Modeling communication overhead: MPI and MPL performance on the IBM SP2

被引：40

作者：

Xu, ZW ^{[1
]}

Hwang, K ^{[1
]}

机构：

[1] UNIV HONG KONG,DEPT ELECT & ELECTR ENGN,HONG KONG,HONG KONG

来源：

IEEE PARALLEL & DISTRIBUTED TECHNOLOGY | 1996年 / 4卷 / 01期

关键词：

D O I：

10.1109/88.481662

中图分类号：

TP301 [理论、方法];

学科分类号：

081202 ;

摘要：

The authors use timing experiments on the IBM SP2 to develop an overhead-quantifying method for evaluating communication performance on message-passing multicomputers. MPP designers and users can apply this method to reveal architectural bottlenecks, and to trade off between computations and communications for parallel applications for parallel applications optimization.

引用

页码：9 / 23

页数：15

共 14 条

[1]

ABELI B, E COMMUNICATION

[2]

ALEXANDROV A, 1995, 7 ANN ACM S PAR ALG, P95

[3] CCL - A PORTABLE AND TUNABLE COLLECTIVE COMMUNICATION LIBRARY FOR SCALABLE PARALLEL COMPUTERS [J].

BALA, V ;

BRUCK, J ;

CYPHER, R ;

ELUSTONDO, P ;

HO, A ;

HO, CT ;

KIPNIS, S ;

SNIR, M .

IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, 1995, 6 (02) :154-164

[4]

CULLER D, 1993, 5TH ACM SIGPLAN S PR, P1

[5]

FRANKE H, 1993, 2ND P WORKSH SCAL PA, P39

[6]

Geist A, 1994, PVM PARALLEL VIRTUAL

[7]

HOCKNEY RW, 1994, PARALLEL COMPUT, V20, P389, DOI 10.1016/0167-8191(94)90095-7

[8]

HWANG K, 1996, IN PRESS IEEE T PARA

[9]

Hwang K., 1993, Advanced Computer Architecture: Parallelism. Scalability

[10]

Mraz R., 1994, Proceedings Supercomputing '94 (Cat. No.94CH34819), P620, DOI 10.1109/SUPERC.1994.344326

← 1 2 →