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
24 * TAK (Tom's lossless Audio Kompressor) decoder
25 * @author Paul B Mahol
28 #include "libavutil/internal.h"
29 #include "libavutil/samplefmt.h"
36 #define MAX_SUBFRAMES 8 ///< max number of subframes per channel
37 #define MAX_PREDICTORS 256
39 typedef struct MCDParam {
40 int8_t present; ///< decorrelation parameter availability for this channel
41 int8_t index; ///< index into array of decorrelation types
46 typedef struct TAKDecContext {
47 AVCodecContext *avctx; ///< parent AVCodecContext
48 AVFrame frame; ///< AVFrame for decoded output
51 GetBitContext gb; ///< bitstream reader initialized to start at the current frame
54 int nb_samples; ///< number of samples in the current frame
55 uint8_t *decode_buffer;
56 unsigned int decode_buffer_size;
57 int32_t *decoded[TAK_MAX_CHANNELS]; ///< decoded samples for each channel
59 int8_t lpc_mode[TAK_MAX_CHANNELS];
60 int8_t sample_shift[TAK_MAX_CHANNELS]; ///< shift applied to every sample in the channel
61 int16_t predictors[MAX_PREDICTORS];
62 int nb_subframes; ///< number of subframes in the current frame
63 int16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples
66 int8_t dmode; ///< channel decorrelation type in the current frame
68 MCDParam mcdparams[TAK_MAX_CHANNELS]; ///< multichannel decorrelation parameters
70 int8_t coding_mode[128];
71 DECLARE_ALIGNED(16, int16_t, filter)[MAX_PREDICTORS];
72 DECLARE_ALIGNED(16, int16_t, residues)[544];
75 static const int8_t mc_dmodes[] = { 1, 3, 4, 6, };
77 static const uint16_t predictor_sizes[] = {
78 4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
81 static const struct CParam {
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 },
140 static int set_bps_params(AVCodecContext *avctx)
142 switch (avctx->bits_per_raw_sample) {
144 avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
147 avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
150 avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
153 av_log(avctx, AV_LOG_ERROR, "invalid/unsupported bits per sample: %d\n",
154 avctx->bits_per_raw_sample);
155 return AVERROR_INVALIDDATA;
161 static void set_sample_rate_params(AVCodecContext *avctx)
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;
170 static av_cold int tak_decode_init(AVCodecContext *avctx)
172 TAKDecContext *s = avctx->priv_data;
175 ff_dsputil_init(&s->dsp, avctx);
178 avcodec_get_frame_defaults(&s->frame);
179 avctx->coded_frame = &s->frame;
180 avctx->bits_per_raw_sample = avctx->bits_per_coded_sample;
182 set_sample_rate_params(avctx);
184 return set_bps_params(avctx);
187 static void decode_lpc(int32_t *coeffs, int mode, int length)
196 for (i = 0; i < length - 1 >> 1; i++) {
198 coeffs[1] += *coeffs;
204 } else if (mode == 2) {
206 int a2 = a1 + *coeffs;
210 for (i = 0; i < length - 2 >> 1; i++) {
211 int a3 = *coeffs + a1;
222 } else if (mode == 3) {
224 int a2 = a1 + *coeffs;
231 for (i = 0; i < length - 3; i++) {
242 static int decode_segment(TAKDecContext *s, int8_t mode, int32_t *decoded, int len)
245 GetBitContext *gb = &s->gb;
249 memset(decoded, 0, len * sizeof(*decoded));
253 if (mode > FF_ARRAY_ELEMS(xcodes))
254 return AVERROR_INVALIDDATA;
255 code = xcodes[mode - 1];
257 for (i = 0; i < len; i++) {
258 int x = get_bits_long(gb, code.init);
259 if (x >= code.escape && get_bits1(gb)) {
261 if (x >= code.aescape) {
262 int scale = get_unary(gb, 1, 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);
