ftp://ftp.cs.berkeley.edu/pub/lapack/SuperLU/superlu.tar.gz SuperLU <title_line>high performance sparse linear system solver <author>James Demmel, Stan Eisenstat, John Gilbert, Xiaoye Li, Joseph Liu, and Jinqchong Teo <contact>Xiaoye Li xiaoye@cs.berkeley.edu John Gilbert gilbert@parc.xerox.com James Demmel demmel@cs.berkeley.edu Jinqchong Teo teo@pasteur.eecs.berkeley.edu <abstract>We are pleased to announce the availability of the SuperLU package, a set of subroutines to solve a sparse linear system A*X=B. It implements Gaussian elimination with partial pivoting, using supernodes and BLAS to optimize performance. It has achieved over 120 Mflops on realistic examples on an IBM RS 6000/590; it is currently the fastest available algorithm for a variety of problems. <environment>SuperLU is implemented in ANSI C, and must be compiled with standard ANSI C compilers. We supply the BLAS in C, but for highest performance optimized BLAS should be used. Currently only the real single-precision and double-precision versions are provided; the complex versions are still under construction. The calling sequence is modeled on LAPACK. There is also a Matlab mex-file interface, test and installation code, a user guide and detailed technical report. <category>numerical-linalg <keywords>sparse linear systems </urc>