https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127610
--- Comment #15 from Drea Pinski <pinskia at gcc dot gnu.org> --- Looking into all of the common decrease back in May 2026; it seems like r17-646-g2508fac25d2722 is what decreased it back then (I could be wrong). Looking into all of the common increase back in Nov 2025, it seems like that was r16-5259-g4cc7bb62566460 ... r16-5248-g38155f105cca2a patch set. which would have caused cfg differences like r17-4133-g230d950810f378 would have caused. I am thinking, DOM, backward jump threader and VRP all matter heavily dependent sometimes on the CFG and IR in terms of speed. So someone would need to look into those passes. I suspect in both the r17-4133-g230d950810f378 and Nov cases; it was the removal of a condition which caused the compile time decrease in those 3 passes. Yes that might be counter intuitive but in the case of DOM and the jump threader; that might allow for more jump threading. I don't understand the VRP case though; maybe more relationships being registered.
