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 Libav.
10 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 #include "libavutil/channel_layout.h"
30 #include "libavutil/common.h"
31 #include "libavutil/float_dsp.h"
32 #include "libavutil/internal.h"
33 #include "libavutil/intreadwrite.h"
34 #include "libavutil/mathematics.h"
35 #include "libavutil/samplefmt.h"
43 #include "dca_parser.h"
45 #include "synth_filter.h"
47 #include "fmtconvert.h"
56 #define DCA_PRIM_CHANNELS_MAX (7)
57 #define DCA_SUBBANDS (32)
58 #define DCA_ABITS_MAX (32) /* Should be 28 */
59 #define DCA_SUBSUBFRAMES_MAX (4)
60 #define DCA_SUBFRAMES_MAX (16)
61 #define DCA_BLOCKS_MAX (16)
62 #define DCA_LFE_MAX (3)
78 /* these are unconfirmed but should be mostly correct */
79 enum DCAExSSSpeakerMask {
80 DCA_EXSS_FRONT_CENTER = 0x0001,
81 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
82 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
83 DCA_EXSS_LFE = 0x0008,
84 DCA_EXSS_REAR_CENTER = 0x0010,
85 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
86 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
87 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
88 DCA_EXSS_OVERHEAD = 0x0100,
89 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
90 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
91 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
92 DCA_EXSS_LFE2 = 0x1000,
93 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
94 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
95 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
98 enum DCAExtensionMask {
99 DCA_EXT_CORE = 0x001, ///< core in core substream
100 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
101 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
102 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
103 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
104 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
105 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
106 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
107 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
108 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
111 /* -1 are reserved or unknown */
112 static const int dca_ext_audio_descr_mask[] = {
116 DCA_EXT_XCH | DCA_EXT_X96,
123 /* extensions that reside in core substream */
124 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
126 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
127 * Some compromises have been made for special configurations. Most configurations
128 * are never used so complete accuracy is not needed.
130 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
131 * S -> side, when both rear and back are configured move one of them to the side channel
133 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
135 static const uint64_t dca_core_channel_layout[] = {
136 AV_CH_FRONT_CENTER, ///< 1, A
137 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
138 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
139 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
140 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
141 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
142 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
143 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
144 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
146 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
147 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
149 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
150 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
152 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
153 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
155 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
156 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
157 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
159 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
160 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
161 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
163 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
164 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
165 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
167 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
168 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
169 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
172 static const int8_t dca_lfe_index[] = {
173 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
176 static const int8_t dca_channel_reorder_lfe[][9] = {
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 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
182 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
183 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
184 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
185 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
186 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
187 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
188 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
189 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
190 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
191 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
192 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
195 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
196 { 0, 2, -1, -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 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
201 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
202 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
203 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
204 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
205 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
206 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
207 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
208 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
209 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
210 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
211 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
214 static const int8_t dca_channel_reorder_nolfe[][9] = {
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 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
220 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
221 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
222 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
223 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
224 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
225 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
226 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
227 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
228 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
229 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
230 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
233 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
234 { 0, 1, -1, -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 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
239 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
240 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
241 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
242 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
243 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
244 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
245 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
246 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
247 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
248 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
249 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
252 #define DCA_DOLBY 101 /* FIXME */
254 #define DCA_CHANNEL_BITS 6
255 #define DCA_CHANNEL_MASK 0x3F
259 #define HEADER_SIZE 14
261 #define DCA_MAX_FRAME_SIZE 16384
262 #define DCA_MAX_EXSS_HEADER_SIZE 4096
264 #define DCA_BUFFER_PADDING_SIZE 1024
266 /** Bit allocation */
268 int offset; ///< code values offset
269 int maxbits[8]; ///< max bits in VLC
270 int wrap; ///< wrap for get_vlc2()
271 VLC vlc[8]; ///< actual codes
274 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
275 static BitAlloc dca_tmode; ///< transition mode VLCs
276 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
277 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
279 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
282 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
287 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
358 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
359 float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
360 float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
361 uint8_t *extra_channels_buffer;
362 unsigned int extra_channels_buffer_size;
364 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
365 int dca_buffer_size; ///< how much data is in the dca_buffer
367 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
369 /* Current position in DCA frame */
370 int current_subframe;
371 int current_subsubframe;
373 int core_ext_mask; ///< present extensions in the core substream
375 /* XCh extension information */
376 int xch_present; ///< XCh extension present and valid
377 int xch_base_channel; ///< index of first (only) channel containing XCH data
379 /* ExSS header parser */
380 int static_fields; ///< static fields present
381 int mix_metadata; ///< mixing metadata present
382 int num_mix_configs; ///< number of mix out configurations
383 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
387 int debug_flag; ///< used for suppressing repeated error messages output
388 AVFloatDSPContext fdsp;
390 SynthFilterContext synth;
391 DCADSPContext dcadsp;
392 FmtConvertContext fmt_conv;
395 static const uint16_t dca_vlc_offs[] = {
396 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
397 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
398 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
399 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
400 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
401 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
404 static av_cold void dca_init_vlcs(void)
406 static int vlcs_initialized = 0;
408 static VLC_TYPE dca_table[23622][2];
410 if (vlcs_initialized)
413 dca_bitalloc_index.offset = 1;
414 dca_bitalloc_index.wrap = 2;
415 for (i = 0; i < 5; i++) {
416 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
417 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
418 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
419 bitalloc_12_bits[i], 1, 1,
420 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
422 dca_scalefactor.offset = -64;
423 dca_scalefactor.wrap = 2;
424 for (i = 0; i < 5; i++) {
425 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
426 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
427 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
428 scales_bits[i], 1, 1,
429 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
431 dca_tmode.offset = 0;
433 for (i = 0; i < 4; i++) {
434 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
435 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
436 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
438 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
441 for (i = 0; i < 10; i++)
442 for (j = 0; j < 7; j++) {
443 if (!bitalloc_codes[i][j])
445 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
446 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
447 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
448 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
450 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
452 bitalloc_bits[i][j], 1, 1,
453 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
456 vlcs_initialized = 1;
459 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
462 *dst++ = get_bits(gb, bits);
465 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
468 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
469 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
470 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
472 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
473 s->prim_channels = s->total_channels;
475 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
476 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
479 for (i = base_channel; i < s->prim_channels; i++) {
480 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
481 if (s->subband_activity[i] > DCA_SUBBANDS)
482 s->subband_activity[i] = DCA_SUBBANDS;
484 for (i = base_channel; i < s->prim_channels; i++) {
485 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
486 if (s->vq_start_subband[i] > DCA_SUBBANDS)
487 s->vq_start_subband[i] = DCA_SUBBANDS;
489 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
490 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
491 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
492 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
494 /* Get codebooks quantization indexes */
496 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
497 for (j = 1; j < 11; j++)
498 for (i = base_channel; i < s->prim_channels; i++)
499 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
501 /* Get scale factor adjustment */
502 for (j = 0; j < 11; j++)
503 for (i = base_channel; i < s->prim_channels; i++)
504 s->scalefactor_adj[i][j] = 1;
506 for (j = 1; j < 11; j++)
507 for (i = base_channel; i < s->prim_channels; i++)
508 if (s->quant_index_huffman[i][j] < thr[j])
509 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
511 if (s->crc_present) {
512 /* Audio header CRC check */
513 get_bits(&s->gb, 16);
516 s->current_subframe = 0;
517 s->current_subsubframe = 0;
520 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
521 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
522 for (i = base_channel; i < s->prim_channels; i++) {
523 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
524 s->subband_activity[i]);
525 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
526 s->vq_start_subband[i]);
527 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
528 s->joint_intensity[i]);
529 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
530 s->transient_huffman[i]);
531 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
532 s->scalefactor_huffman[i]);
533 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
534 s->bitalloc_huffman[i]);
535 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
536 for (j = 0; j < 11; j++)
537 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
538 av_log(s->avctx, AV_LOG_DEBUG, "\n");
539 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
540 for (j = 0; j < 11; j++)
541 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
542 av_log(s->avctx, AV_LOG_DEBUG, "\n");
549 static int dca_parse_frame_header(DCAContext *s)
551 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
554 skip_bits_long(&s->gb, 32);
557 s->frame_type = get_bits(&s->gb, 1);
558 s->samples_deficit = get_bits(&s->gb, 5) + 1;
559 s->crc_present = get_bits(&s->gb, 1);
560 s->sample_blocks = get_bits(&s->gb, 7) + 1;
561 s->frame_size = get_bits(&s->gb, 14) + 1;
562 if (s->frame_size < 95)
563 return AVERROR_INVALIDDATA;
564 s->amode = get_bits(&s->gb, 6);
565 s->sample_rate = avpriv_dca_sample_rates[get_bits(&s->gb, 4)];
567 return AVERROR_INVALIDDATA;
568 s->bit_rate_index = get_bits(&s->gb, 5);
569 s->bit_rate = dca_bit_rates[s->bit_rate_index];
571 return AVERROR_INVALIDDATA;
573 s->downmix = get_bits(&s->gb, 1);
574 s->dynrange = get_bits(&s->gb, 1);
575 s->timestamp = get_bits(&s->gb, 1);
576 s->aux_data = get_bits(&s->gb, 1);
577 s->hdcd = get_bits(&s->gb, 1);
578 s->ext_descr = get_bits(&s->gb, 3);
579 s->ext_coding = get_bits(&s->gb, 1);
580 s->aspf = get_bits(&s->gb, 1);
581 s->lfe = get_bits(&s->gb, 2);
582 s->predictor_history = get_bits(&s->gb, 1);
584 /* TODO: check CRC */
586 s->header_crc = get_bits(&s->gb, 16);
588 s->multirate_inter = get_bits(&s->gb, 1);
589 s->version = get_bits(&s->gb, 4);
590 s->copy_history = get_bits(&s->gb, 2);
591 s->source_pcm_res = get_bits(&s->gb, 3);
592 s->front_sum = get_bits(&s->gb, 1);
593 s->surround_sum = get_bits(&s->gb, 1);
594 s->dialog_norm = get_bits(&s->gb, 4);
596 /* FIXME: channels mixing levels */
597 s->output = s->amode;
599 s->output |= DCA_LFE;
602 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
603 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
604 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
605 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
606 s->sample_blocks, s->sample_blocks * 32);
607 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
608 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
609 s->amode, dca_channels[s->amode]);
610 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
612 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
614 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
615 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
616 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
617 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
618 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
619 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
620 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
621 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
622 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
623 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
624 s->predictor_history);
625 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
626 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
628 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
629 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
630 av_log(s->avctx, AV_LOG_DEBUG,
631 "source pcm resolution: %i (%i bits/sample)\n",
632 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
633 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
634 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
635 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
636 av_log(s->avctx, AV_LOG_DEBUG, "\n");
639 /* Primary audio coding header */
640 s->subframes = get_bits(&s->gb, 4) + 1;
642 return dca_parse_audio_coding_header(s, 0);
646 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
649 /* huffman encoded */
650 value += get_bitalloc(gb, &dca_scalefactor, level);
651 value = av_clip(value, 0, (1 << log2range) - 1);
652 } else if (level < 8) {
653 if (level + 1 > log2range) {
654 skip_bits(gb, level + 1 - log2range);
655 value = get_bits(gb, log2range);
657 value = get_bits(gb, level + 1);
663 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
665 /* Primary audio coding side information */
668 if (get_bits_left(&s->gb) < 0)
669 return AVERROR_INVALIDDATA;
672 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
673 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
676 for (j = base_channel; j < s->prim_channels; j++) {
677 for (k = 0; k < s->subband_activity[j]; k++)
678 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
681 /* Get prediction codebook */
682 for (j = base_channel; j < s->prim_channels; j++) {
683 for (k = 0; k < s->subband_activity[j]; k++) {
684 if (s->prediction_mode[j][k] > 0) {
685 /* (Prediction coefficient VQ address) */
686 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
691 /* Bit allocation index */
692 for (j = base_channel; j < s->prim_channels; j++) {
693 for (k = 0; k < s->vq_start_subband[j]; k++) {
694 if (s->bitalloc_huffman[j] == 6)
695 s->bitalloc[j][k] = get_bits(&s->gb, 5);
696 else if (s->bitalloc_huffman[j] == 5)
697 s->bitalloc[j][k] = get_bits(&s->gb, 4);
698 else if (s->bitalloc_huffman[j] == 7) {
699 av_log(s->avctx, AV_LOG_ERROR,
700 "Invalid bit allocation index\n");
701 return AVERROR_INVALIDDATA;
704 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
707 if (s->bitalloc[j][k] > 26) {
708 av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
709 j, k, s->bitalloc[j][k]);
710 return AVERROR_INVALIDDATA;
715 /* Transition mode */
716 for (j = base_channel; j < s->prim_channels; j++) {
717 for (k = 0; k < s->subband_activity[j]; k++) {
718 s->transition_mode[j][k] = 0;
719 if (s->subsubframes[s->current_subframe] > 1 &&
720 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
721 s->transition_mode[j][k] =
722 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
727 if (get_bits_left(&s->gb) < 0)
728 return AVERROR_INVALIDDATA;
730 for (j = base_channel; j < s->prim_channels; j++) {
731 const uint32_t *scale_table;
732 int scale_sum, log_size;
734 memset(s->scale_factor[j], 0,
735 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
737 if (s->scalefactor_huffman[j] == 6) {
738 scale_table = scale_factor_quant7;
741 scale_table = scale_factor_quant6;
745 /* When huffman coded, only the difference is encoded */
748 for (k = 0; k < s->subband_activity[j]; k++) {
749 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
750 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
751 s->scale_factor[j][k][0] = scale_table[scale_sum];
754 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
