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"
44 #include "synth_filter.h"
46 #include "fmtconvert.h"
55 #define DCA_PRIM_CHANNELS_MAX (7)
56 #define DCA_SUBBANDS (64)
57 #define DCA_ABITS_MAX (32) /* Should be 28 */
58 #define DCA_SUBSUBFRAMES_MAX (4)
59 #define DCA_SUBFRAMES_MAX (16)
60 #define DCA_BLOCKS_MAX (16)
61 #define DCA_LFE_MAX (3)
62 #define DCA_CHSETS_MAX (4)
63 #define DCA_CHSET_CHANS_MAX (8)
79 /* these are unconfirmed but should be mostly correct */
80 enum DCAExSSSpeakerMask {
81 DCA_EXSS_FRONT_CENTER = 0x0001,
82 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
83 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
84 DCA_EXSS_LFE = 0x0008,
85 DCA_EXSS_REAR_CENTER = 0x0010,
86 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
87 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
88 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
89 DCA_EXSS_OVERHEAD = 0x0100,
90 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
91 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
92 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
93 DCA_EXSS_LFE2 = 0x1000,
94 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
95 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
96 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
99 enum DCAXxchSpeakerMask {
100 DCA_XXCH_FRONT_CENTER = 0x0000001,
101 DCA_XXCH_FRONT_LEFT = 0x0000002,
102 DCA_XXCH_FRONT_RIGHT = 0x0000004,
103 DCA_XXCH_SIDE_REAR_LEFT = 0x0000008,
104 DCA_XXCH_SIDE_REAR_RIGHT = 0x0000010,
105 DCA_XXCH_LFE1 = 0x0000020,
106 DCA_XXCH_REAR_CENTER = 0x0000040,
107 DCA_XXCH_SURROUND_REAR_LEFT = 0x0000080,
108 DCA_XXCH_SURROUND_REAR_RIGHT = 0x0000100,
109 DCA_XXCH_SIDE_SURROUND_LEFT = 0x0000200,
110 DCA_XXCH_SIDE_SURROUND_RIGHT = 0x0000400,
111 DCA_XXCH_FRONT_CENTER_LEFT = 0x0000800,
112 DCA_XXCH_FRONT_CENTER_RIGHT = 0x0001000,
113 DCA_XXCH_FRONT_HIGH_LEFT = 0x0002000,
114 DCA_XXCH_FRONT_HIGH_CENTER = 0x0004000,
115 DCA_XXCH_FRONT_HIGH_RIGHT = 0x0008000,
116 DCA_XXCH_LFE2 = 0x0010000,
117 DCA_XXCH_SIDE_FRONT_LEFT = 0x0020000,
118 DCA_XXCH_SIDE_FRONT_RIGHT = 0x0040000,
119 DCA_XXCH_OVERHEAD = 0x0080000,
120 DCA_XXCH_SIDE_HIGH_LEFT = 0x0100000,
121 DCA_XXCH_SIDE_HIGH_RIGHT = 0x0200000,
122 DCA_XXCH_REAR_HIGH_CENTER = 0x0400000,
123 DCA_XXCH_REAR_HIGH_LEFT = 0x0800000,
124 DCA_XXCH_REAR_HIGH_RIGHT = 0x1000000,
125 DCA_XXCH_REAR_LOW_CENTER = 0x2000000,
126 DCA_XXCH_REAR_LOW_LEFT = 0x4000000,
127 DCA_XXCH_REAR_LOW_RIGHT = 0x8000000,
130 static const uint32_t map_xxch_to_native[28] = {
140 AV_CH_SIDE_LEFT, /* side surround left -- dup sur side L */
141 AV_CH_SIDE_RIGHT, /* side surround right -- dup sur side R */
142 AV_CH_FRONT_LEFT_OF_CENTER,
143 AV_CH_FRONT_RIGHT_OF_CENTER,
144 AV_CH_TOP_FRONT_LEFT,
145 AV_CH_TOP_FRONT_CENTER,
146 AV_CH_TOP_FRONT_RIGHT,
147 AV_CH_LOW_FREQUENCY, /* lfe2 -- duplicate lfe1 position */
148 AV_CH_FRONT_LEFT_OF_CENTER, /* side front left -- dup front cntr L */
149 AV_CH_FRONT_RIGHT_OF_CENTER,/* side front right -- dup front cntr R */
150 AV_CH_TOP_CENTER, /* overhead */
151 AV_CH_TOP_FRONT_LEFT, /* side high left -- dup */
152 AV_CH_TOP_FRONT_RIGHT, /* side high right -- dup */
153 AV_CH_TOP_BACK_CENTER,
155 AV_CH_TOP_BACK_RIGHT,
156 AV_CH_BACK_CENTER, /* rear low center -- dup */
157 AV_CH_BACK_LEFT, /* rear low left -- dup */
158 AV_CH_BACK_RIGHT /* read low right -- dup */
161 enum DCAExtensionMask {
162 DCA_EXT_CORE = 0x001, ///< core in core substream
163 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
164 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
165 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
166 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
167 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
168 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
169 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
170 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
171 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
174 /* -1 are reserved or unknown */
175 static const int dca_ext_audio_descr_mask[] = {
179 DCA_EXT_XCH | DCA_EXT_X96,
186 /* extensions that reside in core substream */
187 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
189 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
190 * Some compromises have been made for special configurations. Most configurations
191 * are never used so complete accuracy is not needed.
193 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
194 * S -> side, when both rear and back are configured move one of them to the side channel
196 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
198 static const uint64_t dca_core_channel_layout[] = {
199 AV_CH_FRONT_CENTER, ///< 1, A
200 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
201 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
202 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
203 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
204 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
205 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
206 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
207 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
209 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
210 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
212 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
213 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
215 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
216 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
218 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
219 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
220 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
222 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
223 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
224 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
226 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
227 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
228 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
230 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
231 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
232 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
235 static const int8_t dca_lfe_index[] = {
236 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
239 static const int8_t dca_channel_reorder_lfe[][9] = {
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 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
245 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
246 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
247 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
248 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
249 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
250 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
251 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
252 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
253 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
254 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
255 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
258 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
259 { 0, 2, -1, -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 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
264 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
265 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
266 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
267 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
268 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
269 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
270 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
271 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
272 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
273 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
274 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
277 static const int8_t dca_channel_reorder_nolfe[][9] = {
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 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
283 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
284 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
285 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
286 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
287 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
288 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
289 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
290 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
291 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
292 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
293 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
296 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
297 { 0, 1, -1, -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 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
302 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
303 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
304 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
305 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
306 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
307 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
308 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
309 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
