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/channel_layout.h"
30 #include "libavutil/common.h"
31 #include "libavutil/float_dsp.h"
32 #include "libavutil/intmath.h"
33 #include "libavutil/intreadwrite.h"
34 #include "libavutil/mathematics.h"
35 #include "libavutil/samplefmt.h"
44 #include "dca_parser.h"
45 #include "synth_filter.h"
47 #include "fmtconvert.h"
56 #define DCA_PRIM_CHANNELS_MAX (7)
57 #define DCA_SUBBANDS (32)
58 #define DCA_ABITS_MAX (32) /* Should be 28 */
59 #define DCA_SUBSUBFRAMES_MAX (4)
60 #define DCA_SUBFRAMES_MAX (16)
61 #define DCA_BLOCKS_MAX (16)
62 #define DCA_LFE_MAX (3)
78 /* these are unconfirmed but should be mostly correct */
79 enum DCAExSSSpeakerMask {
80 DCA_EXSS_FRONT_CENTER = 0x0001,
81 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
82 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
83 DCA_EXSS_LFE = 0x0008,
84 DCA_EXSS_REAR_CENTER = 0x0010,
85 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
86 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
87 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
88 DCA_EXSS_OVERHEAD = 0x0100,
89 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
90 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
91 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
92 DCA_EXSS_LFE2 = 0x1000,
93 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
94 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
95 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
98 enum DCAExtensionMask {
99 DCA_EXT_CORE = 0x001, ///< core in core substream
100 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
101 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
102 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
103 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
104 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
105 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
106 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
107 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
108 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
111 /* -1 are reserved or unknown */
112 static const int dca_ext_audio_descr_mask[] = {
116 DCA_EXT_XCH | DCA_EXT_X96,
123 /* extensions that reside in core substream */
124 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
126 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
127 * Some compromises have been made for special configurations. Most configurations
128 * are never used so complete accuracy is not needed.
130 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
131 * S -> side, when both rear and back are configured move one of them to the side channel
133 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
135 static const uint64_t dca_core_channel_layout[] = {
136 AV_CH_FRONT_CENTER, ///< 1, A
137 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
138 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
139 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
140 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
141 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
142 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
143 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
144 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
146 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
147 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
149 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
150 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
152 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
153 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
155 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
156 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
157 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
159 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
160 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
161 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
163 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
164 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
165 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
167 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
168 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
169 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
172 static const int8_t dca_lfe_index[] = {
173 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
176 static const int8_t dca_channel_reorder_lfe[][9] = {
177 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
178 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
179 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
180 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
181 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
182 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
183 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
184 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
185 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
186 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
187 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
188 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
189 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
190 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
191 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
192 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
195 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
196 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
197 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
198 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
199 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
200 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
201 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
202 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
203 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
204 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
205 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
206 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
207 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
208 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
209 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
210 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
211 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
214 static const int8_t dca_channel_reorder_nolfe[][9] = {
215 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
216 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
217 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
218 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
219 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
220 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
221 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
222 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
223 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
224 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
225 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
226 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
227 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
228 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
229 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
230 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
233 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
234 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
235 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
236 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
237 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
238 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
239 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
240 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
241 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
242 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
243 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
