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 FFmpeg.
10 * FFmpeg 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 * FFmpeg 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 FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 #include "libavutil/common.h"
30 #include "libavutil/intmath.h"
31 #include "libavutil/intreadwrite.h"
32 #include "libavutil/mathematics.h"
33 #include "libavutil/audioconvert.h"
42 #include "dca_parser.h"
43 #include "synth_filter.h"
45 #include "fmtconvert.h"
53 #define DCA_PRIM_CHANNELS_MAX (7)
54 #define DCA_SUBBANDS (32)
55 #define DCA_ABITS_MAX (32) /* Should be 28 */
56 #define DCA_SUBSUBFRAMES_MAX (4)
57 #define DCA_SUBFRAMES_MAX (16)
58 #define DCA_BLOCKS_MAX (16)
59 #define DCA_LFE_MAX (3)
75 /* these are unconfirmed but should be mostly correct */
76 enum DCAExSSSpeakerMask {
77 DCA_EXSS_FRONT_CENTER = 0x0001,
78 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
79 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
80 DCA_EXSS_LFE = 0x0008,
81 DCA_EXSS_REAR_CENTER = 0x0010,
82 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
83 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
84 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
85 DCA_EXSS_OVERHEAD = 0x0100,
86 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
87 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
88 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
89 DCA_EXSS_LFE2 = 0x1000,
90 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
91 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
92 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
95 enum DCAExtensionMask {
96 DCA_EXT_CORE = 0x001, ///< core in core substream
97 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
98 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
99 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
100 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
101 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
102 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
103 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
104 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
105 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
108 /* -1 are reserved or unknown */
109 static const int dca_ext_audio_descr_mask[] = {
113 DCA_EXT_XCH | DCA_EXT_X96,
120 /* extensions that reside in core substream */
121 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
123 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
124 * Some compromises have been made for special configurations. Most configurations
125 * are never used so complete accuracy is not needed.
127 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
128 * S -> side, when both rear and back are configured move one of them to the side channel
130 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
132 static const uint64_t dca_core_channel_layout[] = {
133 AV_CH_FRONT_CENTER, ///< 1, A
134 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
135 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
136 AV_CH_LAYOUT_STEREO, ///< 2, (L + R) + (L - R) (sum-difference)
137 AV_CH_LAYOUT_STEREO, ///< 2, LT + RT (left and right total)
138 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER, ///< 3, C + L + R
139 AV_CH_LAYOUT_STEREO | AV_CH_BACK_CENTER, ///< 3, L + R + S
140 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 4, C + L + R + S
141 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 4, L + R + SL + SR
143 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
144 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
146 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
147 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
149 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
150 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
152 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
153 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
154 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
156 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
157 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
158 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
160 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
161 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
162 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
164 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
165 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
166 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
169 static const int8_t dca_lfe_index[] = {
170 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
173 static const int8_t dca_channel_reorder_lfe[][9] = {
174 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
175 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
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 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
180 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
181 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
182 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
183 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
184 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
185 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
186 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
187 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
188 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
189 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
192 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
193 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
194 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
195 { 0, 1, 3, -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 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
199 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
200 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
201 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
202 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
203 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
204 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
205 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
206 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
207 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
208 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
211 static const int8_t dca_channel_reorder_nolfe[][9] = {
212 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
213 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
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 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
218 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
219 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
220 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
221 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
222 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
223 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
224 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
225 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
226 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
227 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
230 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
231 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
232 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
233 { 0, 1, 2, -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 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
237 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
238 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
239 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
240 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
241 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
242 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
243 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
244 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
