Commit | Line | Data |
---|---|---|
2ba45a60 DM |
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 | ||
30 | #include "avcodec.h" | |
31 | #include "get_bits.h" | |
32 | #include "unary.h" | |
33 | #include "mpeg4audio.h" | |
34 | #include "bytestream.h" | |
35 | #include "bgmc.h" | |
36 | #include "bswapdsp.h" | |
37 | #include "internal.h" | |
38 | #include "libavutil/samplefmt.h" | |
39 | #include "libavutil/crc.h" | |
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 | */ | |
47 | static const int8_t parcor_rice_table[3][20][2] = { | |
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 | */ | |
68 | static const int16_t parcor_scaled_values[] = { | |
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 | */ | |
107 | static const uint8_t ltp_gain_values [4][4] = { | |
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 | */ | |
118 | static const int16_t mcc_weightings[] = { | |
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 | */ | |
128 | static const uint8_t tail_code[16][6] = { | |
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 | ||
148 | enum RA_Flag { | |
149 | RA_FLAG_NONE, | |
150 | RA_FLAG_FRAMES, | |
151 | RA_FLAG_HEADER | |
152 | }; | |
153 | ||
154 | ||
155 | typedef struct { | |
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) | |
162 | enum RA_Flag ra_flag; ///< indicates where the size of ra units is stored | |
163 | int adapt_order; ///< adaptive order: 1 = on, 0 = off | |
164 | int coef_table; ///< table index of Rice code parameters | |
165 | int long_term_prediction; ///< long term prediction (LTP): 1 = on, 0 = off | |
166 | int max_order; ///< maximum prediction order (0..1023) | |
167 | int block_switching; ///< number of block switching levels | |
168 | int bgmc; ///< "Block Gilbert-Moore Code": 1 = on, 0 = off (Rice coding only) | |
169 | int sb_part; ///< sub-block partition | |
170 | int joint_stereo; ///< joint stereo: 1 = on, 0 = off | |
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. | |
176 | int *chan_pos; ///< original channel positions | |
177 | int crc_enabled; ///< enable Cyclic Redundancy Checksum | |
178 | } ALSSpecificConfig; | |
179 | ||
180 | ||
181 | typedef struct { | |
182 | int stop_flag; | |
183 | int master_channel; | |
184 | int time_diff_flag; | |
185 | int time_diff_sign; | |
186 | int time_diff_index; | |
187 | int weighting[6]; | |
188 | } ALSChannelData; | |
189 | ||
190 | ||
191 | typedef struct { | |
192 | AVCodecContext *avctx; | |
193 | ALSSpecificConfig sconf; | |
194 | GetBitContext gb; | |
195 | BswapDSPContext bdsp; | |
196 | const AVCRC *crc_table; | |
197 | uint32_t crc_org; ///< CRC value of the original input data | |
198 | uint32_t crc; ///< CRC value calculated from decoded data | |
199 | unsigned int cur_frame_length; ///< length of the current frame to decode | |
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 | |
205 | uint8_t *bgmc_lut; ///< pointer at lookup tables used for BGMC | |
206 | int *bgmc_lut_status; ///< pointer at lookup table status flags used for BGMC | |
207 | int ltp_lag_length; ///< number of bits used for ltp lag value | |
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 | |
211 | int *store_prev_samples; ///< contains store_prev_samples 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 | |
215 | int *ltp_gain_buffer; ///< contains all gain values for ltp 5-tap filter | |
216 | int32_t **quant_cof; ///< quantized parcor coefficients for a channel | |
217 | int32_t *quant_cof_buffer; ///< contains all quantized parcor coefficients | |
218 | int32_t **lpc_cof; ///< coefficients of the direct form prediction filter for a channel | |
219 | int32_t *lpc_cof_buffer; ///< contains all coefficients of the direct form prediction filter | |
220 | int32_t *lpc_cof_reversed_buffer; ///< temporary buffer to set up a reversed versio of lpc_cof_buffer | |
221 | ALSChannelData **chan_data; ///< channel data for multi-channel correlation | |
222 | ALSChannelData *chan_data_buffer; ///< contains channel data for all channels | |
223 | int *reverted_channels; ///< stores a flag for each reverted channel | |
224 | int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block | |
225 | int32_t **raw_samples; ///< decoded raw samples for each channel | |
226 | int32_t *raw_buffer; ///< contains all decoded raw samples including carryover samples | |
227 | uint8_t *crc_buffer; ///< buffer of byte order corrected samples used for CRC check | |
228 | } ALSDecContext; | |
229 | ||
230 | ||
231 | typedef struct { | |
232 | unsigned int block_length; ///< number of samples within the block | |
233 | unsigned int ra_block; ///< if true, this is a random access block | |
234 | int *const_block; ///< if true, this is a constant value 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 | |
238 | int *store_prev_samples;///< if true, carryover samples have to be stored | |
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 | |
242 | int32_t *quant_cof; ///< quantized parcor coefficients | |
243 | int32_t *lpc_cof; ///< coefficients of the direct form prediction | |
244 | int32_t *raw_samples; ///< decoded raw samples / residuals for this block | |
245 | int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block | |
246 | int32_t *raw_other; ///< decoded raw samples of the other channel of a channel pair | |
247 | } ALSBlockData; | |
248 | ||
249 | ||
250 | static av_cold void dprint_specific_config(ALSDecContext *ctx) | |
251 | { | |
252 | #ifdef DEBUG | |
253 | AVCodecContext *avctx = ctx->avctx; | |
254 | ALSSpecificConfig *sconf = &ctx->sconf; | |
255 | ||
256 | av_dlog(avctx, "resolution = %i\n", sconf->resolution); | |
257 | av_dlog(avctx, "floating = %i\n", sconf->floating); | |
258 | av_dlog(avctx, "frame_length = %i\n", sconf->frame_length); | |
259 | av_dlog(avctx, "ra_distance = %i\n", sconf->ra_distance); | |
260 | av_dlog(avctx, "ra_flag = %i\n", sconf->ra_flag); | |
261 | av_dlog(avctx, "adapt_order = %i\n", sconf->adapt_order); | |
262 | av_dlog(avctx, "coef_table = %i\n", sconf->coef_table); | |
263 | av_dlog(avctx, "long_term_prediction = %i\n", sconf->long_term_prediction); | |
264 | av_dlog(avctx, "max_order = %i\n", sconf->max_order); | |
265 | av_dlog(avctx, "block_switching = %i\n", sconf->block_switching); | |
266 | av_dlog(avctx, "bgmc = %i\n", sconf->bgmc); | |
267 | av_dlog(avctx, "sb_part = %i\n", sconf->sb_part); | |
268 | av_dlog(avctx, "joint_stereo = %i\n", sconf->joint_stereo); | |
269 | av_dlog(avctx, "mc_coding = %i\n", sconf->mc_coding); | |
270 | av_dlog(avctx, "chan_config = %i\n", sconf->chan_config); | |
271 | av_dlog(avctx, "chan_sort = %i\n", sconf->chan_sort); | |
272 | av_dlog(avctx, "RLSLMS = %i\n", sconf->rlslms); | |
273 | av_dlog(avctx, "chan_config_info = %i\n", sconf->chan_config_info); | |
274 | #endif | |
275 | } | |
276 | ||
277 | ||
278 | /** Read an ALSSpecificConfig from a buffer into the output struct. | |
