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/common.h"
30 #include "libavutil/intmath.h"
31 #include "libavutil/intreadwrite.h"
32 #include "libavutil/mathematics.h"
33 #include "libavutil/audioconvert.h"
42 #include "dca_parser.h"
43 #include "synth_filter.h"
45 #include "fmtconvert.h"
53 #define DCA_PRIM_CHANNELS_MAX (7)
54 #define DCA_SUBBANDS (64)
55 #define DCA_ABITS_MAX (32) /* Should be 28 */
56 #define DCA_SUBSUBFRAMES_MAX (4)
57 #define DCA_SUBFRAMES_MAX (16)
58 #define DCA_BLOCKS_MAX (16)
59 #define DCA_LFE_MAX (3)
60 #define DCA_CHSETS_MAX (4)
61 #define DCA_CHSET_CHANS_MAX (8)
77 /* these are unconfirmed but should be mostly correct */
78 enum DCAExSSSpeakerMask {
79 DCA_EXSS_FRONT_CENTER = 0x0001,
80 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
81 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
82 DCA_EXSS_LFE = 0x0008,
83 DCA_EXSS_REAR_CENTER = 0x0010,
84 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
85 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
86 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
87 DCA_EXSS_OVERHEAD = 0x0100,
88 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
89 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
90 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
91 DCA_EXSS_LFE2 = 0x1000,
92 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
93 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
94 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
97 enum DCAExtensionMask {
98 DCA_EXT_CORE = 0x001, ///< core in core substream
99 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
100 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
101 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
102 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
103 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
104 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
105 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
106 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
107 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
110 /* -1 are reserved or unknown */
111 static const int dca_ext_audio_descr_mask[] = {
115 DCA_EXT_XCH | DCA_EXT_X96,
122 /* extensions that reside in core substream */
123 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
125 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
126 * Some compromises have been made for special configurations. Most configurations
127 * are never used so complete accuracy is not needed.
129 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
130 * S -> side, when both rear and back are configured move one of them to the side channel
132 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
134 static const uint64_t dca_core_channel_layout[] = {
135 AV_CH_FRONT_CENTER, ///< 1, A
136 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
137 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
138 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
139 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
140 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
141 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
142 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
143 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
145 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
146 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
148 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
149 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
151 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
152 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
154 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
155 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
156 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
158 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
159 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
160 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
162 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
163 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
164 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
166 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
167 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
168 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
171 static const int8_t dca_lfe_index[] = {
172 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
175 static const int8_t dca_channel_reorder_lfe[][9] = {
176 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
177 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
178 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
179 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
180 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
181 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
182 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
183 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
184 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
185 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
186 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
187 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
188 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
189 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
190 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
191 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
194 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
195 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
196 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
197 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
198 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
199 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
200 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
201 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
202 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
203 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
204 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
205 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
206 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
207 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
208 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
209 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
210 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
213 static const int8_t dca_channel_reorder_nolfe[][9] = {
214 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
215 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
216 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
217 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
218 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
219 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
220 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
221 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
222 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
223 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
224 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
225 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
226 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
227 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
228 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
229 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
232 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
233 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
234 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
235 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
236 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
237 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
238 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
