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 avpriv_request_sample(s->avctx, "LFE = 3");
741 return AVERROR_PATCHWELCOME;
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;
1117 int sb_act = s->subband_activity[chans];
1120 scale *= sqrt(1 / 8.0);
1123 if (!s->multirate_inter) /* Non-perfect reconstruction */
1124 prCoeff = fir_32bands_nonperfect;
1125 else /* Perfect reconstruction */
1126 prCoeff = fir_32bands_perfect;
1128 for (i = sb_act; i < 32; i++)
1131 /* Reconstructed channel sample index */
1132 for (subindex = 0; subindex < 8; subindex++) {
1133 /* Load in one sample from each subband and clear inactive subbands */
1134 for (i = 0; i < sb_act; i++) {
1135 unsigned sign = (i - 1) & 2;
1136 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
1137 AV_WN32A(&s->raXin[i], v);
1140 s->synth.synth_filter_float(&s->imdct,
1141 s->subband_fir_hist[chans],
1142 &s->hist_index[chans],
1143 s->subband_fir_noidea[chans], prCoeff,
1144 samples_out, s->raXin, scale);
1149 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
1150 int num_deci_sample, float *samples_in,
1151 float *samples_out, float scale)
1153 /* samples_in: An array holding decimated samples.
1154 * Samples in current subframe starts from samples_in[0],
1155 * while samples_in[-1], samples_in[-2], ..., stores samples
1156 * from last subframe as history.
1158 * samples_out: An array holding interpolated samples
1162 const float *prCoeff;
1165 /* Select decimation filter */
1166 if (decimation_select == 1) {
1168 prCoeff = lfe_fir_128;
1171 prCoeff = lfe_fir_64;
1174 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1175 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1177 samples_out += 2 * decifactor;
1181 /* downmixing routines */
1182 #define MIX_REAR1(samples, s1, rs, coef) \
1183 samples[0][i] += samples[s1][i] * coef[rs][0]; \
1184 samples[1][i] += samples[s1][i] * coef[rs][1];
1186 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1187 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1188 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1190 #define MIX_FRONT3(samples, coef) \
1191 t = samples[c][i]; \
1192 u = samples[l][i]; \
1193 v = samples[r][i]; \
1194 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1195 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1197 #define DOWNMIX_TO_STEREO(op1, op2) \
1198 for (i = 0; i < 256; i++) { \
1203 static void dca_downmix(float **samples, int srcfmt,
1204 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1205 const int8_t *channel_mapping)
1207 int c, l, r, sl, sr, s;
1210 float coef[DCA_PRIM_CHANNELS_MAX][2];
1212 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1213 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1214 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1220 case DCA_STEREO_TOTAL:
1221 case DCA_STEREO_SUMDIFF:
1223 av_log(NULL, AV_LOG_ERROR, "Not implemented!\n");
1228 c = channel_mapping[0];
1229 l = channel_mapping[1];
1230 r = channel_mapping[2];
1231 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1234 s = channel_mapping[2];
1235 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1238 c = channel_mapping[0];
1239 l = channel_mapping[1];
1240 r = channel_mapping[2];
1241 s = channel_mapping[3];
1242 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1243 MIX_REAR1(samples, s, 3, coef));
1246 sl = channel_mapping[2];
1247 sr = channel_mapping[3];
1248 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1251 c = channel_mapping[0];
1252 l = channel_mapping[1];
1253 r = channel_mapping[2];
1254 sl = channel_mapping[3];
1255 sr = channel_mapping[4];
1256 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1257 MIX_REAR2(samples, sl, sr, 3, coef));
1263 #ifndef decode_blockcodes
1264 /* Very compact version of the block code decoder that does not use table
1265 * look-up but is slightly slower */
1266 static int decode_blockcode(int code, int levels, int32_t *values)
1269 int offset = (levels - 1) >> 1;
1271 for (i = 0; i < 4; i++) {
1272 int div = FASTDIV(code, levels);
1273 values[i] = code - offset - div * levels;
1280 static int decode_blockcodes(int code1, int code2, int levels, int32_t *values)
1282 return decode_blockcode(code1, levels, values) |
1283 decode_blockcode(code2, levels, values + 4);
1287 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1288 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1290 #ifndef int8x8_fmul_int32
1291 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1293 float fscale = scale / 16.0;
1295 for (i = 0; i < 8; i++)
1296 dst[i] = src[i] * fscale;
1300 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1303 int subsubframe = s->current_subsubframe;
1305 const float *quant_step_table;
1308 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1309 LOCAL_ALIGNED_16(int32_t, block, [8]);
1315 /* Select quantization step size table */
1316 if (s->bit_rate_index == 0x1f)
1317 quant_step_table = lossless_quant_d;
1319 quant_step_table = lossy_quant_d;
1321 for (k = base_channel; k < s->prim_channels; k++) {
1322 if (get_bits_left(&s->gb) < 0)
1323 return AVERROR_INVALIDDATA;
1325 for (l = 0; l < s->vq_start_subband[k]; l++) {
1328 /* Select the mid-tread linear quantizer */
1329 int abits = s->bitalloc[k][l];
1331 float quant_step_size = quant_step_table[abits];
1334 * Determine quantization index code book and its type
1337 /* Select quantization index code book */
1338 int sel = s->quant_index_huffman[k][abits];
1341 * Extract bits from the bit stream
1344 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1346 /* Deal with transients */
1347 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1348 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1349 s->scalefactor_adj[k][sel];
1351 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1354 int block_code1, block_code2, size, levels, err;
1356 size = abits_sizes[abits - 1];
1357 levels = abits_levels[abits - 1];
1359 block_code1 = get_bits(&s->gb, size);
1360 block_code2 = get_bits(&s->gb, size);
1361 err = decode_blockcodes(block_code1, block_code2,
1364 av_log(s->avctx, AV_LOG_ERROR,
1365 "ERROR: block code look-up failed\n");
1366 return AVERROR_INVALIDDATA;
1370 for (m = 0; m < 8; m++)
1371 block[m] = get_sbits(&s->gb, abits - 3);
1375 for (m = 0; m < 8; m++)
1376 block[m] = get_bitalloc(&s->gb,
1377 &dca_smpl_bitalloc[abits], sel);
1380 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1385 * Inverse ADPCM if in prediction mode
1387 if (s->prediction_mode[k][l]) {
1389 for (m = 0; m < 8; m++) {
1390 for (n = 1; n <= 4; n++)
1392 subband_samples[k][l][m] +=
1393 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1394 subband_samples[k][l][m - n] / 8192);
1395 else if (s->predictor_history)
1396 subband_samples[k][l][m] +=
1397 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1398 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1404 * Decode VQ encoded high frequencies
1406 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1407 /* 1 vector -> 32 samples but we only need the 8 samples
1408 * for this subsubframe. */
1409 int hfvq = s->high_freq_vq[k][l];
1411 if (!s->debug_flag & 0x01) {
1412 av_log(s->avctx, AV_LOG_DEBUG,
1413 "Stream with high frequencies VQ coding\n");
1414 s->debug_flag |= 0x01;
1417 int8x8_fmul_int32(subband_samples[k][l],
1418 &high_freq_vq[hfvq][subsubframe * 8],
1419 s->scale_factor[k][l][0]);
1423 /* Check for DSYNC after subsubframe */
1424 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1425 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1427 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1430 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1431 return AVERROR_INVALIDDATA;
1435 /* Backup predictor history for adpcm */
1436 for (k = base_channel; k < s->prim_channels; k++)
1437 for (l = 0; l < s->vq_start_subband[k]; l++)
1438 memcpy(s->subband_samples_hist[k][l],
1439 &subband_samples[k][l][4],
1440 4 * sizeof(subband_samples[0][0][0]));
1445 static int dca_filter_channels(DCAContext *s, int block_index)
1447 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1450 /* 32 subbands QMF */
1451 for (k = 0; k < s->prim_channels; k++) {
1452 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1453 0, 8388608.0, 8388608.0 };*/
1454 if (s->channel_order_tab[k] >= 0)
1455 qmf_32_subbands(s, k, subband_samples[k],
1456 s->samples_chanptr[s->channel_order_tab[k]],
1457 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1461 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1462 dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
1465 /* Generate LFE samples for this subsubframe FIXME!!! */
1466 if (s->output & DCA_LFE) {
1467 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1468 s->lfe_data + 2 * s->lfe * (block_index + 4),
1469 s->samples_chanptr[s->lfe_index],
1470 1.0 / (256.0 * 32768.0));
1471 /* Outputs 20bits pcm samples */
1478 static int dca_subframe_footer(DCAContext *s, int base_channel)
1480 int aux_data_count = 0, i;
1483 * Unpack optional information
1486 /* presumably optional information only appears in the core? */
1487 if (!base_channel) {
1489 skip_bits_long(&s->gb, 32);
1492 aux_data_count = get_bits(&s->gb, 6);
1494 for (i = 0; i < aux_data_count; i++)
1495 get_bits(&s->gb, 8);
1497 if (s->crc_present && (s->downmix || s->dynrange))
1498 get_bits(&s->gb, 16);
1505 * Decode a dca frame block
1507 * @param s pointer to the DCAContext
