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 "thread.h"
32 #include "avcodec.h"
33 #include "dsputil.h"
34 #include "internal.h"
35 #include "unary.h"
36 
37 #define MAX_SUBFRAMES 8 ///< max number of subframes per channel
38 #define MAX_PREDICTORS 256
39 
40 typedef struct MCDParam {
41  int8_t present; ///< decorrelation parameter availability for this channel
42  int8_t index; ///< index into array of decorrelation types
43  int8_t chan1;
44  int8_t chan2;
45 } MCDParam;
46 
47 typedef struct TAKDecContext {
48  AVCodecContext *avctx; ///< parent AVCodecContext
51  GetBitContext gb; ///< bitstream reader initialized to start at the current frame
52 
53  int uval;
54  int nb_samples; ///< number of samples in the current frame
56  unsigned int decode_buffer_size;
57  int32_t *decoded[TAK_MAX_CHANNELS]; ///< decoded samples for each channel
58 
60  int8_t sample_shift[TAK_MAX_CHANNELS]; ///< shift applied to every sample in the channel
62  int nb_subframes; ///< number of subframes in the current frame
63  int16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples
65 
66  int8_t dmode; ///< channel decorrelation type in the current frame
67 
68  MCDParam mcdparams[TAK_MAX_CHANNELS]; ///< multichannel decorrelation parameters
69 
70  int8_t coding_mode[128];
72  DECLARE_ALIGNED(16, int16_t, residues)[544];
74 
75 static const int8_t mc_dmodes[] = { 1, 3, 4, 6, };
76 
77 static const uint16_t predictor_sizes[] = {
78  4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
79 };
80 
81 static const struct CParam {
82  int init;
83  int escape;
84  int scale;
85  int aescape;
86  int bias;
87 } xcodes[50] = {
88  { 0x01, 0x0000001, 0x0000001, 0x0000003, 0x0000008 },
89  { 0x02, 0x0000003, 0x0000001, 0x0000007, 0x0000006 },
90  { 0x03, 0x0000005, 0x0000002, 0x000000E, 0x000000D },
91  { 0x03, 0x0000003, 0x0000003, 0x000000D, 0x0000018 },
92  { 0x04, 0x000000B, 0x0000004, 0x000001C, 0x0000019 },
93  { 0x04, 0x0000006, 0x0000006, 0x000001A, 0x0000030 },
94  { 0x05, 0x0000016, 0x0000008, 0x0000038, 0x0000032 },
95  { 0x05, 0x000000C, 0x000000C, 0x0000034, 0x0000060 },
96  { 0x06, 0x000002C, 0x0000010, 0x0000070, 0x0000064 },
97  { 0x06, 0x0000018, 0x0000018, 0x0000068, 0x00000C0 },
98  { 0x07, 0x0000058, 0x0000020, 0x00000E0, 0x00000C8 },
99  { 0x07, 0x0000030, 0x0000030, 0x00000D0, 0x0000180 },
100  { 0x08, 0x00000B0, 0x0000040, 0x00001C0, 0x0000190 },
101  { 0x08, 0x0000060, 0x0000060, 0x00001A0, 0x0000300 },
102  { 0x09, 0x0000160, 0x0000080, 0x0000380, 0x0000320 },
103  { 0x09, 0x00000C0, 0x00000C0, 0x0000340, 0x0000600 },
104  { 0x0A, 0x00002C0, 0x0000100, 0x0000700, 0x0000640 },
105  { 0x0A, 0x0000180, 0x0000180, 0x0000680, 0x0000C00 },
106  { 0x0B, 0x0000580, 0x0000200, 0x0000E00, 0x0000C80 },
107  { 0x0B, 0x0000300, 0x0000300, 0x0000D00, 0x0001800 },
108  { 0x0C, 0x0000B00, 0x0000400, 0x0001C00, 0x0001900 },
109  { 0x0C, 0x0000600, 0x0000600, 0x0001A00, 0x0003000 },
110  { 0x0D, 0x0001600, 0x0000800, 0x0003800, 0x0003200 },
111  { 0x0D, 0x0000C00, 0x0000C00, 0x0003400, 0x0006000 },
