Robert Lopez / CMSIS5

Aded CMSIS5 DSP and NN folder. Needs some work

DSP/source/BasicMathFunctions/arm_abs_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_abs_f32.c
robert_lp 0:eedb7d567a5d 4 * Description: Floating-point vector absolute value
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 #include <math.h>
robert_lp 0:eedb7d567a5d 31
robert_lp 0:eedb7d567a5d 32 /**
robert_lp 0:eedb7d567a5d 33 * @ingroup groupMath
robert_lp 0:eedb7d567a5d 34 */
robert_lp 0:eedb7d567a5d 35
robert_lp 0:eedb7d567a5d 36 /**
robert_lp 0:eedb7d567a5d 37 * @defgroup BasicAbs Vector Absolute Value
robert_lp 0:eedb7d567a5d 38 *
robert_lp 0:eedb7d567a5d 39 * Computes the absolute value of a vector on an element-by-element basis.
robert_lp 0:eedb7d567a5d 40 *
robert_lp 0:eedb7d567a5d 41 * <pre>
robert_lp 0:eedb7d567a5d 42 * pDst[n] = abs(pSrc[n]), 0 <= n < blockSize.
robert_lp 0:eedb7d567a5d 43 * </pre>
robert_lp 0:eedb7d567a5d 44 *
robert_lp 0:eedb7d567a5d 45 * The functions support in-place computation allowing the source and
robert_lp 0:eedb7d567a5d 46 * destination pointers to reference the same memory buffer.
robert_lp 0:eedb7d567a5d 47 * There are separate functions for floating-point, Q7, Q15, and Q31 data types.
robert_lp 0:eedb7d567a5d 48 */
robert_lp 0:eedb7d567a5d 49
robert_lp 0:eedb7d567a5d 50 /**
robert_lp 0:eedb7d567a5d 51 * @addtogroup BasicAbs
robert_lp 0:eedb7d567a5d 52 * @{
robert_lp 0:eedb7d567a5d 53 */
robert_lp 0:eedb7d567a5d 54
robert_lp 0:eedb7d567a5d 55 /**
robert_lp 0:eedb7d567a5d 56 * @brief Floating-point vector absolute value.
robert_lp 0:eedb7d567a5d 57 * @param[in] *pSrc points to the input buffer
robert_lp 0:eedb7d567a5d 58 * @param[out] *pDst points to the output buffer
robert_lp 0:eedb7d567a5d 59 * @param[in] blockSize number of samples in each vector
robert_lp 0:eedb7d567a5d 60 * @return none.
robert_lp 0:eedb7d567a5d 61 */
robert_lp 0:eedb7d567a5d 62
robert_lp 0:eedb7d567a5d 63 void arm_abs_f32(
robert_lp 0:eedb7d567a5d 64 float32_t * pSrc,
robert_lp 0:eedb7d567a5d 65 float32_t * pDst,
robert_lp 0:eedb7d567a5d 66 uint32_t blockSize)
robert_lp 0:eedb7d567a5d 67 {
robert_lp 0:eedb7d567a5d 68 uint32_t blkCnt; /* loop counter */
robert_lp 0:eedb7d567a5d 69
robert_lp 0:eedb7d567a5d 70 #if defined (ARM_MATH_DSP)
robert_lp 0:eedb7d567a5d 71
robert_lp 0:eedb7d567a5d 72 /* Run the below code for Cortex-M4 and Cortex-M3 */
robert_lp 0:eedb7d567a5d 73 float32_t in1, in2, in3, in4; /* temporary variables */
robert_lp 0:eedb7d567a5d 74
robert_lp 0:eedb7d567a5d 75 /*loop Unrolling */
robert_lp 0:eedb7d567a5d 76 blkCnt = blockSize >> 2U;
robert_lp 0:eedb7d567a5d 77
robert_lp 0:eedb7d567a5d 78 /* First part of the processing with loop unrolling. Compute 4 outputs at a time.
