Summary of available software for sparse direct methods
Timothy A. Davis · 2006
Table 1 summarizes most of the available software for solving sparse linear systems via direct methods as April 2006. The first column lists the name of the package. The next four columns describe what kinds of factorizations are available: LU, Cholesky, LDL for symmetric indefinite matrices, and QR. If the LDL factorization uses 2-by-2 block pivoting a “2” is listed; a “1” is listed otherwise [10, 11]. The next column states if complex matrices (unsymmetric, symmetric, and/or Hermitian) are supported. The ordering methods available are listed in the next four columns: minimum degree and its variants (minimum fill, column minimum degree, Markowitz, and related methods), nested or one-way dissection (including all graph-based partitionings), permutation to block triangular form, and profile/bandwidth reduction (or related) methods. The next three columns indicate what level of BLAS is used (1: vector, 2: matrix-vector, 3: matrix-matrix), if the package is parallel (“s” for shared-memory or “d” for distributed-memory), and whether or not the package includes an out-of-core option (where most of the factors remain on disk). Most distributedmemory packages can also be used in a shared-memory environment, since most message-passing libraries (MPI in particular) are ported to shared-memory environments. A code is listed as “sd” if it includes two versions, one for shared-memory and the other for distributed-memory. The next column indicates if a MATLAB interface is available. The primary method(s) used in the package are listed in the final column. Table 2 lists the authors of the packages, relevant papers, and where to get the code. An up-to-date table will be maintained at www.cise.ufl.edu/research/sparse/codes. Updates to this document are welcome. Tim Davis, http://www.cise.ufl.edu/∼davis