From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugzilla@gcc.gnu.org>
Received: by sourceware.org (Postfix, from userid 48)
 id 8BBEB3857810; Fri, 22 Oct 2021 08:32:26 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 8BBEB3857810
From: "rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug bootstrap/102681] [12 Regression] AArch64 bootstrap failure
Date: Fri, 22 Oct 2021 08:32:26 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: bootstrap
X-Bugzilla-Version: 12.0
X-Bugzilla-Keywords: build, diagnostic
X-Bugzilla-Severity: normal
X-Bugzilla-Who: rguenth at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P1
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 12.0
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-102681-4-jg4aiMOA3f@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-102681-4@http.gcc.gnu.org/bugzilla/>
References: <bug-102681-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-BeenThere: gcc-bugs@gcc.gnu.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Gcc-bugs mailing list <gcc-bugs.gcc.gnu.org>
List-Unsubscribe: <https://gcc.gnu.org/mailman/options/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=unsubscribe>
List-Archive: <https://gcc.gnu.org/pipermail/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-request@gcc.gnu.org?subject=help>
List-Subscribe: <https://gcc.gnu.org/mailman/listinfo/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=subscribe>
X-List-Received-Date: Fri, 22 Oct 2021 08:32:26 -0000

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D102681

--- Comment #21 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #14)
> Created attachment 51650 [details]
> Little more reduced
>=20
> So FRE is able to figure out for the following:
>   # _20 =3D PHI <0(2), 1(3)>
>   # const_upper_26 =3D PHI <const_upper_27(D)(2), _19(3)>
> ....
>   # _30 =3D PHI <0(12), 1(13)>
>   # const_upper_33 =3D PHI <const_upper_34(D)(12), _29(13)>
>=20
> That _30 is the same as _20 but not _26 is the same as _33 even though it
> does figure out that _19 and _29 are the same as _10. If it is able to
> figure that out, then things would just work.
>=20
> Richi,
>   I assume FRE does not Value number default SSA names (non-parm) the same
> which is why this is happening is that correct?

The issue with CSE here is that with RPO VN I made unvisited vars
VARYING due to on-demand handling.  While vn_visit_phis has special
handling for undefs the hashtable insert/lookup do not.

I am testing the following to rectify this (which then CSEs this PHI).

diff --git a/gcc/tree-ssa-sccvn.c b/gcc/tree-ssa-sccvn.c
index ae0172a143e..893b1d0ddaa 100644
--- a/gcc/tree-ssa-sccvn.c
+++ b/gcc/tree-ssa-sccvn.c
@@ -4499,7 +4499,12 @@ vn_phi_lookup (gimple *phi, bool backedges_varying_p)
       tree def =3D PHI_ARG_DEF_FROM_EDGE (phi, e);
       if (TREE_CODE (def) =3D=3D SSA_NAME
          && (!backedges_varying_p || !(e->flags & EDGE_DFS_BACK)))
-       def =3D SSA_VAL (def);
+       {
+         if (ssa_undefined_value_p (def, false))
+           def =3D VN_TOP;
+         else
+           def =3D SSA_VAL (def);
+       }
       vp1->phiargs[e->dest_idx] =3D def;
     }
   vp1->type =3D TREE_TYPE (gimple_phi_result (phi));
@@ -4543,7 +4548,12 @@ vn_phi_insert (gimple *phi, tree result, bool
backedges_varying_p)
       tree def =3D PHI_ARG_DEF_FROM_EDGE (phi, e);
       if (TREE_CODE (def) =3D=3D SSA_NAME
          && (!backedges_varying_p || !(e->flags & EDGE_DFS_BACK)))
-       def =3D SSA_VAL (def);
+       {
+         if (ssa_undefined_value_p (def, false))
+           def =3D VN_TOP;
+         else
+           def =3D SSA_VAL (def);
+       }
       vp1->phiargs[e->dest_idx] =3D def;
     }
   vp1->value_id =3D VN_INFO (result)->value_id;=