https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122529
--- Comment #2 from Jeffrey A. Law <law at gcc dot gnu.org> --- I thought there was some code in CSE to handle this kind of scenario. Ah, yes, the wider one has to happen first for that code to kick in. Jivan started poking at the opposite case, but I don't think it ever got to a point of functional POC.
