A parallel block-preconditioned GCR method for incompressible flow problems

C. Vuik; J. Frank; A. Segal

doi:10.1016/S0167-739X(00)00073-X

A parallel block-preconditioned GCR method for incompressible flow problems

C. Vuik^*, J. Frank, A. Segal

^*Corresponding author for this work

Research output: Contribution to journal › Article › Academic › peer-review

Abstract

Efficient parallel algorithms are required to simulate incompressible turbulent flows in complex two- and three-dimensional domains. The incompressible Navier-Stokes equations are discretized in general coordinates on a structured grid. For a flow on a general domain we use an unstructured decomposition of the domain into subdomains of simple shape, with a structured grid inside each subdomain. We have developed a parallel block-preconditioned GCR method to solve the resulting systems of linear equations. The method can be smoothly varied between a coarse grain parallel method in which the subdomain problems are solved accurately with an inner iteration process and a fine grain parallel method when only a preconditioner is used to approximate the solution on the blocks. Parallel performance results for Boussinesq flow in a cavity are included.

Original language	English
Pages (from-to)	31-40
Number of pages	10
Journal	Future Generation Computer Systems
Volume	18
Issue number	1
DOIs	https://doi.org/10.1016/S0167-739X(00)00073-X
Publication status	Published - Sept 2001
Externally published	Yes

Keywords

Approximate subdomain solution
Domain decomposition
Incompressible Navier-Stokes equations
Orthogonalization methods
Parallel Krylov subspace methods

Access to Document

10.1016/S0167-739X(00)00073-X

Cite this

@article{5b26eec09ca34225b78675276e64f258,

title = "A parallel block-preconditioned GCR method for incompressible flow problems",

abstract = "Efficient parallel algorithms are required to simulate incompressible turbulent flows in complex two- and three-dimensional domains. The incompressible Navier-Stokes equations are discretized in general coordinates on a structured grid. For a flow on a general domain we use an unstructured decomposition of the domain into subdomains of simple shape, with a structured grid inside each subdomain. We have developed a parallel block-preconditioned GCR method to solve the resulting systems of linear equations. The method can be smoothly varied between a coarse grain parallel method in which the subdomain problems are solved accurately with an inner iteration process and a fine grain parallel method when only a preconditioner is used to approximate the solution on the blocks. Parallel performance results for Boussinesq flow in a cavity are included.",

keywords = "Approximate subdomain solution, Domain decomposition, Incompressible Navier-Stokes equations, Orthogonalization methods, Parallel Krylov subspace methods",

author = "C. Vuik and J. Frank and A. Segal",

year = "2001",

month = sep,

doi = "10.1016/S0167-739X(00)00073-X",

language = "English",

volume = "18",

pages = "31--40",

journal = "Future Generation Computer Systems",

issn = "0167-739X",

publisher = "Elsevier BV",

number = "1",

}

TY - JOUR

T1 - A parallel block-preconditioned GCR method for incompressible flow problems

AU - Vuik, C.

AU - Frank, J.

AU - Segal, A.

PY - 2001/9

Y1 - 2001/9

N2 - Efficient parallel algorithms are required to simulate incompressible turbulent flows in complex two- and three-dimensional domains. The incompressible Navier-Stokes equations are discretized in general coordinates on a structured grid. For a flow on a general domain we use an unstructured decomposition of the domain into subdomains of simple shape, with a structured grid inside each subdomain. We have developed a parallel block-preconditioned GCR method to solve the resulting systems of linear equations. The method can be smoothly varied between a coarse grain parallel method in which the subdomain problems are solved accurately with an inner iteration process and a fine grain parallel method when only a preconditioner is used to approximate the solution on the blocks. Parallel performance results for Boussinesq flow in a cavity are included.

AB - Efficient parallel algorithms are required to simulate incompressible turbulent flows in complex two- and three-dimensional domains. The incompressible Navier-Stokes equations are discretized in general coordinates on a structured grid. For a flow on a general domain we use an unstructured decomposition of the domain into subdomains of simple shape, with a structured grid inside each subdomain. We have developed a parallel block-preconditioned GCR method to solve the resulting systems of linear equations. The method can be smoothly varied between a coarse grain parallel method in which the subdomain problems are solved accurately with an inner iteration process and a fine grain parallel method when only a preconditioner is used to approximate the solution on the blocks. Parallel performance results for Boussinesq flow in a cavity are included.

KW - Approximate subdomain solution

KW - Domain decomposition

KW - Incompressible Navier-Stokes equations

KW - Orthogonalization methods

KW - Parallel Krylov subspace methods

UR - https://www.scopus.com/pages/publications/0035450394

U2 - 10.1016/S0167-739X(00)00073-X

DO - 10.1016/S0167-739X(00)00073-X

M3 - Article

AN - SCOPUS:0035450394

SN - 0167-739X

VL - 18

SP - 31

EP - 40

JO - Future Generation Computer Systems

JF - Future Generation Computer Systems

IS - 1

ER -

A parallel block-preconditioned GCR method for incompressible flow problems

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this