/github/workspace/src/MatrixFunctions/mat_mult/kernels/plp_mat_mult_i16s_rv32im.c
Functions
| Name | |
|---|---|
| void | plp_mat_mult_i16s_rv32im(const int16_t restrict pSrcA, const int16_t restrict pSrcB, uint32_t M, uint32_t N, uint32_t O, int32_t *restrict pDstC) Matrix multiplication of 16-bit integer matrices kernel for RV32IM extension. |
Functions Documentation
function plp_mat_mult_i16s_rv32im
void plp_mat_mult_i16s_rv32im(
const int16_t *__restrict__ pSrcA,
const int16_t *__restrict__ pSrcB,
uint32_t M,
uint32_t N,
uint32_t O,
int32_t *__restrict__ pDstC
)
Matrix multiplication of 16-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 matrix multiplication of a 16-bit integer matrices for RV32IM extension.
Source code
/* =====================================================================
* Project: PULP DSP Library
* Title: plp_mat_mult_i16s_rv32im.c
* Description: 16-bit matrix multiplication kernel for RV32IM
*
* $Date: 18. July 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_i16s_rv32im(const int16_t *__restrict__ pSrcA,
const int16_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[j * O + k];
}
pDstC[i * O + k] = sum;
}
}
}
#else
void plp_mat_mult_i16s_rv32im(const int16_t *__restrict__ pSrcA,
const int16_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 / 2; i++) {
for (k = 0; k < O / 2; k++) {
int32_t sum00 = 0;
int32_t sum01 = 0;
int32_t sum10 = 0;
int32_t sum11 = 0;
for (j = 0; j < N / 2; j++) {
int32_t AVal00 = pSrcA[(i * 2) * N + j * 2];
int32_t AVal10 = pSrcA[(i * 2 + 1) * N + j * 2];
int32_t AVal01 = pSrcA[(i * 2) * N + j * 2 + 1];
int32_t AVal11 = pSrcA[(i * 2 + 1) * N + j * 2 + 1];
int32_t BVal00 = pSrcB[(j * 2) * O + (k * 2)];
int32_t BVal01 = pSrcB[(j * 2) * O + (k * 2 + 1)];
int32_t BVal10 = pSrcB[(j * 2 + 1) * O + (k * 2)];
int32_t BVal11 = pSrcB[(j * 2 + 1) * O + (k * 2 + 1)];
sum00 = sum00 + AVal00 * BVal00;
sum00 = sum00 + AVal01 * BVal10;
sum01 = sum01 + AVal00 * BVal01;
sum01 = sum01 + AVal01 * BVal11;
sum10 = sum10 + AVal10 * BVal00;
sum10 = sum10 + AVal11 * BVal10;
sum11 = sum11 + AVal10 * BVal01;
sum11 = sum11 + AVal11 * BVal11;
}
pDstC[(i * 2) * O + (k * 2)] = sum00;
pDstC[(i * 2) * O + (k * 2 + 1)] = sum01;
pDstC[(i * 2 + 1) * O + (k * 2)] = sum10;
pDstC[(i * 2 + 1) * O + (k * 2 + 1)] = sum11;
}
}
// clean up code
i = i * 2;
j = j * 2;
k = k * 2;
// clean up code
// check if every index is nicely finished
if (i == M && j == N && k == O) {
return;
} else {
uint32_t iEnd = i;
uint32_t jEnd = j;
uint32_t kEnd = k;
// clean up for j
if (jEnd != N) {
for (i = 0; i < iEnd; i++) {
for (k = 0; k < kEnd; k++) {
int32_t sum = 0;
for (j = jEnd; j < N; j++) {
sum += sum + pSrcA[i * N + j] * pSrcB[j * O + k];
}
pDstC[i * O + k] += sum;
}
}
}
// clean up for k
if (kEnd != O) {
for (i = 0; i < iEnd; i++) {
for (k = kEnd; k < O; k++) {
int32_t sum = 0;
for (j = 0; j < N; j++) {
sum = sum + pSrcA[i * N + j] * pSrcB[j * O + k];
}
pDstC[i * O + k] = sum;
}
}
}
// clean up for i
for (i = iEnd; 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[j * O + k];
}
pDstC[i * O + k] = sum;
}
}
}
}
#endif
// #undef BASIC_VERSION
Updated on 2023-03-01 at 16:16:33 +0000