239 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
240 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
241 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
242 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
243 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
244 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
245 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
246 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
247 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
248 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
251 #define DCA_DOLBY 101 /* FIXME */
253 #define DCA_CHANNEL_BITS 6
254 #define DCA_CHANNEL_MASK 0x3F
258 #define HEADER_SIZE 14
260 #define DCA_MAX_FRAME_SIZE 16384
261 #define DCA_MAX_EXSS_HEADER_SIZE 4096
263 #define DCA_BUFFER_PADDING_SIZE 1024
265 /** Bit allocation */
267 int offset; ///< code values offset
268 int maxbits[8]; ///< max bits in VLC
269 int wrap; ///< wrap for get_vlc2()
270 VLC vlc[8]; ///< actual codes
273 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
274 static BitAlloc dca_tmode; ///< transition mode VLCs
275 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
276 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
278 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
281 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
286 AVCodecContext *avctx;
289 int frame_type; ///< type of the current frame
290 int samples_deficit; ///< deficit sample count
291 int crc_present; ///< crc is present in the bitstream
292 int sample_blocks; ///< number of PCM sample blocks
293 int frame_size; ///< primary frame byte size
294 int amode; ///< audio channels arrangement
295 int sample_rate; ///< audio sampling rate
296 int bit_rate; ///< transmission bit rate
297 int bit_rate_index; ///< transmission bit rate index
299 int downmix; ///< embedded downmix enabled
300 int dynrange; ///< embedded dynamic range flag
301 int timestamp; ///< embedded time stamp flag
302 int aux_data; ///< auxiliary data flag
303 int hdcd; ///< source material is mastered in HDCD
304 int ext_descr; ///< extension audio descriptor flag
305 int ext_coding; ///< extended coding flag
306 int aspf; ///< audio sync word insertion flag
307 int lfe; ///< low frequency effects flag
308 int predictor_history; ///< predictor history flag
309 int header_crc; ///< header crc check bytes
310 int multirate_inter; ///< multirate interpolator switch
311 int version; ///< encoder software revision
312 int copy_history; ///< copy history
313 int source_pcm_res; ///< source pcm resolution
314 int front_sum; ///< front sum/difference flag
315 int surround_sum; ///< surround sum/difference flag
316 int dialog_norm; ///< dialog normalisation parameter
318 /* Primary audio coding header */
319 int subframes; ///< number of subframes
320 int total_channels; ///< number of channels including extensions
321 int prim_channels; ///< number of primary audio channels
322 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
323 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
324 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
325 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
326 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
327 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
328 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
329 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
331 /* Primary audio coding side information */
332 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
333 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
334 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
335 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
336 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
337 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
338 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
339 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
340 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
341 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
342 int dynrange_coef; ///< dynamic range coefficient
344 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
346 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
347 int lfe_scale_factor;
349 /* Subband samples history (for ADPCM) */
350 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
351 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
352 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
353 int hist_index[DCA_PRIM_CHANNELS_MAX];
354 DECLARE_ALIGNED(32, float, raXin)[32];
356 int output; ///< type of output
357 float scale_bias; ///< output scale
359 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
360 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX + 1) * 256];
361 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
363 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
364 int dca_buffer_size; ///< how much data is in the dca_buffer
366 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
368 /* Current position in DCA frame */
369 int current_subframe;
370 int current_subsubframe;
372 int core_ext_mask; ///< present extensions in the core substream
374 /* XCh extension information */
375 int xch_present; ///< XCh extension present and valid
376 int xch_base_channel; ///< index of first (only) channel containing XCH data
378 /* ExSS header parser */
379 int static_fields; ///< static fields present
380 int mix_metadata; ///< mixing metadata present
381 int num_mix_configs; ///< number of mix out configurations
382 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
386 int debug_flag; ///< used for suppressing repeated error messages output
389 SynthFilterContext synth;
390 DCADSPContext dcadsp;
391 FmtConvertContext fmt_conv;
394 static const uint16_t dca_vlc_offs[] = {
395 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
396 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
397 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
398 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
399 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
400 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
403 static av_cold void dca_init_vlcs(void)
405 static int vlcs_initialized = 0;
407 static VLC_TYPE dca_table[23622][2];
409 if (vlcs_initialized)
412 dca_bitalloc_index.offset = 1;
413 dca_bitalloc_index.wrap = 2;
414 for (i = 0; i < 5; i++) {
415 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
416 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
417 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
418 bitalloc_12_bits[i], 1, 1,
419 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
421 dca_scalefactor.offset = -64;
422 dca_scalefactor.wrap = 2;
423 for (i = 0; i < 5; i++) {
424 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
425 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
426 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
427 scales_bits[i], 1, 1,
428 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
430 dca_tmode.offset = 0;
432 for (i = 0; i < 4; i++) {
433 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
434 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
435 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
437 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
440 for (i = 0; i < 10; i++)
441 for (j = 0; j < 7; j++) {
442 if (!bitalloc_codes[i][j])
444 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
445 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
446 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
447 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
449 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
451 bitalloc_bits[i][j], 1, 1,
452 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
