 
398  if (filter_order > subframe_size)
399  return AVERROR_INVALIDDATA;
400 
401  lpc_mode = get_bits(gb, 2);
402  if (lpc_mode > 2)
403  return AVERROR_INVALIDDATA;
404 
405  if ((ret = decode_residues(s, decoded, filter_order)) < 0)
406  return ret;
407 
408  if (lpc_mode)
409  decode_lpc(decoded, lpc_mode, filter_order);
410  }
411 
412  dshift = get_bits_esc4(gb);
413  size = get_bits1(gb) + 6;
414 
415  filter_quant = 10;
416  if (get_bits1(gb)) {
417  filter_quant -= get_bits(gb, 3) + 1;
418  if (filter_quant < 3)
419  return AVERROR_INVALIDDATA;
420  }
421 
422  s->predictors[0] = get_sbits(gb, 10);
423  s->predictors[1] = get_sbits(gb, 10);
424  s->predictors[2] = get_sbits(gb, size) << (10 - size);
425  s->predictors[3] = get_sbits(gb, size) << (10 - size);
426  if (filter_order > 4) {
427  int tmp = size - get_bits1(gb);
428 
429  for (i = 4; i < filter_order; i++) {
430  if (!(i & 3))
431  x = tmp - get_bits(gb, 2);
432  s->predictors[i] = get_sbits(gb, x) << (10 - size);
433  }
434  }
435 
436  tfilter[0] = s->predictors[0] << 6;
437  for (i = 1; i < filter_order; i++) {
438  int32_t *p1 = &tfilter[0];
439  int32_t *p2 = &tfilter[i - 1];
440 
441  for (j = 0; j < (i + 1) / 2; j++) {
442  x = *p1 + (s->predictors[i] * *p2 + 256 >> 9);
443  *p2 += s->predictors[i] * *p1 + 256 >> 9;
444  *p1++ = x;
445  p2--;
446  }
447 
448  tfilter[i] = s->predictors[i] << 6;
449  }
450 
451  x = 1 << (32 - (15 - filter_quant));
452  y = 1 << ((15 - filter_quant) - 1);
453  for (i = 0, j = filter_order - 1; i < filter_order / 2; i++, j--) {
454  s->filter[j] = x - ((tfilter[i] + y) >> (15 - filter_quant));
455  s->filter[i] = x - ((tfilter[j] + y) >> (15 - filter_quant));
456  }
457 
458  if ((ret = decode_residues(s, &decoded[filter_order],
459  subframe_size - filter_order)) < 0)
460  return ret;
461 
462  for (i = 0; i < filter_order; i++)
463  s->residues[i] = *decoded++ >> dshift;
464 
465  y = FF_ARRAY_ELEMS(s->residues) - filter_order;
466  x = subframe_size - filter_order;
467  while (x > 0) {
468  int tmp = FFMIN(y, x);
469 
470  for (i = 0; i < tmp; i++) {
471  int v = 1 << (filter_quant - 1);
472 
473  if (filter_order & -16)
474  v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
475  filter_order & -16);
476  for (j = filter_order & -16; j < filter_order; j += 4) {
477  v += s->residues[i + j + 3] * s->filter[j + 3] +
478  s->residues[i + j + 2] * s->filter[j + 2] +
479  s->residues[i + j + 1] * s->filter[j + 1] +
480  s->residues[i + j ] * s->filter[j ];
481  }
482  v = (av_clip_intp2(v >> filter_quant, 13) << dshift) - *decoded;
483  *decoded++ = v;
484  s->residues[filter_order + i] = v >> dshift;
485  }
486 
487  x -= tmp;
488  if (x > 0)
489  memcpy(s->residues, &s->residues[y], 2 * filter_order);
490  }
491 
492  emms_c();
493 
494  return 0;
495 }
496 
497 static int decode_channel(TAKDecContext *s, int chan)
498 {
499  AVCodecContext *avctx = s->avctx;
500  GetBitContext *gb = &s->gb;
501  int32_t *decoded = s->decoded[chan];
502  int left = s->nb_samples - 1;
503  int i = 0, ret, prev = 0;
504 
505  s->sample_shift[chan] = get_bits_esc4(gb);
506  if (s->sample_shift[chan] >= avctx->bits_per_raw_sample)
507  return AVERROR_INVALIDDATA;
508 
509  *decoded++ = get_sbits(gb, avctx->bits_per_raw_sample - s->sample_shift[chan]);
510  s->lpc_mode[chan] = get_bits(gb, 2);
511  s->nb_subframes = get_bits(gb, 3) + 1;
512 
513  if (s->nb_subframes > 1) {
514  if (get_bits_left(gb) < (s->nb_subframes - 1) * 6)
515  return AVERROR_INVALIDDATA;
516 
517  for (; i < s->nb_subframes - 1; i++) {
518  int v = get_bits(gb, 6);
519 
520  s->subframe_len[i] = (v - prev) * s->subframe_scale;
