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/internal.h"
33 #include "libavutil/intreadwrite.h"
34 #include "libavutil/mathematics.h"
35 #include "libavutil/opt.h"
36 #include "libavutil/samplefmt.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 #define DCA_NSYNCAUX 0x9A1105A0
268 /** Bit allocation */
270 int offset; ///< code values offset
271 int maxbits[8]; ///< max bits in VLC
272 int wrap; ///< wrap for get_vlc2()
273 VLC vlc[8]; ///< actual codes
276 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
277 static BitAlloc dca_tmode; ///< transition mode VLCs
278 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
279 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
281 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
284 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
289 AVClass *class; ///< class for AVOptions
290 AVCodecContext *avctx;
292 int frame_type; ///< type of the current frame
293 int samples_deficit; ///< deficit sample count
294 int crc_present; ///< crc is present in the bitstream
295 int sample_blocks; ///< number of PCM sample blocks
296 int frame_size; ///< primary frame byte size
297 int amode; ///< audio channels arrangement
298 int sample_rate; ///< audio sampling rate
299 int bit_rate; ///< transmission bit rate
300 int bit_rate_index; ///< transmission bit rate index
302 int dynrange; ///< embedded dynamic range flag
303 int timestamp; ///< embedded time stamp flag
304 int aux_data; ///< auxiliary data flag
305 int hdcd; ///< source material is mastered in HDCD
306 int ext_descr; ///< extension audio descriptor flag
307 int ext_coding; ///< extended coding flag
308 int aspf; ///< audio sync word insertion flag
309 int lfe; ///< low frequency effects flag
310 int predictor_history; ///< predictor history flag
311 int header_crc; ///< header crc check bytes
312 int multirate_inter; ///< multirate interpolator switch
313 int version; ///< encoder software revision
314 int copy_history; ///< copy history
315 int source_pcm_res; ///< source pcm resolution
316 int front_sum; ///< front sum/difference flag
317 int surround_sum; ///< surround sum/difference flag
318 int dialog_norm; ///< dialog normalisation parameter
320 /* Primary audio coding header */
321 int subframes; ///< number of subframes
322 int total_channels; ///< number of channels including extensions
323 int prim_channels; ///< number of primary audio channels
324 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
325 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
326 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
327 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
328 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
329 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
330 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
331 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
333 /* Primary audio coding side information */
334 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
335 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
336 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
337 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
338 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
339 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
340 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
341 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
342 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
343 float downmix_coef[DCA_PRIM_CHANNELS_MAX + 1][2]; ///< stereo downmix coefficients
344 int dynrange_coef; ///< dynamic range coefficient
346 /* Core substream's embedded downmix coefficients (cf. ETSI TS 102 114 V1.4.1)
347 * Input: primary audio channels (incl. LFE if present)
348 * Output: downmix audio channels (up to 4, no LFE) */
349 uint8_t core_downmix; ///< embedded downmix coefficients available
350 uint8_t core_downmix_amode; ///< audio channel arrangement of embedded downmix
351 uint16_t core_downmix_codes[DCA_PRIM_CHANNELS_MAX + 1][4]; ///< embedded downmix coefficients (9-bit codes)
353 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
355 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
356 int lfe_scale_factor;
358 /* Subband samples history (for ADPCM) */
359 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
360 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
361 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
362 int hist_index[DCA_PRIM_CHANNELS_MAX];
363 DECLARE_ALIGNED(32, float, raXin)[32];
365 int output; ///< type of output
367 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
368 float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
369 float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
370 uint8_t *extra_channels_buffer;
371 unsigned int extra_channels_buffer_size;
373 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
374 int dca_buffer_size; ///< how much data is in the dca_buffer
376 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
378 /* Current position in DCA frame */
379 int current_subframe;
380 int current_subsubframe;
382 int core_ext_mask; ///< present extensions in the core substream
384 /* XCh extension information */
385 int xch_present; ///< XCh extension present and valid
386 int xch_base_channel; ///< index of first (only) channel containing XCH data
387 int xch_disable; ///< whether the XCh extension should be decoded or not
389 /* ExSS header parser */
390 int static_fields; ///< static fields present
391 int mix_metadata; ///< mixing metadata present
392 int num_mix_configs; ///< number of mix out configurations
393 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
397 int debug_flag; ///< used for suppressing repeated error messages output
398 AVFloatDSPContext fdsp;
400 SynthFilterContext synth;
401 DCADSPContext dcadsp;
402 FmtConvertContext fmt_conv;
405 static const uint16_t dca_vlc_offs[] = {
406 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
407 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
408 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
409 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
410 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
411 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
414 static av_cold void dca_init_vlcs(void)
416 static int vlcs_initialized = 0;
418 static VLC_TYPE dca_table[23622][2];
420 if (vlcs_initialized)
423 dca_bitalloc_index.offset = 1;
424 dca_bitalloc_index.wrap = 2;
425 for (i = 0; i < 5; i++) {
426 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
427 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
428 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
429 bitalloc_12_bits[i], 1, 1,
430 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
432 dca_scalefactor.offset = -64;
433 dca_scalefactor.wrap = 2;
434 for (i = 0; i < 5; i++) {
435 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
436 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
437 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
438 scales_bits[i], 1, 1,
439 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
441 dca_tmode.offset = 0;
443 for (i = 0; i < 4; i++) {
444 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
445 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
446 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
448 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
451 for (i = 0; i < 10; i++)
452 for (j = 0; j < 7; j++) {
453 if (!bitalloc_codes[i][j])
455 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
456 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
457 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
