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
8 * This file is part of FFmpeg.
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.
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.
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
29 #include "libavutil/common.h"
30 #include "libavutil/float_dsp.h"
31 #include "libavutil/intmath.h"
32 #include "libavutil/intreadwrite.h"
33 #include "libavutil/mathematics.h"
34 #include "libavutil/audioconvert.h"
43 #include "dca_parser.h"
44 #include "synth_filter.h"
46 #include "fmtconvert.h"
54 #define DCA_PRIM_CHANNELS_MAX (7)
55 #define DCA_SUBBANDS (64)
56 #define DCA_ABITS_MAX (32) /* Should be 28 */
57 #define DCA_SUBSUBFRAMES_MAX (4)
58 #define DCA_SUBFRAMES_MAX (16)
59 #define DCA_BLOCKS_MAX (16)
60 #define DCA_LFE_MAX (3)
61 #define DCA_CHSETS_MAX (4)
62 #define DCA_CHSET_CHANS_MAX (8)
78 /* these are unconfirmed but should be mostly correct */
79 enum DCAExSSSpeakerMask {
80 DCA_EXSS_FRONT_CENTER = 0x0001,
81 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
82 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
83 DCA_EXSS_LFE = 0x0008,
84 DCA_EXSS_REAR_CENTER = 0x0010,
85 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
86 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
87 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
88 DCA_EXSS_OVERHEAD = 0x0100,
89 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
90 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
91 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
92 DCA_EXSS_LFE2 = 0x1000,
93 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
94 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
95 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
98 enum DCAExtensionMask {
99 DCA_EXT_CORE = 0x001, ///< core in core substream
100 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
101 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
102 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
103 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
104 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
105 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
106 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
107 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
108 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
111 /* -1 are reserved or unknown */
112 static const int dca_ext_audio_descr_mask[] = {
116 DCA_EXT_XCH | DCA_EXT_X96,
123 /* extensions that reside in core substream */
124 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
126 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
127 * Some compromises have been made for special configurations. Most configurations
128 * are never used so complete accuracy is not needed.
130 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
131 * S -> side, when both rear and back are configured move one of them to the side channel
133 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
135 static const uint64_t dca_core_channel_layout[] = {
136 AV_CH_FRONT_CENTER, ///< 1, A
137 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
138 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
139 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
140 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
141 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
142 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
143 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
144 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
146 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
147 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
149 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
150 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
152 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
153 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
155 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
156 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
157 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
159 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
160 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
161 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
163 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
164 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
165 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
167 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
168 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
169 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
172 static const int8_t dca_lfe_index[] = {
173 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
176 static const int8_t dca_channel_reorder_lfe[][9] = {
177 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
178 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
179 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
180 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
181 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
182 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
183 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
184 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
185 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
186 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
187 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
188 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
189 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
190 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
191 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
192 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
195 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
196 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
197 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
198 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
199 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
200 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
201 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
202 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
203 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
204 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
205 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
206 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
207 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
208 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
209 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
210 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
211 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
214 static const int8_t dca_channel_reorder_nolfe[][9] = {
215 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
216 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
217 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
218 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
219 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
220 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
221 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
222 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
223 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
224 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
225 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
226 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
227 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
228 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
229 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
230 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
233 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
234 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
235 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
