Horde3D

http://developer.nvidia.com/object/bind ... phics.html
http://developer.download.nvidia.com/op ... aphics.pdf



This is very interesting... Instead of the driver giving back object names/IDs, which can be converted to pointers inside the driver, the driver gives back actual GPU pointer addresses as the object name.

I saw that a few days ago and was quite excited about it. It is great to see that drawcalls are getting cheaper which helps a lot to reduce the complexity (no more texture atlases, extensive batching, instancing, etc. required).

I just hope that ATI / the ARB also feel that this is a good idea

Thankfully ATI, nVidia, and the ARB don't behave like our politicians here in the good ol' dysfunctional states. This sounds like an awesome idea.

One concern could be type safety though. If the pointers are accessible in the application memory space, we will probably need something to differentiate it from C/++ pointers. GLpointer perhaps?