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;
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 total_channels; ///< number of channels including extensions
297 int prim_channels; ///< number of primary audio channels
298 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
299 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
300 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
301 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
302 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
303 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
304 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
305 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
307 /* Primary audio coding side information */
308 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
309 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
310 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
311 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
312 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
313 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
314 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
315 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
316 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
317 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
318 int dynrange_coef; ///< dynamic range coefficient
320 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
322 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
323 int lfe_scale_factor;
325 /* Subband samples history (for ADPCM) */
326 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
327 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
328 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
329 int hist_index[DCA_PRIM_CHANNELS_MAX];
330 DECLARE_ALIGNED(32, float, raXin)[32];
332 int output; ///< type of output
333 float scale_bias; ///< output scale
335 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
336 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
337 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX+1];
339 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
340 int dca_buffer_size; ///< how much data is in the dca_buffer
342 const int8_t* channel_order_tab; ///< channel reordering table, lfe and non lfe
344 /* Current position in DCA frame */
345 int current_subframe;
346 int current_subsubframe;
348 int core_ext_mask; ///< present extensions in the core substream
350 /* XCh extension information */
351 int xch_present; ///< XCh extension present and valid
352 int xch_base_channel; ///< index of first (only) channel containing XCH data
354 /* ExSS header parser */
355 int static_fields; ///< static fields present
356 int mix_metadata; ///< mixing metadata present
357 int num_mix_configs; ///< number of mix out configurations
358 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
362 int debug_flag; ///< used for suppressing repeated error messages output
365 SynthFilterContext synth;
366 DCADSPContext dcadsp;
367 FmtConvertContext fmt_conv;
370 static const uint16_t dca_vlc_offs[] = {
371 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
372 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
373 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
374 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
375 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
376 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
379 static av_cold void dca_init_vlcs(void)
381 static int vlcs_initialized = 0;
383 static VLC_TYPE dca_table[23622][2];
385 if (vlcs_initialized)
388 dca_bitalloc_index.offset = 1;
389 dca_bitalloc_index.wrap = 2;
390 for (i = 0; i < 5; i++) {
391 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
392 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
393 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
394 bitalloc_12_bits[i], 1, 1,
395 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
397 dca_scalefactor.offset = -64;
398 dca_scalefactor.wrap = 2;
399 for (i = 0; i < 5; i++) {
400 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
401 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
402 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
403 scales_bits[i], 1, 1,
404 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
406 dca_tmode.offset = 0;
408 for (i = 0; i < 4; i++) {
409 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
410 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
411 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
413 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
416 for (i = 0; i < 10; i++)
417 for (j = 0; j < 7; j++){
418 if (!bitalloc_codes[i][j]) break;
419 dca_smpl_bitalloc[i+1].offset = bitalloc_offsets[i];
420 dca_smpl_bitalloc[i+1].wrap = 1 + (j > 4);
421 dca_smpl_bitalloc[i+1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
422 dca_smpl_bitalloc[i+1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
423 init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j],
425 bitalloc_bits[i][j], 1, 1,
426 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
429 vlcs_initialized = 1;
432 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
435 *dst++ = get_bits(gb, bits);
438 static int dca_parse_audio_coding_header(DCAContext * s, int base_channel)
441 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
442 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
443 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