1510 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1515 if (s->current_subframe >= s->subframes) {
1516 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1517 s->current_subframe, s->subframes);
1518 return AVERROR_INVALIDDATA;
1521 if (!s->current_subsubframe) {
1523 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1525 /* Read subframe header */
1526 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1530 /* Read subsubframe */
1532 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1534 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1538 s->current_subsubframe++;
1539 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1540 s->current_subsubframe = 0;
1541 s->current_subframe++;
1543 if (s->current_subframe >= s->subframes) {
1545 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1547 /* Read subframe footer */
1548 if ((ret = dca_subframe_footer(s, base_channel)))
1556 * Return the number of channels in an ExSS speaker mask (HD)
1558 static int dca_exss_mask2count(int mask)
1560 /* count bits that mean speaker pairs twice */
1561 return av_popcount(mask) +
1562 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1563 DCA_EXSS_FRONT_LEFT_RIGHT |
1564 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1565 DCA_EXSS_WIDE_LEFT_RIGHT |
1566 DCA_EXSS_SIDE_LEFT_RIGHT |
1567 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1568 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1569 DCA_EXSS_REAR_LEFT_RIGHT |
1570 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1574 * Skip mixing coefficients of a single mix out configuration (HD)
1576 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1580 for (i = 0; i < channels; i++) {
1581 int mix_map_mask = get_bits(gb, out_ch);
1582 int num_coeffs = av_popcount(mix_map_mask);
1583 skip_bits_long(gb, num_coeffs * 6);
1588 * Parse extension substream asset header (HD)
1590 static int dca_exss_parse_asset_header(DCAContext *s)
1592 int header_pos = get_bits_count(&s->gb);
1595 int embedded_stereo = 0;
1596 int embedded_6ch = 0;
1597 int drc_code_present;
1598 int av_uninit(extensions_mask);
1601 if (get_bits_left(&s->gb) < 16)
1604 /* We will parse just enough to get to the extensions bitmask with which
1605 * we can set the profile value. */
1607 header_size = get_bits(&s->gb, 9) + 1;
1608 skip_bits(&s->gb, 3); // asset index
1610 if (s->static_fields) {
1611 if (get_bits1(&s->gb))
1612 skip_bits(&s->gb, 4); // asset type descriptor
1613 if (get_bits1(&s->gb))
1614 skip_bits_long(&s->gb, 24); // language descriptor
1616 if (get_bits1(&s->gb)) {
1617 /* How can one fit 1024 bytes of text here if the maximum value
1618 * for the asset header size field above was 512 bytes? */
1619 int text_length = get_bits(&s->gb, 10) + 1;
1620 if (get_bits_left(&s->gb) < text_length * 8)
1622 skip_bits_long(&s->gb, text_length * 8); // info text
1625 skip_bits(&s->gb, 5); // bit resolution - 1
1626 skip_bits(&s->gb, 4); // max sample rate code
1627 channels = get_bits(&s->gb, 8) + 1;
1629 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1630 int spkr_remap_sets;
1631 int spkr_mask_size = 16;
1635 embedded_stereo = get_bits1(&s->gb);
1637 embedded_6ch = get_bits1(&s->gb);
1639 if (get_bits1(&s->gb)) {
1640 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1641 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1644 spkr_remap_sets = get_bits(&s->gb, 3);
1646 for (i = 0; i < spkr_remap_sets; i++) {
1647 /* std layout mask for each remap set */
1648 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1651 for (i = 0; i < spkr_remap_sets; i++) {
1652 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1653 if (get_bits_left(&s->gb) < 0)
1656 for (j = 0; j < num_spkrs[i]; j++) {
1657 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1658 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1659 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1664 skip_bits(&s->gb, 3); // representation type
1668 drc_code_present = get_bits1(&s->gb);
1669 if (drc_code_present)
1670 get_bits(&s->gb, 8); // drc code
1672 if (get_bits1(&s->gb))
1673 skip_bits(&s->gb, 5); // dialog normalization code
1675 if (drc_code_present && embedded_stereo)
1676 get_bits(&s->gb, 8); // drc stereo code
1678 if (s->mix_metadata && get_bits1(&s->gb)) {
1679 skip_bits(&s->gb, 1); // external mix
1680 skip_bits(&s->gb, 6); // post mix gain code
1682 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1683 skip_bits(&s->gb, 3); // drc limit
1685 skip_bits(&s->gb, 8); // custom drc code
1687 if (get_bits1(&s->gb)) // channel specific scaling
1688 for (i = 0; i < s->num_mix_configs; i++)
1689 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1691 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1693 for (i = 0; i < s->num_mix_configs; i++) {
