https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94267
--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> --- Well, yes .. &TARGET_MEM_REF is a LEA with a specific addressing mode. Folding *&TARGET_MEM_REF would require us to verify the target can handle it (and avoid PR110702 where it's fix was really the only sensible thing I could think of). We _might_ want to fold &TARGET_MEM_REF[0 + ..] to '..', but the whole point of [&]TARGET_MEM_REF is to keep it until RTL expansion. Unless the folding is degenerate of course.
