https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126899
--- Comment #6 from rguenther at suse dot de <rguenther at suse dot de> --- On Tue, 25 Aug 2026, aldyh at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126899 > > --- Comment #5 from Aldy Hernandez <aldyh at gcc dot gnu.org> --- > (In reply to Richard Biener from comment #4) > > As they do not share the cost model they of course differ in detail. I'll > > note we have way too many backwards threader passes compared to DOM thread > > passes, so it's natural the backward threader has to be more careful with > > costing. > > So what do we want to do here going forward, accept this as a regression or > adjust the cost model in the backwards threader? For now accept the regression (but track it). I have played with making the cost models more equal, but that causes regressions on its own. We'll have to tune cost modeling based on "real world data" (benchmarks), rather than only testsuite or absolute counts of # of threads.
