2 * DCA compatible decoder
3 * Copyright (C) 2004 Gildas Bazin
4 * Copyright (C) 2004 Benjamin Zores
5 * Copyright (C) 2006 Benjamin Larsson
6 * Copyright (C) 2007 Konstantin Shishkov
8 * This file is part of Libav.
10 * Libav is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
15 * Libav is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with Libav; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 #include "libavutil/common.h"
30 #include "libavutil/intmath.h"
31 #include "libavutil/intreadwrite.h"
32 #include "libavutil/mathematics.h"
33 #include "libavutil/audioconvert.h"
42 #include "dca_parser.h"
43 #include "synth_filter.h"
45 #include "fmtconvert.h"
53 #define DCA_PRIM_CHANNELS_MAX (7)
54 #define DCA_SUBBANDS (32)
55 #define DCA_ABITS_MAX (32) /* Should be 28 */
56 #define DCA_SUBSUBFRAMES_MAX (4)
57 #define DCA_SUBFRAMES_MAX (16)
58 #define DCA_BLOCKS_MAX (16)
59 #define DCA_LFE_MAX (3)
75 /* these are unconfirmed but should be mostly correct */
76 enum DCAExSSSpeakerMask {
77 DCA_EXSS_FRONT_CENTER = 0x0001,
78 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
79 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
80 DCA_EXSS_LFE = 0x0008,
81 DCA_EXSS_REAR_CENTER = 0x0010,
82 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
83 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
84 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
85 DCA_EXSS_OVERHEAD = 0x0100,
86 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
87 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
88 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
89 DCA_EXSS_LFE2 = 0x1000,
90 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
91 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
92 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
95 enum DCAExtensionMask {
96 DCA_EXT_CORE = 0x001, ///< core in core substream
97 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
98 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
99 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
100 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
101 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
102 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
103 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
104 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
105 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
108 /* -1 are reserved or unknown */
109 static const int dca_ext_audio_descr_mask[] = {
113 DCA_EXT_XCH | DCA_EXT_X96,
120 /* extensions that reside in core substream */
121 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
123 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
124 * Some compromises have been made for special configurations. Most configurations
125 * are never used so complete accuracy is not needed.
127 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
128 * S -> side, when both rear and back are configured move one of them to the side channel
130 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
132 static const uint64_t dca_core_channel_layout[] = {
133 AV_CH_FRONT_CENTER, ///< 1, A
134 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
135 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
136 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
137 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
138 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
139 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
140 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
141 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
143 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
144 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
146 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
147 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
149 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
150 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
152 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
153 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
154 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
156 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
157 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
158 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
160 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
161 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
162 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
164 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
165 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
166 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
169 static const int8_t dca_lfe_index[] = {
170 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
173 static const int8_t dca_channel_reorder_lfe[][9] = {
174 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
175 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
176 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
177 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
178 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
179 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
180 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
181 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
182 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
183 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
184 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
185 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
186 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
187 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
188 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
189 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
192 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
193 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
194 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
195 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
196 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
197 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
198 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
199 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
200 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
201 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
202 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
203 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
204 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
205 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
206 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
207 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
208 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
211 static const int8_t dca_channel_reorder_nolfe[][9] = {
212 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
213 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
214 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
215 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
216 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
217 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
218 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
219 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
220 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
221 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
222 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
223 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
224 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
225 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
226 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
227 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
230 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
231 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
232 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
233 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
234 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
235 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
236 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
237 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
