Fall 2007 Colloquium Series - Ramani Duraiswami

Archived Content

Please Note: The content on this page is not maintained after the colloquium event is completed.  As such, some links may no longer be functional.

Download Adobe PDF Reader

Ramani DuraiswamiRamani Duraiswami
Terascale on the Desktop: Fast Multipole Methods on Graphical Processors

Wednesday, November 7 2007
Building 3 Auditorium - 3:30 PM
(Refreshments at 3:00 PM)

Graphics Processors (GPUs) provide access to significant computational processing resources at low costs. They contain a large number of processing units with access to local and shared memory, and achieve significant speedups vis-a`-vis CPUs on problems that can be mapped to their SPMD architecture. Many applications in molecular dynamics, astrophysics, fluid mechanics and other areas require the O(N^2) computation of mutual Coulombic potentials and forces among N particles. The FMM provides a hierarchical approximate algorithm, to compute these quantities to a specified error at O(NlogN) cost and memory. More generally FMM like algorithms are used to accelerate matrix vector products in applications such as the solution of integral equations, radial basis function interpolation/evaluation and machine learning. I will discuss both the GPU architecture for scientific computing, and the modifications necessary to the FMM for this architecture. To allow easy programming of the GPU architecture, a library callable from a high level programming language was developed. On an NVIDIA 8800 GTX installed on a PC, our FMM code achieves timings that if computed using an O(N^2) algorithm correspond to speeds of 25-45 Tflops (for achieved L2 errors of ~ 10^-6 - 10^-4).

Ramani Duraiswami is an associate professor at the Dept. of Computer Science and the Institute for Advanced Computer Studies at the University of Maryland, College Park. He directs the research at the Perceptual Interfaces and Reality Lab. there and has broad research interests in scientific computing, computational audition, computer vision and machine learning. He received a Ph.D. in Mechanical Engineering from The Johns Hopkins University in 1991, and a B. Tech from the Indian Institute of Technology, Bombay, in 1985. More information on Prof. Duraiswami's research and publications can be obtained at www.umiacs.umd.edu/users/ramani.
(joint work with Nail A. Gumerov. Work supported by NASA)

 

IS&T Colloquium Committee Host: Tony Gualtieri