244 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
245 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
246 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
247 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
248 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
249 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
252 #define DCA_DOLBY 101 /* FIXME */
254 #define DCA_CHANNEL_BITS 6
255 #define DCA_CHANNEL_MASK 0x3F
259 #define HEADER_SIZE 14
261 #define DCA_MAX_FRAME_SIZE 16384
262 #define DCA_MAX_EXSS_HEADER_SIZE 4096
264 #define DCA_BUFFER_PADDING_SIZE 1024
266 /** Bit allocation */
268 int offset; ///< code values offset
269 int maxbits[8]; ///< max bits in VLC
270 int wrap; ///< wrap for get_vlc2()
271 VLC vlc[8]; ///< actual codes
274 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
275 static BitAlloc dca_tmode; ///< transition mode VLCs
276 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
277 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
279 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
282 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
287 AVCodecContext *avctx;
290 int frame_type; ///< type of the current frame
291 int samples_deficit; ///< deficit sample count
292 int crc_present; ///< crc is present in the bitstream
293 int sample_blocks; ///< number of PCM sample blocks
294 int frame_size; ///< primary frame byte size
295 int amode; ///< audio channels arrangement
296 int sample_rate; ///< audio sampling rate
297 int bit_rate; ///< transmission bit rate
298 int bit_rate_index; ///< transmission bit rate index
300 int downmix; ///< embedded downmix enabled
301 int dynrange; ///< embedded dynamic range flag
302 int timestamp; ///< embedded time stamp flag
303 int aux_data; ///< auxiliary data flag
304 int hdcd; ///< source material is mastered in HDCD
305 int ext_descr; ///< extension audio descriptor flag
306 int ext_coding; ///< extended coding flag
307 int aspf; ///< audio sync word insertion flag
308 int lfe; ///< low frequency effects flag
309 int predictor_history; ///< predictor history flag
310 int header_crc; ///< header crc check bytes
311 int multirate_inter; ///< multirate interpolator switch
312 int version; ///< encoder software revision
313 int copy_history; ///< copy history
314 int source_pcm_res; ///< source pcm resolution
315 int front_sum; ///< front sum/difference flag
316 int surround_sum; ///< surround sum/difference flag
317 int dialog_norm; ///< dialog normalisation parameter
319 /* Primary audio coding header */
320 int subframes; ///< number of subframes
321 int total_channels; ///< number of channels including extensions
322 int prim_channels; ///< number of primary audio channels
323 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
324 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
325 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
326 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
327 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
328 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
329 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
330 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
332 /* Primary audio coding side information */
333 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
334 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
335 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
336 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
337 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
338 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
339 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
340 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
341 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
342 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
343 int dynrange_coef; ///< dynamic range coefficient
345 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
347 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
348 int lfe_scale_factor;
350 /* Subband samples history (for ADPCM) */
351 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
352 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
353 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
354 int hist_index[DCA_PRIM_CHANNELS_MAX];
355 DECLARE_ALIGNED(32, float, raXin)[32];
357 int output; ///< type of output
359 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
360 float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
361 float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
362 uint8_t *extra_channels_buffer;
363 unsigned int extra_channels_buffer_size;
365 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
366 int dca_buffer_size; ///< how much data is in the dca_buffer
368 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
370 /* Current position in DCA frame */
371 int current_subframe;
372 int current_subsubframe;
374 int core_ext_mask; ///< present extensions in the core substream
376 /* XCh extension information */
377 int xch_present; ///< XCh extension present and valid
378 int xch_base_channel; ///< index of first (only) channel containing XCH data
380 /* ExSS header parser */
381 int static_fields; ///< static fields present
382 int mix_metadata; ///< mixing metadata present
383 int num_mix_configs; ///< number of mix out configurations
384 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
388 int debug_flag; ///< used for suppressing repeated error messages output
389 AVFloatDSPContext fdsp;
391 SynthFilterContext synth;
392 DCADSPContext dcadsp;
393 FmtConvertContext fmt_conv;
396 static const uint16_t dca_vlc_offs[] = {
397 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
398 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
399 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
400 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
401 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
402 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
405 static av_cold void dca_init_vlcs(void)
407 static int vlcs_initialized = 0;
409 static VLC_TYPE dca_table[23622][2];
411 if (vlcs_initialized)
414 dca_bitalloc_index.offset = 1;
415 dca_bitalloc_index.wrap = 2;
416 for (i = 0; i < 5; i++) {
417 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
418 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
419 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
420 bitalloc_12_bits[i], 1, 1,
421 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
423 dca_scalefactor.offset = -64;
424 dca_scalefactor.wrap = 2;
425 for (i = 0; i < 5; i++) {
426 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
427 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
428 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
429 scales_bits[i], 1, 1,
430 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
432 dca_tmode.offset = 0;
434 for (i = 0; i < 4; i++) {
435 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
436 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
437 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
439 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
442 for (i = 0; i < 10; i++)
443 for (j = 0; j < 7; j++) {
444 if (!bitalloc_codes[i][j])
446 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
447 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
448 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
449 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
451 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
453 bitalloc_bits[i][j], 1, 1,
454 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
457 vlcs_initialized = 1;
460 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
463 *dst++ = get_bits(gb, bits);
466 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
469 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
470 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
471 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
473 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
