A set of sparse matrix kernels for shared memory and message passing
systems, called Spark98, are now available free of charge from

	www.cs.cmu.edu/~quake/spark98.html

The Spark98 kernels perform the core sparse matrix-vector product
operation from a set of finite element simulations of earthquake
induced ground motion in the San Fernando Valley (see
www.cs.cmu.edu/~quake).  Simple yet realistic kernels of this kind
that exhibit highly irregular memory reference patterns have been
difficult to come by in the past. If you build memory systems, SMPs,
communication libraries, thread packages, DSM systems, or
parallelizing compilers, I expect you will find the Spark98 kernels to
be very helpful in understanding the behavior of irregular codes on
your system.

-- David R. O'Hallaron, Research Computer Scientist
-- School of Computer Science, Carnegie Mellon Univ, 
-- 5000 Forbes Avenue, Pittsburgh, PA 15213-3891

-- phone: (412)268-8199	email: droh@cs.cmu.edu
-- fax  : (412)268-5576	web  : http://www.cs.cmu.edu/~droh/