Critical construct

Synchronization is the eternal theme in parallel programming. Check the following code:

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

int main(void)
{    
    int sum = 0;

    #pragma omp parallel for
    for (int index = 1; index <= 10; index++)
    {
        sum += index;
    }

    printf("Sum is %d\n", sum);
    return 0;
}

Compile and run it for several times on my 24-core machine:

# gcc -fopenmp parallel.c
# ./a.out
Sum is 46
# ./a.out
Sum is 48
# ./a.out
Sum is 55

We can see the program didn't output correct result (55) every time. The root cause is multiple threads access the shared variable,sum, simultaneously, so we need the synchronization mechanism to protect it.

The critical construct is used to produce a critical section:

#pragma omp critical [(name) [hint(hint-expression)] ] new-line
    structured-block

Which means there is only 1 thread executing code in section every time. Add "#pragma omp critical" in the above code:

......
#pragma omp parallel for
for (int index = 1; index <= 10; index++)
{
    #pragma omp critical
    sum += index;
}
......

This time the result is always right:

# gcc -fopenmp parallel.c
# ./a.out
Sum is 55
# ./a.out
Sum is 55
# ./a.out
Sum is 55

Please note that the critical section makes threads run actually in sequence, and this will downgrade the power of parallelism, so please pay attention to this caveat.

BTW, the critical construct can have optional name and hint. The name is used to identify the critical construct, and all critical constructs without a name are considered to have the same unspecified name. For hint, I will explain it in following chapters.