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/channel_layout.h"
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/samplefmt.h"
43 #include "synth_filter.h"
45 #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 DCAXxchSpeakerMask {
99 DCA_XXCH_FRONT_CENTER = 0x0000001,
100 DCA_XXCH_FRONT_LEFT = 0x0000002,
101 DCA_XXCH_FRONT_RIGHT = 0x0000004,
102 DCA_XXCH_SIDE_REAR_LEFT = 0x0000008,
103 DCA_XXCH_SIDE_REAR_RIGHT = 0x0000010,
104 DCA_XXCH_LFE1 = 0x0000020,
105 DCA_XXCH_REAR_CENTER = 0x0000040,
106 DCA_XXCH_SURROUND_REAR_LEFT = 0x0000080,
107 DCA_XXCH_SURROUND_REAR_RIGHT = 0x0000100,
108 DCA_XXCH_SIDE_SURROUND_LEFT = 0x0000200,
109 DCA_XXCH_SIDE_SURROUND_RIGHT = 0x0000400,
110 DCA_XXCH_FRONT_CENTER_LEFT = 0x0000800,
111 DCA_XXCH_FRONT_CENTER_RIGHT = 0x0001000,
112 DCA_XXCH_FRONT_HIGH_LEFT = 0x0002000,
113 DCA_XXCH_FRONT_HIGH_CENTER = 0x0004000,
114 DCA_XXCH_FRONT_HIGH_RIGHT = 0x0008000,
115 DCA_XXCH_LFE2 = 0x0010000,
116 DCA_XXCH_SIDE_FRONT_LEFT = 0x0020000,
117 DCA_XXCH_SIDE_FRONT_RIGHT = 0x0040000,
118 DCA_XXCH_OVERHEAD = 0x0080000,
119 DCA_XXCH_SIDE_HIGH_LEFT = 0x0100000,
120 DCA_XXCH_SIDE_HIGH_RIGHT = 0x0200000,
121 DCA_XXCH_REAR_HIGH_CENTER = 0x0400000,
122 DCA_XXCH_REAR_HIGH_LEFT = 0x0800000,
123 DCA_XXCH_REAR_HIGH_RIGHT = 0x1000000,
124 DCA_XXCH_REAR_LOW_CENTER = 0x2000000,
125 DCA_XXCH_REAR_LOW_LEFT = 0x4000000,
126 DCA_XXCH_REAR_LOW_RIGHT = 0x8000000,
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 */
141 AV_CH_FRONT_LEFT_OF_CENTER,
142 AV_CH_FRONT_RIGHT_OF_CENTER,
143 AV_CH_TOP_FRONT_LEFT,
144 AV_CH_TOP_FRONT_CENTER,
145 AV_CH_TOP_FRONT_RIGHT,
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 */
152 AV_CH_TOP_BACK_CENTER,
154 AV_CH_TOP_BACK_RIGHT,
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 */
160 enum DCAExtensionMask {
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
173 /* -1 are reserved or unknown */
174 static const int dca_ext_audio_descr_mask[] = {
178 DCA_EXT_XCH | DCA_EXT_X96,
185 /* extensions that reside in core substream */
186 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
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.
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
195 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
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
208 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
209 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
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
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
217 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
218 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
219 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
221 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
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
225 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
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
229 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
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
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
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},
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},
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},
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},
314 #define DCA_DOLBY 101 /* FIXME */
316 #define DCA_CHANNEL_BITS 6
317 #define DCA_CHANNEL_MASK 0x3F
321 #define HEADER_SIZE 14
323 #define DCA_MAX_FRAME_SIZE 16384
324 #define DCA_MAX_EXSS_HEADER_SIZE 4096
326 #define DCA_BUFFER_PADDING_SIZE 1024
328 /** Bit allocation */
330 int offset; ///< code values offset
331 int maxbits[8]; ///< max bits in VLC
332 int wrap; ///< wrap for get_vlc2()
333 VLC vlc[8]; ///< actual codes
336 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
337 static BitAlloc dca_tmode; ///< transition mode VLCs
338 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
339 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
341 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
344 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
349 AVCodecContext *avctx;
351 int frame_type; ///< type of the current frame
352 int samples_deficit; ///< deficit sample count
353 int crc_present; ///< crc is present in the bitstream
354 int sample_blocks; ///< number of PCM sample blocks
355 int frame_size; ///< primary frame byte size
356 int amode; ///< audio channels arrangement
357 int sample_rate; ///< audio sampling rate
358 int bit_rate; ///< transmission bit rate
359 int bit_rate_index; ///< transmission bit rate index
361 int downmix; ///< embedded downmix enabled
362 int dynrange; ///< embedded dynamic range flag
363 int timestamp; ///< embedded time stamp flag
364 int aux_data; ///< auxiliary data flag
365 int hdcd; ///< source material is mastered in HDCD
366 int ext_descr; ///< extension audio descriptor flag
367 int ext_coding; ///< extended coding flag
368 int aspf; ///< audio sync word insertion flag
369 int lfe; ///< low frequency effects flag
370 int predictor_history; ///< predictor history flag
371 int header_crc; ///< header crc check bytes
372 int multirate_inter; ///< multirate interpolator switch
373 int version; ///< encoder software revision
374 int copy_history; ///< copy history
375 int source_pcm_res; ///< source pcm resolution
376 int front_sum; ///< front sum/difference flag
377 int surround_sum; ///< surround sum/difference flag
378 int dialog_norm; ///< dialog normalisation parameter
380 /* Primary audio coding header */
381 int subframes; ///< number of subframes
382 int total_channels; ///< number of channels including extensions
383 int prim_channels; ///< number of primary audio channels
384 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
385 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
386 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
387 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
388 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
389 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
390 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
391 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
393 /* Primary audio coding side information */
394 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
395 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
396 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
397 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
398 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
399 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
400 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
401 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
402 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
403 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
404 int dynrange_coef; ///< dynamic range coefficient
406 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
408 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
409 int lfe_scale_factor;
411 /* Subband samples history (for ADPCM) */
412 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
413 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
414 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
415 int hist_index[DCA_PRIM_CHANNELS_MAX];
416 DECLARE_ALIGNED(32, float, raXin)[32];
418 int output; ///< type of output
420 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
421 float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
422 float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
423 uint8_t *extra_channels_buffer;
424 unsigned int extra_channels_buffer_size;
426 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
427 int dca_buffer_size; ///< how much data is in the dca_buffer
429 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
431 /* Current position in DCA frame */
432 int current_subframe;
433 int current_subsubframe;
435 int core_ext_mask; ///< present extensions in the core substream
437 /* XCh extension information */
438 int xch_present; ///< XCh extension present and valid
439 int xch_base_channel; ///< index of first (only) channel containing XCH data
441 /* XXCH extension information */
443 int xxch_nbits_spk_mask;
444 uint32_t xxch_core_spkmask;
445 uint32_t xxch_spk_masks[4]; /* speaker masks, last element is core mask */
446 int xxch_chset_nch[4];
447 float xxch_dmix_sf[DCA_CHSETS_MAX];
449 uint32_t xxch_dmix_embedded; /* lower layer has mix pre-embedded, per chset */
450 float xxch_dmix_coeff[DCA_PRIM_CHANNELS_MAX][32]; /* worst case sizing */
452 int8_t xxch_order_tab[32];
455 /* ExSS header parser */
456 int static_fields; ///< static fields present
457 int mix_metadata; ///< mixing metadata present
458 int num_mix_configs; ///< number of mix out configurations
459 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
463 int debug_flag; ///< used for suppressing repeated error messages output
464 AVFloatDSPContext fdsp;
466 SynthFilterContext synth;
467 DCADSPContext dcadsp;
468 FmtConvertContext fmt_conv;
471 static const uint16_t dca_vlc_offs[] = {
472 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
473 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
474 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
475 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
476 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
477 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
480 static av_cold void dca_init_vlcs(void)
482 static int vlcs_initialized = 0;
484 static VLC_TYPE dca_table[23622][2];
486 if (vlcs_initialized)
489 dca_bitalloc_index.offset = 1;
490 dca_bitalloc_index.wrap = 2;
491 for (i = 0; i < 5; i++) {
492 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
493 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
494 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
495 bitalloc_12_bits[i], 1, 1,
496 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
498 dca_scalefactor.offset = -64;
499 dca_scalefactor.wrap = 2;