236 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
237 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
238 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
239 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
240 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
241 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
242 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
243 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
244 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
245 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
246 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
247 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
248 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
249 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
252 #define DCA_DOLBY 101 /* FIXME */
254 #define DCA_CHANNEL_BITS 6
255 #define DCA_CHANNEL_MASK 0x3F
259 #define HEADER_SIZE 14
261 #define DCA_MAX_FRAME_SIZE 16384
262 #define DCA_MAX_EXSS_HEADER_SIZE 4096
264 #define DCA_BUFFER_PADDING_SIZE 1024
266 /** Bit allocation */
268 int offset; ///< code values offset
269 int maxbits[8]; ///< max bits in VLC
270 int wrap; ///< wrap for get_vlc2()
271 VLC vlc[8]; ///< actual codes
274 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
275 static BitAlloc dca_tmode; ///< transition mode VLCs
276 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
277 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
279 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
282 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
287 AVCodecContext *avctx;
290 int frame_type; ///< type of the current frame
291 int samples_deficit; ///< deficit sample count
292 int crc_present; ///< crc is present in the bitstream
293 int sample_blocks; ///< number of PCM sample blocks
294 int frame_size; ///< primary frame byte size
295 int amode; ///< audio channels arrangement
296 int sample_rate; ///< audio sampling rate
297 int bit_rate; ///< transmission bit rate
298 int bit_rate_index; ///< transmission bit rate index
300 int downmix; ///< embedded downmix enabled
301 int dynrange; ///< embedded dynamic range flag
302 int timestamp; ///< embedded time stamp flag
303 int aux_data; ///< auxiliary data flag
304 int hdcd; ///< source material is mastered in HDCD
305 int ext_descr; ///< extension audio descriptor flag
306 int ext_coding; ///< extended coding flag
307 int aspf; ///< audio sync word insertion flag
308 int lfe; ///< low frequency effects flag
309 int predictor_history; ///< predictor history flag
310 int header_crc; ///< header crc check bytes
311 int multirate_inter; ///< multirate interpolator switch
312 int version; ///< encoder software revision
313 int copy_history; ///< copy history
314 int source_pcm_res; ///< source pcm resolution
315 int front_sum; ///< front sum/difference flag
316 int surround_sum; ///< surround sum/difference flag
317 int dialog_norm; ///< dialog normalisation parameter
319 /* Primary audio coding header */
320 int subframes; ///< number of subframes
321 int total_channels; ///< number of channels including extensions
322 int prim_channels; ///< number of primary audio channels
323 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
324 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
325 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
326 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
327 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
328 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
329 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
330 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
332 /* Primary audio coding side information */
333 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
334 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
335 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
336 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
337 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
338 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
339 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
340 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
341 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
342 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
343 int dynrange_coef; ///< dynamic range coefficient
345 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
347 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
348 int lfe_scale_factor;
350 /* Subband samples history (for ADPCM) */
351 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
352 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
353 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
354 int hist_index[DCA_PRIM_CHANNELS_MAX];
355 DECLARE_ALIGNED(32, float, raXin)[32];
357 int output; ///< type of output
358 float scale_bias; ///< output scale
360 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
361 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX + 1) * 256];
362 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
364 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
365 int dca_buffer_size; ///< how much data is in the dca_buffer
367 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
369 /* Current position in DCA frame */
370 int current_subframe;
371 int current_subsubframe;
373 int core_ext_mask; ///< present extensions in the core substream
375 /* XCh extension information */
376 int xch_present; ///< XCh extension present and valid
377 int xch_base_channel; ///< index of first (only) channel containing XCH data
379 /* ExSS header parser */
380 int static_fields; ///< static fields present
381 int mix_metadata; ///< mixing metadata present
382 int num_mix_configs; ///< number of mix out configurations
383 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
387 int debug_flag; ///< used for suppressing repeated error messages output
388 AVFloatDSPContext fdsp;
390 SynthFilterContext synth;
391 DCADSPContext dcadsp;
392 FmtConvertContext fmt_conv;
395 static const uint16_t dca_vlc_offs[] = {
396 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
397 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
398 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
399 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
400 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
401 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
404 static av_cold void dca_init_vlcs(void)
406 static int vlcs_initialized = 0;
408 static VLC_TYPE dca_table[23622][2];
410 if (vlcs_initialized)
413 dca_bitalloc_index.offset = 1;
414 dca_bitalloc_index.wrap = 2;
415 for (i = 0; i < 5; i++) {
416 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
417 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
418 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
419 bitalloc_12_bits[i], 1, 1,
420 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
422 dca_scalefactor.offset = -64;
423 dca_scalefactor.wrap = 2;
424 for (i = 0; i < 5; i++) {
425 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
426 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
427 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
428 scales_bits[i], 1, 1,
429 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
431 dca_tmode.offset = 0;
433 for (i = 0; i < 4; i++) {
434 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
435 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
436 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
438 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
441 for (i = 0; i < 10; i++)
442 for (j = 0; j < 7; j++) {
443 if (!bitalloc_codes[i][j])
445 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
446 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
447 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
