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 for (j = base_channel; j < FFMIN(s->prim_channels, FF_ARRAY_ELEMS(dca_default_coeffs[am])); j++) {
972 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
973 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
978 /* Dynamic range coefficient */
979 if (!base_channel && s->dynrange)
980 s->dynrange_coef = get_bits(&s->gb, 8);
982 /* Side information CRC check word */
983 if (s->crc_present) {
984 get_bits(&s->gb, 16);
988 * Primary audio data arrays
991 /* VQ encoded high frequency subbands */
992 for (j = base_channel; j < s->prim_channels; j++)
993 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
994 /* 1 vector -> 32 samples */
995 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
997 /* Low frequency effect data */
998 if (!base_channel && s->lfe) {
1001 int lfe_samples = 2 * s->lfe * (4 + block_index);
1002 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
1005 for (j = lfe_samples; j < lfe_end_sample; j++) {
1006 /* Signed 8 bits int */
1007 s->lfe_data[j] = get_sbits(&s->gb, 8);
1010 /* Scale factor index */
1011 quant7 = get_bits(&s->gb, 8);
1013 avpriv_request_sample(s->avctx, "LFEScaleIndex larger than 127");
1014 return AVERROR_INVALIDDATA;
1016 s->lfe_scale_factor = scale_factor_quant7[quant7];
1018 /* Quantization step size * scale factor */
1019 lfe_scale = 0.035 * s->lfe_scale_factor;
1021 for (j = lfe_samples; j < lfe_end_sample; j++)
1022 s->lfe_data[j] *= lfe_scale;
1026 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
1027 s->subsubframes[s->current_subframe]);
1028 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
1029 s->partial_samples[s->current_subframe]);
1031 for (j = base_channel; j < s->prim_channels; j++) {
1032 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
1033 for (k = 0; k < s->subband_activity[j]; k++)
1034 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
1035 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1037 for (j = base_channel; j < s->prim_channels; j++) {
1038 for (k = 0; k < s->subband_activity[j]; k++)
1039 av_log(s->avctx, AV_LOG_DEBUG,
1040 "prediction coefs: %f, %f, %f, %f\n",
1041 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
1042 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
1043 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
1044 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
1046 for (j = base_channel; j < s->prim_channels; j++) {
1047 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
1048 for (k = 0; k < s->vq_start_subband[j]; k++)
1049 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
1050 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1052 for (j = base_channel; j < s->prim_channels; j++) {
1053 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
1054 for (k = 0; k < s->subband_activity[j]; k++)
1055 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[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, "Scale factor:");
1060 for (k = 0; k < s->subband_activity[j]; k++) {
1061 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
1062 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
1063 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
1064 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
1066 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1068 for (j = base_channel; j < s->prim_channels; j++) {
1069 if (s->joint_intensity[j] > 0) {
1070 int source_channel = s->joint_intensity[j] - 1;
1071 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
1072 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
1073 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
1074 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1077 if (!base_channel && s->prim_channels > 2 && s->downmix) {
1078 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
1079 for (j = 0; j < s->prim_channels; j++) {
1080 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
1081 dca_downmix_coeffs[s->downmix_coef[j][0]]);
1082 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
1083 dca_downmix_coeffs[s->downmix_coef[j][1]]);
1085 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1087 for (j = base_channel; j < s->prim_channels; j++)
1088 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
1089 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
1090 if (!base_channel && s->lfe) {
1091 int lfe_samples = 2 * s->lfe * (4 + block_index);
1092 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
1094 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
1095 for (j = lfe_samples; j < lfe_end_sample; j++)
1096 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
1097 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1104 static void qmf_32_subbands(DCAContext *s, int chans,
1105 float samples_in[32][8], float *samples_out,
1108 const float *prCoeff;
1111 int sb_act = s->subband_activity[chans];
1114 scale *= sqrt(1 / 8.0);
1117 if (!s->multirate_inter) /* Non-perfect reconstruction */
1118 prCoeff = fir_32bands_nonperfect;
1119 else /* Perfect reconstruction */
1120 prCoeff = fir_32bands_perfect;
1122 for (i = sb_act; i < 32; i++)
1125 /* Reconstructed channel sample index */
1126 for (subindex = 0; subindex < 8; subindex++) {
1127 /* Load in one sample from each subband and clear inactive subbands */
1128 for (i = 0; i < sb_act; i++) {
1129 unsigned sign = (i - 1) & 2;
1130 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
1131 AV_WN32A(&s->raXin[i], v);
1134 s->synth.synth_filter_float(&s->imdct,
1135 s->subband_fir_hist[chans],
1136 &s->hist_index[chans],
1137 s->subband_fir_noidea[chans], prCoeff,
1138 samples_out, s->raXin, scale);
1143 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
1144 int num_deci_sample, float *samples_in,
1145 float *samples_out, float scale)
1147 /* samples_in: An array holding decimated samples.
1148 * Samples in current subframe starts from samples_in[0],
1149 * while samples_in[-1], samples_in[-2], ..., stores samples
1150 * from last subframe as history.
