Tuesday, February 20

MS30
Combinatorial Scientific Computing and Petascale Simulations

9:45 AM - 11:45 AM
Room: Laguna Beach III - B1

Combinatorial algorithms have long played a role in scientific computing, for example in load-balancing and partitioning. Combinatorial scientific computing (CSC) is the study of combinatorial problems and algorithms that arise in scientific computing. Developing algorithms and software tools in the CSC area will be critical to enable future petascale simulations in CS\&E. Researchers from four institutions recently teamed together for a DOE-funded SciDAC institute called CSCAPES. This minisymposium will give an overview of CSC but focus on topics that are part of CSCAPES; in particular, load-balancing, automatic differentiation, and graph coloring.

Organizer: Erik G. Boman
Sandia National Laboratories

9:45-10:10 Graph Coloring Problems for Computing Derivatives: Recent Developments and Future Plans

Alex Pothen and Assefaw H. Gebremedhin, Purdue University

10:15-10:40 Parallel Hypergraph Repartitioning and Load Balancing

Karen D. Devine, Sandia National Laboratories; Umit V. Catalyurek, The Ohio State University; Erik G. Boman, Sandia National Laboratories; Doruk Bozdag, The Ohio State University; Robert Heaphy and Lee Ann Fisk, Sandia National Laboratories

10:45-11:10 How Scalable is Your Load Balancer?

Umit V. Catalyurek and Doru Bozdag, The Ohio State University; Karen D. Devine and Erik G. Boman, Sandia National Laboratories

11:15-11:40 Performance Improvement in a Mesh Quality Optimization Application

Paul D. Hovland, Argonne National Laboratory; Sanjukta Bhowmick, University of Nebraska, Omaha; Todd S. Munson, Argonne National Laboratory; Michelle Strout, Colorado State University