500 for (i = 0; i < 5; i++) {
501 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
502 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
503 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
504 scales_bits[i], 1, 1,
505 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
507 dca_tmode.offset = 0;
509 for (i = 0; i < 4; i++) {
510 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
511 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
512 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
514 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
517 for (i = 0; i < 10; i++)
518 for (j = 0; j < 7; j++) {
519 if (!bitalloc_codes[i][j])
521 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
522 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
523 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
524 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
526 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
528 bitalloc_bits[i][j], 1, 1,
529 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
532 vlcs_initialized = 1;
535 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
538 *dst++ = get_bits(gb, bits);
541 static inline int dca_xxch2index(DCAContext *s, int xxch_ch)
545 /* locate channel set containing the channel */
546 for (i = -1, base = 0, mask = (s->xxch_core_spkmask & ~DCA_XXCH_LFE1);
547 i <= s->xxch_chset && !(mask & xxch_ch); mask = s->xxch_spk_masks[++i])
548 base += av_popcount(mask);
550 return base + av_popcount(mask & (xxch_ch - 1));
553 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel,
557 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
558 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
559 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
560 int hdr_pos = 0, hdr_size = 0;
561 float sign, mag, scale_factor;
562 int this_chans, acc_mask;
563 int embedded_downmix;
567 /* xxch has arbitrary sized audio coding headers */
569 hdr_pos = get_bits_count(&s->gb);
570 hdr_size = get_bits(&s->gb, 7) + 1;
573 nchans = get_bits(&s->gb, 3) + 1;
574 s->total_channels = nchans + base_channel;
575 s->prim_channels = s->total_channels;
577 /* obtain speaker layout mask & downmix coefficients for XXCH */
579 acc_mask = s->xxch_core_spkmask;
581 this_chans = get_bits(&s->gb, s->xxch_nbits_spk_mask - 6) << 6;
582 s->xxch_spk_masks[s->xxch_chset] = this_chans;
583 s->xxch_chset_nch[s->xxch_chset] = nchans;
585 for (i = 0; i <= s->xxch_chset; i++)
586 acc_mask |= s->xxch_spk_masks[i];
588 /* check for downmixing information */
589 if (get_bits1(&s->gb)) {
590 embedded_downmix = get_bits1(&s->gb);
592 1.0f / dca_downmix_scale_factors[(get_bits(&s->gb, 6) - 1) << 2];
594 s->xxch_dmix_sf[s->xxch_chset] = scale_factor;
596 for (i = base_channel; i < s->prim_channels; i++) {
597 mask[i] = get_bits(&s->gb, s->xxch_nbits_spk_mask);
600 for (j = base_channel; j < s->prim_channels; j++) {
601 memset(s->xxch_dmix_coeff[j], 0, sizeof(s->xxch_dmix_coeff[0]));
602 s->xxch_dmix_embedded |= (embedded_downmix << j);
603 for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
604 if (mask[j] & (1 << i)) {
605 if ((1 << i) == DCA_XXCH_LFE1) {
606 av_log(s->avctx, AV_LOG_WARNING,
607 "DCA-XXCH: dmix to LFE1 not supported.\n");
611 coeff = get_bits(&s->gb, 7);
612 sign = (coeff & 64) ? 1.0 : -1.0;
613 mag = dca_downmix_scale_factors[((coeff & 63) - 1) << 2];
614 ichan = dca_xxch2index(s, 1 << i);
615 s->xxch_dmix_coeff[j][ichan] = sign * mag;
622 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
623 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
626 for (i = base_channel; i < s->prim_channels; i++) {
627 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
628 if (s->subband_activity[i] > DCA_SUBBANDS)
629 s->subband_activity[i] = DCA_SUBBANDS;
631 for (i = base_channel; i < s->prim_channels; i++) {
632 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
633 if (s->vq_start_subband[i] > DCA_SUBBANDS)
634 s->vq_start_subband[i] = DCA_SUBBANDS;
636 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
637 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
638 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
639 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
641 /* Get codebooks quantization indexes */
643 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
644 for (j = 1; j < 11; j++)
645 for (i = base_channel; i < s->prim_channels; i++)
646 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
648 /* Get scale factor adjustment */
649 for (j = 0; j < 11; j++)
650 for (i = base_channel; i < s->prim_channels; i++)
651 s->scalefactor_adj[i][j] = 1;
653 for (j = 1; j < 11; j++)
654 for (i = base_channel; i < s->prim_channels; i++)
655 if (s->quant_index_huffman[i][j] < thr[j])
656 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
659 if (s->crc_present) {
660 /* Audio header CRC check */
661 get_bits(&s->gb, 16);
664 /* Skip to the end of the header, also ignore CRC if present */
665 i = get_bits_count(&s->gb);
666 if (hdr_pos + 8 * hdr_size > i)
667 skip_bits_long(&s->gb, hdr_pos + 8 * hdr_size - i);
670 s->current_subframe = 0;
671 s->current_subsubframe = 0;
674 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
675 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
676 for (i = base_channel; i < s->prim_channels; i++) {
677 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
678 s->subband_activity[i]);
679 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
680 s->vq_start_subband[i]);
681 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
682 s->joint_intensity[i]);
683 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
684 s->transient_huffman[i]);
685 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
686 s->scalefactor_huffman[i]);
687 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
688 s->bitalloc_huffman[i]);
689 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
690 for (j = 0; j < 11; j++)
691 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
692 av_log(s->avctx, AV_LOG_DEBUG, "\n");
693 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
694 for (j = 0; j < 11; j++)
695 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
696 av_log(s->avctx, AV_LOG_DEBUG, "\n");
703 static int dca_parse_frame_header(DCAContext *s)
705 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
708 skip_bits_long(&s->gb, 32);
711 s->frame_type = get_bits(&s->gb, 1);
712 s->samples_deficit = get_bits(&s->gb, 5) + 1;
713 s->crc_present = get_bits(&s->gb, 1);
714 s->sample_blocks = get_bits(&s->gb, 7) + 1;
715 s->frame_size = get_bits(&s->gb, 14) + 1;
716 if (s->frame_size < 95)
717 return AVERROR_INVALIDDATA;
718 s->amode = get_bits(&s->gb, 6);
719 s->sample_rate = avpriv_dca_sample_rates[get_bits(&s->gb, 4)];
721 return AVERROR_INVALIDDATA;
722 s->bit_rate_index = get_bits(&s->gb, 5);
723 s->bit_rate = dca_bit_rates[s->bit_rate_index];
725 return AVERROR_INVALIDDATA;
727 s->downmix = get_bits(&s->gb, 1); /* note: this is FixedBit == 0 */
728 s->dynrange = get_bits(&s->gb, 1);
729 s->timestamp = get_bits(&s->gb, 1);
730 s->aux_data = get_bits(&s->gb, 1);
731 s->hdcd = get_bits(&s->gb, 1);
732 s->ext_descr = get_bits(&s->gb, 3);
733 s->ext_coding = get_bits(&s->gb, 1);
734 s->aspf = get_bits(&s->gb, 1);
735 s->lfe = get_bits(&s->gb, 2);
736 s->predictor_history = get_bits(&s->gb, 1);
740 av_log(s->avctx, AV_LOG_ERROR, "Invalid LFE value: %d\n", s->lfe);
741 return AVERROR_INVALIDDATA;
744 /* TODO: check CRC */
746 s->header_crc = get_bits(&s->gb, 16);
748 s->multirate_inter = get_bits(&s->gb, 1);
749 s->version = get_bits(&s->gb, 4);
750 s->copy_history = get_bits(&s->gb, 2);
751 s->source_pcm_res = get_bits(&s->gb, 3);
752 s->front_sum = get_bits(&s->gb, 1);
753 s->surround_sum = get_bits(&s->gb, 1);
754 s->dialog_norm = get_bits(&s->gb, 4);
756 /* FIXME: channels mixing levels */
757 s->output = s->amode;
759 s->output |= DCA_LFE;
762 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
763 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
764 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
765 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
766 s->sample_blocks, s->sample_blocks * 32);
767 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
768 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
769 s->amode, dca_channels[s->amode]);
770 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
772 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
774 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
775 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
776 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
777 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
778 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
779 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
780 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
781 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
782 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
783 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
784 s->predictor_history);
785 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
786 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
788 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
789 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
790 av_log(s->avctx, AV_LOG_DEBUG,
791 "source pcm resolution: %i (%i bits/sample)\n",
792 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
793 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
794 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
795 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
796 av_log(s->avctx, AV_LOG_DEBUG, "\n");
799 /* Primary audio coding header */
800 s->subframes = get_bits(&s->gb, 4) + 1;
802 return dca_parse_audio_coding_header(s, 0, 0);
806 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
809 /* huffman encoded */
810 value += get_bitalloc(gb, &dca_scalefactor, level);
811 value = av_clip(value, 0, (1 << log2range) - 1);
