[Mpi3-hybridpm] Meeting notes

[Rajeev Thakur]

There was a lot of discussion on what is the Advice to Implementors on
what is a sensible number for max_endpoints to be returned by the
implementation. Is it related to the number of cores (and hence
potential number of threads) or the number of (virtual) network
interfaces. If the latter, what exactly is a virtual network interface.
Is each socket a virtual NI?

Someone asked, if you have nodes with 1024 cores connected by some
network, would you want to set max_endpoints to 8 or 128 or 1024 or
65536 or something else?

Rajeev


> -----Original Message-----
> From: mpi3-hybridpm-bounces at lists.mpi-forum.org 
> [mailto:mpi3-hybridpm-bounces at lists.mpi-forum.org] On Behalf 
> Of Pavan Balaji
> Sent: Wednesday, March 10, 2010 12:03 PM
> To: mpi3-hybridpm at lists.mpi-forum.org
> Subject: [Mpi3-hybridpm] Meeting notes
> 
> Folks,
> 
> We had some discussion on Marc's proposal at the Forum. Here are the
> minutes:
> https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Hybrid/n
> otes-2010-03-08
> 
> Overall reaction was that it was getting very messy, and the 
> original elegance it had is getting lost in the corner cases. 
> We need to work on that.
> 
> We'll have the next telecon on 03/16. We have a bunch of 
> things to cleanup for this.
> 
>  -- Pavan
> 
> --
> Pavan Balaji
> http://www.mcs.anl.gov/~balaji
> _______________________________________________
> Mpi3-hybridpm mailing list
> Mpi3-hybridpm at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm
>