1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_rfft_q15.c
4 * Description: RFFT & RIFFT Q15 process function
5 *
6 * $Date: 23 April 2021
7 * $Revision: V1.9.0
8 *
9 * Target Processor: Cortex-M and Cortex-A cores
10 * -------------------------------------------------------------------- */
11 /*
12 * Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
13 *
14 * SPDX-License-Identifier: Apache-2.0
15 *
16 * Licensed under the Apache License, Version 2.0 (the License); you may
17 * not use this file except in compliance with the License.
18 * You may obtain a copy of the License at
19 *
20 * www.apache.org/licenses/LICENSE-2.0
21 *
22 * Unless required by applicable law or agreed to in writing, software
23 * distributed under the License is distributed on an AS IS BASIS, WITHOUT
24 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
25 * See the License for the specific language governing permissions and
26 * limitations under the License.
27 */
28
29 #include "dsp/transform_functions.h"
30
31 /* ----------------------------------------------------------------------
32 * Internal functions prototypes
33 * -------------------------------------------------------------------- */
34
35 void arm_split_rfft_q15(
36 q15_t * pSrc,
37 uint32_t fftLen,
38 const q15_t * pATable,
39 const q15_t * pBTable,
40 q15_t * pDst,
41 uint32_t modifier);
42
43 void arm_split_rifft_q15(
44 q15_t * pSrc,
45 uint32_t fftLen,
46 const q15_t * pATable,
47 const q15_t * pBTable,
48 q15_t * pDst,
49 uint32_t modifier);
50
51 /**
52 @addtogroup RealFFTQ15
53 @{
54 */
55
56 /**
57 @brief Processing function for the Q15 RFFT/RIFFT.
58 @param[in] S points to an instance of the Q15 RFFT/RIFFT structure
59 @param[in] pSrc points to input buffer (Source buffer is modified by this function.)
60 @param[out] pDst points to output buffer
61
62 @par Input an output formats
63 Internally input is downscaled by 2 for every stage to avoid saturations inside CFFT/CIFFT process.
64 Hence the output format is different for different RFFT sizes.
65 The input and output formats for different RFFT sizes and number of bits to upscale are mentioned in the tables below for RFFT and RIFFT:
66 @par Input and Output formats for RFFT Q15
67
68 | RFFT Size | Input Format | Output Format | Number of bits to upscale |
69 | ---------: | ------------: | -------------: | ------------------------: |
70 | 32 | 1.15 | 6.10 | 5 |
71 | 64 | 1.15 | 7.9 | 6 |
72 | 128 | 1.15 | 8.8 | 7 |
73 | 256 | 1.15 | 9.7 | 8 |
74 | 512 | 1.15 | 10.6 | 9 |
75 | 1024 | 1.15 | 11.5 | 10 |
76 | 2048 | 1.15 | 12.4 | 11 |
77 | 4096 | 1.15 | 13.3 | 12 |
78 | 8192 | 1.15 | 14.2 | 13 |
79
80 @par Input and Output formats for RIFFT Q15
81
82 | RIFFT Size | Input Format | Output Format | Number of bits to upscale |
83 | ----------: | ------------: | -------------: | ------------------------: |
84 | 32 | 1.15 | 6.10 | 0 |
85 | 64 | 1.15 | 7.9 | 0 |
86 | 128 | 1.15 | 8.8 | 0 |
87 | 256 | 1.15 | 9.7 | 0 |
88 | 512 | 1.15 | 10.6 | 0 |
89 | 1024 | 1.15 | 11.5 | 0 |
90 | 2048 | 1.15 | 12.4 | 0 |
91 | 4096 | 1.15 | 13.3 | 0 |
92 | 8192 | 1.15 | 14.2 | 0 |
93
94 @par
95 If the input buffer is of length N (fftLenReal), the output buffer must have length 2N
96 since it is containing the conjugate part (except for MVE version where N+2 is enough).
97 The input buffer is modified by this function.
98 @par
99 For the RIFFT, the source buffer must have length N+2 since the Nyquist frequency value
100 is needed but conjugate part is ignored.
