Details | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
4349 | Serge | 1 | /* |
2 | * MPEG-4 ALS decoder |
||
3 | * Copyright (c) 2009 Thilo Borgmann |
||
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 |
||
26 | */ |
||
27 | |||
28 | #include "avcodec.h" |
||
29 | #include "get_bits.h" |
||
30 | #include "unary.h" |
||
31 | #include "mpeg4audio.h" |
||
32 | #include "bytestream.h" |
||
33 | #include "bgmc.h" |
||
34 | #include "dsputil.h" |
||
35 | #include "internal.h" |
||
36 | #include "libavutil/samplefmt.h" |
||
37 | #include "libavutil/crc.h" |
||
38 | |||
39 | #include |
||
40 | |||
41 | /** Rice parameters and corresponding index offsets for decoding the |
||
42 | * indices of scaled PARCOR values. The table chosen is set globally |
||
43 | * by the encoder and stored in ALSSpecificConfig. |
||
44 | */ |
||
45 | static const int8_t parcor_rice_table[3][20][2] = { |
||
46 | { {-52, 4}, {-29, 5}, {-31, 4}, { 19, 4}, {-16, 4}, |
||
47 | { 12, 3}, { -7, 3}, { 9, 3}, { -5, 3}, { 6, 3}, |
||
48 | { -4, 3}, { 3, 3}, { -3, 2}, { 3, 2}, { -2, 2}, |
||
49 | { 3, 2}, { -1, 2}, { 2, 2}, { -1, 2}, { 2, 2} }, |
||
50 | { {-58, 3}, {-42, 4}, {-46, 4}, { 37, 5}, {-36, 4}, |
||
51 | { 29, 4}, {-29, 4}, { 25, 4}, {-23, 4}, { 20, 4}, |
||
52 | {-17, 4}, { 16, 4}, {-12, 4}, { 12, 3}, {-10, 4}, |
||
53 | { 7, 3}, { -4, 4}, { 3, 3}, { -1, 3}, { 1, 3} }, |
||
54 | { {-59, 3}, {-45, 5}, {-50, 4}, { 38, 4}, {-39, 4}, |
||
55 | { 32, 4}, {-30, 4}, { 25, 3}, {-23, 3}, { 20, 3}, |
||
56 | {-20, 3}, { 16, 3}, {-13, 3}, { 10, 3}, { -7, 3}, |
||
57 | { 3, 3}, { 0, 3}, { -1, 3}, { 2, 3}, { -1, 2} } |
||
58 | }; |
||
59 | |||
60 | |||
61 | /** Scaled PARCOR values used for the first two PARCOR coefficients. |
||
62 | * To be indexed by the Rice coded indices. |
||
63 | * Generated by: parcor_scaled_values[i] = 32 + ((i * (i+1)) << 7) - (1 << 20) |
||
64 | * Actual values are divided by 32 in order to be stored in 16 bits. |
||
65 | */ |
||
66 | static const int16_t parcor_scaled_values[] = { |
||
67 | -1048544 / 32, -1048288 / 32, -1047776 / 32, -1047008 / 32, |
||
68 | -1045984 / 32, -1044704 / 32, -1043168 / 32, -1041376 / 32, |
||
69 | -1039328 / 32, -1037024 / 32, -1034464 / 32, -1031648 / 32, |
||
70 | -1028576 / 32, -1025248 / 32, -1021664 / 32, -1017824 / 32, |
||
71 | -1013728 / 32, -1009376 / 32, -1004768 / 32, -999904 / 32, |
||
72 | -994784 / 32, -989408 / 32, -983776 / 32, -977888 / 32, |
||
73 | -971744 / 32, -965344 / 32, -958688 / 32, -951776 / 32, |
||
74 | -944608 / 32, -937184 / 32, -929504 / 32, -921568 / 32, |
||
75 | -913376 / 32, -904928 / 32, -896224 / 32, -887264 / 32, |
||
76 | -878048 / 32, -868576 / 32, -858848 / 32, -848864 / 32, |
||
77 | -838624 / 32, -828128 / 32, -817376 / 32, -806368 / 32, |
||
78 | -795104 / 32, -783584 / 32, -771808 / 32, -759776 / 32, |
||
79 | -747488 / 32, -734944 / 32, -722144 / 32, -709088 / 32, |
||
80 | -695776 / 32, -682208 / 32, -668384 / 32, -654304 / 32, |
||
81 | -639968 / 32, -625376 / 32, -610528 / 32, -595424 / 32, |
||
82 | -580064 / 32, -564448 / 32, -548576 / 32, -532448 / 32, |
||
83 | -516064 / 32, -499424 / 32, -482528 / 32, -465376 / 32, |
||
84 | -447968 / 32, -430304 / 32, -412384 / 32, -394208 / 32, |
||
85 | -375776 / 32, -357088 / 32, -338144 / 32, -318944 / 32, |
||
86 | -299488 / 32, -279776 / 32, -259808 / 32, -239584 / 32, |
||
87 | -219104 / 32, -198368 / 32, -177376 / 32, -156128 / 32, |
||
88 | -134624 / 32, -112864 / 32, -90848 / 32, -68576 / 32, |
||
89 | -46048 / 32, -23264 / 32, -224 / 32, 23072 / 32, |
||
90 | 46624 / 32, 70432 / 32, 94496 / 32, 118816 / 32, |
||
91 | 143392 / 32, 168224 / 32, 193312 / 32, 218656 / 32, |
||
92 | 244256 / 32, 270112 / 32, 296224 / 32, 322592 / 32, |
||
93 | 349216 / 32, 376096 / 32, 403232 / 32, 430624 / 32, |
||
94 | 458272 / 32, 486176 / 32, 514336 / 32, 542752 / 32, |
||
95 | 571424 / 32, 600352 / 32, 629536 / 32, 658976 / 32, |
||
96 | 688672 / 32, 718624 / 32, 748832 / 32, 779296 / 32, |
||
97 | 810016 / 32, 840992 / 32, 872224 / 32, 903712 / 32, |
||
98 | 935456 / 32, 967456 / 32, 999712 / 32, 1032224 / 32 |
||
99 | }; |
||
100 | |||
101 | |||
102 | /** Gain values of p(0) for long-term prediction. |
||
103 | * To be indexed by the Rice coded indices. |
||
104 | */ |
||
105 | static const uint8_t ltp_gain_values [4][4] = { |
||
106 | { 0, 8, 16, 24}, |
||
107 | {32, 40, 48, 56}, |
||
108 | {64, 70, 76, 82}, |
||
109 | {88, 92, 96, 100} |
||
110 | }; |
||
111 | |||
112 | |||
113 | /** Inter-channel weighting factors for multi-channel correlation. |
||
114 | * To be indexed by the Rice coded indices. |
||
115 | */ |
||
116 | static const int16_t mcc_weightings[] = { |
||
117 | 204, 192, 179, 166, 153, 140, 128, 115, |
||
118 | 102, 89, 76, 64, 51, 38, 25, 12, |
||
119 | 0, -12, -25, -38, -51, -64, -76, -89, |
||
120 | -102, -115, -128, -140, -153, -166, -179, -192 |
||
121 | }; |
||
122 | |||
123 | |||
124 | /** Tail codes used in arithmetic coding using block Gilbert-Moore codes. |
||
125 | */ |
||
126 | static const uint8_t tail_code[16][6] = { |
||
127 | { 74, 44, 25, 13, 7, 3}, |
||
128 | { 68, 42, 24, 13, 7, 3}, |
||
129 | { 58, 39, 23, 13, 7, 3}, |
||
130 | {126, 70, 37, 19, 10, 5}, |
||
131 | {132, 70, 37, 20, 10, 5}, |
||
132 | {124, 70, 38, 20, 10, 5}, |
||
133 | {120, 69, 37, 20, 11, 5}, |
||
134 | {116, 67, 37, 20, 11, 5}, |
||
135 | {108, 66, 36, 20, 10, 5}, |
||
136 | {102, 62, 36, 20, 10, 5}, |
||
137 | { 88, 58, 34, 19, 10, 5}, |
||
138 | {162, 89, 49, 25, 13, 7}, |
||
139 | {156, 87, 49, 26, 14, 7}, |
||
140 | {150, 86, 47, 26, 14, 7}, |
||
141 | {142, 84, 47, 26, 14, 7}, |
||
142 | {131, 79, 46, 26, 14, 7} |
||
143 | }; |
||
144 | |||
145 | |||
146 | enum RA_Flag { |
||
147 | RA_FLAG_NONE, |
||
148 | RA_FLAG_FRAMES, |
||
149 | RA_FLAG_HEADER |
||
150 | }; |
||
151 | |||
152 | |||
153 | typedef struct { |
||
154 | uint32_t samples; ///< number of samples, 0xFFFFFFFF if unknown |
||
155 | int resolution; ///< 000 = 8-bit; 001 = 16-bit; 010 = 24-bit; 011 = 32-bit |
||
156 | int floating; ///< 1 = IEEE 32-bit floating-point, 0 = integer |
||
157 | int msb_first; ///< 1 = original CRC calculated on big-endian system, 0 = little-endian |
||
158 | int frame_length; ///< frame length for each frame (last frame may differ) |
||
159 | int ra_distance; ///< distance between RA frames (in frames, 0...255) |
||
160 | enum RA_Flag ra_flag; ///< indicates where the size of ra units is stored |
||
161 | int adapt_order; ///< adaptive order: 1 = on, 0 = off |
||
162 | int coef_table; ///< table index of Rice code parameters |
||
163 | int long_term_prediction; ///< long term prediction (LTP): 1 = on, 0 = off |
||
164 | int max_order; ///< maximum prediction order (0..1023) |
||
165 | int block_switching; ///< number of block switching levels |
||
166 | int bgmc; ///< "Block Gilbert-Moore Code": 1 = on, 0 = off (Rice coding only) |
||
167 | int sb_part; ///< sub-block partition |
||
168 | int joint_stereo; ///< joint stereo: 1 = on, 0 = off |
||
169 | int mc_coding; ///< extended inter-channel coding (multi channel coding): 1 = on, 0 = off |
||
170 | int chan_config; ///< indicates that a chan_config_info field is present |
||
171 | int chan_sort; ///< channel rearrangement: 1 = on, 0 = off |
||
172 | int rlslms; ///< use "Recursive Least Square-Least Mean Square" predictor: 1 = on, 0 = off |
||
173 | int chan_config_info; ///< mapping of channels to loudspeaker locations. Unused until setting channel configuration is implemented. |
||
174 | int *chan_pos; ///< original channel positions |
||
175 | int crc_enabled; ///< enable Cyclic Redundancy Checksum |
||
176 | } ALSSpecificConfig; |
||
177 | |||
178 | |||
179 | typedef struct { |
||
180 | int stop_flag; |
||
181 | int master_channel; |
||
182 | int time_diff_flag; |
||
183 | int time_diff_sign; |
||
184 | int time_diff_index; |
||
185 | int weighting[6]; |
||
186 | } ALSChannelData; |
||
187 | |||
188 | |||
189 | typedef struct { |
||
190 | AVCodecContext *avctx; |
||
191 | ALSSpecificConfig sconf; |
||
192 | GetBitContext gb; |
||
193 | DSPContext dsp; |
||
194 | const AVCRC *crc_table; |
||
195 | uint32_t crc_org; ///< CRC value of the original input data |
||
196 | uint32_t crc; ///< CRC value calculated from decoded data |
||
197 | unsigned int cur_frame_length; ///< length of the current frame to decode |
||
198 | unsigned int frame_id; ///< the frame ID / number of the current frame |
||
199 | unsigned int js_switch; ///< if true, joint-stereo decoding is enforced |
||
200 | unsigned int cs_switch; ///< if true, channel rearrangement is done |
||
201 | unsigned int num_blocks; ///< number of blocks used in the current frame |
||
202 | unsigned int s_max; ///< maximum Rice parameter allowed in entropy coding |
