Question

请添加fork（），join（）的代码，只要它可以添加。我需要将此矩阵乘法并行化。

代码如下：

public static long[][] dc(long[][] M1, long[][] M2, int n)
{

    long[][] M = new long[n][n];

    if (n == 1)
      M[0][0] = M1[0][0] * M2[0][0];
    else 
    {
      int size= n/2;
      long[][] A11 = new long[n/2][n/2];
      long[][] A12 = new long[n/2][n/2];
      long[][] A21 = new long[n/2][n/2];
      long[][] A22 = new long[n/2][n/2];
      long[][] B11 = new long[n/2][n/2];
      long[][] B12 = new long[n/2][n/2];
      long[][] B21 = new long[n/2][n/2];
      long[][] B22 = new long[n/2][n/2];

        copy1(A11,M1,0,0,size);
        copy1(A12,M1,0,(n/2),size);
        copy1(A21,M1,(n/2),0,size);
        copy1(A22,M1,(n/2),(n/2),size);
        copy1(B11,M2,0,0,size);
        copy1(B12,M2,0,(n/2),size);
        copy1(B21,M2,(n/2),0,size);
        copy1(B22,M2,(n/2),(n/2),size);

      long[][] C11 = add( dc(A11, B11,size), dc(A12, B21,size) ,size);//need to fork
      long[][] C12 = add( dc(A11, B12,size), dc(A12, B22,size) ,size);//need to fork
      long[][] C21 = add( dc(A21, B11,size), dc(A22, B21,size) ,size);//need to fork
      long[][] C22 = add( dc(A21, A12,size), dc(A22, B22,size) ,size);//need to fork

        copy2(M,C11,0,0,size);
        copy2(M,C12,0,(n/2),size);
        copy2(M,C21,(n/2),0,size);
        copy2(M,C22,(n/2),(n/2),size);
    }
    return M;
}

我应该使用java还是fork（），exec（），pthread（）？

编辑：

这是我的并行化代码。它不适用于大于N = 32 我得到一个例外 - Java.outofmemory Exception

public static long[][] dc(long[][] M1, long[][] M2, int n)
{
    long[][] M=new long[n][n];
    if (n == 1)
      M[0][0] = M1[0][0] * M2[0][0];
    else 
    {

      int size= n/2;
      long[][] A11 = new long[n/2][n/2];
      long[][] A12 = new long[n/2][n/2];
      long[][] A21 = new long[n/2][n/2];
      long[][] A22 = new long[n/2][n/2];
      long[][] B11 = new long[n/2][n/2];
      long[][] B12 = new long[n/2][n/2];
      long[][] B21 = new long[n/2][n/2];
      long[][] B22 = new long[n/2][n/2];


        copy1(A11,M1,0,0,size);
        copy1(A12,M1,0,(n/2),size);
        copy1(A21,M1,(n/2),0,size);
        copy1(A22,M1,(n/2),(n/2),size);
        copy1(B11,M2,0,0,size);
        copy1(B12,M2,0,(n/2),size);
        copy1(B21,M2,(n/2),0,size);
        copy1(B22,M2,(n/2),(n/2),size);



     class T1 extends Thread
     {
        private long[][] p1;
        public void run()
        {
            p1=dc(A11,B11,size);
            //System.out.println("P1");
            //disp(p1);
        }
     };
     class T2 extends Thread
     {
        private long[][] p2;
        public void run()
        {
            p2=dc(A12,B21,size);
            //System.out.println("P2");
            //disp(p2);
        }
     };
     class T3 extends Thread
     {
        private long[][] p3;
        public void run()
        {
            p3=dc(A11,B12,size);
            //System.out.println("P3");
            //disp(p3);
        }
     };
     class T4 extends Thread
     {
        private long[][] p4;
        public void run()
        {
            p4=dc(A12,B22,size);
            //System.out.println("P4");
            //disp(p4);
        }
     };
     class T5 extends Thread
     {
        private long[][] p5;
        public void run()
        {
            p5=dc(A21,B11,size);
            //System.out.println("P5");
            //disp(p5);
        }
     };
     class T6 extends Thread
     {
        private long[][] p6;
        public void run()
        {
            p6=dc(A22,B21,size);
            //System.out.println("P6");
            //disp(p6);
        }
     };
     class T7 extends Thread
     {
        private long[][] p7;
        public void run()
        {
            p7=dc(A21,B12,size);
            //System.out.println("P7");
            //disp(p7);
        }
     };
     class T8 extends Thread
     {
        private long[][] p8;
        public void run()
        {
            p8=dc(A22,B22,size);
            //System.out.println("P8");
            //disp(p8);
        }
     };

     T1 t1=new T1(); T2 t2=new T2(); T3 t3=new T3(); T4 t4=new T4();
     T5 t5=new T5(); T6 t6=new T6(); T7 t7=new T7(); T8 t8=new T8();
     t1.start();t2.start();t3.start();t4.start();t5.start();t6.start();t7.start();t8.start();
     try 
     {
        t1.join();
        t2.join();
        t3.join();
        t4.join();
        t5.join();
        t6.join();
        t7.join();
        t8.join();
        //System.out.println("Threads completed");
        //disp(t1.p1);
     } 
     catch (InterruptedException e) 
     {
        // TODO Auto-generated catch block
        System.out.println("Exception occured");

     }
     long[][] C11=add(t1.p1,t2.p2,size);
     long[][] C12=add(t3.p3,t4.p4,size);
     long[][] C21=add(t5.p5,t6.p6,size);
     long[][] C22=add(t7.p7,t8.p8,size);

        copy2(M,C11,0,0,size);
        copy2(M,C12,0,(n/2),size);
        copy2(M,C21,(n/2),0,size);
        copy2(M,C22,(n/2),(n/2),size);
    }
    return M;
}

public static long[][] add(long[][] M1, long[][] M2, int size)
{
    long[][] M=new long[size][size];    
    for(int i=0;i<size;i++)
    {
        for(int j=0;j<size;j++)
        {
            M[i][j]=M1[i][j]+M2[i][j];
        }
    }
    //System.out.println("Add method");
    return M;

}

您能告诉我如何让线程适用于大量输入吗？这段代码实际上比没有线程的代码慢。 SO线程实际上根本没用。冷，请你更正代码，让我知道如何有效地并行化，以便它比没有并行化的代码更快？

谢谢！

如何在使用矩阵乘法划分和征服时在递归（Java）中使用fork join？

0 个答案: