2012
DOI: 10.1016/j.jcp.2011.10.013
|View full text |Cite
|
Sign up to set email alerts
|

A fast direct solver for elliptic problems on general meshes in 2D

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

1
75
0

Year Published

2013
2013
2022
2022

Publication Types

Select...
10

Relationship

1
9

Authors

Journals

citations
Cited by 95 publications
(76 citation statements)
references
References 17 publications
1
75
0
Order By: Relevance
“…These solvers do not have optimal complexity (they typically require O(N 3/2 ) for the factorization stage in 2D, and O(N 2 ) in 3D), but are nevertheless popular (especially in 2D) due in part to their robustness, and in part to the unrivaled speed that can be attained for problems involving multiple right hand sides. Very recently, it has been demonstrated that by exploiting structured matrix algebra (such as, e.g., H-matrices, or HSS matrices), to manipulate the dense matrices that arise due to fill-in, direct solvers of linear or close to linear complexity can be constructed [17,28,30,35,40]. The direct solver described in this paper is conceptually similar to these algorithms in that they all rely on hierarchical domain decompositions, and efficient representations of operators that live on the interfaces between sub-domains.…”
Section: Direct Solvers For Sparse Systemsmentioning
confidence: 99%
“…These solvers do not have optimal complexity (they typically require O(N 3/2 ) for the factorization stage in 2D, and O(N 2 ) in 3D), but are nevertheless popular (especially in 2D) due in part to their robustness, and in part to the unrivaled speed that can be attained for problems involving multiple right hand sides. Very recently, it has been demonstrated that by exploiting structured matrix algebra (such as, e.g., H-matrices, or HSS matrices), to manipulate the dense matrices that arise due to fill-in, direct solvers of linear or close to linear complexity can be constructed [17,28,30,35,40]. The direct solver described in this paper is conceptually similar to these algorithms in that they all rely on hierarchical domain decompositions, and efficient representations of operators that live on the interfaces between sub-domains.…”
Section: Direct Solvers For Sparse Systemsmentioning
confidence: 99%
“…Closely related to the concept of H-matrices and its arithmetic are "hierarchically semiseparable matrices", [Xia13,XCGL09,LGWX12] and the idea of "recursive skeletonization", [HG12,GGMR09,HY13a]; for discretizations of PDEs, we mention [HY13a,GM13,SY12,Mar09], and particular applications to boundary integral equations are [MR05,CMZ13,HY13b]. These factorization algorithms aim to exploit that some off-diagonal blocks of certain Schur complements are low rank.…”
Section: Introductionmentioning
confidence: 99%
“…Exploiting the internal structure in these matrices, the complexity of the build stage can be reduced from OðN 1:5 Þ to OðNÞ. This acceleration is analogous to what was done for classical nested dissection for finite-element and finite-difference matrices in [9,21,30,35]. Note that while the acceleration of the solve phase is trivial, it takes some work to exploit the more complicated structure in V s and W s .…”
Section: Linear Complexity Algorithmsmentioning
confidence: 81%