The Computer Science/Computational Science Interface

                                Faisal Saied

                      Department of Computer Science
                 University of Illinois at Urbana-Champaign


In this talk, we take a case study approach and examine the ways in
which the emerging discipline of computational science and computer
science can interact in a mutually beneficial way.

Our two cases come from Bio-Physics and Groundwater Modeling. In each
case, the basic scientific model is expressed in terms of partial
differential equations, that are discretized and lead to large scale
computational problems.  We show how a multiscale algorithm, namely
the multigrid method, can be a very effective solver for both these
applications. The high resolution demanded by by these models makes
parallel computing a very important component of the modeling
process. We describe parallel implementations of our multigrid solvers
and their performance on a variety of parallel architectures.

Finally we discuss how parallel software tools are being used in
current computational science research, where further progress is
needed, and the computer science challenges implied by these needs.

---------------------------------------------------------------------------
Biographical sketch:

B.A. Trinity College, Cambridge University
Diplom Mathematik, Goettingen University
Ph.D. in Computer Science, Yale University (advisor Martin Schultz)
Assistant Professor in Computer Science, University of Illinois at
                                                       Urban-Champaign

Research interests: numerical analysis, multigrid methods, parallel
computing, computational problems in Bio-Physics, groundwater modeling

---------------------------------------------------------------------------