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/samplefmt.h"
44 #include "synth_filter.h"
46 #include "fmtconvert.h"
55 #define DCA_PRIM_CHANNELS_MAX (7)
56 #define DCA_SUBBANDS (32)
57 #define DCA_ABITS_MAX (32) /* Should be 28 */
58 #define DCA_SUBSUBFRAMES_MAX (4)
59 #define DCA_SUBFRAMES_MAX (16)
60 #define DCA_BLOCKS_MAX (16)
61 #define DCA_LFE_MAX (3)
77 /* these are unconfirmed but should be mostly correct */
78 enum DCAExSSSpeakerMask {
79 DCA_EXSS_FRONT_CENTER = 0x0001,
80 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
81 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
82 DCA_EXSS_LFE = 0x0008,
83 DCA_EXSS_REAR_CENTER = 0x0010,
84 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
85 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
86 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
87 DCA_EXSS_OVERHEAD = 0x0100,
88 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
89 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
90 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
91 DCA_EXSS_LFE2 = 0x1000,
92 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
93 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
94 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
97 enum DCAExtensionMask {
98 DCA_EXT_CORE = 0x001, ///< core in core substream
99 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
100 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
101 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
102 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
103 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
104 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
105 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
106 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
107 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
110 /* -1 are reserved or unknown */
111 static const int dca_ext_audio_descr_mask[] = {
115 DCA_EXT_XCH | DCA_EXT_X96,
122 /* extensions that reside in core substream */
123 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
125 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
126 * Some compromises have been made for special configurations. Most configurations
127 * are never used so complete accuracy is not needed.
129 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
130 * S -> side, when both rear and back are configured move one of them to the side channel
132 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
134 static const uint64_t dca_core_channel_layout[] = {
135 AV_CH_FRONT_CENTER, ///< 1, A
136 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
137 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
138 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
139 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
140 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
141 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
142 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
143 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
145 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
146 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
148 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
149 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
151 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
152 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
154 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
155 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
156 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
158 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
159 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
160 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
162 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
163 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
164 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
166 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
167 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
168 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
171 static const int8_t dca_lfe_index[] = {
172 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
175 static const int8_t dca_channel_reorder_lfe[][9] = {
176 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
177 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
178 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
179 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
180 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
181 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
182 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
183 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
184 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
185 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
186 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
187 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
188 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
189 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
190 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
191 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
194 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
195 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
196 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
197 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
198 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
199 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
200 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
201 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
202 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
203 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
204 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
205 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
206 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
207 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
208 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
209 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
210 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
213 static const int8_t dca_channel_reorder_nolfe[][9] = {
214 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
215 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
216 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
217 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
218 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
219 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
220 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
221 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
222 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
223 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
224 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
225 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
226 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
227 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
228 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
229 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
232 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
233 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
234 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
235 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
236 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
237 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
238 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
239 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
240 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
241 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
242 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
243 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
244 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
245 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
246 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