245 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
246 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
249 #define DCA_DOLBY 101 /* FIXME */
251 #define DCA_CHANNEL_BITS 6
252 #define DCA_CHANNEL_MASK 0x3F
256 #define HEADER_SIZE 14
258 #define DCA_MAX_FRAME_SIZE 16384
259 #define DCA_MAX_EXSS_HEADER_SIZE 4096
261 #define DCA_BUFFER_PADDING_SIZE 1024
263 /** Bit allocation */
265 int offset; ///< code values offset
266 int maxbits[8]; ///< max bits in VLC
267 int wrap; ///< wrap for get_vlc2()
268 VLC vlc[8]; ///< actual codes
271 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
272 static BitAlloc dca_tmode; ///< transition mode VLCs
273 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
274 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
276 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
279 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
284 AVCodecContext *avctx;
287 int frame_type; ///< type of the current frame
288 int samples_deficit; ///< deficit sample count
289 int crc_present; ///< crc is present in the bitstream
290 int sample_blocks; ///< number of PCM sample blocks
291 int frame_size; ///< primary frame byte size
292 int amode; ///< audio channels arrangement
293 int sample_rate; ///< audio sampling rate
294 int bit_rate; ///< transmission bit rate
295 int bit_rate_index; ///< transmission bit rate index
297 int downmix; ///< embedded downmix enabled
298 int dynrange; ///< embedded dynamic range flag
299 int timestamp; ///< embedded time stamp flag
300 int aux_data; ///< auxiliary data flag
301 int hdcd; ///< source material is mastered in HDCD
302 int ext_descr; ///< extension audio descriptor flag
303 int ext_coding; ///< extended coding flag
304 int aspf; ///< audio sync word insertion flag
305 int lfe; ///< low frequency effects flag
306 int predictor_history; ///< predictor history flag
307 int header_crc; ///< header crc check bytes
308 int multirate_inter; ///< multirate interpolator switch
309 int version; ///< encoder software revision
310 int copy_history; ///< copy history
311 int source_pcm_res; ///< source pcm resolution
312 int front_sum; ///< front sum/difference flag
313 int surround_sum; ///< surround sum/difference flag
314 int dialog_norm; ///< dialog normalisation parameter
316 /* Primary audio coding header */
317 int subframes; ///< number of subframes
318 int total_channels; ///< number of channels including extensions
319 int prim_channels; ///< number of primary audio channels
320 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
321 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
322 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
323 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
324 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
325 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
326 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
327 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
329 /* Primary audio coding side information */
330 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
331 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
332 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
333 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
334 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
335 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
336 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
337 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
338 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
339 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
340 int dynrange_coef; ///< dynamic range coefficient
342 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
344 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
345 int lfe_scale_factor;
347 /* Subband samples history (for ADPCM) */
348 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
349 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
350 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
351 int hist_index[DCA_PRIM_CHANNELS_MAX];
352 DECLARE_ALIGNED(32, float, raXin)[32];
354 int output; ///< type of output
355 float scale_bias; ///< output scale
357 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
358 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX + 1) * 256];
359 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
361 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
362 int dca_buffer_size; ///< how much data is in the dca_buffer
364 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
366 /* Current position in DCA frame */
367 int current_subframe;
368 int current_subsubframe;
370 int core_ext_mask; ///< present extensions in the core substream
372 /* XCh extension information */
373 int xch_present; ///< XCh extension present and valid
374 int xch_base_channel; ///< index of first (only) channel containing XCH data
376 /* ExSS header parser */
377 int static_fields; ///< static fields present
378 int mix_metadata; ///< mixing metadata present
379 int num_mix_configs; ///< number of mix out configurations
380 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
384 int debug_flag; ///< used for suppressing repeated error messages output
387 SynthFilterContext synth;
388 DCADSPContext dcadsp;
389 FmtConvertContext fmt_conv;
392 static const uint16_t dca_vlc_offs[] = {
393 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
394 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
395 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
396 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
397 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
398 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
401 static av_cold void dca_init_vlcs(void)
403 static int vlcs_initialized = 0;
405 static VLC_TYPE dca_table[23622][2];
407 if (vlcs_initialized)
410 dca_bitalloc_index.offset = 1;
411 dca_bitalloc_index.wrap = 2;
412 for (i = 0; i < 5; i++) {
413 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
414 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
415 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
416 bitalloc_12_bits[i], 1, 1,
417 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
419 dca_scalefactor.offset = -64;
420 dca_scalefactor.wrap = 2;
421 for (i = 0; i < 5; i++) {
422 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
423 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
424 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
425 scales_bits[i], 1, 1,
426 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
428 dca_tmode.offset = 0;
430 for (i = 0; i < 4; i++) {
431 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
432 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
433 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
435 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
438 for (i = 0; i < 10; i++)
439 for (j = 0; j < 7; j++) {
440 if (!bitalloc_codes[i][j])
442 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
443 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
444 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
445 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
447 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
449 bitalloc_bits[i][j], 1, 1,
450 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
453 vlcs_initialized = 1;
456 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
459 *dst++ = get_bits(gb, bits);
462 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
465 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
466 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
467 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
469 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
470 s->prim_channels = s->total_channels;
472 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
473 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
476 for (i = base_channel; i < s->prim_channels; i++) {
477 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
