1 /*
2 * MPEG-4 ALS decoder
3 * Copyright (c) 2009 Thilo Borgmann <thilo.borgmann _at_ mail.de>
4 *
5 * This file is part of FFmpeg.
6 *
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.
11 *
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.
16 *
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
20 */
21
22 /**
23 * @file
24 * MPEG-4 ALS decoder
25 * @author Thilo Borgmann <thilo.borgmann _at_ mail.de>
26 */
27
28 #include <inttypes.h>
29
40
41 #include <stdint.h>
42
43 /** Rice parameters and corresponding index offsets for decoding the
44 * indices of scaled PARCOR values. The table chosen is set globally
45 * by the encoder and stored in ALSSpecificConfig.
46 */
48 { {-52, 4}, {-29, 5}, {-31, 4}, { 19, 4}, {-16, 4},
49 { 12, 3}, { -7, 3}, { 9, 3}, { -5, 3}, { 6, 3},
50 { -4, 3}, { 3, 3}, { -3, 2}, { 3, 2}, { -2, 2},
51 { 3, 2}, { -1, 2}, { 2, 2}, { -1, 2}, { 2, 2} },
52 { {-58, 3}, {-42, 4}, {-46, 4}, { 37, 5}, {-36, 4},
53 { 29, 4}, {-29, 4}, { 25, 4}, {-23, 4}, { 20, 4},
54 {-17, 4}, { 16, 4}, {-12, 4}, { 12, 3}, {-10, 4},
55 { 7, 3}, { -4, 4}, { 3, 3}, { -1, 3}, { 1, 3} },
56 { {-59, 3}, {-45, 5}, {-50, 4}, { 38, 4}, {-39, 4},
57 { 32, 4}, {-30, 4}, { 25, 3}, {-23, 3}, { 20, 3},
58 {-20, 3}, { 16, 3}, {-13, 3}, { 10, 3}, { -7, 3},
59 { 3, 3}, { 0, 3}, { -1, 3}, { 2, 3}, { -1, 2} }
60 };
61
62
63 /** Scaled PARCOR values used for the first two PARCOR coefficients.
64 * To be indexed by the Rice coded indices.
65 * Generated by: parcor_scaled_values[i] = 32 + ((i * (i+1)) << 7) - (1 << 20)
66 * Actual values are divided by 32 in order to be stored in 16 bits.
67 */
69 -1048544 / 32, -1048288 / 32, -1047776 / 32, -1047008 / 32,
70 -1045984 / 32, -1044704 / 32, -1043168 / 32, -1041376 / 32,
71 -1039328 / 32, -1037024 / 32, -1034464 / 32, -1031648 / 32,
72 -1028576 / 32, -1025248 / 32, -1021664 / 32, -1017824 / 32,
73 -1013728 / 32, -1009376 / 32, -1004768 / 32, -999904 / 32,
74 -994784 / 32, -989408 / 32, -983776 / 32, -977888 / 32,
75 -971744 / 32, -965344 / 32, -958688 / 32, -951776 / 32,
76 -944608 / 32, -937184 / 32, -929504 / 32, -921568 / 32,
77 -913376 / 32, -904928 / 32, -896224 / 32, -887264 / 32,
78 -878048 / 32, -868576 / 32, -858848 / 32, -848864 / 32,
79 -838624 / 32, -828128 / 32, -817376 / 32, -806368 / 32,
80 -795104 / 32, -783584 / 32, -771808 / 32, -759776 / 32,
81 -747488 / 32, -734944 / 32, -722144 / 32, -709088 / 32,
82 -695776 / 32, -682208 / 32, -668384 / 32, -654304 / 32,
83 -639968 / 32, -625376 / 32, -610528 / 32, -595424 / 32,
84 -580064 / 32, -564448 / 32, -548576 / 32, -532448 / 32,
85 -516064 / 32, -499424 / 32, -482528 / 32, -465376 / 32,
86 -447968 / 32, -430304 / 32, -412384 / 32, -394208 / 32,
87 -375776 / 32, -357088 / 32, -338144 / 32, -318944 / 32,
88 -299488 / 32, -279776 / 32, -259808 / 32, -239584 / 32,
89 -219104 / 32, -198368 / 32, -177376 / 32, -156128 / 32,
90 -134624 / 32, -112864 / 32, -90848 / 32, -68576 / 32,
91 -46048 / 32, -23264 / 32, -224 / 32, 23072 / 32,
92 46624 / 32, 70432 / 32, 94496 / 32, 118816 / 32,
93 143392 / 32, 168224 / 32, 193312 / 32, 218656 / 32,
94 244256 / 32, 270112 / 32, 296224 / 32, 322592 / 32,
95 349216 / 32, 376096 / 32, 403232 / 32, 430624 / 32,
96 458272 / 32, 486176 / 32, 514336 / 32, 542752 / 32,
97 571424 / 32, 600352 / 32, 629536 / 32, 658976 / 32,
98 688672 / 32, 718624 / 32, 748832 / 32, 779296 / 32,
99 810016 / 32, 840992 / 32, 872224 / 32, 903712 / 32,
100 935456 / 32, 967456 / 32, 999712 / 32, 1032224 / 32
101 };
102
103
104 /** Gain values of p(0) for long-term prediction.