101 It is not using the packing trick of the float version.
102 */
103
arm_rfft_q15(const arm_rfft_instance_q15 * S,q15_t * pSrc,q15_t * pDst)104 void arm_rfft_q15(
105 const arm_rfft_instance_q15 * S,
106 q15_t * pSrc,
107 q15_t * pDst)
108 {
109 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
110 const arm_cfft_instance_q15 *S_CFFT = &(S->cfftInst);
111 #else
112 const arm_cfft_instance_q15 *S_CFFT = S->pCfft;
113 #endif
114 uint32_t L2 = S->fftLenReal >> 1U;
115
116 /* Calculation of RIFFT of input */
117 if (S->ifftFlagR == 1U)
118 {
119 /* Real IFFT core process */
120 arm_split_rifft_q15 (pSrc, L2, S->pTwiddleAReal, S->pTwiddleBReal, pDst, S->twidCoefRModifier);
121
122 /* Complex IFFT process */
123 arm_cfft_q15 (S_CFFT, pDst, S->ifftFlagR, S->bitReverseFlagR);
124
125 arm_shift_q15(pDst, 1, pDst, S->fftLenReal);
126 }
127 else
128 {
129 /* Calculation of RFFT of input */
130
131 /* Complex FFT process */
132 arm_cfft_q15 (S_CFFT, pSrc, S->ifftFlagR, S->bitReverseFlagR);
133
134 /* Real FFT core process */
135 arm_split_rfft_q15 (pSrc, L2, S->pTwiddleAReal, S->pTwiddleBReal, pDst, S->twidCoefRModifier);
136 }
137
138 }
139
140 /**
141 @} end of RealFFTQ15 group
142 */
143
144 /**
145 @brief Core Real FFT process
146 @param[in] pSrc points to input buffer
147 @param[in] fftLen length of FFT
148 @param[in] pATable points to twiddle Coef A buffer
149 @param[in] pBTable points to twiddle Coef B buffer
150 @param[out] pDst points to output buffer
151 @param[in] modifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table
152
153 @par
154 The function implements a Real FFT
155 */
156
157 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
158
159 #include "arm_helium_utils.h"
160 #include "arm_vec_fft.h"
161
162 #if defined(__CMSIS_GCC_H)
163 #define MVE_CMPLX_MULT_FX_AxB_S16(A,B) vqdmladhxq_s16(vqdmlsdhq_s16((__typeof(A))vuninitializedq_s16(), A, B), A, B)
164 #define MVE_CMPLX_MULT_FX_AxConjB_S16(A,B) vqdmladhq_s16(vqdmlsdhxq_s16((__typeof(A))vuninitializedq_s16(), A, B), A, B)
165
166 #endif
167
arm_split_rfft_q15(q15_t * pSrc,uint32_t fftLen,const q15_t * pATable,const q15_t * pBTable,q15_t * pDst,uint32_t modifier)168 void arm_split_rfft_q15(
169 q15_t * pSrc,
170 uint32_t fftLen,
171 const q15_t * pATable,
172 const q15_t * pBTable,
173 q15_t * pDst,
174 uint32_t modifier)
175 {
176 uint32_t i; /* Loop Counter */
177 const q15_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */
178 q15_t *pOut1 = &pDst[2];
179 q15_t *pIn1 = &pSrc[2];
180 uint16x8_t offsetIn = { 6, 7, 4, 5, 2, 3, 0, 1 };
181 uint16x8_t offsetCoef;
182 const uint16_t offsetCoefArr[16] = {
183 0, 0, 2, 2, 4, 4, 6, 6,
184 0, 1, 0, 1, 0, 1, 0, 1
185 };
186
187 offsetCoef = vmulq_n_u16(vld1q_u16(offsetCoefArr), modifier) + vld1q_u16(offsetCoefArr + 8);