269 return AVERROR_INVALIDDATA;
271 scale = get_bits_long(gb, scale_bits) + 1;
272 x += code.scale * scale;
276 x += code.scale * scale - code.escape;
280 decoded[i] = (x >> 1) ^ -(x & 1);
286 static int decode_residues(TAKDecContext *s, int32_t *decoded, int length)
288 GetBitContext *gb = &s->gb;
291 if (length > s->nb_samples)
292 return AVERROR_INVALIDDATA;
297 wlength = length / s->uval;
299 rval = length - (wlength * s->uval);
301 if (rval < s->uval / 2)
306 if (wlength <= 1 || wlength > 128)
307 return AVERROR_INVALIDDATA;
309 s->coding_mode[0] = mode = get_bits(gb, 6);
311 for (i = 1; i < wlength; i++) {
312 int c = get_unary(gb, 1, 6);
316 mode = get_bits(gb, 6);
321 /* mode += sign ? (1 - c) : (c - 1) */
322 int sign = get_bits1(gb);
323 mode += (-sign ^ (c - 1)) + sign;
333 s->coding_mode[i] = mode;
337 while (i < wlength) {
340 mode = s->coding_mode[i];
342 if (i >= wlength - 1)
350 } while (s->coding_mode[i] == mode);
352 if ((ret = decode_segment(s, mode, decoded, len)) < 0)
357 mode = get_bits(gb, 6);
358 if ((ret = decode_segment(s, mode, decoded, length)) < 0)
365 static int get_bits_esc4(GetBitContext *gb)
368 return get_bits(gb, 4) + 1;
373 static int decode_subframe(TAKDecContext *s, int32_t *decoded,
374 int subframe_size, int prev_subframe_size)
376 GetBitContext *gb = &s->gb;
377 int tmp, x, y, i, j, ret = 0;
378 int dshift, size, filter_quant, filter_order;
379 int tfilter[MAX_PREDICTORS];
382 return decode_residues(s, decoded, subframe_size);
384 filter_order = predictor_sizes[get_bits(gb, 4)];
386 if (prev_subframe_size > 0 && get_bits1(gb)) {
387 if (filter_order > prev_subframe_size)
388 return AVERROR_INVALIDDATA;
390 decoded -= filter_order;
391 subframe_size += filter_order;
393 if (filter_order > subframe_size)
394 return AVERROR_INVALIDDATA;
398 if (filter_order > subframe_size)
399 return AVERROR_INVALIDDATA;
401 lpc_mode = get_bits(gb, 2);
403 return AVERROR_INVALIDDATA;
405 if ((ret = decode_residues(s, decoded, filter_order)) < 0)
409 decode_lpc(decoded, lpc_mode, filter_order);
412 dshift = get_bits_esc4(gb);
413 size = get_bits1(gb) + 6;
417 filter_quant -= get_bits(gb, 3) + 1;
418 if (filter_quant < 3)
419 return AVERROR_INVALIDDATA;
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 tmp = size - get_bits1(gb);
429 for (i = 4; i < filter_order; i++) {
431 x = tmp - get_bits(gb, 2);
432 s->predictors[i] = get_sbits(gb, x) << (10 - size);
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];
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;
448 tfilter[i] = s->predictors[i] << 6;
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 tmp = y + tfilter[j];
455 s->filter[j] = x - ((tfilter[i] + y) >> (15 - filter_quant));
456 s->filter[i] = x - ((tfilter[j] + y) >> (15 - filter_quant));
459 if ((ret = decode_residues(s, &decoded[filter_order],
460 subframe_size - filter_order)) < 0)
463 for (i = 0; i < filter_order; i++)
464 s->residues[i] = *decoded++ >> dshift;
466 y = FF_ARRAY_ELEMS(s->residues) - filter_order;
467 x = subframe_size - filter_order;
471 for (i = 0; i < tmp; i++) {
472 int v = 1 << (filter_quant - 1);
474 if (!(filter_order & 15)) {
475 v += s->dsp.scalarproduct_int16(&s->residues[i], s->filter,
477 } else if (filter_order & 4) {
478 for (j = 0; j < filter_order; j += 4) {
479 v += s->residues[i + j + 3] * s->filter[j + 3] +
480 s->residues[i + j + 2] * s->filter[j + 2] +
481 s->residues[i + j + 1] * s->filter[j + 1] +
482 s->residues[i + j ] * s->filter[j ];