310 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
311 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
312 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
315 #define DCA_DOLBY 101 /* FIXME */
317 #define DCA_CHANNEL_BITS 6
318 #define DCA_CHANNEL_MASK 0x3F
322 #define HEADER_SIZE 14
324 #define DCA_MAX_FRAME_SIZE 16384
325 #define DCA_MAX_EXSS_HEADER_SIZE 4096
327 #define DCA_BUFFER_PADDING_SIZE 1024
329 /** Bit allocation */
331 int offset; ///< code values offset
332 int maxbits[8]; ///< max bits in VLC
333 int wrap; ///< wrap for get_vlc2()
334 VLC vlc[8]; ///< actual codes
337 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
338 static BitAlloc dca_tmode; ///< transition mode VLCs
339 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
340 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
342 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
345 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
350 AVCodecContext *avctx;
352 int frame_type; ///< type of the current frame
353 int samples_deficit; ///< deficit sample count
354 int crc_present; ///< crc is present in the bitstream
355 int sample_blocks; ///< number of PCM sample blocks
356 int frame_size; ///< primary frame byte size
357 int amode; ///< audio channels arrangement
358 int sample_rate; ///< audio sampling rate
359 int bit_rate; ///< transmission bit rate
360 int bit_rate_index; ///< transmission bit rate index
362 int downmix; ///< embedded downmix enabled
363 int dynrange; ///< embedded dynamic range flag
364 int timestamp; ///< embedded time stamp flag
365 int aux_data; ///< auxiliary data flag
366 int hdcd; ///< source material is mastered in HDCD
367 int ext_descr; ///< extension audio descriptor flag
368 int ext_coding; ///< extended coding flag
369 int aspf; ///< audio sync word insertion flag
370 int lfe; ///< low frequency effects flag
371 int predictor_history; ///< predictor history flag
372 int header_crc; ///< header crc check bytes
373 int multirate_inter; ///< multirate interpolator switch
374 int version; ///< encoder software revision
375 int copy_history; ///< copy history
376 int source_pcm_res; ///< source pcm resolution
377 int front_sum; ///< front sum/difference flag
378 int surround_sum; ///< surround sum/difference flag
379 int dialog_norm; ///< dialog normalisation parameter
381 /* Primary audio coding header */
382 int subframes; ///< number of subframes
383 int total_channels; ///< number of channels including extensions
384 int prim_channels; ///< number of primary audio channels
385 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
386 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
387 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
388 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
389 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
390 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
391 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
392 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
394 /* Primary audio coding side information */
395 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
396 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
397 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
398 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
399 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
400 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
401 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
402 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
403 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
404 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
405 int dynrange_coef; ///< dynamic range coefficient
407 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
409 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
410 int lfe_scale_factor;
412 /* Subband samples history (for ADPCM) */
413 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
414 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
415 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
416 int hist_index[DCA_PRIM_CHANNELS_MAX];
417 DECLARE_ALIGNED(32, float, raXin)[32];
419 int output; ///< type of output
421 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
422 float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
423 float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
424 uint8_t *extra_channels_buffer;
425 unsigned int extra_channels_buffer_size;
427 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
428 int dca_buffer_size; ///< how much data is in the dca_buffer
430 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
432 /* Current position in DCA frame */
433 int current_subframe;
434 int current_subsubframe;
436 int core_ext_mask; ///< present extensions in the core substream
438 /* XCh extension information */
439 int xch_present; ///< XCh extension present and valid
440 int xch_base_channel; ///< index of first (only) channel containing XCH data
442 /* XXCH extension information */
444 int xxch_nbits_spk_mask;
445 uint32_t xxch_core_spkmask;
446 uint32_t xxch_spk_masks[4]; /* speaker masks, last element is core mask */
447 int xxch_chset_nch[4];
448 float xxch_dmix_sf[DCA_CHSETS_MAX];
450 uint32_t xxch_dmix_embedded; /* lower layer has mix pre-embedded, per chset */
451 float xxch_dmix_coeff[DCA_PRIM_CHANNELS_MAX][32]; /* worst case sizing */
453 int8_t xxch_order_tab[32];
456 /* ExSS header parser */
457 int static_fields; ///< static fields present
458 int mix_metadata; ///< mixing metadata present
459 int num_mix_configs; ///< number of mix out configurations
460 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
464 int debug_flag; ///< used for suppressing repeated error messages output
465 AVFloatDSPContext fdsp;
467 SynthFilterContext synth;
468 DCADSPContext dcadsp;
469 FmtConvertContext fmt_conv;
472 static const uint16_t dca_vlc_offs[] = {
473 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
474 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
475 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
476 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
477 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
478 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
481 static av_cold void dca_init_vlcs(void)
483 static int vlcs_initialized = 0;
485 static VLC_TYPE dca_table[23622][2];
487 if (vlcs_initialized)
490 dca_bitalloc_index.offset = 1;
491 dca_bitalloc_index.wrap = 2;
492 for (i = 0; i < 5; i++) {
493 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
494 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
495 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
496 bitalloc_12_bits[i], 1, 1,
497 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
499 dca_scalefactor.offset = -64;
500 dca_scalefactor.wrap = 2;
501 for (i = 0; i < 5; i++) {
502 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
503 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
504 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
505 scales_bits[i], 1, 1,
506 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
508 dca_tmode.offset = 0;
510 for (i = 0; i < 4; i++) {
511 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
512 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
513 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
515 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
518 for (i = 0; i < 10; i++)
519 for (j = 0; j < 7; j++) {
520 if (!bitalloc_codes[i][j])
522 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
523 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
524 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
525 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
527 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
529 bitalloc_bits[i][j], 1, 1,
530 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
533 vlcs_initialized = 1;
536 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
539 *dst++ = get_bits(gb, bits);
542 static inline int dca_xxch2index(DCAContext *s, int xxch_ch)
546 /* locate channel set containing the channel */
547 for (i = -1, base = 0, mask = (s->xxch_core_spkmask & ~DCA_XXCH_LFE1);
548 i <= s->xxch_chset && !(mask & xxch_ch); mask = s->xxch_spk_masks[++i])
549 base += av_popcount(mask);
551 return base + av_popcount(mask & (xxch_ch - 1));
554 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel,
558 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
559 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
560 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
561 int hdr_pos = 0, hdr_size = 0;
562 float sign, mag, scale_factor;
563 int this_chans, acc_mask;
564 int embedded_downmix;
568 /* xxch has arbitrary sized audio coding headers */
570 hdr_pos = get_bits_count(&s->gb);
571 hdr_size = get_bits(&s->gb, 7) + 1;
574 nchans = get_bits(&s->gb, 3) + 1;
575 s->total_channels = nchans + base_channel;
576 s->prim_channels = s->total_channels;
578 /* obtain speaker layout mask & downmix coefficients for XXCH */
580 acc_mask = s->xxch_core_spkmask;
582 this_chans = get_bits(&s->gb, s->xxch_nbits_spk_mask - 6) << 6;
583 s->xxch_spk_masks[s->xxch_chset] = this_chans;
584 s->xxch_chset_nch[s->xxch_chset] = nchans;
586 for (i = 0; i <= s->xxch_chset; i++)
587 acc_mask |= s->xxch_spk_masks[i];
