The Reproducing Placement Problem With Applications

Abstract

We study a new placement problem: the reproducing placement problem (RPP). In each phase a module (or gate) is decomposed into two (or more) simpler modules. The goal is to find a “good” placement in each phase. The problem, being iterative in nature, requires an iterative algorithm. The problem finds applications in several gate-level placement problems, e.g., in layout-driven logic synthesis. We introduce the notion of minimum floating Steiner trees (MFST). We employ an MFST algorithm as a central step in solving the RPP. A Hanan-like theorem is established for the MFST problem and two approximation algorithms are proposed. Experiments on commonly employed benchmarks verify the effectiveness of the proposed technique.