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 /* Reconstructed channel sample index */
902 for (subindex = 0; subindex < 8; subindex++) {
903 /* Load in one sample from each subband and clear inactive subbands */
904 for (i = 0; i < sb_act; i++){
905 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ ((i-1)&2)<<30;
906 AV_WN32A(&s->raXin[i], v);
911 s->synth.synth_filter_float(&s->imdct,
912 s->subband_fir_hist[chans], &s->hist_index[chans],
913 s->subband_fir_noidea[chans], prCoeff,
914 samples_out, s->raXin, scale);
920 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
921 int num_deci_sample, float *samples_in,
922 float *samples_out, float scale)
924 /* samples_in: An array holding decimated samples.
925 * Samples in current subframe starts from samples_in[0],
926 * while samples_in[-1], samples_in[-2], ..., stores samples
927 * from last subframe as history.
929 * samples_out: An array holding interpolated samples
933 const float *prCoeff;
936 /* Select decimation filter */
937 if (decimation_select == 1) {
939 prCoeff = lfe_fir_128;
942 prCoeff = lfe_fir_64;
945 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
946 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor,
949 samples_out += 2 * decifactor;
953 /* downmixing routines */
954 #define MIX_REAR1(samples, si1, rs, coef) \
955 samples[i] += samples[si1] * coef[rs][0]; \
956 samples[i+256] += samples[si1] * coef[rs][1];
958 #define MIX_REAR2(samples, si1, si2, rs, coef) \
959 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs+1][0]; \
960 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs+1][1];
962 #define MIX_FRONT3(samples, coef) \
966 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
967 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
969 #define DOWNMIX_TO_STEREO(op1, op2) \
970 for (i = 0; i < 256; i++){ \
975 static void dca_downmix(float *samples, int srcfmt,
976 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
977 const int8_t *channel_mapping)
982 float coef[DCA_PRIM_CHANNELS_MAX][2];
984 for (i=0; i<DCA_PRIM_CHANNELS_MAX; i++) {
985 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
986 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
992 case DCA_STEREO_TOTAL:
993 case DCA_STEREO_SUMDIFF:
995 av_log(NULL, 0, "Not implemented!\n");
1000 c = channel_mapping[0] * 256;
1001 l = channel_mapping[1] * 256;
1002 r = channel_mapping[2] * 256;
1003 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),);
1006 s = channel_mapping[2] * 256;
1007 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef),);
1010 c = channel_mapping[0] * 256;
1011 l = channel_mapping[1] * 256;
1012 r = channel_mapping[2] * 256;
1013 s = channel_mapping[3] * 256;
1014 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1015 MIX_REAR1(samples, i + s, 3, coef));
1018 sl = channel_mapping[2] * 256;
1019 sr = channel_mapping[3] * 256;
1020 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef),);
1023 c = channel_mapping[0] * 256;
1024 l = channel_mapping[1] * 256;
1025 r = channel_mapping[2] * 256;
1026 sl = channel_mapping[3] * 256;
1027 sr = channel_mapping[4] * 256;
1028 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1029 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1035 /* Very compact version of the block code decoder that does not use table
1036 * look-up but is slightly slower */
1037 static int decode_blockcode(int code, int levels, int *values)
1040 int offset = (levels - 1) >> 1;
1042 for (i = 0; i < 4; i++) {
1043 int div = FASTDIV(code, levels);
1044 values[i] = code - offset - div*levels;
1051 av_log(NULL, AV_LOG_ERROR, "ERROR: block code look-up failed\n");
1056 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1057 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1059 static int dca_subsubframe(DCAContext * s, int base_channel, int block_index)
1062 int subsubframe = s->current_subsubframe;
1064 const float *quant_step_table;
1067 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1068 LOCAL_ALIGNED_16(int, block, [8]);
1074 /* Select quantization step size table */
1075 if (s->bit_rate_index == 0x1f)
1076 quant_step_table = lossless_quant_d;
1078 quant_step_table = lossy_quant_d;
1080 for (k = base_channel; k < s->prim_channels; k++) {
1081 if (get_bits_left(&s->gb) < 0)
1084 for (l = 0; l < s->vq_start_subband[k]; l++) {