105 * To be indexed by the Rice coded indices.
106 */
108 { 0, 8, 16, 24},
109 {32, 40, 48, 56},
110 {64, 70, 76, 82},
111 {88, 92, 96, 100}
112 };
113
114
115 /** Inter-channel weighting factors for multi-channel correlation.
116 * To be indexed by the Rice coded indices.
117 */
119 204, 192, 179, 166, 153, 140, 128, 115,
120 102, 89, 76, 64, 51, 38, 25, 12,
121 0, -12, -25, -38, -51, -64, -76, -89,
122 -102, -115, -128, -140, -153, -166, -179, -192
123 };
124
125
126 /** Tail codes used in arithmetic coding using block Gilbert-Moore codes.
127 */
129 { 74, 44, 25, 13, 7, 3},
130 { 68, 42, 24, 13, 7, 3},
131 { 58, 39, 23, 13, 7, 3},
132 {126, 70, 37, 19, 10, 5},
133 {132, 70, 37, 20, 10, 5},
134 {124, 70, 38, 20, 10, 5},
135 {120, 69, 37, 20, 11, 5},
136 {116, 67, 37, 20, 11, 5},
137 {108, 66, 36, 20, 10, 5},
138 {102, 62, 36, 20, 10, 5},
139 { 88, 58, 34, 19, 10, 5},
140 {162, 89, 49, 25, 13, 7},
141 {156, 87, 49, 26, 14, 7},
142 {150, 86, 47, 26, 14, 7},
143 {142, 84, 47, 26, 14, 7},
144 {131, 79, 46, 26, 14, 7}
145 };
146
147
152 };
153
154
156 uint32_t
samples;
///< number of samples, 0xFFFFFFFF if unknown
157 int resolution;
///< 000 = 8-bit; 001 = 16-bit; 010 = 24-bit; 011 = 32-bit
158 int floating;
///< 1 = IEEE 32-bit floating-point, 0 = integer
159 int msb_first;
///< 1 = original CRC calculated on big-endian system, 0 = little-endian
160 int frame_length;
///< frame length for each frame (last frame may differ)
161 int ra_distance;
///< distance between RA frames (in frames, 0...255)
168 int bgmc;
///< "Block Gilbert-Moore Code": 1 = on, 0 = off (Rice coding only)
171 int mc_coding;
///< extended inter-channel coding (multi channel coding): 1 = on, 0 = off
172 int chan_config;
///< indicates that a chan_config_info field is present
173 int chan_sort;
///< channel rearrangement: 1 = on, 0 = off
174 int rlslms;
///< use "Recursive Least Square-Least Mean Square" predictor: 1 = on, 0 = off
175 int chan_config_info;
///< mapping of channels to loudspeaker locations. Unused until setting channel configuration is implemented.
179
180
189
190
197 uint32_t
crc_org;
///< CRC value of the original input data
198 uint32_t
crc;
///< CRC value calculated from decoded data
200 unsigned int frame_id;
///< the frame ID / number of the current frame
201 unsigned int js_switch;
///< if true, joint-stereo decoding is enforced
202 unsigned int cs_switch;
///< if true, channel rearrangement is done
203 unsigned int num_blocks;
///< number of blocks used in the current frame
204 unsigned int s_max;
///< maximum Rice parameter allowed in entropy coding
208 int *
const_block;
///< contains const_block flags for all channels
209 unsigned int *
shift_lsbs;
///< contains shift_lsbs flags for all channels
210 unsigned int *
opt_order;
///< contains opt_order flags for all channels
212 int *
use_ltp;
///< contains use_ltp flags for all channels
213 int *
ltp_lag;
///< contains ltp lag values for all channels
214 int **
ltp_gain;
///< gain values for ltp 5-tap filter for a channel
218 int32_t **
lpc_cof;
///< coefficients of the direct form prediction filter for a channel
229
230
233 unsigned int ra_block;
///< if true, this is a random access block
235 int js_blocks;
///< true if this block contains a difference signal
236 unsigned int *
shift_lsbs;
///< shift of values for this block
237 unsigned int *
opt_order;
///< prediction order of this block
239 int *
use_ltp;
///< if true, long-term prediction is used
240 int *
ltp_lag;
///< lag value for long-term prediction
241 int *
ltp_gain;
///< gain values for ltp 5-tap filter
248
249
251 {
252 #ifdef DEBUG
255
274 #endif
275 }
276
277
278 /** Read an ALSSpecificConfig from a buffer into the output struct.
279 */
281 {
283 uint64_t ht_size;
284 int i, config_offset;
288 uint32_t als_id, header_size, trailer_size;
290
293
296
297 if (config_offset < 0)
299
301
304
305 // read the fixed items
311 skip_bits(&gb, 16);
// number of channels already known
332 skip_bits(&gb, 5);
// skip 5 reserved bits
334
335
336 // check for ALSSpecificConfig struct
337 if (als_id !=
MKBETAG(
'A',
'L',
'S',
'0円'))
339
341
342 // read channel config
345 // TODO: use this to set avctx->channel_layout
346
347
348 // read channel sorting
350 int chan_pos_bits = av_ceil_log2(avctx->
channels);
351 int bits_needed = avctx->
channels * chan_pos_bits + 7;
354
357
359
360 for (i = 0; i < avctx->
channels; i++) {
361 int idx;
362
367 break;
368 }
370 }
371
373 }
374
375
376 // read fixed header and trailer sizes,
377 // if size = 0xFFFFFFFF then there is no data field!
380
383 if (header_size == 0xFFFFFFFF)
384 header_size = 0;
385 if (trailer_size == 0xFFFFFFFF)
386 trailer_size = 0;
387
388 ht_size = ((int64_t)(header_size) + (int64_t)(trailer_size)) << 3;
389
390
391 // skip the header and trailer data
394
395 if (ht_size > INT32_MAX)
397
399
400
401 // initialize CRC calculation
405
408 ctx->
crc = 0xFFFFFFFF;
410 } else
412 }
413
414
415 // no need to read the rest of ALSSpecificConfig (ra_unit_size & aux data)
416
418
419 return 0;
420 }
421
422
423 /** Check the ALSSpecificConfig for unsupported features.
424 */
426 {
428 int error = 0;
429
430 // report unsupported feature and set error value
431 #define MISSING_ERR(cond, str, errval) \
432 { \
433 if (cond) { \
434 avpriv_report_missing_feature(ctx->avctx, \
435 str); \
436 error = errval; \
437 } \
438 }
439
442
443 return error;
444 }
445
446
447 /** Parse the bs_info field to extract the block partitioning used in
448 * block switching mode, refer to ISO/IEC 14496-3, section 11.6.2.
449 */
451 unsigned int div, unsigned int **div_blocks,
452 unsigned int *num_blocks)
453 {
454 if (n < 31 && ((bs_info << n) & 0x40000000)) {
455 // if the level is valid and the investigated bit n is set
456 // then recursively check both children at bits (2n+1) and (2n+2)
457 n *= 2;
458 div += 1;
461 } else {
462 // else the bit is not set or the last level has been reached
463 // (bit implicitly not set)
464 **div_blocks = div;
465 (*div_blocks)++;
466 (*num_blocks)++;
467 }
468 }
469
470
471 /** Read and decode a Rice codeword.
472 */
474 {
478
479 if (k > 1) {
480 q <<= (k - 1);
482 } else if (!k) {
483 q >>= 1;
484 }
485 return r ? q : ~q;
486 }
487
488
489 /** Convert PARCOR coefficient k to direct filter coefficient.
490 */
492 {
493 int i, j;
494
495 for (i = 0, j = k - 1; i < j; i++, j--) {
496 int tmp1 = ((
MUL64(par[k], cof[j]) + (1 << 19)) >> 20);
497 cof[j] += ((
MUL64(par[k], cof[i]) + (1 << 19)) >> 20);
498 cof[i] += tmp1;
499 }
500 if (i == j)
501 cof[i] += ((
MUL64(par[k], cof[j]) + (1 << 19)) >> 20);
502
503 cof[k] = par[k];
504 }
505
506
507 /** Read block switching field if necessary and set actual block sizes.
508 * Also assure that the block sizes of the last frame correspond to the
509 * actual number of samples.
510 */
512 uint32_t *bs_info)
513 {
516 unsigned int *ptr_div_blocks = div_blocks;
518
522 *bs_info <<= (32 - bs_info_len);
523 }
524
527
528 // The last frame may have an overdetermined block structure given in
529 // the bitstream. In that case the defined block structure would need
530 // more samples than available to be consistent.
531 // The block structure is actually used but the block sizes are adapted
532 // to fit the actual number of available samples.
533 // Example: 5 samples, 2nd level block sizes: 2 2 2 2.
534 // This results in the actual block sizes: 2 2 1 0.
535 // This is not specified in 14496-3 but actually done by the reference
536 // codec RM22 revision 2.
537 // This appears to happen in case of an odd number of samples in the last
538 // frame which is actually not allowed by the block length switching part
539 // of 14496-3.
540 // The ALS conformance files feature an odd number of samples in the last
541 // frame.
542
545
548
550 if (remaining <= div_blocks[b]) {
551 div_blocks[
b] = remaining;
553 break;
554 }
555
556 remaining -= div_blocks[
b];
557 }
558 }
559 }
560
561
562 /** Read the block data for a constant block
563 */
565 {
569
572
576
577 // skip 5 reserved bits
579
583 }
584
585 // ensure constant block decoding by reusing this field
587
588 return 0;
589 }
590
591
592 /** Decode the block data for a constant block
593 */
595 {
599
600 // write raw samples into buffer
601 for (; smp; smp--)
602 *dst++ = val;
603 }
604
605
606 /** Read the block data for a non-constant block
607 */
609 {
613 unsigned int k;
615 unsigned int sx[8];
616 unsigned int sub_blocks, log2_sub_blocks, sb_length;
617 unsigned int start = 0;
618 unsigned int opt_order;
619 int sb;
622
623
624 // ensure variable block decoding by reusing this field
626
629
631
632 // determine the number of subblocks for entropy decoding
634 log2_sub_blocks = 0;
635 } else {
638 else
640 }
641
642 sub_blocks = 1 << log2_sub_blocks;
643
644 // do not continue in case of a damaged stream since
645 // block_length must be evenly divisible by sub_blocks
648 "Block length is not evenly divisible by the number of subblocks.\n");
650 }
651
653
656 for (k = 1; k < sub_blocks; k++)
658
659 for (k = 0; k < sub_blocks; k++) {
660 sx[k] = s[k] & 0x0F;
661 s [k] >>= 4;
662 }
663 } else {
665 for (k = 1; k < sub_blocks; k++)
667 }
668 for (k = 1; k < sub_blocks; k++)
669 if (s[k] > 32) {
672 }
673
676
678
679
682 int opt_order_length = av_ceil_log2(av_clip((bd->
block_length >> 3) - 1,
689 }
690 } else {
692 }
697 }
699
700 if (opt_order) {
701 int add_base;
702