279 | */ | |
280 | static av_cold int read_specific_config(ALSDecContext *ctx) | |
281 | { | |
282 | GetBitContext gb; | |
283 | uint64_t ht_size; | |
284 | int i, config_offset; | |
285 | MPEG4AudioConfig m4ac = {0}; | |
286 | ALSSpecificConfig *sconf = &ctx->sconf; | |
287 | AVCodecContext *avctx = ctx->avctx; | |
288 | uint32_t als_id, header_size, trailer_size; | |
289 | int ret; | |
290 | ||
291 | if ((ret = init_get_bits8(&gb, avctx->extradata, avctx->extradata_size)) < 0) | |
292 | return ret; | |
293 | ||
294 | config_offset = avpriv_mpeg4audio_get_config(&m4ac, avctx->extradata, | |
295 | avctx->extradata_size * 8, 1); | |
296 | ||
297 | if (config_offset < 0) | |
298 | return AVERROR_INVALIDDATA; | |
299 | ||
300 | skip_bits_long(&gb, config_offset); | |
301 | ||
302 | if (get_bits_left(&gb) < (30 << 3)) | |
303 | return AVERROR_INVALIDDATA; | |
304 | ||
305 | // read the fixed items | |
306 | als_id = get_bits_long(&gb, 32); | |
307 | avctx->sample_rate = m4ac.sample_rate; | |
308 | skip_bits_long(&gb, 32); // sample rate already known | |
309 | sconf->samples = get_bits_long(&gb, 32); | |
310 | avctx->channels = m4ac.channels; | |
311 | skip_bits(&gb, 16); // number of channels already known | |
312 | skip_bits(&gb, 3); // skip file_type | |
313 | sconf->resolution = get_bits(&gb, 3); | |
314 | sconf->floating = get_bits1(&gb); | |
315 | sconf->msb_first = get_bits1(&gb); | |
316 | sconf->frame_length = get_bits(&gb, 16) + 1; | |
317 | sconf->ra_distance = get_bits(&gb, 8); | |
318 | sconf->ra_flag = get_bits(&gb, 2); | |
319 | sconf->adapt_order = get_bits1(&gb); | |
320 | sconf->coef_table = get_bits(&gb, 2); | |
321 | sconf->long_term_prediction = get_bits1(&gb); | |
322 | sconf->max_order = get_bits(&gb, 10); | |
323 | sconf->block_switching = get_bits(&gb, 2); | |
324 | sconf->bgmc = get_bits1(&gb); | |
325 | sconf->sb_part = get_bits1(&gb); | |
326 | sconf->joint_stereo = get_bits1(&gb); | |
327 | sconf->mc_coding = get_bits1(&gb); | |
328 | sconf->chan_config = get_bits1(&gb); | |
329 | sconf->chan_sort = get_bits1(&gb); | |
330 | sconf->crc_enabled = get_bits1(&gb); | |
331 | sconf->rlslms = get_bits1(&gb); | |
332 | skip_bits(&gb, 5); // skip 5 reserved bits | |
333 | skip_bits1(&gb); // skip aux_data_enabled | |
334 | ||
335 | ||
336 | // check for ALSSpecificConfig struct | |
337 | if (als_id != MKBETAG('A','L','S','\0')) | |
338 | return AVERROR_INVALIDDATA; | |
339 | ||
340 | ctx->cur_frame_length = sconf->frame_length; | |
341 | ||
342 | // read channel config | |
343 | if (sconf->chan_config) | |
344 | sconf->chan_config_info = get_bits(&gb, 16); | |
345 | // TODO: use this to set avctx->channel_layout | |
346 | ||
347 | ||
348 | // read channel sorting | |
349 | if (sconf->chan_sort && avctx->channels > 1) { | |
350 | int chan_pos_bits = av_ceil_log2(avctx->channels); | |
351 | int bits_needed = avctx->channels * chan_pos_bits + 7; | |
352 | if (get_bits_left(&gb) < bits_needed) | |
353 | return AVERROR_INVALIDDATA; | |
354 | ||
355 | if (!(sconf->chan_pos = av_malloc(avctx->channels * sizeof(*sconf->chan_pos)))) | |
356 | return AVERROR(ENOMEM); | |
357 | ||
358 | ctx->cs_switch = 1; | |
359 | ||
360 | for (i = 0; i < avctx->channels; i++) { | |
361 | int idx; | |
362 | ||
363 | idx = get_bits(&gb, chan_pos_bits); | |
364 | if (idx >= avctx->channels) { | |
365 | av_log(avctx, AV_LOG_WARNING, "Invalid channel reordering.\n"); | |
366 | ctx->cs_switch = 0; | |
367 | break; | |
368 | } | |
369 | sconf->chan_pos[idx] = i; | |
370 | } | |
371 | ||
372 | align_get_bits(&gb); | |
373 | } | |
374 | ||
375 | ||
376 | // read fixed header and trailer sizes, | |
377 | // if size = 0xFFFFFFFF then there is no data field! | |
378 | if (get_bits_left(&gb) < 64) | |
379 | return AVERROR_INVALIDDATA; | |
380 | ||
381 | header_size = get_bits_long(&gb, 32); | |
382 | trailer_size = get_bits_long(&gb, 32); | |
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 | |
392 | if (get_bits_left(&gb) < ht_size) | |
393 | return AVERROR_INVALIDDATA; | |
394 | ||
395 | if (ht_size > INT32_MAX) | |
396 | return AVERROR_PATCHWELCOME; | |
397 | ||
398 | skip_bits_long(&gb, ht_size); | |
399 | ||
400 | ||
401 | // initialize CRC calculation | |
402 | if (sconf->crc_enabled) { | |
403 | if (get_bits_left(&gb) < 32) | |
404 | return AVERROR_INVALIDDATA; | |
405 | ||
406 | if (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL)) { | |
407 | ctx->crc_table = av_crc_get_table(AV_CRC_32_IEEE_LE); | |
408 | ctx->crc = 0xFFFFFFFF; | |
409 | ctx->crc_org = ~get_bits_long(&gb, 32); | |
410 | } else | |
411 | skip_bits_long(&gb, 32); | |
412 | } | |
413 | ||
414 | ||
415 | // no need to read the rest of ALSSpecificConfig (ra_unit_size & aux data) | |
416 | ||
417 | dprint_specific_config(ctx); | |
418 | ||
419 | return 0; | |
420 | } | |
421 | ||
422 | ||
423 | /** Check the ALSSpecificConfig for unsupported features. | |
424 | */ | |
425 | static int check_specific_config(ALSDecContext *ctx) | |
426 | { | |
427 | ALSSpecificConfig *sconf = &ctx->sconf; | |
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 | ||
440 | MISSING_ERR(sconf->floating, "Floating point decoding", AVERROR_PATCHWELCOME); | |
441 | MISSING_ERR(sconf->rlslms, "Adaptive RLS-LMS prediction", AVERROR_PATCHWELCOME); | |
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 | */ | |
450 | static void parse_bs_info(const uint32_t bs_info, unsigned int n, | |
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; | |
459 | parse_bs_info(bs_info, n + 1, div, div_blocks, num_blocks); | |
460 | parse_bs_info(bs_info, n + 2, div, div_blocks, num_blocks); | |
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 | */ | |
473 | static int32_t decode_rice(GetBitContext *gb, unsigned int k) | |
474 | { | |
475 | int max = get_bits_left(gb) - k; | |
476 | int q = get_unary(gb, 0, max); | |
477 | int r = k ? get_bits1(gb) : !(q & 1); | |
478 | ||
479 | if (k > 1) { | |
480 | q <<= (k - 1); | |
481 | q += get_bits_long(gb, 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 | */ | |
491 | static void parcor_to_lpc(unsigned int k, const int32_t *par, int32_t *cof) | |
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 | */ | |
511 | static void get_block_sizes(ALSDecContext *ctx, unsigned int *div_blocks, | |
512 | uint32_t *bs_info) | |
513 | { | |
514 | ALSSpecificConfig *sconf = &ctx->sconf; | |
515 | GetBitContext *gb = &ctx->gb; | |
516 | unsigned int *ptr_div_blocks = div_blocks; | |
517 | unsigned int b; | |
518 | ||
519 | if (sconf->block_switching) { | |
520 | unsigned int bs_info_len = 1 << (sconf->block_switching + 2); | |
521 | *bs_info = get_bits_long(gb, bs_info_len); | |
522 | *bs_info <<= (32 - bs_info_len); | |
523 | } | |
524 | ||
525 | ctx->num_blocks = 0; | |
526 | parse_bs_info(*bs_info, 0, 0, &ptr_div_blocks, &ctx->num_blocks); | |
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 | ||
543 | for (b = 0; b < ctx->num_blocks; b++) | |
544 | div_blocks[b] = ctx->sconf.frame_length >> div_blocks[b]; | |
545 | ||
546 | if (ctx->cur_frame_length != ctx->sconf.frame_length) { | |
547 | unsigned int remaining = ctx->cur_frame_length; | |
548 | ||
549 | for (b = 0; b < ctx->num_blocks; b++) { | |
550 | if (remaining <= div_blocks[b]) { | |
551 | div_blocks[b] = remaining; | |
552 | ctx->num_blocks = b + 1; | |
553 | break; | |
554 | } | |
555 | ||
556 | remaining -= div_blocks[b]; | |
557 | } | |
558 | } | |
559 | } | |
560 | ||
561 | ||
562 | /** Read the block data for a constant block | |
563 | */ | |
564 | static int read_const_block_data(ALSDecContext *ctx, ALSBlockData *bd) | |
565 | { | |
566 | ALSSpecificConfig *sconf = &ctx->sconf; | |
567 | AVCodecContext *avctx = ctx->avctx; | |
568 | GetBitContext *gb = &ctx->gb; | |
569 | ||
570 | if (bd->block_length <= 0) | |
571 | return AVERROR_INVALIDDATA; | |
572 | ||
573 | *bd->raw_samples = 0; | |
574 | *bd->const_block = get_bits1(gb); // 1 = constant value, 0 = zero block (silence) | |
575 | bd->js_blocks = get_bits1(gb); | |
576 | ||
577 | // skip 5 reserved bits | |
578 | skip_bits(gb, 5); | |
579 | ||
580 | if (*bd->const_block) { | |
581 | unsigned int const_val_bits = sconf->floating ? 24 : avctx->bits_per_raw_sample; | |
582 | *bd->raw_samples = get_sbits_long(gb, const_val_bits); | |
583 | } | |
584 | ||
585 | // ensure constant block decoding by reusing this field | |
586 | *bd->const_block = 1; | |
587 | ||
588 | return 0; | |
589 | } | |
590 | ||
591 | ||
592 | /** Decode the block data for a constant block | |
593 | */ | |
594 | static void decode_const_block_data(ALSDecContext *ctx, ALSBlockData *bd) | |
595 | { | |
596 | int smp = bd->block_length - 1; | |
597 | int32_t val = *bd->raw_samples; | |
598 | int32_t *dst = bd->raw_samples + 1; | |
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 | */ | |
608 | static int read_var_block_data(ALSDecContext *ctx, ALSBlockData *bd) | |
609 | { | |
610 | ALSSpecificConfig *sconf = &ctx->sconf; | |
611 | AVCodecContext *avctx = ctx->avctx; | |
612 | GetBitContext *gb = &ctx->gb; | |
613 | unsigned int k; | |
614 | unsigned int s[8]; | |
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; | |
620 | int32_t *quant_cof = bd->quant_cof; | |
621 | int32_t *current_res; | |
622 | ||
623 | ||
624 | // ensure variable block decoding by reusing this field | |
625 | *bd->const_block = 0; | |
626 | ||
627 | *bd->opt_order = 1; | |
628 | bd->js_blocks = get_bits1(gb); | |
629 | ||
630 | opt_order = *bd->opt_order; | |
631 | ||
632 | // determine the number of subblocks for entropy decoding | |
633 | if (!sconf->bgmc && !sconf->sb_part) { | |
634 | log2_sub_blocks = 0; | |
635 | } else { | |
636 | if (sconf->bgmc && sconf->sb_part) | |
637 | log2_sub_blocks = get_bits(gb, 2); | |
638 | else | |
639 | log2_sub_blocks = 2 * get_bits1(gb); | |
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 | |
646 | if (bd->block_length & (sub_blocks - 1)) { | |
647 | av_log(avctx, AV_LOG_WARNING, | |
648 | "Block length is not evenly divisible by the number of subblocks.\n"); | |
649 | return AVERROR_INVALIDDATA; | |
650 | } | |
651 | ||
652 | sb_length = bd->block_length >> log2_sub_blocks; | |
653 | ||
654 | if (sconf->bgmc) { | |
655 | s[0] = get_bits(gb, 8 + (sconf->resolution > 1)); | |
656 | for (k = 1; k < sub_blocks; k++) | |
657 | s[k] = s[k - 1] + decode_rice(gb, 2); | |
658 | ||
659 | for (k = 0; k < sub_blocks; k++) { | |
660 | sx[k] = s[k] & 0x0F; | |
661 | s [k] >>= 4; | |
662 | } | |
663 | } else { | |
664 | s[0] = get_bits(gb, 4 + (sconf->resolution > 1)); | |
665 | for (k = 1; k < sub_blocks; k++) | |
666 | s[k] = s[k - 1] + decode_rice(gb, 0); | |
667 | } | |
668 | for (k = 1; k < sub_blocks; k++) | |
669 | if (s[k] > 32) { | |
670 | av_log(avctx, AV_LOG_ERROR, "k invalid for rice code.\n"); | |
671 | return AVERROR_INVALIDDATA; | |
672 | } | |
673 | ||
674 | if (get_bits1(gb)) | |
675 | *bd->shift_lsbs = get_bits(gb, 4) + 1; | |
676 | ||
677 | *bd->store_prev_samples = (bd->js_blocks && bd->raw_other) || *bd->shift_lsbs; | |
678 | ||
679 | ||
680 | if (!sconf->rlslms) { | |
681 | if (sconf->adapt_order) { | |
682 | int opt_order_length = av_ceil_log2(av_clip((bd->block_length >> 3) - 1, | |
683 | 2, sconf->max_order + 1)); | |
684 | *bd->opt_order = get_bits(gb, opt_order_length); | |
685 | if (*bd->opt_order > sconf->max_order) { | |
686 | *bd->opt_order = sconf->max_order; | |
687 | av_log(avctx, AV_LOG_ERROR, "Predictor order too large.\n"); | |
688 | return AVERROR_INVALIDDATA; | |
689 | } | |
690 | } else { | |
691 | *bd->opt_order = sconf->max_order; | |
692 | } | |
693 | if (*bd->opt_order > bd->block_length) { | |
694 | *bd->opt_order = bd->block_length; | |
695 | av_log(avctx, AV_LOG_ERROR, "Predictor order too large.\n"); | |
696 | return AVERROR_INVALIDDATA; | |
697 | } | |
698 | opt_order = *bd->opt_order; | |
699 | ||
700 | if (opt_order) { | |
701 | int add_base; | |
702 | ||
703 | if (sconf->coef_table == 3) { | |
704 | add_base = 0x7F; | |
705 | ||
706 | // read coefficient 0 | |
707 | quant_cof[0] = 32 * parcor_scaled_values[get_bits(gb, 7)]; | |
708 | ||
709 | // read coefficient 1 | |
710 | if (opt_order > 1) | |
711 | quant_cof[1] = -32 * parcor_scaled_values[get_bits(gb, 7)]; | |
712 | ||
713 | // read coefficients 2 to opt_order | |
714 | for (k = 2; k < opt_order; k++) | |
715 | quant_cof[k] = get_bits(gb, 7); | |
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++) { | |
723 | int rice_param = parcor_rice_table[sconf->coef_table][k][1]; | |
724 | int offset = parcor_rice_table[sconf->coef_table][k][0]; | |
725 | quant_cof[k] = decode_rice(gb, rice_param) + offset; | |
726 | if (quant_cof[k] < -64 || quant_cof[k] > 63) { | |
727 | av_log(avctx, AV_LOG_ERROR, | |
728 | "quant_cof %"PRIu32" is out of range.\n", | |
729 | quant_cof[k]); | |
730 | return AVERROR_INVALIDDATA; | |
731 | } | |
732 | } | |
733 | ||
734 | // read coefficients 20 to 126 | |
735 | k_max = FFMIN(opt_order, 127); | |
736 | for (; k < k_max; k++) | |
737 | quant_cof[k] = decode_rice(gb, 2) + (k & 1); | |
738 | ||
739 | // read coefficients 127 to opt_order | |
740 | for (; k < opt_order; k++) | |
741 | quant_cof[k] = decode_rice(gb, 1); | |
742 | ||
743 | quant_cof[0] = 32 * parcor_scaled_values[quant_cof[0] + 64]; | |
744 | ||
745 | if (opt_order > 1) | |
746 | quant_cof[1] = -32 * parcor_scaled_values[quant_cof[1] + 64]; | |
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 | |
755 | if (sconf->long_term_prediction) { | |
756 | *bd->use_ltp = get_bits1(gb); | |
757 | ||
758 | if (*bd->use_ltp) { | |
759 | int r, c; | |
760 | ||
761 | bd->ltp_gain[0] = decode_rice(gb, 1) << 3; | |
762 | bd->ltp_gain[1] = decode_rice(gb, 2) << 3; | |
763 | ||
764 | r = get_unary(gb, 0, 3); | |
765 | c = get_bits(gb, 2); | |
