OpenGL working with Direct HMD mode

"tmason101" wrote:
I assume this would have a rather significant performance hit, no?

You are essentially creating two contexts to the graphics card and copying from one buffer/memory location of some kind to another, which Oculus is again doing behind the scenes with their stuff, every frame.

I doubt this would impose a serious performance penalty. While you're creating two contexts, you're essentially using them for discrete tasks. Previously, OpenGL would have been used for both rendering and distortion. Now OpenGL is used for rendering, while Direct3D is used for distortion. Consequently, you're actually reducing the amount of work the OpenGL context has to do.

Furthermore, I doubt that any memory is being copied at all. I don't think nVidia would be so crude as to make a special OpenGL texture area and copy the data to Direct3D texture when OpenGL writing was complete. Instead, nVidia has all the freedome they would require to make sure that OpenGL and Direct3D are sharing the exact same memory. Even if they weren't, the copy would be a block to block transfer directly on the video card, much faster than CPU to GPU transfers.