Design and optimization of algorithms on parallel machines
High performance algebraic solvers
Sparse matrix computations
I do research and provide support for mathematical software on the parallel machines at NERSC. Currently, I am heavily involved in the DOE SciDACprojects. Previously I worked on NSFNPACI project.
I received Ph.D. in Computer Sciencefrom UC Berkeley.
Selected papers:( Full list, over 90 papers)
SuperLU — Sequential and parallel libraries to solve unsymmetric sparse linear systems using LU factorization.
PDSLin — Parallel Domain-decomposition, Schur complement based hybrid sparse linear solver.
XBLAS — A reference implementation for the Extended and Mixed precision BLAS standard.
ARPREC — A C++/F90 package for performing arbitrary precision arithmetic.
CLAPACK — Full set of LAPACK in C.
ieee_except — A set of condition estimation routines, which are faster than those in LAPACK, by utilizing floating-point exception handling. It contains routines to manipulate IEEE exception sticky flags.
SIAM Fellow, Society for Industrial and Applied Mathematics, 2016Elected to the SIAM Council, Jan. 2016 – Dec. 2018Invited Plenary Speaker, SIAM Conference on Applied Linear Algebra, October 26-30, 2015Invited Lecturer, the Fourth Gene Golub SIAM Summer School, July 22 – August 2, 2013Invited Plenary Speaker, SIAM Annual Meeting, July 12-16, 2010
Invited Plenary Speaker, Accelerating Direct Linear Solvers with Algorithmic and Hardware Advances, SIAM Conf. on Applied Linear Algebra, Oct. 26-30, 2015, Atlanta. ( View presentation)
Invited Lecturer, short course onFactorization-based sparse solvers and preconditioners, 4th Gene Golub SIAM Summer School, July 22-Aug. 9, 2013, Shanghai. (Videos and course materials, Extended Summary (in book “Matrix Functions and Matrix Equations”, Z. Bai, W. Gao, Y. Su, editors, Series in Contemporary Applied Mathematics, World Scientific Publisher, Oct. 2015, pp. 109-137)
Towards an optimal-order approximate sparse factorization exploiting data-sparseness in separators, Invited Speaker, Workshop Celebrating 40 Years of Nested Dissection, July 22-23, 2013, Waterloo.
Invited Plenary Speaker, Factorization-based sparse solvers and preconditioners, SIAM Annual Meeting, July 12-16, 2010. (View presentation)
Towards an Optimal Parallel Approximate Factorization Using HSS Structures, Householder Symposium XVIII, June 13-17, 2011, Tahoe City, California.
A Supernodal Approach for ILU with Partial Pivoting, Sparse Days 2010 (Invited), CERFACS, June 15-17, 2010
Sparse matrix methods on high performance computers, Invited Lecture, CS267/Eng233, UCB, March 16, 2010.
Use of Semi-separable Approximate Factorization and Direction-preserving for Constructing Effective Preconditioners, SIAM Conference on Applied Linear Algebra, Monterey Bay-Seaside, Oct. 26-29, 2009.
Performance Modeling Tools for Parallel Sparse Linear Algebra Computations, ParCo 2009, September 1-4, 2009, ENS-Lyon, France.
Scalability Issues in Sparse Factorization and Triangular Solution, Sparse Days, June 23-24, 2008, CERFACS, Toulouse, France.
Evaluation of sparse LU factorization and triangular solution on multicore architectures, VECPAR 2008, June 24-27, 2008, Toulouse, France.
Algebraic Sub-structuring for Large-scale Electromagnetic Application, 16th International Conference on Domain Decomposition Methods, January 12-15, 2005, Courant Institute, New York University.
A Comparison of Three High-Precision Quadrature Schemes, Experimental Math Workshop, March 29-30, 2004. Oakland, CA.
Fill Reduction Algorithm Using Diagonal Markowitz Scheme with Local Symmetrization, SIAM Conference on Computational Science and Engineering, February 10-13, 2003, San Diego.
Selected professional services: