Predictable GPUs Frequency Scaling for Energy and Performance

dc.contributor.authorFan, Kaijie
dc.contributor.authorCosenza, Biagio
dc.contributor.authorJuurlink, Ben
dc.date.accessioned2019-06-18T11:28:05Z
dc.date.available2019-06-18T11:28:05Z
dc.date.issued2019
dc.description.abstractDynamic voltage and frequency scaling (DVFS) is an important solution to balance performance and energy consumption, and hardware vendors provide management libraries that allow the programmer to change both memory and core frequencies. The possibility to manually set these frequencies is a great opportunity for application tuning, which can focus on the best applicationdependent setting. However, this task is not straightforward because of the large set of possible configurations and because of the multi-objective nature of the problem, which minimizes energy consumption and maximizes performance. This paper proposes a method to predict the best core and memory frequency configurations on GPUs for an input OpenCL kernel. Our modeling approach, based on machine learning, first predicts speedup and normalized energy over the default frequency configuration. Then, it combines the two models into a multi-objective one that predicts a Pareto-set of frequency configurations. The approach uses static code features, is built on a set of carefully designed microbenchmarks, and can predict the best frequency settings of a new kernel without executing it. Test results show that our modeling approach is very accurate on predicting extrema points and Pareto set for ten out of twelve test benchmarks, and discover frequency configurations that dominate the default configuration in either energy or performance.en
dc.description.sponsorshipDFG, 360291326, CELERITY: Innovative Modellierung für Skalierbare Verteilte Laufzeitsystemeen
dc.identifier.isbn978-1-4503-6295-5
dc.identifier.urihttps://depositonce.tu-berlin.de/handle/11303/9528
dc.identifier.urihttp://dx.doi.org/10.14279/depositonce-8579
dc.language.isoenen
dc.rights.urihttp://rightsstatements.org/vocab/InC/1.0/en
dc.subject.ddc004 Datenverarbeitung; Informatikde
dc.subject.otherfrequency scalingen
dc.subject.otherenergy efficiencyen
dc.subject.otherGPUsen
dc.subject.othermodelingen
dc.titlePredictable GPUs Frequency Scaling for Energy and Performanceen
dc.typeConference Objecten
dc.type.versionacceptedVersionen
dcterms.bibliographicCitation.doi10.1145/3337821.3337833en
dcterms.bibliographicCitation.originalpublishernameAssociation for Computing Machinery (ACM)en
dcterms.bibliographicCitation.originalpublisherplaceNew York, NY, USAen
dcterms.bibliographicCitation.proceedingstitleProceedings of the 48th International Conference on Parallel Processing (ICPP 2019)en
tub.accessrights.dnbfreeen
tub.affiliationFak. 4 Elektrotechnik und Informatik>Inst. Technische Informatik und Mikroelektronik>FG Architektur eingebetteter Systemede
tub.affiliation.facultyFak. 4 Elektrotechnik und Informatikde
tub.affiliation.groupFG Architektur eingebetteter Systemede
tub.affiliation.instituteInst. Technische Informatik und Mikroelektronikde
tub.publisher.universityorinstitutionTechnische Universität Berlinen
Files
Original bundle
Now showing 1 - 1 of 1
Loading…
Thumbnail Image
Name:
fan_cosenza_juurlink_icpp2019.pdf
Size:
1.45 MB
Format:
Adobe Portable Document Format
Description:
Collections