1087 /* Select the mid-tread linear quantizer */
1088 int abits = s->bitalloc[k][l];
1090 float quant_step_size = quant_step_table[abits];
1093 * Determine quantization index code book and its type
1096 /* Select quantization index code book */
1097 int sel = s->quant_index_huffman[k][abits];
1100 * Extract bits from the bit stream
1103 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1105 /* Deal with transients */
1106 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1107 float rscale = quant_step_size * s->scale_factor[k][l][sfi] * s->scalefactor_adj[k][sel];
1109 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){
1112 int block_code1, block_code2, size, levels;
1114 size = abits_sizes[abits-1];
1115 levels = abits_levels[abits-1];
1117 block_code1 = get_bits(&s->gb, size);
1118 /* FIXME Should test return value */
1119 decode_blockcode(block_code1, levels, block);
1120 block_code2 = get_bits(&s->gb, size);
1121 decode_blockcode(block_code2, levels, &block[4]);
1124 for (m = 0; m < 8; m++)
1125 block[m] = get_sbits(&s->gb, abits - 3);
1129 for (m = 0; m < 8; m++)
1130 block[m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel);
1133 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1138 * Inverse ADPCM if in prediction mode
1140 if (s->prediction_mode[k][l]) {
1142 for (m = 0; m < 8; m++) {
1143 for (n = 1; n <= 4; n++)
1145 subband_samples[k][l][m] +=
1146 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1147 subband_samples[k][l][m - n] / 8192);
1148 else if (s->predictor_history)
1149 subband_samples[k][l][m] +=
1150 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1151 s->subband_samples_hist[k][l][m - n +
1158 * Decode VQ encoded high frequencies
1160 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1161 /* 1 vector -> 32 samples but we only need the 8 samples
1162 * for this subsubframe. */
1165 if (!s->debug_flag & 0x01) {
1166 av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
1167 s->debug_flag |= 0x01;
1170 for (m = 0; m < 8; m++) {
1171 subband_samples[k][l][m] =
1172 high_freq_vq[s->high_freq_vq[k][l]][subsubframe * 8 +
1174 * (float) s->scale_factor[k][l][0] / 16.0;
1179 /* Check for DSYNC after subsubframe */
1180 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1181 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1183 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1186 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1190 /* Backup predictor history for adpcm */
1191 for (k = base_channel; k < s->prim_channels; k++)
1192 for (l = 0; l < s->vq_start_subband[k]; l++)
1193 memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
1194 4 * sizeof(subband_samples[0][0][0]));
1199 static int dca_filter_channels(DCAContext * s, int block_index)
1201 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1204 /* 32 subbands QMF */
1205 for (k = 0; k < s->prim_channels; k++) {
1206 /* static float pcm_to_double[8] =
1207 {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
1208 qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
1209 M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ );
1213 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1214 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1217 /* Generate LFE samples for this subsubframe FIXME!!! */
1218 if (s->output & DCA_LFE) {
1219 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1220 s->lfe_data + 2 * s->lfe * (block_index + 4),
1221 &s->samples[256 * dca_lfe_index[s->amode]],
1222 (1.0/256.0)*s->scale_bias);
1223 /* Outputs 20bits pcm samples */
1230 static int dca_subframe_footer(DCAContext * s, int base_channel)
1232 int aux_data_count = 0, i;
1235 * Unpack optional information
1238 /* presumably optional information only appears in the core? */
1239 if (!base_channel) {
1241 get_bits(&s->gb, 32);
1244 aux_data_count = get_bits(&s->gb, 6);
1246 for (i = 0; i < aux_data_count; i++)
1247 get_bits(&s->gb, 8);
1249 if (s->crc_present && (s->downmix || s->dynrange))
1250 get_bits(&s->gb, 16);
1257 * Decode a dca frame block
1259 * @param s pointer to the DCAContext
1262 static int dca_decode_block(DCAContext * s, int base_channel, int block_index)
1266 if (s->current_subframe >= s->subframes) {
1267 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1268 s->current_subframe, s->subframes);
1272 if (!s->current_subsubframe) {