458 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
460 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
462 bitalloc_bits[i][j], 1, 1,
463 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
466 vlcs_initialized = 1;
469 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
472 *dst++ = get_bits(gb, bits);
475 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
478 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
479 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
480 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
482 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
483 s->prim_channels = s->total_channels;
485 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
486 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
489 for (i = base_channel; i < s->prim_channels; i++) {
490 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
491 if (s->subband_activity[i] > DCA_SUBBANDS)
492 s->subband_activity[i] = DCA_SUBBANDS;
494 for (i = base_channel; i < s->prim_channels; i++) {
495 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
496 if (s->vq_start_subband[i] > DCA_SUBBANDS)
497 s->vq_start_subband[i] = DCA_SUBBANDS;
499 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
500 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
501 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
502 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
504 /* Get codebooks quantization indexes */
506 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
507 for (j = 1; j < 11; j++)
508 for (i = base_channel; i < s->prim_channels; i++)
509 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
511 /* Get scale factor adjustment */
512 for (j = 0; j < 11; j++)
513 for (i = base_channel; i < s->prim_channels; i++)
514 s->scalefactor_adj[i][j] = 1;
516 for (j = 1; j < 11; j++)
517 for (i = base_channel; i < s->prim_channels; i++)
518 if (s->quant_index_huffman[i][j] < thr[j])
519 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
521 if (s->crc_present) {
522 /* Audio header CRC check */
523 get_bits(&s->gb, 16);
526 s->current_subframe = 0;
527 s->current_subsubframe = 0;
530 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
531 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
532 for (i = base_channel; i < s->prim_channels; i++) {
533 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
534 s->subband_activity[i]);
535 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
536 s->vq_start_subband[i]);
537 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
538 s->joint_intensity[i]);
539 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
540 s->transient_huffman[i]);
541 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
542 s->scalefactor_huffman[i]);
543 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
544 s->bitalloc_huffman[i]);
545 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
546 for (j = 0; j < 11; j++)
547 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
548 av_log(s->avctx, AV_LOG_DEBUG, "\n");
549 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
550 for (j = 0; j < 11; j++)
551 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
552 av_log(s->avctx, AV_LOG_DEBUG, "\n");
559 static int dca_parse_frame_header(DCAContext *s)
561 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
564 skip_bits_long(&s->gb, 32);
567 s->frame_type = get_bits(&s->gb, 1);
568 s->samples_deficit = get_bits(&s->gb, 5) + 1;
569 s->crc_present = get_bits(&s->gb, 1);
570 s->sample_blocks = get_bits(&s->gb, 7) + 1;
571 s->frame_size = get_bits(&s->gb, 14) + 1;
572 if (s->frame_size < 95)
573 return AVERROR_INVALIDDATA;
574 s->amode = get_bits(&s->gb, 6);
575 s->sample_rate = avpriv_dca_sample_rates[get_bits(&s->gb, 4)];
577 return AVERROR_INVALIDDATA;
578 s->bit_rate_index = get_bits(&s->gb, 5);
579 s->bit_rate = dca_bit_rates[s->bit_rate_index];
581 return AVERROR_INVALIDDATA;
583 skip_bits1(&s->gb); // always 0 (reserved, cf. ETSI TS 102 114 V1.4.1)
584 s->dynrange = get_bits(&s->gb, 1);
585 s->timestamp = get_bits(&s->gb, 1);
586 s->aux_data = get_bits(&s->gb, 1);
587 s->hdcd = get_bits(&s->gb, 1);
588 s->ext_descr = get_bits(&s->gb, 3);
589 s->ext_coding = get_bits(&s->gb, 1);
590 s->aspf = get_bits(&s->gb, 1);
591 s->lfe = get_bits(&s->gb, 2);
592 s->predictor_history = get_bits(&s->gb, 1);
595 av_log(s->avctx, AV_LOG_ERROR, "Invalid LFE value: %d\n", s->lfe);
596 return AVERROR_INVALIDDATA;
599 /* TODO: check CRC */
601 s->header_crc = get_bits(&s->gb, 16);
603 s->multirate_inter = get_bits(&s->gb, 1);
604 s->version = get_bits(&s->gb, 4);
605 s->copy_history = get_bits(&s->gb, 2);
606 s->source_pcm_res = get_bits(&s->gb, 3);
607 s->front_sum = get_bits(&s->gb, 1);
608 s->surround_sum = get_bits(&s->gb, 1);
609 s->dialog_norm = get_bits(&s->gb, 4);
611 /* FIXME: channels mixing levels */
612 s->output = s->amode;
614 s->output |= DCA_LFE;
617 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
618 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
619 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
620 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
621 s->sample_blocks, s->sample_blocks * 32);
622 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
623 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
624 s->amode, dca_channels[s->amode]);
625 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
627 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
629 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
630 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
631 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
632 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
633 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
634 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
635 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
636 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
637 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
638 s->predictor_history);
639 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
640 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
642 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
643 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
644 av_log(s->avctx, AV_LOG_DEBUG,
645 "source pcm resolution: %i (%i bits/sample)\n",
646 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
647 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
648 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
649 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
650 av_log(s->avctx, AV_LOG_DEBUG, "\n");
653 /* Primary audio coding header */
654 s->subframes = get_bits(&s->gb, 4) + 1;
656 return dca_parse_audio_coding_header(s, 0);
660 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
663 /* huffman encoded */
664 value += get_bitalloc(gb, &dca_scalefactor, level);
665 value = av_clip(value, 0, (1 << log2range) - 1);
666 } else if (level < 8) {
667 if (level + 1 > log2range) {
668 skip_bits(gb, level + 1 - log2range);
669 value = get_bits(gb, log2range);
671 value = get_bits(gb, level + 1);
677 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
679 /* Primary audio coding side information */
682 if (get_bits_left(&s->gb) < 0)
683 return AVERROR_INVALIDDATA;
686 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
687 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
690 for (j = base_channel; j < s->prim_channels; j++) {
691 for (k = 0; k < s->subband_activity[j]; k++)
692 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
695 /* Get prediction codebook */
696 for (j = base_channel; j < s->prim_channels; j++) {
697 for (k = 0; k < s->subband_activity[j]; k++) {
