A global address space framework for irregular applications

Authors:
Kentaro Hara;Kenjiro Taura
Affiliations:
The University of Tokyo;The University of Tokyo
Venue:
Proceedings of the 19th ACM International Symposium on High Performance Distributed Computing
Year:
2010

Citing 2
Cited 0

Advances, Applications and Performance of the Global Arrays Shared Memory Programming Toolkit

International Journal of High Performance Computing Applications
Titanium performance and potential: an NPB experimental study

LCPC'05 Proceedings of the 18th international conference on Languages and Compilers for Parallel Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Practical parallel scientific applications with domain decompositions, such as finite element methods, require irregular domain decompositions of complicated-shaped objects. However, existing PGAS frameworks, such as Global Arrays and XcalableMP, have supported the productive description of exchanging ghost points only for regular domain decompositions. With these backgrounds, we propose, implement and evaluate Distributed Memory Interface (DMI), a global address space framework for irregular applications. DMI provides highly productive APIs called read-write-set for irregular domain decompositions and complicated orderings in practical scientific applications.