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/common.h"
30 #include "libavutil/intmath.h"
31 #include "libavutil/intreadwrite.h"
32 #include "libavutil/audioconvert.h"
41 #include "synth_filter.h"
43 #include "fmtconvert.h"
47 #define DCA_PRIM_CHANNELS_MAX (7)
48 #define DCA_SUBBANDS (32)
49 #define DCA_ABITS_MAX (32) /* Should be 28 */
50 #define DCA_SUBSUBFRAMES_MAX (4)
51 #define DCA_SUBFRAMES_MAX (16)
52 #define DCA_BLOCKS_MAX (16)
53 #define DCA_LFE_MAX (3)
69 /* these are unconfirmed but should be mostly correct */
70 enum DCAExSSSpeakerMask {
71 DCA_EXSS_FRONT_CENTER = 0x0001,
72 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
73 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
74 DCA_EXSS_LFE = 0x0008,
75 DCA_EXSS_REAR_CENTER = 0x0010,
76 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
77 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
78 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
79 DCA_EXSS_OVERHEAD = 0x0100,
80 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
81 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
82 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
83 DCA_EXSS_LFE2 = 0x1000,
84 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
85 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
86 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
89 enum DCAExtensionMask {
90 DCA_EXT_CORE = 0x001, ///< core in core substream
91 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
92 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
93 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
94 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
95 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
96 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
97 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
98 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
99 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
102 /* -1 are reserved or unknown */
103 static const int dca_ext_audio_descr_mask[] = {
107 DCA_EXT_XCH | DCA_EXT_X96,
114 /* extensions that reside in core substream */
115 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
117 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
118 * Some compromises have been made for special configurations. Most configurations
119 * are never used so complete accuracy is not needed.
121 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
122 * S -> side, when both rear and back are configured move one of them to the side channel
124 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
127 static const int64_t dca_core_channel_layout[] = {
128 AV_CH_FRONT_CENTER, ///< 1, A
129 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
130 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
131 AV_CH_LAYOUT_STEREO, ///< 2, (L+R) + (L-R) (sum-difference)
132 AV_CH_LAYOUT_STEREO, ///< 2, LT +RT (left and right total)
133 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER, ///< 3, C+L+R
134 AV_CH_LAYOUT_STEREO|AV_CH_BACK_CENTER, ///< 3, L+R+S
135 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_BACK_CENTER, ///< 4, C + L + R+ S
136 AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 4, L + R +SL+ SR
137 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 5, C + L + R+ SL+SR
138 AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
139 AV_CH_LAYOUT_STEREO|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT|AV_CH_FRONT_CENTER|AV_CH_BACK_CENTER, ///< 6, C + L + R+ LR + RR + OV
140 AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_BACK_CENTER|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT, ///< 6, CF+ CR+LF+ RF+LR + RR
141 AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
142 AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2+ SR1 + SR2
143 AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_BACK_CENTER|AV_CH_SIDE_RIGHT, ///< 8, CL + C+ CR + L + R + SL + S+ SR
146 static const int8_t dca_lfe_index[] = {
147 1,2,2,2,2,3,2,3,2,3,2,3,1,3,2,3
150 static const int8_t dca_channel_reorder_lfe[][9] = {
151 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
152 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
153 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
154 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
155 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
156 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
157 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
158 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
159 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
160 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
161 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
162 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
163 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
164 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
165 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
166 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
169 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
170 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
171 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
172 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
173 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
174 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
175 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
176 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
177 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
178 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
179 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
180 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
181 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
182 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
183 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
184 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
185 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
188 static const int8_t dca_channel_reorder_nolfe[][9] = {
189 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
190 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
191 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
192 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
193 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
194 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
195 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
196 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
197 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
198 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
199 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
200 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
201 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
202 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
203 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
204 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
207 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
208 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
209 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
210 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
211 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
212 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
213 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
214 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
215 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
216 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
217 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
218 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
219 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
220 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
221 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