755 /* Get second scale factor */
756 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
757 s->scale_factor[j][k][1] = scale_table[scale_sum];
762 /* Joint subband scale factor codebook select */
763 for (j = base_channel; j < s->prim_channels; j++) {
764 /* Transmitted only if joint subband coding enabled */
765 if (s->joint_intensity[j] > 0)
766 s->joint_huff[j] = get_bits(&s->gb, 3);
769 if (get_bits_left(&s->gb) < 0)
770 return AVERROR_INVALIDDATA;
772 /* Scale factors for joint subband coding */
773 for (j = base_channel; j < s->prim_channels; j++) {
776 /* Transmitted only if joint subband coding enabled */
777 if (s->joint_intensity[j] > 0) {
779 source_channel = s->joint_intensity[j] - 1;
781 /* When huffman coded, only the difference is encoded
782 * (is this valid as well for joint scales ???) */
784 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
785 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
786 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
789 if (!(s->debug_flag & 0x02)) {
790 av_log(s->avctx, AV_LOG_DEBUG,
791 "Joint stereo coding not supported\n");
792 s->debug_flag |= 0x02;
797 /* Stereo downmix coefficients */
798 if (!base_channel && s->prim_channels > 2) {
800 for (j = base_channel; j < s->prim_channels; j++) {
801 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
802 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
805 int am = s->amode & DCA_CHANNEL_MASK;
806 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
807 av_log(s->avctx, AV_LOG_ERROR,
808 "Invalid channel mode %d\n", am);
809 return AVERROR_INVALIDDATA;
811 for (j = base_channel; j < s->prim_channels; j++) {
812 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
813 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
818 /* Dynamic range coefficient */
819 if (!base_channel && s->dynrange)
820 s->dynrange_coef = get_bits(&s->gb, 8);
822 /* Side information CRC check word */
823 if (s->crc_present) {
824 get_bits(&s->gb, 16);
828 * Primary audio data arrays
831 /* VQ encoded high frequency subbands */
832 for (j = base_channel; j < s->prim_channels; j++)
833 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
834 /* 1 vector -> 32 samples */
835 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
837 /* Low frequency effect data */
838 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 skip_bits(&s->gb, 1);
851 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 7)];
853 /* Quantization step size * scale factor */
854 lfe_scale = 0.035 * s->lfe_scale_factor;
856 for (j = lfe_samples; j < lfe_end_sample; j++)
857 s->lfe_data[j] *= lfe_scale;
861 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
862 s->subsubframes[s->current_subframe]);
863 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
864 s->partial_samples[s->current_subframe]);
866 for (j = base_channel; j < s->prim_channels; j++) {
867 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
868 for (k = 0; k < s->subband_activity[j]; k++)
869 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
870 av_log(s->avctx, AV_LOG_DEBUG, "\n");
872 for (j = base_channel; j < s->prim_channels; j++) {
873 for (k = 0; k < s->subband_activity[j]; k++)
874 av_log(s->avctx, AV_LOG_DEBUG,
875 "prediction coefs: %f, %f, %f, %f\n",
876 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
877 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
878 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
879 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
881 for (j = base_channel; j < s->prim_channels; j++) {
882 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
883 for (k = 0; k < s->vq_start_subband[j]; k++)
884 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
885 av_log(s->avctx, AV_LOG_DEBUG, "\n");
887 for (j = base_channel; j < s->prim_channels; j++) {
888 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
889 for (k = 0; k < s->subband_activity[j]; k++)
890 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
891 av_log(s->avctx, AV_LOG_DEBUG, "\n");
893 for (j = base_channel; j < s->prim_channels; j++) {
894 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
895 for (k = 0; k < s->subband_activity[j]; k++) {
896 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
897 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
898 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
899 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
901 av_log(s->avctx, AV_LOG_DEBUG, "\n");
903 for (j = base_channel; j < s->prim_channels; j++) {
904 if (s->joint_intensity[j] > 0) {
905 int source_channel = s->joint_intensity[j] - 1;
906 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
907 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
908 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
909 av_log(s->avctx, AV_LOG_DEBUG, "\n");
912 if (!base_channel && s->prim_channels > 2 && s->downmix) {
913 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
914 for (j = 0; j < s->prim_channels; j++) {
915 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
916 dca_downmix_coeffs[s->downmix_coef[j][0]]);
917 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
918 dca_downmix_coeffs[s->downmix_coef[j][1]]);
920 av_log(s->avctx, AV_LOG_DEBUG, "\n");
922 for (j = base_channel; j < s->prim_channels; j++)
923 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
924 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
925 if (!base_channel && s->lfe) {
926 int lfe_samples = 2 * s->lfe * (4 + block_index);
927 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
929 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
930 for (j = lfe_samples; j < lfe_end_sample; j++)
931 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
932 av_log(s->avctx, AV_LOG_DEBUG, "\n");
939 static void qmf_32_subbands(DCAContext *s, int chans,
940 float samples_in[32][8], float *samples_out,
943 const float *prCoeff;
946 int sb_act = s->subband_activity[chans];
949 scale *= sqrt(1 / 8.0);
952 if (!s->multirate_inter) /* Non-perfect reconstruction */
953 prCoeff = fir_32bands_nonperfect;
954 else /* Perfect reconstruction */
955 prCoeff = fir_32bands_perfect;
957 for (i = sb_act; i < 32; i++)
960 /* Reconstructed channel sample index */
961 for (subindex = 0; subindex < 8; subindex++) {
962 /* Load in one sample from each subband and clear inactive subbands */
963 for (i = 0; i < sb_act; i++) {
964 unsigned sign = (i - 1) & 2;
965 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
966 AV_WN32A(&s->raXin[i], v);
969 s->synth.synth_filter_float(&s->imdct,
970 s->subband_fir_hist[chans],
971 &s->hist_index[chans],
972 s->subband_fir_noidea[chans], prCoeff,
973 samples_out, s->raXin, scale);
978 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
979 int num_deci_sample, float *samples_in,
980 float *samples_out, float scale)
982 /* samples_in: An array holding decimated samples.
983 * Samples in current subframe starts from samples_in[0],
984 * while samples_in[-1], samples_in[-2], ..., stores samples
985 * from last subframe as history.
