Mercurial > libavcodec.hg
annotate alsdec.c @ 10676:8ee37f5571dc libavcodec
Commit functions used by both AMRNB and SIPR
author | vitor |
---|---|
date | Sat, 12 Dec 2009 13:58:07 +0000 |
parents | 95f3daa991a2 |
children | 997692df50c1 |
rev | line source |
---|---|
10522 | 1 /* |
2 * MPEG-4 ALS decoder | |
3 * Copyright (c) 2009 Thilo Borgmann <thilo.borgmann _at_ googlemail.com> | |
4 * | |
5 * This file is part of FFmpeg. | |
6 * | |
7 * FFmpeg is free software; you can redistribute it and/or | |
8 * modify it under the terms of the GNU Lesser General Public | |
9 * License as published by the Free Software Foundation; either | |
10 * version 2.1 of the License, or (at your option) any later version. | |
11 * | |
12 * FFmpeg is distributed in the hope that it will be useful, | |
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
15 * Lesser General Public License for more details. | |
16 * | |
17 * You should have received a copy of the GNU Lesser General Public | |
18 * License along with FFmpeg; if not, write to the Free Software | |
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
20 */ | |
21 | |
22 /** | |
23 * @file libavcodec/alsdec.c | |
24 * MPEG-4 ALS decoder | |
25 * @author Thilo Borgmann <thilo.borgmann _at_ googlemail.com> | |
26 */ | |
27 | |
28 | |
29 //#define DEBUG | |
30 | |
31 | |
32 #include "avcodec.h" | |
33 #include "get_bits.h" | |
34 #include "unary.h" | |
35 #include "mpeg4audio.h" | |
36 #include "bytestream.h" | |
37 | |
10531
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
38 #include <stdint.h> |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
39 |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
40 /** Rice parameters and corresponding index offsets for decoding the |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
41 * indices of scaled PARCOR values. The table choosen is set globally |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
42 * by the encoder and stored in ALSSpecificConfig. |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
43 */ |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
44 static const int8_t parcor_rice_table[3][20][2] = { |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
45 { {-52, 4}, {-29, 5}, {-31, 4}, { 19, 4}, {-16, 4}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
46 { 12, 3}, { -7, 3}, { 9, 3}, { -5, 3}, { 6, 3}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
47 { -4, 3}, { 3, 3}, { -3, 2}, { 3, 2}, { -2, 2}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
48 { 3, 2}, { -1, 2}, { 2, 2}, { -1, 2}, { 2, 2} }, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
49 { {-58, 3}, {-42, 4}, {-46, 4}, { 37, 5}, {-36, 4}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
50 { 29, 4}, {-29, 4}, { 25, 4}, {-23, 4}, { 20, 4}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
51 {-17, 4}, { 16, 4}, {-12, 4}, { 12, 3}, {-10, 4}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
52 { 7, 3}, { -4, 4}, { 3, 3}, { -1, 3}, { 1, 3} }, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
53 { {-59, 3}, {-45, 5}, {-50, 4}, { 38, 4}, {-39, 4}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
54 { 32, 4}, {-30, 4}, { 25, 3}, {-23, 3}, { 20, 3}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
55 {-20, 3}, { 16, 3}, {-13, 3}, { 10, 3}, { -7, 3}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
56 { 3, 3}, { 0, 3}, { -1, 3}, { 2, 3}, { -1, 2} } |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
57 }; |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
58 |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
59 |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
60 /** Scaled PARCOR values used for the first two PARCOR coefficients. |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
61 * To be indexed by the Rice coded indices. |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
62 * Generated by: parcor_scaled_values[i] = 32 + ((i * (i+1)) << 7) - (1 << 20) |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
63 * Actual values are divided by 32 in order to be stored in 16 bits. |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
64 */ |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
65 static const int16_t parcor_scaled_values[] = { |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
66 -1048544 / 32, -1048288 / 32, -1047776 / 32, -1047008 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
67 -1045984 / 32, -1044704 / 32, -1043168 / 32, -1041376 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
68 -1039328 / 32, -1037024 / 32, -1034464 / 32, -1031648 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
69 -1028576 / 32, -1025248 / 32, -1021664 / 32, -1017824 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
70 -1013728 / 32, -1009376 / 32, -1004768 / 32, -999904 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
71 -994784 / 32, -989408 / 32, -983776 / 32, -977888 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
72 -971744 / 32, -965344 / 32, -958688 / 32, -951776 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
73 -944608 / 32, -937184 / 32, -929504 / 32, -921568 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
74 -913376 / 32, -904928 / 32, -896224 / 32, -887264 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
75 -878048 / 32, -868576 / 32, -858848 / 32, -848864 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
76 -838624 / 32, -828128 / 32, -817376 / 32, -806368 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
77 -795104 / 32, -783584 / 32, -771808 / 32, -759776 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
78 -747488 / 32, -734944 / 32, -722144 / 32, -709088 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
79 -695776 / 32, -682208 / 32, -668384 / 32, -654304 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
80 -639968 / 32, -625376 / 32, -610528 / 32, -595424 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
81 -580064 / 32, -564448 / 32, -548576 / 32, -532448 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
82 -516064 / 32, -499424 / 32, -482528 / 32, -465376 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
83 -447968 / 32, -430304 / 32, -412384 / 32, -394208 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
84 -375776 / 32, -357088 / 32, -338144 / 32, -318944 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
85 -299488 / 32, -279776 / 32, -259808 / 32, -239584 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
86 -219104 / 32, -198368 / 32, -177376 / 32, -156128 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
87 -134624 / 32, -112864 / 32, -90848 / 32, -68576 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