445 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
446 s->prim_channels = s->total_channels;
448 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
449 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
452 for (i = base_channel; i < s->prim_channels; i++) {
453 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
454 if (s->subband_activity[i] > DCA_SUBBANDS)
455 s->subband_activity[i] = DCA_SUBBANDS;
457 for (i = base_channel; i < s->prim_channels; i++) {
458 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
459 if (s->vq_start_subband[i] > DCA_SUBBANDS)
460 s->vq_start_subband[i] = DCA_SUBBANDS;
462 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
463 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
464 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
465 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
467 /* Get codebooks quantization indexes */
469 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
470 for (j = 1; j < 11; j++)
471 for (i = base_channel; i < s->prim_channels; i++)
472 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
474 /* Get scale factor adjustment */
475 for (j = 0; j < 11; j++)
476 for (i = base_channel; i < s->prim_channels; i++)
477 s->scalefactor_adj[i][j] = 1;
479 for (j = 1; j < 11; j++)
480 for (i = base_channel; i < s->prim_channels; i++)
481 if (s->quant_index_huffman[i][j] < thr[j])
482 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
484 if (s->crc_present) {
485 /* Audio header CRC check */
486 get_bits(&s->gb, 16);
489 s->current_subframe = 0;
490 s->current_subsubframe = 0;
493 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
494 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
495 for (i = base_channel; i < s->prim_channels; i++){
496 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", s->subband_activity[i]);
497 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", s->vq_start_subband[i]);
498 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", s->joint_intensity[i]);
499 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", s->transient_huffman[i]);
500 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", s->scalefactor_huffman[i]);
501 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", s->bitalloc_huffman[i]);
502 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
503 for (j = 0; j < 11; j++)
504 av_log(s->avctx, AV_LOG_DEBUG, " %i",
505 s->quant_index_huffman[i][j]);
506 av_log(s->avctx, AV_LOG_DEBUG, "\n");
507 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
508 for (j = 0; j < 11; j++)
509 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
510 av_log(s->avctx, AV_LOG_DEBUG, "\n");
517 static int dca_parse_frame_header(DCAContext * s)
519 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
522 get_bits(&s->gb, 32);
525 s->frame_type = get_bits(&s->gb, 1);
526 s->samples_deficit = get_bits(&s->gb, 5) + 1;
527 s->crc_present = get_bits(&s->gb, 1);
528 s->sample_blocks = get_bits(&s->gb, 7) + 1;
529 s->frame_size = get_bits(&s->gb, 14) + 1;
530 if (s->frame_size < 95)
531 return AVERROR_INVALIDDATA;
532 s->amode = get_bits(&s->gb, 6);
533 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
535 return AVERROR_INVALIDDATA;
536 s->bit_rate_index = get_bits(&s->gb, 5);
537 s->bit_rate = dca_bit_rates[s->bit_rate_index];
539 return AVERROR_INVALIDDATA;
541 s->downmix = get_bits(&s->gb, 1);
542 s->dynrange = get_bits(&s->gb, 1);
543 s->timestamp = get_bits(&s->gb, 1);
544 s->aux_data = get_bits(&s->gb, 1);
545 s->hdcd = get_bits(&s->gb, 1);
546 s->ext_descr = get_bits(&s->gb, 3);
547 s->ext_coding = get_bits(&s->gb, 1);
548 s->aspf = get_bits(&s->gb, 1);
549 s->lfe = get_bits(&s->gb, 2);
550 s->predictor_history = get_bits(&s->gb, 1);
552 /* TODO: check CRC */
554 s->header_crc = get_bits(&s->gb, 16);
556 s->multirate_inter = get_bits(&s->gb, 1);
557 s->version = get_bits(&s->gb, 4);
558 s->copy_history = get_bits(&s->gb, 2);
559 s->source_pcm_res = get_bits(&s->gb, 3);
560 s->front_sum = get_bits(&s->gb, 1);
561 s->surround_sum = get_bits(&s->gb, 1);
562 s->dialog_norm = get_bits(&s->gb, 4);
564 /* FIXME: channels mixing levels */
565 s->output = s->amode;
566 if (s->lfe) s->output |= DCA_LFE;
569 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
570 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
571 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
572 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
573 s->sample_blocks, s->sample_blocks * 32);
574 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
575 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
576 s->amode, dca_channels[s->amode]);
577 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
579 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
581 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
582 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
583 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
584 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
585 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
586 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
587 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
588 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
589 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
590 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
591 s->predictor_history);
592 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
593 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
595 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
