Efficient parabolic solvers scalable across multi-architectural levels

Yu Zhuang, Heng Wu

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

High end computing hardware has been growing fast in both uniprocessor performance and parallel system scales. Steadily advancing but somewhat lagging behind is the speed of memory accesses. Thus, needed are software and algorithms behind software that adapt well with architectural features of high end computing hardware. Stable explicit implicit domain decomposition (SEIDD) is a class of numerical algorithms originally introduced for solving parabolic equations on parallel computers, which has adequately high parallelism, flexible controllability for load balancing, minimal communication cost, and good stability and efficiency. In this paper, we study the effectiveness of SEIDD in harnessing the computing power at the inter-processor level for parallel processing as well as the level of cache memories for fast memory accesses.

Original languageEnglish
Title of host publicationProceedings of the 2012 10th IEEE International Symposium on Parallel and Distributed Processing with Applications, ISPA 2012
Pages111-118
Number of pages8
DOIs
StatePublished - 2012
Event2012 10th IEEE International Symposium on Parallel and Distributed Processing with Applications, ISPA 2012 - Madrid, Spain
Duration: Jul 10 2012Jul 13 2012

Publication series

NameProceedings of the 2012 10th IEEE International Symposium on Parallel and Distributed Processing with Applications, ISPA 2012

Conference

Conference2012 10th IEEE International Symposium on Parallel and Distributed Processing with Applications, ISPA 2012
Country/TerritorySpain
CityMadrid
Period07/10/1207/13/12

Keywords

  • Parallel algorithm
  • cache performance
  • numerical solution
  • partial differential equation

Fingerprint

Dive into the research topics of 'Efficient parabolic solvers scalable across multi-architectural levels'. Together they form a unique fingerprint.

Cite this