ffmpeg / libavcodec / dca.c @ 853daff6
History | View | Annotate | Download (69.5 KB)
| 1 |
/*
|
|---|---|
| 2 |
* DCA compatible decoder
|
| 3 |
* Copyright (C) 2004 Gildas Bazin
|
| 4 |
* Copyright (C) 2004 Benjamin Zores
|
| 5 |
* Copyright (C) 2006 Benjamin Larsson
|
| 6 |
* Copyright (C) 2007 Konstantin Shishkov
|
| 7 |
*
|
| 8 |
* This file is part of FFmpeg.
|
| 9 |
*
|
| 10 |
* FFmpeg is free software; you can redistribute it and/or
|
| 11 |
* modify it under the terms of the GNU Lesser General Public
|
| 12 |
* License as published by the Free Software Foundation; either
|
| 13 |
* version 2.1 of the License, or (at your option) any later version.
|
| 14 |
*
|
| 15 |
* FFmpeg is distributed in the hope that it will be useful,
|
| 16 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 17 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
| 18 |
* Lesser General Public License for more details.
|
| 19 |
*
|
| 20 |
* You should have received a copy of the GNU Lesser General Public
|
| 21 |
* License along with FFmpeg; if not, write to the Free Software
|
| 22 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
| 23 |
*/
|
| 24 |
|
| 25 |
#include <math.h> |
| 26 |
#include <stddef.h> |
| 27 |
#include <stdio.h> |
| 28 |
|
| 29 |
#include "libavutil/common.h" |
| 30 |
#include "libavutil/intmath.h" |
| 31 |
#include "libavutil/intreadwrite.h" |
| 32 |
#include "libavutil/audioconvert.h" |
| 33 |
#include "avcodec.h" |
| 34 |
#include "dsputil.h" |
| 35 |
#include "fft.h" |
| 36 |
#include "get_bits.h" |
| 37 |
#include "put_bits.h" |
| 38 |
#include "dcadata.h" |
| 39 |
#include "dcahuff.h" |
| 40 |
#include "dca.h" |
| 41 |
#include "synth_filter.h" |
| 42 |
#include "dcadsp.h" |
| 43 |
#include "fmtconvert.h" |
| 44 |
|
| 45 |
//#define TRACE
|
| 46 |
|
| 47 |
#define DCA_PRIM_CHANNELS_MAX (7) |
| 48 |
#define DCA_SUBBANDS (32) |
| 49 |
#define DCA_ABITS_MAX (32) /* Should be 28 */ |
| 50 |
#define DCA_SUBSUBFRAMES_MAX (4) |
| 51 |
#define DCA_SUBFRAMES_MAX (16) |
| 52 |
#define DCA_BLOCKS_MAX (16) |
| 53 |
#define DCA_LFE_MAX (3) |
| 54 |
|
| 55 |
enum DCAMode {
|
| 56 |
DCA_MONO = 0,
|
| 57 |
DCA_CHANNEL, |
| 58 |
DCA_STEREO, |
| 59 |
DCA_STEREO_SUMDIFF, |
| 60 |
DCA_STEREO_TOTAL, |
| 61 |
DCA_3F, |
| 62 |
DCA_2F1R, |
| 63 |
DCA_3F1R, |
| 64 |
DCA_2F2R, |
| 65 |
DCA_3F2R, |
| 66 |
DCA_4F2R |
| 67 |
}; |
| 68 |
|
| 69 |
/* these are unconfirmed but should be mostly correct */
|
| 70 |
enum DCAExSSSpeakerMask {
|
| 71 |
DCA_EXSS_FRONT_CENTER = 0x0001,
|
| 72 |
DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
|
| 73 |
DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
|
| 74 |
DCA_EXSS_LFE = 0x0008,
|
| 75 |
DCA_EXSS_REAR_CENTER = 0x0010,
|
| 76 |
DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
|
| 77 |
DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
|
| 78 |
DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
|
| 79 |
DCA_EXSS_OVERHEAD = 0x0100,
|
| 80 |
DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
|
| 81 |
DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
|
| 82 |
DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
|
| 83 |
DCA_EXSS_LFE2 = 0x1000,
|
| 84 |
DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
|
| 85 |
DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
|
| 86 |
DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
|
| 87 |
}; |
| 88 |
|
| 89 |
enum DCAExtensionMask {
|
| 90 |
DCA_EXT_CORE = 0x001, ///< core in core substream |
| 91 |
DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream |
| 92 |
DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream |
| 93 |
DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream |
| 94 |
DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream) |
| 95 |
DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS |
| 96 |
DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS |
| 97 |
DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS |
| 98 |
DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS |
| 99 |
DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS |
| 100 |
}; |
| 101 |
|
| 102 |
/* -1 are reserved or unknown */
|
| 103 |
static const int dca_ext_audio_descr_mask[] = { |
| 104 |
DCA_EXT_XCH, |
| 105 |
-1,
|
| 106 |
DCA_EXT_X96, |
| 107 |
DCA_EXT_XCH | DCA_EXT_X96, |
| 108 |
-1,
|
| 109 |
-1,
|
| 110 |
DCA_EXT_XXCH, |
| 111 |
-1,
|
| 112 |
}; |
| 113 |
|
| 114 |
/* extensions that reside in core substream */
|
| 115 |
#define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
|
| 116 |
|
| 117 |
/* Tables for mapping dts channel configurations to libavcodec multichannel api.
|
| 118 |
* Some compromises have been made for special configurations. Most configurations
|
| 119 |
* are never used so complete accuracy is not needed.
|
| 120 |
*
|
| 121 |
* L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
|
| 122 |
* S -> side, when both rear and back are configured move one of them to the side channel
|
| 123 |
* OV -> center back
|
| 124 |
* All 2 channel configurations -> AV_CH_LAYOUT_STEREO
|
| 125 |
*/
|
| 126 |
|
| 127 |
static const int64_t dca_core_channel_layout[] = { |
| 128 |
AV_CH_FRONT_CENTER, ///< 1, A
|
| 129 |
AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
|
| 130 |
AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
|
| 131 |
AV_CH_LAYOUT_STEREO, ///< 2, (L+R) + (L-R) (sum-difference)
|
| 132 |
AV_CH_LAYOUT_STEREO, ///< 2, LT +RT (left and right total)
|
| 133 |
AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER, ///< 3, C+L+R
|
| 134 |
AV_CH_LAYOUT_STEREO|AV_CH_BACK_CENTER, ///< 3, L+R+S
|
| 135 |
AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_BACK_CENTER, ///< 4, C + L + R+ S
|
| 136 |
AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 4, L + R +SL+ SR
|
| 137 |
AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 5, C + L + R+ SL+SR
|
| 138 |
AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
|
| 139 |
AV_CH_LAYOUT_STEREO|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT|AV_CH_FRONT_CENTER|AV_CH_BACK_CENTER, ///< 6, C + L + R+ LR + RR + OV
|
| 140 |
AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_BACK_CENTER|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT, ///< 6, CF+ CR+LF+ RF+LR + RR
|
| 141 |
AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
|
| 142 |
AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2+ SR1 + SR2
|
| 143 |
AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_BACK_CENTER|AV_CH_SIDE_RIGHT, ///< 8, CL + C+ CR + L + R + SL + S+ SR
|
| 144 |
}; |
| 145 |
|
| 146 |
static const int8_t dca_lfe_index[] = { |
| 147 |
1,2,2,2,2,3,2,3,2,3,2,3,1,3,2,3 |
| 148 |
}; |
| 149 |
|
| 150 |
static const int8_t dca_channel_reorder_lfe[][9] = { |
| 151 |
{ 0, -1, -1, -1, -1, -1, -1, -1, -1},
|
| 152 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 153 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 154 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 155 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 156 |
{ 2, 0, 1, -1, -1, -1, -1, -1, -1},
|
| 157 |
{ 0, 1, 3, -1, -1, -1, -1, -1, -1},
|
| 158 |
{ 2, 0, 1, 4, -1, -1, -1, -1, -1},
|
| 159 |
{ 0, 1, 3, 4, -1, -1, -1, -1, -1},
|
| 160 |
{ 2, 0, 1, 4, 5, -1, -1, -1, -1},
|
| 161 |
{ 3, 4, 0, 1, 5, 6, -1, -1, -1},
|
| 162 |
{ 2, 0, 1, 4, 5, 6, -1, -1, -1},
|
| 163 |
{ 0, 6, 4, 5, 2, 3, -1, -1, -1},
|
| 164 |
{ 4, 2, 5, 0, 1, 6, 7, -1, -1},
|
| 165 |
{ 5, 6, 0, 1, 7, 3, 8, 4, -1},
|
| 166 |
{ 4, 2, 5, 0, 1, 6, 8, 7, -1},
|
| 167 |
}; |
| 168 |
|
| 169 |
static const int8_t dca_channel_reorder_lfe_xch[][9] = { |
| 170 |
{ 0, 2, -1, -1, -1, -1, -1, -1, -1},
|
| 171 |
{ 0, 1, 3, -1, -1, -1, -1, -1, -1},
|
| 172 |
{ 0, 1, 3, -1, -1, -1, -1, -1, -1},
|
| 173 |
{ 0, 1, 3, -1, -1, -1, -1, -1, -1},
|
| 174 |
{ 0, 1, 3, -1, -1, -1, -1, -1, -1},
|
| 175 |
{ 2, 0, 1, 4, -1, -1, -1, -1, -1},
|
| 176 |
{ 0, 1, 3, 4, -1, -1, -1, -1, -1},
|
| 177 |
{ 2, 0, 1, 4, 5, -1, -1, -1, -1},
|
| 178 |
{ 0, 1, 4, 5, 3, -1, -1, -1, -1},
|
| 179 |
{ 2, 0, 1, 5, 6, 4, -1, -1, -1},
|
| 180 |
{ 3, 4, 0, 1, 6, 7, 5, -1, -1},
|
| 181 |
{ 2, 0, 1, 4, 5, 6, 7, -1, -1},
|
| 182 |
{ 0, 6, 4, 5, 2, 3, 7, -1, -1},
|
| 183 |
{ 4, 2, 5, 0, 1, 7, 8, 6, -1},
|
| 184 |
{ 5, 6, 0, 1, 8, 3, 9, 4, 7},
|
| 185 |
{ 4, 2, 5, 0, 1, 6, 9, 8, 7},
|
| 186 |
}; |
| 187 |
|
| 188 |
static const int8_t dca_channel_reorder_nolfe[][9] = { |
| 189 |
{ 0, -1, -1, -1, -1, -1, -1, -1, -1},
|
| 190 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 191 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 192 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 193 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 194 |
{ 2, 0, 1, -1, -1, -1, -1, -1, -1},
|
| 195 |
{ 0, 1, 2, -1, -1, -1, -1, -1, -1},
|
| 196 |
{ 2, 0, 1, 3, -1, -1, -1, -1, -1},
|
| 197 |
{ 0, 1, 2, 3, -1, -1, -1, -1, -1},
|
| 198 |
{ 2, 0, 1, 3, 4, -1, -1, -1, -1},
|
| 199 |
{ 2, 3, 0, 1, 4, 5, -1, -1, -1},
|
| 200 |
{ 2, 0, 1, 3, 4, 5, -1, -1, -1},
|
| 201 |
{ 0, 5, 3, 4, 1, 2, -1, -1, -1},
|
| 202 |
{ 3, 2, 4, 0, 1, 5, 6, -1, -1},
|
| 203 |
{ 4, 5, 0, 1, 6, 2, 7, 3, -1},
|
| 204 |
{ 3, 2, 4, 0, 1, 5, 7, 6, -1},
|
| 205 |
}; |
| 206 |
|
| 207 |
static const int8_t dca_channel_reorder_nolfe_xch[][9] = { |
| 208 |
{ 0, 1, -1, -1, -1, -1, -1, -1, -1},
|
| 209 |
{ 0, 1, 2, -1, -1, -1, -1, -1, -1},
|
| 210 |
{ 0, 1, 2, -1, -1, -1, -1, -1, -1},
|
| 211 |
{ 0, 1, 2, -1, -1, -1, -1, -1, -1},
|
| 212 |
{ 0, 1, 2, -1, -1, -1, -1, -1, -1},
|
| 213 |
{ 2, 0, 1, 3, -1, -1, -1, -1, -1},
|
| 214 |
{ 0, 1, 2, 3, -1, -1, -1, -1, -1},
|
| 215 |
{ 2, 0, 1, 3, 4, -1, -1, -1, -1},
|
| 216 |
{ 0, 1, 3, 4, 2, -1, -1, -1, -1},
|
| 217 |
{ 2, 0, 1, 4, 5, 3, -1, -1, -1},
|
| 218 |
{ 2, 3, 0, 1, 5, 6, 4, -1, -1},
|
| 219 |
{ 2, 0, 1, 3, 4, 5, 6, -1, -1},
|
| 220 |
{ 0, 5, 3, 4, 1, 2, 6, -1, -1},
|
| 221 |
{ 3, 2, 4, 0, 1, 6, 7, 5, -1},
|
| 222 |
{ 4, 5, 0, 1, 7, 2, 8, 3, 6},
|
| 223 |
{ 3, 2, 4, 0, 1, 5, 8, 7, 6},
|
| 224 |
}; |
| 225 |
|
| 226 |
#define DCA_DOLBY 101 /* FIXME */ |
| 227 |
|
| 228 |
#define DCA_CHANNEL_BITS 6 |
| 229 |
#define DCA_CHANNEL_MASK 0x3F |
| 230 |
|
