2 * DCA compatible decoder
3 * Copyright (C) 2004 Gildas Bazin
4 * Copyright (C) 2004 Benjamin Zores
5 * Copyright (C) 2006 Benjamin Larsson
6 * Copyright (C) 2007 Konstantin Shishkov
8 * This file is part of FFmpeg.
10 * FFmpeg is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
15 * FFmpeg is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 #include "libavutil/common.h"
30 #include "libavutil/intmath.h"
31 #include "libavutil/intreadwrite.h"
32 #include "libavutil/audioconvert.h"
41 #include "synth_filter.h"
43 #include "fmtconvert.h"
51 #define DCA_PRIM_CHANNELS_MAX (7)
52 #define DCA_SUBBANDS (32)
53 #define DCA_ABITS_MAX (32) /* Should be 28 */
54 #define DCA_SUBSUBFRAMES_MAX (4)
55 #define DCA_SUBFRAMES_MAX (16)
56 #define DCA_BLOCKS_MAX (16)
57 #define DCA_LFE_MAX (3)
73 /* these are unconfirmed but should be mostly correct */
74 enum DCAExSSSpeakerMask {
75 DCA_EXSS_FRONT_CENTER = 0x0001,
76 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
77 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
78 DCA_EXSS_LFE = 0x0008,
79 DCA_EXSS_REAR_CENTER = 0x0010,
80 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
81 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
82 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
83 DCA_EXSS_OVERHEAD = 0x0100,
84 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
85 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
86 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
87 DCA_EXSS_LFE2 = 0x1000,
88 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
89 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
90 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
93 enum DCAExtensionMask {
94 DCA_EXT_CORE = 0x001, ///< core in core substream
95 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
96 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
97 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
98 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
99 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
100 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
101 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
102 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
103 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
106 /* -1 are reserved or unknown */
107 static const int dca_ext_audio_descr_mask[] = {
111 DCA_EXT_XCH | DCA_EXT_X96,
118 /* extensions that reside in core substream */
119 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
121 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
122 * Some compromises have been made for special configurations. Most configurations
123 * are never used so complete accuracy is not needed.
125 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
126 * S -> side, when both rear and back are configured move one of them to the side channel
128 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
131 static const int64_t dca_core_channel_layout[] = {
132 AV_CH_FRONT_CENTER, ///< 1, A
133 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
134 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
135 AV_CH_LAYOUT_STEREO, ///< 2, (L+R) + (L-R) (sum-difference)
136 AV_CH_LAYOUT_STEREO, ///< 2, LT +RT (left and right total)
137 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER, ///< 3, C+L+R
138 AV_CH_LAYOUT_STEREO|AV_CH_BACK_CENTER, ///< 3, L+R+S
139 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_BACK_CENTER, ///< 4, C + L + R+ S
140 AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 4, L + R +SL+ SR
141 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 5, C + L + R+ SL+SR
142 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
143 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
144 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
145 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
146 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
147 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
150 static const int8_t dca_lfe_index[] = {
151 1,2,2,2,2,3,2,3,2,3,2,3,1,3,2,3
154 static const int8_t dca_channel_reorder_lfe[][9] = {
155 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
156 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
157 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
158 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
159 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
160 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
161 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
162 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
163 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
164 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
165 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
166 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
167 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
168 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
169 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
170 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
173 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
174 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
175 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
176 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
177 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
178 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
179 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
180 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
181 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
182 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
183 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
184 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
185 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
186 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
187 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
188 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
189 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
192 static const int8_t dca_channel_reorder_nolfe[][9] = {
193 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
194 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
195 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
196 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
197 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
198 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
199 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
200 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
201 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
202 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
203 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
204 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
205 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
206 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
207 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
208 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
211 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
212 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
213 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
214 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
215 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
216 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
217 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
218 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
219 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
220 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
221 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
222 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
223 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
224 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