589 /* check for downmixing information */
590 if (get_bits1(&s->gb)) {
591 embedded_downmix = get_bits1(&s->gb);
593 1.0f / dca_downmix_scale_factors[(get_bits(&s->gb, 6) - 1) << 2];
595 s->xxch_dmix_sf[s->xxch_chset] = scale_factor;
597 for (i = base_channel; i < s->prim_channels; i++) {
598 mask[i] = get_bits(&s->gb, s->xxch_nbits_spk_mask);
601 for (j = base_channel; j < s->prim_channels; j++) {
602 memset(s->xxch_dmix_coeff[j], 0, sizeof(s->xxch_dmix_coeff[0]));
603 s->xxch_dmix_embedded |= (embedded_downmix << j);
604 for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
605 if (mask[j] & (1 << i)) {
606 if ((1 << i) == DCA_XXCH_LFE1) {
607 av_log(s->avctx, AV_LOG_WARNING,
608 "DCA-XXCH: dmix to LFE1 not supported.\n");
612 coeff = get_bits(&s->gb, 7);
613 sign = (coeff & 64) ? 1.0 : -1.0;
614 mag = dca_downmix_scale_factors[((coeff & 63) - 1) << 2];
615 ichan = dca_xxch2index(s, 1 << i);
616 s->xxch_dmix_coeff[j][ichan] = sign * mag;
623 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
624 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
627 for (i = base_channel; i < s->prim_channels; i++) {
628 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
629 if (s->subband_activity[i] > DCA_SUBBANDS)
630 s->subband_activity[i] = DCA_SUBBANDS;
632 for (i = base_channel; i < s->prim_channels; i++) {
633 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
634 if (s->vq_start_subband[i] > DCA_SUBBANDS)
635 s->vq_start_subband[i] = DCA_SUBBANDS;
637 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
638 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
639 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
640 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
642 /* Get codebooks quantization indexes */
644 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
645 for (j = 1; j < 11; j++)
646 for (i = base_channel; i < s->prim_channels; i++)
647 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
649 /* Get scale factor adjustment */
650 for (j = 0; j < 11; j++)
651 for (i = base_channel; i < s->prim_channels; i++)
652 s->scalefactor_adj[i][j] = 1;
654 for (j = 1; j < 11; j++)
655 for (i = base_channel; i < s->prim_channels; i++)
656 if (s->quant_index_huffman[i][j] < thr[j])
657 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
660 if (s->crc_present) {
661 /* Audio header CRC check */
662 get_bits(&s->gb, 16);
665 /* Skip to the end of the header, also ignore CRC if present */
666 i = get_bits_count(&s->gb);
667 if (hdr_pos + 8 * hdr_size > i)
668 skip_bits_long(&s->gb, hdr_pos + 8 * hdr_size - i);
671 s->current_subframe = 0;
672 s->current_subsubframe = 0;
675 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
676 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
677 for (i = base_channel; i < s->prim_channels; i++) {
678 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
679 s->subband_activity[i]);
680 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
681 s->vq_start_subband[i]);
682 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
683 s->joint_intensity[i]);
684 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
685 s->transient_huffman[i]);
686 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
687 s->scalefactor_huffman[i]);
688 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
689 s->bitalloc_huffman[i]);
690 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
691 for (j = 0; j < 11; j++)
692 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
693 av_log(s->avctx, AV_LOG_DEBUG, "\n");
694 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
695 for (j = 0; j < 11; j++)
696 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
697 av_log(s->avctx, AV_LOG_DEBUG, "\n");
704 static int dca_parse_frame_header(DCAContext *s)
706 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
709 skip_bits_long(&s->gb, 32);
712 s->frame_type = get_bits(&s->gb, 1);
713 s->samples_deficit = get_bits(&s->gb, 5) + 1;
714 s->crc_present = get_bits(&s->gb, 1);
715 s->sample_blocks = get_bits(&s->gb, 7) + 1;
716 s->frame_size = get_bits(&s->gb, 14) + 1;
717 if (s->frame_size < 95)
718 return AVERROR_INVALIDDATA;
719 s->amode = get_bits(&s->gb, 6);
720 s->sample_rate = avpriv_dca_sample_rates[get_bits(&s->gb, 4)];
722 return AVERROR_INVALIDDATA;
723 s->bit_rate_index = get_bits(&s->gb, 5);
724 s->bit_rate = dca_bit_rates[s->bit_rate_index];
726 return AVERROR_INVALIDDATA;
728 s->downmix = get_bits(&s->gb, 1); /* note: this is FixedBit == 0 */
729 s->dynrange = get_bits(&s->gb, 1);
730 s->timestamp = get_bits(&s->gb, 1);
731 s->aux_data = get_bits(&s->gb, 1);
732 s->hdcd = get_bits(&s->gb, 1);
733 s->ext_descr = get_bits(&s->gb, 3);
734 s->ext_coding = get_bits(&s->gb, 1);
735 s->aspf = get_bits(&s->gb, 1);
736 s->lfe = get_bits(&s->gb, 2);
737 s->predictor_history = get_bits(&s->gb, 1);
741 avpriv_request_sample(s->avctx, "LFE = 3");
742 return AVERROR_PATCHWELCOME;
745 /* TODO: check CRC */
747 s->header_crc = get_bits(&s->gb, 16);
749 s->multirate_inter = get_bits(&s->gb, 1);
750 s->version = get_bits(&s->gb, 4);
751 s->copy_history = get_bits(&s->gb, 2);
752 s->source_pcm_res = get_bits(&s->gb, 3);
753 s->front_sum = get_bits(&s->gb, 1);
754 s->surround_sum = get_bits(&s->gb, 1);
755 s->dialog_norm = get_bits(&s->gb, 4);
757 /* FIXME: channels mixing levels */
758 s->output = s->amode;
760 s->output |= DCA_LFE;
763 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
764 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
765 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
766 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
767 s->sample_blocks, s->sample_blocks * 32);
768 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
769 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
770 s->amode, dca_channels[s->amode]);
771 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
773 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
775 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
776 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
777 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
778 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
779 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
780 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
781 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
782 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
783 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
784 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
785 s->predictor_history);
786 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
787 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
789 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
790 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
791 av_log(s->avctx, AV_LOG_DEBUG,
792 "source pcm resolution: %i (%i bits/sample)\n",
793 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
794 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
795 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
796 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
797 av_log(s->avctx, AV_LOG_DEBUG, "\n");
800 /* Primary audio coding header */
801 s->subframes = get_bits(&s->gb, 4) + 1;
803 return dca_parse_audio_coding_header(s, 0, 0);
807 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
810 /* huffman encoded */
811 value += get_bitalloc(gb, &dca_scalefactor, level);
812 value = av_clip(value, 0, (1 << log2range) - 1);
813 } else if (level < 8) {
814 if (level + 1 > log2range) {
815 skip_bits(gb, level + 1 - log2range);
816 value = get_bits(gb, log2range);
818 value = get_bits(gb, level + 1);
824 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
826 /* Primary audio coding side information */
829 if (get_bits_left(&s->gb) < 0)
830 return AVERROR_INVALIDDATA;
833 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
834 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
837 for (j = base_channel; j < s->prim_channels; j++) {
838 for (k = 0; k < s->subband_activity[j]; k++)
839 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
842 /* Get prediction codebook */
843 for (j = base_channel; j < s->prim_channels; j++) {
844 for (k = 0; k < s->subband_activity[j]; k++) {
845 if (s->prediction_mode[j][k] > 0) {
846 /* (Prediction coefficient VQ address) */
847 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
852 /* Bit allocation index */
853 for (j = base_channel; j < s->prim_channels; j++) {
854 for (k = 0; k < s->vq_start_subband[j]; k++) {
855 if (s->bitalloc_huffman[j] == 6)
856 s->bitalloc[j][k] = get_bits(&s->gb, 5);
857 else if (s->bitalloc_huffman[j] == 5)
858 s->bitalloc[j][k] = get_bits(&s->gb, 4);
859 else if (s->bitalloc_huffman[j] == 7) {
860 av_log(s->avctx, AV_LOG_ERROR,
861 "Invalid bit allocation index\n");
862 return AVERROR_INVALIDDATA;
865 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
868 if (s->bitalloc[j][k] > 26) {
869 av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
870 j, k, s->bitalloc[j][k]);
871 return AVERROR_INVALIDDATA;
876 /* Transition mode */
877 for (j = base_channel; j < s->prim_channels; j++) {
878 for (k = 0; k < s->subband_activity[j]; k++) {
879 s->transition_mode[j][k] = 0;
880 if (s->subsubframes[s->current_subframe] > 1 &&
881 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
882 s->transition_mode[j][k] =
