On Sep 18, 2009, at 5:00 PM, Barend Gehrels wrote:
Gordon Woodhull wrote:
Luke and Barend, I'm sorry but I don't believe either of your benchmarks. I think you are choosing problems that suit each of your libraries. This is understandable, but we will need to build a comprehensive suite of examples from many different domains to learn which algorithms are actually faster in which situations.
Note that the GGL benchmark is Open Source, can be repeated by anyone and can also be extended by anyone. So please add your problem, either directly to the suite or you can suggest it to us such that we can check it.
I surely will, when I have time. To tell the truth, geometry performance has not been much of an issue on most of what I've worked on, where speed is dominated by what's generating the smallish geometries. Based on the presentations of the algorithms given at Boostcon I would expect GGL to be faster on smaller cases and Polygon2D4590 to excel on really huge sets with tons of intersections. Neither algorithm seems to be completely mature. My larger point is that there might be room for many various intersection algorithms for different 2D geometry domains. Looking forward, Gordon