Confused on which variables should be shared or private in nested OpenMP loop

Question

I have a nested loop and im confused on which variables should be shared or private. I believe that i and j should be private but im confused on whether other variables also need to be shared or privated in my code:

    y = p->yMax;
    #pragma omp parallel for private(i,j) 
    for (i = 0; i < p->height; i++) {
        x = p->xMin;
        for (j = 0; j < p->width; j++) {
            p->carray[i * p->width + j] = x + y * I;
            x += p->step;
        }
        y -= p->step;
    }
    
    for (i = 0; i < p->maxIter; i++) {
        p->histogram[i] = 0;
    }
}

Cheers

Answer 1

1. Assuming that you do not need the values of i and j after the loops and observing that you reset i and j to zero at the beginning of the loops, I would simply define the i and j within the loop. If so, you don't have to state the behaviour of i and j when entering and exiting the loop (because it is automatically private).
2. I think that the value of x before entering the loop is not needed as well, so you can also go for private here or define x as a local varieble.
3. If y is treated similar as x and has to be reset to p->yMax in every step, you can either declare y as private and move into the loop, same as x. Or you have to declare y as firstprivate such that the previous value is not lost when entering the parallel region. Using shared is not an option for both x and y since you += and -= change the value of these variables.
4. p should be shared, as you need the values of p->xMin, p->step and p->yMax and you probably also want to use the values of p->carray after exiting the loop.

``

#pragma omp parallel for shared(p)  
    for (size_t i = 0; i < p->height; i++) {
        auto x = p->xMin;
        auto y = p->yMax;
        for (size_t j = 0; j < p->width; j++) {
            p->carray[i * p->width + j] = x + y * I;
            x += p->step;
        }
        y -= p->step;
    }


//This one isn't in the parallel region
    for (size_t i = 0; i < p->maxIter; i++) {
        p->histogram[i] = 0;
    }