238 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
239 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
240 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
241 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
242 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
243 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
244 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
245 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
246 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
249 #define DCA_DOLBY 101 /* FIXME */
251 #define DCA_CHANNEL_BITS 6
252 #define DCA_CHANNEL_MASK 0x3F
256 #define HEADER_SIZE 14
258 #define DCA_MAX_FRAME_SIZE 16384
259 #define DCA_MAX_EXSS_HEADER_SIZE 4096
261 #define DCA_BUFFER_PADDING_SIZE 1024
263 /** Bit allocation */
265 int offset; ///< code values offset
266 int maxbits[8]; ///< max bits in VLC
267 int wrap; ///< wrap for get_vlc2()
268 VLC vlc[8]; ///< actual codes
271 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
272 static BitAlloc dca_tmode; ///< transition mode VLCs
273 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
274 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
276 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
279 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
284 AVCodecContext *avctx;
287 int frame_type; ///< type of the current frame
288 int samples_deficit; ///< deficit sample count
289 int crc_present; ///< crc is present in the bitstream
290 int sample_blocks; ///< number of PCM sample blocks
291 int frame_size; ///< primary frame byte size
292 int amode; ///< audio channels arrangement
293 int sample_rate; ///< audio sampling rate
294 int bit_rate; ///< transmission bit rate
295 int bit_rate_index; ///< transmission bit rate index
297 int downmix; ///< embedded downmix enabled
298 int dynrange; ///< embedded dynamic range flag
299 int timestamp; ///< embedded time stamp flag
300 int aux_data; ///< auxiliary data flag
301 int hdcd; ///< source material is mastered in HDCD
302 int ext_descr; ///< extension audio descriptor flag
303 int ext_coding; ///< extended coding flag
304 int aspf; ///< audio sync word insertion flag
305 int lfe; ///< low frequency effects flag
306 int predictor_history; ///< predictor history flag
307 int header_crc; ///< header crc check bytes
308 int multirate_inter; ///< multirate interpolator switch
309 int version; ///< encoder software revision
310 int copy_history; ///< copy history
311 int source_pcm_res; ///< source pcm resolution
312 int front_sum; ///< front sum/difference flag
313 int surround_sum; ///< surround sum/difference flag
314 int dialog_norm; ///< dialog normalisation parameter
316 /* Primary audio coding header */
317 int subframes; ///< number of subframes
318 int is_channels_set; ///< check for if the channel number is already set
319 int total_channels; ///< number of channels including extensions
320 int prim_channels; ///< number of primary audio channels
321 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
322 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
323 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
324 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
325 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
326 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
327 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
328 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
330 /* Primary audio coding side information */
331 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
332 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
333 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
334 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
335 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
336 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
337 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
338 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
339 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
340 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
341 int dynrange_coef; ///< dynamic range coefficient
343 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
345 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
346 int lfe_scale_factor;
348 /* Subband samples history (for ADPCM) */
349 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
350 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
351 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
352 int hist_index[DCA_PRIM_CHANNELS_MAX];
353 DECLARE_ALIGNED(32, float, raXin)[32];
355 int output; ///< type of output
356 float scale_bias; ///< output scale
358 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
359 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX + 1) * 256];
360 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
362 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
363 int dca_buffer_size; ///< how much data is in the dca_buffer
365 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
367 /* Current position in DCA frame */
368 int current_subframe;
369 int current_subsubframe;
371 int core_ext_mask; ///< present extensions in the core substream
373 /* XCh extension information */
374 int xch_present; ///< XCh extension present and valid
375 int xch_base_channel; ///< index of first (only) channel containing XCH data
377 /* ExSS header parser */
378 int static_fields; ///< static fields present
379 int mix_metadata; ///< mixing metadata present
380 int num_mix_configs; ///< number of mix out configurations
381 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
385 int debug_flag; ///< used for suppressing repeated error messages output
388 SynthFilterContext synth;
389 DCADSPContext dcadsp;
390 FmtConvertContext fmt_conv;
393 static const uint16_t dca_vlc_offs[] = {
394 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
395 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
396 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
397 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
398 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
399 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
402 static av_cold void dca_init_vlcs(void)
404 static int vlcs_initialized = 0;
406 static VLC_TYPE dca_table[23622][2];
408 if (vlcs_initialized)
411 dca_bitalloc_index.offset = 1;
412 dca_bitalloc_index.wrap = 2;
413 for (i = 0; i < 5; i++) {
414 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
415 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
416 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
417 bitalloc_12_bits[i], 1, 1,
418 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
420 dca_scalefactor.offset = -64;
421 dca_scalefactor.wrap = 2;
422 for (i = 0; i < 5; i++) {
423 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
424 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
425 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
426 scales_bits[i], 1, 1,
427 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
429 dca_tmode.offset = 0;
431 for (i = 0; i < 4; i++) {
432 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
433 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
434 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
436 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
439 for (i = 0; i < 10; i++)
440 for (j = 0; j < 7; j++) {
441 if (!bitalloc_codes[i][j])
443 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
444 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
445 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
446 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
448 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
450 bitalloc_bits[i][j], 1, 1,
451 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
454 vlcs_initialized = 1;
457 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
460 *dst++ = get_bits(gb, bits);
463 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
466 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
467 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
468 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
470 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
471 s->prim_channels = s->total_channels;
473 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
474 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
477 for (i = base_channel; i < s->prim_channels; i++) {
478 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
479 if (s->subband_activity[i] > DCA_SUBBANDS)
480 s->subband_activity[i] = DCA_SUBBANDS;
482 for (i = base_channel; i < s->prim_channels; i++) {
483 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
484 if (s->vq_start_subband[i] > DCA_SUBBANDS)
485 s->vq_start_subband[i] = DCA_SUBBANDS;
487 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
488 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
489 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
490 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
492 /* Get codebooks quantization indexes */
494 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
495 for (j = 1; j < 11; j++)
496 for (i = base_channel; i < s->prim_channels; i++)
497 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
499 /* Get scale factor adjustment */
500 for (j = 0; j < 11; j++)
501 for (i = base_channel; i < s->prim_channels; i++)
502 s->scalefactor_adj[i][j] = 1;
504 for (j = 1; j < 11; j++)
505 for (i = base_channel; i < s->prim_channels; i++)
506 if (s->quant_index_huffman[i][j] < thr[j])
507 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
509 if (s->crc_present) {
510 /* Audio header CRC check */
511 get_bits(&s->gb, 16);
514 s->current_subframe = 0;
515 s->current_subsubframe = 0;
518 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
519 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
520 for (i = base_channel; i < s->prim_channels; i++) {
521 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
522 s->subband_activity[i]);
523 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
524 s->vq_start_subband[i]);
525 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
526 s->joint_intensity[i]);
527 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
528 s->transient_huffman[i]);
529 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
530 s->scalefactor_huffman[i]);
531 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
532 s->bitalloc_huffman[i]);
533 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
534 for (j = 0; j < 11; j++)
535 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
536 av_log(s->avctx, AV_LOG_DEBUG, "\n");
537 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
538 for (j = 0; j < 11; j++)
539 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
540 av_log(s->avctx, AV_LOG_DEBUG, "\n");
547 static int dca_parse_frame_header(DCAContext *s)
549 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
552 skip_bits_long(&s->gb, 32);
555 s->frame_type = get_bits(&s->gb, 1);
556 s->samples_deficit = get_bits(&s->gb, 5) + 1;
557 s->crc_present = get_bits(&s->gb, 1);
558 s->sample_blocks = get_bits(&s->gb, 7) + 1;
559 s->frame_size = get_bits(&s->gb, 14) + 1;
560 if (s->frame_size < 95)
561 return AVERROR_INVALIDDATA;
562 s->amode = get_bits(&s->gb, 6);
563 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
565 return AVERROR_INVALIDDATA;
566 s->bit_rate_index = get_bits(&s->gb, 5);
567 s->bit_rate = dca_bit_rates[s->bit_rate_index];
569 return AVERROR_INVALIDDATA;
571 s->downmix = get_bits(&s->gb, 1);
572 s->dynrange = get_bits(&s->gb, 1);
573 s->timestamp = get_bits(&s->gb, 1);
574 s->aux_data = get_bits(&s->gb, 1);
575 s->hdcd = get_bits(&s->gb, 1);
576 s->ext_descr = get_bits(&s->gb, 3);
577 s->ext_coding = get_bits(&s->gb, 1);
578 s->aspf = get_bits(&s->gb, 1);
579 s->lfe = get_bits(&s->gb, 2);
580 s->predictor_history = get_bits(&s->gb, 1);
582 /* TODO: check CRC */
584 s->header_crc = get_bits(&s->gb, 16);
586 s->multirate_inter = get_bits(&s->gb, 1);
587 s->version = get_bits(&s->gb, 4);
588 s->copy_history = get_bits(&s->gb, 2);
589 s->source_pcm_res = get_bits(&s->gb, 3);
590 s->front_sum = get_bits(&s->gb, 1);
591 s->surround_sum = get_bits(&s->gb, 1);
592 s->dialog_norm = get_bits(&s->gb, 4);
594 /* FIXME: channels mixing levels */
595 s->output = s->amode;
597 s->output |= DCA_LFE;
600 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
601 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
602 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
603 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
604 s->sample_blocks, s->sample_blocks * 32);
605 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
606 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
607 s->amode, dca_channels[s->amode]);
608 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
610 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
612 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
613 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
614 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
615 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
616 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
617 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
618 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
619 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
620 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
621 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
622 s->predictor_history);
623 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
624 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
626 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
627 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
628 av_log(s->avctx, AV_LOG_DEBUG,
629 "source pcm resolution: %i (%i bits/sample)\n",
630 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
631 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
632 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
633 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
634 av_log(s->avctx, AV_LOG_DEBUG, "\n");
637 /* Primary audio coding header */
638 s->subframes = get_bits(&s->gb, 4) + 1;
640 return dca_parse_audio_coding_header(s, 0);
644 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
647 /* huffman encoded */
648 value += get_bitalloc(gb, &dca_scalefactor, level);
649 value = av_clip(value, 0, (1 << log2range) - 1);
650 } else if (level < 8) {
651 if (level + 1 > log2range) {
652 skip_bits(gb, level + 1 - log2range);
653 value = get_bits(gb, log2range);
655 value = get_bits(gb, level + 1);
661 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
663 /* Primary audio coding side information */
666 if (get_bits_left(&s->gb) < 0)
667 return AVERROR_INVALIDDATA;
670 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
671 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
674 for (j = base_channel; j < s->prim_channels; j++) {
675 for (k = 0; k < s->subband_activity[j]; k++)
676 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
679 /* Get prediction codebook */
680 for (j = base_channel; j < s->prim_channels; j++) {
681 for (k = 0; k < s->subband_activity[j]; k++) {
682 if (s->prediction_mode[j][k] > 0) {
683 /* (Prediction coefficient VQ address) */
684 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
689 /* Bit allocation index */
690 for (j = base_channel; j < s->prim_channels; j++) {
691 for (k = 0; k < s->vq_start_subband[j]; k++) {
692 if (s->bitalloc_huffman[j] == 6)
693 s->bitalloc[j][k] = get_bits(&s->gb, 5);
694 else if (s->bitalloc_huffman[j] == 5)
695 s->bitalloc[j][k] = get_bits(&s->gb, 4);
696 else if (s->bitalloc_huffman[j] == 7) {
697 av_log(s->avctx, AV_LOG_ERROR,
698 "Invalid bit allocation index\n");
699 return AVERROR_INVALIDDATA;
702 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
705 if (s->bitalloc[j][k] > 26) {
706 // av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index [%i][%i] too big (%i)\n",
707 // j, k, s->bitalloc[j][k]);
708 return AVERROR_INVALIDDATA;
713 /* Transition mode */
714 for (j = base_channel; j < s->prim_channels; j++) {
715 for (k = 0; k < s->subband_activity[j]; k++) {
716 s->transition_mode[j][k] = 0;
717 if (s->subsubframes[s->current_subframe] > 1 &&
718 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
719 s->transition_mode[j][k] =
720 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
725 if (get_bits_left(&s->gb) < 0)
726 return AVERROR_INVALIDDATA;
728 for (j = base_channel; j < s->prim_channels; j++) {
729 const uint32_t *scale_table;
730 int scale_sum, log_size;
732 memset(s->scale_factor[j], 0,
733 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
735 if (s->scalefactor_huffman[j] == 6) {
736 scale_table = scale_factor_quant7;
739 scale_table = scale_factor_quant6;
743 /* When huffman coded, only the difference is encoded */
746 for (k = 0; k < s->subband_activity[j]; k++) {
747 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
748 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
749 s->scale_factor[j][k][0] = scale_table[scale_sum];
752 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
753 /* Get second scale factor */
754 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
755 s->scale_factor[j][k][1] = scale_table[scale_sum];
760 /* Joint subband scale factor codebook select */
761 for (j = base_channel; j < s->prim_channels; j++) {
762 /* Transmitted only if joint subband coding enabled */
763 if (s->joint_intensity[j] > 0)
764 s->joint_huff[j] = get_bits(&s->gb, 3);
767 if (get_bits_left(&s->gb) < 0)
768 return AVERROR_INVALIDDATA;
770 /* Scale factors for joint subband coding */
771 for (j = base_channel; j < s->prim_channels; j++) {
774 /* Transmitted only if joint subband coding enabled */
775 if (s->joint_intensity[j] > 0) {
777 source_channel = s->joint_intensity[j] - 1;
779 /* When huffman coded, only the difference is encoded
780 * (is this valid as well for joint scales ???) */
782 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
783 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
784 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
787 if (!(s->debug_flag & 0x02)) {
788 av_log(s->avctx, AV_LOG_DEBUG,
789 "Joint stereo coding not supported\n");
790 s->debug_flag |= 0x02;
795 /* Stereo downmix coefficients */
796 if (!base_channel && s->prim_channels > 2) {
798 for (j = base_channel; j < s->prim_channels; j++) {
799 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
800 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
803 int am = s->amode & DCA_CHANNEL_MASK;
804 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
805 av_log(s->avctx, AV_LOG_ERROR,
806 "Invalid channel mode %d\n", am);
807 return AVERROR_INVALIDDATA;
809 for (j = base_channel; j < s->prim_channels; j++) {
810 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
811 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
816 /* Dynamic range coefficient */
817 if (!base_channel && s->dynrange)
818 s->dynrange_coef = get_bits(&s->gb, 8);
820 /* Side information CRC check word */
821 if (s->crc_present) {
822 get_bits(&s->gb, 16);
826 * Primary audio data arrays
829 /* VQ encoded high frequency subbands */
830 for (j = base_channel; j < s->prim_channels; j++)
831 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
832 /* 1 vector -> 32 samples */
833 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
835 /* Low frequency effect data */
836 if (!base_channel && s->lfe) {
838 int lfe_samples = 2 * s->lfe * (4 + block_index);
839 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
842 for (j = lfe_samples; j < lfe_end_sample; j++) {
843 /* Signed 8 bits int */
844 s->lfe_data[j] = get_sbits(&s->gb, 8);
847 /* Scale factor index */
848 skip_bits(&s->gb, 1);
849 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 7)];
851 /* Quantization step size * scale factor */
852 lfe_scale = 0.035 * s->lfe_scale_factor;
854 for (j = lfe_samples; j < lfe_end_sample; j++)
855 s->lfe_data[j] *= lfe_scale;
859 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
860 s->subsubframes[s->current_subframe]);
861 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
862 s->partial_samples[s->current_subframe]);
864 for (j = base_channel; j < s->prim_channels; j++) {
865 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
866 for (k = 0; k < s->subband_activity[j]; k++)
867 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
868 av_log(s->avctx, AV_LOG_DEBUG, "\n");
870 for (j = base_channel; j < s->prim_channels; j++) {
871 for (k = 0; k < s->subband_activity[j]; k++)
872 av_log(s->avctx, AV_LOG_DEBUG,
873 "prediction coefs: %f, %f, %f, %f\n",
874 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
875 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
876 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
877 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
879 for (j = base_channel; j < s->prim_channels; j++) {
880 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
881 for (k = 0; k < s->vq_start_subband[j]; k++)
882 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
883 av_log(s->avctx, AV_LOG_DEBUG, "\n");
885 for (j = base_channel; j < s->prim_channels; j++) {
886 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
887 for (k = 0; k < s->subband_activity[j]; k++)
888 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
889 av_log(s->avctx, AV_LOG_DEBUG, "\n");
891 for (j = base_channel; j < s->prim_channels; j++) {
892 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
893 for (k = 0; k < s->subband_activity[j]; k++) {
894 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
895 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
896 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
897 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
899 av_log(s->avctx, AV_LOG_DEBUG, "\n");
901 for (j = base_channel; j < s->prim_channels; j++) {
902 if (s->joint_intensity[j] > 0) {
903 int source_channel = s->joint_intensity[j] - 1;
904 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
905 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
906 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
907 av_log(s->avctx, AV_LOG_DEBUG, "\n");
910 if (!base_channel && s->prim_channels > 2 && s->downmix) {
911 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
912 for (j = 0; j < s->prim_channels; j++) {
913 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
914 dca_downmix_coeffs[s->downmix_coef[j][0]]);
915 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
916 dca_downmix_coeffs[s->downmix_coef[j][1]]);
918 av_log(s->avctx, AV_LOG_DEBUG, "\n");
920 for (j = base_channel; j < s->prim_channels; j++)
921 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
922 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
923 if (!base_channel && s->lfe) {
924 int lfe_samples = 2 * s->lfe * (4 + block_index);
925 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
927 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
928 for (j = lfe_samples; j < lfe_end_sample; j++)
929 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
930 av_log(s->avctx, AV_LOG_DEBUG, "\n");
937 static void qmf_32_subbands(DCAContext *s, int chans,
938 float samples_in[32][8], float *samples_out,
941 const float *prCoeff;
944 int sb_act = s->subband_activity[chans];
947 scale *= sqrt(1 / 8.0);
950 if (!s->multirate_inter) /* Non-perfect reconstruction */
951 prCoeff = fir_32bands_nonperfect;
952 else /* Perfect reconstruction */
953 prCoeff = fir_32bands_perfect;
955 for (i = sb_act; i < 32; i++)
958 /* Reconstructed channel sample index */
959 for (subindex = 0; subindex < 8; subindex++) {
960 /* Load in one sample from each subband and clear inactive subbands */
961 for (i = 0; i < sb_act; i++) {
962 unsigned sign = (i - 1) & 2;
963 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
964 AV_WN32A(&s->raXin[i], v);
967 s->synth.synth_filter_float(&s->imdct,
968 s->subband_fir_hist[chans],
969 &s->hist_index[chans],
970 s->subband_fir_noidea[chans], prCoeff,
971 samples_out, s->raXin, scale);
976 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
977 int num_deci_sample, float *samples_in,
978 float *samples_out, float scale)
980 /* samples_in: An array holding decimated samples.
981 * Samples in current subframe starts from samples_in[0],
982 * while samples_in[-1], samples_in[-2], ..., stores samples
983 * from last subframe as history.