596 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
597 av_log(s->avctx, AV_LOG_DEBUG,
598 "source pcm resolution: %i (%i bits/sample)\n",
599 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
600 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
601 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
602 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
603 av_log(s->avctx, AV_LOG_DEBUG, "\n");
606 /* Primary audio coding header */
607 s->subframes = get_bits(&s->gb, 4) + 1;
609 return dca_parse_audio_coding_header(s, 0);
613 static inline int get_scale(GetBitContext *gb, int level, int value)
616 /* huffman encoded */
617 value += get_bitalloc(gb, &dca_scalefactor, level);
618 } else if (level < 8)
619 value = get_bits(gb, level + 1);
623 static int dca_subframe_header(DCAContext * s, int base_channel, int block_index)
625 /* Primary audio coding side information */
628 if (get_bits_left(&s->gb) < 0)
629 return AVERROR_INVALIDDATA;
632 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
633 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
636 for (j = base_channel; j < s->prim_channels; j++) {
637 for (k = 0; k < s->subband_activity[j]; k++)
638 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
641 /* Get prediction codebook */
642 for (j = base_channel; j < s->prim_channels; j++) {
643 for (k = 0; k < s->subband_activity[j]; k++) {
644 if (s->prediction_mode[j][k] > 0) {
645 /* (Prediction coefficient VQ address) */
646 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
651 /* Bit allocation index */
652 for (j = base_channel; j < s->prim_channels; j++) {
653 for (k = 0; k < s->vq_start_subband[j]; k++) {
654 if (s->bitalloc_huffman[j] == 6)
655 s->bitalloc[j][k] = get_bits(&s->gb, 5);
656 else if (s->bitalloc_huffman[j] == 5)
657 s->bitalloc[j][k] = get_bits(&s->gb, 4);
658 else if (s->bitalloc_huffman[j] == 7) {
659 av_log(s->avctx, AV_LOG_ERROR,
660 "Invalid bit allocation index\n");
661 return AVERROR_INVALIDDATA;
664 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
667 if (s->bitalloc[j][k] > 26) {
668 // av_log(s->avctx,AV_LOG_DEBUG,"bitalloc index [%i][%i] too big (%i)\n",
669 // j, k, s->bitalloc[j][k]);
670 return AVERROR_INVALIDDATA;
675 /* Transition mode */
676 for (j = base_channel; j < s->prim_channels; j++) {
677 for (k = 0; k < s->subband_activity[j]; k++) {
678 s->transition_mode[j][k] = 0;
679 if (s->subsubframes[s->current_subframe] > 1 &&
680 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
681 s->transition_mode[j][k] =
682 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
687 if (get_bits_left(&s->gb) < 0)
688 return AVERROR_INVALIDDATA;
690 for (j = base_channel; j < s->prim_channels; j++) {
691 const uint32_t *scale_table;
694 memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
696 if (s->scalefactor_huffman[j] == 6)
697 scale_table = scale_factor_quant7;
699 scale_table = scale_factor_quant6;
701 /* When huffman coded, only the difference is encoded */
704 for (k = 0; k < s->subband_activity[j]; k++) {
705 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
706 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
707 s->scale_factor[j][k][0] = scale_table[scale_sum];
710 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
711 /* Get second scale factor */
712 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
713 s->scale_factor[j][k][1] = scale_table[scale_sum];
718 /* Joint subband scale factor codebook select */
719 for (j = base_channel; j < s->prim_channels; j++) {
720 /* Transmitted only if joint subband coding enabled */
721 if (s->joint_intensity[j] > 0)
722 s->joint_huff[j] = get_bits(&s->gb, 3);
725 if (get_bits_left(&s->gb) < 0)
726 return AVERROR_INVALIDDATA;
728 /* Scale factors for joint subband coding */
729 for (j = base_channel; j < s->prim_channels; j++) {
732 /* Transmitted only if joint subband coding enabled */
733 if (s->joint_intensity[j] > 0) {
735 source_channel = s->joint_intensity[j] - 1;
737 /* When huffman coded, only the difference is encoded
738 * (is this valid as well for joint scales ???) */
740 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
741 scale = get_scale(&s->gb, s->joint_huff[j], 0);
742 scale += 64; /* bias */
743 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
746 if (!(s->debug_flag & 0x02)) {
747 av_log(s->avctx, AV_LOG_DEBUG,
748 "Joint stereo coding not supported\n");
749 s->debug_flag |= 0x02;
754 /* Stereo downmix coefficients */
755 if (!base_channel && s->prim_channels > 2) {
757 for (j = base_channel; j < s->prim_channels; j++) {
758 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
759 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
762 int am = s->amode & DCA_CHANNEL_MASK;
763 for (j = base_channel; j < s->prim_channels; j++) {
764 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
765 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
770 /* Dynamic range coefficient */
771 if (!base_channel && s->dynrange)
772 s->dynrange_coef = get_bits(&s->gb, 8);
774 /* Side information CRC check word */
775 if (s->crc_present) {
776 get_bits(&s->gb, 16);
780 * Primary audio data arrays
783 /* VQ encoded high frequency subbands */
784 for (j = base_channel; j < s->prim_channels; j++)
785 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
786 /* 1 vector -> 32 samples */
787 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
789 /* Low frequency effect data */
790 if (!base_channel && s->lfe) {