robert_lp 0:eedb7d567a5d 79 ** a second loop below computes the remaining 1 to 3 samples. */
robert_lp 0:eedb7d567a5d 80 while (blkCnt > 0U)
robert_lp 0:eedb7d567a5d 81 {
robert_lp 0:eedb7d567a5d 82 /* C = |A| */
robert_lp 0:eedb7d567a5d 83 /* Calculate absolute and then store the results in the destination buffer. */
robert_lp 0:eedb7d567a5d 84 /* read sample from source */
robert_lp 0:eedb7d567a5d 85 in1 = *pSrc;
robert_lp 0:eedb7d567a5d 86 in2 = *(pSrc + 1);
robert_lp 0:eedb7d567a5d 87 in3 = *(pSrc + 2);
robert_lp 0:eedb7d567a5d 88
robert_lp 0:eedb7d567a5d 89 /* find absolute value */
robert_lp 0:eedb7d567a5d 90 in1 = fabsf(in1);
robert_lp 0:eedb7d567a5d 91
robert_lp 0:eedb7d567a5d 92 /* read sample from source */
robert_lp 0:eedb7d567a5d 93 in4 = *(pSrc + 3);
robert_lp 0:eedb7d567a5d 94
robert_lp 0:eedb7d567a5d 95 /* find absolute value */
robert_lp 0:eedb7d567a5d 96 in2 = fabsf(in2);
robert_lp 0:eedb7d567a5d 97
robert_lp 0:eedb7d567a5d 98 /* read sample from source */
robert_lp 0:eedb7d567a5d 99 *pDst = in1;
robert_lp 0:eedb7d567a5d 100
robert_lp 0:eedb7d567a5d 101 /* find absolute value */
robert_lp 0:eedb7d567a5d 102 in3 = fabsf(in3);
robert_lp 0:eedb7d567a5d 103
robert_lp 0:eedb7d567a5d 104 /* find absolute value */
robert_lp 0:eedb7d567a5d 105 in4 = fabsf(in4);
robert_lp 0:eedb7d567a5d 106
robert_lp 0:eedb7d567a5d 107 /* store result to destination */
robert_lp 0:eedb7d567a5d 108 *(pDst + 1) = in2;
robert_lp 0:eedb7d567a5d 109
robert_lp 0:eedb7d567a5d 110 /* store result to destination */
robert_lp 0:eedb7d567a5d 111 *(pDst + 2) = in3;
robert_lp 0:eedb7d567a5d 112
robert_lp 0:eedb7d567a5d 113 /* store result to destination */
robert_lp 0:eedb7d567a5d 114 *(pDst + 3) = in4;
robert_lp 0:eedb7d567a5d 115
robert_lp 0:eedb7d567a5d 116
robert_lp 0:eedb7d567a5d 117 /* Update source pointer to process next sampels */
robert_lp 0:eedb7d567a5d 118 pSrc += 4U;
robert_lp 0:eedb7d567a5d 119
robert_lp 0:eedb7d567a5d 120 /* Update destination pointer to process next sampels */
robert_lp 0:eedb7d567a5d 121 pDst += 4U;
robert_lp 0:eedb7d567a5d 122
robert_lp 0:eedb7d567a5d 123 /* Decrement the loop counter */
robert_lp 0:eedb7d567a5d 124 blkCnt--;
robert_lp 0:eedb7d567a5d 125 }
robert_lp 0:eedb7d567a5d 126
robert_lp 0:eedb7d567a5d 127 /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
robert_lp 0:eedb7d567a5d 128 ** No loop unrolling is used. */
robert_lp 0:eedb7d567a5d 129 blkCnt = blockSize % 0x4U;
robert_lp 0:eedb7d567a5d 130
robert_lp 0:eedb7d567a5d 131 #else
robert_lp 0:eedb7d567a5d 132
robert_lp 0:eedb7d567a5d 133 /* Run the below code for Cortex-M0 */
robert_lp 0:eedb7d567a5d 134
robert_lp 0:eedb7d567a5d 135 /* Initialize blkCnt with number of samples */
robert_lp 0:eedb7d567a5d 136 blkCnt = blockSize;
robert_lp 0:eedb7d567a5d 137
robert_lp 0:eedb7d567a5d 138 #endif /* #if defined (ARM_MATH_DSP) */
robert_lp 0:eedb7d567a5d 139
robert_lp 0:eedb7d567a5d 140 while (blkCnt > 0U)
robert_lp 0:eedb7d567a5d 141 {
robert_lp 0:eedb7d567a5d 142 /* C = |A| */
robert_lp 0:eedb7d567a5d 143 /* Calculate absolute and then store the results in the destination buffer. */
robert_lp 0:eedb7d567a5d 144 *pDst++ = fabsf(*pSrc++);
robert_lp 0:eedb7d567a5d 145
robert_lp 0:eedb7d567a5d 146 /* Decrement the loop counter */
robert_lp 0:eedb7d567a5d 147 blkCnt--;
robert_lp 0:eedb7d567a5d 148 }
robert_lp 0:eedb7d567a5d 149 }
robert_lp 0:eedb7d567a5d 150
robert_lp 0:eedb7d567a5d 151 /**
robert_lp 0:eedb7d567a5d 152 * @} end of BasicAbs group
robert_lp 0:eedb7d567a5d 153 */
robert_lp 0:eedb7d567a5d 154