OpenMP C++ with Nested For Loops slower than serial

Question

I\'m trying to parallelize the following code:

    int i,j,k;
    float sum;

    for (i=0; i