1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_mean_q31.c
4 * Description: Mean value of a Q31 vector
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/statistics_functions.h"
30
31 /**
32 @ingroup groupStats
33 */
34
35 /**
36 @addtogroup mean
37 @{
38 */
39
40 /**
41 @brief Mean value of a Q31 vector.
42 @param[in] pSrc points to the input vector
43 @param[in] blockSize number of samples in input vector
44 @param[out] pResult mean value returned here
45
46 @par Scaling and Overflow Behavior
47 The function is implemented using a 64-bit internal accumulator.
48 The input is represented in 1.31 format and is accumulated in a 64-bit
49 accumulator in 33.31 format.
50 There is no risk of internal overflow with this approach, and the
51 full precision of intermediate result is preserved.
52 Finally, the accumulator is truncated to yield a result of 1.31 format.
53 */
54 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
arm_mean_q31(const q31_t * pSrc,uint32_t blockSize,q31_t * pResult)55 ARM_DSP_ATTRIBUTE void arm_mean_q31(
56 const q31_t * pSrc,
57 uint32_t blockSize,
58 q31_t * pResult)
59 {
60 uint32_t blkCnt; /* loop counters */
61 q31x4_t vecSrc;
62 q63_t sum = 0LL;
63
64
65 /* Compute 4 outputs at a time */
66 blkCnt = blockSize >> 2U;
67 while (blkCnt > 0U)
68 {
69
70 vecSrc = vldrwq_s32(pSrc);
71 /*
72 * sum lanes
73 */
74 sum = vaddlvaq(sum, vecSrc);
75
76 blkCnt --;
77 pSrc += 4;
78 }
79
80 /* Tail */
81 blkCnt = blockSize & 0x3;
82
83 while (blkCnt > 0U)
84 {
85 /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */
86 sum += *pSrc++;
87 blkCnt --;
88 }
89
90 *pResult = arm_div_int64_to_int32(sum, blockSize);
91 }
92 #else
arm_mean_q31(const q31_t * pSrc,uint32_t blockSize,q31_t * pResult)93 ARM_DSP_ATTRIBUTE void arm_mean_q31(
94 const q31_t * pSrc,
95 uint32_t blockSize,
96 q31_t * pResult)
97 {
98 uint32_t blkCnt; /* Loop counter */
99 q63_t sum = 0; /* Temporary result storage */
100
101 #if defined (ARM_MATH_LOOPUNROLL)
102
103 /* Loop unrolling: Compute 4 outputs at a time */
104 blkCnt = blockSize >> 2U;
105
106 while (blkCnt > 0U)
107 {
108 /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */
109 sum += *pSrc++;
110
111 sum += *pSrc++;
112
113 sum += *pSrc++;
114
115 sum += *pSrc++;
116
117 /* Decrement the loop counter */
118 blkCnt--;
119 }
120
121 /* Loop unrolling: Compute remaining outputs */
122 blkCnt = blockSize % 0x4U;
123
124 #else
125
126 /* Initialize blkCnt with number of samples */
127 blkCnt = blockSize;
128
129 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
130
131 while (blkCnt > 0U)
132 {
133 /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) */
134 sum += *pSrc++;
135
136 /* Decrement loop counter */
137 blkCnt--;
138 }
139
140 /* C = (A[0] + A[1] + A[2] + ... + A[blockSize-1]) / blockSize */
141 /* Store result to destination */
142 *pResult = (q31_t) (sum / blockSize);
143 }
144 #endif /* defined(ARM_MATH_MVEI) */
145
146 /**
147 @} end of mean group
148 */
149
