2 * WavPack lossless audio decoder
3 * Copyright (c) 2006,2011 Konstantin Shishkov
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
22 #define BITSTREAM_READER_LE
24 #include "libavutil/channel_layout.h"
32 * WavPack lossless audio decoder
35 #define WV_MONO 0x00000004
36 #define WV_JOINT_STEREO 0x00000010
37 #define WV_FALSE_STEREO 0x40000000
39 #define WV_HYBRID_MODE 0x00000008
40 #define WV_HYBRID_SHAPE 0x00000008
41 #define WV_HYBRID_BITRATE 0x00000200
42 #define WV_HYBRID_BALANCE 0x00000400
44 #define WV_FLT_SHIFT_ONES 0x01
45 #define WV_FLT_SHIFT_SAME 0x02
46 #define WV_FLT_SHIFT_SENT 0x04
47 #define WV_FLT_ZERO_SENT 0x08
48 #define WV_FLT_ZERO_SIGN 0x10
74 typedef struct SavedContext {
83 typedef struct Decorr {
92 typedef struct WvChannel {
94 int slow_level, error_limit;
95 int bitrate_acc, bitrate_delta;
98 typedef struct WavpackFrameContext {
99 AVCodecContext *avctx;
101 int stereo, stereo_in;
106 uint32_t crc_extra_bits;
107 GetBitContext gb_extra_bits;
108 int data_size; // in bits
111 Decorr decorr[MAX_TERMS];
112 int zero, one, zeroes;
116 int hybrid, hybrid_bitrate;
117 int hybrid_maxclip, hybrid_minclip;
123 SavedContext sc, extra_sc;
124 } WavpackFrameContext;
126 #define WV_MAX_FRAME_DECODERS 14
128 typedef struct WavpackContext {
129 AVCodecContext *avctx;
131 WavpackFrameContext *fdec[WV_MAX_FRAME_DECODERS];
141 // exponent table copied from WavPack source
142 static const uint8_t wp_exp2_table [256] = {
143 0x00, 0x01, 0x01, 0x02, 0x03, 0x03, 0x04, 0x05, 0x06, 0x06, 0x07, 0x08, 0x08, 0x09, 0x0a, 0x0b,
144 0x0b, 0x0c, 0x0d, 0x0e, 0x0e, 0x0f, 0x10, 0x10, 0x11, 0x12, 0x13, 0x13, 0x14, 0x15, 0x16, 0x16,
145 0x17, 0x18, 0x19, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1d, 0x1e, 0x1f, 0x20, 0x20, 0x21, 0x22, 0x23,
146 0x24, 0x24, 0x25, 0x26, 0x27, 0x28, 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2c, 0x2d, 0x2e, 0x2f, 0x30,
147 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3a, 0x3b, 0x3c, 0x3d,
148 0x3e, 0x3f, 0x40, 0x41, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x48, 0x49, 0x4a, 0x4b,
149 0x4c, 0x4d, 0x4e, 0x4f, 0x50, 0x51, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a,
150 0x5b, 0x5c, 0x5d, 0x5e, 0x5e, 0x5f, 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
151 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
152 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x87, 0x88, 0x89, 0x8a,
153 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0x90, 0x91, 0x92, 0x93, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9a, 0x9b,
154 0x9c, 0x9d, 0x9f, 0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad,
155 0xaf, 0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xbc, 0xbd, 0xbe, 0xbf, 0xc0,
156 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc8, 0xc9, 0xca, 0xcb, 0xcd, 0xce, 0xcf, 0xd0, 0xd2, 0xd3, 0xd4,
157 0xd6, 0xd7, 0xd8, 0xd9, 0xdb, 0xdc, 0xdd, 0xde, 0xe0, 0xe1, 0xe2, 0xe4, 0xe5, 0xe6, 0xe8, 0xe9,
158 0xea, 0xec, 0xed, 0xee, 0xf0, 0xf1, 0xf2, 0xf4, 0xf5, 0xf6, 0xf8, 0xf9, 0xfa, 0xfc, 0xfd, 0xff
161 static const uint8_t wp_log2_table [] = {
162 0x00, 0x01, 0x03, 0x04, 0x06, 0x07, 0x09, 0x0a, 0x0b, 0x0d, 0x0e, 0x10, 0x11, 0x12, 0x14, 0x15,
163 0x16, 0x18, 0x19, 0x1a, 0x1c, 0x1d, 0x1e, 0x20, 0x21, 0x22, 0x24, 0x25, 0x26, 0x28, 0x29, 0x2a,
164 0x2c, 0x2d, 0x2e, 0x2f, 0x31, 0x32, 0x33, 0x34, 0x36, 0x37, 0x38, 0x39, 0x3b, 0x3c, 0x3d, 0x3e,
165 0x3f, 0x41, 0x42, 0x43, 0x44, 0x45, 0x47, 0x48, 0x49, 0x4a, 0x4b, 0x4d, 0x4e, 0x4f, 0x50, 0x51,
