FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
dcadec.c
Go to the documentation of this file.
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 
30 #include "libavutil/common.h"
31 #include "libavutil/float_dsp.h"
32 #include "libavutil/internal.h"
33 #include "libavutil/intreadwrite.h"
34 #include "libavutil/mathematics.h"
35 #include "libavutil/opt.h"
36 #include "libavutil/samplefmt.h"
37 #include "avcodec.h"
38 #include "fft.h"
39 #include "get_bits.h"
40 #include "dcadata.h"
41 #include "dcahuff.h"
42 #include "dca.h"
43 #include "mathops.h"
44 #include "synth_filter.h"
45 #include "dcadsp.h"
46 #include "fmtconvert.h"
47 #include "internal.h"
48 
49 #if ARCH_ARM
50 # include "arm/dca.h"
51 #endif
52 
53 //#define TRACE
54 
55 #define DCA_PRIM_CHANNELS_MAX (7)
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)
63 
64 enum DCAMode {
65  DCA_MONO = 0,
76 };
77 
78 /* these are unconfirmed but should be mostly correct */
83  DCA_EXSS_LFE = 0x0008,
92  DCA_EXSS_LFE2 = 0x1000,
96 };
97 
99  DCA_XXCH_FRONT_CENTER = 0x0000001,
100  DCA_XXCH_FRONT_LEFT = 0x0000002,
101  DCA_XXCH_FRONT_RIGHT = 0x0000004,
104  DCA_XXCH_LFE1 = 0x0000020,
105  DCA_XXCH_REAR_CENTER = 0x0000040,
115  DCA_XXCH_LFE2 = 0x0010000,
118  DCA_XXCH_OVERHEAD = 0x0080000,
127 };
128 
129 static const uint32_t map_xxch_to_native[28] = {
139  AV_CH_SIDE_LEFT, /* side surround left -- dup sur side L */
140  AV_CH_SIDE_RIGHT, /* side surround right -- dup sur side R */
146  AV_CH_LOW_FREQUENCY, /* lfe2 -- duplicate lfe1 position */
147  AV_CH_FRONT_LEFT_OF_CENTER, /* side front left -- dup front cntr L */
148  AV_CH_FRONT_RIGHT_OF_CENTER,/* side front right -- dup front cntr R */
149  AV_CH_TOP_CENTER, /* overhead */
150  AV_CH_TOP_FRONT_LEFT, /* side high left -- dup */
151  AV_CH_TOP_FRONT_RIGHT, /* side high right -- dup */
155  AV_CH_BACK_CENTER, /* rear low center -- dup */
156  AV_CH_BACK_LEFT, /* rear low left -- dup */
157  AV_CH_BACK_RIGHT /* read low right -- dup */
158 };
159 
161  DCA_EXT_CORE = 0x001, ///< core in core substream
162  DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
163  DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
164  DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
165  DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
166  DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
167  DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
168  DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
169  DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
170  DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
171 };
172 
173 /* -1 are reserved or unknown */
174 static const int dca_ext_audio_descr_mask[] = {
175  DCA_EXT_XCH,
176  -1,
177  DCA_EXT_X96,
179  -1,
180  -1,
181  DCA_EXT_XXCH,
182  -1,
183 };
184 
185 /* extensions that reside in core substream */
186 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
187 
188 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
189  * Some compromises have been made for special configurations. Most configurations
190  * are never used so complete accuracy is not needed.
191  *
192  * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
193  * S -> side, when both rear and back are configured move one of them to the side channel
194  * OV -> center back
195  * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
196  */
197 static const uint64_t dca_core_channel_layout[] = {
198  AV_CH_FRONT_CENTER, ///< 1, A
199  AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
200  AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
201  AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
202  AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
203  AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
204  AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
205  AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
206  AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
207 
208  AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
209  AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
210 
211  AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
212  AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
213 
214  AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
215  AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
216 
219  AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
220 
222  AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
223  AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
224 
226  AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
227  AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
228 
230  AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
231  AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
232 };
233 
234 static const int8_t dca_lfe_index[] = {
235  1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
236 };
237 
238 static const int8_t dca_channel_reorder_lfe[][9] = {
239  { 0, -1, -1, -1, -1, -1, -1, -1, -1},
240  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
241  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
242  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
243  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
244  { 2, 0, 1, -1, -1, -1, -1, -1, -1},
245  { 0, 1, 3, -1, -1, -1, -1, -1, -1},
246  { 2, 0, 1, 4, -1, -1, -1, -1, -1},
247  { 0, 1, 3, 4, -1, -1, -1, -1, -1},
248  { 2, 0, 1, 4, 5, -1, -1, -1, -1},
249  { 3, 4, 0, 1, 5, 6, -1, -1, -1},
250  { 2, 0, 1, 4, 5, 6, -1, -1, -1},
251  { 0, 6, 4, 5, 2, 3, -1, -1, -1},
252  { 4, 2, 5, 0, 1, 6, 7, -1, -1},
253  { 5, 6, 0, 1, 7, 3, 8, 4, -1},
254  { 4, 2, 5, 0, 1, 6, 8, 7, -1},
255 };
256 
257 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
258  { 0, 2, -1, -1, -1, -1, -1, -1, -1},
259  { 0, 1, 3, -1, -1, -1, -1, -1, -1},
260  { 0, 1, 3, -1, -1, -1, -1, -1, -1},
261  { 0, 1, 3, -1, -1, -1, -1, -1, -1},
262  { 0, 1, 3, -1, -1, -1, -1, -1, -1},
263  { 2, 0, 1, 4, -1, -1, -1, -1, -1},
264  { 0, 1, 3, 4, -1, -1, -1, -1, -1},
265  { 2, 0, 1, 4, 5, -1, -1, -1, -1},
266  { 0, 1, 4, 5, 3, -1, -1, -1, -1},
267  { 2, 0, 1, 5, 6, 4, -1, -1, -1},
268  { 3, 4, 0, 1, 6, 7, 5, -1, -1},
269  { 2, 0, 1, 4, 5, 6, 7, -1, -1},
270  { 0, 6, 4, 5, 2, 3, 7, -1, -1},
271  { 4, 2, 5, 0, 1, 7, 8, 6, -1},
272  { 5, 6, 0, 1, 8, 3, 9, 4, 7},
273  { 4, 2, 5, 0, 1, 6, 9, 8, 7},
274 };
275 
276 static const int8_t dca_channel_reorder_nolfe[][9] = {
277  { 0, -1, -1, -1, -1, -1, -1, -1, -1},
278  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
279  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
280  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
281  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
282  { 2, 0, 1, -1, -1, -1, -1, -1, -1},
283  { 0, 1, 2, -1, -1, -1, -1, -1, -1},
284  { 2, 0, 1, 3, -1, -1, -1, -1, -1},
285  { 0, 1, 2, 3, -1, -1, -1, -1, -1},
286  { 2, 0, 1, 3, 4, -1, -1, -1, -1},
287  { 2, 3, 0, 1, 4, 5, -1, -1, -1},
288  { 2, 0, 1, 3, 4, 5, -1, -1, -1},
289  { 0, 5, 3, 4, 1, 2, -1, -1, -1},
290  { 3, 2, 4, 0, 1, 5, 6, -1, -1},
291  { 4, 5, 0, 1, 6, 2, 7, 3, -1},
292  { 3, 2, 4, 0, 1, 5, 7, 6, -1},
293 };
294 
295 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
296  { 0, 1, -1, -1, -1, -1, -1, -1, -1},
297  { 0, 1, 2, -1, -1, -1, -1, -1, -1},
298  { 0, 1, 2, -1, -1, -1, -1, -1, -1},
299  { 0, 1, 2, -1, -1, -1, -1, -1, -1},
300  { 0, 1, 2, -1, -1, -1, -1, -1, -1},
301  { 2, 0, 1, 3, -1, -1, -1, -1, -1},
302  { 0, 1, 2, 3, -1, -1, -1, -1, -1},
303  { 2, 0, 1, 3, 4, -1, -1, -1, -1},
304  { 0, 1, 3, 4, 2, -1, -1, -1, -1},
305  { 2, 0, 1, 4, 5, 3, -1, -1, -1},
306  { 2, 3, 0, 1, 5, 6, 4, -1, -1},
307  { 2, 0, 1, 3, 4, 5, 6, -1, -1},
308  { 0, 5, 3, 4, 1, 2, 6, -1, -1},
309  { 3, 2, 4, 0, 1, 6, 7, 5, -1},
310  { 4, 5, 0, 1, 7, 2, 8, 3, 6},
311  { 3, 2, 4, 0, 1, 5, 8, 7, 6},
312 };
313 
314 #define DCA_DOLBY 101 /* FIXME */
315 
316 #define DCA_CHANNEL_BITS 6
317 #define DCA_CHANNEL_MASK 0x3F
318 
319 #define DCA_LFE 0x80
320 
321 #define HEADER_SIZE 14
322 
323 #define DCA_MAX_FRAME_SIZE 16384
324 #define DCA_MAX_EXSS_HEADER_SIZE 4096
325 
326 #define DCA_BUFFER_PADDING_SIZE 1024
327 
328 #define DCA_NSYNCAUX 0x9A1105A0
329 
330 /** Bit allocation */
331 typedef struct {
332  int offset; ///< code values offset
333  int maxbits[8]; ///< max bits in VLC
334  int wrap; ///< wrap for get_vlc2()
335  VLC vlc[8]; ///< actual codes
336 } BitAlloc;
337 
338 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
339 static BitAlloc dca_tmode; ///< transition mode VLCs
340 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
341 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
342 
344  int idx)
345 {
346  return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
347  ba->offset;
348 }
349 
350 typedef struct {
351  const AVClass *class; ///< class for AVOptions
353  /* Frame header */
354  int frame_type; ///< type of the current frame
355  int samples_deficit; ///< deficit sample count
356  int crc_present; ///< crc is present in the bitstream
357  int sample_blocks; ///< number of PCM sample blocks
358  int frame_size; ///< primary frame byte size
359  int amode; ///< audio channels arrangement
360  int sample_rate; ///< audio sampling rate
361  int bit_rate; ///< transmission bit rate
362  int bit_rate_index; ///< transmission bit rate index
363 
364  int dynrange; ///< embedded dynamic range flag
365  int timestamp; ///< embedded time stamp flag
366  int aux_data; ///< auxiliary data flag
367  int hdcd; ///< source material is mastered in HDCD
368  int ext_descr; ///< extension audio descriptor flag
369  int ext_coding; ///< extended coding flag
370  int aspf; ///< audio sync word insertion flag
371  int lfe; ///< low frequency effects flag
372  int predictor_history; ///< predictor history flag
373  int header_crc; ///< header crc check bytes
374  int multirate_inter; ///< multirate interpolator switch
375  int version; ///< encoder software revision
376  int copy_history; ///< copy history
377  int source_pcm_res; ///< source pcm resolution
378  int front_sum; ///< front sum/difference flag
379  int surround_sum; ///< surround sum/difference flag
380  int dialog_norm; ///< dialog normalisation parameter
381 
382  /* Primary audio coding header */
383  int subframes; ///< number of subframes
384  int total_channels; ///< number of channels including extensions
385  int prim_channels; ///< number of primary audio channels
386  int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
387  int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
388  int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
389  int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
390  int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
391  int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
392  int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
393  float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
394 
395  /* Primary audio coding side information */
396  int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
397  int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
398  int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
399  int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
400  int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
401  int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
402  int32_t scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2];///< scale factors (2 if transient)
403  int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
404  int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
405  float downmix_coef[DCA_PRIM_CHANNELS_MAX + 1][2]; ///< stereo downmix coefficients
406  int dynrange_coef; ///< dynamic range coefficient
407 
408  /* Core substream's embedded downmix coefficients (cf. ETSI TS 102 114 V1.4.1)
409  * Input: primary audio channels (incl. LFE if present)
410  * Output: downmix audio channels (up to 4, no LFE) */
411  uint8_t core_downmix; ///< embedded downmix coefficients available
412  uint8_t core_downmix_amode; ///< audio channel arrangement of embedded downmix
413  uint16_t core_downmix_codes[DCA_PRIM_CHANNELS_MAX + 1][4]; ///< embedded downmix coefficients (9-bit codes)
414 
415  int32_t high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
416 
417  float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
419 
420  /* Subband samples history (for ADPCM) */
421  DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
422  DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
423  DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
424  int hist_index[DCA_PRIM_CHANNELS_MAX];
425  DECLARE_ALIGNED(32, float, raXin)[32];
426 
427  int output; ///< type of output
428 
430  float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
431  float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
434 
436  int dca_buffer_size; ///< how much data is in the dca_buffer
437 
438  const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
440  /* Current position in DCA frame */
443 
444  int core_ext_mask; ///< present extensions in the core substream
445 
446  /* XCh extension information */
447  int xch_present; ///< XCh extension present and valid
448  int xch_base_channel; ///< index of first (only) channel containing XCH data
449  int xch_disable; ///< whether the XCh extension should be decoded or not
450 
451  /* XXCH extension information */
455  uint32_t xxch_spk_masks[4]; /* speaker masks, last element is core mask */
456  int xxch_chset_nch[4];
457  float xxch_dmix_sf[DCA_CHSETS_MAX];
458 
459  uint32_t xxch_dmix_embedded; /* lower layer has mix pre-embedded, per chset */
460  float xxch_dmix_coeff[DCA_PRIM_CHANNELS_MAX][32]; /* worst case sizing */
461 
462  int8_t xxch_order_tab[32];
463  int8_t lfe_index;
464 
465  /* ExSS header parser */
466  int static_fields; ///< static fields present
467  int mix_metadata; ///< mixing metadata present
468  int num_mix_configs; ///< number of mix out configurations
469  int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
470 
471  int profile;
472 
473  int debug_flag; ///< used for suppressing repeated error messages output
479 } DCAContext;
480 
481 static float dca_dmix_code(unsigned code);
482 
483 static const uint16_t dca_vlc_offs[] = {
484  0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
485  5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
486  5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
487  7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
488  12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
489  18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
490 };
491 
492 static av_cold void dca_init_vlcs(void)
493 {
494  static int vlcs_initialized = 0;
495  int i, j, c = 14;
496  static VLC_TYPE dca_table[23622][2];
497 
498  if (vlcs_initialized)
499  return;
500 
501  dca_bitalloc_index.offset = 1;
502  dca_bitalloc_index.wrap = 2;
503  for (i = 0; i < 5; i++) {
504  dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
505  dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
506  init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
507  bitalloc_12_bits[i], 1, 1,
509  }
510  dca_scalefactor.offset = -64;
511  dca_scalefactor.wrap = 2;
512  for (i = 0; i < 5; i++) {
513  dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
514  dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
515  init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
516  scales_bits[i], 1, 1,
518  }
519  dca_tmode.offset = 0;
520  dca_tmode.wrap = 1;
521  for (i = 0; i < 4; i++) {
522  dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
523  dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
524  init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
525  tmode_bits[i], 1, 1,
527  }
528 
529  for (i = 0; i < 10; i++)
530  for (j = 0; j < 7; j++) {
531  if (!bitalloc_codes[i][j])
532  break;
533  dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
534  dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
535  dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
536  dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
537 
538  init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
539  bitalloc_sizes[i],
540  bitalloc_bits[i][j], 1, 1,
542  c++;
543  }
544  vlcs_initialized = 1;
545 }
546 
547 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
548 {
549  while (len--)
550  *dst++ = get_bits(gb, bits);
551 }
552 
553 static inline int dca_xxch2index(DCAContext *s, int xxch_ch)
554 {
555  int i, base, mask;
556 
557  /* locate channel set containing the channel */
558  for (i = -1, base = 0, mask = (s->xxch_core_spkmask & ~DCA_XXCH_LFE1);
559  i <= s->xxch_chset && !(mask & xxch_ch); mask = s->xxch_spk_masks[++i])
560  base += av_popcount(mask);
561 
562  return base + av_popcount(mask & (xxch_ch - 1));
563 }
564 
565 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel,
566  int xxch)
567 {
568  int i, j;
569  static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
570  static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
571  static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
572  int hdr_pos = 0, hdr_size = 0;
573  float scale_factor;
574  int this_chans, acc_mask;
575  int embedded_downmix;
576  int nchans, mask[8];
577  int coeff, ichan;
578 
579  /* xxch has arbitrary sized audio coding headers */
580  if (xxch) {
581  hdr_pos = get_bits_count(&s->gb);
582  hdr_size = get_bits(&s->gb, 7) + 1;
583  }
584 
585  nchans = get_bits(&s->gb, 3) + 1;
586  s->total_channels = nchans + base_channel;
588 
589  /* obtain speaker layout mask & downmix coefficients for XXCH */
590  if (xxch) {
591  acc_mask = s->xxch_core_spkmask;
592 
593  this_chans = get_bits(&s->gb, s->xxch_nbits_spk_mask - 6) << 6;
594  s->xxch_spk_masks[s->xxch_chset] = this_chans;
595  s->xxch_chset_nch[s->xxch_chset] = nchans;
596 
597  for (i = 0; i <= s->xxch_chset; i++)
598  acc_mask |= s->xxch_spk_masks[i];
599 
600  /* check for downmixing information */
601  if (get_bits1(&s->gb)) {
602  embedded_downmix = get_bits1(&s->gb);
603  coeff = get_bits(&s->gb, 6);
604 
605  if (coeff<1 || coeff>61) {
606  av_log(s->avctx, AV_LOG_ERROR, "6bit coeff %d is out of range\n", coeff);
607  return AVERROR_INVALIDDATA;
608  }
609 
610  scale_factor = -1.0f / dca_dmix_code((coeff<<2)-3);
611 
612  s->xxch_dmix_sf[s->xxch_chset] = scale_factor;
613 
614  for (i = base_channel; i < s->prim_channels; i++) {
615  mask[i] = get_bits(&s->gb, s->xxch_nbits_spk_mask);
616  }
617 
618  for (j = base_channel; j < s->prim_channels; j++) {
619  memset(s->xxch_dmix_coeff[j], 0, sizeof(s->xxch_dmix_coeff[0]));
620  s->xxch_dmix_embedded |= (embedded_downmix << j);
621  for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
622  if (mask[j] & (1 << i)) {
623  if ((1 << i) == DCA_XXCH_LFE1) {
625  "DCA-XXCH: dmix to LFE1 not supported.\n");
626  continue;
627  }
628 
629  coeff = get_bits(&s->gb, 7);
630  ichan = dca_xxch2index(s, 1 << i);
631  if ((coeff&63)<1 || (coeff&63)>61) {
632  av_log(s->avctx, AV_LOG_ERROR, "7bit coeff %d is out of range\n", coeff);
633  return AVERROR_INVALIDDATA;
634  }
635  s->xxch_dmix_coeff[j][ichan] = dca_dmix_code((coeff<<2)-3);
636  }
637  }
638  }
639  }
640  }
641 
644 
645 
646  for (i = base_channel; i < s->prim_channels; i++) {
647  s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
648  if (s->subband_activity[i] > DCA_SUBBANDS)
650  }
651  for (i = base_channel; i < s->prim_channels; i++) {
652  s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
653  if (s->vq_start_subband[i] > DCA_SUBBANDS)
655  }
656  get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
657  get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
658  get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
659  get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
660 
661  /* Get codebooks quantization indexes */
662  if (!base_channel)
663  memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
664  for (j = 1; j < 11; j++)
665  for (i = base_channel; i < s->prim_channels; i++)
666  s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
667 
668  /* Get scale factor adjustment */
669  for (j = 0; j < 11; j++)
670  for (i = base_channel; i < s->prim_channels; i++)
671  s->scalefactor_adj[i][j] = 1;
672 
673  for (j = 1; j < 11; j++)
674  for (i = base_channel; i < s->prim_channels; i++)
675  if (s->quant_index_huffman[i][j] < thr[j])
676  s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
677 
678  if (!xxch) {
679  if (s->crc_present) {
680  /* Audio header CRC check */
681  get_bits(&s->gb, 16);
682  }
683  } else {
684  /* Skip to the end of the header, also ignore CRC if present */
685  i = get_bits_count(&s->gb);
686  if (hdr_pos + 8 * hdr_size > i)
687  skip_bits_long(&s->gb, hdr_pos + 8 * hdr_size - i);
688  }
689 
690  s->current_subframe = 0;
691  s->current_subsubframe = 0;
692 
693 #ifdef TRACE
694  av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
695  av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
696  for (i = base_channel; i < s->prim_channels; i++) {
697  av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
698  s->subband_activity[i]);
699  av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
700  s->vq_start_subband[i]);
701  av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
702  s->joint_intensity[i]);
703  av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
704  s->transient_huffman[i]);
705  av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
706  s->scalefactor_huffman[i]);
707  av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
708  s->bitalloc_huffman[i]);
709  av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
710  for (j = 0; j < 11; j++)
711  av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
712  av_log(s->avctx, AV_LOG_DEBUG, "\n");
713  av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
714  for (j = 0; j < 11; j++)
715  av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
716  av_log(s->avctx, AV_LOG_DEBUG, "\n");
717  }
718 #endif
719 
720  return 0;
721 }
722 
724 {
725  init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
726 
727  /* Sync code */
728  skip_bits_long(&s->gb, 32);
729 
730  /* Frame header */
731  s->frame_type = get_bits(&s->gb, 1);
732  s->samples_deficit = get_bits(&s->gb, 5) + 1;
733  s->crc_present = get_bits(&s->gb, 1);
734  s->sample_blocks = get_bits(&s->gb, 7) + 1;
735  s->frame_size = get_bits(&s->gb, 14) + 1;
736  if (s->frame_size < 95)
737  return AVERROR_INVALIDDATA;
738  s->amode = get_bits(&s->gb, 6);
740  if (!s->sample_rate)
741  return AVERROR_INVALIDDATA;
742  s->bit_rate_index = get_bits(&s->gb, 5);
744  if (!s->bit_rate)
745  return AVERROR_INVALIDDATA;
746 
747  skip_bits1(&s->gb); // always 0 (reserved, cf. ETSI TS 102 114 V1.4.1)
748  s->dynrange = get_bits(&s->gb, 1);
749  s->timestamp = get_bits(&s->gb, 1);
750  s->aux_data = get_bits(&s->gb, 1);
751  s->hdcd = get_bits(&s->gb, 1);
752  s->ext_descr = get_bits(&s->gb, 3);
753  s->ext_coding = get_bits(&s->gb, 1);
754  s->aspf = get_bits(&s->gb, 1);
755  s->lfe = get_bits(&s->gb, 2);
756  s->predictor_history = get_bits(&s->gb, 1);
757 
758  if (s->lfe > 2) {
759  s->lfe = 0;
760  av_log(s->avctx, AV_LOG_ERROR, "Invalid LFE value: %d\n", s->lfe);
761  return AVERROR_INVALIDDATA;
762  }
763 
764  /* TODO: check CRC */
765  if (s->crc_present)
766  s->header_crc = get_bits(&s->gb, 16);
767 
768  s->multirate_inter = get_bits(&s->gb, 1);
769  s->version = get_bits(&s->gb, 4);
770  s->copy_history = get_bits(&s->gb, 2);
771  s->source_pcm_res = get_bits(&s->gb, 3);
772  s->front_sum = get_bits(&s->gb, 1);
773  s->surround_sum = get_bits(&s->gb, 1);
774  s->dialog_norm = get_bits(&s->gb, 4);
