1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_cmplx_dot_prod_q15.c
4 * Description: Processing function for the Q15 Complex Dot product
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/complex_math_functions.h"
30
31 /**
32 @ingroup groupCmplxMath
33 */
34
35 /**
36 @addtogroup cmplx_dot_prod
37 @{
38 */
39
40 /**
41 @brief Q15 complex dot product.
42 @param[in] pSrcA points to the first input vector
43 @param[in] pSrcB points to the second input vector
44 @param[in] numSamples number of samples in each vector
45 @param[out] realResult real part of the result returned here
46 @param[out] imagResult imaginary part of the result returned her
47 @return none
48
49 @par Scaling and Overflow Behavior
50 The function is implemented using an internal 64-bit accumulator.
51 The intermediate 1.15 by 1.15 multiplications are performed with full precision and yield a 2.30 result.
52 These are accumulated in a 64-bit accumulator with 34.30 precision.
53 As a final step, the accumulators are converted to 8.24 format.
54 The return results <code>realResult</code> and <code>imagResult</code> are in 8.24 format.
55 */
56
57 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
arm_cmplx_dot_prod_q15(const q15_t * pSrcA,const q15_t * pSrcB,uint32_t numSamples,q31_t * realResult,q31_t * imagResult)58 void arm_cmplx_dot_prod_q15(
59 const q15_t * pSrcA,
60 const q15_t * pSrcB,
61 uint32_t numSamples,
62 q31_t * realResult,
63 q31_t * imagResult)
64 {
65 int32_t blkCnt;
66 q63_t accReal = 0LL;
67 q63_t accImag = 0LL;
68 q15x8_t vecSrcA, vecSrcB;
69 q15x8_t vecSrcC, vecSrcD;
70
71 blkCnt = (numSamples >> 3);
72 blkCnt -= 1;
73 if (blkCnt > 0) {
74 /* should give more freedom to generate stall free code */
75 vecSrcA = vld1q(pSrcA);
76 vecSrcB = vld1q(pSrcB);
77 pSrcA += 8;
78 pSrcB += 8;
79
80 while (blkCnt > 0) {
81
82 accReal = vmlsldavaq(accReal, vecSrcA, vecSrcB);
83 vecSrcC = vld1q(pSrcA);
84 pSrcA += 8;
85
86 accImag = vmlaldavaxq(accImag, vecSrcA, vecSrcB);
87 vecSrcD = vld1q(pSrcB);
88 pSrcB += 8;
89
90 accReal = vmlsldavaq(accReal, vecSrcC, vecSrcD);
91 vecSrcA = vld1q(pSrcA);
92 pSrcA += 8;
93
94 accImag = vmlaldavaxq(accImag, vecSrcC, vecSrcD);
95 vecSrcB = vld1q(pSrcB);
96 pSrcB += 8;
97 /*
98 * Decrement the blockSize loop counter
99 */
100 blkCnt--;
101 }
102
103 /* process last elements out of the loop avoid the armclang breaking the SW pipeline */
104 accReal = vmlsldavaq(accReal, vecSrcA, vecSrcB);
105 vecSrcC = vld1q(pSrcA);
106
107 accImag = vmlaldavaxq(accImag, vecSrcA, vecSrcB);
108 vecSrcD = vld1q(pSrcB);
109
110 accReal = vmlsldavaq(accReal, vecSrcC, vecSrcD);
111 vecSrcA = vld1q(pSrcA);
112
113 accImag = vmlaldavaxq(accImag, vecSrcC, vecSrcD);
114 vecSrcB = vld1q(pSrcB);
115
116 /*
117 * tail
118 */
119 blkCnt = CMPLX_DIM * (numSamples & 7);
120 do {
121 mve_pred16_t p = vctp16q(blkCnt);
122
123 pSrcA += 8;
124 pSrcB += 8;
125
126 vecSrcA = vldrhq_z_s16(pSrcA, p);
127 vecSrcB = vldrhq_z_s16(pSrcB, p);
128
129 accReal = vmlsldavaq_p(accReal, vecSrcA, vecSrcB, p);
130 accImag = vmlaldavaxq_p(accImag, vecSrcA, vecSrcB, p);
131
132 blkCnt -= 8;
133 }
134 while ((int32_t) blkCnt > 0);
135 } else {
136 blkCnt = numSamples * CMPLX_DIM;
137 while (blkCnt > 0) {
138 mve_pred16_t p = vctp16q(blkCnt);
139
140 vecSrcA = vldrhq_z_s16(pSrcA, p);
141 vecSrcB = vldrhq_z_s16(pSrcB, p);
142
143 accReal = vmlsldavaq_p(accReal, vecSrcA, vecSrcB, p);
144 accImag = vmlaldavaxq_p(accImag, vecSrcA, vecSrcB, p);
145
146 /*
147 * Decrement the blkCnt loop counter
148 * Advance vector source and destination pointers
149 */
150 pSrcA += 8;
151 pSrcB += 8;
152 blkCnt -= 8;
153 }
154 }
155 *realResult = asrl(accReal, (14 - 8));
156 *imagResult = asrl(accImag, (14 - 8));
157 }
158 #else
arm_cmplx_dot_prod_q15(const q15_t * pSrcA,const q15_t * pSrcB,uint32_t numSamples,q31_t * realResult,q31_t * imagResult)159 void arm_cmplx_dot_prod_q15(
160 const q15_t * pSrcA,
161 const q15_t * pSrcB,
162 uint32_t numSamples,
163 q31_t * realResult,
164 q31_t * imagResult)
165 {
166 uint32_t blkCnt; /* Loop counter */
167 q63_t real_sum = 0, imag_sum = 0; /* Temporary result variables */
168 q15_t a0,b0,c0,d0;
169
170 #if defined (ARM_MATH_LOOPUNROLL)
171 /* Loop unrolling: Compute 4 outputs at a time */
172 blkCnt = numSamples >> 2U;
173
174 while (blkCnt > 0U)
175 {
176 a0 = *pSrcA++;
177 b0 = *pSrcA++;
178 c0 = *pSrcB++;
179 d0 = *pSrcB++;
180
181 real_sum += (q31_t)a0 * c0;
182 imag_sum += (q31_t)a0 * d0;
183 real_sum -= (q31_t)b0 * d0;
184 imag_sum += (q31_t)b0 * c0;
185
186 a0 = *pSrcA++;
187 b0 = *pSrcA++;
188 c0 = *pSrcB++;
189 d0 = *pSrcB++;
190
191 real_sum += (q31_t)a0 * c0;
192 imag_sum += (q31_t)a0 * d0;
193 real_sum -= (q31_t)b0 * d0;
194 imag_sum += (q31_t)b0 * c0;
195
196 a0 = *pSrcA++;
197 b0 = *pSrcA++;
198 c0 = *pSrcB++;
199 d0 = *pSrcB++;
200
201 real_sum += (q31_t)a0 * c0;
202 imag_sum += (q31_t)a0 * d0;
203 real_sum -= (q31_t)b0 * d0;
204 imag_sum += (q31_t)b0 * c0;
205
206 a0 = *pSrcA++;
207 b0 = *pSrcA++;
208 c0 = *pSrcB++;
209 d0 = *pSrcB++;
210
211 real_sum += (q31_t)a0 * c0;
212 imag_sum += (q31_t)a0 * d0;
213 real_sum -= (q31_t)b0 * d0;
214 imag_sum += (q31_t)b0 * c0;
215
216 /* Decrement loop counter */
217 blkCnt--;
218 }
219
220 /* Loop unrolling: Compute remaining outputs */
221 blkCnt = numSamples % 0x4U;
222
223 #else
224
225 /* Initialize blkCnt with number of samples */
226 blkCnt = numSamples;
227
228 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
229
230 while (blkCnt > 0U)
231 {
232 a0 = *pSrcA++;
233 b0 = *pSrcA++;
234 c0 = *pSrcB++;
235 d0 = *pSrcB++;
236
237 real_sum += (q31_t)a0 * c0;
238 imag_sum += (q31_t)a0 * d0;
239 real_sum -= (q31_t)b0 * d0;
240 imag_sum += (q31_t)b0 * c0;
241
242 /* Decrement loop counter */
243 blkCnt--;
244 }
245
246 /* Store real and imaginary result in 8.24 format */
247 /* Convert real data in 34.30 to 8.24 by 6 right shifts */
248 *realResult = (q31_t) (real_sum >> 6);
249 /* Convert imaginary data in 34.30 to 8.24 by 6 right shifts */
250 *imagResult = (q31_t) (imag_sum >> 6);
251 }
252 #endif /* defined(ARM_MATH_MVEI) */
253
254 /**
255 @} end of cmplx_dot_prod group
256 */
257