0f7337574fb12129546374e02732ffcd2681fb10
[libav.git] / libavcodec / dcadec.c
1 /*
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
7 *
8 * This file is part of Libav.
9 *
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.
14 *
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.
19 *
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
23 */
24
25 #include <math.h>
26 #include <stddef.h>
27 #include <stdio.h>
28
29 #include "libavutil/channel_layout.h"
30 #include "libavutil/common.h"
31 #include "libavutil/float_dsp.h"
32 #include "libavutil/internal.h"
33 #include "libavutil/intreadwrite.h"
34 #include "libavutil/mathematics.h"
35 #include "libavutil/samplefmt.h"
36 #include "avcodec.h"
37 #include "fft.h"
38 #include "get_bits.h"
39 #include "put_bits.h"
40 #include "dcadata.h"
41 #include "dcahuff.h"
42 #include "dca.h"
43 #include "mathops.h"
44 #include "synth_filter.h"
45 #include "dcadsp.h"
46 #include "fmtconvert.h"
47 #include "internal.h"
48
49 #if ARCH_ARM
50 # include "arm/dca.h"
51 #endif
52
53 //#define TRACE
54
55 #define DCA_PRIM_CHANNELS_MAX (7)
56 #define DCA_SUBBANDS (32)
57 #define DCA_ABITS_MAX (32) /* Should be 28 */
58 #define DCA_SUBSUBFRAMES_MAX (4)
59 #define DCA_SUBFRAMES_MAX (16)
60 #define DCA_BLOCKS_MAX (16)
61 #define DCA_LFE_MAX (3)
62
63 enum DCAMode {
64 DCA_MONO = 0,
65 DCA_CHANNEL,
66 DCA_STEREO,
67 DCA_STEREO_SUMDIFF,
68 DCA_STEREO_TOTAL,
69 DCA_3F,
70 DCA_2F1R,
71 DCA_3F1R,
72 DCA_2F2R,
73 DCA_3F2R,
74 DCA_4F2R
75 };
76
77 /* these are unconfirmed but should be mostly correct */
78 enum DCAExSSSpeakerMask {
79 DCA_EXSS_FRONT_CENTER = 0x0001,
80 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
81 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
82 DCA_EXSS_LFE = 0x0008,
83 DCA_EXSS_REAR_CENTER = 0x0010,
84 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
85 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
86 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
87 DCA_EXSS_OVERHEAD = 0x0100,
88 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
89 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
90 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
91 DCA_EXSS_LFE2 = 0x1000,
92 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
93 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
94 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
95 };
96
97 enum DCAExtensionMask {
98 DCA_EXT_CORE = 0x001, ///< core in core substream
99 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
100 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
101 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
102 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
103 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
104 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
105 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
106 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
107 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
108 };
109
110 /* -1 are reserved or unknown */
111 static const int dca_ext_audio_descr_mask[] = {
112 DCA_EXT_XCH,
113 -1,
114 DCA_EXT_X96,
115 DCA_EXT_XCH | DCA_EXT_X96,
116 -1,
117 -1,
118 DCA_EXT_XXCH,
119 -1,
120 };
121
122 /* extensions that reside in core substream */
123 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
124
125 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
126 * Some compromises have been made for special configurations. Most configurations
127 * are never used so complete accuracy is not needed.
128 *
129 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
130 * S -> side, when both rear and back are configured move one of them to the side channel
131 * OV -> center back
132 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
133 */
134 static const uint64_t dca_core_channel_layout[] = {
135 AV_CH_FRONT_CENTER, ///< 1, A
136 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
137 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
138 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
139 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
140 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
141 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
142 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
143 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
144
145 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
146 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
147
148 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
149 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
150
151 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
152 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
153
154 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
155 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
156 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
157
158 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
159 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
160 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
161
162 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
163 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
164 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
165
166 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
167 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
168 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
169 };
170
171 static const int8_t dca_lfe_index[] = {
172 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
173 };
174
175 static const int8_t dca_channel_reorder_lfe[][9] = {
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 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
180 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
181 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
182 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
183 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
184 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
185 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
186 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
187 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
188 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
189 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
190 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
191 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
192 };
193
194 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
195 { 0, 2, -1, -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 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
199 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
200 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
201 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
202 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
203 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
204 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
205 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
206 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
207 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
208 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
209 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
210 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
211 };
212
213 static const int8_t dca_channel_reorder_nolfe[][9] = {
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 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
218 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
219 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
220 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
221 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
222 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
223 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
224 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
225 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
226 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
227 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
228 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
229 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
230 };
231
232 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
233 { 0, 1, -1, -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 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
237 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
238 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
239 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
240 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
241 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
242 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
243 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
244 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
245 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
246 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
247 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
248 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
249 };
250
251 #define DCA_DOLBY 101 /* FIXME */
252
253 #define DCA_CHANNEL_BITS 6
254 #define DCA_CHANNEL_MASK 0x3F
255
256 #define DCA_LFE 0x80
257
258 #define HEADER_SIZE 14
259
260 #define DCA_MAX_FRAME_SIZE 16384
261 #define DCA_MAX_EXSS_HEADER_SIZE 4096
262
263 #define DCA_BUFFER_PADDING_SIZE 1024
264
265 /** Bit allocation */
266 typedef struct {
267 int offset; ///< code values offset
268 int maxbits[8]; ///< max bits in VLC
269 int wrap; ///< wrap for get_vlc2()
270 VLC vlc[8]; ///< actual codes
271 } BitAlloc;
272
273 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
274 static BitAlloc dca_tmode; ///< transition mode VLCs
275 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
276 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
277
278 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
279 int idx)
280 {
281 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
282 ba->offset;
283 }
284
285 typedef struct {
286 AVCodecContext *avctx;
287 /* Frame header */
288 int frame_type; ///< type of the current frame
289 int samples_deficit; ///< deficit sample count
290 int crc_present; ///< crc is present in the bitstream
291 int sample_blocks; ///< number of PCM sample blocks
292 int frame_size; ///< primary frame byte size
293 int amode; ///< audio channels arrangement
294 int sample_rate; ///< audio sampling rate
295 int bit_rate; ///< transmission bit rate
296 int bit_rate_index; ///< transmission bit rate index
297
298 int downmix; ///< embedded downmix enabled
299 int dynrange; ///< embedded dynamic range flag
300 int timestamp; ///< embedded time stamp flag
301 int aux_data; ///< auxiliary data flag
302 int hdcd; ///< source material is mastered in HDCD
303 int ext_descr; ///< extension audio descriptor flag
304 int ext_coding; ///< extended coding flag
305 int aspf; ///< audio sync word insertion flag
306 int lfe; ///< low frequency effects flag
307 int predictor_history; ///< predictor history flag
308 int header_crc; ///< header crc check bytes
309 int multirate_inter; ///< multirate interpolator switch
310 int version; ///< encoder software revision
311 int copy_history; ///< copy history
312 int source_pcm_res; ///< source pcm resolution
313 int front_sum; ///< front sum/difference flag
314 int surround_sum; ///< surround sum/difference flag
315 int dialog_norm; ///< dialog normalisation parameter
316
317 /* Primary audio coding header */
318 int subframes; ///< number of subframes
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
329
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
342
343 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
344
345 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
346 int lfe_scale_factor;
347
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];
354
355 int output; ///< type of output
356
357 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
358 float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
359 float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
360 uint8_t *extra_channels_buffer;
361 unsigned int extra_channels_buffer_size;
362
363 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
364 int dca_buffer_size; ///< how much data is in the dca_buffer
365
366 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
367 GetBitContext gb;
368 /* Current position in DCA frame */
369 int current_subframe;
370 int current_subsubframe;
371
372 int core_ext_mask; ///< present extensions in the core substream
373
374 /* XCh extension information */
375 int xch_present; ///< XCh extension present and valid
376 int xch_base_channel; ///< index of first (only) channel containing XCH data
377
378 /* ExSS header parser */
379 int static_fields; ///< static fields present
380 int mix_metadata; ///< mixing metadata present
381 int num_mix_configs; ///< number of mix out configurations
382 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
383
384 int profile;
385
386 int debug_flag; ///< used for suppressing repeated error messages output
387 AVFloatDSPContext fdsp;
388 FFTContext imdct;
389 SynthFilterContext synth;
390 DCADSPContext dcadsp;
391 FmtConvertContext fmt_conv;
392 } DCAContext;
393
394 static const uint16_t dca_vlc_offs[] = {
395 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
396 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
397 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
398 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
399 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
400 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
401 };
402
403 static av_cold void dca_init_vlcs(void)
404 {
405 static int vlcs_initialized = 0;
406 int i, j, c = 14;
407 static VLC_TYPE dca_table[23622][2];
408
409 if (vlcs_initialized)
410 return;
411
412 dca_bitalloc_index.offset = 1;
413 dca_bitalloc_index.wrap = 2;
414 for (i = 0; i < 5; i++) {
415 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
416 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
417 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
418 bitalloc_12_bits[i], 1, 1,
419 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
420 }
421 dca_scalefactor.offset = -64;
422 dca_scalefactor.wrap = 2;
423 for (i = 0; i < 5; i++) {
424 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
425 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
426 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
427 scales_bits[i], 1, 1,
428 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
429 }
430 dca_tmode.offset = 0;
431 dca_tmode.wrap = 1;
432 for (i = 0; i < 4; i++) {
433 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
434 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
435 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
436 tmode_bits[i], 1, 1,
437 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
438 }
439
440 for (i = 0; i < 10; i++)
441 for (j = 0; j < 7; j++) {
442 if (!bitalloc_codes[i][j])
443 break;
444 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
445 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
446 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
447 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
448
449 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
450 bitalloc_sizes[i],
451 bitalloc_bits[i][j], 1, 1,
452 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
453 c++;
454 }
455 vlcs_initialized = 1;
456 }
457
458 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
459 {
460 while (len--)
461 *dst++ = get_bits(gb, bits);
462 }
463
464 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
465 {
466 int i, j;
467 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
468 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
469 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
470
471 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
472 s->prim_channels = s->total_channels;
473
474 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
475 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
476
477
478 for (i = base_channel; i < s->prim_channels; i++) {
479 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
480 if (s->subband_activity[i] > DCA_SUBBANDS)
481 s->subband_activity[i] = DCA_SUBBANDS;
482 }
483 for (i = base_channel; i < s->prim_channels; i++) {
484 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
485 if (s->vq_start_subband[i] > DCA_SUBBANDS)
486 s->vq_start_subband[i] = DCA_SUBBANDS;
487 }
488 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
489 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
490 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
491 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
492
493 /* Get codebooks quantization indexes */
494 if (!base_channel)
495 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
496 for (j = 1; j < 11; j++)
497 for (i = base_channel; i < s->prim_channels; i++)
498 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
499
500 /* Get scale factor adjustment */
501 for (j = 0; j < 11; j++)
502 for (i = base_channel; i < s->prim_channels; i++)
503 s->scalefactor_adj[i][j] = 1;
504
505 for (j = 1; j < 11; j++)
506 for (i = base_channel; i < s->prim_channels; i++)
507 if (s->quant_index_huffman[i][j] < thr[j])
508 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
509
510 if (s->crc_present) {
511 /* Audio header CRC check */
512 get_bits(&s->gb, 16);
513 }
514
515 s->current_subframe = 0;
516 s->current_subsubframe = 0;
517
518 #ifdef TRACE
519 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
520 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
521 for (i = base_channel; i < s->prim_channels; i++) {
522 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
523 s->subband_activity[i]);
524 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
525 s->vq_start_subband[i]);
526 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
527 s->joint_intensity[i]);
528 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
529 s->transient_huffman[i]);
530 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
531 s->scalefactor_huffman[i]);
532 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
533 s->bitalloc_huffman[i]);
534 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
535 for (j = 0; j < 11; j++)
536 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
537 av_log(s->avctx, AV_LOG_DEBUG, "\n");
538 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
539 for (j = 0; j < 11; j++)
540 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
541 av_log(s->avctx, AV_LOG_DEBUG, "\n");
542 }
543 #endif
544
545 return 0;
546 }
547
548 static int dca_parse_frame_header(DCAContext *s)
549 {
550 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
551
552 /* Sync code */
553 skip_bits_long(&s->gb, 32);
554
555 /* Frame header */
556 s->frame_type = get_bits(&s->gb, 1);
557 s->samples_deficit = get_bits(&s->gb, 5) + 1;
558 s->crc_present = get_bits(&s->gb, 1);
559 s->sample_blocks = get_bits(&s->gb, 7) + 1;
560 s->frame_size = get_bits(&s->gb, 14) + 1;
561 if (s->frame_size < 95)
562 return AVERROR_INVALIDDATA;
563 s->amode = get_bits(&s->gb, 6);
564 s->sample_rate = avpriv_dca_sample_rates[get_bits(&s->gb, 4)];
565 if (!s->sample_rate)
566 return AVERROR_INVALIDDATA;
567 s->bit_rate_index = get_bits(&s->gb, 5);
568 s->bit_rate = dca_bit_rates[s->bit_rate_index];
569 if (!s->bit_rate)
570 return AVERROR_INVALIDDATA;
571
572 s->downmix = get_bits(&s->gb, 1);
573 s->dynrange = get_bits(&s->gb, 1);
574 s->timestamp = get_bits(&s->gb, 1);
575 s->aux_data = get_bits(&s->gb, 1);
576 s->hdcd = get_bits(&s->gb, 1);
577 s->ext_descr = get_bits(&s->gb, 3);
578 s->ext_coding = get_bits(&s->gb, 1);
579 s->aspf = get_bits(&s->gb, 1);
580 s->lfe = get_bits(&s->gb, 2);
581 s->predictor_history = get_bits(&s->gb, 1);
582
583 /* TODO: check CRC */
584 if (s->crc_present)
585 s->header_crc = get_bits(&s->gb, 16);
586
587 s->multirate_inter = get_bits(&s->gb, 1);
588 s->version = get_bits(&s->gb, 4);
589 s->copy_history = get_bits(&s->gb, 2);
590 s->source_pcm_res = get_bits(&s->gb, 3);
591 s->front_sum = get_bits(&s->gb, 1);
592 s->surround_sum = get_bits(&s->gb, 1);
593 s->dialog_norm = get_bits(&s->gb, 4);
594
595 /* FIXME: channels mixing levels */
596 s->output = s->amode;
597 if (s->lfe)
598 s->output |= DCA_LFE;
599
600 #ifdef TRACE
601 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
602 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
603 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
604 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
605 s->sample_blocks, s->sample_blocks * 32);
606 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
607 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
608 s->amode, dca_channels[s->amode]);
609 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
610 s->sample_rate);
611 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
612 s->bit_rate);
613 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
614 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
615 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
616 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
617 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
618 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
619 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
620 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
621 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
622 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
623 s->predictor_history);
624 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
625 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
626 s->multirate_inter);
627 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
628 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
629 av_log(s->avctx, AV_LOG_DEBUG,
630 "source pcm resolution: %i (%i bits/sample)\n",
631 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
632 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
633 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
634 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
635 av_log(s->avctx, AV_LOG_DEBUG, "\n");
636 #endif
637
638 /* Primary audio coding header */
639 s->subframes = get_bits(&s->gb, 4) + 1;
640
641 return dca_parse_audio_coding_header(s, 0);
642 }
643
644
645 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
646 {
647 if (level < 5) {
648 /* huffman encoded */
649 value += get_bitalloc(gb, &dca_scalefactor, level);
650 value = av_clip(value, 0, (1 << log2range) - 1);
651 } else if (level < 8) {
652 if (level + 1 > log2range) {
653 skip_bits(gb, level + 1 - log2range);
654 value = get_bits(gb, log2range);
655 } else {
656 value = get_bits(gb, level + 1);
657 }
658 }
659 return value;
660 }
661
662 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
663 {
664 /* Primary audio coding side information */
665 int j, k;
666
667 if (get_bits_left(&s->gb) < 0)
668 return AVERROR_INVALIDDATA;
669
670 if (!base_channel) {
671 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
672 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
673 }
674
675 for (j = base_channel; j < s->prim_channels; j++) {
676 for (k = 0; k < s->subband_activity[j]; k++)
677 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
678 }
679
680 /* Get prediction codebook */
681 for (j = base_channel; j < s->prim_channels; j++) {
682 for (k = 0; k < s->subband_activity[j]; k++) {
683 if (s->prediction_mode[j][k] > 0) {
684 /* (Prediction coefficient VQ address) */
685 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
686 }
687 }
688 }
689
690 /* Bit allocation index */
691 for (j = base_channel; j < s->prim_channels; j++) {
692 for (k = 0; k < s->vq_start_subband[j]; k++) {
693 if (s->bitalloc_huffman[j] == 6)
694 s->bitalloc[j][k] = get_bits(&s->gb, 5);
695 else if (s->bitalloc_huffman[j] == 5)
696 s->bitalloc[j][k] = get_bits(&s->gb, 4);
697 else if (s->bitalloc_huffman[j] == 7) {
698 av_log(s->avctx, AV_LOG_ERROR,
699 "Invalid bit allocation index\n");
700 return AVERROR_INVALIDDATA;
701 } else {
702 s->bitalloc[j][k] =
703 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
704 }
705
706 if (s->bitalloc[j][k] > 26) {
707 av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
708 j, k, s->bitalloc[j][k]);
709 return AVERROR_INVALIDDATA;
710 }
711 }
712 }
713
714 /* Transition mode */
715 for (j = base_channel; j < s->prim_channels; j++) {
716 for (k = 0; k < s->subband_activity[j]; k++) {
717 s->transition_mode[j][k] = 0;
718 if (s->subsubframes[s->current_subframe] > 1 &&
719 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
720 s->transition_mode[j][k] =
721 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
722 }
723 }
724 }
725
726 if (get_bits_left(&s->gb) < 0)
727 return AVERROR_INVALIDDATA;
728
729 for (j = base_channel; j < s->prim_channels; j++) {
730 const uint32_t *scale_table;
731 int scale_sum, log_size;
732
733 memset(s->scale_factor[j], 0,
734 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
735
736 if (s->scalefactor_huffman[j] == 6) {
737 scale_table = scale_factor_quant7;
738 log_size = 7;
739 } else {
740 scale_table = scale_factor_quant6;
741 log_size = 6;
742 }
743
744 /* When huffman coded, only the difference is encoded */
745 scale_sum = 0;
746
747 for (k = 0; k < s->subband_activity[j]; k++) {
748 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
749 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
750 s->scale_factor[j][k][0] = scale_table[scale_sum];
751 }
752
753 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
754 /* Get second scale factor */
755 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
756 s->scale_factor[j][k][1] = scale_table[scale_sum];
757 }
758 }
759 }
760
761 /* Joint subband scale factor codebook select */
762 for (j = base_channel; j < s->prim_channels; j++) {
763 /* Transmitted only if joint subband coding enabled */
764 if (s->joint_intensity[j] > 0)
765 s->joint_huff[j] = get_bits(&s->gb, 3);
766 }
767
768 if (get_bits_left(&s->gb) < 0)
769 return AVERROR_INVALIDDATA;
770
771 /* Scale factors for joint subband coding */
772 for (j = base_channel; j < s->prim_channels; j++) {
773 int source_channel;
774
775 /* Transmitted only if joint subband coding enabled */
776 if (s->joint_intensity[j] > 0) {
777 int scale = 0;
778 source_channel = s->joint_intensity[j] - 1;
779
780 /* When huffman coded, only the difference is encoded
781 * (is this valid as well for joint scales ???) */
782
783 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
784 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
785 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
786 }
787
788 if (!(s->debug_flag & 0x02)) {
789 av_log(s->avctx, AV_LOG_DEBUG,
790 "Joint stereo coding not supported\n");
791 s->debug_flag |= 0x02;
792 }
793 }
794 }
795
796 /* Stereo downmix coefficients */
797 if (!base_channel && s->prim_channels > 2) {
798 if (s->downmix) {
799 for (j = base_channel; j < s->prim_channels; j++) {
800 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
801 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
802 }
803 } else {
804 int am = s->amode & DCA_CHANNEL_MASK;
805 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
806 av_log(s->avctx, AV_LOG_ERROR,
807 "Invalid channel mode %d\n", am);
808 return AVERROR_INVALIDDATA;
809 }
810 for (j = base_channel; j < s->prim_channels; j++) {
811 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
812 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
813 }
814 }
815 }
816
817 /* Dynamic range coefficient */
818 if (!base_channel && s->dynrange)
819 s->dynrange_coef = get_bits(&s->gb, 8);
820
821 /* Side information CRC check word */
822 if (s->crc_present) {
823 get_bits(&s->gb, 16);
824 }
825
826 /*
827 * Primary audio data arrays
828 */
829
830 /* VQ encoded high frequency subbands */
831 for (j = base_channel; j < s->prim_channels; j++)
832 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
833 /* 1 vector -> 32 samples */
834 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
835
836 /* Low frequency effect data */
837 if (!base_channel && s->lfe) {
838 /* LFE samples */
839 int lfe_samples = 2 * s->lfe * (4 + block_index);
840 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
841 float lfe_scale;
842
843 for (j = lfe_samples; j < lfe_end_sample; j++) {
844 /* Signed 8 bits int */
845 s->lfe_data[j] = get_sbits(&s->gb, 8);
846 }
847
848 /* Scale factor index */
849 skip_bits(&s->gb, 1);
850 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 7)];
851
852 /* Quantization step size * scale factor */
853 lfe_scale = 0.035 * s->lfe_scale_factor;
854
855 for (j = lfe_samples; j < lfe_end_sample; j++)
856 s->lfe_data[j] *= lfe_scale;
857 }
858
859 #ifdef TRACE
860 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
861 s->subsubframes[s->current_subframe]);
862 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
863 s->partial_samples[s->current_subframe]);
864
865 for (j = base_channel; j < s->prim_channels; j++) {
866 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
867 for (k = 0; k < s->subband_activity[j]; k++)
868 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
869 av_log(s->avctx, AV_LOG_DEBUG, "\n");
870 }
871 for (j = base_channel; j < s->prim_channels; j++) {
872 for (k = 0; k < s->subband_activity[j]; k++)
873 av_log(s->avctx, AV_LOG_DEBUG,
874 "prediction coefs: %f, %f, %f, %f\n",
875 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
876 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
877 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
878 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
879 }
880 for (j = base_channel; j < s->prim_channels; j++) {
881 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
882 for (k = 0; k < s->vq_start_subband[j]; k++)
883 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
884 av_log(s->avctx, AV_LOG_DEBUG, "\n");
885 }
886 for (j = base_channel; j < s->prim_channels; j++) {
887 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
888 for (k = 0; k < s->subband_activity[j]; k++)
889 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
890 av_log(s->avctx, AV_LOG_DEBUG, "\n");
891 }
892 for (j = base_channel; j < s->prim_channels; j++) {
893 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
894 for (k = 0; k < s->subband_activity[j]; k++) {
895 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
896 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
897 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
898 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
899 }
900 av_log(s->avctx, AV_LOG_DEBUG, "\n");
901 }
902 for (j = base_channel; j < s->prim_channels; j++) {
903 if (s->joint_intensity[j] > 0) {
904 int source_channel = s->joint_intensity[j] - 1;
905 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
906 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
907 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
908 av_log(s->avctx, AV_LOG_DEBUG, "\n");
909 }
910 }
911 if (!base_channel && s->prim_channels > 2 && s->downmix) {
912 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
913 for (j = 0; j < s->prim_channels; j++) {
914 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
915 dca_downmix_coeffs[s->downmix_coef[j][0]]);
916 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
917 dca_downmix_coeffs[s->downmix_coef[j][1]]);
918 }
919 av_log(s->avctx, AV_LOG_DEBUG, "\n");
920 }
921 for (j = base_channel; j < s->prim_channels; j++)
922 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
923 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
924 if (!base_channel && s->lfe) {
925 int lfe_samples = 2 * s->lfe * (4 + block_index);
926 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
927
928 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
929 for (j = lfe_samples; j < lfe_end_sample; j++)
930 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
931 av_log(s->avctx, AV_LOG_DEBUG, "\n");
932 }
933 #endif
934
935 return 0;
936 }
937
938 static void qmf_32_subbands(DCAContext *s, int chans,
939 float samples_in[32][8], float *samples_out,
940 float scale)
941 {
942 const float *prCoeff;
943 int i;
944
945 int sb_act = s->subband_activity[chans];
946 int subindex;
947
948 scale *= sqrt(1 / 8.0);
949
950 /* Select filter */
951 if (!s->multirate_inter) /* Non-perfect reconstruction */
952 prCoeff = fir_32bands_nonperfect;
953 else /* Perfect reconstruction */
954 prCoeff = fir_32bands_perfect;
955
956 for (i = sb_act; i < 32; i++)
957 s->raXin[i] = 0.0;
958
959 /* Reconstructed channel sample index */
960 for (subindex = 0; subindex < 8; subindex++) {
961 /* Load in one sample from each subband and clear inactive subbands */
962 for (i = 0; i < sb_act; i++) {
963 unsigned sign = (i - 1) & 2;
964 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
965 AV_WN32A(&s->raXin[i], v);
966 }
967
968 s->synth.synth_filter_float(&s->imdct,
969 s->subband_fir_hist[chans],
970 &s->hist_index[chans],
971 s->subband_fir_noidea[chans], prCoeff,
972 samples_out, s->raXin, scale);
973 samples_out += 32;
974 }
975 }
976
977 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
978 int num_deci_sample, float *samples_in,
979 float *samples_out, float scale)
980 {
981 /* samples_in: An array holding decimated samples.
