• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  *  Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
3  *
4  *  Use of this source code is governed by a BSD-style license
5  *  that can be found in the LICENSE file in the root of the source
6  *  tree. An additional intellectual property rights grant can be found
7  *  in the file PATENTS.  All contributing project authors may
8  *  be found in the AUTHORS file in the root of the source tree.
9  */
10 
11 /*
12  * lpc_masking_model.c
13  *
14  * LPC analysis and filtering functions
15  *
16  */
17 
18 #include "lpc_masking_model.h"
19 
20 #include <limits.h>  /* For LLONG_MAX and LLONG_MIN. */
21 #include "codec.h"
22 #include "entropy_coding.h"
23 #include "settings.h"
24 
25 /* The conversion is implemented by the step-down algorithm */
WebRtcSpl_AToK_JSK(int16_t * a16,int16_t useOrder,int16_t * k16)26 void WebRtcSpl_AToK_JSK(
27     int16_t *a16, /* Q11 */
28     int16_t useOrder,
29     int16_t *k16  /* Q15 */
30                         )
31 {
32   int m, k;
33   int32_t tmp32[MAX_AR_MODEL_ORDER];
34   int32_t tmp32b;
35   int32_t tmp_inv_denum32;
36   int16_t tmp_inv_denum16;
37 
38   k16[useOrder-1] = a16[useOrder] << 4;  // Q11<<4 => Q15
39 
40   for (m=useOrder-1; m>0; m--) {
41     // (1 - k^2) in Q30
42     tmp_inv_denum32 = 1073741823 - k16[m] * k16[m];
43     tmp_inv_denum16 = (int16_t)(tmp_inv_denum32 >> 15);  // (1 - k^2) in Q15.
44 
45     for (k=1; k<=m; k++) {
46       tmp32b = (a16[k] << 16) - ((k16[m] * a16[m - k + 1]) << 1);
47 
48       tmp32[k] = WebRtcSpl_DivW32W16(tmp32b, tmp_inv_denum16); //Q27/Q15 = Q12
49     }
50 
51     for (k=1; k<m; k++) {
52       a16[k] = (int16_t)(tmp32[k] >> 1);  // Q12>>1 => Q11
53     }
54 
55     tmp32[m] = WEBRTC_SPL_SAT(4092, tmp32[m], -4092);
56     k16[m - 1] = (int16_t)(tmp32[m] << 3);  // Q12<<3 => Q15
57   }
58 
59   return;
60 }
61 
62 
63 
64 
65 
WebRtcSpl_LevinsonW32_JSK(int32_t * R,int16_t * A,int16_t * K,int16_t order)66 int16_t WebRtcSpl_LevinsonW32_JSK(
67     int32_t *R,  /* (i) Autocorrelation of length >= order+1 */
68     int16_t *A,  /* (o) A[0..order] LPC coefficients (Q11) */
69     int16_t *K,  /* (o) K[0...order-1] Reflection coefficients (Q15) */
70     int16_t order /* (i) filter order */
71                                         ) {
72   int16_t i, j;
73   int16_t R_hi[LEVINSON_MAX_ORDER+1], R_low[LEVINSON_MAX_ORDER+1];
74   /* Aurocorr coefficients in high precision */
75   int16_t A_hi[LEVINSON_MAX_ORDER+1], A_low[LEVINSON_MAX_ORDER+1];
76   /* LPC coefficients in high precicion */
77   int16_t A_upd_hi[LEVINSON_MAX_ORDER+1], A_upd_low[LEVINSON_MAX_ORDER+1];
78   /* LPC coefficients for next iteration */
79   int16_t K_hi, K_low;      /* reflection coefficient in high precision */
80   int16_t Alpha_hi, Alpha_low, Alpha_exp; /* Prediction gain Alpha in high precision
81                                                    and with scale factor */
82   int16_t tmp_hi, tmp_low;
83   int32_t temp1W32, temp2W32, temp3W32;
84   int16_t norm;
85 
86   /* Normalize the autocorrelation R[0]...R[order+1] */
87 
88   norm = WebRtcSpl_NormW32(R[0]);
89 
90   for (i=order;i>=0;i--) {
91     temp1W32 = R[i] << norm;
92     /* Put R in hi and low format */
93     R_hi[i] = (int16_t)(temp1W32 >> 16);
94     R_low[i] = (int16_t)((temp1W32 - ((int32_t)R_hi[i] << 16)) >> 1);
95   }
96 
97   /* K = A[1] = -R[1] / R[0] */
98 
99   temp2W32 = (R_hi[1] << 16) + (R_low[1] << 1);  /* R[1] in Q31      */
100   temp3W32  = WEBRTC_SPL_ABS_W32(temp2W32);      /* abs R[1]         */
101   temp1W32  = WebRtcSpl_DivW32HiLow(temp3W32, R_hi[0], R_low[0]); /* abs(R[1])/R[0] in Q31 */
102   /* Put back the sign on R[1] */
103   if (temp2W32 > 0) {
104     temp1W32 = -temp1W32;
105   }
106 
107   /* Put K in hi and low format */
108   K_hi = (int16_t)(temp1W32 >> 16);
109   K_low = (int16_t)((temp1W32 - ((int32_t)K_hi << 16)) >> 1);
110 
111   /* Store first reflection coefficient */
112   K[0] = K_hi;
113 
114   temp1W32 >>= 4;  /* A[1] in Q27. */
115 
116   /* Put A[1] in hi and low format */
117   A_hi[1] = (int16_t)(temp1W32 >> 16);
118   A_low[1] = (int16_t)((temp1W32 - ((int32_t)A_hi[1] << 16)) >> 1);
119 
120   /*  Alpha = R[0] * (1-K^2) */
121 
122   temp1W32  = (((K_hi * K_low) >> 14) + K_hi * K_hi) << 1;  /* = k^2 in Q31 */
123 
124   temp1W32 = WEBRTC_SPL_ABS_W32(temp1W32);    /* Guard against <0 */
125   temp1W32 = (int32_t)0x7fffffffL - temp1W32;    /* temp1W32 = (1 - K[0]*K[0]) in Q31 */
126 
127   /* Store temp1W32 = 1 - K[0]*K[0] on hi and low format */
128   tmp_hi = (int16_t)(temp1W32 >> 16);
129   tmp_low = (int16_t)((temp1W32 - ((int32_t)tmp_hi << 16)) >> 1);
130 
131   /* Calculate Alpha in Q31 */
132   temp1W32 = (R_hi[0] * tmp_hi + ((R_hi[0] * tmp_low) >> 15) +
133       ((R_low[0] * tmp_hi) >> 15)) << 1;
134 
135   /* Normalize Alpha and put it in hi and low format */
136 
137   Alpha_exp = WebRtcSpl_NormW32(temp1W32);
