A genetic algorithm for global improvement of macrocell layouts

Abstract

The result of many floorplanning algorithms is a placement of macrocells. A novel technique for the improvement of a given placement is presented which is based on the optimization of the channel densities by refining the cell positions. The authors introduce a distance function for each channel representing the channel width. This width can be altered by shifting adjacent cells along each other by an offset. They present an optimization to find offsets for adjacent cells which lead to a minimal area demand of the total layout. The method is based on a genetic algorithm, an iterative improvement procedure. Results are presented.<>