Online Learning for Dynamic Control of OpenMP Workloads - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2020

Online Learning for Dynamic Control of OpenMP Workloads

Maxime Mirka
Gilles Sassatelli
Abdoulaye Gamatié

Résumé

Optimizing energy-efficiency of modern multicore compute systems through online control is often regarded as both promising and challenging. In this paper, we propose a dynamic control technique for OpenMP workloads that exploits online energy efficiency measures derived from the OpenMP runtime. The proposed strategy relies on an automatic program phase identification which detects workload execution patterns, used by an online learning back-end. We design a synthetic benchmark template that makes it possible to produce benchmarks with controllable characteristics for mimicking a wide range of workload profiles. Experimental results show improvements on a 20-core server when compared to default Linux governors.
Fichier principal
Vignette du fichier
MOCAST_2020_paper_97.pdf (3.6 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02565961 , version 1 (06-05-2020)

Identifiants

Citer

Maxime Mirka, Gilles Sassatelli, Abdoulaye Gamatié. Online Learning for Dynamic Control of OpenMP Workloads. MOCAST 2020 - 9th International Conference on Modern Circuits and Systems Technologies, Sep 2020, Bremen, Germany. ⟨10.1109/MOCAST49295.2020.9200292⟩. ⟨hal-02565961⟩
114 Consultations
63 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More