138   temp1W32 <<= Alpha_exp;
139   Alpha_hi = (int16_t)(temp1W32 >> 16);
140   Alpha_low = (int16_t)((temp1W32 - ((int32_t)Alpha_hi<< 16)) >> 1);
141 
142   /* Perform the iterative calculations in the
143      Levinson Durbin algorithm */
144 
145   for (i=2; i<=order; i++)
146   {
147 
148     /*                    ----
149                           \
150         temp1W32 =  R[i] + > R[j]*A[i-j]
151                           /
152                           ----
153                           j=1..i-1
154     */
155 
156     temp1W32 = 0;
157 
158     for(j=1; j<i; j++) {
159       /* temp1W32 is in Q31 */
160       temp1W32 += ((R_hi[j] * A_hi[i - j]) << 1) +
161           ((((R_hi[j] * A_low[i - j]) >> 15) +
162               ((R_low[j] * A_hi[i - j]) >> 15)) << 1);
163     }
164 
165     temp1W32 <<= 4;
166     temp1W32 += (R_hi[i] << 16) + (R_low[i] << 1);
167 
168     /* K = -temp1W32 / Alpha */
169     temp2W32 = WEBRTC_SPL_ABS_W32(temp1W32);      /* abs(temp1W32) */
170     temp3W32 = WebRtcSpl_DivW32HiLow(temp2W32, Alpha_hi, Alpha_low); /* abs(temp1W32)/Alpha */
171 
172     /* Put the sign of temp1W32 back again */
173     if (temp1W32 > 0) {
174       temp3W32 = -temp3W32;
175     }
176 
177     /* Use the Alpha shifts from earlier to denormalize */
178     norm = WebRtcSpl_NormW32(temp3W32);
179     if ((Alpha_exp <= norm)||(temp3W32==0)) {
180       temp3W32 <<= Alpha_exp;
181     } else {
182       if (temp3W32 > 0)
183       {
184         temp3W32 = (int32_t)0x7fffffffL;
185       } else
186       {
187         temp3W32 = (int32_t)0x80000000L;
188       }
189     }
190 
191     /* Put K on hi and low format */
192     K_hi = (int16_t)(temp3W32 >> 16);
193     K_low = (int16_t)((temp3W32 - ((int32_t)K_hi << 16)) >> 1);
194 
195     /* Store Reflection coefficient in Q15 */
196     K[i-1] = K_hi;
197 
198     /* Test for unstable filter. If unstable return 0 and let the
199        user decide what to do in that case
200     */
201 
202     if ((int32_t)WEBRTC_SPL_ABS_W16(K_hi) > (int32_t)32740) {
203       return(-i); /* Unstable filter */
204     }
205 
206     /*
207       Compute updated LPC coefficient: Anew[i]
208       Anew[j]= A[j] + K*A[i-j]   for j=1..i-1
209       Anew[i]= K
210     */
211 
212     for(j=1; j<i; j++)
213     {
214       temp1W32 = (A_hi[j] << 16) + (A_low[j] << 1);  // temp1W32 = A[j] in Q27
215 
216       temp1W32 += (K_hi * A_hi[i - j] + ((K_hi * A_low[i - j]) >> 15) +
217           ((K_low * A_hi[i - j]) >> 15)) << 1;  // temp1W32 += K*A[i-j] in Q27.
218 
219       /* Put Anew in hi and low format */
220       A_upd_hi[j] = (int16_t)(temp1W32 >> 16);
221       A_upd_low[j] = (int16_t)((temp1W32 - ((int32_t)A_upd_hi[j] << 16)) >> 1);
222     }
223 
224     temp3W32 >>= 4;  /* temp3W32 = K in Q27 (Convert from Q31 to Q27) */
225 
226     /* Store Anew in hi and low format */
227     A_upd_hi[i] = (int16_t)(temp3W32 >> 16);
228     A_upd_low[i] = (int16_t)((temp3W32 - ((int32_t)A_upd_hi[i] << 16)) >> 1);
229 
230     /*  Alpha = Alpha * (1-K^2) */
231 
232     temp1W32 = (((K_hi * K_low) >> 14) + K_hi * K_hi) << 1;  /* K*K in Q31 */
233 
234     temp1W32 = WEBRTC_SPL_ABS_W32(temp1W32);      /* Guard against <0 */
235     temp1W32 = (int32_t)0x7fffffffL - temp1W32;      /* 1 - K*K  in Q31 */
236 
237     /* Convert 1- K^2 in hi and low format */
238     tmp_hi = (int16_t)(temp1W32 >> 16);
239     tmp_low = (int16_t)((temp1W32 - ((int32_t)tmp_hi << 16)) >> 1);
240 
241     /* Calculate Alpha = Alpha * (1-K^2) in Q31 */
242     temp1W32 = (Alpha_hi * tmp_hi + ((Alpha_hi * tmp_low) >> 15) +
243         ((Alpha_low * tmp_hi) >> 15)) << 1;
244 
245     /* Normalize Alpha and store it on hi and low format */
246 
247     norm = WebRtcSpl_NormW32(temp1W32);
248     temp1W32 <<= norm;
249 
250     Alpha_hi = (int16_t)(temp1W32 >> 16);
251     Alpha_low = (int16_t)((temp1W32 - ((int32_t)Alpha_hi << 16)) >> 1);
252 
253     /* Update the total nomalization of Alpha */
254     Alpha_exp = Alpha_exp + norm;
255 
256     /* Update A[] */
257 
258     for(j=1; j<=i; j++)
259     {
260       A_hi[j] =A_upd_hi[j];
261       A_low[j] =A_upd_low[j];
262     }
263   }
264 
265   /*
266     Set A[0] to 1.0 and store the A[i] i=1...order in Q12
267     (Convert from Q27 and use rounding)
268   */
269 
270   A[0] = 2048;
271 
272   for(i=1; i<=order; i++) {
273     /* temp1W32 in Q27 */
274     temp1W32 = (A_hi[i] << 16) + (A_low[i] << 1);
275     /* Round and store upper word */
276     A[i] = (int16_t)((temp1W32 + 32768) >> 16);
277   }
278   return(1); /* Stable filters */
279 }
280 
281 
282 
283 
284 
285 /* window */
286 /* Matlab generation of floating point code:
287  *  t = (1:256)/257; r = 1-(1-t).^.45; w = sin(r*pi).^3; w = w/sum(w); plot((1:256)/8, w); grid;
288  *  for k=1:16, fprintf(1, '%.8f, ', w(k*16 + (-15:0))); fprintf(1, '\n'); end
289  * All values are multiplyed with 2^21 in fixed point code.