982 * Samples in current subframe starts from samples_in[0],
983 * while samples_in[-1], samples_in[-2], ..., stores samples
984 * from last subframe as history.
985 *
986 * samples_out: An array holding interpolated samples
987 */
988
989 int decifactor;
990 const float *prCoeff;
991 int deciindex;
992
993 /* Select decimation filter */
994 if (decimation_select == 1) {
995 decifactor = 64;
996 prCoeff = lfe_fir_128;
997 } else {
998 decifactor = 32;
999 prCoeff = lfe_fir_64;
1000 }
1001 /* Interpolation */
1002 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1003 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1004 samples_in++;
1005 samples_out += 2 * decifactor;
1006 }
1007 }
1008
1009 /* downmixing routines */
1010 #define MIX_REAR1(samples, s1, rs, coef) \
1011 samples[0][i] += samples[s1][i] * coef[rs][0]; \
1012 samples[1][i] += samples[s1][i] * coef[rs][1];
1013
1014 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1015 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1016 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1017
1018 #define MIX_FRONT3(samples, coef) \
1019 t = samples[c][i]; \
1020 u = samples[l][i]; \
1021 v = samples[r][i]; \
1022 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1023 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1024
1025 #define DOWNMIX_TO_STEREO(op1, op2) \
1026 for (i = 0; i < 256; i++) { \
1027 op1 \
1028 op2 \
1029 }
1030
1031 static void dca_downmix(float **samples, int srcfmt,
1032 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1033 const int8_t *channel_mapping)
1034 {
1035 int c, l, r, sl, sr, s;
1036 int i;
1037 float t, u, v;
1038 float coef[DCA_PRIM_CHANNELS_MAX][2];
1039
1040 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1041 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1042 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1043 }
1044
1045 switch (srcfmt) {
1046 case DCA_MONO:
1047 case DCA_CHANNEL:
1048 case DCA_STEREO_TOTAL:
1049 case DCA_STEREO_SUMDIFF:
1050 case DCA_4F2R:
1051 av_log(NULL, 0, "Not implemented!\n");
1052 break;
1053 case DCA_STEREO:
1054 break;
1055 case DCA_3F:
1056 c = channel_mapping[0];
1057 l = channel_mapping[1];
1058 r = channel_mapping[2];
1059 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1060 break;
1061 case DCA_2F1R:
1062 s = channel_mapping[2];
1063 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1064 break;
1065 case DCA_3F1R:
1066 c = channel_mapping[0];
1067 l = channel_mapping[1];
1068 r = channel_mapping[2];
1069 s = channel_mapping[3];
1070 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1071 MIX_REAR1(samples, s, 3, coef));
1072 break;
1073 case DCA_2F2R:
1074 sl = channel_mapping[2];
1075 sr = channel_mapping[3];
1076 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1077 break;
1078 case DCA_3F2R:
1079 c = channel_mapping[0];
1080 l = channel_mapping[1];
1081 r = channel_mapping[2];
1082 sl = channel_mapping[3];
1083 sr = channel_mapping[4];
1084 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1085 MIX_REAR2(samples, sl, sr, 3, coef));
1086 break;
1087 }
1088 }
1089
1090
1091 #ifndef decode_blockcodes
1092 /* Very compact version of the block code decoder that does not use table
1093 * look-up but is slightly slower */
1094 static int decode_blockcode(int code, int levels, int *values)
1095 {
1096 int i;
1097 int offset = (levels - 1) >> 1;
1098
1099 for (i = 0; i < 4; i++) {
1100 int div = FASTDIV(code, levels);
1101 values[i] = code - offset - div * levels;
1102 code = div;
1103 }
1104
1105 return code;
1106 }
1107
1108 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1109 {
1110 return decode_blockcode(code1, levels, values) |
1111 decode_blockcode(code2, levels, values + 4);
1112 }
1113 #endif
1114
1115 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1116 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1117
1118 #ifndef int8x8_fmul_int32
1119 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1120 {
1121 float fscale = scale / 16.0;
1122 int i;
1123 for (i = 0; i < 8; i++)
1124 dst[i] = src[i] * fscale;
1125 }
1126 #endif
1127
1128 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1129 {
1130 int k, l;
1131 int subsubframe = s->current_subsubframe;
1132
1133 const float *quant_step_table;
1134
1135 /* FIXME */
1136 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1137 LOCAL_ALIGNED_16(int, block, [8]);
1138
1139 /*
1140 * Audio data
1141 */
1142
1143 /* Select quantization step size table */
1144 if (s->bit_rate_index == 0x1f)
1145 quant_step_table = lossless_quant_d;
1146 else
1147 quant_step_table = lossy_quant_d;
1148
1149 for (k = base_channel; k < s->prim_channels; k++) {
1150 if (get_bits_left(&s->gb) < 0)
1151 return AVERROR_INVALIDDATA;
1152
1153 for (l = 0; l < s->vq_start_subband[k]; l++) {
1154 int m;
1155
1156 /* Select the mid-tread linear quantizer */
1157 int abits = s->bitalloc[k][l];
1158
1159 float quant_step_size = quant_step_table[abits];
1160
1161 /*
1162 * Determine quantization index code book and its type
1163 */
1164
1165 /* Select quantization index code book */
1166 int sel = s->quant_index_huffman[k][abits];
1167
1168 /*
1169 * Extract bits from the bit stream
1170 */
1171 if (!abits) {
1172 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1173 } else {
1174 /* Deal with transients */
1175 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1176 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1177 s->scalefactor_adj[k][sel];
1178
1179 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1180 if (abits <= 7) {
1181 /* Block code */
1182 int block_code1, block_code2, size, levels, err;
1183
1184 size = abits_sizes[abits - 1];
1185 levels = abits_levels[abits - 1];
1186
1187 block_code1 = get_bits(&s->gb, size);
1188 block_code2 = get_bits(&s->gb, size);
1189 err = decode_blockcodes(block_code1, block_code2,
1190 levels, block);
1191 if (err) {
1192 av_log(s->avctx, AV_LOG_ERROR,
1193 "ERROR: block code look-up failed\n");
1194 return AVERROR_INVALIDDATA;
1195 }
1196 } else {
1197 /* no coding */
1198 for (m = 0; m < 8; m++)
1199 block[m] = get_sbits(&s->gb, abits - 3);
1200 }
1201 } else {
1202 /* Huffman coded */
1203 for (m = 0; m < 8; m++)
1204 block[m] = get_bitalloc(&s->gb,
1205 &dca_smpl_bitalloc[abits], sel);
1206 }
1207
1208 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1209 block, rscale, 8);
1210 }
1211
1212 /*
1213 * Inverse ADPCM if in prediction mode
1214 */
1215 if (s->prediction_mode[k][l]) {
1216 int n;
1217 for (m = 0; m < 8; m++) {
1218 for (n = 1; n <= 4; n++)
1219 if (m >= n)
1220 subband_samples[k][l][m] +=