455 vlcs_initialized = 1;
458 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
461 *dst++ = get_bits(gb, bits);
464 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
467 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
468 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
469 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
471 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
472 s->prim_channels = s->total_channels;
474 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
475 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
478 for (i = base_channel; i < s->prim_channels; i++) {
479 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
480 if (s->subband_activity[i] > DCA_SUBBANDS)
481 s->subband_activity[i] = DCA_SUBBANDS;
483 for (i = base_channel; i < s->prim_channels; i++) {
484 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
485 if (s->vq_start_subband[i] > DCA_SUBBANDS)
486 s->vq_start_subband[i] = DCA_SUBBANDS;
488 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
489 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
490 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
491 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
493 /* Get codebooks quantization indexes */
495 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
496 for (j = 1; j < 11; j++)
497 for (i = base_channel; i < s->prim_channels; i++)
498 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
500 /* Get scale factor adjustment */
501 for (j = 0; j < 11; j++)
502 for (i = base_channel; i < s->prim_channels; i++)
503 s->scalefactor_adj[i][j] = 1;
505 for (j = 1; j < 11; j++)
506 for (i = base_channel; i < s->prim_channels; i++)
507 if (s->quant_index_huffman[i][j] < thr[j])
508 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
510 if (s->crc_present) {
511 /* Audio header CRC check */
512 get_bits(&s->gb, 16);
515 s->current_subframe = 0;
516 s->current_subsubframe = 0;
519 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
520 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
521 for (i = base_channel; i < s->prim_channels; i++) {
522 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
523 s->subband_activity[i]);
524 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
525 s->vq_start_subband[i]);
526 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
527 s->joint_intensity[i]);
528 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
529 s->transient_huffman[i]);
530 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
531 s->scalefactor_huffman[i]);
532 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
533 s->bitalloc_huffman[i]);
534 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
535 for (j = 0; j < 11; j++)
536 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
537 av_log(s->avctx, AV_LOG_DEBUG, "\n");
538 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
539 for (j = 0; j < 11; j++)
540 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
541 av_log(s->avctx, AV_LOG_DEBUG, "\n");
548 static int dca_parse_frame_header(DCAContext *s)
550 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
553 skip_bits_long(&s->gb, 32);
556 s->frame_type = get_bits(&s->gb, 1);
557 s->samples_deficit = get_bits(&s->gb, 5) + 1;
558 s->crc_present = get_bits(&s->gb, 1);
559 s->sample_blocks = get_bits(&s->gb, 7) + 1;
560 s->frame_size = get_bits(&s->gb, 14) + 1;
561 if (s->frame_size < 95)
562 return AVERROR_INVALIDDATA;
563 s->amode = get_bits(&s->gb, 6);
564 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
566 return AVERROR_INVALIDDATA;
567 s->bit_rate_index = get_bits(&s->gb, 5);
568 s->bit_rate = dca_bit_rates[s->bit_rate_index];
570 return AVERROR_INVALIDDATA;
572 s->downmix = get_bits(&s->gb, 1);
573 s->dynrange = get_bits(&s->gb, 1);
574 s->timestamp = get_bits(&s->gb, 1);
575 s->aux_data = get_bits(&s->gb, 1);
576 s->hdcd = get_bits(&s->gb, 1);
577 s->ext_descr = get_bits(&s->gb, 3);
578 s->ext_coding = get_bits(&s->gb, 1);
579 s->aspf = get_bits(&s->gb, 1);
580 s->lfe = get_bits(&s->gb, 2);
581 s->predictor_history = get_bits(&s->gb, 1);
583 /* TODO: check CRC */
585 s->header_crc = get_bits(&s->gb, 16);
587 s->multirate_inter = get_bits(&s->gb, 1);
588 s->version = get_bits(&s->gb, 4);
589 s->copy_history = get_bits(&s->gb, 2);
590 s->source_pcm_res = get_bits(&s->gb, 3);
591 s->front_sum = get_bits(&s->gb, 1);
592 s->surround_sum = get_bits(&s->gb, 1);
593 s->dialog_norm = get_bits(&s->gb, 4);
595 /* FIXME: channels mixing levels */
596 s->output = s->amode;
598 s->output |= DCA_LFE;
601 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
602 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
603 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
604 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
605 s->sample_blocks, s->sample_blocks * 32);
606 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
607 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
608 s->amode, dca_channels[s->amode]);
609 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
611 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
613 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
614 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
615 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
616 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
617 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
618 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
619 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
620 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
621 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
622 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
623 s->predictor_history);
624 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
625 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
627 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
628 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
629 av_log(s->avctx, AV_LOG_DEBUG,
630 "source pcm resolution: %i (%i bits/sample)\n",
631 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
632 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
633 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
634 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
635 av_log(s->avctx, AV_LOG_DEBUG, "\n");
638 /* Primary audio coding header */
639 s->subframes = get_bits(&s->gb, 4) + 1;
641 return dca_parse_audio_coding_header(s, 0);
645 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
648 /* huffman encoded */
649 value += get_bitalloc(gb, &dca_scalefactor, level);
650 value = av_clip(value, 0, (1 << log2range) - 1);
651 } else if (level < 8) {
652 if (level + 1 > log2range) {
653 skip_bits(gb, level + 1 - log2range);
654 value = get_bits(gb, log2range);
656 value = get_bits(gb, level + 1);
662 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
664 /* Primary audio coding side information */
667 if (get_bits_left(&s->gb) < 0)
668 return AVERROR_INVALIDDATA;
671 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
672 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
675 for (j = base_channel; j < s->prim_channels; j++) {
676 for (k = 0; k < s->subband_activity[j]; k++)
677 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
680 /* Get prediction codebook */
681 for (j = base_channel; j < s->prim_channels; j++) {
682 for (k = 0; k < s->subband_activity[j]; k++) {
683 if (s->prediction_mode[j][k] > 0) {
684 /* (Prediction coefficient VQ address) */
685 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
690 /* Bit allocation index */
691 for (j = base_channel; j < s->prim_channels; j++) {
692 for (k = 0; k < s->vq_start_subband[j]; k++) {
693 if (s->bitalloc_huffman[j] == 6)
694 s->bitalloc[j][k] = get_bits(&s->gb, 5);
