That's one of the proposed changes. On 08/26/2013 01:06 PM, Wesley Bland wrote:

Am I mistaken in remembering that MPI doesn't allow an application to go in and out of MPI:

INIT .... FINALIZE .... INIT .... FINALIZE

I've looked at the standard and that's what Section 8.7 says to me. So am I correct in assuming that this will change that text to say that you can start, stop, and start new "MPI epochs"?

Thanks, Wesley

On Mon, Aug 26, 2013 at 12:10 PM, Jeff Squyres (jsquyres) <[email protected] <mailto:[email protected]>> wrote:

The hybrid working group will be hosting a plenary on Friday afternoon in Madrid.

We will be presenting a proposal on MPI_INIT/MPI_FINALIZE regarding interaction with threads (going back to the use cases Jeff Hammond raised at a prior Forum meeting), and nested MPI_INIT/MPI_FINALIZE. This is not a formal reading; we're requesting Forum feedback on the overall ideas.

The slides are attached. See you in Madrid.

-- Jeff Squyres [email protected] <mailto:[email protected]> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/

_______________________________________________ mpi-forum mailing list [email protected] <mailto:[email protected]> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

_______________________________________________ mpi-forum mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

-- Pavan Balaji http://www.mcs.anl.gov/~balaji