projects / libav.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Skip some more code that isnt needed for direct MBs.
[libav.git] / libavcodec / alsdec.c
CommitLineData
99971952
TB		 1/*
2 * MPEG-4 ALS decoder
3 * Copyright (c) 2009 Thilo Borgmann <thilo.borgmann _at_ googlemail.com>
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 libavcodec/alsdec.c
24 * MPEG-4 ALS decoder
25 * @author Thilo Borgmann <thilo.borgmann _at_ googlemail.com>
26 */
27
28
29//#define DEBUG
30
31
32#include "avcodec.h"
33#include "get_bits.h"
34#include "unary.h"
35#include "mpeg4audio.h"
36#include "bytestream.h"
37
25c4fdda
TB		 38#include <stdint.h>
39
40/** Rice parameters and corresponding index offsets for decoding the
41 * indices of scaled PARCOR values. The table choosen is set globally
42 * by the encoder and stored in ALSSpecificConfig.
43 */
44static const int8_t parcor_rice_table[3][20][2] = {
45 { {-52, 4}, {-29, 5}, {-31, 4}, { 19, 4}, {-16, 4},
46 { 12, 3}, { -7, 3}, { 9, 3}, { -5, 3}, { 6, 3},
47 { -4, 3}, { 3, 3}, { -3, 2}, { 3, 2}, { -2, 2},
48 { 3, 2}, { -1, 2}, { 2, 2}, { -1, 2}, { 2, 2} },
49 { {-58, 3}, {-42, 4}, {-46, 4}, { 37, 5}, {-36, 4},
50 { 29, 4}, {-29, 4}, { 25, 4}, {-23, 4}, { 20, 4},
51 {-17, 4}, { 16, 4}, {-12, 4}, { 12, 3}, {-10, 4},
52 { 7, 3}, { -4, 4}, { 3, 3}, { -1, 3}, { 1, 3} },
53 { {-59, 3}, {-45, 5}, {-50, 4}, { 38, 4}, {-39, 4},
54 { 32, 4}, {-30, 4}, { 25, 3}, {-23, 3}, { 20, 3},
55 {-20, 3}, { 16, 3}, {-13, 3}, { 10, 3}, { -7, 3},
56 { 3, 3}, { 0, 3}, { -1, 3}, { 2, 3}, { -1, 2} }
57};
58
59
60/** Scaled PARCOR values used for the first two PARCOR coefficients.
61 * To be indexed by the Rice coded indices.
62 * Generated by: parcor_scaled_values[i] = 32 + ((i * (i+1)) << 7) - (1 << 20)
63 * Actual values are divided by 32 in order to be stored in 16 bits.
64 */
65static const int16_t parcor_scaled_values[] = {
66 -1048544 / 32, -1048288 / 32, -1047776 / 32, -1047008 / 32,
67 -1045984 / 32, -1044704 / 32, -1043168 / 32, -1041376 / 32,
68 -1039328 / 32, -1037024 / 32, -1034464 / 32, -1031648 / 32,
69 -1028576 / 32, -1025248 / 32, -1021664 / 32, -1017824 / 32,
70 -1013728 / 32, -1009376 / 32, -1004768 / 32, -999904 / 32,
71 -994784 / 32, -989408 / 32, -983776 / 32, -977888 / 32,
72 -971744 / 32, -965344 / 32, -958688 / 32, -951776 / 32,
73 -944608 / 32, -937184 / 32, -929504 / 32, -921568 / 32,
74 -913376 / 32, -904928 / 32, -896224 / 32, -887264 / 32,
75 -878048 / 32, -868576 / 32, -858848 / 32, -848864 / 32,
76 -838624 / 32, -828128 / 32, -817376 / 32, -806368 / 32,
77 -795104 / 32, -783584 / 32, -771808 / 32, -759776 / 32,
78 -747488 / 32, -734944 / 32, -722144 / 32, -709088 / 32,
79 -695776 / 32, -682208 / 32, -668384 / 32, -654304 / 32,
80 -639968 / 32, -625376 / 32, -610528 / 32, -595424 / 32,
81 -580064 / 32, -564448 / 32, -548576 / 32, -532448 / 32,
82 -516064 / 32, -499424 / 32, -482528 / 32, -465376 / 32,
83 -447968 / 32, -430304 / 32, -412384 / 32, -394208 / 32,
84 -375776 / 32, -357088 / 32, -338144 / 32, -318944 / 32,
85 -299488 / 32, -279776 / 32, -259808 / 32, -239584 / 32,
86 -219104 / 32, -198368 / 32, -177376 / 32, -156128 / 32,
87 -134624 / 32, -112864 / 32, -90848 / 32, -68576 / 32,
88 -46048 / 32, -23264 / 32, -224 / 32, 23072 / 32,
89 46624 / 32, 70432 / 32, 94496 / 32, 118816 / 32,
90 143392 / 32, 168224 / 32, 193312 / 32, 218656 / 32,
91 244256 / 32, 270112 / 32, 296224 / 32, 322592 / 32,
92 349216 / 32, 376096 / 32, 403232 / 32, 430624 / 32,
93 458272 / 32, 486176 / 32, 514336 / 32, 542752 / 32,
94 571424 / 32, 600352 / 32, 629536 / 32, 658976 / 32,
95 688672 / 32, 718624 / 32, 748832 / 32, 779296 / 32,
96 810016 / 32, 840992 / 32, 872224 / 32, 903712 / 32,
97 935456 / 32, 967456 / 32, 999712 / 32, 1032224 / 32
98};
99
100
101/** Gain values of p(0) for long-term prediction.
102 * To be indexed by the Rice coded indices.
103 */
104static const uint8_t ltp_gain_values [4][4] = {
105 { 0, 8, 16, 24},
106 {32, 40, 48, 56},
107 {64, 70, 76, 82},
108 {88, 92, 96, 100}
109};
110
99971952 111
e38215f2
TB		 112/** Inter-channel weighting factors for multi-channel correlation.
113 * To be indexed by the Rice coded indices.
114 */
115static const int16_t mcc_weightings[] = {
116 204, 192, 179, 166, 153, 140, 128, 115,
117 102, 89, 76, 64, 51, 38, 25, 12,
118 0, -12, -25, -38, -51, -64, -76, -89,
119 -102, -115, -128, -140, -153, -166, -179, -192
120};
121
122
99971952
TB		 123enum RA_Flag {
124 RA_FLAG_NONE,
125 RA_FLAG_FRAMES,
126 RA_FLAG_HEADER
127};
128
129
130typedef struct {
131 uint32_t samples; ///< number of samples, 0xFFFFFFFF if unknown
132 int resolution; ///< 000 = 8-bit; 001 = 16-bit; 010 = 24-bit; 011 = 32-bit
133 int floating; ///< 1 = IEEE 32-bit floating-point, 0 = integer
134 int frame_length; ///< frame length for each frame (last frame may differ)
135 int ra_distance; ///< distance between RA frames (in frames, 0...255)
136 enum RA_Flag ra_flag; ///< indicates where the size of ra units is stored
137 int adapt_order; ///< adaptive order: 1 = on, 0 = off
138 int coef_table; ///< table index of Rice code parameters
139 int long_term_prediction; ///< long term prediction (LTP): 1 = on, 0 = off
140 int max_order; ///< maximum prediction order (0..1023)
141 int block_switching; ///< number of block switching levels
142 int bgmc; ///< "Block Gilbert-Moore Code": 1 = on, 0 = off (Rice coding only)
143 int sb_part; ///< sub-block partition
144 int joint_stereo; ///< joint stereo: 1 = on, 0 = off
145 int mc_coding; ///< extended inter-channel coding (multi channel coding): 1 = on, 0 = off
146 int chan_config; ///< indicates that a chan_config_info field is present
147 int chan_sort; ///< channel rearrangement: 1 = on, 0 = off
148 int rlslms; ///< use "Recursive Least Square-Least Mean Square" predictor: 1 = on, 0 = off
149 int chan_config_info; ///< mapping of channels to loudspeaker locations. Unused until setting channel configuration is implemented.