1274 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1276 /* Read subframe header */
1277 if (dca_subframe_header(s, base_channel, block_index))
1281 /* Read subsubframe */
1283 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1285 if (dca_subsubframe(s, base_channel, block_index))
1289 s->current_subsubframe++;
1290 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1291 s->current_subsubframe = 0;
1292 s->current_subframe++;
1294 if (s->current_subframe >= s->subframes) {
1296 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1298 /* Read subframe footer */
1299 if (dca_subframe_footer(s, base_channel))
1307 * Convert bitstream to one representation based on sync marker
1309 static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst,
1314 const uint16_t *ssrc = (const uint16_t *) src;
1315 uint16_t *sdst = (uint16_t *) dst;
1318 if ((unsigned)src_size > (unsigned)max_size) {
1319 // av_log(NULL, AV_LOG_ERROR, "Input frame size larger than DCA_MAX_FRAME_SIZE!\n");
1321 src_size = max_size;
1326 case DCA_MARKER_RAW_BE:
1327 memcpy(dst, src, src_size);
1329 case DCA_MARKER_RAW_LE:
1330 for (i = 0; i < (src_size + 1) >> 1; i++)
1331 *sdst++ = av_bswap16(*ssrc++);
1333 case DCA_MARKER_14B_BE:
1334 case DCA_MARKER_14B_LE:
1335 init_put_bits(&pb, dst, max_size);
1336 for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) {
1337 tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
1338 put_bits(&pb, 14, tmp);
1340 flush_put_bits(&pb);
1341 return (put_bits_count(&pb) + 7) >> 3;
1348 * Return the number of channels in an ExSS speaker mask (HD)
1350 static int dca_exss_mask2count(int mask)
1352 /* count bits that mean speaker pairs twice */
1353 return av_popcount(mask)
1354 + av_popcount(mask & (
1355 DCA_EXSS_CENTER_LEFT_RIGHT
1356 | DCA_EXSS_FRONT_LEFT_RIGHT
1357 | DCA_EXSS_FRONT_HIGH_LEFT_RIGHT
1358 | DCA_EXSS_WIDE_LEFT_RIGHT
1359 | DCA_EXSS_SIDE_LEFT_RIGHT
1360 | DCA_EXSS_SIDE_HIGH_LEFT_RIGHT
1361 | DCA_EXSS_SIDE_REAR_LEFT_RIGHT
1362 | DCA_EXSS_REAR_LEFT_RIGHT
1363 | DCA_EXSS_REAR_HIGH_LEFT_RIGHT
1368 * Skip mixing coefficients of a single mix out configuration (HD)
1370 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1374 for (i = 0; i < channels; i++) {
1375 int mix_map_mask = get_bits(gb, out_ch);
1376 int num_coeffs = av_popcount(mix_map_mask);
1377 skip_bits_long(gb, num_coeffs * 6);
1382 * Parse extension substream asset header (HD)
1384 static int dca_exss_parse_asset_header(DCAContext *s)
1386 int header_pos = get_bits_count(&s->gb);
1389 int embedded_stereo = 0;
1390 int embedded_6ch = 0;
1391 int drc_code_present;
1392 int extensions_mask;
1395 if (get_bits_left(&s->gb) < 16)
1398 /* We will parse just enough to get to the extensions bitmask with which
1399 * we can set the profile value. */
1401 header_size = get_bits(&s->gb, 9) + 1;
1402 skip_bits(&s->gb, 3); // asset index
1404 if (s->static_fields) {
1405 if (get_bits1(&s->gb))
1406 skip_bits(&s->gb, 4); // asset type descriptor
1407 if (get_bits1(&s->gb))
1408 skip_bits_long(&s->gb, 24); // language descriptor
1410 if (get_bits1(&s->gb)) {
1411 /* How can one fit 1024 bytes of text here if the maximum value
1412 * for the asset header size field above was 512 bytes? */
1413 int text_length = get_bits(&s->gb, 10) + 1;
1414 if (get_bits_left(&s->gb) < text_length * 8)
1416 skip_bits_long(&s->gb, text_length * 8); // info text
1419 skip_bits(&s->gb, 5); // bit resolution - 1
1420 skip_bits(&s->gb, 4); // max sample rate code
1421 channels = get_bits(&s->gb, 8) + 1;
1423 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1424 int spkr_remap_sets;
1425 int spkr_mask_size = 16;
1429 embedded_stereo = get_bits1(&s->gb);
1431 embedded_6ch = get_bits1(&s->gb);
1433 if (get_bits1(&s->gb)) {
1434 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1435 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1438 spkr_remap_sets = get_bits(&s->gb, 3);
1440 for (i = 0; i < spkr_remap_sets; i++) {
1441 /* std layout mask for each remap set */
1442 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1445 for (i = 0; i < spkr_remap_sets; i++) {
1446 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1447 if (get_bits_left(&s->gb) < 0)
1450 for (j = 0; j < num_spkrs[i]; j++) {
