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
|
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_cmplx_mult_cmplx_q31.c
* Description: Q31 complex-by-complex multiplication
*
* $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/complex_math_functions.h"
/**
@ingroup groupCmplxMath
*/
/**
@addtogroup CmplxByCmplxMult
@{
*/
/**
@brief Q31 complex-by-complex multiplication.
@param[in] pSrcA points to first input vector
@param[in] pSrcB points to second input vector
@param[out] pDst points to output vector
@param[in] numSamples number of samples in each vector
@return none
@par Scaling and Overflow Behavior
The function implements 1.31 by 1.31 multiplications and finally output is converted into 3.29 format.
Input down scaling is not required.
*/
#if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
void arm_cmplx_mult_cmplx_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t numSamples)
{
int32_t blkCnt;
q31x4_t vecSrcA, vecSrcB;
q31x4_t vecSrcC, vecSrcD;
q31x4_t vecDst;
blkCnt = numSamples >> 2;
blkCnt -= 1;
if (blkCnt > 0) {
/* should give more freedom to generate stall free code */
vecSrcA = vld1q(pSrcA);
vecSrcB = vld1q(pSrcB);
pSrcA += 4;
pSrcB += 4;
while (blkCnt > 0) {
/* C[2 * i] = A[2 * i] * B[2 * i] - A[2 * i + 1] * B[2 * i + 1]. */
vecDst = vqdmlsdhq(vuninitializedq_s32(), vecSrcA, vecSrcB);
vecSrcC = vld1q(pSrcA);
pSrcA += 4;
/* C[2 * i + 1] = A[2 * i] * B[2 * i + 1] + A[2 * i + 1] * B[2 * i]. */
vecDst = vqdmladhxq(vecDst, vecSrcA, vecSrcB);
vecSrcD = vld1q(pSrcB);
pSrcB += 4;
vst1q(pDst, vshrq(vecDst, 2));
pDst += 4;
vecDst = vqdmlsdhq(vuninitializedq_s32(), vecSrcC, vecSrcD);
vecSrcA = vld1q(pSrcA);
pSrcA += 4;
vecDst = vqdmladhxq(vecDst, vecSrcC, vecSrcD);
vecSrcB = vld1q(pSrcB);
pSrcB += 4;
vst1q(pDst, vshrq(vecDst, 2));
pDst += 4;
/*
* Decrement the blockSize loop counter
*/
blkCnt--;
}
/* process last elements out of the loop avoid the armclang breaking the SW pipeline */
vecDst = vqdmlsdhq(vuninitializedq_s32(), vecSrcA, vecSrcB);
vecSrcC = vld1q(pSrcA);
vecDst = vqdmladhxq(vecDst, vecSrcA, vecSrcB);
vecSrcD = vld1q(pSrcB);
vst1q(pDst, vshrq(vecDst, 2));
pDst += 4;
vecDst = vqdmlsdhq(vuninitializedq_s32(), vecSrcC, vecSrcD);
vecDst = vqdmladhxq(vecDst, vecSrcC, vecSrcD);
vst1q(pDst, vshrq(vecDst, 2));
pDst += 4;
/*
* tail
*/
blkCnt = CMPLX_DIM * (numSamples & 3);
do {
mve_pred16_t p = vctp32q(blkCnt);
pSrcA += 4;
pSrcB += 4;
vecSrcA = vldrwq_z_s32(pSrcA, p);
vecSrcB = vldrwq_z_s32(pSrcB, p);
vecDst = vqdmlsdhq_m(vuninitializedq_s32(), vecSrcA, vecSrcB, p);
vecDst = vqdmladhxq_m(vecDst, vecSrcA, vecSrcB, p);
vecDst = vshrq_m(vuninitializedq_s32(), vecDst, 2, p);