222 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
223 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
226 #define DCA_DOLBY 101 /* FIXME */
228 #define DCA_CHANNEL_BITS 6
229 #define DCA_CHANNEL_MASK 0x3F
233 #define HEADER_SIZE 14
235 #define DCA_MAX_FRAME_SIZE 16384
236 #define DCA_MAX_EXSS_HEADER_SIZE 4096
238 #define DCA_BUFFER_PADDING_SIZE 1024
240 /** Bit allocation */
242 int offset; ///< code values offset
243 int maxbits[8]; ///< max bits in VLC
244 int wrap; ///< wrap for get_vlc2()
245 VLC vlc[8]; ///< actual codes
248 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
249 static BitAlloc dca_tmode; ///< transition mode VLCs
250 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
251 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
253 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx)
255 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) + ba->offset;
259 AVCodecContext *avctx;
261 int frame_type; ///< type of the current frame
262 int samples_deficit; ///< deficit sample count
263 int crc_present; ///< crc is present in the bitstream
264 int sample_blocks; ///< number of PCM sample blocks
265 int frame_size; ///< primary frame byte size
266 int amode; ///< audio channels arrangement
267 int sample_rate; ///< audio sampling rate
268 int bit_rate; ///< transmission bit rate
269 int bit_rate_index; ///< transmission bit rate index
271 int downmix; ///< embedded downmix enabled
272 int dynrange; ///< embedded dynamic range flag
273 int timestamp; ///< embedded time stamp flag
274 int aux_data; ///< auxiliary data flag
275 int hdcd; ///< source material is mastered in HDCD
276 int ext_descr; ///< extension audio descriptor flag
277 int ext_coding; ///< extended coding flag
278 int aspf; ///< audio sync word insertion flag
279 int lfe; ///< low frequency effects flag
280 int predictor_history; ///< predictor history flag
281 int header_crc; ///< header crc check bytes
282 int multirate_inter; ///< multirate interpolator switch
283 int version; ///< encoder software revision
284 int copy_history; ///< copy history
285 int source_pcm_res; ///< source pcm resolution
286 int front_sum; ///< front sum/difference flag
287 int surround_sum; ///< surround sum/difference flag
288 int dialog_norm; ///< dialog normalisation parameter
290 /* Primary audio coding header */
291 int subframes; ///< number of subframes
292 int is_channels_set; ///< check for if the channel number is already set
293 int total_channels; ///< number of channels including extensions
294 int prim_channels; ///< number of primary audio channels
295 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
296 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
297 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
298 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
299 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
300 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
301 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
302 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
304 /* Primary audio coding side information */
305 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
306 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
307 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
308 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
309 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
310 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
311 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
312 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
313 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
314 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
315 int dynrange_coef; ///< dynamic range coefficient
317 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
319 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
320 int lfe_scale_factor;
322 /* Subband samples history (for ADPCM) */
323 float subband_samples_hist[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
324 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
325 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
326 int hist_index[DCA_PRIM_CHANNELS_MAX];
327 DECLARE_ALIGNED(32, float, raXin)[32];
329 int output; ///< type of output
330 float scale_bias; ///< output scale
332 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
333 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
334 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX+1];
336 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
337 int dca_buffer_size; ///< how much data is in the dca_buffer
339 const int8_t* channel_order_tab; ///< channel reordering table, lfe and non lfe
341 /* Current position in DCA frame */
342 int current_subframe;
343 int current_subsubframe;
345 int core_ext_mask; ///< present extensions in the core substream
347 /* XCh extension information */
348 int xch_present; ///< XCh extension present and valid
349 int xch_base_channel; ///< index of first (only) channel containing XCH data
351 /* ExSS header parser */
352 int static_fields; ///< static fields present
353 int mix_metadata; ///< mixing metadata present
354 int num_mix_configs; ///< number of mix out configurations
355 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
359 int debug_flag; ///< used for suppressing repeated error messages output
362 SynthFilterContext synth;
363 DCADSPContext dcadsp;
364 FmtConvertContext fmt_conv;
367 static const uint16_t dca_vlc_offs[] = {
368 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
369 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
370 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
371 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
372 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
373 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
376 static av_cold void dca_init_vlcs(void)
378 static int vlcs_initialized = 0;
380 static VLC_TYPE dca_table[23622][2];
382 if (vlcs_initialized)
385 dca_bitalloc_index.offset = 1;
386 dca_bitalloc_index.wrap = 2;
387 for (i = 0; i < 5; i++) {
388 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
389 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
390 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
391 bitalloc_12_bits[i], 1, 1,
392 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
394 dca_scalefactor.offset = -64;
395 dca_scalefactor.wrap = 2;
396 for (i = 0; i < 5; i++) {
397 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
398 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
399 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
400 scales_bits[i], 1, 1,
401 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
403 dca_tmode.offset = 0;
405 for (i = 0; i < 4; i++) {
406 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
407 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
408 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
410 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
413 for (i = 0; i < 10; i++)
414 for (j = 0; j < 7; j++){
415 if (!bitalloc_codes[i][j]) break;
416 dca_smpl_bitalloc[i+1].offset = bitalloc_offsets[i];
417 dca_smpl_bitalloc[i+1].wrap = 1 + (j > 4);
418 dca_smpl_bitalloc[i+1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
419 dca_smpl_bitalloc[i+1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
420 init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j],
422 bitalloc_bits[i][j], 1, 1,
423 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
426 vlcs_initialized = 1;
429 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
432 *dst++ = get_bits(gb, bits);
435 static int dca_parse_audio_coding_header(DCAContext * s, int base_channel)
438 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
439 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