112  { 0x0E, 0x0002C00, 0x0001000, 0x0007000, 0x0006400 },
113  { 0x0E, 0x0001800, 0x0001800, 0x0006800, 0x000C000 },
114  { 0x0F, 0x0005800, 0x0002000, 0x000E000, 0x000C800 },
115  { 0x0F, 0x0003000, 0x0003000, 0x000D000, 0x0018000 },
116  { 0x10, 0x000B000, 0x0004000, 0x001C000, 0x0019000 },
117  { 0x10, 0x0006000, 0x0006000, 0x001A000, 0x0030000 },
118  { 0x11, 0x0016000, 0x0008000, 0x0038000, 0x0032000 },
119  { 0x11, 0x000C000, 0x000C000, 0x0034000, 0x0060000 },
120  { 0x12, 0x002C000, 0x0010000, 0x0070000, 0x0064000 },
121  { 0x12, 0x0018000, 0x0018000, 0x0068000, 0x00C0000 },
122  { 0x13, 0x0058000, 0x0020000, 0x00E0000, 0x00C8000 },
123  { 0x13, 0x0030000, 0x0030000, 0x00D0000, 0x0180000 },
124  { 0x14, 0x00B0000, 0x0040000, 0x01C0000, 0x0190000 },
125  { 0x14, 0x0060000, 0x0060000, 0x01A0000, 0x0300000 },
126  { 0x15, 0x0160000, 0x0080000, 0x0380000, 0x0320000 },
127  { 0x15, 0x00C0000, 0x00C0000, 0x0340000, 0x0600000 },
128  { 0x16, 0x02C0000, 0x0100000, 0x0700000, 0x0640000 },
129  { 0x16, 0x0180000, 0x0180000, 0x0680000, 0x0C00000 },
130  { 0x17, 0x0580000, 0x0200000, 0x0E00000, 0x0C80000 },
131  { 0x17, 0x0300000, 0x0300000, 0x0D00000, 0x1800000 },
132  { 0x18, 0x0B00000, 0x0400000, 0x1C00000, 0x1900000 },
133  { 0x18, 0x0600000, 0x0600000, 0x1A00000, 0x3000000 },
134  { 0x19, 0x1600000, 0x0800000, 0x3800000, 0x3200000 },
135  { 0x19, 0x0C00000, 0x0C00000, 0x3400000, 0x6000000 },
136  { 0x1A, 0x2C00000, 0x1000000, 0x7000000, 0x6400000 },
137  { 0x1A, 0x1800000, 0x1800000, 0x6800000, 0xC000000 },
138 };
139 
140 static int set_bps_params(AVCodecContext *avctx)
141 {
142  switch (avctx->bits_per_raw_sample) {
143  case 8:
144  avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
145  break;
146  case 16:
148  break;
149  case 24:
151  break;
152  default:
153  av_log(avctx, AV_LOG_ERROR, "invalid/unsupported bits per sample: %d\n",
154  avctx->bits_per_raw_sample);
155  return AVERROR_INVALIDDATA;
156  }
157 
158  return 0;
159 }
160 
162 {
163  TAKDecContext *s = avctx->priv_data;
164  int shift = 3 - (avctx->sample_rate / 11025);
165  shift = FFMAX(0, shift);
166  s->uval = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << shift;
167  s->subframe_scale = FFALIGN(avctx->sample_rate + 511 >> 9, 4) << 1;
168 }
169 
171 {
172  TAKDecContext *s = avctx->priv_data;
173 
174  ff_tak_init_crc();
175  ff_dsputil_init(&s->dsp, avctx);
176 
177  s->avctx = avctx;
179 
180  set_sample_rate_params(avctx);
181 
182  return set_bps_params(avctx);
183 }
184 
185 static void decode_lpc(int32_t *coeffs, int mode, int length)
186 {
187  int i;
188 
189  if (length < 2)
190  return;
191 
192  if (mode == 1) {
193  int a1 = *coeffs++;
194  for (i = 0; i < length - 1 >> 1; i++) {
195  *coeffs += a1;
196  coeffs[1] += *coeffs;
197  a1 = coeffs[1];
198  coeffs += 2;
199  }
200  if (length - 1 & 1)
201  *coeffs += a1;
202  } else if (mode == 2) {
203  int a1 = coeffs[1];
204  int a2 = a1 + *coeffs;
205  coeffs[1] = a2;
206  if (length > 2) {
207  coeffs += 2;
208  for (i = 0; i < length - 2 >> 1; i++) {
209  int a3 = *coeffs + a1;
210  int a4 = a3 + a2;
211  *coeffs = a4;
212  a1 = coeffs[1] + a3;
213  a2 = a1 + a4;
214  coeffs[1] = a2;
215  coeffs += 2;
216  }
217  if (length & 1)
218  *coeffs += a1 + a2;
219  }
220  } else if (mode == 3) {
221  int a1 = coeffs[1];
222  int a2 = a1 + *coeffs;
223  coeffs[1] = a2;
224  if (length > 2) {
225  int a3 = coeffs[2];
226  int a4 = a3 + a1;
227  int a5 = a4 + a2;
228  coeffs += 3;
229  for (i = 0; i < length - 3; i++) {
230  a3 += *coeffs;
231  a4 += a3;
232  a5 += a4;
233  *coeffs = a5;
234  coeffs++;
235  }
236  }
237  }
238 }
239 
240 static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
241 {
242  struct CParam code;
243  GetBitContext *gb = &s->gb;
244  int i;
245 
246  if (!mode) {
247  memset(decoded, 0, len * sizeof(*decoded));
248  return 0;
249  }
250 
251  if (mode > FF_ARRAY_ELEMS(xcodes))
252  return AVERROR_INVALIDDATA;
253  code = xcodes[mode - 1];
254 
255  for (i = 0; i < len; i++) {
256  int x = get_bits_long(gb, code.init);
257  if (x >= code.escape && get_bits1(gb)) {
258  x |= 1 << code.init;
259  if (x >= code.aescape) {
260  int scale = get_unary(gb, 1, 9);
261  if (scale == 9) {
262  int scale_bits = get_bits(gb, 3);
263  if (scale_bits > 0) {
264  if (scale_bits == 7) {
265  scale_bits += get_bits(gb, 5);
266  if (scale_bits > 29)
267  return AVERROR_INVALIDDATA;
268  }
269  scale = get_bits_long(gb, scale_bits) + 1;
270  x += code.scale * scale;
271  }
272  x += code.bias;
273  } else
274  x += code.scale * scale - code.escape;
275  } else
276  x -= code.escape;
277  }
278  decoded[i] = (x >> 1) ^ -(x & 1);
279  }
280 
281  return 0;
282 }
283 
284 static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
285 {
286  GetBitContext *gb = &s->gb;
287  int i, mode, ret;
288 
289  if (length > s->nb_samples)
290  return AVERROR_INVALIDDATA;
291 
292  if (get_bits1(gb)) {
293  int wlength, rval;
294 
295  wlength = length / s->uval;
296 
297  rval = length - (wlength * s->uval);
298 
299  if (rval < s->uval / 2)
300  rval += s->uval;
301  else
302  wlength++;
303 
304  if (wlength <= 1 || wlength > 128)
305  return AVERROR_INVALIDDATA;
306 
307  s->coding_mode[0] = mode = get_bits(gb, 6);
308 
309  for (i = 1; i < wlength; i++) {
310  int c = get_unary(gb, 1, 6);
311 
312  switch (c) {
313  case 6:
314  mode = get_bits(gb, 6);
315  break;
316  case 5:
317  case 4:
318  case 3: {
319  /* mode += sign ? (1 - c) : (c - 1) */
320  int sign = get_bits1(gb);
321  mode += (-sign ^ (c - 1)) + sign;
322  break;
323  }
324  case 2:
325  mode++;
326  break;
327  case 1:
328  mode--;
329  break;
330  }
331  s->coding_mode[i] = mode;
332  }
333 
334  i = 0;
335  while (i < wlength) {
336  int len = 0;
337 
338  mode = s->coding_mode[i];
339  do {
340  if (i >= wlength - 1)
341  len += rval;
342  else
343  len += s->uval;
344  i++;
345 
346  if (i == wlength)
347  break;
348  } while (s->coding_mode[i] == mode);
349 
350  if ((ret = decode_segment(s, mode, decoded, len)) < 0)
351  return ret;
352  decoded += len;
353  }
354  } else {
355  mode = get_bits(gb, 6);
356  if ((ret = decode_segment(s, mode, decoded, length)) < 0)
357  return ret;
358  }
359 
360  return 0;
361 }
362 
364 {
365  if (get_bits1(gb))
366  return get_bits(gb, 4) + 1;
367  else
368  return 0;
369 }
370 
371 static int decode_subframe(TAKDecContext *s, int32_t *decoded,
372  int subframe_size, int prev_subframe_size)
373 {
374  GetBitContext *gb = &s->gb;
375  int tmp, x, y, i, j, ret = 0;
376  int dshift, size, filter_quant, filter_order;
377  int tfilter[MAX_PREDICTORS];
378 
379  if (!get_bits1(gb))
380  return decode_residues(s, decoded, subframe_size);
381 
382  filter_order = predictor_sizes[get_bits(gb, 4)];
383 
384  if (prev_subframe_size > 0 && get_bits1(gb)) {
385  if (filter_order > prev_subframe_size)
386  return AVERROR_INVALIDDATA;
387 
388  decoded -= filter_order;
389  subframe_size += filter_order;
390 
391  if (filter_order > subframe_size)
392  return AVERROR_INVALIDDATA;
393  } else {
394  int lpc_mode;
395 
396  if (filter_order > subframe_size)
397  return AVERROR_INVALIDDATA;
398 
399  lpc_mode = get_bits(gb, 2);
400  if (lpc_mode > 2)
401  return AVERROR_INVALIDDATA;
402 
403  if ((ret = decode_residues(s, decoded, filter_order)) < 0)
404  return ret;
405 
406  if (lpc_mode)
407  decode_lpc(decoded, lpc_mode, filter_order);
408  }
409 
410  dshift = get_bits_esc4(gb);
411  size = get_bits1(gb) + 6;
412 
413  filter_quant = 10;
414  if (get_bits1(gb)) {
415  filter_quant -= get_bits(gb, 3) + 1;
416  if (filter_quant < 3)
417  return AVERROR_INVALIDDATA;
418  }
419 
420  s->predictors[0] = get_sbits(gb, 10);
421  s->predictors[1] = get_sbits(gb, 10);
422  s->predictors[2] = get_sbits(gb, size) << (10 - size);
423  s->predictors[3] = get_sbits(gb, size) << (10 - size);
424  if (filter_order > 4) {
425  tmp = size - get_bits1(gb);
426 
427  for (i = 4; i < filter_order; i++) {
428  if (!(i & 3))
429  x = tmp - get_bits(gb, 2);
430  s->predictors[i] = get_sbits(gb, x) << (10 - size);
431  }
432  }
433 
434  tfilter[0] = s->predictors[0] << 6;
435  for (i = 1; i < filter_order; i++) {
436  int32_t *p1 = &tfilter[0];
437  int32_t *p2 = &tfilter[i - 1];
438 
439  for (j = 0; j < (i + 1) / 2; j++) {
440  x = *p1 + (s->predictors[i] * *p2 + 256 >> 9);
441  *p2 += s->predictors[i] * *p1 + 256 >> 9;
442  *p1++ = x;
443  p2--;
444  }
445 
446  tfilter[i] = s->predictors[i] << 6;
447  }
448 
449  x = 1 << (32 - (15 - filter_quant));
450  y = 1 << ((15 - filter_quant) - 1);
451  for (i = 0, j = filter_order - 1; i < filter_order / 2; i++, j--) {
452  tmp = y + tfilter[j];
453  s->filter[j] = x - ((tfilter[i] + y) >> (15 - filter_quant));
454  s->filter[i] = x - ((tfilter[j] + y) >> (15 - filter_quant));
455  }
456 
457  if ((ret = decode_residues(s, &decoded[filter_order],
458  subframe_size - filter_order)) < 0)
459  return ret;
460 
461  for (i = 0; i < filter_order; i++)
462  s->residues[i] = *decoded++ >> dshift;
463 
464  y = FF_ARRAY_ELEMS(s->residues) - filter_order;
465  x = subframe_size - filter_order;