698 if (s->prediction_mode[j][k] > 0) {
699 /* (Prediction coefficient VQ address) */
700 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
705 /* Bit allocation index */
706 for (j = base_channel; j < s->prim_channels; j++) {
707 for (k = 0; k < s->vq_start_subband[j]; k++) {
708 if (s->bitalloc_huffman[j] == 6)
709 s->bitalloc[j][k] = get_bits(&s->gb, 5);
710 else if (s->bitalloc_huffman[j] == 5)
711 s->bitalloc[j][k] = get_bits(&s->gb, 4);
712 else if (s->bitalloc_huffman[j] == 7) {
713 av_log(s->avctx, AV_LOG_ERROR,
714 "Invalid bit allocation index\n");
715 return AVERROR_INVALIDDATA;
718 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
721 if (s->bitalloc[j][k] > 26) {
722 av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
723 j, k, s->bitalloc[j][k]);
724 return AVERROR_INVALIDDATA;
729 /* Transition mode */
730 for (j = base_channel; j < s->prim_channels; j++) {
731 for (k = 0; k < s->subband_activity[j]; k++) {
732 s->transition_mode[j][k] = 0;
733 if (s->subsubframes[s->current_subframe] > 1 &&
734 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
735 s->transition_mode[j][k] =
736 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
741 if (get_bits_left(&s->gb) < 0)
742 return AVERROR_INVALIDDATA;
744 for (j = base_channel; j < s->prim_channels; j++) {
745 const uint32_t *scale_table;
746 int scale_sum, log_size;
748 memset(s->scale_factor[j], 0,
749 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
751 if (s->scalefactor_huffman[j] == 6) {
752 scale_table = scale_factor_quant7;
755 scale_table = scale_factor_quant6;
759 /* When huffman coded, only the difference is encoded */
762 for (k = 0; k < s->subband_activity[j]; k++) {
763 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
764 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
765 s->scale_factor[j][k][0] = scale_table[scale_sum];
768 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
769 /* Get second scale factor */
770 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
771 s->scale_factor[j][k][1] = scale_table[scale_sum];
776 /* Joint subband scale factor codebook select */
777 for (j = base_channel; j < s->prim_channels; j++) {
778 /* Transmitted only if joint subband coding enabled */
779 if (s->joint_intensity[j] > 0)
780 s->joint_huff[j] = get_bits(&s->gb, 3);
783 if (get_bits_left(&s->gb) < 0)
784 return AVERROR_INVALIDDATA;
786 /* Scale factors for joint subband coding */
787 for (j = base_channel; j < s->prim_channels; j++) {
790 /* Transmitted only if joint subband coding enabled */
791 if (s->joint_intensity[j] > 0) {
793 source_channel = s->joint_intensity[j] - 1;
795 /* When huffman coded, only the difference is encoded
796 * (is this valid as well for joint scales ???) */
798 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
799 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
800 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
803 if (!(s->debug_flag & 0x02)) {
804 av_log(s->avctx, AV_LOG_DEBUG,
805 "Joint stereo coding not supported\n");
806 s->debug_flag |= 0x02;
811 /* Dynamic range coefficient */
812 if (!base_channel && s->dynrange)
813 s->dynrange_coef = get_bits(&s->gb, 8);
815 /* Side information CRC check word */
816 if (s->crc_present) {
817 get_bits(&s->gb, 16);
821 * Primary audio data arrays
824 /* VQ encoded high frequency subbands */
825 for (j = base_channel; j < s->prim_channels; j++)
826 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
827 /* 1 vector -> 32 samples */
828 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
830 /* Low frequency effect data */
831 if (!base_channel && s->lfe) {
833 int lfe_samples = 2 * s->lfe * (4 + block_index);
834 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
837 for (j = lfe_samples; j < lfe_end_sample; j++) {
838 /* Signed 8 bits int */
839 s->lfe_data[j] = get_sbits(&s->gb, 8);
842 /* Scale factor index */
843 skip_bits(&s->gb, 1);
844 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 7)];
846 /* Quantization step size * scale factor */
847 lfe_scale = 0.035 * s->lfe_scale_factor;
849 for (j = lfe_samples; j < lfe_end_sample; j++)
850 s->lfe_data[j] *= lfe_scale;
854 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
855 s->subsubframes[s->current_subframe]);
856 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
857 s->partial_samples[s->current_subframe]);
859 for (j = base_channel; j < s->prim_channels; j++) {
860 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
861 for (k = 0; k < s->subband_activity[j]; k++)
862 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
863 av_log(s->avctx, AV_LOG_DEBUG, "\n");
865 for (j = base_channel; j < s->prim_channels; j++) {
866 for (k = 0; k < s->subband_activity[j]; k++)
867 av_log(s->avctx, AV_LOG_DEBUG,
868 "prediction coefs: %f, %f, %f, %f\n",
869 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
870 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
871 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
872 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
874 for (j = base_channel; j < s->prim_channels; j++) {
875 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
876 for (k = 0; k < s->vq_start_subband[j]; k++)
877 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
878 av_log(s->avctx, AV_LOG_DEBUG, "\n");
880 for (j = base_channel; j < s->prim_channels; j++) {
881 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
882 for (k = 0; k < s->subband_activity[j]; k++)
883 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
884 av_log(s->avctx, AV_LOG_DEBUG, "\n");
886 for (j = base_channel; j < s->prim_channels; j++) {
887 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
888 for (k = 0; k < s->subband_activity[j]; k++) {
889 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
890 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
891 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
892 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
894 av_log(s->avctx, AV_LOG_DEBUG, "\n");
896 for (j = base_channel; j < s->prim_channels; j++) {
897 if (s->joint_intensity[j] > 0) {
898 int source_channel = s->joint_intensity[j] - 1;
899 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
900 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
901 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
902 av_log(s->avctx, AV_LOG_DEBUG, "\n");
905 for (j = base_channel; j < s->prim_channels; j++)
906 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
907 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
908 if (!base_channel && s->lfe) {
909 int lfe_samples = 2 * s->lfe * (4 + block_index);
910 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
912 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
913 for (j = lfe_samples; j < lfe_end_sample; j++)
914 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
915 av_log(s->avctx, AV_LOG_DEBUG, "\n");
922 static void qmf_32_subbands(DCAContext *s, int chans,
923 float samples_in[32][8], float *samples_out,
926 const float *prCoeff;
928 int sb_act = s->subband_activity[chans];
930 scale *= sqrt(1 / 8.0);
933 if (!s->multirate_inter) /* Non-perfect reconstruction */
934 prCoeff = fir_32bands_nonperfect;
935 else /* Perfect reconstruction */
936 prCoeff = fir_32bands_perfect;
938 s->dcadsp.qmf_32_subbands(samples_in, sb_act, &s->synth, &s->imdct,
939 s->subband_fir_hist[chans],
940 &s->hist_index[chans],
941 s->subband_fir_noidea[chans], prCoeff,
942 samples_out, s->raXin, scale);
945 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
946 int num_deci_sample, float *samples_in,
947 float *samples_out, float scale)
949 /* samples_in: An array holding decimated samples.
