https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125388
--- Comment #3 from Matthias Kretz (Vir) <mkretz at gcc dot gnu.org> --- (In reply to Drea Pinski from comment #2) > Don't use floating point for integer division. Plain and simple. That's not a helpful comment. > Especially with -ffast-math. IEC60559 Floating-point division with subsequent truncation is precise and does exactly what is needed. My intent is to use the vdivps / vdivpd instructions. My intent is not necessarily to use float/double division. (Because I understand that the two are not necessarily the same.) > -ffast-math uses reciprocal estimate on x86_64. I know. There's a C++ paper coming that asks for a function to get correctly rounded division *unconditionally*. I'm basically asking for the same thing.