704 add_base = 0x7F;
705
706 // read coefficient 0
708
709 // read coefficient 1
710 if (opt_order > 1)
712
713 // read coefficients 2 to opt_order
714 for (k = 2; k < opt_order; k++)
716 } else {
717 int k_max;
718 add_base = 1;
719
720 // read coefficient 0 to 19
721 k_max =
FFMIN(opt_order, 20);
722 for (k = 0; k < k_max; k++) {
726 if (quant_cof[k] < -64 || quant_cof[k] > 63) {
728 "quant_cof %"PRIu32" is out of range.\n",
729 quant_cof[k]);
731 }
732 }
733
734 // read coefficients 20 to 126
735 k_max =
FFMIN(opt_order, 127);
736 for (; k < k_max; k++)
738
739 // read coefficients 127 to opt_order
740 for (; k < opt_order; k++)
742
744
745 if (opt_order > 1)
747 }
748
749 for (k = 2; k < opt_order; k++)
750 quant_cof[k] = (quant_cof[k] << 14) + (add_base << 13);
751 }
752 }
753
754 // read LTP gain and lag values
757
760
763
767
770
773 }
774 }
775
776 // read first value and residuals in case of a random access block
778 if (opt_order)
780 if (opt_order > 1)
782 if (opt_order > 2)
784
785 start =
FFMIN(opt_order, 3);
786 }
787
788 // read all residuals
791 unsigned int k [8];
792 unsigned int b = av_clip((av_ceil_log2(bd->
block_length) - 3) >> 1, 0, 5);
793
794 // read most significant bits
795 unsigned int high;
796 unsigned int low;
798
800
802
803 for (sb = 0; sb < sub_blocks; sb++) {
804 unsigned int sb_len = sb_length - (sb ? 0 :
start);
805
806 k [sb] = s[sb] > b ? s[sb] - b : 0;
807 delta[sb] = 5 - s[sb] + k[sb];
808
811
812 current_res += sb_len;
813 }
814
816
817
818 // read least significant bits and tails
820
821 for (sb = 0; sb < sub_blocks; sb++, start = 0) {
822 unsigned int cur_tail_code =
tail_code[sx[sb]][delta[sb]];
823 unsigned int cur_k = k[sb];
824 unsigned int cur_s = s[sb];
825
826 for (; start < sb_length; start++) {
828
829 if (res == cur_tail_code) {
830 unsigned int max_msb = (2 + (sx[sb] > 2) + (sx[sb] > 10))
831 << (5 - delta[sb]);
832
834
835 if (res >= 0) {
836 res += (max_msb ) << cur_k;
837 } else {
838 res -= (max_msb - 1) << cur_k;
839 }
840 } else {
841 if (res > cur_tail_code)
842 res--;
843
844 if (res & 1)
845 res = -res;
846
847 res >>= 1;
848
849 if (cur_k) {
850 res <<= cur_k;
852 }
853 }
854
855 *current_res++ = res;
856 }
857 }
858 } else {
860
861 for (sb = 0; sb < sub_blocks; sb++, start = 0)
862 for (; start < sb_length; start++)
864 }
865
868
869 return 0;
870 }
871
872
873 /** Decode the block data for a non-constant block
874 */
876 {
879 unsigned int smp = 0;
880 unsigned int k;
882 int sb;
889
890 // reverse long-term prediction
892 int ltp_smp;
893
894 for (ltp_smp =
FFMAX(*bd->
ltp_lag - 2, 0); ltp_smp < block_length; ltp_smp++) {
895 int center = ltp_smp - *bd->
ltp_lag;
896 int begin =
FFMAX(0, center - 2);
897 int end = center + 3;
898 int tab = 5 - (end - begin);
899 int base;
900
901 y = 1 << 6;
902
903 for (base = begin; base <
end; base++, tab++)
905
906 raw_samples[ltp_smp] += y >> 7;
907 }
908 }
909
910 // reconstruct all samples from residuals
912 for (smp = 0; smp < opt_order; smp++) {
913 y = 1 << 19;
914
915 for (sb = 0; sb < smp; sb++)
916 y +=
MUL64(lpc_cof[sb], raw_samples[-(sb + 1)]);
917
918 *raw_samples++ -= y >> 20;
920 }
921 } else {
922 for (k = 0; k < opt_order; k++)
924
925 // store previous samples in case that they have to be altered
929
930 // reconstruct difference signal for prediction (joint-stereo)
933
934 if (bd->
raw_other > raw_samples) {
// D = R - L
935 left = raw_samples;
937 } else { // D = R - L
939 right = raw_samples;
940 }
941
942 for (sb = -1; sb >= -sconf->
max_order; sb--)
943 raw_samples[sb] = right[sb] - left[sb];
944 }
945
946 // reconstruct shifted signal
948 for (sb = -1; sb >= -sconf->
max_order; sb--)
950 }
951
952 // reverse linear prediction coefficients for efficiency
953 lpc_cof = lpc_cof + opt_order;
954
955 for (sb = 0; sb < opt_order; sb++)
956 lpc_cof_reversed[sb] = lpc_cof[-(sb + 1)];
957
958 // reconstruct raw samples
960 lpc_cof = lpc_cof_reversed + opt_order;
961
962 for (; raw_samples < raw_samples_end; raw_samples++) {
963 y = 1 << 19;
964
965 for (sb = -opt_order; sb < 0; sb++)
966 y +=
MUL64(lpc_cof[sb], raw_samples[sb]);
967
968 *raw_samples -= y >> 20;
969 }
970
972
973 // restore previous samples in case that they have been altered
976 sizeof(*raw_samples) * sconf->
max_order);
977
978 return 0;
979 }
980
981
982 /** Read the block data.
983 */
985 {
988
990 // read block type flag and read the samples accordingly
993 } else {
995 }
996
998 }
999
1000
1001 /** Decode the block data.
1002 */
1004 {
1005 unsigned int smp;
1007
1008 // read block type flag and read the samples accordingly
1011 else
1013
1014 if (ret < 0)
1016
1017 // TODO: read RLSLMS extension data
1018
1022
1023 return 0;
1024 }
1025
1026
1027 /** Read and decode block data successively.
1028 */
1030 {
1032
1035
1037 }
1038
1039
1040 /** Compute the number of samples left to decode for the current frame and
1041 * sets these samples to zero.
1042 */
1044 const unsigned int *div_blocks,
int32_t *
buf)
1045 {
1046 unsigned int count = 0;
1047
1048 while (b < b_max)
1049 count += div_blocks[b++];
1050
1051 if (count)
1052 memset(buf, 0, sizeof(*buf) * count);
1053 }
1054
1055
1056 /** Decode blocks independently.
1057 */
1059 unsigned int c,
const unsigned int *div_blocks,
1060 unsigned int *js_blocks)
1061 {
1065
1078
1079
1082
1084 // damaged block, write zero for the rest of the frame
1087 }
1090 }
1091
1092 return 0;
1093 }
1094
1095
1096 /** Decode blocks dependently.
1097 */
1099 unsigned int c,
const unsigned int *div_blocks,
1100 unsigned int *js_blocks)
1101 {
1107
1120
1133
1134 // decode all blocks
1137
1140
1143
1146
1149 goto fail;
1150
1151 // reconstruct joint-stereo blocks
1152 if (bd[0].js_blocks) {
1153 if (bd[1].js_blocks)
1155
1156 for (s = 0; s < div_blocks[
b]; s++)
1157 bd[0].raw_samples[s] = bd[1].raw_samples[s] - bd[0].raw_samples[s];
1158 } else if (bd[1].js_blocks) {
1159 for (s = 0; s < div_blocks[
b]; s++)
1160 bd[1].raw_samples[s] = bd[1].raw_samples[s] + bd[0].raw_samples[s];
1161 }
1162
1163 offset += div_blocks[
b];
1166 }
1167
1168 // store carryover raw samples,
1169 // the others channel raw samples are stored by the calling function.
1173
1174 return 0;
1175 fail:
1176 // damaged block, write zero for the rest of the frame
1180 }
1181
1183 {
1187 }
1188
1189 /** Read the channel data.
1190 */
1192 {
1196 int entries = 0;
1197
1200
1204 }
1205
1211
1216
1219 }
1220 }
1221
1222 current++;
1223 entries++;
1224 }
1225
1226 if (entries == channels) {
1229 }
1230
1232 return 0;
1233 }
1234
1235
1236 /** Recursively reverts the inter-channel correlation for a block.
