Modelling the Energy Consumption of Soft Real-Time Tasks on Heterogeneous Computing Architectures

Abstract : The problem of reducing the energy consumption of embedded processors is of paramount importance for mobile devices powered by batteries. Many of these devices embed heterogeneous processors like the ARM bitLITTLE for adapting to the performance requirements and energy consumption profiles of modern applications. However, before proposing algorithms for optimally managing the energy consumption of a device, it is necessary to build a realistic model of the performance and power profile of an application. In this paper we address this problem by proposing a simple model for the execution time of soft real-time applications and the energy consumption of the hardware platform. We also propose a simple benchmarking methodology for obtaining the parameters of the model from measures. We identify in the memory access pattern of software tasks a key factor that influences both the performance and the energy consumption. We then show how to apply our methodology on the ODROID development board, and we present some preliminary results of our study. We conclude the paper by discussing current research and future directions.
Type de document :
Communication dans un congrès
Energy Efficiency with Heterogenous Computing, Jan 2016, prague, Czech Republic
Liste complète des métadonnées

Littérature citée [4 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-01242681
Contributeur : Houssam Eddine Zahaf <>
Soumis le : mardi 15 décembre 2015 - 13:10:36
Dernière modification le : vendredi 17 novembre 2017 - 08:50:19
Document(s) archivé(s) le : samedi 29 avril 2017 - 12:21:54

Fichier

EEHCO16_paper_1.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01242681, version 1

Citation

Houssam Eddine Zahaf, Richard Olejnik, Giuseppe Lipari, Abou El Hassen Benyamina. Modelling the Energy Consumption of Soft Real-Time Tasks on Heterogeneous Computing Architectures. Energy Efficiency with Heterogenous Computing, Jan 2016, prague, Czech Republic. 〈hal-01242681〉

Partager

Métriques

Consultations de la notice

257

Téléchargements de fichiers

321