https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127475
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> --- (In reply to Richard Biener from comment #2) > So I'm doing testing of the following (if that succeeds, do the cleanup that > allows). But I'm surely missing something. > > diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc > index c0862577cf8..dd7eb548a3b 100644 > --- a/gcc/tree-vect-loop.cc > +++ b/gcc/tree-vect-loop.cc > @@ -9655,7 +9655,7 @@ vectorizable_induction (loop_vec_info loop_vinfo, > return false; > } > tree stept = TREE_TYPE (step_expr); > - tree step_vectype = get_same_sized_vectype (stept, vectype); > + tree step_vectype = vectype; //get_same_sized_vectype (stept, vectype); > stept = TREE_TYPE (step_vectype); > > /* Check for target support of the vectorized arithmetic used here. */ FAIL: gcc.dg/vect/pr91750.c scan-tree-dump vect "vector\\([][0-9,]*\\) unsigned int" so I'll have to adapt a bit. But get_same_sized_vectype should really go away (even more so the bool special-case ...).
