Coverage Report

Created: 2025-11-16 06:35

next uncovered line (L), next uncovered region (R), next uncovered branch (B)
/src/aac/libAACdec/src/usacdec_lpc.cpp
Line
Count
Source
1
/* -----------------------------------------------------------------------------
2
Software License for The Fraunhofer FDK AAC Codec Library for Android
3
4
© Copyright  1995 - 2019 Fraunhofer-Gesellschaft zur Förderung der angewandten
5
Forschung e.V. All rights reserved.
6
7
 1.    INTRODUCTION
8
The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software
9
that implements the MPEG Advanced Audio Coding ("AAC") encoding and decoding
10
scheme for digital audio. This FDK AAC Codec software is intended to be used on
11
a wide variety of Android devices.
12
13
AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient
14
general perceptual audio codecs. AAC-ELD is considered the best-performing
15
full-bandwidth communications codec by independent studies and is widely
16
deployed. AAC has been standardized by ISO and IEC as part of the MPEG
17
specifications.
18
19
Patent licenses for necessary patent claims for the FDK AAC Codec (including
20
those of Fraunhofer) may be obtained through Via Licensing
21
(www.vialicensing.com) or through the respective patent owners individually for
22
the purpose of encoding or decoding bit streams in products that are compliant
23
with the ISO/IEC MPEG audio standards. Please note that most manufacturers of
24
Android devices already license these patent claims through Via Licensing or
25
directly from the patent owners, and therefore FDK AAC Codec software may
26
already be covered under those patent licenses when it is used for those
27
licensed purposes only.
28
29
Commercially-licensed AAC software libraries, including floating-point versions
30
with enhanced sound quality, are also available from Fraunhofer. Users are
31
encouraged to check the Fraunhofer website for additional applications
32
information and documentation.
33
34
2.    COPYRIGHT LICENSE
35
36
Redistribution and use in source and binary forms, with or without modification,
37
are permitted without payment of copyright license fees provided that you
38
satisfy the following conditions:
39
40
You must retain the complete text of this software license in redistributions of
41
the FDK AAC Codec or your modifications thereto in source code form.
42
43
You must retain the complete text of this software license in the documentation
44
and/or other materials provided with redistributions of the FDK AAC Codec or
45
your modifications thereto in binary form. You must make available free of
46
charge copies of the complete source code of the FDK AAC Codec and your
47
modifications thereto to recipients of copies in binary form.
48
49
The name of Fraunhofer may not be used to endorse or promote products derived
50
from this library without prior written permission.
51
52
You may not charge copyright license fees for anyone to use, copy or distribute
53
the FDK AAC Codec software or your modifications thereto.
54
55
Your modified versions of the FDK AAC Codec must carry prominent notices stating
56
that you changed the software and the date of any change. For modified versions
57
of the FDK AAC Codec, the term "Fraunhofer FDK AAC Codec Library for Android"
58
must be replaced by the term "Third-Party Modified Version of the Fraunhofer FDK
59
AAC Codec Library for Android."
60
61
3.    NO PATENT LICENSE
62
63
NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without
64
limitation the patents of Fraunhofer, ARE GRANTED BY THIS SOFTWARE LICENSE.
65
Fraunhofer provides no warranty of patent non-infringement with respect to this
66
software.
67
68
You may use this FDK AAC Codec software or modifications thereto only for
69
purposes that are authorized by appropriate patent licenses.
70
71
4.    DISCLAIMER
72
73
This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright
74
holders and contributors "AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES,
75
including but not limited to the implied warranties of merchantability and
76
fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
77
CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary,
78
or consequential damages, including but not limited to procurement of substitute
79
goods or services; loss of use, data, or profits, or business interruption,
80
however caused and on any theory of liability, whether in contract, strict
81
liability, or tort (including negligence), arising in any way out of the use of
82
this software, even if advised of the possibility of such damage.
83
84
5.    CONTACT INFORMATION
85
86
Fraunhofer Institute for Integrated Circuits IIS
87
Attention: Audio and Multimedia Departments - FDK AAC LL
88
Am Wolfsmantel 33
89
91058 Erlangen, Germany
90
91
www.iis.fraunhofer.de/amm
92
amm-info@iis.fraunhofer.de
93
----------------------------------------------------------------------------- */
94
95
/**************************** AAC decoder library ******************************
96
97
   Author(s):   Matthias Hildenbrand, Manuel Jander
98
99
   Description: USAC LPC/AVQ decode
100
101
*******************************************************************************/
102
103
#include "usacdec_lpc.h"
104
105
#include "usacdec_rom.h"
106
#include "FDK_trigFcts.h"
107
108
1.86M
#define NQ_MAX 36
109
110
/*
111
 * Helper functions.
112
 */
113
114
/**
115
 * \brief Read unary code.
116
 * \param hBs bitstream handle as data source.
117
 * \return decoded value.
118
 */
119
1.06M
static int get_vlclbf(HANDLE_FDK_BITSTREAM hBs) {
120
1.06M
  int result = 0;
121
122
1.63M
  while (FDKreadBits(hBs, 1) && result <= NQ_MAX) {
123
572k
    result++;
124
572k
  }
125
1.06M
  return result;
126
1.06M
}
127
128
/**
129
 * \brief Read bit count limited unary code.
130
 * \param hBs bitstream handle as data source
131
 * \param n max amount of bits to be read.
132
 * \return decoded value.
