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 uint64_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;
266 int frame_type; ///< type of the current frame
267 int samples_deficit; ///< deficit sample count
268 int crc_present; ///< crc is present in the bitstream
269 int sample_blocks; ///< number of PCM sample blocks
270 int frame_size; ///< primary frame byte size
271 int amode; ///< audio channels arrangement
272 int sample_rate; ///< audio sampling rate
273 int bit_rate; ///< transmission bit rate
274 int bit_rate_index; ///< transmission bit rate index
276 int downmix; ///< embedded downmix enabled
277 int dynrange; ///< embedded dynamic range flag
278 int timestamp; ///< embedded time stamp flag
279 int aux_data; ///< auxiliary data flag
280 int hdcd; ///< source material is mastered in HDCD
281 int ext_descr; ///< extension audio descriptor flag
282 int ext_coding; ///< extended coding flag
283 int aspf; ///< audio sync word insertion flag
284 int lfe; ///< low frequency effects flag
285 int predictor_history; ///< predictor history flag
286 int header_crc; ///< header crc check bytes
287 int multirate_inter; ///< multirate interpolator switch
288 int version; ///< encoder software revision
289 int copy_history; ///< copy history
290 int source_pcm_res; ///< source pcm resolution
291 int front_sum; ///< front sum/difference flag
292 int surround_sum; ///< surround sum/difference flag
293 int dialog_norm; ///< dialog normalisation parameter
295 /* Primary audio coding header */
296 int subframes; ///< number of subframes
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 skip_bits_long(&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)
532 return AVERROR_INVALIDDATA;
533 s->amode = get_bits(&s->gb, 6);
534 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
536 return AVERROR_INVALIDDATA;
537 s->bit_rate_index = get_bits(&s->gb, 5);
538 s->bit_rate = dca_bit_rates[s->bit_rate_index];
540 return AVERROR_INVALIDDATA;
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)
630 return AVERROR_INVALIDDATA;
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");
662 return AVERROR_INVALIDDATA;
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]);
671 return AVERROR_INVALIDDATA;
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)
689 return AVERROR_INVALIDDATA;
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)
727 return AVERROR_INVALIDDATA;
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 unsigned sign = (i - 1) & 2;
913 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
914 AV_WN32A(&s->raXin[i], v);
917 s->synth.synth_filter_float(&s->imdct,
918 s->subband_fir_hist[chans], &s->hist_index[chans],
919 s->subband_fir_noidea[chans], prCoeff,
920 samples_out, s->raXin, scale);
926 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
927 int num_deci_sample, float *samples_in,
928 float *samples_out, float scale)
930 /* samples_in: An array holding decimated samples.
931 * Samples in current subframe starts from samples_in[0],
932 * while samples_in[-1], samples_in[-2], ..., stores samples
933 * from last subframe as history.
935 * samples_out: An array holding interpolated samples
939 const float *prCoeff;
942 /* Select decimation filter */
943 if (decimation_select == 1) {
945 prCoeff = lfe_fir_128;
948 prCoeff = lfe_fir_64;
951 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
952 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor,
955 samples_out += 2 * decifactor;
959 /* downmixing routines */
960 #define MIX_REAR1(samples, si1, rs, coef) \
961 samples[i] += samples[si1] * coef[rs][0]; \
962 samples[i+256] += samples[si1] * coef[rs][1];
964 #define MIX_REAR2(samples, si1, si2, rs, coef) \
965 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs+1][0]; \
966 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs+1][1];
968 #define MIX_FRONT3(samples, coef) \
972 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
973 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
975 #define DOWNMIX_TO_STEREO(op1, op2) \
976 for (i = 0; i < 256; i++){ \
981 static void dca_downmix(float *samples, int srcfmt,
982 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
983 const int8_t *channel_mapping)
988 float coef[DCA_PRIM_CHANNELS_MAX][2];
990 for (i=0; i<DCA_PRIM_CHANNELS_MAX; i++) {
991 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
992 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
998 case DCA_STEREO_TOTAL:
999 case DCA_STEREO_SUMDIFF:
1001 av_log(NULL, 0, "Not implemented!\n");
1006 c = channel_mapping[0] * 256;
1007 l = channel_mapping[1] * 256;
1008 r = channel_mapping[2] * 256;
1009 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),);
1012 s = channel_mapping[2] * 256;
1013 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef),);