290  */
291 static const int16_t kWindowAutocorr[WINLEN] = {
292   0,     0,     0,     0,     0,     1,     1,     2,     2,     3,     5,     6,
293   8,    10,    12,    14,    17,    20,    24,    28,    33,    38,    43,    49,
294   56,    63,    71,    79,    88,    98,   108,   119,   131,   143,   157,   171,
295   186,   202,   219,   237,   256,   275,   296,   318,   341,   365,   390,   416,
296   444,   472,   502,   533,   566,   600,   635,   671,   709,   748,   789,   831,
297   875,   920,   967,  1015,  1065,  1116,  1170,  1224,  1281,  1339,  1399,  1461,
298   1525,  1590,  1657,  1726,  1797,  1870,  1945,  2021,  2100,  2181,  2263,  2348,
299   2434,  2523,  2614,  2706,  2801,  2898,  2997,  3099,  3202,  3307,  3415,  3525,
300   3637,  3751,  3867,  3986,  4106,  4229,  4354,  4481,  4611,  4742,  4876,  5012,
301   5150,  5291,  5433,  5578,  5725,  5874,  6025,  6178,  6333,  6490,  6650,  6811,
302   6974,  7140,  7307,  7476,  7647,  7820,  7995,  8171,  8349,  8529,  8711,  8894,
303   9079,  9265,  9453,  9642,  9833, 10024, 10217, 10412, 10607, 10803, 11000, 11199,
304   11398, 11597, 11797, 11998, 12200, 12401, 12603, 12805, 13008, 13210, 13412, 13614,
305   13815, 14016, 14216, 14416, 14615, 14813, 15009, 15205, 15399, 15591, 15782, 15971,
306   16157, 16342, 16524, 16704, 16881, 17056, 17227, 17395, 17559, 17720, 17877, 18030,
307   18179, 18323, 18462, 18597, 18727, 18851, 18970, 19082, 19189, 19290, 19384, 19471,
308   19551, 19623, 19689, 19746, 19795, 19835, 19867, 19890, 19904, 19908, 19902, 19886,
309   19860, 19823, 19775, 19715, 19644, 19561, 19465, 19357, 19237, 19102, 18955, 18793,
310   18618, 18428, 18223, 18004, 17769, 17518, 17252, 16970, 16672, 16357, 16025, 15677,
311   15311, 14929, 14529, 14111, 13677, 13225, 12755, 12268, 11764, 11243, 10706, 10152,
312   9583,  8998,  8399,  7787,  7162,  6527,  5883,  5231,  4576,  3919,  3265,  2620,
313   1990,  1386,   825,   333
314 };
315 
316 
317 /* By using a hearing threshold level in dB of -28 dB (higher value gives more noise),
318    the H_T_H (in float) can be calculated as:
319    H_T_H = pow(10.0, 0.05 * (-28.0)) = 0.039810717055350
320    In Q19, H_T_H becomes round(0.039810717055350*2^19) ~= 20872, i.e.
321    H_T_H = 20872/524288.0, and H_T_HQ19 = 20872;
322 */
323 
324 
325 /* The bandwidth expansion vectors are created from:
326    kPolyVecLo=[0.900000,0.810000,0.729000,0.656100,0.590490,0.531441,0.478297,0.430467,0.387420,0.348678,0.313811,0.282430];
327    kPolyVecHi=[0.800000,0.640000,0.512000,0.409600,0.327680,0.262144];
328    round(kPolyVecLo*32768)
329    round(kPolyVecHi*32768)
330 */
331 static const int16_t kPolyVecLo[12] = {
332   29491, 26542, 23888, 21499, 19349, 17414, 15673, 14106, 12695, 11425, 10283, 9255
333 };
334 static const int16_t kPolyVecHi[6] = {
335   26214, 20972, 16777, 13422, 10737, 8590
336 };
337 
log2_Q8_LPC(uint32_t x)338 static __inline int32_t log2_Q8_LPC( uint32_t x ) {
339 
340   int32_t zeros;
341   int16_t frac;
342 
343   zeros=WebRtcSpl_NormU32(x);
344   frac = (int16_t)(((x << zeros) & 0x7FFFFFFF) >> 23);
345 
346   /* log2(x) */
347   return ((31 - zeros) << 8) + frac;
348 }
349 
350 static const int16_t kMulPitchGain = -25; /* 200/256 in Q5 */
351 static const int16_t kChngFactor = 3523; /* log10(2)*10/4*0.4/1.4=log10(2)/1.4= 0.2150 in Q14 */
352 static const int16_t kExp2 = 11819; /* 1/log(2) */
353 const int kShiftLowerBand = 11;  /* Shift value for lower band in Q domain. */
354 const int kShiftHigherBand = 12;  /* Shift value for higher band in Q domain. */
355 
WebRtcIsacfix_GetVars(const int16_t * input,const int16_t * pitchGains_Q12,uint32_t * oldEnergy,int16_t * varscale)356 void WebRtcIsacfix_GetVars(const int16_t *input, const int16_t *pitchGains_Q12,
357                            uint32_t *oldEnergy, int16_t *varscale)
358 {
359   int k;
360   uint32_t nrgQ[4];
361   int16_t nrgQlog[4];
362   int16_t tmp16, chng1, chng2, chng3, chng4, tmp, chngQ, oldNrgQlog, pgQ, pg3;
363   int32_t expPg32;
364   int16_t expPg, divVal;
365   int16_t tmp16_1, tmp16_2;
366 
367   /* Calculate energies of first and second frame halfs */
368   nrgQ[0]=0;
369   for (k = QLOOKAHEAD/2; k < (FRAMESAMPLES/4 + QLOOKAHEAD) / 2; k++) {
370     nrgQ[0] += (uint32_t)(input[k] * input[k]);
371   }
372   nrgQ[1]=0;
373   for ( ; k < (FRAMESAMPLES/2 + QLOOKAHEAD) / 2; k++) {
374     nrgQ[1] += (uint32_t)(input[k] * input[k]);
375   }
376   nrgQ[2]=0;
377   for ( ; k < (FRAMESAMPLES * 3 / 4 + QLOOKAHEAD) / 2; k++) {
378     nrgQ[2] += (uint32_t)(input[k] * input[k]);
379   }
380   nrgQ[3]=0;
381   for ( ; k < (FRAMESAMPLES + QLOOKAHEAD) / 2; k++) {
382     nrgQ[3] += (uint32_t)(input[k] * input[k]);
383   }
384 
385   for ( k=0; k<4; k++) {
