A new performance driven macro-cell placement algorithm

Abstract

The authors present a new performance-driven macro-cell placement algorithm. They introduce the concept of a window which is an estimate of the initial placement of a module. There are three phases in the algorithm. In phase one, an initial window for each module is constructed. In phase two, a novel force-directed approach is used to reduce the size of each window in an iterative process so as to determine an initial placement of the modules. In phase three, the same force-directed concept is used to refine the placement. Timing and physical constraints are used in all phases to guide the placement process. The effectiveness of the algorithm is demonstrated by comparing the experimental results with those produced by TimberWolfMC and the GVL algorithm.<>