| 231 |
#define DCA_LFE 0x80 |
| 232 |
|
| 233 |
#define HEADER_SIZE 14 |
| 234 |
|
| 235 |
#define DCA_MAX_FRAME_SIZE 16384 |
| 236 |
#define DCA_MAX_EXSS_HEADER_SIZE 4096 |
| 237 |
|
| 238 |
#define DCA_BUFFER_PADDING_SIZE 1024 |
| 239 |
|
| 240 |
/** Bit allocation */
|
| 241 |
typedef struct { |
| 242 |
int offset; ///< code values offset |
| 243 |
int maxbits[8]; ///< max bits in VLC |
| 244 |
int wrap; ///< wrap for get_vlc2() |
| 245 |
VLC vlc[8]; ///< actual codes |
| 246 |
} BitAlloc; |
| 247 |
|
| 248 |
static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select |
| 249 |
static BitAlloc dca_tmode; ///< transition mode VLCs |
| 250 |
static BitAlloc dca_scalefactor; ///< scalefactor VLCs |
| 251 |
static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs |
| 252 |
|
| 253 |
static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx) |
| 254 |
{
|
| 255 |
return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) + ba->offset;
|
| 256 |
} |
| 257 |
|
| 258 |
typedef struct { |
| 259 |
AVCodecContext *avctx; |
| 260 |
/* Frame header */
|
| 261 |
int frame_type; ///< type of the current frame |
| 262 |
int samples_deficit; ///< deficit sample count |
| 263 |
int crc_present; ///< crc is present in the bitstream |
| 264 |
int sample_blocks; ///< number of PCM sample blocks |
| 265 |
int frame_size; ///< primary frame byte size |
| 266 |
int amode; ///< audio channels arrangement |
| 267 |
int sample_rate; ///< audio sampling rate |
| 268 |
int bit_rate; ///< transmission bit rate |
| 269 |
int bit_rate_index; ///< transmission bit rate index |
| 270 |
|
| 271 |
int downmix; ///< embedded downmix enabled |
| 272 |
int dynrange; ///< embedded dynamic range flag |
| 273 |
int timestamp; ///< embedded time stamp flag |
| 274 |
int aux_data; ///< auxiliary data flag |
| 275 |
int hdcd; ///< source material is mastered in HDCD |
| 276 |
int ext_descr; ///< extension audio descriptor flag |
| 277 |
int ext_coding; ///< extended coding flag |
| 278 |
int aspf; ///< audio sync word insertion flag |
| 279 |
int lfe; ///< low frequency effects flag |
| 280 |
int predictor_history; ///< predictor history flag |
| 281 |
int header_crc; ///< header crc check bytes |
| 282 |
int multirate_inter; ///< multirate interpolator switch |
| 283 |
int version; ///< encoder software revision |
| 284 |
int copy_history; ///< copy history |
| 285 |
int source_pcm_res; ///< source pcm resolution |
| 286 |
int front_sum; ///< front sum/difference flag |
| 287 |
int surround_sum; ///< surround sum/difference flag |
| 288 |
int dialog_norm; ///< dialog normalisation parameter |
| 289 |
|
| 290 |
/* Primary audio coding header */
|
| 291 |
int subframes; ///< number of subframes |
| 292 |
int is_channels_set; ///< check for if the channel number is already set |
| 293 |
int total_channels; ///< number of channels including extensions |
| 294 |
int prim_channels; ///< number of primary audio channels |
| 295 |
int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count |
| 296 |
int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband |
| 297 |
int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index |
| 298 |
int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book |
| 299 |
int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book |
| 300 |
int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select |
| 301 |
int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select |
| 302 |
float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment |
| 303 |
|
| 304 |
/* Primary audio coding side information */
|
| 305 |
int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes |
| 306 |
int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count |
| 307 |
int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not) |
| 308 |
int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs |
| 309 |
int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index |
| 310 |
int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients) |
| 311 |
int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient) |
| 312 |
int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook |
| 313 |
int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors |
| 314 |
int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients |
| 315 |
int dynrange_coef; ///< dynamic range coefficient |
| 316 |
|
| 317 |
int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands |
| 318 |
|
| 319 |
float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data |
| 320 |
int lfe_scale_factor;
|
| 321 |
|
| 322 |
/* Subband samples history (for ADPCM) */
|
| 323 |
float subband_samples_hist[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4]; |
| 324 |
DECLARE_ALIGNED(16, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512]; |
| 325 |
DECLARE_ALIGNED(16, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32]; |
| 326 |
int hist_index[DCA_PRIM_CHANNELS_MAX];
|
| 327 |
DECLARE_ALIGNED(16, float, raXin)[32]; |
| 328 |
|
| 329 |
int output; ///< type of output |
| 330 |
float scale_bias; ///< output scale |
| 331 |
|
| 332 |
DECLARE_ALIGNED(16, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8]; |
| 333 |
DECLARE_ALIGNED(16, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256]; |
| 334 |
const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX+1]; |
| 335 |
|
| 336 |
uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE]; |
| 337 |
int dca_buffer_size; ///< how much data is in the dca_buffer |
| 338 |
|
| 339 |
const int8_t* channel_order_tab; ///< channel reordering table, lfe and non lfe |
| 340 |
GetBitContext gb; |
| 341 |
/* Current position in DCA frame */
|
| 342 |
int current_subframe;
|
| 343 |
int current_subsubframe;
|
| 344 |
|
| 345 |
int core_ext_mask; ///< present extensions in the core substream |
| 346 |
|
| 347 |
/* XCh extension information */
|
| 348 |
int xch_present; ///< XCh extension present and valid |
| 349 |
int xch_base_channel; ///< index of first (only) channel containing XCH data |
| 350 |
|
| 351 |
/* ExSS header parser */
|
| 352 |
int static_fields; ///< static fields present |
| 353 |
int mix_metadata; ///< mixing metadata present |
| 354 |
int num_mix_configs; ///< number of mix out configurations |
| 355 |
int mix_config_num_ch[4]; ///< number of channels in each mix out configuration |
| 356 |
|
| 357 |
int profile;
|
| 358 |
|
| 359 |
int debug_flag; ///< used for suppressing repeated error messages output |
| 360 |
DSPContext dsp; |
| 361 |
FFTContext imdct; |
| 362 |
SynthFilterContext synth; |
| 363 |
DCADSPContext dcadsp; |
| 364 |
FmtConvertContext fmt_conv; |
| 365 |
} DCAContext; |
| 366 |
|
| 367 |
static const uint16_t dca_vlc_offs[] = { |
| 368 |
0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364, |
| 369 |
5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508, |
| 370 |
5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564, |
| 371 |
7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240, |
| 372 |
12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264, |
| 373 |
18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622, |
| 374 |
}; |
| 375 |
|
| 376 |
static av_cold void dca_init_vlcs(void) |
| 377 |
{
|
| 378 |
static int vlcs_initialized = 0; |
| 379 |
int i, j, c = 14; |
| 380 |
static VLC_TYPE dca_table[23622][2]; |
| 381 |
|
| 382 |
if (vlcs_initialized)
|
| 383 |
return;
|
| 384 |
|
| 385 |
dca_bitalloc_index.offset = 1;
|
| 386 |
dca_bitalloc_index.wrap = 2;
|
| 387 |
for (i = 0; i < 5; i++) { |
| 388 |
dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]]; |
| 389 |
dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
|
| 390 |
init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
|
| 391 |
bitalloc_12_bits[i], 1, 1, |
| 392 |
bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC); |
| 393 |
} |
| 394 |
dca_scalefactor.offset = -64;
|
| 395 |
dca_scalefactor.wrap = 2;
|
| 396 |
for (i = 0; i < 5; i++) { |
| 397 |
dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
|
| 398 |
dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5]; |
| 399 |
init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
|
| 400 |
scales_bits[i], 1, 1, |
| 401 |
scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC); |
| 402 |
} |
| 403 |
dca_tmode.offset = 0;
|
| 404 |
dca_tmode.wrap = 1;
|
| 405 |
for (i = 0; i < 4; i++) { |
| 406 |
dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
|
| 407 |
dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10]; |
| 408 |
init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
|
| 409 |
tmode_bits[i], 1, 1, |
| 410 |
tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC); |
| 411 |
} |
| 412 |
|
| 413 |
for (i = 0; i < 10; i++) |
| 414 |
for (j = 0; j < 7; j++){ |
| 415 |
if (!bitalloc_codes[i][j]) break; |
| 416 |
dca_smpl_bitalloc[i+1].offset = bitalloc_offsets[i];
|
| 417 |
dca_smpl_bitalloc[i+1].wrap = 1 + (j > 4); |
| 418 |
dca_smpl_bitalloc[i+1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
|
| 419 |
dca_smpl_bitalloc[i+1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c]; |
| 420 |
init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j],
|
| 421 |
bitalloc_sizes[i], |
| 422 |
bitalloc_bits[i][j], 1, 1, |
| 423 |
bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC); |
| 424 |
c++; |
| 425 |
} |
| 426 |
vlcs_initialized = 1;
|
| 427 |
} |
| 428 |
|
| 429 |
static inline void get_array(GetBitContext *gb, int *dst, int len, int bits) |
| 430 |
{
|
| 431 |
while(len--)
|
| 432 |
*dst++ = get_bits(gb, bits); |
| 433 |
} |
| 434 |
|
| 435 |
static int dca_parse_audio_coding_header(DCAContext * s, int base_channel) |
| 436 |
{
|
| 437 |
int i, j;
|
| 438 |
static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 }; |
| 439 |
static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 }; |
| 440 |
static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 }; |
| 441 |
|
| 442 |
s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel; |
| 443 |
s->prim_channels = s->total_channels; |
| 444 |
|
| 445 |
if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
