1 /* ----------------------------------------------------------------------
2  * Project:      CMSIS DSP Library
3  * Title:        arm_offset_q15.c
4  * Description:  Q15 vector offset
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 BasicOffset
37   @{
38  */
39 
40 /**
41   @brief         Adds a constant offset to a Q15 vector.
42   @param[in]     pSrc       points to the input vector
43   @param[in]     offset     is the offset to be added
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 Q15 range [0x8000 0x7FFF] are saturated.
50  */
51 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
52 
53 #include "arm_helium_utils.h"
54 
arm_offset_q15(const q15_t * pSrc,q15_t offset,q15_t * pDst,uint32_t blockSize)55 ARM_DSP_ATTRIBUTE void arm_offset_q15(
56     const q15_t * pSrc,
57     q15_t   offset,
58     q15_t * pDst,
59     uint32_t blockSize)
60 {
61     uint32_t  blkCnt;           /* loop counters */
62     q15x8_t vecSrc;
63 
64     /* Compute 8 outputs at a time */
65     blkCnt = blockSize >> 3;
66     while (blkCnt > 0U)
67     {
68         /*
69          * C = A + offset
70          * Add offset and then store the result in the destination buffer.
71          */
72         vecSrc = vld1q(pSrc);
73         vst1q(pDst, vqaddq(vecSrc, offset));
74         /*
75          * Decrement the blockSize loop counter
76          */
77         blkCnt--;
78         /*
79          * advance vector source and destination pointers
80          */
81         pSrc += 8;
82         pDst += 8;
83     }
84     /*
85      * tail
86      */
87     blkCnt = blockSize & 7;
88     if (blkCnt > 0U)
89     {
90         mve_pred16_t p0 = vctp16q(blkCnt);
91         vecSrc = vld1q(pSrc);
92         vstrhq_p(pDst, vqaddq(vecSrc, offset), p0);
93     }
94 }
95 
96 
97 #else
arm_offset_q15(const q15_t * pSrc,q15_t offset,q15_t * pDst,uint32_t blockSize)98 ARM_DSP_ATTRIBUTE void arm_offset_q15(
99   const q15_t * pSrc,
100         q15_t offset,
101         q15_t * pDst,
102         uint32_t blockSize)
103 {
104         uint32_t blkCnt;                               /* Loop counter */
105 
106 #if defined (ARM_MATH_LOOPUNROLL)
107 
108 #if defined (ARM_MATH_DSP)
109   q31_t offset_packed;                           /* Offset packed to 32 bit */
110 
111   /* Offset is packed to 32 bit in order to use SIMD32 for addition */
112   offset_packed = __PKHBT(offset, offset, 16);
113 #endif
114 
115   /* Loop unrolling: Compute 4 outputs at a time */
116   blkCnt = blockSize >> 2U;
117 
118   while (blkCnt > 0U)
119   {
120     /* C = A + offset */
121 
122 #if defined (ARM_MATH_DSP)
123     /* Add offset and store result in destination buffer (2 samples at a time). */
124     write_q15x2_ia (&pDst, __QADD16(read_q15x2_ia (&pSrc), offset_packed));
125     write_q15x2_ia (&pDst, __QADD16(read_q15x2_ia (&pSrc), offset_packed));
126 #else
127     *pDst++ = (q15_t) __SSAT(((q31_t) *pSrc++ + offset), 16);
128     *pDst++ = (q15_t) __SSAT(((q31_t) *pSrc++ + offset), 16);
129     *pDst++ = (q15_t) __SSAT(((q31_t) *pSrc++ + offset), 16);
130     *pDst++ = (q15_t) __SSAT(((q31_t) *pSrc++ + offset), 16);
131 #endif
132 
133     /* Decrement loop counter */
134     blkCnt--;
135   }
136 
137   /* Loop unrolling: Compute remaining outputs */
138   blkCnt = blockSize % 0x4U;
139 
140 #else
141 
142   /* Initialize blkCnt with number of samples */
143   blkCnt = blockSize;
144 
145 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
146 
147   while (blkCnt > 0U)
148   {
149     /* C = A + offset */
150 
151     /* Add offset and store result in destination buffer. */
152 #if defined (ARM_MATH_DSP)
153     *pDst++ = (q15_t) __QADD16(*pSrc++, offset);
154 #else
155     *pDst++ = (q15_t) __SSAT(((q31_t) *pSrc++ + offset), 16);
156 #endif
157 
158     /* Decrement loop counter */
159     blkCnt--;
160   }
161 
162 }
163 #endif /* defined(ARM_MATH_MVEI) */
164 
165 /**
166   @} end of BasicOffset group
167  */
168