I’m pretty sure that’s not correct. You can have large collective ops and other operations in the callbacks. HDF5 and PETSc do all sorts of stuff in these callbacks.
— Pavan
On Mar 6, 2014, at 9:39 AM, Jim Dinan <james.dinan(a)gmail.com> wrote:
> I believe that the copy and free attribute callbacks are intended to be local functions that deal with objects the user has attached as attributes, that might need to be copied or freed. Can't find text that confirms this explicitly, though..
>
>
> On Wed, Mar 5, 2014 at 6:31 PM, Balaji, Pavan <balaji(a)anl.gov> wrote:
>
> The endpoints proposal says that Comm_free is not fully synchronizing. What happens if I have an attribute callback on that communicator which does a Bcast? How will that work?
>
> — Pavan
>
> _______________________________________________
> mpiwg-hybridpm mailing list
> mpiwg-hybridpm(a)lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
>
> _______________________________________________
> mpiwg-hybridpm mailing list
> mpiwg-hybridpm(a)lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
I believe that the copy and free attribute callbacks are intended to be
local functions that deal with objects the user has attached as attributes,
that might need to be copied or freed. Can't find text that confirms this
explicitly, though..
On Wed, Mar 5, 2014 at 6:31 PM, Balaji, Pavan <balaji(a)anl.gov> wrote:
>
> The endpoints proposal says that Comm_free is not fully synchronizing.
> What happens if I have an attribute callback on that communicator which
> does a Bcast? How will that work?
>
> -- Pavan
>
> _______________________________________________
> mpiwg-hybridpm mailing list
> mpiwg-hybridpm(a)lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
>
The endpoints proposal says that Comm_free is not fully synchronizing. What happens if I have an attribute callback on that communicator which does a Bcast? How will that work?
— Pavan
Eek, sorry, I made these changes locally and forgot to update the document
and upload to the ticket. Thanks for catching this.
~Jim.
On Tue, Mar 4, 2014 at 6:29 PM, Jeff Squyres (jsquyres)
<jsquyres(a)cisco.com>wrote:
> FYI -- these 2 problems still weren't fixed in the text shown today for
> #380.
>
> Also, page 244 line 25, you need ", ierror" at the end of the parameter
> list.
>
>
> On Feb 14, 2014, at 10:14 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com>
> wrote:
>
> > p244:29: s/Type/TYPE/
> > p244:33: s/OUT_COMM_HDLS/NEW_COMM_HANDLES/
> >
> > Steve, Jeff H., and I stayed on the webex and talked through a potential
> issue with the f08 binding, but we're all convinced that it's ok / you
> don't care. With the minor Type->TYPE fix, and the very important name fix
> (because in Fortran, parameters can be passed by name, and therefore the
> param names are normative), we're happy with the Fortran bindings for
> MPI_COMM_CREATE_ENDPOINTS.
> >
> > --
> > Jeff Squyres
> > jsquyres(a)cisco.com
> > For corporate legal information go to:
> http://www.cisco.com/web/about/doing_business/legal/cri/
> >
> > _______________________________________________
> > mpiwg-hybridpm mailing list
> > mpiwg-hybridpm(a)lists.mpi-forum.org
> > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
>
>
> --
> Jeff Squyres
> jsquyres(a)cisco.com
> For corporate legal information go to:
> http://www.cisco.com/web/about/doing_business/legal/cri/
>
> _______________________________________________
> mpiwg-hybridpm mailing list
> mpiwg-hybridpm(a)lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
>
FYI -- these 2 problems still weren't fixed in the text shown today for #380.
Also, page 244 line 25, you need ", ierror" at the end of the parameter list.
On Feb 14, 2014, at 10:14 AM, Jeff Squyres (jsquyres) <jsquyres(a)cisco.com> wrote:
> p244:29: s/Type/TYPE/
> p244:33: s/OUT_COMM_HDLS/NEW_COMM_HANDLES/
>
> Steve, Jeff H., and I stayed on the webex and talked through a potential issue with the f08 binding, but we're all convinced that it's ok / you don't care. With the minor Type->TYPE fix, and the very important name fix (because in Fortran, parameters can be passed by name, and therefore the param names are normative), we're happy with the Fortran bindings for MPI_COMM_CREATE_ENDPOINTS.
>
> --
> Jeff Squyres
> jsquyres(a)cisco.com
> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
>
> _______________________________________________
> mpiwg-hybridpm mailing list
> mpiwg-hybridpm(a)lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-hybridpm
--
Jeff Squyres
jsquyres(a)cisco.com
For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
