https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68173
--- Comment #14 from rguenther at suse dot de <rguenther at suse dot de> --- On Fri, 20 Nov 2015, vmakarov at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68173 > > --- Comment #13 from Vladimir Makarov <vmakarov at gcc dot gnu.org> --- > (In reply to Richard Biener from comment #12) > > callgrind points at bitmap_set_bit called via process_bb_lives -> > > mark_regno_dead. > > Maybe some code in that (the DCE code?) can be keyed on if (optimize). > > > > in mark_regno_dead callgrind points to > > > > bitmap_set_bit (bb_killed_pseudos, regno); > > > > being the expensive one. > > I've tried to implement bb_killed_pseudos and bb_gen_pseudos as > sparsesets but it gave nothing in term of compiler speed. I think the > major problem is in processing too many pseudos in -O0 mode which are > about 1000 times more than in -O2 mode. Yeah, but without optimization we can't do sth about that (well, do more coalescing maybe, but as said coalescing is O(n^2) because of liveness - so we'd run into the very same issue there).
