Xiaoye Li is a staff computer scientist at the Lawrence Berkeley National Laboratory, Department of Energy, USA. She has over ten years experience working in the research area of high performance scientific computing, particularly in design,

implementation, and performance optimization of parallel numerical algorithms and mathematical software on large-scale parallel computers. She is well recognized for her work on sparse matrix computations, and is the principal author of the widely-used sparse linear solver package SuperLU. Xiaoye received B.S. from Tsinghua University, M.S. from Penn State University, and Ph.D. from University of California at Berkeley. Her research articles are available at http://crd.lbl.gov/~xiaoye.

Abstract

Computer modeling of physical phenomena now plays an increasingly important role in science and engineering. Examples of such modeling include accelerator design, simulation of fusion reactor plasma, and nanostructure calculation. In many of these calculations, a large fraction of the time is usually spent in the solution of sparse linear systems or eigen value problems. In this talk, we will show our experiences in developing scalable parallel sparse matrix algorithms and software that take advantage of the modern parallel computer architectures, as well as meet the applications needs.