https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120752
--- Comment #4 from Jan Hubicka <hubicka at gcc dot gnu.org> --- Hmm, there seems to be no big differences in IPA decisions between the runs, so further investigation is necessary :( The patch attempts to preserve more of profile and here profile is bit counter-productive by disabling vectorizer due to low trip count (there is independent PR for that), so we may be hitting this. I will try to debug it.
