Implications of memory performance for highly efficient supercomputing of scientific applications

Akihiro Musa, Hiroyuki Takizawa, Koki Okabe, Takashi Soga, Hiroaki Kobayashi

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

3 Citations (Scopus)

Abstract

This paper examines the memory performance of the vectorparallel and scalar-parallel computing platforms across five applications of three scientific areas; electromagnetic analysis, CFD/heat analysis, and seismology. Our evaluation results show that the vector platforms can achieve the high computational efficiency and hence significantly outperform the scalar platforms in the areas of these applications. We did exhaustive experiments and quantitatively evaluated representative scalar and vector platforms using real applications from the viewpoint of the system designers and developers. These results demonstrate that the ratio of memory bandwidth to floating-point operation rate needs to reach 4-bytes/flop to preserve the computational performance with hiding the memory access latencies by pipelined vector operations in the vector platforms. We also confirm that the enough number of memory banks to handle stride memory accesses leads to an increase in the execution efficiency. On the scalar platforms, the cache hit rate needs to be almost 100% to achieve the high computational efficiency.

Original languageEnglish
Title of host publicationParallel and Distributed Processing and Applications - 4th International Symposium, ISPA 2006, Proceedings
EditorsFeilong Tang, Minyi Guo, Beniamino Di Martino, Hans P. Zima, Hans P. Zima, Laurence T Yang, Jack Dongarra
PublisherSpringer-Verlag
Pages845-858
Number of pages14
ISBN (Print)9783540680673
Publication statusPublished - 2006 Jan 1
Event4th International Symposium on Parallel and Distributed Processing and Applications, ISPA 2006 - Sorrento, Italy
Duration: 2006 Dec 42006 Dec 6

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4330
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other4th International Symposium on Parallel and Distributed Processing and Applications, ISPA 2006
CountryItaly
CitySorrento
Period06/12/406/12/6

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Implications of memory performance for highly efficient supercomputing of scientific applications'. Together they form a unique fingerprint.

Cite this