1221 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1222 subband_samples[k][l][m - n] / 8192);
1223 else if (s->predictor_history)
1224 subband_samples[k][l][m] +=
1225 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1226 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1227 }
1228 }
1229 }
1230
1231 /*
1232 * Decode VQ encoded high frequencies
1233 */
1234 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1235 /* 1 vector -> 32 samples but we only need the 8 samples
1236 * for this subsubframe. */
1237 int hfvq = s->high_freq_vq[k][l];
1238
1239 if (!s->debug_flag & 0x01) {
1240 av_log(s->avctx, AV_LOG_DEBUG,
1241 "Stream with high frequencies VQ coding\n");
1242 s->debug_flag |= 0x01;
1243 }
1244
1245 int8x8_fmul_int32(subband_samples[k][l],
1246 &high_freq_vq[hfvq][subsubframe * 8],
1247 s->scale_factor[k][l][0]);
1248 }
1249 }
1250
1251 /* Check for DSYNC after subsubframe */
1252 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1253 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1254 #ifdef TRACE
1255 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1256 #endif
1257 } else {
1258 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1259 }
1260 }
1261
1262 /* Backup predictor history for adpcm */
1263 for (k = base_channel; k < s->prim_channels; k++)
1264 for (l = 0; l < s->vq_start_subband[k]; l++)
1265 memcpy(s->subband_samples_hist[k][l],
1266 &subband_samples[k][l][4],
1267 4 * sizeof(subband_samples[0][0][0]));
1268
1269 return 0;
1270 }
1271
1272 static int dca_filter_channels(DCAContext *s, int block_index)
1273 {
1274 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1275 int k;
1276
1277 /* 32 subbands QMF */
1278 for (k = 0; k < s->prim_channels; k++) {
1279 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1280 0, 8388608.0, 8388608.0 };*/
1281 if (s->channel_order_tab[k] >= 0)
1282 qmf_32_subbands(s, k, subband_samples[k],
1283 s->samples_chanptr[s->channel_order_tab[k]],
1284 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1285 }
1286
1287 /* Down mixing */
1288 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1289 dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
1290 }
1291
1292 /* Generate LFE samples for this subsubframe FIXME!!! */
1293 if (s->output & DCA_LFE) {
1294 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1295 s->lfe_data + 2 * s->lfe * (block_index + 4),
1296 s->samples_chanptr[dca_lfe_index[s->amode]],
1297 1.0 / (256.0 * 32768.0));
1298 /* Outputs 20bits pcm samples */
1299 }
1300
1301 return 0;
1302 }
1303
1304
1305 static int dca_subframe_footer(DCAContext *s, int base_channel)
1306 {
1307 int aux_data_count = 0, i;
1308
1309 /*
1310 * Unpack optional information
1311 */
1312
1313 /* presumably optional information only appears in the core? */
1314 if (!base_channel) {
1315 if (s->timestamp)
1316 skip_bits_long(&s->gb, 32);
1317
1318 if (s->aux_data)
1319 aux_data_count = get_bits(&s->gb, 6);
1320
1321 for (i = 0; i < aux_data_count; i++)
1322 get_bits(&s->gb, 8);
1323
1324 if (s->crc_present && (s->downmix || s->dynrange))
1325 get_bits(&s->gb, 16);
1326 }
1327
1328 return 0;
1329 }
1330
1331 /**
1332 * Decode a dca frame block
1333 *
1334 * @param s pointer to the DCAContext
1335 */
1336
1337 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1338 {
1339 int ret;
1340
1341 /* Sanity check */
1342 if (s->current_subframe >= s->subframes) {
1343 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1344 s->current_subframe, s->subframes);
1345 return AVERROR_INVALIDDATA;
1346 }
1347
1348 if (!s->current_subsubframe) {
1349 #ifdef TRACE
1350 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1351 #endif
1352 /* Read subframe header */
1353 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1354 return ret;
1355 }
1356
1357 /* Read subsubframe */
1358 #ifdef TRACE
1359 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1360 #endif
1361 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1362 return ret;
1363
1364 /* Update state */
1365 s->current_subsubframe++;
1366 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1367 s->current_subsubframe = 0;
1368 s->current_subframe++;
1369 }
1370 if (s->current_subframe >= s->subframes) {
1371 #ifdef TRACE
1372 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1373 #endif
1374 /* Read subframe footer */
1375 if ((ret = dca_subframe_footer(s, base_channel)))
1376 return ret;
1377 }
1378
1379 return 0;
1380 }
1381
1382 /**
1383 * Return the number of channels in an ExSS speaker mask (HD)
1384 */
1385 static int dca_exss_mask2count(int mask)
1386 {
1387 /* count bits that mean speaker pairs twice */
1388 return av_popcount(mask) +
1389 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1390 DCA_EXSS_FRONT_LEFT_RIGHT |
1391 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1392 DCA_EXSS_WIDE_LEFT_RIGHT |
1393 DCA_EXSS_SIDE_LEFT_RIGHT |
1394 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1395 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1396 DCA_EXSS_REAR_LEFT_RIGHT |
1397 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1398 }
1399
1400 /**
1401 * Skip mixing coefficients of a single mix out configuration (HD)
1402 */
1403 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1404 {
1405 int i;
1406
1407 for (i = 0; i < channels; i++) {
1408 int mix_map_mask = get_bits(gb, out_ch);
1409 int num_coeffs = av_popcount(mix_map_mask);
1410 skip_bits_long(gb, num_coeffs * 6);
1411 }
1412 }
1413
1414 /**
1415 * Parse extension substream asset header (HD)
1416 */
1417 static int dca_exss_parse_asset_header(DCAContext *s)
1418 {
1419 int header_pos = get_bits_count(&s->gb);
1420 int header_size;
1421 int channels;
1422 int embedded_stereo = 0;
1423 int embedded_6ch = 0;
1424 int drc_code_present;
1425 int extensions_mask;
1426 int i, j;
1427
1428 if (get_bits_left(&s->gb) < 16)
1429 return -1;
1430
1431 /* We will parse just enough to get to the extensions bitmask with which
1432 * we can set the profile value. */
1433
1434 header_size = get_bits(&s->gb, 9) + 1;
1435 skip_bits(&s->gb, 3); // asset index
1436
1437 if (s->static_fields) {
1438 if (get_bits1(&s->gb))
1439 skip_bits(&s->gb, 4); // asset type descriptor
1440 if (get_bits1(&s->gb))
1441 skip_bits_long(&s->gb, 24); // language descriptor
1442
1443 if (get_bits1(&s->gb)) {
1444 /* How can one fit 1024 bytes of text here if the maximum value
1445 * for the asset header size field above was 512 bytes? */
1446 int text_length = get_bits(&s->gb, 10) + 1;
1447 if (get_bits_left(&s->gb) < text_length * 8)
1448 return -1;
1449 skip_bits_long(&s->gb, text_length * 8); // info text
1450 }
1451
1452 skip_bits(&s->gb, 5); // bit resolution - 1
1453 skip_bits(&s->gb, 4); // max sample rate code
1454 channels = get_bits(&s->gb, 8) + 1;
1455
1456 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1457 int spkr_remap_sets;
1458 int spkr_mask_size = 16;
1459 int num_spkrs[7];
1460
1461 if (channels > 2)
1462 embedded_stereo = get_bits1(&s->gb);
1463 if (channels > 6)
1464 embedded_6ch = get_bits1(&s->gb);
1465
1466 if (get_bits1(&s->gb)) {
1467 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1468 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1469 }
1470
1471 spkr_remap_sets = get_bits(&s->gb, 3);
1472
1473 for (i = 0; i < spkr_remap_sets; i++) {
1474 /* std layout mask for each remap set */
1475 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1476 }
1477
1478 for (i = 0; i < spkr_remap_sets; i++) {
1479 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1480 if (get_bits_left(&s->gb) < 0)
1481 return -1;
1482
1483 for (j = 0; j < num_spkrs[i]; j++) {
1484 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1485 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1486 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1487 }
1488 }
1489
1490 } else {
1491 skip_bits(&s->gb, 3); // representation type
1492 }
1493 }
1494
1495 drc_code_present = get_bits1(&s->gb);
1496 if (drc_code_present)
1497 get_bits(&s->gb, 8); // drc code
1498
1499 if (get_bits1(&s->gb))
1500 skip_bits(&s->gb, 5); // dialog normalization code
1501
1502 if (drc_code_present && embedded_stereo)
1503 get_bits(&s->gb, 8); // drc stereo code
1504
1505 if (s->mix_metadata && get_bits1(&s->gb)) {
1506 skip_bits(&s->gb, 1); // external mix
1507 skip_bits(&s->gb, 6); // post mix gain code
1508
1509 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1510 skip_bits(&s->gb, 3); // drc limit
1511 else
1512 skip_bits(&s->gb, 8); // custom drc code
1513
1514 if (get_bits1(&s->gb)) // channel specific scaling
1515 for (i = 0; i < s->num_mix_configs; i++)
1516 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1517 else
1518 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1519
1520 for (i = 0; i < s->num_mix_configs; i++) {
1521 if (get_bits_left(&s->gb) < 0)
1522 return -1;
1523 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1524 if (embedded_6ch)
1525 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1526 if (embedded_stereo)
1527 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1528 }
1529 }
1530
1531 switch (get_bits(&s->gb, 2)) {
1532 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1533 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1534 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1535 case 3: extensions_mask = 0; /* aux coding */ break;
1536 }
1537
1538 /* not parsed further, we were only interested in the extensions mask */
1539
1540 if (get_bits_left(&s->gb) < 0)
1541 return -1;
1542
1543 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1544 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1545 return -1;
1546 }
1547 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1548
1549 if (extensions_mask & DCA_EXT_EXSS_XLL)
1550 s->profile = FF_PROFILE_DTS_HD_MA;
1551 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1552 DCA_EXT_EXSS_XXCH))
1553 s->profile = FF_PROFILE_DTS_HD_HRA;
1554
1555 if (!(extensions_mask & DCA_EXT_CORE))
1556 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1557 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1558 av_log(s->avctx, AV_LOG_WARNING,
1559 "DTS extensions detection mismatch (%d, %d)\n",
1560 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1561
1562 return 0;
1563 }
1564
1565 /**
1566 * Parse extension substream header (HD)
1567 */
1568 static void dca_exss_parse_header(DCAContext *s)
1569 {
1570 int ss_index;
1571 int blownup;
1572 int num_audiop = 1;
1573 int num_assets = 1;
1574 int active_ss_mask[8];
1575 int i, j;
1576
1577 if (get_bits_left(&s->gb) < 52)
1578 return;
1579
1580 skip_bits(&s->gb, 8); // user data
1581 ss_index = get_bits(&s->gb, 2);
1582
1583 blownup = get_bits1(&s->gb);
1584 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1585 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1586
1587 s->static_fields = get_bits1(&s->gb);
1588 if (s->static_fields) {
1589 skip_bits(&s->gb, 2); // reference clock code
1590 skip_bits(&s->gb, 3); // frame duration code
1591
1592 if (get_bits1(&s->gb))
1593 skip_bits_long(&s->gb, 36); // timestamp
1594
1595 /* a single stream can contain multiple audio assets that can be
1596 * combined to form multiple audio presentations */
1597
1598 num_audiop = get_bits(&s->gb, 3) + 1;
1599 if (num_audiop > 1) {
1600 avpriv_request_sample(s->avctx,
1601 "Multiple DTS-HD audio presentations");
1602 /* ignore such streams for now */
1603 return;
1604 }
1605
1606 num_assets = get_bits(&s->gb, 3) + 1;
1607 if (num_assets > 1) {
1608 avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
1609 /* ignore such streams for now */
1610 return;
1611 }
1612
1613 for (i = 0; i < num_audiop; i++)
1614 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1615
1616 for (i = 0; i < num_audiop; i++)
1617 for (j = 0; j <= ss_index; j++)
1618 if (active_ss_mask[i] & (1 << j))
1619 skip_bits(&s->gb, 8); // active asset mask
1620
1621 s->mix_metadata = get_bits1(&s->gb);
1622 if (s->mix_metadata) {
1623 int mix_out_mask_size;
1624
1625 skip_bits(&s->gb, 2); // adjustment level
1626 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1627 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1628
1629 for (i = 0; i < s->num_mix_configs; i++) {
1630 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1631 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1632 }
1633 }
1634 }
1635
1636 for (i = 0; i < num_assets; i++)
1637 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1638
1639 for (i = 0; i < num_assets; i++) {
1640 if (dca_exss_parse_asset_header(s))
1641 return;
1642 }
1643
1644 /* not parsed further, we were only interested in the extensions mask
1645 * from the asset header */
1646 }
1647
1648 /**
1649 * Main frame decoding function
1650 * FIXME add arguments
1651 */
1652 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1653 int *got_frame_ptr, AVPacket *avpkt)
1654 {
1655 AVFrame *frame = data;
1656 const uint8_t *buf = avpkt->data;
1657 int buf_size = avpkt->size;
1658
1659 int lfe_samples;
1660 int num_core_channels = 0;
1661 int i, ret;
1662 float **samples_flt;
1663 DCAContext *s = avctx->priv_data;
1664 int channels, full_channels;
1665 int core_ss_end;
1666
1667
1668 s->xch_present = 0;
1669
