Design and implementation of NX message passing using Shrimp virtual memory mapped communication

R. Alpert; C. Dubnicki; Edward William Felten; Kai Li

doi:10.1109/ICPP.1996.537151

Design and implementation of NX message passing using Shrimp virtual memory mapped communication

R. Alpert, C. Dubnicki, Edward William Felten, Kai Li

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

7 Scopus citations

Abstract

This paper describes the design, implementation and performance of the NX message-passing interface on the Shrimp multicomputer. Unlike traditional methods, our implementation, exploiting Shrimp's virtual memory-mapped communication facility, performs buffer management at user level without using a special message-passing processor, and requires no CPU intervention upon message arrival in the common cases. For a four-byte message, our implementation, achieves a user-To-user latency of 12 microseconds, about factor of four smaller than that on the Intel Paragon. For large messages, our implementation quickly approaches the bandwidth limit imposed by the Shrimp hardware.

Original language	English (US)
Title of host publication	Architecture
Editors	A. Reeves
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	111-119
Number of pages	9
ISBN (Electronic)	081867623X
DOIs	https://doi.org/10.1109/ICPP.1996.537151
State	Published - 1996
Event	25th International Conference on Parallel Processing, ICPP 1996 - Ithaca, United States Duration: Aug 12 1996 → Aug 16 1996

Publication series

Name	Proceedings of the International Conference on Parallel Processing
Volume	1
ISSN (Print)	0190-3918

Other

Other	25th International Conference on Parallel Processing, ICPP 1996
Country/Territory	United States
City	Ithaca
Period	8/12/96 → 8/16/96

All Science Journal Classification (ASJC) codes

Software
General Mathematics
Hardware and Architecture

Access to Document

10.1109/ICPP.1996.537151

Cite this

Alpert, R., Dubnicki, C., Felten, E. W., & Li, K. (1996). Design and implementation of NX message passing using Shrimp virtual memory mapped communication. In A. Reeves (Ed.), Architecture (pp. 111-119). Article 537151 (Proceedings of the International Conference on Parallel Processing; Vol. 1). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICPP.1996.537151

@inproceedings{220268a6805d4b459e24220e6a4cda0f,

title = "Design and implementation of NX message passing using Shrimp virtual memory mapped communication",

abstract = "This paper describes the design, implementation and performance of the NX message-passing interface on the Shrimp multicomputer. Unlike traditional methods, our implementation, exploiting Shrimp's virtual memory-mapped communication facility, performs buffer management at user level without using a special message-passing processor, and requires no CPU intervention upon message arrival in the common cases. For a four-byte message, our implementation, achieves a user-To-user latency of 12 microseconds, about factor of four smaller than that on the Intel Paragon. For large messages, our implementation quickly approaches the bandwidth limit imposed by the Shrimp hardware.",

author = "R. Alpert and C. Dubnicki and Felten, {Edward William} and Kai Li",

note = "Publisher Copyright: {\textcopyright} 1996 IEEE.; 25th International Conference on Parallel Processing, ICPP 1996 ; Conference date: 12-08-1996 Through 16-08-1996",

year = "1996",

doi = "10.1109/ICPP.1996.537151",

language = "English (US)",

series = "Proceedings of the International Conference on Parallel Processing",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "111--119",

editor = "A. Reeves",

booktitle = "Architecture",

address = "United States",

}

Alpert, R, Dubnicki, C, Felten, EW & Li, K 1996, Design and implementation of NX message passing using Shrimp virtual memory mapped communication. in A Reeves (ed.), Architecture., 537151, Proceedings of the International Conference on Parallel Processing, vol. 1, Institute of Electrical and Electronics Engineers Inc., pp. 111-119, 25th International Conference on Parallel Processing, ICPP 1996, Ithaca, United States, 8/12/96. https://doi.org/10.1109/ICPP.1996.537151

Design and implementation of NX message passing using Shrimp virtual memory mapped communication. / Alpert, R.; Dubnicki, C.; Felten, Edward William et al.
Architecture. ed. / A. Reeves. Institute of Electrical and Electronics Engineers Inc., 1996. p. 111-119 537151 (Proceedings of the International Conference on Parallel Processing; Vol. 1).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Design and implementation of NX message passing using Shrimp virtual memory mapped communication

AU - Alpert, R.

AU - Dubnicki, C.

AU - Felten, Edward William

AU - Li, Kai

PY - 1996

Y1 - 1996

N2 - This paper describes the design, implementation and performance of the NX message-passing interface on the Shrimp multicomputer. Unlike traditional methods, our implementation, exploiting Shrimp's virtual memory-mapped communication facility, performs buffer management at user level without using a special message-passing processor, and requires no CPU intervention upon message arrival in the common cases. For a four-byte message, our implementation, achieves a user-To-user latency of 12 microseconds, about factor of four smaller than that on the Intel Paragon. For large messages, our implementation quickly approaches the bandwidth limit imposed by the Shrimp hardware.

AB - This paper describes the design, implementation and performance of the NX message-passing interface on the Shrimp multicomputer. Unlike traditional methods, our implementation, exploiting Shrimp's virtual memory-mapped communication facility, performs buffer management at user level without using a special message-passing processor, and requires no CPU intervention upon message arrival in the common cases. For a four-byte message, our implementation, achieves a user-To-user latency of 12 microseconds, about factor of four smaller than that on the Intel Paragon. For large messages, our implementation quickly approaches the bandwidth limit imposed by the Shrimp hardware.

UR - http://www.scopus.com/inward/record.url?scp=84975293946&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84975293946&partnerID=8YFLogxK

U2 - 10.1109/ICPP.1996.537151

DO - 10.1109/ICPP.1996.537151

M3 - Conference contribution

AN - SCOPUS:84975293946

T3 - Proceedings of the International Conference on Parallel Processing

SP - 111

EP - 119

BT - Architecture

A2 - Reeves, A.

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 25th International Conference on Parallel Processing, ICPP 1996

Y2 - 12 August 1996 through 16 August 1996

ER -

Design and implementation of NX message passing using Shrimp virtual memory mapped communication

Abstract

Publication series

Other

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this