485 for (j = 0; j < filter_order; j += 8) {
486 v += s->residues[i + j + 7] * s->filter[j + 7] +
487 s->residues[i + j + 6] * s->filter[j + 6] +
488 s->residues[i + j + 5] * s->filter[j + 5] +
489 s->residues[i + j + 4] * s->filter[j + 4] +
490 s->residues[i + j + 3] * s->filter[j + 3] +
491 s->residues[i + j + 2] * s->filter[j + 2] +
492 s->residues[i + j + 1] * s->filter[j + 1] +
493 s->residues[i + j ] * s->filter[j ];
496 v = (av_clip(v >> filter_quant, -8192, 8191) << dshift) - *decoded;
498 s->residues[filter_order + i] = v >> dshift;
503 memcpy(s->residues, &s->residues[y], 2 * filter_order);
511 static int decode_channel(TAKDecContext *s, int chan)
513 AVCodecContext *avctx = s->avctx;
514 GetBitContext *gb = &s->gb;
515 int32_t *decoded = s->decoded[chan];
516 int left = s->nb_samples - 1;
517 int i = 0, ret, prev = 0;
519 s->sample_shift[chan] = get_bits_esc4(gb);
520 if (s->sample_shift[chan] >= avctx->bits_per_raw_sample)
521 return AVERROR_INVALIDDATA;
523 *decoded++ = get_sbits(gb, avctx->bits_per_raw_sample - s->sample_shift[chan]);
524 s->lpc_mode[chan] = get_bits(gb, 2);
525 s->nb_subframes = get_bits(gb, 3) + 1;
527 if (s->nb_subframes > 1) {
528 if (get_bits_left(gb) < (s->nb_subframes - 1) * 6)
529 return AVERROR_INVALIDDATA;
531 for (; i < s->nb_subframes - 1; i++) {
532 int v = get_bits(gb, 6);
534 s->subframe_len[i] = (v - prev) * s->subframe_scale;
535 if (s->subframe_len[i] <= 0)
536 return AVERROR_INVALIDDATA;
538 left -= s->subframe_len[i];
543 return AVERROR_INVALIDDATA;
545 s->subframe_len[i] = left;
548 for (i = 0; i < s->nb_subframes; i++) {
549 if ((ret = decode_subframe(s, decoded, s->subframe_len[i], prev)) < 0)
551 decoded += s->subframe_len[i];
552 prev = s->subframe_len[i];
558 static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
560 GetBitContext *gb = &s->gb;
561 int32_t *p1 = s->decoded[c1] + 1;
562 int32_t *p2 = s->decoded[c2] + 1;
567 case 1: /* left/side */
568 for (i = 0; i < length; i++) {
574 case 2: /* side/right */
575 for (i = 0; i < length; i++) {
581 case 3: /* side/mid */
582 for (i = 0; i < length; i++) {
590 case 4: /* side/left with scale factor */
591 FFSWAP(int32_t*, p1, p2);
592 case 5: /* side/right with scale factor */
593 dshift = get_bits_esc4(gb);
594 dfactor = get_sbits(gb, 10);
595 for (i = 0; i < length; i++) {
598 b = dfactor * (b >> dshift) + 128 >> 8 << dshift;
603 FFSWAP(int32_t*, p1, p2);
605 int length2, order_half, filter_order, dval1, dval2;
606 int tmp, x, code_size;
609 return AVERROR_INVALIDDATA;
611 dshift = get_bits_esc4(gb);
612 filter_order = 8 << get_bits1(gb);
613 dval1 = get_bits1(gb);
614 dval2 = get_bits1(gb);
616 for (i = 0; i < filter_order; i++) {
618 code_size = 14 - get_bits(gb, 3);
619 s->filter[i] = get_sbits(gb, code_size);
622 order_half = filter_order / 2;
623 length2 = length - (filter_order - 1);
625 /* decorrelate beginning samples */
627 for (i = 0; i < order_half; i++) {
634 /* decorrelate ending samples */
636 for (i = length2 + order_half; i < length; i++) {
644 for (i = 0; i < filter_order; i++)
645 s->residues[i] = *p2++ >> dshift;
648 x = FF_ARRAY_ELEMS(s->residues) - filter_order;
649 for (; length2 > 0; length2 -= tmp) {
650 tmp = FFMIN(length2, x);
652 for (i = 0; i < tmp; i++)
653 s->residues[filter_order + i] = *p2++ >> dshift;
655 for (i = 0; i < tmp; i++) {
658 if (filter_order == 16) {
659 v += s->dsp.scalarproduct_int16(&s->residues[i], s->filter,
662 v += s->residues[i + 7] * s->filter[7] +
