We may be able to implement an adaptive grid, could give us another decent speed up. [One method](http://wissrech.ins.uni-bonn.de/research/pub/hamaekers/GriebelHamaekersINSpreprint0603.pdf)
