A new linear placement algorithm for cell generation

Many design styles used for automatic cell generation need a linear placement of their components. The authors present a novel two-phase algorithm for that task. In a first step, the net length is globally minimized by quadratic programming with an iterative weight update. Its results are as good as a minimization of the net length by linear programming but are obtained by significantly shorter CPU times. In a second step, track count, net length, and possibility of abutment between components are optimized simultaneously by means of a branch and bound algorithm applied to local subproblems. A comparison to previously published algorithms shows that the results of the proposed algorithm are superior.<>