1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_sub_q7.c
4 * Description: Q7 vector subtraction
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 BasicSub
37 @{
38 */
39
40 /**
41 @brief Q7 vector subtraction.
42 @param[in] pSrcA points to the first input vector
43 @param[in] pSrcB points to the second input vector
44 @param[out] pDst points to the output vector
45 @param[in] blockSize number of samples in each vector
46
47 @par Scaling and Overflow Behavior
48 The function uses saturating arithmetic.
49 Results outside of the allowable Q7 range [0x80 0x7F] will be saturated.
50 */
51 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
52
53 #include "arm_helium_utils.h"
54
arm_sub_q7(const q7_t * pSrcA,const q7_t * pSrcB,q7_t * pDst,uint32_t blockSize)55 ARM_DSP_ATTRIBUTE void arm_sub_q7(
56 const q7_t * pSrcA,
57 const q7_t * pSrcB,
58 q7_t * pDst,
59 uint32_t blockSize)
60 {
61 uint32_t blkCnt; /* loop counters */
62 q7x16_t vecA;
63 q7x16_t vecB;
64
65 /* Compute 16 outputs at a time */
66 blkCnt = blockSize >> 4;
67 while (blkCnt > 0U)
68 {
69 /*
70 * C = A - B
71 * Subtract and then store the results in the destination buffer.
72 */
73 vecA = vld1q(pSrcA);
74 vecB = vld1q(pSrcB);
75 vst1q(pDst, vqsubq(vecA, vecB));
76 /*
77 * Decrement the blockSize loop counter
78 */
79 blkCnt--;
80 /*
81 * advance vector source and destination pointers
82 */
83 pSrcA += 16;
84 pSrcB += 16;
85 pDst += 16;
86 }
87 /*
88 * tail
89 */
90 blkCnt = blockSize & 0xF;
91 if (blkCnt > 0U)
92 {
93 mve_pred16_t p0 = vctp8q(blkCnt);
94 vecA = vld1q(pSrcA);
95 vecB = vld1q(pSrcB);
96 vstrbq_p(pDst, vqsubq(vecA, vecB), p0);
97 }
98 }
99 #else
arm_sub_q7(const q7_t * pSrcA,const q7_t * pSrcB,q7_t * pDst,uint32_t blockSize)100 ARM_DSP_ATTRIBUTE void arm_sub_q7(
101 const q7_t * pSrcA,
102 const q7_t * pSrcB,
103 q7_t * pDst,
104 uint32_t blockSize)
105 {
106 uint32_t blkCnt; /* Loop counter */
107
108 #if defined (ARM_MATH_LOOPUNROLL)
109
110 /* Loop unrolling: Compute 4 outputs at a time */
111 blkCnt = blockSize >> 2U;
112
113 while (blkCnt > 0U)
114 {
115 /* C = A - B */
116
117 #if defined (ARM_MATH_DSP)
118 /* Subtract and store result in destination buffer (4 samples at a time). */
119 write_q7x4_ia (&pDst, __QSUB8(read_q7x4_ia (&pSrcA), read_q7x4_ia (&pSrcB)));
120 #else
121 *pDst++ = (q7_t) __SSAT((q15_t) *pSrcA++ - *pSrcB++, 8);
122 *pDst++ = (q7_t) __SSAT((q15_t) *pSrcA++ - *pSrcB++, 8);
123 *pDst++ = (q7_t) __SSAT((q15_t) *pSrcA++ - *pSrcB++, 8);
124 *pDst++ = (q7_t) __SSAT((q15_t) *pSrcA++ - *pSrcB++, 8);
125 #endif
126
127 /* Decrement loop counter */
128 blkCnt--;
129 }
130
131 /* Loop unrolling: Compute remaining outputs */
132 blkCnt = blockSize % 0x4U;
133
134 #else
135
136 /* Initialize blkCnt with number of samples */
137 blkCnt = blockSize;
138
139 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
140
141 while (blkCnt > 0U)
142 {
143 /* C = A - B */
144
145 /* Subtract and store result in destination buffer. */
146 *pDst++ = (q7_t) __SSAT((q15_t) *pSrcA++ - *pSrcB++, 8);
147
148 /* Decrement loop counter */
149 blkCnt--;
150 }
151
152 }
153 #endif /* defined(ARM_MATH_MVEI) */
154
155 /**
156 @} end of BasicSub group
157 */
158
