Professor David E. Keyes

Ph.D., Applied Mathematics, Harvard 1984

David E. Keyes is the Richard F. Barry Professor of Mathematics & Statistics and Adjunct Professor of Computer Science at Old Dominion University and Acting Director of the Institute for Scientific Computing Research (ISCR) at the Lawrence Livermore National Laboratory. Keyes has accepted an offer to move to Columbia University's Department of Applied Physics and Applied Mathematics and assume the Fu Foundation Chair of Applied Mathematics, effective September 2003.

Schedule of Talks

Monday, April 28, 2003 2:30pm-3:20pm 312 Boucke Building

Scientific Discovery through Advanced Computing

ABSTRACT:  The Scientific Discovery through Advanced Computing (SciDAC) initiative is a web of interconnected projects --- partly research and partly software development --- designed to support simulation, data exploration, and collaboration in many thrust areas of the U.S. Department of Energy, including: climate modeling, fusion energy, chemical and materials science, astrophysics, and high energy and particle physics. SciDAC supports the creation of a new generation of scientific simulation codes for terascale systems. The program also includes research on numerical algorithms and systems software that will allow these codes to use modern parallel computers effectively. SciDAC will be of interest to modelers and researchers for the freely available software it will produce, and its approach is already being imitated in Europe and Asia. The acclaimed solver software "PETSc" is part of the nine-institution "Terascale Optimal PDE Simulations" (TOPS) project within SciDAC. As the director of the TOPS project, the speaker will provide an overview of its scope and scientific goals at a "big picture" level.

Tuesday, April 29, 2003 2:30pm-3:20pm 116 McAllister Building

Scalable Solvers and Software for PDE Applications

ABSTRACT:  Like the theoretical peak performance of a computer system, theoretical efficiency for algorithms is rarely closely approached for real applications. While the quest for the "textbook efficiency" continues on many fronts, real users need to have their solver capabilities upgraded today to exploit the platform potential to run more highly resolved computations. The Terascale Optimal PDE Simulations (TOPS) project of the SciDAC initiative is working on both fronts --- attempting to make fundamental advances in numerical algorithms that will be integrated into tomorrow's scalable solver software while attempting to be of service to SciDAC application developers and others at the outset of the initiative.
   In this talk, we dwell on some practical aspects of migrating from a legacy (usually operator-split) nonlinear solver for evolutionary or equilibrium systems of PDEs to a Jacobian-free Newton-Krylov framework that provides strong controls on splitting error while still incorporating physically-based operator-split methodology (and even legacy subroutines) where possible. It is emphasized that to support even a single application from development through production use on various platforms, contemporary solver libraries must offer a menu of flexibly combinable and tunable components to allow application-specific and architecture-specific trade-offs (e.g., memory versus flops, synchronization frequency versus stability, robustness versus efficiency). We also discuss some experiences with the M3D extended magnetohydrodynamics code of our PPPL-based SciDAC partners, which is designed to underscore the desirability of being able to draw from a broad family of solvers within a single application.