448 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
450 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
452 bitalloc_bits[i][j], 1, 1,
453 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
456 vlcs_initialized = 1;
459 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
462 *dst++ = get_bits(gb, bits);
465 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
468 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
469 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
470 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
472 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
473 s->prim_channels = s->total_channels;
475 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
476 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
479 for (i = base_channel; i < s->prim_channels; i++) {
480 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
481 if (s->subband_activity[i] > DCA_SUBBANDS)
482 s->subband_activity[i] = DCA_SUBBANDS;
484 for (i = base_channel; i < s->prim_channels; i++) {
485 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
486 if (s->vq_start_subband[i] > DCA_SUBBANDS)
487 s->vq_start_subband[i] = DCA_SUBBANDS;
489 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
490 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
491 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
492 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
494 /* Get codebooks quantization indexes */
496 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
497 for (j = 1; j < 11; j++)
498 for (i = base_channel; i < s->prim_channels; i++)
499 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
501 /* Get scale factor adjustment */
502 for (j = 0; j < 11; j++)
503 for (i = base_channel; i < s->prim_channels; i++)
504 s->scalefactor_adj[i][j] = 1;
506 for (j = 1; j < 11; j++)
507 for (i = base_channel; i < s->prim_channels; i++)
508 if (s->quant_index_huffman[i][j] < thr[j])
509 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
511 if (s->crc_present) {
512 /* Audio header CRC check */
513 get_bits(&s->gb, 16);
516 s->current_subframe = 0;
517 s->current_subsubframe = 0;
520 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
521 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
522 for (i = base_channel; i < s->prim_channels; i++) {
523 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
524 s->subband_activity[i]);
525 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
526 s->vq_start_subband[i]);
527 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
528 s->joint_intensity[i]);
529 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
530 s->transient_huffman[i]);
531 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
532 s->scalefactor_huffman[i]);
533 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
534 s->bitalloc_huffman[i]);
535 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
536 for (j = 0; j < 11; j++)
537 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
538 av_log(s->avctx, AV_LOG_DEBUG, "\n");
539 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
540 for (j = 0; j < 11; j++)
541 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
542 av_log(s->avctx, AV_LOG_DEBUG, "\n");
549 static int dca_parse_frame_header(DCAContext *s)
551 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
554 skip_bits_long(&s->gb, 32);
557 s->frame_type = get_bits(&s->gb, 1);
558 s->samples_deficit = get_bits(&s->gb, 5) + 1;
559 s->crc_present = get_bits(&s->gb, 1);
560 s->sample_blocks = get_bits(&s->gb, 7) + 1;
561 s->frame_size = get_bits(&s->gb, 14) + 1;
562 if (s->frame_size < 95)
563 return AVERROR_INVALIDDATA;
564 s->amode = get_bits(&s->gb, 6);
565 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
567 return AVERROR_INVALIDDATA;
568 s->bit_rate_index = get_bits(&s->gb, 5);
569 s->bit_rate = dca_bit_rates[s->bit_rate_index];
571 return AVERROR_INVALIDDATA;
573 s->downmix = get_bits(&s->gb, 1);
574 s->dynrange = get_bits(&s->gb, 1);
575 s->timestamp = get_bits(&s->gb, 1);
576 s->aux_data = get_bits(&s->gb, 1);
577 s->hdcd = get_bits(&s->gb, 1);
578 s->ext_descr = get_bits(&s->gb, 3);
579 s->ext_coding = get_bits(&s->gb, 1);
580 s->aspf = get_bits(&s->gb, 1);
581 s->lfe = get_bits(&s->gb, 2);
582 s->predictor_history = get_bits(&s->gb, 1);
584 /* TODO: check CRC */
586 s->header_crc = get_bits(&s->gb, 16);
588 s->multirate_inter = get_bits(&s->gb, 1);
589 s->version = get_bits(&s->gb, 4);
590 s->copy_history = get_bits(&s->gb, 2);
591 s->source_pcm_res = get_bits(&s->gb, 3);
592 s->front_sum = get_bits(&s->gb, 1);
593 s->surround_sum = get_bits(&s->gb, 1);
594 s->dialog_norm = get_bits(&s->gb, 4);
596 /* FIXME: channels mixing levels */
597 s->output = s->amode;
599 s->output |= DCA_LFE;
602 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
603 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
604 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
605 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
606 s->sample_blocks, s->sample_blocks * 32);
607 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
608 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
609 s->amode, dca_channels[s->amode]);
610 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
612 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
614 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
615 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
616 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
617 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
618 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
619 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
620 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
621 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
622 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
623 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
624 s->predictor_history);
625 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
626 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
628 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
629 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
630 av_log(s->avctx, AV_LOG_DEBUG,
631 "source pcm resolution: %i (%i bits/sample)\n",
632 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
633 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
634 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
635 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
636 av_log(s->avctx, AV_LOG_DEBUG, "\n");
639 /* Primary audio coding header */
640 s->subframes = get_bits(&s->gb, 4) + 1;
642 return dca_parse_audio_coding_header(s, 0);
646 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
649 /* huffman encoded */
650 value += get_bitalloc(gb, &dca_scalefactor, level);
651 value = av_clip(value, 0, (1 << log2range) - 1);
652 } else if (level < 8) {
653 if (level + 1 > log2range) {
654 skip_bits(gb, level + 1 - log2range);
655 value = get_bits(gb, log2range);
657 value = get_bits(gb, level + 1);
663 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
665 /* Primary audio coding side information */
668 if (get_bits_left(&s->gb) < 0)
669 return AVERROR_INVALIDDATA;
672 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
673 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
676 for (j = base_channel; j < s->prim_channels; j++) {
677 for (k = 0; k < s->subband_activity[j]; k++)
678 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
681 /* Get prediction codebook */
682 for (j = base_channel; j < s->prim_channels; j++) {
683 for (k = 0; k < s->subband_activity[j]; k++) {
684 if (s->prediction_mode[j][k] > 0) {
685 /* (Prediction coefficient VQ address) */
686 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
691 /* Bit allocation index */
692 for (j = base_channel; j < s->prim_channels; j++) {
