1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_mse_q31.c
4 * Description: Mean square error between two Q31 vectors
5 *
6 * $Date: 04 April 2022
7 * $Revision: V1.10.0
8 *
9 * Target Processor: Cortex-M and Cortex-A cores
10 * -------------------------------------------------------------------- */
11 /*
12 * Copyright (C) 2010-2022 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/statistics_functions.h"
30
31 /**
32 @ingroup groupStats
33 */
34
35
36 /**
37 @addtogroup MSE
38 @{
39 */
40
41 /**
42 @brief Mean square error between two Q31 vectors.
43 @param[in] pSrcA points to the first input vector
44 @param[in] pSrcB points to the second input vector
45 @param[in] blockSize number of samples in input vector
46 @param[out] pResult mean square error
47 @return none
48 */
49 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
arm_mse_q31(const q31_t * pSrcA,const q31_t * pSrcB,uint32_t blockSize,q31_t * pResult)50 void arm_mse_q31(
51 const q31_t * pSrcA,
52 const q31_t * pSrcB,
53 uint32_t blockSize,
54 q31_t * pResult)
55 {
56 uint32_t blkCnt; /* loop counters */
57 q31x4_t vecSrcA,vecSrcB;
58 q63_t sum = 0LL;
59
60 /* Compute 4 outputs at a time */
61 blkCnt = blockSize >> 2U;
62 while (blkCnt > 0U)
63 {
64 vecSrcA = vld1q(pSrcA);
65 vecSrcB = vld1q(pSrcB);
66
67 vecSrcA = vshrq(vecSrcA,1);
68 vecSrcB = vshrq(vecSrcB,1);
69
70
71 vecSrcA = vqsubq(vecSrcA,vecSrcB);
72 /*
73 * sum lanes
74 */
75 sum = vrmlaldavhaq(sum, vecSrcA, vecSrcA);
76
77 blkCnt--;
78 pSrcA += 4;
79 pSrcB += 4;
80 }
81
82 /*
83 * tail
84 */
85 blkCnt = blockSize & 3;
86 if (blkCnt > 0U)
87 {
88 mve_pred16_t p0 = vctp32q(blkCnt);
89 vecSrcA = vld1q(pSrcA);
90 vecSrcB = vld1q(pSrcB);
91
92 vecSrcA = vshrq(vecSrcA,1);
93 vecSrcB = vshrq(vecSrcB,1);
94
95 vecSrcA = vqsubq(vecSrcA,vecSrcB);
96
97 sum = vrmlaldavhaq_p(sum, vecSrcA, vecSrcA, p0);
98 }
99
100
101 *pResult = (q31_t) ((sum / blockSize)>>21);
102
103 }
104 #else
arm_mse_q31(const q31_t * pSrcA,const q31_t * pSrcB,uint32_t blockSize,q31_t * pResult)105 void arm_mse_q31(
106 const q31_t * pSrcA,
107 const q31_t * pSrcB,
108 uint32_t blockSize,
109 q31_t * pResult)
110 {
111 uint32_t blkCnt; /* Loop counter */
112 q63_t sum = 0; /* Temporary result storage */
113
114 q31_t inA32,inB32; /* Temporary variable to store packed input value */
115
116 #if defined (ARM_MATH_LOOPUNROLL)
117
118 /* Loop unrolling: Compute 4 outputs at a time */
119 blkCnt = blockSize >> 2U;
120
121 while (blkCnt > 0U)
122 {
123 inA32 = *pSrcA++ >> 1;
124 inB32 = *pSrcB++ >> 1;
125 inA32 = __QSUB(inA32, inB32);
126 sum += ((q63_t) inA32 * inA32) >> 14U;
127
128 inA32 = *pSrcA++ >> 1;
129 inB32 = *pSrcB++ >> 1;
130 inA32 = __QSUB(inA32, inB32);
131 sum += ((q63_t) inA32 * inA32) >> 14U;
132
133 inA32 = *pSrcA++ >> 1;
134 inB32 = *pSrcB++ >> 1;
135 inA32 = __QSUB(inA32, inB32);
136 sum += ((q63_t) inA32 * inA32) >> 14U;
137
138 inA32 = *pSrcA++ >> 1;
139 inB32 = *pSrcB++ >> 1;
140 inA32 = __QSUB(inA32, inB32);
141 sum += ((q63_t) inA32 * inA32) >> 14U;
142
143
144 /* Decrement loop counter */
145 blkCnt--;
146 }
147
148 /* Loop unrolling: Compute remaining outputs */
149 blkCnt = blockSize % 0x4U;
150
151 #else
152
153 /* Initialize blkCnt with number of samples */
154 blkCnt = blockSize;
155
156 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
157
158 while (blkCnt > 0U)
159 {
160 inA32 = *pSrcA++ >> 1;
161 inB32 = *pSrcB++ >> 1;
162 inA32 = __QSUB(inA32, inB32);
163 sum += ((q63_t) inA32 * inA32) >> 14U;
164
165 /* Decrement loop counter */
166 blkCnt--;
167 }
168
169 /* Store result in q31 format */
170 *pResult = (q31_t) ((sum / blockSize)>>15);
171 }
172 #endif /* defined(ARM_MATH_MVEI) */
173
174 /**
175 @} end of MSE group
176 */
177