792 int lfe_samples = 2 * s->lfe * (4 + block_index);
793 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
796 for (j = lfe_samples; j < lfe_end_sample; j++) {
797 /* Signed 8 bits int */
798 s->lfe_data[j] = get_sbits(&s->gb, 8);
801 /* Scale factor index */
802 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
804 /* Quantization step size * scale factor */
805 lfe_scale = 0.035 * s->lfe_scale_factor;
807 for (j = lfe_samples; j < lfe_end_sample; j++)
808 s->lfe_data[j] *= lfe_scale;
812 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", s->subsubframes[s->current_subframe]);
813 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
814 s->partial_samples[s->current_subframe]);
815 for (j = base_channel; j < s->prim_channels; j++) {
816 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
817 for (k = 0; k < s->subband_activity[j]; k++)
818 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
819 av_log(s->avctx, AV_LOG_DEBUG, "\n");
821 for (j = base_channel; j < s->prim_channels; j++) {
822 for (k = 0; k < s->subband_activity[j]; k++)
823 av_log(s->avctx, AV_LOG_DEBUG,
824 "prediction coefs: %f, %f, %f, %f\n",
825 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
826 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
827 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
828 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
830 for (j = base_channel; j < s->prim_channels; j++) {
831 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
832 for (k = 0; k < s->vq_start_subband[j]; k++)
833 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
834 av_log(s->avctx, AV_LOG_DEBUG, "\n");
836 for (j = base_channel; j < s->prim_channels; j++) {
837 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
838 for (k = 0; k < s->subband_activity[j]; k++)
839 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
840 av_log(s->avctx, AV_LOG_DEBUG, "\n");
842 for (j = base_channel; j < s->prim_channels; j++) {
843 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
844 for (k = 0; k < s->subband_activity[j]; k++) {
845 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
846 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
847 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
848 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
850 av_log(s->avctx, AV_LOG_DEBUG, "\n");
852 for (j = base_channel; j < s->prim_channels; j++) {
853 if (s->joint_intensity[j] > 0) {
854 int source_channel = s->joint_intensity[j] - 1;
855 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
856 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
857 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
858 av_log(s->avctx, AV_LOG_DEBUG, "\n");
861 if (!base_channel && s->prim_channels > 2 && s->downmix) {
862 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
863 for (j = 0; j < s->prim_channels; j++) {
864 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][0]]);
865 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][1]]);
867 av_log(s->avctx, AV_LOG_DEBUG, "\n");
869 for (j = base_channel; j < s->prim_channels; j++)
870 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
871 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
872 if (!base_channel && s->lfe) {
873 int lfe_samples = 2 * s->lfe * (4 + block_index);
874 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
876 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
877 for (j = lfe_samples; j < lfe_end_sample; j++)
878 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
879 av_log(s->avctx, AV_LOG_DEBUG, "\n");
886 static void qmf_32_subbands(DCAContext * s, int chans,
887 float samples_in[32][8], float *samples_out,
890 const float *prCoeff;
893 int sb_act = s->subband_activity[chans];
896 scale *= sqrt(1/8.0);
899 if (!s->multirate_inter) /* Non-perfect reconstruction */
900 prCoeff = fir_32bands_nonperfect;
901 else /* Perfect reconstruction */
902 prCoeff = fir_32bands_perfect;
904 for (i = sb_act; i < 32; i++)
907 /* Reconstructed channel sample index */
908 for (subindex = 0; subindex < 8; subindex++) {
909 /* Load in one sample from each subband and clear inactive subbands */
910 for (i = 0; i < sb_act; i++){
911 unsigned sign = (i - 1) & 2;
912 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 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 #ifndef decode_blockcodes
1041 /* Very compact version of the block code decoder that does not use table
1042 * look-up but is slightly slower */
1043 static int decode_blockcode(int code, int levels, int *values)
1046 int offset = (levels - 1) >> 1;
1048 for (i = 0; i < 4; i++) {
1049 int div = FASTDIV(code, levels);
1050 values[i] = code - offset - div*levels;
1057 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1059 return decode_blockcode(code1, levels, values) |
1060 decode_blockcode(code2, levels, values + 4);
1064 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1065 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1067 #ifndef int8x8_fmul_int32
1068 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1070 float fscale = scale / 16.0;
1072 for (i = 0; i < 8; i++)
1073 dst[i] = src[i] * fscale;
1077 static int dca_subsubframe(DCAContext * s, int base_channel, int block_index)
1080 int subsubframe = s->current_subsubframe;
1082 const float *quant_step_table;
1085 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1086 LOCAL_ALIGNED_16(int, block, [8]);
1092 /* Select quantization step size table */
1093 if (s->bit_rate_index == 0x1f)
