A consistent local congestion avoiding strategy for iterative improvement of cross point assignment

Abstract

Absrruct: Cross point assignment(CPA) is a critical step between global routing and detailed routing. Usually, A CPA algorithm includes two steps:initial CPA and iterative improvement. In this paper, we proposed a consistent local congestion avoiding strategy for CPA iterative improvement based on four congestion lines which describe accurately the local congested area in a congested GRC. The whole strategy consists of three consistent parts: rip up nets crossing congestion lines, reroute(g1obal routing) nets without crossing congestion lines, and avoid crossing congestion line while re-assigning cross points for ripped up nets. By this strategy, we not only avoid local congestion area. but also make fully use of the routing area in congested GRC's. Experimental results show that our strategy can reduce congested CRCs effectively and achieve high completion rate in detailed routing.