88 -46048 / 32, -23264 / 32, -224 / 32, 23072 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
89 46624 / 32, 70432 / 32, 94496 / 32, 118816 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
90 143392 / 32, 168224 / 32, 193312 / 32, 218656 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
91 244256 / 32, 270112 / 32, 296224 / 32, 322592 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
92 349216 / 32, 376096 / 32, 403232 / 32, 430624 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
93 458272 / 32, 486176 / 32, 514336 / 32, 542752 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
94 571424 / 32, 600352 / 32, 629536 / 32, 658976 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
95 688672 / 32, 718624 / 32, 748832 / 32, 779296 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
96 810016 / 32, 840992 / 32, 872224 / 32, 903712 / 32, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
97 935456 / 32, 967456 / 32, 999712 / 32, 1032224 / 32 |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
98 }; |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
99 |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
100 |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
101 /** Gain values of p(0) for long-term prediction. |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
102 * To be indexed by the Rice coded indices. |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
103 */ |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
104 static const uint8_t ltp_gain_values [4][4] = { |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
105 { 0, 8, 16, 24}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
106 {32, 40, 48, 56}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
107 {64, 70, 76, 82}, |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
108 {88, 92, 96, 100} |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
109 }; |
142645a57180
Merge data tables from als_data.h with the decoder source to reduce
thilo.borgmann
parents:
10530
diff
changeset
|
110 |
10522 | 111 |
112 enum RA_Flag { | |
113 RA_FLAG_NONE, | |
114 RA_FLAG_FRAMES, | |
115 RA_FLAG_HEADER | |
116 }; | |
117 | |
118 | |
119 typedef struct { | |
120 uint32_t samples; ///< number of samples, 0xFFFFFFFF if unknown | |
121 int resolution; ///< 000 = 8-bit; 001 = 16-bit; 010 = 24-bit; 011 = 32-bit | |
122 int floating; ///< 1 = IEEE 32-bit floating-point, 0 = integer | |
123 int frame_length; ///< frame length for each frame (last frame may differ) | |
124 int ra_distance; ///< distance between RA frames (in frames, 0...255) | |
125 enum RA_Flag ra_flag; ///< indicates where the size of ra units is stored | |
126 int adapt_order; ///< adaptive order: 1 = on, 0 = off | |
127 int coef_table; ///< table index of Rice code parameters | |
128 int long_term_prediction; ///< long term prediction (LTP): 1 = on, 0 = off | |
129 int max_order; ///< maximum prediction order (0..1023) | |
130 int block_switching; ///< number of block switching levels | |
131 int bgmc; ///< "Block Gilbert-Moore Code": 1 = on, 0 = off (Rice coding only) | |
132 int sb_part; ///< sub-block partition | |
133 int joint_stereo; ///< joint stereo: 1 = on, 0 = off | |
134 int mc_coding; ///< extended inter-channel coding (multi channel coding): 1 = on, 0 = off | |
135 int chan_config; ///< indicates that a chan_config_info field is present | |
136 int chan_sort; ///< channel rearrangement: 1 = on, 0 = off | |
137 int rlslms; ///< use "Recursive Least Square-Least Mean Square" predictor: 1 = on, 0 = off | |
138 int chan_config_info; ///< mapping of channels to loudspeaker locations. Unused until setting channel configuration is implemented. | |
139 int *chan_pos; ///< original channel positions | |
140 uint32_t header_size; ///< header size of original audio file in bytes, provided for debugging | |
141 uint32_t trailer_size; ///< trailer size of original audio file in bytes, provided for debugging | |
142 } ALSSpecificConfig; | |
143 | |
144 | |
145 typedef struct { | |
146 AVCodecContext *avctx; | |
147 ALSSpecificConfig sconf; | |
148 GetBitContext gb; | |
149 unsigned int cur_frame_length; ///< length of the current frame to decode | |
150 unsigned int frame_id; ///< the frame ID / number of the current frame | |
151 unsigned int js_switch; ///< if true, joint-stereo decoding is enforced | |
152 unsigned int num_blocks; ///< number of blocks used in the current frame | |
10530
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
153 int ltp_lag_length; ///< number of bits used for ltp lag value |
10522 | 154 int32_t *quant_cof; ///< quantized parcor coefficients |
155 int32_t *lpc_cof; ///< coefficients of the direct form prediction filter | |
156 int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block | |
157 int32_t **raw_samples; ///< decoded raw samples for each channel | |
158 int32_t *raw_buffer; ///< contains all decoded raw samples including carryover samples | |
159 } ALSDecContext; | |
160 | |
161 | |
162 static av_cold void dprint_specific_config(ALSDecContext *ctx) | |
163 { | |
164 #ifdef DEBUG | |
165 AVCodecContext *avctx = ctx->avctx; | |
166 ALSSpecificConfig *sconf = &ctx->sconf; | |
167 | |
168 dprintf(avctx, "resolution = %i\n", sconf->resolution); | |
169 dprintf(avctx, "floating = %i\n", sconf->floating); | |
170 dprintf(avctx, "frame_length = %i\n", sconf->frame_length); | |
171 dprintf(avctx, "ra_distance = %i\n", sconf->ra_distance); | |
172 dprintf(avctx, "ra_flag = %i\n", sconf->ra_flag); | |
173 dprintf(avctx, "adapt_order = %i\n", sconf->adapt_order); | |
174 dprintf(avctx, "coef_table = %i\n", sconf->coef_table); | |
175 dprintf(avctx, "long_term_prediction = %i\n", sconf->long_term_prediction); | |
176 dprintf(avctx, "max_order = %i\n", sconf->max_order); | |
177 dprintf(avctx, "block_switching = %i\n", sconf->block_switching); | |
178 dprintf(avctx, "bgmc = %i\n", sconf->bgmc); | |
179 dprintf(avctx, "sb_part = %i\n", sconf->sb_part); | |
180 dprintf(avctx, "joint_stereo = %i\n", sconf->joint_stereo); | |
181 dprintf(avctx, "mc_coding = %i\n", sconf->mc_coding); | |
182 dprintf(avctx, "chan_config = %i\n", sconf->chan_config); | |
183 dprintf(avctx, "chan_sort = %i\n", sconf->chan_sort); | |
184 dprintf(avctx, "RLSLMS = %i\n", sconf->rlslms); | |
185 dprintf(avctx, "chan_config_info = %i\n", sconf->chan_config_info); | |
186 dprintf(avctx, "header_size = %i\n", sconf->header_size); | |
187 dprintf(avctx, "trailer_size = %i\n", sconf->trailer_size); | |
188 #endif | |
189 } | |
190 | |
191 | |
192 /** Reads an ALSSpecificConfig from a buffer into the output struct. | |