440 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
442 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
443 s->prim_channels = s->total_channels;
445 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
446 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
449 for (i = base_channel; i < s->prim_channels; i++) {
450 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
451 if (s->subband_activity[i] > DCA_SUBBANDS)
452 s->subband_activity[i] = DCA_SUBBANDS;
454 for (i = base_channel; i < s->prim_channels; i++) {
455 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
456 if (s->vq_start_subband[i] > DCA_SUBBANDS)
457 s->vq_start_subband[i] = DCA_SUBBANDS;
459 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
460 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
461 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
462 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
464 /* Get codebooks quantization indexes */
466 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
467 for (j = 1; j < 11; j++)
468 for (i = base_channel; i < s->prim_channels; i++)
469 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
471 /* Get scale factor adjustment */
472 for (j = 0; j < 11; j++)
473 for (i = base_channel; i < s->prim_channels; i++)
474 s->scalefactor_adj[i][j] = 1;
476 for (j = 1; j < 11; j++)
477 for (i = base_channel; i < s->prim_channels; i++)
478 if (s->quant_index_huffman[i][j] < thr[j])
479 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
481 if (s->crc_present) {
482 /* Audio header CRC check */
483 get_bits(&s->gb, 16);
486 s->current_subframe = 0;
487 s->current_subsubframe = 0;
490 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
491 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
492 for (i = base_channel; i < s->prim_channels; i++){
493 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", s->subband_activity[i]);
494 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", s->vq_start_subband[i]);
495 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", s->joint_intensity[i]);
496 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", s->transient_huffman[i]);
497 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", s->scalefactor_huffman[i]);
498 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", s->bitalloc_huffman[i]);
499 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
500 for (j = 0; j < 11; j++)
501 av_log(s->avctx, AV_LOG_DEBUG, " %i",
502 s->quant_index_huffman[i][j]);
503 av_log(s->avctx, AV_LOG_DEBUG, "\n");
504 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
505 for (j = 0; j < 11; j++)
506 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
507 av_log(s->avctx, AV_LOG_DEBUG, "\n");
514 static int dca_parse_frame_header(DCAContext * s)
516 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
519 get_bits(&s->gb, 32);
522 s->frame_type = get_bits(&s->gb, 1);
523 s->samples_deficit = get_bits(&s->gb, 5) + 1;
524 s->crc_present = get_bits(&s->gb, 1);
525 s->sample_blocks = get_bits(&s->gb, 7) + 1;
526 s->frame_size = get_bits(&s->gb, 14) + 1;
527 if (s->frame_size < 95)
529 s->amode = get_bits(&s->gb, 6);
530 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
533 s->bit_rate_index = get_bits(&s->gb, 5);
534 s->bit_rate = dca_bit_rates[s->bit_rate_index];
538 s->downmix = get_bits(&s->gb, 1);
539 s->dynrange = get_bits(&s->gb, 1);
540 s->timestamp = get_bits(&s->gb, 1);
541 s->aux_data = get_bits(&s->gb, 1);
542 s->hdcd = get_bits(&s->gb, 1);
543 s->ext_descr = get_bits(&s->gb, 3);
544 s->ext_coding = get_bits(&s->gb, 1);
545 s->aspf = get_bits(&s->gb, 1);
546 s->lfe = get_bits(&s->gb, 2);
547 s->predictor_history = get_bits(&s->gb, 1);
549 /* TODO: check CRC */
551 s->header_crc = get_bits(&s->gb, 16);
553 s->multirate_inter = get_bits(&s->gb, 1);
554 s->version = get_bits(&s->gb, 4);
555 s->copy_history = get_bits(&s->gb, 2);
556 s->source_pcm_res = get_bits(&s->gb, 3);
557 s->front_sum = get_bits(&s->gb, 1);
558 s->surround_sum = get_bits(&s->gb, 1);
559 s->dialog_norm = get_bits(&s->gb, 4);
561 /* FIXME: channels mixing levels */
562 s->output = s->amode;
563 if (s->lfe) s->output |= DCA_LFE;
566 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
567 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
568 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
569 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
570 s->sample_blocks, s->sample_blocks * 32);
571 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
572 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
573 s->amode, dca_channels[s->amode]);
574 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
576 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
578 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
579 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
580 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
581 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
582 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
583 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
584 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
585 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
586 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
587 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
588 s->predictor_history);
589 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
590 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
592 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
593 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
594 av_log(s->avctx, AV_LOG_DEBUG,
595 "source pcm resolution: %i (%i bits/sample)\n",
596 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
597 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
598 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
599 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
600 av_log(s->avctx, AV_LOG_DEBUG, "\n");
603 /* Primary audio coding header */
604 s->subframes = get_bits(&s->gb, 4) + 1;
606 return dca_parse_audio_coding_header(s, 0);
610 static inline int get_scale(GetBitContext *gb, int level, int value)
613 /* huffman encoded */
614 value += get_bitalloc(gb, &dca_scalefactor, level);
615 } else if (level < 8)
616 value = get_bits(gb, level + 1);
620 static int dca_subframe_header(DCAContext * s, int base_channel, int block_index)
622 /* Primary audio coding side information */
625 if (get_bits_left(&s->gb) < 0)
629 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
630 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
633 for (j = base_channel; j < s->prim_channels; j++) {
634 for (k = 0; k < s->subband_activity[j]; k++)
635 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
638 /* Get prediction codebook */
639 for (j = base_channel; j < s->prim_channels; j++) {
640 for (k = 0; k < s->subband_activity[j]; k++) {
641 if (s->prediction_mode[j][k] > 0) {
642 /* (Prediction coefficient VQ address) */
643 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
648 /* Bit allocation index */
649 for (j = base_channel; j < s->prim_channels; j++) {
650 for (k = 0; k < s->vq_start_subband[j]; k++) {
651 if (s->bitalloc_huffman[j] == 6)
652 s->bitalloc[j][k] = get_bits(&s->gb, 5);
653 else if (s->bitalloc_huffman[j] == 5)
654 s->bitalloc[j][k] = get_bits(&s->gb, 4);
655 else if (s->bitalloc_huffman[j] == 7) {
656 av_log(s->avctx, AV_LOG_ERROR,
657 "Invalid bit allocation index\n");
661 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
664 if (s->bitalloc[j][k] > 26) {
665 // av_log(s->avctx,AV_LOG_DEBUG,"bitalloc index [%i][%i] too big (%i)\n",
666 // j, k, s->bitalloc[j][k]);
672 /* Transition mode */
673 for (j = base_channel; j < s->prim_channels; j++) {
674 for (k = 0; k < s->subband_activity[j]; k++) {
675 s->transition_mode[j][k] = 0;
676 if (s->subsubframes[s->current_subframe] > 1 &&
677 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
678 s->transition_mode[j][k] =
679 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
684 if (get_bits_left(&s->gb) < 0)
687 for (j = base_channel; j < s->prim_channels; j++) {
688 const uint32_t *scale_table;
691 memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
693 if (s->scalefactor_huffman[j] == 6)
694 scale_table = scale_factor_quant7;
696 scale_table = scale_factor_quant6;
698 /* When huffman coded, only the difference is encoded */
701 for (k = 0; k < s->subband_activity[j]; k++) {
702 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
703 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
704 s->scale_factor[j][k][0] = scale_table[scale_sum];
707 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
708 /* Get second scale factor */
709 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
710 s->scale_factor[j][k][1] = scale_table[scale_sum];
715 /* Joint subband scale factor codebook select */
716 for (j = base_channel; j < s->prim_channels; j++) {
717 /* Transmitted only if joint subband coding enabled */
718 if (s->joint_intensity[j] > 0)
719 s->joint_huff[j] = get_bits(&s->gb, 3);
722 if (get_bits_left(&s->gb) < 0)
725 /* Scale factors for joint subband coding */
726 for (j = base_channel; j < s->prim_channels; j++) {
729 /* Transmitted only if joint subband coding enabled */
730 if (s->joint_intensity[j] > 0) {
732 source_channel = s->joint_intensity[j] - 1;
734 /* When huffman coded, only the difference is encoded
735 * (is this valid as well for joint scales ???) */
737 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
738 scale = get_scale(&s->gb, s->joint_huff[j], 0);
739 scale += 64; /* bias */
740 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
743 if (!(s->debug_flag & 0x02)) {
744 av_log(s->avctx, AV_LOG_DEBUG,
745 "Joint stereo coding not supported\n");
746 s->debug_flag |= 0x02;
751 /* Stereo downmix coefficients */
752 if (!base_channel && s->prim_channels > 2) {
754 for (j = base_channel; j < s->prim_channels; j++) {
755 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
756 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
759 int am = s->amode & DCA_CHANNEL_MASK;
760 for (j = base_channel; j < s->prim_channels; j++) {
761 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
762 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
767 /* Dynamic range coefficient */
768 if (!base_channel && s->dynrange)
769 s->dynrange_coef = get_bits(&s->gb, 8);
771 /* Side information CRC check word */
772 if (s->crc_present) {
773 get_bits(&s->gb, 16);
777 * Primary audio data arrays
780 /* VQ encoded high frequency subbands */
781 for (j = base_channel; j < s->prim_channels; j++)
782 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
783 /* 1 vector -> 32 samples */
784 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
786 /* Low frequency effect data */
787 if (!base_channel && s->lfe) {
789 int lfe_samples = 2 * s->lfe * (4 + block_index);
790 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
793 for (j = lfe_samples; j < lfe_end_sample; j++) {
794 /* Signed 8 bits int */
795 s->lfe_data[j] = get_sbits(&s->gb, 8);
798 /* Scale factor index */
799 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
801 /* Quantization step size * scale factor */
802 lfe_scale = 0.035 * s->lfe_scale_factor;
804 for (j = lfe_samples; j < lfe_end_sample; j++)
805 s->lfe_data[j] *= lfe_scale;
809 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", s->subsubframes[s->current_subframe]);
810 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
811 s->partial_samples[s->current_subframe]);
812 for (j = base_channel; j < s->prim_channels; j++) {
813 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
814 for (k = 0; k < s->subband_activity[j]; k++)
815 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
816 av_log(s->avctx, AV_LOG_DEBUG, "\n");
818 for (j = base_channel; j < s->prim_channels; j++) {
819 for (k = 0; k < s->subband_activity[j]; k++)
820 av_log(s->avctx, AV_LOG_DEBUG,
821 "prediction coefs: %f, %f, %f, %f\n",
822 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
823 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
824 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
825 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
827 for (j = base_channel; j < s->prim_channels; j++) {
828 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
829 for (k = 0; k < s->vq_start_subband[j]; k++)
830 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
831 av_log(s->avctx, AV_LOG_DEBUG, "\n");
833 for (j = base_channel; j < s->prim_channels; j++) {
834 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
835 for (k = 0; k < s->subband_activity[j]; k++)
836 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
837 av_log(s->avctx, AV_LOG_DEBUG, "\n");
839 for (j = base_channel; j < s->prim_channels; j++) {
840 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
841 for (k = 0; k < s->subband_activity[j]; k++) {
842 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
843 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
844 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
845 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
847 av_log(s->avctx, AV_LOG_DEBUG, "\n");
849 for (j = base_channel; j < s->prim_channels; j++) {
850 if (s->joint_intensity[j] > 0) {
851 int source_channel = s->joint_intensity[j] - 1;
852 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
853 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
854 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
855 av_log(s->avctx, AV_LOG_DEBUG, "\n");
858 if (!base_channel && s->prim_channels > 2 && s->downmix) {
859 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
860 for (j = 0; j < s->prim_channels; j++) {
861 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][0]]);
862 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][1]]);
864 av_log(s->avctx, AV_LOG_DEBUG, "\n");
866 for (j = base_channel; j < s->prim_channels; j++)
867 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
868 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
869 if (!base_channel && s->lfe) {
870 int lfe_samples = 2 * s->lfe * (4 + block_index);
871 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
873 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
874 for (j = lfe_samples; j < lfe_end_sample; j++)
875 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
876 av_log(s->avctx, AV_LOG_DEBUG, "\n");
883 static void qmf_32_subbands(DCAContext * s, int chans,
884 float samples_in[32][8], float *samples_out,
887 const float *prCoeff;
890 int sb_act = s->subband_activity[chans];
893 scale *= sqrt(1/8.0);
896 if (!s->multirate_inter) /* Non-perfect reconstruction */
897 prCoeff = fir_32bands_nonperfect;
898 else /* Perfect reconstruction */
899 prCoeff = fir_32bands_perfect;
901 for (i = sb_act; i < 32; i++)
904 /* Reconstructed channel sample index */
905 for (subindex = 0; subindex < 8; subindex++) {
906 /* Load in one sample from each subband and clear inactive subbands */
907 for (i = 0; i < sb_act; i++){
908 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ ((i-1)&2)<<30;
909 AV_WN32A(&s->raXin[i], v);
912 s->synth.synth_filter_float(&s->imdct,
913 s->subband_fir_hist[chans], &s->hist_index[chans],
914 s->subband_fir_noidea[chans], prCoeff,
915 samples_out, s->raXin, scale);
921 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
922 int num_deci_sample, float *samples_in,
923 float *samples_out, float scale)
925 /* samples_in: An array holding decimated samples.
926 * Samples in current subframe starts from samples_in[0],
927 * while samples_in[-1], samples_in[-2], ..., stores samples
928 * from last subframe as history.