1094 quant_step_table = lossless_quant_d;
1096 quant_step_table = lossy_quant_d;
1098 for (k = base_channel; k < s->prim_channels; k++) {
1099 if (get_bits_left(&s->gb) < 0)
1100 return AVERROR_INVALIDDATA;
1102 for (l = 0; l < s->vq_start_subband[k]; l++) {
1105 /* Select the mid-tread linear quantizer */
1106 int abits = s->bitalloc[k][l];
1108 float quant_step_size = quant_step_table[abits];
1111 * Determine quantization index code book and its type
1114 /* Select quantization index code book */
1115 int sel = s->quant_index_huffman[k][abits];
1118 * Extract bits from the bit stream
1121 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1123 /* Deal with transients */
1124 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1125 float rscale = quant_step_size * s->scale_factor[k][l][sfi] * s->scalefactor_adj[k][sel];
1127 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){
1130 int block_code1, block_code2, size, levels, err;
1132 size = abits_sizes[abits-1];
1133 levels = abits_levels[abits-1];
1135 block_code1 = get_bits(&s->gb, size);
1136 block_code2 = get_bits(&s->gb, size);
1137 err = decode_blockcodes(block_code1, block_code2,
1140 av_log(s->avctx, AV_LOG_ERROR,
1141 "ERROR: block code look-up failed\n");
1142 return AVERROR_INVALIDDATA;
1146 for (m = 0; m < 8; m++)
1147 block[m] = get_sbits(&s->gb, abits - 3);
1151 for (m = 0; m < 8; m++)
1152 block[m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel);
1155 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1160 * Inverse ADPCM if in prediction mode
1162 if (s->prediction_mode[k][l]) {
1164 for (m = 0; m < 8; m++) {
1165 for (n = 1; n <= 4; n++)
1167 subband_samples[k][l][m] +=
1168 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1169 subband_samples[k][l][m - n] / 8192);
1170 else if (s->predictor_history)
1171 subband_samples[k][l][m] +=
1172 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1173 s->subband_samples_hist[k][l][m - n +
1180 * Decode VQ encoded high frequencies
1182 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1183 /* 1 vector -> 32 samples but we only need the 8 samples
1184 * for this subsubframe. */
1185 int hfvq = s->high_freq_vq[k][l];
1187 if (!s->debug_flag & 0x01) {
1188 av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
1189 s->debug_flag |= 0x01;
1192 int8x8_fmul_int32(subband_samples[k][l],
1193 &high_freq_vq[hfvq][subsubframe * 8],
1194 s->scale_factor[k][l][0]);
1198 /* Check for DSYNC after subsubframe */
1199 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1200 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1202 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1205 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1209 /* Backup predictor history for adpcm */
1210 for (k = base_channel; k < s->prim_channels; k++)
1211 for (l = 0; l < s->vq_start_subband[k]; l++)
1212 memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
1213 4 * sizeof(subband_samples[0][0][0]));
1218 static int dca_filter_channels(DCAContext * s, int block_index)
1220 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1223 /* 32 subbands QMF */
1224 for (k = 0; k < s->prim_channels; k++) {
1225 /* static float pcm_to_double[8] =
1226 {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
1227 qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
1228 M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ );
1232 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1233 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1236 /* Generate LFE samples for this subsubframe FIXME!!! */
1237 if (s->output & DCA_LFE) {
1238 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1239 s->lfe_data + 2 * s->lfe * (block_index + 4),
1240 &s->samples[256 * dca_lfe_index[s->amode]],
1241 (1.0/256.0)*s->scale_bias);
1242 /* Outputs 20bits pcm samples */
1249 static int dca_subframe_footer(DCAContext * s, int base_channel)
1251 int aux_data_count = 0, i;
1254 * Unpack optional information
1257 /* presumably optional information only appears in the core? */
1258 if (!base_channel) {
1260 get_bits(&s->gb, 32);
1263 aux_data_count = get_bits(&s->gb, 6);
1265 for (i = 0; i < aux_data_count; i++)
1266 get_bits(&s->gb, 8);
1268 if (s->crc_present && (s->downmix || s->dynrange))
1269 get_bits(&s->gb, 16);
1276 * Decode a dca frame block
1278 * @param s pointer to the DCAContext
1281 static int dca_decode_block(DCAContext * s, int base_channel, int block_index)
1286 if (s->current_subframe >= s->subframes) {
1287 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1288 s->current_subframe, s->subframes);
1289 return AVERROR_INVALIDDATA;
1292 if (!s->current_subsubframe) {
1294 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1296 /* Read subframe header */
1297 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1301 /* Read subsubframe */
1303 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1305 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1309 s->current_subsubframe++;
1310 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1311 s->current_subsubframe = 0;
1312 s->current_subframe++;
1314 if (s->current_subframe >= s->subframes) {
1316 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1318 /* Read subframe footer */
1319 if ((ret = dca_subframe_footer(s, base_channel)))
1327 * Convert bitstream to one representation based on sync marker
1329 static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst,
1334 const uint16_t *ssrc = (const uint16_t *) src;
1335 uint16_t *sdst = (uint16_t *) dst;
