Cuda interaction with SceniX and Managed Code


Is it possible to integrate CUDA, ScenicX AND managed (C#) code and are there any guidlines on this?

I want to program graphics and computational intensive tasks in SceniX and CUDA, but at the same time use C# to access database and create menus etc. Communicating between managed and unmanaged code is not the problem, though there is a need to share (large) datasets between the two. The main issue as of now is that there will be a need to have both SceniX and managed C# eventing run i parallell (on different threads I guess?, but in the same process), as well as performance penalties involved in having both SceniX and (multiple) Cuda routines utilize the same GPU simultaneously whilst sharing the same data chunks on the GPU.

Question really is what is possible, what are the performance penalties and what are the recommended ways of doing it?

I am a bit surprised of the lack of documantation or guidelines on similar matters. Am I just failing to locate such documentation or is really that scarse at this point in time?