1152 * samples_out: An array holding interpolated samples
1156 const float *prCoeff;
1159 /* Select decimation filter */
1160 if (decimation_select == 1) {
1162 prCoeff = lfe_fir_128;
1165 prCoeff = lfe_fir_64;
1168 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1169 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1171 samples_out += 2 * decifactor;
1175 /* downmixing routines */
1176 #define MIX_REAR1(samples, s1, rs, coef) \
1177 samples[0][i] += samples[s1][i] * coef[rs][0]; \
1178 samples[1][i] += samples[s1][i] * coef[rs][1];
1180 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1181 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1182 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1184 #define MIX_FRONT3(samples, coef) \
1185 t = samples[c][i]; \
1186 u = samples[l][i]; \
1187 v = samples[r][i]; \
1188 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1189 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1191 #define DOWNMIX_TO_STEREO(op1, op2) \
1192 for (i = 0; i < 256; i++) { \
1197 static void dca_downmix(float **samples, int srcfmt,
1198 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1199 const int8_t *channel_mapping)
1201 int c, l, r, sl, sr, s;
1204 float coef[DCA_PRIM_CHANNELS_MAX][2];
1206 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1207 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1208 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1214 case DCA_STEREO_TOTAL:
1215 case DCA_STEREO_SUMDIFF:
1217 av_log(NULL, AV_LOG_ERROR, "Not implemented!\n");
1222 c = channel_mapping[0];
1223 l = channel_mapping[1];
1224 r = channel_mapping[2];
1225 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1228 s = channel_mapping[2];
1229 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1232 c = channel_mapping[0];
1233 l = channel_mapping[1];
1234 r = channel_mapping[2];
1235 s = channel_mapping[3];
1236 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1237 MIX_REAR1(samples, s, 3, coef));
1240 sl = channel_mapping[2];
1241 sr = channel_mapping[3];
1242 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1245 c = channel_mapping[0];
1246 l = channel_mapping[1];
1247 r = channel_mapping[2];
1248 sl = channel_mapping[3];
1249 sr = channel_mapping[4];
1250 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1251 MIX_REAR2(samples, sl, sr, 3, coef));
1257 #ifndef decode_blockcodes
1258 /* Very compact version of the block code decoder that does not use table
1259 * look-up but is slightly slower */
1260 static int decode_blockcode(int code, int levels, int *values)
1263 int offset = (levels - 1) >> 1;
1265 for (i = 0; i < 4; i++) {
1266 int div = FASTDIV(code, levels);
1267 values[i] = code - offset - div * levels;
1274 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1276 return decode_blockcode(code1, levels, values) |
1277 decode_blockcode(code2, levels, values + 4);
1281 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1282 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1284 #ifndef int8x8_fmul_int32
1285 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1287 float fscale = scale / 16.0;
1289 for (i = 0; i < 8; i++)
1290 dst[i] = src[i] * fscale;
1294 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1297 int subsubframe = s->current_subsubframe;
1299 const float *quant_step_table;
1302 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1303 LOCAL_ALIGNED_16(int, block, [8]);
1309 /* Select quantization step size table */
1310 if (s->bit_rate_index == 0x1f)
1311 quant_step_table = lossless_quant_d;
1313 quant_step_table = lossy_quant_d;
1315 for (k = base_channel; k < s->prim_channels; k++) {
1316 if (get_bits_left(&s->gb) < 0)
1317 return AVERROR_INVALIDDATA;
1319 for (l = 0; l < s->vq_start_subband[k]; l++) {
1322 /* Select the mid-tread linear quantizer */
1323 int abits = s->bitalloc[k][l];
1325 float quant_step_size = quant_step_table[abits];
1328 * Determine quantization index code book and its type
1331 /* Select quantization index code book */
1332 int sel = s->quant_index_huffman[k][abits];
1335 * Extract bits from the bit stream
1338 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1340 /* Deal with transients */
1341 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1342 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1343 s->scalefactor_adj[k][sel];
1345 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1348 int block_code1, block_code2, size, levels, err;
1350 size = abits_sizes[abits - 1];
1351 levels = abits_levels[abits - 1];
1353 block_code1 = get_bits(&s->gb, size);
1354 block_code2 = get_bits(&s->gb, size);
1355 err = decode_blockcodes(block_code1, block_code2,
1358 av_log(s->avctx, AV_LOG_ERROR,
1359 "ERROR: block code look-up failed\n");
1360 return AVERROR_INVALIDDATA;
1364 for (m = 0; m < 8; m++)
1365 block[m] = get_sbits(&s->gb, abits - 3);
1369 for (m = 0; m < 8; m++)
1370 block[m] = get_bitalloc(&s->gb,