775 
776  /* FIXME: channels mixing levels */
777  s->output = s->amode;
778  if (s->lfe)
779  s->output |= DCA_LFE;
780 
781 #ifdef TRACE
782  av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
783  av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
784  av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
785  av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
786  s->sample_blocks, s->sample_blocks * 32);
787  av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
788  av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
789  s->amode, dca_channels[s->amode]);
790  av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
791  s->sample_rate);
792  av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
793  s->bit_rate);
794  av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
795  av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
796  av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
797  av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
798  av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
799  av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
800  av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
801  av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
802  av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
803  s->predictor_history);
804  av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
805  av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
806  s->multirate_inter);
807  av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
808  av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
810  "source pcm resolution: %i (%i bits/sample)\n",
812  av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
813  av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
814  av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
815  av_log(s->avctx, AV_LOG_DEBUG, "\n");
816 #endif
817 
818  /* Primary audio coding header */
819  s->subframes = get_bits(&s->gb, 4) + 1;
820 
821  return dca_parse_audio_coding_header(s, 0, 0);
822 }
823 
824 
825 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
826 {
827  if (level < 5) {
828  /* huffman encoded */
829  value += get_bitalloc(gb, &dca_scalefactor, level);
830  value = av_clip(value, 0, (1 << log2range) - 1);
831  } else if (level < 8) {
832  if (level + 1 > log2range) {
833  skip_bits(gb, level + 1 - log2range);
834  value = get_bits(gb, log2range);
835  } else {
836  value = get_bits(gb, level + 1);
837  }
838  }
839  return value;
840 }
841 
842 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
843 {
844  /* Primary audio coding side information */
845  int j, k;
846 
847  if (get_bits_left(&s->gb) < 0)
848  return AVERROR_INVALIDDATA;
849 
850  if (!base_channel) {
851  s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
852  s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
853  }
854 
855  for (j = base_channel; j < s->prim_channels; j++) {
856  for (k = 0; k < s->subband_activity[j]; k++)
857  s->prediction_mode[j][k] = get_bits(&s->gb, 1);
858  }
859 
860  /* Get prediction codebook */
861  for (j = base_channel; j < s->prim_channels; j++) {
862  for (k = 0; k < s->subband_activity[j]; k++) {
863  if (s->prediction_mode[j][k] > 0) {
864  /* (Prediction coefficient VQ address) */
865  s->prediction_vq[j][k] = get_bits(&s->gb, 12);
866  }
867  }
868  }
869 
870  /* Bit allocation index */
871  for (j = base_channel; j < s->prim_channels; j++) {
872  for (k = 0; k < s->vq_start_subband[j]; k++) {
873  if (s->bitalloc_huffman[j] == 6)
874  s->bitalloc[j][k] = get_bits(&s->gb, 5);
875  else if (s->bitalloc_huffman[j] == 5)
876  s->bitalloc[j][k] = get_bits(&s->gb, 4);
877  else if (s->bitalloc_huffman[j] == 7) {
879  "Invalid bit allocation index\n");
880  return AVERROR_INVALIDDATA;
881  } else {
882  s->bitalloc[j][k] =
883  get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
884  }
885 
886  if (s->bitalloc[j][k] > 26) {
887  av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
888  j, k, s->bitalloc[j][k]);
889  return AVERROR_INVALIDDATA;
890  }
891  }
892  }
893 
894  /* Transition mode */
895  for (j = base_channel; j < s->prim_channels; j++) {
896  for (k = 0; k < s->subband_activity[j]; k++) {
897  s->transition_mode[j][k] = 0;
898  if (s->subsubframes[s->current_subframe] > 1 &&
899  k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
900  s->transition_mode[j][k] =
901  get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
902  }
903  }
904  }
905 
906  if (get_bits_left(&s->gb) < 0)
907  return AVERROR_INVALIDDATA;
908 
909  for (j = base_channel; j < s->prim_channels; j++) {
910  const uint32_t *scale_table;
911  int scale_sum, log_size;
912 
913  memset(s->scale_factor[j], 0,
914  s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
915 
916  if (s->scalefactor_huffman[j] == 6) {
917  scale_table = scale_factor_quant7;
918  log_size = 7;
919  } else {
920  scale_table = scale_factor_quant6;
921  log_size = 6;
922  }
923 
924  /* When huffman coded, only the difference is encoded */
925  scale_sum = 0;
926 
927  for (k = 0; k < s->subband_activity[j]; k++) {
928  if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
929  scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
930  s->scale_factor[j][k][0] = scale_table[scale_sum];
931  }
932 
933  if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
934  /* Get second scale factor */
935  scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
936  s->scale_factor[j][k][1] = scale_table[scale_sum];
937  }
938  }
939  }
940 
941  /* Joint subband scale factor codebook select */
942  for (j = base_channel; j < s->prim_channels; j++) {
943  /* Transmitted only if joint subband coding enabled */
944  if (s->joint_intensity[j] > 0)
945  s->joint_huff[j] = get_bits(&s->gb, 3);
946  }
947 
948  if (get_bits_left(&s->gb) < 0)
949  return AVERROR_INVALIDDATA;
950 
951  /* Scale factors for joint subband coding */
952  for (j = base_channel; j < s->prim_channels; j++) {
953  int source_channel;
954 
955  /* Transmitted only if joint subband coding enabled */
956  if (s->joint_intensity[j] > 0) {
957  int scale = 0;
958  source_channel = s->joint_intensity[j] - 1;
959 
960  /* When huffman coded, only the difference is encoded
961  * (is this valid as well for joint scales ???) */
962 
963  for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
964  scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
965  s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
966  }
967 
968  if (!(s->debug_flag & 0x02)) {
970  "Joint stereo coding not supported\n");
971  s->debug_flag |= 0x02;
972  }
973  }
974  }
975 
976  /* Dynamic range coefficient */
977  if (!base_channel && s->dynrange)
978  s->dynrange_coef = get_bits(&s->gb, 8);
979 
980  /* Side information CRC check word */
981  if (s->crc_present) {
982  get_bits(&s->gb, 16);
983  }
984 
985  /*
986  * Primary audio data arrays
987  */
988 
989  /* VQ encoded high frequency subbands */
990  for (j = base_channel; j < s->prim_channels; j++)
991  for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
992  /* 1 vector -> 32 samples */
993  s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
994 
995  /* Low frequency effect data */
996  if (!base_channel && s->lfe) {
997  int quant7;
998  /* LFE samples */
999  int lfe_samples = 2 * s->lfe * (4 + block_index);
1000  int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
1001  float lfe_scale;
1002 
1003  for (j = lfe_samples; j < lfe_end_sample; j++) {
1004  /* Signed 8 bits int */
1005  s->lfe_data[j] = get_sbits(&s->gb, 8);
1006  }
1007 
1008  /* Scale factor index */
1009  quant7 = get_bits(&s->gb, 8);
1010  if (quant7 > 127) {
1011  avpriv_request_sample(s->avctx, "LFEScaleIndex larger than 127");
1012  return AVERROR_INVALIDDATA;
1013  }
1015 
1016  /* Quantization step size * scale factor */
1017  lfe_scale = 0.035 * s->lfe_scale_factor;
1018 
1019  for (j = lfe_samples; j < lfe_end_sample; j++)
1020  s->lfe_data[j] *= lfe_scale;
1021  }
1022 
1023 #ifdef TRACE
1024  av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
1026  av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
1028 
1029  for (j = base_channel; j < s->prim_channels; j++) {
1030  av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
1031  for (k = 0; k < s->subband_activity[j]; k++)
1032  av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
1033  av_log(s->avctx, AV_LOG_DEBUG, "\n");
1034  }
1035  for (j = base_channel; j < s->prim_channels; j++) {
1036  for (k = 0; k < s->subband_activity[j]; k++)
1038  "prediction coefs: %f, %f, %f, %f\n",
1039  (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
1040  (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
1041  (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
1042  (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
1043  }
1044  for (j = base_channel; j < s->prim_channels; j++) {
1045  av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
1046  for (k = 0; k < s->vq_start_subband[j]; k++)
1047  av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
1048  av_log(s->avctx, AV_LOG_DEBUG, "\n");
1049  }
1050  for (j = base_channel; j < s->prim_channels; j++) {
1051  av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
1052  for (k = 0; k < s->subband_activity[j]; k++)
1053  av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
1054  av_log(s->avctx, AV_LOG_DEBUG, "\n");
1055  }
1056  for (j = base_channel; j < s->prim_channels; j++) {
1057  av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
1058  for (k = 0; k < s->subband_activity[j]; k++) {
1059  if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
1060  av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
1061  if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
1062  av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
1063  }
1064  av_log(s->avctx, AV_LOG_DEBUG, "\n");
1065  }
1066  for (j = base_channel; j < s->prim_channels; j++) {
1067  if (s->joint_intensity[j] > 0) {
1068  int source_channel = s->joint_intensity[j] - 1;
1069  av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
1070  for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
1071  av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
1072  av_log(s->avctx, AV_LOG_DEBUG, "\n");
1073  }
1074  }
1075  for (j = base_channel; j < s->prim_channels; j++)
1076  for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
1077  av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
1078  if (!base_channel && s->lfe) {
1079  int lfe_samples = 2 * s->lfe * (4 + block_index);
1080  int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
1081 
1082  av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
1083  for (j = lfe_samples; j < lfe_end_sample; j++)
1084  av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
1085  av_log(s->avctx, AV_LOG_DEBUG, "\n");
1086  }
1087 #endif
1088 
1089  return 0;
1090 }
1091 
1092 static void qmf_32_subbands(DCAContext *s, int chans,
1093  float samples_in[32][8], float *samples_out,
1094  float scale)
1095 {
1096  const float *prCoeff;
1097 
1098  int sb_act = s->subband_activity[chans];
1099 
1100  scale *= sqrt(1 / 8.0);
1101 
1102  /* Select filter */
1103  if (!s->multirate_inter) /* Non-perfect reconstruction */
1104  prCoeff = fir_32bands_nonperfect;
1105  else /* Perfect reconstruction */
1106  prCoeff = fir_32bands_perfect;
1107 
1108  s->dcadsp.qmf_32_subbands(samples_in, sb_act, &s->synth, &s->imdct,
1109  s->subband_fir_hist[chans],
1110  &s->hist_index[chans],
1111  s->subband_fir_noidea[chans], prCoeff,
1112  samples_out, s->raXin, scale);
1113 }
1114 
1115 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
1116  int num_deci_sample, float *samples_in,
1117  float *samples_out)
1118 {
1119  /* samples_in: An array holding decimated samples.
1120  * Samples in current subframe starts from samples_in[0],
1121  * while samples_in[-1], samples_in[-2], ..., stores samples
1122  * from last subframe as history.
1123  *
1124  * samples_out: An array holding interpolated samples
1125  */
1126 
1127  int idx;
1128  const float *prCoeff;
1129  int deciindex;
1130 
1131  /* Select decimation filter */
1132  if (decimation_select == 1) {
1133  idx = 1;
1134  prCoeff = lfe_fir_128;
1135  } else {
1136  idx = 0;
1137  prCoeff = lfe_fir_64;
1138  }
1139  /* Interpolation */
1140  for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1141  s->dcadsp.lfe_fir[idx](samples_out, samples_in, prCoeff);
1142  samples_in++;
1143  samples_out += 2 * 32 * (1 + idx);
1144  }
1145 }
1146 
1147 /* downmixing routines */
1148 #define MIX_REAR1(samples, s1, rs, coef) \
1149  samples[0][i] += samples[s1][i] * coef[rs][0]; \
1150  samples[1][i] += samples[s1][i] * coef[rs][1];
1151 
1152 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1153  samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1154  samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1155 
1156 #define MIX_FRONT3(samples, coef) \
1157  t = samples[c][i]; \
1158  u = samples[l][i]; \
1159  v = samples[r][i]; \
1160  samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1161  samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1162 
1163 #define DOWNMIX_TO_STEREO(op1, op2) \
1164  for (i = 0; i < 256; i++) { \
1165  op1 \
1166  op2 \
1167  }
1168 
1169 static void dca_downmix(float **samples, int srcfmt, int lfe_present,
1170  float coef[DCA_PRIM_CHANNELS_MAX + 1][2],
1171  const int8_t *channel_mapping)
1172 {
1173  int c, l, r, sl, sr, s;
1174  int i;
1175  float t, u, v;
1176 
1177  switch (srcfmt) {
1178  case DCA_MONO:
1179  case DCA_4F2R:
1180  av_log(NULL, AV_LOG_ERROR, "Not implemented!\n");
1181  break;
1182  case DCA_CHANNEL:
1183  case DCA_STEREO:
1184  case DCA_STEREO_TOTAL:
1185  case DCA_STEREO_SUMDIFF:
1186  break;
1187  case DCA_3F:
1188  c = channel_mapping[0];
1189  l = channel_mapping[1];
1190  r = channel_mapping[2];
1191  DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1192  break;
1193  case DCA_2F1R:
1194  s = channel_mapping[2];
1195  DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1196  break;
1197  case DCA_3F1R:
1198  c = channel_mapping[0];
1199  l = channel_mapping[1];
1200  r = channel_mapping[2];
1201  s = channel_mapping[3];
1202  DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1203  MIX_REAR1(samples, s, 3, coef));
1204  break;
1205  case DCA_2F2R:
1206  sl = channel_mapping[2];
1207  sr = channel_mapping[3];
1208  DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1209  break;
1210  case DCA_3F2R:
1211  c = channel_mapping[0];
1212  l = channel_mapping[1];
1213  r = channel_mapping[2];
1214  sl = channel_mapping[3];
1215  sr = channel_mapping[4];
1216  DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1217  MIX_REAR2(samples, sl, sr, 3, coef));
1218  break;
1219  }
1220  if (lfe_present) {
1221  int lf_buf = dca_lfe_index[srcfmt];
1222  int lf_idx = dca_channels [srcfmt];
1223  for (i = 0; i < 256; i++) {
1224  samples[0][i] += samples[lf_buf][i] * coef[lf_idx][0];
1225  samples[1][i] += samples[lf_buf][i] * coef[lf_idx][1];
1226  }
1227  }
1228 }
1229 
1230 
1231 #ifndef decode_blockcodes
1232 /* Very compact version of the block code decoder that does not use table
1233  * look-up but is slightly slower */
1234 static int decode_blockcode(int code, int levels, int32_t *values)
1235 {
1236  int i;
1237  int offset = (levels - 1) >> 1;
1238 
1239  for (i = 0; i < 4; i++) {
1240  int div = FASTDIV(code, levels);
1241  values[i] = code - offset - div * levels;
1242  code = div;
1243  }
1244 
1245  return code;
1246 }
1247 
1248 static int decode_blockcodes(int code1, int code2, int levels, int32_t *values)
1249 {
1250  return decode_blockcode(code1, levels, values) |
1251  decode_blockcode(code2, levels, values + 4);
1252 }
1253 #endif
1254 
1255 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1256 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1257 
1258 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1259 {
1260  int k, l;
1261  int subsubframe = s->current_subsubframe;
1262 
1263  const float *quant_step_table;
1264 
1265  /* FIXME */
1266  float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1268 
1269  /*
1270  * Audio data
1271  */
1272 
1273  /* Select quantization step size table */
1274  if (s->bit_rate_index == 0x1f)
1275  quant_step_table = lossless_quant_d;
1276  else
1277  quant_step_table = lossy_quant_d;
1278 
1279  for (k = base_channel; k < s->prim_channels; k++) {
1280  float rscale[DCA_SUBBANDS];
1281 
1282  if (get_bits_left(&s->gb) < 0)
1283  return AVERROR_INVALIDDATA;
1284 
1285  for (l = 0; l < s->vq_start_subband[k]; l++) {
1286  int m;
1287 
1288  /* Select the mid-tread linear quantizer */
1289  int abits = s->bitalloc[k][l];
1290 
1291  float quant_step_size = quant_step_table[abits];
1292 
1293  /*
1294  * Determine quantization index code book and its type
1295  */
1296 
1297  /* Select quantization index code book */
1298  int sel = s->quant_index_huffman[k][abits];
1299 
1300  /*
1301  * Extract bits from the bit stream
1302  */
1303  if (!abits) {
1304  rscale[l] = 0;
1305  memset(block + 8 * l, 0, 8 * sizeof(block[0]));
1306  } else {
1307  /* Deal with transients */
1308  int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1309  rscale[l] = quant_step_size * s->scale_factor[k][l][sfi] *
1310  s->scalefactor_adj[k][sel];
1311 
1312  if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1313  if (abits <= 7) {
1314  /* Block code */
1315  int block_code1, block_code2, size, levels, err;
1316 
1317  size = abits_sizes[abits - 1];
1318  levels = abits_levels[abits - 1];
1319 
1320  block_code1 = get_bits(&s->gb, size);
1321  block_code2 = get_bits(&s->gb, size);
1322  err = decode_blockcodes(block_code1, block_code2,
1323  levels, block + 8 * l);
1324  if (err) {
1326  "ERROR: block code look-up failed\n");
1327  return AVERROR_INVALIDDATA;
1328  }
1329  } else {
1330  /* no coding */
1331  for (m = 0; m < 8; m++)
1332  block[8 * l + m] = get_sbits(&s->gb, abits - 3);
1333  }
1334  } else {
1335  /* Huffman coded */
1336  for (m = 0; m < 8; m++)
1337  block[8 * l + m] = get_bitalloc(&s->gb,
1338  &dca_smpl_bitalloc[abits], sel);
1339  }
1340 
1341  }
1342  }
1343 
1344  s->fmt_conv.int32_to_float_fmul_array8(&s->fmt_conv, subband_samples[k][0],
1345  block, rscale, 8 * s->vq_start_subband[k]);
1346 
1347  for (l = 0; l < s->vq_start_subband[k]; l++) {
1348  int m;
1349  /*
1350  * Inverse ADPCM if in prediction mode
1351  */
1352  if (s->prediction_mode[k][l]) {
1353  int n;
1354  if (s->predictor_history)
1355  subband_samples[k][l][0] += (adpcm_vb[s->prediction_vq[k][l]][0] *
1356  s->subband_samples_hist[k][l][3] +
1357  adpcm_vb[s->prediction_vq[k][l]][1] *
1358  s->subband_samples_hist[k][l][2] +
1359  adpcm_vb[s->prediction_vq[k][l]][2] *
1360  s->subband_samples_hist[k][l][1] +
1361  adpcm_vb[s->prediction_vq[k][l]][3] *
1362  s->subband_samples_hist[k][l][0]) *
1363  (1.0f / 8192);
1364  for (m = 1; m < 8; m++) {
1365  float sum = adpcm_vb[s->prediction_vq[k][l]][0] *
1366  subband_samples[k][l][m - 1];
1367  for (n = 2; n <= 4; n++)
1368  if (m >= n)
1369  sum += adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1370  subband_samples[k][l][m - n];
1371  else if (s->predictor_history)
1372  sum += adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1373  s->subband_samples_hist[k][l][m - n + 4];
1374  subband_samples[k][l][m] += sum * (1.0f / 8192);
1375  }
1376  }
1377  }
1378 
1379  /*
1380  * Decode VQ encoded high frequencies
1381  */
1382  if (s->subband_activity[k] > s->vq_start_subband[k]) {
1383  if (!(s->debug_flag & 0x01)) {
1385  "Stream with high frequencies VQ coding\n");
1386  s->debug_flag |= 0x01;
1387  }
1388  s->dcadsp.decode_hf(subband_samples[k], s->high_freq_vq[k],
1389  high_freq_vq, subsubframe * 8,
1390  s->scale_factor[k], s->vq_start_subband[k],
1391  s->subband_activity[k]);
1392  }
1393  }
1394 
1395  /* Check for DSYNC after subsubframe */
1396  if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1397  if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1398 #ifdef TRACE
1399  av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1400 #endif