166 0x52, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a, 0x5c, 0x5d, 0x5e, 0x5f, 0x60, 0x61, 0x62, 0x63,
167 0x64, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 0x70, 0x71, 0x72, 0x74, 0x75,
168 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x80, 0x81, 0x82, 0x83, 0x84, 0x85,
169 0x86, 0x87, 0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0x90, 0x91, 0x92, 0x93, 0x94, 0x95,
170 0x96, 0x97, 0x98, 0x99, 0x9a, 0x9b, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0xa0, 0xa1, 0xa2, 0xa3, 0xa4,
171 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf, 0xb0, 0xb1, 0xb2, 0xb2,
172 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf, 0xc0, 0xc0,
173 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xcb, 0xcb, 0xcc, 0xcd, 0xce,
174 0xcf, 0xd0, 0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd8, 0xd9, 0xda, 0xdb,
175 0xdc, 0xdc, 0xdd, 0xde, 0xdf, 0xe0, 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe4, 0xe5, 0xe6, 0xe7, 0xe7,
176 0xe8, 0xe9, 0xea, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xee, 0xef, 0xf0, 0xf1, 0xf1, 0xf2, 0xf3, 0xf4,
177 0xf4, 0xf5, 0xf6, 0xf7, 0xf7, 0xf8, 0xf9, 0xf9, 0xfa, 0xfb, 0xfc, 0xfc, 0xfd, 0xfe, 0xff, 0xff
180 static av_always_inline int wp_exp2(int16_t val)
189 res = wp_exp2_table[val & 0xFF] | 0x100;
191 res = (val > 9) ? (res << (val - 9)) : (res >> (9 - val));
192 return neg ? -res : res;
195 static av_always_inline int wp_log2(int32_t val)
204 bits = av_log2(val) + 1;
206 return (bits << 8) + wp_log2_table[(val << (9 - bits)) & 0xFF];
208 return (bits << 8) + wp_log2_table[(val >> (bits - 9)) & 0xFF];
211 #define LEVEL_DECAY(a) ((a + 0x80) >> 8)
213 // macros for manipulating median values
214 #define GET_MED(n) ((c->median[n] >> 4) + 1)
215 #define DEC_MED(n) c->median[n] -= ((c->median[n] + (128 >> n) - 2) / (128 >> n)) * 2
216 #define INC_MED(n) c->median[n] += ((c->median[n] + (128 >> n) ) / (128 >> n)) * 5
218 // macros for applying weight
219 #define UPDATE_WEIGHT_CLIP(weight, delta, samples, in) \
220 if (samples && in) { \
221 if ((samples ^ in) < 0) { \
223 if (weight < -1024) \
233 static av_always_inline int get_tail(GetBitContext *gb, int k)
240 e = (1 << (p + 1)) - k - 1;
241 res = p ? get_bits(gb, p) : 0;
243 res = (res << 1) - e + get_bits1(gb);
247 static void update_error_limit(WavpackFrameContext *ctx)
251 for (i = 0; i <= ctx->stereo_in; i++) {
252 ctx->ch[i].bitrate_acc += ctx->ch[i].bitrate_delta;
253 br[i] = ctx->ch[i].bitrate_acc >> 16;
254 sl[i] = LEVEL_DECAY(ctx->ch[i].slow_level);
256 if (ctx->stereo_in && ctx->hybrid_bitrate) {
257 int balance = (sl[1] - sl[0] + br[1] + 1) >> 1;
258 if (balance > br[0]) {
261 } else if (-balance > br[0]) {
265 br[1] = br[0] + balance;
266 br[0] = br[0] - balance;
269 for (i = 0; i <= ctx->stereo_in; i++) {
270 if (ctx->hybrid_bitrate) {
271 if (sl[i] - br[i] > -0x100)
272 ctx->ch[i].error_limit = wp_exp2(sl[i] - br[i] + 0x100);
274 ctx->ch[i].error_limit = 0;
276 ctx->ch[i].error_limit = wp_exp2(br[i]);
281 static int wv_get_value(WavpackFrameContext *ctx, GetBitContext *gb,
282 int channel, int *last)
285 int sign, base, add, ret;
286 WvChannel *c = &ctx->ch[channel];
290 if ((ctx->ch[0].median[0] < 2U) && (ctx->ch[1].median[0] < 2U) &&
291 !ctx->zero && !ctx->one) {
295 c->slow_level -= LEVEL_DECAY(c->slow_level);
299 t = get_unary_0_33(gb);
301 if (get_bits_left(gb) < t - 1)
303 t = get_bits(gb, t - 1) | (1 << (t-1));
305 if (get_bits_left(gb) < 0)
310 memset(ctx->ch[0].median, 0, sizeof(ctx->ch[0].median));
311 memset(ctx->ch[1].median, 0, sizeof(ctx->ch[1].median));
312 c->slow_level -= LEVEL_DECAY(c->slow_level);
322 t = get_unary_0_33(gb);
323 if (get_bits_left(gb) < 0)
326 t2 = get_unary_0_33(gb);
328 if (get_bits_left(gb) < 0)
332 if (get_bits_left(gb) < t2 - 1)
334 t += get_bits(gb, t2 - 1) | (1 << (t2 - 1));
345 ctx->zero = !ctx->one;
348 if (ctx->hybrid && !channel)
349 update_error_limit(ctx);
353 add = GET_MED(0) - 1;
357 add = GET_MED(1) - 1;
361 base = GET_MED(0) + GET_MED(1);
362 add = GET_MED(2) - 1;
367 base = GET_MED(0) + GET_MED(1) + GET_MED(2) * (t - 2);
368 add = GET_MED(2) - 1;
373 if (!c->error_limit) {
374 ret = base + get_tail(gb, add);
375 if (get_bits_left(gb) <= 0)
378 int mid = (base * 2 + add + 1) >> 1;
379 while (add > c->error_limit) {
380 if (get_bits_left(gb) <= 0)
386 add = mid - base - 1;
387 mid = (base * 2 + add + 1) >> 1;
391 sign = get_bits1(gb);
392 if (ctx->hybrid_bitrate)
393 c->slow_level += wp_log2(ret) - LEVEL_DECAY(c->slow_level);
394 return sign ? ~ret : ret;
401 static inline int wv_get_value_integer(WavpackFrameContext *s, uint32_t *crc,
409 if (s->got_extra_bits && get_bits_left(&s->gb_extra_bits) >= s->extra_bits) {
410 S |= get_bits(&s->gb_extra_bits, s->extra_bits);
411 *crc = *crc * 9 + (S & 0xffff) * 3 + ((unsigned)S >> 16);
415 bit = (S & s->and) | s->or;
416 bit = ((S + bit) << s->shift) - bit;
419 bit = av_clip(bit, s->hybrid_minclip, s->hybrid_maxclip);
421 return bit << s->post_shift;
424 static float wv_get_value_float(WavpackFrameContext *s, uint32_t *crc, int S)
432 int exp = s->float_max_exp;
434 if (s->got_extra_bits) {
435 const int max_bits = 1 + 23 + 8 + 1;
436 const int left_bits = get_bits_left(&s->gb_extra_bits);
438 if (left_bits + 8 * FF_INPUT_BUFFER_PADDING_SIZE < max_bits)
443 S <<= s->float_shift;
447 if (S >= 0x1000000) {
448 if (s->got_extra_bits && get_bits1(&s->gb_extra_bits))
449 S = get_bits(&s->gb_extra_bits, 23);
454 int shift = 23 - av_log2(S);
455 exp = s->float_max_exp;
462 if ((s->float_flag & WV_FLT_SHIFT_ONES) ||
463 (s->got_extra_bits && (s->float_flag & WV_FLT_SHIFT_SAME) &&
464 get_bits1(&s->gb_extra_bits))) {
465 S |= (1 << shift) - 1;
466 } else if (s->got_extra_bits &&
467 (s->float_flag & WV_FLT_SHIFT_SENT)) {
468 S |= get_bits(&s->gb_extra_bits, shift);
472 exp = s->float_max_exp;
478 if (s->got_extra_bits && (s->float_flag & WV_FLT_ZERO_SENT)) {
479 if (get_bits1(&s->gb_extra_bits)) {
480 S = get_bits(&s->gb_extra_bits, 23);
481 if (s->float_max_exp >= 25)
482 exp = get_bits(&s->gb_extra_bits, 8);
483 sign = get_bits1(&s->gb_extra_bits);
485 if (s->float_flag & WV_FLT_ZERO_SIGN)
486 sign = get_bits1(&s->gb_extra_bits);
491 *crc = *crc * 27 + S * 9 + exp * 3 + sign;
493 value.u = (sign << 31) | (exp << 23) | S;
497 static void wv_reset_saved_context(WavpackFrameContext *s)
500 s->sc.crc = s->extra_sc.crc = 0xFFFFFFFF;
503 static inline int wv_check_crc(WavpackFrameContext *s, uint32_t crc,
504 uint32_t crc_extra_bits)
507 av_log(s->avctx, AV_LOG_ERROR, "CRC error\n");
508 return AVERROR_INVALIDDATA;
510 if (s->got_extra_bits && crc_extra_bits != s->crc_extra_bits) {
511 av_log(s->avctx, AV_LOG_ERROR, "Extra bits CRC error\n");
512 return AVERROR_INVALIDDATA;
518 static inline int wv_unpack_stereo(WavpackFrameContext *s, GetBitContext *gb,
519 void *dst, const int type)
523 int A, B, L, L2, R, R2;
525 uint32_t crc = s->sc.crc;
526 uint32_t crc_extra_bits = s->extra_sc.crc;
527 int16_t *dst16 = dst;
528 int32_t *dst32 = dst;
530 const int channel_pad = s->avctx->channels - 2;
532 s->one = s->zero = s->zeroes = 0;
534 L = wv_get_value(s, gb, 0, &last);
537 R = wv_get_value(s, gb, 1, &last);
540 for (i = 0; i < s->terms; i++) {
541 t = s->decorr[i].value;
545 A = 2 * s->decorr[i].samplesA[0] - s->decorr[i].samplesA[1];
546 B = 2 * s->decorr[i].samplesB[0] - s->decorr[i].samplesB[1];
548 A = (3 * s->decorr[i].samplesA[0] - s->decorr[i].samplesA[1]) >> 1;
549 B = (3 * s->decorr[i].samplesB[0] - s->decorr[i].samplesB[1]) >> 1;
551 s->decorr[i].samplesA[1] = s->decorr[i].samplesA[0];
552 s->decorr[i].samplesB[1] = s->decorr[i].samplesB[0];
555 A = s->decorr[i].samplesA[pos];
556 B = s->decorr[i].samplesB[pos];
559 if (type != AV_SAMPLE_FMT_S16) {
560 L2 = L + ((s->decorr[i].weightA * (int64_t)A + 512) >> 10);
561 R2 = R + ((s->decorr[i].weightB * (int64_t)B + 512) >> 10);
563 L2 = L + ((s->decorr[i].weightA * A + 512) >> 10);
564 R2 = R + ((s->decorr[i].weightB * B + 512) >> 10);
566 if (A && L) s->decorr[i].weightA -= ((((L ^ A) >> 30) & 2) - 1) * s->decorr[i].delta;
567 if (B && R) s->decorr[i].weightB -= ((((R ^ B) >> 30) & 2) - 1) * s->decorr[i].delta;
568 s->decorr[i].samplesA[j] = L = L2;
569 s->decorr[i].samplesB[j] = R = R2;
570 } else if (t == -1) {
571 if (type != AV_SAMPLE_FMT_S16)
572 L2 = L + ((s->decorr[i].weightA * (int64_t)s->decorr[i].samplesA[0] + 512) >> 10);
574 L2 = L + ((s->decorr[i].weightA * s->decorr[i].samplesA[0] + 512) >> 10);
575 UPDATE_WEIGHT_CLIP(s->decorr[i].weightA, s->decorr[i].delta, s->decorr[i].samplesA[0], L);
577 if (type != AV_SAMPLE_FMT_S16)
578 R2 = R + ((s->decorr[i].weightB * (int64_t)L2 + 512) >> 10);
580 R2 = R + ((s->decorr[i].weightB * L2 + 512) >> 10);
581 UPDATE_WEIGHT_CLIP(s->decorr[i].weightB, s->decorr[i].delta, L2, R);
583 s->decorr[i].samplesA[0] = R;
585 if (type != AV_SAMPLE_FMT_S16)
586 R2 = R + ((s->decorr[i].weightB * (int64_t)s->decorr[i].samplesB[0] + 512) >> 10);
588 R2 = R + ((s->decorr[i].weightB * s->decorr[i].samplesB[0] + 512) >> 10);
589 UPDATE_WEIGHT_CLIP(s->decorr[i].weightB, s->decorr[i].delta, s->decorr[i].samplesB[0], R);
593 R2 = s->decorr[i].samplesA[0];
594 s->decorr[i].samplesA[0] = R;
597 if (type != AV_SAMPLE_FMT_S16)
598 L2 = L + ((s->decorr[i].weightA * (int64_t)R2 + 512) >> 10);
600 L2 = L + ((s->decorr[i].weightA * R2 + 512) >> 10);
601 UPDATE_WEIGHT_CLIP(s->decorr[i].weightA, s->decorr[i].delta, R2, L);
603 s->decorr[i].samplesB[0] = L;
608 L += (R -= (L >> 1));
609 crc = (crc * 3 + L) * 3 + R;
611 if (type == AV_SAMPLE_FMT_FLT) {
612 *dstfl++ = wv_get_value_float(s, &crc_extra_bits, L);
613 *dstfl++ = wv_get_value_float(s, &crc_extra_bits, R);
614 dstfl += channel_pad;
615 } else if (type == AV_SAMPLE_FMT_S32) {
616 *dst32++ = wv_get_value_integer(s, &crc_extra_bits, L);
617 *dst32++ = wv_get_value_integer(s, &crc_extra_bits, R);
618 dst32 += channel_pad;
620 *dst16++ = wv_get_value_integer(s, &crc_extra_bits, L);
621 *dst16++ = wv_get_value_integer(s, &crc_extra_bits, R);
622 dst16 += channel_pad;
625 } while (!last && count < s->samples);
627 wv_reset_saved_context(s);
628 if ((s->avctx->err_recognition & AV_EF_CRCCHECK) &&
629 wv_check_crc(s, crc, crc_extra_bits))
630 return AVERROR_INVALIDDATA;
635 static inline int wv_unpack_mono(WavpackFrameContext *s, GetBitContext *gb,
636 void *dst, const int type)
642 uint32_t crc = s->sc.crc;