1371 &dca_smpl_bitalloc[abits], sel);
1374 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1379 * Inverse ADPCM if in prediction mode
1381 if (s->prediction_mode[k][l]) {
1383 for (m = 0; m < 8; m++) {
1384 for (n = 1; n <= 4; n++)
1386 subband_samples[k][l][m] +=
1387 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1388 subband_samples[k][l][m - n] / 8192);
1389 else if (s->predictor_history)
1390 subband_samples[k][l][m] +=
1391 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1392 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1398 * Decode VQ encoded high frequencies
1400 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1401 /* 1 vector -> 32 samples but we only need the 8 samples
1402 * for this subsubframe. */
1403 int hfvq = s->high_freq_vq[k][l];
1405 if (!s->debug_flag & 0x01) {
1406 av_log(s->avctx, AV_LOG_DEBUG,
1407 "Stream with high frequencies VQ coding\n");
1408 s->debug_flag |= 0x01;
1411 int8x8_fmul_int32(subband_samples[k][l],
1412 &high_freq_vq[hfvq][subsubframe * 8],
1413 s->scale_factor[k][l][0]);
1417 /* Check for DSYNC after subsubframe */
1418 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1419 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1421 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1424 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1428 /* Backup predictor history for adpcm */
1429 for (k = base_channel; k < s->prim_channels; k++)
1430 for (l = 0; l < s->vq_start_subband[k]; l++)
1431 memcpy(s->subband_samples_hist[k][l],
1432 &subband_samples[k][l][4],
1433 4 * sizeof(subband_samples[0][0][0]));
1438 static int dca_filter_channels(DCAContext *s, int block_index)
1440 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1443 /* 32 subbands QMF */
1444 for (k = 0; k < s->prim_channels; k++) {
1445 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1446 0, 8388608.0, 8388608.0 };*/
1447 if (s->channel_order_tab[k] >= 0)
1448 qmf_32_subbands(s, k, subband_samples[k],
1449 s->samples_chanptr[s->channel_order_tab[k]],
1450 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1454 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1455 dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
1458 /* Generate LFE samples for this subsubframe FIXME!!! */
1459 if (s->output & DCA_LFE) {
1460 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1461 s->lfe_data + 2 * s->lfe * (block_index + 4),
1462 s->samples_chanptr[s->lfe_index],
1463 1.0 / (256.0 * 32768.0));
1464 /* Outputs 20bits pcm samples */
1471 static int dca_subframe_footer(DCAContext *s, int base_channel)
1473 int aux_data_count = 0, i;
1476 * Unpack optional information
1479 /* presumably optional information only appears in the core? */
1480 if (!base_channel) {
1482 skip_bits_long(&s->gb, 32);
1485 aux_data_count = get_bits(&s->gb, 6);
1487 for (i = 0; i < aux_data_count; i++)
1488 get_bits(&s->gb, 8);
1490 if (s->crc_present && (s->downmix || s->dynrange))
1491 get_bits(&s->gb, 16);
1498 * Decode a dca frame block
1500 * @param s pointer to the DCAContext
1503 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1508 if (s->current_subframe >= s->subframes) {
1509 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1510 s->current_subframe, s->subframes);
1511 return AVERROR_INVALIDDATA;
1514 if (!s->current_subsubframe) {
1516 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1518 /* Read subframe header */
1519 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1523 /* Read subsubframe */
1525 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1527 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1531 s->current_subsubframe++;
1532 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1533 s->current_subsubframe = 0;
1534 s->current_subframe++;
1536 if (s->current_subframe >= s->subframes) {
1538 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1540 /* Read subframe footer */
1541 if ((ret = dca_subframe_footer(s, base_channel)))
1549 * Return the number of channels in an ExSS speaker mask (HD)
1551 static int dca_exss_mask2count(int mask)
1553 /* count bits that mean speaker pairs twice */
1554 return av_popcount(mask) +
1555 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1556 DCA_EXSS_FRONT_LEFT_RIGHT |
1557 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1558 DCA_EXSS_WIDE_LEFT_RIGHT |
1559 DCA_EXSS_SIDE_LEFT_RIGHT |
1560 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1561 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1562 DCA_EXSS_REAR_LEFT_RIGHT |
1563 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1567 * Skip mixing coefficients of a single mix out configuration (HD)
1569 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1573 for (i = 0; i < channels; i++) {
1574 int mix_map_mask = get_bits(gb, out_ch);
1575 int num_coeffs = av_popcount(mix_map_mask);
1576 skip_bits_long(gb, num_coeffs * 6);