||
203 | uint8_t *bgmc_lut; ///< pointer at lookup tables used for BGMC |
||
204 | int *bgmc_lut_status; ///< pointer at lookup table status flags used for BGMC |
||
205 | int ltp_lag_length; ///< number of bits used for ltp lag value |
||
206 | int *const_block; ///< contains const_block flags for all channels |
||
207 | unsigned int *shift_lsbs; ///< contains shift_lsbs flags for all channels |
||
208 | unsigned int *opt_order; ///< contains opt_order flags for all channels |
||
209 | int *store_prev_samples; ///< contains store_prev_samples flags for all channels |
||
210 | int *use_ltp; ///< contains use_ltp flags for all channels |
||
211 | int *ltp_lag; ///< contains ltp lag values for all channels |
||
212 | int **ltp_gain; ///< gain values for ltp 5-tap filter for a channel |
||
213 | int *ltp_gain_buffer; ///< contains all gain values for ltp 5-tap filter |
||
214 | int32_t **quant_cof; ///< quantized parcor coefficients for a channel |
||
215 | int32_t *quant_cof_buffer; ///< contains all quantized parcor coefficients |
||
216 | int32_t **lpc_cof; ///< coefficients of the direct form prediction filter for a channel |
||
217 | int32_t *lpc_cof_buffer; ///< contains all coefficients of the direct form prediction filter |
||
218 | int32_t *lpc_cof_reversed_buffer; ///< temporary buffer to set up a reversed versio of lpc_cof_buffer |
||
219 | ALSChannelData **chan_data; ///< channel data for multi-channel correlation |
||
220 | ALSChannelData *chan_data_buffer; ///< contains channel data for all channels |
||
221 | int *reverted_channels; ///< stores a flag for each reverted channel |
||
222 | int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block |
||
223 | int32_t **raw_samples; ///< decoded raw samples for each channel |
||
224 | int32_t *raw_buffer; ///< contains all decoded raw samples including carryover samples |
||
225 | uint8_t *crc_buffer; ///< buffer of byte order corrected samples used for CRC check |
||
226 | } ALSDecContext; |
||
227 | |||
228 | |||
229 | typedef struct { |
||
230 | unsigned int block_length; ///< number of samples within the block |
||
231 | unsigned int ra_block; ///< if true, this is a random access block |
||
232 | int *const_block; ///< if true, this is a constant value block |
||
233 | int js_blocks; ///< true if this block contains a difference signal |
||
234 | unsigned int *shift_lsbs; ///< shift of values for this block |
||
235 | unsigned int *opt_order; ///< prediction order of this block |
||
236 | int *store_prev_samples;///< if true, carryover samples have to be stored |
||
237 | int *use_ltp; ///< if true, long-term prediction is used |
||
238 | int *ltp_lag; ///< lag value for long-term prediction |
||
239 | int *ltp_gain; ///< gain values for ltp 5-tap filter |
||
240 | int32_t *quant_cof; ///< quantized parcor coefficients |
||
241 | int32_t *lpc_cof; ///< coefficients of the direct form prediction |
||
242 | int32_t *raw_samples; ///< decoded raw samples / residuals for this block |
||
243 | int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block |
||
244 | int32_t *raw_other; ///< decoded raw samples of the other channel of a channel pair |
||
245 | } ALSBlockData; |
||
246 | |||
247 | |||
248 | static av_cold void dprint_specific_config(ALSDecContext *ctx) |
||
249 | { |
||
250 | #ifdef DEBUG |
||
251 | AVCodecContext *avctx = ctx->avctx; |
||
252 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
253 | |||
254 | av_dlog(avctx, "resolution = %i\n", sconf->resolution); |
||
255 | av_dlog(avctx, "floating = %i\n", sconf->floating); |
||
256 | av_dlog(avctx, "frame_length = %i\n", sconf->frame_length); |
||
257 | av_dlog(avctx, "ra_distance = %i\n", sconf->ra_distance); |
||
258 | av_dlog(avctx, "ra_flag = %i\n", sconf->ra_flag); |
||
259 | av_dlog(avctx, "adapt_order = %i\n", sconf->adapt_order); |
||
260 | av_dlog(avctx, "coef_table = %i\n", sconf->coef_table); |
||
261 | av_dlog(avctx, "long_term_prediction = %i\n", sconf->long_term_prediction); |
||
262 | av_dlog(avctx, "max_order = %i\n", sconf->max_order); |
||
263 | av_dlog(avctx, "block_switching = %i\n", sconf->block_switching); |
||
264 | av_dlog(avctx, "bgmc = %i\n", sconf->bgmc); |
||
265 | av_dlog(avctx, "sb_part = %i\n", sconf->sb_part); |
||
266 | av_dlog(avctx, "joint_stereo = %i\n", sconf->joint_stereo); |
||
267 | av_dlog(avctx, "mc_coding = %i\n", sconf->mc_coding); |
||
268 | av_dlog(avctx, "chan_config = %i\n", sconf->chan_config); |
||
269 | av_dlog(avctx, "chan_sort = %i\n", sconf->chan_sort); |
||
270 | av_dlog(avctx, "RLSLMS = %i\n", sconf->rlslms); |
||
271 | av_dlog(avctx, "chan_config_info = %i\n", sconf->chan_config_info); |
||
272 | #endif |
||
273 | } |
||
274 | |||
275 | |||
276 | /** Read an ALSSpecificConfig from a buffer into the output struct. |
||
277 | */ |
||
278 | static av_cold int read_specific_config(ALSDecContext *ctx) |
||
279 | { |
||
280 | GetBitContext gb; |
||
281 | uint64_t ht_size; |
||
282 | int i, config_offset; |
||
283 | MPEG4AudioConfig m4ac; |
||
284 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
285 | AVCodecContext *avctx = ctx->avctx; |
||
286 | uint32_t als_id, header_size, trailer_size; |
||
287 | int ret; |
||
288 | |||
289 | if ((ret = init_get_bits8(&gb, avctx->extradata, avctx->extradata_size)) < 0) |
||
290 | return ret; |
||
291 | |||
292 | config_offset = avpriv_mpeg4audio_get_config(&m4ac, avctx->extradata, |
||
293 | avctx->extradata_size * 8, 1); |
||
294 | |||
295 | if (config_offset < 0) |
||
296 | return AVERROR_INVALIDDATA; |
||
297 | |||
298 | skip_bits_long(&gb, config_offset); |
||
299 | |||
300 | if (get_bits_left(&gb) < (30 << 3)) |
||
301 | return AVERROR_INVALIDDATA; |
||
302 | |||
303 | // read the fixed items |
||
304 | als_id = get_bits_long(&gb, 32); |
||
305 | avctx->sample_rate = m4ac.sample_rate; |
||
306 | skip_bits_long(&gb, 32); // sample rate already known |
||
307 | sconf->samples = get_bits_long(&gb, 32); |
||
308 | avctx->channels = m4ac.channels; |
||
309 | skip_bits(&gb, 16); // number of channels already known |
||
310 | skip_bits(&gb, 3); // skip file_type |
||
311 | sconf->resolution = get_bits(&gb, 3); |
||
312 | sconf->floating = get_bits1(&gb); |
||
313 | sconf->msb_first = get_bits1(&gb); |
||
314 | sconf->frame_length = get_bits(&gb, 16) + 1; |
||
315 | sconf->ra_distance = get_bits(&gb, 8); |
||
316 | sconf->ra_flag = get_bits(&gb, 2); |
||
317 | sconf->adapt_order = get_bits1(&gb); |
||
318 | sconf->coef_table = get_bits(&gb, 2); |
||
319 | sconf->long_term_prediction = get_bits1(&gb); |
||
320 | sconf->max_order = get_bits(&gb, 10); |
||
321 | sconf->block_switching = get_bits(&gb, 2); |
||
322 | sconf->bgmc = get_bits1(&gb); |
||
323 | sconf->sb_part = get_bits1(&gb); |
||
324 | sconf->joint_stereo = get_bits1(&gb); |
||
325 | sconf->mc_coding = get_bits1(&gb); |
||
326 | sconf->chan_config = get_bits1(&gb); |
||
327 | sconf->chan_sort = get_bits1(&gb); |
||
328 | sconf->crc_enabled = get_bits1(&gb); |
||
329 | sconf->rlslms = get_bits1(&gb); |
||
330 | skip_bits(&gb, 5); // skip 5 reserved bits |
||
331 | skip_bits1(&gb); // skip aux_data_enabled |
||
332 | |||
333 | |||
334 | // check for ALSSpecificConfig struct |
||
335 | if (als_id != MKBETAG('A','L','S','\0')) |
||
336 | return AVERROR_INVALIDDATA; |
||
337 | |||
338 | ctx->cur_frame_length = sconf->frame_length; |
||
339 | |||
340 | // read channel config |
||
341 | if (sconf->chan_config) |
||
342 | sconf->chan_config_info = get_bits(&gb, 16); |
||
343 | // TODO: use this to set avctx->channel_layout |
||
344 | |||
345 | |||
346 | // read channel sorting |
||
347 | if (sconf->chan_sort && avctx->channels > 1) { |
||
348 | int chan_pos_bits = av_ceil_log2(avctx->channels); |
||
349 | int bits_needed = avctx->channels * chan_pos_bits + 7; |
||
350 | if (get_bits_left(&gb) < bits_needed) |
||
351 | return AVERROR_INVALIDDATA; |
||
352 | |||
353 | if (!(sconf->chan_pos = av_malloc(avctx->channels * sizeof(*sconf->chan_pos)))) |
||
354 | return AVERROR(ENOMEM); |
||
355 | |||
356 | ctx->cs_switch = 1; |
||
357 | |||
358 | for (i = 0; i < avctx->channels; i++) { |
||
359 | int idx; |
||
360 | |||
361 | idx = get_bits(&gb, chan_pos_bits); |
||
362 | if (idx >= avctx->channels) { |
||
363 | av_log(avctx, AV_LOG_WARNING, "Invalid channel reordering.\n"); |
||
364 | ctx->cs_switch = 0; |
||
365 | break; |
||
366 | } |
||
367 | sconf->chan_pos[idx] = i; |
||
368 | } |
||
369 | |||
370 | align_get_bits(&gb); |
||
371 | } |
||
372 | |||
373 | |||
374 | // read fixed header and trailer sizes, |
||
375 | // if size = 0xFFFFFFFF then there is no data field! |
||
376 | if (get_bits_left(&gb) < 64) |
||
377 | return AVERROR_INVALIDDATA; |
||
378 | |||
