zbMATH — the first resource for mathematics

A supernodal approach to sparse partial pivoting. (English) Zbl 0931.65022
This paper constructs a column type method of finding supernodes for large sparse and unsymmetric linear systems of equations. This generalizes one standard method for symmetric sparse Cholesky factorizations to nonsymmetric system matrices. It uses unsymmetric supernodes during a column based LU factorization with symmetric pruning in its symbolic factorization steps and the unsymmetric supernodes in the numeric LU factorization. There are benchmark comparisons with other algorithms.
Reviewer: F.Uhlig (Auburn)

65F05 Direct numerical methods for linear systems and matrix inversion
65F50 Computational methods for sparse matrices
15A23 Factorization of matrices
Full Text: DOI