/github/workspace/src/MatrixFunctions/mat_mult_trans/kernels/plp_mat_mult_trans_i32s_rv32im.c
Functions
| Name | |
|---|---|
| void | plp_mat_mult_trans_i32s_rv32im(const int32_t restrict pSrcA, const int32_t restrict pSrcB, uint32_t M, uint32_t N, uint32_t O, int32_t *restrict pDstC) Matrix multiplication of 32-bit integer matrices kernel for RV32IM extension. |
Functions Documentation
function plp_mat_mult_trans_i32s_rv32im
void plp_mat_mult_trans_i32s_rv32im(
const int32_t *__restrict__ pSrcA,
const int32_t *__restrict__ pSrcB,
uint32_t M,
uint32_t N,
uint32_t O,
int32_t *__restrict__ pDstC
)
Matrix multiplication of 32-bit integer matrices kernel for RV32IM extension.
Parameters:
- pSrcA points to the first input matrix
- pSrcB points to the second input matrix
- M height of the first input matrix
- N width of the first input matrix and hight of the second
- O width of the second input matrix
- pDstC points to the output matrix
Return: none
Matrix transposed matrix multiplication of a 32-bit integer matrices for RV32IM extension.
Source code
/* =====================================================================
* Project: PULP DSP Library
* Title: plp_mat_mult_i32s_rv32im.c
* Description: 32-bit matrix nultiplication kernel for RV32IM
*
* $Date: 22. December 2019
* $Revision: V0
*
* Target Processor: PULP cores
* ===================================================================== */
/*
* Copyright (C) 2019 ETH Zurich and University of Bologna.
*
* Author: Tom Kuchler, ETH Zurich
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "plp_math.h"
// #define BASIC_VERSION // if used don' forget to also use undefine at end of file
#ifdef BASIC_VERSION
void plp_mat_mult_trans_i32s_rv32im(const int32_t *__restrict__ pSrcA,
const int32_t *__restrict__ pSrcB,
uint32_t M,
uint32_t N,
uint32_t O,
int32_t *__restrict__ pDstC) {
uint32_t i = 0; // loop counter
uint32_t j = 0; // loop counter
uint32_t k = 0; // loop counter
for (i = 0; i < M; i++) {
for (k = 0; k < O; k++) {
int32_t sum = 0;
for (j = 0; j < N; j++) {
sum = sum + pSrcA[i * N + j] * pSrcB[k * N + j];
}
pDstC[i * O + k] = sum;
}
}
}
#else
void plp_mat_mult_trans_i32s_rv32im(const int32_t *__restrict__ pSrcA,
const int32_t *__restrict__ pSrcB,
uint32_t M,
uint32_t N,
uint32_t O,
int32_t *__restrict__ pDstC) {
uint32_t i = 0; // loop counter
uint32_t j = 0; // loop counter
uint32_t k = 0; // loop counter
uint32_t mod = N & 0x3;
if (mod == 3) {
for (i = 0; i < M; i++) {
for (k = 0; k < O; k++) {
int32_t sum1 = 0;
int32_t sum2 = 0;
int32_t sum3 = 0;
int32_t sum4 = 0;
for (j = 0; j < N / 4; j++) {
sum1 = sum1 + pSrcA[i * N + j * 4] * pSrcB[k * N + j * 4];
sum2 = sum2 + pSrcA[i * N + j * 4 + 1] * pSrcB[k * N + j * 4 + 1];
sum3 = sum3 + pSrcA[i * N + j * 4 + 2] * pSrcB[k * N + j * 4 + 2];
sum4 = sum4 + pSrcA[i * N + j * 4 + 3] * pSrcB[k * N + j * 4 + 3];
}
int32_t remaining = pSrcA[i * N + N - 1] * pSrcB[k * N + N - 1] +
pSrcA[i * N + N - 2] * pSrcB[k * N + N - 2] +
pSrcA[i * N + N - 3] * pSrcB[k * N + N - 3];
pDstC[i * O + k] = sum1 + sum2 + sum3 + sum4 + remaining;
}
}
} else if (mod == 2) {
for (i = 0; i < M; i++) {
for (k = 0; k < O; k++) {
int32_t sum1 = 0;
int32_t sum2 = 0;
int32_t sum3 = 0;
int32_t sum4 = 0;
for (j = 0; j < N / 4; j++) {
sum1 = sum1 + pSrcA[i * N + j * 4] * pSrcB[k * N + j * 4];
sum2 = sum2 + pSrcA[i * N + j * 4 + 1] * pSrcB[k * N + j * 4 + 1];
sum3 = sum3 + pSrcA[i * N + j * 4 + 2] * pSrcB[k * N + j * 4 + 2];
sum4 = sum4 + pSrcA[i * N + j * 4 + 3] * pSrcB[k * N + j * 4 + 3];
}
int32_t remaining = pSrcA[i * N + N - 1] * pSrcB[k * N + N - 1] +
pSrcA[i * N + N - 2] * pSrcB[k * N + N - 2];
pDstC[i * O + k] = sum1 + sum2 + sum3 + sum4 + remaining;
}
}
} else if (mod == 1) {
for (i = 0; i < M; i++) {
for (k = 0; k < O; k++) {
int32_t sum1 = 0;
int32_t sum2 = 0;
int32_t sum3 = 0;
int32_t sum4 = 0;
for (j = 0; j < N / 4; j++) {
sum1 = sum1 + pSrcA[i * N + j * 4] * pSrcB[k * N + j * 4];
sum2 = sum2 + pSrcA[i * N + j * 4 + 1] * pSrcB[k * N + j * 4 + 1];
sum3 = sum3 + pSrcA[i * N + j * 4 + 2] * pSrcB[k * N + j * 4 + 2];
sum4 = sum4 + pSrcA[i * N + j * 4 + 3] * pSrcB[k * N + j * 4 + 3];
}
pDstC[i * O + k] =
sum1 + sum2 + sum3 + sum4 + pSrcA[i * N + N - 1] * pSrcB[k * N + N - 1];
}
}
} else {
for (i = 0; i < M; i++) {
for (k = 0; k < O; k++) {
int32_t sum1 = 0;
int32_t sum2 = 0;
int32_t sum3 = 0;
int32_t sum4 = 0;
for (j = 0; j < N / 4; j++) {
sum1 = sum1 + pSrcA[i * N + j * 4] * pSrcB[k * N + j * 4];
sum2 = sum2 + pSrcA[i * N + j * 4 + 1] * pSrcB[k * N + j * 4 + 1];
sum3 = sum3 + pSrcA[i * N + j * 4 + 2] * pSrcB[k * N + j * 4 + 2];
sum4 = sum4 + pSrcA[i * N + j * 4 + 3] * pSrcB[k * N + j * 4 + 3];
}
pDstC[i * O + k] = sum1 + sum2 + sum3 + sum4;
}
}
}
}
#endif
// #undef BASIC_VERSION
Updated on 2023-03-01 at 16:16:33 +0000