The SPEED cache coherence protocol for an optical multi-access interconnect architecture

Authors:
Joon-Ho Ha;T. M. Pinkston
Affiliations:
-;-
Venue:
MPPOI '95 Proceedings of the Second Workshop on Massively Parallel Processing Using Optical Interconnections
Year:
1995

Citing 0
Cited 3

A comprehensive bibliography of distributed shared memory

ACM SIGOPS Operating Systems Review
Performance adaptive power-aware reconfigurable optical interconnects for high-performance computing (HPC) systems

Proceedings of the 2007 ACM/IEEE conference on Supercomputing
Two-hop free-space based optical interconnects for chip multiprocessors

NOCS '11 Proceedings of the Fifth ACM/IEEE International Symposium on Networks-on-Chip

Quantified Score

Hi-index	0.00

Visualization

Abstract

The paper presents a low overhead, high performance cache coherence protocol designed to exploit high bandwidth point to point and broadcast features of optics. SPEED integrates the virtues of snoopy based schemes and directory based schemes into one efficient protocol. Directory assist is used exclusively for read traffic to eliminate unnecessary broadcasts while snoopy assist is used exclusively for write and synchronization traffic to reduce directory overhead and synchronization complexities. The proposed protocol has the potential to increase performance as a result of its global independence between read and write operations, concurrency in channel access, reduced contention, and efficient broadcast of coherence and synchronization events.