379 | header_size = get_bits_long(&gb, 32); |
||
380 | trailer_size = get_bits_long(&gb, 32); |
||
381 | if (header_size == 0xFFFFFFFF) |
||
382 | header_size = 0; |
||
383 | if (trailer_size == 0xFFFFFFFF) |
||
384 | trailer_size = 0; |
||
385 | |||
386 | ht_size = ((int64_t)(header_size) + (int64_t)(trailer_size)) << 3; |
||
387 | |||
388 | |||
389 | // skip the header and trailer data |
||
390 | if (get_bits_left(&gb) < ht_size) |
||
391 | return AVERROR_INVALIDDATA; |
||
392 | |||
393 | if (ht_size > INT32_MAX) |
||
394 | return AVERROR_PATCHWELCOME; |
||
395 | |||
396 | skip_bits_long(&gb, ht_size); |
||
397 | |||
398 | |||
399 | // initialize CRC calculation |
||
400 | if (sconf->crc_enabled) { |
||
401 | if (get_bits_left(&gb) < 32) |
||
402 | return AVERROR_INVALIDDATA; |
||
403 | |||
404 | if (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL)) { |
||
405 | ctx->crc_table = av_crc_get_table(AV_CRC_32_IEEE_LE); |
||
406 | ctx->crc = 0xFFFFFFFF; |
||
407 | ctx->crc_org = ~get_bits_long(&gb, 32); |
||
408 | } else |
||
409 | skip_bits_long(&gb, 32); |
||
410 | } |
||
411 | |||
412 | |||
413 | // no need to read the rest of ALSSpecificConfig (ra_unit_size & aux data) |
||
414 | |||
415 | dprint_specific_config(ctx); |
||
416 | |||
417 | return 0; |
||
418 | } |
||
419 | |||
420 | |||
421 | /** Check the ALSSpecificConfig for unsupported features. |
||
422 | */ |
||
423 | static int check_specific_config(ALSDecContext *ctx) |
||
424 | { |
||
425 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
426 | int error = 0; |
||
427 | |||
428 | // report unsupported feature and set error value |
||
429 | #define MISSING_ERR(cond, str, errval) \ |
||
430 | { \ |
||
431 | if (cond) { \ |
||
432 | avpriv_report_missing_feature(ctx->avctx, \ |
||
433 | str); \ |
||
434 | error = errval; \ |
||
435 | } \ |
||
436 | } |
||
437 | |||
438 | MISSING_ERR(sconf->floating, "Floating point decoding", AVERROR_PATCHWELCOME); |
||
439 | MISSING_ERR(sconf->rlslms, "Adaptive RLS-LMS prediction", AVERROR_PATCHWELCOME); |
||
440 | |||
441 | return error; |
||
442 | } |
||
443 | |||
444 | |||
445 | /** Parse the bs_info field to extract the block partitioning used in |
||
446 | * block switching mode, refer to ISO/IEC 14496-3, section 11.6.2. |
||
447 | */ |
||
448 | static void parse_bs_info(const uint32_t bs_info, unsigned int n, |
||
449 | unsigned int div, unsigned int **div_blocks, |
||
450 | unsigned int *num_blocks) |
||
451 | { |
||
452 | if (n < 31 && ((bs_info << n) & 0x40000000)) { |
||
453 | // if the level is valid and the investigated bit n is set |
||
454 | // then recursively check both children at bits (2n+1) and (2n+2) |
||
455 | n *= 2; |
||
456 | div += 1; |
||
457 | parse_bs_info(bs_info, n + 1, div, div_blocks, num_blocks); |
||
458 | parse_bs_info(bs_info, n + 2, div, div_blocks, num_blocks); |
||
459 | } else { |
||
460 | // else the bit is not set or the last level has been reached |
||
461 | // (bit implicitly not set) |
||
462 | **div_blocks = div; |
||
463 | (*div_blocks)++; |
||
464 | (*num_blocks)++; |
||
465 | } |
||
466 | } |
||
467 | |||
468 | |||
469 | /** Read and decode a Rice codeword. |
||
470 | */ |
||
471 | static int32_t decode_rice(GetBitContext *gb, unsigned int k) |
||
472 | { |
||
473 | int max = get_bits_left(gb) - k; |
||
474 | int q = get_unary(gb, 0, max); |
||
475 | int r = k ? get_bits1(gb) : !(q & 1); |
||
476 | |||
477 | if (k > 1) { |
||
478 | q <<= (k - 1); |
||
479 | q += get_bits_long(gb, k - 1); |
||
480 | } else if (!k) { |
||
481 | q >>= 1; |
||
482 | } |
||
483 | return r ? q : ~q; |
||
484 | } |
||
485 | |||
486 | |||
487 | /** Convert PARCOR coefficient k to direct filter coefficient. |
||
488 | */ |
||
489 | static void parcor_to_lpc(unsigned int k, const int32_t *par, int32_t *cof) |
||
490 | { |
||
491 | int i, j; |
||
492 | |||
493 | for (i = 0, j = k - 1; i < j; i++, j--) { |
||
494 | int tmp1 = ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20); |
||
495 | cof[j] += ((MUL64(par[k], cof[i]) + (1 << 19)) >> 20); |
||
496 | cof[i] += tmp1; |
||
497 | } |
||
498 | if (i == j) |
||
499 | cof[i] += ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20); |
||
500 | |||
501 | cof[k] = par[k]; |
||
502 | } |
||
503 | |||
504 | |||
505 | /** Read block switching field if necessary and set actual block sizes. |
||
506 | * Also assure that the block sizes of the last frame correspond to the |
||
507 | * actual number of samples. |
||
508 | */ |
||
509 | static void get_block_sizes(ALSDecContext *ctx, unsigned int *div_blocks, |
||
510 | uint32_t *bs_info) |
||
511 | { |
||
512 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
513 | GetBitContext *gb = &ctx->gb; |
||
514 | unsigned int *ptr_div_blocks = div_blocks; |
||
515 | unsigned int b; |
||
516 | |||
517 | if (sconf->block_switching) { |
||
518 | unsigned int bs_info_len = 1 << (sconf->block_switching + 2); |
||
519 | *bs_info = get_bits_long(gb, bs_info_len); |
||
520 | *bs_info <<= (32 - bs_info_len); |
||
521 | } |
||
522 | |||
523 | ctx->num_blocks = 0; |
||
524 | parse_bs_info(*bs_info, 0, 0, &ptr_div_blocks, &ctx->num_blocks); |
||
525 | |||
526 | // The last frame may have an overdetermined block structure given in |
||
527 | // the bitstream. In that case the defined block structure would need |
||
528 | // more samples than available to be consistent. |
||
529 | // The block structure is actually used but the block sizes are adapted |
||
530 | // to fit the actual number of available samples. |
||
531 | // Example: 5 samples, 2nd level block sizes: 2 2 2 2. |
||
532 | // This results in the actual block sizes: 2 2 1 0. |
||
533 | // This is not specified in 14496-3 but actually done by the reference |
||
534 | // codec RM22 revision 2. |
||
535 | // This appears to happen in case of an odd number of samples in the last |
||
536 | // frame which is actually not allowed by the block length switching part |
||
537 | // of 14496-3. |
||
538 | // The ALS conformance files feature an odd number of samples in the last |
||
539 | // frame. |
||
540 | |||
541 | for (b = 0; b < ctx->num_blocks; b++) |
||
542 | div_blocks[b] = ctx->sconf.frame_length >> div_blocks[b]; |
||
543 | |||
544 | if (ctx->cur_frame_length != ctx->sconf.frame_length) { |
||
545 | unsigned int remaining = ctx->cur_frame_length; |
||
546 | |||
547 | for (b = 0; b < ctx->num_blocks; b++) { |
||
548 | if (remaining <= div_blocks[b]) { |
||
549 | div_blocks[b] = remaining; |
||
550 | ctx->num_blocks = b + 1; |
||
551 | break; |
||
552 | } |
||
553 | |||
554 | remaining -= div_blocks[b]; |
||
555 | } |
||
556 | } |
||
557 | } |
||
558 | |||
559 | |||
560 | /** Read the block data for a constant block |
||
561 | */ |
||
562 | static int read_const_block_data(ALSDecContext *ctx, ALSBlockData *bd) |
||
563 | { |
||
564 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
565 | AVCodecContext *avctx = ctx->avctx; |
||
566 | GetBitContext *gb = &ctx->gb; |
||
567 | |||
568 | if (bd->block_length <= 0) |
||
569 | return AVERROR_INVALIDDATA; |
||
570 | |||
571 | *bd->raw_samples = 0; |
||
572 | *bd->const_block = get_bits1(gb); // 1 = constant value, 0 = zero block (silence) |
||
573 | bd->js_blocks = get_bits1(gb); |
||
574 | |||
575 | // skip 5 reserved bits |
||
576 | skip_bits(gb, 5); |
||
577 | |||
578 | if (*bd->const_block) { |
||
579 | unsigned int const_val_bits = sconf->floating ? 24 : avctx->bits_per_raw_sample; |
||
580 | *bd->raw_samples = get_sbits_long(gb, const_val_bits); |
||
581 | } |
||
582 | |||
583 | // ensure constant block decoding by reusing this field |
||
584 | *bd->const_block = 1; |
||
585 | |||
586 | return 0; |
||
587 | } |
||
588 | |||
589 | |||
590 | /** Decode the block data for a constant block |
||
591 | */ |
||
592 | static void decode_const_block_data(ALSDecContext *ctx, ALSBlockData *bd) |
||
593 | { |
||
594 | int smp = bd->block_length - 1; |
||
595 | int32_t val = *bd->raw_samples; |
||
596 | int32_t *dst = bd->raw_samples + 1; |
||
597 | |||
598 | // write raw samples into buffer |
||
599 | for (; smp; smp--) |
||
600 | *dst++ = val; |
||
601 | } |
||
602 | |||
603 | |||
604 | /** Read the block data for a non-constant block |
||
605 | */ |
||
606 | static int read_var_block_data(ALSDecContext *ctx, ALSBlockData *bd) |
||
607 | { |
||
608 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
609 | AVCodecContext *avctx = ctx->avctx; |
||
610 | GetBitContext *gb = &ctx->gb; |
||
611 | unsigned int k; |
||
612 | unsigned int s[8]; |
||
613 | unsigned int sx[8]; |
||
614 | unsigned int sub_blocks, log2_sub_blocks, sb_length; |
||
615 | unsigned int start = 0; |
||
616 | unsigned int opt_order; |
||
617 | int sb; |
||
618 | int32_t *quant_cof = bd->quant_cof; |
||
619 | int32_t *current_res; |
||
620 | |||
621 | |||
622 | // ensure variable block decoding by reusing this field |
||
623 | *bd->const_block = 0; |
||
624 | |||
625 | *bd->opt_order = 1; |
||
626 | bd->js_blocks = get_bits1(gb); |
||
627 | |||
628 | opt_order = *bd->opt_order; |
||
629 | |||
630 | // determine the number of subblocks for entropy decoding |
||
631 | if (!sconf->bgmc && !sconf->sb_part) { |
||
632 | log2_sub_blocks = 0; |
||
633 | } else { |
||
634 | if (sconf->bgmc && sconf->sb_part) |
||
635 | log2_sub_blocks = get_bits(gb, 2); |
||
636 | else |
||
637 | log2_sub_blocks = 2 * get_bits1(gb); |
||
638 | } |
||
639 | |||
640 | sub_blocks = 1 << log2_sub_blocks; |
||
641 | |||
642 | // do not continue in case of a damaged stream since |
||
643 | // block_length must be evenly divisible by sub_blocks |
||
644 | if (bd->block_length & (sub_blocks - 1)) { |
||
645 | av_log(avctx, AV_LOG_WARNING, |
||
646 | "Block length is not evenly divisible by the number of subblocks.\n"); |
||
647 | return AVERROR_INVALIDDATA; |
||
648 | } |
||
649 | |||
650 | sb_length = bd->block_length >> log2_sub_blocks; |
||
651 | |||
652 | if (sconf->bgmc) { |
||
653 | s[0] = get_bits(gb, 8 + (sconf->resolution > 1)); |
||
654 | for (k = 1; k < sub_blocks; k++) |
||
655 | s[k] = s[k - 1] + decode_rice(gb, 2); |
||
656 | |||
657 | for (k = 0; k < sub_blocks; k++) { |
||
658 | sx[k] = s[k] & 0x0F; |
||
659 | s [k] >>= 4; |
||
660 | } |
||
661 | } else { |
||
662 | s[0] = get_bits(gb, 4 + (sconf->resolution > 1)); |
||
663 | for (k = 1; k < sub_blocks; k++) |
||
664 | s[k] = s[k - 1] + decode_rice(gb, 0); |
||
665 | } |
||
666 | for (k = 1; k < sub_blocks; k++) |
||
667 | if (s[k] > 32) { |
||
668 | av_log(avctx, AV_LOG_ERROR, "k invalid for rice code.\n"); |
||
669 | return AVERROR_INVALIDDATA; |
||
670 | } |
||
671 | |||
672 | if (get_bits1(gb)) |
||
673 | *bd->shift_lsbs = get_bits(gb, 4) + 1; |
||
674 | |||
675 | *bd->store_prev_samples = (bd->js_blocks && bd->raw_other) || *bd->shift_lsbs; |
||
676 | |||
677 | |||
678 | if (!sconf->rlslms) { |
||
679 | if (sconf->adapt_order) { |
||
680 | int opt_order_length = av_ceil_log2(av_clip((bd->block_length >> 3) - 1, |
||
681 | 2, sconf->max_order + 1)); |
||
682 | *bd->opt_order = get_bits(gb, opt_order_length); |
||
683 | if (*bd->opt_order > sconf->max_order) { |
||
684 | *bd->opt_order = sconf->max_order; |
||
685 | av_log(avctx, AV_LOG_ERROR, "Predictor order too large.\n"); |
||
686 | return AVERROR_INVALIDDATA; |
||
687 | } |
||
688 | } else { |
||
689 | *bd->opt_order = sconf->max_order; |
||
690 | } |
||
691 | |||
692 | opt_order = *bd->opt_order; |
||
693 | |||
694 | if (opt_order) { |
||
695 | int add_base; |
||
696 | |||
697 | if (sconf->coef_table == 3) { |
||
698 | add_base = 0x7F; |
||
699 | |||
700 | // read coefficient 0 |
||
701 | quant_cof[0] = 32 * parcor_scaled_values[get_bits(gb, 7)]; |
||
702 | |||
703 | // read coefficient 1 |
||
704 | if (opt_order > 1) |
||
705 | quant_cof[1] = -32 * parcor_scaled_values[get_bits(gb, 7)]; |
||
706 | |||
707 | // read coefficients 2 to opt_order |
||
708 | for (k = 2; k < opt_order; k++) |
||
709 | quant_cof[k] = get_bits(gb, 7); |
||
710 | } else { |
||
711 | int k_max; |
||
712 | add_base = 1; |
||
713 | |||
714 | // read coefficient 0 to 19 |
||
715 | k_max = FFMIN(opt_order, 20); |
||
716 | for (k = 0; k < k_max; k++) { |
||
717 | int rice_param = parcor_rice_table[sconf->coef_table][k][1]; |
||
718 | int offset = parcor_rice_table[sconf->coef_table][k][0]; |
||
719 | quant_cof[k] = decode_rice(gb, rice_param) + offset; |
||
720 | if (quant_cof[k] < -64 || quant_cof[k] > 63) { |
||
721 | av_log(avctx, AV_LOG_ERROR, "quant_cof %d is out of range.\n", quant_cof[k]); |
||
722 | return AVERROR_INVALIDDATA; |
||
723 | } |
||
724 | } |
||
725 | |||
726 | // read coefficients 20 to 126 |
||
727 | k_max = FFMIN(opt_order, 127); |
||
728 | for (; k < k_max; k++) |
||
729 | quant_cof[k] = decode_rice(gb, 2) + (k & 1); |
||
730 | |||
731 | // read coefficients 127 to opt_order |
||
732 | for (; k < opt_order; k++) |
||
733 | quant_cof[k] = decode_rice(gb, 1); |
||
734 | |||
735 | quant_cof[0] = 32 * parcor_scaled_values[quant_cof[0] + 64]; |
||
736 | |||
737 | if (opt_order > 1) |
||
738 | quant_cof[1] = -32 * parcor_scaled_values[quant_cof[1] + 64]; |
||
739 | } |
||
740 | |||
741 | for (k = 2; k < opt_order; k++) |
||
742 | quant_cof[k] = (quant_cof[k] << 14) + (add_base << 13); |
||
743 | } |
||
744 | } |
||
745 | |||
746 | // read LTP gain and lag values |
||
747 | if (sconf->long_term_prediction) { |
||
748 | *bd->use_ltp = get_bits1(gb); |
||
749 | |||
750 | if (*bd->use_ltp) { |
||
751 | int r, c; |
||
752 | |||
753 | bd->ltp_gain[0] = decode_rice(gb, 1) << 3; |
||
754 | bd->ltp_gain[1] = decode_rice(gb, 2) << 3; |
||
755 | |||
756 | r = get_unary(gb, 0, 3); |
||
757 | c = get_bits(gb, 2); |
||
758 | bd->ltp_gain[2] = ltp_gain_values[r][c]; |
||
759 | |||
760 | bd->ltp_gain[3] = decode_rice(gb, 2) << 3; |
||
761 | bd->ltp_gain[4] = decode_rice(gb, 1) << 3; |
||
762 | |||
763 | *bd->ltp_lag = get_bits(gb, ctx->ltp_lag_length); |
||
764 | *bd->ltp_lag += FFMAX(4, opt_order + 1); |
||
765 | } |
||
766 | } |
||
767 | |||
768 | // read first value and residuals in case of a random access block |
||
769 | if (bd->ra_block) { |
||
770 | if (opt_order) |
||
771 | bd->raw_samples[0] = decode_rice(gb, avctx->bits_per_raw_sample - 4); |
||
772 | if (opt_order > 1) |
||
773 | bd->raw_samples[1] = decode_rice(gb, FFMIN(s[0] + 3, ctx->s_max)); |
||
774 | if (opt_order > 2) |
||
775 | bd->raw_samples[2] = decode_rice(gb, FFMIN(s[0] + 1, ctx->s_max)); |
||
776 | |||
777 | start = FFMIN(opt_order, 3); |
||
778 | } |
||
779 | |||
780 | // read all residuals |
||
781 | if (sconf->bgmc) { |
||
782 | int delta[8]; |
||
783 | unsigned int k [8]; |
||
784 | unsigned int b = av_clip((av_ceil_log2(bd->block_length) - 3) >> 1, 0, 5); |
||
785 | |||
786 | // read most significant bits |
||
787 | unsigned int high; |
||
788 | unsigned int low; |
||
789 | unsigned int value; |
||
790 | |||
791 | ff_bgmc_decode_init(gb, &high, &low, &value); |
||
792 | |||
793 | current_res = bd->raw_samples + start; |
||
794 | |||
795 | for (sb = 0; sb < sub_blocks; sb++) { |
||
796 | unsigned int sb_len = sb_length - (sb ? 0 : start); |
||
797 | |||
798 | k [sb] = s[sb] > b ? s[sb] - b : 0; |
||
799 | delta[sb] = 5 - s[sb] + k[sb]; |
||
800 | |||
801 | ff_bgmc_decode(gb, sb_len, current_res, |
||
802 | delta[sb], sx[sb], &high, &low, &value, ctx->bgmc_lut, ctx->bgmc_lut_status); |
||
803 | |||
804 | current_res += sb_len; |
||
805 | } |
||
806 | |||
807 | ff_bgmc_decode_end(gb); |
||
808 | |||
809 | |||
810 | // read least significant bits and tails |
||
811 | current_res = bd->raw_samples + start; |
||
812 | |||
813 | for (sb = 0; sb < sub_blocks; sb++, start = 0) { |
||
814 | unsigned int cur_tail_code = tail_code[sx[sb]][delta[sb]]; |
||
815 | unsigned int cur_k = k[sb]; |
||
816 | unsigned int cur_s = s[sb]; |
||
817 | |||
818 | for (; start < sb_length; start++) { |
||
819 | int32_t res = *current_res; |
||
820 | |||
821 | if (res == cur_tail_code) { |
||
822 | unsigned int max_msb = (2 + (sx[sb] > 2) + (sx[sb] > 10)) |
||
823 | << (5 - delta[sb]); |
||
824 | |||
825 | res = decode_rice(gb, cur_s); |
||
826 | |||
827 | if (res >= 0) { |
||
828 | res += (max_msb ) << cur_k; |
||
829 | } else { |
||
830 | res -= (max_msb - 1) << cur_k; |
||
831 | } |
||
832 | } else { |
||
833 | if (res > cur_tail_code) |
||
834 | res--; |
||
835 | |||
836 | if (res & 1) |
||
837 | res = -res; |
||
838 | |||
839 | res >>= 1; |
||
840 | |||
841 | if (cur_k) { |
||
842 | res <<= cur_k; |
||
843 | res |= get_bits_long(gb, cur_k); |
||
844 | } |
||
845 | } |
||
846 | |||
847 | *current_res++ = res; |
||
848 | } |
||
849 | } |
||
850 | } else { |
||
851 | current_res = bd->raw_samples + start; |
||
852 | |||
853 | for (sb = 0; sb < sub_blocks; sb++, start = 0) |
||
854 | for (; start < sb_length; start++) |
||
855 | *current_res++ = decode_rice(gb, s[sb]); |
||
856 | } |
||
857 | |||
858 | if (!sconf->mc_coding || ctx->js_switch) |
||
859 | align_get_bits(gb); |
||
860 | |||
861 | return 0; |
||
862 | } |
||
863 | |||
864 | |||
865 | /** Decode the block data for a non-constant block |
||
866 | */ |
||
867 | static int decode_var_block_data(ALSDecContext *ctx, ALSBlockData *bd) |
||
868 | { |
||
869 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
870 | unsigned int block_length = bd->block_length; |
||
871 | unsigned int smp = 0; |
||
872 | unsigned int k; |
||
873 | int opt_order = *bd->opt_order; |
||
874 | int sb; |
||
875 | int64_t y; |
||
876 | int32_t *quant_cof = bd->quant_cof; |
||
877 | int32_t *lpc_cof = bd->lpc_cof; |
||
878 | int32_t *raw_samples = bd->raw_samples; |
||
879 | int32_t *raw_samples_end = bd->raw_samples + bd->block_length; |
||
880 | int32_t *lpc_cof_reversed = ctx->lpc_cof_reversed_buffer; |
||
881 | |||
882 | // reverse long-term prediction |
||
883 | if (*bd->use_ltp) { |
||
884 | int ltp_smp; |
||
885 | |||
886 | for (ltp_smp = FFMAX(*bd->ltp_lag - 2, 0); ltp_smp < block_length; ltp_smp++) { |
||
887 | int center = ltp_smp - *bd->ltp_lag; |
||
888 | int begin = FFMAX(0, center - 2); |
||
889 | int end = center + 3; |
||
890 | int tab = 5 - (end - begin); |
||
891 | int base; |
||
892 | |||
893 | y = 1 << 6; |
||
894 | |||
895 | for (base = begin; base < end; base++, tab++) |
||
896 | y += MUL64(bd->ltp_gain[tab], raw_samples[base]); |
||
897 | |||
898 | raw_samples[ltp_smp] += y >> 7; |
||
899 | } |
||
900 | } |
||
901 | |||
902 | // reconstruct all samples from residuals |
||
903 | if (bd->ra_block) { |
||
904 | for (smp = 0; smp < opt_order; smp++) { |
||
905 | y = 1 << 19; |
||
906 | |||
907 | for (sb = 0; sb < smp; sb++) |
||
908 | y += MUL64(lpc_cof[sb], raw_samples[-(sb + 1)]); |
||
909 | |||
910 | *raw_samples++ -= y >> 20; |
||
911 | parcor_to_lpc(smp, quant_cof, lpc_cof); |
||
912 | } |
||
913 | } else { |
||
914 | for (k = 0; k < opt_order; k++) |
||
915 | parcor_to_lpc(k, quant_cof, lpc_cof); |
||
916 | |||
917 | // store previous samples in case that they have to be altered |
||
918 | if (*bd->store_prev_samples) |
||
919 | memcpy(bd->prev_raw_samples, raw_samples - sconf->max_order, |
||
920 | sizeof(*bd->prev_raw_samples) * sconf->max_order); |
||
921 | |||
922 | // reconstruct difference signal for prediction (joint-stereo) |
||
923 | if (bd->js_blocks && bd->raw_other) { |
||
924 | int32_t *left, *right; |
||
925 | |||
926 | if (bd->raw_other > raw_samples) { // D = R - L |
||
927 | left = raw_samples; |
||
928 | right = bd->raw_other; |
||
929 | } else { // D = R - L |
||
930 | left = bd->raw_other; |
||
931 | right = raw_samples; |
||
932 | } |
||
933 | |||
934 | for (sb = -1; sb >= -sconf->max_order; sb--) |
||
935 | raw_samples[sb] = right[sb] - left[sb]; |
||
936 | } |
||
937 | |||
938 | // reconstruct shifted signal |
||
939 | if (*bd->shift_lsbs) |
||
940 | for (sb = -1; sb >= -sconf->max_order; sb--) |
||
941 | raw_samples[sb] >>= *bd->shift_lsbs; |
||
942 | } |
||
943 | |||
944 | // reverse linear prediction coefficients for efficiency |
||
945 | lpc_cof = lpc_cof + opt_order; |
||
946 | |||
947 | for (sb = 0; sb < opt_order; sb++) |
||
948 | lpc_cof_reversed[sb] = lpc_cof[-(sb + 1)]; |
||
949 | |||
950 | // reconstruct raw samples |
||
951 | raw_samples = bd->raw_samples + smp; |
||
952 | lpc_cof = lpc_cof_reversed + opt_order; |
||
953 | |||
954 | for (; raw_samples < raw_samples_end; raw_samples++) { |
||
955 | y = 1 << 19; |
||
956 | |||
957 | for (sb = -opt_order; sb < 0; sb++) |
||
958 | y += MUL64(lpc_cof[sb], raw_samples[sb]); |
||
959 | |||
960 | *raw_samples -= y >> 20; |
||
961 | } |
||
962 | |||
963 | raw_samples = bd->raw_samples; |
||
964 | |||
965 | // restore previous samples in case that they have been altered |
||
966 | if (*bd->store_prev_samples) |
||
967 | memcpy(raw_samples - sconf->max_order, bd->prev_raw_samples, |
||
968 | sizeof(*raw_samples) * sconf->max_order); |
||
969 | |||
970 | return 0; |
||
971 | } |
||
972 | |||
973 | |||
974 | /** Read the block data. |
||
975 | */ |
||
976 | static int read_block(ALSDecContext *ctx, ALSBlockData *bd) |
||
977 | { |
||
978 | int ret; |
||
979 | GetBitContext *gb = &ctx->gb; |
||
980 | |||
981 | *bd->shift_lsbs = 0; |
||
982 | // read block type flag and read the samples accordingly |
||
983 | if (get_bits1(gb)) { |
||
984 | ret = read_var_block_data(ctx, bd); |
||
985 | } else { |
||
986 | ret = read_const_block_data(ctx, bd); |
||
987 | } |
||
988 | |||
989 | return ret; |
||
990 | } |
||
991 | |||
992 | |||
993 | /** Decode the block data. |
||
994 | */ |
||
995 | static int decode_block(ALSDecContext *ctx, ALSBlockData *bd) |
||
996 | { |
||
997 | unsigned int smp; |
||
998 | int ret = 0; |
||
999 | |||
1000 | // read block type flag and read the samples accordingly |
||
1001 | if (*bd->const_block) |
||
1002 | decode_const_block_data(ctx, bd); |
||
1003 | else |
||
1004 | ret = decode_var_block_data(ctx, bd); // always return 0 |
||
1005 | |||
1006 | if (ret < 0) |
||
1007 | return ret; |
||
1008 | |||
1009 | // TODO: read RLSLMS extension data |
||
1010 | |||
1011 | if (*bd->shift_lsbs) |
||
1012 | for (smp = 0; smp < bd->block_length; smp++) |
||
1013 | bd->raw_samples[smp] <<= *bd->shift_lsbs; |
||
1014 | |||
1015 | return 0; |
||
1016 | } |
||
1017 | |||
1018 | |||
1019 | /** Read and decode block data successively. |
||
1020 | */ |
||
1021 | static int read_decode_block(ALSDecContext *ctx, ALSBlockData *bd) |
||
1022 | { |
||
1023 | int ret; |
||
1024 | |||
1025 | if ((ret = read_block(ctx, bd)) < 0) |
||
1026 | return ret; |
||
1027 | |||
1028 | return decode_block(ctx, bd); |
||
1029 | } |
||
1030 | |||
1031 | |||
1032 | /** Compute the number of samples left to decode for the current frame and |
||
1033 | * sets these samples to zero. |
||
1034 | */ |
||
1035 | static void zero_remaining(unsigned int b, unsigned int b_max, |
||
1036 | const unsigned int *div_blocks, int32_t *buf) |
||
1037 | { |
||
1038 | unsigned int count = 0; |
||
1039 | |||
1040 | while (b < b_max) |
||
1041 | count += div_blocks[b++]; |
||
1042 | |||
1043 | if (count) |
||
1044 | memset(buf, 0, sizeof(*buf) * count); |
||
1045 | } |
||
1046 | |||
1047 | |||
1048 | /** Decode blocks independently. |
||
1049 | */ |
||
1050 | static int decode_blocks_ind(ALSDecContext *ctx, unsigned int ra_frame, |
||
1051 | unsigned int c, const unsigned int *div_blocks, |
||
1052 | unsigned int *js_blocks) |
||
1053 | { |
||
1054 | int ret; |
||
1055 | unsigned int b; |
||
1056 | ALSBlockData bd = { 0 }; |
||
1057 | |||
1058 | bd.ra_block = ra_frame; |
||
1059 | bd.const_block = ctx->const_block; |
||
1060 | bd.shift_lsbs = ctx->shift_lsbs; |
||
1061 | bd.opt_order = ctx->opt_order; |
||
1062 | bd.store_prev_samples = ctx->store_prev_samples; |
||
1063 | bd.use_ltp = ctx->use_ltp; |
||
1064 | bd.ltp_lag = ctx->ltp_lag; |
||
1065 | bd.ltp_gain = ctx->ltp_gain[0]; |
||
1066 | bd.quant_cof = ctx->quant_cof[0]; |
||
1067 | bd.lpc_cof = ctx->lpc_cof[0]; |
||
1068 | bd.prev_raw_samples = ctx->prev_raw_samples; |
||
1069 | bd.raw_samples = ctx->raw_samples[c]; |
||
1070 | |||
1071 | |||
1072 | for (b = 0; b < ctx->num_blocks; b++) { |
||
1073 | bd.block_length = div_blocks[b]; |
||
1074 | |||
1075 | if ((ret = read_decode_block(ctx, &bd)) < 0) { |
||
1076 | // damaged block, write zero for the rest of the frame |
||
1077 | zero_remaining(b, ctx->num_blocks, div_blocks, bd.raw_samples); |
||
1078 | return ret; |
||
1079 | } |
||
1080 | bd.raw_samples += div_blocks[b]; |
||
1081 | bd.ra_block = 0; |
||
1082 | } |
||
1083 | |||
1084 | return 0; |
||
1085 | } |
||
1086 | |||
1087 | |||
1088 | /** Decode blocks dependently. |
||
1089 | */ |
||
1090 | static int decode_blocks(ALSDecContext *ctx, unsigned int ra_frame, |
||
1091 | unsigned int c, const unsigned int *div_blocks, |
||
1092 | unsigned int *js_blocks) |
||
1093 | { |
||
1094 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
1095 | unsigned int offset = 0; |
||
1096 | unsigned int b; |
||
1097 | int ret; |
||
1098 | ALSBlockData bd[2] = { { 0 } }; |
||
1099 | |||
1100 | bd[0].ra_block = ra_frame; |
||
1101 | bd[0].const_block = ctx->const_block; |
||
1102 | bd[0].shift_lsbs = ctx->shift_lsbs; |
||
1103 | bd[0].opt_order = ctx->opt_order; |
||
1104 | bd[0].store_prev_samples = ctx->store_prev_samples; |
||
1105 | bd[0].use_ltp = ctx->use_ltp; |
||
1106 | bd[0].ltp_lag = ctx->ltp_lag; |
||
1107 | bd[0].ltp_gain = ctx->ltp_gain[0]; |
||
1108 | bd[0].quant_cof = ctx->quant_cof[0]; |
||
1109 | bd[0].lpc_cof = ctx->lpc_cof[0]; |
||
1110 | bd[0].prev_raw_samples = ctx->prev_raw_samples; |
||
1111 | bd[0].js_blocks = *js_blocks; |
||
1112 | |||
1113 | bd[1].ra_block = ra_frame; |
||
1114 | bd[1].const_block = ctx->const_block; |
||
1115 | bd[1].shift_lsbs = ctx->shift_lsbs; |
||
1116 | bd[1].opt_order = ctx->opt_order; |
||
1117 | bd[1].store_prev_samples = ctx->store_prev_samples; |
||
1118 | bd[1].use_ltp = ctx->use_ltp; |
||
1119 | bd[1].ltp_lag = ctx->ltp_lag; |
||
1120 | bd[1].ltp_gain = ctx->ltp_gain[0]; |
||
1121 | bd[1].quant_cof = ctx->quant_cof[0]; |
||
1122 | bd[1].lpc_cof = ctx->lpc_cof[0]; |
||
1123 | bd[1].prev_raw_samples = ctx->prev_raw_samples; |
||
1124 | bd[1].js_blocks = *(js_blocks + 1); |
||
1125 | |||
1126 | // decode all blocks |
||
1127 | for (b = 0; b < ctx->num_blocks; b++) { |
||
1128 | unsigned int s; |
||
1129 | |||
1130 | bd[0].block_length = div_blocks[b]; |