695 else if (s->bitalloc_huffman[j] == 5)
696 s->bitalloc[j][k] = get_bits(&s->gb, 4);
697 else if (s->bitalloc_huffman[j] == 7) {
698 av_log(s->avctx, AV_LOG_ERROR,
699 "Invalid bit allocation index\n");
700 return AVERROR_INVALIDDATA;
703 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
706 if (s->bitalloc[j][k] > 26) {
707 // av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index [%i][%i] too big (%i)\n",
708 // j, k, s->bitalloc[j][k]);
709 return AVERROR_INVALIDDATA;
714 /* Transition mode */
715 for (j = base_channel; j < s->prim_channels; j++) {
716 for (k = 0; k < s->subband_activity[j]; k++) {
717 s->transition_mode[j][k] = 0;
718 if (s->subsubframes[s->current_subframe] > 1 &&
719 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
720 s->transition_mode[j][k] =
721 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
726 if (get_bits_left(&s->gb) < 0)
727 return AVERROR_INVALIDDATA;
729 for (j = base_channel; j < s->prim_channels; j++) {
730 const uint32_t *scale_table;
731 int scale_sum, log_size;
733 memset(s->scale_factor[j], 0,
734 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
736 if (s->scalefactor_huffman[j] == 6) {
737 scale_table = scale_factor_quant7;
740 scale_table = scale_factor_quant6;
744 /* When huffman coded, only the difference is encoded */
747 for (k = 0; k < s->subband_activity[j]; k++) {
748 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
749 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
750 s->scale_factor[j][k][0] = scale_table[scale_sum];
753 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
754 /* Get second scale factor */
755 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
756 s->scale_factor[j][k][1] = scale_table[scale_sum];
761 /* Joint subband scale factor codebook select */
762 for (j = base_channel; j < s->prim_channels; j++) {
763 /* Transmitted only if joint subband coding enabled */
764 if (s->joint_intensity[j] > 0)
765 s->joint_huff[j] = get_bits(&s->gb, 3);
768 if (get_bits_left(&s->gb) < 0)
769 return AVERROR_INVALIDDATA;
771 /* Scale factors for joint subband coding */
772 for (j = base_channel; j < s->prim_channels; j++) {
775 /* Transmitted only if joint subband coding enabled */
776 if (s->joint_intensity[j] > 0) {
778 source_channel = s->joint_intensity[j] - 1;
780 /* When huffman coded, only the difference is encoded
781 * (is this valid as well for joint scales ???) */
783 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
784 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
785 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
788 if (!(s->debug_flag & 0x02)) {
789 av_log(s->avctx, AV_LOG_DEBUG,
790 "Joint stereo coding not supported\n");
791 s->debug_flag |= 0x02;
796 /* Stereo downmix coefficients */
797 if (!base_channel && s->prim_channels > 2) {
799 for (j = base_channel; j < s->prim_channels; j++) {
800 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
801 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
804 int am = s->amode & DCA_CHANNEL_MASK;
805 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
806 av_log(s->avctx, AV_LOG_ERROR,
807 "Invalid channel mode %d\n", am);
808 return AVERROR_INVALIDDATA;
810 for (j = base_channel; j < FFMIN(s->prim_channels, FF_ARRAY_ELEMS(dca_default_coeffs[am])); j++) {
811 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
812 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
817 /* Dynamic range coefficient */
818 if (!base_channel && s->dynrange)
819 s->dynrange_coef = get_bits(&s->gb, 8);
821 /* Side information CRC check word */
822 if (s->crc_present) {
823 get_bits(&s->gb, 16);
827 * Primary audio data arrays
830 /* VQ encoded high frequency subbands */
831 for (j = base_channel; j < s->prim_channels; j++)
832 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
833 /* 1 vector -> 32 samples */
834 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
836 /* Low frequency effect data */
837 if (!base_channel && s->lfe) {
840 int lfe_samples = 2 * s->lfe * (4 + block_index);
841 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
844 for (j = lfe_samples; j < lfe_end_sample; j++) {
845 /* Signed 8 bits int */
846 s->lfe_data[j] = get_sbits(&s->gb, 8);
849 /* Scale factor index */
850 quant7 = get_bits(&s->gb, 8);
852 av_log_ask_for_sample(s->avctx, "LFEScaleIndex larger than 127\n");
853 return AVERROR_INVALIDDATA;
855 s->lfe_scale_factor = scale_factor_quant7[quant7];
857 /* Quantization step size * scale factor */
858 lfe_scale = 0.035 * s->lfe_scale_factor;
860 for (j = lfe_samples; j < lfe_end_sample; j++)
861 s->lfe_data[j] *= lfe_scale;
865 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
866 s->subsubframes[s->current_subframe]);
867 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
868 s->partial_samples[s->current_subframe]);
870 for (j = base_channel; j < s->prim_channels; j++) {
871 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
872 for (k = 0; k < s->subband_activity[j]; k++)
873 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
874 av_log(s->avctx, AV_LOG_DEBUG, "\n");
876 for (j = base_channel; j < s->prim_channels; j++) {
877 for (k = 0; k < s->subband_activity[j]; k++)
878 av_log(s->avctx, AV_LOG_DEBUG,
879 "prediction coefs: %f, %f, %f, %f\n",
880 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
881 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
882 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
883 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
885 for (j = base_channel; j < s->prim_channels; j++) {
886 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
887 for (k = 0; k < s->vq_start_subband[j]; k++)
888 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
889 av_log(s->avctx, AV_LOG_DEBUG, "\n");
891 for (j = base_channel; j < s->prim_channels; j++) {
892 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
893 for (k = 0; k < s->subband_activity[j]; k++)
894 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
895 av_log(s->avctx, AV_LOG_DEBUG, "\n");
897 for (j = base_channel; j < s->prim_channels; j++) {
898 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
899 for (k = 0; k < s->subband_activity[j]; k++) {
900 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
901 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
902 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
903 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
905 av_log(s->avctx, AV_LOG_DEBUG, "\n");
907 for (j = base_channel; j < s->prim_channels; j++) {
908 if (s->joint_intensity[j] > 0) {
909 int source_channel = s->joint_intensity[j] - 1;
910 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
911 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
912 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
913 av_log(s->avctx, AV_LOG_DEBUG, "\n");
916 if (!base_channel && s->prim_channels > 2 && s->downmix) {
917 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
918 for (j = 0; j < s->prim_channels; j++) {
919 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
920 dca_downmix_coeffs[s->downmix_coef[j][0]]);
921 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
922 dca_downmix_coeffs[s->downmix_coef[j][1]]);
924 av_log(s->avctx, AV_LOG_DEBUG, "\n");
926 for (j = base_channel; j < s->prim_channels; j++)
927 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
928 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
929 if (!base_channel && s->lfe) {
930 int lfe_samples = 2 * s->lfe * (4 + block_index);
931 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
933 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
934 for (j = lfe_samples; j < lfe_end_sample; j++)
935 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
936 av_log(s->avctx, AV_LOG_DEBUG, "\n");
