Reproduce “DeepSoCS: A Neural Scheduler for Heterogeneous System-on-Chip (SoC) Resource Scheduling” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tegg Taekyong Sung, Jeongsoo Ha, Jeewoo Kim, Alex Yahja, Chae-Bong Sohn, Bo Ryu
DeepSoCS, a deep reinforcement learning-based scheduler, optimizes resource allocation for heterogeneous System-on-Chip (SoC) systems by prioritizing continuously arriving hierarchical jobs. This solution achieved 7-9% lower average latency compared to the HEFT heuristic and exhibited superior robustness under simulated noise conditions.
Key insights
- Deep reinforcement learning can overcome the inherent brittleness of heuristic schedulers, enabling adaptive and robust performance in dynamic and noisy heterogeneous SoC environments.
- Effective DRL-based scheduling in complex SoC systems necessitates a comprehensive state representation, integrating both structural graph embeddings and real-time operational features.
- Addressing challenges like delayed consequences and joint actions in real-time DRL requires specialized techniques, such as continuous reward functions and an understanding of action monotonicity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.