tmk_core/tool/mbed/mbed-sdk/libraries/dsp/cmsis_dsp/BasicMathFunctions/arm_dot_prod_q7.c

   1 /* ----------------------------------------------------------------------
   2 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.
   3 *
   4 * $Date:        17. January 2013
   5 * $Revision:    V1.4.1
   6 *
   7 * Project:          CMSIS DSP Library
   8 * Title:                arm_dot_prod_q7.c
   9 *
  10 * Description:  Q7 dot product.
  11 *
  12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
  13 *
  14 * Redistribution and use in source and binary forms, with or without
  15 * modification, are permitted provided that the following conditions
  16 * are met:
  17 *   - Redistributions of source code must retain the above copyright
  18 *     notice, this list of conditions and the following disclaimer.
  19 *   - Redistributions in binary form must reproduce the above copyright
  20 *     notice, this list of conditions and the following disclaimer in
  21 *     the documentation and/or other materials provided with the
  22 *     distribution.
  23 *   - Neither the name of ARM LIMITED nor the names of its contributors
  24 *     may be used to endorse or promote products derived from this
  25 *     software without specific prior written permission.
  26 *
  27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  38 * POSSIBILITY OF SUCH DAMAGE.
  39 * -------------------------------------------------------------------- */
  40
  41 #include "arm_math.h"
  42
  43 /**
  44  * @ingroup groupMath
  45  */
  46
  47 /**
  48  * @addtogroup dot_prod
  49  * @{
  50  */
  51
  52 /**
  53  * @brief Dot product of Q7 vectors.
  54  * @param[in]       *pSrcA points to the first input vector
  55  * @param[in]       *pSrcB points to the second input vector
  56  * @param[in]       blockSize number of samples in each vector
  57  * @param[out]      *result output result returned here
  58  * @return none.
  59  *
  60  * <b>Scaling and Overflow Behavior:</b>
  61  * \par
  62  * The intermediate multiplications are in 1.7 x 1.7 = 2.14 format and these
  63  * results are added to an accumulator in 18.14 format.
  64  * Nonsaturating additions are used and there is no danger of wrap around as long as
  65  * the vectors are less than 2^18 elements long.
  66  * The return result is in 18.14 format.
  67  */
  68
  69 void arm_dot_prod_q7(
  70   q7_t * pSrcA,
  71   q7_t * pSrcB,
  72   uint32_t blockSize,
  73   q31_t * result)
  74 {
  75   uint32_t blkCnt;                               /* loop counter */
  76
  77   q31_t sum = 0;                                 /* Temporary variables to store output */
  78
  79 #ifndef ARM_MATH_CM0_FAMILY
  80
  81 /* Run the below code for Cortex-M4 and Cortex-M3 */
  82
  83   q31_t input1, input2;                          /* Temporary variables to store input */
  84   q31_t inA1, inA2, inB1, inB2;                  /* Temporary variables to store input */
  85
  86
  87
  88   /*loop Unrolling */
  89   blkCnt = blockSize >> 2u;
  90
  91   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
  92    ** a second loop below computes the remaining 1 to 3 samples. */
  93   while(blkCnt > 0u)
  94   {
  95     /* read 4 samples at a time from sourceA */
  96     input1 = *__SIMD32(pSrcA)++;
  97     /* read 4 samples at a time from sourceB */
  98     input2 = *__SIMD32(pSrcB)++;
  99
 100     /* extract two q7_t samples to q15_t samples */
 101     inA1 = __SXTB16(__ROR(input1, 8));
 102     /* extract reminaing two samples */
 103     inA2 = __SXTB16(input1);
 104     /* extract two q7_t samples to q15_t samples */
 105     inB1 = __SXTB16(__ROR(input2, 8));
 106     /* extract reminaing two samples */
 107     inB2 = __SXTB16(input2);
 108
 109     /* multiply and accumulate two samples at a time */
 110     sum = __SMLAD(inA1, inB1, sum);
 111     sum = __SMLAD(inA2, inB2, sum);
 112
 113     /* Decrement the loop counter */
 114     blkCnt--;
 115   }
 116
 117   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
 118    ** No loop unrolling is used. */
 119   blkCnt = blockSize % 0x4u;
 120
 121   while(blkCnt > 0u)
 122   {
 123     /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
 124     /* Dot product and then store the results in a temporary buffer. */
 125     sum = __SMLAD(*pSrcA++, *pSrcB++, sum);
 126
 127     /* Decrement the loop counter */
 128     blkCnt--;
 129   }
 130
 131 #else
 132
 133   /* Run the below code for Cortex-M0 */
 134
 135
 136
 137   /* Initialize blkCnt with number of samples */
 138   blkCnt = blockSize;
 139
 140   while(blkCnt > 0u)
 141   {
 142     /* C = A[0]* B[0] + A[1]* B[1] + A[2]* B[2] + .....+ A[blockSize-1]* B[blockSize-1] */
 143     /* Dot product and then store the results in a temporary buffer. */
 144     sum += (q31_t) ((q15_t) * pSrcA++ * *pSrcB++);
 145
 146     /* Decrement the loop counter */
 147     blkCnt--;
 148   }
 149
 150 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
 151
 152
 153   /* Store the result in the destination buffer in 18.14 format */
 154   *result = sum;
 155 }
 156
 157 /**
 158  * @} end of dot_prod group
 159  */