Possible to have some perf comparison with official bindings?

[YuhengZhi]

Hi Hirolshida,

Thanks for this interesting development!

I think it might be very helpful if there is some quantitative perf comparison with the official bindings as motivations for using these bindings vs. the official ones (and possibly vice versa, depending on specific use cases) ?

P.S. I personally do find pybind11 bindings easier to read / debug / code navigation than some other solutions.

[HiroIshida]

@YuhengZhi

I don't have quantitative measurements, but some reports suggest that Boost.Python (which is used in OMPL) is faster than pybind11 (pybind/pybind11#1825). In my application, where is_valid performs full DoF humanoid collision checking, this difference is not a significant problem. However, in cases such as mobile robot navigation, this performance difference could be quite large (if I were to write everything in C++ for the application).