https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88713
--- Comment #53 from rguenther at suse dot de <rguenther at suse dot de> --- On Thu, 24 Jan 2019, glisse at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88713 > > --- Comment #52 from Marc Glisse <glisse at gcc dot gnu.org> --- > (In reply to Thomas Koenig from comment #49) > > Argh. Sacrificing performance for the sake of bugware... > > But note that in this PR (specifically for avx512 vectors on this cpu), the OP > says that the recip version is slower than calling directly the right insn (it > wasn't clear if that was for inverse or for sqrt). Probably depends on the microarchitecture, yes. But I'd fully expect the two-NR step variant to be slower for a sensible HW implementation (even more so if we need to fend off the exceptional cases)