1581 * Parse extension substream asset header (HD)
1583 static int dca_exss_parse_asset_header(DCAContext *s)
1585 int header_pos = get_bits_count(&s->gb);
1588 int embedded_stereo = 0;
1589 int embedded_6ch = 0;
1590 int drc_code_present;
1591 int av_uninit(extensions_mask);
1594 if (get_bits_left(&s->gb) < 16)
1597 /* We will parse just enough to get to the extensions bitmask with which
1598 * we can set the profile value. */
1600 header_size = get_bits(&s->gb, 9) + 1;
1601 skip_bits(&s->gb, 3); // asset index
1603 if (s->static_fields) {
1604 if (get_bits1(&s->gb))
1605 skip_bits(&s->gb, 4); // asset type descriptor
1606 if (get_bits1(&s->gb))
1607 skip_bits_long(&s->gb, 24); // language descriptor
1609 if (get_bits1(&s->gb)) {
1610 /* How can one fit 1024 bytes of text here if the maximum value
1611 * for the asset header size field above was 512 bytes? */
1612 int text_length = get_bits(&s->gb, 10) + 1;
1613 if (get_bits_left(&s->gb) < text_length * 8)
1615 skip_bits_long(&s->gb, text_length * 8); // info text
1618 skip_bits(&s->gb, 5); // bit resolution - 1
1619 skip_bits(&s->gb, 4); // max sample rate code
1620 channels = get_bits(&s->gb, 8) + 1;
1622 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1623 int spkr_remap_sets;
1624 int spkr_mask_size = 16;
1628 embedded_stereo = get_bits1(&s->gb);
1630 embedded_6ch = get_bits1(&s->gb);
1632 if (get_bits1(&s->gb)) {
1633 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1634 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1637 spkr_remap_sets = get_bits(&s->gb, 3);
1639 for (i = 0; i < spkr_remap_sets; i++) {
1640 /* std layout mask for each remap set */
1641 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1644 for (i = 0; i < spkr_remap_sets; i++) {
1645 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1646 if (get_bits_left(&s->gb) < 0)
1649 for (j = 0; j < num_spkrs[i]; j++) {
1650 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1651 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1652 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1657 skip_bits(&s->gb, 3); // representation type
1661 drc_code_present = get_bits1(&s->gb);
1662 if (drc_code_present)
1663 get_bits(&s->gb, 8); // drc code
1665 if (get_bits1(&s->gb))
1666 skip_bits(&s->gb, 5); // dialog normalization code
1668 if (drc_code_present && embedded_stereo)
1669 get_bits(&s->gb, 8); // drc stereo code
1671 if (s->mix_metadata && get_bits1(&s->gb)) {
1672 skip_bits(&s->gb, 1); // external mix
1673 skip_bits(&s->gb, 6); // post mix gain code
1675 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1676 skip_bits(&s->gb, 3); // drc limit
1678 skip_bits(&s->gb, 8); // custom drc code
1680 if (get_bits1(&s->gb)) // channel specific scaling
1681 for (i = 0; i < s->num_mix_configs; i++)
1682 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1684 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1686 for (i = 0; i < s->num_mix_configs; i++) {
1687 if (get_bits_left(&s->gb) < 0)
1689 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1691 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1692 if (embedded_stereo)
1693 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1697 switch (get_bits(&s->gb, 2)) {
1698 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1699 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1700 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1701 case 3: extensions_mask = 0; /* aux coding */ break;
1704 /* not parsed further, we were only interested in the extensions mask */
1706 if (get_bits_left(&s->gb) < 0)
1709 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1710 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1713 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1715 if (extensions_mask & DCA_EXT_EXSS_XLL)
1716 s->profile = FF_PROFILE_DTS_HD_MA;
1717 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1719 s->profile = FF_PROFILE_DTS_HD_HRA;
1721 if (!(extensions_mask & DCA_EXT_CORE))
1722 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1723 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1724 av_log(s->avctx, AV_LOG_WARNING,
1725 "DTS extensions detection mismatch (%d, %d)\n",
1726 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1731 static int dca_xbr_parse_frame(DCAContext *s)
1733 int scale_table_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS][2];
1734 int active_bands[DCA_CHSETS_MAX][DCA_CHSET_CHANS_MAX];
1735 int abits_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS];
1736 int anctemp[DCA_CHSET_CHANS_MAX];
1737 int chset_fsize[DCA_CHSETS_MAX];
1738 int n_xbr_ch[DCA_CHSETS_MAX];
1739 int hdr_size, num_chsets, xbr_tmode, hdr_pos;
1740 int i, j, k, l, chset, chan_base;
1742 av_log(s->avctx, AV_LOG_DEBUG, "DTS-XBR: decoding XBR extension\n");
1744 /* get bit position of sync header */
1745 hdr_pos = get_bits_count(&s->gb) - 32;
1747 hdr_size = get_bits(&s->gb, 6) + 1;
1748 num_chsets = get_bits(&s->gb, 2) + 1;