225 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
226 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
227 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
230 #define DCA_DOLBY 101 /* FIXME */
232 #define DCA_CHANNEL_BITS 6
233 #define DCA_CHANNEL_MASK 0x3F
237 #define HEADER_SIZE 14
239 #define DCA_MAX_FRAME_SIZE 16384
240 #define DCA_MAX_EXSS_HEADER_SIZE 4096
242 #define DCA_BUFFER_PADDING_SIZE 1024
244 /** Bit allocation */
246 int offset; ///< code values offset
247 int maxbits[8]; ///< max bits in VLC
248 int wrap; ///< wrap for get_vlc2()
249 VLC vlc[8]; ///< actual codes
252 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
253 static BitAlloc dca_tmode; ///< transition mode VLCs
254 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
255 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
257 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx)
259 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) + ba->offset;
263 AVCodecContext *avctx;
265 int frame_type; ///< type of the current frame
266 int samples_deficit; ///< deficit sample count
267 int crc_present; ///< crc is present in the bitstream
268 int sample_blocks; ///< number of PCM sample blocks
269 int frame_size; ///< primary frame byte size
270 int amode; ///< audio channels arrangement
271 int sample_rate; ///< audio sampling rate
272 int bit_rate; ///< transmission bit rate
273 int bit_rate_index; ///< transmission bit rate index
275 int downmix; ///< embedded downmix enabled
276 int dynrange; ///< embedded dynamic range flag
277 int timestamp; ///< embedded time stamp flag
278 int aux_data; ///< auxiliary data flag
279 int hdcd; ///< source material is mastered in HDCD
280 int ext_descr; ///< extension audio descriptor flag
281 int ext_coding; ///< extended coding flag
282 int aspf; ///< audio sync word insertion flag
283 int lfe; ///< low frequency effects flag
284 int predictor_history; ///< predictor history flag
285 int header_crc; ///< header crc check bytes
286 int multirate_inter; ///< multirate interpolator switch
287 int version; ///< encoder software revision
288 int copy_history; ///< copy history
289 int source_pcm_res; ///< source pcm resolution
290 int front_sum; ///< front sum/difference flag
291 int surround_sum; ///< surround sum/difference flag
292 int dialog_norm; ///< dialog normalisation parameter
294 /* Primary audio coding header */
295 int subframes; ///< number of subframes
296 int is_channels_set; ///< check for if the channel number is already set
297 int total_channels; ///< number of channels including extensions
298 int prim_channels; ///< number of primary audio channels
299 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
300 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
301 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
302 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
303 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
304 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
305 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
306 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
308 /* Primary audio coding side information */
309 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
310 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
311 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
312 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
313 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
314 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
315 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
316 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
317 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
318 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
319 int dynrange_coef; ///< dynamic range coefficient
321 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
323 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
324 int lfe_scale_factor;
326 /* Subband samples history (for ADPCM) */
327 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
328 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
329 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
330 int hist_index[DCA_PRIM_CHANNELS_MAX];
331 DECLARE_ALIGNED(32, float, raXin)[32];
333 int output; ///< type of output
334 float scale_bias; ///< output scale
336 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
337 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
338 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX+1];
340 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
341 int dca_buffer_size; ///< how much data is in the dca_buffer
343 const int8_t* channel_order_tab; ///< channel reordering table, lfe and non lfe
345 /* Current position in DCA frame */
346 int current_subframe;
347 int current_subsubframe;
349 int core_ext_mask; ///< present extensions in the core substream
351 /* XCh extension information */
352 int xch_present; ///< XCh extension present and valid
353 int xch_base_channel; ///< index of first (only) channel containing XCH data
355 /* ExSS header parser */
356 int static_fields; ///< static fields present
357 int mix_metadata; ///< mixing metadata present
358 int num_mix_configs; ///< number of mix out configurations
359 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
363 int debug_flag; ///< used for suppressing repeated error messages output
366 SynthFilterContext synth;
367 DCADSPContext dcadsp;
368 FmtConvertContext fmt_conv;
371 static const uint16_t dca_vlc_offs[] = {
372 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
373 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
374 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
375 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
376 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
377 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
380 static av_cold void dca_init_vlcs(void)
382 static int vlcs_initialized = 0;
384 static VLC_TYPE dca_table[23622][2];
386 if (vlcs_initialized)
389 dca_bitalloc_index.offset = 1;
390 dca_bitalloc_index.wrap = 2;
391 for (i = 0; i < 5; i++) {
392 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
393 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
394 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
395 bitalloc_12_bits[i], 1, 1,
396 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
398 dca_scalefactor.offset = -64;
399 dca_scalefactor.wrap = 2;
400 for (i = 0; i < 5; i++) {
401 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
402 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
403 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
404 scales_bits[i], 1, 1,
405 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
407 dca_tmode.offset = 0;
409 for (i = 0; i < 4; i++) {
410 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
411 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
412 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
414 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
417 for (i = 0; i < 10; i++)
418 for (j = 0; j < 7; j++){
419 if (!bitalloc_codes[i][j]) break;
420 dca_smpl_bitalloc[i+1].offset = bitalloc_offsets[i];
421 dca_smpl_bitalloc[i+1].wrap = 1 + (j > 4);
422 dca_smpl_bitalloc[i+1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
423 dca_smpl_bitalloc[i+1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
424 init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j],
426 bitalloc_bits[i][j], 1, 1,
427 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
430 vlcs_initialized = 1;
433 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
436 *dst++ = get_bits(gb, bits);
439 static int dca_parse_audio_coding_header(DCAContext * s, int base_channel)
442 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
443 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