188 offsetIn = vaddq_n_u16(offsetIn, (2 * fftLen - 8));
189
190 /* Init coefficient pointers */
191 pCoefA = &pATable[modifier * 2];
192 pCoefB = &pBTable[modifier * 2];
193
194 const q15_t *pCoefAb, *pCoefBb;
195 pCoefAb = pCoefA;
196 pCoefBb = pCoefB;
197
198 pIn1 = &pSrc[2];
199
200 i = fftLen - 1U;
201 i = i / 4 + 1;
202 while (i > 0U) {
203 q15x8_t in1 = vld1q_s16(pIn1);
204 q15x8_t in2 = vldrhq_gather_shifted_offset_s16(pSrc, offsetIn);
205 q15x8_t coefA = vldrhq_gather_shifted_offset_s16(pCoefAb, offsetCoef);
206 q15x8_t coefB = vldrhq_gather_shifted_offset_s16(pCoefBb, offsetCoef);
207
208 #if defined(__CMSIS_GCC_H)
209 q15x8_t out = vhaddq_s16(MVE_CMPLX_MULT_FX_AxB_S16(in1, coefA),
210 MVE_CMPLX_MULT_FX_AxConjB_S16(coefB, in2));
211 #else
212 q15x8_t out = vhaddq_s16(MVE_CMPLX_MULT_FX_AxB(in1, coefA, q15x8_t),
213 MVE_CMPLX_MULT_FX_AxConjB(coefB, in2, q15x8_t));
214 #endif
215 vst1q_s16(pOut1, out);
216 pOut1 += 8;
217
218 offsetCoef = vaddq_n_u16(offsetCoef, modifier * 8);
219 offsetIn -= 8;
220 pIn1 += 8;
221 i -= 1;
222 }
223
224 pDst[2 * fftLen] = (pSrc[0] - pSrc[1]) >> 1U;
225 pDst[2 * fftLen + 1] = 0;
226
227 pDst[0] = (pSrc[0] + pSrc[1]) >> 1U;
228 pDst[1] = 0;
229 }
230 #else
arm_split_rfft_q15(q15_t * pSrc,uint32_t fftLen,const q15_t * pATable,const q15_t * pBTable,q15_t * pDst,uint32_t modifier)231 void arm_split_rfft_q15(
232 q15_t * pSrc,
233 uint32_t fftLen,
234 const q15_t * pATable,
235 const q15_t * pBTable,
236 q15_t * pDst,
237 uint32_t modifier)
238 {
239 uint32_t i; /* Loop Counter */
240 q31_t outR, outI; /* Temporary variables for output */
241 const q15_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */
242 q15_t *pSrc1, *pSrc2;
243 #if defined (ARM_MATH_DSP)
244 q15_t *pD1, *pD2;
245 #endif
246
247 /* Init coefficient pointers */
248 pCoefA = &pATable[modifier * 2];
249 pCoefB = &pBTable[modifier * 2];
250
251 pSrc1 = &pSrc[2];
252 pSrc2 = &pSrc[(2U * fftLen) - 2U];
253
254 #if defined (ARM_MATH_DSP)
255
256 i = 1U;
257 pD1 = pDst + 2;
258 pD2 = pDst + (4U * fftLen) - 2;
259
260 for (i = fftLen - 1; i > 0; i--)
261 {
262 /*
263 outR = ( pSrc[2 * i] * pATable[2 * i]
264 - pSrc[2 * i + 1] * pATable[2 * i + 1]
265 + pSrc[2 * n - 2 * i] * pBTable[2 * i]
266 + pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]);
267
268 outI = ( pIn[2 * i + 1] * pATable[2 * i]
269 + pIn[2 * i] * pATable[2 * i + 1]
270 + pIn[2 * n - 2 * i] * pBTable[2 * i + 1]
271 - pIn[2 * n - 2 * i + 1] * pBTable[2 * i])
272 */
273
274
275 #ifndef ARM_MATH_BIG_ENDIAN
276 /* pSrc[2 * i] * pATable[2 * i] - pSrc[2 * i + 1] * pATable[2 * i + 1] */
277 outR = __SMUSD(read_q15x2 (pSrc1), read_q15x2((q15_t *) pCoefA));
278 #else
279 /* -(pSrc[2 * i + 1] * pATable[2 * i + 1] - pSrc[2 * i] * pATable[2 * i]) */
280 outR = -(__SMUSD(read_q15x2 (pSrc1), read_q15x2((q15_t *) pCoefA)));
