Comparing doubles

[Robert Dewar]

<<As for compare, I don't claim to know the intricacies of IEEE NaN and
all that, but surely on older float formats at least, compare is
indeed internally just a subtract and test for zero?  Or a normalize
(if applicable) and bitwise compare?  (The latter is definitely the
case on the CDC 6600, to use an old example.)
>>

No, because a subtract and test for zero is subject to problems with both
underflow and overflow and can therefore result in considering unequal
numbers to be equal.