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 /** Bit allocation */
268 int offset; ///< code values offset
269 int maxbits[8]; ///< max bits in VLC
270 int wrap; ///< wrap for get_vlc2()
271 VLC vlc[8]; ///< actual codes
274 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
275 static BitAlloc dca_tmode; ///< transition mode VLCs
276 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
277 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
279 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
282 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
287 AVClass *class; ///< class for AVOptions
288 AVCodecContext *avctx;
290 int frame_type; ///< type of the current frame
291 int samples_deficit; ///< deficit sample count
292 int crc_present; ///< crc is present in the bitstream
293 int sample_blocks; ///< number of PCM sample blocks
294 int frame_size; ///< primary frame byte size
295 int amode; ///< audio channels arrangement
296 int sample_rate; ///< audio sampling rate
297 int bit_rate; ///< transmission bit rate
298 int bit_rate_index; ///< transmission bit rate index
300 int dynrange; ///< embedded dynamic range flag
301 int timestamp; ///< embedded time stamp flag
302 int aux_data; ///< auxiliary data flag
303 int hdcd; ///< source material is mastered in HDCD
304 int ext_descr; ///< extension audio descriptor flag
305 int ext_coding; ///< extended coding flag
306 int aspf; ///< audio sync word insertion flag
307 int lfe; ///< low frequency effects flag
308 int predictor_history; ///< predictor history flag
309 int header_crc; ///< header crc check bytes
310 int multirate_inter; ///< multirate interpolator switch
311 int version; ///< encoder software revision
312 int copy_history; ///< copy history
313 int source_pcm_res; ///< source pcm resolution
314 int front_sum; ///< front sum/difference flag
315 int surround_sum; ///< surround sum/difference flag
316 int dialog_norm; ///< dialog normalisation parameter
318 /* Primary audio coding header */
319 int subframes; ///< number of subframes
320 int total_channels; ///< number of channels including extensions
321 int prim_channels; ///< number of primary audio channels
322 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
323 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
324 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
325 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
326 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
327 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
328 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
329 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
331 /* Primary audio coding side information */
332 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
333 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
334 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
335 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
336 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
337 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
338 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
339 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
340 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
341 float downmix_coef[DCA_PRIM_CHANNELS_MAX + 1][2]; ///< stereo downmix coefficients
342 int dynrange_coef; ///< dynamic range coefficient
344 /* Core substream's embedded downmix coefficients (cf. ETSI TS 102 114 V1.4.1)
345 * Input: primary audio channels (incl. LFE if present)
346 * Output: downmix audio channels (up to 4, no LFE) */
347 uint8_t core_downmix; ///< embedded downmix coefficients available
348 uint8_t core_downmix_amode; ///< audio channel arrangement of embedded downmix
349 uint16_t core_downmix_codes[DCA_PRIM_CHANNELS_MAX + 1][4]; ///< embedded downmix coefficients (9-bit codes)
351 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
353 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
354 int lfe_scale_factor;
356 /* Subband samples history (for ADPCM) */
357 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
358 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
359 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
360 int hist_index[DCA_PRIM_CHANNELS_MAX];
361 DECLARE_ALIGNED(32, float, raXin)[32];
363 int output; ///< type of output
365 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
366 float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
367 float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
368 uint8_t *extra_channels_buffer;
369 unsigned int extra_channels_buffer_size;
371 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
372 int dca_buffer_size; ///< how much data is in the dca_buffer
374 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
376 /* Current position in DCA frame */
377 int current_subframe;
378 int current_subsubframe;
380 int core_ext_mask; ///< present extensions in the core substream
382 /* XCh extension information */
383 int xch_present; ///< XCh extension present and valid
384 int xch_base_channel; ///< index of first (only) channel containing XCH data
385 int xch_disable; ///< whether the XCh extension should be decoded or not
387 /* ExSS header parser */
388 int static_fields; ///< static fields present
389 int mix_metadata; ///< mixing metadata present
390 int num_mix_configs; ///< number of mix out configurations
391 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
395 int debug_flag; ///< used for suppressing repeated error messages output
396 AVFloatDSPContext fdsp;
398 SynthFilterContext synth;
399 DCADSPContext dcadsp;
400 FmtConvertContext fmt_conv;
403 static const uint16_t dca_vlc_offs[] = {
404 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
405 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
406 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
407 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
408 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
409 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
412 static av_cold void dca_init_vlcs(void)
414 static int vlcs_initialized = 0;
416 static VLC_TYPE dca_table[23622][2];
418 if (vlcs_initialized)
421 dca_bitalloc_index.offset = 1;
422 dca_bitalloc_index.wrap = 2;
423 for (i = 0; i < 5; i++) {
424 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
425 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
426 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
427 bitalloc_12_bits[i], 1, 1,
428 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
430 dca_scalefactor.offset = -64;
431 dca_scalefactor.wrap = 2;
432 for (i = 0; i < 5; i++) {
433 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
434 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
435 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
436 scales_bits[i], 1, 1,
437 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
439 dca_tmode.offset = 0;
441 for (i = 0; i < 4; i++) {
442 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
443 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
444 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
446 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
449 for (i = 0; i < 10; i++)
450 for (j = 0; j < 7; j++) {
451 if (!bitalloc_codes[i][j])
453 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
454 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
455 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
456 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
458 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
460 bitalloc_bits[i][j], 1, 1,
