avformat/mpeg: demux ivtv captions
[ffmpeg.git] / libavcodec / takdec.c
blobcfa69f4217b429a2797f57ca8a1cfff9ec0e0e90
1 /*
2 * TAK decoder
3 * Copyright (c) 2012 Paul B Mahol
5 * This file is part of FFmpeg.
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.
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.
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
22 /**
23 * @file
24 * TAK (Tom's lossless Audio Kompressor) decoder
25 * @author Paul B Mahol
28 #include "libavutil/internal.h"
29 #include "libavutil/mem.h"
30 #include "libavutil/mem_internal.h"
31 #include "libavutil/samplefmt.h"
33 #define CACHED_BITSTREAM_READER !ARCH_X86_32
34 #define BITSTREAM_READER_LE
35 #include "audiodsp.h"
36 #include "thread.h"
37 #include "avcodec.h"
38 #include "codec_internal.h"
39 #include "unary.h"
40 #include "tak.h"
41 #include "takdsp.h"
43 #define MAX_SUBFRAMES 8 ///< max number of subframes per channel
44 #define MAX_PREDICTORS 256
46 typedef struct MCDParam {
47 int8_t present; ///< decorrelation parameter availability for this channel
48 int8_t index; ///< index into array of decorrelation types
49 int8_t chan1;
50 int8_t chan2;
51 } MCDParam;
53 typedef struct TAKDecContext {
54 AVCodecContext *avctx; ///< parent AVCodecContext
55 AudioDSPContext adsp;
56 TAKDSPContext tdsp;
57 TAKStreamInfo ti;
58 GetBitContext gb; ///< bitstream reader initialized to start at the current frame
60 int uval;
61 int nb_samples; ///< number of samples in the current frame
62 uint8_t *decode_buffer;
63 unsigned int decode_buffer_size;
64 int32_t *decoded[TAK_MAX_CHANNELS]; ///< decoded samples for each channel
66 int8_t lpc_mode[TAK_MAX_CHANNELS];
67 int8_t sample_shift[TAK_MAX_CHANNELS]; ///< shift applied to every sample in the channel
68 int16_t predictors[MAX_PREDICTORS];
69 int nb_subframes; ///< number of subframes in the current frame
70 int16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples
71 int subframe_scale;
73 int8_t dmode; ///< channel decorrelation type in the current frame
75 MCDParam mcdparams[TAK_MAX_CHANNELS]; ///< multichannel decorrelation parameters
77 int8_t coding_mode[128];
78 DECLARE_ALIGNED(16, int16_t, filter)[MAX_PREDICTORS];
79 DECLARE_ALIGNED(16, int16_t, residues)[544];
80 } TAKDecContext;
82 static const int8_t mc_dmodes[] = { 1, 3, 4, 6, };
84 static const uint16_t predictor_sizes[] = {
85 4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
88 static const struct CParam {
89 int init;
90 int escape;
91 int scale;
92 int aescape;
93 int bias;
94 } xcodes[50] = {
95 { 0x01, 0x0000001, 0x0000001, 0x0000003, 0x0000008 },
96 { 0x02, 0x0000003, 0x0000001, 0x0000007, 0x0000006 },
97 { 0x03, 0x0000005, 0x0000002, 0x000000E, 0x000000D },
98 { 0x03, 0x0000003, 0x0000003, 0x000000D, 0x0000018 },
99 { 0x04, 0x000000B, 0x0000004, 0x000001C, 0x0000019 },
100 { 0x04, 0x0000006, 0x0000006, 0x000001A, 0x0000030 },
101 { 0x05, 0x0000016, 0x0000008, 0x0000038, 0x0000032 },
102 { 0x05, 0x000000C, 0x000000C, 0x0000034, 0x0000060 },
103 { 0x06, 0x000002C, 0x0000010, 0x0000070, 0x0000064 },
104 { 0x06, 0x0000018, 0x0000018, 0x0000068, 0x00000C0 },