1750 for(i = 0; i < num_chsets; i++)
1751 chset_fsize[i] = get_bits(&s->gb, 14) + 1;
1753 xbr_tmode = get_bits1(&s->gb);
1755 for(i = 0; i < num_chsets; i++) {
1756 n_xbr_ch[i] = get_bits(&s->gb, 3) + 1;
1757 k = get_bits(&s->gb, 2) + 5;
1758 for(j = 0; j < n_xbr_ch[i]; j++)
1759 active_bands[i][j] = get_bits(&s->gb, k) + 1;
1762 /* skip to the end of the header */
1763 i = get_bits_count(&s->gb);
1764 if(hdr_pos + hdr_size * 8 > i)
1765 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1767 /* loop over the channel data sets */
1768 /* only decode as many channels as we've decoded base data for */
1769 for(chset = 0, chan_base = 0;
1770 chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->prim_channels;
1771 chan_base += n_xbr_ch[chset++]) {
1772 int start_posn = get_bits_count(&s->gb);
1773 int subsubframe = 0;
1776 /* loop over subframes */
1777 for (k = 0; k < (s->sample_blocks / 8); k++) {
1778 /* parse header if we're on first subsubframe of a block */
1779 if(subsubframe == 0) {
1780 /* Parse subframe header */
1781 for(i = 0; i < n_xbr_ch[chset]; i++) {
1782 anctemp[i] = get_bits(&s->gb, 2) + 2;
1785 for(i = 0; i < n_xbr_ch[chset]; i++) {
1786 get_array(&s->gb, abits_high[i], active_bands[chset][i], anctemp[i]);
1789 for(i = 0; i < n_xbr_ch[chset]; i++) {
1790 anctemp[i] = get_bits(&s->gb, 3);
1791 if(anctemp[i] < 1) {
1792 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: SYNC ERROR\n");
1793 return AVERROR_INVALIDDATA;
1797 /* generate scale factors */
1798 for(i = 0; i < n_xbr_ch[chset]; i++) {
1799 const uint32_t *scale_table;
1802 if (s->scalefactor_huffman[chan_base+i] == 6) {
1803 scale_table = scale_factor_quant7;
1805 scale_table = scale_factor_quant6;
1810 for(j = 0; j < active_bands[chset][i]; j++) {
1811 if(abits_high[i][j] > 0) {
1812 scale_table_high[i][j][0] =
1813 scale_table[get_bits(&s->gb, nbits)];
1815 if(xbr_tmode && s->transition_mode[i][j]) {
1816 scale_table_high[i][j][1] =
1817 scale_table[get_bits(&s->gb, nbits)];
1824 /* decode audio array for this block */
1825 for(i = 0; i < n_xbr_ch[chset]; i++) {
1826 for(j = 0; j < active_bands[chset][i]; j++) {
1827 const int xbr_abits = abits_high[i][j];
1828 const float quant_step_size = lossless_quant_d[xbr_abits];
1829 const int sfi = xbr_tmode && s->transition_mode[i][j] && subsubframe >= s->transition_mode[i][j];
1830 const float rscale = quant_step_size * scale_table_high[i][j][sfi];
1831 float *subband_samples = s->subband_samples[k][chan_base+i][j];
1838 get_array(&s->gb, block, 8, xbr_abits - 3);
1840 int block_code1, block_code2, size, levels, err;
1842 size = abits_sizes[xbr_abits - 1];
1843 levels = abits_levels[xbr_abits - 1];
1845 block_code1 = get_bits(&s->gb, size);
1846 block_code2 = get_bits(&s->gb, size);
1847 err = decode_blockcodes(block_code1, block_code2,
1850 av_log(s->avctx, AV_LOG_ERROR,
1851 "ERROR: DTS-XBR: block code look-up failed\n");
1852 return AVERROR_INVALIDDATA;
1856 /* scale & sum into subband */
1857 for(l = 0; l < 8; l++)
1858 subband_samples[l] += (float)block[l] * rscale;
1862 /* check DSYNC marker */
1863 if(s->aspf || subsubframe == s->subsubframes[subframe] - 1) {
1864 if(get_bits(&s->gb, 16) != 0xffff) {
1865 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: Didn't get subframe DSYNC\n");
1866 return AVERROR_INVALIDDATA;
1870 /* advance sub-sub-frame index */
1871 if(++subsubframe >= s->subsubframes[subframe]) {
1877 /* skip to next channel set */
1878 i = get_bits_count(&s->gb);
1879 if(start_posn + chset_fsize[chset] * 8 != i) {
1880 j = start_posn + chset_fsize[chset] * 8 - i;
1882 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: end of channel set,"
1883 " skipping further than expected (%d bits)\n", j);
1884 skip_bits_long(&s->gb, j);
1891 /* parse initial header for XXCH and dump details */
1892 static int dca_xxch_decode_frame(DCAContext *s)
1894 int hdr_size, spkmsk_bits, num_chsets, core_spk, hdr_pos;
1895 int i, chset, base_channel, chstart, fsize[8];
1897 /* assume header word has already been parsed */
1898 hdr_pos = get_bits_count(&s->gb) - 32;
1899 hdr_size = get_bits(&s->gb, 6) + 1;
1900 /*chhdr_crc =*/ skip_bits1(&s->gb);
1901 spkmsk_bits = get_bits(&s->gb, 5) + 1;
1902 num_chsets = get_bits(&s->gb, 2) + 1;
1904 for (i = 0; i < num_chsets; i++)
1905 fsize[i] = get_bits(&s->gb, 14) + 1;
1907 core_spk = get_bits(&s->gb, spkmsk_bits);
1908 s->xxch_core_spkmask = core_spk;
1909 s->xxch_nbits_spk_mask = spkmsk_bits;
1910 s->xxch_dmix_embedded = 0;
1912 /* skip to the end of the header */
1913 i = get_bits_count(&s->gb);
1914 if (hdr_pos + hdr_size * 8 > i)
1915 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1917 for (chset = 0; chset < num_chsets; chset++) {
1918 chstart = get_bits_count(&s->gb);
1919 base_channel = s->prim_channels;
1920 s->xxch_chset = chset;
1922 /* XXCH and Core headers differ, see 6.4.2 "XXCH Channel Set Header" vs.
