1  /* ----------------------------------------------------------------------
  2   * Project:      CMSIS DSP Library
  3   * Title:        arm_add_f32.c
  4   * Description:  Floating-point vector addition
  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    @defgroup BasicAdd Vector Addition
 37  
 38    Element-by-element addition of two vectors.
 39  
 40    <pre>
 41        pDst[n] = pSrcA[n] + pSrcB[n],   0 <= n < blockSize.
 42    </pre>
 43  
 44    There are separate functions for floating-point, Q7, Q15, and Q31 data types.
 45   */
 46  
 47  /**
 48    @addtogroup BasicAdd
 49    @{
 50   */
 51  
 52  /**
 53    @brief         Floating-point vector addition.
 54    @param[in]     pSrcA      points to first input vector
 55    @param[in]     pSrcB      points to second input vector
 56    @param[out]    pDst       points to output vector
 57    @param[in]     blockSize  number of samples in each vector
 58    @return        none
 59   */
 60  
 61  #if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
 62  
 63  #include "arm_helium_utils.h"
 64  
 65  void arm_add_f32(
 66    const float32_t * pSrcA,
 67    const float32_t * pSrcB,
 68          float32_t * pDst,
 69          uint32_t blockSize)
 70  {
 71      uint32_t blkCnt;                               /* Loop counter */
 72  
 73      f32x4_t vec1;
 74      f32x4_t vec2;
 75      f32x4_t res;
 76  
 77      /* Compute 4 outputs at a time */
 78      blkCnt = blockSize >> 2U;
 79  
 80      while (blkCnt > 0U)
 81      {
 82          /* C = A + B */
 83  
 84          /* Add and then store the results in the destination buffer. */
 85          vec1 = vld1q(pSrcA);
 86          vec2 = vld1q(pSrcB);
 87          res = vaddq(vec1, vec2);
 88          vst1q(pDst, res);
 89  
 90          /* Increment pointers */
 91          pSrcA += 4;
 92          pSrcB += 4; 
 93          pDst += 4;
 94          
 95          /* Decrement the loop counter */
 96          blkCnt--;
 97      }
 98  
 99      /* Tail */
100      blkCnt = blockSize & 0x3;
101  
102      if (blkCnt > 0U)
103      {
104        /* C = A + B */
105        mve_pred16_t p0 = vctp32q(blkCnt);
106        vec1 = vld1q(pSrcA);
107        vec2 = vld1q(pSrcB);
108        vstrwq_p(pDst, vaddq(vec1,vec2), p0);
109      }
110  
111  }
112  
113  #else
114  void arm_add_f32(
115    const float32_t * pSrcA,
116    const float32_t * pSrcB,
117          float32_t * pDst,
118          uint32_t blockSize)
119  {
120          uint32_t blkCnt;                               /* Loop counter */
121  
122  #if defined(ARM_MATH_NEON) && !defined(ARM_MATH_AUTOVECTORIZE)
123      f32x4_t vec1;
124      f32x4_t vec2;
125      f32x4_t res;
126  
127      /* Compute 4 outputs at a time */
128      blkCnt = blockSize >> 2U;
129  
130      while (blkCnt > 0U)
131      {
132          /* C = A + B */
133  
134      	/* Add and then store the results in the destination buffer. */
135          vec1 = vld1q_f32(pSrcA);
136          vec2 = vld1q_f32(pSrcB);
137          res = vaddq_f32(vec1, vec2);
138          vst1q_f32(pDst, res);
139  
140          /* Increment pointers */
141          pSrcA += 4;
142          pSrcB += 4; 
143          pDst += 4;
144          
145          /* Decrement the loop counter */
146          blkCnt--;
147      }
148  
149      /* Tail */
150      blkCnt = blockSize & 0x3;
151  
152  #else
153  #if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
154  
155    /* Loop unrolling: Compute 4 outputs at a time */
156    blkCnt = blockSize >> 2U;
157  
158    while (blkCnt > 0U)
159    {
160      /* C = A + B */
161  
162      /* Add and store result in destination buffer. */
163      *pDst++ = (*pSrcA++) + (*pSrcB++);
164      *pDst++ = (*pSrcA++) + (*pSrcB++);
165      *pDst++ = (*pSrcA++) + (*pSrcB++);
166      *pDst++ = (*pSrcA++) + (*pSrcB++);
167  
168      /* Decrement loop counter */
169      blkCnt--;
170    }
171  
172    /* Loop unrolling: Compute remaining outputs */
173    blkCnt = blockSize % 0x4U;
174  
175  #else
176  
177    /* Initialize blkCnt with number of samples */
178    blkCnt = blockSize;
179  
180  #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
181  #endif /* #if defined(ARM_MATH_NEON) */
182  
183    while (blkCnt > 0U)
184    {
185      /* C = A + B */
186  
187      /* Add and store result in destination buffer. */
188      *pDst++ = (*pSrcA++) + (*pSrcB++);
189  
190      /* Decrement loop counter */
191      blkCnt--;
192    }
193  
194  }
195  #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
196  
197  /**
198    @} end of BasicAdd group
199   */