[Mpi-comments] One comment on MPI-3.0 Draft 2, August 2012

[Richard Graham]

Sébastien,
  To get involved, see meetings.mpi-forum.org, where there are mailing lists you can join, as well as meeting details.  The most effective way to participate it to get involved in person.

Rich

-----Original Message-----
From: mpi-comments-bounces at lists.mpi-forum.org [mailto:mpi-comments-bounces at lists.mpi-forum.org] On Behalf Of Sébastien Boisvert
Sent: Wednesday, August 15, 2012 11:20 AM
To: mpi-comments at lists.mpi-forum.org
Subject: Re: [Mpi-comments] One comment on MPI-3.0 Draft 2, August 2012

Hello Richard,

I really agree that this is a big change. And I am aware that there is no fairness by design [1].

In Open-MPI, George Bosilca provided a patch [2], which I modified [3] for a round-robin policy and tested [4].

For threading, there is MPI_Init_thread that tells MPI if one or more thread will be spawned and if one or more will make MPI calls.

Maybe there could be something like MPI_Init_fairness which could either be MPI_FAIRNESS_NO (the current behavior) or MPI_FAIRNESS_YES (which would do round-robin reception).


How can I get involved ?


[1] http://www.mcs.anl.gov/research/projects/mpi/tutorial/gropp/node92.html

[2]
http://www.open-mpi.org/community/lists/devel/att-9769/iprobe_starvation.patch

[3]
http://www.open-mpi.org/community/lists/devel/att-9770/iprobe_starvation_fixed.patch

[4] http://www.open-mpi.org/community/lists/devel/2011/09/9770.php

Richard Graham a écrit :
> Sébastien,
>    Thanks you for sending in this request, this is truly appreciated.
>    The propose change is a large change to the standard, even though it may not seem so at a first glance.   Today the standard is explicit about lack of fairness, and we would have to think through in detail on the broad impact of such a change.  So, this does not the sort of item we can address for MPI 3.0.
>    However, if you would like to push a change like this into MPI, this can be addressed for the following version of the standard.  What I would suggest, if this is something that you really think belongs in the standard, that you either get involved directly in the process (the standardization effort is open to all), or find someone the is involved to actively push such a request.  As I said, this is a non-trivial change, so that both API changes (if any) and implementation issues need to be addressed.  Please let me know if you need help with figuring out how to get involved.
>
> Thanks again,
> Rich
>
> -----Original Message-----
> From: Sébastien Boisvert [mailto:sebastien.boisvert.3 at ulaval.ca]
> Sent: Tuesday, August 14, 2012 9:47 AM
> To: mpi-comments at mpi-forum.org
> Subject: [Mpi-comments] One comment on MPI-3.0 Draft 2, August 2012
>
> Dear MPI Forum committee members,
>
> I would like to submit a comment on the MPI-3.0 Draft 2, August 2012 for your consideration.
>
> Version: MPI-3.0 Draft 2, August 2012.
>
> The URL of the version of the MPI standard:
> http://meetings.mpi-forum.org/draft_standard/mpi3.0_draft_2.pdf
>
> Page: 65
>
> Line number: 28
>
> Section: 3.8.1
>
> In:
>
> 3. Point-to-Point Communication
>      3.8 Probe and Cancel
>            3.8.1 Probe
>
> Comment:
>
> It says that the source argument of MPI_Iprobe can be MPI_ANY_SOURCE, but it does say anything about fairness. Therefore MPI_ANY_SOURCE can lead to resource starvation.
>
> I think it would be better if probing would be done in a round-robin fashion when the source is MPI_ANY_SOURCE so that any MPI rank has an equal chance of having its message probed and received.
>
> Presently, the MPI standard contains nothing about which source should be probed when MPI_ANY_SOURCE is provided.
>
> I hope you will consider my comment.
>
>
> Sincerely,
>
>
> Sébastien Boisvert
> PhD student
> Université Laval
>
> _______________________________________________
> mpi-comments mailing list
> mpi-comments at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments
>
> _______________________________________________
> mpi-comments mailing list
> mpi-comments at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments

_______________________________________________
mpi-comments mailing list
mpi-comments at lists.mpi-forum.org
http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-comments