193 */ | |
194 static av_cold int read_specific_config(ALSDecContext *ctx) | |
195 { | |
196 GetBitContext gb; | |
197 uint64_t ht_size; | |
198 int i, config_offset, crc_enabled; | |
199 MPEG4AudioConfig m4ac; | |
200 ALSSpecificConfig *sconf = &ctx->sconf; | |
201 AVCodecContext *avctx = ctx->avctx; | |
202 uint32_t als_id; | |
203 | |
204 init_get_bits(&gb, avctx->extradata, avctx->extradata_size * 8); | |
205 | |
206 config_offset = ff_mpeg4audio_get_config(&m4ac, avctx->extradata, | |
207 avctx->extradata_size); | |
208 | |
209 if (config_offset < 0) | |
210 return -1; | |
211 | |
212 skip_bits_long(&gb, config_offset); | |
213 | |
214 if (get_bits_left(&gb) < (30 << 3)) | |
215 return -1; | |
216 | |
217 // read the fixed items | |
218 als_id = get_bits_long(&gb, 32); | |
219 avctx->sample_rate = m4ac.sample_rate; | |
220 skip_bits_long(&gb, 32); // sample rate already known | |
221 sconf->samples = get_bits_long(&gb, 32); | |
222 avctx->channels = m4ac.channels; | |
223 skip_bits(&gb, 16); // number of channels already knwon | |
224 skip_bits(&gb, 3); // skip file_type | |
225 sconf->resolution = get_bits(&gb, 3); | |
226 sconf->floating = get_bits1(&gb); | |
227 skip_bits1(&gb); // skip msb_first | |
228 sconf->frame_length = get_bits(&gb, 16) + 1; | |
229 sconf->ra_distance = get_bits(&gb, 8); | |
230 sconf->ra_flag = get_bits(&gb, 2); | |
231 sconf->adapt_order = get_bits1(&gb); | |
232 sconf->coef_table = get_bits(&gb, 2); | |
233 sconf->long_term_prediction = get_bits1(&gb); | |
234 sconf->max_order = get_bits(&gb, 10); | |
235 sconf->block_switching = get_bits(&gb, 2); | |
236 sconf->bgmc = get_bits1(&gb); | |
237 sconf->sb_part = get_bits1(&gb); | |
238 sconf->joint_stereo = get_bits1(&gb); | |
239 sconf->mc_coding = get_bits1(&gb); | |
240 sconf->chan_config = get_bits1(&gb); | |
241 sconf->chan_sort = get_bits1(&gb); | |
242 crc_enabled = get_bits1(&gb); | |
243 sconf->rlslms = get_bits1(&gb); | |
244 skip_bits(&gb, 5); // skip 5 reserved bits | |
245 skip_bits1(&gb); // skip aux_data_enabled | |
246 | |
247 | |
248 // check for ALSSpecificConfig struct | |
249 if (als_id != MKBETAG('A','L','S','\0')) | |
250 return -1; | |
251 | |
252 ctx->cur_frame_length = sconf->frame_length; | |
253 | |
254 // allocate quantized parcor coefficient buffer | |
255 if (!(ctx->quant_cof = av_malloc(sizeof(*ctx->quant_cof) * sconf->max_order)) || | |
256 !(ctx->lpc_cof = av_malloc(sizeof(*ctx->lpc_cof) * sconf->max_order))) { | |
257 av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
258 return AVERROR(ENOMEM); | |
259 } | |
260 | |
261 // read channel config | |
262 if (sconf->chan_config) | |
263 sconf->chan_config_info = get_bits(&gb, 16); | |
264 // TODO: use this to set avctx->channel_layout | |
265 | |
266 | |
267 // read channel sorting | |
268 if (sconf->chan_sort && avctx->channels > 1) { | |
269 int chan_pos_bits = av_ceil_log2(avctx->channels); | |
270 int bits_needed = avctx->channels * chan_pos_bits + 7; | |
271 if (get_bits_left(&gb) < bits_needed) | |
272 return -1; | |
273 | |
274 if (!(sconf->chan_pos = av_malloc(avctx->channels * sizeof(*sconf->chan_pos)))) | |
275 return AVERROR(ENOMEM); | |
276 | |
277 for (i = 0; i < avctx->channels; i++) | |
278 sconf->chan_pos[i] = get_bits(&gb, chan_pos_bits); | |
279 | |
280 align_get_bits(&gb); | |
281 // TODO: use this to actually do channel sorting | |
282 } else { | |
283 sconf->chan_sort = 0; | |
284 } | |
285 | |
286 | |
287 // read fixed header and trailer sizes, | |
288 // if size = 0xFFFFFFFF then there is no data field! | |
289 if (get_bits_left(&gb) < 64) | |
290 return -1; | |
291 | |
292 sconf->header_size = get_bits_long(&gb, 32); | |
293 sconf->trailer_size = get_bits_long(&gb, 32); | |
294 if (sconf->header_size == 0xFFFFFFFF) | |
295 sconf->header_size = 0; | |
296 if (sconf->trailer_size == 0xFFFFFFFF) | |
297 sconf->trailer_size = 0; | |
298 | |
299 ht_size = ((int64_t)(sconf->header_size) + (int64_t)(sconf->trailer_size)) << 3; | |
300 | |
301 | |
302 // skip the header and trailer data | |
303 if (get_bits_left(&gb) < ht_size) | |
304 return -1; | |
305 | |
306 if (ht_size > INT32_MAX) | |
307 return -1; | |
308 | |
309 skip_bits_long(&gb, ht_size); | |
310 | |
311 | |
312 // skip the crc data | |
313 if (crc_enabled) { | |
314 if (get_bits_left(&gb) < 32) | |
315 return -1; | |
316 | |
317 skip_bits_long(&gb, 32); | |
318 } | |
319 | |
320 | |
321 // no need to read the rest of ALSSpecificConfig (ra_unit_size & aux data) | |
322 | |
323 dprint_specific_config(ctx); | |
324 | |
325 return 0; | |
326 } | |
327 | |
328 | |
329 /** Checks the ALSSpecificConfig for unsupported features. | |
330 */ | |
331 static int check_specific_config(ALSDecContext *ctx) | |
332 { | |
333 ALSSpecificConfig *sconf = &ctx->sconf; | |
334 int error = 0; | |
335 | |
336 // report unsupported feature and set error value | |
337 #define MISSING_ERR(cond, str, errval) \ | |
338 { \ | |
339 if (cond) { \ | |
340 av_log_missing_feature(ctx->avctx, str, 0); \ | |
341 error = errval; \ | |
342 } \ | |
343 } | |
344 | |
345 MISSING_ERR(sconf->floating, "Floating point decoding", -1); | |
346 MISSING_ERR(sconf->bgmc, "BGMC entropy decoding", -1); | |
347 MISSING_ERR(sconf->mc_coding, "Multi-channel correlation", -1); | |
348 MISSING_ERR(sconf->rlslms, "Adaptive RLS-LMS prediction", -1); | |
349 MISSING_ERR(sconf->chan_sort, "Channel sorting", 0); | |
350 | |
351 return error; | |
352 } | |
353 | |
354 | |
355 /** Parses the bs_info field to extract the block partitioning used in | |
356 * block switching mode, refer to ISO/IEC 14496-3, section 11.6.2. | |
357 */ | |
358 static void parse_bs_info(const uint32_t bs_info, unsigned int n, | |
359 unsigned int div, unsigned int **div_blocks, | |
360 unsigned int *num_blocks) | |
361 { | |
362 if (n < 31 && ((bs_info << n) & 0x40000000)) { | |
363 // if the level is valid and the investigated bit n is set | |
364 // then recursively check both children at bits (2n+1) and (2n+2) | |
365 n *= 2; | |
366 div += 1; | |
367 parse_bs_info(bs_info, n + 1, div, div_blocks, num_blocks); | |
368 parse_bs_info(bs_info, n + 2, div, div_blocks, num_blocks); | |
369 } else { | |
370 // else the bit is not set or the last level has been reached | |
371 // (bit implicitly not set) | |
372 **div_blocks = div; | |
373 (*div_blocks)++; | |
374 (*num_blocks)++; | |
375 } | |
376 } | |
377 | |
378 | |
379 /** Reads and decodes a Rice codeword. | |
380 */ | |
381 static int32_t decode_rice(GetBitContext *gb, unsigned int k) | |
382 { | |
10535
95f3daa991a2
Use get_bits_left() instead of size_in_bits - get_bits_count().