133
 */
134
66.2k
static int get_vlclbf_n(HANDLE_FDK_BITSTREAM hBs, int n) {
135
66.2k
  int result = 0;
136
137
97.6k
  while (FDKreadBits(hBs, 1)) {
138
41.2k
    result++;
139
41.2k
    n--;
140
41.2k
    if (n <= 0) {
141
9.88k
      break;
142
9.88k
    }
143
41.2k
  }
144
145
66.2k
  return result;
146
66.2k
}
147
148
/*
149
 * Algebraic Vector Quantizer
150
 */
151
152
/* ZF_SCALE must be greater than (number of FIXP_ZF)/2
153
   because the loss of precision caused by fPow2Div2 in RE8_PPV() */
154
//#define ZF_SCALE ((NQ_MAX-3)>>1)
155
5.95M
#define ZF_SCALE ((DFRACT_BITS / 2))
156
1.63M
#define FIXP_ZF FIXP_DBL
157
4.64M
#define INT2ZF(x, s) (FIXP_ZF)((x) << (ZF_SCALE - (s)))
158
1.30M
#define ZF2INT(x) (INT)((x) >> ZF_SCALE)
159
160
/* 1.0 in ZF format format */
161
1.96M
#define ONEZF ((FIXP_ZF)INT2ZF(1, 0))
162
163
/* static */
164
163k
void nearest_neighbor_2D8(FIXP_ZF x[8], int y[8]) {
165
163k
  FIXP_ZF s, em, e[8];
166
163k
  int i, j, sum;
167
168
  /* round x into 2Z^8 i.e. compute y=(y1,...,y8) such that yi = 2[xi/2]
169
     where [.] is the nearest integer operator
170
     in the mean time, compute sum = y1+...+y8
171
  */
172
163k
  sum = 0;
173
1.47M
  for (i = 0; i < 8; i++) {
174
1.30M
    FIXP_ZF tmp;
175
    /* round to ..., -2, 0, 2, ... ([-1..1[ --> 0) */
176
1.30M
    if (x[i] < (FIXP_ZF)0) {
177
427k
      tmp = ONEZF - x[i];
178
427k
      y[i] = -2 * ((ZF2INT(tmp)) >> 1);
179
881k
    } else {
180
881k
      tmp = ONEZF + x[i];
181
881k
      y[i] = 2 * ((ZF2INT(tmp)) >> 1);
182
881k
    }
183
1.30M
    sum += y[i];
184
1.30M
  }
185
  /* check if y1+...+y8 is a multiple of 4
186
     if not, y is not round xj in the wrong way where j is defined by
187
        j = arg max_i | xi -yi|
188
     (this is called the Wagner rule)
189
  */
190
163k
  if (sum % 4) {
191
    /* find j = arg max_i | xi -yi| */
192
90.2k
    em = (FIXP_SGL)0;
193
90.2k
    j = 0;
194
812k
    for (i = 0; i < 8; i++) {
195
      /* compute ei = xi-yi */
196
721k
      e[i] = x[i] - INT2ZF(y[i], 0);
197
721k
    }
198
812k
    for (i = 0; i < 8; i++) {
199
      /* compute |ei| = | xi-yi | */
200
721k
      if (e[i] < (FIXP_ZF)0) {
201
142k
        s = -e[i];
202
579k
      } else {
203
579k
        s = e[i];
204
579k
      }
205
      /* check if |ei| is maximal, if so, set j=i */
206
721k
      if (em < s) {
207
106k
        em = s;
208
106k
        j = i;
209
106k
      }
210
721k
    }
211
    /* round xj in the "wrong way" */
212
90.2k
    if (e[j] < (FIXP_ZF)0) {
213
43.7k
      y[j] -= 2;
214
46.4k
    } else {
215
46.4k
      y[j] += 2;
216
46.4k
    }
217
90.2k
  }
218
163k
}
219
220
/*--------------------------------------------------------------
221
  RE8_PPV(x,y)
222
  NEAREST NEIGHBOR SEARCH IN INFINITE LATTICE RE8
223
  the algorithm is based on the definition of RE8 as
224
      RE8 = (2D8) U (2D8+[1,1,1,1,1,1,1,1])
225
  it applies the coset decoding of Sloane and Conway
226
  (i) x: point in R^8 in 32-ZF_SCALE.ZF_SCALE format
227
  (o) y: point in RE8 (8-dimensional integer vector)
228
  --------------------------------------------------------------
229
*/
230
/* static */
231
81.8k
void RE8_PPV(FIXP_ZF x[], SHORT y[], int r) {
232
81.8k
  int i, y0[8], y1[8];
233
81.8k
  FIXP_ZF x1[8], tmp;
234
81.8k
  INT64 e;
235
236
  /* find the nearest neighbor y0 of x in 2D8 */
237
81.8k
  nearest_neighbor_2D8(x, y0);
238
  /* find the nearest neighbor y1 of x in 2D8+(1,...,1) (by coset decoding) */
239
736k
  for (i = 0; i < 8; i++) {
240
654k
    x1[i] = x[i] - ONEZF;
241
654k
  }
242
81.8k
  nearest_neighbor_2D8(x1, y1);
243
736k
  for (i = 0; i < 8; i++) {
244
654k
    y1[i] += 1;
245
654k
  }
246
247
  /* compute e0=||x-y0||^2 and e1=||x-y1||^2 */
248
81.8k
  e = 0;
249
736k
  for (i = 0; i < 8; i++) {
250
654k
    tmp = x[i] - INT2ZF(y0[i], 0);
251
654k
    e += (INT64)fPow2Div2(
252
654k
        tmp << r); /* shift left to ensure that no fract part bits get lost. */
253
654k
    tmp = x[i] - INT2ZF(y1[i], 0);
254
654k
    e -= (INT64)fPow2Div2(tmp << r);
255
654k
  }
256
  /* select best candidate y0 or y1 to minimize distortion */
257
81.8k
  if (e < 0) {
258
588k
    for (i = 0; i < 8; i++) {
259
523k
      y[i] = y0[i];
260
523k
    }
261
65.4k
  } else {
262
147k
    for (i = 0; i < 8; i++) {
263
131k
      y[i] = y1[i];
264
131k
    }
265
16.4k
  }
266
81.8k
}
267
268
/* table look-up of unsigned value: find i where index >= table[i]
269
   Note: range must be >= 2, index must be >= table[0] */
270
955k
static int table_lookup(const USHORT *table, unsigned int index, int range) {
271
955k
  int i;
272
273
1.22M
  for (i = 4; i < range; i += 4) {
274
1.09M
    if (index < table[i]) {
275
820k
      break;
276
820k
    }
277
1.09M
  }
278
955k
  if (i > range) {
279
94.9k
    i = range;
280
94.9k
  }
281
282
955k
  if (index < table[i - 2]) {
283
610k
    i -= 2;
284
610k
  }
285
955k
  if (index < table[i - 1]) {
286
528k
    i--;
287
528k
  }
288
955k
  i--;
289
290
955k
  return (i); /* index >= table[i] */
291
955k
}
292
293
/*--------------------------------------------------------------------------
294
  re8_decode_rank_of_permutation(rank, xs, x)
295
  DECODING OF THE RANK OF THE PERMUTATION OF xs
296
  (i) rank: index (rank) of a permutation
297
  (i) xs:   signed leader in RE8 (8-dimensional integer vector)
298
  (o) x:    point in RE8 (8-dimensional integer vector)
299
  --------------------------------------------------------------------------
300
 */
301
477k
static void re8_decode_rank_of_permutation(int rank, int *xs, SHORT x[8]) {
302
477k
  INT a[8], w[8], B, fac, fac_B, target;
303
477k
  int i, j;
304
305
  /* --- pre-processing based on the signed leader xs ---
306
     - compute the alphabet a=[a[0] ... a[q-1]] of x (q elements)
307
       such that a[0]!=...!=a[q-1]
308
       it is assumed that xs is sorted in the form of a signed leader
309
       which can be summarized in 2 requirements:
310
          a) |xs[0]| >= |xs[1]| >= |xs[2]| >= ... >= |xs[7]|
311
          b) if |xs[i]|=|xs[i-1]|, xs[i]>=xs[i+1]
312
       where |.| indicates the absolute value operator
313
     - compute q (the number of symbols in the alphabet)
314
     - compute w[0..q-1] where w[j] counts the number of occurences of
315
       the symbol a[j] in xs
316
     - compute B = prod_j=0..q-1 (w[j]!) where .! is the factorial */
317
  /* xs[i], xs[i-1] and ptr_w/a*/
318
477k
  j = 0;
319
477k
  w[j] = 1;
320
477k
  a[j] = xs[0];
321
477k
  B = 1;
322
3.82M
  for (i = 1; i < 8; i++) {
323
3.34M
    if (xs[i] != xs[i - 1]) {
324
718k
      j++;
325
718k
      w[j] = 1;
326
718k
      a[j] = xs[i];
327
2.62M
    } else {
328
2.62M
      w[j]++;
329
2.62M
      B *= w[j];
330
2.62M
    }
331
3.34M
  }
332
333
  /* --- actual rank decoding ---
334
     the rank of x (where x is a permutation of xs) is based on
335
     Schalkwijk's formula
336
     it is given by rank=sum_{k=0..7} (A_k * fac_k/B_k)
337
     the decoding of this rank is sequential and reconstructs x[0..7]
338
     element by element from x[0] to x[7]
339
     [the tricky part is the inference of A_k for each k...]
340
   */
341
342
477k
  if (w[0] == 8) {
343
460k
    for (i = 0; i < 8; i++) {
344
409k
      x[i] = a[0]; /* avoid fac of 40320 */
345
409k
    }
346
426k
  } else {
347
426k
    target = rank * B;
348
426k
    fac_B = 1;
349
    /* decode x element by element */
350
3.83M
    for (i = 0; i < 8; i++) {
351
3.41M
      fac = fac_B * fdk_dec_tab_factorial[i]; /* fac = 1..5040 */
352
3.41M
      j = -1;
353
6.87M
      do {
354
6.87M
        target -= w[++j] * fac;
355
6.87M
      } while (target >= 0); /* max of 30 tests / SV */
356
3.41M
      x[i] = a[j];
357
      /* update rank, denominator B (B_k) and counter w[j] */
358
3.41M
      target += w[j] * fac; /* target = fac_B*B*rank */
359
3.41M
      fac_B *= w[j];
360
3.41M
      w[j]--;
361
3.41M
    }
362
426k
  }
363
477k
}
364
365
/*--------------------------------------------------------------------------
366
  re8_decode_base_index(n, I, y)
367
  DECODING OF AN INDEX IN Qn (n=0,2,3 or 4)
368
  (i) n: codebook number (*n is an integer defined in {0,2,3,4})
369
  (i) I: index of c (pointer to unsigned 16-bit word)
370
  (o) y: point in RE8 (8-dimensional integer vector)
371
  note: the index I is defined as a 32-bit word, but only
372
  16 bits are required (long can be replaced by unsigned integer)
373
  --------------------------------------------------------------------------
374
 */
375
1.28M
static void re8_decode_base_index(int *n, UINT index, SHORT y[8]) {
376
1.28M
  int i, im, t, sign_code, ka, ks, rank, leader[8];
377
378
1.28M
  if (*n < 2) {
379
7.29M
    for (i = 0; i < 8; i++) {
380
6.48M
      y[i] = 0;
381
6.48M
    }
382
810k
  } else {
383
    // index = (unsigned int)*I;
384
    /* search for the identifier ka of the absolute leader (table-lookup)
385
       Q2 is a subset of Q3 - the two cases are considered in the same branch
386
     */
387
477k
    switch (*n) {
388
231k
      case 2:
389
365k
      case 3:
390
365k
        i = table_lookup(fdk_dec_I3, index, NB_LDQ3);
391
365k
        ka = fdk_dec_A3[i];
392
365k
        break;
393
112k
      case 4:
394
112k
        i = table_lookup(fdk_dec_I4, index, NB_LDQ4);
395
112k
        ka = fdk_dec_A4[i];
396
112k
        break;
397
0
      default:
398
0
        FDK_ASSERT(0);
399
0
        return;
400
477k
    }
401
    /* reconstruct the absolute leader */
402
4.29M
    for (i = 0; i < 8; i++) {
403
3.82M
      leader[i] = fdk_dec_Da[ka][i];
404
3.82M
    }
405
    /* search for the identifier ks of the signed leader (table look-up)
406
       (this search is focused based on the identifier ka of the absolute
407
        leader)*/
408
477k
    t = fdk_dec_Ia[ka];
409
477k
    im = fdk_dec_Ns[ka];
410
477k
    ks = table_lookup(fdk_dec_Is + t, index, im);
411
412
    /* reconstruct the signed leader from its sign code */
413
477k
    sign_code = 2 * fdk_dec_Ds[t + ks];
414
4.29M
    for (i = 7; i >= 0; i--) {
415
3.82M
      leader[i] *= (1 - (sign_code & 2));
416
3.82M
      sign_code >>= 1;
417
3.82M
    }
418
419
    /* compute and decode the rank of the permutation */
420
477k
    rank = index - fdk_dec_Is[t + ks]; /* rank = index - cardinality offset */
421
422
477k
    re8_decode_rank_of_permutation(rank, leader, y);
423
477k
  }
424
1.28M
  return;
425
1.28M
}
426
427
/* re8_y2k(y,m,k)
428
   VORONOI INDEXING (INDEX DECODING) k -> y
429
   (i) k: Voronoi index k[0..7]
430
   (i) m: Voronoi modulo (m = 2^r = 1<<r, where r is integer >=2)
431
   (i) r: Voronoi order  (m = 2^r = 1<<r, where r is integer >=2)
432
   (o) y: 8-dimensional point y[0..7] in RE8
433
 */
434
81.8k
static void re8_k2y(int *k, int r, SHORT *y) {
435
81.8k
  int i, tmp, sum;
436
81.8k
  SHORT v[8];
437
81.8k
  FIXP_ZF zf[8];
438
439
81.8k
  FDK_ASSERT(r <= ZF_SCALE);
440
441
  /* compute y = k M and z=(y-a)/m, where
442
     M = [4        ]
443
         [2 2      ]
444
         [|   \    ]
445
         [2     2  ]
446
         [1 1 _ 1 1]
447
     a=(2,0,...,0)
448
     m = 1<<r
449
  */
450
736k
  for (i = 0; i < 8; i++) {
451
654k
    y[i] = k[7];
452
654k
  }
453
81.8k
  zf[7] = INT2ZF(y[7], r);
454
81.8k
  sum = 0;
455
572k
  for (i = 6; i >= 1; i--) {
456
490k
    tmp = 2 * k[i];
457
490k
    sum += tmp;
458
490k
    y[i] += tmp;
459
490k
    zf[i] = INT2ZF(y[i], r);
460
490k
  }
461
81.8k
  y[0] += (4 * k[0] + sum);
462
81.8k
  zf[0] = INT2ZF(y[0] - 2, r);
463
  /* find nearest neighbor v of z in infinite RE8 */
464
81.8k
  RE8_PPV(zf, v, r);
465
  /* compute y -= m v */
466
736k
  for (i = 0; i < 8; i++) {
467
654k
    y[i] -= (SHORT)(v[i] << r);
468
654k
  }
469
81.8k
}
470
471
/*--------------------------------------------------------------------------
472
  RE8_dec(n, I, k, y)
473
  MULTI-RATE INDEXING OF A POINT y in THE LATTICE RE8 (INDEX DECODING)
474
  (i) n: codebook number (*n is an integer defined in {0,2,3,4,..,n_max}). n_max
475
  = 36 (i) I: index of c (pointer to unsigned 16-bit word) (i) k: index of v
476
  (8-dimensional vector of binary indices) = Voronoi index (o) y: point in RE8
477
  (8-dimensional integer vector) note: the index I is defined as a 32-bit word,
478
  but only 16 bits are required (long can be replaced by unsigned integer)
479
480
  return 0 on success, -1 on error.
