https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68110
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Component|target |tree-optimization
Ever confirmed|0 |1
Status|UNCONFIRMED |NEW
Last reconfirmed| |2021-08-21
--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The testcase is fixed on rtl level since gcc 6 for x86_64.
For aarch64, it was fixed on the rtl level since GCC 9.
On the gimple level we still have:
_4 = .SUB_OVERFLOW (a_2(D), b_3(D));
_1 = IMAGPART_EXPR <_4>;
Maybe a simple match.pd pattern is needed:
(if (TYPE_SIGN (type) == UNSIGNED)
(simplify
(imagpart (sub_overflow:s @0 @1))
(convert (le @0 @1))))