rbultje
parents:
10531
diff
changeset
|
383 int max = get_bits_left(gb) - k; |
10522 | 384 int q = get_unary(gb, 0, max); |
385 int r = k ? get_bits1(gb) : !(q & 1); | |
386 | |
387 if (k > 1) { | |
388 q <<= (k - 1); | |
389 q += get_bits_long(gb, k - 1); | |
390 } else if (!k) { | |
391 q >>= 1; | |
392 } | |
393 return r ? q : ~q; | |
394 } | |
395 | |
396 | |
397 /** Converts PARCOR coefficient k to direct filter coefficient. | |
398 */ | |
399 static void parcor_to_lpc(unsigned int k, const int32_t *par, int32_t *cof) | |
400 { | |
401 int i, j; | |
402 | |
403 for (i = 0, j = k - 1; i < j; i++, j--) { | |
404 int tmp1 = ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20); | |
405 cof[j] += ((MUL64(par[k], cof[i]) + (1 << 19)) >> 20); | |
406 cof[i] += tmp1; | |
407 } | |
408 if (i == j) | |
409 cof[i] += ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20); | |
410 | |
411 cof[k] = par[k]; | |
412 } | |
413 | |
414 | |
415 /** Reads block switching field if necessary and sets actual block sizes. | |
416 * Also assures that the block sizes of the last frame correspond to the | |
417 * actual number of samples. | |
418 */ | |
419 static void get_block_sizes(ALSDecContext *ctx, unsigned int *div_blocks, | |
420 uint32_t *bs_info) | |
421 { | |
422 ALSSpecificConfig *sconf = &ctx->sconf; | |
423 GetBitContext *gb = &ctx->gb; | |
424 unsigned int *ptr_div_blocks = div_blocks; | |
425 unsigned int b; | |
426 | |
427 if (sconf->block_switching) { | |
428 unsigned int bs_info_len = 1 << (sconf->block_switching + 2); | |
429 *bs_info = get_bits_long(gb, bs_info_len); | |
430 *bs_info <<= (32 - bs_info_len); | |
431 } | |
432 | |
433 ctx->num_blocks = 0; | |
434 parse_bs_info(*bs_info, 0, 0, &ptr_div_blocks, &ctx->num_blocks); | |
435 | |
436 // The last frame may have an overdetermined block structure given in | |
437 // the bitstream. In that case the defined block structure would need | |
438 // more samples than available to be consistent. | |
439 // The block structure is actually used but the block sizes are adapted | |
440 // to fit the actual number of available samples. | |
441 // Example: 5 samples, 2nd level block sizes: 2 2 2 2. | |
442 // This results in the actual block sizes: 2 2 1 0. | |
443 // This is not specified in 14496-3 but actually done by the reference | |
444 // codec RM22 revision 2. | |
445 // This appears to happen in case of an odd number of samples in the last | |
446 // frame which is actually not allowed by the block length switching part | |
447 // of 14496-3. | |
448 // The ALS conformance files feature an odd number of samples in the last | |
449 // frame. | |
450 | |
451 for (b = 0; b < ctx->num_blocks; b++) | |
452 div_blocks[b] = ctx->sconf.frame_length >> div_blocks[b]; | |
453 | |
454 if (ctx->cur_frame_length != ctx->sconf.frame_length) { | |
455 unsigned int remaining = ctx->cur_frame_length; | |
456 | |
457 for (b = 0; b < ctx->num_blocks; b++) { | |
458 if (remaining < div_blocks[b]) { | |
459 div_blocks[b] = remaining; | |
460 ctx->num_blocks = b + 1; | |
461 break; | |
462 } | |
463 | |
464 remaining -= div_blocks[b]; | |
465 } | |
466 } | |
467 } | |
468 | |
469 | |
470 /** Reads the block data for a constant block | |
471 */ | |
472 static void read_const_block(ALSDecContext *ctx, int32_t *raw_samples, | |
473 unsigned int block_length, unsigned int *js_blocks) | |
474 { | |
475 ALSSpecificConfig *sconf = &ctx->sconf; | |
476 AVCodecContext *avctx = ctx->avctx; | |
477 GetBitContext *gb = &ctx->gb; | |
478 int32_t const_val = 0; | |
479 unsigned int const_block, k; | |
480 | |
481 const_block = get_bits1(gb); // 1 = constant value, 0 = zero block (silence) | |
482 *js_blocks = get_bits1(gb); | |
483 | |
484 // skip 5 reserved bits | |
485 skip_bits(gb, 5); | |
486 | |
487 if (const_block) { | |
488 unsigned int const_val_bits = sconf->floating ? 24 : avctx->bits_per_raw_sample; | |
489 const_val = get_sbits_long(gb, const_val_bits); | |
490 } | |
491 | |
492 // write raw samples into buffer | |
493 for (k = 0; k < block_length; k++) | |
494 raw_samples[k] = const_val; | |
495 } | |
496 | |
497 | |
498 /** Reads the block data for a non-constant block | |
499 */ | |
500 static int read_var_block(ALSDecContext *ctx, unsigned int ra_block, | |
501 int32_t *raw_samples, unsigned int block_length, | |
502 unsigned int *js_blocks, int32_t *raw_other, | |
503 unsigned int *shift_lsbs) | |
504 { | |
505 ALSSpecificConfig *sconf = &ctx->sconf; | |
506 AVCodecContext *avctx = ctx->avctx; | |
507 GetBitContext *gb = &ctx->gb; | |
508 unsigned int k; | |
509 unsigned int s[8]; | |
510 unsigned int sub_blocks, log2_sub_blocks, sb_length; | |
511 unsigned int opt_order = 1; | |
512 int32_t *quant_cof = ctx->quant_cof; | |
513 int32_t *lpc_cof = ctx->lpc_cof; | |
514 unsigned int start = 0; | |
515 int smp = 0; | |
516 int sb, store_prev_samples; | |
517 int64_t y; | |
10530
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
518 int use_ltp = 0; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
519 int ltp_lag = 0; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
520 int ltp_gain[5]; |
10522 | 521 |
522 *js_blocks = get_bits1(gb); | |
523 | |
524 // determine the number of subblocks for entropy decoding | |
525 if (!sconf->bgmc && !sconf->sb_part) { | |
526 log2_sub_blocks = 0; | |
