Problems with cudaGLRegisterBuffer


When using a pixel/vertex buffer object to copy data between CUDA and GL, it seems, that glBufferData should not be called with a NULL pointer for the data parameter, or the following cudaGLRegisterBuffer command will cause an access violation.

I solved the problem by temporarily allocating a buffer, to pass to glBufferData. Anyone experienced the same thing? Or maybe someone who called cudaGLRegisterBuffer successfully after a call to glBufferData with NULL argument?

Thanks in advance!

It is a known bug.

Was it fixed in CUDA 1.0? I couldn’t find it in the release notes?

Yes, this is fixed in CUDA 1.0.