lib/main/CMSIS/DSP/Source/BasicMathFunctions/arm_mult_q15.c

   1 /* ----------------------------------------------------------------------
   2  * Project:      CMSIS DSP Library
   3  * Title:        arm_mult_q15.c
   4  * Description:  Q15 vector multiplication
   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 BasicMult
  37  * @{
  38  */
  39
  40
  41 /**
  42  * @brief           Q15 vector multiplication
  43  * @param[in]       *pSrcA points to the first input vector
  44  * @param[in]       *pSrcB points to the second input vector
  45  * @param[out]      *pDst points to the output vector
  46  * @param[in]       blockSize number of samples in each vector
  47  * @return none.
  48  *
  49  * <b>Scaling and Overflow Behavior:</b>
  50  * \par
  51  * The function uses saturating arithmetic.
  52  * Results outside of the allowable Q15 range [0x8000 0x7FFF] will be saturated.
  53  */
  54
  55 void arm_mult_q15(
  56   q15_t * pSrcA,
  57   q15_t * pSrcB,
  58   q15_t * pDst,
  59   uint32_t blockSize)
  60 {
  61   uint32_t blkCnt;                               /* loop counters */
  62
  63 #if defined (ARM_MATH_DSP)
  64
  65 /* Run the below code for Cortex-M4 and Cortex-M3 */
  66   q31_t inA1, inA2, inB1, inB2;                  /* temporary input variables */
  67   q15_t out1, out2, out3, out4;                  /* temporary output variables */
  68   q31_t mul1, mul2, mul3, mul4;                  /* temporary variables */
  69
  70   /* loop Unrolling */
  71   blkCnt = blockSize >> 2U;
  72
  73   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
  74    ** a second loop below computes the remaining 1 to 3 samples. */
  75   while (blkCnt > 0U)
  76   {
  77     /* read two samples at a time from sourceA */
  78     inA1 = *__SIMD32(pSrcA)++;
  79     /* read two samples at a time from sourceB */
  80     inB1 = *__SIMD32(pSrcB)++;
  81     /* read two samples at a time from sourceA */
  82     inA2 = *__SIMD32(pSrcA)++;
  83     /* read two samples at a time from sourceB */
  84     inB2 = *__SIMD32(pSrcB)++;
  85
  86     /* multiply mul = sourceA * sourceB */
  87     mul1 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1 >> 16));
  88     mul2 = (q31_t) ((q15_t) inA1 * (q15_t) inB1);
  89     mul3 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) (inB2 >> 16));
  90     mul4 = (q31_t) ((q15_t) inA2 * (q15_t) inB2);
  91
  92     /* saturate result to 16 bit */
  93     out1 = (q15_t) __SSAT(mul1 >> 15, 16);
  94     out2 = (q15_t) __SSAT(mul2 >> 15, 16);
  95     out3 = (q15_t) __SSAT(mul3 >> 15, 16);
  96     out4 = (q15_t) __SSAT(mul4 >> 15, 16);
  97
  98     /* store the result */
  99 #ifndef ARM_MATH_BIG_ENDIAN
 100
 101     *__SIMD32(pDst)++ = __PKHBT(out2, out1, 16);
 102     *__SIMD32(pDst)++ = __PKHBT(out4, out3, 16);
 103
 104 #else
 105
 106     *__SIMD32(pDst)++ = __PKHBT(out2, out1, 16);
 107     *__SIMD32(pDst)++ = __PKHBT(out4, out3, 16);
 108
 109 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
 110
 111     /* Decrement the blockSize loop counter */
 112     blkCnt--;
 113   }
 114
 115   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
 116    ** No loop unrolling is used. */
 117   blkCnt = blockSize % 0x4U;
 118
 119 #else
 120
 121   /* Run the below code for Cortex-M0 */
 122
 123   /* Initialize blkCnt with number of samples */
 124   blkCnt = blockSize;
 125
 126 #endif /* #if defined (ARM_MATH_DSP) */
 127
 128
 129   while (blkCnt > 0U)
 130   {
 131     /* C = A * B */
 132     /* Multiply the inputs and store the result in the destination buffer */
 133     *pDst++ = (q15_t) __SSAT((((q31_t) (*pSrcA++) * (*pSrcB++)) >> 15), 16);
 134
 135     /* Decrement the blockSize loop counter */
 136     blkCnt--;
 137   }
 138 }
 139
 140 /**
 141  * @} end of BasicMult group
 142  */