461 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
464 vlcs_initialized = 1;
467 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
470 *dst++ = get_bits(gb, bits);
473 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
476 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
477 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
478 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
480 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
481 s->prim_channels = s->total_channels;
483 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
484 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
487 for (i = base_channel; i < s->prim_channels; i++) {
488 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
489 if (s->subband_activity[i] > DCA_SUBBANDS)
490 s->subband_activity[i] = DCA_SUBBANDS;
492 for (i = base_channel; i < s->prim_channels; i++) {
493 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
494 if (s->vq_start_subband[i] > DCA_SUBBANDS)
495 s->vq_start_subband[i] = DCA_SUBBANDS;
497 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
498 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
499 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
500 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
502 /* Get codebooks quantization indexes */
504 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
505 for (j = 1; j < 11; j++)
506 for (i = base_channel; i < s->prim_channels; i++)
507 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
509 /* Get scale factor adjustment */
510 for (j = 0; j < 11; j++)
511 for (i = base_channel; i < s->prim_channels; i++)
512 s->scalefactor_adj[i][j] = 1;
514 for (j = 1; j < 11; j++)
515 for (i = base_channel; i < s->prim_channels; i++)
516 if (s->quant_index_huffman[i][j] < thr[j])
517 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
519 if (s->crc_present) {
520 /* Audio header CRC check */
521 get_bits(&s->gb, 16);
524 s->current_subframe = 0;
525 s->current_subsubframe = 0;
528 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
529 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
530 for (i = base_channel; i < s->prim_channels; i++) {
531 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
532 s->subband_activity[i]);
533 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
534 s->vq_start_subband[i]);
535 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
536 s->joint_intensity[i]);
537 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
538 s->transient_huffman[i]);
539 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
540 s->scalefactor_huffman[i]);
541 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
542 s->bitalloc_huffman[i]);
543 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
544 for (j = 0; j < 11; j++)
545 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
546 av_log(s->avctx, AV_LOG_DEBUG, "\n");
547 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
548 for (j = 0; j < 11; j++)
549 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
550 av_log(s->avctx, AV_LOG_DEBUG, "\n");
557 static int dca_parse_frame_header(DCAContext *s)
559 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
562 skip_bits_long(&s->gb, 32);
565 s->frame_type = get_bits(&s->gb, 1);
566 s->samples_deficit = get_bits(&s->gb, 5) + 1;
567 s->crc_present = get_bits(&s->gb, 1);
568 s->sample_blocks = get_bits(&s->gb, 7) + 1;
569 s->frame_size = get_bits(&s->gb, 14) + 1;
570 if (s->frame_size < 95)
571 return AVERROR_INVALIDDATA;
572 s->amode = get_bits(&s->gb, 6);
573 s->sample_rate = avpriv_dca_sample_rates[get_bits(&s->gb, 4)];
575 return AVERROR_INVALIDDATA;
576 s->bit_rate_index = get_bits(&s->gb, 5);
577 s->bit_rate = dca_bit_rates[s->bit_rate_index];
579 return AVERROR_INVALIDDATA;
581 skip_bits1(&s->gb); // always 0 (reserved, cf. ETSI TS 102 114 V1.4.1)
582 s->dynrange = get_bits(&s->gb, 1);
583 s->timestamp = get_bits(&s->gb, 1);
584 s->aux_data = get_bits(&s->gb, 1);
585 s->hdcd = get_bits(&s->gb, 1);
586 s->ext_descr = get_bits(&s->gb, 3);
587 s->ext_coding = get_bits(&s->gb, 1);
588 s->aspf = get_bits(&s->gb, 1);
589 s->lfe = get_bits(&s->gb, 2);
590 s->predictor_history = get_bits(&s->gb, 1);
593 av_log(s->avctx, AV_LOG_ERROR, "Invalid LFE value: %d\n", s->lfe);
594 return AVERROR_INVALIDDATA;
597 /* TODO: check CRC */
599 s->header_crc = get_bits(&s->gb, 16);
601 s->multirate_inter = get_bits(&s->gb, 1);
602 s->version = get_bits(&s->gb, 4);
603 s->copy_history = get_bits(&s->gb, 2);
604 s->source_pcm_res = get_bits(&s->gb, 3);
605 s->front_sum = get_bits(&s->gb, 1);
606 s->surround_sum = get_bits(&s->gb, 1);
607 s->dialog_norm = get_bits(&s->gb, 4);
609 /* FIXME: channels mixing levels */
610 s->output = s->amode;
612 s->output |= DCA_LFE;
615 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
616 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
617 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
618 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
619 s->sample_blocks, s->sample_blocks * 32);
620 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
621 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
622 s->amode, dca_channels[s->amode]);
623 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
625 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
627 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
628 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
629 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
630 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
631 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
632 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
633 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
634 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
635 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
636 s->predictor_history);
637 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
638 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
640 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
641 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
642 av_log(s->avctx, AV_LOG_DEBUG,
643 "source pcm resolution: %i (%i bits/sample)\n",
644 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
645 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
646 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
647 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
648 av_log(s->avctx, AV_LOG_DEBUG, "\n");
651 /* Primary audio coding header */
652 s->subframes = get_bits(&s->gb, 4) + 1;
654 return dca_parse_audio_coding_header(s, 0);
658 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
661 /* huffman encoded */
662 value += get_bitalloc(gb, &dca_scalefactor, level);
663 value = av_clip(value, 0, (1 << log2range) - 1);
664 } else if (level < 8) {
665 if (level + 1 > log2range) {
666 skip_bits(gb, level + 1 - log2range);
667 value = get_bits(gb, log2range);
669 value = get_bits(gb, level + 1);
675 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
677 /* Primary audio coding side information */
680 if (get_bits_left(&s->gb) < 0)
681 return AVERROR_INVALIDDATA;
684 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
685 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
688 for (j = base_channel; j < s->prim_channels; j++) {
689 for (k = 0; k < s->subband_activity[j]; k++)
690 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
693 /* Get prediction codebook */
694 for (j = base_channel; j < s->prim_channels; j++) {
695 for (k = 0; k < s->subband_activity[j]; k++) {
696 if (s->prediction_mode[j][k] > 0) {
697 /* (Prediction coefficient VQ address) */