481
  --------------------------------------------------------------------------
482
 */
483
1.28M
static int RE8_dec(int n, int I, int *k, FIXP_DBL *y) {
484
1.28M
  SHORT v[8];
485
1.28M
  SHORT _y[8];
486
1.28M
  UINT r;
487
1.28M
  int i;
488
489
  /* Check bound of codebook qn */
490
1.28M
  if (n > NQ_MAX) {
491
345
    return -1;
492
345
  }
493
494
  /* decode the sub-indices I and kv[] according to the codebook number n:
495
     if n=0,2,3,4, decode I (no Voronoi extension)
496
     if n>4, Voronoi extension is used, decode I and kv[] */
497
1.28M
  if (n <= 4) {
498
1.20M
    re8_decode_base_index(&n, I, _y);
499
10.8M
    for (i = 0; i < 8; i++) {
500
9.64M
      y[i] = (LONG)_y[i];
501
9.64M
    }
502
1.20M
  } else {
503
    /* compute the Voronoi modulo m = 2^r where r is extension order */
504
81.8k
    r = ((n - 3) >> 1);
505
506
219k
    while (n > 4) {
507
137k
      n -= 2;
508
137k
    }
509
    /* decode base codebook index I into c (c is an element of Q3 or Q4)
510
       [here c is stored in y to save memory] */
511
81.8k
    re8_decode_base_index(&n, I, _y);
512
    /* decode Voronoi index k[] into v */
513
81.8k
    re8_k2y(k, r, v);
514
    /* reconstruct y as y = m c + v (with m=2^r, r integer >=1) */
515
736k
    for (i = 0; i < 8; i++) {
516
654k
      y[i] = (LONG)((_y[i] << r) + v[i]);
517
654k
    }
518
81.8k
  }
519
1.28M
  return 0;
520
1.28M
}
521
522
/**************************/
523
/* start LPC decode stuff */
524
/**************************/
525
//#define M         16
526
#define FREQ_MAX 6400.0f
527
#define FREQ_DIV 400.0f
528
#define LSF_GAP 50.0f
529
530
/**
531
 * \brief calculate inverse weighting factor and add non-weighted residual
532
 *        LSF vector to first stage LSF approximation
533
 * \param lsfq first stage LSF approximation values.
534
 * \param xq weighted residual LSF vector
535
 * \param nk_mode code book number coding mode.
536
 */
537
161k
static void lsf_weight_2st(FIXP_LPC *lsfq, FIXP_DBL *xq, int nk_mode) {
538
161k
  FIXP_LPC d[M_LP_FILTER_ORDER + 1];
539
161k
  FIXP_SGL factor;
540
161k
  LONG w; /* inverse weight factor */
541
161k
  int i;
542
543
  /* compute lsf distance */
544
161k
  d[0] = lsfq[0];
545
161k
  d[M_LP_FILTER_ORDER] =
546
161k
      FL2FXCONST_LPC(FREQ_MAX / (1 << LSF_SCALE)) - lsfq[M_LP_FILTER_ORDER - 1];
547
2.58M
  for (i = 1; i < M_LP_FILTER_ORDER; i++) {
548
2.41M
    d[i] = lsfq[i] - lsfq[i - 1];
549
2.41M
  }
550
551
161k
  switch (nk_mode) {
552
88.4k
    case 0:
553
88.4k
      factor = FL2FXCONST_SGL(2.0f * 60.0f / FREQ_DIV);
554
88.4k
      break; /* abs */
555
25.2k
    case 1:
556
25.2k
      factor = FL2FXCONST_SGL(2.0f * 65.0f / FREQ_DIV);
557
25.2k
      break; /* mid */
558
24.1k
    case 2:
559
24.1k
      factor = FL2FXCONST_SGL(2.0f * 64.0f / FREQ_DIV);
560
24.1k
      break; /* rel1 */
561
23.4k
    default:
562
23.4k
      factor = FL2FXCONST_SGL(2.0f * 63.0f / FREQ_DIV);
563
23.4k
      break; /* rel2 */
564
161k
  }
565
  /* add non-weighted residual LSF vector to LSF1st */
566
2.74M
  for (i = 0; i < M_LP_FILTER_ORDER; i++) {
567
2.58M
    w = (LONG)fMultDiv2(factor, sqrtFixp(fMult(d[i], d[i + 1])));
568
2.58M
    lsfq[i] = fAddSaturate(lsfq[i],
569
2.58M
                           FX_DBL2FX_LPC((FIXP_DBL)((INT64)w * (LONG)xq[i])));
570
2.58M
  }
571
572
161k
  return;
573
161k
}
574
575
/**
576
 * \brief decode nqn amount of code book numbers. These values determine the
577
 * amount of following bits for nqn AVQ RE8 vectors.
578
 * \param nk_mode quantization mode.
579
 * \param nqn amount code book number to read.
580
 * \param qn pointer to output buffer to hold decoded code book numbers qn.
581
 */
582
static void decode_qn(HANDLE_FDK_BITSTREAM hBs, int nk_mode, int nqn,
583
1.12M
                      int qn[]) {
584
1.12M
  int n;
585
586
1.12M
  if (nk_mode == 1) { /* nk mode 1 */
587
    /* Unary code for mid LPC1/LPC3 */
588
    /* Q0=0, Q2=10, Q3=110, ... */
589
2.00M
    for (n = 0; n < nqn; n++) {
590
1.01M
      qn[n] = get_vlclbf(hBs);
591
1.01M
      if (qn[n] > 0) {
592
210k
        qn[n]++;
593
210k
      }
594
1.01M
    }
595
990k
  } else { /* nk_mode 0, 3 and 2 */
596
    /* 2 bits to specify Q2,Q3,Q4,ext */
597
408k
    for (n = 0; n < nqn; n++) {
598
272k
      qn[n] = 2 + FDKreadBits(hBs, 2);
599
272k
    }
600
136k
    if (nk_mode == 2) {
601
      /* Unary code for rel LPC1/LPC3 */
602
      /* Q0 = 0, Q5=10, Q6=110, ... */
603
72.5k
      for (n = 0; n < nqn; n++) {
604
48.3k
        if (qn[n] > 4) {
605
3.97k
          qn[n] = get_vlclbf(hBs);
606
3.97k
          if (qn[n] > 0) qn[n] += 4;
607
3.97k
        }
608
48.3k
      }
609
111k
    } else { /* nk_mode == (0 and 3) */
610
      /* Unary code for abs and rel LPC0/LPC2 */
611
      /* Q5 = 0, Q6=10, Q0=110, Q7=1110, ... */
612
335k
      for (n = 0; n < nqn; n++) {
613
223k
        if (qn[n] > 4) {
614
44.7k
          qn[n] = get_vlclbf(hBs);
615
44.7k
          switch (qn[n]) {
616
22.0k
            case 0:
617
22.0k
              qn[n] = 5;
618
22.0k
              break;
619
8.07k
            case 1:
620
8.07k
              qn[n] = 6;
621
8.07k
              break;
622
2.92k
            case 2:
623
2.92k
              qn[n] = 0;
624
2.92k
              break;
625
11.6k
            default:
626
11.6k
              qn[n] += 4;
627
11.6k
              break;
628
44.7k
          }
629
44.7k
        }
630
223k
      }
631
111k
    }
632
136k
  }
633
1.12M
}
634
635
/**
636
 * \brief reorder LSF coefficients to minimum distance.