883 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
888 if (get_bits_left(&s->gb) < 0)
889 return AVERROR_INVALIDDATA;
891 for (j = base_channel; j < s->prim_channels; j++) {
892 const uint32_t *scale_table;
893 int scale_sum, log_size;
895 memset(s->scale_factor[j], 0,
896 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
898 if (s->scalefactor_huffman[j] == 6) {
899 scale_table = scale_factor_quant7;
902 scale_table = scale_factor_quant6;
906 /* When huffman coded, only the difference is encoded */
909 for (k = 0; k < s->subband_activity[j]; k++) {
910 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
911 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
912 s->scale_factor[j][k][0] = scale_table[scale_sum];
915 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
916 /* Get second scale factor */
917 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
918 s->scale_factor[j][k][1] = scale_table[scale_sum];
923 /* Joint subband scale factor codebook select */
924 for (j = base_channel; j < s->prim_channels; j++) {
925 /* Transmitted only if joint subband coding enabled */
926 if (s->joint_intensity[j] > 0)
927 s->joint_huff[j] = get_bits(&s->gb, 3);
930 if (get_bits_left(&s->gb) < 0)
931 return AVERROR_INVALIDDATA;
933 /* Scale factors for joint subband coding */
934 for (j = base_channel; j < s->prim_channels; j++) {
937 /* Transmitted only if joint subband coding enabled */
938 if (s->joint_intensity[j] > 0) {
940 source_channel = s->joint_intensity[j] - 1;
942 /* When huffman coded, only the difference is encoded
943 * (is this valid as well for joint scales ???) */
945 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
946 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
947 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
950 if (!(s->debug_flag & 0x02)) {
951 av_log(s->avctx, AV_LOG_DEBUG,
952 "Joint stereo coding not supported\n");
953 s->debug_flag |= 0x02;
958 /* Stereo downmix coefficients */
959 if (!base_channel && s->prim_channels > 2) {
961 for (j = base_channel; j < s->prim_channels; j++) {
962 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
963 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
966 int am = s->amode & DCA_CHANNEL_MASK;
967 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
968 av_log(s->avctx, AV_LOG_ERROR,
969 "Invalid channel mode %d\n", am);
970 return AVERROR_INVALIDDATA;
972 for (j = base_channel; j < FFMIN(s->prim_channels, FF_ARRAY_ELEMS(dca_default_coeffs[am])); j++) {
973 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
974 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
979 /* Dynamic range coefficient */
980 if (!base_channel && s->dynrange)
981 s->dynrange_coef = get_bits(&s->gb, 8);
983 /* Side information CRC check word */
984 if (s->crc_present) {
985 get_bits(&s->gb, 16);
989 * Primary audio data arrays
992 /* VQ encoded high frequency subbands */
993 for (j = base_channel; j < s->prim_channels; j++)
994 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
995 /* 1 vector -> 32 samples */
996 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
998 /* Low frequency effect data */
999 if (!base_channel && s->lfe) {
1002 int lfe_samples = 2 * s->lfe * (4 + block_index);
1003 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
1006 for (j = lfe_samples; j < lfe_end_sample; j++) {
1007 /* Signed 8 bits int */
1008 s->lfe_data[j] = get_sbits(&s->gb, 8);
1011 /* Scale factor index */
1012 quant7 = get_bits(&s->gb, 8);
1014 avpriv_request_sample(s->avctx, "LFEScaleIndex larger than 127");
1015 return AVERROR_INVALIDDATA;
1017 s->lfe_scale_factor = scale_factor_quant7[quant7];
1019 /* Quantization step size * scale factor */
1020 lfe_scale = 0.035 * s->lfe_scale_factor;
1022 for (j = lfe_samples; j < lfe_end_sample; j++)
1023 s->lfe_data[j] *= lfe_scale;
1027 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
1028 s->subsubframes[s->current_subframe]);
1029 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
1030 s->partial_samples[s->current_subframe]);
1032 for (j = base_channel; j < s->prim_channels; j++) {
1033 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
1034 for (k = 0; k < s->subband_activity[j]; k++)
1035 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
1036 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1038 for (j = base_channel; j < s->prim_channels; j++) {
1039 for (k = 0; k < s->subband_activity[j]; k++)
1040 av_log(s->avctx, AV_LOG_DEBUG,
1041 "prediction coefs: %f, %f, %f, %f\n",
1042 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
1043 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
1044 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
1045 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
1047 for (j = base_channel; j < s->prim_channels; j++) {
1048 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
1049 for (k = 0; k < s->vq_start_subband[j]; k++)
1050 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
1051 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1053 for (j = base_channel; j < s->prim_channels; j++) {
1054 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
1055 for (k = 0; k < s->subband_activity[j]; k++)
1056 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
1057 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1059 for (j = base_channel; j < s->prim_channels; j++) {
1060 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
1061 for (k = 0; k < s->subband_activity[j]; k++) {
1062 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
1063 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
1064 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
1065 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
1067 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1069 for (j = base_channel; j < s->prim_channels; j++) {
1070 if (s->joint_intensity[j] > 0) {
1071 int source_channel = s->joint_intensity[j] - 1;
1072 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
1073 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
1074 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
1075 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1078 if (!base_channel && s->prim_channels > 2 && s->downmix) {
1079 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
1080 for (j = 0; j < s->prim_channels; j++) {
1081 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
1082 dca_downmix_coeffs[s->downmix_coef[j][0]]);
1083 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
1084 dca_downmix_coeffs[s->downmix_coef[j][1]]);
1086 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1088 for (j = base_channel; j < s->prim_channels; j++)
1089 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
1090 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
1091 if (!base_channel && s->lfe) {
1092 int lfe_samples = 2 * s->lfe * (4 + block_index);
1093 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
1095 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
1096 for (j = lfe_samples; j < lfe_end_sample; j++)
1097 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
1098 av_log(s->avctx, AV_LOG_DEBUG, "\n");
1105 static void qmf_32_subbands(DCAContext *s, int chans,
1106 float samples_in[32][8], float *samples_out,
1109 const float *prCoeff;
1112 int sb_act = s->subband_activity[chans];
1115 scale *= sqrt(1 / 8.0);
1118 if (!s->multirate_inter) /* Non-perfect reconstruction */
1119 prCoeff = fir_32bands_nonperfect;
1120 else /* Perfect reconstruction */
1121 prCoeff = fir_32bands_perfect;
1123 for (i = sb_act; i < 32; i++)
1126 /* Reconstructed channel sample index */
1127 for (subindex = 0; subindex < 8; subindex++) {
1128 /* Load in one sample from each subband and clear inactive subbands */
1129 for (i = 0; i < sb_act; i++) {
1130 unsigned sign = (i - 1) & 2;
1131 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
1132 AV_WN32A(&s->raXin[i], v);
1135 s->synth.synth_filter_float(&s->imdct,
1136 s->subband_fir_hist[chans],
1137 &s->hist_index[chans],
1138 s->subband_fir_noidea[chans], prCoeff,
1139 samples_out, s->raXin, scale);
1144 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
1145 int num_deci_sample, float *samples_in,
1146 float *samples_out, float scale)
1148 /* samples_in: An array holding decimated samples.
1149 * Samples in current subframe starts from samples_in[0],
1150 * while samples_in[-1], samples_in[-2], ..., stores samples
1151 * from last subframe as history.
1153 * samples_out: An array holding interpolated samples
1157 const float *prCoeff;
1160 /* Select decimation filter */
1161 if (decimation_select == 1) {
1163 prCoeff = lfe_fir_128;
1166 prCoeff = lfe_fir_64;
1169 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1170 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1172 samples_out += 2 * decifactor;
1176 /* downmixing routines */
1177 #define MIX_REAR1(samples, s1, rs, coef) \
1178 samples[0][i] += samples[s1][i] * coef[rs][0]; \
1179 samples[1][i] += samples[s1][i] * coef[rs][1];
1181 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1182 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1183 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1185 #define MIX_FRONT3(samples, coef) \
1186 t = samples[c][i]; \
1187 u = samples[l][i]; \
1188 v = samples[r][i]; \
1189 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1190 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1192 #define DOWNMIX_TO_STEREO(op1, op2) \