474 s->prim_channels = s->total_channels;
476 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
477 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
480 for (i = base_channel; i < s->prim_channels; i++) {
481 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
482 if (s->subband_activity[i] > DCA_SUBBANDS)
483 s->subband_activity[i] = DCA_SUBBANDS;
485 for (i = base_channel; i < s->prim_channels; i++) {
486 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
487 if (s->vq_start_subband[i] > DCA_SUBBANDS)
488 s->vq_start_subband[i] = DCA_SUBBANDS;
490 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
491 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
492 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
493 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
495 /* Get codebooks quantization indexes */
497 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
498 for (j = 1; j < 11; j++)
499 for (i = base_channel; i < s->prim_channels; i++)
500 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
502 /* Get scale factor adjustment */
503 for (j = 0; j < 11; j++)
504 for (i = base_channel; i < s->prim_channels; i++)
505 s->scalefactor_adj[i][j] = 1;
507 for (j = 1; j < 11; j++)
508 for (i = base_channel; i < s->prim_channels; i++)
509 if (s->quant_index_huffman[i][j] < thr[j])
510 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
512 if (s->crc_present) {
513 /* Audio header CRC check */
514 get_bits(&s->gb, 16);
517 s->current_subframe = 0;
518 s->current_subsubframe = 0;
521 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
522 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
523 for (i = base_channel; i < s->prim_channels; i++) {
524 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
525 s->subband_activity[i]);
526 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
527 s->vq_start_subband[i]);
528 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
529 s->joint_intensity[i]);
530 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
531 s->transient_huffman[i]);
532 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
533 s->scalefactor_huffman[i]);
534 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
535 s->bitalloc_huffman[i]);
536 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
537 for (j = 0; j < 11; j++)
538 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
539 av_log(s->avctx, AV_LOG_DEBUG, "\n");
540 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
541 for (j = 0; j < 11; j++)
542 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
543 av_log(s->avctx, AV_LOG_DEBUG, "\n");
550 static int dca_parse_frame_header(DCAContext *s)
552 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
555 skip_bits_long(&s->gb, 32);
558 s->frame_type = get_bits(&s->gb, 1);
559 s->samples_deficit = get_bits(&s->gb, 5) + 1;
560 s->crc_present = get_bits(&s->gb, 1);
561 s->sample_blocks = get_bits(&s->gb, 7) + 1;
562 s->frame_size = get_bits(&s->gb, 14) + 1;
563 if (s->frame_size < 95)
564 return AVERROR_INVALIDDATA;
565 s->amode = get_bits(&s->gb, 6);
566 s->sample_rate = avpriv_dca_sample_rates[get_bits(&s->gb, 4)];
568 return AVERROR_INVALIDDATA;
569 s->bit_rate_index = get_bits(&s->gb, 5);
570 s->bit_rate = dca_bit_rates[s->bit_rate_index];
572 return AVERROR_INVALIDDATA;
574 s->downmix = get_bits(&s->gb, 1);
575 s->dynrange = get_bits(&s->gb, 1);
576 s->timestamp = get_bits(&s->gb, 1);
577 s->aux_data = get_bits(&s->gb, 1);
578 s->hdcd = get_bits(&s->gb, 1);
579 s->ext_descr = get_bits(&s->gb, 3);
580 s->ext_coding = get_bits(&s->gb, 1);
581 s->aspf = get_bits(&s->gb, 1);
582 s->lfe = get_bits(&s->gb, 2);
583 s->predictor_history = get_bits(&s->gb, 1);
585 /* TODO: check CRC */
587 s->header_crc = get_bits(&s->gb, 16);
589 s->multirate_inter = get_bits(&s->gb, 1);
590 s->version = get_bits(&s->gb, 4);
591 s->copy_history = get_bits(&s->gb, 2);
592 s->source_pcm_res = get_bits(&s->gb, 3);
593 s->front_sum = get_bits(&s->gb, 1);
594 s->surround_sum = get_bits(&s->gb, 1);
595 s->dialog_norm = get_bits(&s->gb, 4);
597 /* FIXME: channels mixing levels */
598 s->output = s->amode;
600 s->output |= DCA_LFE;
603 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
604 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
605 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
606 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
607 s->sample_blocks, s->sample_blocks * 32);
608 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
609 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
610 s->amode, dca_channels[s->amode]);
611 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
613 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
615 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
616 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
617 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
618 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
619 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
620 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
621 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
622 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
623 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
624 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
625 s->predictor_history);
626 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
627 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
629 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
630 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
631 av_log(s->avctx, AV_LOG_DEBUG,
632 "source pcm resolution: %i (%i bits/sample)\n",
633 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
634 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
635 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
636 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
637 av_log(s->avctx, AV_LOG_DEBUG, "\n");
640 /* Primary audio coding header */
641 s->subframes = get_bits(&s->gb, 4) + 1;
643 return dca_parse_audio_coding_header(s, 0);
647 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
650 /* huffman encoded */
651 value += get_bitalloc(gb, &dca_scalefactor, level);
652 value = av_clip(value, 0, (1 << log2range) - 1);
653 } else if (level < 8) {
654 if (level + 1 > log2range) {
655 skip_bits(gb, level + 1 - log2range);
656 value = get_bits(gb, log2range);
658 value = get_bits(gb, level + 1);
664 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
666 /* Primary audio coding side information */
669 if (get_bits_left(&s->gb) < 0)
670 return AVERROR_INVALIDDATA;
673 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
674 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
677 for (j = base_channel; j < s->prim_channels; j++) {
678 for (k = 0; k < s->subband_activity[j]; k++)
679 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
682 /* Get prediction codebook */
683 for (j = base_channel; j < s->prim_channels; j++) {
684 for (k = 0; k < s->subband_activity[j]; k++) {
685 if (s->prediction_mode[j][k] > 0) {
686 /* (Prediction coefficient VQ address) */
687 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
692 /* Bit allocation index */
693 for (j = base_channel; j < s->prim_channels; j++) {
694 for (k = 0; k < s->vq_start_subband[j]; k++) {
695 if (s->bitalloc_huffman[j] == 6)
696 s->bitalloc[j][k] = get_bits(&s->gb, 5);
697 else if (s->bitalloc_huffman[j] == 5)
698 s->bitalloc[j][k] = get_bits(&s->gb, 4);
699 else if (s->bitalloc_huffman[j] == 7) {
700 av_log(s->avctx, AV_LOG_ERROR,
701 "Invalid bit allocation index\n");
702 return AVERROR_INVALIDDATA;
705 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
708 if (s->bitalloc[j][k] > 26) {
709 av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
710 j, k, s->bitalloc[j][k]);
711 return AVERROR_INVALIDDATA;
716 /* Transition mode */
717 for (j = base_channel; j < s->prim_channels; j++) {
718 for (k = 0; k < s->subband_activity[j]; k++) {
719 s->transition_mode[j][k] = 0;
720 if (s->subsubframes[s->current_subframe] > 1 &&
