How can I use __builtin_bswap64 in NVCC of CUDA?

There are built-in functions in GCC to flip byte contents, e.g. __builtin_bswap64. Is there a similar function under NVCC for CUDA? If not, how do I implement this function on the device?