1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_power_f16.c
4 * Description: Sum of the squares of the elements 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
40 /**
41 @addtogroup power
42 @{
43 */
44
45 /**
46 @brief Sum of the squares of the elements of a floating-point vector.
47 @param[in] pSrc points to the input vector
48 @param[in] blockSize number of samples in input vector
49 @param[out] pResult sum of the squares value returned here
50 @return none
51 */
52 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
53
54 #include "arm_helium_utils.h"
55
arm_power_f16(const float16_t * pSrc,uint32_t blockSize,float16_t * pResult)56 void arm_power_f16(
57 const float16_t * pSrc,
58 uint32_t blockSize,
59 float16_t * pResult)
60 {
61 int32_t blkCnt; /* loop counters */
62 f16x8_t vecSrc;
63 f16x8_t sumVec = vdupq_n_f16(0.0f);
64
65
66 blkCnt = blockSize;
67 do {
68 mve_pred16_t p = vctp16q(blkCnt);
69
70 vecSrc = vldrhq_z_f16((float16_t const *) pSrc, p);
71 /*
72 * sum lanes
73 */
74 sumVec = vfmaq_m(sumVec, vecSrc, vecSrc, p);
75
76 blkCnt -= 8;
77 pSrc += 8;
78 }
79 while (blkCnt > 0);
80
81 *pResult = vecAddAcrossF16Mve(sumVec);
82 }
83 #else
84
arm_power_f16(const float16_t * pSrc,uint32_t blockSize,float16_t * pResult)85 void arm_power_f16(
86 const float16_t * pSrc,
87 uint32_t blockSize,
88 float16_t * pResult)
89 {
90 uint32_t blkCnt; /* Loop counter */
91 _Float16 sum = 0.0f16; /* Temporary result storage */
92 _Float16 in; /* Temporary variable to store input value */
93
94 #if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
95
96 /* Loop unrolling: Compute 4 outputs at a time */
97 blkCnt = blockSize >> 2U;
98
99 while (blkCnt > 0U)
100 {
101 /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */
102
103 /* Compute Power and store result in a temporary variable, sum. */
104 in = *pSrc++;
105 sum += in * in;
106
107 in = *pSrc++;
108 sum += in * in;
109
110 in = *pSrc++;
111 sum += in * in;
112
113 in = *pSrc++;
114 sum += in * in;
115
116 /* Decrement loop counter */
117 blkCnt--;
118 }
119
120 /* Loop unrolling: Compute remaining outputs */
121 blkCnt = blockSize % 0x4U;
122
123 #else
124
125 /* Initialize blkCnt with number of samples */
126 blkCnt = blockSize;
127
128 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
129
130 while (blkCnt > 0U)
131 {
132 /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */
133
134 /* Compute Power and store result in a temporary variable, sum. */
135 in = *pSrc++;
136 sum += in * in;
137
138 /* Decrement loop counter */
139 blkCnt--;
140 }
141
142 /* Store result to destination */
143 *pResult = sum;
144 }
145 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
146
147 /**
148 @} end of power group
149 */
150
151 #endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
152
153