444 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
446 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
447 s->prim_channels = s->total_channels;
449 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
450 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
453 for (i = base_channel; i < s->prim_channels; i++) {
454 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
455 if (s->subband_activity[i] > DCA_SUBBANDS)
456 s->subband_activity[i] = DCA_SUBBANDS;
458 for (i = base_channel; i < s->prim_channels; i++) {
459 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
460 if (s->vq_start_subband[i] > DCA_SUBBANDS)
461 s->vq_start_subband[i] = DCA_SUBBANDS;
463 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
464 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
465 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
466 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
468 /* Get codebooks quantization indexes */
470 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
471 for (j = 1; j < 11; j++)
472 for (i = base_channel; i < s->prim_channels; i++)
473 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
475 /* Get scale factor adjustment */
476 for (j = 0; j < 11; j++)
477 for (i = base_channel; i < s->prim_channels; i++)
478 s->scalefactor_adj[i][j] = 1;
480 for (j = 1; j < 11; j++)
481 for (i = base_channel; i < s->prim_channels; i++)
482 if (s->quant_index_huffman[i][j] < thr[j])
483 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
485 if (s->crc_present) {
486 /* Audio header CRC check */
487 get_bits(&s->gb, 16);
490 s->current_subframe = 0;
491 s->current_subsubframe = 0;
494 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
495 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
496 for (i = base_channel; i < s->prim_channels; i++){
497 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", s->subband_activity[i]);
498 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", s->vq_start_subband[i]);
499 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", s->joint_intensity[i]);
500 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", s->transient_huffman[i]);
501 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", s->scalefactor_huffman[i]);
502 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", s->bitalloc_huffman[i]);
503 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
504 for (j = 0; j < 11; j++)
505 av_log(s->avctx, AV_LOG_DEBUG, " %i",
506 s->quant_index_huffman[i][j]);
507 av_log(s->avctx, AV_LOG_DEBUG, "\n");
508 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
509 for (j = 0; j < 11; j++)
510 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
511 av_log(s->avctx, AV_LOG_DEBUG, "\n");
518 static int dca_parse_frame_header(DCAContext * s)
520 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
523 get_bits(&s->gb, 32);
526 s->frame_type = get_bits(&s->gb, 1);
527 s->samples_deficit = get_bits(&s->gb, 5) + 1;
528 s->crc_present = get_bits(&s->gb, 1);
529 s->sample_blocks = get_bits(&s->gb, 7) + 1;
530 s->frame_size = get_bits(&s->gb, 14) + 1;
531 if (s->frame_size < 95)
533 s->amode = get_bits(&s->gb, 6);
534 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
537 s->bit_rate_index = get_bits(&s->gb, 5);
538 s->bit_rate = dca_bit_rates[s->bit_rate_index];
542 s->downmix = get_bits(&s->gb, 1);
543 s->dynrange = get_bits(&s->gb, 1);
544 s->timestamp = get_bits(&s->gb, 1);
545 s->aux_data = get_bits(&s->gb, 1);
546 s->hdcd = get_bits(&s->gb, 1);
547 s->ext_descr = get_bits(&s->gb, 3);
548 s->ext_coding = get_bits(&s->gb, 1);
549 s->aspf = get_bits(&s->gb, 1);
550 s->lfe = get_bits(&s->gb, 2);
551 s->predictor_history = get_bits(&s->gb, 1);
553 /* TODO: check CRC */
555 s->header_crc = get_bits(&s->gb, 16);
557 s->multirate_inter = get_bits(&s->gb, 1);
558 s->version = get_bits(&s->gb, 4);
559 s->copy_history = get_bits(&s->gb, 2);
560 s->source_pcm_res = get_bits(&s->gb, 3);
561 s->front_sum = get_bits(&s->gb, 1);
562 s->surround_sum = get_bits(&s->gb, 1);
563 s->dialog_norm = get_bits(&s->gb, 4);
565 /* FIXME: channels mixing levels */
566 s->output = s->amode;
567 if (s->lfe) s->output |= DCA_LFE;
570 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
571 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
572 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
573 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
574 s->sample_blocks, s->sample_blocks * 32);
575 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
576 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
577 s->amode, dca_channels[s->amode]);
578 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
580 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
582 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
583 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
584 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
585 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
586 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
587 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
588 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
589 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
590 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
591 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
592 s->predictor_history);
593 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
594 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
596 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
597 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
598 av_log(s->avctx, AV_LOG_DEBUG,
599 "source pcm resolution: %i (%i bits/sample)\n",
600 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
601 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
602 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
603 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
604 av_log(s->avctx, AV_LOG_DEBUG, "\n");
607 /* Primary audio coding header */
608 s->subframes = get_bits(&s->gb, 4) + 1;
610 return dca_parse_audio_coding_header(s, 0);
614 static inline int get_scale(GetBitContext *gb, int level, int value)
617 /* huffman encoded */
618 value += get_bitalloc(gb, &dca_scalefactor, level);
619 } else if (level < 8)
620 value = get_bits(gb, level + 1);
624 static int dca_subframe_header(DCAContext * s, int base_channel, int block_index)
626 /* Primary audio coding side information */
629 if (get_bits_left(&s->gb) < 0)
633 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
634 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
637 for (j = base_channel; j < s->prim_channels; j++) {
638 for (k = 0; k < s->subband_activity[j]; k++)
639 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
642 /* Get prediction codebook */
643 for (j = base_channel; j < s->prim_channels; j++) {
644 for (k = 0; k < s->subband_activity[j]; k++) {
645 if (s->prediction_mode[j][k] > 0) {
646 /* (Prediction coefficient VQ address) */
647 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
652 /* Bit allocation index */
653 for (j = base_channel; j < s->prim_channels; j++) {
654 for (k = 0; k < s->vq_start_subband[j]; k++) {
655 if (s->bitalloc_huffman[j] == 6)
656 s->bitalloc[j][k] = get_bits(&s->gb, 5);
657 else if (s->bitalloc_huffman[j] == 5)
658 s->bitalloc[j][k] = get_bits(&s->gb, 4);
659 else if (s->bitalloc_huffman[j] == 7) {
660 av_log(s->avctx, AV_LOG_ERROR,
661 "Invalid bit allocation index\n");
665 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
668 if (s->bitalloc[j][k] > 26) {
669 // av_log(s->avctx,AV_LOG_DEBUG,"bitalloc index [%i][%i] too big (%i)\n",
670 // j, k, s->bitalloc[j][k]);
676 /* Transition mode */
677 for (j = base_channel; j < s->prim_channels; j++) {
678 for (k = 0; k < s->subband_activity[j]; k++) {
679 s->transition_mode[j][k] = 0;