386     nrgQlog[k] = (int16_t)log2_Q8_LPC(nrgQ[k]); /* log2(nrgQ) */
387   }
388   oldNrgQlog = (int16_t)log2_Q8_LPC(*oldEnergy);
389 
390   /* Calculate average level change */
391   chng1 = WEBRTC_SPL_ABS_W16(nrgQlog[3]-nrgQlog[2]);
392   chng2 = WEBRTC_SPL_ABS_W16(nrgQlog[2]-nrgQlog[1]);
393   chng3 = WEBRTC_SPL_ABS_W16(nrgQlog[1]-nrgQlog[0]);
394   chng4 = WEBRTC_SPL_ABS_W16(nrgQlog[0]-oldNrgQlog);
395   tmp = chng1+chng2+chng3+chng4;
396   chngQ = (int16_t)(tmp * kChngFactor >> 10);  /* Q12 */
397   chngQ += 2926; /* + 1.0/1.4 in Q12 */
398 
399   /* Find average pitch gain */
400   pgQ = 0;
401   for (k=0; k<4; k++)
402   {
403     pgQ += pitchGains_Q12[k];
404   }
405 
406   pg3 = (int16_t)(pgQ * pgQ >> 11);  // pgQ in Q(12+2)=Q14. Q14*Q14>>11 => Q17
407   pg3 = (int16_t)(pgQ * pg3 >> 13);  /* Q14*Q17>>13 =>Q18  */
408   /* kMulPitchGain = -25 = -200 in Q-3. */
409   pg3 = (int16_t)(pg3 * kMulPitchGain >> 5);  // Q10
410   tmp16=(int16_t)WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND(kExp2,pg3,13);/* Q13*Q10>>13 => Q10*/
411   if (tmp16<0) {
412     tmp16_2 = (0x0400 | (tmp16 & 0x03FF));
413     tmp16_1 = ((uint16_t)(tmp16 ^ 0xFFFF) >> 10) - 3;  /* Gives result in Q14 */
414     if (tmp16_1<0)
415       expPg = -(tmp16_2 << -tmp16_1);
416     else
417       expPg = -(tmp16_2 >> tmp16_1);
418   } else
419     expPg = (int16_t) -16384; /* 1 in Q14, since 2^0=1 */
420 
421   expPg32 = (int32_t)expPg << 8;  /* Q22 */
422   divVal = WebRtcSpl_DivW32W16ResW16(expPg32, chngQ); /* Q22/Q12=Q10 */
423 
424   tmp16=(int16_t)WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND(kExp2,divVal,13);/* Q13*Q10>>13 => Q10*/
425   if (tmp16<0) {
426     tmp16_2 = (0x0400 | (tmp16 & 0x03FF));
427     tmp16_1 = ((uint16_t)(tmp16 ^ 0xFFFF) >> 10) - 3;  /* Gives result in Q14 */
428     if (tmp16_1<0)
429       expPg = tmp16_2 << -tmp16_1;
430     else
431       expPg = tmp16_2 >> tmp16_1;
432   } else
433     expPg = (int16_t) 16384; /* 1 in Q14, since 2^0=1 */
434 
435   *varscale = expPg-1;
436   *oldEnergy = nrgQ[3];
437 }
438 
439 
440 
exp2_Q10_T(int16_t x)441 static __inline int16_t  exp2_Q10_T(int16_t x) { // Both in and out in Q10
442 
443   int16_t tmp16_1, tmp16_2;
444 
445   tmp16_2=(int16_t)(0x0400|(x&0x03FF));
446   tmp16_1 = -(x >> 10);
447   if(tmp16_1>0)
448     return tmp16_2 >> tmp16_1;
449   else
450     return tmp16_2 << -tmp16_1;
451 
452 }
453 
454 
455 // Declare function pointers.
456 AutocorrFix WebRtcIsacfix_AutocorrFix;
457 CalculateResidualEnergy WebRtcIsacfix_CalculateResidualEnergy;
458 
459 /* This routine calculates the residual energy for LPC.
460  * Formula as shown in comments inside.
461  */
WebRtcIsacfix_CalculateResidualEnergyC(int lpc_order,int32_t q_val_corr,int q_val_polynomial,int16_t * a_polynomial,int32_t * corr_coeffs,int * q_val_residual_energy)462 int32_t WebRtcIsacfix_CalculateResidualEnergyC(int lpc_order,
463                                                int32_t q_val_corr,
464                                                int q_val_polynomial,
465                                                int16_t* a_polynomial,
466                                                int32_t* corr_coeffs,
467                                                int* q_val_residual_energy) {
468   int i = 0, j = 0;
469   int shift_internal = 0, shift_norm = 0;
470   int32_t tmp32 = 0, word32_high = 0, word32_low = 0, residual_energy = 0;
471   int64_t sum64 = 0, sum64_tmp = 0;
472 
473   for (i = 0; i <= lpc_order; i++) {
474     for (j = i; j <= lpc_order; j++) {
475       /* For the case of i == 0: residual_energy +=
476        *    a_polynomial[j] * corr_coeffs[i] * a_polynomial[j - i];
477        * For the case of i != 0: residual_energy +=
478        *    a_polynomial[j] * corr_coeffs[i] * a_polynomial[j - i] * 2;
479        */
480 
481       tmp32 = a_polynomial[j] * a_polynomial[j - i];
482                                    /* tmp32 in Q(q_val_polynomial * 2). */
483       if (i != 0) {
484         tmp32 <<= 1;
485       }
486       sum64_tmp = (int64_t)tmp32 * (int64_t)corr_coeffs[i];
487       sum64_tmp >>= shift_internal;
488 
489       /* Test overflow and sum the result. */
490       if(((sum64_tmp > 0 && sum64 > 0) && (LLONG_MAX - sum64 < sum64_tmp)) ||
491          ((sum64_tmp < 0 && sum64 < 0) && (LLONG_MIN - sum64 > sum64_tmp))) {
492         /* Shift right for overflow. */
493         shift_internal += 1;
494         sum64 >>= 1;
495         sum64 += sum64_tmp >> 1;
496       } else {
497         sum64 += sum64_tmp;
498       }
499     }
500   }
501 
502   word32_high = (int32_t)(sum64 >> 32);
503   word32_low = (int32_t)sum64;
504 
505   // Calculate the value of shifting (shift_norm) for the 64-bit sum.