693 for (k = 0; k < s->vq_start_subband[j]; k++) {
694 if (s->bitalloc_huffman[j] == 6)
695 s->bitalloc[j][k] = get_bits(&s->gb, 5);
696 else if (s->bitalloc_huffman[j] == 5)
697 s->bitalloc[j][k] = get_bits(&s->gb, 4);
698 else if (s->bitalloc_huffman[j] == 7) {
699 av_log(s->avctx, AV_LOG_ERROR,
700 "Invalid bit allocation index\n");
701 return AVERROR_INVALIDDATA;
704 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
707 if (s->bitalloc[j][k] > 26) {
708 // av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index [%i][%i] too big (%i)\n",
709 // j, k, s->bitalloc[j][k]);
710 return AVERROR_INVALIDDATA;
715 /* Transition mode */
716 for (j = base_channel; j < s->prim_channels; j++) {
717 for (k = 0; k < s->subband_activity[j]; k++) {
718 s->transition_mode[j][k] = 0;
719 if (s->subsubframes[s->current_subframe] > 1 &&
720 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
721 s->transition_mode[j][k] =
722 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
727 if (get_bits_left(&s->gb) < 0)
728 return AVERROR_INVALIDDATA;
730 for (j = base_channel; j < s->prim_channels; j++) {
731 const uint32_t *scale_table;
732 int scale_sum, log_size;
734 memset(s->scale_factor[j], 0,
735 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
737 if (s->scalefactor_huffman[j] == 6) {
738 scale_table = scale_factor_quant7;
741 scale_table = scale_factor_quant6;
745 /* When huffman coded, only the difference is encoded */
748 for (k = 0; k < s->subband_activity[j]; k++) {
749 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
750 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
751 s->scale_factor[j][k][0] = scale_table[scale_sum];
754 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
755 /* Get second scale factor */
756 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
757 s->scale_factor[j][k][1] = scale_table[scale_sum];
762 /* Joint subband scale factor codebook select */
763 for (j = base_channel; j < s->prim_channels; j++) {
764 /* Transmitted only if joint subband coding enabled */
765 if (s->joint_intensity[j] > 0)
766 s->joint_huff[j] = get_bits(&s->gb, 3);
769 if (get_bits_left(&s->gb) < 0)
770 return AVERROR_INVALIDDATA;
772 /* Scale factors for joint subband coding */
773 for (j = base_channel; j < s->prim_channels; j++) {
776 /* Transmitted only if joint subband coding enabled */
777 if (s->joint_intensity[j] > 0) {
779 source_channel = s->joint_intensity[j] - 1;
781 /* When huffman coded, only the difference is encoded
782 * (is this valid as well for joint scales ???) */
784 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
785 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
786 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
789 if (!(s->debug_flag & 0x02)) {
790 av_log(s->avctx, AV_LOG_DEBUG,
791 "Joint stereo coding not supported\n");
792 s->debug_flag |= 0x02;
797 /* Stereo downmix coefficients */
798 if (!base_channel && s->prim_channels > 2) {
800 for (j = base_channel; j < s->prim_channels; j++) {
801 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
802 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
805 int am = s->amode & DCA_CHANNEL_MASK;
806 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
807 av_log(s->avctx, AV_LOG_ERROR,
808 "Invalid channel mode %d\n", am);
809 return AVERROR_INVALIDDATA;
811 for (j = base_channel; j < FFMIN(s->prim_channels, FF_ARRAY_ELEMS(dca_default_coeffs[am])); j++) {
812 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
813 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
818 /* Dynamic range coefficient */
819 if (!base_channel && s->dynrange)
820 s->dynrange_coef = get_bits(&s->gb, 8);
822 /* Side information CRC check word */
823 if (s->crc_present) {
824 get_bits(&s->gb, 16);
828 * Primary audio data arrays
831 /* VQ encoded high frequency subbands */
832 for (j = base_channel; j < s->prim_channels; j++)
833 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
834 /* 1 vector -> 32 samples */
835 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
837 /* Low frequency effect data */
838 if (!base_channel && s->lfe) {
841 int lfe_samples = 2 * s->lfe * (4 + block_index);
842 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
845 for (j = lfe_samples; j < lfe_end_sample; j++) {
846 /* Signed 8 bits int */
847 s->lfe_data[j] = get_sbits(&s->gb, 8);
850 /* Scale factor index */
851 quant7 = get_bits(&s->gb, 8);
853 av_log_ask_for_sample(s->avctx, "LFEScaleIndex larger than 127\n");
854 return AVERROR_INVALIDDATA;
856 s->lfe_scale_factor = scale_factor_quant7[quant7];
858 /* Quantization step size * scale factor */
859 lfe_scale = 0.035 * s->lfe_scale_factor;
861 for (j = lfe_samples; j < lfe_end_sample; j++)
862 s->lfe_data[j] *= lfe_scale;
866 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
867 s->subsubframes[s->current_subframe]);
868 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
869 s->partial_samples[s->current_subframe]);
871 for (j = base_channel; j < s->prim_channels; j++) {
872 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
873 for (k = 0; k < s->subband_activity[j]; k++)
874 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
875 av_log(s->avctx, AV_LOG_DEBUG, "\n");
877 for (j = base_channel; j < s->prim_channels; j++) {
878 for (k = 0; k < s->subband_activity[j]; k++)
879 av_log(s->avctx, AV_LOG_DEBUG,
880 "prediction coefs: %f, %f, %f, %f\n",
881 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
882 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
883 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
884 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
886 for (j = base_channel; j < s->prim_channels; j++) {
887 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
888 for (k = 0; k < s->vq_start_subband[j]; k++)
889 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
890 av_log(s->avctx, AV_LOG_DEBUG, "\n");
892 for (j = base_channel; j < s->prim_channels; j++) {
893 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
894 for (k = 0; k < s->subband_activity[j]; k++)
895 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
896 av_log(s->avctx, AV_LOG_DEBUG, "\n");
898 for (j = base_channel; j < s->prim_channels; j++) {
899 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
900 for (k = 0; k < s->subband_activity[j]; k++) {
901 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
902 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
903 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
904 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
906 av_log(s->avctx, AV_LOG_DEBUG, "\n");
908 for (j = base_channel; j < s->prim_channels; j++) {
909 if (s->joint_intensity[j] > 0) {
910 int source_channel = s->joint_intensity[j] - 1;
911 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
912 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
913 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
914 av_log(s->avctx, AV_LOG_DEBUG, "\n");
917 if (!base_channel && s->prim_channels > 2 && s->downmix) {
918 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
919 for (j = 0; j < s->prim_channels; j++) {
920 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
921 dca_downmix_coeffs[s->downmix_coef[j][0]]);
922 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
923 dca_downmix_coeffs[s->downmix_coef[j][1]]);
925 av_log(s->avctx, AV_LOG_DEBUG, "\n");
927 for (j = base_channel; j < s->prim_channels; j++)
928 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
929 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
930 if (!base_channel && s->lfe) {
931 int lfe_samples = 2 * s->lfe * (4 + block_index);
932 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
934 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
935 for (j = lfe_samples; j < lfe_end_sample; j++)