521  if (s->subframe_len[i] <= 0)
522  return AVERROR_INVALIDDATA;
523 
524  left -= s->subframe_len[i];
525  prev = v;
526  }
527 
528  if (left <= 0)
529  return AVERROR_INVALIDDATA;
530  }
531  s->subframe_len[i] = left;
532 
533  prev = 0;
534  for (i = 0; i < s->nb_subframes; i++) {
535  if ((ret = decode_subframe(s, decoded, s->subframe_len[i], prev)) < 0)
536  return ret;
537  decoded += s->subframe_len[i];
538  prev = s->subframe_len[i];
539  }
540 
541  return 0;
542 }
543 
544 static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
545 {
546  GetBitContext *gb = &s->gb;
547  int32_t *p1 = s->decoded[c1] + (s->dmode > 5);
548  int32_t *p2 = s->decoded[c2] + (s->dmode > 5);
549  int32_t bp1 = p1[0];
550  int32_t bp2 = p2[0];
551  int i;
552  int dshift, dfactor;
553 
554  length += s->dmode < 6;
555 
556  switch (s->dmode) {
557  case 1: /* left/side */
558  s->tdsp.decorrelate_ls(p1, p2, length);
559  break;
560  case 2: /* side/right */
561  s->tdsp.decorrelate_sr(p1, p2, length);
562  break;
563  case 3: /* side/mid */
564  s->tdsp.decorrelate_sm(p1, p2, length);
565  break;
566  case 4: /* side/left with scale factor */
567  FFSWAP(int32_t*, p1, p2);
568  FFSWAP(int32_t, bp1, bp2);
569  case 5: /* side/right with scale factor */
570  dshift = get_bits_esc4(gb);
571  dfactor = get_sbits(gb, 10);
572  s->tdsp.decorrelate_sf(p1, p2, length, dshift, dfactor);
573  break;
574  case 6:
575  FFSWAP(int32_t*, p1, p2);
576  case 7: {
577  int length2, order_half, filter_order, dval1, dval2;
578  int tmp, x, code_size;
579 
580  if (length < 256)
581  return AVERROR_INVALIDDATA;
582 
583  dshift = get_bits_esc4(gb);
584  filter_order = 8 << get_bits1(gb);
585  dval1 = get_bits1(gb);
586  dval2 = get_bits1(gb);
587 
588  for (i = 0; i < filter_order; i++) {
589  if (!(i & 3))
590  code_size = 14 - get_bits(gb, 3);
591  s->filter[i] = get_sbits(gb, code_size);
592  }
593 
594  order_half = filter_order / 2;
595  length2 = length - (filter_order - 1);
596 
597  /* decorrelate beginning samples */
598  if (dval1) {
599  for (i = 0; i < order_half; i++) {
600  int32_t a = p1[i];
601  int32_t b = p2[i];
602  p1[i] = a + b;
603  }
604  }
605 
606  /* decorrelate ending samples */
607  if (dval2) {
608  for (i = length2 + order_half; i < length; i++) {
609  int32_t a = p1[i];
610  int32_t b = p2[i];
611  p1[i] = a + b;
612  }
613  }
614 
615 
616  for (i = 0; i < filter_order; i++)
617  s->residues[i] = *p2++ >> dshift;
618 
619  p1 += order_half;
620  x = FF_ARRAY_ELEMS(s->residues) - filter_order;
621  for (; length2 > 0; length2 -= tmp) {
622  tmp = FFMIN(length2, x);
623 
624  for (i = 0; i < tmp - (tmp == length2); i++)
625  s->residues[filter_order + i] = *p2++ >> dshift;
626 
627  for (i = 0; i < tmp; i++) {
628  int v = 1 << 9;
629 
630  if (filter_order == 16) {
631  v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
632  filter_order);
633  } else {
634  v += s->residues[i + 7] * s->filter[7] +
635  s->residues[i + 6] * s->filter[6] +
636  s->residues[i + 5] * s->filter[5] +
637  s->residues[i + 4] * s->filter[4] +
638  s->residues[i + 3] * s->filter[3] +
639  s->residues[i + 2] * s->filter[2] +
640  s->residues[i + 1] * s->filter[1] +
641  s->residues[i ] * s->filter[0];
642  }
643 
644  v = (av_clip_intp2(v >> 10, 13) << dshift) - *p1;
645  *p1++ = v;
646  }
647 
648  memmove(s->residues, &s->residues[tmp], 2 * filter_order);
649  }
650 
651  emms_c();
652  break;
653  }
654  }
655 
656  if (s->dmode > 0 && s->dmode < 6) {
657  p1[0] = bp1;
658  p2[0] = bp2;
659  }
660 
661  return 0;
662 }
663 
664 static int tak_decode_frame(AVCodecContext *avctx, void *data,
665  int *got_frame_ptr, AVPacket *pkt)
666 {
