1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_abs_f16.c
4 * Description: Floating-point vector absolute value
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/basic_math_functions_f16.h"
30 #include <math.h>
31
32 /**
33 @ingroup groupMath
34 */
35
36
37 /**
38 @addtogroup BasicAbs
39 @{
40 */
41
42 /**
43 @brief Floating-point vector absolute value.
44 @param[in] pSrc points to the input vector
45 @param[out] pDst points to the output vector
46 @param[in] blockSize number of samples in each vector
47 */
48
49
50 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
51
52 #include "arm_helium_utils.h"
53
arm_abs_f16(const float16_t * pSrc,float16_t * pDst,uint32_t blockSize)54 void arm_abs_f16(
55 const float16_t * pSrc,
56 float16_t * pDst,
57 uint32_t blockSize)
58 {
59 uint32_t blkCnt; /* Loop counter */
60 f16x8_t vec1;
61 f16x8_t res;
62
63
64 /* Compute 4 outputs at a time */
65 blkCnt = blockSize >> 3U;
66
67 while (blkCnt > 0U)
68 {
69 /* C = |A| */
70
71 /* Calculate absolute values and then store the results in the destination buffer. */
72 vec1 = vld1q(pSrc);
73 res = vabsq(vec1);
74 vst1q(pDst, res);
75
76 /* Increment pointers */
77 pSrc += 8;
78 pDst += 8;
79
80 /* Decrement the loop counter */
81 blkCnt--;
82 }
83
84 /* Tail */
85 blkCnt = blockSize & 0x7;
86
87
88 if (blkCnt > 0U)
89 {
90 /* C = |A| */
91 mve_pred16_t p0 = vctp16q(blkCnt);
92 vec1 = vld1q(pSrc);
93 vstrhq_p(pDst, vabsq(vec1), p0);
94 }
95
96 }
97
98 #else
99 #if defined(ARM_FLOAT16_SUPPORTED)
arm_abs_f16(const float16_t * pSrc,float16_t * pDst,uint32_t blockSize)100 void arm_abs_f16(
101 const float16_t * pSrc,
102 float16_t * pDst,
103 uint32_t blockSize)
104 {
105 uint32_t blkCnt; /* Loop counter */
106
107 #if defined(ARM_MATH_NEON_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
108 f16x8_t vec1;
109 f16x8_t res;
110
111 /* Compute 4 outputs at a time */
112 blkCnt = blockSize >> 2U;
113
114 while (blkCnt > 0U)
115 {
116 /* C = |A| */
117
118 /* Calculate absolute values and then store the results in the destination buffer. */
119 vec1 = vld1q_f16(pSrc);
120 res = vabsq_f16(vec1);
121 vst1q_f16(pDst, res);
122
123 /* Increment pointers */
124 pSrc += 4;
125 pDst += 4;
126
127 /* Decrement the loop counter */
128 blkCnt--;
129 }
130
131 /* Tail */
132 blkCnt = blockSize & 0x3;
133
134 #else
135 #if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
136
137 /* Loop unrolling: Compute 4 outputs at a time */
138 blkCnt = blockSize >> 2U;
139
140 while (blkCnt > 0U)
141 {
142 /* C = |A| */
143
144 /* Calculate absolute and store result in destination buffer. */
145 *pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
146
147 *pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
148
149 *pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
150
151 *pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
152
153 /* Decrement loop counter */
154 blkCnt--;
155 }
156
157 /* Loop unrolling: Compute remaining outputs */
158 blkCnt = blockSize % 0x4U;
159
160 #else
161
162 /* Initialize blkCnt with number of samples */
163 blkCnt = blockSize;
164
165 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
166 #endif /* #if defined(ARM_MATH_NEON) */
167
168 while (blkCnt > 0U)
169 {
170 /* C = |A| */
171
172 /* Calculate absolute and store result in destination buffer. */
173 *pDst++ = (_Float16)fabsf((float32_t)*pSrc++);
174
175 /* Decrement loop counter */
176 blkCnt--;
177 }
178
179 }
180 #endif /* defined(ARM_FLOAT16_SUPPORTED */
181 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
182 /**
183 @} end of BasicAbs group
184 */
185