1401  } else {
1402  av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1403  return AVERROR_INVALIDDATA;
1404  }
1405  }
1406 
1407  /* Backup predictor history for adpcm */
1408  for (k = base_channel; k < s->prim_channels; k++)
1409  for (l = 0; l < s->vq_start_subband[k]; l++)
1410  AV_COPY128(s->subband_samples_hist[k][l], &subband_samples[k][l][4]);
1411 
1412  return 0;
1413 }
1414 
1415 static int dca_filter_channels(DCAContext *s, int block_index)
1416 {
1417  float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1418  int k;
1419 
1420  /* 32 subbands QMF */
1421  for (k = 0; k < s->prim_channels; k++) {
1422 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1423  0, 8388608.0, 8388608.0 };*/
1424  if (s->channel_order_tab[k] >= 0)
1425  qmf_32_subbands(s, k, subband_samples[k],
1427  M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1428  }
1429 
1430  /* Generate LFE samples for this subsubframe FIXME!!! */
1431  if (s->lfe) {
1432  lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1433  s->lfe_data + 2 * s->lfe * (block_index + 4),
1434  s->samples_chanptr[s->lfe_index]);
1435  /* Outputs 20bits pcm samples */
1436  }
1437 
1438  /* Downmixing to Stereo */
1439  if (s->prim_channels + !!s->lfe > 2 &&
1442  s->channel_order_tab);
1443  }
1444 
1445  return 0;
1446 }
1447 
1448 
1449 static int dca_subframe_footer(DCAContext *s, int base_channel)
1450 {
1451  int in, out, aux_data_count, aux_data_end, reserved;
1452  uint32_t nsyncaux;
1453 
1454  /*
1455  * Unpack optional information
1456  */
1457 
1458  /* presumably optional information only appears in the core? */
1459  if (!base_channel) {
1460  if (s->timestamp)
1461  skip_bits_long(&s->gb, 32);
1462 
1463  if (s->aux_data) {
1464  aux_data_count = get_bits(&s->gb, 6);
1465 
1466  // align (32-bit)
1467  skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1468 
1469  aux_data_end = 8 * aux_data_count + get_bits_count(&s->gb);
1470 
1471  if ((nsyncaux = get_bits_long(&s->gb, 32)) != DCA_NSYNCAUX) {
1472  av_log(s->avctx, AV_LOG_ERROR, "nSYNCAUX mismatch %#"PRIx32"\n",
1473  nsyncaux);
1474  return AVERROR_INVALIDDATA;
1475  }
1476 
1477  if (get_bits1(&s->gb)) { // bAUXTimeStampFlag
1479  "Auxiliary Decode Time Stamp Flag");
1480  // align (4-bit)
1481  skip_bits(&s->gb, (-get_bits_count(&s->gb)) & 4);
1482  // 44 bits: nMSByte (8), nMarker (4), nLSByte (28), nMarker (4)
1483  skip_bits_long(&s->gb, 44);
1484  }
1485 
1486  if ((s->core_downmix = get_bits1(&s->gb))) {
1487  int am = get_bits(&s->gb, 3);
1488  switch (am) {
1489  case 0:
1491  break;
1492  case 1:
1494  break;
1495  case 2:
1497  break;
1498  case 3:
1500  break;
1501  case 4:
1503  break;
1504  case 5:
1506  break;
1507  case 6:
1509  break;
1510  default:
1512  "Invalid mode %d for embedded downmix coefficients\n",
1513  am);
1514  return AVERROR_INVALIDDATA;
1515  }
1516  for (out = 0; out < dca_channels[s->core_downmix_amode]; out++) {
1517  for (in = 0; in < s->prim_channels + !!s->lfe; in++) {
1518  uint16_t tmp = get_bits(&s->gb, 9);
1519  if ((tmp & 0xFF) > 241) {
1521  "Invalid downmix coefficient code %"PRIu16"\n",
1522  tmp);
1523  return AVERROR_INVALIDDATA;
1524  }
1525  s->core_downmix_codes[in][out] = tmp;
1526  }
1527  }
1528  }
1529 
1530  align_get_bits(&s->gb); // byte align
1531  skip_bits(&s->gb, 16); // nAUXCRC16
1532 
1533  // additional data (reserved, cf. ETSI TS 102 114 V1.4.1)
1534  if ((reserved = (aux_data_end - get_bits_count(&s->gb))) < 0) {
1536  "Overread auxiliary data by %d bits\n", -reserved);
1537  return AVERROR_INVALIDDATA;
1538  } else if (reserved) {
1540  "Core auxiliary data reserved content");
1541  skip_bits_long(&s->gb, reserved);
1542  }
1543  }
1544 
1545  if (s->crc_present && s->dynrange)
1546  get_bits(&s->gb, 16);
1547  }
1548 
1549  return 0;
1550 }
1551 
1552 /**
1553  * Decode a dca frame block
1554  *
1555  * @param s pointer to the DCAContext
1556  */
1557 
1558 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1559 {
1560  int ret;
1561 
1562  /* Sanity check */
1563  if (s->current_subframe >= s->subframes) {
1564  av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1565  s->current_subframe, s->subframes);
1566  return AVERROR_INVALIDDATA;
1567  }
1568 
1569  if (!s->current_subsubframe) {
1570 #ifdef TRACE
1571  av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1572 #endif
1573  /* Read subframe header */
1574  if ((ret = dca_subframe_header(s, base_channel, block_index)))
1575  return ret;
1576  }
1577 
1578  /* Read subsubframe */
1579 #ifdef TRACE
1580  av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1581 #endif
1582  if ((ret = dca_subsubframe(s, base_channel, block_index)))
1583  return ret;
1584 
1585  /* Update state */
1586  s->current_subsubframe++;
1588  s->current_subsubframe = 0;
1589  s->current_subframe++;
1590  }
1591  if (s->current_subframe >= s->subframes) {
1592 #ifdef TRACE
1593  av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1594 #endif
1595  /* Read subframe footer */
1596  if ((ret = dca_subframe_footer(s, base_channel)))
1597  return ret;
1598  }
1599 
1600  return 0;
1601 }
1602 
1603 /**
1604  * Return the number of channels in an ExSS speaker mask (HD)
1605  */
1606 static int dca_exss_mask2count(int mask)
1607 {
1608  /* count bits that mean speaker pairs twice */
1609  return av_popcount(mask) +
1610  av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1619 }
1620 
1621 /**
1622  * Skip mixing coefficients of a single mix out configuration (HD)
1623  */
1624 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1625 {
1626  int i;
1627 
1628  for (i = 0; i < channels; i++) {
1629  int mix_map_mask = get_bits(gb, out_ch);
1630  int num_coeffs = av_popcount(mix_map_mask);
1631  skip_bits_long(gb, num_coeffs * 6);
1632  }
1633 }
1634 
1635 /**
1636  * Parse extension substream asset header (HD)
1637  */
1639 {
1640  int header_pos = get_bits_count(&s->gb);
1641  int header_size;
1642  int channels = 0;
1643  int embedded_stereo = 0;
1644  int embedded_6ch = 0;
1645  int drc_code_present;
1646  int av_uninit(extensions_mask);
1647  int i, j;
1648 
1649  if (get_bits_left(&s->gb) < 16)
1650  return -1;
1651 
1652  /* We will parse just enough to get to the extensions bitmask with which
1653  * we can set the profile value. */
1654 
1655  header_size = get_bits(&s->gb, 9) + 1;
1656  skip_bits(&s->gb, 3); // asset index
1657 
1658  if (s->static_fields) {
1659  if (get_bits1(&s->gb))
1660  skip_bits(&s->gb, 4); // asset type descriptor
1661  if (get_bits1(&s->gb))
1662  skip_bits_long(&s->gb, 24); // language descriptor
1663 
1664  if (get_bits1(&s->gb)) {
1665  /* How can one fit 1024 bytes of text here if the maximum value
1666  * for the asset header size field above was 512 bytes? */
1667  int text_length = get_bits(&s->gb, 10) + 1;
1668  if (get_bits_left(&s->gb) < text_length * 8)
1669  return -1;
1670  skip_bits_long(&s->gb, text_length * 8); // info text
1671  }
1672 
1673  skip_bits(&s->gb, 5); // bit resolution - 1
1674  skip_bits(&s->gb, 4); // max sample rate code
1675  channels = get_bits(&s->gb, 8) + 1;
1676 
1677  if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1678  int spkr_remap_sets;
1679  int spkr_mask_size = 16;
1680  int num_spkrs[7];
1681 
1682  if (channels > 2)
1683  embedded_stereo = get_bits1(&s->gb);
1684  if (channels > 6)
1685  embedded_6ch = get_bits1(&s->gb);
1686 
1687  if (get_bits1(&s->gb)) {
1688  spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1689  skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1690  }
1691 
1692  spkr_remap_sets = get_bits(&s->gb, 3);
1693 
1694  for (i = 0; i < spkr_remap_sets; i++) {
1695  /* std layout mask for each remap set */
1696  num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1697  }
1698 
1699  for (i = 0; i < spkr_remap_sets; i++) {
1700  int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1701  if (get_bits_left(&s->gb) < 0)
1702  return -1;
1703 
1704  for (j = 0; j < num_spkrs[i]; j++) {
1705  int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1706  int num_dec_ch = av_popcount(remap_dec_ch_mask);
1707  skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1708  }
1709  }
1710 
1711  } else {
1712  skip_bits(&s->gb, 3); // representation type
1713  }
1714  }
1715 
1716  drc_code_present = get_bits1(&s->gb);
1717  if (drc_code_present)
1718  get_bits(&s->gb, 8); // drc code
1719 
1720  if (get_bits1(&s->gb))
1721  skip_bits(&s->gb, 5); // dialog normalization code
1722 
1723  if (drc_code_present && embedded_stereo)
1724  get_bits(&s->gb, 8); // drc stereo code
1725 
1726  if (s->mix_metadata && get_bits1(&s->gb)) {
1727  skip_bits(&s->gb, 1); // external mix
1728  skip_bits(&s->gb, 6); // post mix gain code
1729 
1730  if (get_bits(&s->gb, 2) != 3) // mixer drc code
1731  skip_bits(&s->gb, 3); // drc limit
1732  else
1733  skip_bits(&s->gb, 8); // custom drc code
1734 
1735  if (get_bits1(&s->gb)) // channel specific scaling
1736  for (i = 0; i < s->num_mix_configs; i++)
1737  skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1738  else
1739  skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1740 
1741  for (i = 0; i < s->num_mix_configs; i++) {
1742  if (get_bits_left(&s->gb) < 0)
1743  return -1;
1744  dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1745  if (embedded_6ch)
1747  if (embedded_stereo)
1749  }
1750  }
1751 
1752  switch (get_bits(&s->gb, 2)) {
1753  case 0: extensions_mask = get_bits(&s->gb, 12); break;
1754  case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1755  case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1756  case 3: extensions_mask = 0; /* aux coding */ break;
1757  }
1758 
1759  /* not parsed further, we were only interested in the extensions mask */
1760 
1761  if (get_bits_left(&s->gb) < 0)
1762  return -1;
1763 
1764  if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1765  av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1766  return -1;
1767  }
1768  skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1769 
1770  if (extensions_mask & DCA_EXT_EXSS_XLL)
1772  else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1775 
1776  if (!(extensions_mask & DCA_EXT_CORE))
1777  av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1778  if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1780  "DTS extensions detection mismatch (%d, %d)\n",
1781  extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1782 
1783  return 0;
1784 }
1785 
1787 {
1788  int scale_table_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS][2];
1789  int active_bands[DCA_CHSETS_MAX][DCA_CHSET_CHANS_MAX];
1790  int abits_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS];
1791  int anctemp[DCA_CHSET_CHANS_MAX];
1792  int chset_fsize[DCA_CHSETS_MAX];
1793  int n_xbr_ch[DCA_CHSETS_MAX];
1794  int hdr_size, num_chsets, xbr_tmode, hdr_pos;
1795  int i, j, k, l, chset, chan_base;
1796 
1797  av_log(s->avctx, AV_LOG_DEBUG, "DTS-XBR: decoding XBR extension\n");
1798 
1799  /* get bit position of sync header */
1800  hdr_pos = get_bits_count(&s->gb) - 32;
1801 
1802  hdr_size = get_bits(&s->gb, 6) + 1;
1803  num_chsets = get_bits(&s->gb, 2) + 1;
1804 
1805  for(i = 0; i < num_chsets; i++)
1806  chset_fsize[i] = get_bits(&s->gb, 14) + 1;
1807 
1808  xbr_tmode = get_bits1(&s->gb);
1809 
1810  for(i = 0; i < num_chsets; i++) {
1811  n_xbr_ch[i] = get_bits(&s->gb, 3) + 1;
1812  k = get_bits(&s->gb, 2) + 5;
1813  for(j = 0; j < n_xbr_ch[i]; j++)
1814  active_bands[i][j] = get_bits(&s->gb, k) + 1;
1815  }
1816 
1817  /* skip to the end of the header */
1818  i = get_bits_count(&s->gb);
1819  if(hdr_pos + hdr_size * 8 > i)
1820  skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1821 
1822  /* loop over the channel data sets */
1823  /* only decode as many channels as we've decoded base data for */
1824  for(chset = 0, chan_base = 0;
1825  chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->prim_channels;
1826  chan_base += n_xbr_ch[chset++]) {
1827  int start_posn = get_bits_count(&s->gb);
1828  int subsubframe = 0;
1829  int subframe = 0;
1830 
1831  /* loop over subframes */
1832  for (k = 0; k < (s->sample_blocks / 8); k++) {
1833  /* parse header if we're on first subsubframe of a block */
1834  if(subsubframe == 0) {
1835  /* Parse subframe header */
1836  for(i = 0; i < n_xbr_ch[chset]; i++) {
1837  anctemp[i] = get_bits(&s->gb, 2) + 2;
1838  }
1839 
1840  for(i = 0; i < n_xbr_ch[chset]; i++) {
1841  get_array(&s->gb, abits_high[i], active_bands[chset][i], anctemp[i]);
1842  }
1843 
1844  for(i = 0; i < n_xbr_ch[chset]; i++) {
1845  anctemp[i] = get_bits(&s->gb, 3);
1846  if(anctemp[i] < 1) {
1847  av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: SYNC ERROR\n");
1848  return AVERROR_INVALIDDATA;
1849  }
1850  }
1851 
1852  /* generate scale factors */
1853  for(i = 0; i < n_xbr_ch[chset]; i++) {
1854  const uint32_t *scale_table;
1855  int nbits;
1856 
1857  if (s->scalefactor_huffman[chan_base+i] == 6) {
1858  scale_table = scale_factor_quant7;
1859  } else {
1860  scale_table = scale_factor_quant6;
1861  }
1862 
1863  nbits = anctemp[i];
1864 
1865  for(j = 0; j < active_bands[chset][i]; j++) {
1866  if(abits_high[i][j] > 0) {
1867  scale_table_high[i][j][0] =
1868  scale_table[get_bits(&s->gb, nbits)];
1869 
1870  if(xbr_tmode && s->transition_mode[i][j]) {
1871  scale_table_high[i][j][1] =
1872  scale_table[get_bits(&s->gb, nbits)];
1873  }
1874  }
1875  }
1876  }
1877  }
1878 
1879  /* decode audio array for this block */
1880  for(i = 0; i < n_xbr_ch[chset]; i++) {
1881  for(j = 0; j < active_bands[chset][i]; j++) {
1882  const int xbr_abits = abits_high[i][j];
1883  const float quant_step_size = lossless_quant_d[xbr_abits];
1884  const int sfi = xbr_tmode && s->transition_mode[i][j] && subsubframe >= s->transition_mode[i][j];
1885  const float rscale = quant_step_size * scale_table_high[i][j][sfi];
1886  float *subband_samples = s->subband_samples[k][chan_base+i][j];
1887  int block[8];
1888 
1889  if(xbr_abits <= 0)
1890  continue;
1891 
1892  if(xbr_abits > 7) {
1893  get_array(&s->gb, block, 8, xbr_abits - 3);
1894  } else {
1895  int block_code1, block_code2, size, levels, err;
1896 
1897  size = abits_sizes[xbr_abits - 1];
1898  levels = abits_levels[xbr_abits - 1];
1899 
1900  block_code1 = get_bits(&s->gb, size);
1901  block_code2 = get_bits(&s->gb, size);
1902  err = decode_blockcodes(block_code1, block_code2,
1903  levels, block);
1904  if (err) {
1906  "ERROR: DTS-XBR: block code look-up failed\n");
1907  return AVERROR_INVALIDDATA;
1908  }
1909  }
1910 
1911  /* scale & sum into subband */
1912  for(l = 0; l < 8; l++)
1913  subband_samples[l] += (float)block[l] * rscale;
1914  }
1915  }
1916 
1917  /* check DSYNC marker */
1918  if(s->aspf || subsubframe == s->subsubframes[subframe] - 1) {
1919  if(get_bits(&s->gb, 16) != 0xffff) {
1920  av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: Didn't get subframe DSYNC\n");
1921  return AVERROR_INVALIDDATA;
1922  }
1923  }
1924 
1925  /* advance sub-sub-frame index */
1926  if(++subsubframe >= s->subsubframes[subframe]) {
1927  subsubframe = 0;
1928  subframe++;
1929  }
1930  }
1931 
1932  /* skip to next channel set */
1933  i = get_bits_count(&s->gb);
1934  if(start_posn + chset_fsize[chset] * 8 != i) {
1935  j = start_posn + chset_fsize[chset] * 8 - i;
1936  if(j < 0 || j >= 8)
1937  av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: end of channel set,"
1938  " skipping further than expected (%d bits)\n", j);
1939  skip_bits_long(&s->gb, j);
1940  }
1941  }
1942 
1943  return 0;
1944 }
1945 
1946 /* parse initial header for XXCH and dump details */
1948 {
1949  int hdr_size, spkmsk_bits, num_chsets, core_spk, hdr_pos;
1950  int i, chset, base_channel, chstart, fsize[8];
1951 
1952  /* assume header word has already been parsed */
1953  hdr_pos = get_bits_count(&s->gb) - 32;
1954  hdr_size = get_bits(&s->gb, 6) + 1;
1955  /*chhdr_crc =*/ skip_bits1(&s->gb);
1956  spkmsk_bits = get_bits(&s->gb, 5) + 1;
1957  num_chsets = get_bits(&s->gb, 2) + 1;
1958 
1959  for (i = 0; i < num_chsets; i++)
1960  fsize[i] = get_bits(&s->gb, 14) + 1;
1961 
1962  core_spk = get_bits(&s->gb, spkmsk_bits);
1963  s->xxch_core_spkmask = core_spk;
1964  s->xxch_nbits_spk_mask = spkmsk_bits;
1965  s->xxch_dmix_embedded = 0;
1966 
1967  /* skip to the end of the header */
1968  i = get_bits_count(&s->gb);
1969  if (hdr_pos + hdr_size * 8 > i)
1970  skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1971 
1972  for (chset = 0; chset < num_chsets; chset++) {
1973  chstart = get_bits_count(&s->gb);
1974  base_channel = s->prim_channels;
1975  s->xxch_chset = chset;
1976 
1977  /* XXCH and Core headers differ, see 6.4.2 "XXCH Channel Set Header" vs.
