如何使用MPI和OpenMP运行并行循环

#include <stdio.h>
#include <stdlib.h>
#include <mpi.h>
#include <omp.h>

#define ARRAY_SIZE 1000000

int main(int argc, char** argv) {
    int rank, size;
    int* array = (int*)malloc(ARRAY_SIZE * sizeof(int));
    long long sum = 0;

    // Initialize MPI
    MPI_Init(&argc, &argv);
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    MPI_Comm_size(MPI_COMM_WORLD, &size);

    // Initialize array
    for (int i = 0; i < ARRAY_SIZE; i++) {
        array[i] = rand() % 100;
    }

    // Calculate local sum using OpenMP
    long long local_sum = 0;
    #pragma omp parallel for reduction(+:local_sum)
    for (int i = rank * (ARRAY_SIZE / size); i < (rank + 1) * (ARRAY_SIZE / size); i++) {
        local_sum += array[i] * array[i];
    }

    // Gather all local sums using MPI
    MPI_Allreduce(&local_sum, &sum, 1, MPI_LONG_LONG, MPI_SUM, MPI_COMM_WORLD);

    // Print result
    if (rank == 0) {
        printf("Total sum of squares: %lld\n", sum);
    }

    // Finalize MPI
    MPI_Finalize();

    free(array);
    return 0;
}