串行代码的并行MPI实现

Parallel MPI Implementation of Serial Code

本文关键字：MPI 实现 并行 代码      更新时间：2023-10-16

我有这样的代码，它使用欧拉方法来解决简单的ODE-dy/dt=2y-2t^2-3。串行代码获取y_0的随机值并计算结果。我想使用MPI向每个节点发送它自己的y_0值并收集结果。我只需要使用MPI_SEND和MPI_RECV吗？这是我的序列号。

#include <iostream>
#include <cmath>
using namespace std;
double tmax = 2.0; //max time value
double h = 0.00000005; //grid spacing
double imax = tmax/h;
double tarray [40000000] = {0}; //array for time values
double xarray [40000000] = {0}; //array for x values
double yarray [40000000] = {0}; //array for y values
//serial calculation for euler's method problem 1.7.1
void final171(double y0)
{
    yarray [0] = y0;
    for (int k = 0; k < imax; k++) {
        tarray[k+1] = tarray[k] + h;
        yarray[k+1] = yarray[k] + h * (2 * yarray[k] - 2 * tarray[k] * tarray[k] - 3);
    }
    cout << "Our Solution Euler: " << yarray[int(imax)];
}

int main(int argc, const char * argv[])
{
    clock_t serialStart;
    double serialTotal;
    srand((unsigned)time(0));
    serialStart = clock();
    for (int i = 0; i < 128; i++) {
        double floor = 0.5, ceiling = 3.5, range = (ceiling - floor);
        double rnd = floor + double((range * rand()) / (RAND_MAX + 1.0));
        final171(rnd);
        cout << ", for y0 = " << rnd << endl;
    }
    serialTotal = (double)(clock() - serialStart) / (double)(CLOCKS_PER_SEC);
    cout << endl << "N = " << imax << endl;
    cout << "Serial Time (s)" << serialTotal << endl;
    return 0;
}

使用MPI_SEND/MPI_RECV当然可以。您也可以使用MPI_SCATTER，因为它可以做同样的事情，但更有效。

相关文章：