766 | bd->ltp_gain[2] = ltp_gain_values[r][c]; | |
767 | ||
768 | bd->ltp_gain[3] = decode_rice(gb, 2) << 3; | |
769 | bd->ltp_gain[4] = decode_rice(gb, 1) << 3; | |
770 | ||
771 | *bd->ltp_lag = get_bits(gb, ctx->ltp_lag_length); | |
772 | *bd->ltp_lag += FFMAX(4, opt_order + 1); | |
773 | } | |
774 | } | |
775 | ||
776 | // read first value and residuals in case of a random access block | |
777 | if (bd->ra_block) { | |
778 | if (opt_order) | |
779 | bd->raw_samples[0] = decode_rice(gb, avctx->bits_per_raw_sample - 4); | |
780 | if (opt_order > 1) | |
781 | bd->raw_samples[1] = decode_rice(gb, FFMIN(s[0] + 3, ctx->s_max)); | |
782 | if (opt_order > 2) | |
783 | bd->raw_samples[2] = decode_rice(gb, FFMIN(s[0] + 1, ctx->s_max)); | |
784 | ||
785 | start = FFMIN(opt_order, 3); | |
786 | } | |
787 | ||
788 | // read all residuals | |
789 | if (sconf->bgmc) { | |
790 | int delta[8]; | |
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; | |
797 | unsigned int value; | |
798 | ||
799 | ff_bgmc_decode_init(gb, &high, &low, &value); | |
800 | ||
801 | current_res = bd->raw_samples + start; | |
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 | ||
809 | ff_bgmc_decode(gb, sb_len, current_res, | |
810 | delta[sb], sx[sb], &high, &low, &value, ctx->bgmc_lut, ctx->bgmc_lut_status); | |
811 | ||
812 | current_res += sb_len; | |
813 | } | |
814 | ||
815 | ff_bgmc_decode_end(gb); | |
816 | ||
817 | ||
818 | // read least significant bits and tails | |
819 | current_res = bd->raw_samples + start; | |
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++) { | |
827 | int32_t res = *current_res; | |
828 | ||
829 | if (res == cur_tail_code) { | |
830 | unsigned int max_msb = (2 + (sx[sb] > 2) + (sx[sb] > 10)) | |
831 | << (5 - delta[sb]); | |
832 | ||
833 | res = decode_rice(gb, cur_s); | |
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; | |
851 | res |= get_bits_long(gb, cur_k); | |
852 | } | |
853 | } | |
854 | ||
855 | *current_res++ = res; | |
856 | } | |
857 | } | |
858 | } else { | |
859 | current_res = bd->raw_samples + start; | |
860 | ||
861 | for (sb = 0; sb < sub_blocks; sb++, start = 0) | |
862 | for (; start < sb_length; start++) | |
863 | *current_res++ = decode_rice(gb, s[sb]); | |
864 | } | |
865 | ||
866 | if (!sconf->mc_coding || ctx->js_switch) | |
867 | align_get_bits(gb); | |
868 | ||
869 | return 0; | |
870 | } | |
871 | ||
872 | ||
873 | /** Decode the block data for a non-constant block | |
874 | */ | |
875 | static int decode_var_block_data(ALSDecContext *ctx, ALSBlockData *bd) | |
876 | { | |
877 | ALSSpecificConfig *sconf = &ctx->sconf; | |
878 | unsigned int block_length = bd->block_length; | |
879 | unsigned int smp = 0; | |
880 | unsigned int k; | |
881 | int opt_order = *bd->opt_order; | |
882 | int sb; | |
883 | int64_t y; | |
884 | int32_t *quant_cof = bd->quant_cof; | |
885 | int32_t *lpc_cof = bd->lpc_cof; | |
886 | int32_t *raw_samples = bd->raw_samples; | |
887 | int32_t *raw_samples_end = bd->raw_samples + bd->block_length; | |
888 | int32_t *lpc_cof_reversed = ctx->lpc_cof_reversed_buffer; | |
889 | ||
890 | // reverse long-term prediction | |
891 | if (*bd->use_ltp) { | |
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++) | |
904 | y += MUL64(bd->ltp_gain[tab], raw_samples[base]); | |
905 | ||
906 | raw_samples[ltp_smp] += y >> 7; | |
907 | } | |
908 | } | |
909 | ||
910 | // reconstruct all samples from residuals | |
911 | if (bd->ra_block) { | |
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; | |
919 | parcor_to_lpc(smp, quant_cof, lpc_cof); | |
920 | } | |
921 | } else { | |
922 | for (k = 0; k < opt_order; k++) | |
923 | parcor_to_lpc(k, quant_cof, lpc_cof); | |
924 | ||
925 | // store previous samples in case that they have to be altered | |
926 | if (*bd->store_prev_samples) | |
927 | memcpy(bd->prev_raw_samples, raw_samples - sconf->max_order, | |
928 | sizeof(*bd->prev_raw_samples) * sconf->max_order); | |
929 | ||
930 | // reconstruct difference signal for prediction (joint-stereo) | |
931 | if (bd->js_blocks && bd->raw_other) { | |
932 | int32_t *left, *right; | |
933 | ||
934 | if (bd->raw_other > raw_samples) { // D = R - L | |
935 | left = raw_samples; | |
936 | right = bd->raw_other; | |
937 | } else { // D = R - L | |
938 | left = bd->raw_other; | |
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 | |
947 | if (*bd->shift_lsbs) | |
948 | for (sb = -1; sb >= -sconf->max_order; sb--) | |
949 | raw_samples[sb] >>= *bd->shift_lsbs; | |
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 | |
959 | raw_samples = bd->raw_samples + smp; | |
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 | ||
971 | raw_samples = bd->raw_samples; | |
972 | ||
973 | // restore previous samples in case that they have been altered | |
974 | if (*bd->store_prev_samples) | |
975 | memcpy(raw_samples - sconf->max_order, bd->prev_raw_samples, | |
976 | sizeof(*raw_samples) * sconf->max_order); | |
977 | ||
978 | return 0; | |
979 | } | |
980 | ||
981 | ||
982 | /** Read the block data. | |
983 | */ | |
984 | static int read_block(ALSDecContext *ctx, ALSBlockData *bd) | |
985 | { | |
986 | int ret; | |
987 | GetBitContext *gb = &ctx->gb; | |
988 | ||
989 | *bd->shift_lsbs = 0; | |
990 | // read block type flag and read the samples accordingly | |
991 | if (get_bits1(gb)) { | |
992 | ret = read_var_block_data(ctx, bd); | |
993 | } else { | |
994 | ret = read_const_block_data(ctx, bd); | |
995 | } | |
996 | ||
997 | return ret; | |
998 | } | |
999 | ||
1000 | ||
1001 | /** Decode the block data. | |
1002 | */ | |
1003 | static int decode_block(ALSDecContext *ctx, ALSBlockData *bd) | |
1004 | { | |
1005 | unsigned int smp; | |
1006 | int ret = 0; | |
1007 | ||
1008 | // read block type flag and read the samples accordingly | |
1009 | if (*bd->const_block) | |
1010 | decode_const_block_data(ctx, bd); | |
1011 | else | |
1012 | ret = decode_var_block_data(ctx, bd); // always return 0 | |
1013 | ||
1014 | if (ret < 0) | |
1015 | return ret; | |
1016 | ||
1017 | // TODO: read RLSLMS extension data | |
1018 | ||
1019 | if (*bd->shift_lsbs) | |
1020 | for (smp = 0; smp < bd->block_length; smp++) | |
1021 | bd->raw_samples[smp] <<= *bd->shift_lsbs; | |
1022 | ||
1023 | return 0; | |
1024 | } | |
1025 | ||
1026 | ||
1027 | /** Read and decode block data successively. | |
1028 | */ | |
1029 | static int read_decode_block(ALSDecContext *ctx, ALSBlockData *bd) | |
1030 | { | |
1031 | int ret; | |
1032 | ||
1033 | if ((ret = read_block(ctx, bd)) < 0) | |
1034 | return ret; | |
1035 | ||
1036 | return decode_block(ctx, bd); | |
1037 | } | |
1038 | ||
1039 | ||
1040 | /** Compute the number of samples left to decode for the current frame and | |
1041 | * sets these samples to zero. | |
1042 | */ | |
1043 | static void zero_remaining(unsigned int b, unsigned int b_max, | |
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 | */ | |