950 * Samples in current subframe starts from samples_in[0],
951 * while samples_in[-1], samples_in[-2], ..., stores samples
952 * from last subframe as history.
954 * samples_out: An array holding interpolated samples
958 const float *prCoeff;
961 /* Select decimation filter */
962 if (decimation_select == 1) {
964 prCoeff = lfe_fir_128;
967 prCoeff = lfe_fir_64;
970 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
971 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
973 samples_out += 2 * decifactor;
977 /* downmixing routines */
978 #define MIX_REAR1(samples, s1, rs, coef) \
979 samples[0][i] += samples[s1][i] * coef[rs][0]; \
980 samples[1][i] += samples[s1][i] * coef[rs][1];
982 #define MIX_REAR2(samples, s1, s2, rs, coef) \
983 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
984 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
986 #define MIX_FRONT3(samples, coef) \
990 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
991 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
993 #define DOWNMIX_TO_STEREO(op1, op2) \
994 for (i = 0; i < 256; i++) { \
999 static void dca_downmix(float **samples, int srcfmt, int lfe_present,
1000 float coef[DCA_PRIM_CHANNELS_MAX + 1][2],
1001 const int8_t *channel_mapping)
1003 int c, l, r, sl, sr, s;
1010 av_log(NULL, 0, "Not implemented!\n");
1014 case DCA_STEREO_TOTAL:
1015 case DCA_STEREO_SUMDIFF:
1018 c = channel_mapping[0];
1019 l = channel_mapping[1];
1020 r = channel_mapping[2];
1021 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1024 s = channel_mapping[2];
1025 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1028 c = channel_mapping[0];
1029 l = channel_mapping[1];
1030 r = channel_mapping[2];
1031 s = channel_mapping[3];
1032 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1033 MIX_REAR1(samples, s, 3, coef));
1036 sl = channel_mapping[2];
1037 sr = channel_mapping[3];
1038 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1041 c = channel_mapping[0];
1042 l = channel_mapping[1];
1043 r = channel_mapping[2];
1044 sl = channel_mapping[3];
1045 sr = channel_mapping[4];
1046 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1047 MIX_REAR2(samples, sl, sr, 3, coef));
1051 int lf_buf = dca_lfe_index[srcfmt];
1052 int lf_idx = dca_channels [srcfmt];
1053 for (i = 0; i < 256; i++) {
1054 samples[0][i] += samples[lf_buf][i] * coef[lf_idx][0];
1055 samples[1][i] += samples[lf_buf][i] * coef[lf_idx][1];
1061 #ifndef decode_blockcodes
1062 /* Very compact version of the block code decoder that does not use table
1063 * look-up but is slightly slower */
1064 static int decode_blockcode(int code, int levels, int32_t *values)
1067 int offset = (levels - 1) >> 1;
1069 for (i = 0; i < 4; i++) {
1070 int div = FASTDIV(code, levels);
1071 values[i] = code - offset - div * levels;
1078 static int decode_blockcodes(int code1, int code2, int levels, int32_t *values)
1080 return decode_blockcode(code1, levels, values) |
1081 decode_blockcode(code2, levels, values + 4);
1085 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1086 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1088 #ifndef int8x8_fmul_int32
1089 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1091 float fscale = scale / 16.0;
1093 for (i = 0; i < 8; i++)
1094 dst[i] = src[i] * fscale;
1098 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1101 int subsubframe = s->current_subsubframe;
1103 const float *quant_step_table;
1106 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1107 LOCAL_ALIGNED_16(int32_t, block, [8 * DCA_SUBBANDS]);
1113 /* Select quantization step size table */
1114 if (s->bit_rate_index == 0x1f)
1115 quant_step_table = lossless_quant_d;
1117 quant_step_table = lossy_quant_d;
1119 for (k = base_channel; k < s->prim_channels; k++) {
1120 float rscale[DCA_SUBBANDS];
1122 if (get_bits_left(&s->gb) < 0)
1123 return AVERROR_INVALIDDATA;
1125 for (l = 0; l < s->vq_start_subband[k]; l++) {
1128 /* Select the mid-tread linear quantizer */
1129 int abits = s->bitalloc[k][l];
1131 float quant_step_size = quant_step_table[abits];
1134 * Determine quantization index code book and its type
1137 /* Select quantization index code book */
1138 int sel = s->quant_index_huffman[k][abits];
1141 * Extract bits from the bit stream
1145 memset(block + 8 * l, 0, 8 * sizeof(block[0]));
1147 /* Deal with transients */
1148 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1149 rscale[l] = quant_step_size * s->scale_factor[k][l][sfi] *
1150 s->scalefactor_adj[k][sel];
1152 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1155 int block_code1, block_code2, size, levels, err;
1157 size = abits_sizes[abits - 1];
1158 levels = abits_levels[abits - 1];
1160 block_code1 = get_bits(&s->gb, size);
1161 block_code2 = get_bits(&s->gb, size);
1162 err = decode_blockcodes(block_code1, block_code2,
1163 levels, block + 8 * l);
1165 av_log(s->avctx, AV_LOG_ERROR,
1166 "ERROR: block code look-up failed\n");
1167 return AVERROR_INVALIDDATA;
1171 for (m = 0; m < 8; m++)
1172 block[8 * l + m] = get_sbits(&s->gb, abits - 3);
1176 for (m = 0; m < 8; m++)
1177 block[8 * l + m] = get_bitalloc(&s->gb,
1178 &dca_smpl_bitalloc[abits], sel);
1184 s->fmt_conv.int32_to_float_fmul_array8(&s->fmt_conv, subband_samples[k][0],
1185 block, rscale, 8 * s->vq_start_subband[k]);
1187 for (l = 0; l < s->vq_start_subband[k]; l++) {
1190 * Inverse ADPCM if in prediction mode
1192 if (s->prediction_mode[k][l]) {
1194 for (m = 0; m < 8; m++) {
1195 for (n = 1; n <= 4; n++)
1197 subband_samples[k][l][m] +=
1198 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1199 subband_samples[k][l][m - n] / 8192);
1200 else if (s->predictor_history)
1201 subband_samples[k][l][m] +=
1202 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1203 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1209 * Decode VQ encoded high frequencies
1211 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1212 /* 1 vector -> 32 samples but we only need the 8 samples
1213 * for this subsubframe. */
