3 * Copyright (c) 2012 Paul B Mahol
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; 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
50 GetBitContext gb
; // bitstream reader initialized to start at the current frame
53 int nb_samples
; // number of samples in the current frame
54 uint8_t *decode_buffer
;
55 unsigned int decode_buffer_size
;
56 int32_t *decoded
[TAK_MAX_CHANNELS
]; // decoded samples for each channel
58 int8_t lpc_mode
[TAK_MAX_CHANNELS
];
59 int8_t sample_shift
[TAK_MAX_CHANNELS
]; // shift applied to every sample in the channel
62 int8_t dmode
; // channel decorrelation type in the current frame
64 MCDParam mcdparams
[TAK_MAX_CHANNELS
]; // multichannel decorrelation parameters
67 unsigned int residues_buf_size
;
70 static const int8_t mc_dmodes
[] = { 1, 3, 4, 6, };
72 static const uint16_t predictor_sizes
[] = {
73 4, 8, 12, 16, 24, 32, 48, 64, 80, 96, 128, 160, 192, 224, 256, 0,
76 static const struct CParam
{
83 { 0x01, 0x0000001, 0x0000001, 0x0000003, 0x0000008 },
84 { 0x02, 0x0000003, 0x0000001, 0x0000007, 0x0000006 },
85 { 0x03, 0x0000005, 0x0000002, 0x000000E, 0x000000D },
86 { 0x03, 0x0000003, 0x0000003, 0x000000D, 0x0000018 },
87 { 0x04, 0x000000B, 0x0000004, 0x000001C, 0x0000019 },
88 { 0x04, 0x0000006, 0x0000006, 0x000001A, 0x0000030 },
89 { 0x05, 0x0000016, 0x0000008, 0x0000038, 0x0000032 },
90 { 0x05, 0x000000C, 0x000000C, 0x0000034, 0x0000060 },
91 { 0x06, 0x000002C, 0x0000010, 0x0000070, 0x0000064 },
92 { 0x06, 0x0000018, 0x0000018, 0x0000068, 0x00000C0 },
93 { 0x07, 0x0000058, 0x0000020, 0x00000E0, 0x00000C8 },
94 { 0x07, 0x0000030, 0x0000030, 0x00000D0, 0x0000180 },
95 { 0x08, 0x00000B0, 0x0000040, 0x00001C0, 0x0000190 },
96 { 0x08, 0x0000060, 0x0000060, 0x00001A0, 0x0000300 },
97 { 0x09, 0x0000160, 0x0000080, 0x0000380, 0x0000320 },
98 { 0x09, 0x00000C0, 0x00000C0, 0x0000340, 0x0000600 },
99 { 0x0A, 0x00002C0, 0x0000100, 0x0000700, 0x0000640 },
100 { 0x0A, 0x0000180, 0x0000180, 0x0000680, 0x0000C00 },
101 { 0x0B, 0x0000580, 0x0000200, 0x0000E00, 0x0000C80 },
102 { 0x0B, 0x0000300, 0x0000300, 0x0000D00, 0x0001800 },
103 { 0x0C, 0x0000B00, 0x0000400, 0x0001C00, 0x0001900 },
104 { 0x0C, 0x0000600, 0x0000600, 0x0001A00, 0x0003000 },
105 { 0x0D, 0x0001600, 0x0000800, 0x0003800, 0x0003200 },
106 { 0x0D, 0x0000C00, 0x0000C00, 0x0003400, 0x0006000 },
107 { 0x0E, 0x0002C00, 0x0001000, 0x0007000, 0x0006400 },
108 { 0x0E, 0x0001800, 0x0001800, 0x0006800, 0x000C000 },
109 { 0x0F, 0x0005800, 0x0002000, 0x000E000, 0x000C800 },
110 { 0x0F, 0x0003000, 0x0003000, 0x000D000, 0x0018000 },
111 { 0x10, 0x000B000, 0x0004000, 0x001C000, 0x0019000 },
112 { 0x10, 0x0006000, 0x0006000, 0x001A000, 0x0030000 },
113 { 0x11, 0x0016000, 0x0008000, 0x0038000, 0x0032000 },
114 { 0x11, 0x000C000, 0x000C000, 0x0034000, 0x0060000 },
115 { 0x12, 0x002C000, 0x0010000, 0x0070000, 0x0064000 },
116 { 0x12, 0x0018000, 0x0018000, 0x0068000, 0x00C0000 },
117 { 0x13, 0x0058000, 0x0020000, 0x00E0000, 0x00C8000 },
118 { 0x13, 0x0030000, 0x0030000, 0x00D0000, 0x0180000 },
119 { 0x14, 0x00B0000, 0x0040000, 0x01C0000, 0x0190000 },
120 { 0x14, 0x0060000, 0x0060000, 0x01A0000, 0x0300000 },
121 { 0x15, 0x0160000, 0x0080000, 0x0380000, 0x0320000 },
122 { 0x15, 0x00C0000, 0x00C0000, 0x0340000, 0x0600000 },
123 { 0x16, 0x02C0000, 0x0100000, 0x0700000, 0x0640000 },
124 { 0x16, 0x0180000, 0x0180000, 0x0680000, 0x0C00000 },
125 { 0x17, 0x0580000, 0x0200000, 0x0E00000, 0x0C80000 },
126 { 0x17, 0x0300000, 0x0300000, 0x0D00000, 0x1800000 },
127 { 0x18, 0x0B00000, 0x0400000, 0x1C00000, 0x1900000 },
128 { 0x18, 0x0600000, 0x0600000, 0x1A00000, 0x3000000 },
129 { 0x19, 0x1600000, 0x0800000, 0x3800000, 0x3200000 },
130 { 0x19, 0x0C00000, 0x0C00000, 0x3400000, 0x6000000 },
131 { 0x1A, 0x2C00000, 0x1000000, 0x7000000, 0x6400000 },
132 { 0x1A, 0x1800000, 0x1800000, 0x6800000, 0xC000000 },
135 static av_cold
void tak_init_static_data(AVCodec
*codec
)
140 static int set_bps_params(AVCodecContext
*avctx
)
142 switch (avctx
->bits_per_coded_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
, "unsupported bits per sample: %d\n",
154 avctx
->bits_per_coded_sample
);
155 return AVERROR_INVALIDDATA
;
157 avctx
->bits_per_raw_sample
= avctx
->bits_per_coded_sample
;
162 static void set_sample_rate_params(AVCodecContext
*avctx
)
164 TAKDecContext
*s
= avctx
->priv_data
;
165 int shift
= 3 - (avctx
->sample_rate
/ 11025);
166 shift
= FFMAX(0, shift
);
167 s
->uval
= FFALIGN(avctx
->sample_rate
+ 511 >> 9, 4) << shift
;
168 s
->subframe_scale
= FFALIGN(avctx
->sample_rate
+ 511 >> 9, 4) << 1;
171 static av_cold
int tak_decode_init(AVCodecContext
*avctx
)
173 TAKDecContext
*s
= avctx
->priv_data
;
175 ff_dsputil_init(&s
->dsp
, avctx
);
179 set_sample_rate_params(avctx
);
181 return set_bps_params(avctx
);
184 static void decode_lpc(int32_t *coeffs
, int mode
, int length
)
193 for (i
= 0; i
< length
- 1 >> 1; i
++) {
195 coeffs
[1] += *coeffs
;
201 } else if (mode
== 2) {
203 int a2
= a1
+ *coeffs
;
207 for (i
= 0; i
< length
- 2 >> 1; i
++) {
208 int a3
= *coeffs
+ a1
;
219 } else if (mode
== 3) {
221 int a2
= a1
+ *coeffs
;
228 for (i
= 0; i
< length
- 3; i
++) {
239 static int decode_segment(GetBitContext
*gb
, int mode
, int32_t *decoded
,
246 memset(decoded
, 0, len
* sizeof(*decoded
));
250 if (mode
> FF_ARRAY_ELEMS(xcodes
))
251 return AVERROR_INVALIDDATA
;
252 code
= xcodes
[mode
- 1];
254 for (i
= 0; i
< len
; i
++) {
255 int x
= get_bits_long(gb
, code
.init
);
256 if (x
>= code
.escape
&& get_bits1(gb
)) {
258 if (x
>= code
.aescape
) {
259 int scale
= get_unary(gb
, 1, 9);
261 int scale_bits
= get_bits(gb
, 3);
262 if (scale_bits
> 0) {
263 if (scale_bits
== 7) {
264 scale_bits
+= get_bits(gb
, 5);
266 return AVERROR_INVALIDDATA
;
268 scale
= get_bits_long(gb
, scale_bits
) + 1;
269 x
+= code
.scale
* scale
;
273 x
+= code
.scale
* scale
- code
.escape
;
277 decoded
[i
] = (x
>> 1) ^ -(x
& 1);
283 static int decode_residues(TAKDecContext
*s
, int32_t *decoded
, int length
)
285 GetBitContext
*gb
= &s
->gb
;
288 if (length
> s
->nb_samples
)
289 return AVERROR_INVALIDDATA
;
293 int coding_mode
[128];
295 wlength
= length
/ s
->uval
;
297 rval
= length
- (wlength
* s
->uval
);
299 if (rval
< s
->uval
/ 2)
304 if (wlength
<= 1 || wlength
> 128)
305 return AVERROR_INVALIDDATA
;
307 coding_mode
[0] = mode
= get_bits(gb
, 6);
309 for (i
= 1; i
< wlength
; i
++) {
310 int c
= get_unary(gb
, 1, 6);
314 mode
= get_bits(gb
, 6);
319 /* mode += sign ? (1 - c) : (c - 1) */
320 int sign
= get_bits1(gb
);
321 mode
+= (-sign
^ (c
- 1)) + sign
;
331 coding_mode
[i
] = mode
;
335 while (i
< wlength
) {
338 mode
= coding_mode
[i
];
340 if (i
>= wlength
- 1)
348 } while (coding_mode
[i
] == mode
);
350 if ((ret
= decode_segment(gb
, mode
, decoded
, len
)) < 0)
355 mode
= get_bits(gb
, 6);
356 if ((ret
= decode_segment(gb
, mode
, decoded
, length
)) < 0)
363 static int get_bits_esc4(GetBitContext
*gb
)
366 return get_bits(gb
, 4) + 1;
371 static void decode_filter_coeffs(TAKDecContext
*s
, int filter_order
, int size
,
372 int filter_quant
, int16_t *filter
)
374 GetBitContext
*gb
= &s
->gb
;
376 int filter_tmp
[MAX_PREDICTORS
];
377 int16_t predictors
[MAX_PREDICTORS
];
379 predictors
[0] = get_sbits(gb
, 10);
380 predictors
[1] = get_sbits(gb
, 10);
381 predictors
[2] = get_sbits(gb
, size
) << (10 - size
);
382 predictors
[3] = get_sbits(gb
, size
) << (10 - size
);
383 if (filter_order
> 4) {
384 int av_uninit(code_size
);
385 int code_size_base
= size
- get_bits1(gb
);
387 for (i
= 4; i
< filter_order
; i
++) {
389 code_size
= code_size_base
- get_bits(gb
, 2);
390 predictors
[i
] = get_sbits(gb
, code_size
) << (10 - size
);
394 filter_tmp
[0] = predictors
[0] << 6;
395 for (i
= 1; i
< filter_order
; i
++) {
396 int *p1
= &filter_tmp
[0];
397 int *p2
= &filter_tmp
[i
- 1];
399 for (j
= 0; j
< (i
+ 1) / 2; j
++) {
400 int tmp
= *p1
+ (predictors
[i
] * *p2
+ 256 >> 9);
401 *p2
= *p2
+ (predictors
[i
] * *p1
+ 256 >> 9);
407 filter_tmp
[i
] = predictors
[i
] << 6;
410 a
= 1 << (32 - (15 - filter_quant
));
411 b
= 1 << ((15 - filter_quant
) - 1);
412 for (i
= 0, j
= filter_order
- 1; i
< filter_order
/ 2; i
++, j
--) {
413 filter
[j
] = a
- ((filter_tmp
[i
] + b
) >> (15 - filter_quant
));
414 filter
[i
] = a
- ((filter_tmp
[j
] + b
) >> (15 - filter_quant
));
418 static int decode_subframe(TAKDecContext
*s
, int32_t *decoded
,
419 int subframe_size
, int prev_subframe_size
)
421 LOCAL_ALIGNED_16(int16_t, filter