105 { 0x07, 0x0000058, 0x0000020, 0x00000E0, 0x00000C8 },
106 { 0x07, 0x0000030, 0x0000030, 0x00000D0, 0x0000180 },
107 { 0x08, 0x00000B0, 0x0000040, 0x00001C0, 0x0000190 },
108 { 0x08, 0x0000060, 0x0000060, 0x00001A0, 0x0000300 },
109 { 0x09, 0x0000160, 0x0000080, 0x0000380, 0x0000320 },
110 { 0x09, 0x00000C0, 0x00000C0, 0x0000340, 0x0000600 },
111 { 0x0A, 0x00002C0, 0x0000100, 0x0000700, 0x0000640 },
112 { 0x0A, 0x0000180, 0x0000180, 0x0000680, 0x0000C00 },
113 { 0x0B, 0x0000580, 0x0000200, 0x0000E00, 0x0000C80 },
114 { 0x0B, 0x0000300, 0x0000300, 0x0000D00, 0x0001800 },
115 { 0x0C, 0x0000B00, 0x0000400, 0x0001C00, 0x0001900 },
116 { 0x0C, 0x0000600, 0x0000600, 0x0001A00, 0x0003000 },
117 { 0x0D, 0x0001600, 0x0000800, 0x0003800, 0x0003200 },
118 { 0x0D, 0x0000C00, 0x0000C00, 0x0003400, 0x0006000 },
119 { 0x0E, 0x0002C00, 0x0001000, 0x0007000, 0x0006400 },
120 { 0x0E, 0x0001800, 0x0001800, 0x0006800, 0x000C000 },
121 { 0x0F, 0x0005800, 0x0002000, 0x000E000, 0x000C800 },
122 { 0x0F, 0x0003000, 0x0003000, 0x000D000, 0x0018000 },
123 { 0x10, 0x000B000, 0x0004000, 0x001C000, 0x0019000 },
124 { 0x10, 0x0006000, 0x0006000, 0x001A000, 0x0030000 },
125 { 0x11, 0x0016000, 0x0008000, 0x0038000, 0x0032000 },
126 { 0x11, 0x000C000, 0x000C000, 0x0034000, 0x0060000 },
127 { 0x12, 0x002C000, 0x0010000, 0x0070000, 0x0064000 },
128 { 0x12, 0x0018000, 0x0018000, 0x0068000, 0x00C0000 },
129 { 0x13, 0x0058000, 0x0020000, 0x00E0000, 0x00C8000 },
130 { 0x13, 0x0030000, 0x0030000, 0x00D0000, 0x0180000 },
131 { 0x14, 0x00B0000, 0x0040000, 0x01C0000, 0x0190000 },
132 { 0x14, 0x0060000, 0x0060000, 0x01A0000, 0x0300000 },
133 { 0x15, 0x0160000, 0x0080000, 0x0380000, 0x0320000 },
134 { 0x15, 0x00C0000, 0x00C0000, 0x0340000, 0x0600000 },
135 { 0x16, 0x02C0000, 0x0100000, 0x0700000, 0x0640000 },
136 { 0x16, 0x0180000, 0x0180000, 0x0680000, 0x0C00000 },
137 { 0x17, 0x0580000, 0x0200000, 0x0E00000, 0x0C80000 },
138 { 0x17, 0x0300000, 0x0300000, 0x0D00000, 0x1800000 },
139 { 0x18, 0x0B00000, 0x0400000, 0x1C00000, 0x1900000 },
140 { 0x18, 0x0600000, 0x0600000, 0x1A00000, 0x3000000 },
141 { 0x19, 0x1600000, 0x0800000, 0x3800000, 0x3200000 },
142 { 0x19, 0x0C00000, 0x0C00000, 0x3400000, 0x6000000 },
143 { 0x1A, 0x2C00000, 0x1000000, 0x7000000, 0x6400000 },
144 { 0x1A, 0x1800000, 0x1800000, 0x6800000, 0xC000000 },
147 static int set_bps_params(AVCodecContext *avctx)
149 switch (avctx->bits_per_raw_sample) {
150 case 8:
151 avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
152 break;
153 case 16:
154 avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
155 break;
156 case 24:
157 avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
158 break;
159 default:
160 av_log(avctx, AV_LOG_ERROR, "invalid/unsupported bits per sample: %d\n",
161 avctx->bits_per_raw_sample);
162 return AVERROR_INVALIDDATA;
165 return 0;
168 static void set_sample_rate_params(AVCodecContext *avctx)
170 TAKDecContext *s = avctx->priv_data;
171 int shift;
173 if (avctx->sample_rate < 11025) {
174 shift = 3;
175 } else if (avctx->sample_rate < 22050) {
176 shift = 2;
177 } else if (avctx->sample_rate < 44100) {
178 shift = 1;
179 } else {
180 shift = 0;