1978  5.3.2 "Primary Audio Coding Header", DTS Spec 1.3.1 */
1979  dca_parse_audio_coding_header(s, base_channel, 1);
1980 
1981  /* decode channel data */
1982  for (i = 0; i < (s->sample_blocks / 8); i++) {
1983  if (dca_decode_block(s, base_channel, i)) {
1985  "Error decoding DTS-XXCH extension\n");
1986  continue;
1987  }
1988  }
1989 
1990  /* skip to end of this section */
1991  i = get_bits_count(&s->gb);
1992  if (chstart + fsize[chset] * 8 > i)
1993  skip_bits_long(&s->gb, chstart + fsize[chset] * 8 - i);
1994  }
1995  s->xxch_chset = num_chsets;
1996 
1997  return 0;
1998 }
1999 
2000 /**
2001  * Parse extension substream header (HD)
2002  */
2004 {
2005  int asset_size[8];
2006  int ss_index;
2007  int blownup;
2008  int num_audiop = 1;
2009  int num_assets = 1;
2010  int active_ss_mask[8];
2011  int i, j;
2012  int start_posn;
2013  int hdrsize;
2014  uint32_t mkr;
2015 
2016  if (get_bits_left(&s->gb) < 52)
2017  return;
2018 
2019  start_posn = get_bits_count(&s->gb) - 32;
2020 
2021  skip_bits(&s->gb, 8); // user data
2022  ss_index = get_bits(&s->gb, 2);
2023 
2024  blownup = get_bits1(&s->gb);
2025  hdrsize = get_bits(&s->gb, 8 + 4 * blownup) + 1; // header_size
2026  skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
2027 
2028  s->static_fields = get_bits1(&s->gb);
2029  if (s->static_fields) {
2030  skip_bits(&s->gb, 2); // reference clock code
2031  skip_bits(&s->gb, 3); // frame duration code
2032 
2033  if (get_bits1(&s->gb))
2034  skip_bits_long(&s->gb, 36); // timestamp
2035 
2036  /* a single stream can contain multiple audio assets that can be
2037  * combined to form multiple audio presentations */
2038 
2039  num_audiop = get_bits(&s->gb, 3) + 1;
2040  if (num_audiop > 1) {
2042  "Multiple DTS-HD audio presentations");
2043  /* ignore such streams for now */
2044  return;
2045  }
2046 
2047  num_assets = get_bits(&s->gb, 3) + 1;
2048  if (num_assets > 1) {
2049  avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
2050  /* ignore such streams for now */
2051  return;
2052  }
2053 
2054  for (i = 0; i < num_audiop; i++)
2055  active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
2056 
2057  for (i = 0; i < num_audiop; i++)
2058  for (j = 0; j <= ss_index; j++)
2059  if (active_ss_mask[i] & (1 << j))
2060  skip_bits(&s->gb, 8); // active asset mask
2061 
2062  s->mix_metadata = get_bits1(&s->gb);
2063  if (s->mix_metadata) {
2064  int mix_out_mask_size;
2065 
2066  skip_bits(&s->gb, 2); // adjustment level
2067  mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
2068  s->num_mix_configs = get_bits(&s->gb, 2) + 1;
2069 
2070  for (i = 0; i < s->num_mix_configs; i++) {
2071  int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
2072  s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
2073  }
2074  }
2075  }
2076 
2077  av_assert0(num_assets > 0); // silence a warning
2078 
2079  for (i = 0; i < num_assets; i++)
2080  asset_size[i] = get_bits_long(&s->gb, 16 + 4 * blownup);
2081 
2082  for (i = 0; i < num_assets; i++) {
2084  return;
2085  }
2086 
2087  /* not parsed further, we were only interested in the extensions mask
2088  * from the asset header */
2089 
2090  j = get_bits_count(&s->gb);
2091  if (start_posn + hdrsize * 8 > j)
2092  skip_bits_long(&s->gb, start_posn + hdrsize * 8 - j);
2093 
2094  for (i = 0; i < num_assets; i++) {
2095  start_posn = get_bits_count(&s->gb);
2096  mkr = get_bits_long(&s->gb, 32);
2097 
2098  /* parse extensions that we know about */
2099  if (mkr == 0x655e315e) {
2101  } else if (mkr == 0x47004a03) {
2103  s->core_ext_mask |= DCA_EXT_XXCH; /* xxx use for chan reordering */
2104  } else {
2106  "DTS-ExSS: unknown marker = 0x%08x\n", mkr);
2107  }
2108 
2109  /* skip to end of block */
2110  j = get_bits_count(&s->gb);
2111  if (start_posn + asset_size[i] * 8 > j)
2112  skip_bits_long(&s->gb, start_posn + asset_size[i] * 8 - j);
2113  }
2114 }
2115 
2116 static float dca_dmix_code(unsigned code)
2117 {
2118  int sign = (code >> 8) - 1;
2119  code &= 0xff;
2120  return ((dca_dmixtable[code] ^ sign) - sign) * (1.0 / (1 << 15));
2121 }
2122 
2123 /**
2124  * Main frame decoding function
2125  * FIXME add arguments
2126  */
2127 static int dca_decode_frame(AVCodecContext *avctx, void *data,
2128  int *got_frame_ptr, AVPacket *avpkt)
2129 {
2130  AVFrame *frame = data;
2131  const uint8_t *buf = avpkt->data;
2132  int buf_size = avpkt->size;
2133  int channel_mask;
2134  int channel_layout;
2135  int lfe_samples;
2136  int num_core_channels = 0;
2137  int i, ret;
2138  float **samples_flt;
2139  float *src_chan;
2140  float *dst_chan;
2141  DCAContext *s = avctx->priv_data;
2142  int core_ss_end;
2143  int channels, full_channels;
2144  float scale;
2145  int achan;
2146  int chset;
2147  int mask;
2148  int lavc;
2149  int posn;
2150  int j, k;
2151  int endch;
2152 
2153  s->xch_present = 0;
2154 
2158  av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
2159  return AVERROR_INVALIDDATA;
2160  }
2161 
2162  if ((ret = dca_parse_frame_header(s)) < 0) {
2163  //seems like the frame is corrupt, try with the next one
2164  return ret;
2165  }
2166  //set AVCodec values with parsed data
2167  avctx->sample_rate = s->sample_rate;
2168  avctx->bit_rate = s->bit_rate;
2169 
2170  s->profile = FF_PROFILE_DTS;
2171 
2172  for (i = 0; i < (s->sample_blocks / 8); i++) {
2173  if ((ret = dca_decode_block(s, 0, i))) {
2174  av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
2175  return ret;
2176  }
2177  }
2178 
2179  /* record number of core channels incase less than max channels are requested */
2180  num_core_channels = s->prim_channels;
2181 
2182  if (s->prim_channels + !!s->lfe > 2 &&
2184  /* Stereo downmix coefficients
2185  *
2186  * The decoder can only downmix to 2-channel, so we need to ensure
2187  * embedded downmix coefficients are actually targeting 2-channel.
2188  */
2189  if (s->core_downmix && (s->core_downmix_amode == DCA_STEREO ||
2191  for (i = 0; i < num_core_channels + !!s->lfe; i++) {
2192  /* Range checked earlier */
2193  s->downmix_coef[i][0] = dca_dmix_code(s->core_downmix_codes[i][0]);
2194  s->downmix_coef[i][1] = dca_dmix_code(s->core_downmix_codes[i][1]);
2195  }
2196  s->output = s->core_downmix_amode;
2197  } else {
2198  int am = s->amode & DCA_CHANNEL_MASK;
2199  if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
2201  "Invalid channel mode %d\n", am);
2202  return AVERROR_INVALIDDATA;
2203  }
2204  if (num_core_channels + !!s->lfe >
2206  avpriv_request_sample(s->avctx, "Downmixing %d channels",
2207  s->prim_channels + !!s->lfe);
2208  return AVERROR_PATCHWELCOME;
2209  }
2210  for (i = 0; i < num_core_channels + !!s->lfe; i++) {
2211  s->downmix_coef[i][0] = dca_default_coeffs[am][i][0];
2212  s->downmix_coef[i][1] = dca_default_coeffs[am][i][1];
2213  }
2214  }
2215  av_dlog(s->avctx, "Stereo downmix coeffs:\n");
2216  for (i = 0; i < num_core_channels + !!s->lfe; i++) {
2217  av_dlog(s->avctx, "L, input channel %d = %f\n", i,
2218  s->downmix_coef[i][0]);
2219  av_dlog(s->avctx, "R, input channel %d = %f\n", i,
2220  s->downmix_coef[i][1]);
2221  }
2222  av_dlog(s->avctx, "\n");
2223  }
2224 
2225  if (s->ext_coding)
2227  else
2228  s->core_ext_mask = 0;
2229 
2230  core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
2231 
2232  /* only scan for extensions if ext_descr was unknown or indicated a
2233  * supported XCh extension */
2234  if (s->core_ext_mask < 0 || s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH)) {
2235 
2236  /* if ext_descr was unknown, clear s->core_ext_mask so that the
2237  * extensions scan can fill it up */
2238  s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
2239 
2240  /* extensions start at 32-bit boundaries into bitstream */
2241  skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2242 
2243  while (core_ss_end - get_bits_count(&s->gb) >= 32) {
2244  uint32_t bits = get_bits_long(&s->gb, 32);
2245 
2246  switch (bits) {
2247  case 0x5a5a5a5a: {
2248  int ext_amode, xch_fsize;
2249 
2251 
2252  /* validate sync word using XCHFSIZE field */
2253  xch_fsize = show_bits(&s->gb, 10);
2254  if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
2255  (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
2256  continue;
2257 
2258  /* skip length-to-end-of-frame field for the moment */
2259  skip_bits(&s->gb, 10);
2260 
2261  s->core_ext_mask |= DCA_EXT_XCH;
2262 
2263  /* extension amode(number of channels in extension) should be 1 */
2264  /* AFAIK XCh is not used for more channels */
2265  if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
2266  av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
2267  " supported!\n", ext_amode);
2268  continue;
2269  }
2270 
2271  if (s->xch_base_channel < 2) {
2272  avpriv_request_sample(avctx, "XCh with fewer than 2 base channels");
2273  continue;
2274  }
2275 
2276  /* much like core primary audio coding header */
2278 
2279  for (i = 0; i < (s->sample_blocks / 8); i++)
2280  if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
2281  av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
2282  continue;
2283  }
2284 
2285  s->xch_present = 1;
2286  break;
2287  }
2288  case 0x47004a03:
2289  /* XXCh: extended channels */
2290  /* usually found either in core or HD part in DTS-HD HRA streams,
2291  * but not in DTS-ES which contains XCh extensions instead */
2294  break;
2295 
2296  case 0x1d95f262: {
2297  int fsize96 = show_bits(&s->gb, 12) + 1;
2298  if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
2299  continue;
2300 
2301  av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
2302  get_bits_count(&s->gb));
2303  skip_bits(&s->gb, 12);
2304  av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
2305  av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
2306 
2307  s->core_ext_mask |= DCA_EXT_X96;
2308  break;
2309  }
2310  }
2311 
2312  skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2313  }
2314  } else {
2315  /* no supported extensions, skip the rest of the core substream */
2316  skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
2317  }
2318 
2319  if (s->core_ext_mask & DCA_EXT_X96)
2321  else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
2323 
2324  /* check for ExSS (HD part) */
2325  if (s->dca_buffer_size - s->frame_size > 32 &&
2326  get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
2328 
2329  avctx->profile = s->profile;
2330 
2331  full_channels = channels = s->prim_channels + !!s->lfe;
2332 
2333  /* If we have XXCH then the channel layout is managed differently */
2334  /* note that XLL will also have another way to do things */
2335  if (!(s->core_ext_mask & DCA_EXT_XXCH)
2336  || (s->core_ext_mask & DCA_EXT_XXCH && avctx->request_channels > 0
2337  && avctx->request_channels
2338  < num_core_channels + !!s->lfe + s->xxch_chset_nch[0]))
2339  { /* xxx should also do MA extensions */
2340  if (s->amode < 16) {
2342 
2343  if (s->prim_channels + !!s->lfe > 2 &&
2345  /*
2346  * Neither the core's auxiliary data nor our default tables contain
2347  * downmix coefficients for the additional channel coded in the XCh
2348  * extension, so when we're doing a Stereo downmix, don't decode it.