680 if (s->subsubframes[s->current_subframe] > 1 &&
681 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
682 s->transition_mode[j][k] =
683 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
688 if (get_bits_left(&s->gb) < 0)
691 for (j = base_channel; j < s->prim_channels; j++) {
692 const uint32_t *scale_table;
695 memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
697 if (s->scalefactor_huffman[j] == 6)
698 scale_table = scale_factor_quant7;
700 scale_table = scale_factor_quant6;
702 /* When huffman coded, only the difference is encoded */
705 for (k = 0; k < s->subband_activity[j]; k++) {
706 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
707 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
708 s->scale_factor[j][k][0] = scale_table[scale_sum];
711 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
712 /* Get second scale factor */
713 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
714 s->scale_factor[j][k][1] = scale_table[scale_sum];
719 /* Joint subband scale factor codebook select */
720 for (j = base_channel; j < s->prim_channels; j++) {
721 /* Transmitted only if joint subband coding enabled */
722 if (s->joint_intensity[j] > 0)
723 s->joint_huff[j] = get_bits(&s->gb, 3);
726 if (get_bits_left(&s->gb) < 0)
729 /* Scale factors for joint subband coding */
730 for (j = base_channel; j < s->prim_channels; j++) {
733 /* Transmitted only if joint subband coding enabled */
734 if (s->joint_intensity[j] > 0) {
736 source_channel = s->joint_intensity[j] - 1;
738 /* When huffman coded, only the difference is encoded
739 * (is this valid as well for joint scales ???) */
741 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
742 scale = get_scale(&s->gb, s->joint_huff[j], 0);
743 scale += 64; /* bias */
744 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
747 if (!(s->debug_flag & 0x02)) {
748 av_log(s->avctx, AV_LOG_DEBUG,
749 "Joint stereo coding not supported\n");
750 s->debug_flag |= 0x02;
755 /* Stereo downmix coefficients */
756 if (!base_channel && s->prim_channels > 2) {
758 for (j = base_channel; j < s->prim_channels; j++) {
759 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
760 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
763 int am = s->amode & DCA_CHANNEL_MASK;
764 for (j = base_channel; j < s->prim_channels; j++) {
765 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
766 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
771 /* Dynamic range coefficient */
772 if (!base_channel && s->dynrange)
773 s->dynrange_coef = get_bits(&s->gb, 8);
775 /* Side information CRC check word */
776 if (s->crc_present) {
777 get_bits(&s->gb, 16);
781 * Primary audio data arrays
784 /* VQ encoded high frequency subbands */
785 for (j = base_channel; j < s->prim_channels; j++)
786 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
787 /* 1 vector -> 32 samples */
788 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
790 /* Low frequency effect data */
791 if (!base_channel && s->lfe) {
793 int lfe_samples = 2 * s->lfe * (4 + block_index);
794 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
797 for (j = lfe_samples; j < lfe_end_sample; j++) {
798 /* Signed 8 bits int */
799 s->lfe_data[j] = get_sbits(&s->gb, 8);
802 /* Scale factor index */
803 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
805 /* Quantization step size * scale factor */
806 lfe_scale = 0.035 * s->lfe_scale_factor;
808 for (j = lfe_samples; j < lfe_end_sample; j++)
809 s->lfe_data[j] *= lfe_scale;
813 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", s->subsubframes[s->current_subframe]);
814 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
815 s->partial_samples[s->current_subframe]);
816 for (j = base_channel; j < s->prim_channels; j++) {
817 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
818 for (k = 0; k < s->subband_activity[j]; k++)
819 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
820 av_log(s->avctx, AV_LOG_DEBUG, "\n");
822 for (j = base_channel; j < s->prim_channels; j++) {
823 for (k = 0; k < s->subband_activity[j]; k++)
824 av_log(s->avctx, AV_LOG_DEBUG,
825 "prediction coefs: %f, %f, %f, %f\n",
826 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
827 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
828 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
829 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
831 for (j = base_channel; j < s->prim_channels; j++) {
832 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
833 for (k = 0; k < s->vq_start_subband[j]; k++)
834 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
835 av_log(s->avctx, AV_LOG_DEBUG, "\n");
837 for (j = base_channel; j < s->prim_channels; j++) {
838 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
839 for (k = 0; k < s->subband_activity[j]; k++)
840 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
841 av_log(s->avctx, AV_LOG_DEBUG, "\n");
843 for (j = base_channel; j < s->prim_channels; j++) {
844 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
845 for (k = 0; k < s->subband_activity[j]; k++) {
846 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
847 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
848 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
849 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
851 av_log(s->avctx, AV_LOG_DEBUG, "\n");
853 for (j = base_channel; j < s->prim_channels; j++) {
854 if (s->joint_intensity[j] > 0) {
855 int source_channel = s->joint_intensity[j] - 1;
856 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
857 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
858 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
859 av_log(s->avctx, AV_LOG_DEBUG, "\n");
862 if (!base_channel && s->prim_channels > 2 && s->downmix) {
863 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
864 for (j = 0; j < s->prim_channels; j++) {
865 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][0]]);
866 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][1]]);
868 av_log(s->avctx, AV_LOG_DEBUG, "\n");
870 for (j = base_channel; j < s->prim_channels; j++)
871 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
872 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
873 if (!base_channel && s->lfe) {
874 int lfe_samples = 2 * s->lfe * (4 + block_index);
875 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
877 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
878 for (j = lfe_samples; j < lfe_end_sample; j++)
879 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
880 av_log(s->avctx, AV_LOG_DEBUG, "\n");
887 static void qmf_32_subbands(DCAContext * s, int chans,
888 float samples_in[32][8], float *samples_out,
891 const float *prCoeff;
894 int sb_act = s->subband_activity[chans];
897 scale *= sqrt(1/8.0);
900 if (!s->multirate_inter) /* Non-perfect reconstruction */
901 prCoeff = fir_32bands_nonperfect;
902 else /* Perfect reconstruction */
903 prCoeff = fir_32bands_perfect;
905 for (i = sb_act; i < 32; i++)
908 /* Reconstructed channel sample index */
909 for (subindex = 0; subindex < 8; subindex++) {
910 /* Load in one sample from each subband and clear inactive subbands */
911 for (i = 0; i < sb_act; i++){
912 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ ((i-1)&2)<<30;
913 AV_WN32A(&s->raXin[i], v);
916 s->synth.synth_filter_float(&s->imdct,
917 s->subband_fir_hist[chans], &s->hist_index[chans],
918 s->subband_fir_noidea[chans], prCoeff,
919 samples_out, s->raXin, scale);
925 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
926 int num_deci_sample, float *samples_in,
927 float *samples_out, float scale)
929 /* samples_in: An array holding decimated samples.
930 * Samples in current subframe starts from samples_in[0],
931 * while samples_in[-1], samples_in[-2], ..., stores samples
932 * from last subframe as history.
