University of Houston

Department of Computer Science

In partial fulfillment of the Requirements for the Degree of
Doctor of Philosophy

Jakub Kurzak
will present his preliminary defense

Massively Parallel Implementation

of Particle Dynamics Fast Multipole Method

for Distributed Memory Systems

Abstract

The Fast Multipole Method is the most sophisticated algorithm for calculating electrostatic forces in molecular systems (solving the underlaying n-body problem). It is characterized by superior O(N) computational complexity, where N is the number of particles, with significant constant coefficient however. In the last decade a lot of effort has been made to alleviate this problem and as a result today the Fast Multipole Method is an attractive alternative to traditional Ewald summation based methods with break-even point for systems of size O(10K) particles.

The focus of this proposal is a massively parallel implementation of the Fast Multipole Method with Periodic Boundary Conditions which facilitates Multiple Time Stepping techniques and preserves computational determinism. Scaling to hundreds of processors for O(10K) systems is achieved by novel approach to balancing the load, overlapping the communication with computation and improving the locality of communication.

Date: Tuesday, August 31, 2004
Time: 4:00 PM
Place: 218-PGH


Faculty, students, and the general public are invited.
Thesis Advisor: Dr. B. Montgomery Pettitt