247 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
248 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
251 #define DCA_DOLBY 101 /* FIXME */
253 #define DCA_CHANNEL_BITS 6
254 #define DCA_CHANNEL_MASK 0x3F
258 #define HEADER_SIZE 14
260 #define DCA_MAX_FRAME_SIZE 16384
261 #define DCA_MAX_EXSS_HEADER_SIZE 4096
263 #define DCA_BUFFER_PADDING_SIZE 1024
265 /** Bit allocation */
267 int offset; ///< code values offset
268 int maxbits[8]; ///< max bits in VLC
269 int wrap; ///< wrap for get_vlc2()
270 VLC vlc[8]; ///< actual codes
273 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
274 static BitAlloc dca_tmode; ///< transition mode VLCs
275 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
276 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
278 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
281 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
286 AVCodecContext *avctx;
288 int frame_type; ///< type of the current frame
289 int samples_deficit; ///< deficit sample count
290 int crc_present; ///< crc is present in the bitstream
291 int sample_blocks; ///< number of PCM sample blocks
292 int frame_size; ///< primary frame byte size
293 int amode; ///< audio channels arrangement
294 int sample_rate; ///< audio sampling rate
295 int bit_rate; ///< transmission bit rate
296 int bit_rate_index; ///< transmission bit rate index
298 int downmix; ///< embedded downmix enabled
299 int dynrange; ///< embedded dynamic range flag
300 int timestamp; ///< embedded time stamp flag
301 int aux_data; ///< auxiliary data flag
302 int hdcd; ///< source material is mastered in HDCD
303 int ext_descr; ///< extension audio descriptor flag
304 int ext_coding; ///< extended coding flag
305 int aspf; ///< audio sync word insertion flag
306 int lfe; ///< low frequency effects flag
307 int predictor_history; ///< predictor history flag
308 int header_crc; ///< header crc check bytes
309 int multirate_inter; ///< multirate interpolator switch
310 int version; ///< encoder software revision
311 int copy_history; ///< copy history
312 int source_pcm_res; ///< source pcm resolution
313 int front_sum; ///< front sum/difference flag
314 int surround_sum; ///< surround sum/difference flag
315 int dialog_norm; ///< dialog normalisation parameter
317 /* Primary audio coding header */
318 int subframes; ///< number of subframes
319 int total_channels; ///< number of channels including extensions
320 int prim_channels; ///< number of primary audio channels
321 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
322 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
323 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
324 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
325 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
326 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
327 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
328 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
330 /* Primary audio coding side information */
331 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
332 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
333 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
334 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
335 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
336 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
337 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
338 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
339 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
340 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
341 int dynrange_coef; ///< dynamic range coefficient
343 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
345 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
346 int lfe_scale_factor;
348 /* Subband samples history (for ADPCM) */
349 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
350 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
351 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
352 int hist_index[DCA_PRIM_CHANNELS_MAX];
353 DECLARE_ALIGNED(32, float, raXin)[32];
355 int output; ///< type of output
357 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
358 float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
359 float *extra_channels[DCA_PRIM_CHANNELS_MAX + 1];
360 uint8_t *extra_channels_buffer;
361 unsigned int extra_channels_buffer_size;
363 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
364 int dca_buffer_size; ///< how much data is in the dca_buffer
366 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
368 /* Current position in DCA frame */
369 int current_subframe;
370 int current_subsubframe;
372 int core_ext_mask; ///< present extensions in the core substream
374 /* XCh extension information */
375 int xch_present; ///< XCh extension present and valid
376 int xch_base_channel; ///< index of first (only) channel containing XCH data
378 /* ExSS header parser */
379 int static_fields; ///< static fields present
380 int mix_metadata; ///< mixing metadata present
381 int num_mix_configs; ///< number of mix out configurations
382 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
386 int debug_flag; ///< used for suppressing repeated error messages output
387 AVFloatDSPContext fdsp;
389 SynthFilterContext synth;
390 DCADSPContext dcadsp;
391 FmtConvertContext fmt_conv;
394 static const uint16_t dca_vlc_offs[] = {
395 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
396 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
397 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
398 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
399 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
400 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
403 static av_cold void dca_init_vlcs(void)
405 static int vlcs_initialized = 0;
407 static VLC_TYPE dca_table[23622][2];
409 if (vlcs_initialized)
412 dca_bitalloc_index.offset = 1;
413 dca_bitalloc_index.wrap = 2;
414 for (i = 0; i < 5; i++) {
415 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
416 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
417 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
418 bitalloc_12_bits[i], 1, 1,
419 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
421 dca_scalefactor.offset = -64;
422 dca_scalefactor.wrap = 2;
423 for (i = 0; i < 5; i++) {
424 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
425 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
426 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
427 scales_bits[i], 1, 1,
428 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
430 dca_tmode.offset = 0;
432 for (i = 0; i < 4; i++) {
433 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
434 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
435 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
437 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
440 for (i = 0; i < 10; i++)
441 for (j = 0; j < 7; j++) {
442 if (!bitalloc_codes[i][j])
444 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
445 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
446 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
447 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
449 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
451 bitalloc_bits[i][j], 1, 1,
452 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
455 vlcs_initialized = 1;
458 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
461 *dst++ = get_bits(gb, bits);
464 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
467 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
468 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
469 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
471 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
472 s->prim_channels = s->total_channels;
474 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
475 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