1338 if ((unsigned)src_size > (unsigned)max_size) {
1339 // av_log(NULL, AV_LOG_ERROR, "Input frame size larger than DCA_MAX_FRAME_SIZE!\n");
1341 src_size = max_size;
1346 case DCA_MARKER_RAW_BE:
1347 memcpy(dst, src, src_size);
1349 case DCA_MARKER_RAW_LE:
1350 for (i = 0; i < (src_size + 1) >> 1; i++)
1351 *sdst++ = av_bswap16(*ssrc++);
1353 case DCA_MARKER_14B_BE:
1354 case DCA_MARKER_14B_LE:
1355 init_put_bits(&pb, dst, max_size);
1356 for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) {
1357 tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
1358 put_bits(&pb, 14, tmp);
1360 flush_put_bits(&pb);
1361 return (put_bits_count(&pb) + 7) >> 3;
1363 return AVERROR_INVALIDDATA;
1368 * Return the number of channels in an ExSS speaker mask (HD)
1370 static int dca_exss_mask2count(int mask)
1372 /* count bits that mean speaker pairs twice */
1373 return av_popcount(mask)
1374 + av_popcount(mask & (
1375 DCA_EXSS_CENTER_LEFT_RIGHT
1376 | DCA_EXSS_FRONT_LEFT_RIGHT
1377 | DCA_EXSS_FRONT_HIGH_LEFT_RIGHT
1378 | DCA_EXSS_WIDE_LEFT_RIGHT
1379 | DCA_EXSS_SIDE_LEFT_RIGHT
1380 | DCA_EXSS_SIDE_HIGH_LEFT_RIGHT
1381 | DCA_EXSS_SIDE_REAR_LEFT_RIGHT
1382 | DCA_EXSS_REAR_LEFT_RIGHT
1383 | DCA_EXSS_REAR_HIGH_LEFT_RIGHT
1388 * Skip mixing coefficients of a single mix out configuration (HD)
1390 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1394 for (i = 0; i < channels; i++) {
1395 int mix_map_mask = get_bits(gb, out_ch);
1396 int num_coeffs = av_popcount(mix_map_mask);
1397 skip_bits_long(gb, num_coeffs * 6);
1402 * Parse extension substream asset header (HD)
1404 static int dca_exss_parse_asset_header(DCAContext *s)
1406 int header_pos = get_bits_count(&s->gb);
1409 int embedded_stereo = 0;
1410 int embedded_6ch = 0;
1411 int drc_code_present;
1412 int extensions_mask;
1415 if (get_bits_left(&s->gb) < 16)
1418 /* We will parse just enough to get to the extensions bitmask with which
1419 * we can set the profile value. */
1421 header_size = get_bits(&s->gb, 9) + 1;
1422 skip_bits(&s->gb, 3); // asset index
1424 if (s->static_fields) {
1425 if (get_bits1(&s->gb))
1426 skip_bits(&s->gb, 4); // asset type descriptor
1427 if (get_bits1(&s->gb))
1428 skip_bits_long(&s->gb, 24); // language descriptor
1430 if (get_bits1(&s->gb)) {
1431 /* How can one fit 1024 bytes of text here if the maximum value
1432 * for the asset header size field above was 512 bytes? */
1433 int text_length = get_bits(&s->gb, 10) + 1;
1434 if (get_bits_left(&s->gb) < text_length * 8)
1436 skip_bits_long(&s->gb, text_length * 8); // info text
1439 skip_bits(&s->gb, 5); // bit resolution - 1
1440 skip_bits(&s->gb, 4); // max sample rate code
1441 channels = get_bits(&s->gb, 8) + 1;
1443 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1444 int spkr_remap_sets;
1445 int spkr_mask_size = 16;
1449 embedded_stereo = get_bits1(&s->gb);
1451 embedded_6ch = get_bits1(&s->gb);
1453 if (get_bits1(&s->gb)) {
1454 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1455 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1458 spkr_remap_sets = get_bits(&s->gb, 3);
1460 for (i = 0; i < spkr_remap_sets; i++) {
1461 /* std layout mask for each remap set */
1462 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1465 for (i = 0; i < spkr_remap_sets; i++) {
1466 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1467 if (get_bits_left(&s->gb) < 0)
1470 for (j = 0; j < num_spkrs[i]; j++) {
1471 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1472 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1473 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1478 skip_bits(&s->gb, 3); // representation type
1482 drc_code_present = get_bits1(&s->gb);
1483 if (drc_code_present)
1484 get_bits(&s->gb, 8); // drc code
1486 if (get_bits1(&s->gb))
1487 skip_bits(&s->gb, 5); // dialog normalization code
1489 if (drc_code_present && embedded_stereo)
1490 get_bits(&s->gb, 8); // drc stereo code
1492 if (s->mix_metadata && get_bits1(&s->gb)) {
1493 skip_bits(&s->gb, 1); // external mix
1494 skip_bits(&s->gb, 6); // post mix gain code
1496 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1497 skip_bits(&s->gb, 3); // drc limit
1499 skip_bits(&s->gb, 8); // custom drc code
1501 if (get_bits1(&s->gb)) // channel specific scaling
1502 for (i = 0; i < s->num_mix_configs; i++)
1503 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1505 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1507 for (i = 0; i < s->num_mix_configs; i++) {
1508 if (get_bits_left(&s->gb) < 0)
1510 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1512 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1513 if (embedded_stereo)
1514 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1518 switch (get_bits(&s->gb, 2)) {
1519 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1520 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1521 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1522 case 3: extensions_mask = 0; /* aux coding */ break;
1525 /* not parsed further, we were only interested in the extensions mask */
1527 if (get_bits_left(&s->gb) < 0)
1530 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1531 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1534 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1536 if (extensions_mask & DCA_EXT_EXSS_XLL)
1537 s->profile = FF_PROFILE_DTS_HD_MA;
1538 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1540 s->profile = FF_PROFILE_DTS_HD_HRA;
1542 if (!(extensions_mask & DCA_EXT_CORE))