1214 int hfvq = s->high_freq_vq[k][l];
1216 if (!s->debug_flag & 0x01) {
1217 av_log(s->avctx, AV_LOG_DEBUG,
1218 "Stream with high frequencies VQ coding\n");
1219 s->debug_flag |= 0x01;
1222 int8x8_fmul_int32(subband_samples[k][l],
1223 &high_freq_vq[hfvq][subsubframe * 8],
1224 s->scale_factor[k][l][0]);
1228 /* Check for DSYNC after subsubframe */
1229 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1230 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1232 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1235 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1236 return AVERROR_INVALIDDATA;
1240 /* Backup predictor history for adpcm */
1241 for (k = base_channel; k < s->prim_channels; k++)
1242 for (l = 0; l < s->vq_start_subband[k]; l++)
1243 memcpy(s->subband_samples_hist[k][l],
1244 &subband_samples[k][l][4],
1245 4 * sizeof(subband_samples[0][0][0]));
1250 static int dca_filter_channels(DCAContext *s, int block_index)
1252 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1255 /* 32 subbands QMF */
1256 for (k = 0; k < s->prim_channels; k++) {
1257 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1258 0, 8388608.0, 8388608.0 };*/
1259 if (s->channel_order_tab[k] >= 0)
1260 qmf_32_subbands(s, k, subband_samples[k],
1261 s->samples_chanptr[s->channel_order_tab[k]],
1262 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1265 /* Generate LFE samples for this subsubframe FIXME!!! */
1267 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1268 s->lfe_data + 2 * s->lfe * (block_index + 4),
1269 s->samples_chanptr[dca_lfe_index[s->amode]],
1270 1.0 / (256.0 * 32768.0));
1271 /* Outputs 20bits pcm samples */
1274 /* Downmixing to Stereo */
1275 if (s->prim_channels + !!s->lfe > 2 &&
1276 s->avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) {
1277 dca_downmix(s->samples_chanptr, s->amode, !!s->lfe, s->downmix_coef,
1278 s->channel_order_tab);
1285 static int dca_subframe_footer(DCAContext *s, int base_channel)
1287 int in, out, aux_data_count, aux_data_end, reserved;
1291 * Unpack optional information
1294 /* presumably optional information only appears in the core? */
1295 if (!base_channel) {
1297 skip_bits_long(&s->gb, 32);
1300 aux_data_count = get_bits(&s->gb, 6);
1303 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1305 aux_data_end = 8 * aux_data_count + get_bits_count(&s->gb);
1307 if ((nsyncaux = get_bits_long(&s->gb, 32)) != DCA_NSYNCAUX) {
1308 av_log(s->avctx, AV_LOG_ERROR, "nSYNCAUX mismatch %#"PRIx32"\n",
1310 return AVERROR_INVALIDDATA;
1313 if (get_bits1(&s->gb)) { // bAUXTimeStampFlag
1314 avpriv_request_sample(s->avctx,
1315 "Auxiliary Decode Time Stamp Flag");
1317 skip_bits(&s->gb, (-get_bits_count(&s->gb)) & 4);
1318 // 44 bits: nMSByte (8), nMarker (4), nLSByte (28), nMarker (4)
1319 skip_bits_long(&s->gb, 44);
1322 if ((s->core_downmix = get_bits1(&s->gb))) {
1323 int am = get_bits(&s->gb, 3);
1326 s->core_downmix_amode = DCA_MONO;
1329 s->core_downmix_amode = DCA_STEREO;
1332 s->core_downmix_amode = DCA_STEREO_TOTAL;
1335 s->core_downmix_amode = DCA_3F;
1338 s->core_downmix_amode = DCA_2F1R;
1341 s->core_downmix_amode = DCA_2F2R;
1344 s->core_downmix_amode = DCA_3F1R;
1347 av_log(s->avctx, AV_LOG_ERROR,
1348 "Invalid mode %d for embedded downmix coefficients\n",
1350 return AVERROR_INVALIDDATA;
1352 for (out = 0; out < dca_channels[s->core_downmix_amode]; out++) {
1353 for (in = 0; in < s->prim_channels + !!s->lfe; in++) {
1354 uint16_t tmp = get_bits(&s->gb, 9);
1355 if ((tmp & 0xFF) > 241) {
1356 av_log(s->avctx, AV_LOG_ERROR,
1357 "Invalid downmix coefficient code %"PRIu16"\n",
1359 return AVERROR_INVALIDDATA;
1361 s->core_downmix_codes[in][out] = tmp;
1366 align_get_bits(&s->gb); // byte align
1367 skip_bits(&s->gb, 16); // nAUXCRC16
1369 // additional data (reserved, cf. ETSI TS 102 114 V1.4.1)
1370 if ((reserved = (aux_data_end - get_bits_count(&s->gb))) < 0) {
1371 av_log(s->avctx, AV_LOG_ERROR,
1372 "Overread auxiliary data by %d bits\n", -reserved);
1373 return AVERROR_INVALIDDATA;
1374 } else if (reserved) {
1375 avpriv_request_sample(s->avctx,
1376 "Core auxiliary data reserved content");
1377 skip_bits_long(&s->gb, reserved);
1381 if (s->crc_present && s->dynrange)
1382 get_bits(&s->gb, 16);
1389 * Decode a dca frame block
1391 * @param s pointer to the DCAContext
1394 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1399 if (s->current_subframe >= s->subframes) {
1400 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1401 s->current_subframe, s->subframes);
1402 return AVERROR_INVALIDDATA;
1405 if (!s->current_subsubframe) {
1407 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1409 /* Read subframe header */
1410 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1414 /* Read subsubframe */
1416 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1418 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1422 s->current_subsubframe++;
1423 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1424 s->current_subsubframe = 0;
1425 s->current_subframe++;
1427 if (s->current_subframe >= s->subframes) {
1429 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1431 /* Read subframe footer */
1432 if ((ret = dca_subframe_footer(s, base_channel)))
1440 * Return the number of channels in an ExSS speaker mask (HD)
1442 static int dca_exss_mask2count(int mask)
1444 /* count bits that mean speaker pairs twice */
1445 return av_popcount(mask) +
1446 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1447 DCA_EXSS_FRONT_LEFT_RIGHT |
1448 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1449 DCA_EXSS_WIDE_LEFT_RIGHT |
1450 DCA_EXSS_SIDE_LEFT_RIGHT |
1451 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1452 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1453 DCA_EXSS_REAR_LEFT_RIGHT |
1454 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1458 * Skip mixing coefficients of a single mix out configuration (HD)
1460 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1464 for (i = 0; i < channels; i++) {
1465 int mix_map_mask = get_bits(gb, out_ch);
1466 int num_coeffs = av_popcount(mix_map_mask);
1467 skip_bits_long(gb, num_coeffs * 6);
1472 * Parse extension substream asset header (HD)
1474 static int dca_exss_parse_asset_header(DCAContext *s)
1476 int header_pos = get_bits_count(&s->gb);
1479 int embedded_stereo = 0;
1480 int embedded_6ch = 0;
1481 int drc_code_present;
1482 int extensions_mask;
1485 if (get_bits_left(&s->gb) < 16)
1488 /* We will parse just enough to get to the extensions bitmask with which
1489 * we can set the profile value. */
1491 header_size = get_bits(&s->gb, 9) + 1;
1492 skip_bits(&s->gb, 3); // asset index
1494 if (s->static_fields) {
1495 if (get_bits1(&s->gb))
1496 skip_bits(&s->gb, 4); // asset type descriptor
1497 if (get_bits1(&s->gb))
1498 skip_bits_long(&s->gb, 24); // language descriptor
1500 if (get_bits1(&s->gb)) {
1501 /* How can one fit 1024 bytes of text here if the maximum value
1502 * for the asset header size field above was 512 bytes? */
1503 int text_length = get_bits(&s->gb, 10) + 1;
1504 if (get_bits_left(&s->gb) < text_length * 8)
1506 skip_bits_long(&s->gb, text_length * 8); // info text
1509 skip_bits(&s->gb, 5); // bit resolution - 1
1510 skip_bits(&s->gb, 4); // max sample rate code
1511 channels = get_bits(&s->gb, 8) + 1;
1513 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1514 int spkr_remap_sets;
1515 int spkr_mask_size = 16;
1519 embedded_stereo = get_bits1(&s->gb);
1521 embedded_6ch = get_bits1(&s->gb);
1523 if (get_bits1(&s->gb)) {
1524 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1525 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1528 spkr_remap_sets = get_bits(&s->gb, 3);
1530 for (i = 0; i < spkr_remap_sets; i++) {
1531 /* std layout mask for each remap set */
1532 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1535 for (i = 0; i < spkr_remap_sets; i++) {
1536 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1537 if (get_bits_left(&s->gb) < 0)
1540 for (j = 0; j < num_spkrs[i]; j++) {
1541 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1542 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1543 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1548 skip_bits(&s->gb, 3); // representation type
1552 drc_code_present = get_bits1(&s->gb);
1553 if (drc_code_present)
1554 get_bits(&s->gb, 8); // drc code
1556 if (get_bits1(&s->gb))
1557 skip_bits(&s->gb, 5); // dialog normalization code
1559 if (drc_code_present && embedded_stereo)
1560 get_bits(&s->gb, 8); // drc stereo code
1562 if (s->mix_metadata && get_bits1(&s->gb)) {
1563 skip_bits(&s->gb, 1); // external mix
1564 skip_bits(&s->gb, 6); // post mix gain code
1566 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1567 skip_bits(&s->gb, 3); // drc limit
1569 skip_bits(&s->gb, 8); // custom drc code
1571 if (get_bits1(&s->gb)) // channel specific scaling
1572 for (i = 0; i < s->num_mix_configs; i++)
1573 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1575 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1577 for (i = 0; i < s->num_mix_configs; i++) {
1578 if (get_bits_left(&s->gb) < 0)
1580 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1582 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1583 if (embedded_stereo)
1584 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1588 switch (get_bits(&s->gb, 2)) {
1589 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1590 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1591 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1592 case 3: extensions_mask = 0; /* aux coding */ break;
1595 /* not parsed further, we were only interested in the extensions mask */
1597 if (get_bits_left(&s->gb) < 0)
1600 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1601 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1604 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1606 if (extensions_mask & DCA_EXT_EXSS_XLL)
1607 s->profile = FF_PROFILE_DTS_HD_MA;
1608 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1610 s->profile = FF_PROFILE_DTS_HD_HRA;
1612 if (!(extensions_mask & DCA_EXT_CORE))
1613 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1614 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1615 av_log(s->avctx, AV_LOG_WARNING,
1616 "DTS extensions detection mismatch (%d, %d)\n",
1617 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1623 * Parse extension substream header (HD)
1625 static void dca_exss_parse_header(DCAContext *s)
1631 int active_ss_mask[8];
1634 if (get_bits_left(&s->gb) < 52)
1637 skip_bits(&s->gb, 8); // user data
1638 ss_index = get_bits(&s->gb, 2);
1640 blownup = get_bits1(&s->gb);
1641 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1642 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1644 s->static_fields = get_bits1(&s->gb);
1645 if (s->static_fields) {
1646 skip_bits(&s->gb, 2); // reference clock code
1647 skip_bits(&s->gb, 3); // frame duration code
1649 if (get_bits1(&s->gb))
1650 skip_bits_long(&s->gb, 36); // timestamp
1652 /* a single stream can contain multiple audio assets that can be
1653 * combined to form multiple audio presentations */
1655 num_audiop = get_bits(&s->gb, 3) + 1;
1656 if (num_audiop > 1) {
1657 avpriv_request_sample(s->avctx,
1658 "Multiple DTS-HD audio presentations");
1659 /* ignore such streams for now */
1663 num_assets = get_bits(&s->gb, 3) + 1;
1664 if (num_assets > 1) {
1665 avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
1666 /* ignore such streams for now */
1670 for (i = 0; i < num_audiop; i++)
1671 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1673 for (i = 0; i < num_audiop; i++)
1674 for (j = 0; j <= ss_index; j++)
1675 if (active_ss_mask[i] & (1 << j))
1676 skip_bits(&s->gb, 8); // active asset mask
1678 s->mix_metadata = get_bits1(&s->gb);
1679 if (s->mix_metadata) {
1680 int mix_out_mask_size;
1682 skip_bits(&s->gb, 2); // adjustment level
1683 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1684 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1686 for (i = 0; i < s->num_mix_configs; i++) {