478 for (i = base_channel; i < s->prim_channels; i++) {
479 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
480 if (s->subband_activity[i] > DCA_SUBBANDS)
481 s->subband_activity[i] = DCA_SUBBANDS;
483 for (i = base_channel; i < s->prim_channels; i++) {
484 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
485 if (s->vq_start_subband[i] > DCA_SUBBANDS)
486 s->vq_start_subband[i] = DCA_SUBBANDS;
488 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
489 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
490 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
491 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
493 /* Get codebooks quantization indexes */
495 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
496 for (j = 1; j < 11; j++)
497 for (i = base_channel; i < s->prim_channels; i++)
498 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
500 /* Get scale factor adjustment */
501 for (j = 0; j < 11; j++)
502 for (i = base_channel; i < s->prim_channels; i++)
503 s->scalefactor_adj[i][j] = 1;
505 for (j = 1; j < 11; j++)
506 for (i = base_channel; i < s->prim_channels; i++)
507 if (s->quant_index_huffman[i][j] < thr[j])
508 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
510 if (s->crc_present) {
511 /* Audio header CRC check */
512 get_bits(&s->gb, 16);
515 s->current_subframe = 0;
516 s->current_subsubframe = 0;
519 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
520 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
521 for (i = base_channel; i < s->prim_channels; i++) {
522 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
523 s->subband_activity[i]);
524 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
525 s->vq_start_subband[i]);
526 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
527 s->joint_intensity[i]);
528 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
529 s->transient_huffman[i]);
530 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
531 s->scalefactor_huffman[i]);
532 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
533 s->bitalloc_huffman[i]);
534 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
535 for (j = 0; j < 11; j++)
536 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
537 av_log(s->avctx, AV_LOG_DEBUG, "\n");
538 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
539 for (j = 0; j < 11; j++)
540 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
541 av_log(s->avctx, AV_LOG_DEBUG, "\n");
548 static int dca_parse_frame_header(DCAContext *s)
550 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
553 skip_bits_long(&s->gb, 32);
556 s->frame_type = get_bits(&s->gb, 1);
557 s->samples_deficit = get_bits(&s->gb, 5) + 1;
558 s->crc_present = get_bits(&s->gb, 1);
559 s->sample_blocks = get_bits(&s->gb, 7) + 1;
560 s->frame_size = get_bits(&s->gb, 14) + 1;
561 if (s->frame_size < 95)
562 return AVERROR_INVALIDDATA;
563 s->amode = get_bits(&s->gb, 6);
564 s->sample_rate = avpriv_dca_sample_rates[get_bits(&s->gb, 4)];
566 return AVERROR_INVALIDDATA;
567 s->bit_rate_index = get_bits(&s->gb, 5);
568 s->bit_rate = dca_bit_rates[s->bit_rate_index];
570 return AVERROR_INVALIDDATA;
572 s->downmix = get_bits(&s->gb, 1);
573 s->dynrange = get_bits(&s->gb, 1);
574 s->timestamp = get_bits(&s->gb, 1);
575 s->aux_data = get_bits(&s->gb, 1);
576 s->hdcd = get_bits(&s->gb, 1);
577 s->ext_descr = get_bits(&s->gb, 3);
578 s->ext_coding = get_bits(&s->gb, 1);
579 s->aspf = get_bits(&s->gb, 1);
580 s->lfe = get_bits(&s->gb, 2);
581 s->predictor_history = get_bits(&s->gb, 1);
584 av_log(s->avctx, AV_LOG_ERROR, "Invalid LFE value: %d\n", s->lfe);
585 return AVERROR_INVALIDDATA;
588 /* TODO: check CRC */
590 s->header_crc = get_bits(&s->gb, 16);
592 s->multirate_inter = get_bits(&s->gb, 1);
593 s->version = get_bits(&s->gb, 4);
594 s->copy_history = get_bits(&s->gb, 2);
595 s->source_pcm_res = get_bits(&s->gb, 3);
596 s->front_sum = get_bits(&s->gb, 1);
597 s->surround_sum = get_bits(&s->gb, 1);
598 s->dialog_norm = get_bits(&s->gb, 4);
600 /* FIXME: channels mixing levels */
601 s->output = s->amode;
603 s->output |= DCA_LFE;
606 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
607 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
608 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
609 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
610 s->sample_blocks, s->sample_blocks * 32);
611 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
612 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
613 s->amode, dca_channels[s->amode]);
614 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
616 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
618 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
619 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
620 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
621 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
622 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
623 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
624 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
625 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
626 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
627 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
628 s->predictor_history);
629 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
630 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
632 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
633 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
634 av_log(s->avctx, AV_LOG_DEBUG,
635 "source pcm resolution: %i (%i bits/sample)\n",
636 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
637 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
638 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
639 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
640 av_log(s->avctx, AV_LOG_DEBUG, "\n");
643 /* Primary audio coding header */
644 s->subframes = get_bits(&s->gb, 4) + 1;
646 return dca_parse_audio_coding_header(s, 0);
650 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
653 /* huffman encoded */
654 value += get_bitalloc(gb, &dca_scalefactor, level);
655 value = av_clip(value, 0, (1 << log2range) - 1);
656 } else if (level < 8) {
657 if (level + 1 > log2range) {
658 skip_bits(gb, level + 1 - log2range);
659 value = get_bits(gb, log2range);
661 value = get_bits(gb, level + 1);
667 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
669 /* Primary audio coding side information */
672 if (get_bits_left(&s->gb) < 0)
673 return AVERROR_INVALIDDATA;
676 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
677 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
680 for (j = base_channel; j < s->prim_channels; j++) {
681 for (k = 0; k < s->subband_activity[j]; k++)
682 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
685 /* Get prediction codebook */
686 for (j = base_channel; j < s->prim_channels; j++) {
687 for (k = 0; k < s->subband_activity[j]; k++) {
688 if (s->prediction_mode[j][k] > 0) {
689 /* (Prediction coefficient VQ address) */
690 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
695 /* Bit allocation index */
696 for (j = base_channel; j < s->prim_channels; j++) {
697 for (k = 0; k < s->vq_start_subband[j]; k++) {
698 if (s->bitalloc_huffman[j] == 6)
699 s->bitalloc[j][k] = get_bits(&s->gb, 5);
700 else if (s->bitalloc_huffman[j] == 5)
701 s->bitalloc[j][k] = get_bits(&s->gb, 4);
702 else if (s->bitalloc_huffman[j] == 7) {
703 av_log(s->avctx, AV_LOG_ERROR,
704 "Invalid bit allocation index\n");
705 return AVERROR_INVALIDDATA;
708 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
711 if (s->bitalloc[j][k] > 26) {
712 av_dlog(s->avctx, "bitalloc index [%i][%i] too big (%i)\n",
713 j, k, s->bitalloc[j][k]);
714 return AVERROR_INVALIDDATA;
719 /* Transition mode */
720 for (j = base_channel; j < s->prim_channels; j++) {
721 for (k = 0; k < s->subband_activity[j]; k++) {
722 s->transition_mode[j][k] = 0;
723 if (s->subsubframes[s->current_subframe] > 1 &&
724 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
