Rajeev
From PNNL it was Sriram Krishnamoorthy.
Others, if you have up-coming system with unique cache coherence or memory access constrains that you think I should explicitly address, please write me a note.
Thanks,
Vinod.
> From: thakur@mcs.anl.gov
> To: mpi3-rma@lists.mpi-forum.org
> Date: Tue, 9 Dec 2008 11:24:13 -0600
> Subject: [Mpi3-rma] Notes from RMA working group concall, Dec 9, 2008
>
> Attendees:
> Rajeev, Darius, Dave Goodell, Argonne
> Bill Gropp, UIUC
> Rich, Vinod, ORNL
> Jesper, Hubert, NEC
> Robert Blackmore, Su Huang, IBM
> Howard Pritchard, Cray
> Christian Bell, Myricom
> Someone from PNNL (missed the name)
> I may have missed others
>
> * Discussion on issues with non-cache-coherent systems. Vinod is
> preparing a presentation for the next Forum mtg. He needs
> input from vendors who have (or will have) systems that are not
> cache coherent or have special memory features.
>
> * Vinod is updating the existing proposal and will send out a new
> version before the Forum mtg.
>
> * Vinod will check with Dan Bonachea if he can attend next mtg and his
> thoughts on current proposal.
>
> * Discussion on how does the target know of completion. Vinod will
> clarify further in new version of proposal and FAQ.
>
> * Long discussion on how to represent the target address. Currently it
> is directly provided as a (void *), which won't work
> everywhere. There was some consensus that it needs to be an object
> of some kind. Care must be taken about scalability issues (don't
> want a process to a hold a list of everyone's addresses).
>
> * Ticket 35 (datatype and Op registration) and 37 (examples describing
> synchronization semantics) were briefly discussed. Jesper will make
> a presentation on 35 at the Forum mtg. Rajeev said having examples
> like in #37 is a good thing; we need to make sure they are correct
> (he thinks they are).
>
> If I have missed anything, please add.
>
> Rajeev
>
> _______________________________________________
> mpi3-rma mailing list
> mpi3-rma@lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-rma