Change search
ReferencesLink to record
Permanent link

Direct link
Software Distributed Shared Memory Using the VIPS Coherence Protocol
Uppsala University, Disciplinary Domain of Science and Technology, Mathematics and Computer Science, Department of Information Technology.
2015 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

A coherent global address space in a distributed system enables shared memory programming in a much larger scale than in a single multicore processor. The solution is to implement a software distributed shared memory (SWDSM) system since hardware support at this scale is non-existent. However, traditional approaches to coherence in SWDSM systems (centralized via 'active' home-node directories) are inherently unfit for such a scenario. Instead, it is crucial to make decisions locally and avoid the long latency imposed by both network and software message-handlers. This thesis investigates the performance of an SWDSM system with a novel and completely distributed coherence protocol that minimizes long-latency communications common in coherence protocols. More specifically, we propose an approach suitable for data race free programs, based on self-invalidation and self-downgrade inspired by the VIPS cache coherence protocol. This thesis tries to exploit the distributed nature of self-invalidation, self-downgrade by using a passive data classification directory that require no message-handlers, thereby incurring no extra latency when issuing coherence requests. The result is an SWDSM system called SVIPS which maximizes local decision making and allows high parallel performance with minimal coherence traffic between nodes in a distributed system.

Place, publisher, year, edition, pages
2015. , 53 p.
UPTEC IT, ISSN 1401-5749 ; 15002
National Category
Engineering and Technology
URN: urn:nbn:se:uu:diva-256975OAI: diva2:827858
Available from: 2015-06-29 Created: 2015-06-29 Last updated: 2015-06-29Bibliographically approved

Open Access in DiVA

fulltext(2824 kB)110 downloads
File information
File name FULLTEXT01.pdfFile size 2824 kBChecksum SHA-512
Type fulltextMimetype application/pdf

By organisation
Department of Information Technology
Engineering and Technology

Search outside of DiVA

GoogleGoogle Scholar
Total: 110 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: 1045 hits
ReferencesLink to record
Permanent link

Direct link