1016 c = channel_mapping[0] * 256;
1017 l = channel_mapping[1] * 256;
1018 r = channel_mapping[2] * 256;
1019 s = channel_mapping[3] * 256;
1020 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1021 MIX_REAR1(samples, i + s, 3, coef));
1024 sl = channel_mapping[2] * 256;
1025 sr = channel_mapping[3] * 256;
1026 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef),);
1029 c = channel_mapping[0] * 256;
1030 l = channel_mapping[1] * 256;
1031 r = channel_mapping[2] * 256;
1032 sl = channel_mapping[3] * 256;
1033 sr = channel_mapping[4] * 256;
1034 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1035 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1041 #ifndef decode_blockcodes
1042 /* Very compact version of the block code decoder that does not use table
1043 * look-up but is slightly slower */
1044 static int decode_blockcode(int code, int levels, int *values)
1047 int offset = (levels - 1) >> 1;
1049 for (i = 0; i < 4; i++) {
1050 int div = FASTDIV(code, levels);
1051 values[i] = code - offset - div*levels;
1058 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1060 return decode_blockcode(code1, levels, values) |
1061 decode_blockcode(code2, levels, values + 4);
1065 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1066 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1068 #ifndef int8x8_fmul_int32
1069 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1071 float fscale = scale / 16.0;
1073 for (i = 0; i < 8; i++)
1074 dst[i] = src[i] * fscale;
1078 static int dca_subsubframe(DCAContext * s, int base_channel, int block_index)
1081 int subsubframe = s->current_subsubframe;
1083 const float *quant_step_table;
1086 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1087 LOCAL_ALIGNED_16(int, block, [8]);
1093 /* Select quantization step size table */
1094 if (s->bit_rate_index == 0x1f)
1095 quant_step_table = lossless_quant_d;
1097 quant_step_table = lossy_quant_d;
1099 for (k = base_channel; k < s->prim_channels; k++) {
1100 if (get_bits_left(&s->gb) < 0)
1101 return AVERROR_INVALIDDATA;
1103 for (l = 0; l < s->vq_start_subband[k]; l++) {
1106 /* Select the mid-tread linear quantizer */
1107 int abits = s->bitalloc[k][l];
1109 float quant_step_size = quant_step_table[abits];
1112 * Determine quantization index code book and its type
1115 /* Select quantization index code book */
1116 int sel = s->quant_index_huffman[k][abits];
1119 * Extract bits from the bit stream
1122 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1124 /* Deal with transients */
1125 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1126 float rscale = quant_step_size * s->scale_factor[k][l][sfi] * s->scalefactor_adj[k][sel];
1128 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){
1131 int block_code1, block_code2, size, levels, err;
1133 size = abits_sizes[abits-1];
1134 levels = abits_levels[abits-1];
1136 block_code1 = get_bits(&s->gb, size);
1137 block_code2 = get_bits(&s->gb, size);
1138 err = decode_blockcodes(block_code1, block_code2,
1141 av_log(s->avctx, AV_LOG_ERROR,
1142 "ERROR: block code look-up failed\n");
1143 return AVERROR_INVALIDDATA;
1147 for (m = 0; m < 8; m++)
1148 block[m] = get_sbits(&s->gb, abits - 3);
1152 for (m = 0; m < 8; m++)
1153 block[m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel);
1156 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1161 * Inverse ADPCM if in prediction mode
1163 if (s->prediction_mode[k][l]) {
1165 for (m = 0; m < 8; m++) {
1166 for (n = 1; n <= 4; n++)
1168 subband_samples[k][l][m] +=
1169 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1170 subband_samples[k][l][m - n] / 8192);
1171 else if (s->predictor_history)
1172 subband_samples[k][l][m] +=
1173 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1174 s->subband_samples_hist[k][l][m - n +
1181 * Decode VQ encoded high frequencies
1183 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1184 /* 1 vector -> 32 samples but we only need the 8 samples
1185 * for this subsubframe. */
1186 int hfvq = s->high_freq_vq[k][l];
1188 if (!s->debug_flag & 0x01) {
1189 av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
1190 s->debug_flag |= 0x01;
1193 int8x8_fmul_int32(subband_samples[k][l],
1194 &high_freq_vq[hfvq][subsubframe * 8],
1195 s->scale_factor[k][l][0]);
1199 /* Check for DSYNC after subsubframe */
1200 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1201 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1203 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1206 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1210 /* Backup predictor history for adpcm */
1211 for (k = base_channel; k < s->prim_channels; k++)
1212 for (l = 0; l < s->vq_start_subband[k]; l++)
1213 memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
1214 4 * sizeof(subband_samples[0][0][0]));
1219 static int dca_filter_channels(DCAContext * s, int block_index)
1221 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1224 /* 32 subbands QMF */
1225 for (k = 0; k < s->prim_channels; k++) {
1226 /* static float pcm_to_double[8] =
1227 {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
1228 qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
1229 M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ );
1233 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1234 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1237 /* Generate LFE samples for this subsubframe FIXME!!! */
1238 if (s->output & DCA_LFE) {
1239 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1240 s->lfe_data + 2 * s->lfe * (block_index + 4),
1241 &s->samples[256 * dca_lfe_index[s->amode]],
1242 (1.0/256.0)*s->scale_bias);
1243 /* Outputs 20bits pcm samples */
1250 static int dca_subframe_footer(DCAContext * s, int base_channel)
1252 int aux_data_count = 0, i;
1255 * Unpack optional information
1258 /* presumably optional information only appears in the core? */
1259 if (!base_channel) {
1261 skip_bits_long(&s->gb, 32);
1264 aux_data_count = get_bits(&s->gb, 6);
1266 for (i = 0; i < aux_data_count; i++)
1267 get_bits(&s->gb, 8);
1269 if (s->crc_present && (s->downmix || s->dynrange))
1270 get_bits(&s->gb, 16);
1277 * Decode a dca frame block
1279 * @param s pointer to the DCAContext
1282 static int dca_decode_block(DCAContext * s, int base_channel, int block_index)
1287 if (s->current_subframe >= s->subframes) {
1288 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1289 s->current_subframe, s->subframes);
1290 return AVERROR_INVALIDDATA;
1293 if (!s->current_subsubframe) {
1295 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1297 /* Read subframe header */
1298 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1302 /* Read subsubframe */
1304 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1306 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1310 s->current_subsubframe++;
1311 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1312 s->current_subsubframe = 0;
1313 s->current_subframe++;
1315 if (s->current_subframe >= s->subframes) {
1317 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1319 /* Read subframe footer */
1320 if ((ret = dca_subframe_footer(s, base_channel)))
1328 * Convert bitstream to one representation based on sync marker
1330 static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst,
1335 const uint16_t *ssrc = (const uint16_t *) src;
1336 uint16_t *sdst = (uint16_t *) dst;
1339 if ((unsigned)src_size > (unsigned)max_size) {
1340 // av_log(NULL, AV_LOG_ERROR, "Input frame size larger than DCA_MAX_FRAME_SIZE!\n");
1342 src_size = max_size;
1347 case DCA_MARKER_RAW_BE:
1348 memcpy(dst, src, src_size);
1350 case DCA_MARKER_RAW_LE:
1351 for (i = 0; i < (src_size + 1) >> 1; i++)
1352 *sdst++ = av_bswap16(*ssrc++);
1354 case DCA_MARKER_14B_BE:
1355 case DCA_MARKER_14B_LE:
1356 init_put_bits(&pb, dst, max_size);
1357 for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) {
1358 tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
1359 put_bits(&pb, 14, tmp);
1361 flush_put_bits(&pb);
1362 return (put_bits_count(&pb) + 7) >> 3;
1364 return AVERROR_INVALIDDATA;
1369 * Return the number of channels in an ExSS speaker mask (HD)
1371 static int dca_exss_mask2count(int mask)
1373 /* count bits that mean speaker pairs twice */
1374 return av_popcount(mask)
1375 + av_popcount(mask & (
1376 DCA_EXSS_CENTER_LEFT_RIGHT
1377 | DCA_EXSS_FRONT_LEFT_RIGHT
1378 | DCA_EXSS_FRONT_HIGH_LEFT_RIGHT
1379 | DCA_EXSS_WIDE_LEFT_RIGHT
1380 | DCA_EXSS_SIDE_LEFT_RIGHT
1381 | DCA_EXSS_SIDE_HIGH_LEFT_RIGHT
1382 | DCA_EXSS_SIDE_REAR_LEFT_RIGHT
1383 | DCA_EXSS_REAR_LEFT_RIGHT
1384 | DCA_EXSS_REAR_HIGH_LEFT_RIGHT
1389 * Skip mixing coefficients of a single mix out configuration (HD)
1391 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1395 for (i = 0; i < channels; i++) {
1396 int mix_map_mask = get_bits(gb, out_ch);
1397 int num_coeffs = av_popcount(mix_map_mask);
1398 skip_bits_long(gb, num_coeffs * 6);
1403 * Parse extension substream asset header (HD)
1405 static int dca_exss_parse_asset_header(DCAContext *s)
1407 int header_pos = get_bits_count(&s->gb);
1410 int embedded_stereo = 0;
1411 int embedded_6ch = 0;
1412 int drc_code_present;
1413 int extensions_mask;
1416 if (get_bits_left(&s->gb) < 16)
1419 /* We will parse just enough to get to the extensions bitmask with which
1420 * we can set the profile value. */
1422 header_size = get_bits(&s->gb, 9) + 1;
1423 skip_bits(&s->gb, 3); // asset index
1425 if (s->static_fields) {
1426 if (get_bits1(&s->gb))
1427 skip_bits(&s->gb, 4); // asset type descriptor
1428 if (get_bits1(&s->gb))
1429 skip_bits_long(&s->gb, 24); // language descriptor
1431 if (get_bits1(&s->gb)) {
