[mpiwg-rma] MPI RMA nit: MPI_Win_create_keyval
thakur at mcs.anl.gov
Fri Mar 4 11:16:07 CST 2016
And for future extensibility — if MPI_Win_dup ever got added to the standard, we won’t need to go back and fix MPI_Win_create_keyval.
> On Mar 4, 2016, at 11:04 AM, William Gropp <wgropp at illinois.edu> wrote:
> As Jeff H suggests, it was for symmetry with the other attributes. There is no way to invoked it in the current standard.
> William Gropp
> Director, Parallel Computing Institute
> Thomas M. Siebel Chair in Computer Science
> Chief Scientist, NCSA
> University of Illinois Urbana-Champaign
> On Mar 4, 2016, at 10:33 AM, Jeff Hammond <jeff.science at gmail.com> wrote:
>> That's before my time. My guess is that it made the interface symmetric wrt other objects.
>> Adding WG list.
>> Sent from my iPhone
>>> On Mar 4, 2016, at 8:07 AM, Jeff Squyres (jsquyres) <jsquyres at cisco.com> wrote:
>>> Jim / JeffH --
>>> Any idea why MPI_WIN_CREATE_KEYVAL includes a win_copy_attr_fn parameter? When can that callback ever get invoked?
>>> Jeff Squyres
>>> jsquyres at cisco.com
>>> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
>> mpiwg-rma mailing list
>> mpiwg-rma at lists.mpi-forum.org
> mpiwg-rma mailing list
> mpiwg-rma at lists.mpi-forum.org
More information about the mpiwg-rma