1058 | static int decode_blocks_ind(ALSDecContext *ctx, unsigned int ra_frame, | |
1059 | unsigned int c, const unsigned int *div_blocks, | |
1060 | unsigned int *js_blocks) | |
1061 | { | |
1062 | int ret; | |
1063 | unsigned int b; | |
1064 | ALSBlockData bd = { 0 }; | |
1065 | ||
1066 | bd.ra_block = ra_frame; | |
1067 | bd.const_block = ctx->const_block; | |
1068 | bd.shift_lsbs = ctx->shift_lsbs; | |
1069 | bd.opt_order = ctx->opt_order; | |
1070 | bd.store_prev_samples = ctx->store_prev_samples; | |
1071 | bd.use_ltp = ctx->use_ltp; | |
1072 | bd.ltp_lag = ctx->ltp_lag; | |
1073 | bd.ltp_gain = ctx->ltp_gain[0]; | |
1074 | bd.quant_cof = ctx->quant_cof[0]; | |
1075 | bd.lpc_cof = ctx->lpc_cof[0]; | |
1076 | bd.prev_raw_samples = ctx->prev_raw_samples; | |
1077 | bd.raw_samples = ctx->raw_samples[c]; | |
1078 | ||
1079 | ||
1080 | for (b = 0; b < ctx->num_blocks; b++) { | |
1081 | bd.block_length = div_blocks[b]; | |
1082 | ||
1083 | if ((ret = read_decode_block(ctx, &bd)) < 0) { | |
1084 | // damaged block, write zero for the rest of the frame | |
1085 | zero_remaining(b, ctx->num_blocks, div_blocks, bd.raw_samples); | |
1086 | return ret; | |
1087 | } | |
1088 | bd.raw_samples += div_blocks[b]; | |
1089 | bd.ra_block = 0; | |
1090 | } | |
1091 | ||
1092 | return 0; | |
1093 | } | |
1094 | ||
1095 | ||
1096 | /** Decode blocks dependently. | |
1097 | */ | |
1098 | static int decode_blocks(ALSDecContext *ctx, unsigned int ra_frame, | |
1099 | unsigned int c, const unsigned int *div_blocks, | |
1100 | unsigned int *js_blocks) | |
1101 | { | |
1102 | ALSSpecificConfig *sconf = &ctx->sconf; | |
1103 | unsigned int offset = 0; | |
1104 | unsigned int b; | |
1105 | int ret; | |
1106 | ALSBlockData bd[2] = { { 0 } }; | |
1107 | ||
1108 | bd[0].ra_block = ra_frame; | |
1109 | bd[0].const_block = ctx->const_block; | |
1110 | bd[0].shift_lsbs = ctx->shift_lsbs; | |
1111 | bd[0].opt_order = ctx->opt_order; | |
1112 | bd[0].store_prev_samples = ctx->store_prev_samples; | |
1113 | bd[0].use_ltp = ctx->use_ltp; | |
1114 | bd[0].ltp_lag = ctx->ltp_lag; | |
1115 | bd[0].ltp_gain = ctx->ltp_gain[0]; | |
1116 | bd[0].quant_cof = ctx->quant_cof[0]; | |
1117 | bd[0].lpc_cof = ctx->lpc_cof[0]; | |
1118 | bd[0].prev_raw_samples = ctx->prev_raw_samples; | |
1119 | bd[0].js_blocks = *js_blocks; | |
1120 | ||
1121 | bd[1].ra_block = ra_frame; | |
1122 | bd[1].const_block = ctx->const_block; | |
1123 | bd[1].shift_lsbs = ctx->shift_lsbs; | |
1124 | bd[1].opt_order = ctx->opt_order; | |
1125 | bd[1].store_prev_samples = ctx->store_prev_samples; | |
1126 | bd[1].use_ltp = ctx->use_ltp; | |
1127 | bd[1].ltp_lag = ctx->ltp_lag; | |
1128 | bd[1].ltp_gain = ctx->ltp_gain[0]; | |
1129 | bd[1].quant_cof = ctx->quant_cof[0]; | |
1130 | bd[1].lpc_cof = ctx->lpc_cof[0]; | |
1131 | bd[1].prev_raw_samples = ctx->prev_raw_samples; | |
1132 | bd[1].js_blocks = *(js_blocks + 1); | |
1133 | ||
1134 | // decode all blocks | |
1135 | for (b = 0; b < ctx->num_blocks; b++) { | |
1136 | unsigned int s; | |
1137 | ||
1138 | bd[0].block_length = div_blocks[b]; | |
1139 | bd[1].block_length = div_blocks[b]; | |
1140 | ||
1141 | bd[0].raw_samples = ctx->raw_samples[c ] + offset; | |
1142 | bd[1].raw_samples = ctx->raw_samples[c + 1] + offset; | |
1143 | ||
1144 | bd[0].raw_other = bd[1].raw_samples; | |
1145 | bd[1].raw_other = bd[0].raw_samples; | |
1146 | ||
1147 | if ((ret = read_decode_block(ctx, &bd[0])) < 0 || | |
1148 | (ret = read_decode_block(ctx, &bd[1])) < 0) | |
1149 | goto fail; | |
1150 | ||
1151 | // reconstruct joint-stereo blocks | |
1152 | if (bd[0].js_blocks) { | |
1153 | if (bd[1].js_blocks) | |
1154 | av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel pair.\n"); | |
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]; | |
1164 | bd[0].ra_block = 0; | |
1165 | bd[1].ra_block = 0; | |
1166 | } | |
1167 | ||
1168 | // store carryover raw samples, | |
1169 | // the others channel raw samples are stored by the calling function. | |
1170 | memmove(ctx->raw_samples[c] - sconf->max_order, | |
1171 | ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
1172 | sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
1173 | ||
1174 | return 0; | |
1175 | fail: | |
1176 | // damaged block, write zero for the rest of the frame | |
1177 | zero_remaining(b, ctx->num_blocks, div_blocks, bd[0].raw_samples); | |
1178 | zero_remaining(b, ctx->num_blocks, div_blocks, bd[1].raw_samples); | |
1179 | return ret; | |
1180 | } | |
1181 | ||
1182 | static inline int als_weighting(GetBitContext *gb, int k, int off) | |
1183 | { | |
1184 | int idx = av_clip(decode_rice(gb, k) + off, | |
1185 | 0, FF_ARRAY_ELEMS(mcc_weightings) - 1); | |
1186 | return mcc_weightings[idx]; | |
1187 | } | |
1188 | ||
1189 | /** Read the channel data. | |
1190 | */ | |
1191 | static int read_channel_data(ALSDecContext *ctx, ALSChannelData *cd, int c) | |
1192 | { | |
1193 | GetBitContext *gb = &ctx->gb; | |
1194 | ALSChannelData *current = cd; | |
1195 | unsigned int channels = ctx->avctx->channels; | |
1196 | int entries = 0; | |
1197 | ||
1198 | while (entries < channels && !(current->stop_flag = get_bits1(gb))) { | |
1199 | current->master_channel = get_bits_long(gb, av_ceil_log2(channels)); | |
1200 | ||
1201 | if (current->master_channel >= channels) { | |
1202 | av_log(ctx->avctx, AV_LOG_ERROR, "Invalid master channel.\n"); | |
1203 | return AVERROR_INVALIDDATA; | |
1204 | } | |
1205 | ||
1206 | if (current->master_channel != c) { | |
1207 | current->time_diff_flag = get_bits1(gb); | |
1208 | current->weighting[0] = als_weighting(gb, 1, 16); | |
1209 | current->weighting[1] = als_weighting(gb, 2, 14); | |
1210 | current->weighting[2] = als_weighting(gb, 1, 16); | |
1211 | ||
1212 | if (current->time_diff_flag) { | |
1213 | current->weighting[3] = als_weighting(gb, 1, 16); | |
1214 | current->weighting[4] = als_weighting(gb, 1, 16); | |
1215 | current->weighting[5] = als_weighting(gb, 1, 16); | |
1216 | ||
1217 | current->time_diff_sign = get_bits1(gb); | |
1218 | current->time_diff_index = get_bits(gb, ctx->ltp_lag_length - 3) + 3; | |
1219 | } | |
1220 | } | |
1221 | ||
1222 | current++; | |
1223 | entries++; | |
1224 | } | |
1225 | ||
1226 | if (entries == channels) { | |
1227 | av_log(ctx->avctx, AV_LOG_ERROR, "Damaged channel data.\n"); | |
1228 | return AVERROR_INVALIDDATA; | |
1229 | } | |
1230 | ||
1231 | align_get_bits(gb); | |
1232 | return 0; | |
1233 | } | |
1234 | ||
1235 | ||
1236 | /** Recursively reverts the inter-channel correlation for a block. | |
1237 | */ | |
1238 | static int revert_channel_correlation(ALSDecContext *ctx, ALSBlockData *bd, | |
1239 | ALSChannelData **cd, int *reverted, | |
1240 | unsigned int offset, int c) | |
1241 | { | |
1242 | ALSChannelData *ch = cd[c]; | |
1243 | unsigned int dep = 0; | |
1244 | unsigned int channels = ctx->avctx->channels; | |
1245 | ||
1246 | if (reverted[c]) | |