1694 if (get_bits_left(&s->gb) < 0)
1696 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1698 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1699 if (embedded_stereo)
1700 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1704 switch (get_bits(&s->gb, 2)) {
1705 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1706 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1707 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1708 case 3: extensions_mask = 0; /* aux coding */ break;
1711 /* not parsed further, we were only interested in the extensions mask */
1713 if (get_bits_left(&s->gb) < 0)
1716 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1717 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1720 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1722 if (extensions_mask & DCA_EXT_EXSS_XLL)
1723 s->profile = FF_PROFILE_DTS_HD_MA;
1724 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1726 s->profile = FF_PROFILE_DTS_HD_HRA;
1728 if (!(extensions_mask & DCA_EXT_CORE))
1729 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1730 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1731 av_log(s->avctx, AV_LOG_WARNING,
1732 "DTS extensions detection mismatch (%d, %d)\n",
1733 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1738 static int dca_xbr_parse_frame(DCAContext *s)
1740 int scale_table_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS][2];
1741 int active_bands[DCA_CHSETS_MAX][DCA_CHSET_CHANS_MAX];
1742 int abits_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS];
1743 int anctemp[DCA_CHSET_CHANS_MAX];
1744 int chset_fsize[DCA_CHSETS_MAX];
1745 int n_xbr_ch[DCA_CHSETS_MAX];
1746 int hdr_size, num_chsets, xbr_tmode, hdr_pos;
1747 int i, j, k, l, chset, chan_base;
1749 av_log(s->avctx, AV_LOG_DEBUG, "DTS-XBR: decoding XBR extension\n");
1751 /* get bit position of sync header */
1752 hdr_pos = get_bits_count(&s->gb) - 32;
1754 hdr_size = get_bits(&s->gb, 6) + 1;
1755 num_chsets = get_bits(&s->gb, 2) + 1;
1757 for(i = 0; i < num_chsets; i++)
1758 chset_fsize[i] = get_bits(&s->gb, 14) + 1;
1760 xbr_tmode = get_bits1(&s->gb);
1762 for(i = 0; i < num_chsets; i++) {
1763 n_xbr_ch[i] = get_bits(&s->gb, 3) + 1;
1764 k = get_bits(&s->gb, 2) + 5;
1765 for(j = 0; j < n_xbr_ch[i]; j++)
1766 active_bands[i][j] = get_bits(&s->gb, k) + 1;
1769 /* skip to the end of the header */
1770 i = get_bits_count(&s->gb);
1771 if(hdr_pos + hdr_size * 8 > i)
1772 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1774 /* loop over the channel data sets */
1775 /* only decode as many channels as we've decoded base data for */
1776 for(chset = 0, chan_base = 0;
1777 chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->prim_channels;
1778 chan_base += n_xbr_ch[chset++]) {
1779 int start_posn = get_bits_count(&s->gb);
1780 int subsubframe = 0;
1783 /* loop over subframes */
1784 for (k = 0; k < (s->sample_blocks / 8); k++) {
1785 /* parse header if we're on first subsubframe of a block */
1786 if(subsubframe == 0) {
1787 /* Parse subframe header */
1788 for(i = 0; i < n_xbr_ch[chset]; i++) {
1789 anctemp[i] = get_bits(&s->gb, 2) + 2;
1792 for(i = 0; i < n_xbr_ch[chset]; i++) {
1793 get_array(&s->gb, abits_high[i], active_bands[chset][i], anctemp[i]);
1796 for(i = 0; i < n_xbr_ch[chset]; i++) {
1797 anctemp[i] = get_bits(&s->gb, 3);
1798 if(anctemp[i] < 1) {
1799 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: SYNC ERROR\n");
1800 return AVERROR_INVALIDDATA;
1804 /* generate scale factors */
1805 for(i = 0; i < n_xbr_ch[chset]; i++) {
1806 const uint32_t *scale_table;
1809 if (s->scalefactor_huffman[chan_base+i] == 6) {
1810 scale_table = scale_factor_quant7;
1812 scale_table = scale_factor_quant6;
1817 for(j = 0; j < active_bands[chset][i]; j++) {
1818 if(abits_high[i][j] > 0) {
1819 scale_table_high[i][j][0] =
1820 scale_table[get_bits(&s->gb, nbits)];
1822 if(xbr_tmode && s->transition_mode[i][j]) {
1823 scale_table_high[i][j][1] =
1824 scale_table[get_bits(&s->gb, nbits)];
1831 /* decode audio array for this block */
1832 for(i = 0; i < n_xbr_ch[chset]; i++) {
1833 for(j = 0; j < active_bands[chset][i]; j++) {
1834 const int xbr_abits = abits_high[i][j];
1835 const float quant_step_size = lossless_quant_d[xbr_abits];
1836 const int sfi = xbr_tmode && s->transition_mode[i][j] && subsubframe >= s->transition_mode[i][j];
1837 const float rscale = quant_step_size * scale_table_high[i][j][sfi];
1838 float *subband_samples = s->subband_samples[k][chan_base+i][j];
1845 get_array(&s->gb, block, 8, xbr_abits - 3);
1847 int block_code1, block_code2, size, levels, err;
1849 size = abits_sizes[xbr_abits - 1];
