Experience with fine-grain communication in EM-X multiprocessor for parallel sparse matrix computation

Mitsuhisa Sato, Yuetsu Kodama, Hirofumi Sakane, Hayato Yamana, Shuichi Sakai, Yoshinori Yamaguchi

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

1 Citation (Scopus)

Abstract

Sparse matrix problems require a communication paradigm different from those used in conventional distributed-memory multiprocessors. We present in this paper how fine-grain communication can help obtain high performance in the experimental distributed-memory multiprocessor, EM-X, developed at ETL, which can handle fine-grain communication very efficiently. The sparse matrix kernel, Conjugate Gradient, is selected for the experiments. Among the steps in CG is the sparse matrix vector multiplications we focus on in the study. Some communication methods are developed for performance comparison, including coarse-grain and fine-grain implementations. Fine-grain communication allows exact data access in an unstructured problem to reduce the amount of communication. While CG presents bottlenecks in terms of a large number of fine-grain remote reads, the multithreaded principles of execution is so designed to tolerate such latency. Experimental results indicate that the performance of fine-grain read implementation is comparable to that of coarse-grain implementation on 64 processors. The results demonstrate that fine-grain communication can be a viable and efficient approach to unstructured sparse matrix problems on large-scale distributed-memory multiprocessors.

Original languageEnglish
Title of host publicationProceedings of the International Parallel Processing Symposium, IPPS
Editors Anon
PublisherIEEE
Pages242-248
Number of pages7
Publication statusPublished - 1997
Externally publishedYes
EventProceedings of the 1997 11th International Parallel Processing Symposium, IPPS 97 - Geneva, Switz
Duration: 1997 Apr 11997 Apr 5

Other

OtherProceedings of the 1997 11th International Parallel Processing Symposium, IPPS 97
CityGeneva, Switz
Period97/4/197/4/5

Fingerprint

Communication
Data storage equipment
Experiments

ASJC Scopus subject areas

  • Hardware and Architecture

Cite this

Sato, M., Kodama, Y., Sakane, H., Yamana, H., Sakai, S., & Yamaguchi, Y. (1997). Experience with fine-grain communication in EM-X multiprocessor for parallel sparse matrix computation. In Anon (Ed.), Proceedings of the International Parallel Processing Symposium, IPPS (pp. 242-248). IEEE.

Experience with fine-grain communication in EM-X multiprocessor for parallel sparse matrix computation. / Sato, Mitsuhisa; Kodama, Yuetsu; Sakane, Hirofumi; Yamana, Hayato; Sakai, Shuichi; Yamaguchi, Yoshinori.

Proceedings of the International Parallel Processing Symposium, IPPS. ed. / Anon. IEEE, 1997. p. 242-248.

