openmp on for loop

Question

I have the following code:

for(i=0; i<num1; i++)
{
 j=i%num2;
 a[j]=do_compute(j);
}

Let's assume num2 << num1. If I wanted to use omp for, is there any need to use j as private? ie as follows:

#pragma omp parallel private(i,j)
{
 #pragma omp for
 for(i=0; i<num1; i++)
 {
  j=i%num2;
  a[j]=do_compute(j);
 }
}

or would this suffice:

#pragma omp parallel
{
 #pragma omp for
 for(i=0; i<num1; i++)
 {
  j=i%num2;
  a[j]=do_compute(j);
 }
}

Thanks.

Answer 1

If j is only used in the scope of the loop, I'd do the following:

[...]
unsigned int i;
#pragma omp parallel for
for(i = 0; i < num1; ++i)
{
    unsigned int j = i % num2;
    a[j] = do_compute(j);
}
[...]

not trying to be picky here, but given that do_compute is deterministic and has no side-effects, you'd be better off with just using

[...]
unsigned int i;
#pragma omp parallel for
for(i = 0; i < num2; ++i)
    a[i] = do_compute(i);
[...]