1193 for (i = 0; i < 256; i++) { \
1198 static void dca_downmix(float **samples, int srcfmt,
1199 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1200 const int8_t *channel_mapping)
1202 int c, l, r, sl, sr, s;
1205 float coef[DCA_PRIM_CHANNELS_MAX][2];
1207 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1208 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1209 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1215 case DCA_STEREO_TOTAL:
1216 case DCA_STEREO_SUMDIFF:
1218 av_log(NULL, AV_LOG_ERROR, "Not implemented!\n");
1223 c = channel_mapping[0];
1224 l = channel_mapping[1];
1225 r = channel_mapping[2];
1226 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1229 s = channel_mapping[2];
1230 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1233 c = channel_mapping[0];
1234 l = channel_mapping[1];
1235 r = channel_mapping[2];
1236 s = channel_mapping[3];
1237 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1238 MIX_REAR1(samples, s, 3, coef));
1241 sl = channel_mapping[2];
1242 sr = channel_mapping[3];
1243 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1246 c = channel_mapping[0];
1247 l = channel_mapping[1];
1248 r = channel_mapping[2];
1249 sl = channel_mapping[3];
1250 sr = channel_mapping[4];
1251 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1252 MIX_REAR2(samples, sl, sr, 3, coef));
1258 #ifndef decode_blockcodes
1259 /* Very compact version of the block code decoder that does not use table
1260 * look-up but is slightly slower */
1261 static int decode_blockcode(int code, int levels, int *values)
1264 int offset = (levels - 1) >> 1;
1266 for (i = 0; i < 4; i++) {
1267 int div = FASTDIV(code, levels);
1268 values[i] = code - offset - div * levels;
1275 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1277 return decode_blockcode(code1, levels, values) |
1278 decode_blockcode(code2, levels, values + 4);
1282 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1283 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1285 #ifndef int8x8_fmul_int32
1286 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1288 float fscale = scale / 16.0;
1290 for (i = 0; i < 8; i++)
1291 dst[i] = src[i] * fscale;
1295 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1298 int subsubframe = s->current_subsubframe;
1300 const float *quant_step_table;
1303 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1304 LOCAL_ALIGNED_16(int, block, [8]);
1310 /* Select quantization step size table */
1311 if (s->bit_rate_index == 0x1f)
1312 quant_step_table = lossless_quant_d;
1314 quant_step_table = lossy_quant_d;
1316 for (k = base_channel; k < s->prim_channels; k++) {
1317 if (get_bits_left(&s->gb) < 0)
1318 return AVERROR_INVALIDDATA;
1320 for (l = 0; l < s->vq_start_subband[k]; l++) {
1323 /* Select the mid-tread linear quantizer */
1324 int abits = s->bitalloc[k][l];
1326 float quant_step_size = quant_step_table[abits];
1329 * Determine quantization index code book and its type
1332 /* Select quantization index code book */
1333 int sel = s->quant_index_huffman[k][abits];
1336 * Extract bits from the bit stream
1339 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1341 /* Deal with transients */
1342 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1343 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1344 s->scalefactor_adj[k][sel];
1346 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1349 int block_code1, block_code2, size, levels, err;
1351 size = abits_sizes[abits - 1];
1352 levels = abits_levels[abits - 1];
1354 block_code1 = get_bits(&s->gb, size);
1355 block_code2 = get_bits(&s->gb, size);
1356 err = decode_blockcodes(block_code1, block_code2,
1359 av_log(s->avctx, AV_LOG_ERROR,
1360 "ERROR: block code look-up failed\n");
1361 return AVERROR_INVALIDDATA;
1365 for (m = 0; m < 8; m++)
1366 block[m] = get_sbits(&s->gb, abits - 3);
1370 for (m = 0; m < 8; m++)
1371 block[m] = get_bitalloc(&s->gb,
1372 &dca_smpl_bitalloc[abits], sel);
1375 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1380 * Inverse ADPCM if in prediction mode
1382 if (s->prediction_mode[k][l]) {
1384 for (m = 0; m < 8; m++) {
1385 for (n = 1; n <= 4; n++)
1387 subband_samples[k][l][m] +=
1388 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1389 subband_samples[k][l][m - n] / 8192);
1390 else if (s->predictor_history)
1391 subband_samples[k][l][m] +=
1392 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1393 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1399 * Decode VQ encoded high frequencies
1401 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1402 /* 1 vector -> 32 samples but we only need the 8 samples
1403 * for this subsubframe. */
1404 int hfvq = s->high_freq_vq[k][l];
1406 if (!s->debug_flag & 0x01) {
1407 av_log(s->avctx, AV_LOG_DEBUG,
1408 "Stream with high frequencies VQ coding\n");
1409 s->debug_flag |= 0x01;
1412 int8x8_fmul_int32(subband_samples[k][l],
1413 &high_freq_vq[hfvq][subsubframe * 8],
1414 s->scale_factor[k][l][0]);
1418 /* Check for DSYNC after subsubframe */
1419 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1420 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1422 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1425 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1429 /* Backup predictor history for adpcm */
1430 for (k = base_channel; k < s->prim_channels; k++)
1431 for (l = 0; l < s->vq_start_subband[k]; l++)
1432 memcpy(s->subband_samples_hist[k][l],
1433 &subband_samples[k][l][4],
1434 4 * sizeof(subband_samples[0][0][0]));
1439 static int dca_filter_channels(DCAContext *s, int block_index)
1441 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1444 /* 32 subbands QMF */
1445 for (k = 0; k < s->prim_channels; k++) {
1446 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1447 0, 8388608.0, 8388608.0 };*/
1448 if (s->channel_order_tab[k] >= 0)
1449 qmf_32_subbands(s, k, subband_samples[k],
1450 s->samples_chanptr[s->channel_order_tab[k]],
1451 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1455 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1456 dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
1459 /* Generate LFE samples for this subsubframe FIXME!!! */
1460 if (s->output & DCA_LFE) {
1461 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1462 s->lfe_data + 2 * s->lfe * (block_index + 4),
1463 s->samples_chanptr[s->lfe_index],
1464 1.0 / (256.0 * 32768.0));
1465 /* Outputs 20bits pcm samples */
1472 static int dca_subframe_footer(DCAContext *s, int base_channel)
1474 int aux_data_count = 0, i;
1477 * Unpack optional information
1480 /* presumably optional information only appears in the core? */
1481 if (!base_channel) {
1483 skip_bits_long(&s->gb, 32);
1486 aux_data_count = get_bits(&s->gb, 6);
1488 for (i = 0; i < aux_data_count; i++)
1489 get_bits(&s->gb, 8);
1491 if (s->crc_present && (s->downmix || s->dynrange))
1492 get_bits(&s->gb, 16);
1499 * Decode a dca frame block
1501 * @param s pointer to the DCAContext
1504 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1509 if (s->current_subframe >= s->subframes) {
1510 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1511 s->current_subframe, s->subframes);
1512 return AVERROR_INVALIDDATA;
1515 if (!s->current_subsubframe) {
1517 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1519 /* Read subframe header */
1520 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1524 /* Read subsubframe */
1526 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1528 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1532 s->current_subsubframe++;
1533 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1534 s->current_subsubframe = 0;
1535 s->current_subframe++;
1537 if (s->current_subframe >= s->subframes) {
1539 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1541 /* Read subframe footer */
1542 if ((ret = dca_subframe_footer(s, base_channel)))
1550 * Return the number of channels in an ExSS speaker mask (HD)
1552 static int dca_exss_mask2count(int mask)
1554 /* count bits that mean speaker pairs twice */
1555 return av_popcount(mask) +
1556 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1557 DCA_EXSS_FRONT_LEFT_RIGHT |
1558 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1559 DCA_EXSS_WIDE_LEFT_RIGHT |
1560 DCA_EXSS_SIDE_LEFT_RIGHT |
1561 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1562 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1563 DCA_EXSS_REAR_LEFT_RIGHT |
1564 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1568 * Skip mixing coefficients of a single mix out configuration (HD)
1570 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1574 for (i = 0; i < channels; i++) {
1575 int mix_map_mask = get_bits(gb, out_ch);
1576 int num_coeffs = av_popcount(mix_map_mask);
1577 skip_bits_long(gb, num_coeffs * 6);
1582 * Parse extension substream asset header (HD)
1584 static int dca_exss_parse_asset_header(DCAContext *s)
1586 int header_pos = get_bits_count(&s->gb);
1589 int embedded_stereo = 0;
1590 int embedded_6ch = 0;
1591 int drc_code_present;
1592 int av_uninit(extensions_mask);
1595 if (get_bits_left(&s->gb) < 16)
1598 /* We will parse just enough to get to the extensions bitmask with which
1599 * we can set the profile value. */
1601 header_size = get_bits(&s->gb, 9) + 1;
