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"
38 #define MAX_SUBFRAMES 8 ///< max number of subframes per channel
39 #define MAX_PREDICTORS 256
41 typedef struct MCDParam {
42 int8_t present; ///< decorrelation parameter availability for this channel
43 int8_t index; ///< index into array of decorrelation types
48 typedef struct TAKDecContext {
49 AVCodecContext *avctx; ///< parent AVCodecContext
53 GetBitContext gb; ///< bitstream reader initialized to start at the current frame
56 int nb_samples; ///< number of samples in the current frame
57 uint8_t *decode_buffer;
58 unsigned int decode_buffer_size;
59 int32_t *decoded[TAK_MAX_CHANNELS]; ///< decoded samples for each channel
61 int8_t lpc_mode[TAK_MAX_CHANNELS];
62 int8_t sample_shift[TAK_MAX_CHANNELS]; ///< shift applied to every sample in the channel
63 int16_t predictors[MAX_PREDICTORS];
64 int nb_subframes; ///< number of subframes in the current frame
65 int16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples
68 int8_t dmode; ///< channel decorrelation type in the current frame
70 MCDParam mcdparams[TAK_MAX_CHANNELS]; ///< multichannel decorrelation parameters
72 int8_t coding_mode[128];
73 DECLARE_ALIGNED(16, int16_t, filter)[MAX_PREDICTORS];
74 DECLARE_ALIGNED(16, int16_t, residues)[544];
77 static const int8_t mc_dmodes[] = { 1, 3, 4, 6, };
79 static const uint16_t predictor_sizes[] = {
80 4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
83 static const struct CParam {
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 },
142 static int set_bps_params(AVCodecContext *avctx)
144 switch (avctx->bits_per_raw_sample) {
146 avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
149 avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
152 avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
155 av_log(avctx, AV_LOG_ERROR, "invalid/unsupported bits per sample: %d\n",
156 avctx->bits_per_raw_sample);
157 return AVERROR_INVALIDDATA;
163 static void set_sample_rate_params(AVCodecContext *avctx)
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;
172 static av_cold int tak_decode_init(AVCodecContext *avctx)
174 TAKDecContext *s = avctx->priv_data;
176 ff_audiodsp_init(&s->adsp);
177 ff_takdsp_init(&s->tdsp);
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 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 int 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 s->filter[j] = x - ((tfilter[i] + y) >> (15 - filter_quant));
455 s->filter[i] = x - ((tfilter[j] + y) >> (15 - filter_quant));
458 if ((ret = decode_residues(s, &decoded[filter_order],
459 subframe_size - filter_order)) < 0)
462 for (i = 0; i < filter_order; i++)
463 s->residues[i] = *decoded++ >> dshift;
465 y = FF_ARRAY_ELEMS(s->residues) - filter_order;
466 x = subframe_size - filter_order;
468 int tmp = FFMIN(y, x);
470 for (i = 0; i < tmp; i++) {
471 int v = 1 << (filter_quant - 1);
473 if (filter_order & -16)
474 v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
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 ];
482 v = (av_clip_intp2(v >> filter_quant, 13) << dshift) - *decoded;
484 s->residues[filter_order + i] = v >> dshift;
489 memcpy(s->residues, &s->residues[y], 2 * filter_order);
497 static int decode_channel(TAKDecContext *s, int chan)
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;
505 s->sample_shift[chan] = get_bits_esc4(gb);
506 if (s->sample_shift[chan] >= avctx->bits_per_raw_sample)
507 return AVERROR_INVALIDDATA;
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;
513 if (s->nb_subframes > 1) {
514 if (get_bits_left(gb) < (s->nb_subframes - 1) * 6)