725 s->transition_mode[j][k] =
726 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
731 if (get_bits_left(&s->gb) < 0)
732 return AVERROR_INVALIDDATA;
734 for (j = base_channel; j < s->prim_channels; j++) {
735 const uint32_t *scale_table;
736 int scale_sum, log_size;
738 memset(s->scale_factor[j], 0,
739 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
741 if (s->scalefactor_huffman[j] == 6) {
742 scale_table = scale_factor_quant7;
745 scale_table = scale_factor_quant6;
749 /* When huffman coded, only the difference is encoded */
752 for (k = 0; k < s->subband_activity[j]; k++) {
753 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
754 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
755 s->scale_factor[j][k][0] = scale_table[scale_sum];
758 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
759 /* Get second scale factor */
760 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
761 s->scale_factor[j][k][1] = scale_table[scale_sum];
766 /* Joint subband scale factor codebook select */
767 for (j = base_channel; j < s->prim_channels; j++) {
768 /* Transmitted only if joint subband coding enabled */
769 if (s->joint_intensity[j] > 0)
770 s->joint_huff[j] = get_bits(&s->gb, 3);
773 if (get_bits_left(&s->gb) < 0)
774 return AVERROR_INVALIDDATA;
776 /* Scale factors for joint subband coding */
777 for (j = base_channel; j < s->prim_channels; j++) {
780 /* Transmitted only if joint subband coding enabled */
781 if (s->joint_intensity[j] > 0) {
783 source_channel = s->joint_intensity[j] - 1;
785 /* When huffman coded, only the difference is encoded
786 * (is this valid as well for joint scales ???) */
788 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
789 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
790 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
793 if (!(s->debug_flag & 0x02)) {
794 av_log(s->avctx, AV_LOG_DEBUG,
795 "Joint stereo coding not supported\n");
796 s->debug_flag |= 0x02;
801 /* Stereo downmix coefficients */
802 if (!base_channel && s->prim_channels > 2) {
804 for (j = base_channel; j < s->prim_channels; j++) {
805 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
806 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
809 int am = s->amode & DCA_CHANNEL_MASK;
810 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
811 av_log(s->avctx, AV_LOG_ERROR,
812 "Invalid channel mode %d\n", am);
813 return AVERROR_INVALIDDATA;
815 if (s->prim_channels > FF_ARRAY_ELEMS(dca_default_coeffs[0])) {
816 avpriv_request_sample(s->avctx, "Downmixing %d channels",
818 return AVERROR_PATCHWELCOME;
821 for (j = base_channel; j < s->prim_channels; j++) {
822 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
823 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
828 /* Dynamic range coefficient */
829 if (!base_channel && s->dynrange)
830 s->dynrange_coef = get_bits(&s->gb, 8);
832 /* Side information CRC check word */
833 if (s->crc_present) {
834 get_bits(&s->gb, 16);
838 * Primary audio data arrays
841 /* VQ encoded high frequency subbands */
842 for (j = base_channel; j < s->prim_channels; j++)
843 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
844 /* 1 vector -> 32 samples */
845 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
847 /* Low frequency effect data */
848 if (!base_channel && s->lfe) {
850 int lfe_samples = 2 * s->lfe * (4 + block_index);
851 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
854 for (j = lfe_samples; j < lfe_end_sample; j++) {
855 /* Signed 8 bits int */
856 s->lfe_data[j] = get_sbits(&s->gb, 8);
859 /* Scale factor index */
860 skip_bits(&s->gb, 1);
861 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 7)];
863 /* Quantization step size * scale factor */
864 lfe_scale = 0.035 * s->lfe_scale_factor;
866 for (j = lfe_samples; j < lfe_end_sample; j++)
867 s->lfe_data[j] *= lfe_scale;
871 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
872 s->subsubframes[s->current_subframe]);
873 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
874 s->partial_samples[s->current_subframe]);
876 for (j = base_channel; j < s->prim_channels; j++) {
877 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
878 for (k = 0; k < s->subband_activity[j]; k++)
879 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
880 av_log(s->avctx, AV_LOG_DEBUG, "\n");
882 for (j = base_channel; j < s->prim_channels; j++) {
883 for (k = 0; k < s->subband_activity[j]; k++)
884 av_log(s->avctx, AV_LOG_DEBUG,
885 "prediction coefs: %f, %f, %f, %f\n",
886 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
887 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
888 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
889 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
891 for (j = base_channel; j < s->prim_channels; j++) {
892 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
893 for (k = 0; k < s->vq_start_subband[j]; k++)
894 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
895 av_log(s->avctx, AV_LOG_DEBUG, "\n");
897 for (j = base_channel; j < s->prim_channels; j++) {
898 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
899 for (k = 0; k < s->subband_activity[j]; k++)
900 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
901 av_log(s->avctx, AV_LOG_DEBUG, "\n");
903 for (j = base_channel; j < s->prim_channels; j++) {
904 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
905 for (k = 0; k < s->subband_activity[j]; k++) {
906 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
907 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
908 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
909 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
911 av_log(s->avctx, AV_LOG_DEBUG, "\n");
913 for (j = base_channel; j < s->prim_channels; j++) {
914 if (s->joint_intensity[j] > 0) {
915 int source_channel = s->joint_intensity[j] - 1;
916 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
917 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
918 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
919 av_log(s->avctx, AV_LOG_DEBUG, "\n");
922 if (!base_channel && s->prim_channels > 2 && s->downmix) {
923 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
924 for (j = 0; j < s->prim_channels; j++) {
925 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
926 dca_downmix_coeffs[s->downmix_coef[j][0]]);
927 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
928 dca_downmix_coeffs[s->downmix_coef[j][1]]);
930 av_log(s->avctx, AV_LOG_DEBUG, "\n");
932 for (j = base_channel; j < s->prim_channels; j++)
933 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
934 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
935 if (!base_channel && s->lfe) {
936 int lfe_samples = 2 * s->lfe * (4 + block_index);
937 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
939 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
940 for (j = lfe_samples; j < lfe_end_sample; j++)
941 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
942 av_log(s->avctx, AV_LOG_DEBUG, "\n");
949 static void qmf_32_subbands(DCAContext *s, int chans,
950 float samples_in[32][8], float *samples_out,
953 const float *prCoeff;
955 int sb_act = s->subband_activity[chans];
957 scale *= sqrt(1 / 8.0);
960 if (!s->multirate_inter) /* Non-perfect reconstruction */
961 prCoeff = fir_32bands_nonperfect;
962 else /* Perfect reconstruction */
963 prCoeff = fir_32bands_perfect;
965 s->dcadsp.qmf_32_subbands(samples_in, sb_act, &s->synth, &s->imdct,
966 s->subband_fir_hist[chans],
967 &s->hist_index[chans],
968 s->subband_fir_noidea[chans], prCoeff,
969 samples_out, s->raXin, scale);
972 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
973 int num_deci_sample, float *samples_in,
974 float *samples_out, float scale)
976 /* samples_in: An array holding decimated samples.
977 * Samples in current subframe starts from samples_in[0],
978 * while samples_in[-1], samples_in[-2], ..., stores samples
979 * from last subframe as history.