1602 skip_bits(&s->gb, 3); // asset index
1604 if (s->static_fields) {
1605 if (get_bits1(&s->gb))
1606 skip_bits(&s->gb, 4); // asset type descriptor
1607 if (get_bits1(&s->gb))
1608 skip_bits_long(&s->gb, 24); // language descriptor
1610 if (get_bits1(&s->gb)) {
1611 /* How can one fit 1024 bytes of text here if the maximum value
1612 * for the asset header size field above was 512 bytes? */
1613 int text_length = get_bits(&s->gb, 10) + 1;
1614 if (get_bits_left(&s->gb) < text_length * 8)
1616 skip_bits_long(&s->gb, text_length * 8); // info text
1619 skip_bits(&s->gb, 5); // bit resolution - 1
1620 skip_bits(&s->gb, 4); // max sample rate code
1621 channels = get_bits(&s->gb, 8) + 1;
1623 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1624 int spkr_remap_sets;
1625 int spkr_mask_size = 16;
1629 embedded_stereo = get_bits1(&s->gb);
1631 embedded_6ch = get_bits1(&s->gb);
1633 if (get_bits1(&s->gb)) {
1634 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1635 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1638 spkr_remap_sets = get_bits(&s->gb, 3);
1640 for (i = 0; i < spkr_remap_sets; i++) {
1641 /* std layout mask for each remap set */
1642 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1645 for (i = 0; i < spkr_remap_sets; i++) {
1646 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1647 if (get_bits_left(&s->gb) < 0)
1650 for (j = 0; j < num_spkrs[i]; j++) {
1651 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1652 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1653 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1658 skip_bits(&s->gb, 3); // representation type
1662 drc_code_present = get_bits1(&s->gb);
1663 if (drc_code_present)
1664 get_bits(&s->gb, 8); // drc code
1666 if (get_bits1(&s->gb))
1667 skip_bits(&s->gb, 5); // dialog normalization code
1669 if (drc_code_present && embedded_stereo)
1670 get_bits(&s->gb, 8); // drc stereo code
1672 if (s->mix_metadata && get_bits1(&s->gb)) {
1673 skip_bits(&s->gb, 1); // external mix
1674 skip_bits(&s->gb, 6); // post mix gain code
1676 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1677 skip_bits(&s->gb, 3); // drc limit
1679 skip_bits(&s->gb, 8); // custom drc code
1681 if (get_bits1(&s->gb)) // channel specific scaling
1682 for (i = 0; i < s->num_mix_configs; i++)
1683 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1685 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1687 for (i = 0; i < s->num_mix_configs; i++) {
1688 if (get_bits_left(&s->gb) < 0)
1690 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1692 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1693 if (embedded_stereo)
1694 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1698 switch (get_bits(&s->gb, 2)) {
1699 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1700 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1701 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1702 case 3: extensions_mask = 0; /* aux coding */ break;
1705 /* not parsed further, we were only interested in the extensions mask */
1707 if (get_bits_left(&s->gb) < 0)
1710 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1711 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1714 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1716 if (extensions_mask & DCA_EXT_EXSS_XLL)
1717 s->profile = FF_PROFILE_DTS_HD_MA;
1718 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1720 s->profile = FF_PROFILE_DTS_HD_HRA;
1722 if (!(extensions_mask & DCA_EXT_CORE))
1723 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1724 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1725 av_log(s->avctx, AV_LOG_WARNING,
1726 "DTS extensions detection mismatch (%d, %d)\n",
1727 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1732 static int dca_xbr_parse_frame(DCAContext *s)
1734 int scale_table_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS][2];
1735 int active_bands[DCA_CHSETS_MAX][DCA_CHSET_CHANS_MAX];
1736 int abits_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS];
1737 int anctemp[DCA_CHSET_CHANS_MAX];
1738 int chset_fsize[DCA_CHSETS_MAX];
1739 int n_xbr_ch[DCA_CHSETS_MAX];
1740 int hdr_size, num_chsets, xbr_tmode, hdr_pos;
1741 int i, j, k, l, chset, chan_base;
1743 av_log(s->avctx, AV_LOG_DEBUG, "DTS-XBR: decoding XBR extension\n");
1745 /* get bit position of sync header */
1746 hdr_pos = get_bits_count(&s->gb) - 32;
1748 hdr_size = get_bits(&s->gb, 6) + 1;
1749 num_chsets = get_bits(&s->gb, 2) + 1;
1751 for(i = 0; i < num_chsets; i++)
1752 chset_fsize[i] = get_bits(&s->gb, 14) + 1;
1754 xbr_tmode = get_bits1(&s->gb);
1756 for(i = 0; i < num_chsets; i++) {
1757 n_xbr_ch[i] = get_bits(&s->gb, 3) + 1;
1758 k = get_bits(&s->gb, 2) + 5;
1759 for(j = 0; j < n_xbr_ch[i]; j++)
1760 active_bands[i][j] = get_bits(&s->gb, k) + 1;
1763 /* skip to the end of the header */
1764 i = get_bits_count(&s->gb);
1765 if(hdr_pos + hdr_size * 8 > i)
1766 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1768 /* loop over the channel data sets */
1769 /* only decode as many channels as we've decoded base data for */
1770 for(chset = 0, chan_base = 0;
1771 chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->prim_channels;
1772 chan_base += n_xbr_ch[chset++]) {
1773 int start_posn = get_bits_count(&s->gb);
1774 int subsubframe = 0;
1777 /* loop over subframes */
1778 for (k = 0; k < (s->sample_blocks / 8); k++) {
1779 /* parse header if we're on first subsubframe of a block */
1780 if(subsubframe == 0) {
1781 /* Parse subframe header */
1782 for(i = 0; i < n_xbr_ch[chset]; i++) {
1783 anctemp[i] = get_bits(&s->gb, 2) + 2;
1786 for(i = 0; i < n_xbr_ch[chset]; i++) {
1787 get_array(&s->gb, abits_high[i], active_bands[chset][i], anctemp[i]);
1790 for(i = 0; i < n_xbr_ch[chset]; i++) {
1791 anctemp[i] = get_bits(&s->gb, 3);
1792 if(anctemp[i] < 1) {
1793 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: SYNC ERROR\n");
1794 return AVERROR_INVALIDDATA;
1798 /* generate scale factors */
1799 for(i = 0; i < n_xbr_ch[chset]; i++) {
1800 const uint32_t *scale_table;
1803 if (s->scalefactor_huffman[chan_base+i] == 6) {
1804 scale_table = scale_factor_quant7;
1806 scale_table = scale_factor_quant6;
1811 for(j = 0; j < active_bands[chset][i]; j++) {
1812 if(abits_high[i][j] > 0) {
1813 scale_table_high[i][j][0] =
1814 scale_table[get_bits(&s->gb, nbits)];
1816 if(xbr_tmode && s->transition_mode[i][j]) {
1817 scale_table_high[i][j][1] =
1818 scale_table[get_bits(&s->gb, nbits)];
1825 /* decode audio array for this block */
1826 for(i = 0; i < n_xbr_ch[chset]; i++) {
1827 for(j = 0; j < active_bands[chset][i]; j++) {
1828 const int xbr_abits = abits_high[i][j];
1829 const float quant_step_size = lossless_quant_d[xbr_abits];
1830 const int sfi = xbr_tmode && s->transition_mode[i][j] && subsubframe >= s->transition_mode[i][j];
1831 const float rscale = quant_step_size * scale_table_high[i][j][sfi];
1832 float *subband_samples = s->subband_samples[k][chan_base+i][j];
1839 get_array(&s->gb, block, 8, xbr_abits - 3);
1841 int block_code1, block_code2, size, levels, err;
1843 size = abits_sizes[xbr_abits - 1];
1844 levels = abits_levels[xbr_abits - 1];
1846 block_code1 = get_bits(&s->gb, size);
1847 block_code2 = get_bits(&s->gb, size);
1848 err = decode_blockcodes(block_code1, block_code2,
1851 av_log(s->avctx, AV_LOG_ERROR,
1852 "ERROR: DTS-XBR: block code look-up failed\n");
1853 return AVERROR_INVALIDDATA;
1857 /* scale & sum into subband */
1858 for(l = 0; l < 8; l++)
1859 subband_samples[l] += (float)block[l] * rscale;
1863 /* check DSYNC marker */
1864 if(s->aspf || subsubframe == s->subsubframes[subframe] - 1) {
1865 if(get_bits(&s->gb, 16) != 0xffff) {
1866 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: Didn't get subframe DSYNC\n");
1867 return AVERROR_INVALIDDATA;
1871 /* advance sub-sub-frame index */
1872 if(++subsubframe >= s->subsubframes[subframe]) {
1878 /* skip to next channel set */
1879 i = get_bits_count(&s->gb);
1880 if(start_posn + chset_fsize[chset] * 8 != i) {
1881 j = start_posn + chset_fsize[chset] * 8 - i;
1883 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: end of channel set,"
1884 " skipping further than expected (%d bits)\n", j);
1885 skip_bits_long(&s->gb, j);
1892 /* parse initial header for XXCH and dump details */
1893 static int dca_xxch_decode_frame(DCAContext *s)
1895 int hdr_size, spkmsk_bits, num_chsets, core_spk, hdr_pos;
1896 int i, chset, base_channel, chstart, fsize[8];
1898 /* assume header word has already been parsed */
1899 hdr_pos = get_bits_count(&s->gb) - 32;
1900 hdr_size = get_bits(&s->gb, 6) + 1;
1901 /*chhdr_crc =*/ skip_bits1(&s->gb);
1902 spkmsk_bits = get_bits(&s->gb, 5) + 1;
1903 num_chsets = get_bits(&s->gb, 2) + 1;
1905 for (i = 0; i < num_chsets; i++)
1906 fsize[i] = get_bits(&s->gb, 14) + 1;
1908 core_spk = get_bits(&s->gb, spkmsk_bits);
1909 s->xxch_core_spkmask = core_spk;
1910 s->xxch_nbits_spk_mask = spkmsk_bits;
1911 s->xxch_dmix_embedded = 0;
1913 /* skip to the end of the header */
1914 i = get_bits_count(&s->gb);
1915 if (hdr_pos + hdr_size * 8 > i)
1916 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1918 for (chset = 0; chset < num_chsets; chset++) {
1919 chstart = get_bits_count(&s->gb);
1920 base_channel = s->prim_channels;
1921 s->xxch_chset = chset;
1923 /* XXCH and Core headers differ, see 6.4.2 "XXCH Channel Set Header" vs.