1850 levels = abits_levels[xbr_abits - 1];
1852 block_code1 = get_bits(&s->gb, size);
1853 block_code2 = get_bits(&s->gb, size);
1854 err = decode_blockcodes(block_code1, block_code2,
1857 av_log(s->avctx, AV_LOG_ERROR,
1858 "ERROR: DTS-XBR: block code look-up failed\n");
1859 return AVERROR_INVALIDDATA;
1863 /* scale & sum into subband */
1864 for(l = 0; l < 8; l++)
1865 subband_samples[l] += (float)block[l] * rscale;
1869 /* check DSYNC marker */
1870 if(s->aspf || subsubframe == s->subsubframes[subframe] - 1) {
1871 if(get_bits(&s->gb, 16) != 0xffff) {
1872 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: Didn't get subframe DSYNC\n");
1873 return AVERROR_INVALIDDATA;
1877 /* advance sub-sub-frame index */
1878 if(++subsubframe >= s->subsubframes[subframe]) {
1884 /* skip to next channel set */
1885 i = get_bits_count(&s->gb);
1886 if(start_posn + chset_fsize[chset] * 8 != i) {
1887 j = start_posn + chset_fsize[chset] * 8 - i;
1889 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: end of channel set,"
1890 " skipping further than expected (%d bits)\n", j);
1891 skip_bits_long(&s->gb, j);
1898 /* parse initial header for XXCH and dump details */
1899 static int dca_xxch_decode_frame(DCAContext *s)
1901 int hdr_size, spkmsk_bits, num_chsets, core_spk, hdr_pos;
1902 int i, chset, base_channel, chstart, fsize[8];
1904 /* assume header word has already been parsed */
1905 hdr_pos = get_bits_count(&s->gb) - 32;
1906 hdr_size = get_bits(&s->gb, 6) + 1;
1907 /*chhdr_crc =*/ skip_bits1(&s->gb);
1908 spkmsk_bits = get_bits(&s->gb, 5) + 1;
1909 num_chsets = get_bits(&s->gb, 2) + 1;
1911 for (i = 0; i < num_chsets; i++)
1912 fsize[i] = get_bits(&s->gb, 14) + 1;
1914 core_spk = get_bits(&s->gb, spkmsk_bits);
1915 s->xxch_core_spkmask = core_spk;
1916 s->xxch_nbits_spk_mask = spkmsk_bits;
1917 s->xxch_dmix_embedded = 0;
1919 /* skip to the end of the header */
1920 i = get_bits_count(&s->gb);
1921 if (hdr_pos + hdr_size * 8 > i)
1922 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1924 for (chset = 0; chset < num_chsets; chset++) {
1925 chstart = get_bits_count(&s->gb);
1926 base_channel = s->prim_channels;
1927 s->xxch_chset = chset;
1929 /* XXCH and Core headers differ, see 6.4.2 "XXCH Channel Set Header" vs.
1930 5.3.2 "Primary Audio Coding Header", DTS Spec 1.3.1 */
1931 dca_parse_audio_coding_header(s, base_channel, 1);
1933 /* decode channel data */
1934 for (i = 0; i < (s->sample_blocks / 8); i++) {
1935 if (dca_decode_block(s, base_channel, i)) {
1936 av_log(s->avctx, AV_LOG_ERROR,
1937 "Error decoding DTS-XXCH extension\n");
1942 /* skip to end of this section */
1943 i = get_bits_count(&s->gb);
1944 if (chstart + fsize[chset] * 8 > i)
1945 skip_bits_long(&s->gb, chstart + fsize[chset] * 8 - i);
1947 s->xxch_chset = num_chsets;
1953 * Parse extension substream header (HD)
1955 static void dca_exss_parse_header(DCAContext *s)
1962 int active_ss_mask[8];
1968 if (get_bits_left(&s->gb) < 52)
1971 start_posn = get_bits_count(&s->gb) - 32;
1973 skip_bits(&s->gb, 8); // user data
1974 ss_index = get_bits(&s->gb, 2);
1976 blownup = get_bits1(&s->gb);
1977 hdrsize = get_bits(&s->gb, 8 + 4 * blownup) + 1; // header_size
1978 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1980 s->static_fields = get_bits1(&s->gb);
1981 if (s->static_fields) {
1982 skip_bits(&s->gb, 2); // reference clock code
1983 skip_bits(&s->gb, 3); // frame duration code
1985 if (get_bits1(&s->gb))
1986 skip_bits_long(&s->gb, 36); // timestamp
1988 /* a single stream can contain multiple audio assets that can be
1989 * combined to form multiple audio presentations */
1991 num_audiop = get_bits(&s->gb, 3) + 1;
1992 if (num_audiop > 1) {
1993 avpriv_request_sample(s->avctx,
1994 "Multiple DTS-HD audio presentations");
1995 /* ignore such streams for now */
1999 num_assets = get_bits(&s->gb, 3) + 1;
2000 if (num_assets > 1) {
2001 avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
2002 /* ignore such streams for now */
2006 for (i = 0; i < num_audiop; i++)
2007 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
2009 for (i = 0; i < num_audiop; i++)
2010 for (j = 0; j <= ss_index; j++)
2011 if (active_ss_mask[i] & (1 << j))
2012 skip_bits(&s->gb, 8); // active asset mask
2014 s->mix_metadata = get_bits1(&s->gb);
2015 if (s->mix_metadata) {
2016 int mix_out_mask_size;
2018 skip_bits(&s->gb, 2); // adjustment level
2019 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
2020 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
2022 for (i = 0; i < s->num_mix_configs; i++) {
2023 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
2024 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
2029 for (i = 0; i < num_assets; i++)