981 * samples_out: An array holding interpolated samples
985 const float *prCoeff;
988 /* Select decimation filter */
989 if (decimation_select == 1) {
991 prCoeff = lfe_fir_128;
994 prCoeff = lfe_fir_64;
997 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
998 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1000 samples_out += 2 * decifactor;
1004 /* downmixing routines */
1005 #define MIX_REAR1(samples, s1, rs, coef) \
1006 samples[0][i] += samples[s1][i] * coef[rs][0]; \
1007 samples[1][i] += samples[s1][i] * coef[rs][1];
1009 #define MIX_REAR2(samples, s1, s2, rs, coef) \
1010 samples[0][i] += samples[s1][i] * coef[rs][0] + samples[s2][i] * coef[rs + 1][0]; \
1011 samples[1][i] += samples[s1][i] * coef[rs][1] + samples[s2][i] * coef[rs + 1][1];
1013 #define MIX_FRONT3(samples, coef) \
1014 t = samples[c][i]; \
1015 u = samples[l][i]; \
1016 v = samples[r][i]; \
1017 samples[0][i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1018 samples[1][i] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1020 #define DOWNMIX_TO_STEREO(op1, op2) \
1021 for (i = 0; i < 256; i++) { \
1026 static void dca_downmix(float **samples, int srcfmt,
1027 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1028 const int8_t *channel_mapping)
1030 int c, l, r, sl, sr, s;
1033 float coef[DCA_PRIM_CHANNELS_MAX][2];
1035 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1036 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1037 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1043 case DCA_STEREO_TOTAL:
1044 case DCA_STEREO_SUMDIFF:
1046 av_log(NULL, 0, "Not implemented!\n");
1051 c = channel_mapping[0];
1052 l = channel_mapping[1];
1053 r = channel_mapping[2];
1054 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1057 s = channel_mapping[2];
1058 DOWNMIX_TO_STEREO(MIX_REAR1(samples, s, 2, coef), );
1061 c = channel_mapping[0];
1062 l = channel_mapping[1];
1063 r = channel_mapping[2];
1064 s = channel_mapping[3];
1065 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1066 MIX_REAR1(samples, s, 3, coef));
1069 sl = channel_mapping[2];
1070 sr = channel_mapping[3];
1071 DOWNMIX_TO_STEREO(MIX_REAR2(samples, sl, sr, 2, coef), );
1074 c = channel_mapping[0];
1075 l = channel_mapping[1];
1076 r = channel_mapping[2];
1077 sl = channel_mapping[3];
1078 sr = channel_mapping[4];
1079 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1080 MIX_REAR2(samples, sl, sr, 3, coef));
1086 #ifndef decode_blockcodes
1087 /* Very compact version of the block code decoder that does not use table
1088 * look-up but is slightly slower */
1089 static int decode_blockcode(int code, int levels, int32_t *values)
1092 int offset = (levels - 1) >> 1;
1094 for (i = 0; i < 4; i++) {
1095 int div = FASTDIV(code, levels);
1096 values[i] = code - offset - div * levels;
1103 static int decode_blockcodes(int code1, int code2, int levels, int32_t *values)
1105 return decode_blockcode(code1, levels, values) |
1106 decode_blockcode(code2, levels, values + 4);
1110 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1111 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1113 #ifndef int8x8_fmul_int32
1114 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1116 float fscale = scale / 16.0;
1118 for (i = 0; i < 8; i++)
1119 dst[i] = src[i] * fscale;
1123 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1126 int subsubframe = s->current_subsubframe;
1128 const float *quant_step_table;
1131 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1132 LOCAL_ALIGNED_16(int32_t, block, [8 * DCA_SUBBANDS]);
1138 /* Select quantization step size table */
1139 if (s->bit_rate_index == 0x1f)
1140 quant_step_table = lossless_quant_d;
1142 quant_step_table = lossy_quant_d;
1144 for (k = base_channel; k < s->prim_channels; k++) {
1145 float rscale[DCA_SUBBANDS];
1147 if (get_bits_left(&s->gb) < 0)
1148 return AVERROR_INVALIDDATA;
1150 for (l = 0; l < s->vq_start_subband[k]; l++) {
1153 /* Select the mid-tread linear quantizer */
1154 int abits = s->bitalloc[k][l];
1156 float quant_step_size = quant_step_table[abits];
1159 * Determine quantization index code book and its type
1162 /* Select quantization index code book */
1163 int sel = s->quant_index_huffman[k][abits];
1166 * Extract bits from the bit stream
1170 memset(block + 8 * l, 0, 8 * sizeof(block[0]));
1172 /* Deal with transients */
1173 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1174 rscale[l] = quant_step_size * s->scale_factor[k][l][sfi] *
1175 s->scalefactor_adj[k][sel];
1177 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1180 int block_code1, block_code2, size, levels, err;
1182 size = abits_sizes[abits - 1];
1183 levels = abits_levels[abits - 1];
1185 block_code1 = get_bits(&s->gb, size);
1186 block_code2 = get_bits(&s->gb, size);
1187 err = decode_blockcodes(block_code1, block_code2,
1188 levels, block + 8 * l);
1190 av_log(s->avctx, AV_LOG_ERROR,
1191 "ERROR: block code look-up failed\n");
1192 return AVERROR_INVALIDDATA;
1196 for (m = 0; m < 8; m++)
1197 block[8 * l + m] = get_sbits(&s->gb, abits - 3);
1201 for (m = 0; m < 8; m++)
1202 block[8 * l + m] = get_bitalloc(&s->gb,
1203 &dca_smpl_bitalloc[abits], sel);
1209 s->fmt_conv.int32_to_float_fmul_array8(&s->fmt_conv, subband_samples[k][0],
1210 block, rscale, 8 * s->vq_start_subband[k]);
1212 for (l = 0; l < s->vq_start_subband[k]; l++) {
1215 * Inverse ADPCM if in prediction mode
1217 if (s->prediction_mode[k][l]) {
1219 for (m = 0; m < 8; m++) {
1220 for (n = 1; n <= 4; n++)
1222 subband_samples[k][l][m] +=
1223 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1224 subband_samples[k][l][m - n] / 8192);
1225 else if (s->predictor_history)
1226 subband_samples[k][l][m] +=
1227 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1228 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1234 * Decode VQ encoded high frequencies
1236 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1237 /* 1 vector -> 32 samples but we only need the 8 samples
1238 * for this subsubframe. */
1239 int hfvq = s->high_freq_vq[k][l];
1241 if (!s->debug_flag & 0x01) {
1242 av_log(s->avctx, AV_LOG_DEBUG,
1243 "Stream with high frequencies VQ coding\n");
1244 s->debug_flag |= 0x01;
1247 int8x8_fmul_int32(subband_samples[k][l],
1248 &high_freq_vq[hfvq][subsubframe * 8],
1249 s->scale_factor[k][l][0]);