936 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
937 av_log(s->avctx, AV_LOG_DEBUG, "\n");
944 static void qmf_32_subbands(DCAContext *s, int chans,
945 float samples_in[32][8], float *samples_out,
948 const float *prCoeff;
951 int sb_act = s->subband_activity[chans];
954 scale *= sqrt(1 / 8.0);
957 if (!s->multirate_inter) /* Non-perfect reconstruction */
958 prCoeff = fir_32bands_nonperfect;
959 else /* Perfect reconstruction */
960 prCoeff = fir_32bands_perfect;
962 for (i = sb_act; i < 32; i++)
965 /* Reconstructed channel sample index */
966 for (subindex = 0; subindex < 8; subindex++) {
967 /* Load in one sample from each subband and clear inactive subbands */
968 for (i = 0; i < sb_act; i++) {
969 unsigned sign = (i - 1) & 2;
970 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
971 AV_WN32A(&s->raXin[i], v);
974 s->synth.synth_filter_float(&s->imdct,
975 s->subband_fir_hist[chans],
976 &s->hist_index[chans],
977 s->subband_fir_noidea[chans], prCoeff,
978 samples_out, s->raXin, scale);
983 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
984 int num_deci_sample, float *samples_in,
985 float *samples_out, float scale)
987 /* samples_in: An array holding decimated samples.
988 * Samples in current subframe starts from samples_in[0],
989 * while samples_in[-1], samples_in[-2], ..., stores samples
990 * from last subframe as history.
992 * samples_out: An array holding interpolated samples
996 const float *prCoeff;
999 /* Select decimation filter */
1000 if (decimation_select == 1) {
1002 prCoeff = lfe_fir_128;
1005 prCoeff = lfe_fir_64;
1008 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1009 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1011 samples_out += 2 * decifactor;
1015 /* downmixing routines */
1016 #define MIX_REAR1(samples, si1, rs, coef) \
1017 samples[i] += samples[si1] * coef[rs][0]; \
1018 samples[i+256] += samples[si1] * coef[rs][1];
1020 #define MIX_REAR2(samples, si1, si2, rs, coef) \
1021 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs + 1][0]; \
1022 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs + 1][1];
1024 #define MIX_FRONT3(samples, coef) \
1025 t = samples[i + c]; \
1026 u = samples[i + l]; \
1027 v = samples[i + r]; \
1028 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1029 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1031 #define DOWNMIX_TO_STEREO(op1, op2) \
1032 for (i = 0; i < 256; i++) { \
1037 static void dca_downmix(float *samples, int srcfmt,
1038 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1039 const int8_t *channel_mapping)
1041 int c, l, r, sl, sr, s;
1044 float coef[DCA_PRIM_CHANNELS_MAX][2];
1046 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1047 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1048 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1054 case DCA_STEREO_TOTAL:
1055 case DCA_STEREO_SUMDIFF:
1057 av_log(NULL, 0, "Not implemented!\n");
1062 c = channel_mapping[0] * 256;
1063 l = channel_mapping[1] * 256;
1064 r = channel_mapping[2] * 256;
1065 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1068 s = channel_mapping[2] * 256;
1069 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef), );
1072 c = channel_mapping[0] * 256;
1073 l = channel_mapping[1] * 256;
1074 r = channel_mapping[2] * 256;
1075 s = channel_mapping[3] * 256;
1076 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1077 MIX_REAR1(samples, i + s, 3, coef));
1080 sl = channel_mapping[2] * 256;
1081 sr = channel_mapping[3] * 256;
1082 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef), );
1085 c = channel_mapping[0] * 256;
1086 l = channel_mapping[1] * 256;
1087 r = channel_mapping[2] * 256;
1088 sl = channel_mapping[3] * 256;
1089 sr = channel_mapping[4] * 256;
1090 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1091 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1097 #ifndef decode_blockcodes
1098 /* Very compact version of the block code decoder that does not use table
1099 * look-up but is slightly slower */
1100 static int decode_blockcode(int code, int levels, int *values)
1103 int offset = (levels - 1) >> 1;
1105 for (i = 0; i < 4; i++) {
1106 int div = FASTDIV(code, levels);
1107 values[i] = code - offset - div * levels;
1114 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1116 return decode_blockcode(code1, levels, values) |
1117 decode_blockcode(code2, levels, values + 4);
1121 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1122 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1124 #ifndef int8x8_fmul_int32
1125 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1127 float fscale = scale / 16.0;
1129 for (i = 0; i < 8; i++)
1130 dst[i] = src[i] * fscale;
1134 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1137 int subsubframe = s->current_subsubframe;
1139 const float *quant_step_table;
1142 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1143 LOCAL_ALIGNED_16(int, block, [8]);
1149 /* Select quantization step size table */
1150 if (s->bit_rate_index == 0x1f)
1151 quant_step_table = lossless_quant_d;
1153 quant_step_table = lossy_quant_d;
1155 for (k = base_channel; k < s->prim_channels; k++) {
1156 if (get_bits_left(&s->gb) < 0)
1157 return AVERROR_INVALIDDATA;
1159 for (l = 0; l < s->vq_start_subband[k]; l++) {
1162 /* Select the mid-tread linear quantizer */
1163 int abits = s->bitalloc[k][l];
1165 float quant_step_size = quant_step_table[abits];
1168 * Determine quantization index code book and its type
1171 /* Select quantization index code book */
1172 int sel = s->quant_index_huffman[k][abits];
1175 * Extract bits from the bit stream
1178 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1180 /* Deal with transients */
1181 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1182 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1183 s->scalefactor_adj[k][sel];
1185 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1188 int block_code1, block_code2, size, levels, err;
1190 size = abits_sizes[abits - 1];
1191 levels = abits_levels[abits - 1];
1193 block_code1 = get_bits(&s->gb, size);
1194 block_code2 = get_bits(&s->gb, size);
1195 err = decode_blockcodes(block_code1, block_code2,
1198 av_log(s->avctx, AV_LOG_ERROR,
1199 "ERROR: block code look-up failed\n");
1200 return AVERROR_INVALIDDATA;
1204 for (m = 0; m < 8; m++)
1205 block[m] = get_sbits(&s->gb, abits - 3);
1209 for (m = 0; m < 8; m++)
1210 block[m] = get_bitalloc(&s->gb,
1211 &dca_smpl_bitalloc[abits], sel);
1214 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1219 * Inverse ADPCM if in prediction mode
1221 if (s->prediction_mode[k][l]) {
1223 for (m = 0; m < 8; m++) {
1224 for (n = 1; n <= 4; n++)
1226 subband_samples[k][l][m] +=
1227 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1228 subband_samples[k][l][m - n] / 8192);
1229 else if (s->predictor_history)
1230 subband_samples[k][l][m] +=
1231 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1232 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1238 * Decode VQ encoded high frequencies
1240 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1241 /* 1 vector -> 32 samples but we only need the 8 samples
1242 * for this subsubframe. */
1243 int hfvq = s->high_freq_vq[k][l];
1245 if (!s->debug_flag & 0x01) {
1246 av_log(s->avctx, AV_LOG_DEBUG,
1247 "Stream with high frequencies VQ coding\n");
1248 s->debug_flag |= 0x01;
1251 int8x8_fmul_int32(subband_samples[k][l],
1252 &high_freq_vq[hfvq][subsubframe * 8],
1253 s->scale_factor[k][l][0]);
1257 /* Check for DSYNC after subsubframe */
1258 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1259 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1261 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1264 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1268 /* Backup predictor history for adpcm */
