Friday, July 11

MS110
Combinatorial Algorithms in Scientific Computing

4:00 PM - 6:00 PM
Room: Royal Palm 5

Combinatorial algorithms play a vital enabling role in computational science and engineering. Many scientific computing problems require tasks that are inherently combinatorial (e.g. data partitioning). Other scientific computing problems have more subtle underlying discrete structures that complement the analytic structures of the problems (e.g. combinatorial structure in discretized PDEs). As computational science problems continue to grow, solutions become more complex, and systems become more parallel, the exploitation of these combinatorial structures should continue to grow in importance. In this session, we present several examples of the development and usage of combinatorial algorithms for the solution of scientific computing problems.

Organizer: Michael M. Wolf
Sandia National Laboratories
Robert Kirby
Texas Tech University

4:00-4:25 Hypergraph-Based Combinatorial Optimization of Matrix-Vector Multiplication

Michael M. Wolf, Sandia National Laboratories

4:30-4:55 Using Sieve for Particle Tracking, Embedded Meshing and Field-Particle Interaction Computations

Dmitry Karpeev and Matthew G. Knepley, Argonne National Laboratory

5:00-5:25 Combinatorial Dataflow Analysis for Differentiation of High-Level PDE Representations

Kevin Long, Texas Tech University; Catherine Beni, California Institute of Technology; Robert Kirby, Texas Tech University

5:30-5:55 Exploitation of Jacobian Scarcity

Andrew Lyons, University of Chicago; Jean Utke and Paul D. Hovland, Argonne National Laboratory