I can find linker/compiler command line options for nvcc and VC/C++ but not for Intel C, not when in 'CUDA runtime' projects. How can I keep them separated without the compilers throwing errors?