527 } else { | |
528 if (sconf->bgmc && sconf->sb_part) | |
529 log2_sub_blocks = get_bits(gb, 2); | |
530 else | |
531 log2_sub_blocks = 2 * get_bits1(gb); | |
532 } | |
533 | |
534 sub_blocks = 1 << log2_sub_blocks; | |
535 | |
536 // do not continue in case of a damaged stream since | |
537 // block_length must be evenly divisible by sub_blocks | |
538 if (block_length & (sub_blocks - 1)) { | |
539 av_log(avctx, AV_LOG_WARNING, | |
540 "Block length is not evenly divisible by the number of subblocks.\n"); | |
541 return -1; | |
542 } | |
543 | |
544 sb_length = block_length >> log2_sub_blocks; | |
545 | |
546 | |
547 if (sconf->bgmc) { | |
548 // TODO: BGMC mode | |
549 } else { | |
550 s[0] = get_bits(gb, 4 + (sconf->resolution > 1)); | |
551 for (k = 1; k < sub_blocks; k++) | |
552 s[k] = s[k - 1] + decode_rice(gb, 0); | |
553 } | |
554 | |
555 if (get_bits1(gb)) | |
556 *shift_lsbs = get_bits(gb, 4) + 1; | |
557 | |
558 store_prev_samples = (*js_blocks && raw_other) || *shift_lsbs; | |
559 | |
560 | |
561 if (!sconf->rlslms) { | |
562 if (sconf->adapt_order) { | |
563 int opt_order_length = av_ceil_log2(av_clip((block_length >> 3) - 1, | |
564 2, sconf->max_order + 1)); | |
565 opt_order = get_bits(gb, opt_order_length); | |
566 } else { | |
567 opt_order = sconf->max_order; | |
568 } | |
569 | |
570 if (opt_order) { | |
571 int add_base; | |
572 | |
573 if (sconf->coef_table == 3) { | |
574 add_base = 0x7F; | |
575 | |
576 // read coefficient 0 | |
577 quant_cof[0] = 32 * parcor_scaled_values[get_bits(gb, 7)]; | |
578 | |
579 // read coefficient 1 | |
580 if (opt_order > 1) | |
581 quant_cof[1] = -32 * parcor_scaled_values[get_bits(gb, 7)]; | |
582 | |
583 // read coefficients 2 to opt_order | |
584 for (k = 2; k < opt_order; k++) | |
585 quant_cof[k] = get_bits(gb, 7); | |
586 } else { | |
587 int k_max; | |
588 add_base = 1; | |
589 | |
590 // read coefficient 0 to 19 | |
591 k_max = FFMIN(opt_order, 20); | |
592 for (k = 0; k < k_max; k++) { | |
593 int rice_param = parcor_rice_table[sconf->coef_table][k][1]; | |
594 int offset = parcor_rice_table[sconf->coef_table][k][0]; | |
595 quant_cof[k] = decode_rice(gb, rice_param) + offset; | |
596 } | |
597 | |
598 // read coefficients 20 to 126 | |
599 k_max = FFMIN(opt_order, 127); | |
600 for (; k < k_max; k++) | |
601 quant_cof[k] = decode_rice(gb, 2) + (k & 1); | |
602 | |
603 // read coefficients 127 to opt_order | |
604 for (; k < opt_order; k++) | |
605 quant_cof[k] = decode_rice(gb, 1); | |
606 | |
607 quant_cof[0] = 32 * parcor_scaled_values[quant_cof[0] + 64]; | |
608 | |
609 if (opt_order > 1) | |
610 quant_cof[1] = -32 * parcor_scaled_values[quant_cof[1] + 64]; | |
611 } | |
612 | |
613 for (k = 2; k < opt_order; k++) | |
614 quant_cof[k] = (quant_cof[k] << 14) + (add_base << 13); | |
615 } | |
616 } | |
617 | |
10530
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
618 // read LTP gain and lag values |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
619 if (sconf->long_term_prediction) { |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
620 use_ltp = get_bits1(gb); |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
621 |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
622 if (use_ltp) { |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
623 ltp_gain[0] = decode_rice(gb, 1) << 3; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
624 ltp_gain[1] = decode_rice(gb, 2) << 3; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
625 |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
626 ltp_gain[2] = ltp_gain_values[get_unary(gb, 0, 4)][get_bits(gb, 2)]; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
627 |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
628 ltp_gain[3] = decode_rice(gb, 2) << 3; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
629 ltp_gain[4] = decode_rice(gb, 1) << 3; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
630 |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
631 ltp_lag = get_bits(gb, ctx->ltp_lag_length); |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
632 ltp_lag += FFMAX(4, opt_order + 1); |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
633 } |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
634 } |
10522 | 635 |
636 // read first value and residuals in case of a random access block | |
637 if (ra_block) { | |
638 if (opt_order) | |
639 raw_samples[0] = decode_rice(gb, avctx->bits_per_raw_sample - 4); | |
640 if (opt_order > 1) | |
641 raw_samples[1] = decode_rice(gb, s[0] + 3); | |
642 if (opt_order > 2) | |
643 raw_samples[2] = decode_rice(gb, s[0] + 1); | |
644 | |
645 start = FFMIN(opt_order, 3); | |
646 } | |
647 | |
648 // read all residuals | |
649 if (sconf->bgmc) { | |
650 // TODO: BGMC mode | |
651 } else { | |
652 int32_t *current_res = raw_samples + start; | |
653 | |
654 for (sb = 0; sb < sub_blocks; sb++, start = 0) | |
655 for (; start < sb_length; start++) | |
656 *current_res++ = decode_rice(gb, s[sb]); | |
657 } | |
658 | |
10530
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
659 // reverse long-term prediction |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
660 if (use_ltp) { |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
661 int ltp_smp; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
662 |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
663 for (ltp_smp = FFMAX(ltp_lag - 2, 0); ltp_smp < block_length; ltp_smp++) { |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
664 int center = ltp_smp - ltp_lag; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
665 int begin = FFMAX(0, center - 2); |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
666 int end = center + 3; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
667 int tab = 5 - (end - begin); |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
668 int base; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
669 |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
670 y = 1 << 6; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
671 |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
672 for (base = begin; base < end; base++, tab++) |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
673 y += MUL64(ltp_gain[tab], raw_samples[base]); |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
674 |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
675 raw_samples[ltp_smp] += y >> 7; |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
676 } |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
677 } |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
678 |
10522 | 679 // reconstruct all samples from residuals |
680 if (ra_block) { | |
681 for (smp = 0; smp < opt_order; smp++) { | |
682 y = 1 << 19; | |
683 | |
684 for (sb = 0; sb < smp; sb++) | |
685 y += MUL64(lpc_cof[sb],raw_samples[smp - (sb + 1)]); | |
686 | |
687 raw_samples[smp] -= y >> 20; | |
688 parcor_to_lpc(smp, quant_cof, lpc_cof); | |
689 } | |
690 } else { | |
691 for (k = 0; k < opt_order; k++) | |
692 parcor_to_lpc(k, quant_cof, lpc_cof); | |
693 | |
694 // store previous samples in case that they have to be altered | |
695 if (store_prev_samples) | |
696 memcpy(ctx->prev_raw_samples, raw_samples - sconf->max_order, | |
697 sizeof(*ctx->prev_raw_samples) * sconf->max_order); | |
698 | |
699 // reconstruct difference signal for prediction (joint-stereo) | |
700 if (*js_blocks && raw_other) { | |
701 int32_t *left, *right; | |
702 | |
703 if (raw_other > raw_samples) { // D = R - L | |
704 left = raw_samples; | |
705 right = raw_other; | |
706 } else { // D = R - L | |
707 left = raw_other; | |
708 right = raw_samples; | |
709 } | |
710 | |
711 for (sb = -1; sb >= -sconf->max_order; sb--) | |
712 raw_samples[sb] = right[sb] - left[sb]; | |
713 } | |
714 | |
715 // reconstruct shifted signal | |
716 if (*shift_lsbs) | |
717 for (sb = -1; sb >= -sconf->max_order; sb--) | |
718 raw_samples[sb] >>= *shift_lsbs; | |
719 } | |
720 | |
721 // reconstruct raw samples | |
722 for (; smp < block_length; smp++) { | |
723 y = 1 << 19; | |
724 | |
725 for (sb = 0; sb < opt_order; sb++) | |
726 y += MUL64(lpc_cof[sb],raw_samples[smp - (sb + 1)]); | |
727 | |
728 raw_samples[smp] -= y >> 20; | |
729 } | |
730 | |
731 // restore previous samples in case that they have been altered | |
732 if (store_prev_samples) | |
733 memcpy(raw_samples - sconf->max_order, ctx->prev_raw_samples, | |
734 sizeof(*raw_samples) * sconf->max_order); | |
735 | |
736 return 0; | |
737 } | |
738 | |
739 | |
740 /** Reads the block data. | |
741 */ | |
742 static int read_block_data(ALSDecContext *ctx, unsigned int ra_block, | |
743 int32_t *raw_samples, unsigned int block_length, | |
744 unsigned int *js_blocks, int32_t *raw_other) | |
745 { | |
746 ALSSpecificConfig *sconf = &ctx->sconf; | |
747 GetBitContext *gb = &ctx->gb; | |
748 unsigned int shift_lsbs = 0; | |
749 unsigned int k; | |
750 | |
751 // read block type flag and read the samples accordingly | |
752 if (get_bits1(gb)) { | |
753 if (read_var_block(ctx, ra_block, raw_samples, block_length, js_blocks, | |
754 raw_other, &shift_lsbs)) | |
755 return -1; | |
756 } else { | |
757 read_const_block(ctx, raw_samples, block_length, js_blocks); | |
758 } | |
759 | |
760 // TODO: read RLSLMS extension data | |
761 | |
762 if (!sconf->mc_coding || ctx->js_switch) | |
763 align_get_bits(gb); | |
764 | |
765 if (shift_lsbs) | |
766 for (k = 0; k < block_length; k++) | |
767 raw_samples[k] <<= shift_lsbs; | |
768 | |
769 return 0; | |
770 } | |
771 | |
772 | |
773 /** Computes the number of samples left to decode for the current frame and | |
774 * sets these samples to zero. | |
775 */ | |
776 static void zero_remaining(unsigned int b, unsigned int b_max, | |
777 const unsigned int *div_blocks, int32_t *buf) | |
778 { | |
779 unsigned int count = 0; | |
780 | |
781 while (b < b_max) | |
782 count += div_blocks[b]; | |
783 | |
10523 | 784 if (count) |
10524 | 785 memset(buf, 0, sizeof(*buf) * count); |
10522 | 786 } |
787 | |
788 | |
789 /** Decodes blocks independently. | |
790 */ | |
791 static int decode_blocks_ind(ALSDecContext *ctx, unsigned int ra_frame, | |
792 unsigned int c, const unsigned int *div_blocks, | |
793 unsigned int *js_blocks) | |
794 { | |
795 int32_t *raw_sample; | |
796 unsigned int b; | |
797 raw_sample = ctx->raw_samples[c]; | |
798 | |
799 for (b = 0; b < ctx->num_blocks; b++) { | |
800 if (read_block_data(ctx, ra_frame, raw_sample, | |
801 div_blocks[b], &js_blocks[0], NULL)) { | |
802 // damaged block, write zero for the rest of the frame | |
803 zero_remaining(b, ctx->num_blocks, div_blocks, raw_sample); | |
804 return -1; | |
805 } | |