|
| 446 |
s->prim_channels = DCA_PRIM_CHANNELS_MAX; |
| 447 |
|
| 448 |
|
| 449 |
for (i = base_channel; i < s->prim_channels; i++) {
|
| 450 |
s->subband_activity[i] = get_bits(&s->gb, 5) + 2; |
| 451 |
if (s->subband_activity[i] > DCA_SUBBANDS)
|
| 452 |
s->subband_activity[i] = DCA_SUBBANDS; |
| 453 |
} |
| 454 |
for (i = base_channel; i < s->prim_channels; i++) {
|
| 455 |
s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1; |
| 456 |
if (s->vq_start_subband[i] > DCA_SUBBANDS)
|
| 457 |
s->vq_start_subband[i] = DCA_SUBBANDS; |
| 458 |
} |
| 459 |
get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
|
| 460 |
get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
|
| 461 |
get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
|
| 462 |
get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
|
| 463 |
|
| 464 |
/* Get codebooks quantization indexes */
|
| 465 |
if (!base_channel)
|
| 466 |
memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman)); |
| 467 |
for (j = 1; j < 11; j++) |
| 468 |
for (i = base_channel; i < s->prim_channels; i++)
|
| 469 |
s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]); |
| 470 |
|
| 471 |
/* Get scale factor adjustment */
|
| 472 |
for (j = 0; j < 11; j++) |
| 473 |
for (i = base_channel; i < s->prim_channels; i++)
|
| 474 |
s->scalefactor_adj[i][j] = 1;
|
| 475 |
|
| 476 |
for (j = 1; j < 11; j++) |
| 477 |
for (i = base_channel; i < s->prim_channels; i++)
|
| 478 |
if (s->quant_index_huffman[i][j] < thr[j])
|
| 479 |
s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
|
| 480 |
|
| 481 |
if (s->crc_present) {
|
| 482 |
/* Audio header CRC check */
|
| 483 |
get_bits(&s->gb, 16);
|
| 484 |
} |
| 485 |
|
| 486 |
s->current_subframe = 0;
|
| 487 |
s->current_subsubframe = 0;
|
| 488 |
|
| 489 |
#ifdef TRACE
|
| 490 |
av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
|
| 491 |
av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
|
| 492 |
for (i = base_channel; i < s->prim_channels; i++){
|
| 493 |
av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", s->subband_activity[i]);
|
| 494 |
av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", s->vq_start_subband[i]);
|
| 495 |
av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", s->joint_intensity[i]);
|
| 496 |
av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", s->transient_huffman[i]);
|
| 497 |
av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", s->scalefactor_huffman[i]);
|
| 498 |
av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", s->bitalloc_huffman[i]);
|
| 499 |
av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
|
| 500 |
for (j = 0; j < 11; j++) |
| 501 |
av_log(s->avctx, AV_LOG_DEBUG, " %i",
|
| 502 |
s->quant_index_huffman[i][j]); |
| 503 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 504 |
av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
|
| 505 |
for (j = 0; j < 11; j++) |
| 506 |
av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
|
| 507 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 508 |
} |
| 509 |
#endif
|
| 510 |
|
| 511 |
return 0; |
| 512 |
} |
| 513 |
|
| 514 |
static int dca_parse_frame_header(DCAContext * s) |
| 515 |
{
|
| 516 |
init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
|
| 517 |
|
| 518 |
/* Sync code */
|
| 519 |
get_bits(&s->gb, 32);
|
| 520 |
|
| 521 |
/* Frame header */
|
| 522 |
s->frame_type = get_bits(&s->gb, 1);
|
| 523 |
s->samples_deficit = get_bits(&s->gb, 5) + 1; |
| 524 |
s->crc_present = get_bits(&s->gb, 1);
|
| 525 |
s->sample_blocks = get_bits(&s->gb, 7) + 1; |
| 526 |
s->frame_size = get_bits(&s->gb, 14) + 1; |
| 527 |
if (s->frame_size < 95) |
| 528 |
return -1; |
| 529 |
s->amode = get_bits(&s->gb, 6);
|
| 530 |
s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
|
| 531 |
if (!s->sample_rate)
|
| 532 |
return -1; |
| 533 |
s->bit_rate_index = get_bits(&s->gb, 5);
|
| 534 |
s->bit_rate = dca_bit_rates[s->bit_rate_index]; |
| 535 |
if (!s->bit_rate)
|
| 536 |
return -1; |
| 537 |
|
| 538 |
s->downmix = get_bits(&s->gb, 1);
|
| 539 |
s->dynrange = get_bits(&s->gb, 1);
|
| 540 |
s->timestamp = get_bits(&s->gb, 1);
|
| 541 |
s->aux_data = get_bits(&s->gb, 1);
|
| 542 |
s->hdcd = get_bits(&s->gb, 1);
|
| 543 |
s->ext_descr = get_bits(&s->gb, 3);
|
| 544 |
s->ext_coding = get_bits(&s->gb, 1);
|
| 545 |
s->aspf = get_bits(&s->gb, 1);
|
| 546 |
s->lfe = get_bits(&s->gb, 2);
|
| 547 |
s->predictor_history = get_bits(&s->gb, 1);
|
| 548 |
|
| 549 |
/* TODO: check CRC */
|
| 550 |
if (s->crc_present)
|
| 551 |
s->header_crc = get_bits(&s->gb, 16);
|
| 552 |
|
| 553 |
s->multirate_inter = get_bits(&s->gb, 1);
|
| 554 |
s->version = get_bits(&s->gb, 4);
|
| 555 |
s->copy_history = get_bits(&s->gb, 2);
|
| 556 |
s->source_pcm_res = get_bits(&s->gb, 3);
|
| 557 |
s->front_sum = get_bits(&s->gb, 1);
|
| 558 |
s->surround_sum = get_bits(&s->gb, 1);
|
| 559 |
s->dialog_norm = get_bits(&s->gb, 4);
|
| 560 |
|
| 561 |
/* FIXME: channels mixing levels */
|
| 562 |
s->output = s->amode; |
| 563 |
if (s->lfe) s->output |= DCA_LFE;
|
| 564 |
|
| 565 |
#ifdef TRACE
|
| 566 |
av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
|
| 567 |
av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
|
| 568 |
av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
|
| 569 |
av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
|
| 570 |
s->sample_blocks, s->sample_blocks * 32);
|
| 571 |
av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
|
| 572 |
av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
|
| 573 |
s->amode, dca_channels[s->amode]); |
| 574 |
av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
|
| 575 |
s->sample_rate); |
| 576 |
av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
|
| 577 |
s->bit_rate); |
| 578 |
av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
|
| 579 |
av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
|
| 580 |
av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
|
| 581 |
av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
|
| 582 |
av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
|
| 583 |
av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
|
| 584 |
av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
|
| 585 |
av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
|
| 586 |
av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
|
| 587 |
av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
|
| 588 |
s->predictor_history); |
| 589 |
av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
|
| 590 |
av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
|
| 591 |
s->multirate_inter); |
| 592 |
av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
|
| 593 |
av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
|
| 594 |
av_log(s->avctx, AV_LOG_DEBUG, |
| 595 |
"source pcm resolution: %i (%i bits/sample)\n",
|
| 596 |
s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]); |
| 597 |
av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
|
| 598 |
av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
|
| 599 |
av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
|
| 600 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 601 |
#endif
|
| 602 |
|
| 603 |
/* Primary audio coding header */
|
| 604 |
s->subframes = get_bits(&s->gb, 4) + 1; |
| 605 |
|
| 606 |
return dca_parse_audio_coding_header(s, 0); |
| 607 |
} |
| 608 |
|
| 609 |
|
| 610 |
static inline int get_scale(GetBitContext *gb, int level, int value) |
| 611 |
{
|
| 612 |
if (level < 5) { |
| 613 |
/* huffman encoded */
|
| 614 |
value += get_bitalloc(gb, &dca_scalefactor, level); |
| 615 |
} else if (level < 8) |
| 616 |
value = get_bits(gb, level + 1);
|
| 617 |
return value;
|
| 618 |
} |
| 619 |
|
| 620 |
static int dca_subframe_header(DCAContext * s, int base_channel, int block_index) |
| 621 |
{
|
| 622 |
/* Primary audio coding side information */
|
| 623 |
int j, k;
|
| 624 |
|
| 625 |
if (get_bits_left(&s->gb) < 0) |
| 626 |
return -1; |
| 627 |
|
| 628 |
if (!base_channel) {
|
| 629 |
s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1; |
| 630 |
s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
|
| 631 |
} |
| 632 |
|
| 633 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 634 |
for (k = 0; k < s->subband_activity[j]; k++) |
| 635 |
s->prediction_mode[j][k] = get_bits(&s->gb, 1);
|
| 636 |
} |
| 637 |
|
| 638 |
/* Get prediction codebook */
|
| 639 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 640 |
for (k = 0; k < s->subband_activity[j]; k++) { |
| 641 |
if (s->prediction_mode[j][k] > 0) { |
| 642 |
/* (Prediction coefficient VQ address) */
|
| 643 |
s->prediction_vq[j][k] = get_bits(&s->gb, 12);
|
| 644 |
} |
| 645 |
} |
| 646 |
} |
| 647 |
|
| 648 |
/* Bit allocation index */
|
| 649 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 650 |
for (k = 0; k < s->vq_start_subband[j]; k++) { |
| 651 |
if (s->bitalloc_huffman[j] == 6) |
| 652 |
s->bitalloc[j][k] = get_bits(&s->gb, 5);
|
| 653 |
else if (s->bitalloc_huffman[j] == 5) |
| 654 |
s->bitalloc[j][k] = get_bits(&s->gb, 4);
|
| 655 |
else if (s->bitalloc_huffman[j] == 7) { |
| 656 |
av_log(s->avctx, AV_LOG_ERROR, |
| 657 |
"Invalid bit allocation index\n");
|
| 658 |
return -1; |
| 659 |
} else {
|
| 660 |
s->bitalloc[j][k] = |
| 661 |
get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]); |
| 662 |
} |
| 663 |
|
| 664 |
if (s->bitalloc[j][k] > 26) { |
| 665 |
// av_log(s->avctx,AV_LOG_DEBUG,"bitalloc index [%i][%i] too big (%i)\n",
|
| 666 |
// j, k, s->bitalloc[j][k]);
|
| 667 |
return -1; |
| 668 |
} |
| 669 |
} |
| 670 |
} |
| 671 |
|
| 672 |
/* Transition mode */
|
| 673 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 674 |
for (k = 0; k < s->subband_activity[j]; k++) { |
| 675 |
s->transition_mode[j][k] = 0;
|
| 676 |
if (s->subsubframes[s->current_subframe] > 1 && |
| 677 |
k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
|
| 678 |
s->transition_mode[j][k] = |
| 679 |
get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]); |
| 680 |
} |
| 681 |
} |
| 682 |