987 * samples_out: An array holding interpolated samples
991 const float *prCoeff;
994 /* Select decimation filter */
995 if (decimation_select == 1) {
997 prCoeff = lfe_fir_128;
1000 prCoeff = lfe_fir_64;
1003 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1004 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1006 samples_out += 2 * decifactor;
1010 /* downmixing routines */
1011 #define MIX_REAR1(samples, s1, rs, coef) \
1012 samples[0][i] += samples[s1][i] * coef[rs][0]; \
1013 samples[1][i] += samples[s1][i] * coef[rs][1];
1015 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1016 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1017 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1019 #define MIX_FRONT3(samples, coef) \
1020 t = samples[c][i]; \
1021 u = samples[l][i]; \
1022 v = samples[r][i]; \
1023 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1024 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1026 #define DOWNMIX_TO_STEREO(op1, op2) \
1027 for (i = 0; i < 256; i++) { \
1032 static void dca_downmix(float **samples, int srcfmt,
1033 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1034 const int8_t *channel_mapping)
1036 int c, l, r, sl, sr, s;
1039 float coef[DCA_PRIM_CHANNELS_MAX][2];
1041 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1042 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1043 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1049 case DCA_STEREO_TOTAL:
1050 case DCA_STEREO_SUMDIFF:
1052 av_log(NULL, 0, "Not implemented!\n");
1057 c = channel_mapping[0];
1058 l = channel_mapping[1];
1059 r = channel_mapping[2];
1060 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1063 s = channel_mapping[2];
1064 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1067 c = channel_mapping[0];
1068 l = channel_mapping[1];
1069 r = channel_mapping[2];
1070 s = channel_mapping[3];
1071 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1072 MIX_REAR1(samples, s, 3, coef));
1075 sl = channel_mapping[2];
1076 sr = channel_mapping[3];
1077 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1080 c = channel_mapping[0];
1081 l = channel_mapping[1];
1082 r = channel_mapping[2];
1083 sl = channel_mapping[3];
1084 sr = channel_mapping[4];
1085 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1086 MIX_REAR2(samples, sl, sr, 3, coef));
1092 #ifndef decode_blockcodes
1093 /* Very compact version of the block code decoder that does not use table
1094 * look-up but is slightly slower */
1095 static int decode_blockcode(int code, int levels, int *values)
1098 int offset = (levels - 1) >> 1;
1100 for (i = 0; i < 4; i++) {
1101 int div = FASTDIV(code, levels);
1102 values[i] = code - offset - div * levels;
1109 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1111 return decode_blockcode(code1, levels, values) |
1112 decode_blockcode(code2, levels, values + 4);
1116 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1117 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1119 #ifndef int8x8_fmul_int32
1120 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1122 float fscale = scale / 16.0;
1124 for (i = 0; i < 8; i++)
1125 dst[i] = src[i] * fscale;
1129 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1132 int subsubframe = s->current_subsubframe;
1134 const float *quant_step_table;
1137 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1138 LOCAL_ALIGNED_16(int, block, [8]);
1144 /* Select quantization step size table */
1145 if (s->bit_rate_index == 0x1f)
1146 quant_step_table = lossless_quant_d;
1148 quant_step_table = lossy_quant_d;
1150 for (k = base_channel; k < s->prim_channels; k++) {
1151 if (get_bits_left(&s->gb) < 0)
1152 return AVERROR_INVALIDDATA;
1154 for (l = 0; l < s->vq_start_subband[k]; l++) {
1157 /* Select the mid-tread linear quantizer */
1158 int abits = s->bitalloc[k][l];
1160 float quant_step_size = quant_step_table[abits];
1163 * Determine quantization index code book and its type
1166 /* Select quantization index code book */
1167 int sel = s->quant_index_huffman[k][abits];
1170 * Extract bits from the bit stream
1173 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1175 /* Deal with transients */
1176 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1177 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1178 s->scalefactor_adj[k][sel];
1180 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1183 int block_code1, block_code2, size, levels, err;
1185 size = abits_sizes[abits - 1];
1186 levels = abits_levels[abits - 1];
1188 block_code1 = get_bits(&s->gb, size);
1189 block_code2 = get_bits(&s->gb, size);
1190 err = decode_blockcodes(block_code1, block_code2,
1193 av_log(s->avctx, AV_LOG_ERROR,
1194 "ERROR: block code look-up failed\n");
1195 return AVERROR_INVALIDDATA;
1199 for (m = 0; m < 8; m++)
1200 block[m] = get_sbits(&s->gb, abits - 3);
1204 for (m = 0; m < 8; m++)
1205 block[m] = get_bitalloc(&s->gb,
1206 &dca_smpl_bitalloc[abits], sel);
1209 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1214 * Inverse ADPCM if in prediction mode
1216 if (s->prediction_mode[k][l]) {
1218 for (m = 0; m < 8; m++) {
1219 for (n = 1; n <= 4; n++)
1221 subband_samples[k][l][m] +=
1222 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1223 subband_samples[k][l][m - n] / 8192);
1224 else if (s->predictor_history)
1225 subband_samples[k][l][m] +=
1226 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1227 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1233 * Decode VQ encoded high frequencies
1235 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1236 /* 1 vector -> 32 samples but we only need the 8 samples
1237 * for this subsubframe. */
1238 int hfvq = s->high_freq_vq[k][l];
1240 if (!s->debug_flag & 0x01) {
1241 av_log(s->avctx, AV_LOG_DEBUG,
1242 "Stream with high frequencies VQ coding\n");
1243 s->debug_flag |= 0x01;
1246 int8x8_fmul_int32(subband_samples[k][l],
1247 &high_freq_vq[hfvq][subsubframe * 8],
1248 s->scale_factor[k][l][0]);
1252 /* Check for DSYNC after subsubframe */
1253 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1254 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1256 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1259 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1263 /* Backup predictor history for adpcm */