2030 asset_size[i] = get_bits_long(&s->gb, 16 + 4 * blownup);
2032 for (i = 0; i < num_assets; i++) {
2033 if (dca_exss_parse_asset_header(s))
2037 /* not parsed further, we were only interested in the extensions mask
2038 * from the asset header */
2040 if (num_assets > 0) {
2041 j = get_bits_count(&s->gb);
2042 if (start_posn + hdrsize * 8 > j)
2043 skip_bits_long(&s->gb, start_posn + hdrsize * 8 - j);
2045 for (i = 0; i < num_assets; i++) {
2046 start_posn = get_bits_count(&s->gb);
2047 mkr = get_bits_long(&s->gb, 32);
2049 /* parse extensions that we know about */
2050 if (mkr == 0x655e315e) {
2051 dca_xbr_parse_frame(s);
2052 } else if (mkr == 0x47004a03) {
2053 dca_xxch_decode_frame(s);
2054 s->core_ext_mask |= DCA_EXT_XXCH; /* xxx use for chan reordering */
2056 av_log(s->avctx, AV_LOG_DEBUG,
2057 "DTS-ExSS: unknown marker = 0x%08x\n", mkr);
2060 /* skip to end of block */
2061 j = get_bits_count(&s->gb);
2062 if (start_posn + asset_size[i] * 8 > j)
2063 skip_bits_long(&s->gb, start_posn + asset_size[i] * 8 - j);
2069 * Main frame decoding function
2070 * FIXME add arguments
2072 static int dca_decode_frame(AVCodecContext *avctx, void *data,
2073 int *got_frame_ptr, AVPacket *avpkt)
2075 AVFrame *frame = data;
2076 const uint8_t *buf = avpkt->data;
2077 int buf_size = avpkt->size;
2081 int num_core_channels = 0;
2083 float **samples_flt;
2086 DCAContext *s = avctx->priv_data;
2088 int channels, full_channels;
2100 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
2101 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
2102 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
2103 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
2104 return AVERROR_INVALIDDATA;
2107 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
2108 if ((ret = dca_parse_frame_header(s)) < 0) {
2109 //seems like the frame is corrupt, try with the next one
2112 //set AVCodec values with parsed data
2113 avctx->sample_rate = s->sample_rate;
2114 avctx->bit_rate = s->bit_rate;
2116 s->profile = FF_PROFILE_DTS;
2118 for (i = 0; i < (s->sample_blocks / 8); i++) {
2119 if ((ret = dca_decode_block(s, 0, i))) {
2120 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
2125 /* record number of core channels incase less than max channels are requested */
2126 num_core_channels = s->prim_channels;
2129 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
2131 s->core_ext_mask = 0;
2133 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
2135 /* only scan for extensions if ext_descr was unknown or indicated a
2136 * supported XCh extension */
2137 if (s->core_ext_mask < 0 || s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH)) {
2139 /* if ext_descr was unknown, clear s->core_ext_mask so that the
2140 * extensions scan can fill it up */
2141 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
2143 /* extensions start at 32-bit boundaries into bitstream */
2144 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2146 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
2147 uint32_t bits = get_bits_long(&s->gb, 32);
2151 int ext_amode, xch_fsize;
2153 s->xch_base_channel = s->prim_channels;
2155 /* validate sync word using XCHFSIZE field */
2156 xch_fsize = show_bits(&s->gb, 10);
2157 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
2158 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
2161 /* skip length-to-end-of-frame field for the moment */
2162 skip_bits(&s->gb, 10);
2164 s->core_ext_mask |= DCA_EXT_XCH;
2166 /* extension amode(number of channels in extension) should be 1 */
2167 /* AFAIK XCh is not used for more channels */
2168 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
2169 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
2170 " supported!\n", ext_amode);
2174 if (s->xch_base_channel < 2) {
2175 avpriv_request_sample(avctx, "XCh with fewer than 2 base channels");
2179 /* much like core primary audio coding header */
2180 dca_parse_audio_coding_header(s, s->xch_base_channel, 0);
2182 for (i = 0; i < (s->sample_blocks / 8); i++)
2183 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
2184 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
2192 /* XXCh: extended channels */
2193 /* usually found either in core or HD part in DTS-HD HRA streams,
2194 * but not in DTS-ES which contains XCh extensions instead */
2195 s->core_ext_mask |= DCA_EXT_XXCH;
2196 dca_xxch_decode_frame(s);
2200 int fsize96 = show_bits(&s->gb, 12) + 1;
2201 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
2204 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
2205 get_bits_count(&s->gb));
2206 skip_bits(&s->gb, 12);
2207 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
2208 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
