mpi_multi_matrice.cpp

#include <cstdlib>
#include <ctime>
#include <random>
#include <mpi.h>

const int limit = 10;

int main(int argc, char* argv[]) {

    int n = strtol(argv[1], NULL, 10);
    // int num_thrs = strtol(argv[2], NULL, 10);
    
    int i, j, k;
    int local_rows;
    
    int* A = NULL;
    int* B = new int[n * n];
    int* C = NULL;
    int rank, size;
    MPI_Init(NULL, NULL);
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    MPI_Comm_size(MPI_COMM_WORLD, &size);

    int rem = n % size;
    local_rows = (rank < rem) ? n / size + 1 : n / size;

    int* local_A = new int[local_rows * n];
    int* local_C = new int[local_rows * n]();

    if (rank == 0) {
        A = new int[n * n];
        C = new int[n * n]();
        srand(static_cast<unsigned>(time(0)));
        for (i = 0; i < n; i++) {
            for (j = 0; j < n; j++) {
                A[i * n + j] = rand() % limit;
                B[i * n + j] = rand() % limit;
            }
        }
        
    }
    
    int* sendcounts = nullptr;
    int* displs = nullptr;
    if (rank == 0) {
        sendcounts = new int[size];
        displs = new int[size];
        int disp = 0;
        for (int r = 0; r < size; r++) {
            sendcounts[r] = ((r < rem) ? n / size + 1 : n / size) * n;
            displs[r] = disp;
            disp += sendcounts[r];
        }
    }

    MPI_Scatterv(A, sendcounts, displs, MPI_INT, local_A, local_rows * n, MPI_INT, 0, MPI_COMM_WORLD);

    MPI_Bcast(B, n * n, MPI_INT, 0, MPI_COMM_WORLD);
    double start, end;
    start = MPI_Wtime();

    for (i = 0; i < local_rows; i++) {
        for (j = 0; j < n; j++) {
            for (k = 0; k < n; k++) {
                local_C[i * n + j] += local_A[i * n + k] * B[k * n + j];
            }
        }
    }

    end = MPI_Wtime();

    MPI_Gatherv(local_C, local_rows * n, MPI_INT, C, sendcounts, displs, MPI_INT, 0, MPI_COMM_WORLD);

    
    double duration = end - start;


    if (rank == 0) {
        printf("cost time: \t%lfs\n", duration);
        delete[] A;
        delete[] C;
        delete[] sendcounts;
        delete[] displs;
    }
    delete[] local_A;
    delete[] local_C;
    delete[] B;

    MPI_Finalize();
    return 0;
}

// mpiexec ./mpi_multi_matrice <the amount of data> -n <the number of processors>