506   if(word32_high != 0) {
507     shift_norm = 32 - WebRtcSpl_NormW32(word32_high);
508     residual_energy = (int32_t)(sum64 >> shift_norm);
509   } else {
510     if((word32_low & 0x80000000) != 0) {
511       shift_norm = 1;
512       residual_energy = (uint32_t)word32_low >> 1;
513     } else {
514       shift_norm = WebRtcSpl_NormW32(word32_low);
515       residual_energy = word32_low << shift_norm;
516       shift_norm = -shift_norm;
517     }
518   }
519 
520   /* Q(q_val_polynomial * 2) * Q(q_val_corr) >> shift_internal >> shift_norm
521    *   = Q(q_val_corr - shift_internal - shift_norm + q_val_polynomial * 2)
522    */
523   *q_val_residual_energy = q_val_corr - shift_internal - shift_norm
524                            + q_val_polynomial * 2;
525 
526   return residual_energy;
527 }
528 
WebRtcIsacfix_GetLpcCoef(int16_t * inLoQ0,int16_t * inHiQ0,MaskFiltstr_enc * maskdata,int16_t snrQ10,const int16_t * pitchGains_Q12,int32_t * gain_lo_hiQ17,int16_t * lo_coeffQ15,int16_t * hi_coeffQ15)529 void WebRtcIsacfix_GetLpcCoef(int16_t *inLoQ0,
530                               int16_t *inHiQ0,
531                               MaskFiltstr_enc *maskdata,
532                               int16_t snrQ10,
533                               const int16_t *pitchGains_Q12,
534                               int32_t *gain_lo_hiQ17,
535                               int16_t *lo_coeffQ15,
536                               int16_t *hi_coeffQ15)
537 {
538   int k, n, ii;
539   int pos1, pos2;
540   int sh_lo, sh_hi, sh, ssh, shMem;
541   int16_t varscaleQ14;
542 
543   int16_t tmpQQlo, tmpQQhi;
544   int32_t tmp32;
545   int16_t tmp16,tmp16b;
546 
547   int16_t polyHI[ORDERHI+1];
548   int16_t rcQ15_lo[ORDERLO], rcQ15_hi[ORDERHI];
549 
550 
551   int16_t DataLoQ6[WINLEN], DataHiQ6[WINLEN];
552   int32_t corrloQQ[ORDERLO+2];
553   int32_t corrhiQQ[ORDERHI+1];
554   int32_t corrlo2QQ[ORDERLO+1];
555   int16_t scale;
556   int16_t QdomLO, QdomHI, newQdomHI, newQdomLO;
557 
558   int32_t res_nrgQQ;
559   int32_t sqrt_nrg;
560 
561   /* less-noise-at-low-frequencies factor */
562   int16_t aaQ14;
563 
564   /* Multiplication with 1/sqrt(12) ~= 0.28901734104046 can be done by convertion to
565      Q15, i.e. round(0.28901734104046*32768) = 9471, and use 9471/32768.0 ~= 0.289032
566   */
567   int16_t snrq;
568   int shft;
569 
570   int16_t tmp16a;
571   int32_t tmp32a, tmp32b, tmp32c;
572 
573   int16_t a_LOQ11[ORDERLO+1];
574   int16_t k_vecloQ15[ORDERLO];
575   int16_t a_HIQ12[ORDERHI+1];
576   int16_t k_vechiQ15[ORDERHI];
577 
578   int16_t stab;
579 
580   snrq=snrQ10;
581 
582   /* SNR= C * 2 ^ (D * snrq) ; C=0.289, D=0.05*log2(10)=0.166 (~=172 in Q10)*/
583   tmp16 = (int16_t)(snrq * 172 >> 10);  // Q10
584   tmp16b = exp2_Q10_T(tmp16); // Q10
585   snrq = (int16_t)(tmp16b * 285 >> 10);  // Q10
586 
587   /* change quallevel depending on pitch gains and level fluctuations */
588   WebRtcIsacfix_GetVars(inLoQ0, pitchGains_Q12, &(maskdata->OldEnergy), &varscaleQ14);
589 
590   /* less-noise-at-low-frequencies factor */
591   /* Calculation of 0.35 * (0.5 + 0.5 * varscale) in fixpoint:
592      With 0.35 in Q16 (0.35 ~= 22938/65536.0 = 0.3500061) and varscaleQ14 in Q14,
593      we get Q16*Q14>>16 = Q14
594   */
595   aaQ14 = (int16_t)((22938 * (8192 + (varscaleQ14 >> 1)) + 32768) >> 16);
596 
597   /* Calculate tmp = (1.0 + aa*aa); in Q12 */
598   tmp16 = (int16_t)(aaQ14 * aaQ14 >> 15);  // Q14*Q14>>15 = Q13
599   tmpQQlo = 4096 + (tmp16 >> 1);  // Q12 + Q13>>1 = Q12.
600 
601   /* Calculate tmp = (1.0+aa) * (1.0+aa); */
602   tmp16 = 8192 + (aaQ14 >> 1);  // 1+a in Q13.
