Questions about nvcc include path

I am trying to extend the vector support by CUDA, e.g, overriding operators in vector structures. I could just do it by modifying the default vector_types.h file. However it is not possible in a multi-user environment where I don’t have the permission to modify it.

Hence, I decided to create a new vector_types.h and do not include the original vector_types.h. But it seems nvcc automatically includes this file even I don’t include it in my code. So my question is that is there anyway to change this behavior of nvcc?

Any help would be appreciated.