||
1131 | bd[1].block_length = div_blocks[b]; |
||
1132 | |||
1133 | bd[0].raw_samples = ctx->raw_samples[c ] + offset; |
||
1134 | bd[1].raw_samples = ctx->raw_samples[c + 1] + offset; |
||
1135 | |||
1136 | bd[0].raw_other = bd[1].raw_samples; |
||
1137 | bd[1].raw_other = bd[0].raw_samples; |
||
1138 | |||
1139 | if ((ret = read_decode_block(ctx, &bd[0])) < 0 || |
||
1140 | (ret = read_decode_block(ctx, &bd[1])) < 0) |
||
1141 | goto fail; |
||
1142 | |||
1143 | // reconstruct joint-stereo blocks |
||
1144 | if (bd[0].js_blocks) { |
||
1145 | if (bd[1].js_blocks) |
||
1146 | av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel pair.\n"); |
||
1147 | |||
1148 | for (s = 0; s < div_blocks[b]; s++) |
||
1149 | bd[0].raw_samples[s] = bd[1].raw_samples[s] - bd[0].raw_samples[s]; |
||
1150 | } else if (bd[1].js_blocks) { |
||
1151 | for (s = 0; s < div_blocks[b]; s++) |
||
1152 | bd[1].raw_samples[s] = bd[1].raw_samples[s] + bd[0].raw_samples[s]; |
||
1153 | } |
||
1154 | |||
1155 | offset += div_blocks[b]; |
||
1156 | bd[0].ra_block = 0; |
||
1157 | bd[1].ra_block = 0; |
||
1158 | } |
||
1159 | |||
1160 | // store carryover raw samples, |
||
1161 | // the others channel raw samples are stored by the calling function. |
||
1162 | memmove(ctx->raw_samples[c] - sconf->max_order, |
||
1163 | ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, |
||
1164 | sizeof(*ctx->raw_samples[c]) * sconf->max_order); |
||
1165 | |||
1166 | return 0; |
||
1167 | fail: |
||
1168 | // damaged block, write zero for the rest of the frame |
||
1169 | zero_remaining(b, ctx->num_blocks, div_blocks, bd[0].raw_samples); |
||
1170 | zero_remaining(b, ctx->num_blocks, div_blocks, bd[1].raw_samples); |
||
1171 | return ret; |
||
1172 | } |
||
1173 | |||
1174 | static inline int als_weighting(GetBitContext *gb, int k, int off) |
||
1175 | { |
||
1176 | int idx = av_clip(decode_rice(gb, k) + off, |
||
1177 | 0, FF_ARRAY_ELEMS(mcc_weightings) - 1); |
||
1178 | return mcc_weightings[idx]; |
||
1179 | } |
||
1180 | |||
1181 | /** Read the channel data. |
||
1182 | */ |
||
1183 | static int read_channel_data(ALSDecContext *ctx, ALSChannelData *cd, int c) |
||
1184 | { |
||
1185 | GetBitContext *gb = &ctx->gb; |
||
1186 | ALSChannelData *current = cd; |
||
1187 | unsigned int channels = ctx->avctx->channels; |
||
1188 | int entries = 0; |
||
1189 | |||
1190 | while (entries < channels && !(current->stop_flag = get_bits1(gb))) { |
||
1191 | current->master_channel = get_bits_long(gb, av_ceil_log2(channels)); |
||
1192 | |||
1193 | if (current->master_channel >= channels) { |
||
1194 | av_log(ctx->avctx, AV_LOG_ERROR, "Invalid master channel.\n"); |
||
1195 | return AVERROR_INVALIDDATA; |
||
1196 | } |
||
1197 | |||
1198 | if (current->master_channel != c) { |
||
1199 | current->time_diff_flag = get_bits1(gb); |
||
1200 | current->weighting[0] = als_weighting(gb, 1, 16); |
||
1201 | current->weighting[1] = als_weighting(gb, 2, 14); |
||
1202 | current->weighting[2] = als_weighting(gb, 1, 16); |
||
1203 | |||
1204 | if (current->time_diff_flag) { |
||
1205 | current->weighting[3] = als_weighting(gb, 1, 16); |
||
1206 | current->weighting[4] = als_weighting(gb, 1, 16); |
||
1207 | current->weighting[5] = als_weighting(gb, 1, 16); |
||
1208 | |||
1209 | current->time_diff_sign = get_bits1(gb); |
||
1210 | current->time_diff_index = get_bits(gb, ctx->ltp_lag_length - 3) + 3; |
||
1211 | } |
||
1212 | } |
||
1213 | |||
1214 | current++; |
||
1215 | entries++; |
||
1216 | } |
||
1217 | |||
1218 | if (entries == channels) { |
||
1219 | av_log(ctx->avctx, AV_LOG_ERROR, "Damaged channel data.\n"); |
||
1220 | return AVERROR_INVALIDDATA; |
||
1221 | } |
||
1222 | |||
1223 | align_get_bits(gb); |
||
1224 | return 0; |
||
1225 | } |
||
1226 | |||
1227 | |||
1228 | /** Recursively reverts the inter-channel correlation for a block. |
||
1229 | */ |
||
1230 | static int revert_channel_correlation(ALSDecContext *ctx, ALSBlockData *bd, |
||
1231 | ALSChannelData **cd, int *reverted, |
||
1232 | unsigned int offset, int c) |
||
1233 | { |
||
1234 | ALSChannelData *ch = cd[c]; |
||
1235 | unsigned int dep = 0; |
||
1236 | unsigned int channels = ctx->avctx->channels; |
||
1237 | |||
1238 | if (reverted[c]) |
||
1239 | return 0; |
||
1240 | |||
1241 | reverted[c] = 1; |
||
1242 | |||
1243 | while (dep < channels && !ch[dep].stop_flag) { |
||
1244 | revert_channel_correlation(ctx, bd, cd, reverted, offset, |
||
1245 | ch[dep].master_channel); |
||
1246 | |||
1247 | dep++; |
||
1248 | } |
||
1249 | |||
1250 | if (dep == channels) { |
||
1251 | av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel correlation.\n"); |
||
1252 | return AVERROR_INVALIDDATA; |
||
1253 | } |
||
1254 | |||
1255 | bd->const_block = ctx->const_block + c; |
||
1256 | bd->shift_lsbs = ctx->shift_lsbs + c; |
||
1257 | bd->opt_order = ctx->opt_order + c; |
||
1258 | bd->store_prev_samples = ctx->store_prev_samples + c; |
||
1259 | bd->use_ltp = ctx->use_ltp + c; |
||
1260 | bd->ltp_lag = ctx->ltp_lag + c; |
||
1261 | bd->ltp_gain = ctx->ltp_gain[c]; |
||
1262 | bd->lpc_cof = ctx->lpc_cof[c]; |
||
1263 | bd->quant_cof = ctx->quant_cof[c]; |
||
1264 | bd->raw_samples = ctx->raw_samples[c] + offset; |
||
1265 | |||
1266 | dep = 0; |
||
1267 | while (!ch[dep].stop_flag) { |
||
1268 | unsigned int smp; |
||
1269 | unsigned int begin = 1; |
||
1270 | unsigned int end = bd->block_length - 1; |
||
1271 | int64_t y; |
||
1272 | int32_t *master = ctx->raw_samples[ch[dep].master_channel] + offset; |
||
1273 | |||
1274 | if (ch[dep].time_diff_flag) { |
||
1275 | int t = ch[dep].time_diff_index; |
||
1276 | |||
1277 | if (ch[dep].time_diff_sign) { |
||
1278 | t = -t; |
||
1279 | begin -= t; |
||
1280 | } else { |
||
1281 | end -= t; |
||
1282 | } |
||
1283 | |||
1284 | for (smp = begin; smp < end; smp++) { |
||
1285 | y = (1 << 6) + |
||
1286 | MUL64(ch[dep].weighting[0], master[smp - 1 ]) + |
||
1287 | MUL64(ch[dep].weighting[1], master[smp ]) + |
||
1288 | MUL64(ch[dep].weighting[2], master[smp + 1 ]) + |
||
1289 | MUL64(ch[dep].weighting[3], master[smp - 1 + t]) + |
||
1290 | MUL64(ch[dep].weighting[4], master[smp + t]) + |
||
1291 | MUL64(ch[dep].weighting[5], master[smp + 1 + t]); |
||
1292 | |||
1293 | bd->raw_samples[smp] += y >> 7; |
||
1294 | } |
||
1295 | } else { |
||
1296 | for (smp = begin; smp < end; smp++) { |
||
1297 | y = (1 << 6) + |
||
1298 | MUL64(ch[dep].weighting[0], master[smp - 1]) + |
||
1299 | MUL64(ch[dep].weighting[1], master[smp ]) + |
||
1300 | MUL64(ch[dep].weighting[2], master[smp + 1]); |
||
1301 | |||
1302 | bd->raw_samples[smp] += y >> 7; |
||
1303 | } |
||
1304 | } |
||
1305 | |||
1306 | dep++; |
||
1307 | } |
||
1308 | |||
1309 | return 0; |
||
1310 | } |
||
1311 | |||
1312 | |||
1313 | /** Read the frame data. |
||
1314 | */ |
||
1315 | static int read_frame_data(ALSDecContext *ctx, unsigned int ra_frame) |
||
1316 | { |
||
1317 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
1318 | AVCodecContext *avctx = ctx->avctx; |
||
1319 | GetBitContext *gb = &ctx->gb; |
||
1320 | unsigned int div_blocks[32]; ///< block sizes. |
||
1321 | unsigned int c; |
||
1322 | unsigned int js_blocks[2]; |
||
1323 | uint32_t bs_info = 0; |
||
1324 | int ret; |
||
1325 | |||
1326 | // skip the size of the ra unit if present in the frame |
||
1327 | if (sconf->ra_flag == RA_FLAG_FRAMES && ra_frame) |
||
1328 | skip_bits_long(gb, 32); |
||
1329 | |||
1330 | if (sconf->mc_coding && sconf->joint_stereo) { |
||
1331 | ctx->js_switch = get_bits1(gb); |
||
1332 | align_get_bits(gb); |
||
1333 | } |
||
1334 | |||
1335 | if (!sconf->mc_coding || ctx->js_switch) { |
||
1336 | int independent_bs = !sconf->joint_stereo; |
||
1337 | |||
1338 | for (c = 0; c < avctx->channels; c++) { |
||
1339 | js_blocks[0] = 0; |
||
1340 | js_blocks[1] = 0; |
||
1341 | |||
1342 | get_block_sizes(ctx, div_blocks, &bs_info); |
||
1343 | |||
1344 | // if joint_stereo and block_switching is set, independent decoding |
||
1345 | // is signaled via the first bit of bs_info |
||
1346 | if (sconf->joint_stereo && sconf->block_switching) |
||
1347 | if (bs_info >> 31) |
||
1348 | independent_bs = 2; |
||
1349 | |||
1350 | // if this is the last channel, it has to be decoded independently |
||
1351 | if (c == avctx->channels - 1) |
||
1352 | independent_bs = 1; |
||
1353 | |||
1354 | if (independent_bs) { |
||
1355 | ret = decode_blocks_ind(ctx, ra_frame, c, |
||
1356 | div_blocks, js_blocks); |
||
1357 | if (ret < 0) |
||
1358 | return ret; |
||
1359 | independent_bs--; |
||
1360 | } else { |
||
1361 | ret = decode_blocks(ctx, ra_frame, c, div_blocks, js_blocks); |
||
1362 | if (ret < 0) |
||
1363 | return ret; |
||
1364 | |||
1365 | c++; |
||
1366 | } |
||
1367 | |||
1368 | // store carryover raw samples |
||
1369 | memmove(ctx->raw_samples[c] - sconf->max_order, |
||
1370 | ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, |
||
1371 | sizeof(*ctx->raw_samples[c]) * sconf->max_order); |
||
1372 | } |
||
1373 | } else { // multi-channel coding |
||
1374 | ALSBlockData bd = { 0 }; |
||
1375 | int b, ret; |
||
1376 | int *reverted_channels = ctx->reverted_channels; |
||
1377 | unsigned int offset = 0; |
||
1378 | |||
1379 | for (c = 0; c < avctx->channels; c++) |
||
1380 | if (ctx->chan_data[c] < ctx->chan_data_buffer) { |
||
1381 | av_log(ctx->avctx, AV_LOG_ERROR, "Invalid channel data.\n"); |
||
1382 | return AVERROR_INVALIDDATA; |
||
1383 | } |
||
1384 | |||
1385 | memset(reverted_channels, 0, sizeof(*reverted_channels) * avctx->channels); |
||
1386 | |||
1387 | bd.ra_block = ra_frame; |
||
1388 | bd.prev_raw_samples = ctx->prev_raw_samples; |
||
1389 | |||
1390 | get_block_sizes(ctx, div_blocks, &bs_info); |
||
1391 | |||
1392 | for (b = 0; b < ctx->num_blocks; b++) { |
||
1393 | bd.block_length = div_blocks[b]; |
||
1394 | |||
1395 | for (c = 0; c < avctx->channels; c++) { |
||
1396 | bd.const_block = ctx->const_block + c; |
||
1397 | bd.shift_lsbs = ctx->shift_lsbs + c; |
||
1398 | bd.opt_order = ctx->opt_order + c; |
||
1399 | bd.store_prev_samples = ctx->store_prev_samples + c; |
||
1400 | bd.use_ltp = ctx->use_ltp + c; |
||
1401 | bd.ltp_lag = ctx->ltp_lag + c; |
||
1402 | bd.ltp_gain = ctx->ltp_gain[c]; |
||
1403 | bd.lpc_cof = ctx->lpc_cof[c]; |
||
1404 | bd.quant_cof = ctx->quant_cof[c]; |
||
1405 | bd.raw_samples = ctx->raw_samples[c] + offset; |
||
1406 | bd.raw_other = NULL; |
||
1407 | |||
1408 | if ((ret = read_block(ctx, &bd)) < 0) |
||
1409 | return ret; |
||
1410 | if ((ret = read_channel_data(ctx, ctx->chan_data[c], c)) < 0) |
||
1411 | return ret; |
||
1412 | } |
||
1413 | |||
1414 | for (c = 0; c < avctx->channels; c++) { |
||
1415 | ret = revert_channel_correlation(ctx, &bd, ctx->chan_data, |
||
1416 | reverted_channels, offset, c); |
||
1417 | if (ret < 0) |
||
1418 | return ret; |
||
1419 | } |
||
1420 | for (c = 0; c < avctx->channels; c++) { |
||
1421 | bd.const_block = ctx->const_block + c; |
||
1422 | bd.shift_lsbs = ctx->shift_lsbs + c; |
||
1423 | bd.opt_order = ctx->opt_order + c; |
||
1424 | bd.store_prev_samples = ctx->store_prev_samples + c; |
||
1425 | bd.use_ltp = ctx->use_ltp + c; |
||
1426 | bd.ltp_lag = ctx->ltp_lag + c; |
||
1427 | bd.ltp_gain = ctx->ltp_gain[c]; |
||
1428 | bd.lpc_cof = ctx->lpc_cof[c]; |
||
1429 | bd.quant_cof = ctx->quant_cof[c]; |
||
1430 | bd.raw_samples = ctx->raw_samples[c] + offset; |
||
1431 | |||
1432 | if ((ret = decode_block(ctx, &bd)) < 0) |
||
1433 | return ret; |
||
1434 | } |
||
1435 | |||
1436 | memset(reverted_channels, 0, avctx->channels * sizeof(*reverted_channels)); |
||
1437 | offset += div_blocks[b]; |
||
1438 | bd.ra_block = 0; |
||
1439 | } |
||
1440 | |||
1441 | // store carryover raw samples |
||
1442 | for (c = 0; c < avctx->channels; c++) |
||
1443 | memmove(ctx->raw_samples[c] - sconf->max_order, |
||
1444 | ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, |
||
1445 | sizeof(*ctx->raw_samples[c]) * sconf->max_order); |
||
1446 | } |
||
1447 | |||
1448 | // TODO: read_diff_float_data |
||
1449 | |||
1450 | return 0; |
||
1451 | } |
||
1452 | |||
1453 | |||
1454 | /** Decode an ALS frame. |
||
1455 | */ |
||
1456 | static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, |
||
1457 | AVPacket *avpkt) |
||
1458 | { |
||
1459 | ALSDecContext *ctx = avctx->priv_data; |
||
1460 | AVFrame *frame = data; |
||
1461 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
1462 | const uint8_t *buffer = avpkt->data; |
||
1463 | int buffer_size = avpkt->size; |
||
1464 | int invalid_frame, ret; |
||
1465 | unsigned int c, sample, ra_frame, bytes_read, shift; |
||
1466 | |||
1467 | init_get_bits(&ctx->gb, buffer, buffer_size * 8); |
||
1468 | |||
1469 | // In the case that the distance between random access frames is set to zero |
||
1470 | // (sconf->ra_distance == 0) no frame is treated as a random access frame. |
||
1471 | // For the first frame, if prediction is used, all samples used from the |
||
1472 | // previous frame are assumed to be zero. |
||
1473 | ra_frame = sconf->ra_distance && !(ctx->frame_id % sconf->ra_distance); |
||
1474 | |||
1475 | // the last frame to decode might have a different length |
||
1476 | if (sconf->samples != 0xFFFFFFFF) |
||
1477 | ctx->cur_frame_length = FFMIN(sconf->samples - ctx->frame_id * (uint64_t) sconf->frame_length, |
||
1478 | sconf->frame_length); |
||
1479 | else |
||
1480 | ctx->cur_frame_length = sconf->frame_length; |
||
1481 | |||
1482 | // decode the frame data |
||
1483 | if ((invalid_frame = read_frame_data(ctx, ra_frame)) < 0) |
||
1484 | av_log(ctx->avctx, AV_LOG_WARNING, |
||
1485 | "Reading frame data failed. Skipping RA unit.\n"); |
||
1486 | |||
1487 | ctx->frame_id++; |
||
1488 | |||
1489 | /* get output buffer */ |
||
1490 | frame->nb_samples = ctx->cur_frame_length; |
||
1491 | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) |
||
1492 | return ret; |
||
1493 | |||
1494 | // transform decoded frame into output format |
||
1495 | #define INTERLEAVE_OUTPUT(bps) \ |
||
1496 | { \ |
||
1497 | int##bps##_t *dest = (int##bps##_t*)frame->data[0]; \ |
||
1498 | shift = bps - ctx->avctx->bits_per_raw_sample; \ |
||
1499 | if (!ctx->cs_switch) { \ |
||
1500 | for (sample = 0; sample < ctx->cur_frame_length; sample++) \ |
||
1501 | for (c = 0; c < avctx->channels; c++) \ |
||
1502 | *dest++ = ctx->raw_samples[c][sample] << shift; \ |
||
1503 | } else { \ |
||
1504 | for (sample = 0; sample < ctx->cur_frame_length; sample++) \ |
||
1505 | for (c = 0; c < avctx->channels; c++) \ |
||
1506 | *dest++ = ctx->raw_samples[sconf->chan_pos[c]][sample] << shift; \ |
||
1507 | } \ |
||
1508 | } |
||
1509 | |||
1510 | if (ctx->avctx->bits_per_raw_sample <= 16) { |
||
1511 | INTERLEAVE_OUTPUT(16) |
||
1512 | } else { |
||
1513 | INTERLEAVE_OUTPUT(32) |
||
1514 | } |
||
1515 | |||
1516 | // update CRC |
||
1517 | if (sconf->crc_enabled && (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL))) { |
||
1518 | int swap = HAVE_BIGENDIAN != sconf->msb_first; |
||
1519 | |||
1520 | if (ctx->avctx->bits_per_raw_sample == 24) { |
||
1521 | int32_t *src = (int32_t *)frame->data[0]; |
||
1522 | |||
1523 | for (sample = 0; |
||
1524 | sample < ctx->cur_frame_length * avctx->channels; |
||
1525 | sample++) { |
||
1526 | int32_t v; |
||
1527 | |||
1528 | if (swap) |
||
1529 | v = av_bswap32(src[sample]); |
||
1530 | else |
||
1531 | v = src[sample]; |
||
1532 | if (!HAVE_BIGENDIAN) |
||
1533 | v >>= 8; |
||
1534 | |||
1535 | ctx->crc = av_crc(ctx->crc_table, ctx->crc, (uint8_t*)(&v), 3); |
||
1536 | } |
||
1537 | } else { |
||
1538 | uint8_t *crc_source; |
||
1539 | |||
1540 | if (swap) { |
||
1541 | if (ctx->avctx->bits_per_raw_sample <= 16) { |
||
1542 | int16_t *src = (int16_t*) frame->data[0]; |
||
1543 | int16_t *dest = (int16_t*) ctx->crc_buffer; |
||
1544 | for (sample = 0; |
||
1545 | sample < ctx->cur_frame_length * avctx->channels; |
||
1546 | sample++) |
||
1547 | *dest++ = av_bswap16(src[sample]); |
||
1548 | } else { |
||
1549 | ctx->dsp.bswap_buf((uint32_t*)ctx->crc_buffer, |
||
1550 | (uint32_t *)frame->data[0], |
||
1551 | ctx->cur_frame_length * avctx->channels); |
||
1552 | } |
||
1553 | crc_source = ctx->crc_buffer; |
||
1554 | } else { |
||
1555 | crc_source = frame->data[0]; |
||
1556 | } |
||
1557 | |||
1558 | ctx->crc = av_crc(ctx->crc_table, ctx->crc, crc_source, |
||
1559 | ctx->cur_frame_length * avctx->channels * |
||
1560 | av_get_bytes_per_sample(avctx->sample_fmt)); |
||
1561 | } |
||
1562 | |||
1563 | |||
1564 | // check CRC sums if this is the last frame |
||
1565 | if (ctx->cur_frame_length != sconf->frame_length && |
||
1566 | ctx->crc_org != ctx->crc) { |
||
1567 | av_log(avctx, AV_LOG_ERROR, "CRC error.\n"); |
||
1568 | } |
||
1569 | } |
||
1570 | |||
1571 | *got_frame_ptr = 1; |
||
1572 | |||
1573 | bytes_read = invalid_frame ? buffer_size : |
||
1574 | (get_bits_count(&ctx->gb) + 7) >> 3; |
||
1575 | |||
1576 | return bytes_read; |
||
1577 | } |
||
1578 | |||
1579 | |||
1580 | /** Uninitialize the ALS decoder. |
||
1581 | */ |
||
1582 | static av_cold int decode_end(AVCodecContext *avctx) |
||
1583 | { |
||
1584 | ALSDecContext *ctx = avctx->priv_data; |
||
1585 | |||
1586 | av_freep(&ctx->sconf.chan_pos); |
||
1587 | |||
1588 | ff_bgmc_end(&ctx->bgmc_lut, &ctx->bgmc_lut_status); |
||
1589 | |||
1590 | av_freep(&ctx->const_block); |
||
1591 | av_freep(&ctx->shift_lsbs); |
||
1592 | av_freep(&ctx->opt_order); |
||
1593 | av_freep(&ctx->store_prev_samples); |
||