721 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
722 s->transition_mode[j][k] =
723 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
728 if (get_bits_left(&s->gb) < 0)
729 return AVERROR_INVALIDDATA;
731 for (j = base_channel; j < s->prim_channels; j++) {
732 const uint32_t *scale_table;
733 int scale_sum, log_size;
735 memset(s->scale_factor[j], 0,
736 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
738 if (s->scalefactor_huffman[j] == 6) {
739 scale_table = scale_factor_quant7;
742 scale_table = scale_factor_quant6;
746 /* When huffman coded, only the difference is encoded */
749 for (k = 0; k < s->subband_activity[j]; k++) {
750 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
751 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
752 s->scale_factor[j][k][0] = scale_table[scale_sum];
755 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
756 /* Get second scale factor */
757 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
758 s->scale_factor[j][k][1] = scale_table[scale_sum];
763 /* Joint subband scale factor codebook select */
764 for (j = base_channel; j < s->prim_channels; j++) {
765 /* Transmitted only if joint subband coding enabled */
766 if (s->joint_intensity[j] > 0)
767 s->joint_huff[j] = get_bits(&s->gb, 3);
770 if (get_bits_left(&s->gb) < 0)
771 return AVERROR_INVALIDDATA;
773 /* Scale factors for joint subband coding */
774 for (j = base_channel; j < s->prim_channels; j++) {
777 /* Transmitted only if joint subband coding enabled */
778 if (s->joint_intensity[j] > 0) {
780 source_channel = s->joint_intensity[j] - 1;
782 /* When huffman coded, only the difference is encoded
783 * (is this valid as well for joint scales ???) */
785 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
786 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
787 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
790 if (!(s->debug_flag & 0x02)) {
791 av_log(s->avctx, AV_LOG_DEBUG,
792 "Joint stereo coding not supported\n");
793 s->debug_flag |= 0x02;
798 /* Stereo downmix coefficients */
799 if (!base_channel && s->prim_channels > 2) {
801 for (j = base_channel; j < s->prim_channels; j++) {
802 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
803 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
806 int am = s->amode & DCA_CHANNEL_MASK;
807 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
808 av_log(s->avctx, AV_LOG_ERROR,
809 "Invalid channel mode %d\n", am);
810 return AVERROR_INVALIDDATA;
812 for (j = base_channel; j < s->prim_channels; j++) {
813 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
814 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
819 /* Dynamic range coefficient */
820 if (!base_channel && s->dynrange)
821 s->dynrange_coef = get_bits(&s->gb, 8);
823 /* Side information CRC check word */
824 if (s->crc_present) {
825 get_bits(&s->gb, 16);
829 * Primary audio data arrays
832 /* VQ encoded high frequency subbands */
833 for (j = base_channel; j < s->prim_channels; j++)
834 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
835 /* 1 vector -> 32 samples */
836 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
838 /* Low frequency effect data */
839 if (!base_channel && s->lfe) {
841 int lfe_samples = 2 * s->lfe * (4 + block_index);
842 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
845 for (j = lfe_samples; j < lfe_end_sample; j++) {
846 /* Signed 8 bits int */
847 s->lfe_data[j] = get_sbits(&s->gb, 8);
850 /* Scale factor index */
851 skip_bits(&s->gb, 1);
852 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 7)];
854 /* Quantization step size * scale factor */
855 lfe_scale = 0.035 * s->lfe_scale_factor;
857 for (j = lfe_samples; j < lfe_end_sample; j++)
858 s->lfe_data[j] *= lfe_scale;
862 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
863 s->subsubframes[s->current_subframe]);
864 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
865 s->partial_samples[s->current_subframe]);
867 for (j = base_channel; j < s->prim_channels; j++) {
868 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
869 for (k = 0; k < s->subband_activity[j]; k++)
870 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
871 av_log(s->avctx, AV_LOG_DEBUG, "\n");
873 for (j = base_channel; j < s->prim_channels; j++) {
874 for (k = 0; k < s->subband_activity[j]; k++)
875 av_log(s->avctx, AV_LOG_DEBUG,
876 "prediction coefs: %f, %f, %f, %f\n",
877 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
878 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
879 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
880 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
882 for (j = base_channel; j < s->prim_channels; j++) {
883 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
884 for (k = 0; k < s->vq_start_subband[j]; k++)
885 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
886 av_log(s->avctx, AV_LOG_DEBUG, "\n");
888 for (j = base_channel; j < s->prim_channels; j++) {
889 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
890 for (k = 0; k < s->subband_activity[j]; k++)
891 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
892 av_log(s->avctx, AV_LOG_DEBUG, "\n");
894 for (j = base_channel; j < s->prim_channels; j++) {
895 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
896 for (k = 0; k < s->subband_activity[j]; k++) {
897 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
898 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
899 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
900 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
902 av_log(s->avctx, AV_LOG_DEBUG, "\n");
904 for (j = base_channel; j < s->prim_channels; j++) {
905 if (s->joint_intensity[j] > 0) {
906 int source_channel = s->joint_intensity[j] - 1;
907 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
908 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
909 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
910 av_log(s->avctx, AV_LOG_DEBUG, "\n");
913 if (!base_channel && s->prim_channels > 2 && s->downmix) {
914 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
915 for (j = 0; j < s->prim_channels; j++) {
916 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
917 dca_downmix_coeffs[s->downmix_coef[j][0]]);
918 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
919 dca_downmix_coeffs[s->downmix_coef[j][1]]);
921 av_log(s->avctx, AV_LOG_DEBUG, "\n");
923 for (j = base_channel; j < s->prim_channels; j++)
924 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
925 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
926 if (!base_channel && s->lfe) {
927 int lfe_samples = 2 * s->lfe * (4 + block_index);
928 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
930 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
931 for (j = lfe_samples; j < lfe_end_sample; j++)
932 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
933 av_log(s->avctx, AV_LOG_DEBUG, "\n");
940 static void qmf_32_subbands(DCAContext *s, int chans,
941 float samples_in[32][8], float *samples_out,
944 const float *prCoeff;
947 int sb_act = s->subband_activity[chans];
950 scale *= sqrt(1 / 8.0);
953 if (!s->multirate_inter) /* Non-perfect reconstruction */
954 prCoeff = fir_32bands_nonperfect;
955 else /* Perfect reconstruction */
956 prCoeff = fir_32bands_perfect;
958 for (i = sb_act; i < 32; i++)
961 /* Reconstructed channel sample index */
962 for (subindex = 0; subindex < 8; subindex++) {
963 /* Load in one sample from each subband and clear inactive subbands */
964 for (i = 0; i < sb_act; i++) {
965 unsigned sign = (i - 1) & 2;
966 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
967 AV_WN32A(&s->raXin[i], v);
970 s->synth.synth_filter_float(&s->imdct,
971 s->subband_fir_hist[chans],
972 &s->hist_index[chans],
973 s->subband_fir_noidea[chans], prCoeff,
974 samples_out, s->raXin, scale);
979 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
980 int num_deci_sample, float *samples_in,
981 float *samples_out, float scale)
983 /* samples_in: An array holding decimated samples.
984 * Samples in current subframe starts from samples_in[0],
985 * while samples_in[-1], samples_in[-2], ..., stores samples
986 * from last subframe as history.
988 * samples_out: An array holding interpolated samples
992 const float *prCoeff;
995 /* Select decimation filter */
996 if (decimation_select == 1) {
998 prCoeff = lfe_fir_128;
1001 prCoeff = lfe_fir_64;
1004 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1005 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1007 samples_out += 2 * decifactor;
1011 /* downmixing routines */
1012 #define MIX_REAR1(samples, s1, rs, coef) \
1013 samples[0][i] += samples[s1][i] * coef[rs][0]; \
1014 samples[1][i] += samples[s1][i] * coef[rs][1];
1016 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1017 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1018 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1020 #define MIX_FRONT3(samples, coef) \
1021 t = samples[c][i]; \
1022 u = samples[l][i]; \
1023 v = samples[r][i]; \
1024 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1025 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1027 #define DOWNMIX_TO_STEREO(op1, op2) \
1028 for (i = 0; i < 256; i++) { \
1033 static void dca_downmix(float **samples, int srcfmt,
1034 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1035 const int8_t *channel_mapping)
1037 int c, l, r, sl, sr, s;
1040 float coef[DCA_PRIM_CHANNELS_MAX][2];
1042 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1043 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1044 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1050 case DCA_STEREO_TOTAL:
1051 case DCA_STEREO_SUMDIFF:
1053 av_log(NULL, 0, "Not implemented!\n");
1058 c = channel_mapping[0];
1059 l = channel_mapping[1];
1060 r = channel_mapping[2];
1061 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1064 s = channel_mapping[2];
1065 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1068 c = channel_mapping[0];
1069 l = channel_mapping[1];
1070 r = channel_mapping[2];
1071 s = channel_mapping[3];
1072 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1073 MIX_REAR1(samples, s, 3, coef));
1076 sl = channel_mapping[2];
1077 sr = channel_mapping[3];
1078 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1081 c = channel_mapping[0];
1082 l = channel_mapping[1];
1083 r = channel_mapping[2];
1084 sl = channel_mapping[3];
1085 sr = channel_mapping[4];
1086 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1087 MIX_REAR2(samples, sl, sr, 3, coef));
1093 #ifndef decode_blockcodes
1094 /* Very compact version of the block code decoder that does not use table
1095 * look-up but is slightly slower */
1096 static int decode_blockcode(int code, int levels, int *values)
1099 int offset = (levels - 1) >> 1;
1101 for (i = 0; i < 4; i++) {
1102 int div = FASTDIV(code, levels);
1103 values[i] = code - offset - div * levels;
1110 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1112 return decode_blockcode(code1, levels, values) |
1113 decode_blockcode(code2, levels, values + 4);
1117 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1118 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1120 #ifndef int8x8_fmul_int32
1121 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1123 float fscale = scale / 16.0;
1125 for (i = 0; i < 8; i++)
1126 dst[i] = src[i] * fscale;
1130 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1133 int subsubframe = s->current_subsubframe;
1135 const float *quant_step_table;
1138 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1139 LOCAL_ALIGNED_16(int, block, [8]);
1145 /* Select quantization step size table */
1146 if (s->bit_rate_index == 0x1f)
1147 quant_step_table = lossless_quant_d;
1149 quant_step_table = lossy_quant_d;
1151 for (k = base_channel; k < s->prim_channels; k++) {
1152 if (get_bits_left(&s->gb) < 0)
1153 return AVERROR_INVALIDDATA;
1155 for (l = 0; l < s->vq_start_subband[k]; l++) {
1158 /* Select the mid-tread linear quantizer */
1159 int abits = s->bitalloc[k][l];
1161 float quant_step_size = quant_step_table[abits];
1164 * Determine quantization index code book and its type
1167 /* Select quantization index code book */
1168 int sel = s->quant_index_huffman[k][abits];
1171 * Extract bits from the bit stream
1174 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1176 /* Deal with transients */
1177 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1178 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1179 s->scalefactor_adj[k][sel];
1181 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1184 int block_code1, block_code2, size, levels, err;
1186 size = abits_sizes[abits - 1];
1187 levels = abits_levels[abits - 1];
1189 block_code1 = get_bits(&s->gb, size);
1190 block_code2 = get_bits(&s->gb, size);
1191 err = decode_blockcodes(block_code1, block_code2,
1194 av_log(s->avctx, AV_LOG_ERROR,
1195 "ERROR: block code look-up failed\n");
1196 return AVERROR_INVALIDDATA;
1200 for (m = 0; m < 8; m++)
1201 block[m] = get_sbits(&s->gb, abits - 3);
1205 for (m = 0; m < 8; m++)
1206 block[m] = get_bitalloc(&s->gb,
1207 &dca_smpl_bitalloc[abits], sel);
1210 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1215 * Inverse ADPCM if in prediction mode
1217 if (s->prediction_mode[k][l]) {
1219 for (m = 0; m < 8; m++) {
1220 for (n = 1; n <= 4; n++)
1222 subband_samples[k][l][m] +=
1223 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1224 subband_samples[k][l][m - n] / 8192);
1225 else if (s->predictor_history)
1226 subband_samples[k][l][m] +=
1227 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1228 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1234 * Decode VQ encoded high frequencies
1236 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1237 /* 1 vector -> 32 samples but we only need the 8 samples
1238 * for this subsubframe. */
1239 int hfvq = s->high_freq_vq[k][l];
1241 if (!s->debug_flag & 0x01) {
1242 av_log(s->avctx, AV_LOG_DEBUG,
1243 "Stream with high frequencies VQ coding\n");
1244 s->debug_flag |= 0x01;
1247 int8x8_fmul_int32(subband_samples[k][l],
1248 &high_freq_vq[hfvq][subsubframe * 8],
1249 s->scale_factor[k][l][0]);
1253 /* Check for DSYNC after subsubframe */
1254 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1255 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1257 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1260 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1264 /* Backup predictor history for adpcm */
1265 for (k = base_channel; k < s->prim_channels; k++)
1266 for (l = 0; l < s->vq_start_subband[k]; l++)
1267 memcpy(s->subband_samples_hist[k][l],
1268 &subband_samples[k][l][4],
1269 4 * sizeof(subband_samples[0][0][0]));
1274 static int dca_filter_channels(DCAContext *s, int block_index)
1276 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1279 /* 32 subbands QMF */
1280 for (k = 0; k < s->prim_channels; k++) {
1281 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1282 0, 8388608.0, 8388608.0 };*/
1283 if (s->channel_order_tab[k] >= 0)
1284 qmf_32_subbands(s, k, subband_samples[k],
1285 s->samples_chanptr[s->channel_order_tab[k]],
1286 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1290 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1291 dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
1294 /* Generate LFE samples for this subsubframe FIXME!!! */
1295 if (s->output & DCA_LFE) {
1296 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1297 s->lfe_data + 2 * s->lfe * (block_index + 4),
1298 s->samples_chanptr[dca_lfe_index[s->amode]],
1299 1.0 / (256.0 * 32768.0));
1300 /* Outputs 20bits pcm samples */
1307 static int dca_subframe_footer(DCAContext *s, int base_channel)
1309 int aux_data_count = 0, i;
1312 * Unpack optional information
1315 /* presumably optional information only appears in the core? */
1316 if (!base_channel) {
1318 skip_bits_long(&s->gb, 32);
1321 aux_data_count = get_bits(&s->gb, 6);
1323 for (i = 0; i < aux_data_count; i++)
1324 get_bits(&s->gb, 8);
1326 if (s->crc_present && (s->downmix || s->dynrange))
1327 get_bits(&s->gb, 16);
1334 * Decode a dca frame block
1336 * @param s pointer to the DCAContext
1339 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1344 if (s->current_subframe >= s->subframes) {
1345 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1346 s->current_subframe, s->subframes);
1347 return AVERROR_INVALIDDATA;
1350 if (!s->current_subsubframe) {
1352 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1354 /* Read subframe header */
1355 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1359 /* Read subsubframe */
1361 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1363 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1367 s->current_subsubframe++;
1368 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1369 s->current_subsubframe = 0;
1370 s->current_subframe++;
1372 if (s->current_subframe >= s->subframes) {
1374 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1376 /* Read subframe footer */
1377 if ((ret = dca_subframe_footer(s, base_channel)))
1385 * Return the number of channels in an ExSS speaker mask (HD)
1387 static int dca_exss_mask2count(int mask)
1389 /* count bits that mean speaker pairs twice */
1390 return av_popcount(mask) +
1391 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1392 DCA_EXSS_FRONT_LEFT_RIGHT |
1393 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1394 DCA_EXSS_WIDE_LEFT_RIGHT |
1395 DCA_EXSS_SIDE_LEFT_RIGHT |
1396 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1397 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1398 DCA_EXSS_REAR_LEFT_RIGHT |
1399 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1403 * Skip mixing coefficients of a single mix out configuration (HD)
1405 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1409 for (i = 0; i < channels; i++) {
1410 int mix_map_mask = get_bits(gb, out_ch);
1411 int num_coeffs = av_popcount(mix_map_mask);
1412 skip_bits_long(gb, num_coeffs * 6);
1417 * Parse extension substream asset header (HD)
1419 static int dca_exss_parse_asset_header(DCAContext *s)
1421 int header_pos = get_bits_count(&s->gb);
1424 int embedded_stereo = 0;
1425 int embedded_6ch = 0;
1426 int drc_code_present;
1427 int extensions_mask;
1430 if (get_bits_left(&s->gb) < 16)
1433 /* We will parse just enough to get to the extensions bitmask with which
1434 * we can set the profile value. */
1436 header_size = get_bits(&s->gb, 9) + 1;
1437 skip_bits(&s->gb, 3); // asset index
1439 if (s->static_fields) {
1440 if (get_bits1(&s->gb))
1441 skip_bits(&s->gb, 4); // asset type descriptor
1442 if (get_bits1(&s->gb))
1443 skip_bits_long(&s->gb, 24); // language descriptor
1445 if (get_bits1(&s->gb)) {
1446 /* How can one fit 1024 bytes of text here if the maximum value
1447 * for the asset header size field above was 512 bytes? */
1448 int text_length = get_bits(&s->gb, 10) + 1;
1449 if (get_bits_left(&s->gb) < text_length * 8)
1451 skip_bits_long(&s->gb, text_length * 8); // info text
1454 skip_bits(&s->gb, 5); // bit resolution - 1
1455 skip_bits(&s->gb, 4); // max sample rate code
1456 channels = get_bits(&s->gb, 8) + 1;
1458 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1459 int spkr_remap_sets;
1460 int spkr_mask_size = 16;
1464 embedded_stereo = get_bits1(&s->gb);
1466 embedded_6ch = get_bits1(&s->gb);
1468 if (get_bits1(&s->gb)) {
1469 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1470 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1473 spkr_remap_sets = get_bits(&s->gb, 3);
1475 for (i = 0; i < spkr_remap_sets; i++) {
1476 /* std layout mask for each remap set */
1477 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1480 for (i = 0; i < spkr_remap_sets; i++) {
1481 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1482 if (get_bits_left(&s->gb) < 0)
1485 for (j = 0; j < num_spkrs[i]; j++) {
1486 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1487 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1488 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1493 skip_bits(&s->gb, 3); // representation type
1497 drc_code_present = get_bits1(&s->gb);
1498 if (drc_code_present)
1499 get_bits(&s->gb, 8); // drc code
1501 if (get_bits1(&s->gb))
1502 skip_bits(&s->gb, 5); // dialog normalization code
1504 if (drc_code_present && embedded_stereo)
1505 get_bits(&s->gb, 8); // drc stereo code
1507 if (s->mix_metadata && get_bits1(&s->gb)) {
1508 skip_bits(&s->gb, 1); // external mix
1509 skip_bits(&s->gb, 6); // post mix gain code