698 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
703 /* Bit allocation index */
704 for (j = base_channel; j < s->prim_channels; j++) {
705 for (k = 0; k < s->vq_start_subband[j]; k++) {
706 if (s->bitalloc_huffman[j] == 6)
707 s->bitalloc[j][k] = get_bits(&s->gb, 5);
708 else if (s->bitalloc_huffman[j] == 5)
709 s->bitalloc[j][k] = get_bits(&s->gb, 4);
710 else if (s->bitalloc_huffman[j] == 7) {
711 av_log(s->avctx, AV_LOG_ERROR,
712 "Invalid bit allocation index\n");
713 return AVERROR_INVALIDDATA;
716 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
719 if (s->bitalloc[j][k] > 26) {
720 av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
721 j, k, s->bitalloc[j][k]);
722 return AVERROR_INVALIDDATA;
727 /* Transition mode */
728 for (j = base_channel; j < s->prim_channels; j++) {
729 for (k = 0; k < s->subband_activity[j]; k++) {
730 s->transition_mode[j][k] = 0;
731 if (s->subsubframes[s->current_subframe] > 1 &&
732 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
733 s->transition_mode[j][k] =
734 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
739 if (get_bits_left(&s->gb) < 0)
740 return AVERROR_INVALIDDATA;
742 for (j = base_channel; j < s->prim_channels; j++) {
743 const uint32_t *scale_table;
744 int scale_sum, log_size;
746 memset(s->scale_factor[j], 0,
747 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
749 if (s->scalefactor_huffman[j] == 6) {
750 scale_table = scale_factor_quant7;
753 scale_table = scale_factor_quant6;
757 /* When huffman coded, only the difference is encoded */
760 for (k = 0; k < s->subband_activity[j]; k++) {
761 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
762 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
763 s->scale_factor[j][k][0] = scale_table[scale_sum];
766 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
767 /* Get second scale factor */
768 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
769 s->scale_factor[j][k][1] = scale_table[scale_sum];
774 /* Joint subband scale factor codebook select */
775 for (j = base_channel; j < s->prim_channels; j++) {
776 /* Transmitted only if joint subband coding enabled */
777 if (s->joint_intensity[j] > 0)
778 s->joint_huff[j] = get_bits(&s->gb, 3);
781 if (get_bits_left(&s->gb) < 0)
782 return AVERROR_INVALIDDATA;
784 /* Scale factors for joint subband coding */
785 for (j = base_channel; j < s->prim_channels; j++) {
788 /* Transmitted only if joint subband coding enabled */
789 if (s->joint_intensity[j] > 0) {
791 source_channel = s->joint_intensity[j] - 1;
793 /* When huffman coded, only the difference is encoded
794 * (is this valid as well for joint scales ???) */
796 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
797 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
798 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
801 if (!(s->debug_flag & 0x02)) {
802 av_log(s->avctx, AV_LOG_DEBUG,
803 "Joint stereo coding not supported\n");
804 s->debug_flag |= 0x02;
809 /* Dynamic range coefficient */
810 if (!base_channel && s->dynrange)
811 s->dynrange_coef = get_bits(&s->gb, 8);
813 /* Side information CRC check word */
814 if (s->crc_present) {
815 get_bits(&s->gb, 16);
819 * Primary audio data arrays
822 /* VQ encoded high frequency subbands */
823 for (j = base_channel; j < s->prim_channels; j++)
824 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
825 /* 1 vector -> 32 samples */
826 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
828 /* Low frequency effect data */
829 if (!base_channel && s->lfe) {
831 int lfe_samples = 2 * s->lfe * (4 + block_index);
832 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
835 for (j = lfe_samples; j < lfe_end_sample; j++) {
836 /* Signed 8 bits int */
837 s->lfe_data[j] = get_sbits(&s->gb, 8);
840 /* Scale factor index */
841 skip_bits(&s->gb, 1);
842 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 7)];
844 /* Quantization step size * scale factor */
845 lfe_scale = 0.035 * s->lfe_scale_factor;
847 for (j = lfe_samples; j < lfe_end_sample; j++)
848 s->lfe_data[j] *= lfe_scale;
852 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
853 s->subsubframes[s->current_subframe]);
854 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
855 s->partial_samples[s->current_subframe]);
857 for (j = base_channel; j < s->prim_channels; j++) {
858 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
859 for (k = 0; k < s->subband_activity[j]; k++)
860 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
861 av_log(s->avctx, AV_LOG_DEBUG, "\n");
863 for (j = base_channel; j < s->prim_channels; j++) {
864 for (k = 0; k < s->subband_activity[j]; k++)
865 av_log(s->avctx, AV_LOG_DEBUG,
866 "prediction coefs: %f, %f, %f, %f\n",
867 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
868 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
869 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
870 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
872 for (j = base_channel; j < s->prim_channels; j++) {
873 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
874 for (k = 0; k < s->vq_start_subband[j]; k++)
875 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
876 av_log(s->avctx, AV_LOG_DEBUG, "\n");
878 for (j = base_channel; j < s->prim_channels; j++) {
879 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
880 for (k = 0; k < s->subband_activity[j]; k++)
881 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
882 av_log(s->avctx, AV_LOG_DEBUG, "\n");
884 for (j = base_channel; j < s->prim_channels; j++) {
885 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
886 for (k = 0; k < s->subband_activity[j]; k++) {
887 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
888 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
889 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
890 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
892 av_log(s->avctx, AV_LOG_DEBUG, "\n");
894 for (j = base_channel; j < s->prim_channels; j++) {
895 if (s->joint_intensity[j] > 0) {
896 int source_channel = s->joint_intensity[j] - 1;
897 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
898 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
899 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
900 av_log(s->avctx, AV_LOG_DEBUG, "\n");
903 for (j = base_channel; j < s->prim_channels; j++)
904 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
905 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
906 if (!base_channel && s->lfe) {
907 int lfe_samples = 2 * s->lfe * (4 + block_index);
908 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
910 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
911 for (j = lfe_samples; j < lfe_end_sample; j++)
912 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
913 av_log(s->avctx, AV_LOG_DEBUG, "\n");
920 static void qmf_32_subbands(DCAContext *s, int chans,
921 float samples_in[32][8], float *samples_out,
924 const float *prCoeff;
926 int sb_act = s->subband_activity[chans];
928 scale *= sqrt(1 / 8.0);
931 if (!s->multirate_inter) /* Non-perfect reconstruction */
932 prCoeff = fir_32bands_nonperfect;
933 else /* Perfect reconstruction */
934 prCoeff = fir_32bands_perfect;
936 s->dcadsp.qmf_32_subbands(samples_in, sb_act, &s->synth, &s->imdct,
937 s->subband_fir_hist[chans],
938 &s->hist_index[chans],
939 s->subband_fir_noidea[chans], prCoeff,
940 samples_out, s->raXin, scale);
943 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
944 int num_deci_sample, float *samples_in,
945 float *samples_out, float scale)
947 /* samples_in: An array holding decimated samples.
