Change search
ReferencesLink to record
Permanent link

Direct link
A Comparative Analysis of Shared Cache Management Techniques for Chip Multiprocessors
Norwegian University of Science and Technology, Faculty of Information Technology, Mathematics and Electrical Engineering, Department of Computer and Information Science.
2013 (English)MasteroppgaveStudent thesis
Abstract [en]

In this thesis we present a comparative analysis of shared cache management techniques for chip multiprocessors. When sharing an unmanaged cache between multiple cores, destructive interference can reduce the performance of the system as the cores compete over limited cache space. This situation is made worse by streamlike applications that exhibit low locality of reference but has high cache demands. Several schemes for dynamically adjusting cache space available to each core has been suggested, and in this work we evaluate 3 such schemes as well as static partitioning and conventional LRU. We deploy a well defined simulation methodology to analyze the performance of the cache management techniques. The gem5 simulator is used to simulate the ARM ISA, and the SPEC2006 benchmark suite is used to create multi-programmed workloads. The simulator has been extended to support cache management schemes and provide detailed simulation statistics. We implement UCP, PIPP, PriSM and static partitioning, and simulate dual core, quad core and 8 core workloads. Our results show that destructive interference is a real issue in many workloads. Static partitioning can work well in scenarios where applications have similar cache demands, by creating private areas in the cache for each core. UCP improves on static partitioning by dynamically adjusting the size of each partition during runtime. PIPP performs decently by trying to maintain a specific cache occupation for each core without strictly enforcing a partition, but does not quite achieve the desired occupation and thus its performance suffers. PriSM fails to perform well, as its effort to determine a target cache allocation and maintain it does not work successfully for our workloads.

Place, publisher, year, edition, pages
Institutt for datateknikk og informasjonsvitenskap , 2013. , 105 p.
URN: urn:nbn:no:ntnu:diva-23386Local ID: ntnudaim:9177OAI: diva2:662350
Available from: 2013-11-06 Created: 2013-11-06 Last updated: 2013-11-06Bibliographically approved

Open Access in DiVA

fulltext(1814 kB)74 downloads
File information
File name FULLTEXT01.pdfFile size 1814 kBChecksum SHA-512
Type fulltextMimetype application/pdf
cover(184 kB)23 downloads
File information
File name COVER01.pdfFile size 184 kBChecksum SHA-512
Type coverMimetype application/pdf

By organisation
Department of Computer and Information Science

Search outside of DiVA

GoogleGoogle Scholar
Total: 74 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

Total: 113 hits
ReferencesLink to record
Permanent link

Direct link