Brief Industry Paper: Retention-Based Energy-Efficient Scheduling of Arbitrary-Deadline DAG Tasks on Multicore Platforms

Retention, with its powerful energy-saving effect and negligible switching delay, offers the advantages of Power Gating (PG) and Clock Gating (CG), and has gradually been adopted into the processor power management systems by both AMD and Intel. However, there has been no research on how to integrate retention to achieve energy-efficient task scheduling. Furthermore, although most commercial chips support per-cluster configurations to achieve energy saving, existing energy-efficient works often underutilize all cores in the cluster, resulting in numerous static power from unused cores. In this paper, we propose a task partitioning algorithm to optimize energy consumption among clusters. We further study energy-efficient scheduling of directed acyclic graph (DAG) tasks with arbitrary deadlines on multicore platforms, which is the first work to integrate the per-cluster retention to reduce power consumption by fully utilizing the used clusters. The experimental results verify that our approach can increase the retention time by 21.24% and reduce the static energy consumption in the idle state by 45.81% compared to retention-based work-conserving scheduling.