} |
| 683 |
|
| 684 |
if (get_bits_left(&s->gb) < 0) |
| 685 |
return -1; |
| 686 |
|
| 687 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 688 |
const uint32_t *scale_table;
|
| 689 |
int scale_sum;
|
| 690 |
|
| 691 |
memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2); |
| 692 |
|
| 693 |
if (s->scalefactor_huffman[j] == 6) |
| 694 |
scale_table = scale_factor_quant7; |
| 695 |
else
|
| 696 |
scale_table = scale_factor_quant6; |
| 697 |
|
| 698 |
/* When huffman coded, only the difference is encoded */
|
| 699 |
scale_sum = 0;
|
| 700 |
|
| 701 |
for (k = 0; k < s->subband_activity[j]; k++) { |
| 702 |
if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) { |
| 703 |
scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum); |
| 704 |
s->scale_factor[j][k][0] = scale_table[scale_sum];
|
| 705 |
} |
| 706 |
|
| 707 |
if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
|
| 708 |
/* Get second scale factor */
|
| 709 |
scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum); |
| 710 |
s->scale_factor[j][k][1] = scale_table[scale_sum];
|
| 711 |
} |
| 712 |
} |
| 713 |
} |
| 714 |
|
| 715 |
/* Joint subband scale factor codebook select */
|
| 716 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 717 |
/* Transmitted only if joint subband coding enabled */
|
| 718 |
if (s->joint_intensity[j] > 0) |
| 719 |
s->joint_huff[j] = get_bits(&s->gb, 3);
|
| 720 |
} |
| 721 |
|
| 722 |
if (get_bits_left(&s->gb) < 0) |
| 723 |
return -1; |
| 724 |
|
| 725 |
/* Scale factors for joint subband coding */
|
| 726 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 727 |
int source_channel;
|
| 728 |
|
| 729 |
/* Transmitted only if joint subband coding enabled */
|
| 730 |
if (s->joint_intensity[j] > 0) { |
| 731 |
int scale = 0; |
| 732 |
source_channel = s->joint_intensity[j] - 1;
|
| 733 |
|
| 734 |
/* When huffman coded, only the difference is encoded
|
| 735 |
* (is this valid as well for joint scales ???) */
|
| 736 |
|
| 737 |
for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
|
| 738 |
scale = get_scale(&s->gb, s->joint_huff[j], 0);
|
| 739 |
scale += 64; /* bias */ |
| 740 |
s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
|
| 741 |
} |
| 742 |
|
| 743 |
if (!(s->debug_flag & 0x02)) { |
| 744 |
av_log(s->avctx, AV_LOG_DEBUG, |
| 745 |
"Joint stereo coding not supported\n");
|
| 746 |
s->debug_flag |= 0x02;
|
| 747 |
} |
| 748 |
} |
| 749 |
} |
| 750 |
|
| 751 |
/* Stereo downmix coefficients */
|
| 752 |
if (!base_channel && s->prim_channels > 2) { |
| 753 |
if (s->downmix) {
|
| 754 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 755 |
s->downmix_coef[j][0] = get_bits(&s->gb, 7); |
| 756 |
s->downmix_coef[j][1] = get_bits(&s->gb, 7); |
| 757 |
} |
| 758 |
} else {
|
| 759 |
int am = s->amode & DCA_CHANNEL_MASK;
|
| 760 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 761 |
s->downmix_coef[j][0] = dca_default_coeffs[am][j][0]; |
| 762 |
s->downmix_coef[j][1] = dca_default_coeffs[am][j][1]; |
| 763 |
} |
| 764 |
} |
| 765 |
} |
| 766 |
|
| 767 |
/* Dynamic range coefficient */
|
| 768 |
if (!base_channel && s->dynrange)
|
| 769 |
s->dynrange_coef = get_bits(&s->gb, 8);
|
| 770 |
|
| 771 |
/* Side information CRC check word */
|
| 772 |
if (s->crc_present) {
|
| 773 |
get_bits(&s->gb, 16);
|
| 774 |
} |
| 775 |
|
| 776 |
/*
|
| 777 |
* Primary audio data arrays
|
| 778 |
*/
|
| 779 |
|
| 780 |
/* VQ encoded high frequency subbands */
|
| 781 |
for (j = base_channel; j < s->prim_channels; j++)
|
| 782 |
for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
|
| 783 |
/* 1 vector -> 32 samples */
|
| 784 |
s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
|
| 785 |
|
| 786 |
/* Low frequency effect data */
|
| 787 |
if (!base_channel && s->lfe) {
|
| 788 |
/* LFE samples */
|
| 789 |
int lfe_samples = 2 * s->lfe * (4 + block_index); |
| 790 |
int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]); |
| 791 |
float lfe_scale;
|
| 792 |
|
| 793 |
for (j = lfe_samples; j < lfe_end_sample; j++) {
|
| 794 |
/* Signed 8 bits int */
|
| 795 |
s->lfe_data[j] = get_sbits(&s->gb, 8);
|
| 796 |
} |
| 797 |
|
| 798 |
/* Scale factor index */
|
| 799 |
s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
|
| 800 |
|
| 801 |
/* Quantization step size * scale factor */
|
| 802 |
lfe_scale = 0.035 * s->lfe_scale_factor; |
| 803 |
|
| 804 |
for (j = lfe_samples; j < lfe_end_sample; j++)
|
| 805 |
s->lfe_data[j] *= lfe_scale; |
| 806 |
} |
| 807 |
|
| 808 |
#ifdef TRACE
|
| 809 |
av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", s->subsubframes[s->current_subframe]);
|
| 810 |
av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
|
| 811 |
s->partial_samples[s->current_subframe]); |
| 812 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 813 |
av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
|
| 814 |
for (k = 0; k < s->subband_activity[j]; k++) |
| 815 |
av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
|
| 816 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 817 |
} |
| 818 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 819 |
for (k = 0; k < s->subband_activity[j]; k++) |
| 820 |
av_log(s->avctx, AV_LOG_DEBUG, |
| 821 |
"prediction coefs: %f, %f, %f, %f\n",
|
| 822 |
(float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192, |
| 823 |
(float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192, |
| 824 |
(float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192, |
| 825 |
(float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192); |
| 826 |
} |
| 827 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 828 |
av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
|
| 829 |
for (k = 0; k < s->vq_start_subband[j]; k++) |
| 830 |
av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
|
| 831 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 832 |
} |
| 833 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 834 |
av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
|
| 835 |
for (k = 0; k < s->subband_activity[j]; k++) |
| 836 |
av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
|
| 837 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 838 |
} |
| 839 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 840 |
av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
|
| 841 |
for (k = 0; k < s->subband_activity[j]; k++) { |
| 842 |
if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) |
| 843 |
av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]); |
| 844 |
if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
|
| 845 |
av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]); |
| 846 |
} |
| 847 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 848 |
} |
| 849 |
for (j = base_channel; j < s->prim_channels; j++) {
|
| 850 |
if (s->joint_intensity[j] > 0) { |
| 851 |
int source_channel = s->joint_intensity[j] - 1; |
| 852 |
av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
|
| 853 |
for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
|
| 854 |
av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
|
| 855 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 856 |
} |
| 857 |
} |
| 858 |
if (!base_channel && s->prim_channels > 2 && s->downmix) { |
| 859 |
av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
|
| 860 |
for (j = 0; j < s->prim_channels; j++) { |
| 861 |
av_log(s->avctx, AV_LOG_DEBUG, "Channel 0,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][0]]); |
| 862 |
av_log(s->avctx, AV_LOG_DEBUG, "Channel 1,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][1]]); |
| 863 |
} |
| 864 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 865 |
} |
| 866 |
for (j = base_channel; j < s->prim_channels; j++)
|
| 867 |
for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
|
| 868 |
av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
|
| 869 |
if (!base_channel && s->lfe) {
|
| 870 |
int lfe_samples = 2 * s->lfe * (4 + block_index); |
| 871 |
int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]); |
| 872 |
|
| 873 |
av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
|
| 874 |
for (j = lfe_samples; j < lfe_end_sample; j++)
|
| 875 |
av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
|
| 876 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
| 877 |
} |
| 878 |
#endif
|
| 879 |
|
| 880 |
return 0; |
| 881 |
} |
| 882 |
|
| 883 |
static void qmf_32_subbands(DCAContext * s, int chans, |
| 884 |
float samples_in[32][8], float *samples_out, |
| 885 |
float scale)
|
| 886 |
{
|
| 887 |
const float *prCoeff; |
| 888 |
int i;
|
| 889 |
|
| 890 |
int sb_act = s->subband_activity[chans];
|
| 891 |
int subindex;
|
| 892 |
|
| 893 |
scale *= sqrt(1/8.0); |
| 894 |
|
| 895 |
/* Select filter */
|
| 896 |
if (!s->multirate_inter) /* Non-perfect reconstruction */ |
| 897 |
prCoeff = fir_32bands_nonperfect; |
| 898 |
else /* Perfect reconstruction */ |
| 899 |
prCoeff = fir_32bands_perfect; |
| 900 |
|
| 901 |
/* Reconstructed channel sample index */
|
| 902 |
for (subindex = 0; subindex < 8; subindex++) { |
| 903 |
/* Load in one sample from each subband and clear inactive subbands */
|
| 904 |
for (i = 0; i < sb_act; i++){ |
| 905 |
uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ ((i-1)&2)<<30; |
| 906 |
AV_WN32A(&s->raXin[i], v); |
| 907 |
} |
| 908 |
for (; i < 32; i++) |
| 909 |
s->raXin[i] = 0.0; |
| 910 |
|
| 911 |
s->synth.synth_filter_float(&s->imdct, |
| 912 |
s->subband_fir_hist[chans], &s->hist_index[chans], |
| 913 |
s->subband_fir_noidea[chans], prCoeff, |
| 914 |
samples_out, s->raXin, scale); |
| 915 |
samples_out+= 32;
|
| 916 |
|
| 917 |
} |
| 918 |
} |
| 919 |
|
| 920 |
static void lfe_interpolation_fir(DCAContext *s, int decimation_select, |
| 921 |
int num_deci_sample, float *samples_in, |
| 922 |
float *samples_out, float scale) |
| 923 |
{
|
| 924 |
/* samples_in: An array holding decimated samples.
|
| 925 |
* Samples in current subframe starts from samples_in[0],
|
| 926 |
* while samples_in[-1], samples_in[-2], ..., stores samples
|
| 927 |
* from last subframe as history.