948 * Samples in current subframe starts from samples_in[0],
949 * while samples_in[-1], samples_in[-2], ..., stores samples
950 * from last subframe as history.
952 * samples_out: An array holding interpolated samples
956 const float *prCoeff;
959 /* Select decimation filter */
960 if (decimation_select == 1) {
962 prCoeff = lfe_fir_128;
965 prCoeff = lfe_fir_64;
968 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
969 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
971 samples_out += 2 * decifactor;
975 /* downmixing routines */
976 #define MIX_REAR1(samples, s1, rs, coef) \
977 samples[0][i] += samples[s1][i] * coef[rs][0]; \
978 samples[1][i] += samples[s1][i] * coef[rs][1];
980 #define MIX_REAR2(samples, s1, s2, rs, coef) \
981 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
982 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
984 #define MIX_FRONT3(samples, coef) \
988 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
989 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
991 #define DOWNMIX_TO_STEREO(op1, op2) \
992 for (i = 0; i < 256; i++) { \
997 static void dca_downmix(float **samples, int srcfmt, int lfe_present,
998 float coef[DCA_PRIM_CHANNELS_MAX + 1][2],
999 const int8_t *channel_mapping)
1001 int c, l, r, sl, sr, s;
1008 av_log(NULL, 0, "Not implemented!\n");
1012 case DCA_STEREO_TOTAL:
1013 case DCA_STEREO_SUMDIFF:
1016 c = channel_mapping[0];
1017 l = channel_mapping[1];
1018 r = channel_mapping[2];
1019 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1022 s = channel_mapping[2];
1023 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1026 c = channel_mapping[0];
1027 l = channel_mapping[1];
1028 r = channel_mapping[2];
1029 s = channel_mapping[3];
1030 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1031 MIX_REAR1(samples, s, 3, coef));
1034 sl = channel_mapping[2];
1035 sr = channel_mapping[3];
1036 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1039 c = channel_mapping[0];
1040 l = channel_mapping[1];
1041 r = channel_mapping[2];
1042 sl = channel_mapping[3];
1043 sr = channel_mapping[4];
1044 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1045 MIX_REAR2(samples, sl, sr, 3, coef));
1049 int lf_buf = dca_lfe_index[srcfmt];
1050 int lf_idx = dca_channels [srcfmt];
1051 for (i = 0; i < 256; i++) {
1052 samples[0][i] += samples[lf_buf][i] * coef[lf_idx][0];
1053 samples[1][i] += samples[lf_buf][i] * coef[lf_idx][1];
1059 #ifndef decode_blockcodes
1060 /* Very compact version of the block code decoder that does not use table
1061 * look-up but is slightly slower */
1062 static int decode_blockcode(int code, int levels, int32_t *values)
1065 int offset = (levels - 1) >> 1;
1067 for (i = 0; i < 4; i++) {
1068 int div = FASTDIV(code, levels);
1069 values[i] = code - offset - div * levels;
1076 static int decode_blockcodes(int code1, int code2, int levels, int32_t *values)
1078 return decode_blockcode(code1, levels, values) |
1079 decode_blockcode(code2, levels, values + 4);
1083 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1084 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1086 #ifndef int8x8_fmul_int32
1087 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1089 float fscale = scale / 16.0;
1091 for (i = 0; i < 8; i++)
1092 dst[i] = src[i] * fscale;
1096 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1099 int subsubframe = s->current_subsubframe;
1101 const float *quant_step_table;
1104 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1105 LOCAL_ALIGNED_16(int32_t, block, [8 * DCA_SUBBANDS]);
1111 /* Select quantization step size table */
1112 if (s->bit_rate_index == 0x1f)
1113 quant_step_table = lossless_quant_d;
1115 quant_step_table = lossy_quant_d;
1117 for (k = base_channel; k < s->prim_channels; k++) {
1118 float rscale[DCA_SUBBANDS];
1120 if (get_bits_left(&s->gb) < 0)
1121 return AVERROR_INVALIDDATA;
1123 for (l = 0; l < s->vq_start_subband[k]; l++) {
1126 /* Select the mid-tread linear quantizer */
1127 int abits = s->bitalloc[k][l];
1129 float quant_step_size = quant_step_table[abits];
1132 * Determine quantization index code book and its type
1135 /* Select quantization index code book */
1136 int sel = s->quant_index_huffman[k][abits];
1139 * Extract bits from the bit stream
1143 memset(block + 8 * l, 0, 8 * sizeof(block[0]));
1145 /* Deal with transients */
1146 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1147 rscale[l] = quant_step_size * s->scale_factor[k][l][sfi] *
1148 s->scalefactor_adj[k][sel];
1150 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1153 int block_code1, block_code2, size, levels, err;
1155 size = abits_sizes[abits - 1];
1156 levels = abits_levels[abits - 1];
1158 block_code1 = get_bits(&s->gb, size);
1159 block_code2 = get_bits(&s->gb, size);
1160 err = decode_blockcodes(block_code1, block_code2,
1161 levels, block + 8 * l);
1163 av_log(s->avctx, AV_LOG_ERROR,
1164 "ERROR: block code look-up failed\n");
1165 return AVERROR_INVALIDDATA;
1169 for (m = 0; m < 8; m++)
1170 block[8 * l + m] = get_sbits(&s->gb, abits - 3);
1174 for (m = 0; m < 8; m++)
1175 block[8 * l + m] = get_bitalloc(&s->gb,
1176 &dca_smpl_bitalloc[abits], sel);
1182 s->fmt_conv.int32_to_float_fmul_array8(&s->fmt_conv, subband_samples[k][0],
1183 block, rscale, 8 * s->vq_start_subband[k]);
1185 for (l = 0; l < s->vq_start_subband[k]; l++) {
1188 * Inverse ADPCM if in prediction mode
1190 if (s->prediction_mode[k][l]) {
1192 for (m = 0; m < 8; m++) {
1193 for (n = 1; n <= 4; n++)
1195 subband_samples[k][l][m] +=
1196 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1197 subband_samples[k][l][m - n] / 8192);
1198 else if (s->predictor_history)
1199 subband_samples[k][l][m] +=
1200 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1201 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1207 * Decode VQ encoded high frequencies
1209 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1210 /* 1 vector -> 32 samples but we only need the 8 samples
