Abstract

As the number of cores in large-scale chip multiprocessors (CMPs) grows, the complexity and overhead of directory-based cache coherence protocols become increasingly important. To reduce the cost of coherence protocol, previous studies classified data into private and shared, and used different coherence protocols for them. On the one hand, private data can be taken out of coherence. On the other hand, because of the spatial locality of private data, a coarse-grained directory can be used to reduce the directory storage. This paper proposes RG_VIPS, a simple coherence protocol based on region granularity directory. We use a write-through protocol for private data and a write-back protocol for shared data. Compared to previous work VIPS, RG_VIPS eliminates OS dependencies and also implements a simple cache coherence protocol (only two stable protocol states, Valid and Invalid) with finer granularity. Simulation of a 16-core CMP with PARSEC shows that the RG_VIPS protocol has little impact on the performance while simplifying the protocol.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.