2210 s->core_ext_mask |= DCA_EXT_X96;
2215 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2218 /* no supported extensions, skip the rest of the core substream */
2219 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
2222 if (s->core_ext_mask & DCA_EXT_X96)
2223 s->profile = FF_PROFILE_DTS_96_24;
2224 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
2225 s->profile = FF_PROFILE_DTS_ES;
2227 /* check for ExSS (HD part) */
2228 if (s->dca_buffer_size - s->frame_size > 32 &&
2229 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
2230 dca_exss_parse_header(s);
2232 avctx->profile = s->profile;
2234 full_channels = channels = s->prim_channels + !!s->lfe;
2236 /* If we have XXCH then the channel layout is managed differently */
2237 /* note that XLL will also have another way to do things */
2238 if (!(s->core_ext_mask & DCA_EXT_XXCH)
2239 || (s->core_ext_mask & DCA_EXT_XXCH && avctx->request_channels > 0
2240 && avctx->request_channels
2241 < num_core_channels + !!s->lfe + s->xxch_chset_nch[0]))
2242 { /* xxx should also do MA extensions */
2243 if (s->amode < 16) {
2244 avctx->channel_layout = dca_core_channel_layout[s->amode];
2246 if (s->xch_present && (!avctx->request_channels ||
2247 avctx->request_channels
2248 > num_core_channels + !!s->lfe)) {
2249 avctx->channel_layout |= AV_CH_BACK_CENTER;
2251 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
2252 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
2254 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
2256 if (s->channel_order_tab[s->xch_base_channel] < 0)
2257 return AVERROR_INVALIDDATA;
2259 channels = num_core_channels + !!s->lfe;
2260 s->xch_present = 0; /* disable further xch processing */
2262 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
2263 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
2265 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
2268 if (channels > !!s->lfe &&
2269 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
2270 return AVERROR_INVALIDDATA;
2272 if (av_get_channel_layout_nb_channels(avctx->channel_layout) != channels) {
2273 av_log(avctx, AV_LOG_ERROR, "Number of channels %d mismatches layout %d\n", channels, av_get_channel_layout_nb_channels(avctx->channel_layout));
2274 return AVERROR_INVALIDDATA;
2277 if (avctx->request_channels == 2 && s->prim_channels > 2) {
2279 s->output = DCA_STEREO;
2280 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
2282 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
2283 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
2284 s->channel_order_tab = dca_channel_order_native;
2286 s->lfe_index = dca_lfe_index[s->amode];
2288 av_log(avctx, AV_LOG_ERROR,
2289 "Non standard configuration %d !\n", s->amode);
2290 return AVERROR_INVALIDDATA;
2293 s->xxch_dmix_embedded = 0;
2295 /* we only get here if an XXCH channel set can be added to the mix */
2296 channel_mask = s->xxch_core_spkmask;
2298 if (avctx->request_channels > 0
2299 && avctx->request_channels < s->prim_channels) {
2300 channels = num_core_channels + !!s->lfe;
2301 for (i = 0; i < s->xxch_chset && channels + s->xxch_chset_nch[i]
2302 <= avctx->request_channels; i++) {
2303 channels += s->xxch_chset_nch[i];
2304 channel_mask |= s->xxch_spk_masks[i];
2307 channels = s->prim_channels + !!s->lfe;
2308 for (i = 0; i < s->xxch_chset; i++) {
2309 channel_mask |= s->xxch_spk_masks[i];
2313 /* Given the DTS spec'ed channel mask, generate an avcodec version */
2315 for (i = 0; i < s->xxch_nbits_spk_mask; ++i) {
2316 if (channel_mask & (1 << i)) {
2317 channel_layout |= map_xxch_to_native[i];
2321 /* make sure that we have managed to get equivelant dts/avcodec channel
2322 * masks in some sense -- unfortunately some channels could overlap */
2323 if (av_popcount(channel_mask) != av_popcount(channel_layout)) {
2324 av_log(avctx, AV_LOG_DEBUG,
2325 "DTS-XXCH: Inconsistant avcodec/dts channel layouts\n");
2326 return AVERROR_INVALIDDATA;
2329 avctx->channel_layout = channel_layout;
2331 if (!(avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE)) {
2332 /* Estimate DTS --> avcodec ordering table */
2333 for (chset = -1, j = 0; chset < s->xxch_chset; ++chset) {
2334 mask = chset >= 0 ? s->xxch_spk_masks[chset]
2335 : s->xxch_core_spkmask;
2336 for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
2337 if (mask & ~(DCA_XXCH_LFE1 | DCA_XXCH_LFE2) & (1 << i)) {
2338 lavc = map_xxch_to_native[i];
2339 posn = av_popcount(channel_layout & (lavc - 1));
2340 s->xxch_order_tab[j++] = posn;
2345 s->lfe_index = av_popcount(channel_layout & (AV_CH_LOW_FREQUENCY-1));
2346 } else { /* native ordering */
2347 for (i = 0; i < channels; i++)
2348 s->xxch_order_tab[i] = i;
2350 s->lfe_index = channels - 1;