1211 * for this subsubframe. */
1212 int hfvq = s->high_freq_vq[k][l];
1214 if (!s->debug_flag & 0x01) {
1215 av_log(s->avctx, AV_LOG_DEBUG,
1216 "Stream with high frequencies VQ coding\n");
1217 s->debug_flag |= 0x01;
1220 int8x8_fmul_int32(subband_samples[k][l],
1221 &high_freq_vq[hfvq][subsubframe * 8],
1222 s->scale_factor[k][l][0]);
1226 /* Check for DSYNC after subsubframe */
1227 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1228 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1230 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1233 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1234 return AVERROR_INVALIDDATA;
1238 /* Backup predictor history for adpcm */
1239 for (k = base_channel; k < s->prim_channels; k++)
1240 for (l = 0; l < s->vq_start_subband[k]; l++)
1241 memcpy(s->subband_samples_hist[k][l],
1242 &subband_samples[k][l][4],
1243 4 * sizeof(subband_samples[0][0][0]));
1248 static int dca_filter_channels(DCAContext *s, int block_index)
1250 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1253 /* 32 subbands QMF */
1254 for (k = 0; k < s->prim_channels; k++) {
1255 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1256 0, 8388608.0, 8388608.0 };*/
1257 if (s->channel_order_tab[k] >= 0)
1258 qmf_32_subbands(s, k, subband_samples[k],
1259 s->samples_chanptr[s->channel_order_tab[k]],
1260 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1263 /* Generate LFE samples for this subsubframe FIXME!!! */
1264 if (s->output & DCA_LFE) {
1265 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1266 s->lfe_data + 2 * s->lfe * (block_index + 4),
1267 s->samples_chanptr[dca_lfe_index[s->amode]],
1268 1.0 / (256.0 * 32768.0));
1269 /* Outputs 20bits pcm samples */
1272 /* Downmixing to Stereo */
1273 if (s->prim_channels + !!s->lfe > 2 &&
1274 s->avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) {
1275 dca_downmix(s->samples_chanptr, s->amode, !!s->lfe, s->downmix_coef,
1276 s->channel_order_tab);
1283 static int dca_subframe_footer(DCAContext *s, int base_channel)
1285 int in, out, aux_data_count, aux_data_end, reserved;
1288 * Unpack optional information
1291 /* presumably optional information only appears in the core? */
1292 if (!base_channel) {
1294 skip_bits_long(&s->gb, 32);
1297 aux_data_count = get_bits(&s->gb, 6);
1300 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1302 aux_data_end = 8 * aux_data_count + get_bits_count(&s->gb);
1304 if (get_bits_long(&s->gb, 32) != 0x9A1105A0) // nSYNCAUX
1305 return AVERROR_INVALIDDATA;
1307 if (get_bits1(&s->gb)) { // bAUXTimeStampFlag
1308 avpriv_request_sample(s->avctx,
1309 "Auxiliary Decode Time Stamp Flag");
1311 skip_bits(&s->gb, (-get_bits_count(&s->gb)) & 4);
1312 // 44 bits: nMSByte (8), nMarker (4), nLSByte (28), nMarker (4)
1313 skip_bits_long(&s->gb, 44);
1316 if ((s->core_downmix = get_bits1(&s->gb))) {
1317 switch (get_bits(&s->gb, 3)) {
1319 s->core_downmix_amode = DCA_MONO;
1322 s->core_downmix_amode = DCA_STEREO;
1325 s->core_downmix_amode = DCA_STEREO_TOTAL;
1328 s->core_downmix_amode = DCA_3F;
1331 s->core_downmix_amode = DCA_2F1R;
1334 s->core_downmix_amode = DCA_2F2R;
1337 s->core_downmix_amode = DCA_3F1R;
1340 return AVERROR_INVALIDDATA;
1342 for (out = 0; out < dca_channels[s->core_downmix_amode]; out++) {
1343 for (in = 0; in < s->prim_channels + !!s->lfe; in++) {
1344 uint16_t tmp = get_bits(&s->gb, 9);
1345 if ((tmp & 0xFF) > 241)
1346 return AVERROR_INVALIDDATA;
1347 s->core_downmix_codes[in][out] = tmp;
1352 align_get_bits(&s->gb); // byte align
1353 skip_bits(&s->gb, 16); // nAUXCRC16
1355 // additional data (reserved, cf. ETSI TS 102 114 V1.4.1)
1356 if ((reserved = (aux_data_end - get_bits_count(&s->gb))) < 0)
1357 return AVERROR_INVALIDDATA;
1358 else if (reserved) {
1359 avpriv_request_sample(s->avctx,
1360 "Core auxiliary data reserved content");
1361 skip_bits_long(&s->gb, reserved);
1365 if (s->crc_present && s->dynrange)
1366 get_bits(&s->gb, 16);
1373 * Decode a dca frame block
1375 * @param s pointer to the DCAContext
1378 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1383 if (s->current_subframe >= s->subframes) {
1384 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1385 s->current_subframe, s->subframes);
1386 return AVERROR_INVALIDDATA;
1389 if (!s->current_subsubframe) {
1391 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1393 /* Read subframe header */
1394 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1398 /* Read subsubframe */
1400 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1402 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1406 s->current_subsubframe++;
1407 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1408 s->current_subsubframe = 0;
1409 s->current_subframe++;
1411 if (s->current_subframe >= s->subframes) {
1413 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1415 /* Read subframe footer */
1416 if ((ret = dca_subframe_footer(s, base_channel)))
1424 * Return the number of channels in an ExSS speaker mask (HD)
1426 static int dca_exss_mask2count(int mask)
1428 /* count bits that mean speaker pairs twice */
1429 return av_popcount(mask) +
1430 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1431 DCA_EXSS_FRONT_LEFT_RIGHT |
1432 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1433 DCA_EXSS_WIDE_LEFT_RIGHT |
1434 DCA_EXSS_SIDE_LEFT_RIGHT |
1435 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1436 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1437 DCA_EXSS_REAR_LEFT_RIGHT |
1438 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1442 * Skip mixing coefficients of a single mix out configuration (HD)
1444 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1448 for (i = 0; i < channels; i++) {
1449 int mix_map_mask = get_bits(gb, out_ch);
1450 int num_coeffs = av_popcount(mix_map_mask);
1451 skip_bits_long(gb, num_coeffs * 6);
1456 * Parse extension substream asset header (HD)
1458 static int dca_exss_parse_asset_header(DCAContext *s)
1460 int header_pos = get_bits_count(&s->gb);
1463 int embedded_stereo = 0;
1464 int embedded_6ch = 0;
1465 int drc_code_present;
1466 int extensions_mask;
1469 if (get_bits_left(&s->gb) < 16)
1472 /* We will parse just enough to get to the extensions bitmask with which
1473 * we can set the profile value. */
1475 header_size = get_bits(&s->gb, 9) + 1;
1476 skip_bits(&s->gb, 3); // asset index
1478 if (s->static_fields) {
1479 if (get_bits1(&s->gb))
1480 skip_bits(&s->gb, 4); // asset type descriptor
1481 if (get_bits1(&s->gb))
1482 skip_bits_long(&s->gb, 24); // language descriptor
1484 if (get_bits1(&s->gb)) {
1485 /* How can one fit 1024 bytes of text here if the maximum value
1486 * for the asset header size field above was 512 bytes? */
1487 int text_length = get_bits(&s->gb, 10) + 1;
1488 if (get_bits_left(&s->gb) < text_length * 8)
1490 skip_bits_long(&s->gb, text_length * 8); // info text
1493 skip_bits(&s->gb, 5); // bit resolution - 1
1494 skip_bits(&s->gb, 4); // max sample rate code
1495 channels = get_bits(&s->gb, 8) + 1;
1497 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1498 int spkr_remap_sets;
1499 int spkr_mask_size = 16;
1503 embedded_stereo = get_bits1(&s->gb);
1505 embedded_6ch = get_bits1(&s->gb);
1507 if (get_bits1(&s->gb)) {
1508 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1509 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1512 spkr_remap_sets = get_bits(&s->gb, 3);
1514 for (i = 0; i < spkr_remap_sets; i++) {
1515 /* std layout mask for each remap set */
1516 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1519 for (i = 0; i < spkr_remap_sets; i++) {
1520 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1521 if (get_bits_left(&s->gb) < 0)
1524 for (j = 0; j < num_spkrs[i]; j++) {
1525 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1526 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1527 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1532 skip_bits(&s->gb, 3); // representation type
1536 drc_code_present = get_bits1(&s->gb);
1537 if (drc_code_present)
1538 get_bits(&s->gb, 8); // drc code
1540 if (get_bits1(&s->gb))
1541 skip_bits(&s->gb, 5); // dialog normalization code
1543 if (drc_code_present && embedded_stereo)
1544 get_bits(&s->gb, 8); // drc stereo code
1546 if (s->mix_metadata && get_bits1(&s->gb)) {
1547 skip_bits(&s->gb, 1); // external mix
1548 skip_bits(&s->gb, 6); // post mix gain code
1550 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1551 skip_bits(&s->gb, 3); // drc limit
1553 skip_bits(&s->gb, 8); // custom drc code
1555 if (get_bits1(&s->gb)) // channel specific scaling
1556 for (i = 0; i < s->num_mix_configs; i++)
1557 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1559 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1561 for (i = 0; i < s->num_mix_configs; i++) {
1562 if (get_bits_left(&s->gb) < 0)
1564 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1566 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1567 if (embedded_stereo)
1568 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1572 switch (get_bits(&s->gb, 2)) {
1573 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1574 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1575 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1576 case 3: extensions_mask = 0; /* aux coding */ break;
1579 /* not parsed further, we were only interested in the extensions mask */
1581 if (get_bits_left(&s->gb) < 0)
1584 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1585 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1588 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1590 if (extensions_mask & DCA_EXT_EXSS_XLL)
1591 s->profile = FF_PROFILE_DTS_HD_MA;
1592 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1594 s->profile = FF_PROFILE_DTS_HD_HRA;
1596 if (!(extensions_mask & DCA_EXT_CORE))
1597 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1598 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1599 av_log(s->avctx, AV_LOG_WARNING,
1600 "DTS extensions detection mismatch (%d, %d)\n",
1601 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1607 * Parse extension substream header (HD)
1609 static void dca_exss_parse_header(DCAContext *s)
1615 int active_ss_mask[8];
1618 if (get_bits_left(&s->gb) < 52)
1621 skip_bits(&s->gb, 8); // user data
1622 ss_index = get_bits(&s->gb, 2);
1624 blownup = get_bits1(&s->gb);
1625 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1626 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1628 s->static_fields = get_bits1(&s->gb);
1629 if (s->static_fields) {
1630 skip_bits(&s->gb, 2); // reference clock code
1631 skip_bits(&s->gb, 3); // frame duration code
1633 if (get_bits1(&s->gb))
1634 skip_bits_long(&s->gb, 36); // timestamp
1636 /* a single stream can contain multiple audio assets that can be
1637 * combined to form multiple audio presentations */
1639 num_audiop = get_bits(&s->gb, 3) + 1;
1640 if (num_audiop > 1) {
1641 avpriv_request_sample(s->avctx,
1642 "Multiple DTS-HD audio presentations");
1643 /* ignore such streams for now */
1647 num_assets = get_bits(&s->gb, 3) + 1;
1648 if (num_assets > 1) {
1649 avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
1650 /* ignore such streams for now */
1654 for (i = 0; i < num_audiop; i++)
1655 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1657 for (i = 0; i < num_audiop; i++)
1658 for (j = 0; j <= ss_index; j++)
1659 if (active_ss_mask[i] & (1 << j))
1660 skip_bits(&s->gb, 8); // active asset mask
1662 s->mix_metadata = get_bits1(&s->gb);
1663 if (s->mix_metadata) {
1664 int mix_out_mask_size;
1666 skip_bits(&s->gb, 2); // adjustment level
1667 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1668 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1670 for (i = 0; i < s->num_mix_configs; i++) {