985 * samples_out: An array holding interpolated samples
989 const float *prCoeff;
992 /* Select decimation filter */
993 if (decimation_select == 1) {
995 prCoeff = lfe_fir_128;
998 prCoeff = lfe_fir_64;
1001 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1002 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1004 samples_out += 2 * decifactor;
1008 /* downmixing routines */
1009 #define MIX_REAR1(samples, si1, rs, coef) \
1010 samples[i] += samples[si1] * coef[rs][0]; \
1011 samples[i+256] += samples[si1] * coef[rs][1];
1013 #define MIX_REAR2(samples, si1, si2, rs, coef) \
1014 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs + 1][0]; \
1015 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs + 1][1];
1017 #define MIX_FRONT3(samples, coef) \
1018 t = samples[i + c]; \
1019 u = samples[i + l]; \
1020 v = samples[i + r]; \
1021 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1022 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1024 #define DOWNMIX_TO_STEREO(op1, op2) \
1025 for (i = 0; i < 256; i++) { \
1030 static void dca_downmix(float *samples, int srcfmt,
1031 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1032 const int8_t *channel_mapping)
1034 int c, l, r, sl, sr, s;
1037 float coef[DCA_PRIM_CHANNELS_MAX][2];
1039 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1040 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1041 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1047 case DCA_STEREO_TOTAL:
1048 case DCA_STEREO_SUMDIFF:
1050 av_log(NULL, 0, "Not implemented!\n");
1055 c = channel_mapping[0] * 256;
1056 l = channel_mapping[1] * 256;
1057 r = channel_mapping[2] * 256;
1058 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1061 s = channel_mapping[2] * 256;
1062 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef), );
1065 c = channel_mapping[0] * 256;
1066 l = channel_mapping[1] * 256;
1067 r = channel_mapping[2] * 256;
1068 s = channel_mapping[3] * 256;
1069 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1070 MIX_REAR1(samples, i + s, 3, coef));
1073 sl = channel_mapping[2] * 256;
1074 sr = channel_mapping[3] * 256;
1075 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef), );
1078 c = channel_mapping[0] * 256;
1079 l = channel_mapping[1] * 256;
1080 r = channel_mapping[2] * 256;
1081 sl = channel_mapping[3] * 256;
1082 sr = channel_mapping[4] * 256;
1083 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1084 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1090 #ifndef decode_blockcodes
1091 /* Very compact version of the block code decoder that does not use table
1092 * look-up but is slightly slower */
1093 static int decode_blockcode(int code, int levels, int *values)
1096 int offset = (levels - 1) >> 1;
1098 for (i = 0; i < 4; i++) {
1099 int div = FASTDIV(code, levels);
1100 values[i] = code - offset - div * levels;
1107 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1109 return decode_blockcode(code1, levels, values) |
1110 decode_blockcode(code2, levels, values + 4);
1114 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1115 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1117 #ifndef int8x8_fmul_int32
1118 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1120 float fscale = scale / 16.0;
1122 for (i = 0; i < 8; i++)
1123 dst[i] = src[i] * fscale;
1127 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1130 int subsubframe = s->current_subsubframe;
1132 const float *quant_step_table;
1135 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1136 LOCAL_ALIGNED_16(int, block, [8]);
1142 /* Select quantization step size table */
1143 if (s->bit_rate_index == 0x1f)
1144 quant_step_table = lossless_quant_d;
1146 quant_step_table = lossy_quant_d;
1148 for (k = base_channel; k < s->prim_channels; k++) {
1149 if (get_bits_left(&s->gb) < 0)
1150 return AVERROR_INVALIDDATA;
1152 for (l = 0; l < s->vq_start_subband[k]; l++) {
1155 /* Select the mid-tread linear quantizer */
1156 int abits = s->bitalloc[k][l];
1158 float quant_step_size = quant_step_table[abits];
1161 * Determine quantization index code book and its type
1164 /* Select quantization index code book */
1165 int sel = s->quant_index_huffman[k][abits];
1168 * Extract bits from the bit stream
1171 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1173 /* Deal with transients */
1174 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1175 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1176 s->scalefactor_adj[k][sel];
1178 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1181 int block_code1, block_code2, size, levels, err;
1183 size = abits_sizes[abits - 1];
1184 levels = abits_levels[abits - 1];
1186 block_code1 = get_bits(&s->gb, size);
1187 block_code2 = get_bits(&s->gb, size);
1188 err = decode_blockcodes(block_code1, block_code2,
1191 av_log(s->avctx, AV_LOG_ERROR,
1192 "ERROR: block code look-up failed\n");
1193 return AVERROR_INVALIDDATA;
1197 for (m = 0; m < 8; m++)
1198 block[m] = get_sbits(&s->gb, abits - 3);
1202 for (m = 0; m < 8; m++)
1203 block[m] = get_bitalloc(&s->gb,
1204 &dca_smpl_bitalloc[abits], sel);
1207 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1212 * Inverse ADPCM if in prediction mode
1214 if (s->prediction_mode[k][l]) {
1216 for (m = 0; m < 8; m++) {
1217 for (n = 1; n <= 4; n++)
1219 subband_samples[k][l][m] +=
1220 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1221 subband_samples[k][l][m - n] / 8192);
1222 else if (s->predictor_history)
1223 subband_samples[k][l][m] +=
1224 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1225 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1231 * Decode VQ encoded high frequencies
1233 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1234 /* 1 vector -> 32 samples but we only need the 8 samples
1235 * for this subsubframe. */
1236 int hfvq = s->high_freq_vq[k][l];
1238 if (!s->debug_flag & 0x01) {
1239 av_log(s->avctx, AV_LOG_DEBUG,
1240 "Stream with high frequencies VQ coding\n");
1241 s->debug_flag |= 0x01;
1244 int8x8_fmul_int32(subband_samples[k][l],
1245 &high_freq_vq[hfvq][subsubframe * 8],
1246 s->scale_factor[k][l][0]);
1250 /* Check for DSYNC after subsubframe */
1251 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1252 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1254 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1257 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1261 /* Backup predictor history for adpcm */
