[Mpi3-rma] MPI_NO_OP and origin buffer

Jim Dinan dinan at mcs.anl.gov
Fri Sep 7 17:34:52 CDT 2012


MPI_Accumulate with MPI_REPLACE accomplishes a put.  NO_OP is not 
permitted in MPI_Accumulate.

On 9/7/12 5:23 PM, Torsten Hoefler wrote:
> All,
>
> So it may be fine for some accumulate ops but not for all because
> MPI_Accumulate with MPI_NO_OP was our back-door to get an atomic put.
> Thanks to Keith for pointing this out!
>
> Thanks,
>    Torsten
>
> On Fri, Sep 07, 2012 at 04:40:54PM -0500, Torsten Hoefler wrote:
>> Hi Jim,
>>
>> I agree, that is a reasonable change.
>>
>> Best,
>>    Torsten
>>
>> On Thu, Sep 06, 2012 at 10:48:47AM -0500, James Dinan wrote:
>>> Hi All,
>>>
>>> It looks like we forgot to specify that for MPI_NO_OP in the new
>>> accumulate ops, the origin buffer is ignored (i.e. it's acceptable to
>>> pass NULL).  We can add this text on page 441, line 30:
>>>
>>> When MPI_NO_OP is specified, the origin buffer is ignored.
>>>
>>> We already have such a use case in the linked list example.
>>>
>>>   ~Jim.
>>> _______________________________________________
>>> mpi3-rma mailing list
>>> mpi3-rma at lists.mpi-forum.org
>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-rma
>>>
>>
>> --
>> ### qreharg rug ebs fv crryF ------------- http://www.unixer.de/ -----
>> Torsten Hoefler           | Assistant Professor
>> Dept. of Computer Science | ETH Zürich
>> Universitätsstrasse 6     | Zurich-8092, Switzerland
>> CAB E 64.1                | Phone: +41 76 309 79 29
>> _______________________________________________
>> mpi3-rma mailing list
>> mpi3-rma at lists.mpi-forum.org
>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-rma
>>
>



More information about the mpiwg-rma mailing list