930 * samples_out: An array holding interpolated samples
934 const float *prCoeff;
937 /* Select decimation filter */
938 if (decimation_select == 1) {
940 prCoeff = lfe_fir_128;
943 prCoeff = lfe_fir_64;
946 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
947 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor,
950 samples_out += 2 * decifactor;
954 /* downmixing routines */
955 #define MIX_REAR1(samples, si1, rs, coef) \
956 samples[i] += samples[si1] * coef[rs][0]; \
957 samples[i+256] += samples[si1] * coef[rs][1];
959 #define MIX_REAR2(samples, si1, si2, rs, coef) \
960 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs+1][0]; \
961 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs+1][1];
963 #define MIX_FRONT3(samples, coef) \
967 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
968 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
970 #define DOWNMIX_TO_STEREO(op1, op2) \
971 for (i = 0; i < 256; i++){ \
976 static void dca_downmix(float *samples, int srcfmt,
977 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
978 const int8_t *channel_mapping)
983 float coef[DCA_PRIM_CHANNELS_MAX][2];
985 for (i=0; i<DCA_PRIM_CHANNELS_MAX; i++) {
986 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
987 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
993 case DCA_STEREO_TOTAL:
994 case DCA_STEREO_SUMDIFF:
996 av_log(NULL, 0, "Not implemented!\n");
1001 c = channel_mapping[0] * 256;
1002 l = channel_mapping[1] * 256;
1003 r = channel_mapping[2] * 256;
1004 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),);
1007 s = channel_mapping[2] * 256;
1008 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef),);
1011 c = channel_mapping[0] * 256;
1012 l = channel_mapping[1] * 256;
1013 r = channel_mapping[2] * 256;
1014 s = channel_mapping[3] * 256;
1015 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1016 MIX_REAR1(samples, i + s, 3, coef));
1019 sl = channel_mapping[2] * 256;
1020 sr = channel_mapping[3] * 256;
1021 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef),);
1024 c = channel_mapping[0] * 256;
1025 l = channel_mapping[1] * 256;
1026 r = channel_mapping[2] * 256;
1027 sl = channel_mapping[3] * 256;
1028 sr = channel_mapping[4] * 256;
1029 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1030 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1036 /* Very compact version of the block code decoder that does not use table
1037 * look-up but is slightly slower */
1038 static int decode_blockcode(int code, int levels, int *values)
1041 int offset = (levels - 1) >> 1;
1043 for (i = 0; i < 4; i++) {
1044 int div = FASTDIV(code, levels);
1045 values[i] = code - offset - div*levels;
1052 av_log(NULL, AV_LOG_ERROR, "ERROR: block code look-up failed\n");
1057 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1058 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1060 static int dca_subsubframe(DCAContext * s, int base_channel, int block_index)
1063 int subsubframe = s->current_subsubframe;
1065 const float *quant_step_table;
1068 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1069 LOCAL_ALIGNED_16(int, block, [8]);
1075 /* Select quantization step size table */
1076 if (s->bit_rate_index == 0x1f)
1077 quant_step_table = lossless_quant_d;
1079 quant_step_table = lossy_quant_d;
1081 for (k = base_channel; k < s->prim_channels; k++) {
1082 if (get_bits_left(&s->gb) < 0)
1085 for (l = 0; l < s->vq_start_subband[k]; l++) {
1088 /* Select the mid-tread linear quantizer */
1089 int abits = s->bitalloc[k][l];
1091 float quant_step_size = quant_step_table[abits];
1094 * Determine quantization index code book and its type
1097 /* Select quantization index code book */
1098 int sel = s->quant_index_huffman[k][abits];
1101 * Extract bits from the bit stream
1104 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1106 /* Deal with transients */
1107 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1108 float rscale = quant_step_size * s->scale_factor[k][l][sfi] * s->scalefactor_adj[k][sel];
1110 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){
1113 int block_code1, block_code2, size, levels;
1115 size = abits_sizes[abits-1];
1116 levels = abits_levels[abits-1];
1118 block_code1 = get_bits(&s->gb, size);
1119 /* FIXME Should test return value */
1120 decode_blockcode(block_code1, levels, block);
1121 block_code2 = get_bits(&s->gb, size);
1122 decode_blockcode(block_code2, levels, &block[4]);
1125 for (m = 0; m < 8; m++)
1126 block[m] = get_sbits(&s->gb, abits - 3);
1130 for (m = 0; m < 8; m++)
1131 block[m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel);
1134 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1139 * Inverse ADPCM if in prediction mode
1141 if (s->prediction_mode[k][l]) {
1143 for (m = 0; m < 8; m++) {
1144 for (n = 1; n <= 4; n++)
1146 subband_samples[k][l][m] +=
1147 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1148 subband_samples[k][l][m - n] / 8192);
1149 else if (s->predictor_history)
1150 subband_samples[k][l][m] +=
1151 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1152 s->subband_samples_hist[k][l][m - n +
1159 * Decode VQ encoded high frequencies
1161 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1162 /* 1 vector -> 32 samples but we only need the 8 samples
1163 * for this subsubframe. */
1166 if (!s->debug_flag & 0x01) {
1167 av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
1168 s->debug_flag |= 0x01;
1171 for (m = 0; m < 8; m++) {
1172 subband_samples[k][l][m] =
1173 high_freq_vq[s->high_freq_vq[k][l]][subsubframe * 8 +
1175 * (float) s->scale_factor[k][l][0] / 16.0;
1180 /* Check for DSYNC after subsubframe */
1181 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1182 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1184 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1187 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1191 /* Backup predictor history for adpcm */
1192 for (k = base_channel; k < s->prim_channels; k++)
1193 for (l = 0; l < s->vq_start_subband[k]; l++)
1194 memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
1195 4 * sizeof(subband_samples[0][0][0]));
1200 static int dca_filter_channels(DCAContext * s, int block_index)
1202 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1205 /* 32 subbands QMF */
