Fastest sparse solver on MATLAB

GRUS SPARSE SOLVER

GSS (GRUS SPARSE SOLVER) is an ANSI C sparse solver for sparse matrix below one million orders. GSS uses new models and many new algorithms at the symbolic-analysis phase. GSS uses multifrontal method to factor A to LU, it supports complete pivoting, partial pivoting ,rook-pivoting and static pivoting .Experiments show that GSS 1.2 is much faster than UMFPACK 4.4. Time of numerical factorization by GSS is less than 1/3 of UMFPACK on average.

GSSMEX is a mex function for MATLAB which uses the core of GSS 1.2. It's much faster than lu function of MATLAB 7.1. Table 1 lists the experimental results. Column “Time of lib” is the time of numerical factorization by GSS 1.2 and UMFPACK 4.4. Column “Time of Matlab” is the time of GSS mex function and lu fuction in Matlab(sec). ‘-’ represents that Matlab 7.1 failed to solve the equations.

GSSMEX 2.0 is ready. Submit application for a free version.

GSSMEX readme.

Any problem, contact gsp@grusoft.com.