1262 for (k = base_channel; k < s->prim_channels; k++)
1263 for (l = 0; l < s->vq_start_subband[k]; l++)
1264 memcpy(s->subband_samples_hist[k][l],
1265 &subband_samples[k][l][4],
1266 4 * sizeof(subband_samples[0][0][0]));
1271 static int dca_filter_channels(DCAContext *s, int block_index)
1273 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1276 /* 32 subbands QMF */
1277 for (k = 0; k < s->prim_channels; k++) {
1278 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1279 0, 8388608.0, 8388608.0 };*/
1280 qmf_32_subbands(s, k, subband_samples[k],
1281 &s->samples[256 * s->channel_order_tab[k]],
1282 M_SQRT1_2 * s->scale_bias /* pcm_to_double[s->source_pcm_res] */);
1286 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1287 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1290 /* Generate LFE samples for this subsubframe FIXME!!! */
1291 if (s->output & DCA_LFE) {
1292 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1293 s->lfe_data + 2 * s->lfe * (block_index + 4),
1294 &s->samples[256 * dca_lfe_index[s->amode]],
1295 (1.0 / 256.0) * s->scale_bias);
1296 /* Outputs 20bits pcm samples */
1303 static int dca_subframe_footer(DCAContext *s, int base_channel)
1305 int aux_data_count = 0, i;
1308 * Unpack optional information
1311 /* presumably optional information only appears in the core? */
1312 if (!base_channel) {
1314 skip_bits_long(&s->gb, 32);
1317 aux_data_count = get_bits(&s->gb, 6);
1319 for (i = 0; i < aux_data_count; i++)
1320 get_bits(&s->gb, 8);
1322 if (s->crc_present && (s->downmix || s->dynrange))
1323 get_bits(&s->gb, 16);
1330 * Decode a dca frame block
1332 * @param s pointer to the DCAContext
1335 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1340 if (s->current_subframe >= s->subframes) {
1341 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1342 s->current_subframe, s->subframes);
1343 return AVERROR_INVALIDDATA;
1346 if (!s->current_subsubframe) {
1348 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1350 /* Read subframe header */
1351 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1355 /* Read subsubframe */
1357 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1359 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1363 s->current_subsubframe++;
1364 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1365 s->current_subsubframe = 0;
1366 s->current_subframe++;
1368 if (s->current_subframe >= s->subframes) {
1370 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1372 /* Read subframe footer */
1373 if ((ret = dca_subframe_footer(s, base_channel)))
1381 * Return the number of channels in an ExSS speaker mask (HD)
1383 static int dca_exss_mask2count(int mask)
1385 /* count bits that mean speaker pairs twice */
1386 return av_popcount(mask) +
1387 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1388 DCA_EXSS_FRONT_LEFT_RIGHT |
1389 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1390 DCA_EXSS_WIDE_LEFT_RIGHT |
1391 DCA_EXSS_SIDE_LEFT_RIGHT |
1392 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1393 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1394 DCA_EXSS_REAR_LEFT_RIGHT |
1395 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1399 * Skip mixing coefficients of a single mix out configuration (HD)
1401 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1405 for (i = 0; i < channels; i++) {
1406 int mix_map_mask = get_bits(gb, out_ch);
1407 int num_coeffs = av_popcount(mix_map_mask);
1408 skip_bits_long(gb, num_coeffs * 6);
1413 * Parse extension substream asset header (HD)
1415 static int dca_exss_parse_asset_header(DCAContext *s)
1417 int header_pos = get_bits_count(&s->gb);
1420 int embedded_stereo = 0;
1421 int embedded_6ch = 0;
1422 int drc_code_present;
1423 int extensions_mask;
1426 if (get_bits_left(&s->gb) < 16)
1429 /* We will parse just enough to get to the extensions bitmask with which
1430 * we can set the profile value. */
1432 header_size = get_bits(&s->gb, 9) + 1;
1433 skip_bits(&s->gb, 3); // asset index
1435 if (s->static_fields) {
1436 if (get_bits1(&s->gb))
1437 skip_bits(&s->gb, 4); // asset type descriptor
1438 if (get_bits1(&s->gb))
1439 skip_bits_long(&s->gb, 24); // language descriptor
1441 if (get_bits1(&s->gb)) {
1442 /* How can one fit 1024 bytes of text here if the maximum value
1443 * for the asset header size field above was 512 bytes? */
1444 int text_length = get_bits(&s->gb, 10) + 1;
1445 if (get_bits_left(&s->gb) < text_length * 8)
1447 skip_bits_long(&s->gb, text_length * 8); // info text
1450 skip_bits(&s->gb, 5); // bit resolution - 1
1451 skip_bits(&s->gb, 4); // max sample rate code
1452 channels = get_bits(&s->gb, 8) + 1;
1454 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1455 int spkr_remap_sets;
1456 int spkr_mask_size = 16;
1460 embedded_stereo = get_bits1(&s->gb);
1462 embedded_6ch = get_bits1(&s->gb);
1464 if (get_bits1(&s->gb)) {
1465 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1466 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1469 spkr_remap_sets = get_bits(&s->gb, 3);
1471 for (i = 0; i < spkr_remap_sets; i++) {
1472 /* std layout mask for each remap set */
1473 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1476 for (i = 0; i < spkr_remap_sets; i++) {
1477 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1478 if (get_bits_left(&s->gb) < 0)
1481 for (j = 0; j < num_spkrs[i]; j++) {
1482 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1483 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1484 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1489 skip_bits(&s->gb, 3); // representation type
1493 drc_code_present = get_bits1(&s->gb);
1494 if (drc_code_present)
1495 get_bits(&s->gb, 8); // drc code
1497 if (get_bits1(&s->gb))
1498 skip_bits(&s->gb, 5); // dialog normalization code
1500 if (drc_code_present && embedded_stereo)
1501 get_bits(&s->gb, 8); // drc stereo code
1503 if (s->mix_metadata && get_bits1(&s->gb)) {
1504 skip_bits(&s->gb, 1); // external mix
1505 skip_bits(&s->gb, 6); // post mix gain code
1507 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1508 skip_bits(&s->gb, 3); // drc limit
1510 skip_bits(&s->gb, 8); // custom drc code
1512 if (get_bits1(&s->gb)) // channel specific scaling
1513 for (i = 0; i < s->num_mix_configs; i++)
