My MPI code runs on many processors. What confuses me is that for the same
amout of job, different processors take very different CPU time to finish it,
sometimes the difference can be 3 times. Anybody knows what I can do to make
the CPU time pretty much the same?