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 25E40386100B; Fri, 16 Oct 2020 14:51:04 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 25E40386100B
From: "jamborm at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug ipa/97456] [10/11 Regression] An incorrect optimization causes
 a function to always return the same value when using -flto
Date: Fri, 16 Oct 2020 14:51:03 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: ipa
X-Bugzilla-Version: 10.1.0
X-Bugzilla-Keywords: wrong-code
X-Bugzilla-Severity: normal
X-Bugzilla-Who: jamborm at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P2
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 10.3
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-97456-4-1s7ND1uisc@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-97456-4@http.gcc.gnu.org/bugzilla/>
References: <bug-97456-4@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, 16 Oct 2020 14:51:04 -0000

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D97456

--- Comment #4 from Martin Jambor <jamborm at gcc dot gnu.org> ---
Looking at Martin's reduced testcase....

(In reply to Richard Biener from comment #2)
> Confirmed with -fwhole-program -O3  IPA SRA messes things up here cloning
> wrong
> and producing the strange
>=20
> wrong.isra (floatD.41 ISRA.85D.75004)
> {
[...]
>=20
>   <bb 6> [local count: 1073741824]:
>=20
>   <bb 2> [local count: 1073741824]:
>   _1 =3D ISRA.85_10(D);
>=20
> eventually mixing up param/replacement here?

...this is simple conversion of a parameter passed by reference to one
passed by value.  I'm good at overlooking things but at least at the
moment I cannot see anything wrong with it.

Rather, I believe it is cplxlower1 which turns

  <bb 2> [local count: 1073741824]:
  a$_M_value_21 =3D COMPLEX_EXPR <ISRA.18_10(D), ISRA.18_10(D)>;

into:

  <bb 2> [local count: 1073741824]:
  a$_M_value_21 =3D COMPLEX_EXPR <a$_M_value$real_10(D), a$_M_value$real_10=
(D)>;

i.e. it replaces two uses of the parameter default-def with two
uninitialized value variable default-defs.

When I disable IPA-SRA, the pass starts with:

  <bb 2> [local count: 1073741824]:
  _1 =3D *pos_6(D);
  a$_M_value_21 =3D COMPLEX_EXPR <_1, _1>;

and changes it into:

  <bb 2> [local count: 1073741824]:
  a$_M_value$real_1 =3D *pos_6(D);
  a$_M_value_21 =3D COMPLEX_EXPR <a$_M_value$real_1, a$_M_value$real_1>;

...so my hypothesis is that cplxlower does not handle good
default-defs properly.=