1514 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1516 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1518 for (i = 0; i < s->num_mix_configs; i++) {
1519 if (get_bits_left(&s->gb) < 0)
1521 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1523 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1524 if (embedded_stereo)
1525 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1529 switch (get_bits(&s->gb, 2)) {
1530 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1531 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1532 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1533 case 3: extensions_mask = 0; /* aux coding */ break;
1536 /* not parsed further, we were only interested in the extensions mask */
1538 if (get_bits_left(&s->gb) < 0)
1541 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1542 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1545 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1547 if (extensions_mask & DCA_EXT_EXSS_XLL)
1548 s->profile = FF_PROFILE_DTS_HD_MA;
1549 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1551 s->profile = FF_PROFILE_DTS_HD_HRA;
1553 if (!(extensions_mask & DCA_EXT_CORE))
1554 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1555 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1556 av_log(s->avctx, AV_LOG_WARNING,
1557 "DTS extensions detection mismatch (%d, %d)\n",
1558 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1564 * Parse extension substream header (HD)
1566 static void dca_exss_parse_header(DCAContext *s)
1572 int active_ss_mask[8];
1575 if (get_bits_left(&s->gb) < 52)
1578 skip_bits(&s->gb, 8); // user data
1579 ss_index = get_bits(&s->gb, 2);
1581 blownup = get_bits1(&s->gb);
1582 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1583 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1585 s->static_fields = get_bits1(&s->gb);
1586 if (s->static_fields) {
1587 skip_bits(&s->gb, 2); // reference clock code
1588 skip_bits(&s->gb, 3); // frame duration code
1590 if (get_bits1(&s->gb))
1591 skip_bits_long(&s->gb, 36); // timestamp
1593 /* a single stream can contain multiple audio assets that can be
1594 * combined to form multiple audio presentations */
1596 num_audiop = get_bits(&s->gb, 3) + 1;
1597 if (num_audiop > 1) {
1598 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1599 /* ignore such streams for now */
1603 num_assets = get_bits(&s->gb, 3) + 1;
1604 if (num_assets > 1) {
1605 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1606 /* ignore such streams for now */
1610 for (i = 0; i < num_audiop; i++)
1611 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1613 for (i = 0; i < num_audiop; i++)
1614 for (j = 0; j <= ss_index; j++)
1615 if (active_ss_mask[i] & (1 << j))
1616 skip_bits(&s->gb, 8); // active asset mask
1618 s->mix_metadata = get_bits1(&s->gb);
1619 if (s->mix_metadata) {
1620 int mix_out_mask_size;
1622 skip_bits(&s->gb, 2); // adjustment level
1623 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1624 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1626 for (i = 0; i < s->num_mix_configs; i++) {
1627 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1628 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1633 for (i = 0; i < num_assets; i++)
1634 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1636 for (i = 0; i < num_assets; i++) {
1637 if (dca_exss_parse_asset_header(s))
1641 /* not parsed further, we were only interested in the extensions mask
1642 * from the asset header */
1646 * Main frame decoding function
1647 * FIXME add arguments
1649 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1650 int *got_frame_ptr, AVPacket *avpkt)
1652 const uint8_t *buf = avpkt->data;
1653 int buf_size = avpkt->size;
1656 int num_core_channels = 0;
1659 int16_t *samples_s16;
1660 DCAContext *s = avctx->priv_data;
1667 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1668 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1669 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1670 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1671 return AVERROR_INVALIDDATA;
1674 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1675 if ((ret = dca_parse_frame_header(s)) < 0) {
1676 //seems like the frame is corrupt, try with the next one
1679 //set AVCodec values with parsed data
1680 avctx->sample_rate = s->sample_rate;
1681 avctx->bit_rate = s->bit_rate;
1683 s->profile = FF_PROFILE_DTS;
1685 for (i = 0; i < (s->sample_blocks / 8); i++) {
1686 if ((ret = dca_decode_block(s, 0, i))) {
1687 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1692 /* record number of core channels incase less than max channels are requested */
1693 num_core_channels = s->prim_channels;
1696 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1698 s->core_ext_mask = 0;
1700 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1702 /* only scan for extensions if ext_descr was unknown or indicated a
1703 * supported XCh extension */
1704 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1706 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1707 * extensions scan can fill it up */
1708 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1710 /* extensions start at 32-bit boundaries into bitstream */
1711 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1713 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1714 uint32_t bits = get_bits_long(&s->gb, 32);
1718 int ext_amode, xch_fsize;
1720 s->xch_base_channel = s->prim_channels;
1722 /* validate sync word using XCHFSIZE field */
1723 xch_fsize = show_bits(&s->gb, 10);
1724 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1725 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1728 /* skip length-to-end-of-frame field for the moment */
1729 skip_bits(&s->gb, 10);
1731 s->core_ext_mask |= DCA_EXT_XCH;
1733 /* extension amode(number of channels in extension) should be 1 */
1734 /* AFAIK XCh is not used for more channels */
1735 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1736 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1737 " supported!\n", ext_amode);
1741 /* much like core primary audio coding header */
1742 dca_parse_audio_coding_header(s, s->xch_base_channel);
1744 for (i = 0; i < (s->sample_blocks / 8); i++)
