A unified approach to topology generation and area optimization of general floorplans

Abstract

In this paper, it is shown that for any rectangularly dualizable graph, a feasible topology can be obtained by using only either straight or Z-cutlines recursively within a bounding rectangle. Given an adjacency graph, a potential topology, which may be nonslicible and is likely to yield an optimally sized floorplan, is produced first in a top-dozen fashion using heuristic search in AND-OR graphs. The advantage of this technique is four-fold: (i) accelerates top-down search phase, (ii) generates a floorplan with minimal number of nonslice cores, (iii) ensures safe routing order without addition of pseudo-modules, and (iv) solves the bottom-up algorithm efficiently for optimal sizing of general floorplans in the second phase.