943 static void qmf_32_subbands(DCAContext *s, int chans,
944 float samples_in[32][8], float *samples_out,
947 const float *prCoeff;
950 int sb_act = s->subband_activity[chans];
953 scale *= sqrt(1 / 8.0);
956 if (!s->multirate_inter) /* Non-perfect reconstruction */
957 prCoeff = fir_32bands_nonperfect;
958 else /* Perfect reconstruction */
959 prCoeff = fir_32bands_perfect;
961 for (i = sb_act; i < 32; i++)
964 /* Reconstructed channel sample index */
965 for (subindex = 0; subindex < 8; subindex++) {
966 /* Load in one sample from each subband and clear inactive subbands */
967 for (i = 0; i < sb_act; i++) {
968 unsigned sign = (i - 1) & 2;
969 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
970 AV_WN32A(&s->raXin[i], v);
973 s->synth.synth_filter_float(&s->imdct,
974 s->subband_fir_hist[chans],
975 &s->hist_index[chans],
976 s->subband_fir_noidea[chans], prCoeff,
977 samples_out, s->raXin, scale);
982 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
983 int num_deci_sample, float *samples_in,
984 float *samples_out, float scale)
986 /* samples_in: An array holding decimated samples.
987 * Samples in current subframe starts from samples_in[0],
988 * while samples_in[-1], samples_in[-2], ..., stores samples
989 * from last subframe as history.
991 * samples_out: An array holding interpolated samples
995 const float *prCoeff;
998 /* Select decimation filter */
999 if (decimation_select == 1) {
1001 prCoeff = lfe_fir_128;
1004 prCoeff = lfe_fir_64;
1007 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1008 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1010 samples_out += 2 * decifactor;
1014 /* downmixing routines */
1015 #define MIX_REAR1(samples, si1, rs, coef) \
1016 samples[i] += samples[si1] * coef[rs][0]; \
1017 samples[i+256] += samples[si1] * coef[rs][1];
1019 #define MIX_REAR2(samples, si1, si2, rs, coef) \
1020 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs + 1][0]; \
1021 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs + 1][1];
1023 #define MIX_FRONT3(samples, coef) \
1024 t = samples[i + c]; \
1025 u = samples[i + l]; \
1026 v = samples[i + r]; \
1027 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1028 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1030 #define DOWNMIX_TO_STEREO(op1, op2) \
1031 for (i = 0; i < 256; i++) { \
1036 static void dca_downmix(float *samples, int srcfmt,
1037 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1038 const int8_t *channel_mapping)
1040 int c, l, r, sl, sr, s;
1043 float coef[DCA_PRIM_CHANNELS_MAX][2];
1045 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1046 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1047 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1053 case DCA_STEREO_TOTAL:
1054 case DCA_STEREO_SUMDIFF:
1056 av_log(NULL, 0, "Not implemented!\n");
1061 c = channel_mapping[0] * 256;
1062 l = channel_mapping[1] * 256;
1063 r = channel_mapping[2] * 256;
1064 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1067 s = channel_mapping[2] * 256;
1068 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef), );
1071 c = channel_mapping[0] * 256;
1072 l = channel_mapping[1] * 256;
1073 r = channel_mapping[2] * 256;
1074 s = channel_mapping[3] * 256;
1075 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1076 MIX_REAR1(samples, i + s, 3, coef));
1079 sl = channel_mapping[2] * 256;
1080 sr = channel_mapping[3] * 256;
1081 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef), );
1084 c = channel_mapping[0] * 256;
1085 l = channel_mapping[1] * 256;
1086 r = channel_mapping[2] * 256;
1087 sl = channel_mapping[3] * 256;
1088 sr = channel_mapping[4] * 256;
1089 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1090 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1096 #ifndef decode_blockcodes
1097 /* Very compact version of the block code decoder that does not use table
1098 * look-up but is slightly slower */
1099 static int decode_blockcode(int code, int levels, int *values)
1102 int offset = (levels - 1) >> 1;
1104 for (i = 0; i < 4; i++) {
1105 int div = FASTDIV(code, levels);
1106 values[i] = code - offset - div * levels;
1113 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1115 return decode_blockcode(code1, levels, values) |
1116 decode_blockcode(code2, levels, values + 4);
1120 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1121 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1123 #ifndef int8x8_fmul_int32
1124 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1126 float fscale = scale / 16.0;
1128 for (i = 0; i < 8; i++)
1129 dst[i] = src[i] * fscale;
1133 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1136 int subsubframe = s->current_subsubframe;
1138 const float *quant_step_table;
1141 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1142 LOCAL_ALIGNED_16(int, block, [8]);
1148 /* Select quantization step size table */
1149 if (s->bit_rate_index == 0x1f)
1150 quant_step_table = lossless_quant_d;
1152 quant_step_table = lossy_quant_d;
1154 for (k = base_channel; k < s->prim_channels; k++) {
1155 if (get_bits_left(&s->gb) < 0)
1156 return AVERROR_INVALIDDATA;
1158 for (l = 0; l < s->vq_start_subband[k]; l++) {
1161 /* Select the mid-tread linear quantizer */
1162 int abits = s->bitalloc[k][l];
1164 float quant_step_size = quant_step_table[abits];
1167 * Determine quantization index code book and its type
1170 /* Select quantization index code book */
1171 int sel = s->quant_index_huffman[k][abits];
1174 * Extract bits from the bit stream
1177 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1179 /* Deal with transients */
1180 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1181 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1182 s->scalefactor_adj[k][sel];
1184 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1187 int block_code1, block_code2, size, levels, err;
1189 size = abits_sizes[abits - 1];
1190 levels = abits_levels[abits - 1];
1192 block_code1 = get_bits(&s->gb, size);
1193 block_code2 = get_bits(&s->gb, size);
1194 err = decode_blockcodes(block_code1, block_code2,
1197 av_log(s->avctx, AV_LOG_ERROR,
1198 "ERROR: block code look-up failed\n");
1199 return AVERROR_INVALIDDATA;
1203 for (m = 0; m < 8; m++)
1204 block[m] = get_sbits(&s->gb, abits - 3);
1208 for (m = 0; m < 8; m++)
1209 block[m] = get_bitalloc(&s->gb,
1210 &dca_smpl_bitalloc[abits], sel);
1213 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1218 * Inverse ADPCM if in prediction mode
1220 if (s->prediction_mode[k][l]) {
1222 for (m = 0; m < 8; m++) {
1223 for (n = 1; n <= 4; n++)
1225 subband_samples[k][l][m] +=
1226 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1227 subband_samples[k][l][m - n] / 8192);
1228 else if (s->predictor_history)
1229 subband_samples[k][l][m] +=
1230 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1231 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1237 * Decode VQ encoded high frequencies
1239 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1240 /* 1 vector -> 32 samples but we only need the 8 samples
1241 * for this subsubframe. */
1242 int hfvq = s->high_freq_vq[k][l];
1244 if (!s->debug_flag & 0x01) {
1245 av_log(s->avctx, AV_LOG_DEBUG,
1246 "Stream with high frequencies VQ coding\n");
1247 s->debug_flag |= 0x01;
1250 int8x8_fmul_int32(subband_samples[k][l],
1251 &high_freq_vq[hfvq][subsubframe * 8],
1252 s->scale_factor[k][l][0]);
1256 /* Check for DSYNC after subsubframe */
1257 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1258 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1260 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1263 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1267 /* Backup predictor history for adpcm */
1268 for (k = base_channel; k < s->prim_channels; k++)
1269 for (l = 0; l < s->vq_start_subband[k]; l++)
