Robert Lopez / CMSIS5

Important changes to repositories hosted on mbed.com

Mbed hosted mercurial repositories are deprecated and are due to be permanently deleted in July 2026.

To keep a copy of this software download the repository Zip archive or clone locally using Mercurial.

It is also possible to export all your personal repositories from the account settings page.

DSP/source/MatrixFunctions/arm_mat_scale_f32.c@0:eedb7d567a5d, 2018-04-12 (annotated)

Committer:
robert_lp
Date:
Thu Apr 12 01:31:58 2018 +0000
Revision:
0:eedb7d567a5d
CMSIS5 Library

Who changed what in which revision?

UserRevisionLine numberNew contents of line
robert_lp 0:eedb7d567a5d 1 /* ----------------------------------------------------------------------
robert_lp 0:eedb7d567a5d 2 * Project: CMSIS DSP Library
robert_lp 0:eedb7d567a5d 3 * Title: arm_mat_scale_f32.c
robert_lp 0:eedb7d567a5d 4 * Description: Multiplies a floating-point matrix by a scalar
robert_lp 0:eedb7d567a5d 5 *
robert_lp 0:eedb7d567a5d 6 * $Date: 27. January 2017
robert_lp 0:eedb7d567a5d 7 * $Revision: V.1.5.1
robert_lp 0:eedb7d567a5d 8 *
robert_lp 0:eedb7d567a5d 9 * Target Processor: Cortex-M cores
robert_lp 0:eedb7d567a5d 10 * -------------------------------------------------------------------- */
robert_lp 0:eedb7d567a5d 11 /*
robert_lp 0:eedb7d567a5d 12 * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved.
robert_lp 0:eedb7d567a5d 13 *
robert_lp 0:eedb7d567a5d 14 * SPDX-License-Identifier: Apache-2.0
robert_lp 0:eedb7d567a5d 15 *
robert_lp 0:eedb7d567a5d 16 * Licensed under the Apache License, Version 2.0 (the License); you may
robert_lp 0:eedb7d567a5d 17 * not use this file except in compliance with the License.
robert_lp 0:eedb7d567a5d 18 * You may obtain a copy of the License at
robert_lp 0:eedb7d567a5d 19 *
robert_lp 0:eedb7d567a5d 20 * www.apache.org/licenses/LICENSE-2.0
robert_lp 0:eedb7d567a5d 21 *
robert_lp 0:eedb7d567a5d 22 * Unless required by applicable law or agreed to in writing, software
robert_lp 0:eedb7d567a5d 23 * distributed under the License is distributed on an AS IS BASIS, WITHOUT
robert_lp 0:eedb7d567a5d 24 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
robert_lp 0:eedb7d567a5d 25 * See the License for the specific language governing permissions and
robert_lp 0:eedb7d567a5d 26 * limitations under the License.
robert_lp 0:eedb7d567a5d 27 */
robert_lp 0:eedb7d567a5d 28
robert_lp 0:eedb7d567a5d 29 #include "arm_math.h"
robert_lp 0:eedb7d567a5d 30
robert_lp 0:eedb7d567a5d 31 /**
robert_lp 0:eedb7d567a5d 32 * @ingroup groupMatrix
robert_lp 0:eedb7d567a5d 33 */
robert_lp 0:eedb7d567a5d 34
robert_lp 0:eedb7d567a5d 35 /**
robert_lp 0:eedb7d567a5d 36 * @defgroup MatrixScale Matrix Scale
robert_lp 0:eedb7d567a5d 37 *
robert_lp 0:eedb7d567a5d 38 * Multiplies a matrix by a scalar. This is accomplished by multiplying each element in the
robert_lp 0:eedb7d567a5d 39 * matrix by the scalar. For example:
robert_lp 0:eedb7d567a5d 40 * \image html MatrixScale.gif "Matrix Scaling of a 3 x 3 matrix"
robert_lp 0:eedb7d567a5d 41 *
robert_lp 0:eedb7d567a5d 42 * The function checks to make sure that the input and output matrices are of the same size.
robert_lp 0:eedb7d567a5d 43 *
robert_lp 0:eedb7d567a5d 44 * In the fixed-point Q15 and Q31 functions, <code>scale</code> is represented by
