Matrix Multiplication using multi-threads gets segmentation fault

[Tosh2pointO]

Hello, I'm trying to write a program that calculates up to a 1024 x 1024 matrix using multi-threading. For example, I need to run a 1024 x 1024 using 256, 64, 16 or 4 threads. Or I need to run a 64 x 64 matrix using 16 or 4 threads. All the Matrices are square. I thought I coded my program correctly, however I get a segmentation fault when I use a 720 x 720 matrix or higher, heres the code.

#include <iostream>
#include <stdio.h>
#include <pthread.h>

using namespace std;


const int   DIM = 720; //works up to 719, crashes at 720
const int   num_of_thr = 4;
int         matrix_A[DIM][DIM];
int         matrix_B[DIM][DIM];
int         c[DIM][DIM];

struct v
{
    int i;
    int j;
};

//worker thread
void* matrix_multi(void* data)
{    
    for(int i = 0; i < DIM; i++)
    {
        for(int j = 0; j < DIM; j++)
        {
            c[i][j] = 0;
            for(int k = 0; k < DIM; k++)
            {
                c[i][j] += matrix_A[i][k] * matrix_B[k][j];
            }
        }
    }
    pthread_exit(0);
}

int main()
{

    pthread_t thr_id[DIM][DIM];
    pthread_attr_t thr_attr;
    pthread_attr_init(&thr_attr);



   //Filling the Matrices
    for(int i = 0; i < DIM; i++)
    {
        for(int j = 0; j < DIM; j++)
        {
            matrix_A[i][j]= i + j;
            matrix_B[i][j] = i + 3;
        }
    }


    //create the threads
    for(int i = 0; i < num_of_thr/2; i++)
    {
        for(int j = 0; j < num_of_thr/2; j++)
        {
            struct v *data = (struct v *) malloc(sizeof(struct v));
            data->i = i;
            data->j = j;
            pthread_create(&thr_id[i][j],NULL,matrix_multi,  &data);
        }
    }

    //joining the threads
    for(int i = 0; i < num_of_thr/2; i++)
    {
        for(int j = 0; j < num_of_thr/2; j++)
        {
        pthread_join(thr_id[i][j],NULL);
        }
    }

     return 0;
}

Any help would be appreciated, thanks in advance.

[ashitpro]

I just ran this code on my CentOS machine (g++ 4.1.2).
It worked pretty well, no seg fault. Even if I increase the number of threads and dimensions.

[newb16]

Seems that your compiler doesn't like huge variables in data segment (matrix_a) and/or stack (thr_id). Allocate them on the heap. (btw - why do you need thr_id be of size [DIM][DIM] when you only fill it to num_of_thr/2 in each dimension?

[Tosh2pointO]

Thanks for the quick help. I changed my code quite a bit since yesterday. However I now get a "invalid conversion from `void*' to `__pthread_t**" on line 63 of my code. Here is the updated code:

#include <pthread.h>
#include <stdlib.h>
#include <stdio.h>

#define SIZE 10			/* Size of matrices */
int N;				/* number of threads */

int A[SIZE][SIZE], B[SIZE][SIZE], C[SIZE][SIZE];

void fill_matrix(int m[SIZE][SIZE])
{
  int i, j, n = 0;
  for (i=0; i<SIZE; i++)
    for (j=0; j<SIZE; j++)
      m[i][j] = n++;
}

void print_matrix(int m[SIZE][SIZE])
{
  int i, j = 0;
  for (i=0; i<SIZE; i++) {
    printf("\n\t| ");
    for (j=0; j<SIZE; j++)
      printf("%2d ", m[i][j]);
    printf("|");
  }
}


void* mmult (void* slice)
{
  int s = (int)slice;
  int from = (s * SIZE)/N;	/* note that this 'slicing' works fine */
  int to = ((s+1) * SIZE)/N;	/* even if SIZE is not divisible by N */
  int i,j,k;

  printf("computing slice %d (from row %d to %d)\n", s, from, to-1);
  for (i=from; i<to; i++)
    for (j=0; j<SIZE; j++) {
      C[i][j]=0;
      for (k=0; k<SIZE; k++)
	C[i][j] += A[i][k]*B[k][j];
    }

  printf("finished slice %d\n", s);
  return 0;
}

int main(int argc, char *argv[])
{
  pthread_t *thread;
  int i;

  if (argc!=2) {
    printf("Usage: %s number_of_threads\n",argv[0]);
    exit(-1);
  }

  N=atoi(argv[1]);
  fill_matrix(A);
  fill_matrix(B);
  thread = malloc(N*sizeof(pthread_t));

  for (i=1; i<N; i++) {
    if (pthread_create (&thread[i], NULL, mmult, (void*)i) != 0 ) {
      perror("Can't create thread");
      exit(-1);
    }
  }

  /* master thread is thread 0 so: */
  mmult(0);

  for (i=1; i<N; i++) pthread_join (thread[i], NULL);

  printf("\n\n");
  print_matrix(A);
  printf("\n\n\t       * \n");
  print_matrix(B);
  printf("\n\n\t       = \n");
  print_matrix(C);
  printf("\n\n");

  return 0;

}

Any help would be appreciated, thanks in advance

[ashitpro]

Explicit type casting should work..

thread = (pthread_t *)malloc(N*sizeof(pthread_t));