https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86270
--- Comment #2 from Marc Glisse <glisse at gcc dot gnu.org> --- :-( So many transforms seem to have this kind of drawback... We could always add a pair of single_use checks, but we are going to miss some optimizations if we do that. Maybe it is slightly relevant that one +1 is in the loop and the other +1 is outside of the loop? It looks like it would work better if +1 and mov were done in the opposite order (kind of like rotating the loop, but maybe that's more like moving one instruction past the other and adapting the addressing in the mov).