, [MAX_PREDICTORS
]);
422 GetBitContext
*gb
= &s
->gb
;
424 int dshift
, size
, filter_quant
, filter_order
;
426 memset(filter
, 0, MAX_PREDICTORS
* sizeof(*filter
));
429 return decode_residues(s
, decoded
, subframe_size
);
431 filter_order
= predictor_sizes
[get_bits(gb
, 4)];
433 if (prev_subframe_size
> 0 && get_bits1(gb
)) {
434 if (filter_order
> prev_subframe_size
)
435 return AVERROR_INVALIDDATA
;
437 decoded
-= filter_order
;
438 subframe_size
+= filter_order
;
440 if (filter_order
> subframe_size
)
441 return AVERROR_INVALIDDATA
;
445 if (filter_order
> subframe_size
)
446 return AVERROR_INVALIDDATA
;
448 lpc_mode
= get_bits(gb
, 2);
450 return AVERROR_INVALIDDATA
;
452 if ((ret
= decode_residues(s
, decoded
, filter_order
)) < 0)
456 decode_lpc(decoded
, lpc_mode
, filter_order
);
459 dshift
= get_bits_esc4(gb
);
460 size
= get_bits1(gb
) + 6;
464 filter_quant
-= get_bits(gb
, 3) + 1;
465 if (filter_quant
< 3)
466 return AVERROR_INVALIDDATA
;
469 decode_filter_coeffs(s
, filter_order
, size
, filter_quant
, filter
);
471 if ((ret
= decode_residues(s
, &decoded
[filter_order
],
472 subframe_size
- filter_order
)) < 0)
475 av_fast_malloc(&s
->residues
, &s
->residues_buf_size
,
476 FFALIGN(subframe_size
+ 16, 16) * sizeof(*s
->residues
));
478 return AVERROR(ENOMEM
);
479 memset(s
->residues
, 0, s
->residues_buf_size
);
481 for (i
= 0; i
< filter_order
; i
++)
482 s
->residues
[i
] = *decoded
++ >> dshift
;
484 for (i
= 0; i
< subframe_size
- filter_order
; i
++) {
485 int v
= 1 << (filter_quant
- 1);
487 v
+= s
->dsp
.scalarproduct_int16(&s
->residues
[i
], filter
,
488 FFALIGN(filter_order
, 16));
490 v
= (av_clip(v
>> filter_quant
, -8192, 8191) << dshift
) - *decoded
;
492 s
->residues
[filter_order
+ i
] = v
>> dshift
;
500 static int decode_channel(TAKDecContext
*s
, int chan
)
502 AVCodecContext
*avctx
= s
->avctx
;
503 GetBitContext
*gb
= &s
->gb
;
504 int32_t *decoded
= s
->decoded
[chan
];
505 int left
= s
->nb_samples
- 1;
506 int i
, prev
, ret
, nb_subframes
;
507 int subframe_len
[MAX_SUBFRAMES
];
509 s
->sample_shift
[chan
] = get_bits_esc4(gb
);
510 if (s
->sample_shift
[chan
] >= avctx
->bits_per_coded_sample
)
511 return AVERROR_INVALIDDATA
;
513 /* NOTE: TAK 2.2.0 appears to set the sample value to 0 if
514 * bits_per_coded_sample - sample_shift is 1, but this produces
515 * non-bit-exact output. Reading the 1 bit using get_sbits() instead
516 * of skipping it produces bit-exact output. This has been reported
517 * to the TAK author. */
518 *decoded
++ = get_sbits(gb
,
519 avctx
->bits_per_coded_sample
-
520 s
->sample_shift
[chan
]);
521 s
->lpc_mode
[chan
] = get_bits(gb
, 2);
522 nb_subframes
= get_bits(gb
, 3) + 1;
525 if (nb_subframes
> 1) {
526 if (get_bits_left(gb
) < (nb_subframes
- 1) * 6)
527 return AVERROR_INVALIDDATA
;
530 for (; i
< nb_subframes
- 1; i
++) {
531 int subframe_end
= get_bits(gb
, 6) * s
->subframe_scale
;
532 if (subframe_end
<= prev
)
533 return AVERROR_INVALIDDATA
;
534 subframe_len
[i
] = subframe_end
- prev
;
535 left
-= subframe_len
[i
];
540 return AVERROR_INVALIDDATA
;
542 subframe_len
[i
] = left
;
545 for (i
= 0; i
< nb_subframes
; i
++) {
546 if ((ret
= decode_subframe(s
, decoded
, subframe_len
[i
], prev
)) < 0)
548 decoded
+= subframe_len
[i
];
549 prev
= subframe_len
[i
];
555 static int decorrelate(TAKDecContext
*s
, int c1
, int c2
, int length
)
557 GetBitContext
*gb
= &s
->gb
;
558 int32_t *p1
= s
->decoded
[c1
] + 1;
559 int32_t *p2
= s
->decoded
[c2
] + 1;
564 case 1: /* left/side */
565 for (i
= 0; i
< length
; i
++) {
571 case 2: /* side/right */
572 for (i
= 0; i
< length
; i
++) {
578 case 3: /* side/mid */
579 for (i
= 0; i
< length
; i
++) {
587 case 4: /* side/left with scale factor */
588 FFSWAP(int32_t*, p1
, p2
);
589 case 5: /* side/right with scale factor */
590 dshift
= get_bits_esc4(gb
);
591 dfactor
= get_sbits(gb
, 10);
592 for (i
= 0; i
< length
; i
++) {
595 b
= dfactor
* (b
>> dshift
) + 128 >> 8 << dshift
;
600 FFSWAP(int32_t*, p1
, p2
);
602 LOCAL_ALIGNED_16(int16_t, filter
, [MAX_PREDICTORS
]);
603 int length2
, order_half
, filter_order
, dval1
, dval2
;
604 int av_uninit(code_size
);
606 memset(filter
, 0, MAX_PREDICTORS
* sizeof(*filter
));
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 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
++) {
643 av_fast_malloc(&s
->residues
, &s
->residues_buf_size
,
644 FFALIGN(length
+ 16, 16) * sizeof(*s
->residues
));
646 return AVERROR(ENOMEM
);
647 memset(s
->residues
, 0, s
->residues_buf_size
);
649 for (i
= 0; i
< length
; i
++)
650 s
->residues
[i
] = p2
[i
] >> dshift
;
654 for (i
= 0; i
< length2
; i
++) {
657 v
+= s
->dsp
.scalarproduct_int16(&s
->residues
[i
], filter
,
658 FFALIGN(filter_order
, 16));
660 p1
[i
] = (av_clip(v
>> 10, -8192, 8191) << dshift
) - p1
[i
];
671 static int tak_decode_frame(AVCodecContext
*avctx
, void *data
,
672 int *got_frame_ptr
, AVPacket
*pkt
)
674 TAKDecContext
*s
= avctx
->priv_data
;
675 AVFrame
*frame
= data
;
676 GetBitContext
*gb
= &s
->gb
;
677 int chan
, i
, ret
, hsize
;
679 if (pkt
->size
< TAK_MIN_FRAME_HEADER_BYTES
)
680 return AVERROR_INVALIDDATA
;
682 init_get_bits(gb
, pkt
->data
, pkt
->size
* 8);
684 if ((ret
= ff_tak_decode_frame_header(avctx
, gb
, &s
->ti
, 0)) < 0)
687 if (s
->ti
.flags
& TAK_FRAME_FLAG_HAS_METADATA
) {
688 av_log_missing_feature(avctx
, "frame metadata", 1);
689 return AVERROR_PATCHWELCOME
;
692 hsize
= get_bits_count(gb
) / 8;
693 if (avctx
->err_recognition
& AV_EF_CRCCHECK
) {
694 if (ff_tak_check_crc(pkt
->data
, hsize
)) {
695 av_log(avctx
, AV_LOG_ERROR
, "CRC error\n");
