Reducing coherence overhead in shared-bus multiprocessors

Sangyeun Cho, Kyung Ho Lee

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

1 Citation (Scopus)

Abstract

To reduce the overhead of cache coherence enforcement in shared-bus multiprocessors, we propose a self-invalidation technique as an extension to write-invalidate protocols. The technique speculatively identifies cache blocks to be invalidated and dynamically determines when to invalidate them locally. We also consider enhancing our selfinvalidation scheme by incorporating read snarfing, to reduce the cache misses due to incorrect prediction. We evaluate our self-invalidation scheme by simulating SPLASH-2 benchmark programs that exhibit various reference patterns, under a realistic shared-bus multiprocessor model. We discuss the effectiveness and hardware complexity of self-invalidation and its enhancement with read snarfing in our extended protocol.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages492-497
Number of pages6
Volume1124
ISBN (Print)3540616276, 9783540616276
Publication statusPublished - 1996
Externally publishedYes
Event2nd International Euro-Par Conference on Parallel Processing, Euro-Par 1996 - Lyon, France
Duration: 1996 Aug 261996 Aug 29

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume1124
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other2nd International Euro-Par Conference on Parallel Processing, Euro-Par 1996
CountryFrance
CityLyon
Period96/8/2696/8/29

Fingerprint

Multiprocessor
Cache
Cache Coherence
Enhancement
Hardware
Benchmark
Evaluate
Prediction
Model

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Cho, S., & Lee, K. H. (1996). Reducing coherence overhead in shared-bus multiprocessors. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1124, pp. 492-497). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1124). Springer Verlag.

Reducing coherence overhead in shared-bus multiprocessors. / Cho, Sangyeun; Lee, Kyung Ho.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1124 Springer Verlag, 1996. p. 492-497 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1124).

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

Cho, S & Lee, KH 1996, Reducing coherence overhead in shared-bus multiprocessors. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 1124, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1124, Springer Verlag, pp. 492-497, 2nd International Euro-Par Conference on Parallel Processing, Euro-Par 1996, Lyon, France, 96/8/26.
Cho S, Lee KH. Reducing coherence overhead in shared-bus multiprocessors. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1124. Springer Verlag. 1996. p. 492-497. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Cho, Sangyeun ; Lee, Kyung Ho. / Reducing coherence overhead in shared-bus multiprocessors. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1124 Springer Verlag, 1996. pp. 492-497 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{3441c2b47e214610bc306481bcdacdd4,
title = "Reducing coherence overhead in shared-bus multiprocessors",
abstract = "To reduce the overhead of cache coherence enforcement in shared-bus multiprocessors, we propose a self-invalidation technique as an extension to write-invalidate protocols. The technique speculatively identifies cache blocks to be invalidated and dynamically determines when to invalidate them locally. We also consider enhancing our selfinvalidation scheme by incorporating read snarfing, to reduce the cache misses due to incorrect prediction. We evaluate our self-invalidation scheme by simulating SPLASH-2 benchmark programs that exhibit various reference patterns, under a realistic shared-bus multiprocessor model. We discuss the effectiveness and hardware complexity of self-invalidation and its enhancement with read snarfing in our extended protocol.",
author = "Sangyeun Cho and Lee, {Kyung Ho}",
year = "1996",
language = "English",
isbn = "3540616276",
volume = "1124",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "492--497",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Reducing coherence overhead in shared-bus multiprocessors

AU - Cho, Sangyeun

AU - Lee, Kyung Ho

PY - 1996

Y1 - 1996

N2 - To reduce the overhead of cache coherence enforcement in shared-bus multiprocessors, we propose a self-invalidation technique as an extension to write-invalidate protocols. The technique speculatively identifies cache blocks to be invalidated and dynamically determines when to invalidate them locally. We also consider enhancing our selfinvalidation scheme by incorporating read snarfing, to reduce the cache misses due to incorrect prediction. We evaluate our self-invalidation scheme by simulating SPLASH-2 benchmark programs that exhibit various reference patterns, under a realistic shared-bus multiprocessor model. We discuss the effectiveness and hardware complexity of self-invalidation and its enhancement with read snarfing in our extended protocol.

AB - To reduce the overhead of cache coherence enforcement in shared-bus multiprocessors, we propose a self-invalidation technique as an extension to write-invalidate protocols. The technique speculatively identifies cache blocks to be invalidated and dynamically determines when to invalidate them locally. We also consider enhancing our selfinvalidation scheme by incorporating read snarfing, to reduce the cache misses due to incorrect prediction. We evaluate our self-invalidation scheme by simulating SPLASH-2 benchmark programs that exhibit various reference patterns, under a realistic shared-bus multiprocessor model. We discuss the effectiveness and hardware complexity of self-invalidation and its enhancement with read snarfing in our extended protocol.

UR - http://www.scopus.com/inward/record.url?scp=84947944324&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84947944324&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84947944324

SN - 3540616276

SN - 9783540616276

VL - 1124

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 492

EP - 497

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -