Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!emory!hubcap!slim
From: slim@cis.ohio-state.edu (Scott Whitman)
Newsgroups: comp.parallel
Subject: Using the Butterfly Parallel Processor
Message-ID: <7426@hubcap.clemson.edu>
Date: 14 Dec 89 15:45:57 GMT
Sender: fpst@hubcap.clemson.edu
Lines: 30
Approved: parallel@hubcap.clemson.edu

Howdy folks,
	
	I am in the process of optimizing my program for running on
the BBN Butterfly GP1000.  I was wondering if anyone has any
suggestions for looking at how to tune the program for good speedup.
I had been working on a small Butterfly (9 nodes) and got reasonable
speedup.  When I finally moved to a larger machine, I noticed that
at 24 nodes, the performance was worse than with 8.  Needless to say,
this is not a desirable affect.

	My first thought was memory contention, but I have scattered
most of the data and gotten rid of any unnecessary data structures.
My feeling is that this still may be the problem, however, I am at a
loss at the moment to figure out where the problem lies.  I do have
the TotalView debugger and gist, but these haven't helped much.  The
program looks like it is evenly load balanced, even on 24 nodes, but
the performance obviously indicates that there is a bottleneck
somewhere.

	If anyone has any suggestions, I'd appreciate it.  Please mail
me at:  slim@cosimo.osgp.osc.edu.

		Thanks,

			-Scott Whitman

--
 Scott Whitman, Department of Computer and Information Science
 slim@cis.ohio-state.edu    or   ...!{pyramid,killer}!osu-cis!slim
The Ohio State University; 2036 Neil Ave. Columbus OH USA 43210-1277