1671 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1672 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1677 for (i = 0; i < num_assets; i++)
1678 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1680 for (i = 0; i < num_assets; i++) {
1681 if (dca_exss_parse_asset_header(s))
1685 /* not parsed further, we were only interested in the extensions mask
1686 * from the asset header */
1690 * Main frame decoding function
1691 * FIXME add arguments
1693 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1694 int *got_frame_ptr, AVPacket *avpkt)
1696 AVFrame *frame = data;
1697 const uint8_t *buf = avpkt->data;
1698 int buf_size = avpkt->size;
1701 int num_core_channels = 0;
1703 float **samples_flt;
1704 DCAContext *s = avctx->priv_data;
1705 int channels, full_channels;
1711 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1712 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1713 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1714 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1715 return AVERROR_INVALIDDATA;
1718 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1719 if ((ret = dca_parse_frame_header(s)) < 0) {
1720 //seems like the frame is corrupt, try with the next one
1723 //set AVCodec values with parsed data
1724 avctx->sample_rate = s->sample_rate;
1725 avctx->bit_rate = s->bit_rate;
1727 s->profile = FF_PROFILE_DTS;
1729 for (i = 0; i < (s->sample_blocks / 8); i++) {
1730 if ((ret = dca_decode_block(s, 0, i))) {
1731 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1736 /* record number of core channels incase less than max channels are requested */
1737 num_core_channels = s->prim_channels;
1740 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1742 s->core_ext_mask = 0;
1744 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1746 /* only scan for extensions if ext_descr was unknown or indicated a
1747 * supported XCh extension */
1748 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1750 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1751 * extensions scan can fill it up */
1752 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1754 /* extensions start at 32-bit boundaries into bitstream */
1755 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1757 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1758 uint32_t bits = get_bits_long(&s->gb, 32);
1762 int ext_amode, xch_fsize;
1764 s->xch_base_channel = s->prim_channels;
1766 /* validate sync word using XCHFSIZE field */
1767 xch_fsize = show_bits(&s->gb, 10);
1768 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1769 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1772 /* skip length-to-end-of-frame field for the moment */
1773 skip_bits(&s->gb, 10);
1775 s->core_ext_mask |= DCA_EXT_XCH;
1777 /* extension amode(number of channels in extension) should be 1 */
1778 /* AFAIK XCh is not used for more channels */
1779 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1780 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1781 " supported!\n", ext_amode);
1785 /* much like core primary audio coding header */
1786 dca_parse_audio_coding_header(s, s->xch_base_channel);
1788 for (i = 0; i < (s->sample_blocks / 8); i++)
1789 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1790 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1798 /* XXCh: extended channels */
1799 /* usually found either in core or HD part in DTS-HD HRA streams,
1800 * but not in DTS-ES which contains XCh extensions instead */
1801 s->core_ext_mask |= DCA_EXT_XXCH;
1805 int fsize96 = show_bits(&s->gb, 12) + 1;
1806 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1809 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1810 get_bits_count(&s->gb));
1811 skip_bits(&s->gb, 12);
1812 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1813 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1815 s->core_ext_mask |= DCA_EXT_X96;
1820 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1823 /* no supported extensions, skip the rest of the core substream */
1824 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1827 if (s->core_ext_mask & DCA_EXT_X96)
1828 s->profile = FF_PROFILE_DTS_96_24;
1829 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1830 s->profile = FF_PROFILE_DTS_ES;
1832 /* check for ExSS (HD part) */
1833 if (s->dca_buffer_size - s->frame_size > 32 &&
1834 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1835 dca_exss_parse_header(s);
1837 avctx->profile = s->profile;
1839 full_channels = channels = s->prim_channels + !!s->lfe;
1841 if (s->amode < 16) {
1842 avctx->channel_layout = dca_core_channel_layout[s->amode];
1844 #if FF_API_REQUEST_CHANNELS
1845 FF_DISABLE_DEPRECATION_WARNINGS
1846 if (s->xch_present && !s->xch_disable &&
1847 (!avctx->request_channels ||
1848 avctx->request_channels > num_core_channels + !!s->lfe)) {
1849 FF_ENABLE_DEPRECATION_WARNINGS
1851 if (s->xch_present && !s->xch_disable) {
1853 avctx->channel_layout |= AV_CH_BACK_CENTER;
1855 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1856 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1858 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1861 channels = num_core_channels + !!s->lfe;
1862 s->xch_present = 0; /* disable further xch processing */
1864 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1865 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1867 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1870 if (channels > !!s->lfe &&
1871 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1872 return AVERROR_INVALIDDATA;
1874 if (s->prim_channels + !!s->lfe > 2 &&
1875 avctx->request_channel_layout == AV_CH_LAYOUT_STEREO) {
1877 s->output = DCA_STEREO;
1878 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1880 /* Stereo downmix coefficients
1882 * The decoder can only downmix to 2-channel, so we need to ensure
1883 * embedded downmix coefficients are actually targeting 2-channel.
1885 if (s->core_downmix && (s->core_downmix_amode == DCA_STEREO ||
1886 s->core_downmix_amode == DCA_STEREO_TOTAL)) {
1888 for (i = 0; i < s->prim_channels + !!s->lfe; i++) {
1889 sign = s->core_downmix_codes[i][0] & 0x100 ? 1 : -1;
1890 code = s->core_downmix_codes[i][0] & 0x0FF;
1891 s->downmix_coef[i][0] = (!code ? 0.0f :
1892 sign * dca_dmixtable[code - 1]);
1893 sign = s->core_downmix_codes[i][1] & 0x100 ? 1 : -1;
1894 code = s->core_downmix_codes[i][1] & 0x0FF;
1895 s->downmix_coef[i][1] = (!code ? 0.0f :
1896 sign * dca_dmixtable[code - 1]);
1899 int am = s->amode & DCA_CHANNEL_MASK;
1900 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
1901 av_log(s->avctx, AV_LOG_ERROR,
1902 "Invalid channel mode %d\n", am);
1903 return AVERROR_INVALIDDATA;
1905 if (s->prim_channels + !!s->lfe >
1906 FF_ARRAY_ELEMS(dca_default_coeffs[0])) {
1907 avpriv_request_sample(s->avctx, "Downmixing %d channels",
1908 s->prim_channels + !!s->lfe);
1909 return AVERROR_PATCHWELCOME;
1911 for (i = 0; i < s->prim_channels + !!s->lfe; i++) {
1912 s->downmix_coef[i][0] = dca_default_coeffs[am][i][0];
1913 s->downmix_coef[i][1] = dca_default_coeffs[am][i][1];
1916 av_dlog(s->avctx, "Stereo downmix coeffs:\n");
1917 for (i = 0; i < s->prim_channels + !!s->lfe; i++) {
1918 av_dlog(s->avctx, "L, input channel %d = %f\n", i,
1919 s->downmix_coef[i][0]);
1920 av_dlog(s->avctx, "R, input channel %d = %f\n", i,
1921 s->downmix_coef[i][1]);
1923 av_dlog(s->avctx, "\n");
1926 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1927 return AVERROR_INVALIDDATA;
1929 avctx->channels = channels;
1931 /* get output buffer */
1932 frame->nb_samples = 256 * (s->sample_blocks / 8);
1933 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
1934 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1937 samples_flt = (float **)frame->extended_data;
1939 /* allocate buffer for extra channels if downmixing */
1940 if (avctx->channels < full_channels) {
1941 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
1943 avctx->sample_fmt, 0);
1947 av_fast_malloc(&s->extra_channels_buffer,
1948 &s->extra_channels_buffer_size, ret);
1949 if (!s->extra_channels_buffer)
1950 return AVERROR(ENOMEM);
1952 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
1953 s->extra_channels_buffer,
1954 full_channels - channels,
1955 frame->nb_samples, avctx->sample_fmt, 0);
1960 /* filter to get final output */
1961 for (i = 0; i < (s->sample_blocks / 8); i++) {
1964 for (ch = 0; ch < channels; ch++)
1965 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
1966 for (; ch < full_channels; ch++)
1967 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
1969 dca_filter_channels(s, i);
1971 /* If this was marked as a DTS-ES stream we need to subtract back- */
1972 /* channel from SL & SR to remove matrixed back-channel signal */
1973 if ((s->source_pcm_res & 1) && s->xch_present) {
1974 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
1975 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
1976 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
1977 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1978 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1982 /* update lfe history */
1983 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1984 for (i = 0; i < 2 * s->lfe * 4; i++)
1985 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1995 * DCA initialization
1997 * @param avctx pointer to the AVCodecContext
2000 static av_cold int dca_decode_init(AVCodecContext *avctx)
2002 DCAContext *s = avctx->priv_data;
2007 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
2008 ff_mdct_init(&s->imdct, 6, 1, 1.0);
2009 ff_synth_filter_init(&s->synth);
2010 ff_dcadsp_init(&s->dcadsp);
2011 ff_fmt_convert_init(&s->fmt_conv, avctx);
2013 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
2015 /* allow downmixing to stereo */
2016 #if FF_API_REQUEST_CHANNELS
2017 FF_DISABLE_DEPRECATION_WARNINGS
2018 if (avctx->request_channels == 2)
2019 avctx->request_channel_layout = AV_CH_LAYOUT_STEREO;
2020 FF_ENABLE_DEPRECATION_WARNINGS
2022 if (avctx->channels > 2 &&
2023 avctx->request_channel_layout == AV_CH_LAYOUT_STEREO)
2024 avctx->channels = 2;
2029 static av_cold int dca_decode_end(AVCodecContext *avctx)
2031 DCAContext *s = avctx->priv_data;
2032 ff_mdct_end(&s->imdct);
2033 av_freep(&s->extra_channels_buffer);
2037 static const AVProfile profiles[] = {
2038 { FF_PROFILE_DTS, "DTS" },
2039 { FF_PROFILE_DTS_ES, "DTS-ES" },
2040 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
2041 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
2042 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
2043 { FF_PROFILE_UNKNOWN },
2046 static const AVOption options[] = {
2047 { "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 },
2051 static const AVClass dca_decoder_class = {
2052 .class_name = "DCA decoder",
2053 .item_name = av_default_item_name,
2055 .version = LIBAVUTIL_VERSION_INT,
2058 AVCodec ff_dca_decoder = {
2060 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
2061 .type = AVMEDIA_TYPE_AUDIO,
2062 .id = AV_CODEC_ID_DTS,
2063 .priv_data_size = sizeof(DCAContext),
2064 .init = dca_decode_init,
2065 .decode = dca_decode_frame,
2066 .close = dca_decode_end,
2067 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
2068 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
2069 AV_SAMPLE_FMT_NONE },
2070 .profiles = NULL_IF_CONFIG_SMALL(profiles),
2071 .priv_class = &dca_decoder_class,