CODEWITHSUNDEEP
copenmp
Romero Azzalini
Romero Azzalini

Reputation: 304

OpenMp C Fileoutput

I remembered to have some trouble with OpenMp and writing into files in parallel, therefore I wrote a simple testprogram that creates files for every for iteration and counts down to zero in the file. Surprisingly it seems to work! If I open the taskmanager with top, I see always just one cpu with a load around 90% and the others around 1%-5%. Is it because the task is not cpu heavy or is there something going wrong (sometimes 2-3 cpus go up to 40%)?

The Testprogram

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

void print2file(char filename[],int i){
  FILE *fp;
  int j;
  fp=fopen(strcat(filename,".txt"),"w");
  for(j=i;j>=0;j--)
    fprintf(fp,"%d\n",j);
  fclose(fp);
}

int main(){
  int i;
  char test[12];
  #pragma omp for
  for(i=0;i<1000000000;i++){
    sprintf(test,"%d",i);
    //printf("test %d",i);
    print2file(test,i);
  }
}

Upvotes: 0

Views: 64

Answers (1)

Daniel Langr
Daniel Langr

Reputation: 23497

#pragma omp for need to be in a parallel region. Either use #pragma omp parallel followed by #pragma omp for inside, or a shortcut #pragma omp parallel for.

Note that you need i and test to be thread-local variables, not shared ones, to prevent data race. There are many ways how to achieve this, such as declaring these variables inside a parallel region, make them private/firstprivate in an OpenMP clause, make them threadprivate, etc.

Upvotes: 2

Related Questions