478 if (s->subband_activity[i] > DCA_SUBBANDS)
479 s->subband_activity[i] = DCA_SUBBANDS;
481 for (i = base_channel; i < s->prim_channels; i++) {
482 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
483 if (s->vq_start_subband[i] > DCA_SUBBANDS)
484 s->vq_start_subband[i] = DCA_SUBBANDS;
486 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
487 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
488 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
489 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
491 /* Get codebooks quantization indexes */
493 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
494 for (j = 1; j < 11; j++)
495 for (i = base_channel; i < s->prim_channels; i++)
496 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
498 /* Get scale factor adjustment */
499 for (j = 0; j < 11; j++)
500 for (i = base_channel; i < s->prim_channels; i++)
501 s->scalefactor_adj[i][j] = 1;
503 for (j = 1; j < 11; j++)
504 for (i = base_channel; i < s->prim_channels; i++)
505 if (s->quant_index_huffman[i][j] < thr[j])
506 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
508 if (s->crc_present) {
509 /* Audio header CRC check */
510 get_bits(&s->gb, 16);
513 s->current_subframe = 0;
514 s->current_subsubframe = 0;
517 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
518 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
519 for (i = base_channel; i < s->prim_channels; i++) {
520 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
521 s->subband_activity[i]);
522 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
523 s->vq_start_subband[i]);
524 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
525 s->joint_intensity[i]);
526 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
527 s->transient_huffman[i]);
528 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
529 s->scalefactor_huffman[i]);
530 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
531 s->bitalloc_huffman[i]);
532 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
533 for (j = 0; j < 11; j++)
534 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
535 av_log(s->avctx, AV_LOG_DEBUG, "\n");
536 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
537 for (j = 0; j < 11; j++)
538 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
539 av_log(s->avctx, AV_LOG_DEBUG, "\n");
546 static int dca_parse_frame_header(DCAContext *s)
548 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
551 skip_bits_long(&s->gb, 32);
554 s->frame_type = get_bits(&s->gb, 1);
555 s->samples_deficit = get_bits(&s->gb, 5) + 1;
556 s->crc_present = get_bits(&s->gb, 1);
557 s->sample_blocks = get_bits(&s->gb, 7) + 1;
558 s->frame_size = get_bits(&s->gb, 14) + 1;
559 if (s->frame_size < 95)
560 return AVERROR_INVALIDDATA;
561 s->amode = get_bits(&s->gb, 6);
562 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
564 return AVERROR_INVALIDDATA;
565 s->bit_rate_index = get_bits(&s->gb, 5);
566 s->bit_rate = dca_bit_rates[s->bit_rate_index];
568 return AVERROR_INVALIDDATA;
570 s->downmix = get_bits(&s->gb, 1);
571 s->dynrange = get_bits(&s->gb, 1);
572 s->timestamp = get_bits(&s->gb, 1);
573 s->aux_data = get_bits(&s->gb, 1);
574 s->hdcd = get_bits(&s->gb, 1);
575 s->ext_descr = get_bits(&s->gb, 3);
576 s->ext_coding = get_bits(&s->gb, 1);
577 s->aspf = get_bits(&s->gb, 1);
578 s->lfe = get_bits(&s->gb, 2);
579 s->predictor_history = get_bits(&s->gb, 1);
581 /* TODO: check CRC */
583 s->header_crc = get_bits(&s->gb, 16);
585 s->multirate_inter = get_bits(&s->gb, 1);
586 s->version = get_bits(&s->gb, 4);
587 s->copy_history = get_bits(&s->gb, 2);
588 s->source_pcm_res = get_bits(&s->gb, 3);
589 s->front_sum = get_bits(&s->gb, 1);
590 s->surround_sum = get_bits(&s->gb, 1);
591 s->dialog_norm = get_bits(&s->gb, 4);
593 /* FIXME: channels mixing levels */
594 s->output = s->amode;
596 s->output |= DCA_LFE;
599 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
600 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
601 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
602 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
603 s->sample_blocks, s->sample_blocks * 32);
604 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
605 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
606 s->amode, dca_channels[s->amode]);
607 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
609 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
611 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
612 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
613 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
614 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
615 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
616 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
617 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
618 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
619 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
620 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
621 s->predictor_history);
622 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
623 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
625 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
626 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
627 av_log(s->avctx, AV_LOG_DEBUG,
628 "source pcm resolution: %i (%i bits/sample)\n",
629 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
630 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
631 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
632 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
633 av_log(s->avctx, AV_LOG_DEBUG, "\n");
636 /* Primary audio coding header */
637 s->subframes = get_bits(&s->gb, 4) + 1;
639 return dca_parse_audio_coding_header(s, 0);
643 static inline int get_scale(GetBitContext *gb, int level, int value, int log2range)
646 /* huffman encoded */
647 value += get_bitalloc(gb, &dca_scalefactor, level);
648 value = av_clip(value, 0, (1 << log2range) - 1);
649 } else if (level < 8) {
650 if (level + 1 > log2range) {
651 skip_bits(gb, level + 1 - log2range);
652 value = get_bits(gb, log2range);
654 value = get_bits(gb, level + 1);
660 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
662 /* Primary audio coding side information */
665 if (get_bits_left(&s->gb) < 0)
666 return AVERROR_INVALIDDATA;
669 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
670 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
673 for (j = base_channel; j < s->prim_channels; j++) {
674 for (k = 0; k < s->subband_activity[j]; k++)
675 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
678 /* Get prediction codebook */
679 for (j = base_channel; j < s->prim_channels; j++) {
680 for (k = 0; k < s->subband_activity[j]; k++) {
681 if (s->prediction_mode[j][k] > 0) {
682 /* (Prediction coefficient VQ address) */
683 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
688 /* Bit allocation index */
689 for (j = base_channel; j < s->prim_channels; j++) {
690 for (k = 0; k < s->vq_start_subband[j]; k++) {
691 if (s->bitalloc_huffman[j] == 6)
692 s->bitalloc[j][k] = get_bits(&s->gb, 5);
693 else if (s->bitalloc_huffman[j] == 5)
694 s->bitalloc[j][k] = get_bits(&s->gb, 4);
695 else if (s->bitalloc_huffman[j] == 7) {
696 av_log(s->avctx, AV_LOG_ERROR,
697 "Invalid bit allocation index\n");
698 return AVERROR_INVALIDDATA;
701 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
704 if (s->bitalloc[j][k] > 26) {
705 // av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index [%i][%i] too big (%i)\n",
706 // j, k, s->bitalloc[j][k]);
707 return AVERROR_INVALIDDATA;
712 /* Transition mode */
713 for (j = base_channel; j < s->prim_channels; j++) {
714 for (k = 0; k < s->subband_activity[j]; k++) {
715 s->transition_mode[j][k] = 0;
716 if (s->subsubframes[s->current_subframe] > 1 &&
717 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
718 s->transition_mode[j][k] =
719 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