1924 5.3.2 "Primary Audio Coding Header", DTS Spec 1.3.1 */
1925 dca_parse_audio_coding_header(s, base_channel, 1);
1927 /* decode channel data */
1928 for (i = 0; i < (s->sample_blocks / 8); i++) {
1929 if (dca_decode_block(s, base_channel, i)) {
1930 av_log(s->avctx, AV_LOG_ERROR,
1931 "Error decoding DTS-XXCH extension\n");
1936 /* skip to end of this section */
1937 i = get_bits_count(&s->gb);
1938 if (chstart + fsize[chset] * 8 > i)
1939 skip_bits_long(&s->gb, chstart + fsize[chset] * 8 - i);
1941 s->xxch_chset = num_chsets;
1947 * Parse extension substream header (HD)
1949 static void dca_exss_parse_header(DCAContext *s)
1956 int active_ss_mask[8];
1962 if (get_bits_left(&s->gb) < 52)
1965 start_posn = get_bits_count(&s->gb) - 32;
1967 skip_bits(&s->gb, 8); // user data
1968 ss_index = get_bits(&s->gb, 2);
1970 blownup = get_bits1(&s->gb);
1971 hdrsize = get_bits(&s->gb, 8 + 4 * blownup) + 1; // header_size
1972 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1974 s->static_fields = get_bits1(&s->gb);
1975 if (s->static_fields) {
1976 skip_bits(&s->gb, 2); // reference clock code
1977 skip_bits(&s->gb, 3); // frame duration code
1979 if (get_bits1(&s->gb))
1980 skip_bits_long(&s->gb, 36); // timestamp
1982 /* a single stream can contain multiple audio assets that can be
1983 * combined to form multiple audio presentations */
1985 num_audiop = get_bits(&s->gb, 3) + 1;
1986 if (num_audiop > 1) {
1987 avpriv_request_sample(s->avctx,
1988 "Multiple DTS-HD audio presentations");
1989 /* ignore such streams for now */
1993 num_assets = get_bits(&s->gb, 3) + 1;
1994 if (num_assets > 1) {
1995 avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
1996 /* ignore such streams for now */
2000 for (i = 0; i < num_audiop; i++)
2001 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
2003 for (i = 0; i < num_audiop; i++)
2004 for (j = 0; j <= ss_index; j++)
2005 if (active_ss_mask[i] & (1 << j))
2006 skip_bits(&s->gb, 8); // active asset mask
2008 s->mix_metadata = get_bits1(&s->gb);
2009 if (s->mix_metadata) {
2010 int mix_out_mask_size;
2012 skip_bits(&s->gb, 2); // adjustment level
2013 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
2014 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
2016 for (i = 0; i < s->num_mix_configs; i++) {
2017 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
2018 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
2023 for (i = 0; i < num_assets; i++)
2024 asset_size[i] = get_bits_long(&s->gb, 16 + 4 * blownup);
2026 for (i = 0; i < num_assets; i++) {
2027 if (dca_exss_parse_asset_header(s))
2031 /* not parsed further, we were only interested in the extensions mask
2032 * from the asset header */
2034 if (num_assets > 0) {
2035 j = get_bits_count(&s->gb);
2036 if (start_posn + hdrsize * 8 > j)
2037 skip_bits_long(&s->gb, start_posn + hdrsize * 8 - j);
2039 for (i = 0; i < num_assets; i++) {
2040 start_posn = get_bits_count(&s->gb);
2041 mkr = get_bits_long(&s->gb, 32);
2043 /* parse extensions that we know about */
2044 if (mkr == 0x655e315e) {
2045 dca_xbr_parse_frame(s);
2046 } else if (mkr == 0x47004a03) {
2047 dca_xxch_decode_frame(s);
2048 s->core_ext_mask |= DCA_EXT_XXCH; /* xxx use for chan reordering */
2050 av_log(s->avctx, AV_LOG_DEBUG,
2051 "DTS-ExSS: unknown marker = 0x%08x\n", mkr);
2054 /* skip to end of block */
2055 j = get_bits_count(&s->gb);
2056 if (start_posn + asset_size[i] * 8 > j)
2057 skip_bits_long(&s->gb, start_posn + asset_size[i] * 8 - j);
2063 * Main frame decoding function
2064 * FIXME add arguments
2066 static int dca_decode_frame(AVCodecContext *avctx, void *data,
2067 int *got_frame_ptr, AVPacket *avpkt)
2069 AVFrame *frame = data;
2070 const uint8_t *buf = avpkt->data;
2071 int buf_size = avpkt->size;
2075 int num_core_channels = 0;
2077 float **samples_flt;
2080 DCAContext *s = avctx->priv_data;
2082 int channels, full_channels;
2094 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
2095 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
2096 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
2097 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
2098 return AVERROR_INVALIDDATA;
2101 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
2102 if ((ret = dca_parse_frame_header(s)) < 0) {
2103 //seems like the frame is corrupt, try with the next one
2106 //set AVCodec values with parsed data
2107 avctx->sample_rate = s->sample_rate;
2108 avctx->bit_rate = s->bit_rate;
2110 s->profile = FF_PROFILE_DTS;
2112 for (i = 0; i < (s->sample_blocks / 8); i++) {
2113 if ((ret = dca_decode_block(s, 0, i))) {
2114 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
2119 /* record number of core channels incase less than max channels are requested */
2120 num_core_channels = s->prim_channels;
2123 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
2125 s->core_ext_mask = 0;
2127 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
2129 /* only scan for extensions if ext_descr was unknown or indicated a
2130 * supported XCh extension */
2131 if (s->core_ext_mask < 0 || s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH)) {
2133 /* if ext_descr was unknown, clear s->core_ext_mask so that the
2134 * extensions scan can fill it up */
2135 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
2137 /* extensions start at 32-bit boundaries into bitstream */
2138 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2140 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
2141 uint32_t bits = get_bits_long(&s->gb, 32);
2145 int ext_amode, xch_fsize;
2147 s->xch_base_channel = s->prim_channels;
2149 /* validate sync word using XCHFSIZE field */
2150 xch_fsize = show_bits(&s->gb, 10);
2151 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
2152 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
2155 /* skip length-to-end-of-frame field for the moment */
2156 skip_bits(&s->gb, 10);
2158 s->core_ext_mask |= DCA_EXT_XCH;
2160 /* extension amode(number of channels in extension) should be 1 */
2161 /* AFAIK XCh is not used for more channels */
2162 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
2163 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
2164 " supported!\n", ext_amode);
2168 if (s->xch_base_channel < 2) {
2169 avpriv_request_sample(avctx, "XCh with fewer than 2 base channels");
2173 /* much like core primary audio coding header */
2174 dca_parse_audio_coding_header(s, s->xch_base_channel, 0);
2176 for (i = 0; i < (s->sample_blocks / 8); i++)
2177 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
2178 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
2186 /* XXCh: extended channels */
2187 /* usually found either in core or HD part in DTS-HD HRA streams,
2188 * but not in DTS-ES which contains XCh extensions instead */
2189 s->core_ext_mask |= DCA_EXT_XXCH;
2190 dca_xxch_decode_frame(s);
2194 int fsize96 = show_bits(&s->gb, 12) + 1;
2195 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
2198 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
2199 get_bits_count(&s->gb));
2200 skip_bits(&s->gb, 12);
2201 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
2202 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
2204 s->core_ext_mask |= DCA_EXT_X96;
2209 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
2212 /* no supported extensions, skip the rest of the core substream */
2213 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
2216 if (s->core_ext_mask & DCA_EXT_X96)
2217 s->profile = FF_PROFILE_DTS_96_24;
2218 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
2219 s->profile = FF_PROFILE_DTS_ES;
2221 /* check for ExSS (HD part) */
2222 if (s->dca_buffer_size - s->frame_size > 32 &&
2223 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
2224 dca_exss_parse_header(s);
2226 avctx->profile = s->profile;
2228 full_channels = channels = s->prim_channels + !!s->lfe;
2230 /* If we have XXCH then the channel layout is managed differently */
2231 /* note that XLL will also have another way to do things */
2232 if (!(s->core_ext_mask & DCA_EXT_XXCH)
2233 || (s->core_ext_mask & DCA_EXT_XXCH && avctx->request_channels > 0
2234 && avctx->request_channels
2235 < num_core_channels + !!s->lfe + s->xxch_chset_nch[0]))
2236 { /* xxx should also do MA extensions */
2237 if (s->amode < 16) {