934 * samples_out: An array holding interpolated samples
938 const float *prCoeff;
941 /* Select decimation filter */
942 if (decimation_select == 1) {
944 prCoeff = lfe_fir_128;
947 prCoeff = lfe_fir_64;
950 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
951 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor,
954 samples_out += 2 * decifactor;
958 /* downmixing routines */
959 #define MIX_REAR1(samples, si1, rs, coef) \
960 samples[i] += samples[si1] * coef[rs][0]; \
961 samples[i+256] += samples[si1] * coef[rs][1];
963 #define MIX_REAR2(samples, si1, si2, rs, coef) \
964 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs+1][0]; \
965 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs+1][1];
967 #define MIX_FRONT3(samples, coef) \
971 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
972 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
974 #define DOWNMIX_TO_STEREO(op1, op2) \
975 for (i = 0; i < 256; i++){ \
980 static void dca_downmix(float *samples, int srcfmt,
981 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
982 const int8_t *channel_mapping)
987 float coef[DCA_PRIM_CHANNELS_MAX][2];
989 for (i=0; i<DCA_PRIM_CHANNELS_MAX; i++) {
990 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
991 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
997 case DCA_STEREO_TOTAL:
998 case DCA_STEREO_SUMDIFF:
1000 av_log(NULL, 0, "Not implemented!\n");
1005 c = channel_mapping[0] * 256;
1006 l = channel_mapping[1] * 256;
1007 r = channel_mapping[2] * 256;
1008 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),);
1011 s = channel_mapping[2] * 256;
1012 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef),);
1015 c = channel_mapping[0] * 256;
1016 l = channel_mapping[1] * 256;
1017 r = channel_mapping[2] * 256;
1018 s = channel_mapping[3] * 256;
1019 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1020 MIX_REAR1(samples, i + s, 3, coef));
1023 sl = channel_mapping[2] * 256;
1024 sr = channel_mapping[3] * 256;
1025 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef),);
1028 c = channel_mapping[0] * 256;
1029 l = channel_mapping[1] * 256;
1030 r = channel_mapping[2] * 256;
1031 sl = channel_mapping[3] * 256;
1032 sr = channel_mapping[4] * 256;
1033 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1034 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1040 /* Very compact version of the block code decoder that does not use table
1041 * look-up but is slightly slower */
1042 static int decode_blockcode(int code, int levels, int *values)
1045 int offset = (levels - 1) >> 1;
1047 for (i = 0; i < 4; i++) {
1048 int div = FASTDIV(code, levels);
1049 values[i] = code - offset - div*levels;
1056 av_log(NULL, AV_LOG_ERROR, "ERROR: block code look-up failed\n");
1061 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1062 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1064 #ifndef int8x8_fmul_int32
1065 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1067 float fscale = scale / 16.0;
1069 for (i = 0; i < 8; i++)
1070 dst[i] = src[i] * fscale;
1074 static int dca_subsubframe(DCAContext * s, int base_channel, int block_index)
1077 int subsubframe = s->current_subsubframe;
1079 const float *quant_step_table;
1082 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1083 LOCAL_ALIGNED_16(int, block, [8]);
1089 /* Select quantization step size table */
1090 if (s->bit_rate_index == 0x1f)
1091 quant_step_table = lossless_quant_d;
1093 quant_step_table = lossy_quant_d;
1095 for (k = base_channel; k < s->prim_channels; k++) {
1096 if (get_bits_left(&s->gb) < 0)
1099 for (l = 0; l < s->vq_start_subband[k]; l++) {
1102 /* Select the mid-tread linear quantizer */
1103 int abits = s->bitalloc[k][l];
1105 float quant_step_size = quant_step_table[abits];
1108 * Determine quantization index code book and its type
1111 /* Select quantization index code book */
1112 int sel = s->quant_index_huffman[k][abits];
1115 * Extract bits from the bit stream
1118 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1120 /* Deal with transients */
1121 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1122 float rscale = quant_step_size * s->scale_factor[k][l][sfi] * s->scalefactor_adj[k][sel];
1124 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){
1127 int block_code1, block_code2, size, levels;
1129 size = abits_sizes[abits-1];
1130 levels = abits_levels[abits-1];
1132 block_code1 = get_bits(&s->gb, size);
1133 /* FIXME Should test return value */
1134 decode_blockcode(block_code1, levels, block);
1135 block_code2 = get_bits(&s->gb, size);
1136 decode_blockcode(block_code2, levels, &block[4]);
1139 for (m = 0; m < 8; m++)
1140 block[m] = get_sbits(&s->gb, abits - 3);
1144 for (m = 0; m < 8; m++)
1145 block[m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel);
1148 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1153 * Inverse ADPCM if in prediction mode
1155 if (s->prediction_mode[k][l]) {
1157 for (m = 0; m < 8; m++) {
1158 for (n = 1; n <= 4; n++)
1160 subband_samples[k][l][m] +=
1161 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1162 subband_samples[k][l][m - n] / 8192);
1163 else if (s->predictor_history)
1164 subband_samples[k][l][m] +=
1165 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1166 s->subband_samples_hist[k][l][m - n +
1173 * Decode VQ encoded high frequencies
1175 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1176 /* 1 vector -> 32 samples but we only need the 8 samples
1177 * for this subsubframe. */
1178 int hfvq = s->high_freq_vq[k][l];
1180 if (!s->debug_flag & 0x01) {
1181 av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
1182 s->debug_flag |= 0x01;
1185 int8x8_fmul_int32(subband_samples[k][l],
1186 &high_freq_vq[hfvq][subsubframe * 8],
1187 s->scale_factor[k][l][0]);
1191 /* Check for DSYNC after subsubframe */
1192 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1193 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1195 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1198 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1202 /* Backup predictor history for adpcm */
1203 for (k = base_channel; k < s->prim_channels; k++)
1204 for (l = 0; l < s->vq_start_subband[k]; l++)
1205 memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
1206 4 * sizeof(subband_samples[0][0][0]));
1211 static int dca_filter_channels(DCAContext * s, int block_index)