1543 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1544 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1545 av_log(s->avctx, AV_LOG_WARNING, "DTS extensions detection mismatch (%d, %d)\n",
1546 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1552 * Parse extension substream header (HD)
1554 static void dca_exss_parse_header(DCAContext *s)
1560 int active_ss_mask[8];
1563 if (get_bits_left(&s->gb) < 52)
1566 skip_bits(&s->gb, 8); // user data
1567 ss_index = get_bits(&s->gb, 2);
1569 blownup = get_bits1(&s->gb);
1570 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1571 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1573 s->static_fields = get_bits1(&s->gb);
1574 if (s->static_fields) {
1575 skip_bits(&s->gb, 2); // reference clock code
1576 skip_bits(&s->gb, 3); // frame duration code
1578 if (get_bits1(&s->gb))
1579 skip_bits_long(&s->gb, 36); // timestamp
1581 /* a single stream can contain multiple audio assets that can be
1582 * combined to form multiple audio presentations */
1584 num_audiop = get_bits(&s->gb, 3) + 1;
1585 if (num_audiop > 1) {
1586 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1587 /* ignore such streams for now */
1591 num_assets = get_bits(&s->gb, 3) + 1;
1592 if (num_assets > 1) {
1593 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1594 /* ignore such streams for now */
1598 for (i = 0; i < num_audiop; i++)
1599 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1601 for (i = 0; i < num_audiop; i++)
1602 for (j = 0; j <= ss_index; j++)
1603 if (active_ss_mask[i] & (1 << j))
1604 skip_bits(&s->gb, 8); // active asset mask
1606 s->mix_metadata = get_bits1(&s->gb);
1607 if (s->mix_metadata) {
1608 int mix_out_mask_size;
1610 skip_bits(&s->gb, 2); // adjustment level
1611 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1612 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1614 for (i = 0; i < s->num_mix_configs; i++) {
1615 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1616 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1621 for (i = 0; i < num_assets; i++)
1622 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1624 for (i = 0; i < num_assets; i++) {
1625 if (dca_exss_parse_asset_header(s))
1629 /* not parsed further, we were only interested in the extensions mask
1630 * from the asset header */
1634 * Main frame decoding function
1635 * FIXME add arguments
1637 static int dca_decode_frame(AVCodecContext * avctx,
1638 void *data, int *data_size,
1641 const uint8_t *buf = avpkt->data;
1642 int buf_size = avpkt->size;
1645 int num_core_channels = 0;
1647 float *samples_flt = data;
1648 int16_t *samples_s16 = data;
1650 DCAContext *s = avctx->priv_data;
1657 s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1658 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1659 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1660 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1661 return AVERROR_INVALIDDATA;
1664 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1665 if ((ret = dca_parse_frame_header(s)) < 0) {
1666 //seems like the frame is corrupt, try with the next one
1669 //set AVCodec values with parsed data
1670 avctx->sample_rate = s->sample_rate;
1671 avctx->bit_rate = s->bit_rate;
1672 avctx->frame_size = s->sample_blocks * 32;
1674 s->profile = FF_PROFILE_DTS;
1676 for (i = 0; i < (s->sample_blocks / 8); i++) {
1677 if ((ret = dca_decode_block(s, 0, i))) {
1678 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1683 /* record number of core channels incase less than max channels are requested */
1684 num_core_channels = s->prim_channels;
1687 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1689 s->core_ext_mask = 0;
1691 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1693 /* only scan for extensions if ext_descr was unknown or indicated a
1694 * supported XCh extension */
1695 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1697 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1698 * extensions scan can fill it up */
1699 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1701 /* extensions start at 32-bit boundaries into bitstream */
1702 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1704 while(core_ss_end - get_bits_count(&s->gb) >= 32) {
1705 uint32_t bits = get_bits_long(&s->gb, 32);
1709 int ext_amode, xch_fsize;
1711 s->xch_base_channel = s->prim_channels;
1713 /* validate sync word using XCHFSIZE field */
1714 xch_fsize = show_bits(&s->gb, 10);
1715 if((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1716 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1719 /* skip length-to-end-of-frame field for the moment */
1720 skip_bits(&s->gb, 10);
1722 s->core_ext_mask |= DCA_EXT_XCH;
1724 /* extension amode should == 1, number of channels in extension */
1725 /* AFAIK XCh is not used for more channels */
1726 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1727 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1728 " supported!\n",ext_amode);
1732 /* much like core primary audio coding header */
1733 dca_parse_audio_coding_header(s, s->xch_base_channel);
1735 for (i = 0; i < (s->sample_blocks / 8); i++) {
1736 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1737 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1746 /* XXCh: extended channels */
1747 /* usually found either in core or HD part in DTS-HD HRA streams,
1748 * but not in DTS-ES which contains XCh extensions instead */
1749 s->core_ext_mask |= DCA_EXT_XXCH;