1270 memcpy(s->subband_samples_hist[k][l],
1271 &subband_samples[k][l][4],
1272 4 * sizeof(subband_samples[0][0][0]));
1277 static int dca_filter_channels(DCAContext *s, int block_index)
1279 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1282 /* 32 subbands QMF */
1283 for (k = 0; k < s->prim_channels; k++) {
1284 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1285 0, 8388608.0, 8388608.0 };*/
1286 qmf_32_subbands(s, k, subband_samples[k],
1287 &s->samples[256 * s->channel_order_tab[k]],
1288 M_SQRT1_2 * s->scale_bias /* pcm_to_double[s->source_pcm_res] */);
1292 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1293 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1296 /* Generate LFE samples for this subsubframe FIXME!!! */
1297 if (s->output & DCA_LFE) {
1298 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1299 s->lfe_data + 2 * s->lfe * (block_index + 4),
1300 &s->samples[256 * dca_lfe_index[s->amode]],
1301 (1.0 / 256.0) * s->scale_bias);
1302 /* Outputs 20bits pcm samples */
1309 static int dca_subframe_footer(DCAContext *s, int base_channel)
1311 int aux_data_count = 0, i;
1314 * Unpack optional information
1317 /* presumably optional information only appears in the core? */
1318 if (!base_channel) {
1320 skip_bits_long(&s->gb, 32);
1323 aux_data_count = get_bits(&s->gb, 6);
1325 for (i = 0; i < aux_data_count; i++)
1326 get_bits(&s->gb, 8);
1328 if (s->crc_present && (s->downmix || s->dynrange))
1329 get_bits(&s->gb, 16);
1336 * Decode a dca frame block
1338 * @param s pointer to the DCAContext
1341 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1346 if (s->current_subframe >= s->subframes) {
1347 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1348 s->current_subframe, s->subframes);
1349 return AVERROR_INVALIDDATA;
1352 if (!s->current_subsubframe) {
1354 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1356 /* Read subframe header */
1357 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1361 /* Read subsubframe */
1363 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1365 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1369 s->current_subsubframe++;
1370 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1371 s->current_subsubframe = 0;
1372 s->current_subframe++;
1374 if (s->current_subframe >= s->subframes) {
1376 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1378 /* Read subframe footer */
1379 if ((ret = dca_subframe_footer(s, base_channel)))
1387 * Return the number of channels in an ExSS speaker mask (HD)
1389 static int dca_exss_mask2count(int mask)
1391 /* count bits that mean speaker pairs twice */
1392 return av_popcount(mask) +
1393 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1394 DCA_EXSS_FRONT_LEFT_RIGHT |
1395 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1396 DCA_EXSS_WIDE_LEFT_RIGHT |
1397 DCA_EXSS_SIDE_LEFT_RIGHT |
1398 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1399 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1400 DCA_EXSS_REAR_LEFT_RIGHT |
1401 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1405 * Skip mixing coefficients of a single mix out configuration (HD)
1407 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1411 for (i = 0; i < channels; i++) {
1412 int mix_map_mask = get_bits(gb, out_ch);
1413 int num_coeffs = av_popcount(mix_map_mask);
1414 skip_bits_long(gb, num_coeffs * 6);
1419 * Parse extension substream asset header (HD)
1421 static int dca_exss_parse_asset_header(DCAContext *s)
1423 int header_pos = get_bits_count(&s->gb);
1426 int embedded_stereo = 0;
1427 int embedded_6ch = 0;
1428 int drc_code_present;
1429 int av_uninit(extensions_mask);
1432 if (get_bits_left(&s->gb) < 16)
1435 /* We will parse just enough to get to the extensions bitmask with which
1436 * we can set the profile value. */
1438 header_size = get_bits(&s->gb, 9) + 1;
1439 skip_bits(&s->gb, 3); // asset index
1441 if (s->static_fields) {
1442 if (get_bits1(&s->gb))
1443 skip_bits(&s->gb, 4); // asset type descriptor
1444 if (get_bits1(&s->gb))
1445 skip_bits_long(&s->gb, 24); // language descriptor
1447 if (get_bits1(&s->gb)) {
1448 /* How can one fit 1024 bytes of text here if the maximum value
1449 * for the asset header size field above was 512 bytes? */
1450 int text_length = get_bits(&s->gb, 10) + 1;
1451 if (get_bits_left(&s->gb) < text_length * 8)
1453 skip_bits_long(&s->gb, text_length * 8); // info text
1456 skip_bits(&s->gb, 5); // bit resolution - 1
1457 skip_bits(&s->gb, 4); // max sample rate code
1458 channels = get_bits(&s->gb, 8) + 1;
1460 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1461 int spkr_remap_sets;
1462 int spkr_mask_size = 16;
1466 embedded_stereo = get_bits1(&s->gb);
1468 embedded_6ch = get_bits1(&s->gb);
1470 if (get_bits1(&s->gb)) {
1471 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1472 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1475 spkr_remap_sets = get_bits(&s->gb, 3);
1477 for (i = 0; i < spkr_remap_sets; i++) {
1478 /* std layout mask for each remap set */
1479 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1482 for (i = 0; i < spkr_remap_sets; i++) {
1483 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1484 if (get_bits_left(&s->gb) < 0)
1487 for (j = 0; j < num_spkrs[i]; j++) {
1488 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1489 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1490 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1495 skip_bits(&s->gb, 3); // representation type
1499 drc_code_present = get_bits1(&s->gb);
1500 if (drc_code_present)
1501 get_bits(&s->gb, 8); // drc code
1503 if (get_bits1(&s->gb))
1504 skip_bits(&s->gb, 5); // dialog normalization code
1506 if (drc_code_present && embedded_stereo)
1507 get_bits(&s->gb, 8); // drc stereo code
1509 if (s->mix_metadata && get_bits1(&s->gb)) {
1510 skip_bits(&s->gb, 1); // external mix
1511 skip_bits(&s->gb, 6); // post mix gain code
1513 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1514 skip_bits(&s->gb, 3); // drc limit
1516 skip_bits(&s->gb, 8); // custom drc code
1518 if (get_bits1(&s->gb)) // channel specific scaling
1519 for (i = 0; i < s->num_mix_configs; i++)
1520 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1522 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1524 for (i = 0; i < s->num_mix_configs; i++) {
1525 if (get_bits_left(&s->gb) < 0)
1527 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1529 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1530 if (embedded_stereo)
1531 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1535 switch (get_bits(&s->gb, 2)) {
1536 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1537 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1538 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1539 case 3: extensions_mask = 0; /* aux coding */ break;
1542 /* not parsed further, we were only interested in the extensions mask */
1544 if (get_bits_left(&s->gb) < 0)
1547 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1548 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1551 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1553 if (extensions_mask & DCA_EXT_EXSS_XLL)
1554 s->profile = FF_PROFILE_DTS_HD_MA;
1555 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1557 s->profile = FF_PROFILE_DTS_HD_HRA;
1559 if (!(extensions_mask & DCA_EXT_CORE))
1560 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1561 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1562 av_log(s->avctx, AV_LOG_WARNING,
1563 "DTS extensions detection mismatch (%d, %d)\n",
1564 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1569 static int dca_xbr_parse_frame(DCAContext *s)
1571 int scale_table_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS][2];
