https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119757
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Li Pan from comment #8)
> Reproduced from today's upstream, looks like out-of-range access, and a
> simple fix may looks like below:
>
> diff --git a/gcc/tree-vect-slp.cc b/gcc/tree-vect-slp.cc
> index ecb4a6521de..93386851089 100644
> --- a/gcc/tree-vect-slp.cc
> +++ b/gcc/tree-vect-slp.cc
> @@ -832,7 +832,7 @@ vect_get_and_check_slp_defs (vec_info *vinfo, unsigned
> char swap,
> /* Now match the operand definition types to that of the first stmt. */
> for (i = 0; i < number_of_oprnds;)
> {
> - if (skip_args[i])
> + if (skip_args[i] || stmt_num >= oprnd_info->def_stmts.length ())
> {
> ++i;
> continue;
>
>
>
> Not sure if my understanding is correct, as well as possibly a more
> complicated reason.
That's for sure only papering over an issue elsewhere. I will have a look.