182 s->uval = FFALIGN(avctx->sample_rate + 511LL >> 9, 4) << shift;
183 s->subframe_scale = FFALIGN(avctx->sample_rate + 511LL >> 9, 4) << 1;
186 static av_cold int tak_decode_init(AVCodecContext *avctx)
188 TAKDecContext *s = avctx->priv_data;
190 ff_audiodsp_init(&s->adsp);
191 ff_takdsp_init(&s->tdsp);
193 s->avctx = avctx;
194 avctx->bits_per_raw_sample = avctx->bits_per_coded_sample;
196 set_sample_rate_params(avctx);
198 return set_bps_params(avctx);
201 static void decode_lpc(int32_t *coeffs, int mode, int length)
203 int i;
205 if (length < 2)
206 return;
208 if (mode == 1) {
209 unsigned a1 = *coeffs++;
210 for (i = 0; i < length - 1 >> 1; i++) {
211 *coeffs += a1;
212 coeffs[1] += (unsigned)*coeffs;
213 a1 = coeffs[1];
214 coeffs += 2;
216 if (length - 1 & 1)
217 *coeffs += a1;
218 } else if (mode == 2) {
219 unsigned a1 = coeffs[1];
220 unsigned a2 = a1 + *coeffs;
221 coeffs[1] = a2;
222 if (length > 2) {
223 coeffs += 2;
224 for (i = 0; i < length - 2 >> 1; i++) {
225 unsigned a3 = *coeffs + a1;
226 unsigned a4 = a3 + a2;
227 *coeffs = a4;
228 a1 = coeffs[1] + a3;
229 a2 = a1 + a4;
230 coeffs[1] = a2;
231 coeffs += 2;
233 if (length & 1)
234 *coeffs += a1 + a2;
236 } else if (mode == 3) {
237 unsigned a1 = coeffs[1];
238 unsigned a2 = a1 + *coeffs;
239 coeffs[1] = a2;
240 if (length > 2) {
241 unsigned a3 = coeffs[2];
242 unsigned a4 = a3 + a1;
243 unsigned a5 = a4 + a2;
244 coeffs[2] = a5;
245 coeffs += 3;
246 for (i = 0; i < length - 3; i++) {
247 a3 += *coeffs;
248 a4 += a3;
249 a5 += a4;
250 *coeffs = a5;
251 coeffs++;
257 static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
259 struct CParam code;
260 GetBitContext *gb = &s->gb;
261 int i;
263 if (!mode) {
264 memset(decoded, 0, len * sizeof(*decoded));
265 return 0;
268 if (mode > FF_ARRAY_ELEMS(xcodes))
269 return AVERROR_INVALIDDATA;
270 code = xcodes[mode - 1];
272 for (i = 0; i < len; i++) {
273 unsigned x = get_bits_long(gb, code.init);
274 if (x >= code.escape && get_bits1(gb)) {
275 x |= 1 << code.init;
276 if (x >= code.aescape) {
277 unsigned scale = get_unary(gb, 1, 9);
278 if (scale == 9) {
279 int scale_bits = get_bits(gb, 3);
280 if (scale_bits > 0) {
281 if (scale_bits == 7) {
282 scale_bits += get_bits(gb, 5);
283 if (scale_bits > 29)
284 return AVERROR_INVALIDDATA;
286 scale = get_bits_long(gb, scale_bits) + 1;
287 x += code.scale * scale;
289 x += code.bias;
290 } else
291 x += code.scale * scale - code.escape;
292 } else
293 x -= code.escape;
295 decoded[i] = (x >> 1) ^ -(x & 1);
298 return 0;
301 static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
303 GetBitContext *gb = &s->gb;
304 int i, mode, ret;
306 if (length > s->nb_samples)
307 return AVERROR_INVALIDDATA;
309 if (get_bits1(gb)) {
310 int wlength, rval;
312 wlength = length / s->uval;
314 rval = length - (wlength * s->uval);
316 if (rval < s->uval / 2)
317 rval += s->uval;
318 else
319 wlength++;
321 if (wlength <= 1 || wlength > 128)
322 return AVERROR_INVALIDDATA;
324 s->coding_mode[0] = mode = get_bits(gb, 6);
326 for (i = 1; i < wlength; i++) {
327 int c = get_unary(gb, 1, 6);
329 switch (c) {
330 case 6:
331 mode = get_bits(gb, 6);