1247 | return 0; | |
1248 | ||
1249 | reverted[c] = 1; | |
1250 | ||
1251 | while (dep < channels && !ch[dep].stop_flag) { | |
1252 | revert_channel_correlation(ctx, bd, cd, reverted, offset, | |
1253 | ch[dep].master_channel); | |
1254 | ||
1255 | dep++; | |
1256 | } | |
1257 | ||
1258 | if (dep == channels) { | |
1259 | av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel correlation.\n"); | |
1260 | return AVERROR_INVALIDDATA; | |
1261 | } | |
1262 | ||
1263 | bd->const_block = ctx->const_block + c; | |
1264 | bd->shift_lsbs = ctx->shift_lsbs + c; | |
1265 | bd->opt_order = ctx->opt_order + c; | |
1266 | bd->store_prev_samples = ctx->store_prev_samples + c; | |
1267 | bd->use_ltp = ctx->use_ltp + c; | |
1268 | bd->ltp_lag = ctx->ltp_lag + c; | |
1269 | bd->ltp_gain = ctx->ltp_gain[c]; | |
1270 | bd->lpc_cof = ctx->lpc_cof[c]; | |
1271 | bd->quant_cof = ctx->quant_cof[c]; | |
1272 | bd->raw_samples = ctx->raw_samples[c] + offset; | |
1273 | ||
1274 | for (dep = 0; !ch[dep].stop_flag; dep++) { | |
1275 | unsigned int smp; | |
1276 | unsigned int begin = 1; | |
1277 | unsigned int end = bd->block_length - 1; | |
1278 | int64_t y; | |
1279 | int32_t *master = ctx->raw_samples[ch[dep].master_channel] + offset; | |
1280 | ||
1281 | if (ch[dep].master_channel == c) | |
1282 | continue; | |
1283 | ||
1284 | if (ch[dep].time_diff_flag) { | |
1285 | int t = ch[dep].time_diff_index; | |
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 | ||
1303 | bd->raw_samples[smp] += y >> 7; | |
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 | ||
1312 | bd->raw_samples[smp] += y >> 7; | |
1313 | } | |
1314 | } | |
1315 | } | |
1316 | ||
1317 | return 0; | |
1318 | } | |
1319 | ||
1320 | ||
1321 | /** Read the frame data. | |
1322 | */ | |
1323 | static int read_frame_data(ALSDecContext *ctx, unsigned int ra_frame) | |
1324 | { | |
1325 | ALSSpecificConfig *sconf = &ctx->sconf; | |
1326 | AVCodecContext *avctx = ctx->avctx; | |
1327 | GetBitContext *gb = &ctx->gb; | |
1328 | unsigned int div_blocks[32]; ///< block sizes. | |
1329 | unsigned int c; | |
1330 | unsigned int js_blocks[2]; | |
1331 | uint32_t bs_info = 0; | |
1332 | int ret; | |
1333 | ||
1334 | // skip the size of the ra unit if present in the frame | |
1335 | if (sconf->ra_flag == RA_FLAG_FRAMES && ra_frame) | |
1336 | skip_bits_long(gb, 32); | |
1337 | ||
1338 | if (sconf->mc_coding && sconf->joint_stereo) { | |
1339 | ctx->js_switch = get_bits1(gb); | |
1340 | align_get_bits(gb); | |
1341 | } | |
1342 | ||
1343 | if (!sconf->mc_coding || ctx->js_switch) { | |
1344 | int independent_bs = !sconf->joint_stereo; | |
1345 | ||
1346 | for (c = 0; c < avctx->channels; c++) { | |
1347 | js_blocks[0] = 0; | |
1348 | js_blocks[1] = 0; | |
1349 | ||
1350 | get_block_sizes(ctx, div_blocks, &bs_info); | |
1351 | ||
1352 | // if joint_stereo and block_switching is set, independent decoding | |
1353 | // is signaled via the first bit of bs_info | |
1354 | if (sconf->joint_stereo && sconf->block_switching) | |
1355 | if (bs_info >> 31) | |
1356 | independent_bs = 2; | |
1357 | ||
1358 | // if this is the last channel, it has to be decoded independently | |
1359 | if (c == avctx->channels - 1) | |
1360 | independent_bs = 1; | |
1361 | ||
1362 | if (independent_bs) { | |
1363 | ret = decode_blocks_ind(ctx, ra_frame, c, | |
1364 | div_blocks, js_blocks); | |
1365 | if (ret < 0) | |
1366 | return ret; | |
1367 | independent_bs--; | |
1368 | } else { | |
1369 | ret = decode_blocks(ctx, ra_frame, c, div_blocks, js_blocks); | |
1370 | if (ret < 0) | |
1371 | return ret; | |
1372 | ||
1373 | c++; | |
1374 | } | |
1375 | ||
1376 | // store carryover raw samples | |
1377 | memmove(ctx->raw_samples[c] - sconf->max_order, | |
1378 | ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
1379 | sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
1380 | } | |
1381 | } else { // multi-channel coding | |
1382 | ALSBlockData bd = { 0 }; | |
1383 | int b, ret; | |
1384 | int *reverted_channels = ctx->reverted_channels; | |
1385 | unsigned int offset = 0; | |
1386 | ||
1387 | for (c = 0; c < avctx->channels; c++) | |
1388 | if (ctx->chan_data[c] < ctx->chan_data_buffer) { | |
1389 | av_log(ctx->avctx, AV_LOG_ERROR, "Invalid channel data.\n"); | |
1390 | return AVERROR_INVALIDDATA; | |
1391 | } | |
1392 | ||
1393 | memset(reverted_channels, 0, sizeof(*reverted_channels) * avctx->channels); | |
1394 | ||
1395 | bd.ra_block = ra_frame; | |
1396 | bd.prev_raw_samples = ctx->prev_raw_samples; | |
1397 | ||
1398 | get_block_sizes(ctx, div_blocks, &bs_info); | |
1399 | ||
1400 | for (b = 0; b < ctx->num_blocks; b++) { | |
1401 | bd.block_length = div_blocks[b]; | |
1402 | if (bd.block_length <= 0) { | |
1403 | av_log(ctx->avctx, AV_LOG_WARNING, | |
1404 | "Invalid block length %u in channel data!\n", | |
1405 | bd.block_length); | |
1406 | continue; | |
1407 | } | |
1408 | ||
1409 | for (c = 0; c < avctx->channels; c++) { | |
1410 | bd.const_block = ctx->const_block + c; | |
1411 | bd.shift_lsbs = ctx->shift_lsbs + c; | |
1412 | bd.opt_order = ctx->opt_order + c; | |
1413 | bd.store_prev_samples = ctx->store_prev_samples + c; | |
1414 | bd.use_ltp = ctx->use_ltp + c; | |
1415 | bd.ltp_lag = ctx->ltp_lag + c; | |
1416 | bd.ltp_gain = ctx->ltp_gain[c]; | |
1417 | bd.lpc_cof = ctx->lpc_cof[c]; | |
1418 | bd.quant_cof = ctx->quant_cof[c]; | |
1419 | bd.raw_samples = ctx->raw_samples[c] + offset; | |
1420 | bd.raw_other = NULL; | |
1421 | ||
1422 | if ((ret = read_block(ctx, &bd)) < 0) | |
1423 | return ret; | |
1424 | if ((ret = read_channel_data(ctx, ctx->chan_data[c], c)) < 0) | |
1425 | return ret; | |
1426 | } | |
1427 | ||
1428 | for (c = 0; c < avctx->channels; c++) { | |
1429 | ret = revert_channel_correlation(ctx, &bd, ctx->chan_data, | |
1430 | reverted_channels, offset, c); | |
1431 | if (ret < 0) | |
1432 | return ret; | |
1433 | } | |
1434 | for (c = 0; c < avctx->channels; c++) { | |
1435 | bd.const_block = ctx->const_block + c; | |
1436 | bd.shift_lsbs = ctx->shift_lsbs + c; | |
1437 | bd.opt_order = ctx->opt_order + c; | |
1438 | bd.store_prev_samples = ctx->store_prev_samples + c; | |
1439 | bd.use_ltp = ctx->use_ltp + c; | |
1440 | bd.ltp_lag = ctx->ltp_lag + c; | |
1441 | bd.ltp_gain = ctx->ltp_gain[c]; | |
1442 | bd.lpc_cof = ctx->lpc_cof[c]; | |
1443 | bd.quant_cof = ctx->quant_cof[c]; | |
1444 | bd.raw_samples = ctx->raw_samples[c] + offset; | |
1445 | ||
1446 | if ((ret = decode_block(ctx, &bd)) < 0) | |
1447 | return ret; | |
1448 | } | |
1449 | ||
1450 | memset(reverted_channels, 0, avctx->channels * sizeof(*reverted_channels)); | |
1451 | offset += div_blocks[b]; | |
1452 | bd.ra_block = 0; | |
1453 | } | |
1454 | ||
1455 | // store carryover raw samples | |
1456 | for (c = 0; c < avctx->channels; c++) | |
1457 | memmove(ctx->raw_samples[c] - sconf->max_order, | |
1458 | ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
