https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86020
--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> --- I spent good part of day today trying to recollect what was motivation for the change. All i can think of is that it was mistaken micro-optimization as mentioned in the mail I sent about reverting the patch. I plan to backport the change if performance turns out to be OK.