281 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
282
283 /* pSrc[2 * n - 2 * i] * pBTable[2 * i] + pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]) */
284 outR = __SMLAD(read_q15x2 (pSrc2), read_q15x2((q15_t *) pCoefB), outR) >> 16U;
285
286 /* pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - pIn[2 * n - 2 * i + 1] * pBTable[2 * i] */
287 #ifndef ARM_MATH_BIG_ENDIAN
288 outI = __SMUSDX(read_q15x2_da (&pSrc2), read_q15x2((q15_t *) pCoefB));
289 #else
290 outI = __SMUSDX(read_q15x2 ((q15_t *) pCoefB), read_q15x2_da (&pSrc2));
291 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
292
293 /* (pIn[2 * i + 1] * pATable[2 * i] + pIn[2 * i] * pATable[2 * i + 1] */
294 outI = __SMLADX(read_q15x2_ia (&pSrc1), read_q15x2 ((q15_t *) pCoefA), outI);
295
296 /* write output */
297 *pD1++ = (q15_t) outR;
298 *pD1++ = outI >> 16U;
299
300 /* write complex conjugate output */
301 pD2[0] = (q15_t) outR;
302 pD2[1] = -(outI >> 16U);
303 pD2 -= 2;
304
305 /* update coefficient pointer */
306 pCoefB = pCoefB + (2U * modifier);
307 pCoefA = pCoefA + (2U * modifier);
308 }
309
310 pDst[2U * fftLen] = (pSrc[0] - pSrc[1]) >> 1U;
311 pDst[2U * fftLen + 1U] = 0;
312
313 pDst[0] = (pSrc[0] + pSrc[1]) >> 1U;
314 pDst[1] = 0;
315
316 #else
317
318 i = 1U;
319
320 while (i < fftLen)
321 {
322 /*
323 outR = ( pSrc[2 * i] * pATable[2 * i]
324 - pSrc[2 * i + 1] * pATable[2 * i + 1]
325 + pSrc[2 * n - 2 * i] * pBTable[2 * i]
326 + pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]);
327 */
328
329 outR = *pSrc1 * *pCoefA;
330 outR = outR - (*(pSrc1 + 1) * *(pCoefA + 1));
331 outR = outR + (*pSrc2 * *pCoefB);
332 outR = (outR + (*(pSrc2 + 1) * *(pCoefB + 1))) >> 16;
333
334 /*
335 outI = ( pIn[2 * i + 1] * pATable[2 * i]
336 + pIn[2 * i] * pATable[2 * i + 1]
337 + pIn[2 * n - 2 * i] * pBTable[2 * i + 1]
338 - pIn[2 * n - 2 * i + 1] * pBTable[2 * i]);
339 */
340
341 outI = *pSrc2 * *(pCoefB + 1);
342 outI = outI - (*(pSrc2 + 1) * *pCoefB);
343 outI = outI + (*(pSrc1 + 1) * *pCoefA);
344 outI = outI + (*pSrc1 * *(pCoefA + 1));
345
346 /* update input pointers */
347 pSrc1 += 2U;
348 pSrc2 -= 2U;
349
350 /* write output */
351 pDst[2U * i] = (q15_t) outR;
352 pDst[2U * i + 1U] = outI >> 16U;
353
354 /* write complex conjugate output */
355 pDst[(4U * fftLen) - (2U * i)] = (q15_t) outR;
356 pDst[((4U * fftLen) - (2U * i)) + 1U] = -(outI >> 16U);
357
358 /* update coefficient pointer */
359 pCoefB = pCoefB + (2U * modifier);
360 pCoefA = pCoefA + (2U * modifier);
361
362 i++;
363 }
364
365 pDst[2U * fftLen] = (pSrc[0] - pSrc[1]) >> 1;
366 pDst[2U * fftLen + 1U] = 0;
367
368 pDst[0] = (pSrc[0] + pSrc[1]) >> 1;
369 pDst[1] = 0;
370
371 #endif /* #if defined (ARM_MATH_DSP) */
372 }
373 #endif /* defined(ARM_MATH_MVEI) */