2353 s->channel_order_tab = s->xxch_order_tab;
2356 if (avctx->channels != channels) {
2357 if (avctx->channels)
2358 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
2359 avctx->channels = channels;
2362 /* get output buffer */
2363 frame->nb_samples = 256 * (s->sample_blocks / 8);
2364 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
2366 samples_flt = (float **)frame->extended_data;
2368 /* allocate buffer for extra channels if downmixing */
2369 if (avctx->channels < full_channels) {
2370 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
2372 avctx->sample_fmt, 0);
2376 av_fast_malloc(&s->extra_channels_buffer,
2377 &s->extra_channels_buffer_size, ret);
2378 if (!s->extra_channels_buffer)
2379 return AVERROR(ENOMEM);
2381 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
2382 s->extra_channels_buffer,
2383 full_channels - channels,
2384 frame->nb_samples, avctx->sample_fmt, 0);
2389 /* filter to get final output */
2390 for (i = 0; i < (s->sample_blocks / 8); i++) {
2393 for (ch = 0; ch < channels; ch++)
2394 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
2395 for (; ch < full_channels; ch++)
2396 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
2398 dca_filter_channels(s, i);
2400 /* If this was marked as a DTS-ES stream we need to subtract back- */
2401 /* channel from SL & SR to remove matrixed back-channel signal */
2402 if ((s->source_pcm_res & 1) && s->xch_present) {
2403 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
2404 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
2405 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
2406 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
2407 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
2410 /* If stream contains XXCH, we might need to undo an embedded downmix */
2411 if (s->xxch_dmix_embedded) {
2412 /* Loop over channel sets in turn */
2413 ch = num_core_channels;
2414 for (chset = 0; chset < s->xxch_chset; chset++) {
2415 endch = ch + s->xxch_chset_nch[chset];
2416 mask = s->xxch_dmix_embedded;
2419 for (j = ch; j < endch; j++) {
2420 if (mask & (1 << j)) { /* this channel has been mixed-out */
2421 src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2422 for (k = 0; k < endch; k++) {
2423 achan = s->channel_order_tab[k];
2424 scale = s->xxch_dmix_coeff[j][k];
2426 dst_chan = s->samples_chanptr[achan];
2427 s->fdsp.vector_fmac_scalar(dst_chan, src_chan,
2434 /* if a downmix has been embedded then undo the pre-scaling */
2435 if ((mask & (1 << ch)) && s->xxch_dmix_sf[chset] != 1.0f) {
2436 scale = s->xxch_dmix_sf[chset];
2438 for (j = 0; j < ch; j++) {
2439 src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2440 for (k = 0; k < 256; k++)
2441 src_chan[k] *= scale;
2444 /* LFE channel is always part of core, scale if it exists */
2446 src_chan = s->samples_chanptr[s->lfe_index];
2447 for (k = 0; k < 256; k++)
2448 src_chan[k] *= scale;
2458 /* update lfe history */
2459 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
2460 for (i = 0; i < 2 * s->lfe * 4; i++)
2461 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
2471 * DCA initialization
2473 * @param avctx pointer to the AVCodecContext
2476 static av_cold int dca_decode_init(AVCodecContext *avctx)
2478 DCAContext *s = avctx->priv_data;
2483 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
2484 ff_mdct_init(&s->imdct, 6, 1, 1.0);
2485 ff_synth_filter_init(&s->synth);
2486 ff_dcadsp_init(&s->dcadsp);
2487 ff_fmt_convert_init(&s->fmt_conv, avctx);
2489 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
2491 /* allow downmixing to stereo */
2492 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
2493 avctx->request_channels == 2) {
2494 avctx->channels = avctx->request_channels;
2500 static av_cold int dca_decode_end(AVCodecContext *avctx)
2502 DCAContext *s = avctx->priv_data;
2503 ff_mdct_end(&s->imdct);
2504 av_freep(&s->extra_channels_buffer);
2508 static const AVProfile profiles[] = {
2509 { FF_PROFILE_DTS, "DTS" },
2510 { FF_PROFILE_DTS_ES, "DTS-ES" },
2511 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2512 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2513 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2514 { FF_PROFILE_UNKNOWN },
2517 AVCodec ff_dca_decoder = {
2519 .type = AVMEDIA_TYPE_AUDIO,
2520 .id = AV_CODEC_ID_DTS,
2521 .priv_data_size = sizeof(DCAContext),
2522 .init = dca_decode_init,
2523 .decode = dca_decode_frame,
2524 .close = dca_decode_end,
2525 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2526 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2527 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
2528 AV_SAMPLE_FMT_NONE },
2529 .profiles = NULL_IF_CONFIG_SMALL(profiles),