1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_max_no_idx_f16.c
4 * Description: Maximum value of a floating-point vector without returning the index
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 #if (defined(ARM_MATH_NEON) || defined(ARM_MATH_MVEF)) && !defined(ARM_MATH_AUTOVECTORIZE)
34 #include <limits.h>
35 #endif
36
37 /**
38 @ingroup groupStats
39 */
40
41
42 /**
43 @addtogroup Max
44 @{
45 */
46
47 /**
48 @brief Maximum value of a floating-point vector.
49 @param[in] pSrc points to the input vector
50 @param[in] blockSize number of samples in input vector
51 @param[out] pResult maximum value returned here
52 */
53
54 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
55
arm_max_no_idx_f16(const float16_t * pSrc,uint32_t blockSize,float16_t * pResult)56 ARM_DSP_ATTRIBUTE void arm_max_no_idx_f16(
57 const float16_t *pSrc,
58 uint32_t blockSize,
59 float16_t *pResult)
60 {
61 f16x8_t vecSrc;
62 f16x8_t curExtremValVec = vdupq_n_f16(F16_MIN);
63 float16_t maxValue = F16_MIN;
64 float16_t newVal;
65 uint32_t blkCnt;
66
67 /* Loop unrolling: Compute 4 outputs at a time */
68 blkCnt = blockSize >> 3U;
69
70 while (blkCnt > 0U)
71 {
72
73 vecSrc = vldrhq_f16(pSrc);
74 /*
75 * update per-lane max.
76 */
77 curExtremValVec = vmaxnmq(vecSrc, curExtremValVec);
78 /*
79 * Decrement the blockSize loop counter
80 * Advance vector source and destination pointers
81 */
82 pSrc += 8;
83 blkCnt --;
84 }
85 /*
86 * Get max value across the vector
87 */
88 maxValue = vmaxnmvq(maxValue, curExtremValVec);
89
90 blkCnt = blockSize & 7;
91
92 while (blkCnt > 0U)
93 {
94 newVal = *pSrc++;
95
96 /* compare for the maximum value */
97 if ((_Float16)maxValue < (_Float16)newVal)
98 {
99 /* Update the maximum value and it's index */
100 maxValue = newVal;
101 }
102
103 blkCnt --;
104 }
105
106 *pResult = maxValue;
107 }
108
109 #else
110
arm_max_no_idx_f16(const float16_t * pSrc,uint32_t blockSize,float16_t * pResult)111 ARM_DSP_ATTRIBUTE void arm_max_no_idx_f16(
112 const float16_t *pSrc,
113 uint32_t blockSize,
114 float16_t *pResult)
115 {
116 float16_t maxValue = F16_MIN;
117 float16_t newVal;
118
119 while (blockSize > 0U)
120 {
121 newVal = *pSrc++;
122
123 /* compare for the maximum value */
124 if ((_Float16)maxValue < (_Float16)newVal)
125 {
126 /* Update the maximum value and it's index */
127 maxValue = newVal;
128 }
129
130 blockSize --;
131 }
132
133 *pResult = maxValue;
134 }
135
136 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
137
138 /**
139 @} end of Max group
140 */
141
142 #endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
143
144
