Robert Lopez / CMSIS5

Aded CMSIS5 DSP and NN folder. Needs some work

DSP/source/BasicMathFunctions/arm_mult_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_mult_f32.c
robert_lp 0:eedb7d567a5d 4 * Description: Floating-point vector multiplication
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 groupMath
robert_lp 0:eedb7d567a5d 33 */
robert_lp 0:eedb7d567a5d 34
robert_lp 0:eedb7d567a5d 35 /**
robert_lp 0:eedb7d567a5d 36 * @defgroup BasicMult Vector Multiplication
robert_lp 0:eedb7d567a5d 37 *
robert_lp 0:eedb7d567a5d 38 * Element-by-element multiplication of two vectors.
robert_lp 0:eedb7d567a5d 39 *
robert_lp 0:eedb7d567a5d 40 * <pre>
robert_lp 0:eedb7d567a5d 41 * pDst[n] = pSrcA[n] * pSrcB[n], 0 <= n < blockSize.
robert_lp 0:eedb7d567a5d 42 * </pre>
robert_lp 0:eedb7d567a5d 43 *
robert_lp 0:eedb7d567a5d 44 * There are separate functions for floating-point, Q7, Q15, and Q31 data types.
robert_lp 0:eedb7d567a5d 45 */
robert_lp 0:eedb7d567a5d 46
robert_lp 0:eedb7d567a5d 47 /**
robert_lp 0:eedb7d567a5d 48 * @addtogroup BasicMult
robert_lp 0:eedb7d567a5d 49 * @{
robert_lp 0:eedb7d567a5d 50 */
robert_lp 0:eedb7d567a5d 51
robert_lp 0:eedb7d567a5d 52 /**
robert_lp 0:eedb7d567a5d 53 * @brief Floating-point vector multiplication.
robert_lp 0:eedb7d567a5d 54 * @param[in] *pSrcA points to the first input vector
robert_lp 0:eedb7d567a5d 55 * @param[in] *pSrcB points to the second input vector
robert_lp 0:eedb7d567a5d 56 * @param[out] *pDst points to the output vector
robert_lp 0:eedb7d567a5d 57 * @param[in] blockSize number of samples in each vector
robert_lp 0:eedb7d567a5d 58 * @return none.
robert_lp 0:eedb7d567a5d 59 */
robert_lp 0:eedb7d567a5d 60
robert_lp 0:eedb7d567a5d 61 void arm_mult_f32(
robert_lp 0:eedb7d567a5d 62 float32_t * pSrcA,
robert_lp 0:eedb7d567a5d 63 float32_t * pSrcB,
robert_lp 0:eedb7d567a5d 64 float32_t * pDst,
robert_lp 0:eedb7d567a5d 65 uint32_t blockSize)
robert_lp 0:eedb7d567a5d 66 {
robert_lp 0:eedb7d567a5d 67 uint32_t blkCnt; /* loop counters */
robert_lp 0:eedb7d567a5d 68 #if defined (ARM_MATH_DSP)
robert_lp 0:eedb7d567a5d 69
robert_lp 0:eedb7d567a5d 70 /* Run the below code for Cortex-M4 and Cortex-M3 */
robert_lp 0:eedb7d567a5d 71 float32_t inA1, inA2, inA3, inA4; /* temporary input variables */
robert_lp 0:eedb7d567a5d 72 float32_t inB1, inB2, inB3, inB4; /* temporary input variables */
robert_lp 0:eedb7d567a5d 73 float32_t out1, out2, out3, out4; /* temporary output 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 * B */
robert_lp 0:eedb7d567a5d 83 /* Multiply the inputs and store the results in output buffer */
robert_lp 0:eedb7d567a5d 84 /* read sample from sourceA */
robert_lp 0:eedb7d567a5d 85 inA1 = *pSrcA;
robert_lp 0:eedb7d567a5d 86 /* read sample from sourceB */
robert_lp 0:eedb7d567a5d 87 inB1 = *pSrcB;
robert_lp 0:eedb7d567a5d 88 /* read sample from sourceA */