1451 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1452 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1453 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1458 skip_bits(&s->gb, 3); // representation type
1462 drc_code_present = get_bits1(&s->gb);
1463 if (drc_code_present)
1464 get_bits(&s->gb, 8); // drc code
1466 if (get_bits1(&s->gb))
1467 skip_bits(&s->gb, 5); // dialog normalization code
1469 if (drc_code_present && embedded_stereo)
1470 get_bits(&s->gb, 8); // drc stereo code
1472 if (s->mix_metadata && get_bits1(&s->gb)) {
1473 skip_bits(&s->gb, 1); // external mix
1474 skip_bits(&s->gb, 6); // post mix gain code
1476 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1477 skip_bits(&s->gb, 3); // drc limit
1479 skip_bits(&s->gb, 8); // custom drc code
1481 if (get_bits1(&s->gb)) // channel specific scaling
1482 for (i = 0; i < s->num_mix_configs; i++)
1483 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1485 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1487 for (i = 0; i < s->num_mix_configs; i++) {
1488 if (get_bits_left(&s->gb) < 0)
1490 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1492 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1493 if (embedded_stereo)
1494 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1498 switch (get_bits(&s->gb, 2)) {
1499 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1500 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1501 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1502 case 3: extensions_mask = 0; /* aux coding */ break;
1505 /* not parsed further, we were only interested in the extensions mask */
1507 if (get_bits_left(&s->gb) < 0)
1510 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1511 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1514 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1516 if (extensions_mask & DCA_EXT_EXSS_XLL)
1517 s->profile = FF_PROFILE_DTS_HD_MA;
1518 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1520 s->profile = FF_PROFILE_DTS_HD_HRA;
1522 if (!(extensions_mask & DCA_EXT_CORE))
1523 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1524 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1525 av_log(s->avctx, AV_LOG_WARNING, "DTS extensions detection mismatch (%d, %d)\n",
1526 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1532 * Parse extension substream header (HD)
1534 static void dca_exss_parse_header(DCAContext *s)
1540 int active_ss_mask[8];
1543 if (get_bits_left(&s->gb) < 52)
1546 skip_bits(&s->gb, 8); // user data
1547 ss_index = get_bits(&s->gb, 2);
1549 blownup = get_bits1(&s->gb);
1550 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1551 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1553 s->static_fields = get_bits1(&s->gb);
1554 if (s->static_fields) {
1555 skip_bits(&s->gb, 2); // reference clock code
1556 skip_bits(&s->gb, 3); // frame duration code
1558 if (get_bits1(&s->gb))
1559 skip_bits_long(&s->gb, 36); // timestamp
1561 /* a single stream can contain multiple audio assets that can be
1562 * combined to form multiple audio presentations */
1564 num_audiop = get_bits(&s->gb, 3) + 1;
1565 if (num_audiop > 1) {
1566 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1567 /* ignore such streams for now */
1571 num_assets = get_bits(&s->gb, 3) + 1;
1572 if (num_assets > 1) {
1573 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1574 /* ignore such streams for now */
1578 for (i = 0; i < num_audiop; i++)
1579 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1581 for (i = 0; i < num_audiop; i++)
1582 for (j = 0; j <= ss_index; j++)
1583 if (active_ss_mask[i] & (1 << j))
1584 skip_bits(&s->gb, 8); // active asset mask
1586 s->mix_metadata = get_bits1(&s->gb);
1587 if (s->mix_metadata) {
1588 int mix_out_mask_size;
1590 skip_bits(&s->gb, 2); // adjustment level
1591 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1592 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1594 for (i = 0; i < s->num_mix_configs; i++) {
1595 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1596 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1601 for (i = 0; i < num_assets; i++)
1602 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1604 for (i = 0; i < num_assets; i++) {
1605 if (dca_exss_parse_asset_header(s))
1609 /* not parsed further, we were only interested in the extensions mask
1610 * from the asset header */