robert_lp 0:eedb7d567a5d 45 * a fractional multiplication <code>scaleFract</code> and an arithmetic shift <code>shift</code>.
robert_lp 0:eedb7d567a5d 46 * The shift allows the gain of the scaling operation to exceed 1.0.
robert_lp 0:eedb7d567a5d 47 * The overall scale factor applied to the fixed-point data is
robert_lp 0:eedb7d567a5d 48 * <pre>
robert_lp 0:eedb7d567a5d 49 * scale = scaleFract * 2^shift.
robert_lp 0:eedb7d567a5d 50 * </pre>
robert_lp 0:eedb7d567a5d 51 */
robert_lp 0:eedb7d567a5d 52
robert_lp 0:eedb7d567a5d 53 /**
robert_lp 0:eedb7d567a5d 54 * @addtogroup MatrixScale
robert_lp 0:eedb7d567a5d 55 * @{
robert_lp 0:eedb7d567a5d 56 */
robert_lp 0:eedb7d567a5d 57
robert_lp 0:eedb7d567a5d 58 /**
robert_lp 0:eedb7d567a5d 59 * @brief Floating-point matrix scaling.
robert_lp 0:eedb7d567a5d 60 * @param[in] *pSrc points to input matrix structure
robert_lp 0:eedb7d567a5d 61 * @param[in] scale scale factor to be applied
robert_lp 0:eedb7d567a5d 62 * @param[out] *pDst points to output matrix structure
robert_lp 0:eedb7d567a5d 63 * @return The function returns either <code>ARM_MATH_SIZE_MISMATCH</code>
robert_lp 0:eedb7d567a5d 64 * or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
robert_lp 0:eedb7d567a5d 65 *
robert_lp 0:eedb7d567a5d 66 */
robert_lp 0:eedb7d567a5d 67
robert_lp 0:eedb7d567a5d 68 arm_status arm_mat_scale_f32(
robert_lp 0:eedb7d567a5d 69 const arm_matrix_instance_f32 * pSrc,
robert_lp 0:eedb7d567a5d 70 float32_t scale,
robert_lp 0:eedb7d567a5d 71 arm_matrix_instance_f32 * pDst)
robert_lp 0:eedb7d567a5d 72 {
robert_lp 0:eedb7d567a5d 73 float32_t *pIn = pSrc->pData; /* input data matrix pointer */
robert_lp 0:eedb7d567a5d 74 float32_t *pOut = pDst->pData; /* output data matrix pointer */
robert_lp 0:eedb7d567a5d 75 uint32_t numSamples; /* total number of elements in the matrix */
robert_lp 0:eedb7d567a5d 76 uint32_t blkCnt; /* loop counters */
robert_lp 0:eedb7d567a5d 77 arm_status status; /* status of matrix scaling */
robert_lp 0:eedb7d567a5d 78
robert_lp 0:eedb7d567a5d 79 #if defined (ARM_MATH_DSP)
robert_lp 0:eedb7d567a5d 80
robert_lp 0:eedb7d567a5d 81 float32_t in1, in2, in3, in4; /* temporary variables */
robert_lp 0:eedb7d567a5d 82 float32_t out1, out2, out3, out4; /* temporary variables */
robert_lp 0:eedb7d567a5d 83
robert_lp 0:eedb7d567a5d 84 #endif // #if defined (ARM_MATH_DSP)
robert_lp 0:eedb7d567a5d 85
robert_lp 0:eedb7d567a5d 86 #ifdef ARM_MATH_MATRIX_CHECK
robert_lp 0:eedb7d567a5d 87 /* Check for matrix mismatch condition */
robert_lp 0:eedb7d567a5d 88 if ((pSrc->numRows != pDst->numRows) || (pSrc->numCols != pDst->numCols))
robert_lp 0:eedb7d567a5d 89 {
robert_lp 0:eedb7d567a5d 90 /* Set status as ARM_MATH_SIZE_MISMATCH */
robert_lp 0:eedb7d567a5d 91 status = ARM_MATH_SIZE_MISMATCH;
robert_lp 0:eedb7d567a5d 92 }
robert_lp 0:eedb7d567a5d 93 else
robert_lp 0:eedb7d567a5d 94 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
robert_lp 0:eedb7d567a5d 95 {
robert_lp 0:eedb7d567a5d 96 /* Total number of samples in the input matrix */
robert_lp 0:eedb7d567a5d 97 numSamples = (uint32_t) pSrc->numRows * pSrc->numCols;
robert_lp 0:eedb7d567a5d 98
robert_lp 0:eedb7d567a5d 99 #if defined (ARM_MATH_DSP)
robert_lp 0:eedb7d567a5d 100
robert_lp 0:eedb7d567a5d 101 /* Run the below code for Cortex-M4 and Cortex-M3 */