1687 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1688 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1693 for (i = 0; i < num_assets; i++)
1694 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1696 for (i = 0; i < num_assets; i++) {
1697 if (dca_exss_parse_asset_header(s))
1701 /* not parsed further, we were only interested in the extensions mask
1702 * from the asset header */
1706 * Main frame decoding function
1707 * FIXME add arguments
1709 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1710 int *got_frame_ptr, AVPacket *avpkt)
1712 AVFrame *frame = data;
1713 const uint8_t *buf = avpkt->data;
1714 int buf_size = avpkt->size;
1717 int num_core_channels = 0;
1719 float **samples_flt;
1720 DCAContext *s = avctx->priv_data;
1721 int channels, full_channels;
1727 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1728 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1729 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1730 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1731 return AVERROR_INVALIDDATA;
1734 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1735 if ((ret = dca_parse_frame_header(s)) < 0) {
1736 //seems like the frame is corrupt, try with the next one
1739 //set AVCodec values with parsed data
1740 avctx->sample_rate = s->sample_rate;
1741 avctx->bit_rate = s->bit_rate;
1743 s->profile = FF_PROFILE_DTS;
1745 for (i = 0; i < (s->sample_blocks / 8); i++) {
1746 if ((ret = dca_decode_block(s, 0, i))) {
1747 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1752 /* record number of core channels incase less than max channels are requested */
1753 num_core_channels = s->prim_channels;
1756 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1758 s->core_ext_mask = 0;
1760 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1762 /* only scan for extensions if ext_descr was unknown or indicated a
1763 * supported XCh extension */
1764 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1766 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1767 * extensions scan can fill it up */
1768 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1770 /* extensions start at 32-bit boundaries into bitstream */
1771 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1773 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1774 uint32_t bits = get_bits_long(&s->gb, 32);
1778 int ext_amode, xch_fsize;
1780 s->xch_base_channel = s->prim_channels;
1782 /* validate sync word using XCHFSIZE field */
1783 xch_fsize = show_bits(&s->gb, 10);
1784 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1785 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1788 /* skip length-to-end-of-frame field for the moment */
1789 skip_bits(&s->gb, 10);
1791 s->core_ext_mask |= DCA_EXT_XCH;
1793 /* extension amode(number of channels in extension) should be 1 */
1794 /* AFAIK XCh is not used for more channels */
1795 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1796 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1797 " supported!\n", ext_amode);
1801 /* much like core primary audio coding header */
1802 dca_parse_audio_coding_header(s, s->xch_base_channel);
1804 for (i = 0; i < (s->sample_blocks / 8); i++)
1805 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1806 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1814 /* XXCh: extended channels */
1815 /* usually found either in core or HD part in DTS-HD HRA streams,
1816 * but not in DTS-ES which contains XCh extensions instead */
1817 s->core_ext_mask |= DCA_EXT_XXCH;
1821 int fsize96 = show_bits(&s->gb, 12) + 1;
1822 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1825 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1826 get_bits_count(&s->gb));
1827 skip_bits(&s->gb, 12);
1828 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1829 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1831 s->core_ext_mask |= DCA_EXT_X96;
1836 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1839 /* no supported extensions, skip the rest of the core substream */
1840 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1843 if (s->core_ext_mask & DCA_EXT_X96)
1844 s->profile = FF_PROFILE_DTS_96_24;
1845 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1846 s->profile = FF_PROFILE_DTS_ES;
1848 /* check for ExSS (HD part) */
1849 if (s->dca_buffer_size - s->frame_size > 32 &&
1850 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1851 dca_exss_parse_header(s);
1853 avctx->profile = s->profile;
1855 full_channels = channels = s->prim_channels + !!s->lfe;
1857 if (s->amode < 16) {
1858 avctx->channel_layout = dca_core_channel_layout[s->amode];
1860 #if FF_API_REQUEST_CHANNELS
1861 FF_DISABLE_DEPRECATION_WARNINGS
1862 if (s->xch_present && !s->xch_disable &&
1863 (!avctx->request_channels ||
1864 avctx->request_channels > num_core_channels + !!s->lfe)) {
1865 FF_ENABLE_DEPRECATION_WARNINGS
1867 if (s->xch_present && !s->xch_disable) {
1869 avctx->channel_layout |= AV_CH_BACK_CENTER;
1871 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1872 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1874 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1877 channels = num_core_channels + !!s->lfe;
1878 s->xch_present = 0; /* disable further xch processing */
1880 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1881 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1883 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1886 if (channels > !!s->lfe &&
1887 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1888 return AVERROR_INVALIDDATA;
1890 if (s->prim_channels + !!s->lfe > 2 &&
1891 avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) {
1893 s->output = DCA_STEREO;
1894 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1896 /* Stereo downmix coefficients
1898 * The decoder can only downmix to 2-channel, so we need to ensure
1899 * embedded downmix coefficients are actually targeting 2-channel.
