lib/main/CMSIS/DSP/Source/BasicMathFunctions/arm_dot_prod_q7.c

   1 /* ----------------------------------------------------------------------
   2  * Project:      CMSIS DSP Library
   3  * Title:        arm_dot_prod_q7.c
   4  * Description:  Q7 dot product
   5  *
   6  * $Date:        27. January 2017
   7  * $Revision:    V.1.5.1
   8  *
   9  * Target Processor: Cortex-M cores
  10  * -------------------------------------------------------------------- */
  11 /*
  12  * Copyright (C) 2010-2017 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 "arm_math.h"
  30
  31 /**
  32  * @ingroup groupMath
  33  */
  34
  35 /**
  36  * @addtogroup dot_prod
  37  * @{
  38  */
  39
  40 /**
  41  * @brief Dot product of Q7 vectors.
  42  * @param[in]       *pSrcA points to the first input vector
  43  * @param[in]       *pSrcB points to the second input vector
  44  * @param[in]       blockSize number of samples in each vector
  45  * @param[out]      *result output result returned here
  46  * @return none.
  47  *
  48  * <b>Scaling and Overflow Behavior:</b>
  49  * \par
  50  * The intermediate multiplications are in 1.7 x 1.7 = 2.14 format and these
  51  * results are added to an accumulator in 18.14 format.
  52  * Nonsaturating additions are used and there is no danger of wrap around as long as
  53  * the vectors are less than 2^18 elements long.
  54  * The return result is in 18.14 format.
  55  */
  56
  57 void arm_dot_prod_q7(
  58   q7_t * pSrcA,
  59   q7_t * pSrcB,
  60   uint32_t blockSize,
  61   q31_t * result)
  62 {
  63   uint32_t blkCnt;                               /* loop counter */
  64
  65   q31_t sum = 0;                                 /* Temporary variables to store output */
  66
  67 #if defined (ARM_MATH_DSP)
  68
  69 /* Run the below code for Cortex-M4 and Cortex-M3 */
  70
  71   q31_t input1, input2;                          /* Temporary variables to store input */
  72   q31_t inA1, inA2, inB1, inB2;                  /* Temporary variables to store input */
  73
  74
  75
  76   /*loop Unrolling */
  77   blkCnt = blockSize >> 2U;
  78
  79   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
  80    ** a second loop below computes the remaining 1 to 3 samples. */
  81   while (blkCnt > 0U)
  82   {
  83     /* read 4 samples at a time from sourceA */
  84     input1 = *__SIMD32(pSrcA)++;
  85     /* read 4 samples at a time from sourceB */
  86     input2 = *__SIMD32(pSrcB)++;
  87
  88     /* extract two q7_t samples to q15_t samples */
  89     inA1 = __SXTB16(__ROR(input1, 8));
  90     /* extract reminaing two samples */
  91     inA2 = __SXTB16(input1);
  92     /* extract two q7_t samples to q15_t samples */
  93     inB1 = __SXTB16(__ROR(input2, 8));
  94     /* extract reminaing two samples */
  95     inB2 = __SXTB16(input2);
  96
  97     /* multiply and accumulate two samples at a time */
  98     sum = __SMLAD(inA1, inB1, sum);
  99     sum = __SMLAD(inA2, inB2, sum);
 100
 101     /* Decrement the loop counter */
 102     blkCnt--;
 103   }
 104
 105   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
 106    ** No loop unrolling is used. */
 107   blkCnt = blockSize % 0x4U;
 108
 109   while (blkCnt > 0U)
 110   {
 111     /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
 112     /* Dot product and then store the results in a temporary buffer. */
 113     sum = __SMLAD(*pSrcA++, *pSrcB++, sum);
 114
 115     /* Decrement the loop counter */
 116     blkCnt--;
 117   }
 118
 119 #else
 120
 121   /* Run the below code for Cortex-M0 */
 122
 123
 124
 125   /* Initialize blkCnt with number of samples */
 126   blkCnt = blockSize;
 127
 128   while (blkCnt > 0U)
 129   {
 130     /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
 131     /* Dot product and then store the results in a temporary buffer. */
 132     sum += (q31_t) ((q15_t) * pSrcA++ * *pSrcB++);
 133
 134     /* Decrement the loop counter */
 135     blkCnt--;
 136   }
 137
 138 #endif /* #if defined (ARM_MATH_DSP) */
 139
 140
 141   /* Store the result in the destination buffer in 18.14 format */
 142   *result = sum;
 143 }
 144
 145 /**
 146  * @} end of dot_prod group
 147  */