No documentation yet. If you want to understand how it works and what
exactly "highly optimized" means please look for the collectives
papers on this page (http://www.netlib.org/utk/people/JackDongarra/
papers.htm). In few words, we have multiples algorithms and we tune
them based on the network characteristics.
We have a home made tools that help us with that. It's far from
complete, but it's enough for what we're doing right now. If you
want, I can give you the source code and the documentation and you
can play around with it.
Thanks,
george.
On Oct 27, 2006, at 2:56 PM, Tony Ladd wrote:
George
Thanks for the info. When you say "highly optimized" do you
algorithmically,
tuning, or both? In particular I wonder if OMPI optimized
collectives use
the divide and conquer strategy to maximize network bandwidth.
Sorry to be dense but I could not find documanetation on how to
access the
optimized collectives. I would be willing to fiddle with the
parameters a
bit by hand if I had some guidance as to how to set things and what
I might
vary.
The optimization I was talking about (divide and conquer) would
work better
than the basic strategy regardless of network; only message size
might have
some effect.
Thanks
Tony
-------------------------------
Tony Ladd
Chemical Engineering
University of Florida
PO Box 116005
Gainesville, FL 32611-6005
Tel: 352-392-6509
FAX: 352-392-9513
Email: tl...@che.ufl.edu
Web: http://ladd.che.ufl.edu
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
