1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_abs_q31.c
4 * Description: Q31 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.h"
30
31 /**
32 @ingroup groupMath
33 */
34
35 /**
36 @addtogroup BasicAbs
37 @{
38 */
39
40 /**
41 @brief Q31 vector absolute value.
42 @param[in] pSrc points to the input vector
43 @param[out] pDst points to the output vector
44 @param[in] blockSize number of samples in each vector
45
46 @par Scaling and Overflow Behavior
47 The function uses saturating arithmetic.
48 The Q31 value -1 (0x80000000) will be saturated to the maximum allowable positive value 0x7FFFFFFF.
49 */
50
51 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
52
53 #include "arm_helium_utils.h"
54
arm_abs_q31(const q31_t * pSrc,q31_t * pDst,uint32_t blockSize)55 void arm_abs_q31(
56 const q31_t * pSrc,
57 q31_t * pDst,
58 uint32_t blockSize)
59 {
60 uint32_t blkCnt; /* Loop counters */
61 q31x4_t vecSrc;
62
63 /* Compute 4 outputs at a time */
64 blkCnt = blockSize >> 2;
65
66 while (blkCnt > 0U)
67 {
68 /*
69 * C = |A|
70 * Calculate absolute and then store the results in the destination buffer.
71 */
72 vecSrc = vld1q(pSrc);
73 vst1q(pDst, vqabsq(vecSrc));
74 /*
75 * Decrement the blockSize loop counter
76 */
77 blkCnt--;
78 /*
79 * Advance vector source and destination pointers
80 */
81 pSrc += 4;
82 pDst += 4;
83 }
84 /*
85 * Tail
86 */
87 blkCnt = blockSize & 3;
88
89 if (blkCnt > 0U)
90 {
91 mve_pred16_t p0 = vctp32q(blkCnt);
92 vecSrc = vld1q(pSrc);
93 vstrwq_p(pDst, vqabsq(vecSrc), p0);
94 }
95 }
96
97 #else
arm_abs_q31(const q31_t * pSrc,q31_t * pDst,uint32_t blockSize)98 void arm_abs_q31(
99 const q31_t * pSrc,
100 q31_t * pDst,
101 uint32_t blockSize)
102 {
103 uint32_t blkCnt; /* Loop counter */
104 q31_t in; /* Temporary variable */
105
106 #if defined(ARM_MATH_NEON)
107 int32x4_t vec1;
108 int32x4_t res;
109
110 /* Compute 4 outputs at a time */
111 blkCnt = blockSize >> 2U;
112
113 while (blkCnt > 0U)
114 {
115 /* C = |A| */
116 /* Calculate absolute and then store the results in the destination buffer. */
117
118 vec1 = vld1q_s32(pSrc);
119 res = vqabsq_s32(vec1);
120 vst1q_s32(pDst, res);
121
122 /* Increment pointers */
123 pSrc += 4;
124 pDst += 4;
125
126 /* Decrement the blockSize loop counter */
127 blkCnt--;
128 }
129
130 /* Tail */
131 blkCnt = blockSize & 0x3;
132
133 #else
134 #if defined (ARM_MATH_LOOPUNROLL)
135
136 /* Loop unrolling: Compute 4 outputs at a time */
137 blkCnt = blockSize >> 2U;
138
139 while (blkCnt > 0U)
140 {
141 /* C = |A| */
142
143 /* Calculate absolute of input (if -1 then saturated to 0x7fffffff) and store result in destination buffer. */
144 in = *pSrc++;
145 #if defined (ARM_MATH_DSP)
146 *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
147 #else
148 *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
149 #endif
150
151 in = *pSrc++;
152 #if defined (ARM_MATH_DSP)
153 *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
154 #else
155 *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
156 #endif
157
158 in = *pSrc++;
159 #if defined (ARM_MATH_DSP)
160 *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
161 #else
162 *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
163 #endif
164
165 in = *pSrc++;
166 #if defined (ARM_MATH_DSP)
167 *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
168 #else
169 *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
170 #endif
171
172 /* Decrement loop counter */
173 blkCnt--;
174 }
175
176 /* Loop unrolling: Compute remaining outputs */
177 blkCnt = blockSize % 0x4U;
178
179 #else
180
181 /* Initialize blkCnt with number of samples */
182 blkCnt = blockSize;
183
184 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
185 #endif /* #if defined (ARM_MATH_NEON) */
186
187 while (blkCnt > 0U)
188 {
189 /* C = |A| */
190
191 /* Calculate absolute of input (if -1 then saturated to 0x7fffffff) and store result in destination buffer. */
192 in = *pSrc++;
193 #if defined (ARM_MATH_DSP)
194 *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
195 #else
196 *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
197 #endif
198
199 /* Decrement loop counter */
200 blkCnt--;
201 }
202
203 }
204 #endif /* #if defined (ARM_MATH_MVEI) */
205 /**
206 @} end of BasicAbs group
207 */
208
