I am a research assistant in the Computer Science Department at Vanderbilt University. I work for the Computational Science research group, which focuses on the development of software for large-scale atomic structure calculations. The computational problem involved is solving a partial differential equation of eigenvalue type. To solve this problem, the variational method is used, and the nature of the variational function divides the problem into two major constituents -- solving linear integro-differential equations, and finding a few lowest eigenvalue/eigenvector pairs of a large, sparse, symmetric matrix. The matrix computation and diagonalization are the most computationally expensive tasks, thus parallel algorithms are employed in their execution.

The calculations of the research group are performed on a variety of distributed memory parallel computers, including the NERSC Cray T3E and IBM SP. My current project is to benchmark the performance of the parallel relativistic atomic structure software on a local 4-node PC cluster.

In the workshop, I hope to get acquainted with the state-of-the-art linear algebra software for distributed memory parallel computers; in particular, I am interested in the solution of the large, sparse, symmetric eigenvalue problem. I would like to compare the performance of the eigensolvers from the NERSC ACTS Toolkit with that of the algorithms currently used in parallel atomic structure calculations.

I am also interested in learning about the run-time support of parallel computing in distributed-memory environments (in particular, performance analysis and fault tolerance tools), with the prospect of using these tools to facilitate the execution of our parallel programs.