667  TAKDecContext *s = avctx->priv_data;
668  AVFrame *frame = data;
669  ThreadFrame tframe = { .f = data };
670  GetBitContext *gb = &s->gb;
671  int chan, i, ret, hsize;
672 
673  if (pkt->size < TAK_MIN_FRAME_HEADER_BYTES)
674  return AVERROR_INVALIDDATA;
675 
676  if ((ret = init_get_bits8(gb, pkt->data, pkt->size)) < 0)
677  return ret;
678 
679  if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
680  return ret;
681 
682  hsize = get_bits_count(gb) / 8;
684  if (ff_tak_check_crc(pkt->data, hsize)) {
685  av_log(avctx, AV_LOG_ERROR, "CRC error\n");
686  if (avctx->err_recognition & AV_EF_EXPLODE)
687  return AVERROR_INVALIDDATA;
688  }
689  }
690 
691  if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
693  av_log(avctx, AV_LOG_ERROR, "unsupported codec: %d\n", s->ti.codec);
694  return AVERROR_PATCHWELCOME;
695  }
696  if (s->ti.data_type) {
697  av_log(avctx, AV_LOG_ERROR,
698  "unsupported data type: %d\n", s->ti.data_type);
699  return AVERROR_INVALIDDATA;
700  }
701  if (s->ti.codec == TAK_CODEC_MONO_STEREO && s->ti.channels > 2) {
702  av_log(avctx, AV_LOG_ERROR,
703  "invalid number of channels: %d\n", s->ti.channels);
704  return AVERROR_INVALIDDATA;
705  }
706  if (s->ti.channels > 6) {
707  av_log(avctx, AV_LOG_ERROR,
708  "unsupported number of channels: %d\n", s->ti.channels);
709  return AVERROR_INVALIDDATA;
710  }
711 
712  if (s->ti.frame_samples <= 0) {
713  av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
714  return AVERROR_INVALIDDATA;
715  }
716 
717  avctx->bits_per_raw_sample = s->ti.bps;
718  if ((ret = set_bps_params(avctx)) < 0)
719  return ret;
720  if (s->ti.sample_rate != avctx->sample_rate) {
721  avctx->sample_rate = s->ti.sample_rate;
722  set_sample_rate_params(avctx);
723  }
724  if (s->ti.ch_layout)
725  avctx->channel_layout = s->ti.ch_layout;
726  avctx->channels = s->ti.channels;
727 
729  : s->ti.frame_samples;
730 
731  frame->nb_samples = s->nb_samples;
732  if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0)
733  return ret;
734  ff_thread_finish_setup(avctx);
735 
736  if (avctx->bits_per_raw_sample <= 16) {
737  int buf_size = av_samples_get_buffer_size(NULL, avctx->channels,
738  s->nb_samples,
739  AV_SAMPLE_FMT_S32P, 0);
740  if (buf_size < 0)
741  return buf_size;
742  av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
743  if (!s->decode_buffer)
744  return AVERROR(ENOMEM);
746  s->decode_buffer, avctx->channels,
748  if (ret < 0)
749  return ret;
750  } else {
751  for (chan = 0; chan < avctx->channels; chan++)
752  s->decoded[chan] = (int32_t *)frame->extended_data[chan];
753  }
754 
755  if (s->nb_samples < 16) {
756  for (chan = 0; chan < avctx->channels; chan++) {
757  int32_t *decoded = s->decoded[chan];
758  for (i = 0; i < s->nb_samples; i++)
759  decoded[i] = get_sbits(gb, avctx->bits_per_raw_sample);
760  }
761  } else {
762  if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
763  for (chan = 0; chan < avctx->channels; chan++)
764  if (ret = decode_channel(s, chan))
765  return ret;
766 
767  if (avctx->channels == 2) {
768  s->nb_subframes = get_bits(gb, 1) + 1;
769  if (s->nb_subframes > 1) {
770  s->subframe_len[1] = get_bits(gb, 6);
771  }
772 
773  s->dmode = get_bits(gb, 3);
774  if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
775  return ret;
776  }
777  } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
778  if (get_bits1(gb)) {
779  int ch_mask = 0;
780 
781  chan = get_bits(gb, 4) + 1;
782  if (chan > avctx->channels)
783  return AVERROR_INVALIDDATA;
784 
785  for (i = 0; i < chan; i++) {
786  int nbit = get_bits(gb, 4);
787 
788  if (nbit >= avctx->channels)
789  return AVERROR_INVALIDDATA;
