Question

Here is a serial program in C and parallel program in OpenMP that takes in a string as input and counts the number of occurrences of a character you choose.

Why is the runtime for the output for the OpenMP parallel program much longer?

Serial Program

#include <stdio.h>
#include <string.h>
#include <time.h>

int main(){
   char str[1000], ch;
   int i, frequency = 0;
   clock_t t;
struct timespec ts, ts2;
  

   printf("Enter a string: ");
   gets(str);

   int len = strlen(str);

   printf("Enter a character to find the frequency: ");
   scanf("%c",&ch);

   t = clock();
   timespec_get(&ts, TIME_UTC);

   for(i = 0; i < len; ++i)
   {
   if(ch == str[i])
   frequency += 1;
   }

   t = clock() - t;
   timespec_get(&ts2, TIME_UTC);
double time_taken = ((double)t)/CLOCKS_PER_SEC; // in seconds
   printf("Frequency of %c = %d\n", ch, frequency);
   printf("Time required = %ld \n",(ts2.tv_nsec - ts.tv_nsec));

   return 0;
}

Parallel Program using OpenMP directives

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

int main(){
   char str[1000], ch;
   int i, frequency = 0;

   printf("Enter a string: ");
   gets(str);

   int len = strlen(str);

   printf("Enter a character to find the frequency: ");
   scanf("%c",&ch);

   double start = omp_get_wtime();

   #pragma omp parallel for reduction(+ : frequency)
   for(i = 0; i < len; ++i)
   {
      if(ch == str[i])
   frequency += 1;
   }

   double end = omp_get_wtime();
   printf("Frequency of %c = %d\n", ch, frequency);
   printf("Time required = %lf \n", end - start);

   return 0;
}

Answer 1

I compared the time taken by the two methodologies used but it seems that the parallel approach is taking very less time compared to serial approach.Its only taking 0.000002 while the other one is taking 300

Serial

Enter a string: hello Enter a character to find the frequency: 1 requency of 1 2 Time required = 300

Parallel

Enter a string: hello Enter a character to find the frequency: 1 Frequency of 1 - 2 Time required-0.000002