637
 * \param lsf pointer to buffer containing LSF coefficients and where reordered
638
 * LSF coefficients will be stored into, scaled by LSF_SCALE.
639
 * \param min_dist min distance scaled by LSF_SCALE
640
 * \param n number of LSF/LSP coefficients.
641
 */
642
161k
static void reorder_lsf(FIXP_LPC *lsf, FIXP_LPC min_dist, int n) {
643
161k
  FIXP_LPC lsf_min;
644
161k
  int i;
645
646
161k
  lsf_min = min_dist;
647
2.74M
  for (i = 0; i < n; i++) {
648
2.58M
    if (lsf[i] < lsf_min) {
649
121k
      lsf[i] = lsf_min;
650
121k
    }
651
2.58M
    lsf_min = fAddSaturate(lsf[i], min_dist);
652
2.58M
  }
653
654
  /* reverse */
655
161k
  lsf_min = FL2FXCONST_LPC(FREQ_MAX / (1 << LSF_SCALE)) - min_dist;
656
2.74M
  for (i = n - 1; i >= 0; i--) {
657
2.58M
    if (lsf[i] > lsf_min) {
658
18.7k
      lsf[i] = lsf_min;
659
18.7k
    }
660
661
2.58M
    lsf_min = lsf[i] - min_dist;
662
2.58M
  }
663
161k
}
664
665
/**
666
 * \brief First stage approximation
667
 * \param hBs bitstream handle as data source
668
 * \param lsfq pointer to output buffer to hold LPC coefficients scaled by
669
 * LSF_SCALE.
670
 */
671
static void vlpc_1st_dec(
672
    HANDLE_FDK_BITSTREAM hBs, /* input:  codebook index                  */
673
    FIXP_LPC *lsfq            /* i/o:    i:prediction   o:quantized lsf  */
674
88.4k
) {
675
88.4k
  const FIXP_LPC *p_dico;
676
88.4k
  int i, index;
677
678
88.4k
  index = FDKreadBits(hBs, 8);
679
88.4k
  p_dico = &fdk_dec_dico_lsf_abs_8b[index * M_LP_FILTER_ORDER];
680
1.50M
  for (i = 0; i < M_LP_FILTER_ORDER; i++) {
681
1.41M
    lsfq[i] = p_dico[i];
682
1.41M
  }
683
88.4k
}
684
685
/**
686
 * \brief Do first stage approximation weighting and multiply with AVQ
687
 * refinement.
688
 * \param hBs bitstream handle data ssource.
689
 * \param lsfq buffer holding 1st stage approx, 2nd stage approx is added to
690
 * this values.
691
 * \param nk_mode quantization mode.
692
 * \return 0 on success, -1 on error.
693
 */
694
static int vlpc_2st_dec(
695
    HANDLE_FDK_BITSTREAM hBs,
696
    FIXP_LPC *lsfq, /* i/o:    i:1st stage   o:1st+2nd stage   */
697
    int nk_mode     /* input:  0=abs, >0=rel                   */
698
161k
) {
699
161k
  int err;
700
161k
  FIXP_DBL xq[M_LP_FILTER_ORDER]; /* weighted residual LSF vector */
701
702
  /* Decode AVQ refinement */
703
161k
  { err = CLpc_DecodeAVQ(hBs, xq, nk_mode, 2, 8); }
704
161k
  if (err != 0) {
705
12
    return -1;
706
12
  }
707
708
  /* add non-weighted residual LSF vector to LSF1st */
709
161k
  lsf_weight_2st(lsfq, xq, nk_mode);
710
711
  /* reorder */
712
161k
  reorder_lsf(lsfq, FL2FXCONST_LPC(LSF_GAP / (1 << LSF_SCALE)),
713
161k
              M_LP_FILTER_ORDER);
714
715
161k
  return 0;
716
161k
}
717
718
/*
719
 * Externally visible functions
720
 */
721
722
int CLpc_DecodeAVQ(HANDLE_FDK_BITSTREAM hBs, FIXP_DBL *pOutput, int nk_mode,
723
247k
                   int no_qn, int length) {
724
247k
  int i, l;
725
726
1.37M
  for (i = 0; i < length; i += 8 * no_qn) {
727
1.12M
    int qn[2], nk, n, I;
728
1.12M
    int kv[8] = {0};
729
730
1.12M
    decode_qn(hBs, nk_mode, no_qn, qn);
731
732
2.41M
    for (l = 0; l < no_qn; l++) {
733
1.28M
      if (qn[l] == 0) {
734
810k
        FDKmemclear(&pOutput[i + l * 8], 8 * sizeof(FIXP_DBL));
735
810k
      }
736
737
      /* Voronoi extension order ( nk ) */
738
1.28M
      nk = 0;
739
1.28M
      n = qn[l];
740
1.28M
      if (qn[l] > 4) {
741
82.1k
        nk = (qn[l] - 3) >> 1;
742
82.1k
        n = qn[l] - nk * 2;
743
82.1k
      }
744
745
      /* Base codebook index, in reverse bit group order (!) */
746
1.28M
      I = FDKreadBits(hBs, 4 * n);
747
748
1.28M
      if (nk > 0) {
749
82.1k
        int j;
750
751
739k
        for (j = 0; j < 8; j++) {
752
657k
          kv[j] = FDKreadBits(hBs, nk);
753
657k
        }
754
82.1k
      }
755
756
1.28M
      if (RE8_dec(qn[l], I, kv, &pOutput[i + l * 8]) != 0) {
757
345
        return -1;
758
345
      }
759
1.28M
    }
760
1.12M
  }
761
247k
  return 0;
762
247k
}
763
764
int CLpc_Read(HANDLE_FDK_BITSTREAM hBs, FIXP_LPC lsp[][M_LP_FILTER_ORDER],
765
              FIXP_LPC lpc4_lsf[M_LP_FILTER_ORDER],
766
              FIXP_LPC lsf_adaptive_mean_cand[M_LP_FILTER_ORDER],
767
              FIXP_SGL pStability[], UCHAR *mod, int first_lpd_flag,
768
42.2k
              int last_lpc_lost, int last_frame_ok) {
769
42.2k
  int i, k, err;
770
42.2k
  int mode_lpc_bin = 0; /* mode_lpc bitstream representation */
771
42.2k
  int lpc_present[5] = {0, 0, 0, 0, 0};
772
42.2k
  int lpc0_available = 1;
773
42.2k
  int s = 0;
774
42.2k
  int l = 3;
775
42.2k
  const int nbDiv = NB_DIV;
776
777
42.2k
  lpc_present[4 >> s] = 1; /* LPC4 */
778
779
  /* Decode LPC filters in the following order: LPC 4,0,2,1,3 */
780
781
  /*** Decode LPC4 ***/
782
42.2k
  vlpc_1st_dec(hBs, lsp[4 >> s]);
783
42.2k
  err = vlpc_2st_dec(hBs, lsp[4 >> s], 0); /* nk_mode = 0 */
784
42.2k
  if (err != 0) {
785
5
    return err;
786
5
  }
787
788
  /*** Decode LPC0 and LPC2 ***/
789
42.2k
  k = 0;
790
42.2k
  if (!first_lpd_flag) {
791
24.0k
    lpc_present[0] = 1;
792
24.0k
    lpc0_available = !last_lpc_lost;
793
    /* old LPC4 is new LPC0 */
794
408k
    for (i = 0; i < M_LP_FILTER_ORDER; i++) {
795
384k
      lsp[0][i] = lpc4_lsf[i];
796
384k
    }
797
    /* skip LPC0 and continue with LPC2 */
798
24.0k
    k = 2;
799
24.0k
  }
800
801
102k
  for (; k < l; k += 2) {
802
60.4k
    int nk_mode = 0;
803
804
60.4k
    if ((k == 2) && (mod[0] == 3)) {
805
487
      break; /* skip LPC2 */
806
487
    }
807
808
59.9k
    lpc_present[k >> s] = 1;
809
810
59.9k
    mode_lpc_bin = FDKreadBit(hBs);
811
812
59.9k
    if (mode_lpc_bin == 0) {
813
      /* LPC0/LPC2: Abs */
814
36.5k
      vlpc_1st_dec(hBs, lsp[k >> s]);
815
36.5k
    } else {
816
      /* LPC0/LPC2: RelR */
817
398k
      for (i = 0; i < M_LP_FILTER_ORDER; i++) {
818
374k
        lsp[k >> s][i] = lsp[4 >> s][i];
819
374k
      }
820
23.4k
      nk_mode = 3;
821
23.4k
    }
822
823
59.9k
    err = vlpc_2st_dec(hBs, lsp[k >> s], nk_mode);
824
59.9k
    if (err != 0) {
825
3
      return err;
826
3
    }
827
59.9k
  }
828
829
  /*** Decode LPC1 ***/
830
42.2k
  if (mod[0] < 2) { /* else: skip LPC1 */
831
27.5k
    lpc_present[1] = 1;
832
27.5k
    mode_lpc_bin = get_vlclbf_n(hBs, 2);
833
834
27.5k
    switch (mode_lpc_bin) {
835
3.53k
      case 1:
836
        /* LPC1: abs */
837
3.53k
        vlpc_1st_dec(hBs, lsp[1]);
838
3.53k
        err = vlpc_2st_dec(hBs, lsp[1], 0);
839
3.53k
        if (err != 0) {
840
1
          return err;
841
1
        }
842
3.52k
        break;
843
7.11k
      case 2:
844
        /* LPC1: mid0 (no second stage AVQ quantizer in this case) */
845
7.11k
        if (lpc0_available) { /* LPC0/lsf[0] might be zero some times */
846
120k
          for (i = 0; i < M_LP_FILTER_ORDER; i++) {
847
113k
            lsp[1][i] = (lsp[0][i] >> 1) + (lsp[2][i] >> 1);
848
113k
          }
849
7.11k
        } else {
850
17
          for (i = 0; i < M_LP_FILTER_ORDER; i++) {
851
16
            lsp[1][i] = lsp[2][i];
852
16
          }
853
1
        }
854
7.11k
        break;
855
16.9k
      case 0:
856
        /* LPC1: RelR */
857
287k
        for (i = 0; i < M_LP_FILTER_ORDER; i++) {
858
270k
          lsp[1][i] = lsp[2][i];
859
270k
        }
860
16.9k
        err = vlpc_2st_dec(hBs, lsp[1], 2 << s);
861
16.9k
        if (err != 0) {
862
1
          return err;
863
1
        }
864
16.9k
        break;
865
27.5k
    }
866
27.5k
  }
867
868
  /*** Decode LPC3 ***/
869
42.2k
  if ((mod[2] < 2)) { /* else: skip LPC3 */
870
38.7k
    int nk_mode = 0;
871
38.7k
    lpc_present[3] = 1;
872
873
38.7k
    mode_lpc_bin = get_vlclbf_n(hBs, 3);
874
875
38.7k
    switch (mode_lpc_bin) {
876
6.22k
      case 1:
877
        /* LPC3: abs */
878
6.22k
        vlpc_1st_dec(hBs, lsp[3]);
879
6.22k
        break;
880
25.2k
      case 0:
881
        /* LPC3: mid */
882
429k
        for (i = 0; i < M_LP_FILTER_ORDER; i++) {
883
403k
          lsp[3][i] = (lsp[2][i] >> 1) + (lsp[4][i] >> 1);
884
403k
        }
885
25.2k
        nk_mode = 1;
886
25.2k
        break;
887
4.47k
      case 2:
888
        /* LPC3: relL */
889
76.0k
        for (i = 0; i < M_LP_FILTER_ORDER; i++) {
890
71.5k
          lsp[3][i] = lsp[2][i];
891
71.5k
        }
892
4.47k
        nk_mode = 2;
893
4.47k
        break;
894
2.76k
      case 3:
895
        /* LPC3: relR */
896
47.0k
        for (i = 0; i < M_LP_FILTER_ORDER; i++) {
897
44.3k
          lsp[3][i] = lsp[4][i];
898
44.3k
        }
899
2.76k
        nk_mode = 2;
900
2.76k
        break;
901
38.7k
    }
902
38.7k
    err = vlpc_2st_dec(hBs, lsp[3], nk_mode);
903
38.7k
    if (err != 0) {
904
2
      return err;
905
2
    }
906
38.7k
  }
907
908
42.2k
  if (!lpc0_available && !last_frame_ok) {
909
    /* LPC(0) was lost. Use next available LPC(k) instead */
910
0
    for (k = 1; k < (nbDiv + 1); k++) {
911
0
      if (lpc_present[k]) {
912
0
        for (i = 0; i < M_LP_FILTER_ORDER; i++) {
913
0
#define LSF_INIT_TILT (0.25f)
914
0
          if (mod[0] > 0) {
915
0
            lsp[0][i] = FX_DBL2FX_LPC(
916
0
                fMult(lsp[k][i], FL2FXCONST_SGL(1.0f - LSF_INIT_TILT)) +
917
0
                fMult(fdk_dec_lsf_init[i], FL2FXCONST_SGL(LSF_INIT_TILT)));
918
0
          } else {
919
0
            lsp[0][i] = lsp[k][i];
920
0
          }
921
0
        }
922
0
        break;
923
0
      }
924
0
    }
925
0
  }
926
927
717k
  for (i = 0; i < M_LP_FILTER_ORDER; i++) {
928
675k
    lpc4_lsf[i] = lsp[4 >> s][i];
929
675k
  }
930
931
42.2k
  {
932
42.2k
    FIXP_DBL divFac;
933
42.2k
    int last, numLpc = 0;
934
935
42.2k
    i = nbDiv;
936
130k
    do {
937
130k
      numLpc += lpc_present[i--];
938
130k
