Detailed Differences in Simulation Results

For numerical simulations, if you run the same case twice using the exact same setup you should get the exact same results. If an element in your setup changes, you might get detailed differences in your result. This difference should not affect the overall convergence behavior. These elements in the setup include:

• Different number of processes
• Different process topology (so say you run on 4 nodes with 64 processes each or 8 nodes with 32 processes each, even though that is 256 processes in both cases, it is a different topology)
• Different CPUs (more precisely, different instruction sets)
• Different Operating Systems