Do we know how this compares to native OpenGL implementations form AMD/Nvidia ? It would seem like these optimizations are already implemented on their native drivers.