790 
791  if (ch_mask & 1 << nbit)
792  return AVERROR_INVALIDDATA;
793 
794  s->mcdparams[i].present = get_bits1(gb);
795  if (s->mcdparams[i].present) {
796  s->mcdparams[i].index = get_bits(gb, 2);
797  s->mcdparams[i].chan2 = get_bits(gb, 4);
798  if (s->mcdparams[i].chan2 >= avctx->channels) {
799  av_log(avctx, AV_LOG_ERROR,
800  "invalid channel 2 (%d) for %d channel(s)\n",
801  s->mcdparams[i].chan2, avctx->channels);
802  return AVERROR_INVALIDDATA;
803  }
804  if (s->mcdparams[i].index == 1) {
805  if ((nbit == s->mcdparams[i].chan2) ||
806  (ch_mask & 1 << s->mcdparams[i].chan2))
807  return AVERROR_INVALIDDATA;
808 
809  ch_mask |= 1 << s->mcdparams[i].chan2;
810  } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
811  return AVERROR_INVALIDDATA;
812  }
813  }
814  s->mcdparams[i].chan1 = nbit;
815 
816  ch_mask |= 1 << nbit;
817  }
818  } else {
819  chan = avctx->channels;
820  for (i = 0; i < chan; i++) {
821  s->mcdparams[i].present = 0;
822  s->mcdparams[i].chan1 = i;
823  }
824  }
825 
826  for (i = 0; i < chan; i++) {
827  if (s->mcdparams[i].present && s->mcdparams[i].index == 1)
828  if (ret = decode_channel(s, s->mcdparams[i].chan2))
829  return ret;
830 
831  if (ret = decode_channel(s, s->mcdparams[i].chan1))
832  return ret;
833 
834  if (s->mcdparams[i].present) {
835  s->dmode = mc_dmodes[s->mcdparams[i].index];
836  if (ret = decorrelate(s,
837  s->mcdparams[i].chan2,
838  s->mcdparams[i].chan1,
839  s->nb_samples - 1))
840  return ret;
841  }
842  }
843  }
844 
845  for (chan = 0; chan < avctx->channels; chan++) {
846  int32_t *decoded = s->decoded[chan];
847 
848  if (s->lpc_mode[chan])
849  decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
850 
851  if (s->sample_shift[chan] > 0)
852  for (i = 0; i < s->nb_samples; i++)
853  decoded[i] <<= s->sample_shift[chan];
854  }
855  }
856 
857  align_get_bits(gb);
858  skip_bits(gb, 24);
859  if (get_bits_left(gb) < 0)
860  av_log(avctx, AV_LOG_DEBUG, "overread\n");
861  else if (get_bits_left(gb) > 0)
862  av_log(avctx, AV_LOG_DEBUG, "underread\n");
863 
865  if (ff_tak_check_crc(pkt->data + hsize,
866  get_bits_count(gb) / 8 - hsize)) {
867  av_log(avctx, AV_LOG_ERROR, "CRC error\n");
868  if (avctx->err_recognition & AV_EF_EXPLODE)
869  return AVERROR_INVALIDDATA;
870  }
871  }
872 
873  /* convert to output buffer */
874  switch (avctx->sample_fmt) {
875  case AV_SAMPLE_FMT_U8P:
876  for (chan = 0; chan < avctx->channels; chan++) {
877  uint8_t *samples = (uint8_t *)frame->extended_data[chan];
878  int32_t *decoded = s->decoded[chan];
879  for (i = 0; i < s->nb_samples; i++)
880  samples[i] = decoded[i] + 0x80;
881  }
882  break;
883  case AV_SAMPLE_FMT_S16P:
884  for (chan = 0; chan < avctx->channels; chan++) {
885  int16_t *samples = (int16_t *)frame->extended_data[chan];
886  int32_t *decoded = s->decoded[chan];
887  for (i = 0; i < s->nb_samples; i++)
888  samples[i] = decoded[i];
889  }
890  break;
891  case AV_SAMPLE_FMT_S32P:
892  for (chan = 0; chan < avctx->channels; chan++) {
893  int32_t *samples = (int32_t *)frame->extended_data[chan];
894  for (i = 0; i < s->nb_samples; i++)
895  samples[i] <<= 8;
896  }
897  break;
898  }
899 
900  *got_frame_ptr = 1;
901 
902  return pkt->size;
903 }
904 
905 #if HAVE_THREADS
906 static int init_thread_copy(AVCodecContext *avctx)
907 {
908  TAKDecContext *s = avctx->priv_data;
909  s->avctx = avctx;
910  return 0;
911 }
912 
913 static int update_thread_context(AVCodecContext *dst,
914  const AVCodecContext *src)
915 {
916  TAKDecContext *tsrc = src->priv_data;
917  TAKDecContext *tdst = dst->priv_data;
918 
919  if (dst == src)
920  return 0;
921  memcpy(&tdst->ti, &tsrc->ti, sizeof(TAKStreamInfo));