806 raw_sample += div_blocks[b]; | |
807 ra_frame = 0; | |
808 } | |
809 | |
810 return 0; | |
811 } | |
812 | |
813 | |
814 /** Decodes blocks dependently. | |
815 */ | |
816 static int decode_blocks(ALSDecContext *ctx, unsigned int ra_frame, | |
817 unsigned int c, const unsigned int *div_blocks, | |
818 unsigned int *js_blocks) | |
819 { | |
820 ALSSpecificConfig *sconf = &ctx->sconf; | |
821 unsigned int offset = 0; | |
822 int32_t *raw_samples_R; | |
823 int32_t *raw_samples_L; | |
824 unsigned int b; | |
825 | |
826 // decode all blocks | |
827 for (b = 0; b < ctx->num_blocks; b++) { | |
828 unsigned int s; | |
829 raw_samples_L = ctx->raw_samples[c ] + offset; | |
830 raw_samples_R = ctx->raw_samples[c + 1] + offset; | |
831 if (read_block_data(ctx, ra_frame, raw_samples_L, div_blocks[b], | |
832 &js_blocks[0], raw_samples_R) || | |
833 read_block_data(ctx, ra_frame, raw_samples_R, div_blocks[b], | |
834 &js_blocks[1], raw_samples_L)) { | |
835 // damaged block, write zero for the rest of the frame | |
836 zero_remaining(b, ctx->num_blocks, div_blocks, raw_samples_L); | |
837 zero_remaining(b, ctx->num_blocks, div_blocks, raw_samples_R); | |
838 return -1; | |
839 } | |
840 | |
841 // reconstruct joint-stereo blocks | |
842 if (js_blocks[0]) { | |
843 if (js_blocks[1]) | |
844 av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel pair!\n"); | |
845 | |
846 for (s = 0; s < div_blocks[b]; s++) | |
847 raw_samples_L[s] = raw_samples_R[s] - raw_samples_L[s]; | |
848 } else if (js_blocks[1]) { | |
849 for (s = 0; s < div_blocks[b]; s++) | |
850 raw_samples_R[s] = raw_samples_R[s] + raw_samples_L[s]; | |
851 } | |
852 | |
853 offset += div_blocks[b]; | |
854 ra_frame = 0; | |
855 } | |
856 | |
857 // store carryover raw samples, | |
858 // the others channel raw samples are stored by the calling function. | |
859 memmove(ctx->raw_samples[c] - sconf->max_order, | |
860 ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
861 sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
862 | |
863 return 0; | |
864 } | |
865 | |
866 | |
867 /** Reads the frame data. | |
868 */ | |
869 static int read_frame_data(ALSDecContext *ctx, unsigned int ra_frame) | |
870 { | |
871 ALSSpecificConfig *sconf = &ctx->sconf; | |
872 AVCodecContext *avctx = ctx->avctx; | |
873 GetBitContext *gb = &ctx->gb; | |
874 unsigned int div_blocks[32]; ///< block sizes. | |
875 unsigned int c; | |
876 unsigned int js_blocks[2]; | |
877 | |
878 uint32_t bs_info = 0; | |
879 | |
880 // skip the size of the ra unit if present in the frame | |
881 if (sconf->ra_flag == RA_FLAG_FRAMES && ra_frame) | |
882 skip_bits_long(gb, 32); | |
883 | |
884 if (sconf->mc_coding && sconf->joint_stereo) { | |
885 ctx->js_switch = get_bits1(gb); | |
886 align_get_bits(gb); | |
887 } | |
888 | |
889 if (!sconf->mc_coding || ctx->js_switch) { | |
890 int independent_bs = !sconf->joint_stereo; | |
891 | |
892 for (c = 0; c < avctx->channels; c++) { | |
893 js_blocks[0] = 0; | |
894 js_blocks[1] = 0; | |
895 | |
896 get_block_sizes(ctx, div_blocks, &bs_info); | |
897 | |
898 // if joint_stereo and block_switching is set, independent decoding | |
899 // is signaled via the first bit of bs_info | |
900 if (sconf->joint_stereo && sconf->block_switching) | |
901 if (bs_info >> 31) | |
902 independent_bs = 2; | |
903 | |
904 // if this is the last channel, it has to be decoded independently | |
905 if (c == avctx->channels - 1) | |
906 independent_bs = 1; | |
907 | |
908 if (independent_bs) { | |
909 if (decode_blocks_ind(ctx, ra_frame, c, div_blocks, js_blocks)) | |
910 return -1; | |
911 | |
912 independent_bs--; | |
913 } else { | |
914 if (decode_blocks(ctx, ra_frame, c, div_blocks, js_blocks)) | |
915 return -1; | |
916 | |
917 c++; | |
918 } | |
919 | |
920 // store carryover raw samples | |
921 memmove(ctx->raw_samples[c] - sconf->max_order, | |
922 ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
923 sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
924 } | |
925 } else { // multi-channel coding | |
926 get_block_sizes(ctx, div_blocks, &bs_info); | |
927 | |
928 // TODO: multi channel coding might use a temporary buffer instead as | |
929 // the actual channel is not known when read_block-data is called | |
930 if (decode_blocks_ind(ctx, ra_frame, 0, div_blocks, js_blocks)) | |
931 return -1; | |
932 // TODO: read_channel_data | |
933 } | |
934 | |
935 // TODO: read_diff_float_data | |
936 | |
937 return 0; | |
938 } | |
939 | |
940 | |
941 /** Decodes an ALS frame. | |
942 */ | |
943 static int decode_frame(AVCodecContext *avctx, | |
944 void *data, int *data_size, | |
945 AVPacket *avpkt) | |
946 { | |
947 ALSDecContext *ctx = avctx->priv_data; | |
948 ALSSpecificConfig *sconf = &ctx->sconf; | |
949 const uint8_t *buffer = avpkt->data; | |
950 int buffer_size = avpkt->size; | |
951 int invalid_frame, size; | |
952 unsigned int c, sample, ra_frame, bytes_read, shift; | |
953 | |
954 init_get_bits(&ctx->gb, buffer, buffer_size * 8); | |
955 | |
956 // In the case that the distance between random access frames is set to zero | |
957 // (sconf->ra_distance == 0) no frame is treated as a random access frame. | |
958 // For the first frame, if prediction is used, all samples used from the | |
959 // previous frame are assumed to be zero. | |