1572 int active_bands[DCA_CHSETS_MAX][DCA_CHSET_CHANS_MAX];
1573 int abits_high[DCA_CHSET_CHANS_MAX][DCA_SUBBANDS];
1574 int anctemp[DCA_CHSET_CHANS_MAX];
1575 int chset_fsize[DCA_CHSETS_MAX];
1576 int n_xbr_ch[DCA_CHSETS_MAX];
1577 int hdr_size, num_chsets, xbr_tmode, hdr_pos;
1578 int i, j, k, l, chset, chan_base;
1580 av_log(s->avctx, AV_LOG_DEBUG, "DTS-XBR: decoding XBR extension\n");
1582 /* get bit position of sync header */
1583 hdr_pos = get_bits_count(&s->gb) - 32;
1585 hdr_size = get_bits(&s->gb, 6) + 1;
1586 num_chsets = get_bits(&s->gb, 2) + 1;
1588 for(i = 0; i < num_chsets; i++)
1589 chset_fsize[i] = get_bits(&s->gb, 14) + 1;
1591 xbr_tmode = get_bits1(&s->gb);
1593 for(i = 0; i < num_chsets; i++) {
1594 n_xbr_ch[i] = get_bits(&s->gb, 3) + 1;
1595 k = get_bits(&s->gb, 2) + 5;
1596 for(j = 0; j < n_xbr_ch[i]; j++)
1597 active_bands[i][j] = get_bits(&s->gb, k) + 1;
1600 /* skip to the end of the header */
1601 i = get_bits_count(&s->gb);
1602 if(hdr_pos + hdr_size * 8 > i)
1603 skip_bits_long(&s->gb, hdr_pos + hdr_size * 8 - i);
1605 /* loop over the channel data sets */
1606 /* only decode as many channels as we've decoded base data for */
1607 for(chset = 0, chan_base = 0;
1608 chset < num_chsets && chan_base + n_xbr_ch[chset] <= s->prim_channels;
1609 chan_base += n_xbr_ch[chset++]) {
1610 int start_posn = get_bits_count(&s->gb);
1611 int subsubframe = 0;
1614 /* loop over subframes */
1615 for (k = 0; k < (s->sample_blocks / 8); k++) {
1616 /* parse header if we're on first subsubframe of a block */
1617 if(subsubframe == 0) {
1618 /* Parse subframe header */
1619 for(i = 0; i < n_xbr_ch[chset]; i++) {
1620 anctemp[i] = get_bits(&s->gb, 2) + 2;
1623 for(i = 0; i < n_xbr_ch[chset]; i++) {
1624 get_array(&s->gb, abits_high[i], active_bands[chset][i], anctemp[i]);
1627 for(i = 0; i < n_xbr_ch[chset]; i++) {
1628 anctemp[i] = get_bits(&s->gb, 3);
1629 if(anctemp[i] < 1) {
1630 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: SYNC ERROR\n");
1631 return AVERROR_INVALIDDATA;
1635 /* generate scale factors */
1636 for(i = 0; i < n_xbr_ch[chset]; i++) {
1637 const uint32_t *scale_table;
1640 if (s->scalefactor_huffman[chan_base+i] == 6) {
1641 scale_table = scale_factor_quant7;
1643 scale_table = scale_factor_quant6;
1648 for(j = 0; j < active_bands[chset][i]; j++) {
1649 if(abits_high[i][j] > 0) {
1650 scale_table_high[i][j][0] =
1651 scale_table[get_bits(&s->gb, nbits)];
1653 if(xbr_tmode && s->transition_mode[i][j]) {
1654 scale_table_high[i][j][1] =
1655 scale_table[get_bits(&s->gb, nbits)];
1662 /* decode audio array for this block */
1663 for(i = 0; i < n_xbr_ch[chset]; i++) {
1664 for(j = 0; j < active_bands[chset][i]; j++) {
1665 const int xbr_abits = abits_high[i][j];
1666 const float quant_step_size = lossless_quant_d[xbr_abits];
1667 const int sfi = xbr_tmode && s->transition_mode[i][j] && subsubframe >= s->transition_mode[i][j];
1668 const float rscale = quant_step_size * scale_table_high[i][j][sfi];
1669 float *subband_samples = s->subband_samples[k][chan_base+i][j];
1676 get_array(&s->gb, block, 8, xbr_abits - 3);
1678 int block_code1, block_code2, size, levels, err;
1680 size = abits_sizes[xbr_abits - 1];
1681 levels = abits_levels[xbr_abits - 1];
1683 block_code1 = get_bits(&s->gb, size);
1684 block_code2 = get_bits(&s->gb, size);
1685 err = decode_blockcodes(block_code1, block_code2,
1688 av_log(s->avctx, AV_LOG_ERROR,
1689 "ERROR: DTS-XBR: block code look-up failed\n");
1690 return AVERROR_INVALIDDATA;
1694 /* scale & sum into subband */
1695 for(l = 0; l < 8; l++)
1696 subband_samples[l] += (float)block[l] * rscale;
1700 /* check DSYNC marker */
1701 if(s->aspf || subsubframe == s->subsubframes[subframe] - 1) {
1702 if(get_bits(&s->gb, 16) != 0xffff) {
1703 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: Didn't get subframe DSYNC\n");
1704 return AVERROR_INVALIDDATA;
1708 /* advance sub-sub-frame index */
1709 if(++subsubframe >= s->subsubframes[subframe]) {
1715 /* skip to next channel set */
1716 i = get_bits_count(&s->gb);
1717 if(start_posn + chset_fsize[chset] * 8 != i) {
1718 j = start_posn + chset_fsize[chset] * 8 - i;
1720 av_log(s->avctx, AV_LOG_ERROR, "DTS-XBR: end of channel set,"
1721 " skipping further than expected (%d bits)\n", j);
1722 skip_bits_long(&s->gb, j);
1730 * Parse extension substream header (HD)
1732 static void dca_exss_parse_header(DCAContext *s)
1738 int active_ss_mask[8];
1744 if (get_bits_left(&s->gb) < 52)
1747 start_posn = get_bits_count(&s->gb) - 32;
1749 skip_bits(&s->gb, 8); // user data
1750 ss_index = get_bits(&s->gb, 2);
1752 blownup = get_bits1(&s->gb);
1753 hdrsize = get_bits(&s->gb, 8 + 4 * blownup) + 1; // header_size
1754 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1756 s->static_fields = get_bits1(&s->gb);
1757 if (s->static_fields) {
1758 skip_bits(&s->gb, 2); // reference clock code
1759 skip_bits(&s->gb, 3); // frame duration code
1761 if (get_bits1(&s->gb))
1762 skip_bits_long(&s->gb, 36); // timestamp
1764 /* a single stream can contain multiple audio assets that can be
1765 * combined to form multiple audio presentations */
1767 num_audiop = get_bits(&s->gb, 3) + 1;
1768 if (num_audiop > 1) {
1769 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1770 /* ignore such streams for now */
1774 num_assets = get_bits(&s->gb, 3) + 1;
1775 if (num_assets > 1) {
1776 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1777 /* ignore such streams for now */
1781 for (i = 0; i < num_audiop; i++)
1782 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1784 for (i = 0; i < num_audiop; i++)
1785 for (j = 0; j <= ss_index; j++)
1786 if (active_ss_mask[i] & (1 << j))
1787 skip_bits(&s->gb, 8); // active asset mask
1789 s->mix_metadata = get_bits1(&s->gb);
1790 if (s->mix_metadata) {
1791 int mix_out_mask_size;
1793 skip_bits(&s->gb, 2); // adjustment level
1794 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1795 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1797 for (i = 0; i < s->num_mix_configs; i++) {
1798 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1799 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1804 for (i = 0; i < num_assets; i++)
1805 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1807 for (i = 0; i < num_assets; i++) {
1808 if (dca_exss_parse_asset_header(s))
1812 /* not parsed further, we were only interested in the extensions mask
1813 * from the asset header */
1815 if(num_assets > 0) {
1816 j = get_bits_count(&s->gb);
1817 if(start_posn + hdrsize * 8 > j)
1818 skip_bits_long(&s->gb, start_posn + hdrsize * 8 - j);
1820 /* check first asset for XBR -- should also check extension mask! */
1821 mkr = get_bits_long(&s->gb, 32);
1823 if(mkr == 0x655e315e)
1824 dca_xbr_parse_frame(s);
1829 * Main frame decoding function
1830 * FIXME add arguments
1832 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1833 int *got_frame_ptr, AVPacket *avpkt)
1835 const uint8_t *buf = avpkt->data;
1836 int buf_size = avpkt->size;
1839 int num_core_channels = 0;
1842 int16_t *samples_s16;
1843 DCAContext *s = avctx->priv_data;
1850 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1851 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1852 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1853 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1854 return AVERROR_INVALIDDATA;
1857 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1858 if ((ret = dca_parse_frame_header(s)) < 0) {
1859 //seems like the frame is corrupt, try with the next one
1862 //set AVCodec values with parsed data
1863 avctx->sample_rate = s->sample_rate;
1864 avctx->bit_rate = s->bit_rate;
1866 s->profile = FF_PROFILE_DTS;
1868 for (i = 0; i < (s->sample_blocks / 8); i++) {
1869 if ((ret = dca_decode_block(s, 0, i))) {
