Poster De Conférence Année : 2024

How to prove the existence of an optimal frequency assignment algorithm for CPU and memory accesses in absence of an appropriate energy model

Résumé

While targeting a reduced energy consumption for the embedded real-time systems community, different techniques exist and, in this paper, we are interested in Dynamic Voltage and Frequency Scaling. Quan and Hu propose an algorithm ensuring an energy-optimal CPU frequency assignment for executing independent programs on a single-core processor under realtime constraints. Since its optimality is proposed for CPU variable frequencies, we propose to add variable frequency assignment for memory accesses to the problem solved by Quan and Hu. Based on numerical evaluation of energy consumption of TACLeBench programs, we study the impact of both CPU and memory accesses frequencies on the energy consumption but no existing energy model considers these two factors. Moving back to the existence of an optimal algorithm to assign both frequencies, one may propose to find the appropriate pair of CPU-memory accesses frequencies, but an appropriate energy model is required to compare two different pairs.

Mots clés

Fichier principal
Vignette du fichier
RTNS-JW-2024.pdf (306.93 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-04999233 , version 1 (20-03-2025)

Licence

Identifiants

  • HAL Id : hal-04999233 , version 1

Citer

Myriam Mabrouki, Liliana Cucu-Grosjean, Stéphan Plassart. How to prove the existence of an optimal frequency assignment algorithm for CPU and memory accesses in absence of an appropriate energy model. JRWRTC-RTNS 2024 - Junior Researcher Workshop on Real-Time Computing at the 32nd International Conference on Real-Time Networks and Systems, Nov 2024, Porto, Portugal. https://cister-labs.pt/rtns24/JRWRTC2024-proceedings.pdf. ⟨hal-04999233⟩
195 Consultations
172 Téléchargements

Partager

  • More