It currently only has a lower and upper bound, so it loses information about gaps in the range. LLVM supports outputting the full set of information.