150 int *chan_pos; ///< original channel positions
99971952
TB		 151} ALSSpecificConfig;
152
153
154typedef struct {
e38215f2
TB		 155 int stop_flag;
156 int master_channel;
157 int time_diff_flag;
158 int time_diff_sign;
159 int time_diff_index;
160 int weighting[6];
161} ALSChannelData;
162
163
164typedef struct {
99971952
TB		 165 AVCodecContext *avctx;
166 ALSSpecificConfig sconf;
167 GetBitContext gb;
168 unsigned int cur_frame_length; ///< length of the current frame to decode
169 unsigned int frame_id; ///< the frame ID / number of the current frame
170 unsigned int js_switch; ///< if true, joint-stereo decoding is enforced
171 unsigned int num_blocks; ///< number of blocks used in the current frame
93d38cf6 172 int ltp_lag_length; ///< number of bits used for ltp lag value
1261b07f
TB		 173 int *use_ltp; ///< contains use_ltp flags for all channels
174 int *ltp_lag; ///< contains ltp lag values for all channels
175 int **ltp_gain; ///< gain values for ltp 5-tap filter for a channel
176 int *ltp_gain_buffer; ///< contains all gain values for ltp 5-tap filter
e38215f2
TB		 177 int32_t **quant_cof; ///< quantized parcor coefficients for a channel
178 int32_t *quant_cof_buffer; ///< contains all quantized parcor coefficients
179 int32_t **lpc_cof; ///< coefficients of the direct form prediction filter for a channel
180 int32_t *lpc_cof_buffer; ///< contains all coefficients of the direct form prediction filter
ff9ea0b7 181 int32_t *lpc_cof_reversed_buffer; ///< temporary buffer to set up a reversed versio of lpc_cof_buffer
e38215f2
TB		 182 ALSChannelData **chan_data; ///< channel data for multi-channel correlation
183 ALSChannelData *chan_data_buffer; ///< contains channel data for all channels
184 int *reverted_channels; ///< stores a flag for each reverted channel
99971952
TB		 185 int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block
186 int32_t **raw_samples; ///< decoded raw samples for each channel
187 int32_t *raw_buffer; ///< contains all decoded raw samples including carryover samples
188} ALSDecContext;
189
190
1261b07f
TB		 191typedef struct {
192 unsigned int block_length; ///< number of samples within the block
193 unsigned int ra_block; ///< if true, this is a random access block
194 int const_block; ///< if true, this is a constant value block
195 int32_t const_val; ///< the sample value of a constant block
196 int js_blocks; ///< true if this block contains a difference signal
197 unsigned int shift_lsbs; ///< shift of values for this block
198 unsigned int opt_order; ///< prediction order of this block
199 int store_prev_samples;///< if true, carryover samples have to be stored
200 int *use_ltp; ///< if true, long-term prediction is used
201 int *ltp_lag; ///< lag value for long-term prediction
202 int *ltp_gain; ///< gain values for ltp 5-tap filter
203 int32_t *quant_cof; ///< quantized parcor coefficients
204 int32_t *lpc_cof; ///< coefficients of the direct form prediction
205 int32_t *raw_samples; ///< decoded raw samples / residuals for this block
206 int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block
207 int32_t *raw_other; ///< decoded raw samples of the other channel of a channel pair
208} ALSBlockData;
209
210
99971952
TB		 211static av_cold void dprint_specific_config(ALSDecContext *ctx)
212{
213#ifdef DEBUG
214 AVCodecContext *avctx = ctx->avctx;
215 ALSSpecificConfig *sconf = &ctx->sconf;
216
217 dprintf(avctx, "resolution = %i\n", sconf->resolution);
218 dprintf(avctx, "floating = %i\n", sconf->floating);
219 dprintf(avctx, "frame_length = %i\n", sconf->frame_length);
220 dprintf(avctx, "ra_distance = %i\n", sconf->ra_distance);
221 dprintf(avctx, "ra_flag = %i\n", sconf->ra_flag);
222 dprintf(avctx, "adapt_order = %i\n", sconf->adapt_order);
223 dprintf(avctx, "coef_table = %i\n", sconf->coef_table);
224 dprintf(avctx, "long_term_prediction = %i\n", sconf->long_term_prediction);
225 dprintf(avctx, "max_order = %i\n", sconf->max_order);
226 dprintf(avctx, "block_switching = %i\n", sconf->block_switching);
227 dprintf(avctx, "bgmc = %i\n", sconf->bgmc);
228 dprintf(avctx, "sb_part = %i\n", sconf->sb_part);
229 dprintf(avctx, "joint_stereo = %i\n", sconf->joint_stereo);
230 dprintf(avctx, "mc_coding = %i\n", sconf->mc_coding);
231 dprintf(avctx, "chan_config = %i\n", sconf->chan_config);
232 dprintf(avctx, "chan_sort = %i\n", sconf->chan_sort);
233 dprintf(avctx, "RLSLMS = %i\n", sconf->rlslms);
234 dprintf(avctx, "chan_config_info = %i\n", sconf->chan_config_info);
99971952
TB		 235#endif
236}
237
238
239/** Reads an ALSSpecificConfig from a buffer into the output struct.
240 */
241static av_cold int read_specific_config(ALSDecContext *ctx)
242{
243 GetBitContext gb;
244 uint64_t ht_size;
245 int i, config_offset, crc_enabled;
246 MPEG4AudioConfig m4ac;
247 ALSSpecificConfig *sconf = &ctx->sconf;
248 AVCodecContext *avctx = ctx->avctx;
8e14fbe8 249 uint32_t als_id, header_size, trailer_size;
99971952
TB		 250
251 init_get_bits(&gb, avctx->extradata, avctx->extradata_size * 8);
252
253 config_offset = ff_mpeg4audio_get_config(&m4ac, avctx->extradata,
254 avctx->extradata_size);
255
256 if (config_offset < 0)
257 return -1;
258
259 skip_bits_long(&gb, config_offset);
260
261 if (get_bits_left(&gb) < (30 << 3))
262 return -1;
263
264 // read the fixed items
265 als_id = get_bits_long(&gb, 32);
266 avctx->sample_rate = m4ac.sample_rate;
267 skip_bits_long(&gb, 32); // sample rate already known
268 sconf->samples = get_bits_long(&gb, 32);
269 avctx->channels = m4ac.channels;
270 skip_bits(&gb, 16); // number of channels already knwon
271 skip_bits(&gb, 3); // skip file_type
272 sconf->resolution = get_bits(&gb, 3);
273 sconf->floating = get_bits1(&gb);
274 skip_bits1(&gb); // skip msb_first
275 sconf->frame_length = get_bits(&gb, 16) + 1;
276 sconf->ra_distance = get_bits(&gb, 8);
277 sconf->ra_flag = get_bits(&gb, 2);
278 sconf->adapt_order = get_bits1(&gb);
279 sconf->coef_table = get_bits(&gb, 2);
280 sconf->long_term_prediction = get_bits1(&gb);
281 sconf->max_order = get_bits(&gb, 10);
282 sconf->block_switching = get_bits(&gb, 2);
283 sconf->bgmc = get_bits1(&gb);
284 sconf->sb_part = get_bits1(&gb);
285 sconf->joint_stereo = get_bits1(&gb);
286 sconf->mc_coding = get_bits1(&gb);
287 sconf->chan_config = get_bits1(&gb);
288 sconf->chan_sort = get_bits1(&gb);
289 crc_enabled = get_bits1(&gb);
290 sconf->rlslms = get_bits1(&gb);
291 skip_bits(&gb, 5); // skip 5 reserved bits
292 skip_bits1(&gb); // skip aux_data_enabled
293
294
295 // check for ALSSpecificConfig struct
296 if (als_id != MKBETAG('A','L','S','\0'))
297 return -1;
298
299 ctx->cur_frame_length = sconf->frame_length;
300
99971952
TB		 301 // read channel config
302 if (sconf->chan_config)
303 sconf->chan_config_info = get_bits(&gb, 16);
304 // TODO: use this to set avctx->channel_layout
305
306
307 // read channel sorting
308 if (sconf->chan_sort && avctx->channels > 1) {
309 int chan_pos_bits = av_ceil_log2(avctx->channels);
310 int bits_needed = avctx->channels * chan_pos_bits + 7;
311 if (get_bits_left(&gb) < bits_needed)
312 return -1;
313
314 if (!(sconf->chan_pos = av_malloc(avctx->channels * sizeof(*sconf->chan_pos))))
315 return AVERROR(ENOMEM);
316
317 for (i = 0; i < avctx->channels; i++)
318 sconf->chan_pos[i] = get_bits(&gb, chan_pos_bits);
319
320 align_get_bits(&gb);
321 // TODO: use this to actually do channel sorting
322 } else {
323 sconf->chan_sort = 0;
324 }
325
326
327 // read fixed header and trailer sizes,
328 // if size = 0xFFFFFFFF then there is no data field!
