[mpiwg-hybridpm] Hybrid telecon fiasco

[Jim Dinan]

That is another option...  Consider a case with cooperatively scheduled
user-level threads.  It could be hard to meet the concurrency requirement
for collectives even if you have MPI_THREAD_MULTIPLE.


On Mon, Apr 7, 2014 at 2:09 PM, Jeff Hammond <jeff.science at gmail.com> wrote:

> Would it make it easier to just require MPI_THREAD_MULTIPLE here?  Is
> that not the only _reasonable_ use case?
>
> Sorry if I am bringing a dead horse back to life so we can beat it to
> death again.
>
> Jeff
>
> On Mon, Apr 7, 2014 at 1:07 PM, Jim Dinan <james.dinan at gmail.com> wrote:
> > Every process in a communicator has its own set of attributes.  Seems
> > natural that this would be extended to endpoints.  They don't propagate
> when
> > the endpoints communicator is created, so you would need to set them on
> each
> > endpoint.
> >
> >  ~Jim.
> >
> >
> > On Mon, Apr 7, 2014 at 1:27 PM, Jeff Squyres (jsquyres) <
> jsquyres at cisco.com>
> > wrote:
> >>
> >> More to the point -- did we decide that if you create_endpoints and get
> N
> >> local communicators out, are there N local copies of each of the
> attributes?
> >> Or just 1 copy?
> >>
> >> I.e., does the attribute copy callback get invoked N times or 1 time?
> >>
> >> This may be in the text already -- forgive me if it is -- but I should
> >> think that there should be a symmetry between attribute copy/creation
> and
> >> destruction with regards to endpoint-created communicator handles.
> >>
> >>
> >>
> >> On Apr 7, 2014, at 1:21 PM, "Balaji, Pavan" <balaji at anl.gov> wrote:
> >>
> >> >
> >> > On Apr 7, 2014, at 12:09 PM, Jim Dinan <james.dinan at gmail.com> wrote:
> >> >> Dan suggested a third approach where we leave the Comm_free semantic
> >> >> unchanged and add a new function, MPI_Comm_free_endpoints(MPI_Comm
> >> >> comm_handles[], int my_num_ep), that has symmetry with
> >> >> Comm_create_endpoints.  An endpoints communicator can be freed with
> either
> >> >> Comm_free or Comm_free_endpoints.  They differ in their concurrency
> >> >> requirement.
> >> >
> >> > What happens to the attribute callbacks when you use
> >> > MPI_Comm_free_endpoints?
> >> >
> >> >  -- Pavan
> >> >
> >> > _______________________________________________
> >> > mpiwg-hybridpm mailing list
> >> > mpiwg-hybridpm at lists.mpi-forum.org
> >> > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
> >>
> >>
> >> --
> >> Jeff Squyres
> >> jsquyres at cisco.com
> >> For corporate legal information go to:
> >> http://www.cisco.com/web/about/doing_business/legal/cri/
> >>
> >> _______________________________________________
> >> mpiwg-hybridpm mailing list
> >> mpiwg-hybridpm at lists.mpi-forum.org
> >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
> >
> >
> >
> > _______________________________________________
> > mpiwg-hybridpm mailing list
> > mpiwg-hybridpm at lists.mpi-forum.org
> > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
>
>
>
> --
> Jeff Hammond
> jeff.science at gmail.com
> _______________________________________________
> mpiwg-hybridpm mailing list
> mpiwg-hybridpm at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-hybridpm/attachments/20140407/55beb61b/attachment-0001.html>