1253 /* Check for DSYNC after subsubframe */
1254 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1255 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1257 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1260 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1261 return AVERROR_INVALIDDATA;
1265 /* Backup predictor history for adpcm */
1266 for (k = base_channel; k < s->prim_channels; k++)
1267 for (l = 0; l < s->vq_start_subband[k]; l++)
1268 memcpy(s->subband_samples_hist[k][l],
1269 &subband_samples[k][l][4],
1270 4 * sizeof(subband_samples[0][0][0]));
1275 static int dca_filter_channels(DCAContext *s, int block_index)
1277 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1280 /* 32 subbands QMF */
1281 for (k = 0; k < s->prim_channels; k++) {
1282 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1283 0, 8388608.0, 8388608.0 };*/
1284 if (s->channel_order_tab[k] >= 0)
1285 qmf_32_subbands(s, k, subband_samples[k],
1286 s->samples_chanptr[s->channel_order_tab[k]],
1287 M_SQRT1_2 / 32768.0 /* pcm_to_double[s->source_pcm_res] */);
1291 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1292 dca_downmix(s->samples_chanptr, s->amode, s->downmix_coef, s->channel_order_tab);
1295 /* Generate LFE samples for this subsubframe FIXME!!! */
1296 if (s->output & DCA_LFE) {
1297 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1298 s->lfe_data + 2 * s->lfe * (block_index + 4),
1299 s->samples_chanptr[dca_lfe_index[s->amode]],
1300 1.0 / (256.0 * 32768.0));
1301 /* Outputs 20bits pcm samples */
1308 static int dca_subframe_footer(DCAContext *s, int base_channel)
1310 int aux_data_count = 0, i;
1313 * Unpack optional information
1316 /* presumably optional information only appears in the core? */
1317 if (!base_channel) {
1319 skip_bits_long(&s->gb, 32);
1322 aux_data_count = get_bits(&s->gb, 6);
1324 for (i = 0; i < aux_data_count; i++)
1325 get_bits(&s->gb, 8);
1327 if (s->crc_present && (s->downmix || s->dynrange))
1328 get_bits(&s->gb, 16);
1335 * Decode a dca frame block
1337 * @param s pointer to the DCAContext
1340 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1345 if (s->current_subframe >= s->subframes) {
1346 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1347 s->current_subframe, s->subframes);
1348 return AVERROR_INVALIDDATA;
1351 if (!s->current_subsubframe) {
1353 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1355 /* Read subframe header */
1356 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1360 /* Read subsubframe */
1362 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1364 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1368 s->current_subsubframe++;
1369 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1370 s->current_subsubframe = 0;
1371 s->current_subframe++;
1373 if (s->current_subframe >= s->subframes) {
1375 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1377 /* Read subframe footer */
1378 if ((ret = dca_subframe_footer(s, base_channel)))
1386 * Return the number of channels in an ExSS speaker mask (HD)
1388 static int dca_exss_mask2count(int mask)
1390 /* count bits that mean speaker pairs twice */
1391 return av_popcount(mask) +
1392 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1393 DCA_EXSS_FRONT_LEFT_RIGHT |
1394 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1395 DCA_EXSS_WIDE_LEFT_RIGHT |
1396 DCA_EXSS_SIDE_LEFT_RIGHT |
1397 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1398 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1399 DCA_EXSS_REAR_LEFT_RIGHT |
1400 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1404 * Skip mixing coefficients of a single mix out configuration (HD)
1406 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1410 for (i = 0; i < channels; i++) {
1411 int mix_map_mask = get_bits(gb, out_ch);
1412 int num_coeffs = av_popcount(mix_map_mask);
1413 skip_bits_long(gb, num_coeffs * 6);
1418 * Parse extension substream asset header (HD)
1420 static int dca_exss_parse_asset_header(DCAContext *s)
1422 int header_pos = get_bits_count(&s->gb);
1425 int embedded_stereo = 0;
1426 int embedded_6ch = 0;
1427 int drc_code_present;
1428 int extensions_mask;
1431 if (get_bits_left(&s->gb) < 16)
1434 /* We will parse just enough to get to the extensions bitmask with which
1435 * we can set the profile value. */
1437 header_size = get_bits(&s->gb, 9) + 1;
1438 skip_bits(&s->gb, 3); // asset index
1440 if (s->static_fields) {
1441 if (get_bits1(&s->gb))
1442 skip_bits(&s->gb, 4); // asset type descriptor
1443 if (get_bits1(&s->gb))
1444 skip_bits_long(&s->gb, 24); // language descriptor
1446 if (get_bits1(&s->gb)) {
1447 /* How can one fit 1024 bytes of text here if the maximum value
1448 * for the asset header size field above was 512 bytes? */
1449 int text_length = get_bits(&s->gb, 10) + 1;
1450 if (get_bits_left(&s->gb) < text_length * 8)
1452 skip_bits_long(&s->gb, text_length * 8); // info text
1455 skip_bits(&s->gb, 5); // bit resolution - 1
1456 skip_bits(&s->gb, 4); // max sample rate code
1457 channels = get_bits(&s->gb, 8) + 1;
1459 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1460 int spkr_remap_sets;
1461 int spkr_mask_size = 16;
1465 embedded_stereo = get_bits1(&s->gb);
1467 embedded_6ch = get_bits1(&s->gb);
1469 if (get_bits1(&s->gb)) {
1470 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1471 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1474 spkr_remap_sets = get_bits(&s->gb, 3);
1476 for (i = 0; i < spkr_remap_sets; i++) {
1477 /* std layout mask for each remap set */
1478 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1481 for (i = 0; i < spkr_remap_sets; i++) {
1482 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1483 if (get_bits_left(&s->gb) < 0)
1486 for (j = 0; j < num_spkrs[i]; j++) {
1487 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1488 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1489 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1494 skip_bits(&s->gb, 3); // representation type
1498 drc_code_present = get_bits1(&s->gb);
1499 if (drc_code_present)
1500 get_bits(&s->gb, 8); // drc code
1502 if (get_bits1(&s->gb))
1503 skip_bits(&s->gb, 5); // dialog normalization code
1505 if (drc_code_present && embedded_stereo)
1506 get_bits(&s->gb, 8); // drc stereo code
1508 if (s->mix_metadata && get_bits1(&s->gb)) {