1206 for (k = 0; k < s->prim_channels; k++) {
1207 /* static float pcm_to_double[8] =
1208 {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
1209 qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
1210 M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ );
1214 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1215 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1218 /* Generate LFE samples for this subsubframe FIXME!!! */
1219 if (s->output & DCA_LFE) {
1220 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1221 s->lfe_data + 2 * s->lfe * (block_index + 4),
1222 &s->samples[256 * dca_lfe_index[s->amode]],
1223 (1.0/256.0)*s->scale_bias);
1224 /* Outputs 20bits pcm samples */
1231 static int dca_subframe_footer(DCAContext * s, int base_channel)
1233 int aux_data_count = 0, i;
1236 * Unpack optional information
1239 /* presumably optional information only appears in the core? */
1240 if (!base_channel) {
1242 get_bits(&s->gb, 32);
1245 aux_data_count = get_bits(&s->gb, 6);
1247 for (i = 0; i < aux_data_count; i++)
1248 get_bits(&s->gb, 8);
1250 if (s->crc_present && (s->downmix || s->dynrange))
1251 get_bits(&s->gb, 16);
1258 * Decode a dca frame block
1260 * @param s pointer to the DCAContext
1263 static int dca_decode_block(DCAContext * s, int base_channel, int block_index)
1267 if (s->current_subframe >= s->subframes) {
1268 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1269 s->current_subframe, s->subframes);
1273 if (!s->current_subsubframe) {
1275 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1277 /* Read subframe header */
1278 if (dca_subframe_header(s, base_channel, block_index))
1282 /* Read subsubframe */
1284 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1286 if (dca_subsubframe(s, base_channel, block_index))
1290 s->current_subsubframe++;
1291 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1292 s->current_subsubframe = 0;
1293 s->current_subframe++;
1295 if (s->current_subframe >= s->subframes) {
1297 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1299 /* Read subframe footer */
1300 if (dca_subframe_footer(s, base_channel))
1308 * Convert bitstream to one representation based on sync marker
1310 static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst,
1315 const uint16_t *ssrc = (const uint16_t *) src;
1316 uint16_t *sdst = (uint16_t *) dst;
1319 if ((unsigned)src_size > (unsigned)max_size) {
1320 // av_log(NULL, AV_LOG_ERROR, "Input frame size larger than DCA_MAX_FRAME_SIZE!\n");
1322 src_size = max_size;
1327 case DCA_MARKER_RAW_BE:
1328 memcpy(dst, src, src_size);
1330 case DCA_MARKER_RAW_LE:
1331 for (i = 0; i < (src_size + 1) >> 1; i++)
1332 *sdst++ = av_bswap16(*ssrc++);
1334 case DCA_MARKER_14B_BE:
1335 case DCA_MARKER_14B_LE:
1336 init_put_bits(&pb, dst, max_size);
1337 for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) {
1338 tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
1339 put_bits(&pb, 14, tmp);
1341 flush_put_bits(&pb);
1342 return (put_bits_count(&pb) + 7) >> 3;
1349 * Return the number of channels in an ExSS speaker mask (HD)
1351 static int dca_exss_mask2count(int mask)
1353 /* count bits that mean speaker pairs twice */
1354 return av_popcount(mask)
1355 + av_popcount(mask & (
1356 DCA_EXSS_CENTER_LEFT_RIGHT
1357 | DCA_EXSS_FRONT_LEFT_RIGHT
1358 | DCA_EXSS_FRONT_HIGH_LEFT_RIGHT
1359 | DCA_EXSS_WIDE_LEFT_RIGHT
1360 | DCA_EXSS_SIDE_LEFT_RIGHT
1361 | DCA_EXSS_SIDE_HIGH_LEFT_RIGHT
1362 | DCA_EXSS_SIDE_REAR_LEFT_RIGHT
1363 | DCA_EXSS_REAR_LEFT_RIGHT
1364 | DCA_EXSS_REAR_HIGH_LEFT_RIGHT
1369 * Skip mixing coefficients of a single mix out configuration (HD)
1371 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1375 for (i = 0; i < channels; i++) {
1376 int mix_map_mask = get_bits(gb, out_ch);
1377 int num_coeffs = av_popcount(mix_map_mask);
1378 skip_bits_long(gb, num_coeffs * 6);
1383 * Parse extension substream asset header (HD)
1385 static int dca_exss_parse_asset_header(DCAContext *s)
1387 int header_pos = get_bits_count(&s->gb);
1390 int embedded_stereo = 0;
1391 int embedded_6ch = 0;
1392 int drc_code_present;
1393 int extensions_mask;
1396 if (get_bits_left(&s->gb) < 16)
1399 /* We will parse just enough to get to the extensions bitmask with which
1400 * we can set the profile value. */
1402 header_size = get_bits(&s->gb, 9) + 1;
1403 skip_bits(&s->gb, 3); // asset index
1405 if (s->static_fields) {
1406 if (get_bits1(&s->gb))
1407 skip_bits(&s->gb, 4); // asset type descriptor
1408 if (get_bits1(&s->gb))
1409 skip_bits_long(&s->gb, 24); // language descriptor
1411 if (get_bits1(&s->gb)) {
1412 /* How can one fit 1024 bytes of text here if the maximum value
1413 * for the asset header size field above was 512 bytes? */
1414 int text_length = get_bits(&s->gb, 10) + 1;
1415 if (get_bits_left(&s->gb) < text_length * 8)
1417 skip_bits_long(&s->gb, text_length * 8); // info text
1420 skip_bits(&s->gb, 5); // bit resolution - 1
1421 skip_bits(&s->gb, 4); // max sample rate code
1422 channels = get_bits(&s->gb, 8) + 1;
1424 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1425 int spkr_remap_sets;
1426 int spkr_mask_size = 16;
1430 embedded_stereo = get_bits1(&s->gb);
1432 embedded_6ch = get_bits1(&s->gb);
1434 if (get_bits1(&s->gb)) {
1435 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1436 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1439 spkr_remap_sets = get_bits(&s->gb, 3);
1441 for (i = 0; i < spkr_remap_sets; i++) {
1442 /* std layout mask for each remap set */
1443 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1446 for (i = 0; i < spkr_remap_sets; i++) {
1447 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1448 if (get_bits_left(&s->gb) < 0)
1451 for (j = 0; j < num_spkrs[i]; j++) {
1452 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1453 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1454 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1459 skip_bits(&s->gb, 3); // representation type
1463 drc_code_present = get_bits1(&s->gb);
1464 if (drc_code_present)
1465 get_bits(&s->gb, 8); // drc code
1467 if (get_bits1(&s->gb))
1468 skip_bits(&s->gb, 5); // dialog normalization code
1470 if (drc_code_present && embedded_stereo)
1471 get_bits(&s->gb, 8); // drc stereo code