922  return 0;
923 }
924 #endif
925 
927 {
928  TAKDecContext *s = avctx->priv_data;
929 
930  av_freep(&s->decode_buffer);
931 
932  return 0;
933 }
934 
936  .name = "tak",
937  .long_name = NULL_IF_CONFIG_SMALL("TAK (Tom's lossless Audio Kompressor)"),
938  .type = AVMEDIA_TYPE_AUDIO,
939  .id = AV_CODEC_ID_TAK,
940  .priv_data_size = sizeof(TAKDecContext),
942  .close = tak_decode_close,
945  .update_thread_context = ONLY_IF_THREADS_ENABLED(update_thread_context),
947  .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
951 };
#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:181
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:1468
const char * b
Definition: vf_curves.c:109
#define AV_EF_COMPLIANT
consider all spec non compliances as errors
Definition: avcodec.h:2826
static int decode_subframe(TAKDecContext *s, int32_t *decoded, int subframe_size, int prev_subframe_size)
Definition: takdec.c:373
#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:2924
static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
Definition: takdec.c:242
void(* decorrelate_ls)(int32_t *p1, int32_t *p2, int length)
Definition: takdsp.h:25
AVCodec.
Definition: avcodec.h:3392
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:2295
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:926
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:1467
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:2917
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:497
#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:3399
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:286
#define FFMAX(a, b)
Definition: common.h:94
#define AV_CODEC_CAP_FRAME_THREADS
Codec supports frame-level multithreading.
Definition: avcodec.h:919
uint64_t channel_layout
Audio channel layout.
Definition: avcodec.h:2338
#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:172
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:2811
#define FFMIN(a, b)
Definition: common.h:96
signed 32 bits, planar
Definition: samplefmt.h:69
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
int32_t
#define a2
Definition: regdef.h:48
static int get_bits_esc4(GetBitContext *gb)
Definition: takdec.c:365
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:2822
unsigned 8 bits, planar
Definition: samplefmt.h:67
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:544
static void decode_lpc(int32_t *coeffs, int mode, int length)
Definition: takdec.c:187
#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:935
Definition: takdec.c:83
#define a5
Definition: regdef.h:51
Libavcodec external API header.
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:59
#define TAK_MIN_FRAME_HEADER_BYTES
Definition: tak.h:98
int sample_rate
samples per second
Definition: avcodec.h:2287
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:1532
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:2819
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:572
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:1574
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:664
static const int16_t coeffs[]
int len
int channels
number of audio channels
Definition: avcodec.h:2288
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:68
#define FFSWAP(type, a, b)
Definition: common.h:99
uint8_t ** extended_data
pointers to the data planes/channels.
Definition: frame.h:225
This structure stores compressed data.
Definition: avcodec.h:1444
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:235
#define AV_CODEC_CAP_DR1
Codec uses get_buffer() for allocating buffers and supports custom allocators.
Definition: avcodec.h:856
for(j=16;j >0;--j)