I am currently using a cuda kernel to generate geometry (basically modifying the marching cubes example from the sdk). However, in that example they just allocate one giant vbo and write to that. However, in my code I want to create many VBOs and write to them in cuda kernels using cudaGLRegisterBufferObject. The problem I am encountering right now is I get an out of memory error that is way too premature, that happens on the cudaGLRegisterBufferObject() call. Is there a problem with registering too many VBOs at once?
I can generate about 100 times more geometry with one giant VBO, the problem is there is a lot wasted space (0 padding between objects for complicated reasons), so I want to use many VBOs and let the card handle memory.