https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127606
--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> --- Thanks for the input Vlad, this was originally my thought as well - the only inconvenience is that two uses of the same pseudo (with the same reaching definition) can get a different value (for undefined bits) from a spill/reload which is indeed "inconvenient". I will post my patch to combine to the list to expose this to a wider audience.
