[Bug rtl-optimization/82692] [8 Regression] Ordered comparisons used for unordered built-ins

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82692

--- Comment #15 from Segher Boessenkool <segher at gcc dot gnu.org> ---
My point is that doing this only for FLOAT_MODE_P makes no real sense.
If we can describe ordered comparisons with special CC modes, we should
do tests with those modes only here.