Presentation
Late Breaking Results: RL-LPO: Reinforcement Learning Based Leakage Power Optimization Framework with Graph Neural Network
DescriptionLeakage power optimization based on threshold voltage (𝑽𝑡ℎ) assignment poses great challenge in circuit design. In this paper, a Reinforcement Learning-based Leakage Power Optimization framework (RL-LPO) is proposed to formulate 𝑽𝑡ℎ assignment as a reinforcement learning (RL) process by learning characteristics of circuit instances with Graph Neural Networks (GNN). The RL-LPO was validated by the IWLS2005 and Opencores benchmark circuits and experimental results demonstrate that our work achieves better leakage power optimization by additional 3% reduction on average than the commercial tool 𝑃𝑟𝑖𝑚𝑒𝑇𝑖𝑚𝑒 with 6.7× speed up when being transferred to unseen circuits with negligible timing degradation
Event Type
Late Breaking Results Poster
TimeWednesday, July 12th6:00pm - 7:00pm PDT
LocationLevel 2 Lobby
AI
Autonomous Systems
Cloud
Design
EDA
Embedded Systems
IP
Security