Close

Presentation

Aging-aware Critical Path Selection via Graph Attentional Networks
TimeWednesday, July 13th6pm - 7pm PDT
LocationLevel 2 Lobby
Event Type
Networking Reception
Work-in-Progress Poster
DescriptionThe conventional critical path selection tools cannot figure out critical paths accurately considering aging for dynamic timing analysis (DTA). This paper proposes an aging-aware critical path selection flow. The Graph-Attention networks are used to predict the critical gates in the aged circuits. Then the path criticality computation algorithm can achieve path ranking. Experimental results show the proposed GAT model has superior accuracy to classical machine learning models on critical gate detection. Compared with the commercial tool, the proposed aging-aware flow achieves 99.51%, 98.56%, and 97.00% accuracy on average for top-1%, top-5%, and top-10% path sets under different aging conditions.