4.7 Article

Fast Sparse Cholesky Decomposition and Inversion using Nested Dissection Matrix Reordering

Journal

JOURNAL OF CHEMICAL THEORY AND COMPUTATION
Volume 7, Issue 2, Pages 351-368

Publisher

AMER CHEMICAL SOC
DOI: 10.1021/ct100618s

Keywords

-

Funding

  1. Office of Energy Research, Office of Basic Energy Sciences, Chemical Sciences Division of the U.S. Department of Energy [DE-AC0376SF00098]

Ask authors/readers for more resources

Here we present an efficient, yet nonlinear scaling, algorithm for the computation of Cholesky factors of sparse symmetric positive definite matrices and their inverses. The key feature of this implementation is the separation of the task into an algebraic and a numeric part. The algebraic part of the algorithm attempts to find a reordering of the rows and columns which preserves at least some degree of sparsity and afterward determines the exact nonzero structure of both the Cholesky factor and its corresponding inverse. It is based on graph theory and does not involve any kind of numerical thresholding. This preprocessing then allows for a very efficient implementation of the numerical factorization step. Furthermore this approach even allows use of highly optimized dense linear algebra kernels which leads to yet another performance boost. We will show some illustrative timings of our sparse code and compare it to the standard library implementation and a recent sparse implementation using thresholding. We conclude with some comments on how to deal with positive semidefinite matrices.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.7
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available