812 } else if (level < 8) {
813 if (level + 1 > log2range) {
814 skip_bits(gb, level + 1 - log2range);
815 value = get_bits(gb, log2range);
817 value = get_bits(gb, level + 1);
823 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
825 /* Primary audio coding side information */
828 if (get_bits_left(&s->gb) < 0)
829 return AVERROR_INVALIDDATA;
832 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
833 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
836 for (j = base_channel; j < s->prim_channels; j++) {
837 for (k = 0; k < s->subband_activity[j]; k++)
838 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
841 /* Get prediction codebook */
842 for (j = base_channel; j < s->prim_channels; j++) {
843 for (k = 0; k < s->subband_activity[j]; k++) {
844 if (s->prediction_mode[j][k] > 0) {
845 /* (Prediction coefficient VQ address) */
846 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
851 /* Bit allocation index */
852 for (j = base_channel; j < s->prim_channels; j++) {
853 for (k = 0; k < s->vq_start_subband[j]; k++) {
854 if (s->bitalloc_huffman[j] == 6)
855 s->bitalloc[j][k] = get_bits(&s->gb, 5);
856 else if (s->bitalloc_huffman[j] == 5)
857 s->bitalloc[j][k] = get_bits(&s->gb, 4);
858 else if (s->bitalloc_huffman[j] == 7) {
859 av_log(s->avctx, AV_LOG_ERROR,
860 "Invalid bit allocation index\n");
861 return AVERROR_INVALIDDATA;
864 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
867 if (s->bitalloc[j][k] > 26) {
868 av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
869 j, k, s->bitalloc[j][k]);
870 return AVERROR_INVALIDDATA;
875 /* Transition mode */
876 for (j = base_channel; j < s->prim_channels; j++) {
877 for (k = 0; k < s->subband_activity[j]; k++) {
878 s->transition_mode[j][k] = 0;
879 if (s->subsubframes[s->current_subframe] > 1 &&
880 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
881 s->transition_mode[j][k] =
882 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
887 if (get_bits_left(&s->gb) < 0)
888 return AVERROR_INVALIDDATA;
890 for (j = base_channel; j < s->prim_channels; j++) {
891 const uint32_t *scale_table;
892 int scale_sum, log_size;
894 memset(s->scale_factor[j], 0,
895 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
897 if (s->scalefactor_huffman[j] == 6) {
898 scale_table = scale_factor_quant7;
901 scale_table = scale_factor_quant6;
905 /* When huffman coded, only the difference is encoded */
908 for (k = 0; k < s->subband_activity[j]; k++) {
909 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
910 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
911 s->scale_factor[j][k][0] = scale_table[scale_sum];
914 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
915 /* Get second scale factor */
916 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
917 s->scale_factor[j][k][1] = scale_table[scale_sum];
922 /* Joint subband scale factor codebook select */
923 for (j = base_channel; j < s->prim_channels; j++) {
924 /* Transmitted only if joint subband coding enabled */
925 if (s->joint_intensity[j] > 0)
926 s->joint_huff[j] = get_bits(&s->gb, 3);
929 if (get_bits_left(&s->gb) < 0)
930 return AVERROR_INVALIDDATA;
932 /* Scale factors for joint subband coding */
933 for (j = base_channel; j < s->prim_channels; j++) {
936 /* Transmitted only if joint subband coding enabled */
937 if (s->joint_intensity[j] > 0) {
939 source_channel = s->joint_intensity[j] - 1;
941 /* When huffman coded, only the difference is encoded
942 * (is this valid as well for joint scales ???) */
944 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
945 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
946 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
949 if (!(s->debug_flag & 0x02)) {
950 av_log(s->avctx, AV_LOG_DEBUG,
951 "Joint stereo coding not supported\n");
952 s->debug_flag |= 0x02;
957 /* Stereo downmix coefficients */
958 if (!base_channel && s->prim_channels > 2) {
960 for (j = base_channel; j < s->prim_channels; j++) {
961 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
962 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
965 int am = s->amode & DCA_CHANNEL_MASK;
966 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
967 av_log(s->avctx, AV_LOG_ERROR,
968 "Invalid channel mode %d\n", am);
969 return AVERROR_INVALIDDATA;
971 if (s->prim_channels > FF_ARRAY_ELEMS(dca_default_coeffs[0])) {
972 avpriv_request_sample(s->avctx, "Downmixing %d channels",
974 return AVERROR_PATCHWELCOME;
977 for (j = base_channel; j < s->prim_channels; j++) {
978 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
979 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
984 /* Dynamic range coefficient */
985 if (!base_channel && s->dynrange)
986 s->dynrange_coef = get_bits(&s->gb, 8);
988 /* Side information CRC check word */
989 if (s->crc_present) {
990 get_bits(&s->gb, 16);
994 * Primary audio data arrays
997 /* VQ encoded high frequency subbands */
998 for (j = base_channel; j < s->prim_channels; j++)
999 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
1000 /* 1 vector -> 32 samples */
1001 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
1003 /* Low frequency effect data */
1004 if (!base_channel && s->lfe) {
1007 int lfe_samples = 2 * s->lfe * (4 + block_index);
1008 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
1011 for (j = lfe_samples; j < lfe_end_sample; j++) {
1012 /* Signed 8 bits int */
1013 s->lfe_data[j] = get_sbits(&s->gb, 8);
1016 /* Scale factor index */
1017 quant7 = get_bits(&s->gb, 8);
1019 avpriv_request_sample(s->avctx, "LFEScaleIndex larger than 127");
1020 return AVERROR_INVALIDDATA;
1022 s->lfe_scale_factor = scale_factor_quant7[quant7];
1024 /* Quantization step size * scale factor */
1025 lfe_scale = 0.035 * s->lfe_scale_factor;
1027 for (j = lfe_samples; j < lfe_end_sample; j++)
1028 s->lfe_data[j] *= lfe_scale;
1032 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
1033 s->subsubframes[s->current_subframe]);
1034 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
1035 s->partial_samples[s->current_subframe]);
1037 for (j = base_channel; j < s->prim_channels; j++) {
1038 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
1039 for (k = 0; k < s->subband_activity[j]; k++)
1040 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
1041 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1043 for (j = base_channel; j < s->prim_channels; j++) {
1044 for (k = 0; k < s->subband_activity[j]; k++)
1045 av_log(s->avctx, AV_LOG_DEBUG,
1046 "prediction coefs: %f, %f, %f, %f\n",
1047 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
1048 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
1049 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
1050 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
1052 for (j = base_channel; j < s->prim_channels; j++) {
1053 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
1054 for (k = 0; k < s->vq_start_subband[j]; k++)
1055 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
1056 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1058 for (j = base_channel; j < s->prim_channels; j++) {
1059 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
1060 for (k = 0; k < s->subband_activity[j]; k++)
1061 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
1062 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1064 for (j = base_channel; j < s->prim_channels; j++) {
1065 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
1066 for (k = 0; k < s->subband_activity[j]; k++) {
1067 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
1068 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
1069 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
1070 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
1072 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1074 for (j = base_channel; j < s->prim_channels; j++) {
1075 if (s->joint_intensity[j] > 0) {
1076 int source_channel = s->joint_intensity[j] - 1;
1077 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
1078 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
1079 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
1080 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1083 if (!base_channel && s->prim_channels > 2 && s->downmix) {
1084 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
1085 for (j = 0; j < s->prim_channels; j++) {
1086 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
1087 dca_downmix_coeffs[s->downmix_coef[j][0]]);
1088 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
1089 dca_downmix_coeffs[s->downmix_coef[j][1]]);
1091 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1093 for (j = base_channel; j < s->prim_channels; j++)
1094 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
1095 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
1096 if (!base_channel && s->lfe) {
1097 int lfe_samples = 2 * s->lfe * (4 + block_index);
1098 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
1100 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
1101 for (j = lfe_samples; j < lfe_end_sample; j++)
1102 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
1103 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1110 static void qmf_32_subbands(DCAContext *s, int chans,
1111 float samples_in[32][8], float *samples_out,
1114 const float *prCoeff;
1116 int sb_act = s->subband_activity[chans];
1118 scale *= sqrt(1 / 8.0);
1121 if (!s->multirate_inter) /* Non-perfect reconstruction */
1122 prCoeff = fir_32bands_nonperfect;
1123 else /* Perfect reconstruction */
1124 prCoeff = fir_32bands_perfect;
1126 s->dcadsp.qmf_32_subbands(samples_in, sb_act, &s->synth, &s->imdct,
1127 s->subband_fir_hist[chans],
1128 &s->hist_index[chans],
1129 s->subband_fir_noidea[chans], prCoeff,
1130 samples_out, s->raXin, scale);
1133 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
1134 int num_deci_sample, float *samples_in,
1135 float *samples_out, float scale)
1137 /* samples_in: An array holding decimated samples.
