Classical and intelligent approaches for VLSI static timing analysis: an overview
Nikolay O. Vasilyev, Mariya A. Zapletina, Галина Алексеевна Иванова · Radioelectronics Nanosystems Information Technologies · 2022
The paper systematizes the modern scientific and technical experience of VLSI timing analysis. It provides a comparative overview of some existing classical and intelligent methods of static timing analysis, as well as identifies actual tasks and directions for future investigation. The special attention is paid to the intelligent methods for reducing the pessimism of the integrated circuits’ static timing analysis results. In particular, an elimination of the false paths and the technological and operational parameters variations are considered. According to recent publications (issued in 2013-2021), various types of neural networks and regression analysis tools are often used to build intelligent methods for estimating the timing characteristics of integrated circuits. In a number of works aimed at reducing the pessimism of classical static timing analysis, the advantage of ensemble methods is argued. It is noted that methods of this type demonstrate more accurate results compared to single methods in conditions of insufficient training data.