(Core dumped) ¿A too large array?

[Isolda_]

I'm writing a code which implementes the cell-to-cell mapping method.

Finally I know why my compiler is saying me there is a segmentation fault. Apparently it's because I defined a matrix of DIMENSION^2xDIMENSION^2, where DIMENSION is 100.

I need that matrix; i'm declaring the matrix as

 int prob[DIMENSION*DIMENSION][DIMENSION*DIMENSION]

¿There is a way to declare it in a way there is no error?

[Salem]

Think about what the size is - it's 400MB

You could replace 'int' with 'char', but perhaps you need to explore ways of partitioning your problem into smaller data sets.

[matsp]

That's probably OK if you make it a global/static array, but not if you have it as an automatic local variable - it uses 10000 x 10000 = 100000000 cells *4 bytes, which is far too much for stack storage.

--
Mats

[Isolda_]

ok. I don't see how to write in another way my code. Here is what I'm triyng to do.

I divided and x-y plane in little cell, each of 0.04*0.04. In each cell I started an initial condition of a dynamical system (Duffing-van Der Pol oscilator), and I solve it with RK4. In each cell I chose a number of random intial conditions; clearlñy each cell is characterized by two numbers, i and j. So, I use n to describe such a number.

for(i=0;i<DIMENSION;i++){
  for(j=0;j<DIMENSION;j++){
   n=i*200+j;
   l=j*200+i;

   
  prob[n][l]=0;
   
   for(k=0;k<RANDOM;k++){
      
	xout[0]= (drand48()+i-DIMENSION/2)/25.0;
	xout[1]= (drand48()+j-DIMENSION/2)/25.0;

	

	for(m = 0; m< NSTEPS; m++){
    	  for(u=0;u<NDIM;u++) xin[u]=xout[u];
	  runge_kutta (xin, tau, param, xout);
	         }
		salidax[n][k]=xout[0];
		saliday[n][k]=xout[1];		
           }
	}
}

Then I verified in wich cell the output is. Finally, if cell "n" leds to cell "l", probability of n->l increses by one.

   for(n=0;n<DIMENSION*DIMENSION;n++){
	   for(i=0;i<DIMENSION;i++){
  		for(j=0;j<DIMENSION;j++){
		l=i*200+j;
		
		for(k=0;k<RANDOM;k++){

	         if((i-DIMENSION/2)/25.0<salidax[n][k]<(i+1-DIMENSION/2)/25.0)
		   if((j-DIMENSION/2)/25.0<saliday[n][k]<(j+1-DIMENSION/2)/25.0) prob[n][l]++;
                  }
           
	    }
          }
        }

And that's it. But I'm not clever enogh to see another solution. ¿Do you have sugestions?

Thanks for your answers!

[Isolda_]

¿There is a way to define the array dynamically? ¿Can I declare it like

int *prob

, without giving the size it will have?

[Salem]

Sure, if the minor size is constant, then it's one easy declaration and malloc to allocate the whole array.

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

#define DIMENSION 100

int main() {
    int (*prob)[DIMENSION*DIMENSION];
    printf("Allocating &#37;ld\n", DIMENSION*DIMENSION * sizeof *prob );
    prob = malloc( DIMENSION*DIMENSION * sizeof *prob );

    prob[0][0] = 0;

    free( prob );

    return 0;
}

[Isolda_]

I'm trying with this, and at the moment it seems to work fine.

  
int **prob=(int **)malloc(DIMENSION*DIMENSION*sizeof (int*));

  for(i=0;i<DIMENSION*DIMENSION;i++){
    prob[i]=(int*)malloc(DIMENSION*DIMENSION*sizeof(int));
  if (prob == NULL) {
    printf("No hay ubicación de memoria suficiente");
    exit(0);
              }
      }

[Salem]

Have you read the FAQ yet on why casting malloc is bad ?

[Isolda_]

Sorry, no.

I changed my code:

 int **prob;
  prob=malloc (DIMENSION * DIMENSION*sizeof ( int* ));

   if ( prob == NULL ) {
    printf("No hay ubicación de memoria suficiente");
    exit(0);
   }


  for (i=0; i < DIMENSION * DIMENSION; i++ ){

    prob[i]=malloc(DIMENSION * DIMENSION * sizeof(int));

  if (prob[i] == NULL) {
    printf("No hay ubicaci&#243;n de memoria suficiente");
    exit(0);

  }
  }

Thanks for the advise... I didn't know about casting.