|
| 928 |
*
|
| 929 |
* samples_out: An array holding interpolated samples
|
| 930 |
*/
|
| 931 |
|
| 932 |
int decifactor;
|
| 933 |
const float *prCoeff; |
| 934 |
int deciindex;
|
| 935 |
|
| 936 |
/* Select decimation filter */
|
| 937 |
if (decimation_select == 1) { |
| 938 |
decifactor = 64;
|
| 939 |
prCoeff = lfe_fir_128; |
| 940 |
} else {
|
| 941 |
decifactor = 32;
|
| 942 |
prCoeff = lfe_fir_64; |
| 943 |
} |
| 944 |
/* Interpolation */
|
| 945 |
for (deciindex = 0; deciindex < num_deci_sample; deciindex++) { |
| 946 |
s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, |
| 947 |
scale); |
| 948 |
samples_in++; |
| 949 |
samples_out += 2 * decifactor;
|
| 950 |
} |
| 951 |
} |
| 952 |
|
| 953 |
/* downmixing routines */
|
| 954 |
#define MIX_REAR1(samples, si1, rs, coef) \
|
| 955 |
samples[i] += samples[si1] * coef[rs][0]; \
|
| 956 |
samples[i+256] += samples[si1] * coef[rs][1]; |
| 957 |
|
| 958 |
#define MIX_REAR2(samples, si1, si2, rs, coef) \
|
| 959 |
samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs+1][0]; \ |
| 960 |
samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs+1][1]; |
| 961 |
|
| 962 |
#define MIX_FRONT3(samples, coef) \
|
| 963 |
t = samples[i+c]; \ |
| 964 |
u = samples[i+l]; \ |
| 965 |
v = samples[i+r]; \ |
| 966 |
samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \ |
| 967 |
samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1]; |
| 968 |
|
| 969 |
#define DOWNMIX_TO_STEREO(op1, op2) \
|
| 970 |
for (i = 0; i < 256; i++){ \ |
| 971 |
op1 \ |
| 972 |
op2 \ |
| 973 |
} |
| 974 |
|
| 975 |
static void dca_downmix(float *samples, int srcfmt, |
| 976 |
int downmix_coef[DCA_PRIM_CHANNELS_MAX][2], |
| 977 |
const int8_t *channel_mapping)
|
| 978 |
{
|
| 979 |
int c,l,r,sl,sr,s;
|
| 980 |
int i;
|
| 981 |
float t, u, v;
|
| 982 |
float coef[DCA_PRIM_CHANNELS_MAX][2]; |
| 983 |
|
| 984 |
for (i=0; i<DCA_PRIM_CHANNELS_MAX; i++) { |
| 985 |
coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]]; |
| 986 |
coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]]; |
| 987 |
} |
| 988 |
|
| 989 |
switch (srcfmt) {
|
| 990 |
case DCA_MONO:
|
| 991 |
case DCA_CHANNEL:
|
| 992 |
case DCA_STEREO_TOTAL:
|
| 993 |
case DCA_STEREO_SUMDIFF:
|
| 994 |
case DCA_4F2R:
|
| 995 |
av_log(NULL, 0, "Not implemented!\n"); |
| 996 |
break;
|
| 997 |
case DCA_STEREO:
|
| 998 |
break;
|
| 999 |
case DCA_3F:
|
| 1000 |
c = channel_mapping[0] * 256; |
| 1001 |
l = channel_mapping[1] * 256; |
| 1002 |
r = channel_mapping[2] * 256; |
| 1003 |
DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),); |
| 1004 |
break;
|
| 1005 |
case DCA_2F1R:
|
| 1006 |
s = channel_mapping[2] * 256; |
| 1007 |
DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef),);
|
| 1008 |
break;
|
| 1009 |
case DCA_3F1R:
|
| 1010 |
c = channel_mapping[0] * 256; |
| 1011 |
l = channel_mapping[1] * 256; |
| 1012 |
r = channel_mapping[2] * 256; |
| 1013 |
s = channel_mapping[3] * 256; |
| 1014 |
DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), |
| 1015 |
MIX_REAR1(samples, i + s, 3, coef));
|
| 1016 |
break;
|
| 1017 |
case DCA_2F2R:
|
| 1018 |
sl = channel_mapping[2] * 256; |
| 1019 |
sr = channel_mapping[3] * 256; |
| 1020 |
DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef),);
|
| 1021 |
break;
|
| 1022 |
case DCA_3F2R:
|
| 1023 |
c = channel_mapping[0] * 256; |
| 1024 |
l = channel_mapping[1] * 256; |
| 1025 |
r = channel_mapping[2] * 256; |
| 1026 |
sl = channel_mapping[3] * 256; |
| 1027 |
sr = channel_mapping[4] * 256; |
| 1028 |
DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), |
| 1029 |
MIX_REAR2(samples, i + sl, i + sr, 3, coef));
|
| 1030 |
break;
|
| 1031 |
} |
| 1032 |
} |
| 1033 |
|
| 1034 |
|
| 1035 |
/* Very compact version of the block code decoder that does not use table
|
| 1036 |
* look-up but is slightly slower */
|
| 1037 |
static int decode_blockcode(int code, int levels, int *values) |
| 1038 |
{
|
| 1039 |
int i;
|
| 1040 |
int offset = (levels - 1) >> 1; |
| 1041 |
|
| 1042 |
for (i = 0; i < 4; i++) { |
| 1043 |
int div = FASTDIV(code, levels);
|
| 1044 |
values[i] = code - offset - div*levels; |
| 1045 |
code = div; |
| 1046 |
} |
| 1047 |
|
| 1048 |
if (code == 0) |
| 1049 |
return 0; |
| 1050 |
else {
|
| 1051 |
av_log(NULL, AV_LOG_ERROR, "ERROR: block code look-up failed\n"); |
| 1052 |
return -1; |
| 1053 |
} |
| 1054 |
} |
| 1055 |
|
| 1056 |
static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 }; |
| 1057 |
static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 }; |
| 1058 |
|
| 1059 |
static int dca_subsubframe(DCAContext * s, int base_channel, int block_index) |
| 1060 |
{
|
| 1061 |
int k, l;
|
| 1062 |
int subsubframe = s->current_subsubframe;
|
| 1063 |
|
| 1064 |
const float *quant_step_table; |
| 1065 |
|
| 1066 |
/* FIXME */
|
| 1067 |
float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index]; |
| 1068 |
LOCAL_ALIGNED_16(int, block, [8]); |
| 1069 |
|
| 1070 |
/*
|
| 1071 |
* Audio data
|
| 1072 |
*/
|
| 1073 |
|
| 1074 |
/* Select quantization step size table */
|
| 1075 |
if (s->bit_rate_index == 0x1f) |
| 1076 |
quant_step_table = lossless_quant_d; |
| 1077 |
else
|
| 1078 |
quant_step_table = lossy_quant_d; |
| 1079 |
|
| 1080 |
for (k = base_channel; k < s->prim_channels; k++) {
|
| 1081 |
if (get_bits_left(&s->gb) < 0) |
| 1082 |
return -1; |
| 1083 |
|
| 1084 |
for (l = 0; l < s->vq_start_subband[k]; l++) { |
| 1085 |
int m;
|
| 1086 |
|
| 1087 |
/* Select the mid-tread linear quantizer */
|
| 1088 |
int abits = s->bitalloc[k][l];
|
| 1089 |
|
| 1090 |
float quant_step_size = quant_step_table[abits];
|
| 1091 |
|
| 1092 |
/*
|
| 1093 |
* Determine quantization index code book and its type
|
| 1094 |
*/
|
| 1095 |
|
| 1096 |
/* Select quantization index code book */
|
| 1097 |
int sel = s->quant_index_huffman[k][abits];
|
| 1098 |
|
| 1099 |
/*
|
| 1100 |
* Extract bits from the bit stream
|
| 1101 |
*/
|
| 1102 |
if (!abits){
|
| 1103 |
memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0])); |
| 1104 |
} else {
|
| 1105 |
/* Deal with transients */
|
| 1106 |
int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
|
| 1107 |
float rscale = quant_step_size * s->scale_factor[k][l][sfi] * s->scalefactor_adj[k][sel];
|
| 1108 |
|
| 1109 |
if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){ |
| 1110 |
if (abits <= 7){ |
| 1111 |
/* Block code */
|
| 1112 |
int block_code1, block_code2, size, levels;
|
| 1113 |
|
| 1114 |
size = abits_sizes[abits-1];
|
| 1115 |
levels = abits_levels[abits-1];
|
| 1116 |
|
| 1117 |
block_code1 = get_bits(&s->gb, size); |
| 1118 |
/* FIXME Should test return value */
|
| 1119 |
decode_blockcode(block_code1, levels, block); |
| 1120 |
block_code2 = get_bits(&s->gb, size); |
| 1121 |
decode_blockcode(block_code2, levels, &block[4]);
|
| 1122 |
}else{
|
| 1123 |
/* no coding */
|
| 1124 |
for (m = 0; m < 8; m++) |
| 1125 |
block[m] = get_sbits(&s->gb, abits - 3);
|
| 1126 |
} |
| 1127 |
}else{
|
| 1128 |
/* Huffman coded */
|
| 1129 |
for (m = 0; m < 8; m++) |
| 1130 |
block[m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel); |
| 1131 |
} |
| 1132 |
|
| 1133 |
s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l], |
| 1134 |
block, rscale, 8);
|
| 1135 |
} |
| 1136 |
|
| 1137 |
/*
|
| 1138 |
* Inverse ADPCM if in prediction mode
|
| 1139 |
*/
|
| 1140 |
if (s->prediction_mode[k][l]) {
|
| 1141 |
int n;
|
| 1142 |
for (m = 0; m < 8; m++) { |
| 1143 |
for (n = 1; n <= 4; n++) |
| 1144 |
if (m >= n)
|
| 1145 |
subband_samples[k][l][m] += |
| 1146 |
(adpcm_vb[s->prediction_vq[k][l]][n - 1] *
|
| 1147 |
subband_samples[k][l][m - n] / 8192);
|
| 1148 |
else if (s->predictor_history) |
| 1149 |
subband_samples[k][l][m] += |
| 1150 |
(adpcm_vb[s->prediction_vq[k][l]][n - 1] *
|
| 1151 |
s->subband_samples_hist[k][l][m - n + |
| 1152 |
4] / 8192); |
| 1153 |
} |
| 1154 |
} |
| 1155 |
} |
| 1156 |
|
| 1157 |
/*
|
| 1158 |
* Decode VQ encoded high frequencies
|
| 1159 |
*/
|
| 1160 |
for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
|
| 1161 |
/* 1 vector -> 32 samples but we only need the 8 samples
|
| 1162 |
* for this subsubframe. */
|
| 1163 |
int m;
|
| 1164 |
|
| 1165 |
if (!s->debug_flag & 0x01) { |
| 1166 |
av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
|
| 1167 |
s->debug_flag |= 0x01;
|
| 1168 |
} |
| 1169 |
|
| 1170 |
for (m = 0; m < 8; m++) { |
| 1171 |
subband_samples[k][l][m] = |
| 1172 |
high_freq_vq[s->high_freq_vq[k][l]][subsubframe * 8 +
|
| 1173 |
m] |
| 1174 |
* (float) s->scale_factor[k][l][0] / 16.0; |
| 1175 |
} |
| 1176 |
} |
| 1177 |
} |
| 1178 |
|
| 1179 |
/* Check for DSYNC after subsubframe */
|
| 1180 |
if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) { |
| 1181 |
if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */ |
| 1182 |
#ifdef TRACE
|
| 1183 |
av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
|
| 1184 |
#endif
|
| 1185 |
} else {
|
| 1186 |