1138 * Samples in current subframe starts from samples_in[0],
1139 * while samples_in[-1], samples_in[-2], ..., stores samples
1140 * from last subframe as history.
1142 * samples_out: An array holding interpolated samples
1146 const float *prCoeff;
1149 /* Select decimation filter */
1150 if (decimation_select == 1) {
1152 prCoeff = lfe_fir_128;
1155 prCoeff = lfe_fir_64;
1158 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1159 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1161 samples_out += 2 * decifactor;
1165 /* downmixing routines */
1166 #define MIX_REAR1(samples, s1, rs, coef) \
1167 samples[0][i] += samples[s1][i] * coef[rs][0]; \
1168 samples[1][i] += samples[s1][i] * coef[rs][1];
1170 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1171 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1172 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1174 #define MIX_FRONT3(samples, coef) \
1175 t = samples[c][i]; \
1176 u = samples[l][i]; \
1177 v = samples[r][i]; \
1178 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1179 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1181 #define DOWNMIX_TO_STEREO(op1, op2) \
1182 for (i = 0; i < 256; i++) { \
1187 static void dca_downmix(float **samples, int srcfmt,
1188 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1189 const int8_t *channel_mapping)
1191 int c, l, r, sl, sr, s;
1194 float coef[DCA_PRIM_CHANNELS_MAX][2];
1196 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1197 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1198 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1204 case DCA_STEREO_TOTAL:
1205 case DCA_STEREO_SUMDIFF:
1207 av_log(NULL, AV_LOG_ERROR, "Not implemented!\n");
1212 c = channel_mapping[0];
1213 l = channel_mapping[1];
1214 r = channel_mapping[2];
1215 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1218 s = channel_mapping[2];
1219 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1222 c = channel_mapping[0];
1223 l = channel_mapping[1];
1224 r = channel_mapping[2];
1225 s = channel_mapping[3];
1226 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1227 MIX_REAR1(samples, s, 3, coef));
1230 sl = channel_mapping[2];
1231 sr = channel_mapping[3];
1232 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1235 c = channel_mapping[0];
1236 l = channel_mapping[1];
1237 r = channel_mapping[2];
1238 sl = channel_mapping[3];
1239 sr = channel_mapping[4];
1240 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1241 MIX_REAR2(samples, sl, sr, 3, coef));
1247 #ifndef decode_blockcodes
1248 /* Very compact version of the block code decoder that does not use table
1249 * look-up but is slightly slower */
1250 static int decode_blockcode(int code, int levels, int32_t *values)
1253 int offset = (levels - 1) >> 1;
1255 for (i = 0; i < 4; i++) {
1256 int div = FASTDIV(code, levels);
1257 values[i] = code - offset - div * levels;
1264 static int decode_blockcodes(int code1, int code2, int levels, int32_t *values)
1266 return decode_blockcode(code1, levels, values) |
1267 decode_blockcode(code2, levels, values + 4);
1271 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1272 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1274 #ifndef int8x8_fmul_int32
1275 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1277 float fscale = scale / 16.0;
1279 for (i = 0; i < 8; i++)
1280 dst[i] = src[i] * fscale;
1284 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1287 int subsubframe = s->current_subsubframe;
1289 const float *quant_step_table;
1292 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1293 LOCAL_ALIGNED_16(int32_t, block, [8 * DCA_SUBBANDS]);
1299 /* Select quantization step size table */
1300 if (s->bit_rate_index == 0x1f)
1301 quant_step_table = lossless_quant_d;
1303 quant_step_table = lossy_quant_d;
1305 for (k = base_channel; k < s->prim_channels; k++) {
1306 float rscale[DCA_SUBBANDS];
1308 if (get_bits_left(&s->gb) < 0)
1309 return AVERROR_INVALIDDATA;
1311 for (l = 0; l < s->vq_start_subband[k]; l++) {
1314 /* Select the mid-tread linear quantizer */
1315 int abits = s->bitalloc[k][l];
1317 float quant_step_size = quant_step_table[abits];
1320 * Determine quantization index code book and its type
1323 /* Select quantization index code book */
1324 int sel = s->quant_index_huffman[k][abits];
1327 * Extract bits from the bit stream
1331 memset(block + 8 * l, 0, 8 * sizeof(block[0]));
1333 /* Deal with transients */
1334 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1335 rscale[l] = quant_step_size * s->scale_factor[k][l][sfi] *
1336 s->scalefactor_adj[k][sel];
1338 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1341 int block_code1, block_code2, size, levels, err;
1343 size = abits_sizes[abits - 1];
1344 levels = abits_levels[abits - 1];
1346 block_code1 = get_bits(&s->gb, size);
1347 block_code2 = get_bits(&s->gb, size);
1348 err = decode_blockcodes(block_code1, block_code2,
1349 levels, block + 8 * l);
1351 av_log(s->avctx, AV_LOG_ERROR,
1352 "ERROR: block code look-up failed\n");
1353 return AVERROR_INVALIDDATA;
1357 for (m = 0; m < 8; m++)
1358 block[8 * l + m] = get_sbits(&s->gb, abits - 3);
1362 for (m = 0; m < 8; m++)
1363 block[8 * l + m] = get_bitalloc(&s->gb,
1364 &dca_smpl_bitalloc[abits], sel);
1370 s->fmt_conv.int32_to_float_fmul_array8(&s->fmt_conv, subband_samples[k][0],
1371 block, rscale, 8 * s->vq_start_subband[k]);
1373 for (l = 0; l < s->vq_start_subband[k]; l++) {
1376 * Inverse ADPCM if in prediction mode
1378 if (s->prediction_mode[k][l]) {
1380 for (m = 0; m < 8; m++) {
1381 for (n = 1; n <= 4; n++)
1383 subband_samples[k][l][m] +=
1384 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1385 subband_samples[k][l][m - n] / 8192);
1386 else if (s->predictor_history)
1387 subband_samples[k][l][m] +=
1388 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1389 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1395 * Decode VQ encoded high frequencies
1397 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1398 /* 1 vector -> 32 samples but we only need the 8 samples
1399 * for this subsubframe. */
1400 int hfvq = s->high_freq_vq[k][l];
1402 if (!s->debug_flag & 0x01) {
1403 av_log(s->avctx, AV_LOG_DEBUG,
1404 "Stream with high frequencies VQ coding\n");
1405 s->debug_flag |= 0x01;
1408 int8x8_fmul_int32(subband_samples[k][l],
1409 &high_freq_vq[hfvq][subsubframe * 8],
1410 s->scale_factor[k][l][0]);
1414 /* Check for DSYNC after subsubframe */
1415 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1416 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1418 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1421 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1422 return AVERROR_INVALIDDATA;