1614 * Main frame decoding function
1615 * FIXME add arguments
1617 static int dca_decode_frame(AVCodecContext * avctx,
1618 void *data, int *data_size,
1621 const uint8_t *buf = avpkt->data;
1622 int buf_size = avpkt->size;
1625 int num_core_channels = 0;
1627 float *samples_flt = data;
1628 int16_t *samples_s16 = data;
1630 DCAContext *s = avctx->priv_data;
1637 s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1638 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1639 if (s->dca_buffer_size == -1) {
1640 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1644 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1645 if (dca_parse_frame_header(s) < 0) {
1646 //seems like the frame is corrupt, try with the next one
1650 //set AVCodec values with parsed data
1651 avctx->sample_rate = s->sample_rate;
1652 avctx->bit_rate = s->bit_rate;
1653 avctx->frame_size = s->sample_blocks * 32;
1655 s->profile = FF_PROFILE_DTS;
1657 for (i = 0; i < (s->sample_blocks / 8); i++) {
1658 dca_decode_block(s, 0, i);
1661 /* record number of core channels incase less than max channels are requested */
1662 num_core_channels = s->prim_channels;
1665 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1667 s->core_ext_mask = 0;
1669 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1671 /* only scan for extensions if ext_descr was unknown or indicated a
1672 * supported XCh extension */
1673 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1675 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1676 * extensions scan can fill it up */
1677 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1679 /* extensions start at 32-bit boundaries into bitstream */
1680 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1682 while(core_ss_end - get_bits_count(&s->gb) >= 32) {
1683 uint32_t bits = get_bits_long(&s->gb, 32);
1687 int ext_amode, xch_fsize;
1689 s->xch_base_channel = s->prim_channels;
1691 /* validate sync word using XCHFSIZE field */
1692 xch_fsize = show_bits(&s->gb, 10);
1693 if((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1694 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1697 /* skip length-to-end-of-frame field for the moment */
1698 skip_bits(&s->gb, 10);
1700 s->core_ext_mask |= DCA_EXT_XCH;
1702 /* extension amode should == 1, number of channels in extension */
1703 /* AFAIK XCh is not used for more channels */
1704 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1705 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1706 " supported!\n",ext_amode);
1710 /* much like core primary audio coding header */
1711 dca_parse_audio_coding_header(s, s->xch_base_channel);
1713 for (i = 0; i < (s->sample_blocks / 8); i++) {
1714 dca_decode_block(s, s->xch_base_channel, i);
1721 /* XXCh: extended channels */
1722 /* usually found either in core or HD part in DTS-HD HRA streams,
1723 * but not in DTS-ES which contains XCh extensions instead */
1724 s->core_ext_mask |= DCA_EXT_XXCH;
1728 int fsize96 = show_bits(&s->gb, 12) + 1;
1729 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1732 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n", get_bits_count(&s->gb));
1733 skip_bits(&s->gb, 12);
1734 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1735 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1737 s->core_ext_mask |= DCA_EXT_X96;
1742 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1746 /* no supported extensions, skip the rest of the core substream */
1747 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1750 if (s->core_ext_mask & DCA_EXT_X96)
1751 s->profile = FF_PROFILE_DTS_96_24;
1752 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1753 s->profile = FF_PROFILE_DTS_ES;
1755 /* check for ExSS (HD part) */
1756 if (s->dca_buffer_size - s->frame_size > 32
1757 && get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1758 dca_exss_parse_header(s);
1760 avctx->profile = s->profile;
1762 channels = s->prim_channels + !!s->lfe;
1765 avctx->channel_layout = dca_core_channel_layout[s->amode];
1767 if (s->xch_present && (!avctx->request_channels ||
1768 avctx->request_channels > num_core_channels + !!s->lfe)) {
1769 avctx->channel_layout |= AV_CH_BACK_CENTER;
1771 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1772 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1774 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1777 channels = num_core_channels + !!s->lfe;