332 break;
333 case 5:
334 case 4:
335 case 3: {
336 /* mode += sign ? (1 - c) : (c - 1) */
337 int sign = get_bits1(gb);
338 mode += (-sign ^ (c - 1)) + sign;
339 break;
341 case 2:
342 mode++;
343 break;
344 case 1:
345 mode--;
346 break;
348 s->coding_mode[i] = mode;
351 i = 0;
352 while (i < wlength) {
353 int len = 0;
355 mode = s->coding_mode[i];
356 do {
357 if (i >= wlength - 1)
358 len += rval;
359 else
360 len += s->uval;
361 i++;
363 if (i == wlength)
364 break;
365 } while (s->coding_mode[i] == mode);
367 if ((ret = decode_segment(s, mode, decoded, len)) < 0)
368 return ret;
369 decoded += len;
371 } else {
372 mode = get_bits(gb, 6);
373 if ((ret = decode_segment(s, mode, decoded, length)) < 0)
374 return ret;
377 return 0;
380 static int get_bits_esc4(GetBitContext *gb)
382 if (get_bits1(gb))
383 return get_bits(gb, 4) + 1;
384 else
385 return 0;
388 static int decode_subframe(TAKDecContext *s, int32_t *decoded,
389 int subframe_size, int prev_subframe_size)
391 GetBitContext *gb = &s->gb;
392 int x, y, i, j, ret = 0;
393 int dshift, size, filter_quant, filter_order;
394 int tfilter[MAX_PREDICTORS];
396 if (!get_bits1(gb))
397 return decode_residues(s, decoded, subframe_size);
399 filter_order = predictor_sizes[get_bits(gb, 4)];
401 if (prev_subframe_size > 0 && get_bits1(gb)) {
402 if (filter_order > prev_subframe_size)
403 return AVERROR_INVALIDDATA;
405 decoded -= filter_order;
406 subframe_size += filter_order;
408 if (filter_order > subframe_size)
409 return AVERROR_INVALIDDATA;
410 } else {
411 int lpc_mode;
413 if (filter_order > subframe_size)
414 return AVERROR_INVALIDDATA;
416 lpc_mode = get_bits(gb, 2);
417 if (lpc_mode > 2)
418 return AVERROR_INVALIDDATA;
420 if ((ret = decode_residues(s, decoded, filter_order)) < 0)
421 return ret;
423 if (lpc_mode)
424 decode_lpc(decoded, lpc_mode, filter_order);
427 dshift = get_bits_esc4(gb);
428 size = get_bits1(gb) + 6;
430 filter_quant = 10;
431 if (get_bits1(gb)) {
432 filter_quant -= get_bits(gb, 3) + 1;
433 if (filter_quant < 3)
434 return AVERROR_INVALIDDATA;
437 s->predictors[0] = get_sbits(gb, 10);
438 s->predictors[1] = get_sbits(gb, 10);
439 s->predictors[2] = get_sbits(gb, size) * (1 << (10 - size));
440 s->predictors[3] = get_sbits(gb, size) * (1 << (10 - size));
441 if (filter_order > 4) {
442 int tmp = size - get_bits1(gb);
444 for (i = 4; i < filter_order; i++) {
445 if (!(i & 3))
446 x = tmp - get_bits(gb, 2);
447 s->predictors[i] = get_sbits(gb, x) * (1 << (10 - size));
451 tfilter[0] = s->predictors[0] * 64;
452 for (i = 1; i < filter_order; i++) {
453 uint32_t *p1 = &tfilter[0];
454 uint32_t *p2 = &tfilter[i - 1];
456 for (j = 0; j < (i + 1) / 2; j++) {
457 x = *p1 + ((int32_t)(s->predictors[i] * *p2 + 256) >> 9);
458 *p2 += (int32_t)(s->predictors[i] * *p1 + 256) >> 9;
459 *p1++ = x;
460 p2--;
463 tfilter[i] = s->predictors[i] * 64;
466 x = 1 << (32 - (15 - filter_quant));
467 y = 1 << ((15 - filter_quant) - 1);
468 for (i = 0, j = filter_order - 1; i < filter_order / 2; i++, j--) {
469 s->filter[j] = x - ((tfilter[i] + y) >> (15 - filter_quant));
470 s->filter[i] = x - ((tfilter[j] + y) >> (15 - filter_quant));
