Anyone test on different MPI implementations?

Report all good and bad behavior here
Post Reply
User avatar
gavin.ridley.utk
Posts: 95
Joined: Wed Mar 16, 2016 7:07 am
Security question 1: No
Security question 2: 40

Anyone test on different MPI implementations?

Post by gavin.ridley.utk » Wed Sep 26, 2018 7:50 pm

Hi all,

I've read that different MPI implementations can have different performance depending on the application, e.g., https://pdfs.semanticscholar.org/d851/6 ... cd7971.pdf.

Has anyone compared open MPI, MPICH, MVAPICH, etc? It'd be nice to know which is best to pick when installing on HPC platforms.

Thanks,
Gavin Ridley

Yellowstone Energy

Ville Valtavirta
Posts: 457
Joined: Fri Sep 07, 2012 1:43 pm
Security question 1: No
Security question 2: 92

Re: Anyone test on different MPI implementations?

Post by Ville Valtavirta » Mon Oct 01, 2018 1:51 pm

There have been some tests by users but this may be platform dependent.

Some combinations of the MPI-implementation and OpenMP library (and compiler?) will also lead to random crashes meaning that it is probably best to run a suite of tests using different combinations when first starting to use Serpent in a new HPC-environment.

-Ville

Ville Valtavirta
Posts: 457
Joined: Fri Sep 07, 2012 1:43 pm
Security question 1: No
Security question 2: 92

Re: Anyone test on different MPI implementations?

Post by Ville Valtavirta » Mon Oct 01, 2018 1:52 pm


User avatar
gavin.ridley.utk
Posts: 95
Joined: Wed Mar 16, 2016 7:07 am
Security question 1: No
Security question 2: 40

Re: Anyone test on different MPI implementations?

Post by gavin.ridley.utk » Mon Oct 15, 2018 4:22 pm

Thank you Dr. Valtavirta!
Gavin Ridley

Yellowstone Energy

Post Reply