1459 | sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
1460 | } | |
1461 | ||
1462 | // TODO: read_diff_float_data | |
1463 | ||
1464 | return 0; | |
1465 | } | |
1466 | ||
1467 | ||
1468 | /** Decode an ALS frame. | |
1469 | */ | |
1470 | static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, | |
1471 | AVPacket *avpkt) | |
1472 | { | |
1473 | ALSDecContext *ctx = avctx->priv_data; | |
1474 | AVFrame *frame = data; | |
1475 | ALSSpecificConfig *sconf = &ctx->sconf; | |
1476 | const uint8_t *buffer = avpkt->data; | |
1477 | int buffer_size = avpkt->size; | |
1478 | int invalid_frame, ret; | |
1479 | unsigned int c, sample, ra_frame, bytes_read, shift; | |
1480 | ||
1481 | init_get_bits(&ctx->gb, buffer, buffer_size * 8); | |
1482 | ||
1483 | // In the case that the distance between random access frames is set to zero | |
1484 | // (sconf->ra_distance == 0) no frame is treated as a random access frame. | |
1485 | // For the first frame, if prediction is used, all samples used from the | |
1486 | // previous frame are assumed to be zero. | |
1487 | ra_frame = sconf->ra_distance && !(ctx->frame_id % sconf->ra_distance); | |
1488 | ||
1489 | // the last frame to decode might have a different length | |
1490 | if (sconf->samples != 0xFFFFFFFF) | |
1491 | ctx->cur_frame_length = FFMIN(sconf->samples - ctx->frame_id * (uint64_t) sconf->frame_length, | |
1492 | sconf->frame_length); | |
1493 | else | |
1494 | ctx->cur_frame_length = sconf->frame_length; | |
1495 | ||
1496 | // decode the frame data | |
1497 | if ((invalid_frame = read_frame_data(ctx, ra_frame)) < 0) | |
1498 | av_log(ctx->avctx, AV_LOG_WARNING, | |
1499 | "Reading frame data failed. Skipping RA unit.\n"); | |
1500 | ||
1501 | ctx->frame_id++; | |
1502 | ||
1503 | /* get output buffer */ | |
1504 | frame->nb_samples = ctx->cur_frame_length; | |
1505 | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) | |
1506 | return ret; | |
1507 | ||
1508 | // transform decoded frame into output format | |
1509 | #define INTERLEAVE_OUTPUT(bps) \ | |
1510 | { \ | |
1511 | int##bps##_t *dest = (int##bps##_t*)frame->data[0]; \ | |
1512 | shift = bps - ctx->avctx->bits_per_raw_sample; \ | |
1513 | if (!ctx->cs_switch) { \ | |
1514 | for (sample = 0; sample < ctx->cur_frame_length; sample++) \ | |
1515 | for (c = 0; c < avctx->channels; c++) \ | |
1516 | *dest++ = ctx->raw_samples[c][sample] << shift; \ | |
1517 | } else { \ | |
1518 | for (sample = 0; sample < ctx->cur_frame_length; sample++) \ | |
1519 | for (c = 0; c < avctx->channels; c++) \ | |
1520 | *dest++ = ctx->raw_samples[sconf->chan_pos[c]][sample] << shift; \ | |
1521 | } \ | |
1522 | } | |
1523 | ||
1524 | if (ctx->avctx->bits_per_raw_sample <= 16) { | |
1525 | INTERLEAVE_OUTPUT(16) | |
1526 | } else { | |
1527 | INTERLEAVE_OUTPUT(32) | |
1528 | } | |
1529 | ||
1530 | // update CRC | |
1531 | if (sconf->crc_enabled && (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL))) { | |
1532 | int swap = HAVE_BIGENDIAN != sconf->msb_first; | |
1533 | ||
1534 | if (ctx->avctx->bits_per_raw_sample == 24) { | |
1535 | int32_t *src = (int32_t *)frame->data[0]; | |
1536 | ||
1537 | for (sample = 0; | |
1538 | sample < ctx->cur_frame_length * avctx->channels; | |
1539 | sample++) { | |
1540 | int32_t v; | |
1541 | ||
1542 | if (swap) | |
1543 | v = av_bswap32(src[sample]); | |
1544 | else | |
1545 | v = src[sample]; | |
1546 | if (!HAVE_BIGENDIAN) | |
1547 | v >>= 8; | |
1548 | ||
1549 | ctx->crc = av_crc(ctx->crc_table, ctx->crc, (uint8_t*)(&v), 3); | |
1550 | } | |
1551 | } else { | |
1552 | uint8_t *crc_source; | |
1553 | ||
1554 | if (swap) { | |
1555 | if (ctx->avctx->bits_per_raw_sample <= 16) { | |
1556 | int16_t *src = (int16_t*) frame->data[0]; | |
1557 | int16_t *dest = (int16_t*) ctx->crc_buffer; | |
1558 | for (sample = 0; | |
1559 | sample < ctx->cur_frame_length * avctx->channels; | |
1560 | sample++) | |
1561 | *dest++ = av_bswap16(src[sample]); | |
1562 | } else { | |
1563 | ctx->bdsp.bswap_buf((uint32_t *) ctx->crc_buffer, | |
1564 | (uint32_t *) frame->data[0], | |
1565 | ctx->cur_frame_length * avctx->channels); | |
1566 | } | |
1567 | crc_source = ctx->crc_buffer; | |
1568 | } else { | |
1569 | crc_source = frame->data[0]; | |
1570 | } | |
1571 | ||
1572 | ctx->crc = av_crc(ctx->crc_table, ctx->crc, crc_source, | |
1573 | ctx->cur_frame_length * avctx->channels * | |
1574 | av_get_bytes_per_sample(avctx->sample_fmt)); | |
1575 | } | |
1576 | ||
1577 | ||
1578 | // check CRC sums if this is the last frame | |
1579 | if (ctx->cur_frame_length != sconf->frame_length && | |
1580 | ctx->crc_org != ctx->crc) { | |
1581 | av_log(avctx, AV_LOG_ERROR, "CRC error.\n"); | |
1582 | if (avctx->err_recognition & AV_EF_EXPLODE) | |
1583 | return AVERROR_INVALIDDATA; | |
1584 | } | |
1585 | } | |
1586 | ||
1587 | *got_frame_ptr = 1; | |
1588 | ||
1589 | bytes_read = invalid_frame ? buffer_size : | |
1590 | (get_bits_count(&ctx->gb) + 7) >> 3; | |
1591 | ||
1592 | return bytes_read; | |
1593 | } | |
1594 | ||
1595 | ||
1596 | /** Uninitialize the ALS decoder. | |
1597 | */ | |
1598 | static av_cold int decode_end(AVCodecContext *avctx) | |
1599 | { | |
1600 | ALSDecContext *ctx = avctx->priv_data; | |
1601 | ||
1602 | av_freep(&ctx->sconf.chan_pos); | |
1603 | ||
1604 | ff_bgmc_end(&ctx->bgmc_lut, &ctx->bgmc_lut_status); | |
1605 | ||
1606 | av_freep(&ctx->const_block); | |
1607 | av_freep(&ctx->shift_lsbs); | |
1608 | av_freep(&ctx->opt_order); | |
1609 | av_freep(&ctx->store_prev_samples); | |
1610 | av_freep(&ctx->use_ltp); | |
1611 | av_freep(&ctx->ltp_lag); | |
1612 | av_freep(&ctx->ltp_gain); | |
1613 | av_freep(&ctx->ltp_gain_buffer); | |
1614 | av_freep(&ctx->quant_cof); | |
1615 | av_freep(&ctx->lpc_cof); | |
1616 | av_freep(&ctx->quant_cof_buffer); | |
1617 | av_freep(&ctx->lpc_cof_buffer); | |
1618 | av_freep(&ctx->lpc_cof_reversed_buffer); | |
1619 | av_freep(&ctx->prev_raw_samples); | |
1620 | av_freep(&ctx->raw_samples); | |
1621 | av_freep(&ctx->raw_buffer); | |
1622 | av_freep(&ctx->chan_data); | |
1623 | av_freep(&ctx->chan_data_buffer); | |
1624 | av_freep(&ctx->reverted_channels); | |
1625 | av_freep(&ctx->crc_buffer); | |
1626 | ||
1627 | return 0; | |
1628 | } | |
1629 | ||
1630 | ||
1631 | /** Initialize the ALS decoder. | |
1632 | */ | |
1633 | static av_cold int decode_init(AVCodecContext *avctx) | |
1634 | { | |
1635 | unsigned int c; | |
1636 | unsigned int channel_size; | |
1637 | int num_buffers, ret; | |
1638 | ALSDecContext *ctx = avctx->priv_data; | |
1639 | ALSSpecificConfig *sconf = &ctx->sconf; | |
1640 | ctx->avctx = avctx; | |
1641 | ||
1642 | if (!avctx->extradata) { | |
1643 | av_log(avctx, AV_LOG_ERROR, "Missing required ALS extradata.\n"); | |
1644 | return AVERROR_INVALIDDATA; | |
1645 | } | |
1646 | ||
1647 | if ((ret = read_specific_config(ctx)) < 0) { | |
1648 | av_log(avctx, AV_LOG_ERROR, "Reading ALSSpecificConfig failed.\n"); | |