643 uint32_t crc_extra_bits = s->extra_sc.crc;
644 int16_t *dst16 = dst;
645 int32_t *dst32 = dst;
647 const int channel_stride = s->avctx->channels;
649 s->one = s->zero = s->zeroes = 0;
651 T = wv_get_value(s, gb, 0, &last);
655 for (i = 0; i < s->terms; i++) {
656 t = s->decorr[i].value;
659 A = 2 * s->decorr[i].samplesA[0] - s->decorr[i].samplesA[1];
661 A = (3 * s->decorr[i].samplesA[0] - s->decorr[i].samplesA[1]) >> 1;
662 s->decorr[i].samplesA[1] = s->decorr[i].samplesA[0];
665 A = s->decorr[i].samplesA[pos];
668 if (type != AV_SAMPLE_FMT_S16)
669 S = T + ((s->decorr[i].weightA * (int64_t)A + 512) >> 10);
671 S = T + ((s->decorr[i].weightA * A + 512) >> 10);
673 s->decorr[i].weightA -= ((((T ^ A) >> 30) & 2) - 1) * s->decorr[i].delta;
674 s->decorr[i].samplesA[j] = T = S;
679 if (type == AV_SAMPLE_FMT_FLT) {
680 *dstfl = wv_get_value_float(s, &crc_extra_bits, S);
681 dstfl += channel_stride;
682 } else if (type == AV_SAMPLE_FMT_S32) {
683 *dst32 = wv_get_value_integer(s, &crc_extra_bits, S);
684 dst32 += channel_stride;
686 *dst16 = wv_get_value_integer(s, &crc_extra_bits, S);
687 dst16 += channel_stride;
690 } while (!last && count < s->samples);
692 wv_reset_saved_context(s);
693 if ((s->avctx->err_recognition & AV_EF_CRCCHECK) &&
694 wv_check_crc(s, crc, crc_extra_bits))
695 return AVERROR_INVALIDDATA;
700 static av_cold int wv_alloc_frame_context(WavpackContext *c)
703 if (c->fdec_num == WV_MAX_FRAME_DECODERS)
706 c->fdec[c->fdec_num] = av_mallocz(sizeof(**c->fdec));
707 if (!c->fdec[c->fdec_num])
710 c->fdec[c->fdec_num - 1]->avctx = c->avctx;
711 wv_reset_saved_context(c->fdec[c->fdec_num - 1]);
716 static av_cold int wavpack_decode_init(AVCodecContext *avctx)
718 WavpackContext *s = avctx->priv_data;
721 if (avctx->bits_per_coded_sample <= 16)
722 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
724 avctx->sample_fmt = AV_SAMPLE_FMT_S32;
725 if (avctx->channels <= 2 && !avctx->channel_layout)
726 avctx->channel_layout = (avctx->channels == 2) ? AV_CH_LAYOUT_STEREO :
729 s->multichannel = avctx->channels > 2;
730 /* lavf demuxer does not provide extradata, Matroska stores 0x403
731 there, use this to detect decoding mode for multichannel */
733 if (s->multichannel && avctx->extradata && avctx->extradata_size == 2) {
734 int ver = AV_RL16(avctx->extradata);
735 if (ver >= 0x402 && ver <= 0x410)
744 static av_cold int wavpack_decode_end(AVCodecContext *avctx)
746 WavpackContext *s = avctx->priv_data;
749 for (i = 0; i < s->fdec_num; i++)
750 av_freep(&s->fdec[i]);
756 static int wavpack_decode_block(AVCodecContext *avctx, int block_no,
757 void *data, int *got_frame_ptr,
758 const uint8_t *buf, int buf_size)
760 WavpackContext *wc = avctx->priv_data;
761 WavpackFrameContext *s;
762 void *samples = data;
764 int got_terms = 0, got_weights = 0, got_samples = 0,
765 got_entropy = 0, got_bs = 0, got_float = 0, got_hybrid = 0;
766 const uint8_t *orig_buf = buf;
767 const uint8_t *buf_end = buf + buf_size;
768 int i, j, id, size, ssize, weights, t;
769 int bpp, chan, chmask, orig_bpp;
776 if (block_no >= wc->fdec_num && wv_alloc_frame_context(wc) < 0) {
777 av_log(avctx, AV_LOG_ERROR, "Error creating frame decode context\n");
781 s = wc->fdec[block_no];
783 av_log(avctx, AV_LOG_ERROR, "Context for block %d is not present\n", block_no);
787 memset(s->decorr, 0, MAX_TERMS * sizeof(Decorr));
788 memset(s->ch, 0, sizeof(s->ch));
790 s->and = s->or = s->shift = 0;