515 return AVERROR_INVALIDDATA;
517 for (; i < s->nb_subframes - 1; i++) {
518 int v = get_bits(gb, 6);
520 s->subframe_len[i] = (v - prev) * s->subframe_scale;
521 if (s->subframe_len[i] <= 0)
522 return AVERROR_INVALIDDATA;
524 left -= s->subframe_len[i];
529 return AVERROR_INVALIDDATA;
531 s->subframe_len[i] = left;
534 for (i = 0; i < s->nb_subframes; i++) {
535 if ((ret = decode_subframe(s, decoded, s->subframe_len[i], prev)) < 0)
537 decoded += s->subframe_len[i];
538 prev = s->subframe_len[i];
544 static int decorrelate(TAKDecContext *s, int c1, int c2, int length)
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);
554 length += s->dmode < 6;
557 case 1: /* left/side */
558 s->tdsp.decorrelate_ls(p1, p2, length);
560 case 2: /* side/right */
561 s->tdsp.decorrelate_sr(p1, p2, length);
563 case 3: /* side/mid */
564 s->tdsp.decorrelate_sm(p1, p2, length);
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);
575 FFSWAP(int32_t*, p1, p2);
577 int length2, order_half, filter_order, dval1, dval2;
578 int tmp, x, code_size;
581 return AVERROR_INVALIDDATA;
583 dshift = get_bits_esc4(gb);
584 filter_order = 8 << get_bits1(gb);
585 dval1 = get_bits1(gb);
586 dval2 = get_bits1(gb);
588 for (i = 0; i < filter_order; i++) {
590 code_size = 14 - get_bits(gb, 3);
591 s->filter[i] = get_sbits(gb, code_size);
594 order_half = filter_order / 2;
595 length2 = length - (filter_order - 1);
597 /* decorrelate beginning samples */
599 for (i = 0; i < order_half; i++) {
606 /* decorrelate ending samples */
608 for (i = length2 + order_half; i < length; i++) {
616 for (i = 0; i < filter_order; i++)
617 s->residues[i] = *p2++ >> dshift;
620 x = FF_ARRAY_ELEMS(s->residues) - filter_order;
621 for (; length2 > 0; length2 -= tmp) {
622 tmp = FFMIN(length2, x);
624 for (i = 0; i < tmp; i++)
625 s->residues[filter_order + i] = *p2++ >> dshift;
627 for (i = 0; i < tmp; i++) {
630 if (filter_order == 16) {
631 v += s->adsp.scalarproduct_int16(&s->residues[i], s->filter,
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];
644 v = (av_clip_intp2(v >> 10, 13) << dshift) - *p1;
648 memcpy(s->residues, &s->residues[tmp], 2 * filter_order);
656 if (s->dmode > 0 && s->dmode < 6) {
664 static int tak_decode_frame(AVCodecContext *avctx, void *data,
665 int *got_frame_ptr, AVPacket *pkt)
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;
673 if (pkt->size < TAK_MIN_FRAME_HEADER_BYTES)
674 return AVERROR_INVALIDDATA;
676 if ((ret = init_get_bits8(gb, pkt->data, pkt->size)) < 0)
679 if ((ret = ff_tak_decode_frame_header(avctx, gb, &s->ti, 0)) < 0)
682 hsize = get_bits_count(gb) / 8;
683 if (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_COMPLIANT)) {
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;
691 if (s->ti.codec != TAK_CODEC_MONO_STEREO &&
692 s->ti.codec != TAK_CODEC_MULTICHANNEL) {
693 av_log(avctx, AV_LOG_ERROR, "unsupported codec: %d\n", s->ti.codec);
694 return AVERROR_PATCHWELCOME;
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;
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;
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;
712 if (s->ti.frame_samples <= 0) {
713 av_log(avctx, AV_LOG_ERROR, "unsupported/invalid number of samples\n");
714 return AVERROR_INVALIDDATA;
717 avctx->bits_per_raw_sample = s->ti.bps;
718 if ((ret = set_bps_params(avctx)) < 0)
720 if (s->ti.sample_rate != avctx->sample_rate) {
721 avctx->sample_rate = s->ti.sample_rate;
722 set_sample_rate_params(avctx);
725 avctx->channel_layout = s->ti.ch_layout;
726 avctx->channels = s->ti.channels;