1649 | goto fail; | |
1650 | } | |
1651 | ||
1652 | if ((ret = check_specific_config(ctx)) < 0) { | |
1653 | goto fail; | |
1654 | } | |
1655 | ||
1656 | if (sconf->bgmc) { | |
1657 | ret = ff_bgmc_init(avctx, &ctx->bgmc_lut, &ctx->bgmc_lut_status); | |
1658 | if (ret < 0) | |
1659 | goto fail; | |
1660 | } | |
1661 | if (sconf->floating) { | |
1662 | avctx->sample_fmt = AV_SAMPLE_FMT_FLT; | |
1663 | avctx->bits_per_raw_sample = 32; | |
1664 | } else { | |
1665 | avctx->sample_fmt = sconf->resolution > 1 | |
1666 | ? AV_SAMPLE_FMT_S32 : AV_SAMPLE_FMT_S16; | |
1667 | avctx->bits_per_raw_sample = (sconf->resolution + 1) * 8; | |
1668 | } | |
1669 | ||
1670 | // set maximum Rice parameter for progressive decoding based on resolution | |
1671 | // This is not specified in 14496-3 but actually done by the reference | |
1672 | // codec RM22 revision 2. | |
1673 | ctx->s_max = sconf->resolution > 1 ? 31 : 15; | |
1674 | ||
1675 | // set lag value for long-term prediction | |
1676 | ctx->ltp_lag_length = 8 + (avctx->sample_rate >= 96000) + | |
1677 | (avctx->sample_rate >= 192000); | |
1678 | ||
1679 | // allocate quantized parcor coefficient buffer | |
1680 | num_buffers = sconf->mc_coding ? avctx->channels : 1; | |
1681 | ||
1682 | ctx->quant_cof = av_malloc(sizeof(*ctx->quant_cof) * num_buffers); | |
1683 | ctx->lpc_cof = av_malloc(sizeof(*ctx->lpc_cof) * num_buffers); | |
1684 | ctx->quant_cof_buffer = av_malloc(sizeof(*ctx->quant_cof_buffer) * | |
1685 | num_buffers * sconf->max_order); | |
1686 | ctx->lpc_cof_buffer = av_malloc(sizeof(*ctx->lpc_cof_buffer) * | |
1687 | num_buffers * sconf->max_order); | |
1688 | ctx->lpc_cof_reversed_buffer = av_malloc(sizeof(*ctx->lpc_cof_buffer) * | |
1689 | sconf->max_order); | |
1690 | ||
1691 | if (!ctx->quant_cof || !ctx->lpc_cof || | |
1692 | !ctx->quant_cof_buffer || !ctx->lpc_cof_buffer || | |
1693 | !ctx->lpc_cof_reversed_buffer) { | |
1694 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
1695 | ret = AVERROR(ENOMEM); | |
1696 | goto fail; | |
1697 | } | |
1698 | ||
1699 | // assign quantized parcor coefficient buffers | |
1700 | for (c = 0; c < num_buffers; c++) { | |
1701 | ctx->quant_cof[c] = ctx->quant_cof_buffer + c * sconf->max_order; | |
1702 | ctx->lpc_cof[c] = ctx->lpc_cof_buffer + c * sconf->max_order; | |
1703 | } | |
1704 | ||
1705 | // allocate and assign lag and gain data buffer for ltp mode | |
1706 | ctx->const_block = av_malloc (sizeof(*ctx->const_block) * num_buffers); | |
1707 | ctx->shift_lsbs = av_malloc (sizeof(*ctx->shift_lsbs) * num_buffers); | |
1708 | ctx->opt_order = av_malloc (sizeof(*ctx->opt_order) * num_buffers); | |
1709 | ctx->store_prev_samples = av_malloc(sizeof(*ctx->store_prev_samples) * num_buffers); | |
1710 | ctx->use_ltp = av_mallocz(sizeof(*ctx->use_ltp) * num_buffers); | |
1711 | ctx->ltp_lag = av_malloc (sizeof(*ctx->ltp_lag) * num_buffers); | |
1712 | ctx->ltp_gain = av_malloc (sizeof(*ctx->ltp_gain) * num_buffers); | |
1713 | ctx->ltp_gain_buffer = av_malloc (sizeof(*ctx->ltp_gain_buffer) * | |
1714 | num_buffers * 5); | |
1715 | ||
1716 | if (!ctx->const_block || !ctx->shift_lsbs || | |
1717 | !ctx->opt_order || !ctx->store_prev_samples || | |
1718 | !ctx->use_ltp || !ctx->ltp_lag || | |
1719 | !ctx->ltp_gain || !ctx->ltp_gain_buffer) { | |
1720 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
1721 | ret = AVERROR(ENOMEM); | |
1722 | goto fail; | |
1723 | } | |
1724 | ||
1725 | for (c = 0; c < num_buffers; c++) | |
1726 | ctx->ltp_gain[c] = ctx->ltp_gain_buffer + c * 5; | |
1727 | ||
1728 | // allocate and assign channel data buffer for mcc mode | |
1729 | if (sconf->mc_coding) { | |
1730 | ctx->chan_data_buffer = av_malloc(sizeof(*ctx->chan_data_buffer) * | |
1731 | num_buffers * num_buffers); | |
1732 | ctx->chan_data = av_malloc(sizeof(*ctx->chan_data) * | |
1733 | num_buffers); | |
1734 | ctx->reverted_channels = av_malloc(sizeof(*ctx->reverted_channels) * | |
1735 | num_buffers); | |
1736 | ||
1737 | if (!ctx->chan_data_buffer || !ctx->chan_data || !ctx->reverted_channels) { | |
1738 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
1739 | ret = AVERROR(ENOMEM); | |
1740 | goto fail; | |
1741 | } | |
1742 | ||
1743 | for (c = 0; c < num_buffers; c++) | |
1744 | ctx->chan_data[c] = ctx->chan_data_buffer + c * num_buffers; | |
1745 | } else { | |
1746 | ctx->chan_data = NULL; | |
1747 | ctx->chan_data_buffer = NULL; | |
1748 | ctx->reverted_channels = NULL; | |
1749 | } | |
1750 | ||
1751 | channel_size = sconf->frame_length + sconf->max_order; | |
1752 | ||
1753 | ctx->prev_raw_samples = av_malloc (sizeof(*ctx->prev_raw_samples) * sconf->max_order); | |
1754 | ctx->raw_buffer = av_mallocz(sizeof(*ctx-> raw_buffer) * avctx->channels * channel_size); | |
1755 | ctx->raw_samples = av_malloc (sizeof(*ctx-> raw_samples) * avctx->channels); | |
1756 | ||
1757 | // allocate previous raw sample buffer | |
1758 | if (!ctx->prev_raw_samples || !ctx->raw_buffer|| !ctx->raw_samples) { | |
1759 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
1760 | ret = AVERROR(ENOMEM); | |
1761 | goto fail; | |
1762 | } | |
1763 | ||
1764 | // assign raw samples buffers | |
1765 | ctx->raw_samples[0] = ctx->raw_buffer + sconf->max_order; | |
1766 | for (c = 1; c < avctx->channels; c++) | |
1767 | ctx->raw_samples[c] = ctx->raw_samples[c - 1] + channel_size; | |
1768 | ||
1769 | // allocate crc buffer | |
1770 | if (HAVE_BIGENDIAN != sconf->msb_first && sconf->crc_enabled && | |
1771 | (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL))) { | |
1772 | ctx->crc_buffer = av_malloc(sizeof(*ctx->crc_buffer) * | |
1773 | ctx->cur_frame_length * | |
1774 | avctx->channels * | |
1775 | av_get_bytes_per_sample(avctx->sample_fmt)); | |
1776 | if (!ctx->crc_buffer) { | |
1777 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
1778 | ret = AVERROR(ENOMEM); | |
1779 | goto fail; | |
1780 | } | |
1781 | } | |
1782 | ||
1783 | ff_bswapdsp_init(&ctx->bdsp); | |
1784 | ||
1785 | return 0; | |
1786 | ||
1787 | fail: | |
1788 | decode_end(avctx); | |
1789 | return ret; | |
1790 | } | |
1791 | ||
1792 | ||
1793 | /** Flush (reset) the frame ID after seeking. | |
1794 | */ | |
1795 | static av_cold void flush(AVCodecContext *avctx) | |
1796 | { | |
1797 | ALSDecContext *ctx = avctx->priv_data; | |
1798 | ||
1799 | ctx->frame_id = 0; | |
1800 | } | |
1801 | ||
1802 | ||
1803 | AVCodec ff_als_decoder = { | |
1804 | .name = "als", | |
1805 | .long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Audio Lossless Coding (ALS)"), | |
1806 | .type = AVMEDIA_TYPE_AUDIO, | |
1807 | .id = AV_CODEC_ID_MP4ALS, | |
1808 | .priv_data_size = sizeof(ALSDecContext), | |
1809 | .init = decode_init, | |
1810 | .close = decode_end, | |
1811 | .decode = decode_frame, | |
1812 | .flush = flush, | |
1813 | .capabilities = CODEC_CAP_SUBFRAMES | CODEC_CAP_DR1, | |
1814 | }; |