1923 5.3.2 "Primary Audio Coding Header", DTS Spec 1.3.1 */
1924 dca_parse_audio_coding_header(s, base_channel, 1);
1926 /* decode channel data */
1927 for (i = 0; i < (s->sample_blocks / 8); i++) {
1928 if (dca_decode_block(s, base_channel, i)) {
1929 av_log(s->avctx, AV_LOG_ERROR,
1930 "Error decoding DTS-XXCH extension\n");
1935 /* skip to end of this section */
1936 i = get_bits_count(&s->gb);
1937 if (chstart + fsize[chset] * 8 > i)
1938 skip_bits_long(&s->gb, chstart + fsize[chset] * 8 - i);
1940 s->xxch_chset = num_chsets;
1946 * Parse extension substream header (HD)
1948 static void dca_exss_parse_header(DCAContext *s)
1955 int active_ss_mask[8];
1961 if (get_bits_left(&s->gb) < 52)
1964 start_posn = get_bits_count(&s->gb) - 32;
1966 skip_bits(&s->gb, 8); // user data
1967 ss_index = get_bits(&s->gb, 2);
1969 blownup = get_bits1(&s->gb);
1970 hdrsize = get_bits(&s->gb, 8 + 4 * blownup) + 1; // header_size
1971 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1973 s->static_fields = get_bits1(&s->gb);
1974 if (s->static_fields) {
1975 skip_bits(&s->gb, 2); // reference clock code
1976 skip_bits(&s->gb, 3); // frame duration code
1978 if (get_bits1(&s->gb))
1979 skip_bits_long(&s->gb, 36); // timestamp
1981 /* a single stream can contain multiple audio assets that can be
1982 * combined to form multiple audio presentations */
1984 num_audiop = get_bits(&s->gb, 3) + 1;
1985 if (num_audiop > 1) {
1986 avpriv_request_sample(s->avctx,
1987 "Multiple DTS-HD audio presentations");
1988 /* ignore such streams for now */
1992 num_assets = get_bits(&s->gb, 3) + 1;
1993 if (num_assets > 1) {
1994 avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
1995 /* ignore such streams for now */
1999 for (i = 0; i < num_audiop; i++)
2000 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
2002 for (i = 0; i < num_audiop; i++)
2003 for (j = 0; j <= ss_index; j++)
2004 if (active_ss_mask[i] & (1 << j))
2005 skip_bits(&s->gb, 8); // active asset mask
2007 s->mix_metadata = get_bits1(&s->gb);
2008 if (s->mix_metadata) {
2009 int mix_out_mask_size;
2011 skip_bits(&s->gb, 2); // adjustment level
2012 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
2013 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
2015 for (i = 0; i < s->num_mix_configs; i++) {
2016 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
2017 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
2022 for (i = 0; i < num_assets; i++)
2023 asset_size[i] = get_bits_long(&s->gb, 16 + 4 * blownup);
2025 for (i = 0; i < num_assets; i++) {
2026 if (dca_exss_parse_asset_header(s))
2030 /* not parsed further, we were only interested in the extensions mask
2031 * from the asset header */
2033 if (num_assets > 0) {
2034 j = get_bits_count(&s->gb);
2035 if (start_posn + hdrsize * 8 > j)
2036 skip_bits_long(&s->gb, start_posn + hdrsize * 8 - j);
2038 for (i = 0; i < num_assets; i++) {
2039 start_posn = get_bits_count(&s->gb);
2040 mkr = get_bits_long(&s->gb, 32);
2042 /* parse extensions that we know about */
2043 if (mkr == 0x655e315e) {
2044 dca_xbr_parse_frame(s);
2045 } else if (mkr == 0x47004a03) {
2046 dca_xxch_decode_frame(s);
2047 s->core_ext_mask |= DCA_EXT_XXCH; /* xxx use for chan reordering */
2049 av_log(s->avctx, AV_LOG_DEBUG,
2050 "DTS-ExSS: unknown marker = 0x%08x\n", mkr);
2053 /* skip to end of block */
2054 j = get_bits_count(&s->gb);
2055 if (start_posn + asset_size[i] * 8 > j)
2056 skip_bits_long(&s->gb, start_posn + asset_size[i] * 8 - j);
2062 * Main frame decoding function
2063 * FIXME add arguments
2065 static int dca_decode_frame(AVCodecContext *avctx, void *data,
2066 int *got_frame_ptr, AVPacket *avpkt)
2068 AVFrame *frame = data;
2069 const uint8_t *buf = avpkt->data;
2070 int buf_size = avpkt->size;
2074 int num_core_channels = 0;
2076 float **samples_flt;
2079 DCAContext *s = avctx->priv_data;
2081 int channels, full_channels;
2093 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
2094 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
2095 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
2096 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
2097 return AVERROR_INVALIDDATA;
2100 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
2101 if ((ret = dca_parse_frame_header(s)) < 0) {
2102 //seems like the frame is corrupt, try with the next one
2105 //set AVCodec values with parsed data
2106 avctx->sample_rate = s->sample_rate;
2107 avctx->bit_rate = s->bit_rate;
2109 s->profile = FF_PROFILE_DTS;
2111 for (i = 0; i < (s->sample_blocks / 8); i++) {
2112 if ((ret = dca_decode_block(s, 0, i))) {
2113 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
2118 /* record number of core channels incase less than max channels are requested */
2119 num_core_channels = s->prim_channels;
2122 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
2124 s->core_ext_mask = 0;
2126 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
2128 /* only scan for extensions if ext_descr was unknown or indicated a
2129 * supported XCh extension */
2130 if (s->core_ext_mask < 0 || s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH)) {
2132 /* if ext_descr was unknown, clear s->core_ext_mask so that the
2133 * extensions scan can fill it up */
2134 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
2136 /* extensions start at 32-bit boundaries into bitstream */
2137 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2139 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
2140 uint32_t bits = get_bits_long(&s->gb, 32);
2144 int ext_amode, xch_fsize;
