On Wed, 18 Oct 2017, Richard Biener wrote: > When overflow/underflow can be disregarded is there any reason remaining to > make this guarded by flag_unsafe_math_optimizations? Are there any cases > where rounding issues can flip the comparison result?
Yes. E.g. (in round-to-nearest) 3.0f * 0xfffffep0f == 3.0f * 0xfffffdp0f (generically, multiplication / division by any non-power-of-2 constant should be expected sometimes to map two different values to the same value even when overflow and underflow cannot occur). -- Joseph S. Myers jos...@codesourcery.com