https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91154
--- Comment #20 from rguenther at suse dot de <rguenther at suse dot de> --- On Wed, 7 Aug 2019, segher at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91154 > > Segher Boessenkool <segher at gcc dot gnu.org> changed: > > What |Removed |Added > ---------------------------------------------------------------------------- > CC| |segher at gcc dot gnu.org > > --- Comment #19 from Segher Boessenkool <segher at gcc dot gnu.org> --- > So how does this cause 12% degradation (20% by some other measurements) > on power9? Pretty much everything is the *opposite* way around for us: > we do have cheap conditional moves, we do prefer integer registers. Might be that speculating the jump (which is very very well predicted) is still a lot faster than the conditional move.