Parallelization of large scale ocean models by data decomposition

H. -P Kersken; B Fritzsch; O Schenk; W Hiller; J Behrens; E Krauße

doi:10.1007/BFb0020393

Back

Book chapter

Parallelization of large scale ocean models by data decomposition

H. -P Kersken, B Fritzsch, O Schenk, W Hiller, J Behrens and E Krauße

High-Performance Computing and Networking, pp.323-328

Lecture Notes in Computer Science, Springer Berlin Heidelberg

06/10/2005

DOI: 10.1007/BFb0020393

View Online

Abstract

There exist two widely used types of models for the investigation of flows in the world oceans: quasi-geostrophic (QG) models and models based on the primitive equations of motion (PE-models). Starting from existing sequential programs we investigate different strategies for their parallelization. Implementation strategies for the QG-model on a multiprocessor with shared memory and a few nodes are discussed. For the PE-model we describe an implementation on a distributed memory machine with special emphasis on the load balancing issue and communication costs.

Primitive Equation

Friction Term

Shared Memory

Elliptic Problem

Helmholtz Equation

Details

Title: Subtitle: Parallelization of large scale ocean models by data decomposition
Creators: H. -P Kersken
B Fritzsch
O Schenk
W Hiller
J Behrens
E Krauße
Resource Type: Book chapter
Publication Details: High-Performance Computing and Networking, pp.323-328
Series: Lecture Notes in Computer Science
DOI: 10.1007/BFb0020393
eISSN: 1611-3349
ISSN: 0302-9743
Publisher: Springer Berlin Heidelberg; Berlin, Heidelberg
Language: English
Date published: 06/10/2005
Academic Unit: Iowa Neuroscience Institute; Biology; Craniofacial Anomalies Research Center
Record Identifier: 9984070900202771

Metrics

21 Record Views