466  while (x > 0) {
467  tmp = FFMIN(y, x);
468 
469  for (i = 0; i < tmp; i++) {
470  int v = 1 << (filter_quant - 1);
471 
472  if (filter_order & -16)
473  v += s->dsp.scalarproduct_int16(&s->residues[i], s->filter,
474  filter_order & -16);
475  for (j = filter_order & -16; j < filter_order; j += 4) {
476  v += s->residues[i + j + 3] * s->filter[j + 3] +
477  s->residues[i + j + 2] * s->filter[j + 2] +
478  s->residues[i + j + 1] * s->filter[j + 1] +
479  s->residues[i + j ] * s->filter[j ];
480  }
481  v = (av_clip(v >> filter_quant, -8192, 8191) << dshift) - *decoded;
482  *decoded++ = v;
483  s->residues[filter_order + i] = v >> dshift;
484  }
485 
486  x -= tmp;
487  if (x > 0)
488  memcpy(s->residues, &s->residues[y], 2 * filter_order);
489  }
490 
491  emms_c();
492 
493  return 0;
494 }
495 
496 static int decode_channel(TAKDecContext *s, int chan)
497 {
498  AVCodecContext *avctx = s->avctx;
499  GetBitContext *gb = &s->gb;
500  int32_t *decoded = s->decoded[chan];
501  int left = s->nb_samples - 1;
502  int i = 0, ret, prev = 0;
503 
504  s->sample_shift[chan] = get_bits_esc4(gb);
505  if (s->sample_shift[chan] >= avctx->bits_per_raw_sample)
506  return AVERROR_INVALIDDATA;
507 
508  *decoded++ = get_sbits(gb, avctx->bits_per_raw_sample - s->sample_shift[chan]);
509  s->lpc_mode[chan] = get_bits(gb, 2);
510  s->nb_subframes = get_bits(gb, 3) + 1;
511 
512  if (s->nb_subframes > 1) {
513  if (get_bits_left(gb) < (s->nb_subframes - 1) * 6)
514  return AVERROR_INVALIDDATA;
515 
516  for (; i < s->nb_subframes - 1; i++) {
517  int v = get_bits(gb, 6);
518 
519  s->subframe_len[i] = (v - prev) * s->subframe_scale;
520  if (s->subframe_len[i] <= 0)
521  return AVERROR_INVALIDDATA;
522 
523  left -= s->subframe_len[i];
524  prev = v;
525  }
526 
527  if (left <= 0)
528  return AVERROR_INVALIDDATA;
529  }
530  s->subframe_len[i] = left;
531 
532  prev = 0;
533  for (i = 0; i < s->nb_subframes; i++) {
534  if ((ret = decode_subframe(s, decoded, s->subframe_len[i], prev)) < 0)
535  return ret;
536  decoded += s->subframe_len[i];
537  prev = s->subframe_len[i];
538  }
539 
540  return 0;
541 }
542 
543 static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
544 {
545  GetBitContext *gb = &s->gb;
546  int32_t *p1 = s->decoded[c1] + 1;
547  int32_t *p2 = s->decoded[c2] + 1;
548  int i;
549  int dshift, dfactor;
550 
551  switch (s->dmode) {
552  case 1: /* left/side */
553  for (i = 0; i < length; i++) {
554  int32_t a = p1[i];
555  int32_t b = p2[i];
556  p2[i] = a + b;
557  }
558  break;
559  case 2: /* side/right */
560  for (i = 0; i < length; i++) {
561  int32_t a = p1[i];
562  int32_t b = p2[i];
563  p1[i] = b - a;
564  }
565  break;
566  case 3: /* side/mid */
567  for (i = 0; i < length; i++) {
568  int32_t a = p1[i];
569  int32_t b = p2[i];
570  a -= b >> 1;
571  p1[i] = a;
572  p2[i] = a + b;
573  }
574  break;
575  case 4: /* side/left with scale factor */
576  FFSWAP(int32_t*, p1, p2);
577  case 5: /* side/right with scale factor */
578  dshift = get_bits_esc4(gb);
