I'd like to see the percentage improvement over the "naive" approach (just stacking them edge-to-edge) for each of them. I can't tell if the improvements are significant enough for this to have obviously practical benefits.