https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120795
--- Comment #3 from Jeffrey A. Law <law at gcc dot gnu.org> --- It'd be a surprised if it's the ext-dce change, unless it's that second parameter to the remove_reg_equal_equiv_notes call. Looking at it again, I may have inverted the desired value. If bisection lands on the ext-dce change, that's the first I'd look at.