1594 | av_freep(&ctx->use_ltp); |
||
1595 | av_freep(&ctx->ltp_lag); |
||
1596 | av_freep(&ctx->ltp_gain); |
||
1597 | av_freep(&ctx->ltp_gain_buffer); |
||
1598 | av_freep(&ctx->quant_cof); |
||
1599 | av_freep(&ctx->lpc_cof); |
||
1600 | av_freep(&ctx->quant_cof_buffer); |
||
1601 | av_freep(&ctx->lpc_cof_buffer); |
||
1602 | av_freep(&ctx->lpc_cof_reversed_buffer); |
||
1603 | av_freep(&ctx->prev_raw_samples); |
||
1604 | av_freep(&ctx->raw_samples); |
||
1605 | av_freep(&ctx->raw_buffer); |
||
1606 | av_freep(&ctx->chan_data); |
||
1607 | av_freep(&ctx->chan_data_buffer); |
||
1608 | av_freep(&ctx->reverted_channels); |
||
1609 | av_freep(&ctx->crc_buffer); |
||
1610 | |||
1611 | return 0; |
||
1612 | } |
||
1613 | |||
1614 | |||
1615 | /** Initialize the ALS decoder. |
||
1616 | */ |
||
1617 | static av_cold int decode_init(AVCodecContext *avctx) |
||
1618 | { |
||
1619 | unsigned int c; |
||
1620 | unsigned int channel_size; |
||
1621 | int num_buffers, ret; |
||
1622 | ALSDecContext *ctx = avctx->priv_data; |
||
1623 | ALSSpecificConfig *sconf = &ctx->sconf; |
||
1624 | ctx->avctx = avctx; |
||
1625 | |||
1626 | if (!avctx->extradata) { |
||
1627 | av_log(avctx, AV_LOG_ERROR, "Missing required ALS extradata.\n"); |
||
1628 | return AVERROR_INVALIDDATA; |
||
1629 | } |
||
1630 | |||
1631 | if ((ret = read_specific_config(ctx)) < 0) { |
||
1632 | av_log(avctx, AV_LOG_ERROR, "Reading ALSSpecificConfig failed.\n"); |
||
1633 | goto fail; |
||
1634 | } |
||
1635 | |||
1636 | if ((ret = check_specific_config(ctx)) < 0) { |
||
1637 | goto fail; |
||
1638 | } |
||
1639 | |||
1640 | if (sconf->bgmc) { |
||
1641 | ret = ff_bgmc_init(avctx, &ctx->bgmc_lut, &ctx->bgmc_lut_status); |
||
1642 | if (ret < 0) |
||
1643 | goto fail; |
||
1644 | } |
||
1645 | if (sconf->floating) { |
||
1646 | avctx->sample_fmt = AV_SAMPLE_FMT_FLT; |
||
1647 | avctx->bits_per_raw_sample = 32; |
||
1648 | } else { |
||
1649 | avctx->sample_fmt = sconf->resolution > 1 |
||
1650 | ? AV_SAMPLE_FMT_S32 : AV_SAMPLE_FMT_S16; |
||
1651 | avctx->bits_per_raw_sample = (sconf->resolution + 1) * 8; |
||
1652 | } |
||
1653 | |||
1654 | // set maximum Rice parameter for progressive decoding based on resolution |
||
1655 | // This is not specified in 14496-3 but actually done by the reference |
||
1656 | // codec RM22 revision 2. |
||
1657 | ctx->s_max = sconf->resolution > 1 ? 31 : 15; |
||
1658 | |||
1659 | // set lag value for long-term prediction |
||
1660 | ctx->ltp_lag_length = 8 + (avctx->sample_rate >= 96000) + |
||
1661 | (avctx->sample_rate >= 192000); |
||
1662 | |||
1663 | // allocate quantized parcor coefficient buffer |
||
1664 | num_buffers = sconf->mc_coding ? avctx->channels : 1; |
||
1665 | |||
1666 | ctx->quant_cof = av_malloc(sizeof(*ctx->quant_cof) * num_buffers); |
||
1667 | ctx->lpc_cof = av_malloc(sizeof(*ctx->lpc_cof) * num_buffers); |
||
1668 | ctx->quant_cof_buffer = av_malloc(sizeof(*ctx->quant_cof_buffer) * |
||
1669 | num_buffers * sconf->max_order); |
||
1670 | ctx->lpc_cof_buffer = av_malloc(sizeof(*ctx->lpc_cof_buffer) * |
||
1671 | num_buffers * sconf->max_order); |
||
1672 | ctx->lpc_cof_reversed_buffer = av_malloc(sizeof(*ctx->lpc_cof_buffer) * |
||
1673 | sconf->max_order); |
||
1674 | |||
1675 | if (!ctx->quant_cof || !ctx->lpc_cof || |
||
1676 | !ctx->quant_cof_buffer || !ctx->lpc_cof_buffer || |
||
1677 | !ctx->lpc_cof_reversed_buffer) { |
||
1678 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); |
||
1679 | ret = AVERROR(ENOMEM); |
||
1680 | goto fail; |
||
1681 | } |
||
1682 | |||
1683 | // assign quantized parcor coefficient buffers |
||
1684 | for (c = 0; c < num_buffers; c++) { |
||
1685 | ctx->quant_cof[c] = ctx->quant_cof_buffer + c * sconf->max_order; |
||
1686 | ctx->lpc_cof[c] = ctx->lpc_cof_buffer + c * sconf->max_order; |
||
1687 | } |
||
1688 | |||
1689 | // allocate and assign lag and gain data buffer for ltp mode |
||
1690 | ctx->const_block = av_malloc (sizeof(*ctx->const_block) * num_buffers); |
||
1691 | ctx->shift_lsbs = av_malloc (sizeof(*ctx->shift_lsbs) * num_buffers); |
||
1692 | ctx->opt_order = av_malloc (sizeof(*ctx->opt_order) * num_buffers); |
||
1693 | ctx->store_prev_samples = av_malloc(sizeof(*ctx->store_prev_samples) * num_buffers); |
||
1694 | ctx->use_ltp = av_mallocz(sizeof(*ctx->use_ltp) * num_buffers); |
||
1695 | ctx->ltp_lag = av_malloc (sizeof(*ctx->ltp_lag) * num_buffers); |
||
1696 | ctx->ltp_gain = av_malloc (sizeof(*ctx->ltp_gain) * num_buffers); |
||
1697 | ctx->ltp_gain_buffer = av_malloc (sizeof(*ctx->ltp_gain_buffer) * |
||
1698 | num_buffers * 5); |
||
1699 | |||
1700 | if (!ctx->const_block || !ctx->shift_lsbs || |
||
1701 | !ctx->opt_order || !ctx->store_prev_samples || |
||
1702 | !ctx->use_ltp || !ctx->ltp_lag || |
||
1703 | !ctx->ltp_gain || !ctx->ltp_gain_buffer) { |
||
1704 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); |
||
1705 | ret = AVERROR(ENOMEM); |
||
1706 | goto fail; |
||
1707 | } |
||
1708 | |||
1709 | for (c = 0; c < num_buffers; c++) |
||
1710 | ctx->ltp_gain[c] = ctx->ltp_gain_buffer + c * 5; |
||
1711 | |||
1712 | // allocate and assign channel data buffer for mcc mode |
||
1713 | if (sconf->mc_coding) { |
||
1714 | ctx->chan_data_buffer = av_malloc(sizeof(*ctx->chan_data_buffer) * |
||
1715 | num_buffers * num_buffers); |
||
1716 | ctx->chan_data = av_malloc(sizeof(*ctx->chan_data) * |
||
1717 | num_buffers); |
||
1718 | ctx->reverted_channels = av_malloc(sizeof(*ctx->reverted_channels) * |
||
1719 | num_buffers); |
||
1720 | |||
1721 | if (!ctx->chan_data_buffer || !ctx->chan_data || !ctx->reverted_channels) { |
||
1722 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); |
||
1723 | ret = AVERROR(ENOMEM); |
||
1724 | goto fail; |
||
1725 | } |
||
1726 | |||
1727 | for (c = 0; c < num_buffers; c++) |
||
1728 | ctx->chan_data[c] = ctx->chan_data_buffer + c * num_buffers; |
||
1729 | } else { |
||
1730 | ctx->chan_data = NULL; |
||
1731 | ctx->chan_data_buffer = NULL; |
||
1732 | ctx->reverted_channels = NULL; |
||
1733 | } |
||
1734 | |||
1735 | channel_size = sconf->frame_length + sconf->max_order; |
||
1736 | |||
1737 | ctx->prev_raw_samples = av_malloc (sizeof(*ctx->prev_raw_samples) * sconf->max_order); |
||
1738 | ctx->raw_buffer = av_mallocz(sizeof(*ctx-> raw_buffer) * avctx->channels * channel_size); |
||
1739 | ctx->raw_samples = av_malloc (sizeof(*ctx-> raw_samples) * avctx->channels); |
||
1740 | |||
1741 | // allocate previous raw sample buffer |
||
1742 | if (!ctx->prev_raw_samples || !ctx->raw_buffer|| !ctx->raw_samples) { |
||
1743 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); |
||
1744 | ret = AVERROR(ENOMEM); |
||
1745 | goto fail; |
||
1746 | } |
||
1747 | |||
1748 | // assign raw samples buffers |
||
1749 | ctx->raw_samples[0] = ctx->raw_buffer + sconf->max_order; |
||
1750 | for (c = 1; c < avctx->channels; c++) |
||
1751 | ctx->raw_samples[c] = ctx->raw_samples[c - 1] + channel_size; |
||
1752 | |||
1753 | // allocate crc buffer |
||
1754 | if (HAVE_BIGENDIAN != sconf->msb_first && sconf->crc_enabled && |
||
1755 | (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL))) { |
||
1756 | ctx->crc_buffer = av_malloc(sizeof(*ctx->crc_buffer) * |
||
1757 | ctx->cur_frame_length * |
||
1758 | avctx->channels * |
||
1759 | av_get_bytes_per_sample(avctx->sample_fmt)); |
||
1760 | if (!ctx->crc_buffer) { |
||
1761 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); |
||
1762 | ret = AVERROR(ENOMEM); |
||
1763 | goto fail; |
||
1764 | } |
||
1765 | } |
||
1766 | |||
1767 | ff_dsputil_init(&ctx->dsp, avctx); |
||
1768 | |||
1769 | return 0; |
||
1770 | |||
1771 | fail: |
||
1772 | decode_end(avctx); |
||
1773 | return ret; |
||
1774 | } |
||
1775 | |||
1776 | |||
1777 | /** Flush (reset) the frame ID after seeking. |
||
1778 | */ |
||
1779 | static av_cold void flush(AVCodecContext *avctx) |
||
1780 | { |
||
1781 | ALSDecContext *ctx = avctx->priv_data; |
||
1782 | |||
1783 | ctx->frame_id = 0; |
||
1784 | } |
||
1785 | |||
1786 | |||
1787 | AVCodec ff_als_decoder = { |
||
1788 | .name = "als", |
||
1789 | .long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Audio Lossless Coding (ALS)"), |
||
1790 | .type = AVMEDIA_TYPE_AUDIO, |
||
1791 | .id = AV_CODEC_ID_MP4ALS, |
||
1792 | .priv_data_size = sizeof(ALSDecContext), |
||
1793 | .init = decode_init, |
||
1794 | .close = decode_end, |
||
1795 | .decode = decode_frame, |
||
1796 | .flush = flush, |
||
1797 | .capabilities = CODEC_CAP_SUBFRAMES | CODEC_CAP_DR1, |
||
1798 | };>>>>>>>>=>>=>><>>>><>>>>>>>>>>>>>>>>>>>>><>>><>>>>>>>>>>>>>=><=>>>>><>>>>>><>>>><>>>>=><=>><>><>><>>>>><>><>><>><>><>><>>>>>>>>>>>><>=>=>>>=><=>><>><>><>><>>=><=>><>>>>><>>>>><>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><>><> |