724 if (get_bits_left(&s->gb) < 0)
725 return AVERROR_INVALIDDATA;
727 for (j = base_channel; j < s->prim_channels; j++) {
728 const uint32_t *scale_table;
729 int scale_sum, log_size;
731 memset(s->scale_factor[j], 0,
732 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
734 if (s->scalefactor_huffman[j] == 6) {
735 scale_table = scale_factor_quant7;
738 scale_table = scale_factor_quant6;
742 /* When huffman coded, only the difference is encoded */
745 for (k = 0; k < s->subband_activity[j]; k++) {
746 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
747 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
748 s->scale_factor[j][k][0] = scale_table[scale_sum];
751 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
752 /* Get second scale factor */
753 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum, log_size);
754 s->scale_factor[j][k][1] = scale_table[scale_sum];
759 /* Joint subband scale factor codebook select */
760 for (j = base_channel; j < s->prim_channels; j++) {
761 /* Transmitted only if joint subband coding enabled */
762 if (s->joint_intensity[j] > 0)
763 s->joint_huff[j] = get_bits(&s->gb, 3);
766 if (get_bits_left(&s->gb) < 0)
767 return AVERROR_INVALIDDATA;
769 /* Scale factors for joint subband coding */
770 for (j = base_channel; j < s->prim_channels; j++) {
773 /* Transmitted only if joint subband coding enabled */
774 if (s->joint_intensity[j] > 0) {
776 source_channel = s->joint_intensity[j] - 1;
778 /* When huffman coded, only the difference is encoded
779 * (is this valid as well for joint scales ???) */
781 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
782 scale = get_scale(&s->gb, s->joint_huff[j], 64 /* bias */, 7);
783 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
786 if (!(s->debug_flag & 0x02)) {
787 av_log(s->avctx, AV_LOG_DEBUG,
788 "Joint stereo coding not supported\n");
789 s->debug_flag |= 0x02;
794 /* Stereo downmix coefficients */
795 if (!base_channel && s->prim_channels > 2) {
797 for (j = base_channel; j < s->prim_channels; j++) {
798 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
799 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
802 int am = s->amode & DCA_CHANNEL_MASK;
803 if (am >= FF_ARRAY_ELEMS(dca_default_coeffs)) {
804 av_log(s->avctx, AV_LOG_ERROR,
805 "Invalid channel mode %d\n", am);
806 return AVERROR_INVALIDDATA;
808 for (j = base_channel; j < s->prim_channels; j++) {
809 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
810 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
815 /* Dynamic range coefficient */
816 if (!base_channel && s->dynrange)
817 s->dynrange_coef = get_bits(&s->gb, 8);
819 /* Side information CRC check word */
820 if (s->crc_present) {
821 get_bits(&s->gb, 16);
825 * Primary audio data arrays
828 /* VQ encoded high frequency subbands */
829 for (j = base_channel; j < s->prim_channels; j++)
830 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
831 /* 1 vector -> 32 samples */
832 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
834 /* Low frequency effect data */
835 if (!base_channel && s->lfe) {
838 int lfe_samples = 2 * s->lfe * (4 + block_index);
839 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
842 for (j = lfe_samples; j < lfe_end_sample; j++) {
843 /* Signed 8 bits int */
844 s->lfe_data[j] = get_sbits(&s->gb, 8);
847 /* Scale factor index */
848 quant7 = get_bits(&s->gb, 8);
850 av_log_ask_for_sample(s->avctx, "LFEScaleIndex larger than 127\n");
851 return AVERROR_INVALIDDATA;
853 s->lfe_scale_factor = scale_factor_quant7[quant7];
855 /* Quantization step size * scale factor */
856 lfe_scale = 0.035 * s->lfe_scale_factor;
858 for (j = lfe_samples; j < lfe_end_sample; j++)
859 s->lfe_data[j] *= lfe_scale;
863 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
864 s->subsubframes[s->current_subframe]);
865 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
866 s->partial_samples[s->current_subframe]);
868 for (j = base_channel; j < s->prim_channels; j++) {
869 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
870 for (k = 0; k < s->subband_activity[j]; k++)
871 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
872 av_log(s->avctx, AV_LOG_DEBUG, "\n");
874 for (j = base_channel; j < s->prim_channels; j++) {
875 for (k = 0; k < s->subband_activity[j]; k++)
876 av_log(s->avctx, AV_LOG_DEBUG,
877 "prediction coefs: %f, %f, %f, %f\n",
878 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
879 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
880 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
881 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
883 for (j = base_channel; j < s->prim_channels; j++) {
884 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
885 for (k = 0; k < s->vq_start_subband[j]; k++)
886 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
887 av_log(s->avctx, AV_LOG_DEBUG, "\n");
889 for (j = base_channel; j < s->prim_channels; j++) {
890 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
891 for (k = 0; k < s->subband_activity[j]; k++)
892 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
893 av_log(s->avctx, AV_LOG_DEBUG, "\n");
895 for (j = base_channel; j < s->prim_channels; j++) {
896 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
897 for (k = 0; k < s->subband_activity[j]; k++) {
898 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
899 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
900 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
901 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
903 av_log(s->avctx, AV_LOG_DEBUG, "\n");
905 for (j = base_channel; j < s->prim_channels; j++) {
906 if (s->joint_intensity[j] > 0) {
907 int source_channel = s->joint_intensity[j] - 1;
908 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
909 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
910 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
911 av_log(s->avctx, AV_LOG_DEBUG, "\n");
914 if (!base_channel && s->prim_channels > 2 && s->downmix) {
915 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
916 for (j = 0; j < s->prim_channels; j++) {
917 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
918 dca_downmix_coeffs[s->downmix_coef[j][0]]);
919 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
920 dca_downmix_coeffs[s->downmix_coef[j][1]]);
922 av_log(s->avctx, AV_LOG_DEBUG, "\n");
924 for (j = base_channel; j < s->prim_channels; j++)
925 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
926 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
927 if (!base_channel && s->lfe) {
928 int lfe_samples = 2 * s->lfe * (4 + block_index);
929 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
931 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
932 for (j = lfe_samples; j < lfe_end_sample; j++)
933 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
934 av_log(s->avctx, AV_LOG_DEBUG, "\n");
941 static void qmf_32_subbands(DCAContext *s, int chans,
942 float samples_in[32][8], float *samples_out,
945 const float *prCoeff;
948 int sb_act = s->subband_activity[chans];
951 scale *= sqrt(1 / 8.0);
954 if (!s->multirate_inter) /* Non-perfect reconstruction */
955 prCoeff = fir_32bands_nonperfect;
956 else /* Perfect reconstruction */
957 prCoeff = fir_32bands_perfect;
959 for (i = sb_act; i < 32; i++)
962 /* Reconstructed channel sample index */
963 for (subindex = 0; subindex < 8; subindex++) {
964 /* Load in one sample from each subband and clear inactive subbands */
965 for (i = 0; i < sb_act; i++) {
966 unsigned sign = (i - 1) & 2;
967 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
968 AV_WN32A(&s->raXin[i], v);
971 s->synth.synth_filter_float(&s->imdct,
972 s->subband_fir_hist[chans],
973 &s->hist_index[chans],
974 s->subband_fir_noidea[chans], prCoeff,
975 samples_out, s->raXin, scale);
980 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
981 int num_deci_sample, float *samples_in,
982 float *samples_out, float scale)
984 /* samples_in: An array holding decimated samples.
