lib/main/CMSIS/DSP/Source/MatrixFunctions/arm_mat_scale_q15.c

   1 /* ----------------------------------------------------------------------
   2  * Project:      CMSIS DSP Library
   3  * Title:        arm_mat_scale_q15.c
   4  * Description:  Multiplies a Q15 matrix by a scalar
   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 groupMatrix
  33  */
  34
  35 /**
  36  * @addtogroup MatrixScale
  37  * @{
  38  */
  39
  40 /**
  41  * @brief Q15 matrix scaling.
  42  * @param[in]       *pSrc points to input matrix
  43  * @param[in]       scaleFract fractional portion of the scale factor
  44  * @param[in]       shift number of bits to shift the result by
  45  * @param[out]      *pDst points to output matrix structure
  46  * @return              The function returns either
  47  * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
  48  *
  49  * @details
  50  * <b>Scaling and Overflow Behavior:</b>
  51  * \par
  52  * The input data <code>*pSrc</code> and <code>scaleFract</code> are in 1.15 format.
  53  * These are multiplied to yield a 2.30 intermediate result and this is shifted with saturation to 1.15 format.
  54  */
  55
  56 arm_status arm_mat_scale_q15(
  57   const arm_matrix_instance_q15 * pSrc,
  58   q15_t scaleFract,
  59   int32_t shift,
  60   arm_matrix_instance_q15 * pDst)
  61 {
  62   q15_t *pIn = pSrc->pData;                      /* input data matrix pointer */
  63   q15_t *pOut = pDst->pData;                     /* output data matrix pointer */
  64   uint32_t numSamples;                           /* total number of elements in the matrix */
  65   int32_t totShift = 15 - shift;                 /* total shift to apply after scaling */
  66   uint32_t blkCnt;                               /* loop counters */
  67   arm_status status;                             /* status of matrix scaling     */
  68
  69 #if defined (ARM_MATH_DSP)
  70
  71   q15_t in1, in2, in3, in4;
  72   q31_t out1, out2, out3, out4;
  73   q31_t inA1, inA2;
  74
  75 #endif //     #if defined (ARM_MATH_DSP)
  76
  77 #ifdef ARM_MATH_MATRIX_CHECK
  78   /* Check for matrix mismatch */
  79   if ((pSrc->numRows != pDst->numRows) || (pSrc->numCols != pDst->numCols))
  80   {
  81     /* Set status as ARM_MATH_SIZE_MISMATCH */
  82     status = ARM_MATH_SIZE_MISMATCH;
  83   }
  84   else
  85 #endif //    #ifdef ARM_MATH_MATRIX_CHECK
  86   {
  87     /* Total number of samples in the input matrix */
  88     numSamples = (uint32_t) pSrc->numRows * pSrc->numCols;
  89
  90 #if defined (ARM_MATH_DSP)
  91
  92     /* Run the below code for Cortex-M4 and Cortex-M3 */
  93     /* Loop Unrolling */
  94     blkCnt = numSamples >> 2;
  95
  96     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
  97      ** a second loop below computes the remaining 1 to 3 samples. */
  98     while (blkCnt > 0U)
  99     {
 100       /* C(m,n) = A(m,n) * k */
 101       /* Scale, saturate and then store the results in the destination buffer. */
 102       /* Reading 2 inputs from memory */
 103       inA1 = _SIMD32_OFFSET(pIn);
 104       inA2 = _SIMD32_OFFSET(pIn + 2);
 105
 106       /* C = A * scale */
 107       /* Scale the inputs and then store the 2 results in the destination buffer
 108        * in single cycle by packing the outputs */
 109       out1 = (q31_t) ((q15_t) (inA1 >> 16) * scaleFract);
 110       out2 = (q31_t) ((q15_t) inA1 * scaleFract);
 111       out3 = (q31_t) ((q15_t) (inA2 >> 16) * scaleFract);
 112       out4 = (q31_t) ((q15_t) inA2 * scaleFract);
 113
 114       out1 = out1 >> totShift;
 115       inA1 = _SIMD32_OFFSET(pIn + 4);
 116       out2 = out2 >> totShift;
 117       inA2 = _SIMD32_OFFSET(pIn + 6);
 118       out3 = out3 >> totShift;
 119       out4 = out4 >> totShift;
 120
 121       in1 = (q15_t) (__SSAT(out1, 16));
 122       in2 = (q15_t) (__SSAT(out2, 16));
 123       in3 = (q15_t) (__SSAT(out3, 16));
 124       in4 = (q15_t) (__SSAT(out4, 16));
 125
 126       _SIMD32_OFFSET(pOut) = __PKHBT(in2, in1, 16);
 127       _SIMD32_OFFSET(pOut + 2) = __PKHBT(in4, in3, 16);
 128
 129       /* update pointers to process next sampels */
 130       pIn += 4U;
 131       pOut += 4U;
 132
 133
 134       /* Decrement the numSamples loop counter */
 135       blkCnt--;
 136     }
 137
 138     /* If the numSamples is not a multiple of 4, compute any remaining output samples here.
 139      ** No loop unrolling is used. */
 140     blkCnt = numSamples % 0x4U;
 141
 142 #else
 143
 144     /* Run the below code for Cortex-M0 */
 145
 146     /* Initialize blkCnt with number of samples */
 147     blkCnt = numSamples;
 148
 149 #endif /* #if defined (ARM_MATH_DSP) */
 150
 151     while (blkCnt > 0U)
 152     {
 153       /* C(m,n) = A(m,n) * k */
 154       /* Scale, saturate and then store the results in the destination buffer. */
 155       *pOut++ =
 156         (q15_t) (__SSAT(((q31_t) (*pIn++) * scaleFract) >> totShift, 16));
 157
 158       /* Decrement the numSamples loop counter */
 159       blkCnt--;
 160     }
 161     /* Set status as ARM_MATH_SUCCESS */
 162     status = ARM_MATH_SUCCESS;
 163   }
 164
 165   /* Return to application */
 166   return (status);
 167 }
 168
 169 /**
 170  * @} end of MatrixScale group
 171  */