374
375 /**
376 @brief Core Real IFFT process
377 @param[in] pSrc points to input buffer
378 @param[in] fftLen length of FFT
379 @param[in] pATable points to twiddle Coef A buffer
380 @param[in] pBTable points to twiddle Coef B buffer
381 @param[out] pDst points to output buffer
382 @param[in] modifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table
383
384 @par
385 The function implements a Real IFFT
386 */
387
388 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
389
390 #include "arm_helium_utils.h"
391 #include "arm_vec_fft.h"
392
arm_split_rifft_q15(q15_t * pSrc,uint32_t fftLen,const q15_t * pATable,const q15_t * pBTable,q15_t * pDst,uint32_t modifier)393 void arm_split_rifft_q15(
394 q15_t * pSrc,
395 uint32_t fftLen,
396 const q15_t * pATable,
397 const q15_t * pBTable,
398 q15_t * pDst,
399 uint32_t modifier)
400 {
401 uint32_t i; /* Loop Counter */
402 const q15_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */
403 q15_t *pIn1;
404 uint16x8_t offset = { 6, 7, 4, 5, 2, 3, 0, 1 };
405 uint16x8_t offsetCoef;
406 int16x8_t conj = { 1, -1, 1, -1, 1, -1, 1, -1 }; /* conjugate */
407 const uint16_t offsetCoefArr[16] = {
408 0, 0, 2, 2, 4, 4, 6, 6,
409 0, 1, 0, 1, 0, 1, 0, 1
410 };
411
412 offsetCoef = vmulq_n_u16(vld1q_u16(offsetCoefArr), modifier) + vld1q_u16(offsetCoefArr + 8);
413
414 offset = vaddq_n_u16(offset, (2 * fftLen - 6));
415
416 /* Init coefficient pointers */
417 pCoefA = &pATable[0];
418 pCoefB = &pBTable[0];
419
420 const q15_t *pCoefAb, *pCoefBb;
421 pCoefAb = pCoefA;
422 pCoefBb = pCoefB;
423
424 pIn1 = &pSrc[0];
425
426 i = fftLen;
427 i = i / 4;
428
429 while (i > 0U) {
430 q15x8_t in1 = vld1q_s16(pIn1);
431 q15x8_t in2 = vldrhq_gather_shifted_offset_s16(pSrc, offset);
432 q15x8_t coefA = vldrhq_gather_shifted_offset_s16(pCoefAb, offsetCoef);
433 q15x8_t coefB = vldrhq_gather_shifted_offset_s16(pCoefBb, offsetCoef);
434
435 /* can we avoid the conjugate here ? */
436 q15x8_t out = vhaddq_s16(MVE_CMPLX_MULT_FX_AxConjB(in1, coefA, q15x8_t),
437 vmulq(conj, MVE_CMPLX_MULT_FX_AxB(in2, coefB, q15x8_t)));
438
439 vst1q_s16(pDst, out);
440 pDst += 8;
441
442 offsetCoef = vaddq_n_u16(offsetCoef, modifier * 8);
443 offset -= 8;
444
445 pIn1 += 8;
446 i -= 1;
447 }
448 }
449 #else
arm_split_rifft_q15(q15_t * pSrc,uint32_t fftLen,const q15_t * pATable,const q15_t * pBTable,q15_t * pDst,uint32_t modifier)450 void arm_split_rifft_q15(
451 q15_t * pSrc,
452 uint32_t fftLen,
453 const q15_t * pATable,
454 const q15_t * pBTable,
455 q15_t * pDst,
456 uint32_t modifier)
457 {
458 uint32_t i; /* Loop Counter */
459 q31_t outR, outI; /* Temporary variables for output */
460 const q15_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */
461 q15_t *pSrc1, *pSrc2;
462 q15_t *pDst1 = &pDst[0];
463
464 pCoefA = &pATable[0];
