Program 2Matrix Multiplication 1 Wrtie a PThreads program to

Program 2-Matrix Multiplication

1) Wrtie a PThreads program to calculate matrix multiplication

            a) You can start with the code from the examples page

            b) It should work with a varialble number of threads reguardless of matrix size

            c) From the command line It should accept the number of threads to use

                     1) For quick debugging purposes use a default number of threads

2) Compare to serial version

          a) Log into Cheetah, then SSH into amdquad01

          b) Run the serial program for matirces of size NxN, Where N=1024

          c) Run the parellel program for matrices of same size N

                       1)Time for the following number of threads: 1,2,3,4, 8, 16, 32

3) Deliverables

           a) Source Code

           b) Graph of Results

Solution

Source Code in Pthreads for matrix multiplication

#include<pthread.h>

#include<stdio.h>

#include<stdlib.h>

#define A 3

#define B 2

#define C 3

#define NUM_T 10 //number of threads

int x[A][B]={{1,2},{1,2},{3,2}};

int y[B][C]={{1,2,3},{4,5,6}};

int z[A][C];

struct s{

int m; //row

int n;//column

};

void * thread1(void *p); // the thread

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

{

int i,j,count=0;

for(i=0;i<A;i++){

for(j=0;j<C;j++){

//Assig a row and column for each thread

struct s* data=(struct s*)malloc(sizeof(struct s));

data ->m=i;

data ->n=j;

//Now create the thread

pthread_t tid; //thread Id

pthread_attr_t attr;// set of thread attributes

//default attributes

pthread_attr_init(&attr);

//finally create the thread

pthread_create(&tid,&attr,runner,data);

//use join method so that parent wait untill all threads complete

pthread_join(tid,NULL);

count++;

}

}

//resulting matrix

for(i=0;i<A;i++){

for(j=0;j<C;j++){

printf(\"\ %d\",z[i][j]);

}

}

//thread will begin control in this function

void * thread1(void *p)

{

struct s* data=p;//structure that holds data

int n1,sum1=0;//the counter and sum;

//row*column loop

for(n1=0;n1<B;n1++)

{

sum1+=A[data->m][n1]*B[n1][data->n];

}

//assign the sum in result

z[data->m][data->n]=sum1;

pthread_exit(0);

}

Pthread is POSIX thread is a parallel execution model. As given in condition that the program should work with variable number of threads regardless of matrix size