Open vatai opened 5 years ago
We should make something like a "benchmark" or "performance measurement framework". But first we should plan it properly. The goal would probably be finding the point where MPI is faster (e.g. the matrix/vector doesn't fit into memory).
Is this done? If so can we close the issue?
Nope. If you can, try recreating the OpenMP experiment on the many core system. To have the MPI version measured, I think we need to finish #16 first.
We should make something like a "benchmark" or "performance measurement framework". But first we should plan it properly. The goal would probably be finding the point where MPI is faster (e.g. the matrix/vector doesn't fit into memory).