1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_mse_q7.c
4 * Description: Mean square error between two Q7 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 @defgroup MSE Mean Square Error
37
38 Calculates the mean square error between two vectors.
39
40 */
41
42 /**
43 @addtogroup MSE
44 @{
45 */
46
47 /**
48 @brief Mean square error between two Q7 vectors.
49 @param[in] pSrcA points to the first input vector
50 @param[in] pSrcB points to the second input vector
51 @param[in] blockSize number of samples in input vector
52 @param[out] pResult mean square error
53 */
54 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
arm_mse_q7(const q7_t * pSrcA,const q7_t * pSrcB,uint32_t blockSize,q7_t * pResult)55 void arm_mse_q7(
56 const q7_t * pSrcA,
57 const q7_t * pSrcB,
58 uint32_t blockSize,
59 q7_t * pResult)
60 {
61 uint32_t blkCnt; /* loop counters */
62 q7x16_t vecSrcA,vecSrcB;
63 q31_t sum = 0LL;
64
65 /* Compute 16 outputs at a time */
66 blkCnt = blockSize >> 4U;
67 while (blkCnt > 0U)
68 {
69 vecSrcA = vld1q(pSrcA);
70 vecSrcB = vld1q(pSrcB);
71
72 vecSrcA = vshrq(vecSrcA,1);
73 vecSrcB = vshrq(vecSrcB,1);
74
75 vecSrcA = vqsubq(vecSrcA,vecSrcB);
76 /*
77 * sum lanes
78 */
79 sum = vmladavaq(sum, vecSrcA, vecSrcA);
80
81 blkCnt--;
82 pSrcA += 16;
83 pSrcB += 16;
84 }
85
86 /*
87 * tail
88 */
89 blkCnt = blockSize & 0xF;
90 if (blkCnt > 0U)
91 {
92 mve_pred16_t p0 = vctp8q(blkCnt);
93 vecSrcA = vld1q(pSrcA);
94 vecSrcB = vld1q(pSrcB);
95
96 vecSrcA = vshrq(vecSrcA,1);
97 vecSrcB = vshrq(vecSrcB,1);
98
99 vecSrcA = vqsubq(vecSrcA,vecSrcB);
100
101 sum = vmladavaq_p(sum, vecSrcA, vecSrcA, p0);
102 }
103
104 *pResult = (q7_t) __SSAT((q15_t) (sum / blockSize)>>5, 8);
105 }
106 #else
arm_mse_q7(const q7_t * pSrcA,const q7_t * pSrcB,uint32_t blockSize,q7_t * pResult)107 void arm_mse_q7(
108 const q7_t * pSrcA,
109 const q7_t * pSrcB,
110 uint32_t blockSize,
111 q7_t * pResult)
112 {
113 uint32_t blkCnt; /* Loop counter */
114 q31_t sum = 0; /* Temporary result storage */
115 q7_t inA,inB; /* Temporary variable to store input value */
116
117
118 #if defined (ARM_MATH_LOOPUNROLL)
119
120 /* Loop unrolling: Compute 4 outputs at a time */
121 blkCnt = blockSize >> 2U;
122
123 while (blkCnt > 0U)
124 {
125 inA = *pSrcA++ >> 1;
126 inB = *pSrcB++ >> 1;
127 inA = (q7_t) __SSAT((q15_t) inA - (q15_t)inB, 8);
128 sum += ((q15_t) inA * inA);
129
130 inA = *pSrcA++ >> 1;
131 inB = *pSrcB++ >> 1;
132 inA = (q7_t) __SSAT((q15_t) inA - (q15_t)inB, 8);
133 sum += ((q15_t) inA * inA);
134
135 inA = *pSrcA++ >> 1;
136 inB = *pSrcB++ >> 1;
137 inA = (q7_t) __SSAT((q15_t) inA - (q15_t)inB, 8);
138 sum += ((q15_t) inA * inA);
139
140 inA = *pSrcA++ >> 1;
141 inB = *pSrcB++ >> 1;
142 inA = (q7_t) __SSAT((q15_t) inA - (q15_t)inB, 8);
143 sum += ((q15_t) inA * inA);
144
145 /* Decrement loop counter */
146 blkCnt--;
147 }
148
149 /* Loop unrolling: Compute remaining outputs */
150 blkCnt = blockSize % 0x4U;
151
152 #else
153
154 /* Initialize blkCnt with number of samples */
155 blkCnt = blockSize;
156
157 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
158
159 while (blkCnt > 0U)
160 {
161 inA = *pSrcA++ >> 1;
162 inB = *pSrcB++ >> 1;
163
164 inA = (q7_t) __SSAT((q15_t) inA - (q15_t)inB, 8);
165 sum += ((q15_t) inA * inA);
166
167 /* Decrement loop counter */
168 blkCnt--;
169 }
170
171 /* Store result in q7 format */
172 *pResult = (q7_t) __SSAT((q15_t) (sum / blockSize)>>5, 8);;
173 }
174 #endif /* defined(ARM_MATH_MVEI) */
175
176 /**
177 @} end of MSE group
178 */
179