465 pCoefB = &pBTable[0];
466
467 pSrc1 = &pSrc[0];
468 pSrc2 = &pSrc[2 * fftLen];
469
470 i = fftLen;
471 while (i > 0U)
472 {
473 /*
474 outR = ( pIn[2 * i] * pATable[2 * i]
475 + pIn[2 * i + 1] * pATable[2 * i + 1]
476 + pIn[2 * n - 2 * i] * pBTable[2 * i]
477 - pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1]);
478
479 outI = ( pIn[2 * i + 1] * pATable[2 * i]
480 - pIn[2 * i] * pATable[2 * i + 1]
481 - pIn[2 * n - 2 * i] * pBTable[2 * i + 1]
482 - pIn[2 * n - 2 * i + 1] * pBTable[2 * i]);
483 */
484
485 #if defined (ARM_MATH_DSP)
486
487 #ifndef ARM_MATH_BIG_ENDIAN
488 /* pIn[2 * n - 2 * i] * pBTable[2 * i] - pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1]) */
489 outR = __SMUSD(read_q15x2(pSrc2), read_q15x2((q15_t *) pCoefB));
490 #else
491 /* -(-pIn[2 * n - 2 * i] * pBTable[2 * i] + pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1])) */
492 outR = -(__SMUSD(read_q15x2(pSrc2), read_q15x2((q15_t *) pCoefB)));
493 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
494
495 /* pIn[2 * i] * pATable[2 * i] + pIn[2 * i + 1] * pATable[2 * i + 1] + pIn[2 * n - 2 * i] * pBTable[2 * i] */
496 outR = __SMLAD(read_q15x2(pSrc1), read_q15x2 ((q15_t *) pCoefA), outR) >> 16U;
497
498 /* -pIn[2 * n - 2 * i] * pBTable[2 * i + 1] + pIn[2 * n - 2 * i + 1] * pBTable[2 * i] */
499 outI = __SMUADX(read_q15x2_da (&pSrc2), read_q15x2((q15_t *) pCoefB));
500
501 /* pIn[2 * i + 1] * pATable[2 * i] - pIn[2 * i] * pATable[2 * i + 1] */
502 #ifndef ARM_MATH_BIG_ENDIAN
503 outI = __SMLSDX(read_q15x2 ((q15_t *) pCoefA), read_q15x2_ia (&pSrc1), -outI);
504 #else
505 outI = __SMLSDX(read_q15x2_ia (&pSrc1), read_q15x2 ((q15_t *) pCoefA), -outI);
506 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
507
508 /* write output */
509 #ifndef ARM_MATH_BIG_ENDIAN
510 write_q15x2_ia (&pDst1, __PKHBT(outR, (outI >> 16U), 16));
511 #else
512 write_q15x2_ia (&pDst1, __PKHBT((outI >> 16U), outR, 16));
513 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
514
515
516 #else /* #if defined (ARM_MATH_DSP) */
517
518 outR = *pSrc2 * *pCoefB;
519 outR = outR - (*(pSrc2 + 1) * *(pCoefB + 1));
520 outR = outR + (*pSrc1 * *pCoefA);
521 outR = (outR + (*(pSrc1 + 1) * *(pCoefA + 1))) >> 16;
522
523 outI = *(pSrc1 + 1) * *pCoefA;
524 outI = outI - (*pSrc1 * *(pCoefA + 1));
525 outI = outI - (*pSrc2 * *(pCoefB + 1));
526 outI = outI - (*(pSrc2 + 1) * *(pCoefB));
527
528 /* update input pointers */
529 pSrc1 += 2U;
530 pSrc2 -= 2U;
531
532 /* write output */
533 *pDst1++ = (q15_t) outR;
534 *pDst1++ = (q15_t) (outI >> 16);
535
536 #endif /* #if defined (ARM_MATH_DSP) */
537
538 /* update coefficient pointer */
539 pCoefB = pCoefB + (2 * modifier);
540 pCoefA = pCoefA + (2 * modifier);
541
542 i--;
543 }
544
545 }
546 #endif /* defined(ARM_MATH_MVEI) */
547