1511 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1512 skip_bits(&s->gb, 3); // drc limit
1514 skip_bits(&s->gb, 8); // custom drc code
1516 if (get_bits1(&s->gb)) // channel specific scaling
1517 for (i = 0; i < s->num_mix_configs; i++)
1518 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1520 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1522 for (i = 0; i < s->num_mix_configs; i++) {
1523 if (get_bits_left(&s->gb) < 0)
1525 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1527 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1528 if (embedded_stereo)
1529 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1533 switch (get_bits(&s->gb, 2)) {
1534 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1535 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1536 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1537 case 3: extensions_mask = 0; /* aux coding */ break;
1540 /* not parsed further, we were only interested in the extensions mask */
1542 if (get_bits_left(&s->gb) < 0)
1545 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1546 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1549 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1551 if (extensions_mask & DCA_EXT_EXSS_XLL)
1552 s->profile = FF_PROFILE_DTS_HD_MA;
1553 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1555 s->profile = FF_PROFILE_DTS_HD_HRA;
1557 if (!(extensions_mask & DCA_EXT_CORE))
1558 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1559 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1560 av_log(s->avctx, AV_LOG_WARNING,
1561 "DTS extensions detection mismatch (%d, %d)\n",
1562 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1568 * Parse extension substream header (HD)
1570 static void dca_exss_parse_header(DCAContext *s)
1576 int active_ss_mask[8];
1579 if (get_bits_left(&s->gb) < 52)
1582 skip_bits(&s->gb, 8); // user data
1583 ss_index = get_bits(&s->gb, 2);
1585 blownup = get_bits1(&s->gb);
1586 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1587 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1589 s->static_fields = get_bits1(&s->gb);
1590 if (s->static_fields) {
1591 skip_bits(&s->gb, 2); // reference clock code
1592 skip_bits(&s->gb, 3); // frame duration code
1594 if (get_bits1(&s->gb))
1595 skip_bits_long(&s->gb, 36); // timestamp
1597 /* a single stream can contain multiple audio assets that can be
1598 * combined to form multiple audio presentations */
1600 num_audiop = get_bits(&s->gb, 3) + 1;
1601 if (num_audiop > 1) {
1602 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1603 /* ignore such streams for now */
1607 num_assets = get_bits(&s->gb, 3) + 1;
1608 if (num_assets > 1) {
1609 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1610 /* ignore such streams for now */
1614 for (i = 0; i < num_audiop; i++)
1615 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1617 for (i = 0; i < num_audiop; i++)
1618 for (j = 0; j <= ss_index; j++)
1619 if (active_ss_mask[i] & (1 << j))
1620 skip_bits(&s->gb, 8); // active asset mask
1622 s->mix_metadata = get_bits1(&s->gb);
1623 if (s->mix_metadata) {
1624 int mix_out_mask_size;
1626 skip_bits(&s->gb, 2); // adjustment level
1627 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1628 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1630 for (i = 0; i < s->num_mix_configs; i++) {
1631 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1632 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1637 for (i = 0; i < num_assets; i++)
1638 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1640 for (i = 0; i < num_assets; i++) {
1641 if (dca_exss_parse_asset_header(s))
1645 /* not parsed further, we were only interested in the extensions mask
1646 * from the asset header */
1650 * Main frame decoding function
1651 * FIXME add arguments
1653 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1654 int *got_frame_ptr, AVPacket *avpkt)
1656 const uint8_t *buf = avpkt->data;
1657 int buf_size = avpkt->size;
1660 int num_core_channels = 0;
1662 float **samples_flt;
1663 DCAContext *s = avctx->priv_data;
1664 int channels, full_channels;
1670 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1671 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1672 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1673 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1674 return AVERROR_INVALIDDATA;
1677 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1678 if ((ret = dca_parse_frame_header(s)) < 0) {
1679 //seems like the frame is corrupt, try with the next one
1682 //set AVCodec values with parsed data
1683 avctx->sample_rate = s->sample_rate;
1684 avctx->bit_rate = s->bit_rate;
1686 s->profile = FF_PROFILE_DTS;
1688 for (i = 0; i < (s->sample_blocks / 8); i++) {
1689 if ((ret = dca_decode_block(s, 0, i))) {
1690 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1695 /* record number of core channels incase less than max channels are requested */
1696 num_core_channels = s->prim_channels;
1699 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1701 s->core_ext_mask = 0;
1703 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1705 /* only scan for extensions if ext_descr was unknown or indicated a
1706 * supported XCh extension */
1707 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1709 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1710 * extensions scan can fill it up */
1711 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1713 /* extensions start at 32-bit boundaries into bitstream */
1714 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1716 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1717 uint32_t bits = get_bits_long(&s->gb, 32);
1721 int ext_amode, xch_fsize;
1723 s->xch_base_channel = s->prim_channels;
1725 /* validate sync word using XCHFSIZE field */
1726 xch_fsize = show_bits(&s->gb, 10);
1727 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1728 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1731 /* skip length-to-end-of-frame field for the moment */
1732 skip_bits(&s->gb, 10);
1734 s->core_ext_mask |= DCA_EXT_XCH;
1736 /* extension amode(number of channels in extension) should be 1 */
1737 /* AFAIK XCh is not used for more channels */
