Performance Implications of Nonuniform Device Topologies in Scalable Heterogeneous Architectures

Author

Meredith, Jeremy S. ; Roth, Philip C. ; Spafford, Kyle L. ; Vetter, Jeffrey S.

Author_Institution

Oak Ridge Nat. Lab., Oak Ridge, TN, USA

Volume

31

Issue

5

fYear

2011

Firstpage

66

Lastpage

75

Abstract

This article considers trends in heterogeneous system design, particularly for GPUs. Using the Keeneland Initial Delivery System, the authors examine the performance implications of increased parallelism and specialized hardware on parallel scientific applications. They examine how nonuniform data-transfer performance across the node-level topology can impact performance. Finally, they help users of GPU-based systems avoid performance problems related to this nonuniformity.

Keywords

coprocessors; multiprocessing systems; natural sciences computing; parallel architectures; GPU; Keeneland initial delivery system; heterogeneous system design; node-level topology; nonuniform data-transfer performance; nonuniform device topology; parallel scientific application; scalable heterogeneous architecture; Bandwidth; Computer architecture; Graphics processing unit; Nework topology; Performance evaluation; Programming; Random access memory; GPU; data-transfer performance; heterogeneous GPUs; nonuniformity;

fLanguage

English

Journal_Title

Micro, IEEE

Publisher

ieee

ISSN

0272-1732

Type

jour

DOI

10.1109/MM.2011.79

Filename

5989784