As CMP platforms are widely adopted, more and more cores are integrated on to the die. To reduce the off-chip memory access, the last level cache is usually organized as a distributed shared cache. In order to avoid hot-spots, cache lines are interleaved across the distributed shared cache slices using a hash function. However, as we increase the number of cores and cache slices in the platform, this also implies that most of data references go to remote cache slices, thereby increasing the access latency significantly. In this paper, we propose a hybrid last level cache, which has some amount of private space and some amount of shared space on each cache slice. For workloads with no sharing, the goal is to provide more hits into the local slice while still keeping the overall miss rate low. For workloads with sufficient sharing, the goal is to allow more sharing in the last-level cache slice. We present hybrid last-level cache design options and study its hit/miss rate behavior for a number of important server applications and multi-programmed workloads. Our simulation results on running multi-programmed workloads based on SPEC CINT2000 as well as multithreaded workloads based on commercial server benchmarks (TPCC, SPECjbb, SAP and TPCE) show that this architecture is advantageous especially since it can improve the local hit rate significantly while keeping the overall miss rate similar to the shared cache.
End of preview. The entire article is 8 pages. To view the full-text, please rent this article to continue.
/lp/association-for-computing-machinery/towards-hybrid-last-level-caches-for-chip-multiprocessors-0JS189V7Mx
Welcome to DeepDyve! Rent Premier Research Articles and Save Up to 90%
Towards hybrid last level caches for chip-multiprocessors
Zhao, Li; Iyer, Ravi; Upton, Mike; Newell, Don
Follow ACM SIGARCH Computer Architecture News
, Volume 36 (2)
Association for Computing Machinery – May 1, 2008
More Info
More Like This Article
View All dataSource[]=actageo&dataSource[]=aspet&dataSource[]=aaos&dataSource[]=aacc&dataSource[]=aacr&dataSource[]=aea&dataSource[]=aip&dataSource[]=ajnr&dataSource[]=ams&dataSource[]=aps_physical&dataSource[]=appi_book&dataSource[]=appi_journal&dataSource[]=apha&dataSource[]=asip&dataSource[]=asm&dataSource[]=asn&dataSource[]=aspb&dataSource[]=avs&dataSource[]=annual_reviews&dataSource[]=arxiv&dataSource[]=acm&dataSource[]=berghahn&dataSource[]=cabi&dataSource[]=clinical_trials&dataSource[]=dailymed&dataSource[]=degruyter&dataSource[]=du_press&dataSource[]=esa&dataSource[]=eu_press&dataSource[]=elsevier&dataSource[]=emerald&dataSource[]=ejtr&dataSource[]=emea&dataSource[]=epo&dataSource[]=faseb&dataSource[]=gsa&dataSource[]=health_affairs&dataSource[]=hindawi&dataSource[]=imanager&dataSource[]=imedpub&dataSource[]=informa_healthcare&dataSource[]=informs&dataSource[]=iop&dataSource[]=iucr&dataSource[]=iospress&dataSource[]=jbjs&dataSource[]=leftcoast&dataSource[]=lu_press&dataSource[]=mesharpe&dataSource[]=mary_ann_liebert&dataSource[]=medline&dataSource[]=mit_press&dataSource[]=nature&dataSource[]=oxford&dataSource[]=pier_professional&dataSource[]=pnas&dataSource[]=portlandpress&dataSource[]=psyc_articles&dataSource[]=psyc_books&dataSource[]=psyc_critiques&dataSource[]=plos_journal&dataSource[]=pubmed_central&dataSource[]=rsna&dataSource[]=rockefeller&dataSource[]=rcn&dataSource[]=ria&dataSource[]=rsc&dataSource[]=sage&dataSource[]=spie&dataSource[]=springer_journal&dataSource[]=springer&dataSource[]=taylor_francis&dataSource[]=aps&dataSource[]=the_scientist&dataSource[]=uc_press&dataSource[]=uspto_abstract&dataSource[]=wiley&dataSource[]=pct
© 2012 DeepDyve, Inc. All rights reserved.
Terms of Service | Privacy Policy