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

Sato, M, Kodama, Y, Sakane, H, Yamana, H, Sakai, S & Yamaguchi, Y 1997, Experience with fine-grain communication in EM-X multiprocessor for parallel sparse matrix computation. in Anon (ed.), Proceedings of the International Parallel Processing Symposium, IPPS. IEEE, pp. 242-248, Proceedings of the 1997 11th International Parallel Processing Symposium, IPPS 97, Geneva, Switz, 97/4/1.
Sato M, Kodama Y, Sakane H, Yamana H, Sakai S, Yamaguchi Y. Experience with fine-grain communication in EM-X multiprocessor for parallel sparse matrix computation. In Anon, editor, Proceedings of the International Parallel Processing Symposium, IPPS. IEEE. 1997. p. 242-248
Sato, Mitsuhisa ; Kodama, Yuetsu ; Sakane, Hirofumi ; Yamana, Hayato ; Sakai, Shuichi ; Yamaguchi, Yoshinori. / Experience with fine-grain communication in EM-X multiprocessor for parallel sparse matrix computation. Proceedings of the International Parallel Processing Symposium, IPPS. editor / Anon. IEEE, 1997. pp. 242-248
@inproceedings{ab81d9c6b39f4753a1166bd77525d3a4,
title = "Experience with fine-grain communication in EM-X multiprocessor for parallel sparse matrix computation",
abstract = "Sparse matrix problems require a communication paradigm different from those used in conventional distributed-memory multiprocessors. We present in this paper how fine-grain communication can help obtain high performance in the experimental distributed-memory multiprocessor, EM-X, developed at ETL, which can handle fine-grain communication very efficiently. The sparse matrix kernel, Conjugate Gradient, is selected for the experiments. Among the steps in CG is the sparse matrix vector multiplications we focus on in the study. Some communication methods are developed for performance comparison, including coarse-grain and fine-grain implementations. Fine-grain communication allows exact data access in an unstructured problem to reduce the amount of communication. While CG presents bottlenecks in terms of a large number of fine-grain remote reads, the multithreaded principles of execution is so designed to tolerate such latency. Experimental results indicate that the performance of fine-grain read implementation is comparable to that of coarse-grain implementation on 64 processors. The results demonstrate that fine-grain communication can be a viable and efficient approach to unstructured sparse matrix problems on large-scale distributed-memory multiprocessors.",
author = "Mitsuhisa Sato and Yuetsu Kodama and Hirofumi Sakane and Hayato Yamana and Shuichi Sakai and Yoshinori Yamaguchi",
year = "1997",
language = "English",
pages = "242--248",
editor = "Anon",
booktitle = "Proceedings of the International Parallel Processing Symposium, IPPS",
publisher = "IEEE",

}

TY - GEN

T1 - Experience with fine-grain communication in EM-X multiprocessor for parallel sparse matrix computation

AU - Sato, Mitsuhisa

AU - Kodama, Yuetsu

AU - Sakane, Hirofumi

AU - Yamana, Hayato

AU - Sakai, Shuichi

AU - Yamaguchi, Yoshinori

PY - 1997

Y1 - 1997

N2 - Sparse matrix problems require a communication paradigm different from those used in conventional distributed-memory multiprocessors. We present in this paper how fine-grain communication can help obtain high performance in the experimental distributed-memory multiprocessor, EM-X, developed at ETL, which can handle fine-grain communication very efficiently. The sparse matrix kernel, Conjugate Gradient, is selected for the experiments. Among the steps in CG is the sparse matrix vector multiplications we focus on in the study. Some communication methods are developed for performance comparison, including coarse-grain and fine-grain implementations. Fine-grain communication allows exact data access in an unstructured problem to reduce the amount of communication. While CG presents bottlenecks in terms of a large number of fine-grain remote reads, the multithreaded principles of execution is so designed to tolerate such latency. Experimental results indicate that the performance of fine-grain read implementation is comparable to that of coarse-grain implementation on 64 processors. The results demonstrate that fine-grain communication can be a viable and efficient approach to unstructured sparse matrix problems on large-scale distributed-memory multiprocessors.

AB - Sparse matrix problems require a communication paradigm different from those used in conventional distributed-memory multiprocessors. We present in this paper how fine-grain communication can help obtain high performance in the experimental distributed-memory multiprocessor, EM-X, developed at ETL, which can handle fine-grain communication very efficiently. The sparse matrix kernel, Conjugate Gradient, is selected for the experiments. Among the steps in CG is the sparse matrix vector multiplications we focus on in the study. Some communication methods are developed for performance comparison, including coarse-grain and fine-grain implementations. Fine-grain communication allows exact data access in an unstructured problem to reduce the amount of communication. While CG presents bottlenecks in terms of a large number of fine-grain remote reads, the multithreaded principles of execution is so designed to tolerate such latency. Experimental results indicate that the performance of fine-grain read implementation is comparable to that of coarse-grain implementation on 64 processors. The results demonstrate that fine-grain communication can be a viable and efficient approach to unstructured sparse matrix problems on large-scale distributed-memory multiprocessors.

UR - http://www.scopus.com/inward/record.url?scp=0030646771&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0030646771&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0030646771

SP - 242

EP - 248

BT - Proceedings of the International Parallel Processing Symposium, IPPS

A2 - Anon, null

PB - IEEE

ER -