2349  */
2350  s->xch_disable = 1;
2351  }
2352 
2353 #if FF_API_REQUEST_CHANNELS
2355  if (s->xch_present && !s->xch_disable &&
2356  (!avctx->request_channels ||
2357  avctx->request_channels > num_core_channels + !!s->lfe)) {
2359 #else
2360  if (s->xch_present && !s->xch_disable) {
2361 #endif
2363  if (s->lfe) {
2366  } else {
2368  }
2369  if (s->channel_order_tab[s->xch_base_channel] < 0)
2370  return AVERROR_INVALIDDATA;
2371  } else {
2372  channels = num_core_channels + !!s->lfe;
2373  s->xch_present = 0; /* disable further xch processing */
2374  if (s->lfe) {
2377  } else
2379  }
2380 
2381  if (channels > !!s->lfe &&
2382  s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
2383  return AVERROR_INVALIDDATA;
2384 
2385  if (av_get_channel_layout_nb_channels(avctx->channel_layout) != channels) {
2386  av_log(avctx, AV_LOG_ERROR, "Number of channels %d mismatches layout %d\n", channels, av_get_channel_layout_nb_channels(avctx->channel_layout));
2387  return AVERROR_INVALIDDATA;
2388  }
2389 
2390  if (num_core_channels + !!s->lfe > 2 &&
2392  channels = 2;
2393  s->output = s->prim_channels == 2 ? s->amode : DCA_STEREO;
2395  }
2396  else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
2397  static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
2398  s->channel_order_tab = dca_channel_order_native;
2399  }
2400  s->lfe_index = dca_lfe_index[s->amode];
2401  } else {
2402  av_log(avctx, AV_LOG_ERROR,
2403  "Non standard configuration %d !\n", s->amode);
2404  return AVERROR_INVALIDDATA;
2405  }
2406 
2407  s->xxch_dmix_embedded = 0;
2408  } else {
2409  /* we only get here if an XXCH channel set can be added to the mix */
2410  channel_mask = s->xxch_core_spkmask;
2411 
2412  if (avctx->request_channels > 0
2413  && avctx->request_channels < s->prim_channels) {
2414  channels = num_core_channels + !!s->lfe;
2415  for (i = 0; i < s->xxch_chset && channels + s->xxch_chset_nch[i]
2416  <= avctx->request_channels; i++) {
2417  channels += s->xxch_chset_nch[i];
2418  channel_mask |= s->xxch_spk_masks[i];
2419  }
2420  } else {
2421  channels = s->prim_channels + !!s->lfe;
2422  for (i = 0; i < s->xxch_chset; i++) {
2423  channel_mask |= s->xxch_spk_masks[i];
2424  }
2425  }
2426 
2427  /* Given the DTS spec'ed channel mask, generate an avcodec version */
2428  channel_layout = 0;
2429  for (i = 0; i < s->xxch_nbits_spk_mask; ++i) {
2430  if (channel_mask & (1 << i)) {
2431  channel_layout |= map_xxch_to_native[i];
2432  }
2433  }
2434 
2435  /* make sure that we have managed to get equivalent dts/avcodec channel
2436  * masks in some sense -- unfortunately some channels could overlap */
2437  if (av_popcount(channel_mask) != av_popcount(channel_layout)) {
2438  av_log(avctx, AV_LOG_DEBUG,
2439  "DTS-XXCH: Inconsistent avcodec/dts channel layouts\n");
2440  return AVERROR_INVALIDDATA;
2441  }
2442 
2443  avctx->channel_layout = channel_layout;
2444 
2445  if (!(avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE)) {
2446  /* Estimate DTS --> avcodec ordering table */
2447  for (chset = -1, j = 0; chset < s->xxch_chset; ++chset) {
2448  mask = chset >= 0 ? s->xxch_spk_masks[chset]
2449  : s->xxch_core_spkmask;
2450  for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
2451  if (mask & ~(DCA_XXCH_LFE1 | DCA_XXCH_LFE2) & (1 << i)) {
2452  lavc = map_xxch_to_native[i];
2453  posn = av_popcount(channel_layout & (lavc - 1));
2454  s->xxch_order_tab[j++] = posn;
2455  }
2456  }
2457 
2458  }
2459 
2460  s->lfe_index = av_popcount(channel_layout & (AV_CH_LOW_FREQUENCY-1));
2461  } else { /* native ordering */
2462  for (i = 0; i < channels; i++)
2463  s->xxch_order_tab[i] = i;
2464 
2465  s->lfe_index = channels - 1;
2466  }
2467 
2469  }
2470 
2471  if (avctx->channels != channels) {
2472  if (avctx->channels)
2473  av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
2474  avctx->channels = channels;
2475  }
2476 
2477  /* get output buffer */
2478  frame->nb_samples = 256 * (s->sample_blocks / 8);
2479  if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
2480  return ret;
2481  samples_flt = (float **)frame->extended_data;
2482 
2483  /* allocate buffer for extra channels if downmixing */
2484  if (avctx->channels < full_channels) {
2485  ret = av_samples_get_buffer_size(NULL, full_channels - channels,
2486  frame->nb_samples,
2487  avctx->sample_fmt, 0);
2488  if (ret < 0)
2489  return ret;
2490 
2492  &s->extra_channels_buffer_size, ret);
2493  if (!s->extra_channels_buffer)
2494  return AVERROR(ENOMEM);
2495 
2496  ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
2498  full_channels - channels,
2499  frame->nb_samples, avctx->sample_fmt, 0);
2500  if (ret < 0)
2501  return ret;
2502  }
2503 
2504  /* filter to get final output */
2505  for (i = 0; i < (s->sample_blocks / 8); i++) {
2506  int ch;
2507 
2508  for (ch = 0; ch < channels; ch++)
2509  s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
2510  for (; ch < full_channels; ch++)
2511  s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
2512 
2513  dca_filter_channels(s, i);
2514 
2515  /* If this was marked as a DTS-ES stream we need to subtract back- */
2516  /* channel from SL & SR to remove matrixed back-channel signal */
2517  if ((s->source_pcm_res & 1) && s->xch_present) {
2518  float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
2519  float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
2520  float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
2521  s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
2522  s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
2523  }
2524 
2525  /* If stream contains XXCH, we might need to undo an embedded downmix */
2526  if (s->xxch_dmix_embedded) {
2527  /* Loop over channel sets in turn */
2528  ch = num_core_channels;
2529  for (chset = 0; chset < s->xxch_chset; chset++) {
2530  endch = ch + s->xxch_chset_nch[chset];
2531  mask = s->xxch_dmix_embedded;
2532 
2533  /* undo downmix */
2534  for (j = ch; j < endch; j++) {
2535  if (mask & (1 << j)) { /* this channel has been mixed-out */
2536  src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2537  for (k = 0; k < endch; k++) {
2538  achan = s->channel_order_tab[k];
2539  scale = s->xxch_dmix_coeff[j][k];
2540  if (scale != 0.0) {
2541  dst_chan = s->samples_chanptr[achan];
2542  s->fdsp.vector_fmac_scalar(dst_chan, src_chan,
2543  -scale, 256);
2544  }
2545  }
2546  }
2547  }
2548 
2549  /* if a downmix has been embedded then undo the pre-scaling */
2550  if ((mask & (1 << ch)) && s->xxch_dmix_sf[chset] != 1.0f) {
2551  scale = s->xxch_dmix_sf[chset];
2552 
2553  for (j = 0; j < ch; j++) {
2554  src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2555  for (k = 0; k < 256; k++)
2556  src_chan[k] *= scale;
2557  }
2558 
2559  /* LFE channel is always part of core, scale if it exists */
2560  if (s->lfe) {
2561  src_chan = s->samples_chanptr[s->lfe_index];
2562  for (k = 0; k < 256; k++)
2563  src_chan[k] *= scale;
2564  }
2565  }
2566 
2567  ch = endch;
2568  }
2569 
2570  }
2571  }
2572 
2573  /* update lfe history */
2574  lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
2575  for (i = 0; i < 2 * s->lfe * 4; i++)
2576  s->lfe_data[i] = s->lfe_data[i + lfe_samples];
2577 
2578  /* AVMatrixEncoding
2579  *
2580  * DCA_STEREO_TOTAL (Lt/Rt) is equivalent to Dolby Surround */
2582  (s->output & ~DCA_LFE) == DCA_STEREO_TOTAL ?
2584  if (ret < 0)
2585  return ret;
2586 
2587  *got_frame_ptr = 1;
2588 
2589  return buf_size;
2590 }
2591 
2592 
2593 
2594 /**
2595  * DCA initialization
2596  *
2597  * @param avctx pointer to the AVCodecContext
2598  */
2599 
2601 {
2602  DCAContext *s = avctx->priv_data;
2603 
2604  s->avctx = avctx;
2605  dca_init_vlcs();
2606 
2608  ff_mdct_init(&s->imdct, 6, 1, 1.0);
2610  ff_dcadsp_init(&s->dcadsp);
2611  ff_fmt_convert_init(&s->fmt_conv, avctx);
2612 
2613  avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
2614 
2615  /* allow downmixing to stereo */
2616 #if FF_API_REQUEST_CHANNELS
2618  if (avctx->request_channels == 2)
2621 #endif
2622  if (avctx->channels > 2 &&
2624  avctx->channels = 2;
2625 
2626  return 0;
2627 }
2628 
2630 {
2631  DCAContext *s = avctx->priv_data;
2632  ff_mdct_end(&s->imdct);
2634  return 0;
2635 }
2636 
2637 static const AVProfile profiles[] = {
2638  { FF_PROFILE_DTS, "DTS" },
2639  { FF_PROFILE_DTS_ES, "DTS-ES" },
2640  { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2641  { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2642  { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2643  { FF_PROFILE_UNKNOWN },
2644 };
2645 
2646 static const AVOption options[] = {
2647  { "disable_xch", "disable decoding of the XCh extension", offsetof(DCAContext, xch_disable), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_DECODING_PARAM|AV_OPT_FLAG_AUDIO_PARAM },
2648  { NULL },
2649 };
2650 
2651 static const AVClass dca_decoder_class = {
2652  .class_name = "DCA decoder",
2653  .item_name = av_default_item_name,
2654  .option = options,
2655  .version = LIBAVUTIL_VERSION_INT,
2656  .category = AV_CLASS_CATEGORY_DECODER,
2657 };
2658 
2660  .name = "dca",
2661  .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2662  .type = AVMEDIA_TYPE_AUDIO,
2663  .id = AV_CODEC_ID_DTS,
2664  .priv_data_size = sizeof(DCAContext),
2665  .init = dca_decode_init,
2667  .close = dca_decode_end,
2668  .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2669  .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
2671  .profiles = NULL_IF_CONFIG_SMALL(profiles),
2672  .priv_class = &dca_decoder_class,
2673 };