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 F38F43858CDB; Wed, 17 Jan 2024 08:48:59 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org F38F43858CDB
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1705481340;
	bh=pPObHV4NrCndYi4PQmpFVxiTmR6zxPlzYggpHajGkjU=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=qcOvhjMILk4cz9d4SksLw6fE4j35mC2zU1k9bJm6bao/2k1EkA8za0u2sKQi7iGdZ
	 skn6O4wY4oA5RTpRrInKQZrcMSIeObJfD60/zrOloF/JfwLsaz3LJGgTRELGVxX063
	 UOKkxLPLdFHwfZvuwECmd83tXtHkI15TmLNhTFPY=
From: "rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/113431] [14 Regression] Wrong code at -O3
Date: Wed, 17 Jan 2024 08:48:55 +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: 14.0
X-Bugzilla-Keywords: needs-bisection, wrong-code
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: 14.0
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-113431-4-zDRCE4UQEK@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-113431-4@http.gcc.gnu.org/bugzilla/>
References: <bug-113431-4@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=3D113431

--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to JuzheZhong from comment #11)
> I think this following:
>=20
> https://godbolt.org/z/5sWEWWGox
>=20
> ARM SVE GCC-11 correctly vectorize this codes.
>=20
> But both GCC-12 and GCC-13 failed to vectorize it.
>=20
> GCC-14 vectorize it in a wrong way.
>=20
> Would it be possible to recover it back to GCC-11 ?

With GCC 11 it probably SLP vectorizes the completely unrolled nest.  That
can be recovered with -fno-tree-loop-vectorize on trunk which on x86_64
produces

main:
.LFB0:
        .cfi_startproc
        movl    a+4(%rip), %edx
        xorl    %eax, %eax
        movl    $2, b(%rip)
        testl   %edx, %edx
        sete    %al
        movd    %eax, %xmm0
        setne   %al
        movzbl  %al, %eax
        pshufd  $0, %xmm0, %xmm0
        movups  %xmm0, a+4(%rip)
        movd    %eax, %xmm1
        movups  %xmm0, a+20(%rip)
        pshufd  $0, %xmm1, %xmm0
        movups  %xmm0, a+40(%rip)
        movups  %xmm0, a+56(%rip)
        ret=