arXiv:2601.06425v2 Announce Type: replace-cross Abstract: Leakage power in multicore embedded systems now rivals dynamic power, so DVFS schedulers must respect deadlines and thermal limits, not just average makespan. Existing heuristics lack per-core, temperature-aware control and overlook the irregular execution of OpenMP DAGs. We propose HiDVFS, a general, extensible hierarchical multi-agent DVFS scheduler: a profiler agent selects cores and frequencies, a thermal agent groups cores by temperature, and a priority agent orders tasks under contention, all trained with a makespan-focused reward
The increasing power density and complexity of multi-core embedded systems for real-time AI workloads are making thermal management and power efficiency critical constraints.
Efficient power management in high-performance computing directly impacts performance, cost, and environmental footprint, especially for AI applications where power consumption is a growing bottleneck.
This advancement offers a more sophisticated and adaptive approach to Dynamic Voltage and Frequency Scaling (DVFS), moving beyond static heuristics to optimize real-time system performance under thermal and deadline constraints.
- · AI hardware developers
- · Embedded systems manufacturers
- · High-performance computing sector
- · Data center operators
- · Less efficient DVFS solutions
- · Developers neglecting thermal management
Improved energy efficiency and reliability in AI-specific hardware, reducing operational costs and extending component lifespan.
Enables the deployment of more powerful AI models on edge devices and embedded systems by overcoming thermal and power limitations.
Could contribute to the broader adoption of AI across various industrial applications where power and real-time performance are paramount.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI