[mpich-discuss] MPI_Allgatherv complexity

[Pavan Balaji]

On 11/02/2009 06:21 PM, Justin Luitjens wrote:
> Could you point me to a place that documents the change in algorithm so
> I can verify with NICS and Cray if Kraken has this updated algorithm?

The algorithm is described here:
http://www.mcs.anl.gov/~balaji/pubs/2008/europvm/europvm08.allgatherv.pdf

The actual code is present here:
http://trac.mcs.anl.gov/projects/mpich2/browser/mpich2/trunk/src/mpi/coll/allgatherv.c?rev=5355#L698

 -- Pavan

-- 
Pavan Balaji
http://www.mcs.anl.gov/~balaji