https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117557
--- Comment #10 from Tamar Christina <tnfchris at gcc dot gnu.org> --- The remaining failure I think is in find_reloads, where the mask feeding into the store gets ANDed with the wrong mask. Before it was using the mask coming from w1 < 0, and new one from w0 < 0; So it looks like a loop next and old one picked the mask from the inner loop and new one from the outer one. But checking with #novector to confirm.
