Presentation
TOTAL: Multi-Corners Timing Optimization Based on Transfer and Active Learning
SessionAll Routes to Timing Closure
DescriptionDesigner normally needs to progressively optimize the design fro Static timing analysis (STA) of full corners, which is computationally expensive. We propose TOTAL, a multi-corner time optimization assistant that guides accurate and efficient time closure under multi-corner. TOTAL tries to catch the closed connections between the paths correlations during an optimization process in long and short terms. Most importantly, TOTAL is equipped with a Bayesian decision strategy to deliver reliable predictions with accuracy close to 100%, putting the machine-learning-based STA in practice in design optimization. Experimental results show that TOTAL's accuracy outperforms the SOTA competitors by up to 2x.
Event Type
Research Manuscript
TimeTuesday, July 11th2:25pm - 2:40pm PDT
Location3002, 3rd Floor
EDA
RISC-V
Timing and Low Power Design