时间戳

首页 > 解决方案 > 在java中的main之外初始化MPI

java - 在java中的main之外初始化MPI

问题描述

我试图在 java 中创建一个分布式 MPI 类，它将为我做一些计算。不幸的是， MPI.Init 似乎需要主类中的 argv,argc。有没有办法在这个类中初始化 MPI？

public class distributed {

    public distributed(int mat[][],int n,int m) throws MPIException {

        MPI.Init(argv); // issue is here,i dont have "argv" to initialize with,causing null pointer exceptions in variables like "myrank"

        final int myrank = MPI.COMM_WORLD.Rank();
        final int size = MPI.COMM_WORLD.Size();

        final int rows = n;
        final int rowChunk = (rows+size-1)/size;

        final int startRow = myrank *rowChunk;
        int endRow = (myrank+1)*rowChunk;

        int[] newRow = new int[m];
           }
}

标签： javampimpj-express

解决方案

为什么不在您的主要方法中进行myrank初始化。size然后将它们作为Distributed类构造函数的参数传递？

public class Distributed {

    public Distributed(int mat[][],int n,int m, int myrank, int size) {
        // These two variables are now parameters of the constructor
        // final int myrank = MPI.COMM_WORLD.Rank();
        // final int size = MPI.COMM_WORLD.Size();

        final int rows = n;
        final int rowChunk = (rows+size-1)/size;

        final int startRow = myrank *rowChunk;
        int endRow = (myrank+1)*rowChunk;

        int[] newRow = new int[m];
    }

    public static void main (String [] args) {
        MPI.Init(args, args.length); 
        final int myrank = MPI.COMM_WORLD.Rank();
        final int size = MPI.COMM_WORLD.Size();

        Distributed d = new Distributed (..., myrank, size);
    }
}

推荐阅读