8:30 AM-9:15 AM
Room: Capitol Center/South
Chair: Jack J. Dongarra, University of Tennessee, Knoxville; and Oak Ridge National Laboratory
Larger and harder problems, new computers, and an expanding user community continue to raise new challenges for numerical linear algebra. The speaker will present recent progress and open problems in these areas. These include automatic performance tuning of algorithms to match architectures, new parallel direct and iterative algorithms for large sparse linear systems, new parallel eigenvalue/eigenvector algorithms with optimal parallel complexity, and a new book to help users find the best algorithm and software for their eigenvalue problems.
James W. Demmel
Department of Computer Science
University of California, Berkeley