985 * Samples in current subframe starts from samples_in[0],
986 * while samples_in[-1], samples_in[-2], ..., stores samples
987 * from last subframe as history.
989 * samples_out: An array holding interpolated samples
993 const float *prCoeff;
996 /* Select decimation filter */
997 if (decimation_select == 1) {
999 prCoeff = lfe_fir_128;
1002 prCoeff = lfe_fir_64;
1005 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
1006 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
1008 samples_out += 2 * decifactor;
1012 /* downmixing routines */
1013 #define MIX_REAR1(samples, si1, rs, coef) \
1014 samples[i] += samples[si1] * coef[rs][0]; \
1015 samples[i+256] += samples[si1] * coef[rs][1];
1017 #define MIX_REAR2(samples, si1, si2, rs, coef) \
1018 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs + 1][0]; \
1019 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs + 1][1];
1021 #define MIX_FRONT3(samples, coef) \
1022 t = samples[i + c]; \
1023 u = samples[i + l]; \
1024 v = samples[i + r]; \
1025 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1026 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1028 #define DOWNMIX_TO_STEREO(op1, op2) \
1029 for (i = 0; i < 256; i++) { \
1034 static void dca_downmix(float *samples, int srcfmt,
1035 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1036 const int8_t *channel_mapping)
1038 int c, l, r, sl, sr, s;
1041 float coef[DCA_PRIM_CHANNELS_MAX][2];
1043 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1044 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1045 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1051 case DCA_STEREO_TOTAL:
1052 case DCA_STEREO_SUMDIFF:
1054 av_log(NULL, 0, "Not implemented!\n");
1059 c = channel_mapping[0] * 256;
1060 l = channel_mapping[1] * 256;
1061 r = channel_mapping[2] * 256;
1062 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1065 s = channel_mapping[2] * 256;
1066 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef), );
1069 c = channel_mapping[0] * 256;
1070 l = channel_mapping[1] * 256;
1071 r = channel_mapping[2] * 256;
1072 s = channel_mapping[3] * 256;
1073 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1074 MIX_REAR1(samples, i + s, 3, coef));
1077 sl = channel_mapping[2] * 256;
1078 sr = channel_mapping[3] * 256;
1079 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef), );
1082 c = channel_mapping[0] * 256;
1083 l = channel_mapping[1] * 256;
1084 r = channel_mapping[2] * 256;
1085 sl = channel_mapping[3] * 256;
1086 sr = channel_mapping[4] * 256;
1087 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1088 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1094 #ifndef decode_blockcodes
1095 /* Very compact version of the block code decoder that does not use table
1096 * look-up but is slightly slower */
1097 static int decode_blockcode(int code, int levels, int *values)
1100 int offset = (levels - 1) >> 1;
1102 for (i = 0; i < 4; i++) {
1103 int div = FASTDIV(code, levels);
1104 values[i] = code - offset - div * levels;
1111 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1113 return decode_blockcode(code1, levels, values) |
1114 decode_blockcode(code2, levels, values + 4);
1118 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1119 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1121 #ifndef int8x8_fmul_int32
1122 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1124 float fscale = scale / 16.0;
1126 for (i = 0; i < 8; i++)
1127 dst[i] = src[i] * fscale;
1131 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1134 int subsubframe = s->current_subsubframe;
1136 const float *quant_step_table;
1139 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1140 LOCAL_ALIGNED_16(int, block, [8]);
1146 /* Select quantization step size table */
1147 if (s->bit_rate_index == 0x1f)
1148 quant_step_table = lossless_quant_d;
1150 quant_step_table = lossy_quant_d;
1152 for (k = base_channel; k < s->prim_channels; k++) {
1153 if (get_bits_left(&s->gb) < 0)
1154 return AVERROR_INVALIDDATA;
1156 for (l = 0; l < s->vq_start_subband[k]; l++) {
1159 /* Select the mid-tread linear quantizer */
1160 int abits = s->bitalloc[k][l];
1162 float quant_step_size = quant_step_table[abits];
1165 * Determine quantization index code book and its type
1168 /* Select quantization index code book */
1169 int sel = s->quant_index_huffman[k][abits];
1172 * Extract bits from the bit stream
1175 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1177 /* Deal with transients */
1178 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1179 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1180 s->scalefactor_adj[k][sel];
1182 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1185 int block_code1, block_code2, size, levels, err;
1187 size = abits_sizes[abits - 1];
1188 levels = abits_levels[abits - 1];
1190 block_code1 = get_bits(&s->gb, size);
1191 block_code2 = get_bits(&s->gb, size);
1192 err = decode_blockcodes(block_code1, block_code2,
1195 av_log(s->avctx, AV_LOG_ERROR,
1196 "ERROR: block code look-up failed\n");
1197 return AVERROR_INVALIDDATA;
1201 for (m = 0; m < 8; m++)
1202 block[m] = get_sbits(&s->gb, abits - 3);
1206 for (m = 0; m < 8; m++)
1207 block[m] = get_bitalloc(&s->gb,
1208 &dca_smpl_bitalloc[abits], sel);
1211 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1216 * Inverse ADPCM if in prediction mode
1218 if (s->prediction_mode[k][l]) {
1220 for (m = 0; m < 8; m++) {
1221 for (n = 1; n <= 4; n++)
1223 subband_samples[k][l][m] +=
1224 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1225 subband_samples[k][l][m - n] / 8192);
1226 else if (s->predictor_history)
1227 subband_samples[k][l][m] +=
1228 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1229 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1235 * Decode VQ encoded high frequencies
1237 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1238 /* 1 vector -> 32 samples but we only need the 8 samples
1239 * for this subsubframe. */
1240 int hfvq = s->high_freq_vq[k][l];
1242 if (!s->debug_flag & 0x01) {
1243 av_log(s->avctx, AV_LOG_DEBUG,
1244 "Stream with high frequencies VQ coding\n");
1245 s->debug_flag |= 0x01;
1248 int8x8_fmul_int32(subband_samples[k][l],