791 s->got_extra_bits = 0;
794 s->samples = AV_RL32(buf); buf += 4;
800 s->samples = wc->samples;
802 s->frame_flags = AV_RL32(buf); buf += 4;
803 bpp = av_get_bytes_per_sample(avctx->sample_fmt);
804 samples = (uint8_t*)samples + bpp * wc->ch_offset;
805 orig_bpp = ((s->frame_flags & 0x03) + 1) << 3;
807 s->stereo = !(s->frame_flags & WV_MONO);
808 s->stereo_in = (s->frame_flags & WV_FALSE_STEREO) ? 0 : s->stereo;
809 s->joint = s->frame_flags & WV_JOINT_STEREO;
810 s->hybrid = s->frame_flags & WV_HYBRID_MODE;
811 s->hybrid_bitrate = s->frame_flags & WV_HYBRID_BITRATE;
812 s->post_shift = bpp * 8 - orig_bpp + ((s->frame_flags >> 13) & 0x1f);
813 s->hybrid_maxclip = (( 1LL << (orig_bpp - 1)) - 1);
814 s->hybrid_minclip = ((-1LL << (orig_bpp - 1)));
815 s->CRC = AV_RL32(buf); buf += 4;
817 buf += 4; //skip block size;
819 wc->ch_offset += 1 + s->stereo;
821 // parse metadata blocks
822 while (buf < buf_end) {
825 if (id & WP_IDF_LONG) {
826 size |= (*buf++) << 8;
827 size |= (*buf++) << 16;
829 size <<= 1; // size is specified in words
834 av_log(avctx, AV_LOG_ERROR, "Got incorrect block %02X with size %i\n", id, size);
837 if (buf + ssize > buf_end) {
838 av_log(avctx, AV_LOG_ERROR, "Block size %i is out of bounds\n", size);
841 if (id & WP_IDF_IGNORE) {
845 switch (id & WP_IDF_MASK) {
847 if (size > MAX_TERMS) {
848 av_log(avctx, AV_LOG_ERROR, "Too many decorrelation terms\n");
854 for (i = 0; i < s->terms; i++) {
855 s->decorr[s->terms - i - 1].value = (*buf & 0x1F) - 5;
856 s->decorr[s->terms - i - 1].delta = *buf >> 5;
861 case WP_ID_DECWEIGHTS:
863 av_log(avctx, AV_LOG_ERROR, "No decorrelation terms met\n");
866 weights = size >> s->stereo_in;
867 if (weights > MAX_TERMS || weights > s->terms) {
868 av_log(avctx, AV_LOG_ERROR, "Too many decorrelation weights\n");
872 for (i = 0; i < weights; i++) {
873 t = (int8_t)(*buf++);
874 s->decorr[s->terms - i - 1].weightA = t << 3;
875 if (s->decorr[s->terms - i - 1].weightA > 0)
876 s->decorr[s->terms - i - 1].weightA +=
877 (s->decorr[s->terms - i - 1].weightA + 64) >> 7;
879 t = (int8_t)(*buf++);
880 s->decorr[s->terms - i - 1].weightB = t << 3;
881 if (s->decorr[s->terms - i - 1].weightB > 0)
882 s->decorr[s->terms - i - 1].weightB +=
883 (s->decorr[s->terms - i - 1].weightB + 64) >> 7;
888 case WP_ID_DECSAMPLES:
890 av_log(avctx, AV_LOG_ERROR, "No decorrelation terms met\n");
894 for (i = s->terms - 1; (i >= 0) && (t < size); i--) {
895 if (s->decorr[i].value > 8) {
896 s->decorr[i].samplesA[0] = wp_exp2(AV_RL16(buf)); buf += 2;
897 s->decorr[i].samplesA[1] = wp_exp2(AV_RL16(buf)); buf += 2;
899 s->decorr[i].samplesB[0] = wp_exp2(AV_RL16(buf)); buf += 2;
900 s->decorr[i].samplesB[1] = wp_exp2(AV_RL16(buf)); buf += 2;
904 } else if (s->decorr[i].value < 0) {
905 s->decorr[i].samplesA[0] = wp_exp2(AV_RL16(buf)); buf += 2;
906 s->decorr[i].samplesB[0] = wp_exp2(AV_RL16(buf)); buf += 2;
909 for (j = 0; j < s->decorr[i].value; j++) {
910 s->decorr[i].samplesA[j] = wp_exp2(AV_RL16(buf)); buf += 2;
912 s->decorr[i].samplesB[j] = wp_exp2(AV_RL16(buf)); buf += 2;
915 t += s->decorr[i].value * 2 * (s->stereo_in + 1);
921 if (size != 6 * (s->stereo_in + 1)) {
922 av_log(avctx, AV_LOG_ERROR, "Entropy vars size should be %i, "
923 "got %i", 6 * (s->stereo_in + 1), size);
927 for (j = 0; j <= s->stereo_in; j++) {
928 for (i = 0; i < 3; i++) {
929 s->ch[j].median[i] = wp_exp2(AV_RL16(buf));
936 if (s->hybrid_bitrate) {
937 for (i = 0; i <= s->stereo_in; i++) {
