Highly scalable parallel algorithms for sparse matrix factorization

Abstract

In this paper, we describe scalable parallel algorithms for sparse matrix factorization, analyze their performance and scalability, and present experimental results for up to 1024 processors on a Cray T3D parallel computer. Through our analysis and experimental results, we demonstrate that our algorithms substantially improve the state of the art in parallel direct solution of sparse linear systems-both in terms of scalability and overall performance. It is a well known fact that dense matrix factorization scales well and can be implemented efficiently on parallel computers. In this paper, we present the first algorithms to factor a wide class of sparse matrices (including those arising from two-and three-dimensional finite element problems) that are asymptotically as scalable as dense matrix factorization algorithms on a variety of parallel architectures. Our algorithms incur less communication overhead and are more scalable than any previously known parallel formulation of sparse matrix factorization. Although, in this paper, we discuss Cholesky factorization of symmetric positive definite matrices, the algorithms can be adapted for solving sparse linear least squares problems and for Gaussian elimination of diagonally dominant matrices that are almost symmetric in structure. An implementation of one of our sparse Cholesky factorization algorithms delivers up to 20 GFlops on a Cray T3D for medium-size structural engineering and linear programming problems. To the best of our knowledge, this is the highest performance ever obtained for sparse Cholesky factorization on any supercomputer.

SLIM3D: A tool for three-dimensional thermomechanical modeling of lithospheric deformation with elasto-visco-plastic rheology

“…This fact is explained mostly by the following: (i) the code is still sequential (ii) the code utilizes a linear solver which is not numerically scalable (Farhat et al, 2000). To remove these disadvantages, we plan to extend our code in the following mainstream directions of linear solver technology: (i) Direct parallel multifrontal solvers (Gupta et al, 1997;Amestoy et al, 2000) (ii) Multigrid methods with adaptation for plasticity (Adams, 2000;Ekevid et al, 2004) (iii) Dual-primal domain decomposition methods (Farhat et al, 2000) At the same time, we are planning to parallelize the entire finite element routines using the MPI package.…”

Locality-sensitive and Re-use Promoting Personalized PageRank computations

“…Sub-task 4 involves decomposition of the compensation matrix M 0 . Since M 0 is sparse, this step can also be parallelized effectively (Gupta et al, 1997). Finally, Sub-tasks 5 and 6 involve matrix multiplications and inversions.…”

Towards Lazy Data Association in SLAM

“…The mean and covariance are the estimate of the SLAM posterior at time t. The classical solution to this problem involves the inversion of a spare matrix, which is costly [8]. However, there exists a number of efficient approximations, such as loopy belief propagation [17] and tree-based approximation techniques [27] for approximating these quantities; all of those techniques can exploit an existing solution when modifying the set of equality constraints.…”