1426 /* Backup predictor history for adpcm */
1427 for (k = base_channel; k < s->prim_channels; k++)
1428 for (l = 0; l < s->vq_start_subband[k]; l++)
1429 memcpy(s->subband_samples_hist[k][l],
1430 &subband_samples[k][l][4],
1431 4 * sizeof(subband_samples[0][0][0]));
1436 static int dca_filter_channels(DCAContext *s, int block_index)
1438 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1441 /* 32 subbands QMF */
1442 for (k = 0; k < s->prim_channels; k++) {
1443 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1444 0, 8388608.0, 8388608.0 };*/
1445 if (s->channel_order_tab[k] >= 0)
1446 qmf_32_subbands(s, k, subband_samples[k],
1447 s->samples_chanptr[s->channel_order_tab[k]],
1448 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1452 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1453 dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
1456 /* Generate LFE samples for this subsubframe FIXME!!! */
1457 if (s->output & DCA_LFE) {
1458 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1459 s->lfe_data + 2 * s->lfe * (block_index + 4),
1460 s->samples_chanptr[s->lfe_index],
1461 1.0 / (256.0 * 32768.0));
1462 /* Outputs 20bits pcm samples */
1469 static int dca_subframe_footer(DCAContext *s, int base_channel)
1471 int aux_data_count = 0, i;
1474 * Unpack optional information
1477 /* presumably optional information only appears in the core? */
1478 if (!base_channel) {
1480 skip_bits_long(&s->gb, 32);
1483 aux_data_count = get_bits(&s->gb, 6);
1485 for (i = 0; i < aux_data_count; i++)
1486 get_bits(&s->gb, 8);
1488 if (s->crc_present && (s->downmix || s->dynrange))
1489 get_bits(&s->gb, 16);
1496 * Decode a dca frame block
1498 * @param s pointer to the DCAContext
1501 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1506 if (s->current_subframe >= s->subframes) {
1507 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1508 s->current_subframe, s->subframes);
1509 return AVERROR_INVALIDDATA;
1512 if (!s->current_subsubframe) {
1514 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1516 /* Read subframe header */
1517 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1521 /* Read subsubframe */
1523 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1525 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1529 s->current_subsubframe++;
1530 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1531 s->current_subsubframe = 0;
1532 s->current_subframe++;
1534 if (s->current_subframe >= s->subframes) {
1536 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1538 /* Read subframe footer */
1539 if ((ret = dca_subframe_footer(s, base_channel)))
1547 * Return the number of channels in an ExSS speaker mask (HD)
1549 static int dca_exss_mask2count(int mask)
1551 /* count bits that mean speaker pairs twice */
1552 return av_popcount(mask) +
1553 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1554 DCA_EXSS_FRONT_LEFT_RIGHT |
1555 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1556 DCA_EXSS_WIDE_LEFT_RIGHT |
1557 DCA_EXSS_SIDE_LEFT_RIGHT |
1558 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1559 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1560 DCA_EXSS_REAR_LEFT_RIGHT |
1561 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1565 * Skip mixing coefficients of a single mix out configuration (HD)
1567 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1571 for (i = 0; i < channels; i++) {
1572 int mix_map_mask = get_bits(gb, out_ch);
1573 int num_coeffs = av_popcount(mix_map_mask);
1574 skip_bits_long(gb, num_coeffs * 6);
1579 * Parse extension substream asset header (HD)
1581 static int dca_exss_parse_asset_header(DCAContext *s)
1583 int header_pos = get_bits_count(&s->gb);
1586 int embedded_stereo = 0;
1587 int embedded_6ch = 0;
1588 int drc_code_present;
1589 int av_uninit(extensions_mask);
1592 if (get_bits_left(&s->gb) < 16)
1595 /* We will parse just enough to get to the extensions bitmask with which
1596 * we can set the profile value. */
1598 header_size = get_bits(&s->gb, 9) + 1;
1599 skip_bits(&s->gb, 3); // asset index
1601 if (s->static_fields) {
1602 if (get_bits1(&s->gb))
1603 skip_bits(&s->gb, 4); // asset type descriptor
1604 if (get_bits1(&s->gb))
1605 skip_bits_long(&s->gb, 24); // language descriptor
1607 if (get_bits1(&s->gb)) {
1608 /* How can one fit 1024 bytes of text here if the maximum value
1609 * for the asset header size field above was 512 bytes? */
1610 int text_length = get_bits(&s->gb, 10) + 1;
1611 if (get_bits_left(&s->gb) < text_length * 8)
1613 skip_bits_long(&s->gb, text_length * 8); // info text
1616 skip_bits(&s->gb, 5); // bit resolution - 1
1617 skip_bits(&s->gb, 4); // max sample rate code
1618 channels = get_bits(&s->gb, 8) + 1;
1620 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1621 int spkr_remap_sets;
1622 int spkr_mask_size = 16;
1626 embedded_stereo = get_bits1(&s->gb);
1628 embedded_6ch = get_bits1(&s->gb);
1630 if (get_bits1(&s->gb)) {
1631 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1632 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1635 spkr_remap_sets = get_bits(&s->gb, 3);
1637 for (i = 0; i < spkr_remap_sets; i++) {
1638 /* std layout mask for each remap set */
1639 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1642 for (i = 0; i < spkr_remap_sets; i++) {
1643 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1644 if (get_bits_left(&s->gb) < 0)
1647 for (j = 0; j < num_spkrs[i]; j++) {
1648 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1649 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1650 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1655 skip_bits(&s->gb, 3); // representation type
1659 drc_code_present = get_bits1(&s->gb);
1660 if (drc_code_present)
1661 get_bits(&s->gb, 8); // drc code
1663 if (get_bits1(&s->gb))
1664 skip_bits(&s->gb, 5); // dialog normalization code
1666 if (drc_code_present && embedded_stereo)
1667 get_bits(&s->gb, 8); // drc stereo code
1669 if (s->mix_metadata && get_bits1(&s->gb)) {
1670 skip_bits(&s->gb, 1); // external mix
1671 skip_bits(&s->gb, 6); // post mix gain code
1673 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1674 skip_bits(&s->gb, 3); // drc limit
1676 skip_bits(&s->gb, 8); // custom drc code
1678 if (get_bits1(&s->gb)) // channel specific scaling
1679 for (i = 0; i < s->num_mix_configs; i++)
1680 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1682 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1684 for (i = 0; i < s->num_mix_configs; i++) {
1685 if (get_bits_left(&s->gb) < 0)
1687 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1689 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1690 if (embedded_stereo)