vstrwq_p_s32(pDst, vecDst, p);
pDst += 4;
blkCnt -= 4;
}
while ((int32_t) blkCnt > 0);
} else {
blkCnt = numSamples * CMPLX_DIM;
while (blkCnt > 0) {
mve_pred16_t p = vctp32q(blkCnt);
vecSrcA = vldrwq_z_s32(pSrcA, p);
vecSrcB = vldrwq_z_s32(pSrcB, p);
vecDst = vqdmlsdhq_m(vuninitializedq_s32(), vecSrcA, vecSrcB, p);
vecDst = vqdmladhxq_m(vecDst, vecSrcA, vecSrcB, p);
vecDst = vshrq_m(vuninitializedq_s32(), vecDst, 2, p);
vstrwq_p_s32(pDst, vecDst, p);
pDst += 4;
pSrcA += 4;
pSrcB += 4;
blkCnt -= 4;
}
}
}
#else
void arm_cmplx_mult_cmplx_q31(
const q31_t * pSrcA,
const q31_t * pSrcB,
q31_t * pDst,
uint32_t numSamples)
{
uint32_t blkCnt; /* Loop counter */
q31_t a, b, c, d; /* Temporary variables */
#if defined (ARM_MATH_LOOPUNROLL)
/* Loop unrolling: Compute 4 outputs at a time */
blkCnt = numSamples >> 2U;
while (blkCnt > 0U)
{
/* C[2 * i ] = A[2 * i] * B[2 * i ] - A[2 * i + 1] * B[2 * i + 1]. */
/* C[2 * i + 1] = A[2 * i] * B[2 * i + 1] + A[2 * i + 1] * B[2 * i ]. */
a = *pSrcA++;
b = *pSrcA++;
c = *pSrcB++;
d = *pSrcB++;
/* store result in 3.29 format in destination buffer. */
*pDst++ = (q31_t) ( (((q63_t) a * c) >> 33) - (((q63_t) b * d) >> 33) );
*pDst++ = (q31_t) ( (((q63_t) a * d) >> 33) + (((q63_t) b * c) >> 33) );
a = *pSrcA++;
b = *pSrcA++;
c = *pSrcB++;
d = *pSrcB++;
*pDst++ = (q31_t) ( (((q63_t) a * c) >> 33) - (((q63_t) b * d) >> 33) );
*pDst++ = (q31_t) ( (((q63_t) a * d) >> 33) + (((q63_t) b * c) >> 33) );
a = *pSrcA++;
b = *pSrcA++;
c = *pSrcB++;
d = *pSrcB++;
*pDst++ = (q31_t) ( (((q63_t) a * c) >> 33) - (((q63_t) b * d) >> 33) );
*pDst++ = (q31_t) ( (((q63_t) a * d) >> 33) + (((q63_t) b * c) >> 33) );
a = *pSrcA++;
b = *pSrcA++;
c = *pSrcB++;
d = *pSrcB++;
*pDst++ = (q31_t) ( (((q63_t) a * c) >> 33) - (((q63_t) b * d) >> 33) );
*pDst++ = (q31_t) ( (((q63_t) a * d) >> 33) + (((q63_t) b * c) >> 33) );
/* Decrement loop counter */
blkCnt--;
}
/* Loop unrolling: Compute remaining outputs */
blkCnt = numSamples % 0x4U;
#else
/* Initialize blkCnt with number of samples */
blkCnt = numSamples;
#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
while (blkCnt > 0U)
{
/* C[2 * i ] = A[2 * i] * B[2 * i ] - A[2 * i + 1] * B[2 * i + 1]. */
/* C[2 * i + 1] = A[2 * i] * B[2 * i + 1] + A[2 * i + 1] * B[2 * i ]. */
a = *pSrcA++;
b = *pSrcA++;
c = *pSrcB++;
d = *pSrcB++;
/* store result in 3.29 format in destination buffer. */
*pDst++ = (q31_t) ( (((q63_t) a * c) >> 33) - (((q63_t) b * d) >> 33) );
*pDst++ = (q31_t) ( (((q63_t) a * d) >> 33) + (((q63_t) b * c) >> 33) );
/* Decrement loop counter */
blkCnt--;
}
}
#endif /* defined(ARM_MATH_MVEI) */
/**
@} end of CmplxByCmplxMult group
*/
|