2146 s->xch_base_channel = s->prim_channels;
2148 /* validate sync word using XCHFSIZE field */
2149 xch_fsize = show_bits(&s->gb, 10);
2150 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
2151 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
2154 /* skip length-to-end-of-frame field for the moment */
2155 skip_bits(&s->gb, 10);
2157 s->core_ext_mask |= DCA_EXT_XCH;
2159 /* extension amode(number of channels in extension) should be 1 */
2160 /* AFAIK XCh is not used for more channels */
2161 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
2162 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
2163 " supported!\n", ext_amode);
2167 if (s->xch_base_channel < 2) {
2168 avpriv_request_sample(avctx, "XCh with fewer than 2 base channels");
2172 /* much like core primary audio coding header */
2173 dca_parse_audio_coding_header(s, s->xch_base_channel, 0);
2175 for (i = 0; i < (s->sample_blocks / 8); i++)
2176 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
2177 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
2185 /* XXCh: extended channels */
2186 /* usually found either in core or HD part in DTS-HD HRA streams,
2187 * but not in DTS-ES which contains XCh extensions instead */
2188 s->core_ext_mask |= DCA_EXT_XXCH;
2189 dca_xxch_decode_frame(s);
2193 int fsize96 = show_bits(&s->gb, 12) + 1;
2194 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
2197 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
2198 get_bits_count(&s->gb));
2199 skip_bits(&s->gb, 12);
2200 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
2201 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
2203 s->core_ext_mask |= DCA_EXT_X96;
2208 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2211 /* no supported extensions, skip the rest of the core substream */
2212 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
2215 if (s->core_ext_mask & DCA_EXT_X96)
2216 s->profile = FF_PROFILE_DTS_96_24;
2217 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
2218 s->profile = FF_PROFILE_DTS_ES;
2220 /* check for ExSS (HD part) */
2221 if (s->dca_buffer_size - s->frame_size > 32 &&
2222 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
2223 dca_exss_parse_header(s);
2225 avctx->profile = s->profile;
2227 full_channels = channels = s->prim_channels + !!s->lfe;
2229 /* If we have XXCH then the channel layout is managed differently */
2230 /* note that XLL will also have another way to do things */
2231 if (!(s->core_ext_mask & DCA_EXT_XXCH)
2232 || (s->core_ext_mask & DCA_EXT_XXCH && avctx->request_channels > 0
2233 && avctx->request_channels
2234 < num_core_channels + !!s->lfe + s->xxch_chset_nch[0]))
2235 { /* xxx should also do MA extensions */
2236 if (s->amode < 16) {
2237 avctx->channel_layout = dca_core_channel_layout[s->amode];
2239 if (s->xch_present && (!avctx->request_channels ||
2240 avctx->request_channels
2241 > num_core_channels + !!s->lfe)) {
2242 avctx->channel_layout |= AV_CH_BACK_CENTER;
2244 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
2245 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
2247 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
2249 if (s->channel_order_tab[s->xch_base_channel] < 0)
2250 return AVERROR_INVALIDDATA;
2252 channels = num_core_channels + !!s->lfe;
2253 s->xch_present = 0; /* disable further xch processing */
2255 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
2256 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
2258 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
2261 if (channels > !!s->lfe &&
2262 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
2263 return AVERROR_INVALIDDATA;
2265 if (av_get_channel_layout_nb_channels(avctx->channel_layout) != channels) {
2266 av_log(avctx, AV_LOG_ERROR, "Number of channels %d mismatches layout %d\n", channels, av_get_channel_layout_nb_channels(avctx->channel_layout));
2267 return AVERROR_INVALIDDATA;
2270 if (avctx->request_channels == 2 && s->prim_channels > 2) {
2272 s->output = DCA_STEREO;
2273 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
2275 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
2276 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
2277 s->channel_order_tab = dca_channel_order_native;
2279 s->lfe_index = dca_lfe_index[s->amode];
2281 av_log(avctx, AV_LOG_ERROR,
2282 "Non standard configuration %d !\n", s->amode);
2283 return AVERROR_INVALIDDATA;
2286 s->xxch_dmix_embedded = 0;
2288 /* we only get here if an XXCH channel set can be added to the mix */
2289 channel_mask = s->xxch_core_spkmask;
2291 if (avctx->request_channels > 0
2292 && avctx->request_channels < s->prim_channels) {
2293 channels = num_core_channels + !!s->lfe;
2294 for (i = 0; i < s->xxch_chset && channels + s->xxch_chset_nch[i]
2295 <= avctx->request_channels; i++) {
2296 channels += s->xxch_chset_nch[i];
2297 channel_mask |= s->xxch_spk_masks[i];
2300 channels = s->prim_channels + !!s->lfe;
2301 for (i = 0; i < s->xxch_chset; i++) {
2302 channel_mask |= s->xxch_spk_masks[i];
2306 /* Given the DTS spec'ed channel mask, generate an avcodec version */
2308 for (i = 0; i < s->xxch_nbits_spk_mask; ++i) {
2309 if (channel_mask & (1 << i)) {
2310 channel_layout |= map_xxch_to_native[i];
2314 /* make sure that we have managed to get equivelant dts/avcodec channel
2315 * masks in some sense -- unfortunately some channels could overlap */
2316 if (av_popcount(channel_mask) != av_popcount(channel_layout)) {
2317 av_log(avctx, AV_LOG_DEBUG,
2318 "DTS-XXCH: Inconsistant avcodec/dts channel layouts\n");
