2003
DOI: 10.1007/978-3-540-24596-4_41
|View full text |Cite
|
Sign up to set email alerts
|

Performance Analysis of a Hybrid Overset Multi-block Application on Multiple Architectures

Abstract: This paper presents a detailed performance analysis of a multi-block overset grid computational fluid dynamics app!ication on multiple state-of-the-art computer architectures. The application is implemented using a hybrid MPI+OpenMP programming paradigm that exploits both coarse and fine-grain parallelism; the former via MPI message passing and the latter via OpenMP directives. The hybrid model also extends the applicability of multi-block programs to large clusters of SNIP nodes by overcoming the restriction … Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
4
1

Citation Types

0
5
0

Year Published

2006
2006
2020
2020

Publication Types

Select...
2
2

Relationship

1
3

Authors

Journals

citations
Cited by 4 publications
(5 citation statements)
references
References 5 publications
0
5
0
Order By: Relevance
“…Scalability on Seaborg exceeded all others, with computational efficiency decreasing for a larger number of MPI tasks primarily due to load imbalance. A hybrid MPI+OpenMP implementation [9] showed similar performance as a pure MPI approach on all systems except for Seaborg where the multilevel results were significantly better. Note that adding more OpenMP threads beyond an optimal number, depending on the number of MPI processes, does not improve performance.…”
Section: Fluid Dynamicsmentioning
confidence: 84%
“…Scalability on Seaborg exceeded all others, with computational efficiency decreasing for a larger number of MPI tasks primarily due to load imbalance. A hybrid MPI+OpenMP implementation [9] showed similar performance as a pure MPI approach on all systems except for Seaborg where the multilevel results were significantly better. Note that adding more OpenMP threads beyond an optimal number, depending on the number of MPI processes, does not improve performance.…”
Section: Fluid Dynamicsmentioning
confidence: 84%
“…The hybrid MPI+OpenMP version of OVERFLOW-D takes advantage of the overset grid system, which offers a natural coarse-grain parallelism [5]. A binpacking algorithm clusters individual grids into groups, each of which is then assigned to an MPI process.…”
Section: Overflow-d: Rotor Vortex Simulationsmentioning
confidence: 99%
“…Because Columbia is a cache-based superscalar architecture, modifications were necessary to improve performance. The linear solver of the application, called LU-SGS, was reimplemented using a pipeline algorithm [5] to enhance efficiency which is dictated by the type of data dependencies inherent in the solution algorithm.…”
Section: Overflow-d: Rotor Vortex Simulationsmentioning
confidence: 99%
“…The hybrid MPI+OpenMP version of OVERFLOW-D takes advantage of the overset grid system, which offers a natural coarse-grain parallelism. 8 A bin-packing algorithm clusters individual grids into groups, each of which is then assigned to an MPI process. The grouping strategy uses a connectivity test that inspects for an overlap between a pair of grids before assigning them to the same group, regardless of the size of the boundary data or their connectivity to other grids.…”
Section: Overflow-d: Rotor Vortex Simulationsmentioning
confidence: 99%
“…Because Columbia is a cache-based superscalar architecture, modifications were necessary to improve performance. The linear solver of the application, called LU-SGS, was re-implemented using a pipeline algorithm 8 to enhance efficiency which is dictated by the type of data dependencies inherent in the solution algorithm.…”
Section: Overflow-d: Rotor Vortex Simulationsmentioning
confidence: 99%