1993
DOI: 10.1142/s0129053393000025
|View full text |Cite
|
Sign up to set email alerts
|

Parallelization of a Class of Implicit Finite Difference Schemes in Computational Fluid Dynamics

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
35
0
1

Year Published

1996
1996
2019
2019

Publication Types

Select...
6
2

Relationship

0
8

Authors

Journals

citations
Cited by 62 publications
(37 citation statements)
references
References 0 publications
0
35
0
1
Order By: Relevance
“…However, when such a method was used, strong filtering was needed in order to maintain computational stability, and as a result, the resolution of the calculation in the entire region was severely degraded. Therefore, in the present study, a pipelining algorithm (22) , in which sweep is performed over all processes in each direction, is adopted to solve the band matrices.…”
Section: Governing Equations and Numerical Methodsmentioning
confidence: 99%
“…However, when such a method was used, strong filtering was needed in order to maintain computational stability, and as a result, the resolution of the calculation in the entire region was severely degraded. Therefore, in the present study, a pipelining algorithm (22) , in which sweep is performed over all processes in each direction, is adopted to solve the band matrices.…”
Section: Governing Equations and Numerical Methodsmentioning
confidence: 99%
“…Alternating Direction Implicit (ADI) integration is a common technique for solving partial differential equations that uses this solution style [NNN93]. Two of the NAS parallel benchmarks [BHS+95], SP and BT, use ADI integration to solve the Navier-Stokes equation in three dimensions.…”
Section: Compiler-support For Advanced Partitioningsmentioning
confidence: 99%
“…Two of the NAS parallel benchmarks [BHS+95], SP and BT, use ADI integration to solve the Navier-Stokes equation in three dimensions. Fractional step methods and other solution techniques that use line sweeps are described by Naik et al [NNN93]. For this class of computations, applying a standard block partitioning to any of the spatial dimensions is problematic-recurrences along the partitioned dimension partially serialize execution.…”
Section: Compiler-support For Advanced Partitioningsmentioning
confidence: 99%
“…Finally, the dHPF compiler provides novel compiler support for multipartitioning [14,15], a family of sophisticated skewed-cyclic block distributions that were originally developed for hand-coded parallelization of tightly-coupled multi-dimensional line-sweep computations [8,9,11]. Multipartitioning enables decomposition of arrays of d ≥ 2 dimensions among a set of processors so that for a line sweep computation along any dimension of an array, all processors are active in each step of the computation, load-balance is nearly perfect, and only coarse-grain communication is needed.…”
Section: The Dhpf Compilermentioning
confidence: 99%
“…Line sweep computations are used to solve one-dimensional recurrences along a dimension of a multi-dimensional hyper-rectangular domain. This strategy is at the heart of a variety of numerical methods and solution techniques [8]. SP and BT employ line sweeps to perform Alternating Direction Implicit (ADI) integration-a widely-used numerical technique for solving partial differential equations such as the Navier-Stokes equation [8,9,10,11].…”
Section: Introductionmentioning
confidence: 99%