329 if (get_bits_left(&gb) < 64)
330 return -1;
331
8e14fbe8
TB		 332 header_size = get_bits_long(&gb, 32);
333 trailer_size = get_bits_long(&gb, 32);
334 if (header_size == 0xFFFFFFFF)
335 header_size = 0;
336 if (trailer_size == 0xFFFFFFFF)
337 trailer_size = 0;
99971952 338
8e14fbe8 339 ht_size = ((int64_t)(header_size) + (int64_t)(trailer_size)) << 3;
99971952
TB		 340
341
342 // skip the header and trailer data
343 if (get_bits_left(&gb) < ht_size)
344 return -1;
345
346 if (ht_size > INT32_MAX)
347 return -1;
348
349 skip_bits_long(&gb, ht_size);
350
351
352 // skip the crc data
353 if (crc_enabled) {
354 if (get_bits_left(&gb) < 32)
355 return -1;
356
357 skip_bits_long(&gb, 32);
358 }
359
360
361 // no need to read the rest of ALSSpecificConfig (ra_unit_size & aux data)
362
363 dprint_specific_config(ctx);
364
365 return 0;
366}
367
368
369/** Checks the ALSSpecificConfig for unsupported features.
370 */
371static int check_specific_config(ALSDecContext *ctx)
372{
373 ALSSpecificConfig *sconf = &ctx->sconf;
374 int error = 0;
375
376 // report unsupported feature and set error value
377 #define MISSING_ERR(cond, str, errval) \
378 { \
379 if (cond) { \
380 av_log_missing_feature(ctx->avctx, str, 0); \
381 error = errval; \
382 } \
383 }
384
385 MISSING_ERR(sconf->floating, "Floating point decoding", -1);
99971952 386 MISSING_ERR(sconf->bgmc, "BGMC entropy decoding", -1);
99971952
TB		 387 MISSING_ERR(sconf->rlslms, "Adaptive RLS-LMS prediction", -1);
388 MISSING_ERR(sconf->chan_sort, "Channel sorting", 0);
389
390 return error;
391}
392
393
394/** Parses the bs_info field to extract the block partitioning used in
395 * block switching mode, refer to ISO/IEC 14496-3, section 11.6.2.
396 */
397static void parse_bs_info(const uint32_t bs_info, unsigned int n,
398 unsigned int div, unsigned int **div_blocks,
399 unsigned int *num_blocks)
400{
401 if (n < 31 && ((bs_info << n) & 0x40000000)) {
402 // if the level is valid and the investigated bit n is set
403 // then recursively check both children at bits (2n+1) and (2n+2)
404 n *= 2;
405 div += 1;
406 parse_bs_info(bs_info, n + 1, div, div_blocks, num_blocks);
407 parse_bs_info(bs_info, n + 2, div, div_blocks, num_blocks);
408 } else {
409 // else the bit is not set or the last level has been reached
410 // (bit implicitly not set)
411 **div_blocks = div;
412 (*div_blocks)++;
413 (*num_blocks)++;
414 }
415}
416
417
418/** Reads and decodes a Rice codeword.
419 */
420static int32_t decode_rice(GetBitContext *gb, unsigned int k)
421{
6e44ba15 422 int max = get_bits_left(gb) - k;
99971952
TB		 423 int q = get_unary(gb, 0, max);
424 int r = k ? get_bits1(gb) : !(q & 1);
425
426 if (k > 1) {
427 q <<= (k - 1);
428 q += get_bits_long(gb, k - 1);
429 } else if (!k) {
430 q >>= 1;
431 }
432 return r ? q : ~q;
433}
434
435
436/** Converts PARCOR coefficient k to direct filter coefficient.
437 */
438static void parcor_to_lpc(unsigned int k, const int32_t *par, int32_t *cof)
439{
440 int i, j;
441
442 for (i = 0, j = k - 1; i < j; i++, j--) {
443 int tmp1 = ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20);
444 cof[j] += ((MUL64(par[k], cof[i]) + (1 << 19)) >> 20);
445 cof[i] += tmp1;
446 }
447 if (i == j)
448 cof[i] += ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20);
449
450 cof[k] = par[k];
451}
452
453
454/** Reads block switching field if necessary and sets actual block sizes.
455 * Also assures that the block sizes of the last frame correspond to the
456 * actual number of samples.
457 */
458static void get_block_sizes(ALSDecContext *ctx, unsigned int *div_blocks,
459 uint32_t *bs_info)
460{
461 ALSSpecificConfig *sconf = &ctx->sconf;
462 GetBitContext *gb = &ctx->gb;
463 unsigned int *ptr_div_blocks = div_blocks;
464 unsigned int b;
465
466 if (sconf->block_switching) {
467 unsigned int bs_info_len = 1 << (sconf->block_switching + 2);
468 *bs_info = get_bits_long(gb, bs_info_len);
469 *bs_info <<= (32 - bs_info_len);
470 }
471
472 ctx->num_blocks = 0;
473 parse_bs_info(*bs_info, 0, 0, &ptr_div_blocks, &ctx->num_blocks);
474
475 // The last frame may have an overdetermined block structure given in
476 // the bitstream. In that case the defined block structure would need
477 // more samples than available to be consistent.
478 // The block structure is actually used but the block sizes are adapted
479 // to fit the actual number of available samples.
480 // Example: 5 samples, 2nd level block sizes: 2 2 2 2.
481 // This results in the actual block sizes: 2 2 1 0.
482 // This is not specified in 14496-3 but actually done by the reference
483 // codec RM22 revision 2.
484 // This appears to happen in case of an odd number of samples in the last
485 // frame which is actually not allowed by the block length switching part
486 // of 14496-3.
487 // The ALS conformance files feature an odd number of samples in the last
488 // frame.
489
490 for (b = 0; b < ctx->num_blocks; b++)
491 div_blocks[b] = ctx->sconf.frame_length >> div_blocks[b];
492
493 if (ctx->cur_frame_length != ctx->sconf.frame_length) {
494 unsigned int remaining = ctx->cur_frame_length;
495
496 for (b = 0; b < ctx->num_blocks; b++) {
497 if (remaining < div_blocks[b]) {
498 div_blocks[b] = remaining;
499 ctx->num_blocks = b + 1;
500 break;
501 }
502
503 remaining -= div_blocks[b];
504 }
505 }
506}
507
508
509/** Reads the block data for a constant block
510 */
1261b07f 511static void read_const_block_data(ALSDecContext *ctx, ALSBlockData *bd)
99971952
TB		 512{
513 ALSSpecificConfig *sconf = &ctx->sconf;
514 AVCodecContext *avctx = ctx->avctx;
515 GetBitContext *gb = &ctx->gb;
99971952 516
1261b07f
TB		 517 bd->const_val = 0;
518 bd->const_block = get_bits1(gb); // 1 = constant value, 0 = zero block (silence)
519 bd->js_blocks = get_bits1(gb);
99971952
TB		 520
521 // skip 5 reserved bits
522 skip_bits(gb, 5);
523
1261b07f 524 if (bd->const_block) {
99971952 525 unsigned int const_val_bits = sconf->floating ? 24 : avctx->bits_per_raw_sample;
1261b07f 526 bd->const_val = get_sbits_long(gb, const_val_bits);
99971952
TB		 527 }
528
1261b07f
TB		 529 // ensure constant block decoding by reusing this field
530 bd->const_block = 1;
531}
532
533
534/** Decodes the block data for a constant block
535 */
536static void decode_const_block_data(ALSDecContext *ctx, ALSBlockData *bd)
537{
538 int smp = bd->block_length;
539 int32_t val = bd->const_val;
540 int32_t *dst = bd->raw_samples;
541
99971952 542 // write raw samples into buffer
1261b07f
TB		 543 for (; smp; smp--)
544 *dst++ = val;
99971952
TB		 545}
546
547
548/** Reads the block data for a non-constant block
549 */
1261b07f 550static int read_var_block_data(ALSDecContext *ctx, ALSBlockData *bd)
99971952
TB		 551{
552 ALSSpecificConfig *sconf = &ctx->sconf;
553 AVCodecContext *avctx = ctx->avctx;
554 GetBitContext *gb = &ctx->gb;
555 unsigned int k;
556 unsigned int s[8];
557 unsigned int sub_blocks, log2_sub_blocks, sb_length;
99971952 558 unsigned int start = 0;
1261b07f
TB		 559 unsigned int opt_order;
560 int sb;
561 int32_t *quant_cof = bd->quant_cof;
99971952 562
1261b07f
TB		 563
564 // ensure variable block decoding by reusing this field
565 bd->const_block = 0;
566
567 bd->opt_order = 1;
568 bd->js_blocks = get_bits1(gb);
569
570 opt_order = bd->opt_order;
99971952
TB		 571
572 // determine the number of subblocks for entropy decoding
573 if (!sconf->bgmc && !sconf->sb_part) {
574 log2_sub_blocks = 0;
575 } else {
576 if (sconf->bgmc && sconf->sb_part)
577 log2_sub_blocks = get_bits(gb, 2);
578 else
579 log2_sub_blocks = 2 * get_bits1(gb);
580 }
581
582 sub_blocks = 1 << log2_sub_blocks;
583
584 // do not continue in case of a damaged stream since
585 // block_length must be evenly divisible by sub_blocks
1261b07f 586 if (bd->block_length & (sub_blocks - 1)) {
99971952
TB		 587 av_log(avctx, AV_LOG_WARNING,
588 "Block length is not evenly divisible by the number of subblocks.\n");
589 return -1;
590 }
591
1261b07f 592 sb_length = bd->block_length >> log2_sub_blocks;
99971952
TB		 593
594
595 if (sconf->bgmc) {
596 // TODO: BGMC mode
597 } else {
598 s[0] = get_bits(gb, 4 + (sconf->resolution > 1));
599 for (k = 1; k < sub_blocks; k++)
600 s[k] = s[k - 1] + decode_rice(gb, 0);
601 }
602
603 if (get_bits1(gb))
1261b07f 604 bd->shift_lsbs = get_bits(gb, 4) + 1;
99971952 605
1261b07f 606 bd->store_prev_samples = (bd->js_blocks && bd->raw_other) || bd->shift_lsbs;
99971952
TB		 607
608
609 if (!sconf->rlslms) {
610 if (sconf->adapt_order) {
1261b07f 611 int opt_order_length = av_ceil_log2(av_clip((bd->block_length >> 3) - 1,
99971952 612 2, sconf->max_order + 1));
1261b07f 613 bd->opt_order = get_bits(gb, opt_order_length);
99971952 614 } else {
1261b07f 615 bd->opt_order = sconf->max_order;
99971952
TB		 616 }
617
1261b07f
TB		 618 opt_order = bd->opt_order;
619
99971952
TB		 620 if (opt_order) {
621 int add_base;
622
623 if (sconf->coef_table == 3) {
624 add_base = 0x7F;
625
626 // read coefficient 0
627 quant_cof[0] = 32 * parcor_scaled_values[get_bits(gb, 7)];
628
629 // read coefficient 1
630 if (opt_order > 1)
631 quant_cof[1] = -32 * parcor_scaled_values[get_bits(gb, 7)];
632
633 // read coefficients 2 to opt_order
634 for (k = 2; k < opt_order; k++)
635 quant_cof[k] = get_bits(gb, 7);
636 } else {
637 int k_max;
638 add_base = 1;
639
640 // read coefficient 0 to 19
641 k_max = FFMIN(opt_order, 20);
642 for (k = 0; k < k_max; k++) {
643 int rice_param = parcor_rice_table[sconf->coef_table][k][1];
644 int offset = parcor_rice_table[sconf->coef_table][k][0];
645 quant_cof[k] = decode_rice(gb, rice_param) + offset;
646 }
647
648 // read coefficients 20 to 126
649 k_max = FFMIN(opt_order, 127);
650 for (; k < k_max; k++)
651 quant_cof[k] = decode_rice(gb, 2) + (k & 1);
652
653 // read coefficients 127 to opt_order
654 for (; k < opt_order; k++)
655 quant_cof[k] = decode_rice(gb, 1);
656
657 quant_cof[0] = 32 * parcor_scaled_values[quant_cof[0] + 64];
658
659 if (opt_order > 1)
660 quant_cof[1] = -32 * parcor_scaled_values[quant_cof[1] + 64];
661 }
662
663 for (k = 2; k < opt_order; k++)
664 quant_cof[k] = (quant_cof[k] << 14) + (add_base << 13);
665 }
666 }
667
93d38cf6
TB		 668 // read LTP gain and lag values
669 if (sconf->long_term_prediction) {
1261b07f 670 *bd->use_ltp = get_bits1(gb);
93d38cf6 671
1261b07f
TB		 672 if (*bd->use_ltp) {
673 bd->ltp_gain[0] = decode_rice(gb, 1) << 3;
674 bd->ltp_gain[1] = decode_rice(gb, 2) << 3;
93d38cf6 675
1261b07f 676 bd->ltp_gain[2] = ltp_gain_values[get_unary(gb, 0, 4)][get_bits(gb, 2)];
93d38cf6 677
1261b07f
TB		 678 bd->ltp_gain[3] = decode_rice(gb, 2) << 3;
679 bd->ltp_gain[4] = decode_rice(gb, 1) << 3;
93d38cf6 680
1261b07f
TB		 681 *bd->ltp_lag = get_bits(gb, ctx->ltp_lag_length);
682 *bd->ltp_lag += FFMAX(4, opt_order + 1);
93d38cf6
TB		 683 }
684 }
99971952
TB		 685
686 // read first value and residuals in case of a random access block
1261b07f 687 if (bd->ra_block) {
99971952 688 if (opt_order)
1261b07f 689 bd->raw_samples[0] = decode_rice(gb, avctx->bits_per_raw_sample - 4);
99971952 690 if (opt_order > 1)
1261b07f 691 bd->raw_samples[1] = decode_rice(gb, s[0] + 3);
99971952 692 if (opt_order > 2)
1261b07f 693 bd->raw_samples[2] = decode_rice(gb, s[0] + 1);
99971952
TB		 694
695 start = FFMIN(opt_order, 3);
696 }
697
698 // read all residuals
699 if (sconf->bgmc) {
700 // TODO: BGMC mode
701 } else {
1261b07f 702 int32_t *current_res = bd->raw_samples + start;
99971952
TB		 703
704 for (sb = 0; sb < sub_blocks; sb++, start = 0)
705 for (; start < sb_length; start++)
706 *current_res++ = decode_rice(gb, s[sb]);