1473 if (s->mix_metadata && get_bits1(&s->gb)) {
1474 skip_bits(&s->gb, 1); // external mix
1475 skip_bits(&s->gb, 6); // post mix gain code
1477 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1478 skip_bits(&s->gb, 3); // drc limit
1480 skip_bits(&s->gb, 8); // custom drc code
1482 if (get_bits1(&s->gb)) // channel specific scaling
1483 for (i = 0; i < s->num_mix_configs; i++)
1484 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1486 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1488 for (i = 0; i < s->num_mix_configs; i++) {
1489 if (get_bits_left(&s->gb) < 0)
1491 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1493 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1494 if (embedded_stereo)
1495 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1499 switch (get_bits(&s->gb, 2)) {
1500 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1501 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1502 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1503 case 3: extensions_mask = 0; /* aux coding */ break;
1506 /* not parsed further, we were only interested in the extensions mask */
1508 if (get_bits_left(&s->gb) < 0)
1511 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1512 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1515 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1517 if (extensions_mask & DCA_EXT_EXSS_XLL)
1518 s->profile = FF_PROFILE_DTS_HD_MA;
1519 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1521 s->profile = FF_PROFILE_DTS_HD_HRA;
1523 if (!(extensions_mask & DCA_EXT_CORE))
1524 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1525 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1526 av_log(s->avctx, AV_LOG_WARNING, "DTS extensions detection mismatch (%d, %d)\n",
1527 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1533 * Parse extension substream header (HD)
1535 static void dca_exss_parse_header(DCAContext *s)
1541 int active_ss_mask[8];
1544 if (get_bits_left(&s->gb) < 52)
1547 skip_bits(&s->gb, 8); // user data
1548 ss_index = get_bits(&s->gb, 2);
1550 blownup = get_bits1(&s->gb);
1551 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1552 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1554 s->static_fields = get_bits1(&s->gb);
1555 if (s->static_fields) {
1556 skip_bits(&s->gb, 2); // reference clock code
1557 skip_bits(&s->gb, 3); // frame duration code
1559 if (get_bits1(&s->gb))
1560 skip_bits_long(&s->gb, 36); // timestamp
1562 /* a single stream can contain multiple audio assets that can be
1563 * combined to form multiple audio presentations */
1565 num_audiop = get_bits(&s->gb, 3) + 1;
1566 if (num_audiop > 1) {
1567 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1568 /* ignore such streams for now */
1572 num_assets = get_bits(&s->gb, 3) + 1;
1573 if (num_assets > 1) {
1574 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1575 /* ignore such streams for now */
1579 for (i = 0; i < num_audiop; i++)
1580 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1582 for (i = 0; i < num_audiop; i++)
1583 for (j = 0; j <= ss_index; j++)
1584 if (active_ss_mask[i] & (1 << j))
1585 skip_bits(&s->gb, 8); // active asset mask
1587 s->mix_metadata = get_bits1(&s->gb);
1588 if (s->mix_metadata) {
1589 int mix_out_mask_size;
1591 skip_bits(&s->gb, 2); // adjustment level
1592 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1593 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1595 for (i = 0; i < s->num_mix_configs; i++) {
1596 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1597 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1602 for (i = 0; i < num_assets; i++)
1603 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1605 for (i = 0; i < num_assets; i++) {
1606 if (dca_exss_parse_asset_header(s))
1610 /* not parsed further, we were only interested in the extensions mask
1611 * from the asset header */
1615 * Main frame decoding function
1616 * FIXME add arguments
1618 static int dca_decode_frame(AVCodecContext * avctx,
1619 void *data, int *data_size,
1622 const uint8_t *buf = avpkt->data;
1623 int buf_size = avpkt->size;
1626 int num_core_channels = 0;
1628 float *samples_flt = data;
1629 int16_t *samples_s16 = data;
1631 DCAContext *s = avctx->priv_data;
1638 s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1639 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1640 if (s->dca_buffer_size == -1) {
1641 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1645 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1646 if (dca_parse_frame_header(s) < 0) {
1647 //seems like the frame is corrupt, try with the next one
1651 //set AVCodec values with parsed data
1652 avctx->sample_rate = s->sample_rate;
1653 avctx->bit_rate = s->bit_rate;
1654 avctx->frame_size = s->sample_blocks * 32;
1656 s->profile = FF_PROFILE_DTS;
1658 for (i = 0; i < (s->sample_blocks / 8); i++) {
1659 dca_decode_block(s, 0, i);
1662 /* record number of core channels incase less than max channels are requested */
1663 num_core_channels = s->prim_channels;
1666 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1668 s->core_ext_mask = 0;
1670 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1672 /* only scan for extensions if ext_descr was unknown or indicated a
1673 * supported XCh extension */
1674 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1676 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1677 * extensions scan can fill it up */
1678 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1680 /* extensions start at 32-bit boundaries into bitstream */
1681 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1683 while(core_ss_end - get_bits_count(&s->gb) >= 32) {
1684 uint32_t bits = get_bits_long(&s->gb, 32);
1688 int ext_amode, xch_fsize;
1690 s->xch_base_channel = s->prim_channels;
1692 /* validate sync word using XCHFSIZE field */
1693 xch_fsize = show_bits(&s->gb, 10);
1694 if((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1695 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1698 /* skip length-to-end-of-frame field for the moment */
1699 skip_bits(&s->gb, 10);
1701 s->core_ext_mask |= DCA_EXT_XCH;
1703 /* extension amode should == 1, number of channels in extension */
1704 /* AFAIK XCh is not used for more channels */
1705 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1706 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1707 " supported!\n",ext_amode);
1711 /* much like core primary audio coding header */
1712 dca_parse_audio_coding_header(s, s->xch_base_channel);
1714 for (i = 0; i < (s->sample_blocks / 8); i++) {
1715 dca_decode_block(s, s->xch_base_channel, i);
1722 /* XXCh: extended channels */
1723 /* usually found either in core or HD part in DTS-HD HRA streams,
1724 * but not in DTS-ES which contains XCh extensions instead */
1725 s->core_ext_mask |= DCA_EXT_XXCH;
1729 int fsize96 = show_bits(&s->gb, 12) + 1;
1730 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1733 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n", get_bits_count(&s->gb));
1734 skip_bits(&s->gb, 12);
1735 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1736 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1738 s->core_ext_mask |= DCA_EXT_X96;
1743 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1747 /* no supported extensions, skip the rest of the core substream */
1748 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1751 if (s->core_ext_mask & DCA_EXT_X96)
1752 s->profile = FF_PROFILE_DTS_96_24;
1753 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1754 s->profile = FF_PROFILE_DTS_ES;
1756 /* check for ExSS (HD part) */
1757 if (s->dca_buffer_size - s->frame_size > 32
1758 && get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1759 dca_exss_parse_header(s);
1761 avctx->profile = s->profile;
1763 channels = s->prim_channels + !!s->lfe;
1766 avctx->channel_layout = dca_core_channel_layout[s->amode];
1768 if (s->xch_present && (!avctx->request_channels ||
1769 avctx->request_channels > num_core_channels + !!s->lfe)) {
1770 avctx->channel_layout |= AV_CH_BACK_CENTER;
1772 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1773 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1775 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1778 channels = num_core_channels + !!s->lfe;
1779 s->xch_present = 0; /* disable further xch processing */
1781 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1782 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1784 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1787 if (channels > !!s->lfe &&
1788 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1791 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1793 s->output = DCA_STEREO;
1794 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1797 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
1802 /* There is nothing that prevents a dts frame to change channel configuration
1803 but Libav doesn't support that so only set the channels if it is previously
1804 unset. Ideally during the first probe for channels the crc should be checked
1805 and only set avctx->channels when the crc is ok. Right now the decoder could
1806 set the channels based on a broken first frame.*/
1807 if (s->is_channels_set == 0) {
1808 s->is_channels_set = 1;
1809 avctx->channels = channels;
1811 if (avctx->channels != channels) {
1812 av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of "
1813 "channels changing in stream. Skipping frame.\n");
1817 out_size = 256 / 8 * s->sample_blocks * channels *
1818 av_get_bytes_per_sample(avctx->sample_fmt);
1819 if (*data_size < out_size)
1821 *data_size = out_size;
1823 /* filter to get final output */
1824 for (i = 0; i < (s->sample_blocks / 8); i++) {
1825 dca_filter_channels(s, i);
1827 /* If this was marked as a DTS-ES stream we need to subtract back- */
1828 /* channel from SL & SR to remove matrixed back-channel signal */
1829 if((s->source_pcm_res & 1) && s->xch_present) {
1830 float* back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1831 float* lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1832 float* rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1833 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1834 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1837 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1838 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1840 samples_flt += 256 * channels;
1842 s->fmt_conv.float_to_int16_interleave(samples_s16,
1843 s->samples_chanptr, 256,
1845 samples_s16 += 256 * channels;
1849 /* update lfe history */
1850 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1851 for (i = 0; i < 2 * s->lfe * 4; i++) {
1852 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1861 * DCA initialization
1863 * @param avctx pointer to the AVCodecContext
1866 static av_cold int dca_decode_init(AVCodecContext * avctx)
1868 DCAContext *s = avctx->priv_data;
1874 dsputil_init(&s->dsp, avctx);
1875 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1876 ff_synth_filter_init(&s->synth);
1877 ff_dcadsp_init(&s->dcadsp);
1878 ff_fmt_convert_init(&s->fmt_conv, avctx);
1880 for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++)
1881 s->samples_chanptr[i] = s->samples + i * 256;
1883 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1884 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1885 s->scale_bias = 1.0 / 32768.0;
1887 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1888 s->scale_bias = 1.0;
1891 /* allow downmixing to stereo */
1892 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1893 avctx->request_channels == 2) {
1894 avctx->channels = avctx->request_channels;
1900 static av_cold int dca_decode_end(AVCodecContext * avctx)
1902 DCAContext *s = avctx->priv_data;
1903 ff_mdct_end(&s->imdct);
1907 static const AVProfile profiles[] = {
1908 { FF_PROFILE_DTS, "DTS" },
1909 { FF_PROFILE_DTS_ES, "DTS-ES" },
1910 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1911 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1912 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1913 { FF_PROFILE_UNKNOWN },
1916 AVCodec ff_dca_decoder = {
1918 .type = AVMEDIA_TYPE_AUDIO,
1920 .priv_data_size = sizeof(DCAContext),
1921 .init = dca_decode_init,
1922 .decode = dca_decode_frame,
1923 .close = dca_decode_end,
1924 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1925 .capabilities = CODEC_CAP_CHANNEL_CONF,
1926 .sample_fmts = (const enum AVSampleFormat[]) {
1927 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
1929 .profiles = NULL_IF_CONFIG_SMALL(profiles),