1745 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1746 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1754 /* XXCh: extended channels */
1755 /* usually found either in core or HD part in DTS-HD HRA streams,
1756 * but not in DTS-ES which contains XCh extensions instead */
1757 s->core_ext_mask |= DCA_EXT_XXCH;
1761 int fsize96 = show_bits(&s->gb, 12) + 1;
1762 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1765 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1766 get_bits_count(&s->gb));
1767 skip_bits(&s->gb, 12);
1768 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1769 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1771 s->core_ext_mask |= DCA_EXT_X96;
1776 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1779 /* no supported extensions, skip the rest of the core substream */
1780 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1783 if (s->core_ext_mask & DCA_EXT_X96)
1784 s->profile = FF_PROFILE_DTS_96_24;
1785 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1786 s->profile = FF_PROFILE_DTS_ES;
1788 /* check for ExSS (HD part) */
1789 if (s->dca_buffer_size - s->frame_size > 32 &&
1790 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1791 dca_exss_parse_header(s);
1793 avctx->profile = s->profile;
1795 channels = s->prim_channels + !!s->lfe;
1797 if (s->amode < 16) {
1798 avctx->channel_layout = dca_core_channel_layout[s->amode];
1800 if (s->xch_present && (!avctx->request_channels ||
1801 avctx->request_channels > num_core_channels + !!s->lfe)) {
1802 avctx->channel_layout |= AV_CH_BACK_CENTER;
1804 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1805 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1807 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1810 channels = num_core_channels + !!s->lfe;
1811 s->xch_present = 0; /* disable further xch processing */
1813 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1814 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1816 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1819 if (channels > !!s->lfe &&
1820 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1821 return AVERROR_INVALIDDATA;
1823 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1825 s->output = DCA_STEREO;
1826 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1829 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1830 return AVERROR_INVALIDDATA;
1834 /* There is nothing that prevents a dts frame to change channel configuration
1835 but Libav doesn't support that so only set the channels if it is previously
1836 unset. Ideally during the first probe for channels the crc should be checked
1837 and only set avctx->channels when the crc is ok. Right now the decoder could
1838 set the channels based on a broken first frame.*/
1839 if (s->is_channels_set == 0) {
1840 s->is_channels_set = 1;
1841 avctx->channels = channels;
1843 if (avctx->channels != channels) {
1844 av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of "
1845 "channels changing in stream. Skipping frame.\n");
1846 return AVERROR_PATCHWELCOME;
1849 /* get output buffer */
1850 s->frame.nb_samples = 256 * (s->sample_blocks / 8);
1851 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
1852 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1855 samples_flt = (float *) s->frame.data[0];
1856 samples_s16 = (int16_t *) s->frame.data[0];
1858 /* filter to get final output */
1859 for (i = 0; i < (s->sample_blocks / 8); i++) {
1860 dca_filter_channels(s, i);
1862 /* If this was marked as a DTS-ES stream we need to subtract back- */
1863 /* channel from SL & SR to remove matrixed back-channel signal */
1864 if ((s->source_pcm_res & 1) && s->xch_present) {
1865 float *back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1866 float *lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1867 float *rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1868 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1869 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1872 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1873 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1875 samples_flt += 256 * channels;
1877 s->fmt_conv.float_to_int16_interleave(samples_s16,
1878 s->samples_chanptr, 256,
1880 samples_s16 += 256 * channels;
1884 /* update lfe history */
1885 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1886 for (i = 0; i < 2 * s->lfe * 4; i++)
1887 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1890 *(AVFrame *) data = s->frame;
1898 * DCA initialization
1900 * @param avctx pointer to the AVCodecContext
1903 static av_cold int dca_decode_init(AVCodecContext *avctx)
1905 DCAContext *s = avctx->priv_data;
1911 ff_dsputil_init(&s->dsp, avctx);
1912 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1913 ff_synth_filter_init(&s->synth);
1914 ff_dcadsp_init(&s->dcadsp);
1915 ff_fmt_convert_init(&s->fmt_conv, avctx);
1917 for (i = 0; i < DCA_PRIM_CHANNELS_MAX + 1; i++)
1918 s->samples_chanptr[i] = s->samples + i * 256;
1920 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1921 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1922 s->scale_bias = 1.0 / 32768.0;
1924 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1925 s->scale_bias = 1.0;
1928 /* allow downmixing to stereo */
1929 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1930 avctx->request_channels == 2) {
1931 avctx->channels = avctx->request_channels;
1934 avcodec_get_frame_defaults(&s->frame);
1935 avctx->coded_frame = &s->frame;
1940 static av_cold int dca_decode_end(AVCodecContext *avctx)
1942 DCAContext *s = avctx->priv_data;
1943 ff_mdct_end(&s->imdct);
1947 static const AVProfile profiles[] = {
1948 { FF_PROFILE_DTS, "DTS" },
1949 { FF_PROFILE_DTS_ES, "DTS-ES" },
1950 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1951 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1952 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1953 { FF_PROFILE_UNKNOWN },
1956 AVCodec ff_dca_decoder = {
1958 .type = AVMEDIA_TYPE_AUDIO,
1960 .priv_data_size = sizeof(DCAContext),
1961 .init = dca_decode_init,
1962 .decode = dca_decode_frame,
1963 .close = dca_decode_end,
1964 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1965 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1966 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
1968 AV_SAMPLE_FMT_NONE },
1969 .profiles = NULL_IF_CONFIG_SMALL(profiles),