The
lack of physical information at the early gate-level Netlist
makes timing accuracy a significant challenge. Advancing in the physical
design flow reduces the timing inaccuracy gap in the cell’s
delay estimation, where the physical layout information is no longer
estimated. This paper aims to improve the cell delay accuracy of Oasys-RTL Synthesis by predicting the cell’s output
capacitance using Machine Learning algorithms. We trained and tested
various ML Regression algorithms to predict the Oasys-RTL cell’s
output capacitance using several 16 nm designs. With the help of a
state-of-art tool, APRISA P
&R by Siemens
EDA, we derived accurate output capacitance values and compared
the inaccuracy and error margins between postplaced APRISA and postplaced
Oasys-RTL Netlists before and after applying ML models. We reached
the highest accuracy using the Random Forest model.
The RF model has achieved a total average accuracy of 93.04% and has reduced the RMSE from 2.537 to 0.655.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.