1691 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1695 switch (get_bits(&s->gb, 2)) {
1696 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1697 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1698 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1699 case 3: extensions_mask = 0; /* aux coding */ break;
1702 /* not parsed further, we were only interested in the extensions mask */
1704 if (get_bits_left(&s->gb) < 0)
1707 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1708 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1711 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1713 if (extensions_mask & DCA_EXT_EXSS_XLL)
1714 s->profile = FF_PROFILE_DTS_HD_MA;
1715 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1717 s->profile = FF_PROFILE_DTS_HD_HRA;
1719 if (!(extensions_mask & DCA_EXT_CORE))
1720 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1721 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1722 av_log(s->avctx, AV_LOG_WARNING,
1723 "DTS extensions detection mismatch (%d, %d)\n",
1724 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1729 static int dca_xbr_parse_frame(DCAContext *s)
1731 int scale_table_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS][2];
1732 int active_bands[DCA_CHSETS_MAX][DCA_CHSET_CHANS_MAX];
1733 int abits_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS];
1734 int anctemp[DCA_CHSET_CHANS_MAX];
1735 int chset_fsize[DCA_CHSETS_MAX];
1736 int n_xbr_ch[DCA_CHSETS_MAX];
1737 int hdr_size, num_chsets, xbr_tmode, hdr_pos;
1738 int i, j, k, l, chset, chan_base;
1740 av_log(s->avctx, AV_LOG_DEBUG, "DTS-XBR: decoding XBR extension\n");
1742 /* get bit position of sync header */
1743 hdr_pos = get_bits_count(&s->gb) - 32;
1745 hdr_size = get_bits(&s->gb, 6) + 1;
1746 num_chsets = get_bits(&s->gb, 2) + 1;
1748 for(i = 0; i < num_chsets; i++)
1749 chset_fsize[i] = get_bits(&s->gb, 14) + 1;
1751 xbr_tmode = get_bits1(&s->gb);
1753 for(i = 0; i < num_chsets; i++) {
1754 n_xbr_ch[i] = get_bits(&s->gb, 3) + 1;
1755 k = get_bits(&s->gb, 2) + 5;
1756 for(j = 0; j < n_xbr_ch[i]; j++)
1757 active_bands[i][j] = get_bits(&s->gb, k) + 1;
1760 /* skip to the end of the header */
1761 i = get_bits_count(&s->gb);
1762 if(hdr_pos + hdr_size * 8 > i)
1763 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1765 /* loop over the channel data sets */
1766 /* only decode as many channels as we've decoded base data for */
1767 for(chset = 0, chan_base = 0;
1768 chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->prim_channels;
1769 chan_base += n_xbr_ch[chset++]) {
1770 int start_posn = get_bits_count(&s->gb);
1771 int subsubframe = 0;
1774 /* loop over subframes */
1775 for (k = 0; k < (s->sample_blocks / 8); k++) {
1776 /* parse header if we're on first subsubframe of a block */
1777 if(subsubframe == 0) {
1778 /* Parse subframe header */
1779 for(i = 0; i < n_xbr_ch[chset]; i++) {
1780 anctemp[i] = get_bits(&s->gb, 2) + 2;
1783 for(i = 0; i < n_xbr_ch[chset]; i++) {
1784 get_array(&s->gb, abits_high[i], active_bands[chset][i], anctemp[i]);
1787 for(i = 0; i < n_xbr_ch[chset]; i++) {
1788 anctemp[i] = get_bits(&s->gb, 3);
1789 if(anctemp[i] < 1) {
1790 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: SYNC ERROR\n");
1791 return AVERROR_INVALIDDATA;
1795 /* generate scale factors */
1796 for(i = 0; i < n_xbr_ch[chset]; i++) {
1797 const uint32_t *scale_table;
1800 if (s->scalefactor_huffman[chan_base+i] == 6) {
1801 scale_table = scale_factor_quant7;
1803 scale_table = scale_factor_quant6;
1808 for(j = 0; j < active_bands[chset][i]; j++) {
1809 if(abits_high[i][j] > 0) {
1810 scale_table_high[i][j][0] =
1811 scale_table[get_bits(&s->gb, nbits)];
1813 if(xbr_tmode && s->transition_mode[i][j]) {
1814 scale_table_high[i][j][1] =
1815 scale_table[get_bits(&s->gb, nbits)];
1822 /* decode audio array for this block */
1823 for(i = 0; i < n_xbr_ch[chset]; i++) {
1824 for(j = 0; j < active_bands[chset][i]; j++) {
1825 const int xbr_abits = abits_high[i][j];
1826 const float quant_step_size = lossless_quant_d[xbr_abits];
1827 const int sfi = xbr_tmode && s->transition_mode[i][j] && subsubframe >= s->transition_mode[i][j];
1828 const float rscale = quant_step_size * scale_table_high[i][j][sfi];
1829 float *subband_samples = s->subband_samples[k][chan_base+i][j];
1836 get_array(&s->gb, block, 8, xbr_abits - 3);
1838 int block_code1, block_code2, size, levels, err;
1840 size = abits_sizes[xbr_abits - 1];
1841 levels = abits_levels[xbr_abits - 1];
1843 block_code1 = get_bits(&s->gb, size);
1844 block_code2 = get_bits(&s->gb, size);
1845 err = decode_blockcodes(block_code1, block_code2,
1848 av_log(s->avctx, AV_LOG_ERROR,
1849 "ERROR: DTS-XBR: block code look-up failed\n");
1850 return AVERROR_INVALIDDATA;
1854 /* scale & sum into subband */
1855 for(l = 0; l < 8; l++)
1856 subband_samples[l] += (float)block[l] * rscale;
1860 /* check DSYNC marker */
1861 if(s->aspf || subsubframe == s->subsubframes[subframe] - 1) {
1862 if(get_bits(&s->gb, 16) != 0xffff) {
1863 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: Didn't get subframe DSYNC\n");
1864 return AVERROR_INVALIDDATA;
1868 /* advance sub-sub-frame index */
1869 if(++subsubframe >= s->subsubframes[subframe]) {
1875 /* skip to next channel set */
1876 i = get_bits_count(&s->gb);
1877 if(start_posn + chset_fsize[chset] * 8 != i) {
1878 j = start_posn + chset_fsize[chset] * 8 - i;
1880 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: end of channel set,"
1881 " skipping further than expected (%d bits)\n", j);
1882 skip_bits_long(&s->gb, j);
1889 /* parse initial header for XXCH and dump details */
1890 static int dca_xxch_decode_frame(DCAContext *s)
1892 int hdr_size, spkmsk_bits, num_chsets, core_spk, hdr_pos;
1893 int i, chset, base_channel, chstart, fsize[8];
1895 /* assume header word has already been parsed */
1896 hdr_pos = get_bits_count(&s->gb) - 32;
1897 hdr_size = get_bits(&s->gb, 6) + 1;
1898 /*chhdr_crc =*/ skip_bits1(&s->gb);
1899 spkmsk_bits = get_bits(&s->gb, 5) + 1;
1900 num_chsets = get_bits(&s->gb, 2) + 1;
1902 for (i = 0; i < num_chsets; i++)
1903 fsize[i] = get_bits(&s->gb, 14) + 1;
1905 core_spk = get_bits(&s->gb, spkmsk_bits);
1906 s->xxch_core_spkmask = core_spk;
1907 s->xxch_nbits_spk_mask = spkmsk_bits;
1908 s->xxch_dmix_embedded = 0;
1910 /* skip to the end of the header */
1911 i = get_bits_count(&s->gb);
1912 if (hdr_pos + hdr_size * 8 > i)
1913 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1915 for (chset = 0; chset < num_chsets; chset++) {
1916 chstart = get_bits_count(&s->gb);
1917 base_channel = s->prim_channels;
1918 s->xxch_chset = chset;
1920 /* XXCH and Core headers differ, see 6.4.2 "XXCH Channel Set Header" vs.