707 }
708
1261b07f
TB		 709 if (!sconf->mc_coding || ctx->js_switch)
710 align_get_bits(gb);
711
712 return 0;
713}
714
715
716/** Decodes the block data for a non-constant block
717 */
718static int decode_var_block_data(ALSDecContext *ctx, ALSBlockData *bd)
719{
720 ALSSpecificConfig *sconf = &ctx->sconf;
721 unsigned int block_length = bd->block_length;
722 unsigned int smp = 0;
723 unsigned int k;
b2f4b043 724 int opt_order = bd->opt_order;
1261b07f
TB		 725 int sb;
726 int64_t y;
727 int32_t *quant_cof = bd->quant_cof;
728 int32_t *lpc_cof = bd->lpc_cof;
729 int32_t *raw_samples = bd->raw_samples;
99c5f5cc 730 int32_t *raw_samples_end = bd->raw_samples + bd->block_length;
ff9ea0b7 731 int32_t *lpc_cof_reversed = ctx->lpc_cof_reversed_buffer;
1261b07f 732
93d38cf6 733 // reverse long-term prediction
1261b07f 734 if (*bd->use_ltp) {
93d38cf6
TB		 735 int ltp_smp;
736
1261b07f
TB		 737 for (ltp_smp = FFMAX(*bd->ltp_lag - 2, 0); ltp_smp < block_length; ltp_smp++) {
738 int center = ltp_smp - *bd->ltp_lag;
93d38cf6
TB		 739 int begin = FFMAX(0, center - 2);
740 int end = center + 3;
741 int tab = 5 - (end - begin);
742 int base;
743
744 y = 1 << 6;
745
746 for (base = begin; base < end; base++, tab++)
1261b07f 747 y += MUL64(bd->ltp_gain[tab], raw_samples[base]);
93d38cf6
TB		 748
749 raw_samples[ltp_smp] += y >> 7;
750 }
751 }
752
99971952 753 // reconstruct all samples from residuals
1261b07f 754 if (bd->ra_block) {
99971952
TB		 755 for (smp = 0; smp < opt_order; smp++) {
756 y = 1 << 19;
757
758 for (sb = 0; sb < smp; sb++)
99c5f5cc 759 y += MUL64(lpc_cof[sb], raw_samples[-(sb + 1)]);
99971952 760
99c5f5cc 761 *raw_samples++ -= y >> 20;
99971952
TB		 762 parcor_to_lpc(smp, quant_cof, lpc_cof);
763 }
764 } else {
765 for (k = 0; k < opt_order; k++)
766 parcor_to_lpc(k, quant_cof, lpc_cof);
767
768 // store previous samples in case that they have to be altered
1261b07f
TB		 769 if (bd->store_prev_samples)
770 memcpy(bd->prev_raw_samples, raw_samples - sconf->max_order,
771 sizeof(*bd->prev_raw_samples) * sconf->max_order);
99971952
TB		 772
773 // reconstruct difference signal for prediction (joint-stereo)
1261b07f 774 if (bd->js_blocks && bd->raw_other) {
99971952
TB		 775 int32_t *left, *right;
776
1261b07f 777 if (bd->raw_other > raw_samples) { // D = R - L
99971952 778 left = raw_samples;
1261b07f 779 right = bd->raw_other;
99971952 780 } else { // D = R - L
1261b07f 781 left = bd->raw_other;
99971952
TB		 782 right = raw_samples;
783 }
784
785 for (sb = -1; sb >= -sconf->max_order; sb--)
786 raw_samples[sb] = right[sb] - left[sb];
787 }
788
789 // reconstruct shifted signal
1261b07f 790 if (bd->shift_lsbs)
99971952 791 for (sb = -1; sb >= -sconf->max_order; sb--)
1261b07f 792 raw_samples[sb] >>= bd->shift_lsbs;
99971952
TB		 793 }
794
99c5f5cc
TB		 795 // reverse linear prediction coefficients for efficiency
796 lpc_cof = lpc_cof + opt_order;
797
798 for (sb = 0; sb < opt_order; sb++)
799 lpc_cof_reversed[sb] = lpc_cof[-(sb + 1)];
800
99971952 801 // reconstruct raw samples
99c5f5cc
TB		 802 raw_samples = bd->raw_samples + smp;
803 lpc_cof = lpc_cof_reversed + opt_order;
804
805 for (; raw_samples < raw_samples_end; raw_samples++) {
99971952
TB		 806 y = 1 << 19;
807
99c5f5cc
TB		 808 for (sb = -opt_order; sb < 0; sb++)
809 y += MUL64(lpc_cof[sb], raw_samples[sb]);
99971952 810
99c5f5cc 811 *raw_samples -= y >> 20;
99971952
TB		 812 }
813
99c5f5cc
TB		 814 raw_samples = bd->raw_samples;
815
99971952 816 // restore previous samples in case that they have been altered
1261b07f
TB		 817 if (bd->store_prev_samples)
818 memcpy(raw_samples - sconf->max_order, bd->prev_raw_samples,
99971952
TB		 819 sizeof(*raw_samples) * sconf->max_order);
820
821 return 0;
822}
823
824
825/** Reads the block data.
826 */
1261b07f 827static int read_block(ALSDecContext *ctx, ALSBlockData *bd)
99971952 828{
99971952 829 GetBitContext *gb = &ctx->gb;
99971952
TB		 830
831 // read block type flag and read the samples accordingly
832 if (get_bits1(gb)) {
1261b07f 833 if (read_var_block_data(ctx, bd))
99971952
TB		 834 return -1;
835 } else {
1261b07f 836 read_const_block_data(ctx, bd);
99971952
TB		 837 }
838
1261b07f
TB		 839 return 0;
840}
99971952 841
99971952 842
1261b07f
TB		 843/** Decodes the block data.
844 */
845static int decode_block(ALSDecContext *ctx, ALSBlockData *bd)
846{
847 unsigned int smp;
848
849 // read block type flag and read the samples accordingly
850 if (bd->const_block)
851 decode_const_block_data(ctx, bd);
852 else if (decode_var_block_data(ctx, bd))
853 return -1;
854
855 // TODO: read RLSLMS extension data
856
857 if (bd->shift_lsbs)
858 for (smp = 0; smp < bd->block_length; smp++)
859 bd->raw_samples[smp] <<= bd->shift_lsbs;
99971952
TB		 860
861 return 0;
862}
863
864
1261b07f
TB		 865/** Reads and decodes block data successively.
866 */
867static int read_decode_block(ALSDecContext *ctx, ALSBlockData *bd)
868{
869 int ret;
870
871 ret = read_block(ctx, bd);
872
873 if (ret)
874 return ret;
875
876 ret = decode_block(ctx, bd);
877
878 return ret;
879}
880
881
99971952
TB		 882/** Computes the number of samples left to decode for the current frame and
883 * sets these samples to zero.
884 */
885static void zero_remaining(unsigned int b, unsigned int b_max,
886 const unsigned int *div_blocks, int32_t *buf)
887{
888 unsigned int count = 0;
889
890 while (b < b_max)
891 count += div_blocks[b];
892
0bb622ba 893 if (count)
9349e558 894 memset(buf, 0, sizeof(*buf) * count);
99971952
TB		 895}
896
897
898/** Decodes blocks independently.
899 */
900static int decode_blocks_ind(ALSDecContext *ctx, unsigned int ra_frame,
901 unsigned int c, const unsigned int *div_blocks,
902 unsigned int *js_blocks)
903{
99971952 904 unsigned int b;
1261b07f
TB		 905 ALSBlockData bd;
906
907 memset(&bd, 0, sizeof(ALSBlockData));
908
909 bd.ra_block = ra_frame;
910 bd.use_ltp = ctx->use_ltp;
911 bd.ltp_lag = ctx->ltp_lag;
912 bd.ltp_gain = ctx->ltp_gain[0];
e38215f2
TB		 913 bd.quant_cof = ctx->quant_cof[0];
914 bd.lpc_cof = ctx->lpc_cof[0];
1261b07f
TB		 915 bd.prev_raw_samples = ctx->prev_raw_samples;
916 bd.raw_samples = ctx->raw_samples[c];
917
99971952
TB		 918
919 for (b = 0; b < ctx->num_blocks; b++) {
1261b07f
TB		 920 bd.shift_lsbs = 0;
921 bd.block_length = div_blocks[b];
922
923 if (read_decode_block(ctx, &bd)) {
99971952 924 // damaged block, write zero for the rest of the frame
1261b07f 925 zero_remaining(b, ctx->num_blocks, div_blocks, bd.raw_samples);
99971952
TB		 926 return -1;
927 }
1261b07f
TB		 928 bd.raw_samples += div_blocks[b];
929 bd.ra_block = 0;
99971952
TB		 930 }
931
932 return 0;
933}
934
935
936/** Decodes blocks dependently.
