[mpich-discuss] Asynchronous All_Reduce

[Jeff Hammond]

Nonblocking collectives were voted into MPI-3.  I believe that the
latest version of MPICH2 has MPIX_I(all)reduce, which will become
MPI_I(all)reduce once the MPI-3 standard is ratified.

If you are on a machine for which installing the latest MPICH2 is
imprudent (e.g. a supercomputer), then consider Torsten Hoeffler's NBC
(Non-Blocking Collectives) library, which sits on top of MPI and was
the implementation prototype for the MPI-3 standard.

Generalized requests are one way to implement nonblocking operations
that are not provided by MPI, but you may find this challenging.

Jeff

On Wed, Mar 28, 2012 at 5:09 PM, Bharath Pattabiraman
<bharath650 at gmail.com> wrote:
> Hi,
>
> I am wondering if there is an asynchronouus way to do Reduce/Allreduce operations, and for other collective communication calls too.
>
> Regards,
> Bharath
> _______________________________________________
> mpich-discuss mailing list     mpich-discuss at mcs.anl.gov
> To manage subscription options or unsubscribe:
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss



-- 
Jeff Hammond
Argonne Leadership Computing Facility
University of Chicago Computation Institute
jhammond at alcf.anl.gov / (630) 252-5381
http://www.linkedin.com/in/jeffhammond
https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond (in-progress)
https://wiki.alcf.anl.gov/old/index.php/User:Jhammond (deprecated)
https://wiki-old.alcf.anl.gov/index.php/User:Jhammond(deprecated)