1213 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1216 /* 32 subbands QMF */
1217 for (k = 0; k < s->prim_channels; k++) {
1218 /* static float pcm_to_double[8] =
1219 {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
1220 qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
1221 M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ );
1225 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1226 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1229 /* Generate LFE samples for this subsubframe FIXME!!! */
1230 if (s->output & DCA_LFE) {
1231 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1232 s->lfe_data + 2 * s->lfe * (block_index + 4),
1233 &s->samples[256 * dca_lfe_index[s->amode]],
1234 (1.0/256.0)*s->scale_bias);
1235 /* Outputs 20bits pcm samples */
1242 static int dca_subframe_footer(DCAContext * s, int base_channel)
1244 int aux_data_count = 0, i;
1247 * Unpack optional information
1250 /* presumably optional information only appears in the core? */
1251 if (!base_channel) {
1253 get_bits(&s->gb, 32);
1256 aux_data_count = get_bits(&s->gb, 6);
1258 for (i = 0; i < aux_data_count; i++)
1259 get_bits(&s->gb, 8);
1261 if (s->crc_present && (s->downmix || s->dynrange))
1262 get_bits(&s->gb, 16);
1269 * Decode a dca frame block
1271 * @param s pointer to the DCAContext
1274 static int dca_decode_block(DCAContext * s, int base_channel, int block_index)
1278 if (s->current_subframe >= s->subframes) {
1279 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1280 s->current_subframe, s->subframes);
1284 if (!s->current_subsubframe) {
1286 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1288 /* Read subframe header */
1289 if (dca_subframe_header(s, base_channel, block_index))
1293 /* Read subsubframe */
1295 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1297 if (dca_subsubframe(s, base_channel, block_index))
1301 s->current_subsubframe++;
1302 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1303 s->current_subsubframe = 0;
1304 s->current_subframe++;
1306 if (s->current_subframe >= s->subframes) {
1308 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1310 /* Read subframe footer */
1311 if (dca_subframe_footer(s, base_channel))
1319 * Convert bitstream to one representation based on sync marker
1321 static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst,
1326 const uint16_t *ssrc = (const uint16_t *) src;
1327 uint16_t *sdst = (uint16_t *) dst;
1330 if ((unsigned)src_size > (unsigned)max_size) {
1331 // av_log(NULL, AV_LOG_ERROR, "Input frame size larger than DCA_MAX_FRAME_SIZE!\n");
1333 src_size = max_size;
1338 case DCA_MARKER_RAW_BE:
1339 memcpy(dst, src, src_size);
1341 case DCA_MARKER_RAW_LE:
1342 for (i = 0; i < (src_size + 1) >> 1; i++)
1343 *sdst++ = av_bswap16(*ssrc++);
1345 case DCA_MARKER_14B_BE:
1346 case DCA_MARKER_14B_LE:
1347 init_put_bits(&pb, dst, max_size);
1348 for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) {
1349 tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
1350 put_bits(&pb, 14, tmp);
1352 flush_put_bits(&pb);
1353 return (put_bits_count(&pb) + 7) >> 3;
1360 * Return the number of channels in an ExSS speaker mask (HD)
1362 static int dca_exss_mask2count(int mask)
1364 /* count bits that mean speaker pairs twice */
1365 return av_popcount(mask)
1366 + av_popcount(mask & (
1367 DCA_EXSS_CENTER_LEFT_RIGHT
1368 | DCA_EXSS_FRONT_LEFT_RIGHT
1369 | DCA_EXSS_FRONT_HIGH_LEFT_RIGHT
1370 | DCA_EXSS_WIDE_LEFT_RIGHT
1371 | DCA_EXSS_SIDE_LEFT_RIGHT
1372 | DCA_EXSS_SIDE_HIGH_LEFT_RIGHT
1373 | DCA_EXSS_SIDE_REAR_LEFT_RIGHT
1374 | DCA_EXSS_REAR_LEFT_RIGHT
1375 | DCA_EXSS_REAR_HIGH_LEFT_RIGHT
1380 * Skip mixing coefficients of a single mix out configuration (HD)
1382 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1386 for (i = 0; i < channels; i++) {
1387 int mix_map_mask = get_bits(gb, out_ch);
1388 int num_coeffs = av_popcount(mix_map_mask);
1389 skip_bits_long(gb, num_coeffs * 6);
1394 * Parse extension substream asset header (HD)
1396 static int dca_exss_parse_asset_header(DCAContext *s)
1398 int header_pos = get_bits_count(&s->gb);
1401 int embedded_stereo = 0;
1402 int embedded_6ch = 0;
1403 int drc_code_present;
1404 int extensions_mask;
1407 if (get_bits_left(&s->gb) < 16)
1410 /* We will parse just enough to get to the extensions bitmask with which
1411 * we can set the profile value. */
1413 header_size = get_bits(&s->gb, 9) + 1;
1414 skip_bits(&s->gb, 3); // asset index
1416 if (s->static_fields) {
1417 if (get_bits1(&s->gb))
1418 skip_bits(&s->gb, 4); // asset type descriptor
1419 if (get_bits1(&s->gb))
1420 skip_bits_long(&s->gb, 24); // language descriptor
1422 if (get_bits1(&s->gb)) {
1423 /* How can one fit 1024 bytes of text here if the maximum value
1424 * for the asset header size field above was 512 bytes? */
1425 int text_length = get_bits(&s->gb, 10) + 1;
1426 if (get_bits_left(&s->gb) < text_length * 8)
1428 skip_bits_long(&s->gb, text_length * 8); // info text
1431 skip_bits(&s->gb, 5); // bit resolution - 1
1432 skip_bits(&s->gb, 4); // max sample rate code
1433 channels = get_bits(&s->gb, 8) + 1;
1435 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1436 int spkr_remap_sets;
1437 int spkr_mask_size = 16;
1441 embedded_stereo = get_bits1(&s->gb);
1443 embedded_6ch = get_bits1(&s->gb);
1445 if (get_bits1(&s->gb)) {
1446 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1447 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1450 spkr_remap_sets = get_bits(&s->gb, 3);
1452 for (i = 0; i < spkr_remap_sets; i++) {
1453 /* std layout mask for each remap set */
1454 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1457 for (i = 0; i < spkr_remap_sets; i++) {
1458 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1459 if (get_bits_left(&s->gb) < 0)
1462 for (j = 0; j < num_spkrs[i]; j++) {
1463 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1464 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1465 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1470 skip_bits(&s->gb, 3); // representation type
1474 drc_code_present = get_bits1(&s->gb);
1475 if (drc_code_present)
1476 get_bits(&s->gb, 8); // drc code
1478 if (get_bits1(&s->gb))
1479 skip_bits(&s->gb, 5); // dialog normalization code
1481 if (drc_code_present && embedded_stereo)
1482 get_bits(&s->gb, 8); // drc stereo code