1432 /* How can one fit 1024 bytes of text here if the maximum value
1433 * for the asset header size field above was 512 bytes? */
1434 int text_length = get_bits(&s->gb, 10) + 1;
1435 if (get_bits_left(&s->gb) < text_length * 8)
1437 skip_bits_long(&s->gb, text_length * 8); // info text
1440 skip_bits(&s->gb, 5); // bit resolution - 1
1441 skip_bits(&s->gb, 4); // max sample rate code
1442 channels = get_bits(&s->gb, 8) + 1;
1444 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1445 int spkr_remap_sets;
1446 int spkr_mask_size = 16;
1450 embedded_stereo = get_bits1(&s->gb);
1452 embedded_6ch = get_bits1(&s->gb);
1454 if (get_bits1(&s->gb)) {
1455 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1456 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1459 spkr_remap_sets = get_bits(&s->gb, 3);
1461 for (i = 0; i < spkr_remap_sets; i++) {
1462 /* std layout mask for each remap set */
1463 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1466 for (i = 0; i < spkr_remap_sets; i++) {
1467 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1468 if (get_bits_left(&s->gb) < 0)
1471 for (j = 0; j < num_spkrs[i]; j++) {
1472 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1473 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1474 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1479 skip_bits(&s->gb, 3); // representation type
1483 drc_code_present = get_bits1(&s->gb);
1484 if (drc_code_present)
1485 get_bits(&s->gb, 8); // drc code
1487 if (get_bits1(&s->gb))
1488 skip_bits(&s->gb, 5); // dialog normalization code
1490 if (drc_code_present && embedded_stereo)
1491 get_bits(&s->gb, 8); // drc stereo code
1493 if (s->mix_metadata && get_bits1(&s->gb)) {
1494 skip_bits(&s->gb, 1); // external mix
1495 skip_bits(&s->gb, 6); // post mix gain code
1497 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1498 skip_bits(&s->gb, 3); // drc limit
1500 skip_bits(&s->gb, 8); // custom drc code
1502 if (get_bits1(&s->gb)) // channel specific scaling
1503 for (i = 0; i < s->num_mix_configs; i++)
1504 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1506 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1508 for (i = 0; i < s->num_mix_configs; i++) {
1509 if (get_bits_left(&s->gb) < 0)
1511 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1513 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1514 if (embedded_stereo)
1515 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1519 switch (get_bits(&s->gb, 2)) {
1520 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1521 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1522 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1523 case 3: extensions_mask = 0; /* aux coding */ break;
1526 /* not parsed further, we were only interested in the extensions mask */
1528 if (get_bits_left(&s->gb) < 0)
1531 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1532 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1535 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1537 if (extensions_mask & DCA_EXT_EXSS_XLL)
1538 s->profile = FF_PROFILE_DTS_HD_MA;
1539 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1541 s->profile = FF_PROFILE_DTS_HD_HRA;
1543 if (!(extensions_mask & DCA_EXT_CORE))
1544 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1545 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1546 av_log(s->avctx, AV_LOG_WARNING, "DTS extensions detection mismatch (%d, %d)\n",
1547 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1553 * Parse extension substream header (HD)
1555 static void dca_exss_parse_header(DCAContext *s)
1561 int active_ss_mask[8];
1564 if (get_bits_left(&s->gb) < 52)
1567 skip_bits(&s->gb, 8); // user data
1568 ss_index = get_bits(&s->gb, 2);
1570 blownup = get_bits1(&s->gb);
1571 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1572 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1574 s->static_fields = get_bits1(&s->gb);
1575 if (s->static_fields) {
1576 skip_bits(&s->gb, 2); // reference clock code
1577 skip_bits(&s->gb, 3); // frame duration code
1579 if (get_bits1(&s->gb))
1580 skip_bits_long(&s->gb, 36); // timestamp
1582 /* a single stream can contain multiple audio assets that can be
1583 * combined to form multiple audio presentations */
1585 num_audiop = get_bits(&s->gb, 3) + 1;
1586 if (num_audiop > 1) {
1587 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1588 /* ignore such streams for now */
1592 num_assets = get_bits(&s->gb, 3) + 1;
1593 if (num_assets > 1) {
1594 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1595 /* ignore such streams for now */
1599 for (i = 0; i < num_audiop; i++)
1600 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1602 for (i = 0; i < num_audiop; i++)
1603 for (j = 0; j <= ss_index; j++)
1604 if (active_ss_mask[i] & (1 << j))