robert_lp 0:eedb7d567a5d 89 inA2 = *(pSrcA + 1);
robert_lp 0:eedb7d567a5d 90 /* read sample from sourceB */
robert_lp 0:eedb7d567a5d 91 inB2 = *(pSrcB + 1);
robert_lp 0:eedb7d567a5d 92
robert_lp 0:eedb7d567a5d 93 /* out = sourceA * sourceB */
robert_lp 0:eedb7d567a5d 94 out1 = inA1 * inB1;
robert_lp 0:eedb7d567a5d 95
robert_lp 0:eedb7d567a5d 96 /* read sample from sourceA */
robert_lp 0:eedb7d567a5d 97 inA3 = *(pSrcA + 2);
robert_lp 0:eedb7d567a5d 98 /* read sample from sourceB */
robert_lp 0:eedb7d567a5d 99 inB3 = *(pSrcB + 2);
robert_lp 0:eedb7d567a5d 100
robert_lp 0:eedb7d567a5d 101 /* out = sourceA * sourceB */
robert_lp 0:eedb7d567a5d 102 out2 = inA2 * inB2;
robert_lp 0:eedb7d567a5d 103
robert_lp 0:eedb7d567a5d 104 /* read sample from sourceA */
robert_lp 0:eedb7d567a5d 105 inA4 = *(pSrcA + 3);
robert_lp 0:eedb7d567a5d 106
robert_lp 0:eedb7d567a5d 107 /* store result to destination buffer */
robert_lp 0:eedb7d567a5d 108 *pDst = out1;
robert_lp 0:eedb7d567a5d 109
robert_lp 0:eedb7d567a5d 110 /* read sample from sourceB */
robert_lp 0:eedb7d567a5d 111 inB4 = *(pSrcB + 3);
robert_lp 0:eedb7d567a5d 112
robert_lp 0:eedb7d567a5d 113 /* out = sourceA * sourceB */
robert_lp 0:eedb7d567a5d 114 out3 = inA3 * inB3;
robert_lp 0:eedb7d567a5d 115
robert_lp 0:eedb7d567a5d 116 /* store result to destination buffer */
robert_lp 0:eedb7d567a5d 117 *(pDst + 1) = out2;
robert_lp 0:eedb7d567a5d 118
robert_lp 0:eedb7d567a5d 119 /* out = sourceA * sourceB */
robert_lp 0:eedb7d567a5d 120 out4 = inA4 * inB4;
robert_lp 0:eedb7d567a5d 121 /* store result to destination buffer */
robert_lp 0:eedb7d567a5d 122 *(pDst + 2) = out3;
robert_lp 0:eedb7d567a5d 123 /* store result to destination buffer */
robert_lp 0:eedb7d567a5d 124 *(pDst + 3) = out4;
robert_lp 0:eedb7d567a5d 125
robert_lp 0:eedb7d567a5d 126
robert_lp 0:eedb7d567a5d 127 /* update pointers to process next samples */
robert_lp 0:eedb7d567a5d 128 pSrcA += 4U;
robert_lp 0:eedb7d567a5d 129 pSrcB += 4U;
robert_lp 0:eedb7d567a5d 130 pDst += 4U;
robert_lp 0:eedb7d567a5d 131
robert_lp 0:eedb7d567a5d 132 /* Decrement the blockSize 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 blockSize 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 = blockSize % 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 = blockSize;
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 = A * B */
robert_lp 0:eedb7d567a5d 152 /* Multiply the inputs and store the results in output buffer */
robert_lp 0:eedb7d567a5d 153 *pDst++ = (*pSrcA++) * (*pSrcB++);
robert_lp 0:eedb7d567a5d 154
robert_lp 0:eedb7d567a5d 155 /* Decrement the blockSize loop counter */
robert_lp 0:eedb7d567a5d 156 blkCnt--;
robert_lp 0:eedb7d567a5d 157 }
robert_lp 0:eedb7d567a5d 158 }
robert_lp 0:eedb7d567a5d 159
robert_lp 0:eedb7d567a5d 160 /**
robert_lp 0:eedb7d567a5d 161 * @} end of BasicMult group
robert_lp 0:eedb7d567a5d 162 */
robert_lp 0:eedb7d567a5d 163