579  dfactor = get_sbits(gb, 10);
580  for (i = 0; i < length; i++) {
581  int32_t a = p1[i];
582  int32_t b = p2[i];
583  b = dfactor * (b >> dshift) + 128 >> 8 << dshift;
584  p1[i] = b - a;
585  }
586  break;
587  case 6:
588  FFSWAP(int32_t*, p1, p2);
589  case 7: {
590  int length2, order_half, filter_order, dval1, dval2;
591  int tmp, x, code_size;
592 
593  if (length < 256)
594  return AVERROR_INVALIDDATA;
595 
596  dshift = get_bits_esc4(gb);
597  filter_order = 8 << get_bits1(gb);
598  dval1 = get_bits1(gb);
599  dval2 = get_bits1(gb);
600 
601  for (i = 0; i < filter_order; i++) {
602  if (!(i & 3))
603  code_size = 14 - get_bits(gb, 3);
604  s->filter[i] = get_sbits(gb, code_size);
605  }
606 
607  order_half = filter_order / 2;
608  length2 = length - (filter_order - 1);
609 
610  /* decorrelate beginning samples */
611  if (dval1) {
612  for (i = 0; i < order_half; i++) {
613  int32_t a = p1[i];
614  int32_t b = p2[i];
615  p1[i] = a + b;
616  }
617  }
618 
619  /* decorrelate ending samples */
620  if (dval2) {
621  for (i = length2 + order_half; i < length; i++) {
622  int32_t a = p1[i];
623  int32_t b = p2[i];
624  p1[i] = a + b;
625  }
626  }
627 
628 
629  for (i = 0; i < filter_order; i++)
630  s->residues[i] = *p2++ >> dshift;
631 
632  p1 += order_half;
633  x = FF_ARRAY_ELEMS(s->residues) - filter_order;
634  for (; length2 > 0; length2 -= tmp) {
635  tmp = FFMIN(length2, x);
636 
637  for (i = 0; i < tmp; i++)
638  s->residues[filter_order + i] = *p2++ >> dshift;
639 
640  for (i = 0; i < tmp; i++) {
641  int v = 1 << 9;
642 
643  if (filter_order == 16) {
644  v += s->dsp.scalarproduct_int16(&s->residues[i], s->filter,
645  filter_order);
646  } else {
647  v += s->residues[i + 7] * s->filter[7] +
648  s->residues[i + 6] * s->filter[6] +
649  s->residues[i + 5] * s->filter[5] +
650  s->residues[i + 4] * s->filter[4] +
651  s->residues[i + 3] * s->filter[3] +
652  s->residues[i + 2] * s->filter[2] +
653  s->residues[i + 1] * s->filter[1] +
654  s->residues[i ] * s->filter[0];
655  }
656 
657  v = (av_clip(v >> 10, -8192, 8191) << dshift) - *p1;
658  *p1++ = v;
659  }
660 
661  memcpy(s->residues, &s->residues[tmp], 2 * filter_order);
662  }
663 
664  emms_c();
665  break;
666  }
667  }
668 
669  return 0;
670 }
671 
672 static int tak_decode_frame(AVCodecContext *avctx, void *data,
673  int *got_frame_ptr, AVPacket *pkt)
674 {
675  TAKDecContext *s = avctx->priv_data;
676  AVFrame *frame = data;
677  ThreadFrame tframe = { .f = data };
678  GetBitContext *gb = &s->gb;
679  int chan, i, ret, hsize;
680 
681  if (pkt->size < TAK_MIN_FRAME_HEADER_BYTES)
682  return AVERROR_INVALIDDATA;
683 
684  if ((ret = init_get_bits8(gb, pkt->data, pkt->size)) < 0)
685  return ret;
686 
687  if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
688  return ret;
689 
690  if (avctx->err_recognition & AV_EF_CRCCHECK) {
691  hsize = get_bits_count(gb) / 8;
692  if (ff_tak_check_crc(pkt->data, hsize)) {
693  av_log(avctx, AV_LOG_ERROR, "CRC error\n");
694  return AVERROR_INVALIDDATA;