1269 for (k = base_channel; k < s->prim_channels; k++)
1270 for (l = 0; l < s->vq_start_subband[k]; l++)
1271 memcpy(s->subband_samples_hist[k][l],
1272 &subband_samples[k][l][4],
1273 4 * sizeof(subband_samples[0][0][0]));
1278 static int dca_filter_channels(DCAContext *s, int block_index)
1280 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1283 /* 32 subbands QMF */
1284 for (k = 0; k < s->prim_channels; k++) {
1285 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1286 0, 8388608.0, 8388608.0 };*/
1287 qmf_32_subbands(s, k, subband_samples[k],
1288 &s->samples[256 * s->channel_order_tab[k]],
1289 M_SQRT1_2 * s->scale_bias /* pcm_to_double[s->source_pcm_res] */);
1293 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1294 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1297 /* Generate LFE samples for this subsubframe FIXME!!! */
1298 if (s->output & DCA_LFE) {
1299 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1300 s->lfe_data + 2 * s->lfe * (block_index + 4),
1301 &s->samples[256 * dca_lfe_index[s->amode]],
1302 (1.0 / 256.0) * s->scale_bias);
1303 /* Outputs 20bits pcm samples */
1310 static int dca_subframe_footer(DCAContext *s, int base_channel)
1312 int aux_data_count = 0, i;
1315 * Unpack optional information
1318 /* presumably optional information only appears in the core? */
1319 if (!base_channel) {
1321 skip_bits_long(&s->gb, 32);
1324 aux_data_count = get_bits(&s->gb, 6);
1326 for (i = 0; i < aux_data_count; i++)
1327 get_bits(&s->gb, 8);
1329 if (s->crc_present && (s->downmix || s->dynrange))
1330 get_bits(&s->gb, 16);
1337 * Decode a dca frame block
1339 * @param s pointer to the DCAContext
1342 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1347 if (s->current_subframe >= s->subframes) {
1348 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1349 s->current_subframe, s->subframes);
1350 return AVERROR_INVALIDDATA;
1353 if (!s->current_subsubframe) {
1355 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1357 /* Read subframe header */
1358 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1362 /* Read subsubframe */
1364 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1366 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1370 s->current_subsubframe++;
1371 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1372 s->current_subsubframe = 0;
1373 s->current_subframe++;
1375 if (s->current_subframe >= s->subframes) {
1377 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1379 /* Read subframe footer */
1380 if ((ret = dca_subframe_footer(s, base_channel)))
1388 * Return the number of channels in an ExSS speaker mask (HD)
1390 static int dca_exss_mask2count(int mask)
1392 /* count bits that mean speaker pairs twice */
1393 return av_popcount(mask) +
1394 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1395 DCA_EXSS_FRONT_LEFT_RIGHT |
1396 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1397 DCA_EXSS_WIDE_LEFT_RIGHT |
1398 DCA_EXSS_SIDE_LEFT_RIGHT |
1399 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1400 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1401 DCA_EXSS_REAR_LEFT_RIGHT |
1402 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1406 * Skip mixing coefficients of a single mix out configuration (HD)
1408 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1412 for (i = 0; i < channels; i++) {
1413 int mix_map_mask = get_bits(gb, out_ch);
1414 int num_coeffs = av_popcount(mix_map_mask);
1415 skip_bits_long(gb, num_coeffs * 6);
1420 * Parse extension substream asset header (HD)
1422 static int dca_exss_parse_asset_header(DCAContext *s)
1424 int header_pos = get_bits_count(&s->gb);
1427 int embedded_stereo = 0;
1428 int embedded_6ch = 0;
1429 int drc_code_present;
1430 int av_uninit(extensions_mask);
1433 if (get_bits_left(&s->gb) < 16)
1436 /* We will parse just enough to get to the extensions bitmask with which
1437 * we can set the profile value. */
1439 header_size = get_bits(&s->gb, 9) + 1;
1440 skip_bits(&s->gb, 3); // asset index
1442 if (s->static_fields) {
1443 if (get_bits1(&s->gb))
1444 skip_bits(&s->gb, 4); // asset type descriptor
1445 if (get_bits1(&s->gb))
1446 skip_bits_long(&s->gb, 24); // language descriptor
1448 if (get_bits1(&s->gb)) {
1449 /* How can one fit 1024 bytes of text here if the maximum value
1450 * for the asset header size field above was 512 bytes? */
1451 int text_length = get_bits(&s->gb, 10) + 1;
1452 if (get_bits_left(&s->gb) < text_length * 8)
1454 skip_bits_long(&s->gb, text_length * 8); // info text
1457 skip_bits(&s->gb, 5); // bit resolution - 1
1458 skip_bits(&s->gb, 4); // max sample rate code
1459 channels = get_bits(&s->gb, 8) + 1;
1461 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1462 int spkr_remap_sets;
1463 int spkr_mask_size = 16;
1467 embedded_stereo = get_bits1(&s->gb);
1469 embedded_6ch = get_bits1(&s->gb);
1471 if (get_bits1(&s->gb)) {
1472 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1473 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1476 spkr_remap_sets = get_bits(&s->gb, 3);
1478 for (i = 0; i < spkr_remap_sets; i++) {
1479 /* std layout mask for each remap set */
1480 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1483 for (i = 0; i < spkr_remap_sets; i++) {
1484 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1485 if (get_bits_left(&s->gb) < 0)
1488 for (j = 0; j < num_spkrs[i]; j++) {
1489 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1490 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1491 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1496 skip_bits(&s->gb, 3); // representation type
1500 drc_code_present = get_bits1(&s->gb);
1501 if (drc_code_present)
1502 get_bits(&s->gb, 8); // drc code
1504 if (get_bits1(&s->gb))
1505 skip_bits(&s->gb, 5); // dialog normalization code
1507 if (drc_code_present && embedded_stereo)
1508 get_bits(&s->gb, 8); // drc stereo code
1510 if (s->mix_metadata && get_bits1(&s->gb)) {
1511 skip_bits(&s->gb, 1); // external mix
1512 skip_bits(&s->gb, 6); // post mix gain code
1514 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1515 skip_bits(&s->gb, 3); // drc limit
1517 skip_bits(&s->gb, 8); // custom drc code
1519 if (get_bits1(&s->gb)) // channel specific scaling
1520 for (i = 0; i < s->num_mix_configs; i++)
1521 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1523 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1525 for (i = 0; i < s->num_mix_configs; i++) {
1526 if (get_bits_left(&s->gb) < 0)
1528 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1530 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1531 if (embedded_stereo)
1532 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1536 switch (get_bits(&s->gb, 2)) {
1537 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1538 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1539 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1540 case 3: extensions_mask = 0; /* aux coding */ break;
1543 /* not parsed further, we were only interested in the extensions mask */
1545 if (get_bits_left(&s->gb) < 0)
1548 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1549 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1552 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1554 if (extensions_mask & DCA_EXT_EXSS_XLL)
1555 s->profile = FF_PROFILE_DTS_HD_MA;
1556 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1558 s->profile = FF_PROFILE_DTS_HD_HRA;
1560 if (!(extensions_mask & DCA_EXT_CORE))
1561 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1562 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1563 av_log(s->avctx, AV_LOG_WARNING,
1564 "DTS extensions detection mismatch (%d, %d)\n",
1565 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1570 static int dca_xbr_parse_frame(DCAContext *s)
1572 int scale_table_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS][2];
