Currently it is not possible to compile the host and device side of *cuda_memtest* with the clang compiler. Maybe it is enough to update the `CMakeList.txt` and take care of the differences to create the binary with nvcc and clang.