Sharing GPU memory between CUDA and DirectX ?

I am trying to find a way to reduce what seems like unneeded memory movement.

Current flow:
Updates on CUDA allocated memory -> CPU memory -> DirectX buffers for instanced drawing

Would like to try to just run the CUDA kernels to update the GPU memory then have DirectX use the information already on the GPU for drawing.

Is there a way to do this, or does anyone know if there is a good chance this could work?


Take a look at the CUDA/DirectX interop APIs:

and sample codes:


Is it possible to retrieve too DX7 surface memory via CUDA ?
The goal is to be able to copy from DX7 surface to DX11 ressource without passing via the system memory (too slow)

Thanks in advance,