1753 int fsize96 = show_bits(&s->gb, 12) + 1;
1754 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1757 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n", get_bits_count(&s->gb));
1758 skip_bits(&s->gb, 12);
1759 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1760 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1762 s->core_ext_mask |= DCA_EXT_X96;
1767 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1771 /* no supported extensions, skip the rest of the core substream */
1772 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1775 if (s->core_ext_mask & DCA_EXT_X96)
1776 s->profile = FF_PROFILE_DTS_96_24;
1777 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1778 s->profile = FF_PROFILE_DTS_ES;
1780 /* check for ExSS (HD part) */
1781 if (s->dca_buffer_size - s->frame_size > 32
1782 && get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1783 dca_exss_parse_header(s);
1785 avctx->profile = s->profile;
1787 channels = s->prim_channels + !!s->lfe;
1790 avctx->channel_layout = dca_core_channel_layout[s->amode];
1792 if (s->xch_present && (!avctx->request_channels ||
1793 avctx->request_channels > num_core_channels + !!s->lfe)) {
1794 avctx->channel_layout |= AV_CH_BACK_CENTER;
1796 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1797 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1799 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1802 channels = num_core_channels + !!s->lfe;
1803 s->xch_present = 0; /* disable further xch processing */
1805 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1806 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1808 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1811 if (channels > !!s->lfe &&
1812 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1813 return AVERROR_INVALIDDATA;
1815 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1817 s->output = DCA_STEREO;
1818 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1820 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
1821 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
1822 s->channel_order_tab = dca_channel_order_native;
1825 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
1826 return AVERROR_INVALIDDATA;
1829 if (avctx->channels != channels) {
1830 if (avctx->channels)
1831 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
1832 avctx->channels = channels;
1835 out_size = 256 / 8 * s->sample_blocks * channels *
1836 av_get_bytes_per_sample(avctx->sample_fmt);
1837 if (*data_size < out_size)
1838 return AVERROR(EINVAL);
1839 *data_size = out_size;
1841 /* filter to get final output */
1842 for (i = 0; i < (s->sample_blocks / 8); i++) {
1843 dca_filter_channels(s, i);
1845 /* If this was marked as a DTS-ES stream we need to subtract back- */
1846 /* channel from SL & SR to remove matrixed back-channel signal */
1847 if((s->source_pcm_res & 1) && s->xch_present) {
1848 float* back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1849 float* lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1850 float* rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1851 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1852 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1855 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1856 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1858 samples_flt += 256 * channels;
1860 s->fmt_conv.float_to_int16_interleave(samples_s16,
1861 s->samples_chanptr, 256,
1863 samples_s16 += 256 * channels;
1867 /* update lfe history */
1868 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1869 for (i = 0; i < 2 * s->lfe * 4; i++) {
1870 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1879 * DCA initialization
1881 * @param avctx pointer to the AVCodecContext
1884 static av_cold int dca_decode_init(AVCodecContext * avctx)
1886 DCAContext *s = avctx->priv_data;
1892 dsputil_init(&s->dsp, avctx);
1893 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1894 ff_synth_filter_init(&s->synth);
1895 ff_dcadsp_init(&s->dcadsp);
1896 ff_fmt_convert_init(&s->fmt_conv, avctx);
1898 for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++)
1899 s->samples_chanptr[i] = s->samples + i * 256;
1901 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1902 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1903 s->scale_bias = 1.0 / 32768.0;
1905 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1906 s->scale_bias = 1.0;
1909 /* allow downmixing to stereo */
1910 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1911 avctx->request_channels == 2) {
1912 avctx->channels = avctx->request_channels;
1918 static av_cold int dca_decode_end(AVCodecContext * avctx)
1920 DCAContext *s = avctx->priv_data;
1921 ff_mdct_end(&s->imdct);
1925 static const AVProfile profiles[] = {
1926 { FF_PROFILE_DTS, "DTS" },
1927 { FF_PROFILE_DTS_ES, "DTS-ES" },
1928 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1929 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1930 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1931 { FF_PROFILE_UNKNOWN },
1934 AVCodec ff_dca_decoder = {
1936 .type = AVMEDIA_TYPE_AUDIO,
1938 .priv_data_size = sizeof(DCAContext),
1939 .init = dca_decode_init,
1940 .decode = dca_decode_frame,
1941 .close = dca_decode_end,
1942 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1943 .capabilities = CODEC_CAP_CHANNEL_CONF,
1944 .sample_fmts = (const enum AVSampleFormat[]) {
1945 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
1947 .profiles = NULL_IF_CONFIG_SMALL(profiles),