Researchers have developed HiDVFS, a novel hierarchical multi-agent DVFS scheduler designed for real-time OpenMP DAG workloads on multicore embedded systems. This system addresses the challenge of leakage power by incorporating per-core, temperature-aware control and prioritizing deadlines alongside thermal limits. HiDVFS utilizes distinct agents for profiling, thermal management, and task prioritization, trained with a makespan-focused reward and a calibrated conformal shield for predicted response times. Benchmarks on various platforms demonstrate significant improvements in makespan, energy reduction, and deadline adherence compared to existing methods. AI
IMPACT This research introduces advanced scheduling techniques that could improve the efficiency and reliability of AI workloads running on embedded systems.
RANK_REASON Academic paper detailing a new scheduling algorithm. [lever_c_demoted from research: ic=1 ai=0.7]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
