1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_mean_f16.c
4 * Description: Mean value of a floating-point 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_f16.h"
30
31 #if defined(ARM_FLOAT16_SUPPORTED)
32
33
34 /**
35 @ingroup groupStats
36 */
37
38 /**
39 @defgroup mean Mean
40
41 Calculates the mean of the input vector. Mean is defined as the average of the elements in the vector.
42 The underlying algorithm is used:
43
44 <pre>
45 Result = (pSrc[0] + pSrc[1] + pSrc[2] + ... + pSrc[blockSize-1]) / blockSize;
46 </pre>
47
48 There are separate functions for floating-point, Q31, Q15, and Q7 data types.
49 */
50
51 /**
52 @addtogroup mean
53 @{
54 */
55
56 /**
57 @brief Mean value of a floating-point vector.
58 @param[in] pSrc points to the input vector.
59 @param[in] blockSize number of samples in input vector.
60 @param[out] pResult mean value returned here.
61 */
62 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
63
64 #include "arm_helium_utils.h"
65
arm_mean_f16(const float16_t * pSrc,uint32_t blockSize,float16_t * pResult)66 void arm_mean_f16(
67 const float16_t * pSrc,
68 uint32_t blockSize,
69 float16_t * pResult)
70 {
71 int32_t blkCnt; /* loop counters */
72 f16x8_t vecSrc;
73 f16x8_t sumVec = vdupq_n_f16(0.0f16);
74
75 blkCnt = blockSize;
76 do {
77 mve_pred16_t p = vctp16q(blkCnt);
78
79 vecSrc = vldrhq_z_f16((float16_t const *) pSrc, p);
80 sumVec = vaddq_m_f16(sumVec, sumVec, vecSrc, p);
81
82 blkCnt -= 8;
83 pSrc += 8;
84 }
85 while (blkCnt > 0);
86
87 *pResult = (_Float16)vecAddAcrossF16Mve(sumVec) / (_Float16) blockSize;
88 }
89
90
91 #else
92
arm_mean_f16(const float16_t * pSrc,uint32_t blockSize,float16_t * pResult)93 void arm_mean_f16(
94 const float16_t * pSrc,
95 uint32_t blockSize,
96 float16_t * pResult)
97 {
98 uint32_t blkCnt; /* Loop counter */
99 float16_t sum = 0.0f; /* Temporary result storage */
100
101 #if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
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 += (_Float16)*pSrc++;
110
111 sum += (_Float16)*pSrc++;
112
113 sum += (_Float16)*pSrc++;
114
115 sum += (_Float16)*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 += (_Float16)*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 = ((_Float16)sum / (_Float16)blockSize);
143 }
144 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
145
146 /**
147 @} end of mean group
148 */
149
150 #endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
151
152
