Bandwidth bandit: Quantitative characterization of memory contention
2012 (English)In: Parallel Architectures and Compilation Techniques - Conference Proceedings, PACT, 2012, 457-458 p.Conference paper (Refereed)
Applications that are co-scheduled on a multi-core compete for shared resources, such as cache capacity and memory bandwidth. The performance degradation resulting from this contention can be substantial, which makes it important to effectively manage these shared resources. This, however, requires quantitative insight into how applications are impacted by such contention. In this paper we present a quantitative method to measure applications' sensitivities to different degrees of contention for off-chip memory bandwidth on real hardware. We then use the data captured with our profiling method to estimate the throughput of a set of co-running instances of a single threaded application.
Place, publisher, year, edition, pages
2012. 457-458 p.
Memory contention, Performance prediction, Cache capacity, Memory bandwidths, Multi core, Off-chip memories, Performance degradation, Profiling methods, Quantitative characterization, Quantitative method, Shared resources, Single-threaded, Parallel architectures
IdentifiersURN: urn:nbn:se:uu:diva-186828DOI: 10.1145/2370816.2370894ISBN: 978-1-4503-1182-3OAI: oai:DiVA.org:uu-186828DiVA: diva2:575216
21st International Conference on Parallel Architectures and Compilation Techniques, PACT 2012, 19 September 2012 through 23 September 2012, Minneapolis, MN, USA