1  /* ----------------------------------------------------------------------
  2   * Project:      CMSIS DSP Library
  3   * Title:        arm_mult_f16.c
  4   * Description:  Floating-point vector multiplication
  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 BasicMult Vector Multiplication
 37  
 38    Element-by-element multiplication 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 BasicMult
 49    @{
 50   */
 51  
 52  /**
 53    @brief         Floating-point vector multiplication.
 54    @param[in]     pSrcA      points to the first input vector.
 55    @param[in]     pSrcB      points to the second input vector.
 56    @param[out]    pDst       points to the output vector.
 57    @param[in]     blockSize  number of samples in each vector.
 58    @return        none
 59   */
 60  
 61  
 62  #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE) 
 63  
 64  #include "arm_helium_utils.h"
 65  
 66  void arm_mult_f16(
 67    const float16_t * pSrcA,
 68    const float16_t * pSrcB,
 69          float16_t * pDst,
 70          uint32_t blockSize)
 71  {
 72      uint32_t blkCnt;                               /* Loop counter */
 73  
 74      f16x8_t vec1;
 75      f16x8_t vec2;
 76      f16x8_t res;
 77  
 78      /* Compute 4 outputs at a time */
 79      blkCnt = blockSize >> 3U;
 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 = vmulq(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      if (blkCnt > 0U)
102      {
103        /* C = A + B */
104        mve_pred16_t p0 = vctp16q(blkCnt);
105        vec1 = vld1q(pSrcA);
106        vec2 = vld1q(pSrcB);
107        vstrhq_p(pDst, vmulq(vec1,vec2), p0);
108      }
109  
110  }
111  
112  #else
113  #if defined(ARM_FLOAT16_SUPPORTED)
114  void arm_mult_f16(
115    const float16_t * pSrcA,
116    const float16_t * pSrcB,
117          float16_t * pDst,
118          uint32_t blockSize)
119  {
120      uint32_t blkCnt;                               /* Loop counter */
121  
122  #if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
123  
124    /* Loop unrolling: Compute 4 outputs at a time */
125    blkCnt = blockSize >> 2U;
126  
127    while (blkCnt > 0U)
128    {
129      /* C = A * B */
130  
131      /* Multiply inputs and store result in destination buffer. */
132      *pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
133  
134      *pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
135  
136      *pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
137  
138      *pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
139  
140      /* Decrement loop counter */
141      blkCnt--;
142    }
143  
144    /* Loop unrolling: Compute remaining outputs */
145    blkCnt = blockSize % 0x4U;
146  
147  #else
148  
149    /* Initialize blkCnt with number of samples */
150    blkCnt = blockSize;
151  
152  #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
153  
154    while (blkCnt > 0U)
155    {
156      /* C = A * B */
157  
158      /* Multiply input and store result in destination buffer. */
159      *pDst++ = (_Float16)(*pSrcA++) * (_Float16)(*pSrcB++);
160  
161      /* Decrement loop counter */
162      blkCnt--;
163    }
164  
165  }
166  #endif
167  #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
168  
169  /**
170    @} end of BasicMult group
171   */