1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
|
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_rfft_q15.c
* Description: RFFT & RIFFT Q15 process function
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/transform_functions.h"
/* ----------------------------------------------------------------------
* Internal functions prototypes
* -------------------------------------------------------------------- */
void arm_split_rfft_q15(
q15_t * pSrc,
uint32_t fftLen,
const q15_t * pATable,
const q15_t * pBTable,
q15_t * pDst,
uint32_t modifier);
void arm_split_rifft_q15(
q15_t * pSrc,
uint32_t fftLen,
const q15_t * pATable,
const q15_t * pBTable,
q15_t * pDst,
uint32_t modifier);
/**
@addtogroup RealFFT
@{
*/
/**
@brief Processing function for the Q15 RFFT/RIFFT.
@param[in] S points to an instance of the Q15 RFFT/RIFFT structure
@param[in] pSrc points to input buffer (Source buffer is modified by this function.)
@param[out] pDst points to output buffer
@return none
@par Input an output formats
Internally input is downscaled by 2 for every stage to avoid saturations inside CFFT/CIFFT process.
Hence the output format is different for different RFFT sizes.
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:
@par
\image html RFFTQ15.gif "Input and Output Formats for Q15 RFFT"
@par
\image html RIFFTQ15.gif "Input and Output Formats for Q15 RIFFT"
@par
If the input buffer is of length N, the output buffer must have length 2*N.
The input buffer is modified by this function.
@par
For the RIFFT, the source buffer must at least have length
fftLenReal + 2.
The last two elements must be equal to what would be generated
by the RFFT:
(pSrc[0] - pSrc[1]) >> 1 and 0
*/
void arm_rfft_q15(
const arm_rfft_instance_q15 * S,
q15_t * pSrc,
q15_t * pDst)
{
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
const arm_cfft_instance_q15 *S_CFFT = &(S->cfftInst);
#else
const arm_cfft_instance_q15 *S_CFFT = S->pCfft;
#endif
uint32_t L2 = S->fftLenReal >> 1U;
/* Calculation of RIFFT of input */
if (S->ifftFlagR == 1U)
{
/* Real IFFT core process */
arm_split_rifft_q15 (pSrc, L2, S->pTwiddleAReal, S->pTwiddleBReal, pDst, S->twidCoefRModifier);
/* Complex IFFT process */
arm_cfft_q15 (S_CFFT, pDst, S->ifftFlagR, S->bitReverseFlagR);
arm_shift_q15(pDst, 1, pDst, S->fftLenReal);
}
else
{
/* Calculation of RFFT of input */
/* Complex FFT process */
arm_cfft_q15 (S_CFFT, pSrc, S->ifftFlagR, S->bitReverseFlagR);
/* Real FFT core process */
arm_split_rfft_q15 (pSrc, L2, S->pTwiddleAReal, S->pTwiddleBReal, pDst, S->twidCoefRModifier);
}
}
/**
@} end of RealFFT group
*/
/**
@brief Core Real FFT process
@param[in] pSrc points to input buffer
@param[in] fftLen length of FFT
@param[in] pATable points to twiddle Coef A buffer
@param[in] pBTable points to twiddle Coef B buffer
@param[out] pDst points to output buffer
@param[in] modifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table
@return none
@par
The function implements a Real FFT
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
#include "arm_vec_fft.h"
#if defined(__CMSIS_GCC_H)
#define MVE_CMPLX_MULT_FX_AxB_S16(A,B) vqdmladhxq_s16(vqdmlsdhq_s16((__typeof(A))vuninitializedq_s16(), A, B), A, B)
#define MVE_CMPLX_MULT_FX_AxConjB_S16(A,B) vqdmladhq_s16(vqdmlsdhxq_s16((__typeof(A))vuninitializedq_s16(), A, B), A, B)
#endif
void arm_split_rfft_q15(
q15_t * pSrc,
uint32_t fftLen,
const q15_t * pATable,
const q15_t * pBTable,
q15_t * pDst,
uint32_t modifier)
{
uint32_t i; /* Loop Counter */
const q15_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */
q15_t *pOut1 = &pDst[2];
q15_t *pIn1 = &pSrc[2];
uint16x8_t offsetIn = { 6, 7, 4, 5, 2, 3, 0, 1 };
uint16x8_t offsetCoef;
const uint16_t offsetCoefArr[16] = {
0, 0, 2, 2, 4, 4, 6, 6,
0, 1, 0, 1, 0, 1, 0, 1
};
offsetCoef = vmulq_n_u16(vld1q_u16(offsetCoefArr), modifier) + vld1q_u16(offsetCoefArr + 8);
offsetIn = vaddq_n_u16(offsetIn, (2 * fftLen - 8));
/* Init coefficient pointers */
pCoefA = &pATable[modifier * 2];
pCoefB = &pBTable[modifier * 2];
const q15_t *pCoefAb, *pCoefBb;
pCoefAb = pCoefA;
pCoefBb = pCoefB;
pIn1 = &pSrc[2];
i = fftLen - 1U;
i = i / 4 + 1;
while (i > 0U) {
q15x8_t in1 = vld1q_s16(pIn1);
q15x8_t in2 = vldrhq_gather_shifted_offset_s16(pSrc, offsetIn);
q15x8_t coefA = vldrhq_gather_shifted_offset_s16(pCoefAb, offsetCoef);
q15x8_t coefB = vldrhq_gather_shifted_offset_s16(pCoefBb, offsetCoef);
#if defined(__CMSIS_GCC_H)
q15x8_t out = vhaddq_s16(MVE_CMPLX_MULT_FX_AxB_S16(in1, coefA),
MVE_CMPLX_MULT_FX_AxConjB_S16(coefB, in2));
#else
q15x8_t out = vhaddq_s16(MVE_CMPLX_MULT_FX_AxB(in1, coefA, q15x8_t),
MVE_CMPLX_MULT_FX_AxConjB(coefB, in2, q15x8_t));
#endif
vst1q_s16(pOut1, out);
pOut1 += 8;
offsetCoef = vaddq_n_u16(offsetCoef, modifier * 8);
offsetIn -= 8;
pIn1 += 8;
i -= 1;
}
pDst[2 * fftLen] = (pSrc[0] - pSrc[1]) >> 1U;
pDst[2 * fftLen + 1] = 0;
pDst[0] = (pSrc[0] + pSrc[1]) >> 1U;
pDst[1] = 0;
}
#else
void arm_split_rfft_q15(
q15_t * pSrc,
uint32_t fftLen,
const q15_t * pATable,
const q15_t * pBTable,
q15_t * pDst,
uint32_t modifier)
{
uint32_t i; /* Loop Counter */
q31_t outR, outI; /* Temporary variables for output */
const q15_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */
q15_t *pSrc1, *pSrc2;
#if defined (ARM_MATH_DSP)
q15_t *pD1, *pD2;
#endif
/* Init coefficient pointers */
pCoefA = &pATable[modifier * 2];
pCoefB = &pBTable[modifier * 2];
pSrc1 = &pSrc[2];
pSrc2 = &pSrc[(2U * fftLen) - 2U];
#if defined (ARM_MATH_DSP)
i = 1U;
pD1 = pDst + 2;
pD2 = pDst + (4U * fftLen) - 2;
for (i = fftLen - 1; i > 0; i--)
{
/*
outR = ( pSrc[2 * i] * pATable[2 * i]
- pSrc[2 * i + 1] * pATable[2 * i + 1]
+ pSrc[2 * n - 2 * i] * pBTable[2 * i]
+ pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]);
outI = ( pIn[2 * i + 1] * pATable[2 * i]
+ pIn[2 * i] * pATable[2 * i + 1]
+ pIn[2 * n - 2 * i] * pBTable[2 * i + 1]
- pIn[2 * n - 2 * i + 1] * pBTable[2 * i])
*/
#ifndef ARM_MATH_BIG_ENDIAN
/* pSrc[2 * i] * pATable[2 * i] - pSrc[2 * i + 1] * pATable[2 * i + 1] */
outR = __SMUSD(read_q15x2 (pSrc1), read_q15x2((q15_t *) pCoefA));
#else
/* -(pSrc[2 * i + 1] * pATable[2 * i + 1] - pSrc[2 * i] * pATable[2 * i]) */
outR = -(__SMUSD(read_q15x2 (pSrc1), read_q15x2((q15_t *) pCoefA)));
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
/* pSrc[2 * n - 2 * i] * pBTable[2 * i] + pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]) */
outR = __SMLAD(read_q15x2 (pSrc2), read_q15x2((q15_t *) pCoefB), outR) >> 16U;
/* pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - pIn[2 * n - 2 * i + 1] * pBTable[2 * i] */
#ifndef ARM_MATH_BIG_ENDIAN
outI = __SMUSDX(read_q15x2_da (&pSrc2), read_q15x2((q15_t *) pCoefB));
#else
outI = __SMUSDX(read_q15x2 ((q15_t *) pCoefB), read_q15x2_da (&pSrc2));
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
/* (pIn[2 * i + 1] * pATable[2 * i] + pIn[2 * i] * pATable[2 * i + 1] */
outI = __SMLADX(read_q15x2_ia (&pSrc1), read_q15x2 ((q15_t *) pCoefA), outI);
/* write output */
*pD1++ = (q15_t) outR;
*pD1++ = outI >> 16U;
/* write complex conjugate output */
pD2[0] = (q15_t) outR;
pD2[1] = -(outI >> 16U);
pD2 -= 2;
/* update coefficient pointer */
pCoefB = pCoefB + (2U * modifier);
pCoefA = pCoefA + (2U * modifier);
}
pDst[2U * fftLen] = (pSrc[0] - pSrc[1]) >> 1U;
pDst[2U * fftLen + 1U] = 0;
pDst[0] = (pSrc[0] + pSrc[1]) >> 1U;
pDst[1] = 0;
#else
i = 1U;
while (i < fftLen)
{
/*
outR = ( pSrc[2 * i] * pATable[2 * i]
- pSrc[2 * i + 1] * pATable[2 * i + 1]
+ pSrc[2 * n - 2 * i] * pBTable[2 * i]
+ pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]);
*/
outR = *pSrc1 * *pCoefA;
outR = outR - (*(pSrc1 + 1) * *(pCoefA + 1));
outR = outR + (*pSrc2 * *pCoefB);
outR = (outR + (*(pSrc2 + 1) * *(pCoefB + 1))) >> 16;
/*
outI = ( pIn[2 * i + 1] * pATable[2 * i]
+ pIn[2 * i] * pATable[2 * i + 1]
+ pIn[2 * n - 2 * i] * pBTable[2 * i + 1]
- pIn[2 * n - 2 * i + 1] * pBTable[2 * i]);
*/
outI = *pSrc2 * *(pCoefB + 1);
outI = outI - (*(pSrc2 + 1) * *pCoefB);
outI = outI + (*(pSrc1 + 1) * *pCoefA);
outI = outI + (*pSrc1 * *(pCoefA + 1));
/* update input pointers */
pSrc1 += 2U;
pSrc2 -= 2U;
/* write output */
pDst[2U * i] = (q15_t) outR;
pDst[2U * i + 1U] = outI >> 16U;
/* write complex conjugate output */
pDst[(4U * fftLen) - (2U * i)] = (q15_t) outR;
pDst[((4U * fftLen) - (2U * i)) + 1U] = -(outI >> 16U);
/* update coefficient pointer */
pCoefB = pCoefB + (2U * modifier);
pCoefA = pCoefA + (2U * modifier);
i++;
}
pDst[2U * fftLen] = (pSrc[0] - pSrc[1]) >> 1;
pDst[2U * fftLen + 1U] = 0;
pDst[0] = (pSrc[0] + pSrc[1]) >> 1;
pDst[1] = 0;
#endif /* #if defined (ARM_MATH_DSP) */
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@brief Core Real IFFT process
@param[in] pSrc points to input buffer
@param[in] fftLen length of FFT
@param[in] pATable points to twiddle Coef A buffer
@param[in] pBTable points to twiddle Coef B buffer
@param[out] pDst points to output buffer
@param[in] modifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table
@return none
@par
The function implements a Real IFFT
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_helium_utils.h"
#include "arm_vec_fft.h"
void arm_split_rifft_q15(
q15_t * pSrc,
uint32_t fftLen,
const q15_t * pATable,
const q15_t * pBTable,
q15_t * pDst,
uint32_t modifier)
{
uint32_t i; /* Loop Counter */
const q15_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */
q15_t *pIn1;
uint16x8_t offset = { 6, 7, 4, 5, 2, 3, 0, 1 };
uint16x8_t offsetCoef;
int16x8_t conj = { 1, -1, 1, -1, 1, -1, 1, -1 }; /* conjugate */
const uint16_t offsetCoefArr[16] = {
0, 0, 2, 2, 4, 4, 6, 6,
0, 1, 0, 1, 0, 1, 0, 1
};
offsetCoef = vmulq_n_u16(vld1q_u16(offsetCoefArr), modifier) + vld1q_u16(offsetCoefArr + 8);
offset = vaddq_n_u16(offset, (2 * fftLen - 6));
/* Init coefficient pointers */
pCoefA = &pATable[0];
pCoefB = &pBTable[0];
const q15_t *pCoefAb, *pCoefBb;
pCoefAb = pCoefA;
pCoefBb = pCoefB;
pIn1 = &pSrc[0];
i = fftLen;
i = i / 4;
while (i > 0U) {
q15x8_t in1 = vld1q_s16(pIn1);
q15x8_t in2 = vldrhq_gather_shifted_offset_s16(pSrc, offset);
q15x8_t coefA = vldrhq_gather_shifted_offset_s16(pCoefAb, offsetCoef);
q15x8_t coefB = vldrhq_gather_shifted_offset_s16(pCoefBb, offsetCoef);
/* can we avoid the conjugate here ? */
q15x8_t out = vhaddq_s16(MVE_CMPLX_MULT_FX_AxConjB(in1, coefA, q15x8_t),
vmulq(conj, MVE_CMPLX_MULT_FX_AxB(in2, coefB, q15x8_t)));
vst1q_s16(pDst, out);
pDst += 8;
offsetCoef = vaddq_n_u16(offsetCoef, modifier * 8);
offset -= 8;
pIn1 += 8;
i -= 1;
}
}
#else
void arm_split_rifft_q15(
q15_t * pSrc,
uint32_t fftLen,
const q15_t * pATable,
const q15_t * pBTable,
q15_t * pDst,
uint32_t modifier)
{
uint32_t i; /* Loop Counter */
q31_t outR, outI; /* Temporary variables for output */
const q15_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */
q15_t *pSrc1, *pSrc2;
q15_t *pDst1 = &pDst[0];
pCoefA = &pATable[0];
pCoefB = &pBTable[0];
pSrc1 = &pSrc[0];
pSrc2 = &pSrc[2 * fftLen];
i = fftLen;
while (i > 0U)
{
/*
outR = ( pIn[2 * i] * pATable[2 * i]
+ pIn[2 * i + 1] * pATable[2 * i + 1]
+ pIn[2 * n - 2 * i] * pBTable[2 * i]
- pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1]);
outI = ( pIn[2 * i + 1] * pATable[2 * i]
- pIn[2 * i] * pATable[2 * i + 1]
- pIn[2 * n - 2 * i] * pBTable[2 * i + 1]
- pIn[2 * n - 2 * i + 1] * pBTable[2 * i]);
*/
#if defined (ARM_MATH_DSP)
#ifndef ARM_MATH_BIG_ENDIAN
/* pIn[2 * n - 2 * i] * pBTable[2 * i] - pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1]) */
outR = __SMUSD(read_q15x2(pSrc2), read_q15x2((q15_t *) pCoefB));
#else
/* -(-pIn[2 * n - 2 * i] * pBTable[2 * i] + pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1])) */
outR = -(__SMUSD(read_q15x2(pSrc2), read_q15x2((q15_t *) pCoefB)));
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
/* pIn[2 * i] * pATable[2 * i] + pIn[2 * i + 1] * pATable[2 * i + 1] + pIn[2 * n - 2 * i] * pBTable[2 * i] */
outR = __SMLAD(read_q15x2(pSrc1), read_q15x2 ((q15_t *) pCoefA), outR) >> 16U;
/* -pIn[2 * n - 2 * i] * pBTable[2 * i + 1] + pIn[2 * n - 2 * i + 1] * pBTable[2 * i] */
outI = __SMUADX(read_q15x2_da (&pSrc2), read_q15x2((q15_t *) pCoefB));
/* pIn[2 * i + 1] * pATable[2 * i] - pIn[2 * i] * pATable[2 * i + 1] */
#ifndef ARM_MATH_BIG_ENDIAN
outI = __SMLSDX(read_q15x2 ((q15_t *) pCoefA), read_q15x2_ia (&pSrc1), -outI);
#else
outI = __SMLSDX(read_q15x2_ia (&pSrc1), read_q15x2 ((q15_t *) pCoefA), -outI);
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
/* write output */
#ifndef ARM_MATH_BIG_ENDIAN
write_q15x2_ia (&pDst1, __PKHBT(outR, (outI >> 16U), 16));
#else
write_q15x2_ia (&pDst1, __PKHBT((outI >> 16U), outR, 16));
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
#else /* #if defined (ARM_MATH_DSP) */
outR = *pSrc2 * *pCoefB;
outR = outR - (*(pSrc2 + 1) * *(pCoefB + 1));
outR = outR + (*pSrc1 * *pCoefA);
outR = (outR + (*(pSrc1 + 1) * *(pCoefA + 1))) >> 16;
outI = *(pSrc1 + 1) * *pCoefA;
outI = outI - (*pSrc1 * *(pCoefA + 1));
outI = outI - (*pSrc2 * *(pCoefB + 1));
outI = outI - (*(pSrc2 + 1) * *(pCoefB));
/* update input pointers */
pSrc1 += 2U;
pSrc2 -= 2U;
/* write output */
*pDst1++ = (q15_t) outR;
*pDst1++ = (q15_t) (outI >> 16);
#endif /* #if defined (ARM_MATH_DSP) */
/* update coefficient pointer */
pCoefB = pCoefB + (2 * modifier);
pCoefA = pCoefA + (2 * modifier);
i--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
|
