1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_abs_q7.c
4 * Description: Q7 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 Q7 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 Conditions for optimum performance
47 Input and output buffers should be aligned by 32-bit
48 @par Scaling and Overflow Behavior
49 The function uses saturating arithmetic.
50 The Q7 value -1 (0x80) will be saturated to the maximum allowable positive value 0x7F.
51 */
52
53 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
54
55 #include "arm_helium_utils.h"
56
arm_abs_q7(const q7_t * pSrc,q7_t * pDst,uint32_t blockSize)57 void arm_abs_q7(
58 const q7_t * pSrc,
59 q7_t * pDst,
60 uint32_t blockSize)
61 {
62 uint32_t blkCnt; /* loop counters */
63 q7x16_t vecSrc;
64
65 /* Compute 16 outputs at a time */
66 blkCnt = blockSize >> 4;
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 += 16;
83 pDst += 16;
84 }
85 /*
86 * tail
87 */
88 blkCnt = blockSize & 0xF;
89 if (blkCnt > 0U)
90 {
91 mve_pred16_t p0 = vctp8q(blkCnt);
92 vecSrc = vld1q(pSrc);
93 vstrbq_p(pDst, vqabsq(vecSrc), p0);
94 }
95 }
96
97 #else
arm_abs_q7(const q7_t * pSrc,q7_t * pDst,uint32_t blockSize)98 void arm_abs_q7(
99 const q7_t * pSrc,
100 q7_t * pDst,
101 uint32_t blockSize)
102 {
103 uint32_t blkCnt; /* Loop counter */
104 q7_t in; /* Temporary input variable */
105
106 #if defined (ARM_MATH_LOOPUNROLL)
107
108 /* Loop unrolling: Compute 4 outputs at a time */
109 blkCnt = blockSize >> 2U;
110
111 while (blkCnt > 0U)
112 {
113 /* C = |A| */
114
115 /* Calculate absolute of input (if -1 then saturated to 0x7f) and store result in destination buffer. */
116 in = *pSrc++;
117 #if defined (ARM_MATH_DSP)
118 *pDst++ = (in > 0) ? in : (q7_t)__QSUB8(0, in);
119 #else
120 *pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
121 #endif
122
123 in = *pSrc++;
124 #if defined (ARM_MATH_DSP)
125 *pDst++ = (in > 0) ? in : (q7_t)__QSUB8(0, in);
126 #else
127 *pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
128 #endif
129
130 in = *pSrc++;
131 #if defined (ARM_MATH_DSP)
132 *pDst++ = (in > 0) ? in : (q7_t)__QSUB8(0, in);
133 #else
134 *pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
135 #endif
136
137 in = *pSrc++;
138 #if defined (ARM_MATH_DSP)
139 *pDst++ = (in > 0) ? in : (q7_t)__QSUB8(0, in);
140 #else
141 *pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
142 #endif
143
144 /* Decrement loop counter */
145 blkCnt--;
146 }
147
148 /* Loop unrolling: Compute remaining outputs */
149 blkCnt = blockSize % 0x4U;
150
151 #else
152
153 /* Initialize blkCnt with number of samples */
154 blkCnt = blockSize;
155
156 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
157
158 while (blkCnt > 0U)
159 {
160 /* C = |A| */
161
162 /* Calculate absolute of input (if -1 then saturated to 0x7f) and store result in destination buffer. */
163 in = *pSrc++;
164 #if defined (ARM_MATH_DSP)
165 *pDst++ = (in > 0) ? in : (q7_t) __QSUB8(0, in);
166 #else
167 *pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? (q7_t) 0x7f : -in);
168 #endif
169
170 /* Decrement loop counter */
171 blkCnt--;
172 }
173
174 }
175 #endif /* defined(ARM_MATH_MVEI) */
176
177 /**
178 @} end of BasicAbs group
179 */
180
