https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117922
--- Comment #14 from rguenther at suse dot de <rguenther at suse dot de> --- On Fri, 6 Dec 2024, ebotcazou at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117922 > > Eric Botcazou <ebotcazou at gcc dot gnu.org> changed: > > What |Removed |Added > ---------------------------------------------------------------------------- > CC| |ebotcazou at gcc dot gnu.org > > --- Comment #13 from Eric Botcazou <ebotcazou at gcc dot gnu.org> --- > > One issue in fold-mem-offsets is that it seems to do per-BB operation but > > relies on global DF and looks at uses/defs also not within a BB? If it's > > supposed to be "global" then a better order than FOR_ALL_BB_FN might improve > > things. > > It also seems to do the same work multiple times: for each "root" in a given > BB, it goes up to its def in the BB and then goes down to all the uses of this > def, but I presume that this could be done only once per def? Yes. Trying to understand how it works I also believe the overall setup is overly complex - something like get_single_def_in_bb shouldn't be necessary. Walking interesting memory uses from interesting reg defs in a BB should be better.
