Use intrinsics for vector operations as arithmetic overloading is not supported by the Intel compiler yet.