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 1B7AC387606A; Thu, 27 Jun 2024 06:30:34 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 1B7AC387606A
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1719469835;
	bh=Aoe7KUO4skJ925KvYNYqqwONEugeKTgWOpIOZN0px+0=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=WofBIe05j/vbdnitO++ie5YH3n5zW+Z4TFOEgxBoudE6c+WG6P3EHRdvh4ppTOKT3
	 kiDgYOuLbhuAvANTjB5UmMJBTqWXqRpUDE10qL0YM0ZBmUp8khGM+xwl7IK43cjnvb
	 1wfzSdi6nKNf7Jt/87J//we1uC89sYCE9NMdSJow=
From: "rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/115669] [15 Regression] rv64gcv/aarch64+sve
 -fwrapv miscompile since r15-1006-gd93353e6423
Date: Thu, 27 Jun 2024 06:30:34 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: tree-optimization
X-Bugzilla-Version: 15.0
X-Bugzilla-Keywords: wrong-code
X-Bugzilla-Severity: normal
X-Bugzilla-Who: rguenth at gcc dot gnu.org
X-Bugzilla-Status: ASSIGNED
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: rguenth at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 15.0
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-115669-4-dveOkbeet3@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-115669-4@http.gcc.gnu.org/bugzilla/>
References: <bug-115669-4@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
List-Id: <gcc-bugs.sourceware.org>

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D115669

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
I can reproduce on aarch64 with -O2 -march=3Darmv8.3-a+sve -fno-vect-cost-m=
odel
-fwrapv.  The reduction is

  <bb 3> [local count: 955630224]:
  # b_lsm.9_11 =3D PHI <_9(7), pretmp_42(6)>
  _4 =3D *_3;
  _5 =3D (signed int) _4;
  b.2_7 =3D (signed int) b_lsm.9_11;
  _15 =3D b.2_7 + -1;
  _8 =3D _15 - _5;
  _9 =3D (unsigned int) _8;

with -fwrapv and

  # b_lsm.9_11 =3D PHI <_7(7), pretmp_40(6)>
  _4 =3D *_3;
  _5 =3D (unsigned int) _4;
  _15 =3D b_lsm.9_11 + 4294967295;
  _7 =3D _15 - _5;

without.  So it seems the support for singed/unsigned converted reductions
is somehow broken.  At least for VLA vectors.

The epilog does

  # vect__9.22_59 =3D PHI <vect__9.22_58(3)>
  _60 =3D (vector([4,4]) signed int) vect__9.22_59;
  _61 =3D .REDUC_PLUS (_60);

I don't see where it goes wrong yet.=