HiLITE: Hierarchical and Lightweight Imitation Learning for Power Management of Embedded SoCs

Anderson L. Sartor, Anish Krishnakumar, Samet E. Arda, Umit Y. Ogras, Radu Marculescu

Research output: Contribution to journalArticlepeer-review

Abstract

Modern systems-on-chip (SoCs) use dynamic power management (DPM) techniques to improve energy efficiency. However, existing techniques are unable to efficiently adapt the runtime decisions considering multiple objectives (e.g., energy and real-time requirements) simultaneously on heterogeneous platforms. To address this need, we propose HiLITE, a hierarchical imitation learning framework that maximizes the energy efficiency while satisfying soft real-time constraints on embedded SoCs. Our approach first trains DPM policies using imitation learning; then, it applies a regression policy at runtime to minimize deadline misses. HiLITE improves the energy-delay product by 40 percent on average, and reduces deadline misses by up to 76 percent, compared to state-of-the-art approaches. In addition, we show that the trained policies not only achieve high accuracy, but also have negligible prediction time overhead and small memory footprint.

Original languageEnglish (US)
Article number9085952
Pages (from-to)63-67
Number of pages5
JournalIEEE Computer Architecture Letters
Volume19
Issue number1
DOIs
StatePublished - Jan 1 2020
Externally publishedYes

Keywords

  • Dynamic power management
  • Hierarchical imitation learning
  • Machine learning
  • Real-time
  • SoC

ASJC Scopus subject areas

  • Hardware and Architecture

Fingerprint Dive into the research topics of 'HiLITE: Hierarchical and Lightweight Imitation Learning for Power Management of Embedded SoCs'. Together they form a unique fingerprint.

Cite this