The compute time for the XOR operation follows a power curve as a function of the number of input vertices. One deals with this by subdividing the layout into multiple windows so that each processor in a workstation can work on an optimal number of polygons/vertices.
If you have a good idea of the "optimal" number of vertices per tile, you can divide the layout into a number of tiles so that no tile contains more than the optimal number.
