Dear DuMuX experts,

Working with https://git.iws.uni-stuttgart.de/andrian/rate-sensitivity using 
direct linear solver UMFPackBackend, I observed almost a 10-fold degradation in 
performance of a parallel version with Open MPI as compared to a serial version!

On a 40 cores Ubuntu desktop 1 time step with 6 Newton iterations took 27 sec 
wall time with a serial version vs. 224 sec with a parallel version..

Your feedback is greatly appreciated.

Many thanks,
Nikolai

_______________________________________________
Dumux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux

Reply via email to