696 return AVERROR_INVALIDDATA
;
700 if (s
->ti
.codec
!= TAK_CODEC_MONO_STEREO
&&
701 s
->ti
.codec
!= TAK_CODEC_MULTICHANNEL
) {
702 av_log(avctx
, AV_LOG_ERROR
, "unsupported codec: %d\n", 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 if (s
->ti
.bps
!= avctx
->bits_per_coded_sample
) {
727 avctx
->bits_per_coded_sample
= s
->ti
.bps
;
728 if ((ret
= set_bps_params(avctx
)) < 0)
731 if (s
->ti
.sample_rate
!= avctx
->sample_rate
) {
732 avctx
->sample_rate
= s
->ti
.sample_rate
;
733 set_sample_rate_params(avctx
);
736 avctx
->channel_layout
= s
->ti
.ch_layout
;
737 avctx
->channels
= s
->ti
.channels
;
739 s
->nb_samples
= s
->ti
.last_frame_samples
? s
->ti
.last_frame_samples
740 : s
->ti
.frame_samples
;
742 frame
->nb_samples
= s
->nb_samples
;
743 if ((ret
= ff_get_buffer(avctx
, frame
, 0)) < 0)
746 if (avctx
->bits_per_coded_sample
<= 16) {
747 int buf_size
= av_samples_get_buffer_size(NULL
, avctx
->channels
,
749 AV_SAMPLE_FMT_S32P
, 0);
750 av_fast_malloc(&s
->decode_buffer
, &s
->decode_buffer_size
, buf_size
);
751 if (!s
->decode_buffer
)
752 return AVERROR(ENOMEM
);
753 ret
= av_samples_fill_arrays((uint8_t **)s
->decoded
, NULL
,
754 s
->decode_buffer
, avctx
->channels
,
755 s
->nb_samples
, AV_SAMPLE_FMT_S32P
, 0);
759 for (chan
= 0; chan
< avctx
->channels
; chan
++)
760 s
->decoded
[chan
] = (int32_t *)frame
->extended_data
[chan
];
763 if (s
->nb_samples
< 16) {
764 for (chan
= 0; chan
< avctx
->channels
; chan
++) {
765 int32_t *decoded
= s
->decoded
[chan
];
766 for (i
= 0; i
< s
->nb_samples
; i
++)
767 decoded
[i
] = get_sbits(gb
, avctx
->bits_per_coded_sample
);
770 if (s
->ti
.codec
== TAK_CODEC_MONO_STEREO
) {
771 for (chan
= 0; chan
< avctx
->channels
; chan
++)
772 if (ret
= decode_channel(s
, chan
))
775 if (avctx
->channels
== 2) {
777 // some kind of subframe length, but it seems to be unused
781 s
->dmode
= get_bits(gb
, 3);
782 if (ret
= decorrelate(s
, 0, 1, s
->nb_samples
- 1))
785 } else if (s
->ti
.codec
== TAK_CODEC_MULTICHANNEL
) {
789 chan
= get_bits(gb
, 4) + 1;
790 if (chan
> avctx
->channels
)
791 return AVERROR_INVALIDDATA
;
793 for (i
= 0; i
< chan
; i
++) {
794 int nbit
= get_bits(gb
, 4);
796 if (nbit
>= avctx
->channels
)
797 return AVERROR_INVALIDDATA
;
799 if (ch_mask
& 1 << nbit
)
800 return AVERROR_INVALIDDATA
;
802 s
->mcdparams
[i
].present
= get_bits1(gb
);
803 if (s
->mcdparams
[i
].present
) {
804 s
->mcdparams
[i
].index
= get_bits(gb
, 2);
805 s
->mcdparams
[i
].chan2
= get_bits(gb
, 4);
806 if (s
->mcdparams
[i
].index
== 1) {
807 if ((nbit
== s
->mcdparams
[i
].chan2
) ||
808 (ch_mask
& 1 << s
->mcdparams
[i
].chan2
))
809 return AVERROR_INVALIDDATA
;
811 ch_mask
|= 1 << s
->mcdparams
[i
].chan2
;
812 } else if (!(ch_mask
& 1 << s
->mcdparams
[i
].chan2
)) {
813 return AVERROR_INVALIDDATA
;
816 s
->mcdparams
[i
].chan1
= nbit
;
818 ch_mask
|= 1 << nbit
;
821 chan
= avctx
->channels
;
822 for (i
= 0; i
< chan
; i
++) {
823 s
->mcdparams
[i
].present
= 0;
824 s
->mcdparams
[i
].chan1
= i
;
828 for (i
= 0; i
< chan
; i
++) {
829 if (s
->mcdparams
[i
].present
&& s
->mcdparams
[i
].index
== 1)
830 if (ret
= decode_channel(s
, s
->mcdparams
[i
].chan2
))
833 if (ret
= decode_channel(s
, s
->mcdparams
[i
].chan1
))
836 if (s
->mcdparams
[i
].present
) {
837 s
->dmode
= mc_dmodes
[s
->mcdparams
[i
].index
];
838 if (ret
= decorrelate(s
,
839 s
->mcdparams
[i
].chan2
,
840 s
->mcdparams
[i
].chan1
,
847 for (chan
= 0; chan
< avctx
->channels
; chan
++) {
848 int32_t *decoded
= s
->decoded
[chan
];
850 if (s
->lpc_mode
[chan
])
851 decode_lpc(decoded
, s
->lpc_mode
[chan
], s
->nb_samples
);
853 if (s
->sample_shift
[chan
] > 0)
854 for (i
= 0; i
< s
->nb_samples
; i
++)
855 decoded
[i
] <<= s
->sample_shift
[chan
];
861 if (get_bits_left(gb
) < 0)
862 av_log(avctx
, AV_LOG_DEBUG
, "overread\n");
863 else if (get_bits_left(gb
) > 0)
864 av_log(avctx
, AV_LOG_DEBUG
, "underread\n");
866 if (avctx
->err_recognition
& AV_EF_CRCCHECK
) {
867 if (ff_tak_check_crc(pkt
->data
+ hsize
,
868 get_bits_count(gb
) / 8 - hsize
)) {
869 av_log(avctx
, AV_LOG_ERROR
, "CRC error\n");
870 return AVERROR_INVALIDDATA
;
874 /* convert to output buffer */
875 switch (avctx
->sample_fmt
) {
876 case AV_SAMPLE_FMT_U8P
:
877 for (chan
= 0; chan
< avctx
->channels
; chan
++) {
878 uint8_t *samples
= (uint8_t *)frame
->extended_data
[chan
];
879 int32_t *decoded
= s
->decoded
[chan
];
880 for (i
= 0; i
< s
->nb_samples
; i
++)
881 samples
[i
] = decoded
[i
] + 0x80;
884 case AV_SAMPLE_FMT_S16P
:
885 for (chan
= 0; chan
< avctx
->channels
; chan
++) {
886 int16_t *samples
= (int16_t *)frame
->extended_data
[chan
];
887 int32_t *decoded
= s
->decoded
[chan
];
888 for (i
= 0; i
< s
->nb_samples
; i
++)
889 samples
[i
] = decoded
[i
];
892 case AV_SAMPLE_FMT_S32P
:
893 for (chan
= 0; chan
< avctx
->channels
; chan
++) {
894 int32_t *samples
= (int32_t *)frame
->extended_data
[chan
];
895 for (i
= 0; i
< s
->nb_samples
; i
++)
906 static av_cold
int tak_decode_close(AVCodecContext
*avctx
)
908 TAKDecContext
*s
= avctx
->priv_data
;
910 av_freep(&s
->decode_buffer
);
911 av_freep(&s
->residues
);
916 AVCodec ff_tak_decoder
= {
918 .type
= AVMEDIA_TYPE_AUDIO
,
919 .id
= AV_CODEC_ID_TAK
,
920 .priv_data_size
= sizeof(TAKDecContext
),
921 .init
= tak_decode_init
,
922 .init_static_data
= tak_init_static_data
,
923 .close
= tak_decode_close
,
924 .decode
= tak_decode_frame
,
925 .capabilities
= CODEC_CAP_DR1
,
926 .long_name
= NULL_IF_CONFIG_SMALL("TAK (Tom's lossless Audio Kompressor)"),
927 .sample_fmts
= (const enum AVSampleFormat
[]) { AV_SAMPLE_FMT_U8P
,
930 AV_SAMPLE_FMT_NONE
},