https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98982
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> --- loop distribution is our pass for this, but then the load/store idioms are matched by the bswap pass which relies on unrolled loops where the unrolling will only happen if it doesn't seem to increase code size. Note your testcase is incomplete and cannot be compiled.