1484 if (s->mix_metadata && get_bits1(&s->gb)) {
1485 skip_bits(&s->gb, 1); // external mix
1486 skip_bits(&s->gb, 6); // post mix gain code
1488 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1489 skip_bits(&s->gb, 3); // drc limit
1491 skip_bits(&s->gb, 8); // custom drc code
1493 if (get_bits1(&s->gb)) // channel specific scaling
1494 for (i = 0; i < s->num_mix_configs; i++)
1495 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1497 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1499 for (i = 0; i < s->num_mix_configs; i++) {
1500 if (get_bits_left(&s->gb) < 0)
1502 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1504 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1505 if (embedded_stereo)
1506 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1510 switch (get_bits(&s->gb, 2)) {
1511 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1512 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1513 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1514 case 3: extensions_mask = 0; /* aux coding */ break;
1517 /* not parsed further, we were only interested in the extensions mask */
1519 if (get_bits_left(&s->gb) < 0)
1522 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1523 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1526 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1528 if (extensions_mask & DCA_EXT_EXSS_XLL)
1529 s->profile = FF_PROFILE_DTS_HD_MA;
1530 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1532 s->profile = FF_PROFILE_DTS_HD_HRA;
1534 if (!(extensions_mask & DCA_EXT_CORE))
1535 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1536 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1537 av_log(s->avctx, AV_LOG_WARNING, "DTS extensions detection mismatch (%d, %d)\n",
1538 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1544 * Parse extension substream header (HD)
1546 static void dca_exss_parse_header(DCAContext *s)
1552 int active_ss_mask[8];
1555 if (get_bits_left(&s->gb) < 52)
1558 skip_bits(&s->gb, 8); // user data
1559 ss_index = get_bits(&s->gb, 2);
1561 blownup = get_bits1(&s->gb);
1562 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1563 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1565 s->static_fields = get_bits1(&s->gb);
1566 if (s->static_fields) {
1567 skip_bits(&s->gb, 2); // reference clock code
1568 skip_bits(&s->gb, 3); // frame duration code
1570 if (get_bits1(&s->gb))
1571 skip_bits_long(&s->gb, 36); // timestamp
1573 /* a single stream can contain multiple audio assets that can be
1574 * combined to form multiple audio presentations */
1576 num_audiop = get_bits(&s->gb, 3) + 1;
1577 if (num_audiop > 1) {
1578 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1579 /* ignore such streams for now */
1583 num_assets = get_bits(&s->gb, 3) + 1;
1584 if (num_assets > 1) {
1585 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1586 /* ignore such streams for now */
1590 for (i = 0; i < num_audiop; i++)
1591 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1593 for (i = 0; i < num_audiop; i++)
1594 for (j = 0; j <= ss_index; j++)
1595 if (active_ss_mask[i] & (1 << j))
1596 skip_bits(&s->gb, 8); // active asset mask
1598 s->mix_metadata = get_bits1(&s->gb);
1599 if (s->mix_metadata) {
1600 int mix_out_mask_size;
1602 skip_bits(&s->gb, 2); // adjustment level
1603 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1604 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1606 for (i = 0; i < s->num_mix_configs; i++) {
1607 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1608 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1613 for (i = 0; i < num_assets; i++)
1614 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1616 for (i = 0; i < num_assets; i++) {
1617 if (dca_exss_parse_asset_header(s))
1621 /* not parsed further, we were only interested in the extensions mask
1622 * from the asset header */
1626 * Main frame decoding function
1627 * FIXME add arguments
1629 static int dca_decode_frame(AVCodecContext * avctx,
1630 void *data, int *data_size,
1633 const uint8_t *buf = avpkt->data;
1634 int buf_size = avpkt->size;
1637 int num_core_channels = 0;
1639 float *samples_flt = data;
1640 int16_t *samples_s16 = data;
1642 DCAContext *s = avctx->priv_data;
1649 s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1650 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1651 if (s->dca_buffer_size == -1) {
1652 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1656 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1657 if (dca_parse_frame_header(s) < 0) {
1658 //seems like the frame is corrupt, try with the next one
1662 //set AVCodec values with parsed data
1663 avctx->sample_rate = s->sample_rate;
1664 avctx->bit_rate = s->bit_rate;
1665 avctx->frame_size = s->sample_blocks * 32;
1667 s->profile = FF_PROFILE_DTS;
1669 for (i = 0; i < (s->sample_blocks / 8); i++) {
1670 dca_decode_block(s, 0, i);
1673 /* record number of core channels incase less than max channels are requested */
1674 num_core_channels = s->prim_channels;
1677 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1679 s->core_ext_mask = 0;
1681 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1683 /* only scan for extensions if ext_descr was unknown or indicated a
1684 * supported XCh extension */
1685 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1687 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1688 * extensions scan can fill it up */
1689 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1691 /* extensions start at 32-bit boundaries into bitstream */
1692 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1694 while(core_ss_end - get_bits_count(&s->gb) >= 32) {
1695 uint32_t bits = get_bits_long(&s->gb, 32);
1699 int ext_amode, xch_fsize;
1701 s->xch_base_channel = s->prim_channels;
1703 /* validate sync word using XCHFSIZE field */
1704 xch_fsize = show_bits(&s->gb, 10);
1705 if((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1706 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1709 /* skip length-to-end-of-frame field for the moment */
1710 skip_bits(&s->gb, 10);
1712 s->core_ext_mask |= DCA_EXT_XCH;
1714 /* extension amode should == 1, number of channels in extension */
1715 /* AFAIK XCh is not used for more channels */
1716 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1717 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1718 " supported!\n",ext_amode);