1921 5.3.2 "Primary Audio Coding Header", DTS Spec 1.3.1 */
1922 dca_parse_audio_coding_header(s, base_channel, 1);
1924 /* decode channel data */
1925 for (i = 0; i < (s->sample_blocks / 8); i++) {
1926 if (dca_decode_block(s, base_channel, i)) {
1927 av_log(s->avctx, AV_LOG_ERROR,
1928 "Error decoding DTS-XXCH extension\n");
1933 /* skip to end of this section */
1934 i = get_bits_count(&s->gb);
1935 if (chstart + fsize[chset] * 8 > i)
1936 skip_bits_long(&s->gb, chstart + fsize[chset] * 8 - i);
1938 s->xxch_chset = num_chsets;
1944 * Parse extension substream header (HD)
1946 static void dca_exss_parse_header(DCAContext *s)
1953 int active_ss_mask[8];
1959 if (get_bits_left(&s->gb) < 52)
1962 start_posn = get_bits_count(&s->gb) - 32;
1964 skip_bits(&s->gb, 8); // user data
1965 ss_index = get_bits(&s->gb, 2);
1967 blownup = get_bits1(&s->gb);
1968 hdrsize = get_bits(&s->gb, 8 + 4 * blownup) + 1; // header_size
1969 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1971 s->static_fields = get_bits1(&s->gb);
1972 if (s->static_fields) {
1973 skip_bits(&s->gb, 2); // reference clock code
1974 skip_bits(&s->gb, 3); // frame duration code
1976 if (get_bits1(&s->gb))
1977 skip_bits_long(&s->gb, 36); // timestamp
1979 /* a single stream can contain multiple audio assets that can be
1980 * combined to form multiple audio presentations */
1982 num_audiop = get_bits(&s->gb, 3) + 1;
1983 if (num_audiop > 1) {
1984 avpriv_request_sample(s->avctx,
1985 "Multiple DTS-HD audio presentations");
1986 /* ignore such streams for now */
1990 num_assets = get_bits(&s->gb, 3) + 1;
1991 if (num_assets > 1) {
1992 avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
1993 /* ignore such streams for now */
1997 for (i = 0; i < num_audiop; i++)
1998 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
2000 for (i = 0; i < num_audiop; i++)
2001 for (j = 0; j <= ss_index; j++)
2002 if (active_ss_mask[i] & (1 << j))
2003 skip_bits(&s->gb, 8); // active asset mask
2005 s->mix_metadata = get_bits1(&s->gb);
2006 if (s->mix_metadata) {
2007 int mix_out_mask_size;
2009 skip_bits(&s->gb, 2); // adjustment level
2010 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
2011 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
2013 for (i = 0; i < s->num_mix_configs; i++) {
2014 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
2015 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
2020 for (i = 0; i < num_assets; i++)
2021 asset_size[i] = get_bits_long(&s->gb, 16 + 4 * blownup);
2023 for (i = 0; i < num_assets; i++) {
2024 if (dca_exss_parse_asset_header(s))
2028 /* not parsed further, we were only interested in the extensions mask
2029 * from the asset header */
2031 if (num_assets > 0) {
2032 j = get_bits_count(&s->gb);
2033 if (start_posn + hdrsize * 8 > j)
2034 skip_bits_long(&s->gb, start_posn + hdrsize * 8 - j);
2036 for (i = 0; i < num_assets; i++) {
2037 start_posn = get_bits_count(&s->gb);
2038 mkr = get_bits_long(&s->gb, 32);
2040 /* parse extensions that we know about */
2041 if (mkr == 0x655e315e) {
2042 dca_xbr_parse_frame(s);
2043 } else if (mkr == 0x47004a03) {
2044 dca_xxch_decode_frame(s);
2045 s->core_ext_mask |= DCA_EXT_XXCH; /* xxx use for chan reordering */
2047 av_log(s->avctx, AV_LOG_DEBUG,
2048 "DTS-ExSS: unknown marker = 0x%08x\n", mkr);
2051 /* skip to end of block */
2052 j = get_bits_count(&s->gb);
2053 if (start_posn + asset_size[i] * 8 > j)
2054 skip_bits_long(&s->gb, start_posn + asset_size[i] * 8 - j);
2060 * Main frame decoding function
2061 * FIXME add arguments
2063 static int dca_decode_frame(AVCodecContext *avctx, void *data,
2064 int *got_frame_ptr, AVPacket *avpkt)
2066 AVFrame *frame = data;
2067 const uint8_t *buf = avpkt->data;
2068 int buf_size = avpkt->size;
2072 int num_core_channels = 0;
2074 float **samples_flt;
2077 DCAContext *s = avctx->priv_data;
2079 int channels, full_channels;
2091 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
2092 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
2093 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
2094 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
2095 return AVERROR_INVALIDDATA;
2098 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
2099 if ((ret = dca_parse_frame_header(s)) < 0) {
2100 //seems like the frame is corrupt, try with the next one
2103 //set AVCodec values with parsed data
2104 avctx->sample_rate = s->sample_rate;
2105 avctx->bit_rate = s->bit_rate;
2107 s->profile = FF_PROFILE_DTS;
2109 for (i = 0; i < (s->sample_blocks / 8); i++) {
2110 if ((ret = dca_decode_block(s, 0, i))) {
2111 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
2116 /* record number of core channels incase less than max channels are requested */
2117 num_core_channels = s->prim_channels;
2120 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
2122 s->core_ext_mask = 0;
2124 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
2126 /* only scan for extensions if ext_descr was unknown or indicated a
2127 * supported XCh extension */
2128 if (s->core_ext_mask < 0 || s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH)) {
2130 /* if ext_descr was unknown, clear s->core_ext_mask so that the
2131 * extensions scan can fill it up */
2132 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
2134 /* extensions start at 32-bit boundaries into bitstream */
2135 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2137 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
2138 uint32_t bits = get_bits_long(&s->gb, 32);
2142 int ext_amode, xch_fsize;
2144 s->xch_base_channel = s->prim_channels;
2146 /* validate sync word using XCHFSIZE field */
2147 xch_fsize = show_bits(&s->gb, 10);
2148 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
2149 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
2152 /* skip length-to-end-of-frame field for the moment */
2153 skip_bits(&s->gb, 10);
2155 s->core_ext_mask |= DCA_EXT_XCH;
2157 /* extension amode(number of channels in extension) should be 1 */
2158 /* AFAIK XCh is not used for more channels */
2159 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
2160 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
2161 " supported!\n", ext_amode);
2165 if (s->xch_base_channel < 2) {
2166 avpriv_request_sample(avctx, "XCh with fewer than 2 base channels");
2170 /* much like core primary audio coding header */
2171 dca_parse_audio_coding_header(s, s->xch_base_channel, 0);
2173 for (i = 0; i < (s->sample_blocks / 8); i++)
2174 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
2175 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
2183 /* XXCh: extended channels */
2184 /* usually found either in core or HD part in DTS-HD HRA streams,
2185 * but not in DTS-ES which contains XCh extensions instead */
2186 s->core_ext_mask |= DCA_EXT_XXCH;
2187 dca_xxch_decode_frame(s);
2191 int fsize96 = show_bits(&s->gb, 12) + 1;
2192 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
2195 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
2196 get_bits_count(&s->gb));
2197 skip_bits(&s->gb, 12);
2198 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
2199 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
2201 s->core_ext_mask |= DCA_EXT_X96;
2206 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2209 /* no supported extensions, skip the rest of the core substream */
2210 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
2213 if (s->core_ext_mask & DCA_EXT_X96)
2214 s->profile = FF_PROFILE_DTS_96_24;
2215 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
2216 s->profile = FF_PROFILE_DTS_ES;
2218 /* check for ExSS (HD part) */
2219 if (s->dca_buffer_size - s->frame_size > 32 &&
2220 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
2221 dca_exss_parse_header(s);
2223 avctx->profile = s->profile;
2225 full_channels = channels = s->prim_channels + !!s->lfe;
2227 /* If we have XXCH then the channel layout is managed differently */
2228 /* note that XLL will also have another way to do things */
2229 if (!(s->core_ext_mask & DCA_EXT_XXCH)
2230 || (s->core_ext_mask & DCA_EXT_XXCH && avctx->request_channels > 0
2231 && avctx->request_channels
2232 < num_core_channels + !!s->lfe + s->xxch_chset_nch[0]))
2233 { /* xxx should also do MA extensions */
2234 if (s->amode < 16) {
