[Mpi-forum] MPI-3 One-Sided Communications

[Anthony Skjellum]

Can you point me at the proposed solutions? :-)

On Apr 23, 2009, at 8:07 PM, Vinod tipparaju wrote:

>
> First, as far as my understanding goes, no, that is not why we have  
> the active message working group.
>
> I will not comment on if the MPI-2 missed its mark. You are more  
> familiar with what went on that I am. In my opinion, it was a design  
> that approached the problem of defining one-sided data access in a  
> ubiquitous way on both CC and non-CC machines -- I wasn't  
> participating in the forum when these choices where made.
>
> Lower latency than two sided means many things. One of them is tag  
> matching ( as you know very well, this can have good and bad effects  
> based on the scenario). There are may others such as lower latency  
> single element updates and lesser than the currently necessary  
> synchronization for remote data access and update.
>
> All of these have been addressed with a proposed potential solution  
> in MPI forum RMA working group in _two_ different ways with-in the  
> last one year. If you think neither of these two approaches have  
> satisfied the "starting over" criterion, I would like to know, why?
>
> Bandwidth is an entirely different story.
>
> Vinod Tipparaju.
>
> > To: mpi-forum at lists.mpi-forum.org
> > From: tony at cis.uab.edu
> > Date: Fri, 24 Apr 2009 00:32:54 +0000
> > Subject: Re: [Mpi-forum] MPI-3 One-Sided Communications
> >
> > Hi, I was giving my opinion...
> >
> > The main requirements for one sided was achieve lower latency than  
> two sided (as well as one sidedness), and allow for potentially  
> lower bandwidth on long transfers as a side effect... What one sided  
> achieved is higher latency and higher bandwidth in typical (quality)  
> implementations. The api is consequently a mismatch to its original  
> purpose... It diverged from its design center.. Missed the mark...  
> Way too complex compared to simple put and get.
> >
> > Starting again from first principles with the goal of very low  
> latency remote put and get without the legacy of existing one sided  
> is my suggestion.
> >
> > In other words : Start over.
> >
> > Perhaps that is why now we have an active messages group?
> > Not sure.
> >
> >
> >
> > Tony
> > Anthony Skjellum, PhD
> > Professor and Chair
> > Dept. of Computer and Information Sciences
> > University of Alabama at Birmingham
> > +1-205-807-4968
> >
> > -----Original Message-----
> > From: Vinod tipparaju <tipparajuv at hotmail.com>
> >
> > Date: Thu, 23 Apr 2009 20:09:38
> > To: <mpi-forum at lists.mpi-forum.org>
> > Subject: Re: [Mpi-forum] MPI-3 One-Sided Communications
> >
> >
> > _______________________________________________
> > mpi-forum mailing list
> > mpi-forum at lists.mpi-forum.org
> > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
> >
> >
> > _______________________________________________
> > mpi-forum mailing list
> > mpi-forum at lists.mpi-forum.org
> > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
> _______________________________________________
> mpi-forum mailing list
> mpi-forum at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpi-forum/attachments/20090423/ce77c65e/attachment-0001.html>