473 if ((ret = decode_residues(s, &decoded[filter_order],
474 subframe_size - filter_order)) < 0)
475 return ret;
477 for (i = 0; i < filter_order; i++)
478 s->residues[i] = *decoded++ >> dshift;
480 y = FF_ARRAY_ELEMS(s->residues) - filter_order;
481 x = subframe_size - filter_order;
482 while (x > 0) {
483 int tmp = FFMIN(y, x);
485 for (i = 0; i < tmp; i++) {
486 int v = 1 << (filter_quant - 1);
488 if (filter_order & -16)
489 v += (unsigned)s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
490 filter_order & -16);
491 for (j = filter_order & -16; j < filter_order; j += 4) {
492 v += s->residues[i + j + 3] * (unsigned)s->filter[j + 3] +
493 s->residues[i + j + 2] * (unsigned)s->filter[j + 2] +
494 s->residues[i + j + 1] * (unsigned)s->filter[j + 1] +
495 s->residues[i + j ] * (unsigned)s->filter[j ];
497 v = (av_clip_intp2(v >> filter_quant, 13) * (1 << dshift)) - (unsigned)*decoded;
498 *decoded++ = v;
499 s->residues[filter_order + i] = v >> dshift;
502 x -= tmp;
503 if (x > 0)
504 memcpy(s->residues, &s->residues[y], 2 * filter_order);
507 return 0;
510 static int decode_channel(TAKDecContext *s, int chan)
512 AVCodecContext *avctx = s->avctx;
513 GetBitContext *gb = &s->gb;
514 int32_t *decoded = s->decoded[chan];
515 int left = s->nb_samples - 1;
516 int i = 0, ret, prev = 0;
518 s->sample_shift[chan] = get_bits_esc4(gb);
519 if (s->sample_shift[chan] >= avctx->bits_per_raw_sample)
520 return AVERROR_INVALIDDATA;
522 *decoded++ = get_sbits(gb, avctx->bits_per_raw_sample - s->sample_shift[chan]);
523 s->lpc_mode[chan] = get_bits(gb, 2);
524 s->nb_subframes = get_bits(gb, 3) + 1;
526 if (s->nb_subframes > 1) {
527 if (get_bits_left(gb) < (s->nb_subframes - 1) * 6)
528 return AVERROR_INVALIDDATA;
530 for (; i < s->nb_subframes - 1; i++) {
531 int v = get_bits(gb, 6);
533 s->subframe_len[i] = (v - prev) * s->subframe_scale;
534 if (s->subframe_len[i] <= 0)
535 return AVERROR_INVALIDDATA;
537 left -= s->subframe_len[i];
538 prev = v;
541 if (left <= 0)
542 return AVERROR_INVALIDDATA;
544 s->subframe_len[i] = left;
546 prev = 0;
547 for (i = 0; i < s->nb_subframes; i++) {
548 if ((ret = decode_subframe(s, decoded, s->subframe_len[i], prev)) < 0)
549 return ret;
550 decoded += s->subframe_len[i];
551 prev = s->subframe_len[i];
554 return 0;
557 static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
559 GetBitContext *gb = &s->gb;
560 int32_t *p1 = s->decoded[c1] + (s->dmode > 5);
561 int32_t *p2 = s->decoded[c2] + (s->dmode > 5);
562 int32_t bp1 = p1[0];
563 int32_t bp2 = p2[0];
564 int i;
565 int dshift, dfactor;
567 length += s->dmode < 6;
569 switch (s->dmode) {
570 case 1: /* left/side */
571 s->tdsp.decorrelate_ls(p1, p2, length);
572 break;
573 case 2: /* side/right */
574 s->tdsp.decorrelate_sr(p1, p2, length);
575 break;
576 case 3: /* side/mid */
577 s->tdsp.decorrelate_sm(p1, p2, length);
578 break;
579 case 4: /* side/left with scale factor */
580 FFSWAP(int32_t*, p1, p2);
581 FFSWAP(int32_t, bp1, bp2);
582 case 5: /* side/right with scale factor */
583 dshift = get_bits_esc4(gb);
584 dfactor = get_sbits(gb, 10);
585 s->tdsp.decorrelate_sf(p1, p2, length, dshift, dfactor);
