Multiple Frequency Selection in DVFS-Enabled Processors to Minimize Energy Consumption

TL;DR

This paper introduces MFS-DVFS, a novel technique for reducing energy consumption in DVFS-enabled processors by executing tasks with multiple frequencies to optimize slack time utilization.

Contribution

It proposes a new slack reclamation method using multiple frequency selection to minimize energy, outperforming previous approaches.

Findings

Achieves energy savings close to the optimal

Effective on both random and real-world task graphs

Outperforms previous slack reclamation techniques

Abstract

In this chapter we focus on slack reclamation and propose a new slack reclamation technique, Multiple Frequency Selection DVFS (MFS-DVFS). The key idea is to execute each task with a linear combination of more than one frequency such that this combination results in using the lowest energy by covering the whole slack time of the task. We have tested our algorithm with both random and real-world application task graphs and compared with the results in previous researches in [9] and [12-13]. The experimental results show that our approach can achieve energy almost identical to the optimum energy saving.