Wirelength Optimal Rectangle Packings

Finding wirelength optimal packings of rectangles is a well known problem in VLSI design. We propose a branch and bound algorithm for this problem that is based on the rectangle packing algorithm of Moffitt and Pollack. It makes use of a very efficient implementation of an incremental network simplex algorithm. Our algorithm allows for the first time to find optimum solutions of three instances of the well known MCNC block packing benchmark and optimally solves real-world instances with up to 15 rectangles in 1 hour. The largest instance so far for which an optimum solution has been computed contained 6 blocks.