Use uint64_t type for floating point registers
as compiler don't know about uint128_t yet.

Discussed with: theraven, kib