C – Parallel for loop in OpenMP

Prerequisites:

1. OpenMP | Introduction with Installation Guide
2. OpenMP Hello World Program

We have the for loops in the programs and we can parallelize them easily with the help of OpenMP Library. Add #include<omp.h>, as all the prototypes for these parallelization techniques of OpenMP lie in that header file.

Syntax:

#pragma omp parallel for [clause[[,] clause] ...] new-line
for-loop

Explanation:

• The Threads split the range of iterations evenly and execute in parallel.
• For Example,
  • There are 4 threads
  • The for loop goes from 1 to 1000. Total 1000 iterations.
  • So, thread 0 executes 1 to 250
  • Thread 1 executes 251 to 500
  • Thread 2 executes 501 to 750
  • Thread 3 executes 751 to 1000
• So in this way, the time taken would reduce as the work is now not done in sequential order. Rather it is executed parallelly.
• This is just an example for visualization, the actual range may vary a bit, but at the most, it divides evenly.

Algorithm:

1.  Start the program
2. There are many for loops in the program.
3. Add the for loop construct before all the for loops.
4. num_threads( n ) needs to be mentioned to get n threads
5. If not mentioned, by default, the no. of processor’s scores threads are formed.
6. So therefore parallelized.

Example:

C

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

int main()
{

    #pragma omp parallel for num_threads(4)
    for (int i = 1; i <= 10; i++) {
        int tid = omp_get_thread_num();
        printf("The thread %d  executes i = %d\n", tid, i);
    }

    return 0;
}

Output:

Output

• Here are 10 iterations, and 4 threads. Each thread has 2/3 iterations with it.
• Another proof that parallelization has happened is the order of iteration:
  • Here in the sequential or serial program, we have iterations i=1 to i=10 in the order one after another.
  • But here you can see that the i=4 comes first and i=1 at 3rd last place.
  • As in the above example, thread 1 finished before thread 0

Time Complexity: O(N)

Space Complexity: O(1)