1  /* ----------------------------------------------------------------------
  2   * Project:      CMSIS DSP Library
  3   * Title:        arm_add_f16.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_f16.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_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
 62  
 63  #include "arm_helium_utils.h"
 64  
 65  void arm_add_f16(
 66    const float16_t * pSrcA,
 67    const float16_t * pSrcB,
 68          float16_t * pDst,
 69          uint32_t blockSize)
 70  {
 71      uint32_t blkCnt;                               /* Loop counter */
 72  
 73      f16x8_t vec1;
 74      f16x8_t vec2;
 75      f16x8_t res;
 76  
 77      /* Compute 4 outputs at a time */
 78      blkCnt = blockSize >> 3U;
 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 += 8;
 92          pSrcB += 8; 
 93          pDst += 8;
 94          
 95          /* Decrement the loop counter */
 96          blkCnt--;
 97      }
 98  
 99      /* Tail */
100      blkCnt = blockSize & 0x7;
101  
102      if (blkCnt > 0U)
103      {
104        /* C = A + B */
105        mve_pred16_t p0 = vctp16q(blkCnt);
106        vec1 = vld1q(pSrcA);
107        vec2 = vld1q(pSrcB);
108        vstrhq_p(pDst, vaddq(vec1,vec2), p0);
109      }
110  
111  }
112  
113  #else
114  #if defined(ARM_FLOAT16_SUPPORTED)
115  void arm_add_f16(
116    const float16_t * pSrcA,
117    const float16_t * pSrcB,
118          float16_t * pDst,
119          uint32_t blockSize)
120  {
121          uint32_t blkCnt;                               /* Loop counter */
122  
123  #if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
124  
125    /* Loop unrolling: Compute 4 outputs at a time */
126    blkCnt = blockSize >> 2U;
127  
128    while (blkCnt > 0U)
129    {
130      /* C = A + B */
131  
132      /* Add and store result in destination buffer. */
133      *pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
134      *pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
135      *pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
136      *pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
137  
138      /* Decrement loop counter */
139      blkCnt--;
140    }
141  
142    /* Loop unrolling: Compute remaining outputs */
143    blkCnt = blockSize % 0x4U;
144  
145  #else
146  
147    /* Initialize blkCnt with number of samples */
148    blkCnt = blockSize;
149  
150  #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
151  
152    while (blkCnt > 0U)
153    {
154      /* C = A + B */
155  
156      /* Add and store result in destination buffer. */
157      *pDst++ = (_Float16)(*pSrcA++) + (_Float16)(*pSrcB++);
158  
159      /* Decrement loop counter */
160      blkCnt--;
161    }
162  
163  }
164  #endif /* defined(ARM_FLOAT16_SUPPORTED) */
165  #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
166  
167  /**
168    @} end of BasicAdd group
169   */