yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

Hi,

I've been trying to parallelize some loops with openMP in some code used 
by FlowEngine (independant compilation out of Yade). I could get 
no-so-bad speedup with little efforts. Now I'm wondering how to get even 
better speedup and I have a few general questions.
- Is there already a mechanism in Yade to run more than one engine in 
parallel? For instance, flow engine and interactionsDispatcher are doing 
two independant things and could be run at the same time without 
interfering. It implies nested omp directives.
- If no, is that possible?
- Is there any limitation in the use of MPI and OpenMP in the same 
program (mostly thinking to ScaLapack,MUMPS and others, using MPI for 
linear aglebra)?

My impression is no/yes/no. Could you confirm?

Thanks.

Bruno