603   tmpQQhi = (int16_t)(tmp16 * tmp16 >> 14);  // Q13*Q13>>14 = Q12
604 
605   /* replace data in buffer by new look-ahead data */
606   for (pos1 = 0; pos1 < QLOOKAHEAD; pos1++) {
607     maskdata->DataBufferLoQ0[pos1 + WINLEN - QLOOKAHEAD] = inLoQ0[pos1];
608   }
609 
610   for (k = 0; k < SUBFRAMES; k++) {
611 
612     /* Update input buffer and multiply signal with window */
613     for (pos1 = 0; pos1 < WINLEN - UPDATE/2; pos1++) {
614       maskdata->DataBufferLoQ0[pos1] = maskdata->DataBufferLoQ0[pos1 + UPDATE/2];
615       maskdata->DataBufferHiQ0[pos1] = maskdata->DataBufferHiQ0[pos1 + UPDATE/2];
616       DataLoQ6[pos1] = (int16_t)(maskdata->DataBufferLoQ0[pos1] *
617           kWindowAutocorr[pos1] >> 15);  // Q0*Q21>>15 = Q6
618       DataHiQ6[pos1] = (int16_t)(maskdata->DataBufferHiQ0[pos1] *
619           kWindowAutocorr[pos1] >> 15);  // Q0*Q21>>15 = Q6
620     }
621     pos2 = (int16_t)(k * UPDATE / 2);
622     for (n = 0; n < UPDATE/2; n++, pos1++) {
623       maskdata->DataBufferLoQ0[pos1] = inLoQ0[QLOOKAHEAD + pos2];
624       maskdata->DataBufferHiQ0[pos1] = inHiQ0[pos2++];
625       DataLoQ6[pos1] = (int16_t)(maskdata->DataBufferLoQ0[pos1] *
626           kWindowAutocorr[pos1] >> 15);  // Q0*Q21>>15 = Q6
627       DataHiQ6[pos1] = (int16_t)(maskdata->DataBufferHiQ0[pos1] *
628           kWindowAutocorr[pos1] >> 15);  // Q0*Q21>>15 = Q6
629     }
630 
631     /* Get correlation coefficients */
632     /* The highest absolute value measured inside DataLo in the test set
633        For DataHi, corresponding value was 160.
634 
635        This means that it should be possible to represent the input values
636        to WebRtcSpl_AutoCorrelation() as Q6 values (since 307*2^6 =
637        19648). Of course, Q0 will also work, but due to the low energy in
638        DataLo and DataHi, the outputted autocorrelation will be more accurate
639        and mimic the floating point code better, by being in an high as possible
640        Q-domain.
641     */
642 
643     WebRtcIsacfix_AutocorrFix(corrloQQ,DataLoQ6,WINLEN, ORDERLO+1, &scale);
644     QdomLO = 12-scale; // QdomLO is the Q-domain of corrloQQ
645     sh_lo = WebRtcSpl_NormW32(corrloQQ[0]);
646     QdomLO += sh_lo;
647     for (ii=0; ii<ORDERLO+2; ii++) {
648       corrloQQ[ii] <<= sh_lo;
649     }
650     /* It is investigated whether it was possible to use 16 bits for the
651        32-bit vector corrloQQ, but it didn't work. */
652 
653     WebRtcIsacfix_AutocorrFix(corrhiQQ,DataHiQ6,WINLEN, ORDERHI, &scale);
654 
655     QdomHI = 12-scale; // QdomHI is the Q-domain of corrhiQQ
656     sh_hi = WebRtcSpl_NormW32(corrhiQQ[0]);
657     QdomHI += sh_hi;
658     for (ii=0; ii<ORDERHI+1; ii++) {
659       corrhiQQ[ii] <<= sh_hi;
660     }
661 
662     /* less noise for lower frequencies, by filtering/scaling autocorrelation sequences */
663 
664     /* Calculate corrlo2[0] = tmpQQlo * corrlo[0] - 2.0*tmpQQlo * corrlo[1];*/
665     // |corrlo2QQ| in Q(QdomLO-5).
666     corrlo2QQ[0] = (WEBRTC_SPL_MUL_16_32_RSFT16(tmpQQlo, corrloQQ[0]) >> 1) -
667         (WEBRTC_SPL_MUL_16_32_RSFT16(aaQ14, corrloQQ[1]) >> 2);
668 
669     /* Calculate corrlo2[n] = tmpQQlo * corrlo[n] - tmpQQlo * (corrlo[n-1] + corrlo[n+1]);*/
670     for (n = 1; n <= ORDERLO; n++) {
671 
672       tmp32 = (corrloQQ[n - 1] >> 1) + (corrloQQ[n + 1] >> 1);  // Q(QdomLO-1).
673       corrlo2QQ[n] = (WEBRTC_SPL_MUL_16_32_RSFT16(tmpQQlo, corrloQQ[n]) >> 1) -
674           (WEBRTC_SPL_MUL_16_32_RSFT16(aaQ14, tmp32) >> 2);
675     }
676     QdomLO -= 5;
677 
678     /* Calculate corrhi[n] = tmpQQhi * corrhi[n]; */
679     for (n = 0; n <= ORDERHI; n++) {
680       corrhiQQ[n] = WEBRTC_SPL_MUL_16_32_RSFT16(tmpQQhi, corrhiQQ[n]); // Q(12+QdomHI-16) = Q(QdomHI-4)
681     }
682     QdomHI -= 4;
683 
684     /* add white noise floor */
685     /* corrlo2QQ is in Q(QdomLO) and corrhiQQ is in Q(QdomHI) */
686     /* Calculate corrlo2[0] += 9.5367431640625e-7; and
687        corrhi[0]  += 9.5367431640625e-7, where the constant is 1/2^20 */
688 
689     tmp32 = WEBRTC_SPL_SHIFT_W32((int32_t) 1, QdomLO-20);
690     corrlo2QQ[0] += tmp32;
691     tmp32 = WEBRTC_SPL_SHIFT_W32((int32_t) 1, QdomHI-20);
692     corrhiQQ[0]  += tmp32;
693 
694     /* corrlo2QQ is in Q(QdomLO) and corrhiQQ is in Q(QdomHI) before the following
695        code segment, where we want to make sure we get a 1-bit margin */
696     for (n = 0; n <= ORDERLO; n++) {
697       corrlo2QQ[n] >>= 1;  // Make sure we have a 1-bit margin.
698     }
699     QdomLO -= 1; // Now, corrlo2QQ is in Q(QdomLO), with a 1-bit margin
700 
701     for (n = 0; n <= ORDERHI; n++) {
702       corrhiQQ[n] >>= 1;  // Make sure we have a 1-bit margin.
