基于OpenMP矩阵并行化

OpenMP C++代码
将矩阵A按行分块，放入六个内核中计算Ci=Ai*B,i=(0,1,2,3,4,5)。

#include <iostream>  
#include <omp.h> // OpenMP编程需要包含的头文件
#include <time.h>
#include <stdlib.h>

using namespace std;

#define MatrixOrder 1000

int A[MatrixOrder][MatrixOrder] = { 0 };
int B[MatrixOrder][MatrixOrder] = { 0 };
int C[MatrixOrder][MatrixOrder] = { 0 };

void matrixInit()
{
    for (int row = 0; row < MatrixOrder; row++) {
        for (int col = 0; col < MatrixOrder; col++) {
            srand(row + col);
            A[row][col] = (rand() % 10);
            B[row][col] = (rand() % 10);
        }
    }
    //#pragma omp barrier
}




int main()
{
    matrixInit();

    clock_t t1 = clock(); //开始计时；
    for (int row = 0; row < MatrixOrder; row++) {
        for (int col = 0; col < MatrixOrder; col++) {
            for (int trans = 0; trans < MatrixOrder; trans++) {
                C[row][col] += A[row][trans] * B[trans][col];


            }
            // printf("%d ", C[row][col]);
            C[row][col] = 0;
        }
        // printf("\n");
    }
    clock_t t2 = clock(); //结束计时
    cout << "串行time: " << t2 - t1 << endl;
    clock_t t3 = clock(); //开始计时；
    omp_set_num_threads(6);

#pragma omp parallel

    {
       


#pragma omp sections 

        {

#pragma omp section

            for (int row = 0; row < MatrixOrder / 6; row++) {
                for (int col = 0; col < MatrixOrder; col++) {
                    for (int trans = 0; trans < MatrixOrder; trans++) {
                        C[row][col] += A[row][trans] * B[trans][col];
                    }
                }
            }





#pragma omp section



            for (int row = MatrixOrder / 6; row < MatrixOrder / 3; row++) {
                for (int col = 0; col < MatrixOrder; col++) {
                    for (int trans = 0; trans < MatrixOrder; trans++) {
                        C[row][col] += A[row][trans] * B[trans][col];
                    }
                }
            }



#pragma omp section



            for (int row = MatrixOrder / 3; row < MatrixOrder / 2; row++) {
                for (int col = 0; col < MatrixOrder; col++) {
                    for (int trans = 0; trans < MatrixOrder; trans++) {
                        C[row][col] += A[row][trans] * B[trans][col];
                    }
                }
            }



#pragma omp section



            for (int row = MatrixOrder / 2; row < 2 * MatrixOrder / 3; row++) {
                for (int col = 0; col < MatrixOrder; col++) {
                    for (int trans = 0; trans < MatrixOrder; trans++) {
                        C[row][col] += A[row][trans] * B[trans][col];
                    }
                }
            }

#pragma omp section



            for (int row = 2 * MatrixOrder / 3; row < 5 * MatrixOrder / 6; row++) {
                for (int col = 0; col < MatrixOrder; col++) {
                    for (int trans = 0; trans < MatrixOrder; trans++) {
                        C[row][col] += A[row][trans] * B[trans][col];
                    }
                }
            }

#pragma omp section



            for (int row = 5 * MatrixOrder / 6; row < MatrixOrder; row++) {
                for (int col = 0; col < MatrixOrder; col++) {
                    for (int trans = 0; trans < MatrixOrder; trans++) {
                        C[row][col] += A[row][trans] * B[trans][col];
                    }
                }
            }


        }


    }
    clock_t t4 = clock(); //结束计时
    cout << "并行time: " << t4 - t3 << endl;
    cout << "加速比：" << double((t2-t1)/double((t4-t3)))<<endl;
       //for (int row = 0; row < MatrixOrder; row++) {
       //    for (int col = 0; col < MatrixOrder; col++) {
       //        for (int trans = 0; trans < MatrixOrder; trans++) {


       //        }
       //        printf("%d ", C[row][col]);
       //    }
       //    printf("\n");
       //}
    system("pause");

    return 0;
}