    } while (i >= 0 && numLpc < 3);
939
940
42.2k
    last = i;
941
942
42.2k
    switch (numLpc) {
943
41.7k
      case 3:
944
41.7k
        divFac = FL2FXCONST_DBL(1.0f / 3.0f);
945
41.7k
        break;
946
487
      case 2:
947
487
        divFac = FL2FXCONST_DBL(1.0f / 2.0f);
948
487
        break;
949
0
      default:
950
0
        divFac = FL2FXCONST_DBL(1.0f);
951
0
        break;
952
42.2k
    }
953
954
    /* get the adaptive mean for the next (bad) frame */
955
717k
    for (k = 0; k < M_LP_FILTER_ORDER; k++) {
956
675k
      FIXP_DBL tmp = (FIXP_DBL)0;
957
2.76M
      for (i = nbDiv; i > last; i--) {
958
2.09M
        if (lpc_present[i]) {
959
2.01M
          tmp = fMultAdd(tmp >> 1, lsp[i][k], divFac);
960
2.01M
        }
961
2.09M
      }
962
675k
      lsf_adaptive_mean_cand[k] = FX_DBL2FX_LPC(tmp);
963
675k
    }
964
42.2k
  }
965
966
  /* calculate stability factor Theta. Needed for ACELP decoder and concealment
967
   */
968
0
  {
969
42.2k
    FIXP_LPC *lsf_prev, *lsf_curr;
970
42.2k
    k = 0;
971
972
42.2k
    FDK_ASSERT(lpc_present[0] == 1 && lpc_present[4 >> s] == 1);
973
42.2k
    lsf_prev = lsp[0];
974
211k
    for (i = 1; i < (nbDiv + 1); i++) {
975
168k
      if (lpc_present[i]) {
976
150k
        FIXP_DBL tmp = (FIXP_DBL)0;
977
150k
        int j;
978
150k
        lsf_curr = lsp[i];
979
980
        /* sum = tmp * 2^(LSF_SCALE*2 + 4) */
981
2.55M
        for (j = 0; j < M_LP_FILTER_ORDER; j++) {
982
2.40M
          tmp += fPow2Div2((FIXP_SGL)(lsf_curr[j] - lsf_prev[j])) >> 3;
983
2.40M
        }
984
985
        /* tmp = (float)(FL2FXCONST_DBL(1.25f) - fMult(tmp,
986
         * FL2FXCONST_DBL(1/400000.0f))); */
987
150k
        tmp = FL2FXCONST_DBL(1.25f / (1 << LSF_SCALE)) -
988
150k
              fMult(tmp, FL2FXCONST_DBL((1 << (LSF_SCALE + 4)) / 400000.0f));
989
150k
        if (tmp >= FL2FXCONST_DBL(1.0f / (1 << LSF_SCALE))) {
990
37.6k
          pStability[k] = FL2FXCONST_SGL(1.0f / 2.0f);
991
112k
        } else if (tmp < FL2FXCONST_DBL(0.0f)) {
992
72.5k
          pStability[k] = FL2FXCONST_SGL(0.0f);
993
72.5k
        } else {
994
39.9k
          pStability[k] = FX_DBL2FX_SGL(tmp << (LSF_SCALE - 1));
995
39.9k
        }
996
997
150k
        lsf_prev = lsf_curr;
998
150k
        k = i;
999
150k
      } else {
1000
        /* Mark stability value as undefined. */
1001
18.6k
        pStability[i] = (FIXP_SGL)-1;
1002
18.6k
      }
1003
168k
    }
1004
42.2k
  }
1005
1006
  /* convert into LSP domain */
1007
253k
  for (i = 0; i < (nbDiv + 1); i++) {
1008
211k
    if (lpc_present[i]) {
1009
3.27M
      for (k = 0; k < M_LP_FILTER_ORDER; k++) {
1010
3.07M
        lsp[i][k] = FX_DBL2FX_LPC(
1011
3.07M
            fixp_cos(fMult(lsp[i][k],
1012
3.07M
                           FL2FXCONST_SGL((1 << LSPARG_SCALE) * M_PI / 6400.0)),
1013
3.07M
                     LSF_SCALE - LSPARG_SCALE));
1014
3.07M
      }
1015
192k
    }
1016
211k
  }
1017
1018
42.2k
  return 0;
1019
42.2k
}
1020
1021
void CLpc_Conceal(FIXP_LPC lsp[][M_LP_FILTER_ORDER],
1022
                  FIXP_LPC lpc4_lsf[M_LP_FILTER_ORDER],
1023
                  FIXP_LPC lsf_adaptive_mean[M_LP_FILTER_ORDER],
1024
7.56k
                  const int first_lpd_flag) {
1025
7.56k
  int i, j;
1026
1027
7.56k
#define BETA (FL2FXCONST_SGL(0.25f))
1028
7.56k
#define ONE_BETA (FL2FXCONST_SGL(0.75f))
1029
7.56k
#define BFI_FAC (FL2FXCONST_SGL(0.90f))
1030
7.56k
#define ONE_BFI_FAC (FL2FXCONST_SGL(0.10f))
1031
1032
  /* Frame loss concealment (could be improved) */
1033
1034
7.56k
  if (first_lpd_flag) {
1035
    /* Reset past LSF values */
1036
110k
    for (i = 0; i < M_LP_FILTER_ORDER; i++) {
1037
103k
      lsp[0][i] = lpc4_lsf[i] = fdk_dec_lsf_init[i];
1038
103k
    }
1039
6.49k
  } else {
1040
    /* old LPC4 is new LPC0 */
1041
18.1k
    for (i = 0; i < M_LP_FILTER_ORDER; i++) {
1042
17.1k
      lsp[0][i] = lpc4_lsf[i];
1043
17.1k
    }
1044
1.07k
  }
1045
1046
  /* LPC1 */
1047
128k
  for (i = 0; i < M_LP_FILTER_ORDER; i++) {
1048
121k
    FIXP_LPC lsf_mean = FX_DBL2FX_LPC(fMult(BETA, fdk_dec_lsf_init[i]) +
1049
121k
                                      fMult(ONE_BETA, lsf_adaptive_mean[i]));
1050
1051
121k
    lsp[1][i] = FX_DBL2FX_LPC(fMult(BFI_FAC, lpc4_lsf[i]) +
1052
121k
                              fMult(ONE_BFI_FAC, lsf_mean));
1053
121k
  }
1054
1055
  /* LPC2 - LPC4 */
1056
30.2k
  for (j = 2; j <= 4; j++) {
1057
385k
    for (i = 0; i < M_LP_FILTER_ORDER; i++) {
1058
      /* lsf_mean[i] =  FX_DBL2FX_LPC(fMult((FIXP_LPC)(BETA + j *
1059
         FL2FXCONST_LPC(0.1f)), fdk_dec_lsf_init[i])
1060
                                    + fMult((FIXP_LPC)(ONE_BETA - j *
1061
         FL2FXCONST_LPC(0.1f)), lsf_adaptive_mean[i])); */
1062
1063
363k
      FIXP_LPC lsf_mean = FX_DBL2FX_LPC(
1064
363k
          fMult((FIXP_SGL)(BETA + (FIXP_SGL)(j * (INT)FL2FXCONST_SGL(0.1f))),
1065
363k