1237 */
1241 {
1243 unsigned int dep = 0;
1245
1246 if (reverted[c])
1247 return 0;
1248
1250
1251 while (dep < channels && !ch[dep].stop_flag) {
1253 ch[dep].master_channel);
1254
1255 dep++;
1256 }
1257
1258 if (dep == channels) {
1261 }
1262
1273
1274 for (dep = 0; !ch[dep].
stop_flag; dep++) {
1275 unsigned int smp;
1276 unsigned int begin = 1;
1280
1281 if (ch[dep].master_channel == c)
1282 continue;
1283
1284 if (ch[dep].time_diff_flag) {
1286
1287 if (ch[dep].time_diff_sign) {
1288 t = -t;
1289 begin -= t;
1290 } else {
1291 end -= t;
1292 }
1293
1294 for (smp = begin; smp <
end; smp++) {
1295 y = (1 << 6) +
1296 MUL64(ch[dep].weighting[0], master[smp - 1 ]) +
1297 MUL64(ch[dep].weighting[1], master[smp ]) +
1298 MUL64(ch[dep].weighting[2], master[smp + 1 ]) +
1299 MUL64(ch[dep].weighting[3], master[smp - 1 + t]) +
1300 MUL64(ch[dep].weighting[4], master[smp + t]) +
1301 MUL64(ch[dep].weighting[5], master[smp + 1 + t]);
1302
1304 }
1305 } else {
1306 for (smp = begin; smp <
end; smp++) {
1307 y = (1 << 6) +
1308 MUL64(ch[dep].weighting[0], master[smp - 1]) +
1309 MUL64(ch[dep].weighting[1], master[smp ]) +
1310 MUL64(ch[dep].weighting[2], master[smp + 1]);
1311
1313 }
1314 }
1315 }
1316
1317 return 0;
1318 }
1319
1320
1321 /** Read the frame data.
1322 */
1324 {
1328 unsigned int div_blocks[32]; ///< block sizes.
1330 unsigned int js_blocks[2];
1331 uint32_t bs_info = 0;
1333
1334 // skip the size of the ra unit if present in the frame
1337
1341 }
1342
1345
1346 for (c = 0; c < avctx->
channels; c++) {
1347 js_blocks[0] = 0;
1348 js_blocks[1] = 0;
1349
1351
1352 // if joint_stereo and block_switching is set, independent decoding
1353 // is signaled via the first bit of bs_info
1355 if (bs_info >> 31)
1356 independent_bs = 2;
1357
1358 // if this is the last channel, it has to be decoded independently
1360 independent_bs = 1;
1361
1362 if (independent_bs) {
1364 div_blocks, js_blocks);
1365 if (ret < 0)
1367 independent_bs--;
1368 } else {
1369 ret =
decode_blocks(ctx, ra_frame, c, div_blocks, js_blocks);
1370 if (ret < 0)
1372
1373 c++;
1374 }
1375
1376 // store carryover raw samples
1380 }
1381 } else { // multi-channel coding
1386
1387 for (c = 0; c < avctx->
channels; c++)
1391 }
1392
1393 memset(reverted_channels, 0,
sizeof(*reverted_channels) * avctx->
channels);
1394
1397
1399
1404 "Invalid block length %u in channel data!\n",
1406 continue;
1407 }
1408
1409 for (c = 0; c < avctx->
channels; c++) {
1421
1426 }
1427
1428 for (c = 0; c < avctx->
channels; c++) {
1430 reverted_channels, offset, c);
1431 if (ret < 0)
1433 }
1434 for (c = 0; c < avctx->
channels; c++) {
1445
1448 }
1449
1450 memset(reverted_channels, 0, avctx->
channels *
sizeof(*reverted_channels));
1451 offset += div_blocks[
b];
1453 }
1454
1455 // store carryover raw samples
1456 for (c = 0; c < avctx->
channels; c++)
1460 }
1461
1462 // TODO: read_diff_float_data
1463
1464 return 0;
1465 }
1466
1467
1468 /** Decode an ALS frame.
1469 */
1472 {
1477 int buffer_size = avpkt->
size;
1478 int invalid_frame,
ret;
1480
1483
1484 // In the case that the distance between random access frames is set to zero
1485 // (sconf->ra_distance == 0) no frame is treated as a random access frame.
