https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117244
--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Andrew Pinski from comment #3) > vect_get_range_info/vect_determine_precisions_from_range definitely needs to > be improved here. It does not support sign extended types into unsigned types at all since it still uses the legacy range which only supports min/max and those expressions need 2 ranges to express the sign extension into unsigned types.
