Skip to content

/github/workspace/src/MatrixFunctions/mat_mult_trans/kernels/plp_mat_mult_trans_q8s_rv32im.c

Functions

Name
void plp_mat_mult_trans_q8s_rv32im(const int8_t restrict pSrcA, const int8_t restrict pSrcB, uint32_t M, uint32_t N, uint32_t O, uint32_t shift, int8_t *restrict pDstC)
Matrix multiplication of 8-bit integer matrices kernel for RV32IM extension.

Defines

Name
BASIC_VERSION

Functions Documentation

function plp_mat_mult_trans_q8s_rv32im

void plp_mat_mult_trans_q8s_rv32im(
    const int8_t *__restrict__ pSrcA,
    const int8_t *__restrict__ pSrcB,
    uint32_t M,
    uint32_t N,
    uint32_t O,
    uint32_t shift,
    int8_t *__restrict__ pDstC
)

Matrix multiplication of 8-bit integer matrices kernel for RV32IM extension.

Parameters:

  • pSrcA points to the first input matrix
  • pSrcB points to the second input matrix, stored transposed in memory
  • 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
  • shift Amount to shift the result of each multiplication.
  • pDstC points to the output matrix

Return: none

Par: Fix-Point and Shifting

The result will be shifted by the parameter shift to the right (multiplied by 2^-shift). Assume that matrix A is represented as pSrcA * 2^-x, and matrix B as pSrcB * 2^-y (in other words, A has it's x last digits after the binary point). Then, the output is represented as pDstC * 2^-(x + y - shift).

matrix transposed matrix multiplication of a 8-bit fix-point matrices for RV32IM extension.

The output of the matrix multiplication will also be stored as an 8-bit array. Set the shift parameter such that no overflow ocurrs.

Macros Documentation

define BASIC_VERSION

#define BASIC_VERSION 

Source code

/* =====================================================================
 * Project:      PULP DSP Library
 * Title:        plp_mat_mult_trans_q8s_rv32im.c
 * Description:  32-bit integer matrix multiplication for XPULPV2
 *
 * $Date:        18. July 2019
 * $Revision:    V0
 *
 * Target Processor: PULP cores
 * ===================================================================== */
/*
 * Copyright (C) 2020 ETH Zurich and University of Bologna.
 *
 * Author: Tibor Schneider, 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"

void plp_mat_mult_trans_q8s_rv32im(const int8_t *__restrict__ pSrcA,
                                   const int8_t *__restrict__ pSrcB,
                                   uint32_t M,
                                   uint32_t N,
                                   uint32_t O,
                                   uint32_t shift,
                                   int8_t *__restrict__ pDstC) {

#define BASIC_VERSION // if used don't forget to also use the undefine at end of file
#ifdef BASIC_VERSION

    int32_t round = 1 << (shift - 1);

    uint32_t m; // loop counter
    uint32_t n; // loop counter
    uint32_t o; // loop counter

    for (m = 0; m < M; m++) {
        for (o = 0; o < O; o++) {
            int32_t sum = 0;
            for (n = 0; n < N; n++) {
                int32_t valA = (int32_t)pSrcA[m * N + n];
                int32_t valB = (int32_t)pSrcB[o * N + n];
                sum += (valA * valB + round) >> shift;
            }
            pDstC[m * O + o] = (int8_t)sum;
        }
    }

#else

    // TODO hackathon

#endif
#undef BASIC_VERSION
}

Updated on 2023-03-01 at 16:16:33 +0000