While working on the implementation of the MPI version of the STREAM benchmark, I realized that there were some subtleties in timing that could easily lead to inaccurate and/or misleading results. This post is a transcription of my notes as I looked at the issues…. Primary requirement: I want a… read more