1722 /* much like core primary audio coding header */
1723 dca_parse_audio_coding_header(s, s->xch_base_channel);
1725 for (i = 0; i < (s->sample_blocks / 8); i++) {
1726 dca_decode_block(s, s->xch_base_channel, i);
1733 /* XXCh: extended channels */
1734 /* usually found either in core or HD part in DTS-HD HRA streams,
1735 * but not in DTS-ES which contains XCh extensions instead */
1736 s->core_ext_mask |= DCA_EXT_XXCH;
1740 int fsize96 = show_bits(&s->gb, 12) + 1;
1741 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1744 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n", get_bits_count(&s->gb));
1745 skip_bits(&s->gb, 12);
1746 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1747 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1749 s->core_ext_mask |= DCA_EXT_X96;
1754 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1758 /* no supported extensions, skip the rest of the core substream */
1759 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1762 if (s->core_ext_mask & DCA_EXT_X96)
1763 s->profile = FF_PROFILE_DTS_96_24;
1764 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1765 s->profile = FF_PROFILE_DTS_ES;
1767 /* check for ExSS (HD part) */
1768 if (s->dca_buffer_size - s->frame_size > 32
1769 && get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1770 dca_exss_parse_header(s);
1772 avctx->profile = s->profile;
1774 channels = s->prim_channels + !!s->lfe;
1777 avctx->channel_layout = dca_core_channel_layout[s->amode];
1779 if (s->xch_present && (!avctx->request_channels ||
1780 avctx->request_channels > num_core_channels + !!s->lfe)) {
1781 avctx->channel_layout |= AV_CH_BACK_CENTER;
1783 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1784 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1786 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1789 channels = num_core_channels + !!s->lfe;
1790 s->xch_present = 0; /* disable further xch processing */
1792 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1793 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1795 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1798 if (channels > !!s->lfe &&
1799 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1802 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1804 s->output = DCA_STEREO;
1805 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1807 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
1808 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
1809 s->channel_order_tab = dca_channel_order_native;
1812 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
1817 /* There is nothing that prevents a dts frame to change channel configuration
1818 but FFmpeg doesn't support that so only set the channels if it is previously
1819 unset. Ideally during the first probe for channels the crc should be checked
1820 and only set avctx->channels when the crc is ok. Right now the decoder could
1821 set the channels based on a broken first frame.*/
1822 if (s->is_channels_set == 0) {
1823 s->is_channels_set = 1;
1824 avctx->channels = channels;
1826 if (avctx->channels != channels) {
1827 av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of "
1828 "channels changing in stream. Skipping frame.\n");
1832 out_size = 256 / 8 * s->sample_blocks * channels *
1833 av_get_bytes_per_sample(avctx->sample_fmt);
1834 if (*data_size < out_size)
1836 *data_size = out_size;
1838 /* filter to get final output */
1839 for (i = 0; i < (s->sample_blocks / 8); i++) {
1840 dca_filter_channels(s, i);
1842 /* If this was marked as a DTS-ES stream we need to subtract back- */
1843 /* channel from SL & SR to remove matrixed back-channel signal */
1844 if((s->source_pcm_res & 1) && s->xch_present) {
1845 float* back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1846 float* lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1847 float* rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1848 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1849 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1852 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1853 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1855 samples_flt += 256 * channels;
1857 s->fmt_conv.float_to_int16_interleave(samples_s16,
1858 s->samples_chanptr, 256,
1860 samples_s16 += 256 * channels;
1864 /* update lfe history */
1865 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1866 for (i = 0; i < 2 * s->lfe * 4; i++) {
1867 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1876 * DCA initialization
1878 * @param avctx pointer to the AVCodecContext
1881 static av_cold int dca_decode_init(AVCodecContext * avctx)
1883 DCAContext *s = avctx->priv_data;
1889 dsputil_init(&s->dsp, avctx);
1890 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1891 ff_synth_filter_init(&s->synth);
1892 ff_dcadsp_init(&s->dcadsp);
1893 ff_fmt_convert_init(&s->fmt_conv, avctx);
1895 for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++)
1896 s->samples_chanptr[i] = s->samples + i * 256;
1898 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1899 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1900 s->scale_bias = 1.0 / 32768.0;
1902 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1903 s->scale_bias = 1.0;
1906 /* allow downmixing to stereo */
1907 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1908 avctx->request_channels == 2) {
1909 avctx->channels = avctx->request_channels;
1915 static av_cold int dca_decode_end(AVCodecContext * avctx)
1917 DCAContext *s = avctx->priv_data;
1918 ff_mdct_end(&s->imdct);
1922 static const AVProfile profiles[] = {
1923 { FF_PROFILE_DTS, "DTS" },
1924 { FF_PROFILE_DTS_ES, "DTS-ES" },
1925 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1926 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1927 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1928 { FF_PROFILE_UNKNOWN },
1931 AVCodec ff_dca_decoder = {
1933 .type = AVMEDIA_TYPE_AUDIO,
1935 .priv_data_size = sizeof(DCAContext),
1936 .init = dca_decode_init,
1937 .decode = dca_decode_frame,
1938 .close = dca_decode_end,
1939 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1940 .capabilities = CODEC_CAP_CHANNEL_CONF,
1941 .sample_fmts = (const enum AVSampleFormat[]) {
1942 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
1944 .profiles = NULL_IF_CONFIG_SMALL(profiles),