Partitioning-oriented placement using advanced genetic algorithm BOA

placement problem, hypergraph bisectioning, simple and advanced GA, estimation of distribution, BOA algorithm, Bayesian network, top-down placer

This paper deals with an adaptation of the genetic partitioning algorithm BOA, based on the estimation of distribution of promising solution, for the placement of the hypergraph nodes into the regular structure of allocations. This task is a simplification of the placement problem encountered in PCB design or more complex case of physical layout of gate array on VLSI chip level. We present top-down placer based on the recursive bisectioning of the hypergraph/circuits. The hypergraph is repeatedly divided into densely connected subgraphs so the number of nets among them is minimized. We compare performance of our placer with the one based on the hybrid GA algorithm and Breuer's force-directed algorithm on artificial and real hypergraphs.