1605 skip_bits(&s->gb, 8); // active asset mask
1607 s->mix_metadata = get_bits1(&s->gb);
1608 if (s->mix_metadata) {
1609 int mix_out_mask_size;
1611 skip_bits(&s->gb, 2); // adjustment level
1612 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1613 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1615 for (i = 0; i < s->num_mix_configs; i++) {
1616 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1617 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1622 for (i = 0; i < num_assets; i++)
1623 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1625 for (i = 0; i < num_assets; i++) {
1626 if (dca_exss_parse_asset_header(s))
1630 /* not parsed further, we were only interested in the extensions mask
1631 * from the asset header */
1635 * Main frame decoding function
1636 * FIXME add arguments
1638 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1639 int *got_frame_ptr, AVPacket *avpkt)
1641 const uint8_t *buf = avpkt->data;
1642 int buf_size = avpkt->size;
1645 int num_core_channels = 0;
1648 int16_t *samples_s16;
1649 DCAContext *s = avctx->priv_data;
1656 s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1657 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1658 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1659 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1660 return AVERROR_INVALIDDATA;
1663 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1664 if ((ret = dca_parse_frame_header(s)) < 0) {
1665 //seems like the frame is corrupt, try with the next one
1668 //set AVCodec values with parsed data
1669 avctx->sample_rate = s->sample_rate;
1670 avctx->bit_rate = s->bit_rate;
1671 avctx->frame_size = s->sample_blocks * 32;
1673 s->profile = FF_PROFILE_DTS;
1675 for (i = 0; i < (s->sample_blocks / 8); i++) {
1676 if ((ret = dca_decode_block(s, 0, i))) {
1677 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1682 /* record number of core channels incase less than max channels are requested */
1683 num_core_channels = s->prim_channels;
1686 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1688 s->core_ext_mask = 0;
1690 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1692 /* only scan for extensions if ext_descr was unknown or indicated a
1693 * supported XCh extension */
1694 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1696 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1697 * extensions scan can fill it up */
1698 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1700 /* extensions start at 32-bit boundaries into bitstream */
1701 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1703 while(core_ss_end - get_bits_count(&s->gb) >= 32) {
1704 uint32_t bits = get_bits_long(&s->gb, 32);
1708 int ext_amode, xch_fsize;
1710 s->xch_base_channel = s->prim_channels;
1712 /* validate sync word using XCHFSIZE field */
1713 xch_fsize = show_bits(&s->gb, 10);
1714 if((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1715 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1718 /* skip length-to-end-of-frame field for the moment */
1719 skip_bits(&s->gb, 10);
1721 s->core_ext_mask |= DCA_EXT_XCH;
1723 /* extension amode should == 1, number of channels in extension */
1724 /* AFAIK XCh is not used for more channels */
1725 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1726 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1727 " supported!\n",ext_amode);
1731 /* much like core primary audio coding header */
1732 dca_parse_audio_coding_header(s, s->xch_base_channel);
1734 for (i = 0; i < (s->sample_blocks / 8); i++) {
1735 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1736 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1745 /* XXCh: extended channels */
1746 /* usually found either in core or HD part in DTS-HD HRA streams,
1747 * but not in DTS-ES which contains XCh extensions instead */
1748 s->core_ext_mask |= DCA_EXT_XXCH;
1752 int fsize96 = show_bits(&s->gb, 12) + 1;
1753 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1756 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n", get_bits_count(&s->gb));
1757 skip_bits(&s->gb, 12);
1758 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1759 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1761 s->core_ext_mask |= DCA_EXT_X96;
1766 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1770 /* no supported extensions, skip the rest of the core substream */
1771 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1774 if (s->core_ext_mask & DCA_EXT_X96)
1775 s->profile = FF_PROFILE_DTS_96_24;
1776 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1777 s->profile = FF_PROFILE_DTS_ES;
1779 /* check for ExSS (HD part) */
1780 if (s->dca_buffer_size - s->frame_size > 32
1781 && get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1782 dca_exss_parse_header(s);