2238 avctx->channel_layout = dca_core_channel_layout[s->amode];
2240 if (s->xch_present && (!avctx->request_channels ||
2241 avctx->request_channels
2242 > num_core_channels + !!s->lfe)) {
2243 avctx->channel_layout |= AV_CH_BACK_CENTER;
2245 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
2246 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
2248 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
2250 if (s->channel_order_tab[s->xch_base_channel] < 0)
2251 return AVERROR_INVALIDDATA;
2253 channels = num_core_channels + !!s->lfe;
2254 s->xch_present = 0; /* disable further xch processing */
2256 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
2257 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
2259 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
2262 if (channels > !!s->lfe &&
2263 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
2264 return AVERROR_INVALIDDATA;
2266 if (av_get_channel_layout_nb_channels(avctx->channel_layout) != channels) {
2267 av_log(avctx, AV_LOG_ERROR, "Number of channels %d mismatches layout %d\n", channels, av_get_channel_layout_nb_channels(avctx->channel_layout));
2268 return AVERROR_INVALIDDATA;
2271 if (avctx->request_channels == 2 && s->prim_channels > 2) {
2273 s->output = DCA_STEREO;
2274 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
2276 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
2277 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
2278 s->channel_order_tab = dca_channel_order_native;
2280 s->lfe_index = dca_lfe_index[s->amode];
2282 av_log(avctx, AV_LOG_ERROR,
2283 "Non standard configuration %d !\n", s->amode);
2284 return AVERROR_INVALIDDATA;
2287 s->xxch_dmix_embedded = 0;
2289 /* we only get here if an XXCH channel set can be added to the mix */
2290 channel_mask = s->xxch_core_spkmask;
2292 if (avctx->request_channels > 0
2293 && avctx->request_channels < s->prim_channels) {
2294 channels = num_core_channels + !!s->lfe;
2295 for (i = 0; i < s->xxch_chset && channels + s->xxch_chset_nch[i]
2296 <= avctx->request_channels; i++) {
2297 channels += s->xxch_chset_nch[i];
2298 channel_mask |= s->xxch_spk_masks[i];
2301 channels = s->prim_channels + !!s->lfe;
2302 for (i = 0; i < s->xxch_chset; i++) {
2303 channel_mask |= s->xxch_spk_masks[i];
2307 /* Given the DTS spec'ed channel mask, generate an avcodec version */
2309 for (i = 0; i < s->xxch_nbits_spk_mask; ++i) {
2310 if (channel_mask & (1 << i)) {
2311 channel_layout |= map_xxch_to_native[i];
2315 /* make sure that we have managed to get equivelant dts/avcodec channel
2316 * masks in some sense -- unfortunately some channels could overlap */
2317 if (av_popcount(channel_mask) != av_popcount(channel_layout)) {
2318 av_log(avctx, AV_LOG_DEBUG,
2319 "DTS-XXCH: Inconsistant avcodec/dts channel layouts\n");
2320 return AVERROR_INVALIDDATA;
2323 avctx->channel_layout = channel_layout;
2325 if (!(avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE)) {
2326 /* Estimate DTS --> avcodec ordering table */
2327 for (chset = -1, j = 0; chset < s->xxch_chset; ++chset) {
2328 mask = chset >= 0 ? s->xxch_spk_masks[chset]
2329 : s->xxch_core_spkmask;
2330 for (i = 0; i < s->xxch_nbits_spk_mask; i++) {
2331 if (mask & ~(DCA_XXCH_LFE1 | DCA_XXCH_LFE2) & (1 << i)) {
2332 lavc = map_xxch_to_native[i];
2333 posn = av_popcount(channel_layout & (lavc - 1));
2334 s->xxch_order_tab[j++] = posn;
2339 s->lfe_index = av_popcount(channel_layout & (AV_CH_LOW_FREQUENCY-1));
2340 } else { /* native ordering */
2341 for (i = 0; i < channels; i++)
2342 s->xxch_order_tab[i] = i;
2344 s->lfe_index = channels - 1;
2347 s->channel_order_tab = s->xxch_order_tab;
2350 if (avctx->channels != channels) {
2351 if (avctx->channels)
2352 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
2353 avctx->channels = channels;
2356 /* get output buffer */
2357 frame->nb_samples = 256 * (s->sample_blocks / 8);
2358 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
2360 samples_flt = (float **)frame->extended_data;
2362 /* allocate buffer for extra channels if downmixing */
2363 if (avctx->channels < full_channels) {
2364 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
2366 avctx->sample_fmt, 0);
2370 av_fast_malloc(&s->extra_channels_buffer,
2371 &s->extra_channels_buffer_size, ret);
2372 if (!s->extra_channels_buffer)
2373 return AVERROR(ENOMEM);
2375 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
2376 s->extra_channels_buffer,
2377 full_channels - channels,
2378 frame->nb_samples, avctx->sample_fmt, 0);
2383 /* filter to get final output */
2384 for (i = 0; i < (s->sample_blocks / 8); i++) {
2387 for (ch = 0; ch < channels; ch++)
2388 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
2389 for (; ch < full_channels; ch++)
2390 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
2392 dca_filter_channels(s, i);
2394 /* If this was marked as a DTS-ES stream we need to subtract back- */
2395 /* channel from SL & SR to remove matrixed back-channel signal */
2396 if ((s->source_pcm_res & 1) && s->xch_present) {
2397 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
2398 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
2399 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
2400 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
2401 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
2404 /* If stream contains XXCH, we might need to undo an embedded downmix */
2405 if (s->xxch_dmix_embedded) {
2406 /* Loop over channel sets in turn */
2407 ch = num_core_channels;
2408 for (chset = 0; chset < s->xxch_chset; chset++) {
2409 endch = ch + s->xxch_chset_nch[chset];
2410 mask = s->xxch_dmix_embedded;
2413 for (j = ch; j < endch; j++) {
2414 if (mask & (1 << j)) { /* this channel has been mixed-out */
2415 src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2416 for (k = 0; k < endch; k++) {
2417 achan = s->channel_order_tab[k];
2418 scale = s->xxch_dmix_coeff[j][k];
2420 dst_chan = s->samples_chanptr[achan];
2421 s->fdsp.vector_fmac_scalar(dst_chan, src_chan,
2428 /* if a downmix has been embedded then undo the pre-scaling */
2429 if ((mask & (1 << ch)) && s->xxch_dmix_sf[chset] != 1.0f) {
2430 scale = s->xxch_dmix_sf[chset];
2432 for (j = 0; j < ch; j++) {
2433 src_chan = s->samples_chanptr[s->channel_order_tab[j]];
2434 for (k = 0; k < 256; k++)
2435 src_chan[k] *= scale;
2438 /* LFE channel is always part of core, scale if it exists */
2440 src_chan = s->samples_chanptr[s->lfe_index];
2441 for (k = 0; k < 256; k++)
2442 src_chan[k] *= scale;
2452 /* update lfe history */
2453 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
2454 for (i = 0; i < 2 * s->lfe * 4; i++)
2455 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
2465 * DCA initialization
2467 * @param avctx pointer to the AVCodecContext
2470 static av_cold int dca_decode_init(AVCodecContext *avctx)
2472 DCAContext *s = avctx->priv_data;
2477 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
2478 ff_mdct_init(&s->imdct, 6, 1, 1.0);
2479 ff_synth_filter_init(&s->synth);
2480 ff_dcadsp_init(&s->dcadsp);
2481 ff_fmt_convert_init(&s->fmt_conv, avctx);
2483 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
2485 /* allow downmixing to stereo */
2486 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
2487 avctx->request_channels == 2) {
2488 avctx->channels = avctx->request_channels;
2494 static av_cold int dca_decode_end(AVCodecContext *avctx)
2496 DCAContext *s = avctx->priv_data;
2497 ff_mdct_end(&s->imdct);
2498 av_freep(&s->extra_channels_buffer);
2502 static const AVProfile profiles[] = {
2503 { FF_PROFILE_DTS, "DTS" },
2504 { FF_PROFILE_DTS_ES, "DTS-ES" },
2505 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2506 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2507 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2508 { FF_PROFILE_UNKNOWN },
2511 AVCodec ff_dca_decoder = {
2513 .type = AVMEDIA_TYPE_AUDIO,
2514 .id = AV_CODEC_ID_DTS,
2515 .priv_data_size = sizeof(DCAContext),
2516 .init = dca_decode_init,
2517 .decode = dca_decode_frame,
2518 .close = dca_decode_end,
2519 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2520 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2521 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
2522 AV_SAMPLE_FMT_NONE },
2523 .profiles = NULL_IF_CONFIG_SMALL(profiles),