PolyMorph engine (Tessellation) - cuda/OpenCL

i would like to know if i have only access to the PolyMorph engine with DirectX or OpenGL.
In that case i have no access with Cuda and OpenCL.

I would like to use hardware-Tessellation with my Cuda or OpenCL programm but it seems to be not possibly.
Than i have to implement Tessellation on my own if my presumption is rigtht.

So is it correct or incorrect?

thanks for your attention.

Yes, hardware tessellation is a graphics feature and is not accessible from CUDA or OpenCL.

It theory it would be possible to do tessellation in OpenGL or DX11, stream out the results to a vertex buffer, and then map this buffer in CUDA, but I wouldn’t recommend it.

Why not?

I don’t have any hard data to back this up, just intuition - stream out isn’t that efficient and the cuda / graphics context switch is still quite expensive, so I think you’d probably be better off just implementing a tessellator in CUDA.

Tessellation on Fermi is really fast, however, so I’d be happy to be proved wrong!