Thread: Cygwin memory allocation / stackdumpfile

I have my model in C which runs an evolutionary process, but once I input too large a population or too long an evolutionary period, it fails. For example, I can evolve a (constant) population of 20 individuals over 50 years in 14 seconds, but 30 individuals won't run. The error message is cygwin_exceptionpen_stackdumpfile: Dumping stack ... , which I believe is to do with memory allocation, but I tried using pflags to assign up to 4GB and it still doesn't run.

The code is fairly long, but a simplified version is below (i know the functions all run as intended)

Code:

int main(){


//structs initialised, controls like population size loaded
//output file opened


// initialise with artificial 'chromosome' using random number generation for each of N 'genes'
Initialise(N, ctrl.Population, &chrom);



            for (int Generation = 1; Generation <= ctrl.Generations; Generation++){


                    for (int Individual = 1; Individual <= ctrl.Population; Individual++){


                        //calls function to assess fitness


                    } //END OF INDIVIDUAL LOOP


                    if(Generation < ctrl.Generations){
                            Selection(Generation, ctrl.Population, ctrl.MutationRate, N, &chrom);
                    }


            } // END OF GENERATION LOOP
 
    printf("Finished sim!");


} /// END OF MAIN FUNCTION


void Initialise(int N, int Population, chrom_struct *chrom)
{
    double rand_double(); /* 0.0 to 1.0, unif. dist. */


    chrom->Genes  = (double **)malloc(Population * sizeof(double *));
    for(int i = 0; i < N; i++){
        chrom->Genes[i] = (double *)malloc(N * sizeof(double));
    }
    chrom->Fitness = (double *)malloc(Population * sizeof(double));


    struct timeval tm;
    gettimeofday(&tm, NULL);
    srandom(tm.tv_sec + tm.tv_usec * 1000000ul);


    for(int Individual = 1; Individual <= Population; Individual++){
        chrom->Genes[Individual][0] = 0;
        for(int i = 1; i<= N; i++){
            chrom->Genes[Individual][i] = rand_double();
        }
    }


} // END OF INITIALISE


double rand_double(){
    struct timeval tm;
    gettimeofday(&tm, NULL);
    srandom(tm.tv_sec + tm.tv_usec * 1000000ul);
    return random()/(double)RAND_MAX;
}


void Selection(int Generation, int Population, double MutationRate, int N, chrom_struct *chrom)
{
    // copies current generation into parental array
    double **Parents;
    Parents = (double **)malloc(Population * sizeof(double *));
    for(int i = 0; i < soil->N; i++){
        Parents[i] = (double *)malloc(N * sizeof(double));
    }


    for(int Individual = 1; Individual <= Population; Individual++){
        for(int i = 1; i<= N; i++){
            Parents[Individual][i] = chrom->Genes[Individual][i];
        }
    }


    // ranks individuals  in current genes array depending on fitness


    // copies fittest individual over to parental array directly


    // for rest of pop, fills parental array from top half of individuals

    // overwrites genes array using randomly selected parents, recombines, mutates


} // END OF SELECTION FUNCTION

I'm trying to write a makefile etc to run it on the university's supercomputing cluster, but it's slow progress since I've never done it before. I was wondering if there's a way to get it to run on my laptop? Even 100 individuals x 100 generations would be fine.

Thanks for your help!

Originally Posted by john.c

No you don't. You think they all run as intended.

For example, your rand_double function seems to be incorrect.
You should only seed the rng once in the entire run of the program.
So you should move the seeding to main.

Thanks - I've moved that out, tidies it up a bit.

Hi, thanks for replying. I overwrite those arrays continuously - it should work so that the current chromosome array is copied to the parents, then filled up, and after the fitness function is called, it's copied to parents again - so I don't think that's the issue. However, should I be freeing them at the end of the program? I don't know how long 'due course' means, don't want my computer storing a load of unnecessary info!

Thank you!!! Oh my goodness this worked! I'm used to working in VBA, so if I declare an array of size 10, it automatically creates an array of size 11, from 0-10. Technically I guess this makes more sense...