1509 skip_bits(&s->gb, 1); // external mix
1510 skip_bits(&s->gb, 6); // post mix gain code
1512 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1513 skip_bits(&s->gb, 3); // drc limit
1515 skip_bits(&s->gb, 8); // custom drc code
1517 if (get_bits1(&s->gb)) // channel specific scaling
1518 for (i = 0; i < s->num_mix_configs; i++)
1519 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1521 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1523 for (i = 0; i < s->num_mix_configs; i++) {
1524 if (get_bits_left(&s->gb) < 0)
1526 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1528 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1529 if (embedded_stereo)
1530 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1534 switch (get_bits(&s->gb, 2)) {
1535 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1536 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1537 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1538 case 3: extensions_mask = 0; /* aux coding */ break;
1541 /* not parsed further, we were only interested in the extensions mask */
1543 if (get_bits_left(&s->gb) < 0)
1546 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1547 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1550 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1552 if (extensions_mask & DCA_EXT_EXSS_XLL)
1553 s->profile = FF_PROFILE_DTS_HD_MA;
1554 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1556 s->profile = FF_PROFILE_DTS_HD_HRA;
1558 if (!(extensions_mask & DCA_EXT_CORE))
1559 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1560 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1561 av_log(s->avctx, AV_LOG_WARNING,
1562 "DTS extensions detection mismatch (%d, %d)\n",
1563 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1569 * Parse extension substream header (HD)
1571 static void dca_exss_parse_header(DCAContext *s)
1577 int active_ss_mask[8];
1580 if (get_bits_left(&s->gb) < 52)
1583 skip_bits(&s->gb, 8); // user data
1584 ss_index = get_bits(&s->gb, 2);
1586 blownup = get_bits1(&s->gb);
1587 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1588 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1590 s->static_fields = get_bits1(&s->gb);
1591 if (s->static_fields) {
1592 skip_bits(&s->gb, 2); // reference clock code
1593 skip_bits(&s->gb, 3); // frame duration code
1595 if (get_bits1(&s->gb))
1596 skip_bits_long(&s->gb, 36); // timestamp
1598 /* a single stream can contain multiple audio assets that can be
1599 * combined to form multiple audio presentations */
1601 num_audiop = get_bits(&s->gb, 3) + 1;
1602 if (num_audiop > 1) {
1603 avpriv_request_sample(s->avctx,
1604 "Multiple DTS-HD audio presentations");
1605 /* ignore such streams for now */
1609 num_assets = get_bits(&s->gb, 3) + 1;
1610 if (num_assets > 1) {
1611 avpriv_request_sample(s->avctx, "Multiple DTS-HD audio assets");
1612 /* ignore such streams for now */
1616 for (i = 0; i < num_audiop; i++)
1617 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1619 for (i = 0; i < num_audiop; i++)
1620 for (j = 0; j <= ss_index; j++)
1621 if (active_ss_mask[i] & (1 << j))
1622 skip_bits(&s->gb, 8); // active asset mask
1624 s->mix_metadata = get_bits1(&s->gb);
1625 if (s->mix_metadata) {
1626 int mix_out_mask_size;
1628 skip_bits(&s->gb, 2); // adjustment level
1629 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1630 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1632 for (i = 0; i < s->num_mix_configs; i++) {
1633 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1634 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1639 for (i = 0; i < num_assets; i++)
1640 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1642 for (i = 0; i < num_assets; i++) {
1643 if (dca_exss_parse_asset_header(s))
1647 /* not parsed further, we were only interested in the extensions mask
1648 * from the asset header */
1652 * Main frame decoding function
1653 * FIXME add arguments
1655 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1656 int *got_frame_ptr, AVPacket *avpkt)
1658 AVFrame *frame = data;
1659 const uint8_t *buf = avpkt->data;
1660 int buf_size = avpkt->size;
1663 int num_core_channels = 0;
1665 float **samples_flt;
1666 DCAContext *s = avctx->priv_data;
1667 int channels, full_channels;
1673 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1674 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1675 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1676 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1677 return AVERROR_INVALIDDATA;
1680 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1681 if ((ret = dca_parse_frame_header(s)) < 0) {
1682 //seems like the frame is corrupt, try with the next one
1685 //set AVCodec values with parsed data
1686 avctx->sample_rate = s->sample_rate;
1687 avctx->bit_rate = s->bit_rate;
1689 s->profile = FF_PROFILE_DTS;
1691 for (i = 0; i < (s->sample_blocks / 8); i++) {
1692 if ((ret = dca_decode_block(s, 0, i))) {
1693 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1698 /* record number of core channels incase less than max channels are requested */
1699 num_core_channels = s->prim_channels;
1702 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1704 s->core_ext_mask = 0;
1706 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1708 /* only scan for extensions if ext_descr was unknown or indicated a
1709 * supported XCh extension */
1710 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1712 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1713 * extensions scan can fill it up */
1714 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1716 /* extensions start at 32-bit boundaries into bitstream */
1717 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1719 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1720 uint32_t bits = get_bits_long(&s->gb, 32);
1724 int ext_amode, xch_fsize;
1726 s->xch_base_channel = s->prim_channels;
1728 /* validate sync word using XCHFSIZE field */
1729 xch_fsize = show_bits(&s->gb, 10);
1730 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1731 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1734 /* skip length-to-end-of-frame field for the moment */
1735 skip_bits(&s->gb, 10);
1737 s->core_ext_mask |= DCA_EXT_XCH;
1739 /* extension amode(number of channels in extension) should be 1 */
1740 /* AFAIK XCh is not used for more channels */
1741 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1742 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1743 " supported!\n", ext_amode);
1747 /* much like core primary audio coding header */
1748 dca_parse_audio_coding_header(s, s->xch_base_channel);
1750 for (i = 0; i < (s->sample_blocks / 8); i++)
1751 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1752 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1760 /* XXCh: extended channels */
1761 /* usually found either in core or HD part in DTS-HD HRA streams,
1762 * but not in DTS-ES which contains XCh extensions instead */
1763 s->core_ext_mask |= DCA_EXT_XXCH;
1767 int fsize96 = show_bits(&s->gb, 12) + 1;
1768 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1771 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1772 get_bits_count(&s->gb));
1773 skip_bits(&s->gb, 12);
1774 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1775 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1777 s->core_ext_mask |= DCA_EXT_X96;
1782 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1785 /* no supported extensions, skip the rest of the core substream */
1786 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1789 if (s->core_ext_mask & DCA_EXT_X96)
1790 s->profile = FF_PROFILE_DTS_96_24;
1791 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1792 s->profile = FF_PROFILE_DTS_ES;
1794 /* check for ExSS (HD part) */
1795 if (s->dca_buffer_size - s->frame_size > 32 &&
1796 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1797 dca_exss_parse_header(s);
1799 avctx->profile = s->profile;
1801 full_channels = channels = s->prim_channels + !!s->lfe;
1803 if (s->amode < 16) {
1804 avctx->channel_layout = dca_core_channel_layout[s->amode];
1806 if (s->xch_present && (!avctx->request_channels ||
1807 avctx->request_channels > num_core_channels + !!s->lfe)) {
1808 avctx->channel_layout |= AV_CH_BACK_CENTER;
1810 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1811 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1813 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1816 channels = num_core_channels + !!s->lfe;
1817 s->xch_present = 0; /* disable further xch processing */
1819 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1820 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1822 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1825 if (channels > !!s->lfe &&
1826 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1827 return AVERROR_INVALIDDATA;
1829 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1831 s->output = DCA_STEREO;
1832 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1835 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1836 return AVERROR_INVALIDDATA;
1838 avctx->channels = channels;
1840 /* get output buffer */
1841 frame->nb_samples = 256 * (s->sample_blocks / 8);
1842 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
1843 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1846 samples_flt = (float **)frame->extended_data;
1848 /* allocate buffer for extra channels if downmixing */
1849 if (avctx->channels < full_channels) {
1850 ret = av_samples_get_buffer_size(NULL, full_channels - channels,
1852 avctx->sample_fmt, 0);
1856 av_fast_malloc(&s->extra_channels_buffer,
1857 &s->extra_channels_buffer_size, ret);
1858 if (!s->extra_channels_buffer)
1859 return AVERROR(ENOMEM);
1861 ret = av_samples_fill_arrays((uint8_t **)s->extra_channels, NULL,
1862 s->extra_channels_buffer,
1863 full_channels - channels,
1864 frame->nb_samples, avctx->sample_fmt, 0);
1869 /* filter to get final output */
1870 for (i = 0; i < (s->sample_blocks / 8); i++) {
1873 for (ch = 0; ch < channels; ch++)
1874 s->samples_chanptr[ch] = samples_flt[ch] + i * 256;
1875 for (; ch < full_channels; ch++)
1876 s->samples_chanptr[ch] = s->extra_channels[ch - channels] + i * 256;
1878 dca_filter_channels(s, i);
1880 /* If this was marked as a DTS-ES stream we need to subtract back- */
1881 /* channel from SL & SR to remove matrixed back-channel signal */
1882 if ((s->source_pcm_res & 1) && s->xch_present) {
1883 float *back_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel]];
1884 float *lt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 2]];
1885 float *rt_chan = s->samples_chanptr[s->channel_order_tab[s->xch_base_channel - 1]];
1886 s->fdsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1887 s->fdsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1891 /* update lfe history */
1892 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1893 for (i = 0; i < 2 * s->lfe * 4; i++)
1894 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1904 * DCA initialization
1906 * @param avctx pointer to the AVCodecContext
1909 static av_cold int dca_decode_init(AVCodecContext *avctx)
1911 DCAContext *s = avctx->priv_data;
1916 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
1917 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1918 ff_synth_filter_init(&s->synth);
1919 ff_dcadsp_init(&s->dcadsp);
1920 ff_fmt_convert_init(&s->fmt_conv, avctx);
1922 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
1924 /* allow downmixing to stereo */
1925 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1926 avctx->request_channels == 2) {
1927 avctx->channels = avctx->request_channels;
1933 static av_cold int dca_decode_end(AVCodecContext *avctx)
1935 DCAContext *s = avctx->priv_data;
1936 ff_mdct_end(&s->imdct);
1937 av_freep(&s->extra_channels_buffer);
1941 static const AVProfile profiles[] = {
1942 { FF_PROFILE_DTS, "DTS" },
1943 { FF_PROFILE_DTS_ES, "DTS-ES" },
1944 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1945 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1946 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1947 { FF_PROFILE_UNKNOWN },
1950 AVCodec ff_dca_decoder = {
1952 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1953 .type = AVMEDIA_TYPE_AUDIO,
1954 .id = AV_CODEC_ID_DTS,
1955 .priv_data_size = sizeof(DCAContext),
1956 .init = dca_decode_init,
1957 .decode = dca_decode_frame,
1958 .close = dca_decode_end,
1959 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1960 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
1961 AV_SAMPLE_FMT_NONE },
1962 .profiles = NULL_IF_CONFIG_SMALL(profiles),