robert_lp 0:eedb7d567a5d 102
robert_lp 0:eedb7d567a5d 103 /* Loop Unrolling */
robert_lp 0:eedb7d567a5d 104 blkCnt = numSamples >> 2;
robert_lp 0:eedb7d567a5d 105
robert_lp 0:eedb7d567a5d 106 /* First part of the processing with loop unrolling. Compute 4 outputs at a time.
robert_lp 0:eedb7d567a5d 107 ** a second loop below computes the remaining 1 to 3 samples. */
robert_lp 0:eedb7d567a5d 108 while (blkCnt > 0U)
robert_lp 0:eedb7d567a5d 109 {
robert_lp 0:eedb7d567a5d 110 /* C(m,n) = A(m,n) * scale */
robert_lp 0:eedb7d567a5d 111 /* Scaling and results are stored in the destination buffer. */
robert_lp 0:eedb7d567a5d 112 in1 = pIn[0];
robert_lp 0:eedb7d567a5d 113 in2 = pIn[1];
robert_lp 0:eedb7d567a5d 114 in3 = pIn[2];
robert_lp 0:eedb7d567a5d 115 in4 = pIn[3];
robert_lp 0:eedb7d567a5d 116
robert_lp 0:eedb7d567a5d 117 out1 = in1 * scale;
robert_lp 0:eedb7d567a5d 118 out2 = in2 * scale;
robert_lp 0:eedb7d567a5d 119 out3 = in3 * scale;
robert_lp 0:eedb7d567a5d 120 out4 = in4 * scale;
robert_lp 0:eedb7d567a5d 121
robert_lp 0:eedb7d567a5d 122
robert_lp 0:eedb7d567a5d 123 pOut[0] = out1;
robert_lp 0:eedb7d567a5d 124 pOut[1] = out2;
robert_lp 0:eedb7d567a5d 125 pOut[2] = out3;
robert_lp 0:eedb7d567a5d 126 pOut[3] = out4;
robert_lp 0:eedb7d567a5d 127
robert_lp 0:eedb7d567a5d 128 /* update pointers to process next sampels */
robert_lp 0:eedb7d567a5d 129 pIn += 4U;
robert_lp 0:eedb7d567a5d 130 pOut += 4U;
robert_lp 0:eedb7d567a5d 131
robert_lp 0:eedb7d567a5d 132 /* Decrement the numSamples loop counter */
robert_lp 0:eedb7d567a5d 133 blkCnt--;
robert_lp 0:eedb7d567a5d 134 }
robert_lp 0:eedb7d567a5d 135
robert_lp 0:eedb7d567a5d 136 /* If the numSamples is not a multiple of 4, compute any remaining output samples here.
robert_lp 0:eedb7d567a5d 137 ** No loop unrolling is used. */
robert_lp 0:eedb7d567a5d 138 blkCnt = numSamples % 0x4U;
robert_lp 0:eedb7d567a5d 139
robert_lp 0:eedb7d567a5d 140 #else
robert_lp 0:eedb7d567a5d 141
robert_lp 0:eedb7d567a5d 142 /* Run the below code for Cortex-M0 */
robert_lp 0:eedb7d567a5d 143
robert_lp 0:eedb7d567a5d 144 /* Initialize blkCnt with number of samples */
robert_lp 0:eedb7d567a5d 145 blkCnt = numSamples;
robert_lp 0:eedb7d567a5d 146
robert_lp 0:eedb7d567a5d 147 #endif /* #if defined (ARM_MATH_DSP) */
robert_lp 0:eedb7d567a5d 148
robert_lp 0:eedb7d567a5d 149 while (blkCnt > 0U)
robert_lp 0:eedb7d567a5d 150 {
robert_lp 0:eedb7d567a5d 151 /* C(m,n) = A(m,n) * scale */
robert_lp 0:eedb7d567a5d 152 /* The results are stored in the destination buffer. */
robert_lp 0:eedb7d567a5d 153 *pOut++ = (*pIn++) * scale;
robert_lp 0:eedb7d567a5d 154
robert_lp 0:eedb7d567a5d 155 /* Decrement the loop counter */
robert_lp 0:eedb7d567a5d 156 blkCnt--;
robert_lp 0:eedb7d567a5d 157 }
robert_lp 0:eedb7d567a5d 158
robert_lp 0:eedb7d567a5d 159 /* Set status as ARM_MATH_SUCCESS */
robert_lp 0:eedb7d567a5d 160 status = ARM_MATH_SUCCESS;
robert_lp 0:eedb7d567a5d 161 }
robert_lp 0:eedb7d567a5d 162
robert_lp 0:eedb7d567a5d 163 /* Return to application */
robert_lp 0:eedb7d567a5d 164 return (status);
robert_lp 0:eedb7d567a5d 165 }
robert_lp 0:eedb7d567a5d 166
robert_lp 0:eedb7d567a5d 167 /**
robert_lp 0:eedb7d567a5d 168 * @} end of MatrixScale group
robert_lp 0:eedb7d567a5d 169 */
robert_lp 0:eedb7d567a5d 170