1778 s->xch_present = 0; /* disable further xch processing */
1780 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1781 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1783 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1786 if (channels > !!s->lfe &&
1787 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1790 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1792 s->output = DCA_STEREO;
1793 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1796 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
1801 /* There is nothing that prevents a dts frame to change channel configuration
1802 but Libav doesn't support that so only set the channels if it is previously
1803 unset. Ideally during the first probe for channels the crc should be checked
1804 and only set avctx->channels when the crc is ok. Right now the decoder could
1805 set the channels based on a broken first frame.*/
1806 if (s->is_channels_set == 0) {
1807 s->is_channels_set = 1;
1808 avctx->channels = channels;
1810 if (avctx->channels != channels) {
1811 av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of "
1812 "channels changing in stream. Skipping frame.\n");
1816 out_size = 256 / 8 * s->sample_blocks * channels *
1817 av_get_bytes_per_sample(avctx->sample_fmt);
1818 if (*data_size < out_size)
1820 *data_size = out_size;
1822 /* filter to get final output */
1823 for (i = 0; i < (s->sample_blocks / 8); i++) {
1824 dca_filter_channels(s, i);
1826 /* If this was marked as a DTS-ES stream we need to subtract back- */
1827 /* channel from SL & SR to remove matrixed back-channel signal */
1828 if((s->source_pcm_res & 1) && s->xch_present) {
1829 float* back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1830 float* lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1831 float* rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1833 for(j = 0; j < 256; ++j) {
1834 lt_chan[j] -= back_chan[j] * M_SQRT1_2;
1835 rt_chan[j] -= back_chan[j] * M_SQRT1_2;
1839 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1840 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1842 samples_flt += 256 * channels;
1844 s->fmt_conv.float_to_int16_interleave(samples_s16,
1845 s->samples_chanptr, 256,
1847 samples_s16 += 256 * channels;
1851 /* update lfe history */
1852 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1853 for (i = 0; i < 2 * s->lfe * 4; i++) {
1854 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1863 * DCA initialization
1865 * @param avctx pointer to the AVCodecContext
1868 static av_cold int dca_decode_init(AVCodecContext * avctx)
1870 DCAContext *s = avctx->priv_data;
1876 dsputil_init(&s->dsp, avctx);
1877 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1878 ff_synth_filter_init(&s->synth);
1879 ff_dcadsp_init(&s->dcadsp);
1880 ff_fmt_convert_init(&s->fmt_conv, avctx);
1882 for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++)
1883 s->samples_chanptr[i] = s->samples + i * 256;
1885 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1886 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1887 s->scale_bias = 1.0 / 32768.0;
1889 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1890 s->scale_bias = 1.0;
1893 /* allow downmixing to stereo */
1894 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1895 avctx->request_channels == 2) {
1896 avctx->channels = avctx->request_channels;
1902 static av_cold int dca_decode_end(AVCodecContext * avctx)
1904 DCAContext *s = avctx->priv_data;
1905 ff_mdct_end(&s->imdct);
1909 static const AVProfile profiles[] = {
1910 { FF_PROFILE_DTS, "DTS" },
1911 { FF_PROFILE_DTS_ES, "DTS-ES" },
1912 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1913 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1914 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1915 { FF_PROFILE_UNKNOWN },
1918 AVCodec ff_dca_decoder = {
1920 .type = AVMEDIA_TYPE_AUDIO,
1922 .priv_data_size = sizeof(DCAContext),
1923 .init = dca_decode_init,
1924 .decode = dca_decode_frame,
1925 .close = dca_decode_end,
1926 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1927 .capabilities = CODEC_CAP_CHANNEL_CONF,
1928 .sample_fmts = (const enum AVSampleFormat[]) {
1929 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
1931 .profiles = NULL_IF_CONFIG_SMALL(profiles),