2235 avctx->channel_layout = dca_core_channel_layout[s->amode];
2237 if (s->xch_present && (!avctx->request_channels ||
2238 avctx->request_channels
2239 > num_core_channels + !!s->lfe)) {
2240 avctx->channel_layout |= AV_CH_BACK_CENTER;
2242 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
2243 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
2245 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
2247 if (s->channel_order_tab[s->xch_base_channel] < 0)
2248 return AVERROR_INVALIDDATA;
2250 channels = num_core_channels + !!s->lfe;
2251 s->xch_present = 0; /* disable further xch processing */
2253 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
2254 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
2256 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
2259 if (channels > !!s->lfe &&
2260 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
2261 return AVERROR_INVALIDDATA;
2263 if (av_get_channel_layout_nb_channels(avctx->channel_layout) != channels) {
2264 av_log(avctx, AV_LOG_ERROR, "Number of channels %d mismatches layout %d\n", channels, av_get_channel_layout_nb_channels(avctx->channel_layout));
2265 return AVERROR_INVALIDDATA;
2268 if (avctx->request_channels == 2 && s->prim_channels > 2) {
2270 s->output = DCA_STEREO;
2271 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
2273 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
2274 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
2275 s->channel_order_tab = dca_channel_order_native;
2277 s->lfe_index = dca_lfe_index[s->amode];
2279 av_log(avctx, AV_LOG_ERROR,
2280 "Non standard configuration %d !\n", s->amode);
2281 return AVERROR_INVALIDDATA;
2284 s->xxch_dmix_embedded = 0;
2286 /* we only get here if an XXCH channel set can be added to the mix */
2287 channel_mask = s->xxch_core_spkmask;
2289 if (avctx->request_channels > 0
2290 && avctx->request_channels < s->prim_channels) {
2291 channels = num_core_channels + !!s->lfe;
2292 for (i = 0; i < s->xxch_chset && channels + s->xxch_chset_nch[i]
2293 <= avctx->request_channels; i++) {
2294 channels += s->xxch_chset_nch[i];
2295 channel_mask |= s->xxch_spk_masks[i];
2298 channels = s->prim_channels + !!s->lfe;
2299 for (i = 0; i < s->xxch_chset; i++) {
2300 channel_mask |= s->xxch_spk_masks[i];
2304 /* Given the DTS spec'ed channel mask, generate an avcodec version */
2306 for (i = 0; i < s->xxch_nbits_spk_mask; ++i) {
2307 if (channel_mask & (1 << i)) {
2308 channel_layout |= map_xxch_to_native[i];
2312 /* make sure that we have managed to get equivelant dts/avcodec channel
2313 * masks in some sense -- unfortunately some channels could overlap */
2314 if (av_popcount(channel_mask) != av_popcount(channel_layout)) {
2315 av_log(avctx, AV_LOG_DEBUG,
2316 "DTS-XXCH: Inconsistant avcodec/dts channel layouts\n");
2317 return AVERROR_INVALIDDATA;
2320 avctx->channel_layout = channel_layout;
2322 if (!(avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE)) {
2323 /* Estimate DTS --> avcodec ordering table */
2324 for (chset = -1, j = 0; chset < s->xxch_chset; ++chset) {
2325 mask = chset >= 0 ? s->xxch_spk_masks[chset]
2326 : s->xxch_core_spkmask;
2327 for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
2328 if (mask & ~(DCA_XXCH_LFE1 | DCA_XXCH_LFE2) & (1 << i)) {
2329 lavc = map_xxch_to_native[i];
2330 posn = av_popcount(channel_layout & (lavc - 1));
2331 s->xxch_order_tab[j++] = posn;
2336 s->lfe_index = av_popcount(channel_layout & (AV_CH_LOW_FREQUENCY-1));
2337 } else { /* native ordering */
2338 for (i = 0; i < channels; i++)
2339 s->xxch_order_tab[i] = i;
2341 s->lfe_index = channels - 1;
2344 s->channel_order_tab = s->xxch_order_tab;
2347 if (avctx->channels != channels) {
2348 if (avctx->channels)
2349 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
2350 avctx->channels = channels;
2353 /* get output buffer */
2354 frame->nb_samples = 256 * (s->sample_blocks / 8);
2355 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
2357 samples_flt = (float **)frame->extended_data;
2359 /* allocate buffer for extra channels if downmixing */
2360 if (avctx->channels < full_channels) {
2361 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
2363 avctx->sample_fmt, 0);
2367 av_fast_malloc(&s->extra_channels_buffer,
2368 &s->extra_channels_buffer_size, ret);
2369 if (!s->extra_channels_buffer)
2370 return AVERROR(ENOMEM);
2372 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
2373 s->extra_channels_buffer,
2374 full_channels - channels,
2375 frame->nb_samples, avctx->sample_fmt, 0);
2380 /* filter to get final output */
2381 for (i = 0; i < (s->sample_blocks / 8); i++) {
2384 for (ch = 0; ch < channels; ch++)
2385 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
2386 for (; ch < full_channels; ch++)
2387 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
2389 dca_filter_channels(s, i);
2391 /* If this was marked as a DTS-ES stream we need to subtract back- */
2392 /* channel from SL & SR to remove matrixed back-channel signal */
2393 if ((s->source_pcm_res & 1) && s->xch_present) {
2394 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
2395 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
2396 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
2397 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
2398 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
2401 /* If stream contains XXCH, we might need to undo an embedded downmix */
2402 if (s->xxch_dmix_embedded) {
2403 /* Loop over channel sets in turn */
2404 ch = num_core_channels;
2405 for (chset = 0; chset < s->xxch_chset; chset++) {
2406 endch = ch + s->xxch_chset_nch[chset];
2407 mask = s->xxch_dmix_embedded;
2410 for (j = ch; j < endch; j++) {
2411 if (mask & (1 << j)) { /* this channel has been mixed-out */
2412 src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2413 for (k = 0; k < endch; k++) {
2414 achan = s->channel_order_tab[k];
2415 scale = s->xxch_dmix_coeff[j][k];
2417 dst_chan = s->samples_chanptr[achan];
2418 s->fdsp.vector_fmac_scalar(dst_chan, src_chan,
2425 /* if a downmix has been embedded then undo the pre-scaling */
2426 if ((mask & (1 << ch)) && s->xxch_dmix_sf[chset] != 1.0f) {
2427 scale = s->xxch_dmix_sf[chset];
2429 for (j = 0; j < ch; j++) {
2430 src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2431 for (k = 0; k < 256; k++)
2432 src_chan[k] *= scale;
2435 /* LFE channel is always part of core, scale if it exists */
2437 src_chan = s->samples_chanptr[s->lfe_index];
2438 for (k = 0; k < 256; k++)
2439 src_chan[k] *= scale;
2449 /* update lfe history */
2450 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
2451 for (i = 0; i < 2 * s->lfe * 4; i++)
2452 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
2462 * DCA initialization
2464 * @param avctx pointer to the AVCodecContext
2467 static av_cold int dca_decode_init(AVCodecContext *avctx)
2469 DCAContext *s = avctx->priv_data;
2474 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
2475 ff_mdct_init(&s->imdct, 6, 1, 1.0);
2476 ff_synth_filter_init(&s->synth);
2477 ff_dcadsp_init(&s->dcadsp);
2478 ff_fmt_convert_init(&s->fmt_conv, avctx);
2480 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
2482 /* allow downmixing to stereo */
2483 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
2484 avctx->request_channels == 2) {
2485 avctx->channels = avctx->request_channels;
2491 static av_cold int dca_decode_end(AVCodecContext *avctx)
2493 DCAContext *s = avctx->priv_data;
2494 ff_mdct_end(&s->imdct);
2495 av_freep(&s->extra_channels_buffer);
2499 static const AVProfile profiles[] = {
2500 { FF_PROFILE_DTS, "DTS" },
2501 { FF_PROFILE_DTS_ES, "DTS-ES" },
2502 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2503 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2504 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2505 { FF_PROFILE_UNKNOWN },
2508 AVCodec ff_dca_decoder = {
2510 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2511 .type = AVMEDIA_TYPE_AUDIO,
2512 .id = AV_CODEC_ID_DTS,
2513 .priv_data_size = sizeof(DCAContext),
2514 .init = dca_decode_init,
2515 .decode = dca_decode_frame,
2516 .close = dca_decode_end,
2517 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2518 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
2519 AV_SAMPLE_FMT_NONE },
2520 .profiles = NULL_IF_CONFIG_SMALL(profiles),