938 s->ch[i].slow_level = wp_exp2(AV_RL16(buf));
943 for (i = 0; i < (s->stereo_in + 1); i++) {
944 s->ch[i].bitrate_acc = AV_RL16(buf) << 16;
949 for (i = 0; i < (s->stereo_in + 1); i++) {
950 s->ch[i].bitrate_delta = wp_exp2((int16_t)AV_RL16(buf));
954 for (i = 0; i < (s->stereo_in + 1); i++)
955 s->ch[i].bitrate_delta = 0;
959 case WP_ID_INT32INFO:
961 av_log(avctx, AV_LOG_ERROR, "Invalid INT32INFO, size = %i, sent_bits = %i\n", size, *buf);
966 s->extra_bits = buf[0];
976 /* original WavPack decoder forces 32-bit lossy sound to be treated
977 * as 24-bit one in order to have proper clipping
979 if (s->hybrid && bpp == 4 && s->post_shift < 8 && s->shift > 8) {
982 s->hybrid_maxclip >>= 8;
983 s->hybrid_minclip >>= 8;
987 case WP_ID_FLOATINFO:
989 av_log(avctx, AV_LOG_ERROR, "Invalid FLOATINFO, size = %i\n", size);
993 s->float_flag = buf[0];
994 s->float_shift = buf[1];
995 s->float_max_exp = buf[2];
1000 s->sc.offset = buf - orig_buf;
1001 s->sc.size = size * 8;
1002 init_get_bits(&s->gb, buf, size * 8);
1003 s->data_size = size * 8;
1007 case WP_ID_EXTRABITS:
1009 av_log(avctx, AV_LOG_ERROR, "Invalid EXTRABITS, size = %i\n",
1014 s->extra_sc.offset = buf - orig_buf;
1015 s->extra_sc.size = size * 8;
1016 init_get_bits(&s->gb_extra_bits, buf, size * 8);
1017 s->crc_extra_bits = get_bits_long(&s->gb_extra_bits, 32);
1019 s->got_extra_bits = 1;
1021 case WP_ID_CHANINFO:
1023 av_log(avctx, AV_LOG_ERROR, "Insufficient channel information\n");
1028 case 0: chmask = *buf; break;
1029 case 1: chmask = AV_RL16(buf); break;
1030 case 2: chmask = AV_RL24(buf); break;
1031 case 3: chmask = AV_RL32(buf); break;
1033 chan |= (buf[1] & 0xF) << 8;
1034 chmask = AV_RL24(buf + 2);
1037 av_log(avctx, AV_LOG_ERROR, "Invalid channel info size %d\n",
1039 chan = avctx->channels;
1040 chmask = avctx->channel_layout;
1042 if (chan != avctx->channels) {
1043 av_log(avctx, AV_LOG_ERROR, "Block reports total %d channels, "
1044 "decoder believes it's %d channels\n", chan,
1048 if (!avctx->channel_layout)
1049 avctx->channel_layout = chmask;
1055 if (id & WP_IDF_ODD)
1060 av_log(avctx, AV_LOG_ERROR, "No block with decorrelation terms\n");
1064 av_log(avctx, AV_LOG_ERROR, "No block with decorrelation weights\n");
1068 av_log(avctx, AV_LOG_ERROR, "No block with decorrelation samples\n");
1072 av_log(avctx, AV_LOG_ERROR, "No block with entropy info\n");
1075 if (s->hybrid && !got_hybrid) {
1076 av_log(avctx, AV_LOG_ERROR, "Hybrid config not found\n");
1080 av_log(avctx, AV_LOG_ERROR, "Packed samples not found\n");
1083 if (!got_float && avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1084 av_log(avctx, AV_LOG_ERROR, "Float information not found\n");
1087 if (s->got_extra_bits && avctx->sample_fmt != AV_SAMPLE_FMT_FLT) {
1088 const int size = get_bits_left(&s->gb_extra_bits);
1089 const int wanted = s->samples * s->extra_bits << s->stereo_in;
1090 if (size < wanted) {
1091 av_log(avctx, AV_LOG_ERROR, "Too small EXTRABITS\n");
1092 s->got_extra_bits = 0;
1097 if (avctx->sample_fmt == AV_SAMPLE_FMT_S16)
1098 samplecount = wv_unpack_stereo(s, &s->gb, samples, AV_SAMPLE_FMT_S16);
1099 else if (avctx->sample_fmt == AV_SAMPLE_FMT_S32)
1100 samplecount = wv_unpack_stereo(s, &s->gb, samples, AV_SAMPLE_FMT_S32);
1102 samplecount = wv_unpack_stereo(s, &s->gb, samples, AV_SAMPLE_FMT_FLT);
1104 if (samplecount < 0)
1109 const int channel_stride = avctx->channels;
1111 if (avctx->sample_fmt == AV_SAMPLE_FMT_S16)