av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
|
| 1187 |
} |
| 1188 |
} |
| 1189 |
|
| 1190 |
/* Backup predictor history for adpcm */
|
| 1191 |
for (k = base_channel; k < s->prim_channels; k++)
|
| 1192 |
for (l = 0; l < s->vq_start_subband[k]; l++) |
| 1193 |
memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
|
| 1194 |
4 * sizeof(subband_samples[0][0][0])); |
| 1195 |
|
| 1196 |
return 0; |
| 1197 |
} |
| 1198 |
|
| 1199 |
static int dca_filter_channels(DCAContext * s, int block_index) |
| 1200 |
{
|
| 1201 |
float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index]; |
| 1202 |
int k;
|
| 1203 |
|
| 1204 |
/* 32 subbands QMF */
|
| 1205 |
for (k = 0; k < s->prim_channels; k++) { |
| 1206 |
/* static float pcm_to_double[8] =
|
| 1207 |
{32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
|
| 1208 |
qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
|
| 1209 |
M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ );
|
| 1210 |
} |
| 1211 |
|
| 1212 |
/* Down mixing */
|
| 1213 |
if (s->avctx->request_channels == 2 && s->prim_channels > 2) { |
| 1214 |
dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab); |
| 1215 |
} |
| 1216 |
|
| 1217 |
/* Generate LFE samples for this subsubframe FIXME!!! */
|
| 1218 |
if (s->output & DCA_LFE) {
|
| 1219 |
lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
|
| 1220 |
s->lfe_data + 2 * s->lfe * (block_index + 4), |
| 1221 |
&s->samples[256 * dca_lfe_index[s->amode]],
|
| 1222 |
(1.0/256.0)*s->scale_bias); |
| 1223 |
/* Outputs 20bits pcm samples */
|
| 1224 |
} |
| 1225 |
|
| 1226 |
return 0; |
| 1227 |
} |
| 1228 |
|
| 1229 |
|
| 1230 |
static int dca_subframe_footer(DCAContext * s, int base_channel) |
| 1231 |
{
|
| 1232 |
int aux_data_count = 0, i; |
| 1233 |
|
| 1234 |
/*
|
| 1235 |
* Unpack optional information
|
| 1236 |
*/
|
| 1237 |
|
| 1238 |
/* presumably optional information only appears in the core? */
|
| 1239 |
if (!base_channel) {
|
| 1240 |
if (s->timestamp)
|
| 1241 |
get_bits(&s->gb, 32);
|
| 1242 |
|
| 1243 |
if (s->aux_data)
|
| 1244 |
aux_data_count = get_bits(&s->gb, 6);
|
| 1245 |
|
| 1246 |
for (i = 0; i < aux_data_count; i++) |
| 1247 |
get_bits(&s->gb, 8);
|
| 1248 |
|
| 1249 |
if (s->crc_present && (s->downmix || s->dynrange))
|
| 1250 |
get_bits(&s->gb, 16);
|
| 1251 |
} |
| 1252 |
|
| 1253 |
return 0; |
| 1254 |
} |
| 1255 |
|
| 1256 |
/**
|
| 1257 |
* Decode a dca frame block
|
| 1258 |
*
|
| 1259 |
* @param s pointer to the DCAContext
|
| 1260 |
*/
|
| 1261 |
|
| 1262 |
static int dca_decode_block(DCAContext * s, int base_channel, int block_index) |
| 1263 |
{
|
| 1264 |
|
| 1265 |
/* Sanity check */
|
| 1266 |
if (s->current_subframe >= s->subframes) {
|
| 1267 |
av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
|
| 1268 |
s->current_subframe, s->subframes); |
| 1269 |
return -1; |
| 1270 |
} |
| 1271 |
|
| 1272 |
if (!s->current_subsubframe) {
|
| 1273 |
#ifdef TRACE
|
| 1274 |
av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
|
| 1275 |
#endif
|
| 1276 |
/* Read subframe header */
|
| 1277 |
if (dca_subframe_header(s, base_channel, block_index))
|
| 1278 |
return -1; |
| 1279 |
} |
| 1280 |
|
| 1281 |
/* Read subsubframe */
|
| 1282 |
#ifdef TRACE
|
| 1283 |
av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
|
| 1284 |
#endif
|
| 1285 |
if (dca_subsubframe(s, base_channel, block_index))
|
| 1286 |
return -1; |
| 1287 |
|
| 1288 |
/* Update state */
|
| 1289 |
s->current_subsubframe++; |
| 1290 |
if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
|
| 1291 |
s->current_subsubframe = 0;
|
| 1292 |
s->current_subframe++; |
| 1293 |
} |
| 1294 |
if (s->current_subframe >= s->subframes) {
|
| 1295 |
#ifdef TRACE
|
| 1296 |
av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
|
| 1297 |
#endif
|
| 1298 |
/* Read subframe footer */
|
| 1299 |
if (dca_subframe_footer(s, base_channel))
|
| 1300 |
return -1; |
| 1301 |
} |
| 1302 |
|
| 1303 |
return 0; |
| 1304 |
} |
| 1305 |
|
| 1306 |
/**
|
| 1307 |
* Convert bitstream to one representation based on sync marker
|
| 1308 |
*/
|
| 1309 |
static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst, |
| 1310 |
int max_size)
|
| 1311 |
{
|
| 1312 |
uint32_t mrk; |
| 1313 |
int i, tmp;
|
| 1314 |
const uint16_t *ssrc = (const uint16_t *) src; |
| 1315 |
uint16_t *sdst = (uint16_t *) dst; |
| 1316 |
PutBitContext pb; |
| 1317 |
|
| 1318 |
if ((unsigned)src_size > (unsigned)max_size) { |
| 1319 |
// av_log(NULL, AV_LOG_ERROR, "Input frame size larger then DCA_MAX_FRAME_SIZE!\n");
|
| 1320 |
// return -1;
|
| 1321 |
src_size = max_size; |
| 1322 |
} |
| 1323 |
|
| 1324 |
mrk = AV_RB32(src); |
| 1325 |
switch (mrk) {
|
| 1326 |
case DCA_MARKER_RAW_BE:
|
| 1327 |
memcpy(dst, src, src_size); |
| 1328 |
return src_size;
|
| 1329 |
case DCA_MARKER_RAW_LE:
|
| 1330 |
for (i = 0; i < (src_size + 1) >> 1; i++) |
| 1331 |
*sdst++ = av_bswap16(*ssrc++); |
| 1332 |
return src_size;
|
| 1333 |
case DCA_MARKER_14B_BE:
|
| 1334 |
case DCA_MARKER_14B_LE:
|
| 1335 |
init_put_bits(&pb, dst, max_size); |
| 1336 |
for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) { |
| 1337 |
tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
|
| 1338 |
put_bits(&pb, 14, tmp);
|
| 1339 |
} |
| 1340 |
flush_put_bits(&pb); |
| 1341 |
return (put_bits_count(&pb) + 7) >> 3; |
| 1342 |
default:
|
| 1343 |
return -1; |
| 1344 |
} |
| 1345 |
} |
| 1346 |
|
| 1347 |
/**
|
| 1348 |
* Return the number of channels in an ExSS speaker mask (HD)
|
| 1349 |
*/
|
| 1350 |
static int dca_exss_mask2count(int mask) |
| 1351 |
{
|
| 1352 |
/* count bits that mean speaker pairs twice */
|
| 1353 |
return av_popcount(mask)
|
| 1354 |
+ av_popcount(mask & ( |
| 1355 |
DCA_EXSS_CENTER_LEFT_RIGHT |
| 1356 |
| DCA_EXSS_FRONT_LEFT_RIGHT |
| 1357 |
| DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
| 1358 |
| DCA_EXSS_WIDE_LEFT_RIGHT |
| 1359 |
| DCA_EXSS_SIDE_LEFT_RIGHT |
| 1360 |
| DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
| 1361 |
| DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
| 1362 |
| DCA_EXSS_REAR_LEFT_RIGHT |
| 1363 |
| DCA_EXSS_REAR_HIGH_LEFT_RIGHT |
| 1364 |
)); |
| 1365 |
} |
| 1366 |
|
| 1367 |
/**
|
| 1368 |
* Skip mixing coefficients of a single mix out configuration (HD)
|
| 1369 |
*/
|
| 1370 |
static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch) |
| 1371 |
{
|
| 1372 |
int i;
|
| 1373 |
|
| 1374 |
for (i = 0; i < channels; i++) { |
| 1375 |
int mix_map_mask = get_bits(gb, out_ch);
|
| 1376 |
int num_coeffs = av_popcount(mix_map_mask);
|
| 1377 |
skip_bits_long(gb, num_coeffs * 6);
|
| 1378 |
} |
| 1379 |
} |
| 1380 |
|
| 1381 |
/**
|
| 1382 |
* Parse extension substream asset header (HD)
|
| 1383 |
*/
|
| 1384 |
static int dca_exss_parse_asset_header(DCAContext *s) |
| 1385 |
{
|
| 1386 |
int header_pos = get_bits_count(&s->gb);
|
| 1387 |
int header_size;
|
| 1388 |
int channels;
|
| 1389 |
int embedded_stereo = 0; |
| 1390 |
int embedded_6ch = 0; |
| 1391 |
int drc_code_present;
|
| 1392 |
int extensions_mask;
|
| 1393 |
int i, j;
|
| 1394 |
|
| 1395 |
if (get_bits_left(&s->gb) < 16) |
| 1396 |
return -1; |
| 1397 |
|
| 1398 |
/* We will parse just enough to get to the extensions bitmask with which
|
| 1399 |
* we can set the profile value. */
|
| 1400 |
|
| 1401 |
header_size = get_bits(&s->gb, 9) + 1; |
| 1402 |
skip_bits(&s->gb, 3); // asset index |
| 1403 |
|
| 1404 |
if (s->static_fields) {
|
| 1405 |
if (get_bits1(&s->gb))
|
| 1406 |
skip_bits(&s->gb, 4); // asset type descriptor |
| 1407 |
if (get_bits1(&s->gb))
|
| 1408 |
skip_bits_long(&s->gb, 24); // language descriptor |
| 1409 |
|
| 1410 |
if (get_bits1(&s->gb)) {
|
| 1411 |
/* How can one fit 1024 bytes of text here if the maximum value
|
| 1412 |
* for the asset header size field above was 512 bytes? */
|
| 1413 |
int text_length = get_bits(&s->gb, 10) + 1; |
| 1414 |
if (get_bits_left(&s->gb) < text_length * 8) |
| 1415 |
return -1; |
| 1416 |
skip_bits_long(&s->gb, text_length * 8); // info text |
| 1417 |
} |
| 1418 |
|
| 1419 |
skip_bits(&s->gb, 5); // bit resolution - 1 |
| 1420 |
skip_bits(&s->gb, 4); // max sample rate code |
| 1421 |
channels = get_bits(&s->gb, 8) + 1; |
| 1422 |
|
| 1423 |
if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers |
| 1424 |
int spkr_remap_sets;
|
| 1425 |
int spkr_mask_size = 16; |
| 1426 |
int num_spkrs[7]; |
| 1427 |
|
| 1428 |
if (channels > 2) |
| 1429 |
embedded_stereo = get_bits1(&s->gb); |
| 1430 |
if (channels > 6) |
| 1431 |
embedded_6ch = get_bits1(&s->gb); |
| 1432 |
|
| 1433 |
if (get_bits1(&s->gb)) {
|
| 1434 |
spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2; |
| 1435 |
skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
|
| 1436 |
} |
| 1437 |
|
| 1438 |
spkr_remap_sets = get_bits(&s->gb, 3);
|
| 1439 |
|
| 1440 |
for (i = 0; i < spkr_remap_sets; i++) { |
| 1441 |
/* std layout mask for each remap set */
|
| 1442 |