695  }
696  }
697 
698  if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
700  av_log(avctx, AV_LOG_ERROR, "unsupported codec: %d\n", s->ti.codec);
701  return AVERROR_PATCHWELCOME;
702  }
703  if (s->ti.data_type) {
704  av_log(avctx, AV_LOG_ERROR,
705  "unsupported data type: %d\n", s->ti.data_type);
706  return AVERROR_INVALIDDATA;
707  }
708  if (s->ti.codec == TAK_CODEC_MONO_STEREO && s->ti.channels > 2) {
709  av_log(avctx, AV_LOG_ERROR,
710  "invalid number of channels: %d\n", s->ti.channels);
711  return AVERROR_INVALIDDATA;
712  }
713  if (s->ti.channels > 6) {
714  av_log(avctx, AV_LOG_ERROR,
715  "unsupported number of channels: %d\n", s->ti.channels);
716  return AVERROR_INVALIDDATA;
717  }
718 
719  if (s->ti.frame_samples <= 0) {
720  av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
721  return AVERROR_INVALIDDATA;
722  }
723 
724  if (s->ti.bps != avctx->bits_per_raw_sample) {
725  avctx->bits_per_raw_sample = s->ti.bps;
726  if ((ret = set_bps_params(avctx)) < 0)
727  return ret;
728  }
729  if (s->ti.sample_rate != avctx->sample_rate) {
730  avctx->sample_rate = s->ti.sample_rate;
731  set_sample_rate_params(avctx);
732  }
733  if (s->ti.ch_layout)
734  avctx->channel_layout = s->ti.ch_layout;
735  avctx->channels = s->ti.channels;
736 
738  : s->ti.frame_samples;
739 
740  frame->nb_samples = s->nb_samples;
741  if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0)
742  return ret;
743  ff_thread_finish_setup(avctx);
744 
745  if (avctx->bits_per_raw_sample <= 16) {
746  int buf_size = av_samples_get_buffer_size(NULL, avctx->channels,
747  s->nb_samples,
748  AV_SAMPLE_FMT_S32P, 0);
749  av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
750  if (!s->decode_buffer)
751  return AVERROR(ENOMEM);
752  ret = av_samples_fill_arrays((uint8_t **)s->decoded, NULL,
753  s->decode_buffer, avctx->channels,
755  if (ret < 0)
756  return ret;
757  } else {
758  for (chan = 0; chan < avctx->channels; chan++)
759  s->decoded[chan] = (int32_t *)frame->extended_data[chan];
760  }
761 
762  if (s->nb_samples < 16) {
763  for (chan = 0; chan < avctx->channels; chan++) {
764  int32_t *decoded = s->decoded[chan];
765  for (i = 0; i < s->nb_samples; i++)
766  decoded[i] = get_sbits(gb, avctx->bits_per_raw_sample);
767  }
768  } else {
769  if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
770  for (chan = 0; chan < avctx->channels; chan++)
771  if (ret = decode_channel(s, chan))
772  return ret;
773 
774  if (avctx->channels == 2) {
775  s->nb_subframes = get_bits(gb, 1) + 1;
776  if (s->nb_subframes > 1) {
777  s->subframe_len[1] = get_bits(gb, 6);
778  }
779 
780  s->dmode = get_bits(gb, 3);
781  if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
782  return ret;
783  }
784  } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
785  if (get_bits1(gb)) {
786  int ch_mask = 0;
787 
788  chan = get_bits(gb, 4) + 1;
789  if (chan > avctx->channels)
790  return AVERROR_INVALIDDATA;
791 
792  for (i = 0; i < chan; i++) {
793  int nbit = get_bits(gb, 4);