703     }
704     QdomHI -= 1; // Now, corrhiQQ is in Q(QdomHI), with a 1-bit margin
705 
706 
707     newQdomLO = QdomLO;
708 
709     for (n = 0; n <= ORDERLO; n++) {
710       int32_t tmp, tmpB, tmpCorr;
711       int16_t alpha=328; //0.01 in Q15
712       int16_t beta=324; //(1-0.01)*0.01=0.0099 in Q15
713       int16_t gamma=32440; //(1-0.01)=0.99 in Q15
714 
715       if (maskdata->CorrBufLoQQ[n] != 0) {
716         shMem=WebRtcSpl_NormW32(maskdata->CorrBufLoQQ[n]);
717         sh = QdomLO - maskdata->CorrBufLoQdom[n];
718         if (sh<=shMem) {
719           tmp = WEBRTC_SPL_SHIFT_W32(maskdata->CorrBufLoQQ[n], sh); // Get CorrBufLoQQ to same domain as corrlo2
720           tmp = WEBRTC_SPL_MUL_16_32_RSFT15(alpha, tmp);
721         } else if ((sh-shMem)<7){
722           tmp = WEBRTC_SPL_SHIFT_W32(maskdata->CorrBufLoQQ[n], shMem); // Shift up CorrBufLoQQ as much as possible
723           // Shift |alpha| the number of times required to get |tmp| in QdomLO.
724           tmp = WEBRTC_SPL_MUL_16_32_RSFT15(alpha << (sh - shMem), tmp);
725         } else {
726           tmp = WEBRTC_SPL_SHIFT_W32(maskdata->CorrBufLoQQ[n], shMem); // Shift up CorrBufHiQQ as much as possible
727           // Shift |alpha| as much as possible without overflow the number of
728           // times required to get |tmp| in QdomLO.
729           tmp = WEBRTC_SPL_MUL_16_32_RSFT15(alpha << 6, tmp);
730           tmpCorr = corrloQQ[n] >> (sh - shMem - 6);
731           tmp = tmp + tmpCorr;
732           maskdata->CorrBufLoQQ[n] = tmp;
733           newQdomLO = QdomLO-(sh-shMem-6);
734           maskdata->CorrBufLoQdom[n] = newQdomLO;
735         }
736       } else
737         tmp = 0;
738 
739       tmp = tmp + corrlo2QQ[n];
740 
741       maskdata->CorrBufLoQQ[n] = tmp;
742       maskdata->CorrBufLoQdom[n] = QdomLO;
743 
744       tmp=WEBRTC_SPL_MUL_16_32_RSFT15(beta, tmp);
745       tmpB=WEBRTC_SPL_MUL_16_32_RSFT15(gamma, corrlo2QQ[n]);
746       corrlo2QQ[n] = tmp + tmpB;
747     }
748     if( newQdomLO!=QdomLO) {
749       for (n = 0; n <= ORDERLO; n++) {
750         if (maskdata->CorrBufLoQdom[n] != newQdomLO)
751           corrloQQ[n] >>= maskdata->CorrBufLoQdom[n] - newQdomLO;
752       }
753       QdomLO = newQdomLO;
754     }
755 
756 
757     newQdomHI = QdomHI;
758 
759     for (n = 0; n <= ORDERHI; n++) {
760       int32_t tmp, tmpB, tmpCorr;
761       int16_t alpha=328; //0.01 in Q15
762       int16_t beta=324; //(1-0.01)*0.01=0.0099 in Q15
763       int16_t gamma=32440; //(1-0.01)=0.99 in Q1
764       if (maskdata->CorrBufHiQQ[n] != 0) {
765         shMem=WebRtcSpl_NormW32(maskdata->CorrBufHiQQ[n]);
766         sh = QdomHI - maskdata->CorrBufHiQdom[n];
767         if (sh<=shMem) {
768           tmp = WEBRTC_SPL_SHIFT_W32(maskdata->CorrBufHiQQ[n], sh); // Get CorrBufHiQQ to same domain as corrhi
769           tmp = WEBRTC_SPL_MUL_16_32_RSFT15(alpha, tmp);
770           tmpCorr = corrhiQQ[n];
771           tmp = tmp + tmpCorr;
772           maskdata->CorrBufHiQQ[n] = tmp;
773           maskdata->CorrBufHiQdom[n] = QdomHI;
774         } else if ((sh-shMem)<7) {
775           tmp = WEBRTC_SPL_SHIFT_W32(maskdata->CorrBufHiQQ[n], shMem); // Shift up CorrBufHiQQ as much as possible
776           // Shift |alpha| the number of times required to get |tmp| in QdomHI.
777           tmp = WEBRTC_SPL_MUL_16_32_RSFT15(alpha << (sh - shMem), tmp);
778           tmpCorr = corrhiQQ[n];
779           tmp = tmp + tmpCorr;
780           maskdata->CorrBufHiQQ[n] = tmp;
781           maskdata->CorrBufHiQdom[n] = QdomHI;
782         } else {
783           tmp = WEBRTC_SPL_SHIFT_W32(maskdata->CorrBufHiQQ[n], shMem); // Shift up CorrBufHiQQ as much as possible
784           // Shift |alpha| as much as possible without overflow the number of
785           // times required to get |tmp| in QdomHI.