1249 &high_freq_vq[hfvq][subsubframe * 8],
1250 s->scale_factor[k][l][0]);
1254 /* Check for DSYNC after subsubframe */
1255 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1256 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1258 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1261 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
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 qmf_32_subbands(s, k, subband_samples[k],
1285 &s->samples[256 * s->channel_order_tab[k]],
1286 M_SQRT1_2 * s->scale_bias /* pcm_to_double[s->source_pcm_res] */);
1290 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1291 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1294 /* Generate LFE samples for this subsubframe FIXME!!! */
1295 if (s->output & DCA_LFE) {
1296 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1297 s->lfe_data + 2 * s->lfe * (block_index + 4),
1298 &s->samples[256 * dca_lfe_index[s->amode]],
1299 (1.0 / 256.0) * s->scale_bias);
1300 /* Outputs 20bits pcm samples */
1307 static int dca_subframe_footer(DCAContext *s, int base_channel)
1309 int aux_data_count = 0, i;
1312 * Unpack optional information
1315 /* presumably optional information only appears in the core? */
1316 if (!base_channel) {
1318 skip_bits_long(&s->gb, 32);
1321 aux_data_count = get_bits(&s->gb, 6);
1323 for (i = 0; i < aux_data_count; i++)
1324 get_bits(&s->gb, 8);
1326 if (s->crc_present && (s->downmix || s->dynrange))
1327 get_bits(&s->gb, 16);
1334 * Decode a dca frame block
1336 * @param s pointer to the DCAContext
1339 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1344 if (s->current_subframe >= s->subframes) {
1345 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1346 s->current_subframe, s->subframes);
1347 return AVERROR_INVALIDDATA;
1350 if (!s->current_subsubframe) {
1352 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1354 /* Read subframe header */
1355 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1359 /* Read subsubframe */
1361 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1363 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1367 s->current_subsubframe++;
1368 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1369 s->current_subsubframe = 0;
1370 s->current_subframe++;
1372 if (s->current_subframe >= s->subframes) {
1374 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1376 /* Read subframe footer */
1377 if ((ret = dca_subframe_footer(s, base_channel)))
1385 * Return the number of channels in an ExSS speaker mask (HD)
1387 static int dca_exss_mask2count(int mask)
1389 /* count bits that mean speaker pairs twice */
1390 return av_popcount(mask) +
1391 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1392 DCA_EXSS_FRONT_LEFT_RIGHT |
1393 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1394 DCA_EXSS_WIDE_LEFT_RIGHT |
1395 DCA_EXSS_SIDE_LEFT_RIGHT |
1396 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1397 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1398 DCA_EXSS_REAR_LEFT_RIGHT |
1399 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1403 * Skip mixing coefficients of a single mix out configuration (HD)
1405 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1409 for (i = 0; i < channels; i++) {
1410 int mix_map_mask = get_bits(gb, out_ch);
1411 int num_coeffs = av_popcount(mix_map_mask);
1412 skip_bits_long(gb, num_coeffs * 6);
1417 * Parse extension substream asset header (HD)
1419 static int dca_exss_parse_asset_header(DCAContext *s)
1421 int header_pos = get_bits_count(&s->gb);
1424 int embedded_stereo = 0;
1425 int embedded_6ch = 0;
1426 int drc_code_present;
1427 int av_uninit(extensions_mask);
1430 if (get_bits_left(&s->gb) < 16)
1433 /* We will parse just enough to get to the extensions bitmask with which
1434 * we can set the profile value. */
1436 header_size = get_bits(&s->gb, 9) + 1;
1437 skip_bits(&s->gb, 3); // asset index
1439 if (s->static_fields) {
1440 if (get_bits1(&s->gb))
1441 skip_bits(&s->gb, 4); // asset type descriptor
1442 if (get_bits1(&s->gb))
1443 skip_bits_long(&s->gb, 24); // language descriptor
1445 if (get_bits1(&s->gb)) {
1446 /* How can one fit 1024 bytes of text here if the maximum value
1447 * for the asset header size field above was 512 bytes? */
1448 int text_length = get_bits(&s->gb, 10) + 1;
1449 if (get_bits_left(&s->gb) < text_length * 8)
1451 skip_bits_long(&s->gb, text_length * 8); // info text
1454 skip_bits(&s->gb, 5); // bit resolution - 1
1455 skip_bits(&s->gb, 4); // max sample rate code
1456 channels = get_bits(&s->gb, 8) + 1;
1458 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1459 int spkr_remap_sets;
1460 int spkr_mask_size = 16;
1464 embedded_stereo = get_bits1(&s->gb);
1466 embedded_6ch = get_bits1(&s->gb);
1468 if (get_bits1(&s->gb)) {
1469 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1470 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1473 spkr_remap_sets = get_bits(&s->gb, 3);
1475 for (i = 0; i < spkr_remap_sets; i++) {
1476 /* std layout mask for each remap set */
1477 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1480 for (i = 0; i < spkr_remap_sets; i++) {
1481 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1482 if (get_bits_left(&s->gb) < 0)
1485 for (j = 0; j < num_spkrs[i]; j++) {
1486 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1487 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1488 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1493 skip_bits(&s->gb, 3); // representation type
1497 drc_code_present = get_bits1(&s->gb);
1498 if (drc_code_present)
1499 get_bits(&s->gb, 8); // drc code
1501 if (get_bits1(&s->gb))
1502 skip_bits(&s->gb, 5); // dialog normalization code
1504 if (drc_code_present && embedded_stereo)
1505 get_bits(&s->gb, 8); // drc stereo code
1507 if (s->mix_metadata && get_bits1(&s->gb)) {
1508 skip_bits(&s->gb, 1); // external mix
1509 skip_bits(&s->gb, 6); // post mix gain code
1511 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1512 skip_bits(&s->gb, 3); // drc limit
1514 skip_bits(&s->gb, 8); // custom drc code
1516 if (get_bits1(&s->gb)) // channel specific scaling
1517 for (i = 0; i < s->num_mix_configs; i++)