1870 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1875 /* record number of core channels incase less than max channels are requested */
1876 num_core_channels = s->prim_channels;
1879 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1881 s->core_ext_mask = 0;
1883 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1885 /* only scan for extensions if ext_descr was unknown or indicated a
1886 * supported XCh extension */
1887 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1889 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1890 * extensions scan can fill it up */
1891 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1893 /* extensions start at 32-bit boundaries into bitstream */
1894 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1896 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1897 uint32_t bits = get_bits_long(&s->gb, 32);
1901 int ext_amode, xch_fsize;
1903 s->xch_base_channel = s->prim_channels;
1905 /* validate sync word using XCHFSIZE field */
1906 xch_fsize = show_bits(&s->gb, 10);
1907 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1908 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1911 /* skip length-to-end-of-frame field for the moment */
1912 skip_bits(&s->gb, 10);
1914 s->core_ext_mask |= DCA_EXT_XCH;
1916 /* extension amode(number of channels in extension) should be 1 */
1917 /* AFAIK XCh is not used for more channels */
1918 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1919 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1920 " supported!\n", ext_amode);
1924 /* much like core primary audio coding header */
1925 dca_parse_audio_coding_header(s, s->xch_base_channel);
1927 for (i = 0; i < (s->sample_blocks / 8); i++)
1928 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1929 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1937 /* XXCh: extended channels */
1938 /* usually found either in core or HD part in DTS-HD HRA streams,
1939 * but not in DTS-ES which contains XCh extensions instead */
1940 s->core_ext_mask |= DCA_EXT_XXCH;
1944 int fsize96 = show_bits(&s->gb, 12) + 1;
1945 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1948 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1949 get_bits_count(&s->gb));
1950 skip_bits(&s->gb, 12);
1951 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1952 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1954 s->core_ext_mask |= DCA_EXT_X96;
1959 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1962 /* no supported extensions, skip the rest of the core substream */
1963 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1966 if (s->core_ext_mask & DCA_EXT_X96)
1967 s->profile = FF_PROFILE_DTS_96_24;
1968 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1969 s->profile = FF_PROFILE_DTS_ES;
1971 /* check for ExSS (HD part) */
1972 if (s->dca_buffer_size - s->frame_size > 32 &&
1973 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1974 dca_exss_parse_header(s);
1976 avctx->profile = s->profile;
1978 channels = s->prim_channels + !!s->lfe;
1980 if (s->amode < 16) {
1981 avctx->channel_layout = dca_core_channel_layout[s->amode];
1983 if (s->xch_present && (!avctx->request_channels ||
1984 avctx->request_channels > num_core_channels + !!s->lfe)) {
1985 avctx->channel_layout |= AV_CH_BACK_CENTER;
1987 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1988 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1990 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1993 channels = num_core_channels + !!s->lfe;
1994 s->xch_present = 0; /* disable further xch processing */
1996 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1997 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1999 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
2002 if (channels > !!s->lfe &&
2003 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
2004 return AVERROR_INVALIDDATA;
2006 if (avctx->request_channels == 2 && s->prim_channels > 2) {
2008 s->output = DCA_STEREO;
2009 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
2011 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
2012 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
2013 s->channel_order_tab = dca_channel_order_native;
2016 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
2017 return AVERROR_INVALIDDATA;
2020 if (avctx->channels != channels) {
2021 if (avctx->channels)
2022 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
2023 avctx->channels = channels;
2026 /* get output buffer */
2027 s->frame.nb_samples = 256 * (s->sample_blocks / 8);
2028 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
2029 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
2032 samples_flt = (float *) s->frame.data[0];
2033 samples_s16 = (int16_t *) s->frame.data[0];
2035 /* filter to get final output */
2036 for (i = 0; i < (s->sample_blocks / 8); i++) {
2037 dca_filter_channels(s, i);
2039 /* If this was marked as a DTS-ES stream we need to subtract back- */
2040 /* channel from SL & SR to remove matrixed back-channel signal */
2041 if ((s->source_pcm_res & 1) && s->xch_present) {
2042 float *back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
2043 float *lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
2044 float *rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
2045 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
2046 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
2049 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
2050 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
2052 samples_flt += 256 * channels;
2054 s->fmt_conv.float_to_int16_interleave(samples_s16,
2055 s->samples_chanptr, 256,
2057 samples_s16 += 256 * channels;
2061 /* update lfe history */
2062 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
2063 for (i = 0; i < 2 * s->lfe * 4; i++)
2064 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
2067 *(AVFrame *) data = s->frame;
2075 * DCA initialization
2077 * @param avctx pointer to the AVCodecContext
2080 static av_cold int dca_decode_init(AVCodecContext *avctx)
2082 DCAContext *s = avctx->priv_data;
2088 ff_dsputil_init(&s->dsp, avctx);
2089 ff_mdct_init(&s->imdct, 6, 1, 1.0);
2090 ff_synth_filter_init(&s->synth);
2091 ff_dcadsp_init(&s->dcadsp);
2092 ff_fmt_convert_init(&s->fmt_conv, avctx);
2094 for (i = 0; i < DCA_PRIM_CHANNELS_MAX + 1; i++)
2095 s->samples_chanptr[i] = s->samples + i * 256;
2097 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
2098 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
2099 s->scale_bias = 1.0 / 32768.0;
2101 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
2102 s->scale_bias = 1.0;
2105 /* allow downmixing to stereo */
2106 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
2107 avctx->request_channels == 2) {
2108 avctx->channels = avctx->request_channels;
2111 avcodec_get_frame_defaults(&s->frame);
2112 avctx->coded_frame = &s->frame;
2117 static av_cold int dca_decode_end(AVCodecContext *avctx)
2119 DCAContext *s = avctx->priv_data;
2120 ff_mdct_end(&s->imdct);
2124 static const AVProfile profiles[] = {
2125 { FF_PROFILE_DTS, "DTS" },
2126 { FF_PROFILE_DTS_ES, "DTS-ES" },
2127 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2128 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2129 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2130 { FF_PROFILE_UNKNOWN },
2133 AVCodec ff_dca_decoder = {
2135 .type = AVMEDIA_TYPE_AUDIO,
2137 .priv_data_size = sizeof(DCAContext),
2138 .init = dca_decode_init,
2139 .decode = dca_decode_frame,
2140 .close = dca_decode_end,
2141 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2142 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2143 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
2145 AV_SAMPLE_FMT_NONE },
2146 .profiles = NULL_IF_CONFIG_SMALL(profiles),