1738 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1739 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1740 " supported!\n", ext_amode);
1744 /* much like core primary audio coding header */
1745 dca_parse_audio_coding_header(s, s->xch_base_channel);
1747 for (i = 0; i < (s->sample_blocks / 8); i++)
1748 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1749 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1757 /* XXCh: extended channels */
1758 /* usually found either in core or HD part in DTS-HD HRA streams,
1759 * but not in DTS-ES which contains XCh extensions instead */
1760 s->core_ext_mask |= DCA_EXT_XXCH;
1764 int fsize96 = show_bits(&s->gb, 12) + 1;
1765 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1768 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1769 get_bits_count(&s->gb));
1770 skip_bits(&s->gb, 12);
1771 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1772 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1774 s->core_ext_mask |= DCA_EXT_X96;
1779 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1782 /* no supported extensions, skip the rest of the core substream */
1783 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1786 if (s->core_ext_mask & DCA_EXT_X96)
1787 s->profile = FF_PROFILE_DTS_96_24;
1788 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1789 s->profile = FF_PROFILE_DTS_ES;
1791 /* check for ExSS (HD part) */
1792 if (s->dca_buffer_size - s->frame_size > 32 &&
1793 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1794 dca_exss_parse_header(s);
1796 avctx->profile = s->profile;
1798 full_channels = channels = s->prim_channels + !!s->lfe;
1800 if (s->amode < 16) {
1801 avctx->channel_layout = dca_core_channel_layout[s->amode];
1803 if (s->xch_present && (!avctx->request_channels ||
1804 avctx->request_channels > num_core_channels + !!s->lfe)) {
1805 avctx->channel_layout |= AV_CH_BACK_CENTER;
1807 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1808 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1810 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1813 channels = num_core_channels + !!s->lfe;
1814 s->xch_present = 0; /* disable further xch processing */
1816 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1817 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1819 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1822 if (channels > !!s->lfe &&
1823 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1824 return AVERROR_INVALIDDATA;
1826 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1828 s->output = DCA_STEREO;
1829 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1832 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1833 return AVERROR_INVALIDDATA;
1835 avctx->channels = channels;
1837 /* get output buffer */
1838 s->frame.nb_samples = 256 * (s->sample_blocks / 8);
1839 if ((ret = ff_get_buffer(avctx, &s->frame)) < 0) {
1840 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1843 samples_flt = (float **) s->frame.extended_data;
1845 /* allocate buffer for extra channels if downmixing */
1846 if (avctx->channels < full_channels) {
1847 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
1848 s->frame.nb_samples,
1849 avctx->sample_fmt, 0);
1853 av_fast_malloc(&s->extra_channels_buffer,
1854 &s->extra_channels_buffer_size, ret);
1855 if (!s->extra_channels_buffer)
1856 return AVERROR(ENOMEM);
1858 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
1859 s->extra_channels_buffer,
1860 full_channels - channels,
1861 s->frame.nb_samples, avctx->sample_fmt, 0);
1866 /* filter to get final output */
1867 for (i = 0; i < (s->sample_blocks / 8); i++) {
1870 for (ch = 0; ch < channels; ch++)
1871 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
1872 for (; ch < full_channels; ch++)
1873 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
1875 dca_filter_channels(s, i);
1877 /* If this was marked as a DTS-ES stream we need to subtract back- */
1878 /* channel from SL & SR to remove matrixed back-channel signal */
1879 if ((s->source_pcm_res & 1) && s->xch_present) {
1880 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
1881 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
1882 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
1883 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1884 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1888 /* update lfe history */
1889 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1890 for (i = 0; i < 2 * s->lfe * 4; i++)
1891 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1894 *(AVFrame *) data = s->frame;
1902 * DCA initialization
1904 * @param avctx pointer to the AVCodecContext
1907 static av_cold int dca_decode_init(AVCodecContext *avctx)
1909 DCAContext *s = avctx->priv_data;
1914 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
1915 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1916 ff_synth_filter_init(&s->synth);
1917 ff_dcadsp_init(&s->dcadsp);
1918 ff_fmt_convert_init(&s->fmt_conv, avctx);
1920 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
1922 /* allow downmixing to stereo */
1923 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1924 avctx->request_channels == 2) {
1925 avctx->channels = avctx->request_channels;
1928 avcodec_get_frame_defaults(&s->frame);
1929 avctx->coded_frame = &s->frame;
1934 static av_cold int dca_decode_end(AVCodecContext *avctx)
1936 DCAContext *s = avctx->priv_data;
1937 ff_mdct_end(&s->imdct);
1938 av_freep(&s->extra_channels_buffer);
1942 static const AVProfile profiles[] = {
1943 { FF_PROFILE_DTS, "DTS" },
1944 { FF_PROFILE_DTS_ES, "DTS-ES" },
1945 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1946 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1947 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1948 { FF_PROFILE_UNKNOWN },
1951 AVCodec ff_dca_decoder = {
1953 .type = AVMEDIA_TYPE_AUDIO,
1954 .id = AV_CODEC_ID_DTS,
1955 .priv_data_size = sizeof(DCAContext),
1956 .init = dca_decode_init,
1957 .decode = dca_decode_frame,
1958 .close = dca_decode_end,
1959 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1960 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1961 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
1962 AV_SAMPLE_FMT_NONE },
1963 .profiles = NULL_IF_CONFIG_SMALL(profiles),