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;
593 if (length < 256)
594 return AVERROR_INVALIDDATA;
596 dshift = get_bits_esc4(gb);
597 filter_order = 8 << get_bits1(gb);
598 dval1 = get_bits1(gb);
599 dval2 = get_bits1(gb);
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);
607 order_half = filter_order / 2;
608 length2 = length - (filter_order - 1);
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;
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;
629 for (i = 0; i < filter_order; i++)
630 s->residues[i] = *p2++ >> dshift;
632 p1 += order_half;
633 x = FF_ARRAY_ELEMS(s->residues) - filter_order;
634 for (; length2 > 0; length2 -= tmp) {
635 tmp = FFMIN(length2, x);
637 for (i = 0; i < tmp - (tmp == length2); i++)
638 s->residues[filter_order + i] = *p2++ >> dshift;
640 for (i = 0; i < tmp; i++) {
641 int v = 1 << 9;
643 if (filter_order == 16) {
644 v += s->adsp.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];
657 v = av_clip_intp2(v >> 10, 13) * (1U << dshift) - *p1;
658 *p1++ = v;
661 memmove(s->residues, &s->residues[tmp], 2 * filter_order);
663 break;
667 if (s->dmode > 0 && s->dmode < 6) {
668 p1[0] = bp1;
669 p2[0] = bp2;
672 return 0;
675 static int tak_decode_frame(AVCodecContext *avctx, AVFrame *frame,
676 int *got_frame_ptr, AVPacket *pkt)
678 TAKDecContext *s = avctx->priv_data;
679 GetBitContext *gb = &s->gb;
680 int chan, i, ret, hsize;
682 if (pkt->size < TAK_MIN_FRAME_HEADER_BYTES)
683 return AVERROR_INVALIDDATA;
685 if ((ret = init_get_bits8(gb, pkt->data, pkt->size)) < 0)
686 return ret;
688 if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
689 return ret;
691 hsize = get_bits_count(gb) / 8;
692 if (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_COMPLIANT)) {
693 if (ff_tak_check_crc(pkt->data, hsize)) {
694 av_log(avctx, AV_LOG_ERROR, "CRC error\n");
695 if (avctx->err_recognition & AV_EF_EXPLODE)
696 return AVERROR_INVALIDDATA;
700 if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
701 s->ti.codec != TAK_CODEC_MULTICHANNEL) {
702 avpriv_report_missing_feature(avctx, "TAK codec type %d", s->ti.codec);
703 return AVERROR_PATCHWELCOME;
705 if (s->ti.data_type) {
706 av_log(avctx, AV_LOG_ERROR,
707 "unsupported data type: %d\n", s->ti.data_type);
708 return AVERROR_INVALIDDATA;
710 if (s->ti.codec == TAK_CODEC_MONO_STEREO && s->ti.channels > 2) {
711 av_log(avctx, AV_LOG_ERROR,
712 "invalid number of channels: %d\n", s->ti.channels);
713 return AVERROR_INVALIDDATA;
715 if (s->ti.channels > 6) {
716 av_log(avctx, AV_LOG_ERROR,
717 "unsupported number of channels: %d\n", s->ti.channels);
718 return AVERROR_INVALIDDATA;
721 if (s->ti.frame_samples <= 0) {
722 av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
723 return AVERROR_INVALIDDATA;
726 avctx->bits_per_raw_sample = s->ti.bps;
727 if ((ret = set_bps_params(avctx)) < 0)
728 return ret;
729 if (s->ti.sample_rate != avctx->sample_rate) {
730 avctx->sample_rate = s->ti.sample_rate;
731 set_sample_rate_params(avctx);
734 av_channel_layout_uninit(&avctx->ch_layout);
735 if (s->ti.ch_layout) {
736 av_channel_layout_from_mask(&avctx->ch_layout, s->ti.ch_layout);