794 
795  if (nbit >= avctx->channels)
796  return AVERROR_INVALIDDATA;
797 
798  if (ch_mask & 1 << nbit)
799  return AVERROR_INVALIDDATA;
800 
801  s->mcdparams[i].present = get_bits1(gb);
802  if (s->mcdparams[i].present) {
803  s->mcdparams[i].index = get_bits(gb, 2);
804  s->mcdparams[i].chan2 = get_bits(gb, 4);
805  if (s->mcdparams[i].index == 1) {
806  if ((nbit == s->mcdparams[i].chan2) ||
807  (ch_mask & 1 << s->mcdparams[i].chan2))
808  return AVERROR_INVALIDDATA;
809 
810  ch_mask |= 1 << s->mcdparams[i].chan2;
811  } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
812  return AVERROR_INVALIDDATA;
813  }
814  }
815  s->mcdparams[i].chan1 = nbit;
816 
817  ch_mask |= 1 << nbit;
818  }
819  } else {
820  chan = avctx->channels;
821  for (i = 0; i < chan; i++) {
822  s->mcdparams[i].present = 0;
823  s->mcdparams[i].chan1 = i;
824  }
825  }
826 
827  for (i = 0; i < chan; i++) {
828  if (s->mcdparams[i].present && s->mcdparams[i].index == 1)
829  if (ret = decode_channel(s, s->mcdparams[i].chan2))
830  return ret;
831 
832  if (ret = decode_channel(s, s->mcdparams[i].chan1))
833  return ret;
834 
835  if (s->mcdparams[i].present) {
836  s->dmode = mc_dmodes[s->mcdparams[i].index];
837  if (ret = decorrelate(s,
838  s->mcdparams[i].chan2,
839  s->mcdparams[i].chan1,
840  s->nb_samples - 1))
841  return ret;
842  }
843  }
844  }
845 
846  for (chan = 0; chan < avctx->channels; chan++) {
847  int32_t *decoded = s->decoded[chan];
848 
849  if (s->lpc_mode[chan])
850  decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
851 
852  if (s->sample_shift[chan] > 0)
853  for (i = 0; i < s->nb_samples; i++)
854  decoded[i] <<= s->sample_shift[chan];
855  }
856  }
857 
858  align_get_bits(gb);
859  skip_bits(gb, 24);
860  if (get_bits_left(gb) < 0)
861  av_log(avctx, AV_LOG_DEBUG, "overread\n");
862  else if (get_bits_left(gb) > 0)
863  av_log(avctx, AV_LOG_DEBUG, "underread\n");
864 
865  if (avctx->err_recognition & AV_EF_CRCCHECK) {
866  if (ff_tak_check_crc(pkt->data + hsize,
867  get_bits_count(gb) / 8 - hsize)) {
868  av_log(avctx, AV_LOG_ERROR, "CRC error\n");
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 
906 {
907  TAKDecContext *s = avctx->priv_data;
908  s->avctx = avctx;
909  return 0;
910 }
911 
913  const AVCodecContext *src)
914 {
915  TAKDecContext *tsrc = src->priv_data;
916  TAKDecContext *tdst = dst->priv_data;
917 
918  if (dst == src)
919  return 0;
920  memcpy(&tdst->ti, &tsrc->ti, sizeof(TAKStreamInfo));
921  return 0;
922 }
923 
925 {
926  TAKDecContext *s = avctx->priv_data;
927 
928  av_freep(&s->decode_buffer);
929 
930  return 0;
931 }
932 
934  .name = "tak",
935  .type = AVMEDIA_TYPE_AUDIO,
936  .id = AV_CODEC_ID_TAK,
937  .priv_data_size = sizeof(TAKDecContext),
942  .update_thread_context = ONLY_IF_THREADS_ENABLED(update_thread_context),
943  .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
944  .long_name = NULL_IF_CONFIG_SMALL("TAK (Tom's lossless Audio Kompressor)"),
945  .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
949 };