num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size)); |
| 1443 |
} |
| 1444 |
|
| 1445 |
for (i = 0; i < spkr_remap_sets; i++) { |
| 1446 |
int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1; |
| 1447 |
if (get_bits_left(&s->gb) < 0) |
| 1448 |
return -1; |
| 1449 |
|
| 1450 |
for (j = 0; j < num_spkrs[i]; j++) { |
| 1451 |
int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
|
| 1452 |
int num_dec_ch = av_popcount(remap_dec_ch_mask);
|
| 1453 |
skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes |
| 1454 |
} |
| 1455 |
} |
| 1456 |
|
| 1457 |
} else {
|
| 1458 |
skip_bits(&s->gb, 3); // representation type |
| 1459 |
} |
| 1460 |
} |
| 1461 |
|
| 1462 |
drc_code_present = get_bits1(&s->gb); |
| 1463 |
if (drc_code_present)
|
| 1464 |
get_bits(&s->gb, 8); // drc code |
| 1465 |
|
| 1466 |
if (get_bits1(&s->gb))
|
| 1467 |
skip_bits(&s->gb, 5); // dialog normalization code |
| 1468 |
|
| 1469 |
if (drc_code_present && embedded_stereo)
|
| 1470 |
get_bits(&s->gb, 8); // drc stereo code |
| 1471 |
|
| 1472 |
if (s->mix_metadata && get_bits1(&s->gb)) {
|
| 1473 |
skip_bits(&s->gb, 1); // external mix |
| 1474 |
skip_bits(&s->gb, 6); // post mix gain code |
| 1475 |
|
| 1476 |
if (get_bits(&s->gb, 2) != 3) // mixer drc code |
| 1477 |
skip_bits(&s->gb, 3); // drc limit |
| 1478 |
else
|
| 1479 |
skip_bits(&s->gb, 8); // custom drc code |
| 1480 |
|
| 1481 |
if (get_bits1(&s->gb)) // channel specific scaling |
| 1482 |
for (i = 0; i < s->num_mix_configs; i++) |
| 1483 |
skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes |
| 1484 |
else
|
| 1485 |
skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes |
| 1486 |
|
| 1487 |
for (i = 0; i < s->num_mix_configs; i++) { |
| 1488 |
if (get_bits_left(&s->gb) < 0) |
| 1489 |
return -1; |
| 1490 |
dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]); |
| 1491 |
if (embedded_6ch)
|
| 1492 |
dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
|
| 1493 |
if (embedded_stereo)
|
| 1494 |
dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
|
| 1495 |
} |
| 1496 |
} |
| 1497 |
|
| 1498 |
switch (get_bits(&s->gb, 2)) { |
| 1499 |
case 0: extensions_mask = get_bits(&s->gb, 12); break; |
| 1500 |
case 1: extensions_mask = DCA_EXT_EXSS_XLL; break; |
| 1501 |
case 2: extensions_mask = DCA_EXT_EXSS_LBR; break; |
| 1502 |
case 3: extensions_mask = 0; /* aux coding */ break; |
| 1503 |
} |
| 1504 |
|
| 1505 |
/* not parsed further, we were only interested in the extensions mask */
|
| 1506 |
|
| 1507 |
if (get_bits_left(&s->gb) < 0) |
| 1508 |
return -1; |
| 1509 |
|
| 1510 |
if (get_bits_count(&s->gb) - header_pos > header_size * 8) { |
| 1511 |
av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
|
| 1512 |
return -1; |
| 1513 |
} |
| 1514 |
skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
|
| 1515 |
|
| 1516 |
if (extensions_mask & DCA_EXT_EXSS_XLL)
|
| 1517 |
s->profile = FF_PROFILE_DTS_HD_MA; |
| 1518 |
else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 | |
| 1519 |
DCA_EXT_EXSS_XXCH)) |
| 1520 |
s->profile = FF_PROFILE_DTS_HD_HRA; |
| 1521 |
|
| 1522 |
if (!(extensions_mask & DCA_EXT_CORE))
|
| 1523 |
av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
|
| 1524 |
if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
|
| 1525 |
av_log(s->avctx, AV_LOG_WARNING, "DTS extensions detection mismatch (%d, %d)\n",
|
| 1526 |
extensions_mask & DCA_CORE_EXTS, s->core_ext_mask); |
| 1527 |
|
| 1528 |
return 0; |
| 1529 |
} |
| 1530 |
|
| 1531 |
/**
|
| 1532 |
* Parse extension substream header (HD)
|
| 1533 |
*/
|
| 1534 |
static void dca_exss_parse_header(DCAContext *s) |
| 1535 |
{
|
| 1536 |
int ss_index;
|
| 1537 |
int blownup;
|
| 1538 |
int header_size;
|
| 1539 |
int hd_size;
|
| 1540 |
int num_audiop = 1; |
| 1541 |
int num_assets = 1; |
| 1542 |
int active_ss_mask[8]; |
| 1543 |
int i, j;
|
| 1544 |
|
| 1545 |
if (get_bits_left(&s->gb) < 52) |
| 1546 |
return;
|
| 1547 |
|
| 1548 |
skip_bits(&s->gb, 8); // user data |
| 1549 |
ss_index = get_bits(&s->gb, 2);
|
| 1550 |
|
| 1551 |
blownup = get_bits1(&s->gb); |
| 1552 |
header_size = get_bits(&s->gb, 8 + 4 * blownup) + 1; |
| 1553 |
hd_size = get_bits_long(&s->gb, 16 + 4 * blownup) + 1; |
| 1554 |
|
| 1555 |
s->static_fields = get_bits1(&s->gb); |
| 1556 |
if (s->static_fields) {
|
| 1557 |
skip_bits(&s->gb, 2); // reference clock code |
| 1558 |
skip_bits(&s->gb, 3); // frame duration code |
| 1559 |
|
| 1560 |
if (get_bits1(&s->gb))
|
| 1561 |
skip_bits_long(&s->gb, 36); // timestamp |
| 1562 |
|
| 1563 |
/* a single stream can contain multiple audio assets that can be
|
| 1564 |
* combined to form multiple audio presentations */
|
| 1565 |
|
| 1566 |
num_audiop = get_bits(&s->gb, 3) + 1; |
| 1567 |
if (num_audiop > 1) { |
| 1568 |
av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
|
| 1569 |
/* ignore such streams for now */
|
| 1570 |
return;
|
| 1571 |
} |
| 1572 |
|
| 1573 |
num_assets = get_bits(&s->gb, 3) + 1; |
| 1574 |
if (num_assets > 1) { |
| 1575 |
av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
|
| 1576 |
/* ignore such streams for now */
|
| 1577 |
return;
|
| 1578 |
} |
| 1579 |
|
| 1580 |
for (i = 0; i < num_audiop; i++) |
| 1581 |
active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
|
| 1582 |
|
| 1583 |
for (i = 0; i < num_audiop; i++) |
| 1584 |
for (j = 0; j <= ss_index; j++) |
| 1585 |
if (active_ss_mask[i] & (1 << j)) |
| 1586 |
skip_bits(&s->gb, 8); // active asset mask |
| 1587 |
|
| 1588 |
s->mix_metadata = get_bits1(&s->gb); |
| 1589 |
if (s->mix_metadata) {
|
| 1590 |
int mix_out_mask_size;
|
| 1591 |
|
| 1592 |
skip_bits(&s->gb, 2); // adjustment level |
| 1593 |
mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2; |
| 1594 |
s->num_mix_configs = get_bits(&s->gb, 2) + 1; |
| 1595 |
|
| 1596 |
for (i = 0; i < s->num_mix_configs; i++) { |
| 1597 |
int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
|
| 1598 |
s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask); |
| 1599 |
} |
| 1600 |
} |
| 1601 |
} |
| 1602 |
|
| 1603 |
for (i = 0; i < num_assets; i++) |
| 1604 |
skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size |
| 1605 |
|
| 1606 |
for (i = 0; i < num_assets; i++) { |
| 1607 |
if (dca_exss_parse_asset_header(s))
|
| 1608 |
return;
|
| 1609 |
} |
| 1610 |
|
| 1611 |
/* not parsed further, we were only interested in the extensions mask
|
| 1612 |
* from the asset header */
|
| 1613 |
} |
| 1614 |
|
| 1615 |
/**
|
| 1616 |
* Main frame decoding function
|
| 1617 |
* FIXME add arguments
|
| 1618 |
*/
|
| 1619 |
static int dca_decode_frame(AVCodecContext * avctx, |
| 1620 |
void *data, int *data_size, |
| 1621 |
AVPacket *avpkt) |
| 1622 |
{
|
| 1623 |
const uint8_t *buf = avpkt->data;
|
| 1624 |
int buf_size = avpkt->size;
|
| 1625 |
|
| 1626 |
int lfe_samples;
|
| 1627 |
int num_core_channels = 0; |
| 1628 |
int i;
|
| 1629 |
int16_t *samples = data; |
| 1630 |
DCAContext *s = avctx->priv_data; |
| 1631 |
int channels;
|
| 1632 |
int core_ss_end;
|
| 1633 |
|
| 1634 |
|
| 1635 |
s->xch_present = 0;
|
| 1636 |
|
| 1637 |
s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer, |
| 1638 |
DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE); |
| 1639 |
if (s->dca_buffer_size == -1) { |
| 1640 |
av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
|
| 1641 |
return -1; |
| 1642 |
} |
| 1643 |
|
| 1644 |
init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
|
| 1645 |
if (dca_parse_frame_header(s) < 0) { |
| 1646 |
//seems like the frame is corrupt, try with the next one
|
| 1647 |
*data_size=0;
|
| 1648 |
return buf_size;
|
| 1649 |
} |
| 1650 |
//set AVCodec values with parsed data
|
| 1651 |
avctx->sample_rate = s->sample_rate; |
| 1652 |
avctx->bit_rate = s->bit_rate; |
| 1653 |
|
| 1654 |
s->profile = FF_PROFILE_DTS; |
| 1655 |
|
| 1656 |
for (i = 0; i < (s->sample_blocks / 8); i++) { |
| 1657 |
dca_decode_block(s, 0, i);
|
| 1658 |
} |
| 1659 |
|
| 1660 |
/* record number of core channels incase less than max channels are requested */
|
| 1661 |
num_core_channels = s->prim_channels; |
| 1662 |
|
| 1663 |
if (s->ext_coding)
|
| 1664 |
s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr]; |
| 1665 |
else
|
| 1666 |
s->core_ext_mask = 0;
|
| 1667 |
|
| 1668 |
core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
|
| 1669 |
|
| 1670 |
/* only scan for extensions if ext_descr was unknown or indicated a
|
| 1671 |
* supported XCh extension */
|
| 1672 |
if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) { |
| 1673 |
|
| 1674 |
/* if ext_descr was unknown, clear s->core_ext_mask so that the
|
| 1675 |
* extensions scan can fill it up */
|
| 1676 |
s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
|
| 1677 |
|
| 1678 |
/* extensions start at 32-bit boundaries into bitstream */
|
| 1679 |
skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
|
| 1680 |
|
| 1681 |
while(core_ss_end - get_bits_count(&s->gb) >= 32) { |