737 } else {
738 avctx->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC;
739 avctx->ch_layout.nb_channels = s->ti.channels;
742 s->nb_samples = s->ti.last_frame_samples ? s->ti.last_frame_samples
743 : s->ti.frame_samples;
745 frame->nb_samples = s->nb_samples;
746 if ((ret = ff_thread_get_buffer(avctx, frame, 0)) < 0)
747 return ret;
748 ff_thread_finish_setup(avctx);
750 if (avctx->bits_per_raw_sample <= 16) {
751 int buf_size = av_samples_get_buffer_size(NULL, avctx->ch_layout.nb_channels,
752 s->nb_samples,
753 AV_SAMPLE_FMT_S32P, 0);
754 if (buf_size < 0)
755 return buf_size;
756 av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
757 if (!s->decode_buffer)
758 return AVERROR(ENOMEM);
759 ret = av_samples_fill_arrays((uint8_t **)s->decoded, NULL,
760 s->decode_buffer, avctx->ch_layout.nb_channels,
761 s->nb_samples, AV_SAMPLE_FMT_S32P, 0);
762 if (ret < 0)
763 return ret;
764 } else {
765 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++)
766 s->decoded[chan] = (int32_t *)frame->extended_data[chan];
769 if (s->nb_samples < 16) {
770 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
771 int32_t *decoded = s->decoded[chan];
772 for (i = 0; i < s->nb_samples; i++)
773 decoded[i] = get_sbits(gb, avctx->bits_per_raw_sample);
775 } else {
776 if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
777 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++)
778 if (ret = decode_channel(s, chan))
779 return ret;
781 if (avctx->ch_layout.nb_channels == 2) {
782 s->nb_subframes = get_bits(gb, 1) + 1;
783 if (s->nb_subframes > 1) {
784 s->subframe_len[1] = get_bits(gb, 6);
787 s->dmode = get_bits(gb, 3);
788 if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
789 return ret;
791 } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
792 if (get_bits1(gb)) {
793 int ch_mask = 0;
795 chan = get_bits(gb, 4) + 1;
796 if (chan > avctx->ch_layout.nb_channels)
797 return AVERROR_INVALIDDATA;
799 for (i = 0; i < chan; i++) {
800 int nbit = get_bits(gb, 4);
802 if (nbit >= avctx->ch_layout.nb_channels)
803 return AVERROR_INVALIDDATA;
805 if (ch_mask & 1 << nbit)
806 return AVERROR_INVALIDDATA;
808 s->mcdparams[i].present = get_bits1(gb);
809 if (s->mcdparams[i].present) {
810 s->mcdparams[i].index = get_bits(gb, 2);
811 s->mcdparams[i].chan2 = get_bits(gb, 4);
812 if (s->mcdparams[i].chan2 >= avctx->ch_layout.nb_channels) {
813 av_log(avctx, AV_LOG_ERROR,
814 "invalid channel 2 (%d) for %d channel(s)\n",
815 s->mcdparams[i].chan2, avctx->ch_layout.nb_channels);
816 return AVERROR_INVALIDDATA;
818 if (s->mcdparams[i].index == 1) {
819 if ((nbit == s->mcdparams[i].chan2) ||
820 (ch_mask & 1 << s->mcdparams[i].chan2))
821 return AVERROR_INVALIDDATA;
823 ch_mask |= 1 << s->mcdparams[i].chan2;
824 } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
825 return AVERROR_INVALIDDATA;
828 s->mcdparams[i].chan1 = nbit;
830 ch_mask |= 1 << nbit;
832 } else {
833 chan = avctx->ch_layout.nb_channels;
834 for (i = 0; i < chan; i++) {
835 s->mcdparams[i].present = 0;
836 s->mcdparams[i].chan1 = i;
840 for (i = 0; i < chan; i++) {
841 if (s->mcdparams[i].present && s->mcdparams[i].index == 1)