786           tmp = WEBRTC_SPL_MUL_16_32_RSFT15(alpha << 6, tmp);
787           tmpCorr = corrhiQQ[n] >> (sh - shMem - 6);
788           tmp = tmp + tmpCorr;
789           maskdata->CorrBufHiQQ[n] = tmp;
790           newQdomHI = QdomHI-(sh-shMem-6);
791           maskdata->CorrBufHiQdom[n] = newQdomHI;
792         }
793       } else {
794         tmp = corrhiQQ[n];
795         tmpCorr = tmp;
796         maskdata->CorrBufHiQQ[n] = tmp;
797         maskdata->CorrBufHiQdom[n] = QdomHI;
798       }
799 
800       tmp=WEBRTC_SPL_MUL_16_32_RSFT15(beta, tmp);
801       tmpB=WEBRTC_SPL_MUL_16_32_RSFT15(gamma, tmpCorr);
802       corrhiQQ[n] = tmp + tmpB;
803     }
804 
805     if( newQdomHI!=QdomHI) {
806       for (n = 0; n <= ORDERHI; n++) {
807         if (maskdata->CorrBufHiQdom[n] != newQdomHI)
808           corrhiQQ[n] >>= maskdata->CorrBufHiQdom[n] - newQdomHI;
809       }
810       QdomHI = newQdomHI;
811     }
812 
813     stab=WebRtcSpl_LevinsonW32_JSK(corrlo2QQ, a_LOQ11, k_vecloQ15, ORDERLO);
814 
815     if (stab<0) {  // If unstable use lower order
816       a_LOQ11[0]=2048;
817       for (n = 1; n <= ORDERLO; n++) {
818         a_LOQ11[n]=0;
819       }
820 
821       stab=WebRtcSpl_LevinsonW32_JSK(corrlo2QQ, a_LOQ11, k_vecloQ15, 8);
822     }
823 
824 
825     WebRtcSpl_LevinsonDurbin(corrhiQQ,  a_HIQ12,  k_vechiQ15, ORDERHI);
826 
827     /* bandwidth expansion */
828     for (n = 1; n <= ORDERLO; n++) {
829       a_LOQ11[n] = (int16_t)((kPolyVecLo[n - 1] * a_LOQ11[n] + (1 << 14)) >>
830           15);
831     }
832 
833 
834     polyHI[0] = a_HIQ12[0];
835     for (n = 1; n <= ORDERHI; n++) {
836       a_HIQ12[n] = (int16_t)(((int32_t)(kPolyVecHi[n - 1] * a_HIQ12[n]) +
837         (1 << 14)) >> 15);
838       polyHI[n] = a_HIQ12[n];
839     }
840 
841     /* Normalize the corrlo2 vector */
842     sh = WebRtcSpl_NormW32(corrlo2QQ[0]);
843     for (n = 0; n <= ORDERLO; n++) {
844       corrlo2QQ[n] <<= sh;
845     }
846     QdomLO += sh; /* Now, corrlo2QQ is still in Q(QdomLO) */
847 
848 
849     /* residual energy */
850 
851     sh_lo = 31;
852     res_nrgQQ = WebRtcIsacfix_CalculateResidualEnergy(ORDERLO, QdomLO,
853         kShiftLowerBand, a_LOQ11, corrlo2QQ, &sh_lo);
854 
855     /* Convert to reflection coefficients */
856     WebRtcSpl_AToK_JSK(a_LOQ11, ORDERLO, rcQ15_lo);
857 
858     if (sh_lo & 0x0001) {
859       res_nrgQQ >>= 1;
860       sh_lo-=1;
861     }
862 
863 
864     if( res_nrgQQ > 0 )
865     {
866       sqrt_nrg=WebRtcSpl_Sqrt(res_nrgQQ);
867 
868       /* add hearing threshold and compute the gain */
869       /* lo_coeff = varscale * S_N_R / (sqrt_nrg + varscale * H_T_H); */
870 
871       tmp32a = varscaleQ14 >> 1;  // H_T_HQ19=65536 (16-17=-1)
872       ssh = sh_lo >> 1;  // sqrt_nrg is in Qssh.
873       sh = ssh - 14;
874       tmp32b = WEBRTC_SPL_SHIFT_W32(tmp32a, sh); // Q14->Qssh
875       tmp32c = sqrt_nrg + tmp32b;  // Qssh  (denominator)
876       tmp32a = varscaleQ14 * snrq;  // Q24 (numerator)
877 
878       sh = WebRtcSpl_NormW32(tmp32c);
879       shft = 16 - sh;
880       tmp16a = (int16_t) WEBRTC_SPL_SHIFT_W32(tmp32c, -shft); // Q(ssh-shft)  (denominator)
881 
882       tmp32b = WebRtcSpl_DivW32W16(tmp32a, tmp16a); // Q(24-ssh+shft)
883       sh = ssh-shft-7;
884       *gain_lo_hiQ17 = WEBRTC_SPL_SHIFT_W32(tmp32b, sh);  // Gains in Q17
885     }
886     else
887     {
888       *gain_lo_hiQ17 = 100;  // Gains in Q17
889     }
890     gain_lo_hiQ17++;
891 
892     /* copy coefficients to output array */
893     for (n = 0; n < ORDERLO; n++) {
894       *lo_coeffQ15 = (int16_t) (rcQ15_lo[n]);
895       lo_coeffQ15++;
896     }
897     /* residual energy */
898     sh_hi = 31;
899     res_nrgQQ = WebRtcIsacfix_CalculateResidualEnergy(ORDERHI, QdomHI,
900         kShiftHigherBand, a_HIQ12, corrhiQQ, &sh_hi);
901 
902     /* Convert to reflection coefficients */
903     WebRtcSpl_LpcToReflCoef(polyHI, ORDERHI, rcQ15_hi);
904 
905     if (sh_hi & 0x0001) {
906       res_nrgQQ >>= 1;
907       sh_hi-=1;
908     }
909 
910 
911     if( res_nrgQQ > 0 )
912     {
913       sqrt_nrg=WebRtcSpl_Sqrt(res_nrgQQ);
914 
915 
916       /* add hearing threshold and compute the gain */
917       /* hi_coeff = varscale * S_N_R / (sqrt_nrg + varscale * H_T_H); */
918 
919       tmp32a = varscaleQ14 >> 1;  // H_T_HQ19=65536 (16-17=-1)
920 
921       ssh = sh_hi >> 1;  // |sqrt_nrg| is in Qssh.
922       sh = ssh - 14;
923       tmp32b = WEBRTC_SPL_SHIFT_W32(tmp32a, sh); // Q14->Qssh
924       tmp32c = sqrt_nrg + tmp32b;  // Qssh  (denominator)
925       tmp32a = varscaleQ14 * snrq;  // Q24 (numerator)
926 
927       sh = WebRtcSpl_NormW32(tmp32c);
928       shft = 16 - sh;
929       tmp16a = (int16_t) WEBRTC_SPL_SHIFT_W32(tmp32c, -shft); // Q(ssh-shft)  (denominator)
930 
931       tmp32b = WebRtcSpl_DivW32W16(tmp32a, tmp16a); // Q(24-ssh+shft)
932       sh = ssh-shft-7;
933       *gain_lo_hiQ17 = WEBRTC_SPL_SHIFT_W32(tmp32b, sh);  // Gains in Q17
934     }
935     else
936     {
937       *gain_lo_hiQ17 = 100;  // Gains in Q17
938     }
939     gain_lo_hiQ17++;
940 
941 
942     /* copy coefficients to output array */
943     for (n = 0; n < ORDERHI; n++) {
944       *hi_coeffQ15 = rcQ15_hi[n];
945       hi_coeffQ15++;
946     }
947   }
948 }
949