https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125291
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2026-05-13
Ever confirmed|0 |1
--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Xi Ruoyao from comment #4)
> Tentative untested patch:
>
> diff --git a/gcc/tree-scalar-evolution.cc b/gcc/tree-scalar-evolution.cc
> index ecdef7529a6..b55ae0586c8 100644
> --- a/gcc/tree-scalar-evolution.cc
> +++ b/gcc/tree-scalar-evolution.cc
> @@ -1377,10 +1377,16 @@ simplify_peeled_chrec (class loop *loop, tree arg,
> tree init_cond)
> && wi::to_widest (init_cond) == wi::to_widest (left_before)
> && !scev_probably_wraps_p (NULL_TREE, left_before, right, NULL,
> loop, false))
> - return build_polynomial_chrec (loop->num, init_cond,
> - chrec_convert (TREE_TYPE (ev),
> - right, NULL,
> - false, NULL_TREE));
> + {
> + tree r = build_polynomial_chrec (loop->num, init_cond,
> + chrec_convert (TREE_TYPE (ev),
> + right, NULL,
> + false, NULL_TREE));
> + /* (u8)[6, 260, 514] is [6, 4, 2], not [6, 260, 514]. */
> + if (tree_int_cst_sign_bit (right)
> + == (scev_direction (r) == EV_DIR_DECREASES))
> + return r;
Hmm, wouldn't it be better to always sign-extend 'right'? Thanks for
working on a fix. I'm not sure right is always an INTEGER_CST btw.
> + }
> return chrec_dont_know;
> }
