Network partitioning of data parallel computations

Jon B. Weissman, Andrew S. Grimshaw

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Scopus citations

Abstract

Partitioning data parallel computations across a network of heterogeneous workstations is a difficult problem for the user. We have developed a runtime partitioning method for choosing the number and type of processors to apply to a data parallel computation, and a decomposition of the data domain in order to achieve reduced completion time. The partitioning method utilizes information about the problem in the form of callback functions and uses a set of topology-specific communication functions to estimate communication costs. We show that the method makes effective partitioning decisions and has runtime overhead that is easily tolerated. In particular, we show that for two implementations of a canonical stencil application, minimum elapsed times are obtained for a range of problem sizes on a network of heterogeneous workstations.

Original languageEnglish (US)
Title of host publicationProceedings of the 3rd IEEE International Symposium on High Performance Distributed Computing
Pages149-156
Number of pages8
StatePublished - Dec 1 1994
EventProceedings of the 3rd IEEE International Symposium on High Performance Distributed Computing - San Francisco, CA, USA
Duration: Apr 2 1994Apr 5 1994

Other

OtherProceedings of the 3rd IEEE International Symposium on High Performance Distributed Computing
CitySan Francisco, CA, USA
Period4/2/944/5/94

Fingerprint Dive into the research topics of 'Network partitioning of data parallel computations'. Together they form a unique fingerprint.

Cite this