1518 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1520 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1522 for (i = 0; i < s->num_mix_configs; i++) {
1523 if (get_bits_left(&s->gb) < 0)
1525 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1527 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1528 if (embedded_stereo)
1529 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1533 switch (get_bits(&s->gb, 2)) {
1534 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1535 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1536 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1537 case 3: extensions_mask = 0; /* aux coding */ break;
1540 /* not parsed further, we were only interested in the extensions mask */
1542 if (get_bits_left(&s->gb) < 0)
1545 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1546 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1549 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1551 if (extensions_mask & DCA_EXT_EXSS_XLL)
1552 s->profile = FF_PROFILE_DTS_HD_MA;
1553 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1555 s->profile = FF_PROFILE_DTS_HD_HRA;
1557 if (!(extensions_mask & DCA_EXT_CORE))
1558 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1559 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1560 av_log(s->avctx, AV_LOG_WARNING,
1561 "DTS extensions detection mismatch (%d, %d)\n",
1562 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1568 * Parse extension substream header (HD)
1570 static void dca_exss_parse_header(DCAContext *s)
1576 int active_ss_mask[8];
1579 if (get_bits_left(&s->gb) < 52)
1582 skip_bits(&s->gb, 8); // user data
1583 ss_index = get_bits(&s->gb, 2);
1585 blownup = get_bits1(&s->gb);
1586 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1587 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1589 s->static_fields = get_bits1(&s->gb);
1590 if (s->static_fields) {
1591 skip_bits(&s->gb, 2); // reference clock code
1592 skip_bits(&s->gb, 3); // frame duration code
1594 if (get_bits1(&s->gb))
1595 skip_bits_long(&s->gb, 36); // timestamp
1597 /* a single stream can contain multiple audio assets that can be
1598 * combined to form multiple audio presentations */
1600 num_audiop = get_bits(&s->gb, 3) + 1;
1601 if (num_audiop > 1) {
1602 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1603 /* ignore such streams for now */
1607 num_assets = get_bits(&s->gb, 3) + 1;
1608 if (num_assets > 1) {
1609 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1610 /* ignore such streams for now */
1614 for (i = 0; i < num_audiop; i++)
1615 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1617 for (i = 0; i < num_audiop; i++)
1618 for (j = 0; j <= ss_index; j++)
1619 if (active_ss_mask[i] & (1 << j))
1620 skip_bits(&s->gb, 8); // active asset mask
1622 s->mix_metadata = get_bits1(&s->gb);
1623 if (s->mix_metadata) {
1624 int mix_out_mask_size;
1626 skip_bits(&s->gb, 2); // adjustment level
1627 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1628 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1630 for (i = 0; i < s->num_mix_configs; i++) {
1631 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1632 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1637 for (i = 0; i < num_assets; i++)
1638 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1640 for (i = 0; i < num_assets; i++) {
1641 if (dca_exss_parse_asset_header(s))
1645 /* not parsed further, we were only interested in the extensions mask
1646 * from the asset header */
1650 * Main frame decoding function
1651 * FIXME add arguments
1653 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1654 int *got_frame_ptr, AVPacket *avpkt)
1656 const uint8_t *buf = avpkt->data;
1657 int buf_size = avpkt->size;
1660 int num_core_channels = 0;
1663 int16_t *samples_s16;
1664 DCAContext *s = avctx->priv_data;
1671 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1672 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1673 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1674 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1675 return AVERROR_INVALIDDATA;
1678 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1679 if ((ret = dca_parse_frame_header(s)) < 0) {
1680 //seems like the frame is corrupt, try with the next one
1683 //set AVCodec values with parsed data
1684 avctx->sample_rate = s->sample_rate;
1685 avctx->bit_rate = s->bit_rate;
1687 s->profile = FF_PROFILE_DTS;
1689 for (i = 0; i < (s->sample_blocks / 8); i++) {
1690 if ((ret = dca_decode_block(s, 0, i))) {
1691 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1696 /* record number of core channels incase less than max channels are requested */
1697 num_core_channels = s->prim_channels;
1700 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1702 s->core_ext_mask = 0;
1704 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1706 /* only scan for extensions if ext_descr was unknown or indicated a
1707 * supported XCh extension */
1708 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1710 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1711 * extensions scan can fill it up */
1712 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1714 /* extensions start at 32-bit boundaries into bitstream */
1715 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1717 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1718 uint32_t bits = get_bits_long(&s->gb, 32);
1722 int ext_amode, xch_fsize;
1724 s->xch_base_channel = s->prim_channels;
1726 /* validate sync word using XCHFSIZE field */
1727 xch_fsize = show_bits(&s->gb, 10);
1728 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1729 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1732 /* skip length-to-end-of-frame field for the moment */
1733 skip_bits(&s->gb, 10);
1735 s->core_ext_mask |= DCA_EXT_XCH;
1737 /* extension amode(number of channels in extension) should be 1 */
1738 /* AFAIK XCh is not used for more channels */
1739 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1740 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1741 " supported!\n", ext_amode);
1745 /* much like core primary audio coding header */
1746 dca_parse_audio_coding_header(s, s->xch_base_channel);