2319 return AVERROR_INVALIDDATA;
2322 avctx->channel_layout = channel_layout;
2324 if (!(avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE)) {
2325 /* Estimate DTS --> avcodec ordering table */
2326 for (chset = -1, j = 0; chset < s->xxch_chset; ++chset) {
2327 mask = chset >= 0 ? s->xxch_spk_masks[chset]
2328 : s->xxch_core_spkmask;
2329 for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
2330 if (mask & ~(DCA_XXCH_LFE1 | DCA_XXCH_LFE2) & (1 << i)) {
2331 lavc = map_xxch_to_native[i];
2332 posn = av_popcount(channel_layout & (lavc - 1));
2333 s->xxch_order_tab[j++] = posn;
2338 s->lfe_index = av_popcount(channel_layout & (AV_CH_LOW_FREQUENCY-1));
2339 } else { /* native ordering */
2340 for (i = 0; i < channels; i++)
2341 s->xxch_order_tab[i] = i;
2343 s->lfe_index = channels - 1;
2346 s->channel_order_tab = s->xxch_order_tab;
2349 if (avctx->channels != channels) {
2350 if (avctx->channels)
2351 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
2352 avctx->channels = channels;
2355 /* get output buffer */
2356 frame->nb_samples = 256 * (s->sample_blocks / 8);
2357 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
2359 samples_flt = (float **)frame->extended_data;
2361 /* allocate buffer for extra channels if downmixing */
2362 if (avctx->channels < full_channels) {
2363 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
2365 avctx->sample_fmt, 0);
2369 av_fast_malloc(&s->extra_channels_buffer,
2370 &s->extra_channels_buffer_size, ret);
2371 if (!s->extra_channels_buffer)
2372 return AVERROR(ENOMEM);
2374 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
2375 s->extra_channels_buffer,
2376 full_channels - channels,
2377 frame->nb_samples, avctx->sample_fmt, 0);
2382 /* filter to get final output */
2383 for (i = 0; i < (s->sample_blocks / 8); i++) {
2386 for (ch = 0; ch < channels; ch++)
2387 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
2388 for (; ch < full_channels; ch++)
2389 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
2391 dca_filter_channels(s, i);
2393 /* If this was marked as a DTS-ES stream we need to subtract back- */
2394 /* channel from SL & SR to remove matrixed back-channel signal */
2395 if ((s->source_pcm_res & 1) && s->xch_present) {
2396 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
2397 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
2398 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
2399 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
2400 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
2403 /* If stream contains XXCH, we might need to undo an embedded downmix */
2404 if (s->xxch_dmix_embedded) {
2405 /* Loop over channel sets in turn */
2406 ch = num_core_channels;
2407 for (chset = 0; chset < s->xxch_chset; chset++) {
2408 endch = ch + s->xxch_chset_nch[chset];
2409 mask = s->xxch_dmix_embedded;
2412 for (j = ch; j < endch; j++) {
2413 if (mask & (1 << j)) { /* this channel has been mixed-out */
2414 src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2415 for (k = 0; k < endch; k++) {
2416 achan = s->channel_order_tab[k];
2417 scale = s->xxch_dmix_coeff[j][k];
2419 dst_chan = s->samples_chanptr[achan];
2420 s->fdsp.vector_fmac_scalar(dst_chan, src_chan,
2427 /* if a downmix has been embedded then undo the pre-scaling */
2428 if ((mask & (1 << ch)) && s->xxch_dmix_sf[chset] != 1.0f) {
2429 scale = s->xxch_dmix_sf[chset];
2431 for (j = 0; j < ch; j++) {
2432 src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2433 for (k = 0; k < 256; k++)
2434 src_chan[k] *= scale;
2437 /* LFE channel is always part of core, scale if it exists */
2439 src_chan = s->samples_chanptr[s->lfe_index];
2440 for (k = 0; k < 256; k++)
2441 src_chan[k] *= scale;
2451 /* update lfe history */
2452 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
2453 for (i = 0; i < 2 * s->lfe * 4; i++)
2454 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
2464 * DCA initialization
2466 * @param avctx pointer to the AVCodecContext
2469 static av_cold int dca_decode_init(AVCodecContext *avctx)
2471 DCAContext *s = avctx->priv_data;
2476 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
2477 ff_mdct_init(&s->imdct, 6, 1, 1.0);
2478 ff_synth_filter_init(&s->synth);
2479 ff_dcadsp_init(&s->dcadsp);
2480 ff_fmt_convert_init(&s->fmt_conv, avctx);
2482 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
2484 /* allow downmixing to stereo */
2485 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
2486 avctx->request_channels == 2) {
2487 avctx->channels = avctx->request_channels;
2493 static av_cold int dca_decode_end(AVCodecContext *avctx)
2495 DCAContext *s = avctx->priv_data;
2496 ff_mdct_end(&s->imdct);
2497 av_freep(&s->extra_channels_buffer);
2501 static const AVProfile profiles[] = {
2502 { FF_PROFILE_DTS, "DTS" },
2503 { FF_PROFILE_DTS_ES, "DTS-ES" },
2504 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2505 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2506 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2507 { FF_PROFILE_UNKNOWN },
2510 AVCodec ff_dca_decoder = {
2512 .type = AVMEDIA_TYPE_AUDIO,
2513 .id = AV_CODEC_ID_DTS,
2514 .priv_data_size = sizeof(DCAContext),
2515 .init = dca_decode_init,
2516 .decode = dca_decode_frame,
2517 .close = dca_decode_end,
2518 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2519 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2520 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
2521 AV_SAMPLE_FMT_NONE },
2522 .profiles = NULL_IF_CONFIG_SMALL(profiles),