I'm testing the time my program spends in parallel and scalar regions by using omp_get_wtime and I made a suprising observation.
When I use only one core, the program spends about 10 seconds in the parallel region and about 1 second in the scalar region.
When I use 4 cores, the parallel region scales to 3 seconds, but the time spend in the scalar region grows up to about 5 seconds.
What could that be? Has someone an advice for me?
I use IVF14 with Windows on an i7-Board.