[Mpi3-ft] reenable_anysource

Josh Hursey jjhursey at open-mpi.org
Mon Dec 12 07:38:49 CST 2011

Point-to-point operations over an intercommunicator go from the caller
(in the local group) to some process in the remote group. So isn't it
sufficient to say that the group returned by
MPI_Comm_reenable_any_source() using an intercommunicator is
representative of the failures in the remote group? We do a similar
trick for MPI_Comm_validate().

So the failures in the local group in relation to the point-to-point
operations for an intercommunicator do not matter when reenabling
ANY_SOURCE receives since two processes in the local group cannot use
that intercommunicator to communicate with one another. If the user
needs to know the failures in the local group they can use the
separate accessor function for it, MPI_Comm_group_failed().

Do you think that is sufficient functionality for the user?

Good point though, we need to clarify this in the text.

-- Josh

On Fri, Dec 9, 2011 at 5:32 PM, Sur, Sayantan <sayantan.sur at intel.com> wrote:
>> -----Original Message-----
>> From: mpi3-ft-bounces at lists.mpi-forum.org [mailto:mpi3-ft-
>> bounces at lists.mpi-forum.org] On Behalf Of Darius Buntinas
>> Sent: Friday, December 09, 2011 2:22 PM
>> To: MPI 3.0 Fault Tolerance and Dynamic Process Control working Group
>> Subject: [Mpi3-ft] reenable_anysource
>> We've proposed MPI_Comm_reenable_anysource that returns a group
>> containing the failed processes known to the process at the time of the
>> failure.
>> Don't we need another function for intercommunicators that returns two
>> groups of failed processes: one for the local group and another for the
>> remote group?
>> MPI_Comm_reenable_anysource_inter? (now we're definitely over 30
>> characters).
> Yes, but as long as it is < 140 characters, do we really care? I mean, we could still tweet about the function ;)
> (sorry, couldn't resist)
> How about we have the same name and have some arg ignored?
> MPI_Comm_reenable_anysource(comm, failed_local_group, failed_remote_group)
> With failed_remote_group only valid if comm is inter? Or does that violate some naming convention?
>> -d
>> _______________________________________________
>> mpi3-ft mailing list
>> mpi3-ft at lists.mpi-forum.org
>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft
> _______________________________________________
> mpi3-ft mailing list
> mpi3-ft at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-ft

Joshua Hursey
Postdoctoral Research Associate
Oak Ridge National Laboratory

More information about the mpiwg-ft mailing list