728 s->nb_samples = s->ti.last_frame_samples ? s->ti.last_frame_samples
729 : s->ti.frame_samples;
731 frame->nb_samples = s->nb_samples;
732 if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0)
734 ff_thread_finish_setup(avctx);
736 if (avctx->bits_per_raw_sample <= 16) {
737 int buf_size = av_samples_get_buffer_size(NULL, avctx->channels,
739 AV_SAMPLE_FMT_S32P, 0);
742 av_fast_malloc(&s->decode_buffer, &s->decode_buffer_size, buf_size);
743 if (!s->decode_buffer)
744 return AVERROR(ENOMEM);
745 ret = av_samples_fill_arrays((uint8_t **)s->decoded, NULL,
746 s->decode_buffer, avctx->channels,
747 s->nb_samples, AV_SAMPLE_FMT_S32P, 0);
751 for (chan = 0; chan < avctx->channels; chan++)
752 s->decoded[chan] = (int32_t *)frame->extended_data[chan];
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);
762 if (s->ti.codec == TAK_CODEC_MONO_STEREO) {
763 for (chan = 0; chan < avctx->channels; chan++)
764 if (ret = decode_channel(s, chan))
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);
773 s->dmode = get_bits(gb, 3);
774 if (ret = decorrelate(s, 0, 1, s->nb_samples - 1))
777 } else if (s->ti.codec == TAK_CODEC_MULTICHANNEL) {
781 chan = get_bits(gb, 4) + 1;
782 if (chan > avctx->channels)
783 return AVERROR_INVALIDDATA;
785 for (i = 0; i < chan; i++) {
786 int nbit = get_bits(gb, 4);
788 if (nbit >= avctx->channels)
789 return AVERROR_INVALIDDATA;
791 if (ch_mask & 1 << nbit)
792 return AVERROR_INVALIDDATA;
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;
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;
809 ch_mask |= 1 << s->mcdparams[i].chan2;
810 } else if (!(ch_mask & 1 << s->mcdparams[i].chan2)) {
811 return AVERROR_INVALIDDATA;
814 s->mcdparams[i].chan1 = nbit;
816 ch_mask |= 1 << nbit;
819 chan = avctx->channels;
820 for (i = 0; i < chan; i++) {
821 s->mcdparams[i].present = 0;
822 s->mcdparams[i].chan1 = i;
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))
831 if (ret = decode_channel(s, s->mcdparams[i].chan1))
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,
845 for (chan = 0; chan < avctx->channels; chan++) {
846 int32_t *decoded = s->decoded[chan];
848 if (s->lpc_mode[chan])
849 decode_lpc(decoded, s->lpc_mode[chan], s->nb_samples);
851 if (s->sample_shift[chan] > 0)
852 for (i = 0; i < s->nb_samples; i++)
853 decoded[i] <<= s->sample_shift[chan];
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");
864 if (avctx->err_recognition & (AV_EF_CRCCHECK | AV_EF_COMPLIANT)) {
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;
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;
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];
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++)
906 static int init_thread_copy(AVCodecContext *avctx)
908 TAKDecContext *s = avctx->priv_data;
913 static int update_thread_context(AVCodecContext *dst,
914 const AVCodecContext *src)
916 TAKDecContext *tsrc = src->priv_data;
917 TAKDecContext *tdst = dst->priv_data;
921 memcpy(&tdst->ti, &tsrc->ti, sizeof(TAKStreamInfo));
926 static av_cold int tak_decode_close(AVCodecContext *avctx)
928 TAKDecContext *s = avctx->priv_data;
930 av_freep(&s->decode_buffer);
935 AVCodec ff_tak_decoder = {
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),
941 .init = tak_decode_init,
942 .close = tak_decode_close,
943 .decode = tak_decode_frame,
944 .init_thread_copy = ONLY_IF_THREADS_ENABLED(init_thread_copy),
945 .update_thread_context = ONLY_IF_THREADS_ENABLED(update_thread_context),
946 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
947 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
950 AV_SAMPLE_FMT_NONE },