960 ra_frame = sconf->ra_distance && !(ctx->frame_id % sconf->ra_distance); | |
961 | |
962 // the last frame to decode might have a different length | |
963 if (sconf->samples != 0xFFFFFFFF) | |
964 ctx->cur_frame_length = FFMIN(sconf->samples - ctx->frame_id * (uint64_t) sconf->frame_length, | |
965 sconf->frame_length); | |
966 else | |
967 ctx->cur_frame_length = sconf->frame_length; | |
968 | |
969 // decode the frame data | |
970 if ((invalid_frame = read_frame_data(ctx, ra_frame) < 0)) | |
971 av_log(ctx->avctx, AV_LOG_WARNING, | |
972 "Reading frame data failed. Skipping RA unit.\n"); | |
973 | |
974 ctx->frame_id++; | |
975 | |
976 // check for size of decoded data | |
977 size = ctx->cur_frame_length * avctx->channels * | |
978 (av_get_bits_per_sample_format(avctx->sample_fmt) >> 3); | |
979 | |
980 if (size > *data_size) { | |
981 av_log(avctx, AV_LOG_ERROR, "Decoded data exceeds buffer size.\n"); | |
982 return -1; | |
983 } | |
984 | |
985 *data_size = size; | |
986 | |
987 // transform decoded frame into output format | |
988 #define INTERLEAVE_OUTPUT(bps) \ | |
989 { \ | |
990 int##bps##_t *dest = (int##bps##_t*) data; \ | |
991 shift = bps - ctx->avctx->bits_per_raw_sample; \ | |
992 for (sample = 0; sample < ctx->cur_frame_length; sample++) \ | |
993 for (c = 0; c < avctx->channels; c++) \ | |
994 *dest++ = ctx->raw_samples[c][sample] << shift; \ | |
995 } | |
996 | |
997 if (ctx->avctx->bits_per_raw_sample <= 16) { | |
998 INTERLEAVE_OUTPUT(16) | |
999 } else { | |
1000 INTERLEAVE_OUTPUT(32) | |
1001 } | |
1002 | |
1003 bytes_read = invalid_frame ? buffer_size : | |
1004 (get_bits_count(&ctx->gb) + 7) >> 3; | |
1005 | |
1006 return bytes_read; | |
1007 } | |
1008 | |
1009 | |
1010 /** Uninitializes the ALS decoder. | |
1011 */ | |
1012 static av_cold int decode_end(AVCodecContext *avctx) | |
1013 { | |
1014 ALSDecContext *ctx = avctx->priv_data; | |
1015 | |
1016 av_freep(&ctx->sconf.chan_pos); | |
1017 | |
1018 av_freep(&ctx->quant_cof); | |
1019 av_freep(&ctx->lpc_cof); | |
1020 av_freep(&ctx->prev_raw_samples); | |
1021 av_freep(&ctx->raw_samples); | |
1022 av_freep(&ctx->raw_buffer); | |
1023 | |
1024 return 0; | |
1025 } | |
1026 | |
1027 | |
1028 /** Initializes the ALS decoder. | |
1029 */ | |
1030 static av_cold int decode_init(AVCodecContext *avctx) | |
1031 { | |
1032 unsigned int c; | |
1033 unsigned int channel_size; | |
1034 ALSDecContext *ctx = avctx->priv_data; | |
1035 ALSSpecificConfig *sconf = &ctx->sconf; | |
1036 ctx->avctx = avctx; | |
1037 | |
1038 if (!avctx->extradata) { | |
1039 av_log(avctx, AV_LOG_ERROR, "Missing required ALS extradata.\n"); | |
1040 return -1; | |
1041 } | |
1042 | |
1043 if (read_specific_config(ctx)) { | |
1044 av_log(avctx, AV_LOG_ERROR, "Reading ALSSpecificConfig failed.\n"); | |
1045 decode_end(avctx); | |
1046 return -1; | |
1047 } | |
1048 | |
1049 if (check_specific_config(ctx)) { | |
1050 decode_end(avctx); | |
1051 return -1; | |
1052 } | |
1053 | |
1054 if (sconf->floating) { | |
1055 avctx->sample_fmt = SAMPLE_FMT_FLT; | |
1056 avctx->bits_per_raw_sample = 32; | |
1057 } else { | |
1058 avctx->sample_fmt = sconf->resolution > 1 | |
1059 ? SAMPLE_FMT_S32 : SAMPLE_FMT_S16; | |
1060 avctx->bits_per_raw_sample = (sconf->resolution + 1) * 8; | |
1061 } | |
1062 | |
10530
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
1063 // set lag value for long-term prediction |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
1064 ctx->ltp_lag_length = 8 + (avctx->sample_rate >= 96000) + |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
1065 (avctx->sample_rate >= 192000); |
d428e57f14c6
Add long-term prediction to the ALS decoder.
thilo.borgmann
parents:
10524
diff
changeset
|
1066 |
10522 | 1067 avctx->frame_size = sconf->frame_length; |
1068 channel_size = sconf->frame_length + sconf->max_order; | |
1069 | |
1070 ctx->prev_raw_samples = av_malloc (sizeof(*ctx->prev_raw_samples) * sconf->max_order); | |
1071 ctx->raw_buffer = av_mallocz(sizeof(*ctx-> raw_buffer) * avctx->channels * channel_size); | |
1072 ctx->raw_samples = av_malloc (sizeof(*ctx-> raw_samples) * avctx->channels); | |
1073 | |
1074 // allocate previous raw sample buffer | |
1075 if (!ctx->prev_raw_samples || !ctx->raw_buffer|| !ctx->raw_samples) { | |
1076 av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
1077 decode_end(avctx); | |
1078 return AVERROR(ENOMEM); | |
1079 } | |
1080 | |
1081 // assign raw samples buffers | |
1082 ctx->raw_samples[0] = ctx->raw_buffer + sconf->max_order; | |
1083 for (c = 1; c < avctx->channels; c++) | |
1084 ctx->raw_samples[c] = ctx->raw_samples[c - 1] + channel_size; | |
1085 | |
1086 return 0; | |
1087 } | |
1088 | |
1089 | |
1090 /** Flushes (resets) the frame ID after seeking. | |
1091 */ | |
1092 static av_cold void flush(AVCodecContext *avctx) | |
1093 { | |
1094 ALSDecContext *ctx = avctx->priv_data; | |
1095 | |
1096 ctx->frame_id = 0; | |
1097 } | |
1098 | |
1099 | |
1100 AVCodec als_decoder = { | |
1101 "als", | |
1102 CODEC_TYPE_AUDIO, | |
1103 CODEC_ID_MP4ALS, | |
1104 sizeof(ALSDecContext), | |
1105 decode_init, | |
1106 NULL, | |
1107 decode_end, | |
1108 decode_frame, | |
1109 .flush = flush, | |
1110 .capabilities = CODEC_CAP_SUBFRAMES, | |
1111 .long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Audio Lossless Coding (ALS)"), | |
1112 }; | |
1113 |