1784 avctx->profile = s->profile;
1786 channels = s->prim_channels + !!s->lfe;
1789 avctx->channel_layout = dca_core_channel_layout[s->amode];
1791 if (s->xch_present && (!avctx->request_channels ||
1792 avctx->request_channels > num_core_channels + !!s->lfe)) {
1793 avctx->channel_layout |= AV_CH_BACK_CENTER;
1795 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1796 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1798 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1801 channels = num_core_channels + !!s->lfe;
1802 s->xch_present = 0; /* disable further xch processing */
1804 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1805 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1807 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1810 if (channels > !!s->lfe &&
1811 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1812 return AVERROR_INVALIDDATA;
1814 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1816 s->output = DCA_STEREO;
1817 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1819 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
1820 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
1821 s->channel_order_tab = dca_channel_order_native;
1824 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
1825 return AVERROR_INVALIDDATA;
1828 if (avctx->channels != channels) {
1829 if (avctx->channels)
1830 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
1831 avctx->channels = channels;
1834 /* get output buffer */
1835 s->frame.nb_samples = 256 * (s->sample_blocks / 8);
1836 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
1837 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1840 samples_flt = (float *)s->frame.data[0];
1841 samples_s16 = (int16_t *)s->frame.data[0];
1843 /* filter to get final output */
1844 for (i = 0; i < (s->sample_blocks / 8); i++) {
1845 dca_filter_channels(s, i);
1847 /* If this was marked as a DTS-ES stream we need to subtract back- */
1848 /* channel from SL & SR to remove matrixed back-channel signal */
1849 if((s->source_pcm_res & 1) && s->xch_present) {
1850 float* back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1851 float* lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1852 float* rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1853 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1854 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1857 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1858 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1860 samples_flt += 256 * channels;
1862 s->fmt_conv.float_to_int16_interleave(samples_s16,
1863 s->samples_chanptr, 256,
1865 samples_s16 += 256 * channels;
1869 /* update lfe history */
1870 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1871 for (i = 0; i < 2 * s->lfe * 4; i++) {
1872 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1876 *(AVFrame *)data = s->frame;
1884 * DCA initialization
1886 * @param avctx pointer to the AVCodecContext
1889 static av_cold int dca_decode_init(AVCodecContext * avctx)
1891 DCAContext *s = avctx->priv_data;
1897 dsputil_init(&s->dsp, avctx);
1898 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1899 ff_synth_filter_init(&s->synth);
1900 ff_dcadsp_init(&s->dcadsp);
1901 ff_fmt_convert_init(&s->fmt_conv, avctx);
1903 for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++)
1904 s->samples_chanptr[i] = s->samples + i * 256;
1906 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1907 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1908 s->scale_bias = 1.0 / 32768.0;
1910 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1911 s->scale_bias = 1.0;
1914 /* allow downmixing to stereo */
1915 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1916 avctx->request_channels == 2) {
1917 avctx->channels = avctx->request_channels;
1920 avcodec_get_frame_defaults(&s->frame);
1921 avctx->coded_frame = &s->frame;
1926 static av_cold int dca_decode_end(AVCodecContext * avctx)
1928 DCAContext *s = avctx->priv_data;
1929 ff_mdct_end(&s->imdct);
1933 static const AVProfile profiles[] = {
1934 { FF_PROFILE_DTS, "DTS" },
1935 { FF_PROFILE_DTS_ES, "DTS-ES" },
1936 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1937 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1938 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1939 { FF_PROFILE_UNKNOWN },
1942 AVCodec ff_dca_decoder = {
1944 .type = AVMEDIA_TYPE_AUDIO,
1946 .priv_data_size = sizeof(DCAContext),
1947 .init = dca_decode_init,
1948 .decode = dca_decode_frame,
1949 .close = dca_decode_end,
1950 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1951 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1952 .sample_fmts = (const enum AVSampleFormat[]) {
1953 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
1955 .profiles = NULL_IF_CONFIG_SMALL(profiles),