1573 int active_bands[DCA_CHSETS_MAX][DCA_CHSET_CHANS_MAX];
1574 int abits_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS];
1575 int anctemp[DCA_CHSET_CHANS_MAX];
1576 int chset_fsize[DCA_CHSETS_MAX];
1577 int n_xbr_ch[DCA_CHSETS_MAX];
1578 int hdr_size, num_chsets, xbr_tmode, hdr_pos;
1579 int i, j, k, l, chset, chan_base;
1581 av_log(s->avctx, AV_LOG_DEBUG, "DTS-XBR: decoding XBR extension\n");
1583 /* get bit position of sync header */
1584 hdr_pos = get_bits_count(&s->gb) - 32;
1586 hdr_size = get_bits(&s->gb, 6) + 1;
1587 num_chsets = get_bits(&s->gb, 2) + 1;
1589 for(i = 0; i < num_chsets; i++)
1590 chset_fsize[i] = get_bits(&s->gb, 14) + 1;
1592 xbr_tmode = get_bits1(&s->gb);
1594 for(i = 0; i < num_chsets; i++) {
1595 n_xbr_ch[i] = get_bits(&s->gb, 3) + 1;
1596 k = get_bits(&s->gb, 2) + 5;
1597 for(j = 0; j < n_xbr_ch[i]; j++)
1598 active_bands[i][j] = get_bits(&s->gb, k) + 1;
1601 /* skip to the end of the header */
1602 i = get_bits_count(&s->gb);
1603 if(hdr_pos + hdr_size * 8 > i)
1604 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1606 /* loop over the channel data sets */
1607 /* only decode as many channels as we've decoded base data for */
1608 for(chset = 0, chan_base = 0;
1609 chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->prim_channels;
1610 chan_base += n_xbr_ch[chset++]) {
1611 int start_posn = get_bits_count(&s->gb);
1612 int subsubframe = 0;
1615 /* loop over subframes */
1616 for (k = 0; k < (s->sample_blocks / 8); k++) {
1617 /* parse header if we're on first subsubframe of a block */
1618 if(subsubframe == 0) {
1619 /* Parse subframe header */
1620 for(i = 0; i < n_xbr_ch[chset]; i++) {
1621 anctemp[i] = get_bits(&s->gb, 2) + 2;
1624 for(i = 0; i < n_xbr_ch[chset]; i++) {
1625 get_array(&s->gb, abits_high[i], active_bands[chset][i], anctemp[i]);
1628 for(i = 0; i < n_xbr_ch[chset]; i++) {
1629 anctemp[i] = get_bits(&s->gb, 3);
1630 if(anctemp[i] < 1) {
1631 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: SYNC ERROR\n");
1632 return AVERROR_INVALIDDATA;
1636 /* generate scale factors */
1637 for(i = 0; i < n_xbr_ch[chset]; i++) {
1638 const uint32_t *scale_table;
1641 if (s->scalefactor_huffman[chan_base+i] == 6) {
1642 scale_table = scale_factor_quant7;
1644 scale_table = scale_factor_quant6;
1649 for(j = 0; j < active_bands[chset][i]; j++) {
1650 if(abits_high[i][j] > 0) {
1651 scale_table_high[i][j][0] =
1652 scale_table[get_bits(&s->gb, nbits)];
1654 if(xbr_tmode && s->transition_mode[i][j]) {
1655 scale_table_high[i][j][1] =
1656 scale_table[get_bits(&s->gb, nbits)];
1663 /* decode audio array for this block */
1664 for(i = 0; i < n_xbr_ch[chset]; i++) {
1665 for(j = 0; j < active_bands[chset][i]; j++) {
1666 const int xbr_abits = abits_high[i][j];
1667 const float quant_step_size = lossless_quant_d[xbr_abits];
1668 const int sfi = xbr_tmode && s->transition_mode[i][j] && subsubframe >= s->transition_mode[i][j];
1669 const float rscale = quant_step_size * scale_table_high[i][j][sfi];
1670 float *subband_samples = s->subband_samples[k][chan_base+i][j];
1677 get_array(&s->gb, block, 8, xbr_abits - 3);
1679 int block_code1, block_code2, size, levels, err;
1681 size = abits_sizes[xbr_abits - 1];
1682 levels = abits_levels[xbr_abits - 1];
1684 block_code1 = get_bits(&s->gb, size);
1685 block_code2 = get_bits(&s->gb, size);
1686 err = decode_blockcodes(block_code1, block_code2,
1689 av_log(s->avctx, AV_LOG_ERROR,
1690 "ERROR: DTS-XBR: block code look-up failed\n");
1691 return AVERROR_INVALIDDATA;
1695 /* scale & sum into subband */
1696 for(l = 0; l < 8; l++)
1697 subband_samples[l] += (float)block[l] * rscale;
1701 /* check DSYNC marker */
1702 if(s->aspf || subsubframe == s->subsubframes[subframe] - 1) {
1703 if(get_bits(&s->gb, 16) != 0xffff) {
1704 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: Didn't get subframe DSYNC\n");
1705 return AVERROR_INVALIDDATA;
1709 /* advance sub-sub-frame index */
1710 if(++subsubframe >= s->subsubframes[subframe]) {
1716 /* skip to next channel set */
1717 i = get_bits_count(&s->gb);
1718 if(start_posn + chset_fsize[chset] * 8 != i) {
1719 j = start_posn + chset_fsize[chset] * 8 - i;
1721 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: end of channel set,"
1722 " skipping further than expected (%d bits)\n", j);
1723 skip_bits_long(&s->gb, j);
1731 * Parse extension substream header (HD)
1733 static void dca_exss_parse_header(DCAContext *s)
1739 int active_ss_mask[8];
1745 if (get_bits_left(&s->gb) < 52)
1748 start_posn = get_bits_count(&s->gb) - 32;
1750 skip_bits(&s->gb, 8); // user data
1751 ss_index = get_bits(&s->gb, 2);
1753 blownup = get_bits1(&s->gb);
1754 hdrsize = get_bits(&s->gb, 8 + 4 * blownup) + 1; // header_size
1755 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1757 s->static_fields = get_bits1(&s->gb);
1758 if (s->static_fields) {
1759 skip_bits(&s->gb, 2); // reference clock code
1760 skip_bits(&s->gb, 3); // frame duration code
1762 if (get_bits1(&s->gb))
1763 skip_bits_long(&s->gb, 36); // timestamp
1765 /* a single stream can contain multiple audio assets that can be
1766 * combined to form multiple audio presentations */
1768 num_audiop = get_bits(&s->gb, 3) + 1;
1769 if (num_audiop > 1) {
1770 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1771 /* ignore such streams for now */
1775 num_assets = get_bits(&s->gb, 3) + 1;
1776 if (num_assets > 1) {
1777 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1778 /* ignore such streams for now */
1782 for (i = 0; i < num_audiop; i++)
1783 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1785 for (i = 0; i < num_audiop; i++)
1786 for (j = 0; j <= ss_index; j++)
1787 if (active_ss_mask[i] & (1 << j))
1788 skip_bits(&s->gb, 8); // active asset mask
1790 s->mix_metadata = get_bits1(&s->gb);
1791 if (s->mix_metadata) {
1792 int mix_out_mask_size;
1794 skip_bits(&s->gb, 2); // adjustment level
1795 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1796 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1798 for (i = 0; i < s->num_mix_configs; i++) {
1799 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1800 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1805 for (i = 0; i < num_assets; i++)
1806 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1808 for (i = 0; i < num_assets; i++) {
1809 if (dca_exss_parse_asset_header(s))
1813 /* not parsed further, we were only interested in the extensions mask
1814 * from the asset header */
1816 if(num_assets > 0) {
1817 j = get_bits_count(&s->gb);
1818 if(start_posn + hdrsize * 8 > j)
1819 skip_bits_long(&s->gb, start_posn + hdrsize * 8 - j);
1821 /* check first asset for XBR -- should also check extension mask! */
1822 mkr = get_bits_long(&s->gb, 32);
1824 if(mkr == 0x655e315e)
1825 dca_xbr_parse_frame(s);
1830 * Main frame decoding function
1831 * FIXME add arguments
1833 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1834 int *got_frame_ptr, AVPacket *avpkt)
1836 const uint8_t *buf = avpkt->data;
1837 int buf_size = avpkt->size;
1840 int num_core_channels = 0;
1843 int16_t *samples_s16;
1844 DCAContext *s = avctx->priv_data;
1851 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1852 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1853 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1854 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1855 return AVERROR_INVALIDDATA;
1858 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1859 if ((ret = dca_parse_frame_header(s)) < 0) {
1860 //seems like the frame is corrupt, try with the next one
1863 //set AVCodec values with parsed data
1864 avctx->sample_rate = s->sample_rate;
1865 avctx->bit_rate = s->bit_rate;
1867 s->profile = FF_PROFILE_DTS;
1869 for (i = 0; i < (s->sample_blocks / 8); i++) {
1870 if ((ret = dca_decode_block(s, 0, i))) {