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;
1675 }
1676
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
1680 return ret;
1681 }
1682 //set AVCodec values with parsed data
1683 avctx->sample_rate = s->sample_rate;
1684 avctx->bit_rate = s->bit_rate;
1685
1686 s->profile = FF_PROFILE_DTS;
1687
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");
1691 return ret;
1692 }
1693 }
1694
1695 /* record number of core channels incase less than max channels are requested */
1696 num_core_channels = s->prim_channels;
1697
1698 if (s->ext_coding)
1699 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1700 else
1701 s->core_ext_mask = 0;
1702
1703 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1704
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) {
1708
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);
1712
1713 /* extensions start at 32-bit boundaries into bitstream */
1714 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1715
1716 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1717 uint32_t bits = get_bits_long(&s->gb, 32);
1718
1719 switch (bits) {
1720 case 0x5a5a5a5a: {
1721 int ext_amode, xch_fsize;
1722
1723 s->xch_base_channel = s->prim_channels;
1724
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))
1729 continue;
1730
1731 /* skip length-to-end-of-frame field for the moment */
1732 skip_bits(&s->gb, 10);
1733
1734 s->core_ext_mask |= DCA_EXT_XCH;
1735
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);
1741 continue;
1742 }
1743
1744 /* much like core primary audio coding header */
1745 dca_parse_audio_coding_header(s, s->xch_base_channel);
1746
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");
1750 continue;
1751 }
1752
1753 s->xch_present = 1;
1754 break;
1755 }
1756 case 0x47004a03:
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;
1761 break;
1762
1763 case 0x1d95f262: {
1764 int fsize96 = show_bits(&s->gb, 12) + 1;
1765 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1766 continue;
1767
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));
1773
1774 s->core_ext_mask |= DCA_EXT_X96;
1775 break;
1776 }
1777 }
1778
1779 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1780 }
1781 } else {
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));
1784 }
1785
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;
1790
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);
1795
1796 avctx->profile = s->profile;
1797
1798 full_channels = channels = s->prim_channels + !!s->lfe;
1799
1800 if (s->amode < 16) {
1801 avctx->channel_layout = dca_core_channel_layout[s->amode];
1802
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;
1806 if (s->lfe) {
1807 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1808 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1809 } else {
1810 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1811 }
1812 } else {
1813 channels = num_core_channels + !!s->lfe;
1814 s->xch_present = 0; /* disable further xch processing */
1815 if (s->lfe) {
1816 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1817 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1818 } else
1819 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1820 }
1821
1822 if (channels > !!s->lfe &&
1823 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1824 return AVERROR_INVALIDDATA;
1825
1826 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1827 channels = 2;
1828 s->output = DCA_STEREO;
1829 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1830 }
1831 } else {
1832 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1833 return AVERROR_INVALIDDATA;
1834 }
1835 avctx->channels = channels;
1836
1837 /* get output buffer */
1838 frame->nb_samples = 256 * (s->sample_blocks / 8);
1839 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
1840 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1841 return ret;
1842 }
1843 samples_flt = (float **)frame->extended_data;
1844
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 frame->nb_samples,
1849 avctx->sample_fmt, 0);
1850 if (ret < 0)
1851 return ret;
1852
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);
1857
1858 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
1859 s->extra_channels_buffer,
1860 full_channels - channels,
1861 frame->nb_samples, avctx->sample_fmt, 0);
1862 if (ret < 0)
1863 return ret;
1864 }
1865
1866 /* filter to get final output */
1867 for (i = 0; i < (s->sample_blocks / 8); i++) {
1868 int ch;
1869
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;
1874
1875 dca_filter_channels(s, i);
1876
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);
1885 }
1886 }
1887
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];
1892
1893 *got_frame_ptr = 1;
1894
1895 return buf_size;
1896 }
1897
1898
1899
1900 /**
1901 * DCA initialization
1902 *
1903 * @param avctx pointer to the AVCodecContext
1904 */
1905
1906 static av_cold int dca_decode_init(AVCodecContext *avctx)
1907 {
1908 DCAContext *s = avctx->priv_data;
1909
1910 s->avctx = avctx;
1911 dca_init_vlcs();
1912
1913 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
1914 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1915 ff_synth_filter_init(&s->synth);
1916 ff_dcadsp_init(&s->dcadsp);
1917 ff_fmt_convert_init(&s->fmt_conv, avctx);
1918
1919 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
1920
1921 /* allow downmixing to stereo */
1922 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1923 avctx->request_channels == 2) {
1924 avctx->channels = avctx->request_channels;
1925 }
1926
1927 return 0;
1928 }
1929
1930 static av_cold int dca_decode_end(AVCodecContext *avctx)
1931 {
1932 DCAContext *s = avctx->priv_data;
1933 ff_mdct_end(&s->imdct);
1934 av_freep(&s->extra_channels_buffer);
1935 return 0;
1936 }
1937
1938 static const AVProfile profiles[] = {
1939 { FF_PROFILE_DTS, "DTS" },
1940 { FF_PROFILE_DTS_ES, "DTS-ES" },
1941 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1942 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1943 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1944 { FF_PROFILE_UNKNOWN },
1945 };
1946
1947 AVCodec ff_dca_decoder = {
1948 .name = "dca",
1949 .type = AVMEDIA_TYPE_AUDIO,
1950 .id = AV_CODEC_ID_DTS,
1951 .priv_data_size = sizeof(DCAContext),
1952 .init = dca_decode_init,
1953 .decode = dca_decode_frame,
1954 .close = dca_decode_end,
1955 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1956 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1957 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
1958 AV_SAMPLE_FMT_NONE },
1959 .profiles = NULL_IF_CONFIG_SMALL(profiles),
1960 };