842 if (ret = decode_channel(s, s->mcdparams[i].chan2))
843 return ret;
845 if (ret = decode_channel(s, s->mcdparams[i].chan1))
846 return ret;
848 if (s->mcdparams[i].present) {
849 s->dmode = mc_dmodes[s->mcdparams[i].index];
850 if (ret = decorrelate(s,
851 s->mcdparams[i].chan2,
852 s->mcdparams[i].chan1,
853 s->nb_samples - 1))
854 return ret;
859 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
860 int32_t *decoded = s->decoded[chan];
862 if (s->lpc_mode[chan])
863 decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
865 if (s->sample_shift[chan] > 0)
866 for (i = 0; i < s->nb_samples; i++)
867 decoded[i] *= 1U << s->sample_shift[chan];
871 align_get_bits(gb);
872 skip_bits(gb, 24);
873 if (get_bits_left(gb) < 0)
874 av_log(avctx, AV_LOG_DEBUG, "overread\n");
875 else if (get_bits_left(gb) > 0)
876 av_log(avctx, AV_LOG_DEBUG, "underread\n");
878 if (avctx->err_recognition & (AV_EF_CRCCHECK | AV_EF_COMPLIANT)) {
879 if (ff_tak_check_crc(pkt->data + hsize,
880 get_bits_count(gb) / 8 - hsize)) {
881 av_log(avctx, AV_LOG_ERROR, "CRC error\n");
882 if (avctx->err_recognition & AV_EF_EXPLODE)
883 return AVERROR_INVALIDDATA;
887 /* convert to output buffer */
888 switch (avctx->sample_fmt) {
889 case AV_SAMPLE_FMT_U8P:
890 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
891 uint8_t *samples = (uint8_t *)frame->extended_data[chan];
892 int32_t *decoded = s->decoded[chan];
893 for (i = 0; i < s->nb_samples; i++)
894 samples[i] = decoded[i] + 0x80U;
896 break;
897 case AV_SAMPLE_FMT_S16P:
898 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
899 int16_t *samples = (int16_t *)frame->extended_data[chan];
900 int32_t *decoded = s->decoded[chan];
901 for (i = 0; i < s->nb_samples; i++)
902 samples[i] = decoded[i];
904 break;
905 case AV_SAMPLE_FMT_S32P:
906 for (chan = 0; chan < avctx->ch_layout.nb_channels; chan++) {
907 int32_t *samples = (int32_t *)frame->extended_data[chan];
908 for (i = 0; i < s->nb_samples; i++)
909 samples[i] *= 1U << 8;
911 break;
914 *got_frame_ptr = 1;
916 return pkt->size;
919 #if HAVE_THREADS
920 static int update_thread_context(AVCodecContext *dst,
921 const AVCodecContext *src)
923 TAKDecContext *tsrc = src->priv_data;
924 TAKDecContext *tdst = dst->priv_data;
926 if (dst == src)
927 return 0;
928 memcpy(&tdst->ti, &tsrc->ti, sizeof(TAKStreamInfo));
929 return 0;
931 #endif
933 static av_cold int tak_decode_close(AVCodecContext *avctx)
935 TAKDecContext *s = avctx->priv_data;
937 av_freep(&s->decode_buffer);
939 return 0;
942 const FFCodec ff_tak_decoder = {
943 .p.name = "tak",
944 CODEC_LONG_NAME("TAK (Tom's lossless Audio Kompressor)"),
945 .p.type = AVMEDIA_TYPE_AUDIO,
946 .p.id = AV_CODEC_ID_TAK,
947 .priv_data_size = sizeof(TAKDecContext),
948 .init = tak_decode_init,
949 .close = tak_decode_close,
950 FF_CODEC_DECODE_CB(tak_decode_frame),
951 UPDATE_THREAD_CONTEXT(update_thread_context),
952 .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS | AV_CODEC_CAP_CHANNEL_CONF,
953 .p.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
954 AV_SAMPLE_FMT_S16P,
955 AV_SAMPLE_FMT_S32P,
956 AV_SAMPLE_FMT_NONE },