1871 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1876 /* record number of core channels incase less than max channels are requested */
1877 num_core_channels = s->prim_channels;
1880 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1882 s->core_ext_mask = 0;
1884 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1886 /* only scan for extensions if ext_descr was unknown or indicated a
1887 * supported XCh extension */
1888 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1890 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1891 * extensions scan can fill it up */
1892 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1894 /* extensions start at 32-bit boundaries into bitstream */
1895 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1897 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1898 uint32_t bits = get_bits_long(&s->gb, 32);
1902 int ext_amode, xch_fsize;
1904 s->xch_base_channel = s->prim_channels;
1906 /* validate sync word using XCHFSIZE field */
1907 xch_fsize = show_bits(&s->gb, 10);
1908 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1909 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1912 /* skip length-to-end-of-frame field for the moment */
1913 skip_bits(&s->gb, 10);
1915 s->core_ext_mask |= DCA_EXT_XCH;
1917 /* extension amode(number of channels in extension) should be 1 */
1918 /* AFAIK XCh is not used for more channels */
1919 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1920 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1921 " supported!\n", ext_amode);
1925 /* much like core primary audio coding header */
1926 dca_parse_audio_coding_header(s, s->xch_base_channel);
1928 for (i = 0; i < (s->sample_blocks / 8); i++)
1929 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1930 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1938 /* XXCh: extended channels */
1939 /* usually found either in core or HD part in DTS-HD HRA streams,
1940 * but not in DTS-ES which contains XCh extensions instead */
1941 s->core_ext_mask |= DCA_EXT_XXCH;
1945 int fsize96 = show_bits(&s->gb, 12) + 1;
1946 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1949 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1950 get_bits_count(&s->gb));
1951 skip_bits(&s->gb, 12);
1952 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1953 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1955 s->core_ext_mask |= DCA_EXT_X96;
1960 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1963 /* no supported extensions, skip the rest of the core substream */
1964 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1967 if (s->core_ext_mask & DCA_EXT_X96)
1968 s->profile = FF_PROFILE_DTS_96_24;
1969 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1970 s->profile = FF_PROFILE_DTS_ES;
1972 /* check for ExSS (HD part) */
1973 if (s->dca_buffer_size - s->frame_size > 32 &&
1974 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1975 dca_exss_parse_header(s);
1977 avctx->profile = s->profile;
1979 channels = s->prim_channels + !!s->lfe;
1981 if (s->amode < 16) {
1982 avctx->channel_layout = dca_core_channel_layout[s->amode];
1984 if (s->xch_present && (!avctx->request_channels ||
1985 avctx->request_channels > num_core_channels + !!s->lfe)) {
1986 avctx->channel_layout |= AV_CH_BACK_CENTER;
1988 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1989 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1991 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1994 channels = num_core_channels + !!s->lfe;
1995 s->xch_present = 0; /* disable further xch processing */
1997 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1998 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
2000 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
2003 if (channels > !!s->lfe &&
2004 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
2005 return AVERROR_INVALIDDATA;
2007 if (avctx->request_channels == 2 && s->prim_channels > 2) {
2009 s->output = DCA_STEREO;
2010 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
2012 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
2013 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
2014 s->channel_order_tab = dca_channel_order_native;
2017 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
2018 return AVERROR_INVALIDDATA;
2021 if (avctx->channels != channels) {
2022 if (avctx->channels)
2023 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
2024 avctx->channels = channels;
2027 /* get output buffer */
2028 s->frame.nb_samples = 256 * (s->sample_blocks / 8);
2029 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
2030 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
2033 samples_flt = (float *) s->frame.data[0];
2034 samples_s16 = (int16_t *) s->frame.data[0];
2036 /* filter to get final output */
2037 for (i = 0; i < (s->sample_blocks / 8); i++) {
2038 dca_filter_channels(s, i);
2040 /* If this was marked as a DTS-ES stream we need to subtract back- */
2041 /* channel from SL & SR to remove matrixed back-channel signal */
2042 if ((s->source_pcm_res & 1) && s->xch_present) {
2043 float *back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
2044 float *lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
2045 float *rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
2046 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
2047 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
2050 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
2051 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
2053 samples_flt += 256 * channels;
2055 s->fmt_conv.float_to_int16_interleave(samples_s16,
2056 s->samples_chanptr, 256,
2058 samples_s16 += 256 * channels;
2062 /* update lfe history */
2063 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
2064 for (i = 0; i < 2 * s->lfe * 4; i++)
2065 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
2068 *(AVFrame *) data = s->frame;
2076 * DCA initialization
2078 * @param avctx pointer to the AVCodecContext
2081 static av_cold int dca_decode_init(AVCodecContext *avctx)
2083 DCAContext *s = avctx->priv_data;
2089 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
2090 ff_mdct_init(&s->imdct, 6, 1, 1.0);
2091 ff_synth_filter_init(&s->synth);
2092 ff_dcadsp_init(&s->dcadsp);
2093 ff_fmt_convert_init(&s->fmt_conv, avctx);
2095 for (i = 0; i < DCA_PRIM_CHANNELS_MAX + 1; i++)
2096 s->samples_chanptr[i] = s->samples + i * 256;
2098 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
2099 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
2100 s->scale_bias = 1.0 / 32768.0;
2102 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
2103 s->scale_bias = 1.0;
2106 /* allow downmixing to stereo */
2107 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
2108 avctx->request_channels == 2) {
2109 avctx->channels = avctx->request_channels;
2112 avcodec_get_frame_defaults(&s->frame);
2113 avctx->coded_frame = &s->frame;
2118 static av_cold int dca_decode_end(AVCodecContext *avctx)
2120 DCAContext *s = avctx->priv_data;
2121 ff_mdct_end(&s->imdct);
2125 static const AVProfile profiles[] = {
2126 { FF_PROFILE_DTS, "DTS" },
2127 { FF_PROFILE_DTS_ES, "DTS-ES" },
2128 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2129 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2130 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2131 { FF_PROFILE_UNKNOWN },
2134 AVCodec ff_dca_decoder = {
2136 .type = AVMEDIA_TYPE_AUDIO,
2138 .priv_data_size = sizeof(DCAContext),
2139 .init = dca_decode_init,
2140 .decode = dca_decode_frame,
2141 .close = dca_decode_end,
2142 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2143 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2144 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
2146 AV_SAMPLE_FMT_NONE },
2147 .profiles = NULL_IF_CONFIG_SMALL(profiles),