Movatterモバイル変換


[0]ホーム

URL:


Skip to content

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up

Structured Matrix Package (LBNL)

License

NotificationsYou must be signed in to change notification settings

pghysels/STRUMPACK

Repository files navigation

STRUMPACK -- STRUctured Matrix PACKage, Copyright (c) 2014-2021, TheRegents of the University of California, through Lawrence BerkeleyNational Laboratory (subject to receipt of any required approvals fromthe U.S. Dept. of Energy). All rights reserved.

Documentation & Installation instructions

http://portal.nersc.gov/project/sparse/strumpack/master/

http://portal.nersc.gov/project/sparse/strumpack/v7.2.0/

Website

http://portal.nersc.gov/project/sparse/strumpack/

Current developers - Lawrence Berkeley National Laboratory

Past contributors

  • Ryan Synk
  • Lucy Guo
  • Gustavo Chávez
  • Liza Rebrova
  • François-Henry Rouet
  • Theo Mary
  • Christopher Gorman
  • Jonas Actor
  • Michael Neuder

Overview

STRUMPACK - STRUctured Matrix PACKage - is a software libraryproviding linear algebra routines and linear system solvers for sparseand for dense rank-structured linear systems. Many large densematrices are rank structured, meaning they exhibit some kind oflow-rank property, for instance in hierarchically definedsub-blocks. In sparse direct solvers based on LU factorization, the LUfactors can often also be approximated well using rank-structuredmatrix compression, leading to robust preconditioners. The sparsesolver in STRUMPACK can also be used as an exact direct solver, inwhich case it functions similarly as for instanceSuperLU orsuperlu_dist. TheSTRUMPACK sparse direct solver delivers good performance anddistributed memory scalability and provides excellent CUDA support.

Currently, STRUMPACK has support for the Hierarchically Semi-Separable(HSS), Block Low Rank (BLR), Hierachically Off-Diagonal Low Rank(HODLR), Butterfly and Hierarchically Off-Diagonal Butterfly (HODBF)rank-structured matrix formats. Such matrices appear in manyapplications, e.g., the Boundary Element Method for discretization ofintegral equations, structured matrices like Toeplitz and Cauchy,kernel and covariance matrices etc. In the LU factorization of sparselinear systems arising from the discretization of partial differentialequations, the fill-in in the triangular factors often has low-rankstructure. Hence, the sparse linear solve algorithms in STRUMPACKexploit the different dense rank-structured matrix formats to compressthe fill-in. This leads to purely algebraic, fast and scalable (bothwith problem size and compute cores) approximate direct solvers orpreconditioners. These preconditioners are mostly aimed at largesparse linear systems which result from the discretization of apartial differential equation, but are not limited to any particulartype of problem. STRUMPACK also provides preconditioned GMRES andBiCGStab iterative solvers.

Apart from rank-structured compression, the STRUMPACK sparse solveralso support compression of the factors using theZFPlibrary, a general purpose compression algorithm tuned for floatingpoint data. This can be used with a specified precision, or withlossless compression.

The HODLR and Butterfly functionality in STRUMPACK is implementedthrough interfaces to the ButterflyPACK package:https://github.com/liuyangzhuan/ButterflyPACK

NOTICE

This software is owned by the U.S. Department of Energy. Assuch, the U.S. Government has been granted for itself and othersacting on its behalf a paid-up, nonexclusive, irrevocable, worldwidelicense in the Software to reproduce, prepare derivative works, andperform publicly and display publicly. Beginning five (5) years afterthe date permission to assert copyright is obtained from theU.S. Department of Energy, and subject to any subsequent five (5) yearrenewals, the U.S. Government is granted for itself and others actingon its behalf a paid-up, nonexclusive, irrevocable, worldwide licensein the Software to reproduce, prepare derivative works, distributecopies to the public, perform publicly and display publicly, and topermit others to do so.

If you have questions about your rights to use or distribute thissoftware, please contact Berkeley Lab's Technology Transfer DepartmentatTTD@lbl.gov.


[8]ページ先頭

©2009-2025 Movatter.jp