| 1682 |
uint32_t bits = get_bits_long(&s->gb, 32);
|
| 1683 |
|
| 1684 |
switch(bits) {
|
| 1685 |
case 0x5a5a5a5a: { |
| 1686 |
int ext_amode, xch_fsize;
|
| 1687 |
|
| 1688 |
s->xch_base_channel = s->prim_channels; |
| 1689 |
|
| 1690 |
/* validate sync word using XCHFSIZE field */
|
| 1691 |
xch_fsize = show_bits(&s->gb, 10);
|
| 1692 |
if((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) && |
| 1693 |
(s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1)) |
| 1694 |
continue;
|
| 1695 |
|
| 1696 |
/* skip length-to-end-of-frame field for the moment */
|
| 1697 |
skip_bits(&s->gb, 10);
|
| 1698 |
|
| 1699 |
s->core_ext_mask |= DCA_EXT_XCH; |
| 1700 |
|
| 1701 |
/* extension amode should == 1, number of channels in extension */
|
| 1702 |
/* AFAIK XCh is not used for more channels */
|
| 1703 |
if ((ext_amode = get_bits(&s->gb, 4)) != 1) { |
| 1704 |
av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
|
| 1705 |
" supported!\n",ext_amode);
|
| 1706 |
continue;
|
| 1707 |
} |
| 1708 |
|
| 1709 |
/* much like core primary audio coding header */
|
| 1710 |
dca_parse_audio_coding_header(s, s->xch_base_channel); |
| 1711 |
|
| 1712 |
for (i = 0; i < (s->sample_blocks / 8); i++) { |
| 1713 |
dca_decode_block(s, s->xch_base_channel, i); |
| 1714 |
} |
| 1715 |
|
| 1716 |
s->xch_present = 1;
|
| 1717 |
break;
|
| 1718 |
} |
| 1719 |
case 0x47004a03: |
| 1720 |
/* XXCh: extended channels */
|
| 1721 |
/* usually found either in core or HD part in DTS-HD HRA streams,
|
| 1722 |
* but not in DTS-ES which contains XCh extensions instead */
|
| 1723 |
s->core_ext_mask |= DCA_EXT_XXCH; |
| 1724 |
break;
|
| 1725 |
|
| 1726 |
case 0x1d95f262: { |
| 1727 |
int fsize96 = show_bits(&s->gb, 12) + 1; |
| 1728 |
if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96) |
| 1729 |
continue;
|
| 1730 |
|
| 1731 |
av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n", get_bits_count(&s->gb));
|
| 1732 |
skip_bits(&s->gb, 12);
|
| 1733 |
av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
|
| 1734 |
av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4)); |
| 1735 |
|
| 1736 |
s->core_ext_mask |= DCA_EXT_X96; |
| 1737 |
break;
|
| 1738 |
} |
| 1739 |
} |
| 1740 |
|
| 1741 |
skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
|
| 1742 |
} |
| 1743 |
|
| 1744 |
} else {
|
| 1745 |
/* no supported extensions, skip the rest of the core substream */
|
| 1746 |
skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb)); |
| 1747 |
} |
| 1748 |
|
| 1749 |
if (s->core_ext_mask & DCA_EXT_X96)
|
| 1750 |
s->profile = FF_PROFILE_DTS_96_24; |
| 1751 |
else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH)) |
| 1752 |
s->profile = FF_PROFILE_DTS_ES; |
| 1753 |
|
| 1754 |
/* check for ExSS (HD part) */
|
| 1755 |
if (s->dca_buffer_size - s->frame_size > 32 |
| 1756 |
&& get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
|
| 1757 |
dca_exss_parse_header(s); |
| 1758 |
|
| 1759 |
avctx->profile = s->profile; |
| 1760 |
|
| 1761 |
channels = s->prim_channels + !!s->lfe; |
| 1762 |
|
| 1763 |
if (s->amode<16) { |
| 1764 |
avctx->channel_layout = dca_core_channel_layout[s->amode]; |
| 1765 |
|
| 1766 |
if (s->xch_present && (!avctx->request_channels ||
|
| 1767 |
avctx->request_channels > num_core_channels + !!s->lfe)) {
|
| 1768 |
avctx->channel_layout |= AV_CH_BACK_CENTER; |
| 1769 |
if (s->lfe) {
|
| 1770 |
avctx->channel_layout |= AV_CH_LOW_FREQUENCY; |
| 1771 |
s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode]; |
| 1772 |
} else {
|
| 1773 |
s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode]; |
| 1774 |
} |
| 1775 |
} else {
|
| 1776 |
channels = num_core_channels + !!s->lfe; |
| 1777 |
s->xch_present = 0; /* disable further xch processing */ |
| 1778 |
if (s->lfe) {
|
| 1779 |
avctx->channel_layout |= AV_CH_LOW_FREQUENCY; |
| 1780 |
s->channel_order_tab = dca_channel_reorder_lfe[s->amode]; |
| 1781 |
} else
|
| 1782 |
s->channel_order_tab = dca_channel_reorder_nolfe[s->amode]; |
| 1783 |
} |
| 1784 |
|
| 1785 |
if (channels > !!s->lfe &&
|
| 1786 |
s->channel_order_tab[channels - 1 - !!s->lfe] < 0) |
| 1787 |
return -1; |
| 1788 |
|
| 1789 |
if (avctx->request_channels == 2 && s->prim_channels > 2) { |
| 1790 |
channels = 2;
|
| 1791 |
s->output = DCA_STEREO; |
| 1792 |
avctx->channel_layout = AV_CH_LAYOUT_STEREO; |
| 1793 |
} |
| 1794 |
} else {
|
| 1795 |
av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
|
| 1796 |
return -1; |
| 1797 |
} |
| 1798 |
|
| 1799 |
|
| 1800 |
/* There is nothing that prevents a dts frame to change channel configuration
|
| 1801 |
but FFmpeg doesn't support that so only set the channels if it is previously
|
| 1802 |
unset. Ideally during the first probe for channels the crc should be checked
|
| 1803 |
and only set avctx->channels when the crc is ok. Right now the decoder could
|
| 1804 |
set the channels based on a broken first frame.*/
|
| 1805 |
if (s->is_channels_set == 0) { |
| 1806 |
s->is_channels_set = 1;
|
| 1807 |
avctx->channels = channels; |
| 1808 |
} |
| 1809 |
if (avctx->channels != channels) {
|
| 1810 |
av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of "
|
| 1811 |
"channels changing in stream. Skipping frame.\n");
|
| 1812 |
return -1; |
| 1813 |
} |
| 1814 |
|
| 1815 |
if (*data_size < (s->sample_blocks / 8) * 256 * sizeof(int16_t) * channels) |
| 1816 |
return -1; |
| 1817 |
*data_size = 256 / 8 * s->sample_blocks * sizeof(int16_t) * channels; |
| 1818 |
|
| 1819 |
/* filter to get final output */
|
| 1820 |
for (i = 0; i < (s->sample_blocks / 8); i++) { |
| 1821 |
dca_filter_channels(s, i); |
| 1822 |
|
| 1823 |
/* If this was marked as a DTS-ES stream we need to subtract back- */
|
| 1824 |
/* channel from SL & SR to remove matrixed back-channel signal */
|
| 1825 |
if((s->source_pcm_res & 1) && s->xch_present) { |
| 1826 |
float* back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256; |
| 1827 |
float* lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256; |
| 1828 |
float* rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256; |
| 1829 |
int j;
|
| 1830 |
for(j = 0; j < 256; ++j) { |
| 1831 |
lt_chan[j] -= back_chan[j] * M_SQRT1_2; |
| 1832 |
rt_chan[j] -= back_chan[j] * M_SQRT1_2; |
| 1833 |
} |
| 1834 |
} |
| 1835 |
|
| 1836 |
s->fmt_conv.float_to_int16_interleave(samples, s->samples_chanptr, 256, channels);
|
| 1837 |
samples += 256 * channels;
|
| 1838 |
} |
| 1839 |
|
| 1840 |
/* update lfe history */
|
| 1841 |
lfe_samples = 2 * s->lfe * (s->sample_blocks / 8); |
| 1842 |
for (i = 0; i < 2 * s->lfe * 4; i++) { |
| 1843 |
s->lfe_data[i] = s->lfe_data[i + lfe_samples]; |
| 1844 |
} |
| 1845 |
|
| 1846 |
return buf_size;
|
| 1847 |
} |
| 1848 |
|
| 1849 |
|
| 1850 |
|
| 1851 |
/**
|
| 1852 |
* DCA initialization
|
| 1853 |
*
|
| 1854 |
* @param avctx pointer to the AVCodecContext
|
| 1855 |
*/
|
| 1856 |
|
| 1857 |
static av_cold int dca_decode_init(AVCodecContext * avctx) |
| 1858 |
{
|
| 1859 |
DCAContext *s = avctx->priv_data; |
| 1860 |
int i;
|
| 1861 |
|
| 1862 |
s->avctx = avctx; |
| 1863 |
dca_init_vlcs(); |
| 1864 |
|
| 1865 |
dsputil_init(&s->dsp, avctx); |
| 1866 |
ff_mdct_init(&s->imdct, 6, 1, 1.0); |
| 1867 |
ff_synth_filter_init(&s->synth); |
| 1868 |
ff_dcadsp_init(&s->dcadsp); |
| 1869 |
ff_fmt_convert_init(&s->fmt_conv, avctx); |
| 1870 |
|
| 1871 |
for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++) |
| 1872 |
s->samples_chanptr[i] = s->samples + i * 256;
|
| 1873 |
avctx->sample_fmt = AV_SAMPLE_FMT_S16; |
| 1874 |
|
| 1875 |
s->scale_bias = 1.0; |
| 1876 |
|
| 1877 |
/* allow downmixing to stereo */
|
| 1878 |
if (avctx->channels > 0 && avctx->request_channels < avctx->channels && |
| 1879 |
avctx->request_channels == 2) {
|
| 1880 |
avctx->channels = avctx->request_channels; |
| 1881 |
} |
| 1882 |
|
| 1883 |
return 0; |
| 1884 |
} |
| 1885 |
|
| 1886 |
static av_cold int dca_decode_end(AVCodecContext * avctx) |
| 1887 |
{
|
| 1888 |
DCAContext *s = avctx->priv_data; |
| 1889 |
ff_mdct_end(&s->imdct); |
| 1890 |
return 0; |
| 1891 |
} |
| 1892 |
|
| 1893 |
static const AVProfile profiles[] = { |
| 1894 |
{ FF_PROFILE_DTS, "DTS" },
|
| 1895 |
{ FF_PROFILE_DTS_ES, "DTS-ES" },
|
| 1896 |
{ FF_PROFILE_DTS_96_24, "DTS 96/24" },
|
| 1897 |
{ FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
|
| 1898 |
{ FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
|
| 1899 |
{ FF_PROFILE_UNKNOWN },
|
| 1900 |
}; |
| 1901 |
|
| 1902 |
AVCodec ff_dca_decoder = {
|
| 1903 |
.name = "dca",
|
| 1904 |
.type = AVMEDIA_TYPE_AUDIO, |
| 1905 |
.id = CODEC_ID_DTS, |
| 1906 |
.priv_data_size = sizeof(DCAContext),
|
| 1907 |
.init = dca_decode_init, |
| 1908 |
.decode = dca_decode_frame, |
| 1909 |
.close = dca_decode_end, |
| 1910 |
.long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
|
| 1911 |
.capabilities = CODEC_CAP_CHANNEL_CONF, |
| 1912 |
.profiles = NULL_IF_CONFIG_SMALL(profiles), |
| 1913 |
}; |