RL-CCD: Concurrent Clock and Data Optimization using Attention-Based Self-Supervised Reinforcement Learning*
DescriptionConcurrent Clock and Data (CCD) optimization is a well-adopted technique that resolves timing violations using a mixture of clock and delay fixing strategies. However, existing CCD algorithms fail to prioritize violating endpoints for either clock-path or data-path optimization correctly, leading to globally sub-optimal results. In this paper, we present RL-CCD, a Reinforcement Learning (RL) agent that improves CCD quality by prioritizing endpoints for useful skew optimization using a self-supervised attention mechanism. Experimental results on 18 industrial designs in 5-12nm technologies demonstrate that RL-CCD improves a native commercial CCD engine by up to 64% in Total Negative Slack (TNS) .
TimeTuesday, July 11th1:40pm - 1:55pm PDT
Location3002, 3rd Floor
Timing and Low Power Design