1112 samplecount = wv_unpack_mono(s, &s->gb, samples, AV_SAMPLE_FMT_S16);
1113 else if (avctx->sample_fmt == AV_SAMPLE_FMT_S32)
1114 samplecount = wv_unpack_mono(s, &s->gb, samples, AV_SAMPLE_FMT_S32);
1116 samplecount = wv_unpack_mono(s, &s->gb, samples, AV_SAMPLE_FMT_FLT);
1118 if (samplecount < 0)
1121 if (s->stereo && avctx->sample_fmt == AV_SAMPLE_FMT_S16) {
1122 int16_t *dst = (int16_t*)samples + 1;
1123 int16_t *src = (int16_t*)samples;
1124 int cnt = samplecount;
1127 src += channel_stride;
1128 dst += channel_stride;
1130 } else if (s->stereo && avctx->sample_fmt == AV_SAMPLE_FMT_S32) {
1131 int32_t *dst = (int32_t*)samples + 1;
1132 int32_t *src = (int32_t*)samples;
1133 int cnt = samplecount;
1136 src += channel_stride;
1137 dst += channel_stride;
1139 } else if (s->stereo) {
1140 float *dst = (float*)samples + 1;
1141 float *src = (float*)samples;
1142 int cnt = samplecount;
1145 src += channel_stride;
1146 dst += channel_stride;
1153 return samplecount * bpp;
1156 static void wavpack_decode_flush(AVCodecContext *avctx)
1158 WavpackContext *s = avctx->priv_data;
1161 for (i = 0; i < s->fdec_num; i++)
1162 wv_reset_saved_context(s->fdec[i]);
1165 static int wavpack_decode_frame(AVCodecContext *avctx, void *data,
1166 int *got_frame_ptr, AVPacket *avpkt)
1168 WavpackContext *s = avctx->priv_data;
1169 const uint8_t *buf = avpkt->data;
1170 int buf_size = avpkt->size;
1171 AVFrame *frame = data;
1172 int frame_size, ret, frame_flags;
1173 int samplecount = 0;
1178 /* determine number of samples */
1180 s->samples = AV_RL32(buf); buf += 4;
1181 frame_flags = AV_RL32(buf);
1183 if (s->multichannel) {
1184 s->samples = AV_RL32(buf + 4);
1185 frame_flags = AV_RL32(buf + 8);
1187 s->samples = AV_RL32(buf);
1188 frame_flags = AV_RL32(buf + 4);
1191 if (s->samples <= 0) {
1192 av_log(avctx, AV_LOG_ERROR, "Invalid number of samples: %d\n",
1194 return AVERROR(EINVAL);
1197 if (frame_flags & 0x80) {
1198 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1199 } else if ((frame_flags & 0x03) <= 1) {
1200 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1202 avctx->sample_fmt = AV_SAMPLE_FMT_S32;
1203 avctx->bits_per_raw_sample = ((frame_flags & 0x03) + 1) << 3;
1206 /* get output buffer */
1207 frame->nb_samples = s->samples;
1208 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
1209 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1213 while (buf_size > 0) {
1214 if (!s->multichannel) {
1215 frame_size = buf_size;
1218 frame_size = AV_RL32(buf) - 12; buf += 4; buf_size -= 4;
1220 if (buf_size < 12) //MKV files can have zero flags after last block
1222 frame_size = AV_RL32(buf + 8) + 12;
1225 if (frame_size < 0 || frame_size > buf_size) {
1226 av_log(avctx, AV_LOG_ERROR, "Block %d has invalid size (size %d "
1227 "vs. %d bytes left)\n", s->block, frame_size, buf_size);
1228 wavpack_decode_flush(avctx);
1231 if ((samplecount = wavpack_decode_block(avctx, s->block,
1232 frame->data[0], got_frame_ptr,
1233 buf, frame_size)) < 0) {
1234 wavpack_decode_flush(avctx);
1238 buf += frame_size; buf_size -= frame_size;
1244 AVCodec ff_wavpack_decoder = {
1246 .type = AVMEDIA_TYPE_AUDIO,
1247 .id = AV_CODEC_ID_WAVPACK,
1248 .priv_data_size = sizeof(WavpackContext),
1249 .init = wavpack_decode_init,
1250 .close = wavpack_decode_end,
1251 .decode = wavpack_decode_frame,
1252 .flush = wavpack_decode_flush,
1253 .capabilities = CODEC_CAP_SUBFRAMES | CODEC_CAP_DR1,
1254 .long_name = NULL_IF_CONFIG_SMALL("WavPack"),