937 */
938static int decode_blocks(ALSDecContext *ctx, unsigned int ra_frame,
939 unsigned int c, const unsigned int *div_blocks,
940 unsigned int *js_blocks)
941{
942 ALSSpecificConfig *sconf = &ctx->sconf;
943 unsigned int offset = 0;
99971952 944 unsigned int b;
1261b07f
TB		 945 ALSBlockData bd[2];
946
947 memset(bd, 0, 2 * sizeof(ALSBlockData));
948
949 bd[0].ra_block = ra_frame;
950 bd[0].use_ltp = ctx->use_ltp;
951 bd[0].ltp_lag = ctx->ltp_lag;
952 bd[0].ltp_gain = ctx->ltp_gain[0];
e38215f2
TB		 953 bd[0].quant_cof = ctx->quant_cof[0];
954 bd[0].lpc_cof = ctx->lpc_cof[0];
1261b07f
TB		 955 bd[0].prev_raw_samples = ctx->prev_raw_samples;
956 bd[0].js_blocks = *js_blocks;
957
958 bd[1].ra_block = ra_frame;
959 bd[1].use_ltp = ctx->use_ltp;
960 bd[1].ltp_lag = ctx->ltp_lag;
961 bd[1].ltp_gain = ctx->ltp_gain[0];
e38215f2
TB		 962 bd[1].quant_cof = ctx->quant_cof[0];
963 bd[1].lpc_cof = ctx->lpc_cof[0];
1261b07f
TB		 964 bd[1].prev_raw_samples = ctx->prev_raw_samples;
965 bd[1].js_blocks = *(js_blocks + 1);
99971952
TB		 966
967 // decode all blocks
968 for (b = 0; b < ctx->num_blocks; b++) {
969 unsigned int s;
1261b07f
TB		 970
971 bd[0].shift_lsbs = 0;
972 bd[1].shift_lsbs = 0;
973
974 bd[0].block_length = div_blocks[b];
975 bd[1].block_length = div_blocks[b];
976
977 bd[0].raw_samples = ctx->raw_samples[c ] + offset;
978 bd[1].raw_samples = ctx->raw_samples[c + 1] + offset;
979
980 bd[0].raw_other = bd[1].raw_samples;
981 bd[1].raw_other = bd[0].raw_samples;
982
983 if(read_decode_block(ctx, &bd[0]) || read_decode_block(ctx, &bd[1])) {
99971952 984 // damaged block, write zero for the rest of the frame
1261b07f
TB		 985 zero_remaining(b, ctx->num_blocks, div_blocks, bd[0].raw_samples);
986 zero_remaining(b, ctx->num_blocks, div_blocks, bd[1].raw_samples);
99971952
TB		 987 return -1;
988 }
989
990 // reconstruct joint-stereo blocks
1261b07f
TB		 991 if (bd[0].js_blocks) {
992 if (bd[1].js_blocks)
99971952
TB		 993 av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel pair!\n");
994
995 for (s = 0; s < div_blocks[b]; s++)
1261b07f
TB		 996 bd[0].raw_samples[s] = bd[1].raw_samples[s] - bd[0].raw_samples[s];
997 } else if (bd[1].js_blocks) {
99971952 998 for (s = 0; s < div_blocks[b]; s++)
1261b07f 999 bd[1].raw_samples[s] = bd[1].raw_samples[s] + bd[0].raw_samples[s];
99971952
TB		 1000 }
1001
1002 offset += div_blocks[b];
1261b07f
TB		 1003 bd[0].ra_block = 0;
1004 bd[1].ra_block = 0;
99971952
TB		 1005 }
1006
1007 // store carryover raw samples,
1008 // the others channel raw samples are stored by the calling function.
1009 memmove(ctx->raw_samples[c] - sconf->max_order,
1010 ctx->raw_samples[c] - sconf->max_order + sconf->frame_length,
1011 sizeof(*ctx->raw_samples[c]) * sconf->max_order);
1012
1013 return 0;
1014}
1015
1016
e38215f2
TB		 1017/** Reads the channel data.
1018 */
1019static int read_channel_data(ALSDecContext *ctx, ALSChannelData *cd, int c)
1020{
1021 GetBitContext *gb = &ctx->gb;
1022 ALSChannelData *current = cd;
1023 unsigned int channels = ctx->avctx->channels;
1024 int entries = 0;
1025
1026 while (entries < channels && !(current->stop_flag = get_bits1(gb))) {
1027 current->master_channel = get_bits_long(gb, av_ceil_log2(channels));
1028
1029 if (current->master_channel >= channels) {
1030 av_log(ctx->avctx, AV_LOG_ERROR, "Invalid master channel!\n");
1031 return -1;
1032 }
1033
1034 if (current->master_channel != c) {
1035 current->time_diff_flag = get_bits1(gb);
1036 current->weighting[0] = mcc_weightings[av_clip(decode_rice(gb, 1) + 16, 0, 32)];
1037 current->weighting[1] = mcc_weightings[av_clip(decode_rice(gb, 2) + 14, 0, 32)];
1038 current->weighting[2] = mcc_weightings[av_clip(decode_rice(gb, 1) + 16, 0, 32)];
1039
1040 if (current->time_diff_flag) {
1041 current->weighting[3] = mcc_weightings[av_clip(decode_rice(gb, 1) + 16, 0, 32)];
1042 current->weighting[4] = mcc_weightings[av_clip(decode_rice(gb, 1) + 16, 0, 32)];
1043 current->weighting[5] = mcc_weightings[av_clip(decode_rice(gb, 1) + 16, 0, 32)];
1044
1045 current->time_diff_sign = get_bits1(gb);
1046 current->time_diff_index = get_bits(gb, ctx->ltp_lag_length - 3) + 3;
1047 }
1048 }
1049
1050 current++;
1051 entries++;
1052 }
1053
1054 if (entries == channels) {
1055 av_log(ctx->avctx, AV_LOG_ERROR, "Damaged channel data!\n");
1056 return -1;
1057 }
1058
1059 align_get_bits(gb);
1060 return 0;
1061}
1062
1063
1064/** Recursively reverts the inter-channel correlation for a block.
1065 */
1066static int revert_channel_correlation(ALSDecContext *ctx, ALSBlockData *bd,
1067 ALSChannelData **cd, int *reverted,
1068 unsigned int offset, int c)
1069{
1070 ALSChannelData *ch = cd[c];
1071 unsigned int dep = 0;
1072 unsigned int channels = ctx->avctx->channels;
1073
1074 if (reverted[c])
1075 return 0;
1076
1077 reverted[c] = 1;
1078
1079 while (dep < channels && !ch[dep].stop_flag) {
1080 revert_channel_correlation(ctx, bd, cd, reverted, offset,
1081 ch[dep].master_channel);
1082
1083 dep++;
1084 }
1085
1086 if (dep == channels) {
1087 av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel correlation!\n");
1088 return -1;
1089 }
1090
1091 bd->use_ltp = ctx->use_ltp + c;
1092 bd->ltp_lag = ctx->ltp_lag + c;
1093 bd->ltp_gain = ctx->ltp_gain[c];
1094 bd->lpc_cof = ctx->lpc_cof[c];
1095 bd->quant_cof = ctx->quant_cof[c];
1096 bd->raw_samples = ctx->raw_samples[c] + offset;
1097
1098 dep = 0;
1099 while (!ch[dep].stop_flag) {
1100 unsigned int smp;
1101 unsigned int begin = 1;
1102 unsigned int end = bd->block_length - 1;
1103 int64_t y;
1104 int32_t *master = ctx->raw_samples[ch[dep].master_channel] + offset;
1105
1106 if (ch[dep].time_diff_flag) {
1107 int t = ch[dep].time_diff_index;
1108
1109 if (ch[dep].time_diff_sign) {
1110 t = -t;
1111 begin -= t;
1112 } else {
1113 end -= t;
1114 }
1115
1116 for (smp = begin; smp < end; smp++) {
1117 y = (1 << 6) +
1118 MUL64(ch[dep].weighting[0], master[smp - 1 ]) +
1119 MUL64(ch[dep].weighting[1], master[smp ]) +
1120 MUL64(ch[dep].weighting[2], master[smp + 1 ]) +
1121 MUL64(ch[dep].weighting[3], master[smp - 1 + t]) +
1122 MUL64(ch[dep].weighting[4], master[smp + t]) +
1123 MUL64(ch[dep].weighting[5], master[smp + 1 + t]);
1124
1125 bd->raw_samples[smp] += y >> 7;
1126 }
1127 } else {
1128 for (smp = begin; smp < end; smp++) {
1129 y = (1 << 6) +
1130 MUL64(ch[dep].weighting[0], master[smp - 1]) +
1131 MUL64(ch[dep].weighting[1], master[smp ]) +
1132 MUL64(ch[dep].weighting[2], master[smp + 1]);
1133
1134 bd->raw_samples[smp] += y >> 7;
1135 }
1136 }
1137
1138 dep++;
1139 }
1140
1141 return 0;
1142}
1143
1144
99971952
TB		 1145/** Reads the frame data.