1748 for (i = 0; i < (s->sample_blocks / 8); i++)
1749 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1750 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1758 /* XXCh: extended channels */
1759 /* usually found either in core or HD part in DTS-HD HRA streams,
1760 * but not in DTS-ES which contains XCh extensions instead */
1761 s->core_ext_mask |= DCA_EXT_XXCH;
1765 int fsize96 = show_bits(&s->gb, 12) + 1;
1766 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1769 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1770 get_bits_count(&s->gb));
1771 skip_bits(&s->gb, 12);
1772 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1773 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1775 s->core_ext_mask |= DCA_EXT_X96;
1780 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1783 /* no supported extensions, skip the rest of the core substream */
1784 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1787 if (s->core_ext_mask & DCA_EXT_X96)
1788 s->profile = FF_PROFILE_DTS_96_24;
1789 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1790 s->profile = FF_PROFILE_DTS_ES;
1792 /* check for ExSS (HD part) */
1793 if (s->dca_buffer_size - s->frame_size > 32 &&
1794 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1795 dca_exss_parse_header(s);
1797 avctx->profile = s->profile;
1799 channels = s->prim_channels + !!s->lfe;
1801 if (s->amode < 16) {
1802 avctx->channel_layout = dca_core_channel_layout[s->amode];
1804 if (s->xch_present && (!avctx->request_channels ||
1805 avctx->request_channels > num_core_channels + !!s->lfe)) {
1806 avctx->channel_layout |= AV_CH_BACK_CENTER;
1808 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1809 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1811 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1814 channels = num_core_channels + !!s->lfe;
1815 s->xch_present = 0; /* disable further xch processing */
1817 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1818 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1820 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1823 if (channels > !!s->lfe &&
1824 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1825 return AVERROR_INVALIDDATA;
1827 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1829 s->output = DCA_STEREO;
1830 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1832 else if (avctx->request_channel_layout & AV_CH_LAYOUT_NATIVE) {
1833 static const int8_t dca_channel_order_native[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 };
1834 s->channel_order_tab = dca_channel_order_native;
1837 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1838 return AVERROR_INVALIDDATA;
1841 if (avctx->channels != channels) {
1842 if (avctx->channels)
1843 av_log(avctx, AV_LOG_INFO, "Number of channels changed in DCA decoder (%d -> %d)\n", avctx->channels, channels);
1844 avctx->channels = channels;
1847 /* get output buffer */
1848 s->frame.nb_samples = 256 * (s->sample_blocks / 8);
1849 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
1850 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1853 samples_flt = (float *) s->frame.data[0];
1854 samples_s16 = (int16_t *) s->frame.data[0];
1856 /* filter to get final output */
1857 for (i = 0; i < (s->sample_blocks / 8); i++) {
1858 dca_filter_channels(s, i);
1860 /* If this was marked as a DTS-ES stream we need to subtract back- */
1861 /* channel from SL & SR to remove matrixed back-channel signal */
1862 if ((s->source_pcm_res & 1) && s->xch_present) {
1863 float *back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1864 float *lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1865 float *rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1866 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1867 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1870 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1871 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1873 samples_flt += 256 * channels;
1875 s->fmt_conv.float_to_int16_interleave(samples_s16,
1876 s->samples_chanptr, 256,
1878 samples_s16 += 256 * channels;
1882 /* update lfe history */
1883 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1884 for (i = 0; i < 2 * s->lfe * 4; i++)
1885 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1888 *(AVFrame *) data = s->frame;
1896 * DCA initialization
1898 * @param avctx pointer to the AVCodecContext
1901 static av_cold int dca_decode_init(AVCodecContext *avctx)
1903 DCAContext *s = avctx->priv_data;
1909 ff_dsputil_init(&s->dsp, avctx);
1910 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1911 ff_synth_filter_init(&s->synth);
1912 ff_dcadsp_init(&s->dcadsp);
1913 ff_fmt_convert_init(&s->fmt_conv, avctx);
1915 for (i = 0; i < DCA_PRIM_CHANNELS_MAX + 1; i++)
1916 s->samples_chanptr[i] = s->samples + i * 256;
1918 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1919 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1920 s->scale_bias = 1.0 / 32768.0;
1922 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1923 s->scale_bias = 1.0;
1926 /* allow downmixing to stereo */
1927 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1928 avctx->request_channels == 2) {
1929 avctx->channels = avctx->request_channels;
1932 avcodec_get_frame_defaults(&s->frame);
1933 avctx->coded_frame = &s->frame;
1938 static av_cold int dca_decode_end(AVCodecContext *avctx)
1940 DCAContext *s = avctx->priv_data;
1941 ff_mdct_end(&s->imdct);
1945 static const AVProfile profiles[] = {
1946 { FF_PROFILE_DTS, "DTS" },
1947 { FF_PROFILE_DTS_ES, "DTS-ES" },
1948 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1949 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1950 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1951 { FF_PROFILE_UNKNOWN },
1954 AVCodec ff_dca_decoder = {
1956 .type = AVMEDIA_TYPE_AUDIO,
1958 .priv_data_size = sizeof(DCAContext),
1959 .init = dca_decode_init,
1960 .decode = dca_decode_frame,
1961 .close = dca_decode_end,
1962 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1963 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1964 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
1966 AV_SAMPLE_FMT_NONE },
1967 .profiles = NULL_IF_CONFIG_SMALL(profiles),