https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118832
--- Comment #6 from Robin Dapp <rdapp at gcc dot gnu.org> --- Thanks for laying it out so clearly. Helps to put things into perspective. I believe all our insn_and_split patterns already have can_create_pseudo_p in their condition so shouldn't match after reload. Warrants an audit as well of course. A back entrance is via the mov expander that might call expand_const_vector. So I would hope we're largely good.