1146 */
1147static int read_frame_data(ALSDecContext *ctx, unsigned int ra_frame)
1148{
1149 ALSSpecificConfig *sconf = &ctx->sconf;
1150 AVCodecContext *avctx = ctx->avctx;
1151 GetBitContext *gb = &ctx->gb;
1152 unsigned int div_blocks[32]; ///< block sizes.
1153 unsigned int c;
1154 unsigned int js_blocks[2];
1155
1156 uint32_t bs_info = 0;
1157
1158 // skip the size of the ra unit if present in the frame
1159 if (sconf->ra_flag == RA_FLAG_FRAMES && ra_frame)
1160 skip_bits_long(gb, 32);
1161
1162 if (sconf->mc_coding && sconf->joint_stereo) {
1163 ctx->js_switch = get_bits1(gb);
1164 align_get_bits(gb);
1165 }
1166
1167 if (!sconf->mc_coding || ctx->js_switch) {
1168 int independent_bs = !sconf->joint_stereo;
1169
1170 for (c = 0; c < avctx->channels; c++) {
1171 js_blocks[0] = 0;
1172 js_blocks[1] = 0;
1173
1174 get_block_sizes(ctx, div_blocks, &bs_info);
1175
1176 // if joint_stereo and block_switching is set, independent decoding
1177 // is signaled via the first bit of bs_info
1178 if (sconf->joint_stereo && sconf->block_switching)
1179 if (bs_info >> 31)
1180 independent_bs = 2;
1181
1182 // if this is the last channel, it has to be decoded independently
1183 if (c == avctx->channels - 1)
1184 independent_bs = 1;
1185
1186 if (independent_bs) {
1187 if (decode_blocks_ind(ctx, ra_frame, c, div_blocks, js_blocks))
1188 return -1;
1189
1190 independent_bs--;
1191 } else {
1192 if (decode_blocks(ctx, ra_frame, c, div_blocks, js_blocks))
1193 return -1;
1194
1195 c++;
1196 }
1197
1198 // store carryover raw samples
1199 memmove(ctx->raw_samples[c] - sconf->max_order,
1200 ctx->raw_samples[c] - sconf->max_order + sconf->frame_length,
1201 sizeof(*ctx->raw_samples[c]) * sconf->max_order);
1202 }
1203 } else { // multi-channel coding
e38215f2
TB		 1204 ALSBlockData bd;
1205 int b;
1206 int *reverted_channels = ctx->reverted_channels;
1207 unsigned int offset = 0;
1208
1209 for (c = 0; c < avctx->channels; c++)
1210 if (ctx->chan_data[c] < ctx->chan_data_buffer) {
1211 av_log(ctx->avctx, AV_LOG_ERROR, "Invalid channel data!\n");
1212 return -1;
1213 }
1214
1215 memset(&bd, 0, sizeof(ALSBlockData));
1216 memset(reverted_channels, 0, sizeof(*reverted_channels) * avctx->channels);
1217
1218 bd.ra_block = ra_frame;
1219 bd.prev_raw_samples = ctx->prev_raw_samples;
1220
99971952
TB		 1221 get_block_sizes(ctx, div_blocks, &bs_info);
1222
e38215f2
TB		 1223 for (b = 0; b < ctx->num_blocks; b++) {
1224 bd.shift_lsbs = 0;
1225 bd.block_length = div_blocks[b];
1226
1227 for (c = 0; c < avctx->channels; c++) {
1228 bd.use_ltp = ctx->use_ltp + c;
1229 bd.ltp_lag = ctx->ltp_lag + c;
1230 bd.ltp_gain = ctx->ltp_gain[c];
1231 bd.lpc_cof = ctx->lpc_cof[c];
1232 bd.quant_cof = ctx->quant_cof[c];
1233 bd.raw_samples = ctx->raw_samples[c] + offset;
1234 bd.raw_other = NULL;
1235
1236 read_block(ctx, &bd);
1237 if (read_channel_data(ctx, ctx->chan_data[c], c))
1238 return -1;
1239 }
1240
1241 for (c = 0; c < avctx->channels; c++)
1242 if (revert_channel_correlation(ctx, &bd, ctx->chan_data,
1243 reverted_channels, offset, c))
1244 return -1;
1245
1246 for (c = 0; c < avctx->channels; c++) {
1247 bd.use_ltp = ctx->use_ltp + c;
1248 bd.ltp_lag = ctx->ltp_lag + c;
1249 bd.ltp_gain = ctx->ltp_gain[c];
1250 bd.lpc_cof = ctx->lpc_cof[c];
1251 bd.quant_cof = ctx->quant_cof[c];
1252 bd.raw_samples = ctx->raw_samples[c] + offset;
1253 decode_block(ctx, &bd);
1254 }
1255
1256 memset(reverted_channels, 0, avctx->channels * sizeof(*reverted_channels));
1257 offset += div_blocks[b];
1258 bd.ra_block = 0;
1259 }
1260
1261 // store carryover raw samples
1262 for (c = 0; c < avctx->channels; c++)
1263 memmove(ctx->raw_samples[c] - sconf->max_order,
1264 ctx->raw_samples[c] - sconf->max_order + sconf->frame_length,
1265 sizeof(*ctx->raw_samples[c]) * sconf->max_order);
99971952
TB		 1266 }
1267
1268 // TODO: read_diff_float_data
1269
1270 return 0;
1271}
1272
1273
1274/** Decodes an ALS frame.
1275 */
1276static int decode_frame(AVCodecContext *avctx,
1277 void *data, int *data_size,
1278 AVPacket *avpkt)
1279{
1280 ALSDecContext *ctx = avctx->priv_data;
1281 ALSSpecificConfig *sconf = &ctx->sconf;
1282 const uint8_t *buffer = avpkt->data;
1283 int buffer_size = avpkt->size;
1284 int invalid_frame, size;
1285 unsigned int c, sample, ra_frame, bytes_read, shift;
1286
1287 init_get_bits(&ctx->gb, buffer, buffer_size * 8);
1288
1289 // In the case that the distance between random access frames is set to zero
1290 // (sconf->ra_distance == 0) no frame is treated as a random access frame.
1291 // For the first frame, if prediction is used, all samples used from the
1292 // previous frame are assumed to be zero.
1293 ra_frame = sconf->ra_distance && !(ctx->frame_id % sconf->ra_distance);
1294
1295 // the last frame to decode might have a different length
1296 if (sconf->samples != 0xFFFFFFFF)
1297 ctx->cur_frame_length = FFMIN(sconf->samples - ctx->frame_id * (uint64_t) sconf->frame_length,
1298 sconf->frame_length);
1299 else
1300 ctx->cur_frame_length = sconf->frame_length;
1301
1302 // decode the frame data
1303 if ((invalid_frame = read_frame_data(ctx, ra_frame) < 0))
1304 av_log(ctx->avctx, AV_LOG_WARNING,
1305 "Reading frame data failed. Skipping RA unit.\n");
1306
1307 ctx->frame_id++;
1308
1309 // check for size of decoded data
1310 size = ctx->cur_frame_length * avctx->channels *
1311 (av_get_bits_per_sample_format(avctx->sample_fmt) >> 3);
1312
1313 if (size > *data_size) {
1314 av_log(avctx, AV_LOG_ERROR, "Decoded data exceeds buffer size.\n");
1315 return -1;
1316 }
1317
1318 *data_size = size;
1319
1320 // transform decoded frame into output format
1321 #define INTERLEAVE_OUTPUT(bps) \
1322 { \
1323 int##bps##_t *dest = (int##bps##_t*) data; \
1324 shift = bps - ctx->avctx->bits_per_raw_sample; \
1325 for (sample = 0; sample < ctx->cur_frame_length; sample++) \
1326 for (c = 0; c < avctx->channels; c++) \
1327 *dest++ = ctx->raw_samples[c][sample] << shift; \
1328 }
1329
1330 if (ctx->avctx->bits_per_raw_sample <= 16) {
1331 INTERLEAVE_OUTPUT(16)
1332 } else {
1333 INTERLEAVE_OUTPUT(32)
1334 }
1335
1336 bytes_read = invalid_frame ? buffer_size :
1337 (get_bits_count(&ctx->gb) + 7) >> 3;
1338
1339 return bytes_read;
1340}
1341
1342
1343/** Uninitializes the ALS decoder.
1344 */
1345static av_cold int decode_end(AVCodecContext *avctx)
1346{
1347 ALSDecContext *ctx = avctx->priv_data;
1348
1349 av_freep(&ctx->sconf.chan_pos);
1350
1261b07f
TB		 1351 av_freep(&ctx->use_ltp);
1352 av_freep(&ctx->ltp_lag);
1353 av_freep(&ctx->ltp_gain);
1354 av_freep(&ctx->ltp_gain_buffer);
99971952
TB		 1355 av_freep(&ctx->quant_cof);
1356 av_freep(&ctx->lpc_cof);
e38215f2
TB		 1357 av_freep(&ctx->quant_cof_buffer);
1358 av_freep(&ctx->lpc_cof_buffer);
ff9ea0b7 1359 av_freep(&ctx->lpc_cof_reversed_buffer);
99971952
TB		 1360 av_freep(&ctx->prev_raw_samples);
1361 av_freep(&ctx->raw_samples);
1362 av_freep(&ctx->raw_buffer);
e38215f2
TB		 1363 av_freep(&ctx->chan_data);
1364 av_freep(&ctx->chan_data_buffer);
1365 av_freep(&ctx->reverted_channels);
99971952
TB		 1366
1367 return 0;
1368}
1369
1370
1371/** Initializes the ALS decoder.
1372 */
1373static av_cold int decode_init(AVCodecContext *avctx)
1374{
1375 unsigned int c;
1376 unsigned int channel_size;
1261b07f 1377 int num_buffers;
99971952
TB		 1378 ALSDecContext *ctx = avctx->priv_data;
1379 ALSSpecificConfig *sconf = &ctx->sconf;
1380 ctx->avctx = avctx;
1381
1382 if (!avctx->extradata) {
1383 av_log(avctx, AV_LOG_ERROR, "Missing required ALS extradata.\n");
1384 return -1;
1385 }
1386
1387 if (read_specific_config(ctx)) {
1388 av_log(avctx, AV_LOG_ERROR, "Reading ALSSpecificConfig failed.\n");
1389 decode_end(avctx);
1390 return -1;
1391 }
1392
1393 if (check_specific_config(ctx)) {
1394 decode_end(avctx);
1395 return -1;
1396 }
1397
1398 if (sconf->floating) {
1399 avctx->sample_fmt = SAMPLE_FMT_FLT;
1400 avctx->bits_per_raw_sample = 32;
1401 } else {
1402 avctx->sample_fmt = sconf->resolution > 1
1403 ? SAMPLE_FMT_S32 : SAMPLE_FMT_S16;
1404 avctx->bits_per_raw_sample = (sconf->resolution + 1) * 8;
1405 }
1406
93d38cf6
TB		 1407 // set lag value for long-term prediction
1408 ctx->ltp_lag_length = 8 + (avctx->sample_rate >= 96000) +
1409 (avctx->sample_rate >= 192000);
1410
1261b07f
TB		 1411 // allocate quantized parcor coefficient buffer
1412 num_buffers = sconf->mc_coding ? avctx->channels : 1;
1413
e38215f2
TB		 1414 ctx->quant_cof = av_malloc(sizeof(*ctx->quant_cof) * num_buffers);
1415 ctx->lpc_cof = av_malloc(sizeof(*ctx->lpc_cof) * num_buffers);
1416 ctx->quant_cof_buffer = av_malloc(sizeof(*ctx->quant_cof_buffer) *
1417 num_buffers * sconf->max_order);
1418 ctx->lpc_cof_buffer = av_malloc(sizeof(*ctx->lpc_cof_buffer) *
1419 num_buffers * sconf->max_order);
ff9ea0b7
TB		 1420 ctx->lpc_cof_reversed_buffer = av_malloc(sizeof(*ctx->lpc_cof_buffer) *
1421 sconf->max_order);
e38215f2 1422
099809d1
TB		 1423 if (!ctx->quant_cof || !ctx->lpc_cof ||
1424 !ctx->quant_cof_buffer || !ctx->lpc_cof_buffer ||
ff9ea0b7 1425 !ctx->lpc_cof_reversed_buffer) {
e38215f2
TB		 1426 av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
1427 return AVERROR(ENOMEM);
1428 }
1429
1430 // assign quantized parcor coefficient buffers
1431 for (c = 0; c < num_buffers; c++) {
1432 ctx->quant_cof[c] = ctx->quant_cof_buffer + c * sconf->max_order;
1433 ctx->lpc_cof[c] = ctx->lpc_cof_buffer + c * sconf->max_order;
1434 }
1435
1261b07f
TB		 1436 // allocate and assign lag and gain data buffer for ltp mode
1437 ctx->use_ltp = av_mallocz(sizeof(*ctx->use_ltp) * num_buffers);
1438 ctx->ltp_lag = av_malloc (sizeof(*ctx->ltp_lag) * num_buffers);
1439 ctx->ltp_gain = av_malloc (sizeof(*ctx->ltp_gain) * num_buffers);
1440 ctx->ltp_gain_buffer = av_malloc (sizeof(*ctx->ltp_gain_buffer) *
1441 num_buffers * 5);
1442
1443 if (!ctx->use_ltp || !ctx->ltp_lag ||
1444 !ctx->ltp_gain || !ctx->ltp_gain_buffer) {
1445 av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
1446 decode_end(avctx);
1447 return AVERROR(ENOMEM);
1448 }
1449
1450 for (c = 0; c < num_buffers; c++)
1451 ctx->ltp_gain[c] = ctx->ltp_gain_buffer + c * 5;
1452
e38215f2
TB		 1453 // allocate and assign channel data buffer for mcc mode
1454 if (sconf->mc_coding) {
1455 ctx->chan_data_buffer = av_malloc(sizeof(*ctx->chan_data_buffer) *
1456 num_buffers);
1457 ctx->chan_data = av_malloc(sizeof(ALSChannelData) *
1458 num_buffers);
1459 ctx->reverted_channels = av_malloc(sizeof(*ctx->reverted_channels) *
1460 num_buffers);
1461
1462 if (!ctx->chan_data_buffer || !ctx->chan_data || !ctx->reverted_channels) {
1463 av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
1464 decode_end(avctx);
1465 return AVERROR(ENOMEM);
1466 }
1467
1468 for (c = 0; c < num_buffers; c++)
1469 ctx->chan_data[c] = ctx->chan_data_buffer + c;
1470 } else {
1471 ctx->chan_data = NULL;
1472 ctx->chan_data_buffer = NULL;
1473 ctx->reverted_channels = NULL;
1474 }
1475
99971952
TB		 1476 avctx->frame_size = sconf->frame_length;
1477 channel_size = sconf->frame_length + sconf->max_order;
1478
1479 ctx->prev_raw_samples = av_malloc (sizeof(*ctx->prev_raw_samples) * sconf->max_order);
1480 ctx->raw_buffer = av_mallocz(sizeof(*ctx-> raw_buffer) * avctx->channels * channel_size);
1481 ctx->raw_samples = av_malloc (sizeof(*ctx-> raw_samples) * avctx->channels);
1482
1483 // allocate previous raw sample buffer
1484 if (!ctx->prev_raw_samples || !ctx->raw_buffer|| !ctx->raw_samples) {
1485 av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
1486 decode_end(avctx);
1487 return AVERROR(ENOMEM);
1488 }
1489
1490 // assign raw samples buffers
1491 ctx->raw_samples[0] = ctx->raw_buffer + sconf->max_order;
1492 for (c = 1; c < avctx->channels; c++)
1493 ctx->raw_samples[c] = ctx->raw_samples[c - 1] + channel_size;
1494
1495 return 0;
1496}
1497
1498
1499/** Flushes (resets) the frame ID after seeking.
1500 */
1501static av_cold void flush(AVCodecContext *avctx)
1502{
1503 ALSDecContext *ctx = avctx->priv_data;
1504
1505 ctx->frame_id = 0;
1506}
1507
1508
1509AVCodec als_decoder = {
1510 "als",
1511 CODEC_TYPE_AUDIO,
1512 CODEC_ID_MP4ALS,
1513 sizeof(ALSDecContext),
1514 decode_init,
1515 NULL,
1516 decode_end,
1517 decode_frame,
1518 .flush = flush,
1519 .capabilities = CODEC_CAP_SUBFRAMES,
1520 .long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Audio Lossless Coding (ALS)"),
1521};
1522
Libav master