Algorithmic Aspects of Physical Design Problems

University Of California-Los Angeles, Los Angeles CA

Investigators

Abstract

Automation of a given design process requires an algorithmic analysis of it. The availability of fast and easily implementable algorithms is essential to the discipline. This proposal focuses on the physical design problems and their interaction with higher and lower-levels of the design hierarchy. Two classes of problems are being studied. First, the interaction between various cost functions in the placement problem is being investigated. In particular, relationship among net-cut, wirelength, congestion, and timing is being researched. Next, the class of algorithmic predictors and statistical predictors are being studied. One goal of this project is to show that it is possible to obtain accurate prediction very fast using floorplan and placement predictors. CAD tools associated with the above projects are under development.