663 s->residues[i + 6] * s->filter[6] +
664 s->residues[i + 5] * s->filter[5] +
665 s->residues[i + 4] * s->filter[4] +
666 s->residues[i + 3] * s->filter[3] +
667 s->residues[i + 2] * s->filter[2] +
668 s->residues[i + 1] * s->filter[1] +
669 s->residues[i ] * s->filter[0];
672 v = (av_clip(v >> 10, -8192, 8191) << dshift) - *p1;
676 memcpy(s->residues, &s->residues[tmp], 2 * filter_order);
687 static int tak_decode_frame(AVCodecContext *avctx, void *data,
688 int *got_frame_ptr, AVPacket *pkt)
690 TAKDecContext *s = avctx->priv_data;
691 GetBitContext *gb = &s->gb;
692 int chan, i, ret, hsize;
694 if (pkt->size < TAK_MIN_FRAME_HEADER_BYTES)
695 return AVERROR_INVALIDDATA;
697 if ((ret = init_get_bits8(gb, pkt->data, pkt->size)) < 0)
700 if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
703 if (avctx->err_recognition & AV_EF_CRCCHECK) {
704 hsize = get_bits_count(gb) / 8;
705 if (ff_tak_check_crc(pkt->data, hsize)) {
706 av_log(avctx, AV_LOG_ERROR, "CRC error\n");
707 return AVERROR_INVALIDDATA;
711 if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
712 s->ti.codec != TAK_CODEC_MULTICHANNEL) {
713 av_log(avctx, AV_LOG_ERROR, "unsupported codec: %d\n", s->ti.codec);
714 return AVERROR_PATCHWELCOME;
716 if (s->ti.data_type) {
717 av_log(avctx, AV_LOG_ERROR,
718 "unsupported data type: %d\n", s->ti.data_type);
719 return AVERROR_INVALIDDATA;
721 if (s->ti.codec == TAK_CODEC_MONO_STEREO && s->ti.channels > 2) {
722 av_log(avctx, AV_LOG_ERROR,
723 "invalid number of channels: %d\n", s->ti.channels);
724 return AVERROR_INVALIDDATA;
726 if (s->ti.channels > 6) {
727 av_log(avctx, AV_LOG_ERROR,
728 "unsupported number of channels: %d\n", s->ti.channels);
729 return AVERROR_INVALIDDATA;
732 if (s->ti.frame_samples <= 0) {
733 av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
734 return AVERROR_INVALIDDATA;
737 if (s->ti.bps != avctx->bits_per_raw_sample) {
738 avctx->bits_per_raw_sample = s->ti.bps;
739 if ((ret = set_bps_params(avctx)) < 0)
742 if (s->ti.sample_rate != avctx->sample_rate) {
743 avctx->sample_rate = s->ti.sample_rate;
744 set_sample_rate_params(avctx);
747 avctx->channel_layout = s->ti.ch_layout;
748 avctx->channels = s->ti.channels;
750 s->nb_samples = s->ti.last_frame_samples ? s->ti.last_frame_samples
751 : s->ti.frame_samples;
753 s->frame.nb_samples = s->nb_samples;
754 if ((ret = ff_get_buffer(avctx, &s->frame)) < 0)
757 if (avctx->bits_per_raw_sample <= 16) {
758 int buf_size = av_samples_get_buffer_size(NULL, avctx->channels,
760 AV_SAMPLE_FMT_S32P, 0);
761 av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
762 if (!s->decode_buffer)
763 return AVERROR(ENOMEM);
764 ret = av_samples_fill_arrays((uint8_t **)s->decoded, NULL,
765 s->decode_buffer, avctx->channels,
766 s->nb_samples, AV_SAMPLE_FMT_S32P, 0);
770 for (chan = 0; chan < avctx->channels; chan++)
771 s->decoded[chan] = (int32_t *)s->frame.extended_data[chan];
774 if (s->nb_samples < 16) {
775 for (chan = 0; chan < avctx->channels; chan++) {
776 int32_t *decoded = s->decoded[chan];
777 for (i = 0; i < s->nb_samples; i++)
778 decoded[i] = get_sbits(gb, avctx->bits_per_raw_sample);
781 if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
782 for (chan = 0; chan < avctx->channels; chan++)
783 if (ret = decode_channel(s, chan))
786 if (avctx->channels == 2) {
787 s->nb_subframes = get_bits(gb, 1) + 1;
788 if (s->nb_subframes > 1) {
789 s->subframe_len[1] = get_bits(gb, 6);
792 s->dmode = get_bits(gb, 3);
793 if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
796 } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
800 chan = get_bits(gb, 4) + 1;
801 if (chan > avctx->channels)
802 return AVERROR_INVALIDDATA;
804 for (i = 0; i < chan; i++) {
805 int nbit = get_bits(gb, 4);
807 if (nbit >= avctx->channels)
808 return AVERROR_INVALIDDATA;
810 if (ch_mask & 1 << nbit)
811 return AVERROR_INVALIDDATA;
813 s->mcdparams[i].present = get_bits1(gb);
814 if (s->mcdparams[i].present) {
815 s->mcdparams[i].index = get_bits(gb, 2);
816 s->mcdparams[i].chan2 = get_bits(gb, 4);
817 if (s->mcdparams[i].index == 1) {
818 if ((nbit == s->mcdparams[i].chan2) ||
819 (ch_mask & 1 << s->mcdparams[i].chan2))
820 return AVERROR_INVALIDDATA;
822 ch_mask |= 1 << s->mcdparams[i].chan2;
823 } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
824 return AVERROR_INVALIDDATA;
827 s->mcdparams[i].chan1 = nbit;
829 ch_mask |= 1 << nbit;
832 chan = avctx->channels;
833 for (i = 0; i < chan; i++) {
834 s->mcdparams[i].present = 0;
835 s->mcdparams[i].chan1 = i;
839 for (i = 0; i < chan; i++) {
840 if (s->mcdparams[i].present && s->mcdparams[i].index == 1)
841 if (ret = decode_channel(s, s->mcdparams[i].chan2))
844 if (ret = decode_channel(s, s->mcdparams[i].chan1))
847 if (s->mcdparams[i].present) {
848 s->dmode = mc_dmodes[s->mcdparams[i].index];
849 if (ret = decorrelate(s,
850 s->mcdparams[i].chan2,
851 s->mcdparams[i].chan1,
858 for (chan = 0; chan < avctx->channels; chan++) {
859 int32_t *decoded = s->decoded[chan];
861 if (s->lpc_mode[chan])
862 decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
864 if (s->sample_shift[chan] > 0)
865 for (i = 0; i < s->nb_samples; i++)
866 decoded[i] <<= s->sample_shift[chan];
872 if (get_bits_left(gb) < 0)
873 av_log(avctx, AV_LOG_DEBUG, "overread\n");
874 else if (get_bits_left(gb) > 0)
875 av_log(avctx, AV_LOG_DEBUG, "underread\n");
877 if (avctx->err_recognition & AV_EF_CRCCHECK) {
878 if (ff_tak_check_crc(pkt->data + hsize,
879 get_bits_count(gb) / 8 - hsize)) {
880 av_log(avctx, AV_LOG_ERROR, "CRC error\n");
881 return AVERROR_INVALIDDATA;
885 /* convert to output buffer */
886 switch (avctx->sample_fmt) {
887 case AV_SAMPLE_FMT_U8P:
888 for (chan = 0; chan < avctx->channels; chan++) {
889 uint8_t *samples = (uint8_t *)s->frame.extended_data[chan];
890 int32_t *decoded = s->decoded[chan];
891 for (i = 0; i < s->nb_samples; i++)
892 samples[i] = decoded[i] + 0x80;
895 case AV_SAMPLE_FMT_S16P:
896 for (chan = 0; chan < avctx->channels; chan++) {
897 int16_t *samples = (int16_t *)s->frame.extended_data[chan];
898 int32_t *decoded = s->decoded[chan];
899 for (i = 0; i < s->nb_samples; i++)
900 samples[i] = decoded[i];
903 case AV_SAMPLE_FMT_S32P:
904 for (chan = 0; chan < avctx->channels; chan++) {
905 int32_t *samples = (int32_t *)s->frame.extended_data[chan];
906 for (i = 0; i < s->nb_samples; i++)
913 *(AVFrame *)data = s->frame;
918 static av_cold int tak_decode_close(AVCodecContext *avctx)
920 TAKDecContext *s = avctx->priv_data;
922 av_freep(&s->decode_buffer);
927 AVCodec ff_tak_decoder = {
929 .type = AVMEDIA_TYPE_AUDIO,
930 .id = AV_CODEC_ID_TAK,
931 .priv_data_size = sizeof(TAKDecContext),
932 .init = tak_decode_init,
933 .close = tak_decode_close,
934 .decode = tak_decode_frame,
935 .capabilities = CODEC_CAP_DR1,
936 .long_name = NULL_IF_CONFIG_SMALL("TAK (Tom's lossless Audio Kompressor)"),
937 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
940 AV_SAMPLE_FMT_NONE },