STRUMPACK -- STRUctured Matrices PACKage, Copyright (c) 2014-2017, The Regents of the University of California, through Lawrence Berkeley National Laboratory (subject to receipt of any required approvals from the U.S. Dept. of Energy). All rights reserved.
http://portal.nersc.gov/project/sparse/strumpack/master/
http://portal.nersc.gov/project/sparse/strumpack/
- Pieter Ghysels - [email protected] (Lawrence Berkeley National Laboratory)
- Xiaoye S. Li - [email protected] (Lawrence Berkeley National Laboratory)
- Gustavo Chávez - [email protected] (Lawrence Berkeley National Laboratory)
- Yang Liu - [email protected] (Lawrence Berkeley National Laboratory)
- Lucy Guo - [email protected]
- Liza Rebrova - [email protected] (University of Michigan)
- François-Henry Rouet - [email protected],[email protected] (Livermore Software Technology Corp., Lawrence Berkeley National Laboratory)
- Theo Mary - [email protected] (University of Manchester)
- Christopher Gorman - (UC Santa Barbara)
- Jonas Actor - (Rice University)
STRUMPACK - STRUctured Matrix PACKage - is a software library providing linear algebra routines for sparse matrices and for dense rank-structured matrices, i.e., matrices that exhibit some kind of low-rank property. In particular, STRUMPACK uses the Hierarchically Semi-Separable matrix format (HSS). Such matrices appear in many applications, e.g., Finite Element Methods, Boundary Element Methods ... In sparse matrix factorization, the fill-in in the triangular factors often has a low-rank structure. Hence, the sparse linear solve in STRUMPACK exploits the HSS matrix format to compress the fill-in. Exploiting this structure using a compression algorithm allows for fast solution of linear systems and/or fast computation of matrix-vector products, which are two of the main building blocks of matrix computations. STRUMPACK has two main components: a distributed-memory dense matrix computations package (for dense matrices that have the HSS structure) and a distributed memory fully algebraic sparse general solver and preconditioner. The preconditioner is mostly aimed at large sparse linear systems which result from the discretization of a partial differential equation, but is not limited to any particular type of problem. STRUMPACK also provides preconditioned GMRES and BiCGStab iterative solvers.
- The sparse solver is documented in doc/manual.pdf. STRUMPACK-sparse can be used as a direct solver for sparse linear systems or as a preconditioner. It also includes GMRes and BiCGStab iterative solvers that can use the preconditioner. The preconditioning strategy is based on applying low-rank approximations to the fill-in of a sparse multifrontal LU factorization. The code uses MPI+OpenMP for hybrid distributed and shared memory parallelism. The main point of contact is: Pieter Ghysels ([email protected]).
- The dense distributed-memory package can be found in the src/HSS directory. This is currently not well documented.
If you have questions about your rights to use or distribute this software, please contact Berkeley Lab's Technology Transfer Department at [email protected].
This software is owned by the U.S. Department of Energy. As such, the U.S. Government has been granted for itself and others acting on its behalf a paid-up, nonexclusive, irrevocable, worldwide license in the Software to reproduce, prepare derivative works, and perform publicly and display publicly. Beginning five (5) years after the date permission to assert copyright is obtained from the U.S. Department of Energy, and subject to any subsequent five (5) year renewals, the U.S. Government is granted for itself and others acting on its behalf a paid-up, nonexclusive, irrevocable, worldwide license in the Software to reproduce, prepare derivative works, distribute copies to the public, perform publicly and display publicly, and to permit others to do so.