I posted last week about a problem getting cuGLCtxCreate to work for me - I was getting an out of memory error from CUDA every time I registered a buffer in Vista. I figured out the problem - I was using the OpenGL extension NV_vertex_buffer_unified_memory and making all of my OpenGL buffers resident when they were created (before registering some of them with CUDA). Apparently, this step causes an interaction with CUDA that prevents the buffer from later being registered - in Vista / Win7 only. It works on XP.
I have not checked yet to see if making a buffer non-resident in OpenGL, and then registering / mapping the buffer in CUDA works.
I don’t know if this is a bug, but I figured I’d at least post it here in case I’m not the only one trying to use CUDA and vbum together.