P. Hénon and P. Ramet.
PaStiX: Un solveur parallèle direct pour des matrices creuses symétriques définies positives basé sur un ordonnancement statique performant et sur une gestion mémoire efficace.
In ACTES RenPar'2001,
Paris, France,
April 2001.
Keyword(s): Sparse.
Abstract: |
La r\'esolution de grands syst\`emes lin\'eaires creux est un point crucial dans de nombreuses applications industrielles et scientifiques. Notre travail porte sur le partionnement et la distribution de grandes matrices creuses pour la factorisation $LDL^t$ en parall\`ele sur machine de type MIMD. Nous pr\'esentons dans cet article notre technique de factorisation parall\`ele bas\'ee sur un ordonnancement statique des calculs et des communications, et nous la validons sur des syst\`emes de plus d'un million d'inconnues pour des probl\`emes d'\'el\'ements finis 3D. |
@InProceedings{c:LaBRI::HR01,
author = {P. H\'enon and P. Ramet},
title = {{PaStiX}: Un solveur parall\`ele direct pour des matrices creuses sym\'etriques d\'efinies positives bas\'e sur un ordonnancement statique performant et sur une gestion m\'emoire efficace},
booktitle = "ACTES RenPar'2001",
OPTcrossref = {},
OPTkey = {},
OPTpages = {},
year = {2001},
OPTeditor = {},
OPTvolume = {},
OPTnumber = {},
OPTseries = {},
address = {Paris, France},
month = apr,
OPTorganization = {},
OPTpublisher = {},
OPTnote = {},
OPTannote = {},
URL = {http://www.labri.fr/~ramet/restricted/renpar01.ps},
KEYWORDS = "Sparse",
ABSTRACT = {La r\'esolution de grands syst\`emes lin\'eaires creux est un point crucial dans de nombreuses applications industrielles et scientifiques. Notre travail porte sur le partionnement et la distribution de grandes matrices creuses pour la factorisation $LDL^t$ en parall\`ele sur machine de type MIMD. Nous pr\'esentons dans cet article notre technique de factorisation parall\`ele bas\'ee sur un ordonnancement statique des calculs et des communications, et nous la validons sur des syst\`emes de plus d'un million d'inconnues pour des probl\`emes d'\'el\'ements finis 3D.}
}
P. Hénon,
P. Ramet,
and J. Roman.
PaStiX: A Parallel Direct Solver for Sparse SPD Matrices based on Efficient Static Scheduling and Memory Managment.
In Tenth SIAM Conference on Parallel Processing for Scientific Computing,
Portsmouth, USA,
March 2001.
Keyword(s): Sparse.
Abstract: |
Solving large sparse symmetric positive definite systems of linear equations is a crucial and time-consuming step, arising in many scientific and engineering applications. In this work, we consider the block partitioning and scheduling problem for sparse parallel factorization without pivoting. We focus on the scalability of the parallel solver, and on the compromise between memory overhead and efficiency. We validate this study with parallel experiments on a large collection of irregular industrial problems. |
@InProceedings{C:LaBRI::siam2001,
author = {H\'enon, P. and Ramet, P. and Roman, J.},
title = {{PaStiX}: {A} {P}arallel {D}irect {S}olver for {S}parse {SPD} {M}atrices based on {E}fficient {S}tatic {S}cheduling and {M}emory {M}anagment},
booktitle = {Tenth {SIAM} Conference on Parallel Processing for Scientific Computing},
OPTcrossref = {},
OPTkey = {},
OPTpages = {},
year = {2001},
OPTeditor = {},
OPTvolume = {},
OPTnumber = {},
OPTseries = {},
address = {Portsmouth, USA},
month = mar,
OPTorganization = {},
OPTpublisher = {},
OPTnote = {},
OPTannote = {},
URL = {http://www.labri.fr/~ramet/restricted/ppsc01.ps},
KEYWORDS = "Sparse",
ABSTRACT = {Solving large sparse symmetric positive definite systems of linear equations is a crucial and time-consuming step, arising in many scientific and engineering applications. In this work, we consider the block partitioning and scheduling problem for sparse parallel factorization without pivoting. We focus on the scalability of the parallel solver, and on the compromise between memory overhead and efficiency. We validate this study with parallel experiments on a large collection of irregular industrial problems.}
}