1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_weighted_average_f16.c
4 * Description: Weighted average
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 <limits.h>
30 #include <math.h>
31
32 #include "dsp/support_functions_f16.h"
33
34 #if defined(ARM_FLOAT16_SUPPORTED)
35
36 /**
37 @ingroup groupSupport
38 */
39
40 /**
41 @defgroup weightedaverage Weighted Average
42
43 Weighted average of values
44 */
45
46
47 /**
48 * @addtogroup weightedaverage
49 * @{
50 */
51
52
53 /**
54 * @brief Weighted average
55 *
56 *
57 * @param[in] *in Array of input values.
58 * @param[in] *weigths Weights
59 * @param[in] blockSize Number of samples in the input array.
60 *
61 * @return Weighted average
62 *
63 */
64
65 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
66
67 #include "arm_helium_utils.h"
68
arm_weighted_average_f16(const float16_t * in,const float16_t * weigths,uint32_t blockSize)69 ARM_DSP_ATTRIBUTE float16_t arm_weighted_average_f16(const float16_t *in,const float16_t *weigths, uint32_t blockSize)
70 {
71 _Float16 accum1, accum2;
72 float16x8_t accum1V, accum2V;
73 float16x8_t inV, wV;
74 const float16_t *pIn, *pW;
75 uint32_t blkCnt;
76
77
78 pIn = in;
79 pW = weigths;
80
81
82 accum1V = vdupq_n_f16(0.0f16);
83 accum2V = vdupq_n_f16(0.0f16);
84
85 blkCnt = blockSize >> 3;
86 while (blkCnt > 0)
87 {
88 inV = vld1q(pIn);
89 wV = vld1q(pW);
90
91 pIn += 4;
92 pW += 4;
93
94 accum1V = vfmaq(accum1V, inV, wV);
95 accum2V = vaddq(accum2V, wV);
96 blkCnt--;
97 }
98
99 accum1 = vecAddAcrossF16Mve(accum1V);
100 accum2 = vecAddAcrossF16Mve(accum2V);
101
102 blkCnt = blockSize & 7;
103 while(blkCnt > 0)
104 {
105 accum1 += (_Float16)*pIn++ * (_Float16)*pW;
106 accum2 += (_Float16)*pW++;
107 blkCnt--;
108 }
109
110
111 return (accum1 / accum2);
112 }
113
114 #else
115
arm_weighted_average_f16(const float16_t * in,const float16_t * weigths,uint32_t blockSize)116 ARM_DSP_ATTRIBUTE float16_t arm_weighted_average_f16(const float16_t *in, const float16_t *weigths, uint32_t blockSize)
117 {
118
119 _Float16 accum1, accum2;
120 const float16_t *pIn, *pW;
121 uint32_t blkCnt;
122
123
124 pIn = in;
125 pW = weigths;
126
127 accum1=0.0f16;
128 accum2=0.0f16;
129
130 blkCnt = blockSize;
131 while(blkCnt > 0)
132 {
133 accum1 += (_Float16)*pIn++ * (_Float16)*pW;
134 accum2 += (_Float16)*pW++;
135 blkCnt--;
136 }
137
138 return(accum1 / accum2);
139 }
140 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
141
142 /**
143 * @} end of weightedaverage group
144 */
145
146 #endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
147
148