1486 // For the first frame, if prediction is used, all samples used from the
1487 // previous frame are assumed to be zero.
1489
1490 // the last frame to decode might have a different length
1491 if (sconf->
samples != 0xFFFFFFFF)
1494 else
1496
1497 // decode the frame data
1500 "Reading frame data failed. Skipping RA unit.\n");
1501
1503
1504 /* get output buffer */
1508
1509 // transform decoded frame into output format
1510 #define INTERLEAVE_OUTPUT(bps) \
1511 { \
1512 int##bps##_t *dest = (int##bps##_t*)frame->data[0]; \
1513 shift = bps - ctx->avctx->bits_per_raw_sample; \
1514 if (!ctx->cs_switch) { \
1515 for (sample = 0; sample < ctx->cur_frame_length; sample++) \
1516 for (c = 0; c < avctx->channels; c++) \
1517 *dest++ = ctx->raw_samples[c][sample] << shift; \
1518 } else { \
1519 for (sample = 0; sample < ctx->cur_frame_length; sample++) \
1520 for (c = 0; c < avctx->channels; c++) \
1521 *dest++ = ctx->raw_samples[sconf->chan_pos[c]][sample] << shift; \
1522 } \
1523 }
1524
1527 } else {
1529 }
1530
1531 // update CRC
1533 int swap = HAVE_BIGENDIAN != sconf->
msb_first;
1534
1537
1540 sample++) {
1542
1543 if (swap)
1545 else
1547 if (!HAVE_BIGENDIAN)
1548 v >>= 8;
1549
1551 }
1552 } else {
1554
1555 if (swap) {
1557 int16_t *
src = (int16_t*) frame->
data[0];
1559 for (sample = 0;
1561 sample++)
1563 } else {
1565 (uint32_t *) frame->
data[0],
1567 }
1569 } else {
1570 crc_source = frame->
data[0];
1571 }
1572
1576 }
1577
1578
1579 // check CRC sums if this is the last frame
1585 }
1586 }
1587
1588 *got_frame_ptr = 1;
1589
1590 bytes_read = invalid_frame ? buffer_size :
1592
1593 return bytes_read;
1594 }
1595
1596
1597 /** Uninitialize the ALS decoder.
1598 */
1600 {
1602
1604
1606
1627
1628 return 0;
1629 }
1630
1631
1632 /** Initialize the ALS decoder.
1633 */
1635 {
1637 unsigned int channel_size;
1638 int num_buffers,
ret;
1642
1646 }
1647
1650 goto fail;
1651 }
1652
1654 goto fail;
1655 }
1656
1659 if (ret < 0)
1660 goto fail;
1661 }
1665 } else {
1669 }
1670
1671 // set maximum Rice parameter for progressive decoding based on resolution
1672 // This is not specified in 14496-3 but actually done by the reference
1673 // codec RM22 revision 2.
1675
1676 // set lag value for long-term prediction
1679
1680 // allocate quantized parcor coefficient buffer
1682
1691
1697 goto fail;
1698 }
1699
1700 // assign quantized parcor coefficient buffers
1701 for (c = 0; c < num_buffers; c++) {
1704 }
1705
1706 // allocate and assign lag and gain data buffer for ltp mode
1715 num_buffers * 5);
1716
1723 goto fail;
1724 }
1725
1726 for (c = 0; c < num_buffers; c++)
1728
1729 // allocate and assign channel data buffer for mcc mode
1732 num_buffers * num_buffers);
1734 num_buffers);
1736 num_buffers);
1737
1741 goto fail;
1742 }
1743
1744 for (c = 0; c < num_buffers; c++)
1746 } else {
1750 }
1751
1753
1757
1758 // allocate previous raw sample buffer
1762 goto fail;
1763 }
1764
1765 // assign raw samples buffers
1767 for (c = 1; c < avctx->
channels; c++)
1769
1770 // allocate crc buffer
1780 goto fail;
1781 }
1782 }
1783
1785
1786 return 0;
1787
1788 fail:
1791 }
1792
1793
1794 /** Flush (reset) the frame ID after seeking.
1795 */
1797 {
1799
1801 }
1802
1803
1815 };