                (FIXP_SGL)fdk_dec_lsf_init[i]) +
1066
363k
          fMult(
1067
363k
              (FIXP_SGL)(ONE_BETA - (FIXP_SGL)(j * (INT)FL2FXCONST_SGL(0.1f))),
1068
363k
              lsf_adaptive_mean[i]));
1069
1070
363k
      lsp[j][i] = FX_DBL2FX_LPC(fMult(BFI_FAC, lsp[j - 1][i]) +
1071
363k
                                fMult(ONE_BFI_FAC, lsf_mean));
1072
363k
    }
1073
22.6k
  }
1074
1075
  /* Update past values for the future */
1076
128k
  for (i = 0; i < M_LP_FILTER_ORDER; i++) {
1077
121k
    lpc4_lsf[i] = lsp[4][i];
1078
121k
  }
1079
1080
  /* convert into LSP domain */
1081
45.3k
  for (j = 0; j < 5; j++) {
1082
643k
    for (i = 0; i < M_LP_FILTER_ORDER; i++) {
1083
605k
      lsp[j][i] = FX_DBL2FX_LPC(fixp_cos(
1084
605k
          fMult(lsp[j][i], FL2FXCONST_SGL((1 << LSPARG_SCALE) * M_PI / 6400.0)),
1085
605k
          LSF_SCALE - LSPARG_SCALE));
1086
605k
    }
1087
37.8k
  }
1088
7.56k
}
1089
1090
73.0k
void E_LPC_a_weight(FIXP_LPC *wA, const FIXP_LPC *A, int m) {
1091
73.0k
  FIXP_DBL f;
1092
73.0k
  int i;
1093
1094
73.0k
  f = FL2FXCONST_DBL(0.92f);
1095
1.24M
  for (i = 0; i < m; i++) {
1096
1.16M
    wA[i] = FX_DBL2FX_LPC(fMult(A[i], f));
1097
1.16M
    f = fMult(f, FL2FXCONST_DBL(0.92f));
1098
1.16M
  }
1099
73.0k
}
1100
1101
90.9k
void CLpd_DecodeGain(FIXP_DBL *gain, INT *gain_e, int gain_code) {
1102
  /* gain * 2^(gain_e) = 10^(gain_code/28) */
1103
90.9k
  *gain = fLdPow(
1104
90.9k
      FL2FXCONST_DBL(3.3219280948873623478703194294894 / 4.0), /* log2(10)*/
1105
90.9k
      2,
1106
90.9k
      fMultDiv2((FIXP_DBL)gain_code << (DFRACT_BITS - 1 - 7),
1107
90.9k
                FL2FXCONST_DBL(2.0f / 28.0f)),
1108
90.9k
      7, gain_e);
1109
90.9k
}
1110
1111
  /**
1112
   * \brief *   Find the polynomial F1(z) or F2(z) from the LSPs.
1113
   * This is performed by expanding the product polynomials:
1114
   *
1115
   * F1(z) =   product   ( 1 - 2 LSP_i z^-1 + z^-2 )
1116
   *         i=0,2,4,6,8
1117
   * F2(z) =   product   ( 1 - 2 LSP_i z^-1 + z^-2 )
1118
   *         i=1,3,5,7,9
1119
   *
1120
   * where LSP_i are the LSPs in the cosine domain.
1121
   * R.A.Salami    October 1990
1122
   * \param lsp input, line spectral freq. (cosine domain)
1123
   * \param f output, the coefficients of F1 or F2, scaled by 8 bits
1124
   * \param n no of coefficients (m/2)
1125
   * \param flag 1 : F1(z) ; 2 : F2(z)
1126
   */
1127
1128
8.16M
#define SF_F 8
1129
1130
960k
static void get_lsppol(FIXP_LPC lsp[], FIXP_DBL f[], int n, int flag) {
1131
960k
  FIXP_DBL b;
1132
960k
  FIXP_LPC *plsp;
1133
960k
  int i, j;
1134
1135
960k
  plsp = lsp + flag - 1;
1136
960k
  f[0] = FL2FXCONST_DBL(1.0f / (1 << SF_F));
1137
960k
  b = -FX_LPC2FX_DBL(*plsp);
1138
960k
  f[1] = b >> (SF_F - 1);
1139
7.68M
  for (i = 2; i <= n; i++) {
1140
6.72M
    plsp += 2;
1141
6.72M
    b = -FX_LPC2FX_DBL(*plsp);
1142
6.72M
    f[i] = SATURATE_LEFT_SHIFT((fMultDiv2(b, f[i - 1]) + (f[i - 2] >> 1)), 2,
1143
6.72M
                               DFRACT_BITS);
1144
26.8M
    for (j = i - 1; j > 1; j--) {
1145
20.1M
      f[j] = SATURATE_LEFT_SHIFT(
1146
20.1M
          ((f[j] >> 2) + fMultDiv2(b, f[j - 1]) + (f[j - 2] >> 2)), 2,
1147
20.1M
          DFRACT_BITS);
1148
20.1M
    }
1149
6.72M
    f[1] = f[1] + (b >> (SF_F - 1));
1150
6.72M
  }
1151
960k
  return;
1152
960k
}
1153
1154
6.72M
#define NC M_LP_FILTER_ORDER / 2
1155
1156
/**
1157
 * \brief lsp input LSP vector
1158
 * \brief a output LP filter coefficient vector scaled by SF_A_COEFFS.
1159
 */
1160
480k
void E_LPC_f_lsp_a_conversion(FIXP_LPC *lsp, FIXP_LPC *a, INT *a_exp) {
1161
480k
  FIXP_DBL f1[NC + 1], f2[NC + 1];
1162
480k
  int i, k;
1163
1164
  /*-----------------------------------------------------*
1165
   *  Find the polynomials F1(z) and F2(z)               *
1166
   *-----------------------------------------------------*/
1167
1168
480k
  get_lsppol(lsp, f1, NC, 1);
1169
480k
  get_lsppol(lsp, f2, NC, 2);
1170
1171
  /*-----------------------------------------------------*
1172
   *  Multiply F1(z) by (1+z^-1) and F2(z) by (1-z^-1)   *
1173
   *-----------------------------------------------------*/
1174
480k
  scaleValues(f1, NC + 1, -2);
1175
480k
  scaleValues(f2, NC + 1, -2);
1176
1177
4.32M
  for (i = NC; i > 0; i--) {
1178
3.84M
    f1[i] += f1[i - 1];
1179
3.84M
    f2[i] -= f2[i - 1];
1180
3.84M
  }
1181
1182
480k
  FIXP_DBL aDBL[M_LP_FILTER_ORDER];
1183
1184
4.32M
  for (i = 1, k = M_LP_FILTER_ORDER - 1; i <= NC; i++, k--) {
1185
3.84M
    aDBL[i - 1] = f1[i] + f2[i];
1186
3.84M
    aDBL[k] = f1[i] - f2[i];
1187
3.84M
  }
1188
1189
480k
  int headroom_a = getScalefactor(aDBL, M_LP_FILTER_ORDER);
1190
1191
8.16M
  for (i = 0; i < M_LP_FILTER_ORDER; i++) {
1192
7.68M
    a[i] = FX_DBL2FX_LPC(aDBL[i] << headroom_a);
1193
7.68M
  }
1194
1195
480k
  *a_exp = SF_F + (2 - 1) - headroom_a;
1196
480k
}