> The problem I have is wanting to test different versions of the > same simulation for performance. All the data is computed and > the output is identical.
How about profiling the code and using the call counts? I think this would give a lot of useful information.