• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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   @return        none
46 
47   @par           Scaling and Overflow Behavior
48                    The function uses saturating arithmetic.
49                    The Q31 value -1 (0x80000000) will be saturated to the maximum allowable positive value 0x7FFFFFFF.
50  */
51 
52 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
53 
54 #include "arm_helium_utils.h"
55 
arm_abs_q31(const q31_t * pSrc,q31_t * pDst,uint32_t blockSize)56 void arm_abs_q31(
57     const q31_t * pSrc,
58     q31_t * pDst,
59     uint32_t blockSize)
60 {
61     uint32_t  blkCnt;           /* Loop counters */
62     q31x4_t vecSrc;
63 
64     /* Compute 4 outputs at a time */
65     blkCnt = blockSize >> 2;
66 
67     while (blkCnt > 0U)
68     {
69         /*
70          * C = |A|
71          * Calculate absolute and then store the results in the destination buffer.
72          */
73         vecSrc = vld1q(pSrc);
74         vst1q(pDst, vqabsq(vecSrc));
75         /*
76          * Decrement the blockSize loop counter
77          */
78         blkCnt--;
79         /*
80          * Advance vector source and destination pointers
81          */
82         pSrc += 4;
83         pDst += 4;
84     }
85     /*
86      * Tail
87      */
88     blkCnt = blockSize & 3;
89 
90     if (blkCnt > 0U)
91     {
92         mve_pred16_t p0 = vctp32q(blkCnt);
93         vecSrc = vld1q(pSrc);
94         vstrwq_p(pDst, vqabsq(vecSrc), p0);
95     }
96 }
97 
98 #else
arm_abs_q31(const q31_t * pSrc,q31_t * pDst,uint32_t blockSize)99 void arm_abs_q31(
100   const q31_t * pSrc,
101         q31_t * pDst,
102         uint32_t blockSize)
103 {
104         uint32_t blkCnt;                               /* Loop counter */
105         q31_t in;                                      /* Temporary variable */
106 
107 #if defined(ARM_MATH_NEON)
108     int32x4_t vec1;
109     int32x4_t res;
110 
111     /* Compute 4 outputs at a time */
112     blkCnt = blockSize >> 2U;
113 
114     while (blkCnt > 0U)
115     {
116         /* C = |A| */
117         /* Calculate absolute and then store the results in the destination buffer. */
118 
119         vec1 = vld1q_s32(pSrc);
120         res = vqabsq_s32(vec1);
121         vst1q_s32(pDst, res);
122 
123         /* Increment pointers */
124         pSrc += 4;
125         pDst += 4;
126 
127         /* Decrement the blockSize loop counter */
128         blkCnt--;
129     }
130 
131     /* Tail */
132     blkCnt = blockSize & 0x3;
133 
134 #else
135 #if defined (ARM_MATH_LOOPUNROLL)
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 of input (if -1 then saturated to 0x7fffffff) and store result in destination buffer. */
145     in = *pSrc++;
146 #if defined (ARM_MATH_DSP)
147     *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
148 #else
149     *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
150 #endif
151 
152     in = *pSrc++;
153 #if defined (ARM_MATH_DSP)
154     *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
155 #else
156     *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
157 #endif
158 
159     in = *pSrc++;
160 #if defined (ARM_MATH_DSP)
161     *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
162 #else
163     *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
164 #endif
165 
166     in = *pSrc++;
167 #if defined (ARM_MATH_DSP)
168     *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
169 #else
170     *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
171 #endif
172 
173     /* Decrement loop counter */
174     blkCnt--;
175   }
176 
177   /* Loop unrolling: Compute remaining outputs */
178   blkCnt = blockSize % 0x4U;
179 
180 #else
181 
182   /* Initialize blkCnt with number of samples */
183   blkCnt = blockSize;
184 
185 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
186 #endif /* #if defined (ARM_MATH_NEON) */
187 
188   while (blkCnt > 0U)
189   {
190     /* C = |A| */
191 
192     /* Calculate absolute of input (if -1 then saturated to 0x7fffffff) and store result in destination buffer. */
193     in = *pSrc++;
194 #if defined (ARM_MATH_DSP)
195     *pDst++ = (in > 0) ? in : (q31_t)__QSUB(0, in);
196 #else
197     *pDst++ = (in > 0) ? in : ((in == INT32_MIN) ? INT32_MAX : -in);
198 #endif
199 
200     /* Decrement loop counter */
201     blkCnt--;
202   }
203 
204 }
205 #endif /* #if defined (ARM_MATH_MVEI) */
206 /**
207   @} end of BasicAbs group
208  */
209