1264 for (k = base_channel; k < s->prim_channels; k++)
1265 for (l = 0; l < s->vq_start_subband[k]; l++)
1266 memcpy(s->subband_samples_hist[k][l],
1267 &subband_samples[k][l][4],
1268 4 * sizeof(subband_samples[0][0][0]));
1273 static int dca_filter_channels(DCAContext *s, int block_index)
1275 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1278 /* 32 subbands QMF */
1279 for (k = 0; k < s->prim_channels; k++) {
1280 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1281 0, 8388608.0, 8388608.0 };*/
1282 if (s->channel_order_tab[k] >= 0)
1283 qmf_32_subbands(s, k, subband_samples[k],
1284 s->samples_chanptr[s->channel_order_tab[k]],
1285 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1289 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1290 dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
1293 /* Generate LFE samples for this subsubframe FIXME!!! */
1294 if (s->output & DCA_LFE) {
1295 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1296 s->lfe_data + 2 * s->lfe * (block_index + 4),
1297 s->samples_chanptr[dca_lfe_index[s->amode]],
1298 1.0 / (256.0 * 32768.0));
1299 /* Outputs 20bits pcm samples */
1306 static int dca_subframe_footer(DCAContext *s, int base_channel)
1308 int aux_data_count = 0, i;
1311 * Unpack optional information
1314 /* presumably optional information only appears in the core? */
1315 if (!base_channel) {
1317 skip_bits_long(&s->gb, 32);
1320 aux_data_count = get_bits(&s->gb, 6);
1322 for (i = 0; i < aux_data_count; i++)
1323 get_bits(&s->gb, 8);
1325 if (s->crc_present && (s->downmix || s->dynrange))
1326 get_bits(&s->gb, 16);
1333 * Decode a dca frame block
1335 * @param s pointer to the DCAContext
1338 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1343 if (s->current_subframe >= s->subframes) {
1344 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1345 s->current_subframe, s->subframes);
1346 return AVERROR_INVALIDDATA;
1349 if (!s->current_subsubframe) {
1351 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1353 /* Read subframe header */
1354 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1358 /* Read subsubframe */
1360 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1362 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1366 s->current_subsubframe++;
1367 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1368 s->current_subsubframe = 0;
1369 s->current_subframe++;
1371 if (s->current_subframe >= s->subframes) {
1373 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1375 /* Read subframe footer */
1376 if ((ret = dca_subframe_footer(s, base_channel)))
1384 * Return the number of channels in an ExSS speaker mask (HD)
1386 static int dca_exss_mask2count(int mask)
1388 /* count bits that mean speaker pairs twice */
1389 return av_popcount(mask) +
1390 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1391 DCA_EXSS_FRONT_LEFT_RIGHT |
1392 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1393 DCA_EXSS_WIDE_LEFT_RIGHT |
1394 DCA_EXSS_SIDE_LEFT_RIGHT |
1395 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1396 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1397 DCA_EXSS_REAR_LEFT_RIGHT |
1398 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1402 * Skip mixing coefficients of a single mix out configuration (HD)
1404 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1408 for (i = 0; i < channels; i++) {
1409 int mix_map_mask = get_bits(gb, out_ch);
1410 int num_coeffs = av_popcount(mix_map_mask);
1411 skip_bits_long(gb, num_coeffs * 6);
1416 * Parse extension substream asset header (HD)
1418 static int dca_exss_parse_asset_header(DCAContext *s)
1420 int header_pos = get_bits_count(&s->gb);
1423 int embedded_stereo = 0;
1424 int embedded_6ch = 0;
1425 int drc_code_present;
1426 int extensions_mask;
1429 if (get_bits_left(&s->gb) < 16)
1432 /* We will parse just enough to get to the extensions bitmask with which
1433 * we can set the profile value. */
1435 header_size = get_bits(&s->gb, 9) + 1;
1436 skip_bits(&s->gb, 3); // asset index
1438 if (s->static_fields) {
1439 if (get_bits1(&s->gb))
1440 skip_bits(&s->gb, 4); // asset type descriptor
1441 if (get_bits1(&s->gb))
1442 skip_bits_long(&s->gb, 24); // language descriptor
1444 if (get_bits1(&s->gb)) {
1445 /* How can one fit 1024 bytes of text here if the maximum value
1446 * for the asset header size field above was 512 bytes? */
1447 int text_length = get_bits(&s->gb, 10) + 1;
1448 if (get_bits_left(&s->gb) < text_length * 8)
1450 skip_bits_long(&s->gb, text_length * 8); // info text
1453 skip_bits(&s->gb, 5); // bit resolution - 1
1454 skip_bits(&s->gb, 4); // max sample rate code
1455 channels = get_bits(&s->gb, 8) + 1;
1457 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1458 int spkr_remap_sets;
1459 int spkr_mask_size = 16;
1463 embedded_stereo = get_bits1(&s->gb);
1465 embedded_6ch = get_bits1(&s->gb);
1467 if (get_bits1(&s->gb)) {
1468 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1469 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1472 spkr_remap_sets = get_bits(&s->gb, 3);
1474 for (i = 0; i < spkr_remap_sets; i++) {
1475 /* std layout mask for each remap set */
1476 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1479 for (i = 0; i < spkr_remap_sets; i++) {
1480 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1481 if (get_bits_left(&s->gb) < 0)
1484 for (j = 0; j < num_spkrs[i]; j++) {
1485 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1486 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1487 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1492 skip_bits(&s->gb, 3); // representation type
1496 drc_code_present = get_bits1(&s->gb);
1497 if (drc_code_present)
1498 get_bits(&s->gb, 8); // drc code
1500 if (get_bits1(&s->gb))
1501 skip_bits(&s->gb, 5); // dialog normalization code
1503 if (drc_code_present && embedded_stereo)
1504 get_bits(&s->gb, 8); // drc stereo code
1506 if (s->mix_metadata && get_bits1(&s->gb)) {
1507 skip_bits(&s->gb, 1); // external mix
1508 skip_bits(&s->gb, 6); // post mix gain code
1510 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1511 skip_bits(&s->gb, 3); // drc limit
1513 skip_bits(&s->gb, 8); // custom drc code
1515 if (get_bits1(&s->gb)) // channel specific scaling
1516 for (i = 0; i < s->num_mix_configs; i++)
1517 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1519 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1521 for (i = 0; i < s->num_mix_configs; i++) {
1522 if (get_bits_left(&s->gb) < 0)
1524 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1526 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1527 if (embedded_stereo)
1528 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1532 switch (get_bits(&s->gb, 2)) {
1533 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1534 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1535 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1536 case 3: extensions_mask = 0; /* aux coding */ break;
1539 /* not parsed further, we were only interested in the extensions mask */
1541 if (get_bits_left(&s->gb) < 0)
1544 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1545 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1548 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1550 if (extensions_mask & DCA_EXT_EXSS_XLL)
1551 s->profile = FF_PROFILE_DTS_HD_MA;
1552 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1554 s->profile = FF_PROFILE_DTS_HD_HRA;
1556 if (!(extensions_mask & DCA_EXT_CORE))
1557 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1558 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1559 av_log(s->avctx, AV_LOG_WARNING,
1560 "DTS extensions detection mismatch (%d, %d)\n",
1561 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1567 * Parse extension substream header (HD)
1569 static void dca_exss_parse_header(DCAContext *s)
1575 int active_ss_mask[8];
1578 if (get_bits_left(&s->gb) < 52)
1581 skip_bits(&s->gb, 8); // user data
1582 ss_index = get_bits(&s->gb, 2);
1584 blownup = get_bits1(&s->gb);
1585 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1586 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1588 s->static_fields = get_bits1(&s->gb);
1589 if (s->static_fields) {
1590 skip_bits(&s->gb, 2); // reference clock code
1591 skip_bits(&s->gb, 3); // frame duration code
1593 if (get_bits1(&s->gb))
1594 skip_bits_long(&s->gb, 36); // timestamp
1596 /* a single stream can contain multiple audio assets that can be
1597 * combined to form multiple audio presentations */
1599 num_audiop = get_bits(&s->gb, 3) + 1;
1600 if (num_audiop > 1) {
1601 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1602 /* ignore such streams for now */
1606 num_assets = get_bits(&s->gb, 3) + 1;
1607 if (num_assets > 1) {
1608 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1609 /* ignore such streams for now */
1613 for (i = 0; i < num_audiop; i++)
1614 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1616 for (i = 0; i < num_audiop; i++)
1617 for (j = 0; j <= ss_index; j++)
1618 if (active_ss_mask[i] & (1 << j))
1619 skip_bits(&s->gb, 8); // active asset mask
1621 s->mix_metadata = get_bits1(&s->gb);
1622 if (s->mix_metadata) {
1623 int mix_out_mask_size;
1625 skip_bits(&s->gb, 2); // adjustment level
1626 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1627 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1629 for (i = 0; i < s->num_mix_configs; i++) {
1630 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1631 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1636 for (i = 0; i < num_assets; i++)
1637 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1639 for (i = 0; i < num_assets; i++) {
1640 if (dca_exss_parse_asset_header(s))
1644 /* not parsed further, we were only interested in the extensions mask
1645 * from the asset header */
1649 * Main frame decoding function
1650 * FIXME add arguments
1652 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1653 int *got_frame_ptr, AVPacket *avpkt)
1655 AVFrame *frame = data;
1656 const uint8_t *buf = avpkt->data;
1657 int buf_size = avpkt->size;
1660 int num_core_channels = 0;
1662 float **samples_flt;
1663 DCAContext *s = avctx->priv_data;
1664 int channels, full_channels;
1670 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1671 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1672 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1673 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1674 return AVERROR_INVALIDDATA;
1677 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1678 if ((ret = dca_parse_frame_header(s)) < 0) {
1679 //seems like the frame is corrupt, try with the next one
1682 //set AVCodec values with parsed data
1683 avctx->sample_rate = s->sample_rate;
1684 avctx->bit_rate = s->bit_rate;
1686 s->profile = FF_PROFILE_DTS;
1688 for (i = 0; i < (s->sample_blocks / 8); i++) {
1689 if ((ret = dca_decode_block(s, 0, i))) {
1690 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1695 /* record number of core channels incase less than max channels are requested */
1696 num_core_channels = s->prim_channels;
1699 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1701 s->core_ext_mask = 0;
1703 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1705 /* only scan for extensions if ext_descr was unknown or indicated a
1706 * supported XCh extension */
1707 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1709 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1710 * extensions scan can fill it up */
1711 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1713 /* extensions start at 32-bit boundaries into bitstream */
1714 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1716 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1717 uint32_t bits = get_bits_long(&s->gb, 32);
1721 int ext_amode, xch_fsize;
1723 s->xch_base_channel = s->prim_channels;
1725 /* validate sync word using XCHFSIZE field */
1726 xch_fsize = show_bits(&s->gb, 10);
1727 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1728 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1731 /* skip length-to-end-of-frame field for the moment */
1732 skip_bits(&s->gb, 10);
1734 s->core_ext_mask |= DCA_EXT_XCH;
1736 /* extension amode(number of channels in extension) should be 1 */
1737 /* AFAIK XCh is not used for more channels */
1738 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1739 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1740 " supported!\n", ext_amode);
1744 /* much like core primary audio coding header */
1745 dca_parse_audio_coding_header(s, s->xch_base_channel);
1747 for (i = 0; i < (s->sample_blocks / 8); i++)
1748 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1749 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1757 /* XXCh: extended channels */
1758 /* usually found either in core or HD part in DTS-HD HRA streams,
1759 * but not in DTS-ES which contains XCh extensions instead */
1760 s->core_ext_mask |= DCA_EXT_XXCH;
1764 int fsize96 = show_bits(&s->gb, 12) + 1;
1765 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1768 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1769 get_bits_count(&s->gb));
1770 skip_bits(&s->gb, 12);
1771 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1772 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1774 s->core_ext_mask |= DCA_EXT_X96;
1779 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1782 /* no supported extensions, skip the rest of the core substream */
1783 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1786 if (s->core_ext_mask & DCA_EXT_X96)
1787 s->profile = FF_PROFILE_DTS_96_24;
1788 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1789 s->profile = FF_PROFILE_DTS_ES;
1791 /* check for ExSS (HD part) */
1792 if (s->dca_buffer_size - s->frame_size > 32 &&
1793 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1794 dca_exss_parse_header(s);
1796 avctx->profile = s->profile;
1798 full_channels = channels = s->prim_channels + !!s->lfe;
1800 if (s->amode < 16) {
1801 avctx->channel_layout = dca_core_channel_layout[s->amode];
1803 if (s->xch_present && (!avctx->request_channels ||
1804 avctx->request_channels > num_core_channels + !!s->lfe)) {
1805 avctx->channel_layout |= AV_CH_BACK_CENTER;
1807 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1808 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1810 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1813 channels = num_core_channels + !!s->lfe;
1814 s->xch_present = 0; /* disable further xch processing */
1816 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1817 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1819 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1822 if (channels > !!s->lfe &&
1823 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1824 return AVERROR_INVALIDDATA;
1826 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1828 s->output = DCA_STEREO;
1829 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1832 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1833 return AVERROR_INVALIDDATA;
1835 avctx->channels = channels;
1837 /* get output buffer */
1838 frame->nb_samples = 256 * (s->sample_blocks / 8);
1839 if ((ret = ff_get_buffer(avctx, frame)) < 0) {
1840 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1843 samples_flt = (float **)frame->extended_data;
1845 /* allocate buffer for extra channels if downmixing */
1846 if (avctx->channels < full_channels) {
1847 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
1849 avctx->sample_fmt, 0);
1853 av_fast_malloc(&s->extra_channels_buffer,
1854 &s->extra_channels_buffer_size, ret);
1855 if (!s->extra_channels_buffer)
1856 return AVERROR(ENOMEM);
1858 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
1859 s->extra_channels_buffer,
1860 full_channels - channels,
1861 frame->nb_samples, avctx->sample_fmt, 0);
1866 /* filter to get final output */
1867 for (i = 0; i < (s->sample_blocks / 8); i++) {
1870 for (ch = 0; ch < channels; ch++)
1871 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
1872 for (; ch < full_channels; ch++)
1873 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
1875 dca_filter_channels(s, i);
1877 /* If this was marked as a DTS-ES stream we need to subtract back- */
1878 /* channel from SL & SR to remove matrixed back-channel signal */
1879 if ((s->source_pcm_res & 1) && s->xch_present) {
1880 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
1881 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
1882 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
1883 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1884 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1888 /* update lfe history */
1889 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1890 for (i = 0; i < 2 * s->lfe * 4; i++)
1891 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1901 * DCA initialization
1903 * @param avctx pointer to the AVCodecContext
1906 static av_cold int dca_decode_init(AVCodecContext *avctx)
1908 DCAContext *s = avctx->priv_data;
1913 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
1914 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1915 ff_synth_filter_init(&s->synth);
1916 ff_dcadsp_init(&s->dcadsp);
1917 ff_fmt_convert_init(&s->fmt_conv, avctx);
1919 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
1921 /* allow downmixing to stereo */
1922 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1923 avctx->request_channels == 2) {
1924 avctx->channels = avctx->request_channels;
1930 static av_cold int dca_decode_end(AVCodecContext *avctx)
1932 DCAContext *s = avctx->priv_data;
1933 ff_mdct_end(&s->imdct);
1934 av_freep(&s->extra_channels_buffer);
1938 static const AVProfile profiles[] = {
1939 { FF_PROFILE_DTS, "DTS" },
1940 { FF_PROFILE_DTS_ES, "DTS-ES" },
1941 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1942 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1943 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1944 { FF_PROFILE_UNKNOWN },
1947 AVCodec ff_dca_decoder = {
1949 .type = AVMEDIA_TYPE_AUDIO,
1950 .id = AV_CODEC_ID_DTS,
1951 .priv_data_size = sizeof(DCAContext),
1952 .init = dca_decode_init,
1953 .decode = dca_decode_frame,
1954 .close = dca_decode_end,
1955 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1956 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1957 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
1958 AV_SAMPLE_FMT_NONE },
1959 .profiles = NULL_IF_CONFIG_SMALL(profiles),