1901 if (s->core_downmix && (s->core_downmix_amode == DCA_STEREO ||
1902 s->core_downmix_amode == DCA_STEREO_TOTAL)) {
1904 for (i = 0; i < s->prim_channels + !!s->lfe; i++) {
1905 sign = s->core_downmix_codes[i][0] & 0x100 ? 1 : -1;
1906 code = s->core_downmix_codes[i][0] & 0x0FF;
1907 s->downmix_coef[i][0] = (!code ? 0.0f :
1908 sign * dca_dmixtable[code - 1]);
1909 sign = s->core_downmix_codes[i][1] & 0x100 ? 1 : -1;
1910 code = s->core_downmix_codes[i][1] & 0x0FF;
1911 s->downmix_coef[i][1] = (!code ? 0.0f :
1912 sign * dca_dmixtable[code - 1]);
1915 int am = s->amode & DCA_CHANNEL_MASK;
1916 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
1917 av_log(s->avctx, AV_LOG_ERROR,
1918 "Invalid channel mode %d\n", am);
1919 return AVERROR_INVALIDDATA;
1921 if (s->prim_channels + !!s->lfe >
1922 FF_ARRAY_ELEMS(dca_default_coeffs[0])) {
1923 avpriv_request_sample(s->avctx, "Downmixing %d channels",
1924 s->prim_channels + !!s->lfe);
1925 return AVERROR_PATCHWELCOME;
1927 for (i = 0; i < s->prim_channels + !!s->lfe; i++) {
1928 s->downmix_coef[i][0] = dca_default_coeffs[am][i][0];
1929 s->downmix_coef[i][1] = dca_default_coeffs[am][i][1];
1932 av_dlog(s->avctx, "Stereo downmix coeffs:\n");
1933 for (i = 0; i < s->prim_channels + !!s->lfe; i++) {
1934 av_dlog(s->avctx, "L, input channel %d = %f\n", i,
1935 s->downmix_coef[i][0]);
1936 av_dlog(s->avctx, "R, input channel %d = %f\n", i,
1937 s->downmix_coef[i][1]);
1939 av_dlog(s->avctx, "\n");
1942 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1943 return AVERROR_INVALIDDATA;
1945 avctx->channels = channels;
1947 /* get output buffer */
1948 frame->nb_samples = 256 * (s->sample_blocks / 8);
1949 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
1950 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1953 samples_flt = (float **)frame->extended_data;
1955 /* allocate buffer for extra channels if downmixing */
1956 if (avctx->channels < full_channels) {
1957 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
1959 avctx->sample_fmt, 0);
1963 av_fast_malloc(&s->extra_channels_buffer,
1964 &s->extra_channels_buffer_size, ret);
1965 if (!s->extra_channels_buffer)
1966 return AVERROR(ENOMEM);
1968 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
1969 s->extra_channels_buffer,
1970 full_channels - channels,
1971 frame->nb_samples, avctx->sample_fmt, 0);
1976 /* filter to get final output */
1977 for (i = 0; i < (s->sample_blocks / 8); i++) {
1980 for (ch = 0; ch < channels; ch++)
1981 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
1982 for (; ch < full_channels; ch++)
1983 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
1985 dca_filter_channels(s, i);
1987 /* If this was marked as a DTS-ES stream we need to subtract back- */
1988 /* channel from SL & SR to remove matrixed back-channel signal */
1989 if ((s->source_pcm_res & 1) && s->xch_present) {
1990 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
1991 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
1992 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
1993 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1994 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1998 /* update lfe history */
1999 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
2000 for (i = 0; i < 2 * s->lfe * 4; i++)
2001 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
2011 * DCA initialization
2013 * @param avctx pointer to the AVCodecContext
2016 static av_cold int dca_decode_init(AVCodecContext *avctx)
2018 DCAContext *s = avctx->priv_data;
2023 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
2024 ff_mdct_init(&s->imdct, 6, 1, 1.0);
2025 ff_synth_filter_init(&s->synth);
2026 ff_dcadsp_init(&s->dcadsp);
2027 ff_fmt_convert_init(&s->fmt_conv, avctx);
2029 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
2031 /* allow downmixing to stereo */
2032 #if FF_API_REQUEST_CHANNELS
2033 FF_DISABLE_DEPRECATION_WARNINGS
2034 if (avctx->request_channels == 2)
2035 avctx->request_channel_layout = AV_CH_LAYOUT_STEREO;
2036 FF_ENABLE_DEPRECATION_WARNINGS
2038 if (avctx->channels > 2 &&
2039 avctx->request_channel_layout == AV_CH_LAYOUT_STEREO)
2040 avctx->channels = 2;
2045 static av_cold int dca_decode_end(AVCodecContext *avctx)
2047 DCAContext *s = avctx->priv_data;
2048 ff_mdct_end(&s->imdct);
2049 av_freep(&s->extra_channels_buffer);
2053 static const AVProfile profiles[] = {
2054 { FF_PROFILE_DTS, "DTS" },
2055 { FF_PROFILE_DTS_ES, "DTS-ES" },
2056 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2057 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2058 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2059 { FF_PROFILE_UNKNOWN },
2062 static const AVOption options[] = {
2063 { "disable_xch", "disable decoding of the XCh extension", offsetof(DCAContext, xch_disable), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_DECODING_PARAM|AV_OPT_FLAG_AUDIO_PARAM },
2067 static const AVClass dca_decoder_class = {
2068 .class_name = "DCA decoder",
2069 .item_name = av_default_item_name,
2071 .version = LIBAVUTIL_VERSION_INT,
2074 AVCodec ff_dca_decoder = {
2076 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2077 .type = AVMEDIA_TYPE_AUDIO,
2078 .id = AV_CODEC_ID_DTS,
2079 .priv_data_size = sizeof(DCAContext),
2080 .init = dca_decode_init,
2081 .decode = dca_decode_frame,
2082 .close = dca_decode_end,
2083 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2084 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
2085 AV_SAMPLE_FMT_NONE },
2086 .profiles = NULL_IF_CONFIG_SMALL(profiles),
2087 .priv_class = &dca_decoder_class,