Parallelizing for loop with unknown increment condition using OpenMP

Question

I am finding all primes using the Sieve of Eratosthenes algorithm. I attempted to parallelize this algorithm. However, speedup stops increasing after two threads!